{ "best_metric": null, "best_model_checkpoint": null, "epoch": 3.963038180341186, "eval_steps": 500, "global_step": 39392, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 6e-06, "loss": 0.9611, "step": 1 }, { "epoch": 0.0, "learning_rate": 1.2e-05, "loss": 0.6909, "step": 2 }, { "epoch": 0.0, "learning_rate": 1.8e-05, "loss": 1.0511, "step": 3 }, { "epoch": 0.0, "learning_rate": 2.4e-05, "loss": 0.8962, "step": 4 }, { "epoch": 0.0, "learning_rate": 3e-05, "loss": 0.9467, "step": 5 }, { "epoch": 0.0, "learning_rate": 3.6e-05, "loss": 0.6699, "step": 6 }, { "epoch": 0.0, "learning_rate": 4.2e-05, "loss": 0.899, "step": 7 }, { "epoch": 0.0, "learning_rate": 4.8e-05, "loss": 0.8717, "step": 8 }, { "epoch": 0.0, "learning_rate": 5.4000000000000005e-05, "loss": 0.6804, "step": 9 }, { "epoch": 0.0, "learning_rate": 6e-05, "loss": 1.1548, "step": 10 }, { "epoch": 0.0, "learning_rate": 5.9999999904545704e-05, "loss": 1.0779, "step": 11 }, { "epoch": 0.0, "learning_rate": 5.9999999618182824e-05, "loss": 1.2244, "step": 12 }, { "epoch": 0.0, "learning_rate": 5.999999914091136e-05, "loss": 0.9183, "step": 13 }, { "epoch": 0.0, "learning_rate": 5.999999847273132e-05, "loss": 0.842, "step": 14 }, { "epoch": 0.0, "learning_rate": 5.999999761364269e-05, "loss": 1.057, "step": 15 }, { "epoch": 0.0, "learning_rate": 5.9999996563645495e-05, "loss": 1.0065, "step": 16 }, { "epoch": 0.0, "learning_rate": 5.999999532273973e-05, "loss": 0.2884, "step": 17 }, { "epoch": 0.0, "learning_rate": 5.999999389092541e-05, "loss": 1.0586, "step": 18 }, { "epoch": 0.0, "learning_rate": 5.999999226820254e-05, "loss": 1.1559, "step": 19 }, { "epoch": 0.0, "learning_rate": 5.999999045457113e-05, "loss": 0.3994, "step": 20 }, { "epoch": 0.0, "learning_rate": 5.9999988450031195e-05, "loss": 0.6816, "step": 21 }, { "epoch": 0.0, "learning_rate": 5.999998625458275e-05, "loss": 1.3041, "step": 22 }, { "epoch": 0.0, "learning_rate": 5.999998386822581e-05, "loss": 1.0317, "step": 23 }, { "epoch": 0.0, "learning_rate": 5.9999981290960374e-05, "loss": 0.4954, "step": 24 }, { "epoch": 0.0, "learning_rate": 5.9999978522786466e-05, "loss": 0.9628, "step": 25 }, { "epoch": 0.0, "learning_rate": 5.999997556370412e-05, "loss": 0.2448, "step": 26 }, { "epoch": 0.0, "learning_rate": 5.999997241371333e-05, "loss": 1.1264, "step": 27 }, { "epoch": 0.0, "learning_rate": 5.9999969072814134e-05, "loss": 0.5052, "step": 28 }, { "epoch": 0.0, "learning_rate": 5.9999965541006556e-05, "loss": 0.5601, "step": 29 }, { "epoch": 0.0, "learning_rate": 5.999996181829059e-05, "loss": 0.6859, "step": 30 }, { "epoch": 0.0, "learning_rate": 5.99999579046663e-05, "loss": 1.6138, "step": 31 }, { "epoch": 0.0, "learning_rate": 5.999995380013368e-05, "loss": 0.5146, "step": 32 }, { "epoch": 0.0, "learning_rate": 5.999994950469277e-05, "loss": 0.3252, "step": 33 }, { "epoch": 0.0, "learning_rate": 5.999994501834359e-05, "loss": 1.9287, "step": 34 }, { "epoch": 0.0, "learning_rate": 5.999994034108618e-05, "loss": 0.9413, "step": 35 }, { "epoch": 0.0, "learning_rate": 5.999993547292055e-05, "loss": 0.6465, "step": 36 }, { "epoch": 0.0, "learning_rate": 5.999993041384675e-05, "loss": 1.1896, "step": 37 }, { "epoch": 0.0, "learning_rate": 5.999992516386481e-05, "loss": 0.9641, "step": 38 }, { "epoch": 0.0, "learning_rate": 5.9999919722974753e-05, "loss": 0.8682, "step": 39 }, { "epoch": 0.0, "learning_rate": 5.999991409117662e-05, "loss": 0.4005, "step": 40 }, { "epoch": 0.0, "learning_rate": 5.999990826847045e-05, "loss": 0.9974, "step": 41 }, { "epoch": 0.0, "learning_rate": 5.9999902254856276e-05, "loss": 0.6602, "step": 42 }, { "epoch": 0.0, "learning_rate": 5.9999896050334123e-05, "loss": 0.3048, "step": 43 }, { "epoch": 0.0, "learning_rate": 5.999988965490406e-05, "loss": 0.8209, "step": 44 }, { "epoch": 0.0, "learning_rate": 5.9999883068566106e-05, "loss": 0.7759, "step": 45 }, { "epoch": 0.0, "learning_rate": 5.9999876291320315e-05, "loss": 0.3947, "step": 46 }, { "epoch": 0.0, "learning_rate": 5.9999869323166715e-05, "loss": 0.4532, "step": 47 }, { "epoch": 0.0, "learning_rate": 5.999986216410537e-05, "loss": 0.4846, "step": 48 }, { "epoch": 0.0, "learning_rate": 5.99998548141363e-05, "loss": 0.6761, "step": 49 }, { "epoch": 0.01, "learning_rate": 5.9999847273259574e-05, "loss": 0.5133, "step": 50 }, { "epoch": 0.01, "learning_rate": 5.9999839541475225e-05, "loss": 1.4704, "step": 51 }, { "epoch": 0.01, "learning_rate": 5.9999831618783324e-05, "loss": 1.3765, "step": 52 }, { "epoch": 0.01, "learning_rate": 5.99998235051839e-05, "loss": 0.7539, "step": 53 }, { "epoch": 0.01, "learning_rate": 5.999981520067701e-05, "loss": 0.9062, "step": 54 }, { "epoch": 0.01, "learning_rate": 5.999980670526271e-05, "loss": 0.7255, "step": 55 }, { "epoch": 0.01, "learning_rate": 5.9999798018941055e-05, "loss": 0.2341, "step": 56 }, { "epoch": 0.01, "learning_rate": 5.999978914171209e-05, "loss": 0.6138, "step": 57 }, { "epoch": 0.01, "learning_rate": 5.999978007357589e-05, "loss": 1.0273, "step": 58 }, { "epoch": 0.01, "learning_rate": 5.9999770814532495e-05, "loss": 1.0243, "step": 59 }, { "epoch": 0.01, "learning_rate": 5.9999761364581975e-05, "loss": 0.7568, "step": 60 }, { "epoch": 0.01, "learning_rate": 5.9999751723724385e-05, "loss": 0.7743, "step": 61 }, { "epoch": 0.01, "learning_rate": 5.9999741891959785e-05, "loss": 0.205, "step": 62 }, { "epoch": 0.01, "learning_rate": 5.999973186928825e-05, "loss": 0.8313, "step": 63 }, { "epoch": 0.01, "learning_rate": 5.9999721655709826e-05, "loss": 0.5768, "step": 64 }, { "epoch": 0.01, "learning_rate": 5.9999711251224576e-05, "loss": 0.5365, "step": 65 }, { "epoch": 0.01, "learning_rate": 5.999970065583259e-05, "loss": 1.1081, "step": 66 }, { "epoch": 0.01, "learning_rate": 5.999968986953392e-05, "loss": 0.588, "step": 67 }, { "epoch": 0.01, "learning_rate": 5.999967889232863e-05, "loss": 0.5623, "step": 68 }, { "epoch": 0.01, "learning_rate": 5.9999667724216806e-05, "loss": 1.3136, "step": 69 }, { "epoch": 0.01, "learning_rate": 5.999965636519849e-05, "loss": 0.4415, "step": 70 }, { "epoch": 0.01, "learning_rate": 5.99996448152738e-05, "loss": 0.2088, "step": 71 }, { "epoch": 0.01, "learning_rate": 5.999963307444276e-05, "loss": 0.884, "step": 72 }, { "epoch": 0.01, "learning_rate": 5.999962114270548e-05, "loss": 1.1938, "step": 73 }, { "epoch": 0.01, "learning_rate": 5.9999609020062015e-05, "loss": 0.7319, "step": 74 }, { "epoch": 0.01, "learning_rate": 5.999959670651246e-05, "loss": 0.4003, "step": 75 }, { "epoch": 0.01, "learning_rate": 5.999958420205688e-05, "loss": 0.4763, "step": 76 }, { "epoch": 0.01, "learning_rate": 5.999957150669536e-05, "loss": 0.5517, "step": 77 }, { "epoch": 0.01, "learning_rate": 5.999955862042798e-05, "loss": 0.9328, "step": 78 }, { "epoch": 0.01, "learning_rate": 5.9999545543254816e-05, "loss": 0.7181, "step": 79 }, { "epoch": 0.01, "learning_rate": 5.9999532275175954e-05, "loss": 0.9047, "step": 80 }, { "epoch": 0.01, "learning_rate": 5.999951881619149e-05, "loss": 0.7952, "step": 81 }, { "epoch": 0.01, "learning_rate": 5.9999505166301496e-05, "loss": 1.162, "step": 82 }, { "epoch": 0.01, "learning_rate": 5.9999491325506063e-05, "loss": 0.2519, "step": 83 }, { "epoch": 0.01, "learning_rate": 5.999947729380529e-05, "loss": 0.9357, "step": 84 }, { "epoch": 0.01, "learning_rate": 5.9999463071199244e-05, "loss": 0.5643, "step": 85 }, { "epoch": 0.01, "learning_rate": 5.999944865768802e-05, "loss": 1.2378, "step": 86 }, { "epoch": 0.01, "learning_rate": 5.999943405327173e-05, "loss": 1.237, "step": 87 }, { "epoch": 0.01, "learning_rate": 5.999941925795046e-05, "loss": 0.6347, "step": 88 }, { "epoch": 0.01, "learning_rate": 5.9999404271724284e-05, "loss": 0.9233, "step": 89 }, { "epoch": 0.01, "learning_rate": 5.999938909459332e-05, "loss": 0.5314, "step": 90 }, { "epoch": 0.01, "learning_rate": 5.9999373726557655e-05, "loss": 0.861, "step": 91 }, { "epoch": 0.01, "learning_rate": 5.9999358167617383e-05, "loss": 1.0264, "step": 92 }, { "epoch": 0.01, "learning_rate": 5.999934241777261e-05, "loss": 0.2721, "step": 93 }, { "epoch": 0.01, "learning_rate": 5.9999326477023434e-05, "loss": 1.1545, "step": 94 }, { "epoch": 0.01, "learning_rate": 5.999931034536997e-05, "loss": 1.0487, "step": 95 }, { "epoch": 0.01, "learning_rate": 5.9999294022812295e-05, "loss": 1.117, "step": 96 }, { "epoch": 0.01, "learning_rate": 5.999927750935052e-05, "loss": 1.0498, "step": 97 }, { "epoch": 0.01, "learning_rate": 5.9999260804984765e-05, "loss": 0.6774, "step": 98 }, { "epoch": 0.01, "learning_rate": 5.999924390971511e-05, "loss": 0.2515, "step": 99 }, { "epoch": 0.01, "learning_rate": 5.9999226823541706e-05, "loss": 0.6544, "step": 100 }, { "epoch": 0.01, "learning_rate": 5.999920954646462e-05, "loss": 0.9319, "step": 101 }, { "epoch": 0.01, "learning_rate": 5.999919207848396e-05, "loss": 0.8882, "step": 102 }, { "epoch": 0.01, "learning_rate": 5.9999174419599877e-05, "loss": 0.3299, "step": 103 }, { "epoch": 0.01, "learning_rate": 5.999915656981246e-05, "loss": 1.0048, "step": 104 }, { "epoch": 0.01, "learning_rate": 5.9999138529121814e-05, "loss": 0.3582, "step": 105 }, { "epoch": 0.01, "learning_rate": 5.999912029752807e-05, "loss": 0.3013, "step": 106 }, { "epoch": 0.01, "learning_rate": 5.9999101875031325e-05, "loss": 0.4037, "step": 107 }, { "epoch": 0.01, "learning_rate": 5.999908326163172e-05, "loss": 0.7843, "step": 108 }, { "epoch": 0.01, "learning_rate": 5.999906445732935e-05, "loss": 1.2967, "step": 109 }, { "epoch": 0.01, "learning_rate": 5.999904546212436e-05, "loss": 0.6817, "step": 110 }, { "epoch": 0.01, "learning_rate": 5.999902627601685e-05, "loss": 0.6538, "step": 111 }, { "epoch": 0.01, "learning_rate": 5.9999006899006946e-05, "loss": 0.4924, "step": 112 }, { "epoch": 0.01, "learning_rate": 5.9998987331094775e-05, "loss": 0.2237, "step": 113 }, { "epoch": 0.01, "learning_rate": 5.9998967572280466e-05, "loss": 0.5863, "step": 114 }, { "epoch": 0.01, "learning_rate": 5.999894762256414e-05, "loss": 1.1475, "step": 115 }, { "epoch": 0.01, "learning_rate": 5.999892748194591e-05, "loss": 0.7886, "step": 116 }, { "epoch": 0.01, "learning_rate": 5.9998907150425935e-05, "loss": 0.4083, "step": 117 }, { "epoch": 0.01, "learning_rate": 5.9998886628004316e-05, "loss": 0.2351, "step": 118 }, { "epoch": 0.01, "learning_rate": 5.99988659146812e-05, "loss": 0.4066, "step": 119 }, { "epoch": 0.01, "learning_rate": 5.999884501045672e-05, "loss": 0.7706, "step": 120 }, { "epoch": 0.01, "learning_rate": 5.999882391533099e-05, "loss": 0.99, "step": 121 }, { "epoch": 0.01, "learning_rate": 5.9998802629304156e-05, "loss": 1.1271, "step": 122 }, { "epoch": 0.01, "learning_rate": 5.9998781152376355e-05, "loss": 1.2219, "step": 123 }, { "epoch": 0.01, "learning_rate": 5.999875948454773e-05, "loss": 1.3663, "step": 124 }, { "epoch": 0.01, "learning_rate": 5.9998737625818414e-05, "loss": 0.5026, "step": 125 }, { "epoch": 0.01, "learning_rate": 5.999871557618853e-05, "loss": 0.6587, "step": 126 }, { "epoch": 0.01, "learning_rate": 5.9998693335658245e-05, "loss": 0.5226, "step": 127 }, { "epoch": 0.01, "learning_rate": 5.9998670904227685e-05, "loss": 0.6075, "step": 128 }, { "epoch": 0.01, "learning_rate": 5.999864828189699e-05, "loss": 1.0565, "step": 129 }, { "epoch": 0.01, "learning_rate": 5.999862546866632e-05, "loss": 0.5124, "step": 130 }, { "epoch": 0.01, "learning_rate": 5.9998602464535796e-05, "loss": 0.5832, "step": 131 }, { "epoch": 0.01, "learning_rate": 5.999857926950559e-05, "loss": 1.0389, "step": 132 }, { "epoch": 0.01, "learning_rate": 5.9998555883575834e-05, "loss": 1.1801, "step": 133 }, { "epoch": 0.01, "learning_rate": 5.999853230674668e-05, "loss": 0.3315, "step": 134 }, { "epoch": 0.01, "learning_rate": 5.999850853901827e-05, "loss": 1.2234, "step": 135 }, { "epoch": 0.01, "learning_rate": 5.999848458039078e-05, "loss": 0.4588, "step": 136 }, { "epoch": 0.01, "learning_rate": 5.999846043086433e-05, "loss": 0.6374, "step": 137 }, { "epoch": 0.01, "learning_rate": 5.999843609043909e-05, "loss": 0.9688, "step": 138 }, { "epoch": 0.01, "learning_rate": 5.999841155911523e-05, "loss": 0.8913, "step": 139 }, { "epoch": 0.01, "learning_rate": 5.999838683689288e-05, "loss": 0.3771, "step": 140 }, { "epoch": 0.01, "learning_rate": 5.9998361923772206e-05, "loss": 0.973, "step": 141 }, { "epoch": 0.01, "learning_rate": 5.999833681975338e-05, "loss": 0.5952, "step": 142 }, { "epoch": 0.01, "learning_rate": 5.9998311524836543e-05, "loss": 1.0392, "step": 143 }, { "epoch": 0.01, "learning_rate": 5.9998286039021866e-05, "loss": 0.233, "step": 144 }, { "epoch": 0.01, "learning_rate": 5.99982603623095e-05, "loss": 0.5669, "step": 145 }, { "epoch": 0.01, "learning_rate": 5.999823449469963e-05, "loss": 1.0502, "step": 146 }, { "epoch": 0.01, "learning_rate": 5.99982084361924e-05, "loss": 0.2361, "step": 147 }, { "epoch": 0.02, "learning_rate": 5.999818218678798e-05, "loss": 0.3378, "step": 148 }, { "epoch": 0.02, "learning_rate": 5.9998155746486546e-05, "loss": 1.2308, "step": 149 }, { "epoch": 0.02, "learning_rate": 5.999812911528826e-05, "loss": 0.7853, "step": 150 }, { "epoch": 0.02, "learning_rate": 5.999810229319329e-05, "loss": 0.9563, "step": 151 }, { "epoch": 0.02, "learning_rate": 5.9998075280201815e-05, "loss": 0.9404, "step": 152 }, { "epoch": 0.02, "learning_rate": 5.9998048076314e-05, "loss": 0.5781, "step": 153 }, { "epoch": 0.02, "learning_rate": 5.999802068153001e-05, "loss": 0.3761, "step": 154 }, { "epoch": 0.02, "learning_rate": 5.999799309585003e-05, "loss": 1.2978, "step": 155 }, { "epoch": 0.02, "learning_rate": 5.9997965319274244e-05, "loss": 0.3358, "step": 156 }, { "epoch": 0.02, "learning_rate": 5.9997937351802814e-05, "loss": 0.7192, "step": 157 }, { "epoch": 0.02, "learning_rate": 5.999790919343592e-05, "loss": 0.543, "step": 158 }, { "epoch": 0.02, "learning_rate": 5.9997880844173746e-05, "loss": 0.456, "step": 159 }, { "epoch": 0.02, "learning_rate": 5.999785230401647e-05, "loss": 0.6113, "step": 160 }, { "epoch": 0.02, "learning_rate": 5.999782357296426e-05, "loss": 0.6918, "step": 161 }, { "epoch": 0.02, "learning_rate": 5.999779465101733e-05, "loss": 0.811, "step": 162 }, { "epoch": 0.02, "learning_rate": 5.999776553817584e-05, "loss": 1.1079, "step": 163 }, { "epoch": 0.02, "learning_rate": 5.999773623443998e-05, "loss": 0.7662, "step": 164 }, { "epoch": 0.02, "learning_rate": 5.9997706739809946e-05, "loss": 0.64, "step": 165 }, { "epoch": 0.02, "learning_rate": 5.999767705428591e-05, "loss": 0.5964, "step": 166 }, { "epoch": 0.02, "learning_rate": 5.9997647177868074e-05, "loss": 0.7939, "step": 167 }, { "epoch": 0.02, "learning_rate": 5.999761711055663e-05, "loss": 0.3607, "step": 168 }, { "epoch": 0.02, "learning_rate": 5.999758685235174e-05, "loss": 1.1062, "step": 169 }, { "epoch": 0.02, "learning_rate": 5.999755640325364e-05, "loss": 0.689, "step": 170 }, { "epoch": 0.02, "learning_rate": 5.99975257632625e-05, "loss": 0.4168, "step": 171 }, { "epoch": 0.02, "learning_rate": 5.99974949323785e-05, "loss": 0.7292, "step": 172 }, { "epoch": 0.02, "learning_rate": 5.999746391060188e-05, "loss": 0.4276, "step": 173 }, { "epoch": 0.02, "learning_rate": 5.9997432697932784e-05, "loss": 1.1441, "step": 174 }, { "epoch": 0.02, "learning_rate": 5.999740129437146e-05, "loss": 1.1062, "step": 175 }, { "epoch": 0.02, "learning_rate": 5.9997369699918074e-05, "loss": 0.5886, "step": 176 }, { "epoch": 0.02, "learning_rate": 5.999733791457284e-05, "loss": 0.7991, "step": 177 }, { "epoch": 0.02, "learning_rate": 5.9997305938335966e-05, "loss": 0.7667, "step": 178 }, { "epoch": 0.02, "learning_rate": 5.9997273771207644e-05, "loss": 1.0188, "step": 179 }, { "epoch": 0.02, "learning_rate": 5.999724141318808e-05, "loss": 0.912, "step": 180 }, { "epoch": 0.02, "learning_rate": 5.999720886427749e-05, "loss": 0.7402, "step": 181 }, { "epoch": 0.02, "learning_rate": 5.999717612447606e-05, "loss": 0.5352, "step": 182 }, { "epoch": 0.02, "learning_rate": 5.9997143193784035e-05, "loss": 0.2948, "step": 183 }, { "epoch": 0.02, "learning_rate": 5.9997110072201586e-05, "loss": 0.7372, "step": 184 }, { "epoch": 0.02, "learning_rate": 5.9997076759728946e-05, "loss": 0.4266, "step": 185 }, { "epoch": 0.02, "learning_rate": 5.9997043256366325e-05, "loss": 0.4558, "step": 186 }, { "epoch": 0.02, "learning_rate": 5.999700956211392e-05, "loss": 0.307, "step": 187 }, { "epoch": 0.02, "learning_rate": 5.999697567697196e-05, "loss": 0.3302, "step": 188 }, { "epoch": 0.02, "learning_rate": 5.999694160094067e-05, "loss": 0.7018, "step": 189 }, { "epoch": 0.02, "learning_rate": 5.9996907334020254e-05, "loss": 0.5692, "step": 190 }, { "epoch": 0.02, "learning_rate": 5.999687287621092e-05, "loss": 0.3531, "step": 191 }, { "epoch": 0.02, "learning_rate": 5.99968382275129e-05, "loss": 0.8143, "step": 192 }, { "epoch": 0.02, "learning_rate": 5.999680338792642e-05, "loss": 0.7237, "step": 193 }, { "epoch": 0.02, "learning_rate": 5.999676835745169e-05, "loss": 1.2443, "step": 194 }, { "epoch": 0.02, "learning_rate": 5.999673313608894e-05, "loss": 0.5115, "step": 195 }, { "epoch": 0.02, "learning_rate": 5.999669772383839e-05, "loss": 0.5782, "step": 196 }, { "epoch": 0.02, "learning_rate": 5.999666212070027e-05, "loss": 0.853, "step": 197 }, { "epoch": 0.02, "learning_rate": 5.9996626326674796e-05, "loss": 0.7049, "step": 198 }, { "epoch": 0.02, "learning_rate": 5.999659034176221e-05, "loss": 1.0725, "step": 199 }, { "epoch": 0.02, "learning_rate": 5.999655416596273e-05, "loss": 1.0408, "step": 200 }, { "epoch": 0.02, "learning_rate": 5.999651779927659e-05, "loss": 0.7292, "step": 201 }, { "epoch": 0.02, "learning_rate": 5.999648124170403e-05, "loss": 0.9249, "step": 202 }, { "epoch": 0.02, "learning_rate": 5.999644449324526e-05, "loss": 0.6987, "step": 203 }, { "epoch": 0.02, "learning_rate": 5.999640755390055e-05, "loss": 0.4626, "step": 204 }, { "epoch": 0.02, "learning_rate": 5.9996370423670097e-05, "loss": 1.3488, "step": 205 }, { "epoch": 0.02, "learning_rate": 5.999633310255415e-05, "loss": 1.1553, "step": 206 }, { "epoch": 0.02, "learning_rate": 5.999629559055297e-05, "loss": 0.4537, "step": 207 }, { "epoch": 0.02, "learning_rate": 5.999625788766676e-05, "loss": 0.7393, "step": 208 }, { "epoch": 0.02, "learning_rate": 5.999621999389578e-05, "loss": 0.9267, "step": 209 }, { "epoch": 0.02, "learning_rate": 5.9996181909240275e-05, "loss": 0.7131, "step": 210 }, { "epoch": 0.02, "learning_rate": 5.9996143633700467e-05, "loss": 0.2273, "step": 211 }, { "epoch": 0.02, "learning_rate": 5.999610516727662e-05, "loss": 1.0285, "step": 212 }, { "epoch": 0.02, "learning_rate": 5.9996066509968966e-05, "loss": 0.7143, "step": 213 }, { "epoch": 0.02, "learning_rate": 5.9996027661777755e-05, "loss": 0.7384, "step": 214 }, { "epoch": 0.02, "learning_rate": 5.9995988622703244e-05, "loss": 0.6515, "step": 215 }, { "epoch": 0.02, "learning_rate": 5.999594939274567e-05, "loss": 0.4473, "step": 216 }, { "epoch": 0.02, "learning_rate": 5.999590997190528e-05, "loss": 0.2267, "step": 217 }, { "epoch": 0.02, "learning_rate": 5.9995870360182327e-05, "loss": 1.1585, "step": 218 }, { "epoch": 0.02, "learning_rate": 5.999583055757707e-05, "loss": 0.9453, "step": 219 }, { "epoch": 0.02, "learning_rate": 5.999579056408976e-05, "loss": 0.9381, "step": 220 }, { "epoch": 0.02, "learning_rate": 5.999575037972065e-05, "loss": 0.1904, "step": 221 }, { "epoch": 0.02, "learning_rate": 5.9995710004469996e-05, "loss": 0.6304, "step": 222 }, { "epoch": 0.02, "learning_rate": 5.9995669438338053e-05, "loss": 0.5582, "step": 223 }, { "epoch": 0.02, "learning_rate": 5.9995628681325075e-05, "loss": 0.3723, "step": 224 }, { "epoch": 0.02, "learning_rate": 5.999558773343133e-05, "loss": 0.8, "step": 225 }, { "epoch": 0.02, "learning_rate": 5.999554659465708e-05, "loss": 1.1826, "step": 226 }, { "epoch": 0.02, "learning_rate": 5.999550526500258e-05, "loss": 1.1407, "step": 227 }, { "epoch": 0.02, "learning_rate": 5.99954637444681e-05, "loss": 0.3101, "step": 228 }, { "epoch": 0.02, "learning_rate": 5.9995422033053894e-05, "loss": 0.3289, "step": 229 }, { "epoch": 0.02, "learning_rate": 5.9995380130760234e-05, "loss": 0.5929, "step": 230 }, { "epoch": 0.02, "learning_rate": 5.999533803758739e-05, "loss": 0.7245, "step": 231 }, { "epoch": 0.02, "learning_rate": 5.999529575353562e-05, "loss": 0.8261, "step": 232 }, { "epoch": 0.02, "learning_rate": 5.99952532786052e-05, "loss": 0.4927, "step": 233 }, { "epoch": 0.02, "learning_rate": 5.99952106127964e-05, "loss": 0.3573, "step": 234 }, { "epoch": 0.02, "learning_rate": 5.9995167756109484e-05, "loss": 0.5373, "step": 235 }, { "epoch": 0.02, "learning_rate": 5.999512470854474e-05, "loss": 0.6602, "step": 236 }, { "epoch": 0.02, "learning_rate": 5.9995081470102424e-05, "loss": 0.695, "step": 237 }, { "epoch": 0.02, "learning_rate": 5.999503804078283e-05, "loss": 0.3144, "step": 238 }, { "epoch": 0.02, "learning_rate": 5.9994994420586215e-05, "loss": 0.9949, "step": 239 }, { "epoch": 0.02, "learning_rate": 5.9994950609512875e-05, "loss": 0.4096, "step": 240 }, { "epoch": 0.02, "learning_rate": 5.999490660756308e-05, "loss": 1.0615, "step": 241 }, { "epoch": 0.02, "learning_rate": 5.9994862414737106e-05, "loss": 0.6429, "step": 242 }, { "epoch": 0.02, "learning_rate": 5.999481803103523e-05, "loss": 0.4857, "step": 243 }, { "epoch": 0.02, "learning_rate": 5.9994773456457764e-05, "loss": 0.333, "step": 244 }, { "epoch": 0.02, "learning_rate": 5.999472869100495e-05, "loss": 0.6271, "step": 245 }, { "epoch": 0.02, "learning_rate": 5.999468373467711e-05, "loss": 1.3449, "step": 246 }, { "epoch": 0.03, "learning_rate": 5.99946385874745e-05, "loss": 0.3742, "step": 247 }, { "epoch": 0.03, "learning_rate": 5.9994593249397424e-05, "loss": 0.9754, "step": 248 }, { "epoch": 0.03, "learning_rate": 5.999454772044617e-05, "loss": 0.6475, "step": 249 }, { "epoch": 0.03, "learning_rate": 5.999450200062102e-05, "loss": 0.4562, "step": 250 }, { "epoch": 0.03, "learning_rate": 5.9994456089922274e-05, "loss": 1.127, "step": 251 }, { "epoch": 0.03, "learning_rate": 5.999440998835022e-05, "loss": 0.6281, "step": 252 }, { "epoch": 0.03, "learning_rate": 5.999436369590515e-05, "loss": 0.3564, "step": 253 }, { "epoch": 0.03, "learning_rate": 5.999431721258736e-05, "loss": 1.2867, "step": 254 }, { "epoch": 0.03, "learning_rate": 5.999427053839714e-05, "loss": 0.8598, "step": 255 }, { "epoch": 0.03, "learning_rate": 5.99942236733348e-05, "loss": 0.8063, "step": 256 }, { "epoch": 0.03, "learning_rate": 5.9994176617400625e-05, "loss": 0.7529, "step": 257 }, { "epoch": 0.03, "learning_rate": 5.9994129370594934e-05, "loss": 0.9318, "step": 258 }, { "epoch": 0.03, "learning_rate": 5.9994081932918e-05, "loss": 0.3904, "step": 259 }, { "epoch": 0.03, "learning_rate": 5.999403430437014e-05, "loss": 0.5806, "step": 260 }, { "epoch": 0.03, "learning_rate": 5.999398648495166e-05, "loss": 0.9175, "step": 261 }, { "epoch": 0.03, "learning_rate": 5.9993938474662856e-05, "loss": 0.9559, "step": 262 }, { "epoch": 0.03, "learning_rate": 5.9993890273504046e-05, "loss": 0.3542, "step": 263 }, { "epoch": 0.03, "learning_rate": 5.999384188147553e-05, "loss": 1.1604, "step": 264 }, { "epoch": 0.03, "learning_rate": 5.99937932985776e-05, "loss": 1.2079, "step": 265 }, { "epoch": 0.03, "learning_rate": 5.9993744524810584e-05, "loss": 1.6522, "step": 266 }, { "epoch": 0.03, "learning_rate": 5.99936955601748e-05, "loss": 0.8846, "step": 267 }, { "epoch": 0.03, "learning_rate": 5.999364640467054e-05, "loss": 0.7613, "step": 268 }, { "epoch": 0.03, "learning_rate": 5.9993597058298126e-05, "loss": 0.4491, "step": 269 }, { "epoch": 0.03, "learning_rate": 5.999354752105786e-05, "loss": 0.5709, "step": 270 }, { "epoch": 0.03, "learning_rate": 5.999349779295008e-05, "loss": 1.2629, "step": 271 }, { "epoch": 0.03, "learning_rate": 5.999344787397509e-05, "loss": 0.2976, "step": 272 }, { "epoch": 0.03, "learning_rate": 5.99933977641332e-05, "loss": 1.5506, "step": 273 }, { "epoch": 0.03, "learning_rate": 5.9993347463424745e-05, "loss": 1.0612, "step": 274 }, { "epoch": 0.03, "learning_rate": 5.999329697185003e-05, "loss": 1.0229, "step": 275 }, { "epoch": 0.03, "learning_rate": 5.999324628940939e-05, "loss": 0.898, "step": 276 }, { "epoch": 0.03, "learning_rate": 5.999319541610313e-05, "loss": 0.9809, "step": 277 }, { "epoch": 0.03, "learning_rate": 5.99931443519316e-05, "loss": 0.5301, "step": 278 }, { "epoch": 0.03, "learning_rate": 5.9993093096895095e-05, "loss": 0.9113, "step": 279 }, { "epoch": 0.03, "learning_rate": 5.999304165099396e-05, "loss": 1.0725, "step": 280 }, { "epoch": 0.03, "learning_rate": 5.999299001422852e-05, "loss": 0.7417, "step": 281 }, { "epoch": 0.03, "learning_rate": 5.9992938186599105e-05, "loss": 1.0214, "step": 282 }, { "epoch": 0.03, "learning_rate": 5.999288616810603e-05, "loss": 0.5221, "step": 283 }, { "epoch": 0.03, "learning_rate": 5.9992833958749643e-05, "loss": 0.8052, "step": 284 }, { "epoch": 0.03, "learning_rate": 5.999278155853027e-05, "loss": 1.3484, "step": 285 }, { "epoch": 0.03, "learning_rate": 5.9992728967448244e-05, "loss": 1.0481, "step": 286 }, { "epoch": 0.03, "learning_rate": 5.999267618550391e-05, "loss": 0.7045, "step": 287 }, { "epoch": 0.03, "learning_rate": 5.9992623212697585e-05, "loss": 0.6667, "step": 288 }, { "epoch": 0.03, "learning_rate": 5.9992570049029625e-05, "loss": 0.1496, "step": 289 }, { "epoch": 0.03, "learning_rate": 5.999251669450035e-05, "loss": 1.2328, "step": 290 }, { "epoch": 0.03, "learning_rate": 5.9992463149110105e-05, "loss": 0.6702, "step": 291 }, { "epoch": 0.03, "learning_rate": 5.9992409412859244e-05, "loss": 1.5594, "step": 292 }, { "epoch": 0.03, "learning_rate": 5.9992355485748095e-05, "loss": 1.2045, "step": 293 }, { "epoch": 0.03, "learning_rate": 5.9992301367777004e-05, "loss": 1.2895, "step": 294 }, { "epoch": 0.03, "learning_rate": 5.999224705894632e-05, "loss": 0.4864, "step": 295 }, { "epoch": 0.03, "learning_rate": 5.999219255925638e-05, "loss": 0.5734, "step": 296 }, { "epoch": 0.03, "learning_rate": 5.9992137868707544e-05, "loss": 0.8902, "step": 297 }, { "epoch": 0.03, "learning_rate": 5.999208298730014e-05, "loss": 0.6732, "step": 298 }, { "epoch": 0.03, "learning_rate": 5.999202791503454e-05, "loss": 0.34, "step": 299 }, { "epoch": 0.03, "learning_rate": 5.999197265191109e-05, "loss": 0.6457, "step": 300 }, { "epoch": 0.03, "learning_rate": 5.999191719793012e-05, "loss": 0.3467, "step": 301 }, { "epoch": 0.03, "learning_rate": 5.9991861553092014e-05, "loss": 0.5989, "step": 302 }, { "epoch": 0.03, "learning_rate": 5.999180571739709e-05, "loss": 0.9266, "step": 303 }, { "epoch": 0.03, "learning_rate": 5.999174969084574e-05, "loss": 0.8037, "step": 304 }, { "epoch": 0.03, "learning_rate": 5.9991693473438296e-05, "loss": 0.6816, "step": 305 }, { "epoch": 0.03, "learning_rate": 5.999163706517513e-05, "loss": 0.7889, "step": 306 }, { "epoch": 0.03, "learning_rate": 5.99915804660566e-05, "loss": 0.8056, "step": 307 }, { "epoch": 0.03, "learning_rate": 5.9991523676083056e-05, "loss": 1.0151, "step": 308 }, { "epoch": 0.03, "learning_rate": 5.999146669525486e-05, "loss": 0.435, "step": 309 }, { "epoch": 0.03, "learning_rate": 5.9991409523572384e-05, "loss": 0.3535, "step": 310 }, { "epoch": 0.03, "learning_rate": 5.9991352161035985e-05, "loss": 1.1287, "step": 311 }, { "epoch": 0.03, "learning_rate": 5.999129460764603e-05, "loss": 0.3947, "step": 312 }, { "epoch": 0.03, "learning_rate": 5.999123686340289e-05, "loss": 0.4871, "step": 313 }, { "epoch": 0.03, "learning_rate": 5.999117892830692e-05, "loss": 1.2137, "step": 314 }, { "epoch": 0.03, "learning_rate": 5.99911208023585e-05, "loss": 0.2265, "step": 315 }, { "epoch": 0.03, "learning_rate": 5.9991062485557996e-05, "loss": 0.2442, "step": 316 }, { "epoch": 0.03, "learning_rate": 5.999100397790578e-05, "loss": 0.3961, "step": 317 }, { "epoch": 0.03, "learning_rate": 5.9990945279402225e-05, "loss": 1.456, "step": 318 }, { "epoch": 0.03, "learning_rate": 5.9990886390047705e-05, "loss": 0.8236, "step": 319 }, { "epoch": 0.03, "learning_rate": 5.999082730984259e-05, "loss": 0.5154, "step": 320 }, { "epoch": 0.03, "learning_rate": 5.999076803878726e-05, "loss": 1.3931, "step": 321 }, { "epoch": 0.03, "learning_rate": 5.999070857688209e-05, "loss": 0.3635, "step": 322 }, { "epoch": 0.03, "learning_rate": 5.999064892412746e-05, "loss": 1.1244, "step": 323 }, { "epoch": 0.03, "learning_rate": 5.999058908052375e-05, "loss": 0.5527, "step": 324 }, { "epoch": 0.03, "learning_rate": 5.999052904607134e-05, "loss": 1.3549, "step": 325 }, { "epoch": 0.03, "learning_rate": 5.9990468820770616e-05, "loss": 0.559, "step": 326 }, { "epoch": 0.03, "learning_rate": 5.999040840462195e-05, "loss": 0.3449, "step": 327 }, { "epoch": 0.03, "learning_rate": 5.999034779762574e-05, "loss": 0.3137, "step": 328 }, { "epoch": 0.03, "learning_rate": 5.999028699978237e-05, "loss": 0.8157, "step": 329 }, { "epoch": 0.03, "learning_rate": 5.999022601109221e-05, "loss": 1.4109, "step": 330 }, { "epoch": 0.03, "learning_rate": 5.9990164831555665e-05, "loss": 1.2802, "step": 331 }, { "epoch": 0.03, "learning_rate": 5.999010346117313e-05, "loss": 0.3149, "step": 332 }, { "epoch": 0.03, "learning_rate": 5.9990041899944966e-05, "loss": 1.3209, "step": 333 }, { "epoch": 0.03, "learning_rate": 5.9989980147871595e-05, "loss": 0.8334, "step": 334 }, { "epoch": 0.03, "learning_rate": 5.99899182049534e-05, "loss": 0.9313, "step": 335 }, { "epoch": 0.03, "learning_rate": 5.9989856071190774e-05, "loss": 0.5031, "step": 336 }, { "epoch": 0.03, "learning_rate": 5.998979374658411e-05, "loss": 0.7843, "step": 337 }, { "epoch": 0.03, "learning_rate": 5.9989731231133815e-05, "loss": 0.4697, "step": 338 }, { "epoch": 0.03, "learning_rate": 5.998966852484027e-05, "loss": 0.6147, "step": 339 }, { "epoch": 0.03, "learning_rate": 5.998960562770389e-05, "loss": 0.3488, "step": 340 }, { "epoch": 0.03, "learning_rate": 5.9989542539725064e-05, "loss": 1.2117, "step": 341 }, { "epoch": 0.03, "learning_rate": 5.99894792609042e-05, "loss": 0.8119, "step": 342 }, { "epoch": 0.03, "learning_rate": 5.998941579124169e-05, "loss": 0.0953, "step": 343 }, { "epoch": 0.03, "learning_rate": 5.998935213073795e-05, "loss": 1.0257, "step": 344 }, { "epoch": 0.04, "learning_rate": 5.998928827939339e-05, "loss": 0.6113, "step": 345 }, { "epoch": 0.04, "learning_rate": 5.99892242372084e-05, "loss": 0.7502, "step": 346 }, { "epoch": 0.04, "learning_rate": 5.99891600041834e-05, "loss": 0.5541, "step": 347 }, { "epoch": 0.04, "learning_rate": 5.99890955803188e-05, "loss": 0.4839, "step": 348 }, { "epoch": 0.04, "learning_rate": 5.998903096561499e-05, "loss": 1.1221, "step": 349 }, { "epoch": 0.04, "learning_rate": 5.99889661600724e-05, "loss": 0.7463, "step": 350 }, { "epoch": 0.04, "learning_rate": 5.998890116369145e-05, "loss": 0.6099, "step": 351 }, { "epoch": 0.04, "learning_rate": 5.998883597647253e-05, "loss": 0.8686, "step": 352 }, { "epoch": 0.04, "learning_rate": 5.998877059841607e-05, "loss": 0.3367, "step": 353 }, { "epoch": 0.04, "learning_rate": 5.998870502952247e-05, "loss": 0.3661, "step": 354 }, { "epoch": 0.04, "learning_rate": 5.9988639269792174e-05, "loss": 0.4266, "step": 355 }, { "epoch": 0.04, "learning_rate": 5.998857331922558e-05, "loss": 0.7553, "step": 356 }, { "epoch": 0.04, "learning_rate": 5.998850717782312e-05, "loss": 1.442, "step": 357 }, { "epoch": 0.04, "learning_rate": 5.99884408455852e-05, "loss": 1.8646, "step": 358 }, { "epoch": 0.04, "learning_rate": 5.998837432251225e-05, "loss": 0.5022, "step": 359 }, { "epoch": 0.04, "learning_rate": 5.99883076086047e-05, "loss": 1.053, "step": 360 }, { "epoch": 0.04, "learning_rate": 5.9988240703862965e-05, "loss": 0.993, "step": 361 }, { "epoch": 0.04, "learning_rate": 5.998817360828748e-05, "loss": 0.386, "step": 362 }, { "epoch": 0.04, "learning_rate": 5.998810632187866e-05, "loss": 1.0764, "step": 363 }, { "epoch": 0.04, "learning_rate": 5.9988038844636935e-05, "loss": 0.6977, "step": 364 }, { "epoch": 0.04, "learning_rate": 5.9987971176562754e-05, "loss": 0.4798, "step": 365 }, { "epoch": 0.04, "learning_rate": 5.9987903317656515e-05, "loss": 0.8181, "step": 366 }, { "epoch": 0.04, "learning_rate": 5.998783526791868e-05, "loss": 0.5464, "step": 367 }, { "epoch": 0.04, "learning_rate": 5.998776702734966e-05, "loss": 0.8875, "step": 368 }, { "epoch": 0.04, "learning_rate": 5.9987698595949914e-05, "loss": 1.2418, "step": 369 }, { "epoch": 0.04, "learning_rate": 5.9987629973719844e-05, "loss": 0.7488, "step": 370 }, { "epoch": 0.04, "learning_rate": 5.998756116065992e-05, "loss": 0.8514, "step": 371 }, { "epoch": 0.04, "learning_rate": 5.998749215677055e-05, "loss": 0.3395, "step": 372 }, { "epoch": 0.04, "learning_rate": 5.9987422962052196e-05, "loss": 1.1217, "step": 373 }, { "epoch": 0.04, "learning_rate": 5.9987353576505294e-05, "loss": 1.093, "step": 374 }, { "epoch": 0.04, "learning_rate": 5.998728400013028e-05, "loss": 0.2827, "step": 375 }, { "epoch": 0.04, "learning_rate": 5.998721423292759e-05, "loss": 1.1022, "step": 376 }, { "epoch": 0.04, "learning_rate": 5.998714427489768e-05, "loss": 0.2735, "step": 377 }, { "epoch": 0.04, "learning_rate": 5.998707412604099e-05, "loss": 0.5297, "step": 378 }, { "epoch": 0.04, "learning_rate": 5.998700378635797e-05, "loss": 0.2697, "step": 379 }, { "epoch": 0.04, "learning_rate": 5.998693325584907e-05, "loss": 0.4194, "step": 380 }, { "epoch": 0.04, "learning_rate": 5.9986862534514737e-05, "loss": 0.374, "step": 381 }, { "epoch": 0.04, "learning_rate": 5.9986791622355407e-05, "loss": 0.8277, "step": 382 }, { "epoch": 0.04, "learning_rate": 5.998672051937154e-05, "loss": 1.4296, "step": 383 }, { "epoch": 0.04, "learning_rate": 5.998664922556361e-05, "loss": 0.3553, "step": 384 }, { "epoch": 0.04, "learning_rate": 5.998657774093203e-05, "loss": 0.2792, "step": 385 }, { "epoch": 0.04, "learning_rate": 5.9986506065477296e-05, "loss": 0.7573, "step": 386 }, { "epoch": 0.04, "learning_rate": 5.998643419919984e-05, "loss": 0.8796, "step": 387 }, { "epoch": 0.04, "learning_rate": 5.998636214210011e-05, "loss": 0.995, "step": 388 }, { "epoch": 0.04, "learning_rate": 5.99862898941786e-05, "loss": 0.3223, "step": 389 }, { "epoch": 0.04, "learning_rate": 5.998621745543573e-05, "loss": 0.274, "step": 390 }, { "epoch": 0.04, "learning_rate": 5.998614482587199e-05, "loss": 1.292, "step": 391 }, { "epoch": 0.04, "learning_rate": 5.9986072005487826e-05, "loss": 1.0212, "step": 392 }, { "epoch": 0.04, "learning_rate": 5.9985998994283715e-05, "loss": 1.0004, "step": 393 }, { "epoch": 0.04, "learning_rate": 5.99859257922601e-05, "loss": 1.0147, "step": 394 }, { "epoch": 0.04, "learning_rate": 5.998585239941747e-05, "loss": 0.4574, "step": 395 }, { "epoch": 0.04, "learning_rate": 5.998577881575628e-05, "loss": 0.9685, "step": 396 }, { "epoch": 0.04, "learning_rate": 5.998570504127701e-05, "loss": 0.79, "step": 397 }, { "epoch": 0.04, "learning_rate": 5.998563107598011e-05, "loss": 0.951, "step": 398 }, { "epoch": 0.04, "learning_rate": 5.998555691986606e-05, "loss": 0.7738, "step": 399 }, { "epoch": 0.04, "learning_rate": 5.998548257293534e-05, "loss": 0.2533, "step": 400 }, { "epoch": 0.04, "learning_rate": 5.998540803518842e-05, "loss": 0.5555, "step": 401 }, { "epoch": 0.04, "learning_rate": 5.998533330662576e-05, "loss": 0.6275, "step": 402 }, { "epoch": 0.04, "learning_rate": 5.9985258387247846e-05, "loss": 0.6821, "step": 403 }, { "epoch": 0.04, "learning_rate": 5.9985183277055156e-05, "loss": 0.8291, "step": 404 }, { "epoch": 0.04, "learning_rate": 5.998510797604816e-05, "loss": 0.7464, "step": 405 }, { "epoch": 0.04, "learning_rate": 5.998503248422736e-05, "loss": 1.3062, "step": 406 }, { "epoch": 0.04, "learning_rate": 5.998495680159321e-05, "loss": 0.8968, "step": 407 }, { "epoch": 0.04, "learning_rate": 5.9984880928146206e-05, "loss": 0.513, "step": 408 }, { "epoch": 0.04, "learning_rate": 5.998480486388682e-05, "loss": 1.023, "step": 409 }, { "epoch": 0.04, "learning_rate": 5.9984728608815554e-05, "loss": 1.3292, "step": 410 }, { "epoch": 0.04, "learning_rate": 5.998465216293287e-05, "loss": 0.8223, "step": 411 }, { "epoch": 0.04, "learning_rate": 5.998457552623927e-05, "loss": 0.431, "step": 412 }, { "epoch": 0.04, "learning_rate": 5.9984498698735235e-05, "loss": 0.7604, "step": 413 }, { "epoch": 0.04, "learning_rate": 5.998442168042127e-05, "loss": 0.9758, "step": 414 }, { "epoch": 0.04, "learning_rate": 5.9984344471297834e-05, "loss": 0.8111, "step": 415 }, { "epoch": 0.04, "learning_rate": 5.998426707136545e-05, "loss": 1.031, "step": 416 }, { "epoch": 0.04, "learning_rate": 5.998418948062459e-05, "loss": 0.5566, "step": 417 }, { "epoch": 0.04, "learning_rate": 5.998411169907576e-05, "loss": 1.3899, "step": 418 }, { "epoch": 0.04, "learning_rate": 5.998403372671945e-05, "loss": 0.3331, "step": 419 }, { "epoch": 0.04, "learning_rate": 5.998395556355614e-05, "loss": 1.7484, "step": 420 }, { "epoch": 0.04, "learning_rate": 5.998387720958637e-05, "loss": 0.7298, "step": 421 }, { "epoch": 0.04, "learning_rate": 5.998379866481059e-05, "loss": 0.1869, "step": 422 }, { "epoch": 0.04, "learning_rate": 5.9983719929229334e-05, "loss": 0.8926, "step": 423 }, { "epoch": 0.04, "learning_rate": 5.998364100284308e-05, "loss": 0.6108, "step": 424 }, { "epoch": 0.04, "learning_rate": 5.998356188565235e-05, "loss": 0.9012, "step": 425 }, { "epoch": 0.04, "learning_rate": 5.9983482577657635e-05, "loss": 1.1116, "step": 426 }, { "epoch": 0.04, "learning_rate": 5.9983403078859446e-05, "loss": 0.2982, "step": 427 }, { "epoch": 0.04, "learning_rate": 5.998332338925828e-05, "loss": 0.4509, "step": 428 }, { "epoch": 0.04, "learning_rate": 5.998324350885466e-05, "loss": 0.4404, "step": 429 }, { "epoch": 0.04, "learning_rate": 5.998316343764908e-05, "loss": 1.7052, "step": 430 }, { "epoch": 0.04, "learning_rate": 5.998308317564205e-05, "loss": 0.2172, "step": 431 }, { "epoch": 0.04, "learning_rate": 5.998300272283409e-05, "loss": 0.4058, "step": 432 }, { "epoch": 0.04, "learning_rate": 5.99829220792257e-05, "loss": 0.3381, "step": 433 }, { "epoch": 0.04, "learning_rate": 5.998284124481741e-05, "loss": 0.9059, "step": 434 }, { "epoch": 0.04, "learning_rate": 5.998276021960971e-05, "loss": 0.6692, "step": 435 }, { "epoch": 0.04, "learning_rate": 5.998267900360315e-05, "loss": 0.5322, "step": 436 }, { "epoch": 0.04, "learning_rate": 5.998259759679821e-05, "loss": 0.4405, "step": 437 }, { "epoch": 0.04, "learning_rate": 5.998251599919543e-05, "loss": 0.475, "step": 438 }, { "epoch": 0.04, "learning_rate": 5.998243421079532e-05, "loss": 0.8556, "step": 439 }, { "epoch": 0.04, "learning_rate": 5.998235223159841e-05, "loss": 1.2924, "step": 440 }, { "epoch": 0.04, "learning_rate": 5.9982270061605216e-05, "loss": 1.229, "step": 441 }, { "epoch": 0.04, "learning_rate": 5.9982187700816264e-05, "loss": 0.2064, "step": 442 }, { "epoch": 0.04, "learning_rate": 5.998210514923207e-05, "loss": 0.6372, "step": 443 }, { "epoch": 0.05, "learning_rate": 5.998202240685316e-05, "loss": 0.5239, "step": 444 }, { "epoch": 0.05, "learning_rate": 5.998193947368008e-05, "loss": 0.5744, "step": 445 }, { "epoch": 0.05, "learning_rate": 5.9981856349713336e-05, "loss": 0.9508, "step": 446 }, { "epoch": 0.05, "learning_rate": 5.998177303495346e-05, "loss": 0.6277, "step": 447 }, { "epoch": 0.05, "learning_rate": 5.9981689529400985e-05, "loss": 0.5385, "step": 448 }, { "epoch": 0.05, "learning_rate": 5.998160583305645e-05, "loss": 0.1886, "step": 449 }, { "epoch": 0.05, "learning_rate": 5.998152194592038e-05, "loss": 0.8452, "step": 450 }, { "epoch": 0.05, "learning_rate": 5.998143786799331e-05, "loss": 0.8216, "step": 451 }, { "epoch": 0.05, "learning_rate": 5.998135359927577e-05, "loss": 0.5168, "step": 452 }, { "epoch": 0.05, "learning_rate": 5.99812691397683e-05, "loss": 0.9705, "step": 453 }, { "epoch": 0.05, "learning_rate": 5.998118448947144e-05, "loss": 0.7036, "step": 454 }, { "epoch": 0.05, "learning_rate": 5.998109964838574e-05, "loss": 0.8518, "step": 455 }, { "epoch": 0.05, "learning_rate": 5.9981014616511716e-05, "loss": 1.0004, "step": 456 }, { "epoch": 0.05, "learning_rate": 5.9980929393849924e-05, "loss": 1.5896, "step": 457 }, { "epoch": 0.05, "learning_rate": 5.99808439804009e-05, "loss": 1.1848, "step": 458 }, { "epoch": 0.05, "learning_rate": 5.998075837616519e-05, "loss": 0.9662, "step": 459 }, { "epoch": 0.05, "learning_rate": 5.998067258114334e-05, "loss": 0.3593, "step": 460 }, { "epoch": 0.05, "learning_rate": 5.998058659533589e-05, "loss": 0.6831, "step": 461 }, { "epoch": 0.05, "learning_rate": 5.9980500418743395e-05, "loss": 1.1698, "step": 462 }, { "epoch": 0.05, "learning_rate": 5.9980414051366406e-05, "loss": 0.2089, "step": 463 }, { "epoch": 0.05, "learning_rate": 5.9980327493205465e-05, "loss": 0.7581, "step": 464 }, { "epoch": 0.05, "learning_rate": 5.998024074426112e-05, "loss": 1.469, "step": 465 }, { "epoch": 0.05, "learning_rate": 5.9980153804533927e-05, "loss": 0.57, "step": 466 }, { "epoch": 0.05, "learning_rate": 5.998006667402445e-05, "loss": 0.3228, "step": 467 }, { "epoch": 0.05, "learning_rate": 5.9979979352733224e-05, "loss": 0.8751, "step": 468 }, { "epoch": 0.05, "learning_rate": 5.997989184066082e-05, "loss": 0.6136, "step": 469 }, { "epoch": 0.05, "learning_rate": 5.9979804137807784e-05, "loss": 0.7279, "step": 470 }, { "epoch": 0.05, "learning_rate": 5.9979716244174684e-05, "loss": 1.1365, "step": 471 }, { "epoch": 0.05, "learning_rate": 5.997962815976207e-05, "loss": 0.6784, "step": 472 }, { "epoch": 0.05, "learning_rate": 5.997953988457051e-05, "loss": 1.0689, "step": 473 }, { "epoch": 0.05, "learning_rate": 5.997945141860056e-05, "loss": 0.6053, "step": 474 }, { "epoch": 0.05, "learning_rate": 5.997936276185279e-05, "loss": 0.5844, "step": 475 }, { "epoch": 0.05, "learning_rate": 5.9979273914327766e-05, "loss": 0.3803, "step": 476 }, { "epoch": 0.05, "learning_rate": 5.9979184876026044e-05, "loss": 0.4299, "step": 477 }, { "epoch": 0.05, "learning_rate": 5.997909564694818e-05, "loss": 1.2806, "step": 478 }, { "epoch": 0.05, "learning_rate": 5.997900622709478e-05, "loss": 0.9993, "step": 479 }, { "epoch": 0.05, "learning_rate": 5.997891661646637e-05, "loss": 0.4537, "step": 480 }, { "epoch": 0.05, "learning_rate": 5.997882681506355e-05, "loss": 0.5847, "step": 481 }, { "epoch": 0.05, "learning_rate": 5.997873682288688e-05, "loss": 0.9249, "step": 482 }, { "epoch": 0.05, "learning_rate": 5.997864663993692e-05, "loss": 0.4477, "step": 483 }, { "epoch": 0.05, "learning_rate": 5.997855626621427e-05, "loss": 0.665, "step": 484 }, { "epoch": 0.05, "learning_rate": 5.997846570171949e-05, "loss": 0.8012, "step": 485 }, { "epoch": 0.05, "learning_rate": 5.997837494645316e-05, "loss": 0.5784, "step": 486 }, { "epoch": 0.05, "learning_rate": 5.997828400041586e-05, "loss": 1.209, "step": 487 }, { "epoch": 0.05, "learning_rate": 5.9978192863608155e-05, "loss": 1.1913, "step": 488 }, { "epoch": 0.05, "learning_rate": 5.9978101536030635e-05, "loss": 0.418, "step": 489 }, { "epoch": 0.05, "learning_rate": 5.997801001768389e-05, "loss": 0.4858, "step": 490 }, { "epoch": 0.05, "learning_rate": 5.997791830856848e-05, "loss": 1.2531, "step": 491 }, { "epoch": 0.05, "learning_rate": 5.997782640868501e-05, "loss": 0.4787, "step": 492 }, { "epoch": 0.05, "learning_rate": 5.997773431803406e-05, "loss": 0.5832, "step": 493 }, { "epoch": 0.05, "learning_rate": 5.997764203661622e-05, "loss": 0.2085, "step": 494 }, { "epoch": 0.05, "learning_rate": 5.997754956443205e-05, "loss": 0.4369, "step": 495 }, { "epoch": 0.05, "learning_rate": 5.9977456901482176e-05, "loss": 0.415, "step": 496 }, { "epoch": 0.05, "learning_rate": 5.997736404776716e-05, "loss": 1.1403, "step": 497 }, { "epoch": 0.05, "learning_rate": 5.9977271003287606e-05, "loss": 0.4629, "step": 498 }, { "epoch": 0.05, "learning_rate": 5.9977177768044105e-05, "loss": 0.519, "step": 499 }, { "epoch": 0.05, "learning_rate": 5.9977084342037245e-05, "loss": 0.3958, "step": 500 }, { "epoch": 0.05, "learning_rate": 5.997699072526763e-05, "loss": 0.9738, "step": 501 }, { "epoch": 0.05, "learning_rate": 5.9976896917735845e-05, "loss": 1.2351, "step": 502 }, { "epoch": 0.05, "learning_rate": 5.997680291944249e-05, "loss": 0.3609, "step": 503 }, { "epoch": 0.05, "learning_rate": 5.997670873038817e-05, "loss": 0.5062, "step": 504 }, { "epoch": 0.05, "learning_rate": 5.9976614350573475e-05, "loss": 0.199, "step": 505 }, { "epoch": 0.05, "learning_rate": 5.997651977999901e-05, "loss": 0.471, "step": 506 }, { "epoch": 0.05, "learning_rate": 5.9976425018665386e-05, "loss": 1.3523, "step": 507 }, { "epoch": 0.05, "learning_rate": 5.9976330066573184e-05, "loss": 1.383, "step": 508 }, { "epoch": 0.05, "learning_rate": 5.997623492372303e-05, "loss": 0.6397, "step": 509 }, { "epoch": 0.05, "learning_rate": 5.997613959011552e-05, "loss": 0.9004, "step": 510 }, { "epoch": 0.05, "learning_rate": 5.9976044065751254e-05, "loss": 0.6136, "step": 511 }, { "epoch": 0.05, "learning_rate": 5.997594835063085e-05, "loss": 0.198, "step": 512 }, { "epoch": 0.05, "learning_rate": 5.9975852444754915e-05, "loss": 0.4103, "step": 513 }, { "epoch": 0.05, "learning_rate": 5.9975756348124066e-05, "loss": 0.4988, "step": 514 }, { "epoch": 0.05, "learning_rate": 5.9975660060738896e-05, "loss": 1.1275, "step": 515 }, { "epoch": 0.05, "learning_rate": 5.9975563582600036e-05, "loss": 0.8931, "step": 516 }, { "epoch": 0.05, "learning_rate": 5.997546691370809e-05, "loss": 0.8093, "step": 517 }, { "epoch": 0.05, "learning_rate": 5.9975370054063666e-05, "loss": 1.2331, "step": 518 }, { "epoch": 0.05, "learning_rate": 5.99752730036674e-05, "loss": 0.8776, "step": 519 }, { "epoch": 0.05, "learning_rate": 5.997517576251991e-05, "loss": 0.5471, "step": 520 }, { "epoch": 0.05, "learning_rate": 5.997507833062179e-05, "loss": 1.1267, "step": 521 }, { "epoch": 0.05, "learning_rate": 5.9974980707973684e-05, "loss": 0.9827, "step": 522 }, { "epoch": 0.05, "learning_rate": 5.997488289457619e-05, "loss": 0.4691, "step": 523 }, { "epoch": 0.05, "learning_rate": 5.997478489042996e-05, "loss": 0.6836, "step": 524 }, { "epoch": 0.05, "learning_rate": 5.997468669553559e-05, "loss": 0.476, "step": 525 }, { "epoch": 0.05, "learning_rate": 5.9974588309893726e-05, "loss": 0.459, "step": 526 }, { "epoch": 0.05, "learning_rate": 5.997448973350498e-05, "loss": 0.4088, "step": 527 }, { "epoch": 0.05, "learning_rate": 5.997439096636998e-05, "loss": 1.0805, "step": 528 }, { "epoch": 0.05, "learning_rate": 5.997429200848937e-05, "loss": 0.9229, "step": 529 }, { "epoch": 0.05, "learning_rate": 5.997419285986376e-05, "loss": 0.4924, "step": 530 }, { "epoch": 0.05, "learning_rate": 5.9974093520493784e-05, "loss": 0.8146, "step": 531 }, { "epoch": 0.05, "learning_rate": 5.9973993990380087e-05, "loss": 1.4679, "step": 532 }, { "epoch": 0.05, "learning_rate": 5.99738942695233e-05, "loss": 0.5055, "step": 533 }, { "epoch": 0.05, "learning_rate": 5.9973794357924047e-05, "loss": 0.4325, "step": 534 }, { "epoch": 0.05, "learning_rate": 5.9973694255582964e-05, "loss": 0.6653, "step": 535 }, { "epoch": 0.05, "learning_rate": 5.99735939625007e-05, "loss": 0.2578, "step": 536 }, { "epoch": 0.05, "learning_rate": 5.9973493478677894e-05, "loss": 1.2039, "step": 537 }, { "epoch": 0.05, "learning_rate": 5.9973392804115166e-05, "loss": 0.9276, "step": 538 }, { "epoch": 0.05, "learning_rate": 5.9973291938813174e-05, "loss": 0.9127, "step": 539 }, { "epoch": 0.05, "learning_rate": 5.997319088277254e-05, "loss": 0.8766, "step": 540 }, { "epoch": 0.05, "learning_rate": 5.997308963599394e-05, "loss": 1.2223, "step": 541 }, { "epoch": 0.06, "learning_rate": 5.9972988198478e-05, "loss": 0.3587, "step": 542 }, { "epoch": 0.06, "learning_rate": 5.9972886570225356e-05, "loss": 1.2694, "step": 543 }, { "epoch": 0.06, "learning_rate": 5.9972784751236664e-05, "loss": 2.0048, "step": 544 }, { "epoch": 0.06, "learning_rate": 5.9972682741512585e-05, "loss": 1.2022, "step": 545 }, { "epoch": 0.06, "learning_rate": 5.997258054105374e-05, "loss": 0.2993, "step": 546 }, { "epoch": 0.06, "learning_rate": 5.9972478149860805e-05, "loss": 0.3411, "step": 547 }, { "epoch": 0.06, "learning_rate": 5.997237556793441e-05, "loss": 1.1513, "step": 548 }, { "epoch": 0.06, "learning_rate": 5.997227279527523e-05, "loss": 0.6251, "step": 549 }, { "epoch": 0.06, "learning_rate": 5.997216983188391e-05, "loss": 1.1043, "step": 550 }, { "epoch": 0.06, "learning_rate": 5.997206667776109e-05, "loss": 1.415, "step": 551 }, { "epoch": 0.06, "learning_rate": 5.997196333290744e-05, "loss": 0.3739, "step": 552 }, { "epoch": 0.06, "learning_rate": 5.997185979732363e-05, "loss": 1.4098, "step": 553 }, { "epoch": 0.06, "learning_rate": 5.99717560710103e-05, "loss": 1.183, "step": 554 }, { "epoch": 0.06, "learning_rate": 5.997165215396812e-05, "loss": 0.5319, "step": 555 }, { "epoch": 0.06, "learning_rate": 5.9971548046197744e-05, "loss": 1.1395, "step": 556 }, { "epoch": 0.06, "learning_rate": 5.997144374769984e-05, "loss": 0.1463, "step": 557 }, { "epoch": 0.06, "learning_rate": 5.997133925847506e-05, "loss": 1.1007, "step": 558 }, { "epoch": 0.06, "learning_rate": 5.9971234578524085e-05, "loss": 0.9355, "step": 559 }, { "epoch": 0.06, "learning_rate": 5.9971129707847584e-05, "loss": 0.7259, "step": 560 }, { "epoch": 0.06, "learning_rate": 5.99710246464462e-05, "loss": 0.4962, "step": 561 }, { "epoch": 0.06, "learning_rate": 5.997091939432063e-05, "loss": 1.1245, "step": 562 }, { "epoch": 0.06, "learning_rate": 5.997081395147152e-05, "loss": 1.3628, "step": 563 }, { "epoch": 0.06, "learning_rate": 5.997070831789955e-05, "loss": 1.073, "step": 564 }, { "epoch": 0.06, "learning_rate": 5.99706024936054e-05, "loss": 1.2823, "step": 565 }, { "epoch": 0.06, "learning_rate": 5.9970496478589736e-05, "loss": 0.7573, "step": 566 }, { "epoch": 0.06, "learning_rate": 5.9970390272853236e-05, "loss": 0.4683, "step": 567 }, { "epoch": 0.06, "learning_rate": 5.997028387639657e-05, "loss": 0.3766, "step": 568 }, { "epoch": 0.06, "learning_rate": 5.9970177289220424e-05, "loss": 0.5968, "step": 569 }, { "epoch": 0.06, "learning_rate": 5.997007051132546e-05, "loss": 0.9696, "step": 570 }, { "epoch": 0.06, "learning_rate": 5.996996354271238e-05, "loss": 1.3313, "step": 571 }, { "epoch": 0.06, "learning_rate": 5.9969856383381846e-05, "loss": 0.9367, "step": 572 }, { "epoch": 0.06, "learning_rate": 5.996974903333456e-05, "loss": 0.7735, "step": 573 }, { "epoch": 0.06, "learning_rate": 5.996964149257118e-05, "loss": 1.2414, "step": 574 }, { "epoch": 0.06, "learning_rate": 5.99695337610924e-05, "loss": 0.8832, "step": 575 }, { "epoch": 0.06, "learning_rate": 5.9969425838898914e-05, "loss": 0.7956, "step": 576 }, { "epoch": 0.06, "learning_rate": 5.99693177259914e-05, "loss": 0.4441, "step": 577 }, { "epoch": 0.06, "learning_rate": 5.996920942237055e-05, "loss": 0.3953, "step": 578 }, { "epoch": 0.06, "learning_rate": 5.996910092803706e-05, "loss": 1.4373, "step": 579 }, { "epoch": 0.06, "learning_rate": 5.99689922429916e-05, "loss": 0.7617, "step": 580 }, { "epoch": 0.06, "learning_rate": 5.996888336723488e-05, "loss": 0.3258, "step": 581 }, { "epoch": 0.06, "learning_rate": 5.9968774300767586e-05, "loss": 0.9018, "step": 582 }, { "epoch": 0.06, "learning_rate": 5.9968665043590414e-05, "loss": 0.7324, "step": 583 }, { "epoch": 0.06, "learning_rate": 5.996855559570406e-05, "loss": 0.5455, "step": 584 }, { "epoch": 0.06, "learning_rate": 5.996844595710922e-05, "loss": 1.1078, "step": 585 }, { "epoch": 0.06, "learning_rate": 5.996833612780658e-05, "loss": 0.4002, "step": 586 }, { "epoch": 0.06, "learning_rate": 5.9968226107796866e-05, "loss": 0.9084, "step": 587 }, { "epoch": 0.06, "learning_rate": 5.996811589708075e-05, "loss": 1.6678, "step": 588 }, { "epoch": 0.06, "learning_rate": 5.996800549565895e-05, "loss": 0.7559, "step": 589 }, { "epoch": 0.06, "learning_rate": 5.9967894903532155e-05, "loss": 0.2455, "step": 590 }, { "epoch": 0.06, "learning_rate": 5.9967784120701085e-05, "loss": 0.2278, "step": 591 }, { "epoch": 0.06, "learning_rate": 5.9967673147166436e-05, "loss": 1.1137, "step": 592 }, { "epoch": 0.06, "learning_rate": 5.996756198292891e-05, "loss": 1.0953, "step": 593 }, { "epoch": 0.06, "learning_rate": 5.9967450627989226e-05, "loss": 0.6262, "step": 594 }, { "epoch": 0.06, "learning_rate": 5.996733908234809e-05, "loss": 1.0849, "step": 595 }, { "epoch": 0.06, "learning_rate": 5.9967227346006194e-05, "loss": 0.5934, "step": 596 }, { "epoch": 0.06, "learning_rate": 5.996711541896428e-05, "loss": 1.4646, "step": 597 }, { "epoch": 0.06, "learning_rate": 5.996700330122303e-05, "loss": 0.6864, "step": 598 }, { "epoch": 0.06, "learning_rate": 5.9966890992783176e-05, "loss": 0.7407, "step": 599 }, { "epoch": 0.06, "learning_rate": 5.996677849364543e-05, "loss": 0.4342, "step": 600 }, { "epoch": 0.06, "learning_rate": 5.99666658038105e-05, "loss": 0.9907, "step": 601 }, { "epoch": 0.06, "learning_rate": 5.996655292327912e-05, "loss": 0.5706, "step": 602 }, { "epoch": 0.06, "learning_rate": 5.9966439852051986e-05, "loss": 0.8936, "step": 603 }, { "epoch": 0.06, "learning_rate": 5.996632659012983e-05, "loss": 0.5444, "step": 604 }, { "epoch": 0.06, "learning_rate": 5.996621313751338e-05, "loss": 0.6518, "step": 605 }, { "epoch": 0.06, "learning_rate": 5.9966099494203344e-05, "loss": 0.5314, "step": 606 }, { "epoch": 0.06, "learning_rate": 5.9965985660200444e-05, "loss": 0.7321, "step": 607 }, { "epoch": 0.06, "learning_rate": 5.996587163550542e-05, "loss": 0.4493, "step": 608 }, { "epoch": 0.06, "learning_rate": 5.996575742011898e-05, "loss": 0.1487, "step": 609 }, { "epoch": 0.06, "learning_rate": 5.9965643014041866e-05, "loss": 0.8076, "step": 610 }, { "epoch": 0.06, "learning_rate": 5.99655284172748e-05, "loss": 0.9151, "step": 611 }, { "epoch": 0.06, "learning_rate": 5.9965413629818506e-05, "loss": 0.5829, "step": 612 }, { "epoch": 0.06, "learning_rate": 5.9965298651673723e-05, "loss": 1.1176, "step": 613 }, { "epoch": 0.06, "learning_rate": 5.996518348284117e-05, "loss": 0.5693, "step": 614 }, { "epoch": 0.06, "learning_rate": 5.9965068123321606e-05, "loss": 0.9186, "step": 615 }, { "epoch": 0.06, "learning_rate": 5.996495257311574e-05, "loss": 0.5334, "step": 616 }, { "epoch": 0.06, "learning_rate": 5.996483683222431e-05, "loss": 0.8471, "step": 617 }, { "epoch": 0.06, "learning_rate": 5.996472090064806e-05, "loss": 1.1946, "step": 618 }, { "epoch": 0.06, "learning_rate": 5.996460477838772e-05, "loss": 0.5363, "step": 619 }, { "epoch": 0.06, "learning_rate": 5.996448846544405e-05, "loss": 1.471, "step": 620 }, { "epoch": 0.06, "learning_rate": 5.996437196181776e-05, "loss": 0.5724, "step": 621 }, { "epoch": 0.06, "learning_rate": 5.996425526750961e-05, "loss": 0.6465, "step": 622 }, { "epoch": 0.06, "learning_rate": 5.9964138382520345e-05, "loss": 0.7636, "step": 623 }, { "epoch": 0.06, "learning_rate": 5.996402130685069e-05, "loss": 0.6175, "step": 624 }, { "epoch": 0.06, "learning_rate": 5.9963904040501403e-05, "loss": 0.0177, "step": 625 }, { "epoch": 0.06, "learning_rate": 5.996378658347324e-05, "loss": 1.4002, "step": 626 }, { "epoch": 0.06, "learning_rate": 5.996366893576693e-05, "loss": 1.0862, "step": 627 }, { "epoch": 0.06, "learning_rate": 5.996355109738323e-05, "loss": 0.2438, "step": 628 }, { "epoch": 0.06, "learning_rate": 5.99634330683229e-05, "loss": 0.8852, "step": 629 }, { "epoch": 0.06, "learning_rate": 5.996331484858667e-05, "loss": 0.606, "step": 630 }, { "epoch": 0.06, "learning_rate": 5.9963196438175305e-05, "loss": 1.5637, "step": 631 }, { "epoch": 0.06, "learning_rate": 5.996307783708956e-05, "loss": 0.868, "step": 632 }, { "epoch": 0.06, "learning_rate": 5.9962959045330184e-05, "loss": 0.755, "step": 633 }, { "epoch": 0.06, "learning_rate": 5.996284006289793e-05, "loss": 1.3792, "step": 634 }, { "epoch": 0.06, "learning_rate": 5.9962720889793565e-05, "loss": 1.6055, "step": 635 }, { "epoch": 0.06, "learning_rate": 5.9962601526017845e-05, "loss": 0.815, "step": 636 }, { "epoch": 0.06, "learning_rate": 5.9962481971571526e-05, "loss": 0.9183, "step": 637 }, { "epoch": 0.06, "learning_rate": 5.9962362226455366e-05, "loss": 0.5483, "step": 638 }, { "epoch": 0.06, "learning_rate": 5.996224229067013e-05, "loss": 0.6956, "step": 639 }, { "epoch": 0.06, "learning_rate": 5.996212216421659e-05, "loss": 0.2355, "step": 640 }, { "epoch": 0.07, "learning_rate": 5.9962001847095494e-05, "loss": 1.359, "step": 641 }, { "epoch": 0.07, "learning_rate": 5.996188133930763e-05, "loss": 1.1783, "step": 642 }, { "epoch": 0.07, "learning_rate": 5.996176064085373e-05, "loss": 0.3781, "step": 643 }, { "epoch": 0.07, "learning_rate": 5.99616397517346e-05, "loss": 1.2409, "step": 644 }, { "epoch": 0.07, "learning_rate": 5.996151867195099e-05, "loss": 0.4251, "step": 645 }, { "epoch": 0.07, "learning_rate": 5.996139740150367e-05, "loss": 1.4537, "step": 646 }, { "epoch": 0.07, "learning_rate": 5.9961275940393416e-05, "loss": 1.319, "step": 647 }, { "epoch": 0.07, "learning_rate": 5.9961154288621e-05, "loss": 1.0258, "step": 648 }, { "epoch": 0.07, "learning_rate": 5.9961032446187193e-05, "loss": 0.4076, "step": 649 }, { "epoch": 0.07, "learning_rate": 5.996091041309278e-05, "loss": 0.9052, "step": 650 }, { "epoch": 0.07, "learning_rate": 5.9960788189338526e-05, "loss": 0.4149, "step": 651 }, { "epoch": 0.07, "learning_rate": 5.996066577492521e-05, "loss": 0.7334, "step": 652 }, { "epoch": 0.07, "learning_rate": 5.996054316985362e-05, "loss": 0.4792, "step": 653 }, { "epoch": 0.07, "learning_rate": 5.996042037412453e-05, "loss": 0.3469, "step": 654 }, { "epoch": 0.07, "learning_rate": 5.9960297387738725e-05, "loss": 0.9926, "step": 655 }, { "epoch": 0.07, "learning_rate": 5.9960174210696984e-05, "loss": 0.8215, "step": 656 }, { "epoch": 0.07, "learning_rate": 5.9960050843000086e-05, "loss": 0.9677, "step": 657 }, { "epoch": 0.07, "learning_rate": 5.9959927284648826e-05, "loss": 1.1923, "step": 658 }, { "epoch": 0.07, "learning_rate": 5.995980353564399e-05, "loss": 0.09, "step": 659 }, { "epoch": 0.07, "learning_rate": 5.995967959598635e-05, "loss": 0.6642, "step": 660 }, { "epoch": 0.07, "learning_rate": 5.995955546567672e-05, "loss": 0.4872, "step": 661 }, { "epoch": 0.07, "learning_rate": 5.9959431144715876e-05, "loss": 1.0875, "step": 662 }, { "epoch": 0.07, "learning_rate": 5.99593066331046e-05, "loss": 0.7099, "step": 663 }, { "epoch": 0.07, "learning_rate": 5.9959181930843694e-05, "loss": 1.2392, "step": 664 }, { "epoch": 0.07, "learning_rate": 5.995905703793395e-05, "loss": 1.2947, "step": 665 }, { "epoch": 0.07, "learning_rate": 5.995893195437617e-05, "loss": 1.2215, "step": 666 }, { "epoch": 0.07, "learning_rate": 5.9958806680171145e-05, "loss": 1.0522, "step": 667 }, { "epoch": 0.07, "learning_rate": 5.9958681215319665e-05, "loss": 0.6942, "step": 668 }, { "epoch": 0.07, "learning_rate": 5.9958555559822536e-05, "loss": 0.6779, "step": 669 }, { "epoch": 0.07, "learning_rate": 5.9958429713680564e-05, "loss": 0.6918, "step": 670 }, { "epoch": 0.07, "learning_rate": 5.9958303676894535e-05, "loss": 0.5521, "step": 671 }, { "epoch": 0.07, "learning_rate": 5.995817744946526e-05, "loss": 0.448, "step": 672 }, { "epoch": 0.07, "learning_rate": 5.9958051031393545e-05, "loss": 0.3696, "step": 673 }, { "epoch": 0.07, "learning_rate": 5.995792442268018e-05, "loss": 0.3717, "step": 674 }, { "epoch": 0.07, "learning_rate": 5.9957797623325985e-05, "loss": 0.4186, "step": 675 }, { "epoch": 0.07, "learning_rate": 5.995767063333176e-05, "loss": 1.0991, "step": 676 }, { "epoch": 0.07, "learning_rate": 5.9957543452698325e-05, "loss": 1.1699, "step": 677 }, { "epoch": 0.07, "learning_rate": 5.995741608142647e-05, "loss": 0.9235, "step": 678 }, { "epoch": 0.07, "learning_rate": 5.995728851951702e-05, "loss": 0.5803, "step": 679 }, { "epoch": 0.07, "learning_rate": 5.9957160766970776e-05, "loss": 0.1788, "step": 680 }, { "epoch": 0.07, "learning_rate": 5.9957032823788564e-05, "loss": 0.4068, "step": 681 }, { "epoch": 0.07, "learning_rate": 5.995690468997119e-05, "loss": 1.1909, "step": 682 }, { "epoch": 0.07, "learning_rate": 5.995677636551947e-05, "loss": 0.9389, "step": 683 }, { "epoch": 0.07, "learning_rate": 5.995664785043423e-05, "loss": 0.9354, "step": 684 }, { "epoch": 0.07, "learning_rate": 5.995651914471626e-05, "loss": 0.6715, "step": 685 }, { "epoch": 0.07, "learning_rate": 5.995639024836642e-05, "loss": 0.3613, "step": 686 }, { "epoch": 0.07, "learning_rate": 5.995626116138549e-05, "loss": 0.3665, "step": 687 }, { "epoch": 0.07, "learning_rate": 5.995613188377432e-05, "loss": 1.2575, "step": 688 }, { "epoch": 0.07, "learning_rate": 5.9956002415533715e-05, "loss": 0.3688, "step": 689 }, { "epoch": 0.07, "learning_rate": 5.995587275666451e-05, "loss": 0.5994, "step": 690 }, { "epoch": 0.07, "learning_rate": 5.995574290716753e-05, "loss": 1.268, "step": 691 }, { "epoch": 0.07, "learning_rate": 5.99556128670436e-05, "loss": 0.2947, "step": 692 }, { "epoch": 0.07, "learning_rate": 5.995548263629354e-05, "loss": 0.2899, "step": 693 }, { "epoch": 0.07, "learning_rate": 5.9955352214918185e-05, "loss": 0.9178, "step": 694 }, { "epoch": 0.07, "learning_rate": 5.995522160291836e-05, "loss": 0.6208, "step": 695 }, { "epoch": 0.07, "learning_rate": 5.9955090800294907e-05, "loss": 0.6876, "step": 696 }, { "epoch": 0.07, "learning_rate": 5.995495980704865e-05, "loss": 0.8101, "step": 697 }, { "epoch": 0.07, "learning_rate": 5.995482862318042e-05, "loss": 0.9553, "step": 698 }, { "epoch": 0.07, "learning_rate": 5.9954697248691064e-05, "loss": 0.2213, "step": 699 }, { "epoch": 0.07, "learning_rate": 5.99545656835814e-05, "loss": 0.9857, "step": 700 }, { "epoch": 0.07, "learning_rate": 5.9954433927852284e-05, "loss": 0.9924, "step": 701 }, { "epoch": 0.07, "learning_rate": 5.995430198150454e-05, "loss": 0.4246, "step": 702 }, { "epoch": 0.07, "learning_rate": 5.995416984453902e-05, "loss": 0.8086, "step": 703 }, { "epoch": 0.07, "learning_rate": 5.995403751695656e-05, "loss": 0.737, "step": 704 }, { "epoch": 0.07, "learning_rate": 5.995390499875799e-05, "loss": 0.2854, "step": 705 }, { "epoch": 0.07, "learning_rate": 5.995377228994417e-05, "loss": 0.4219, "step": 706 }, { "epoch": 0.07, "learning_rate": 5.9953639390515936e-05, "loss": 0.159, "step": 707 }, { "epoch": 0.07, "learning_rate": 5.9953506300474134e-05, "loss": 0.5283, "step": 708 }, { "epoch": 0.07, "learning_rate": 5.995337301981962e-05, "loss": 0.9114, "step": 709 }, { "epoch": 0.07, "learning_rate": 5.995323954855323e-05, "loss": 0.4824, "step": 710 }, { "epoch": 0.07, "learning_rate": 5.995310588667581e-05, "loss": 0.5147, "step": 711 }, { "epoch": 0.07, "learning_rate": 5.995297203418823e-05, "loss": 0.5746, "step": 712 }, { "epoch": 0.07, "learning_rate": 5.995283799109133e-05, "loss": 0.9441, "step": 713 }, { "epoch": 0.07, "learning_rate": 5.995270375738596e-05, "loss": 0.4538, "step": 714 }, { "epoch": 0.07, "learning_rate": 5.995256933307298e-05, "loss": 0.4202, "step": 715 }, { "epoch": 0.07, "learning_rate": 5.9952434718153244e-05, "loss": 0.4775, "step": 716 }, { "epoch": 0.07, "learning_rate": 5.995229991262761e-05, "loss": 0.9116, "step": 717 }, { "epoch": 0.07, "learning_rate": 5.995216491649693e-05, "loss": 0.6272, "step": 718 }, { "epoch": 0.07, "learning_rate": 5.9952029729762066e-05, "loss": 0.9807, "step": 719 }, { "epoch": 0.07, "learning_rate": 5.995189435242388e-05, "loss": 0.6383, "step": 720 }, { "epoch": 0.07, "learning_rate": 5.995175878448323e-05, "loss": 0.8616, "step": 721 }, { "epoch": 0.07, "learning_rate": 5.995162302594099e-05, "loss": 1.1022, "step": 722 }, { "epoch": 0.07, "learning_rate": 5.9951487076798e-05, "loss": 0.4044, "step": 723 }, { "epoch": 0.07, "learning_rate": 5.9951350937055155e-05, "loss": 0.5404, "step": 724 }, { "epoch": 0.07, "learning_rate": 5.995121460671331e-05, "loss": 1.1615, "step": 725 }, { "epoch": 0.07, "learning_rate": 5.995107808577332e-05, "loss": 0.6072, "step": 726 }, { "epoch": 0.07, "learning_rate": 5.9950941374236064e-05, "loss": 0.9487, "step": 727 }, { "epoch": 0.07, "learning_rate": 5.9950804472102405e-05, "loss": 0.6474, "step": 728 }, { "epoch": 0.07, "learning_rate": 5.995066737937323e-05, "loss": 1.1497, "step": 729 }, { "epoch": 0.07, "learning_rate": 5.99505300960494e-05, "loss": 0.7856, "step": 730 }, { "epoch": 0.07, "learning_rate": 5.995039262213179e-05, "loss": 1.1286, "step": 731 }, { "epoch": 0.07, "learning_rate": 5.995025495762126e-05, "loss": 1.1544, "step": 732 }, { "epoch": 0.07, "learning_rate": 5.995011710251872e-05, "loss": 0.1973, "step": 733 }, { "epoch": 0.07, "learning_rate": 5.9949979056825024e-05, "loss": 1.2409, "step": 734 }, { "epoch": 0.07, "learning_rate": 5.994984082054105e-05, "loss": 0.2227, "step": 735 }, { "epoch": 0.07, "learning_rate": 5.994970239366769e-05, "loss": 0.8987, "step": 736 }, { "epoch": 0.07, "learning_rate": 5.994956377620581e-05, "loss": 0.5902, "step": 737 }, { "epoch": 0.07, "learning_rate": 5.9949424968156305e-05, "loss": 1.4954, "step": 738 }, { "epoch": 0.08, "learning_rate": 5.994928596952005e-05, "loss": 0.3285, "step": 739 }, { "epoch": 0.08, "learning_rate": 5.994914678029793e-05, "loss": 0.9254, "step": 740 }, { "epoch": 0.08, "learning_rate": 5.994900740049084e-05, "loss": 0.6972, "step": 741 }, { "epoch": 0.08, "learning_rate": 5.9948867830099655e-05, "loss": 0.6307, "step": 742 }, { "epoch": 0.08, "learning_rate": 5.9948728069125275e-05, "loss": 0.7056, "step": 743 }, { "epoch": 0.08, "learning_rate": 5.994858811756857e-05, "loss": 0.4869, "step": 744 }, { "epoch": 0.08, "learning_rate": 5.994844797543045e-05, "loss": 0.1569, "step": 745 }, { "epoch": 0.08, "learning_rate": 5.9948307642711805e-05, "loss": 0.9653, "step": 746 }, { "epoch": 0.08, "learning_rate": 5.994816711941351e-05, "loss": 0.5782, "step": 747 }, { "epoch": 0.08, "learning_rate": 5.994802640553649e-05, "loss": 0.4263, "step": 748 }, { "epoch": 0.08, "learning_rate": 5.994788550108161e-05, "loss": 1.4481, "step": 749 }, { "epoch": 0.08, "learning_rate": 5.994774440604979e-05, "loss": 1.0927, "step": 750 }, { "epoch": 0.08, "learning_rate": 5.994760312044191e-05, "loss": 0.5295, "step": 751 }, { "epoch": 0.08, "learning_rate": 5.994746164425887e-05, "loss": 1.6251, "step": 752 }, { "epoch": 0.08, "learning_rate": 5.994731997750159e-05, "loss": 0.8541, "step": 753 }, { "epoch": 0.08, "learning_rate": 5.9947178120170943e-05, "loss": 1.2485, "step": 754 }, { "epoch": 0.08, "learning_rate": 5.994703607226787e-05, "loss": 0.1555, "step": 755 }, { "epoch": 0.08, "learning_rate": 5.994689383379323e-05, "loss": 1.304, "step": 756 }, { "epoch": 0.08, "learning_rate": 5.9946751404747964e-05, "loss": 0.6735, "step": 757 }, { "epoch": 0.08, "learning_rate": 5.994660878513295e-05, "loss": 1.1749, "step": 758 }, { "epoch": 0.08, "learning_rate": 5.994646597494912e-05, "loss": 0.8336, "step": 759 }, { "epoch": 0.08, "learning_rate": 5.9946322974197366e-05, "loss": 1.3057, "step": 760 }, { "epoch": 0.08, "learning_rate": 5.994617978287862e-05, "loss": 1.2463, "step": 761 }, { "epoch": 0.08, "learning_rate": 5.994603640099376e-05, "loss": 0.8107, "step": 762 }, { "epoch": 0.08, "learning_rate": 5.9945892828543725e-05, "loss": 1.2675, "step": 763 }, { "epoch": 0.08, "learning_rate": 5.994574906552942e-05, "loss": 1.3613, "step": 764 }, { "epoch": 0.08, "learning_rate": 5.994560511195176e-05, "loss": 1.1386, "step": 765 }, { "epoch": 0.08, "learning_rate": 5.9945460967811654e-05, "loss": 0.9878, "step": 766 }, { "epoch": 0.08, "learning_rate": 5.994531663311004e-05, "loss": 0.839, "step": 767 }, { "epoch": 0.08, "learning_rate": 5.994517210784781e-05, "loss": 0.9265, "step": 768 }, { "epoch": 0.08, "learning_rate": 5.99450273920259e-05, "loss": 0.8044, "step": 769 }, { "epoch": 0.08, "learning_rate": 5.9944882485645236e-05, "loss": 0.2424, "step": 770 }, { "epoch": 0.08, "learning_rate": 5.9944737388706716e-05, "loss": 0.4956, "step": 771 }, { "epoch": 0.08, "learning_rate": 5.9944592101211296e-05, "loss": 0.6453, "step": 772 }, { "epoch": 0.08, "learning_rate": 5.994444662315987e-05, "loss": 1.1616, "step": 773 }, { "epoch": 0.08, "learning_rate": 5.994430095455338e-05, "loss": 0.9172, "step": 774 }, { "epoch": 0.08, "learning_rate": 5.994415509539275e-05, "loss": 1.275, "step": 775 }, { "epoch": 0.08, "learning_rate": 5.9944009045678907e-05, "loss": 0.9774, "step": 776 }, { "epoch": 0.08, "learning_rate": 5.994386280541279e-05, "loss": 0.8119, "step": 777 }, { "epoch": 0.08, "learning_rate": 5.994371637459531e-05, "loss": 0.78, "step": 778 }, { "epoch": 0.08, "learning_rate": 5.9943569753227425e-05, "loss": 0.4432, "step": 779 }, { "epoch": 0.08, "learning_rate": 5.9943422941310045e-05, "loss": 0.9889, "step": 780 }, { "epoch": 0.08, "learning_rate": 5.9943275938844106e-05, "loss": 0.8009, "step": 781 }, { "epoch": 0.08, "learning_rate": 5.994312874583056e-05, "loss": 0.6015, "step": 782 }, { "epoch": 0.08, "learning_rate": 5.994298136227032e-05, "loss": 1.2429, "step": 783 }, { "epoch": 0.08, "learning_rate": 5.9942833788164356e-05, "loss": 0.9151, "step": 784 }, { "epoch": 0.08, "learning_rate": 5.994268602351358e-05, "loss": 0.3077, "step": 785 }, { "epoch": 0.08, "learning_rate": 5.9942538068318935e-05, "loss": 0.8845, "step": 786 }, { "epoch": 0.08, "learning_rate": 5.9942389922581375e-05, "loss": 0.2527, "step": 787 }, { "epoch": 0.08, "learning_rate": 5.9942241586301836e-05, "loss": 0.6106, "step": 788 }, { "epoch": 0.08, "learning_rate": 5.994209305948126e-05, "loss": 0.8529, "step": 789 }, { "epoch": 0.08, "learning_rate": 5.9941944342120595e-05, "loss": 0.6385, "step": 790 }, { "epoch": 0.08, "learning_rate": 5.9941795434220784e-05, "loss": 0.8685, "step": 791 }, { "epoch": 0.08, "learning_rate": 5.994164633578278e-05, "loss": 0.4938, "step": 792 }, { "epoch": 0.08, "learning_rate": 5.9941497046807536e-05, "loss": 1.0358, "step": 793 }, { "epoch": 0.08, "learning_rate": 5.994134756729598e-05, "loss": 1.0513, "step": 794 }, { "epoch": 0.08, "learning_rate": 5.99411978972491e-05, "loss": 0.5238, "step": 795 }, { "epoch": 0.08, "learning_rate": 5.9941048036667804e-05, "loss": 0.8088, "step": 796 }, { "epoch": 0.08, "learning_rate": 5.994089798555308e-05, "loss": 0.7233, "step": 797 }, { "epoch": 0.08, "learning_rate": 5.994074774390587e-05, "loss": 1.0202, "step": 798 }, { "epoch": 0.08, "learning_rate": 5.994059731172713e-05, "loss": 1.108, "step": 799 }, { "epoch": 0.08, "learning_rate": 5.9940446689017825e-05, "loss": 0.6646, "step": 800 }, { "epoch": 0.08, "learning_rate": 5.99402958757789e-05, "loss": 0.6832, "step": 801 }, { "epoch": 0.08, "learning_rate": 5.9940144872011325e-05, "loss": 1.2159, "step": 802 }, { "epoch": 0.08, "learning_rate": 5.9939993677716055e-05, "loss": 0.8498, "step": 803 }, { "epoch": 0.08, "learning_rate": 5.993984229289405e-05, "loss": 0.4032, "step": 804 }, { "epoch": 0.08, "learning_rate": 5.993969071754629e-05, "loss": 0.7459, "step": 805 }, { "epoch": 0.08, "learning_rate": 5.993953895167372e-05, "loss": 0.2889, "step": 806 }, { "epoch": 0.08, "learning_rate": 5.993938699527732e-05, "loss": 1.0238, "step": 807 }, { "epoch": 0.08, "learning_rate": 5.993923484835804e-05, "loss": 0.9424, "step": 808 }, { "epoch": 0.08, "learning_rate": 5.993908251091687e-05, "loss": 0.7954, "step": 809 }, { "epoch": 0.08, "learning_rate": 5.993892998295477e-05, "loss": 1.1163, "step": 810 }, { "epoch": 0.08, "learning_rate": 5.99387772644727e-05, "loss": 1.0896, "step": 811 }, { "epoch": 0.08, "learning_rate": 5.993862435547164e-05, "loss": 0.7397, "step": 812 }, { "epoch": 0.08, "learning_rate": 5.993847125595258e-05, "loss": 0.4138, "step": 813 }, { "epoch": 0.08, "learning_rate": 5.9938317965916456e-05, "loss": 0.3417, "step": 814 }, { "epoch": 0.08, "learning_rate": 5.9938164485364275e-05, "loss": 0.2841, "step": 815 }, { "epoch": 0.08, "learning_rate": 5.9938010814297016e-05, "loss": 0.4741, "step": 816 }, { "epoch": 0.08, "learning_rate": 5.993785695271563e-05, "loss": 0.2632, "step": 817 }, { "epoch": 0.08, "learning_rate": 5.993770290062111e-05, "loss": 0.5257, "step": 818 }, { "epoch": 0.08, "learning_rate": 5.9937548658014446e-05, "loss": 1.1948, "step": 819 }, { "epoch": 0.08, "learning_rate": 5.993739422489661e-05, "loss": 0.8593, "step": 820 }, { "epoch": 0.08, "learning_rate": 5.993723960126858e-05, "loss": 0.9148, "step": 821 }, { "epoch": 0.08, "learning_rate": 5.993708478713135e-05, "loss": 0.5111, "step": 822 }, { "epoch": 0.08, "learning_rate": 5.9936929782485904e-05, "loss": 1.4134, "step": 823 }, { "epoch": 0.08, "learning_rate": 5.9936774587333216e-05, "loss": 0.7602, "step": 824 }, { "epoch": 0.08, "learning_rate": 5.993661920167429e-05, "loss": 1.3578, "step": 825 }, { "epoch": 0.08, "learning_rate": 5.993646362551011e-05, "loss": 1.2169, "step": 826 }, { "epoch": 0.08, "learning_rate": 5.993630785884167e-05, "loss": 0.2906, "step": 827 }, { "epoch": 0.08, "learning_rate": 5.993615190166994e-05, "loss": 0.6531, "step": 828 }, { "epoch": 0.08, "learning_rate": 5.9935995753995936e-05, "loss": 1.164, "step": 829 }, { "epoch": 0.08, "learning_rate": 5.993583941582064e-05, "loss": 0.4303, "step": 830 }, { "epoch": 0.08, "learning_rate": 5.9935682887145046e-05, "loss": 0.8911, "step": 831 }, { "epoch": 0.08, "learning_rate": 5.993552616797017e-05, "loss": 0.4636, "step": 832 }, { "epoch": 0.08, "learning_rate": 5.993536925829698e-05, "loss": 0.3854, "step": 833 }, { "epoch": 0.08, "learning_rate": 5.993521215812649e-05, "loss": 0.3658, "step": 834 }, { "epoch": 0.08, "learning_rate": 5.99350548674597e-05, "loss": 0.5457, "step": 835 }, { "epoch": 0.08, "learning_rate": 5.993489738629761e-05, "loss": 0.5916, "step": 836 }, { "epoch": 0.08, "learning_rate": 5.993473971464122e-05, "loss": 0.5056, "step": 837 }, { "epoch": 0.09, "learning_rate": 5.993458185249153e-05, "loss": 1.1208, "step": 838 }, { "epoch": 0.09, "learning_rate": 5.993442379984955e-05, "loss": 1.0001, "step": 839 }, { "epoch": 0.09, "learning_rate": 5.9934265556716285e-05, "loss": 0.7269, "step": 840 }, { "epoch": 0.09, "learning_rate": 5.993410712309274e-05, "loss": 1.1036, "step": 841 }, { "epoch": 0.09, "learning_rate": 5.993394849897994e-05, "loss": 0.8246, "step": 842 }, { "epoch": 0.09, "learning_rate": 5.993378968437886e-05, "loss": 0.2627, "step": 843 }, { "epoch": 0.09, "learning_rate": 5.9933630679290534e-05, "loss": 0.6711, "step": 844 }, { "epoch": 0.09, "learning_rate": 5.993347148371597e-05, "loss": 1.3401, "step": 845 }, { "epoch": 0.09, "learning_rate": 5.993331209765619e-05, "loss": 0.4429, "step": 846 }, { "epoch": 0.09, "learning_rate": 5.993315252111219e-05, "loss": 0.3356, "step": 847 }, { "epoch": 0.09, "learning_rate": 5.9932992754084995e-05, "loss": 0.8205, "step": 848 }, { "epoch": 0.09, "learning_rate": 5.9932832796575626e-05, "loss": 1.0614, "step": 849 }, { "epoch": 0.09, "learning_rate": 5.993267264858509e-05, "loss": 1.1596, "step": 850 }, { "epoch": 0.09, "learning_rate": 5.993251231011441e-05, "loss": 0.3754, "step": 851 }, { "epoch": 0.09, "learning_rate": 5.993235178116462e-05, "loss": 1.0934, "step": 852 }, { "epoch": 0.09, "learning_rate": 5.993219106173672e-05, "loss": 0.5334, "step": 853 }, { "epoch": 0.09, "learning_rate": 5.993203015183175e-05, "loss": 0.6592, "step": 854 }, { "epoch": 0.09, "learning_rate": 5.993186905145072e-05, "loss": 1.2703, "step": 855 }, { "epoch": 0.09, "learning_rate": 5.9931707760594665e-05, "loss": 0.6539, "step": 856 }, { "epoch": 0.09, "learning_rate": 5.9931546279264614e-05, "loss": 0.7534, "step": 857 }, { "epoch": 0.09, "learning_rate": 5.993138460746158e-05, "loss": 0.4736, "step": 858 }, { "epoch": 0.09, "learning_rate": 5.993122274518661e-05, "loss": 0.5078, "step": 859 }, { "epoch": 0.09, "learning_rate": 5.993106069244072e-05, "loss": 0.5637, "step": 860 }, { "epoch": 0.09, "learning_rate": 5.993089844922494e-05, "loss": 1.3652, "step": 861 }, { "epoch": 0.09, "learning_rate": 5.9930736015540314e-05, "loss": 0.5501, "step": 862 }, { "epoch": 0.09, "learning_rate": 5.9930573391387875e-05, "loss": 0.9075, "step": 863 }, { "epoch": 0.09, "learning_rate": 5.9930410576768646e-05, "loss": 0.9839, "step": 864 }, { "epoch": 0.09, "learning_rate": 5.993024757168367e-05, "loss": 0.376, "step": 865 }, { "epoch": 0.09, "learning_rate": 5.9930084376133986e-05, "loss": 0.8899, "step": 866 }, { "epoch": 0.09, "learning_rate": 5.9929920990120636e-05, "loss": 1.2575, "step": 867 }, { "epoch": 0.09, "learning_rate": 5.992975741364465e-05, "loss": 0.3999, "step": 868 }, { "epoch": 0.09, "learning_rate": 5.992959364670708e-05, "loss": 1.0932, "step": 869 }, { "epoch": 0.09, "learning_rate": 5.992942968930895e-05, "loss": 1.2193, "step": 870 }, { "epoch": 0.09, "learning_rate": 5.9929265541451324e-05, "loss": 0.4893, "step": 871 }, { "epoch": 0.09, "learning_rate": 5.9929101203135236e-05, "loss": 1.2102, "step": 872 }, { "epoch": 0.09, "learning_rate": 5.992893667436174e-05, "loss": 1.0117, "step": 873 }, { "epoch": 0.09, "learning_rate": 5.992877195513187e-05, "loss": 1.0674, "step": 874 }, { "epoch": 0.09, "learning_rate": 5.992860704544668e-05, "loss": 0.7328, "step": 875 }, { "epoch": 0.09, "learning_rate": 5.992844194530721e-05, "loss": 0.4814, "step": 876 }, { "epoch": 0.09, "learning_rate": 5.992827665471453e-05, "loss": 1.2574, "step": 877 }, { "epoch": 0.09, "learning_rate": 5.9928111173669696e-05, "loss": 0.3544, "step": 878 }, { "epoch": 0.09, "learning_rate": 5.992794550217373e-05, "loss": 0.8544, "step": 879 }, { "epoch": 0.09, "learning_rate": 5.9927779640227714e-05, "loss": 0.4075, "step": 880 }, { "epoch": 0.09, "learning_rate": 5.992761358783269e-05, "loss": 0.7004, "step": 881 }, { "epoch": 0.09, "learning_rate": 5.9927447344989716e-05, "loss": 1.0467, "step": 882 }, { "epoch": 0.09, "learning_rate": 5.992728091169985e-05, "loss": 0.2034, "step": 883 }, { "epoch": 0.09, "learning_rate": 5.992711428796416e-05, "loss": 0.5745, "step": 884 }, { "epoch": 0.09, "learning_rate": 5.992694747378369e-05, "loss": 0.2963, "step": 885 }, { "epoch": 0.09, "learning_rate": 5.9926780469159524e-05, "loss": 1.6303, "step": 886 }, { "epoch": 0.09, "learning_rate": 5.9926613274092705e-05, "loss": 0.9505, "step": 887 }, { "epoch": 0.09, "learning_rate": 5.9926445888584303e-05, "loss": 1.1239, "step": 888 }, { "epoch": 0.09, "learning_rate": 5.992627831263539e-05, "loss": 0.6669, "step": 889 }, { "epoch": 0.09, "learning_rate": 5.992611054624702e-05, "loss": 0.934, "step": 890 }, { "epoch": 0.09, "learning_rate": 5.9925942589420267e-05, "loss": 0.6273, "step": 891 }, { "epoch": 0.09, "learning_rate": 5.9925774442156196e-05, "loss": 1.3053, "step": 892 }, { "epoch": 0.09, "learning_rate": 5.992560610445589e-05, "loss": 1.1022, "step": 893 }, { "epoch": 0.09, "learning_rate": 5.9925437576320406e-05, "loss": 0.1624, "step": 894 }, { "epoch": 0.09, "learning_rate": 5.992526885775083e-05, "loss": 1.0971, "step": 895 }, { "epoch": 0.09, "learning_rate": 5.992509994874821e-05, "loss": 0.4757, "step": 896 }, { "epoch": 0.09, "learning_rate": 5.992493084931365e-05, "loss": 0.4453, "step": 897 }, { "epoch": 0.09, "learning_rate": 5.9924761559448205e-05, "loss": 1.22, "step": 898 }, { "epoch": 0.09, "learning_rate": 5.992459207915297e-05, "loss": 1.5914, "step": 899 }, { "epoch": 0.09, "learning_rate": 5.992442240842902e-05, "loss": 0.6316, "step": 900 }, { "epoch": 0.09, "learning_rate": 5.992425254727742e-05, "loss": 1.5475, "step": 901 }, { "epoch": 0.09, "learning_rate": 5.992408249569926e-05, "loss": 0.4901, "step": 902 }, { "epoch": 0.09, "learning_rate": 5.992391225369563e-05, "loss": 0.3896, "step": 903 }, { "epoch": 0.09, "learning_rate": 5.99237418212676e-05, "loss": 1.3218, "step": 904 }, { "epoch": 0.09, "learning_rate": 5.9923571198416266e-05, "loss": 1.0489, "step": 905 }, { "epoch": 0.09, "learning_rate": 5.99234003851427e-05, "loss": 0.6329, "step": 906 }, { "epoch": 0.09, "learning_rate": 5.9923229381448e-05, "loss": 0.5065, "step": 907 }, { "epoch": 0.09, "learning_rate": 5.992305818733326e-05, "loss": 0.0689, "step": 908 }, { "epoch": 0.09, "learning_rate": 5.992288680279955e-05, "loss": 0.5334, "step": 909 }, { "epoch": 0.09, "learning_rate": 5.992271522784798e-05, "loss": 1.6699, "step": 910 }, { "epoch": 0.09, "learning_rate": 5.992254346247963e-05, "loss": 0.7227, "step": 911 }, { "epoch": 0.09, "learning_rate": 5.99223715066956e-05, "loss": 0.5982, "step": 912 }, { "epoch": 0.09, "learning_rate": 5.9922199360496976e-05, "loss": 0.6213, "step": 913 }, { "epoch": 0.09, "learning_rate": 5.992202702388486e-05, "loss": 0.6863, "step": 914 }, { "epoch": 0.09, "learning_rate": 5.992185449686034e-05, "loss": 0.3016, "step": 915 }, { "epoch": 0.09, "learning_rate": 5.992168177942453e-05, "loss": 0.4865, "step": 916 }, { "epoch": 0.09, "learning_rate": 5.992150887157852e-05, "loss": 0.7168, "step": 917 }, { "epoch": 0.09, "learning_rate": 5.99213357733234e-05, "loss": 1.0368, "step": 918 }, { "epoch": 0.09, "learning_rate": 5.99211624846603e-05, "loss": 0.933, "step": 919 }, { "epoch": 0.09, "learning_rate": 5.992098900559029e-05, "loss": 0.4797, "step": 920 }, { "epoch": 0.09, "learning_rate": 5.992081533611449e-05, "loss": 1.3121, "step": 921 }, { "epoch": 0.09, "learning_rate": 5.992064147623401e-05, "loss": 0.9277, "step": 922 }, { "epoch": 0.09, "learning_rate": 5.9920467425949945e-05, "loss": 1.0576, "step": 923 }, { "epoch": 0.09, "learning_rate": 5.992029318526341e-05, "loss": 1.3692, "step": 924 }, { "epoch": 0.09, "learning_rate": 5.99201187541755e-05, "loss": 0.3829, "step": 925 }, { "epoch": 0.09, "learning_rate": 5.9919944132687354e-05, "loss": 0.3077, "step": 926 }, { "epoch": 0.09, "learning_rate": 5.991976932080006e-05, "loss": 0.8876, "step": 927 }, { "epoch": 0.09, "learning_rate": 5.991959431851473e-05, "loss": 0.4811, "step": 928 }, { "epoch": 0.09, "learning_rate": 5.991941912583249e-05, "loss": 0.337, "step": 929 }, { "epoch": 0.09, "learning_rate": 5.991924374275445e-05, "loss": 1.2518, "step": 930 }, { "epoch": 0.09, "learning_rate": 5.9919068169281716e-05, "loss": 0.619, "step": 931 }, { "epoch": 0.09, "learning_rate": 5.9918892405415425e-05, "loss": 0.4147, "step": 932 }, { "epoch": 0.09, "learning_rate": 5.991871645115668e-05, "loss": 0.6999, "step": 933 }, { "epoch": 0.09, "learning_rate": 5.99185403065066e-05, "loss": 0.8617, "step": 934 }, { "epoch": 0.09, "learning_rate": 5.991836397146632e-05, "loss": 2.0689, "step": 935 }, { "epoch": 0.1, "learning_rate": 5.991818744603694e-05, "loss": 1.1669, "step": 936 }, { "epoch": 0.1, "learning_rate": 5.991801073021961e-05, "loss": 0.3879, "step": 937 }, { "epoch": 0.1, "learning_rate": 5.991783382401544e-05, "loss": 0.4889, "step": 938 }, { "epoch": 0.1, "learning_rate": 5.991765672742555e-05, "loss": 0.0959, "step": 939 }, { "epoch": 0.1, "learning_rate": 5.9917479440451074e-05, "loss": 1.3007, "step": 940 }, { "epoch": 0.1, "learning_rate": 5.991730196309315e-05, "loss": 0.9245, "step": 941 }, { "epoch": 0.1, "learning_rate": 5.991712429535288e-05, "loss": 0.5462, "step": 942 }, { "epoch": 0.1, "learning_rate": 5.991694643723142e-05, "loss": 0.5576, "step": 943 }, { "epoch": 0.1, "learning_rate": 5.99167683887299e-05, "loss": 0.7575, "step": 944 }, { "epoch": 0.1, "learning_rate": 5.991659014984945e-05, "loss": 0.7763, "step": 945 }, { "epoch": 0.1, "learning_rate": 5.9916411720591194e-05, "loss": 0.2509, "step": 946 }, { "epoch": 0.1, "learning_rate": 5.991623310095627e-05, "loss": 0.7057, "step": 947 }, { "epoch": 0.1, "learning_rate": 5.9916054290945824e-05, "loss": 0.3836, "step": 948 }, { "epoch": 0.1, "learning_rate": 5.991587529056099e-05, "loss": 0.635, "step": 949 }, { "epoch": 0.1, "learning_rate": 5.9915696099802906e-05, "loss": 0.4033, "step": 950 }, { "epoch": 0.1, "learning_rate": 5.991551671867271e-05, "loss": 0.2431, "step": 951 }, { "epoch": 0.1, "learning_rate": 5.991533714717155e-05, "loss": 1.0649, "step": 952 }, { "epoch": 0.1, "learning_rate": 5.9915157385300564e-05, "loss": 1.3552, "step": 953 }, { "epoch": 0.1, "learning_rate": 5.991497743306089e-05, "loss": 0.6536, "step": 954 }, { "epoch": 0.1, "learning_rate": 5.9914797290453686e-05, "loss": 0.5932, "step": 955 }, { "epoch": 0.1, "learning_rate": 5.9914616957480095e-05, "loss": 0.4215, "step": 956 }, { "epoch": 0.1, "learning_rate": 5.991443643414126e-05, "loss": 1.0358, "step": 957 }, { "epoch": 0.1, "learning_rate": 5.991425572043833e-05, "loss": 1.0602, "step": 958 }, { "epoch": 0.1, "learning_rate": 5.9914074816372454e-05, "loss": 0.8767, "step": 959 }, { "epoch": 0.1, "learning_rate": 5.9913893721944785e-05, "loss": 0.5317, "step": 960 }, { "epoch": 0.1, "learning_rate": 5.991371243715648e-05, "loss": 1.2056, "step": 961 }, { "epoch": 0.1, "learning_rate": 5.991353096200869e-05, "loss": 0.6534, "step": 962 }, { "epoch": 0.1, "learning_rate": 5.9913349296502564e-05, "loss": 0.9767, "step": 963 }, { "epoch": 0.1, "learning_rate": 5.991316744063926e-05, "loss": 0.9296, "step": 964 }, { "epoch": 0.1, "learning_rate": 5.991298539441995e-05, "loss": 0.9538, "step": 965 }, { "epoch": 0.1, "learning_rate": 5.991280315784578e-05, "loss": 1.4234, "step": 966 }, { "epoch": 0.1, "learning_rate": 5.9912620730917904e-05, "loss": 1.4448, "step": 967 }, { "epoch": 0.1, "learning_rate": 5.991243811363749e-05, "loss": 0.3306, "step": 968 }, { "epoch": 0.1, "learning_rate": 5.991225530600569e-05, "loss": 0.542, "step": 969 }, { "epoch": 0.1, "learning_rate": 5.991207230802369e-05, "loss": 0.579, "step": 970 }, { "epoch": 0.1, "learning_rate": 5.9911889119692635e-05, "loss": 0.438, "step": 971 }, { "epoch": 0.1, "learning_rate": 5.9911705741013694e-05, "loss": 1.1109, "step": 972 }, { "epoch": 0.1, "learning_rate": 5.991152217198804e-05, "loss": 1.5141, "step": 973 }, { "epoch": 0.1, "learning_rate": 5.991133841261684e-05, "loss": 0.6465, "step": 974 }, { "epoch": 0.1, "learning_rate": 5.991115446290126e-05, "loss": 1.3102, "step": 975 }, { "epoch": 0.1, "learning_rate": 5.9910970322842475e-05, "loss": 0.9687, "step": 976 }, { "epoch": 0.1, "learning_rate": 5.9910785992441645e-05, "loss": 0.5502, "step": 977 }, { "epoch": 0.1, "learning_rate": 5.991060147169995e-05, "loss": 0.3456, "step": 978 }, { "epoch": 0.1, "learning_rate": 5.991041676061858e-05, "loss": 1.4105, "step": 979 }, { "epoch": 0.1, "learning_rate": 5.991023185919868e-05, "loss": 0.8106, "step": 980 }, { "epoch": 0.1, "learning_rate": 5.991004676744145e-05, "loss": 0.4834, "step": 981 }, { "epoch": 0.1, "learning_rate": 5.990986148534806e-05, "loss": 1.2451, "step": 982 }, { "epoch": 0.1, "learning_rate": 5.99096760129197e-05, "loss": 0.4089, "step": 983 }, { "epoch": 0.1, "learning_rate": 5.990949035015753e-05, "loss": 1.3578, "step": 984 }, { "epoch": 0.1, "learning_rate": 5.990930449706274e-05, "loss": 1.5009, "step": 985 }, { "epoch": 0.1, "learning_rate": 5.990911845363652e-05, "loss": 1.1916, "step": 986 }, { "epoch": 0.1, "learning_rate": 5.990893221988005e-05, "loss": 0.2532, "step": 987 }, { "epoch": 0.1, "learning_rate": 5.990874579579451e-05, "loss": 1.0601, "step": 988 }, { "epoch": 0.1, "learning_rate": 5.990855918138108e-05, "loss": 0.4567, "step": 989 }, { "epoch": 0.1, "learning_rate": 5.9908372376640966e-05, "loss": 0.9463, "step": 990 }, { "epoch": 0.1, "learning_rate": 5.990818538157535e-05, "loss": 1.0196, "step": 991 }, { "epoch": 0.1, "learning_rate": 5.990799819618541e-05, "loss": 0.564, "step": 992 }, { "epoch": 0.1, "learning_rate": 5.9907810820472356e-05, "loss": 0.7695, "step": 993 }, { "epoch": 0.1, "learning_rate": 5.990762325443737e-05, "loss": 0.3217, "step": 994 }, { "epoch": 0.1, "learning_rate": 5.9907435498081656e-05, "loss": 0.3681, "step": 995 }, { "epoch": 0.1, "learning_rate": 5.9907247551406383e-05, "loss": 0.5123, "step": 996 }, { "epoch": 0.1, "learning_rate": 5.9907059414412776e-05, "loss": 0.9781, "step": 997 }, { "epoch": 0.1, "learning_rate": 5.990687108710201e-05, "loss": 0.8282, "step": 998 }, { "epoch": 0.1, "learning_rate": 5.99066825694753e-05, "loss": 1.0481, "step": 999 }, { "epoch": 0.1, "learning_rate": 5.990649386153384e-05, "loss": 1.1435, "step": 1000 }, { "epoch": 0.1, "learning_rate": 5.990630496327883e-05, "loss": 1.1217, "step": 1001 }, { "epoch": 0.1, "learning_rate": 5.990611587471147e-05, "loss": 0.4725, "step": 1002 }, { "epoch": 0.1, "learning_rate": 5.990592659583297e-05, "loss": 0.2599, "step": 1003 }, { "epoch": 0.1, "learning_rate": 5.9905737126644523e-05, "loss": 0.2956, "step": 1004 }, { "epoch": 0.1, "learning_rate": 5.9905547467147344e-05, "loss": 0.2303, "step": 1005 }, { "epoch": 0.1, "learning_rate": 5.9905357617342636e-05, "loss": 1.0216, "step": 1006 }, { "epoch": 0.1, "learning_rate": 5.9905167577231615e-05, "loss": 0.4012, "step": 1007 }, { "epoch": 0.1, "learning_rate": 5.990497734681548e-05, "loss": 0.9425, "step": 1008 }, { "epoch": 0.1, "learning_rate": 5.990478692609545e-05, "loss": 1.1702, "step": 1009 }, { "epoch": 0.1, "learning_rate": 5.990459631507272e-05, "loss": 0.7712, "step": 1010 }, { "epoch": 0.1, "learning_rate": 5.9904405513748524e-05, "loss": 0.8091, "step": 1011 }, { "epoch": 0.1, "learning_rate": 5.990421452212407e-05, "loss": 0.548, "step": 1012 }, { "epoch": 0.1, "learning_rate": 5.990402334020057e-05, "loss": 1.1248, "step": 1013 }, { "epoch": 0.1, "learning_rate": 5.9903831967979236e-05, "loss": 1.0739, "step": 1014 }, { "epoch": 0.1, "learning_rate": 5.990364040546129e-05, "loss": 1.4044, "step": 1015 }, { "epoch": 0.1, "learning_rate": 5.990344865264796e-05, "loss": 0.9397, "step": 1016 }, { "epoch": 0.1, "learning_rate": 5.990325670954046e-05, "loss": 0.6002, "step": 1017 }, { "epoch": 0.1, "learning_rate": 5.990306457614e-05, "loss": 2.0723, "step": 1018 }, { "epoch": 0.1, "learning_rate": 5.9902872252447825e-05, "loss": 0.3973, "step": 1019 }, { "epoch": 0.1, "learning_rate": 5.990267973846514e-05, "loss": 0.6684, "step": 1020 }, { "epoch": 0.1, "learning_rate": 5.990248703419318e-05, "loss": 0.9518, "step": 1021 }, { "epoch": 0.1, "learning_rate": 5.990229413963316e-05, "loss": 0.5254, "step": 1022 }, { "epoch": 0.1, "learning_rate": 5.990210105478633e-05, "loss": 0.9565, "step": 1023 }, { "epoch": 0.1, "learning_rate": 5.9901907779653884e-05, "loss": 0.8618, "step": 1024 }, { "epoch": 0.1, "learning_rate": 5.990171431423709e-05, "loss": 0.4014, "step": 1025 }, { "epoch": 0.1, "learning_rate": 5.9901520658537146e-05, "loss": 0.6232, "step": 1026 }, { "epoch": 0.1, "learning_rate": 5.990132681255531e-05, "loss": 0.9021, "step": 1027 }, { "epoch": 0.1, "learning_rate": 5.99011327762928e-05, "loss": 0.9832, "step": 1028 }, { "epoch": 0.1, "learning_rate": 5.990093854975086e-05, "loss": 1.5473, "step": 1029 }, { "epoch": 0.1, "learning_rate": 5.990074413293073e-05, "loss": 0.8174, "step": 1030 }, { "epoch": 0.1, "learning_rate": 5.990054952583362e-05, "loss": 0.5064, "step": 1031 }, { "epoch": 0.1, "learning_rate": 5.9900354728460804e-05, "loss": 0.4166, "step": 1032 }, { "epoch": 0.1, "learning_rate": 5.99001597408135e-05, "loss": 0.8929, "step": 1033 }, { "epoch": 0.1, "learning_rate": 5.989996456289295e-05, "loss": 0.7764, "step": 1034 }, { "epoch": 0.11, "learning_rate": 5.98997691947004e-05, "loss": 2.2356, "step": 1035 }, { "epoch": 0.11, "learning_rate": 5.98995736362371e-05, "loss": 1.2962, "step": 1036 }, { "epoch": 0.11, "learning_rate": 5.989937788750428e-05, "loss": 0.4391, "step": 1037 }, { "epoch": 0.11, "learning_rate": 5.98991819485032e-05, "loss": 0.9117, "step": 1038 }, { "epoch": 0.11, "learning_rate": 5.98989858192351e-05, "loss": 0.2917, "step": 1039 }, { "epoch": 0.11, "learning_rate": 5.989878949970122e-05, "loss": 1.262, "step": 1040 }, { "epoch": 0.11, "learning_rate": 5.989859298990283e-05, "loss": 0.941, "step": 1041 }, { "epoch": 0.11, "learning_rate": 5.9898396289841154e-05, "loss": 1.0029, "step": 1042 }, { "epoch": 0.11, "learning_rate": 5.989819939951747e-05, "loss": 0.2558, "step": 1043 }, { "epoch": 0.11, "learning_rate": 5.989800231893302e-05, "loss": 1.1906, "step": 1044 }, { "epoch": 0.11, "learning_rate": 5.989780504808904e-05, "loss": 0.599, "step": 1045 }, { "epoch": 0.11, "learning_rate": 5.9897607586986826e-05, "loss": 0.9021, "step": 1046 }, { "epoch": 0.11, "learning_rate": 5.98974099356276e-05, "loss": 0.8136, "step": 1047 }, { "epoch": 0.11, "learning_rate": 5.9897212094012626e-05, "loss": 1.3501, "step": 1048 }, { "epoch": 0.11, "learning_rate": 5.989701406214317e-05, "loss": 0.6223, "step": 1049 }, { "epoch": 0.11, "learning_rate": 5.989681584002049e-05, "loss": 1.1058, "step": 1050 }, { "epoch": 0.11, "learning_rate": 5.989661742764585e-05, "loss": 0.3952, "step": 1051 }, { "epoch": 0.11, "learning_rate": 5.9896418825020506e-05, "loss": 1.1468, "step": 1052 }, { "epoch": 0.11, "learning_rate": 5.989622003214573e-05, "loss": 0.4558, "step": 1053 }, { "epoch": 0.11, "learning_rate": 5.989602104902278e-05, "loss": 1.1686, "step": 1054 }, { "epoch": 0.11, "learning_rate": 5.9895821875652924e-05, "loss": 1.2434, "step": 1055 }, { "epoch": 0.11, "learning_rate": 5.989562251203744e-05, "loss": 1.3086, "step": 1056 }, { "epoch": 0.11, "learning_rate": 5.9895422958177584e-05, "loss": 1.0195, "step": 1057 }, { "epoch": 0.11, "learning_rate": 5.989522321407462e-05, "loss": 0.5461, "step": 1058 }, { "epoch": 0.11, "learning_rate": 5.989502327972983e-05, "loss": 0.6049, "step": 1059 }, { "epoch": 0.11, "learning_rate": 5.989482315514449e-05, "loss": 0.7353, "step": 1060 }, { "epoch": 0.11, "learning_rate": 5.9894622840319867e-05, "loss": 0.9036, "step": 1061 }, { "epoch": 0.11, "learning_rate": 5.9894422335257233e-05, "loss": 1.1176, "step": 1062 }, { "epoch": 0.11, "learning_rate": 5.989422163995787e-05, "loss": 1.0178, "step": 1063 }, { "epoch": 0.11, "learning_rate": 5.989402075442306e-05, "loss": 0.9615, "step": 1064 }, { "epoch": 0.11, "learning_rate": 5.9893819678654064e-05, "loss": 1.0475, "step": 1065 }, { "epoch": 0.11, "learning_rate": 5.989361841265217e-05, "loss": 0.8589, "step": 1066 }, { "epoch": 0.11, "learning_rate": 5.989341695641867e-05, "loss": 1.1158, "step": 1067 }, { "epoch": 0.11, "learning_rate": 5.989321530995483e-05, "loss": 0.3571, "step": 1068 }, { "epoch": 0.11, "learning_rate": 5.9893013473261946e-05, "loss": 0.4898, "step": 1069 }, { "epoch": 0.11, "learning_rate": 5.9892811446341295e-05, "loss": 0.2291, "step": 1070 }, { "epoch": 0.11, "learning_rate": 5.989260922919416e-05, "loss": 1.3025, "step": 1071 }, { "epoch": 0.11, "learning_rate": 5.989240682182183e-05, "loss": 0.6572, "step": 1072 }, { "epoch": 0.11, "learning_rate": 5.98922042242256e-05, "loss": 0.2266, "step": 1073 }, { "epoch": 0.11, "learning_rate": 5.989200143640675e-05, "loss": 1.293, "step": 1074 }, { "epoch": 0.11, "learning_rate": 5.9891798458366576e-05, "loss": 0.7097, "step": 1075 }, { "epoch": 0.11, "learning_rate": 5.989159529010636e-05, "loss": 0.8102, "step": 1076 }, { "epoch": 0.11, "learning_rate": 5.989139193162741e-05, "loss": 0.5723, "step": 1077 }, { "epoch": 0.11, "learning_rate": 5.989118838293102e-05, "loss": 0.6829, "step": 1078 }, { "epoch": 0.11, "learning_rate": 5.9890984644018465e-05, "loss": 0.5598, "step": 1079 }, { "epoch": 0.11, "learning_rate": 5.989078071489107e-05, "loss": 0.6182, "step": 1080 }, { "epoch": 0.11, "learning_rate": 5.9890576595550105e-05, "loss": 1.3274, "step": 1081 }, { "epoch": 0.11, "learning_rate": 5.989037228599688e-05, "loss": 1.0231, "step": 1082 }, { "epoch": 0.11, "learning_rate": 5.9890167786232707e-05, "loss": 0.0848, "step": 1083 }, { "epoch": 0.11, "learning_rate": 5.9889963096258874e-05, "loss": 0.6395, "step": 1084 }, { "epoch": 0.11, "learning_rate": 5.9889758216076684e-05, "loss": 1.3573, "step": 1085 }, { "epoch": 0.11, "learning_rate": 5.9889553145687446e-05, "loss": 0.7138, "step": 1086 }, { "epoch": 0.11, "learning_rate": 5.988934788509246e-05, "loss": 0.6657, "step": 1087 }, { "epoch": 0.11, "learning_rate": 5.9889142434293034e-05, "loss": 0.8916, "step": 1088 }, { "epoch": 0.11, "learning_rate": 5.9888936793290476e-05, "loss": 0.6092, "step": 1089 }, { "epoch": 0.11, "learning_rate": 5.98887309620861e-05, "loss": 1.4271, "step": 1090 }, { "epoch": 0.11, "learning_rate": 5.98885249406812e-05, "loss": 1.186, "step": 1091 }, { "epoch": 0.11, "learning_rate": 5.988831872907711e-05, "loss": 1.1247, "step": 1092 }, { "epoch": 0.11, "learning_rate": 5.988811232727512e-05, "loss": 0.9706, "step": 1093 }, { "epoch": 0.11, "learning_rate": 5.988790573527656e-05, "loss": 1.2722, "step": 1094 }, { "epoch": 0.11, "learning_rate": 5.9887698953082726e-05, "loss": 0.2973, "step": 1095 }, { "epoch": 0.11, "learning_rate": 5.988749198069496e-05, "loss": 0.9726, "step": 1096 }, { "epoch": 0.11, "learning_rate": 5.988728481811455e-05, "loss": 0.4296, "step": 1097 }, { "epoch": 0.11, "learning_rate": 5.988707746534284e-05, "loss": 0.5022, "step": 1098 }, { "epoch": 0.11, "learning_rate": 5.988686992238114e-05, "loss": 0.5961, "step": 1099 }, { "epoch": 0.11, "learning_rate": 5.988666218923077e-05, "loss": 0.6511, "step": 1100 }, { "epoch": 0.11, "learning_rate": 5.988645426589305e-05, "loss": 1.2256, "step": 1101 }, { "epoch": 0.11, "learning_rate": 5.9886246152369296e-05, "loss": 1.1772, "step": 1102 }, { "epoch": 0.11, "learning_rate": 5.9886037848660846e-05, "loss": 1.7857, "step": 1103 }, { "epoch": 0.11, "learning_rate": 5.9885829354769024e-05, "loss": 0.8413, "step": 1104 }, { "epoch": 0.11, "learning_rate": 5.9885620670695144e-05, "loss": 0.3054, "step": 1105 }, { "epoch": 0.11, "learning_rate": 5.9885411796440555e-05, "loss": 0.9743, "step": 1106 }, { "epoch": 0.11, "learning_rate": 5.9885202732006563e-05, "loss": 0.4139, "step": 1107 }, { "epoch": 0.11, "learning_rate": 5.988499347739452e-05, "loss": 0.4964, "step": 1108 }, { "epoch": 0.11, "learning_rate": 5.9884784032605736e-05, "loss": 0.7091, "step": 1109 }, { "epoch": 0.11, "learning_rate": 5.988457439764156e-05, "loss": 0.9931, "step": 1110 }, { "epoch": 0.11, "learning_rate": 5.988436457250333e-05, "loss": 0.8346, "step": 1111 }, { "epoch": 0.11, "learning_rate": 5.9884154557192367e-05, "loss": 0.5494, "step": 1112 }, { "epoch": 0.11, "learning_rate": 5.988394435171001e-05, "loss": 1.3061, "step": 1113 }, { "epoch": 0.11, "learning_rate": 5.98837339560576e-05, "loss": 0.5401, "step": 1114 }, { "epoch": 0.11, "learning_rate": 5.9883523370236475e-05, "loss": 1.0709, "step": 1115 }, { "epoch": 0.11, "learning_rate": 5.9883312594247974e-05, "loss": 0.6027, "step": 1116 }, { "epoch": 0.11, "learning_rate": 5.988310162809344e-05, "loss": 1.5028, "step": 1117 }, { "epoch": 0.11, "learning_rate": 5.9882890471774226e-05, "loss": 0.3135, "step": 1118 }, { "epoch": 0.11, "learning_rate": 5.9882679125291654e-05, "loss": 0.7944, "step": 1119 }, { "epoch": 0.11, "learning_rate": 5.988246758864708e-05, "loss": 0.6967, "step": 1120 }, { "epoch": 0.11, "learning_rate": 5.988225586184185e-05, "loss": 1.0496, "step": 1121 }, { "epoch": 0.11, "learning_rate": 5.9882043944877326e-05, "loss": 0.4604, "step": 1122 }, { "epoch": 0.11, "learning_rate": 5.9881831837754824e-05, "loss": 0.6298, "step": 1123 }, { "epoch": 0.11, "learning_rate": 5.9881619540475716e-05, "loss": 1.084, "step": 1124 }, { "epoch": 0.11, "learning_rate": 5.988140705304136e-05, "loss": 0.7396, "step": 1125 }, { "epoch": 0.11, "learning_rate": 5.988119437545309e-05, "loss": 1.3351, "step": 1126 }, { "epoch": 0.11, "learning_rate": 5.988098150771225e-05, "loss": 0.6039, "step": 1127 }, { "epoch": 0.11, "learning_rate": 5.988076844982023e-05, "loss": 0.2427, "step": 1128 }, { "epoch": 0.11, "learning_rate": 5.988055520177837e-05, "loss": 0.4775, "step": 1129 }, { "epoch": 0.11, "learning_rate": 5.9880341763588014e-05, "loss": 1.3579, "step": 1130 }, { "epoch": 0.11, "learning_rate": 5.9880128135250534e-05, "loss": 1.0938, "step": 1131 }, { "epoch": 0.11, "learning_rate": 5.987991431676728e-05, "loss": 0.6036, "step": 1132 }, { "epoch": 0.12, "learning_rate": 5.987970030813962e-05, "loss": 0.8429, "step": 1133 }, { "epoch": 0.12, "learning_rate": 5.987948610936892e-05, "loss": 0.6885, "step": 1134 }, { "epoch": 0.12, "learning_rate": 5.987927172045653e-05, "loss": 0.5668, "step": 1135 }, { "epoch": 0.12, "learning_rate": 5.987905714140383e-05, "loss": 1.3051, "step": 1136 }, { "epoch": 0.12, "learning_rate": 5.987884237221217e-05, "loss": 0.3961, "step": 1137 }, { "epoch": 0.12, "learning_rate": 5.9878627412882924e-05, "loss": 0.6899, "step": 1138 }, { "epoch": 0.12, "learning_rate": 5.987841226341746e-05, "loss": 1.5431, "step": 1139 }, { "epoch": 0.12, "learning_rate": 5.987819692381715e-05, "loss": 0.9564, "step": 1140 }, { "epoch": 0.12, "learning_rate": 5.987798139408335e-05, "loss": 0.3516, "step": 1141 }, { "epoch": 0.12, "learning_rate": 5.987776567421746e-05, "loss": 1.0985, "step": 1142 }, { "epoch": 0.12, "learning_rate": 5.9877549764220824e-05, "loss": 0.6131, "step": 1143 }, { "epoch": 0.12, "learning_rate": 5.987733366409483e-05, "loss": 0.5717, "step": 1144 }, { "epoch": 0.12, "learning_rate": 5.987711737384085e-05, "loss": 0.5964, "step": 1145 }, { "epoch": 0.12, "learning_rate": 5.987690089346026e-05, "loss": 1.8654, "step": 1146 }, { "epoch": 0.12, "learning_rate": 5.9876684222954445e-05, "loss": 0.8197, "step": 1147 }, { "epoch": 0.12, "learning_rate": 5.9876467362324765e-05, "loss": 1.0402, "step": 1148 }, { "epoch": 0.12, "learning_rate": 5.987625031157262e-05, "loss": 1.0689, "step": 1149 }, { "epoch": 0.12, "learning_rate": 5.987603307069938e-05, "loss": 0.9582, "step": 1150 }, { "epoch": 0.12, "learning_rate": 5.9875815639706436e-05, "loss": 1.1618, "step": 1151 }, { "epoch": 0.12, "learning_rate": 5.9875598018595166e-05, "loss": 0.4754, "step": 1152 }, { "epoch": 0.12, "learning_rate": 5.987538020736695e-05, "loss": 0.6072, "step": 1153 }, { "epoch": 0.12, "learning_rate": 5.987516220602318e-05, "loss": 0.8993, "step": 1154 }, { "epoch": 0.12, "learning_rate": 5.987494401456525e-05, "loss": 1.3039, "step": 1155 }, { "epoch": 0.12, "learning_rate": 5.987472563299454e-05, "loss": 0.3788, "step": 1156 }, { "epoch": 0.12, "learning_rate": 5.9874507061312436e-05, "loss": 0.3224, "step": 1157 }, { "epoch": 0.12, "learning_rate": 5.987428829952034e-05, "loss": 1.7597, "step": 1158 }, { "epoch": 0.12, "learning_rate": 5.9874069347619637e-05, "loss": 0.6044, "step": 1159 }, { "epoch": 0.12, "learning_rate": 5.987385020561172e-05, "loss": 1.4585, "step": 1160 }, { "epoch": 0.12, "learning_rate": 5.987363087349798e-05, "loss": 0.7983, "step": 1161 }, { "epoch": 0.12, "learning_rate": 5.987341135127982e-05, "loss": 0.3511, "step": 1162 }, { "epoch": 0.12, "learning_rate": 5.987319163895863e-05, "loss": 1.0326, "step": 1163 }, { "epoch": 0.12, "learning_rate": 5.987297173653583e-05, "loss": 1.1401, "step": 1164 }, { "epoch": 0.12, "learning_rate": 5.9872751644012777e-05, "loss": 0.7799, "step": 1165 }, { "epoch": 0.12, "learning_rate": 5.9872531361390915e-05, "loss": 0.8379, "step": 1166 }, { "epoch": 0.12, "learning_rate": 5.987231088867162e-05, "loss": 1.3704, "step": 1167 }, { "epoch": 0.12, "learning_rate": 5.9872090225856306e-05, "loss": 1.2137, "step": 1168 }, { "epoch": 0.12, "learning_rate": 5.987186937294638e-05, "loss": 0.7455, "step": 1169 }, { "epoch": 0.12, "learning_rate": 5.9871648329943235e-05, "loss": 0.3902, "step": 1170 }, { "epoch": 0.12, "learning_rate": 5.987142709684828e-05, "loss": 1.2054, "step": 1171 }, { "epoch": 0.12, "learning_rate": 5.987120567366293e-05, "loss": 0.3369, "step": 1172 }, { "epoch": 0.12, "learning_rate": 5.987098406038859e-05, "loss": 0.7286, "step": 1173 }, { "epoch": 0.12, "learning_rate": 5.987076225702667e-05, "loss": 0.5244, "step": 1174 }, { "epoch": 0.12, "learning_rate": 5.9870540263578586e-05, "loss": 0.6615, "step": 1175 }, { "epoch": 0.12, "learning_rate": 5.987031808004574e-05, "loss": 0.4507, "step": 1176 }, { "epoch": 0.12, "learning_rate": 5.987009570642956e-05, "loss": 0.8982, "step": 1177 }, { "epoch": 0.12, "learning_rate": 5.9869873142731455e-05, "loss": 0.8075, "step": 1178 }, { "epoch": 0.12, "learning_rate": 5.986965038895284e-05, "loss": 1.0292, "step": 1179 }, { "epoch": 0.12, "learning_rate": 5.9869427445095126e-05, "loss": 1.0694, "step": 1180 }, { "epoch": 0.12, "learning_rate": 5.9869204311159746e-05, "loss": 1.2786, "step": 1181 }, { "epoch": 0.12, "learning_rate": 5.9868980987148105e-05, "loss": 1.1012, "step": 1182 }, { "epoch": 0.12, "learning_rate": 5.9868757473061634e-05, "loss": 0.8826, "step": 1183 }, { "epoch": 0.12, "learning_rate": 5.986853376890176e-05, "loss": 0.5006, "step": 1184 }, { "epoch": 0.12, "learning_rate": 5.986830987466989e-05, "loss": 0.7524, "step": 1185 }, { "epoch": 0.12, "learning_rate": 5.986808579036747e-05, "loss": 1.1605, "step": 1186 }, { "epoch": 0.12, "learning_rate": 5.986786151599591e-05, "loss": 1.1111, "step": 1187 }, { "epoch": 0.12, "learning_rate": 5.9867637051556636e-05, "loss": 0.333, "step": 1188 }, { "epoch": 0.12, "learning_rate": 5.986741239705108e-05, "loss": 0.8858, "step": 1189 }, { "epoch": 0.12, "learning_rate": 5.986718755248068e-05, "loss": 0.5438, "step": 1190 }, { "epoch": 0.12, "learning_rate": 5.9866962517846864e-05, "loss": 0.3823, "step": 1191 }, { "epoch": 0.12, "learning_rate": 5.986673729315106e-05, "loss": 1.0548, "step": 1192 }, { "epoch": 0.12, "learning_rate": 5.9866511878394685e-05, "loss": 1.6192, "step": 1193 }, { "epoch": 0.12, "learning_rate": 5.98662862735792e-05, "loss": 1.1931, "step": 1194 }, { "epoch": 0.12, "learning_rate": 5.986606047870605e-05, "loss": 0.353, "step": 1195 }, { "epoch": 0.12, "learning_rate": 5.986583449377663e-05, "loss": 0.9614, "step": 1196 }, { "epoch": 0.12, "learning_rate": 5.986560831879241e-05, "loss": 0.3753, "step": 1197 }, { "epoch": 0.12, "learning_rate": 5.9865381953754815e-05, "loss": 1.5116, "step": 1198 }, { "epoch": 0.12, "learning_rate": 5.9865155398665294e-05, "loss": 0.3428, "step": 1199 }, { "epoch": 0.12, "learning_rate": 5.986492865352528e-05, "loss": 0.7692, "step": 1200 }, { "epoch": 0.12, "learning_rate": 5.986470171833623e-05, "loss": 0.2597, "step": 1201 }, { "epoch": 0.12, "learning_rate": 5.9864474593099576e-05, "loss": 0.9104, "step": 1202 }, { "epoch": 0.12, "learning_rate": 5.986424727781677e-05, "loss": 0.8567, "step": 1203 }, { "epoch": 0.12, "learning_rate": 5.986401977248925e-05, "loss": 0.9326, "step": 1204 }, { "epoch": 0.12, "learning_rate": 5.9863792077118466e-05, "loss": 1.2911, "step": 1205 }, { "epoch": 0.12, "learning_rate": 5.986356419170587e-05, "loss": 0.9643, "step": 1206 }, { "epoch": 0.12, "learning_rate": 5.986333611625292e-05, "loss": 0.4945, "step": 1207 }, { "epoch": 0.12, "learning_rate": 5.986310785076105e-05, "loss": 0.7052, "step": 1208 }, { "epoch": 0.12, "learning_rate": 5.986287939523173e-05, "loss": 1.0412, "step": 1209 }, { "epoch": 0.12, "learning_rate": 5.986265074966641e-05, "loss": 0.982, "step": 1210 }, { "epoch": 0.12, "learning_rate": 5.9862421914066534e-05, "loss": 0.5036, "step": 1211 }, { "epoch": 0.12, "learning_rate": 5.986219288843356e-05, "loss": 1.0965, "step": 1212 }, { "epoch": 0.12, "learning_rate": 5.9861963672768965e-05, "loss": 0.4125, "step": 1213 }, { "epoch": 0.12, "learning_rate": 5.986173426707418e-05, "loss": 0.7608, "step": 1214 }, { "epoch": 0.12, "learning_rate": 5.9861504671350684e-05, "loss": 0.7943, "step": 1215 }, { "epoch": 0.12, "learning_rate": 5.9861274885599934e-05, "loss": 0.6643, "step": 1216 }, { "epoch": 0.12, "learning_rate": 5.986104490982339e-05, "loss": 0.6512, "step": 1217 }, { "epoch": 0.12, "learning_rate": 5.9860814744022505e-05, "loss": 1.1138, "step": 1218 }, { "epoch": 0.12, "learning_rate": 5.986058438819877e-05, "loss": 0.8373, "step": 1219 }, { "epoch": 0.12, "learning_rate": 5.986035384235363e-05, "loss": 1.023, "step": 1220 }, { "epoch": 0.12, "learning_rate": 5.986012310648855e-05, "loss": 0.9845, "step": 1221 }, { "epoch": 0.12, "learning_rate": 5.985989218060502e-05, "loss": 0.908, "step": 1222 }, { "epoch": 0.12, "learning_rate": 5.985966106470448e-05, "loss": 0.4896, "step": 1223 }, { "epoch": 0.12, "learning_rate": 5.985942975878843e-05, "loss": 1.1295, "step": 1224 }, { "epoch": 0.12, "learning_rate": 5.985919826285832e-05, "loss": 0.5718, "step": 1225 }, { "epoch": 0.12, "learning_rate": 5.985896657691563e-05, "loss": 1.528, "step": 1226 }, { "epoch": 0.12, "learning_rate": 5.985873470096185e-05, "loss": 1.1541, "step": 1227 }, { "epoch": 0.12, "learning_rate": 5.985850263499842e-05, "loss": 0.4132, "step": 1228 }, { "epoch": 0.12, "learning_rate": 5.9858270379026855e-05, "loss": 0.8125, "step": 1229 }, { "epoch": 0.12, "learning_rate": 5.98580379330486e-05, "loss": 1.1453, "step": 1230 }, { "epoch": 0.12, "learning_rate": 5.985780529706516e-05, "loss": 0.6032, "step": 1231 }, { "epoch": 0.13, "learning_rate": 5.985757247107801e-05, "loss": 0.4604, "step": 1232 }, { "epoch": 0.13, "learning_rate": 5.9857339455088624e-05, "loss": 1.1665, "step": 1233 }, { "epoch": 0.13, "learning_rate": 5.985710624909848e-05, "loss": 1.0531, "step": 1234 }, { "epoch": 0.13, "learning_rate": 5.985687285310908e-05, "loss": 0.9177, "step": 1235 }, { "epoch": 0.13, "learning_rate": 5.9856639267121895e-05, "loss": 0.4208, "step": 1236 }, { "epoch": 0.13, "learning_rate": 5.985640549113842e-05, "loss": 0.2211, "step": 1237 }, { "epoch": 0.13, "learning_rate": 5.9856171525160135e-05, "loss": 0.302, "step": 1238 }, { "epoch": 0.13, "learning_rate": 5.9855937369188524e-05, "loss": 0.6353, "step": 1239 }, { "epoch": 0.13, "learning_rate": 5.98557030232251e-05, "loss": 0.3606, "step": 1240 }, { "epoch": 0.13, "learning_rate": 5.9855468487271324e-05, "loss": 0.6197, "step": 1241 }, { "epoch": 0.13, "learning_rate": 5.9855233761328716e-05, "loss": 0.7696, "step": 1242 }, { "epoch": 0.13, "learning_rate": 5.985499884539875e-05, "loss": 0.3138, "step": 1243 }, { "epoch": 0.13, "learning_rate": 5.985476373948294e-05, "loss": 0.3279, "step": 1244 }, { "epoch": 0.13, "learning_rate": 5.985452844358276e-05, "loss": 0.7071, "step": 1245 }, { "epoch": 0.13, "learning_rate": 5.985429295769973e-05, "loss": 1.2222, "step": 1246 }, { "epoch": 0.13, "learning_rate": 5.985405728183533e-05, "loss": 1.096, "step": 1247 }, { "epoch": 0.13, "learning_rate": 5.9853821415991064e-05, "loss": 0.2738, "step": 1248 }, { "epoch": 0.13, "learning_rate": 5.9853585360168436e-05, "loss": 0.7136, "step": 1249 }, { "epoch": 0.13, "learning_rate": 5.985334911436894e-05, "loss": 0.5939, "step": 1250 }, { "epoch": 0.13, "learning_rate": 5.985311267859411e-05, "loss": 0.6351, "step": 1251 }, { "epoch": 0.13, "learning_rate": 5.98528760528454e-05, "loss": 1.6408, "step": 1252 }, { "epoch": 0.13, "learning_rate": 5.985263923712437e-05, "loss": 0.846, "step": 1253 }, { "epoch": 0.13, "learning_rate": 5.985240223143248e-05, "loss": 0.559, "step": 1254 }, { "epoch": 0.13, "learning_rate": 5.9852165035771265e-05, "loss": 1.3103, "step": 1255 }, { "epoch": 0.13, "learning_rate": 5.9851927650142235e-05, "loss": 1.171, "step": 1256 }, { "epoch": 0.13, "learning_rate": 5.985169007454689e-05, "loss": 1.3802, "step": 1257 }, { "epoch": 0.13, "learning_rate": 5.985145230898674e-05, "loss": 1.1768, "step": 1258 }, { "epoch": 0.13, "learning_rate": 5.985121435346331e-05, "loss": 0.2226, "step": 1259 }, { "epoch": 0.13, "learning_rate": 5.98509762079781e-05, "loss": 1.1928, "step": 1260 }, { "epoch": 0.13, "learning_rate": 5.985073787253264e-05, "loss": 0.7502, "step": 1261 }, { "epoch": 0.13, "learning_rate": 5.9850499347128437e-05, "loss": 0.3657, "step": 1262 }, { "epoch": 0.13, "learning_rate": 5.9850260631767014e-05, "loss": 0.6444, "step": 1263 }, { "epoch": 0.13, "learning_rate": 5.985002172644989e-05, "loss": 0.8802, "step": 1264 }, { "epoch": 0.13, "learning_rate": 5.984978263117858e-05, "loss": 0.568, "step": 1265 }, { "epoch": 0.13, "learning_rate": 5.9849543345954616e-05, "loss": 0.4587, "step": 1266 }, { "epoch": 0.13, "learning_rate": 5.98493038707795e-05, "loss": 1.204, "step": 1267 }, { "epoch": 0.13, "learning_rate": 5.984906420565478e-05, "loss": 0.2801, "step": 1268 }, { "epoch": 0.13, "learning_rate": 5.9848824350581964e-05, "loss": 0.6, "step": 1269 }, { "epoch": 0.13, "learning_rate": 5.9848584305562587e-05, "loss": 0.7783, "step": 1270 }, { "epoch": 0.13, "learning_rate": 5.984834407059818e-05, "loss": 0.5533, "step": 1271 }, { "epoch": 0.13, "learning_rate": 5.9848103645690257e-05, "loss": 0.6091, "step": 1272 }, { "epoch": 0.13, "learning_rate": 5.984786303084037e-05, "loss": 0.407, "step": 1273 }, { "epoch": 0.13, "learning_rate": 5.984762222605001e-05, "loss": 1.3683, "step": 1274 }, { "epoch": 0.13, "learning_rate": 5.984738123132076e-05, "loss": 0.3727, "step": 1275 }, { "epoch": 0.13, "learning_rate": 5.984714004665412e-05, "loss": 1.2181, "step": 1276 }, { "epoch": 0.13, "learning_rate": 5.984689867205165e-05, "loss": 0.4414, "step": 1277 }, { "epoch": 0.13, "learning_rate": 5.984665710751486e-05, "loss": 1.3013, "step": 1278 }, { "epoch": 0.13, "learning_rate": 5.98464153530453e-05, "loss": 0.2121, "step": 1279 }, { "epoch": 0.13, "learning_rate": 5.9846173408644516e-05, "loss": 0.7102, "step": 1280 }, { "epoch": 0.13, "learning_rate": 5.9845931274314026e-05, "loss": 0.6573, "step": 1281 }, { "epoch": 0.13, "learning_rate": 5.984568895005539e-05, "loss": 0.5891, "step": 1282 }, { "epoch": 0.13, "learning_rate": 5.984544643587014e-05, "loss": 0.7248, "step": 1283 }, { "epoch": 0.13, "learning_rate": 5.984520373175982e-05, "loss": 1.017, "step": 1284 }, { "epoch": 0.13, "learning_rate": 5.9844960837725984e-05, "loss": 1.2089, "step": 1285 }, { "epoch": 0.13, "learning_rate": 5.984471775377016e-05, "loss": 1.1071, "step": 1286 }, { "epoch": 0.13, "learning_rate": 5.984447447989391e-05, "loss": 1.3618, "step": 1287 }, { "epoch": 0.13, "learning_rate": 5.9844231016098786e-05, "loss": 1.0378, "step": 1288 }, { "epoch": 0.13, "learning_rate": 5.984398736238633e-05, "loss": 0.9968, "step": 1289 }, { "epoch": 0.13, "learning_rate": 5.984374351875808e-05, "loss": 0.6314, "step": 1290 }, { "epoch": 0.13, "learning_rate": 5.984349948521561e-05, "loss": 0.7591, "step": 1291 }, { "epoch": 0.13, "learning_rate": 5.9843255261760456e-05, "loss": 0.737, "step": 1292 }, { "epoch": 0.13, "learning_rate": 5.984301084839418e-05, "loss": 0.2742, "step": 1293 }, { "epoch": 0.13, "learning_rate": 5.9842766245118334e-05, "loss": 0.7152, "step": 1294 }, { "epoch": 0.13, "learning_rate": 5.9842521451934476e-05, "loss": 0.4293, "step": 1295 }, { "epoch": 0.13, "learning_rate": 5.984227646884417e-05, "loss": 0.6558, "step": 1296 }, { "epoch": 0.13, "learning_rate": 5.984203129584896e-05, "loss": 0.6589, "step": 1297 }, { "epoch": 0.13, "learning_rate": 5.9841785932950426e-05, "loss": 0.9688, "step": 1298 }, { "epoch": 0.13, "learning_rate": 5.984154038015011e-05, "loss": 0.2537, "step": 1299 }, { "epoch": 0.13, "learning_rate": 5.984129463744959e-05, "loss": 1.0172, "step": 1300 }, { "epoch": 0.13, "learning_rate": 5.984104870485042e-05, "loss": 0.9822, "step": 1301 }, { "epoch": 0.13, "learning_rate": 5.984080258235417e-05, "loss": 1.304, "step": 1302 }, { "epoch": 0.13, "learning_rate": 5.984055626996241e-05, "loss": 1.0327, "step": 1303 }, { "epoch": 0.13, "learning_rate": 5.984030976767669e-05, "loss": 0.7037, "step": 1304 }, { "epoch": 0.13, "learning_rate": 5.984006307549859e-05, "loss": 0.416, "step": 1305 }, { "epoch": 0.13, "learning_rate": 5.9839816193429684e-05, "loss": 1.271, "step": 1306 }, { "epoch": 0.13, "learning_rate": 5.983956912147154e-05, "loss": 0.5846, "step": 1307 }, { "epoch": 0.13, "learning_rate": 5.983932185962574e-05, "loss": 0.4476, "step": 1308 }, { "epoch": 0.13, "learning_rate": 5.9839074407893835e-05, "loss": 0.6944, "step": 1309 }, { "epoch": 0.13, "learning_rate": 5.983882676627741e-05, "loss": 0.4762, "step": 1310 }, { "epoch": 0.13, "learning_rate": 5.9838578934778046e-05, "loss": 1.2798, "step": 1311 }, { "epoch": 0.13, "learning_rate": 5.9838330913397316e-05, "loss": 1.4649, "step": 1312 }, { "epoch": 0.13, "learning_rate": 5.98380827021368e-05, "loss": 0.9865, "step": 1313 }, { "epoch": 0.13, "learning_rate": 5.983783430099808e-05, "loss": 0.2177, "step": 1314 }, { "epoch": 0.13, "learning_rate": 5.983758570998273e-05, "loss": 0.5438, "step": 1315 }, { "epoch": 0.13, "learning_rate": 5.9837336929092336e-05, "loss": 0.2122, "step": 1316 }, { "epoch": 0.13, "learning_rate": 5.983708795832849e-05, "loss": 0.6341, "step": 1317 }, { "epoch": 0.13, "learning_rate": 5.983683879769275e-05, "loss": 0.7423, "step": 1318 }, { "epoch": 0.13, "learning_rate": 5.983658944718673e-05, "loss": 1.2097, "step": 1319 }, { "epoch": 0.13, "learning_rate": 5.9836339906811994e-05, "loss": 0.6911, "step": 1320 }, { "epoch": 0.13, "learning_rate": 5.983609017657015e-05, "loss": 0.4651, "step": 1321 }, { "epoch": 0.13, "learning_rate": 5.983584025646278e-05, "loss": 0.3754, "step": 1322 }, { "epoch": 0.13, "learning_rate": 5.983559014649147e-05, "loss": 0.2986, "step": 1323 }, { "epoch": 0.13, "learning_rate": 5.983533984665781e-05, "loss": 0.7459, "step": 1324 }, { "epoch": 0.13, "learning_rate": 5.98350893569634e-05, "loss": 0.9162, "step": 1325 }, { "epoch": 0.13, "learning_rate": 5.983483867740983e-05, "loss": 1.1588, "step": 1326 }, { "epoch": 0.13, "learning_rate": 5.983458780799871e-05, "loss": 0.2967, "step": 1327 }, { "epoch": 0.13, "learning_rate": 5.9834336748731605e-05, "loss": 1.1421, "step": 1328 }, { "epoch": 0.13, "learning_rate": 5.9834085499610135e-05, "loss": 0.419, "step": 1329 }, { "epoch": 0.14, "learning_rate": 5.98338340606359e-05, "loss": 1.0198, "step": 1330 }, { "epoch": 0.14, "learning_rate": 5.9833582431810485e-05, "loss": 0.9191, "step": 1331 }, { "epoch": 0.14, "learning_rate": 5.983333061313551e-05, "loss": 0.3507, "step": 1332 }, { "epoch": 0.14, "learning_rate": 5.9833078604612566e-05, "loss": 0.5406, "step": 1333 }, { "epoch": 0.14, "learning_rate": 5.9832826406243255e-05, "loss": 0.4176, "step": 1334 }, { "epoch": 0.14, "learning_rate": 5.983257401802919e-05, "loss": 0.9748, "step": 1335 }, { "epoch": 0.14, "learning_rate": 5.983232143997197e-05, "loss": 1.2234, "step": 1336 }, { "epoch": 0.14, "learning_rate": 5.983206867207321e-05, "loss": 0.98, "step": 1337 }, { "epoch": 0.14, "learning_rate": 5.983181571433451e-05, "loss": 0.6652, "step": 1338 }, { "epoch": 0.14, "learning_rate": 5.983156256675748e-05, "loss": 1.0509, "step": 1339 }, { "epoch": 0.14, "learning_rate": 5.983130922934374e-05, "loss": 1.2525, "step": 1340 }, { "epoch": 0.14, "learning_rate": 5.983105570209489e-05, "loss": 1.2173, "step": 1341 }, { "epoch": 0.14, "learning_rate": 5.9830801985012554e-05, "loss": 0.6914, "step": 1342 }, { "epoch": 0.14, "learning_rate": 5.983054807809834e-05, "loss": 1.4776, "step": 1343 }, { "epoch": 0.14, "learning_rate": 5.983029398135387e-05, "loss": 0.8809, "step": 1344 }, { "epoch": 0.14, "learning_rate": 5.9830039694780755e-05, "loss": 0.21, "step": 1345 }, { "epoch": 0.14, "learning_rate": 5.982978521838062e-05, "loss": 1.771, "step": 1346 }, { "epoch": 0.14, "learning_rate": 5.982953055215508e-05, "loss": 1.1672, "step": 1347 }, { "epoch": 0.14, "learning_rate": 5.982927569610575e-05, "loss": 1.3271, "step": 1348 }, { "epoch": 0.14, "learning_rate": 5.982902065023425e-05, "loss": 1.0076, "step": 1349 }, { "epoch": 0.14, "learning_rate": 5.9828765414542224e-05, "loss": 1.2501, "step": 1350 }, { "epoch": 0.14, "learning_rate": 5.9828509989031275e-05, "loss": 0.4928, "step": 1351 }, { "epoch": 0.14, "learning_rate": 5.982825437370303e-05, "loss": 0.4342, "step": 1352 }, { "epoch": 0.14, "learning_rate": 5.9827998568559134e-05, "loss": 1.0488, "step": 1353 }, { "epoch": 0.14, "learning_rate": 5.9827742573601196e-05, "loss": 0.6044, "step": 1354 }, { "epoch": 0.14, "learning_rate": 5.982748638883085e-05, "loss": 0.8611, "step": 1355 }, { "epoch": 0.14, "learning_rate": 5.982723001424973e-05, "loss": 0.5924, "step": 1356 }, { "epoch": 0.14, "learning_rate": 5.9826973449859464e-05, "loss": 1.0922, "step": 1357 }, { "epoch": 0.14, "learning_rate": 5.982671669566169e-05, "loss": 1.445, "step": 1358 }, { "epoch": 0.14, "learning_rate": 5.982645975165803e-05, "loss": 0.4924, "step": 1359 }, { "epoch": 0.14, "learning_rate": 5.982620261785013e-05, "loss": 1.2243, "step": 1360 }, { "epoch": 0.14, "learning_rate": 5.9825945294239624e-05, "loss": 0.294, "step": 1361 }, { "epoch": 0.14, "learning_rate": 5.982568778082815e-05, "loss": 0.502, "step": 1362 }, { "epoch": 0.14, "learning_rate": 5.982543007761735e-05, "loss": 1.02, "step": 1363 }, { "epoch": 0.14, "learning_rate": 5.9825172184608847e-05, "loss": 0.9695, "step": 1364 }, { "epoch": 0.14, "learning_rate": 5.9824914101804296e-05, "loss": 1.2078, "step": 1365 }, { "epoch": 0.14, "learning_rate": 5.982465582920535e-05, "loss": 1.2938, "step": 1366 }, { "epoch": 0.14, "learning_rate": 5.982439736681363e-05, "loss": 0.2824, "step": 1367 }, { "epoch": 0.14, "learning_rate": 5.982413871463079e-05, "loss": 1.1618, "step": 1368 }, { "epoch": 0.14, "learning_rate": 5.982387987265848e-05, "loss": 1.1872, "step": 1369 }, { "epoch": 0.14, "learning_rate": 5.9823620840898344e-05, "loss": 0.6632, "step": 1370 }, { "epoch": 0.14, "learning_rate": 5.982336161935203e-05, "loss": 0.1458, "step": 1371 }, { "epoch": 0.14, "learning_rate": 5.982310220802119e-05, "loss": 0.4226, "step": 1372 }, { "epoch": 0.14, "learning_rate": 5.982284260690747e-05, "loss": 0.6113, "step": 1373 }, { "epoch": 0.14, "learning_rate": 5.982258281601252e-05, "loss": 1.1409, "step": 1374 }, { "epoch": 0.14, "learning_rate": 5.9822322835338e-05, "loss": 0.9909, "step": 1375 }, { "epoch": 0.14, "learning_rate": 5.982206266488557e-05, "loss": 1.2272, "step": 1376 }, { "epoch": 0.14, "learning_rate": 5.9821802304656876e-05, "loss": 1.1472, "step": 1377 }, { "epoch": 0.14, "learning_rate": 5.982154175465358e-05, "loss": 0.65, "step": 1378 }, { "epoch": 0.14, "learning_rate": 5.982128101487733e-05, "loss": 1.066, "step": 1379 }, { "epoch": 0.14, "learning_rate": 5.982102008532979e-05, "loss": 0.9501, "step": 1380 }, { "epoch": 0.14, "learning_rate": 5.982075896601263e-05, "loss": 1.5551, "step": 1381 }, { "epoch": 0.14, "learning_rate": 5.98204976569275e-05, "loss": 0.9529, "step": 1382 }, { "epoch": 0.14, "learning_rate": 5.982023615807608e-05, "loss": 1.534, "step": 1383 }, { "epoch": 0.14, "learning_rate": 5.9819974469460004e-05, "loss": 1.1275, "step": 1384 }, { "epoch": 0.14, "learning_rate": 5.9819712591080955e-05, "loss": 0.5649, "step": 1385 }, { "epoch": 0.14, "learning_rate": 5.981945052294061e-05, "loss": 0.9163, "step": 1386 }, { "epoch": 0.14, "learning_rate": 5.9819188265040626e-05, "loss": 0.3621, "step": 1387 }, { "epoch": 0.14, "learning_rate": 5.981892581738266e-05, "loss": 0.5598, "step": 1388 }, { "epoch": 0.14, "learning_rate": 5.98186631799684e-05, "loss": 0.6739, "step": 1389 }, { "epoch": 0.14, "learning_rate": 5.9818400352799505e-05, "loss": 1.2023, "step": 1390 }, { "epoch": 0.14, "learning_rate": 5.9818137335877665e-05, "loss": 0.6143, "step": 1391 }, { "epoch": 0.14, "learning_rate": 5.981787412920454e-05, "loss": 1.1682, "step": 1392 }, { "epoch": 0.14, "learning_rate": 5.9817610732781805e-05, "loss": 1.1159, "step": 1393 }, { "epoch": 0.14, "learning_rate": 5.981734714661114e-05, "loss": 1.2014, "step": 1394 }, { "epoch": 0.14, "learning_rate": 5.981708337069422e-05, "loss": 0.625, "step": 1395 }, { "epoch": 0.14, "learning_rate": 5.9816819405032715e-05, "loss": 0.5495, "step": 1396 }, { "epoch": 0.14, "learning_rate": 5.9816555249628324e-05, "loss": 0.4957, "step": 1397 }, { "epoch": 0.14, "learning_rate": 5.981629090448272e-05, "loss": 1.1734, "step": 1398 }, { "epoch": 0.14, "learning_rate": 5.981602636959758e-05, "loss": 1.5285, "step": 1399 }, { "epoch": 0.14, "learning_rate": 5.9815761644974584e-05, "loss": 0.9433, "step": 1400 }, { "epoch": 0.14, "learning_rate": 5.981549673061544e-05, "loss": 0.6394, "step": 1401 }, { "epoch": 0.14, "learning_rate": 5.9815231626521805e-05, "loss": 0.4987, "step": 1402 }, { "epoch": 0.14, "learning_rate": 5.9814966332695374e-05, "loss": 0.9429, "step": 1403 }, { "epoch": 0.14, "learning_rate": 5.981470084913785e-05, "loss": 0.1641, "step": 1404 }, { "epoch": 0.14, "learning_rate": 5.981443517585091e-05, "loss": 0.5173, "step": 1405 }, { "epoch": 0.14, "learning_rate": 5.981416931283624e-05, "loss": 0.737, "step": 1406 }, { "epoch": 0.14, "learning_rate": 5.981390326009555e-05, "loss": 0.8649, "step": 1407 }, { "epoch": 0.14, "learning_rate": 5.981363701763051e-05, "loss": 1.3453, "step": 1408 }, { "epoch": 0.14, "learning_rate": 5.981337058544283e-05, "loss": 0.9963, "step": 1409 }, { "epoch": 0.14, "learning_rate": 5.98131039635342e-05, "loss": 1.185, "step": 1410 }, { "epoch": 0.14, "learning_rate": 5.9812837151906325e-05, "loss": 0.341, "step": 1411 }, { "epoch": 0.14, "learning_rate": 5.981257015056089e-05, "loss": 1.0465, "step": 1412 }, { "epoch": 0.14, "learning_rate": 5.98123029594996e-05, "loss": 0.3065, "step": 1413 }, { "epoch": 0.14, "learning_rate": 5.9812035578724146e-05, "loss": 0.3551, "step": 1414 }, { "epoch": 0.14, "learning_rate": 5.981176800823625e-05, "loss": 0.9341, "step": 1415 }, { "epoch": 0.14, "learning_rate": 5.98115002480376e-05, "loss": 1.7064, "step": 1416 }, { "epoch": 0.14, "learning_rate": 5.981123229812991e-05, "loss": 1.5043, "step": 1417 }, { "epoch": 0.14, "learning_rate": 5.9810964158514866e-05, "loss": 0.9056, "step": 1418 }, { "epoch": 0.14, "learning_rate": 5.9810695829194196e-05, "loss": 0.5721, "step": 1419 }, { "epoch": 0.14, "learning_rate": 5.98104273101696e-05, "loss": 0.7939, "step": 1420 }, { "epoch": 0.14, "learning_rate": 5.9810158601442775e-05, "loss": 0.7, "step": 1421 }, { "epoch": 0.14, "learning_rate": 5.9809889703015453e-05, "loss": 0.9804, "step": 1422 }, { "epoch": 0.14, "learning_rate": 5.980962061488932e-05, "loss": 0.423, "step": 1423 }, { "epoch": 0.14, "learning_rate": 5.980935133706611e-05, "loss": 1.2736, "step": 1424 }, { "epoch": 0.14, "learning_rate": 5.980908186954753e-05, "loss": 1.3671, "step": 1425 }, { "epoch": 0.14, "learning_rate": 5.980881221233529e-05, "loss": 1.6594, "step": 1426 }, { "epoch": 0.14, "learning_rate": 5.980854236543111e-05, "loss": 1.369, "step": 1427 }, { "epoch": 0.15, "learning_rate": 5.98082723288367e-05, "loss": 1.0586, "step": 1428 }, { "epoch": 0.15, "learning_rate": 5.9808002102553795e-05, "loss": 0.978, "step": 1429 }, { "epoch": 0.15, "learning_rate": 5.9807731686584095e-05, "loss": 0.8498, "step": 1430 }, { "epoch": 0.15, "learning_rate": 5.9807461080929336e-05, "loss": 0.7775, "step": 1431 }, { "epoch": 0.15, "learning_rate": 5.9807190285591224e-05, "loss": 1.278, "step": 1432 }, { "epoch": 0.15, "learning_rate": 5.98069193005715e-05, "loss": 0.9634, "step": 1433 }, { "epoch": 0.15, "learning_rate": 5.980664812587189e-05, "loss": 0.3161, "step": 1434 }, { "epoch": 0.15, "learning_rate": 5.980637676149409e-05, "loss": 1.0848, "step": 1435 }, { "epoch": 0.15, "learning_rate": 5.980610520743986e-05, "loss": 0.314, "step": 1436 }, { "epoch": 0.15, "learning_rate": 5.980583346371092e-05, "loss": 0.43, "step": 1437 }, { "epoch": 0.15, "learning_rate": 5.980556153030899e-05, "loss": 1.035, "step": 1438 }, { "epoch": 0.15, "learning_rate": 5.9805289407235794e-05, "loss": 0.448, "step": 1439 }, { "epoch": 0.15, "learning_rate": 5.9805017094493085e-05, "loss": 0.5084, "step": 1440 }, { "epoch": 0.15, "learning_rate": 5.980474459208259e-05, "loss": 0.4192, "step": 1441 }, { "epoch": 0.15, "learning_rate": 5.980447190000603e-05, "loss": 0.5493, "step": 1442 }, { "epoch": 0.15, "learning_rate": 5.9804199018265144e-05, "loss": 0.4352, "step": 1443 }, { "epoch": 0.15, "learning_rate": 5.980392594686168e-05, "loss": 1.202, "step": 1444 }, { "epoch": 0.15, "learning_rate": 5.980365268579738e-05, "loss": 0.4058, "step": 1445 }, { "epoch": 0.15, "learning_rate": 5.9803379235073964e-05, "loss": 1.2179, "step": 1446 }, { "epoch": 0.15, "learning_rate": 5.9803105594693176e-05, "loss": 0.4104, "step": 1447 }, { "epoch": 0.15, "learning_rate": 5.980283176465676e-05, "loss": 0.4885, "step": 1448 }, { "epoch": 0.15, "learning_rate": 5.980255774496647e-05, "loss": 0.8567, "step": 1449 }, { "epoch": 0.15, "learning_rate": 5.980228353562403e-05, "loss": 1.1628, "step": 1450 }, { "epoch": 0.15, "learning_rate": 5.980200913663121e-05, "loss": 1.2568, "step": 1451 }, { "epoch": 0.15, "learning_rate": 5.980173454798973e-05, "loss": 1.3571, "step": 1452 }, { "epoch": 0.15, "learning_rate": 5.980145976970134e-05, "loss": 1.1845, "step": 1453 }, { "epoch": 0.15, "learning_rate": 5.9801184801767814e-05, "loss": 0.4744, "step": 1454 }, { "epoch": 0.15, "learning_rate": 5.9800909644190874e-05, "loss": 0.8657, "step": 1455 }, { "epoch": 0.15, "learning_rate": 5.980063429697229e-05, "loss": 0.2679, "step": 1456 }, { "epoch": 0.15, "learning_rate": 5.9800358760113804e-05, "loss": 1.4714, "step": 1457 }, { "epoch": 0.15, "learning_rate": 5.980008303361716e-05, "loss": 0.615, "step": 1458 }, { "epoch": 0.15, "learning_rate": 5.9799807117484136e-05, "loss": 0.3973, "step": 1459 }, { "epoch": 0.15, "learning_rate": 5.979953101171647e-05, "loss": 0.2275, "step": 1460 }, { "epoch": 0.15, "learning_rate": 5.979925471631594e-05, "loss": 1.2127, "step": 1461 }, { "epoch": 0.15, "learning_rate": 5.9798978231284265e-05, "loss": 1.2615, "step": 1462 }, { "epoch": 0.15, "learning_rate": 5.9798701556623246e-05, "loss": 0.5551, "step": 1463 }, { "epoch": 0.15, "learning_rate": 5.9798424692334616e-05, "loss": 0.481, "step": 1464 }, { "epoch": 0.15, "learning_rate": 5.979814763842015e-05, "loss": 1.4045, "step": 1465 }, { "epoch": 0.15, "learning_rate": 5.9797870394881614e-05, "loss": 1.1194, "step": 1466 }, { "epoch": 0.15, "learning_rate": 5.979759296172077e-05, "loss": 0.4279, "step": 1467 }, { "epoch": 0.15, "learning_rate": 5.979731533893937e-05, "loss": 1.0279, "step": 1468 }, { "epoch": 0.15, "learning_rate": 5.979703752653919e-05, "loss": 1.4037, "step": 1469 }, { "epoch": 0.15, "learning_rate": 5.9796759524522006e-05, "loss": 0.7757, "step": 1470 }, { "epoch": 0.15, "learning_rate": 5.9796481332889576e-05, "loss": 0.5507, "step": 1471 }, { "epoch": 0.15, "learning_rate": 5.979620295164367e-05, "loss": 0.9744, "step": 1472 }, { "epoch": 0.15, "learning_rate": 5.979592438078607e-05, "loss": 0.5733, "step": 1473 }, { "epoch": 0.15, "learning_rate": 5.9795645620318545e-05, "loss": 0.3869, "step": 1474 }, { "epoch": 0.15, "learning_rate": 5.979536667024285e-05, "loss": 1.2466, "step": 1475 }, { "epoch": 0.15, "learning_rate": 5.979508753056079e-05, "loss": 0.4342, "step": 1476 }, { "epoch": 0.15, "learning_rate": 5.979480820127413e-05, "loss": 0.3171, "step": 1477 }, { "epoch": 0.15, "learning_rate": 5.9794528682384645e-05, "loss": 0.6888, "step": 1478 }, { "epoch": 0.15, "learning_rate": 5.97942489738941e-05, "loss": 1.0982, "step": 1479 }, { "epoch": 0.15, "learning_rate": 5.97939690758043e-05, "loss": 1.3033, "step": 1480 }, { "epoch": 0.15, "learning_rate": 5.979368898811701e-05, "loss": 1.2741, "step": 1481 }, { "epoch": 0.15, "learning_rate": 5.979340871083402e-05, "loss": 0.8522, "step": 1482 }, { "epoch": 0.15, "learning_rate": 5.979312824395711e-05, "loss": 0.2461, "step": 1483 }, { "epoch": 0.15, "learning_rate": 5.979284758748807e-05, "loss": 0.8648, "step": 1484 }, { "epoch": 0.15, "learning_rate": 5.9792566741428675e-05, "loss": 1.0591, "step": 1485 }, { "epoch": 0.15, "learning_rate": 5.979228570578072e-05, "loss": 0.8973, "step": 1486 }, { "epoch": 0.15, "learning_rate": 5.9792004480546e-05, "loss": 0.6757, "step": 1487 }, { "epoch": 0.15, "learning_rate": 5.979172306572628e-05, "loss": 0.8219, "step": 1488 }, { "epoch": 0.15, "learning_rate": 5.9791441461323386e-05, "loss": 0.9454, "step": 1489 }, { "epoch": 0.15, "learning_rate": 5.979115966733909e-05, "loss": 0.407, "step": 1490 }, { "epoch": 0.15, "learning_rate": 5.9790877683775174e-05, "loss": 0.9106, "step": 1491 }, { "epoch": 0.15, "learning_rate": 5.9790595510633456e-05, "loss": 0.3629, "step": 1492 }, { "epoch": 0.15, "learning_rate": 5.9790313147915715e-05, "loss": 0.6982, "step": 1493 }, { "epoch": 0.15, "learning_rate": 5.979003059562377e-05, "loss": 0.9565, "step": 1494 }, { "epoch": 0.15, "learning_rate": 5.978974785375938e-05, "loss": 0.5486, "step": 1495 }, { "epoch": 0.15, "learning_rate": 5.978946492232438e-05, "loss": 0.9852, "step": 1496 }, { "epoch": 0.15, "learning_rate": 5.978918180132056e-05, "loss": 1.2247, "step": 1497 }, { "epoch": 0.15, "learning_rate": 5.978889849074971e-05, "loss": 1.9676, "step": 1498 }, { "epoch": 0.15, "learning_rate": 5.978861499061365e-05, "loss": 0.5637, "step": 1499 }, { "epoch": 0.15, "learning_rate": 5.978833130091417e-05, "loss": 0.5143, "step": 1500 }, { "epoch": 0.15, "learning_rate": 5.978804742165308e-05, "loss": 1.1102, "step": 1501 }, { "epoch": 0.15, "learning_rate": 5.97877633528322e-05, "loss": 0.374, "step": 1502 }, { "epoch": 0.15, "learning_rate": 5.9787479094453314e-05, "loss": 1.2437, "step": 1503 }, { "epoch": 0.15, "learning_rate": 5.978719464651825e-05, "loss": 1.2112, "step": 1504 }, { "epoch": 0.15, "learning_rate": 5.978691000902881e-05, "loss": 1.9989, "step": 1505 }, { "epoch": 0.15, "learning_rate": 5.978662518198681e-05, "loss": 0.4426, "step": 1506 }, { "epoch": 0.15, "learning_rate": 5.978634016539405e-05, "loss": 0.8182, "step": 1507 }, { "epoch": 0.15, "learning_rate": 5.978605495925236e-05, "loss": 1.0572, "step": 1508 }, { "epoch": 0.15, "learning_rate": 5.978576956356354e-05, "loss": 1.3305, "step": 1509 }, { "epoch": 0.15, "learning_rate": 5.9785483978329416e-05, "loss": 0.7135, "step": 1510 }, { "epoch": 0.15, "learning_rate": 5.97851982035518e-05, "loss": 0.3329, "step": 1511 }, { "epoch": 0.15, "learning_rate": 5.978491223923253e-05, "loss": 0.5997, "step": 1512 }, { "epoch": 0.15, "learning_rate": 5.9784626085373394e-05, "loss": 0.9828, "step": 1513 }, { "epoch": 0.15, "learning_rate": 5.978433974197623e-05, "loss": 0.3034, "step": 1514 }, { "epoch": 0.15, "learning_rate": 5.978405320904286e-05, "loss": 1.2557, "step": 1515 }, { "epoch": 0.15, "learning_rate": 5.9783766486575104e-05, "loss": 1.1023, "step": 1516 }, { "epoch": 0.15, "learning_rate": 5.978347957457479e-05, "loss": 1.0233, "step": 1517 }, { "epoch": 0.15, "learning_rate": 5.978319247304374e-05, "loss": 0.192, "step": 1518 }, { "epoch": 0.15, "learning_rate": 5.978290518198379e-05, "loss": 1.1028, "step": 1519 }, { "epoch": 0.15, "learning_rate": 5.978261770139675e-05, "loss": 0.6244, "step": 1520 }, { "epoch": 0.15, "learning_rate": 5.978233003128447e-05, "loss": 0.4375, "step": 1521 }, { "epoch": 0.15, "learning_rate": 5.978204217164877e-05, "loss": 0.3215, "step": 1522 }, { "epoch": 0.15, "learning_rate": 5.9781754122491486e-05, "loss": 0.4045, "step": 1523 }, { "epoch": 0.15, "learning_rate": 5.978146588381444e-05, "loss": 0.57, "step": 1524 }, { "epoch": 0.15, "learning_rate": 5.978117745561948e-05, "loss": 0.4967, "step": 1525 }, { "epoch": 0.15, "learning_rate": 5.978088883790843e-05, "loss": 0.8931, "step": 1526 }, { "epoch": 0.16, "learning_rate": 5.978060003068314e-05, "loss": 0.8107, "step": 1527 }, { "epoch": 0.16, "learning_rate": 5.978031103394544e-05, "loss": 1.0542, "step": 1528 }, { "epoch": 0.16, "learning_rate": 5.978002184769717e-05, "loss": 0.3544, "step": 1529 }, { "epoch": 0.16, "learning_rate": 5.977973247194017e-05, "loss": 0.9444, "step": 1530 }, { "epoch": 0.16, "learning_rate": 5.977944290667627e-05, "loss": 0.4681, "step": 1531 }, { "epoch": 0.16, "learning_rate": 5.977915315190733e-05, "loss": 1.0468, "step": 1532 }, { "epoch": 0.16, "learning_rate": 5.9778863207635196e-05, "loss": 1.656, "step": 1533 }, { "epoch": 0.16, "learning_rate": 5.97785730738617e-05, "loss": 1.3162, "step": 1534 }, { "epoch": 0.16, "learning_rate": 5.97782827505887e-05, "loss": 1.266, "step": 1535 }, { "epoch": 0.16, "learning_rate": 5.9777992237818026e-05, "loss": 0.8591, "step": 1536 }, { "epoch": 0.16, "learning_rate": 5.977770153555154e-05, "loss": 0.5086, "step": 1537 }, { "epoch": 0.16, "learning_rate": 5.97774106437911e-05, "loss": 1.5029, "step": 1538 }, { "epoch": 0.16, "learning_rate": 5.977711956253853e-05, "loss": 1.0147, "step": 1539 }, { "epoch": 0.16, "learning_rate": 5.977682829179571e-05, "loss": 0.5487, "step": 1540 }, { "epoch": 0.16, "learning_rate": 5.9776536831564484e-05, "loss": 0.5364, "step": 1541 }, { "epoch": 0.16, "learning_rate": 5.977624518184669e-05, "loss": 0.6845, "step": 1542 }, { "epoch": 0.16, "learning_rate": 5.977595334264422e-05, "loss": 0.0891, "step": 1543 }, { "epoch": 0.16, "learning_rate": 5.977566131395889e-05, "loss": 1.3858, "step": 1544 }, { "epoch": 0.16, "learning_rate": 5.977536909579259e-05, "loss": 0.5517, "step": 1545 }, { "epoch": 0.16, "learning_rate": 5.9775076688147175e-05, "loss": 0.6742, "step": 1546 }, { "epoch": 0.16, "learning_rate": 5.977478409102448e-05, "loss": 0.3489, "step": 1547 }, { "epoch": 0.16, "learning_rate": 5.97744913044264e-05, "loss": 1.0084, "step": 1548 }, { "epoch": 0.16, "learning_rate": 5.977419832835478e-05, "loss": 0.7246, "step": 1549 }, { "epoch": 0.16, "learning_rate": 5.977390516281149e-05, "loss": 0.298, "step": 1550 }, { "epoch": 0.16, "learning_rate": 5.97736118077984e-05, "loss": 0.6454, "step": 1551 }, { "epoch": 0.16, "learning_rate": 5.977331826331736e-05, "loss": 0.8041, "step": 1552 }, { "epoch": 0.16, "learning_rate": 5.9773024529370246e-05, "loss": 0.869, "step": 1553 }, { "epoch": 0.16, "learning_rate": 5.977273060595895e-05, "loss": 0.6331, "step": 1554 }, { "epoch": 0.16, "learning_rate": 5.9772436493085304e-05, "loss": 0.0996, "step": 1555 }, { "epoch": 0.16, "learning_rate": 5.97721421907512e-05, "loss": 1.5555, "step": 1556 }, { "epoch": 0.16, "learning_rate": 5.9771847698958514e-05, "loss": 1.0741, "step": 1557 }, { "epoch": 0.16, "learning_rate": 5.9771553017709115e-05, "loss": 1.0844, "step": 1558 }, { "epoch": 0.16, "learning_rate": 5.977125814700488e-05, "loss": 1.0447, "step": 1559 }, { "epoch": 0.16, "learning_rate": 5.977096308684768e-05, "loss": 0.5347, "step": 1560 }, { "epoch": 0.16, "learning_rate": 5.97706678372394e-05, "loss": 0.6057, "step": 1561 }, { "epoch": 0.16, "learning_rate": 5.9770372398181904e-05, "loss": 0.6366, "step": 1562 }, { "epoch": 0.16, "learning_rate": 5.9770076769677094e-05, "loss": 0.3858, "step": 1563 }, { "epoch": 0.16, "learning_rate": 5.976978095172684e-05, "loss": 0.8321, "step": 1564 }, { "epoch": 0.16, "learning_rate": 5.9769484944333025e-05, "loss": 0.8182, "step": 1565 }, { "epoch": 0.16, "learning_rate": 5.9769188747497534e-05, "loss": 0.5166, "step": 1566 }, { "epoch": 0.16, "learning_rate": 5.976889236122224e-05, "loss": 0.6648, "step": 1567 }, { "epoch": 0.16, "learning_rate": 5.976859578550905e-05, "loss": 0.3111, "step": 1568 }, { "epoch": 0.16, "learning_rate": 5.976829902035984e-05, "loss": 0.8296, "step": 1569 }, { "epoch": 0.16, "learning_rate": 5.97680020657765e-05, "loss": 0.523, "step": 1570 }, { "epoch": 0.16, "learning_rate": 5.9767704921760916e-05, "loss": 0.854, "step": 1571 }, { "epoch": 0.16, "learning_rate": 5.9767407588314986e-05, "loss": 1.471, "step": 1572 }, { "epoch": 0.16, "learning_rate": 5.976711006544059e-05, "loss": 1.1114, "step": 1573 }, { "epoch": 0.16, "learning_rate": 5.9766812353139635e-05, "loss": 1.3511, "step": 1574 }, { "epoch": 0.16, "learning_rate": 5.9766514451414015e-05, "loss": 0.5587, "step": 1575 }, { "epoch": 0.16, "learning_rate": 5.976621636026561e-05, "loss": 1.2171, "step": 1576 }, { "epoch": 0.16, "learning_rate": 5.976591807969634e-05, "loss": 0.6853, "step": 1577 }, { "epoch": 0.16, "learning_rate": 5.976561960970808e-05, "loss": 0.8367, "step": 1578 }, { "epoch": 0.16, "learning_rate": 5.976532095030274e-05, "loss": 0.7848, "step": 1579 }, { "epoch": 0.16, "learning_rate": 5.9765022101482226e-05, "loss": 0.4049, "step": 1580 }, { "epoch": 0.16, "learning_rate": 5.9764723063248435e-05, "loss": 0.9316, "step": 1581 }, { "epoch": 0.16, "learning_rate": 5.976442383560327e-05, "loss": 0.9644, "step": 1582 }, { "epoch": 0.16, "learning_rate": 5.976412441854864e-05, "loss": 0.2059, "step": 1583 }, { "epoch": 0.16, "learning_rate": 5.976382481208643e-05, "loss": 1.0027, "step": 1584 }, { "epoch": 0.16, "learning_rate": 5.976352501621857e-05, "loss": 1.2819, "step": 1585 }, { "epoch": 0.16, "learning_rate": 5.9763225030946964e-05, "loss": 0.4628, "step": 1586 }, { "epoch": 0.16, "learning_rate": 5.976292485627351e-05, "loss": 1.1785, "step": 1587 }, { "epoch": 0.16, "learning_rate": 5.976262449220013e-05, "loss": 0.4949, "step": 1588 }, { "epoch": 0.16, "learning_rate": 5.976232393872872e-05, "loss": 0.3875, "step": 1589 }, { "epoch": 0.16, "learning_rate": 5.976202319586121e-05, "loss": 0.7528, "step": 1590 }, { "epoch": 0.16, "learning_rate": 5.97617222635995e-05, "loss": 0.3443, "step": 1591 }, { "epoch": 0.16, "learning_rate": 5.976142114194553e-05, "loss": 0.9777, "step": 1592 }, { "epoch": 0.16, "learning_rate": 5.9761119830901176e-05, "loss": 0.5877, "step": 1593 }, { "epoch": 0.16, "learning_rate": 5.976081833046839e-05, "loss": 0.391, "step": 1594 }, { "epoch": 0.16, "learning_rate": 5.976051664064907e-05, "loss": 1.1929, "step": 1595 }, { "epoch": 0.16, "learning_rate": 5.976021476144515e-05, "loss": 0.7854, "step": 1596 }, { "epoch": 0.16, "learning_rate": 5.975991269285854e-05, "loss": 1.5092, "step": 1597 }, { "epoch": 0.16, "learning_rate": 5.975961043489116e-05, "loss": 0.3309, "step": 1598 }, { "epoch": 0.16, "learning_rate": 5.9759307987544954e-05, "loss": 1.1042, "step": 1599 }, { "epoch": 0.16, "learning_rate": 5.9759005350821824e-05, "loss": 0.3911, "step": 1600 }, { "epoch": 0.16, "learning_rate": 5.975870252472372e-05, "loss": 1.4275, "step": 1601 }, { "epoch": 0.16, "learning_rate": 5.9758399509252536e-05, "loss": 0.7206, "step": 1602 }, { "epoch": 0.16, "learning_rate": 5.975809630441022e-05, "loss": 1.8159, "step": 1603 }, { "epoch": 0.16, "learning_rate": 5.9757792910198705e-05, "loss": 0.3662, "step": 1604 }, { "epoch": 0.16, "learning_rate": 5.975748932661992e-05, "loss": 1.221, "step": 1605 }, { "epoch": 0.16, "learning_rate": 5.975718555367579e-05, "loss": 0.3864, "step": 1606 }, { "epoch": 0.16, "learning_rate": 5.975688159136824e-05, "loss": 0.3457, "step": 1607 }, { "epoch": 0.16, "learning_rate": 5.9756577439699235e-05, "loss": 1.1481, "step": 1608 }, { "epoch": 0.16, "learning_rate": 5.975627309867068e-05, "loss": 0.66, "step": 1609 }, { "epoch": 0.16, "learning_rate": 5.975596856828452e-05, "loss": 0.835, "step": 1610 }, { "epoch": 0.16, "learning_rate": 5.9755663848542705e-05, "loss": 0.6892, "step": 1611 }, { "epoch": 0.16, "learning_rate": 5.975535893944716e-05, "loss": 0.3613, "step": 1612 }, { "epoch": 0.16, "learning_rate": 5.975505384099983e-05, "loss": 0.778, "step": 1613 }, { "epoch": 0.16, "learning_rate": 5.9754748553202656e-05, "loss": 1.2153, "step": 1614 }, { "epoch": 0.16, "learning_rate": 5.9754443076057585e-05, "loss": 0.2522, "step": 1615 }, { "epoch": 0.16, "learning_rate": 5.975413740956655e-05, "loss": 0.6208, "step": 1616 }, { "epoch": 0.16, "learning_rate": 5.9753831553731515e-05, "loss": 0.4896, "step": 1617 }, { "epoch": 0.16, "learning_rate": 5.975352550855441e-05, "loss": 0.3359, "step": 1618 }, { "epoch": 0.16, "learning_rate": 5.975321927403719e-05, "loss": 1.0462, "step": 1619 }, { "epoch": 0.16, "learning_rate": 5.975291285018179e-05, "loss": 0.338, "step": 1620 }, { "epoch": 0.16, "learning_rate": 5.975260623699019e-05, "loss": 0.7782, "step": 1621 }, { "epoch": 0.16, "learning_rate": 5.9752299434464304e-05, "loss": 1.2811, "step": 1622 }, { "epoch": 0.16, "learning_rate": 5.9751992442606114e-05, "loss": 0.9133, "step": 1623 }, { "epoch": 0.16, "learning_rate": 5.9751685261417554e-05, "loss": 0.5137, "step": 1624 }, { "epoch": 0.17, "learning_rate": 5.9751377890900596e-05, "loss": 0.3997, "step": 1625 }, { "epoch": 0.17, "learning_rate": 5.975107033105717e-05, "loss": 1.2322, "step": 1626 }, { "epoch": 0.17, "learning_rate": 5.975076258188927e-05, "loss": 0.4079, "step": 1627 }, { "epoch": 0.17, "learning_rate": 5.9750454643398824e-05, "loss": 0.3107, "step": 1628 }, { "epoch": 0.17, "learning_rate": 5.975014651558781e-05, "loss": 1.2167, "step": 1629 }, { "epoch": 0.17, "learning_rate": 5.974983819845818e-05, "loss": 0.3223, "step": 1630 }, { "epoch": 0.17, "learning_rate": 5.974952969201189e-05, "loss": 1.3834, "step": 1631 }, { "epoch": 0.17, "learning_rate": 5.9749220996250915e-05, "loss": 1.122, "step": 1632 }, { "epoch": 0.17, "learning_rate": 5.9748912111177214e-05, "loss": 0.3034, "step": 1633 }, { "epoch": 0.17, "learning_rate": 5.9748603036792754e-05, "loss": 0.5497, "step": 1634 }, { "epoch": 0.17, "learning_rate": 5.97482937730995e-05, "loss": 0.9487, "step": 1635 }, { "epoch": 0.17, "learning_rate": 5.974798432009942e-05, "loss": 1.1578, "step": 1636 }, { "epoch": 0.17, "learning_rate": 5.97476746777945e-05, "loss": 0.9865, "step": 1637 }, { "epoch": 0.17, "learning_rate": 5.974736484618667e-05, "loss": 0.5843, "step": 1638 }, { "epoch": 0.17, "learning_rate": 5.974705482527794e-05, "loss": 0.9233, "step": 1639 }, { "epoch": 0.17, "learning_rate": 5.974674461507027e-05, "loss": 0.8369, "step": 1640 }, { "epoch": 0.17, "learning_rate": 5.9746434215565624e-05, "loss": 1.2071, "step": 1641 }, { "epoch": 0.17, "learning_rate": 5.9746123626766004e-05, "loss": 0.7656, "step": 1642 }, { "epoch": 0.17, "learning_rate": 5.974581284867335e-05, "loss": 1.2365, "step": 1643 }, { "epoch": 0.17, "learning_rate": 5.974550188128967e-05, "loss": 0.7545, "step": 1644 }, { "epoch": 0.17, "learning_rate": 5.974519072461694e-05, "loss": 0.6029, "step": 1645 }, { "epoch": 0.17, "learning_rate": 5.974487937865711e-05, "loss": 0.6524, "step": 1646 }, { "epoch": 0.17, "learning_rate": 5.97445678434122e-05, "loss": 1.4184, "step": 1647 }, { "epoch": 0.17, "learning_rate": 5.9744256118884174e-05, "loss": 0.9958, "step": 1648 }, { "epoch": 0.17, "learning_rate": 5.974394420507501e-05, "loss": 0.6584, "step": 1649 }, { "epoch": 0.17, "learning_rate": 5.97436321019867e-05, "loss": 1.3797, "step": 1650 }, { "epoch": 0.17, "learning_rate": 5.9743319809621236e-05, "loss": 0.6361, "step": 1651 }, { "epoch": 0.17, "learning_rate": 5.9743007327980596e-05, "loss": 0.7146, "step": 1652 }, { "epoch": 0.17, "learning_rate": 5.974269465706678e-05, "loss": 0.7646, "step": 1653 }, { "epoch": 0.17, "learning_rate": 5.974238179688177e-05, "loss": 0.8582, "step": 1654 }, { "epoch": 0.17, "learning_rate": 5.974206874742754e-05, "loss": 1.0352, "step": 1655 }, { "epoch": 0.17, "learning_rate": 5.9741755508706116e-05, "loss": 0.5479, "step": 1656 }, { "epoch": 0.17, "learning_rate": 5.974144208071947e-05, "loss": 0.7961, "step": 1657 }, { "epoch": 0.17, "learning_rate": 5.9741128463469594e-05, "loss": 0.4908, "step": 1658 }, { "epoch": 0.17, "learning_rate": 5.97408146569585e-05, "loss": 1.0629, "step": 1659 }, { "epoch": 0.17, "learning_rate": 5.974050066118817e-05, "loss": 0.7654, "step": 1660 }, { "epoch": 0.17, "learning_rate": 5.974018647616061e-05, "loss": 0.608, "step": 1661 }, { "epoch": 0.17, "learning_rate": 5.973987210187782e-05, "loss": 2.1207, "step": 1662 }, { "epoch": 0.17, "learning_rate": 5.973955753834179e-05, "loss": 0.6886, "step": 1663 }, { "epoch": 0.17, "learning_rate": 5.9739242785554524e-05, "loss": 0.8175, "step": 1664 }, { "epoch": 0.17, "learning_rate": 5.973892784351805e-05, "loss": 1.1177, "step": 1665 }, { "epoch": 0.17, "learning_rate": 5.973861271223434e-05, "loss": 0.797, "step": 1666 }, { "epoch": 0.17, "learning_rate": 5.9738297391705415e-05, "loss": 0.5413, "step": 1667 }, { "epoch": 0.17, "learning_rate": 5.973798188193327e-05, "loss": 1.2156, "step": 1668 }, { "epoch": 0.17, "learning_rate": 5.973766618291993e-05, "loss": 0.4501, "step": 1669 }, { "epoch": 0.17, "learning_rate": 5.973735029466739e-05, "loss": 0.4866, "step": 1670 }, { "epoch": 0.17, "learning_rate": 5.9737034217177676e-05, "loss": 0.9219, "step": 1671 }, { "epoch": 0.17, "learning_rate": 5.973671795045277e-05, "loss": 0.5281, "step": 1672 }, { "epoch": 0.17, "learning_rate": 5.973640149449472e-05, "loss": 1.1787, "step": 1673 }, { "epoch": 0.17, "learning_rate": 5.973608484930552e-05, "loss": 1.8833, "step": 1674 }, { "epoch": 0.17, "learning_rate": 5.9735768014887184e-05, "loss": 1.384, "step": 1675 }, { "epoch": 0.17, "learning_rate": 5.973545099124174e-05, "loss": 0.4122, "step": 1676 }, { "epoch": 0.17, "learning_rate": 5.973513377837119e-05, "loss": 1.2693, "step": 1677 }, { "epoch": 0.17, "learning_rate": 5.9734816376277554e-05, "loss": 0.3397, "step": 1678 }, { "epoch": 0.17, "learning_rate": 5.973449878496287e-05, "loss": 0.9303, "step": 1679 }, { "epoch": 0.17, "learning_rate": 5.973418100442915e-05, "loss": 0.676, "step": 1680 }, { "epoch": 0.17, "learning_rate": 5.973386303467841e-05, "loss": 0.8367, "step": 1681 }, { "epoch": 0.17, "learning_rate": 5.973354487571267e-05, "loss": 0.5123, "step": 1682 }, { "epoch": 0.17, "learning_rate": 5.9733226527533966e-05, "loss": 0.8957, "step": 1683 }, { "epoch": 0.17, "learning_rate": 5.9732907990144325e-05, "loss": 1.4406, "step": 1684 }, { "epoch": 0.17, "learning_rate": 5.9732589263545765e-05, "loss": 0.3705, "step": 1685 }, { "epoch": 0.17, "learning_rate": 5.973227034774031e-05, "loss": 1.2025, "step": 1686 }, { "epoch": 0.17, "learning_rate": 5.973195124273001e-05, "loss": 0.4599, "step": 1687 }, { "epoch": 0.17, "learning_rate": 5.973163194851687e-05, "loss": 1.2127, "step": 1688 }, { "epoch": 0.17, "learning_rate": 5.973131246510294e-05, "loss": 0.1462, "step": 1689 }, { "epoch": 0.17, "learning_rate": 5.973099279249025e-05, "loss": 1.0677, "step": 1690 }, { "epoch": 0.17, "learning_rate": 5.9730672930680836e-05, "loss": 0.5677, "step": 1691 }, { "epoch": 0.17, "learning_rate": 5.9730352879676725e-05, "loss": 0.5824, "step": 1692 }, { "epoch": 0.17, "learning_rate": 5.973003263947996e-05, "loss": 1.0222, "step": 1693 }, { "epoch": 0.17, "learning_rate": 5.9729712210092575e-05, "loss": 0.2273, "step": 1694 }, { "epoch": 0.17, "learning_rate": 5.972939159151661e-05, "loss": 1.3486, "step": 1695 }, { "epoch": 0.17, "learning_rate": 5.972907078375411e-05, "loss": 1.2736, "step": 1696 }, { "epoch": 0.17, "learning_rate": 5.972874978680711e-05, "loss": 1.1287, "step": 1697 }, { "epoch": 0.17, "learning_rate": 5.972842860067766e-05, "loss": 0.2947, "step": 1698 }, { "epoch": 0.17, "learning_rate": 5.972810722536779e-05, "loss": 1.2875, "step": 1699 }, { "epoch": 0.17, "learning_rate": 5.972778566087957e-05, "loss": 1.3454, "step": 1700 }, { "epoch": 0.17, "learning_rate": 5.972746390721502e-05, "loss": 1.3171, "step": 1701 }, { "epoch": 0.17, "learning_rate": 5.97271419643762e-05, "loss": 0.8134, "step": 1702 }, { "epoch": 0.17, "learning_rate": 5.972681983236515e-05, "loss": 1.9027, "step": 1703 }, { "epoch": 0.17, "learning_rate": 5.972649751118394e-05, "loss": 0.9059, "step": 1704 }, { "epoch": 0.17, "learning_rate": 5.97261750008346e-05, "loss": 0.8212, "step": 1705 }, { "epoch": 0.17, "learning_rate": 5.9725852301319197e-05, "loss": 1.1876, "step": 1706 }, { "epoch": 0.17, "learning_rate": 5.972552941263978e-05, "loss": 0.692, "step": 1707 }, { "epoch": 0.17, "learning_rate": 5.972520633479839e-05, "loss": 0.5219, "step": 1708 }, { "epoch": 0.17, "learning_rate": 5.97248830677971e-05, "loss": 0.9708, "step": 1709 }, { "epoch": 0.17, "learning_rate": 5.972455961163797e-05, "loss": 1.39, "step": 1710 }, { "epoch": 0.17, "learning_rate": 5.9724235966323046e-05, "loss": 0.7811, "step": 1711 }, { "epoch": 0.17, "learning_rate": 5.972391213185439e-05, "loss": 0.5763, "step": 1712 }, { "epoch": 0.17, "learning_rate": 5.972358810823407e-05, "loss": 0.2155, "step": 1713 }, { "epoch": 0.17, "learning_rate": 5.972326389546414e-05, "loss": 0.7395, "step": 1714 }, { "epoch": 0.17, "learning_rate": 5.972293949354666e-05, "loss": 0.4732, "step": 1715 }, { "epoch": 0.17, "learning_rate": 5.972261490248371e-05, "loss": 1.476, "step": 1716 }, { "epoch": 0.17, "learning_rate": 5.972229012227734e-05, "loss": 0.6725, "step": 1717 }, { "epoch": 0.17, "learning_rate": 5.972196515292962e-05, "loss": 1.0131, "step": 1718 }, { "epoch": 0.17, "learning_rate": 5.972163999444263e-05, "loss": 0.9768, "step": 1719 }, { "epoch": 0.17, "learning_rate": 5.9721314646818426e-05, "loss": 1.0049, "step": 1720 }, { "epoch": 0.17, "learning_rate": 5.972098911005908e-05, "loss": 0.3555, "step": 1721 }, { "epoch": 0.17, "learning_rate": 5.972066338416666e-05, "loss": 0.2915, "step": 1722 }, { "epoch": 0.17, "learning_rate": 5.972033746914325e-05, "loss": 0.1314, "step": 1723 }, { "epoch": 0.18, "learning_rate": 5.972001136499092e-05, "loss": 0.5318, "step": 1724 }, { "epoch": 0.18, "learning_rate": 5.971968507171174e-05, "loss": 0.6849, "step": 1725 }, { "epoch": 0.18, "learning_rate": 5.971935858930779e-05, "loss": 0.6314, "step": 1726 }, { "epoch": 0.18, "learning_rate": 5.971903191778116e-05, "loss": 0.4396, "step": 1727 }, { "epoch": 0.18, "learning_rate": 5.9718705057133905e-05, "loss": 0.6359, "step": 1728 }, { "epoch": 0.18, "learning_rate": 5.9718378007368117e-05, "loss": 0.5555, "step": 1729 }, { "epoch": 0.18, "learning_rate": 5.971805076848588e-05, "loss": 0.8444, "step": 1730 }, { "epoch": 0.18, "learning_rate": 5.971772334048926e-05, "loss": 0.3761, "step": 1731 }, { "epoch": 0.18, "learning_rate": 5.971739572338037e-05, "loss": 1.2886, "step": 1732 }, { "epoch": 0.18, "learning_rate": 5.971706791716128e-05, "loss": 1.5773, "step": 1733 }, { "epoch": 0.18, "learning_rate": 5.971673992183407e-05, "loss": 1.0312, "step": 1734 }, { "epoch": 0.18, "learning_rate": 5.9716411737400836e-05, "loss": 0.6728, "step": 1735 }, { "epoch": 0.18, "learning_rate": 5.971608336386365e-05, "loss": 0.3699, "step": 1736 }, { "epoch": 0.18, "learning_rate": 5.971575480122462e-05, "loss": 0.5057, "step": 1737 }, { "epoch": 0.18, "learning_rate": 5.971542604948584e-05, "loss": 1.6902, "step": 1738 }, { "epoch": 0.18, "learning_rate": 5.9715097108649376e-05, "loss": 0.6212, "step": 1739 }, { "epoch": 0.18, "learning_rate": 5.971476797871735e-05, "loss": 0.8131, "step": 1740 }, { "epoch": 0.18, "learning_rate": 5.971443865969184e-05, "loss": 1.0371, "step": 1741 }, { "epoch": 0.18, "learning_rate": 5.971410915157495e-05, "loss": 0.8871, "step": 1742 }, { "epoch": 0.18, "learning_rate": 5.971377945436877e-05, "loss": 0.4749, "step": 1743 }, { "epoch": 0.18, "learning_rate": 5.971344956807541e-05, "loss": 1.3856, "step": 1744 }, { "epoch": 0.18, "learning_rate": 5.971311949269695e-05, "loss": 0.8585, "step": 1745 }, { "epoch": 0.18, "learning_rate": 5.9712789228235506e-05, "loss": 0.8252, "step": 1746 }, { "epoch": 0.18, "learning_rate": 5.971245877469317e-05, "loss": 0.9857, "step": 1747 }, { "epoch": 0.18, "learning_rate": 5.971212813207205e-05, "loss": 0.899, "step": 1748 }, { "epoch": 0.18, "learning_rate": 5.971179730037425e-05, "loss": 0.6917, "step": 1749 }, { "epoch": 0.18, "learning_rate": 5.9711466279601875e-05, "loss": 0.5349, "step": 1750 }, { "epoch": 0.18, "learning_rate": 5.971113506975704e-05, "loss": 0.5645, "step": 1751 }, { "epoch": 0.18, "learning_rate": 5.971080367084183e-05, "loss": 0.5357, "step": 1752 }, { "epoch": 0.18, "learning_rate": 5.971047208285838e-05, "loss": 1.3394, "step": 1753 }, { "epoch": 0.18, "learning_rate": 5.971014030580879e-05, "loss": 0.5505, "step": 1754 }, { "epoch": 0.18, "learning_rate": 5.970980833969516e-05, "loss": 0.5869, "step": 1755 }, { "epoch": 0.18, "learning_rate": 5.970947618451962e-05, "loss": 0.4634, "step": 1756 }, { "epoch": 0.18, "learning_rate": 5.970914384028426e-05, "loss": 1.0956, "step": 1757 }, { "epoch": 0.18, "learning_rate": 5.9708811306991225e-05, "loss": 1.302, "step": 1758 }, { "epoch": 0.18, "learning_rate": 5.970847858464262e-05, "loss": 0.6571, "step": 1759 }, { "epoch": 0.18, "learning_rate": 5.9708145673240545e-05, "loss": 0.5773, "step": 1760 }, { "epoch": 0.18, "learning_rate": 5.9707812572787136e-05, "loss": 1.7692, "step": 1761 }, { "epoch": 0.18, "learning_rate": 5.970747928328451e-05, "loss": 1.4014, "step": 1762 }, { "epoch": 0.18, "learning_rate": 5.970714580473479e-05, "loss": 0.3864, "step": 1763 }, { "epoch": 0.18, "learning_rate": 5.9706812137140104e-05, "loss": 1.1486, "step": 1764 }, { "epoch": 0.18, "learning_rate": 5.970647828050256e-05, "loss": 1.6944, "step": 1765 }, { "epoch": 0.18, "learning_rate": 5.970614423482428e-05, "loss": 0.8864, "step": 1766 }, { "epoch": 0.18, "learning_rate": 5.970581000010741e-05, "loss": 0.8034, "step": 1767 }, { "epoch": 0.18, "learning_rate": 5.970547557635406e-05, "loss": 1.378, "step": 1768 }, { "epoch": 0.18, "learning_rate": 5.9705140963566366e-05, "loss": 0.8189, "step": 1769 }, { "epoch": 0.18, "learning_rate": 5.970480616174645e-05, "loss": 0.8623, "step": 1770 }, { "epoch": 0.18, "learning_rate": 5.970447117089645e-05, "loss": 1.4196, "step": 1771 }, { "epoch": 0.18, "learning_rate": 5.9704135991018505e-05, "loss": 0.3533, "step": 1772 }, { "epoch": 0.18, "learning_rate": 5.970380062211473e-05, "loss": 1.1769, "step": 1773 }, { "epoch": 0.18, "learning_rate": 5.970346506418727e-05, "loss": 1.2751, "step": 1774 }, { "epoch": 0.18, "learning_rate": 5.970312931723826e-05, "loss": 2.772, "step": 1775 }, { "epoch": 0.18, "learning_rate": 5.9702793381269826e-05, "loss": 0.492, "step": 1776 }, { "epoch": 0.18, "learning_rate": 5.9702457256284125e-05, "loss": 1.2609, "step": 1777 }, { "epoch": 0.18, "learning_rate": 5.970212094228328e-05, "loss": 0.8036, "step": 1778 }, { "epoch": 0.18, "learning_rate": 5.970178443926944e-05, "loss": 1.4552, "step": 1779 }, { "epoch": 0.18, "learning_rate": 5.9701447747244734e-05, "loss": 0.1135, "step": 1780 }, { "epoch": 0.18, "learning_rate": 5.970111086621133e-05, "loss": 0.5206, "step": 1781 }, { "epoch": 0.18, "learning_rate": 5.9700773796171345e-05, "loss": 1.5116, "step": 1782 }, { "epoch": 0.18, "learning_rate": 5.9700436537126936e-05, "loss": 1.0096, "step": 1783 }, { "epoch": 0.18, "learning_rate": 5.9700099089080244e-05, "loss": 1.0872, "step": 1784 }, { "epoch": 0.18, "learning_rate": 5.9699761452033424e-05, "loss": 0.5911, "step": 1785 }, { "epoch": 0.18, "learning_rate": 5.9699423625988616e-05, "loss": 0.9611, "step": 1786 }, { "epoch": 0.18, "learning_rate": 5.969908561094797e-05, "loss": 0.5215, "step": 1787 }, { "epoch": 0.18, "learning_rate": 5.969874740691365e-05, "loss": 0.6323, "step": 1788 }, { "epoch": 0.18, "learning_rate": 5.9698409013887805e-05, "loss": 1.1345, "step": 1789 }, { "epoch": 0.18, "learning_rate": 5.969807043187257e-05, "loss": 1.1169, "step": 1790 }, { "epoch": 0.18, "learning_rate": 5.9697731660870116e-05, "loss": 0.4197, "step": 1791 }, { "epoch": 0.18, "learning_rate": 5.96973927008826e-05, "loss": 0.4604, "step": 1792 }, { "epoch": 0.18, "learning_rate": 5.9697053551912173e-05, "loss": 1.4395, "step": 1793 }, { "epoch": 0.18, "learning_rate": 5.9696714213960996e-05, "loss": 0.2961, "step": 1794 }, { "epoch": 0.18, "learning_rate": 5.9696374687031224e-05, "loss": 0.634, "step": 1795 }, { "epoch": 0.18, "learning_rate": 5.9696034971125014e-05, "loss": 0.4418, "step": 1796 }, { "epoch": 0.18, "learning_rate": 5.969569506624455e-05, "loss": 1.0772, "step": 1797 }, { "epoch": 0.18, "learning_rate": 5.969535497239197e-05, "loss": 0.8826, "step": 1798 }, { "epoch": 0.18, "learning_rate": 5.969501468956944e-05, "loss": 1.3446, "step": 1799 }, { "epoch": 0.18, "learning_rate": 5.969467421777915e-05, "loss": 0.812, "step": 1800 }, { "epoch": 0.18, "learning_rate": 5.969433355702324e-05, "loss": 0.5114, "step": 1801 }, { "epoch": 0.18, "learning_rate": 5.969399270730389e-05, "loss": 1.2678, "step": 1802 }, { "epoch": 0.18, "learning_rate": 5.9693651668623265e-05, "loss": 0.6259, "step": 1803 }, { "epoch": 0.18, "learning_rate": 5.9693310440983534e-05, "loss": 2.0559, "step": 1804 }, { "epoch": 0.18, "learning_rate": 5.969296902438687e-05, "loss": 1.2513, "step": 1805 }, { "epoch": 0.18, "learning_rate": 5.969262741883546e-05, "loss": 0.5999, "step": 1806 }, { "epoch": 0.18, "learning_rate": 5.969228562433145e-05, "loss": 0.4483, "step": 1807 }, { "epoch": 0.18, "learning_rate": 5.969194364087704e-05, "loss": 0.5862, "step": 1808 }, { "epoch": 0.18, "learning_rate": 5.9691601468474386e-05, "loss": 1.0727, "step": 1809 }, { "epoch": 0.18, "learning_rate": 5.969125910712569e-05, "loss": 0.5698, "step": 1810 }, { "epoch": 0.18, "learning_rate": 5.969091655683311e-05, "loss": 0.5367, "step": 1811 }, { "epoch": 0.18, "learning_rate": 5.969057381759882e-05, "loss": 1.3601, "step": 1812 }, { "epoch": 0.18, "learning_rate": 5.969023088942503e-05, "loss": 0.5868, "step": 1813 }, { "epoch": 0.18, "learning_rate": 5.968988777231389e-05, "loss": 0.7297, "step": 1814 }, { "epoch": 0.18, "learning_rate": 5.968954446626761e-05, "loss": 0.5578, "step": 1815 }, { "epoch": 0.18, "learning_rate": 5.968920097128836e-05, "loss": 0.476, "step": 1816 }, { "epoch": 0.18, "learning_rate": 5.9688857287378336e-05, "loss": 0.9506, "step": 1817 }, { "epoch": 0.18, "learning_rate": 5.9688513414539705e-05, "loss": 0.9572, "step": 1818 }, { "epoch": 0.18, "learning_rate": 5.968816935277468e-05, "loss": 0.398, "step": 1819 }, { "epoch": 0.18, "learning_rate": 5.968782510208543e-05, "loss": 0.2044, "step": 1820 }, { "epoch": 0.18, "learning_rate": 5.968748066247416e-05, "loss": 0.3861, "step": 1821 }, { "epoch": 0.19, "learning_rate": 5.968713603394306e-05, "loss": 0.3498, "step": 1822 }, { "epoch": 0.19, "learning_rate": 5.968679121649432e-05, "loss": 0.9798, "step": 1823 }, { "epoch": 0.19, "learning_rate": 5.968644621013013e-05, "loss": 1.0584, "step": 1824 }, { "epoch": 0.19, "learning_rate": 5.968610101485269e-05, "loss": 1.4247, "step": 1825 }, { "epoch": 0.19, "learning_rate": 5.9685755630664196e-05, "loss": 1.4017, "step": 1826 }, { "epoch": 0.19, "learning_rate": 5.968541005756684e-05, "loss": 0.9637, "step": 1827 }, { "epoch": 0.19, "learning_rate": 5.968506429556283e-05, "loss": 0.4174, "step": 1828 }, { "epoch": 0.19, "learning_rate": 5.9684718344654376e-05, "loss": 0.3928, "step": 1829 }, { "epoch": 0.19, "learning_rate": 5.968437220484366e-05, "loss": 1.15, "step": 1830 }, { "epoch": 0.19, "learning_rate": 5.968402587613289e-05, "loss": 0.9854, "step": 1831 }, { "epoch": 0.19, "learning_rate": 5.9683679358524274e-05, "loss": 0.748, "step": 1832 }, { "epoch": 0.19, "learning_rate": 5.9683332652020016e-05, "loss": 0.2382, "step": 1833 }, { "epoch": 0.19, "learning_rate": 5.9682985756622314e-05, "loss": 1.0285, "step": 1834 }, { "epoch": 0.19, "learning_rate": 5.968263867233339e-05, "loss": 0.7253, "step": 1835 }, { "epoch": 0.19, "learning_rate": 5.9682291399155445e-05, "loss": 1.0218, "step": 1836 }, { "epoch": 0.19, "learning_rate": 5.968194393709068e-05, "loss": 0.4681, "step": 1837 }, { "epoch": 0.19, "learning_rate": 5.968159628614132e-05, "loss": 1.1544, "step": 1838 }, { "epoch": 0.19, "learning_rate": 5.968124844630958e-05, "loss": 0.9235, "step": 1839 }, { "epoch": 0.19, "learning_rate": 5.968090041759766e-05, "loss": 0.5093, "step": 1840 }, { "epoch": 0.19, "learning_rate": 5.9680552200007784e-05, "loss": 0.9123, "step": 1841 }, { "epoch": 0.19, "learning_rate": 5.968020379354216e-05, "loss": 0.6413, "step": 1842 }, { "epoch": 0.19, "learning_rate": 5.9679855198203014e-05, "loss": 0.3583, "step": 1843 }, { "epoch": 0.19, "learning_rate": 5.967950641399256e-05, "loss": 0.3314, "step": 1844 }, { "epoch": 0.19, "learning_rate": 5.967915744091302e-05, "loss": 1.4601, "step": 1845 }, { "epoch": 0.19, "learning_rate": 5.9678808278966605e-05, "loss": 0.6474, "step": 1846 }, { "epoch": 0.19, "learning_rate": 5.967845892815555e-05, "loss": 0.9649, "step": 1847 }, { "epoch": 0.19, "learning_rate": 5.967810938848207e-05, "loss": 0.6192, "step": 1848 }, { "epoch": 0.19, "learning_rate": 5.967775965994839e-05, "loss": 1.2495, "step": 1849 }, { "epoch": 0.19, "learning_rate": 5.967740974255674e-05, "loss": 0.2614, "step": 1850 }, { "epoch": 0.19, "learning_rate": 5.967705963630935e-05, "loss": 0.5897, "step": 1851 }, { "epoch": 0.19, "learning_rate": 5.9676709341208434e-05, "loss": 0.9869, "step": 1852 }, { "epoch": 0.19, "learning_rate": 5.967635885725624e-05, "loss": 1.766, "step": 1853 }, { "epoch": 0.19, "learning_rate": 5.9676008184454975e-05, "loss": 1.3281, "step": 1854 }, { "epoch": 0.19, "learning_rate": 5.967565732280689e-05, "loss": 1.4757, "step": 1855 }, { "epoch": 0.19, "learning_rate": 5.9675306272314205e-05, "loss": 1.2879, "step": 1856 }, { "epoch": 0.19, "learning_rate": 5.9674955032979164e-05, "loss": 1.0235, "step": 1857 }, { "epoch": 0.19, "learning_rate": 5.9674603604804e-05, "loss": 0.8651, "step": 1858 }, { "epoch": 0.19, "learning_rate": 5.967425198779095e-05, "loss": 2.2621, "step": 1859 }, { "epoch": 0.19, "learning_rate": 5.967390018194224e-05, "loss": 1.1933, "step": 1860 }, { "epoch": 0.19, "learning_rate": 5.967354818726012e-05, "loss": 1.2548, "step": 1861 }, { "epoch": 0.19, "learning_rate": 5.967319600374684e-05, "loss": 1.1124, "step": 1862 }, { "epoch": 0.19, "learning_rate": 5.967284363140462e-05, "loss": 0.7509, "step": 1863 }, { "epoch": 0.19, "learning_rate": 5.967249107023571e-05, "loss": 0.8679, "step": 1864 }, { "epoch": 0.19, "learning_rate": 5.967213832024235e-05, "loss": 0.3401, "step": 1865 }, { "epoch": 0.19, "learning_rate": 5.96717853814268e-05, "loss": 0.3983, "step": 1866 }, { "epoch": 0.19, "learning_rate": 5.967143225379129e-05, "loss": 0.4677, "step": 1867 }, { "epoch": 0.19, "learning_rate": 5.967107893733807e-05, "loss": 1.1463, "step": 1868 }, { "epoch": 0.19, "learning_rate": 5.96707254320694e-05, "loss": 1.109, "step": 1869 }, { "epoch": 0.19, "learning_rate": 5.967037173798751e-05, "loss": 1.2674, "step": 1870 }, { "epoch": 0.19, "learning_rate": 5.9670017855094664e-05, "loss": 0.7873, "step": 1871 }, { "epoch": 0.19, "learning_rate": 5.966966378339312e-05, "loss": 1.0374, "step": 1872 }, { "epoch": 0.19, "learning_rate": 5.966930952288512e-05, "loss": 1.5313, "step": 1873 }, { "epoch": 0.19, "learning_rate": 5.966895507357292e-05, "loss": 0.4504, "step": 1874 }, { "epoch": 0.19, "learning_rate": 5.966860043545877e-05, "loss": 1.2439, "step": 1875 }, { "epoch": 0.19, "learning_rate": 5.966824560854494e-05, "loss": 1.1837, "step": 1876 }, { "epoch": 0.19, "learning_rate": 5.9667890592833675e-05, "loss": 1.2827, "step": 1877 }, { "epoch": 0.19, "learning_rate": 5.966753538832725e-05, "loss": 0.9065, "step": 1878 }, { "epoch": 0.19, "learning_rate": 5.9667179995027905e-05, "loss": 0.6224, "step": 1879 }, { "epoch": 0.19, "learning_rate": 5.966682441293792e-05, "loss": 0.3902, "step": 1880 }, { "epoch": 0.19, "learning_rate": 5.9666468642059554e-05, "loss": 0.5243, "step": 1881 }, { "epoch": 0.19, "learning_rate": 5.966611268239505e-05, "loss": 0.6128, "step": 1882 }, { "epoch": 0.19, "learning_rate": 5.966575653394671e-05, "loss": 0.6561, "step": 1883 }, { "epoch": 0.19, "learning_rate": 5.966540019671677e-05, "loss": 1.3146, "step": 1884 }, { "epoch": 0.19, "learning_rate": 5.966504367070751e-05, "loss": 0.4909, "step": 1885 }, { "epoch": 0.19, "learning_rate": 5.96646869559212e-05, "loss": 0.485, "step": 1886 }, { "epoch": 0.19, "learning_rate": 5.966433005236011e-05, "loss": 0.6364, "step": 1887 }, { "epoch": 0.19, "learning_rate": 5.96639729600265e-05, "loss": 1.1639, "step": 1888 }, { "epoch": 0.19, "learning_rate": 5.9663615678922654e-05, "loss": 0.8256, "step": 1889 }, { "epoch": 0.19, "learning_rate": 5.966325820905085e-05, "loss": 1.5423, "step": 1890 }, { "epoch": 0.19, "learning_rate": 5.966290055041335e-05, "loss": 0.9053, "step": 1891 }, { "epoch": 0.19, "learning_rate": 5.966254270301243e-05, "loss": 0.3285, "step": 1892 }, { "epoch": 0.19, "learning_rate": 5.966218466685037e-05, "loss": 1.1316, "step": 1893 }, { "epoch": 0.19, "learning_rate": 5.9661826441929455e-05, "loss": 1.3753, "step": 1894 }, { "epoch": 0.19, "learning_rate": 5.9661468028251975e-05, "loss": 0.5123, "step": 1895 }, { "epoch": 0.19, "learning_rate": 5.9661109425820175e-05, "loss": 1.2879, "step": 1896 }, { "epoch": 0.19, "learning_rate": 5.9660750634636366e-05, "loss": 0.8628, "step": 1897 }, { "epoch": 0.19, "learning_rate": 5.966039165470283e-05, "loss": 1.0976, "step": 1898 }, { "epoch": 0.19, "learning_rate": 5.9660032486021836e-05, "loss": 0.438, "step": 1899 }, { "epoch": 0.19, "learning_rate": 5.965967312859568e-05, "loss": 1.4286, "step": 1900 }, { "epoch": 0.19, "learning_rate": 5.965931358242666e-05, "loss": 0.4039, "step": 1901 }, { "epoch": 0.19, "learning_rate": 5.9658953847517034e-05, "loss": 0.94, "step": 1902 }, { "epoch": 0.19, "learning_rate": 5.9658593923869114e-05, "loss": 0.9372, "step": 1903 }, { "epoch": 0.19, "learning_rate": 5.9658233811485185e-05, "loss": 0.5633, "step": 1904 }, { "epoch": 0.19, "learning_rate": 5.965787351036754e-05, "loss": 0.2773, "step": 1905 }, { "epoch": 0.19, "learning_rate": 5.965751302051847e-05, "loss": 0.8922, "step": 1906 }, { "epoch": 0.19, "learning_rate": 5.9657152341940266e-05, "loss": 1.583, "step": 1907 }, { "epoch": 0.19, "learning_rate": 5.965679147463524e-05, "loss": 0.8569, "step": 1908 }, { "epoch": 0.19, "learning_rate": 5.9656430418605664e-05, "loss": 0.6852, "step": 1909 }, { "epoch": 0.19, "learning_rate": 5.9656069173853846e-05, "loss": 0.6526, "step": 1910 }, { "epoch": 0.19, "learning_rate": 5.965570774038209e-05, "loss": 0.7228, "step": 1911 }, { "epoch": 0.19, "learning_rate": 5.9655346118192684e-05, "loss": 0.9298, "step": 1912 }, { "epoch": 0.19, "learning_rate": 5.965498430728794e-05, "loss": 1.0741, "step": 1913 }, { "epoch": 0.19, "learning_rate": 5.9654622307670166e-05, "loss": 0.9284, "step": 1914 }, { "epoch": 0.19, "learning_rate": 5.965426011934165e-05, "loss": 1.0235, "step": 1915 }, { "epoch": 0.19, "learning_rate": 5.96538977423047e-05, "loss": 0.5426, "step": 1916 }, { "epoch": 0.19, "learning_rate": 5.9653535176561636e-05, "loss": 0.2982, "step": 1917 }, { "epoch": 0.19, "learning_rate": 5.9653172422114746e-05, "loss": 0.6485, "step": 1918 }, { "epoch": 0.19, "learning_rate": 5.965280947896635e-05, "loss": 0.7931, "step": 1919 }, { "epoch": 0.19, "learning_rate": 5.9652446347118755e-05, "loss": 0.7334, "step": 1920 }, { "epoch": 0.2, "learning_rate": 5.965208302657427e-05, "loss": 1.4552, "step": 1921 }, { "epoch": 0.2, "learning_rate": 5.9651719517335214e-05, "loss": 0.5567, "step": 1922 }, { "epoch": 0.2, "learning_rate": 5.965135581940389e-05, "loss": 1.0262, "step": 1923 }, { "epoch": 0.2, "learning_rate": 5.9650991932782616e-05, "loss": 1.3212, "step": 1924 }, { "epoch": 0.2, "learning_rate": 5.9650627857473714e-05, "loss": 1.0592, "step": 1925 }, { "epoch": 0.2, "learning_rate": 5.965026359347949e-05, "loss": 0.9559, "step": 1926 }, { "epoch": 0.2, "learning_rate": 5.964989914080227e-05, "loss": 0.3213, "step": 1927 }, { "epoch": 0.2, "learning_rate": 5.964953449944438e-05, "loss": 1.4297, "step": 1928 }, { "epoch": 0.2, "learning_rate": 5.9649169669408126e-05, "loss": 0.2667, "step": 1929 }, { "epoch": 0.2, "learning_rate": 5.9648804650695834e-05, "loss": 0.4263, "step": 1930 }, { "epoch": 0.2, "learning_rate": 5.964843944330982e-05, "loss": 1.3483, "step": 1931 }, { "epoch": 0.2, "learning_rate": 5.964807404725243e-05, "loss": 0.3982, "step": 1932 }, { "epoch": 0.2, "learning_rate": 5.964770846252597e-05, "loss": 0.2619, "step": 1933 }, { "epoch": 0.2, "learning_rate": 5.9647342689132774e-05, "loss": 1.4046, "step": 1934 }, { "epoch": 0.2, "learning_rate": 5.9646976727075167e-05, "loss": 1.1999, "step": 1935 }, { "epoch": 0.2, "learning_rate": 5.964661057635548e-05, "loss": 0.1572, "step": 1936 }, { "epoch": 0.2, "learning_rate": 5.964624423697603e-05, "loss": 1.6078, "step": 1937 }, { "epoch": 0.2, "learning_rate": 5.9645877708939175e-05, "loss": 1.2117, "step": 1938 }, { "epoch": 0.2, "learning_rate": 5.964551099224722e-05, "loss": 0.7846, "step": 1939 }, { "epoch": 0.2, "learning_rate": 5.9645144086902514e-05, "loss": 0.3061, "step": 1940 }, { "epoch": 0.2, "learning_rate": 5.964477699290739e-05, "loss": 0.4388, "step": 1941 }, { "epoch": 0.2, "learning_rate": 5.964440971026419e-05, "loss": 1.1041, "step": 1942 }, { "epoch": 0.2, "learning_rate": 5.964404223897524e-05, "loss": 0.9268, "step": 1943 }, { "epoch": 0.2, "learning_rate": 5.964367457904287e-05, "loss": 0.8486, "step": 1944 }, { "epoch": 0.2, "learning_rate": 5.9643306730469445e-05, "loss": 0.9467, "step": 1945 }, { "epoch": 0.2, "learning_rate": 5.964293869325729e-05, "loss": 0.2279, "step": 1946 }, { "epoch": 0.2, "learning_rate": 5.9642570467408754e-05, "loss": 1.0068, "step": 1947 }, { "epoch": 0.2, "learning_rate": 5.9642202052926164e-05, "loss": 1.4953, "step": 1948 }, { "epoch": 0.2, "learning_rate": 5.964183344981189e-05, "loss": 1.4955, "step": 1949 }, { "epoch": 0.2, "learning_rate": 5.9641464658068255e-05, "loss": 0.5641, "step": 1950 }, { "epoch": 0.2, "learning_rate": 5.9641095677697624e-05, "loss": 0.3844, "step": 1951 }, { "epoch": 0.2, "learning_rate": 5.964072650870233e-05, "loss": 1.1599, "step": 1952 }, { "epoch": 0.2, "learning_rate": 5.9640357151084726e-05, "loss": 0.6373, "step": 1953 }, { "epoch": 0.2, "learning_rate": 5.963998760484717e-05, "loss": 0.572, "step": 1954 }, { "epoch": 0.2, "learning_rate": 5.9639617869992e-05, "loss": 1.1169, "step": 1955 }, { "epoch": 0.2, "learning_rate": 5.963924794652159e-05, "loss": 1.1567, "step": 1956 }, { "epoch": 0.2, "learning_rate": 5.963887783443827e-05, "loss": 0.4087, "step": 1957 }, { "epoch": 0.2, "learning_rate": 5.963850753374442e-05, "loss": 0.4372, "step": 1958 }, { "epoch": 0.2, "learning_rate": 5.963813704444237e-05, "loss": 1.3178, "step": 1959 }, { "epoch": 0.2, "learning_rate": 5.9637766366534494e-05, "loss": 0.9917, "step": 1960 }, { "epoch": 0.2, "learning_rate": 5.963739550002315e-05, "loss": 1.1084, "step": 1961 }, { "epoch": 0.2, "learning_rate": 5.96370244449107e-05, "loss": 1.0593, "step": 1962 }, { "epoch": 0.2, "learning_rate": 5.963665320119949e-05, "loss": 1.485, "step": 1963 }, { "epoch": 0.2, "learning_rate": 5.96362817688919e-05, "loss": 0.5759, "step": 1964 }, { "epoch": 0.2, "learning_rate": 5.9635910147990286e-05, "loss": 0.2528, "step": 1965 }, { "epoch": 0.2, "learning_rate": 5.963553833849702e-05, "loss": 1.1895, "step": 1966 }, { "epoch": 0.2, "learning_rate": 5.963516634041445e-05, "loss": 1.013, "step": 1967 }, { "epoch": 0.2, "learning_rate": 5.963479415374497e-05, "loss": 0.8271, "step": 1968 }, { "epoch": 0.2, "learning_rate": 5.963442177849092e-05, "loss": 1.1133, "step": 1969 }, { "epoch": 0.2, "learning_rate": 5.963404921465469e-05, "loss": 1.2698, "step": 1970 }, { "epoch": 0.2, "learning_rate": 5.9633676462238645e-05, "loss": 0.35, "step": 1971 }, { "epoch": 0.2, "learning_rate": 5.9633303521245155e-05, "loss": 0.7272, "step": 1972 }, { "epoch": 0.2, "learning_rate": 5.963293039167659e-05, "loss": 1.0201, "step": 1973 }, { "epoch": 0.2, "learning_rate": 5.963255707353533e-05, "loss": 0.4115, "step": 1974 }, { "epoch": 0.2, "learning_rate": 5.963218356682376e-05, "loss": 1.143, "step": 1975 }, { "epoch": 0.2, "learning_rate": 5.9631809871544236e-05, "loss": 0.814, "step": 1976 }, { "epoch": 0.2, "learning_rate": 5.963143598769915e-05, "loss": 0.4163, "step": 1977 }, { "epoch": 0.2, "learning_rate": 5.963106191529088e-05, "loss": 0.5441, "step": 1978 }, { "epoch": 0.2, "learning_rate": 5.9630687654321794e-05, "loss": 0.9346, "step": 1979 }, { "epoch": 0.2, "learning_rate": 5.963031320479429e-05, "loss": 0.6711, "step": 1980 }, { "epoch": 0.2, "learning_rate": 5.962993856671075e-05, "loss": 0.9891, "step": 1981 }, { "epoch": 0.2, "learning_rate": 5.9629563740073545e-05, "loss": 0.5022, "step": 1982 }, { "epoch": 0.2, "learning_rate": 5.9629188724885066e-05, "loss": 1.4512, "step": 1983 }, { "epoch": 0.2, "learning_rate": 5.962881352114772e-05, "loss": 1.269, "step": 1984 }, { "epoch": 0.2, "learning_rate": 5.9628438128863856e-05, "loss": 0.6845, "step": 1985 }, { "epoch": 0.2, "learning_rate": 5.962806254803589e-05, "loss": 1.4855, "step": 1986 }, { "epoch": 0.2, "learning_rate": 5.96276867786662e-05, "loss": 0.6426, "step": 1987 }, { "epoch": 0.2, "learning_rate": 5.962731082075719e-05, "loss": 1.1786, "step": 1988 }, { "epoch": 0.2, "learning_rate": 5.962693467431124e-05, "loss": 0.6797, "step": 1989 }, { "epoch": 0.2, "learning_rate": 5.962655833933075e-05, "loss": 0.6506, "step": 1990 }, { "epoch": 0.2, "learning_rate": 5.962618181581811e-05, "loss": 0.6533, "step": 1991 }, { "epoch": 0.2, "learning_rate": 5.962580510377573e-05, "loss": 1.2443, "step": 1992 }, { "epoch": 0.2, "learning_rate": 5.9625428203205986e-05, "loss": 0.307, "step": 1993 }, { "epoch": 0.2, "learning_rate": 5.9625051114111295e-05, "loss": 1.0541, "step": 1994 }, { "epoch": 0.2, "learning_rate": 5.962467383649404e-05, "loss": 0.4458, "step": 1995 }, { "epoch": 0.2, "learning_rate": 5.962429637035663e-05, "loss": 0.3133, "step": 1996 }, { "epoch": 0.2, "learning_rate": 5.962391871570148e-05, "loss": 1.1313, "step": 1997 }, { "epoch": 0.2, "learning_rate": 5.9623540872530975e-05, "loss": 1.0685, "step": 1998 }, { "epoch": 0.2, "learning_rate": 5.9623162840847524e-05, "loss": 0.659, "step": 1999 }, { "epoch": 0.2, "learning_rate": 5.962278462065353e-05, "loss": 1.1681, "step": 2000 }, { "epoch": 0.2, "learning_rate": 5.96224062119514e-05, "loss": 0.7629, "step": 2001 }, { "epoch": 0.2, "learning_rate": 5.962202761474355e-05, "loss": 0.2156, "step": 2002 }, { "epoch": 0.2, "learning_rate": 5.962164882903239e-05, "loss": 0.4361, "step": 2003 }, { "epoch": 0.2, "learning_rate": 5.9621269854820324e-05, "loss": 0.4392, "step": 2004 }, { "epoch": 0.2, "learning_rate": 5.9620890692109764e-05, "loss": 0.6316, "step": 2005 }, { "epoch": 0.2, "learning_rate": 5.962051134090313e-05, "loss": 0.3045, "step": 2006 }, { "epoch": 0.2, "learning_rate": 5.962013180120282e-05, "loss": 0.9845, "step": 2007 }, { "epoch": 0.2, "learning_rate": 5.961975207301126e-05, "loss": 0.755, "step": 2008 }, { "epoch": 0.2, "learning_rate": 5.961937215633088e-05, "loss": 0.218, "step": 2009 }, { "epoch": 0.2, "learning_rate": 5.961899205116407e-05, "loss": 0.9749, "step": 2010 }, { "epoch": 0.2, "learning_rate": 5.961861175751326e-05, "loss": 1.2353, "step": 2011 }, { "epoch": 0.2, "learning_rate": 5.9618231275380876e-05, "loss": 1.1462, "step": 2012 }, { "epoch": 0.2, "learning_rate": 5.9617850604769334e-05, "loss": 0.3959, "step": 2013 }, { "epoch": 0.2, "learning_rate": 5.961746974568107e-05, "loss": 1.19, "step": 2014 }, { "epoch": 0.2, "learning_rate": 5.9617088698118486e-05, "loss": 1.2067, "step": 2015 }, { "epoch": 0.2, "learning_rate": 5.961670746208401e-05, "loss": 1.4503, "step": 2016 }, { "epoch": 0.2, "learning_rate": 5.9616326037580085e-05, "loss": 0.6058, "step": 2017 }, { "epoch": 0.2, "learning_rate": 5.9615944424609126e-05, "loss": 0.5747, "step": 2018 }, { "epoch": 0.21, "learning_rate": 5.961556262317356e-05, "loss": 0.259, "step": 2019 }, { "epoch": 0.21, "learning_rate": 5.961518063327582e-05, "loss": 0.9001, "step": 2020 }, { "epoch": 0.21, "learning_rate": 5.961479845491834e-05, "loss": 1.545, "step": 2021 }, { "epoch": 0.21, "learning_rate": 5.961441608810355e-05, "loss": 0.4146, "step": 2022 }, { "epoch": 0.21, "learning_rate": 5.9614033532833874e-05, "loss": 0.3089, "step": 2023 }, { "epoch": 0.21, "learning_rate": 5.9613650789111755e-05, "loss": 0.469, "step": 2024 }, { "epoch": 0.21, "learning_rate": 5.961326785693963e-05, "loss": 0.1887, "step": 2025 }, { "epoch": 0.21, "learning_rate": 5.961288473631994e-05, "loss": 0.691, "step": 2026 }, { "epoch": 0.21, "learning_rate": 5.961250142725511e-05, "loss": 0.7155, "step": 2027 }, { "epoch": 0.21, "learning_rate": 5.961211792974759e-05, "loss": 0.5071, "step": 2028 }, { "epoch": 0.21, "learning_rate": 5.961173424379981e-05, "loss": 0.4681, "step": 2029 }, { "epoch": 0.21, "learning_rate": 5.961135036941422e-05, "loss": 0.4226, "step": 2030 }, { "epoch": 0.21, "learning_rate": 5.9610966306593275e-05, "loss": 0.8446, "step": 2031 }, { "epoch": 0.21, "learning_rate": 5.961058205533939e-05, "loss": 1.4307, "step": 2032 }, { "epoch": 0.21, "learning_rate": 5.961019761565503e-05, "loss": 0.9168, "step": 2033 }, { "epoch": 0.21, "learning_rate": 5.9609812987542634e-05, "loss": 0.969, "step": 2034 }, { "epoch": 0.21, "learning_rate": 5.960942817100466e-05, "loss": 1.3414, "step": 2035 }, { "epoch": 0.21, "learning_rate": 5.960904316604354e-05, "loss": 1.0957, "step": 2036 }, { "epoch": 0.21, "learning_rate": 5.9608657972661736e-05, "loss": 0.8218, "step": 2037 }, { "epoch": 0.21, "learning_rate": 5.96082725908617e-05, "loss": 0.3387, "step": 2038 }, { "epoch": 0.21, "learning_rate": 5.960788702064588e-05, "loss": 0.9256, "step": 2039 }, { "epoch": 0.21, "learning_rate": 5.960750126201672e-05, "loss": 0.392, "step": 2040 }, { "epoch": 0.21, "learning_rate": 5.96071153149767e-05, "loss": 0.3964, "step": 2041 }, { "epoch": 0.21, "learning_rate": 5.960672917952825e-05, "loss": 0.7331, "step": 2042 }, { "epoch": 0.21, "learning_rate": 5.960634285567384e-05, "loss": 0.9969, "step": 2043 }, { "epoch": 0.21, "learning_rate": 5.960595634341594e-05, "loss": 1.0108, "step": 2044 }, { "epoch": 0.21, "learning_rate": 5.960556964275698e-05, "loss": 0.3684, "step": 2045 }, { "epoch": 0.21, "learning_rate": 5.960518275369944e-05, "loss": 1.0517, "step": 2046 }, { "epoch": 0.21, "learning_rate": 5.960479567624579e-05, "loss": 0.3014, "step": 2047 }, { "epoch": 0.21, "learning_rate": 5.960440841039847e-05, "loss": 0.4952, "step": 2048 }, { "epoch": 0.21, "learning_rate": 5.9604020956159965e-05, "loss": 0.4108, "step": 2049 }, { "epoch": 0.21, "learning_rate": 5.9603633313532725e-05, "loss": 0.6244, "step": 2050 }, { "epoch": 0.21, "learning_rate": 5.960324548251923e-05, "loss": 0.5555, "step": 2051 }, { "epoch": 0.21, "learning_rate": 5.9602857463121946e-05, "loss": 1.1432, "step": 2052 }, { "epoch": 0.21, "learning_rate": 5.960246925534333e-05, "loss": 0.9252, "step": 2053 }, { "epoch": 0.21, "learning_rate": 5.960208085918587e-05, "loss": 0.4488, "step": 2054 }, { "epoch": 0.21, "learning_rate": 5.9601692274652026e-05, "loss": 0.8799, "step": 2055 }, { "epoch": 0.21, "learning_rate": 5.9601303501744265e-05, "loss": 0.5533, "step": 2056 }, { "epoch": 0.21, "learning_rate": 5.9600914540465085e-05, "loss": 0.4137, "step": 2057 }, { "epoch": 0.21, "learning_rate": 5.960052539081693e-05, "loss": 0.7381, "step": 2058 }, { "epoch": 0.21, "learning_rate": 5.96001360528023e-05, "loss": 0.3563, "step": 2059 }, { "epoch": 0.21, "learning_rate": 5.959974652642366e-05, "loss": 0.9662, "step": 2060 }, { "epoch": 0.21, "learning_rate": 5.959935681168351e-05, "loss": 1.5496, "step": 2061 }, { "epoch": 0.21, "learning_rate": 5.9598966908584294e-05, "loss": 1.221, "step": 2062 }, { "epoch": 0.21, "learning_rate": 5.959857681712852e-05, "loss": 0.2996, "step": 2063 }, { "epoch": 0.21, "learning_rate": 5.959818653731866e-05, "loss": 0.3, "step": 2064 }, { "epoch": 0.21, "learning_rate": 5.95977960691572e-05, "loss": 2.2458, "step": 2065 }, { "epoch": 0.21, "learning_rate": 5.959740541264663e-05, "loss": 0.9467, "step": 2066 }, { "epoch": 0.21, "learning_rate": 5.9597014567789424e-05, "loss": 1.102, "step": 2067 }, { "epoch": 0.21, "learning_rate": 5.959662353458809e-05, "loss": 0.5661, "step": 2068 }, { "epoch": 0.21, "learning_rate": 5.9596232313045086e-05, "loss": 1.0639, "step": 2069 }, { "epoch": 0.21, "learning_rate": 5.9595840903162926e-05, "loss": 0.4242, "step": 2070 }, { "epoch": 0.21, "learning_rate": 5.95954493049441e-05, "loss": 1.2749, "step": 2071 }, { "epoch": 0.21, "learning_rate": 5.959505751839109e-05, "loss": 0.7759, "step": 2072 }, { "epoch": 0.21, "learning_rate": 5.959466554350639e-05, "loss": 1.6358, "step": 2073 }, { "epoch": 0.21, "learning_rate": 5.95942733802925e-05, "loss": 1.2132, "step": 2074 }, { "epoch": 0.21, "learning_rate": 5.95938810287519e-05, "loss": 0.5293, "step": 2075 }, { "epoch": 0.21, "learning_rate": 5.9593488488887114e-05, "loss": 1.5332, "step": 2076 }, { "epoch": 0.21, "learning_rate": 5.959309576070062e-05, "loss": 0.5561, "step": 2077 }, { "epoch": 0.21, "learning_rate": 5.959270284419492e-05, "loss": 0.4297, "step": 2078 }, { "epoch": 0.21, "learning_rate": 5.9592309739372525e-05, "loss": 0.4421, "step": 2079 }, { "epoch": 0.21, "learning_rate": 5.959191644623592e-05, "loss": 1.2804, "step": 2080 }, { "epoch": 0.21, "learning_rate": 5.9591522964787624e-05, "loss": 0.3882, "step": 2081 }, { "epoch": 0.21, "learning_rate": 5.959112929503012e-05, "loss": 0.8107, "step": 2082 }, { "epoch": 0.21, "learning_rate": 5.9590735436965935e-05, "loss": 0.2377, "step": 2083 }, { "epoch": 0.21, "learning_rate": 5.9590341390597565e-05, "loss": 1.4244, "step": 2084 }, { "epoch": 0.21, "learning_rate": 5.958994715592752e-05, "loss": 0.3204, "step": 2085 }, { "epoch": 0.21, "learning_rate": 5.958955273295831e-05, "loss": 0.2708, "step": 2086 }, { "epoch": 0.21, "learning_rate": 5.958915812169244e-05, "loss": 0.3905, "step": 2087 }, { "epoch": 0.21, "learning_rate": 5.9588763322132423e-05, "loss": 0.3961, "step": 2088 }, { "epoch": 0.21, "learning_rate": 5.958836833428078e-05, "loss": 0.7883, "step": 2089 }, { "epoch": 0.21, "learning_rate": 5.9587973158140007e-05, "loss": 0.7464, "step": 2090 }, { "epoch": 0.21, "learning_rate": 5.9587577793712625e-05, "loss": 0.4365, "step": 2091 }, { "epoch": 0.21, "learning_rate": 5.9587182241001163e-05, "loss": 1.231, "step": 2092 }, { "epoch": 0.21, "learning_rate": 5.958678650000813e-05, "loss": 1.4578, "step": 2093 }, { "epoch": 0.21, "learning_rate": 5.958639057073604e-05, "loss": 1.534, "step": 2094 }, { "epoch": 0.21, "learning_rate": 5.958599445318741e-05, "loss": 1.3917, "step": 2095 }, { "epoch": 0.21, "learning_rate": 5.9585598147364775e-05, "loss": 0.2313, "step": 2096 }, { "epoch": 0.21, "learning_rate": 5.958520165327064e-05, "loss": 0.8112, "step": 2097 }, { "epoch": 0.21, "learning_rate": 5.958480497090755e-05, "loss": 0.503, "step": 2098 }, { "epoch": 0.21, "learning_rate": 5.958440810027801e-05, "loss": 0.688, "step": 2099 }, { "epoch": 0.21, "learning_rate": 5.958401104138455e-05, "loss": 0.9918, "step": 2100 }, { "epoch": 0.21, "learning_rate": 5.958361379422969e-05, "loss": 1.2182, "step": 2101 }, { "epoch": 0.21, "learning_rate": 5.958321635881597e-05, "loss": 0.6189, "step": 2102 }, { "epoch": 0.21, "learning_rate": 5.958281873514592e-05, "loss": 0.5386, "step": 2103 }, { "epoch": 0.21, "learning_rate": 5.9582420923222065e-05, "loss": 0.7895, "step": 2104 }, { "epoch": 0.21, "learning_rate": 5.958202292304694e-05, "loss": 0.3653, "step": 2105 }, { "epoch": 0.21, "learning_rate": 5.958162473462306e-05, "loss": 0.6737, "step": 2106 }, { "epoch": 0.21, "learning_rate": 5.958122635795299e-05, "loss": 1.5867, "step": 2107 }, { "epoch": 0.21, "learning_rate": 5.958082779303924e-05, "loss": 0.4472, "step": 2108 }, { "epoch": 0.21, "learning_rate": 5.958042903988436e-05, "loss": 2.1094, "step": 2109 }, { "epoch": 0.21, "learning_rate": 5.958003009849088e-05, "loss": 0.1871, "step": 2110 }, { "epoch": 0.21, "learning_rate": 5.9579630968861334e-05, "loss": 1.2405, "step": 2111 }, { "epoch": 0.21, "learning_rate": 5.9579231650998276e-05, "loss": 0.8308, "step": 2112 }, { "epoch": 0.21, "learning_rate": 5.9578832144904244e-05, "loss": 0.6184, "step": 2113 }, { "epoch": 0.21, "learning_rate": 5.957843245058178e-05, "loss": 1.3217, "step": 2114 }, { "epoch": 0.21, "learning_rate": 5.957803256803341e-05, "loss": 0.3041, "step": 2115 }, { "epoch": 0.21, "learning_rate": 5.95776324972617e-05, "loss": 0.7624, "step": 2116 }, { "epoch": 0.21, "learning_rate": 5.957723223826919e-05, "loss": 0.8077, "step": 2117 }, { "epoch": 0.22, "learning_rate": 5.957683179105843e-05, "loss": 0.898, "step": 2118 }, { "epoch": 0.22, "learning_rate": 5.957643115563195e-05, "loss": 1.7453, "step": 2119 }, { "epoch": 0.22, "learning_rate": 5.957603033199232e-05, "loss": 0.5137, "step": 2120 }, { "epoch": 0.22, "learning_rate": 5.957562932014209e-05, "loss": 0.3131, "step": 2121 }, { "epoch": 0.22, "learning_rate": 5.957522812008381e-05, "loss": 0.2675, "step": 2122 }, { "epoch": 0.22, "learning_rate": 5.9574826731820016e-05, "loss": 0.315, "step": 2123 }, { "epoch": 0.22, "learning_rate": 5.957442515535329e-05, "loss": 1.3645, "step": 2124 }, { "epoch": 0.22, "learning_rate": 5.957402339068616e-05, "loss": 0.8875, "step": 2125 }, { "epoch": 0.22, "learning_rate": 5.95736214378212e-05, "loss": 0.5362, "step": 2126 }, { "epoch": 0.22, "learning_rate": 5.957321929676097e-05, "loss": 0.8996, "step": 2127 }, { "epoch": 0.22, "learning_rate": 5.957281696750801e-05, "loss": 0.3836, "step": 2128 }, { "epoch": 0.22, "learning_rate": 5.9572414450064896e-05, "loss": 0.3687, "step": 2129 }, { "epoch": 0.22, "learning_rate": 5.957201174443419e-05, "loss": 1.0334, "step": 2130 }, { "epoch": 0.22, "learning_rate": 5.957160885061845e-05, "loss": 1.0832, "step": 2131 }, { "epoch": 0.22, "learning_rate": 5.957120576862024e-05, "loss": 0.1588, "step": 2132 }, { "epoch": 0.22, "learning_rate": 5.9570802498442136e-05, "loss": 0.5114, "step": 2133 }, { "epoch": 0.22, "learning_rate": 5.957039904008667e-05, "loss": 1.0364, "step": 2134 }, { "epoch": 0.22, "learning_rate": 5.9569995393556464e-05, "loss": 0.6059, "step": 2135 }, { "epoch": 0.22, "learning_rate": 5.956959155885404e-05, "loss": 0.9489, "step": 2136 }, { "epoch": 0.22, "learning_rate": 5.956918753598199e-05, "loss": 0.9418, "step": 2137 }, { "epoch": 0.22, "learning_rate": 5.9568783324942863e-05, "loss": 0.7637, "step": 2138 }, { "epoch": 0.22, "learning_rate": 5.956837892573926e-05, "loss": 0.6017, "step": 2139 }, { "epoch": 0.22, "learning_rate": 5.9567974338373743e-05, "loss": 0.9605, "step": 2140 }, { "epoch": 0.22, "learning_rate": 5.9567569562848884e-05, "loss": 1.0633, "step": 2141 }, { "epoch": 0.22, "learning_rate": 5.956716459916726e-05, "loss": 0.4589, "step": 2142 }, { "epoch": 0.22, "learning_rate": 5.956675944733145e-05, "loss": 1.3311, "step": 2143 }, { "epoch": 0.22, "learning_rate": 5.956635410734403e-05, "loss": 0.4576, "step": 2144 }, { "epoch": 0.22, "learning_rate": 5.9565948579207583e-05, "loss": 1.1327, "step": 2145 }, { "epoch": 0.22, "learning_rate": 5.956554286292468e-05, "loss": 1.196, "step": 2146 }, { "epoch": 0.22, "learning_rate": 5.956513695849791e-05, "loss": 1.9873, "step": 2147 }, { "epoch": 0.22, "learning_rate": 5.9564730865929864e-05, "loss": 1.2422, "step": 2148 }, { "epoch": 0.22, "learning_rate": 5.956432458522311e-05, "loss": 0.9793, "step": 2149 }, { "epoch": 0.22, "learning_rate": 5.956391811638024e-05, "loss": 0.5048, "step": 2150 }, { "epoch": 0.22, "learning_rate": 5.956351145940385e-05, "loss": 1.1631, "step": 2151 }, { "epoch": 0.22, "learning_rate": 5.9563104614296506e-05, "loss": 0.8936, "step": 2152 }, { "epoch": 0.22, "learning_rate": 5.9562697581060824e-05, "loss": 0.3804, "step": 2153 }, { "epoch": 0.22, "learning_rate": 5.9562290359699375e-05, "loss": 0.4076, "step": 2154 }, { "epoch": 0.22, "learning_rate": 5.956188295021476e-05, "loss": 1.1775, "step": 2155 }, { "epoch": 0.22, "learning_rate": 5.9561475352609564e-05, "loss": 1.2598, "step": 2156 }, { "epoch": 0.22, "learning_rate": 5.956106756688638e-05, "loss": 0.3611, "step": 2157 }, { "epoch": 0.22, "learning_rate": 5.9560659593047816e-05, "loss": 0.3868, "step": 2158 }, { "epoch": 0.22, "learning_rate": 5.956025143109646e-05, "loss": 0.608, "step": 2159 }, { "epoch": 0.22, "learning_rate": 5.9559843081034906e-05, "loss": 0.7251, "step": 2160 }, { "epoch": 0.22, "learning_rate": 5.955943454286576e-05, "loss": 0.8895, "step": 2161 }, { "epoch": 0.22, "learning_rate": 5.955902581659161e-05, "loss": 1.3419, "step": 2162 }, { "epoch": 0.22, "learning_rate": 5.955861690221507e-05, "loss": 0.4093, "step": 2163 }, { "epoch": 0.22, "learning_rate": 5.9558207799738735e-05, "loss": 1.1101, "step": 2164 }, { "epoch": 0.22, "learning_rate": 5.955779850916521e-05, "loss": 0.35, "step": 2165 }, { "epoch": 0.22, "learning_rate": 5.95573890304971e-05, "loss": 0.7798, "step": 2166 }, { "epoch": 0.22, "learning_rate": 5.955697936373701e-05, "loss": 0.3641, "step": 2167 }, { "epoch": 0.22, "learning_rate": 5.955656950888755e-05, "loss": 1.0015, "step": 2168 }, { "epoch": 0.22, "learning_rate": 5.955615946595132e-05, "loss": 1.0058, "step": 2169 }, { "epoch": 0.22, "learning_rate": 5.955574923493094e-05, "loss": 1.3674, "step": 2170 }, { "epoch": 0.22, "learning_rate": 5.955533881582901e-05, "loss": 0.8125, "step": 2171 }, { "epoch": 0.22, "learning_rate": 5.955492820864815e-05, "loss": 0.9914, "step": 2172 }, { "epoch": 0.22, "learning_rate": 5.955451741339097e-05, "loss": 0.8793, "step": 2173 }, { "epoch": 0.22, "learning_rate": 5.9554106430060084e-05, "loss": 1.4865, "step": 2174 }, { "epoch": 0.22, "learning_rate": 5.955369525865811e-05, "loss": 0.4402, "step": 2175 }, { "epoch": 0.22, "learning_rate": 5.955328389918766e-05, "loss": 0.4447, "step": 2176 }, { "epoch": 0.22, "learning_rate": 5.955287235165135e-05, "loss": 0.3639, "step": 2177 }, { "epoch": 0.22, "learning_rate": 5.955246061605181e-05, "loss": 0.7062, "step": 2178 }, { "epoch": 0.22, "learning_rate": 5.955204869239165e-05, "loss": 1.1076, "step": 2179 }, { "epoch": 0.22, "learning_rate": 5.9551636580673496e-05, "loss": 1.1117, "step": 2180 }, { "epoch": 0.22, "learning_rate": 5.955122428089997e-05, "loss": 0.347, "step": 2181 }, { "epoch": 0.22, "learning_rate": 5.955081179307369e-05, "loss": 1.0347, "step": 2182 }, { "epoch": 0.22, "learning_rate": 5.955039911719729e-05, "loss": 1.5676, "step": 2183 }, { "epoch": 0.22, "learning_rate": 5.954998625327339e-05, "loss": 0.4467, "step": 2184 }, { "epoch": 0.22, "learning_rate": 5.954957320130461e-05, "loss": 0.469, "step": 2185 }, { "epoch": 0.22, "learning_rate": 5.95491599612936e-05, "loss": 0.6529, "step": 2186 }, { "epoch": 0.22, "learning_rate": 5.954874653324297e-05, "loss": 0.0643, "step": 2187 }, { "epoch": 0.22, "learning_rate": 5.9548332917155364e-05, "loss": 0.8411, "step": 2188 }, { "epoch": 0.22, "learning_rate": 5.954791911303341e-05, "loss": 0.7156, "step": 2189 }, { "epoch": 0.22, "learning_rate": 5.9547505120879725e-05, "loss": 0.3546, "step": 2190 }, { "epoch": 0.22, "learning_rate": 5.9547090940696973e-05, "loss": 1.4946, "step": 2191 }, { "epoch": 0.22, "learning_rate": 5.954667657248777e-05, "loss": 0.6065, "step": 2192 }, { "epoch": 0.22, "learning_rate": 5.954626201625475e-05, "loss": 0.7337, "step": 2193 }, { "epoch": 0.22, "learning_rate": 5.9545847272000565e-05, "loss": 0.324, "step": 2194 }, { "epoch": 0.22, "learning_rate": 5.954543233972785e-05, "loss": 0.9723, "step": 2195 }, { "epoch": 0.22, "learning_rate": 5.954501721943924e-05, "loss": 0.8342, "step": 2196 }, { "epoch": 0.22, "learning_rate": 5.954460191113738e-05, "loss": 1.2135, "step": 2197 }, { "epoch": 0.22, "learning_rate": 5.9544186414824906e-05, "loss": 1.2936, "step": 2198 }, { "epoch": 0.22, "learning_rate": 5.954377073050448e-05, "loss": 0.183, "step": 2199 }, { "epoch": 0.22, "learning_rate": 5.954335485817873e-05, "loss": 1.1609, "step": 2200 }, { "epoch": 0.22, "learning_rate": 5.954293879785031e-05, "loss": 0.9869, "step": 2201 }, { "epoch": 0.22, "learning_rate": 5.954252254952186e-05, "loss": 0.9749, "step": 2202 }, { "epoch": 0.22, "learning_rate": 5.954210611319606e-05, "loss": 1.0105, "step": 2203 }, { "epoch": 0.22, "learning_rate": 5.9541689488875504e-05, "loss": 1.124, "step": 2204 }, { "epoch": 0.22, "learning_rate": 5.954127267656289e-05, "loss": 0.9871, "step": 2205 }, { "epoch": 0.22, "learning_rate": 5.9540855676260854e-05, "loss": 0.9169, "step": 2206 }, { "epoch": 0.22, "learning_rate": 5.954043848797205e-05, "loss": 1.1402, "step": 2207 }, { "epoch": 0.22, "learning_rate": 5.954002111169913e-05, "loss": 0.9834, "step": 2208 }, { "epoch": 0.22, "learning_rate": 5.953960354744475e-05, "loss": 0.2368, "step": 2209 }, { "epoch": 0.22, "learning_rate": 5.953918579521158e-05, "loss": 0.2871, "step": 2210 }, { "epoch": 0.22, "learning_rate": 5.953876785500226e-05, "loss": 1.3669, "step": 2211 }, { "epoch": 0.22, "learning_rate": 5.953834972681946e-05, "loss": 1.3795, "step": 2212 }, { "epoch": 0.22, "learning_rate": 5.953793141066585e-05, "loss": 0.9443, "step": 2213 }, { "epoch": 0.22, "learning_rate": 5.953751290654407e-05, "loss": 0.7215, "step": 2214 }, { "epoch": 0.22, "learning_rate": 5.9537094214456796e-05, "loss": 0.8953, "step": 2215 }, { "epoch": 0.23, "learning_rate": 5.953667533440669e-05, "loss": 0.3258, "step": 2216 }, { "epoch": 0.23, "learning_rate": 5.953625626639642e-05, "loss": 0.7486, "step": 2217 }, { "epoch": 0.23, "learning_rate": 5.953583701042865e-05, "loss": 0.3926, "step": 2218 }, { "epoch": 0.23, "learning_rate": 5.9535417566506056e-05, "loss": 1.6139, "step": 2219 }, { "epoch": 0.23, "learning_rate": 5.9534997934631296e-05, "loss": 0.2754, "step": 2220 }, { "epoch": 0.23, "learning_rate": 5.9534578114807046e-05, "loss": 0.4352, "step": 2221 }, { "epoch": 0.23, "learning_rate": 5.953415810703598e-05, "loss": 0.2472, "step": 2222 }, { "epoch": 0.23, "learning_rate": 5.953373791132076e-05, "loss": 0.49, "step": 2223 }, { "epoch": 0.23, "learning_rate": 5.953331752766407e-05, "loss": 1.022, "step": 2224 }, { "epoch": 0.23, "learning_rate": 5.953289695606858e-05, "loss": 1.6402, "step": 2225 }, { "epoch": 0.23, "learning_rate": 5.953247619653697e-05, "loss": 0.7377, "step": 2226 }, { "epoch": 0.23, "learning_rate": 5.953205524907192e-05, "loss": 0.4341, "step": 2227 }, { "epoch": 0.23, "learning_rate": 5.9531634113676095e-05, "loss": 1.3683, "step": 2228 }, { "epoch": 0.23, "learning_rate": 5.9531212790352196e-05, "loss": 0.8495, "step": 2229 }, { "epoch": 0.23, "learning_rate": 5.953079127910289e-05, "loss": 1.0899, "step": 2230 }, { "epoch": 0.23, "learning_rate": 5.953036957993086e-05, "loss": 1.1959, "step": 2231 }, { "epoch": 0.23, "learning_rate": 5.952994769283879e-05, "loss": 0.4037, "step": 2232 }, { "epoch": 0.23, "learning_rate": 5.952952561782938e-05, "loss": 1.4337, "step": 2233 }, { "epoch": 0.23, "learning_rate": 5.952910335490528e-05, "loss": 1.2475, "step": 2234 }, { "epoch": 0.23, "learning_rate": 5.9528680904069215e-05, "loss": 1.426, "step": 2235 }, { "epoch": 0.23, "learning_rate": 5.952825826532385e-05, "loss": 1.3955, "step": 2236 }, { "epoch": 0.23, "learning_rate": 5.952783543867189e-05, "loss": 0.7031, "step": 2237 }, { "epoch": 0.23, "learning_rate": 5.952741242411602e-05, "loss": 0.9248, "step": 2238 }, { "epoch": 0.23, "learning_rate": 5.952698922165892e-05, "loss": 0.4671, "step": 2239 }, { "epoch": 0.23, "learning_rate": 5.9526565831303306e-05, "loss": 0.9758, "step": 2240 }, { "epoch": 0.23, "learning_rate": 5.9526142253051846e-05, "loss": 1.3609, "step": 2241 }, { "epoch": 0.23, "learning_rate": 5.9525718486907255e-05, "loss": 0.5554, "step": 2242 }, { "epoch": 0.23, "learning_rate": 5.952529453287223e-05, "loss": 0.4166, "step": 2243 }, { "epoch": 0.23, "learning_rate": 5.9524870390949454e-05, "loss": 1.0726, "step": 2244 }, { "epoch": 0.23, "learning_rate": 5.952444606114164e-05, "loss": 0.9166, "step": 2245 }, { "epoch": 0.23, "learning_rate": 5.952402154345148e-05, "loss": 0.9977, "step": 2246 }, { "epoch": 0.23, "learning_rate": 5.952359683788168e-05, "loss": 0.6456, "step": 2247 }, { "epoch": 0.23, "learning_rate": 5.952317194443494e-05, "loss": 0.3907, "step": 2248 }, { "epoch": 0.23, "learning_rate": 5.952274686311397e-05, "loss": 0.542, "step": 2249 }, { "epoch": 0.23, "learning_rate": 5.952232159392145e-05, "loss": 1.4145, "step": 2250 }, { "epoch": 0.23, "learning_rate": 5.952189613686013e-05, "loss": 0.7416, "step": 2251 }, { "epoch": 0.23, "learning_rate": 5.952147049193269e-05, "loss": 1.193, "step": 2252 }, { "epoch": 0.23, "learning_rate": 5.952104465914183e-05, "loss": 0.2797, "step": 2253 }, { "epoch": 0.23, "learning_rate": 5.9520618638490274e-05, "loss": 0.8197, "step": 2254 }, { "epoch": 0.23, "learning_rate": 5.952019242998074e-05, "loss": 0.4167, "step": 2255 }, { "epoch": 0.23, "learning_rate": 5.951976603361593e-05, "loss": 0.9374, "step": 2256 }, { "epoch": 0.23, "learning_rate": 5.951933944939856e-05, "loss": 1.0152, "step": 2257 }, { "epoch": 0.23, "learning_rate": 5.9518912677331336e-05, "loss": 0.8476, "step": 2258 }, { "epoch": 0.23, "learning_rate": 5.9518485717416986e-05, "loss": 0.7759, "step": 2259 }, { "epoch": 0.23, "learning_rate": 5.9518058569658226e-05, "loss": 0.8008, "step": 2260 }, { "epoch": 0.23, "learning_rate": 5.951763123405777e-05, "loss": 0.2715, "step": 2261 }, { "epoch": 0.23, "learning_rate": 5.9517203710618327e-05, "loss": 0.5598, "step": 2262 }, { "epoch": 0.23, "learning_rate": 5.951677599934264e-05, "loss": 0.6027, "step": 2263 }, { "epoch": 0.23, "learning_rate": 5.9516348100233416e-05, "loss": 0.4162, "step": 2264 }, { "epoch": 0.23, "learning_rate": 5.9515920013293374e-05, "loss": 0.3498, "step": 2265 }, { "epoch": 0.23, "learning_rate": 5.9515491738525254e-05, "loss": 1.4992, "step": 2266 }, { "epoch": 0.23, "learning_rate": 5.951506327593177e-05, "loss": 0.8462, "step": 2267 }, { "epoch": 0.23, "learning_rate": 5.951463462551565e-05, "loss": 0.865, "step": 2268 }, { "epoch": 0.23, "learning_rate": 5.951420578727963e-05, "loss": 1.3235, "step": 2269 }, { "epoch": 0.23, "learning_rate": 5.951377676122643e-05, "loss": 0.7111, "step": 2270 }, { "epoch": 0.23, "learning_rate": 5.951334754735878e-05, "loss": 0.7908, "step": 2271 }, { "epoch": 0.23, "learning_rate": 5.951291814567941e-05, "loss": 0.8306, "step": 2272 }, { "epoch": 0.23, "learning_rate": 5.9512488556191056e-05, "loss": 0.3779, "step": 2273 }, { "epoch": 0.23, "learning_rate": 5.951205877889646e-05, "loss": 0.9137, "step": 2274 }, { "epoch": 0.23, "learning_rate": 5.951162881379835e-05, "loss": 0.7706, "step": 2275 }, { "epoch": 0.23, "learning_rate": 5.9511198660899444e-05, "loss": 1.3054, "step": 2276 }, { "epoch": 0.23, "learning_rate": 5.9510768320202515e-05, "loss": 0.6213, "step": 2277 }, { "epoch": 0.23, "learning_rate": 5.951033779171027e-05, "loss": 0.7545, "step": 2278 }, { "epoch": 0.23, "learning_rate": 5.9509907075425475e-05, "loss": 0.7413, "step": 2279 }, { "epoch": 0.23, "learning_rate": 5.9509476171350846e-05, "loss": 0.5276, "step": 2280 }, { "epoch": 0.23, "learning_rate": 5.9509045079489135e-05, "loss": 0.9164, "step": 2281 }, { "epoch": 0.23, "learning_rate": 5.95086137998431e-05, "loss": 0.6579, "step": 2282 }, { "epoch": 0.23, "learning_rate": 5.950818233241546e-05, "loss": 0.6903, "step": 2283 }, { "epoch": 0.23, "learning_rate": 5.950775067720898e-05, "loss": 0.8204, "step": 2284 }, { "epoch": 0.23, "learning_rate": 5.950731883422639e-05, "loss": 0.9505, "step": 2285 }, { "epoch": 0.23, "learning_rate": 5.950688680347046e-05, "loss": 1.2598, "step": 2286 }, { "epoch": 0.23, "learning_rate": 5.950645458494393e-05, "loss": 0.5954, "step": 2287 }, { "epoch": 0.23, "learning_rate": 5.950602217864954e-05, "loss": 0.8422, "step": 2288 }, { "epoch": 0.23, "learning_rate": 5.950558958459005e-05, "loss": 1.491, "step": 2289 }, { "epoch": 0.23, "learning_rate": 5.950515680276821e-05, "loss": 1.3455, "step": 2290 }, { "epoch": 0.23, "learning_rate": 5.950472383318678e-05, "loss": 1.0867, "step": 2291 }, { "epoch": 0.23, "learning_rate": 5.9504290675848515e-05, "loss": 1.3041, "step": 2292 }, { "epoch": 0.23, "learning_rate": 5.950385733075616e-05, "loss": 0.6913, "step": 2293 }, { "epoch": 0.23, "learning_rate": 5.950342379791249e-05, "loss": 1.1731, "step": 2294 }, { "epoch": 0.23, "learning_rate": 5.950299007732025e-05, "loss": 0.984, "step": 2295 }, { "epoch": 0.23, "learning_rate": 5.950255616898221e-05, "loss": 1.2216, "step": 2296 }, { "epoch": 0.23, "learning_rate": 5.950212207290111e-05, "loss": 1.2232, "step": 2297 }, { "epoch": 0.23, "learning_rate": 5.950168778907974e-05, "loss": 0.0649, "step": 2298 }, { "epoch": 0.23, "learning_rate": 5.950125331752085e-05, "loss": 0.5823, "step": 2299 }, { "epoch": 0.23, "learning_rate": 5.950081865822721e-05, "loss": 0.8986, "step": 2300 }, { "epoch": 0.23, "learning_rate": 5.950038381120159e-05, "loss": 1.0603, "step": 2301 }, { "epoch": 0.23, "learning_rate": 5.949994877644673e-05, "loss": 0.9665, "step": 2302 }, { "epoch": 0.23, "learning_rate": 5.9499513553965433e-05, "loss": 0.3427, "step": 2303 }, { "epoch": 0.23, "learning_rate": 5.9499078143760445e-05, "loss": 0.5818, "step": 2304 }, { "epoch": 0.23, "learning_rate": 5.949864254583455e-05, "loss": 1.01, "step": 2305 }, { "epoch": 0.23, "learning_rate": 5.949820676019052e-05, "loss": 0.4451, "step": 2306 }, { "epoch": 0.23, "learning_rate": 5.949777078683112e-05, "loss": 0.2554, "step": 2307 }, { "epoch": 0.23, "learning_rate": 5.9497334625759125e-05, "loss": 0.9999, "step": 2308 }, { "epoch": 0.23, "learning_rate": 5.949689827697731e-05, "loss": 0.7491, "step": 2309 }, { "epoch": 0.23, "learning_rate": 5.949646174048847e-05, "loss": 0.4519, "step": 2310 }, { "epoch": 0.23, "learning_rate": 5.949602501629536e-05, "loss": 1.385, "step": 2311 }, { "epoch": 0.23, "learning_rate": 5.949558810440076e-05, "loss": 1.1231, "step": 2312 }, { "epoch": 0.23, "learning_rate": 5.9495151004807466e-05, "loss": 0.8794, "step": 2313 }, { "epoch": 0.23, "learning_rate": 5.949471371751826e-05, "loss": 0.6092, "step": 2314 }, { "epoch": 0.24, "learning_rate": 5.94942762425359e-05, "loss": 1.1576, "step": 2315 }, { "epoch": 0.24, "learning_rate": 5.94938385798632e-05, "loss": 0.4212, "step": 2316 }, { "epoch": 0.24, "learning_rate": 5.9493400729502927e-05, "loss": 1.2769, "step": 2317 }, { "epoch": 0.24, "learning_rate": 5.949296269145787e-05, "loss": 1.2863, "step": 2318 }, { "epoch": 0.24, "learning_rate": 5.949252446573082e-05, "loss": 0.9786, "step": 2319 }, { "epoch": 0.24, "learning_rate": 5.949208605232457e-05, "loss": 0.9793, "step": 2320 }, { "epoch": 0.24, "learning_rate": 5.949164745124188e-05, "loss": 1.0596, "step": 2321 }, { "epoch": 0.24, "learning_rate": 5.949120866248559e-05, "loss": 1.2119, "step": 2322 }, { "epoch": 0.24, "learning_rate": 5.9490769686058454e-05, "loss": 0.7602, "step": 2323 }, { "epoch": 0.24, "learning_rate": 5.949033052196328e-05, "loss": 0.5408, "step": 2324 }, { "epoch": 0.24, "learning_rate": 5.948989117020287e-05, "loss": 0.3976, "step": 2325 }, { "epoch": 0.24, "learning_rate": 5.948945163078e-05, "loss": 0.3997, "step": 2326 }, { "epoch": 0.24, "learning_rate": 5.948901190369749e-05, "loss": 0.5914, "step": 2327 }, { "epoch": 0.24, "learning_rate": 5.9488571988958114e-05, "loss": 0.7604, "step": 2328 }, { "epoch": 0.24, "learning_rate": 5.948813188656469e-05, "loss": 0.2832, "step": 2329 }, { "epoch": 0.24, "learning_rate": 5.948769159652001e-05, "loss": 0.3986, "step": 2330 }, { "epoch": 0.24, "learning_rate": 5.9487251118826886e-05, "loss": 1.3259, "step": 2331 }, { "epoch": 0.24, "learning_rate": 5.9486810453488106e-05, "loss": 0.56, "step": 2332 }, { "epoch": 0.24, "learning_rate": 5.9486369600506483e-05, "loss": 0.5352, "step": 2333 }, { "epoch": 0.24, "learning_rate": 5.948592855988482e-05, "loss": 0.4722, "step": 2334 }, { "epoch": 0.24, "learning_rate": 5.9485487331625924e-05, "loss": 0.2505, "step": 2335 }, { "epoch": 0.24, "learning_rate": 5.9485045915732605e-05, "loss": 1.2138, "step": 2336 }, { "epoch": 0.24, "learning_rate": 5.948460431220767e-05, "loss": 0.8894, "step": 2337 }, { "epoch": 0.24, "learning_rate": 5.9484162521053935e-05, "loss": 1.1538, "step": 2338 }, { "epoch": 0.24, "learning_rate": 5.94837205422742e-05, "loss": 1.9655, "step": 2339 }, { "epoch": 0.24, "learning_rate": 5.9483278375871284e-05, "loss": 0.4408, "step": 2340 }, { "epoch": 0.24, "learning_rate": 5.9482836021848006e-05, "loss": 1.1423, "step": 2341 }, { "epoch": 0.24, "learning_rate": 5.948239348020717e-05, "loss": 0.0983, "step": 2342 }, { "epoch": 0.24, "learning_rate": 5.94819507509516e-05, "loss": 0.5239, "step": 2343 }, { "epoch": 0.24, "learning_rate": 5.9481507834084106e-05, "loss": 0.4176, "step": 2344 }, { "epoch": 0.24, "learning_rate": 5.948106472960751e-05, "loss": 1.2511, "step": 2345 }, { "epoch": 0.24, "learning_rate": 5.948062143752464e-05, "loss": 0.5899, "step": 2346 }, { "epoch": 0.24, "learning_rate": 5.948017795783831e-05, "loss": 0.5648, "step": 2347 }, { "epoch": 0.24, "learning_rate": 5.947973429055135e-05, "loss": 1.3003, "step": 2348 }, { "epoch": 0.24, "learning_rate": 5.947929043566656e-05, "loss": 0.5513, "step": 2349 }, { "epoch": 0.24, "learning_rate": 5.947884639318679e-05, "loss": 0.4099, "step": 2350 }, { "epoch": 0.24, "learning_rate": 5.947840216311486e-05, "loss": 0.6228, "step": 2351 }, { "epoch": 0.24, "learning_rate": 5.9477957745453576e-05, "loss": 0.9686, "step": 2352 }, { "epoch": 0.24, "learning_rate": 5.947751314020579e-05, "loss": 0.43, "step": 2353 }, { "epoch": 0.24, "learning_rate": 5.947706834737433e-05, "loss": 0.7032, "step": 2354 }, { "epoch": 0.24, "learning_rate": 5.947662336696202e-05, "loss": 0.6318, "step": 2355 }, { "epoch": 0.24, "learning_rate": 5.947617819897168e-05, "loss": 1.0296, "step": 2356 }, { "epoch": 0.24, "learning_rate": 5.9475732843406174e-05, "loss": 0.9949, "step": 2357 }, { "epoch": 0.24, "learning_rate": 5.947528730026831e-05, "loss": 0.3617, "step": 2358 }, { "epoch": 0.24, "learning_rate": 5.9474841569560926e-05, "loss": 1.3906, "step": 2359 }, { "epoch": 0.24, "learning_rate": 5.947439565128686e-05, "loss": 1.2923, "step": 2360 }, { "epoch": 0.24, "learning_rate": 5.947394954544896e-05, "loss": 0.9679, "step": 2361 }, { "epoch": 0.24, "learning_rate": 5.9473503252050056e-05, "loss": 1.1516, "step": 2362 }, { "epoch": 0.24, "learning_rate": 5.947305677109299e-05, "loss": 0.5929, "step": 2363 }, { "epoch": 0.24, "learning_rate": 5.947261010258061e-05, "loss": 0.644, "step": 2364 }, { "epoch": 0.24, "learning_rate": 5.947216324651574e-05, "loss": 0.5442, "step": 2365 }, { "epoch": 0.24, "learning_rate": 5.947171620290124e-05, "loss": 0.3594, "step": 2366 }, { "epoch": 0.24, "learning_rate": 5.947126897173995e-05, "loss": 0.4067, "step": 2367 }, { "epoch": 0.24, "learning_rate": 5.947082155303471e-05, "loss": 0.4321, "step": 2368 }, { "epoch": 0.24, "learning_rate": 5.9470373946788375e-05, "loss": 0.2963, "step": 2369 }, { "epoch": 0.24, "learning_rate": 5.946992615300379e-05, "loss": 0.9423, "step": 2370 }, { "epoch": 0.24, "learning_rate": 5.946947817168381e-05, "loss": 0.4505, "step": 2371 }, { "epoch": 0.24, "learning_rate": 5.9469030002831284e-05, "loss": 0.4056, "step": 2372 }, { "epoch": 0.24, "learning_rate": 5.946858164644906e-05, "loss": 1.2519, "step": 2373 }, { "epoch": 0.24, "learning_rate": 5.946813310253999e-05, "loss": 1.0287, "step": 2374 }, { "epoch": 0.24, "learning_rate": 5.946768437110693e-05, "loss": 0.7399, "step": 2375 }, { "epoch": 0.24, "learning_rate": 5.946723545215274e-05, "loss": 1.2247, "step": 2376 }, { "epoch": 0.24, "learning_rate": 5.946678634568027e-05, "loss": 0.5443, "step": 2377 }, { "epoch": 0.24, "learning_rate": 5.946633705169239e-05, "loss": 0.4095, "step": 2378 }, { "epoch": 0.24, "learning_rate": 5.9465887570191945e-05, "loss": 0.9901, "step": 2379 }, { "epoch": 0.24, "learning_rate": 5.9465437901181795e-05, "loss": 0.9318, "step": 2380 }, { "epoch": 0.24, "learning_rate": 5.9464988044664814e-05, "loss": 0.2323, "step": 2381 }, { "epoch": 0.24, "learning_rate": 5.946453800064386e-05, "loss": 0.8907, "step": 2382 }, { "epoch": 0.24, "learning_rate": 5.946408776912179e-05, "loss": 0.8393, "step": 2383 }, { "epoch": 0.24, "learning_rate": 5.9463637350101484e-05, "loss": 0.3813, "step": 2384 }, { "epoch": 0.24, "learning_rate": 5.946318674358578e-05, "loss": 0.9322, "step": 2385 }, { "epoch": 0.24, "learning_rate": 5.946273594957758e-05, "loss": 0.4559, "step": 2386 }, { "epoch": 0.24, "learning_rate": 5.9462284968079727e-05, "loss": 0.3484, "step": 2387 }, { "epoch": 0.24, "learning_rate": 5.946183379909511e-05, "loss": 0.3934, "step": 2388 }, { "epoch": 0.24, "learning_rate": 5.9461382442626584e-05, "loss": 0.7687, "step": 2389 }, { "epoch": 0.24, "learning_rate": 5.946093089867703e-05, "loss": 1.3251, "step": 2390 }, { "epoch": 0.24, "learning_rate": 5.946047916724932e-05, "loss": 1.1644, "step": 2391 }, { "epoch": 0.24, "learning_rate": 5.9460027248346323e-05, "loss": 0.6371, "step": 2392 }, { "epoch": 0.24, "learning_rate": 5.945957514197092e-05, "loss": 0.377, "step": 2393 }, { "epoch": 0.24, "learning_rate": 5.9459122848125994e-05, "loss": 0.9156, "step": 2394 }, { "epoch": 0.24, "learning_rate": 5.945867036681441e-05, "loss": 0.5349, "step": 2395 }, { "epoch": 0.24, "learning_rate": 5.9458217698039065e-05, "loss": 0.6317, "step": 2396 }, { "epoch": 0.24, "learning_rate": 5.9457764841802824e-05, "loss": 0.7928, "step": 2397 }, { "epoch": 0.24, "learning_rate": 5.9457311798108576e-05, "loss": 1.4559, "step": 2398 }, { "epoch": 0.24, "learning_rate": 5.945685856695919e-05, "loss": 0.9948, "step": 2399 }, { "epoch": 0.24, "learning_rate": 5.945640514835757e-05, "loss": 0.2423, "step": 2400 }, { "epoch": 0.24, "learning_rate": 5.94559515423066e-05, "loss": 1.5212, "step": 2401 }, { "epoch": 0.24, "learning_rate": 5.945549774880915e-05, "loss": 0.771, "step": 2402 }, { "epoch": 0.24, "learning_rate": 5.945504376786812e-05, "loss": 0.5958, "step": 2403 }, { "epoch": 0.24, "learning_rate": 5.9454589599486404e-05, "loss": 0.476, "step": 2404 }, { "epoch": 0.24, "learning_rate": 5.9454135243666876e-05, "loss": 0.585, "step": 2405 }, { "epoch": 0.24, "learning_rate": 5.945368070041244e-05, "loss": 2.3113, "step": 2406 }, { "epoch": 0.24, "learning_rate": 5.9453225969725985e-05, "loss": 2.065, "step": 2407 }, { "epoch": 0.24, "learning_rate": 5.9452771051610404e-05, "loss": 0.6814, "step": 2408 }, { "epoch": 0.24, "learning_rate": 5.945231594606859e-05, "loss": 1.164, "step": 2409 }, { "epoch": 0.24, "learning_rate": 5.945186065310344e-05, "loss": 0.5236, "step": 2410 }, { "epoch": 0.24, "learning_rate": 5.9451405172717856e-05, "loss": 0.6364, "step": 2411 }, { "epoch": 0.24, "learning_rate": 5.945094950491474e-05, "loss": 0.8572, "step": 2412 }, { "epoch": 0.25, "learning_rate": 5.9450493649696975e-05, "loss": 0.8268, "step": 2413 }, { "epoch": 0.25, "learning_rate": 5.945003760706747e-05, "loss": 1.0304, "step": 2414 }, { "epoch": 0.25, "learning_rate": 5.944958137702914e-05, "loss": 0.5459, "step": 2415 }, { "epoch": 0.25, "learning_rate": 5.944912495958487e-05, "loss": 0.5306, "step": 2416 }, { "epoch": 0.25, "learning_rate": 5.944866835473758e-05, "loss": 0.9092, "step": 2417 }, { "epoch": 0.25, "learning_rate": 5.944821156249016e-05, "loss": 0.4681, "step": 2418 }, { "epoch": 0.25, "learning_rate": 5.9447754582845524e-05, "loss": 0.5539, "step": 2419 }, { "epoch": 0.25, "learning_rate": 5.9447297415806584e-05, "loss": 0.3933, "step": 2420 }, { "epoch": 0.25, "learning_rate": 5.944684006137624e-05, "loss": 1.0416, "step": 2421 }, { "epoch": 0.25, "learning_rate": 5.944638251955741e-05, "loss": 0.2931, "step": 2422 }, { "epoch": 0.25, "learning_rate": 5.9445924790353004e-05, "loss": 1.4223, "step": 2423 }, { "epoch": 0.25, "learning_rate": 5.944546687376594e-05, "loss": 0.4252, "step": 2424 }, { "epoch": 0.25, "learning_rate": 5.944500876979912e-05, "loss": 0.9621, "step": 2425 }, { "epoch": 0.25, "learning_rate": 5.944455047845547e-05, "loss": 0.2691, "step": 2426 }, { "epoch": 0.25, "learning_rate": 5.9444091999737895e-05, "loss": 0.7919, "step": 2427 }, { "epoch": 0.25, "learning_rate": 5.9443633333649325e-05, "loss": 1.3725, "step": 2428 }, { "epoch": 0.25, "learning_rate": 5.944317448019268e-05, "loss": 1.4259, "step": 2429 }, { "epoch": 0.25, "learning_rate": 5.9442715439370864e-05, "loss": 0.7938, "step": 2430 }, { "epoch": 0.25, "learning_rate": 5.944225621118681e-05, "loss": 0.9755, "step": 2431 }, { "epoch": 0.25, "learning_rate": 5.944179679564344e-05, "loss": 1.4638, "step": 2432 }, { "epoch": 0.25, "learning_rate": 5.9441337192743666e-05, "loss": 1.3943, "step": 2433 }, { "epoch": 0.25, "learning_rate": 5.944087740249043e-05, "loss": 1.7849, "step": 2434 }, { "epoch": 0.25, "learning_rate": 5.9440417424886655e-05, "loss": 0.8695, "step": 2435 }, { "epoch": 0.25, "learning_rate": 5.943995725993526e-05, "loss": 0.4596, "step": 2436 }, { "epoch": 0.25, "learning_rate": 5.9439496907639166e-05, "loss": 0.7793, "step": 2437 }, { "epoch": 0.25, "learning_rate": 5.943903636800132e-05, "loss": 1.1398, "step": 2438 }, { "epoch": 0.25, "learning_rate": 5.943857564102464e-05, "loss": 1.1037, "step": 2439 }, { "epoch": 0.25, "learning_rate": 5.943811472671208e-05, "loss": 0.2739, "step": 2440 }, { "epoch": 0.25, "learning_rate": 5.943765362506654e-05, "loss": 1.152, "step": 2441 }, { "epoch": 0.25, "learning_rate": 5.943719233609098e-05, "loss": 1.4002, "step": 2442 }, { "epoch": 0.25, "learning_rate": 5.943673085978832e-05, "loss": 0.3812, "step": 2443 }, { "epoch": 0.25, "learning_rate": 5.943626919616152e-05, "loss": 0.8041, "step": 2444 }, { "epoch": 0.25, "learning_rate": 5.9435807345213485e-05, "loss": 0.6315, "step": 2445 }, { "epoch": 0.25, "learning_rate": 5.9435345306947176e-05, "loss": 0.3944, "step": 2446 }, { "epoch": 0.25, "learning_rate": 5.943488308136552e-05, "loss": 1.0586, "step": 2447 }, { "epoch": 0.25, "learning_rate": 5.943442066847147e-05, "loss": 0.7982, "step": 2448 }, { "epoch": 0.25, "learning_rate": 5.943395806826796e-05, "loss": 0.3336, "step": 2449 }, { "epoch": 0.25, "learning_rate": 5.943349528075795e-05, "loss": 1.1432, "step": 2450 }, { "epoch": 0.25, "learning_rate": 5.943303230594436e-05, "loss": 1.0904, "step": 2451 }, { "epoch": 0.25, "learning_rate": 5.9432569143830165e-05, "loss": 1.3246, "step": 2452 }, { "epoch": 0.25, "learning_rate": 5.9432105794418286e-05, "loss": 0.3778, "step": 2453 }, { "epoch": 0.25, "learning_rate": 5.943164225771168e-05, "loss": 1.0095, "step": 2454 }, { "epoch": 0.25, "learning_rate": 5.9431178533713306e-05, "loss": 0.4772, "step": 2455 }, { "epoch": 0.25, "learning_rate": 5.94307146224261e-05, "loss": 0.3371, "step": 2456 }, { "epoch": 0.25, "learning_rate": 5.9430250523853024e-05, "loss": 0.8545, "step": 2457 }, { "epoch": 0.25, "learning_rate": 5.9429786237997034e-05, "loss": 0.7143, "step": 2458 }, { "epoch": 0.25, "learning_rate": 5.942932176486108e-05, "loss": 1.2538, "step": 2459 }, { "epoch": 0.25, "learning_rate": 5.942885710444812e-05, "loss": 0.4512, "step": 2460 }, { "epoch": 0.25, "learning_rate": 5.94283922567611e-05, "loss": 1.2608, "step": 2461 }, { "epoch": 0.25, "learning_rate": 5.9427927221803e-05, "loss": 0.5212, "step": 2462 }, { "epoch": 0.25, "learning_rate": 5.9427461999576744e-05, "loss": 1.7004, "step": 2463 }, { "epoch": 0.25, "learning_rate": 5.9426996590085334e-05, "loss": 0.3162, "step": 2464 }, { "epoch": 0.25, "learning_rate": 5.9426530993331705e-05, "loss": 1.3919, "step": 2465 }, { "epoch": 0.25, "learning_rate": 5.942606520931883e-05, "loss": 0.5451, "step": 2466 }, { "epoch": 0.25, "learning_rate": 5.9425599238049665e-05, "loss": 0.7366, "step": 2467 }, { "epoch": 0.25, "learning_rate": 5.942513307952719e-05, "loss": 0.3898, "step": 2468 }, { "epoch": 0.25, "learning_rate": 5.942466673375435e-05, "loss": 0.4571, "step": 2469 }, { "epoch": 0.25, "learning_rate": 5.9424200200734136e-05, "loss": 0.6988, "step": 2470 }, { "epoch": 0.25, "learning_rate": 5.94237334804695e-05, "loss": 1.0383, "step": 2471 }, { "epoch": 0.25, "learning_rate": 5.9423266572963414e-05, "loss": 0.9577, "step": 2472 }, { "epoch": 0.25, "learning_rate": 5.942279947821886e-05, "loss": 1.1399, "step": 2473 }, { "epoch": 0.25, "learning_rate": 5.9422332196238794e-05, "loss": 1.4027, "step": 2474 }, { "epoch": 0.25, "learning_rate": 5.942186472702621e-05, "loss": 1.0917, "step": 2475 }, { "epoch": 0.25, "learning_rate": 5.942139707058407e-05, "loss": 0.7248, "step": 2476 }, { "epoch": 0.25, "learning_rate": 5.942092922691535e-05, "loss": 0.5327, "step": 2477 }, { "epoch": 0.25, "learning_rate": 5.942046119602302e-05, "loss": 0.2194, "step": 2478 }, { "epoch": 0.25, "learning_rate": 5.9419992977910076e-05, "loss": 0.6992, "step": 2479 }, { "epoch": 0.25, "learning_rate": 5.941952457257948e-05, "loss": 1.1535, "step": 2480 }, { "epoch": 0.25, "learning_rate": 5.941905598003423e-05, "loss": 0.6142, "step": 2481 }, { "epoch": 0.25, "learning_rate": 5.941858720027729e-05, "loss": 1.3552, "step": 2482 }, { "epoch": 0.25, "learning_rate": 5.941811823331166e-05, "loss": 0.4111, "step": 2483 }, { "epoch": 0.25, "learning_rate": 5.941764907914032e-05, "loss": 0.7859, "step": 2484 }, { "epoch": 0.25, "learning_rate": 5.941717973776625e-05, "loss": 0.8723, "step": 2485 }, { "epoch": 0.25, "learning_rate": 5.9416710209192436e-05, "loss": 0.459, "step": 2486 }, { "epoch": 0.25, "learning_rate": 5.941624049342187e-05, "loss": 0.4722, "step": 2487 }, { "epoch": 0.25, "learning_rate": 5.9415770590457546e-05, "loss": 0.5883, "step": 2488 }, { "epoch": 0.25, "learning_rate": 5.9415300500302436e-05, "loss": 1.0449, "step": 2489 }, { "epoch": 0.25, "learning_rate": 5.941483022295955e-05, "loss": 0.379, "step": 2490 }, { "epoch": 0.25, "learning_rate": 5.941435975843187e-05, "loss": 0.4455, "step": 2491 }, { "epoch": 0.25, "learning_rate": 5.94138891067224e-05, "loss": 1.1085, "step": 2492 }, { "epoch": 0.25, "learning_rate": 5.9413418267834134e-05, "loss": 0.8996, "step": 2493 }, { "epoch": 0.25, "learning_rate": 5.9412947241770054e-05, "loss": 0.2513, "step": 2494 }, { "epoch": 0.25, "learning_rate": 5.9412476028533174e-05, "loss": 0.2914, "step": 2495 }, { "epoch": 0.25, "learning_rate": 5.941200462812648e-05, "loss": 1.319, "step": 2496 }, { "epoch": 0.25, "learning_rate": 5.9411533040552985e-05, "loss": 0.6258, "step": 2497 }, { "epoch": 0.25, "learning_rate": 5.9411061265815674e-05, "loss": 0.632, "step": 2498 }, { "epoch": 0.25, "learning_rate": 5.9410589303917556e-05, "loss": 1.1774, "step": 2499 }, { "epoch": 0.25, "learning_rate": 5.941011715486165e-05, "loss": 1.4072, "step": 2500 }, { "epoch": 0.25, "learning_rate": 5.9409644818650935e-05, "loss": 0.9624, "step": 2501 }, { "epoch": 0.25, "learning_rate": 5.940917229528843e-05, "loss": 1.5124, "step": 2502 }, { "epoch": 0.25, "learning_rate": 5.940869958477714e-05, "loss": 1.4012, "step": 2503 }, { "epoch": 0.25, "learning_rate": 5.9408226687120075e-05, "loss": 0.3687, "step": 2504 }, { "epoch": 0.25, "learning_rate": 5.940775360232024e-05, "loss": 0.4577, "step": 2505 }, { "epoch": 0.25, "learning_rate": 5.940728033038065e-05, "loss": 1.2154, "step": 2506 }, { "epoch": 0.25, "learning_rate": 5.9406806871304315e-05, "loss": 0.9035, "step": 2507 }, { "epoch": 0.25, "learning_rate": 5.940633322509425e-05, "loss": 0.6519, "step": 2508 }, { "epoch": 0.25, "learning_rate": 5.9405859391753464e-05, "loss": 0.6558, "step": 2509 }, { "epoch": 0.25, "learning_rate": 5.9405385371284984e-05, "loss": 0.3812, "step": 2510 }, { "epoch": 0.25, "learning_rate": 5.940491116369181e-05, "loss": 0.3227, "step": 2511 }, { "epoch": 0.26, "learning_rate": 5.940443676897697e-05, "loss": 0.4254, "step": 2512 }, { "epoch": 0.26, "learning_rate": 5.940396218714347e-05, "loss": 0.765, "step": 2513 }, { "epoch": 0.26, "learning_rate": 5.940348741819436e-05, "loss": 1.15, "step": 2514 }, { "epoch": 0.26, "learning_rate": 5.9403012462132624e-05, "loss": 0.9, "step": 2515 }, { "epoch": 0.26, "learning_rate": 5.940253731896131e-05, "loss": 0.508, "step": 2516 }, { "epoch": 0.26, "learning_rate": 5.940206198868344e-05, "loss": 0.6364, "step": 2517 }, { "epoch": 0.26, "learning_rate": 5.9401586471302026e-05, "loss": 1.0135, "step": 2518 }, { "epoch": 0.26, "learning_rate": 5.9401110766820096e-05, "loss": 0.4942, "step": 2519 }, { "epoch": 0.26, "learning_rate": 5.9400634875240684e-05, "loss": 0.4724, "step": 2520 }, { "epoch": 0.26, "learning_rate": 5.9400158796566825e-05, "loss": 1.7387, "step": 2521 }, { "epoch": 0.26, "learning_rate": 5.9399682530801535e-05, "loss": 0.8669, "step": 2522 }, { "epoch": 0.26, "learning_rate": 5.939920607794785e-05, "loss": 0.5612, "step": 2523 }, { "epoch": 0.26, "learning_rate": 5.93987294380088e-05, "loss": 0.7745, "step": 2524 }, { "epoch": 0.26, "learning_rate": 5.9398252610987415e-05, "loss": 0.5662, "step": 2525 }, { "epoch": 0.26, "learning_rate": 5.939777559688674e-05, "loss": 1.0908, "step": 2526 }, { "epoch": 0.26, "learning_rate": 5.93972983957098e-05, "loss": 0.8797, "step": 2527 }, { "epoch": 0.26, "learning_rate": 5.939682100745965e-05, "loss": 0.6313, "step": 2528 }, { "epoch": 0.26, "learning_rate": 5.939634343213929e-05, "loss": 1.2741, "step": 2529 }, { "epoch": 0.26, "learning_rate": 5.93958656697518e-05, "loss": 0.7943, "step": 2530 }, { "epoch": 0.26, "learning_rate": 5.9395387720300205e-05, "loss": 1.0341, "step": 2531 }, { "epoch": 0.26, "learning_rate": 5.9394909583787536e-05, "loss": 0.3303, "step": 2532 }, { "epoch": 0.26, "learning_rate": 5.939443126021685e-05, "loss": 0.3907, "step": 2533 }, { "epoch": 0.26, "learning_rate": 5.939395274959119e-05, "loss": 0.3209, "step": 2534 }, { "epoch": 0.26, "learning_rate": 5.939347405191359e-05, "loss": 1.2374, "step": 2535 }, { "epoch": 0.26, "learning_rate": 5.9392995167187104e-05, "loss": 1.4128, "step": 2536 }, { "epoch": 0.26, "learning_rate": 5.939251609541478e-05, "loss": 1.1299, "step": 2537 }, { "epoch": 0.26, "learning_rate": 5.9392036836599665e-05, "loss": 0.6629, "step": 2538 }, { "epoch": 0.26, "learning_rate": 5.939155739074481e-05, "loss": 0.7476, "step": 2539 }, { "epoch": 0.26, "learning_rate": 5.939107775785326e-05, "loss": 0.78, "step": 2540 }, { "epoch": 0.26, "learning_rate": 5.939059793792808e-05, "loss": 0.2316, "step": 2541 }, { "epoch": 0.26, "learning_rate": 5.939011793097231e-05, "loss": 1.1474, "step": 2542 }, { "epoch": 0.26, "learning_rate": 5.9389637736989015e-05, "loss": 0.5111, "step": 2543 }, { "epoch": 0.26, "learning_rate": 5.9389157355981235e-05, "loss": 1.0598, "step": 2544 }, { "epoch": 0.26, "learning_rate": 5.9388676787952044e-05, "loss": 1.5312, "step": 2545 }, { "epoch": 0.26, "learning_rate": 5.9388196032904497e-05, "loss": 0.5879, "step": 2546 }, { "epoch": 0.26, "learning_rate": 5.938771509084164e-05, "loss": 0.8647, "step": 2547 }, { "epoch": 0.26, "learning_rate": 5.938723396176656e-05, "loss": 0.9654, "step": 2548 }, { "epoch": 0.26, "learning_rate": 5.9386752645682286e-05, "loss": 1.0643, "step": 2549 }, { "epoch": 0.26, "learning_rate": 5.9386271142591917e-05, "loss": 0.4193, "step": 2550 }, { "epoch": 0.26, "learning_rate": 5.938578945249848e-05, "loss": 0.3744, "step": 2551 }, { "epoch": 0.26, "learning_rate": 5.938530757540506e-05, "loss": 0.688, "step": 2552 }, { "epoch": 0.26, "learning_rate": 5.938482551131473e-05, "loss": 0.5685, "step": 2553 }, { "epoch": 0.26, "learning_rate": 5.938434326023054e-05, "loss": 0.8555, "step": 2554 }, { "epoch": 0.26, "learning_rate": 5.938386082215558e-05, "loss": 1.0364, "step": 2555 }, { "epoch": 0.26, "learning_rate": 5.938337819709289e-05, "loss": 1.0913, "step": 2556 }, { "epoch": 0.26, "learning_rate": 5.9382895385045575e-05, "loss": 0.5181, "step": 2557 }, { "epoch": 0.26, "learning_rate": 5.938241238601669e-05, "loss": 0.6151, "step": 2558 }, { "epoch": 0.26, "learning_rate": 5.938192920000931e-05, "loss": 1.0, "step": 2559 }, { "epoch": 0.26, "learning_rate": 5.93814458270265e-05, "loss": 0.7786, "step": 2560 }, { "epoch": 0.26, "learning_rate": 5.938096226707135e-05, "loss": 0.6949, "step": 2561 }, { "epoch": 0.26, "learning_rate": 5.938047852014694e-05, "loss": 0.8818, "step": 2562 }, { "epoch": 0.26, "learning_rate": 5.937999458625634e-05, "loss": 0.5558, "step": 2563 }, { "epoch": 0.26, "learning_rate": 5.937951046540262e-05, "loss": 0.836, "step": 2564 }, { "epoch": 0.26, "learning_rate": 5.937902615758888e-05, "loss": 0.7766, "step": 2565 }, { "epoch": 0.26, "learning_rate": 5.9378541662818195e-05, "loss": 0.7394, "step": 2566 }, { "epoch": 0.26, "learning_rate": 5.9378056981093644e-05, "loss": 0.3822, "step": 2567 }, { "epoch": 0.26, "learning_rate": 5.937757211241832e-05, "loss": 0.2547, "step": 2568 }, { "epoch": 0.26, "learning_rate": 5.9377087056795296e-05, "loss": 1.2686, "step": 2569 }, { "epoch": 0.26, "learning_rate": 5.937660181422767e-05, "loss": 0.5114, "step": 2570 }, { "epoch": 0.26, "learning_rate": 5.937611638471852e-05, "loss": 1.1328, "step": 2571 }, { "epoch": 0.26, "learning_rate": 5.937563076827095e-05, "loss": 0.6315, "step": 2572 }, { "epoch": 0.26, "learning_rate": 5.9375144964888035e-05, "loss": 0.8797, "step": 2573 }, { "epoch": 0.26, "learning_rate": 5.937465897457287e-05, "loss": 0.1253, "step": 2574 }, { "epoch": 0.26, "learning_rate": 5.9374172797328554e-05, "loss": 1.3569, "step": 2575 }, { "epoch": 0.26, "learning_rate": 5.937368643315818e-05, "loss": 0.5742, "step": 2576 }, { "epoch": 0.26, "learning_rate": 5.937319988206483e-05, "loss": 0.5999, "step": 2577 }, { "epoch": 0.26, "learning_rate": 5.937271314405162e-05, "loss": 1.2341, "step": 2578 }, { "epoch": 0.26, "learning_rate": 5.9372226219121635e-05, "loss": 0.5089, "step": 2579 }, { "epoch": 0.26, "learning_rate": 5.937173910727797e-05, "loss": 0.1828, "step": 2580 }, { "epoch": 0.26, "learning_rate": 5.937125180852374e-05, "loss": 1.1294, "step": 2581 }, { "epoch": 0.26, "learning_rate": 5.937076432286203e-05, "loss": 0.99, "step": 2582 }, { "epoch": 0.26, "learning_rate": 5.937027665029595e-05, "loss": 1.1303, "step": 2583 }, { "epoch": 0.26, "learning_rate": 5.93697887908286e-05, "loss": 1.1076, "step": 2584 }, { "epoch": 0.26, "learning_rate": 5.936930074446309e-05, "loss": 1.3479, "step": 2585 }, { "epoch": 0.26, "learning_rate": 5.936881251120252e-05, "loss": 0.7112, "step": 2586 }, { "epoch": 0.26, "learning_rate": 5.936832409105001e-05, "loss": 0.635, "step": 2587 }, { "epoch": 0.26, "learning_rate": 5.936783548400865e-05, "loss": 0.5022, "step": 2588 }, { "epoch": 0.26, "learning_rate": 5.936734669008156e-05, "loss": 0.6093, "step": 2589 }, { "epoch": 0.26, "learning_rate": 5.9366857709271846e-05, "loss": 1.2218, "step": 2590 }, { "epoch": 0.26, "learning_rate": 5.9366368541582614e-05, "loss": 0.6383, "step": 2591 }, { "epoch": 0.26, "learning_rate": 5.9365879187017e-05, "loss": 1.3004, "step": 2592 }, { "epoch": 0.26, "learning_rate": 5.936538964557809e-05, "loss": 0.5613, "step": 2593 }, { "epoch": 0.26, "learning_rate": 5.9364899917269016e-05, "loss": 1.2688, "step": 2594 }, { "epoch": 0.26, "learning_rate": 5.936441000209288e-05, "loss": 0.871, "step": 2595 }, { "epoch": 0.26, "learning_rate": 5.936391990005282e-05, "loss": 0.6956, "step": 2596 }, { "epoch": 0.26, "learning_rate": 5.9363429611151945e-05, "loss": 0.8977, "step": 2597 }, { "epoch": 0.26, "learning_rate": 5.936293913539337e-05, "loss": 0.8609, "step": 2598 }, { "epoch": 0.26, "learning_rate": 5.9362448472780226e-05, "loss": 0.8019, "step": 2599 }, { "epoch": 0.26, "learning_rate": 5.936195762331562e-05, "loss": 1.2905, "step": 2600 }, { "epoch": 0.26, "learning_rate": 5.93614665870027e-05, "loss": 0.8435, "step": 2601 }, { "epoch": 0.26, "learning_rate": 5.936097536384456e-05, "loss": 1.3059, "step": 2602 }, { "epoch": 0.26, "learning_rate": 5.936048395384435e-05, "loss": 1.6494, "step": 2603 }, { "epoch": 0.26, "learning_rate": 5.9359992357005196e-05, "loss": 0.8236, "step": 2604 }, { "epoch": 0.26, "learning_rate": 5.9359500573330215e-05, "loss": 0.4517, "step": 2605 }, { "epoch": 0.26, "learning_rate": 5.9359008602822534e-05, "loss": 0.5884, "step": 2606 }, { "epoch": 0.26, "learning_rate": 5.9358516445485305e-05, "loss": 0.7392, "step": 2607 }, { "epoch": 0.26, "learning_rate": 5.935802410132163e-05, "loss": 0.4826, "step": 2608 }, { "epoch": 0.26, "learning_rate": 5.9357531570334667e-05, "loss": 1.5875, "step": 2609 }, { "epoch": 0.27, "learning_rate": 5.935703885252754e-05, "loss": 0.3892, "step": 2610 }, { "epoch": 0.27, "learning_rate": 5.935654594790339e-05, "loss": 0.8682, "step": 2611 }, { "epoch": 0.27, "learning_rate": 5.9356052856465345e-05, "loss": 0.6221, "step": 2612 }, { "epoch": 0.27, "learning_rate": 5.9355559578216544e-05, "loss": 0.4781, "step": 2613 }, { "epoch": 0.27, "learning_rate": 5.935506611316013e-05, "loss": 1.0887, "step": 2614 }, { "epoch": 0.27, "learning_rate": 5.935457246129925e-05, "loss": 0.6299, "step": 2615 }, { "epoch": 0.27, "learning_rate": 5.9354078622637034e-05, "loss": 0.831, "step": 2616 }, { "epoch": 0.27, "learning_rate": 5.935358459717663e-05, "loss": 0.8885, "step": 2617 }, { "epoch": 0.27, "learning_rate": 5.935309038492117e-05, "loss": 0.9677, "step": 2618 }, { "epoch": 0.27, "learning_rate": 5.935259598587382e-05, "loss": 1.1161, "step": 2619 }, { "epoch": 0.27, "learning_rate": 5.93521014000377e-05, "loss": 0.3266, "step": 2620 }, { "epoch": 0.27, "learning_rate": 5.935160662741599e-05, "loss": 1.3458, "step": 2621 }, { "epoch": 0.27, "learning_rate": 5.935111166801182e-05, "loss": 1.2525, "step": 2622 }, { "epoch": 0.27, "learning_rate": 5.935061652182834e-05, "loss": 0.4709, "step": 2623 }, { "epoch": 0.27, "learning_rate": 5.935012118886869e-05, "loss": 0.9347, "step": 2624 }, { "epoch": 0.27, "learning_rate": 5.9349625669136045e-05, "loss": 1.0765, "step": 2625 }, { "epoch": 0.27, "learning_rate": 5.934912996263355e-05, "loss": 0.4198, "step": 2626 }, { "epoch": 0.27, "learning_rate": 5.934863406936435e-05, "loss": 0.3387, "step": 2627 }, { "epoch": 0.27, "learning_rate": 5.934813798933162e-05, "loss": 0.8128, "step": 2628 }, { "epoch": 0.27, "learning_rate": 5.93476417225385e-05, "loss": 0.386, "step": 2629 }, { "epoch": 0.27, "learning_rate": 5.9347145268988136e-05, "loss": 0.3452, "step": 2630 }, { "epoch": 0.27, "learning_rate": 5.934664862868372e-05, "loss": 0.6041, "step": 2631 }, { "epoch": 0.27, "learning_rate": 5.9346151801628394e-05, "loss": 0.2839, "step": 2632 }, { "epoch": 0.27, "learning_rate": 5.9345654787825315e-05, "loss": 1.3644, "step": 2633 }, { "epoch": 0.27, "learning_rate": 5.9345157587277654e-05, "loss": 0.7991, "step": 2634 }, { "epoch": 0.27, "learning_rate": 5.934466019998858e-05, "loss": 1.184, "step": 2635 }, { "epoch": 0.27, "learning_rate": 5.9344162625961254e-05, "loss": 1.3064, "step": 2636 }, { "epoch": 0.27, "learning_rate": 5.934366486519883e-05, "loss": 1.1136, "step": 2637 }, { "epoch": 0.27, "learning_rate": 5.934316691770448e-05, "loss": 0.7883, "step": 2638 }, { "epoch": 0.27, "learning_rate": 5.9342668783481396e-05, "loss": 0.6127, "step": 2639 }, { "epoch": 0.27, "learning_rate": 5.934217046253272e-05, "loss": 0.4445, "step": 2640 }, { "epoch": 0.27, "learning_rate": 5.934167195486164e-05, "loss": 1.2362, "step": 2641 }, { "epoch": 0.27, "learning_rate": 5.934117326047132e-05, "loss": 0.8911, "step": 2642 }, { "epoch": 0.27, "learning_rate": 5.934067437936493e-05, "loss": 1.1379, "step": 2643 }, { "epoch": 0.27, "learning_rate": 5.934017531154565e-05, "loss": 0.3782, "step": 2644 }, { "epoch": 0.27, "learning_rate": 5.9339676057016666e-05, "loss": 0.9548, "step": 2645 }, { "epoch": 0.27, "learning_rate": 5.933917661578113e-05, "loss": 1.3717, "step": 2646 }, { "epoch": 0.27, "learning_rate": 5.933867698784224e-05, "loss": 0.4404, "step": 2647 }, { "epoch": 0.27, "learning_rate": 5.9338177173203185e-05, "loss": 0.254, "step": 2648 }, { "epoch": 0.27, "learning_rate": 5.933767717186712e-05, "loss": 0.4807, "step": 2649 }, { "epoch": 0.27, "learning_rate": 5.933717698383724e-05, "loss": 0.3947, "step": 2650 }, { "epoch": 0.27, "learning_rate": 5.9336676609116725e-05, "loss": 0.345, "step": 2651 }, { "epoch": 0.27, "learning_rate": 5.933617604770876e-05, "loss": 1.328, "step": 2652 }, { "epoch": 0.27, "learning_rate": 5.933567529961653e-05, "loss": 1.644, "step": 2653 }, { "epoch": 0.27, "learning_rate": 5.9335174364843216e-05, "loss": 0.5056, "step": 2654 }, { "epoch": 0.27, "learning_rate": 5.933467324339203e-05, "loss": 1.2747, "step": 2655 }, { "epoch": 0.27, "learning_rate": 5.9334171935266126e-05, "loss": 1.108, "step": 2656 }, { "epoch": 0.27, "learning_rate": 5.9333670440468715e-05, "loss": 0.5009, "step": 2657 }, { "epoch": 0.27, "learning_rate": 5.9333168759002984e-05, "loss": 0.9502, "step": 2658 }, { "epoch": 0.27, "learning_rate": 5.9332666890872136e-05, "loss": 1.2955, "step": 2659 }, { "epoch": 0.27, "learning_rate": 5.933216483607934e-05, "loss": 0.5542, "step": 2660 }, { "epoch": 0.27, "learning_rate": 5.933166259462782e-05, "loss": 0.903, "step": 2661 }, { "epoch": 0.27, "learning_rate": 5.933116016652075e-05, "loss": 0.2534, "step": 2662 }, { "epoch": 0.27, "learning_rate": 5.933065755176133e-05, "loss": 0.6614, "step": 2663 }, { "epoch": 0.27, "learning_rate": 5.933015475035277e-05, "loss": 0.7016, "step": 2664 }, { "epoch": 0.27, "learning_rate": 5.9329651762298266e-05, "loss": 1.5718, "step": 2665 }, { "epoch": 0.27, "learning_rate": 5.9329148587601005e-05, "loss": 1.3658, "step": 2666 }, { "epoch": 0.27, "learning_rate": 5.932864522626421e-05, "loss": 0.5261, "step": 2667 }, { "epoch": 0.27, "learning_rate": 5.9328141678291066e-05, "loss": 0.7204, "step": 2668 }, { "epoch": 0.27, "learning_rate": 5.9327637943684786e-05, "loss": 1.6531, "step": 2669 }, { "epoch": 0.27, "learning_rate": 5.932713402244858e-05, "loss": 1.0488, "step": 2670 }, { "epoch": 0.27, "learning_rate": 5.932662991458565e-05, "loss": 0.7499, "step": 2671 }, { "epoch": 0.27, "learning_rate": 5.9326125620099196e-05, "loss": 0.2094, "step": 2672 }, { "epoch": 0.27, "learning_rate": 5.9325621138992446e-05, "loss": 0.5847, "step": 2673 }, { "epoch": 0.27, "learning_rate": 5.932511647126859e-05, "loss": 0.9503, "step": 2674 }, { "epoch": 0.27, "learning_rate": 5.932461161693086e-05, "loss": 0.3795, "step": 2675 }, { "epoch": 0.27, "learning_rate": 5.9324106575982445e-05, "loss": 1.3708, "step": 2676 }, { "epoch": 0.27, "learning_rate": 5.9323601348426574e-05, "loss": 0.4262, "step": 2677 }, { "epoch": 0.27, "learning_rate": 5.932309593426646e-05, "loss": 0.6167, "step": 2678 }, { "epoch": 0.27, "learning_rate": 5.932259033350532e-05, "loss": 0.9506, "step": 2679 }, { "epoch": 0.27, "learning_rate": 5.932208454614637e-05, "loss": 0.8584, "step": 2680 }, { "epoch": 0.27, "learning_rate": 5.9321578572192824e-05, "loss": 0.3826, "step": 2681 }, { "epoch": 0.27, "learning_rate": 5.9321072411647916e-05, "loss": 0.6214, "step": 2682 }, { "epoch": 0.27, "learning_rate": 5.932056606451484e-05, "loss": 0.5605, "step": 2683 }, { "epoch": 0.27, "learning_rate": 5.932005953079686e-05, "loss": 0.4148, "step": 2684 }, { "epoch": 0.27, "learning_rate": 5.931955281049715e-05, "loss": 0.8016, "step": 2685 }, { "epoch": 0.27, "learning_rate": 5.931904590361898e-05, "loss": 0.9365, "step": 2686 }, { "epoch": 0.27, "learning_rate": 5.931853881016555e-05, "loss": 0.9676, "step": 2687 }, { "epoch": 0.27, "learning_rate": 5.931803153014009e-05, "loss": 0.7064, "step": 2688 }, { "epoch": 0.27, "learning_rate": 5.931752406354583e-05, "loss": 0.5693, "step": 2689 }, { "epoch": 0.27, "learning_rate": 5.9317016410386e-05, "loss": 0.3687, "step": 2690 }, { "epoch": 0.27, "learning_rate": 5.931650857066383e-05, "loss": 0.8766, "step": 2691 }, { "epoch": 0.27, "learning_rate": 5.931600054438255e-05, "loss": 0.373, "step": 2692 }, { "epoch": 0.27, "learning_rate": 5.93154923315454e-05, "loss": 0.5874, "step": 2693 }, { "epoch": 0.27, "learning_rate": 5.93149839321556e-05, "loss": 0.7334, "step": 2694 }, { "epoch": 0.27, "learning_rate": 5.9314475346216404e-05, "loss": 0.6727, "step": 2695 }, { "epoch": 0.27, "learning_rate": 5.931396657373104e-05, "loss": 0.6347, "step": 2696 }, { "epoch": 0.27, "learning_rate": 5.9313457614702736e-05, "loss": 0.5914, "step": 2697 }, { "epoch": 0.27, "learning_rate": 5.9312948469134745e-05, "loss": 0.9814, "step": 2698 }, { "epoch": 0.27, "learning_rate": 5.93124391370303e-05, "loss": 1.0294, "step": 2699 }, { "epoch": 0.27, "learning_rate": 5.931192961839264e-05, "loss": 0.7637, "step": 2700 }, { "epoch": 0.27, "learning_rate": 5.931141991322501e-05, "loss": 0.7036, "step": 2701 }, { "epoch": 0.27, "learning_rate": 5.9310910021530665e-05, "loss": 0.4761, "step": 2702 }, { "epoch": 0.27, "learning_rate": 5.931039994331282e-05, "loss": 1.2104, "step": 2703 }, { "epoch": 0.27, "learning_rate": 5.930988967857476e-05, "loss": 0.2889, "step": 2704 }, { "epoch": 0.27, "learning_rate": 5.93093792273197e-05, "loss": 0.8682, "step": 2705 }, { "epoch": 0.27, "learning_rate": 5.9308868589550906e-05, "loss": 1.0814, "step": 2706 }, { "epoch": 0.27, "learning_rate": 5.930835776527162e-05, "loss": 0.7622, "step": 2707 }, { "epoch": 0.27, "learning_rate": 5.930784675448509e-05, "loss": 1.4267, "step": 2708 }, { "epoch": 0.28, "learning_rate": 5.930733555719458e-05, "loss": 0.795, "step": 2709 }, { "epoch": 0.28, "learning_rate": 5.9306824173403337e-05, "loss": 0.3121, "step": 2710 }, { "epoch": 0.28, "learning_rate": 5.93063126031146e-05, "loss": 1.3382, "step": 2711 }, { "epoch": 0.28, "learning_rate": 5.9305800846331655e-05, "loss": 0.4616, "step": 2712 }, { "epoch": 0.28, "learning_rate": 5.9305288903057736e-05, "loss": 1.2397, "step": 2713 }, { "epoch": 0.28, "learning_rate": 5.930477677329611e-05, "loss": 1.1048, "step": 2714 }, { "epoch": 0.28, "learning_rate": 5.930426445705002e-05, "loss": 0.5483, "step": 2715 }, { "epoch": 0.28, "learning_rate": 5.930375195432275e-05, "loss": 0.8359, "step": 2716 }, { "epoch": 0.28, "learning_rate": 5.930323926511755e-05, "loss": 0.3722, "step": 2717 }, { "epoch": 0.28, "learning_rate": 5.930272638943768e-05, "loss": 0.8784, "step": 2718 }, { "epoch": 0.28, "learning_rate": 5.930221332728641e-05, "loss": 0.5223, "step": 2719 }, { "epoch": 0.28, "learning_rate": 5.930170007866699e-05, "loss": 0.8484, "step": 2720 }, { "epoch": 0.28, "learning_rate": 5.930118664358272e-05, "loss": 1.3115, "step": 2721 }, { "epoch": 0.28, "learning_rate": 5.930067302203682e-05, "loss": 0.7204, "step": 2722 }, { "epoch": 0.28, "learning_rate": 5.9300159214032595e-05, "loss": 0.774, "step": 2723 }, { "epoch": 0.28, "learning_rate": 5.92996452195733e-05, "loss": 0.5616, "step": 2724 }, { "epoch": 0.28, "learning_rate": 5.929913103866221e-05, "loss": 0.2959, "step": 2725 }, { "epoch": 0.28, "learning_rate": 5.92986166713026e-05, "loss": 1.1404, "step": 2726 }, { "epoch": 0.28, "learning_rate": 5.929810211749774e-05, "loss": 1.476, "step": 2727 }, { "epoch": 0.28, "learning_rate": 5.9297587377250894e-05, "loss": 1.35, "step": 2728 }, { "epoch": 0.28, "learning_rate": 5.929707245056536e-05, "loss": 0.4244, "step": 2729 }, { "epoch": 0.28, "learning_rate": 5.9296557337444394e-05, "loss": 0.829, "step": 2730 }, { "epoch": 0.28, "learning_rate": 5.9296042037891285e-05, "loss": 1.1587, "step": 2731 }, { "epoch": 0.28, "learning_rate": 5.929552655190931e-05, "loss": 0.8993, "step": 2732 }, { "epoch": 0.28, "learning_rate": 5.929501087950175e-05, "loss": 0.9386, "step": 2733 }, { "epoch": 0.28, "learning_rate": 5.929449502067188e-05, "loss": 0.2318, "step": 2734 }, { "epoch": 0.28, "learning_rate": 5.929397897542299e-05, "loss": 0.3798, "step": 2735 }, { "epoch": 0.28, "learning_rate": 5.929346274375836e-05, "loss": 0.4701, "step": 2736 }, { "epoch": 0.28, "learning_rate": 5.929294632568128e-05, "loss": 1.2082, "step": 2737 }, { "epoch": 0.28, "learning_rate": 5.929242972119503e-05, "loss": 1.2994, "step": 2738 }, { "epoch": 0.28, "learning_rate": 5.92919129303029e-05, "loss": 0.3465, "step": 2739 }, { "epoch": 0.28, "learning_rate": 5.9291395953008186e-05, "loss": 0.5373, "step": 2740 }, { "epoch": 0.28, "learning_rate": 5.929087878931416e-05, "loss": 1.2316, "step": 2741 }, { "epoch": 0.28, "learning_rate": 5.9290361439224135e-05, "loss": 1.1344, "step": 2742 }, { "epoch": 0.28, "learning_rate": 5.928984390274139e-05, "loss": 0.5771, "step": 2743 }, { "epoch": 0.28, "learning_rate": 5.9289326179869214e-05, "loss": 0.7135, "step": 2744 }, { "epoch": 0.28, "learning_rate": 5.928880827061092e-05, "loss": 1.198, "step": 2745 }, { "epoch": 0.28, "learning_rate": 5.9288290174969785e-05, "loss": 0.3488, "step": 2746 }, { "epoch": 0.28, "learning_rate": 5.928777189294911e-05, "loss": 1.2442, "step": 2747 }, { "epoch": 0.28, "learning_rate": 5.92872534245522e-05, "loss": 1.5295, "step": 2748 }, { "epoch": 0.28, "learning_rate": 5.928673476978235e-05, "loss": 0.3148, "step": 2749 }, { "epoch": 0.28, "learning_rate": 5.9286215928642874e-05, "loss": 1.3636, "step": 2750 }, { "epoch": 0.28, "learning_rate": 5.9285696901137044e-05, "loss": 1.1427, "step": 2751 }, { "epoch": 0.28, "learning_rate": 5.9285177687268185e-05, "loss": 0.3014, "step": 2752 }, { "epoch": 0.28, "learning_rate": 5.92846582870396e-05, "loss": 1.1612, "step": 2753 }, { "epoch": 0.28, "learning_rate": 5.928413870045458e-05, "loss": 1.7744, "step": 2754 }, { "epoch": 0.28, "learning_rate": 5.9283618927516444e-05, "loss": 1.3985, "step": 2755 }, { "epoch": 0.28, "learning_rate": 5.9283098968228505e-05, "loss": 0.7399, "step": 2756 }, { "epoch": 0.28, "learning_rate": 5.9282578822594055e-05, "loss": 0.9183, "step": 2757 }, { "epoch": 0.28, "learning_rate": 5.928205849061642e-05, "loss": 1.4553, "step": 2758 }, { "epoch": 0.28, "learning_rate": 5.92815379722989e-05, "loss": 0.3375, "step": 2759 }, { "epoch": 0.28, "learning_rate": 5.928101726764482e-05, "loss": 1.4664, "step": 2760 }, { "epoch": 0.28, "learning_rate": 5.928049637665748e-05, "loss": 1.0055, "step": 2761 }, { "epoch": 0.28, "learning_rate": 5.92799752993402e-05, "loss": 0.247, "step": 2762 }, { "epoch": 0.28, "learning_rate": 5.927945403569629e-05, "loss": 0.9494, "step": 2763 }, { "epoch": 0.28, "learning_rate": 5.9278932585729084e-05, "loss": 0.8962, "step": 2764 }, { "epoch": 0.28, "learning_rate": 5.927841094944188e-05, "loss": 0.6279, "step": 2765 }, { "epoch": 0.28, "learning_rate": 5.9277889126838e-05, "loss": 1.1228, "step": 2766 }, { "epoch": 0.28, "learning_rate": 5.927736711792079e-05, "loss": 0.5425, "step": 2767 }, { "epoch": 0.28, "learning_rate": 5.9276844922693544e-05, "loss": 0.8397, "step": 2768 }, { "epoch": 0.28, "learning_rate": 5.927632254115959e-05, "loss": 0.6224, "step": 2769 }, { "epoch": 0.28, "learning_rate": 5.927579997332226e-05, "loss": 1.4062, "step": 2770 }, { "epoch": 0.28, "learning_rate": 5.927527721918488e-05, "loss": 1.0868, "step": 2771 }, { "epoch": 0.28, "learning_rate": 5.9274754278750775e-05, "loss": 0.9379, "step": 2772 }, { "epoch": 0.28, "learning_rate": 5.9274231152023265e-05, "loss": 0.3872, "step": 2773 }, { "epoch": 0.28, "learning_rate": 5.9273707839005684e-05, "loss": 0.4311, "step": 2774 }, { "epoch": 0.28, "learning_rate": 5.927318433970137e-05, "loss": 1.2329, "step": 2775 }, { "epoch": 0.28, "learning_rate": 5.927266065411363e-05, "loss": 0.2967, "step": 2776 }, { "epoch": 0.28, "learning_rate": 5.9272136782245826e-05, "loss": 0.526, "step": 2777 }, { "epoch": 0.28, "learning_rate": 5.9271612724101276e-05, "loss": 0.5499, "step": 2778 }, { "epoch": 0.28, "learning_rate": 5.927108847968332e-05, "loss": 0.9568, "step": 2779 }, { "epoch": 0.28, "learning_rate": 5.92705640489953e-05, "loss": 0.957, "step": 2780 }, { "epoch": 0.28, "learning_rate": 5.9270039432040525e-05, "loss": 0.6194, "step": 2781 }, { "epoch": 0.28, "learning_rate": 5.926951462882237e-05, "loss": 0.4289, "step": 2782 }, { "epoch": 0.28, "learning_rate": 5.9268989639344156e-05, "loss": 0.9951, "step": 2783 }, { "epoch": 0.28, "learning_rate": 5.926846446360923e-05, "loss": 1.0308, "step": 2784 }, { "epoch": 0.28, "learning_rate": 5.926793910162092e-05, "loss": 0.8851, "step": 2785 }, { "epoch": 0.28, "learning_rate": 5.9267413553382586e-05, "loss": 0.5885, "step": 2786 }, { "epoch": 0.28, "learning_rate": 5.9266887818897567e-05, "loss": 0.8935, "step": 2787 }, { "epoch": 0.28, "learning_rate": 5.9266361898169206e-05, "loss": 0.4752, "step": 2788 }, { "epoch": 0.28, "learning_rate": 5.926583579120085e-05, "loss": 0.5286, "step": 2789 }, { "epoch": 0.28, "learning_rate": 5.926530949799585e-05, "loss": 0.8765, "step": 2790 }, { "epoch": 0.28, "learning_rate": 5.926478301855755e-05, "loss": 0.2945, "step": 2791 }, { "epoch": 0.28, "learning_rate": 5.926425635288931e-05, "loss": 0.4005, "step": 2792 }, { "epoch": 0.28, "learning_rate": 5.9263729500994475e-05, "loss": 0.7594, "step": 2793 }, { "epoch": 0.28, "learning_rate": 5.926320246287639e-05, "loss": 1.1652, "step": 2794 }, { "epoch": 0.28, "learning_rate": 5.926267523853842e-05, "loss": 1.0239, "step": 2795 }, { "epoch": 0.28, "learning_rate": 5.926214782798393e-05, "loss": 1.2178, "step": 2796 }, { "epoch": 0.28, "learning_rate": 5.926162023121625e-05, "loss": 1.3344, "step": 2797 }, { "epoch": 0.28, "learning_rate": 5.926109244823875e-05, "loss": 1.4422, "step": 2798 }, { "epoch": 0.28, "learning_rate": 5.9260564479054785e-05, "loss": 0.5753, "step": 2799 }, { "epoch": 0.28, "learning_rate": 5.926003632366773e-05, "loss": 0.954, "step": 2800 }, { "epoch": 0.28, "learning_rate": 5.925950798208093e-05, "loss": 1.0034, "step": 2801 }, { "epoch": 0.28, "learning_rate": 5.925897945429775e-05, "loss": 0.3456, "step": 2802 }, { "epoch": 0.28, "learning_rate": 5.9258450740321567e-05, "loss": 0.6127, "step": 2803 }, { "epoch": 0.28, "learning_rate": 5.9257921840155715e-05, "loss": 0.4492, "step": 2804 }, { "epoch": 0.28, "learning_rate": 5.925739275380359e-05, "loss": 0.9029, "step": 2805 }, { "epoch": 0.28, "learning_rate": 5.9256863481268546e-05, "loss": 0.6804, "step": 2806 }, { "epoch": 0.29, "learning_rate": 5.925633402255395e-05, "loss": 1.4224, "step": 2807 }, { "epoch": 0.29, "learning_rate": 5.925580437766318e-05, "loss": 0.7975, "step": 2808 }, { "epoch": 0.29, "learning_rate": 5.92552745465996e-05, "loss": 0.5334, "step": 2809 }, { "epoch": 0.29, "learning_rate": 5.925474452936658e-05, "loss": 1.1681, "step": 2810 }, { "epoch": 0.29, "learning_rate": 5.925421432596749e-05, "loss": 0.7034, "step": 2811 }, { "epoch": 0.29, "learning_rate": 5.9253683936405716e-05, "loss": 1.7508, "step": 2812 }, { "epoch": 0.29, "learning_rate": 5.925315336068462e-05, "loss": 0.7097, "step": 2813 }, { "epoch": 0.29, "learning_rate": 5.9252622598807593e-05, "loss": 0.4401, "step": 2814 }, { "epoch": 0.29, "learning_rate": 5.9252091650778e-05, "loss": 0.7476, "step": 2815 }, { "epoch": 0.29, "learning_rate": 5.925156051659922e-05, "loss": 1.0557, "step": 2816 }, { "epoch": 0.29, "learning_rate": 5.9251029196274636e-05, "loss": 1.6201, "step": 2817 }, { "epoch": 0.29, "learning_rate": 5.925049768980764e-05, "loss": 0.9322, "step": 2818 }, { "epoch": 0.29, "learning_rate": 5.924996599720159e-05, "loss": 1.6787, "step": 2819 }, { "epoch": 0.29, "learning_rate": 5.924943411845989e-05, "loss": 0.4933, "step": 2820 }, { "epoch": 0.29, "learning_rate": 5.924890205358593e-05, "loss": 1.4084, "step": 2821 }, { "epoch": 0.29, "learning_rate": 5.9248369802583066e-05, "loss": 0.1279, "step": 2822 }, { "epoch": 0.29, "learning_rate": 5.924783736545471e-05, "loss": 0.9762, "step": 2823 }, { "epoch": 0.29, "learning_rate": 5.924730474220424e-05, "loss": 0.3286, "step": 2824 }, { "epoch": 0.29, "learning_rate": 5.924677193283506e-05, "loss": 0.4072, "step": 2825 }, { "epoch": 0.29, "learning_rate": 5.924623893735054e-05, "loss": 0.693, "step": 2826 }, { "epoch": 0.29, "learning_rate": 5.924570575575408e-05, "loss": 0.2016, "step": 2827 }, { "epoch": 0.29, "learning_rate": 5.924517238804908e-05, "loss": 0.4246, "step": 2828 }, { "epoch": 0.29, "learning_rate": 5.9244638834238924e-05, "loss": 1.0796, "step": 2829 }, { "epoch": 0.29, "learning_rate": 5.9244105094327017e-05, "loss": 0.5491, "step": 2830 }, { "epoch": 0.29, "learning_rate": 5.9243571168316746e-05, "loss": 0.9373, "step": 2831 }, { "epoch": 0.29, "learning_rate": 5.924303705621151e-05, "loss": 0.916, "step": 2832 }, { "epoch": 0.29, "learning_rate": 5.924250275801472e-05, "loss": 0.3933, "step": 2833 }, { "epoch": 0.29, "learning_rate": 5.924196827372976e-05, "loss": 1.198, "step": 2834 }, { "epoch": 0.29, "learning_rate": 5.924143360336004e-05, "loss": 0.3912, "step": 2835 }, { "epoch": 0.29, "learning_rate": 5.924089874690896e-05, "loss": 1.0365, "step": 2836 }, { "epoch": 0.29, "learning_rate": 5.924036370437992e-05, "loss": 1.036, "step": 2837 }, { "epoch": 0.29, "learning_rate": 5.923982847577634e-05, "loss": 0.9326, "step": 2838 }, { "epoch": 0.29, "learning_rate": 5.923929306110161e-05, "loss": 1.3797, "step": 2839 }, { "epoch": 0.29, "learning_rate": 5.9238757460359136e-05, "loss": 0.9531, "step": 2840 }, { "epoch": 0.29, "learning_rate": 5.923822167355234e-05, "loss": 0.4707, "step": 2841 }, { "epoch": 0.29, "learning_rate": 5.923768570068462e-05, "loss": 1.4376, "step": 2842 }, { "epoch": 0.29, "learning_rate": 5.9237149541759396e-05, "loss": 2.0054, "step": 2843 }, { "epoch": 0.29, "learning_rate": 5.923661319678006e-05, "loss": 0.6978, "step": 2844 }, { "epoch": 0.29, "learning_rate": 5.923607666575006e-05, "loss": 0.7265, "step": 2845 }, { "epoch": 0.29, "learning_rate": 5.923553994867278e-05, "loss": 0.8122, "step": 2846 }, { "epoch": 0.29, "learning_rate": 5.923500304555165e-05, "loss": 1.5095, "step": 2847 }, { "epoch": 0.29, "learning_rate": 5.923446595639008e-05, "loss": 0.2637, "step": 2848 }, { "epoch": 0.29, "learning_rate": 5.923392868119148e-05, "loss": 0.4773, "step": 2849 }, { "epoch": 0.29, "learning_rate": 5.923339121995929e-05, "loss": 0.2558, "step": 2850 }, { "epoch": 0.29, "learning_rate": 5.923285357269692e-05, "loss": 1.096, "step": 2851 }, { "epoch": 0.29, "learning_rate": 5.923231573940778e-05, "loss": 1.2387, "step": 2852 }, { "epoch": 0.29, "learning_rate": 5.923177772009531e-05, "loss": 0.7443, "step": 2853 }, { "epoch": 0.29, "learning_rate": 5.923123951476293e-05, "loss": 0.3131, "step": 2854 }, { "epoch": 0.29, "learning_rate": 5.923070112341406e-05, "loss": 1.7054, "step": 2855 }, { "epoch": 0.29, "learning_rate": 5.923016254605213e-05, "loss": 1.2221, "step": 2856 }, { "epoch": 0.29, "learning_rate": 5.9229623782680563e-05, "loss": 0.5848, "step": 2857 }, { "epoch": 0.29, "learning_rate": 5.922908483330278e-05, "loss": 0.5196, "step": 2858 }, { "epoch": 0.29, "learning_rate": 5.9228545697922237e-05, "loss": 0.2584, "step": 2859 }, { "epoch": 0.29, "learning_rate": 5.922800637654234e-05, "loss": 0.6905, "step": 2860 }, { "epoch": 0.29, "learning_rate": 5.922746686916652e-05, "loss": 1.3144, "step": 2861 }, { "epoch": 0.29, "learning_rate": 5.922692717579824e-05, "loss": 0.7933, "step": 2862 }, { "epoch": 0.29, "learning_rate": 5.9226387296440895e-05, "loss": 0.7884, "step": 2863 }, { "epoch": 0.29, "learning_rate": 5.922584723109796e-05, "loss": 0.8868, "step": 2864 }, { "epoch": 0.29, "learning_rate": 5.9225306979772836e-05, "loss": 0.4059, "step": 2865 }, { "epoch": 0.29, "learning_rate": 5.922476654246898e-05, "loss": 0.6388, "step": 2866 }, { "epoch": 0.29, "learning_rate": 5.922422591918983e-05, "loss": 0.6643, "step": 2867 }, { "epoch": 0.29, "learning_rate": 5.922368510993882e-05, "loss": 0.5397, "step": 2868 }, { "epoch": 0.29, "learning_rate": 5.922314411471939e-05, "loss": 1.9526, "step": 2869 }, { "epoch": 0.29, "learning_rate": 5.9222602933535e-05, "loss": 1.3379, "step": 2870 }, { "epoch": 0.29, "learning_rate": 5.9222061566389075e-05, "loss": 1.0582, "step": 2871 }, { "epoch": 0.29, "learning_rate": 5.9221520013285065e-05, "loss": 1.4439, "step": 2872 }, { "epoch": 0.29, "learning_rate": 5.922097827422642e-05, "loss": 0.1558, "step": 2873 }, { "epoch": 0.29, "learning_rate": 5.9220436349216585e-05, "loss": 0.5585, "step": 2874 }, { "epoch": 0.29, "learning_rate": 5.921989423825901e-05, "loss": 0.9535, "step": 2875 }, { "epoch": 0.29, "learning_rate": 5.9219351941357144e-05, "loss": 0.5393, "step": 2876 }, { "epoch": 0.29, "learning_rate": 5.921880945851444e-05, "loss": 0.647, "step": 2877 }, { "epoch": 0.29, "learning_rate": 5.921826678973435e-05, "loss": 0.4546, "step": 2878 }, { "epoch": 0.29, "learning_rate": 5.921772393502032e-05, "loss": 0.3729, "step": 2879 }, { "epoch": 0.29, "learning_rate": 5.921718089437581e-05, "loss": 0.7666, "step": 2880 }, { "epoch": 0.29, "learning_rate": 5.921663766780427e-05, "loss": 0.7895, "step": 2881 }, { "epoch": 0.29, "learning_rate": 5.921609425530918e-05, "loss": 0.2832, "step": 2882 }, { "epoch": 0.29, "learning_rate": 5.9215550656893965e-05, "loss": 0.7428, "step": 2883 }, { "epoch": 0.29, "learning_rate": 5.9215006872562105e-05, "loss": 0.4901, "step": 2884 }, { "epoch": 0.29, "learning_rate": 5.9214462902317056e-05, "loss": 0.5311, "step": 2885 }, { "epoch": 0.29, "learning_rate": 5.9213918746162275e-05, "loss": 0.4672, "step": 2886 }, { "epoch": 0.29, "learning_rate": 5.921337440410123e-05, "loss": 0.4962, "step": 2887 }, { "epoch": 0.29, "learning_rate": 5.921282987613739e-05, "loss": 0.714, "step": 2888 }, { "epoch": 0.29, "learning_rate": 5.92122851622742e-05, "loss": 0.9423, "step": 2889 }, { "epoch": 0.29, "learning_rate": 5.921174026251515e-05, "loss": 0.9459, "step": 2890 }, { "epoch": 0.29, "learning_rate": 5.92111951768637e-05, "loss": 0.7184, "step": 2891 }, { "epoch": 0.29, "learning_rate": 5.9210649905323314e-05, "loss": 0.9499, "step": 2892 }, { "epoch": 0.29, "learning_rate": 5.921010444789746e-05, "loss": 0.5797, "step": 2893 }, { "epoch": 0.29, "learning_rate": 5.920955880458962e-05, "loss": 0.6779, "step": 2894 }, { "epoch": 0.29, "learning_rate": 5.920901297540326e-05, "loss": 1.0592, "step": 2895 }, { "epoch": 0.29, "learning_rate": 5.920846696034184e-05, "loss": 0.9393, "step": 2896 }, { "epoch": 0.29, "learning_rate": 5.920792075940887e-05, "loss": 0.5555, "step": 2897 }, { "epoch": 0.29, "learning_rate": 5.920737437260778e-05, "loss": 0.3911, "step": 2898 }, { "epoch": 0.29, "learning_rate": 5.920682779994208e-05, "loss": 0.8527, "step": 2899 }, { "epoch": 0.29, "learning_rate": 5.920628104141524e-05, "loss": 0.476, "step": 2900 }, { "epoch": 0.29, "learning_rate": 5.920573409703074e-05, "loss": 1.3307, "step": 2901 }, { "epoch": 0.29, "learning_rate": 5.9205186966792065e-05, "loss": 0.5184, "step": 2902 }, { "epoch": 0.29, "learning_rate": 5.920463965070268e-05, "loss": 0.4153, "step": 2903 }, { "epoch": 0.29, "learning_rate": 5.920409214876608e-05, "loss": 0.392, "step": 2904 }, { "epoch": 0.29, "learning_rate": 5.9203544460985745e-05, "loss": 0.7283, "step": 2905 }, { "epoch": 0.3, "learning_rate": 5.920299658736516e-05, "loss": 0.7322, "step": 2906 }, { "epoch": 0.3, "learning_rate": 5.920244852790782e-05, "loss": 0.5474, "step": 2907 }, { "epoch": 0.3, "learning_rate": 5.920190028261721e-05, "loss": 0.4088, "step": 2908 }, { "epoch": 0.3, "learning_rate": 5.920135185149681e-05, "loss": 1.005, "step": 2909 }, { "epoch": 0.3, "learning_rate": 5.9200803234550124e-05, "loss": 0.8917, "step": 2910 }, { "epoch": 0.3, "learning_rate": 5.9200254431780626e-05, "loss": 1.3583, "step": 2911 }, { "epoch": 0.3, "learning_rate": 5.919970544319182e-05, "loss": 0.9509, "step": 2912 }, { "epoch": 0.3, "learning_rate": 5.91991562687872e-05, "loss": 0.7563, "step": 2913 }, { "epoch": 0.3, "learning_rate": 5.919860690857025e-05, "loss": 1.3739, "step": 2914 }, { "epoch": 0.3, "learning_rate": 5.919805736254447e-05, "loss": 0.9318, "step": 2915 }, { "epoch": 0.3, "learning_rate": 5.919750763071337e-05, "loss": 1.1346, "step": 2916 }, { "epoch": 0.3, "learning_rate": 5.919695771308044e-05, "loss": 0.6463, "step": 2917 }, { "epoch": 0.3, "learning_rate": 5.919640760964917e-05, "loss": 1.083, "step": 2918 }, { "epoch": 0.3, "learning_rate": 5.919585732042307e-05, "loss": 0.6667, "step": 2919 }, { "epoch": 0.3, "learning_rate": 5.919530684540565e-05, "loss": 0.541, "step": 2920 }, { "epoch": 0.3, "learning_rate": 5.919475618460039e-05, "loss": 1.3594, "step": 2921 }, { "epoch": 0.3, "learning_rate": 5.919420533801082e-05, "loss": 1.1744, "step": 2922 }, { "epoch": 0.3, "learning_rate": 5.919365430564043e-05, "loss": 0.3907, "step": 2923 }, { "epoch": 0.3, "learning_rate": 5.9193103087492726e-05, "loss": 1.0834, "step": 2924 }, { "epoch": 0.3, "learning_rate": 5.919255168357122e-05, "loss": 1.0398, "step": 2925 }, { "epoch": 0.3, "learning_rate": 5.9192000093879425e-05, "loss": 0.532, "step": 2926 }, { "epoch": 0.3, "learning_rate": 5.919144831842084e-05, "loss": 1.3751, "step": 2927 }, { "epoch": 0.3, "learning_rate": 5.919089635719899e-05, "loss": 1.4594, "step": 2928 }, { "epoch": 0.3, "learning_rate": 5.919034421021738e-05, "loss": 0.7526, "step": 2929 }, { "epoch": 0.3, "learning_rate": 5.918979187747952e-05, "loss": 1.0144, "step": 2930 }, { "epoch": 0.3, "learning_rate": 5.918923935898893e-05, "loss": 0.5603, "step": 2931 }, { "epoch": 0.3, "learning_rate": 5.918868665474912e-05, "loss": 0.7797, "step": 2932 }, { "epoch": 0.3, "learning_rate": 5.9188133764763626e-05, "loss": 1.1483, "step": 2933 }, { "epoch": 0.3, "learning_rate": 5.918758068903594e-05, "loss": 0.8425, "step": 2934 }, { "epoch": 0.3, "learning_rate": 5.9187027427569596e-05, "loss": 1.1235, "step": 2935 }, { "epoch": 0.3, "learning_rate": 5.918647398036811e-05, "loss": 1.4581, "step": 2936 }, { "epoch": 0.3, "learning_rate": 5.918592034743501e-05, "loss": 0.5403, "step": 2937 }, { "epoch": 0.3, "learning_rate": 5.918536652877382e-05, "loss": 0.3219, "step": 2938 }, { "epoch": 0.3, "learning_rate": 5.9184812524388064e-05, "loss": 0.2147, "step": 2939 }, { "epoch": 0.3, "learning_rate": 5.9184258334281255e-05, "loss": 0.1997, "step": 2940 }, { "epoch": 0.3, "learning_rate": 5.918370395845693e-05, "loss": 1.1778, "step": 2941 }, { "epoch": 0.3, "learning_rate": 5.9183149396918616e-05, "loss": 0.455, "step": 2942 }, { "epoch": 0.3, "learning_rate": 5.9182594649669834e-05, "loss": 0.9209, "step": 2943 }, { "epoch": 0.3, "learning_rate": 5.9182039716714127e-05, "loss": 1.2356, "step": 2944 }, { "epoch": 0.3, "learning_rate": 5.9181484598055024e-05, "loss": 0.9291, "step": 2945 }, { "epoch": 0.3, "learning_rate": 5.9180929293696044e-05, "loss": 0.749, "step": 2946 }, { "epoch": 0.3, "learning_rate": 5.918037380364074e-05, "loss": 1.2336, "step": 2947 }, { "epoch": 0.3, "learning_rate": 5.917981812789264e-05, "loss": 0.4786, "step": 2948 }, { "epoch": 0.3, "learning_rate": 5.917926226645528e-05, "loss": 1.2246, "step": 2949 }, { "epoch": 0.3, "learning_rate": 5.917870621933219e-05, "loss": 0.7015, "step": 2950 }, { "epoch": 0.3, "learning_rate": 5.91781499865269e-05, "loss": 0.9072, "step": 2951 }, { "epoch": 0.3, "learning_rate": 5.917759356804298e-05, "loss": 1.2744, "step": 2952 }, { "epoch": 0.3, "learning_rate": 5.917703696388395e-05, "loss": 0.4463, "step": 2953 }, { "epoch": 0.3, "learning_rate": 5.9176480174053355e-05, "loss": 0.4358, "step": 2954 }, { "epoch": 0.3, "learning_rate": 5.917592319855474e-05, "loss": 0.4953, "step": 2955 }, { "epoch": 0.3, "learning_rate": 5.9175366037391656e-05, "loss": 0.8504, "step": 2956 }, { "epoch": 0.3, "learning_rate": 5.9174808690567635e-05, "loss": 1.3182, "step": 2957 }, { "epoch": 0.3, "learning_rate": 5.917425115808624e-05, "loss": 1.1662, "step": 2958 }, { "epoch": 0.3, "learning_rate": 5.9173693439950986e-05, "loss": 1.1064, "step": 2959 }, { "epoch": 0.3, "learning_rate": 5.9173135536165466e-05, "loss": 0.4998, "step": 2960 }, { "epoch": 0.3, "learning_rate": 5.91725774467332e-05, "loss": 0.6833, "step": 2961 }, { "epoch": 0.3, "learning_rate": 5.917201917165775e-05, "loss": 1.3538, "step": 2962 }, { "epoch": 0.3, "learning_rate": 5.917146071094267e-05, "loss": 0.2013, "step": 2963 }, { "epoch": 0.3, "learning_rate": 5.917090206459152e-05, "loss": 0.3491, "step": 2964 }, { "epoch": 0.3, "learning_rate": 5.917034323260783e-05, "loss": 1.3042, "step": 2965 }, { "epoch": 0.3, "learning_rate": 5.916978421499518e-05, "loss": 0.4853, "step": 2966 }, { "epoch": 0.3, "learning_rate": 5.916922501175712e-05, "loss": 0.4612, "step": 2967 }, { "epoch": 0.3, "learning_rate": 5.916866562289721e-05, "loss": 1.3429, "step": 2968 }, { "epoch": 0.3, "learning_rate": 5.916810604841901e-05, "loss": 0.2759, "step": 2969 }, { "epoch": 0.3, "learning_rate": 5.916754628832607e-05, "loss": 0.212, "step": 2970 }, { "epoch": 0.3, "learning_rate": 5.916698634262197e-05, "loss": 0.5836, "step": 2971 }, { "epoch": 0.3, "learning_rate": 5.916642621131026e-05, "loss": 1.2393, "step": 2972 }, { "epoch": 0.3, "learning_rate": 5.916586589439451e-05, "loss": 1.1625, "step": 2973 }, { "epoch": 0.3, "learning_rate": 5.916530539187828e-05, "loss": 1.1571, "step": 2974 }, { "epoch": 0.3, "learning_rate": 5.916474470376515e-05, "loss": 0.8985, "step": 2975 }, { "epoch": 0.3, "learning_rate": 5.916418383005868e-05, "loss": 1.3267, "step": 2976 }, { "epoch": 0.3, "learning_rate": 5.916362277076243e-05, "loss": 0.4606, "step": 2977 }, { "epoch": 0.3, "learning_rate": 5.916306152587998e-05, "loss": 0.3659, "step": 2978 }, { "epoch": 0.3, "learning_rate": 5.916250009541491e-05, "loss": 0.5498, "step": 2979 }, { "epoch": 0.3, "learning_rate": 5.9161938479370786e-05, "loss": 0.9286, "step": 2980 }, { "epoch": 0.3, "learning_rate": 5.916137667775116e-05, "loss": 0.9443, "step": 2981 }, { "epoch": 0.3, "learning_rate": 5.916081469055964e-05, "loss": 0.6, "step": 2982 }, { "epoch": 0.3, "learning_rate": 5.916025251779979e-05, "loss": 0.475, "step": 2983 }, { "epoch": 0.3, "learning_rate": 5.9159690159475184e-05, "loss": 1.767, "step": 2984 }, { "epoch": 0.3, "learning_rate": 5.9159127615589395e-05, "loss": 1.5092, "step": 2985 }, { "epoch": 0.3, "learning_rate": 5.915856488614602e-05, "loss": 0.8064, "step": 2986 }, { "epoch": 0.3, "learning_rate": 5.915800197114863e-05, "loss": 0.9938, "step": 2987 }, { "epoch": 0.3, "learning_rate": 5.91574388706008e-05, "loss": 0.849, "step": 2988 }, { "epoch": 0.3, "learning_rate": 5.9156875584506125e-05, "loss": 1.6894, "step": 2989 }, { "epoch": 0.3, "learning_rate": 5.915631211286819e-05, "loss": 1.5108, "step": 2990 }, { "epoch": 0.3, "learning_rate": 5.915574845569058e-05, "loss": 1.2392, "step": 2991 }, { "epoch": 0.3, "learning_rate": 5.915518461297687e-05, "loss": 0.4963, "step": 2992 }, { "epoch": 0.3, "learning_rate": 5.915462058473066e-05, "loss": 0.9816, "step": 2993 }, { "epoch": 0.3, "learning_rate": 5.9154056370955534e-05, "loss": 1.2292, "step": 2994 }, { "epoch": 0.3, "learning_rate": 5.915349197165507e-05, "loss": 0.9534, "step": 2995 }, { "epoch": 0.3, "learning_rate": 5.9152927386832894e-05, "loss": 0.9726, "step": 2996 }, { "epoch": 0.3, "learning_rate": 5.915236261649258e-05, "loss": 1.3728, "step": 2997 }, { "epoch": 0.3, "learning_rate": 5.915179766063771e-05, "loss": 0.6356, "step": 2998 }, { "epoch": 0.3, "learning_rate": 5.9151232519271886e-05, "loss": 1.3041, "step": 2999 }, { "epoch": 0.3, "learning_rate": 5.9150667192398715e-05, "loss": 0.2374, "step": 3000 }, { "epoch": 0.3, "learning_rate": 5.915010168002179e-05, "loss": 1.088, "step": 3001 }, { "epoch": 0.3, "learning_rate": 5.91495359821447e-05, "loss": 0.8862, "step": 3002 }, { "epoch": 0.3, "learning_rate": 5.914897009877106e-05, "loss": 1.3785, "step": 3003 }, { "epoch": 0.31, "learning_rate": 5.914840402990445e-05, "loss": 0.5526, "step": 3004 }, { "epoch": 0.31, "learning_rate": 5.9147837775548494e-05, "loss": 0.4009, "step": 3005 }, { "epoch": 0.31, "learning_rate": 5.914727133570679e-05, "loss": 1.3364, "step": 3006 }, { "epoch": 0.31, "learning_rate": 5.9146704710382936e-05, "loss": 1.1238, "step": 3007 }, { "epoch": 0.31, "learning_rate": 5.914613789958054e-05, "loss": 0.5144, "step": 3008 }, { "epoch": 0.31, "learning_rate": 5.914557090330321e-05, "loss": 0.6594, "step": 3009 }, { "epoch": 0.31, "learning_rate": 5.9145003721554556e-05, "loss": 0.9201, "step": 3010 }, { "epoch": 0.31, "learning_rate": 5.914443635433818e-05, "loss": 0.1461, "step": 3011 }, { "epoch": 0.31, "learning_rate": 5.914386880165771e-05, "loss": 1.1804, "step": 3012 }, { "epoch": 0.31, "learning_rate": 5.914330106351674e-05, "loss": 1.3649, "step": 3013 }, { "epoch": 0.31, "learning_rate": 5.914273313991889e-05, "loss": 1.2584, "step": 3014 }, { "epoch": 0.31, "learning_rate": 5.914216503086777e-05, "loss": 1.1371, "step": 3015 }, { "epoch": 0.31, "learning_rate": 5.9141596736366994e-05, "loss": 1.1143, "step": 3016 }, { "epoch": 0.31, "learning_rate": 5.914102825642019e-05, "loss": 0.8118, "step": 3017 }, { "epoch": 0.31, "learning_rate": 5.914045959103097e-05, "loss": 0.5805, "step": 3018 }, { "epoch": 0.31, "learning_rate": 5.913989074020294e-05, "loss": 0.4873, "step": 3019 }, { "epoch": 0.31, "learning_rate": 5.913932170393974e-05, "loss": 1.5788, "step": 3020 }, { "epoch": 0.31, "learning_rate": 5.913875248224498e-05, "loss": 0.9594, "step": 3021 }, { "epoch": 0.31, "learning_rate": 5.913818307512228e-05, "loss": 1.2666, "step": 3022 }, { "epoch": 0.31, "learning_rate": 5.913761348257527e-05, "loss": 1.2405, "step": 3023 }, { "epoch": 0.31, "learning_rate": 5.9137043704607566e-05, "loss": 1.2411, "step": 3024 }, { "epoch": 0.31, "learning_rate": 5.9136473741222816e-05, "loss": 1.494, "step": 3025 }, { "epoch": 0.31, "learning_rate": 5.913590359242462e-05, "loss": 1.2451, "step": 3026 }, { "epoch": 0.31, "learning_rate": 5.913533325821662e-05, "loss": 0.7439, "step": 3027 }, { "epoch": 0.31, "learning_rate": 5.913476273860244e-05, "loss": 1.1388, "step": 3028 }, { "epoch": 0.31, "learning_rate": 5.9134192033585715e-05, "loss": 0.7016, "step": 3029 }, { "epoch": 0.31, "learning_rate": 5.913362114317008e-05, "loss": 1.0882, "step": 3030 }, { "epoch": 0.31, "learning_rate": 5.913305006735916e-05, "loss": 1.5777, "step": 3031 }, { "epoch": 0.31, "learning_rate": 5.9132478806156584e-05, "loss": 1.4504, "step": 3032 }, { "epoch": 0.31, "learning_rate": 5.913190735956601e-05, "loss": 0.6834, "step": 3033 }, { "epoch": 0.31, "learning_rate": 5.9131335727591046e-05, "loss": 0.3708, "step": 3034 }, { "epoch": 0.31, "learning_rate": 5.913076391023535e-05, "loss": 1.2969, "step": 3035 }, { "epoch": 0.31, "learning_rate": 5.913019190750256e-05, "loss": 0.5318, "step": 3036 }, { "epoch": 0.31, "learning_rate": 5.9129619719396306e-05, "loss": 0.6766, "step": 3037 }, { "epoch": 0.31, "learning_rate": 5.9129047345920224e-05, "loss": 0.8981, "step": 3038 }, { "epoch": 0.31, "learning_rate": 5.912847478707798e-05, "loss": 0.8921, "step": 3039 }, { "epoch": 0.31, "learning_rate": 5.9127902042873195e-05, "loss": 0.7779, "step": 3040 }, { "epoch": 0.31, "learning_rate": 5.912732911330952e-05, "loss": 0.8777, "step": 3041 }, { "epoch": 0.31, "learning_rate": 5.9126755998390614e-05, "loss": 0.6166, "step": 3042 }, { "epoch": 0.31, "learning_rate": 5.9126182698120104e-05, "loss": 0.5984, "step": 3043 }, { "epoch": 0.31, "learning_rate": 5.912560921250165e-05, "loss": 0.4443, "step": 3044 }, { "epoch": 0.31, "learning_rate": 5.9125035541538897e-05, "loss": 1.0342, "step": 3045 }, { "epoch": 0.31, "learning_rate": 5.91244616852355e-05, "loss": 1.4234, "step": 3046 }, { "epoch": 0.31, "learning_rate": 5.912388764359511e-05, "loss": 0.6511, "step": 3047 }, { "epoch": 0.31, "learning_rate": 5.9123313416621374e-05, "loss": 0.1923, "step": 3048 }, { "epoch": 0.31, "learning_rate": 5.912273900431796e-05, "loss": 1.0033, "step": 3049 }, { "epoch": 0.31, "learning_rate": 5.912216440668851e-05, "loss": 0.8234, "step": 3050 }, { "epoch": 0.31, "learning_rate": 5.9121589623736675e-05, "loss": 0.364, "step": 3051 }, { "epoch": 0.31, "learning_rate": 5.912101465546613e-05, "loss": 1.0198, "step": 3052 }, { "epoch": 0.31, "learning_rate": 5.9120439501880524e-05, "loss": 0.9399, "step": 3053 }, { "epoch": 0.31, "learning_rate": 5.9119864162983516e-05, "loss": 1.3259, "step": 3054 }, { "epoch": 0.31, "learning_rate": 5.911928863877878e-05, "loss": 1.3955, "step": 3055 }, { "epoch": 0.31, "learning_rate": 5.911871292926996e-05, "loss": 1.1932, "step": 3056 }, { "epoch": 0.31, "learning_rate": 5.911813703446073e-05, "loss": 1.1828, "step": 3057 }, { "epoch": 0.31, "learning_rate": 5.9117560954354754e-05, "loss": 0.4412, "step": 3058 }, { "epoch": 0.31, "learning_rate": 5.9116984688955694e-05, "loss": 1.1149, "step": 3059 }, { "epoch": 0.31, "learning_rate": 5.911640823826722e-05, "loss": 0.3981, "step": 3060 }, { "epoch": 0.31, "learning_rate": 5.911583160229301e-05, "loss": 0.6181, "step": 3061 }, { "epoch": 0.31, "learning_rate": 5.9115254781036714e-05, "loss": 1.0785, "step": 3062 }, { "epoch": 0.31, "learning_rate": 5.911467777450201e-05, "loss": 1.1694, "step": 3063 }, { "epoch": 0.31, "learning_rate": 5.911410058269258e-05, "loss": 0.5186, "step": 3064 }, { "epoch": 0.31, "learning_rate": 5.911352320561209e-05, "loss": 1.2296, "step": 3065 }, { "epoch": 0.31, "learning_rate": 5.911294564326421e-05, "loss": 0.2419, "step": 3066 }, { "epoch": 0.31, "learning_rate": 5.911236789565262e-05, "loss": 0.6656, "step": 3067 }, { "epoch": 0.31, "learning_rate": 5.9111789962781e-05, "loss": 1.0874, "step": 3068 }, { "epoch": 0.31, "learning_rate": 5.9111211844653016e-05, "loss": 0.8867, "step": 3069 }, { "epoch": 0.31, "learning_rate": 5.9110633541272356e-05, "loss": 0.4455, "step": 3070 }, { "epoch": 0.31, "learning_rate": 5.91100550526427e-05, "loss": 1.3425, "step": 3071 }, { "epoch": 0.31, "learning_rate": 5.9109476378767725e-05, "loss": 1.2011, "step": 3072 }, { "epoch": 0.31, "learning_rate": 5.9108897519651126e-05, "loss": 1.4782, "step": 3073 }, { "epoch": 0.31, "learning_rate": 5.910831847529657e-05, "loss": 1.5563, "step": 3074 }, { "epoch": 0.31, "learning_rate": 5.910773924570774e-05, "loss": 0.5571, "step": 3075 }, { "epoch": 0.31, "learning_rate": 5.910715983088834e-05, "loss": 1.087, "step": 3076 }, { "epoch": 0.31, "learning_rate": 5.9106580230842044e-05, "loss": 1.0741, "step": 3077 }, { "epoch": 0.31, "learning_rate": 5.9106000445572554e-05, "loss": 0.8211, "step": 3078 }, { "epoch": 0.31, "learning_rate": 5.910542047508354e-05, "loss": 1.5822, "step": 3079 }, { "epoch": 0.31, "learning_rate": 5.91048403193787e-05, "loss": 0.721, "step": 3080 }, { "epoch": 0.31, "learning_rate": 5.910425997846174e-05, "loss": 1.2061, "step": 3081 }, { "epoch": 0.31, "learning_rate": 5.9103679452336326e-05, "loss": 0.5612, "step": 3082 }, { "epoch": 0.31, "learning_rate": 5.9103098741006175e-05, "loss": 1.1252, "step": 3083 }, { "epoch": 0.31, "learning_rate": 5.910251784447498e-05, "loss": 1.2018, "step": 3084 }, { "epoch": 0.31, "learning_rate": 5.910193676274643e-05, "loss": 1.0212, "step": 3085 }, { "epoch": 0.31, "learning_rate": 5.910135549582422e-05, "loss": 0.9368, "step": 3086 }, { "epoch": 0.31, "learning_rate": 5.910077404371206e-05, "loss": 0.829, "step": 3087 }, { "epoch": 0.31, "learning_rate": 5.9100192406413644e-05, "loss": 1.3476, "step": 3088 }, { "epoch": 0.31, "learning_rate": 5.909961058393267e-05, "loss": 1.1211, "step": 3089 }, { "epoch": 0.31, "learning_rate": 5.909902857627284e-05, "loss": 0.8948, "step": 3090 }, { "epoch": 0.31, "learning_rate": 5.909844638343787e-05, "loss": 1.4558, "step": 3091 }, { "epoch": 0.31, "learning_rate": 5.9097864005431464e-05, "loss": 1.1084, "step": 3092 }, { "epoch": 0.31, "learning_rate": 5.909728144225732e-05, "loss": 1.228, "step": 3093 }, { "epoch": 0.31, "learning_rate": 5.909669869391913e-05, "loss": 1.0005, "step": 3094 }, { "epoch": 0.31, "learning_rate": 5.909611576042062e-05, "loss": 0.7323, "step": 3095 }, { "epoch": 0.31, "learning_rate": 5.909553264176551e-05, "loss": 0.4748, "step": 3096 }, { "epoch": 0.31, "learning_rate": 5.9094949337957496e-05, "loss": 0.217, "step": 3097 }, { "epoch": 0.31, "learning_rate": 5.909436584900029e-05, "loss": 0.3674, "step": 3098 }, { "epoch": 0.31, "learning_rate": 5.90937821748976e-05, "loss": 0.4721, "step": 3099 }, { "epoch": 0.31, "learning_rate": 5.9093198315653163e-05, "loss": 1.0575, "step": 3100 }, { "epoch": 0.31, "learning_rate": 5.909261427127067e-05, "loss": 0.5585, "step": 3101 }, { "epoch": 0.31, "learning_rate": 5.909203004175385e-05, "loss": 0.7611, "step": 3102 }, { "epoch": 0.32, "learning_rate": 5.909144562710642e-05, "loss": 0.3245, "step": 3103 }, { "epoch": 0.32, "learning_rate": 5.909086102733209e-05, "loss": 0.7421, "step": 3104 }, { "epoch": 0.32, "learning_rate": 5.909027624243459e-05, "loss": 1.1548, "step": 3105 }, { "epoch": 0.32, "learning_rate": 5.908969127241764e-05, "loss": 1.1382, "step": 3106 }, { "epoch": 0.32, "learning_rate": 5.9089106117284965e-05, "loss": 1.3959, "step": 3107 }, { "epoch": 0.32, "learning_rate": 5.908852077704029e-05, "loss": 0.434, "step": 3108 }, { "epoch": 0.32, "learning_rate": 5.908793525168732e-05, "loss": 1.1544, "step": 3109 }, { "epoch": 0.32, "learning_rate": 5.90873495412298e-05, "loss": 1.088, "step": 3110 }, { "epoch": 0.32, "learning_rate": 5.9086763645671454e-05, "loss": 0.3729, "step": 3111 }, { "epoch": 0.32, "learning_rate": 5.9086177565016005e-05, "loss": 0.1659, "step": 3112 }, { "epoch": 0.32, "learning_rate": 5.9085591299267194e-05, "loss": 0.7377, "step": 3113 }, { "epoch": 0.32, "learning_rate": 5.908500484842874e-05, "loss": 0.8454, "step": 3114 }, { "epoch": 0.32, "learning_rate": 5.908441821250438e-05, "loss": 1.091, "step": 3115 }, { "epoch": 0.32, "learning_rate": 5.908383139149786e-05, "loss": 1.4637, "step": 3116 }, { "epoch": 0.32, "learning_rate": 5.9083244385412884e-05, "loss": 0.3594, "step": 3117 }, { "epoch": 0.32, "learning_rate": 5.9082657194253215e-05, "loss": 0.7178, "step": 3118 }, { "epoch": 0.32, "learning_rate": 5.9082069818022574e-05, "loss": 0.7441, "step": 3119 }, { "epoch": 0.32, "learning_rate": 5.9081482256724706e-05, "loss": 0.648, "step": 3120 }, { "epoch": 0.32, "learning_rate": 5.908089451036335e-05, "loss": 1.4934, "step": 3121 }, { "epoch": 0.32, "learning_rate": 5.9080306578942236e-05, "loss": 1.2529, "step": 3122 }, { "epoch": 0.32, "learning_rate": 5.907971846246512e-05, "loss": 1.028, "step": 3123 }, { "epoch": 0.32, "learning_rate": 5.9079130160935735e-05, "loss": 0.5571, "step": 3124 }, { "epoch": 0.32, "learning_rate": 5.907854167435783e-05, "loss": 0.467, "step": 3125 }, { "epoch": 0.32, "learning_rate": 5.907795300273514e-05, "loss": 1.0174, "step": 3126 }, { "epoch": 0.32, "learning_rate": 5.9077364146071434e-05, "loss": 0.3982, "step": 3127 }, { "epoch": 0.32, "learning_rate": 5.907677510437044e-05, "loss": 0.6749, "step": 3128 }, { "epoch": 0.32, "learning_rate": 5.907618587763591e-05, "loss": 0.26, "step": 3129 }, { "epoch": 0.32, "learning_rate": 5.907559646587159e-05, "loss": 1.3655, "step": 3130 }, { "epoch": 0.32, "learning_rate": 5.9075006869081224e-05, "loss": 0.5134, "step": 3131 }, { "epoch": 0.32, "learning_rate": 5.9074417087268596e-05, "loss": 0.7376, "step": 3132 }, { "epoch": 0.32, "learning_rate": 5.907382712043742e-05, "loss": 0.4792, "step": 3133 }, { "epoch": 0.32, "learning_rate": 5.9073236968591486e-05, "loss": 1.0223, "step": 3134 }, { "epoch": 0.32, "learning_rate": 5.9072646631734526e-05, "loss": 1.7371, "step": 3135 }, { "epoch": 0.32, "learning_rate": 5.90720561098703e-05, "loss": 1.1529, "step": 3136 }, { "epoch": 0.32, "learning_rate": 5.9071465403002565e-05, "loss": 1.8652, "step": 3137 }, { "epoch": 0.32, "learning_rate": 5.907087451113509e-05, "loss": 0.6044, "step": 3138 }, { "epoch": 0.32, "learning_rate": 5.9070283434271626e-05, "loss": 0.9829, "step": 3139 }, { "epoch": 0.32, "learning_rate": 5.906969217241594e-05, "loss": 0.6484, "step": 3140 }, { "epoch": 0.32, "learning_rate": 5.906910072557179e-05, "loss": 0.4032, "step": 3141 }, { "epoch": 0.32, "learning_rate": 5.906850909374294e-05, "loss": 0.6202, "step": 3142 }, { "epoch": 0.32, "learning_rate": 5.9067917276933164e-05, "loss": 1.2752, "step": 3143 }, { "epoch": 0.32, "learning_rate": 5.906732527514621e-05, "loss": 0.5075, "step": 3144 }, { "epoch": 0.32, "learning_rate": 5.906673308838586e-05, "loss": 1.4452, "step": 3145 }, { "epoch": 0.32, "learning_rate": 5.906614071665587e-05, "loss": 1.2531, "step": 3146 }, { "epoch": 0.32, "learning_rate": 5.906554815996003e-05, "loss": 1.2336, "step": 3147 }, { "epoch": 0.32, "learning_rate": 5.906495541830209e-05, "loss": 0.2719, "step": 3148 }, { "epoch": 0.32, "learning_rate": 5.906436249168584e-05, "loss": 0.8961, "step": 3149 }, { "epoch": 0.32, "learning_rate": 5.906376938011504e-05, "loss": 1.121, "step": 3150 }, { "epoch": 0.32, "learning_rate": 5.906317608359346e-05, "loss": 0.5384, "step": 3151 }, { "epoch": 0.32, "learning_rate": 5.906258260212489e-05, "loss": 0.8803, "step": 3152 }, { "epoch": 0.32, "learning_rate": 5.90619889357131e-05, "loss": 0.8505, "step": 3153 }, { "epoch": 0.32, "learning_rate": 5.906139508436187e-05, "loss": 0.5178, "step": 3154 }, { "epoch": 0.32, "learning_rate": 5.906080104807498e-05, "loss": 0.686, "step": 3155 }, { "epoch": 0.32, "learning_rate": 5.90602068268562e-05, "loss": 1.7751, "step": 3156 }, { "epoch": 0.32, "learning_rate": 5.9059612420709324e-05, "loss": 1.4534, "step": 3157 }, { "epoch": 0.32, "learning_rate": 5.905901782963813e-05, "loss": 0.9519, "step": 3158 }, { "epoch": 0.32, "learning_rate": 5.905842305364639e-05, "loss": 1.1705, "step": 3159 }, { "epoch": 0.32, "learning_rate": 5.905782809273791e-05, "loss": 0.7248, "step": 3160 }, { "epoch": 0.32, "learning_rate": 5.905723294691647e-05, "loss": 0.614, "step": 3161 }, { "epoch": 0.32, "learning_rate": 5.905663761618584e-05, "loss": 1.3354, "step": 3162 }, { "epoch": 0.32, "learning_rate": 5.905604210054983e-05, "loss": 0.944, "step": 3163 }, { "epoch": 0.32, "learning_rate": 5.905544640001223e-05, "loss": 1.4183, "step": 3164 }, { "epoch": 0.32, "learning_rate": 5.9054850514576804e-05, "loss": 0.1765, "step": 3165 }, { "epoch": 0.32, "learning_rate": 5.905425444424738e-05, "loss": 0.5379, "step": 3166 }, { "epoch": 0.32, "learning_rate": 5.905365818902773e-05, "loss": 1.2533, "step": 3167 }, { "epoch": 0.32, "learning_rate": 5.9053061748921635e-05, "loss": 0.7301, "step": 3168 }, { "epoch": 0.32, "learning_rate": 5.905246512393293e-05, "loss": 0.6845, "step": 3169 }, { "epoch": 0.32, "learning_rate": 5.905186831406538e-05, "loss": 1.4804, "step": 3170 }, { "epoch": 0.32, "learning_rate": 5.905127131932279e-05, "loss": 0.624, "step": 3171 }, { "epoch": 0.32, "learning_rate": 5.905067413970896e-05, "loss": 0.5738, "step": 3172 }, { "epoch": 0.32, "learning_rate": 5.9050076775227694e-05, "loss": 1.8128, "step": 3173 }, { "epoch": 0.32, "learning_rate": 5.904947922588278e-05, "loss": 0.9165, "step": 3174 }, { "epoch": 0.32, "learning_rate": 5.904888149167805e-05, "loss": 0.1917, "step": 3175 }, { "epoch": 0.32, "learning_rate": 5.904828357261728e-05, "loss": 0.6092, "step": 3176 }, { "epoch": 0.32, "learning_rate": 5.9047685468704285e-05, "loss": 1.4007, "step": 3177 }, { "epoch": 0.32, "learning_rate": 5.904708717994286e-05, "loss": 0.712, "step": 3178 }, { "epoch": 0.32, "learning_rate": 5.9046488706336836e-05, "loss": 0.8811, "step": 3179 }, { "epoch": 0.32, "learning_rate": 5.9045890047890005e-05, "loss": 0.323, "step": 3180 }, { "epoch": 0.32, "learning_rate": 5.904529120460617e-05, "loss": 0.9542, "step": 3181 }, { "epoch": 0.32, "learning_rate": 5.9044692176489166e-05, "loss": 0.4523, "step": 3182 }, { "epoch": 0.32, "learning_rate": 5.904409296354278e-05, "loss": 1.2917, "step": 3183 }, { "epoch": 0.32, "learning_rate": 5.904349356577083e-05, "loss": 1.1141, "step": 3184 }, { "epoch": 0.32, "learning_rate": 5.9042893983177154e-05, "loss": 1.5931, "step": 3185 }, { "epoch": 0.32, "learning_rate": 5.9042294215765535e-05, "loss": 1.8676, "step": 3186 }, { "epoch": 0.32, "learning_rate": 5.904169426353981e-05, "loss": 0.9426, "step": 3187 }, { "epoch": 0.32, "learning_rate": 5.9041094126503784e-05, "loss": 0.5241, "step": 3188 }, { "epoch": 0.32, "learning_rate": 5.904049380466129e-05, "loss": 0.5904, "step": 3189 }, { "epoch": 0.32, "learning_rate": 5.903989329801614e-05, "loss": 0.4892, "step": 3190 }, { "epoch": 0.32, "learning_rate": 5.903929260657215e-05, "loss": 0.3979, "step": 3191 }, { "epoch": 0.32, "learning_rate": 5.9038691730333156e-05, "loss": 1.6905, "step": 3192 }, { "epoch": 0.32, "learning_rate": 5.903809066930297e-05, "loss": 0.8208, "step": 3193 }, { "epoch": 0.32, "learning_rate": 5.9037489423485427e-05, "loss": 1.7912, "step": 3194 }, { "epoch": 0.32, "learning_rate": 5.9036887992884345e-05, "loss": 0.9327, "step": 3195 }, { "epoch": 0.32, "learning_rate": 5.9036286377503554e-05, "loss": 0.9727, "step": 3196 }, { "epoch": 0.32, "learning_rate": 5.9035684577346884e-05, "loss": 0.1443, "step": 3197 }, { "epoch": 0.32, "learning_rate": 5.9035082592418154e-05, "loss": 0.599, "step": 3198 }, { "epoch": 0.32, "learning_rate": 5.9034480422721215e-05, "loss": 0.8083, "step": 3199 }, { "epoch": 0.32, "learning_rate": 5.903387806825988e-05, "loss": 1.5149, "step": 3200 }, { "epoch": 0.33, "learning_rate": 5.9033275529037996e-05, "loss": 0.4109, "step": 3201 }, { "epoch": 0.33, "learning_rate": 5.9032672805059386e-05, "loss": 0.6354, "step": 3202 }, { "epoch": 0.33, "learning_rate": 5.90320698963279e-05, "loss": 0.3985, "step": 3203 }, { "epoch": 0.33, "learning_rate": 5.903146680284736e-05, "loss": 1.0418, "step": 3204 }, { "epoch": 0.33, "learning_rate": 5.903086352462161e-05, "loss": 1.495, "step": 3205 }, { "epoch": 0.33, "learning_rate": 5.903026006165448e-05, "loss": 0.6177, "step": 3206 }, { "epoch": 0.33, "learning_rate": 5.902965641394983e-05, "loss": 0.754, "step": 3207 }, { "epoch": 0.33, "learning_rate": 5.9029052581511485e-05, "loss": 0.5661, "step": 3208 }, { "epoch": 0.33, "learning_rate": 5.90284485643433e-05, "loss": 0.2879, "step": 3209 }, { "epoch": 0.33, "learning_rate": 5.90278443624491e-05, "loss": 1.2565, "step": 3210 }, { "epoch": 0.33, "learning_rate": 5.9027239975832746e-05, "loss": 1.1427, "step": 3211 }, { "epoch": 0.33, "learning_rate": 5.902663540449808e-05, "loss": 0.458, "step": 3212 }, { "epoch": 0.33, "learning_rate": 5.902603064844895e-05, "loss": 0.496, "step": 3213 }, { "epoch": 0.33, "learning_rate": 5.9025425707689205e-05, "loss": 0.8751, "step": 3214 }, { "epoch": 0.33, "learning_rate": 5.902482058222268e-05, "loss": 0.6154, "step": 3215 }, { "epoch": 0.33, "learning_rate": 5.902421527205325e-05, "loss": 0.5362, "step": 3216 }, { "epoch": 0.33, "learning_rate": 5.9023609777184754e-05, "loss": 1.0153, "step": 3217 }, { "epoch": 0.33, "learning_rate": 5.902300409762104e-05, "loss": 1.0864, "step": 3218 }, { "epoch": 0.33, "learning_rate": 5.902239823336598e-05, "loss": 0.203, "step": 3219 }, { "epoch": 0.33, "learning_rate": 5.902179218442341e-05, "loss": 1.5964, "step": 3220 }, { "epoch": 0.33, "learning_rate": 5.9021185950797196e-05, "loss": 0.6535, "step": 3221 }, { "epoch": 0.33, "learning_rate": 5.90205795324912e-05, "loss": 0.3327, "step": 3222 }, { "epoch": 0.33, "learning_rate": 5.9019972929509275e-05, "loss": 1.3292, "step": 3223 }, { "epoch": 0.33, "learning_rate": 5.901936614185527e-05, "loss": 0.794, "step": 3224 }, { "epoch": 0.33, "learning_rate": 5.901875916953307e-05, "loss": 1.209, "step": 3225 }, { "epoch": 0.33, "learning_rate": 5.901815201254653e-05, "loss": 0.5191, "step": 3226 }, { "epoch": 0.33, "learning_rate": 5.90175446708995e-05, "loss": 0.669, "step": 3227 }, { "epoch": 0.33, "learning_rate": 5.901693714459586e-05, "loss": 0.9227, "step": 3228 }, { "epoch": 0.33, "learning_rate": 5.9016329433639465e-05, "loss": 1.1135, "step": 3229 }, { "epoch": 0.33, "learning_rate": 5.9015721538034185e-05, "loss": 0.7475, "step": 3230 }, { "epoch": 0.33, "learning_rate": 5.9015113457783895e-05, "loss": 0.7241, "step": 3231 }, { "epoch": 0.33, "learning_rate": 5.901450519289247e-05, "loss": 1.2028, "step": 3232 }, { "epoch": 0.33, "learning_rate": 5.901389674336376e-05, "loss": 0.8122, "step": 3233 }, { "epoch": 0.33, "learning_rate": 5.9013288109201656e-05, "loss": 1.0695, "step": 3234 }, { "epoch": 0.33, "learning_rate": 5.901267929041002e-05, "loss": 0.9754, "step": 3235 }, { "epoch": 0.33, "learning_rate": 5.9012070286992735e-05, "loss": 1.2693, "step": 3236 }, { "epoch": 0.33, "learning_rate": 5.901146109895367e-05, "loss": 0.5327, "step": 3237 }, { "epoch": 0.33, "learning_rate": 5.9010851726296695e-05, "loss": 1.1508, "step": 3238 }, { "epoch": 0.33, "learning_rate": 5.9010242169025704e-05, "loss": 0.2984, "step": 3239 }, { "epoch": 0.33, "learning_rate": 5.9009632427144566e-05, "loss": 0.6157, "step": 3240 }, { "epoch": 0.33, "learning_rate": 5.9009022500657164e-05, "loss": 0.3528, "step": 3241 }, { "epoch": 0.33, "learning_rate": 5.900841238956738e-05, "loss": 0.4593, "step": 3242 }, { "epoch": 0.33, "learning_rate": 5.900780209387909e-05, "loss": 1.2156, "step": 3243 }, { "epoch": 0.33, "learning_rate": 5.900719161359618e-05, "loss": 0.3539, "step": 3244 }, { "epoch": 0.33, "learning_rate": 5.900658094872254e-05, "loss": 1.2163, "step": 3245 }, { "epoch": 0.33, "learning_rate": 5.900597009926206e-05, "loss": 1.167, "step": 3246 }, { "epoch": 0.33, "learning_rate": 5.900535906521861e-05, "loss": 0.4452, "step": 3247 }, { "epoch": 0.33, "learning_rate": 5.90047478465961e-05, "loss": 0.6498, "step": 3248 }, { "epoch": 0.33, "learning_rate": 5.90041364433984e-05, "loss": 0.3825, "step": 3249 }, { "epoch": 0.33, "learning_rate": 5.9003524855629414e-05, "loss": 1.3789, "step": 3250 }, { "epoch": 0.33, "learning_rate": 5.900291308329303e-05, "loss": 0.5492, "step": 3251 }, { "epoch": 0.33, "learning_rate": 5.900230112639314e-05, "loss": 0.3856, "step": 3252 }, { "epoch": 0.33, "learning_rate": 5.900168898493364e-05, "loss": 1.3306, "step": 3253 }, { "epoch": 0.33, "learning_rate": 5.9001076658918416e-05, "loss": 1.0986, "step": 3254 }, { "epoch": 0.33, "learning_rate": 5.9000464148351383e-05, "loss": 0.7819, "step": 3255 }, { "epoch": 0.33, "learning_rate": 5.8999851453236424e-05, "loss": 0.4933, "step": 3256 }, { "epoch": 0.33, "learning_rate": 5.899923857357744e-05, "loss": 0.6041, "step": 3257 }, { "epoch": 0.33, "learning_rate": 5.8998625509378336e-05, "loss": 0.3275, "step": 3258 }, { "epoch": 0.33, "learning_rate": 5.8998012260643015e-05, "loss": 1.1732, "step": 3259 }, { "epoch": 0.33, "learning_rate": 5.899739882737537e-05, "loss": 0.7959, "step": 3260 }, { "epoch": 0.33, "learning_rate": 5.899678520957931e-05, "loss": 0.6991, "step": 3261 }, { "epoch": 0.33, "learning_rate": 5.899617140725874e-05, "loss": 1.1714, "step": 3262 }, { "epoch": 0.33, "learning_rate": 5.899555742041757e-05, "loss": 1.0148, "step": 3263 }, { "epoch": 0.33, "learning_rate": 5.89949432490597e-05, "loss": 0.7366, "step": 3264 }, { "epoch": 0.33, "learning_rate": 5.899432889318904e-05, "loss": 0.5095, "step": 3265 }, { "epoch": 0.33, "learning_rate": 5.8993714352809506e-05, "loss": 1.2302, "step": 3266 }, { "epoch": 0.33, "learning_rate": 5.8993099627925006e-05, "loss": 0.5373, "step": 3267 }, { "epoch": 0.33, "learning_rate": 5.899248471853943e-05, "loss": 1.2059, "step": 3268 }, { "epoch": 0.33, "learning_rate": 5.899186962465673e-05, "loss": 0.5253, "step": 3269 }, { "epoch": 0.33, "learning_rate": 5.89912543462808e-05, "loss": 1.1642, "step": 3270 }, { "epoch": 0.33, "learning_rate": 5.899063888341555e-05, "loss": 0.9109, "step": 3271 }, { "epoch": 0.33, "learning_rate": 5.8990023236064896e-05, "loss": 0.5766, "step": 3272 }, { "epoch": 0.33, "learning_rate": 5.898940740423277e-05, "loss": 1.2439, "step": 3273 }, { "epoch": 0.33, "learning_rate": 5.898879138792309e-05, "loss": 0.1807, "step": 3274 }, { "epoch": 0.33, "learning_rate": 5.898817518713976e-05, "loss": 1.3828, "step": 3275 }, { "epoch": 0.33, "learning_rate": 5.898755880188672e-05, "loss": 0.9087, "step": 3276 }, { "epoch": 0.33, "learning_rate": 5.8986942232167866e-05, "loss": 0.9561, "step": 3277 }, { "epoch": 0.33, "learning_rate": 5.898632547798715e-05, "loss": 1.4656, "step": 3278 }, { "epoch": 0.33, "learning_rate": 5.8985708539348495e-05, "loss": 0.2311, "step": 3279 }, { "epoch": 0.33, "learning_rate": 5.89850914162558e-05, "loss": 1.9623, "step": 3280 }, { "epoch": 0.33, "learning_rate": 5.8984474108713026e-05, "loss": 1.042, "step": 3281 }, { "epoch": 0.33, "learning_rate": 5.8983856616724075e-05, "loss": 0.6724, "step": 3282 }, { "epoch": 0.33, "learning_rate": 5.8983238940292886e-05, "loss": 0.9655, "step": 3283 }, { "epoch": 0.33, "learning_rate": 5.898262107942339e-05, "loss": 0.6492, "step": 3284 }, { "epoch": 0.33, "learning_rate": 5.898200303411953e-05, "loss": 1.3109, "step": 3285 }, { "epoch": 0.33, "learning_rate": 5.898138480438522e-05, "loss": 0.4461, "step": 3286 }, { "epoch": 0.33, "learning_rate": 5.8980766390224396e-05, "loss": 0.761, "step": 3287 }, { "epoch": 0.33, "learning_rate": 5.8980147791641014e-05, "loss": 1.3946, "step": 3288 }, { "epoch": 0.33, "learning_rate": 5.897952900863898e-05, "loss": 0.7648, "step": 3289 }, { "epoch": 0.33, "learning_rate": 5.8978910041222254e-05, "loss": 1.0262, "step": 3290 }, { "epoch": 0.33, "learning_rate": 5.8978290889394776e-05, "loss": 1.2347, "step": 3291 }, { "epoch": 0.33, "learning_rate": 5.897767155316047e-05, "loss": 0.3197, "step": 3292 }, { "epoch": 0.33, "learning_rate": 5.897705203252329e-05, "loss": 0.5322, "step": 3293 }, { "epoch": 0.33, "learning_rate": 5.897643232748718e-05, "loss": 1.2205, "step": 3294 }, { "epoch": 0.33, "learning_rate": 5.8975812438056063e-05, "loss": 0.4588, "step": 3295 }, { "epoch": 0.33, "learning_rate": 5.89751923642339e-05, "loss": 0.9987, "step": 3296 }, { "epoch": 0.33, "learning_rate": 5.897457210602465e-05, "loss": 1.7143, "step": 3297 }, { "epoch": 0.33, "learning_rate": 5.8973951663432234e-05, "loss": 1.0155, "step": 3298 }, { "epoch": 0.33, "learning_rate": 5.897333103646062e-05, "loss": 0.4432, "step": 3299 }, { "epoch": 0.34, "learning_rate": 5.8972710225113743e-05, "loss": 0.36, "step": 3300 }, { "epoch": 0.34, "learning_rate": 5.897208922939556e-05, "loss": 0.1236, "step": 3301 }, { "epoch": 0.34, "learning_rate": 5.8971468049310025e-05, "loss": 1.41, "step": 3302 }, { "epoch": 0.34, "learning_rate": 5.8970846684861083e-05, "loss": 0.8196, "step": 3303 }, { "epoch": 0.34, "learning_rate": 5.8970225136052704e-05, "loss": 1.3211, "step": 3304 }, { "epoch": 0.34, "learning_rate": 5.896960340288883e-05, "loss": 1.0442, "step": 3305 }, { "epoch": 0.34, "learning_rate": 5.896898148537342e-05, "loss": 0.684, "step": 3306 }, { "epoch": 0.34, "learning_rate": 5.896835938351043e-05, "loss": 0.7679, "step": 3307 }, { "epoch": 0.34, "learning_rate": 5.896773709730382e-05, "loss": 0.5166, "step": 3308 }, { "epoch": 0.34, "learning_rate": 5.8967114626757555e-05, "loss": 0.4887, "step": 3309 }, { "epoch": 0.34, "learning_rate": 5.896649197187559e-05, "loss": 0.6267, "step": 3310 }, { "epoch": 0.34, "learning_rate": 5.896586913266188e-05, "loss": 0.8373, "step": 3311 }, { "epoch": 0.34, "learning_rate": 5.896524610912041e-05, "loss": 1.1324, "step": 3312 }, { "epoch": 0.34, "learning_rate": 5.896462290125513e-05, "loss": 0.1647, "step": 3313 }, { "epoch": 0.34, "learning_rate": 5.8963999509070004e-05, "loss": 1.1119, "step": 3314 }, { "epoch": 0.34, "learning_rate": 5.896337593256901e-05, "loss": 0.9758, "step": 3315 }, { "epoch": 0.34, "learning_rate": 5.8962752171756116e-05, "loss": 1.5248, "step": 3316 }, { "epoch": 0.34, "learning_rate": 5.896212822663527e-05, "loss": 0.8213, "step": 3317 }, { "epoch": 0.34, "learning_rate": 5.896150409721047e-05, "loss": 0.8422, "step": 3318 }, { "epoch": 0.34, "learning_rate": 5.896087978348567e-05, "loss": 1.0109, "step": 3319 }, { "epoch": 0.34, "learning_rate": 5.8960255285464844e-05, "loss": 0.7699, "step": 3320 }, { "epoch": 0.34, "learning_rate": 5.895963060315198e-05, "loss": 1.4057, "step": 3321 }, { "epoch": 0.34, "learning_rate": 5.8959005736551034e-05, "loss": 1.0115, "step": 3322 }, { "epoch": 0.34, "learning_rate": 5.8958380685666e-05, "loss": 1.2941, "step": 3323 }, { "epoch": 0.34, "learning_rate": 5.895775545050084e-05, "loss": 0.8889, "step": 3324 }, { "epoch": 0.34, "learning_rate": 5.895713003105955e-05, "loss": 0.9885, "step": 3325 }, { "epoch": 0.34, "learning_rate": 5.895650442734609e-05, "loss": 1.2369, "step": 3326 }, { "epoch": 0.34, "learning_rate": 5.895587863936446e-05, "loss": 0.2925, "step": 3327 }, { "epoch": 0.34, "learning_rate": 5.895525266711863e-05, "loss": 1.4893, "step": 3328 }, { "epoch": 0.34, "learning_rate": 5.8954626510612586e-05, "loss": 1.3968, "step": 3329 }, { "epoch": 0.34, "learning_rate": 5.895400016985032e-05, "loss": 1.0884, "step": 3330 }, { "epoch": 0.34, "learning_rate": 5.895337364483581e-05, "loss": 0.6071, "step": 3331 }, { "epoch": 0.34, "learning_rate": 5.895274693557303e-05, "loss": 0.4025, "step": 3332 }, { "epoch": 0.34, "learning_rate": 5.8952120042065996e-05, "loss": 0.5125, "step": 3333 }, { "epoch": 0.34, "learning_rate": 5.895149296431869e-05, "loss": 0.7736, "step": 3334 }, { "epoch": 0.34, "learning_rate": 5.8950865702335084e-05, "loss": 0.9958, "step": 3335 }, { "epoch": 0.34, "learning_rate": 5.895023825611918e-05, "loss": 0.7308, "step": 3336 }, { "epoch": 0.34, "learning_rate": 5.894961062567499e-05, "loss": 0.4879, "step": 3337 }, { "epoch": 0.34, "learning_rate": 5.894898281100647e-05, "loss": 0.742, "step": 3338 }, { "epoch": 0.34, "learning_rate": 5.894835481211766e-05, "loss": 0.3114, "step": 3339 }, { "epoch": 0.34, "learning_rate": 5.894772662901251e-05, "loss": 1.3347, "step": 3340 }, { "epoch": 0.34, "learning_rate": 5.894709826169505e-05, "loss": 0.2736, "step": 3341 }, { "epoch": 0.34, "learning_rate": 5.894646971016927e-05, "loss": 1.0715, "step": 3342 }, { "epoch": 0.34, "learning_rate": 5.894584097443917e-05, "loss": 1.1559, "step": 3343 }, { "epoch": 0.34, "learning_rate": 5.8945212054508744e-05, "loss": 0.8747, "step": 3344 }, { "epoch": 0.34, "learning_rate": 5.894458295038201e-05, "loss": 1.0809, "step": 3345 }, { "epoch": 0.34, "learning_rate": 5.894395366206294e-05, "loss": 0.6553, "step": 3346 }, { "epoch": 0.34, "learning_rate": 5.894332418955558e-05, "loss": 2.2535, "step": 3347 }, { "epoch": 0.34, "learning_rate": 5.8942694532863905e-05, "loss": 1.229, "step": 3348 }, { "epoch": 0.34, "learning_rate": 5.8942064691991936e-05, "loss": 0.8173, "step": 3349 }, { "epoch": 0.34, "learning_rate": 5.894143466694368e-05, "loss": 0.5435, "step": 3350 }, { "epoch": 0.34, "learning_rate": 5.894080445772314e-05, "loss": 0.5655, "step": 3351 }, { "epoch": 0.34, "learning_rate": 5.894017406433433e-05, "loss": 0.4995, "step": 3352 }, { "epoch": 0.34, "learning_rate": 5.893954348678127e-05, "loss": 0.4179, "step": 3353 }, { "epoch": 0.34, "learning_rate": 5.8938912725067953e-05, "loss": 0.7089, "step": 3354 }, { "epoch": 0.34, "learning_rate": 5.89382817791984e-05, "loss": 1.1017, "step": 3355 }, { "epoch": 0.34, "learning_rate": 5.893765064917664e-05, "loss": 1.301, "step": 3356 }, { "epoch": 0.34, "learning_rate": 5.893701933500668e-05, "loss": 1.9343, "step": 3357 }, { "epoch": 0.34, "learning_rate": 5.8936387836692537e-05, "loss": 1.3105, "step": 3358 }, { "epoch": 0.34, "learning_rate": 5.893575615423823e-05, "loss": 0.4783, "step": 3359 }, { "epoch": 0.34, "learning_rate": 5.893512428764777e-05, "loss": 1.6913, "step": 3360 }, { "epoch": 0.34, "learning_rate": 5.893449223692519e-05, "loss": 0.4916, "step": 3361 }, { "epoch": 0.34, "learning_rate": 5.893386000207452e-05, "loss": 0.7971, "step": 3362 }, { "epoch": 0.34, "learning_rate": 5.8933227583099765e-05, "loss": 0.366, "step": 3363 }, { "epoch": 0.34, "learning_rate": 5.893259498000496e-05, "loss": 0.8438, "step": 3364 }, { "epoch": 0.34, "learning_rate": 5.893196219279411e-05, "loss": 0.428, "step": 3365 }, { "epoch": 0.34, "learning_rate": 5.8931329221471274e-05, "loss": 0.7396, "step": 3366 }, { "epoch": 0.34, "learning_rate": 5.8930696066040466e-05, "loss": 1.1552, "step": 3367 }, { "epoch": 0.34, "learning_rate": 5.893006272650571e-05, "loss": 1.8914, "step": 3368 }, { "epoch": 0.34, "learning_rate": 5.8929429202871036e-05, "loss": 1.8577, "step": 3369 }, { "epoch": 0.34, "learning_rate": 5.892879549514049e-05, "loss": 1.3033, "step": 3370 }, { "epoch": 0.34, "learning_rate": 5.8928161603318085e-05, "loss": 0.5055, "step": 3371 }, { "epoch": 0.34, "learning_rate": 5.892752752740787e-05, "loss": 1.0713, "step": 3372 }, { "epoch": 0.34, "learning_rate": 5.892689326741388e-05, "loss": 0.56, "step": 3373 }, { "epoch": 0.34, "learning_rate": 5.892625882334013e-05, "loss": 0.9094, "step": 3374 }, { "epoch": 0.34, "learning_rate": 5.892562419519069e-05, "loss": 1.6608, "step": 3375 }, { "epoch": 0.34, "learning_rate": 5.892498938296957e-05, "loss": 1.2266, "step": 3376 }, { "epoch": 0.34, "learning_rate": 5.892435438668082e-05, "loss": 0.9834, "step": 3377 }, { "epoch": 0.34, "learning_rate": 5.892371920632848e-05, "loss": 0.6629, "step": 3378 }, { "epoch": 0.34, "learning_rate": 5.89230838419166e-05, "loss": 1.7338, "step": 3379 }, { "epoch": 0.34, "learning_rate": 5.8922448293449226e-05, "loss": 0.7707, "step": 3380 }, { "epoch": 0.34, "learning_rate": 5.8921812560930386e-05, "loss": 0.8709, "step": 3381 }, { "epoch": 0.34, "learning_rate": 5.892117664436413e-05, "loss": 1.7744, "step": 3382 }, { "epoch": 0.34, "learning_rate": 5.892054054375451e-05, "loss": 1.0392, "step": 3383 }, { "epoch": 0.34, "learning_rate": 5.891990425910557e-05, "loss": 1.0236, "step": 3384 }, { "epoch": 0.34, "learning_rate": 5.8919267790421365e-05, "loss": 0.8636, "step": 3385 }, { "epoch": 0.34, "learning_rate": 5.891863113770595e-05, "loss": 1.4785, "step": 3386 }, { "epoch": 0.34, "learning_rate": 5.891799430096335e-05, "loss": 0.4754, "step": 3387 }, { "epoch": 0.34, "learning_rate": 5.891735728019765e-05, "loss": 0.5487, "step": 3388 }, { "epoch": 0.34, "learning_rate": 5.8916720075412874e-05, "loss": 0.7881, "step": 3389 }, { "epoch": 0.34, "learning_rate": 5.891608268661311e-05, "loss": 1.1368, "step": 3390 }, { "epoch": 0.34, "learning_rate": 5.891544511380238e-05, "loss": 1.1579, "step": 3391 }, { "epoch": 0.34, "learning_rate": 5.891480735698476e-05, "loss": 0.4639, "step": 3392 }, { "epoch": 0.34, "learning_rate": 5.891416941616431e-05, "loss": 0.9145, "step": 3393 }, { "epoch": 0.34, "learning_rate": 5.891353129134509e-05, "loss": 1.0996, "step": 3394 }, { "epoch": 0.34, "learning_rate": 5.891289298253114e-05, "loss": 0.6822, "step": 3395 }, { "epoch": 0.34, "learning_rate": 5.891225448972655e-05, "loss": 1.0082, "step": 3396 }, { "epoch": 0.34, "learning_rate": 5.891161581293537e-05, "loss": 1.2106, "step": 3397 }, { "epoch": 0.35, "learning_rate": 5.891097695216166e-05, "loss": 1.1716, "step": 3398 }, { "epoch": 0.35, "learning_rate": 5.89103379074095e-05, "loss": 1.0257, "step": 3399 }, { "epoch": 0.35, "learning_rate": 5.8909698678682944e-05, "loss": 0.741, "step": 3400 }, { "epoch": 0.35, "learning_rate": 5.890905926598606e-05, "loss": 1.56, "step": 3401 }, { "epoch": 0.35, "learning_rate": 5.8908419669322926e-05, "loss": 0.8407, "step": 3402 }, { "epoch": 0.35, "learning_rate": 5.8907779888697605e-05, "loss": 0.6922, "step": 3403 }, { "epoch": 0.35, "learning_rate": 5.890713992411417e-05, "loss": 0.2846, "step": 3404 }, { "epoch": 0.35, "learning_rate": 5.8906499775576685e-05, "loss": 0.8283, "step": 3405 }, { "epoch": 0.35, "learning_rate": 5.8905859443089246e-05, "loss": 1.2255, "step": 3406 }, { "epoch": 0.35, "learning_rate": 5.89052189266559e-05, "loss": 0.4036, "step": 3407 }, { "epoch": 0.35, "learning_rate": 5.890457822628075e-05, "loss": 1.1326, "step": 3408 }, { "epoch": 0.35, "learning_rate": 5.890393734196785e-05, "loss": 0.8437, "step": 3409 }, { "epoch": 0.35, "learning_rate": 5.89032962737213e-05, "loss": 0.6786, "step": 3410 }, { "epoch": 0.35, "learning_rate": 5.8902655021545155e-05, "loss": 0.9078, "step": 3411 }, { "epoch": 0.35, "learning_rate": 5.890201358544352e-05, "loss": 0.9606, "step": 3412 }, { "epoch": 0.35, "learning_rate": 5.890137196542045e-05, "loss": 0.2947, "step": 3413 }, { "epoch": 0.35, "learning_rate": 5.8900730161480054e-05, "loss": 1.3145, "step": 3414 }, { "epoch": 0.35, "learning_rate": 5.89000881736264e-05, "loss": 0.616, "step": 3415 }, { "epoch": 0.35, "learning_rate": 5.889944600186359e-05, "loss": 1.4381, "step": 3416 }, { "epoch": 0.35, "learning_rate": 5.88988036461957e-05, "loss": 0.4314, "step": 3417 }, { "epoch": 0.35, "learning_rate": 5.88981611066268e-05, "loss": 1.0807, "step": 3418 }, { "epoch": 0.35, "learning_rate": 5.8897518383161004e-05, "loss": 0.3365, "step": 3419 }, { "epoch": 0.35, "learning_rate": 5.88968754758024e-05, "loss": 1.0352, "step": 3420 }, { "epoch": 0.35, "learning_rate": 5.889623238455507e-05, "loss": 0.2433, "step": 3421 }, { "epoch": 0.35, "learning_rate": 5.8895589109423107e-05, "loss": 0.8161, "step": 3422 }, { "epoch": 0.35, "learning_rate": 5.8894945650410614e-05, "loss": 0.7082, "step": 3423 }, { "epoch": 0.35, "learning_rate": 5.8894302007521675e-05, "loss": 0.4755, "step": 3424 }, { "epoch": 0.35, "learning_rate": 5.889365818076039e-05, "loss": 0.4169, "step": 3425 }, { "epoch": 0.35, "learning_rate": 5.8893014170130855e-05, "loss": 0.2656, "step": 3426 }, { "epoch": 0.35, "learning_rate": 5.889236997563717e-05, "loss": 0.2342, "step": 3427 }, { "epoch": 0.35, "learning_rate": 5.889172559728343e-05, "loss": 0.5268, "step": 3428 }, { "epoch": 0.35, "learning_rate": 5.889108103507375e-05, "loss": 0.4554, "step": 3429 }, { "epoch": 0.35, "learning_rate": 5.8890436289012215e-05, "loss": 0.584, "step": 3430 }, { "epoch": 0.35, "learning_rate": 5.8889791359102934e-05, "loss": 1.1995, "step": 3431 }, { "epoch": 0.35, "learning_rate": 5.888914624535001e-05, "loss": 1.3482, "step": 3432 }, { "epoch": 0.35, "learning_rate": 5.888850094775755e-05, "loss": 0.6695, "step": 3433 }, { "epoch": 0.35, "learning_rate": 5.888785546632966e-05, "loss": 1.3015, "step": 3434 }, { "epoch": 0.35, "learning_rate": 5.8887209801070445e-05, "loss": 0.549, "step": 3435 }, { "epoch": 0.35, "learning_rate": 5.8886563951984014e-05, "loss": 0.6433, "step": 3436 }, { "epoch": 0.35, "learning_rate": 5.888591791907448e-05, "loss": 1.1518, "step": 3437 }, { "epoch": 0.35, "learning_rate": 5.888527170234597e-05, "loss": 0.3103, "step": 3438 }, { "epoch": 0.35, "learning_rate": 5.888462530180255e-05, "loss": 0.6997, "step": 3439 }, { "epoch": 0.35, "learning_rate": 5.8883978717448386e-05, "loss": 0.7393, "step": 3440 }, { "epoch": 0.35, "learning_rate": 5.888333194928756e-05, "loss": 0.3248, "step": 3441 }, { "epoch": 0.35, "learning_rate": 5.88826849973242e-05, "loss": 1.3735, "step": 3442 }, { "epoch": 0.35, "learning_rate": 5.8882037861562426e-05, "loss": 0.7454, "step": 3443 }, { "epoch": 0.35, "learning_rate": 5.8881390542006335e-05, "loss": 0.3714, "step": 3444 }, { "epoch": 0.35, "learning_rate": 5.888074303866007e-05, "loss": 0.2779, "step": 3445 }, { "epoch": 0.35, "learning_rate": 5.8880095351527745e-05, "loss": 1.2573, "step": 3446 }, { "epoch": 0.35, "learning_rate": 5.887944748061348e-05, "loss": 0.2671, "step": 3447 }, { "epoch": 0.35, "learning_rate": 5.887879942592139e-05, "loss": 0.384, "step": 3448 }, { "epoch": 0.35, "learning_rate": 5.8878151187455616e-05, "loss": 1.1886, "step": 3449 }, { "epoch": 0.35, "learning_rate": 5.887750276522027e-05, "loss": 0.9315, "step": 3450 }, { "epoch": 0.35, "learning_rate": 5.887685415921948e-05, "loss": 0.2024, "step": 3451 }, { "epoch": 0.35, "learning_rate": 5.887620536945738e-05, "loss": 0.4159, "step": 3452 }, { "epoch": 0.35, "learning_rate": 5.8875556395938086e-05, "loss": 0.6584, "step": 3453 }, { "epoch": 0.35, "learning_rate": 5.887490723866574e-05, "loss": 0.2768, "step": 3454 }, { "epoch": 0.35, "learning_rate": 5.887425789764447e-05, "loss": 0.1503, "step": 3455 }, { "epoch": 0.35, "learning_rate": 5.8873608372878405e-05, "loss": 1.3673, "step": 3456 }, { "epoch": 0.35, "learning_rate": 5.887295866437169e-05, "loss": 0.7006, "step": 3457 }, { "epoch": 0.35, "learning_rate": 5.887230877212844e-05, "loss": 1.0991, "step": 3458 }, { "epoch": 0.35, "learning_rate": 5.8871658696152804e-05, "loss": 0.5982, "step": 3459 }, { "epoch": 0.35, "learning_rate": 5.887100843644891e-05, "loss": 0.5703, "step": 3460 }, { "epoch": 0.35, "learning_rate": 5.887035799302091e-05, "loss": 1.1949, "step": 3461 }, { "epoch": 0.35, "learning_rate": 5.886970736587293e-05, "loss": 0.3864, "step": 3462 }, { "epoch": 0.35, "learning_rate": 5.886905655500912e-05, "loss": 0.9145, "step": 3463 }, { "epoch": 0.35, "learning_rate": 5.8868405560433604e-05, "loss": 1.3092, "step": 3464 }, { "epoch": 0.35, "learning_rate": 5.886775438215055e-05, "loss": 1.4215, "step": 3465 }, { "epoch": 0.35, "learning_rate": 5.886710302016408e-05, "loss": 0.4753, "step": 3466 }, { "epoch": 0.35, "learning_rate": 5.8866451474478355e-05, "loss": 1.5962, "step": 3467 }, { "epoch": 0.35, "learning_rate": 5.8865799745097504e-05, "loss": 1.1006, "step": 3468 }, { "epoch": 0.35, "learning_rate": 5.8865147832025694e-05, "loss": 1.0095, "step": 3469 }, { "epoch": 0.35, "learning_rate": 5.886449573526706e-05, "loss": 1.5328, "step": 3470 }, { "epoch": 0.35, "learning_rate": 5.8863843454825754e-05, "loss": 0.705, "step": 3471 }, { "epoch": 0.35, "learning_rate": 5.886319099070594e-05, "loss": 0.6299, "step": 3472 }, { "epoch": 0.35, "learning_rate": 5.8862538342911746e-05, "loss": 0.9109, "step": 3473 }, { "epoch": 0.35, "learning_rate": 5.886188551144734e-05, "loss": 0.8463, "step": 3474 }, { "epoch": 0.35, "learning_rate": 5.886123249631688e-05, "loss": 1.3161, "step": 3475 }, { "epoch": 0.35, "learning_rate": 5.886057929752452e-05, "loss": 0.2994, "step": 3476 }, { "epoch": 0.35, "learning_rate": 5.8859925915074396e-05, "loss": 1.3466, "step": 3477 }, { "epoch": 0.35, "learning_rate": 5.8859272348970694e-05, "loss": 1.375, "step": 3478 }, { "epoch": 0.35, "learning_rate": 5.8858618599217556e-05, "loss": 1.2225, "step": 3479 }, { "epoch": 0.35, "learning_rate": 5.885796466581915e-05, "loss": 1.042, "step": 3480 }, { "epoch": 0.35, "learning_rate": 5.885731054877964e-05, "loss": 0.4315, "step": 3481 }, { "epoch": 0.35, "learning_rate": 5.8856656248103175e-05, "loss": 1.1288, "step": 3482 }, { "epoch": 0.35, "learning_rate": 5.885600176379393e-05, "loss": 0.5521, "step": 3483 }, { "epoch": 0.35, "learning_rate": 5.885534709585607e-05, "loss": 1.3983, "step": 3484 }, { "epoch": 0.35, "learning_rate": 5.885469224429376e-05, "loss": 1.2684, "step": 3485 }, { "epoch": 0.35, "learning_rate": 5.885403720911116e-05, "loss": 0.4875, "step": 3486 }, { "epoch": 0.35, "learning_rate": 5.885338199031244e-05, "loss": 1.3461, "step": 3487 }, { "epoch": 0.35, "learning_rate": 5.885272658790179e-05, "loss": 0.4987, "step": 3488 }, { "epoch": 0.35, "learning_rate": 5.8852071001883346e-05, "loss": 1.1231, "step": 3489 }, { "epoch": 0.35, "learning_rate": 5.885141523226131e-05, "loss": 0.7517, "step": 3490 }, { "epoch": 0.35, "learning_rate": 5.8850759279039836e-05, "loss": 1.6189, "step": 3491 }, { "epoch": 0.35, "learning_rate": 5.88501031422231e-05, "loss": 1.3794, "step": 3492 }, { "epoch": 0.35, "learning_rate": 5.88494468218153e-05, "loss": 0.6777, "step": 3493 }, { "epoch": 0.35, "learning_rate": 5.884879031782058e-05, "loss": 0.4401, "step": 3494 }, { "epoch": 0.35, "learning_rate": 5.884813363024314e-05, "loss": 0.2282, "step": 3495 }, { "epoch": 0.35, "learning_rate": 5.884747675908715e-05, "loss": 1.2768, "step": 3496 }, { "epoch": 0.36, "learning_rate": 5.884681970435679e-05, "loss": 0.972, "step": 3497 }, { "epoch": 0.36, "learning_rate": 5.884616246605625e-05, "loss": 1.2212, "step": 3498 }, { "epoch": 0.36, "learning_rate": 5.88455050441897e-05, "loss": 1.0517, "step": 3499 }, { "epoch": 0.36, "learning_rate": 5.884484743876134e-05, "loss": 0.3974, "step": 3500 }, { "epoch": 0.36, "learning_rate": 5.884418964977532e-05, "loss": 1.1341, "step": 3501 }, { "epoch": 0.36, "learning_rate": 5.8843531677235874e-05, "loss": 0.5975, "step": 3502 }, { "epoch": 0.36, "learning_rate": 5.8842873521147156e-05, "loss": 0.3165, "step": 3503 }, { "epoch": 0.36, "learning_rate": 5.8842215181513365e-05, "loss": 0.3344, "step": 3504 }, { "epoch": 0.36, "learning_rate": 5.884155665833868e-05, "loss": 0.6578, "step": 3505 }, { "epoch": 0.36, "learning_rate": 5.884089795162731e-05, "loss": 0.268, "step": 3506 }, { "epoch": 0.36, "learning_rate": 5.884023906138344e-05, "loss": 1.1272, "step": 3507 }, { "epoch": 0.36, "learning_rate": 5.8839579987611245e-05, "loss": 1.0217, "step": 3508 }, { "epoch": 0.36, "learning_rate": 5.883892073031495e-05, "loss": 0.9645, "step": 3509 }, { "epoch": 0.36, "learning_rate": 5.8838261289498726e-05, "loss": 1.587, "step": 3510 }, { "epoch": 0.36, "learning_rate": 5.8837601665166784e-05, "loss": 0.5394, "step": 3511 }, { "epoch": 0.36, "learning_rate": 5.883694185732331e-05, "loss": 0.7039, "step": 3512 }, { "epoch": 0.36, "learning_rate": 5.883628186597251e-05, "loss": 0.7427, "step": 3513 }, { "epoch": 0.36, "learning_rate": 5.8835621691118577e-05, "loss": 0.5334, "step": 3514 }, { "epoch": 0.36, "learning_rate": 5.8834961332765724e-05, "loss": 0.5265, "step": 3515 }, { "epoch": 0.36, "learning_rate": 5.8834300790918146e-05, "loss": 1.021, "step": 3516 }, { "epoch": 0.36, "learning_rate": 5.883364006558005e-05, "loss": 1.9651, "step": 3517 }, { "epoch": 0.36, "learning_rate": 5.8832979156755624e-05, "loss": 0.2336, "step": 3518 }, { "epoch": 0.36, "learning_rate": 5.8832318064449095e-05, "loss": 0.7351, "step": 3519 }, { "epoch": 0.36, "learning_rate": 5.883165678866467e-05, "loss": 1.213, "step": 3520 }, { "epoch": 0.36, "learning_rate": 5.883099532940654e-05, "loss": 1.4893, "step": 3521 }, { "epoch": 0.36, "learning_rate": 5.8830333686678926e-05, "loss": 0.7263, "step": 3522 }, { "epoch": 0.36, "learning_rate": 5.882967186048604e-05, "loss": 1.2321, "step": 3523 }, { "epoch": 0.36, "learning_rate": 5.882900985083208e-05, "loss": 0.2234, "step": 3524 }, { "epoch": 0.36, "learning_rate": 5.8828347657721276e-05, "loss": 0.9439, "step": 3525 }, { "epoch": 0.36, "learning_rate": 5.882768528115783e-05, "loss": 1.1727, "step": 3526 }, { "epoch": 0.36, "learning_rate": 5.882702272114596e-05, "loss": 1.3027, "step": 3527 }, { "epoch": 0.36, "learning_rate": 5.882635997768989e-05, "loss": 1.285, "step": 3528 }, { "epoch": 0.36, "learning_rate": 5.882569705079383e-05, "loss": 1.7336, "step": 3529 }, { "epoch": 0.36, "learning_rate": 5.8825033940462e-05, "loss": 0.7748, "step": 3530 }, { "epoch": 0.36, "learning_rate": 5.882437064669862e-05, "loss": 0.8646, "step": 3531 }, { "epoch": 0.36, "learning_rate": 5.8823707169507904e-05, "loss": 1.2713, "step": 3532 }, { "epoch": 0.36, "learning_rate": 5.882304350889409e-05, "loss": 1.1937, "step": 3533 }, { "epoch": 0.36, "learning_rate": 5.8822379664861385e-05, "loss": 1.4386, "step": 3534 }, { "epoch": 0.36, "learning_rate": 5.8821715637414025e-05, "loss": 1.7115, "step": 3535 }, { "epoch": 0.36, "learning_rate": 5.882105142655622e-05, "loss": 1.3327, "step": 3536 }, { "epoch": 0.36, "learning_rate": 5.8820387032292224e-05, "loss": 0.2965, "step": 3537 }, { "epoch": 0.36, "learning_rate": 5.881972245462624e-05, "loss": 1.668, "step": 3538 }, { "epoch": 0.36, "learning_rate": 5.8819057693562507e-05, "loss": 0.6563, "step": 3539 }, { "epoch": 0.36, "learning_rate": 5.8818392749105256e-05, "loss": 1.3599, "step": 3540 }, { "epoch": 0.36, "learning_rate": 5.881772762125872e-05, "loss": 0.4142, "step": 3541 }, { "epoch": 0.36, "learning_rate": 5.8817062310027116e-05, "loss": 1.4215, "step": 3542 }, { "epoch": 0.36, "learning_rate": 5.88163968154147e-05, "loss": 1.2715, "step": 3543 }, { "epoch": 0.36, "learning_rate": 5.881573113742569e-05, "loss": 1.2776, "step": 3544 }, { "epoch": 0.36, "learning_rate": 5.881506527606433e-05, "loss": 1.1619, "step": 3545 }, { "epoch": 0.36, "learning_rate": 5.881439923133487e-05, "loss": 0.3802, "step": 3546 }, { "epoch": 0.36, "learning_rate": 5.8813733003241527e-05, "loss": 0.6063, "step": 3547 }, { "epoch": 0.36, "learning_rate": 5.881306659178854e-05, "loss": 0.5244, "step": 3548 }, { "epoch": 0.36, "learning_rate": 5.8812399996980164e-05, "loss": 0.8083, "step": 3549 }, { "epoch": 0.36, "learning_rate": 5.8811733218820637e-05, "loss": 0.4158, "step": 3550 }, { "epoch": 0.36, "learning_rate": 5.88110662573142e-05, "loss": 1.8306, "step": 3551 }, { "epoch": 0.36, "learning_rate": 5.88103991124651e-05, "loss": 1.1793, "step": 3552 }, { "epoch": 0.36, "learning_rate": 5.880973178427757e-05, "loss": 1.2929, "step": 3553 }, { "epoch": 0.36, "learning_rate": 5.880906427275587e-05, "loss": 0.8139, "step": 3554 }, { "epoch": 0.36, "learning_rate": 5.880839657790425e-05, "loss": 1.5976, "step": 3555 }, { "epoch": 0.36, "learning_rate": 5.880772869972695e-05, "loss": 0.8458, "step": 3556 }, { "epoch": 0.36, "learning_rate": 5.880706063822823e-05, "loss": 1.1844, "step": 3557 }, { "epoch": 0.36, "learning_rate": 5.880639239341233e-05, "loss": 0.9646, "step": 3558 }, { "epoch": 0.36, "learning_rate": 5.8805723965283505e-05, "loss": 2.0395, "step": 3559 }, { "epoch": 0.36, "learning_rate": 5.8805055353846015e-05, "loss": 1.2708, "step": 3560 }, { "epoch": 0.36, "learning_rate": 5.880438655910411e-05, "loss": 0.7165, "step": 3561 }, { "epoch": 0.36, "learning_rate": 5.880371758106204e-05, "loss": 1.1802, "step": 3562 }, { "epoch": 0.36, "learning_rate": 5.8803048419724074e-05, "loss": 0.9779, "step": 3563 }, { "epoch": 0.36, "learning_rate": 5.8802379075094465e-05, "loss": 1.7173, "step": 3564 }, { "epoch": 0.36, "learning_rate": 5.880170954717747e-05, "loss": 1.3611, "step": 3565 }, { "epoch": 0.36, "learning_rate": 5.8801039835977366e-05, "loss": 1.0277, "step": 3566 }, { "epoch": 0.36, "learning_rate": 5.880036994149838e-05, "loss": 0.7042, "step": 3567 }, { "epoch": 0.36, "learning_rate": 5.879969986374481e-05, "loss": 0.3993, "step": 3568 }, { "epoch": 0.36, "learning_rate": 5.8799029602720907e-05, "loss": 1.4045, "step": 3569 }, { "epoch": 0.36, "learning_rate": 5.879835915843093e-05, "loss": 0.6966, "step": 3570 }, { "epoch": 0.36, "learning_rate": 5.879768853087915e-05, "loss": 1.7418, "step": 3571 }, { "epoch": 0.36, "learning_rate": 5.879701772006983e-05, "loss": 0.6589, "step": 3572 }, { "epoch": 0.36, "learning_rate": 5.879634672600725e-05, "loss": 0.6995, "step": 3573 }, { "epoch": 0.36, "learning_rate": 5.879567554869569e-05, "loss": 0.8704, "step": 3574 }, { "epoch": 0.36, "learning_rate": 5.879500418813937e-05, "loss": 0.7963, "step": 3575 }, { "epoch": 0.36, "learning_rate": 5.879433264434263e-05, "loss": 2.1028, "step": 3576 }, { "epoch": 0.36, "learning_rate": 5.87936609173097e-05, "loss": 1.0323, "step": 3577 }, { "epoch": 0.36, "learning_rate": 5.879298900704486e-05, "loss": 0.7014, "step": 3578 }, { "epoch": 0.36, "learning_rate": 5.8792316913552396e-05, "loss": 0.5964, "step": 3579 }, { "epoch": 0.36, "learning_rate": 5.879164463683659e-05, "loss": 0.5055, "step": 3580 }, { "epoch": 0.36, "learning_rate": 5.879097217690169e-05, "loss": 0.7304, "step": 3581 }, { "epoch": 0.36, "learning_rate": 5.8790299533752e-05, "loss": 0.5561, "step": 3582 }, { "epoch": 0.36, "learning_rate": 5.878962670739181e-05, "loss": 0.6476, "step": 3583 }, { "epoch": 0.36, "learning_rate": 5.8788953697825384e-05, "loss": 0.4813, "step": 3584 }, { "epoch": 0.36, "learning_rate": 5.8788280505056994e-05, "loss": 0.6544, "step": 3585 }, { "epoch": 0.36, "learning_rate": 5.878760712909096e-05, "loss": 1.7263, "step": 3586 }, { "epoch": 0.36, "learning_rate": 5.878693356993153e-05, "loss": 0.9668, "step": 3587 }, { "epoch": 0.36, "learning_rate": 5.878625982758301e-05, "loss": 0.4634, "step": 3588 }, { "epoch": 0.36, "learning_rate": 5.878558590204968e-05, "loss": 0.4708, "step": 3589 }, { "epoch": 0.36, "learning_rate": 5.878491179333584e-05, "loss": 1.2585, "step": 3590 }, { "epoch": 0.36, "learning_rate": 5.878423750144577e-05, "loss": 0.6445, "step": 3591 }, { "epoch": 0.36, "learning_rate": 5.878356302638375e-05, "loss": 1.2615, "step": 3592 }, { "epoch": 0.36, "learning_rate": 5.8782888368154105e-05, "loss": 1.2658, "step": 3593 }, { "epoch": 0.36, "learning_rate": 5.8782213526761086e-05, "loss": 0.2323, "step": 3594 }, { "epoch": 0.37, "learning_rate": 5.878153850220903e-05, "loss": 1.3082, "step": 3595 }, { "epoch": 0.37, "learning_rate": 5.87808632945022e-05, "loss": 0.5207, "step": 3596 }, { "epoch": 0.37, "learning_rate": 5.878018790364491e-05, "loss": 0.7276, "step": 3597 }, { "epoch": 0.37, "learning_rate": 5.877951232964145e-05, "loss": 0.8394, "step": 3598 }, { "epoch": 0.37, "learning_rate": 5.877883657249612e-05, "loss": 1.0189, "step": 3599 }, { "epoch": 0.37, "learning_rate": 5.8778160632213225e-05, "loss": 1.4657, "step": 3600 }, { "epoch": 0.37, "learning_rate": 5.8777484508797066e-05, "loss": 1.6905, "step": 3601 }, { "epoch": 0.37, "learning_rate": 5.8776808202251937e-05, "loss": 1.842, "step": 3602 }, { "epoch": 0.37, "learning_rate": 5.877613171258215e-05, "loss": 0.4138, "step": 3603 }, { "epoch": 0.37, "learning_rate": 5.877545503979202e-05, "loss": 1.1951, "step": 3604 }, { "epoch": 0.37, "learning_rate": 5.877477818388582e-05, "loss": 0.9745, "step": 3605 }, { "epoch": 0.37, "learning_rate": 5.877410114486789e-05, "loss": 1.4838, "step": 3606 }, { "epoch": 0.37, "learning_rate": 5.877342392274252e-05, "loss": 0.9232, "step": 3607 }, { "epoch": 0.37, "learning_rate": 5.877274651751404e-05, "loss": 0.8079, "step": 3608 }, { "epoch": 0.37, "learning_rate": 5.877206892918673e-05, "loss": 0.5593, "step": 3609 }, { "epoch": 0.37, "learning_rate": 5.877139115776493e-05, "loss": 1.0786, "step": 3610 }, { "epoch": 0.37, "learning_rate": 5.877071320325294e-05, "loss": 0.646, "step": 3611 }, { "epoch": 0.37, "learning_rate": 5.877003506565507e-05, "loss": 0.7549, "step": 3612 }, { "epoch": 0.37, "learning_rate": 5.876935674497564e-05, "loss": 0.3868, "step": 3613 }, { "epoch": 0.37, "learning_rate": 5.8768678241218976e-05, "loss": 1.3986, "step": 3614 }, { "epoch": 0.37, "learning_rate": 5.876799955438939e-05, "loss": 0.2583, "step": 3615 }, { "epoch": 0.37, "learning_rate": 5.876732068449119e-05, "loss": 1.1882, "step": 3616 }, { "epoch": 0.37, "learning_rate": 5.876664163152871e-05, "loss": 1.1835, "step": 3617 }, { "epoch": 0.37, "learning_rate": 5.876596239550627e-05, "loss": 0.4136, "step": 3618 }, { "epoch": 0.37, "learning_rate": 5.876528297642818e-05, "loss": 0.4532, "step": 3619 }, { "epoch": 0.37, "learning_rate": 5.8764603374298775e-05, "loss": 0.8457, "step": 3620 }, { "epoch": 0.37, "learning_rate": 5.876392358912238e-05, "loss": 0.4777, "step": 3621 }, { "epoch": 0.37, "learning_rate": 5.876324362090332e-05, "loss": 0.376, "step": 3622 }, { "epoch": 0.37, "learning_rate": 5.8762563469645906e-05, "loss": 0.9026, "step": 3623 }, { "epoch": 0.37, "learning_rate": 5.8761883135354494e-05, "loss": 1.2817, "step": 3624 }, { "epoch": 0.37, "learning_rate": 5.876120261803339e-05, "loss": 0.4474, "step": 3625 }, { "epoch": 0.37, "learning_rate": 5.876052191768694e-05, "loss": 0.7016, "step": 3626 }, { "epoch": 0.37, "learning_rate": 5.875984103431946e-05, "loss": 1.6526, "step": 3627 }, { "epoch": 0.37, "learning_rate": 5.875915996793531e-05, "loss": 1.0542, "step": 3628 }, { "epoch": 0.37, "learning_rate": 5.875847871853879e-05, "loss": 1.511, "step": 3629 }, { "epoch": 0.37, "learning_rate": 5.8757797286134255e-05, "loss": 0.8639, "step": 3630 }, { "epoch": 0.37, "learning_rate": 5.875711567072605e-05, "loss": 0.4637, "step": 3631 }, { "epoch": 0.37, "learning_rate": 5.875643387231849e-05, "loss": 0.8728, "step": 3632 }, { "epoch": 0.37, "learning_rate": 5.8755751890915925e-05, "loss": 0.9051, "step": 3633 }, { "epoch": 0.37, "learning_rate": 5.8755069726522694e-05, "loss": 0.7845, "step": 3634 }, { "epoch": 0.37, "learning_rate": 5.875438737914314e-05, "loss": 0.5266, "step": 3635 }, { "epoch": 0.37, "learning_rate": 5.875370484878161e-05, "loss": 0.5504, "step": 3636 }, { "epoch": 0.37, "learning_rate": 5.875302213544242e-05, "loss": 0.8169, "step": 3637 }, { "epoch": 0.37, "learning_rate": 5.8752339239129954e-05, "loss": 0.3925, "step": 3638 }, { "epoch": 0.37, "learning_rate": 5.8751656159848536e-05, "loss": 1.655, "step": 3639 }, { "epoch": 0.37, "learning_rate": 5.875097289760251e-05, "loss": 0.6487, "step": 3640 }, { "epoch": 0.37, "learning_rate": 5.875028945239624e-05, "loss": 0.5784, "step": 3641 }, { "epoch": 0.37, "learning_rate": 5.874960582423406e-05, "loss": 1.1099, "step": 3642 }, { "epoch": 0.37, "learning_rate": 5.8748922013120326e-05, "loss": 0.1282, "step": 3643 }, { "epoch": 0.37, "learning_rate": 5.8748238019059385e-05, "loss": 1.3003, "step": 3644 }, { "epoch": 0.37, "learning_rate": 5.874755384205559e-05, "loss": 0.3525, "step": 3645 }, { "epoch": 0.37, "learning_rate": 5.874686948211331e-05, "loss": 0.6549, "step": 3646 }, { "epoch": 0.37, "learning_rate": 5.874618493923689e-05, "loss": 0.3201, "step": 3647 }, { "epoch": 0.37, "learning_rate": 5.874550021343067e-05, "loss": 1.429, "step": 3648 }, { "epoch": 0.37, "learning_rate": 5.874481530469902e-05, "loss": 0.6597, "step": 3649 }, { "epoch": 0.37, "learning_rate": 5.8744130213046324e-05, "loss": 1.4156, "step": 3650 }, { "epoch": 0.37, "learning_rate": 5.87434449384769e-05, "loss": 1.4723, "step": 3651 }, { "epoch": 0.37, "learning_rate": 5.874275948099513e-05, "loss": 1.3907, "step": 3652 }, { "epoch": 0.37, "learning_rate": 5.874207384060538e-05, "loss": 1.3793, "step": 3653 }, { "epoch": 0.37, "learning_rate": 5.8741388017312e-05, "loss": 0.9297, "step": 3654 }, { "epoch": 0.37, "learning_rate": 5.8740702011119364e-05, "loss": 0.5218, "step": 3655 }, { "epoch": 0.37, "learning_rate": 5.8740015822031826e-05, "loss": 0.9069, "step": 3656 }, { "epoch": 0.37, "learning_rate": 5.8739329450053766e-05, "loss": 0.2779, "step": 3657 }, { "epoch": 0.37, "learning_rate": 5.8738642895189544e-05, "loss": 1.1787, "step": 3658 }, { "epoch": 0.37, "learning_rate": 5.873795615744353e-05, "loss": 0.2867, "step": 3659 }, { "epoch": 0.37, "learning_rate": 5.8737269236820105e-05, "loss": 1.5357, "step": 3660 }, { "epoch": 0.37, "learning_rate": 5.873658213332362e-05, "loss": 0.61, "step": 3661 }, { "epoch": 0.37, "learning_rate": 5.873589484695847e-05, "loss": 0.3386, "step": 3662 }, { "epoch": 0.37, "learning_rate": 5.8735207377729e-05, "loss": 0.3407, "step": 3663 }, { "epoch": 0.37, "learning_rate": 5.873451972563961e-05, "loss": 0.5919, "step": 3664 }, { "epoch": 0.37, "learning_rate": 5.873383189069468e-05, "loss": 1.395, "step": 3665 }, { "epoch": 0.37, "learning_rate": 5.8733143872898555e-05, "loss": 1.8252, "step": 3666 }, { "epoch": 0.37, "learning_rate": 5.8732455672255645e-05, "loss": 0.2498, "step": 3667 }, { "epoch": 0.37, "learning_rate": 5.8731767288770305e-05, "loss": 0.2169, "step": 3668 }, { "epoch": 0.37, "learning_rate": 5.873107872244695e-05, "loss": 1.2684, "step": 3669 }, { "epoch": 0.37, "learning_rate": 5.873038997328992e-05, "loss": 0.4098, "step": 3670 }, { "epoch": 0.37, "learning_rate": 5.872970104130362e-05, "loss": 0.8841, "step": 3671 }, { "epoch": 0.37, "learning_rate": 5.872901192649245e-05, "loss": 1.6248, "step": 3672 }, { "epoch": 0.37, "learning_rate": 5.872832262886076e-05, "loss": 0.9103, "step": 3673 }, { "epoch": 0.37, "learning_rate": 5.872763314841295e-05, "loss": 0.6577, "step": 3674 }, { "epoch": 0.37, "learning_rate": 5.872694348515343e-05, "loss": 0.9259, "step": 3675 }, { "epoch": 0.37, "learning_rate": 5.8726253639086554e-05, "loss": 0.995, "step": 3676 }, { "epoch": 0.37, "learning_rate": 5.8725563610216736e-05, "loss": 0.3373, "step": 3677 }, { "epoch": 0.37, "learning_rate": 5.872487339854835e-05, "loss": 0.362, "step": 3678 }, { "epoch": 0.37, "learning_rate": 5.87241830040858e-05, "loss": 0.805, "step": 3679 }, { "epoch": 0.37, "learning_rate": 5.872349242683349e-05, "loss": 1.0607, "step": 3680 }, { "epoch": 0.37, "learning_rate": 5.8722801666795784e-05, "loss": 0.9683, "step": 3681 }, { "epoch": 0.37, "learning_rate": 5.8722110723977104e-05, "loss": 1.0269, "step": 3682 }, { "epoch": 0.37, "learning_rate": 5.8721419598381826e-05, "loss": 0.5952, "step": 3683 }, { "epoch": 0.37, "learning_rate": 5.872072829001437e-05, "loss": 1.232, "step": 3684 }, { "epoch": 0.37, "learning_rate": 5.872003679887913e-05, "loss": 1.4078, "step": 3685 }, { "epoch": 0.37, "learning_rate": 5.87193451249805e-05, "loss": 0.5056, "step": 3686 }, { "epoch": 0.37, "learning_rate": 5.8718653268322876e-05, "loss": 0.9677, "step": 3687 }, { "epoch": 0.37, "learning_rate": 5.871796122891067e-05, "loss": 0.3723, "step": 3688 }, { "epoch": 0.37, "learning_rate": 5.871726900674828e-05, "loss": 0.6786, "step": 3689 }, { "epoch": 0.37, "learning_rate": 5.8716576601840125e-05, "loss": 0.6834, "step": 3690 }, { "epoch": 0.37, "learning_rate": 5.8715884014190595e-05, "loss": 1.4157, "step": 3691 }, { "epoch": 0.37, "learning_rate": 5.87151912438041e-05, "loss": 0.6111, "step": 3692 }, { "epoch": 0.38, "learning_rate": 5.8714498290685054e-05, "loss": 0.6755, "step": 3693 }, { "epoch": 0.38, "learning_rate": 5.871380515483787e-05, "loss": 0.4524, "step": 3694 }, { "epoch": 0.38, "learning_rate": 5.8713111836266944e-05, "loss": 1.3105, "step": 3695 }, { "epoch": 0.38, "learning_rate": 5.8712418334976695e-05, "loss": 1.1404, "step": 3696 }, { "epoch": 0.38, "learning_rate": 5.8711724650971554e-05, "loss": 0.6821, "step": 3697 }, { "epoch": 0.38, "learning_rate": 5.871103078425591e-05, "loss": 0.478, "step": 3698 }, { "epoch": 0.38, "learning_rate": 5.871033673483419e-05, "loss": 1.6746, "step": 3699 }, { "epoch": 0.38, "learning_rate": 5.870964250271081e-05, "loss": 1.072, "step": 3700 }, { "epoch": 0.38, "learning_rate": 5.870894808789018e-05, "loss": 1.0828, "step": 3701 }, { "epoch": 0.38, "learning_rate": 5.870825349037673e-05, "loss": 1.1821, "step": 3702 }, { "epoch": 0.38, "learning_rate": 5.870755871017488e-05, "loss": 1.4565, "step": 3703 }, { "epoch": 0.38, "learning_rate": 5.8706863747289035e-05, "loss": 0.2149, "step": 3704 }, { "epoch": 0.38, "learning_rate": 5.8706168601723643e-05, "loss": 0.6362, "step": 3705 }, { "epoch": 0.38, "learning_rate": 5.8705473273483115e-05, "loss": 0.7691, "step": 3706 }, { "epoch": 0.38, "learning_rate": 5.870477776257186e-05, "loss": 0.8962, "step": 3707 }, { "epoch": 0.38, "learning_rate": 5.870408206899434e-05, "loss": 0.611, "step": 3708 }, { "epoch": 0.38, "learning_rate": 5.8703386192754944e-05, "loss": 0.3463, "step": 3709 }, { "epoch": 0.38, "learning_rate": 5.870269013385812e-05, "loss": 0.2274, "step": 3710 }, { "epoch": 0.38, "learning_rate": 5.8701993892308304e-05, "loss": 0.5887, "step": 3711 }, { "epoch": 0.38, "learning_rate": 5.870129746810992e-05, "loss": 1.3637, "step": 3712 }, { "epoch": 0.38, "learning_rate": 5.870060086126739e-05, "loss": 0.4533, "step": 3713 }, { "epoch": 0.38, "learning_rate": 5.8699904071785156e-05, "loss": 2.1436, "step": 3714 }, { "epoch": 0.38, "learning_rate": 5.8699207099667644e-05, "loss": 0.63, "step": 3715 }, { "epoch": 0.38, "learning_rate": 5.8698509944919305e-05, "loss": 0.8836, "step": 3716 }, { "epoch": 0.38, "learning_rate": 5.8697812607544565e-05, "loss": 0.2451, "step": 3717 }, { "epoch": 0.38, "learning_rate": 5.8697115087547853e-05, "loss": 1.109, "step": 3718 }, { "epoch": 0.38, "learning_rate": 5.8696417384933624e-05, "loss": 1.0371, "step": 3719 }, { "epoch": 0.38, "learning_rate": 5.8695719499706315e-05, "loss": 1.0683, "step": 3720 }, { "epoch": 0.38, "learning_rate": 5.8695021431870364e-05, "loss": 1.0357, "step": 3721 }, { "epoch": 0.38, "learning_rate": 5.86943231814302e-05, "loss": 0.8074, "step": 3722 }, { "epoch": 0.38, "learning_rate": 5.869362474839029e-05, "loss": 0.4524, "step": 3723 }, { "epoch": 0.38, "learning_rate": 5.8692926132755074e-05, "loss": 0.5748, "step": 3724 }, { "epoch": 0.38, "learning_rate": 5.869222733452898e-05, "loss": 1.3749, "step": 3725 }, { "epoch": 0.38, "learning_rate": 5.869152835371647e-05, "loss": 1.3399, "step": 3726 }, { "epoch": 0.38, "learning_rate": 5.869082919032199e-05, "loss": 0.8184, "step": 3727 }, { "epoch": 0.38, "learning_rate": 5.869012984434998e-05, "loss": 0.9761, "step": 3728 }, { "epoch": 0.38, "learning_rate": 5.868943031580491e-05, "loss": 1.1398, "step": 3729 }, { "epoch": 0.38, "learning_rate": 5.868873060469121e-05, "loss": 1.0444, "step": 3730 }, { "epoch": 0.38, "learning_rate": 5.868803071101335e-05, "loss": 0.5229, "step": 3731 }, { "epoch": 0.38, "learning_rate": 5.868733063477577e-05, "loss": 1.3627, "step": 3732 }, { "epoch": 0.38, "learning_rate": 5.868663037598293e-05, "loss": 0.982, "step": 3733 }, { "epoch": 0.38, "learning_rate": 5.8685929934639294e-05, "loss": 0.7201, "step": 3734 }, { "epoch": 0.38, "learning_rate": 5.868522931074931e-05, "loss": 0.9566, "step": 3735 }, { "epoch": 0.38, "learning_rate": 5.8684528504317434e-05, "loss": 0.2415, "step": 3736 }, { "epoch": 0.38, "learning_rate": 5.868382751534813e-05, "loss": 0.6577, "step": 3737 }, { "epoch": 0.38, "learning_rate": 5.8683126343845874e-05, "loss": 0.3232, "step": 3738 }, { "epoch": 0.38, "learning_rate": 5.86824249898151e-05, "loss": 1.4851, "step": 3739 }, { "epoch": 0.38, "learning_rate": 5.8681723453260294e-05, "loss": 0.4136, "step": 3740 }, { "epoch": 0.38, "learning_rate": 5.86810217341859e-05, "loss": 1.6428, "step": 3741 }, { "epoch": 0.38, "learning_rate": 5.8680319832596395e-05, "loss": 1.0078, "step": 3742 }, { "epoch": 0.38, "learning_rate": 5.8679617748496256e-05, "loss": 1.1266, "step": 3743 }, { "epoch": 0.38, "learning_rate": 5.867891548188993e-05, "loss": 0.6882, "step": 3744 }, { "epoch": 0.38, "learning_rate": 5.86782130327819e-05, "loss": 0.395, "step": 3745 }, { "epoch": 0.38, "learning_rate": 5.867751040117664e-05, "loss": 1.5079, "step": 3746 }, { "epoch": 0.38, "learning_rate": 5.8676807587078596e-05, "loss": 0.2764, "step": 3747 }, { "epoch": 0.38, "learning_rate": 5.8676104590492265e-05, "loss": 1.143, "step": 3748 }, { "epoch": 0.38, "learning_rate": 5.8675401411422124e-05, "loss": 0.539, "step": 3749 }, { "epoch": 0.38, "learning_rate": 5.8674698049872626e-05, "loss": 0.2572, "step": 3750 }, { "epoch": 0.38, "learning_rate": 5.867399450584826e-05, "loss": 0.4065, "step": 3751 }, { "epoch": 0.38, "learning_rate": 5.8673290779353496e-05, "loss": 0.2989, "step": 3752 }, { "epoch": 0.38, "learning_rate": 5.867258687039282e-05, "loss": 1.1238, "step": 3753 }, { "epoch": 0.38, "learning_rate": 5.867188277897072e-05, "loss": 1.0109, "step": 3754 }, { "epoch": 0.38, "learning_rate": 5.867117850509165e-05, "loss": 0.9305, "step": 3755 }, { "epoch": 0.38, "learning_rate": 5.867047404876011e-05, "loss": 0.6781, "step": 3756 }, { "epoch": 0.38, "learning_rate": 5.866976940998059e-05, "loss": 0.93, "step": 3757 }, { "epoch": 0.38, "learning_rate": 5.866906458875756e-05, "loss": 0.5819, "step": 3758 }, { "epoch": 0.38, "learning_rate": 5.86683595850955e-05, "loss": 0.8689, "step": 3759 }, { "epoch": 0.38, "learning_rate": 5.866765439899892e-05, "loss": 0.5258, "step": 3760 }, { "epoch": 0.38, "learning_rate": 5.866694903047228e-05, "loss": 1.3705, "step": 3761 }, { "epoch": 0.38, "learning_rate": 5.866624347952009e-05, "loss": 0.4102, "step": 3762 }, { "epoch": 0.38, "learning_rate": 5.8665537746146834e-05, "loss": 1.4762, "step": 3763 }, { "epoch": 0.38, "learning_rate": 5.8664831830356996e-05, "loss": 1.1074, "step": 3764 }, { "epoch": 0.38, "learning_rate": 5.866412573215507e-05, "loss": 1.4841, "step": 3765 }, { "epoch": 0.38, "learning_rate": 5.8663419451545556e-05, "loss": 1.0682, "step": 3766 }, { "epoch": 0.38, "learning_rate": 5.866271298853294e-05, "loss": 0.8948, "step": 3767 }, { "epoch": 0.38, "learning_rate": 5.866200634312173e-05, "loss": 0.8067, "step": 3768 }, { "epoch": 0.38, "learning_rate": 5.866129951531641e-05, "loss": 0.5117, "step": 3769 }, { "epoch": 0.38, "learning_rate": 5.866059250512149e-05, "loss": 1.3818, "step": 3770 }, { "epoch": 0.38, "learning_rate": 5.865988531254145e-05, "loss": 1.2533, "step": 3771 }, { "epoch": 0.38, "learning_rate": 5.865917793758081e-05, "loss": 1.3202, "step": 3772 }, { "epoch": 0.38, "learning_rate": 5.8658470380244066e-05, "loss": 0.741, "step": 3773 }, { "epoch": 0.38, "learning_rate": 5.865776264053572e-05, "loss": 1.4049, "step": 3774 }, { "epoch": 0.38, "learning_rate": 5.865705471846026e-05, "loss": 1.0439, "step": 3775 }, { "epoch": 0.38, "learning_rate": 5.865634661402222e-05, "loss": 0.505, "step": 3776 }, { "epoch": 0.38, "learning_rate": 5.865563832722609e-05, "loss": 0.9405, "step": 3777 }, { "epoch": 0.38, "learning_rate": 5.865492985807638e-05, "loss": 1.1547, "step": 3778 }, { "epoch": 0.38, "learning_rate": 5.865422120657759e-05, "loss": 0.4077, "step": 3779 }, { "epoch": 0.38, "learning_rate": 5.865351237273424e-05, "loss": 0.8069, "step": 3780 }, { "epoch": 0.38, "learning_rate": 5.8652803356550845e-05, "loss": 0.679, "step": 3781 }, { "epoch": 0.38, "learning_rate": 5.86520941580319e-05, "loss": 1.0285, "step": 3782 }, { "epoch": 0.38, "learning_rate": 5.865138477718194e-05, "loss": 0.3399, "step": 3783 }, { "epoch": 0.38, "learning_rate": 5.865067521400545e-05, "loss": 1.1662, "step": 3784 }, { "epoch": 0.38, "learning_rate": 5.8649965468506974e-05, "loss": 0.3791, "step": 3785 }, { "epoch": 0.38, "learning_rate": 5.8649255540691005e-05, "loss": 1.4079, "step": 3786 }, { "epoch": 0.38, "learning_rate": 5.864854543056208e-05, "loss": 0.2902, "step": 3787 }, { "epoch": 0.38, "learning_rate": 5.864783513812472e-05, "loss": 0.5728, "step": 3788 }, { "epoch": 0.38, "learning_rate": 5.864712466338342e-05, "loss": 0.6557, "step": 3789 }, { "epoch": 0.38, "learning_rate": 5.864641400634272e-05, "loss": 1.0501, "step": 3790 }, { "epoch": 0.38, "learning_rate": 5.864570316700714e-05, "loss": 1.1548, "step": 3791 }, { "epoch": 0.39, "learning_rate": 5.864499214538121e-05, "loss": 0.7386, "step": 3792 }, { "epoch": 0.39, "learning_rate": 5.864428094146944e-05, "loss": 0.4735, "step": 3793 }, { "epoch": 0.39, "learning_rate": 5.864356955527637e-05, "loss": 0.3725, "step": 3794 }, { "epoch": 0.39, "learning_rate": 5.8642857986806505e-05, "loss": 0.4199, "step": 3795 }, { "epoch": 0.39, "learning_rate": 5.8642146236064394e-05, "loss": 0.7338, "step": 3796 }, { "epoch": 0.39, "learning_rate": 5.864143430305457e-05, "loss": 0.7488, "step": 3797 }, { "epoch": 0.39, "learning_rate": 5.864072218778154e-05, "loss": 1.6355, "step": 3798 }, { "epoch": 0.39, "learning_rate": 5.864000989024986e-05, "loss": 0.5452, "step": 3799 }, { "epoch": 0.39, "learning_rate": 5.863929741046405e-05, "loss": 0.9438, "step": 3800 }, { "epoch": 0.39, "learning_rate": 5.863858474842864e-05, "loss": 1.1654, "step": 3801 }, { "epoch": 0.39, "learning_rate": 5.863787190414818e-05, "loss": 1.0736, "step": 3802 }, { "epoch": 0.39, "learning_rate": 5.86371588776272e-05, "loss": 1.0099, "step": 3803 }, { "epoch": 0.39, "learning_rate": 5.863644566887023e-05, "loss": 0.5217, "step": 3804 }, { "epoch": 0.39, "learning_rate": 5.863573227788181e-05, "loss": 0.8162, "step": 3805 }, { "epoch": 0.39, "learning_rate": 5.863501870466648e-05, "loss": 1.9419, "step": 3806 }, { "epoch": 0.39, "learning_rate": 5.863430494922879e-05, "loss": 1.2912, "step": 3807 }, { "epoch": 0.39, "learning_rate": 5.863359101157328e-05, "loss": 0.2574, "step": 3808 }, { "epoch": 0.39, "learning_rate": 5.863287689170448e-05, "loss": 1.2719, "step": 3809 }, { "epoch": 0.39, "learning_rate": 5.863216258962695e-05, "loss": 0.1854, "step": 3810 }, { "epoch": 0.39, "learning_rate": 5.863144810534524e-05, "loss": 0.4131, "step": 3811 }, { "epoch": 0.39, "learning_rate": 5.8630733438863875e-05, "loss": 0.7183, "step": 3812 }, { "epoch": 0.39, "learning_rate": 5.8630018590187406e-05, "loss": 0.8501, "step": 3813 }, { "epoch": 0.39, "learning_rate": 5.86293035593204e-05, "loss": 1.4531, "step": 3814 }, { "epoch": 0.39, "learning_rate": 5.8628588346267404e-05, "loss": 0.5166, "step": 3815 }, { "epoch": 0.39, "learning_rate": 5.862787295103295e-05, "loss": 0.6888, "step": 3816 }, { "epoch": 0.39, "learning_rate": 5.8627157373621614e-05, "loss": 1.2378, "step": 3817 }, { "epoch": 0.39, "learning_rate": 5.862644161403793e-05, "loss": 1.5222, "step": 3818 }, { "epoch": 0.39, "learning_rate": 5.8625725672286464e-05, "loss": 1.3888, "step": 3819 }, { "epoch": 0.39, "learning_rate": 5.862500954837177e-05, "loss": 1.0282, "step": 3820 }, { "epoch": 0.39, "learning_rate": 5.8624293242298404e-05, "loss": 1.0886, "step": 3821 }, { "epoch": 0.39, "learning_rate": 5.862357675407093e-05, "loss": 0.4746, "step": 3822 }, { "epoch": 0.39, "learning_rate": 5.86228600836939e-05, "loss": 1.2721, "step": 3823 }, { "epoch": 0.39, "learning_rate": 5.862214323117187e-05, "loss": 0.9985, "step": 3824 }, { "epoch": 0.39, "learning_rate": 5.862142619650942e-05, "loss": 0.8326, "step": 3825 }, { "epoch": 0.39, "learning_rate": 5.862070897971109e-05, "loss": 1.0627, "step": 3826 }, { "epoch": 0.39, "learning_rate": 5.861999158078146e-05, "loss": 1.221, "step": 3827 }, { "epoch": 0.39, "learning_rate": 5.86192739997251e-05, "loss": 0.2158, "step": 3828 }, { "epoch": 0.39, "learning_rate": 5.8618556236546556e-05, "loss": 0.9124, "step": 3829 }, { "epoch": 0.39, "learning_rate": 5.861783829125042e-05, "loss": 1.0485, "step": 3830 }, { "epoch": 0.39, "learning_rate": 5.861712016384124e-05, "loss": 1.5361, "step": 3831 }, { "epoch": 0.39, "learning_rate": 5.861640185432358e-05, "loss": 0.7173, "step": 3832 }, { "epoch": 0.39, "learning_rate": 5.8615683362702044e-05, "loss": 0.796, "step": 3833 }, { "epoch": 0.39, "learning_rate": 5.861496468898118e-05, "loss": 0.9636, "step": 3834 }, { "epoch": 0.39, "learning_rate": 5.861424583316556e-05, "loss": 0.5986, "step": 3835 }, { "epoch": 0.39, "learning_rate": 5.8613526795259774e-05, "loss": 0.9335, "step": 3836 }, { "epoch": 0.39, "learning_rate": 5.861280757526839e-05, "loss": 1.5167, "step": 3837 }, { "epoch": 0.39, "learning_rate": 5.861208817319598e-05, "loss": 1.4352, "step": 3838 }, { "epoch": 0.39, "learning_rate": 5.861136858904713e-05, "loss": 1.7685, "step": 3839 }, { "epoch": 0.39, "learning_rate": 5.861064882282641e-05, "loss": 0.4209, "step": 3840 }, { "epoch": 0.39, "learning_rate": 5.86099288745384e-05, "loss": 0.888, "step": 3841 }, { "epoch": 0.39, "learning_rate": 5.860920874418769e-05, "loss": 1.3025, "step": 3842 }, { "epoch": 0.39, "learning_rate": 5.860848843177886e-05, "loss": 0.9764, "step": 3843 }, { "epoch": 0.39, "learning_rate": 5.860776793731649e-05, "loss": 0.9002, "step": 3844 }, { "epoch": 0.39, "learning_rate": 5.860704726080517e-05, "loss": 1.3592, "step": 3845 }, { "epoch": 0.39, "learning_rate": 5.860632640224948e-05, "loss": 1.0964, "step": 3846 }, { "epoch": 0.39, "learning_rate": 5.860560536165402e-05, "loss": 1.5486, "step": 3847 }, { "epoch": 0.39, "learning_rate": 5.860488413902336e-05, "loss": 0.7009, "step": 3848 }, { "epoch": 0.39, "learning_rate": 5.8604162734362115e-05, "loss": 0.7518, "step": 3849 }, { "epoch": 0.39, "learning_rate": 5.860344114767485e-05, "loss": 0.5252, "step": 3850 }, { "epoch": 0.39, "learning_rate": 5.860271937896616e-05, "loss": 1.3633, "step": 3851 }, { "epoch": 0.39, "learning_rate": 5.860199742824065e-05, "loss": 0.7052, "step": 3852 }, { "epoch": 0.39, "learning_rate": 5.860127529550291e-05, "loss": 0.7411, "step": 3853 }, { "epoch": 0.39, "learning_rate": 5.860055298075754e-05, "loss": 1.3901, "step": 3854 }, { "epoch": 0.39, "learning_rate": 5.8599830484009124e-05, "loss": 0.1513, "step": 3855 }, { "epoch": 0.39, "learning_rate": 5.859910780526227e-05, "loss": 1.7545, "step": 3856 }, { "epoch": 0.39, "learning_rate": 5.859838494452157e-05, "loss": 0.5164, "step": 3857 }, { "epoch": 0.39, "learning_rate": 5.8597661901791625e-05, "loss": 1.2791, "step": 3858 }, { "epoch": 0.39, "learning_rate": 5.8596938677077045e-05, "loss": 1.433, "step": 3859 }, { "epoch": 0.39, "learning_rate": 5.8596215270382425e-05, "loss": 1.1559, "step": 3860 }, { "epoch": 0.39, "learning_rate": 5.859549168171237e-05, "loss": 1.1875, "step": 3861 }, { "epoch": 0.39, "learning_rate": 5.8594767911071484e-05, "loss": 0.3293, "step": 3862 }, { "epoch": 0.39, "learning_rate": 5.859404395846437e-05, "loss": 0.3274, "step": 3863 }, { "epoch": 0.39, "learning_rate": 5.859331982389564e-05, "loss": 1.0109, "step": 3864 }, { "epoch": 0.39, "learning_rate": 5.85925955073699e-05, "loss": 0.9844, "step": 3865 }, { "epoch": 0.39, "learning_rate": 5.8591871008891763e-05, "loss": 1.2243, "step": 3866 }, { "epoch": 0.39, "learning_rate": 5.8591146328465826e-05, "loss": 1.0069, "step": 3867 }, { "epoch": 0.39, "learning_rate": 5.8590421466096716e-05, "loss": 1.384, "step": 3868 }, { "epoch": 0.39, "learning_rate": 5.858969642178904e-05, "loss": 0.6034, "step": 3869 }, { "epoch": 0.39, "learning_rate": 5.858897119554742e-05, "loss": 1.0258, "step": 3870 }, { "epoch": 0.39, "learning_rate": 5.858824578737645e-05, "loss": 0.8122, "step": 3871 }, { "epoch": 0.39, "learning_rate": 5.8587520197280766e-05, "loss": 0.8688, "step": 3872 }, { "epoch": 0.39, "learning_rate": 5.858679442526498e-05, "loss": 1.4872, "step": 3873 }, { "epoch": 0.39, "learning_rate": 5.858606847133371e-05, "loss": 1.5541, "step": 3874 }, { "epoch": 0.39, "learning_rate": 5.8585342335491574e-05, "loss": 0.9364, "step": 3875 }, { "epoch": 0.39, "learning_rate": 5.8584616017743184e-05, "loss": 0.7491, "step": 3876 }, { "epoch": 0.39, "learning_rate": 5.858388951809319e-05, "loss": 1.5177, "step": 3877 }, { "epoch": 0.39, "learning_rate": 5.8583162836546186e-05, "loss": 0.5462, "step": 3878 }, { "epoch": 0.39, "learning_rate": 5.85824359731068e-05, "loss": 1.1993, "step": 3879 }, { "epoch": 0.39, "learning_rate": 5.8581708927779674e-05, "loss": 0.6877, "step": 3880 }, { "epoch": 0.39, "learning_rate": 5.8580981700569426e-05, "loss": 0.88, "step": 3881 }, { "epoch": 0.39, "learning_rate": 5.8580254291480686e-05, "loss": 0.354, "step": 3882 }, { "epoch": 0.39, "learning_rate": 5.857952670051807e-05, "loss": 0.192, "step": 3883 }, { "epoch": 0.39, "learning_rate": 5.857879892768623e-05, "loss": 1.3075, "step": 3884 }, { "epoch": 0.39, "learning_rate": 5.8578070972989774e-05, "loss": 0.2202, "step": 3885 }, { "epoch": 0.39, "learning_rate": 5.8577342836433354e-05, "loss": 0.8967, "step": 3886 }, { "epoch": 0.39, "learning_rate": 5.8576614518021596e-05, "loss": 0.8661, "step": 3887 }, { "epoch": 0.39, "learning_rate": 5.857588601775913e-05, "loss": 0.6478, "step": 3888 }, { "epoch": 0.39, "learning_rate": 5.857515733565059e-05, "loss": 0.9476, "step": 3889 }, { "epoch": 0.4, "learning_rate": 5.857442847170063e-05, "loss": 0.6176, "step": 3890 }, { "epoch": 0.4, "learning_rate": 5.857369942591388e-05, "loss": 0.4338, "step": 3891 }, { "epoch": 0.4, "learning_rate": 5.857297019829497e-05, "loss": 1.361, "step": 3892 }, { "epoch": 0.4, "learning_rate": 5.8572240788848544e-05, "loss": 0.9187, "step": 3893 }, { "epoch": 0.4, "learning_rate": 5.8571511197579254e-05, "loss": 0.4831, "step": 3894 }, { "epoch": 0.4, "learning_rate": 5.8570781424491734e-05, "loss": 0.5339, "step": 3895 }, { "epoch": 0.4, "learning_rate": 5.857005146959063e-05, "loss": 0.9001, "step": 3896 }, { "epoch": 0.4, "learning_rate": 5.856932133288059e-05, "loss": 0.7823, "step": 3897 }, { "epoch": 0.4, "learning_rate": 5.8568591014366244e-05, "loss": 1.3509, "step": 3898 }, { "epoch": 0.4, "learning_rate": 5.8567860514052264e-05, "loss": 1.0888, "step": 3899 }, { "epoch": 0.4, "learning_rate": 5.856712983194329e-05, "loss": 0.7027, "step": 3900 }, { "epoch": 0.4, "learning_rate": 5.856639896804396e-05, "loss": 1.2971, "step": 3901 }, { "epoch": 0.4, "learning_rate": 5.856566792235895e-05, "loss": 1.3293, "step": 3902 }, { "epoch": 0.4, "learning_rate": 5.856493669489288e-05, "loss": 0.6546, "step": 3903 }, { "epoch": 0.4, "learning_rate": 5.856420528565043e-05, "loss": 0.2241, "step": 3904 }, { "epoch": 0.4, "learning_rate": 5.856347369463624e-05, "loss": 0.4564, "step": 3905 }, { "epoch": 0.4, "learning_rate": 5.8562741921854965e-05, "loss": 0.3811, "step": 3906 }, { "epoch": 0.4, "learning_rate": 5.856200996731127e-05, "loss": 0.7942, "step": 3907 }, { "epoch": 0.4, "learning_rate": 5.85612778310098e-05, "loss": 1.345, "step": 3908 }, { "epoch": 0.4, "learning_rate": 5.8560545512955244e-05, "loss": 0.572, "step": 3909 }, { "epoch": 0.4, "learning_rate": 5.855981301315223e-05, "loss": 1.493, "step": 3910 }, { "epoch": 0.4, "learning_rate": 5.855908033160543e-05, "loss": 0.2382, "step": 3911 }, { "epoch": 0.4, "learning_rate": 5.85583474683195e-05, "loss": 0.2887, "step": 3912 }, { "epoch": 0.4, "learning_rate": 5.8557614423299127e-05, "loss": 0.3698, "step": 3913 }, { "epoch": 0.4, "learning_rate": 5.8556881196548943e-05, "loss": 1.6532, "step": 3914 }, { "epoch": 0.4, "learning_rate": 5.855614778807365e-05, "loss": 0.2811, "step": 3915 }, { "epoch": 0.4, "learning_rate": 5.855541419787788e-05, "loss": 0.4652, "step": 3916 }, { "epoch": 0.4, "learning_rate": 5.855468042596632e-05, "loss": 0.5276, "step": 3917 }, { "epoch": 0.4, "learning_rate": 5.855394647234364e-05, "loss": 1.4757, "step": 3918 }, { "epoch": 0.4, "learning_rate": 5.85532123370145e-05, "loss": 0.6666, "step": 3919 }, { "epoch": 0.4, "learning_rate": 5.8552478019983586e-05, "loss": 1.3196, "step": 3920 }, { "epoch": 0.4, "learning_rate": 5.8551743521255566e-05, "loss": 0.2956, "step": 3921 }, { "epoch": 0.4, "learning_rate": 5.85510088408351e-05, "loss": 0.7341, "step": 3922 }, { "epoch": 0.4, "learning_rate": 5.8550273978726886e-05, "loss": 1.1432, "step": 3923 }, { "epoch": 0.4, "learning_rate": 5.85495389349356e-05, "loss": 1.4311, "step": 3924 }, { "epoch": 0.4, "learning_rate": 5.8548803709465885e-05, "loss": 1.2239, "step": 3925 }, { "epoch": 0.4, "learning_rate": 5.854806830232245e-05, "loss": 1.1793, "step": 3926 }, { "epoch": 0.4, "learning_rate": 5.8547332713509975e-05, "loss": 1.2584, "step": 3927 }, { "epoch": 0.4, "learning_rate": 5.8546596943033136e-05, "loss": 0.6804, "step": 3928 }, { "epoch": 0.4, "learning_rate": 5.854586099089661e-05, "loss": 0.9044, "step": 3929 }, { "epoch": 0.4, "learning_rate": 5.8545124857105086e-05, "loss": 0.8026, "step": 3930 }, { "epoch": 0.4, "learning_rate": 5.854438854166325e-05, "loss": 1.1375, "step": 3931 }, { "epoch": 0.4, "learning_rate": 5.854365204457577e-05, "loss": 0.2905, "step": 3932 }, { "epoch": 0.4, "learning_rate": 5.854291536584735e-05, "loss": 0.3244, "step": 3933 }, { "epoch": 0.4, "learning_rate": 5.854217850548269e-05, "loss": 1.9312, "step": 3934 }, { "epoch": 0.4, "learning_rate": 5.854144146348645e-05, "loss": 1.3267, "step": 3935 }, { "epoch": 0.4, "learning_rate": 5.8540704239863337e-05, "loss": 0.6767, "step": 3936 }, { "epoch": 0.4, "learning_rate": 5.853996683461804e-05, "loss": 0.4931, "step": 3937 }, { "epoch": 0.4, "learning_rate": 5.8539229247755256e-05, "loss": 0.8532, "step": 3938 }, { "epoch": 0.4, "learning_rate": 5.853849147927967e-05, "loss": 0.5892, "step": 3939 }, { "epoch": 0.4, "learning_rate": 5.853775352919598e-05, "loss": 0.505, "step": 3940 }, { "epoch": 0.4, "learning_rate": 5.853701539750889e-05, "loss": 0.9093, "step": 3941 }, { "epoch": 0.4, "learning_rate": 5.853627708422308e-05, "loss": 0.8243, "step": 3942 }, { "epoch": 0.4, "learning_rate": 5.8535538589343264e-05, "loss": 0.8791, "step": 3943 }, { "epoch": 0.4, "learning_rate": 5.853479991287414e-05, "loss": 0.7056, "step": 3944 }, { "epoch": 0.4, "learning_rate": 5.853406105482039e-05, "loss": 0.9335, "step": 3945 }, { "epoch": 0.4, "learning_rate": 5.8533322015186736e-05, "loss": 1.2047, "step": 3946 }, { "epoch": 0.4, "learning_rate": 5.853258279397789e-05, "loss": 0.3136, "step": 3947 }, { "epoch": 0.4, "learning_rate": 5.8531843391198526e-05, "loss": 0.4756, "step": 3948 }, { "epoch": 0.4, "learning_rate": 5.853110380685337e-05, "loss": 1.3142, "step": 3949 }, { "epoch": 0.4, "learning_rate": 5.8530364040947124e-05, "loss": 0.7319, "step": 3950 }, { "epoch": 0.4, "learning_rate": 5.852962409348449e-05, "loss": 0.7475, "step": 3951 }, { "epoch": 0.4, "learning_rate": 5.8528883964470184e-05, "loss": 0.7048, "step": 3952 }, { "epoch": 0.4, "learning_rate": 5.852814365390892e-05, "loss": 0.5577, "step": 3953 }, { "epoch": 0.4, "learning_rate": 5.852740316180539e-05, "loss": 0.6738, "step": 3954 }, { "epoch": 0.4, "learning_rate": 5.852666248816433e-05, "loss": 0.7314, "step": 3955 }, { "epoch": 0.4, "learning_rate": 5.852592163299044e-05, "loss": 1.2109, "step": 3956 }, { "epoch": 0.4, "learning_rate": 5.8525180596288425e-05, "loss": 0.9447, "step": 3957 }, { "epoch": 0.4, "learning_rate": 5.8524439378063024e-05, "loss": 0.5493, "step": 3958 }, { "epoch": 0.4, "learning_rate": 5.852369797831894e-05, "loss": 0.4364, "step": 3959 }, { "epoch": 0.4, "learning_rate": 5.852295639706089e-05, "loss": 0.4578, "step": 3960 }, { "epoch": 0.4, "learning_rate": 5.8522214634293594e-05, "loss": 1.6854, "step": 3961 }, { "epoch": 0.4, "learning_rate": 5.852147269002178e-05, "loss": 1.1755, "step": 3962 }, { "epoch": 0.4, "learning_rate": 5.852073056425016e-05, "loss": 0.9071, "step": 3963 }, { "epoch": 0.4, "learning_rate": 5.851998825698347e-05, "loss": 1.4036, "step": 3964 }, { "epoch": 0.4, "learning_rate": 5.8519245768226415e-05, "loss": 0.3117, "step": 3965 }, { "epoch": 0.4, "learning_rate": 5.851850309798373e-05, "loss": 0.6643, "step": 3966 }, { "epoch": 0.4, "learning_rate": 5.8517760246260147e-05, "loss": 0.4695, "step": 3967 }, { "epoch": 0.4, "learning_rate": 5.851701721306037e-05, "loss": 0.7685, "step": 3968 }, { "epoch": 0.4, "learning_rate": 5.851627399838916e-05, "loss": 1.2259, "step": 3969 }, { "epoch": 0.4, "learning_rate": 5.851553060225123e-05, "loss": 0.6485, "step": 3970 }, { "epoch": 0.4, "learning_rate": 5.851478702465131e-05, "loss": 0.9917, "step": 3971 }, { "epoch": 0.4, "learning_rate": 5.8514043265594126e-05, "loss": 0.7956, "step": 3972 }, { "epoch": 0.4, "learning_rate": 5.8513299325084426e-05, "loss": 1.1551, "step": 3973 }, { "epoch": 0.4, "learning_rate": 5.851255520312693e-05, "loss": 0.5984, "step": 3974 }, { "epoch": 0.4, "learning_rate": 5.851181089972638e-05, "loss": 1.1834, "step": 3975 }, { "epoch": 0.4, "learning_rate": 5.851106641488751e-05, "loss": 1.3141, "step": 3976 }, { "epoch": 0.4, "learning_rate": 5.851032174861506e-05, "loss": 0.5376, "step": 3977 }, { "epoch": 0.4, "learning_rate": 5.850957690091377e-05, "loss": 1.1534, "step": 3978 }, { "epoch": 0.4, "learning_rate": 5.850883187178838e-05, "loss": 1.2222, "step": 3979 }, { "epoch": 0.4, "learning_rate": 5.850808666124363e-05, "loss": 1.4715, "step": 3980 }, { "epoch": 0.4, "learning_rate": 5.8507341269284256e-05, "loss": 0.8175, "step": 3981 }, { "epoch": 0.4, "learning_rate": 5.850659569591502e-05, "loss": 0.5986, "step": 3982 }, { "epoch": 0.4, "learning_rate": 5.8505849941140644e-05, "loss": 0.8444, "step": 3983 }, { "epoch": 0.4, "learning_rate": 5.850510400496588e-05, "loss": 0.5559, "step": 3984 }, { "epoch": 0.4, "learning_rate": 5.850435788739548e-05, "loss": 1.3231, "step": 3985 }, { "epoch": 0.4, "learning_rate": 5.850361158843419e-05, "loss": 0.9525, "step": 3986 }, { "epoch": 0.4, "learning_rate": 5.850286510808676e-05, "loss": 0.3699, "step": 3987 }, { "epoch": 0.4, "learning_rate": 5.850211844635794e-05, "loss": 1.1132, "step": 3988 }, { "epoch": 0.41, "learning_rate": 5.8501371603252483e-05, "loss": 0.9205, "step": 3989 }, { "epoch": 0.41, "learning_rate": 5.850062457877514e-05, "loss": 1.28, "step": 3990 }, { "epoch": 0.41, "learning_rate": 5.849987737293065e-05, "loss": 1.3916, "step": 3991 }, { "epoch": 0.41, "learning_rate": 5.849912998572381e-05, "loss": 0.682, "step": 3992 }, { "epoch": 0.41, "learning_rate": 5.849838241715933e-05, "loss": 0.2956, "step": 3993 }, { "epoch": 0.41, "learning_rate": 5.849763466724199e-05, "loss": 0.8347, "step": 3994 }, { "epoch": 0.41, "learning_rate": 5.8496886735976534e-05, "loss": 0.2978, "step": 3995 }, { "epoch": 0.41, "learning_rate": 5.849613862336774e-05, "loss": 1.2135, "step": 3996 }, { "epoch": 0.41, "learning_rate": 5.849539032942036e-05, "loss": 0.7591, "step": 3997 }, { "epoch": 0.41, "learning_rate": 5.849464185413916e-05, "loss": 1.4763, "step": 3998 }, { "epoch": 0.41, "learning_rate": 5.8493893197528896e-05, "loss": 1.5159, "step": 3999 }, { "epoch": 0.41, "learning_rate": 5.849314435959433e-05, "loss": 1.5775, "step": 4000 }, { "epoch": 0.41, "learning_rate": 5.8492395340340246e-05, "loss": 0.8129, "step": 4001 }, { "epoch": 0.41, "learning_rate": 5.849164613977138e-05, "loss": 0.7379, "step": 4002 }, { "epoch": 0.41, "learning_rate": 5.849089675789252e-05, "loss": 0.6237, "step": 4003 }, { "epoch": 0.41, "learning_rate": 5.8490147194708446e-05, "loss": 0.3609, "step": 4004 }, { "epoch": 0.41, "learning_rate": 5.8489397450223905e-05, "loss": 0.735, "step": 4005 }, { "epoch": 0.41, "learning_rate": 5.848864752444367e-05, "loss": 1.3592, "step": 4006 }, { "epoch": 0.41, "learning_rate": 5.848789741737253e-05, "loss": 1.2962, "step": 4007 }, { "epoch": 0.41, "learning_rate": 5.848714712901524e-05, "loss": 1.0638, "step": 4008 }, { "epoch": 0.41, "learning_rate": 5.848639665937659e-05, "loss": 1.0956, "step": 4009 }, { "epoch": 0.41, "learning_rate": 5.8485646008461345e-05, "loss": 0.7366, "step": 4010 }, { "epoch": 0.41, "learning_rate": 5.848489517627429e-05, "loss": 0.7213, "step": 4011 }, { "epoch": 0.41, "learning_rate": 5.848414416282019e-05, "loss": 1.3403, "step": 4012 }, { "epoch": 0.41, "learning_rate": 5.848339296810383e-05, "loss": 0.5295, "step": 4013 }, { "epoch": 0.41, "learning_rate": 5.848264159213e-05, "loss": 0.8344, "step": 4014 }, { "epoch": 0.41, "learning_rate": 5.8481890034903476e-05, "loss": 0.8821, "step": 4015 }, { "epoch": 0.41, "learning_rate": 5.8481138296429035e-05, "loss": 0.8297, "step": 4016 }, { "epoch": 0.41, "learning_rate": 5.848038637671147e-05, "loss": 0.6218, "step": 4017 }, { "epoch": 0.41, "learning_rate": 5.8479634275755555e-05, "loss": 0.6468, "step": 4018 }, { "epoch": 0.41, "learning_rate": 5.8478881993566084e-05, "loss": 0.9067, "step": 4019 }, { "epoch": 0.41, "learning_rate": 5.8478129530147836e-05, "loss": 0.451, "step": 4020 }, { "epoch": 0.41, "learning_rate": 5.8477376885505615e-05, "loss": 0.8239, "step": 4021 }, { "epoch": 0.41, "learning_rate": 5.84766240596442e-05, "loss": 0.3918, "step": 4022 }, { "epoch": 0.41, "learning_rate": 5.847587105256837e-05, "loss": 0.5075, "step": 4023 }, { "epoch": 0.41, "learning_rate": 5.8475117864282934e-05, "loss": 0.6064, "step": 4024 }, { "epoch": 0.41, "learning_rate": 5.847436449479269e-05, "loss": 0.9601, "step": 4025 }, { "epoch": 0.41, "learning_rate": 5.847361094410242e-05, "loss": 0.9824, "step": 4026 }, { "epoch": 0.41, "learning_rate": 5.847285721221691e-05, "loss": 1.1474, "step": 4027 }, { "epoch": 0.41, "learning_rate": 5.847210329914098e-05, "loss": 1.4379, "step": 4028 }, { "epoch": 0.41, "learning_rate": 5.847134920487942e-05, "loss": 0.49, "step": 4029 }, { "epoch": 0.41, "learning_rate": 5.847059492943702e-05, "loss": 0.5068, "step": 4030 }, { "epoch": 0.41, "learning_rate": 5.846984047281857e-05, "loss": 0.3126, "step": 4031 }, { "epoch": 0.41, "learning_rate": 5.84690858350289e-05, "loss": 1.543, "step": 4032 }, { "epoch": 0.41, "learning_rate": 5.8468331016072805e-05, "loss": 1.5415, "step": 4033 }, { "epoch": 0.41, "learning_rate": 5.846757601595507e-05, "loss": 1.0241, "step": 4034 }, { "epoch": 0.41, "learning_rate": 5.846682083468051e-05, "loss": 0.3624, "step": 4035 }, { "epoch": 0.41, "learning_rate": 5.846606547225393e-05, "loss": 1.074, "step": 4036 }, { "epoch": 0.41, "learning_rate": 5.846530992868015e-05, "loss": 0.5837, "step": 4037 }, { "epoch": 0.41, "learning_rate": 5.846455420396396e-05, "loss": 0.369, "step": 4038 }, { "epoch": 0.41, "learning_rate": 5.846379829811017e-05, "loss": 1.031, "step": 4039 }, { "epoch": 0.41, "learning_rate": 5.84630422111236e-05, "loss": 2.1472, "step": 4040 }, { "epoch": 0.41, "learning_rate": 5.846228594300906e-05, "loss": 1.2481, "step": 4041 }, { "epoch": 0.41, "learning_rate": 5.846152949377136e-05, "loss": 1.023, "step": 4042 }, { "epoch": 0.41, "learning_rate": 5.846077286341531e-05, "loss": 1.1317, "step": 4043 }, { "epoch": 0.41, "learning_rate": 5.846001605194573e-05, "loss": 1.1744, "step": 4044 }, { "epoch": 0.41, "learning_rate": 5.8459259059367434e-05, "loss": 1.3358, "step": 4045 }, { "epoch": 0.41, "learning_rate": 5.845850188568524e-05, "loss": 1.3313, "step": 4046 }, { "epoch": 0.41, "learning_rate": 5.8457744530903976e-05, "loss": 0.6484, "step": 4047 }, { "epoch": 0.41, "learning_rate": 5.8456986995028436e-05, "loss": 1.4232, "step": 4048 }, { "epoch": 0.41, "learning_rate": 5.8456229278063465e-05, "loss": 0.2406, "step": 4049 }, { "epoch": 0.41, "learning_rate": 5.845547138001387e-05, "loss": 1.502, "step": 4050 }, { "epoch": 0.41, "learning_rate": 5.8454713300884484e-05, "loss": 0.8618, "step": 4051 }, { "epoch": 0.41, "learning_rate": 5.845395504068013e-05, "loss": 1.1782, "step": 4052 }, { "epoch": 0.41, "learning_rate": 5.845319659940563e-05, "loss": 0.4912, "step": 4053 }, { "epoch": 0.41, "learning_rate": 5.8452437977065804e-05, "loss": 1.4948, "step": 4054 }, { "epoch": 0.41, "learning_rate": 5.8451679173665495e-05, "loss": 0.4744, "step": 4055 }, { "epoch": 0.41, "learning_rate": 5.845092018920952e-05, "loss": 0.5421, "step": 4056 }, { "epoch": 0.41, "learning_rate": 5.845016102370271e-05, "loss": 0.5288, "step": 4057 }, { "epoch": 0.41, "learning_rate": 5.84494016771499e-05, "loss": 1.4919, "step": 4058 }, { "epoch": 0.41, "learning_rate": 5.8448642149555915e-05, "loss": 1.6738, "step": 4059 }, { "epoch": 0.41, "learning_rate": 5.844788244092561e-05, "loss": 0.6233, "step": 4060 }, { "epoch": 0.41, "learning_rate": 5.8447122551263786e-05, "loss": 1.7061, "step": 4061 }, { "epoch": 0.41, "learning_rate": 5.84463624805753e-05, "loss": 0.7954, "step": 4062 }, { "epoch": 0.41, "learning_rate": 5.844560222886499e-05, "loss": 1.2033, "step": 4063 }, { "epoch": 0.41, "learning_rate": 5.844484179613769e-05, "loss": 0.7024, "step": 4064 }, { "epoch": 0.41, "learning_rate": 5.844408118239823e-05, "loss": 1.4189, "step": 4065 }, { "epoch": 0.41, "learning_rate": 5.8443320387651464e-05, "loss": 0.7944, "step": 4066 }, { "epoch": 0.41, "learning_rate": 5.8442559411902225e-05, "loss": 0.4602, "step": 4067 }, { "epoch": 0.41, "learning_rate": 5.8441798255155355e-05, "loss": 0.7906, "step": 4068 }, { "epoch": 0.41, "learning_rate": 5.8441036917415703e-05, "loss": 0.3257, "step": 4069 }, { "epoch": 0.41, "learning_rate": 5.844027539868812e-05, "loss": 1.8791, "step": 4070 }, { "epoch": 0.41, "learning_rate": 5.8439513698977427e-05, "loss": 1.6566, "step": 4071 }, { "epoch": 0.41, "learning_rate": 5.8438751818288504e-05, "loss": 0.8907, "step": 4072 }, { "epoch": 0.41, "learning_rate": 5.843798975662617e-05, "loss": 0.8768, "step": 4073 }, { "epoch": 0.41, "learning_rate": 5.84372275139953e-05, "loss": 0.9634, "step": 4074 }, { "epoch": 0.41, "learning_rate": 5.843646509040072e-05, "loss": 1.3436, "step": 4075 }, { "epoch": 0.41, "learning_rate": 5.84357024858473e-05, "loss": 1.5085, "step": 4076 }, { "epoch": 0.41, "learning_rate": 5.843493970033989e-05, "loss": 0.3009, "step": 4077 }, { "epoch": 0.41, "learning_rate": 5.8434176733883336e-05, "loss": 1.1047, "step": 4078 }, { "epoch": 0.41, "learning_rate": 5.843341358648249e-05, "loss": 0.9482, "step": 4079 }, { "epoch": 0.41, "learning_rate": 5.843265025814224e-05, "loss": 0.607, "step": 4080 }, { "epoch": 0.41, "learning_rate": 5.84318867488674e-05, "loss": 1.0146, "step": 4081 }, { "epoch": 0.41, "learning_rate": 5.843112305866285e-05, "loss": 1.09, "step": 4082 }, { "epoch": 0.41, "learning_rate": 5.8430359187533464e-05, "loss": 1.3043, "step": 4083 }, { "epoch": 0.41, "learning_rate": 5.842959513548407e-05, "loss": 0.5756, "step": 4084 }, { "epoch": 0.41, "learning_rate": 5.842883090251956e-05, "loss": 1.2676, "step": 4085 }, { "epoch": 0.41, "learning_rate": 5.842806648864478e-05, "loss": 0.9291, "step": 4086 }, { "epoch": 0.42, "learning_rate": 5.84273018938646e-05, "loss": 0.8726, "step": 4087 }, { "epoch": 0.42, "learning_rate": 5.8426537118183886e-05, "loss": 0.8327, "step": 4088 }, { "epoch": 0.42, "learning_rate": 5.8425772161607505e-05, "loss": 0.4033, "step": 4089 }, { "epoch": 0.42, "learning_rate": 5.842500702414032e-05, "loss": 1.1266, "step": 4090 }, { "epoch": 0.42, "learning_rate": 5.84242417057872e-05, "loss": 0.4242, "step": 4091 }, { "epoch": 0.42, "learning_rate": 5.8423476206553024e-05, "loss": 1.5827, "step": 4092 }, { "epoch": 0.42, "learning_rate": 5.842271052644266e-05, "loss": 0.5406, "step": 4093 }, { "epoch": 0.42, "learning_rate": 5.842194466546097e-05, "loss": 1.2537, "step": 4094 }, { "epoch": 0.42, "learning_rate": 5.8421178623612844e-05, "loss": 0.4328, "step": 4095 }, { "epoch": 0.42, "learning_rate": 5.842041240090315e-05, "loss": 1.5114, "step": 4096 }, { "epoch": 0.42, "learning_rate": 5.8419645997336764e-05, "loss": 1.9024, "step": 4097 }, { "epoch": 0.42, "learning_rate": 5.841887941291855e-05, "loss": 1.171, "step": 4098 }, { "epoch": 0.42, "learning_rate": 5.841811264765342e-05, "loss": 0.7626, "step": 4099 }, { "epoch": 0.42, "learning_rate": 5.841734570154622e-05, "loss": 1.5871, "step": 4100 }, { "epoch": 0.42, "learning_rate": 5.841657857460185e-05, "loss": 2.0151, "step": 4101 }, { "epoch": 0.42, "learning_rate": 5.841581126682517e-05, "loss": 0.9318, "step": 4102 }, { "epoch": 0.42, "learning_rate": 5.841504377822108e-05, "loss": 0.822, "step": 4103 }, { "epoch": 0.42, "learning_rate": 5.841427610879447e-05, "loss": 0.4919, "step": 4104 }, { "epoch": 0.42, "learning_rate": 5.84135082585502e-05, "loss": 0.4683, "step": 4105 }, { "epoch": 0.42, "learning_rate": 5.841274022749319e-05, "loss": 1.4384, "step": 4106 }, { "epoch": 0.42, "learning_rate": 5.8411972015628306e-05, "loss": 0.5501, "step": 4107 }, { "epoch": 0.42, "learning_rate": 5.841120362296044e-05, "loss": 1.0473, "step": 4108 }, { "epoch": 0.42, "learning_rate": 5.8410435049494474e-05, "loss": 1.3487, "step": 4109 }, { "epoch": 0.42, "learning_rate": 5.840966629523532e-05, "loss": 0.8396, "step": 4110 }, { "epoch": 0.42, "learning_rate": 5.8408897360187846e-05, "loss": 1.5856, "step": 4111 }, { "epoch": 0.42, "learning_rate": 5.840812824435696e-05, "loss": 0.5829, "step": 4112 }, { "epoch": 0.42, "learning_rate": 5.8407358947747554e-05, "loss": 0.9163, "step": 4113 }, { "epoch": 0.42, "learning_rate": 5.8406589470364525e-05, "loss": 1.4817, "step": 4114 }, { "epoch": 0.42, "learning_rate": 5.8405819812212766e-05, "loss": 0.2666, "step": 4115 }, { "epoch": 0.42, "learning_rate": 5.8405049973297175e-05, "loss": 0.7987, "step": 4116 }, { "epoch": 0.42, "learning_rate": 5.840427995362265e-05, "loss": 1.1555, "step": 4117 }, { "epoch": 0.42, "learning_rate": 5.840350975319409e-05, "loss": 0.8843, "step": 4118 }, { "epoch": 0.42, "learning_rate": 5.8402739372016405e-05, "loss": 0.5758, "step": 4119 }, { "epoch": 0.42, "learning_rate": 5.840196881009449e-05, "loss": 0.2847, "step": 4120 }, { "epoch": 0.42, "learning_rate": 5.8401198067433256e-05, "loss": 1.089, "step": 4121 }, { "epoch": 0.42, "learning_rate": 5.840042714403759e-05, "loss": 0.3474, "step": 4122 }, { "epoch": 0.42, "learning_rate": 5.839965603991242e-05, "loss": 0.7069, "step": 4123 }, { "epoch": 0.42, "learning_rate": 5.839888475506263e-05, "loss": 0.4063, "step": 4124 }, { "epoch": 0.42, "learning_rate": 5.839811328949315e-05, "loss": 0.6072, "step": 4125 }, { "epoch": 0.42, "learning_rate": 5.8397341643208885e-05, "loss": 0.7783, "step": 4126 }, { "epoch": 0.42, "learning_rate": 5.8396569816214735e-05, "loss": 1.3911, "step": 4127 }, { "epoch": 0.42, "learning_rate": 5.8395797808515614e-05, "loss": 0.3689, "step": 4128 }, { "epoch": 0.42, "learning_rate": 5.839502562011645e-05, "loss": 0.8781, "step": 4129 }, { "epoch": 0.42, "learning_rate": 5.839425325102213e-05, "loss": 1.3232, "step": 4130 }, { "epoch": 0.42, "learning_rate": 5.839348070123759e-05, "loss": 0.4425, "step": 4131 }, { "epoch": 0.42, "learning_rate": 5.8392707970767744e-05, "loss": 1.1528, "step": 4132 }, { "epoch": 0.42, "learning_rate": 5.83919350596175e-05, "loss": 0.422, "step": 4133 }, { "epoch": 0.42, "learning_rate": 5.839116196779179e-05, "loss": 1.1722, "step": 4134 }, { "epoch": 0.42, "learning_rate": 5.8390388695295516e-05, "loss": 0.9734, "step": 4135 }, { "epoch": 0.42, "learning_rate": 5.838961524213361e-05, "loss": 1.3144, "step": 4136 }, { "epoch": 0.42, "learning_rate": 5.838884160831099e-05, "loss": 1.2016, "step": 4137 }, { "epoch": 0.42, "learning_rate": 5.838806779383259e-05, "loss": 1.2265, "step": 4138 }, { "epoch": 0.42, "learning_rate": 5.8387293798703324e-05, "loss": 1.0876, "step": 4139 }, { "epoch": 0.42, "learning_rate": 5.838651962292813e-05, "loss": 1.4529, "step": 4140 }, { "epoch": 0.42, "learning_rate": 5.8385745266511906e-05, "loss": 1.5215, "step": 4141 }, { "epoch": 0.42, "learning_rate": 5.8384970729459606e-05, "loss": 1.3575, "step": 4142 }, { "epoch": 0.42, "learning_rate": 5.838419601177615e-05, "loss": 1.1861, "step": 4143 }, { "epoch": 0.42, "learning_rate": 5.8383421113466465e-05, "loss": 0.5589, "step": 4144 }, { "epoch": 0.42, "learning_rate": 5.838264603453549e-05, "loss": 0.7368, "step": 4145 }, { "epoch": 0.42, "learning_rate": 5.838187077498815e-05, "loss": 1.2384, "step": 4146 }, { "epoch": 0.42, "learning_rate": 5.8381095334829375e-05, "loss": 1.5568, "step": 4147 }, { "epoch": 0.42, "learning_rate": 5.8380319714064115e-05, "loss": 0.4862, "step": 4148 }, { "epoch": 0.42, "learning_rate": 5.8379543912697296e-05, "loss": 0.366, "step": 4149 }, { "epoch": 0.42, "learning_rate": 5.837876793073386e-05, "loss": 1.1435, "step": 4150 }, { "epoch": 0.42, "learning_rate": 5.837799176817873e-05, "loss": 1.5426, "step": 4151 }, { "epoch": 0.42, "learning_rate": 5.8377215425036866e-05, "loss": 1.0456, "step": 4152 }, { "epoch": 0.42, "learning_rate": 5.837643890131319e-05, "loss": 0.5984, "step": 4153 }, { "epoch": 0.42, "learning_rate": 5.8375662197012654e-05, "loss": 0.7974, "step": 4154 }, { "epoch": 0.42, "learning_rate": 5.83748853121402e-05, "loss": 0.4154, "step": 4155 }, { "epoch": 0.42, "learning_rate": 5.837410824670077e-05, "loss": 0.5432, "step": 4156 }, { "epoch": 0.42, "learning_rate": 5.83733310006993e-05, "loss": 0.4187, "step": 4157 }, { "epoch": 0.42, "learning_rate": 5.837255357414076e-05, "loss": 1.2893, "step": 4158 }, { "epoch": 0.42, "learning_rate": 5.837177596703008e-05, "loss": 0.7774, "step": 4159 }, { "epoch": 0.42, "learning_rate": 5.8370998179372205e-05, "loss": 0.2779, "step": 4160 }, { "epoch": 0.42, "learning_rate": 5.837022021117209e-05, "loss": 1.4558, "step": 4161 }, { "epoch": 0.42, "learning_rate": 5.83694420624347e-05, "loss": 1.2215, "step": 4162 }, { "epoch": 0.42, "learning_rate": 5.836866373316496e-05, "loss": 0.8183, "step": 4163 }, { "epoch": 0.42, "learning_rate": 5.836788522336784e-05, "loss": 0.4592, "step": 4164 }, { "epoch": 0.42, "learning_rate": 5.8367106533048285e-05, "loss": 0.7108, "step": 4165 }, { "epoch": 0.42, "learning_rate": 5.836632766221126e-05, "loss": 1.289, "step": 4166 }, { "epoch": 0.42, "learning_rate": 5.836554861086172e-05, "loss": 1.0196, "step": 4167 }, { "epoch": 0.42, "learning_rate": 5.836476937900461e-05, "loss": 1.4813, "step": 4168 }, { "epoch": 0.42, "learning_rate": 5.8363989966644905e-05, "loss": 1.0259, "step": 4169 }, { "epoch": 0.42, "learning_rate": 5.836321037378755e-05, "loss": 0.3535, "step": 4170 }, { "epoch": 0.42, "learning_rate": 5.836243060043753e-05, "loss": 0.8456, "step": 4171 }, { "epoch": 0.42, "learning_rate": 5.8361650646599775e-05, "loss": 0.5973, "step": 4172 }, { "epoch": 0.42, "learning_rate": 5.836087051227927e-05, "loss": 0.5287, "step": 4173 }, { "epoch": 0.42, "learning_rate": 5.836009019748097e-05, "loss": 0.3003, "step": 4174 }, { "epoch": 0.42, "learning_rate": 5.8359309702209856e-05, "loss": 0.9623, "step": 4175 }, { "epoch": 0.42, "learning_rate": 5.835852902647088e-05, "loss": 0.9523, "step": 4176 }, { "epoch": 0.42, "learning_rate": 5.835774817026901e-05, "loss": 0.7222, "step": 4177 }, { "epoch": 0.42, "learning_rate": 5.835696713360922e-05, "loss": 0.6019, "step": 4178 }, { "epoch": 0.42, "learning_rate": 5.8356185916496474e-05, "loss": 0.4807, "step": 4179 }, { "epoch": 0.42, "learning_rate": 5.8355404518935756e-05, "loss": 1.2053, "step": 4180 }, { "epoch": 0.42, "learning_rate": 5.835462294093203e-05, "loss": 1.2986, "step": 4181 }, { "epoch": 0.42, "learning_rate": 5.835384118249027e-05, "loss": 1.1645, "step": 4182 }, { "epoch": 0.42, "learning_rate": 5.835305924361544e-05, "loss": 0.761, "step": 4183 }, { "epoch": 0.42, "learning_rate": 5.835227712431253e-05, "loss": 0.7344, "step": 4184 }, { "epoch": 0.42, "learning_rate": 5.835149482458653e-05, "loss": 1.5296, "step": 4185 }, { "epoch": 0.43, "learning_rate": 5.835071234444239e-05, "loss": 0.0704, "step": 4186 }, { "epoch": 0.43, "learning_rate": 5.8349929683885106e-05, "loss": 0.3132, "step": 4187 }, { "epoch": 0.43, "learning_rate": 5.834914684291965e-05, "loss": 1.417, "step": 4188 }, { "epoch": 0.43, "learning_rate": 5.834836382155101e-05, "loss": 1.2959, "step": 4189 }, { "epoch": 0.43, "learning_rate": 5.834758061978416e-05, "loss": 1.2256, "step": 4190 }, { "epoch": 0.43, "learning_rate": 5.83467972376241e-05, "loss": 0.488, "step": 4191 }, { "epoch": 0.43, "learning_rate": 5.834601367507581e-05, "loss": 0.6859, "step": 4192 }, { "epoch": 0.43, "learning_rate": 5.834522993214426e-05, "loss": 0.7648, "step": 4193 }, { "epoch": 0.43, "learning_rate": 5.834444600883446e-05, "loss": 1.0991, "step": 4194 }, { "epoch": 0.43, "learning_rate": 5.834366190515139e-05, "loss": 0.3456, "step": 4195 }, { "epoch": 0.43, "learning_rate": 5.834287762110003e-05, "loss": 1.7089, "step": 4196 }, { "epoch": 0.43, "learning_rate": 5.8342093156685373e-05, "loss": 1.4981, "step": 4197 }, { "epoch": 0.43, "learning_rate": 5.834130851191244e-05, "loss": 0.6337, "step": 4198 }, { "epoch": 0.43, "learning_rate": 5.8340523686786183e-05, "loss": 1.299, "step": 4199 }, { "epoch": 0.43, "learning_rate": 5.833973868131162e-05, "loss": 1.2893, "step": 4200 }, { "epoch": 0.43, "learning_rate": 5.833895349549373e-05, "loss": 0.7145, "step": 4201 }, { "epoch": 0.43, "learning_rate": 5.833816812933753e-05, "loss": 1.1853, "step": 4202 }, { "epoch": 0.43, "learning_rate": 5.833738258284801e-05, "loss": 0.6302, "step": 4203 }, { "epoch": 0.43, "learning_rate": 5.8336596856030174e-05, "loss": 1.333, "step": 4204 }, { "epoch": 0.43, "learning_rate": 5.8335810948889e-05, "loss": 1.8101, "step": 4205 }, { "epoch": 0.43, "learning_rate": 5.8335024861429514e-05, "loss": 0.3582, "step": 4206 }, { "epoch": 0.43, "learning_rate": 5.833423859365671e-05, "loss": 0.6428, "step": 4207 }, { "epoch": 0.43, "learning_rate": 5.8333452145575584e-05, "loss": 1.2129, "step": 4208 }, { "epoch": 0.43, "learning_rate": 5.833266551719115e-05, "loss": 1.4624, "step": 4209 }, { "epoch": 0.43, "learning_rate": 5.8331878708508416e-05, "loss": 0.8217, "step": 4210 }, { "epoch": 0.43, "learning_rate": 5.8331091719532374e-05, "loss": 1.4818, "step": 4211 }, { "epoch": 0.43, "learning_rate": 5.8330304550268046e-05, "loss": 0.4478, "step": 4212 }, { "epoch": 0.43, "learning_rate": 5.8329517200720446e-05, "loss": 1.5228, "step": 4213 }, { "epoch": 0.43, "learning_rate": 5.8328729670894555e-05, "loss": 0.7577, "step": 4214 }, { "epoch": 0.43, "learning_rate": 5.832794196079542e-05, "loss": 0.7379, "step": 4215 }, { "epoch": 0.43, "learning_rate": 5.832715407042804e-05, "loss": 0.8559, "step": 4216 }, { "epoch": 0.43, "learning_rate": 5.832636599979743e-05, "loss": 0.2582, "step": 4217 }, { "epoch": 0.43, "learning_rate": 5.832557774890859e-05, "loss": 1.4149, "step": 4218 }, { "epoch": 0.43, "learning_rate": 5.832478931776656e-05, "loss": 1.1833, "step": 4219 }, { "epoch": 0.43, "learning_rate": 5.832400070637634e-05, "loss": 0.9916, "step": 4220 }, { "epoch": 0.43, "learning_rate": 5.832321191474296e-05, "loss": 0.6812, "step": 4221 }, { "epoch": 0.43, "learning_rate": 5.8322422942871435e-05, "loss": 0.5962, "step": 4222 }, { "epoch": 0.43, "learning_rate": 5.832163379076678e-05, "loss": 0.5274, "step": 4223 }, { "epoch": 0.43, "learning_rate": 5.832084445843402e-05, "loss": 0.0901, "step": 4224 }, { "epoch": 0.43, "learning_rate": 5.832005494587818e-05, "loss": 1.2744, "step": 4225 }, { "epoch": 0.43, "learning_rate": 5.831926525310429e-05, "loss": 0.9169, "step": 4226 }, { "epoch": 0.43, "learning_rate": 5.8318475380117364e-05, "loss": 0.427, "step": 4227 }, { "epoch": 0.43, "learning_rate": 5.8317685326922433e-05, "loss": 0.7651, "step": 4228 }, { "epoch": 0.43, "learning_rate": 5.831689509352453e-05, "loss": 0.6317, "step": 4229 }, { "epoch": 0.43, "learning_rate": 5.8316104679928675e-05, "loss": 0.3594, "step": 4230 }, { "epoch": 0.43, "learning_rate": 5.83153140861399e-05, "loss": 0.2327, "step": 4231 }, { "epoch": 0.43, "learning_rate": 5.831452331216324e-05, "loss": 0.8213, "step": 4232 }, { "epoch": 0.43, "learning_rate": 5.831373235800373e-05, "loss": 1.099, "step": 4233 }, { "epoch": 0.43, "learning_rate": 5.8312941223666396e-05, "loss": 0.9793, "step": 4234 }, { "epoch": 0.43, "learning_rate": 5.831214990915626e-05, "loss": 0.7167, "step": 4235 }, { "epoch": 0.43, "learning_rate": 5.831135841447839e-05, "loss": 1.0419, "step": 4236 }, { "epoch": 0.43, "learning_rate": 5.83105667396378e-05, "loss": 0.8918, "step": 4237 }, { "epoch": 0.43, "learning_rate": 5.8309774884639534e-05, "loss": 0.8127, "step": 4238 }, { "epoch": 0.43, "learning_rate": 5.830898284948863e-05, "loss": 1.3018, "step": 4239 }, { "epoch": 0.43, "learning_rate": 5.8308190634190127e-05, "loss": 0.5224, "step": 4240 }, { "epoch": 0.43, "learning_rate": 5.8307398238749075e-05, "loss": 1.1143, "step": 4241 }, { "epoch": 0.43, "learning_rate": 5.8306605663170496e-05, "loss": 0.5994, "step": 4242 }, { "epoch": 0.43, "learning_rate": 5.8305812907459464e-05, "loss": 0.4508, "step": 4243 }, { "epoch": 0.43, "learning_rate": 5.8305019971620994e-05, "loss": 0.837, "step": 4244 }, { "epoch": 0.43, "learning_rate": 5.830422685566015e-05, "loss": 0.5042, "step": 4245 }, { "epoch": 0.43, "learning_rate": 5.830343355958197e-05, "loss": 1.2279, "step": 4246 }, { "epoch": 0.43, "learning_rate": 5.830264008339151e-05, "loss": 1.0839, "step": 4247 }, { "epoch": 0.43, "learning_rate": 5.830184642709381e-05, "loss": 0.6157, "step": 4248 }, { "epoch": 0.43, "learning_rate": 5.830105259069393e-05, "loss": 1.2016, "step": 4249 }, { "epoch": 0.43, "learning_rate": 5.830025857419691e-05, "loss": 0.962, "step": 4250 }, { "epoch": 0.43, "learning_rate": 5.829946437760782e-05, "loss": 0.6216, "step": 4251 }, { "epoch": 0.43, "learning_rate": 5.82986700009317e-05, "loss": 0.5361, "step": 4252 }, { "epoch": 0.43, "learning_rate": 5.829787544417361e-05, "loss": 1.1081, "step": 4253 }, { "epoch": 0.43, "learning_rate": 5.8297080707338607e-05, "loss": 0.259, "step": 4254 }, { "epoch": 0.43, "learning_rate": 5.829628579043175e-05, "loss": 1.4929, "step": 4255 }, { "epoch": 0.43, "learning_rate": 5.829549069345809e-05, "loss": 0.5148, "step": 4256 }, { "epoch": 0.43, "learning_rate": 5.829469541642269e-05, "loss": 0.7714, "step": 4257 }, { "epoch": 0.43, "learning_rate": 5.8293899959330615e-05, "loss": 0.4428, "step": 4258 }, { "epoch": 0.43, "learning_rate": 5.829310432218693e-05, "loss": 0.5385, "step": 4259 }, { "epoch": 0.43, "learning_rate": 5.829230850499668e-05, "loss": 0.7465, "step": 4260 }, { "epoch": 0.43, "learning_rate": 5.829151250776494e-05, "loss": 1.1664, "step": 4261 }, { "epoch": 0.43, "learning_rate": 5.829071633049679e-05, "loss": 0.6402, "step": 4262 }, { "epoch": 0.43, "learning_rate": 5.8289919973197276e-05, "loss": 1.358, "step": 4263 }, { "epoch": 0.43, "learning_rate": 5.828912343587148e-05, "loss": 1.1341, "step": 4264 }, { "epoch": 0.43, "learning_rate": 5.8288326718524456e-05, "loss": 1.4698, "step": 4265 }, { "epoch": 0.43, "learning_rate": 5.828752982116129e-05, "loss": 0.641, "step": 4266 }, { "epoch": 0.43, "learning_rate": 5.828673274378704e-05, "loss": 1.2515, "step": 4267 }, { "epoch": 0.43, "learning_rate": 5.828593548640679e-05, "loss": 0.595, "step": 4268 }, { "epoch": 0.43, "learning_rate": 5.8285138049025595e-05, "loss": 0.3685, "step": 4269 }, { "epoch": 0.43, "learning_rate": 5.8284340431648555e-05, "loss": 0.449, "step": 4270 }, { "epoch": 0.43, "learning_rate": 5.828354263428072e-05, "loss": 1.3685, "step": 4271 }, { "epoch": 0.43, "learning_rate": 5.8282744656927194e-05, "loss": 0.9106, "step": 4272 }, { "epoch": 0.43, "learning_rate": 5.828194649959302e-05, "loss": 0.9129, "step": 4273 }, { "epoch": 0.43, "learning_rate": 5.828114816228332e-05, "loss": 1.3988, "step": 4274 }, { "epoch": 0.43, "learning_rate": 5.828034964500314e-05, "loss": 0.9929, "step": 4275 }, { "epoch": 0.43, "learning_rate": 5.827955094775757e-05, "loss": 1.911, "step": 4276 }, { "epoch": 0.43, "learning_rate": 5.827875207055171e-05, "loss": 1.1147, "step": 4277 }, { "epoch": 0.43, "learning_rate": 5.827795301339061e-05, "loss": 0.6811, "step": 4278 }, { "epoch": 0.43, "learning_rate": 5.827715377627939e-05, "loss": 0.7264, "step": 4279 }, { "epoch": 0.43, "learning_rate": 5.827635435922312e-05, "loss": 0.4307, "step": 4280 }, { "epoch": 0.43, "learning_rate": 5.827555476222687e-05, "loss": 0.5567, "step": 4281 }, { "epoch": 0.43, "learning_rate": 5.827475498529577e-05, "loss": 1.2304, "step": 4282 }, { "epoch": 0.43, "learning_rate": 5.8273955028434864e-05, "loss": 0.9962, "step": 4283 }, { "epoch": 0.44, "learning_rate": 5.827315489164928e-05, "loss": 0.2936, "step": 4284 }, { "epoch": 0.44, "learning_rate": 5.827235457494409e-05, "loss": 0.796, "step": 4285 }, { "epoch": 0.44, "learning_rate": 5.8271554078324385e-05, "loss": 0.3606, "step": 4286 }, { "epoch": 0.44, "learning_rate": 5.827075340179527e-05, "loss": 1.3008, "step": 4287 }, { "epoch": 0.44, "learning_rate": 5.8269952545361836e-05, "loss": 0.9632, "step": 4288 }, { "epoch": 0.44, "learning_rate": 5.826915150902917e-05, "loss": 1.151, "step": 4289 }, { "epoch": 0.44, "learning_rate": 5.826835029280239e-05, "loss": 1.3181, "step": 4290 }, { "epoch": 0.44, "learning_rate": 5.826754889668658e-05, "loss": 0.5716, "step": 4291 }, { "epoch": 0.44, "learning_rate": 5.8266747320686834e-05, "loss": 1.0764, "step": 4292 }, { "epoch": 0.44, "learning_rate": 5.826594556480827e-05, "loss": 0.5152, "step": 4293 }, { "epoch": 0.44, "learning_rate": 5.826514362905598e-05, "loss": 0.5841, "step": 4294 }, { "epoch": 0.44, "learning_rate": 5.826434151343506e-05, "loss": 0.9992, "step": 4295 }, { "epoch": 0.44, "learning_rate": 5.826353921795064e-05, "loss": 1.4805, "step": 4296 }, { "epoch": 0.44, "learning_rate": 5.82627367426078e-05, "loss": 1.2048, "step": 4297 }, { "epoch": 0.44, "learning_rate": 5.8261934087411644e-05, "loss": 1.0214, "step": 4298 }, { "epoch": 0.44, "learning_rate": 5.82611312523673e-05, "loss": 0.4267, "step": 4299 }, { "epoch": 0.44, "learning_rate": 5.826032823747987e-05, "loss": 1.2248, "step": 4300 }, { "epoch": 0.44, "learning_rate": 5.825952504275446e-05, "loss": 0.9772, "step": 4301 }, { "epoch": 0.44, "learning_rate": 5.825872166819618e-05, "loss": 0.8259, "step": 4302 }, { "epoch": 0.44, "learning_rate": 5.825791811381015e-05, "loss": 1.1414, "step": 4303 }, { "epoch": 0.44, "learning_rate": 5.8257114379601474e-05, "loss": 0.6575, "step": 4304 }, { "epoch": 0.44, "learning_rate": 5.8256310465575285e-05, "loss": 0.9216, "step": 4305 }, { "epoch": 0.44, "learning_rate": 5.8255506371736666e-05, "loss": 0.5151, "step": 4306 }, { "epoch": 0.44, "learning_rate": 5.825470209809077e-05, "loss": 0.544, "step": 4307 }, { "epoch": 0.44, "learning_rate": 5.8253897644642685e-05, "loss": 1.3926, "step": 4308 }, { "epoch": 0.44, "learning_rate": 5.8253093011397554e-05, "loss": 0.5925, "step": 4309 }, { "epoch": 0.44, "learning_rate": 5.825228819836048e-05, "loss": 0.3784, "step": 4310 }, { "epoch": 0.44, "learning_rate": 5.82514832055366e-05, "loss": 0.4083, "step": 4311 }, { "epoch": 0.44, "learning_rate": 5.825067803293102e-05, "loss": 1.8083, "step": 4312 }, { "epoch": 0.44, "learning_rate": 5.8249872680548875e-05, "loss": 0.6315, "step": 4313 }, { "epoch": 0.44, "learning_rate": 5.824906714839529e-05, "loss": 0.5723, "step": 4314 }, { "epoch": 0.44, "learning_rate": 5.824826143647538e-05, "loss": 0.7026, "step": 4315 }, { "epoch": 0.44, "learning_rate": 5.824745554479429e-05, "loss": 1.2999, "step": 4316 }, { "epoch": 0.44, "learning_rate": 5.824664947335713e-05, "loss": 0.8583, "step": 4317 }, { "epoch": 0.44, "learning_rate": 5.8245843222169046e-05, "loss": 1.0799, "step": 4318 }, { "epoch": 0.44, "learning_rate": 5.8245036791235154e-05, "loss": 0.8851, "step": 4319 }, { "epoch": 0.44, "learning_rate": 5.824423018056059e-05, "loss": 1.2272, "step": 4320 }, { "epoch": 0.44, "learning_rate": 5.82434233901505e-05, "loss": 0.4859, "step": 4321 }, { "epoch": 0.44, "learning_rate": 5.824261642001e-05, "loss": 1.4541, "step": 4322 }, { "epoch": 0.44, "learning_rate": 5.8241809270144245e-05, "loss": 1.0099, "step": 4323 }, { "epoch": 0.44, "learning_rate": 5.824100194055836e-05, "loss": 0.579, "step": 4324 }, { "epoch": 0.44, "learning_rate": 5.824019443125746e-05, "loss": 1.0271, "step": 4325 }, { "epoch": 0.44, "learning_rate": 5.8239386742246723e-05, "loss": 0.633, "step": 4326 }, { "epoch": 0.44, "learning_rate": 5.823857887353127e-05, "loss": 1.353, "step": 4327 }, { "epoch": 0.44, "learning_rate": 5.8237770825116245e-05, "loss": 1.0469, "step": 4328 }, { "epoch": 0.44, "learning_rate": 5.823696259700678e-05, "loss": 0.8279, "step": 4329 }, { "epoch": 0.44, "learning_rate": 5.8236154189208035e-05, "loss": 0.9286, "step": 4330 }, { "epoch": 0.44, "learning_rate": 5.823534560172515e-05, "loss": 0.475, "step": 4331 }, { "epoch": 0.44, "learning_rate": 5.823453683456325e-05, "loss": 0.1999, "step": 4332 }, { "epoch": 0.44, "learning_rate": 5.823372788772751e-05, "loss": 1.0367, "step": 4333 }, { "epoch": 0.44, "learning_rate": 5.8232918761223074e-05, "loss": 0.3271, "step": 4334 }, { "epoch": 0.44, "learning_rate": 5.823210945505507e-05, "loss": 0.7751, "step": 4335 }, { "epoch": 0.44, "learning_rate": 5.8231299969228675e-05, "loss": 0.4809, "step": 4336 }, { "epoch": 0.44, "learning_rate": 5.8230490303749014e-05, "loss": 0.4573, "step": 4337 }, { "epoch": 0.44, "learning_rate": 5.8229680458621254e-05, "loss": 0.5104, "step": 4338 }, { "epoch": 0.44, "learning_rate": 5.822887043385055e-05, "loss": 0.2644, "step": 4339 }, { "epoch": 0.44, "learning_rate": 5.822806022944205e-05, "loss": 0.3403, "step": 4340 }, { "epoch": 0.44, "learning_rate": 5.8227249845400916e-05, "loss": 0.9038, "step": 4341 }, { "epoch": 0.44, "learning_rate": 5.822643928173231e-05, "loss": 0.6959, "step": 4342 }, { "epoch": 0.44, "learning_rate": 5.8225628538441365e-05, "loss": 1.0042, "step": 4343 }, { "epoch": 0.44, "learning_rate": 5.822481761553327e-05, "loss": 1.5711, "step": 4344 }, { "epoch": 0.44, "learning_rate": 5.822400651301317e-05, "loss": 0.5293, "step": 4345 }, { "epoch": 0.44, "learning_rate": 5.822319523088623e-05, "loss": 0.9098, "step": 4346 }, { "epoch": 0.44, "learning_rate": 5.822238376915761e-05, "loss": 0.8033, "step": 4347 }, { "epoch": 0.44, "learning_rate": 5.822157212783247e-05, "loss": 0.9025, "step": 4348 }, { "epoch": 0.44, "learning_rate": 5.8220760306916e-05, "loss": 1.2007, "step": 4349 }, { "epoch": 0.44, "learning_rate": 5.8219948306413335e-05, "loss": 0.3532, "step": 4350 }, { "epoch": 0.44, "learning_rate": 5.821913612632965e-05, "loss": 1.5123, "step": 4351 }, { "epoch": 0.44, "learning_rate": 5.821832376667013e-05, "loss": 1.2188, "step": 4352 }, { "epoch": 0.44, "learning_rate": 5.821751122743992e-05, "loss": 0.3207, "step": 4353 }, { "epoch": 0.44, "learning_rate": 5.82166985086442e-05, "loss": 1.0521, "step": 4354 }, { "epoch": 0.44, "learning_rate": 5.8215885610288165e-05, "loss": 0.8636, "step": 4355 }, { "epoch": 0.44, "learning_rate": 5.821507253237695e-05, "loss": 1.6131, "step": 4356 }, { "epoch": 0.44, "learning_rate": 5.821425927491575e-05, "loss": 0.5584, "step": 4357 }, { "epoch": 0.44, "learning_rate": 5.821344583790974e-05, "loss": 1.2205, "step": 4358 }, { "epoch": 0.44, "learning_rate": 5.82126322213641e-05, "loss": 0.4372, "step": 4359 }, { "epoch": 0.44, "learning_rate": 5.821181842528399e-05, "loss": 1.4569, "step": 4360 }, { "epoch": 0.44, "learning_rate": 5.8211004449674606e-05, "loss": 0.3378, "step": 4361 }, { "epoch": 0.44, "learning_rate": 5.8210190294541125e-05, "loss": 0.7785, "step": 4362 }, { "epoch": 0.44, "learning_rate": 5.8209375959888714e-05, "loss": 0.7586, "step": 4363 }, { "epoch": 0.44, "learning_rate": 5.820856144572257e-05, "loss": 0.7355, "step": 4364 }, { "epoch": 0.44, "learning_rate": 5.820774675204788e-05, "loss": 0.2721, "step": 4365 }, { "epoch": 0.44, "learning_rate": 5.820693187886981e-05, "loss": 1.0751, "step": 4366 }, { "epoch": 0.44, "learning_rate": 5.820611682619356e-05, "loss": 0.7535, "step": 4367 }, { "epoch": 0.44, "learning_rate": 5.8205301594024314e-05, "loss": 1.2515, "step": 4368 }, { "epoch": 0.44, "learning_rate": 5.820448618236726e-05, "loss": 0.4917, "step": 4369 }, { "epoch": 0.44, "learning_rate": 5.820367059122758e-05, "loss": 0.5659, "step": 4370 }, { "epoch": 0.44, "learning_rate": 5.820285482061047e-05, "loss": 0.6074, "step": 4371 }, { "epoch": 0.44, "learning_rate": 5.8202038870521123e-05, "loss": 0.4681, "step": 4372 }, { "epoch": 0.44, "learning_rate": 5.820122274096473e-05, "loss": 1.4503, "step": 4373 }, { "epoch": 0.44, "learning_rate": 5.820040643194649e-05, "loss": 1.2674, "step": 4374 }, { "epoch": 0.44, "learning_rate": 5.819958994347158e-05, "loss": 1.0212, "step": 4375 }, { "epoch": 0.44, "learning_rate": 5.819877327554521e-05, "loss": 0.9742, "step": 4376 }, { "epoch": 0.44, "learning_rate": 5.819795642817258e-05, "loss": 0.3151, "step": 4377 }, { "epoch": 0.44, "learning_rate": 5.819713940135888e-05, "loss": 1.1802, "step": 4378 }, { "epoch": 0.44, "learning_rate": 5.8196322195109314e-05, "loss": 1.1028, "step": 4379 }, { "epoch": 0.44, "learning_rate": 5.8195504809429075e-05, "loss": 1.5855, "step": 4380 }, { "epoch": 0.44, "learning_rate": 5.819468724432337e-05, "loss": 0.4004, "step": 4381 }, { "epoch": 0.44, "learning_rate": 5.8193869499797395e-05, "loss": 1.813, "step": 4382 }, { "epoch": 0.45, "learning_rate": 5.819305157585637e-05, "loss": 0.2971, "step": 4383 }, { "epoch": 0.45, "learning_rate": 5.819223347250549e-05, "loss": 1.514, "step": 4384 }, { "epoch": 0.45, "learning_rate": 5.819141518974996e-05, "loss": 0.4015, "step": 4385 }, { "epoch": 0.45, "learning_rate": 5.819059672759498e-05, "loss": 0.9506, "step": 4386 }, { "epoch": 0.45, "learning_rate": 5.818977808604578e-05, "loss": 1.0014, "step": 4387 }, { "epoch": 0.45, "learning_rate": 5.818895926510754e-05, "loss": 0.4924, "step": 4388 }, { "epoch": 0.45, "learning_rate": 5.8188140264785503e-05, "loss": 0.4966, "step": 4389 }, { "epoch": 0.45, "learning_rate": 5.8187321085084854e-05, "loss": 1.3572, "step": 4390 }, { "epoch": 0.45, "learning_rate": 5.8186501726010816e-05, "loss": 0.4825, "step": 4391 }, { "epoch": 0.45, "learning_rate": 5.8185682187568614e-05, "loss": 1.0268, "step": 4392 }, { "epoch": 0.45, "learning_rate": 5.818486246976345e-05, "loss": 1.0225, "step": 4393 }, { "epoch": 0.45, "learning_rate": 5.818404257260054e-05, "loss": 0.8759, "step": 4394 }, { "epoch": 0.45, "learning_rate": 5.8183222496085095e-05, "loss": 0.4399, "step": 4395 }, { "epoch": 0.45, "learning_rate": 5.818240224022236e-05, "loss": 1.0998, "step": 4396 }, { "epoch": 0.45, "learning_rate": 5.818158180501753e-05, "loss": 0.4764, "step": 4397 }, { "epoch": 0.45, "learning_rate": 5.818076119047584e-05, "loss": 0.5198, "step": 4398 }, { "epoch": 0.45, "learning_rate": 5.81799403966025e-05, "loss": 0.4705, "step": 4399 }, { "epoch": 0.45, "learning_rate": 5.817911942340274e-05, "loss": 1.0517, "step": 4400 }, { "epoch": 0.45, "learning_rate": 5.817829827088178e-05, "loss": 0.5805, "step": 4401 }, { "epoch": 0.45, "learning_rate": 5.817747693904487e-05, "loss": 0.81, "step": 4402 }, { "epoch": 0.45, "learning_rate": 5.817665542789719e-05, "loss": 1.2291, "step": 4403 }, { "epoch": 0.45, "learning_rate": 5.817583373744401e-05, "loss": 0.7772, "step": 4404 }, { "epoch": 0.45, "learning_rate": 5.817501186769053e-05, "loss": 1.1513, "step": 4405 }, { "epoch": 0.45, "learning_rate": 5.817418981864201e-05, "loss": 1.1218, "step": 4406 }, { "epoch": 0.45, "learning_rate": 5.817336759030365e-05, "loss": 1.6565, "step": 4407 }, { "epoch": 0.45, "learning_rate": 5.81725451826807e-05, "loss": 0.6271, "step": 4408 }, { "epoch": 0.45, "learning_rate": 5.8171722595778395e-05, "loss": 0.772, "step": 4409 }, { "epoch": 0.45, "learning_rate": 5.817089982960195e-05, "loss": 1.0942, "step": 4410 }, { "epoch": 0.45, "learning_rate": 5.8170076884156634e-05, "loss": 0.8823, "step": 4411 }, { "epoch": 0.45, "learning_rate": 5.816925375944765e-05, "loss": 1.2732, "step": 4412 }, { "epoch": 0.45, "learning_rate": 5.8168430455480264e-05, "loss": 0.3544, "step": 4413 }, { "epoch": 0.45, "learning_rate": 5.816760697225969e-05, "loss": 1.0212, "step": 4414 }, { "epoch": 0.45, "learning_rate": 5.8166783309791194e-05, "loss": 0.4155, "step": 4415 }, { "epoch": 0.45, "learning_rate": 5.8165959468079995e-05, "loss": 0.4673, "step": 4416 }, { "epoch": 0.45, "learning_rate": 5.816513544713135e-05, "loss": 1.4242, "step": 4417 }, { "epoch": 0.45, "learning_rate": 5.816431124695049e-05, "loss": 0.4091, "step": 4418 }, { "epoch": 0.45, "learning_rate": 5.816348686754267e-05, "loss": 0.5582, "step": 4419 }, { "epoch": 0.45, "learning_rate": 5.816266230891315e-05, "loss": 1.4113, "step": 4420 }, { "epoch": 0.45, "learning_rate": 5.816183757106713e-05, "loss": 0.8423, "step": 4421 }, { "epoch": 0.45, "learning_rate": 5.816101265400991e-05, "loss": 0.9216, "step": 4422 }, { "epoch": 0.45, "learning_rate": 5.816018755774671e-05, "loss": 0.4651, "step": 4423 }, { "epoch": 0.45, "learning_rate": 5.8159362282282797e-05, "loss": 0.4871, "step": 4424 }, { "epoch": 0.45, "learning_rate": 5.815853682762341e-05, "loss": 0.4178, "step": 4425 }, { "epoch": 0.45, "learning_rate": 5.815771119377381e-05, "loss": 0.4096, "step": 4426 }, { "epoch": 0.45, "learning_rate": 5.815688538073924e-05, "loss": 0.7131, "step": 4427 }, { "epoch": 0.45, "learning_rate": 5.815605938852497e-05, "loss": 0.5843, "step": 4428 }, { "epoch": 0.45, "learning_rate": 5.815523321713625e-05, "loss": 0.512, "step": 4429 }, { "epoch": 0.45, "learning_rate": 5.8154406866578324e-05, "loss": 0.5134, "step": 4430 }, { "epoch": 0.45, "learning_rate": 5.8153580336856476e-05, "loss": 1.0096, "step": 4431 }, { "epoch": 0.45, "learning_rate": 5.815275362797595e-05, "loss": 1.1955, "step": 4432 }, { "epoch": 0.45, "learning_rate": 5.815192673994201e-05, "loss": 0.8036, "step": 4433 }, { "epoch": 0.45, "learning_rate": 5.815109967275992e-05, "loss": 0.8903, "step": 4434 }, { "epoch": 0.45, "learning_rate": 5.815027242643494e-05, "loss": 1.2878, "step": 4435 }, { "epoch": 0.45, "learning_rate": 5.814944500097233e-05, "loss": 0.4881, "step": 4436 }, { "epoch": 0.45, "learning_rate": 5.814861739637736e-05, "loss": 0.5356, "step": 4437 }, { "epoch": 0.45, "learning_rate": 5.81477896126553e-05, "loss": 1.3708, "step": 4438 }, { "epoch": 0.45, "learning_rate": 5.814696164981142e-05, "loss": 1.2849, "step": 4439 }, { "epoch": 0.45, "learning_rate": 5.814613350785098e-05, "loss": 1.5471, "step": 4440 }, { "epoch": 0.45, "learning_rate": 5.8145305186779256e-05, "loss": 0.8752, "step": 4441 }, { "epoch": 0.45, "learning_rate": 5.814447668660151e-05, "loss": 0.5189, "step": 4442 }, { "epoch": 0.45, "learning_rate": 5.814364800732303e-05, "loss": 0.5611, "step": 4443 }, { "epoch": 0.45, "learning_rate": 5.814281914894907e-05, "loss": 0.9727, "step": 4444 }, { "epoch": 0.45, "learning_rate": 5.814199011148492e-05, "loss": 1.5502, "step": 4445 }, { "epoch": 0.45, "learning_rate": 5.8141160894935854e-05, "loss": 1.2189, "step": 4446 }, { "epoch": 0.45, "learning_rate": 5.8140331499307147e-05, "loss": 0.9115, "step": 4447 }, { "epoch": 0.45, "learning_rate": 5.8139501924604066e-05, "loss": 0.8241, "step": 4448 }, { "epoch": 0.45, "learning_rate": 5.8138672170831905e-05, "loss": 1.3197, "step": 4449 }, { "epoch": 0.45, "learning_rate": 5.813784223799594e-05, "loss": 0.562, "step": 4450 }, { "epoch": 0.45, "learning_rate": 5.813701212610145e-05, "loss": 1.3518, "step": 4451 }, { "epoch": 0.45, "learning_rate": 5.813618183515371e-05, "loss": 0.3428, "step": 4452 }, { "epoch": 0.45, "learning_rate": 5.813535136515803e-05, "loss": 0.5, "step": 4453 }, { "epoch": 0.45, "learning_rate": 5.813452071611967e-05, "loss": 1.05, "step": 4454 }, { "epoch": 0.45, "learning_rate": 5.8133689888043916e-05, "loss": 0.8166, "step": 4455 }, { "epoch": 0.45, "learning_rate": 5.813285888093607e-05, "loss": 1.0936, "step": 4456 }, { "epoch": 0.45, "learning_rate": 5.813202769480141e-05, "loss": 1.4965, "step": 4457 }, { "epoch": 0.45, "learning_rate": 5.8131196329645225e-05, "loss": 0.6444, "step": 4458 }, { "epoch": 0.45, "learning_rate": 5.813036478547281e-05, "loss": 0.2491, "step": 4459 }, { "epoch": 0.45, "learning_rate": 5.812953306228946e-05, "loss": 0.9075, "step": 4460 }, { "epoch": 0.45, "learning_rate": 5.8128701160100456e-05, "loss": 0.5219, "step": 4461 }, { "epoch": 0.45, "learning_rate": 5.8127869078911105e-05, "loss": 1.431, "step": 4462 }, { "epoch": 0.45, "learning_rate": 5.812703681872669e-05, "loss": 1.1114, "step": 4463 }, { "epoch": 0.45, "learning_rate": 5.812620437955251e-05, "loss": 0.3306, "step": 4464 }, { "epoch": 0.45, "learning_rate": 5.812537176139387e-05, "loss": 0.2944, "step": 4465 }, { "epoch": 0.45, "learning_rate": 5.812453896425606e-05, "loss": 0.8628, "step": 4466 }, { "epoch": 0.45, "learning_rate": 5.812370598814439e-05, "loss": 1.2377, "step": 4467 }, { "epoch": 0.45, "learning_rate": 5.8122872833064156e-05, "loss": 1.2914, "step": 4468 }, { "epoch": 0.45, "learning_rate": 5.812203949902064e-05, "loss": 0.1842, "step": 4469 }, { "epoch": 0.45, "learning_rate": 5.812120598601918e-05, "loss": 1.3441, "step": 4470 }, { "epoch": 0.45, "learning_rate": 5.812037229406505e-05, "loss": 1.5614, "step": 4471 }, { "epoch": 0.45, "learning_rate": 5.8119538423163575e-05, "loss": 0.5738, "step": 4472 }, { "epoch": 0.45, "learning_rate": 5.811870437332006e-05, "loss": 0.6658, "step": 4473 }, { "epoch": 0.45, "learning_rate": 5.81178701445398e-05, "loss": 0.5116, "step": 4474 }, { "epoch": 0.45, "learning_rate": 5.8117035736828115e-05, "loss": 0.5589, "step": 4475 }, { "epoch": 0.45, "learning_rate": 5.811620115019031e-05, "loss": 0.3154, "step": 4476 }, { "epoch": 0.45, "learning_rate": 5.8115366384631694e-05, "loss": 1.1496, "step": 4477 }, { "epoch": 0.45, "learning_rate": 5.811453144015759e-05, "loss": 0.7989, "step": 4478 }, { "epoch": 0.45, "learning_rate": 5.8113696316773295e-05, "loss": 0.265, "step": 4479 }, { "epoch": 0.45, "learning_rate": 5.811286101448413e-05, "loss": 1.1255, "step": 4480 }, { "epoch": 0.46, "learning_rate": 5.8112025533295424e-05, "loss": 1.2684, "step": 4481 }, { "epoch": 0.46, "learning_rate": 5.811118987321247e-05, "loss": 1.2147, "step": 4482 }, { "epoch": 0.46, "learning_rate": 5.81103540342406e-05, "loss": 1.0347, "step": 4483 }, { "epoch": 0.46, "learning_rate": 5.810951801638514e-05, "loss": 0.3934, "step": 4484 }, { "epoch": 0.46, "learning_rate": 5.810868181965139e-05, "loss": 0.9942, "step": 4485 }, { "epoch": 0.46, "learning_rate": 5.8107845444044686e-05, "loss": 0.5255, "step": 4486 }, { "epoch": 0.46, "learning_rate": 5.810700888957034e-05, "loss": 0.928, "step": 4487 }, { "epoch": 0.46, "learning_rate": 5.810617215623369e-05, "loss": 1.4162, "step": 4488 }, { "epoch": 0.46, "learning_rate": 5.810533524404005e-05, "loss": 1.0642, "step": 4489 }, { "epoch": 0.46, "learning_rate": 5.8104498152994744e-05, "loss": 1.1991, "step": 4490 }, { "epoch": 0.46, "learning_rate": 5.810366088310311e-05, "loss": 0.6571, "step": 4491 }, { "epoch": 0.46, "learning_rate": 5.810282343437046e-05, "loss": 0.9152, "step": 4492 }, { "epoch": 0.46, "learning_rate": 5.810198580680214e-05, "loss": 0.6199, "step": 4493 }, { "epoch": 0.46, "learning_rate": 5.8101148000403475e-05, "loss": 1.3667, "step": 4494 }, { "epoch": 0.46, "learning_rate": 5.810031001517979e-05, "loss": 0.741, "step": 4495 }, { "epoch": 0.46, "learning_rate": 5.8099471851136424e-05, "loss": 1.4875, "step": 4496 }, { "epoch": 0.46, "learning_rate": 5.809863350827871e-05, "loss": 0.9647, "step": 4497 }, { "epoch": 0.46, "learning_rate": 5.809779498661198e-05, "loss": 0.9572, "step": 4498 }, { "epoch": 0.46, "learning_rate": 5.809695628614157e-05, "loss": 0.3792, "step": 4499 }, { "epoch": 0.46, "learning_rate": 5.809611740687282e-05, "loss": 0.8118, "step": 4500 }, { "epoch": 0.46, "learning_rate": 5.8095278348811074e-05, "loss": 1.1146, "step": 4501 }, { "epoch": 0.46, "learning_rate": 5.809443911196166e-05, "loss": 0.6798, "step": 4502 }, { "epoch": 0.46, "learning_rate": 5.8093599696329915e-05, "loss": 1.2027, "step": 4503 }, { "epoch": 0.46, "learning_rate": 5.809276010192119e-05, "loss": 0.5569, "step": 4504 }, { "epoch": 0.46, "learning_rate": 5.809192032874083e-05, "loss": 1.3029, "step": 4505 }, { "epoch": 0.46, "learning_rate": 5.8091080376794186e-05, "loss": 1.1712, "step": 4506 }, { "epoch": 0.46, "learning_rate": 5.809024024608658e-05, "loss": 0.7223, "step": 4507 }, { "epoch": 0.46, "learning_rate": 5.8089399936623375e-05, "loss": 1.0885, "step": 4508 }, { "epoch": 0.46, "learning_rate": 5.808855944840992e-05, "loss": 0.4174, "step": 4509 }, { "epoch": 0.46, "learning_rate": 5.808771878145155e-05, "loss": 1.2145, "step": 4510 }, { "epoch": 0.46, "learning_rate": 5.808687793575363e-05, "loss": 0.9159, "step": 4511 }, { "epoch": 0.46, "learning_rate": 5.808603691132149e-05, "loss": 1.6274, "step": 4512 }, { "epoch": 0.46, "learning_rate": 5.808519570816051e-05, "loss": 0.6929, "step": 4513 }, { "epoch": 0.46, "learning_rate": 5.8084354326276024e-05, "loss": 1.3744, "step": 4514 }, { "epoch": 0.46, "learning_rate": 5.808351276567339e-05, "loss": 0.6713, "step": 4515 }, { "epoch": 0.46, "learning_rate": 5.8082671026357964e-05, "loss": 1.009, "step": 4516 }, { "epoch": 0.46, "learning_rate": 5.8081829108335103e-05, "loss": 0.8544, "step": 4517 }, { "epoch": 0.46, "learning_rate": 5.8080987011610165e-05, "loss": 0.9086, "step": 4518 }, { "epoch": 0.46, "learning_rate": 5.8080144736188515e-05, "loss": 0.4151, "step": 4519 }, { "epoch": 0.46, "learning_rate": 5.80793022820755e-05, "loss": 0.4483, "step": 4520 }, { "epoch": 0.46, "learning_rate": 5.807845964927648e-05, "loss": 1.0366, "step": 4521 }, { "epoch": 0.46, "learning_rate": 5.807761683779684e-05, "loss": 1.2418, "step": 4522 }, { "epoch": 0.46, "learning_rate": 5.8076773847641915e-05, "loss": 0.8193, "step": 4523 }, { "epoch": 0.46, "learning_rate": 5.807593067881709e-05, "loss": 0.3889, "step": 4524 }, { "epoch": 0.46, "learning_rate": 5.807508733132771e-05, "loss": 0.6067, "step": 4525 }, { "epoch": 0.46, "learning_rate": 5.8074243805179164e-05, "loss": 0.705, "step": 4526 }, { "epoch": 0.46, "learning_rate": 5.807340010037681e-05, "loss": 1.2983, "step": 4527 }, { "epoch": 0.46, "learning_rate": 5.8072556216926015e-05, "loss": 1.2973, "step": 4528 }, { "epoch": 0.46, "learning_rate": 5.8071712154832156e-05, "loss": 0.4752, "step": 4529 }, { "epoch": 0.46, "learning_rate": 5.80708679141006e-05, "loss": 1.4291, "step": 4530 }, { "epoch": 0.46, "learning_rate": 5.807002349473672e-05, "loss": 0.4292, "step": 4531 }, { "epoch": 0.46, "learning_rate": 5.8069178896745876e-05, "loss": 1.3923, "step": 4532 }, { "epoch": 0.46, "learning_rate": 5.806833412013347e-05, "loss": 0.6371, "step": 4533 }, { "epoch": 0.46, "learning_rate": 5.8067489164904856e-05, "loss": 1.1667, "step": 4534 }, { "epoch": 0.46, "learning_rate": 5.8066644031065426e-05, "loss": 1.4652, "step": 4535 }, { "epoch": 0.46, "learning_rate": 5.806579871862055e-05, "loss": 1.4126, "step": 4536 }, { "epoch": 0.46, "learning_rate": 5.806495322757559e-05, "loss": 0.3559, "step": 4537 }, { "epoch": 0.46, "learning_rate": 5.806410755793596e-05, "loss": 1.3343, "step": 4538 }, { "epoch": 0.46, "learning_rate": 5.806326170970703e-05, "loss": 1.3992, "step": 4539 }, { "epoch": 0.46, "learning_rate": 5.806241568289417e-05, "loss": 0.9757, "step": 4540 }, { "epoch": 0.46, "learning_rate": 5.806156947750277e-05, "loss": 0.382, "step": 4541 }, { "epoch": 0.46, "learning_rate": 5.806072309353822e-05, "loss": 0.9225, "step": 4542 }, { "epoch": 0.46, "learning_rate": 5.805987653100591e-05, "loss": 0.9262, "step": 4543 }, { "epoch": 0.46, "learning_rate": 5.8059029789911214e-05, "loss": 0.4299, "step": 4544 }, { "epoch": 0.46, "learning_rate": 5.805818287025952e-05, "loss": 0.4727, "step": 4545 }, { "epoch": 0.46, "learning_rate": 5.805733577205623e-05, "loss": 1.1527, "step": 4546 }, { "epoch": 0.46, "learning_rate": 5.805648849530673e-05, "loss": 1.2572, "step": 4547 }, { "epoch": 0.46, "learning_rate": 5.805564104001641e-05, "loss": 1.0621, "step": 4548 }, { "epoch": 0.46, "learning_rate": 5.8054793406190665e-05, "loss": 0.4635, "step": 4549 }, { "epoch": 0.46, "learning_rate": 5.8053945593834884e-05, "loss": 0.3694, "step": 4550 }, { "epoch": 0.46, "learning_rate": 5.805309760295447e-05, "loss": 0.7575, "step": 4551 }, { "epoch": 0.46, "learning_rate": 5.8052249433554804e-05, "loss": 0.9029, "step": 4552 }, { "epoch": 0.46, "learning_rate": 5.8051401085641306e-05, "loss": 0.5052, "step": 4553 }, { "epoch": 0.46, "learning_rate": 5.8050552559219356e-05, "loss": 1.3405, "step": 4554 }, { "epoch": 0.46, "learning_rate": 5.804970385429437e-05, "loss": 0.3251, "step": 4555 }, { "epoch": 0.46, "learning_rate": 5.804885497087173e-05, "loss": 0.3268, "step": 4556 }, { "epoch": 0.46, "learning_rate": 5.804800590895685e-05, "loss": 0.8816, "step": 4557 }, { "epoch": 0.46, "learning_rate": 5.8047156668555125e-05, "loss": 1.0109, "step": 4558 }, { "epoch": 0.46, "learning_rate": 5.804630724967197e-05, "loss": 1.5707, "step": 4559 }, { "epoch": 0.46, "learning_rate": 5.804545765231279e-05, "loss": 0.4072, "step": 4560 }, { "epoch": 0.46, "learning_rate": 5.804460787648298e-05, "loss": 0.6141, "step": 4561 }, { "epoch": 0.46, "learning_rate": 5.8043757922187955e-05, "loss": 1.9025, "step": 4562 }, { "epoch": 0.46, "learning_rate": 5.804290778943312e-05, "loss": 0.4705, "step": 4563 }, { "epoch": 0.46, "learning_rate": 5.804205747822389e-05, "loss": 0.2603, "step": 4564 }, { "epoch": 0.46, "learning_rate": 5.804120698856569e-05, "loss": 1.4475, "step": 4565 }, { "epoch": 0.46, "learning_rate": 5.804035632046389e-05, "loss": 0.9056, "step": 4566 }, { "epoch": 0.46, "learning_rate": 5.803950547392394e-05, "loss": 0.4449, "step": 4567 }, { "epoch": 0.46, "learning_rate": 5.803865444895125e-05, "loss": 1.3844, "step": 4568 }, { "epoch": 0.46, "learning_rate": 5.8037803245551216e-05, "loss": 1.3092, "step": 4569 }, { "epoch": 0.46, "learning_rate": 5.803695186372928e-05, "loss": 0.5906, "step": 4570 }, { "epoch": 0.46, "learning_rate": 5.803610030349085e-05, "loss": 1.4666, "step": 4571 }, { "epoch": 0.46, "learning_rate": 5.803524856484133e-05, "loss": 0.3632, "step": 4572 }, { "epoch": 0.46, "learning_rate": 5.803439664778617e-05, "loss": 0.8167, "step": 4573 }, { "epoch": 0.46, "learning_rate": 5.803354455233076e-05, "loss": 0.5114, "step": 4574 }, { "epoch": 0.46, "learning_rate": 5.803269227848055e-05, "loss": 0.5058, "step": 4575 }, { "epoch": 0.46, "learning_rate": 5.8031839826240933e-05, "loss": 0.5727, "step": 4576 }, { "epoch": 0.46, "learning_rate": 5.8030987195617365e-05, "loss": 0.9648, "step": 4577 }, { "epoch": 0.46, "learning_rate": 5.803013438661525e-05, "loss": 1.0088, "step": 4578 }, { "epoch": 0.46, "learning_rate": 5.802928139924003e-05, "loss": 1.0925, "step": 4579 }, { "epoch": 0.47, "learning_rate": 5.8028428233497126e-05, "loss": 0.4521, "step": 4580 }, { "epoch": 0.47, "learning_rate": 5.802757488939197e-05, "loss": 1.5301, "step": 4581 }, { "epoch": 0.47, "learning_rate": 5.8026721366929986e-05, "loss": 0.7829, "step": 4582 }, { "epoch": 0.47, "learning_rate": 5.802586766611661e-05, "loss": 0.5005, "step": 4583 }, { "epoch": 0.47, "learning_rate": 5.802501378695727e-05, "loss": 0.3935, "step": 4584 }, { "epoch": 0.47, "learning_rate": 5.80241597294574e-05, "loss": 0.9835, "step": 4585 }, { "epoch": 0.47, "learning_rate": 5.802330549362245e-05, "loss": 0.4869, "step": 4586 }, { "epoch": 0.47, "learning_rate": 5.802245107945783e-05, "loss": 2.0095, "step": 4587 }, { "epoch": 0.47, "learning_rate": 5.8021596486969e-05, "loss": 1.2369, "step": 4588 }, { "epoch": 0.47, "learning_rate": 5.802074171616139e-05, "loss": 0.4486, "step": 4589 }, { "epoch": 0.47, "learning_rate": 5.8019886767040435e-05, "loss": 0.3502, "step": 4590 }, { "epoch": 0.47, "learning_rate": 5.8019031639611586e-05, "loss": 0.6838, "step": 4591 }, { "epoch": 0.47, "learning_rate": 5.8018176333880276e-05, "loss": 0.3897, "step": 4592 }, { "epoch": 0.47, "learning_rate": 5.801732084985195e-05, "loss": 0.4246, "step": 4593 }, { "epoch": 0.47, "learning_rate": 5.801646518753206e-05, "loss": 0.8474, "step": 4594 }, { "epoch": 0.47, "learning_rate": 5.801560934692604e-05, "loss": 1.7773, "step": 4595 }, { "epoch": 0.47, "learning_rate": 5.801475332803933e-05, "loss": 1.4932, "step": 4596 }, { "epoch": 0.47, "learning_rate": 5.8013897130877394e-05, "loss": 0.8464, "step": 4597 }, { "epoch": 0.47, "learning_rate": 5.801304075544567e-05, "loss": 2.1701, "step": 4598 }, { "epoch": 0.47, "learning_rate": 5.8012184201749624e-05, "loss": 1.2559, "step": 4599 }, { "epoch": 0.47, "learning_rate": 5.801132746979468e-05, "loss": 0.4527, "step": 4600 }, { "epoch": 0.47, "learning_rate": 5.801047055958631e-05, "loss": 0.7374, "step": 4601 }, { "epoch": 0.47, "learning_rate": 5.8009613471129954e-05, "loss": 1.5675, "step": 4602 }, { "epoch": 0.47, "learning_rate": 5.800875620443108e-05, "loss": 1.2448, "step": 4603 }, { "epoch": 0.47, "learning_rate": 5.800789875949513e-05, "loss": 0.8076, "step": 4604 }, { "epoch": 0.47, "learning_rate": 5.800704113632758e-05, "loss": 1.5001, "step": 4605 }, { "epoch": 0.47, "learning_rate": 5.800618333493387e-05, "loss": 1.7026, "step": 4606 }, { "epoch": 0.47, "learning_rate": 5.800532535531946e-05, "loss": 0.7612, "step": 4607 }, { "epoch": 0.47, "learning_rate": 5.800446719748981e-05, "loss": 0.5772, "step": 4608 }, { "epoch": 0.47, "learning_rate": 5.800360886145039e-05, "loss": 0.735, "step": 4609 }, { "epoch": 0.47, "learning_rate": 5.8002750347206655e-05, "loss": 0.6182, "step": 4610 }, { "epoch": 0.47, "learning_rate": 5.800189165476407e-05, "loss": 0.7969, "step": 4611 }, { "epoch": 0.47, "learning_rate": 5.8001032784128096e-05, "loss": 0.3371, "step": 4612 }, { "epoch": 0.47, "learning_rate": 5.8000173735304206e-05, "loss": 1.4578, "step": 4613 }, { "epoch": 0.47, "learning_rate": 5.799931450829786e-05, "loss": 0.868, "step": 4614 }, { "epoch": 0.47, "learning_rate": 5.799845510311453e-05, "loss": 0.192, "step": 4615 }, { "epoch": 0.47, "learning_rate": 5.799759551975968e-05, "loss": 0.7148, "step": 4616 }, { "epoch": 0.47, "learning_rate": 5.7996735758238784e-05, "loss": 0.6043, "step": 4617 }, { "epoch": 0.47, "learning_rate": 5.799587581855731e-05, "loss": 1.2046, "step": 4618 }, { "epoch": 0.47, "learning_rate": 5.799501570072074e-05, "loss": 0.387, "step": 4619 }, { "epoch": 0.47, "learning_rate": 5.799415540473453e-05, "loss": 1.2669, "step": 4620 }, { "epoch": 0.47, "learning_rate": 5.799329493060417e-05, "loss": 0.7032, "step": 4621 }, { "epoch": 0.47, "learning_rate": 5.799243427833513e-05, "loss": 0.4525, "step": 4622 }, { "epoch": 0.47, "learning_rate": 5.7991573447932884e-05, "loss": 1.7875, "step": 4623 }, { "epoch": 0.47, "learning_rate": 5.799071243940291e-05, "loss": 1.5165, "step": 4624 }, { "epoch": 0.47, "learning_rate": 5.79898512527507e-05, "loss": 0.9732, "step": 4625 }, { "epoch": 0.47, "learning_rate": 5.798898988798171e-05, "loss": 0.3329, "step": 4626 }, { "epoch": 0.47, "learning_rate": 5.798812834510145e-05, "loss": 0.8181, "step": 4627 }, { "epoch": 0.47, "learning_rate": 5.798726662411539e-05, "loss": 0.7309, "step": 4628 }, { "epoch": 0.47, "learning_rate": 5.798640472502901e-05, "loss": 1.7384, "step": 4629 }, { "epoch": 0.47, "learning_rate": 5.7985542647847794e-05, "loss": 0.7101, "step": 4630 }, { "epoch": 0.47, "learning_rate": 5.7984680392577226e-05, "loss": 0.3681, "step": 4631 }, { "epoch": 0.47, "learning_rate": 5.79838179592228e-05, "loss": 0.668, "step": 4632 }, { "epoch": 0.47, "learning_rate": 5.798295534779001e-05, "loss": 0.3078, "step": 4633 }, { "epoch": 0.47, "learning_rate": 5.798209255828433e-05, "loss": 0.4378, "step": 4634 }, { "epoch": 0.47, "learning_rate": 5.798122959071126e-05, "loss": 1.048, "step": 4635 }, { "epoch": 0.47, "learning_rate": 5.798036644507628e-05, "loss": 0.5578, "step": 4636 }, { "epoch": 0.47, "learning_rate": 5.797950312138491e-05, "loss": 0.6624, "step": 4637 }, { "epoch": 0.47, "learning_rate": 5.797863961964261e-05, "loss": 0.4924, "step": 4638 }, { "epoch": 0.47, "learning_rate": 5.79777759398549e-05, "loss": 1.247, "step": 4639 }, { "epoch": 0.47, "learning_rate": 5.7976912082027264e-05, "loss": 0.6043, "step": 4640 }, { "epoch": 0.47, "learning_rate": 5.7976048046165206e-05, "loss": 0.7782, "step": 4641 }, { "epoch": 0.47, "learning_rate": 5.797518383227422e-05, "loss": 1.2017, "step": 4642 }, { "epoch": 0.47, "learning_rate": 5.79743194403598e-05, "loss": 1.2905, "step": 4643 }, { "epoch": 0.47, "learning_rate": 5.797345487042746e-05, "loss": 0.9793, "step": 4644 }, { "epoch": 0.47, "learning_rate": 5.79725901224827e-05, "loss": 1.4879, "step": 4645 }, { "epoch": 0.47, "learning_rate": 5.7971725196531005e-05, "loss": 0.6704, "step": 4646 }, { "epoch": 0.47, "learning_rate": 5.7970860092577904e-05, "loss": 1.93, "step": 4647 }, { "epoch": 0.47, "learning_rate": 5.796999481062888e-05, "loss": 1.2962, "step": 4648 }, { "epoch": 0.47, "learning_rate": 5.796912935068946e-05, "loss": 0.7179, "step": 4649 }, { "epoch": 0.47, "learning_rate": 5.796826371276513e-05, "loss": 0.3409, "step": 4650 }, { "epoch": 0.47, "learning_rate": 5.796739789686141e-05, "loss": 0.2462, "step": 4651 }, { "epoch": 0.47, "learning_rate": 5.7966531902983817e-05, "loss": 1.2003, "step": 4652 }, { "epoch": 0.47, "learning_rate": 5.796566573113785e-05, "loss": 0.4994, "step": 4653 }, { "epoch": 0.47, "learning_rate": 5.7964799381329024e-05, "loss": 1.1204, "step": 4654 }, { "epoch": 0.47, "learning_rate": 5.796393285356285e-05, "loss": 1.291, "step": 4655 }, { "epoch": 0.47, "learning_rate": 5.7963066147844845e-05, "loss": 1.4831, "step": 4656 }, { "epoch": 0.47, "learning_rate": 5.7962199264180526e-05, "loss": 1.1702, "step": 4657 }, { "epoch": 0.47, "learning_rate": 5.7961332202575415e-05, "loss": 1.0728, "step": 4658 }, { "epoch": 0.47, "learning_rate": 5.796046496303502e-05, "loss": 0.1574, "step": 4659 }, { "epoch": 0.47, "learning_rate": 5.795959754556486e-05, "loss": 1.0623, "step": 4660 }, { "epoch": 0.47, "learning_rate": 5.795872995017046e-05, "loss": 1.4677, "step": 4661 }, { "epoch": 0.47, "learning_rate": 5.795786217685734e-05, "loss": 0.6109, "step": 4662 }, { "epoch": 0.47, "learning_rate": 5.7956994225631015e-05, "loss": 0.8004, "step": 4663 }, { "epoch": 0.47, "learning_rate": 5.795612609649702e-05, "loss": 0.5664, "step": 4664 }, { "epoch": 0.47, "learning_rate": 5.795525778946087e-05, "loss": 1.0942, "step": 4665 }, { "epoch": 0.47, "learning_rate": 5.7954389304528096e-05, "loss": 0.3384, "step": 4666 }, { "epoch": 0.47, "learning_rate": 5.795352064170422e-05, "loss": 0.3932, "step": 4667 }, { "epoch": 0.47, "learning_rate": 5.795265180099478e-05, "loss": 1.216, "step": 4668 }, { "epoch": 0.47, "learning_rate": 5.795178278240529e-05, "loss": 1.339, "step": 4669 }, { "epoch": 0.47, "learning_rate": 5.7950913585941286e-05, "loss": 0.4119, "step": 4670 }, { "epoch": 0.47, "learning_rate": 5.795004421160831e-05, "loss": 1.0588, "step": 4671 }, { "epoch": 0.47, "learning_rate": 5.794917465941188e-05, "loss": 1.8467, "step": 4672 }, { "epoch": 0.47, "learning_rate": 5.7948304929357536e-05, "loss": 0.9383, "step": 4673 }, { "epoch": 0.47, "learning_rate": 5.794743502145081e-05, "loss": 1.1007, "step": 4674 }, { "epoch": 0.47, "learning_rate": 5.794656493569724e-05, "loss": 1.4818, "step": 4675 }, { "epoch": 0.47, "learning_rate": 5.794569467210236e-05, "loss": 0.6596, "step": 4676 }, { "epoch": 0.47, "learning_rate": 5.794482423067172e-05, "loss": 0.2362, "step": 4677 }, { "epoch": 0.48, "learning_rate": 5.7943953611410835e-05, "loss": 0.3762, "step": 4678 }, { "epoch": 0.48, "learning_rate": 5.794308281432527e-05, "loss": 1.2673, "step": 4679 }, { "epoch": 0.48, "learning_rate": 5.794221183942055e-05, "loss": 1.2109, "step": 4680 }, { "epoch": 0.48, "learning_rate": 5.794134068670223e-05, "loss": 0.3356, "step": 4681 }, { "epoch": 0.48, "learning_rate": 5.7940469356175844e-05, "loss": 0.7109, "step": 4682 }, { "epoch": 0.48, "learning_rate": 5.793959784784694e-05, "loss": 0.4231, "step": 4683 }, { "epoch": 0.48, "learning_rate": 5.7938726161721066e-05, "loss": 1.0656, "step": 4684 }, { "epoch": 0.48, "learning_rate": 5.793785429780376e-05, "loss": 0.8498, "step": 4685 }, { "epoch": 0.48, "learning_rate": 5.793698225610058e-05, "loss": 1.2623, "step": 4686 }, { "epoch": 0.48, "learning_rate": 5.793611003661708e-05, "loss": 1.2107, "step": 4687 }, { "epoch": 0.48, "learning_rate": 5.79352376393588e-05, "loss": 0.5182, "step": 4688 }, { "epoch": 0.48, "learning_rate": 5.793436506433129e-05, "loss": 0.6568, "step": 4689 }, { "epoch": 0.48, "learning_rate": 5.793349231154011e-05, "loss": 0.3186, "step": 4690 }, { "epoch": 0.48, "learning_rate": 5.793261938099081e-05, "loss": 1.1711, "step": 4691 }, { "epoch": 0.48, "learning_rate": 5.793174627268895e-05, "loss": 0.6561, "step": 4692 }, { "epoch": 0.48, "learning_rate": 5.7930872986640086e-05, "loss": 1.081, "step": 4693 }, { "epoch": 0.48, "learning_rate": 5.792999952284976e-05, "loss": 1.2555, "step": 4694 }, { "epoch": 0.48, "learning_rate": 5.792912588132355e-05, "loss": 0.4434, "step": 4695 }, { "epoch": 0.48, "learning_rate": 5.792825206206701e-05, "loss": 0.3451, "step": 4696 }, { "epoch": 0.48, "learning_rate": 5.79273780650857e-05, "loss": 1.4397, "step": 4697 }, { "epoch": 0.48, "learning_rate": 5.792650389038517e-05, "loss": 1.2494, "step": 4698 }, { "epoch": 0.48, "learning_rate": 5.792562953797099e-05, "loss": 0.6159, "step": 4699 }, { "epoch": 0.48, "learning_rate": 5.792475500784874e-05, "loss": 0.976, "step": 4700 }, { "epoch": 0.48, "learning_rate": 5.7923880300023964e-05, "loss": 0.455, "step": 4701 }, { "epoch": 0.48, "learning_rate": 5.792300541450224e-05, "loss": 0.6373, "step": 4702 }, { "epoch": 0.48, "learning_rate": 5.7922130351289135e-05, "loss": 0.6214, "step": 4703 }, { "epoch": 0.48, "learning_rate": 5.792125511039021e-05, "loss": 0.4587, "step": 4704 }, { "epoch": 0.48, "learning_rate": 5.792037969181104e-05, "loss": 1.5292, "step": 4705 }, { "epoch": 0.48, "learning_rate": 5.79195040955572e-05, "loss": 0.1956, "step": 4706 }, { "epoch": 0.48, "learning_rate": 5.791862832163425e-05, "loss": 0.7634, "step": 4707 }, { "epoch": 0.48, "learning_rate": 5.7917752370047775e-05, "loss": 1.0414, "step": 4708 }, { "epoch": 0.48, "learning_rate": 5.791687624080334e-05, "loss": 1.4056, "step": 4709 }, { "epoch": 0.48, "learning_rate": 5.7915999933906526e-05, "loss": 0.9966, "step": 4710 }, { "epoch": 0.48, "learning_rate": 5.791512344936291e-05, "loss": 0.2772, "step": 4711 }, { "epoch": 0.48, "learning_rate": 5.791424678717807e-05, "loss": 1.0258, "step": 4712 }, { "epoch": 0.48, "learning_rate": 5.791336994735757e-05, "loss": 0.5563, "step": 4713 }, { "epoch": 0.48, "learning_rate": 5.791249292990702e-05, "loss": 0.7782, "step": 4714 }, { "epoch": 0.48, "learning_rate": 5.791161573483197e-05, "loss": 0.8417, "step": 4715 }, { "epoch": 0.48, "learning_rate": 5.7910738362138025e-05, "loss": 1.5645, "step": 4716 }, { "epoch": 0.48, "learning_rate": 5.7909860811830756e-05, "loss": 1.0947, "step": 4717 }, { "epoch": 0.48, "learning_rate": 5.790898308391575e-05, "loss": 1.3276, "step": 4718 }, { "epoch": 0.48, "learning_rate": 5.790810517839859e-05, "loss": 0.6551, "step": 4719 }, { "epoch": 0.48, "learning_rate": 5.790722709528487e-05, "loss": 0.4704, "step": 4720 }, { "epoch": 0.48, "learning_rate": 5.790634883458017e-05, "loss": 1.4095, "step": 4721 }, { "epoch": 0.48, "learning_rate": 5.790547039629008e-05, "loss": 1.1531, "step": 4722 }, { "epoch": 0.48, "learning_rate": 5.790459178042019e-05, "loss": 0.4542, "step": 4723 }, { "epoch": 0.48, "learning_rate": 5.7903712986976104e-05, "loss": 0.8381, "step": 4724 }, { "epoch": 0.48, "learning_rate": 5.79028340159634e-05, "loss": 0.3418, "step": 4725 }, { "epoch": 0.48, "learning_rate": 5.7901954867387674e-05, "loss": 1.1713, "step": 4726 }, { "epoch": 0.48, "learning_rate": 5.790107554125452e-05, "loss": 1.1185, "step": 4727 }, { "epoch": 0.48, "learning_rate": 5.790019603756953e-05, "loss": 0.4483, "step": 4728 }, { "epoch": 0.48, "learning_rate": 5.789931635633831e-05, "loss": 1.1353, "step": 4729 }, { "epoch": 0.48, "learning_rate": 5.789843649756646e-05, "loss": 1.0254, "step": 4730 }, { "epoch": 0.48, "learning_rate": 5.789755646125957e-05, "loss": 1.5113, "step": 4731 }, { "epoch": 0.48, "learning_rate": 5.789667624742325e-05, "loss": 1.5753, "step": 4732 }, { "epoch": 0.48, "learning_rate": 5.789579585606308e-05, "loss": 1.4674, "step": 4733 }, { "epoch": 0.48, "learning_rate": 5.789491528718469e-05, "loss": 0.8589, "step": 4734 }, { "epoch": 0.48, "learning_rate": 5.789403454079366e-05, "loss": 0.5104, "step": 4735 }, { "epoch": 0.48, "learning_rate": 5.7893153616895615e-05, "loss": 0.6218, "step": 4736 }, { "epoch": 0.48, "learning_rate": 5.789227251549615e-05, "loss": 0.6215, "step": 4737 }, { "epoch": 0.48, "learning_rate": 5.789139123660087e-05, "loss": 0.3307, "step": 4738 }, { "epoch": 0.48, "learning_rate": 5.7890509780215386e-05, "loss": 0.6158, "step": 4739 }, { "epoch": 0.48, "learning_rate": 5.788962814634531e-05, "loss": 0.6867, "step": 4740 }, { "epoch": 0.48, "learning_rate": 5.788874633499625e-05, "loss": 0.2887, "step": 4741 }, { "epoch": 0.48, "learning_rate": 5.788786434617382e-05, "loss": 0.9147, "step": 4742 }, { "epoch": 0.48, "learning_rate": 5.788698217988363e-05, "loss": 1.1697, "step": 4743 }, { "epoch": 0.48, "learning_rate": 5.788609983613129e-05, "loss": 0.5777, "step": 4744 }, { "epoch": 0.48, "learning_rate": 5.788521731492242e-05, "loss": 1.278, "step": 4745 }, { "epoch": 0.48, "learning_rate": 5.788433461626264e-05, "loss": 1.1223, "step": 4746 }, { "epoch": 0.48, "learning_rate": 5.7883451740157564e-05, "loss": 1.0627, "step": 4747 }, { "epoch": 0.48, "learning_rate": 5.7882568686612794e-05, "loss": 1.325, "step": 4748 }, { "epoch": 0.48, "learning_rate": 5.7881685455633975e-05, "loss": 0.4178, "step": 4749 }, { "epoch": 0.48, "learning_rate": 5.7880802047226724e-05, "loss": 2.2962, "step": 4750 }, { "epoch": 0.48, "learning_rate": 5.787991846139664e-05, "loss": 0.2396, "step": 4751 }, { "epoch": 0.48, "learning_rate": 5.787903469814937e-05, "loss": 1.0295, "step": 4752 }, { "epoch": 0.48, "learning_rate": 5.787815075749053e-05, "loss": 0.4328, "step": 4753 }, { "epoch": 0.48, "learning_rate": 5.787726663942573e-05, "loss": 1.3856, "step": 4754 }, { "epoch": 0.48, "learning_rate": 5.787638234396062e-05, "loss": 0.3812, "step": 4755 }, { "epoch": 0.48, "learning_rate": 5.787549787110082e-05, "loss": 0.6686, "step": 4756 }, { "epoch": 0.48, "learning_rate": 5.7874613220851954e-05, "loss": 0.4168, "step": 4757 }, { "epoch": 0.48, "learning_rate": 5.787372839321965e-05, "loss": 1.2049, "step": 4758 }, { "epoch": 0.48, "learning_rate": 5.787284338820955e-05, "loss": 0.3735, "step": 4759 }, { "epoch": 0.48, "learning_rate": 5.7871958205827274e-05, "loss": 0.3737, "step": 4760 }, { "epoch": 0.48, "learning_rate": 5.7871072846078455e-05, "loss": 1.1924, "step": 4761 }, { "epoch": 0.48, "learning_rate": 5.787018730896873e-05, "loss": 1.323, "step": 4762 }, { "epoch": 0.48, "learning_rate": 5.786930159450375e-05, "loss": 0.3736, "step": 4763 }, { "epoch": 0.48, "learning_rate": 5.786841570268912e-05, "loss": 1.8889, "step": 4764 }, { "epoch": 0.48, "learning_rate": 5.78675296335305e-05, "loss": 1.4234, "step": 4765 }, { "epoch": 0.48, "learning_rate": 5.786664338703353e-05, "loss": 0.2356, "step": 4766 }, { "epoch": 0.48, "learning_rate": 5.786575696320383e-05, "loss": 1.1219, "step": 4767 }, { "epoch": 0.48, "learning_rate": 5.7864870362047056e-05, "loss": 1.1053, "step": 4768 }, { "epoch": 0.48, "learning_rate": 5.786398358356886e-05, "loss": 1.1597, "step": 4769 }, { "epoch": 0.48, "learning_rate": 5.786309662777486e-05, "loss": 1.5782, "step": 4770 }, { "epoch": 0.48, "learning_rate": 5.786220949467071e-05, "loss": 1.3343, "step": 4771 }, { "epoch": 0.48, "learning_rate": 5.7861322184262065e-05, "loss": 0.2612, "step": 4772 }, { "epoch": 0.48, "learning_rate": 5.786043469655456e-05, "loss": 0.9108, "step": 4773 }, { "epoch": 0.48, "learning_rate": 5.785954703155386e-05, "loss": 0.3211, "step": 4774 }, { "epoch": 0.48, "learning_rate": 5.7858659189265585e-05, "loss": 0.3762, "step": 4775 }, { "epoch": 0.48, "learning_rate": 5.7857771169695415e-05, "loss": 0.6296, "step": 4776 }, { "epoch": 0.49, "learning_rate": 5.7856882972848974e-05, "loss": 0.6665, "step": 4777 }, { "epoch": 0.49, "learning_rate": 5.785599459873194e-05, "loss": 1.0123, "step": 4778 }, { "epoch": 0.49, "learning_rate": 5.785510604734994e-05, "loss": 1.2571, "step": 4779 }, { "epoch": 0.49, "learning_rate": 5.785421731870865e-05, "loss": 1.1353, "step": 4780 }, { "epoch": 0.49, "learning_rate": 5.785332841281371e-05, "loss": 0.8392, "step": 4781 }, { "epoch": 0.49, "learning_rate": 5.785243932967079e-05, "loss": 0.4735, "step": 4782 }, { "epoch": 0.49, "learning_rate": 5.785155006928554e-05, "loss": 1.0473, "step": 4783 }, { "epoch": 0.49, "learning_rate": 5.785066063166363e-05, "loss": 0.8019, "step": 4784 }, { "epoch": 0.49, "learning_rate": 5.7849771016810694e-05, "loss": 1.5673, "step": 4785 }, { "epoch": 0.49, "learning_rate": 5.784888122473242e-05, "loss": 0.2352, "step": 4786 }, { "epoch": 0.49, "learning_rate": 5.784799125543446e-05, "loss": 0.3802, "step": 4787 }, { "epoch": 0.49, "learning_rate": 5.7847101108922474e-05, "loss": 0.3795, "step": 4788 }, { "epoch": 0.49, "learning_rate": 5.784621078520213e-05, "loss": 0.2803, "step": 4789 }, { "epoch": 0.49, "learning_rate": 5.784532028427909e-05, "loss": 1.1455, "step": 4790 }, { "epoch": 0.49, "learning_rate": 5.7844429606159034e-05, "loss": 0.8081, "step": 4791 }, { "epoch": 0.49, "learning_rate": 5.784353875084762e-05, "loss": 0.5294, "step": 4792 }, { "epoch": 0.49, "learning_rate": 5.784264771835051e-05, "loss": 0.5222, "step": 4793 }, { "epoch": 0.49, "learning_rate": 5.784175650867339e-05, "loss": 0.5129, "step": 4794 }, { "epoch": 0.49, "learning_rate": 5.784086512182191e-05, "loss": 0.7854, "step": 4795 }, { "epoch": 0.49, "learning_rate": 5.783997355780177e-05, "loss": 0.617, "step": 4796 }, { "epoch": 0.49, "learning_rate": 5.7839081816618625e-05, "loss": 0.5389, "step": 4797 }, { "epoch": 0.49, "learning_rate": 5.783818989827816e-05, "loss": 1.7378, "step": 4798 }, { "epoch": 0.49, "learning_rate": 5.7837297802786034e-05, "loss": 0.073, "step": 4799 }, { "epoch": 0.49, "learning_rate": 5.7836405530147935e-05, "loss": 1.4077, "step": 4800 }, { "epoch": 0.49, "learning_rate": 5.783551308036954e-05, "loss": 1.4334, "step": 4801 }, { "epoch": 0.49, "learning_rate": 5.783462045345653e-05, "loss": 1.1322, "step": 4802 }, { "epoch": 0.49, "learning_rate": 5.78337276494146e-05, "loss": 0.5742, "step": 4803 }, { "epoch": 0.49, "learning_rate": 5.78328346682494e-05, "loss": 1.0242, "step": 4804 }, { "epoch": 0.49, "learning_rate": 5.7831941509966625e-05, "loss": 0.3774, "step": 4805 }, { "epoch": 0.49, "learning_rate": 5.7831048174571966e-05, "loss": 0.4758, "step": 4806 }, { "epoch": 0.49, "learning_rate": 5.783015466207111e-05, "loss": 1.3982, "step": 4807 }, { "epoch": 0.49, "learning_rate": 5.7829260972469734e-05, "loss": 1.1374, "step": 4808 }, { "epoch": 0.49, "learning_rate": 5.782836710577353e-05, "loss": 0.5847, "step": 4809 }, { "epoch": 0.49, "learning_rate": 5.7827473061988176e-05, "loss": 0.4615, "step": 4810 }, { "epoch": 0.49, "learning_rate": 5.782657884111937e-05, "loss": 1.2651, "step": 4811 }, { "epoch": 0.49, "learning_rate": 5.7825684443172816e-05, "loss": 0.9409, "step": 4812 }, { "epoch": 0.49, "learning_rate": 5.782478986815418e-05, "loss": 0.6181, "step": 4813 }, { "epoch": 0.49, "learning_rate": 5.782389511606917e-05, "loss": 1.0193, "step": 4814 }, { "epoch": 0.49, "learning_rate": 5.782300018692347e-05, "loss": 0.9539, "step": 4815 }, { "epoch": 0.49, "learning_rate": 5.7822105080722786e-05, "loss": 0.6991, "step": 4816 }, { "epoch": 0.49, "learning_rate": 5.7821209797472805e-05, "loss": 1.1151, "step": 4817 }, { "epoch": 0.49, "learning_rate": 5.7820314337179236e-05, "loss": 0.4767, "step": 4818 }, { "epoch": 0.49, "learning_rate": 5.7819418699847766e-05, "loss": 0.6986, "step": 4819 }, { "epoch": 0.49, "learning_rate": 5.78185228854841e-05, "loss": 0.3872, "step": 4820 }, { "epoch": 0.49, "learning_rate": 5.781762689409394e-05, "loss": 0.9576, "step": 4821 }, { "epoch": 0.49, "learning_rate": 5.7816730725682976e-05, "loss": 0.9833, "step": 4822 }, { "epoch": 0.49, "learning_rate": 5.7815834380256925e-05, "loss": 0.8369, "step": 4823 }, { "epoch": 0.49, "learning_rate": 5.7814937857821486e-05, "loss": 0.6203, "step": 4824 }, { "epoch": 0.49, "learning_rate": 5.7814041158382365e-05, "loss": 1.4902, "step": 4825 }, { "epoch": 0.49, "learning_rate": 5.781314428194527e-05, "loss": 1.2907, "step": 4826 }, { "epoch": 0.49, "learning_rate": 5.781224722851589e-05, "loss": 0.358, "step": 4827 }, { "epoch": 0.49, "learning_rate": 5.7811349998099966e-05, "loss": 0.4896, "step": 4828 }, { "epoch": 0.49, "learning_rate": 5.7810452590703186e-05, "loss": 0.6444, "step": 4829 }, { "epoch": 0.49, "learning_rate": 5.780955500633126e-05, "loss": 0.4796, "step": 4830 }, { "epoch": 0.49, "learning_rate": 5.780865724498991e-05, "loss": 0.5271, "step": 4831 }, { "epoch": 0.49, "learning_rate": 5.780775930668485e-05, "loss": 0.3365, "step": 4832 }, { "epoch": 0.49, "learning_rate": 5.780686119142178e-05, "loss": 0.7048, "step": 4833 }, { "epoch": 0.49, "learning_rate": 5.780596289920642e-05, "loss": 0.7213, "step": 4834 }, { "epoch": 0.49, "learning_rate": 5.78050644300445e-05, "loss": 0.7906, "step": 4835 }, { "epoch": 0.49, "learning_rate": 5.780416578394172e-05, "loss": 0.5022, "step": 4836 }, { "epoch": 0.49, "learning_rate": 5.7803266960903804e-05, "loss": 1.2211, "step": 4837 }, { "epoch": 0.49, "learning_rate": 5.780236796093649e-05, "loss": 1.2595, "step": 4838 }, { "epoch": 0.49, "learning_rate": 5.780146878404546e-05, "loss": 1.7927, "step": 4839 }, { "epoch": 0.49, "learning_rate": 5.780056943023647e-05, "loss": 1.7147, "step": 4840 }, { "epoch": 0.49, "learning_rate": 5.7799669899515234e-05, "loss": 0.6825, "step": 4841 }, { "epoch": 0.49, "learning_rate": 5.779877019188747e-05, "loss": 0.5487, "step": 4842 }, { "epoch": 0.49, "learning_rate": 5.779787030735891e-05, "loss": 0.4929, "step": 4843 }, { "epoch": 0.49, "learning_rate": 5.7796970245935285e-05, "loss": 0.6925, "step": 4844 }, { "epoch": 0.49, "learning_rate": 5.779607000762231e-05, "loss": 0.594, "step": 4845 }, { "epoch": 0.49, "learning_rate": 5.779516959242572e-05, "loss": 1.2907, "step": 4846 }, { "epoch": 0.49, "learning_rate": 5.779426900035124e-05, "loss": 0.5538, "step": 4847 }, { "epoch": 0.49, "learning_rate": 5.779336823140461e-05, "loss": 0.8529, "step": 4848 }, { "epoch": 0.49, "learning_rate": 5.779246728559156e-05, "loss": 1.025, "step": 4849 }, { "epoch": 0.49, "learning_rate": 5.779156616291782e-05, "loss": 0.5485, "step": 4850 }, { "epoch": 0.49, "learning_rate": 5.779066486338912e-05, "loss": 0.4452, "step": 4851 }, { "epoch": 0.49, "learning_rate": 5.77897633870112e-05, "loss": 0.437, "step": 4852 }, { "epoch": 0.49, "learning_rate": 5.778886173378981e-05, "loss": 0.7462, "step": 4853 }, { "epoch": 0.49, "learning_rate": 5.778795990373067e-05, "loss": 0.7456, "step": 4854 }, { "epoch": 0.49, "learning_rate": 5.778705789683952e-05, "loss": 0.5226, "step": 4855 }, { "epoch": 0.49, "learning_rate": 5.77861557131221e-05, "loss": 1.5325, "step": 4856 }, { "epoch": 0.49, "learning_rate": 5.778525335258416e-05, "loss": 0.2201, "step": 4857 }, { "epoch": 0.49, "learning_rate": 5.7784350815231446e-05, "loss": 0.8737, "step": 4858 }, { "epoch": 0.49, "learning_rate": 5.778344810106968e-05, "loss": 1.7831, "step": 4859 }, { "epoch": 0.49, "learning_rate": 5.7782545210104626e-05, "loss": 1.7092, "step": 4860 }, { "epoch": 0.49, "learning_rate": 5.778164214234203e-05, "loss": 0.8803, "step": 4861 }, { "epoch": 0.49, "learning_rate": 5.778073889778762e-05, "loss": 1.358, "step": 4862 }, { "epoch": 0.49, "learning_rate": 5.7779835476447164e-05, "loss": 1.0897, "step": 4863 }, { "epoch": 0.49, "learning_rate": 5.7778931878326405e-05, "loss": 0.8862, "step": 4864 }, { "epoch": 0.49, "learning_rate": 5.777802810343108e-05, "loss": 1.0778, "step": 4865 }, { "epoch": 0.49, "learning_rate": 5.777712415176696e-05, "loss": 0.6084, "step": 4866 }, { "epoch": 0.49, "learning_rate": 5.777622002333979e-05, "loss": 1.3334, "step": 4867 }, { "epoch": 0.49, "learning_rate": 5.7775315718155316e-05, "loss": 0.2925, "step": 4868 }, { "epoch": 0.49, "learning_rate": 5.77744112362193e-05, "loss": 1.1014, "step": 4869 }, { "epoch": 0.49, "learning_rate": 5.7773506577537504e-05, "loss": 1.0409, "step": 4870 }, { "epoch": 0.49, "learning_rate": 5.777260174211567e-05, "loss": 0.9795, "step": 4871 }, { "epoch": 0.49, "learning_rate": 5.777169672995957e-05, "loss": 0.9712, "step": 4872 }, { "epoch": 0.49, "learning_rate": 5.7770791541074944e-05, "loss": 0.4286, "step": 4873 }, { "epoch": 0.49, "learning_rate": 5.7769886175467583e-05, "loss": 1.3075, "step": 4874 }, { "epoch": 0.5, "learning_rate": 5.776898063314322e-05, "loss": 0.1199, "step": 4875 }, { "epoch": 0.5, "learning_rate": 5.776807491410763e-05, "loss": 0.5432, "step": 4876 }, { "epoch": 0.5, "learning_rate": 5.776716901836657e-05, "loss": 0.9994, "step": 4877 }, { "epoch": 0.5, "learning_rate": 5.776626294592581e-05, "loss": 1.1785, "step": 4878 }, { "epoch": 0.5, "learning_rate": 5.776535669679113e-05, "loss": 1.4669, "step": 4879 }, { "epoch": 0.5, "learning_rate": 5.7764450270968275e-05, "loss": 0.5472, "step": 4880 }, { "epoch": 0.5, "learning_rate": 5.776354366846301e-05, "loss": 1.3927, "step": 4881 }, { "epoch": 0.5, "learning_rate": 5.776263688928113e-05, "loss": 1.1615, "step": 4882 }, { "epoch": 0.5, "learning_rate": 5.776172993342838e-05, "loss": 0.7844, "step": 4883 }, { "epoch": 0.5, "learning_rate": 5.7760822800910544e-05, "loss": 0.906, "step": 4884 }, { "epoch": 0.5, "learning_rate": 5.77599154917334e-05, "loss": 1.3282, "step": 4885 }, { "epoch": 0.5, "learning_rate": 5.7759008005902704e-05, "loss": 1.5549, "step": 4886 }, { "epoch": 0.5, "learning_rate": 5.775810034342424e-05, "loss": 0.39, "step": 4887 }, { "epoch": 0.5, "learning_rate": 5.7757192504303794e-05, "loss": 0.9538, "step": 4888 }, { "epoch": 0.5, "learning_rate": 5.775628448854713e-05, "loss": 1.4803, "step": 4889 }, { "epoch": 0.5, "learning_rate": 5.7755376296160036e-05, "loss": 0.5494, "step": 4890 }, { "epoch": 0.5, "learning_rate": 5.775446792714828e-05, "loss": 0.8385, "step": 4891 }, { "epoch": 0.5, "learning_rate": 5.775355938151764e-05, "loss": 1.0005, "step": 4892 }, { "epoch": 0.5, "learning_rate": 5.775265065927392e-05, "loss": 0.5242, "step": 4893 }, { "epoch": 0.5, "learning_rate": 5.775174176042289e-05, "loss": 1.2586, "step": 4894 }, { "epoch": 0.5, "learning_rate": 5.775083268497032e-05, "loss": 1.0432, "step": 4895 }, { "epoch": 0.5, "learning_rate": 5.774992343292201e-05, "loss": 1.6142, "step": 4896 }, { "epoch": 0.5, "learning_rate": 5.774901400428375e-05, "loss": 1.428, "step": 4897 }, { "epoch": 0.5, "learning_rate": 5.7748104399061326e-05, "loss": 1.4019, "step": 4898 }, { "epoch": 0.5, "learning_rate": 5.774719461726051e-05, "loss": 0.3962, "step": 4899 }, { "epoch": 0.5, "learning_rate": 5.774628465888711e-05, "loss": 1.2255, "step": 4900 }, { "epoch": 0.5, "learning_rate": 5.7745374523946914e-05, "loss": 0.5231, "step": 4901 }, { "epoch": 0.5, "learning_rate": 5.77444642124457e-05, "loss": 1.1656, "step": 4902 }, { "epoch": 0.5, "learning_rate": 5.774355372438927e-05, "loss": 0.4493, "step": 4903 }, { "epoch": 0.5, "learning_rate": 5.774264305978342e-05, "loss": 0.0934, "step": 4904 }, { "epoch": 0.5, "learning_rate": 5.7741732218633946e-05, "loss": 0.8359, "step": 4905 }, { "epoch": 0.5, "learning_rate": 5.774082120094664e-05, "loss": 1.3524, "step": 4906 }, { "epoch": 0.5, "learning_rate": 5.773991000672731e-05, "loss": 0.2861, "step": 4907 }, { "epoch": 0.5, "learning_rate": 5.773899863598173e-05, "loss": 0.326, "step": 4908 }, { "epoch": 0.5, "learning_rate": 5.773808708871573e-05, "loss": 0.7796, "step": 4909 }, { "epoch": 0.5, "learning_rate": 5.7737175364935084e-05, "loss": 0.6997, "step": 4910 }, { "epoch": 0.5, "learning_rate": 5.773626346464561e-05, "loss": 0.7389, "step": 4911 }, { "epoch": 0.5, "learning_rate": 5.773535138785311e-05, "loss": 0.5787, "step": 4912 }, { "epoch": 0.5, "learning_rate": 5.7734439134563375e-05, "loss": 1.0017, "step": 4913 }, { "epoch": 0.5, "learning_rate": 5.773352670478223e-05, "loss": 0.9165, "step": 4914 }, { "epoch": 0.5, "learning_rate": 5.773261409851547e-05, "loss": 0.8888, "step": 4915 }, { "epoch": 0.5, "learning_rate": 5.7731701315768907e-05, "loss": 0.7062, "step": 4916 }, { "epoch": 0.5, "learning_rate": 5.7730788356548335e-05, "loss": 1.1247, "step": 4917 }, { "epoch": 0.5, "learning_rate": 5.772987522085959e-05, "loss": 1.3558, "step": 4918 }, { "epoch": 0.5, "learning_rate": 5.7728961908708454e-05, "loss": 0.8621, "step": 4919 }, { "epoch": 0.5, "learning_rate": 5.7728048420100754e-05, "loss": 1.495, "step": 4920 }, { "epoch": 0.5, "learning_rate": 5.772713475504231e-05, "loss": 0.759, "step": 4921 }, { "epoch": 0.5, "learning_rate": 5.772622091353893e-05, "loss": 1.662, "step": 4922 }, { "epoch": 0.5, "learning_rate": 5.772530689559642e-05, "loss": 0.9571, "step": 4923 }, { "epoch": 0.5, "learning_rate": 5.772439270122061e-05, "loss": 0.3256, "step": 4924 }, { "epoch": 0.5, "learning_rate": 5.772347833041731e-05, "loss": 1.0355, "step": 4925 }, { "epoch": 0.5, "learning_rate": 5.772256378319234e-05, "loss": 0.4115, "step": 4926 }, { "epoch": 0.5, "learning_rate": 5.772164905955151e-05, "loss": 0.6201, "step": 4927 }, { "epoch": 0.5, "learning_rate": 5.7720734159500665e-05, "loss": 1.086, "step": 4928 }, { "epoch": 0.5, "learning_rate": 5.771981908304561e-05, "loss": 0.5675, "step": 4929 }, { "epoch": 0.5, "learning_rate": 5.7718903830192166e-05, "loss": 0.2062, "step": 4930 }, { "epoch": 0.5, "learning_rate": 5.7717988400946175e-05, "loss": 1.3657, "step": 4931 }, { "epoch": 0.5, "learning_rate": 5.771707279531344e-05, "loss": 0.5173, "step": 4932 }, { "epoch": 0.5, "learning_rate": 5.7716157013299804e-05, "loss": 1.3414, "step": 4933 }, { "epoch": 0.5, "learning_rate": 5.7715241054911084e-05, "loss": 1.0449, "step": 4934 }, { "epoch": 0.5, "learning_rate": 5.771432492015312e-05, "loss": 0.1589, "step": 4935 }, { "epoch": 0.5, "learning_rate": 5.771340860903172e-05, "loss": 0.3882, "step": 4936 }, { "epoch": 0.5, "learning_rate": 5.771249212155275e-05, "loss": 0.6262, "step": 4937 }, { "epoch": 0.5, "learning_rate": 5.7711575457722005e-05, "loss": 0.493, "step": 4938 }, { "epoch": 0.5, "learning_rate": 5.771065861754534e-05, "loss": 1.0549, "step": 4939 }, { "epoch": 0.5, "learning_rate": 5.770974160102859e-05, "loss": 1.2237, "step": 4940 }, { "epoch": 0.5, "learning_rate": 5.770882440817759e-05, "loss": 0.5894, "step": 4941 }, { "epoch": 0.5, "learning_rate": 5.770790703899817e-05, "loss": 1.0943, "step": 4942 }, { "epoch": 0.5, "learning_rate": 5.7706989493496165e-05, "loss": 1.4908, "step": 4943 }, { "epoch": 0.5, "learning_rate": 5.770607177167742e-05, "loss": 0.8523, "step": 4944 }, { "epoch": 0.5, "learning_rate": 5.770515387354778e-05, "loss": 1.2132, "step": 4945 }, { "epoch": 0.5, "learning_rate": 5.770423579911308e-05, "loss": 2.0764, "step": 4946 }, { "epoch": 0.5, "learning_rate": 5.770331754837915e-05, "loss": 0.5671, "step": 4947 }, { "epoch": 0.5, "learning_rate": 5.770239912135186e-05, "loss": 0.4404, "step": 4948 }, { "epoch": 0.5, "learning_rate": 5.770148051803704e-05, "loss": 1.4097, "step": 4949 }, { "epoch": 0.5, "learning_rate": 5.7700561738440526e-05, "loss": 0.844, "step": 4950 }, { "epoch": 0.5, "learning_rate": 5.769964278256818e-05, "loss": 0.7752, "step": 4951 }, { "epoch": 0.5, "learning_rate": 5.7698723650425846e-05, "loss": 2.336, "step": 4952 }, { "epoch": 0.5, "learning_rate": 5.7697804342019374e-05, "loss": 1.4555, "step": 4953 }, { "epoch": 0.5, "learning_rate": 5.76968848573546e-05, "loss": 1.033, "step": 4954 }, { "epoch": 0.5, "learning_rate": 5.76959651964374e-05, "loss": 0.3658, "step": 4955 }, { "epoch": 0.5, "learning_rate": 5.769504535927361e-05, "loss": 0.6996, "step": 4956 }, { "epoch": 0.5, "learning_rate": 5.769412534586909e-05, "loss": 1.4427, "step": 4957 }, { "epoch": 0.5, "learning_rate": 5.769320515622968e-05, "loss": 1.0644, "step": 4958 }, { "epoch": 0.5, "learning_rate": 5.769228479036126e-05, "loss": 0.3576, "step": 4959 }, { "epoch": 0.5, "learning_rate": 5.769136424826967e-05, "loss": 1.1142, "step": 4960 }, { "epoch": 0.5, "learning_rate": 5.7690443529960777e-05, "loss": 2.0422, "step": 4961 }, { "epoch": 0.5, "learning_rate": 5.768952263544043e-05, "loss": 1.3762, "step": 4962 }, { "epoch": 0.5, "learning_rate": 5.768860156471449e-05, "loss": 1.0491, "step": 4963 }, { "epoch": 0.5, "learning_rate": 5.768768031778883e-05, "loss": 1.1335, "step": 4964 }, { "epoch": 0.5, "learning_rate": 5.7686758894669304e-05, "loss": 0.943, "step": 4965 }, { "epoch": 0.5, "learning_rate": 5.7685837295361775e-05, "loss": 0.595, "step": 4966 }, { "epoch": 0.5, "learning_rate": 5.768491551987211e-05, "loss": 0.3978, "step": 4967 }, { "epoch": 0.5, "learning_rate": 5.7683993568206174e-05, "loss": 1.1256, "step": 4968 }, { "epoch": 0.5, "learning_rate": 5.7683071440369845e-05, "loss": 0.7414, "step": 4969 }, { "epoch": 0.5, "learning_rate": 5.7682149136368966e-05, "loss": 1.775, "step": 4970 }, { "epoch": 0.5, "learning_rate": 5.7681226656209425e-05, "loss": 0.5167, "step": 4971 }, { "epoch": 0.5, "learning_rate": 5.768030399989709e-05, "loss": 1.3664, "step": 4972 }, { "epoch": 0.5, "learning_rate": 5.767938116743783e-05, "loss": 0.9651, "step": 4973 }, { "epoch": 0.51, "learning_rate": 5.767845815883752e-05, "loss": 0.8368, "step": 4974 }, { "epoch": 0.51, "learning_rate": 5.767753497410203e-05, "loss": 0.4746, "step": 4975 }, { "epoch": 0.51, "learning_rate": 5.767661161323723e-05, "loss": 0.4058, "step": 4976 }, { "epoch": 0.51, "learning_rate": 5.767568807624901e-05, "loss": 0.4593, "step": 4977 }, { "epoch": 0.51, "learning_rate": 5.767476436314324e-05, "loss": 0.6497, "step": 4978 }, { "epoch": 0.51, "learning_rate": 5.76738404739258e-05, "loss": 1.4118, "step": 4979 }, { "epoch": 0.51, "learning_rate": 5.767291640860256e-05, "loss": 1.4202, "step": 4980 }, { "epoch": 0.51, "learning_rate": 5.767199216717941e-05, "loss": 0.737, "step": 4981 }, { "epoch": 0.51, "learning_rate": 5.767106774966223e-05, "loss": 0.4569, "step": 4982 }, { "epoch": 0.51, "learning_rate": 5.76701431560569e-05, "loss": 0.4252, "step": 4983 }, { "epoch": 0.51, "learning_rate": 5.7669218386369304e-05, "loss": 0.6445, "step": 4984 }, { "epoch": 0.51, "learning_rate": 5.766829344060533e-05, "loss": 1.3421, "step": 4985 }, { "epoch": 0.51, "learning_rate": 5.7667368318770863e-05, "loss": 0.2338, "step": 4986 }, { "epoch": 0.51, "learning_rate": 5.7666443020871785e-05, "loss": 1.4329, "step": 4987 }, { "epoch": 0.51, "learning_rate": 5.7665517546913994e-05, "loss": 0.7924, "step": 4988 }, { "epoch": 0.51, "learning_rate": 5.7664591896903366e-05, "loss": 1.0294, "step": 4989 }, { "epoch": 0.51, "learning_rate": 5.766366607084581e-05, "loss": 1.0354, "step": 4990 }, { "epoch": 0.51, "learning_rate": 5.7662740068747195e-05, "loss": 0.7682, "step": 4991 }, { "epoch": 0.51, "learning_rate": 5.7661813890613436e-05, "loss": 1.3985, "step": 4992 }, { "epoch": 0.51, "learning_rate": 5.7660887536450414e-05, "loss": 0.909, "step": 4993 }, { "epoch": 0.51, "learning_rate": 5.765996100626402e-05, "loss": 1.8707, "step": 4994 }, { "epoch": 0.51, "learning_rate": 5.7659034300060166e-05, "loss": 0.9548, "step": 4995 }, { "epoch": 0.51, "learning_rate": 5.765810741784473e-05, "loss": 0.6611, "step": 4996 }, { "epoch": 0.51, "learning_rate": 5.7657180359623626e-05, "loss": 1.6092, "step": 4997 }, { "epoch": 0.51, "learning_rate": 5.765625312540275e-05, "loss": 0.3906, "step": 4998 }, { "epoch": 0.51, "learning_rate": 5.765532571518799e-05, "loss": 0.4726, "step": 4999 }, { "epoch": 0.51, "learning_rate": 5.765439812898527e-05, "loss": 0.4486, "step": 5000 }, { "epoch": 0.51, "learning_rate": 5.765347036680047e-05, "loss": 0.9736, "step": 5001 }, { "epoch": 0.51, "learning_rate": 5.765254242863952e-05, "loss": 1.1078, "step": 5002 }, { "epoch": 0.51, "learning_rate": 5.7651614314508294e-05, "loss": 1.3432, "step": 5003 }, { "epoch": 0.51, "learning_rate": 5.7650686024412725e-05, "loss": 1.0822, "step": 5004 }, { "epoch": 0.51, "learning_rate": 5.76497575583587e-05, "loss": 0.5394, "step": 5005 }, { "epoch": 0.51, "learning_rate": 5.764882891635214e-05, "loss": 1.0446, "step": 5006 }, { "epoch": 0.51, "learning_rate": 5.764790009839895e-05, "loss": 0.8231, "step": 5007 }, { "epoch": 0.51, "learning_rate": 5.764697110450504e-05, "loss": 1.4414, "step": 5008 }, { "epoch": 0.51, "learning_rate": 5.764604193467632e-05, "loss": 0.9528, "step": 5009 }, { "epoch": 0.51, "learning_rate": 5.764511258891872e-05, "loss": 0.52, "step": 5010 }, { "epoch": 0.51, "learning_rate": 5.764418306723813e-05, "loss": 0.312, "step": 5011 }, { "epoch": 0.51, "learning_rate": 5.764325336964048e-05, "loss": 0.9049, "step": 5012 }, { "epoch": 0.51, "learning_rate": 5.764232349613168e-05, "loss": 1.3152, "step": 5013 }, { "epoch": 0.51, "learning_rate": 5.764139344671764e-05, "loss": 0.6079, "step": 5014 }, { "epoch": 0.51, "learning_rate": 5.76404632214043e-05, "loss": 1.4229, "step": 5015 }, { "epoch": 0.51, "learning_rate": 5.763953282019756e-05, "loss": 0.8083, "step": 5016 }, { "epoch": 0.51, "learning_rate": 5.763860224310335e-05, "loss": 0.1521, "step": 5017 }, { "epoch": 0.51, "learning_rate": 5.7637671490127584e-05, "loss": 2.2171, "step": 5018 }, { "epoch": 0.51, "learning_rate": 5.7636740561276204e-05, "loss": 0.6658, "step": 5019 }, { "epoch": 0.51, "learning_rate": 5.763580945655511e-05, "loss": 1.0195, "step": 5020 }, { "epoch": 0.51, "learning_rate": 5.763487817597024e-05, "loss": 1.1228, "step": 5021 }, { "epoch": 0.51, "learning_rate": 5.763394671952752e-05, "loss": 0.1929, "step": 5022 }, { "epoch": 0.51, "learning_rate": 5.763301508723287e-05, "loss": 1.1595, "step": 5023 }, { "epoch": 0.51, "learning_rate": 5.7632083279092233e-05, "loss": 1.5415, "step": 5024 }, { "epoch": 0.51, "learning_rate": 5.763115129511152e-05, "loss": 0.4868, "step": 5025 }, { "epoch": 0.51, "learning_rate": 5.7630219135296676e-05, "loss": 0.2463, "step": 5026 }, { "epoch": 0.51, "learning_rate": 5.7629286799653626e-05, "loss": 0.8982, "step": 5027 }, { "epoch": 0.51, "learning_rate": 5.7628354288188316e-05, "loss": 0.5718, "step": 5028 }, { "epoch": 0.51, "learning_rate": 5.7627421600906654e-05, "loss": 1.3579, "step": 5029 }, { "epoch": 0.51, "learning_rate": 5.76264887378146e-05, "loss": 1.0292, "step": 5030 }, { "epoch": 0.51, "learning_rate": 5.7625555698918086e-05, "loss": 0.7215, "step": 5031 }, { "epoch": 0.51, "learning_rate": 5.7624622484223044e-05, "loss": 0.8515, "step": 5032 }, { "epoch": 0.51, "learning_rate": 5.76236890937354e-05, "loss": 0.4018, "step": 5033 }, { "epoch": 0.51, "learning_rate": 5.762275552746112e-05, "loss": 0.7852, "step": 5034 }, { "epoch": 0.51, "learning_rate": 5.7621821785406125e-05, "loss": 0.8766, "step": 5035 }, { "epoch": 0.51, "learning_rate": 5.762088786757637e-05, "loss": 0.1946, "step": 5036 }, { "epoch": 0.51, "learning_rate": 5.7619953773977784e-05, "loss": 0.2827, "step": 5037 }, { "epoch": 0.51, "learning_rate": 5.7619019504616325e-05, "loss": 0.459, "step": 5038 }, { "epoch": 0.51, "learning_rate": 5.7618085059497934e-05, "loss": 1.6701, "step": 5039 }, { "epoch": 0.51, "learning_rate": 5.7617150438628556e-05, "loss": 1.0389, "step": 5040 }, { "epoch": 0.51, "learning_rate": 5.761621564201414e-05, "loss": 0.9136, "step": 5041 }, { "epoch": 0.51, "learning_rate": 5.7615280669660625e-05, "loss": 0.5971, "step": 5042 }, { "epoch": 0.51, "learning_rate": 5.761434552157398e-05, "loss": 1.6564, "step": 5043 }, { "epoch": 0.51, "learning_rate": 5.761341019776014e-05, "loss": 1.4292, "step": 5044 }, { "epoch": 0.51, "learning_rate": 5.761247469822506e-05, "loss": 0.7629, "step": 5045 }, { "epoch": 0.51, "learning_rate": 5.76115390229747e-05, "loss": 1.6347, "step": 5046 }, { "epoch": 0.51, "learning_rate": 5.7610603172015e-05, "loss": 1.0636, "step": 5047 }, { "epoch": 0.51, "learning_rate": 5.760966714535194e-05, "loss": 0.8174, "step": 5048 }, { "epoch": 0.51, "learning_rate": 5.760873094299145e-05, "loss": 0.186, "step": 5049 }, { "epoch": 0.51, "learning_rate": 5.7607794564939505e-05, "loss": 0.7599, "step": 5050 }, { "epoch": 0.51, "learning_rate": 5.7606858011202056e-05, "loss": 0.9816, "step": 5051 }, { "epoch": 0.51, "learning_rate": 5.760592128178507e-05, "loss": 1.5171, "step": 5052 }, { "epoch": 0.51, "learning_rate": 5.760498437669449e-05, "loss": 0.0942, "step": 5053 }, { "epoch": 0.51, "learning_rate": 5.76040472959363e-05, "loss": 1.6352, "step": 5054 }, { "epoch": 0.51, "learning_rate": 5.760311003951645e-05, "loss": 0.9512, "step": 5055 }, { "epoch": 0.51, "learning_rate": 5.7602172607440915e-05, "loss": 0.4512, "step": 5056 }, { "epoch": 0.51, "learning_rate": 5.760123499971564e-05, "loss": 1.3386, "step": 5057 }, { "epoch": 0.51, "learning_rate": 5.760029721634662e-05, "loss": 0.4706, "step": 5058 }, { "epoch": 0.51, "learning_rate": 5.75993592573398e-05, "loss": 1.1595, "step": 5059 }, { "epoch": 0.51, "learning_rate": 5.7598421122701164e-05, "loss": 0.5836, "step": 5060 }, { "epoch": 0.51, "learning_rate": 5.7597482812436675e-05, "loss": 0.4987, "step": 5061 }, { "epoch": 0.51, "learning_rate": 5.75965443265523e-05, "loss": 1.0494, "step": 5062 }, { "epoch": 0.51, "learning_rate": 5.759560566505401e-05, "loss": 1.2527, "step": 5063 }, { "epoch": 0.51, "learning_rate": 5.759466682794779e-05, "loss": 0.4006, "step": 5064 }, { "epoch": 0.51, "learning_rate": 5.759372781523961e-05, "loss": 1.2279, "step": 5065 }, { "epoch": 0.51, "learning_rate": 5.7592788626935436e-05, "loss": 0.4716, "step": 5066 }, { "epoch": 0.51, "learning_rate": 5.759184926304126e-05, "loss": 1.0103, "step": 5067 }, { "epoch": 0.51, "learning_rate": 5.759090972356305e-05, "loss": 1.2849, "step": 5068 }, { "epoch": 0.51, "learning_rate": 5.758997000850678e-05, "loss": 1.1956, "step": 5069 }, { "epoch": 0.51, "learning_rate": 5.758903011787844e-05, "loss": 0.4327, "step": 5070 }, { "epoch": 0.51, "learning_rate": 5.758809005168401e-05, "loss": 0.6985, "step": 5071 }, { "epoch": 0.52, "learning_rate": 5.758714980992947e-05, "loss": 0.1752, "step": 5072 }, { "epoch": 0.52, "learning_rate": 5.7586209392620804e-05, "loss": 0.796, "step": 5073 }, { "epoch": 0.52, "learning_rate": 5.758526879976399e-05, "loss": 1.1676, "step": 5074 }, { "epoch": 0.52, "learning_rate": 5.758432803136503e-05, "loss": 1.5201, "step": 5075 }, { "epoch": 0.52, "learning_rate": 5.758338708742989e-05, "loss": 0.8356, "step": 5076 }, { "epoch": 0.52, "learning_rate": 5.758244596796457e-05, "loss": 0.4342, "step": 5077 }, { "epoch": 0.52, "learning_rate": 5.758150467297506e-05, "loss": 1.5442, "step": 5078 }, { "epoch": 0.52, "learning_rate": 5.7580563202467344e-05, "loss": 0.5475, "step": 5079 }, { "epoch": 0.52, "learning_rate": 5.757962155644741e-05, "loss": 1.0952, "step": 5080 }, { "epoch": 0.52, "learning_rate": 5.757867973492127e-05, "loss": 1.5005, "step": 5081 }, { "epoch": 0.52, "learning_rate": 5.75777377378949e-05, "loss": 0.5011, "step": 5082 }, { "epoch": 0.52, "learning_rate": 5.7576795565374286e-05, "loss": 1.3398, "step": 5083 }, { "epoch": 0.52, "learning_rate": 5.757585321736545e-05, "loss": 1.4767, "step": 5084 }, { "epoch": 0.52, "learning_rate": 5.7574910693874365e-05, "loss": 0.9357, "step": 5085 }, { "epoch": 0.52, "learning_rate": 5.757396799490705e-05, "loss": 0.4791, "step": 5086 }, { "epoch": 0.52, "learning_rate": 5.757302512046948e-05, "loss": 0.3293, "step": 5087 }, { "epoch": 0.52, "learning_rate": 5.757208207056767e-05, "loss": 0.7206, "step": 5088 }, { "epoch": 0.52, "learning_rate": 5.757113884520762e-05, "loss": 0.3994, "step": 5089 }, { "epoch": 0.52, "learning_rate": 5.7570195444395334e-05, "loss": 1.277, "step": 5090 }, { "epoch": 0.52, "learning_rate": 5.756925186813681e-05, "loss": 1.4609, "step": 5091 }, { "epoch": 0.52, "learning_rate": 5.756830811643806e-05, "loss": 0.7356, "step": 5092 }, { "epoch": 0.52, "learning_rate": 5.7567364189305076e-05, "loss": 0.2459, "step": 5093 }, { "epoch": 0.52, "learning_rate": 5.756642008674387e-05, "loss": 0.9969, "step": 5094 }, { "epoch": 0.52, "learning_rate": 5.756547580876047e-05, "loss": 0.7941, "step": 5095 }, { "epoch": 0.52, "learning_rate": 5.756453135536086e-05, "loss": 0.7048, "step": 5096 }, { "epoch": 0.52, "learning_rate": 5.7563586726551056e-05, "loss": 0.3802, "step": 5097 }, { "epoch": 0.52, "learning_rate": 5.756264192233708e-05, "loss": 1.1052, "step": 5098 }, { "epoch": 0.52, "learning_rate": 5.756169694272492e-05, "loss": 0.4648, "step": 5099 }, { "epoch": 0.52, "learning_rate": 5.7560751787720616e-05, "loss": 0.7399, "step": 5100 }, { "epoch": 0.52, "learning_rate": 5.755980645733017e-05, "loss": 1.2383, "step": 5101 }, { "epoch": 0.52, "learning_rate": 5.755886095155961e-05, "loss": 0.7643, "step": 5102 }, { "epoch": 0.52, "learning_rate": 5.755791527041493e-05, "loss": 0.6845, "step": 5103 }, { "epoch": 0.52, "learning_rate": 5.755696941390217e-05, "loss": 0.1658, "step": 5104 }, { "epoch": 0.52, "learning_rate": 5.7556023382027335e-05, "loss": 0.969, "step": 5105 }, { "epoch": 0.52, "learning_rate": 5.755507717479645e-05, "loss": 1.596, "step": 5106 }, { "epoch": 0.52, "learning_rate": 5.755413079221553e-05, "loss": 1.645, "step": 5107 }, { "epoch": 0.52, "learning_rate": 5.755318423429061e-05, "loss": 1.5731, "step": 5108 }, { "epoch": 0.52, "learning_rate": 5.755223750102771e-05, "loss": 1.4538, "step": 5109 }, { "epoch": 0.52, "learning_rate": 5.7551290592432846e-05, "loss": 0.7246, "step": 5110 }, { "epoch": 0.52, "learning_rate": 5.755034350851205e-05, "loss": 1.4946, "step": 5111 }, { "epoch": 0.52, "learning_rate": 5.754939624927135e-05, "loss": 1.3124, "step": 5112 }, { "epoch": 0.52, "learning_rate": 5.754844881471678e-05, "loss": 0.9159, "step": 5113 }, { "epoch": 0.52, "learning_rate": 5.7547501204854356e-05, "loss": 0.4296, "step": 5114 }, { "epoch": 0.52, "learning_rate": 5.754655341969012e-05, "loss": 1.4805, "step": 5115 }, { "epoch": 0.52, "learning_rate": 5.754560545923009e-05, "loss": 1.5005, "step": 5116 }, { "epoch": 0.52, "learning_rate": 5.7544657323480304e-05, "loss": 0.847, "step": 5117 }, { "epoch": 0.52, "learning_rate": 5.7543709012446804e-05, "loss": 0.4244, "step": 5118 }, { "epoch": 0.52, "learning_rate": 5.7542760526135604e-05, "loss": 0.7499, "step": 5119 }, { "epoch": 0.52, "learning_rate": 5.754181186455277e-05, "loss": 0.907, "step": 5120 }, { "epoch": 0.52, "learning_rate": 5.754086302770432e-05, "loss": 0.5179, "step": 5121 }, { "epoch": 0.52, "learning_rate": 5.753991401559629e-05, "loss": 0.1907, "step": 5122 }, { "epoch": 0.52, "learning_rate": 5.7538964828234725e-05, "loss": 1.2311, "step": 5123 }, { "epoch": 0.52, "learning_rate": 5.7538015465625664e-05, "loss": 0.8646, "step": 5124 }, { "epoch": 0.52, "learning_rate": 5.753706592777515e-05, "loss": 0.4851, "step": 5125 }, { "epoch": 0.52, "learning_rate": 5.7536116214689225e-05, "loss": 1.3596, "step": 5126 }, { "epoch": 0.52, "learning_rate": 5.753516632637393e-05, "loss": 0.2206, "step": 5127 }, { "epoch": 0.52, "learning_rate": 5.753421626283531e-05, "loss": 1.467, "step": 5128 }, { "epoch": 0.52, "learning_rate": 5.753326602407942e-05, "loss": 0.3642, "step": 5129 }, { "epoch": 0.52, "learning_rate": 5.7532315610112295e-05, "loss": 0.6297, "step": 5130 }, { "epoch": 0.52, "learning_rate": 5.753136502093998e-05, "loss": 1.2405, "step": 5131 }, { "epoch": 0.52, "learning_rate": 5.7530414256568535e-05, "loss": 0.8759, "step": 5132 }, { "epoch": 0.52, "learning_rate": 5.752946331700401e-05, "loss": 0.9191, "step": 5133 }, { "epoch": 0.52, "learning_rate": 5.752851220225244e-05, "loss": 1.4759, "step": 5134 }, { "epoch": 0.52, "learning_rate": 5.7527560912319914e-05, "loss": 0.4893, "step": 5135 }, { "epoch": 0.52, "learning_rate": 5.752660944721245e-05, "loss": 0.1707, "step": 5136 }, { "epoch": 0.52, "learning_rate": 5.7525657806936115e-05, "loss": 0.3747, "step": 5137 }, { "epoch": 0.52, "learning_rate": 5.7524705991496966e-05, "loss": 1.3061, "step": 5138 }, { "epoch": 0.52, "learning_rate": 5.752375400090106e-05, "loss": 1.1225, "step": 5139 }, { "epoch": 0.52, "learning_rate": 5.7522801835154456e-05, "loss": 0.2298, "step": 5140 }, { "epoch": 0.52, "learning_rate": 5.75218494942632e-05, "loss": 1.4647, "step": 5141 }, { "epoch": 0.52, "learning_rate": 5.752089697823338e-05, "loss": 0.7321, "step": 5142 }, { "epoch": 0.52, "learning_rate": 5.751994428707103e-05, "loss": 1.3454, "step": 5143 }, { "epoch": 0.52, "learning_rate": 5.751899142078222e-05, "loss": 1.2707, "step": 5144 }, { "epoch": 0.52, "learning_rate": 5.751803837937304e-05, "loss": 1.0334, "step": 5145 }, { "epoch": 0.52, "learning_rate": 5.751708516284951e-05, "loss": 1.6906, "step": 5146 }, { "epoch": 0.52, "learning_rate": 5.751613177121772e-05, "loss": 0.7854, "step": 5147 }, { "epoch": 0.52, "learning_rate": 5.751517820448374e-05, "loss": 0.6072, "step": 5148 }, { "epoch": 0.52, "learning_rate": 5.7514224462653636e-05, "loss": 0.3824, "step": 5149 }, { "epoch": 0.52, "learning_rate": 5.751327054573347e-05, "loss": 1.183, "step": 5150 }, { "epoch": 0.52, "learning_rate": 5.751231645372932e-05, "loss": 0.5085, "step": 5151 }, { "epoch": 0.52, "learning_rate": 5.751136218664725e-05, "loss": 1.1385, "step": 5152 }, { "epoch": 0.52, "learning_rate": 5.751040774449334e-05, "loss": 0.614, "step": 5153 }, { "epoch": 0.52, "learning_rate": 5.750945312727366e-05, "loss": 1.0926, "step": 5154 }, { "epoch": 0.52, "learning_rate": 5.750849833499429e-05, "loss": 1.3532, "step": 5155 }, { "epoch": 0.52, "learning_rate": 5.7507543367661294e-05, "loss": 1.2455, "step": 5156 }, { "epoch": 0.52, "learning_rate": 5.750658822528076e-05, "loss": 1.0926, "step": 5157 }, { "epoch": 0.52, "learning_rate": 5.750563290785876e-05, "loss": 0.5375, "step": 5158 }, { "epoch": 0.52, "learning_rate": 5.750467741540138e-05, "loss": 1.3504, "step": 5159 }, { "epoch": 0.52, "learning_rate": 5.750372174791469e-05, "loss": 0.5499, "step": 5160 }, { "epoch": 0.52, "learning_rate": 5.7502765905404785e-05, "loss": 0.5532, "step": 5161 }, { "epoch": 0.52, "learning_rate": 5.750180988787773e-05, "loss": 1.0962, "step": 5162 }, { "epoch": 0.52, "learning_rate": 5.750085369533963e-05, "loss": 0.6333, "step": 5163 }, { "epoch": 0.52, "learning_rate": 5.749989732779655e-05, "loss": 0.5717, "step": 5164 }, { "epoch": 0.52, "learning_rate": 5.749894078525459e-05, "loss": 0.4978, "step": 5165 }, { "epoch": 0.52, "learning_rate": 5.749798406771984e-05, "loss": 0.3178, "step": 5166 }, { "epoch": 0.52, "learning_rate": 5.7497027175198366e-05, "loss": 0.2397, "step": 5167 }, { "epoch": 0.52, "learning_rate": 5.749607010769628e-05, "loss": 1.4312, "step": 5168 }, { "epoch": 0.52, "learning_rate": 5.7495112865219663e-05, "loss": 0.4739, "step": 5169 }, { "epoch": 0.52, "learning_rate": 5.74941554477746e-05, "loss": 1.1106, "step": 5170 }, { "epoch": 0.53, "learning_rate": 5.74931978553672e-05, "loss": 0.6613, "step": 5171 }, { "epoch": 0.53, "learning_rate": 5.749224008800354e-05, "loss": 0.8123, "step": 5172 }, { "epoch": 0.53, "learning_rate": 5.7491282145689735e-05, "loss": 0.3222, "step": 5173 }, { "epoch": 0.53, "learning_rate": 5.749032402843186e-05, "loss": 0.4851, "step": 5174 }, { "epoch": 0.53, "learning_rate": 5.748936573623602e-05, "loss": 1.0068, "step": 5175 }, { "epoch": 0.53, "learning_rate": 5.748840726910832e-05, "loss": 0.8183, "step": 5176 }, { "epoch": 0.53, "learning_rate": 5.7487448627054844e-05, "loss": 1.3195, "step": 5177 }, { "epoch": 0.53, "learning_rate": 5.748648981008171e-05, "loss": 0.8792, "step": 5178 }, { "epoch": 0.53, "learning_rate": 5.7485530818195005e-05, "loss": 0.9176, "step": 5179 }, { "epoch": 0.53, "learning_rate": 5.748457165140084e-05, "loss": 0.6012, "step": 5180 }, { "epoch": 0.53, "learning_rate": 5.748361230970532e-05, "loss": 0.5491, "step": 5181 }, { "epoch": 0.53, "learning_rate": 5.7482652793114536e-05, "loss": 1.9012, "step": 5182 }, { "epoch": 0.53, "learning_rate": 5.748169310163462e-05, "loss": 0.8786, "step": 5183 }, { "epoch": 0.53, "learning_rate": 5.748073323527165e-05, "loss": 0.3814, "step": 5184 }, { "epoch": 0.53, "learning_rate": 5.747977319403175e-05, "loss": 2.0822, "step": 5185 }, { "epoch": 0.53, "learning_rate": 5.7478812977921034e-05, "loss": 1.0314, "step": 5186 }, { "epoch": 0.53, "learning_rate": 5.747785258694561e-05, "loss": 1.517, "step": 5187 }, { "epoch": 0.53, "learning_rate": 5.7476892021111575e-05, "loss": 0.5671, "step": 5188 }, { "epoch": 0.53, "learning_rate": 5.747593128042506e-05, "loss": 0.5629, "step": 5189 }, { "epoch": 0.53, "learning_rate": 5.747497036489216e-05, "loss": 0.4282, "step": 5190 }, { "epoch": 0.53, "learning_rate": 5.7474009274519004e-05, "loss": 0.6205, "step": 5191 }, { "epoch": 0.53, "learning_rate": 5.7473048009311706e-05, "loss": 0.3221, "step": 5192 }, { "epoch": 0.53, "learning_rate": 5.7472086569276387e-05, "loss": 1.0929, "step": 5193 }, { "epoch": 0.53, "learning_rate": 5.747112495441915e-05, "loss": 0.6129, "step": 5194 }, { "epoch": 0.53, "learning_rate": 5.7470163164746136e-05, "loss": 0.5697, "step": 5195 }, { "epoch": 0.53, "learning_rate": 5.746920120026344e-05, "loss": 0.8895, "step": 5196 }, { "epoch": 0.53, "learning_rate": 5.746823906097722e-05, "loss": 0.1982, "step": 5197 }, { "epoch": 0.53, "learning_rate": 5.746727674689355e-05, "loss": 0.5806, "step": 5198 }, { "epoch": 0.53, "learning_rate": 5.7466314258018596e-05, "loss": 1.4763, "step": 5199 }, { "epoch": 0.53, "learning_rate": 5.7465351594358465e-05, "loss": 1.3875, "step": 5200 }, { "epoch": 0.53, "learning_rate": 5.746438875591928e-05, "loss": 0.8067, "step": 5201 }, { "epoch": 0.53, "learning_rate": 5.746342574270718e-05, "loss": 1.1169, "step": 5202 }, { "epoch": 0.53, "learning_rate": 5.746246255472828e-05, "loss": 0.5057, "step": 5203 }, { "epoch": 0.53, "learning_rate": 5.7461499191988714e-05, "loss": 1.471, "step": 5204 }, { "epoch": 0.53, "learning_rate": 5.746053565449461e-05, "loss": 1.3069, "step": 5205 }, { "epoch": 0.53, "learning_rate": 5.745957194225211e-05, "loss": 0.9134, "step": 5206 }, { "epoch": 0.53, "learning_rate": 5.745860805526734e-05, "loss": 0.765, "step": 5207 }, { "epoch": 0.53, "learning_rate": 5.745764399354644e-05, "loss": 0.4046, "step": 5208 }, { "epoch": 0.53, "learning_rate": 5.7456679757095535e-05, "loss": 1.146, "step": 5209 }, { "epoch": 0.53, "learning_rate": 5.745571534592076e-05, "loss": 1.4679, "step": 5210 }, { "epoch": 0.53, "learning_rate": 5.745475076002826e-05, "loss": 1.3294, "step": 5211 }, { "epoch": 0.53, "learning_rate": 5.745378599942417e-05, "loss": 1.0397, "step": 5212 }, { "epoch": 0.53, "learning_rate": 5.745282106411463e-05, "loss": 0.5218, "step": 5213 }, { "epoch": 0.53, "learning_rate": 5.745185595410579e-05, "loss": 0.7614, "step": 5214 }, { "epoch": 0.53, "learning_rate": 5.7450890669403764e-05, "loss": 1.1768, "step": 5215 }, { "epoch": 0.53, "learning_rate": 5.744992521001473e-05, "loss": 0.4899, "step": 5216 }, { "epoch": 0.53, "learning_rate": 5.74489595759448e-05, "loss": 0.3363, "step": 5217 }, { "epoch": 0.53, "learning_rate": 5.744799376720014e-05, "loss": 0.4786, "step": 5218 }, { "epoch": 0.53, "learning_rate": 5.7447027783786896e-05, "loss": 0.5619, "step": 5219 }, { "epoch": 0.53, "learning_rate": 5.7446061625711206e-05, "loss": 1.4779, "step": 5220 }, { "epoch": 0.53, "learning_rate": 5.744509529297921e-05, "loss": 0.6889, "step": 5221 }, { "epoch": 0.53, "learning_rate": 5.7444128785597076e-05, "loss": 1.3041, "step": 5222 }, { "epoch": 0.53, "learning_rate": 5.744316210357095e-05, "loss": 0.2338, "step": 5223 }, { "epoch": 0.53, "learning_rate": 5.744219524690697e-05, "loss": 0.4597, "step": 5224 }, { "epoch": 0.53, "learning_rate": 5.7441228215611306e-05, "loss": 0.3761, "step": 5225 }, { "epoch": 0.53, "learning_rate": 5.7440261009690104e-05, "loss": 0.2165, "step": 5226 }, { "epoch": 0.53, "learning_rate": 5.7439293629149515e-05, "loss": 1.196, "step": 5227 }, { "epoch": 0.53, "learning_rate": 5.743832607399571e-05, "loss": 1.1989, "step": 5228 }, { "epoch": 0.53, "learning_rate": 5.743735834423483e-05, "loss": 0.7689, "step": 5229 }, { "epoch": 0.53, "learning_rate": 5.7436390439873036e-05, "loss": 1.7019, "step": 5230 }, { "epoch": 0.53, "learning_rate": 5.74354223609165e-05, "loss": 0.8981, "step": 5231 }, { "epoch": 0.53, "learning_rate": 5.743445410737136e-05, "loss": 1.6555, "step": 5232 }, { "epoch": 0.53, "learning_rate": 5.743348567924381e-05, "loss": 1.2861, "step": 5233 }, { "epoch": 0.53, "learning_rate": 5.743251707653997e-05, "loss": 0.2468, "step": 5234 }, { "epoch": 0.53, "learning_rate": 5.743154829926604e-05, "loss": 1.1048, "step": 5235 }, { "epoch": 0.53, "learning_rate": 5.743057934742817e-05, "loss": 1.1077, "step": 5236 }, { "epoch": 0.53, "learning_rate": 5.7429610221032526e-05, "loss": 0.3837, "step": 5237 }, { "epoch": 0.53, "learning_rate": 5.742864092008529e-05, "loss": 1.4546, "step": 5238 }, { "epoch": 0.53, "learning_rate": 5.7427671444592605e-05, "loss": 0.7057, "step": 5239 }, { "epoch": 0.53, "learning_rate": 5.742670179456066e-05, "loss": 0.6651, "step": 5240 }, { "epoch": 0.53, "learning_rate": 5.742573196999561e-05, "loss": 0.9309, "step": 5241 }, { "epoch": 0.53, "learning_rate": 5.7424761970903634e-05, "loss": 1.2617, "step": 5242 }, { "epoch": 0.53, "learning_rate": 5.742379179729091e-05, "loss": 0.8255, "step": 5243 }, { "epoch": 0.53, "learning_rate": 5.742282144916361e-05, "loss": 0.7807, "step": 5244 }, { "epoch": 0.53, "learning_rate": 5.742185092652791e-05, "loss": 0.6661, "step": 5245 }, { "epoch": 0.53, "learning_rate": 5.742088022938997e-05, "loss": 1.1492, "step": 5246 }, { "epoch": 0.53, "learning_rate": 5.741990935775599e-05, "loss": 1.3604, "step": 5247 }, { "epoch": 0.53, "learning_rate": 5.741893831163213e-05, "loss": 0.6377, "step": 5248 }, { "epoch": 0.53, "learning_rate": 5.741796709102458e-05, "loss": 0.7951, "step": 5249 }, { "epoch": 0.53, "learning_rate": 5.7416995695939524e-05, "loss": 0.898, "step": 5250 }, { "epoch": 0.53, "learning_rate": 5.7416024126383136e-05, "loss": 0.3118, "step": 5251 }, { "epoch": 0.53, "learning_rate": 5.7415052382361596e-05, "loss": 1.6568, "step": 5252 }, { "epoch": 0.53, "learning_rate": 5.741408046388109e-05, "loss": 0.5093, "step": 5253 }, { "epoch": 0.53, "learning_rate": 5.7413108370947816e-05, "loss": 1.5717, "step": 5254 }, { "epoch": 0.53, "learning_rate": 5.741213610356793e-05, "loss": 0.4469, "step": 5255 }, { "epoch": 0.53, "learning_rate": 5.741116366174766e-05, "loss": 1.7029, "step": 5256 }, { "epoch": 0.53, "learning_rate": 5.7410191045493154e-05, "loss": 0.5213, "step": 5257 }, { "epoch": 0.53, "learning_rate": 5.7409218254810626e-05, "loss": 0.9326, "step": 5258 }, { "epoch": 0.53, "learning_rate": 5.740824528970626e-05, "loss": 0.3739, "step": 5259 }, { "epoch": 0.53, "learning_rate": 5.740727215018625e-05, "loss": 0.512, "step": 5260 }, { "epoch": 0.53, "learning_rate": 5.7406298836256785e-05, "loss": 0.2063, "step": 5261 }, { "epoch": 0.53, "learning_rate": 5.7405325347924054e-05, "loss": 1.3163, "step": 5262 }, { "epoch": 0.53, "learning_rate": 5.740435168519427e-05, "loss": 0.8674, "step": 5263 }, { "epoch": 0.53, "learning_rate": 5.74033778480736e-05, "loss": 0.952, "step": 5264 }, { "epoch": 0.53, "learning_rate": 5.7402403836568275e-05, "loss": 0.3169, "step": 5265 }, { "epoch": 0.53, "learning_rate": 5.740142965068447e-05, "loss": 0.8278, "step": 5266 }, { "epoch": 0.53, "learning_rate": 5.740045529042839e-05, "loss": 1.3784, "step": 5267 }, { "epoch": 0.53, "learning_rate": 5.7399480755806235e-05, "loss": 0.6364, "step": 5268 }, { "epoch": 0.54, "learning_rate": 5.739850604682421e-05, "loss": 0.8567, "step": 5269 }, { "epoch": 0.54, "learning_rate": 5.739753116348852e-05, "loss": 1.3387, "step": 5270 }, { "epoch": 0.54, "learning_rate": 5.739655610580536e-05, "loss": 1.0708, "step": 5271 }, { "epoch": 0.54, "learning_rate": 5.739558087378093e-05, "loss": 1.5922, "step": 5272 }, { "epoch": 0.54, "learning_rate": 5.739460546742146e-05, "loss": 0.7124, "step": 5273 }, { "epoch": 0.54, "learning_rate": 5.7393629886733136e-05, "loss": 0.4345, "step": 5274 }, { "epoch": 0.54, "learning_rate": 5.739265413172217e-05, "loss": 0.6896, "step": 5275 }, { "epoch": 0.54, "learning_rate": 5.739167820239478e-05, "loss": 1.0459, "step": 5276 }, { "epoch": 0.54, "learning_rate": 5.739070209875717e-05, "loss": 0.4152, "step": 5277 }, { "epoch": 0.54, "learning_rate": 5.738972582081555e-05, "loss": 0.6307, "step": 5278 }, { "epoch": 0.54, "learning_rate": 5.738874936857612e-05, "loss": 1.1596, "step": 5279 }, { "epoch": 0.54, "learning_rate": 5.738777274204513e-05, "loss": 0.7807, "step": 5280 }, { "epoch": 0.54, "learning_rate": 5.738679594122876e-05, "loss": 0.9352, "step": 5281 }, { "epoch": 0.54, "learning_rate": 5.7385818966133245e-05, "loss": 0.4838, "step": 5282 }, { "epoch": 0.54, "learning_rate": 5.738484181676479e-05, "loss": 0.3394, "step": 5283 }, { "epoch": 0.54, "learning_rate": 5.7383864493129625e-05, "loss": 0.479, "step": 5284 }, { "epoch": 0.54, "learning_rate": 5.738288699523396e-05, "loss": 1.2525, "step": 5285 }, { "epoch": 0.54, "learning_rate": 5.738190932308402e-05, "loss": 0.6336, "step": 5286 }, { "epoch": 0.54, "learning_rate": 5.7380931476686036e-05, "loss": 1.4065, "step": 5287 }, { "epoch": 0.54, "learning_rate": 5.7379953456046206e-05, "loss": 0.5024, "step": 5288 }, { "epoch": 0.54, "learning_rate": 5.737897526117077e-05, "loss": 0.5721, "step": 5289 }, { "epoch": 0.54, "learning_rate": 5.7377996892065956e-05, "loss": 1.0342, "step": 5290 }, { "epoch": 0.54, "learning_rate": 5.7377018348737985e-05, "loss": 0.517, "step": 5291 }, { "epoch": 0.54, "learning_rate": 5.737603963119308e-05, "loss": 1.0513, "step": 5292 }, { "epoch": 0.54, "learning_rate": 5.737506073943748e-05, "loss": 1.7179, "step": 5293 }, { "epoch": 0.54, "learning_rate": 5.7374081673477406e-05, "loss": 0.4172, "step": 5294 }, { "epoch": 0.54, "learning_rate": 5.7373102433319084e-05, "loss": 1.4269, "step": 5295 }, { "epoch": 0.54, "learning_rate": 5.7372123018968756e-05, "loss": 0.2776, "step": 5296 }, { "epoch": 0.54, "learning_rate": 5.737114343043264e-05, "loss": 1.389, "step": 5297 }, { "epoch": 0.54, "learning_rate": 5.737016366771699e-05, "loss": 1.2982, "step": 5298 }, { "epoch": 0.54, "learning_rate": 5.736918373082803e-05, "loss": 1.1387, "step": 5299 }, { "epoch": 0.54, "learning_rate": 5.7368203619771994e-05, "loss": 1.0778, "step": 5300 }, { "epoch": 0.54, "learning_rate": 5.736722333455512e-05, "loss": 1.3609, "step": 5301 }, { "epoch": 0.54, "learning_rate": 5.7366242875183655e-05, "loss": 0.4157, "step": 5302 }, { "epoch": 0.54, "learning_rate": 5.7365262241663827e-05, "loss": 0.5608, "step": 5303 }, { "epoch": 0.54, "learning_rate": 5.736428143400188e-05, "loss": 0.2312, "step": 5304 }, { "epoch": 0.54, "learning_rate": 5.736330045220406e-05, "loss": 0.4858, "step": 5305 }, { "epoch": 0.54, "learning_rate": 5.736231929627659e-05, "loss": 0.9325, "step": 5306 }, { "epoch": 0.54, "learning_rate": 5.736133796622575e-05, "loss": 0.3019, "step": 5307 }, { "epoch": 0.54, "learning_rate": 5.7360356462057754e-05, "loss": 0.6096, "step": 5308 }, { "epoch": 0.54, "learning_rate": 5.735937478377886e-05, "loss": 1.2555, "step": 5309 }, { "epoch": 0.54, "learning_rate": 5.73583929313953e-05, "loss": 1.3346, "step": 5310 }, { "epoch": 0.54, "learning_rate": 5.735741090491335e-05, "loss": 0.2781, "step": 5311 }, { "epoch": 0.54, "learning_rate": 5.735642870433924e-05, "loss": 1.4316, "step": 5312 }, { "epoch": 0.54, "learning_rate": 5.7355446329679225e-05, "loss": 1.5764, "step": 5313 }, { "epoch": 0.54, "learning_rate": 5.7354463780939556e-05, "loss": 0.5442, "step": 5314 }, { "epoch": 0.54, "learning_rate": 5.7353481058126484e-05, "loss": 0.7837, "step": 5315 }, { "epoch": 0.54, "learning_rate": 5.7352498161246265e-05, "loss": 1.3522, "step": 5316 }, { "epoch": 0.54, "learning_rate": 5.735151509030516e-05, "loss": 0.762, "step": 5317 }, { "epoch": 0.54, "learning_rate": 5.735053184530941e-05, "loss": 1.3382, "step": 5318 }, { "epoch": 0.54, "learning_rate": 5.734954842626528e-05, "loss": 1.181, "step": 5319 }, { "epoch": 0.54, "learning_rate": 5.734856483317903e-05, "loss": 1.2884, "step": 5320 }, { "epoch": 0.54, "learning_rate": 5.7347581066056916e-05, "loss": 0.9174, "step": 5321 }, { "epoch": 0.54, "learning_rate": 5.73465971249052e-05, "loss": 0.3203, "step": 5322 }, { "epoch": 0.54, "learning_rate": 5.734561300973014e-05, "loss": 0.7196, "step": 5323 }, { "epoch": 0.54, "learning_rate": 5.734462872053802e-05, "loss": 1.2237, "step": 5324 }, { "epoch": 0.54, "learning_rate": 5.7343644257335065e-05, "loss": 1.1445, "step": 5325 }, { "epoch": 0.54, "learning_rate": 5.734265962012757e-05, "loss": 0.3809, "step": 5326 }, { "epoch": 0.54, "learning_rate": 5.7341674808921786e-05, "loss": 0.6845, "step": 5327 }, { "epoch": 0.54, "learning_rate": 5.7340689823723984e-05, "loss": 1.0162, "step": 5328 }, { "epoch": 0.54, "learning_rate": 5.733970466454043e-05, "loss": 0.5561, "step": 5329 }, { "epoch": 0.54, "learning_rate": 5.733871933137741e-05, "loss": 0.8231, "step": 5330 }, { "epoch": 0.54, "learning_rate": 5.733773382424117e-05, "loss": 0.493, "step": 5331 }, { "epoch": 0.54, "learning_rate": 5.733674814313799e-05, "loss": 1.2845, "step": 5332 }, { "epoch": 0.54, "learning_rate": 5.7335762288074154e-05, "loss": 1.1984, "step": 5333 }, { "epoch": 0.54, "learning_rate": 5.733477625905592e-05, "loss": 1.2675, "step": 5334 }, { "epoch": 0.54, "learning_rate": 5.733379005608956e-05, "loss": 0.3411, "step": 5335 }, { "epoch": 0.54, "learning_rate": 5.7332803679181364e-05, "loss": 0.4623, "step": 5336 }, { "epoch": 0.54, "learning_rate": 5.73318171283376e-05, "loss": 0.5768, "step": 5337 }, { "epoch": 0.54, "learning_rate": 5.7330830403564564e-05, "loss": 0.9943, "step": 5338 }, { "epoch": 0.54, "learning_rate": 5.732984350486851e-05, "loss": 0.8107, "step": 5339 }, { "epoch": 0.54, "learning_rate": 5.732885643225573e-05, "loss": 0.5727, "step": 5340 }, { "epoch": 0.54, "learning_rate": 5.7327869185732505e-05, "loss": 1.5054, "step": 5341 }, { "epoch": 0.54, "learning_rate": 5.732688176530511e-05, "loss": 0.8214, "step": 5342 }, { "epoch": 0.54, "learning_rate": 5.7325894170979844e-05, "loss": 0.9082, "step": 5343 }, { "epoch": 0.54, "learning_rate": 5.732490640276298e-05, "loss": 0.9414, "step": 5344 }, { "epoch": 0.54, "learning_rate": 5.7323918460660804e-05, "loss": 0.7423, "step": 5345 }, { "epoch": 0.54, "learning_rate": 5.73229303446796e-05, "loss": 0.7322, "step": 5346 }, { "epoch": 0.54, "learning_rate": 5.732194205482568e-05, "loss": 0.9761, "step": 5347 }, { "epoch": 0.54, "learning_rate": 5.7320953591105304e-05, "loss": 0.9507, "step": 5348 }, { "epoch": 0.54, "learning_rate": 5.731996495352478e-05, "loss": 1.4479, "step": 5349 }, { "epoch": 0.54, "learning_rate": 5.731897614209038e-05, "loss": 0.3216, "step": 5350 }, { "epoch": 0.54, "learning_rate": 5.7317987156808404e-05, "loss": 1.0809, "step": 5351 }, { "epoch": 0.54, "learning_rate": 5.731699799768516e-05, "loss": 1.2037, "step": 5352 }, { "epoch": 0.54, "learning_rate": 5.7316008664726934e-05, "loss": 1.1168, "step": 5353 }, { "epoch": 0.54, "learning_rate": 5.731501915794002e-05, "loss": 0.5549, "step": 5354 }, { "epoch": 0.54, "learning_rate": 5.731402947733071e-05, "loss": 0.5056, "step": 5355 }, { "epoch": 0.54, "learning_rate": 5.731303962290532e-05, "loss": 0.3824, "step": 5356 }, { "epoch": 0.54, "learning_rate": 5.731204959467012e-05, "loss": 0.6601, "step": 5357 }, { "epoch": 0.54, "learning_rate": 5.731105939263143e-05, "loss": 0.5515, "step": 5358 }, { "epoch": 0.54, "learning_rate": 5.731006901679555e-05, "loss": 0.4723, "step": 5359 }, { "epoch": 0.54, "learning_rate": 5.730907846716878e-05, "loss": 0.8134, "step": 5360 }, { "epoch": 0.54, "learning_rate": 5.7308087743757424e-05, "loss": 1.2668, "step": 5361 }, { "epoch": 0.54, "learning_rate": 5.730709684656779e-05, "loss": 0.9245, "step": 5362 }, { "epoch": 0.54, "learning_rate": 5.730610577560617e-05, "loss": 1.0989, "step": 5363 }, { "epoch": 0.54, "learning_rate": 5.7305114530878895e-05, "loss": 0.3165, "step": 5364 }, { "epoch": 0.54, "learning_rate": 5.730412311239225e-05, "loss": 0.6695, "step": 5365 }, { "epoch": 0.54, "learning_rate": 5.7303131520152553e-05, "loss": 0.4067, "step": 5366 }, { "epoch": 0.54, "learning_rate": 5.730213975416611e-05, "loss": 1.3367, "step": 5367 }, { "epoch": 0.55, "learning_rate": 5.730114781443924e-05, "loss": 0.635, "step": 5368 }, { "epoch": 0.55, "learning_rate": 5.730015570097825e-05, "loss": 1.1641, "step": 5369 }, { "epoch": 0.55, "learning_rate": 5.729916341378946e-05, "loss": 1.8184, "step": 5370 }, { "epoch": 0.55, "learning_rate": 5.729817095287917e-05, "loss": 1.0689, "step": 5371 }, { "epoch": 0.55, "learning_rate": 5.729717831825371e-05, "loss": 0.616, "step": 5372 }, { "epoch": 0.55, "learning_rate": 5.729618550991939e-05, "loss": 0.7754, "step": 5373 }, { "epoch": 0.55, "learning_rate": 5.729519252788253e-05, "loss": 0.5778, "step": 5374 }, { "epoch": 0.55, "learning_rate": 5.729419937214945e-05, "loss": 1.302, "step": 5375 }, { "epoch": 0.55, "learning_rate": 5.729320604272647e-05, "loss": 0.7887, "step": 5376 }, { "epoch": 0.55, "learning_rate": 5.72922125396199e-05, "loss": 1.0606, "step": 5377 }, { "epoch": 0.55, "learning_rate": 5.729121886283608e-05, "loss": 0.3292, "step": 5378 }, { "epoch": 0.55, "learning_rate": 5.729022501238133e-05, "loss": 0.5953, "step": 5379 }, { "epoch": 0.55, "learning_rate": 5.728923098826196e-05, "loss": 0.3727, "step": 5380 }, { "epoch": 0.55, "learning_rate": 5.7288236790484306e-05, "loss": 0.3981, "step": 5381 }, { "epoch": 0.55, "learning_rate": 5.7287242419054695e-05, "loss": 1.004, "step": 5382 }, { "epoch": 0.55, "learning_rate": 5.728624787397946e-05, "loss": 0.6403, "step": 5383 }, { "epoch": 0.55, "learning_rate": 5.7285253155264924e-05, "loss": 1.2202, "step": 5384 }, { "epoch": 0.55, "learning_rate": 5.728425826291741e-05, "loss": 1.0034, "step": 5385 }, { "epoch": 0.55, "learning_rate": 5.728326319694325e-05, "loss": 0.4511, "step": 5386 }, { "epoch": 0.55, "learning_rate": 5.72822679573488e-05, "loss": 0.7494, "step": 5387 }, { "epoch": 0.55, "learning_rate": 5.728127254414036e-05, "loss": 0.5985, "step": 5388 }, { "epoch": 0.55, "learning_rate": 5.728027695732429e-05, "loss": 0.7684, "step": 5389 }, { "epoch": 0.55, "learning_rate": 5.727928119690691e-05, "loss": 1.2555, "step": 5390 }, { "epoch": 0.55, "learning_rate": 5.727828526289456e-05, "loss": 0.8953, "step": 5391 }, { "epoch": 0.55, "learning_rate": 5.727728915529358e-05, "loss": 1.377, "step": 5392 }, { "epoch": 0.55, "learning_rate": 5.727629287411032e-05, "loss": 0.4294, "step": 5393 }, { "epoch": 0.55, "learning_rate": 5.727529641935109e-05, "loss": 0.5936, "step": 5394 }, { "epoch": 0.55, "learning_rate": 5.727429979102227e-05, "loss": 0.6275, "step": 5395 }, { "epoch": 0.55, "learning_rate": 5.727330298913017e-05, "loss": 0.3373, "step": 5396 }, { "epoch": 0.55, "learning_rate": 5.727230601368114e-05, "loss": 0.9276, "step": 5397 }, { "epoch": 0.55, "learning_rate": 5.727130886468154e-05, "loss": 0.5001, "step": 5398 }, { "epoch": 0.55, "learning_rate": 5.7270311542137695e-05, "loss": 1.402, "step": 5399 }, { "epoch": 0.55, "learning_rate": 5.7269314046055976e-05, "loss": 0.4343, "step": 5400 }, { "epoch": 0.55, "learning_rate": 5.7268316376442705e-05, "loss": 0.5003, "step": 5401 }, { "epoch": 0.55, "learning_rate": 5.7267318533304245e-05, "loss": 1.0835, "step": 5402 }, { "epoch": 0.55, "learning_rate": 5.726632051664694e-05, "loss": 0.2552, "step": 5403 }, { "epoch": 0.55, "learning_rate": 5.726532232647715e-05, "loss": 0.7406, "step": 5404 }, { "epoch": 0.55, "learning_rate": 5.726432396280121e-05, "loss": 0.779, "step": 5405 }, { "epoch": 0.55, "learning_rate": 5.72633254256255e-05, "loss": 0.9489, "step": 5406 }, { "epoch": 0.55, "learning_rate": 5.726232671495635e-05, "loss": 0.6779, "step": 5407 }, { "epoch": 0.55, "learning_rate": 5.7261327830800124e-05, "loss": 1.0155, "step": 5408 }, { "epoch": 0.55, "learning_rate": 5.726032877316317e-05, "loss": 1.3712, "step": 5409 }, { "epoch": 0.55, "learning_rate": 5.7259329542051865e-05, "loss": 0.2887, "step": 5410 }, { "epoch": 0.55, "learning_rate": 5.7258330137472555e-05, "loss": 1.6252, "step": 5411 }, { "epoch": 0.55, "learning_rate": 5.7257330559431604e-05, "loss": 1.4843, "step": 5412 }, { "epoch": 0.55, "learning_rate": 5.725633080793536e-05, "loss": 1.6219, "step": 5413 }, { "epoch": 0.55, "learning_rate": 5.725533088299021e-05, "loss": 0.5592, "step": 5414 }, { "epoch": 0.55, "learning_rate": 5.725433078460249e-05, "loss": 1.1503, "step": 5415 }, { "epoch": 0.55, "learning_rate": 5.725333051277859e-05, "loss": 0.5382, "step": 5416 }, { "epoch": 0.55, "learning_rate": 5.725233006752485e-05, "loss": 0.2903, "step": 5417 }, { "epoch": 0.55, "learning_rate": 5.7251329448847656e-05, "loss": 1.0825, "step": 5418 }, { "epoch": 0.55, "learning_rate": 5.725032865675337e-05, "loss": 0.9619, "step": 5419 }, { "epoch": 0.55, "learning_rate": 5.724932769124835e-05, "loss": 1.7268, "step": 5420 }, { "epoch": 0.55, "learning_rate": 5.724832655233898e-05, "loss": 0.5806, "step": 5421 }, { "epoch": 0.55, "learning_rate": 5.7247325240031626e-05, "loss": 0.3407, "step": 5422 }, { "epoch": 0.55, "learning_rate": 5.724632375433266e-05, "loss": 1.0794, "step": 5423 }, { "epoch": 0.55, "learning_rate": 5.7245322095248454e-05, "loss": 0.7496, "step": 5424 }, { "epoch": 0.55, "learning_rate": 5.724432026278538e-05, "loss": 0.7125, "step": 5425 }, { "epoch": 0.55, "learning_rate": 5.724331825694981e-05, "loss": 0.3637, "step": 5426 }, { "epoch": 0.55, "learning_rate": 5.724231607774815e-05, "loss": 0.3754, "step": 5427 }, { "epoch": 0.55, "learning_rate": 5.724131372518673e-05, "loss": 0.4595, "step": 5428 }, { "epoch": 0.55, "learning_rate": 5.7240311199271966e-05, "loss": 0.4615, "step": 5429 }, { "epoch": 0.55, "learning_rate": 5.723930850001022e-05, "loss": 1.1547, "step": 5430 }, { "epoch": 0.55, "learning_rate": 5.723830562740787e-05, "loss": 1.1979, "step": 5431 }, { "epoch": 0.55, "learning_rate": 5.7237302581471314e-05, "loss": 0.4228, "step": 5432 }, { "epoch": 0.55, "learning_rate": 5.7236299362206925e-05, "loss": 1.0828, "step": 5433 }, { "epoch": 0.55, "learning_rate": 5.723529596962108e-05, "loss": 0.9788, "step": 5434 }, { "epoch": 0.55, "learning_rate": 5.723429240372018e-05, "loss": 1.3334, "step": 5435 }, { "epoch": 0.55, "learning_rate": 5.723328866451061e-05, "loss": 1.1388, "step": 5436 }, { "epoch": 0.55, "learning_rate": 5.723228475199874e-05, "loss": 0.7258, "step": 5437 }, { "epoch": 0.55, "learning_rate": 5.723128066619097e-05, "loss": 0.753, "step": 5438 }, { "epoch": 0.55, "learning_rate": 5.723027640709369e-05, "loss": 0.8152, "step": 5439 }, { "epoch": 0.55, "learning_rate": 5.72292719747133e-05, "loss": 0.6176, "step": 5440 }, { "epoch": 0.55, "learning_rate": 5.722826736905618e-05, "loss": 0.8456, "step": 5441 }, { "epoch": 0.55, "learning_rate": 5.722726259012872e-05, "loss": 1.2093, "step": 5442 }, { "epoch": 0.55, "learning_rate": 5.7226257637937315e-05, "loss": 1.096, "step": 5443 }, { "epoch": 0.55, "learning_rate": 5.7225252512488365e-05, "loss": 1.1909, "step": 5444 }, { "epoch": 0.55, "learning_rate": 5.7224247213788265e-05, "loss": 0.7148, "step": 5445 }, { "epoch": 0.55, "learning_rate": 5.722324174184341e-05, "loss": 0.3738, "step": 5446 }, { "epoch": 0.55, "learning_rate": 5.722223609666021e-05, "loss": 0.2815, "step": 5447 }, { "epoch": 0.55, "learning_rate": 5.722123027824505e-05, "loss": 0.5374, "step": 5448 }, { "epoch": 0.55, "learning_rate": 5.722022428660433e-05, "loss": 1.2861, "step": 5449 }, { "epoch": 0.55, "learning_rate": 5.7219218121744464e-05, "loss": 0.7391, "step": 5450 }, { "epoch": 0.55, "learning_rate": 5.721821178367185e-05, "loss": 0.5965, "step": 5451 }, { "epoch": 0.55, "learning_rate": 5.7217205272392886e-05, "loss": 0.9767, "step": 5452 }, { "epoch": 0.55, "learning_rate": 5.721619858791398e-05, "loss": 0.3625, "step": 5453 }, { "epoch": 0.55, "learning_rate": 5.721519173024155e-05, "loss": 1.1092, "step": 5454 }, { "epoch": 0.55, "learning_rate": 5.7214184699381994e-05, "loss": 0.7242, "step": 5455 }, { "epoch": 0.55, "learning_rate": 5.7213177495341696e-05, "loss": 0.5163, "step": 5456 }, { "epoch": 0.55, "learning_rate": 5.721217011812711e-05, "loss": 1.4254, "step": 5457 }, { "epoch": 0.55, "learning_rate": 5.7211162567744624e-05, "loss": 1.1584, "step": 5458 }, { "epoch": 0.55, "learning_rate": 5.721015484420065e-05, "loss": 0.5658, "step": 5459 }, { "epoch": 0.55, "learning_rate": 5.72091469475016e-05, "loss": 1.8828, "step": 5460 }, { "epoch": 0.55, "learning_rate": 5.720813887765389e-05, "loss": 1.308, "step": 5461 }, { "epoch": 0.55, "learning_rate": 5.720713063466393e-05, "loss": 1.6957, "step": 5462 }, { "epoch": 0.55, "learning_rate": 5.720612221853815e-05, "loss": 1.4021, "step": 5463 }, { "epoch": 0.55, "learning_rate": 5.720511362928294e-05, "loss": 1.067, "step": 5464 }, { "epoch": 0.55, "learning_rate": 5.7204104866904755e-05, "loss": 1.041, "step": 5465 }, { "epoch": 0.56, "learning_rate": 5.7203095931409986e-05, "loss": 0.9776, "step": 5466 }, { "epoch": 0.56, "learning_rate": 5.720208682280507e-05, "loss": 0.8097, "step": 5467 }, { "epoch": 0.56, "learning_rate": 5.720107754109642e-05, "loss": 1.0081, "step": 5468 }, { "epoch": 0.56, "learning_rate": 5.720006808629046e-05, "loss": 0.6186, "step": 5469 }, { "epoch": 0.56, "learning_rate": 5.719905845839361e-05, "loss": 1.5912, "step": 5470 }, { "epoch": 0.56, "learning_rate": 5.7198048657412304e-05, "loss": 0.8953, "step": 5471 }, { "epoch": 0.56, "learning_rate": 5.7197038683352956e-05, "loss": 0.7486, "step": 5472 }, { "epoch": 0.56, "learning_rate": 5.7196028536222e-05, "loss": 0.6646, "step": 5473 }, { "epoch": 0.56, "learning_rate": 5.719501821602587e-05, "loss": 0.6672, "step": 5474 }, { "epoch": 0.56, "learning_rate": 5.719400772277099e-05, "loss": 0.3792, "step": 5475 }, { "epoch": 0.56, "learning_rate": 5.71929970564638e-05, "loss": 0.4935, "step": 5476 }, { "epoch": 0.56, "learning_rate": 5.71919862171107e-05, "loss": 1.3422, "step": 5477 }, { "epoch": 0.56, "learning_rate": 5.719097520471816e-05, "loss": 0.3579, "step": 5478 }, { "epoch": 0.56, "learning_rate": 5.718996401929259e-05, "loss": 1.4375, "step": 5479 }, { "epoch": 0.56, "learning_rate": 5.7188952660840434e-05, "loss": 0.7536, "step": 5480 }, { "epoch": 0.56, "learning_rate": 5.718794112936813e-05, "loss": 1.4927, "step": 5481 }, { "epoch": 0.56, "learning_rate": 5.71869294248821e-05, "loss": 1.1227, "step": 5482 }, { "epoch": 0.56, "learning_rate": 5.7185917547388805e-05, "loss": 0.9996, "step": 5483 }, { "epoch": 0.56, "learning_rate": 5.718490549689467e-05, "loss": 0.2974, "step": 5484 }, { "epoch": 0.56, "learning_rate": 5.7183893273406136e-05, "loss": 0.6944, "step": 5485 }, { "epoch": 0.56, "learning_rate": 5.718288087692965e-05, "loss": 0.8951, "step": 5486 }, { "epoch": 0.56, "learning_rate": 5.7181868307471647e-05, "loss": 0.4342, "step": 5487 }, { "epoch": 0.56, "learning_rate": 5.718085556503858e-05, "loss": 0.4309, "step": 5488 }, { "epoch": 0.56, "learning_rate": 5.7179842649636885e-05, "loss": 0.3188, "step": 5489 }, { "epoch": 0.56, "learning_rate": 5.7178829561273014e-05, "loss": 0.9428, "step": 5490 }, { "epoch": 0.56, "learning_rate": 5.7177816299953405e-05, "loss": 1.5583, "step": 5491 }, { "epoch": 0.56, "learning_rate": 5.7176802865684514e-05, "loss": 0.5776, "step": 5492 }, { "epoch": 0.56, "learning_rate": 5.717578925847279e-05, "loss": 1.5792, "step": 5493 }, { "epoch": 0.56, "learning_rate": 5.717477547832468e-05, "loss": 0.5854, "step": 5494 }, { "epoch": 0.56, "learning_rate": 5.717376152524664e-05, "loss": 0.9549, "step": 5495 }, { "epoch": 0.56, "learning_rate": 5.7172747399245115e-05, "loss": 0.9187, "step": 5496 }, { "epoch": 0.56, "learning_rate": 5.7171733100326564e-05, "loss": 0.2944, "step": 5497 }, { "epoch": 0.56, "learning_rate": 5.7170718628497445e-05, "loss": 0.8494, "step": 5498 }, { "epoch": 0.56, "learning_rate": 5.71697039837642e-05, "loss": 0.4518, "step": 5499 }, { "epoch": 0.56, "learning_rate": 5.71686891661333e-05, "loss": 1.6186, "step": 5500 }, { "epoch": 0.56, "learning_rate": 5.7167674175611194e-05, "loss": 0.84, "step": 5501 }, { "epoch": 0.56, "learning_rate": 5.7166659012204346e-05, "loss": 1.5556, "step": 5502 }, { "epoch": 0.56, "learning_rate": 5.716564367591922e-05, "loss": 1.3827, "step": 5503 }, { "epoch": 0.56, "learning_rate": 5.716462816676227e-05, "loss": 0.8374, "step": 5504 }, { "epoch": 0.56, "learning_rate": 5.716361248473996e-05, "loss": 0.6958, "step": 5505 }, { "epoch": 0.56, "learning_rate": 5.716259662985875e-05, "loss": 1.5111, "step": 5506 }, { "epoch": 0.56, "learning_rate": 5.716158060212511e-05, "loss": 0.6608, "step": 5507 }, { "epoch": 0.56, "learning_rate": 5.716056440154551e-05, "loss": 1.6615, "step": 5508 }, { "epoch": 0.56, "learning_rate": 5.71595480281264e-05, "loss": 1.1886, "step": 5509 }, { "epoch": 0.56, "learning_rate": 5.715853148187427e-05, "loss": 0.4311, "step": 5510 }, { "epoch": 0.56, "learning_rate": 5.715751476279556e-05, "loss": 0.4094, "step": 5511 }, { "epoch": 0.56, "learning_rate": 5.715649787089677e-05, "loss": 0.2764, "step": 5512 }, { "epoch": 0.56, "learning_rate": 5.715548080618436e-05, "loss": 1.0156, "step": 5513 }, { "epoch": 0.56, "learning_rate": 5.7154463568664793e-05, "loss": 1.3476, "step": 5514 }, { "epoch": 0.56, "learning_rate": 5.715344615834455e-05, "loss": 0.4534, "step": 5515 }, { "epoch": 0.56, "learning_rate": 5.715242857523011e-05, "loss": 0.889, "step": 5516 }, { "epoch": 0.56, "learning_rate": 5.715141081932795e-05, "loss": 1.2035, "step": 5517 }, { "epoch": 0.56, "learning_rate": 5.715039289064453e-05, "loss": 0.2107, "step": 5518 }, { "epoch": 0.56, "learning_rate": 5.7149374789186344e-05, "loss": 1.5279, "step": 5519 }, { "epoch": 0.56, "learning_rate": 5.714835651495986e-05, "loss": 1.0305, "step": 5520 }, { "epoch": 0.56, "learning_rate": 5.7147338067971566e-05, "loss": 0.9832, "step": 5521 }, { "epoch": 0.56, "learning_rate": 5.714631944822794e-05, "loss": 1.2064, "step": 5522 }, { "epoch": 0.56, "learning_rate": 5.7145300655735465e-05, "loss": 1.0173, "step": 5523 }, { "epoch": 0.56, "learning_rate": 5.7144281690500624e-05, "loss": 0.7399, "step": 5524 }, { "epoch": 0.56, "learning_rate": 5.714326255252989e-05, "loss": 1.2819, "step": 5525 }, { "epoch": 0.56, "learning_rate": 5.714224324182977e-05, "loss": 0.9599, "step": 5526 }, { "epoch": 0.56, "learning_rate": 5.714122375840674e-05, "loss": 0.9305, "step": 5527 }, { "epoch": 0.56, "learning_rate": 5.7140204102267285e-05, "loss": 0.6007, "step": 5528 }, { "epoch": 0.56, "learning_rate": 5.713918427341789e-05, "loss": 0.408, "step": 5529 }, { "epoch": 0.56, "learning_rate": 5.713816427186505e-05, "loss": 0.1217, "step": 5530 }, { "epoch": 0.56, "learning_rate": 5.713714409761527e-05, "loss": 1.4237, "step": 5531 }, { "epoch": 0.56, "learning_rate": 5.713612375067502e-05, "loss": 0.354, "step": 5532 }, { "epoch": 0.56, "learning_rate": 5.71351032310508e-05, "loss": 1.1894, "step": 5533 }, { "epoch": 0.56, "learning_rate": 5.713408253874911e-05, "loss": 0.3575, "step": 5534 }, { "epoch": 0.56, "learning_rate": 5.7133061673776434e-05, "loss": 1.1832, "step": 5535 }, { "epoch": 0.56, "learning_rate": 5.713204063613928e-05, "loss": 0.5375, "step": 5536 }, { "epoch": 0.56, "learning_rate": 5.713101942584414e-05, "loss": 0.9021, "step": 5537 }, { "epoch": 0.56, "learning_rate": 5.712999804289752e-05, "loss": 0.7339, "step": 5538 }, { "epoch": 0.56, "learning_rate": 5.7128976487305904e-05, "loss": 0.3484, "step": 5539 }, { "epoch": 0.56, "learning_rate": 5.7127954759075806e-05, "loss": 1.1318, "step": 5540 }, { "epoch": 0.56, "learning_rate": 5.712693285821372e-05, "loss": 2.6407, "step": 5541 }, { "epoch": 0.56, "learning_rate": 5.712591078472616e-05, "loss": 1.2409, "step": 5542 }, { "epoch": 0.56, "learning_rate": 5.712488853861961e-05, "loss": 0.9054, "step": 5543 }, { "epoch": 0.56, "learning_rate": 5.712386611990061e-05, "loss": 1.3265, "step": 5544 }, { "epoch": 0.56, "learning_rate": 5.712284352857563e-05, "loss": 0.7057, "step": 5545 }, { "epoch": 0.56, "learning_rate": 5.712182076465119e-05, "loss": 0.3383, "step": 5546 }, { "epoch": 0.56, "learning_rate": 5.71207978281338e-05, "loss": 1.2238, "step": 5547 }, { "epoch": 0.56, "learning_rate": 5.7119774719029976e-05, "loss": 1.4968, "step": 5548 }, { "epoch": 0.56, "learning_rate": 5.711875143734622e-05, "loss": 0.9457, "step": 5549 }, { "epoch": 0.56, "learning_rate": 5.711772798308905e-05, "loss": 0.2897, "step": 5550 }, { "epoch": 0.56, "learning_rate": 5.711670435626497e-05, "loss": 1.3448, "step": 5551 }, { "epoch": 0.56, "learning_rate": 5.71156805568805e-05, "loss": 0.397, "step": 5552 }, { "epoch": 0.56, "learning_rate": 5.711465658494216e-05, "loss": 0.4717, "step": 5553 }, { "epoch": 0.56, "learning_rate": 5.711363244045645e-05, "loss": 0.5015, "step": 5554 }, { "epoch": 0.56, "learning_rate": 5.7112608123429907e-05, "loss": 0.3747, "step": 5555 }, { "epoch": 0.56, "learning_rate": 5.711158363386904e-05, "loss": 0.5755, "step": 5556 }, { "epoch": 0.56, "learning_rate": 5.711055897178037e-05, "loss": 0.549, "step": 5557 }, { "epoch": 0.56, "learning_rate": 5.7109534137170405e-05, "loss": 0.8004, "step": 5558 }, { "epoch": 0.56, "learning_rate": 5.710850913004568e-05, "loss": 0.2033, "step": 5559 }, { "epoch": 0.56, "learning_rate": 5.7107483950412724e-05, "loss": 1.0782, "step": 5560 }, { "epoch": 0.56, "learning_rate": 5.7106458598278044e-05, "loss": 0.4822, "step": 5561 }, { "epoch": 0.56, "learning_rate": 5.7105433073648174e-05, "loss": 0.4518, "step": 5562 }, { "epoch": 0.56, "learning_rate": 5.710440737652965e-05, "loss": 0.9231, "step": 5563 }, { "epoch": 0.56, "learning_rate": 5.710338150692897e-05, "loss": 1.2473, "step": 5564 }, { "epoch": 0.57, "learning_rate": 5.71023554648527e-05, "loss": 1.1316, "step": 5565 }, { "epoch": 0.57, "learning_rate": 5.7101329250307344e-05, "loss": 0.5806, "step": 5566 }, { "epoch": 0.57, "learning_rate": 5.710030286329944e-05, "loss": 1.0795, "step": 5567 }, { "epoch": 0.57, "learning_rate": 5.7099276303835514e-05, "loss": 1.132, "step": 5568 }, { "epoch": 0.57, "learning_rate": 5.70982495719221e-05, "loss": 0.5017, "step": 5569 }, { "epoch": 0.57, "learning_rate": 5.709722266756574e-05, "loss": 0.7667, "step": 5570 }, { "epoch": 0.57, "learning_rate": 5.709619559077296e-05, "loss": 0.6383, "step": 5571 }, { "epoch": 0.57, "learning_rate": 5.7095168341550296e-05, "loss": 0.9983, "step": 5572 }, { "epoch": 0.57, "learning_rate": 5.709414091990429e-05, "loss": 0.7349, "step": 5573 }, { "epoch": 0.57, "learning_rate": 5.7093113325841485e-05, "loss": 0.6384, "step": 5574 }, { "epoch": 0.57, "learning_rate": 5.709208555936841e-05, "loss": 0.4994, "step": 5575 }, { "epoch": 0.57, "learning_rate": 5.7091057620491616e-05, "loss": 1.2089, "step": 5576 }, { "epoch": 0.57, "learning_rate": 5.709002950921762e-05, "loss": 0.3206, "step": 5577 }, { "epoch": 0.57, "learning_rate": 5.708900122555299e-05, "loss": 0.2144, "step": 5578 }, { "epoch": 0.57, "learning_rate": 5.708797276950426e-05, "loss": 1.3518, "step": 5579 }, { "epoch": 0.57, "learning_rate": 5.708694414107799e-05, "loss": 0.4965, "step": 5580 }, { "epoch": 0.57, "learning_rate": 5.70859153402807e-05, "loss": 1.5926, "step": 5581 }, { "epoch": 0.57, "learning_rate": 5.708488636711895e-05, "loss": 1.7128, "step": 5582 }, { "epoch": 0.57, "learning_rate": 5.708385722159928e-05, "loss": 1.0239, "step": 5583 }, { "epoch": 0.57, "learning_rate": 5.708282790372826e-05, "loss": 1.2725, "step": 5584 }, { "epoch": 0.57, "learning_rate": 5.7081798413512405e-05, "loss": 0.3942, "step": 5585 }, { "epoch": 0.57, "learning_rate": 5.7080768750958306e-05, "loss": 1.1826, "step": 5586 }, { "epoch": 0.57, "learning_rate": 5.707973891607249e-05, "loss": 0.7228, "step": 5587 }, { "epoch": 0.57, "learning_rate": 5.7078708908861516e-05, "loss": 0.6506, "step": 5588 }, { "epoch": 0.57, "learning_rate": 5.707767872933195e-05, "loss": 0.7497, "step": 5589 }, { "epoch": 0.57, "learning_rate": 5.707664837749033e-05, "loss": 0.5559, "step": 5590 }, { "epoch": 0.57, "learning_rate": 5.707561785334322e-05, "loss": 0.4259, "step": 5591 }, { "epoch": 0.57, "learning_rate": 5.7074587156897175e-05, "loss": 0.7337, "step": 5592 }, { "epoch": 0.57, "learning_rate": 5.7073556288158753e-05, "loss": 0.8813, "step": 5593 }, { "epoch": 0.57, "learning_rate": 5.707252524713452e-05, "loss": 0.9626, "step": 5594 }, { "epoch": 0.57, "learning_rate": 5.7071494033831046e-05, "loss": 0.6137, "step": 5595 }, { "epoch": 0.57, "learning_rate": 5.707046264825488e-05, "loss": 0.5052, "step": 5596 }, { "epoch": 0.57, "learning_rate": 5.7069431090412586e-05, "loss": 1.6634, "step": 5597 }, { "epoch": 0.57, "learning_rate": 5.706839936031073e-05, "loss": 0.3044, "step": 5598 }, { "epoch": 0.57, "learning_rate": 5.7067367457955876e-05, "loss": 1.2811, "step": 5599 }, { "epoch": 0.57, "learning_rate": 5.7066335383354585e-05, "loss": 0.2438, "step": 5600 }, { "epoch": 0.57, "learning_rate": 5.706530313651344e-05, "loss": 0.6855, "step": 5601 }, { "epoch": 0.57, "learning_rate": 5.706427071743901e-05, "loss": 0.4839, "step": 5602 }, { "epoch": 0.57, "learning_rate": 5.7063238126137846e-05, "loss": 0.8254, "step": 5603 }, { "epoch": 0.57, "learning_rate": 5.706220536261653e-05, "loss": 1.2558, "step": 5604 }, { "epoch": 0.57, "learning_rate": 5.706117242688164e-05, "loss": 0.6644, "step": 5605 }, { "epoch": 0.57, "learning_rate": 5.706013931893974e-05, "loss": 0.8314, "step": 5606 }, { "epoch": 0.57, "learning_rate": 5.705910603879741e-05, "loss": 1.2764, "step": 5607 }, { "epoch": 0.57, "learning_rate": 5.705807258646122e-05, "loss": 0.3476, "step": 5608 }, { "epoch": 0.57, "learning_rate": 5.705703896193775e-05, "loss": 2.5087, "step": 5609 }, { "epoch": 0.57, "learning_rate": 5.705600516523358e-05, "loss": 0.7595, "step": 5610 }, { "epoch": 0.57, "learning_rate": 5.705497119635528e-05, "loss": 1.338, "step": 5611 }, { "epoch": 0.57, "learning_rate": 5.705393705530944e-05, "loss": 0.8056, "step": 5612 }, { "epoch": 0.57, "learning_rate": 5.705290274210264e-05, "loss": 1.7067, "step": 5613 }, { "epoch": 0.57, "learning_rate": 5.7051868256741454e-05, "loss": 1.5051, "step": 5614 }, { "epoch": 0.57, "learning_rate": 5.705083359923246e-05, "loss": 0.8481, "step": 5615 }, { "epoch": 0.57, "learning_rate": 5.7049798769582265e-05, "loss": 0.2993, "step": 5616 }, { "epoch": 0.57, "learning_rate": 5.7048763767797434e-05, "loss": 0.1732, "step": 5617 }, { "epoch": 0.57, "learning_rate": 5.704772859388456e-05, "loss": 1.4769, "step": 5618 }, { "epoch": 0.57, "learning_rate": 5.704669324785024e-05, "loss": 1.1502, "step": 5619 }, { "epoch": 0.57, "learning_rate": 5.7045657729701046e-05, "loss": 0.2654, "step": 5620 }, { "epoch": 0.57, "learning_rate": 5.704462203944357e-05, "loss": 1.1663, "step": 5621 }, { "epoch": 0.57, "learning_rate": 5.704358617708442e-05, "loss": 0.4602, "step": 5622 }, { "epoch": 0.57, "learning_rate": 5.704255014263017e-05, "loss": 0.4528, "step": 5623 }, { "epoch": 0.57, "learning_rate": 5.704151393608741e-05, "loss": 0.4903, "step": 5624 }, { "epoch": 0.57, "learning_rate": 5.7040477557462755e-05, "loss": 0.4473, "step": 5625 }, { "epoch": 0.57, "learning_rate": 5.703944100676278e-05, "loss": 0.6829, "step": 5626 }, { "epoch": 0.57, "learning_rate": 5.703840428399409e-05, "loss": 1.4403, "step": 5627 }, { "epoch": 0.57, "learning_rate": 5.703736738916328e-05, "loss": 1.0598, "step": 5628 }, { "epoch": 0.57, "learning_rate": 5.703633032227695e-05, "loss": 1.0886, "step": 5629 }, { "epoch": 0.57, "learning_rate": 5.7035293083341706e-05, "loss": 0.2775, "step": 5630 }, { "epoch": 0.57, "learning_rate": 5.7034255672364136e-05, "loss": 1.3862, "step": 5631 }, { "epoch": 0.57, "learning_rate": 5.7033218089350847e-05, "loss": 1.0593, "step": 5632 }, { "epoch": 0.57, "learning_rate": 5.703218033430844e-05, "loss": 0.8662, "step": 5633 }, { "epoch": 0.57, "learning_rate": 5.703114240724352e-05, "loss": 1.0916, "step": 5634 }, { "epoch": 0.57, "learning_rate": 5.70301043081627e-05, "loss": 1.39, "step": 5635 }, { "epoch": 0.57, "learning_rate": 5.702906603707257e-05, "loss": 1.6559, "step": 5636 }, { "epoch": 0.57, "learning_rate": 5.702802759397976e-05, "loss": 0.3985, "step": 5637 }, { "epoch": 0.57, "learning_rate": 5.702698897889086e-05, "loss": 1.2727, "step": 5638 }, { "epoch": 0.57, "learning_rate": 5.702595019181248e-05, "loss": 0.444, "step": 5639 }, { "epoch": 0.57, "learning_rate": 5.7024911232751224e-05, "loss": 0.9676, "step": 5640 }, { "epoch": 0.57, "learning_rate": 5.702387210171373e-05, "loss": 0.879, "step": 5641 }, { "epoch": 0.57, "learning_rate": 5.702283279870658e-05, "loss": 1.03, "step": 5642 }, { "epoch": 0.57, "learning_rate": 5.702179332373641e-05, "loss": 0.3608, "step": 5643 }, { "epoch": 0.57, "learning_rate": 5.7020753676809824e-05, "loss": 1.1861, "step": 5644 }, { "epoch": 0.57, "learning_rate": 5.701971385793345e-05, "loss": 0.6451, "step": 5645 }, { "epoch": 0.57, "learning_rate": 5.7018673867113886e-05, "loss": 1.088, "step": 5646 }, { "epoch": 0.57, "learning_rate": 5.701763370435777e-05, "loss": 1.0112, "step": 5647 }, { "epoch": 0.57, "learning_rate": 5.70165933696717e-05, "loss": 2.0302, "step": 5648 }, { "epoch": 0.57, "learning_rate": 5.7015552863062306e-05, "loss": 1.1629, "step": 5649 }, { "epoch": 0.57, "learning_rate": 5.701451218453622e-05, "loss": 0.3219, "step": 5650 }, { "epoch": 0.57, "learning_rate": 5.701347133410006e-05, "loss": 0.7919, "step": 5651 }, { "epoch": 0.57, "learning_rate": 5.701243031176043e-05, "loss": 0.6249, "step": 5652 }, { "epoch": 0.57, "learning_rate": 5.701138911752398e-05, "loss": 1.0739, "step": 5653 }, { "epoch": 0.57, "learning_rate": 5.701034775139732e-05, "loss": 1.1371, "step": 5654 }, { "epoch": 0.57, "learning_rate": 5.700930621338708e-05, "loss": 1.4479, "step": 5655 }, { "epoch": 0.57, "learning_rate": 5.700826450349989e-05, "loss": 0.8592, "step": 5656 }, { "epoch": 0.57, "learning_rate": 5.700722262174239e-05, "loss": 1.2611, "step": 5657 }, { "epoch": 0.57, "learning_rate": 5.700618056812119e-05, "loss": 0.9291, "step": 5658 }, { "epoch": 0.57, "learning_rate": 5.700513834264293e-05, "loss": 0.9458, "step": 5659 }, { "epoch": 0.57, "learning_rate": 5.700409594531425e-05, "loss": 0.4446, "step": 5660 }, { "epoch": 0.57, "learning_rate": 5.700305337614177e-05, "loss": 1.1317, "step": 5661 }, { "epoch": 0.57, "learning_rate": 5.700201063513213e-05, "loss": 0.7725, "step": 5662 }, { "epoch": 0.58, "learning_rate": 5.700096772229197e-05, "loss": 1.7083, "step": 5663 }, { "epoch": 0.58, "learning_rate": 5.6999924637627926e-05, "loss": 0.8609, "step": 5664 }, { "epoch": 0.58, "learning_rate": 5.699888138114663e-05, "loss": 1.7691, "step": 5665 }, { "epoch": 0.58, "learning_rate": 5.6997837952854724e-05, "loss": 1.5053, "step": 5666 }, { "epoch": 0.58, "learning_rate": 5.699679435275885e-05, "loss": 1.0364, "step": 5667 }, { "epoch": 0.58, "learning_rate": 5.699575058086564e-05, "loss": 1.882, "step": 5668 }, { "epoch": 0.58, "learning_rate": 5.699470663718174e-05, "loss": 0.3045, "step": 5669 }, { "epoch": 0.58, "learning_rate": 5.699366252171381e-05, "loss": 1.7116, "step": 5670 }, { "epoch": 0.58, "learning_rate": 5.6992618234468474e-05, "loss": 0.9153, "step": 5671 }, { "epoch": 0.58, "learning_rate": 5.6991573775452386e-05, "loss": 1.7081, "step": 5672 }, { "epoch": 0.58, "learning_rate": 5.6990529144672186e-05, "loss": 0.5214, "step": 5673 }, { "epoch": 0.58, "learning_rate": 5.698948434213453e-05, "loss": 0.627, "step": 5674 }, { "epoch": 0.58, "learning_rate": 5.698843936784606e-05, "loss": 0.4578, "step": 5675 }, { "epoch": 0.58, "learning_rate": 5.6987394221813425e-05, "loss": 0.2324, "step": 5676 }, { "epoch": 0.58, "learning_rate": 5.6986348904043284e-05, "loss": 1.9524, "step": 5677 }, { "epoch": 0.58, "learning_rate": 5.6985303414542295e-05, "loss": 1.2751, "step": 5678 }, { "epoch": 0.58, "learning_rate": 5.698425775331709e-05, "loss": 0.3034, "step": 5679 }, { "epoch": 0.58, "learning_rate": 5.6983211920374335e-05, "loss": 1.7728, "step": 5680 }, { "epoch": 0.58, "learning_rate": 5.698216591572068e-05, "loss": 0.644, "step": 5681 }, { "epoch": 0.58, "learning_rate": 5.6981119739362796e-05, "loss": 0.4912, "step": 5682 }, { "epoch": 0.58, "learning_rate": 5.698007339130733e-05, "loss": 0.7877, "step": 5683 }, { "epoch": 0.58, "learning_rate": 5.697902687156093e-05, "loss": 1.071, "step": 5684 }, { "epoch": 0.58, "learning_rate": 5.697798018013027e-05, "loss": 0.6309, "step": 5685 }, { "epoch": 0.58, "learning_rate": 5.6976933317022006e-05, "loss": 1.1134, "step": 5686 }, { "epoch": 0.58, "learning_rate": 5.69758862822428e-05, "loss": 1.2, "step": 5687 }, { "epoch": 0.58, "learning_rate": 5.697483907579932e-05, "loss": 0.8697, "step": 5688 }, { "epoch": 0.58, "learning_rate": 5.697379169769823e-05, "loss": 0.6722, "step": 5689 }, { "epoch": 0.58, "learning_rate": 5.697274414794619e-05, "loss": 0.739, "step": 5690 }, { "epoch": 0.58, "learning_rate": 5.6971696426549866e-05, "loss": 1.3874, "step": 5691 }, { "epoch": 0.58, "learning_rate": 5.697064853351592e-05, "loss": 0.4477, "step": 5692 }, { "epoch": 0.58, "learning_rate": 5.696960046885102e-05, "loss": 0.703, "step": 5693 }, { "epoch": 0.58, "learning_rate": 5.696855223256186e-05, "loss": 1.4703, "step": 5694 }, { "epoch": 0.58, "learning_rate": 5.696750382465508e-05, "loss": 0.3912, "step": 5695 }, { "epoch": 0.58, "learning_rate": 5.6966455245137374e-05, "loss": 0.8088, "step": 5696 }, { "epoch": 0.58, "learning_rate": 5.6965406494015403e-05, "loss": 0.5255, "step": 5697 }, { "epoch": 0.58, "learning_rate": 5.696435757129584e-05, "loss": 0.6061, "step": 5698 }, { "epoch": 0.58, "learning_rate": 5.696330847698537e-05, "loss": 1.155, "step": 5699 }, { "epoch": 0.58, "learning_rate": 5.696225921109066e-05, "loss": 0.804, "step": 5700 }, { "epoch": 0.58, "learning_rate": 5.696120977361838e-05, "loss": 1.4233, "step": 5701 }, { "epoch": 0.58, "learning_rate": 5.696016016457523e-05, "loss": 1.39, "step": 5702 }, { "epoch": 0.58, "learning_rate": 5.695911038396787e-05, "loss": 0.9486, "step": 5703 }, { "epoch": 0.58, "learning_rate": 5.6958060431802994e-05, "loss": 0.3415, "step": 5704 }, { "epoch": 0.58, "learning_rate": 5.695701030808727e-05, "loss": 0.5608, "step": 5705 }, { "epoch": 0.58, "learning_rate": 5.695596001282739e-05, "loss": 0.4752, "step": 5706 }, { "epoch": 0.58, "learning_rate": 5.695490954603003e-05, "loss": 1.4026, "step": 5707 }, { "epoch": 0.58, "learning_rate": 5.6953858907701894e-05, "loss": 1.575, "step": 5708 }, { "epoch": 0.58, "learning_rate": 5.695280809784964e-05, "loss": 1.9203, "step": 5709 }, { "epoch": 0.58, "learning_rate": 5.695175711647997e-05, "loss": 0.7779, "step": 5710 }, { "epoch": 0.58, "learning_rate": 5.695070596359957e-05, "loss": 0.4403, "step": 5711 }, { "epoch": 0.58, "learning_rate": 5.6949654639215144e-05, "loss": 0.6952, "step": 5712 }, { "epoch": 0.58, "learning_rate": 5.694860314333335e-05, "loss": 0.7791, "step": 5713 }, { "epoch": 0.58, "learning_rate": 5.69475514759609e-05, "loss": 0.908, "step": 5714 }, { "epoch": 0.58, "learning_rate": 5.6946499637104496e-05, "loss": 0.3795, "step": 5715 }, { "epoch": 0.58, "learning_rate": 5.6945447626770805e-05, "loss": 1.0961, "step": 5716 }, { "epoch": 0.58, "learning_rate": 5.694439544496654e-05, "loss": 0.4447, "step": 5717 }, { "epoch": 0.58, "learning_rate": 5.6943343091698396e-05, "loss": 0.8668, "step": 5718 }, { "epoch": 0.58, "learning_rate": 5.694229056697307e-05, "loss": 0.7153, "step": 5719 }, { "epoch": 0.58, "learning_rate": 5.694123787079725e-05, "loss": 1.6123, "step": 5720 }, { "epoch": 0.58, "learning_rate": 5.6940185003177634e-05, "loss": 0.9787, "step": 5721 }, { "epoch": 0.58, "learning_rate": 5.6939131964120944e-05, "loss": 1.0661, "step": 5722 }, { "epoch": 0.58, "learning_rate": 5.693807875363386e-05, "loss": 0.9264, "step": 5723 }, { "epoch": 0.58, "learning_rate": 5.6937025371723095e-05, "loss": 0.9515, "step": 5724 }, { "epoch": 0.58, "learning_rate": 5.6935971818395345e-05, "loss": 1.6019, "step": 5725 }, { "epoch": 0.58, "learning_rate": 5.693491809365731e-05, "loss": 0.3626, "step": 5726 }, { "epoch": 0.58, "learning_rate": 5.6933864197515705e-05, "loss": 0.8399, "step": 5727 }, { "epoch": 0.58, "learning_rate": 5.6932810129977245e-05, "loss": 1.1431, "step": 5728 }, { "epoch": 0.58, "learning_rate": 5.693175589104862e-05, "loss": 0.8841, "step": 5729 }, { "epoch": 0.58, "learning_rate": 5.693070148073655e-05, "loss": 1.7796, "step": 5730 }, { "epoch": 0.58, "learning_rate": 5.6929646899047726e-05, "loss": 1.4296, "step": 5731 }, { "epoch": 0.58, "learning_rate": 5.692859214598889e-05, "loss": 1.7357, "step": 5732 }, { "epoch": 0.58, "learning_rate": 5.692753722156673e-05, "loss": 1.665, "step": 5733 }, { "epoch": 0.58, "learning_rate": 5.6926482125787966e-05, "loss": 1.3204, "step": 5734 }, { "epoch": 0.58, "learning_rate": 5.692542685865932e-05, "loss": 1.8925, "step": 5735 }, { "epoch": 0.58, "learning_rate": 5.692437142018749e-05, "loss": 0.5917, "step": 5736 }, { "epoch": 0.58, "learning_rate": 5.6923315810379215e-05, "loss": 0.5545, "step": 5737 }, { "epoch": 0.58, "learning_rate": 5.69222600292412e-05, "loss": 0.4938, "step": 5738 }, { "epoch": 0.58, "learning_rate": 5.692120407678015e-05, "loss": 1.0391, "step": 5739 }, { "epoch": 0.58, "learning_rate": 5.692014795300281e-05, "loss": 0.8713, "step": 5740 }, { "epoch": 0.58, "learning_rate": 5.6919091657915885e-05, "loss": 1.2651, "step": 5741 }, { "epoch": 0.58, "learning_rate": 5.69180351915261e-05, "loss": 1.4765, "step": 5742 }, { "epoch": 0.58, "learning_rate": 5.6916978553840194e-05, "loss": 0.7038, "step": 5743 }, { "epoch": 0.58, "learning_rate": 5.6915921744864855e-05, "loss": 0.4008, "step": 5744 }, { "epoch": 0.58, "learning_rate": 5.691486476460684e-05, "loss": 1.1486, "step": 5745 }, { "epoch": 0.58, "learning_rate": 5.691380761307287e-05, "loss": 1.2095, "step": 5746 }, { "epoch": 0.58, "learning_rate": 5.691275029026966e-05, "loss": 1.1781, "step": 5747 }, { "epoch": 0.58, "learning_rate": 5.691169279620394e-05, "loss": 0.6709, "step": 5748 }, { "epoch": 0.58, "learning_rate": 5.6910635130882446e-05, "loss": 0.6966, "step": 5749 }, { "epoch": 0.58, "learning_rate": 5.6909577294311916e-05, "loss": 1.0951, "step": 5750 }, { "epoch": 0.58, "learning_rate": 5.690851928649907e-05, "loss": 0.8828, "step": 5751 }, { "epoch": 0.58, "learning_rate": 5.690746110745064e-05, "loss": 0.2439, "step": 5752 }, { "epoch": 0.58, "learning_rate": 5.690640275717337e-05, "loss": 0.3139, "step": 5753 }, { "epoch": 0.58, "learning_rate": 5.690534423567398e-05, "loss": 0.2369, "step": 5754 }, { "epoch": 0.58, "learning_rate": 5.690428554295922e-05, "loss": 0.3923, "step": 5755 }, { "epoch": 0.58, "learning_rate": 5.6903226679035826e-05, "loss": 0.6166, "step": 5756 }, { "epoch": 0.58, "learning_rate": 5.6902167643910525e-05, "loss": 1.3613, "step": 5757 }, { "epoch": 0.58, "learning_rate": 5.6901108437590067e-05, "loss": 0.4435, "step": 5758 }, { "epoch": 0.58, "learning_rate": 5.690004906008118e-05, "loss": 1.3317, "step": 5759 }, { "epoch": 0.58, "learning_rate": 5.689898951139062e-05, "loss": 1.5087, "step": 5760 }, { "epoch": 0.58, "learning_rate": 5.6897929791525125e-05, "loss": 1.2838, "step": 5761 }, { "epoch": 0.59, "learning_rate": 5.6896869900491435e-05, "loss": 0.1656, "step": 5762 }, { "epoch": 0.59, "learning_rate": 5.689580983829631e-05, "loss": 1.1415, "step": 5763 }, { "epoch": 0.59, "learning_rate": 5.689474960494647e-05, "loss": 0.6956, "step": 5764 }, { "epoch": 0.59, "learning_rate": 5.689368920044868e-05, "loss": 0.7441, "step": 5765 }, { "epoch": 0.59, "learning_rate": 5.689262862480967e-05, "loss": 1.0611, "step": 5766 }, { "epoch": 0.59, "learning_rate": 5.6891567878036216e-05, "loss": 1.0622, "step": 5767 }, { "epoch": 0.59, "learning_rate": 5.6890506960135044e-05, "loss": 0.2891, "step": 5768 }, { "epoch": 0.59, "learning_rate": 5.688944587111292e-05, "loss": 0.815, "step": 5769 }, { "epoch": 0.59, "learning_rate": 5.68883846109766e-05, "loss": 0.2338, "step": 5770 }, { "epoch": 0.59, "learning_rate": 5.688732317973282e-05, "loss": 0.9837, "step": 5771 }, { "epoch": 0.59, "learning_rate": 5.688626157738834e-05, "loss": 0.9247, "step": 5772 }, { "epoch": 0.59, "learning_rate": 5.6885199803949924e-05, "loss": 0.4493, "step": 5773 }, { "epoch": 0.59, "learning_rate": 5.688413785942433e-05, "loss": 1.1066, "step": 5774 }, { "epoch": 0.59, "learning_rate": 5.6883075743818306e-05, "loss": 0.8022, "step": 5775 }, { "epoch": 0.59, "learning_rate": 5.688201345713861e-05, "loss": 0.6386, "step": 5776 }, { "epoch": 0.59, "learning_rate": 5.6880950999392006e-05, "loss": 0.644, "step": 5777 }, { "epoch": 0.59, "learning_rate": 5.687988837058526e-05, "loss": 0.8315, "step": 5778 }, { "epoch": 0.59, "learning_rate": 5.687882557072513e-05, "loss": 0.384, "step": 5779 }, { "epoch": 0.59, "learning_rate": 5.687776259981838e-05, "loss": 0.6321, "step": 5780 }, { "epoch": 0.59, "learning_rate": 5.687669945787177e-05, "loss": 0.4365, "step": 5781 }, { "epoch": 0.59, "learning_rate": 5.6875636144892073e-05, "loss": 0.9721, "step": 5782 }, { "epoch": 0.59, "learning_rate": 5.6874572660886055e-05, "loss": 0.838, "step": 5783 }, { "epoch": 0.59, "learning_rate": 5.687350900586048e-05, "loss": 0.7815, "step": 5784 }, { "epoch": 0.59, "learning_rate": 5.687244517982211e-05, "loss": 0.9233, "step": 5785 }, { "epoch": 0.59, "learning_rate": 5.687138118277772e-05, "loss": 0.7479, "step": 5786 }, { "epoch": 0.59, "learning_rate": 5.687031701473409e-05, "loss": 0.3083, "step": 5787 }, { "epoch": 0.59, "learning_rate": 5.6869252675697985e-05, "loss": 0.3645, "step": 5788 }, { "epoch": 0.59, "learning_rate": 5.6868188165676174e-05, "loss": 0.7661, "step": 5789 }, { "epoch": 0.59, "learning_rate": 5.686712348467543e-05, "loss": 0.9, "step": 5790 }, { "epoch": 0.59, "learning_rate": 5.686605863270254e-05, "loss": 1.001, "step": 5791 }, { "epoch": 0.59, "learning_rate": 5.686499360976427e-05, "loss": 0.2662, "step": 5792 }, { "epoch": 0.59, "learning_rate": 5.6863928415867406e-05, "loss": 0.5987, "step": 5793 }, { "epoch": 0.59, "learning_rate": 5.686286305101872e-05, "loss": 1.5484, "step": 5794 }, { "epoch": 0.59, "learning_rate": 5.6861797515224985e-05, "loss": 1.2344, "step": 5795 }, { "epoch": 0.59, "learning_rate": 5.686073180849299e-05, "loss": 0.8579, "step": 5796 }, { "epoch": 0.59, "learning_rate": 5.685966593082952e-05, "loss": 0.6234, "step": 5797 }, { "epoch": 0.59, "learning_rate": 5.685859988224136e-05, "loss": 0.4626, "step": 5798 }, { "epoch": 0.59, "learning_rate": 5.6857533662735274e-05, "loss": 1.5989, "step": 5799 }, { "epoch": 0.59, "learning_rate": 5.6856467272318074e-05, "loss": 1.7775, "step": 5800 }, { "epoch": 0.59, "learning_rate": 5.6855400710996526e-05, "loss": 0.5077, "step": 5801 }, { "epoch": 0.59, "learning_rate": 5.685433397877742e-05, "loss": 0.9288, "step": 5802 }, { "epoch": 0.59, "learning_rate": 5.6853267075667564e-05, "loss": 0.7881, "step": 5803 }, { "epoch": 0.59, "learning_rate": 5.6852200001673714e-05, "loss": 0.3545, "step": 5804 }, { "epoch": 0.59, "learning_rate": 5.685113275680269e-05, "loss": 0.5752, "step": 5805 }, { "epoch": 0.59, "learning_rate": 5.685006534106127e-05, "loss": 1.136, "step": 5806 }, { "epoch": 0.59, "learning_rate": 5.684899775445625e-05, "loss": 0.8839, "step": 5807 }, { "epoch": 0.59, "learning_rate": 5.684792999699442e-05, "loss": 1.4557, "step": 5808 }, { "epoch": 0.59, "learning_rate": 5.684686206868258e-05, "loss": 1.1723, "step": 5809 }, { "epoch": 0.59, "learning_rate": 5.684579396952751e-05, "loss": 0.7178, "step": 5810 }, { "epoch": 0.59, "learning_rate": 5.684472569953603e-05, "loss": 0.6496, "step": 5811 }, { "epoch": 0.59, "learning_rate": 5.6843657258714926e-05, "loss": 1.1902, "step": 5812 }, { "epoch": 0.59, "learning_rate": 5.6842588647071e-05, "loss": 0.291, "step": 5813 }, { "epoch": 0.59, "learning_rate": 5.684151986461105e-05, "loss": 0.4515, "step": 5814 }, { "epoch": 0.59, "learning_rate": 5.684045091134188e-05, "loss": 1.1356, "step": 5815 }, { "epoch": 0.59, "learning_rate": 5.6839381787270285e-05, "loss": 1.0267, "step": 5816 }, { "epoch": 0.59, "learning_rate": 5.683831249240308e-05, "loss": 0.7825, "step": 5817 }, { "epoch": 0.59, "learning_rate": 5.683724302674707e-05, "loss": 0.48, "step": 5818 }, { "epoch": 0.59, "learning_rate": 5.6836173390309044e-05, "loss": 0.9207, "step": 5819 }, { "epoch": 0.59, "learning_rate": 5.683510358309583e-05, "loss": 1.2187, "step": 5820 }, { "epoch": 0.59, "learning_rate": 5.6834033605114214e-05, "loss": 1.026, "step": 5821 }, { "epoch": 0.59, "learning_rate": 5.6832963456371027e-05, "loss": 1.2459, "step": 5822 }, { "epoch": 0.59, "learning_rate": 5.6831893136873065e-05, "loss": 1.2534, "step": 5823 }, { "epoch": 0.59, "learning_rate": 5.6830822646627146e-05, "loss": 0.6048, "step": 5824 }, { "epoch": 0.59, "learning_rate": 5.6829751985640065e-05, "loss": 1.0512, "step": 5825 }, { "epoch": 0.59, "learning_rate": 5.682868115391866e-05, "loss": 0.9363, "step": 5826 }, { "epoch": 0.59, "learning_rate": 5.682761015146973e-05, "loss": 0.4894, "step": 5827 }, { "epoch": 0.59, "learning_rate": 5.68265389783001e-05, "loss": 0.5617, "step": 5828 }, { "epoch": 0.59, "learning_rate": 5.6825467634416574e-05, "loss": 0.6398, "step": 5829 }, { "epoch": 0.59, "learning_rate": 5.6824396119825983e-05, "loss": 1.3741, "step": 5830 }, { "epoch": 0.59, "learning_rate": 5.682332443453514e-05, "loss": 0.5512, "step": 5831 }, { "epoch": 0.59, "learning_rate": 5.682225257855086e-05, "loss": 0.8193, "step": 5832 }, { "epoch": 0.59, "learning_rate": 5.682118055187996e-05, "loss": 1.4753, "step": 5833 }, { "epoch": 0.59, "learning_rate": 5.682010835452929e-05, "loss": 0.709, "step": 5834 }, { "epoch": 0.59, "learning_rate": 5.681903598650564e-05, "loss": 1.5394, "step": 5835 }, { "epoch": 0.59, "learning_rate": 5.681796344781586e-05, "loss": 0.6146, "step": 5836 }, { "epoch": 0.59, "learning_rate": 5.681689073846675e-05, "loss": 0.3132, "step": 5837 }, { "epoch": 0.59, "learning_rate": 5.681581785846515e-05, "loss": 0.8819, "step": 5838 }, { "epoch": 0.59, "learning_rate": 5.6814744807817894e-05, "loss": 0.9898, "step": 5839 }, { "epoch": 0.59, "learning_rate": 5.68136715865318e-05, "loss": 0.3892, "step": 5840 }, { "epoch": 0.59, "learning_rate": 5.681259819461369e-05, "loss": 0.5958, "step": 5841 }, { "epoch": 0.59, "learning_rate": 5.681152463207042e-05, "loss": 0.4382, "step": 5842 }, { "epoch": 0.59, "learning_rate": 5.681045089890881e-05, "loss": 0.4666, "step": 5843 }, { "epoch": 0.59, "learning_rate": 5.680937699513567e-05, "loss": 0.8457, "step": 5844 }, { "epoch": 0.59, "learning_rate": 5.680830292075786e-05, "loss": 0.4549, "step": 5845 }, { "epoch": 0.59, "learning_rate": 5.6807228675782215e-05, "loss": 0.8345, "step": 5846 }, { "epoch": 0.59, "learning_rate": 5.680615426021556e-05, "loss": 0.7075, "step": 5847 }, { "epoch": 0.59, "learning_rate": 5.6805079674064745e-05, "loss": 1.0557, "step": 5848 }, { "epoch": 0.59, "learning_rate": 5.680400491733659e-05, "loss": 0.9292, "step": 5849 }, { "epoch": 0.59, "learning_rate": 5.6802929990037944e-05, "loss": 1.3479, "step": 5850 }, { "epoch": 0.59, "learning_rate": 5.680185489217566e-05, "loss": 0.9998, "step": 5851 }, { "epoch": 0.59, "learning_rate": 5.680077962375656e-05, "loss": 0.6032, "step": 5852 }, { "epoch": 0.59, "learning_rate": 5.67997041847875e-05, "loss": 0.4986, "step": 5853 }, { "epoch": 0.59, "learning_rate": 5.6798628575275306e-05, "loss": 0.661, "step": 5854 }, { "epoch": 0.59, "learning_rate": 5.6797552795226835e-05, "loss": 0.4311, "step": 5855 }, { "epoch": 0.59, "learning_rate": 5.679647684464894e-05, "loss": 1.4179, "step": 5856 }, { "epoch": 0.59, "learning_rate": 5.679540072354846e-05, "loss": 1.5904, "step": 5857 }, { "epoch": 0.59, "learning_rate": 5.679432443193224e-05, "loss": 0.5479, "step": 5858 }, { "epoch": 0.59, "learning_rate": 5.679324796980713e-05, "loss": 0.5094, "step": 5859 }, { "epoch": 0.6, "learning_rate": 5.6792171337179986e-05, "loss": 0.9289, "step": 5860 }, { "epoch": 0.6, "learning_rate": 5.6791094534057664e-05, "loss": 0.7946, "step": 5861 }, { "epoch": 0.6, "learning_rate": 5.6790017560446996e-05, "loss": 0.9128, "step": 5862 }, { "epoch": 0.6, "learning_rate": 5.678894041635485e-05, "loss": 1.7565, "step": 5863 }, { "epoch": 0.6, "learning_rate": 5.678786310178808e-05, "loss": 0.5971, "step": 5864 }, { "epoch": 0.6, "learning_rate": 5.678678561675354e-05, "loss": 1.3655, "step": 5865 }, { "epoch": 0.6, "learning_rate": 5.6785707961258085e-05, "loss": 1.1436, "step": 5866 }, { "epoch": 0.6, "learning_rate": 5.678463013530858e-05, "loss": 0.4513, "step": 5867 }, { "epoch": 0.6, "learning_rate": 5.678355213891188e-05, "loss": 0.2947, "step": 5868 }, { "epoch": 0.6, "learning_rate": 5.678247397207483e-05, "loss": 0.873, "step": 5869 }, { "epoch": 0.6, "learning_rate": 5.678139563480431e-05, "loss": 1.5916, "step": 5870 }, { "epoch": 0.6, "learning_rate": 5.678031712710718e-05, "loss": 1.3721, "step": 5871 }, { "epoch": 0.6, "learning_rate": 5.6779238448990296e-05, "loss": 1.0304, "step": 5872 }, { "epoch": 0.6, "learning_rate": 5.6778159600460524e-05, "loss": 0.8263, "step": 5873 }, { "epoch": 0.6, "learning_rate": 5.677708058152474e-05, "loss": 0.7089, "step": 5874 }, { "epoch": 0.6, "learning_rate": 5.6776001392189796e-05, "loss": 0.3621, "step": 5875 }, { "epoch": 0.6, "learning_rate": 5.677492203246256e-05, "loss": 1.3328, "step": 5876 }, { "epoch": 0.6, "learning_rate": 5.6773842502349915e-05, "loss": 0.3279, "step": 5877 }, { "epoch": 0.6, "learning_rate": 5.677276280185872e-05, "loss": 1.1275, "step": 5878 }, { "epoch": 0.6, "learning_rate": 5.677168293099585e-05, "loss": 0.2921, "step": 5879 }, { "epoch": 0.6, "learning_rate": 5.677060288976816e-05, "loss": 0.7068, "step": 5880 }, { "epoch": 0.6, "learning_rate": 5.676952267818255e-05, "loss": 0.6564, "step": 5881 }, { "epoch": 0.6, "learning_rate": 5.676844229624588e-05, "loss": 1.0692, "step": 5882 }, { "epoch": 0.6, "learning_rate": 5.6767361743965024e-05, "loss": 1.3113, "step": 5883 }, { "epoch": 0.6, "learning_rate": 5.676628102134686e-05, "loss": 1.1011, "step": 5884 }, { "epoch": 0.6, "learning_rate": 5.676520012839827e-05, "loss": 0.469, "step": 5885 }, { "epoch": 0.6, "learning_rate": 5.676411906512613e-05, "loss": 0.3714, "step": 5886 }, { "epoch": 0.6, "learning_rate": 5.6763037831537306e-05, "loss": 1.4667, "step": 5887 }, { "epoch": 0.6, "learning_rate": 5.67619564276387e-05, "loss": 0.2494, "step": 5888 }, { "epoch": 0.6, "learning_rate": 5.676087485343718e-05, "loss": 0.9756, "step": 5889 }, { "epoch": 0.6, "learning_rate": 5.675979310893964e-05, "loss": 0.7266, "step": 5890 }, { "epoch": 0.6, "learning_rate": 5.675871119415295e-05, "loss": 1.5261, "step": 5891 }, { "epoch": 0.6, "learning_rate": 5.6757629109084e-05, "loss": 1.4315, "step": 5892 }, { "epoch": 0.6, "learning_rate": 5.6756546853739677e-05, "loss": 0.2147, "step": 5893 }, { "epoch": 0.6, "learning_rate": 5.675546442812688e-05, "loss": 1.8262, "step": 5894 }, { "epoch": 0.6, "learning_rate": 5.675438183225247e-05, "loss": 0.513, "step": 5895 }, { "epoch": 0.6, "learning_rate": 5.6753299066123364e-05, "loss": 0.7795, "step": 5896 }, { "epoch": 0.6, "learning_rate": 5.675221612974643e-05, "loss": 0.2262, "step": 5897 }, { "epoch": 0.6, "learning_rate": 5.6751133023128577e-05, "loss": 0.7952, "step": 5898 }, { "epoch": 0.6, "learning_rate": 5.675004974627669e-05, "loss": 0.5958, "step": 5899 }, { "epoch": 0.6, "learning_rate": 5.674896629919766e-05, "loss": 0.5676, "step": 5900 }, { "epoch": 0.6, "learning_rate": 5.674788268189839e-05, "loss": 0.6505, "step": 5901 }, { "epoch": 0.6, "learning_rate": 5.6746798894385767e-05, "loss": 0.2099, "step": 5902 }, { "epoch": 0.6, "learning_rate": 5.674571493666669e-05, "loss": 0.4556, "step": 5903 }, { "epoch": 0.6, "learning_rate": 5.674463080874805e-05, "loss": 0.4131, "step": 5904 }, { "epoch": 0.6, "learning_rate": 5.674354651063677e-05, "loss": 0.5844, "step": 5905 }, { "epoch": 0.6, "learning_rate": 5.674246204233972e-05, "loss": 0.514, "step": 5906 }, { "epoch": 0.6, "learning_rate": 5.674137740386382e-05, "loss": 1.6101, "step": 5907 }, { "epoch": 0.6, "learning_rate": 5.674029259521596e-05, "loss": 0.8004, "step": 5908 }, { "epoch": 0.6, "learning_rate": 5.673920761640307e-05, "loss": 0.4193, "step": 5909 }, { "epoch": 0.6, "learning_rate": 5.673812246743201e-05, "loss": 0.4904, "step": 5910 }, { "epoch": 0.6, "learning_rate": 5.6737037148309726e-05, "loss": 1.4728, "step": 5911 }, { "epoch": 0.6, "learning_rate": 5.6735951659043105e-05, "loss": 0.9755, "step": 5912 }, { "epoch": 0.6, "learning_rate": 5.6734865999639054e-05, "loss": 0.3369, "step": 5913 }, { "epoch": 0.6, "learning_rate": 5.673378017010449e-05, "loss": 0.2127, "step": 5914 }, { "epoch": 0.6, "learning_rate": 5.673269417044633e-05, "loss": 0.8488, "step": 5915 }, { "epoch": 0.6, "learning_rate": 5.673160800067146e-05, "loss": 0.5997, "step": 5916 }, { "epoch": 0.6, "learning_rate": 5.673052166078681e-05, "loss": 1.2389, "step": 5917 }, { "epoch": 0.6, "learning_rate": 5.672943515079928e-05, "loss": 0.7618, "step": 5918 }, { "epoch": 0.6, "learning_rate": 5.672834847071581e-05, "loss": 1.2743, "step": 5919 }, { "epoch": 0.6, "learning_rate": 5.672726162054329e-05, "loss": 0.562, "step": 5920 }, { "epoch": 0.6, "learning_rate": 5.6726174600288646e-05, "loss": 1.375, "step": 5921 }, { "epoch": 0.6, "learning_rate": 5.6725087409958784e-05, "loss": 1.0981, "step": 5922 }, { "epoch": 0.6, "learning_rate": 5.6724000049560644e-05, "loss": 0.9115, "step": 5923 }, { "epoch": 0.6, "learning_rate": 5.672291251910114e-05, "loss": 0.5471, "step": 5924 }, { "epoch": 0.6, "learning_rate": 5.672182481858717e-05, "loss": 0.7044, "step": 5925 }, { "epoch": 0.6, "learning_rate": 5.672073694802568e-05, "loss": 0.3439, "step": 5926 }, { "epoch": 0.6, "learning_rate": 5.671964890742359e-05, "loss": 1.3953, "step": 5927 }, { "epoch": 0.6, "learning_rate": 5.6718560696787815e-05, "loss": 0.684, "step": 5928 }, { "epoch": 0.6, "learning_rate": 5.671747231612528e-05, "loss": 0.9548, "step": 5929 }, { "epoch": 0.6, "learning_rate": 5.6716383765442924e-05, "loss": 1.0242, "step": 5930 }, { "epoch": 0.6, "learning_rate": 5.6715295044747655e-05, "loss": 0.502, "step": 5931 }, { "epoch": 0.6, "learning_rate": 5.671420615404641e-05, "loss": 1.6149, "step": 5932 }, { "epoch": 0.6, "learning_rate": 5.671311709334613e-05, "loss": 1.2437, "step": 5933 }, { "epoch": 0.6, "learning_rate": 5.6712027862653735e-05, "loss": 0.4356, "step": 5934 }, { "epoch": 0.6, "learning_rate": 5.671093846197615e-05, "loss": 2.0876, "step": 5935 }, { "epoch": 0.6, "learning_rate": 5.670984889132032e-05, "loss": 1.1936, "step": 5936 }, { "epoch": 0.6, "learning_rate": 5.670875915069317e-05, "loss": 0.5022, "step": 5937 }, { "epoch": 0.6, "learning_rate": 5.670766924010163e-05, "loss": 1.8642, "step": 5938 }, { "epoch": 0.6, "learning_rate": 5.670657915955265e-05, "loss": 1.2173, "step": 5939 }, { "epoch": 0.6, "learning_rate": 5.6705488909053155e-05, "loss": 0.8038, "step": 5940 }, { "epoch": 0.6, "learning_rate": 5.670439848861009e-05, "loss": 0.4372, "step": 5941 }, { "epoch": 0.6, "learning_rate": 5.67033078982304e-05, "loss": 0.9632, "step": 5942 }, { "epoch": 0.6, "learning_rate": 5.6702217137921007e-05, "loss": 1.2591, "step": 5943 }, { "epoch": 0.6, "learning_rate": 5.670112620768886e-05, "loss": 1.1659, "step": 5944 }, { "epoch": 0.6, "learning_rate": 5.670003510754092e-05, "loss": 1.3879, "step": 5945 }, { "epoch": 0.6, "learning_rate": 5.669894383748409e-05, "loss": 1.1379, "step": 5946 }, { "epoch": 0.6, "learning_rate": 5.669785239752536e-05, "loss": 1.1201, "step": 5947 }, { "epoch": 0.6, "learning_rate": 5.669676078767164e-05, "loss": 0.4084, "step": 5948 }, { "epoch": 0.6, "learning_rate": 5.66956690079299e-05, "loss": 1.2045, "step": 5949 }, { "epoch": 0.6, "learning_rate": 5.669457705830707e-05, "loss": 0.6868, "step": 5950 }, { "epoch": 0.6, "learning_rate": 5.669348493881011e-05, "loss": 0.4063, "step": 5951 }, { "epoch": 0.6, "learning_rate": 5.669239264944597e-05, "loss": 0.719, "step": 5952 }, { "epoch": 0.6, "learning_rate": 5.6691300190221595e-05, "loss": 0.4751, "step": 5953 }, { "epoch": 0.6, "learning_rate": 5.669020756114393e-05, "loss": 1.4838, "step": 5954 }, { "epoch": 0.6, "learning_rate": 5.6689114762219956e-05, "loss": 2.4898, "step": 5955 }, { "epoch": 0.6, "learning_rate": 5.668802179345659e-05, "loss": 0.9804, "step": 5956 }, { "epoch": 0.6, "learning_rate": 5.6686928654860813e-05, "loss": 0.4448, "step": 5957 }, { "epoch": 0.6, "learning_rate": 5.668583534643958e-05, "loss": 1.0114, "step": 5958 }, { "epoch": 0.61, "learning_rate": 5.668474186819984e-05, "loss": 1.5527, "step": 5959 }, { "epoch": 0.61, "learning_rate": 5.668364822014856e-05, "loss": 0.5925, "step": 5960 }, { "epoch": 0.61, "learning_rate": 5.668255440229268e-05, "loss": 0.8547, "step": 5961 }, { "epoch": 0.61, "learning_rate": 5.668146041463918e-05, "loss": 0.5173, "step": 5962 }, { "epoch": 0.61, "learning_rate": 5.668036625719502e-05, "loss": 1.2925, "step": 5963 }, { "epoch": 0.61, "learning_rate": 5.6679271929967155e-05, "loss": 1.1241, "step": 5964 }, { "epoch": 0.61, "learning_rate": 5.667817743296256e-05, "loss": 1.6034, "step": 5965 }, { "epoch": 0.61, "learning_rate": 5.6677082766188176e-05, "loss": 0.7199, "step": 5966 }, { "epoch": 0.61, "learning_rate": 5.667598792965099e-05, "loss": 1.1353, "step": 5967 }, { "epoch": 0.61, "learning_rate": 5.667489292335798e-05, "loss": 0.9507, "step": 5968 }, { "epoch": 0.61, "learning_rate": 5.6673797747316086e-05, "loss": 0.666, "step": 5969 }, { "epoch": 0.61, "learning_rate": 5.66727024015323e-05, "loss": 0.9618, "step": 5970 }, { "epoch": 0.61, "learning_rate": 5.667160688601358e-05, "loss": 0.1339, "step": 5971 }, { "epoch": 0.61, "learning_rate": 5.6670511200766896e-05, "loss": 1.4209, "step": 5972 }, { "epoch": 0.61, "learning_rate": 5.666941534579922e-05, "loss": 0.4786, "step": 5973 }, { "epoch": 0.61, "learning_rate": 5.6668319321117535e-05, "loss": 0.8637, "step": 5974 }, { "epoch": 0.61, "learning_rate": 5.6667223126728816e-05, "loss": 0.9312, "step": 5975 }, { "epoch": 0.61, "learning_rate": 5.666612676264003e-05, "loss": 0.9957, "step": 5976 }, { "epoch": 0.61, "learning_rate": 5.6665030228858154e-05, "loss": 0.773, "step": 5977 }, { "epoch": 0.61, "learning_rate": 5.666393352539018e-05, "loss": 1.6589, "step": 5978 }, { "epoch": 0.61, "learning_rate": 5.666283665224307e-05, "loss": 0.8296, "step": 5979 }, { "epoch": 0.61, "learning_rate": 5.666173960942381e-05, "loss": 0.6297, "step": 5980 }, { "epoch": 0.61, "learning_rate": 5.666064239693938e-05, "loss": 1.2658, "step": 5981 }, { "epoch": 0.61, "learning_rate": 5.665954501479676e-05, "loss": 1.0322, "step": 5982 }, { "epoch": 0.61, "learning_rate": 5.665844746300295e-05, "loss": 1.6497, "step": 5983 }, { "epoch": 0.61, "learning_rate": 5.6657349741564915e-05, "loss": 1.1513, "step": 5984 }, { "epoch": 0.61, "learning_rate": 5.665625185048965e-05, "loss": 0.5982, "step": 5985 }, { "epoch": 0.61, "learning_rate": 5.665515378978413e-05, "loss": 1.5975, "step": 5986 }, { "epoch": 0.61, "learning_rate": 5.665405555945536e-05, "loss": 0.3873, "step": 5987 }, { "epoch": 0.61, "learning_rate": 5.6652957159510314e-05, "loss": 0.4571, "step": 5988 }, { "epoch": 0.61, "learning_rate": 5.6651858589956e-05, "loss": 1.2526, "step": 5989 }, { "epoch": 0.61, "learning_rate": 5.665075985079938e-05, "loss": 0.2897, "step": 5990 }, { "epoch": 0.61, "learning_rate": 5.664966094204747e-05, "loss": 1.2023, "step": 5991 }, { "epoch": 0.61, "learning_rate": 5.6648561863707263e-05, "loss": 1.1513, "step": 5992 }, { "epoch": 0.61, "learning_rate": 5.6647462615785735e-05, "loss": 0.7323, "step": 5993 }, { "epoch": 0.61, "learning_rate": 5.6646363198289904e-05, "loss": 0.6973, "step": 5994 }, { "epoch": 0.61, "learning_rate": 5.6645263611226744e-05, "loss": 1.1355, "step": 5995 }, { "epoch": 0.61, "learning_rate": 5.664416385460327e-05, "loss": 1.3136, "step": 5996 }, { "epoch": 0.61, "learning_rate": 5.664306392842646e-05, "loss": 0.8591, "step": 5997 }, { "epoch": 0.61, "learning_rate": 5.664196383270334e-05, "loss": 0.492, "step": 5998 }, { "epoch": 0.61, "learning_rate": 5.66408635674409e-05, "loss": 1.1306, "step": 5999 }, { "epoch": 0.61, "learning_rate": 5.663976313264613e-05, "loss": 1.3323, "step": 6000 }, { "epoch": 0.61, "learning_rate": 5.663866252832604e-05, "loss": 0.7347, "step": 6001 }, { "epoch": 0.61, "learning_rate": 5.663756175448764e-05, "loss": 1.1456, "step": 6002 }, { "epoch": 0.61, "learning_rate": 5.663646081113793e-05, "loss": 1.3731, "step": 6003 }, { "epoch": 0.61, "learning_rate": 5.663535969828392e-05, "loss": 0.8794, "step": 6004 }, { "epoch": 0.61, "learning_rate": 5.6634258415932614e-05, "loss": 0.2429, "step": 6005 }, { "epoch": 0.61, "learning_rate": 5.663315696409101e-05, "loss": 0.7331, "step": 6006 }, { "epoch": 0.61, "learning_rate": 5.663205534276613e-05, "loss": 0.6181, "step": 6007 }, { "epoch": 0.61, "learning_rate": 5.663095355196499e-05, "loss": 0.363, "step": 6008 }, { "epoch": 0.61, "learning_rate": 5.6629851591694585e-05, "loss": 1.8659, "step": 6009 }, { "epoch": 0.61, "learning_rate": 5.662874946196193e-05, "loss": 0.9001, "step": 6010 }, { "epoch": 0.61, "learning_rate": 5.662764716277406e-05, "loss": 0.3511, "step": 6011 }, { "epoch": 0.61, "learning_rate": 5.6626544694137955e-05, "loss": 0.3914, "step": 6012 }, { "epoch": 0.61, "learning_rate": 5.6625442056060655e-05, "loss": 0.6525, "step": 6013 }, { "epoch": 0.61, "learning_rate": 5.662433924854917e-05, "loss": 1.0475, "step": 6014 }, { "epoch": 0.61, "learning_rate": 5.662323627161052e-05, "loss": 1.7327, "step": 6015 }, { "epoch": 0.61, "learning_rate": 5.662213312525173e-05, "loss": 0.8056, "step": 6016 }, { "epoch": 0.61, "learning_rate": 5.66210298094798e-05, "loss": 0.9889, "step": 6017 }, { "epoch": 0.61, "learning_rate": 5.661992632430177e-05, "loss": 0.2814, "step": 6018 }, { "epoch": 0.61, "learning_rate": 5.661882266972466e-05, "loss": 0.9707, "step": 6019 }, { "epoch": 0.61, "learning_rate": 5.6617718845755473e-05, "loss": 0.9816, "step": 6020 }, { "epoch": 0.61, "learning_rate": 5.661661485240126e-05, "loss": 0.5029, "step": 6021 }, { "epoch": 0.61, "learning_rate": 5.661551068966904e-05, "loss": 0.5035, "step": 6022 }, { "epoch": 0.61, "learning_rate": 5.661440635756583e-05, "loss": 0.5934, "step": 6023 }, { "epoch": 0.61, "learning_rate": 5.661330185609866e-05, "loss": 1.5582, "step": 6024 }, { "epoch": 0.61, "learning_rate": 5.6612197185274564e-05, "loss": 0.2936, "step": 6025 }, { "epoch": 0.61, "learning_rate": 5.661109234510058e-05, "loss": 0.5038, "step": 6026 }, { "epoch": 0.61, "learning_rate": 5.660998733558371e-05, "loss": 1.9003, "step": 6027 }, { "epoch": 0.61, "learning_rate": 5.660888215673101e-05, "loss": 1.7493, "step": 6028 }, { "epoch": 0.61, "learning_rate": 5.660777680854951e-05, "loss": 1.5624, "step": 6029 }, { "epoch": 0.61, "learning_rate": 5.6606671291046235e-05, "loss": 1.5129, "step": 6030 }, { "epoch": 0.61, "learning_rate": 5.6605565604228225e-05, "loss": 1.3333, "step": 6031 }, { "epoch": 0.61, "learning_rate": 5.660445974810252e-05, "loss": 1.1664, "step": 6032 }, { "epoch": 0.61, "learning_rate": 5.660335372267615e-05, "loss": 1.4176, "step": 6033 }, { "epoch": 0.61, "learning_rate": 5.660224752795616e-05, "loss": 1.3779, "step": 6034 }, { "epoch": 0.61, "learning_rate": 5.660114116394959e-05, "loss": 1.2812, "step": 6035 }, { "epoch": 0.61, "learning_rate": 5.660003463066347e-05, "loss": 0.416, "step": 6036 }, { "epoch": 0.61, "learning_rate": 5.659892792810486e-05, "loss": 0.7885, "step": 6037 }, { "epoch": 0.61, "learning_rate": 5.6597821056280785e-05, "loss": 1.368, "step": 6038 }, { "epoch": 0.61, "learning_rate": 5.659671401519829e-05, "loss": 0.5841, "step": 6039 }, { "epoch": 0.61, "learning_rate": 5.6595606804864436e-05, "loss": 0.1272, "step": 6040 }, { "epoch": 0.61, "learning_rate": 5.659449942528625e-05, "loss": 1.835, "step": 6041 }, { "epoch": 0.61, "learning_rate": 5.659339187647079e-05, "loss": 1.2013, "step": 6042 }, { "epoch": 0.61, "learning_rate": 5.65922841584251e-05, "loss": 1.1956, "step": 6043 }, { "epoch": 0.61, "learning_rate": 5.659117627115623e-05, "loss": 0.8319, "step": 6044 }, { "epoch": 0.61, "learning_rate": 5.659006821467122e-05, "loss": 0.5544, "step": 6045 }, { "epoch": 0.61, "learning_rate": 5.658895998897714e-05, "loss": 0.5232, "step": 6046 }, { "epoch": 0.61, "learning_rate": 5.658785159408104e-05, "loss": 0.8796, "step": 6047 }, { "epoch": 0.61, "learning_rate": 5.6586743029989956e-05, "loss": 0.6335, "step": 6048 }, { "epoch": 0.61, "learning_rate": 5.658563429671096e-05, "loss": 1.5022, "step": 6049 }, { "epoch": 0.61, "learning_rate": 5.6584525394251094e-05, "loss": 0.7392, "step": 6050 }, { "epoch": 0.61, "learning_rate": 5.658341632261743e-05, "loss": 0.6644, "step": 6051 }, { "epoch": 0.61, "learning_rate": 5.658230708181702e-05, "loss": 1.5677, "step": 6052 }, { "epoch": 0.61, "learning_rate": 5.658119767185691e-05, "loss": 0.9774, "step": 6053 }, { "epoch": 0.61, "learning_rate": 5.6580088092744176e-05, "loss": 0.1809, "step": 6054 }, { "epoch": 0.61, "learning_rate": 5.657897834448587e-05, "loss": 0.8408, "step": 6055 }, { "epoch": 0.61, "learning_rate": 5.657786842708906e-05, "loss": 1.2159, "step": 6056 }, { "epoch": 0.62, "learning_rate": 5.6576758340560806e-05, "loss": 0.9985, "step": 6057 }, { "epoch": 0.62, "learning_rate": 5.6575648084908174e-05, "loss": 0.4619, "step": 6058 }, { "epoch": 0.62, "learning_rate": 5.657453766013822e-05, "loss": 0.7533, "step": 6059 }, { "epoch": 0.62, "learning_rate": 5.6573427066258024e-05, "loss": 1.0876, "step": 6060 }, { "epoch": 0.62, "learning_rate": 5.657231630327464e-05, "loss": 0.3164, "step": 6061 }, { "epoch": 0.62, "learning_rate": 5.657120537119516e-05, "loss": 0.1562, "step": 6062 }, { "epoch": 0.62, "learning_rate": 5.6570094270026615e-05, "loss": 1.4003, "step": 6063 }, { "epoch": 0.62, "learning_rate": 5.656898299977611e-05, "loss": 0.9301, "step": 6064 }, { "epoch": 0.62, "learning_rate": 5.6567871560450705e-05, "loss": 0.7472, "step": 6065 }, { "epoch": 0.62, "learning_rate": 5.6566759952057474e-05, "loss": 0.4716, "step": 6066 }, { "epoch": 0.62, "learning_rate": 5.6565648174603486e-05, "loss": 1.3471, "step": 6067 }, { "epoch": 0.62, "learning_rate": 5.656453622809581e-05, "loss": 0.8369, "step": 6068 }, { "epoch": 0.62, "learning_rate": 5.656342411254155e-05, "loss": 0.3544, "step": 6069 }, { "epoch": 0.62, "learning_rate": 5.656231182794775e-05, "loss": 1.2762, "step": 6070 }, { "epoch": 0.62, "learning_rate": 5.6561199374321505e-05, "loss": 0.8883, "step": 6071 }, { "epoch": 0.62, "learning_rate": 5.6560086751669894e-05, "loss": 0.9657, "step": 6072 }, { "epoch": 0.62, "learning_rate": 5.6558973959999995e-05, "loss": 1.0775, "step": 6073 }, { "epoch": 0.62, "learning_rate": 5.655786099931889e-05, "loss": 0.6965, "step": 6074 }, { "epoch": 0.62, "learning_rate": 5.655674786963366e-05, "loss": 1.4892, "step": 6075 }, { "epoch": 0.62, "learning_rate": 5.6555634570951385e-05, "loss": 1.8003, "step": 6076 }, { "epoch": 0.62, "learning_rate": 5.655452110327916e-05, "loss": 0.3875, "step": 6077 }, { "epoch": 0.62, "learning_rate": 5.6553407466624066e-05, "loss": 0.8097, "step": 6078 }, { "epoch": 0.62, "learning_rate": 5.6552293660993183e-05, "loss": 0.6936, "step": 6079 }, { "epoch": 0.62, "learning_rate": 5.655117968639361e-05, "loss": 0.3351, "step": 6080 }, { "epoch": 0.62, "learning_rate": 5.655006554283242e-05, "loss": 0.8507, "step": 6081 }, { "epoch": 0.62, "learning_rate": 5.654895123031672e-05, "loss": 0.837, "step": 6082 }, { "epoch": 0.62, "learning_rate": 5.6547836748853595e-05, "loss": 1.2849, "step": 6083 }, { "epoch": 0.62, "learning_rate": 5.654672209845013e-05, "loss": 0.1406, "step": 6084 }, { "epoch": 0.62, "learning_rate": 5.6545607279113425e-05, "loss": 1.2844, "step": 6085 }, { "epoch": 0.62, "learning_rate": 5.654449229085058e-05, "loss": 1.6662, "step": 6086 }, { "epoch": 0.62, "learning_rate": 5.6543377133668686e-05, "loss": 0.2071, "step": 6087 }, { "epoch": 0.62, "learning_rate": 5.654226180757483e-05, "loss": 0.956, "step": 6088 }, { "epoch": 0.62, "learning_rate": 5.6541146312576115e-05, "loss": 1.2552, "step": 6089 }, { "epoch": 0.62, "learning_rate": 5.654003064867965e-05, "loss": 1.0291, "step": 6090 }, { "epoch": 0.62, "learning_rate": 5.653891481589252e-05, "loss": 0.2277, "step": 6091 }, { "epoch": 0.62, "learning_rate": 5.653779881422184e-05, "loss": 0.9308, "step": 6092 }, { "epoch": 0.62, "learning_rate": 5.653668264367469e-05, "loss": 1.3527, "step": 6093 }, { "epoch": 0.62, "learning_rate": 5.65355663042582e-05, "loss": 1.1593, "step": 6094 }, { "epoch": 0.62, "learning_rate": 5.653444979597946e-05, "loss": 0.6772, "step": 6095 }, { "epoch": 0.62, "learning_rate": 5.6533333118845566e-05, "loss": 1.0868, "step": 6096 }, { "epoch": 0.62, "learning_rate": 5.653221627286364e-05, "loss": 1.1908, "step": 6097 }, { "epoch": 0.62, "learning_rate": 5.653109925804078e-05, "loss": 1.0598, "step": 6098 }, { "epoch": 0.62, "learning_rate": 5.65299820743841e-05, "loss": 0.7799, "step": 6099 }, { "epoch": 0.62, "learning_rate": 5.652886472190071e-05, "loss": 1.0747, "step": 6100 }, { "epoch": 0.62, "learning_rate": 5.652774720059771e-05, "loss": 0.9056, "step": 6101 }, { "epoch": 0.62, "learning_rate": 5.652662951048223e-05, "loss": 0.5535, "step": 6102 }, { "epoch": 0.62, "learning_rate": 5.6525511651561354e-05, "loss": 1.1167, "step": 6103 }, { "epoch": 0.62, "learning_rate": 5.652439362384222e-05, "loss": 1.3301, "step": 6104 }, { "epoch": 0.62, "learning_rate": 5.652327542733194e-05, "loss": 0.6306, "step": 6105 }, { "epoch": 0.62, "learning_rate": 5.652215706203763e-05, "loss": 0.8933, "step": 6106 }, { "epoch": 0.62, "learning_rate": 5.652103852796639e-05, "loss": 1.5261, "step": 6107 }, { "epoch": 0.62, "learning_rate": 5.6519919825125354e-05, "loss": 1.5811, "step": 6108 }, { "epoch": 0.62, "learning_rate": 5.6518800953521636e-05, "loss": 1.0721, "step": 6109 }, { "epoch": 0.62, "learning_rate": 5.651768191316236e-05, "loss": 0.9919, "step": 6110 }, { "epoch": 0.62, "learning_rate": 5.651656270405464e-05, "loss": 0.6845, "step": 6111 }, { "epoch": 0.62, "learning_rate": 5.6515443326205604e-05, "loss": 1.4347, "step": 6112 }, { "epoch": 0.62, "learning_rate": 5.651432377962238e-05, "loss": 0.6122, "step": 6113 }, { "epoch": 0.62, "learning_rate": 5.651320406431207e-05, "loss": 0.9826, "step": 6114 }, { "epoch": 0.62, "learning_rate": 5.651208418028183e-05, "loss": 0.3004, "step": 6115 }, { "epoch": 0.62, "learning_rate": 5.6510964127538764e-05, "loss": 1.3043, "step": 6116 }, { "epoch": 0.62, "learning_rate": 5.650984390609001e-05, "loss": 1.4562, "step": 6117 }, { "epoch": 0.62, "learning_rate": 5.65087235159427e-05, "loss": 0.596, "step": 6118 }, { "epoch": 0.62, "learning_rate": 5.650760295710396e-05, "loss": 0.3155, "step": 6119 }, { "epoch": 0.62, "learning_rate": 5.6506482229580904e-05, "loss": 1.5468, "step": 6120 }, { "epoch": 0.62, "learning_rate": 5.650536133338069e-05, "loss": 0.4311, "step": 6121 }, { "epoch": 0.62, "learning_rate": 5.6504240268510437e-05, "loss": 0.272, "step": 6122 }, { "epoch": 0.62, "learning_rate": 5.650311903497728e-05, "loss": 1.6208, "step": 6123 }, { "epoch": 0.62, "learning_rate": 5.650199763278836e-05, "loss": 0.8028, "step": 6124 }, { "epoch": 0.62, "learning_rate": 5.6500876061950805e-05, "loss": 1.152, "step": 6125 }, { "epoch": 0.62, "learning_rate": 5.649975432247176e-05, "loss": 0.6293, "step": 6126 }, { "epoch": 0.62, "learning_rate": 5.649863241435835e-05, "loss": 1.0029, "step": 6127 }, { "epoch": 0.62, "learning_rate": 5.649751033761773e-05, "loss": 0.8597, "step": 6128 }, { "epoch": 0.62, "learning_rate": 5.649638809225704e-05, "loss": 1.1461, "step": 6129 }, { "epoch": 0.62, "learning_rate": 5.649526567828341e-05, "loss": 0.2165, "step": 6130 }, { "epoch": 0.62, "learning_rate": 5.6494143095703985e-05, "loss": 0.6225, "step": 6131 }, { "epoch": 0.62, "learning_rate": 5.649302034452592e-05, "loss": 0.407, "step": 6132 }, { "epoch": 0.62, "learning_rate": 5.649189742475635e-05, "loss": 1.0684, "step": 6133 }, { "epoch": 0.62, "learning_rate": 5.649077433640242e-05, "loss": 0.4818, "step": 6134 }, { "epoch": 0.62, "learning_rate": 5.6489651079471275e-05, "loss": 0.2643, "step": 6135 }, { "epoch": 0.62, "learning_rate": 5.6488527653970076e-05, "loss": 0.2153, "step": 6136 }, { "epoch": 0.62, "learning_rate": 5.648740405990596e-05, "loss": 0.3958, "step": 6137 }, { "epoch": 0.62, "learning_rate": 5.648628029728609e-05, "loss": 0.4886, "step": 6138 }, { "epoch": 0.62, "learning_rate": 5.6485156366117596e-05, "loss": 1.5124, "step": 6139 }, { "epoch": 0.62, "learning_rate": 5.648403226640765e-05, "loss": 1.2617, "step": 6140 }, { "epoch": 0.62, "learning_rate": 5.648290799816339e-05, "loss": 0.6342, "step": 6141 }, { "epoch": 0.62, "learning_rate": 5.648178356139199e-05, "loss": 1.2589, "step": 6142 }, { "epoch": 0.62, "learning_rate": 5.6480658956100584e-05, "loss": 0.7498, "step": 6143 }, { "epoch": 0.62, "learning_rate": 5.647953418229634e-05, "loss": 0.9843, "step": 6144 }, { "epoch": 0.62, "learning_rate": 5.6478409239986405e-05, "loss": 0.4235, "step": 6145 }, { "epoch": 0.62, "learning_rate": 5.647728412917797e-05, "loss": 0.1186, "step": 6146 }, { "epoch": 0.62, "learning_rate": 5.647615884987815e-05, "loss": 1.5359, "step": 6147 }, { "epoch": 0.62, "learning_rate": 5.647503340209414e-05, "loss": 1.1158, "step": 6148 }, { "epoch": 0.62, "learning_rate": 5.6473907785833076e-05, "loss": 1.314, "step": 6149 }, { "epoch": 0.62, "learning_rate": 5.647278200110214e-05, "loss": 0.6088, "step": 6150 }, { "epoch": 0.62, "learning_rate": 5.64716560479085e-05, "loss": 0.6301, "step": 6151 }, { "epoch": 0.62, "learning_rate": 5.64705299262593e-05, "loss": 0.6789, "step": 6152 }, { "epoch": 0.62, "learning_rate": 5.6469403636161717e-05, "loss": 1.3092, "step": 6153 }, { "epoch": 0.62, "learning_rate": 5.6468277177622924e-05, "loss": 1.2365, "step": 6154 }, { "epoch": 0.62, "learning_rate": 5.646715055065008e-05, "loss": 0.4071, "step": 6155 }, { "epoch": 0.63, "learning_rate": 5.6466023755250355e-05, "loss": 1.3398, "step": 6156 }, { "epoch": 0.63, "learning_rate": 5.646489679143093e-05, "loss": 0.3797, "step": 6157 }, { "epoch": 0.63, "learning_rate": 5.6463769659198975e-05, "loss": 1.0731, "step": 6158 }, { "epoch": 0.63, "learning_rate": 5.646264235856165e-05, "loss": 0.8029, "step": 6159 }, { "epoch": 0.63, "learning_rate": 5.646151488952613e-05, "loss": 1.1356, "step": 6160 }, { "epoch": 0.63, "learning_rate": 5.646038725209961e-05, "loss": 1.6227, "step": 6161 }, { "epoch": 0.63, "learning_rate": 5.645925944628924e-05, "loss": 1.3399, "step": 6162 }, { "epoch": 0.63, "learning_rate": 5.645813147210221e-05, "loss": 0.578, "step": 6163 }, { "epoch": 0.63, "learning_rate": 5.64570033295457e-05, "loss": 0.3176, "step": 6164 }, { "epoch": 0.63, "learning_rate": 5.6455875018626886e-05, "loss": 0.4335, "step": 6165 }, { "epoch": 0.63, "learning_rate": 5.6454746539352945e-05, "loss": 0.368, "step": 6166 }, { "epoch": 0.63, "learning_rate": 5.6453617891731056e-05, "loss": 1.4584, "step": 6167 }, { "epoch": 0.63, "learning_rate": 5.645248907576842e-05, "loss": 0.789, "step": 6168 }, { "epoch": 0.63, "learning_rate": 5.6451360091472196e-05, "loss": 0.7411, "step": 6169 }, { "epoch": 0.63, "learning_rate": 5.645023093884958e-05, "loss": 0.5136, "step": 6170 }, { "epoch": 0.63, "learning_rate": 5.6449101617907754e-05, "loss": 1.0173, "step": 6171 }, { "epoch": 0.63, "learning_rate": 5.644797212865391e-05, "loss": 1.5889, "step": 6172 }, { "epoch": 0.63, "learning_rate": 5.644684247109524e-05, "loss": 0.2098, "step": 6173 }, { "epoch": 0.63, "learning_rate": 5.6445712645238915e-05, "loss": 1.6506, "step": 6174 }, { "epoch": 0.63, "learning_rate": 5.644458265109214e-05, "loss": 0.9493, "step": 6175 }, { "epoch": 0.63, "learning_rate": 5.64434524886621e-05, "loss": 0.4819, "step": 6176 }, { "epoch": 0.63, "learning_rate": 5.6442322157955985e-05, "loss": 0.4348, "step": 6177 }, { "epoch": 0.63, "learning_rate": 5.6441191658981e-05, "loss": 0.5789, "step": 6178 }, { "epoch": 0.63, "learning_rate": 5.6440060991744324e-05, "loss": 0.9506, "step": 6179 }, { "epoch": 0.63, "learning_rate": 5.643893015625316e-05, "loss": 0.228, "step": 6180 }, { "epoch": 0.63, "learning_rate": 5.64377991525147e-05, "loss": 0.7396, "step": 6181 }, { "epoch": 0.63, "learning_rate": 5.6436667980536145e-05, "loss": 0.9672, "step": 6182 }, { "epoch": 0.63, "learning_rate": 5.6435536640324684e-05, "loss": 1.1125, "step": 6183 }, { "epoch": 0.63, "learning_rate": 5.643440513188754e-05, "loss": 1.36, "step": 6184 }, { "epoch": 0.63, "learning_rate": 5.643327345523189e-05, "loss": 0.3176, "step": 6185 }, { "epoch": 0.63, "learning_rate": 5.643214161036494e-05, "loss": 1.4026, "step": 6186 }, { "epoch": 0.63, "learning_rate": 5.64310095972939e-05, "loss": 0.8793, "step": 6187 }, { "epoch": 0.63, "learning_rate": 5.642987741602597e-05, "loss": 1.7102, "step": 6188 }, { "epoch": 0.63, "learning_rate": 5.6428745066568354e-05, "loss": 1.2541, "step": 6189 }, { "epoch": 0.63, "learning_rate": 5.6427612548928254e-05, "loss": 1.4419, "step": 6190 }, { "epoch": 0.63, "learning_rate": 5.642647986311289e-05, "loss": 1.3076, "step": 6191 }, { "epoch": 0.63, "learning_rate": 5.6425347009129455e-05, "loss": 1.0822, "step": 6192 }, { "epoch": 0.63, "learning_rate": 5.642421398698517e-05, "loss": 1.1118, "step": 6193 }, { "epoch": 0.63, "learning_rate": 5.642308079668723e-05, "loss": 1.4162, "step": 6194 }, { "epoch": 0.63, "learning_rate": 5.6421947438242864e-05, "loss": 0.5287, "step": 6195 }, { "epoch": 0.63, "learning_rate": 5.642081391165927e-05, "loss": 0.2226, "step": 6196 }, { "epoch": 0.63, "learning_rate": 5.641968021694367e-05, "loss": 0.7299, "step": 6197 }, { "epoch": 0.63, "learning_rate": 5.641854635410328e-05, "loss": 1.3356, "step": 6198 }, { "epoch": 0.63, "learning_rate": 5.6417412323145305e-05, "loss": 0.794, "step": 6199 }, { "epoch": 0.63, "learning_rate": 5.641627812407697e-05, "loss": 1.6208, "step": 6200 }, { "epoch": 0.63, "learning_rate": 5.641514375690549e-05, "loss": 0.5447, "step": 6201 }, { "epoch": 0.63, "learning_rate": 5.641400922163808e-05, "loss": 0.7537, "step": 6202 }, { "epoch": 0.63, "learning_rate": 5.6412874518281976e-05, "loss": 0.4673, "step": 6203 }, { "epoch": 0.63, "learning_rate": 5.6411739646844375e-05, "loss": 0.4501, "step": 6204 }, { "epoch": 0.63, "learning_rate": 5.641060460733252e-05, "loss": 1.2563, "step": 6205 }, { "epoch": 0.63, "learning_rate": 5.640946939975361e-05, "loss": 1.8313, "step": 6206 }, { "epoch": 0.63, "learning_rate": 5.640833402411489e-05, "loss": 0.6923, "step": 6207 }, { "epoch": 0.63, "learning_rate": 5.640719848042358e-05, "loss": 0.9366, "step": 6208 }, { "epoch": 0.63, "learning_rate": 5.6406062768686904e-05, "loss": 1.362, "step": 6209 }, { "epoch": 0.63, "learning_rate": 5.6404926888912096e-05, "loss": 0.2759, "step": 6210 }, { "epoch": 0.63, "learning_rate": 5.640379084110637e-05, "loss": 0.3592, "step": 6211 }, { "epoch": 0.63, "learning_rate": 5.640265462527697e-05, "loss": 0.5104, "step": 6212 }, { "epoch": 0.63, "learning_rate": 5.640151824143112e-05, "loss": 0.9671, "step": 6213 }, { "epoch": 0.63, "learning_rate": 5.6400381689576045e-05, "loss": 1.0221, "step": 6214 }, { "epoch": 0.63, "learning_rate": 5.639924496971898e-05, "loss": 0.519, "step": 6215 }, { "epoch": 0.63, "learning_rate": 5.639810808186718e-05, "loss": 0.4127, "step": 6216 }, { "epoch": 0.63, "learning_rate": 5.639697102602785e-05, "loss": 0.6564, "step": 6217 }, { "epoch": 0.63, "learning_rate": 5.639583380220824e-05, "loss": 1.4322, "step": 6218 }, { "epoch": 0.63, "learning_rate": 5.6394696410415584e-05, "loss": 1.1088, "step": 6219 }, { "epoch": 0.63, "learning_rate": 5.6393558850657123e-05, "loss": 0.4565, "step": 6220 }, { "epoch": 0.63, "learning_rate": 5.639242112294009e-05, "loss": 0.623, "step": 6221 }, { "epoch": 0.63, "learning_rate": 5.639128322727174e-05, "loss": 0.4743, "step": 6222 }, { "epoch": 0.63, "learning_rate": 5.639014516365929e-05, "loss": 1.4127, "step": 6223 }, { "epoch": 0.63, "learning_rate": 5.638900693211001e-05, "loss": 1.4016, "step": 6224 }, { "epoch": 0.63, "learning_rate": 5.638786853263112e-05, "loss": 1.4884, "step": 6225 }, { "epoch": 0.63, "learning_rate": 5.6386729965229876e-05, "loss": 0.8859, "step": 6226 }, { "epoch": 0.63, "learning_rate": 5.6385591229913514e-05, "loss": 0.582, "step": 6227 }, { "epoch": 0.63, "learning_rate": 5.638445232668929e-05, "loss": 0.387, "step": 6228 }, { "epoch": 0.63, "learning_rate": 5.638331325556445e-05, "loss": 0.9849, "step": 6229 }, { "epoch": 0.63, "learning_rate": 5.638217401654624e-05, "loss": 0.4577, "step": 6230 }, { "epoch": 0.63, "learning_rate": 5.638103460964192e-05, "loss": 1.2742, "step": 6231 }, { "epoch": 0.63, "learning_rate": 5.637989503485872e-05, "loss": 0.2713, "step": 6232 }, { "epoch": 0.63, "learning_rate": 5.637875529220391e-05, "loss": 0.6826, "step": 6233 }, { "epoch": 0.63, "learning_rate": 5.637761538168473e-05, "loss": 0.3943, "step": 6234 }, { "epoch": 0.63, "learning_rate": 5.637647530330845e-05, "loss": 0.2878, "step": 6235 }, { "epoch": 0.63, "learning_rate": 5.63753350570823e-05, "loss": 0.1971, "step": 6236 }, { "epoch": 0.63, "learning_rate": 5.637419464301356e-05, "loss": 0.1383, "step": 6237 }, { "epoch": 0.63, "learning_rate": 5.6373054061109475e-05, "loss": 0.6154, "step": 6238 }, { "epoch": 0.63, "learning_rate": 5.637191331137732e-05, "loss": 1.2454, "step": 6239 }, { "epoch": 0.63, "learning_rate": 5.6370772393824326e-05, "loss": 0.5883, "step": 6240 }, { "epoch": 0.63, "learning_rate": 5.6369631308457776e-05, "loss": 1.1718, "step": 6241 }, { "epoch": 0.63, "learning_rate": 5.6368490055284926e-05, "loss": 1.0096, "step": 6242 }, { "epoch": 0.63, "learning_rate": 5.6367348634313026e-05, "loss": 0.6467, "step": 6243 }, { "epoch": 0.63, "learning_rate": 5.636620704554936e-05, "loss": 0.9159, "step": 6244 }, { "epoch": 0.63, "learning_rate": 5.636506528900118e-05, "loss": 1.5398, "step": 6245 }, { "epoch": 0.63, "learning_rate": 5.636392336467576e-05, "loss": 0.634, "step": 6246 }, { "epoch": 0.63, "learning_rate": 5.6362781272580356e-05, "loss": 0.2874, "step": 6247 }, { "epoch": 0.63, "learning_rate": 5.636163901272223e-05, "loss": 0.9195, "step": 6248 }, { "epoch": 0.63, "learning_rate": 5.6360496585108675e-05, "loss": 0.5631, "step": 6249 }, { "epoch": 0.63, "learning_rate": 5.635935398974694e-05, "loss": 0.9091, "step": 6250 }, { "epoch": 0.63, "learning_rate": 5.6358211226644314e-05, "loss": 1.5332, "step": 6251 }, { "epoch": 0.63, "learning_rate": 5.6357068295808044e-05, "loss": 0.8041, "step": 6252 }, { "epoch": 0.63, "learning_rate": 5.6355925197245434e-05, "loss": 0.4109, "step": 6253 }, { "epoch": 0.64, "learning_rate": 5.635478193096373e-05, "loss": 1.3048, "step": 6254 }, { "epoch": 0.64, "learning_rate": 5.635363849697023e-05, "loss": 0.6252, "step": 6255 }, { "epoch": 0.64, "learning_rate": 5.6352494895272195e-05, "loss": 0.4526, "step": 6256 }, { "epoch": 0.64, "learning_rate": 5.6351351125876905e-05, "loss": 0.7332, "step": 6257 }, { "epoch": 0.64, "learning_rate": 5.6350207188791644e-05, "loss": 0.7963, "step": 6258 }, { "epoch": 0.64, "learning_rate": 5.634906308402368e-05, "loss": 1.2466, "step": 6259 }, { "epoch": 0.64, "learning_rate": 5.6347918811580314e-05, "loss": 0.8124, "step": 6260 }, { "epoch": 0.64, "learning_rate": 5.6346774371468816e-05, "loss": 0.4097, "step": 6261 }, { "epoch": 0.64, "learning_rate": 5.634562976369646e-05, "loss": 0.9418, "step": 6262 }, { "epoch": 0.64, "learning_rate": 5.6344484988270547e-05, "loss": 0.6416, "step": 6263 }, { "epoch": 0.64, "learning_rate": 5.634334004519835e-05, "loss": 1.0823, "step": 6264 }, { "epoch": 0.64, "learning_rate": 5.6342194934487155e-05, "loss": 0.5055, "step": 6265 }, { "epoch": 0.64, "learning_rate": 5.634104965614426e-05, "loss": 0.5404, "step": 6266 }, { "epoch": 0.64, "learning_rate": 5.6339904210176945e-05, "loss": 0.9994, "step": 6267 }, { "epoch": 0.64, "learning_rate": 5.633875859659249e-05, "loss": 0.5997, "step": 6268 }, { "epoch": 0.64, "learning_rate": 5.6337612815398205e-05, "loss": 0.463, "step": 6269 }, { "epoch": 0.64, "learning_rate": 5.633646686660137e-05, "loss": 1.6371, "step": 6270 }, { "epoch": 0.64, "learning_rate": 5.633532075020927e-05, "loss": 0.7464, "step": 6271 }, { "epoch": 0.64, "learning_rate": 5.633417446622922e-05, "loss": 1.5413, "step": 6272 }, { "epoch": 0.64, "learning_rate": 5.633302801466849e-05, "loss": 0.7445, "step": 6273 }, { "epoch": 0.64, "learning_rate": 5.63318813955344e-05, "loss": 0.7785, "step": 6274 }, { "epoch": 0.64, "learning_rate": 5.6330734608834224e-05, "loss": 0.4942, "step": 6275 }, { "epoch": 0.64, "learning_rate": 5.6329587654575265e-05, "loss": 0.6486, "step": 6276 }, { "epoch": 0.64, "learning_rate": 5.632844053276484e-05, "loss": 0.4273, "step": 6277 }, { "epoch": 0.64, "learning_rate": 5.632729324341023e-05, "loss": 0.821, "step": 6278 }, { "epoch": 0.64, "learning_rate": 5.6326145786518744e-05, "loss": 0.951, "step": 6279 }, { "epoch": 0.64, "learning_rate": 5.632499816209768e-05, "loss": 0.8361, "step": 6280 }, { "epoch": 0.64, "learning_rate": 5.6323850370154334e-05, "loss": 0.8834, "step": 6281 }, { "epoch": 0.64, "learning_rate": 5.632270241069603e-05, "loss": 0.8048, "step": 6282 }, { "epoch": 0.64, "learning_rate": 5.632155428373005e-05, "loss": 1.2304, "step": 6283 }, { "epoch": 0.64, "learning_rate": 5.6320405989263715e-05, "loss": 1.5066, "step": 6284 }, { "epoch": 0.64, "learning_rate": 5.631925752730433e-05, "loss": 0.6731, "step": 6285 }, { "epoch": 0.64, "learning_rate": 5.631810889785921e-05, "loss": 1.1658, "step": 6286 }, { "epoch": 0.64, "learning_rate": 5.631696010093565e-05, "loss": 0.7289, "step": 6287 }, { "epoch": 0.64, "learning_rate": 5.631581113654096e-05, "loss": 0.456, "step": 6288 }, { "epoch": 0.64, "learning_rate": 5.6314662004682473e-05, "loss": 1.144, "step": 6289 }, { "epoch": 0.64, "learning_rate": 5.631351270536748e-05, "loss": 0.7272, "step": 6290 }, { "epoch": 0.64, "learning_rate": 5.6312363238603305e-05, "loss": 0.984, "step": 6291 }, { "epoch": 0.64, "learning_rate": 5.631121360439726e-05, "loss": 1.5511, "step": 6292 }, { "epoch": 0.64, "learning_rate": 5.631006380275666e-05, "loss": 1.721, "step": 6293 }, { "epoch": 0.64, "learning_rate": 5.6308913833688825e-05, "loss": 1.0011, "step": 6294 }, { "epoch": 0.64, "learning_rate": 5.630776369720107e-05, "loss": 1.0073, "step": 6295 }, { "epoch": 0.64, "learning_rate": 5.630661339330071e-05, "loss": 0.6425, "step": 6296 }, { "epoch": 0.64, "learning_rate": 5.630546292199507e-05, "loss": 1.3539, "step": 6297 }, { "epoch": 0.64, "learning_rate": 5.630431228329147e-05, "loss": 0.9227, "step": 6298 }, { "epoch": 0.64, "learning_rate": 5.630316147719724e-05, "loss": 0.8609, "step": 6299 }, { "epoch": 0.64, "learning_rate": 5.6302010503719704e-05, "loss": 0.568, "step": 6300 }, { "epoch": 0.64, "learning_rate": 5.630085936286617e-05, "loss": 2.0707, "step": 6301 }, { "epoch": 0.64, "learning_rate": 5.629970805464397e-05, "loss": 0.4586, "step": 6302 }, { "epoch": 0.64, "learning_rate": 5.629855657906044e-05, "loss": 1.3759, "step": 6303 }, { "epoch": 0.64, "learning_rate": 5.629740493612289e-05, "loss": 1.0056, "step": 6304 }, { "epoch": 0.64, "learning_rate": 5.629625312583866e-05, "loss": 1.6187, "step": 6305 }, { "epoch": 0.64, "learning_rate": 5.6295101148215084e-05, "loss": 0.3366, "step": 6306 }, { "epoch": 0.64, "learning_rate": 5.6293949003259485e-05, "loss": 1.9024, "step": 6307 }, { "epoch": 0.64, "learning_rate": 5.6292796690979195e-05, "loss": 1.2255, "step": 6308 }, { "epoch": 0.64, "learning_rate": 5.6291644211381567e-05, "loss": 1.0099, "step": 6309 }, { "epoch": 0.64, "learning_rate": 5.62904915644739e-05, "loss": 1.6197, "step": 6310 }, { "epoch": 0.64, "learning_rate": 5.628933875026355e-05, "loss": 1.3089, "step": 6311 }, { "epoch": 0.64, "learning_rate": 5.628818576875784e-05, "loss": 0.5829, "step": 6312 }, { "epoch": 0.64, "learning_rate": 5.628703261996413e-05, "loss": 1.3878, "step": 6313 }, { "epoch": 0.64, "learning_rate": 5.628587930388973e-05, "loss": 1.1376, "step": 6314 }, { "epoch": 0.64, "learning_rate": 5.628472582054201e-05, "loss": 1.0727, "step": 6315 }, { "epoch": 0.64, "learning_rate": 5.628357216992828e-05, "loss": 1.0586, "step": 6316 }, { "epoch": 0.64, "learning_rate": 5.6282418352055906e-05, "loss": 1.3552, "step": 6317 }, { "epoch": 0.64, "learning_rate": 5.6281264366932214e-05, "loss": 1.3235, "step": 6318 }, { "epoch": 0.64, "learning_rate": 5.628011021456456e-05, "loss": 0.9424, "step": 6319 }, { "epoch": 0.64, "learning_rate": 5.627895589496028e-05, "loss": 0.7549, "step": 6320 }, { "epoch": 0.64, "learning_rate": 5.627780140812672e-05, "loss": 1.6227, "step": 6321 }, { "epoch": 0.64, "learning_rate": 5.6276646754071225e-05, "loss": 0.6483, "step": 6322 }, { "epoch": 0.64, "learning_rate": 5.627549193280115e-05, "loss": 0.4267, "step": 6323 }, { "epoch": 0.64, "learning_rate": 5.627433694432384e-05, "loss": 0.3274, "step": 6324 }, { "epoch": 0.64, "learning_rate": 5.6273181788646644e-05, "loss": 0.8429, "step": 6325 }, { "epoch": 0.64, "learning_rate": 5.6272026465776916e-05, "loss": 0.8522, "step": 6326 }, { "epoch": 0.64, "learning_rate": 5.6270870975722e-05, "loss": 1.4739, "step": 6327 }, { "epoch": 0.64, "learning_rate": 5.626971531848926e-05, "loss": 0.588, "step": 6328 }, { "epoch": 0.64, "learning_rate": 5.626855949408605e-05, "loss": 0.3539, "step": 6329 }, { "epoch": 0.64, "learning_rate": 5.626740350251971e-05, "loss": 1.0584, "step": 6330 }, { "epoch": 0.64, "learning_rate": 5.626624734379761e-05, "loss": 0.6476, "step": 6331 }, { "epoch": 0.64, "learning_rate": 5.6265091017927107e-05, "loss": 1.2366, "step": 6332 }, { "epoch": 0.64, "learning_rate": 5.6263934524915555e-05, "loss": 1.1558, "step": 6333 }, { "epoch": 0.64, "learning_rate": 5.6262777864770315e-05, "loss": 1.5881, "step": 6334 }, { "epoch": 0.64, "learning_rate": 5.626162103749874e-05, "loss": 0.4436, "step": 6335 }, { "epoch": 0.64, "learning_rate": 5.626046404310821e-05, "loss": 0.5822, "step": 6336 }, { "epoch": 0.64, "learning_rate": 5.625930688160608e-05, "loss": 0.8051, "step": 6337 }, { "epoch": 0.64, "learning_rate": 5.62581495529997e-05, "loss": 1.3764, "step": 6338 }, { "epoch": 0.64, "learning_rate": 5.6256992057296444e-05, "loss": 1.3079, "step": 6339 }, { "epoch": 0.64, "learning_rate": 5.625583439450369e-05, "loss": 1.1928, "step": 6340 }, { "epoch": 0.64, "learning_rate": 5.625467656462879e-05, "loss": 0.3634, "step": 6341 }, { "epoch": 0.64, "learning_rate": 5.625351856767911e-05, "loss": 0.3933, "step": 6342 }, { "epoch": 0.64, "learning_rate": 5.6252360403662036e-05, "loss": 0.5002, "step": 6343 }, { "epoch": 0.64, "learning_rate": 5.625120207258492e-05, "loss": 1.0119, "step": 6344 }, { "epoch": 0.64, "learning_rate": 5.625004357445514e-05, "loss": 0.5159, "step": 6345 }, { "epoch": 0.64, "learning_rate": 5.624888490928007e-05, "loss": 1.4671, "step": 6346 }, { "epoch": 0.64, "learning_rate": 5.624772607706709e-05, "loss": 0.696, "step": 6347 }, { "epoch": 0.64, "learning_rate": 5.624656707782355e-05, "loss": 1.6991, "step": 6348 }, { "epoch": 0.64, "learning_rate": 5.6245407911556855e-05, "loss": 0.5661, "step": 6349 }, { "epoch": 0.64, "learning_rate": 5.6244248578274364e-05, "loss": 0.5338, "step": 6350 }, { "epoch": 0.64, "learning_rate": 5.624308907798346e-05, "loss": 1.6533, "step": 6351 }, { "epoch": 0.65, "learning_rate": 5.6241929410691514e-05, "loss": 1.1044, "step": 6352 }, { "epoch": 0.65, "learning_rate": 5.624076957640592e-05, "loss": 1.0101, "step": 6353 }, { "epoch": 0.65, "learning_rate": 5.6239609575134044e-05, "loss": 0.111, "step": 6354 }, { "epoch": 0.65, "learning_rate": 5.623844940688328e-05, "loss": 1.0517, "step": 6355 }, { "epoch": 0.65, "learning_rate": 5.6237289071661014e-05, "loss": 0.5678, "step": 6356 }, { "epoch": 0.65, "learning_rate": 5.623612856947461e-05, "loss": 0.3213, "step": 6357 }, { "epoch": 0.65, "learning_rate": 5.623496790033147e-05, "loss": 0.8694, "step": 6358 }, { "epoch": 0.65, "learning_rate": 5.623380706423897e-05, "loss": 1.3078, "step": 6359 }, { "epoch": 0.65, "learning_rate": 5.62326460612045e-05, "loss": 1.3102, "step": 6360 }, { "epoch": 0.65, "learning_rate": 5.623148489123545e-05, "loss": 1.0477, "step": 6361 }, { "epoch": 0.65, "learning_rate": 5.623032355433922e-05, "loss": 1.2834, "step": 6362 }, { "epoch": 0.65, "learning_rate": 5.622916205052318e-05, "loss": 0.9289, "step": 6363 }, { "epoch": 0.65, "learning_rate": 5.622800037979473e-05, "loss": 1.4863, "step": 6364 }, { "epoch": 0.65, "learning_rate": 5.622683854216127e-05, "loss": 0.3838, "step": 6365 }, { "epoch": 0.65, "learning_rate": 5.622567653763018e-05, "loss": 1.7641, "step": 6366 }, { "epoch": 0.65, "learning_rate": 5.622451436620886e-05, "loss": 1.7018, "step": 6367 }, { "epoch": 0.65, "learning_rate": 5.622335202790472e-05, "loss": 1.6582, "step": 6368 }, { "epoch": 0.65, "learning_rate": 5.6222189522725135e-05, "loss": 1.0784, "step": 6369 }, { "epoch": 0.65, "learning_rate": 5.622102685067751e-05, "loss": 0.9793, "step": 6370 }, { "epoch": 0.65, "learning_rate": 5.621986401176923e-05, "loss": 1.542, "step": 6371 }, { "epoch": 0.65, "learning_rate": 5.6218701006007734e-05, "loss": 0.284, "step": 6372 }, { "epoch": 0.65, "learning_rate": 5.621753783340039e-05, "loss": 0.6686, "step": 6373 }, { "epoch": 0.65, "learning_rate": 5.621637449395461e-05, "loss": 1.4099, "step": 6374 }, { "epoch": 0.65, "learning_rate": 5.621521098767779e-05, "loss": 0.8683, "step": 6375 }, { "epoch": 0.65, "learning_rate": 5.621404731457735e-05, "loss": 0.6836, "step": 6376 }, { "epoch": 0.65, "learning_rate": 5.621288347466068e-05, "loss": 1.5801, "step": 6377 }, { "epoch": 0.65, "learning_rate": 5.6211719467935194e-05, "loss": 1.3176, "step": 6378 }, { "epoch": 0.65, "learning_rate": 5.621055529440829e-05, "loss": 1.9497, "step": 6379 }, { "epoch": 0.65, "learning_rate": 5.620939095408739e-05, "loss": 1.5606, "step": 6380 }, { "epoch": 0.65, "learning_rate": 5.6208226446979896e-05, "loss": 1.6519, "step": 6381 }, { "epoch": 0.65, "learning_rate": 5.620706177309321e-05, "loss": 1.4848, "step": 6382 }, { "epoch": 0.65, "learning_rate": 5.6205896932434765e-05, "loss": 1.589, "step": 6383 }, { "epoch": 0.65, "learning_rate": 5.620473192501196e-05, "loss": 1.7186, "step": 6384 }, { "epoch": 0.65, "learning_rate": 5.62035667508322e-05, "loss": 1.058, "step": 6385 }, { "epoch": 0.65, "learning_rate": 5.6202401409902925e-05, "loss": 1.2388, "step": 6386 }, { "epoch": 0.65, "learning_rate": 5.6201235902231525e-05, "loss": 1.3671, "step": 6387 }, { "epoch": 0.65, "learning_rate": 5.6200070227825424e-05, "loss": 1.7794, "step": 6388 }, { "epoch": 0.65, "learning_rate": 5.619890438669205e-05, "loss": 0.9476, "step": 6389 }, { "epoch": 0.65, "learning_rate": 5.61977383788388e-05, "loss": 1.1631, "step": 6390 }, { "epoch": 0.65, "learning_rate": 5.619657220427313e-05, "loss": 1.0551, "step": 6391 }, { "epoch": 0.65, "learning_rate": 5.619540586300243e-05, "loss": 0.8781, "step": 6392 }, { "epoch": 0.65, "learning_rate": 5.6194239355034136e-05, "loss": 0.8164, "step": 6393 }, { "epoch": 0.65, "learning_rate": 5.619307268037566e-05, "loss": 1.2482, "step": 6394 }, { "epoch": 0.65, "learning_rate": 5.619190583903444e-05, "loss": 1.3014, "step": 6395 }, { "epoch": 0.65, "learning_rate": 5.619073883101789e-05, "loss": 0.2435, "step": 6396 }, { "epoch": 0.65, "learning_rate": 5.6189571656333456e-05, "loss": 0.4683, "step": 6397 }, { "epoch": 0.65, "learning_rate": 5.618840431498854e-05, "loss": 1.4003, "step": 6398 }, { "epoch": 0.65, "learning_rate": 5.618723680699058e-05, "loss": 1.537, "step": 6399 }, { "epoch": 0.65, "learning_rate": 5.618606913234701e-05, "loss": 1.1421, "step": 6400 }, { "epoch": 0.65, "learning_rate": 5.6184901291065255e-05, "loss": 1.2, "step": 6401 }, { "epoch": 0.65, "learning_rate": 5.618373328315275e-05, "loss": 0.7342, "step": 6402 }, { "epoch": 0.65, "learning_rate": 5.6182565108616936e-05, "loss": 1.1436, "step": 6403 }, { "epoch": 0.65, "learning_rate": 5.618139676746523e-05, "loss": 1.5893, "step": 6404 }, { "epoch": 0.65, "learning_rate": 5.618022825970507e-05, "loss": 0.6623, "step": 6405 }, { "epoch": 0.65, "learning_rate": 5.61790595853439e-05, "loss": 1.3067, "step": 6406 }, { "epoch": 0.65, "learning_rate": 5.6177890744389165e-05, "loss": 0.5241, "step": 6407 }, { "epoch": 0.65, "learning_rate": 5.6176721736848284e-05, "loss": 1.4501, "step": 6408 }, { "epoch": 0.65, "learning_rate": 5.61755525627287e-05, "loss": 0.6225, "step": 6409 }, { "epoch": 0.65, "learning_rate": 5.617438322203786e-05, "loss": 1.4515, "step": 6410 }, { "epoch": 0.65, "learning_rate": 5.61732137147832e-05, "loss": 1.5321, "step": 6411 }, { "epoch": 0.65, "learning_rate": 5.6172044040972173e-05, "loss": 0.9542, "step": 6412 }, { "epoch": 0.65, "learning_rate": 5.61708742006122e-05, "loss": 1.0295, "step": 6413 }, { "epoch": 0.65, "learning_rate": 5.616970419371075e-05, "loss": 0.2409, "step": 6414 }, { "epoch": 0.65, "learning_rate": 5.616853402027526e-05, "loss": 0.8066, "step": 6415 }, { "epoch": 0.65, "learning_rate": 5.6167363680313166e-05, "loss": 0.4539, "step": 6416 }, { "epoch": 0.65, "learning_rate": 5.6166193173831935e-05, "loss": 0.9763, "step": 6417 }, { "epoch": 0.65, "learning_rate": 5.6165022500838996e-05, "loss": 1.9294, "step": 6418 }, { "epoch": 0.65, "learning_rate": 5.6163851661341816e-05, "loss": 1.6288, "step": 6419 }, { "epoch": 0.65, "learning_rate": 5.616268065534783e-05, "loss": 1.2014, "step": 6420 }, { "epoch": 0.65, "learning_rate": 5.616150948286449e-05, "loss": 0.4049, "step": 6421 }, { "epoch": 0.65, "learning_rate": 5.6160338143899265e-05, "loss": 1.455, "step": 6422 }, { "epoch": 0.65, "learning_rate": 5.6159166638459605e-05, "loss": 0.5951, "step": 6423 }, { "epoch": 0.65, "learning_rate": 5.615799496655295e-05, "loss": 0.7681, "step": 6424 }, { "epoch": 0.65, "learning_rate": 5.615682312818677e-05, "loss": 0.203, "step": 6425 }, { "epoch": 0.65, "learning_rate": 5.615565112336852e-05, "loss": 1.3271, "step": 6426 }, { "epoch": 0.65, "learning_rate": 5.6154478952105655e-05, "loss": 0.8279, "step": 6427 }, { "epoch": 0.65, "learning_rate": 5.615330661440563e-05, "loss": 1.537, "step": 6428 }, { "epoch": 0.65, "learning_rate": 5.6152134110275924e-05, "loss": 0.8629, "step": 6429 }, { "epoch": 0.65, "learning_rate": 5.615096143972398e-05, "loss": 0.9861, "step": 6430 }, { "epoch": 0.65, "learning_rate": 5.614978860275726e-05, "loss": 0.6395, "step": 6431 }, { "epoch": 0.65, "learning_rate": 5.6148615599383235e-05, "loss": 0.3561, "step": 6432 }, { "epoch": 0.65, "learning_rate": 5.614744242960937e-05, "loss": 0.7941, "step": 6433 }, { "epoch": 0.65, "learning_rate": 5.6146269093443126e-05, "loss": 0.7337, "step": 6434 }, { "epoch": 0.65, "learning_rate": 5.614509559089198e-05, "loss": 1.4709, "step": 6435 }, { "epoch": 0.65, "learning_rate": 5.614392192196338e-05, "loss": 1.2375, "step": 6436 }, { "epoch": 0.65, "learning_rate": 5.614274808666481e-05, "loss": 0.773, "step": 6437 }, { "epoch": 0.65, "learning_rate": 5.614157408500374e-05, "loss": 1.3818, "step": 6438 }, { "epoch": 0.65, "learning_rate": 5.6140399916987636e-05, "loss": 0.4316, "step": 6439 }, { "epoch": 0.65, "learning_rate": 5.613922558262398e-05, "loss": 0.3345, "step": 6440 }, { "epoch": 0.65, "learning_rate": 5.613805108192022e-05, "loss": 1.4309, "step": 6441 }, { "epoch": 0.65, "learning_rate": 5.6136876414883855e-05, "loss": 1.2788, "step": 6442 }, { "epoch": 0.65, "learning_rate": 5.613570158152235e-05, "loss": 1.1755, "step": 6443 }, { "epoch": 0.65, "learning_rate": 5.613452658184318e-05, "loss": 0.7149, "step": 6444 }, { "epoch": 0.65, "learning_rate": 5.6133351415853834e-05, "loss": 0.4039, "step": 6445 }, { "epoch": 0.65, "learning_rate": 5.613217608356177e-05, "loss": 0.5002, "step": 6446 }, { "epoch": 0.65, "learning_rate": 5.613100058497448e-05, "loss": 1.2786, "step": 6447 }, { "epoch": 0.65, "learning_rate": 5.612982492009945e-05, "loss": 0.4081, "step": 6448 }, { "epoch": 0.65, "learning_rate": 5.6128649088944156e-05, "loss": 0.6751, "step": 6449 }, { "epoch": 0.65, "learning_rate": 5.612747309151606e-05, "loss": 0.7339, "step": 6450 }, { "epoch": 0.66, "learning_rate": 5.612629692782269e-05, "loss": 1.1287, "step": 6451 }, { "epoch": 0.66, "learning_rate": 5.612512059787149e-05, "loss": 0.8674, "step": 6452 }, { "epoch": 0.66, "learning_rate": 5.612394410166997e-05, "loss": 1.1727, "step": 6453 }, { "epoch": 0.66, "learning_rate": 5.6122767439225595e-05, "loss": 0.7976, "step": 6454 }, { "epoch": 0.66, "learning_rate": 5.6121590610545876e-05, "loss": 0.7385, "step": 6455 }, { "epoch": 0.66, "learning_rate": 5.6120413615638286e-05, "loss": 0.8185, "step": 6456 }, { "epoch": 0.66, "learning_rate": 5.6119236454510325e-05, "loss": 0.4082, "step": 6457 }, { "epoch": 0.66, "learning_rate": 5.611805912716948e-05, "loss": 1.1989, "step": 6458 }, { "epoch": 0.66, "learning_rate": 5.611688163362324e-05, "loss": 0.4916, "step": 6459 }, { "epoch": 0.66, "learning_rate": 5.611570397387911e-05, "loss": 1.145, "step": 6460 }, { "epoch": 0.66, "learning_rate": 5.611452614794456e-05, "loss": 0.5928, "step": 6461 }, { "epoch": 0.66, "learning_rate": 5.61133481558271e-05, "loss": 1.0088, "step": 6462 }, { "epoch": 0.66, "learning_rate": 5.611216999753424e-05, "loss": 1.0074, "step": 6463 }, { "epoch": 0.66, "learning_rate": 5.611099167307345e-05, "loss": 0.4567, "step": 6464 }, { "epoch": 0.66, "learning_rate": 5.610981318245225e-05, "loss": 1.2913, "step": 6465 }, { "epoch": 0.66, "learning_rate": 5.6108634525678134e-05, "loss": 1.2668, "step": 6466 }, { "epoch": 0.66, "learning_rate": 5.610745570275859e-05, "loss": 0.1553, "step": 6467 }, { "epoch": 0.66, "learning_rate": 5.610627671370114e-05, "loss": 0.841, "step": 6468 }, { "epoch": 0.66, "learning_rate": 5.610509755851327e-05, "loss": 1.2989, "step": 6469 }, { "epoch": 0.66, "learning_rate": 5.6103918237202496e-05, "loss": 0.5727, "step": 6470 }, { "epoch": 0.66, "learning_rate": 5.610273874977631e-05, "loss": 0.1386, "step": 6471 }, { "epoch": 0.66, "learning_rate": 5.610155909624223e-05, "loss": 0.7813, "step": 6472 }, { "epoch": 0.66, "learning_rate": 5.610037927660775e-05, "loss": 1.296, "step": 6473 }, { "epoch": 0.66, "learning_rate": 5.609919929088039e-05, "loss": 0.4123, "step": 6474 }, { "epoch": 0.66, "learning_rate": 5.6098019139067654e-05, "loss": 0.5474, "step": 6475 }, { "epoch": 0.66, "learning_rate": 5.6096838821177055e-05, "loss": 0.4822, "step": 6476 }, { "epoch": 0.66, "learning_rate": 5.60956583372161e-05, "loss": 0.8494, "step": 6477 }, { "epoch": 0.66, "learning_rate": 5.60944776871923e-05, "loss": 1.192, "step": 6478 }, { "epoch": 0.66, "learning_rate": 5.609329687111317e-05, "loss": 0.4175, "step": 6479 }, { "epoch": 0.66, "learning_rate": 5.609211588898623e-05, "loss": 0.4691, "step": 6480 }, { "epoch": 0.66, "learning_rate": 5.6090934740818995e-05, "loss": 0.6943, "step": 6481 }, { "epoch": 0.66, "learning_rate": 5.608975342661896e-05, "loss": 0.3324, "step": 6482 }, { "epoch": 0.66, "learning_rate": 5.608857194639368e-05, "loss": 1.3532, "step": 6483 }, { "epoch": 0.66, "learning_rate": 5.608739030015064e-05, "loss": 1.1129, "step": 6484 }, { "epoch": 0.66, "learning_rate": 5.6086208487897375e-05, "loss": 1.0825, "step": 6485 }, { "epoch": 0.66, "learning_rate": 5.60850265096414e-05, "loss": 1.1576, "step": 6486 }, { "epoch": 0.66, "learning_rate": 5.608384436539024e-05, "loss": 1.1193, "step": 6487 }, { "epoch": 0.66, "learning_rate": 5.608266205515142e-05, "loss": 1.0857, "step": 6488 }, { "epoch": 0.66, "learning_rate": 5.608147957893246e-05, "loss": 0.659, "step": 6489 }, { "epoch": 0.66, "learning_rate": 5.608029693674088e-05, "loss": 1.0954, "step": 6490 }, { "epoch": 0.66, "learning_rate": 5.6079114128584224e-05, "loss": 0.449, "step": 6491 }, { "epoch": 0.66, "learning_rate": 5.6077931154469996e-05, "loss": 1.4029, "step": 6492 }, { "epoch": 0.66, "learning_rate": 5.607674801440574e-05, "loss": 1.4294, "step": 6493 }, { "epoch": 0.66, "learning_rate": 5.607556470839898e-05, "loss": 1.2586, "step": 6494 }, { "epoch": 0.66, "learning_rate": 5.607438123645725e-05, "loss": 0.6851, "step": 6495 }, { "epoch": 0.66, "learning_rate": 5.607319759858807e-05, "loss": 0.5736, "step": 6496 }, { "epoch": 0.66, "learning_rate": 5.607201379479898e-05, "loss": 0.4408, "step": 6497 }, { "epoch": 0.66, "learning_rate": 5.607082982509751e-05, "loss": 1.5142, "step": 6498 }, { "epoch": 0.66, "learning_rate": 5.60696456894912e-05, "loss": 0.6696, "step": 6499 }, { "epoch": 0.66, "learning_rate": 5.606846138798758e-05, "loss": 0.2442, "step": 6500 }, { "epoch": 0.66, "learning_rate": 5.60672769205942e-05, "loss": 1.1904, "step": 6501 }, { "epoch": 0.66, "learning_rate": 5.6066092287318576e-05, "loss": 0.499, "step": 6502 }, { "epoch": 0.66, "learning_rate": 5.606490748816826e-05, "loss": 1.1361, "step": 6503 }, { "epoch": 0.66, "learning_rate": 5.606372252315079e-05, "loss": 0.1875, "step": 6504 }, { "epoch": 0.66, "learning_rate": 5.606253739227369e-05, "loss": 0.371, "step": 6505 }, { "epoch": 0.66, "learning_rate": 5.606135209554454e-05, "loss": 0.9752, "step": 6506 }, { "epoch": 0.66, "learning_rate": 5.6060166632970845e-05, "loss": 1.1526, "step": 6507 }, { "epoch": 0.66, "learning_rate": 5.605898100456016e-05, "loss": 0.9699, "step": 6508 }, { "epoch": 0.66, "learning_rate": 5.605779521032004e-05, "loss": 0.2661, "step": 6509 }, { "epoch": 0.66, "learning_rate": 5.6056609250258025e-05, "loss": 0.5471, "step": 6510 }, { "epoch": 0.66, "learning_rate": 5.6055423124381654e-05, "loss": 1.2119, "step": 6511 }, { "epoch": 0.66, "learning_rate": 5.605423683269849e-05, "loss": 1.3029, "step": 6512 }, { "epoch": 0.66, "learning_rate": 5.605305037521608e-05, "loss": 1.4082, "step": 6513 }, { "epoch": 0.66, "learning_rate": 5.605186375194196e-05, "loss": 1.1538, "step": 6514 }, { "epoch": 0.66, "learning_rate": 5.6050676962883686e-05, "loss": 0.667, "step": 6515 }, { "epoch": 0.66, "learning_rate": 5.604949000804882e-05, "loss": 0.5963, "step": 6516 }, { "epoch": 0.66, "learning_rate": 5.604830288744491e-05, "loss": 0.6737, "step": 6517 }, { "epoch": 0.66, "learning_rate": 5.60471156010795e-05, "loss": 0.892, "step": 6518 }, { "epoch": 0.66, "learning_rate": 5.6045928148960164e-05, "loss": 1.2642, "step": 6519 }, { "epoch": 0.66, "learning_rate": 5.6044740531094454e-05, "loss": 0.3753, "step": 6520 }, { "epoch": 0.66, "learning_rate": 5.6043552747489915e-05, "loss": 1.2083, "step": 6521 }, { "epoch": 0.66, "learning_rate": 5.604236479815412e-05, "loss": 0.4105, "step": 6522 }, { "epoch": 0.66, "learning_rate": 5.6041176683094615e-05, "loss": 1.3277, "step": 6523 }, { "epoch": 0.66, "learning_rate": 5.603998840231897e-05, "loss": 0.8825, "step": 6524 }, { "epoch": 0.66, "learning_rate": 5.603879995583474e-05, "loss": 1.7342, "step": 6525 }, { "epoch": 0.66, "learning_rate": 5.603761134364951e-05, "loss": 0.7554, "step": 6526 }, { "epoch": 0.66, "learning_rate": 5.603642256577081e-05, "loss": 0.6979, "step": 6527 }, { "epoch": 0.66, "learning_rate": 5.6035233622206224e-05, "loss": 0.7239, "step": 6528 }, { "epoch": 0.66, "learning_rate": 5.603404451296333e-05, "loss": 0.9194, "step": 6529 }, { "epoch": 0.66, "learning_rate": 5.6032855238049666e-05, "loss": 2.2672, "step": 6530 }, { "epoch": 0.66, "learning_rate": 5.603166579747282e-05, "loss": 1.5719, "step": 6531 }, { "epoch": 0.66, "learning_rate": 5.603047619124036e-05, "loss": 0.9453, "step": 6532 }, { "epoch": 0.66, "learning_rate": 5.602928641935984e-05, "loss": 1.0777, "step": 6533 }, { "epoch": 0.66, "learning_rate": 5.602809648183885e-05, "loss": 1.4784, "step": 6534 }, { "epoch": 0.66, "learning_rate": 5.602690637868496e-05, "loss": 0.4299, "step": 6535 }, { "epoch": 0.66, "learning_rate": 5.602571610990573e-05, "loss": 1.3521, "step": 6536 }, { "epoch": 0.66, "learning_rate": 5.602452567550875e-05, "loss": 1.0249, "step": 6537 }, { "epoch": 0.66, "learning_rate": 5.602333507550158e-05, "loss": 1.6518, "step": 6538 }, { "epoch": 0.66, "learning_rate": 5.602214430989181e-05, "loss": 2.4295, "step": 6539 }, { "epoch": 0.66, "learning_rate": 5.602095337868702e-05, "loss": 1.3493, "step": 6540 }, { "epoch": 0.66, "learning_rate": 5.601976228189477e-05, "loss": 1.7021, "step": 6541 }, { "epoch": 0.66, "learning_rate": 5.6018571019522666e-05, "loss": 0.3882, "step": 6542 }, { "epoch": 0.66, "learning_rate": 5.601737959157826e-05, "loss": 1.1686, "step": 6543 }, { "epoch": 0.66, "learning_rate": 5.601618799806915e-05, "loss": 0.8255, "step": 6544 }, { "epoch": 0.66, "learning_rate": 5.6014996239002923e-05, "loss": 0.7266, "step": 6545 }, { "epoch": 0.66, "learning_rate": 5.6013804314387154e-05, "loss": 0.2213, "step": 6546 }, { "epoch": 0.66, "learning_rate": 5.6012612224229434e-05, "loss": 0.9447, "step": 6547 }, { "epoch": 0.66, "learning_rate": 5.601141996853734e-05, "loss": 0.8943, "step": 6548 }, { "epoch": 0.67, "learning_rate": 5.601022754731847e-05, "loss": 0.6118, "step": 6549 }, { "epoch": 0.67, "learning_rate": 5.6009034960580395e-05, "loss": 0.5461, "step": 6550 }, { "epoch": 0.67, "learning_rate": 5.6007842208330726e-05, "loss": 1.0844, "step": 6551 }, { "epoch": 0.67, "learning_rate": 5.600664929057704e-05, "loss": 1.6315, "step": 6552 }, { "epoch": 0.67, "learning_rate": 5.600545620732693e-05, "loss": 0.8147, "step": 6553 }, { "epoch": 0.67, "learning_rate": 5.600426295858799e-05, "loss": 0.7028, "step": 6554 }, { "epoch": 0.67, "learning_rate": 5.600306954436782e-05, "loss": 0.6697, "step": 6555 }, { "epoch": 0.67, "learning_rate": 5.600187596467401e-05, "loss": 0.6428, "step": 6556 }, { "epoch": 0.67, "learning_rate": 5.6000682219514136e-05, "loss": 1.8998, "step": 6557 }, { "epoch": 0.67, "learning_rate": 5.599948830889582e-05, "loss": 0.7806, "step": 6558 }, { "epoch": 0.67, "learning_rate": 5.599829423282665e-05, "loss": 1.0993, "step": 6559 }, { "epoch": 0.67, "learning_rate": 5.5997099991314226e-05, "loss": 0.9135, "step": 6560 }, { "epoch": 0.67, "learning_rate": 5.599590558436615e-05, "loss": 0.5895, "step": 6561 }, { "epoch": 0.67, "learning_rate": 5.599471101199002e-05, "loss": 0.284, "step": 6562 }, { "epoch": 0.67, "learning_rate": 5.599351627419344e-05, "loss": 0.3841, "step": 6563 }, { "epoch": 0.67, "learning_rate": 5.5992321370984005e-05, "loss": 0.9825, "step": 6564 }, { "epoch": 0.67, "learning_rate": 5.599112630236933e-05, "loss": 0.7597, "step": 6565 }, { "epoch": 0.67, "learning_rate": 5.598993106835701e-05, "loss": 1.4918, "step": 6566 }, { "epoch": 0.67, "learning_rate": 5.598873566895466e-05, "loss": 0.398, "step": 6567 }, { "epoch": 0.67, "learning_rate": 5.5987540104169875e-05, "loss": 1.1451, "step": 6568 }, { "epoch": 0.67, "learning_rate": 5.598634437401029e-05, "loss": 1.0337, "step": 6569 }, { "epoch": 0.67, "learning_rate": 5.5985148478483476e-05, "loss": 1.181, "step": 6570 }, { "epoch": 0.67, "learning_rate": 5.598395241759707e-05, "loss": 1.5482, "step": 6571 }, { "epoch": 0.67, "learning_rate": 5.5982756191358677e-05, "loss": 0.725, "step": 6572 }, { "epoch": 0.67, "learning_rate": 5.59815597997759e-05, "loss": 1.1179, "step": 6573 }, { "epoch": 0.67, "learning_rate": 5.598036324285637e-05, "loss": 1.2265, "step": 6574 }, { "epoch": 0.67, "learning_rate": 5.597916652060769e-05, "loss": 0.4688, "step": 6575 }, { "epoch": 0.67, "learning_rate": 5.597796963303747e-05, "loss": 0.1925, "step": 6576 }, { "epoch": 0.67, "learning_rate": 5.5976772580153336e-05, "loss": 0.3375, "step": 6577 }, { "epoch": 0.67, "learning_rate": 5.597557536196291e-05, "loss": 0.4067, "step": 6578 }, { "epoch": 0.67, "learning_rate": 5.59743779784738e-05, "loss": 0.4389, "step": 6579 }, { "epoch": 0.67, "learning_rate": 5.597318042969363e-05, "loss": 1.5486, "step": 6580 }, { "epoch": 0.67, "learning_rate": 5.597198271563002e-05, "loss": 0.3599, "step": 6581 }, { "epoch": 0.67, "learning_rate": 5.597078483629059e-05, "loss": 0.3729, "step": 6582 }, { "epoch": 0.67, "learning_rate": 5.5969586791682965e-05, "loss": 1.0455, "step": 6583 }, { "epoch": 0.67, "learning_rate": 5.596838858181477e-05, "loss": 1.1502, "step": 6584 }, { "epoch": 0.67, "learning_rate": 5.596719020669363e-05, "loss": 0.4308, "step": 6585 }, { "epoch": 0.67, "learning_rate": 5.596599166632718e-05, "loss": 1.0201, "step": 6586 }, { "epoch": 0.67, "learning_rate": 5.5964792960723025e-05, "loss": 1.1048, "step": 6587 }, { "epoch": 0.67, "learning_rate": 5.59635940898888e-05, "loss": 1.1499, "step": 6588 }, { "epoch": 0.67, "learning_rate": 5.596239505383215e-05, "loss": 0.9997, "step": 6589 }, { "epoch": 0.67, "learning_rate": 5.5961195852560686e-05, "loss": 0.6846, "step": 6590 }, { "epoch": 0.67, "learning_rate": 5.595999648608205e-05, "loss": 0.9657, "step": 6591 }, { "epoch": 0.67, "learning_rate": 5.595879695440388e-05, "loss": 0.3483, "step": 6592 }, { "epoch": 0.67, "learning_rate": 5.5957597257533794e-05, "loss": 0.5793, "step": 6593 }, { "epoch": 0.67, "learning_rate": 5.5956397395479436e-05, "loss": 0.239, "step": 6594 }, { "epoch": 0.67, "learning_rate": 5.5955197368248435e-05, "loss": 0.8349, "step": 6595 }, { "epoch": 0.67, "learning_rate": 5.595399717584844e-05, "loss": 0.7325, "step": 6596 }, { "epoch": 0.67, "learning_rate": 5.595279681828708e-05, "loss": 0.6636, "step": 6597 }, { "epoch": 0.67, "learning_rate": 5.595159629557199e-05, "loss": 0.426, "step": 6598 }, { "epoch": 0.67, "learning_rate": 5.5950395607710813e-05, "loss": 0.4585, "step": 6599 }, { "epoch": 0.67, "learning_rate": 5.5949194754711194e-05, "loss": 0.6015, "step": 6600 }, { "epoch": 0.67, "learning_rate": 5.5947993736580774e-05, "loss": 0.5082, "step": 6601 }, { "epoch": 0.67, "learning_rate": 5.594679255332718e-05, "loss": 1.1892, "step": 6602 }, { "epoch": 0.67, "learning_rate": 5.594559120495808e-05, "loss": 0.2274, "step": 6603 }, { "epoch": 0.67, "learning_rate": 5.59443896914811e-05, "loss": 0.6143, "step": 6604 }, { "epoch": 0.67, "learning_rate": 5.59431880129039e-05, "loss": 0.4675, "step": 6605 }, { "epoch": 0.67, "learning_rate": 5.5941986169234116e-05, "loss": 1.5255, "step": 6606 }, { "epoch": 0.67, "learning_rate": 5.5940784160479405e-05, "loss": 0.5801, "step": 6607 }, { "epoch": 0.67, "learning_rate": 5.593958198664741e-05, "loss": 1.1085, "step": 6608 }, { "epoch": 0.67, "learning_rate": 5.593837964774578e-05, "loss": 0.799, "step": 6609 }, { "epoch": 0.67, "learning_rate": 5.593717714378218e-05, "loss": 0.8621, "step": 6610 }, { "epoch": 0.67, "learning_rate": 5.593597447476425e-05, "loss": 0.2916, "step": 6611 }, { "epoch": 0.67, "learning_rate": 5.593477164069963e-05, "loss": 1.2889, "step": 6612 }, { "epoch": 0.67, "learning_rate": 5.5933568641596e-05, "loss": 0.6115, "step": 6613 }, { "epoch": 0.67, "learning_rate": 5.593236547746101e-05, "loss": 0.8539, "step": 6614 }, { "epoch": 0.67, "learning_rate": 5.59311621483023e-05, "loss": 1.0, "step": 6615 }, { "epoch": 0.67, "learning_rate": 5.592995865412754e-05, "loss": 1.3169, "step": 6616 }, { "epoch": 0.67, "learning_rate": 5.59287549949444e-05, "loss": 1.4179, "step": 6617 }, { "epoch": 0.67, "learning_rate": 5.5927551170760516e-05, "loss": 1.2758, "step": 6618 }, { "epoch": 0.67, "learning_rate": 5.592634718158356e-05, "loss": 1.1877, "step": 6619 }, { "epoch": 0.67, "learning_rate": 5.5925143027421196e-05, "loss": 0.6576, "step": 6620 }, { "epoch": 0.67, "learning_rate": 5.592393870828108e-05, "loss": 0.3708, "step": 6621 }, { "epoch": 0.67, "learning_rate": 5.592273422417089e-05, "loss": 1.258, "step": 6622 }, { "epoch": 0.67, "learning_rate": 5.5921529575098274e-05, "loss": 1.5016, "step": 6623 }, { "epoch": 0.67, "learning_rate": 5.59203247610709e-05, "loss": 0.4456, "step": 6624 }, { "epoch": 0.67, "learning_rate": 5.5919119782096444e-05, "loss": 1.0385, "step": 6625 }, { "epoch": 0.67, "learning_rate": 5.5917914638182575e-05, "loss": 0.7887, "step": 6626 }, { "epoch": 0.67, "learning_rate": 5.591670932933696e-05, "loss": 1.0313, "step": 6627 }, { "epoch": 0.67, "learning_rate": 5.5915503855567257e-05, "loss": 1.2183, "step": 6628 }, { "epoch": 0.67, "learning_rate": 5.591429821688115e-05, "loss": 0.7771, "step": 6629 }, { "epoch": 0.67, "learning_rate": 5.5913092413286304e-05, "loss": 0.7827, "step": 6630 }, { "epoch": 0.67, "learning_rate": 5.5911886444790395e-05, "loss": 0.6964, "step": 6631 }, { "epoch": 0.67, "learning_rate": 5.5910680311401104e-05, "loss": 0.4145, "step": 6632 }, { "epoch": 0.67, "learning_rate": 5.5909474013126096e-05, "loss": 0.573, "step": 6633 }, { "epoch": 0.67, "learning_rate": 5.5908267549973056e-05, "loss": 0.4061, "step": 6634 }, { "epoch": 0.67, "learning_rate": 5.590706092194965e-05, "loss": 0.2768, "step": 6635 }, { "epoch": 0.67, "learning_rate": 5.590585412906357e-05, "loss": 0.8308, "step": 6636 }, { "epoch": 0.67, "learning_rate": 5.5904647171322485e-05, "loss": 0.8467, "step": 6637 }, { "epoch": 0.67, "learning_rate": 5.590344004873408e-05, "loss": 1.3283, "step": 6638 }, { "epoch": 0.67, "learning_rate": 5.590223276130604e-05, "loss": 1.0815, "step": 6639 }, { "epoch": 0.67, "learning_rate": 5.590102530904604e-05, "loss": 0.7376, "step": 6640 }, { "epoch": 0.67, "learning_rate": 5.589981769196178e-05, "loss": 1.0255, "step": 6641 }, { "epoch": 0.67, "learning_rate": 5.5898609910060925e-05, "loss": 0.3677, "step": 6642 }, { "epoch": 0.67, "learning_rate": 5.5897401963351165e-05, "loss": 0.2478, "step": 6643 }, { "epoch": 0.67, "learning_rate": 5.5896193851840194e-05, "loss": 0.5134, "step": 6644 }, { "epoch": 0.67, "learning_rate": 5.5894985575535704e-05, "loss": 0.6294, "step": 6645 }, { "epoch": 0.67, "learning_rate": 5.5893777134445366e-05, "loss": 0.5355, "step": 6646 }, { "epoch": 0.67, "learning_rate": 5.5892568528576884e-05, "loss": 0.5519, "step": 6647 }, { "epoch": 0.68, "learning_rate": 5.589135975793795e-05, "loss": 1.1547, "step": 6648 }, { "epoch": 0.68, "learning_rate": 5.5890150822536245e-05, "loss": 0.6924, "step": 6649 }, { "epoch": 0.68, "learning_rate": 5.588894172237948e-05, "loss": 0.6212, "step": 6650 }, { "epoch": 0.68, "learning_rate": 5.588773245747532e-05, "loss": 1.8063, "step": 6651 }, { "epoch": 0.68, "learning_rate": 5.58865230278315e-05, "loss": 0.3519, "step": 6652 }, { "epoch": 0.68, "learning_rate": 5.588531343345568e-05, "loss": 1.265, "step": 6653 }, { "epoch": 0.68, "learning_rate": 5.588410367435558e-05, "loss": 1.6286, "step": 6654 }, { "epoch": 0.68, "learning_rate": 5.588289375053889e-05, "loss": 1.4976, "step": 6655 }, { "epoch": 0.68, "learning_rate": 5.588168366201331e-05, "loss": 1.3441, "step": 6656 }, { "epoch": 0.68, "learning_rate": 5.588047340878654e-05, "loss": 1.5602, "step": 6657 }, { "epoch": 0.68, "learning_rate": 5.587926299086628e-05, "loss": 0.5258, "step": 6658 }, { "epoch": 0.68, "learning_rate": 5.587805240826024e-05, "loss": 0.035, "step": 6659 }, { "epoch": 0.68, "learning_rate": 5.5876841660976115e-05, "loss": 0.6742, "step": 6660 }, { "epoch": 0.68, "learning_rate": 5.587563074902162e-05, "loss": 0.6306, "step": 6661 }, { "epoch": 0.68, "learning_rate": 5.5874419672404454e-05, "loss": 0.9669, "step": 6662 }, { "epoch": 0.68, "learning_rate": 5.5873208431132324e-05, "loss": 0.5492, "step": 6663 }, { "epoch": 0.68, "learning_rate": 5.587199702521293e-05, "loss": 0.9843, "step": 6664 }, { "epoch": 0.68, "learning_rate": 5.5870785454654e-05, "loss": 0.3885, "step": 6665 }, { "epoch": 0.68, "learning_rate": 5.586957371946323e-05, "loss": 1.1536, "step": 6666 }, { "epoch": 0.68, "learning_rate": 5.5868361819648327e-05, "loss": 0.5456, "step": 6667 }, { "epoch": 0.68, "learning_rate": 5.586714975521701e-05, "loss": 1.7256, "step": 6668 }, { "epoch": 0.68, "learning_rate": 5.5865937526177e-05, "loss": 0.4804, "step": 6669 }, { "epoch": 0.68, "learning_rate": 5.5864725132536e-05, "loss": 1.5284, "step": 6670 }, { "epoch": 0.68, "learning_rate": 5.586351257430173e-05, "loss": 0.6295, "step": 6671 }, { "epoch": 0.68, "learning_rate": 5.586229985148191e-05, "loss": 0.3191, "step": 6672 }, { "epoch": 0.68, "learning_rate": 5.586108696408424e-05, "loss": 1.6732, "step": 6673 }, { "epoch": 0.68, "learning_rate": 5.585987391211646e-05, "loss": 1.4483, "step": 6674 }, { "epoch": 0.68, "learning_rate": 5.585866069558628e-05, "loss": 0.3998, "step": 6675 }, { "epoch": 0.68, "learning_rate": 5.5857447314501416e-05, "loss": 0.3373, "step": 6676 }, { "epoch": 0.68, "learning_rate": 5.58562337688696e-05, "loss": 1.1602, "step": 6677 }, { "epoch": 0.68, "learning_rate": 5.585502005869854e-05, "loss": 1.1035, "step": 6678 }, { "epoch": 0.68, "learning_rate": 5.585380618399597e-05, "loss": 1.3542, "step": 6679 }, { "epoch": 0.68, "learning_rate": 5.5852592144769626e-05, "loss": 1.0772, "step": 6680 }, { "epoch": 0.68, "learning_rate": 5.5851377941027206e-05, "loss": 0.7515, "step": 6681 }, { "epoch": 0.68, "learning_rate": 5.5850163572776454e-05, "loss": 0.9162, "step": 6682 }, { "epoch": 0.68, "learning_rate": 5.584894904002509e-05, "loss": 0.848, "step": 6683 }, { "epoch": 0.68, "learning_rate": 5.584773434278086e-05, "loss": 1.1858, "step": 6684 }, { "epoch": 0.68, "learning_rate": 5.584651948105147e-05, "loss": 0.3852, "step": 6685 }, { "epoch": 0.68, "learning_rate": 5.584530445484467e-05, "loss": 0.6094, "step": 6686 }, { "epoch": 0.68, "learning_rate": 5.584408926416818e-05, "loss": 0.7623, "step": 6687 }, { "epoch": 0.68, "learning_rate": 5.584287390902975e-05, "loss": 0.6181, "step": 6688 }, { "epoch": 0.68, "learning_rate": 5.584165838943709e-05, "loss": 0.7254, "step": 6689 }, { "epoch": 0.68, "learning_rate": 5.584044270539795e-05, "loss": 0.8814, "step": 6690 }, { "epoch": 0.68, "learning_rate": 5.5839226856920055e-05, "loss": 0.4072, "step": 6691 }, { "epoch": 0.68, "learning_rate": 5.583801084401116e-05, "loss": 1.3228, "step": 6692 }, { "epoch": 0.68, "learning_rate": 5.583679466667899e-05, "loss": 0.3818, "step": 6693 }, { "epoch": 0.68, "learning_rate": 5.5835578324931276e-05, "loss": 0.97, "step": 6694 }, { "epoch": 0.68, "learning_rate": 5.583436181877579e-05, "loss": 0.5019, "step": 6695 }, { "epoch": 0.68, "learning_rate": 5.5833145148220244e-05, "loss": 1.0001, "step": 6696 }, { "epoch": 0.68, "learning_rate": 5.5831928313272384e-05, "loss": 0.5888, "step": 6697 }, { "epoch": 0.68, "learning_rate": 5.583071131393997e-05, "loss": 0.6838, "step": 6698 }, { "epoch": 0.68, "learning_rate": 5.582949415023072e-05, "loss": 1.1049, "step": 6699 }, { "epoch": 0.68, "learning_rate": 5.582827682215241e-05, "loss": 0.4327, "step": 6700 }, { "epoch": 0.68, "learning_rate": 5.582705932971277e-05, "loss": 0.9336, "step": 6701 }, { "epoch": 0.68, "learning_rate": 5.582584167291955e-05, "loss": 1.2149, "step": 6702 }, { "epoch": 0.68, "learning_rate": 5.5824623851780496e-05, "loss": 0.7307, "step": 6703 }, { "epoch": 0.68, "learning_rate": 5.582340586630335e-05, "loss": 0.1788, "step": 6704 }, { "epoch": 0.68, "learning_rate": 5.582218771649589e-05, "loss": 1.5952, "step": 6705 }, { "epoch": 0.68, "learning_rate": 5.5820969402365846e-05, "loss": 0.5463, "step": 6706 }, { "epoch": 0.68, "learning_rate": 5.581975092392097e-05, "loss": 0.3929, "step": 6707 }, { "epoch": 0.68, "learning_rate": 5.5818532281169024e-05, "loss": 0.5124, "step": 6708 }, { "epoch": 0.68, "learning_rate": 5.581731347411776e-05, "loss": 1.6207, "step": 6709 }, { "epoch": 0.68, "learning_rate": 5.581609450277493e-05, "loss": 1.2481, "step": 6710 }, { "epoch": 0.68, "learning_rate": 5.5814875367148306e-05, "loss": 1.3283, "step": 6711 }, { "epoch": 0.68, "learning_rate": 5.5813656067245624e-05, "loss": 0.2363, "step": 6712 }, { "epoch": 0.68, "learning_rate": 5.5812436603074664e-05, "loss": 0.3779, "step": 6713 }, { "epoch": 0.68, "learning_rate": 5.581121697464317e-05, "loss": 0.9896, "step": 6714 }, { "epoch": 0.68, "learning_rate": 5.5809997181958914e-05, "loss": 1.3792, "step": 6715 }, { "epoch": 0.68, "learning_rate": 5.580877722502966e-05, "loss": 0.5194, "step": 6716 }, { "epoch": 0.68, "learning_rate": 5.580755710386316e-05, "loss": 0.6583, "step": 6717 }, { "epoch": 0.68, "learning_rate": 5.580633681846718e-05, "loss": 1.6219, "step": 6718 }, { "epoch": 0.68, "learning_rate": 5.580511636884949e-05, "loss": 1.5013, "step": 6719 }, { "epoch": 0.68, "learning_rate": 5.580389575501787e-05, "loss": 1.0613, "step": 6720 }, { "epoch": 0.68, "learning_rate": 5.580267497698007e-05, "loss": 0.456, "step": 6721 }, { "epoch": 0.68, "learning_rate": 5.5801454034743847e-05, "loss": 1.5988, "step": 6722 }, { "epoch": 0.68, "learning_rate": 5.5800232928316996e-05, "loss": 1.13, "step": 6723 }, { "epoch": 0.68, "learning_rate": 5.5799011657707276e-05, "loss": 0.9465, "step": 6724 }, { "epoch": 0.68, "learning_rate": 5.5797790222922465e-05, "loss": 1.4394, "step": 6725 }, { "epoch": 0.68, "learning_rate": 5.579656862397033e-05, "loss": 1.4034, "step": 6726 }, { "epoch": 0.68, "learning_rate": 5.579534686085864e-05, "loss": 1.1475, "step": 6727 }, { "epoch": 0.68, "learning_rate": 5.579412493359518e-05, "loss": 1.0163, "step": 6728 }, { "epoch": 0.68, "learning_rate": 5.579290284218771e-05, "loss": 1.5692, "step": 6729 }, { "epoch": 0.68, "learning_rate": 5.579168058664404e-05, "loss": 1.6803, "step": 6730 }, { "epoch": 0.68, "learning_rate": 5.579045816697191e-05, "loss": 0.4754, "step": 6731 }, { "epoch": 0.68, "learning_rate": 5.578923558317912e-05, "loss": 1.1201, "step": 6732 }, { "epoch": 0.68, "learning_rate": 5.578801283527345e-05, "loss": 0.5535, "step": 6733 }, { "epoch": 0.68, "learning_rate": 5.5786789923262676e-05, "loss": 0.4212, "step": 6734 }, { "epoch": 0.68, "learning_rate": 5.578556684715457e-05, "loss": 0.4988, "step": 6735 }, { "epoch": 0.68, "learning_rate": 5.578434360695694e-05, "loss": 0.3254, "step": 6736 }, { "epoch": 0.68, "learning_rate": 5.578312020267755e-05, "loss": 1.4116, "step": 6737 }, { "epoch": 0.68, "learning_rate": 5.578189663432419e-05, "loss": 0.6849, "step": 6738 }, { "epoch": 0.68, "learning_rate": 5.578067290190465e-05, "loss": 1.6321, "step": 6739 }, { "epoch": 0.68, "learning_rate": 5.5779449005426714e-05, "loss": 0.8629, "step": 6740 }, { "epoch": 0.68, "learning_rate": 5.577822494489817e-05, "loss": 1.2138, "step": 6741 }, { "epoch": 0.68, "learning_rate": 5.5777000720326805e-05, "loss": 0.5457, "step": 6742 }, { "epoch": 0.68, "learning_rate": 5.577577633172042e-05, "loss": 0.7467, "step": 6743 }, { "epoch": 0.68, "learning_rate": 5.57745517790868e-05, "loss": 0.6213, "step": 6744 }, { "epoch": 0.68, "learning_rate": 5.577332706243374e-05, "loss": 0.8625, "step": 6745 }, { "epoch": 0.69, "learning_rate": 5.577210218176902e-05, "loss": 0.3999, "step": 6746 }, { "epoch": 0.69, "learning_rate": 5.577087713710046e-05, "loss": 0.9574, "step": 6747 }, { "epoch": 0.69, "learning_rate": 5.576965192843583e-05, "loss": 1.4109, "step": 6748 }, { "epoch": 0.69, "learning_rate": 5.576842655578295e-05, "loss": 1.0735, "step": 6749 }, { "epoch": 0.69, "learning_rate": 5.57672010191496e-05, "loss": 0.8159, "step": 6750 }, { "epoch": 0.69, "learning_rate": 5.576597531854358e-05, "loss": 0.5662, "step": 6751 }, { "epoch": 0.69, "learning_rate": 5.57647494539727e-05, "loss": 1.2101, "step": 6752 }, { "epoch": 0.69, "learning_rate": 5.576352342544475e-05, "loss": 0.8105, "step": 6753 }, { "epoch": 0.69, "learning_rate": 5.576229723296755e-05, "loss": 0.2478, "step": 6754 }, { "epoch": 0.69, "learning_rate": 5.576107087654888e-05, "loss": 1.3228, "step": 6755 }, { "epoch": 0.69, "learning_rate": 5.5759844356196566e-05, "loss": 1.0222, "step": 6756 }, { "epoch": 0.69, "learning_rate": 5.57586176719184e-05, "loss": 1.1169, "step": 6757 }, { "epoch": 0.69, "learning_rate": 5.575739082372218e-05, "loss": 1.48, "step": 6758 }, { "epoch": 0.69, "learning_rate": 5.575616381161573e-05, "loss": 0.9083, "step": 6759 }, { "epoch": 0.69, "learning_rate": 5.575493663560685e-05, "loss": 0.4419, "step": 6760 }, { "epoch": 0.69, "learning_rate": 5.5753709295703355e-05, "loss": 0.4147, "step": 6761 }, { "epoch": 0.69, "learning_rate": 5.575248179191305e-05, "loss": 0.717, "step": 6762 }, { "epoch": 0.69, "learning_rate": 5.575125412424374e-05, "loss": 0.4649, "step": 6763 }, { "epoch": 0.69, "learning_rate": 5.575002629270325e-05, "loss": 1.178, "step": 6764 }, { "epoch": 0.69, "learning_rate": 5.574879829729939e-05, "loss": 0.7845, "step": 6765 }, { "epoch": 0.69, "learning_rate": 5.574757013803997e-05, "loss": 1.0524, "step": 6766 }, { "epoch": 0.69, "learning_rate": 5.5746341814932815e-05, "loss": 0.5381, "step": 6767 }, { "epoch": 0.69, "learning_rate": 5.5745113327985724e-05, "loss": 1.3406, "step": 6768 }, { "epoch": 0.69, "learning_rate": 5.574388467720654e-05, "loss": 0.8985, "step": 6769 }, { "epoch": 0.69, "learning_rate": 5.574265586260306e-05, "loss": 1.1663, "step": 6770 }, { "epoch": 0.69, "learning_rate": 5.574142688418311e-05, "loss": 1.4574, "step": 6771 }, { "epoch": 0.69, "learning_rate": 5.574019774195451e-05, "loss": 1.4898, "step": 6772 }, { "epoch": 0.69, "learning_rate": 5.573896843592508e-05, "loss": 0.4722, "step": 6773 }, { "epoch": 0.69, "learning_rate": 5.5737738966102656e-05, "loss": 1.2256, "step": 6774 }, { "epoch": 0.69, "learning_rate": 5.5736509332495054e-05, "loss": 1.1624, "step": 6775 }, { "epoch": 0.69, "learning_rate": 5.57352795351101e-05, "loss": 0.4647, "step": 6776 }, { "epoch": 0.69, "learning_rate": 5.5734049573955596e-05, "loss": 0.8919, "step": 6777 }, { "epoch": 0.69, "learning_rate": 5.573281944903941e-05, "loss": 1.1792, "step": 6778 }, { "epoch": 0.69, "learning_rate": 5.573158916036934e-05, "loss": 1.3449, "step": 6779 }, { "epoch": 0.69, "learning_rate": 5.573035870795323e-05, "loss": 0.628, "step": 6780 }, { "epoch": 0.69, "learning_rate": 5.5729128091798915e-05, "loss": 1.3161, "step": 6781 }, { "epoch": 0.69, "learning_rate": 5.5727897311914195e-05, "loss": 0.3327, "step": 6782 }, { "epoch": 0.69, "learning_rate": 5.572666636830694e-05, "loss": 0.8652, "step": 6783 }, { "epoch": 0.69, "learning_rate": 5.572543526098496e-05, "loss": 1.0646, "step": 6784 }, { "epoch": 0.69, "learning_rate": 5.572420398995611e-05, "loss": 0.4997, "step": 6785 }, { "epoch": 0.69, "learning_rate": 5.57229725552282e-05, "loss": 0.9058, "step": 6786 }, { "epoch": 0.69, "learning_rate": 5.572174095680907e-05, "loss": 0.8136, "step": 6787 }, { "epoch": 0.69, "learning_rate": 5.572050919470658e-05, "loss": 0.671, "step": 6788 }, { "epoch": 0.69, "learning_rate": 5.571927726892855e-05, "loss": 0.9299, "step": 6789 }, { "epoch": 0.69, "learning_rate": 5.571804517948281e-05, "loss": 0.322, "step": 6790 }, { "epoch": 0.69, "learning_rate": 5.5716812926377235e-05, "loss": 0.8561, "step": 6791 }, { "epoch": 0.69, "learning_rate": 5.571558050961964e-05, "loss": 0.4681, "step": 6792 }, { "epoch": 0.69, "learning_rate": 5.571434792921787e-05, "loss": 2.0253, "step": 6793 }, { "epoch": 0.69, "learning_rate": 5.5713115185179774e-05, "loss": 0.8585, "step": 6794 }, { "epoch": 0.69, "learning_rate": 5.5711882277513196e-05, "loss": 1.0871, "step": 6795 }, { "epoch": 0.69, "learning_rate": 5.571064920622597e-05, "loss": 0.6074, "step": 6796 }, { "epoch": 0.69, "learning_rate": 5.5709415971325966e-05, "loss": 0.8199, "step": 6797 }, { "epoch": 0.69, "learning_rate": 5.570818257282101e-05, "loss": 1.2441, "step": 6798 }, { "epoch": 0.69, "learning_rate": 5.5706949010718966e-05, "loss": 1.0695, "step": 6799 }, { "epoch": 0.69, "learning_rate": 5.5705715285027676e-05, "loss": 0.2298, "step": 6800 }, { "epoch": 0.69, "learning_rate": 5.5704481395755e-05, "loss": 1.0657, "step": 6801 }, { "epoch": 0.69, "learning_rate": 5.570324734290877e-05, "loss": 0.2195, "step": 6802 }, { "epoch": 0.69, "learning_rate": 5.570201312649686e-05, "loss": 1.5363, "step": 6803 }, { "epoch": 0.69, "learning_rate": 5.5700778746527115e-05, "loss": 1.3029, "step": 6804 }, { "epoch": 0.69, "learning_rate": 5.5699544203007395e-05, "loss": 1.2274, "step": 6805 }, { "epoch": 0.69, "learning_rate": 5.569830949594554e-05, "loss": 0.5465, "step": 6806 }, { "epoch": 0.69, "learning_rate": 5.569707462534943e-05, "loss": 1.0731, "step": 6807 }, { "epoch": 0.69, "learning_rate": 5.5695839591226906e-05, "loss": 0.5973, "step": 6808 }, { "epoch": 0.69, "learning_rate": 5.5694604393585844e-05, "loss": 1.501, "step": 6809 }, { "epoch": 0.69, "learning_rate": 5.5693369032434084e-05, "loss": 0.8963, "step": 6810 }, { "epoch": 0.69, "learning_rate": 5.56921335077795e-05, "loss": 1.3337, "step": 6811 }, { "epoch": 0.69, "learning_rate": 5.569089781962996e-05, "loss": 0.4143, "step": 6812 }, { "epoch": 0.69, "learning_rate": 5.56896619679933e-05, "loss": 1.7526, "step": 6813 }, { "epoch": 0.69, "learning_rate": 5.568842595287743e-05, "loss": 0.948, "step": 6814 }, { "epoch": 0.69, "learning_rate": 5.568718977429018e-05, "loss": 0.3008, "step": 6815 }, { "epoch": 0.69, "learning_rate": 5.5685953432239415e-05, "loss": 1.2517, "step": 6816 }, { "epoch": 0.69, "learning_rate": 5.568471692673303e-05, "loss": 0.8701, "step": 6817 }, { "epoch": 0.69, "learning_rate": 5.568348025777887e-05, "loss": 1.3627, "step": 6818 }, { "epoch": 0.69, "learning_rate": 5.568224342538481e-05, "loss": 1.431, "step": 6819 }, { "epoch": 0.69, "learning_rate": 5.568100642955872e-05, "loss": 1.2611, "step": 6820 }, { "epoch": 0.69, "learning_rate": 5.5679769270308485e-05, "loss": 0.3546, "step": 6821 }, { "epoch": 0.69, "learning_rate": 5.567853194764196e-05, "loss": 0.4261, "step": 6822 }, { "epoch": 0.69, "learning_rate": 5.567729446156703e-05, "loss": 0.9388, "step": 6823 }, { "epoch": 0.69, "learning_rate": 5.5676056812091564e-05, "loss": 1.7609, "step": 6824 }, { "epoch": 0.69, "learning_rate": 5.5674818999223436e-05, "loss": 1.0226, "step": 6825 }, { "epoch": 0.69, "learning_rate": 5.567358102297053e-05, "loss": 1.0928, "step": 6826 }, { "epoch": 0.69, "learning_rate": 5.5672342883340724e-05, "loss": 0.0844, "step": 6827 }, { "epoch": 0.69, "learning_rate": 5.56711045803419e-05, "loss": 1.2041, "step": 6828 }, { "epoch": 0.69, "learning_rate": 5.566986611398191e-05, "loss": 0.425, "step": 6829 }, { "epoch": 0.69, "learning_rate": 5.5668627484268677e-05, "loss": 0.8692, "step": 6830 }, { "epoch": 0.69, "learning_rate": 5.5667388691210065e-05, "loss": 1.1633, "step": 6831 }, { "epoch": 0.69, "learning_rate": 5.566614973481395e-05, "loss": 0.6814, "step": 6832 }, { "epoch": 0.69, "learning_rate": 5.566491061508822e-05, "loss": 2.1678, "step": 6833 }, { "epoch": 0.69, "learning_rate": 5.5663671332040765e-05, "loss": 0.7793, "step": 6834 }, { "epoch": 0.69, "learning_rate": 5.5662431885679465e-05, "loss": 0.9263, "step": 6835 }, { "epoch": 0.69, "learning_rate": 5.566119227601221e-05, "loss": 0.5012, "step": 6836 }, { "epoch": 0.69, "learning_rate": 5.56599525030469e-05, "loss": 0.3793, "step": 6837 }, { "epoch": 0.69, "learning_rate": 5.56587125667914e-05, "loss": 1.5004, "step": 6838 }, { "epoch": 0.69, "learning_rate": 5.565747246725362e-05, "loss": 0.9004, "step": 6839 }, { "epoch": 0.69, "learning_rate": 5.565623220444145e-05, "loss": 1.3938, "step": 6840 }, { "epoch": 0.69, "learning_rate": 5.565499177836278e-05, "loss": 0.4137, "step": 6841 }, { "epoch": 0.69, "learning_rate": 5.56537511890255e-05, "loss": 0.9833, "step": 6842 }, { "epoch": 0.69, "learning_rate": 5.56525104364375e-05, "loss": 1.2897, "step": 6843 }, { "epoch": 0.69, "learning_rate": 5.565126952060668e-05, "loss": 1.3202, "step": 6844 }, { "epoch": 0.7, "learning_rate": 5.565002844154095e-05, "loss": 0.9739, "step": 6845 }, { "epoch": 0.7, "learning_rate": 5.564878719924819e-05, "loss": 0.8044, "step": 6846 }, { "epoch": 0.7, "learning_rate": 5.564754579373631e-05, "loss": 0.6042, "step": 6847 }, { "epoch": 0.7, "learning_rate": 5.56463042250132e-05, "loss": 1.0278, "step": 6848 }, { "epoch": 0.7, "learning_rate": 5.564506249308677e-05, "loss": 1.4285, "step": 6849 }, { "epoch": 0.7, "learning_rate": 5.564382059796492e-05, "loss": 1.1268, "step": 6850 }, { "epoch": 0.7, "learning_rate": 5.564257853965554e-05, "loss": 0.6814, "step": 6851 }, { "epoch": 0.7, "learning_rate": 5.564133631816656e-05, "loss": 0.5567, "step": 6852 }, { "epoch": 0.7, "learning_rate": 5.564009393350586e-05, "loss": 0.5883, "step": 6853 }, { "epoch": 0.7, "learning_rate": 5.5638851385681354e-05, "loss": 0.5838, "step": 6854 }, { "epoch": 0.7, "learning_rate": 5.563760867470096e-05, "loss": 0.3098, "step": 6855 }, { "epoch": 0.7, "learning_rate": 5.563636580057258e-05, "loss": 1.3047, "step": 6856 }, { "epoch": 0.7, "learning_rate": 5.563512276330412e-05, "loss": 0.7914, "step": 6857 }, { "epoch": 0.7, "learning_rate": 5.563387956290348e-05, "loss": 1.0754, "step": 6858 }, { "epoch": 0.7, "learning_rate": 5.563263619937859e-05, "loss": 0.3486, "step": 6859 }, { "epoch": 0.7, "learning_rate": 5.563139267273735e-05, "loss": 0.9506, "step": 6860 }, { "epoch": 0.7, "learning_rate": 5.5630148982987684e-05, "loss": 0.497, "step": 6861 }, { "epoch": 0.7, "learning_rate": 5.56289051301375e-05, "loss": 0.5302, "step": 6862 }, { "epoch": 0.7, "learning_rate": 5.562766111419471e-05, "loss": 0.6315, "step": 6863 }, { "epoch": 0.7, "learning_rate": 5.562641693516723e-05, "loss": 0.4337, "step": 6864 }, { "epoch": 0.7, "learning_rate": 5.562517259306299e-05, "loss": 1.6163, "step": 6865 }, { "epoch": 0.7, "learning_rate": 5.5623928087889885e-05, "loss": 0.34, "step": 6866 }, { "epoch": 0.7, "learning_rate": 5.562268341965586e-05, "loss": 0.9775, "step": 6867 }, { "epoch": 0.7, "learning_rate": 5.562143858836882e-05, "loss": 0.9578, "step": 6868 }, { "epoch": 0.7, "learning_rate": 5.56201935940367e-05, "loss": 0.7484, "step": 6869 }, { "epoch": 0.7, "learning_rate": 5.561894843666741e-05, "loss": 0.6229, "step": 6870 }, { "epoch": 0.7, "learning_rate": 5.561770311626889e-05, "loss": 0.4374, "step": 6871 }, { "epoch": 0.7, "learning_rate": 5.561645763284903e-05, "loss": 1.3066, "step": 6872 }, { "epoch": 0.7, "learning_rate": 5.5615211986415796e-05, "loss": 0.2013, "step": 6873 }, { "epoch": 0.7, "learning_rate": 5.561396617697709e-05, "loss": 1.3034, "step": 6874 }, { "epoch": 0.7, "learning_rate": 5.561272020454085e-05, "loss": 1.2428, "step": 6875 }, { "epoch": 0.7, "learning_rate": 5.5611474069115005e-05, "loss": 1.4791, "step": 6876 }, { "epoch": 0.7, "learning_rate": 5.561022777070747e-05, "loss": 0.6689, "step": 6877 }, { "epoch": 0.7, "learning_rate": 5.56089813093262e-05, "loss": 1.1546, "step": 6878 }, { "epoch": 0.7, "learning_rate": 5.5607734684979104e-05, "loss": 1.5065, "step": 6879 }, { "epoch": 0.7, "learning_rate": 5.560648789767414e-05, "loss": 1.053, "step": 6880 }, { "epoch": 0.7, "learning_rate": 5.560524094741922e-05, "loss": 0.2877, "step": 6881 }, { "epoch": 0.7, "learning_rate": 5.560399383422229e-05, "loss": 1.6477, "step": 6882 }, { "epoch": 0.7, "learning_rate": 5.560274655809129e-05, "loss": 1.2981, "step": 6883 }, { "epoch": 0.7, "learning_rate": 5.560149911903413e-05, "loss": 1.4068, "step": 6884 }, { "epoch": 0.7, "learning_rate": 5.5600251517058784e-05, "loss": 0.6455, "step": 6885 }, { "epoch": 0.7, "learning_rate": 5.559900375217318e-05, "loss": 0.3772, "step": 6886 }, { "epoch": 0.7, "learning_rate": 5.559775582438525e-05, "loss": 1.2083, "step": 6887 }, { "epoch": 0.7, "learning_rate": 5.559650773370294e-05, "loss": 1.5028, "step": 6888 }, { "epoch": 0.7, "learning_rate": 5.559525948013419e-05, "loss": 0.5938, "step": 6889 }, { "epoch": 0.7, "learning_rate": 5.5594011063686944e-05, "loss": 1.2538, "step": 6890 }, { "epoch": 0.7, "learning_rate": 5.5592762484369154e-05, "loss": 1.436, "step": 6891 }, { "epoch": 0.7, "learning_rate": 5.5591513742188754e-05, "loss": 0.7956, "step": 6892 }, { "epoch": 0.7, "learning_rate": 5.5590264837153694e-05, "loss": 1.1416, "step": 6893 }, { "epoch": 0.7, "learning_rate": 5.558901576927193e-05, "loss": 0.7964, "step": 6894 }, { "epoch": 0.7, "learning_rate": 5.5587766538551405e-05, "loss": 0.5113, "step": 6895 }, { "epoch": 0.7, "learning_rate": 5.558651714500007e-05, "loss": 0.1793, "step": 6896 }, { "epoch": 0.7, "learning_rate": 5.5585267588625856e-05, "loss": 0.2384, "step": 6897 }, { "epoch": 0.7, "learning_rate": 5.5584017869436753e-05, "loss": 1.1092, "step": 6898 }, { "epoch": 0.7, "learning_rate": 5.558276798744068e-05, "loss": 1.2541, "step": 6899 }, { "epoch": 0.7, "learning_rate": 5.5581517942645616e-05, "loss": 1.6584, "step": 6900 }, { "epoch": 0.7, "learning_rate": 5.558026773505949e-05, "loss": 0.4468, "step": 6901 }, { "epoch": 0.7, "learning_rate": 5.557901736469028e-05, "loss": 1.777, "step": 6902 }, { "epoch": 0.7, "learning_rate": 5.557776683154593e-05, "loss": 0.1909, "step": 6903 }, { "epoch": 0.7, "learning_rate": 5.55765161356344e-05, "loss": 0.5696, "step": 6904 }, { "epoch": 0.7, "learning_rate": 5.5575265276963654e-05, "loss": 1.5832, "step": 6905 }, { "epoch": 0.7, "learning_rate": 5.557401425554165e-05, "loss": 0.9129, "step": 6906 }, { "epoch": 0.7, "learning_rate": 5.5572763071376346e-05, "loss": 0.4653, "step": 6907 }, { "epoch": 0.7, "learning_rate": 5.557151172447571e-05, "loss": 0.3108, "step": 6908 }, { "epoch": 0.7, "learning_rate": 5.557026021484771e-05, "loss": 0.9693, "step": 6909 }, { "epoch": 0.7, "learning_rate": 5.556900854250028e-05, "loss": 0.4599, "step": 6910 }, { "epoch": 0.7, "learning_rate": 5.5567756707441416e-05, "loss": 0.5409, "step": 6911 }, { "epoch": 0.7, "learning_rate": 5.556650470967908e-05, "loss": 0.9869, "step": 6912 }, { "epoch": 0.7, "learning_rate": 5.5565252549221235e-05, "loss": 0.6921, "step": 6913 }, { "epoch": 0.7, "learning_rate": 5.5564000226075837e-05, "loss": 0.8927, "step": 6914 }, { "epoch": 0.7, "learning_rate": 5.5562747740250876e-05, "loss": 1.0008, "step": 6915 }, { "epoch": 0.7, "learning_rate": 5.55614950917543e-05, "loss": 0.3182, "step": 6916 }, { "epoch": 0.7, "learning_rate": 5.556024228059411e-05, "loss": 0.3258, "step": 6917 }, { "epoch": 0.7, "learning_rate": 5.5558989306778255e-05, "loss": 1.6387, "step": 6918 }, { "epoch": 0.7, "learning_rate": 5.555773617031472e-05, "loss": 0.2936, "step": 6919 }, { "epoch": 0.7, "learning_rate": 5.555648287121147e-05, "loss": 1.2429, "step": 6920 }, { "epoch": 0.7, "learning_rate": 5.555522940947649e-05, "loss": 1.076, "step": 6921 }, { "epoch": 0.7, "learning_rate": 5.555397578511774e-05, "loss": 0.6693, "step": 6922 }, { "epoch": 0.7, "learning_rate": 5.555272199814322e-05, "loss": 1.3831, "step": 6923 }, { "epoch": 0.7, "learning_rate": 5.5551468048560884e-05, "loss": 0.831, "step": 6924 }, { "epoch": 0.7, "learning_rate": 5.5550213936378746e-05, "loss": 0.9701, "step": 6925 }, { "epoch": 0.7, "learning_rate": 5.554895966160476e-05, "loss": 1.0473, "step": 6926 }, { "epoch": 0.7, "learning_rate": 5.55477052242469e-05, "loss": 1.6604, "step": 6927 }, { "epoch": 0.7, "learning_rate": 5.554645062431317e-05, "loss": 0.9059, "step": 6928 }, { "epoch": 0.7, "learning_rate": 5.554519586181155e-05, "loss": 1.0997, "step": 6929 }, { "epoch": 0.7, "learning_rate": 5.554394093675003e-05, "loss": 0.8012, "step": 6930 }, { "epoch": 0.7, "learning_rate": 5.5542685849136576e-05, "loss": 0.652, "step": 6931 }, { "epoch": 0.7, "learning_rate": 5.5541430598979185e-05, "loss": 0.7191, "step": 6932 }, { "epoch": 0.7, "learning_rate": 5.554017518628585e-05, "loss": 0.4698, "step": 6933 }, { "epoch": 0.7, "learning_rate": 5.5538919611064554e-05, "loss": 0.9184, "step": 6934 }, { "epoch": 0.7, "learning_rate": 5.553766387332329e-05, "loss": 0.2947, "step": 6935 }, { "epoch": 0.7, "learning_rate": 5.5536407973070045e-05, "loss": 1.0942, "step": 6936 }, { "epoch": 0.7, "learning_rate": 5.553515191031282e-05, "loss": 1.7388, "step": 6937 }, { "epoch": 0.7, "learning_rate": 5.553389568505961e-05, "loss": 0.6733, "step": 6938 }, { "epoch": 0.7, "learning_rate": 5.553263929731838e-05, "loss": 0.7658, "step": 6939 }, { "epoch": 0.7, "learning_rate": 5.553138274709717e-05, "loss": 0.6323, "step": 6940 }, { "epoch": 0.7, "learning_rate": 5.5530126034403934e-05, "loss": 1.7257, "step": 6941 }, { "epoch": 0.7, "learning_rate": 5.55288691592467e-05, "loss": 0.2714, "step": 6942 }, { "epoch": 0.71, "learning_rate": 5.552761212163345e-05, "loss": 0.6782, "step": 6943 }, { "epoch": 0.71, "learning_rate": 5.5526354921572184e-05, "loss": 0.538, "step": 6944 }, { "epoch": 0.71, "learning_rate": 5.552509755907091e-05, "loss": 0.4969, "step": 6945 }, { "epoch": 0.71, "learning_rate": 5.5523840034137625e-05, "loss": 1.389, "step": 6946 }, { "epoch": 0.71, "learning_rate": 5.552258234678033e-05, "loss": 1.307, "step": 6947 }, { "epoch": 0.71, "learning_rate": 5.5521324497007036e-05, "loss": 1.4707, "step": 6948 }, { "epoch": 0.71, "learning_rate": 5.552006648482573e-05, "loss": 0.4425, "step": 6949 }, { "epoch": 0.71, "learning_rate": 5.551880831024444e-05, "loss": 0.4528, "step": 6950 }, { "epoch": 0.71, "learning_rate": 5.551754997327116e-05, "loss": 1.3308, "step": 6951 }, { "epoch": 0.71, "learning_rate": 5.55162914739139e-05, "loss": 0.4962, "step": 6952 }, { "epoch": 0.71, "learning_rate": 5.551503281218066e-05, "loss": 1.1567, "step": 6953 }, { "epoch": 0.71, "learning_rate": 5.551377398807946e-05, "loss": 1.3808, "step": 6954 }, { "epoch": 0.71, "learning_rate": 5.5512515001618316e-05, "loss": 0.6121, "step": 6955 }, { "epoch": 0.71, "learning_rate": 5.5511255852805235e-05, "loss": 0.5543, "step": 6956 }, { "epoch": 0.71, "learning_rate": 5.550999654164821e-05, "loss": 0.7303, "step": 6957 }, { "epoch": 0.71, "learning_rate": 5.550873706815529e-05, "loss": 0.3553, "step": 6958 }, { "epoch": 0.71, "learning_rate": 5.550747743233446e-05, "loss": 0.209, "step": 6959 }, { "epoch": 0.71, "learning_rate": 5.550621763419375e-05, "loss": 0.2756, "step": 6960 }, { "epoch": 0.71, "learning_rate": 5.550495767374117e-05, "loss": 0.9449, "step": 6961 }, { "epoch": 0.71, "learning_rate": 5.550369755098475e-05, "loss": 0.9764, "step": 6962 }, { "epoch": 0.71, "learning_rate": 5.550243726593249e-05, "loss": 1.7178, "step": 6963 }, { "epoch": 0.71, "learning_rate": 5.550117681859243e-05, "loss": 0.818, "step": 6964 }, { "epoch": 0.71, "learning_rate": 5.549991620897258e-05, "loss": 0.398, "step": 6965 }, { "epoch": 0.71, "learning_rate": 5.549865543708096e-05, "loss": 1.1128, "step": 6966 }, { "epoch": 0.71, "learning_rate": 5.549739450292561e-05, "loss": 0.6232, "step": 6967 }, { "epoch": 0.71, "learning_rate": 5.549613340651452e-05, "loss": 1.1919, "step": 6968 }, { "epoch": 0.71, "learning_rate": 5.549487214785575e-05, "loss": 0.8345, "step": 6969 }, { "epoch": 0.71, "learning_rate": 5.549361072695731e-05, "loss": 0.4702, "step": 6970 }, { "epoch": 0.71, "learning_rate": 5.549234914382723e-05, "loss": 0.7218, "step": 6971 }, { "epoch": 0.71, "learning_rate": 5.549108739847354e-05, "loss": 0.2592, "step": 6972 }, { "epoch": 0.71, "learning_rate": 5.548982549090426e-05, "loss": 1.1147, "step": 6973 }, { "epoch": 0.71, "learning_rate": 5.548856342112742e-05, "loss": 1.9616, "step": 6974 }, { "epoch": 0.71, "learning_rate": 5.548730118915108e-05, "loss": 0.2028, "step": 6975 }, { "epoch": 0.71, "learning_rate": 5.548603879498325e-05, "loss": 0.6868, "step": 6976 }, { "epoch": 0.71, "learning_rate": 5.5484776238631945e-05, "loss": 0.3983, "step": 6977 }, { "epoch": 0.71, "learning_rate": 5.5483513520105234e-05, "loss": 1.1311, "step": 6978 }, { "epoch": 0.71, "learning_rate": 5.5482250639411124e-05, "loss": 0.9592, "step": 6979 }, { "epoch": 0.71, "learning_rate": 5.548098759655768e-05, "loss": 0.4212, "step": 6980 }, { "epoch": 0.71, "learning_rate": 5.5479724391552915e-05, "loss": 0.5652, "step": 6981 }, { "epoch": 0.71, "learning_rate": 5.5478461024404884e-05, "loss": 1.2669, "step": 6982 }, { "epoch": 0.71, "learning_rate": 5.547719749512161e-05, "loss": 0.6763, "step": 6983 }, { "epoch": 0.71, "learning_rate": 5.5475933803711155e-05, "loss": 0.7539, "step": 6984 }, { "epoch": 0.71, "learning_rate": 5.547466995018154e-05, "loss": 0.7651, "step": 6985 }, { "epoch": 0.71, "learning_rate": 5.547340593454082e-05, "loss": 0.85, "step": 6986 }, { "epoch": 0.71, "learning_rate": 5.5472141756797036e-05, "loss": 0.3096, "step": 6987 }, { "epoch": 0.71, "learning_rate": 5.547087741695823e-05, "loss": 1.3079, "step": 6988 }, { "epoch": 0.71, "learning_rate": 5.546961291503245e-05, "loss": 0.8287, "step": 6989 }, { "epoch": 0.71, "learning_rate": 5.5468348251027745e-05, "loss": 0.8854, "step": 6990 }, { "epoch": 0.71, "learning_rate": 5.5467083424952164e-05, "loss": 0.4501, "step": 6991 }, { "epoch": 0.71, "learning_rate": 5.5465818436813735e-05, "loss": 0.6999, "step": 6992 }, { "epoch": 0.71, "learning_rate": 5.546455328662054e-05, "loss": 0.4037, "step": 6993 }, { "epoch": 0.71, "learning_rate": 5.546328797438061e-05, "loss": 1.4137, "step": 6994 }, { "epoch": 0.71, "learning_rate": 5.5462022500102e-05, "loss": 0.4679, "step": 6995 }, { "epoch": 0.71, "learning_rate": 5.546075686379278e-05, "loss": 0.5804, "step": 6996 }, { "epoch": 0.71, "learning_rate": 5.545949106546098e-05, "loss": 0.9177, "step": 6997 }, { "epoch": 0.71, "learning_rate": 5.545822510511466e-05, "loss": 1.2433, "step": 6998 }, { "epoch": 0.71, "learning_rate": 5.545695898276188e-05, "loss": 1.0579, "step": 6999 }, { "epoch": 0.71, "learning_rate": 5.54556926984107e-05, "loss": 0.2855, "step": 7000 }, { "epoch": 0.71, "learning_rate": 5.545442625206917e-05, "loss": 0.7411, "step": 7001 }, { "epoch": 0.71, "learning_rate": 5.545315964374536e-05, "loss": 0.873, "step": 7002 }, { "epoch": 0.71, "learning_rate": 5.5451892873447326e-05, "loss": 1.8317, "step": 7003 }, { "epoch": 0.71, "learning_rate": 5.545062594118313e-05, "loss": 0.3324, "step": 7004 }, { "epoch": 0.71, "learning_rate": 5.544935884696083e-05, "loss": 1.186, "step": 7005 }, { "epoch": 0.71, "learning_rate": 5.544809159078849e-05, "loss": 0.722, "step": 7006 }, { "epoch": 0.71, "learning_rate": 5.544682417267417e-05, "loss": 1.1284, "step": 7007 }, { "epoch": 0.71, "learning_rate": 5.5445556592625954e-05, "loss": 1.2203, "step": 7008 }, { "epoch": 0.71, "learning_rate": 5.5444288850651886e-05, "loss": 0.8377, "step": 7009 }, { "epoch": 0.71, "learning_rate": 5.544302094676005e-05, "loss": 0.3096, "step": 7010 }, { "epoch": 0.71, "learning_rate": 5.5441752880958504e-05, "loss": 0.8905, "step": 7011 }, { "epoch": 0.71, "learning_rate": 5.5440484653255324e-05, "loss": 0.5375, "step": 7012 }, { "epoch": 0.71, "learning_rate": 5.543921626365858e-05, "loss": 0.3191, "step": 7013 }, { "epoch": 0.71, "learning_rate": 5.543794771217634e-05, "loss": 0.1588, "step": 7014 }, { "epoch": 0.71, "learning_rate": 5.543667899881667e-05, "loss": 1.3858, "step": 7015 }, { "epoch": 0.71, "learning_rate": 5.543541012358766e-05, "loss": 0.4347, "step": 7016 }, { "epoch": 0.71, "learning_rate": 5.543414108649737e-05, "loss": 1.739, "step": 7017 }, { "epoch": 0.71, "learning_rate": 5.543287188755389e-05, "loss": 0.9179, "step": 7018 }, { "epoch": 0.71, "learning_rate": 5.543160252676528e-05, "loss": 1.5859, "step": 7019 }, { "epoch": 0.71, "learning_rate": 5.543033300413963e-05, "loss": 0.4124, "step": 7020 }, { "epoch": 0.71, "learning_rate": 5.542906331968501e-05, "loss": 0.4991, "step": 7021 }, { "epoch": 0.71, "learning_rate": 5.542779347340951e-05, "loss": 1.5182, "step": 7022 }, { "epoch": 0.71, "learning_rate": 5.542652346532121e-05, "loss": 0.6855, "step": 7023 }, { "epoch": 0.71, "learning_rate": 5.5425253295428176e-05, "loss": 1.4006, "step": 7024 }, { "epoch": 0.71, "learning_rate": 5.54239829637385e-05, "loss": 0.9883, "step": 7025 }, { "epoch": 0.71, "learning_rate": 5.5422712470260285e-05, "loss": 0.4197, "step": 7026 }, { "epoch": 0.71, "learning_rate": 5.542144181500159e-05, "loss": 0.8756, "step": 7027 }, { "epoch": 0.71, "learning_rate": 5.542017099797052e-05, "loss": 0.8349, "step": 7028 }, { "epoch": 0.71, "learning_rate": 5.541890001917513e-05, "loss": 0.5529, "step": 7029 }, { "epoch": 0.71, "learning_rate": 5.541762887862355e-05, "loss": 0.8696, "step": 7030 }, { "epoch": 0.71, "learning_rate": 5.541635757632384e-05, "loss": 1.2997, "step": 7031 }, { "epoch": 0.71, "learning_rate": 5.54150861122841e-05, "loss": 0.6513, "step": 7032 }, { "epoch": 0.71, "learning_rate": 5.541381448651242e-05, "loss": 0.8974, "step": 7033 }, { "epoch": 0.71, "learning_rate": 5.54125426990169e-05, "loss": 1.3835, "step": 7034 }, { "epoch": 0.71, "learning_rate": 5.541127074980562e-05, "loss": 0.8045, "step": 7035 }, { "epoch": 0.71, "learning_rate": 5.540999863888668e-05, "loss": 0.8937, "step": 7036 }, { "epoch": 0.71, "learning_rate": 5.5408726366268174e-05, "loss": 1.7266, "step": 7037 }, { "epoch": 0.71, "learning_rate": 5.5407453931958204e-05, "loss": 1.368, "step": 7038 }, { "epoch": 0.71, "learning_rate": 5.540618133596486e-05, "loss": 1.464, "step": 7039 }, { "epoch": 0.71, "learning_rate": 5.540490857829624e-05, "loss": 0.4197, "step": 7040 }, { "epoch": 0.71, "learning_rate": 5.5403635658960444e-05, "loss": 1.2378, "step": 7041 }, { "epoch": 0.72, "learning_rate": 5.540236257796558e-05, "loss": 0.8301, "step": 7042 }, { "epoch": 0.72, "learning_rate": 5.540108933531974e-05, "loss": 0.5708, "step": 7043 }, { "epoch": 0.72, "learning_rate": 5.539981593103104e-05, "loss": 0.7135, "step": 7044 }, { "epoch": 0.72, "learning_rate": 5.5398542365107565e-05, "loss": 0.4143, "step": 7045 }, { "epoch": 0.72, "learning_rate": 5.5397268637557435e-05, "loss": 0.5538, "step": 7046 }, { "epoch": 0.72, "learning_rate": 5.5395994748388744e-05, "loss": 1.1467, "step": 7047 }, { "epoch": 0.72, "learning_rate": 5.539472069760961e-05, "loss": 0.7514, "step": 7048 }, { "epoch": 0.72, "learning_rate": 5.539344648522812e-05, "loss": 1.0036, "step": 7049 }, { "epoch": 0.72, "learning_rate": 5.539217211125241e-05, "loss": 0.3074, "step": 7050 }, { "epoch": 0.72, "learning_rate": 5.539089757569057e-05, "loss": 1.1816, "step": 7051 }, { "epoch": 0.72, "learning_rate": 5.538962287855072e-05, "loss": 1.6286, "step": 7052 }, { "epoch": 0.72, "learning_rate": 5.5388348019840965e-05, "loss": 1.1077, "step": 7053 }, { "epoch": 0.72, "learning_rate": 5.5387072999569426e-05, "loss": 0.535, "step": 7054 }, { "epoch": 0.72, "learning_rate": 5.5385797817744214e-05, "loss": 0.3429, "step": 7055 }, { "epoch": 0.72, "learning_rate": 5.5384522474373435e-05, "loss": 0.4094, "step": 7056 }, { "epoch": 0.72, "learning_rate": 5.538324696946521e-05, "loss": 1.4067, "step": 7057 }, { "epoch": 0.72, "learning_rate": 5.538197130302766e-05, "loss": 1.4668, "step": 7058 }, { "epoch": 0.72, "learning_rate": 5.5380695475068914e-05, "loss": 1.4496, "step": 7059 }, { "epoch": 0.72, "learning_rate": 5.537941948559706e-05, "loss": 1.489, "step": 7060 }, { "epoch": 0.72, "learning_rate": 5.5378143334620246e-05, "loss": 0.7069, "step": 7061 }, { "epoch": 0.72, "learning_rate": 5.5376867022146576e-05, "loss": 0.5297, "step": 7062 }, { "epoch": 0.72, "learning_rate": 5.537559054818418e-05, "loss": 1.6449, "step": 7063 }, { "epoch": 0.72, "learning_rate": 5.537431391274118e-05, "loss": 1.0363, "step": 7064 }, { "epoch": 0.72, "learning_rate": 5.537303711582569e-05, "loss": 0.5599, "step": 7065 }, { "epoch": 0.72, "learning_rate": 5.5371760157445855e-05, "loss": 0.8884, "step": 7066 }, { "epoch": 0.72, "learning_rate": 5.537048303760979e-05, "loss": 0.2133, "step": 7067 }, { "epoch": 0.72, "learning_rate": 5.536920575632562e-05, "loss": 0.3093, "step": 7068 }, { "epoch": 0.72, "learning_rate": 5.536792831360147e-05, "loss": 1.1708, "step": 7069 }, { "epoch": 0.72, "learning_rate": 5.5366650709445485e-05, "loss": 1.2101, "step": 7070 }, { "epoch": 0.72, "learning_rate": 5.5365372943865774e-05, "loss": 0.209, "step": 7071 }, { "epoch": 0.72, "learning_rate": 5.536409501687049e-05, "loss": 1.2353, "step": 7072 }, { "epoch": 0.72, "learning_rate": 5.536281692846775e-05, "loss": 0.6951, "step": 7073 }, { "epoch": 0.72, "learning_rate": 5.5361538678665687e-05, "loss": 0.3968, "step": 7074 }, { "epoch": 0.72, "learning_rate": 5.536026026747245e-05, "loss": 1.1645, "step": 7075 }, { "epoch": 0.72, "learning_rate": 5.535898169489615e-05, "loss": 0.3109, "step": 7076 }, { "epoch": 0.72, "learning_rate": 5.5357702960944946e-05, "loss": 0.3595, "step": 7077 }, { "epoch": 0.72, "learning_rate": 5.5356424065626964e-05, "loss": 1.4257, "step": 7078 }, { "epoch": 0.72, "learning_rate": 5.535514500895035e-05, "loss": 1.5936, "step": 7079 }, { "epoch": 0.72, "learning_rate": 5.535386579092324e-05, "loss": 0.8805, "step": 7080 }, { "epoch": 0.72, "learning_rate": 5.535258641155377e-05, "loss": 0.8273, "step": 7081 }, { "epoch": 0.72, "learning_rate": 5.535130687085009e-05, "loss": 0.1925, "step": 7082 }, { "epoch": 0.72, "learning_rate": 5.5350027168820325e-05, "loss": 1.1507, "step": 7083 }, { "epoch": 0.72, "learning_rate": 5.5348747305472643e-05, "loss": 1.0427, "step": 7084 }, { "epoch": 0.72, "learning_rate": 5.534746728081516e-05, "loss": 1.2519, "step": 7085 }, { "epoch": 0.72, "learning_rate": 5.5346187094856055e-05, "loss": 1.0971, "step": 7086 }, { "epoch": 0.72, "learning_rate": 5.534490674760345e-05, "loss": 0.3667, "step": 7087 }, { "epoch": 0.72, "learning_rate": 5.5343626239065496e-05, "loss": 0.3974, "step": 7088 }, { "epoch": 0.72, "learning_rate": 5.534234556925036e-05, "loss": 0.92, "step": 7089 }, { "epoch": 0.72, "learning_rate": 5.534106473816616e-05, "loss": 1.0964, "step": 7090 }, { "epoch": 0.72, "learning_rate": 5.533978374582108e-05, "loss": 1.8586, "step": 7091 }, { "epoch": 0.72, "learning_rate": 5.533850259222325e-05, "loss": 1.206, "step": 7092 }, { "epoch": 0.72, "learning_rate": 5.533722127738082e-05, "loss": 0.9404, "step": 7093 }, { "epoch": 0.72, "learning_rate": 5.533593980130197e-05, "loss": 0.8921, "step": 7094 }, { "epoch": 0.72, "learning_rate": 5.533465816399482e-05, "loss": 1.0224, "step": 7095 }, { "epoch": 0.72, "learning_rate": 5.533337636546755e-05, "loss": 0.7905, "step": 7096 }, { "epoch": 0.72, "learning_rate": 5.533209440572831e-05, "loss": 0.3539, "step": 7097 }, { "epoch": 0.72, "learning_rate": 5.533081228478526e-05, "loss": 0.2559, "step": 7098 }, { "epoch": 0.72, "learning_rate": 5.532953000264656e-05, "loss": 0.6717, "step": 7099 }, { "epoch": 0.72, "learning_rate": 5.5328247559320357e-05, "loss": 0.4438, "step": 7100 }, { "epoch": 0.72, "learning_rate": 5.5326964954814836e-05, "loss": 0.7181, "step": 7101 }, { "epoch": 0.72, "learning_rate": 5.532568218913813e-05, "loss": 0.3152, "step": 7102 }, { "epoch": 0.72, "learning_rate": 5.532439926229842e-05, "loss": 0.6807, "step": 7103 }, { "epoch": 0.72, "learning_rate": 5.532311617430387e-05, "loss": 0.7021, "step": 7104 }, { "epoch": 0.72, "learning_rate": 5.532183292516265e-05, "loss": 1.4047, "step": 7105 }, { "epoch": 0.72, "learning_rate": 5.532054951488291e-05, "loss": 1.0079, "step": 7106 }, { "epoch": 0.72, "learning_rate": 5.531926594347282e-05, "loss": 1.1846, "step": 7107 }, { "epoch": 0.72, "learning_rate": 5.5317982210940564e-05, "loss": 0.8386, "step": 7108 }, { "epoch": 0.72, "learning_rate": 5.531669831729429e-05, "loss": 0.6712, "step": 7109 }, { "epoch": 0.72, "learning_rate": 5.531541426254219e-05, "loss": 0.4872, "step": 7110 }, { "epoch": 0.72, "learning_rate": 5.531413004669242e-05, "loss": 0.2066, "step": 7111 }, { "epoch": 0.72, "learning_rate": 5.531284566975316e-05, "loss": 0.9164, "step": 7112 }, { "epoch": 0.72, "learning_rate": 5.531156113173257e-05, "loss": 0.7059, "step": 7113 }, { "epoch": 0.72, "learning_rate": 5.5310276432638834e-05, "loss": 1.1112, "step": 7114 }, { "epoch": 0.72, "learning_rate": 5.530899157248014e-05, "loss": 0.3148, "step": 7115 }, { "epoch": 0.72, "learning_rate": 5.530770655126465e-05, "loss": 0.9391, "step": 7116 }, { "epoch": 0.72, "learning_rate": 5.530642136900053e-05, "loss": 1.2556, "step": 7117 }, { "epoch": 0.72, "learning_rate": 5.5305136025695976e-05, "loss": 0.4481, "step": 7118 }, { "epoch": 0.72, "learning_rate": 5.530385052135917e-05, "loss": 0.8587, "step": 7119 }, { "epoch": 0.72, "learning_rate": 5.530256485599828e-05, "loss": 1.0873, "step": 7120 }, { "epoch": 0.72, "learning_rate": 5.5301279029621495e-05, "loss": 0.4169, "step": 7121 }, { "epoch": 0.72, "learning_rate": 5.5299993042236996e-05, "loss": 1.5678, "step": 7122 }, { "epoch": 0.72, "learning_rate": 5.529870689385296e-05, "loss": 0.8752, "step": 7123 }, { "epoch": 0.72, "learning_rate": 5.529742058447759e-05, "loss": 0.1749, "step": 7124 }, { "epoch": 0.72, "learning_rate": 5.5296134114119035e-05, "loss": 0.7879, "step": 7125 }, { "epoch": 0.72, "learning_rate": 5.529484748278553e-05, "loss": 0.5776, "step": 7126 }, { "epoch": 0.72, "learning_rate": 5.529356069048522e-05, "loss": 1.3807, "step": 7127 }, { "epoch": 0.72, "learning_rate": 5.5292273737226325e-05, "loss": 1.1882, "step": 7128 }, { "epoch": 0.72, "learning_rate": 5.529098662301702e-05, "loss": 0.5871, "step": 7129 }, { "epoch": 0.72, "learning_rate": 5.528969934786549e-05, "loss": 0.3965, "step": 7130 }, { "epoch": 0.72, "learning_rate": 5.528841191177994e-05, "loss": 1.5386, "step": 7131 }, { "epoch": 0.72, "learning_rate": 5.528712431476856e-05, "loss": 1.4961, "step": 7132 }, { "epoch": 0.72, "learning_rate": 5.5285836556839545e-05, "loss": 1.2395, "step": 7133 }, { "epoch": 0.72, "learning_rate": 5.528454863800107e-05, "loss": 1.5138, "step": 7134 }, { "epoch": 0.72, "learning_rate": 5.528326055826135e-05, "loss": 1.2934, "step": 7135 }, { "epoch": 0.72, "learning_rate": 5.528197231762858e-05, "loss": 0.4124, "step": 7136 }, { "epoch": 0.72, "learning_rate": 5.528068391611096e-05, "loss": 1.1477, "step": 7137 }, { "epoch": 0.72, "learning_rate": 5.527939535371668e-05, "loss": 0.3678, "step": 7138 }, { "epoch": 0.72, "learning_rate": 5.5278106630453946e-05, "loss": 0.6193, "step": 7139 }, { "epoch": 0.73, "learning_rate": 5.527681774633096e-05, "loss": 1.1669, "step": 7140 }, { "epoch": 0.73, "learning_rate": 5.527552870135592e-05, "loss": 0.3487, "step": 7141 }, { "epoch": 0.73, "learning_rate": 5.527423949553703e-05, "loss": 0.8647, "step": 7142 }, { "epoch": 0.73, "learning_rate": 5.5272950128882494e-05, "loss": 1.1614, "step": 7143 }, { "epoch": 0.73, "learning_rate": 5.5271660601400527e-05, "loss": 0.5618, "step": 7144 }, { "epoch": 0.73, "learning_rate": 5.527037091309932e-05, "loss": 0.4825, "step": 7145 }, { "epoch": 0.73, "learning_rate": 5.526908106398708e-05, "loss": 0.6453, "step": 7146 }, { "epoch": 0.73, "learning_rate": 5.5267791054072024e-05, "loss": 0.4295, "step": 7147 }, { "epoch": 0.73, "learning_rate": 5.5266500883362357e-05, "loss": 0.9906, "step": 7148 }, { "epoch": 0.73, "learning_rate": 5.526521055186629e-05, "loss": 2.011, "step": 7149 }, { "epoch": 0.73, "learning_rate": 5.526392005959205e-05, "loss": 0.8436, "step": 7150 }, { "epoch": 0.73, "learning_rate": 5.52626294065478e-05, "loss": 0.4051, "step": 7151 }, { "epoch": 0.73, "learning_rate": 5.5261338592741815e-05, "loss": 0.9076, "step": 7152 }, { "epoch": 0.73, "learning_rate": 5.526004761818227e-05, "loss": 1.0304, "step": 7153 }, { "epoch": 0.73, "learning_rate": 5.5258756482877396e-05, "loss": 0.7397, "step": 7154 }, { "epoch": 0.73, "learning_rate": 5.525746518683541e-05, "loss": 0.6432, "step": 7155 }, { "epoch": 0.73, "learning_rate": 5.5256173730064515e-05, "loss": 0.3718, "step": 7156 }, { "epoch": 0.73, "learning_rate": 5.525488211257294e-05, "loss": 0.5059, "step": 7157 }, { "epoch": 0.73, "learning_rate": 5.52535903343689e-05, "loss": 0.3556, "step": 7158 }, { "epoch": 0.73, "learning_rate": 5.525229839546062e-05, "loss": 1.1792, "step": 7159 }, { "epoch": 0.73, "learning_rate": 5.525100629585632e-05, "loss": 1.1312, "step": 7160 }, { "epoch": 0.73, "learning_rate": 5.524971403556422e-05, "loss": 0.6264, "step": 7161 }, { "epoch": 0.73, "learning_rate": 5.524842161459255e-05, "loss": 1.7106, "step": 7162 }, { "epoch": 0.73, "learning_rate": 5.524712903294953e-05, "loss": 1.2104, "step": 7163 }, { "epoch": 0.73, "learning_rate": 5.524583629064338e-05, "loss": 1.1801, "step": 7164 }, { "epoch": 0.73, "learning_rate": 5.524454338768234e-05, "loss": 1.3597, "step": 7165 }, { "epoch": 0.73, "learning_rate": 5.524325032407462e-05, "loss": 0.2944, "step": 7166 }, { "epoch": 0.73, "learning_rate": 5.5241957099828464e-05, "loss": 0.1444, "step": 7167 }, { "epoch": 0.73, "learning_rate": 5.52406637149521e-05, "loss": 1.3542, "step": 7168 }, { "epoch": 0.73, "learning_rate": 5.523937016945374e-05, "loss": 0.3489, "step": 7169 }, { "epoch": 0.73, "learning_rate": 5.5238076463341635e-05, "loss": 0.3052, "step": 7170 }, { "epoch": 0.73, "learning_rate": 5.523678259662402e-05, "loss": 0.4693, "step": 7171 }, { "epoch": 0.73, "learning_rate": 5.5235488569309117e-05, "loss": 0.2865, "step": 7172 }, { "epoch": 0.73, "learning_rate": 5.5234194381405164e-05, "loss": 1.1279, "step": 7173 }, { "epoch": 0.73, "learning_rate": 5.52329000329204e-05, "loss": 1.0331, "step": 7174 }, { "epoch": 0.73, "learning_rate": 5.523160552386305e-05, "loss": 1.7445, "step": 7175 }, { "epoch": 0.73, "learning_rate": 5.5230310854241366e-05, "loss": 1.4827, "step": 7176 }, { "epoch": 0.73, "learning_rate": 5.522901602406358e-05, "loss": 0.5408, "step": 7177 }, { "epoch": 0.73, "learning_rate": 5.522772103333794e-05, "loss": 0.749, "step": 7178 }, { "epoch": 0.73, "learning_rate": 5.522642588207268e-05, "loss": 0.9927, "step": 7179 }, { "epoch": 0.73, "learning_rate": 5.522513057027605e-05, "loss": 0.6844, "step": 7180 }, { "epoch": 0.73, "learning_rate": 5.522383509795627e-05, "loss": 0.4886, "step": 7181 }, { "epoch": 0.73, "learning_rate": 5.52225394651216e-05, "loss": 0.4856, "step": 7182 }, { "epoch": 0.73, "learning_rate": 5.522124367178029e-05, "loss": 0.872, "step": 7183 }, { "epoch": 0.73, "learning_rate": 5.521994771794058e-05, "loss": 0.3288, "step": 7184 }, { "epoch": 0.73, "learning_rate": 5.5218651603610714e-05, "loss": 1.1558, "step": 7185 }, { "epoch": 0.73, "learning_rate": 5.521735532879895e-05, "loss": 1.1447, "step": 7186 }, { "epoch": 0.73, "learning_rate": 5.5216058893513514e-05, "loss": 0.9355, "step": 7187 }, { "epoch": 0.73, "learning_rate": 5.521476229776269e-05, "loss": 1.2608, "step": 7188 }, { "epoch": 0.73, "learning_rate": 5.52134655415547e-05, "loss": 0.9046, "step": 7189 }, { "epoch": 0.73, "learning_rate": 5.521216862489781e-05, "loss": 0.3276, "step": 7190 }, { "epoch": 0.73, "learning_rate": 5.5210871547800274e-05, "loss": 1.1083, "step": 7191 }, { "epoch": 0.73, "learning_rate": 5.520957431027033e-05, "loss": 2.0614, "step": 7192 }, { "epoch": 0.73, "learning_rate": 5.520827691231626e-05, "loss": 1.6775, "step": 7193 }, { "epoch": 0.73, "learning_rate": 5.5206979353946294e-05, "loss": 0.2525, "step": 7194 }, { "epoch": 0.73, "learning_rate": 5.52056816351687e-05, "loss": 0.3693, "step": 7195 }, { "epoch": 0.73, "learning_rate": 5.520438375599175e-05, "loss": 1.3885, "step": 7196 }, { "epoch": 0.73, "learning_rate": 5.5203085716423685e-05, "loss": 0.6099, "step": 7197 }, { "epoch": 0.73, "learning_rate": 5.5201787516472764e-05, "loss": 1.2095, "step": 7198 }, { "epoch": 0.73, "learning_rate": 5.5200489156147255e-05, "loss": 0.9155, "step": 7199 }, { "epoch": 0.73, "learning_rate": 5.5199190635455424e-05, "loss": 0.24, "step": 7200 }, { "epoch": 0.73, "learning_rate": 5.5197891954405525e-05, "loss": 1.333, "step": 7201 }, { "epoch": 0.73, "learning_rate": 5.519659311300583e-05, "loss": 1.035, "step": 7202 }, { "epoch": 0.73, "learning_rate": 5.519529411126461e-05, "loss": 1.0005, "step": 7203 }, { "epoch": 0.73, "learning_rate": 5.519399494919011e-05, "loss": 1.5008, "step": 7204 }, { "epoch": 0.73, "learning_rate": 5.519269562679062e-05, "loss": 1.1113, "step": 7205 }, { "epoch": 0.73, "learning_rate": 5.519139614407439e-05, "loss": 0.5381, "step": 7206 }, { "epoch": 0.73, "learning_rate": 5.519009650104971e-05, "loss": 1.3021, "step": 7207 }, { "epoch": 0.73, "learning_rate": 5.518879669772483e-05, "loss": 0.3883, "step": 7208 }, { "epoch": 0.73, "learning_rate": 5.5187496734108035e-05, "loss": 0.8429, "step": 7209 }, { "epoch": 0.73, "learning_rate": 5.518619661020759e-05, "loss": 1.4151, "step": 7210 }, { "epoch": 0.73, "learning_rate": 5.518489632603177e-05, "loss": 0.622, "step": 7211 }, { "epoch": 0.73, "learning_rate": 5.518359588158885e-05, "loss": 0.8837, "step": 7212 }, { "epoch": 0.73, "learning_rate": 5.518229527688711e-05, "loss": 0.8925, "step": 7213 }, { "epoch": 0.73, "learning_rate": 5.518099451193482e-05, "loss": 1.0575, "step": 7214 }, { "epoch": 0.73, "learning_rate": 5.517969358674026e-05, "loss": 0.4703, "step": 7215 }, { "epoch": 0.73, "learning_rate": 5.5178392501311705e-05, "loss": 0.3966, "step": 7216 }, { "epoch": 0.73, "learning_rate": 5.5177091255657447e-05, "loss": 1.408, "step": 7217 }, { "epoch": 0.73, "learning_rate": 5.5175789849785755e-05, "loss": 0.6846, "step": 7218 }, { "epoch": 0.73, "learning_rate": 5.517448828370491e-05, "loss": 1.3446, "step": 7219 }, { "epoch": 0.73, "learning_rate": 5.51731865574232e-05, "loss": 0.7279, "step": 7220 }, { "epoch": 0.73, "learning_rate": 5.51718846709489e-05, "loss": 1.0735, "step": 7221 }, { "epoch": 0.73, "learning_rate": 5.517058262429032e-05, "loss": 0.3261, "step": 7222 }, { "epoch": 0.73, "learning_rate": 5.5169280417455715e-05, "loss": 0.4093, "step": 7223 }, { "epoch": 0.73, "learning_rate": 5.5167978050453386e-05, "loss": 1.6621, "step": 7224 }, { "epoch": 0.73, "learning_rate": 5.516667552329163e-05, "loss": 1.4565, "step": 7225 }, { "epoch": 0.73, "learning_rate": 5.516537283597871e-05, "loss": 0.4581, "step": 7226 }, { "epoch": 0.73, "learning_rate": 5.5164069988522934e-05, "loss": 0.943, "step": 7227 }, { "epoch": 0.73, "learning_rate": 5.516276698093259e-05, "loss": 0.7166, "step": 7228 }, { "epoch": 0.73, "learning_rate": 5.5161463813215966e-05, "loss": 0.9479, "step": 7229 }, { "epoch": 0.73, "learning_rate": 5.5160160485381366e-05, "loss": 0.2588, "step": 7230 }, { "epoch": 0.73, "learning_rate": 5.515885699743707e-05, "loss": 1.4632, "step": 7231 }, { "epoch": 0.73, "learning_rate": 5.515755334939138e-05, "loss": 0.7583, "step": 7232 }, { "epoch": 0.73, "learning_rate": 5.5156249541252596e-05, "loss": 1.0456, "step": 7233 }, { "epoch": 0.73, "learning_rate": 5.515494557302901e-05, "loss": 1.2823, "step": 7234 }, { "epoch": 0.73, "learning_rate": 5.5153641444728915e-05, "loss": 0.5043, "step": 7235 }, { "epoch": 0.73, "learning_rate": 5.515233715636062e-05, "loss": 1.3023, "step": 7236 }, { "epoch": 0.73, "learning_rate": 5.515103270793241e-05, "loss": 1.6947, "step": 7237 }, { "epoch": 0.73, "learning_rate": 5.5149728099452606e-05, "loss": 0.7058, "step": 7238 }, { "epoch": 0.74, "learning_rate": 5.514842333092949e-05, "loss": 1.4827, "step": 7239 }, { "epoch": 0.74, "learning_rate": 5.514711840237138e-05, "loss": 1.1632, "step": 7240 }, { "epoch": 0.74, "learning_rate": 5.514581331378658e-05, "loss": 1.5525, "step": 7241 }, { "epoch": 0.74, "learning_rate": 5.514450806518338e-05, "loss": 0.6164, "step": 7242 }, { "epoch": 0.74, "learning_rate": 5.514320265657011e-05, "loss": 1.1026, "step": 7243 }, { "epoch": 0.74, "learning_rate": 5.5141897087955054e-05, "loss": 0.577, "step": 7244 }, { "epoch": 0.74, "learning_rate": 5.514059135934652e-05, "loss": 0.7111, "step": 7245 }, { "epoch": 0.74, "learning_rate": 5.513928547075284e-05, "loss": 0.5421, "step": 7246 }, { "epoch": 0.74, "learning_rate": 5.513797942218231e-05, "loss": 1.2371, "step": 7247 }, { "epoch": 0.74, "learning_rate": 5.5136673213643234e-05, "loss": 2.5505, "step": 7248 }, { "epoch": 0.74, "learning_rate": 5.513536684514394e-05, "loss": 1.4979, "step": 7249 }, { "epoch": 0.74, "learning_rate": 5.513406031669273e-05, "loss": 0.541, "step": 7250 }, { "epoch": 0.74, "learning_rate": 5.5132753628297924e-05, "loss": 1.6297, "step": 7251 }, { "epoch": 0.74, "learning_rate": 5.513144677996783e-05, "loss": 0.3789, "step": 7252 }, { "epoch": 0.74, "learning_rate": 5.5130139771710775e-05, "loss": 1.5657, "step": 7253 }, { "epoch": 0.74, "learning_rate": 5.512883260353507e-05, "loss": 1.5037, "step": 7254 }, { "epoch": 0.74, "learning_rate": 5.512752527544903e-05, "loss": 1.2615, "step": 7255 }, { "epoch": 0.74, "learning_rate": 5.512621778746098e-05, "loss": 0.7891, "step": 7256 }, { "epoch": 0.74, "learning_rate": 5.512491013957924e-05, "loss": 1.3383, "step": 7257 }, { "epoch": 0.74, "learning_rate": 5.512360233181213e-05, "loss": 1.0505, "step": 7258 }, { "epoch": 0.74, "learning_rate": 5.512229436416797e-05, "loss": 0.5613, "step": 7259 }, { "epoch": 0.74, "learning_rate": 5.512098623665508e-05, "loss": 1.5283, "step": 7260 }, { "epoch": 0.74, "learning_rate": 5.51196779492818e-05, "loss": 1.059, "step": 7261 }, { "epoch": 0.74, "learning_rate": 5.511836950205644e-05, "loss": 0.7795, "step": 7262 }, { "epoch": 0.74, "learning_rate": 5.511706089498733e-05, "loss": 0.2271, "step": 7263 }, { "epoch": 0.74, "learning_rate": 5.511575212808281e-05, "loss": 1.0856, "step": 7264 }, { "epoch": 0.74, "learning_rate": 5.511444320135119e-05, "loss": 1.4058, "step": 7265 }, { "epoch": 0.74, "learning_rate": 5.5113134114800814e-05, "loss": 0.5043, "step": 7266 }, { "epoch": 0.74, "learning_rate": 5.511182486843999e-05, "loss": 0.5936, "step": 7267 }, { "epoch": 0.74, "learning_rate": 5.511051546227708e-05, "loss": 0.5414, "step": 7268 }, { "epoch": 0.74, "learning_rate": 5.51092058963204e-05, "loss": 0.8094, "step": 7269 }, { "epoch": 0.74, "learning_rate": 5.5107896170578284e-05, "loss": 1.2924, "step": 7270 }, { "epoch": 0.74, "learning_rate": 5.5106586285059066e-05, "loss": 0.2085, "step": 7271 }, { "epoch": 0.74, "learning_rate": 5.510527623977109e-05, "loss": 0.5964, "step": 7272 }, { "epoch": 0.74, "learning_rate": 5.5103966034722686e-05, "loss": 0.537, "step": 7273 }, { "epoch": 0.74, "learning_rate": 5.5102655669922194e-05, "loss": 0.9855, "step": 7274 }, { "epoch": 0.74, "learning_rate": 5.510134514537794e-05, "loss": 1.0757, "step": 7275 }, { "epoch": 0.74, "learning_rate": 5.510003446109828e-05, "loss": 0.2866, "step": 7276 }, { "epoch": 0.74, "learning_rate": 5.509872361709155e-05, "loss": 0.8452, "step": 7277 }, { "epoch": 0.74, "learning_rate": 5.5097412613366086e-05, "loss": 1.4019, "step": 7278 }, { "epoch": 0.74, "learning_rate": 5.5096101449930246e-05, "loss": 0.42, "step": 7279 }, { "epoch": 0.74, "learning_rate": 5.5094790126792354e-05, "loss": 0.7051, "step": 7280 }, { "epoch": 0.74, "learning_rate": 5.509347864396077e-05, "loss": 1.2753, "step": 7281 }, { "epoch": 0.74, "learning_rate": 5.509216700144383e-05, "loss": 1.3615, "step": 7282 }, { "epoch": 0.74, "learning_rate": 5.509085519924989e-05, "loss": 0.6427, "step": 7283 }, { "epoch": 0.74, "learning_rate": 5.508954323738729e-05, "loss": 1.2987, "step": 7284 }, { "epoch": 0.74, "learning_rate": 5.5088231115864373e-05, "loss": 1.3577, "step": 7285 }, { "epoch": 0.74, "learning_rate": 5.508691883468951e-05, "loss": 1.237, "step": 7286 }, { "epoch": 0.74, "learning_rate": 5.5085606393871035e-05, "loss": 0.9248, "step": 7287 }, { "epoch": 0.74, "learning_rate": 5.5084293793417305e-05, "loss": 0.5969, "step": 7288 }, { "epoch": 0.74, "learning_rate": 5.508298103333667e-05, "loss": 0.9855, "step": 7289 }, { "epoch": 0.74, "learning_rate": 5.5081668113637483e-05, "loss": 0.4465, "step": 7290 }, { "epoch": 0.74, "learning_rate": 5.508035503432811e-05, "loss": 0.4236, "step": 7291 }, { "epoch": 0.74, "learning_rate": 5.507904179541689e-05, "loss": 0.499, "step": 7292 }, { "epoch": 0.74, "learning_rate": 5.507772839691219e-05, "loss": 1.9008, "step": 7293 }, { "epoch": 0.74, "learning_rate": 5.507641483882237e-05, "loss": 0.4713, "step": 7294 }, { "epoch": 0.74, "learning_rate": 5.507510112115578e-05, "loss": 0.3819, "step": 7295 }, { "epoch": 0.74, "learning_rate": 5.507378724392079e-05, "loss": 1.591, "step": 7296 }, { "epoch": 0.74, "learning_rate": 5.507247320712576e-05, "loss": 0.5914, "step": 7297 }, { "epoch": 0.74, "learning_rate": 5.507115901077904e-05, "loss": 1.2186, "step": 7298 }, { "epoch": 0.74, "learning_rate": 5.506984465488901e-05, "loss": 0.4954, "step": 7299 }, { "epoch": 0.74, "learning_rate": 5.506853013946402e-05, "loss": 1.3613, "step": 7300 }, { "epoch": 0.74, "learning_rate": 5.5067215464512436e-05, "loss": 0.9348, "step": 7301 }, { "epoch": 0.74, "learning_rate": 5.506590063004263e-05, "loss": 0.6012, "step": 7302 }, { "epoch": 0.74, "learning_rate": 5.506458563606297e-05, "loss": 0.3912, "step": 7303 }, { "epoch": 0.74, "learning_rate": 5.506327048258183e-05, "loss": 0.8297, "step": 7304 }, { "epoch": 0.74, "learning_rate": 5.506195516960757e-05, "loss": 0.9288, "step": 7305 }, { "epoch": 0.74, "learning_rate": 5.5060639697148556e-05, "loss": 1.4175, "step": 7306 }, { "epoch": 0.74, "learning_rate": 5.505932406521317e-05, "loss": 1.7531, "step": 7307 }, { "epoch": 0.74, "learning_rate": 5.505800827380977e-05, "loss": 1.3611, "step": 7308 }, { "epoch": 0.74, "learning_rate": 5.505669232294674e-05, "loss": 1.2297, "step": 7309 }, { "epoch": 0.74, "learning_rate": 5.505537621263246e-05, "loss": 1.2011, "step": 7310 }, { "epoch": 0.74, "learning_rate": 5.505405994287529e-05, "loss": 0.2656, "step": 7311 }, { "epoch": 0.74, "learning_rate": 5.505274351368362e-05, "loss": 1.3839, "step": 7312 }, { "epoch": 0.74, "learning_rate": 5.5051426925065814e-05, "loss": 1.1269, "step": 7313 }, { "epoch": 0.74, "learning_rate": 5.505011017703026e-05, "loss": 0.8002, "step": 7314 }, { "epoch": 0.74, "learning_rate": 5.504879326958534e-05, "loss": 1.4843, "step": 7315 }, { "epoch": 0.74, "learning_rate": 5.504747620273942e-05, "loss": 1.1399, "step": 7316 }, { "epoch": 0.74, "learning_rate": 5.504615897650089e-05, "loss": 0.2812, "step": 7317 }, { "epoch": 0.74, "learning_rate": 5.504484159087813e-05, "loss": 0.3666, "step": 7318 }, { "epoch": 0.74, "learning_rate": 5.504352404587954e-05, "loss": 0.5798, "step": 7319 }, { "epoch": 0.74, "learning_rate": 5.5042206341513475e-05, "loss": 0.8566, "step": 7320 }, { "epoch": 0.74, "learning_rate": 5.5040888477788345e-05, "loss": 0.3414, "step": 7321 }, { "epoch": 0.74, "learning_rate": 5.503957045471252e-05, "loss": 0.601, "step": 7322 }, { "epoch": 0.74, "learning_rate": 5.50382522722944e-05, "loss": 0.643, "step": 7323 }, { "epoch": 0.74, "learning_rate": 5.5036933930542364e-05, "loss": 0.8416, "step": 7324 }, { "epoch": 0.74, "learning_rate": 5.5035615429464806e-05, "loss": 0.3331, "step": 7325 }, { "epoch": 0.74, "learning_rate": 5.5034296769070115e-05, "loss": 1.4041, "step": 7326 }, { "epoch": 0.74, "learning_rate": 5.503297794936668e-05, "loss": 0.8626, "step": 7327 }, { "epoch": 0.74, "learning_rate": 5.50316589703629e-05, "loss": 0.6573, "step": 7328 }, { "epoch": 0.74, "learning_rate": 5.503033983206717e-05, "loss": 0.5973, "step": 7329 }, { "epoch": 0.74, "learning_rate": 5.5029020534487865e-05, "loss": 1.3462, "step": 7330 }, { "epoch": 0.74, "learning_rate": 5.5027701077633407e-05, "loss": 1.4108, "step": 7331 }, { "epoch": 0.74, "learning_rate": 5.502638146151218e-05, "loss": 0.574, "step": 7332 }, { "epoch": 0.74, "learning_rate": 5.502506168613258e-05, "loss": 0.4938, "step": 7333 }, { "epoch": 0.74, "learning_rate": 5.5023741751503005e-05, "loss": 1.0156, "step": 7334 }, { "epoch": 0.74, "learning_rate": 5.5022421657631855e-05, "loss": 0.4943, "step": 7335 }, { "epoch": 0.74, "learning_rate": 5.502110140452753e-05, "loss": 0.9206, "step": 7336 }, { "epoch": 0.75, "learning_rate": 5.501978099219844e-05, "loss": 0.8593, "step": 7337 }, { "epoch": 0.75, "learning_rate": 5.5018460420652986e-05, "loss": 0.4293, "step": 7338 }, { "epoch": 0.75, "learning_rate": 5.5017139689899566e-05, "loss": 0.4895, "step": 7339 }, { "epoch": 0.75, "learning_rate": 5.501581879994658e-05, "loss": 1.3424, "step": 7340 }, { "epoch": 0.75, "learning_rate": 5.501449775080245e-05, "loss": 0.9853, "step": 7341 }, { "epoch": 0.75, "learning_rate": 5.5013176542475554e-05, "loss": 0.5601, "step": 7342 }, { "epoch": 0.75, "learning_rate": 5.501185517497433e-05, "loss": 1.5324, "step": 7343 }, { "epoch": 0.75, "learning_rate": 5.5010533648307174e-05, "loss": 0.8319, "step": 7344 }, { "epoch": 0.75, "learning_rate": 5.50092119624825e-05, "loss": 1.5009, "step": 7345 }, { "epoch": 0.75, "learning_rate": 5.500789011750871e-05, "loss": 1.1178, "step": 7346 }, { "epoch": 0.75, "learning_rate": 5.500656811339423e-05, "loss": 0.7342, "step": 7347 }, { "epoch": 0.75, "learning_rate": 5.500524595014745e-05, "loss": 0.8661, "step": 7348 }, { "epoch": 0.75, "learning_rate": 5.50039236277768e-05, "loss": 0.1422, "step": 7349 }, { "epoch": 0.75, "learning_rate": 5.50026011462907e-05, "loss": 1.4273, "step": 7350 }, { "epoch": 0.75, "learning_rate": 5.5001278505697546e-05, "loss": 1.4001, "step": 7351 }, { "epoch": 0.75, "learning_rate": 5.4999955706005775e-05, "loss": 0.4429, "step": 7352 }, { "epoch": 0.75, "learning_rate": 5.4998632747223795e-05, "loss": 1.216, "step": 7353 }, { "epoch": 0.75, "learning_rate": 5.499730962936002e-05, "loss": 0.676, "step": 7354 }, { "epoch": 0.75, "learning_rate": 5.4995986352422886e-05, "loss": 1.1608, "step": 7355 }, { "epoch": 0.75, "learning_rate": 5.49946629164208e-05, "loss": 0.7192, "step": 7356 }, { "epoch": 0.75, "learning_rate": 5.499333932136218e-05, "loss": 0.3984, "step": 7357 }, { "epoch": 0.75, "learning_rate": 5.499201556725546e-05, "loss": 0.4416, "step": 7358 }, { "epoch": 0.75, "learning_rate": 5.4990691654109075e-05, "loss": 0.344, "step": 7359 }, { "epoch": 0.75, "learning_rate": 5.4989367581931424e-05, "loss": 0.7133, "step": 7360 }, { "epoch": 0.75, "learning_rate": 5.4988043350730946e-05, "loss": 0.7942, "step": 7361 }, { "epoch": 0.75, "learning_rate": 5.498671896051606e-05, "loss": 1.5574, "step": 7362 }, { "epoch": 0.75, "learning_rate": 5.4985394411295205e-05, "loss": 1.1749, "step": 7363 }, { "epoch": 0.75, "learning_rate": 5.498406970307681e-05, "loss": 0.5732, "step": 7364 }, { "epoch": 0.75, "learning_rate": 5.498274483586928e-05, "loss": 1.4778, "step": 7365 }, { "epoch": 0.75, "learning_rate": 5.4981419809681085e-05, "loss": 0.7788, "step": 7366 }, { "epoch": 0.75, "learning_rate": 5.498009462452064e-05, "loss": 1.6339, "step": 7367 }, { "epoch": 0.75, "learning_rate": 5.4978769280396364e-05, "loss": 0.7673, "step": 7368 }, { "epoch": 0.75, "learning_rate": 5.497744377731671e-05, "loss": 0.8582, "step": 7369 }, { "epoch": 0.75, "learning_rate": 5.49761181152901e-05, "loss": 1.2027, "step": 7370 }, { "epoch": 0.75, "learning_rate": 5.497479229432498e-05, "loss": 1.483, "step": 7371 }, { "epoch": 0.75, "learning_rate": 5.497346631442978e-05, "loss": 0.4193, "step": 7372 }, { "epoch": 0.75, "learning_rate": 5.497214017561294e-05, "loss": 1.1486, "step": 7373 }, { "epoch": 0.75, "learning_rate": 5.49708138778829e-05, "loss": 1.594, "step": 7374 }, { "epoch": 0.75, "learning_rate": 5.4969487421248106e-05, "loss": 0.7568, "step": 7375 }, { "epoch": 0.75, "learning_rate": 5.4968160805716995e-05, "loss": 0.6739, "step": 7376 }, { "epoch": 0.75, "learning_rate": 5.496683403129799e-05, "loss": 1.152, "step": 7377 }, { "epoch": 0.75, "learning_rate": 5.496550709799957e-05, "loss": 1.1865, "step": 7378 }, { "epoch": 0.75, "learning_rate": 5.4964180005830145e-05, "loss": 0.6176, "step": 7379 }, { "epoch": 0.75, "learning_rate": 5.4962852754798185e-05, "loss": 0.4857, "step": 7380 }, { "epoch": 0.75, "learning_rate": 5.496152534491212e-05, "loss": 1.6724, "step": 7381 }, { "epoch": 0.75, "learning_rate": 5.4960197776180404e-05, "loss": 0.2651, "step": 7382 }, { "epoch": 0.75, "learning_rate": 5.495887004861149e-05, "loss": 0.9353, "step": 7383 }, { "epoch": 0.75, "learning_rate": 5.495754216221382e-05, "loss": 0.8185, "step": 7384 }, { "epoch": 0.75, "learning_rate": 5.495621411699585e-05, "loss": 1.3677, "step": 7385 }, { "epoch": 0.75, "learning_rate": 5.495488591296602e-05, "loss": 0.7358, "step": 7386 }, { "epoch": 0.75, "learning_rate": 5.495355755013279e-05, "loss": 1.1783, "step": 7387 }, { "epoch": 0.75, "learning_rate": 5.4952229028504615e-05, "loss": 0.8779, "step": 7388 }, { "epoch": 0.75, "learning_rate": 5.495090034808995e-05, "loss": 0.1161, "step": 7389 }, { "epoch": 0.75, "learning_rate": 5.494957150889724e-05, "loss": 0.474, "step": 7390 }, { "epoch": 0.75, "learning_rate": 5.494824251093496e-05, "loss": 1.0989, "step": 7391 }, { "epoch": 0.75, "learning_rate": 5.494691335421154e-05, "loss": 1.7675, "step": 7392 }, { "epoch": 0.75, "learning_rate": 5.494558403873546e-05, "loss": 1.1771, "step": 7393 }, { "epoch": 0.75, "learning_rate": 5.494425456451517e-05, "loss": 0.5864, "step": 7394 }, { "epoch": 0.75, "learning_rate": 5.494292493155914e-05, "loss": 1.6968, "step": 7395 }, { "epoch": 0.75, "learning_rate": 5.4941595139875825e-05, "loss": 0.6608, "step": 7396 }, { "epoch": 0.75, "learning_rate": 5.4940265189473685e-05, "loss": 0.428, "step": 7397 }, { "epoch": 0.75, "learning_rate": 5.493893508036118e-05, "loss": 0.8658, "step": 7398 }, { "epoch": 0.75, "learning_rate": 5.4937604812546785e-05, "loss": 0.4771, "step": 7399 }, { "epoch": 0.75, "learning_rate": 5.493627438603896e-05, "loss": 0.608, "step": 7400 }, { "epoch": 0.75, "learning_rate": 5.493494380084617e-05, "loss": 0.6793, "step": 7401 }, { "epoch": 0.75, "learning_rate": 5.493361305697688e-05, "loss": 1.3231, "step": 7402 }, { "epoch": 0.75, "learning_rate": 5.493228215443956e-05, "loss": 1.0503, "step": 7403 }, { "epoch": 0.75, "learning_rate": 5.493095109324268e-05, "loss": 0.873, "step": 7404 }, { "epoch": 0.75, "learning_rate": 5.492961987339472e-05, "loss": 1.8314, "step": 7405 }, { "epoch": 0.75, "learning_rate": 5.492828849490414e-05, "loss": 1.1116, "step": 7406 }, { "epoch": 0.75, "learning_rate": 5.492695695777942e-05, "loss": 1.6006, "step": 7407 }, { "epoch": 0.75, "learning_rate": 5.492562526202902e-05, "loss": 1.3, "step": 7408 }, { "epoch": 0.75, "learning_rate": 5.492429340766142e-05, "loss": 0.4358, "step": 7409 }, { "epoch": 0.75, "learning_rate": 5.492296139468511e-05, "loss": 0.397, "step": 7410 }, { "epoch": 0.75, "learning_rate": 5.4921629223108545e-05, "loss": 0.6473, "step": 7411 }, { "epoch": 0.75, "learning_rate": 5.492029689294022e-05, "loss": 1.1457, "step": 7412 }, { "epoch": 0.75, "learning_rate": 5.49189644041886e-05, "loss": 0.9215, "step": 7413 }, { "epoch": 0.75, "learning_rate": 5.491763175686218e-05, "loss": 0.9728, "step": 7414 }, { "epoch": 0.75, "learning_rate": 5.491629895096942e-05, "loss": 0.9329, "step": 7415 }, { "epoch": 0.75, "learning_rate": 5.491496598651882e-05, "loss": 0.8327, "step": 7416 }, { "epoch": 0.75, "learning_rate": 5.491363286351884e-05, "loss": 1.6604, "step": 7417 }, { "epoch": 0.75, "learning_rate": 5.491229958197799e-05, "loss": 1.234, "step": 7418 }, { "epoch": 0.75, "learning_rate": 5.491096614190475e-05, "loss": 0.4967, "step": 7419 }, { "epoch": 0.75, "learning_rate": 5.490963254330758e-05, "loss": 0.8653, "step": 7420 }, { "epoch": 0.75, "learning_rate": 5.4908298786194994e-05, "loss": 1.0819, "step": 7421 }, { "epoch": 0.75, "learning_rate": 5.4906964870575465e-05, "loss": 1.0113, "step": 7422 }, { "epoch": 0.75, "learning_rate": 5.49056307964575e-05, "loss": 0.7656, "step": 7423 }, { "epoch": 0.75, "learning_rate": 5.490429656384956e-05, "loss": 0.4124, "step": 7424 }, { "epoch": 0.75, "learning_rate": 5.4902962172760164e-05, "loss": 1.5976, "step": 7425 }, { "epoch": 0.75, "learning_rate": 5.4901627623197786e-05, "loss": 1.0677, "step": 7426 }, { "epoch": 0.75, "learning_rate": 5.490029291517092e-05, "loss": 0.593, "step": 7427 }, { "epoch": 0.75, "learning_rate": 5.489895804868805e-05, "loss": 1.2255, "step": 7428 }, { "epoch": 0.75, "learning_rate": 5.48976230237577e-05, "loss": 0.6784, "step": 7429 }, { "epoch": 0.75, "learning_rate": 5.489628784038834e-05, "loss": 0.4693, "step": 7430 }, { "epoch": 0.75, "learning_rate": 5.4894952498588486e-05, "loss": 0.9099, "step": 7431 }, { "epoch": 0.75, "learning_rate": 5.489361699836662e-05, "loss": 1.0175, "step": 7432 }, { "epoch": 0.75, "learning_rate": 5.4892281339731245e-05, "loss": 0.8882, "step": 7433 }, { "epoch": 0.75, "learning_rate": 5.4890945522690855e-05, "loss": 1.3266, "step": 7434 }, { "epoch": 0.75, "learning_rate": 5.488960954725396e-05, "loss": 0.7025, "step": 7435 }, { "epoch": 0.76, "learning_rate": 5.4888273413429064e-05, "loss": 1.7204, "step": 7436 }, { "epoch": 0.76, "learning_rate": 5.488693712122466e-05, "loss": 0.8148, "step": 7437 }, { "epoch": 0.76, "learning_rate": 5.488560067064926e-05, "loss": 0.3027, "step": 7438 }, { "epoch": 0.76, "learning_rate": 5.488426406171136e-05, "loss": 0.6446, "step": 7439 }, { "epoch": 0.76, "learning_rate": 5.488292729441947e-05, "loss": 1.2667, "step": 7440 }, { "epoch": 0.76, "learning_rate": 5.4881590368782085e-05, "loss": 0.7546, "step": 7441 }, { "epoch": 0.76, "learning_rate": 5.488025328480775e-05, "loss": 0.2973, "step": 7442 }, { "epoch": 0.76, "learning_rate": 5.4878916042504924e-05, "loss": 1.2887, "step": 7443 }, { "epoch": 0.76, "learning_rate": 5.4877578641882155e-05, "loss": 0.4528, "step": 7444 }, { "epoch": 0.76, "learning_rate": 5.487624108294793e-05, "loss": 0.3147, "step": 7445 }, { "epoch": 0.76, "learning_rate": 5.487490336571078e-05, "loss": 0.4805, "step": 7446 }, { "epoch": 0.76, "learning_rate": 5.48735654901792e-05, "loss": 0.4988, "step": 7447 }, { "epoch": 0.76, "learning_rate": 5.4872227456361714e-05, "loss": 1.193, "step": 7448 }, { "epoch": 0.76, "learning_rate": 5.487088926426683e-05, "loss": 1.1394, "step": 7449 }, { "epoch": 0.76, "learning_rate": 5.486955091390308e-05, "loss": 0.6152, "step": 7450 }, { "epoch": 0.76, "learning_rate": 5.486821240527896e-05, "loss": 0.7913, "step": 7451 }, { "epoch": 0.76, "learning_rate": 5.4866873738403e-05, "loss": 0.7029, "step": 7452 }, { "epoch": 0.76, "learning_rate": 5.486553491328371e-05, "loss": 1.1572, "step": 7453 }, { "epoch": 0.76, "learning_rate": 5.4864195929929624e-05, "loss": 1.1858, "step": 7454 }, { "epoch": 0.76, "learning_rate": 5.486285678834925e-05, "loss": 1.851, "step": 7455 }, { "epoch": 0.76, "learning_rate": 5.4861517488551104e-05, "loss": 1.3557, "step": 7456 }, { "epoch": 0.76, "learning_rate": 5.486017803054373e-05, "loss": 0.8899, "step": 7457 }, { "epoch": 0.76, "learning_rate": 5.485883841433564e-05, "loss": 0.5555, "step": 7458 }, { "epoch": 0.76, "learning_rate": 5.485749863993536e-05, "loss": 1.1227, "step": 7459 }, { "epoch": 0.76, "learning_rate": 5.4856158707351404e-05, "loss": 0.9609, "step": 7460 }, { "epoch": 0.76, "learning_rate": 5.485481861659232e-05, "loss": 0.8208, "step": 7461 }, { "epoch": 0.76, "learning_rate": 5.485347836766663e-05, "loss": 0.3911, "step": 7462 }, { "epoch": 0.76, "learning_rate": 5.485213796058285e-05, "loss": 1.361, "step": 7463 }, { "epoch": 0.76, "learning_rate": 5.4850797395349516e-05, "loss": 1.7423, "step": 7464 }, { "epoch": 0.76, "learning_rate": 5.484945667197517e-05, "loss": 1.2149, "step": 7465 }, { "epoch": 0.76, "learning_rate": 5.4848115790468326e-05, "loss": 1.2689, "step": 7466 }, { "epoch": 0.76, "learning_rate": 5.484677475083752e-05, "loss": 1.1077, "step": 7467 }, { "epoch": 0.76, "learning_rate": 5.484543355309131e-05, "loss": 1.173, "step": 7468 }, { "epoch": 0.76, "learning_rate": 5.48440921972382e-05, "loss": 0.3627, "step": 7469 }, { "epoch": 0.76, "learning_rate": 5.484275068328674e-05, "loss": 1.0101, "step": 7470 }, { "epoch": 0.76, "learning_rate": 5.484140901124548e-05, "loss": 1.2901, "step": 7471 }, { "epoch": 0.76, "learning_rate": 5.484006718112292e-05, "loss": 0.7954, "step": 7472 }, { "epoch": 0.76, "learning_rate": 5.483872519292763e-05, "loss": 1.0681, "step": 7473 }, { "epoch": 0.76, "learning_rate": 5.4837383046668146e-05, "loss": 1.2877, "step": 7474 }, { "epoch": 0.76, "learning_rate": 5.483604074235301e-05, "loss": 0.4526, "step": 7475 }, { "epoch": 0.76, "learning_rate": 5.4834698279990745e-05, "loss": 1.334, "step": 7476 }, { "epoch": 0.76, "learning_rate": 5.4833355659589915e-05, "loss": 0.3523, "step": 7477 }, { "epoch": 0.76, "learning_rate": 5.483201288115906e-05, "loss": 0.4896, "step": 7478 }, { "epoch": 0.76, "learning_rate": 5.483066994470672e-05, "loss": 1.2445, "step": 7479 }, { "epoch": 0.76, "learning_rate": 5.482932685024144e-05, "loss": 0.4842, "step": 7480 }, { "epoch": 0.76, "learning_rate": 5.482798359777177e-05, "loss": 1.3156, "step": 7481 }, { "epoch": 0.76, "learning_rate": 5.4826640187306256e-05, "loss": 0.709, "step": 7482 }, { "epoch": 0.76, "learning_rate": 5.482529661885345e-05, "loss": 0.3022, "step": 7483 }, { "epoch": 0.76, "learning_rate": 5.482395289242191e-05, "loss": 0.1986, "step": 7484 }, { "epoch": 0.76, "learning_rate": 5.482260900802016e-05, "loss": 1.5512, "step": 7485 }, { "epoch": 0.76, "learning_rate": 5.482126496565678e-05, "loss": 0.2509, "step": 7486 }, { "epoch": 0.76, "learning_rate": 5.4819920765340314e-05, "loss": 0.5115, "step": 7487 }, { "epoch": 0.76, "learning_rate": 5.48185764070793e-05, "loss": 1.7899, "step": 7488 }, { "epoch": 0.76, "learning_rate": 5.4817231890882324e-05, "loss": 0.57, "step": 7489 }, { "epoch": 0.76, "learning_rate": 5.4815887216757916e-05, "loss": 0.7195, "step": 7490 }, { "epoch": 0.76, "learning_rate": 5.481454238471464e-05, "loss": 0.407, "step": 7491 }, { "epoch": 0.76, "learning_rate": 5.4813197394761064e-05, "loss": 1.5386, "step": 7492 }, { "epoch": 0.76, "learning_rate": 5.481185224690573e-05, "loss": 1.1244, "step": 7493 }, { "epoch": 0.76, "learning_rate": 5.4810506941157216e-05, "loss": 1.5645, "step": 7494 }, { "epoch": 0.76, "learning_rate": 5.480916147752408e-05, "loss": 0.7817, "step": 7495 }, { "epoch": 0.76, "learning_rate": 5.480781585601486e-05, "loss": 1.0056, "step": 7496 }, { "epoch": 0.76, "learning_rate": 5.480647007663815e-05, "loss": 0.5909, "step": 7497 }, { "epoch": 0.76, "learning_rate": 5.48051241394025e-05, "loss": 1.1215, "step": 7498 }, { "epoch": 0.76, "learning_rate": 5.480377804431648e-05, "loss": 1.2559, "step": 7499 }, { "epoch": 0.76, "learning_rate": 5.4802431791388636e-05, "loss": 0.4832, "step": 7500 }, { "epoch": 0.76, "learning_rate": 5.480108538062756e-05, "loss": 0.8764, "step": 7501 }, { "epoch": 0.76, "learning_rate": 5.479973881204182e-05, "loss": 0.5197, "step": 7502 }, { "epoch": 0.76, "learning_rate": 5.479839208563997e-05, "loss": 1.0768, "step": 7503 }, { "epoch": 0.76, "learning_rate": 5.479704520143058e-05, "loss": 2.0009, "step": 7504 }, { "epoch": 0.76, "learning_rate": 5.479569815942224e-05, "loss": 1.5703, "step": 7505 }, { "epoch": 0.76, "learning_rate": 5.4794350959623496e-05, "loss": 1.0272, "step": 7506 }, { "epoch": 0.76, "learning_rate": 5.4793003602042955e-05, "loss": 1.6256, "step": 7507 }, { "epoch": 0.76, "learning_rate": 5.479165608668915e-05, "loss": 0.6903, "step": 7508 }, { "epoch": 0.76, "learning_rate": 5.4790308413570685e-05, "loss": 1.3513, "step": 7509 }, { "epoch": 0.76, "learning_rate": 5.4788960582696134e-05, "loss": 0.3763, "step": 7510 }, { "epoch": 0.76, "learning_rate": 5.478761259407406e-05, "loss": 0.2484, "step": 7511 }, { "epoch": 0.76, "learning_rate": 5.478626444771305e-05, "loss": 1.1988, "step": 7512 }, { "epoch": 0.76, "learning_rate": 5.478491614362169e-05, "loss": 0.6532, "step": 7513 }, { "epoch": 0.76, "learning_rate": 5.4783567681808544e-05, "loss": 0.199, "step": 7514 }, { "epoch": 0.76, "learning_rate": 5.4782219062282194e-05, "loss": 1.0466, "step": 7515 }, { "epoch": 0.76, "learning_rate": 5.478087028505124e-05, "loss": 1.5196, "step": 7516 }, { "epoch": 0.76, "learning_rate": 5.477952135012426e-05, "loss": 0.5463, "step": 7517 }, { "epoch": 0.76, "learning_rate": 5.477817225750983e-05, "loss": 0.4053, "step": 7518 }, { "epoch": 0.76, "learning_rate": 5.477682300721653e-05, "loss": 1.2824, "step": 7519 }, { "epoch": 0.76, "learning_rate": 5.477547359925296e-05, "loss": 0.2343, "step": 7520 }, { "epoch": 0.76, "learning_rate": 5.477412403362769e-05, "loss": 0.9973, "step": 7521 }, { "epoch": 0.76, "learning_rate": 5.4772774310349335e-05, "loss": 0.9631, "step": 7522 }, { "epoch": 0.76, "learning_rate": 5.477142442942646e-05, "loss": 0.9548, "step": 7523 }, { "epoch": 0.76, "learning_rate": 5.477007439086767e-05, "loss": 0.6665, "step": 7524 }, { "epoch": 0.76, "learning_rate": 5.476872419468154e-05, "loss": 0.3607, "step": 7525 }, { "epoch": 0.76, "learning_rate": 5.476737384087668e-05, "loss": 1.0008, "step": 7526 }, { "epoch": 0.76, "learning_rate": 5.476602332946167e-05, "loss": 0.5914, "step": 7527 }, { "epoch": 0.76, "learning_rate": 5.476467266044511e-05, "loss": 1.2999, "step": 7528 }, { "epoch": 0.76, "learning_rate": 5.476332183383559e-05, "loss": 1.4162, "step": 7529 }, { "epoch": 0.76, "learning_rate": 5.476197084964172e-05, "loss": 1.7668, "step": 7530 }, { "epoch": 0.76, "learning_rate": 5.4760619707872086e-05, "loss": 0.6571, "step": 7531 }, { "epoch": 0.76, "learning_rate": 5.4759268408535275e-05, "loss": 0.5498, "step": 7532 }, { "epoch": 0.76, "learning_rate": 5.475791695163991e-05, "loss": 1.0757, "step": 7533 }, { "epoch": 0.77, "learning_rate": 5.475656533719458e-05, "loss": 1.1372, "step": 7534 }, { "epoch": 0.77, "learning_rate": 5.4755213565207884e-05, "loss": 1.1364, "step": 7535 }, { "epoch": 0.77, "learning_rate": 5.4753861635688426e-05, "loss": 0.2849, "step": 7536 }, { "epoch": 0.77, "learning_rate": 5.47525095486448e-05, "loss": 0.9046, "step": 7537 }, { "epoch": 0.77, "learning_rate": 5.4751157304085635e-05, "loss": 0.5815, "step": 7538 }, { "epoch": 0.77, "learning_rate": 5.4749804902019514e-05, "loss": 0.9711, "step": 7539 }, { "epoch": 0.77, "learning_rate": 5.4748452342455056e-05, "loss": 0.5323, "step": 7540 }, { "epoch": 0.77, "learning_rate": 5.4747099625400854e-05, "loss": 1.5477, "step": 7541 }, { "epoch": 0.77, "learning_rate": 5.474574675086553e-05, "loss": 1.4479, "step": 7542 }, { "epoch": 0.77, "learning_rate": 5.4744393718857684e-05, "loss": 1.433, "step": 7543 }, { "epoch": 0.77, "learning_rate": 5.4743040529385926e-05, "loss": 0.7479, "step": 7544 }, { "epoch": 0.77, "learning_rate": 5.4741687182458875e-05, "loss": 1.8094, "step": 7545 }, { "epoch": 0.77, "learning_rate": 5.474033367808514e-05, "loss": 0.7825, "step": 7546 }, { "epoch": 0.77, "learning_rate": 5.473898001627334e-05, "loss": 0.55, "step": 7547 }, { "epoch": 0.77, "learning_rate": 5.473762619703207e-05, "loss": 0.2618, "step": 7548 }, { "epoch": 0.77, "learning_rate": 5.473627222036996e-05, "loss": 1.4449, "step": 7549 }, { "epoch": 0.77, "learning_rate": 5.473491808629562e-05, "loss": 1.1887, "step": 7550 }, { "epoch": 0.77, "learning_rate": 5.473356379481769e-05, "loss": 1.2635, "step": 7551 }, { "epoch": 0.77, "learning_rate": 5.473220934594475e-05, "loss": 0.1796, "step": 7552 }, { "epoch": 0.77, "learning_rate": 5.473085473968544e-05, "loss": 0.3097, "step": 7553 }, { "epoch": 0.77, "learning_rate": 5.472949997604839e-05, "loss": 0.3389, "step": 7554 }, { "epoch": 0.77, "learning_rate": 5.47281450550422e-05, "loss": 1.3103, "step": 7555 }, { "epoch": 0.77, "learning_rate": 5.4726789976675496e-05, "loss": 1.4228, "step": 7556 }, { "epoch": 0.77, "learning_rate": 5.472543474095692e-05, "loss": 1.0454, "step": 7557 }, { "epoch": 0.77, "learning_rate": 5.4724079347895076e-05, "loss": 1.4184, "step": 7558 }, { "epoch": 0.77, "learning_rate": 5.47227237974986e-05, "loss": 0.7862, "step": 7559 }, { "epoch": 0.77, "learning_rate": 5.4721368089776114e-05, "loss": 0.5774, "step": 7560 }, { "epoch": 0.77, "learning_rate": 5.472001222473625e-05, "loss": 0.7694, "step": 7561 }, { "epoch": 0.77, "learning_rate": 5.471865620238763e-05, "loss": 1.6764, "step": 7562 }, { "epoch": 0.77, "learning_rate": 5.471730002273888e-05, "loss": 0.2778, "step": 7563 }, { "epoch": 0.77, "learning_rate": 5.471594368579864e-05, "loss": 1.4638, "step": 7564 }, { "epoch": 0.77, "learning_rate": 5.471458719157553e-05, "loss": 1.0877, "step": 7565 }, { "epoch": 0.77, "learning_rate": 5.471323054007821e-05, "loss": 0.4922, "step": 7566 }, { "epoch": 0.77, "learning_rate": 5.471187373131527e-05, "loss": 1.2725, "step": 7567 }, { "epoch": 0.77, "learning_rate": 5.4710516765295374e-05, "loss": 1.9185, "step": 7568 }, { "epoch": 0.77, "learning_rate": 5.470915964202715e-05, "loss": 0.5981, "step": 7569 }, { "epoch": 0.77, "learning_rate": 5.470780236151923e-05, "loss": 0.2997, "step": 7570 }, { "epoch": 0.77, "learning_rate": 5.4706444923780254e-05, "loss": 0.7375, "step": 7571 }, { "epoch": 0.77, "learning_rate": 5.470508732881887e-05, "loss": 0.291, "step": 7572 }, { "epoch": 0.77, "learning_rate": 5.47037295766437e-05, "loss": 0.8458, "step": 7573 }, { "epoch": 0.77, "learning_rate": 5.4702371667263405e-05, "loss": 0.5251, "step": 7574 }, { "epoch": 0.77, "learning_rate": 5.47010136006866e-05, "loss": 0.7405, "step": 7575 }, { "epoch": 0.77, "learning_rate": 5.469965537692195e-05, "loss": 1.2186, "step": 7576 }, { "epoch": 0.77, "learning_rate": 5.469829699597809e-05, "loss": 1.3863, "step": 7577 }, { "epoch": 0.77, "learning_rate": 5.469693845786366e-05, "loss": 0.4555, "step": 7578 }, { "epoch": 0.77, "learning_rate": 5.469557976258731e-05, "loss": 0.3635, "step": 7579 }, { "epoch": 0.77, "learning_rate": 5.469422091015768e-05, "loss": 0.4563, "step": 7580 }, { "epoch": 0.77, "learning_rate": 5.469286190058343e-05, "loss": 0.9036, "step": 7581 }, { "epoch": 0.77, "learning_rate": 5.469150273387319e-05, "loss": 1.143, "step": 7582 }, { "epoch": 0.77, "learning_rate": 5.469014341003563e-05, "loss": 1.4128, "step": 7583 }, { "epoch": 0.77, "learning_rate": 5.468878392907938e-05, "loss": 1.0662, "step": 7584 }, { "epoch": 0.77, "learning_rate": 5.468742429101312e-05, "loss": 0.2236, "step": 7585 }, { "epoch": 0.77, "learning_rate": 5.4686064495845464e-05, "loss": 0.5563, "step": 7586 }, { "epoch": 0.77, "learning_rate": 5.4684704543585094e-05, "loss": 1.5614, "step": 7587 }, { "epoch": 0.77, "learning_rate": 5.468334443424065e-05, "loss": 1.237, "step": 7588 }, { "epoch": 0.77, "learning_rate": 5.4681984167820793e-05, "loss": 0.6686, "step": 7589 }, { "epoch": 0.77, "learning_rate": 5.468062374433418e-05, "loss": 0.2802, "step": 7590 }, { "epoch": 0.77, "learning_rate": 5.4679263163789476e-05, "loss": 0.8445, "step": 7591 }, { "epoch": 0.77, "learning_rate": 5.467790242619531e-05, "loss": 0.6786, "step": 7592 }, { "epoch": 0.77, "learning_rate": 5.4676541531560374e-05, "loss": 1.199, "step": 7593 }, { "epoch": 0.77, "learning_rate": 5.467518047989331e-05, "loss": 1.3418, "step": 7594 }, { "epoch": 0.77, "learning_rate": 5.467381927120279e-05, "loss": 0.851, "step": 7595 }, { "epoch": 0.77, "learning_rate": 5.467245790549747e-05, "loss": 0.777, "step": 7596 }, { "epoch": 0.77, "learning_rate": 5.467109638278601e-05, "loss": 0.8653, "step": 7597 }, { "epoch": 0.77, "learning_rate": 5.466973470307708e-05, "loss": 1.3495, "step": 7598 }, { "epoch": 0.77, "learning_rate": 5.466837286637933e-05, "loss": 0.4959, "step": 7599 }, { "epoch": 0.77, "learning_rate": 5.466701087270146e-05, "loss": 0.6112, "step": 7600 }, { "epoch": 0.77, "learning_rate": 5.466564872205211e-05, "loss": 0.5123, "step": 7601 }, { "epoch": 0.77, "learning_rate": 5.466428641443995e-05, "loss": 0.4205, "step": 7602 }, { "epoch": 0.77, "learning_rate": 5.4662923949873654e-05, "loss": 0.4381, "step": 7603 }, { "epoch": 0.77, "learning_rate": 5.4661561328361895e-05, "loss": 1.2781, "step": 7604 }, { "epoch": 0.77, "learning_rate": 5.466019854991334e-05, "loss": 0.9419, "step": 7605 }, { "epoch": 0.77, "learning_rate": 5.465883561453667e-05, "loss": 0.387, "step": 7606 }, { "epoch": 0.77, "learning_rate": 5.465747252224055e-05, "loss": 0.943, "step": 7607 }, { "epoch": 0.77, "learning_rate": 5.465610927303364e-05, "loss": 1.0142, "step": 7608 }, { "epoch": 0.77, "learning_rate": 5.465474586692464e-05, "loss": 0.9513, "step": 7609 }, { "epoch": 0.77, "learning_rate": 5.465338230392222e-05, "loss": 0.4581, "step": 7610 }, { "epoch": 0.77, "learning_rate": 5.4652018584035045e-05, "loss": 1.0182, "step": 7611 }, { "epoch": 0.77, "learning_rate": 5.4650654707271816e-05, "loss": 0.9925, "step": 7612 }, { "epoch": 0.77, "learning_rate": 5.464929067364118e-05, "loss": 0.2227, "step": 7613 }, { "epoch": 0.77, "learning_rate": 5.464792648315185e-05, "loss": 0.8632, "step": 7614 }, { "epoch": 0.77, "learning_rate": 5.464656213581249e-05, "loss": 0.4488, "step": 7615 }, { "epoch": 0.77, "learning_rate": 5.464519763163178e-05, "loss": 1.0253, "step": 7616 }, { "epoch": 0.77, "learning_rate": 5.464383297061841e-05, "loss": 0.6912, "step": 7617 }, { "epoch": 0.77, "learning_rate": 5.464246815278105e-05, "loss": 0.271, "step": 7618 }, { "epoch": 0.77, "learning_rate": 5.4641103178128416e-05, "loss": 0.8689, "step": 7619 }, { "epoch": 0.77, "learning_rate": 5.4639738046669166e-05, "loss": 0.501, "step": 7620 }, { "epoch": 0.77, "learning_rate": 5.4638372758412e-05, "loss": 0.9996, "step": 7621 }, { "epoch": 0.77, "learning_rate": 5.4637007313365603e-05, "loss": 0.7385, "step": 7622 }, { "epoch": 0.77, "learning_rate": 5.4635641711538665e-05, "loss": 0.4902, "step": 7623 }, { "epoch": 0.77, "learning_rate": 5.4634275952939867e-05, "loss": 0.5677, "step": 7624 }, { "epoch": 0.77, "learning_rate": 5.4632910037577916e-05, "loss": 0.8376, "step": 7625 }, { "epoch": 0.77, "learning_rate": 5.463154396546149e-05, "loss": 0.5541, "step": 7626 }, { "epoch": 0.77, "learning_rate": 5.463017773659929e-05, "loss": 1.4834, "step": 7627 }, { "epoch": 0.77, "learning_rate": 5.4628811351000016e-05, "loss": 0.7777, "step": 7628 }, { "epoch": 0.77, "learning_rate": 5.462744480867235e-05, "loss": 0.813, "step": 7629 }, { "epoch": 0.77, "learning_rate": 5.4626078109625e-05, "loss": 0.7133, "step": 7630 }, { "epoch": 0.77, "learning_rate": 5.462471125386665e-05, "loss": 0.5751, "step": 7631 }, { "epoch": 0.77, "learning_rate": 5.4623344241406e-05, "loss": 1.4216, "step": 7632 }, { "epoch": 0.78, "learning_rate": 5.462197707225176e-05, "loss": 0.7613, "step": 7633 }, { "epoch": 0.78, "learning_rate": 5.462060974641263e-05, "loss": 0.4706, "step": 7634 }, { "epoch": 0.78, "learning_rate": 5.4619242263897304e-05, "loss": 1.2121, "step": 7635 }, { "epoch": 0.78, "learning_rate": 5.461787462471448e-05, "loss": 0.6323, "step": 7636 }, { "epoch": 0.78, "learning_rate": 5.461650682887287e-05, "loss": 1.0027, "step": 7637 }, { "epoch": 0.78, "learning_rate": 5.461513887638118e-05, "loss": 0.3081, "step": 7638 }, { "epoch": 0.78, "learning_rate": 5.46137707672481e-05, "loss": 1.6936, "step": 7639 }, { "epoch": 0.78, "learning_rate": 5.461240250148235e-05, "loss": 0.9454, "step": 7640 }, { "epoch": 0.78, "learning_rate": 5.4611034079092646e-05, "loss": 0.7598, "step": 7641 }, { "epoch": 0.78, "learning_rate": 5.460966550008767e-05, "loss": 1.7008, "step": 7642 }, { "epoch": 0.78, "learning_rate": 5.460829676447615e-05, "loss": 1.2902, "step": 7643 }, { "epoch": 0.78, "learning_rate": 5.4606927872266785e-05, "loss": 0.9428, "step": 7644 }, { "epoch": 0.78, "learning_rate": 5.46055588234683e-05, "loss": 1.5316, "step": 7645 }, { "epoch": 0.78, "learning_rate": 5.460418961808939e-05, "loss": 0.2228, "step": 7646 }, { "epoch": 0.78, "learning_rate": 5.460282025613878e-05, "loss": 0.8376, "step": 7647 }, { "epoch": 0.78, "learning_rate": 5.460145073762519e-05, "loss": 1.0419, "step": 7648 }, { "epoch": 0.78, "learning_rate": 5.4600081062557314e-05, "loss": 1.1961, "step": 7649 }, { "epoch": 0.78, "learning_rate": 5.459871123094389e-05, "loss": 0.7232, "step": 7650 }, { "epoch": 0.78, "learning_rate": 5.459734124279362e-05, "loss": 0.6538, "step": 7651 }, { "epoch": 0.78, "learning_rate": 5.459597109811523e-05, "loss": 0.858, "step": 7652 }, { "epoch": 0.78, "learning_rate": 5.459460079691744e-05, "loss": 0.6788, "step": 7653 }, { "epoch": 0.78, "learning_rate": 5.459323033920897e-05, "loss": 0.9315, "step": 7654 }, { "epoch": 0.78, "learning_rate": 5.4591859724998515e-05, "loss": 0.34, "step": 7655 }, { "epoch": 0.78, "learning_rate": 5.459048895429484e-05, "loss": 1.7858, "step": 7656 }, { "epoch": 0.78, "learning_rate": 5.4589118027106645e-05, "loss": 0.9331, "step": 7657 }, { "epoch": 0.78, "learning_rate": 5.4587746943442656e-05, "loss": 1.2352, "step": 7658 }, { "epoch": 0.78, "learning_rate": 5.4586375703311596e-05, "loss": 1.3237, "step": 7659 }, { "epoch": 0.78, "learning_rate": 5.4585004306722195e-05, "loss": 0.4273, "step": 7660 }, { "epoch": 0.78, "learning_rate": 5.4583632753683185e-05, "loss": 1.2241, "step": 7661 }, { "epoch": 0.78, "learning_rate": 5.458226104420329e-05, "loss": 1.274, "step": 7662 }, { "epoch": 0.78, "learning_rate": 5.458088917829123e-05, "loss": 0.8864, "step": 7663 }, { "epoch": 0.78, "learning_rate": 5.457951715595575e-05, "loss": 0.5766, "step": 7664 }, { "epoch": 0.78, "learning_rate": 5.457814497720556e-05, "loss": 1.2423, "step": 7665 }, { "epoch": 0.78, "learning_rate": 5.4576772642049415e-05, "loss": 0.6063, "step": 7666 }, { "epoch": 0.78, "learning_rate": 5.4575400150496036e-05, "loss": 1.7397, "step": 7667 }, { "epoch": 0.78, "learning_rate": 5.457402750255416e-05, "loss": 0.4742, "step": 7668 }, { "epoch": 0.78, "learning_rate": 5.457265469823252e-05, "loss": 0.8437, "step": 7669 }, { "epoch": 0.78, "learning_rate": 5.4571281737539856e-05, "loss": 0.7008, "step": 7670 }, { "epoch": 0.78, "learning_rate": 5.45699086204849e-05, "loss": 0.5287, "step": 7671 }, { "epoch": 0.78, "learning_rate": 5.45685353470764e-05, "loss": 0.5809, "step": 7672 }, { "epoch": 0.78, "learning_rate": 5.456716191732308e-05, "loss": 1.7316, "step": 7673 }, { "epoch": 0.78, "learning_rate": 5.456578833123368e-05, "loss": 0.5465, "step": 7674 }, { "epoch": 0.78, "learning_rate": 5.456441458881696e-05, "loss": 0.8573, "step": 7675 }, { "epoch": 0.78, "learning_rate": 5.456304069008165e-05, "loss": 1.332, "step": 7676 }, { "epoch": 0.78, "learning_rate": 5.456166663503649e-05, "loss": 1.0754, "step": 7677 }, { "epoch": 0.78, "learning_rate": 5.456029242369022e-05, "loss": 0.5583, "step": 7678 }, { "epoch": 0.78, "learning_rate": 5.4558918056051605e-05, "loss": 0.8916, "step": 7679 }, { "epoch": 0.78, "learning_rate": 5.455754353212937e-05, "loss": 0.7886, "step": 7680 }, { "epoch": 0.78, "learning_rate": 5.455616885193228e-05, "loss": 1.6418, "step": 7681 }, { "epoch": 0.78, "learning_rate": 5.455479401546906e-05, "loss": 0.3761, "step": 7682 }, { "epoch": 0.78, "learning_rate": 5.455341902274848e-05, "loss": 0.7092, "step": 7683 }, { "epoch": 0.78, "learning_rate": 5.455204387377928e-05, "loss": 0.4483, "step": 7684 }, { "epoch": 0.78, "learning_rate": 5.4550668568570214e-05, "loss": 1.1842, "step": 7685 }, { "epoch": 0.78, "learning_rate": 5.454929310713003e-05, "loss": 0.7693, "step": 7686 }, { "epoch": 0.78, "learning_rate": 5.454791748946749e-05, "loss": 0.9506, "step": 7687 }, { "epoch": 0.78, "learning_rate": 5.454654171559133e-05, "loss": 1.4189, "step": 7688 }, { "epoch": 0.78, "learning_rate": 5.454516578551033e-05, "loss": 1.3552, "step": 7689 }, { "epoch": 0.78, "learning_rate": 5.454378969923322e-05, "loss": 0.537, "step": 7690 }, { "epoch": 0.78, "learning_rate": 5.4542413456768775e-05, "loss": 0.7817, "step": 7691 }, { "epoch": 0.78, "learning_rate": 5.454103705812575e-05, "loss": 0.3059, "step": 7692 }, { "epoch": 0.78, "learning_rate": 5.45396605033129e-05, "loss": 0.3333, "step": 7693 }, { "epoch": 0.78, "learning_rate": 5.453828379233898e-05, "loss": 1.0496, "step": 7694 }, { "epoch": 0.78, "learning_rate": 5.453690692521276e-05, "loss": 2.8696, "step": 7695 }, { "epoch": 0.78, "learning_rate": 5.4535529901943e-05, "loss": 1.4245, "step": 7696 }, { "epoch": 0.78, "learning_rate": 5.453415272253846e-05, "loss": 1.1942, "step": 7697 }, { "epoch": 0.78, "learning_rate": 5.453277538700791e-05, "loss": 0.7269, "step": 7698 }, { "epoch": 0.78, "learning_rate": 5.453139789536011e-05, "loss": 0.8486, "step": 7699 }, { "epoch": 0.78, "learning_rate": 5.453002024760383e-05, "loss": 0.3999, "step": 7700 }, { "epoch": 0.78, "learning_rate": 5.4528642443747815e-05, "loss": 1.3425, "step": 7701 }, { "epoch": 0.78, "learning_rate": 5.452726448380086e-05, "loss": 1.3893, "step": 7702 }, { "epoch": 0.78, "learning_rate": 5.452588636777173e-05, "loss": 0.9966, "step": 7703 }, { "epoch": 0.78, "learning_rate": 5.4524508095669185e-05, "loss": 1.1207, "step": 7704 }, { "epoch": 0.78, "learning_rate": 5.4523129667502e-05, "loss": 1.0178, "step": 7705 }, { "epoch": 0.78, "learning_rate": 5.452175108327894e-05, "loss": 0.6767, "step": 7706 }, { "epoch": 0.78, "learning_rate": 5.4520372343008795e-05, "loss": 0.5579, "step": 7707 }, { "epoch": 0.78, "learning_rate": 5.4518993446700326e-05, "loss": 1.0496, "step": 7708 }, { "epoch": 0.78, "learning_rate": 5.451761439436231e-05, "loss": 0.2836, "step": 7709 }, { "epoch": 0.78, "learning_rate": 5.451623518600352e-05, "loss": 0.3425, "step": 7710 }, { "epoch": 0.78, "learning_rate": 5.451485582163274e-05, "loss": 0.4493, "step": 7711 }, { "epoch": 0.78, "learning_rate": 5.451347630125874e-05, "loss": 0.4476, "step": 7712 }, { "epoch": 0.78, "learning_rate": 5.451209662489031e-05, "loss": 1.0658, "step": 7713 }, { "epoch": 0.78, "learning_rate": 5.451071679253621e-05, "loss": 0.9154, "step": 7714 }, { "epoch": 0.78, "learning_rate": 5.450933680420524e-05, "loss": 0.4304, "step": 7715 }, { "epoch": 0.78, "learning_rate": 5.450795665990617e-05, "loss": 1.011, "step": 7716 }, { "epoch": 0.78, "learning_rate": 5.4506576359647796e-05, "loss": 0.3095, "step": 7717 }, { "epoch": 0.78, "learning_rate": 5.4505195903438885e-05, "loss": 0.5686, "step": 7718 }, { "epoch": 0.78, "learning_rate": 5.4503815291288226e-05, "loss": 0.5335, "step": 7719 }, { "epoch": 0.78, "learning_rate": 5.4502434523204616e-05, "loss": 1.5615, "step": 7720 }, { "epoch": 0.78, "learning_rate": 5.450105359919684e-05, "loss": 0.852, "step": 7721 }, { "epoch": 0.78, "learning_rate": 5.449967251927366e-05, "loss": 0.9216, "step": 7722 }, { "epoch": 0.78, "learning_rate": 5.44982912834439e-05, "loss": 1.2663, "step": 7723 }, { "epoch": 0.78, "learning_rate": 5.449690989171633e-05, "loss": 0.904, "step": 7724 }, { "epoch": 0.78, "learning_rate": 5.4495528344099745e-05, "loss": 0.694, "step": 7725 }, { "epoch": 0.78, "learning_rate": 5.449414664060294e-05, "loss": 1.317, "step": 7726 }, { "epoch": 0.78, "learning_rate": 5.44927647812347e-05, "loss": 0.9592, "step": 7727 }, { "epoch": 0.78, "learning_rate": 5.449138276600381e-05, "loss": 1.3575, "step": 7728 }, { "epoch": 0.78, "learning_rate": 5.449000059491909e-05, "loss": 0.4803, "step": 7729 }, { "epoch": 0.78, "learning_rate": 5.448861826798932e-05, "loss": 2.281, "step": 7730 }, { "epoch": 0.79, "learning_rate": 5.4487235785223305e-05, "loss": 1.3058, "step": 7731 }, { "epoch": 0.79, "learning_rate": 5.448585314662983e-05, "loss": 1.4282, "step": 7732 }, { "epoch": 0.79, "learning_rate": 5.448447035221771e-05, "loss": 0.437, "step": 7733 }, { "epoch": 0.79, "learning_rate": 5.448308740199572e-05, "loss": 0.9242, "step": 7734 }, { "epoch": 0.79, "learning_rate": 5.448170429597268e-05, "loss": 0.4241, "step": 7735 }, { "epoch": 0.79, "learning_rate": 5.448032103415739e-05, "loss": 0.7811, "step": 7736 }, { "epoch": 0.79, "learning_rate": 5.4478937616558656e-05, "loss": 0.3816, "step": 7737 }, { "epoch": 0.79, "learning_rate": 5.4477554043185266e-05, "loss": 0.6351, "step": 7738 }, { "epoch": 0.79, "learning_rate": 5.4476170314046034e-05, "loss": 0.1885, "step": 7739 }, { "epoch": 0.79, "learning_rate": 5.4474786429149775e-05, "loss": 1.3148, "step": 7740 }, { "epoch": 0.79, "learning_rate": 5.447340238850528e-05, "loss": 1.0773, "step": 7741 }, { "epoch": 0.79, "learning_rate": 5.4472018192121356e-05, "loss": 0.4786, "step": 7742 }, { "epoch": 0.79, "learning_rate": 5.4470633840006826e-05, "loss": 0.3997, "step": 7743 }, { "epoch": 0.79, "learning_rate": 5.4469249332170487e-05, "loss": 0.4643, "step": 7744 }, { "epoch": 0.79, "learning_rate": 5.446786466862116e-05, "loss": 1.4358, "step": 7745 }, { "epoch": 0.79, "learning_rate": 5.446647984936764e-05, "loss": 0.332, "step": 7746 }, { "epoch": 0.79, "learning_rate": 5.446509487441876e-05, "loss": 1.4571, "step": 7747 }, { "epoch": 0.79, "learning_rate": 5.446370974378332e-05, "loss": 0.3488, "step": 7748 }, { "epoch": 0.79, "learning_rate": 5.446232445747013e-05, "loss": 0.6997, "step": 7749 }, { "epoch": 0.79, "learning_rate": 5.4460939015488024e-05, "loss": 0.6536, "step": 7750 }, { "epoch": 0.79, "learning_rate": 5.4459553417845806e-05, "loss": 0.5671, "step": 7751 }, { "epoch": 0.79, "learning_rate": 5.445816766455229e-05, "loss": 1.0696, "step": 7752 }, { "epoch": 0.79, "learning_rate": 5.44567817556163e-05, "loss": 1.0669, "step": 7753 }, { "epoch": 0.79, "learning_rate": 5.445539569104665e-05, "loss": 0.7679, "step": 7754 }, { "epoch": 0.79, "learning_rate": 5.445400947085218e-05, "loss": 1.1727, "step": 7755 }, { "epoch": 0.79, "learning_rate": 5.445262309504169e-05, "loss": 0.6453, "step": 7756 }, { "epoch": 0.79, "learning_rate": 5.4451236563624e-05, "loss": 0.9781, "step": 7757 }, { "epoch": 0.79, "learning_rate": 5.444984987660795e-05, "loss": 0.382, "step": 7758 }, { "epoch": 0.79, "learning_rate": 5.444846303400236e-05, "loss": 1.2814, "step": 7759 }, { "epoch": 0.79, "learning_rate": 5.444707603581605e-05, "loss": 1.7151, "step": 7760 }, { "epoch": 0.79, "learning_rate": 5.444568888205785e-05, "loss": 1.485, "step": 7761 }, { "epoch": 0.79, "learning_rate": 5.444430157273657e-05, "loss": 0.5288, "step": 7762 }, { "epoch": 0.79, "learning_rate": 5.444291410786107e-05, "loss": 0.8388, "step": 7763 }, { "epoch": 0.79, "learning_rate": 5.444152648744016e-05, "loss": 0.9124, "step": 7764 }, { "epoch": 0.79, "learning_rate": 5.4440138711482676e-05, "loss": 1.5737, "step": 7765 }, { "epoch": 0.79, "learning_rate": 5.443875077999744e-05, "loss": 1.1806, "step": 7766 }, { "epoch": 0.79, "learning_rate": 5.443736269299329e-05, "loss": 0.4046, "step": 7767 }, { "epoch": 0.79, "learning_rate": 5.4435974450479065e-05, "loss": 0.1773, "step": 7768 }, { "epoch": 0.79, "learning_rate": 5.443458605246359e-05, "loss": 0.5307, "step": 7769 }, { "epoch": 0.79, "learning_rate": 5.443319749895571e-05, "loss": 0.968, "step": 7770 }, { "epoch": 0.79, "learning_rate": 5.4431808789964256e-05, "loss": 0.5755, "step": 7771 }, { "epoch": 0.79, "learning_rate": 5.443041992549806e-05, "loss": 0.184, "step": 7772 }, { "epoch": 0.79, "learning_rate": 5.4429030905565965e-05, "loss": 1.004, "step": 7773 }, { "epoch": 0.79, "learning_rate": 5.442764173017682e-05, "loss": 1.0085, "step": 7774 }, { "epoch": 0.79, "learning_rate": 5.442625239933945e-05, "loss": 1.4446, "step": 7775 }, { "epoch": 0.79, "learning_rate": 5.44248629130627e-05, "loss": 1.8081, "step": 7776 }, { "epoch": 0.79, "learning_rate": 5.442347327135542e-05, "loss": 0.7652, "step": 7777 }, { "epoch": 0.79, "learning_rate": 5.442208347422644e-05, "loss": 0.5422, "step": 7778 }, { "epoch": 0.79, "learning_rate": 5.442069352168462e-05, "loss": 0.5973, "step": 7779 }, { "epoch": 0.79, "learning_rate": 5.441930341373879e-05, "loss": 0.6014, "step": 7780 }, { "epoch": 0.79, "learning_rate": 5.44179131503978e-05, "loss": 1.1201, "step": 7781 }, { "epoch": 0.79, "learning_rate": 5.4416522731670506e-05, "loss": 0.8449, "step": 7782 }, { "epoch": 0.79, "learning_rate": 5.441513215756575e-05, "loss": 1.0216, "step": 7783 }, { "epoch": 0.79, "learning_rate": 5.441374142809238e-05, "loss": 1.3502, "step": 7784 }, { "epoch": 0.79, "learning_rate": 5.441235054325924e-05, "loss": 0.799, "step": 7785 }, { "epoch": 0.79, "learning_rate": 5.4410959503075194e-05, "loss": 0.6983, "step": 7786 }, { "epoch": 0.79, "learning_rate": 5.440956830754909e-05, "loss": 1.1353, "step": 7787 }, { "epoch": 0.79, "learning_rate": 5.440817695668977e-05, "loss": 0.868, "step": 7788 }, { "epoch": 0.79, "learning_rate": 5.4406785450506116e-05, "loss": 1.3584, "step": 7789 }, { "epoch": 0.79, "learning_rate": 5.440539378900695e-05, "loss": 1.3392, "step": 7790 }, { "epoch": 0.79, "learning_rate": 5.440400197220114e-05, "loss": 0.4588, "step": 7791 }, { "epoch": 0.79, "learning_rate": 5.440261000009755e-05, "loss": 1.3119, "step": 7792 }, { "epoch": 0.79, "learning_rate": 5.4401217872705035e-05, "loss": 0.3558, "step": 7793 }, { "epoch": 0.79, "learning_rate": 5.4399825590032454e-05, "loss": 0.8545, "step": 7794 }, { "epoch": 0.79, "learning_rate": 5.4398433152088656e-05, "loss": 1.1835, "step": 7795 }, { "epoch": 0.79, "learning_rate": 5.439704055888252e-05, "loss": 1.311, "step": 7796 }, { "epoch": 0.79, "learning_rate": 5.439564781042289e-05, "loss": 0.4757, "step": 7797 }, { "epoch": 0.79, "learning_rate": 5.4394254906718646e-05, "loss": 0.9875, "step": 7798 }, { "epoch": 0.79, "learning_rate": 5.439286184777864e-05, "loss": 0.4688, "step": 7799 }, { "epoch": 0.79, "learning_rate": 5.439146863361174e-05, "loss": 0.4591, "step": 7800 }, { "epoch": 0.79, "learning_rate": 5.4390075264226816e-05, "loss": 0.2062, "step": 7801 }, { "epoch": 0.79, "learning_rate": 5.438868173963273e-05, "loss": 0.5334, "step": 7802 }, { "epoch": 0.79, "learning_rate": 5.438728805983836e-05, "loss": 0.372, "step": 7803 }, { "epoch": 0.79, "learning_rate": 5.4385894224852554e-05, "loss": 1.5262, "step": 7804 }, { "epoch": 0.79, "learning_rate": 5.4384500234684194e-05, "loss": 0.8283, "step": 7805 }, { "epoch": 0.79, "learning_rate": 5.438310608934215e-05, "loss": 1.0415, "step": 7806 }, { "epoch": 0.79, "learning_rate": 5.43817117888353e-05, "loss": 0.6792, "step": 7807 }, { "epoch": 0.79, "learning_rate": 5.4380317333172514e-05, "loss": 0.8211, "step": 7808 }, { "epoch": 0.79, "learning_rate": 5.4378922722362656e-05, "loss": 0.834, "step": 7809 }, { "epoch": 0.79, "learning_rate": 5.4377527956414614e-05, "loss": 1.2949, "step": 7810 }, { "epoch": 0.79, "learning_rate": 5.437613303533726e-05, "loss": 1.602, "step": 7811 }, { "epoch": 0.79, "learning_rate": 5.437473795913946e-05, "loss": 0.5808, "step": 7812 }, { "epoch": 0.79, "learning_rate": 5.437334272783011e-05, "loss": 0.4494, "step": 7813 }, { "epoch": 0.79, "learning_rate": 5.4371947341418065e-05, "loss": 0.6189, "step": 7814 }, { "epoch": 0.79, "learning_rate": 5.437055179991223e-05, "loss": 1.2179, "step": 7815 }, { "epoch": 0.79, "learning_rate": 5.436915610332147e-05, "loss": 1.4154, "step": 7816 }, { "epoch": 0.79, "learning_rate": 5.436776025165467e-05, "loss": 0.5402, "step": 7817 }, { "epoch": 0.79, "learning_rate": 5.436636424492072e-05, "loss": 0.3707, "step": 7818 }, { "epoch": 0.79, "learning_rate": 5.4364968083128496e-05, "loss": 0.2659, "step": 7819 }, { "epoch": 0.79, "learning_rate": 5.436357176628688e-05, "loss": 1.3988, "step": 7820 }, { "epoch": 0.79, "learning_rate": 5.4362175294404764e-05, "loss": 0.7681, "step": 7821 }, { "epoch": 0.79, "learning_rate": 5.436077866749103e-05, "loss": 0.6287, "step": 7822 }, { "epoch": 0.79, "learning_rate": 5.435938188555457e-05, "loss": 0.498, "step": 7823 }, { "epoch": 0.79, "learning_rate": 5.4357984948604266e-05, "loss": 0.1572, "step": 7824 }, { "epoch": 0.79, "learning_rate": 5.435658785664902e-05, "loss": 1.0118, "step": 7825 }, { "epoch": 0.79, "learning_rate": 5.435519060969771e-05, "loss": 1.5963, "step": 7826 }, { "epoch": 0.79, "learning_rate": 5.4353793207759234e-05, "loss": 1.8438, "step": 7827 }, { "epoch": 0.79, "learning_rate": 5.4352395650842476e-05, "loss": 0.5332, "step": 7828 }, { "epoch": 0.79, "learning_rate": 5.435099793895635e-05, "loss": 1.4772, "step": 7829 }, { "epoch": 0.8, "learning_rate": 5.434960007210972e-05, "loss": 0.7505, "step": 7830 }, { "epoch": 0.8, "learning_rate": 5.434820205031151e-05, "loss": 1.0182, "step": 7831 }, { "epoch": 0.8, "learning_rate": 5.43468038735706e-05, "loss": 0.589, "step": 7832 }, { "epoch": 0.8, "learning_rate": 5.434540554189589e-05, "loss": 0.3422, "step": 7833 }, { "epoch": 0.8, "learning_rate": 5.434400705529629e-05, "loss": 1.4574, "step": 7834 }, { "epoch": 0.8, "learning_rate": 5.4342608413780676e-05, "loss": 0.8071, "step": 7835 }, { "epoch": 0.8, "learning_rate": 5.434120961735797e-05, "loss": 0.471, "step": 7836 }, { "epoch": 0.8, "learning_rate": 5.433981066603707e-05, "loss": 1.3124, "step": 7837 }, { "epoch": 0.8, "learning_rate": 5.433841155982687e-05, "loss": 1.1233, "step": 7838 }, { "epoch": 0.8, "learning_rate": 5.4337012298736274e-05, "loss": 0.3206, "step": 7839 }, { "epoch": 0.8, "learning_rate": 5.4335612882774195e-05, "loss": 0.8859, "step": 7840 }, { "epoch": 0.8, "learning_rate": 5.433421331194953e-05, "loss": 1.0777, "step": 7841 }, { "epoch": 0.8, "learning_rate": 5.4332813586271196e-05, "loss": 0.9512, "step": 7842 }, { "epoch": 0.8, "learning_rate": 5.433141370574809e-05, "loss": 0.7367, "step": 7843 }, { "epoch": 0.8, "learning_rate": 5.433001367038912e-05, "loss": 0.1866, "step": 7844 }, { "epoch": 0.8, "learning_rate": 5.43286134802032e-05, "loss": 1.6369, "step": 7845 }, { "epoch": 0.8, "learning_rate": 5.4327213135199245e-05, "loss": 0.7235, "step": 7846 }, { "epoch": 0.8, "learning_rate": 5.4325812635386146e-05, "loss": 0.2457, "step": 7847 }, { "epoch": 0.8, "learning_rate": 5.432441198077284e-05, "loss": 0.2886, "step": 7848 }, { "epoch": 0.8, "learning_rate": 5.432301117136822e-05, "loss": 0.9059, "step": 7849 }, { "epoch": 0.8, "learning_rate": 5.432161020718122e-05, "loss": 1.1998, "step": 7850 }, { "epoch": 0.8, "learning_rate": 5.4320209088220745e-05, "loss": 0.4752, "step": 7851 }, { "epoch": 0.8, "learning_rate": 5.431880781449571e-05, "loss": 0.3286, "step": 7852 }, { "epoch": 0.8, "learning_rate": 5.431740638601503e-05, "loss": 0.8782, "step": 7853 }, { "epoch": 0.8, "learning_rate": 5.431600480278763e-05, "loss": 1.5158, "step": 7854 }, { "epoch": 0.8, "learning_rate": 5.431460306482242e-05, "loss": 0.4376, "step": 7855 }, { "epoch": 0.8, "learning_rate": 5.431320117212833e-05, "loss": 0.5405, "step": 7856 }, { "epoch": 0.8, "learning_rate": 5.431179912471428e-05, "loss": 0.513, "step": 7857 }, { "epoch": 0.8, "learning_rate": 5.4310396922589184e-05, "loss": 1.1754, "step": 7858 }, { "epoch": 0.8, "learning_rate": 5.430899456576198e-05, "loss": 1.1858, "step": 7859 }, { "epoch": 0.8, "learning_rate": 5.4307592054241564e-05, "loss": 0.361, "step": 7860 }, { "epoch": 0.8, "learning_rate": 5.4306189388036894e-05, "loss": 1.354, "step": 7861 }, { "epoch": 0.8, "learning_rate": 5.4304786567156874e-05, "loss": 1.2638, "step": 7862 }, { "epoch": 0.8, "learning_rate": 5.4303383591610446e-05, "loss": 1.1521, "step": 7863 }, { "epoch": 0.8, "learning_rate": 5.430198046140652e-05, "loss": 0.8269, "step": 7864 }, { "epoch": 0.8, "learning_rate": 5.430057717655404e-05, "loss": 0.9327, "step": 7865 }, { "epoch": 0.8, "learning_rate": 5.429917373706193e-05, "loss": 1.0695, "step": 7866 }, { "epoch": 0.8, "learning_rate": 5.4297770142939126e-05, "loss": 0.3439, "step": 7867 }, { "epoch": 0.8, "learning_rate": 5.4296366394194556e-05, "loss": 1.2796, "step": 7868 }, { "epoch": 0.8, "learning_rate": 5.4294962490837145e-05, "loss": 0.9088, "step": 7869 }, { "epoch": 0.8, "learning_rate": 5.4293558432875845e-05, "loss": 0.6802, "step": 7870 }, { "epoch": 0.8, "learning_rate": 5.429215422031958e-05, "loss": 1.017, "step": 7871 }, { "epoch": 0.8, "learning_rate": 5.429074985317728e-05, "loss": 0.471, "step": 7872 }, { "epoch": 0.8, "learning_rate": 5.4289345331457896e-05, "loss": 0.3953, "step": 7873 }, { "epoch": 0.8, "learning_rate": 5.428794065517035e-05, "loss": 1.2691, "step": 7874 }, { "epoch": 0.8, "learning_rate": 5.428653582432359e-05, "loss": 1.0207, "step": 7875 }, { "epoch": 0.8, "learning_rate": 5.428513083892657e-05, "loss": 1.3453, "step": 7876 }, { "epoch": 0.8, "learning_rate": 5.42837256989882e-05, "loss": 1.3783, "step": 7877 }, { "epoch": 0.8, "learning_rate": 5.4282320404517445e-05, "loss": 0.1248, "step": 7878 }, { "epoch": 0.8, "learning_rate": 5.428091495552324e-05, "loss": 0.0738, "step": 7879 }, { "epoch": 0.8, "learning_rate": 5.4279509352014525e-05, "loss": 1.3072, "step": 7880 }, { "epoch": 0.8, "learning_rate": 5.427810359400025e-05, "loss": 0.1925, "step": 7881 }, { "epoch": 0.8, "learning_rate": 5.4276697681489363e-05, "loss": 0.6924, "step": 7882 }, { "epoch": 0.8, "learning_rate": 5.4275291614490806e-05, "loss": 1.3856, "step": 7883 }, { "epoch": 0.8, "learning_rate": 5.4273885393013524e-05, "loss": 1.0657, "step": 7884 }, { "epoch": 0.8, "learning_rate": 5.427247901706647e-05, "loss": 0.322, "step": 7885 }, { "epoch": 0.8, "learning_rate": 5.427107248665859e-05, "loss": 0.6099, "step": 7886 }, { "epoch": 0.8, "learning_rate": 5.426966580179885e-05, "loss": 1.3483, "step": 7887 }, { "epoch": 0.8, "learning_rate": 5.4268258962496184e-05, "loss": 1.4156, "step": 7888 }, { "epoch": 0.8, "learning_rate": 5.426685196875955e-05, "loss": 0.8041, "step": 7889 }, { "epoch": 0.8, "learning_rate": 5.42654448205979e-05, "loss": 0.8556, "step": 7890 }, { "epoch": 0.8, "learning_rate": 5.426403751802019e-05, "loss": 1.3496, "step": 7891 }, { "epoch": 0.8, "learning_rate": 5.426263006103538e-05, "loss": 0.8241, "step": 7892 }, { "epoch": 0.8, "learning_rate": 5.426122244965242e-05, "loss": 0.8954, "step": 7893 }, { "epoch": 0.8, "learning_rate": 5.425981468388027e-05, "loss": 1.0842, "step": 7894 }, { "epoch": 0.8, "learning_rate": 5.4258406763727886e-05, "loss": 0.4119, "step": 7895 }, { "epoch": 0.8, "learning_rate": 5.4256998689204235e-05, "loss": 0.4666, "step": 7896 }, { "epoch": 0.8, "learning_rate": 5.425559046031827e-05, "loss": 1.377, "step": 7897 }, { "epoch": 0.8, "learning_rate": 5.425418207707895e-05, "loss": 0.8928, "step": 7898 }, { "epoch": 0.8, "learning_rate": 5.4252773539495236e-05, "loss": 1.0988, "step": 7899 }, { "epoch": 0.8, "learning_rate": 5.4251364847576114e-05, "loss": 0.6634, "step": 7900 }, { "epoch": 0.8, "learning_rate": 5.424995600133053e-05, "loss": 1.6473, "step": 7901 }, { "epoch": 0.8, "learning_rate": 5.4248547000767435e-05, "loss": 0.6737, "step": 7902 }, { "epoch": 0.8, "learning_rate": 5.424713784589582e-05, "loss": 1.1921, "step": 7903 }, { "epoch": 0.8, "learning_rate": 5.424572853672465e-05, "loss": 0.5401, "step": 7904 }, { "epoch": 0.8, "learning_rate": 5.4244319073262885e-05, "loss": 0.5734, "step": 7905 }, { "epoch": 0.8, "learning_rate": 5.424290945551949e-05, "loss": 1.6514, "step": 7906 }, { "epoch": 0.8, "learning_rate": 5.424149968350344e-05, "loss": 0.7325, "step": 7907 }, { "epoch": 0.8, "learning_rate": 5.4240089757223724e-05, "loss": 1.1355, "step": 7908 }, { "epoch": 0.8, "learning_rate": 5.4238679676689284e-05, "loss": 0.9198, "step": 7909 }, { "epoch": 0.8, "learning_rate": 5.4237269441909115e-05, "loss": 1.5555, "step": 7910 }, { "epoch": 0.8, "learning_rate": 5.423585905289218e-05, "loss": 0.4066, "step": 7911 }, { "epoch": 0.8, "learning_rate": 5.423444850964746e-05, "loss": 0.7665, "step": 7912 }, { "epoch": 0.8, "learning_rate": 5.423303781218392e-05, "loss": 0.7332, "step": 7913 }, { "epoch": 0.8, "learning_rate": 5.423162696051055e-05, "loss": 1.0494, "step": 7914 }, { "epoch": 0.8, "learning_rate": 5.423021595463633e-05, "loss": 0.713, "step": 7915 }, { "epoch": 0.8, "learning_rate": 5.422880479457023e-05, "loss": 0.6219, "step": 7916 }, { "epoch": 0.8, "learning_rate": 5.422739348032124e-05, "loss": 1.335, "step": 7917 }, { "epoch": 0.8, "learning_rate": 5.422598201189832e-05, "loss": 0.634, "step": 7918 }, { "epoch": 0.8, "learning_rate": 5.422457038931048e-05, "loss": 0.5648, "step": 7919 }, { "epoch": 0.8, "learning_rate": 5.4223158612566683e-05, "loss": 1.4604, "step": 7920 }, { "epoch": 0.8, "learning_rate": 5.422174668167592e-05, "loss": 0.8438, "step": 7921 }, { "epoch": 0.8, "learning_rate": 5.4220334596647175e-05, "loss": 1.4154, "step": 7922 }, { "epoch": 0.8, "learning_rate": 5.421892235748944e-05, "loss": 1.2414, "step": 7923 }, { "epoch": 0.8, "learning_rate": 5.4217509964211694e-05, "loss": 0.8741, "step": 7924 }, { "epoch": 0.8, "learning_rate": 5.4216097416822924e-05, "loss": 1.2684, "step": 7925 }, { "epoch": 0.8, "learning_rate": 5.421468471533213e-05, "loss": 1.0277, "step": 7926 }, { "epoch": 0.8, "learning_rate": 5.421327185974828e-05, "loss": 0.7958, "step": 7927 }, { "epoch": 0.81, "learning_rate": 5.421185885008039e-05, "loss": 0.8278, "step": 7928 }, { "epoch": 0.81, "learning_rate": 5.4210445686337436e-05, "loss": 1.2474, "step": 7929 }, { "epoch": 0.81, "learning_rate": 5.420903236852842e-05, "loss": 1.2804, "step": 7930 }, { "epoch": 0.81, "learning_rate": 5.4207618896662324e-05, "loss": 1.8779, "step": 7931 }, { "epoch": 0.81, "learning_rate": 5.420620527074816e-05, "loss": 0.7605, "step": 7932 }, { "epoch": 0.81, "learning_rate": 5.420479149079491e-05, "loss": 0.9035, "step": 7933 }, { "epoch": 0.81, "learning_rate": 5.4203377556811576e-05, "loss": 0.7107, "step": 7934 }, { "epoch": 0.81, "learning_rate": 5.420196346880715e-05, "loss": 1.2094, "step": 7935 }, { "epoch": 0.81, "learning_rate": 5.4200549226790644e-05, "loss": 1.6169, "step": 7936 }, { "epoch": 0.81, "learning_rate": 5.419913483077104e-05, "loss": 0.3594, "step": 7937 }, { "epoch": 0.81, "learning_rate": 5.419772028075735e-05, "loss": 1.2239, "step": 7938 }, { "epoch": 0.81, "learning_rate": 5.419630557675858e-05, "loss": 1.4103, "step": 7939 }, { "epoch": 0.81, "learning_rate": 5.419489071878372e-05, "loss": 1.5846, "step": 7940 }, { "epoch": 0.81, "learning_rate": 5.419347570684178e-05, "loss": 1.6411, "step": 7941 }, { "epoch": 0.81, "learning_rate": 5.4192060540941765e-05, "loss": 1.4559, "step": 7942 }, { "epoch": 0.81, "learning_rate": 5.4190645221092674e-05, "loss": 1.1319, "step": 7943 }, { "epoch": 0.81, "learning_rate": 5.418922974730354e-05, "loss": 1.2278, "step": 7944 }, { "epoch": 0.81, "learning_rate": 5.418781411958333e-05, "loss": 0.95, "step": 7945 }, { "epoch": 0.81, "learning_rate": 5.4186398337941075e-05, "loss": 1.82, "step": 7946 }, { "epoch": 0.81, "learning_rate": 5.418498240238579e-05, "loss": 1.2915, "step": 7947 }, { "epoch": 0.81, "learning_rate": 5.418356631292647e-05, "loss": 1.1523, "step": 7948 }, { "epoch": 0.81, "learning_rate": 5.418215006957213e-05, "loss": 0.9161, "step": 7949 }, { "epoch": 0.81, "learning_rate": 5.41807336723318e-05, "loss": 1.4016, "step": 7950 }, { "epoch": 0.81, "learning_rate": 5.417931712121447e-05, "loss": 1.1108, "step": 7951 }, { "epoch": 0.81, "learning_rate": 5.4177900416229166e-05, "loss": 1.229, "step": 7952 }, { "epoch": 0.81, "learning_rate": 5.417648355738491e-05, "loss": 0.4911, "step": 7953 }, { "epoch": 0.81, "learning_rate": 5.417506654469069e-05, "loss": 0.96, "step": 7954 }, { "epoch": 0.81, "learning_rate": 5.417364937815556e-05, "loss": 0.7018, "step": 7955 }, { "epoch": 0.81, "learning_rate": 5.4172232057788515e-05, "loss": 1.0075, "step": 7956 }, { "epoch": 0.81, "learning_rate": 5.4170814583598584e-05, "loss": 1.5251, "step": 7957 }, { "epoch": 0.81, "learning_rate": 5.416939695559478e-05, "loss": 0.7368, "step": 7958 }, { "epoch": 0.81, "learning_rate": 5.416797917378613e-05, "loss": 1.21, "step": 7959 }, { "epoch": 0.81, "learning_rate": 5.416656123818165e-05, "loss": 0.946, "step": 7960 }, { "epoch": 0.81, "learning_rate": 5.416514314879038e-05, "loss": 0.48, "step": 7961 }, { "epoch": 0.81, "learning_rate": 5.4163724905621325e-05, "loss": 0.4094, "step": 7962 }, { "epoch": 0.81, "learning_rate": 5.416230650868352e-05, "loss": 1.0381, "step": 7963 }, { "epoch": 0.81, "learning_rate": 5.4160887957985986e-05, "loss": 0.241, "step": 7964 }, { "epoch": 0.81, "learning_rate": 5.415946925353775e-05, "loss": 1.0136, "step": 7965 }, { "epoch": 0.81, "learning_rate": 5.415805039534784e-05, "loss": 0.7864, "step": 7966 }, { "epoch": 0.81, "learning_rate": 5.4156631383425294e-05, "loss": 1.1305, "step": 7967 }, { "epoch": 0.81, "learning_rate": 5.415521221777913e-05, "loss": 0.5577, "step": 7968 }, { "epoch": 0.81, "learning_rate": 5.4153792898418385e-05, "loss": 1.4346, "step": 7969 }, { "epoch": 0.81, "learning_rate": 5.41523734253521e-05, "loss": 0.2732, "step": 7970 }, { "epoch": 0.81, "learning_rate": 5.415095379858929e-05, "loss": 1.3386, "step": 7971 }, { "epoch": 0.81, "learning_rate": 5.4149534018138994e-05, "loss": 0.4168, "step": 7972 }, { "epoch": 0.81, "learning_rate": 5.414811408401026e-05, "loss": 1.1688, "step": 7973 }, { "epoch": 0.81, "learning_rate": 5.414669399621211e-05, "loss": 0.468, "step": 7974 }, { "epoch": 0.81, "learning_rate": 5.414527375475358e-05, "loss": 1.019, "step": 7975 }, { "epoch": 0.81, "learning_rate": 5.4143853359643725e-05, "loss": 0.4293, "step": 7976 }, { "epoch": 0.81, "learning_rate": 5.414243281089156e-05, "loss": 1.3681, "step": 7977 }, { "epoch": 0.81, "learning_rate": 5.414101210850614e-05, "loss": 0.4144, "step": 7978 }, { "epoch": 0.81, "learning_rate": 5.4139591252496504e-05, "loss": 0.5815, "step": 7979 }, { "epoch": 0.81, "learning_rate": 5.4138170242871696e-05, "loss": 1.3386, "step": 7980 }, { "epoch": 0.81, "learning_rate": 5.413674907964075e-05, "loss": 1.6683, "step": 7981 }, { "epoch": 0.81, "learning_rate": 5.4135327762812716e-05, "loss": 0.4048, "step": 7982 }, { "epoch": 0.81, "learning_rate": 5.413390629239664e-05, "loss": 0.6095, "step": 7983 }, { "epoch": 0.81, "learning_rate": 5.4132484668401566e-05, "loss": 0.7967, "step": 7984 }, { "epoch": 0.81, "learning_rate": 5.413106289083654e-05, "loss": 0.6255, "step": 7985 }, { "epoch": 0.81, "learning_rate": 5.41296409597106e-05, "loss": 0.577, "step": 7986 }, { "epoch": 0.81, "learning_rate": 5.412821887503281e-05, "loss": 0.5035, "step": 7987 }, { "epoch": 0.81, "learning_rate": 5.412679663681222e-05, "loss": 1.8637, "step": 7988 }, { "epoch": 0.81, "learning_rate": 5.412537424505787e-05, "loss": 0.9849, "step": 7989 }, { "epoch": 0.81, "learning_rate": 5.412395169977882e-05, "loss": 0.6128, "step": 7990 }, { "epoch": 0.81, "learning_rate": 5.412252900098412e-05, "loss": 0.948, "step": 7991 }, { "epoch": 0.81, "learning_rate": 5.412110614868281e-05, "loss": 0.7777, "step": 7992 }, { "epoch": 0.81, "learning_rate": 5.411968314288397e-05, "loss": 1.3183, "step": 7993 }, { "epoch": 0.81, "learning_rate": 5.4118259983596635e-05, "loss": 0.9609, "step": 7994 }, { "epoch": 0.81, "learning_rate": 5.411683667082987e-05, "loss": 1.5081, "step": 7995 }, { "epoch": 0.81, "learning_rate": 5.4115413204592734e-05, "loss": 0.4772, "step": 7996 }, { "epoch": 0.81, "learning_rate": 5.4113989584894275e-05, "loss": 0.3241, "step": 7997 }, { "epoch": 0.81, "learning_rate": 5.411256581174356e-05, "loss": 0.3224, "step": 7998 }, { "epoch": 0.81, "learning_rate": 5.411114188514965e-05, "loss": 1.313, "step": 7999 }, { "epoch": 0.81, "learning_rate": 5.410971780512161e-05, "loss": 1.2215, "step": 8000 }, { "epoch": 0.81, "learning_rate": 5.41082935716685e-05, "loss": 0.7152, "step": 8001 }, { "epoch": 0.81, "learning_rate": 5.410686918479937e-05, "loss": 1.2261, "step": 8002 }, { "epoch": 0.81, "learning_rate": 5.410544464452331e-05, "loss": 0.5345, "step": 8003 }, { "epoch": 0.81, "learning_rate": 5.410401995084936e-05, "loss": 1.0952, "step": 8004 }, { "epoch": 0.81, "learning_rate": 5.4102595103786594e-05, "loss": 0.4921, "step": 8005 }, { "epoch": 0.81, "learning_rate": 5.410117010334409e-05, "loss": 0.613, "step": 8006 }, { "epoch": 0.81, "learning_rate": 5.409974494953091e-05, "loss": 0.4927, "step": 8007 }, { "epoch": 0.81, "learning_rate": 5.409831964235612e-05, "loss": 0.6398, "step": 8008 }, { "epoch": 0.81, "learning_rate": 5.409689418182878e-05, "loss": 1.2965, "step": 8009 }, { "epoch": 0.81, "learning_rate": 5.409546856795798e-05, "loss": 1.2399, "step": 8010 }, { "epoch": 0.81, "learning_rate": 5.4094042800752784e-05, "loss": 1.2545, "step": 8011 }, { "epoch": 0.81, "learning_rate": 5.409261688022226e-05, "loss": 1.9152, "step": 8012 }, { "epoch": 0.81, "learning_rate": 5.40911908063755e-05, "loss": 1.1297, "step": 8013 }, { "epoch": 0.81, "learning_rate": 5.408976457922155e-05, "loss": 1.7507, "step": 8014 }, { "epoch": 0.81, "learning_rate": 5.408833819876952e-05, "loss": 0.805, "step": 8015 }, { "epoch": 0.81, "learning_rate": 5.408691166502846e-05, "loss": 1.3426, "step": 8016 }, { "epoch": 0.81, "learning_rate": 5.408548497800747e-05, "loss": 0.8976, "step": 8017 }, { "epoch": 0.81, "learning_rate": 5.4084058137715595e-05, "loss": 0.331, "step": 8018 }, { "epoch": 0.81, "learning_rate": 5.4082631144161946e-05, "loss": 0.3786, "step": 8019 }, { "epoch": 0.81, "learning_rate": 5.4081203997355594e-05, "loss": 1.7615, "step": 8020 }, { "epoch": 0.81, "learning_rate": 5.407977669730563e-05, "loss": 0.3824, "step": 8021 }, { "epoch": 0.81, "learning_rate": 5.407834924402111e-05, "loss": 1.5615, "step": 8022 }, { "epoch": 0.81, "learning_rate": 5.407692163751114e-05, "loss": 1.2049, "step": 8023 }, { "epoch": 0.81, "learning_rate": 5.407549387778481e-05, "loss": 0.3999, "step": 8024 }, { "epoch": 0.81, "learning_rate": 5.407406596485118e-05, "loss": 1.3984, "step": 8025 }, { "epoch": 0.81, "learning_rate": 5.407263789871937e-05, "loss": 1.1626, "step": 8026 }, { "epoch": 0.82, "learning_rate": 5.407120967939844e-05, "loss": 0.7105, "step": 8027 }, { "epoch": 0.82, "learning_rate": 5.4069781306897496e-05, "loss": 0.3416, "step": 8028 }, { "epoch": 0.82, "learning_rate": 5.406835278122561e-05, "loss": 0.3097, "step": 8029 }, { "epoch": 0.82, "learning_rate": 5.406692410239189e-05, "loss": 0.7526, "step": 8030 }, { "epoch": 0.82, "learning_rate": 5.4065495270405415e-05, "loss": 1.2324, "step": 8031 }, { "epoch": 0.82, "learning_rate": 5.4064066285275286e-05, "loss": 0.7623, "step": 8032 }, { "epoch": 0.82, "learning_rate": 5.406263714701059e-05, "loss": 0.4622, "step": 8033 }, { "epoch": 0.82, "learning_rate": 5.4061207855620434e-05, "loss": 0.8212, "step": 8034 }, { "epoch": 0.82, "learning_rate": 5.40597784111139e-05, "loss": 0.483, "step": 8035 }, { "epoch": 0.82, "learning_rate": 5.405834881350009e-05, "loss": 0.3937, "step": 8036 }, { "epoch": 0.82, "learning_rate": 5.40569190627881e-05, "loss": 0.5071, "step": 8037 }, { "epoch": 0.82, "learning_rate": 5.405548915898702e-05, "loss": 1.071, "step": 8038 }, { "epoch": 0.82, "learning_rate": 5.4054059102105975e-05, "loss": 0.9234, "step": 8039 }, { "epoch": 0.82, "learning_rate": 5.4052628892154035e-05, "loss": 1.2087, "step": 8040 }, { "epoch": 0.82, "learning_rate": 5.4051198529140326e-05, "loss": 0.273, "step": 8041 }, { "epoch": 0.82, "learning_rate": 5.4049768013073924e-05, "loss": 0.8835, "step": 8042 }, { "epoch": 0.82, "learning_rate": 5.4048337343963956e-05, "loss": 0.5004, "step": 8043 }, { "epoch": 0.82, "learning_rate": 5.404690652181952e-05, "loss": 0.6181, "step": 8044 }, { "epoch": 0.82, "learning_rate": 5.404547554664971e-05, "loss": 0.515, "step": 8045 }, { "epoch": 0.82, "learning_rate": 5.404404441846365e-05, "loss": 1.4714, "step": 8046 }, { "epoch": 0.82, "learning_rate": 5.404261313727044e-05, "loss": 0.3183, "step": 8047 }, { "epoch": 0.82, "learning_rate": 5.404118170307918e-05, "loss": 0.8924, "step": 8048 }, { "epoch": 0.82, "learning_rate": 5.403975011589899e-05, "loss": 0.9814, "step": 8049 }, { "epoch": 0.82, "learning_rate": 5.403831837573897e-05, "loss": 1.2936, "step": 8050 }, { "epoch": 0.82, "learning_rate": 5.4036886482608236e-05, "loss": 0.3139, "step": 8051 }, { "epoch": 0.82, "learning_rate": 5.4035454436515904e-05, "loss": 0.554, "step": 8052 }, { "epoch": 0.82, "learning_rate": 5.4034022237471086e-05, "loss": 0.9922, "step": 8053 }, { "epoch": 0.82, "learning_rate": 5.4032589885482895e-05, "loss": 0.4481, "step": 8054 }, { "epoch": 0.82, "learning_rate": 5.403115738056044e-05, "loss": 0.8808, "step": 8055 }, { "epoch": 0.82, "learning_rate": 5.402972472271284e-05, "loss": 0.9388, "step": 8056 }, { "epoch": 0.82, "learning_rate": 5.4028291911949215e-05, "loss": 1.3513, "step": 8057 }, { "epoch": 0.82, "learning_rate": 5.4026858948278675e-05, "loss": 0.8861, "step": 8058 }, { "epoch": 0.82, "learning_rate": 5.402542583171035e-05, "loss": 0.3328, "step": 8059 }, { "epoch": 0.82, "learning_rate": 5.4023992562253357e-05, "loss": 0.4163, "step": 8060 }, { "epoch": 0.82, "learning_rate": 5.4022559139916813e-05, "loss": 0.8689, "step": 8061 }, { "epoch": 0.82, "learning_rate": 5.4021125564709844e-05, "loss": 0.8072, "step": 8062 }, { "epoch": 0.82, "learning_rate": 5.401969183664156e-05, "loss": 0.8861, "step": 8063 }, { "epoch": 0.82, "learning_rate": 5.401825795572111e-05, "loss": 0.3351, "step": 8064 }, { "epoch": 0.82, "learning_rate": 5.401682392195759e-05, "loss": 1.0313, "step": 8065 }, { "epoch": 0.82, "learning_rate": 5.4015389735360145e-05, "loss": 1.4335, "step": 8066 }, { "epoch": 0.82, "learning_rate": 5.40139553959379e-05, "loss": 0.4098, "step": 8067 }, { "epoch": 0.82, "learning_rate": 5.4012520903699966e-05, "loss": 1.5752, "step": 8068 }, { "epoch": 0.82, "learning_rate": 5.4011086258655494e-05, "loss": 0.3557, "step": 8069 }, { "epoch": 0.82, "learning_rate": 5.40096514608136e-05, "loss": 0.9846, "step": 8070 }, { "epoch": 0.82, "learning_rate": 5.4008216510183414e-05, "loss": 1.1083, "step": 8071 }, { "epoch": 0.82, "learning_rate": 5.400678140677408e-05, "loss": 1.4603, "step": 8072 }, { "epoch": 0.82, "learning_rate": 5.4005346150594715e-05, "loss": 0.5966, "step": 8073 }, { "epoch": 0.82, "learning_rate": 5.4003910741654466e-05, "loss": 1.2424, "step": 8074 }, { "epoch": 0.82, "learning_rate": 5.400247517996245e-05, "loss": 1.4373, "step": 8075 }, { "epoch": 0.82, "learning_rate": 5.4001039465527824e-05, "loss": 1.0376, "step": 8076 }, { "epoch": 0.82, "learning_rate": 5.399960359835971e-05, "loss": 0.2754, "step": 8077 }, { "epoch": 0.82, "learning_rate": 5.3998167578467244e-05, "loss": 0.97, "step": 8078 }, { "epoch": 0.82, "learning_rate": 5.3996731405859566e-05, "loss": 1.3274, "step": 8079 }, { "epoch": 0.82, "learning_rate": 5.399529508054583e-05, "loss": 0.3478, "step": 8080 }, { "epoch": 0.82, "learning_rate": 5.3993858602535156e-05, "loss": 0.4266, "step": 8081 }, { "epoch": 0.82, "learning_rate": 5.39924219718367e-05, "loss": 1.2323, "step": 8082 }, { "epoch": 0.82, "learning_rate": 5.39909851884596e-05, "loss": 1.0284, "step": 8083 }, { "epoch": 0.82, "learning_rate": 5.398954825241298e-05, "loss": 0.1579, "step": 8084 }, { "epoch": 0.82, "learning_rate": 5.398811116370601e-05, "loss": 1.4546, "step": 8085 }, { "epoch": 0.82, "learning_rate": 5.398667392234783e-05, "loss": 1.075, "step": 8086 }, { "epoch": 0.82, "learning_rate": 5.398523652834758e-05, "loss": 0.5349, "step": 8087 }, { "epoch": 0.82, "learning_rate": 5.398379898171441e-05, "loss": 0.2661, "step": 8088 }, { "epoch": 0.82, "learning_rate": 5.398236128245747e-05, "loss": 1.1196, "step": 8089 }, { "epoch": 0.82, "learning_rate": 5.3980923430585905e-05, "loss": 0.8033, "step": 8090 }, { "epoch": 0.82, "learning_rate": 5.397948542610886e-05, "loss": 0.1257, "step": 8091 }, { "epoch": 0.82, "learning_rate": 5.397804726903549e-05, "loss": 1.3423, "step": 8092 }, { "epoch": 0.82, "learning_rate": 5.397660895937495e-05, "loss": 0.7701, "step": 8093 }, { "epoch": 0.82, "learning_rate": 5.39751704971364e-05, "loss": 1.4494, "step": 8094 }, { "epoch": 0.82, "learning_rate": 5.3973731882328986e-05, "loss": 1.1779, "step": 8095 }, { "epoch": 0.82, "learning_rate": 5.397229311496186e-05, "loss": 1.5842, "step": 8096 }, { "epoch": 0.82, "learning_rate": 5.397085419504418e-05, "loss": 2.1032, "step": 8097 }, { "epoch": 0.82, "learning_rate": 5.39694151225851e-05, "loss": 0.2669, "step": 8098 }, { "epoch": 0.82, "learning_rate": 5.3967975897593774e-05, "loss": 1.0311, "step": 8099 }, { "epoch": 0.82, "learning_rate": 5.396653652007938e-05, "loss": 0.8566, "step": 8100 }, { "epoch": 0.82, "learning_rate": 5.396509699005106e-05, "loss": 1.274, "step": 8101 }, { "epoch": 0.82, "learning_rate": 5.3963657307517983e-05, "loss": 0.5878, "step": 8102 }, { "epoch": 0.82, "learning_rate": 5.39622174724893e-05, "loss": 1.3545, "step": 8103 }, { "epoch": 0.82, "learning_rate": 5.396077748497418e-05, "loss": 1.1924, "step": 8104 }, { "epoch": 0.82, "learning_rate": 5.39593373449818e-05, "loss": 0.9488, "step": 8105 }, { "epoch": 0.82, "learning_rate": 5.39578970525213e-05, "loss": 1.3453, "step": 8106 }, { "epoch": 0.82, "learning_rate": 5.3956456607601857e-05, "loss": 1.056, "step": 8107 }, { "epoch": 0.82, "learning_rate": 5.3955016010232644e-05, "loss": 0.8688, "step": 8108 }, { "epoch": 0.82, "learning_rate": 5.395357526042281e-05, "loss": 1.3315, "step": 8109 }, { "epoch": 0.82, "learning_rate": 5.395213435818156e-05, "loss": 1.4185, "step": 8110 }, { "epoch": 0.82, "learning_rate": 5.395069330351802e-05, "loss": 1.5236, "step": 8111 }, { "epoch": 0.82, "learning_rate": 5.394925209644138e-05, "loss": 1.0306, "step": 8112 }, { "epoch": 0.82, "learning_rate": 5.394781073696081e-05, "loss": 0.7501, "step": 8113 }, { "epoch": 0.82, "learning_rate": 5.3946369225085485e-05, "loss": 1.3369, "step": 8114 }, { "epoch": 0.82, "learning_rate": 5.394492756082458e-05, "loss": 0.6079, "step": 8115 }, { "epoch": 0.82, "learning_rate": 5.394348574418725e-05, "loss": 1.7226, "step": 8116 }, { "epoch": 0.82, "learning_rate": 5.394204377518271e-05, "loss": 1.4371, "step": 8117 }, { "epoch": 0.82, "learning_rate": 5.3940601653820096e-05, "loss": 1.565, "step": 8118 }, { "epoch": 0.82, "learning_rate": 5.39391593801086e-05, "loss": 1.0849, "step": 8119 }, { "epoch": 0.82, "learning_rate": 5.3937716954057404e-05, "loss": 1.247, "step": 8120 }, { "epoch": 0.82, "learning_rate": 5.393627437567568e-05, "loss": 1.67, "step": 8121 }, { "epoch": 0.82, "learning_rate": 5.393483164497262e-05, "loss": 0.6812, "step": 8122 }, { "epoch": 0.82, "learning_rate": 5.3933388761957386e-05, "loss": 0.3219, "step": 8123 }, { "epoch": 0.82, "learning_rate": 5.393194572663918e-05, "loss": 0.9128, "step": 8124 }, { "epoch": 0.83, "learning_rate": 5.393050253902717e-05, "loss": 0.3757, "step": 8125 }, { "epoch": 0.83, "learning_rate": 5.3929059199130554e-05, "loss": 0.4561, "step": 8126 }, { "epoch": 0.83, "learning_rate": 5.3927615706958495e-05, "loss": 1.0718, "step": 8127 }, { "epoch": 0.83, "learning_rate": 5.39261720625202e-05, "loss": 0.5135, "step": 8128 }, { "epoch": 0.83, "learning_rate": 5.392472826582485e-05, "loss": 0.6971, "step": 8129 }, { "epoch": 0.83, "learning_rate": 5.392328431688162e-05, "loss": 0.2041, "step": 8130 }, { "epoch": 0.83, "learning_rate": 5.392184021569972e-05, "loss": 1.3082, "step": 8131 }, { "epoch": 0.83, "learning_rate": 5.392039596228833e-05, "loss": 1.5793, "step": 8132 }, { "epoch": 0.83, "learning_rate": 5.391895155665663e-05, "loss": 1.1927, "step": 8133 }, { "epoch": 0.83, "learning_rate": 5.3917506998813835e-05, "loss": 1.3903, "step": 8134 }, { "epoch": 0.83, "learning_rate": 5.391606228876912e-05, "loss": 0.77, "step": 8135 }, { "epoch": 0.83, "learning_rate": 5.3914617426531674e-05, "loss": 1.1476, "step": 8136 }, { "epoch": 0.83, "learning_rate": 5.391317241211071e-05, "loss": 0.7352, "step": 8137 }, { "epoch": 0.83, "learning_rate": 5.391172724551541e-05, "loss": 1.6502, "step": 8138 }, { "epoch": 0.83, "learning_rate": 5.391028192675497e-05, "loss": 1.4729, "step": 8139 }, { "epoch": 0.83, "learning_rate": 5.39088364558386e-05, "loss": 0.4136, "step": 8140 }, { "epoch": 0.83, "learning_rate": 5.390739083277548e-05, "loss": 1.1907, "step": 8141 }, { "epoch": 0.83, "learning_rate": 5.3905945057574827e-05, "loss": 0.571, "step": 8142 }, { "epoch": 0.83, "learning_rate": 5.390449913024583e-05, "loss": 1.0514, "step": 8143 }, { "epoch": 0.83, "learning_rate": 5.3903053050797705e-05, "loss": 0.2428, "step": 8144 }, { "epoch": 0.83, "learning_rate": 5.390160681923963e-05, "loss": 1.1869, "step": 8145 }, { "epoch": 0.83, "learning_rate": 5.3900160435580825e-05, "loss": 0.1945, "step": 8146 }, { "epoch": 0.83, "learning_rate": 5.389871389983049e-05, "loss": 0.4948, "step": 8147 }, { "epoch": 0.83, "learning_rate": 5.389726721199784e-05, "loss": 1.0626, "step": 8148 }, { "epoch": 0.83, "learning_rate": 5.389582037209207e-05, "loss": 0.9271, "step": 8149 }, { "epoch": 0.83, "learning_rate": 5.389437338012238e-05, "loss": 0.6779, "step": 8150 }, { "epoch": 0.83, "learning_rate": 5.3892926236098004e-05, "loss": 1.3031, "step": 8151 }, { "epoch": 0.83, "learning_rate": 5.3891478940028116e-05, "loss": 0.3579, "step": 8152 }, { "epoch": 0.83, "learning_rate": 5.389003149192196e-05, "loss": 1.3825, "step": 8153 }, { "epoch": 0.83, "learning_rate": 5.3888583891788716e-05, "loss": 0.223, "step": 8154 }, { "epoch": 0.83, "learning_rate": 5.388713613963762e-05, "loss": 0.4513, "step": 8155 }, { "epoch": 0.83, "learning_rate": 5.388568823547788e-05, "loss": 0.0817, "step": 8156 }, { "epoch": 0.83, "learning_rate": 5.388424017931871e-05, "loss": 1.2321, "step": 8157 }, { "epoch": 0.83, "learning_rate": 5.388279197116931e-05, "loss": 0.9499, "step": 8158 }, { "epoch": 0.83, "learning_rate": 5.3881343611038915e-05, "loss": 1.5718, "step": 8159 }, { "epoch": 0.83, "learning_rate": 5.387989509893673e-05, "loss": 0.9514, "step": 8160 }, { "epoch": 0.83, "learning_rate": 5.387844643487199e-05, "loss": 0.9839, "step": 8161 }, { "epoch": 0.83, "learning_rate": 5.3876997618853884e-05, "loss": 0.7536, "step": 8162 }, { "epoch": 0.83, "learning_rate": 5.3875548650891654e-05, "loss": 0.697, "step": 8163 }, { "epoch": 0.83, "learning_rate": 5.387409953099452e-05, "loss": 0.454, "step": 8164 }, { "epoch": 0.83, "learning_rate": 5.387265025917169e-05, "loss": 1.0411, "step": 8165 }, { "epoch": 0.83, "learning_rate": 5.387120083543241e-05, "loss": 1.0971, "step": 8166 }, { "epoch": 0.83, "learning_rate": 5.386975125978587e-05, "loss": 1.1411, "step": 8167 }, { "epoch": 0.83, "learning_rate": 5.386830153224133e-05, "loss": 0.5457, "step": 8168 }, { "epoch": 0.83, "learning_rate": 5.3866851652807984e-05, "loss": 0.773, "step": 8169 }, { "epoch": 0.83, "learning_rate": 5.386540162149508e-05, "loss": 0.5156, "step": 8170 }, { "epoch": 0.83, "learning_rate": 5.3863951438311846e-05, "loss": 0.2995, "step": 8171 }, { "epoch": 0.83, "learning_rate": 5.386250110326749e-05, "loss": 1.1059, "step": 8172 }, { "epoch": 0.83, "learning_rate": 5.3861050616371274e-05, "loss": 0.4383, "step": 8173 }, { "epoch": 0.83, "learning_rate": 5.3859599977632396e-05, "loss": 0.6272, "step": 8174 }, { "epoch": 0.83, "learning_rate": 5.38581491870601e-05, "loss": 1.6697, "step": 8175 }, { "epoch": 0.83, "learning_rate": 5.385669824466362e-05, "loss": 0.5715, "step": 8176 }, { "epoch": 0.83, "learning_rate": 5.385524715045219e-05, "loss": 0.2641, "step": 8177 }, { "epoch": 0.83, "learning_rate": 5.385379590443503e-05, "loss": 1.2799, "step": 8178 }, { "epoch": 0.83, "learning_rate": 5.38523445066214e-05, "loss": 0.2954, "step": 8179 }, { "epoch": 0.83, "learning_rate": 5.385089295702052e-05, "loss": 0.2329, "step": 8180 }, { "epoch": 0.83, "learning_rate": 5.384944125564164e-05, "loss": 1.5349, "step": 8181 }, { "epoch": 0.83, "learning_rate": 5.384798940249398e-05, "loss": 0.8667, "step": 8182 }, { "epoch": 0.83, "learning_rate": 5.3846537397586784e-05, "loss": 0.6316, "step": 8183 }, { "epoch": 0.83, "learning_rate": 5.38450852409293e-05, "loss": 1.3842, "step": 8184 }, { "epoch": 0.83, "learning_rate": 5.384363293253077e-05, "loss": 0.9685, "step": 8185 }, { "epoch": 0.83, "learning_rate": 5.3842180472400415e-05, "loss": 1.2023, "step": 8186 }, { "epoch": 0.83, "learning_rate": 5.384072786054751e-05, "loss": 0.7942, "step": 8187 }, { "epoch": 0.83, "learning_rate": 5.383927509698127e-05, "loss": 0.8329, "step": 8188 }, { "epoch": 0.83, "learning_rate": 5.383782218171096e-05, "loss": 0.6563, "step": 8189 }, { "epoch": 0.83, "learning_rate": 5.3836369114745826e-05, "loss": 0.4649, "step": 8190 }, { "epoch": 0.83, "learning_rate": 5.3834915896095084e-05, "loss": 0.2362, "step": 8191 }, { "epoch": 0.83, "learning_rate": 5.383346252576803e-05, "loss": 0.7928, "step": 8192 }, { "epoch": 0.83, "learning_rate": 5.383200900377387e-05, "loss": 0.3597, "step": 8193 }, { "epoch": 0.83, "learning_rate": 5.383055533012188e-05, "loss": 0.4464, "step": 8194 }, { "epoch": 0.83, "learning_rate": 5.382910150482129e-05, "loss": 0.6837, "step": 8195 }, { "epoch": 0.83, "learning_rate": 5.382764752788136e-05, "loss": 0.8022, "step": 8196 }, { "epoch": 0.83, "learning_rate": 5.3826193399311366e-05, "loss": 0.6209, "step": 8197 }, { "epoch": 0.83, "learning_rate": 5.382473911912053e-05, "loss": 1.1217, "step": 8198 }, { "epoch": 0.83, "learning_rate": 5.3823284687318115e-05, "loss": 0.886, "step": 8199 }, { "epoch": 0.83, "learning_rate": 5.382183010391339e-05, "loss": 0.3072, "step": 8200 }, { "epoch": 0.83, "learning_rate": 5.382037536891559e-05, "loss": 0.699, "step": 8201 }, { "epoch": 0.83, "learning_rate": 5.381892048233398e-05, "loss": 0.6168, "step": 8202 }, { "epoch": 0.83, "learning_rate": 5.381746544417783e-05, "loss": 0.6521, "step": 8203 }, { "epoch": 0.83, "learning_rate": 5.381601025445639e-05, "loss": 0.6118, "step": 8204 }, { "epoch": 0.83, "learning_rate": 5.3814554913178914e-05, "loss": 0.9958, "step": 8205 }, { "epoch": 0.83, "learning_rate": 5.381309942035468e-05, "loss": 0.546, "step": 8206 }, { "epoch": 0.83, "learning_rate": 5.3811643775992934e-05, "loss": 0.4145, "step": 8207 }, { "epoch": 0.83, "learning_rate": 5.381018798010294e-05, "loss": 0.4685, "step": 8208 }, { "epoch": 0.83, "learning_rate": 5.3808732032693974e-05, "loss": 0.3517, "step": 8209 }, { "epoch": 0.83, "learning_rate": 5.38072759337753e-05, "loss": 1.1924, "step": 8210 }, { "epoch": 0.83, "learning_rate": 5.3805819683356174e-05, "loss": 0.6019, "step": 8211 }, { "epoch": 0.83, "learning_rate": 5.380436328144586e-05, "loss": 1.1517, "step": 8212 }, { "epoch": 0.83, "learning_rate": 5.380290672805365e-05, "loss": 0.2242, "step": 8213 }, { "epoch": 0.83, "learning_rate": 5.380145002318878e-05, "loss": 0.8152, "step": 8214 }, { "epoch": 0.83, "learning_rate": 5.379999316686055e-05, "loss": 0.4976, "step": 8215 }, { "epoch": 0.83, "learning_rate": 5.3798536159078204e-05, "loss": 1.383, "step": 8216 }, { "epoch": 0.83, "learning_rate": 5.3797078999851034e-05, "loss": 1.1082, "step": 8217 }, { "epoch": 0.83, "learning_rate": 5.379562168918831e-05, "loss": 1.332, "step": 8218 }, { "epoch": 0.83, "learning_rate": 5.37941642270993e-05, "loss": 1.2521, "step": 8219 }, { "epoch": 0.83, "learning_rate": 5.3792706613593276e-05, "loss": 0.7029, "step": 8220 }, { "epoch": 0.83, "learning_rate": 5.379124884867953e-05, "loss": 0.7549, "step": 8221 }, { "epoch": 0.83, "learning_rate": 5.378979093236731e-05, "loss": 0.6627, "step": 8222 }, { "epoch": 0.83, "learning_rate": 5.378833286466592e-05, "loss": 0.8923, "step": 8223 }, { "epoch": 0.84, "learning_rate": 5.378687464558463e-05, "loss": 0.7376, "step": 8224 }, { "epoch": 0.84, "learning_rate": 5.378541627513272e-05, "loss": 0.9852, "step": 8225 }, { "epoch": 0.84, "learning_rate": 5.3783957753319466e-05, "loss": 0.8107, "step": 8226 }, { "epoch": 0.84, "learning_rate": 5.3782499080154144e-05, "loss": 0.5132, "step": 8227 }, { "epoch": 0.84, "learning_rate": 5.378104025564606e-05, "loss": 1.0686, "step": 8228 }, { "epoch": 0.84, "learning_rate": 5.377958127980448e-05, "loss": 0.492, "step": 8229 }, { "epoch": 0.84, "learning_rate": 5.3778122152638684e-05, "loss": 1.0281, "step": 8230 }, { "epoch": 0.84, "learning_rate": 5.377666287415797e-05, "loss": 0.8901, "step": 8231 }, { "epoch": 0.84, "learning_rate": 5.377520344437162e-05, "loss": 0.9127, "step": 8232 }, { "epoch": 0.84, "learning_rate": 5.377374386328891e-05, "loss": 1.2276, "step": 8233 }, { "epoch": 0.84, "learning_rate": 5.377228413091915e-05, "loss": 0.5884, "step": 8234 }, { "epoch": 0.84, "learning_rate": 5.377082424727161e-05, "loss": 1.6628, "step": 8235 }, { "epoch": 0.84, "learning_rate": 5.376936421235558e-05, "loss": 1.2376, "step": 8236 }, { "epoch": 0.84, "learning_rate": 5.3767904026180374e-05, "loss": 0.5759, "step": 8237 }, { "epoch": 0.84, "learning_rate": 5.3766443688755265e-05, "loss": 0.4092, "step": 8238 }, { "epoch": 0.84, "learning_rate": 5.376498320008954e-05, "loss": 1.566, "step": 8239 }, { "epoch": 0.84, "learning_rate": 5.376352256019252e-05, "loss": 1.2276, "step": 8240 }, { "epoch": 0.84, "learning_rate": 5.376206176907346e-05, "loss": 1.0346, "step": 8241 }, { "epoch": 0.84, "learning_rate": 5.376060082674169e-05, "loss": 1.1194, "step": 8242 }, { "epoch": 0.84, "learning_rate": 5.37591397332065e-05, "loss": 0.879, "step": 8243 }, { "epoch": 0.84, "learning_rate": 5.375767848847718e-05, "loss": 0.6193, "step": 8244 }, { "epoch": 0.84, "learning_rate": 5.375621709256303e-05, "loss": 0.4481, "step": 8245 }, { "epoch": 0.84, "learning_rate": 5.3754755545473354e-05, "loss": 1.3265, "step": 8246 }, { "epoch": 0.84, "learning_rate": 5.375329384721745e-05, "loss": 0.8484, "step": 8247 }, { "epoch": 0.84, "learning_rate": 5.3751831997804616e-05, "loss": 1.2202, "step": 8248 }, { "epoch": 0.84, "learning_rate": 5.375036999724416e-05, "loss": 0.8404, "step": 8249 }, { "epoch": 0.84, "learning_rate": 5.374890784554539e-05, "loss": 0.3952, "step": 8250 }, { "epoch": 0.84, "learning_rate": 5.37474455427176e-05, "loss": 1.1035, "step": 8251 }, { "epoch": 0.84, "learning_rate": 5.37459830887701e-05, "loss": 0.9759, "step": 8252 }, { "epoch": 0.84, "learning_rate": 5.3744520483712207e-05, "loss": 1.035, "step": 8253 }, { "epoch": 0.84, "learning_rate": 5.374305772755321e-05, "loss": 0.7936, "step": 8254 }, { "epoch": 0.84, "learning_rate": 5.374159482030242e-05, "loss": 1.2119, "step": 8255 }, { "epoch": 0.84, "learning_rate": 5.374013176196916e-05, "loss": 1.337, "step": 8256 }, { "epoch": 0.84, "learning_rate": 5.3738668552562736e-05, "loss": 0.5665, "step": 8257 }, { "epoch": 0.84, "learning_rate": 5.373720519209245e-05, "loss": 1.408, "step": 8258 }, { "epoch": 0.84, "learning_rate": 5.373574168056762e-05, "loss": 1.1812, "step": 8259 }, { "epoch": 0.84, "learning_rate": 5.3734278017997564e-05, "loss": 0.7788, "step": 8260 }, { "epoch": 0.84, "learning_rate": 5.3732814204391584e-05, "loss": 1.4514, "step": 8261 }, { "epoch": 0.84, "learning_rate": 5.373135023975901e-05, "loss": 0.438, "step": 8262 }, { "epoch": 0.84, "learning_rate": 5.372988612410915e-05, "loss": 0.3641, "step": 8263 }, { "epoch": 0.84, "learning_rate": 5.372842185745132e-05, "loss": 0.4293, "step": 8264 }, { "epoch": 0.84, "learning_rate": 5.372695743979484e-05, "loss": 1.9528, "step": 8265 }, { "epoch": 0.84, "learning_rate": 5.3725492871149036e-05, "loss": 0.7908, "step": 8266 }, { "epoch": 0.84, "learning_rate": 5.37240281515232e-05, "loss": 0.5781, "step": 8267 }, { "epoch": 0.84, "learning_rate": 5.37225632809267e-05, "loss": 1.3558, "step": 8268 }, { "epoch": 0.84, "learning_rate": 5.3721098259368825e-05, "loss": 0.3672, "step": 8269 }, { "epoch": 0.84, "learning_rate": 5.371963308685889e-05, "loss": 1.5859, "step": 8270 }, { "epoch": 0.84, "learning_rate": 5.3718167763406245e-05, "loss": 0.7089, "step": 8271 }, { "epoch": 0.84, "learning_rate": 5.3716702289020204e-05, "loss": 1.5599, "step": 8272 }, { "epoch": 0.84, "learning_rate": 5.371523666371009e-05, "loss": 1.5552, "step": 8273 }, { "epoch": 0.84, "learning_rate": 5.371377088748523e-05, "loss": 1.1627, "step": 8274 }, { "epoch": 0.84, "learning_rate": 5.371230496035496e-05, "loss": 0.8455, "step": 8275 }, { "epoch": 0.84, "learning_rate": 5.371083888232859e-05, "loss": 1.3767, "step": 8276 }, { "epoch": 0.84, "learning_rate": 5.370937265341547e-05, "loss": 0.679, "step": 8277 }, { "epoch": 0.84, "learning_rate": 5.370790627362492e-05, "loss": 0.4338, "step": 8278 }, { "epoch": 0.84, "learning_rate": 5.3706439742966274e-05, "loss": 0.4543, "step": 8279 }, { "epoch": 0.84, "learning_rate": 5.370497306144886e-05, "loss": 1.3578, "step": 8280 }, { "epoch": 0.84, "learning_rate": 5.370350622908202e-05, "loss": 0.4613, "step": 8281 }, { "epoch": 0.84, "learning_rate": 5.3702039245875086e-05, "loss": 0.2544, "step": 8282 }, { "epoch": 0.84, "learning_rate": 5.370057211183739e-05, "loss": 0.9867, "step": 8283 }, { "epoch": 0.84, "learning_rate": 5.369910482697826e-05, "loss": 2.096, "step": 8284 }, { "epoch": 0.84, "learning_rate": 5.369763739130705e-05, "loss": 1.6217, "step": 8285 }, { "epoch": 0.84, "learning_rate": 5.369616980483309e-05, "loss": 0.4812, "step": 8286 }, { "epoch": 0.84, "learning_rate": 5.369470206756572e-05, "loss": 0.6908, "step": 8287 }, { "epoch": 0.84, "learning_rate": 5.369323417951428e-05, "loss": 0.4236, "step": 8288 }, { "epoch": 0.84, "learning_rate": 5.369176614068811e-05, "loss": 0.3672, "step": 8289 }, { "epoch": 0.84, "learning_rate": 5.3690297951096544e-05, "loss": 1.5754, "step": 8290 }, { "epoch": 0.84, "learning_rate": 5.3688829610748945e-05, "loss": 0.4998, "step": 8291 }, { "epoch": 0.84, "learning_rate": 5.368736111965465e-05, "loss": 1.4565, "step": 8292 }, { "epoch": 0.84, "learning_rate": 5.3685892477822993e-05, "loss": 1.0141, "step": 8293 }, { "epoch": 0.84, "learning_rate": 5.368442368526333e-05, "loss": 0.2994, "step": 8294 }, { "epoch": 0.84, "learning_rate": 5.3682954741985e-05, "loss": 1.6018, "step": 8295 }, { "epoch": 0.84, "learning_rate": 5.368148564799736e-05, "loss": 0.8759, "step": 8296 }, { "epoch": 0.84, "learning_rate": 5.3680016403309755e-05, "loss": 1.6051, "step": 8297 }, { "epoch": 0.84, "learning_rate": 5.367854700793153e-05, "loss": 0.3145, "step": 8298 }, { "epoch": 0.84, "learning_rate": 5.367707746187204e-05, "loss": 1.167, "step": 8299 }, { "epoch": 0.84, "learning_rate": 5.367560776514064e-05, "loss": 1.4416, "step": 8300 }, { "epoch": 0.84, "learning_rate": 5.3674137917746676e-05, "loss": 0.6476, "step": 8301 }, { "epoch": 0.84, "learning_rate": 5.36726679196995e-05, "loss": 1.2872, "step": 8302 }, { "epoch": 0.84, "learning_rate": 5.3671197771008475e-05, "loss": 0.5675, "step": 8303 }, { "epoch": 0.84, "learning_rate": 5.366972747168296e-05, "loss": 0.4553, "step": 8304 }, { "epoch": 0.84, "learning_rate": 5.366825702173229e-05, "loss": 0.4087, "step": 8305 }, { "epoch": 0.84, "learning_rate": 5.366678642116585e-05, "loss": 1.0998, "step": 8306 }, { "epoch": 0.84, "learning_rate": 5.366531566999297e-05, "loss": 1.3622, "step": 8307 }, { "epoch": 0.84, "learning_rate": 5.366384476822304e-05, "loss": 0.7, "step": 8308 }, { "epoch": 0.84, "learning_rate": 5.366237371586539e-05, "loss": 0.4471, "step": 8309 }, { "epoch": 0.84, "learning_rate": 5.3660902512929406e-05, "loss": 0.7711, "step": 8310 }, { "epoch": 0.84, "learning_rate": 5.365943115942443e-05, "loss": 1.2572, "step": 8311 }, { "epoch": 0.84, "learning_rate": 5.365795965535984e-05, "loss": 0.9504, "step": 8312 }, { "epoch": 0.84, "learning_rate": 5.3656488000744993e-05, "loss": 1.3816, "step": 8313 }, { "epoch": 0.84, "learning_rate": 5.3655016195589256e-05, "loss": 0.949, "step": 8314 }, { "epoch": 0.84, "learning_rate": 5.3653544239902e-05, "loss": 1.1444, "step": 8315 }, { "epoch": 0.84, "learning_rate": 5.3652072133692574e-05, "loss": 0.9097, "step": 8316 }, { "epoch": 0.84, "learning_rate": 5.365059987697037e-05, "loss": 1.2181, "step": 8317 }, { "epoch": 0.84, "learning_rate": 5.364912746974474e-05, "loss": 0.8571, "step": 8318 }, { "epoch": 0.84, "learning_rate": 5.3647654912025066e-05, "loss": 0.4328, "step": 8319 }, { "epoch": 0.84, "learning_rate": 5.3646182203820703e-05, "loss": 1.5133, "step": 8320 }, { "epoch": 0.84, "learning_rate": 5.364470934514103e-05, "loss": 0.8117, "step": 8321 }, { "epoch": 0.85, "learning_rate": 5.364323633599543e-05, "loss": 0.4109, "step": 8322 }, { "epoch": 0.85, "learning_rate": 5.3641763176393274e-05, "loss": 0.1737, "step": 8323 }, { "epoch": 0.85, "learning_rate": 5.364028986634392e-05, "loss": 1.0912, "step": 8324 }, { "epoch": 0.85, "learning_rate": 5.3638816405856754e-05, "loss": 0.7446, "step": 8325 }, { "epoch": 0.85, "learning_rate": 5.3637342794941163e-05, "loss": 0.4407, "step": 8326 }, { "epoch": 0.85, "learning_rate": 5.363586903360651e-05, "loss": 1.43, "step": 8327 }, { "epoch": 0.85, "learning_rate": 5.363439512186217e-05, "loss": 0.6075, "step": 8328 }, { "epoch": 0.85, "learning_rate": 5.363292105971754e-05, "loss": 1.4124, "step": 8329 }, { "epoch": 0.85, "learning_rate": 5.3631446847181986e-05, "loss": 0.9967, "step": 8330 }, { "epoch": 0.85, "learning_rate": 5.3629972484264894e-05, "loss": 0.8709, "step": 8331 }, { "epoch": 0.85, "learning_rate": 5.3628497970975655e-05, "loss": 0.7139, "step": 8332 }, { "epoch": 0.85, "learning_rate": 5.3627023307323634e-05, "loss": 0.9695, "step": 8333 }, { "epoch": 0.85, "learning_rate": 5.362554849331823e-05, "loss": 0.8265, "step": 8334 }, { "epoch": 0.85, "learning_rate": 5.3624073528968816e-05, "loss": 0.4142, "step": 8335 }, { "epoch": 0.85, "learning_rate": 5.362259841428479e-05, "loss": 1.3335, "step": 8336 }, { "epoch": 0.85, "learning_rate": 5.3621123149275534e-05, "loss": 1.1933, "step": 8337 }, { "epoch": 0.85, "learning_rate": 5.361964773395044e-05, "loss": 1.5012, "step": 8338 }, { "epoch": 0.85, "learning_rate": 5.361817216831889e-05, "loss": 0.7413, "step": 8339 }, { "epoch": 0.85, "learning_rate": 5.3616696452390275e-05, "loss": 0.6879, "step": 8340 }, { "epoch": 0.85, "learning_rate": 5.361522058617399e-05, "loss": 0.9353, "step": 8341 }, { "epoch": 0.85, "learning_rate": 5.361374456967943e-05, "loss": 0.7212, "step": 8342 }, { "epoch": 0.85, "learning_rate": 5.361226840291597e-05, "loss": 0.5763, "step": 8343 }, { "epoch": 0.85, "learning_rate": 5.361079208589302e-05, "loss": 1.2229, "step": 8344 }, { "epoch": 0.85, "learning_rate": 5.360931561861998e-05, "loss": 0.3827, "step": 8345 }, { "epoch": 0.85, "learning_rate": 5.360783900110623e-05, "loss": 0.4196, "step": 8346 }, { "epoch": 0.85, "learning_rate": 5.3606362233361173e-05, "loss": 1.2371, "step": 8347 }, { "epoch": 0.85, "learning_rate": 5.360488531539421e-05, "loss": 1.4126, "step": 8348 }, { "epoch": 0.85, "learning_rate": 5.3603408247214735e-05, "loss": 0.8794, "step": 8349 }, { "epoch": 0.85, "learning_rate": 5.360193102883215e-05, "loss": 0.8617, "step": 8350 }, { "epoch": 0.85, "learning_rate": 5.3600453660255854e-05, "loss": 1.4716, "step": 8351 }, { "epoch": 0.85, "learning_rate": 5.359897614149525e-05, "loss": 0.7823, "step": 8352 }, { "epoch": 0.85, "learning_rate": 5.359749847255974e-05, "loss": 0.7597, "step": 8353 }, { "epoch": 0.85, "learning_rate": 5.359602065345873e-05, "loss": 0.9841, "step": 8354 }, { "epoch": 0.85, "learning_rate": 5.359454268420161e-05, "loss": 0.2977, "step": 8355 }, { "epoch": 0.85, "learning_rate": 5.359306456479781e-05, "loss": 0.4144, "step": 8356 }, { "epoch": 0.85, "learning_rate": 5.359158629525671e-05, "loss": 0.5834, "step": 8357 }, { "epoch": 0.85, "learning_rate": 5.359010787558773e-05, "loss": 0.3406, "step": 8358 }, { "epoch": 0.85, "learning_rate": 5.358862930580028e-05, "loss": 1.181, "step": 8359 }, { "epoch": 0.85, "learning_rate": 5.358715058590377e-05, "loss": 1.2065, "step": 8360 }, { "epoch": 0.85, "learning_rate": 5.358567171590761e-05, "loss": 0.8934, "step": 8361 }, { "epoch": 0.85, "learning_rate": 5.35841926958212e-05, "loss": 1.4711, "step": 8362 }, { "epoch": 0.85, "learning_rate": 5.358271352565396e-05, "loss": 1.3827, "step": 8363 }, { "epoch": 0.85, "learning_rate": 5.358123420541531e-05, "loss": 0.7511, "step": 8364 }, { "epoch": 0.85, "learning_rate": 5.357975473511465e-05, "loss": 1.1469, "step": 8365 }, { "epoch": 0.85, "learning_rate": 5.3578275114761404e-05, "loss": 0.3824, "step": 8366 }, { "epoch": 0.85, "learning_rate": 5.357679534436498e-05, "loss": 0.384, "step": 8367 }, { "epoch": 0.85, "learning_rate": 5.357531542393481e-05, "loss": 0.6761, "step": 8368 }, { "epoch": 0.85, "learning_rate": 5.3573835353480294e-05, "loss": 0.6949, "step": 8369 }, { "epoch": 0.85, "learning_rate": 5.357235513301086e-05, "loss": 0.8299, "step": 8370 }, { "epoch": 0.85, "learning_rate": 5.3570874762535916e-05, "loss": 0.3872, "step": 8371 }, { "epoch": 0.85, "learning_rate": 5.35693942420649e-05, "loss": 1.6778, "step": 8372 }, { "epoch": 0.85, "learning_rate": 5.356791357160723e-05, "loss": 0.6663, "step": 8373 }, { "epoch": 0.85, "learning_rate": 5.3566432751172314e-05, "loss": 1.1177, "step": 8374 }, { "epoch": 0.85, "learning_rate": 5.3564951780769586e-05, "loss": 0.7646, "step": 8375 }, { "epoch": 0.85, "learning_rate": 5.356347066040847e-05, "loss": 0.7312, "step": 8376 }, { "epoch": 0.85, "learning_rate": 5.356198939009839e-05, "loss": 0.6116, "step": 8377 }, { "epoch": 0.85, "learning_rate": 5.356050796984878e-05, "loss": 0.5153, "step": 8378 }, { "epoch": 0.85, "learning_rate": 5.3559026399669065e-05, "loss": 0.8224, "step": 8379 }, { "epoch": 0.85, "learning_rate": 5.3557544679568656e-05, "loss": 0.8319, "step": 8380 }, { "epoch": 0.85, "learning_rate": 5.3556062809557e-05, "loss": 1.8766, "step": 8381 }, { "epoch": 0.85, "learning_rate": 5.355458078964352e-05, "loss": 0.205, "step": 8382 }, { "epoch": 0.85, "learning_rate": 5.355309861983764e-05, "loss": 0.5681, "step": 8383 }, { "epoch": 0.85, "learning_rate": 5.355161630014881e-05, "loss": 0.9869, "step": 8384 }, { "epoch": 0.85, "learning_rate": 5.355013383058645e-05, "loss": 0.7785, "step": 8385 }, { "epoch": 0.85, "learning_rate": 5.3548651211160005e-05, "loss": 1.489, "step": 8386 }, { "epoch": 0.85, "learning_rate": 5.354716844187889e-05, "loss": 0.5513, "step": 8387 }, { "epoch": 0.85, "learning_rate": 5.354568552275256e-05, "loss": 0.9085, "step": 8388 }, { "epoch": 0.85, "learning_rate": 5.354420245379045e-05, "loss": 0.7716, "step": 8389 }, { "epoch": 0.85, "learning_rate": 5.3542719235001976e-05, "loss": 1.0735, "step": 8390 }, { "epoch": 0.85, "learning_rate": 5.35412358663966e-05, "loss": 1.4566, "step": 8391 }, { "epoch": 0.85, "learning_rate": 5.353975234798376e-05, "loss": 1.1874, "step": 8392 }, { "epoch": 0.85, "learning_rate": 5.353826867977289e-05, "loss": 1.1612, "step": 8393 }, { "epoch": 0.85, "learning_rate": 5.3536784861773436e-05, "loss": 0.3396, "step": 8394 }, { "epoch": 0.85, "learning_rate": 5.353530089399483e-05, "loss": 0.9394, "step": 8395 }, { "epoch": 0.85, "learning_rate": 5.3533816776446516e-05, "loss": 0.645, "step": 8396 }, { "epoch": 0.85, "learning_rate": 5.353233250913795e-05, "loss": 0.9081, "step": 8397 }, { "epoch": 0.85, "learning_rate": 5.3530848092078584e-05, "loss": 1.0612, "step": 8398 }, { "epoch": 0.85, "learning_rate": 5.352936352527784e-05, "loss": 0.7362, "step": 8399 }, { "epoch": 0.85, "learning_rate": 5.3527878808745174e-05, "loss": 1.1324, "step": 8400 }, { "epoch": 0.85, "learning_rate": 5.352639394249005e-05, "loss": 1.2015, "step": 8401 }, { "epoch": 0.85, "learning_rate": 5.3524908926521895e-05, "loss": 0.167, "step": 8402 }, { "epoch": 0.85, "learning_rate": 5.352342376085018e-05, "loss": 0.593, "step": 8403 }, { "epoch": 0.85, "learning_rate": 5.352193844548434e-05, "loss": 1.1316, "step": 8404 }, { "epoch": 0.85, "learning_rate": 5.3520452980433825e-05, "loss": 1.0999, "step": 8405 }, { "epoch": 0.85, "learning_rate": 5.35189673657081e-05, "loss": 0.3469, "step": 8406 }, { "epoch": 0.85, "learning_rate": 5.3517481601316616e-05, "loss": 0.6117, "step": 8407 }, { "epoch": 0.85, "learning_rate": 5.3515995687268825e-05, "loss": 0.4625, "step": 8408 }, { "epoch": 0.85, "learning_rate": 5.351450962357418e-05, "loss": 1.0954, "step": 8409 }, { "epoch": 0.85, "learning_rate": 5.351302341024214e-05, "loss": 1.8786, "step": 8410 }, { "epoch": 0.85, "learning_rate": 5.351153704728217e-05, "loss": 0.4191, "step": 8411 }, { "epoch": 0.85, "learning_rate": 5.351005053470372e-05, "loss": 0.8929, "step": 8412 }, { "epoch": 0.85, "learning_rate": 5.350856387251625e-05, "loss": 0.5899, "step": 8413 }, { "epoch": 0.85, "learning_rate": 5.350707706072922e-05, "loss": 1.6129, "step": 8414 }, { "epoch": 0.85, "learning_rate": 5.35055900993521e-05, "loss": 0.5545, "step": 8415 }, { "epoch": 0.85, "learning_rate": 5.3504102988394344e-05, "loss": 1.1637, "step": 8416 }, { "epoch": 0.85, "learning_rate": 5.350261572786542e-05, "loss": 1.0819, "step": 8417 }, { "epoch": 0.85, "learning_rate": 5.350112831777478e-05, "loss": 0.6801, "step": 8418 }, { "epoch": 0.85, "learning_rate": 5.349964075813192e-05, "loss": 0.6826, "step": 8419 }, { "epoch": 0.85, "learning_rate": 5.3498153048946273e-05, "loss": 1.5029, "step": 8420 }, { "epoch": 0.86, "learning_rate": 5.349666519022732e-05, "loss": 0.2415, "step": 8421 }, { "epoch": 0.86, "learning_rate": 5.349517718198453e-05, "loss": 0.9872, "step": 8422 }, { "epoch": 0.86, "learning_rate": 5.3493689024227374e-05, "loss": 0.678, "step": 8423 }, { "epoch": 0.86, "learning_rate": 5.349220071696531e-05, "loss": 0.794, "step": 8424 }, { "epoch": 0.86, "learning_rate": 5.349071226020782e-05, "loss": 0.4654, "step": 8425 }, { "epoch": 0.86, "learning_rate": 5.3489223653964376e-05, "loss": 0.6321, "step": 8426 }, { "epoch": 0.86, "learning_rate": 5.348773489824444e-05, "loss": 1.2325, "step": 8427 }, { "epoch": 0.86, "learning_rate": 5.348624599305751e-05, "loss": 1.5838, "step": 8428 }, { "epoch": 0.86, "learning_rate": 5.348475693841303e-05, "loss": 0.9588, "step": 8429 }, { "epoch": 0.86, "learning_rate": 5.34832677343205e-05, "loss": 1.2477, "step": 8430 }, { "epoch": 0.86, "learning_rate": 5.348177838078938e-05, "loss": 1.2008, "step": 8431 }, { "epoch": 0.86, "learning_rate": 5.3480288877829165e-05, "loss": 1.1091, "step": 8432 }, { "epoch": 0.86, "learning_rate": 5.347879922544932e-05, "loss": 1.0644, "step": 8433 }, { "epoch": 0.86, "learning_rate": 5.3477309423659326e-05, "loss": 0.7742, "step": 8434 }, { "epoch": 0.86, "learning_rate": 5.3475819472468675e-05, "loss": 0.4222, "step": 8435 }, { "epoch": 0.86, "learning_rate": 5.347432937188683e-05, "loss": 0.3056, "step": 8436 }, { "epoch": 0.86, "learning_rate": 5.347283912192328e-05, "loss": 0.9835, "step": 8437 }, { "epoch": 0.86, "learning_rate": 5.347134872258752e-05, "loss": 0.4974, "step": 8438 }, { "epoch": 0.86, "learning_rate": 5.3469858173889026e-05, "loss": 1.1752, "step": 8439 }, { "epoch": 0.86, "learning_rate": 5.346836747583728e-05, "loss": 0.8958, "step": 8440 }, { "epoch": 0.86, "learning_rate": 5.346687662844177e-05, "loss": 0.323, "step": 8441 }, { "epoch": 0.86, "learning_rate": 5.3465385631711985e-05, "loss": 0.6192, "step": 8442 }, { "epoch": 0.86, "learning_rate": 5.346389448565741e-05, "loss": 0.1356, "step": 8443 }, { "epoch": 0.86, "learning_rate": 5.346240319028754e-05, "loss": 0.9847, "step": 8444 }, { "epoch": 0.86, "learning_rate": 5.346091174561186e-05, "loss": 0.6762, "step": 8445 }, { "epoch": 0.86, "learning_rate": 5.345942015163986e-05, "loss": 0.3406, "step": 8446 }, { "epoch": 0.86, "learning_rate": 5.3457928408381044e-05, "loss": 1.4562, "step": 8447 }, { "epoch": 0.86, "learning_rate": 5.3456436515844885e-05, "loss": 0.8358, "step": 8448 }, { "epoch": 0.86, "learning_rate": 5.34549444740409e-05, "loss": 1.4738, "step": 8449 }, { "epoch": 0.86, "learning_rate": 5.345345228297856e-05, "loss": 1.906, "step": 8450 }, { "epoch": 0.86, "learning_rate": 5.345195994266738e-05, "loss": 0.9956, "step": 8451 }, { "epoch": 0.86, "learning_rate": 5.3450467453116835e-05, "loss": 0.2785, "step": 8452 }, { "epoch": 0.86, "learning_rate": 5.344897481433645e-05, "loss": 0.6076, "step": 8453 }, { "epoch": 0.86, "learning_rate": 5.34474820263357e-05, "loss": 1.0478, "step": 8454 }, { "epoch": 0.86, "learning_rate": 5.3445989089124095e-05, "loss": 0.9223, "step": 8455 }, { "epoch": 0.86, "learning_rate": 5.3444496002711144e-05, "loss": 0.9917, "step": 8456 }, { "epoch": 0.86, "learning_rate": 5.344300276710633e-05, "loss": 0.4175, "step": 8457 }, { "epoch": 0.86, "learning_rate": 5.3441509382319174e-05, "loss": 0.2222, "step": 8458 }, { "epoch": 0.86, "learning_rate": 5.344001584835916e-05, "loss": 0.516, "step": 8459 }, { "epoch": 0.86, "learning_rate": 5.34385221652358e-05, "loss": 1.799, "step": 8460 }, { "epoch": 0.86, "learning_rate": 5.3437028332958616e-05, "loss": 0.7417, "step": 8461 }, { "epoch": 0.86, "learning_rate": 5.343553435153709e-05, "loss": 0.6801, "step": 8462 }, { "epoch": 0.86, "learning_rate": 5.343404022098074e-05, "loss": 0.875, "step": 8463 }, { "epoch": 0.86, "learning_rate": 5.3432545941299065e-05, "loss": 0.8025, "step": 8464 }, { "epoch": 0.86, "learning_rate": 5.34310515125016e-05, "loss": 0.2095, "step": 8465 }, { "epoch": 0.86, "learning_rate": 5.3429556934597825e-05, "loss": 0.9834, "step": 8466 }, { "epoch": 0.86, "learning_rate": 5.3428062207597266e-05, "loss": 0.3541, "step": 8467 }, { "epoch": 0.86, "learning_rate": 5.342656733150942e-05, "loss": 1.1525, "step": 8468 }, { "epoch": 0.86, "learning_rate": 5.3425072306343826e-05, "loss": 1.2495, "step": 8469 }, { "epoch": 0.86, "learning_rate": 5.342357713210998e-05, "loss": 0.5237, "step": 8470 }, { "epoch": 0.86, "learning_rate": 5.3422081808817395e-05, "loss": 0.2721, "step": 8471 }, { "epoch": 0.86, "learning_rate": 5.3420586336475595e-05, "loss": 1.1047, "step": 8472 }, { "epoch": 0.86, "learning_rate": 5.341909071509409e-05, "loss": 1.0645, "step": 8473 }, { "epoch": 0.86, "learning_rate": 5.3417594944682414e-05, "loss": 1.3329, "step": 8474 }, { "epoch": 0.86, "learning_rate": 5.341609902525006e-05, "loss": 1.47, "step": 8475 }, { "epoch": 0.86, "learning_rate": 5.3414602956806554e-05, "loss": 1.13, "step": 8476 }, { "epoch": 0.86, "learning_rate": 5.341310673936143e-05, "loss": 0.6966, "step": 8477 }, { "epoch": 0.86, "learning_rate": 5.34116103729242e-05, "loss": 1.3827, "step": 8478 }, { "epoch": 0.86, "learning_rate": 5.341011385750439e-05, "loss": 1.3341, "step": 8479 }, { "epoch": 0.86, "learning_rate": 5.340861719311152e-05, "loss": 0.9619, "step": 8480 }, { "epoch": 0.86, "learning_rate": 5.340712037975512e-05, "loss": 1.1482, "step": 8481 }, { "epoch": 0.86, "learning_rate": 5.340562341744471e-05, "loss": 0.8814, "step": 8482 }, { "epoch": 0.86, "learning_rate": 5.340412630618981e-05, "loss": 1.1832, "step": 8483 }, { "epoch": 0.86, "learning_rate": 5.340262904599995e-05, "loss": 1.072, "step": 8484 }, { "epoch": 0.86, "learning_rate": 5.3401131636884674e-05, "loss": 0.7486, "step": 8485 }, { "epoch": 0.86, "learning_rate": 5.339963407885349e-05, "loss": 0.4645, "step": 8486 }, { "epoch": 0.86, "learning_rate": 5.339813637191594e-05, "loss": 0.4199, "step": 8487 }, { "epoch": 0.86, "learning_rate": 5.339663851608155e-05, "loss": 1.0042, "step": 8488 }, { "epoch": 0.86, "learning_rate": 5.339514051135985e-05, "loss": 0.7424, "step": 8489 }, { "epoch": 0.86, "learning_rate": 5.339364235776038e-05, "loss": 1.6581, "step": 8490 }, { "epoch": 0.86, "learning_rate": 5.3392144055292664e-05, "loss": 0.5362, "step": 8491 }, { "epoch": 0.86, "learning_rate": 5.339064560396625e-05, "loss": 0.1856, "step": 8492 }, { "epoch": 0.86, "learning_rate": 5.3389147003790664e-05, "loss": 0.4335, "step": 8493 }, { "epoch": 0.86, "learning_rate": 5.338764825477544e-05, "loss": 1.0073, "step": 8494 }, { "epoch": 0.86, "learning_rate": 5.3386149356930124e-05, "loss": 0.8742, "step": 8495 }, { "epoch": 0.86, "learning_rate": 5.3384650310264246e-05, "loss": 0.8585, "step": 8496 }, { "epoch": 0.86, "learning_rate": 5.338315111478735e-05, "loss": 0.313, "step": 8497 }, { "epoch": 0.86, "learning_rate": 5.338165177050897e-05, "loss": 0.4755, "step": 8498 }, { "epoch": 0.86, "learning_rate": 5.3380152277438665e-05, "loss": 1.3651, "step": 8499 }, { "epoch": 0.86, "learning_rate": 5.337865263558596e-05, "loss": 0.3689, "step": 8500 }, { "epoch": 0.86, "learning_rate": 5.3377152844960396e-05, "loss": 1.313, "step": 8501 }, { "epoch": 0.86, "learning_rate": 5.337565290557153e-05, "loss": 0.4701, "step": 8502 }, { "epoch": 0.86, "learning_rate": 5.337415281742891e-05, "loss": 0.4087, "step": 8503 }, { "epoch": 0.86, "learning_rate": 5.337265258054207e-05, "loss": 1.9655, "step": 8504 }, { "epoch": 0.86, "learning_rate": 5.3371152194920554e-05, "loss": 0.4534, "step": 8505 }, { "epoch": 0.86, "learning_rate": 5.336965166057392e-05, "loss": 0.7515, "step": 8506 }, { "epoch": 0.86, "learning_rate": 5.336815097751171e-05, "loss": 2.0942, "step": 8507 }, { "epoch": 0.86, "learning_rate": 5.336665014574348e-05, "loss": 0.2163, "step": 8508 }, { "epoch": 0.86, "learning_rate": 5.336514916527878e-05, "loss": 0.9064, "step": 8509 }, { "epoch": 0.86, "learning_rate": 5.336364803612715e-05, "loss": 1.9561, "step": 8510 }, { "epoch": 0.86, "learning_rate": 5.336214675829816e-05, "loss": 1.4806, "step": 8511 }, { "epoch": 0.86, "learning_rate": 5.3360645331801354e-05, "loss": 0.7158, "step": 8512 }, { "epoch": 0.86, "learning_rate": 5.3359143756646285e-05, "loss": 1.5858, "step": 8513 }, { "epoch": 0.86, "learning_rate": 5.335764203284252e-05, "loss": 0.5258, "step": 8514 }, { "epoch": 0.86, "learning_rate": 5.335614016039959e-05, "loss": 0.4125, "step": 8515 }, { "epoch": 0.86, "learning_rate": 5.335463813932709e-05, "loss": 0.6966, "step": 8516 }, { "epoch": 0.86, "learning_rate": 5.335313596963455e-05, "loss": 0.3971, "step": 8517 }, { "epoch": 0.86, "learning_rate": 5.3351633651331524e-05, "loss": 1.2828, "step": 8518 }, { "epoch": 0.87, "learning_rate": 5.3350131184427595e-05, "loss": 0.6863, "step": 8519 }, { "epoch": 0.87, "learning_rate": 5.334862856893232e-05, "loss": 0.4915, "step": 8520 }, { "epoch": 0.87, "learning_rate": 5.334712580485525e-05, "loss": 0.4495, "step": 8521 }, { "epoch": 0.87, "learning_rate": 5.3345622892205946e-05, "loss": 0.635, "step": 8522 }, { "epoch": 0.87, "learning_rate": 5.334411983099398e-05, "loss": 0.2127, "step": 8523 }, { "epoch": 0.87, "learning_rate": 5.334261662122892e-05, "loss": 1.2459, "step": 8524 }, { "epoch": 0.87, "learning_rate": 5.3341113262920333e-05, "loss": 1.5985, "step": 8525 }, { "epoch": 0.87, "learning_rate": 5.3339609756077775e-05, "loss": 0.8442, "step": 8526 }, { "epoch": 0.87, "learning_rate": 5.333810610071082e-05, "loss": 0.623, "step": 8527 }, { "epoch": 0.87, "learning_rate": 5.333660229682904e-05, "loss": 0.5507, "step": 8528 }, { "epoch": 0.87, "learning_rate": 5.3335098344442e-05, "loss": 0.8626, "step": 8529 }, { "epoch": 0.87, "learning_rate": 5.333359424355926e-05, "loss": 0.3284, "step": 8530 }, { "epoch": 0.87, "learning_rate": 5.333208999419041e-05, "loss": 0.6669, "step": 8531 }, { "epoch": 0.87, "learning_rate": 5.333058559634502e-05, "loss": 1.1455, "step": 8532 }, { "epoch": 0.87, "learning_rate": 5.3329081050032656e-05, "loss": 1.36, "step": 8533 }, { "epoch": 0.87, "learning_rate": 5.332757635526289e-05, "loss": 0.9497, "step": 8534 }, { "epoch": 0.87, "learning_rate": 5.3326071512045306e-05, "loss": 1.1202, "step": 8535 }, { "epoch": 0.87, "learning_rate": 5.3324566520389485e-05, "loss": 1.2617, "step": 8536 }, { "epoch": 0.87, "learning_rate": 5.3323061380304984e-05, "loss": 1.0476, "step": 8537 }, { "epoch": 0.87, "learning_rate": 5.33215560918014e-05, "loss": 1.1937, "step": 8538 }, { "epoch": 0.87, "learning_rate": 5.33200506548883e-05, "loss": 1.189, "step": 8539 }, { "epoch": 0.87, "learning_rate": 5.3318545069575264e-05, "loss": 1.1352, "step": 8540 }, { "epoch": 0.87, "learning_rate": 5.331703933587189e-05, "loss": 1.0038, "step": 8541 }, { "epoch": 0.87, "learning_rate": 5.331553345378774e-05, "loss": 0.5976, "step": 8542 }, { "epoch": 0.87, "learning_rate": 5.331402742333241e-05, "loss": 0.7877, "step": 8543 }, { "epoch": 0.87, "learning_rate": 5.331252124451547e-05, "loss": 0.3272, "step": 8544 }, { "epoch": 0.87, "learning_rate": 5.331101491734653e-05, "loss": 0.3159, "step": 8545 }, { "epoch": 0.87, "learning_rate": 5.330950844183514e-05, "loss": 1.4679, "step": 8546 }, { "epoch": 0.87, "learning_rate": 5.3308001817990926e-05, "loss": 0.2675, "step": 8547 }, { "epoch": 0.87, "learning_rate": 5.3306495045823434e-05, "loss": 0.6838, "step": 8548 }, { "epoch": 0.87, "learning_rate": 5.330498812534229e-05, "loss": 1.7569, "step": 8549 }, { "epoch": 0.87, "learning_rate": 5.330348105655706e-05, "loss": 0.4169, "step": 8550 }, { "epoch": 0.87, "learning_rate": 5.330197383947734e-05, "loss": 0.9801, "step": 8551 }, { "epoch": 0.87, "learning_rate": 5.330046647411272e-05, "loss": 0.8416, "step": 8552 }, { "epoch": 0.87, "learning_rate": 5.32989589604728e-05, "loss": 0.235, "step": 8553 }, { "epoch": 0.87, "learning_rate": 5.329745129856717e-05, "loss": 1.1779, "step": 8554 }, { "epoch": 0.87, "learning_rate": 5.329594348840542e-05, "loss": 1.1538, "step": 8555 }, { "epoch": 0.87, "learning_rate": 5.329443552999715e-05, "loss": 0.9087, "step": 8556 }, { "epoch": 0.87, "learning_rate": 5.329292742335195e-05, "loss": 0.7772, "step": 8557 }, { "epoch": 0.87, "learning_rate": 5.3291419168479435e-05, "loss": 0.3254, "step": 8558 }, { "epoch": 0.87, "learning_rate": 5.3289910765389176e-05, "loss": 0.7375, "step": 8559 }, { "epoch": 0.87, "learning_rate": 5.328840221409079e-05, "loss": 0.836, "step": 8560 }, { "epoch": 0.87, "learning_rate": 5.328689351459387e-05, "loss": 0.2472, "step": 8561 }, { "epoch": 0.87, "learning_rate": 5.328538466690801e-05, "loss": 0.5237, "step": 8562 }, { "epoch": 0.87, "learning_rate": 5.328387567104282e-05, "loss": 0.9084, "step": 8563 }, { "epoch": 0.87, "learning_rate": 5.328236652700792e-05, "loss": 1.0607, "step": 8564 }, { "epoch": 0.87, "learning_rate": 5.328085723481289e-05, "loss": 0.3572, "step": 8565 }, { "epoch": 0.87, "learning_rate": 5.3279347794467335e-05, "loss": 0.7411, "step": 8566 }, { "epoch": 0.87, "learning_rate": 5.3277838205980866e-05, "loss": 0.6567, "step": 8567 }, { "epoch": 0.87, "learning_rate": 5.32763284693631e-05, "loss": 0.4957, "step": 8568 }, { "epoch": 0.87, "learning_rate": 5.327481858462363e-05, "loss": 0.881, "step": 8569 }, { "epoch": 0.87, "learning_rate": 5.327330855177206e-05, "loss": 1.0526, "step": 8570 }, { "epoch": 0.87, "learning_rate": 5.327179837081802e-05, "loss": 0.2502, "step": 8571 }, { "epoch": 0.87, "learning_rate": 5.3270288041771095e-05, "loss": 0.9869, "step": 8572 }, { "epoch": 0.87, "learning_rate": 5.326877756464092e-05, "loss": 0.7086, "step": 8573 }, { "epoch": 0.87, "learning_rate": 5.326726693943709e-05, "loss": 0.8163, "step": 8574 }, { "epoch": 0.87, "learning_rate": 5.326575616616923e-05, "loss": 0.7865, "step": 8575 }, { "epoch": 0.87, "learning_rate": 5.326424524484695e-05, "loss": 0.6772, "step": 8576 }, { "epoch": 0.87, "learning_rate": 5.3262734175479864e-05, "loss": 1.1853, "step": 8577 }, { "epoch": 0.87, "learning_rate": 5.3261222958077575e-05, "loss": 0.4949, "step": 8578 }, { "epoch": 0.87, "learning_rate": 5.325971159264973e-05, "loss": 0.4286, "step": 8579 }, { "epoch": 0.87, "learning_rate": 5.3258200079205906e-05, "loss": 0.9415, "step": 8580 }, { "epoch": 0.87, "learning_rate": 5.3256688417755764e-05, "loss": 1.3172, "step": 8581 }, { "epoch": 0.87, "learning_rate": 5.32551766083089e-05, "loss": 0.5207, "step": 8582 }, { "epoch": 0.87, "learning_rate": 5.325366465087493e-05, "loss": 1.6625, "step": 8583 }, { "epoch": 0.87, "learning_rate": 5.32521525454635e-05, "loss": 1.1126, "step": 8584 }, { "epoch": 0.87, "learning_rate": 5.3250640292084205e-05, "loss": 0.9775, "step": 8585 }, { "epoch": 0.87, "learning_rate": 5.324912789074668e-05, "loss": 1.1991, "step": 8586 }, { "epoch": 0.87, "learning_rate": 5.324761534146056e-05, "loss": 0.2263, "step": 8587 }, { "epoch": 0.87, "learning_rate": 5.324610264423544e-05, "loss": 1.079, "step": 8588 }, { "epoch": 0.87, "learning_rate": 5.324458979908099e-05, "loss": 0.2535, "step": 8589 }, { "epoch": 0.87, "learning_rate": 5.32430768060068e-05, "loss": 0.3387, "step": 8590 }, { "epoch": 0.87, "learning_rate": 5.324156366502251e-05, "loss": 0.5898, "step": 8591 }, { "epoch": 0.87, "learning_rate": 5.3240050376137755e-05, "loss": 0.3619, "step": 8592 }, { "epoch": 0.87, "learning_rate": 5.323853693936216e-05, "loss": 0.5719, "step": 8593 }, { "epoch": 0.87, "learning_rate": 5.3237023354705355e-05, "loss": 0.8759, "step": 8594 }, { "epoch": 0.87, "learning_rate": 5.323550962217697e-05, "loss": 0.9666, "step": 8595 }, { "epoch": 0.87, "learning_rate": 5.323399574178664e-05, "loss": 0.589, "step": 8596 }, { "epoch": 0.87, "learning_rate": 5.3232481713544014e-05, "loss": 1.2946, "step": 8597 }, { "epoch": 0.87, "learning_rate": 5.32309675374587e-05, "loss": 1.2963, "step": 8598 }, { "epoch": 0.87, "learning_rate": 5.322945321354035e-05, "loss": 0.4245, "step": 8599 }, { "epoch": 0.87, "learning_rate": 5.3227938741798593e-05, "loss": 0.381, "step": 8600 }, { "epoch": 0.87, "learning_rate": 5.322642412224307e-05, "loss": 0.0991, "step": 8601 }, { "epoch": 0.87, "learning_rate": 5.3224909354883424e-05, "loss": 0.3615, "step": 8602 }, { "epoch": 0.87, "learning_rate": 5.322339443972929e-05, "loss": 1.6935, "step": 8603 }, { "epoch": 0.87, "learning_rate": 5.322187937679031e-05, "loss": 0.5715, "step": 8604 }, { "epoch": 0.87, "learning_rate": 5.3220364166076115e-05, "loss": 1.1057, "step": 8605 }, { "epoch": 0.87, "learning_rate": 5.3218848807596365e-05, "loss": 1.182, "step": 8606 }, { "epoch": 0.87, "learning_rate": 5.3217333301360695e-05, "loss": 0.9134, "step": 8607 }, { "epoch": 0.87, "learning_rate": 5.3215817647378736e-05, "loss": 0.5385, "step": 8608 }, { "epoch": 0.87, "learning_rate": 5.321430184566016e-05, "loss": 1.5268, "step": 8609 }, { "epoch": 0.87, "learning_rate": 5.3212785896214586e-05, "loss": 1.2683, "step": 8610 }, { "epoch": 0.87, "learning_rate": 5.321126979905168e-05, "loss": 0.6503, "step": 8611 }, { "epoch": 0.87, "learning_rate": 5.3209753554181075e-05, "loss": 1.0115, "step": 8612 }, { "epoch": 0.87, "learning_rate": 5.320823716161244e-05, "loss": 0.9253, "step": 8613 }, { "epoch": 0.87, "learning_rate": 5.3206720621355406e-05, "loss": 1.1875, "step": 8614 }, { "epoch": 0.87, "learning_rate": 5.320520393341963e-05, "loss": 0.8763, "step": 8615 }, { "epoch": 0.87, "learning_rate": 5.3203687097814766e-05, "loss": 0.4326, "step": 8616 }, { "epoch": 0.88, "learning_rate": 5.320217011455046e-05, "loss": 1.5884, "step": 8617 }, { "epoch": 0.88, "learning_rate": 5.320065298363638e-05, "loss": 1.3745, "step": 8618 }, { "epoch": 0.88, "learning_rate": 5.319913570508216e-05, "loss": 1.0333, "step": 8619 }, { "epoch": 0.88, "learning_rate": 5.3197618278897466e-05, "loss": 1.3999, "step": 8620 }, { "epoch": 0.88, "learning_rate": 5.319610070509196e-05, "loss": 0.3954, "step": 8621 }, { "epoch": 0.88, "learning_rate": 5.3194582983675284e-05, "loss": 0.6648, "step": 8622 }, { "epoch": 0.88, "learning_rate": 5.319306511465712e-05, "loss": 1.3404, "step": 8623 }, { "epoch": 0.88, "learning_rate": 5.31915470980471e-05, "loss": 0.8706, "step": 8624 }, { "epoch": 0.88, "learning_rate": 5.31900289338549e-05, "loss": 0.3933, "step": 8625 }, { "epoch": 0.88, "learning_rate": 5.318851062209019e-05, "loss": 1.5368, "step": 8626 }, { "epoch": 0.88, "learning_rate": 5.318699216276261e-05, "loss": 0.2797, "step": 8627 }, { "epoch": 0.88, "learning_rate": 5.3185473555881825e-05, "loss": 0.379, "step": 8628 }, { "epoch": 0.88, "learning_rate": 5.318395480145752e-05, "loss": 0.3192, "step": 8629 }, { "epoch": 0.88, "learning_rate": 5.318243589949934e-05, "loss": 1.2887, "step": 8630 }, { "epoch": 0.88, "learning_rate": 5.318091685001695e-05, "loss": 0.9943, "step": 8631 }, { "epoch": 0.88, "learning_rate": 5.317939765302003e-05, "loss": 1.5074, "step": 8632 }, { "epoch": 0.88, "learning_rate": 5.317787830851825e-05, "loss": 0.995, "step": 8633 }, { "epoch": 0.88, "learning_rate": 5.317635881652126e-05, "loss": 1.1952, "step": 8634 }, { "epoch": 0.88, "learning_rate": 5.3174839177038734e-05, "loss": 1.3277, "step": 8635 }, { "epoch": 0.88, "learning_rate": 5.3173319390080356e-05, "loss": 0.6386, "step": 8636 }, { "epoch": 0.88, "learning_rate": 5.317179945565579e-05, "loss": 1.0686, "step": 8637 }, { "epoch": 0.88, "learning_rate": 5.3170279373774705e-05, "loss": 1.4648, "step": 8638 }, { "epoch": 0.88, "learning_rate": 5.316875914444678e-05, "loss": 0.5397, "step": 8639 }, { "epoch": 0.88, "learning_rate": 5.316723876768168e-05, "loss": 1.1908, "step": 8640 }, { "epoch": 0.88, "learning_rate": 5.316571824348909e-05, "loss": 0.834, "step": 8641 }, { "epoch": 0.88, "learning_rate": 5.316419757187868e-05, "loss": 0.7619, "step": 8642 }, { "epoch": 0.88, "learning_rate": 5.316267675286012e-05, "loss": 0.992, "step": 8643 }, { "epoch": 0.88, "learning_rate": 5.316115578644311e-05, "loss": 0.7952, "step": 8644 }, { "epoch": 0.88, "learning_rate": 5.3159634672637305e-05, "loss": 0.9075, "step": 8645 }, { "epoch": 0.88, "learning_rate": 5.31581134114524e-05, "loss": 0.4979, "step": 8646 }, { "epoch": 0.88, "learning_rate": 5.315659200289807e-05, "loss": 0.6953, "step": 8647 }, { "epoch": 0.88, "learning_rate": 5.3155070446984e-05, "loss": 0.7369, "step": 8648 }, { "epoch": 0.88, "learning_rate": 5.3153548743719874e-05, "loss": 0.1854, "step": 8649 }, { "epoch": 0.88, "learning_rate": 5.315202689311537e-05, "loss": 1.2053, "step": 8650 }, { "epoch": 0.88, "learning_rate": 5.3150504895180167e-05, "loss": 1.3465, "step": 8651 }, { "epoch": 0.88, "learning_rate": 5.314898274992396e-05, "loss": 1.0665, "step": 8652 }, { "epoch": 0.88, "learning_rate": 5.314746045735644e-05, "loss": 0.7616, "step": 8653 }, { "epoch": 0.88, "learning_rate": 5.314593801748728e-05, "loss": 1.663, "step": 8654 }, { "epoch": 0.88, "learning_rate": 5.314441543032618e-05, "loss": 0.7361, "step": 8655 }, { "epoch": 0.88, "learning_rate": 5.3142892695882825e-05, "loss": 0.5449, "step": 8656 }, { "epoch": 0.88, "learning_rate": 5.3141369814166904e-05, "loss": 1.3313, "step": 8657 }, { "epoch": 0.88, "learning_rate": 5.3139846785188105e-05, "loss": 1.1283, "step": 8658 }, { "epoch": 0.88, "learning_rate": 5.313832360895613e-05, "loss": 1.0833, "step": 8659 }, { "epoch": 0.88, "learning_rate": 5.313680028548066e-05, "loss": 1.311, "step": 8660 }, { "epoch": 0.88, "learning_rate": 5.31352768147714e-05, "loss": 0.377, "step": 8661 }, { "epoch": 0.88, "learning_rate": 5.313375319683804e-05, "loss": 0.8984, "step": 8662 }, { "epoch": 0.88, "learning_rate": 5.313222943169027e-05, "loss": 0.5059, "step": 8663 }, { "epoch": 0.88, "learning_rate": 5.313070551933779e-05, "loss": 1.468, "step": 8664 }, { "epoch": 0.88, "learning_rate": 5.312918145979031e-05, "loss": 1.0401, "step": 8665 }, { "epoch": 0.88, "learning_rate": 5.312765725305751e-05, "loss": 0.3747, "step": 8666 }, { "epoch": 0.88, "learning_rate": 5.3126132899149096e-05, "loss": 1.2767, "step": 8667 }, { "epoch": 0.88, "learning_rate": 5.312460839807477e-05, "loss": 0.4348, "step": 8668 }, { "epoch": 0.88, "learning_rate": 5.312308374984424e-05, "loss": 0.6982, "step": 8669 }, { "epoch": 0.88, "learning_rate": 5.31215589544672e-05, "loss": 0.5766, "step": 8670 }, { "epoch": 0.88, "learning_rate": 5.312003401195335e-05, "loss": 1.2753, "step": 8671 }, { "epoch": 0.88, "learning_rate": 5.3118508922312396e-05, "loss": 1.2208, "step": 8672 }, { "epoch": 0.88, "learning_rate": 5.311698368555406e-05, "loss": 0.9879, "step": 8673 }, { "epoch": 0.88, "learning_rate": 5.311545830168802e-05, "loss": 1.0936, "step": 8674 }, { "epoch": 0.88, "learning_rate": 5.3113932770724e-05, "loss": 1.7386, "step": 8675 }, { "epoch": 0.88, "learning_rate": 5.3112407092671716e-05, "loss": 0.9846, "step": 8676 }, { "epoch": 0.88, "learning_rate": 5.3110881267540847e-05, "loss": 1.3518, "step": 8677 }, { "epoch": 0.88, "learning_rate": 5.310935529534114e-05, "loss": 1.1701, "step": 8678 }, { "epoch": 0.88, "learning_rate": 5.310782917608227e-05, "loss": 0.5285, "step": 8679 }, { "epoch": 0.88, "learning_rate": 5.310630290977398e-05, "loss": 0.5269, "step": 8680 }, { "epoch": 0.88, "learning_rate": 5.310477649642596e-05, "loss": 1.0488, "step": 8681 }, { "epoch": 0.88, "learning_rate": 5.3103249936047935e-05, "loss": 0.8648, "step": 8682 }, { "epoch": 0.88, "learning_rate": 5.3101723228649615e-05, "loss": 0.989, "step": 8683 }, { "epoch": 0.88, "learning_rate": 5.310019637424072e-05, "loss": 0.2363, "step": 8684 }, { "epoch": 0.88, "learning_rate": 5.309866937283095e-05, "loss": 0.5448, "step": 8685 }, { "epoch": 0.88, "learning_rate": 5.309714222443006e-05, "loss": 1.0885, "step": 8686 }, { "epoch": 0.88, "learning_rate": 5.309561492904772e-05, "loss": 0.3706, "step": 8687 }, { "epoch": 0.88, "learning_rate": 5.309408748669369e-05, "loss": 1.4526, "step": 8688 }, { "epoch": 0.88, "learning_rate": 5.309255989737767e-05, "loss": 1.5615, "step": 8689 }, { "epoch": 0.88, "learning_rate": 5.3091032161109364e-05, "loss": 1.2609, "step": 8690 }, { "epoch": 0.88, "learning_rate": 5.308950427789853e-05, "loss": 0.8662, "step": 8691 }, { "epoch": 0.88, "learning_rate": 5.308797624775487e-05, "loss": 1.0346, "step": 8692 }, { "epoch": 0.88, "learning_rate": 5.3086448070688117e-05, "loss": 0.9806, "step": 8693 }, { "epoch": 0.88, "learning_rate": 5.308491974670798e-05, "loss": 0.6534, "step": 8694 }, { "epoch": 0.88, "learning_rate": 5.3083391275824214e-05, "loss": 0.6255, "step": 8695 }, { "epoch": 0.88, "learning_rate": 5.3081862658046514e-05, "loss": 0.2166, "step": 8696 }, { "epoch": 0.88, "learning_rate": 5.308033389338463e-05, "loss": 1.6356, "step": 8697 }, { "epoch": 0.88, "learning_rate": 5.307880498184827e-05, "loss": 2.0639, "step": 8698 }, { "epoch": 0.88, "learning_rate": 5.3077275923447183e-05, "loss": 1.4723, "step": 8699 }, { "epoch": 0.88, "learning_rate": 5.3075746718191094e-05, "loss": 0.575, "step": 8700 }, { "epoch": 0.88, "learning_rate": 5.307421736608972e-05, "loss": 0.2657, "step": 8701 }, { "epoch": 0.88, "learning_rate": 5.307268786715281e-05, "loss": 1.5434, "step": 8702 }, { "epoch": 0.88, "learning_rate": 5.307115822139009e-05, "loss": 1.3821, "step": 8703 }, { "epoch": 0.88, "learning_rate": 5.306962842881129e-05, "loss": 1.7784, "step": 8704 }, { "epoch": 0.88, "learning_rate": 5.306809848942617e-05, "loss": 0.9313, "step": 8705 }, { "epoch": 0.88, "learning_rate": 5.306656840324443e-05, "loss": 1.4978, "step": 8706 }, { "epoch": 0.88, "learning_rate": 5.306503817027583e-05, "loss": 0.8073, "step": 8707 }, { "epoch": 0.88, "learning_rate": 5.30635077905301e-05, "loss": 1.1686, "step": 8708 }, { "epoch": 0.88, "learning_rate": 5.306197726401698e-05, "loss": 0.9419, "step": 8709 }, { "epoch": 0.88, "learning_rate": 5.306044659074622e-05, "loss": 0.5803, "step": 8710 }, { "epoch": 0.88, "learning_rate": 5.305891577072754e-05, "loss": 0.3925, "step": 8711 }, { "epoch": 0.88, "learning_rate": 5.305738480397069e-05, "loss": 0.1291, "step": 8712 }, { "epoch": 0.88, "learning_rate": 5.3055853690485415e-05, "loss": 1.4934, "step": 8713 }, { "epoch": 0.88, "learning_rate": 5.305432243028146e-05, "loss": 1.0401, "step": 8714 }, { "epoch": 0.88, "learning_rate": 5.305279102336857e-05, "loss": 0.8786, "step": 8715 }, { "epoch": 0.89, "learning_rate": 5.305125946975648e-05, "loss": 0.8793, "step": 8716 }, { "epoch": 0.89, "learning_rate": 5.3049727769454957e-05, "loss": 0.4525, "step": 8717 }, { "epoch": 0.89, "learning_rate": 5.3048195922473724e-05, "loss": 0.6013, "step": 8718 }, { "epoch": 0.89, "learning_rate": 5.3046663928822544e-05, "loss": 0.5719, "step": 8719 }, { "epoch": 0.89, "learning_rate": 5.3045131788511166e-05, "loss": 0.9392, "step": 8720 }, { "epoch": 0.89, "learning_rate": 5.304359950154933e-05, "loss": 0.6069, "step": 8721 }, { "epoch": 0.89, "learning_rate": 5.304206706794679e-05, "loss": 1.7225, "step": 8722 }, { "epoch": 0.89, "learning_rate": 5.304053448771331e-05, "loss": 0.4662, "step": 8723 }, { "epoch": 0.89, "learning_rate": 5.303900176085863e-05, "loss": 1.3998, "step": 8724 }, { "epoch": 0.89, "learning_rate": 5.30374688873925e-05, "loss": 1.3371, "step": 8725 }, { "epoch": 0.89, "learning_rate": 5.303593586732469e-05, "loss": 1.331, "step": 8726 }, { "epoch": 0.89, "learning_rate": 5.3034402700664956e-05, "loss": 0.5971, "step": 8727 }, { "epoch": 0.89, "learning_rate": 5.303286938742304e-05, "loss": 0.4752, "step": 8728 }, { "epoch": 0.89, "learning_rate": 5.3031335927608695e-05, "loss": 0.7533, "step": 8729 }, { "epoch": 0.89, "learning_rate": 5.302980232123169e-05, "loss": 0.3201, "step": 8730 }, { "epoch": 0.89, "learning_rate": 5.3028268568301795e-05, "loss": 0.5268, "step": 8731 }, { "epoch": 0.89, "learning_rate": 5.302673466882875e-05, "loss": 0.6059, "step": 8732 }, { "epoch": 0.89, "learning_rate": 5.302520062282233e-05, "loss": 1.0474, "step": 8733 }, { "epoch": 0.89, "learning_rate": 5.30236664302923e-05, "loss": 0.4715, "step": 8734 }, { "epoch": 0.89, "learning_rate": 5.30221320912484e-05, "loss": 1.9104, "step": 8735 }, { "epoch": 0.89, "learning_rate": 5.302059760570042e-05, "loss": 1.1856, "step": 8736 }, { "epoch": 0.89, "learning_rate": 5.301906297365811e-05, "loss": 1.4771, "step": 8737 }, { "epoch": 0.89, "learning_rate": 5.3017528195131246e-05, "loss": 0.2806, "step": 8738 }, { "epoch": 0.89, "learning_rate": 5.3015993270129584e-05, "loss": 0.7656, "step": 8739 }, { "epoch": 0.89, "learning_rate": 5.3014458198662894e-05, "loss": 1.089, "step": 8740 }, { "epoch": 0.89, "learning_rate": 5.3012922980740956e-05, "loss": 0.298, "step": 8741 }, { "epoch": 0.89, "learning_rate": 5.3011387616373526e-05, "loss": 1.2168, "step": 8742 }, { "epoch": 0.89, "learning_rate": 5.3009852105570376e-05, "loss": 0.6159, "step": 8743 }, { "epoch": 0.89, "learning_rate": 5.3008316448341284e-05, "loss": 1.5137, "step": 8744 }, { "epoch": 0.89, "learning_rate": 5.3006780644696015e-05, "loss": 0.7687, "step": 8745 }, { "epoch": 0.89, "learning_rate": 5.300524469464435e-05, "loss": 1.8436, "step": 8746 }, { "epoch": 0.89, "learning_rate": 5.300370859819606e-05, "loss": 0.2758, "step": 8747 }, { "epoch": 0.89, "learning_rate": 5.300217235536092e-05, "loss": 0.2511, "step": 8748 }, { "epoch": 0.89, "learning_rate": 5.30006359661487e-05, "loss": 1.0109, "step": 8749 }, { "epoch": 0.89, "learning_rate": 5.2999099430569184e-05, "loss": 1.0968, "step": 8750 }, { "epoch": 0.89, "learning_rate": 5.299756274863215e-05, "loss": 0.7085, "step": 8751 }, { "epoch": 0.89, "learning_rate": 5.2996025920347376e-05, "loss": 1.411, "step": 8752 }, { "epoch": 0.89, "learning_rate": 5.299448894572464e-05, "loss": 1.0465, "step": 8753 }, { "epoch": 0.89, "learning_rate": 5.299295182477373e-05, "loss": 1.1579, "step": 8754 }, { "epoch": 0.89, "learning_rate": 5.2991414557504415e-05, "loss": 0.4207, "step": 8755 }, { "epoch": 0.89, "learning_rate": 5.2989877143926486e-05, "loss": 0.1477, "step": 8756 }, { "epoch": 0.89, "learning_rate": 5.298833958404972e-05, "loss": 1.5197, "step": 8757 }, { "epoch": 0.89, "learning_rate": 5.2986801877883915e-05, "loss": 1.1373, "step": 8758 }, { "epoch": 0.89, "learning_rate": 5.298526402543884e-05, "loss": 0.4349, "step": 8759 }, { "epoch": 0.89, "learning_rate": 5.298372602672429e-05, "loss": 1.3458, "step": 8760 }, { "epoch": 0.89, "learning_rate": 5.2982187881750056e-05, "loss": 0.4663, "step": 8761 }, { "epoch": 0.89, "learning_rate": 5.298064959052592e-05, "loss": 1.2405, "step": 8762 }, { "epoch": 0.89, "learning_rate": 5.297911115306167e-05, "loss": 0.7087, "step": 8763 }, { "epoch": 0.89, "learning_rate": 5.29775725693671e-05, "loss": 0.3902, "step": 8764 }, { "epoch": 0.89, "learning_rate": 5.2976033839451995e-05, "loss": 0.5532, "step": 8765 }, { "epoch": 0.89, "learning_rate": 5.2974494963326156e-05, "loss": 1.1318, "step": 8766 }, { "epoch": 0.89, "learning_rate": 5.297295594099937e-05, "loss": 0.4861, "step": 8767 }, { "epoch": 0.89, "learning_rate": 5.2971416772481427e-05, "loss": 1.3855, "step": 8768 }, { "epoch": 0.89, "learning_rate": 5.2969877457782134e-05, "loss": 0.8064, "step": 8769 }, { "epoch": 0.89, "learning_rate": 5.2968337996911274e-05, "loss": 0.2755, "step": 8770 }, { "epoch": 0.89, "learning_rate": 5.296679838987865e-05, "loss": 1.2272, "step": 8771 }, { "epoch": 0.89, "learning_rate": 5.296525863669406e-05, "loss": 0.7966, "step": 8772 }, { "epoch": 0.89, "learning_rate": 5.29637187373673e-05, "loss": 0.9946, "step": 8773 }, { "epoch": 0.89, "learning_rate": 5.2962178691908166e-05, "loss": 1.5824, "step": 8774 }, { "epoch": 0.89, "learning_rate": 5.296063850032647e-05, "loss": 1.0706, "step": 8775 }, { "epoch": 0.89, "learning_rate": 5.295909816263199e-05, "loss": 0.9574, "step": 8776 }, { "epoch": 0.89, "learning_rate": 5.295755767883456e-05, "loss": 0.4045, "step": 8777 }, { "epoch": 0.89, "learning_rate": 5.2956017048943955e-05, "loss": 0.3512, "step": 8778 }, { "epoch": 0.89, "learning_rate": 5.295447627296999e-05, "loss": 0.788, "step": 8779 }, { "epoch": 0.89, "learning_rate": 5.295293535092248e-05, "loss": 0.6182, "step": 8780 }, { "epoch": 0.89, "learning_rate": 5.2951394282811216e-05, "loss": 1.3137, "step": 8781 }, { "epoch": 0.89, "learning_rate": 5.2949853068646014e-05, "loss": 0.3243, "step": 8782 }, { "epoch": 0.89, "learning_rate": 5.294831170843667e-05, "loss": 0.5123, "step": 8783 }, { "epoch": 0.89, "learning_rate": 5.2946770202193e-05, "loss": 0.5667, "step": 8784 }, { "epoch": 0.89, "learning_rate": 5.2945228549924824e-05, "loss": 0.9874, "step": 8785 }, { "epoch": 0.89, "learning_rate": 5.294368675164194e-05, "loss": 0.9727, "step": 8786 }, { "epoch": 0.89, "learning_rate": 5.294214480735415e-05, "loss": 0.7047, "step": 8787 }, { "epoch": 0.89, "learning_rate": 5.294060271707129e-05, "loss": 1.0119, "step": 8788 }, { "epoch": 0.89, "learning_rate": 5.293906048080316e-05, "loss": 0.8232, "step": 8789 }, { "epoch": 0.89, "learning_rate": 5.2937518098559577e-05, "loss": 0.4567, "step": 8790 }, { "epoch": 0.89, "learning_rate": 5.293597557035034e-05, "loss": 1.1119, "step": 8791 }, { "epoch": 0.89, "learning_rate": 5.2934432896185294e-05, "loss": 1.1228, "step": 8792 }, { "epoch": 0.89, "learning_rate": 5.293289007607424e-05, "loss": 0.5632, "step": 8793 }, { "epoch": 0.89, "learning_rate": 5.293134711002699e-05, "loss": 1.4215, "step": 8794 }, { "epoch": 0.89, "learning_rate": 5.292980399805338e-05, "loss": 0.6716, "step": 8795 }, { "epoch": 0.89, "learning_rate": 5.292826074016322e-05, "loss": 0.643, "step": 8796 }, { "epoch": 0.89, "learning_rate": 5.292671733636633e-05, "loss": 1.5089, "step": 8797 }, { "epoch": 0.89, "learning_rate": 5.2925173786672524e-05, "loss": 0.7985, "step": 8798 }, { "epoch": 0.89, "learning_rate": 5.292363009109164e-05, "loss": 1.2015, "step": 8799 }, { "epoch": 0.89, "learning_rate": 5.29220862496335e-05, "loss": 1.1374, "step": 8800 }, { "epoch": 0.89, "learning_rate": 5.292054226230792e-05, "loss": 1.3312, "step": 8801 }, { "epoch": 0.89, "learning_rate": 5.291899812912472e-05, "loss": 0.9082, "step": 8802 }, { "epoch": 0.89, "learning_rate": 5.291745385009374e-05, "loss": 1.1349, "step": 8803 }, { "epoch": 0.89, "learning_rate": 5.291590942522481e-05, "loss": 0.951, "step": 8804 }, { "epoch": 0.89, "learning_rate": 5.291436485452773e-05, "loss": 0.5819, "step": 8805 }, { "epoch": 0.89, "learning_rate": 5.2912820138012364e-05, "loss": 0.7954, "step": 8806 }, { "epoch": 0.89, "learning_rate": 5.2911275275688524e-05, "loss": 0.8069, "step": 8807 }, { "epoch": 0.89, "learning_rate": 5.290973026756605e-05, "loss": 0.4893, "step": 8808 }, { "epoch": 0.89, "learning_rate": 5.290818511365476e-05, "loss": 0.7576, "step": 8809 }, { "epoch": 0.89, "learning_rate": 5.29066398139645e-05, "loss": 1.405, "step": 8810 }, { "epoch": 0.89, "learning_rate": 5.290509436850509e-05, "loss": 0.5089, "step": 8811 }, { "epoch": 0.89, "learning_rate": 5.290354877728639e-05, "loss": 1.2916, "step": 8812 }, { "epoch": 0.89, "learning_rate": 5.290200304031821e-05, "loss": 0.399, "step": 8813 }, { "epoch": 0.9, "learning_rate": 5.290045715761039e-05, "loss": 0.8151, "step": 8814 }, { "epoch": 0.9, "learning_rate": 5.2898911129172775e-05, "loss": 1.0303, "step": 8815 }, { "epoch": 0.9, "learning_rate": 5.28973649550152e-05, "loss": 0.4334, "step": 8816 }, { "epoch": 0.9, "learning_rate": 5.28958186351475e-05, "loss": 1.0678, "step": 8817 }, { "epoch": 0.9, "learning_rate": 5.289427216957952e-05, "loss": 0.3638, "step": 8818 }, { "epoch": 0.9, "learning_rate": 5.2892725558321114e-05, "loss": 0.2905, "step": 8819 }, { "epoch": 0.9, "learning_rate": 5.2891178801382106e-05, "loss": 1.4761, "step": 8820 }, { "epoch": 0.9, "learning_rate": 5.2889631898772336e-05, "loss": 0.3012, "step": 8821 }, { "epoch": 0.9, "learning_rate": 5.288808485050167e-05, "loss": 1.5339, "step": 8822 }, { "epoch": 0.9, "learning_rate": 5.288653765657993e-05, "loss": 1.4925, "step": 8823 }, { "epoch": 0.9, "learning_rate": 5.288499031701698e-05, "loss": 0.6301, "step": 8824 }, { "epoch": 0.9, "learning_rate": 5.2883442831822645e-05, "loss": 1.7662, "step": 8825 }, { "epoch": 0.9, "learning_rate": 5.288189520100679e-05, "loss": 1.3331, "step": 8826 }, { "epoch": 0.9, "learning_rate": 5.288034742457926e-05, "loss": 2.3506, "step": 8827 }, { "epoch": 0.9, "learning_rate": 5.28787995025499e-05, "loss": 1.1578, "step": 8828 }, { "epoch": 0.9, "learning_rate": 5.287725143492857e-05, "loss": 1.1053, "step": 8829 }, { "epoch": 0.9, "learning_rate": 5.287570322172511e-05, "loss": 0.9858, "step": 8830 }, { "epoch": 0.9, "learning_rate": 5.287415486294938e-05, "loss": 1.1212, "step": 8831 }, { "epoch": 0.9, "learning_rate": 5.287260635861123e-05, "loss": 1.1469, "step": 8832 }, { "epoch": 0.9, "learning_rate": 5.287105770872051e-05, "loss": 0.4533, "step": 8833 }, { "epoch": 0.9, "learning_rate": 5.286950891328708e-05, "loss": 1.0604, "step": 8834 }, { "epoch": 0.9, "learning_rate": 5.28679599723208e-05, "loss": 1.4104, "step": 8835 }, { "epoch": 0.9, "learning_rate": 5.286641088583152e-05, "loss": 1.0967, "step": 8836 }, { "epoch": 0.9, "learning_rate": 5.28648616538291e-05, "loss": 0.3835, "step": 8837 }, { "epoch": 0.9, "learning_rate": 5.2863312276323396e-05, "loss": 1.5914, "step": 8838 }, { "epoch": 0.9, "learning_rate": 5.286176275332427e-05, "loss": 0.2631, "step": 8839 }, { "epoch": 0.9, "learning_rate": 5.2860213084841586e-05, "loss": 1.0917, "step": 8840 }, { "epoch": 0.9, "learning_rate": 5.2858663270885204e-05, "loss": 1.1273, "step": 8841 }, { "epoch": 0.9, "learning_rate": 5.285711331146498e-05, "loss": 0.4735, "step": 8842 }, { "epoch": 0.9, "learning_rate": 5.285556320659079e-05, "loss": 1.4781, "step": 8843 }, { "epoch": 0.9, "learning_rate": 5.2854012956272474e-05, "loss": 1.0526, "step": 8844 }, { "epoch": 0.9, "learning_rate": 5.285246256051993e-05, "loss": 0.9881, "step": 8845 }, { "epoch": 0.9, "learning_rate": 5.2850912019343006e-05, "loss": 0.2752, "step": 8846 }, { "epoch": 0.9, "learning_rate": 5.284936133275156e-05, "loss": 0.6091, "step": 8847 }, { "epoch": 0.9, "learning_rate": 5.2847810500755483e-05, "loss": 1.2725, "step": 8848 }, { "epoch": 0.9, "learning_rate": 5.2846259523364626e-05, "loss": 1.3827, "step": 8849 }, { "epoch": 0.9, "learning_rate": 5.284470840058886e-05, "loss": 0.9733, "step": 8850 }, { "epoch": 0.9, "learning_rate": 5.284315713243807e-05, "loss": 1.6768, "step": 8851 }, { "epoch": 0.9, "learning_rate": 5.2841605718922114e-05, "loss": 1.1631, "step": 8852 }, { "epoch": 0.9, "learning_rate": 5.284005416005087e-05, "loss": 0.9271, "step": 8853 }, { "epoch": 0.9, "learning_rate": 5.283850245583421e-05, "loss": 1.356, "step": 8854 }, { "epoch": 0.9, "learning_rate": 5.2836950606282005e-05, "loss": 1.2213, "step": 8855 }, { "epoch": 0.9, "learning_rate": 5.2835398611404136e-05, "loss": 0.4237, "step": 8856 }, { "epoch": 0.9, "learning_rate": 5.283384647121047e-05, "loss": 0.9159, "step": 8857 }, { "epoch": 0.9, "learning_rate": 5.2832294185710906e-05, "loss": 0.8184, "step": 8858 }, { "epoch": 0.9, "learning_rate": 5.2830741754915296e-05, "loss": 0.6115, "step": 8859 }, { "epoch": 0.9, "learning_rate": 5.282918917883354e-05, "loss": 0.9838, "step": 8860 }, { "epoch": 0.9, "learning_rate": 5.28276364574755e-05, "loss": 0.7537, "step": 8861 }, { "epoch": 0.9, "learning_rate": 5.282608359085107e-05, "loss": 0.5085, "step": 8862 }, { "epoch": 0.9, "learning_rate": 5.2824530578970124e-05, "loss": 0.4322, "step": 8863 }, { "epoch": 0.9, "learning_rate": 5.2822977421842556e-05, "loss": 1.3102, "step": 8864 }, { "epoch": 0.9, "learning_rate": 5.282142411947823e-05, "loss": 1.3863, "step": 8865 }, { "epoch": 0.9, "learning_rate": 5.2819870671887056e-05, "loss": 0.1738, "step": 8866 }, { "epoch": 0.9, "learning_rate": 5.2818317079078896e-05, "loss": 0.6667, "step": 8867 }, { "epoch": 0.9, "learning_rate": 5.281676334106366e-05, "loss": 0.8854, "step": 8868 }, { "epoch": 0.9, "learning_rate": 5.281520945785121e-05, "loss": 0.8235, "step": 8869 }, { "epoch": 0.9, "learning_rate": 5.2813655429451446e-05, "loss": 0.6613, "step": 8870 }, { "epoch": 0.9, "learning_rate": 5.281210125587426e-05, "loss": 1.0052, "step": 8871 }, { "epoch": 0.9, "learning_rate": 5.281054693712954e-05, "loss": 1.2996, "step": 8872 }, { "epoch": 0.9, "learning_rate": 5.280899247322718e-05, "loss": 0.529, "step": 8873 }, { "epoch": 0.9, "learning_rate": 5.280743786417706e-05, "loss": 0.7576, "step": 8874 }, { "epoch": 0.9, "learning_rate": 5.280588310998909e-05, "loss": 0.5204, "step": 8875 }, { "epoch": 0.9, "learning_rate": 5.2804328210673156e-05, "loss": 0.4918, "step": 8876 }, { "epoch": 0.9, "learning_rate": 5.280277316623915e-05, "loss": 0.382, "step": 8877 }, { "epoch": 0.9, "learning_rate": 5.280121797669697e-05, "loss": 1.4663, "step": 8878 }, { "epoch": 0.9, "learning_rate": 5.2799662642056515e-05, "loss": 0.4323, "step": 8879 }, { "epoch": 0.9, "learning_rate": 5.2798107162327675e-05, "loss": 0.6713, "step": 8880 }, { "epoch": 0.9, "learning_rate": 5.2796551537520354e-05, "loss": 0.7243, "step": 8881 }, { "epoch": 0.9, "learning_rate": 5.279499576764446e-05, "loss": 0.9232, "step": 8882 }, { "epoch": 0.9, "learning_rate": 5.2793439852709886e-05, "loss": 0.7566, "step": 8883 }, { "epoch": 0.9, "learning_rate": 5.279188379272652e-05, "loss": 1.2127, "step": 8884 }, { "epoch": 0.9, "learning_rate": 5.279032758770428e-05, "loss": 1.0054, "step": 8885 }, { "epoch": 0.9, "learning_rate": 5.278877123765306e-05, "loss": 0.3863, "step": 8886 }, { "epoch": 0.9, "learning_rate": 5.278721474258279e-05, "loss": 0.5101, "step": 8887 }, { "epoch": 0.9, "learning_rate": 5.2785658102503337e-05, "loss": 1.6634, "step": 8888 }, { "epoch": 0.9, "learning_rate": 5.278410131742462e-05, "loss": 0.4758, "step": 8889 }, { "epoch": 0.9, "learning_rate": 5.2782544387356554e-05, "loss": 0.7815, "step": 8890 }, { "epoch": 0.9, "learning_rate": 5.278098731230904e-05, "loss": 1.1249, "step": 8891 }, { "epoch": 0.9, "learning_rate": 5.2779430092292e-05, "loss": 1.1297, "step": 8892 }, { "epoch": 0.9, "learning_rate": 5.277787272731533e-05, "loss": 0.967, "step": 8893 }, { "epoch": 0.9, "learning_rate": 5.277631521738894e-05, "loss": 1.0263, "step": 8894 }, { "epoch": 0.9, "learning_rate": 5.2774757562522736e-05, "loss": 1.9381, "step": 8895 }, { "epoch": 0.9, "learning_rate": 5.277319976272665e-05, "loss": 1.3704, "step": 8896 }, { "epoch": 0.9, "learning_rate": 5.277164181801058e-05, "loss": 0.8538, "step": 8897 }, { "epoch": 0.9, "learning_rate": 5.277008372838445e-05, "loss": 0.8346, "step": 8898 }, { "epoch": 0.9, "learning_rate": 5.276852549385816e-05, "loss": 0.8937, "step": 8899 }, { "epoch": 0.9, "learning_rate": 5.276696711444164e-05, "loss": 0.59, "step": 8900 }, { "epoch": 0.9, "learning_rate": 5.2765408590144805e-05, "loss": 1.1348, "step": 8901 }, { "epoch": 0.9, "learning_rate": 5.276384992097756e-05, "loss": 1.6267, "step": 8902 }, { "epoch": 0.9, "learning_rate": 5.2762291106949846e-05, "loss": 1.3771, "step": 8903 }, { "epoch": 0.9, "learning_rate": 5.276073214807157e-05, "loss": 0.8343, "step": 8904 }, { "epoch": 0.9, "learning_rate": 5.275917304435264e-05, "loss": 0.7975, "step": 8905 }, { "epoch": 0.9, "learning_rate": 5.2757613795803006e-05, "loss": 1.0168, "step": 8906 }, { "epoch": 0.9, "learning_rate": 5.2756054402432565e-05, "loss": 0.5149, "step": 8907 }, { "epoch": 0.9, "learning_rate": 5.275449486425126e-05, "loss": 0.6633, "step": 8908 }, { "epoch": 0.9, "learning_rate": 5.2752935181269e-05, "loss": 0.2915, "step": 8909 }, { "epoch": 0.9, "learning_rate": 5.275137535349572e-05, "loss": 1.4685, "step": 8910 }, { "epoch": 0.9, "learning_rate": 5.2749815380941334e-05, "loss": 0.2085, "step": 8911 }, { "epoch": 0.9, "learning_rate": 5.274825526361578e-05, "loss": 1.0505, "step": 8912 }, { "epoch": 0.91, "learning_rate": 5.274669500152899e-05, "loss": 0.3694, "step": 8913 }, { "epoch": 0.91, "learning_rate": 5.274513459469088e-05, "loss": 0.4124, "step": 8914 }, { "epoch": 0.91, "learning_rate": 5.2743574043111385e-05, "loss": 1.3168, "step": 8915 }, { "epoch": 0.91, "learning_rate": 5.274201334680044e-05, "loss": 1.3299, "step": 8916 }, { "epoch": 0.91, "learning_rate": 5.274045250576798e-05, "loss": 0.7681, "step": 8917 }, { "epoch": 0.91, "learning_rate": 5.273889152002392e-05, "loss": 1.3976, "step": 8918 }, { "epoch": 0.91, "learning_rate": 5.273733038957821e-05, "loss": 0.3105, "step": 8919 }, { "epoch": 0.91, "learning_rate": 5.273576911444078e-05, "loss": 0.5565, "step": 8920 }, { "epoch": 0.91, "learning_rate": 5.2734207694621566e-05, "loss": 0.4298, "step": 8921 }, { "epoch": 0.91, "learning_rate": 5.2732646130130495e-05, "loss": 0.8933, "step": 8922 }, { "epoch": 0.91, "learning_rate": 5.273108442097752e-05, "loss": 1.2773, "step": 8923 }, { "epoch": 0.91, "learning_rate": 5.272952256717257e-05, "loss": 1.6664, "step": 8924 }, { "epoch": 0.91, "learning_rate": 5.272796056872558e-05, "loss": 1.5357, "step": 8925 }, { "epoch": 0.91, "learning_rate": 5.2726398425646496e-05, "loss": 0.4457, "step": 8926 }, { "epoch": 0.91, "learning_rate": 5.272483613794526e-05, "loss": 1.2902, "step": 8927 }, { "epoch": 0.91, "learning_rate": 5.272327370563181e-05, "loss": 0.5826, "step": 8928 }, { "epoch": 0.91, "learning_rate": 5.27217111287161e-05, "loss": 1.4559, "step": 8929 }, { "epoch": 0.91, "learning_rate": 5.2720148407208055e-05, "loss": 0.9352, "step": 8930 }, { "epoch": 0.91, "learning_rate": 5.2718585541117626e-05, "loss": 0.3802, "step": 8931 }, { "epoch": 0.91, "learning_rate": 5.2717022530454767e-05, "loss": 0.2862, "step": 8932 }, { "epoch": 0.91, "learning_rate": 5.271545937522941e-05, "loss": 0.8186, "step": 8933 }, { "epoch": 0.91, "learning_rate": 5.271389607545153e-05, "loss": 0.4052, "step": 8934 }, { "epoch": 0.91, "learning_rate": 5.2712332631131047e-05, "loss": 1.3829, "step": 8935 }, { "epoch": 0.91, "learning_rate": 5.271076904227791e-05, "loss": 1.2602, "step": 8936 }, { "epoch": 0.91, "learning_rate": 5.270920530890208e-05, "loss": 0.2728, "step": 8937 }, { "epoch": 0.91, "learning_rate": 5.270764143101352e-05, "loss": 1.5871, "step": 8938 }, { "epoch": 0.91, "learning_rate": 5.2706077408622155e-05, "loss": 0.7477, "step": 8939 }, { "epoch": 0.91, "learning_rate": 5.270451324173796e-05, "loss": 0.6348, "step": 8940 }, { "epoch": 0.91, "learning_rate": 5.270294893037087e-05, "loss": 1.0546, "step": 8941 }, { "epoch": 0.91, "learning_rate": 5.2701384474530855e-05, "loss": 0.9659, "step": 8942 }, { "epoch": 0.91, "learning_rate": 5.269981987422787e-05, "loss": 0.2231, "step": 8943 }, { "epoch": 0.91, "learning_rate": 5.269825512947186e-05, "loss": 0.3356, "step": 8944 }, { "epoch": 0.91, "learning_rate": 5.26966902402728e-05, "loss": 0.8337, "step": 8945 }, { "epoch": 0.91, "learning_rate": 5.2695125206640624e-05, "loss": 1.0444, "step": 8946 }, { "epoch": 0.91, "learning_rate": 5.2693560028585306e-05, "loss": 1.0983, "step": 8947 }, { "epoch": 0.91, "learning_rate": 5.269199470611681e-05, "loss": 0.6675, "step": 8948 }, { "epoch": 0.91, "learning_rate": 5.269042923924509e-05, "loss": 1.2636, "step": 8949 }, { "epoch": 0.91, "learning_rate": 5.268886362798012e-05, "loss": 0.6617, "step": 8950 }, { "epoch": 0.91, "learning_rate": 5.268729787233184e-05, "loss": 0.5473, "step": 8951 }, { "epoch": 0.91, "learning_rate": 5.268573197231023e-05, "loss": 0.7573, "step": 8952 }, { "epoch": 0.91, "learning_rate": 5.268416592792526e-05, "loss": 0.8589, "step": 8953 }, { "epoch": 0.91, "learning_rate": 5.268259973918688e-05, "loss": 1.3842, "step": 8954 }, { "epoch": 0.91, "learning_rate": 5.2681033406105064e-05, "loss": 1.0212, "step": 8955 }, { "epoch": 0.91, "learning_rate": 5.267946692868978e-05, "loss": 1.5402, "step": 8956 }, { "epoch": 0.91, "learning_rate": 5.2677900306951e-05, "loss": 0.8937, "step": 8957 }, { "epoch": 0.91, "learning_rate": 5.2676333540898685e-05, "loss": 0.8236, "step": 8958 }, { "epoch": 0.91, "learning_rate": 5.267476663054282e-05, "loss": 0.8544, "step": 8959 }, { "epoch": 0.91, "learning_rate": 5.267319957589335e-05, "loss": 0.8008, "step": 8960 }, { "epoch": 0.91, "learning_rate": 5.267163237696027e-05, "loss": 0.2774, "step": 8961 }, { "epoch": 0.91, "learning_rate": 5.267006503375355e-05, "loss": 0.5305, "step": 8962 }, { "epoch": 0.91, "learning_rate": 5.2668497546283156e-05, "loss": 0.2403, "step": 8963 }, { "epoch": 0.91, "learning_rate": 5.266692991455907e-05, "loss": 1.0951, "step": 8964 }, { "epoch": 0.91, "learning_rate": 5.266536213859126e-05, "loss": 0.547, "step": 8965 }, { "epoch": 0.91, "learning_rate": 5.266379421838971e-05, "loss": 1.2244, "step": 8966 }, { "epoch": 0.91, "learning_rate": 5.2662226153964396e-05, "loss": 0.3114, "step": 8967 }, { "epoch": 0.91, "learning_rate": 5.26606579453253e-05, "loss": 1.069, "step": 8968 }, { "epoch": 0.91, "learning_rate": 5.26590895924824e-05, "loss": 1.4219, "step": 8969 }, { "epoch": 0.91, "learning_rate": 5.2657521095445664e-05, "loss": 1.433, "step": 8970 }, { "epoch": 0.91, "learning_rate": 5.2655952454225085e-05, "loss": 0.5106, "step": 8971 }, { "epoch": 0.91, "learning_rate": 5.265438366883065e-05, "loss": 0.2291, "step": 8972 }, { "epoch": 0.91, "learning_rate": 5.265281473927233e-05, "loss": 1.3646, "step": 8973 }, { "epoch": 0.91, "learning_rate": 5.2651245665560114e-05, "loss": 1.2571, "step": 8974 }, { "epoch": 0.91, "learning_rate": 5.264967644770399e-05, "loss": 1.7006, "step": 8975 }, { "epoch": 0.91, "learning_rate": 5.2648107085713935e-05, "loss": 1.3921, "step": 8976 }, { "epoch": 0.91, "learning_rate": 5.264653757959995e-05, "loss": 1.2693, "step": 8977 }, { "epoch": 0.91, "learning_rate": 5.2644967929372015e-05, "loss": 1.3703, "step": 8978 }, { "epoch": 0.91, "learning_rate": 5.264339813504012e-05, "loss": 0.9159, "step": 8979 }, { "epoch": 0.91, "learning_rate": 5.264182819661424e-05, "loss": 1.2254, "step": 8980 }, { "epoch": 0.91, "learning_rate": 5.2640258114104394e-05, "loss": 0.2748, "step": 8981 }, { "epoch": 0.91, "learning_rate": 5.263868788752056e-05, "loss": 1.7443, "step": 8982 }, { "epoch": 0.91, "learning_rate": 5.2637117516872716e-05, "loss": 1.1554, "step": 8983 }, { "epoch": 0.91, "learning_rate": 5.2635547002170874e-05, "loss": 0.4669, "step": 8984 }, { "epoch": 0.91, "learning_rate": 5.263397634342502e-05, "loss": 1.3705, "step": 8985 }, { "epoch": 0.91, "learning_rate": 5.2632405540645146e-05, "loss": 1.6229, "step": 8986 }, { "epoch": 0.91, "learning_rate": 5.2630834593841264e-05, "loss": 0.7592, "step": 8987 }, { "epoch": 0.91, "learning_rate": 5.2629263503023357e-05, "loss": 0.5683, "step": 8988 }, { "epoch": 0.91, "learning_rate": 5.262769226820142e-05, "loss": 0.4523, "step": 8989 }, { "epoch": 0.91, "learning_rate": 5.2626120889385464e-05, "loss": 0.8036, "step": 8990 }, { "epoch": 0.91, "learning_rate": 5.2624549366585476e-05, "loss": 0.2639, "step": 8991 }, { "epoch": 0.91, "learning_rate": 5.2622977699811456e-05, "loss": 0.8009, "step": 8992 }, { "epoch": 0.91, "learning_rate": 5.2621405889073426e-05, "loss": 0.3436, "step": 8993 }, { "epoch": 0.91, "learning_rate": 5.2619833934381374e-05, "loss": 0.9218, "step": 8994 }, { "epoch": 0.91, "learning_rate": 5.26182618357453e-05, "loss": 1.4242, "step": 8995 }, { "epoch": 0.91, "learning_rate": 5.261668959317521e-05, "loss": 1.3436, "step": 8996 }, { "epoch": 0.91, "learning_rate": 5.261511720668112e-05, "loss": 1.162, "step": 8997 }, { "epoch": 0.91, "learning_rate": 5.261354467627301e-05, "loss": 0.8268, "step": 8998 }, { "epoch": 0.91, "learning_rate": 5.2611972001960916e-05, "loss": 0.9813, "step": 8999 }, { "epoch": 0.91, "learning_rate": 5.261039918375483e-05, "loss": 1.2191, "step": 9000 }, { "epoch": 0.91, "learning_rate": 5.260882622166477e-05, "loss": 0.4748, "step": 9001 }, { "epoch": 0.91, "learning_rate": 5.260725311570074e-05, "loss": 0.9121, "step": 9002 }, { "epoch": 0.91, "learning_rate": 5.2605679865872755e-05, "loss": 1.4946, "step": 9003 }, { "epoch": 0.91, "learning_rate": 5.260410647219081e-05, "loss": 0.2715, "step": 9004 }, { "epoch": 0.91, "learning_rate": 5.260253293466495e-05, "loss": 1.427, "step": 9005 }, { "epoch": 0.91, "learning_rate": 5.260095925330515e-05, "loss": 0.953, "step": 9006 }, { "epoch": 0.91, "learning_rate": 5.2599385428121455e-05, "loss": 0.724, "step": 9007 }, { "epoch": 0.91, "learning_rate": 5.2597811459123865e-05, "loss": 1.5922, "step": 9008 }, { "epoch": 0.91, "learning_rate": 5.2596237346322395e-05, "loss": 0.4602, "step": 9009 }, { "epoch": 0.91, "learning_rate": 5.2594663089727066e-05, "loss": 1.4475, "step": 9010 }, { "epoch": 0.92, "learning_rate": 5.25930886893479e-05, "loss": 1.77, "step": 9011 }, { "epoch": 0.92, "learning_rate": 5.2591514145194916e-05, "loss": 0.676, "step": 9012 }, { "epoch": 0.92, "learning_rate": 5.258993945727813e-05, "loss": 1.5272, "step": 9013 }, { "epoch": 0.92, "learning_rate": 5.258836462560756e-05, "loss": 1.1446, "step": 9014 }, { "epoch": 0.92, "learning_rate": 5.258678965019322e-05, "loss": 1.2161, "step": 9015 }, { "epoch": 0.92, "learning_rate": 5.258521453104517e-05, "loss": 0.8314, "step": 9016 }, { "epoch": 0.92, "learning_rate": 5.2583639268173385e-05, "loss": 0.3141, "step": 9017 }, { "epoch": 0.92, "learning_rate": 5.258206386158792e-05, "loss": 1.3058, "step": 9018 }, { "epoch": 0.92, "learning_rate": 5.258048831129879e-05, "loss": 0.2445, "step": 9019 }, { "epoch": 0.92, "learning_rate": 5.257891261731602e-05, "loss": 0.968, "step": 9020 }, { "epoch": 0.92, "learning_rate": 5.2577336779649634e-05, "loss": 0.3787, "step": 9021 }, { "epoch": 0.92, "learning_rate": 5.257576079830967e-05, "loss": 1.5701, "step": 9022 }, { "epoch": 0.92, "learning_rate": 5.2574184673306156e-05, "loss": 1.1299, "step": 9023 }, { "epoch": 0.92, "learning_rate": 5.257260840464912e-05, "loss": 1.2656, "step": 9024 }, { "epoch": 0.92, "learning_rate": 5.2571031992348585e-05, "loss": 0.6583, "step": 9025 }, { "epoch": 0.92, "learning_rate": 5.2569455436414585e-05, "loss": 1.8331, "step": 9026 }, { "epoch": 0.92, "learning_rate": 5.2567878736857165e-05, "loss": 0.3843, "step": 9027 }, { "epoch": 0.92, "learning_rate": 5.256630189368635e-05, "loss": 0.9113, "step": 9028 }, { "epoch": 0.92, "learning_rate": 5.2564724906912174e-05, "loss": 1.4699, "step": 9029 }, { "epoch": 0.92, "learning_rate": 5.256314777654467e-05, "loss": 0.6012, "step": 9030 }, { "epoch": 0.92, "learning_rate": 5.256157050259388e-05, "loss": 0.873, "step": 9031 }, { "epoch": 0.92, "learning_rate": 5.2559993085069835e-05, "loss": 1.4037, "step": 9032 }, { "epoch": 0.92, "learning_rate": 5.255841552398258e-05, "loss": 1.11, "step": 9033 }, { "epoch": 0.92, "learning_rate": 5.255683781934214e-05, "loss": 0.8857, "step": 9034 }, { "epoch": 0.92, "learning_rate": 5.2555259971158574e-05, "loss": 1.2527, "step": 9035 }, { "epoch": 0.92, "learning_rate": 5.2553681979441916e-05, "loss": 1.236, "step": 9036 }, { "epoch": 0.92, "learning_rate": 5.2552103844202207e-05, "loss": 0.5017, "step": 9037 }, { "epoch": 0.92, "learning_rate": 5.2550525565449474e-05, "loss": 1.9464, "step": 9038 }, { "epoch": 0.92, "learning_rate": 5.254894714319379e-05, "loss": 1.4055, "step": 9039 }, { "epoch": 0.92, "learning_rate": 5.254736857744518e-05, "loss": 0.249, "step": 9040 }, { "epoch": 0.92, "learning_rate": 5.2545789868213695e-05, "loss": 2.212, "step": 9041 }, { "epoch": 0.92, "learning_rate": 5.254421101550938e-05, "loss": 0.5562, "step": 9042 }, { "epoch": 0.92, "learning_rate": 5.254263201934227e-05, "loss": 0.7717, "step": 9043 }, { "epoch": 0.92, "learning_rate": 5.254105287972244e-05, "loss": 0.2851, "step": 9044 }, { "epoch": 0.92, "learning_rate": 5.253947359665992e-05, "loss": 1.6245, "step": 9045 }, { "epoch": 0.92, "learning_rate": 5.2537894170164766e-05, "loss": 0.6508, "step": 9046 }, { "epoch": 0.92, "learning_rate": 5.2536314600247024e-05, "loss": 1.5624, "step": 9047 }, { "epoch": 0.92, "learning_rate": 5.253473488691676e-05, "loss": 1.203, "step": 9048 }, { "epoch": 0.92, "learning_rate": 5.2533155030184e-05, "loss": 1.5275, "step": 9049 }, { "epoch": 0.92, "learning_rate": 5.253157503005883e-05, "loss": 0.4007, "step": 9050 }, { "epoch": 0.92, "learning_rate": 5.252999488655127e-05, "loss": 0.7301, "step": 9051 }, { "epoch": 0.92, "learning_rate": 5.2528414599671403e-05, "loss": 1.4098, "step": 9052 }, { "epoch": 0.92, "learning_rate": 5.2526834169429276e-05, "loss": 1.3062, "step": 9053 }, { "epoch": 0.92, "learning_rate": 5.252525359583495e-05, "loss": 1.1551, "step": 9054 }, { "epoch": 0.92, "learning_rate": 5.252367287889847e-05, "loss": 1.1522, "step": 9055 }, { "epoch": 0.92, "learning_rate": 5.252209201862991e-05, "loss": 0.2927, "step": 9056 }, { "epoch": 0.92, "learning_rate": 5.2520511015039324e-05, "loss": 0.6505, "step": 9057 }, { "epoch": 0.92, "learning_rate": 5.251892986813677e-05, "loss": 1.1236, "step": 9058 }, { "epoch": 0.92, "learning_rate": 5.251734857793231e-05, "loss": 1.2071, "step": 9059 }, { "epoch": 0.92, "learning_rate": 5.2515767144436024e-05, "loss": 1.4233, "step": 9060 }, { "epoch": 0.92, "learning_rate": 5.2514185567657946e-05, "loss": 0.7891, "step": 9061 }, { "epoch": 0.92, "learning_rate": 5.251260384760816e-05, "loss": 0.5789, "step": 9062 }, { "epoch": 0.92, "learning_rate": 5.2511021984296734e-05, "loss": 1.9225, "step": 9063 }, { "epoch": 0.92, "learning_rate": 5.250943997773372e-05, "loss": 0.6397, "step": 9064 }, { "epoch": 0.92, "learning_rate": 5.250785782792919e-05, "loss": 1.2291, "step": 9065 }, { "epoch": 0.92, "learning_rate": 5.250627553489323e-05, "loss": 0.5315, "step": 9066 }, { "epoch": 0.92, "learning_rate": 5.250469309863588e-05, "loss": 1.2944, "step": 9067 }, { "epoch": 0.92, "learning_rate": 5.250311051916723e-05, "loss": 0.7556, "step": 9068 }, { "epoch": 0.92, "learning_rate": 5.250152779649734e-05, "loss": 0.3066, "step": 9069 }, { "epoch": 0.92, "learning_rate": 5.24999449306363e-05, "loss": 1.4759, "step": 9070 }, { "epoch": 0.92, "learning_rate": 5.249836192159416e-05, "loss": 0.5405, "step": 9071 }, { "epoch": 0.92, "learning_rate": 5.249677876938101e-05, "loss": 1.0194, "step": 9072 }, { "epoch": 0.92, "learning_rate": 5.2495195474006906e-05, "loss": 0.6526, "step": 9073 }, { "epoch": 0.92, "learning_rate": 5.249361203548195e-05, "loss": 1.2072, "step": 9074 }, { "epoch": 0.92, "learning_rate": 5.24920284538162e-05, "loss": 0.7979, "step": 9075 }, { "epoch": 0.92, "learning_rate": 5.249044472901973e-05, "loss": 0.8848, "step": 9076 }, { "epoch": 0.92, "learning_rate": 5.2488860861102634e-05, "loss": 0.4019, "step": 9077 }, { "epoch": 0.92, "learning_rate": 5.248727685007498e-05, "loss": 0.4762, "step": 9078 }, { "epoch": 0.92, "learning_rate": 5.248569269594684e-05, "loss": 0.6421, "step": 9079 }, { "epoch": 0.92, "learning_rate": 5.2484108398728315e-05, "loss": 1.2451, "step": 9080 }, { "epoch": 0.92, "learning_rate": 5.248252395842948e-05, "loss": 0.5422, "step": 9081 }, { "epoch": 0.92, "learning_rate": 5.248093937506042e-05, "loss": 0.9763, "step": 9082 }, { "epoch": 0.92, "learning_rate": 5.2479354648631206e-05, "loss": 0.9476, "step": 9083 }, { "epoch": 0.92, "learning_rate": 5.2477769779151926e-05, "loss": 1.6218, "step": 9084 }, { "epoch": 0.92, "learning_rate": 5.2476184766632675e-05, "loss": 1.0697, "step": 9085 }, { "epoch": 0.92, "learning_rate": 5.2474599611083535e-05, "loss": 1.0921, "step": 9086 }, { "epoch": 0.92, "learning_rate": 5.24730143125146e-05, "loss": 1.1023, "step": 9087 }, { "epoch": 0.92, "learning_rate": 5.2471428870935936e-05, "loss": 1.3467, "step": 9088 }, { "epoch": 0.92, "learning_rate": 5.246984328635766e-05, "loss": 1.1369, "step": 9089 }, { "epoch": 0.92, "learning_rate": 5.246825755878985e-05, "loss": 1.3417, "step": 9090 }, { "epoch": 0.92, "learning_rate": 5.2466671688242585e-05, "loss": 0.4622, "step": 9091 }, { "epoch": 0.92, "learning_rate": 5.246508567472598e-05, "loss": 0.2983, "step": 9092 }, { "epoch": 0.92, "learning_rate": 5.24634995182501e-05, "loss": 1.0136, "step": 9093 }, { "epoch": 0.92, "learning_rate": 5.246191321882507e-05, "loss": 1.678, "step": 9094 }, { "epoch": 0.92, "learning_rate": 5.246032677646096e-05, "loss": 0.3785, "step": 9095 }, { "epoch": 0.92, "learning_rate": 5.2458740191167885e-05, "loss": 0.8939, "step": 9096 }, { "epoch": 0.92, "learning_rate": 5.2457153462955915e-05, "loss": 0.7554, "step": 9097 }, { "epoch": 0.92, "learning_rate": 5.245556659183518e-05, "loss": 0.398, "step": 9098 }, { "epoch": 0.92, "learning_rate": 5.2453979577815755e-05, "loss": 1.487, "step": 9099 }, { "epoch": 0.92, "learning_rate": 5.245239242090775e-05, "loss": 0.2675, "step": 9100 }, { "epoch": 0.92, "learning_rate": 5.2450805121121256e-05, "loss": 0.9447, "step": 9101 }, { "epoch": 0.92, "learning_rate": 5.244921767846637e-05, "loss": 0.4605, "step": 9102 }, { "epoch": 0.92, "learning_rate": 5.244763009295321e-05, "loss": 0.7205, "step": 9103 }, { "epoch": 0.92, "learning_rate": 5.244604236459188e-05, "loss": 0.8534, "step": 9104 }, { "epoch": 0.92, "learning_rate": 5.2444454493392473e-05, "loss": 1.0087, "step": 9105 }, { "epoch": 0.92, "learning_rate": 5.2442866479365096e-05, "loss": 0.6857, "step": 9106 }, { "epoch": 0.92, "learning_rate": 5.2441278322519845e-05, "loss": 0.783, "step": 9107 }, { "epoch": 0.92, "learning_rate": 5.2439690022866845e-05, "loss": 0.2813, "step": 9108 }, { "epoch": 0.92, "learning_rate": 5.2438101580416186e-05, "loss": 1.6909, "step": 9109 }, { "epoch": 0.93, "learning_rate": 5.243651299517799e-05, "loss": 0.5392, "step": 9110 }, { "epoch": 0.93, "learning_rate": 5.243492426716236e-05, "loss": 1.5544, "step": 9111 }, { "epoch": 0.93, "learning_rate": 5.24333353963794e-05, "loss": 1.4295, "step": 9112 }, { "epoch": 0.93, "learning_rate": 5.243174638283924e-05, "loss": 0.9942, "step": 9113 }, { "epoch": 0.93, "learning_rate": 5.243015722655197e-05, "loss": 1.4108, "step": 9114 }, { "epoch": 0.93, "learning_rate": 5.242856792752772e-05, "loss": 1.4451, "step": 9115 }, { "epoch": 0.93, "learning_rate": 5.242697848577659e-05, "loss": 0.5996, "step": 9116 }, { "epoch": 0.93, "learning_rate": 5.2425388901308706e-05, "loss": 0.9739, "step": 9117 }, { "epoch": 0.93, "learning_rate": 5.242379917413418e-05, "loss": 0.7195, "step": 9118 }, { "epoch": 0.93, "learning_rate": 5.242220930426312e-05, "loss": 0.3031, "step": 9119 }, { "epoch": 0.93, "learning_rate": 5.242061929170565e-05, "loss": 1.3052, "step": 9120 }, { "epoch": 0.93, "learning_rate": 5.241902913647189e-05, "loss": 0.5649, "step": 9121 }, { "epoch": 0.93, "learning_rate": 5.241743883857195e-05, "loss": 0.3519, "step": 9122 }, { "epoch": 0.93, "learning_rate": 5.241584839801596e-05, "loss": 0.6406, "step": 9123 }, { "epoch": 0.93, "learning_rate": 5.241425781481405e-05, "loss": 0.6562, "step": 9124 }, { "epoch": 0.93, "learning_rate": 5.241266708897632e-05, "loss": 0.7615, "step": 9125 }, { "epoch": 0.93, "learning_rate": 5.24110762205129e-05, "loss": 0.9924, "step": 9126 }, { "epoch": 0.93, "learning_rate": 5.2409485209433923e-05, "loss": 1.4847, "step": 9127 }, { "epoch": 0.93, "learning_rate": 5.240789405574951e-05, "loss": 0.635, "step": 9128 }, { "epoch": 0.93, "learning_rate": 5.240630275946977e-05, "loss": 0.4786, "step": 9129 }, { "epoch": 0.93, "learning_rate": 5.240471132060486e-05, "loss": 1.6586, "step": 9130 }, { "epoch": 0.93, "learning_rate": 5.240311973916488e-05, "loss": 0.3803, "step": 9131 }, { "epoch": 0.93, "learning_rate": 5.240152801515997e-05, "loss": 1.1271, "step": 9132 }, { "epoch": 0.93, "learning_rate": 5.239993614860026e-05, "loss": 0.8611, "step": 9133 }, { "epoch": 0.93, "learning_rate": 5.2398344139495874e-05, "loss": 0.8747, "step": 9134 }, { "epoch": 0.93, "learning_rate": 5.239675198785695e-05, "loss": 0.8829, "step": 9135 }, { "epoch": 0.93, "learning_rate": 5.2395159693693615e-05, "loss": 0.4235, "step": 9136 }, { "epoch": 0.93, "learning_rate": 5.239356725701601e-05, "loss": 1.4292, "step": 9137 }, { "epoch": 0.93, "learning_rate": 5.239197467783425e-05, "loss": 1.1647, "step": 9138 }, { "epoch": 0.93, "learning_rate": 5.2390381956158486e-05, "loss": 0.9386, "step": 9139 }, { "epoch": 0.93, "learning_rate": 5.238878909199885e-05, "loss": 0.7377, "step": 9140 }, { "epoch": 0.93, "learning_rate": 5.238719608536548e-05, "loss": 0.6963, "step": 9141 }, { "epoch": 0.93, "learning_rate": 5.2385602936268506e-05, "loss": 0.3261, "step": 9142 }, { "epoch": 0.93, "learning_rate": 5.238400964471808e-05, "loss": 0.5432, "step": 9143 }, { "epoch": 0.93, "learning_rate": 5.238241621072432e-05, "loss": 0.9908, "step": 9144 }, { "epoch": 0.93, "learning_rate": 5.238082263429738e-05, "loss": 0.8207, "step": 9145 }, { "epoch": 0.93, "learning_rate": 5.2379228915447403e-05, "loss": 0.5277, "step": 9146 }, { "epoch": 0.93, "learning_rate": 5.237763505418453e-05, "loss": 0.433, "step": 9147 }, { "epoch": 0.93, "learning_rate": 5.2376041050518895e-05, "loss": 1.2384, "step": 9148 }, { "epoch": 0.93, "learning_rate": 5.237444690446064e-05, "loss": 1.418, "step": 9149 }, { "epoch": 0.93, "learning_rate": 5.237285261601993e-05, "loss": 1.3065, "step": 9150 }, { "epoch": 0.93, "learning_rate": 5.237125818520689e-05, "loss": 0.2452, "step": 9151 }, { "epoch": 0.93, "learning_rate": 5.236966361203167e-05, "loss": 0.5139, "step": 9152 }, { "epoch": 0.93, "learning_rate": 5.236806889650443e-05, "loss": 0.2986, "step": 9153 }, { "epoch": 0.93, "learning_rate": 5.236647403863531e-05, "loss": 1.8616, "step": 9154 }, { "epoch": 0.93, "learning_rate": 5.2364879038434445e-05, "loss": 0.848, "step": 9155 }, { "epoch": 0.93, "learning_rate": 5.2363283895912e-05, "loss": 1.462, "step": 9156 }, { "epoch": 0.93, "learning_rate": 5.2361688611078134e-05, "loss": 1.2683, "step": 9157 }, { "epoch": 0.93, "learning_rate": 5.236009318394298e-05, "loss": 0.9099, "step": 9158 }, { "epoch": 0.93, "learning_rate": 5.23584976145167e-05, "loss": 0.7706, "step": 9159 }, { "epoch": 0.93, "learning_rate": 5.235690190280945e-05, "loss": 0.5262, "step": 9160 }, { "epoch": 0.93, "learning_rate": 5.235530604883138e-05, "loss": 1.5169, "step": 9161 }, { "epoch": 0.93, "learning_rate": 5.2353710052592656e-05, "loss": 1.6796, "step": 9162 }, { "epoch": 0.93, "learning_rate": 5.235211391410341e-05, "loss": 0.4546, "step": 9163 }, { "epoch": 0.93, "learning_rate": 5.235051763337382e-05, "loss": 0.178, "step": 9164 }, { "epoch": 0.93, "learning_rate": 5.234892121041404e-05, "loss": 0.7998, "step": 9165 }, { "epoch": 0.93, "learning_rate": 5.2347324645234224e-05, "loss": 0.3213, "step": 9166 }, { "epoch": 0.93, "learning_rate": 5.234572793784454e-05, "loss": 2.0099, "step": 9167 }, { "epoch": 0.93, "learning_rate": 5.2344131088255145e-05, "loss": 1.8673, "step": 9168 }, { "epoch": 0.93, "learning_rate": 5.23425340964762e-05, "loss": 1.3644, "step": 9169 }, { "epoch": 0.93, "learning_rate": 5.234093696251786e-05, "loss": 0.7414, "step": 9170 }, { "epoch": 0.93, "learning_rate": 5.23393396863903e-05, "loss": 1.1463, "step": 9171 }, { "epoch": 0.93, "learning_rate": 5.2337742268103675e-05, "loss": 1.0375, "step": 9172 }, { "epoch": 0.93, "learning_rate": 5.233614470766816e-05, "loss": 1.1955, "step": 9173 }, { "epoch": 0.93, "learning_rate": 5.233454700509392e-05, "loss": 0.4331, "step": 9174 }, { "epoch": 0.93, "learning_rate": 5.233294916039112e-05, "loss": 0.9388, "step": 9175 }, { "epoch": 0.93, "learning_rate": 5.233135117356993e-05, "loss": 1.1269, "step": 9176 }, { "epoch": 0.93, "learning_rate": 5.232975304464051e-05, "loss": 0.7144, "step": 9177 }, { "epoch": 0.93, "learning_rate": 5.232815477361304e-05, "loss": 1.0065, "step": 9178 }, { "epoch": 0.93, "learning_rate": 5.232655636049768e-05, "loss": 1.4296, "step": 9179 }, { "epoch": 0.93, "learning_rate": 5.2324957805304615e-05, "loss": 1.2285, "step": 9180 }, { "epoch": 0.93, "learning_rate": 5.232335910804401e-05, "loss": 0.8864, "step": 9181 }, { "epoch": 0.93, "learning_rate": 5.232176026872604e-05, "loss": 0.4105, "step": 9182 }, { "epoch": 0.93, "learning_rate": 5.232016128736088e-05, "loss": 1.6309, "step": 9183 }, { "epoch": 0.93, "learning_rate": 5.2318562163958704e-05, "loss": 0.9614, "step": 9184 }, { "epoch": 0.93, "learning_rate": 5.2316962898529694e-05, "loss": 1.4208, "step": 9185 }, { "epoch": 0.93, "learning_rate": 5.231536349108402e-05, "loss": 0.5556, "step": 9186 }, { "epoch": 0.93, "learning_rate": 5.231376394163185e-05, "loss": 1.8698, "step": 9187 }, { "epoch": 0.93, "learning_rate": 5.2312164250183384e-05, "loss": 1.2483, "step": 9188 }, { "epoch": 0.93, "learning_rate": 5.23105644167488e-05, "loss": 1.7317, "step": 9189 }, { "epoch": 0.93, "learning_rate": 5.2308964441338255e-05, "loss": 0.3493, "step": 9190 }, { "epoch": 0.93, "learning_rate": 5.230736432396196e-05, "loss": 1.2244, "step": 9191 }, { "epoch": 0.93, "learning_rate": 5.2305764064630076e-05, "loss": 0.7977, "step": 9192 }, { "epoch": 0.93, "learning_rate": 5.23041636633528e-05, "loss": 0.4889, "step": 9193 }, { "epoch": 0.93, "learning_rate": 5.2302563120140316e-05, "loss": 0.4647, "step": 9194 }, { "epoch": 0.93, "learning_rate": 5.23009624350028e-05, "loss": 1.3975, "step": 9195 }, { "epoch": 0.93, "learning_rate": 5.229936160795044e-05, "loss": 0.9286, "step": 9196 }, { "epoch": 0.93, "learning_rate": 5.2297760638993425e-05, "loss": 0.7366, "step": 9197 }, { "epoch": 0.93, "learning_rate": 5.2296159528141946e-05, "loss": 0.499, "step": 9198 }, { "epoch": 0.93, "learning_rate": 5.2294558275406194e-05, "loss": 1.4431, "step": 9199 }, { "epoch": 0.93, "learning_rate": 5.229295688079635e-05, "loss": 1.0149, "step": 9200 }, { "epoch": 0.93, "learning_rate": 5.229135534432261e-05, "loss": 0.5099, "step": 9201 }, { "epoch": 0.93, "learning_rate": 5.2289753665995166e-05, "loss": 0.2506, "step": 9202 }, { "epoch": 0.93, "learning_rate": 5.228815184582421e-05, "loss": 0.7027, "step": 9203 }, { "epoch": 0.93, "learning_rate": 5.228654988381993e-05, "loss": 1.0922, "step": 9204 }, { "epoch": 0.93, "learning_rate": 5.2284947779992526e-05, "loss": 1.4646, "step": 9205 }, { "epoch": 0.93, "learning_rate": 5.228334553435219e-05, "loss": 1.5018, "step": 9206 }, { "epoch": 0.93, "learning_rate": 5.228174314690913e-05, "loss": 0.4234, "step": 9207 }, { "epoch": 0.94, "learning_rate": 5.2280140617673526e-05, "loss": 0.7628, "step": 9208 }, { "epoch": 0.94, "learning_rate": 5.227853794665558e-05, "loss": 0.4332, "step": 9209 }, { "epoch": 0.94, "learning_rate": 5.2276935133865504e-05, "loss": 0.1316, "step": 9210 }, { "epoch": 0.94, "learning_rate": 5.227533217931348e-05, "loss": 0.3806, "step": 9211 }, { "epoch": 0.94, "learning_rate": 5.227372908300971e-05, "loss": 1.3854, "step": 9212 }, { "epoch": 0.94, "learning_rate": 5.227212584496441e-05, "loss": 0.7921, "step": 9213 }, { "epoch": 0.94, "learning_rate": 5.2270522465187764e-05, "loss": 1.1903, "step": 9214 }, { "epoch": 0.94, "learning_rate": 5.226891894369e-05, "loss": 0.5612, "step": 9215 }, { "epoch": 0.94, "learning_rate": 5.22673152804813e-05, "loss": 1.1227, "step": 9216 }, { "epoch": 0.94, "learning_rate": 5.226571147557188e-05, "loss": 1.6045, "step": 9217 }, { "epoch": 0.94, "learning_rate": 5.226410752897193e-05, "loss": 0.9788, "step": 9218 }, { "epoch": 0.94, "learning_rate": 5.2262503440691684e-05, "loss": 0.8055, "step": 9219 }, { "epoch": 0.94, "learning_rate": 5.2260899210741327e-05, "loss": 0.5635, "step": 9220 }, { "epoch": 0.94, "learning_rate": 5.225929483913108e-05, "loss": 0.8346, "step": 9221 }, { "epoch": 0.94, "learning_rate": 5.2257690325871144e-05, "loss": 1.3221, "step": 9222 }, { "epoch": 0.94, "learning_rate": 5.2256085670971737e-05, "loss": 0.6405, "step": 9223 }, { "epoch": 0.94, "learning_rate": 5.225448087444307e-05, "loss": 0.9413, "step": 9224 }, { "epoch": 0.94, "learning_rate": 5.2252875936295353e-05, "loss": 1.4452, "step": 9225 }, { "epoch": 0.94, "learning_rate": 5.2251270856538795e-05, "loss": 1.1516, "step": 9226 }, { "epoch": 0.94, "learning_rate": 5.224966563518362e-05, "loss": 0.9392, "step": 9227 }, { "epoch": 0.94, "learning_rate": 5.224806027224003e-05, "loss": 0.3912, "step": 9228 }, { "epoch": 0.94, "learning_rate": 5.224645476771825e-05, "loss": 1.3754, "step": 9229 }, { "epoch": 0.94, "learning_rate": 5.2244849121628496e-05, "loss": 0.4059, "step": 9230 }, { "epoch": 0.94, "learning_rate": 5.224324333398098e-05, "loss": 0.4721, "step": 9231 }, { "epoch": 0.94, "learning_rate": 5.224163740478593e-05, "loss": 0.2775, "step": 9232 }, { "epoch": 0.94, "learning_rate": 5.224003133405356e-05, "loss": 0.326, "step": 9233 }, { "epoch": 0.94, "learning_rate": 5.2238425121794095e-05, "loss": 1.5397, "step": 9234 }, { "epoch": 0.94, "learning_rate": 5.223681876801774e-05, "loss": 0.9822, "step": 9235 }, { "epoch": 0.94, "learning_rate": 5.223521227273473e-05, "loss": 1.1306, "step": 9236 }, { "epoch": 0.94, "learning_rate": 5.2233605635955306e-05, "loss": 1.4208, "step": 9237 }, { "epoch": 0.94, "learning_rate": 5.2231998857689664e-05, "loss": 0.9126, "step": 9238 }, { "epoch": 0.94, "learning_rate": 5.223039193794803e-05, "loss": 1.2747, "step": 9239 }, { "epoch": 0.94, "learning_rate": 5.2228784876740656e-05, "loss": 0.4148, "step": 9240 }, { "epoch": 0.94, "learning_rate": 5.222717767407774e-05, "loss": 1.1851, "step": 9241 }, { "epoch": 0.94, "learning_rate": 5.2225570329969526e-05, "loss": 0.7391, "step": 9242 }, { "epoch": 0.94, "learning_rate": 5.222396284442623e-05, "loss": 1.4322, "step": 9243 }, { "epoch": 0.94, "learning_rate": 5.222235521745809e-05, "loss": 0.5651, "step": 9244 }, { "epoch": 0.94, "learning_rate": 5.222074744907535e-05, "loss": 1.3433, "step": 9245 }, { "epoch": 0.94, "learning_rate": 5.221913953928821e-05, "loss": 0.9353, "step": 9246 }, { "epoch": 0.94, "learning_rate": 5.221753148810692e-05, "loss": 1.8434, "step": 9247 }, { "epoch": 0.94, "learning_rate": 5.2215923295541714e-05, "loss": 1.5776, "step": 9248 }, { "epoch": 0.94, "learning_rate": 5.221431496160283e-05, "loss": 0.57, "step": 9249 }, { "epoch": 0.94, "learning_rate": 5.221270648630049e-05, "loss": 0.887, "step": 9250 }, { "epoch": 0.94, "learning_rate": 5.2211097869644936e-05, "loss": 1.0261, "step": 9251 }, { "epoch": 0.94, "learning_rate": 5.22094891116464e-05, "loss": 0.706, "step": 9252 }, { "epoch": 0.94, "learning_rate": 5.2207880212315135e-05, "loss": 1.3507, "step": 9253 }, { "epoch": 0.94, "learning_rate": 5.220627117166137e-05, "loss": 0.9499, "step": 9254 }, { "epoch": 0.94, "learning_rate": 5.220466198969534e-05, "loss": 0.9164, "step": 9255 }, { "epoch": 0.94, "learning_rate": 5.220305266642728e-05, "loss": 1.5605, "step": 9256 }, { "epoch": 0.94, "learning_rate": 5.220144320186744e-05, "loss": 0.7523, "step": 9257 }, { "epoch": 0.94, "learning_rate": 5.219983359602607e-05, "loss": 1.183, "step": 9258 }, { "epoch": 0.94, "learning_rate": 5.21982238489134e-05, "loss": 0.5825, "step": 9259 }, { "epoch": 0.94, "learning_rate": 5.219661396053967e-05, "loss": 0.6448, "step": 9260 }, { "epoch": 0.94, "learning_rate": 5.219500393091514e-05, "loss": 1.4241, "step": 9261 }, { "epoch": 0.94, "learning_rate": 5.219339376005005e-05, "loss": 0.8386, "step": 9262 }, { "epoch": 0.94, "learning_rate": 5.2191783447954645e-05, "loss": 1.3191, "step": 9263 }, { "epoch": 0.94, "learning_rate": 5.2190172994639176e-05, "loss": 1.0695, "step": 9264 }, { "epoch": 0.94, "learning_rate": 5.218856240011388e-05, "loss": 0.8367, "step": 9265 }, { "epoch": 0.94, "learning_rate": 5.218695166438902e-05, "loss": 1.1173, "step": 9266 }, { "epoch": 0.94, "learning_rate": 5.2185340787474835e-05, "loss": 1.5143, "step": 9267 }, { "epoch": 0.94, "learning_rate": 5.218372976938158e-05, "loss": 0.7593, "step": 9268 }, { "epoch": 0.94, "learning_rate": 5.218211861011951e-05, "loss": 1.3815, "step": 9269 }, { "epoch": 0.94, "learning_rate": 5.218050730969887e-05, "loss": 0.4972, "step": 9270 }, { "epoch": 0.94, "learning_rate": 5.217889586812994e-05, "loss": 0.5378, "step": 9271 }, { "epoch": 0.94, "learning_rate": 5.217728428542294e-05, "loss": 1.0166, "step": 9272 }, { "epoch": 0.94, "learning_rate": 5.2175672561588136e-05, "loss": 0.2729, "step": 9273 }, { "epoch": 0.94, "learning_rate": 5.2174060696635786e-05, "loss": 1.28, "step": 9274 }, { "epoch": 0.94, "learning_rate": 5.217244869057616e-05, "loss": 0.8253, "step": 9275 }, { "epoch": 0.94, "learning_rate": 5.2170836543419493e-05, "loss": 0.3146, "step": 9276 }, { "epoch": 0.94, "learning_rate": 5.216922425517607e-05, "loss": 0.7488, "step": 9277 }, { "epoch": 0.94, "learning_rate": 5.216761182585612e-05, "loss": 1.0938, "step": 9278 }, { "epoch": 0.94, "learning_rate": 5.216599925546993e-05, "loss": 1.0029, "step": 9279 }, { "epoch": 0.94, "learning_rate": 5.216438654402777e-05, "loss": 1.3227, "step": 9280 }, { "epoch": 0.94, "learning_rate": 5.216277369153986e-05, "loss": 0.9495, "step": 9281 }, { "epoch": 0.94, "learning_rate": 5.2161160698016496e-05, "loss": 0.3403, "step": 9282 }, { "epoch": 0.94, "learning_rate": 5.215954756346795e-05, "loss": 0.7143, "step": 9283 }, { "epoch": 0.94, "learning_rate": 5.215793428790446e-05, "loss": 1.2945, "step": 9284 }, { "epoch": 0.94, "learning_rate": 5.215632087133631e-05, "loss": 0.9278, "step": 9285 }, { "epoch": 0.94, "learning_rate": 5.2154707313773756e-05, "loss": 1.2882, "step": 9286 }, { "epoch": 0.94, "learning_rate": 5.215309361522708e-05, "loss": 0.3137, "step": 9287 }, { "epoch": 0.94, "learning_rate": 5.2151479775706545e-05, "loss": 1.5972, "step": 9288 }, { "epoch": 0.94, "learning_rate": 5.214986579522241e-05, "loss": 0.1135, "step": 9289 }, { "epoch": 0.94, "learning_rate": 5.2148251673784965e-05, "loss": 1.1302, "step": 9290 }, { "epoch": 0.94, "learning_rate": 5.214663741140447e-05, "loss": 1.0133, "step": 9291 }, { "epoch": 0.94, "learning_rate": 5.21450230080912e-05, "loss": 1.1104, "step": 9292 }, { "epoch": 0.94, "learning_rate": 5.214340846385543e-05, "loss": 1.7541, "step": 9293 }, { "epoch": 0.94, "learning_rate": 5.2141793778707424e-05, "loss": 0.7414, "step": 9294 }, { "epoch": 0.94, "learning_rate": 5.214017895265747e-05, "loss": 0.9572, "step": 9295 }, { "epoch": 0.94, "learning_rate": 5.213856398571584e-05, "loss": 0.5402, "step": 9296 }, { "epoch": 0.94, "learning_rate": 5.213694887789281e-05, "loss": 0.6889, "step": 9297 }, { "epoch": 0.94, "learning_rate": 5.213533362919866e-05, "loss": 1.5859, "step": 9298 }, { "epoch": 0.94, "learning_rate": 5.213371823964366e-05, "loss": 0.8493, "step": 9299 }, { "epoch": 0.94, "learning_rate": 5.21321027092381e-05, "loss": 0.7474, "step": 9300 }, { "epoch": 0.94, "learning_rate": 5.213048703799227e-05, "loss": 1.0385, "step": 9301 }, { "epoch": 0.94, "learning_rate": 5.212887122591643e-05, "loss": 0.3738, "step": 9302 }, { "epoch": 0.94, "learning_rate": 5.212725527302088e-05, "loss": 0.8126, "step": 9303 }, { "epoch": 0.94, "learning_rate": 5.212563917931588e-05, "loss": 1.0017, "step": 9304 }, { "epoch": 0.94, "learning_rate": 5.212402294481173e-05, "loss": 0.8813, "step": 9305 }, { "epoch": 0.94, "learning_rate": 5.212240656951873e-05, "loss": 1.5942, "step": 9306 }, { "epoch": 0.95, "learning_rate": 5.212079005344714e-05, "loss": 1.8198, "step": 9307 }, { "epoch": 0.95, "learning_rate": 5.2119173396607253e-05, "loss": 1.4598, "step": 9308 }, { "epoch": 0.95, "learning_rate": 5.211755659900937e-05, "loss": 0.579, "step": 9309 }, { "epoch": 0.95, "learning_rate": 5.211593966066377e-05, "loss": 1.5321, "step": 9310 }, { "epoch": 0.95, "learning_rate": 5.211432258158075e-05, "loss": 0.8542, "step": 9311 }, { "epoch": 0.95, "learning_rate": 5.211270536177058e-05, "loss": 0.492, "step": 9312 }, { "epoch": 0.95, "learning_rate": 5.211108800124357e-05, "loss": 1.1744, "step": 9313 }, { "epoch": 0.95, "learning_rate": 5.2109470500010004e-05, "loss": 0.8243, "step": 9314 }, { "epoch": 0.95, "learning_rate": 5.210785285808018e-05, "loss": 0.9317, "step": 9315 }, { "epoch": 0.95, "learning_rate": 5.210623507546439e-05, "loss": 1.0162, "step": 9316 }, { "epoch": 0.95, "learning_rate": 5.210461715217295e-05, "loss": 0.5932, "step": 9317 }, { "epoch": 0.95, "learning_rate": 5.210299908821611e-05, "loss": 1.2023, "step": 9318 }, { "epoch": 0.95, "learning_rate": 5.21013808836042e-05, "loss": 0.5222, "step": 9319 }, { "epoch": 0.95, "learning_rate": 5.2099762538347505e-05, "loss": 0.5327, "step": 9320 }, { "epoch": 0.95, "learning_rate": 5.209814405245633e-05, "loss": 0.5074, "step": 9321 }, { "epoch": 0.95, "learning_rate": 5.209652542594098e-05, "loss": 0.2668, "step": 9322 }, { "epoch": 0.95, "learning_rate": 5.209490665881175e-05, "loss": 1.1127, "step": 9323 }, { "epoch": 0.95, "learning_rate": 5.209328775107893e-05, "loss": 1.019, "step": 9324 }, { "epoch": 0.95, "learning_rate": 5.209166870275284e-05, "loss": 1.017, "step": 9325 }, { "epoch": 0.95, "learning_rate": 5.209004951384377e-05, "loss": 0.2418, "step": 9326 }, { "epoch": 0.95, "learning_rate": 5.208843018436202e-05, "loss": 0.3299, "step": 9327 }, { "epoch": 0.95, "learning_rate": 5.208681071431791e-05, "loss": 1.3808, "step": 9328 }, { "epoch": 0.95, "learning_rate": 5.208519110372174e-05, "loss": 0.4408, "step": 9329 }, { "epoch": 0.95, "learning_rate": 5.208357135258382e-05, "loss": 1.1576, "step": 9330 }, { "epoch": 0.95, "learning_rate": 5.208195146091445e-05, "loss": 1.3646, "step": 9331 }, { "epoch": 0.95, "learning_rate": 5.208033142872394e-05, "loss": 1.0283, "step": 9332 }, { "epoch": 0.95, "learning_rate": 5.2078711256022596e-05, "loss": 1.3717, "step": 9333 }, { "epoch": 0.95, "learning_rate": 5.207709094282074e-05, "loss": 0.6106, "step": 9334 }, { "epoch": 0.95, "learning_rate": 5.207547048912867e-05, "loss": 0.2562, "step": 9335 }, { "epoch": 0.95, "learning_rate": 5.2073849894956706e-05, "loss": 1.3189, "step": 9336 }, { "epoch": 0.95, "learning_rate": 5.2072229160315167e-05, "loss": 1.072, "step": 9337 }, { "epoch": 0.95, "learning_rate": 5.207060828521435e-05, "loss": 0.6247, "step": 9338 }, { "epoch": 0.95, "learning_rate": 5.206898726966458e-05, "loss": 0.2452, "step": 9339 }, { "epoch": 0.95, "learning_rate": 5.206736611367617e-05, "loss": 0.1633, "step": 9340 }, { "epoch": 0.95, "learning_rate": 5.206574481725943e-05, "loss": 0.7155, "step": 9341 }, { "epoch": 0.95, "learning_rate": 5.206412338042469e-05, "loss": 1.0541, "step": 9342 }, { "epoch": 0.95, "learning_rate": 5.206250180318227e-05, "loss": 1.4499, "step": 9343 }, { "epoch": 0.95, "learning_rate": 5.206088008554247e-05, "loss": 0.6212, "step": 9344 }, { "epoch": 0.95, "learning_rate": 5.205925822751563e-05, "loss": 1.2714, "step": 9345 }, { "epoch": 0.95, "learning_rate": 5.205763622911205e-05, "loss": 0.2695, "step": 9346 }, { "epoch": 0.95, "learning_rate": 5.205601409034208e-05, "loss": 1.2524, "step": 9347 }, { "epoch": 0.95, "learning_rate": 5.205439181121601e-05, "loss": 1.3333, "step": 9348 }, { "epoch": 0.95, "learning_rate": 5.205276939174419e-05, "loss": 0.7565, "step": 9349 }, { "epoch": 0.95, "learning_rate": 5.2051146831936936e-05, "loss": 1.1277, "step": 9350 }, { "epoch": 0.95, "learning_rate": 5.204952413180458e-05, "loss": 0.821, "step": 9351 }, { "epoch": 0.95, "learning_rate": 5.204790129135742e-05, "loss": 0.5083, "step": 9352 }, { "epoch": 0.95, "learning_rate": 5.204627831060582e-05, "loss": 1.6287, "step": 9353 }, { "epoch": 0.95, "learning_rate": 5.204465518956008e-05, "loss": 0.765, "step": 9354 }, { "epoch": 0.95, "learning_rate": 5.204303192823055e-05, "loss": 1.8056, "step": 9355 }, { "epoch": 0.95, "learning_rate": 5.2041408526627544e-05, "loss": 1.1256, "step": 9356 }, { "epoch": 0.95, "learning_rate": 5.203978498476141e-05, "loss": 1.0763, "step": 9357 }, { "epoch": 0.95, "learning_rate": 5.203816130264245e-05, "loss": 1.5271, "step": 9358 }, { "epoch": 0.95, "learning_rate": 5.2036537480281034e-05, "loss": 0.8029, "step": 9359 }, { "epoch": 0.95, "learning_rate": 5.203491351768747e-05, "loss": 0.7394, "step": 9360 }, { "epoch": 0.95, "learning_rate": 5.20332894148721e-05, "loss": 0.5542, "step": 9361 }, { "epoch": 0.95, "learning_rate": 5.203166517184525e-05, "loss": 0.4019, "step": 9362 }, { "epoch": 0.95, "learning_rate": 5.203004078861728e-05, "loss": 1.2052, "step": 9363 }, { "epoch": 0.95, "learning_rate": 5.2028416265198496e-05, "loss": 0.9673, "step": 9364 }, { "epoch": 0.95, "learning_rate": 5.202679160159926e-05, "loss": 0.8838, "step": 9365 }, { "epoch": 0.95, "learning_rate": 5.202516679782989e-05, "loss": 1.0275, "step": 9366 }, { "epoch": 0.95, "learning_rate": 5.202354185390075e-05, "loss": 1.2125, "step": 9367 }, { "epoch": 0.95, "learning_rate": 5.202191676982216e-05, "loss": 0.679, "step": 9368 }, { "epoch": 0.95, "learning_rate": 5.202029154560447e-05, "loss": 1.4272, "step": 9369 }, { "epoch": 0.95, "learning_rate": 5.201866618125803e-05, "loss": 0.8241, "step": 9370 }, { "epoch": 0.95, "learning_rate": 5.2017040676793165e-05, "loss": 0.8922, "step": 9371 }, { "epoch": 0.95, "learning_rate": 5.201541503222023e-05, "loss": 1.301, "step": 9372 }, { "epoch": 0.95, "learning_rate": 5.2013789247549577e-05, "loss": 0.9078, "step": 9373 }, { "epoch": 0.95, "learning_rate": 5.201216332279153e-05, "loss": 1.1279, "step": 9374 }, { "epoch": 0.95, "learning_rate": 5.2010537257956465e-05, "loss": 0.4406, "step": 9375 }, { "epoch": 0.95, "learning_rate": 5.200891105305471e-05, "loss": 1.4764, "step": 9376 }, { "epoch": 0.95, "learning_rate": 5.2007284708096615e-05, "loss": 0.9131, "step": 9377 }, { "epoch": 0.95, "learning_rate": 5.200565822309253e-05, "loss": 1.1854, "step": 9378 }, { "epoch": 0.95, "learning_rate": 5.200403159805281e-05, "loss": 0.5784, "step": 9379 }, { "epoch": 0.95, "learning_rate": 5.20024048329878e-05, "loss": 0.4045, "step": 9380 }, { "epoch": 0.95, "learning_rate": 5.200077792790785e-05, "loss": 0.6539, "step": 9381 }, { "epoch": 0.95, "learning_rate": 5.199915088282334e-05, "loss": 1.2983, "step": 9382 }, { "epoch": 0.95, "learning_rate": 5.199752369774458e-05, "loss": 1.6962, "step": 9383 }, { "epoch": 0.95, "learning_rate": 5.199589637268196e-05, "loss": 0.7831, "step": 9384 }, { "epoch": 0.95, "learning_rate": 5.199426890764582e-05, "loss": 0.3353, "step": 9385 }, { "epoch": 0.95, "learning_rate": 5.1992641302646525e-05, "loss": 1.5914, "step": 9386 }, { "epoch": 0.95, "learning_rate": 5.199101355769442e-05, "loss": 0.8058, "step": 9387 }, { "epoch": 0.95, "learning_rate": 5.198938567279987e-05, "loss": 0.7855, "step": 9388 }, { "epoch": 0.95, "learning_rate": 5.198775764797324e-05, "loss": 0.5302, "step": 9389 }, { "epoch": 0.95, "learning_rate": 5.198612948322488e-05, "loss": 1.3224, "step": 9390 }, { "epoch": 0.95, "learning_rate": 5.198450117856516e-05, "loss": 0.5286, "step": 9391 }, { "epoch": 0.95, "learning_rate": 5.1982872734004434e-05, "loss": 0.9811, "step": 9392 }, { "epoch": 0.95, "learning_rate": 5.1981244149553076e-05, "loss": 0.8411, "step": 9393 }, { "epoch": 0.95, "learning_rate": 5.1979615425221436e-05, "loss": 0.6797, "step": 9394 }, { "epoch": 0.95, "learning_rate": 5.197798656101988e-05, "loss": 0.4352, "step": 9395 }, { "epoch": 0.95, "learning_rate": 5.1976357556958784e-05, "loss": 1.4353, "step": 9396 }, { "epoch": 0.95, "learning_rate": 5.1974728413048514e-05, "loss": 0.3998, "step": 9397 }, { "epoch": 0.95, "learning_rate": 5.1973099129299426e-05, "loss": 1.4069, "step": 9398 }, { "epoch": 0.95, "learning_rate": 5.1971469705721896e-05, "loss": 0.3133, "step": 9399 }, { "epoch": 0.95, "learning_rate": 5.19698401423263e-05, "loss": 0.4594, "step": 9400 }, { "epoch": 0.95, "learning_rate": 5.1968210439122983e-05, "loss": 0.8068, "step": 9401 }, { "epoch": 0.95, "learning_rate": 5.196658059612233e-05, "loss": 1.0244, "step": 9402 }, { "epoch": 0.95, "learning_rate": 5.196495061333473e-05, "loss": 0.6425, "step": 9403 }, { "epoch": 0.95, "learning_rate": 5.196332049077053e-05, "loss": 0.5793, "step": 9404 }, { "epoch": 0.96, "learning_rate": 5.196169022844012e-05, "loss": 0.167, "step": 9405 }, { "epoch": 0.96, "learning_rate": 5.196005982635386e-05, "loss": 0.8849, "step": 9406 }, { "epoch": 0.96, "learning_rate": 5.195842928452214e-05, "loss": 0.3344, "step": 9407 }, { "epoch": 0.96, "learning_rate": 5.195679860295533e-05, "loss": 0.6725, "step": 9408 }, { "epoch": 0.96, "learning_rate": 5.1955167781663806e-05, "loss": 1.0257, "step": 9409 }, { "epoch": 0.96, "learning_rate": 5.195353682065794e-05, "loss": 1.1112, "step": 9410 }, { "epoch": 0.96, "learning_rate": 5.195190571994812e-05, "loss": 1.0032, "step": 9411 }, { "epoch": 0.96, "learning_rate": 5.195027447954473e-05, "loss": 0.5755, "step": 9412 }, { "epoch": 0.96, "learning_rate": 5.194864309945814e-05, "loss": 0.8744, "step": 9413 }, { "epoch": 0.96, "learning_rate": 5.194701157969874e-05, "loss": 0.9723, "step": 9414 }, { "epoch": 0.96, "learning_rate": 5.19453799202769e-05, "loss": 0.3438, "step": 9415 }, { "epoch": 0.96, "learning_rate": 5.1943748121203e-05, "loss": 1.2005, "step": 9416 }, { "epoch": 0.96, "learning_rate": 5.194211618248745e-05, "loss": 0.7818, "step": 9417 }, { "epoch": 0.96, "learning_rate": 5.194048410414061e-05, "loss": 0.5942, "step": 9418 }, { "epoch": 0.96, "learning_rate": 5.193885188617289e-05, "loss": 1.8797, "step": 9419 }, { "epoch": 0.96, "learning_rate": 5.193721952859465e-05, "loss": 0.7385, "step": 9420 }, { "epoch": 0.96, "learning_rate": 5.193558703141629e-05, "loss": 1.1164, "step": 9421 }, { "epoch": 0.96, "learning_rate": 5.193395439464821e-05, "loss": 1.2405, "step": 9422 }, { "epoch": 0.96, "learning_rate": 5.193232161830078e-05, "loss": 0.859, "step": 9423 }, { "epoch": 0.96, "learning_rate": 5.19306887023844e-05, "loss": 0.9709, "step": 9424 }, { "epoch": 0.96, "learning_rate": 5.192905564690946e-05, "loss": 0.634, "step": 9425 }, { "epoch": 0.96, "learning_rate": 5.192742245188636e-05, "loss": 0.2272, "step": 9426 }, { "epoch": 0.96, "learning_rate": 5.192578911732547e-05, "loss": 0.8561, "step": 9427 }, { "epoch": 0.96, "learning_rate": 5.19241556432372e-05, "loss": 1.5499, "step": 9428 }, { "epoch": 0.96, "learning_rate": 5.192252202963196e-05, "loss": 0.2306, "step": 9429 }, { "epoch": 0.96, "learning_rate": 5.192088827652012e-05, "loss": 1.2188, "step": 9430 }, { "epoch": 0.96, "learning_rate": 5.191925438391209e-05, "loss": 1.0327, "step": 9431 }, { "epoch": 0.96, "learning_rate": 5.191762035181825e-05, "loss": 1.2183, "step": 9432 }, { "epoch": 0.96, "learning_rate": 5.191598618024903e-05, "loss": 1.3315, "step": 9433 }, { "epoch": 0.96, "learning_rate": 5.191435186921481e-05, "loss": 0.9823, "step": 9434 }, { "epoch": 0.96, "learning_rate": 5.1912717418725975e-05, "loss": 1.2583, "step": 9435 }, { "epoch": 0.96, "learning_rate": 5.191108282879295e-05, "loss": 0.7431, "step": 9436 }, { "epoch": 0.96, "learning_rate": 5.190944809942613e-05, "loss": 0.637, "step": 9437 }, { "epoch": 0.96, "learning_rate": 5.190781323063592e-05, "loss": 0.8823, "step": 9438 }, { "epoch": 0.96, "learning_rate": 5.1906178222432716e-05, "loss": 1.3934, "step": 9439 }, { "epoch": 0.96, "learning_rate": 5.190454307482693e-05, "loss": 0.3935, "step": 9440 }, { "epoch": 0.96, "learning_rate": 5.190290778782897e-05, "loss": 0.8117, "step": 9441 }, { "epoch": 0.96, "learning_rate": 5.190127236144923e-05, "loss": 1.1562, "step": 9442 }, { "epoch": 0.96, "learning_rate": 5.189963679569813e-05, "loss": 1.6289, "step": 9443 }, { "epoch": 0.96, "learning_rate": 5.189800109058606e-05, "loss": 1.2711, "step": 9444 }, { "epoch": 0.96, "learning_rate": 5.189636524612345e-05, "loss": 1.6769, "step": 9445 }, { "epoch": 0.96, "learning_rate": 5.1894729262320705e-05, "loss": 0.72, "step": 9446 }, { "epoch": 0.96, "learning_rate": 5.1893093139188235e-05, "loss": 0.77, "step": 9447 }, { "epoch": 0.96, "learning_rate": 5.189145687673644e-05, "loss": 0.5718, "step": 9448 }, { "epoch": 0.96, "learning_rate": 5.1889820474975745e-05, "loss": 1.1961, "step": 9449 }, { "epoch": 0.96, "learning_rate": 5.1888183933916555e-05, "loss": 1.1234, "step": 9450 }, { "epoch": 0.96, "learning_rate": 5.188654725356929e-05, "loss": 1.3516, "step": 9451 }, { "epoch": 0.96, "learning_rate": 5.1884910433944373e-05, "loss": 0.815, "step": 9452 }, { "epoch": 0.96, "learning_rate": 5.1883273475052206e-05, "loss": 0.9356, "step": 9453 }, { "epoch": 0.96, "learning_rate": 5.188163637690321e-05, "loss": 0.1239, "step": 9454 }, { "epoch": 0.96, "learning_rate": 5.1879999139507806e-05, "loss": 1.1321, "step": 9455 }, { "epoch": 0.96, "learning_rate": 5.187836176287641e-05, "loss": 0.5891, "step": 9456 }, { "epoch": 0.96, "learning_rate": 5.187672424701944e-05, "loss": 0.4845, "step": 9457 }, { "epoch": 0.96, "learning_rate": 5.187508659194733e-05, "loss": 0.5946, "step": 9458 }, { "epoch": 0.96, "learning_rate": 5.187344879767048e-05, "loss": 1.3222, "step": 9459 }, { "epoch": 0.96, "learning_rate": 5.187181086419933e-05, "loss": 0.4406, "step": 9460 }, { "epoch": 0.96, "learning_rate": 5.1870172791544293e-05, "loss": 1.5507, "step": 9461 }, { "epoch": 0.96, "learning_rate": 5.186853457971579e-05, "loss": 0.6741, "step": 9462 }, { "epoch": 0.96, "learning_rate": 5.186689622872426e-05, "loss": 1.6075, "step": 9463 }, { "epoch": 0.96, "learning_rate": 5.186525773858012e-05, "loss": 0.4757, "step": 9464 }, { "epoch": 0.96, "learning_rate": 5.186361910929379e-05, "loss": 1.44, "step": 9465 }, { "epoch": 0.96, "learning_rate": 5.1861980340875715e-05, "loss": 0.4532, "step": 9466 }, { "epoch": 0.96, "learning_rate": 5.186034143333631e-05, "loss": 0.257, "step": 9467 }, { "epoch": 0.96, "learning_rate": 5.185870238668601e-05, "loss": 1.1001, "step": 9468 }, { "epoch": 0.96, "learning_rate": 5.1857063200935244e-05, "loss": 1.3327, "step": 9469 }, { "epoch": 0.96, "learning_rate": 5.185542387609444e-05, "loss": 0.3364, "step": 9470 }, { "epoch": 0.96, "learning_rate": 5.1853784412174036e-05, "loss": 1.1547, "step": 9471 }, { "epoch": 0.96, "learning_rate": 5.185214480918446e-05, "loss": 0.4335, "step": 9472 }, { "epoch": 0.96, "learning_rate": 5.1850505067136144e-05, "loss": 1.2574, "step": 9473 }, { "epoch": 0.96, "learning_rate": 5.1848865186039527e-05, "loss": 1.3169, "step": 9474 }, { "epoch": 0.96, "learning_rate": 5.184722516590504e-05, "loss": 0.3332, "step": 9475 }, { "epoch": 0.96, "learning_rate": 5.184558500674313e-05, "loss": 0.2686, "step": 9476 }, { "epoch": 0.96, "learning_rate": 5.184394470856423e-05, "loss": 1.2821, "step": 9477 }, { "epoch": 0.96, "learning_rate": 5.184230427137877e-05, "loss": 0.5542, "step": 9478 }, { "epoch": 0.96, "learning_rate": 5.1840663695197194e-05, "loss": 1.2793, "step": 9479 }, { "epoch": 0.96, "learning_rate": 5.1839022980029944e-05, "loss": 1.2526, "step": 9480 }, { "epoch": 0.96, "learning_rate": 5.183738212588747e-05, "loss": 0.5104, "step": 9481 }, { "epoch": 0.96, "learning_rate": 5.183574113278019e-05, "loss": 0.9514, "step": 9482 }, { "epoch": 0.96, "learning_rate": 5.183410000071857e-05, "loss": 0.9115, "step": 9483 }, { "epoch": 0.96, "learning_rate": 5.183245872971304e-05, "loss": 0.9116, "step": 9484 }, { "epoch": 0.96, "learning_rate": 5.1830817319774045e-05, "loss": 1.3409, "step": 9485 }, { "epoch": 0.96, "learning_rate": 5.182917577091203e-05, "loss": 0.9508, "step": 9486 }, { "epoch": 0.96, "learning_rate": 5.1827534083137454e-05, "loss": 1.7372, "step": 9487 }, { "epoch": 0.96, "learning_rate": 5.1825892256460746e-05, "loss": 0.8797, "step": 9488 }, { "epoch": 0.96, "learning_rate": 5.1824250290892366e-05, "loss": 0.2964, "step": 9489 }, { "epoch": 0.96, "learning_rate": 5.182260818644276e-05, "loss": 0.4564, "step": 9490 }, { "epoch": 0.96, "learning_rate": 5.182096594312238e-05, "loss": 0.4207, "step": 9491 }, { "epoch": 0.96, "learning_rate": 5.181932356094167e-05, "loss": 1.5769, "step": 9492 }, { "epoch": 0.96, "learning_rate": 5.181768103991109e-05, "loss": 0.2515, "step": 9493 }, { "epoch": 0.96, "learning_rate": 5.181603838004109e-05, "loss": 1.2359, "step": 9494 }, { "epoch": 0.96, "learning_rate": 5.1814395581342106e-05, "loss": 0.7672, "step": 9495 }, { "epoch": 0.96, "learning_rate": 5.181275264382462e-05, "loss": 0.2322, "step": 9496 }, { "epoch": 0.96, "learning_rate": 5.181110956749908e-05, "loss": 1.5011, "step": 9497 }, { "epoch": 0.96, "learning_rate": 5.180946635237592e-05, "loss": 0.4165, "step": 9498 }, { "epoch": 0.96, "learning_rate": 5.180782299846562e-05, "loss": 0.7958, "step": 9499 }, { "epoch": 0.96, "learning_rate": 5.1806179505778634e-05, "loss": 0.9906, "step": 9500 }, { "epoch": 0.96, "learning_rate": 5.180453587432541e-05, "loss": 0.363, "step": 9501 }, { "epoch": 0.96, "learning_rate": 5.180289210411642e-05, "loss": 0.5523, "step": 9502 }, { "epoch": 0.96, "learning_rate": 5.180124819516212e-05, "loss": 0.3675, "step": 9503 }, { "epoch": 0.97, "learning_rate": 5.179960414747296e-05, "loss": 1.8386, "step": 9504 }, { "epoch": 0.97, "learning_rate": 5.179795996105942e-05, "loss": 1.118, "step": 9505 }, { "epoch": 0.97, "learning_rate": 5.179631563593196e-05, "loss": 0.7019, "step": 9506 }, { "epoch": 0.97, "learning_rate": 5.179467117210103e-05, "loss": 1.457, "step": 9507 }, { "epoch": 0.97, "learning_rate": 5.179302656957711e-05, "loss": 1.6583, "step": 9508 }, { "epoch": 0.97, "learning_rate": 5.179138182837066e-05, "loss": 0.251, "step": 9509 }, { "epoch": 0.97, "learning_rate": 5.178973694849214e-05, "loss": 1.1674, "step": 9510 }, { "epoch": 0.97, "learning_rate": 5.1788091929952026e-05, "loss": 0.3161, "step": 9511 }, { "epoch": 0.97, "learning_rate": 5.1786446772760784e-05, "loss": 1.4493, "step": 9512 }, { "epoch": 0.97, "learning_rate": 5.178480147692888e-05, "loss": 0.825, "step": 9513 }, { "epoch": 0.97, "learning_rate": 5.178315604246679e-05, "loss": 0.6046, "step": 9514 }, { "epoch": 0.97, "learning_rate": 5.178151046938498e-05, "loss": 0.7481, "step": 9515 }, { "epoch": 0.97, "learning_rate": 5.1779864757693926e-05, "loss": 1.2064, "step": 9516 }, { "epoch": 0.97, "learning_rate": 5.17782189074041e-05, "loss": 0.6017, "step": 9517 }, { "epoch": 0.97, "learning_rate": 5.177657291852597e-05, "loss": 0.89, "step": 9518 }, { "epoch": 0.97, "learning_rate": 5.177492679107001e-05, "loss": 0.5963, "step": 9519 }, { "epoch": 0.97, "learning_rate": 5.177328052504671e-05, "loss": 1.0645, "step": 9520 }, { "epoch": 0.97, "learning_rate": 5.177163412046653e-05, "loss": 1.7362, "step": 9521 }, { "epoch": 0.97, "learning_rate": 5.1769987577339946e-05, "loss": 1.2048, "step": 9522 }, { "epoch": 0.97, "learning_rate": 5.1768340895677446e-05, "loss": 1.4417, "step": 9523 }, { "epoch": 0.97, "learning_rate": 5.176669407548951e-05, "loss": 0.5712, "step": 9524 }, { "epoch": 0.97, "learning_rate": 5.176504711678662e-05, "loss": 0.6358, "step": 9525 }, { "epoch": 0.97, "learning_rate": 5.176340001957924e-05, "loss": 0.9451, "step": 9526 }, { "epoch": 0.97, "learning_rate": 5.1761752783877864e-05, "loss": 0.5711, "step": 9527 }, { "epoch": 0.97, "learning_rate": 5.176010540969298e-05, "loss": 0.212, "step": 9528 }, { "epoch": 0.97, "learning_rate": 5.175845789703505e-05, "loss": 1.4155, "step": 9529 }, { "epoch": 0.97, "learning_rate": 5.175681024591457e-05, "loss": 1.3472, "step": 9530 }, { "epoch": 0.97, "learning_rate": 5.1755162456342034e-05, "loss": 0.2936, "step": 9531 }, { "epoch": 0.97, "learning_rate": 5.175351452832792e-05, "loss": 0.9938, "step": 9532 }, { "epoch": 0.97, "learning_rate": 5.175186646188272e-05, "loss": 0.5665, "step": 9533 }, { "epoch": 0.97, "learning_rate": 5.17502182570169e-05, "loss": 0.3097, "step": 9534 }, { "epoch": 0.97, "learning_rate": 5.1748569913740976e-05, "loss": 0.9427, "step": 9535 }, { "epoch": 0.97, "learning_rate": 5.174692143206542e-05, "loss": 1.2111, "step": 9536 }, { "epoch": 0.97, "learning_rate": 5.1745272812000746e-05, "loss": 0.5763, "step": 9537 }, { "epoch": 0.97, "learning_rate": 5.1743624053557415e-05, "loss": 0.4022, "step": 9538 }, { "epoch": 0.97, "learning_rate": 5.174197515674594e-05, "loss": 0.4813, "step": 9539 }, { "epoch": 0.97, "learning_rate": 5.17403261215768e-05, "loss": 0.8326, "step": 9540 }, { "epoch": 0.97, "learning_rate": 5.173867694806049e-05, "loss": 0.4002, "step": 9541 }, { "epoch": 0.97, "learning_rate": 5.173702763620752e-05, "loss": 1.1904, "step": 9542 }, { "epoch": 0.97, "learning_rate": 5.173537818602837e-05, "loss": 0.8321, "step": 9543 }, { "epoch": 0.97, "learning_rate": 5.1733728597533544e-05, "loss": 1.5736, "step": 9544 }, { "epoch": 0.97, "learning_rate": 5.173207887073353e-05, "loss": 0.2619, "step": 9545 }, { "epoch": 0.97, "learning_rate": 5.173042900563885e-05, "loss": 1.0467, "step": 9546 }, { "epoch": 0.97, "learning_rate": 5.172877900225997e-05, "loss": 1.067, "step": 9547 }, { "epoch": 0.97, "learning_rate": 5.1727128860607414e-05, "loss": 0.7854, "step": 9548 }, { "epoch": 0.97, "learning_rate": 5.172547858069167e-05, "loss": 1.2759, "step": 9549 }, { "epoch": 0.97, "learning_rate": 5.1723828162523245e-05, "loss": 0.3659, "step": 9550 }, { "epoch": 0.97, "learning_rate": 5.172217760611265e-05, "loss": 0.7365, "step": 9551 }, { "epoch": 0.97, "learning_rate": 5.172052691147037e-05, "loss": 0.3654, "step": 9552 }, { "epoch": 0.97, "learning_rate": 5.1718876078606926e-05, "loss": 1.2956, "step": 9553 }, { "epoch": 0.97, "learning_rate": 5.171722510753281e-05, "loss": 0.7628, "step": 9554 }, { "epoch": 0.97, "learning_rate": 5.171557399825855e-05, "loss": 0.8533, "step": 9555 }, { "epoch": 0.97, "learning_rate": 5.171392275079463e-05, "loss": 0.5982, "step": 9556 }, { "epoch": 0.97, "learning_rate": 5.1712271365151564e-05, "loss": 1.5803, "step": 9557 }, { "epoch": 0.97, "learning_rate": 5.171061984133986e-05, "loss": 0.8061, "step": 9558 }, { "epoch": 0.97, "learning_rate": 5.1708968179370036e-05, "loss": 0.5928, "step": 9559 }, { "epoch": 0.97, "learning_rate": 5.17073163792526e-05, "loss": 0.8754, "step": 9560 }, { "epoch": 0.97, "learning_rate": 5.170566444099805e-05, "loss": 1.8494, "step": 9561 }, { "epoch": 0.97, "learning_rate": 5.170401236461692e-05, "loss": 0.9685, "step": 9562 }, { "epoch": 0.97, "learning_rate": 5.170236015011971e-05, "loss": 0.8769, "step": 9563 }, { "epoch": 0.97, "learning_rate": 5.170070779751693e-05, "loss": 1.3737, "step": 9564 }, { "epoch": 0.97, "learning_rate": 5.16990553068191e-05, "loss": 0.8722, "step": 9565 }, { "epoch": 0.97, "learning_rate": 5.1697402678036745e-05, "loss": 1.6677, "step": 9566 }, { "epoch": 0.97, "learning_rate": 5.169574991118037e-05, "loss": 0.5448, "step": 9567 }, { "epoch": 0.97, "learning_rate": 5.169409700626049e-05, "loss": 0.8019, "step": 9568 }, { "epoch": 0.97, "learning_rate": 5.1692443963287634e-05, "loss": 0.6113, "step": 9569 }, { "epoch": 0.97, "learning_rate": 5.1690790782272314e-05, "loss": 0.4637, "step": 9570 }, { "epoch": 0.97, "learning_rate": 5.168913746322506e-05, "loss": 1.1905, "step": 9571 }, { "epoch": 0.97, "learning_rate": 5.1687484006156383e-05, "loss": 0.6645, "step": 9572 }, { "epoch": 0.97, "learning_rate": 5.168583041107681e-05, "loss": 0.7581, "step": 9573 }, { "epoch": 0.97, "learning_rate": 5.168417667799686e-05, "loss": 1.0021, "step": 9574 }, { "epoch": 0.97, "learning_rate": 5.1682522806927054e-05, "loss": 1.1021, "step": 9575 }, { "epoch": 0.97, "learning_rate": 5.1680868797877925e-05, "loss": 1.5758, "step": 9576 }, { "epoch": 0.97, "learning_rate": 5.167921465086e-05, "loss": 1.2901, "step": 9577 }, { "epoch": 0.97, "learning_rate": 5.1677560365883795e-05, "loss": 0.9724, "step": 9578 }, { "epoch": 0.97, "learning_rate": 5.167590594295984e-05, "loss": 1.199, "step": 9579 }, { "epoch": 0.97, "learning_rate": 5.167425138209867e-05, "loss": 0.7664, "step": 9580 }, { "epoch": 0.97, "learning_rate": 5.1672596683310816e-05, "loss": 1.0848, "step": 9581 }, { "epoch": 0.97, "learning_rate": 5.167094184660679e-05, "loss": 1.3555, "step": 9582 }, { "epoch": 0.97, "learning_rate": 5.1669286871997134e-05, "loss": 0.6705, "step": 9583 }, { "epoch": 0.97, "learning_rate": 5.166763175949238e-05, "loss": 0.6672, "step": 9584 }, { "epoch": 0.97, "learning_rate": 5.1665976509103065e-05, "loss": 0.4661, "step": 9585 }, { "epoch": 0.97, "learning_rate": 5.1664321120839716e-05, "loss": 1.2991, "step": 9586 }, { "epoch": 0.97, "learning_rate": 5.1662665594712864e-05, "loss": 0.359, "step": 9587 }, { "epoch": 0.97, "learning_rate": 5.1661009930733054e-05, "loss": 0.3798, "step": 9588 }, { "epoch": 0.97, "learning_rate": 5.165935412891082e-05, "loss": 1.4588, "step": 9589 }, { "epoch": 0.97, "learning_rate": 5.16576981892567e-05, "loss": 0.6304, "step": 9590 }, { "epoch": 0.97, "learning_rate": 5.165604211178121e-05, "loss": 0.4929, "step": 9591 }, { "epoch": 0.97, "learning_rate": 5.165438589649492e-05, "loss": 0.6867, "step": 9592 }, { "epoch": 0.97, "learning_rate": 5.165272954340835e-05, "loss": 1.3112, "step": 9593 }, { "epoch": 0.97, "learning_rate": 5.165107305253206e-05, "loss": 0.3117, "step": 9594 }, { "epoch": 0.97, "learning_rate": 5.164941642387656e-05, "loss": 1.4872, "step": 9595 }, { "epoch": 0.97, "learning_rate": 5.1647759657452416e-05, "loss": 0.4116, "step": 9596 }, { "epoch": 0.97, "learning_rate": 5.1646102753270166e-05, "loss": 1.3974, "step": 9597 }, { "epoch": 0.97, "learning_rate": 5.1644445711340356e-05, "loss": 0.9462, "step": 9598 }, { "epoch": 0.97, "learning_rate": 5.164278853167353e-05, "loss": 0.2833, "step": 9599 }, { "epoch": 0.97, "learning_rate": 5.164113121428022e-05, "loss": 1.5098, "step": 9600 }, { "epoch": 0.97, "learning_rate": 5.1639473759170984e-05, "loss": 1.3405, "step": 9601 }, { "epoch": 0.98, "learning_rate": 5.1637816166356376e-05, "loss": 0.759, "step": 9602 }, { "epoch": 0.98, "learning_rate": 5.163615843584694e-05, "loss": 1.4082, "step": 9603 }, { "epoch": 0.98, "learning_rate": 5.163450056765321e-05, "loss": 1.1367, "step": 9604 }, { "epoch": 0.98, "learning_rate": 5.163284256178576e-05, "loss": 0.7303, "step": 9605 }, { "epoch": 0.98, "learning_rate": 5.163118441825512e-05, "loss": 0.5455, "step": 9606 }, { "epoch": 0.98, "learning_rate": 5.162952613707186e-05, "loss": 1.2521, "step": 9607 }, { "epoch": 0.98, "learning_rate": 5.162786771824652e-05, "loss": 0.4107, "step": 9608 }, { "epoch": 0.98, "learning_rate": 5.162620916178966e-05, "loss": 0.333, "step": 9609 }, { "epoch": 0.98, "learning_rate": 5.162455046771183e-05, "loss": 1.1975, "step": 9610 }, { "epoch": 0.98, "learning_rate": 5.1622891636023595e-05, "loss": 0.8829, "step": 9611 }, { "epoch": 0.98, "learning_rate": 5.1621232666735487e-05, "loss": 0.4806, "step": 9612 }, { "epoch": 0.98, "learning_rate": 5.1619573559858084e-05, "loss": 0.665, "step": 9613 }, { "epoch": 0.98, "learning_rate": 5.161791431540195e-05, "loss": 0.9461, "step": 9614 }, { "epoch": 0.98, "learning_rate": 5.1616254933377626e-05, "loss": 0.3718, "step": 9615 }, { "epoch": 0.98, "learning_rate": 5.161459541379568e-05, "loss": 1.078, "step": 9616 }, { "epoch": 0.98, "learning_rate": 5.161293575666667e-05, "loss": 1.8842, "step": 9617 }, { "epoch": 0.98, "learning_rate": 5.1611275962001166e-05, "loss": 1.0842, "step": 9618 }, { "epoch": 0.98, "learning_rate": 5.160961602980971e-05, "loss": 0.8823, "step": 9619 }, { "epoch": 0.98, "learning_rate": 5.160795596010289e-05, "loss": 0.4129, "step": 9620 }, { "epoch": 0.98, "learning_rate": 5.160629575289125e-05, "loss": 1.3962, "step": 9621 }, { "epoch": 0.98, "learning_rate": 5.160463540818537e-05, "loss": 1.062, "step": 9622 }, { "epoch": 0.98, "learning_rate": 5.160297492599581e-05, "loss": 1.0955, "step": 9623 }, { "epoch": 0.98, "learning_rate": 5.160131430633312e-05, "loss": 1.3604, "step": 9624 }, { "epoch": 0.98, "learning_rate": 5.15996535492079e-05, "loss": 1.0935, "step": 9625 }, { "epoch": 0.98, "learning_rate": 5.1597992654630695e-05, "loss": 1.026, "step": 9626 }, { "epoch": 0.98, "learning_rate": 5.159633162261209e-05, "loss": 0.7001, "step": 9627 }, { "epoch": 0.98, "learning_rate": 5.1594670453162634e-05, "loss": 0.1525, "step": 9628 }, { "epoch": 0.98, "learning_rate": 5.159300914629292e-05, "loss": 0.3052, "step": 9629 }, { "epoch": 0.98, "learning_rate": 5.1591347702013504e-05, "loss": 1.4164, "step": 9630 }, { "epoch": 0.98, "learning_rate": 5.158968612033496e-05, "loss": 0.5024, "step": 9631 }, { "epoch": 0.98, "learning_rate": 5.158802440126788e-05, "loss": 0.4127, "step": 9632 }, { "epoch": 0.98, "learning_rate": 5.1586362544822814e-05, "loss": 0.9235, "step": 9633 }, { "epoch": 0.98, "learning_rate": 5.1584700551010356e-05, "loss": 1.2838, "step": 9634 }, { "epoch": 0.98, "learning_rate": 5.158303841984107e-05, "loss": 1.6532, "step": 9635 }, { "epoch": 0.98, "learning_rate": 5.158137615132554e-05, "loss": 0.6589, "step": 9636 }, { "epoch": 0.98, "learning_rate": 5.157971374547434e-05, "loss": 1.4652, "step": 9637 }, { "epoch": 0.98, "learning_rate": 5.157805120229806e-05, "loss": 1.21, "step": 9638 }, { "epoch": 0.98, "learning_rate": 5.157638852180726e-05, "loss": 1.5413, "step": 9639 }, { "epoch": 0.98, "learning_rate": 5.157472570401253e-05, "loss": 0.605, "step": 9640 }, { "epoch": 0.98, "learning_rate": 5.157306274892447e-05, "loss": 0.9825, "step": 9641 }, { "epoch": 0.98, "learning_rate": 5.157139965655363e-05, "loss": 0.3103, "step": 9642 }, { "epoch": 0.98, "learning_rate": 5.1569736426910616e-05, "loss": 0.4315, "step": 9643 }, { "epoch": 0.98, "learning_rate": 5.1568073060006e-05, "loss": 1.2204, "step": 9644 }, { "epoch": 0.98, "learning_rate": 5.1566409555850375e-05, "loss": 0.449, "step": 9645 }, { "epoch": 0.98, "learning_rate": 5.156474591445432e-05, "loss": 0.6308, "step": 9646 }, { "epoch": 0.98, "learning_rate": 5.156308213582843e-05, "loss": 0.1714, "step": 9647 }, { "epoch": 0.98, "learning_rate": 5.156141821998329e-05, "loss": 0.4504, "step": 9648 }, { "epoch": 0.98, "learning_rate": 5.155975416692948e-05, "loss": 1.4226, "step": 9649 }, { "epoch": 0.98, "learning_rate": 5.15580899766776e-05, "loss": 1.7002, "step": 9650 }, { "epoch": 0.98, "learning_rate": 5.155642564923824e-05, "loss": 0.6787, "step": 9651 }, { "epoch": 0.98, "learning_rate": 5.155476118462198e-05, "loss": 0.3269, "step": 9652 }, { "epoch": 0.98, "learning_rate": 5.155309658283942e-05, "loss": 0.6628, "step": 9653 }, { "epoch": 0.98, "learning_rate": 5.155143184390116e-05, "loss": 1.0312, "step": 9654 }, { "epoch": 0.98, "learning_rate": 5.1549766967817785e-05, "loss": 1.0238, "step": 9655 }, { "epoch": 0.98, "learning_rate": 5.154810195459988e-05, "loss": 1.2264, "step": 9656 }, { "epoch": 0.98, "learning_rate": 5.1546436804258065e-05, "loss": 0.3442, "step": 9657 }, { "epoch": 0.98, "learning_rate": 5.1544771516802925e-05, "loss": 0.4819, "step": 9658 }, { "epoch": 0.98, "learning_rate": 5.154310609224504e-05, "loss": 0.2766, "step": 9659 }, { "epoch": 0.98, "learning_rate": 5.1541440530595035e-05, "loss": 1.5943, "step": 9660 }, { "epoch": 0.98, "learning_rate": 5.153977483186349e-05, "loss": 0.8354, "step": 9661 }, { "epoch": 0.98, "learning_rate": 5.153810899606102e-05, "loss": 0.549, "step": 9662 }, { "epoch": 0.98, "learning_rate": 5.153644302319821e-05, "loss": 1.1099, "step": 9663 }, { "epoch": 0.98, "learning_rate": 5.153477691328567e-05, "loss": 0.5344, "step": 9664 }, { "epoch": 0.98, "learning_rate": 5.1533110666334e-05, "loss": 0.9287, "step": 9665 }, { "epoch": 0.98, "learning_rate": 5.153144428235382e-05, "loss": 0.824, "step": 9666 }, { "epoch": 0.98, "learning_rate": 5.152977776135571e-05, "loss": 0.3883, "step": 9667 }, { "epoch": 0.98, "learning_rate": 5.152811110335028e-05, "loss": 1.6209, "step": 9668 }, { "epoch": 0.98, "learning_rate": 5.152644430834815e-05, "loss": 0.9873, "step": 9669 }, { "epoch": 0.98, "learning_rate": 5.1524777376359915e-05, "loss": 0.4897, "step": 9670 }, { "epoch": 0.98, "learning_rate": 5.152311030739617e-05, "loss": 0.309, "step": 9671 }, { "epoch": 0.98, "learning_rate": 5.152144310146756e-05, "loss": 1.4091, "step": 9672 }, { "epoch": 0.98, "learning_rate": 5.151977575858466e-05, "loss": 0.8556, "step": 9673 }, { "epoch": 0.98, "learning_rate": 5.151810827875811e-05, "loss": 1.1161, "step": 9674 }, { "epoch": 0.98, "learning_rate": 5.151644066199849e-05, "loss": 1.4445, "step": 9675 }, { "epoch": 0.98, "learning_rate": 5.1514772908316425e-05, "loss": 0.2701, "step": 9676 }, { "epoch": 0.98, "learning_rate": 5.151310501772254e-05, "loss": 0.6194, "step": 9677 }, { "epoch": 0.98, "learning_rate": 5.1511436990227426e-05, "loss": 1.0162, "step": 9678 }, { "epoch": 0.98, "learning_rate": 5.150976882584172e-05, "loss": 0.5695, "step": 9679 }, { "epoch": 0.98, "learning_rate": 5.1508100524576024e-05, "loss": 0.3182, "step": 9680 }, { "epoch": 0.98, "learning_rate": 5.1506432086440965e-05, "loss": 1.2526, "step": 9681 }, { "epoch": 0.98, "learning_rate": 5.150476351144715e-05, "loss": 0.5867, "step": 9682 }, { "epoch": 0.98, "learning_rate": 5.15030947996052e-05, "loss": 1.233, "step": 9683 }, { "epoch": 0.98, "learning_rate": 5.1501425950925736e-05, "loss": 1.1774, "step": 9684 }, { "epoch": 0.98, "learning_rate": 5.149975696541937e-05, "loss": 0.5524, "step": 9685 }, { "epoch": 0.98, "learning_rate": 5.1498087843096736e-05, "loss": 0.9604, "step": 9686 }, { "epoch": 0.98, "learning_rate": 5.1496418583968454e-05, "loss": 0.4201, "step": 9687 }, { "epoch": 0.98, "learning_rate": 5.149474918804514e-05, "loss": 1.0797, "step": 9688 }, { "epoch": 0.98, "learning_rate": 5.14930796553374e-05, "loss": 0.8316, "step": 9689 }, { "epoch": 0.98, "learning_rate": 5.1491409985855905e-05, "loss": 1.201, "step": 9690 }, { "epoch": 0.98, "learning_rate": 5.148974017961124e-05, "loss": 0.6438, "step": 9691 }, { "epoch": 0.98, "learning_rate": 5.1488070236614036e-05, "loss": 0.9138, "step": 9692 }, { "epoch": 0.98, "learning_rate": 5.148640015687494e-05, "loss": 0.508, "step": 9693 }, { "epoch": 0.98, "learning_rate": 5.1484729940404565e-05, "loss": 0.9358, "step": 9694 }, { "epoch": 0.98, "learning_rate": 5.1483059587213535e-05, "loss": 1.2207, "step": 9695 }, { "epoch": 0.98, "learning_rate": 5.14813890973125e-05, "loss": 1.7415, "step": 9696 }, { "epoch": 0.98, "learning_rate": 5.1479718470712055e-05, "loss": 0.9441, "step": 9697 }, { "epoch": 0.98, "learning_rate": 5.1478047707422876e-05, "loss": 0.2872, "step": 9698 }, { "epoch": 0.98, "learning_rate": 5.147637680745556e-05, "loss": 0.5388, "step": 9699 }, { "epoch": 0.98, "learning_rate": 5.1474705770820755e-05, "loss": 0.6982, "step": 9700 }, { "epoch": 0.99, "learning_rate": 5.147303459752909e-05, "loss": 1.3192, "step": 9701 }, { "epoch": 0.99, "learning_rate": 5.1471363287591194e-05, "loss": 0.8154, "step": 9702 }, { "epoch": 0.99, "learning_rate": 5.146969184101773e-05, "loss": 1.5027, "step": 9703 }, { "epoch": 0.99, "learning_rate": 5.1468020257819296e-05, "loss": 1.675, "step": 9704 }, { "epoch": 0.99, "learning_rate": 5.146634853800656e-05, "loss": 0.599, "step": 9705 }, { "epoch": 0.99, "learning_rate": 5.146467668159014e-05, "loss": 0.3209, "step": 9706 }, { "epoch": 0.99, "learning_rate": 5.1463004688580685e-05, "loss": 0.5416, "step": 9707 }, { "epoch": 0.99, "learning_rate": 5.1461332558988845e-05, "loss": 0.4022, "step": 9708 }, { "epoch": 0.99, "learning_rate": 5.1459660292825235e-05, "loss": 0.9296, "step": 9709 }, { "epoch": 0.99, "learning_rate": 5.145798789010051e-05, "loss": 1.7327, "step": 9710 }, { "epoch": 0.99, "learning_rate": 5.145631535082532e-05, "loss": 1.4516, "step": 9711 }, { "epoch": 0.99, "learning_rate": 5.1454642675010295e-05, "loss": 1.0753, "step": 9712 }, { "epoch": 0.99, "learning_rate": 5.145296986266609e-05, "loss": 1.1785, "step": 9713 }, { "epoch": 0.99, "learning_rate": 5.145129691380335e-05, "loss": 0.7905, "step": 9714 }, { "epoch": 0.99, "learning_rate": 5.1449623828432706e-05, "loss": 1.5552, "step": 9715 }, { "epoch": 0.99, "learning_rate": 5.144795060656482e-05, "loss": 0.6411, "step": 9716 }, { "epoch": 0.99, "learning_rate": 5.144627724821034e-05, "loss": 1.1137, "step": 9717 }, { "epoch": 0.99, "learning_rate": 5.144460375337991e-05, "loss": 1.4045, "step": 9718 }, { "epoch": 0.99, "learning_rate": 5.144293012208418e-05, "loss": 1.407, "step": 9719 }, { "epoch": 0.99, "learning_rate": 5.1441256354333795e-05, "loss": 1.4755, "step": 9720 }, { "epoch": 0.99, "learning_rate": 5.143958245013941e-05, "loss": 0.7295, "step": 9721 }, { "epoch": 0.99, "learning_rate": 5.1437908409511685e-05, "loss": 0.5378, "step": 9722 }, { "epoch": 0.99, "learning_rate": 5.143623423246127e-05, "loss": 0.899, "step": 9723 }, { "epoch": 0.99, "learning_rate": 5.14345599189988e-05, "loss": 0.6188, "step": 9724 }, { "epoch": 0.99, "learning_rate": 5.143288546913496e-05, "loss": 0.4155, "step": 9725 }, { "epoch": 0.99, "learning_rate": 5.1431210882880375e-05, "loss": 1.0904, "step": 9726 }, { "epoch": 0.99, "learning_rate": 5.142953616024573e-05, "loss": 1.8238, "step": 9727 }, { "epoch": 0.99, "learning_rate": 5.142786130124166e-05, "loss": 0.3747, "step": 9728 }, { "epoch": 0.99, "learning_rate": 5.142618630587884e-05, "loss": 0.5401, "step": 9729 }, { "epoch": 0.99, "learning_rate": 5.1424511174167916e-05, "loss": 1.4254, "step": 9730 }, { "epoch": 0.99, "learning_rate": 5.1422835906119554e-05, "loss": 1.3384, "step": 9731 }, { "epoch": 0.99, "learning_rate": 5.142116050174442e-05, "loss": 0.694, "step": 9732 }, { "epoch": 0.99, "learning_rate": 5.1419484961053164e-05, "loss": 0.7597, "step": 9733 }, { "epoch": 0.99, "learning_rate": 5.141780928405645e-05, "loss": 1.4272, "step": 9734 }, { "epoch": 0.99, "learning_rate": 5.141613347076495e-05, "loss": 0.5274, "step": 9735 }, { "epoch": 0.99, "learning_rate": 5.141445752118932e-05, "loss": 1.6496, "step": 9736 }, { "epoch": 0.99, "learning_rate": 5.141278143534024e-05, "loss": 0.1233, "step": 9737 }, { "epoch": 0.99, "learning_rate": 5.1411105213228354e-05, "loss": 1.281, "step": 9738 }, { "epoch": 0.99, "learning_rate": 5.140942885486435e-05, "loss": 1.2136, "step": 9739 }, { "epoch": 0.99, "learning_rate": 5.1407752360258875e-05, "loss": 0.4499, "step": 9740 }, { "epoch": 0.99, "learning_rate": 5.1406075729422606e-05, "loss": 0.3481, "step": 9741 }, { "epoch": 0.99, "learning_rate": 5.140439896236623e-05, "loss": 1.1807, "step": 9742 }, { "epoch": 0.99, "learning_rate": 5.140272205910038e-05, "loss": 0.5802, "step": 9743 }, { "epoch": 0.99, "learning_rate": 5.140104501963576e-05, "loss": 1.3998, "step": 9744 }, { "epoch": 0.99, "learning_rate": 5.139936784398304e-05, "loss": 0.9448, "step": 9745 }, { "epoch": 0.99, "learning_rate": 5.139769053215287e-05, "loss": 0.4285, "step": 9746 }, { "epoch": 0.99, "learning_rate": 5.1396013084155945e-05, "loss": 0.6406, "step": 9747 }, { "epoch": 0.99, "learning_rate": 5.139433550000293e-05, "loss": 0.7449, "step": 9748 }, { "epoch": 0.99, "learning_rate": 5.1392657779704496e-05, "loss": 0.5227, "step": 9749 }, { "epoch": 0.99, "learning_rate": 5.139097992327134e-05, "loss": 0.8275, "step": 9750 }, { "epoch": 0.99, "learning_rate": 5.1389301930714106e-05, "loss": 0.6038, "step": 9751 }, { "epoch": 0.99, "learning_rate": 5.138762380204351e-05, "loss": 1.592, "step": 9752 }, { "epoch": 0.99, "learning_rate": 5.13859455372702e-05, "loss": 0.9661, "step": 9753 }, { "epoch": 0.99, "learning_rate": 5.138426713640488e-05, "loss": 1.2595, "step": 9754 }, { "epoch": 0.99, "learning_rate": 5.138258859945822e-05, "loss": 0.6822, "step": 9755 }, { "epoch": 0.99, "learning_rate": 5.138090992644089e-05, "loss": 0.841, "step": 9756 }, { "epoch": 0.99, "learning_rate": 5.137923111736358e-05, "loss": 1.3556, "step": 9757 }, { "epoch": 0.99, "learning_rate": 5.1377552172236983e-05, "loss": 1.2377, "step": 9758 }, { "epoch": 0.99, "learning_rate": 5.137587309107177e-05, "loss": 0.9584, "step": 9759 }, { "epoch": 0.99, "learning_rate": 5.137419387387864e-05, "loss": 0.6075, "step": 9760 }, { "epoch": 0.99, "learning_rate": 5.1372514520668274e-05, "loss": 1.4536, "step": 9761 }, { "epoch": 0.99, "learning_rate": 5.137083503145135e-05, "loss": 1.0504, "step": 9762 }, { "epoch": 0.99, "learning_rate": 5.136915540623857e-05, "loss": 0.9709, "step": 9763 }, { "epoch": 0.99, "learning_rate": 5.1367475645040605e-05, "loss": 0.8049, "step": 9764 }, { "epoch": 0.99, "learning_rate": 5.136579574786816e-05, "loss": 1.2269, "step": 9765 }, { "epoch": 0.99, "learning_rate": 5.136411571473191e-05, "loss": 0.4046, "step": 9766 }, { "epoch": 0.99, "learning_rate": 5.136243554564256e-05, "loss": 0.4088, "step": 9767 }, { "epoch": 0.99, "learning_rate": 5.136075524061079e-05, "loss": 0.7947, "step": 9768 }, { "epoch": 0.99, "learning_rate": 5.135907479964732e-05, "loss": 0.3527, "step": 9769 }, { "epoch": 0.99, "learning_rate": 5.13573942227628e-05, "loss": 1.0743, "step": 9770 }, { "epoch": 0.99, "learning_rate": 5.1355713509967966e-05, "loss": 1.2227, "step": 9771 }, { "epoch": 0.99, "learning_rate": 5.135403266127348e-05, "loss": 0.8219, "step": 9772 }, { "epoch": 0.99, "learning_rate": 5.135235167669006e-05, "loss": 0.9886, "step": 9773 }, { "epoch": 0.99, "learning_rate": 5.1350670556228404e-05, "loss": 0.5332, "step": 9774 }, { "epoch": 0.99, "learning_rate": 5.134898929989919e-05, "loss": 0.7607, "step": 9775 }, { "epoch": 0.99, "learning_rate": 5.134730790771314e-05, "loss": 0.4543, "step": 9776 }, { "epoch": 0.99, "learning_rate": 5.1345626379680936e-05, "loss": 1.5176, "step": 9777 }, { "epoch": 0.99, "learning_rate": 5.134394471581329e-05, "loss": 1.4569, "step": 9778 }, { "epoch": 0.99, "learning_rate": 5.13422629161209e-05, "loss": 0.2274, "step": 9779 }, { "epoch": 0.99, "learning_rate": 5.134058098061447e-05, "loss": 0.4708, "step": 9780 }, { "epoch": 0.99, "learning_rate": 5.1338898909304694e-05, "loss": 0.5952, "step": 9781 }, { "epoch": 0.99, "learning_rate": 5.133721670220229e-05, "loss": 1.3616, "step": 9782 }, { "epoch": 0.99, "learning_rate": 5.133553435931795e-05, "loss": 1.3742, "step": 9783 }, { "epoch": 0.99, "learning_rate": 5.133385188066239e-05, "loss": 1.3254, "step": 9784 }, { "epoch": 0.99, "learning_rate": 5.133216926624631e-05, "loss": 0.6484, "step": 9785 }, { "epoch": 0.99, "learning_rate": 5.133048651608042e-05, "loss": 0.7062, "step": 9786 }, { "epoch": 0.99, "learning_rate": 5.132880363017543e-05, "loss": 0.5231, "step": 9787 }, { "epoch": 0.99, "learning_rate": 5.132712060854205e-05, "loss": 1.3441, "step": 9788 }, { "epoch": 0.99, "learning_rate": 5.132543745119098e-05, "loss": 0.6204, "step": 9789 }, { "epoch": 0.99, "learning_rate": 5.1323754158132946e-05, "loss": 1.3462, "step": 9790 }, { "epoch": 0.99, "learning_rate": 5.132207072937865e-05, "loss": 1.0065, "step": 9791 }, { "epoch": 0.99, "learning_rate": 5.1320387164938804e-05, "loss": 0.9118, "step": 9792 }, { "epoch": 0.99, "learning_rate": 5.131870346482413e-05, "loss": 0.7603, "step": 9793 }, { "epoch": 0.99, "learning_rate": 5.131701962904534e-05, "loss": 1.0233, "step": 9794 }, { "epoch": 0.99, "learning_rate": 5.131533565761313e-05, "loss": 1.1877, "step": 9795 }, { "epoch": 0.99, "learning_rate": 5.131365155053825e-05, "loss": 0.5582, "step": 9796 }, { "epoch": 0.99, "learning_rate": 5.13119673078314e-05, "loss": 1.5619, "step": 9797 }, { "epoch": 0.99, "learning_rate": 5.131028292950329e-05, "loss": 2.6783, "step": 9798 }, { "epoch": 1.0, "learning_rate": 5.130859841556465e-05, "loss": 0.6939, "step": 9799 }, { "epoch": 1.0, "learning_rate": 5.130691376602619e-05, "loss": 0.5525, "step": 9800 }, { "epoch": 1.0, "learning_rate": 5.130522898089864e-05, "loss": 0.4851, "step": 9801 }, { "epoch": 1.0, "learning_rate": 5.130354406019272e-05, "loss": 1.4872, "step": 9802 }, { "epoch": 1.0, "learning_rate": 5.130185900391915e-05, "loss": 0.6113, "step": 9803 }, { "epoch": 1.0, "learning_rate": 5.1300173812088644e-05, "loss": 1.0235, "step": 9804 }, { "epoch": 1.0, "learning_rate": 5.129848848471194e-05, "loss": 0.388, "step": 9805 }, { "epoch": 1.0, "learning_rate": 5.129680302179976e-05, "loss": 1.3518, "step": 9806 }, { "epoch": 1.0, "learning_rate": 5.1295117423362825e-05, "loss": 1.1895, "step": 9807 }, { "epoch": 1.0, "learning_rate": 5.129343168941186e-05, "loss": 1.2792, "step": 9808 }, { "epoch": 1.0, "learning_rate": 5.1291745819957605e-05, "loss": 0.3232, "step": 9809 }, { "epoch": 1.0, "learning_rate": 5.129005981501078e-05, "loss": 1.1598, "step": 9810 }, { "epoch": 1.0, "learning_rate": 5.12883736745821e-05, "loss": 0.4241, "step": 9811 }, { "epoch": 1.0, "learning_rate": 5.1286687398682324e-05, "loss": 0.8171, "step": 9812 }, { "epoch": 1.0, "learning_rate": 5.1285000987322164e-05, "loss": 1.8208, "step": 9813 }, { "epoch": 1.0, "learning_rate": 5.128331444051234e-05, "loss": 0.7778, "step": 9814 }, { "epoch": 1.0, "learning_rate": 5.128162775826362e-05, "loss": 0.7202, "step": 9815 }, { "epoch": 1.0, "learning_rate": 5.127994094058671e-05, "loss": 0.7287, "step": 9816 }, { "epoch": 1.0, "learning_rate": 5.1278253987492346e-05, "loss": 0.7039, "step": 9817 }, { "epoch": 1.0, "learning_rate": 5.127656689899127e-05, "loss": 0.9916, "step": 9818 }, { "epoch": 1.0, "learning_rate": 5.1274879675094226e-05, "loss": 0.376, "step": 9819 }, { "epoch": 1.0, "learning_rate": 5.1273192315811925e-05, "loss": 1.0307, "step": 9820 }, { "epoch": 1.0, "learning_rate": 5.127150482115513e-05, "loss": 0.689, "step": 9821 }, { "epoch": 1.0, "learning_rate": 5.126981719113457e-05, "loss": 0.4435, "step": 9822 }, { "epoch": 1.0, "learning_rate": 5.126812942576098e-05, "loss": 0.3826, "step": 9823 }, { "epoch": 1.0, "learning_rate": 5.1266441525045115e-05, "loss": 1.8374, "step": 9824 }, { "epoch": 1.0, "learning_rate": 5.12647534889977e-05, "loss": 1.6093, "step": 9825 }, { "epoch": 1.0, "learning_rate": 5.1263065317629486e-05, "loss": 2.2007, "step": 9826 }, { "epoch": 1.0, "learning_rate": 5.12613770109512e-05, "loss": 0.4435, "step": 9827 }, { "epoch": 1.0, "learning_rate": 5.125968856897362e-05, "loss": 0.1576, "step": 9828 }, { "epoch": 1.0, "learning_rate": 5.1257999991707455e-05, "loss": 0.5218, "step": 9829 }, { "epoch": 1.0, "learning_rate": 5.125631127916347e-05, "loss": 0.1557, "step": 9830 }, { "epoch": 1.0, "learning_rate": 5.125462243135241e-05, "loss": 1.3982, "step": 9831 }, { "epoch": 1.0, "learning_rate": 5.1252933448285013e-05, "loss": 1.3734, "step": 9832 }, { "epoch": 1.0, "learning_rate": 5.1251244329972034e-05, "loss": 0.9552, "step": 9833 }, { "epoch": 1.0, "learning_rate": 5.124955507642423e-05, "loss": 1.7075, "step": 9834 }, { "epoch": 1.0, "learning_rate": 5.124786568765232e-05, "loss": 1.5997, "step": 9835 }, { "epoch": 1.0, "learning_rate": 5.1246176163667086e-05, "loss": 0.7346, "step": 9836 }, { "epoch": 1.0, "learning_rate": 5.124448650447927e-05, "loss": 1.4396, "step": 9837 }, { "epoch": 1.0, "learning_rate": 5.124279671009963e-05, "loss": 1.1977, "step": 9838 }, { "epoch": 1.0, "learning_rate": 5.1241106780538905e-05, "loss": 1.2189, "step": 9839 }, { "epoch": 1.0, "learning_rate": 5.123941671580786e-05, "loss": 0.9315, "step": 9840 }, { "epoch": 1.0, "learning_rate": 5.123772651591725e-05, "loss": 0.671, "step": 9841 }, { "epoch": 1.0, "learning_rate": 5.123603618087782e-05, "loss": 0.9388, "step": 9842 }, { "epoch": 1.0, "learning_rate": 5.123434571070034e-05, "loss": 0.8185, "step": 9843 }, { "epoch": 1.0, "learning_rate": 5.1232655105395556e-05, "loss": 0.5843, "step": 9844 }, { "epoch": 1.0, "learning_rate": 5.1230964364974245e-05, "loss": 0.7122, "step": 9845 }, { "epoch": 1.0, "learning_rate": 5.122927348944714e-05, "loss": 1.5117, "step": 9846 }, { "epoch": 1.0, "learning_rate": 5.1227582478825016e-05, "loss": 0.4915, "step": 9847 }, { "epoch": 1.0, "learning_rate": 5.122589133311863e-05, "loss": 0.9005, "step": 9848 }, { "epoch": 1.0, "learning_rate": 5.1224200052338765e-05, "loss": 0.7782, "step": 9849 }, { "epoch": 1.0, "learning_rate": 5.122250863649615e-05, "loss": 1.7332, "step": 9850 }, { "epoch": 1.0, "learning_rate": 5.122081708560156e-05, "loss": 1.0507, "step": 9851 }, { "epoch": 1.0, "learning_rate": 5.1219125399665775e-05, "loss": 1.1134, "step": 9852 }, { "epoch": 1.0, "learning_rate": 5.1217433578699536e-05, "loss": 0.2872, "step": 9853 }, { "epoch": 1.0, "learning_rate": 5.121574162271363e-05, "loss": 1.1051, "step": 9854 }, { "epoch": 1.0, "learning_rate": 5.1214049531718816e-05, "loss": 1.1271, "step": 9855 }, { "epoch": 1.0, "learning_rate": 5.1212357305725856e-05, "loss": 0.4226, "step": 9856 }, { "epoch": 1.0, "learning_rate": 5.1210664944745525e-05, "loss": 1.2312, "step": 9857 }, { "epoch": 1.0, "learning_rate": 5.120897244878858e-05, "loss": 0.3025, "step": 9858 }, { "epoch": 1.0, "learning_rate": 5.120727981786583e-05, "loss": 0.4082, "step": 9859 }, { "epoch": 1.0, "learning_rate": 5.1205587051987997e-05, "loss": 1.3968, "step": 9860 }, { "epoch": 1.0, "learning_rate": 5.1203894151165874e-05, "loss": 1.3312, "step": 9861 }, { "epoch": 1.0, "learning_rate": 5.120220111541024e-05, "loss": 1.0046, "step": 9862 }, { "epoch": 1.0, "learning_rate": 5.120050794473187e-05, "loss": 0.7127, "step": 9863 }, { "epoch": 1.0, "learning_rate": 5.119881463914153e-05, "loss": 0.2777, "step": 9864 }, { "epoch": 1.0, "learning_rate": 5.119712119865e-05, "loss": 0.7565, "step": 9865 }, { "epoch": 1.0, "learning_rate": 5.119542762326805e-05, "loss": 0.6742, "step": 9866 }, { "epoch": 1.0, "learning_rate": 5.119373391300646e-05, "loss": 0.2829, "step": 9867 }, { "epoch": 1.0, "learning_rate": 5.1192040067876006e-05, "loss": 0.5579, "step": 9868 }, { "epoch": 1.0, "learning_rate": 5.1190346087887484e-05, "loss": 1.3076, "step": 9869 }, { "epoch": 1.0, "learning_rate": 5.118865197305165e-05, "loss": 0.062, "step": 9870 }, { "epoch": 1.0, "learning_rate": 5.11869577233793e-05, "loss": 1.8487, "step": 9871 }, { "epoch": 1.0, "learning_rate": 5.118526333888122e-05, "loss": 1.1424, "step": 9872 }, { "epoch": 1.0, "learning_rate": 5.118356881956817e-05, "loss": 1.1942, "step": 9873 }, { "epoch": 1.0, "learning_rate": 5.1181874165450946e-05, "loss": 0.419, "step": 9874 }, { "epoch": 1.0, "learning_rate": 5.118017937654034e-05, "loss": 0.9841, "step": 9875 }, { "epoch": 1.0, "learning_rate": 5.1178484452847125e-05, "loss": 0.4205, "step": 9876 }, { "epoch": 1.0, "learning_rate": 5.117678939438209e-05, "loss": 1.1439, "step": 9877 }, { "epoch": 1.0, "learning_rate": 5.117509420115603e-05, "loss": 0.8497, "step": 9878 }, { "epoch": 1.0, "learning_rate": 5.117339887317972e-05, "loss": 1.1761, "step": 9879 }, { "epoch": 1.0, "learning_rate": 5.1171703410463965e-05, "loss": 2.2265, "step": 9880 }, { "epoch": 1.0, "learning_rate": 5.1170007813019525e-05, "loss": 1.2195, "step": 9881 }, { "epoch": 1.0, "learning_rate": 5.1168312080857225e-05, "loss": 1.1706, "step": 9882 }, { "epoch": 1.0, "learning_rate": 5.116661621398784e-05, "loss": 1.5886, "step": 9883 }, { "epoch": 1.0, "learning_rate": 5.1164920212422145e-05, "loss": 0.7485, "step": 9884 }, { "epoch": 1.0, "learning_rate": 5.1163224076170966e-05, "loss": 1.3228, "step": 9885 }, { "epoch": 1.0, "learning_rate": 5.116152780524507e-05, "loss": 1.3356, "step": 9886 }, { "epoch": 1.0, "learning_rate": 5.115983139965527e-05, "loss": 0.3603, "step": 9887 }, { "epoch": 1.0, "learning_rate": 5.115813485941235e-05, "loss": 0.485, "step": 9888 }, { "epoch": 1.0, "learning_rate": 5.1156438184527104e-05, "loss": 0.7495, "step": 9889 }, { "epoch": 1.0, "learning_rate": 5.115474137501033e-05, "loss": 1.3115, "step": 9890 }, { "epoch": 1.0, "learning_rate": 5.115304443087284e-05, "loss": 1.76, "step": 9891 }, { "epoch": 1.0, "learning_rate": 5.115134735212542e-05, "loss": 0.3462, "step": 9892 }, { "epoch": 1.0, "learning_rate": 5.1149650138778867e-05, "loss": 0.4644, "step": 9893 }, { "epoch": 1.0, "learning_rate": 5.1147952790843994e-05, "loss": 1.3975, "step": 9894 }, { "epoch": 1.0, "learning_rate": 5.1146255308331586e-05, "loss": 0.6397, "step": 9895 }, { "epoch": 1.0, "learning_rate": 5.114455769125246e-05, "loss": 1.1477, "step": 9896 }, { "epoch": 1.0, "learning_rate": 5.114285993961741e-05, "loss": 0.9682, "step": 9897 }, { "epoch": 1.01, "learning_rate": 5.114116205343724e-05, "loss": 1.2461, "step": 9898 }, { "epoch": 1.01, "learning_rate": 5.113946403272276e-05, "loss": 0.3756, "step": 9899 }, { "epoch": 1.01, "learning_rate": 5.113776587748478e-05, "loss": 0.67, "step": 9900 }, { "epoch": 1.01, "learning_rate": 5.113606758773409e-05, "loss": 1.4258, "step": 9901 }, { "epoch": 1.01, "learning_rate": 5.113436916348151e-05, "loss": 1.0243, "step": 9902 }, { "epoch": 1.01, "learning_rate": 5.113267060473784e-05, "loss": 0.8218, "step": 9903 }, { "epoch": 1.01, "learning_rate": 5.113097191151389e-05, "loss": 1.6141, "step": 9904 }, { "epoch": 1.01, "learning_rate": 5.112927308382049e-05, "loss": 0.9089, "step": 9905 }, { "epoch": 1.01, "learning_rate": 5.112757412166842e-05, "loss": 1.1069, "step": 9906 }, { "epoch": 1.01, "learning_rate": 5.112587502506851e-05, "loss": 0.6788, "step": 9907 }, { "epoch": 1.01, "learning_rate": 5.112417579403157e-05, "loss": 0.9826, "step": 9908 }, { "epoch": 1.01, "learning_rate": 5.11224764285684e-05, "loss": 0.2788, "step": 9909 }, { "epoch": 1.01, "learning_rate": 5.1120776928689845e-05, "loss": 1.2236, "step": 9910 }, { "epoch": 1.01, "learning_rate": 5.1119077294406684e-05, "loss": 1.1181, "step": 9911 }, { "epoch": 1.01, "learning_rate": 5.111737752572976e-05, "loss": 0.6004, "step": 9912 }, { "epoch": 1.01, "learning_rate": 5.111567762266987e-05, "loss": 0.709, "step": 9913 }, { "epoch": 1.01, "learning_rate": 5.111397758523784e-05, "loss": 1.0871, "step": 9914 }, { "epoch": 1.01, "learning_rate": 5.111227741344449e-05, "loss": 1.0037, "step": 9915 }, { "epoch": 1.01, "learning_rate": 5.111057710730065e-05, "loss": 0.3645, "step": 9916 }, { "epoch": 1.01, "learning_rate": 5.110887666681711e-05, "loss": 1.6155, "step": 9917 }, { "epoch": 1.01, "learning_rate": 5.1107176092004724e-05, "loss": 1.3604, "step": 9918 }, { "epoch": 1.01, "learning_rate": 5.1105475382874293e-05, "loss": 1.1534, "step": 9919 }, { "epoch": 1.01, "learning_rate": 5.1103774539436645e-05, "loss": 0.6965, "step": 9920 }, { "epoch": 1.01, "learning_rate": 5.110207356170261e-05, "loss": 0.5282, "step": 9921 }, { "epoch": 1.01, "learning_rate": 5.1100372449682995e-05, "loss": 0.3575, "step": 9922 }, { "epoch": 1.01, "learning_rate": 5.1098671203388636e-05, "loss": 1.3258, "step": 9923 }, { "epoch": 1.01, "learning_rate": 5.1096969822830374e-05, "loss": 0.5754, "step": 9924 }, { "epoch": 1.01, "learning_rate": 5.1095268308019016e-05, "loss": 1.2384, "step": 9925 }, { "epoch": 1.01, "learning_rate": 5.10935666589654e-05, "loss": 1.2308, "step": 9926 }, { "epoch": 1.01, "learning_rate": 5.1091864875680346e-05, "loss": 0.3503, "step": 9927 }, { "epoch": 1.01, "learning_rate": 5.1090162958174685e-05, "loss": 0.4054, "step": 9928 }, { "epoch": 1.01, "learning_rate": 5.108846090645925e-05, "loss": 1.3902, "step": 9929 }, { "epoch": 1.01, "learning_rate": 5.1086758720544865e-05, "loss": 1.4759, "step": 9930 }, { "epoch": 1.01, "learning_rate": 5.1085056400442384e-05, "loss": 0.8938, "step": 9931 }, { "epoch": 1.01, "learning_rate": 5.108335394616261e-05, "loss": 0.8889, "step": 9932 }, { "epoch": 1.01, "learning_rate": 5.108165135771641e-05, "loss": 1.2016, "step": 9933 }, { "epoch": 1.01, "learning_rate": 5.1079948635114584e-05, "loss": 0.56, "step": 9934 }, { "epoch": 1.01, "learning_rate": 5.1078245778368e-05, "loss": 1.4449, "step": 9935 }, { "epoch": 1.01, "learning_rate": 5.107654278748746e-05, "loss": 1.6206, "step": 9936 }, { "epoch": 1.01, "learning_rate": 5.1074839662483836e-05, "loss": 1.081, "step": 9937 }, { "epoch": 1.01, "learning_rate": 5.107313640336794e-05, "loss": 0.4163, "step": 9938 }, { "epoch": 1.01, "learning_rate": 5.107143301015063e-05, "loss": 0.4586, "step": 9939 }, { "epoch": 1.01, "learning_rate": 5.106972948284273e-05, "loss": 1.1839, "step": 9940 }, { "epoch": 1.01, "learning_rate": 5.106802582145509e-05, "loss": 1.3409, "step": 9941 }, { "epoch": 1.01, "learning_rate": 5.106632202599855e-05, "loss": 0.3297, "step": 9942 }, { "epoch": 1.01, "learning_rate": 5.106461809648394e-05, "loss": 0.6593, "step": 9943 }, { "epoch": 1.01, "learning_rate": 5.1062914032922124e-05, "loss": 0.7712, "step": 9944 }, { "epoch": 1.01, "learning_rate": 5.1061209835323926e-05, "loss": 0.8544, "step": 9945 }, { "epoch": 1.01, "learning_rate": 5.10595055037002e-05, "loss": 0.7551, "step": 9946 }, { "epoch": 1.01, "learning_rate": 5.105780103806181e-05, "loss": 0.9756, "step": 9947 }, { "epoch": 1.01, "learning_rate": 5.105609643841957e-05, "loss": 0.6333, "step": 9948 }, { "epoch": 1.01, "learning_rate": 5.105439170478434e-05, "loss": 0.3182, "step": 9949 }, { "epoch": 1.01, "learning_rate": 5.105268683716698e-05, "loss": 0.5325, "step": 9950 }, { "epoch": 1.01, "learning_rate": 5.1050981835578335e-05, "loss": 0.2913, "step": 9951 }, { "epoch": 1.01, "learning_rate": 5.1049276700029234e-05, "loss": 0.8316, "step": 9952 }, { "epoch": 1.01, "learning_rate": 5.104757143053055e-05, "loss": 1.41, "step": 9953 }, { "epoch": 1.01, "learning_rate": 5.1045866027093134e-05, "loss": 1.0803, "step": 9954 }, { "epoch": 1.01, "learning_rate": 5.104416048972783e-05, "loss": 0.5318, "step": 9955 }, { "epoch": 1.01, "learning_rate": 5.1042454818445496e-05, "loss": 0.9802, "step": 9956 }, { "epoch": 1.01, "learning_rate": 5.104074901325698e-05, "loss": 1.267, "step": 9957 }, { "epoch": 1.01, "learning_rate": 5.1039043074173145e-05, "loss": 0.3265, "step": 9958 }, { "epoch": 1.01, "learning_rate": 5.1037337001204835e-05, "loss": 1.1303, "step": 9959 }, { "epoch": 1.01, "learning_rate": 5.1035630794362926e-05, "loss": 0.805, "step": 9960 }, { "epoch": 1.01, "learning_rate": 5.103392445365826e-05, "loss": 0.3073, "step": 9961 }, { "epoch": 1.01, "learning_rate": 5.103221797910169e-05, "loss": 0.8807, "step": 9962 }, { "epoch": 1.01, "learning_rate": 5.103051137070411e-05, "loss": 0.6903, "step": 9963 }, { "epoch": 1.01, "learning_rate": 5.1028804628476335e-05, "loss": 0.6977, "step": 9964 }, { "epoch": 1.01, "learning_rate": 5.102709775242926e-05, "loss": 0.232, "step": 9965 }, { "epoch": 1.01, "learning_rate": 5.102539074257373e-05, "loss": 1.3969, "step": 9966 }, { "epoch": 1.01, "learning_rate": 5.1023683598920616e-05, "loss": 1.1918, "step": 9967 }, { "epoch": 1.01, "learning_rate": 5.1021976321480765e-05, "loss": 1.3124, "step": 9968 }, { "epoch": 1.01, "learning_rate": 5.102026891026506e-05, "loss": 0.3084, "step": 9969 }, { "epoch": 1.01, "learning_rate": 5.101856136528437e-05, "loss": 0.8225, "step": 9970 }, { "epoch": 1.01, "learning_rate": 5.101685368654954e-05, "loss": 0.4507, "step": 9971 }, { "epoch": 1.01, "learning_rate": 5.101514587407145e-05, "loss": 1.734, "step": 9972 }, { "epoch": 1.01, "learning_rate": 5.101343792786097e-05, "loss": 0.4456, "step": 9973 }, { "epoch": 1.01, "learning_rate": 5.1011729847928957e-05, "loss": 1.531, "step": 9974 }, { "epoch": 1.01, "learning_rate": 5.101002163428629e-05, "loss": 0.5912, "step": 9975 }, { "epoch": 1.01, "learning_rate": 5.100831328694385e-05, "loss": 1.4128, "step": 9976 }, { "epoch": 1.0, "learning_rate": 5.1006604805912484e-05, "loss": 0.591, "step": 9977 }, { "epoch": 1.0, "learning_rate": 5.1004896191203084e-05, "loss": 0.6218, "step": 9978 }, { "epoch": 1.0, "learning_rate": 5.100318744282651e-05, "loss": 0.7344, "step": 9979 }, { "epoch": 1.0, "learning_rate": 5.100147856079364e-05, "loss": 0.6549, "step": 9980 }, { "epoch": 1.0, "learning_rate": 5.099976954511535e-05, "loss": 0.5124, "step": 9981 }, { "epoch": 1.0, "learning_rate": 5.0998060395802524e-05, "loss": 0.3462, "step": 9982 }, { "epoch": 1.0, "learning_rate": 5.099635111286602e-05, "loss": 0.228, "step": 9983 }, { "epoch": 1.0, "learning_rate": 5.099464169631673e-05, "loss": 0.1761, "step": 9984 }, { "epoch": 1.0, "learning_rate": 5.099293214616552e-05, "loss": 0.7124, "step": 9985 }, { "epoch": 1.0, "learning_rate": 5.0991222462423286e-05, "loss": 0.2299, "step": 9986 }, { "epoch": 1.0, "learning_rate": 5.098951264510089e-05, "loss": 0.4541, "step": 9987 }, { "epoch": 1.0, "learning_rate": 5.098780269420922e-05, "loss": 0.3822, "step": 9988 }, { "epoch": 1.0, "learning_rate": 5.098609260975917e-05, "loss": 0.5111, "step": 9989 }, { "epoch": 1.0, "learning_rate": 5.09843823917616e-05, "loss": 0.881, "step": 9990 }, { "epoch": 1.0, "learning_rate": 5.098267204022741e-05, "loss": 0.5658, "step": 9991 }, { "epoch": 1.0, "learning_rate": 5.098096155516747e-05, "loss": 1.3317, "step": 9992 }, { "epoch": 1.0, "learning_rate": 5.097925093659267e-05, "loss": 0.8239, "step": 9993 }, { "epoch": 1.0, "learning_rate": 5.0977540184513904e-05, "loss": 0.2992, "step": 9994 }, { "epoch": 1.0, "learning_rate": 5.097582929894205e-05, "loss": 0.8668, "step": 9995 }, { "epoch": 1.0, "learning_rate": 5.0974118279888003e-05, "loss": 0.2838, "step": 9996 }, { "epoch": 1.0, "learning_rate": 5.097240712736264e-05, "loss": 0.363, "step": 9997 }, { "epoch": 1.0, "learning_rate": 5.0970695841376864e-05, "loss": 0.9666, "step": 9998 }, { "epoch": 1.0, "learning_rate": 5.0968984421941544e-05, "loss": 0.3689, "step": 9999 }, { "epoch": 1.0, "learning_rate": 5.09672728690676e-05, "loss": 0.5873, "step": 10000 }, { "epoch": 1.0, "learning_rate": 5.09655611827659e-05, "loss": 0.5405, "step": 10001 }, { "epoch": 1.0, "learning_rate": 5.0963849363047336e-05, "loss": 0.2207, "step": 10002 }, { "epoch": 1.0, "learning_rate": 5.096213740992283e-05, "loss": 0.4964, "step": 10003 }, { "epoch": 1.0, "learning_rate": 5.0960425323403226e-05, "loss": 0.4009, "step": 10004 }, { "epoch": 1.0, "learning_rate": 5.095871310349948e-05, "loss": 0.7538, "step": 10005 }, { "epoch": 1.0, "learning_rate": 5.0957000750222435e-05, "loss": 0.6273, "step": 10006 }, { "epoch": 1.0, "learning_rate": 5.095528826358302e-05, "loss": 0.6393, "step": 10007 }, { "epoch": 1.0, "learning_rate": 5.0953575643592125e-05, "loss": 0.2944, "step": 10008 }, { "epoch": 1.0, "learning_rate": 5.0951862890260635e-05, "loss": 1.0878, "step": 10009 }, { "epoch": 1.0, "learning_rate": 5.095015000359947e-05, "loss": 0.9526, "step": 10010 }, { "epoch": 1.0, "learning_rate": 5.094843698361952e-05, "loss": 0.8185, "step": 10011 }, { "epoch": 1.0, "learning_rate": 5.094672383033169e-05, "loss": 0.969, "step": 10012 }, { "epoch": 1.0, "learning_rate": 5.094501054374685e-05, "loss": 0.3704, "step": 10013 }, { "epoch": 1.0, "learning_rate": 5.094329712387596e-05, "loss": 0.3481, "step": 10014 }, { "epoch": 1.0, "learning_rate": 5.094158357072989e-05, "loss": 0.3721, "step": 10015 }, { "epoch": 1.0, "learning_rate": 5.0939869884319546e-05, "loss": 0.8935, "step": 10016 }, { "epoch": 1.0, "learning_rate": 5.093815606465583e-05, "loss": 0.3944, "step": 10017 }, { "epoch": 1.0, "learning_rate": 5.0936442111749664e-05, "loss": 0.6017, "step": 10018 }, { "epoch": 1.0, "learning_rate": 5.0934728025611934e-05, "loss": 0.6502, "step": 10019 }, { "epoch": 1.0, "learning_rate": 5.0933013806253566e-05, "loss": 0.8329, "step": 10020 }, { "epoch": 1.0, "learning_rate": 5.0931299453685465e-05, "loss": 0.2279, "step": 10021 }, { "epoch": 1.0, "learning_rate": 5.092958496791853e-05, "loss": 1.1017, "step": 10022 }, { "epoch": 1.0, "learning_rate": 5.092787034896368e-05, "loss": 0.4913, "step": 10023 }, { "epoch": 1.0, "learning_rate": 5.0926155596831826e-05, "loss": 0.3977, "step": 10024 }, { "epoch": 1.0, "learning_rate": 5.092444071153388e-05, "loss": 0.3856, "step": 10025 }, { "epoch": 1.01, "learning_rate": 5.0922725693080746e-05, "loss": 1.3985, "step": 10026 }, { "epoch": 1.01, "learning_rate": 5.0921010541483345e-05, "loss": 0.2736, "step": 10027 }, { "epoch": 1.01, "learning_rate": 5.0919295256752605e-05, "loss": 1.0319, "step": 10028 }, { "epoch": 1.01, "learning_rate": 5.0917579838899416e-05, "loss": 0.4758, "step": 10029 }, { "epoch": 1.01, "learning_rate": 5.091586428793471e-05, "loss": 0.7653, "step": 10030 }, { "epoch": 1.01, "learning_rate": 5.0914148603869404e-05, "loss": 0.1313, "step": 10031 }, { "epoch": 1.01, "learning_rate": 5.091243278671441e-05, "loss": 0.6156, "step": 10032 }, { "epoch": 1.01, "learning_rate": 5.091071683648064e-05, "loss": 0.9544, "step": 10033 }, { "epoch": 1.01, "learning_rate": 5.090900075317904e-05, "loss": 0.8978, "step": 10034 }, { "epoch": 1.01, "learning_rate": 5.09072845368205e-05, "loss": 0.9273, "step": 10035 }, { "epoch": 1.01, "learning_rate": 5.090556818741596e-05, "loss": 0.4643, "step": 10036 }, { "epoch": 1.01, "learning_rate": 5.0903851704976345e-05, "loss": 0.8113, "step": 10037 }, { "epoch": 1.01, "learning_rate": 5.0902135089512556e-05, "loss": 0.2269, "step": 10038 }, { "epoch": 1.01, "learning_rate": 5.0900418341035537e-05, "loss": 0.7946, "step": 10039 }, { "epoch": 1.01, "learning_rate": 5.0898701459556205e-05, "loss": 0.2674, "step": 10040 }, { "epoch": 1.01, "learning_rate": 5.089698444508549e-05, "loss": 0.6678, "step": 10041 }, { "epoch": 1.01, "learning_rate": 5.08952672976343e-05, "loss": 0.4897, "step": 10042 }, { "epoch": 1.01, "learning_rate": 5.08935500172136e-05, "loss": 0.3919, "step": 10043 }, { "epoch": 1.01, "learning_rate": 5.089183260383429e-05, "loss": 0.3093, "step": 10044 }, { "epoch": 1.01, "learning_rate": 5.08901150575073e-05, "loss": 0.756, "step": 10045 }, { "epoch": 1.01, "learning_rate": 5.088839737824357e-05, "loss": 0.3428, "step": 10046 }, { "epoch": 1.01, "learning_rate": 5.0886679566054024e-05, "loss": 0.322, "step": 10047 }, { "epoch": 1.01, "learning_rate": 5.088496162094959e-05, "loss": 0.7348, "step": 10048 }, { "epoch": 1.01, "learning_rate": 5.0883243542941206e-05, "loss": 0.2929, "step": 10049 }, { "epoch": 1.01, "learning_rate": 5.0881525332039815e-05, "loss": 0.7876, "step": 10050 }, { "epoch": 1.01, "learning_rate": 5.087980698825633e-05, "loss": 0.4651, "step": 10051 }, { "epoch": 1.01, "learning_rate": 5.08780885116017e-05, "loss": 0.2443, "step": 10052 }, { "epoch": 1.01, "learning_rate": 5.0876369902086845e-05, "loss": 0.7783, "step": 10053 }, { "epoch": 1.01, "learning_rate": 5.0874651159722735e-05, "loss": 0.582, "step": 10054 }, { "epoch": 1.01, "learning_rate": 5.087293228452027e-05, "loss": 0.7584, "step": 10055 }, { "epoch": 1.01, "learning_rate": 5.0871213276490416e-05, "loss": 0.5003, "step": 10056 }, { "epoch": 1.01, "learning_rate": 5.0869494135644095e-05, "loss": 1.1308, "step": 10057 }, { "epoch": 1.01, "learning_rate": 5.086777486199225e-05, "loss": 1.0992, "step": 10058 }, { "epoch": 1.01, "learning_rate": 5.086605545554583e-05, "loss": 0.3556, "step": 10059 }, { "epoch": 1.01, "learning_rate": 5.086433591631577e-05, "loss": 0.6386, "step": 10060 }, { "epoch": 1.01, "learning_rate": 5.0862616244313007e-05, "loss": 0.8664, "step": 10061 }, { "epoch": 1.01, "learning_rate": 5.086089643954849e-05, "loss": 0.8306, "step": 10062 }, { "epoch": 1.01, "learning_rate": 5.085917650203318e-05, "loss": 0.7114, "step": 10063 }, { "epoch": 1.01, "learning_rate": 5.085745643177799e-05, "loss": 0.702, "step": 10064 }, { "epoch": 1.01, "learning_rate": 5.0855736228793886e-05, "loss": 0.6478, "step": 10065 }, { "epoch": 1.01, "learning_rate": 5.085401589309181e-05, "loss": 0.8485, "step": 10066 }, { "epoch": 1.01, "learning_rate": 5.085229542468271e-05, "loss": 0.6555, "step": 10067 }, { "epoch": 1.01, "learning_rate": 5.0850574823577534e-05, "loss": 0.133, "step": 10068 }, { "epoch": 1.01, "learning_rate": 5.0848854089787235e-05, "loss": 0.5247, "step": 10069 }, { "epoch": 1.01, "learning_rate": 5.0847133223322764e-05, "loss": 0.0793, "step": 10070 }, { "epoch": 1.01, "learning_rate": 5.0845412224195065e-05, "loss": 0.7157, "step": 10071 }, { "epoch": 1.01, "learning_rate": 5.084369109241509e-05, "loss": 0.8116, "step": 10072 }, { "epoch": 1.01, "learning_rate": 5.0841969827993795e-05, "loss": 0.3707, "step": 10073 }, { "epoch": 1.01, "learning_rate": 5.0840248430942126e-05, "loss": 0.6092, "step": 10074 }, { "epoch": 1.01, "learning_rate": 5.083852690127105e-05, "loss": 1.0852, "step": 10075 }, { "epoch": 1.01, "learning_rate": 5.0836805238991526e-05, "loss": 0.763, "step": 10076 }, { "epoch": 1.01, "learning_rate": 5.083508344411448e-05, "loss": 0.5091, "step": 10077 }, { "epoch": 1.01, "learning_rate": 5.08333615166509e-05, "loss": 0.359, "step": 10078 }, { "epoch": 1.01, "learning_rate": 5.0831639456611734e-05, "loss": 0.2773, "step": 10079 }, { "epoch": 1.01, "learning_rate": 5.0829917264007944e-05, "loss": 0.3257, "step": 10080 }, { "epoch": 1.01, "learning_rate": 5.082819493885047e-05, "loss": 1.0977, "step": 10081 }, { "epoch": 1.01, "learning_rate": 5.08264724811503e-05, "loss": 0.4139, "step": 10082 }, { "epoch": 1.01, "learning_rate": 5.082474989091838e-05, "loss": 0.373, "step": 10083 }, { "epoch": 1.01, "learning_rate": 5.0823027168165664e-05, "loss": 0.4046, "step": 10084 }, { "epoch": 1.01, "learning_rate": 5.082130431290314e-05, "loss": 1.0448, "step": 10085 }, { "epoch": 1.01, "learning_rate": 5.081958132514174e-05, "loss": 0.459, "step": 10086 }, { "epoch": 1.01, "learning_rate": 5.081785820489245e-05, "loss": 0.423, "step": 10087 }, { "epoch": 1.01, "learning_rate": 5.081613495216623e-05, "loss": 0.1563, "step": 10088 }, { "epoch": 1.01, "learning_rate": 5.0814411566974056e-05, "loss": 0.2905, "step": 10089 }, { "epoch": 1.01, "learning_rate": 5.081268804932687e-05, "loss": 0.4489, "step": 10090 }, { "epoch": 1.01, "learning_rate": 5.0810964399235664e-05, "loss": 0.3112, "step": 10091 }, { "epoch": 1.01, "learning_rate": 5.08092406167114e-05, "loss": 0.6607, "step": 10092 }, { "epoch": 1.01, "learning_rate": 5.0807516701765036e-05, "loss": 1.0025, "step": 10093 }, { "epoch": 1.01, "learning_rate": 5.080579265440756e-05, "loss": 0.2369, "step": 10094 }, { "epoch": 1.01, "learning_rate": 5.0804068474649924e-05, "loss": 0.4472, "step": 10095 }, { "epoch": 1.01, "learning_rate": 5.080234416250312e-05, "loss": 0.4706, "step": 10096 }, { "epoch": 1.01, "learning_rate": 5.080061971797811e-05, "loss": 0.1908, "step": 10097 }, { "epoch": 1.01, "learning_rate": 5.079889514108586e-05, "loss": 1.0172, "step": 10098 }, { "epoch": 1.01, "learning_rate": 5.0797170431837366e-05, "loss": 0.4588, "step": 10099 }, { "epoch": 1.01, "learning_rate": 5.079544559024358e-05, "loss": 1.1428, "step": 10100 }, { "epoch": 1.01, "learning_rate": 5.079372061631549e-05, "loss": 0.6937, "step": 10101 }, { "epoch": 1.01, "learning_rate": 5.079199551006409e-05, "loss": 0.5343, "step": 10102 }, { "epoch": 1.01, "learning_rate": 5.079027027150032e-05, "loss": 1.0934, "step": 10103 }, { "epoch": 1.01, "learning_rate": 5.078854490063519e-05, "loss": 1.1579, "step": 10104 }, { "epoch": 1.01, "learning_rate": 5.078681939747967e-05, "loss": 0.3908, "step": 10105 }, { "epoch": 1.01, "learning_rate": 5.078509376204473e-05, "loss": 1.2466, "step": 10106 }, { "epoch": 1.01, "learning_rate": 5.0783367994341364e-05, "loss": 1.3814, "step": 10107 }, { "epoch": 1.01, "learning_rate": 5.078164209438055e-05, "loss": 0.158, "step": 10108 }, { "epoch": 1.01, "learning_rate": 5.077991606217328e-05, "loss": 0.3552, "step": 10109 }, { "epoch": 1.01, "learning_rate": 5.0778189897730516e-05, "loss": 0.6154, "step": 10110 }, { "epoch": 1.01, "learning_rate": 5.077646360106326e-05, "loss": 0.8448, "step": 10111 }, { "epoch": 1.01, "learning_rate": 5.07747371721825e-05, "loss": 0.7085, "step": 10112 }, { "epoch": 1.01, "learning_rate": 5.077301061109921e-05, "loss": 0.4467, "step": 10113 }, { "epoch": 1.01, "learning_rate": 5.077128391782438e-05, "loss": 0.8404, "step": 10114 }, { "epoch": 1.01, "learning_rate": 5.076955709236901e-05, "loss": 0.3927, "step": 10115 }, { "epoch": 1.01, "learning_rate": 5.0767830134744064e-05, "loss": 0.7232, "step": 10116 }, { "epoch": 1.01, "learning_rate": 5.0766103044960566e-05, "loss": 0.3606, "step": 10117 }, { "epoch": 1.01, "learning_rate": 5.076437582302947e-05, "loss": 0.2111, "step": 10118 }, { "epoch": 1.01, "learning_rate": 5.07626484689618e-05, "loss": 0.4691, "step": 10119 }, { "epoch": 1.01, "learning_rate": 5.076092098276853e-05, "loss": 0.9441, "step": 10120 }, { "epoch": 1.01, "learning_rate": 5.075919336446065e-05, "loss": 0.5829, "step": 10121 }, { "epoch": 1.01, "learning_rate": 5.075746561404918e-05, "loss": 0.0021, "step": 10122 }, { "epoch": 1.01, "learning_rate": 5.075573773154508e-05, "loss": 0.2177, "step": 10123 }, { "epoch": 1.02, "learning_rate": 5.0754009716959356e-05, "loss": 0.3777, "step": 10124 }, { "epoch": 1.02, "learning_rate": 5.075228157030302e-05, "loss": 0.4935, "step": 10125 }, { "epoch": 1.02, "learning_rate": 5.075055329158705e-05, "loss": 0.5433, "step": 10126 }, { "epoch": 1.02, "learning_rate": 5.0748824880822464e-05, "loss": 1.4765, "step": 10127 }, { "epoch": 1.02, "learning_rate": 5.0747096338020244e-05, "loss": 0.3681, "step": 10128 }, { "epoch": 1.02, "learning_rate": 5.074536766319139e-05, "loss": 0.6527, "step": 10129 }, { "epoch": 1.02, "learning_rate": 5.074363885634692e-05, "loss": 0.5299, "step": 10130 }, { "epoch": 1.02, "learning_rate": 5.0741909917497815e-05, "loss": 0.5141, "step": 10131 }, { "epoch": 1.02, "learning_rate": 5.074018084665509e-05, "loss": 0.3091, "step": 10132 }, { "epoch": 1.02, "learning_rate": 5.073845164382974e-05, "loss": 0.7562, "step": 10133 }, { "epoch": 1.02, "learning_rate": 5.0736722309032766e-05, "loss": 0.2037, "step": 10134 }, { "epoch": 1.02, "learning_rate": 5.0734992842275196e-05, "loss": 0.81, "step": 10135 }, { "epoch": 1.02, "learning_rate": 5.073326324356801e-05, "loss": 0.2386, "step": 10136 }, { "epoch": 1.02, "learning_rate": 5.073153351292222e-05, "loss": 0.2342, "step": 10137 }, { "epoch": 1.02, "learning_rate": 5.072980365034885e-05, "loss": 0.3002, "step": 10138 }, { "epoch": 1.02, "learning_rate": 5.072807365585889e-05, "loss": 0.5784, "step": 10139 }, { "epoch": 1.02, "learning_rate": 5.0726343529463355e-05, "loss": 0.6272, "step": 10140 }, { "epoch": 1.02, "learning_rate": 5.0724613271173256e-05, "loss": 0.4387, "step": 10141 }, { "epoch": 1.02, "learning_rate": 5.07228828809996e-05, "loss": 0.4251, "step": 10142 }, { "epoch": 1.02, "learning_rate": 5.072115235895339e-05, "loss": 0.7883, "step": 10143 }, { "epoch": 1.02, "learning_rate": 5.071942170504566e-05, "loss": 0.254, "step": 10144 }, { "epoch": 1.02, "learning_rate": 5.0717690919287416e-05, "loss": 0.4601, "step": 10145 }, { "epoch": 1.02, "learning_rate": 5.071596000168966e-05, "loss": 0.7286, "step": 10146 }, { "epoch": 1.02, "learning_rate": 5.071422895226342e-05, "loss": 0.8424, "step": 10147 }, { "epoch": 1.02, "learning_rate": 5.071249777101971e-05, "loss": 0.3765, "step": 10148 }, { "epoch": 1.02, "learning_rate": 5.0710766457969535e-05, "loss": 0.7274, "step": 10149 }, { "epoch": 1.02, "learning_rate": 5.070903501312392e-05, "loss": 0.5758, "step": 10150 }, { "epoch": 1.02, "learning_rate": 5.0707303436493896e-05, "loss": 0.3297, "step": 10151 }, { "epoch": 1.02, "learning_rate": 5.070557172809047e-05, "loss": 0.5537, "step": 10152 }, { "epoch": 1.02, "learning_rate": 5.0703839887924656e-05, "loss": 0.4006, "step": 10153 }, { "epoch": 1.02, "learning_rate": 5.070210791600748e-05, "loss": 0.8023, "step": 10154 }, { "epoch": 1.02, "learning_rate": 5.070037581234997e-05, "loss": 0.1171, "step": 10155 }, { "epoch": 1.02, "learning_rate": 5.069864357696313e-05, "loss": 0.4394, "step": 10156 }, { "epoch": 1.02, "learning_rate": 5.069691120985801e-05, "loss": 1.0676, "step": 10157 }, { "epoch": 1.02, "learning_rate": 5.0695178711045625e-05, "loss": 0.5881, "step": 10158 }, { "epoch": 1.02, "learning_rate": 5.069344608053698e-05, "loss": 0.5611, "step": 10159 }, { "epoch": 1.02, "learning_rate": 5.0691713318343134e-05, "loss": 0.4503, "step": 10160 }, { "epoch": 1.02, "learning_rate": 5.068998042447508e-05, "loss": 0.512, "step": 10161 }, { "epoch": 1.02, "learning_rate": 5.068824739894387e-05, "loss": 0.9037, "step": 10162 }, { "epoch": 1.02, "learning_rate": 5.0686514241760525e-05, "loss": 1.2131, "step": 10163 }, { "epoch": 1.02, "learning_rate": 5.068478095293606e-05, "loss": 0.6347, "step": 10164 }, { "epoch": 1.02, "learning_rate": 5.0683047532481534e-05, "loss": 0.8507, "step": 10165 }, { "epoch": 1.02, "learning_rate": 5.068131398040795e-05, "loss": 0.1671, "step": 10166 }, { "epoch": 1.02, "learning_rate": 5.067958029672636e-05, "loss": 0.4552, "step": 10167 }, { "epoch": 1.02, "learning_rate": 5.067784648144778e-05, "loss": 1.06, "step": 10168 }, { "epoch": 1.02, "learning_rate": 5.0676112534583256e-05, "loss": 0.253, "step": 10169 }, { "epoch": 1.02, "learning_rate": 5.0674378456143814e-05, "loss": 0.2595, "step": 10170 }, { "epoch": 1.02, "learning_rate": 5.067264424614049e-05, "loss": 0.6038, "step": 10171 }, { "epoch": 1.02, "learning_rate": 5.067090990458434e-05, "loss": 0.3082, "step": 10172 }, { "epoch": 1.02, "learning_rate": 5.066917543148636e-05, "loss": 0.2636, "step": 10173 }, { "epoch": 1.02, "learning_rate": 5.066744082685763e-05, "loss": 0.2718, "step": 10174 }, { "epoch": 1.02, "learning_rate": 5.066570609070915e-05, "loss": 0.5854, "step": 10175 }, { "epoch": 1.02, "learning_rate": 5.0663971223051994e-05, "loss": 0.3889, "step": 10176 }, { "epoch": 1.02, "learning_rate": 5.066223622389718e-05, "loss": 0.7633, "step": 10177 }, { "epoch": 1.02, "learning_rate": 5.066050109325575e-05, "loss": 0.4341, "step": 10178 }, { "epoch": 1.02, "learning_rate": 5.065876583113876e-05, "loss": 0.7809, "step": 10179 }, { "epoch": 1.02, "learning_rate": 5.065703043755723e-05, "loss": 0.3464, "step": 10180 }, { "epoch": 1.02, "learning_rate": 5.0655294912522234e-05, "loss": 0.5887, "step": 10181 }, { "epoch": 1.02, "learning_rate": 5.065355925604479e-05, "loss": 0.4433, "step": 10182 }, { "epoch": 1.02, "learning_rate": 5.065182346813595e-05, "loss": 0.7175, "step": 10183 }, { "epoch": 1.02, "learning_rate": 5.065008754880677e-05, "loss": 0.2528, "step": 10184 }, { "epoch": 1.02, "learning_rate": 5.0648351498068275e-05, "loss": 1.2162, "step": 10185 }, { "epoch": 1.02, "learning_rate": 5.064661531593154e-05, "loss": 0.1739, "step": 10186 }, { "epoch": 1.02, "learning_rate": 5.064487900240759e-05, "loss": 1.035, "step": 10187 }, { "epoch": 1.02, "learning_rate": 5.064314255750749e-05, "loss": 0.8564, "step": 10188 }, { "epoch": 1.02, "learning_rate": 5.0641405981242275e-05, "loss": 0.5293, "step": 10189 }, { "epoch": 1.02, "learning_rate": 5.0639669273623014e-05, "loss": 0.4415, "step": 10190 }, { "epoch": 1.02, "learning_rate": 5.063793243466074e-05, "loss": 0.3849, "step": 10191 }, { "epoch": 1.02, "learning_rate": 5.063619546436653e-05, "loss": 1.0432, "step": 10192 }, { "epoch": 1.02, "learning_rate": 5.063445836275141e-05, "loss": 0.8654, "step": 10193 }, { "epoch": 1.02, "learning_rate": 5.0632721129826445e-05, "loss": 0.8203, "step": 10194 }, { "epoch": 1.02, "learning_rate": 5.06309837656027e-05, "loss": 0.6858, "step": 10195 }, { "epoch": 1.02, "learning_rate": 5.062924627009122e-05, "loss": 0.5771, "step": 10196 }, { "epoch": 1.02, "learning_rate": 5.062750864330306e-05, "loss": 0.4082, "step": 10197 }, { "epoch": 1.02, "learning_rate": 5.062577088524929e-05, "loss": 0.4818, "step": 10198 }, { "epoch": 1.02, "learning_rate": 5.062403299594096e-05, "loss": 1.0309, "step": 10199 }, { "epoch": 1.02, "learning_rate": 5.0622294975389123e-05, "loss": 0.5321, "step": 10200 }, { "epoch": 1.02, "learning_rate": 5.0620556823604843e-05, "loss": 0.368, "step": 10201 }, { "epoch": 1.02, "learning_rate": 5.0618818540599195e-05, "loss": 0.1661, "step": 10202 }, { "epoch": 1.02, "learning_rate": 5.0617080126383223e-05, "loss": 1.0948, "step": 10203 }, { "epoch": 1.02, "learning_rate": 5.061534158096801e-05, "loss": 0.578, "step": 10204 }, { "epoch": 1.02, "learning_rate": 5.061360290436458e-05, "loss": 0.8261, "step": 10205 }, { "epoch": 1.02, "learning_rate": 5.061186409658405e-05, "loss": 0.6574, "step": 10206 }, { "epoch": 1.02, "learning_rate": 5.061012515763744e-05, "loss": 0.2241, "step": 10207 }, { "epoch": 1.02, "learning_rate": 5.060838608753583e-05, "loss": 0.3823, "step": 10208 }, { "epoch": 1.02, "learning_rate": 5.0606646886290305e-05, "loss": 0.6246, "step": 10209 }, { "epoch": 1.02, "learning_rate": 5.060490755391191e-05, "loss": 0.6068, "step": 10210 }, { "epoch": 1.02, "learning_rate": 5.0603168090411724e-05, "loss": 0.9702, "step": 10211 }, { "epoch": 1.02, "learning_rate": 5.060142849580082e-05, "loss": 1.1419, "step": 10212 }, { "epoch": 1.02, "learning_rate": 5.059968877009026e-05, "loss": 0.2074, "step": 10213 }, { "epoch": 1.02, "learning_rate": 5.059794891329111e-05, "loss": 0.5106, "step": 10214 }, { "epoch": 1.02, "learning_rate": 5.059620892541445e-05, "loss": 0.3434, "step": 10215 }, { "epoch": 1.02, "learning_rate": 5.059446880647136e-05, "loss": 0.4384, "step": 10216 }, { "epoch": 1.02, "learning_rate": 5.0592728556472895e-05, "loss": 0.2835, "step": 10217 }, { "epoch": 1.02, "learning_rate": 5.059098817543015e-05, "loss": 1.0224, "step": 10218 }, { "epoch": 1.02, "learning_rate": 5.058924766335419e-05, "loss": 0.4721, "step": 10219 }, { "epoch": 1.02, "learning_rate": 5.058750702025608e-05, "loss": 0.2663, "step": 10220 }, { "epoch": 1.02, "learning_rate": 5.0585766246146914e-05, "loss": 0.9152, "step": 10221 }, { "epoch": 1.02, "learning_rate": 5.058402534103776e-05, "loss": 0.308, "step": 10222 }, { "epoch": 1.03, "learning_rate": 5.0582284304939696e-05, "loss": 0.3882, "step": 10223 }, { "epoch": 1.03, "learning_rate": 5.0580543137863806e-05, "loss": 0.563, "step": 10224 }, { "epoch": 1.03, "learning_rate": 5.0578801839821174e-05, "loss": 0.5876, "step": 10225 }, { "epoch": 1.03, "learning_rate": 5.057706041082287e-05, "loss": 0.8034, "step": 10226 }, { "epoch": 1.03, "learning_rate": 5.057531885087999e-05, "loss": 0.3528, "step": 10227 }, { "epoch": 1.03, "learning_rate": 5.05735771600036e-05, "loss": 0.2807, "step": 10228 }, { "epoch": 1.03, "learning_rate": 5.057183533820479e-05, "loss": 0.8577, "step": 10229 }, { "epoch": 1.03, "learning_rate": 5.057009338549465e-05, "loss": 0.4081, "step": 10230 }, { "epoch": 1.03, "learning_rate": 5.0568351301884265e-05, "loss": 1.0542, "step": 10231 }, { "epoch": 1.03, "learning_rate": 5.056660908738471e-05, "loss": 0.4256, "step": 10232 }, { "epoch": 1.03, "learning_rate": 5.056486674200708e-05, "loss": 0.672, "step": 10233 }, { "epoch": 1.03, "learning_rate": 5.056312426576246e-05, "loss": 0.3157, "step": 10234 }, { "epoch": 1.03, "learning_rate": 5.056138165866194e-05, "loss": 0.6333, "step": 10235 }, { "epoch": 1.03, "learning_rate": 5.0559638920716615e-05, "loss": 1.0913, "step": 10236 }, { "epoch": 1.03, "learning_rate": 5.055789605193756e-05, "loss": 0.3652, "step": 10237 }, { "epoch": 1.03, "learning_rate": 5.055615305233589e-05, "loss": 0.2342, "step": 10238 }, { "epoch": 1.03, "learning_rate": 5.055440992192267e-05, "loss": 1.0344, "step": 10239 }, { "epoch": 1.03, "learning_rate": 5.0552666660709e-05, "loss": 0.5826, "step": 10240 }, { "epoch": 1.03, "learning_rate": 5.055092326870599e-05, "loss": 0.4049, "step": 10241 }, { "epoch": 1.03, "learning_rate": 5.0549179745924714e-05, "loss": 0.7971, "step": 10242 }, { "epoch": 1.03, "learning_rate": 5.054743609237629e-05, "loss": 1.142, "step": 10243 }, { "epoch": 1.03, "learning_rate": 5.054569230807178e-05, "loss": 1.1656, "step": 10244 }, { "epoch": 1.03, "learning_rate": 5.054394839302231e-05, "loss": 0.2325, "step": 10245 }, { "epoch": 1.03, "learning_rate": 5.054220434723896e-05, "loss": 1.086, "step": 10246 }, { "epoch": 1.03, "learning_rate": 5.0540460170732845e-05, "loss": 0.7537, "step": 10247 }, { "epoch": 1.03, "learning_rate": 5.0538715863515055e-05, "loss": 0.5937, "step": 10248 }, { "epoch": 1.03, "learning_rate": 5.0536971425596686e-05, "loss": 1.0458, "step": 10249 }, { "epoch": 1.03, "learning_rate": 5.0535226856988845e-05, "loss": 0.7467, "step": 10250 }, { "epoch": 1.03, "learning_rate": 5.053348215770264e-05, "loss": 0.5183, "step": 10251 }, { "epoch": 1.03, "learning_rate": 5.053173732774916e-05, "loss": 0.3886, "step": 10252 }, { "epoch": 1.03, "learning_rate": 5.052999236713951e-05, "loss": 0.8589, "step": 10253 }, { "epoch": 1.03, "learning_rate": 5.05282472758848e-05, "loss": 0.1291, "step": 10254 }, { "epoch": 1.03, "learning_rate": 5.052650205399614e-05, "loss": 0.8201, "step": 10255 }, { "epoch": 1.03, "learning_rate": 5.052475670148462e-05, "loss": 0.4659, "step": 10256 }, { "epoch": 1.03, "learning_rate": 5.052301121836136e-05, "loss": 0.436, "step": 10257 }, { "epoch": 1.03, "learning_rate": 5.052126560463747e-05, "loss": 0.27, "step": 10258 }, { "epoch": 1.03, "learning_rate": 5.051951986032404e-05, "loss": 0.4934, "step": 10259 }, { "epoch": 1.03, "learning_rate": 5.05177739854322e-05, "loss": 1.0503, "step": 10260 }, { "epoch": 1.03, "learning_rate": 5.051602797997306e-05, "loss": 0.7446, "step": 10261 }, { "epoch": 1.03, "learning_rate": 5.051428184395771e-05, "loss": 0.3243, "step": 10262 }, { "epoch": 1.03, "learning_rate": 5.051253557739728e-05, "loss": 0.4549, "step": 10263 }, { "epoch": 1.03, "learning_rate": 5.0510789180302876e-05, "loss": 0.6121, "step": 10264 }, { "epoch": 1.03, "learning_rate": 5.050904265268561e-05, "loss": 0.324, "step": 10265 }, { "epoch": 1.03, "learning_rate": 5.05072959945566e-05, "loss": 0.2267, "step": 10266 }, { "epoch": 1.03, "learning_rate": 5.050554920592696e-05, "loss": 0.8707, "step": 10267 }, { "epoch": 1.03, "learning_rate": 5.0503802286807804e-05, "loss": 0.7329, "step": 10268 }, { "epoch": 1.03, "learning_rate": 5.050205523721025e-05, "loss": 0.2611, "step": 10269 }, { "epoch": 1.03, "learning_rate": 5.0500308057145413e-05, "loss": 0.2458, "step": 10270 }, { "epoch": 1.03, "learning_rate": 5.0498560746624424e-05, "loss": 0.6123, "step": 10271 }, { "epoch": 1.03, "learning_rate": 5.049681330565839e-05, "loss": 0.4566, "step": 10272 }, { "epoch": 1.03, "learning_rate": 5.0495065734258426e-05, "loss": 0.4486, "step": 10273 }, { "epoch": 1.03, "learning_rate": 5.049331803243567e-05, "loss": 0.5308, "step": 10274 }, { "epoch": 1.03, "learning_rate": 5.0491570200201225e-05, "loss": 0.6848, "step": 10275 }, { "epoch": 1.03, "learning_rate": 5.0489822237566225e-05, "loss": 0.3961, "step": 10276 }, { "epoch": 1.03, "learning_rate": 5.04880741445418e-05, "loss": 0.1836, "step": 10277 }, { "epoch": 1.03, "learning_rate": 5.048632592113905e-05, "loss": 0.3256, "step": 10278 }, { "epoch": 1.03, "learning_rate": 5.048457756736913e-05, "loss": 0.8988, "step": 10279 }, { "epoch": 1.03, "learning_rate": 5.048282908324314e-05, "loss": 0.1426, "step": 10280 }, { "epoch": 1.03, "learning_rate": 5.048108046877222e-05, "loss": 1.1225, "step": 10281 }, { "epoch": 1.03, "learning_rate": 5.04793317239675e-05, "loss": 0.3264, "step": 10282 }, { "epoch": 1.03, "learning_rate": 5.0477582848840106e-05, "loss": 1.2936, "step": 10283 }, { "epoch": 1.03, "learning_rate": 5.047583384340116e-05, "loss": 0.3126, "step": 10284 }, { "epoch": 1.03, "learning_rate": 5.047408470766179e-05, "loss": 0.9387, "step": 10285 }, { "epoch": 1.03, "learning_rate": 5.0472335441633136e-05, "loss": 0.9754, "step": 10286 }, { "epoch": 1.03, "learning_rate": 5.047058604532634e-05, "loss": 0.6865, "step": 10287 }, { "epoch": 1.03, "learning_rate": 5.046883651875251e-05, "loss": 0.7369, "step": 10288 }, { "epoch": 1.03, "learning_rate": 5.0467086861922795e-05, "loss": 0.9364, "step": 10289 }, { "epoch": 1.03, "learning_rate": 5.0465337074848325e-05, "loss": 0.2679, "step": 10290 }, { "epoch": 1.03, "learning_rate": 5.0463587157540234e-05, "loss": 0.7892, "step": 10291 }, { "epoch": 1.03, "learning_rate": 5.0461837110009664e-05, "loss": 0.4701, "step": 10292 }, { "epoch": 1.03, "learning_rate": 5.046008693226774e-05, "loss": 0.2721, "step": 10293 }, { "epoch": 1.03, "learning_rate": 5.045833662432561e-05, "loss": 1.109, "step": 10294 }, { "epoch": 1.03, "learning_rate": 5.0456586186194405e-05, "loss": 0.6354, "step": 10295 }, { "epoch": 1.03, "learning_rate": 5.045483561788527e-05, "loss": 0.1856, "step": 10296 }, { "epoch": 1.03, "learning_rate": 5.045308491940935e-05, "loss": 0.3809, "step": 10297 }, { "epoch": 1.03, "learning_rate": 5.045133409077777e-05, "loss": 0.8119, "step": 10298 }, { "epoch": 1.03, "learning_rate": 5.044958313200168e-05, "loss": 0.444, "step": 10299 }, { "epoch": 1.03, "learning_rate": 5.044783204309222e-05, "loss": 0.3898, "step": 10300 }, { "epoch": 1.03, "learning_rate": 5.0446080824060536e-05, "loss": 0.1918, "step": 10301 }, { "epoch": 1.03, "learning_rate": 5.0444329474917774e-05, "loss": 0.9976, "step": 10302 }, { "epoch": 1.03, "learning_rate": 5.044257799567508e-05, "loss": 0.4958, "step": 10303 }, { "epoch": 1.03, "learning_rate": 5.04408263863436e-05, "loss": 0.2457, "step": 10304 }, { "epoch": 1.03, "learning_rate": 5.043907464693447e-05, "loss": 0.4215, "step": 10305 }, { "epoch": 1.03, "learning_rate": 5.0437322777458846e-05, "loss": 0.5914, "step": 10306 }, { "epoch": 1.03, "learning_rate": 5.043557077792787e-05, "loss": 0.4312, "step": 10307 }, { "epoch": 1.03, "learning_rate": 5.04338186483527e-05, "loss": 0.4737, "step": 10308 }, { "epoch": 1.03, "learning_rate": 5.043206638874448e-05, "loss": 1.3769, "step": 10309 }, { "epoch": 1.03, "learning_rate": 5.0430313999114365e-05, "loss": 0.3462, "step": 10310 }, { "epoch": 1.03, "learning_rate": 5.0428561479473496e-05, "loss": 0.5189, "step": 10311 }, { "epoch": 1.03, "learning_rate": 5.042680882983304e-05, "loss": 1.0464, "step": 10312 }, { "epoch": 1.03, "learning_rate": 5.042505605020414e-05, "loss": 0.8318, "step": 10313 }, { "epoch": 1.03, "learning_rate": 5.042330314059796e-05, "loss": 1.0021, "step": 10314 }, { "epoch": 1.03, "learning_rate": 5.042155010102564e-05, "loss": 0.6883, "step": 10315 }, { "epoch": 1.03, "learning_rate": 5.041979693149835e-05, "loss": 0.5571, "step": 10316 }, { "epoch": 1.03, "learning_rate": 5.041804363202724e-05, "loss": 0.988, "step": 10317 }, { "epoch": 1.03, "learning_rate": 5.041629020262347e-05, "loss": 0.2588, "step": 10318 }, { "epoch": 1.03, "learning_rate": 5.041453664329819e-05, "loss": 1.0105, "step": 10319 }, { "epoch": 1.03, "learning_rate": 5.0412782954062564e-05, "loss": 0.2047, "step": 10320 }, { "epoch": 1.04, "learning_rate": 5.041102913492776e-05, "loss": 0.493, "step": 10321 }, { "epoch": 1.04, "learning_rate": 5.040927518590493e-05, "loss": 0.3234, "step": 10322 }, { "epoch": 1.04, "learning_rate": 5.0407521107005224e-05, "loss": 0.8206, "step": 10323 }, { "epoch": 1.04, "learning_rate": 5.0405766898239834e-05, "loss": 0.5055, "step": 10324 }, { "epoch": 1.04, "learning_rate": 5.0404012559619896e-05, "loss": 0.8629, "step": 10325 }, { "epoch": 1.04, "learning_rate": 5.040225809115658e-05, "loss": 0.246, "step": 10326 }, { "epoch": 1.04, "learning_rate": 5.040050349286107e-05, "loss": 0.9106, "step": 10327 }, { "epoch": 1.04, "learning_rate": 5.039874876474451e-05, "loss": 0.6678, "step": 10328 }, { "epoch": 1.04, "learning_rate": 5.039699390681807e-05, "loss": 0.4568, "step": 10329 }, { "epoch": 1.04, "learning_rate": 5.0395238919092926e-05, "loss": 0.2327, "step": 10330 }, { "epoch": 1.04, "learning_rate": 5.039348380158023e-05, "loss": 0.5546, "step": 10331 }, { "epoch": 1.04, "learning_rate": 5.0391728554291175e-05, "loss": 0.4553, "step": 10332 }, { "epoch": 1.04, "learning_rate": 5.0389973177236904e-05, "loss": 0.7332, "step": 10333 }, { "epoch": 1.04, "learning_rate": 5.0388217670428603e-05, "loss": 0.6485, "step": 10334 }, { "epoch": 1.04, "learning_rate": 5.038646203387745e-05, "loss": 0.5464, "step": 10335 }, { "epoch": 1.04, "learning_rate": 5.0384706267594597e-05, "loss": 0.2304, "step": 10336 }, { "epoch": 1.04, "learning_rate": 5.038295037159123e-05, "loss": 0.8039, "step": 10337 }, { "epoch": 1.04, "learning_rate": 5.0381194345878526e-05, "loss": 0.1913, "step": 10338 }, { "epoch": 1.04, "learning_rate": 5.0379438190467655e-05, "loss": 0.2989, "step": 10339 }, { "epoch": 1.04, "learning_rate": 5.0377681905369785e-05, "loss": 0.8582, "step": 10340 }, { "epoch": 1.04, "learning_rate": 5.0375925490596104e-05, "loss": 0.3412, "step": 10341 }, { "epoch": 1.04, "learning_rate": 5.0374168946157786e-05, "loss": 0.6572, "step": 10342 }, { "epoch": 1.04, "learning_rate": 5.0372412272066e-05, "loss": 0.5232, "step": 10343 }, { "epoch": 1.04, "learning_rate": 5.037065546833194e-05, "loss": 0.3945, "step": 10344 }, { "epoch": 1.04, "learning_rate": 5.036889853496677e-05, "loss": 0.278, "step": 10345 }, { "epoch": 1.04, "learning_rate": 5.036714147198169e-05, "loss": 0.78, "step": 10346 }, { "epoch": 1.04, "learning_rate": 5.036538427938787e-05, "loss": 1.0641, "step": 10347 }, { "epoch": 1.04, "learning_rate": 5.0363626957196485e-05, "loss": 0.4555, "step": 10348 }, { "epoch": 1.04, "learning_rate": 5.0361869505418724e-05, "loss": 0.682, "step": 10349 }, { "epoch": 1.04, "learning_rate": 5.036011192406577e-05, "loss": 0.6749, "step": 10350 }, { "epoch": 1.04, "learning_rate": 5.035835421314881e-05, "loss": 0.4365, "step": 10351 }, { "epoch": 1.04, "learning_rate": 5.035659637267904e-05, "loss": 0.57, "step": 10352 }, { "epoch": 1.04, "learning_rate": 5.035483840266762e-05, "loss": 0.1148, "step": 10353 }, { "epoch": 1.04, "learning_rate": 5.035308030312575e-05, "loss": 1.0775, "step": 10354 }, { "epoch": 1.04, "learning_rate": 5.0351322074064636e-05, "loss": 0.5804, "step": 10355 }, { "epoch": 1.04, "learning_rate": 5.034956371549544e-05, "loss": 0.4529, "step": 10356 }, { "epoch": 1.04, "learning_rate": 5.034780522742937e-05, "loss": 0.4458, "step": 10357 }, { "epoch": 1.04, "learning_rate": 5.0346046609877594e-05, "loss": 0.5786, "step": 10358 }, { "epoch": 1.04, "learning_rate": 5.034428786285133e-05, "loss": 0.1756, "step": 10359 }, { "epoch": 1.04, "learning_rate": 5.034252898636175e-05, "loss": 0.5859, "step": 10360 }, { "epoch": 1.04, "learning_rate": 5.0340769980420056e-05, "loss": 0.0872, "step": 10361 }, { "epoch": 1.04, "learning_rate": 5.033901084503744e-05, "loss": 0.849, "step": 10362 }, { "epoch": 1.04, "learning_rate": 5.03372515802251e-05, "loss": 0.6316, "step": 10363 }, { "epoch": 1.04, "learning_rate": 5.033549218599422e-05, "loss": 0.4374, "step": 10364 }, { "epoch": 1.04, "learning_rate": 5.033373266235601e-05, "loss": 0.5916, "step": 10365 }, { "epoch": 1.04, "learning_rate": 5.0331973009321657e-05, "loss": 0.9598, "step": 10366 }, { "epoch": 1.04, "learning_rate": 5.033021322690236e-05, "loss": 1.0637, "step": 10367 }, { "epoch": 1.04, "learning_rate": 5.032845331510933e-05, "loss": 1.0161, "step": 10368 }, { "epoch": 1.04, "learning_rate": 5.032669327395375e-05, "loss": 0.6925, "step": 10369 }, { "epoch": 1.04, "learning_rate": 5.032493310344683e-05, "loss": 0.3385, "step": 10370 }, { "epoch": 1.04, "learning_rate": 5.0323172803599764e-05, "loss": 0.3529, "step": 10371 }, { "epoch": 1.04, "learning_rate": 5.032141237442376e-05, "loss": 0.8665, "step": 10372 }, { "epoch": 1.04, "learning_rate": 5.031965181593001e-05, "loss": 0.5831, "step": 10373 }, { "epoch": 1.04, "learning_rate": 5.0317891128129736e-05, "loss": 0.6963, "step": 10374 }, { "epoch": 1.04, "learning_rate": 5.031613031103413e-05, "loss": 0.4591, "step": 10375 }, { "epoch": 1.04, "learning_rate": 5.0314369364654395e-05, "loss": 0.1708, "step": 10376 }, { "epoch": 1.04, "learning_rate": 5.0312608289001745e-05, "loss": 0.5352, "step": 10377 }, { "epoch": 1.04, "learning_rate": 5.031084708408739e-05, "loss": 0.4121, "step": 10378 }, { "epoch": 1.04, "learning_rate": 5.030908574992252e-05, "loss": 0.6844, "step": 10379 }, { "epoch": 1.04, "learning_rate": 5.030732428651835e-05, "loss": 0.5823, "step": 10380 }, { "epoch": 1.04, "learning_rate": 5.030556269388611e-05, "loss": 0.1917, "step": 10381 }, { "epoch": 1.04, "learning_rate": 5.030380097203698e-05, "loss": 1.4822, "step": 10382 }, { "epoch": 1.04, "learning_rate": 5.030203912098219e-05, "loss": 0.2486, "step": 10383 }, { "epoch": 1.04, "learning_rate": 5.030027714073295e-05, "loss": 1.6959, "step": 10384 }, { "epoch": 1.04, "learning_rate": 5.029851503130046e-05, "loss": 1.0296, "step": 10385 }, { "epoch": 1.04, "learning_rate": 5.029675279269596e-05, "loss": 0.7319, "step": 10386 }, { "epoch": 1.04, "learning_rate": 5.029499042493063e-05, "loss": 0.4674, "step": 10387 }, { "epoch": 1.04, "learning_rate": 5.0293227928015696e-05, "loss": 1.1336, "step": 10388 }, { "epoch": 1.04, "learning_rate": 5.0291465301962396e-05, "loss": 0.7222, "step": 10389 }, { "epoch": 1.04, "learning_rate": 5.028970254678192e-05, "loss": 0.7988, "step": 10390 }, { "epoch": 1.04, "learning_rate": 5.02879396624855e-05, "loss": 0.6473, "step": 10391 }, { "epoch": 1.04, "learning_rate": 5.028617664908435e-05, "loss": 0.1734, "step": 10392 }, { "epoch": 1.04, "learning_rate": 5.028441350658968e-05, "loss": 0.3634, "step": 10393 }, { "epoch": 1.04, "learning_rate": 5.0282650235012736e-05, "loss": 0.6701, "step": 10394 }, { "epoch": 1.04, "learning_rate": 5.028088683436471e-05, "loss": 0.8428, "step": 10395 }, { "epoch": 1.04, "learning_rate": 5.0279123304656834e-05, "loss": 1.3539, "step": 10396 }, { "epoch": 1.04, "learning_rate": 5.027735964590034e-05, "loss": 0.7585, "step": 10397 }, { "epoch": 1.04, "learning_rate": 5.0275595858106436e-05, "loss": 0.2514, "step": 10398 }, { "epoch": 1.04, "learning_rate": 5.027383194128635e-05, "loss": 0.4513, "step": 10399 }, { "epoch": 1.04, "learning_rate": 5.027206789545132e-05, "loss": 0.5555, "step": 10400 }, { "epoch": 1.04, "learning_rate": 5.027030372061256e-05, "loss": 0.7958, "step": 10401 }, { "epoch": 1.04, "learning_rate": 5.0268539416781296e-05, "loss": 0.7481, "step": 10402 }, { "epoch": 1.04, "learning_rate": 5.026677498396876e-05, "loss": 0.5977, "step": 10403 }, { "epoch": 1.04, "learning_rate": 5.026501042218617e-05, "loss": 1.0214, "step": 10404 }, { "epoch": 1.04, "learning_rate": 5.0263245731444764e-05, "loss": 1.2057, "step": 10405 }, { "epoch": 1.04, "learning_rate": 5.026148091175578e-05, "loss": 0.5556, "step": 10406 }, { "epoch": 1.04, "learning_rate": 5.025971596313044e-05, "loss": 0.3752, "step": 10407 }, { "epoch": 1.04, "learning_rate": 5.025795088557997e-05, "loss": 0.4961, "step": 10408 }, { "epoch": 1.04, "learning_rate": 5.02561856791156e-05, "loss": 0.4864, "step": 10409 }, { "epoch": 1.04, "learning_rate": 5.025442034374858e-05, "loss": 0.597, "step": 10410 }, { "epoch": 1.04, "learning_rate": 5.0252654879490133e-05, "loss": 0.453, "step": 10411 }, { "epoch": 1.04, "learning_rate": 5.025088928635149e-05, "loss": 0.3445, "step": 10412 }, { "epoch": 1.04, "learning_rate": 5.024912356434389e-05, "loss": 0.6791, "step": 10413 }, { "epoch": 1.04, "learning_rate": 5.0247357713478584e-05, "loss": 1.019, "step": 10414 }, { "epoch": 1.04, "learning_rate": 5.0245591733766774e-05, "loss": 0.9432, "step": 10415 }, { "epoch": 1.04, "learning_rate": 5.024382562521974e-05, "loss": 0.847, "step": 10416 }, { "epoch": 1.04, "learning_rate": 5.0242059387848695e-05, "loss": 0.2165, "step": 10417 }, { "epoch": 1.04, "learning_rate": 5.024029302166488e-05, "loss": 0.9091, "step": 10418 }, { "epoch": 1.04, "learning_rate": 5.0238526526679545e-05, "loss": 1.0303, "step": 10419 }, { "epoch": 1.05, "learning_rate": 5.0236759902903926e-05, "loss": 0.205, "step": 10420 }, { "epoch": 1.05, "learning_rate": 5.0234993150349266e-05, "loss": 1.1179, "step": 10421 }, { "epoch": 1.05, "learning_rate": 5.02332262690268e-05, "loss": 0.5659, "step": 10422 }, { "epoch": 1.05, "learning_rate": 5.023145925894779e-05, "loss": 0.401, "step": 10423 }, { "epoch": 1.05, "learning_rate": 5.022969212012346e-05, "loss": 0.2808, "step": 10424 }, { "epoch": 1.05, "learning_rate": 5.022792485256507e-05, "loss": 1.1608, "step": 10425 }, { "epoch": 1.05, "learning_rate": 5.022615745628386e-05, "loss": 0.4096, "step": 10426 }, { "epoch": 1.05, "learning_rate": 5.022438993129108e-05, "loss": 0.7042, "step": 10427 }, { "epoch": 1.05, "learning_rate": 5.0222622277597975e-05, "loss": 0.7149, "step": 10428 }, { "epoch": 1.05, "learning_rate": 5.02208544952158e-05, "loss": 1.1236, "step": 10429 }, { "epoch": 1.05, "learning_rate": 5.0219086584155794e-05, "loss": 0.4129, "step": 10430 }, { "epoch": 1.05, "learning_rate": 5.021731854442921e-05, "loss": 0.384, "step": 10431 }, { "epoch": 1.05, "learning_rate": 5.02155503760473e-05, "loss": 0.4483, "step": 10432 }, { "epoch": 1.05, "learning_rate": 5.021378207902133e-05, "loss": 0.3687, "step": 10433 }, { "epoch": 1.05, "learning_rate": 5.021201365336253e-05, "loss": 0.6542, "step": 10434 }, { "epoch": 1.05, "learning_rate": 5.0210245099082166e-05, "loss": 0.2706, "step": 10435 }, { "epoch": 1.05, "learning_rate": 5.0208476416191495e-05, "loss": 0.8705, "step": 10436 }, { "epoch": 1.05, "learning_rate": 5.020670760470177e-05, "loss": 0.9053, "step": 10437 }, { "epoch": 1.05, "learning_rate": 5.020493866462423e-05, "loss": 0.9306, "step": 10438 }, { "epoch": 1.05, "learning_rate": 5.020316959597016e-05, "loss": 0.2036, "step": 10439 }, { "epoch": 1.05, "learning_rate": 5.0201400398750806e-05, "loss": 1.2977, "step": 10440 }, { "epoch": 1.05, "learning_rate": 5.0199631072977416e-05, "loss": 0.7169, "step": 10441 }, { "epoch": 1.05, "learning_rate": 5.0197861618661266e-05, "loss": 0.8363, "step": 10442 }, { "epoch": 1.05, "learning_rate": 5.01960920358136e-05, "loss": 1.032, "step": 10443 }, { "epoch": 1.05, "learning_rate": 5.019432232444569e-05, "loss": 0.2934, "step": 10444 }, { "epoch": 1.05, "learning_rate": 5.01925524845688e-05, "loss": 0.1846, "step": 10445 }, { "epoch": 1.05, "learning_rate": 5.0190782516194175e-05, "loss": 0.6609, "step": 10446 }, { "epoch": 1.05, "learning_rate": 5.01890124193331e-05, "loss": 0.7334, "step": 10447 }, { "epoch": 1.05, "learning_rate": 5.018724219399684e-05, "loss": 0.6791, "step": 10448 }, { "epoch": 1.05, "learning_rate": 5.018547184019664e-05, "loss": 0.2289, "step": 10449 }, { "epoch": 1.05, "learning_rate": 5.018370135794377e-05, "loss": 0.6751, "step": 10450 }, { "epoch": 1.05, "learning_rate": 5.018193074724951e-05, "loss": 0.2467, "step": 10451 }, { "epoch": 1.05, "learning_rate": 5.018016000812513e-05, "loss": 1.1761, "step": 10452 }, { "epoch": 1.05, "learning_rate": 5.017838914058187e-05, "loss": 0.4193, "step": 10453 }, { "epoch": 1.05, "learning_rate": 5.017661814463103e-05, "loss": 0.3672, "step": 10454 }, { "epoch": 1.05, "learning_rate": 5.0174847020283865e-05, "loss": 1.17, "step": 10455 }, { "epoch": 1.05, "learning_rate": 5.0173075767551645e-05, "loss": 0.7152, "step": 10456 }, { "epoch": 1.05, "learning_rate": 5.0171304386445646e-05, "loss": 0.619, "step": 10457 }, { "epoch": 1.05, "learning_rate": 5.016953287697715e-05, "loss": 0.3784, "step": 10458 }, { "epoch": 1.05, "learning_rate": 5.0167761239157406e-05, "loss": 1.1764, "step": 10459 }, { "epoch": 1.05, "learning_rate": 5.016598947299771e-05, "loss": 0.4139, "step": 10460 }, { "epoch": 1.05, "learning_rate": 5.016421757850932e-05, "loss": 0.7697, "step": 10461 }, { "epoch": 1.05, "learning_rate": 5.016244555570352e-05, "loss": 0.6424, "step": 10462 }, { "epoch": 1.05, "learning_rate": 5.0160673404591605e-05, "loss": 1.1283, "step": 10463 }, { "epoch": 1.05, "learning_rate": 5.015890112518481e-05, "loss": 1.1086, "step": 10464 }, { "epoch": 1.05, "learning_rate": 5.0157128717494455e-05, "loss": 0.3292, "step": 10465 }, { "epoch": 1.05, "learning_rate": 5.015535618153179e-05, "loss": 0.3448, "step": 10466 }, { "epoch": 1.05, "learning_rate": 5.0153583517308114e-05, "loss": 0.3924, "step": 10467 }, { "epoch": 1.05, "learning_rate": 5.015181072483469e-05, "loss": 1.1099, "step": 10468 }, { "epoch": 1.05, "learning_rate": 5.0150037804122814e-05, "loss": 0.247, "step": 10469 }, { "epoch": 1.05, "learning_rate": 5.0148264755183754e-05, "loss": 0.2456, "step": 10470 }, { "epoch": 1.05, "learning_rate": 5.014649157802882e-05, "loss": 0.6563, "step": 10471 }, { "epoch": 1.05, "learning_rate": 5.014471827266927e-05, "loss": 0.2269, "step": 10472 }, { "epoch": 1.05, "learning_rate": 5.014294483911639e-05, "loss": 0.7084, "step": 10473 }, { "epoch": 1.05, "learning_rate": 5.014117127738148e-05, "loss": 0.2591, "step": 10474 }, { "epoch": 1.05, "learning_rate": 5.013939758747581e-05, "loss": 0.7979, "step": 10475 }, { "epoch": 1.05, "learning_rate": 5.013762376941069e-05, "loss": 0.5993, "step": 10476 }, { "epoch": 1.05, "learning_rate": 5.0135849823197375e-05, "loss": 0.6543, "step": 10477 }, { "epoch": 1.05, "learning_rate": 5.0134075748847175e-05, "loss": 1.008, "step": 10478 }, { "epoch": 1.05, "learning_rate": 5.013230154637139e-05, "loss": 0.1162, "step": 10479 }, { "epoch": 1.05, "learning_rate": 5.013052721578128e-05, "loss": 0.7011, "step": 10480 }, { "epoch": 1.05, "learning_rate": 5.012875275708817e-05, "loss": 0.8838, "step": 10481 }, { "epoch": 1.05, "learning_rate": 5.012697817030332e-05, "loss": 0.2152, "step": 10482 }, { "epoch": 1.05, "learning_rate": 5.012520345543804e-05, "loss": 0.9321, "step": 10483 }, { "epoch": 1.05, "learning_rate": 5.0123428612503624e-05, "loss": 0.713, "step": 10484 }, { "epoch": 1.05, "learning_rate": 5.012165364151136e-05, "loss": 0.2553, "step": 10485 }, { "epoch": 1.05, "learning_rate": 5.0119878542472555e-05, "loss": 0.2819, "step": 10486 }, { "epoch": 1.05, "learning_rate": 5.011810331539849e-05, "loss": 0.2759, "step": 10487 }, { "epoch": 1.05, "learning_rate": 5.011632796030048e-05, "loss": 0.6116, "step": 10488 }, { "epoch": 1.05, "learning_rate": 5.011455247718979e-05, "loss": 0.5032, "step": 10489 }, { "epoch": 1.05, "learning_rate": 5.011277686607776e-05, "loss": 0.2942, "step": 10490 }, { "epoch": 1.05, "learning_rate": 5.0111001126975664e-05, "loss": 0.1109, "step": 10491 }, { "epoch": 1.05, "learning_rate": 5.01092252598948e-05, "loss": 0.5424, "step": 10492 }, { "epoch": 1.05, "learning_rate": 5.010744926484648e-05, "loss": 0.9242, "step": 10493 }, { "epoch": 1.05, "learning_rate": 5.0105673141842006e-05, "loss": 0.8399, "step": 10494 }, { "epoch": 1.05, "learning_rate": 5.010389689089268e-05, "loss": 0.6905, "step": 10495 }, { "epoch": 1.05, "learning_rate": 5.01021205120098e-05, "loss": 0.4386, "step": 10496 }, { "epoch": 1.05, "learning_rate": 5.010034400520467e-05, "loss": 0.949, "step": 10497 }, { "epoch": 1.05, "learning_rate": 5.009856737048859e-05, "loss": 0.8283, "step": 10498 }, { "epoch": 1.05, "learning_rate": 5.009679060787287e-05, "loss": 0.3934, "step": 10499 }, { "epoch": 1.05, "learning_rate": 5.0095013717368835e-05, "loss": 0.8797, "step": 10500 }, { "epoch": 1.05, "learning_rate": 5.009323669898778e-05, "loss": 0.7128, "step": 10501 }, { "epoch": 1.05, "learning_rate": 5.0091459552741e-05, "loss": 0.7095, "step": 10502 }, { "epoch": 1.05, "learning_rate": 5.008968227863981e-05, "loss": 0.2018, "step": 10503 }, { "epoch": 1.05, "learning_rate": 5.008790487669554e-05, "loss": 0.1102, "step": 10504 }, { "epoch": 1.05, "learning_rate": 5.0086127346919476e-05, "loss": 0.41, "step": 10505 }, { "epoch": 1.05, "learning_rate": 5.008434968932293e-05, "loss": 0.4993, "step": 10506 }, { "epoch": 1.05, "learning_rate": 5.0082571903917235e-05, "loss": 0.3965, "step": 10507 }, { "epoch": 1.05, "learning_rate": 5.0080793990713693e-05, "loss": 0.797, "step": 10508 }, { "epoch": 1.05, "learning_rate": 5.007901594972361e-05, "loss": 0.3133, "step": 10509 }, { "epoch": 1.05, "learning_rate": 5.0077237780958316e-05, "loss": 0.5933, "step": 10510 }, { "epoch": 1.05, "learning_rate": 5.007545948442911e-05, "loss": 0.1941, "step": 10511 }, { "epoch": 1.05, "learning_rate": 5.0073681060147324e-05, "loss": 0.3023, "step": 10512 }, { "epoch": 1.05, "learning_rate": 5.007190250812426e-05, "loss": 0.4751, "step": 10513 }, { "epoch": 1.05, "learning_rate": 5.007012382837125e-05, "loss": 0.8028, "step": 10514 }, { "epoch": 1.05, "learning_rate": 5.0068345020899616e-05, "loss": 0.2429, "step": 10515 }, { "epoch": 1.05, "learning_rate": 5.006656608572065e-05, "loss": 0.811, "step": 10516 }, { "epoch": 1.05, "learning_rate": 5.0064787022845705e-05, "loss": 0.3839, "step": 10517 }, { "epoch": 1.06, "learning_rate": 5.006300783228608e-05, "loss": 0.1683, "step": 10518 }, { "epoch": 1.06, "learning_rate": 5.006122851405311e-05, "loss": 0.9179, "step": 10519 }, { "epoch": 1.06, "learning_rate": 5.005944906815811e-05, "loss": 0.7096, "step": 10520 }, { "epoch": 1.06, "learning_rate": 5.005766949461241e-05, "loss": 0.8386, "step": 10521 }, { "epoch": 1.06, "learning_rate": 5.0055889793427335e-05, "loss": 0.3895, "step": 10522 }, { "epoch": 1.06, "learning_rate": 5.00541099646142e-05, "loss": 0.5396, "step": 10523 }, { "epoch": 1.06, "learning_rate": 5.005233000818434e-05, "loss": 0.3231, "step": 10524 }, { "epoch": 1.06, "learning_rate": 5.005054992414908e-05, "loss": 0.51, "step": 10525 }, { "epoch": 1.06, "learning_rate": 5.0048769712519754e-05, "loss": 1.1703, "step": 10526 }, { "epoch": 1.06, "learning_rate": 5.004698937330768e-05, "loss": 1.0277, "step": 10527 }, { "epoch": 1.06, "learning_rate": 5.004520890652419e-05, "loss": 0.7165, "step": 10528 }, { "epoch": 1.06, "learning_rate": 5.004342831218061e-05, "loss": 0.1632, "step": 10529 }, { "epoch": 1.06, "learning_rate": 5.00416475902883e-05, "loss": 0.955, "step": 10530 }, { "epoch": 1.06, "learning_rate": 5.003986674085855e-05, "loss": 0.538, "step": 10531 }, { "epoch": 1.06, "learning_rate": 5.003808576390271e-05, "loss": 2.7574, "step": 10532 }, { "epoch": 1.06, "learning_rate": 5.003630465943213e-05, "loss": 0.9855, "step": 10533 }, { "epoch": 1.06, "learning_rate": 5.003452342745812e-05, "loss": 0.2297, "step": 10534 }, { "epoch": 1.06, "learning_rate": 5.003274206799202e-05, "loss": 0.6326, "step": 10535 }, { "epoch": 1.06, "learning_rate": 5.003096058104518e-05, "loss": 0.8606, "step": 10536 }, { "epoch": 1.06, "learning_rate": 5.0029178966628916e-05, "loss": 0.0779, "step": 10537 }, { "epoch": 1.06, "learning_rate": 5.002739722475458e-05, "loss": 0.267, "step": 10538 }, { "epoch": 1.06, "learning_rate": 5.0025615355433516e-05, "loss": 0.785, "step": 10539 }, { "epoch": 1.06, "learning_rate": 5.002383335867704e-05, "loss": 0.8529, "step": 10540 }, { "epoch": 1.06, "learning_rate": 5.002205123449652e-05, "loss": 0.3882, "step": 10541 }, { "epoch": 1.06, "learning_rate": 5.002026898290327e-05, "loss": 0.4287, "step": 10542 }, { "epoch": 1.06, "learning_rate": 5.001848660390866e-05, "loss": 0.3055, "step": 10543 }, { "epoch": 1.06, "learning_rate": 5.0016704097524e-05, "loss": 0.2223, "step": 10544 }, { "epoch": 1.06, "learning_rate": 5.0014921463760654e-05, "loss": 0.4023, "step": 10545 }, { "epoch": 1.06, "learning_rate": 5.001313870262997e-05, "loss": 0.8692, "step": 10546 }, { "epoch": 1.06, "learning_rate": 5.001135581414328e-05, "loss": 0.4496, "step": 10547 }, { "epoch": 1.06, "learning_rate": 5.000957279831193e-05, "loss": 1.0565, "step": 10548 }, { "epoch": 1.06, "learning_rate": 5.000778965514727e-05, "loss": 0.4907, "step": 10549 }, { "epoch": 1.06, "learning_rate": 5.000600638466066e-05, "loss": 0.1665, "step": 10550 }, { "epoch": 1.06, "learning_rate": 5.000422298686342e-05, "loss": 0.2186, "step": 10551 }, { "epoch": 1.06, "learning_rate": 5.000243946176693e-05, "loss": 1.3032, "step": 10552 }, { "epoch": 1.06, "learning_rate": 5.0000655809382516e-05, "loss": 0.7544, "step": 10553 }, { "epoch": 1.06, "learning_rate": 4.9998872029721536e-05, "loss": 0.6438, "step": 10554 }, { "epoch": 1.06, "learning_rate": 4.9997088122795345e-05, "loss": 0.3006, "step": 10555 }, { "epoch": 1.06, "learning_rate": 4.9995304088615296e-05, "loss": 0.7043, "step": 10556 }, { "epoch": 1.06, "learning_rate": 4.999351992719274e-05, "loss": 0.3279, "step": 10557 }, { "epoch": 1.06, "learning_rate": 4.999173563853902e-05, "loss": 0.5959, "step": 10558 }, { "epoch": 1.06, "learning_rate": 4.998995122266551e-05, "loss": 1.2114, "step": 10559 }, { "epoch": 1.06, "learning_rate": 4.998816667958355e-05, "loss": 1.2984, "step": 10560 }, { "epoch": 1.06, "learning_rate": 4.998638200930449e-05, "loss": 0.2947, "step": 10561 }, { "epoch": 1.06, "learning_rate": 4.9984597211839716e-05, "loss": 0.6929, "step": 10562 }, { "epoch": 1.06, "learning_rate": 4.998281228720055e-05, "loss": 0.2138, "step": 10563 }, { "epoch": 1.06, "learning_rate": 4.998102723539838e-05, "loss": 0.3674, "step": 10564 }, { "epoch": 1.06, "learning_rate": 4.997924205644456e-05, "loss": 0.8755, "step": 10565 }, { "epoch": 1.06, "learning_rate": 4.997745675035043e-05, "loss": 1.0242, "step": 10566 }, { "epoch": 1.06, "learning_rate": 4.997567131712737e-05, "loss": 0.9707, "step": 10567 }, { "epoch": 1.06, "learning_rate": 4.997388575678674e-05, "loss": 0.2132, "step": 10568 }, { "epoch": 1.06, "learning_rate": 4.997210006933991e-05, "loss": 0.7552, "step": 10569 }, { "epoch": 1.06, "learning_rate": 4.9970314254798214e-05, "loss": 0.2816, "step": 10570 }, { "epoch": 1.06, "learning_rate": 4.9968528313173035e-05, "loss": 0.9465, "step": 10571 }, { "epoch": 1.06, "learning_rate": 4.996674224447575e-05, "loss": 0.9379, "step": 10572 }, { "epoch": 1.06, "learning_rate": 4.9964956048717716e-05, "loss": 0.7775, "step": 10573 }, { "epoch": 1.06, "learning_rate": 4.996316972591028e-05, "loss": 0.4899, "step": 10574 }, { "epoch": 1.06, "learning_rate": 4.996138327606484e-05, "loss": 0.4327, "step": 10575 }, { "epoch": 1.06, "learning_rate": 4.995959669919274e-05, "loss": 0.7534, "step": 10576 }, { "epoch": 1.06, "learning_rate": 4.995780999530537e-05, "loss": 0.2767, "step": 10577 }, { "epoch": 1.06, "learning_rate": 4.995602316441408e-05, "loss": 0.2154, "step": 10578 }, { "epoch": 1.06, "learning_rate": 4.995423620653026e-05, "loss": 1.2596, "step": 10579 }, { "epoch": 1.06, "learning_rate": 4.9952449121665265e-05, "loss": 1.0785, "step": 10580 }, { "epoch": 1.06, "learning_rate": 4.995066190983048e-05, "loss": 0.6354, "step": 10581 }, { "epoch": 1.06, "learning_rate": 4.9948874571037264e-05, "loss": 0.3576, "step": 10582 }, { "epoch": 1.06, "learning_rate": 4.9947087105297e-05, "loss": 0.354, "step": 10583 }, { "epoch": 1.06, "learning_rate": 4.9945299512621075e-05, "loss": 0.8239, "step": 10584 }, { "epoch": 1.06, "learning_rate": 4.994351179302084e-05, "loss": 0.4215, "step": 10585 }, { "epoch": 1.06, "learning_rate": 4.994172394650768e-05, "loss": 0.7366, "step": 10586 }, { "epoch": 1.06, "learning_rate": 4.993993597309298e-05, "loss": 0.4457, "step": 10587 }, { "epoch": 1.06, "learning_rate": 4.993814787278811e-05, "loss": 0.2771, "step": 10588 }, { "epoch": 1.06, "learning_rate": 4.993635964560446e-05, "loss": 0.7196, "step": 10589 }, { "epoch": 1.06, "learning_rate": 4.99345712915534e-05, "loss": 0.7369, "step": 10590 }, { "epoch": 1.06, "learning_rate": 4.9932782810646304e-05, "loss": 0.2629, "step": 10591 }, { "epoch": 1.06, "learning_rate": 4.993099420289456e-05, "loss": 0.7898, "step": 10592 }, { "epoch": 1.06, "learning_rate": 4.992920546830956e-05, "loss": 0.4851, "step": 10593 }, { "epoch": 1.06, "learning_rate": 4.992741660690267e-05, "loss": 0.8362, "step": 10594 }, { "epoch": 1.06, "learning_rate": 4.992562761868529e-05, "loss": 0.0735, "step": 10595 }, { "epoch": 1.06, "learning_rate": 4.992383850366879e-05, "loss": 0.872, "step": 10596 }, { "epoch": 1.06, "learning_rate": 4.992204926186457e-05, "loss": 0.3864, "step": 10597 }, { "epoch": 1.06, "learning_rate": 4.9920259893283994e-05, "loss": 1.2985, "step": 10598 }, { "epoch": 1.06, "learning_rate": 4.991847039793847e-05, "loss": 1.381, "step": 10599 }, { "epoch": 1.06, "learning_rate": 4.9916680775839375e-05, "loss": 1.4789, "step": 10600 }, { "epoch": 1.06, "learning_rate": 4.99148910269981e-05, "loss": 0.3146, "step": 10601 }, { "epoch": 1.06, "learning_rate": 4.9913101151426036e-05, "loss": 0.9968, "step": 10602 }, { "epoch": 1.06, "learning_rate": 4.9911311149134565e-05, "loss": 0.1737, "step": 10603 }, { "epoch": 1.06, "learning_rate": 4.990952102013509e-05, "loss": 0.7324, "step": 10604 }, { "epoch": 1.06, "learning_rate": 4.9907730764439e-05, "loss": 1.0239, "step": 10605 }, { "epoch": 1.06, "learning_rate": 4.9905940382057676e-05, "loss": 0.8292, "step": 10606 }, { "epoch": 1.06, "learning_rate": 4.990414987300253e-05, "loss": 0.9381, "step": 10607 }, { "epoch": 1.06, "learning_rate": 4.990235923728494e-05, "loss": 0.9053, "step": 10608 }, { "epoch": 1.06, "learning_rate": 4.990056847491631e-05, "loss": 0.336, "step": 10609 }, { "epoch": 1.06, "learning_rate": 4.989877758590803e-05, "loss": 0.8831, "step": 10610 }, { "epoch": 1.06, "learning_rate": 4.9896986570271507e-05, "loss": 0.7811, "step": 10611 }, { "epoch": 1.06, "learning_rate": 4.9895195428018124e-05, "loss": 0.5484, "step": 10612 }, { "epoch": 1.06, "learning_rate": 4.989340415915929e-05, "loss": 0.3922, "step": 10613 }, { "epoch": 1.06, "learning_rate": 4.9891612763706394e-05, "loss": 0.2514, "step": 10614 }, { "epoch": 1.06, "learning_rate": 4.988982124167084e-05, "loss": 0.532, "step": 10615 }, { "epoch": 1.06, "learning_rate": 4.9888029593064037e-05, "loss": 0.8124, "step": 10616 }, { "epoch": 1.07, "learning_rate": 4.9886237817897384e-05, "loss": 0.3584, "step": 10617 }, { "epoch": 1.07, "learning_rate": 4.988444591618227e-05, "loss": 0.4082, "step": 10618 }, { "epoch": 1.07, "learning_rate": 4.9882653887930104e-05, "loss": 0.9684, "step": 10619 }, { "epoch": 1.07, "learning_rate": 4.98808617331523e-05, "loss": 0.6321, "step": 10620 }, { "epoch": 1.07, "learning_rate": 4.9879069451860246e-05, "loss": 0.489, "step": 10621 }, { "epoch": 1.07, "learning_rate": 4.987727704406536e-05, "loss": 0.756, "step": 10622 }, { "epoch": 1.07, "learning_rate": 4.987548450977905e-05, "loss": 0.2886, "step": 10623 }, { "epoch": 1.07, "learning_rate": 4.987369184901271e-05, "loss": 0.6981, "step": 10624 }, { "epoch": 1.07, "learning_rate": 4.987189906177776e-05, "loss": 0.7174, "step": 10625 }, { "epoch": 1.07, "learning_rate": 4.9870106148085596e-05, "loss": 1.1657, "step": 10626 }, { "epoch": 1.07, "learning_rate": 4.9868313107947655e-05, "loss": 0.2026, "step": 10627 }, { "epoch": 1.07, "learning_rate": 4.9866519941375314e-05, "loss": 0.4548, "step": 10628 }, { "epoch": 1.07, "learning_rate": 4.986472664838e-05, "loss": 0.4855, "step": 10629 }, { "epoch": 1.07, "learning_rate": 4.986293322897312e-05, "loss": 0.4624, "step": 10630 }, { "epoch": 1.07, "learning_rate": 4.986113968316609e-05, "loss": 0.2149, "step": 10631 }, { "epoch": 1.07, "learning_rate": 4.985934601097032e-05, "loss": 0.731, "step": 10632 }, { "epoch": 1.07, "learning_rate": 4.985755221239724e-05, "loss": 0.4376, "step": 10633 }, { "epoch": 1.07, "learning_rate": 4.985575828745824e-05, "loss": 0.5148, "step": 10634 }, { "epoch": 1.07, "learning_rate": 4.985396423616475e-05, "loss": 0.5957, "step": 10635 }, { "epoch": 1.07, "learning_rate": 4.985217005852818e-05, "loss": 0.5412, "step": 10636 }, { "epoch": 1.07, "learning_rate": 4.985037575455996e-05, "loss": 0.5491, "step": 10637 }, { "epoch": 1.07, "learning_rate": 4.984858132427149e-05, "loss": 0.5454, "step": 10638 }, { "epoch": 1.07, "learning_rate": 4.9846786767674214e-05, "loss": 0.1873, "step": 10639 }, { "epoch": 1.07, "learning_rate": 4.984499208477953e-05, "loss": 0.3985, "step": 10640 }, { "epoch": 1.07, "learning_rate": 4.984319727559886e-05, "loss": 0.5021, "step": 10641 }, { "epoch": 1.07, "learning_rate": 4.9841402340143636e-05, "loss": 0.1912, "step": 10642 }, { "epoch": 1.07, "learning_rate": 4.9839607278425275e-05, "loss": 0.4085, "step": 10643 }, { "epoch": 1.07, "learning_rate": 4.98378120904552e-05, "loss": 0.799, "step": 10644 }, { "epoch": 1.07, "learning_rate": 4.9836016776244844e-05, "loss": 1.0066, "step": 10645 }, { "epoch": 1.07, "learning_rate": 4.983422133580561e-05, "loss": 0.2364, "step": 10646 }, { "epoch": 1.07, "learning_rate": 4.983242576914895e-05, "loss": 0.5505, "step": 10647 }, { "epoch": 1.07, "learning_rate": 4.9830630076286266e-05, "loss": 0.2545, "step": 10648 }, { "epoch": 1.07, "learning_rate": 4.9828834257229e-05, "loss": 0.8202, "step": 10649 }, { "epoch": 1.07, "learning_rate": 4.982703831198858e-05, "loss": 0.5677, "step": 10650 }, { "epoch": 1.07, "learning_rate": 4.982524224057642e-05, "loss": 0.864, "step": 10651 }, { "epoch": 1.07, "learning_rate": 4.982344604300397e-05, "loss": 0.5153, "step": 10652 }, { "epoch": 1.07, "learning_rate": 4.982164971928265e-05, "loss": 0.2382, "step": 10653 }, { "epoch": 1.07, "learning_rate": 4.9819853269423886e-05, "loss": 0.9489, "step": 10654 }, { "epoch": 1.07, "learning_rate": 4.981805669343912e-05, "loss": 0.7845, "step": 10655 }, { "epoch": 1.07, "learning_rate": 4.981625999133978e-05, "loss": 0.8421, "step": 10656 }, { "epoch": 1.07, "learning_rate": 4.98144631631373e-05, "loss": 0.5454, "step": 10657 }, { "epoch": 1.07, "learning_rate": 4.981266620884311e-05, "loss": 0.5422, "step": 10658 }, { "epoch": 1.07, "learning_rate": 4.981086912846865e-05, "loss": 0.2758, "step": 10659 }, { "epoch": 1.07, "learning_rate": 4.980907192202536e-05, "loss": 0.3487, "step": 10660 }, { "epoch": 1.07, "learning_rate": 4.9807274589524674e-05, "loss": 0.9183, "step": 10661 }, { "epoch": 1.07, "learning_rate": 4.980547713097802e-05, "loss": 0.2819, "step": 10662 }, { "epoch": 1.07, "learning_rate": 4.980367954639685e-05, "loss": 0.6005, "step": 10663 }, { "epoch": 1.07, "learning_rate": 4.98018818357926e-05, "loss": 0.2639, "step": 10664 }, { "epoch": 1.07, "learning_rate": 4.98000839991767e-05, "loss": 0.7066, "step": 10665 }, { "epoch": 1.07, "learning_rate": 4.97982860365606e-05, "loss": 0.2709, "step": 10666 }, { "epoch": 1.07, "learning_rate": 4.979648794795574e-05, "loss": 0.5904, "step": 10667 }, { "epoch": 1.07, "learning_rate": 4.9794689733373556e-05, "loss": 0.2669, "step": 10668 }, { "epoch": 1.07, "learning_rate": 4.9792891392825506e-05, "loss": 0.5316, "step": 10669 }, { "epoch": 1.07, "learning_rate": 4.979109292632302e-05, "loss": 1.2823, "step": 10670 }, { "epoch": 1.07, "learning_rate": 4.978929433387754e-05, "loss": 0.7404, "step": 10671 }, { "epoch": 1.07, "learning_rate": 4.9787495615500533e-05, "loss": 0.7878, "step": 10672 }, { "epoch": 1.07, "learning_rate": 4.978569677120343e-05, "loss": 0.9868, "step": 10673 }, { "epoch": 1.07, "learning_rate": 4.978389780099766e-05, "loss": 0.8239, "step": 10674 }, { "epoch": 1.07, "learning_rate": 4.9782098704894713e-05, "loss": 0.4293, "step": 10675 }, { "epoch": 1.07, "learning_rate": 4.978029948290602e-05, "loss": 0.1422, "step": 10676 }, { "epoch": 1.07, "learning_rate": 4.9778500135043e-05, "loss": 0.9344, "step": 10677 }, { "epoch": 1.07, "learning_rate": 4.9776700661317146e-05, "loss": 0.9696, "step": 10678 }, { "epoch": 1.07, "learning_rate": 4.97749010617399e-05, "loss": 0.5348, "step": 10679 }, { "epoch": 1.07, "learning_rate": 4.977310133632269e-05, "loss": 0.6436, "step": 10680 }, { "epoch": 1.07, "learning_rate": 4.9771301485076983e-05, "loss": 0.3787, "step": 10681 }, { "epoch": 1.07, "learning_rate": 4.976950150801425e-05, "loss": 0.514, "step": 10682 }, { "epoch": 1.07, "learning_rate": 4.976770140514592e-05, "loss": 0.6755, "step": 10683 }, { "epoch": 1.07, "learning_rate": 4.976590117648347e-05, "loss": 0.4442, "step": 10684 }, { "epoch": 1.07, "learning_rate": 4.9764100822038326e-05, "loss": 0.5663, "step": 10685 }, { "epoch": 1.07, "learning_rate": 4.976230034182198e-05, "loss": 0.4531, "step": 10686 }, { "epoch": 1.07, "learning_rate": 4.976049973584586e-05, "loss": 0.3131, "step": 10687 }, { "epoch": 1.07, "learning_rate": 4.975869900412145e-05, "loss": 0.3208, "step": 10688 }, { "epoch": 1.07, "learning_rate": 4.975689814666018e-05, "loss": 0.5806, "step": 10689 }, { "epoch": 1.07, "learning_rate": 4.9755097163473536e-05, "loss": 0.6136, "step": 10690 }, { "epoch": 1.07, "learning_rate": 4.975329605457296e-05, "loss": 0.1405, "step": 10691 }, { "epoch": 1.07, "learning_rate": 4.9751494819969936e-05, "loss": 0.3457, "step": 10692 }, { "epoch": 1.07, "learning_rate": 4.974969345967591e-05, "loss": 1.1219, "step": 10693 }, { "epoch": 1.07, "learning_rate": 4.9747891973702334e-05, "loss": 0.5511, "step": 10694 }, { "epoch": 1.07, "learning_rate": 4.97460903620607e-05, "loss": 0.1872, "step": 10695 }, { "epoch": 1.07, "learning_rate": 4.974428862476246e-05, "loss": 0.648, "step": 10696 }, { "epoch": 1.07, "learning_rate": 4.974248676181906e-05, "loss": 1.0592, "step": 10697 }, { "epoch": 1.07, "learning_rate": 4.974068477324201e-05, "loss": 0.7096, "step": 10698 }, { "epoch": 1.07, "learning_rate": 4.9738882659042736e-05, "loss": 0.3926, "step": 10699 }, { "epoch": 1.07, "learning_rate": 4.973708041923273e-05, "loss": 0.3493, "step": 10700 }, { "epoch": 1.07, "learning_rate": 4.973527805382345e-05, "loss": 0.6357, "step": 10701 }, { "epoch": 1.07, "learning_rate": 4.973347556282636e-05, "loss": 0.8198, "step": 10702 }, { "epoch": 1.07, "learning_rate": 4.973167294625295e-05, "loss": 0.4711, "step": 10703 }, { "epoch": 1.07, "learning_rate": 4.972987020411467e-05, "loss": 0.8163, "step": 10704 }, { "epoch": 1.07, "learning_rate": 4.9728067336423013e-05, "loss": 0.5387, "step": 10705 }, { "epoch": 1.07, "learning_rate": 4.972626434318943e-05, "loss": 0.7699, "step": 10706 }, { "epoch": 1.07, "learning_rate": 4.9724461224425406e-05, "loss": 0.2828, "step": 10707 }, { "epoch": 1.07, "learning_rate": 4.972265798014242e-05, "loss": 0.742, "step": 10708 }, { "epoch": 1.07, "learning_rate": 4.972085461035194e-05, "loss": 0.4575, "step": 10709 }, { "epoch": 1.07, "learning_rate": 4.9719051115065444e-05, "loss": 0.3503, "step": 10710 }, { "epoch": 1.07, "learning_rate": 4.97172474942944e-05, "loss": 1.1134, "step": 10711 }, { "epoch": 1.07, "learning_rate": 4.97154437480503e-05, "loss": 0.8609, "step": 10712 }, { "epoch": 1.07, "learning_rate": 4.9713639876344615e-05, "loss": 0.6312, "step": 10713 }, { "epoch": 1.07, "learning_rate": 4.9711835879188835e-05, "loss": 0.2567, "step": 10714 }, { "epoch": 1.08, "learning_rate": 4.9710031756594414e-05, "loss": 1.4042, "step": 10715 }, { "epoch": 1.08, "learning_rate": 4.9708227508572866e-05, "loss": 0.7853, "step": 10716 }, { "epoch": 1.08, "learning_rate": 4.970642313513564e-05, "loss": 0.2706, "step": 10717 }, { "epoch": 1.08, "learning_rate": 4.970461863629424e-05, "loss": 0.8251, "step": 10718 }, { "epoch": 1.08, "learning_rate": 4.970281401206014e-05, "loss": 1.1491, "step": 10719 }, { "epoch": 1.08, "learning_rate": 4.9701009262444826e-05, "loss": 0.2773, "step": 10720 }, { "epoch": 1.08, "learning_rate": 4.96992043874598e-05, "loss": 1.2666, "step": 10721 }, { "epoch": 1.08, "learning_rate": 4.969739938711651e-05, "loss": 0.5829, "step": 10722 }, { "epoch": 1.08, "learning_rate": 4.969559426142647e-05, "loss": 0.8262, "step": 10723 }, { "epoch": 1.08, "learning_rate": 4.969378901040117e-05, "loss": 0.3069, "step": 10724 }, { "epoch": 1.08, "learning_rate": 4.969198363405208e-05, "loss": 0.6972, "step": 10725 }, { "epoch": 1.08, "learning_rate": 4.9690178132390695e-05, "loss": 0.6907, "step": 10726 }, { "epoch": 1.08, "learning_rate": 4.9688372505428524e-05, "loss": 0.2347, "step": 10727 }, { "epoch": 1.08, "learning_rate": 4.968656675317702e-05, "loss": 1.2195, "step": 10728 }, { "epoch": 1.08, "learning_rate": 4.96847608756477e-05, "loss": 1.0903, "step": 10729 }, { "epoch": 1.08, "learning_rate": 4.968295487285205e-05, "loss": 0.8708, "step": 10730 }, { "epoch": 1.08, "learning_rate": 4.9681148744801565e-05, "loss": 0.4261, "step": 10731 }, { "epoch": 1.08, "learning_rate": 4.967934249150774e-05, "loss": 1.2892, "step": 10732 }, { "epoch": 1.08, "learning_rate": 4.967753611298206e-05, "loss": 0.5874, "step": 10733 }, { "epoch": 1.08, "learning_rate": 4.9675729609236025e-05, "loss": 0.7388, "step": 10734 }, { "epoch": 1.08, "learning_rate": 4.9673922980281136e-05, "loss": 0.326, "step": 10735 }, { "epoch": 1.08, "learning_rate": 4.967211622612889e-05, "loss": 0.409, "step": 10736 }, { "epoch": 1.08, "learning_rate": 4.967030934679077e-05, "loss": 0.7098, "step": 10737 }, { "epoch": 1.08, "learning_rate": 4.966850234227829e-05, "loss": 0.7869, "step": 10738 }, { "epoch": 1.08, "learning_rate": 4.966669521260294e-05, "loss": 0.3045, "step": 10739 }, { "epoch": 1.08, "learning_rate": 4.966488795777623e-05, "loss": 0.1557, "step": 10740 }, { "epoch": 1.08, "learning_rate": 4.9663080577809645e-05, "loss": 0.3726, "step": 10741 }, { "epoch": 1.08, "learning_rate": 4.966127307271469e-05, "loss": 0.5358, "step": 10742 }, { "epoch": 1.08, "learning_rate": 4.9659465442502885e-05, "loss": 0.8704, "step": 10743 }, { "epoch": 1.08, "learning_rate": 4.965765768718573e-05, "loss": 0.741, "step": 10744 }, { "epoch": 1.08, "learning_rate": 4.96558498067747e-05, "loss": 0.2704, "step": 10745 }, { "epoch": 1.08, "learning_rate": 4.9654041801281325e-05, "loss": 0.3544, "step": 10746 }, { "epoch": 1.08, "learning_rate": 4.965223367071711e-05, "loss": 0.8946, "step": 10747 }, { "epoch": 1.08, "learning_rate": 4.9650425415093554e-05, "loss": 0.7712, "step": 10748 }, { "epoch": 1.08, "learning_rate": 4.964861703442216e-05, "loss": 0.257, "step": 10749 }, { "epoch": 1.08, "learning_rate": 4.964680852871446e-05, "loss": 0.7135, "step": 10750 }, { "epoch": 1.08, "learning_rate": 4.9644999897981925e-05, "loss": 0.2326, "step": 10751 }, { "epoch": 1.08, "learning_rate": 4.9643191142236085e-05, "loss": 0.6048, "step": 10752 }, { "epoch": 1.08, "learning_rate": 4.964138226148846e-05, "loss": 0.8506, "step": 10753 }, { "epoch": 1.08, "learning_rate": 4.963957325575054e-05, "loss": 0.7821, "step": 10754 }, { "epoch": 1.08, "learning_rate": 4.963776412503385e-05, "loss": 0.3478, "step": 10755 }, { "epoch": 1.08, "learning_rate": 4.9635954869349905e-05, "loss": 0.5735, "step": 10756 }, { "epoch": 1.08, "learning_rate": 4.963414548871021e-05, "loss": 0.7432, "step": 10757 }, { "epoch": 1.08, "learning_rate": 4.963233598312628e-05, "loss": 0.7337, "step": 10758 }, { "epoch": 1.08, "learning_rate": 4.9630526352609634e-05, "loss": 0.3074, "step": 10759 }, { "epoch": 1.08, "learning_rate": 4.9628716597171786e-05, "loss": 0.1087, "step": 10760 }, { "epoch": 1.08, "learning_rate": 4.962690671682425e-05, "loss": 0.9537, "step": 10761 }, { "epoch": 1.08, "learning_rate": 4.962509671157855e-05, "loss": 0.7403, "step": 10762 }, { "epoch": 1.08, "learning_rate": 4.9623286581446205e-05, "loss": 0.4678, "step": 10763 }, { "epoch": 1.08, "learning_rate": 4.9621476326438726e-05, "loss": 0.6941, "step": 10764 }, { "epoch": 1.08, "learning_rate": 4.961966594656763e-05, "loss": 0.6884, "step": 10765 }, { "epoch": 1.08, "learning_rate": 4.961785544184445e-05, "loss": 1.5487, "step": 10766 }, { "epoch": 1.08, "learning_rate": 4.96160448122807e-05, "loss": 0.2515, "step": 10767 }, { "epoch": 1.08, "learning_rate": 4.9614234057887895e-05, "loss": 0.7235, "step": 10768 }, { "epoch": 1.08, "learning_rate": 4.961242317867758e-05, "loss": 0.8091, "step": 10769 }, { "epoch": 1.08, "learning_rate": 4.961061217466126e-05, "loss": 0.6205, "step": 10770 }, { "epoch": 1.08, "learning_rate": 4.960880104585045e-05, "loss": 0.7742, "step": 10771 }, { "epoch": 1.08, "learning_rate": 4.96069897922567e-05, "loss": 0.4415, "step": 10772 }, { "epoch": 1.08, "learning_rate": 4.9605178413891534e-05, "loss": 0.2676, "step": 10773 }, { "epoch": 1.08, "learning_rate": 4.960336691076647e-05, "loss": 0.7751, "step": 10774 }, { "epoch": 1.08, "learning_rate": 4.960155528289302e-05, "loss": 0.2462, "step": 10775 }, { "epoch": 1.08, "learning_rate": 4.959974353028275e-05, "loss": 1.1449, "step": 10776 }, { "epoch": 1.08, "learning_rate": 4.959793165294716e-05, "loss": 0.4936, "step": 10777 }, { "epoch": 1.08, "learning_rate": 4.9596119650897775e-05, "loss": 0.5228, "step": 10778 }, { "epoch": 1.08, "learning_rate": 4.9594307524146156e-05, "loss": 0.5308, "step": 10779 }, { "epoch": 1.08, "learning_rate": 4.959249527270381e-05, "loss": 0.2808, "step": 10780 }, { "epoch": 1.08, "learning_rate": 4.9590682896582284e-05, "loss": 0.4311, "step": 10781 }, { "epoch": 1.08, "learning_rate": 4.95888703957931e-05, "loss": 1.1572, "step": 10782 }, { "epoch": 1.08, "learning_rate": 4.9587057770347796e-05, "loss": 0.4618, "step": 10783 }, { "epoch": 1.08, "learning_rate": 4.958524502025791e-05, "loss": 0.4415, "step": 10784 }, { "epoch": 1.08, "learning_rate": 4.958343214553497e-05, "loss": 0.4944, "step": 10785 }, { "epoch": 1.08, "learning_rate": 4.9581619146190526e-05, "loss": 1.0815, "step": 10786 }, { "epoch": 1.08, "learning_rate": 4.95798060222361e-05, "loss": 0.277, "step": 10787 }, { "epoch": 1.08, "learning_rate": 4.9577992773683245e-05, "loss": 0.7589, "step": 10788 }, { "epoch": 1.08, "learning_rate": 4.9576179400543494e-05, "loss": 0.4072, "step": 10789 }, { "epoch": 1.08, "learning_rate": 4.957436590282837e-05, "loss": 0.6962, "step": 10790 }, { "epoch": 1.08, "learning_rate": 4.957255228054943e-05, "loss": 0.367, "step": 10791 }, { "epoch": 1.08, "learning_rate": 4.9570738533718225e-05, "loss": 0.4102, "step": 10792 }, { "epoch": 1.08, "learning_rate": 4.956892466234628e-05, "loss": 0.2331, "step": 10793 }, { "epoch": 1.08, "learning_rate": 4.956711066644514e-05, "loss": 0.4849, "step": 10794 }, { "epoch": 1.08, "learning_rate": 4.956529654602635e-05, "loss": 0.5266, "step": 10795 }, { "epoch": 1.08, "learning_rate": 4.956348230110146e-05, "loss": 1.0626, "step": 10796 }, { "epoch": 1.08, "learning_rate": 4.956166793168201e-05, "loss": 1.5334, "step": 10797 }, { "epoch": 1.08, "learning_rate": 4.9559853437779544e-05, "loss": 0.5352, "step": 10798 }, { "epoch": 1.08, "learning_rate": 4.955803881940562e-05, "loss": 0.4046, "step": 10799 }, { "epoch": 1.08, "learning_rate": 4.955622407657177e-05, "loss": 0.4193, "step": 10800 }, { "epoch": 1.08, "learning_rate": 4.955440920928955e-05, "loss": 0.2971, "step": 10801 }, { "epoch": 1.08, "learning_rate": 4.955259421757051e-05, "loss": 1.01, "step": 10802 }, { "epoch": 1.08, "learning_rate": 4.9550779101426194e-05, "loss": 0.7225, "step": 10803 }, { "epoch": 1.08, "learning_rate": 4.954896386086817e-05, "loss": 0.4328, "step": 10804 }, { "epoch": 1.08, "learning_rate": 4.954714849590797e-05, "loss": 0.5269, "step": 10805 }, { "epoch": 1.08, "learning_rate": 4.9545333006557145e-05, "loss": 0.9235, "step": 10806 }, { "epoch": 1.08, "learning_rate": 4.954351739282726e-05, "loss": 0.1727, "step": 10807 }, { "epoch": 1.08, "learning_rate": 4.9541701654729875e-05, "loss": 0.6238, "step": 10808 }, { "epoch": 1.08, "learning_rate": 4.9539885792276525e-05, "loss": 0.5277, "step": 10809 }, { "epoch": 1.08, "learning_rate": 4.953806980547877e-05, "loss": 0.9283, "step": 10810 }, { "epoch": 1.08, "learning_rate": 4.9536253694348184e-05, "loss": 1.1293, "step": 10811 }, { "epoch": 1.08, "learning_rate": 4.953443745889631e-05, "loss": 0.244, "step": 10812 }, { "epoch": 1.08, "learning_rate": 4.95326210991347e-05, "loss": 1.3642, "step": 10813 }, { "epoch": 1.09, "learning_rate": 4.953080461507492e-05, "loss": 0.3105, "step": 10814 }, { "epoch": 1.09, "learning_rate": 4.952898800672853e-05, "loss": 0.6282, "step": 10815 }, { "epoch": 1.09, "learning_rate": 4.95271712741071e-05, "loss": 0.7181, "step": 10816 }, { "epoch": 1.09, "learning_rate": 4.952535441722217e-05, "loss": 0.3026, "step": 10817 }, { "epoch": 1.09, "learning_rate": 4.952353743608532e-05, "loss": 0.799, "step": 10818 }, { "epoch": 1.09, "learning_rate": 4.9521720330708095e-05, "loss": 0.5047, "step": 10819 }, { "epoch": 1.09, "learning_rate": 4.951990310110208e-05, "loss": 0.5248, "step": 10820 }, { "epoch": 1.09, "learning_rate": 4.951808574727882e-05, "loss": 0.3487, "step": 10821 }, { "epoch": 1.09, "learning_rate": 4.9516268269249885e-05, "loss": 1.3681, "step": 10822 }, { "epoch": 1.09, "learning_rate": 4.951445066702684e-05, "loss": 1.4164, "step": 10823 }, { "epoch": 1.09, "learning_rate": 4.9512632940621264e-05, "loss": 0.9617, "step": 10824 }, { "epoch": 1.09, "learning_rate": 4.951081509004471e-05, "loss": 0.8219, "step": 10825 }, { "epoch": 1.09, "learning_rate": 4.950899711530876e-05, "loss": 1.3257, "step": 10826 }, { "epoch": 1.09, "learning_rate": 4.950717901642497e-05, "loss": 0.4501, "step": 10827 }, { "epoch": 1.09, "learning_rate": 4.9505360793404905e-05, "loss": 0.5327, "step": 10828 }, { "epoch": 1.09, "learning_rate": 4.9503542446260155e-05, "loss": 0.5889, "step": 10829 }, { "epoch": 1.09, "learning_rate": 4.9501723975002275e-05, "loss": 0.4939, "step": 10830 }, { "epoch": 1.09, "learning_rate": 4.9499905379642834e-05, "loss": 1.4237, "step": 10831 }, { "epoch": 1.09, "learning_rate": 4.949808666019342e-05, "loss": 0.7219, "step": 10832 }, { "epoch": 1.09, "learning_rate": 4.94962678166656e-05, "loss": 0.6337, "step": 10833 }, { "epoch": 1.09, "learning_rate": 4.949444884907096e-05, "loss": 0.3246, "step": 10834 }, { "epoch": 1.09, "learning_rate": 4.949262975742105e-05, "loss": 0.6146, "step": 10835 }, { "epoch": 1.09, "learning_rate": 4.949081054172746e-05, "loss": 0.6038, "step": 10836 }, { "epoch": 1.09, "learning_rate": 4.9488991202001775e-05, "loss": 0.9464, "step": 10837 }, { "epoch": 1.09, "learning_rate": 4.948717173825556e-05, "loss": 0.3658, "step": 10838 }, { "epoch": 1.09, "learning_rate": 4.9485352150500395e-05, "loss": 0.9913, "step": 10839 }, { "epoch": 1.09, "learning_rate": 4.9483532438747856e-05, "loss": 1.0141, "step": 10840 }, { "epoch": 1.09, "learning_rate": 4.948171260300954e-05, "loss": 0.575, "step": 10841 }, { "epoch": 1.09, "learning_rate": 4.9479892643297005e-05, "loss": 0.8263, "step": 10842 }, { "epoch": 1.09, "learning_rate": 4.947807255962186e-05, "loss": 0.4259, "step": 10843 }, { "epoch": 1.09, "learning_rate": 4.947625235199566e-05, "loss": 0.3676, "step": 10844 }, { "epoch": 1.09, "learning_rate": 4.947443202043e-05, "loss": 0.4531, "step": 10845 }, { "epoch": 1.09, "learning_rate": 4.947261156493646e-05, "loss": 0.9271, "step": 10846 }, { "epoch": 1.09, "learning_rate": 4.947079098552664e-05, "loss": 0.4719, "step": 10847 }, { "epoch": 1.09, "learning_rate": 4.94689702822121e-05, "loss": 0.5289, "step": 10848 }, { "epoch": 1.09, "learning_rate": 4.9467149455004436e-05, "loss": 0.5879, "step": 10849 }, { "epoch": 1.09, "learning_rate": 4.9465328503915254e-05, "loss": 0.7326, "step": 10850 }, { "epoch": 1.09, "learning_rate": 4.946350742895612e-05, "loss": 0.4099, "step": 10851 }, { "epoch": 1.09, "learning_rate": 4.9461686230138624e-05, "loss": 0.5649, "step": 10852 }, { "epoch": 1.09, "learning_rate": 4.9459864907474364e-05, "loss": 0.1757, "step": 10853 }, { "epoch": 1.09, "learning_rate": 4.945804346097493e-05, "loss": 0.3067, "step": 10854 }, { "epoch": 1.09, "learning_rate": 4.9456221890651905e-05, "loss": 1.2162, "step": 10855 }, { "epoch": 1.09, "learning_rate": 4.945440019651688e-05, "loss": 0.7671, "step": 10856 }, { "epoch": 1.09, "learning_rate": 4.9452578378581465e-05, "loss": 0.8557, "step": 10857 }, { "epoch": 1.09, "learning_rate": 4.945075643685723e-05, "loss": 0.3609, "step": 10858 }, { "epoch": 1.09, "learning_rate": 4.94489343713558e-05, "loss": 0.2201, "step": 10859 }, { "epoch": 1.09, "learning_rate": 4.9447112182088735e-05, "loss": 1.083, "step": 10860 }, { "epoch": 1.09, "learning_rate": 4.9445289869067644e-05, "loss": 0.6943, "step": 10861 }, { "epoch": 1.09, "learning_rate": 4.944346743230413e-05, "loss": 0.8649, "step": 10862 }, { "epoch": 1.09, "learning_rate": 4.944164487180979e-05, "loss": 0.7458, "step": 10863 }, { "epoch": 1.09, "learning_rate": 4.943982218759621e-05, "loss": 0.4211, "step": 10864 }, { "epoch": 1.09, "learning_rate": 4.9437999379675e-05, "loss": 0.4548, "step": 10865 }, { "epoch": 1.09, "learning_rate": 4.9436176448057756e-05, "loss": 0.4205, "step": 10866 }, { "epoch": 1.09, "learning_rate": 4.9434353392756086e-05, "loss": 0.5417, "step": 10867 }, { "epoch": 1.09, "learning_rate": 4.943253021378158e-05, "loss": 1.0289, "step": 10868 }, { "epoch": 1.09, "learning_rate": 4.943070691114584e-05, "loss": 1.146, "step": 10869 }, { "epoch": 1.09, "learning_rate": 4.9428883484860474e-05, "loss": 0.8122, "step": 10870 }, { "epoch": 1.09, "learning_rate": 4.9427059934937086e-05, "loss": 0.6946, "step": 10871 }, { "epoch": 1.09, "learning_rate": 4.942523626138728e-05, "loss": 0.4538, "step": 10872 }, { "epoch": 1.09, "learning_rate": 4.942341246422266e-05, "loss": 0.7757, "step": 10873 }, { "epoch": 1.09, "learning_rate": 4.942158854345483e-05, "loss": 0.4947, "step": 10874 }, { "epoch": 1.09, "learning_rate": 4.9419764499095406e-05, "loss": 0.5483, "step": 10875 }, { "epoch": 1.09, "learning_rate": 4.9417940331155983e-05, "loss": 0.7873, "step": 10876 }, { "epoch": 1.09, "learning_rate": 4.941611603964817e-05, "loss": 0.2957, "step": 10877 }, { "epoch": 1.09, "learning_rate": 4.9414291624583583e-05, "loss": 1.1563, "step": 10878 }, { "epoch": 1.09, "learning_rate": 4.9412467085973835e-05, "loss": 0.7107, "step": 10879 }, { "epoch": 1.09, "learning_rate": 4.9410642423830525e-05, "loss": 1.3699, "step": 10880 }, { "epoch": 1.09, "learning_rate": 4.9408817638165275e-05, "loss": 0.6007, "step": 10881 }, { "epoch": 1.09, "learning_rate": 4.94069927289897e-05, "loss": 0.4656, "step": 10882 }, { "epoch": 1.09, "learning_rate": 4.94051676963154e-05, "loss": 0.7896, "step": 10883 }, { "epoch": 1.09, "learning_rate": 4.940334254015399e-05, "loss": 0.5249, "step": 10884 }, { "epoch": 1.09, "learning_rate": 4.940151726051709e-05, "loss": 0.6774, "step": 10885 }, { "epoch": 1.09, "learning_rate": 4.939969185741632e-05, "loss": 0.7333, "step": 10886 }, { "epoch": 1.09, "learning_rate": 4.939786633086329e-05, "loss": 0.9451, "step": 10887 }, { "epoch": 1.09, "learning_rate": 4.939604068086961e-05, "loss": 1.159, "step": 10888 }, { "epoch": 1.09, "learning_rate": 4.9394214907446915e-05, "loss": 0.5887, "step": 10889 }, { "epoch": 1.09, "learning_rate": 4.939238901060681e-05, "loss": 0.6702, "step": 10890 }, { "epoch": 1.09, "learning_rate": 4.939056299036092e-05, "loss": 0.5116, "step": 10891 }, { "epoch": 1.09, "learning_rate": 4.938873684672086e-05, "loss": 0.767, "step": 10892 }, { "epoch": 1.09, "learning_rate": 4.938691057969827e-05, "loss": 1.2168, "step": 10893 }, { "epoch": 1.09, "learning_rate": 4.938508418930474e-05, "loss": 0.7456, "step": 10894 }, { "epoch": 1.09, "learning_rate": 4.9383257675551914e-05, "loss": 0.4642, "step": 10895 }, { "epoch": 1.09, "learning_rate": 4.938143103845141e-05, "loss": 0.3829, "step": 10896 }, { "epoch": 1.09, "learning_rate": 4.937960427801484e-05, "loss": 0.8614, "step": 10897 }, { "epoch": 1.09, "learning_rate": 4.9377777394253854e-05, "loss": 0.9668, "step": 10898 }, { "epoch": 1.09, "learning_rate": 4.937595038718007e-05, "loss": 0.8873, "step": 10899 }, { "epoch": 1.09, "learning_rate": 4.9374123256805104e-05, "loss": 0.2017, "step": 10900 }, { "epoch": 1.09, "learning_rate": 4.9372296003140585e-05, "loss": 0.4963, "step": 10901 }, { "epoch": 1.09, "learning_rate": 4.937046862619814e-05, "loss": 0.7243, "step": 10902 }, { "epoch": 1.09, "learning_rate": 4.936864112598941e-05, "loss": 0.552, "step": 10903 }, { "epoch": 1.09, "learning_rate": 4.936681350252601e-05, "loss": 0.835, "step": 10904 }, { "epoch": 1.09, "learning_rate": 4.936498575581957e-05, "loss": 0.9164, "step": 10905 }, { "epoch": 1.09, "learning_rate": 4.936315788588174e-05, "loss": 1.0357, "step": 10906 }, { "epoch": 1.09, "learning_rate": 4.936132989272413e-05, "loss": 0.48, "step": 10907 }, { "epoch": 1.09, "learning_rate": 4.935950177635839e-05, "loss": 0.4085, "step": 10908 }, { "epoch": 1.09, "learning_rate": 4.935767353679614e-05, "loss": 0.7084, "step": 10909 }, { "epoch": 1.09, "learning_rate": 4.9355845174049024e-05, "loss": 0.4794, "step": 10910 }, { "epoch": 1.09, "learning_rate": 4.9354016688128664e-05, "loss": 1.0126, "step": 10911 }, { "epoch": 1.1, "learning_rate": 4.935218807904672e-05, "loss": 0.5599, "step": 10912 }, { "epoch": 1.1, "learning_rate": 4.9350359346814796e-05, "loss": 0.3456, "step": 10913 }, { "epoch": 1.1, "learning_rate": 4.9348530491444564e-05, "loss": 0.4505, "step": 10914 }, { "epoch": 1.1, "learning_rate": 4.9346701512947624e-05, "loss": 0.3012, "step": 10915 }, { "epoch": 1.1, "learning_rate": 4.934487241133566e-05, "loss": 1.1704, "step": 10916 }, { "epoch": 1.1, "learning_rate": 4.934304318662026e-05, "loss": 0.6534, "step": 10917 }, { "epoch": 1.1, "learning_rate": 4.934121383881311e-05, "loss": 0.6258, "step": 10918 }, { "epoch": 1.1, "learning_rate": 4.933938436792583e-05, "loss": 0.4361, "step": 10919 }, { "epoch": 1.1, "learning_rate": 4.933755477397006e-05, "loss": 0.608, "step": 10920 }, { "epoch": 1.1, "learning_rate": 4.933572505695745e-05, "loss": 0.8117, "step": 10921 }, { "epoch": 1.1, "learning_rate": 4.9333895216899644e-05, "loss": 0.6192, "step": 10922 }, { "epoch": 1.1, "learning_rate": 4.933206525380828e-05, "loss": 0.2774, "step": 10923 }, { "epoch": 1.1, "learning_rate": 4.933023516769501e-05, "loss": 0.9193, "step": 10924 }, { "epoch": 1.1, "learning_rate": 4.9328404958571484e-05, "loss": 0.6713, "step": 10925 }, { "epoch": 1.1, "learning_rate": 4.932657462644933e-05, "loss": 0.5319, "step": 10926 }, { "epoch": 1.1, "learning_rate": 4.932474417134021e-05, "loss": 0.2154, "step": 10927 }, { "epoch": 1.1, "learning_rate": 4.9322913593255775e-05, "loss": 0.4897, "step": 10928 }, { "epoch": 1.1, "learning_rate": 4.9321082892207656e-05, "loss": 0.2448, "step": 10929 }, { "epoch": 1.1, "learning_rate": 4.931925206820753e-05, "loss": 0.661, "step": 10930 }, { "epoch": 1.1, "learning_rate": 4.9317421121267024e-05, "loss": 0.2826, "step": 10931 }, { "epoch": 1.1, "learning_rate": 4.93155900513978e-05, "loss": 0.4856, "step": 10932 }, { "epoch": 1.1, "learning_rate": 4.931375885861151e-05, "loss": 0.8591, "step": 10933 }, { "epoch": 1.1, "learning_rate": 4.9311927542919806e-05, "loss": 0.8223, "step": 10934 }, { "epoch": 1.1, "learning_rate": 4.9310096104334344e-05, "loss": 0.3379, "step": 10935 }, { "epoch": 1.1, "learning_rate": 4.9308264542866775e-05, "loss": 0.9231, "step": 10936 }, { "epoch": 1.1, "learning_rate": 4.930643285852876e-05, "loss": 1.0973, "step": 10937 }, { "epoch": 1.1, "learning_rate": 4.9304601051331936e-05, "loss": 0.441, "step": 10938 }, { "epoch": 1.1, "learning_rate": 4.930276912128798e-05, "loss": 0.9609, "step": 10939 }, { "epoch": 1.1, "learning_rate": 4.930093706840856e-05, "loss": 0.7108, "step": 10940 }, { "epoch": 1.1, "learning_rate": 4.929910489270531e-05, "loss": 0.2236, "step": 10941 }, { "epoch": 1.1, "learning_rate": 4.9297272594189896e-05, "loss": 0.4066, "step": 10942 }, { "epoch": 1.1, "learning_rate": 4.9295440172873975e-05, "loss": 0.5684, "step": 10943 }, { "epoch": 1.1, "learning_rate": 4.9293607628769224e-05, "loss": 0.2356, "step": 10944 }, { "epoch": 1.1, "learning_rate": 4.929177496188729e-05, "loss": 0.9778, "step": 10945 }, { "epoch": 1.1, "learning_rate": 4.928994217223983e-05, "loss": 0.0971, "step": 10946 }, { "epoch": 1.1, "learning_rate": 4.9288109259838535e-05, "loss": 0.2206, "step": 10947 }, { "epoch": 1.1, "learning_rate": 4.9286276224695036e-05, "loss": 0.3483, "step": 10948 }, { "epoch": 1.1, "learning_rate": 4.9284443066821016e-05, "loss": 0.274, "step": 10949 }, { "epoch": 1.1, "learning_rate": 4.928260978622814e-05, "loss": 0.959, "step": 10950 }, { "epoch": 1.1, "learning_rate": 4.928077638292806e-05, "loss": 0.7869, "step": 10951 }, { "epoch": 1.1, "learning_rate": 4.927894285693246e-05, "loss": 0.3074, "step": 10952 }, { "epoch": 1.1, "learning_rate": 4.9277109208253005e-05, "loss": 0.3501, "step": 10953 }, { "epoch": 1.1, "learning_rate": 4.927527543690136e-05, "loss": 0.7466, "step": 10954 }, { "epoch": 1.1, "learning_rate": 4.9273441542889186e-05, "loss": 0.8443, "step": 10955 }, { "epoch": 1.1, "learning_rate": 4.927160752622817e-05, "loss": 0.3896, "step": 10956 }, { "epoch": 1.1, "learning_rate": 4.9269773386929976e-05, "loss": 0.4985, "step": 10957 }, { "epoch": 1.1, "learning_rate": 4.926793912500627e-05, "loss": 0.5569, "step": 10958 }, { "epoch": 1.1, "learning_rate": 4.926610474046873e-05, "loss": 0.9189, "step": 10959 }, { "epoch": 1.1, "learning_rate": 4.926427023332903e-05, "loss": 1.2543, "step": 10960 }, { "epoch": 1.1, "learning_rate": 4.926243560359884e-05, "loss": 0.9913, "step": 10961 }, { "epoch": 1.1, "learning_rate": 4.9260600851289844e-05, "loss": 0.7554, "step": 10962 }, { "epoch": 1.1, "learning_rate": 4.925876597641371e-05, "loss": 0.3904, "step": 10963 }, { "epoch": 1.1, "learning_rate": 4.925693097898211e-05, "loss": 0.5573, "step": 10964 }, { "epoch": 1.1, "learning_rate": 4.925509585900673e-05, "loss": 0.3571, "step": 10965 }, { "epoch": 1.1, "learning_rate": 4.925326061649924e-05, "loss": 0.7757, "step": 10966 }, { "epoch": 1.1, "learning_rate": 4.925142525147134e-05, "loss": 0.4367, "step": 10967 }, { "epoch": 1.1, "learning_rate": 4.924958976393468e-05, "loss": 0.846, "step": 10968 }, { "epoch": 1.1, "learning_rate": 4.9247754153900954e-05, "loss": 0.5707, "step": 10969 }, { "epoch": 1.1, "learning_rate": 4.924591842138185e-05, "loss": 0.5635, "step": 10970 }, { "epoch": 1.1, "learning_rate": 4.9244082566389033e-05, "loss": 0.3522, "step": 10971 }, { "epoch": 1.1, "learning_rate": 4.92422465889342e-05, "loss": 0.839, "step": 10972 }, { "epoch": 1.1, "learning_rate": 4.924041048902903e-05, "loss": 0.6527, "step": 10973 }, { "epoch": 1.1, "learning_rate": 4.923857426668521e-05, "loss": 0.6239, "step": 10974 }, { "epoch": 1.1, "learning_rate": 4.923673792191442e-05, "loss": 0.8204, "step": 10975 }, { "epoch": 1.1, "learning_rate": 4.9234901454728345e-05, "loss": 0.2869, "step": 10976 }, { "epoch": 1.1, "learning_rate": 4.923306486513867e-05, "loss": 0.9033, "step": 10977 }, { "epoch": 1.1, "learning_rate": 4.9231228153157096e-05, "loss": 0.2796, "step": 10978 }, { "epoch": 1.1, "learning_rate": 4.92293913187953e-05, "loss": 0.3882, "step": 10979 }, { "epoch": 1.1, "learning_rate": 4.9227554362064976e-05, "loss": 0.2882, "step": 10980 }, { "epoch": 1.1, "learning_rate": 4.92257172829778e-05, "loss": 0.1245, "step": 10981 }, { "epoch": 1.1, "learning_rate": 4.9223880081545475e-05, "loss": 0.8291, "step": 10982 }, { "epoch": 1.1, "learning_rate": 4.922204275777971e-05, "loss": 0.1762, "step": 10983 }, { "epoch": 1.1, "learning_rate": 4.922020531169216e-05, "loss": 0.4162, "step": 10984 }, { "epoch": 1.1, "learning_rate": 4.921836774329453e-05, "loss": 0.6876, "step": 10985 }, { "epoch": 1.1, "learning_rate": 4.921653005259853e-05, "loss": 0.959, "step": 10986 }, { "epoch": 1.1, "learning_rate": 4.921469223961584e-05, "loss": 0.6742, "step": 10987 }, { "epoch": 1.1, "learning_rate": 4.9212854304358154e-05, "loss": 0.2403, "step": 10988 }, { "epoch": 1.1, "learning_rate": 4.921101624683718e-05, "loss": 0.6834, "step": 10989 }, { "epoch": 1.1, "learning_rate": 4.9209178067064595e-05, "loss": 0.7732, "step": 10990 }, { "epoch": 1.1, "learning_rate": 4.920733976505212e-05, "loss": 0.2612, "step": 10991 }, { "epoch": 1.1, "learning_rate": 4.9205501340811436e-05, "loss": 0.5882, "step": 10992 }, { "epoch": 1.1, "learning_rate": 4.9203662794354254e-05, "loss": 0.7579, "step": 10993 }, { "epoch": 1.1, "learning_rate": 4.920182412569226e-05, "loss": 0.8843, "step": 10994 }, { "epoch": 1.1, "learning_rate": 4.919998533483716e-05, "loss": 0.599, "step": 10995 }, { "epoch": 1.1, "learning_rate": 4.919814642180066e-05, "loss": 0.7577, "step": 10996 }, { "epoch": 1.1, "learning_rate": 4.9196307386594464e-05, "loss": 0.4344, "step": 10997 }, { "epoch": 1.1, "learning_rate": 4.919446822923027e-05, "loss": 0.2813, "step": 10998 }, { "epoch": 1.1, "learning_rate": 4.919262894971978e-05, "loss": 0.6112, "step": 10999 }, { "epoch": 1.1, "learning_rate": 4.91907895480747e-05, "loss": 0.2073, "step": 11000 }, { "epoch": 1.1, "learning_rate": 4.918895002430674e-05, "loss": 1.0044, "step": 11001 }, { "epoch": 1.1, "learning_rate": 4.91871103784276e-05, "loss": 0.5255, "step": 11002 }, { "epoch": 1.1, "learning_rate": 4.9185270610448984e-05, "loss": 0.2084, "step": 11003 }, { "epoch": 1.1, "learning_rate": 4.91834307203826e-05, "loss": 0.2364, "step": 11004 }, { "epoch": 1.1, "learning_rate": 4.918159070824018e-05, "loss": 0.2311, "step": 11005 }, { "epoch": 1.1, "learning_rate": 4.917975057403339e-05, "loss": 0.8981, "step": 11006 }, { "epoch": 1.1, "learning_rate": 4.9177910317773974e-05, "loss": 0.4499, "step": 11007 }, { "epoch": 1.1, "learning_rate": 4.917606993947364e-05, "loss": 0.4689, "step": 11008 }, { "epoch": 1.1, "learning_rate": 4.917422943914408e-05, "loss": 0.5403, "step": 11009 }, { "epoch": 1.1, "learning_rate": 4.917238881679703e-05, "loss": 0.4505, "step": 11010 }, { "epoch": 1.11, "learning_rate": 4.9170548072444184e-05, "loss": 0.8988, "step": 11011 }, { "epoch": 1.11, "learning_rate": 4.9168707206097266e-05, "loss": 1.0479, "step": 11012 }, { "epoch": 1.11, "learning_rate": 4.916686621776798e-05, "loss": 0.2827, "step": 11013 }, { "epoch": 1.11, "learning_rate": 4.9165025107468053e-05, "loss": 0.2717, "step": 11014 }, { "epoch": 1.11, "learning_rate": 4.9163183875209204e-05, "loss": 0.8115, "step": 11015 }, { "epoch": 1.11, "learning_rate": 4.916134252100314e-05, "loss": 1.1754, "step": 11016 }, { "epoch": 1.11, "learning_rate": 4.9159501044861574e-05, "loss": 1.0684, "step": 11017 }, { "epoch": 1.11, "learning_rate": 4.915765944679624e-05, "loss": 1.0041, "step": 11018 }, { "epoch": 1.11, "learning_rate": 4.915581772681885e-05, "loss": 0.5712, "step": 11019 }, { "epoch": 1.11, "learning_rate": 4.9153975884941116e-05, "loss": 0.7427, "step": 11020 }, { "epoch": 1.11, "learning_rate": 4.915213392117476e-05, "loss": 0.3926, "step": 11021 }, { "epoch": 1.11, "learning_rate": 4.9150291835531525e-05, "loss": 0.528, "step": 11022 }, { "epoch": 1.11, "learning_rate": 4.9148449628023105e-05, "loss": 0.3622, "step": 11023 }, { "epoch": 1.11, "learning_rate": 4.914660729866125e-05, "loss": 0.7504, "step": 11024 }, { "epoch": 1.11, "learning_rate": 4.914476484745766e-05, "loss": 0.346, "step": 11025 }, { "epoch": 1.11, "learning_rate": 4.914292227442407e-05, "loss": 0.9208, "step": 11026 }, { "epoch": 1.11, "learning_rate": 4.91410795795722e-05, "loss": 0.8767, "step": 11027 }, { "epoch": 1.11, "learning_rate": 4.913923676291379e-05, "loss": 0.6796, "step": 11028 }, { "epoch": 1.11, "learning_rate": 4.913739382446056e-05, "loss": 0.8975, "step": 11029 }, { "epoch": 1.11, "learning_rate": 4.9135550764224236e-05, "loss": 0.605, "step": 11030 }, { "epoch": 1.11, "learning_rate": 4.913370758221654e-05, "loss": 1.0378, "step": 11031 }, { "epoch": 1.11, "learning_rate": 4.913186427844921e-05, "loss": 0.1511, "step": 11032 }, { "epoch": 1.11, "learning_rate": 4.913002085293398e-05, "loss": 0.41, "step": 11033 }, { "epoch": 1.11, "learning_rate": 4.912817730568256e-05, "loss": 0.1678, "step": 11034 }, { "epoch": 1.11, "learning_rate": 4.912633363670671e-05, "loss": 0.4933, "step": 11035 }, { "epoch": 1.11, "learning_rate": 4.912448984601815e-05, "loss": 0.2997, "step": 11036 }, { "epoch": 1.11, "learning_rate": 4.91226459336286e-05, "loss": 0.3889, "step": 11037 }, { "epoch": 1.11, "learning_rate": 4.912080189954982e-05, "loss": 0.484, "step": 11038 }, { "epoch": 1.11, "learning_rate": 4.911895774379351e-05, "loss": 0.6977, "step": 11039 }, { "epoch": 1.11, "learning_rate": 4.911711346637144e-05, "loss": 0.554, "step": 11040 }, { "epoch": 1.11, "learning_rate": 4.9115269067295334e-05, "loss": 0.6839, "step": 11041 }, { "epoch": 1.11, "learning_rate": 4.9113424546576926e-05, "loss": 0.1745, "step": 11042 }, { "epoch": 1.11, "learning_rate": 4.9111579904227954e-05, "loss": 1.1247, "step": 11043 }, { "epoch": 1.11, "learning_rate": 4.9109735140260156e-05, "loss": 0.676, "step": 11044 }, { "epoch": 1.11, "learning_rate": 4.9107890254685274e-05, "loss": 0.4921, "step": 11045 }, { "epoch": 1.11, "learning_rate": 4.910604524751505e-05, "loss": 0.5925, "step": 11046 }, { "epoch": 1.11, "learning_rate": 4.910420011876122e-05, "loss": 0.3733, "step": 11047 }, { "epoch": 1.11, "learning_rate": 4.910235486843553e-05, "loss": 0.7926, "step": 11048 }, { "epoch": 1.11, "learning_rate": 4.910050949654973e-05, "loss": 1.4553, "step": 11049 }, { "epoch": 1.11, "learning_rate": 4.9098664003115536e-05, "loss": 0.5921, "step": 11050 }, { "epoch": 1.11, "learning_rate": 4.909681838814472e-05, "loss": 0.432, "step": 11051 }, { "epoch": 1.11, "learning_rate": 4.909497265164901e-05, "loss": 0.539, "step": 11052 }, { "epoch": 1.11, "learning_rate": 4.9093126793640155e-05, "loss": 0.6065, "step": 11053 }, { "epoch": 1.11, "learning_rate": 4.909128081412992e-05, "loss": 0.4236, "step": 11054 }, { "epoch": 1.11, "learning_rate": 4.9089434713130026e-05, "loss": 0.6112, "step": 11055 }, { "epoch": 1.11, "learning_rate": 4.908758849065223e-05, "loss": 0.6273, "step": 11056 }, { "epoch": 1.11, "learning_rate": 4.9085742146708287e-05, "loss": 0.5346, "step": 11057 }, { "epoch": 1.11, "learning_rate": 4.908389568130995e-05, "loss": 0.0773, "step": 11058 }, { "epoch": 1.11, "learning_rate": 4.9082049094468944e-05, "loss": 0.2556, "step": 11059 }, { "epoch": 1.11, "learning_rate": 4.908020238619704e-05, "loss": 0.651, "step": 11060 }, { "epoch": 1.11, "learning_rate": 4.907835555650599e-05, "loss": 0.4486, "step": 11061 }, { "epoch": 1.11, "learning_rate": 4.907650860540754e-05, "loss": 0.818, "step": 11062 }, { "epoch": 1.11, "learning_rate": 4.9074661532913455e-05, "loss": 0.2239, "step": 11063 }, { "epoch": 1.11, "learning_rate": 4.9072814339035475e-05, "loss": 0.3221, "step": 11064 }, { "epoch": 1.11, "learning_rate": 4.907096702378536e-05, "loss": 1.0196, "step": 11065 }, { "epoch": 1.11, "learning_rate": 4.906911958717486e-05, "loss": 0.4649, "step": 11066 }, { "epoch": 1.11, "learning_rate": 4.906727202921575e-05, "loss": 0.2785, "step": 11067 }, { "epoch": 1.11, "learning_rate": 4.9065424349919765e-05, "loss": 0.8688, "step": 11068 }, { "epoch": 1.11, "learning_rate": 4.9063576549298674e-05, "loss": 0.5049, "step": 11069 }, { "epoch": 1.11, "learning_rate": 4.906172862736423e-05, "loss": 0.7247, "step": 11070 }, { "epoch": 1.11, "learning_rate": 4.9059880584128204e-05, "loss": 1.3893, "step": 11071 }, { "epoch": 1.11, "learning_rate": 4.905803241960235e-05, "loss": 1.285, "step": 11072 }, { "epoch": 1.11, "learning_rate": 4.905618413379842e-05, "loss": 0.8336, "step": 11073 }, { "epoch": 1.11, "learning_rate": 4.9054335726728185e-05, "loss": 0.951, "step": 11074 }, { "epoch": 1.11, "learning_rate": 4.905248719840342e-05, "loss": 0.8431, "step": 11075 }, { "epoch": 1.11, "learning_rate": 4.905063854883586e-05, "loss": 0.4709, "step": 11076 }, { "epoch": 1.11, "learning_rate": 4.904878977803728e-05, "loss": 0.6233, "step": 11077 }, { "epoch": 1.11, "learning_rate": 4.904694088601945e-05, "loss": 0.7212, "step": 11078 }, { "epoch": 1.11, "learning_rate": 4.9045091872794144e-05, "loss": 0.5444, "step": 11079 }, { "epoch": 1.11, "learning_rate": 4.904324273837311e-05, "loss": 0.6232, "step": 11080 }, { "epoch": 1.11, "learning_rate": 4.904139348276813e-05, "loss": 1.1098, "step": 11081 }, { "epoch": 1.11, "learning_rate": 4.903954410599096e-05, "loss": 0.7099, "step": 11082 }, { "epoch": 1.11, "learning_rate": 4.903769460805337e-05, "loss": 0.9632, "step": 11083 }, { "epoch": 1.11, "learning_rate": 4.9035844988967144e-05, "loss": 1.3079, "step": 11084 }, { "epoch": 1.11, "learning_rate": 4.9033995248744044e-05, "loss": 0.817, "step": 11085 }, { "epoch": 1.11, "learning_rate": 4.903214538739583e-05, "loss": 0.2615, "step": 11086 }, { "epoch": 1.11, "learning_rate": 4.903029540493428e-05, "loss": 0.9084, "step": 11087 }, { "epoch": 1.11, "learning_rate": 4.902844530137119e-05, "loss": 0.4529, "step": 11088 }, { "epoch": 1.11, "learning_rate": 4.902659507671829e-05, "loss": 1.2411, "step": 11089 }, { "epoch": 1.11, "learning_rate": 4.90247447309874e-05, "loss": 0.3234, "step": 11090 }, { "epoch": 1.11, "learning_rate": 4.9022894264190245e-05, "loss": 0.6122, "step": 11091 }, { "epoch": 1.11, "learning_rate": 4.9021043676338646e-05, "loss": 0.5062, "step": 11092 }, { "epoch": 1.11, "learning_rate": 4.901919296744436e-05, "loss": 1.318, "step": 11093 }, { "epoch": 1.11, "learning_rate": 4.901734213751915e-05, "loss": 0.7722, "step": 11094 }, { "epoch": 1.11, "learning_rate": 4.9015491186574816e-05, "loss": 1.0172, "step": 11095 }, { "epoch": 1.11, "learning_rate": 4.901364011462314e-05, "loss": 0.8358, "step": 11096 }, { "epoch": 1.11, "learning_rate": 4.901178892167589e-05, "loss": 1.0101, "step": 11097 }, { "epoch": 1.11, "learning_rate": 4.900993760774484e-05, "loss": 0.8039, "step": 11098 }, { "epoch": 1.11, "learning_rate": 4.900808617284178e-05, "loss": 1.1027, "step": 11099 }, { "epoch": 1.11, "learning_rate": 4.90062346169785e-05, "loss": 0.2487, "step": 11100 }, { "epoch": 1.11, "learning_rate": 4.9004382940166766e-05, "loss": 0.3034, "step": 11101 }, { "epoch": 1.11, "learning_rate": 4.900253114241835e-05, "loss": 1.2203, "step": 11102 }, { "epoch": 1.11, "learning_rate": 4.900067922374509e-05, "loss": 0.5848, "step": 11103 }, { "epoch": 1.11, "learning_rate": 4.8998827184158715e-05, "loss": 0.5407, "step": 11104 }, { "epoch": 1.11, "learning_rate": 4.899697502367103e-05, "loss": 0.9315, "step": 11105 }, { "epoch": 1.11, "learning_rate": 4.899512274229383e-05, "loss": 1.1382, "step": 11106 }, { "epoch": 1.11, "learning_rate": 4.89932703400389e-05, "loss": 0.8457, "step": 11107 }, { "epoch": 1.11, "learning_rate": 4.8991417816918005e-05, "loss": 0.752, "step": 11108 }, { "epoch": 1.12, "learning_rate": 4.898956517294296e-05, "loss": 0.3529, "step": 11109 }, { "epoch": 1.12, "learning_rate": 4.898771240812555e-05, "loss": 0.8772, "step": 11110 }, { "epoch": 1.12, "learning_rate": 4.898585952247756e-05, "loss": 1.1247, "step": 11111 }, { "epoch": 1.12, "learning_rate": 4.8984006516010785e-05, "loss": 0.8604, "step": 11112 }, { "epoch": 1.12, "learning_rate": 4.898215338873702e-05, "loss": 1.1622, "step": 11113 }, { "epoch": 1.12, "learning_rate": 4.8980300140668045e-05, "loss": 0.17, "step": 11114 }, { "epoch": 1.12, "learning_rate": 4.8978446771815653e-05, "loss": 0.1854, "step": 11115 }, { "epoch": 1.12, "learning_rate": 4.897659328219166e-05, "loss": 0.7579, "step": 11116 }, { "epoch": 1.12, "learning_rate": 4.8974739671807835e-05, "loss": 0.3263, "step": 11117 }, { "epoch": 1.12, "learning_rate": 4.897288594067598e-05, "loss": 0.4327, "step": 11118 }, { "epoch": 1.12, "learning_rate": 4.897103208880791e-05, "loss": 0.2055, "step": 11119 }, { "epoch": 1.12, "learning_rate": 4.8969178116215406e-05, "loss": 0.3571, "step": 11120 }, { "epoch": 1.12, "learning_rate": 4.896732402291027e-05, "loss": 0.4576, "step": 11121 }, { "epoch": 1.12, "learning_rate": 4.8965469808904294e-05, "loss": 0.4125, "step": 11122 }, { "epoch": 1.12, "learning_rate": 4.896361547420929e-05, "loss": 0.5556, "step": 11123 }, { "epoch": 1.12, "learning_rate": 4.896176101883704e-05, "loss": 0.6056, "step": 11124 }, { "epoch": 1.12, "learning_rate": 4.895990644279936e-05, "loss": 0.767, "step": 11125 }, { "epoch": 1.12, "learning_rate": 4.895805174610806e-05, "loss": 0.6715, "step": 11126 }, { "epoch": 1.12, "learning_rate": 4.895619692877492e-05, "loss": 0.2536, "step": 11127 }, { "epoch": 1.12, "learning_rate": 4.8954341990811754e-05, "loss": 0.4187, "step": 11128 }, { "epoch": 1.12, "learning_rate": 4.895248693223037e-05, "loss": 0.7401, "step": 11129 }, { "epoch": 1.12, "learning_rate": 4.895063175304256e-05, "loss": 0.8805, "step": 11130 }, { "epoch": 1.12, "learning_rate": 4.8948776453260146e-05, "loss": 0.563, "step": 11131 }, { "epoch": 1.12, "learning_rate": 4.894692103289493e-05, "loss": 0.7644, "step": 11132 }, { "epoch": 1.12, "learning_rate": 4.894506549195871e-05, "loss": 0.6929, "step": 11133 }, { "epoch": 1.12, "learning_rate": 4.8943209830463296e-05, "loss": 0.4158, "step": 11134 }, { "epoch": 1.12, "learning_rate": 4.8941354048420516e-05, "loss": 0.7907, "step": 11135 }, { "epoch": 1.12, "learning_rate": 4.893949814584216e-05, "loss": 1.0021, "step": 11136 }, { "epoch": 1.12, "learning_rate": 4.893764212274003e-05, "loss": 1.0911, "step": 11137 }, { "epoch": 1.12, "learning_rate": 4.8935785979125966e-05, "loss": 0.6067, "step": 11138 }, { "epoch": 1.12, "learning_rate": 4.893392971501176e-05, "loss": 0.5453, "step": 11139 }, { "epoch": 1.12, "learning_rate": 4.893207333040922e-05, "loss": 0.9399, "step": 11140 }, { "epoch": 1.12, "learning_rate": 4.893021682533017e-05, "loss": 0.9289, "step": 11141 }, { "epoch": 1.12, "learning_rate": 4.892836019978643e-05, "loss": 0.7099, "step": 11142 }, { "epoch": 1.12, "learning_rate": 4.89265034537898e-05, "loss": 1.251, "step": 11143 }, { "epoch": 1.12, "learning_rate": 4.89246465873521e-05, "loss": 0.9128, "step": 11144 }, { "epoch": 1.12, "learning_rate": 4.892278960048515e-05, "loss": 0.7141, "step": 11145 }, { "epoch": 1.12, "learning_rate": 4.892093249320077e-05, "loss": 0.549, "step": 11146 }, { "epoch": 1.12, "learning_rate": 4.891907526551077e-05, "loss": 0.3366, "step": 11147 }, { "epoch": 1.12, "learning_rate": 4.8917217917426975e-05, "loss": 0.8483, "step": 11148 }, { "epoch": 1.12, "learning_rate": 4.8915360448961194e-05, "loss": 0.9362, "step": 11149 }, { "epoch": 1.12, "learning_rate": 4.8913502860125255e-05, "loss": 0.9424, "step": 11150 }, { "epoch": 1.12, "learning_rate": 4.891164515093099e-05, "loss": 0.2651, "step": 11151 }, { "epoch": 1.12, "learning_rate": 4.890978732139021e-05, "loss": 0.8614, "step": 11152 }, { "epoch": 1.12, "learning_rate": 4.890792937151472e-05, "loss": 0.8236, "step": 11153 }, { "epoch": 1.12, "learning_rate": 4.890607130131637e-05, "loss": 0.781, "step": 11154 }, { "epoch": 1.12, "learning_rate": 4.890421311080698e-05, "loss": 1.0825, "step": 11155 }, { "epoch": 1.12, "learning_rate": 4.890235479999837e-05, "loss": 1.0252, "step": 11156 }, { "epoch": 1.12, "learning_rate": 4.890049636890235e-05, "loss": 0.8282, "step": 11157 }, { "epoch": 1.12, "learning_rate": 4.889863781753078e-05, "loss": 1.4027, "step": 11158 }, { "epoch": 1.12, "learning_rate": 4.8896779145895454e-05, "loss": 0.454, "step": 11159 }, { "epoch": 1.12, "learning_rate": 4.889492035400822e-05, "loss": 1.1326, "step": 11160 }, { "epoch": 1.12, "learning_rate": 4.8893061441880894e-05, "loss": 0.6843, "step": 11161 }, { "epoch": 1.12, "learning_rate": 4.889120240952532e-05, "loss": 0.863, "step": 11162 }, { "epoch": 1.12, "learning_rate": 4.8889343256953324e-05, "loss": 0.2237, "step": 11163 }, { "epoch": 1.12, "learning_rate": 4.888748398417672e-05, "loss": 0.6313, "step": 11164 }, { "epoch": 1.12, "learning_rate": 4.888562459120736e-05, "loss": 0.1438, "step": 11165 }, { "epoch": 1.12, "learning_rate": 4.888376507805707e-05, "loss": 0.7576, "step": 11166 }, { "epoch": 1.12, "learning_rate": 4.888190544473769e-05, "loss": 1.4292, "step": 11167 }, { "epoch": 1.12, "learning_rate": 4.888004569126104e-05, "loss": 0.6256, "step": 11168 }, { "epoch": 1.12, "learning_rate": 4.8878185817638954e-05, "loss": 0.8476, "step": 11169 }, { "epoch": 1.12, "learning_rate": 4.8876325823883286e-05, "loss": 0.7571, "step": 11170 }, { "epoch": 1.12, "learning_rate": 4.8874465710005865e-05, "loss": 0.5526, "step": 11171 }, { "epoch": 1.12, "learning_rate": 4.887260547601851e-05, "loss": 1.0663, "step": 11172 }, { "epoch": 1.12, "learning_rate": 4.8870745121933086e-05, "loss": 0.726, "step": 11173 }, { "epoch": 1.12, "learning_rate": 4.886888464776141e-05, "loss": 0.3133, "step": 11174 }, { "epoch": 1.12, "learning_rate": 4.8867024053515335e-05, "loss": 0.2047, "step": 11175 }, { "epoch": 1.12, "learning_rate": 4.88651633392067e-05, "loss": 0.1152, "step": 11176 }, { "epoch": 1.12, "learning_rate": 4.886330250484733e-05, "loss": 0.3746, "step": 11177 }, { "epoch": 1.12, "learning_rate": 4.8861441550449096e-05, "loss": 0.2688, "step": 11178 }, { "epoch": 1.12, "learning_rate": 4.88595804760238e-05, "loss": 0.172, "step": 11179 }, { "epoch": 1.12, "learning_rate": 4.8857719281583334e-05, "loss": 0.7309, "step": 11180 }, { "epoch": 1.12, "learning_rate": 4.8855857967139505e-05, "loss": 0.4211, "step": 11181 }, { "epoch": 1.12, "learning_rate": 4.885399653270416e-05, "loss": 0.9428, "step": 11182 }, { "epoch": 1.12, "learning_rate": 4.8852134978289165e-05, "loss": 1.0327, "step": 11183 }, { "epoch": 1.12, "learning_rate": 4.8850273303906357e-05, "loss": 0.1833, "step": 11184 }, { "epoch": 1.12, "learning_rate": 4.884841150956757e-05, "loss": 0.402, "step": 11185 }, { "epoch": 1.12, "learning_rate": 4.884654959528467e-05, "loss": 0.433, "step": 11186 }, { "epoch": 1.12, "learning_rate": 4.8844687561069496e-05, "loss": 0.5171, "step": 11187 }, { "epoch": 1.12, "learning_rate": 4.88428254069339e-05, "loss": 0.5062, "step": 11188 }, { "epoch": 1.12, "learning_rate": 4.8840963132889726e-05, "loss": 1.1196, "step": 11189 }, { "epoch": 1.12, "learning_rate": 4.883910073894884e-05, "loss": 0.2772, "step": 11190 }, { "epoch": 1.12, "learning_rate": 4.883723822512307e-05, "loss": 0.5992, "step": 11191 }, { "epoch": 1.12, "learning_rate": 4.883537559142429e-05, "loss": 0.9107, "step": 11192 }, { "epoch": 1.12, "learning_rate": 4.883351283786435e-05, "loss": 0.2602, "step": 11193 }, { "epoch": 1.12, "learning_rate": 4.8831649964455094e-05, "loss": 1.1379, "step": 11194 }, { "epoch": 1.12, "learning_rate": 4.882978697120838e-05, "loss": 1.06, "step": 11195 }, { "epoch": 1.12, "learning_rate": 4.882792385813607e-05, "loss": 1.3195, "step": 11196 }, { "epoch": 1.12, "learning_rate": 4.8826060625250015e-05, "loss": 0.2822, "step": 11197 }, { "epoch": 1.12, "learning_rate": 4.882419727256207e-05, "loss": 0.4396, "step": 11198 }, { "epoch": 1.12, "learning_rate": 4.8822333800084085e-05, "loss": 0.5465, "step": 11199 }, { "epoch": 1.12, "learning_rate": 4.882047020782794e-05, "loss": 0.466, "step": 11200 }, { "epoch": 1.12, "learning_rate": 4.881860649580548e-05, "loss": 0.5672, "step": 11201 }, { "epoch": 1.12, "learning_rate": 4.881674266402856e-05, "loss": 0.7448, "step": 11202 }, { "epoch": 1.12, "learning_rate": 4.881487871250905e-05, "loss": 0.2199, "step": 11203 }, { "epoch": 1.12, "learning_rate": 4.881301464125882e-05, "loss": 0.6137, "step": 11204 }, { "epoch": 1.12, "learning_rate": 4.881115045028971e-05, "loss": 1.377, "step": 11205 }, { "epoch": 1.12, "learning_rate": 4.8809286139613604e-05, "loss": 0.9579, "step": 11206 }, { "epoch": 1.12, "learning_rate": 4.880742170924235e-05, "loss": 0.3754, "step": 11207 }, { "epoch": 1.13, "learning_rate": 4.880555715918782e-05, "loss": 0.5302, "step": 11208 }, { "epoch": 1.13, "learning_rate": 4.8803692489461885e-05, "loss": 1.1177, "step": 11209 }, { "epoch": 1.13, "learning_rate": 4.8801827700076395e-05, "loss": 1.0518, "step": 11210 }, { "epoch": 1.13, "learning_rate": 4.8799962791043226e-05, "loss": 0.9315, "step": 11211 }, { "epoch": 1.13, "learning_rate": 4.879809776237425e-05, "loss": 0.338, "step": 11212 }, { "epoch": 1.13, "learning_rate": 4.8796232614081336e-05, "loss": 0.5766, "step": 11213 }, { "epoch": 1.13, "learning_rate": 4.879436734617634e-05, "loss": 0.7422, "step": 11214 }, { "epoch": 1.13, "learning_rate": 4.879250195867114e-05, "loss": 0.428, "step": 11215 }, { "epoch": 1.13, "learning_rate": 4.879063645157761e-05, "loss": 0.5994, "step": 11216 }, { "epoch": 1.13, "learning_rate": 4.878877082490762e-05, "loss": 1.049, "step": 11217 }, { "epoch": 1.13, "learning_rate": 4.878690507867303e-05, "loss": 0.153, "step": 11218 }, { "epoch": 1.13, "learning_rate": 4.878503921288573e-05, "loss": 0.9688, "step": 11219 }, { "epoch": 1.13, "learning_rate": 4.878317322755759e-05, "loss": 0.3276, "step": 11220 }, { "epoch": 1.13, "learning_rate": 4.8781307122700474e-05, "loss": 0.4321, "step": 11221 }, { "epoch": 1.13, "learning_rate": 4.8779440898326266e-05, "loss": 0.2641, "step": 11222 }, { "epoch": 1.13, "learning_rate": 4.877757455444684e-05, "loss": 2.2926, "step": 11223 }, { "epoch": 1.13, "learning_rate": 4.8775708091074083e-05, "loss": 0.2669, "step": 11224 }, { "epoch": 1.13, "learning_rate": 4.8773841508219846e-05, "loss": 0.981, "step": 11225 }, { "epoch": 1.13, "learning_rate": 4.8771974805896035e-05, "loss": 0.2126, "step": 11226 }, { "epoch": 1.13, "learning_rate": 4.877010798411452e-05, "loss": 0.4249, "step": 11227 }, { "epoch": 1.13, "learning_rate": 4.876824104288717e-05, "loss": 0.8038, "step": 11228 }, { "epoch": 1.13, "learning_rate": 4.876637398222587e-05, "loss": 1.028, "step": 11229 }, { "epoch": 1.13, "learning_rate": 4.876450680214252e-05, "loss": 0.9861, "step": 11230 }, { "epoch": 1.13, "learning_rate": 4.876263950264897e-05, "loss": 0.8818, "step": 11231 }, { "epoch": 1.13, "learning_rate": 4.876077208375713e-05, "loss": 0.6171, "step": 11232 }, { "epoch": 1.13, "learning_rate": 4.875890454547888e-05, "loss": 0.3778, "step": 11233 }, { "epoch": 1.13, "learning_rate": 4.875703688782608e-05, "loss": 0.5271, "step": 11234 }, { "epoch": 1.13, "learning_rate": 4.8755169110810645e-05, "loss": 0.3891, "step": 11235 }, { "epoch": 1.13, "learning_rate": 4.8753301214444445e-05, "loss": 0.8934, "step": 11236 }, { "epoch": 1.13, "learning_rate": 4.8751433198739367e-05, "loss": 0.8198, "step": 11237 }, { "epoch": 1.13, "learning_rate": 4.8749565063707316e-05, "loss": 0.1212, "step": 11238 }, { "epoch": 1.13, "learning_rate": 4.874769680936015e-05, "loss": 0.9863, "step": 11239 }, { "epoch": 1.13, "learning_rate": 4.874582843570978e-05, "loss": 1.0137, "step": 11240 }, { "epoch": 1.13, "learning_rate": 4.874395994276808e-05, "loss": 0.5507, "step": 11241 }, { "epoch": 1.13, "learning_rate": 4.874209133054697e-05, "loss": 0.1969, "step": 11242 }, { "epoch": 1.13, "learning_rate": 4.874022259905831e-05, "loss": 0.6109, "step": 11243 }, { "epoch": 1.13, "learning_rate": 4.8738353748313994e-05, "loss": 0.4693, "step": 11244 }, { "epoch": 1.13, "learning_rate": 4.8736484778325924e-05, "loss": 0.4986, "step": 11245 }, { "epoch": 1.13, "learning_rate": 4.8734615689106006e-05, "loss": 0.6214, "step": 11246 }, { "epoch": 1.13, "learning_rate": 4.873274648066611e-05, "loss": 0.7593, "step": 11247 }, { "epoch": 1.13, "learning_rate": 4.8730877153018143e-05, "loss": 0.3554, "step": 11248 }, { "epoch": 1.13, "learning_rate": 4.8729007706174004e-05, "loss": 0.1933, "step": 11249 }, { "epoch": 1.13, "learning_rate": 4.8727138140145586e-05, "loss": 0.4605, "step": 11250 }, { "epoch": 1.13, "learning_rate": 4.8725268454944776e-05, "loss": 0.5799, "step": 11251 }, { "epoch": 1.13, "learning_rate": 4.8723398650583485e-05, "loss": 0.3448, "step": 11252 }, { "epoch": 1.13, "learning_rate": 4.8721528727073615e-05, "loss": 0.7925, "step": 11253 }, { "epoch": 1.13, "learning_rate": 4.871965868442705e-05, "loss": 0.2884, "step": 11254 }, { "epoch": 1.13, "learning_rate": 4.87177885226557e-05, "loss": 0.4385, "step": 11255 }, { "epoch": 1.13, "learning_rate": 4.871591824177146e-05, "loss": 0.5114, "step": 11256 }, { "epoch": 1.13, "learning_rate": 4.8714047841786246e-05, "loss": 0.551, "step": 11257 }, { "epoch": 1.13, "learning_rate": 4.8712177322711945e-05, "loss": 0.5835, "step": 11258 }, { "epoch": 1.13, "learning_rate": 4.8710306684560464e-05, "loss": 0.3865, "step": 11259 }, { "epoch": 1.13, "learning_rate": 4.870843592734371e-05, "loss": 0.8611, "step": 11260 }, { "epoch": 1.13, "learning_rate": 4.870656505107359e-05, "loss": 0.1952, "step": 11261 }, { "epoch": 1.13, "learning_rate": 4.870469405576201e-05, "loss": 0.4664, "step": 11262 }, { "epoch": 1.13, "learning_rate": 4.870282294142087e-05, "loss": 0.7424, "step": 11263 }, { "epoch": 1.13, "learning_rate": 4.8700951708062074e-05, "loss": 0.737, "step": 11264 }, { "epoch": 1.13, "learning_rate": 4.8699080355697535e-05, "loss": 0.4903, "step": 11265 }, { "epoch": 1.13, "learning_rate": 4.869720888433917e-05, "loss": 0.6276, "step": 11266 }, { "epoch": 1.13, "learning_rate": 4.869533729399888e-05, "loss": 0.4953, "step": 11267 }, { "epoch": 1.13, "learning_rate": 4.869346558468857e-05, "loss": 0.5794, "step": 11268 }, { "epoch": 1.13, "learning_rate": 4.8691593756420155e-05, "loss": 1.1817, "step": 11269 }, { "epoch": 1.13, "learning_rate": 4.8689721809205553e-05, "loss": 0.2882, "step": 11270 }, { "epoch": 1.13, "learning_rate": 4.868784974305667e-05, "loss": 0.4207, "step": 11271 }, { "epoch": 1.13, "learning_rate": 4.868597755798542e-05, "loss": 1.0092, "step": 11272 }, { "epoch": 1.13, "learning_rate": 4.8684105254003725e-05, "loss": 0.486, "step": 11273 }, { "epoch": 1.13, "learning_rate": 4.868223283112347e-05, "loss": 0.3, "step": 11274 }, { "epoch": 1.13, "learning_rate": 4.868036028935661e-05, "loss": 0.166, "step": 11275 }, { "epoch": 1.13, "learning_rate": 4.867848762871504e-05, "loss": 0.322, "step": 11276 }, { "epoch": 1.13, "learning_rate": 4.867661484921069e-05, "loss": 1.2728, "step": 11277 }, { "epoch": 1.13, "learning_rate": 4.867474195085546e-05, "loss": 1.1758, "step": 11278 }, { "epoch": 1.13, "learning_rate": 4.8672868933661275e-05, "loss": 1.2457, "step": 11279 }, { "epoch": 1.13, "learning_rate": 4.867099579764005e-05, "loss": 0.3804, "step": 11280 }, { "epoch": 1.13, "learning_rate": 4.866912254280372e-05, "loss": 0.5948, "step": 11281 }, { "epoch": 1.13, "learning_rate": 4.8667249169164183e-05, "loss": 0.3551, "step": 11282 }, { "epoch": 1.13, "learning_rate": 4.8665375676733383e-05, "loss": 0.6472, "step": 11283 }, { "epoch": 1.13, "learning_rate": 4.866350206552324e-05, "loss": 0.4173, "step": 11284 }, { "epoch": 1.13, "learning_rate": 4.866162833554566e-05, "loss": 0.5875, "step": 11285 }, { "epoch": 1.13, "learning_rate": 4.8659754486812575e-05, "loss": 0.2274, "step": 11286 }, { "epoch": 1.13, "learning_rate": 4.8657880519335916e-05, "loss": 1.4335, "step": 11287 }, { "epoch": 1.13, "learning_rate": 4.865600643312761e-05, "loss": 0.8436, "step": 11288 }, { "epoch": 1.13, "learning_rate": 4.8654132228199566e-05, "loss": 0.1511, "step": 11289 }, { "epoch": 1.13, "learning_rate": 4.865225790456372e-05, "loss": 0.9055, "step": 11290 }, { "epoch": 1.13, "learning_rate": 4.865038346223201e-05, "loss": 0.3509, "step": 11291 }, { "epoch": 1.13, "learning_rate": 4.8648508901216346e-05, "loss": 0.352, "step": 11292 }, { "epoch": 1.13, "learning_rate": 4.864663422152867e-05, "loss": 0.2697, "step": 11293 }, { "epoch": 1.13, "learning_rate": 4.8644759423180906e-05, "loss": 0.5591, "step": 11294 }, { "epoch": 1.13, "learning_rate": 4.864288450618499e-05, "loss": 0.4828, "step": 11295 }, { "epoch": 1.13, "learning_rate": 4.864100947055284e-05, "loss": 0.3253, "step": 11296 }, { "epoch": 1.13, "learning_rate": 4.863913431629641e-05, "loss": 0.1824, "step": 11297 }, { "epoch": 1.13, "learning_rate": 4.863725904342761e-05, "loss": 0.6745, "step": 11298 }, { "epoch": 1.13, "learning_rate": 4.8635383651958385e-05, "loss": 0.9355, "step": 11299 }, { "epoch": 1.13, "learning_rate": 4.8633508141900654e-05, "loss": 0.5584, "step": 11300 }, { "epoch": 1.13, "learning_rate": 4.863163251326638e-05, "loss": 0.2347, "step": 11301 }, { "epoch": 1.13, "learning_rate": 4.862975676606749e-05, "loss": 0.7101, "step": 11302 }, { "epoch": 1.13, "learning_rate": 4.8627880900315915e-05, "loss": 0.2539, "step": 11303 }, { "epoch": 1.13, "learning_rate": 4.862600491602358e-05, "loss": 0.1993, "step": 11304 }, { "epoch": 1.13, "learning_rate": 4.862412881320244e-05, "loss": 0.4028, "step": 11305 }, { "epoch": 1.14, "learning_rate": 4.8622252591864434e-05, "loss": 0.8125, "step": 11306 }, { "epoch": 1.14, "learning_rate": 4.86203762520215e-05, "loss": 0.2619, "step": 11307 }, { "epoch": 1.14, "learning_rate": 4.861849979368556e-05, "loss": 0.9084, "step": 11308 }, { "epoch": 1.14, "learning_rate": 4.8616623216868584e-05, "loss": 1.2909, "step": 11309 }, { "epoch": 1.14, "learning_rate": 4.86147465215825e-05, "loss": 1.0769, "step": 11310 }, { "epoch": 1.14, "learning_rate": 4.861286970783924e-05, "loss": 1.3379, "step": 11311 }, { "epoch": 1.14, "learning_rate": 4.861099277565077e-05, "loss": 0.7611, "step": 11312 }, { "epoch": 1.14, "learning_rate": 4.860911572502901e-05, "loss": 1.16, "step": 11313 }, { "epoch": 1.14, "learning_rate": 4.860723855598592e-05, "loss": 1.5837, "step": 11314 }, { "epoch": 1.14, "learning_rate": 4.8605361268533446e-05, "loss": 1.0229, "step": 11315 }, { "epoch": 1.14, "learning_rate": 4.860348386268353e-05, "loss": 0.8782, "step": 11316 }, { "epoch": 1.14, "learning_rate": 4.8601606338448117e-05, "loss": 0.3468, "step": 11317 }, { "epoch": 1.14, "learning_rate": 4.8599728695839163e-05, "loss": 0.6495, "step": 11318 }, { "epoch": 1.14, "learning_rate": 4.8597850934868605e-05, "loss": 0.465, "step": 11319 }, { "epoch": 1.14, "learning_rate": 4.85959730555484e-05, "loss": 0.6848, "step": 11320 }, { "epoch": 1.14, "learning_rate": 4.85940950578905e-05, "loss": 0.7013, "step": 11321 }, { "epoch": 1.14, "learning_rate": 4.8592216941906854e-05, "loss": 1.2167, "step": 11322 }, { "epoch": 1.14, "learning_rate": 4.859033870760941e-05, "loss": 0.7371, "step": 11323 }, { "epoch": 1.14, "learning_rate": 4.858846035501012e-05, "loss": 0.9062, "step": 11324 }, { "epoch": 1.14, "learning_rate": 4.8586581884120943e-05, "loss": 0.277, "step": 11325 }, { "epoch": 1.14, "learning_rate": 4.858470329495383e-05, "loss": 0.2635, "step": 11326 }, { "epoch": 1.14, "learning_rate": 4.858282458752074e-05, "loss": 1.0683, "step": 11327 }, { "epoch": 1.14, "learning_rate": 4.858094576183361e-05, "loss": 0.212, "step": 11328 }, { "epoch": 1.14, "learning_rate": 4.857906681790441e-05, "loss": 0.863, "step": 11329 }, { "epoch": 1.14, "learning_rate": 4.8577187755745114e-05, "loss": 0.5405, "step": 11330 }, { "epoch": 1.14, "learning_rate": 4.8575308575367645e-05, "loss": 0.8945, "step": 11331 }, { "epoch": 1.14, "learning_rate": 4.8573429276783985e-05, "loss": 0.278, "step": 11332 }, { "epoch": 1.14, "learning_rate": 4.8571549860006085e-05, "loss": 0.994, "step": 11333 }, { "epoch": 1.14, "learning_rate": 4.8569670325045913e-05, "loss": 1.191, "step": 11334 }, { "epoch": 1.14, "learning_rate": 4.856779067191542e-05, "loss": 0.8015, "step": 11335 }, { "epoch": 1.14, "learning_rate": 4.856591090062657e-05, "loss": 0.322, "step": 11336 }, { "epoch": 1.14, "learning_rate": 4.856403101119133e-05, "loss": 0.179, "step": 11337 }, { "epoch": 1.14, "learning_rate": 4.856215100362165e-05, "loss": 0.3079, "step": 11338 }, { "epoch": 1.14, "learning_rate": 4.856027087792951e-05, "loss": 0.8167, "step": 11339 }, { "epoch": 1.14, "learning_rate": 4.8558390634126864e-05, "loss": 0.5127, "step": 11340 }, { "epoch": 1.14, "learning_rate": 4.855651027222568e-05, "loss": 0.8371, "step": 11341 }, { "epoch": 1.14, "learning_rate": 4.8554629792237915e-05, "loss": 0.4173, "step": 11342 }, { "epoch": 1.14, "learning_rate": 4.855274919417556e-05, "loss": 0.3386, "step": 11343 }, { "epoch": 1.14, "learning_rate": 4.855086847805056e-05, "loss": 0.4224, "step": 11344 }, { "epoch": 1.14, "learning_rate": 4.854898764387488e-05, "loss": 1.8516, "step": 11345 }, { "epoch": 1.14, "learning_rate": 4.854710669166052e-05, "loss": 0.196, "step": 11346 }, { "epoch": 1.14, "learning_rate": 4.854522562141942e-05, "loss": 0.957, "step": 11347 }, { "epoch": 1.14, "learning_rate": 4.8543344433163556e-05, "loss": 0.4128, "step": 11348 }, { "epoch": 1.14, "learning_rate": 4.85414631269049e-05, "loss": 0.5342, "step": 11349 }, { "epoch": 1.14, "learning_rate": 4.853958170265543e-05, "loss": 0.6714, "step": 11350 }, { "epoch": 1.14, "learning_rate": 4.8537700160427106e-05, "loss": 0.2732, "step": 11351 }, { "epoch": 1.14, "learning_rate": 4.853581850023192e-05, "loss": 0.3368, "step": 11352 }, { "epoch": 1.14, "learning_rate": 4.8533936722081837e-05, "loss": 1.2711, "step": 11353 }, { "epoch": 1.14, "learning_rate": 4.853205482598883e-05, "loss": 0.7858, "step": 11354 }, { "epoch": 1.14, "learning_rate": 4.853017281196488e-05, "loss": 0.8779, "step": 11355 }, { "epoch": 1.14, "learning_rate": 4.852829068002195e-05, "loss": 0.7426, "step": 11356 }, { "epoch": 1.14, "learning_rate": 4.852640843017204e-05, "loss": 0.2225, "step": 11357 }, { "epoch": 1.14, "learning_rate": 4.85245260624271e-05, "loss": 0.5783, "step": 11358 }, { "epoch": 1.14, "learning_rate": 4.852264357679913e-05, "loss": 0.4949, "step": 11359 }, { "epoch": 1.14, "learning_rate": 4.852076097330011e-05, "loss": 0.2701, "step": 11360 }, { "epoch": 1.14, "learning_rate": 4.8518878251941995e-05, "loss": 1.3436, "step": 11361 }, { "epoch": 1.14, "learning_rate": 4.8516995412736796e-05, "loss": 0.9805, "step": 11362 }, { "epoch": 1.14, "learning_rate": 4.851511245569648e-05, "loss": 0.3244, "step": 11363 }, { "epoch": 1.14, "learning_rate": 4.851322938083303e-05, "loss": 0.9843, "step": 11364 }, { "epoch": 1.14, "learning_rate": 4.851134618815843e-05, "loss": 0.4319, "step": 11365 }, { "epoch": 1.14, "learning_rate": 4.850946287768467e-05, "loss": 0.8301, "step": 11366 }, { "epoch": 1.14, "learning_rate": 4.8507579449423734e-05, "loss": 0.4183, "step": 11367 }, { "epoch": 1.14, "learning_rate": 4.850569590338759e-05, "loss": 1.0348, "step": 11368 }, { "epoch": 1.14, "learning_rate": 4.850381223958825e-05, "loss": 0.8828, "step": 11369 }, { "epoch": 1.14, "learning_rate": 4.8501928458037675e-05, "loss": 0.8723, "step": 11370 }, { "epoch": 1.14, "learning_rate": 4.850004455874787e-05, "loss": 0.6621, "step": 11371 }, { "epoch": 1.14, "learning_rate": 4.8498160541730824e-05, "loss": 0.1689, "step": 11372 }, { "epoch": 1.14, "learning_rate": 4.849627640699852e-05, "loss": 0.5118, "step": 11373 }, { "epoch": 1.14, "learning_rate": 4.8494392154562946e-05, "loss": 0.4932, "step": 11374 }, { "epoch": 1.14, "learning_rate": 4.8492507784436104e-05, "loss": 0.9109, "step": 11375 }, { "epoch": 1.14, "learning_rate": 4.8490623296629965e-05, "loss": 0.2767, "step": 11376 }, { "epoch": 1.14, "learning_rate": 4.848873869115654e-05, "loss": 0.3038, "step": 11377 }, { "epoch": 1.14, "learning_rate": 4.848685396802782e-05, "loss": 0.5626, "step": 11378 }, { "epoch": 1.14, "learning_rate": 4.8484969127255786e-05, "loss": 0.7676, "step": 11379 }, { "epoch": 1.14, "learning_rate": 4.8483084168852447e-05, "loss": 1.2732, "step": 11380 }, { "epoch": 1.14, "learning_rate": 4.848119909282979e-05, "loss": 0.462, "step": 11381 }, { "epoch": 1.14, "learning_rate": 4.847931389919982e-05, "loss": 0.16, "step": 11382 }, { "epoch": 1.14, "learning_rate": 4.847742858797451e-05, "loss": 0.7254, "step": 11383 }, { "epoch": 1.14, "learning_rate": 4.8475543159165884e-05, "loss": 0.6961, "step": 11384 }, { "epoch": 1.14, "learning_rate": 4.8473657612785923e-05, "loss": 0.3181, "step": 11385 }, { "epoch": 1.14, "learning_rate": 4.847177194884664e-05, "loss": 0.3739, "step": 11386 }, { "epoch": 1.14, "learning_rate": 4.8469886167360026e-05, "loss": 0.5156, "step": 11387 }, { "epoch": 1.14, "learning_rate": 4.846800026833808e-05, "loss": 0.2179, "step": 11388 }, { "epoch": 1.14, "learning_rate": 4.846611425179281e-05, "loss": 0.7831, "step": 11389 }, { "epoch": 1.14, "learning_rate": 4.84642281177362e-05, "loss": 0.6532, "step": 11390 }, { "epoch": 1.14, "learning_rate": 4.846234186618029e-05, "loss": 0.1255, "step": 11391 }, { "epoch": 1.14, "learning_rate": 4.846045549713704e-05, "loss": 0.3844, "step": 11392 }, { "epoch": 1.14, "learning_rate": 4.845856901061848e-05, "loss": 0.7802, "step": 11393 }, { "epoch": 1.14, "learning_rate": 4.845668240663661e-05, "loss": 0.4046, "step": 11394 }, { "epoch": 1.14, "learning_rate": 4.8454795685203435e-05, "loss": 0.1606, "step": 11395 }, { "epoch": 1.14, "learning_rate": 4.845290884633096e-05, "loss": 0.3858, "step": 11396 }, { "epoch": 1.14, "learning_rate": 4.845102189003119e-05, "loss": 0.5897, "step": 11397 }, { "epoch": 1.14, "learning_rate": 4.844913481631614e-05, "loss": 0.3837, "step": 11398 }, { "epoch": 1.14, "learning_rate": 4.844724762519781e-05, "loss": 0.8948, "step": 11399 }, { "epoch": 1.14, "learning_rate": 4.8445360316688224e-05, "loss": 0.6779, "step": 11400 }, { "epoch": 1.14, "learning_rate": 4.844347289079937e-05, "loss": 0.8034, "step": 11401 }, { "epoch": 1.14, "learning_rate": 4.8441585347543276e-05, "loss": 0.2827, "step": 11402 }, { "epoch": 1.14, "learning_rate": 4.843969768693195e-05, "loss": 0.8499, "step": 11403 }, { "epoch": 1.15, "learning_rate": 4.84378099089774e-05, "loss": 0.5981, "step": 11404 }, { "epoch": 1.15, "learning_rate": 4.843592201369164e-05, "loss": 0.3113, "step": 11405 }, { "epoch": 1.15, "learning_rate": 4.8434034001086686e-05, "loss": 1.1814, "step": 11406 }, { "epoch": 1.15, "learning_rate": 4.8432145871174555e-05, "loss": 0.5209, "step": 11407 }, { "epoch": 1.15, "learning_rate": 4.843025762396726e-05, "loss": 0.348, "step": 11408 }, { "epoch": 1.15, "learning_rate": 4.842836925947682e-05, "loss": 0.1223, "step": 11409 }, { "epoch": 1.15, "learning_rate": 4.8426480777715233e-05, "loss": 0.3053, "step": 11410 }, { "epoch": 1.15, "learning_rate": 4.842459217869455e-05, "loss": 0.7506, "step": 11411 }, { "epoch": 1.15, "learning_rate": 4.842270346242677e-05, "loss": 0.4313, "step": 11412 }, { "epoch": 1.15, "learning_rate": 4.84208146289239e-05, "loss": 0.6462, "step": 11413 }, { "epoch": 1.15, "learning_rate": 4.841892567819798e-05, "loss": 0.9075, "step": 11414 }, { "epoch": 1.15, "learning_rate": 4.8417036610261034e-05, "loss": 0.4294, "step": 11415 }, { "epoch": 1.15, "learning_rate": 4.841514742512506e-05, "loss": 0.5671, "step": 11416 }, { "epoch": 1.15, "learning_rate": 4.84132581228021e-05, "loss": 0.3768, "step": 11417 }, { "epoch": 1.15, "learning_rate": 4.8411368703304176e-05, "loss": 0.4847, "step": 11418 }, { "epoch": 1.15, "learning_rate": 4.840947916664329e-05, "loss": 1.2849, "step": 11419 }, { "epoch": 1.15, "learning_rate": 4.840758951283149e-05, "loss": 0.2843, "step": 11420 }, { "epoch": 1.15, "learning_rate": 4.84056997418808e-05, "loss": 1.0095, "step": 11421 }, { "epoch": 1.15, "learning_rate": 4.8403809853803224e-05, "loss": 1.0818, "step": 11422 }, { "epoch": 1.15, "learning_rate": 4.840191984861081e-05, "loss": 0.6752, "step": 11423 }, { "epoch": 1.15, "learning_rate": 4.8400029726315584e-05, "loss": 0.2628, "step": 11424 }, { "epoch": 1.15, "learning_rate": 4.839813948692956e-05, "loss": 0.8636, "step": 11425 }, { "epoch": 1.15, "learning_rate": 4.8396249130464775e-05, "loss": 1.261, "step": 11426 }, { "epoch": 1.15, "learning_rate": 4.839435865693327e-05, "loss": 1.0082, "step": 11427 }, { "epoch": 1.15, "learning_rate": 4.8392468066347054e-05, "loss": 0.778, "step": 11428 }, { "epoch": 1.15, "learning_rate": 4.839057735871817e-05, "loss": 0.2664, "step": 11429 }, { "epoch": 1.15, "learning_rate": 4.838868653405866e-05, "loss": 0.853, "step": 11430 }, { "epoch": 1.15, "learning_rate": 4.8386795592380524e-05, "loss": 0.9751, "step": 11431 }, { "epoch": 1.15, "learning_rate": 4.8384904533695825e-05, "loss": 0.4572, "step": 11432 }, { "epoch": 1.15, "learning_rate": 4.83830133580166e-05, "loss": 0.3841, "step": 11433 }, { "epoch": 1.15, "learning_rate": 4.838112206535484e-05, "loss": 0.2552, "step": 11434 }, { "epoch": 1.15, "learning_rate": 4.8379230655722636e-05, "loss": 0.5728, "step": 11435 }, { "epoch": 1.15, "learning_rate": 4.8377339129132e-05, "loss": 0.666, "step": 11436 }, { "epoch": 1.15, "learning_rate": 4.837544748559496e-05, "loss": 0.7419, "step": 11437 }, { "epoch": 1.15, "learning_rate": 4.8373555725123565e-05, "loss": 0.7436, "step": 11438 }, { "epoch": 1.15, "learning_rate": 4.837166384772985e-05, "loss": 0.4495, "step": 11439 }, { "epoch": 1.15, "learning_rate": 4.8369771853425864e-05, "loss": 0.9016, "step": 11440 }, { "epoch": 1.15, "learning_rate": 4.836787974222363e-05, "loss": 0.4296, "step": 11441 }, { "epoch": 1.15, "learning_rate": 4.836598751413519e-05, "loss": 0.1662, "step": 11442 }, { "epoch": 1.15, "learning_rate": 4.83640951691726e-05, "loss": 0.4954, "step": 11443 }, { "epoch": 1.15, "learning_rate": 4.83622027073479e-05, "loss": 1.076, "step": 11444 }, { "epoch": 1.15, "learning_rate": 4.836031012867312e-05, "loss": 0.8958, "step": 11445 }, { "epoch": 1.15, "learning_rate": 4.835841743316032e-05, "loss": 0.163, "step": 11446 }, { "epoch": 1.15, "learning_rate": 4.8356524620821526e-05, "loss": 0.2797, "step": 11447 }, { "epoch": 1.15, "learning_rate": 4.835463169166879e-05, "loss": 0.3504, "step": 11448 }, { "epoch": 1.15, "learning_rate": 4.835273864571416e-05, "loss": 0.4508, "step": 11449 }, { "epoch": 1.15, "learning_rate": 4.83508454829697e-05, "loss": 0.3406, "step": 11450 }, { "epoch": 1.15, "learning_rate": 4.834895220344742e-05, "loss": 0.4949, "step": 11451 }, { "epoch": 1.15, "learning_rate": 4.8347058807159395e-05, "loss": 0.2949, "step": 11452 }, { "epoch": 1.15, "learning_rate": 4.834516529411767e-05, "loss": 0.9062, "step": 11453 }, { "epoch": 1.15, "learning_rate": 4.834327166433429e-05, "loss": 0.5607, "step": 11454 }, { "epoch": 1.15, "learning_rate": 4.83413779178213e-05, "loss": 1.0292, "step": 11455 }, { "epoch": 1.15, "learning_rate": 4.8339484054590775e-05, "loss": 0.3558, "step": 11456 }, { "epoch": 1.15, "learning_rate": 4.8337590074654735e-05, "loss": 0.4024, "step": 11457 }, { "epoch": 1.15, "learning_rate": 4.833569597802526e-05, "loss": 0.382, "step": 11458 }, { "epoch": 1.15, "learning_rate": 4.833380176471439e-05, "loss": 0.6206, "step": 11459 }, { "epoch": 1.15, "learning_rate": 4.8331907434734174e-05, "loss": 0.7998, "step": 11460 }, { "epoch": 1.15, "learning_rate": 4.8330012988096675e-05, "loss": 0.8299, "step": 11461 }, { "epoch": 1.15, "learning_rate": 4.832811842481394e-05, "loss": 0.429, "step": 11462 }, { "epoch": 1.15, "learning_rate": 4.832622374489804e-05, "loss": 0.9602, "step": 11463 }, { "epoch": 1.15, "learning_rate": 4.832432894836102e-05, "loss": 0.2925, "step": 11464 }, { "epoch": 1.15, "learning_rate": 4.8322434035214944e-05, "loss": 0.1985, "step": 11465 }, { "epoch": 1.15, "learning_rate": 4.832053900547187e-05, "loss": 0.1888, "step": 11466 }, { "epoch": 1.15, "learning_rate": 4.831864385914385e-05, "loss": 0.3651, "step": 11467 }, { "epoch": 1.15, "learning_rate": 4.831674859624295e-05, "loss": 0.371, "step": 11468 }, { "epoch": 1.15, "learning_rate": 4.8314853216781236e-05, "loss": 0.9125, "step": 11469 }, { "epoch": 1.15, "learning_rate": 4.831295772077076e-05, "loss": 1.017, "step": 11470 }, { "epoch": 1.15, "learning_rate": 4.831106210822358e-05, "loss": 0.3499, "step": 11471 }, { "epoch": 1.15, "learning_rate": 4.8309166379151776e-05, "loss": 0.5011, "step": 11472 }, { "epoch": 1.15, "learning_rate": 4.830727053356741e-05, "loss": 0.6407, "step": 11473 }, { "epoch": 1.15, "learning_rate": 4.830537457148251e-05, "loss": 0.3915, "step": 11474 }, { "epoch": 1.15, "learning_rate": 4.8303478492909196e-05, "loss": 0.4894, "step": 11475 }, { "epoch": 1.15, "learning_rate": 4.830158229785951e-05, "loss": 0.7144, "step": 11476 }, { "epoch": 1.15, "learning_rate": 4.8299685986345506e-05, "loss": 0.6727, "step": 11477 }, { "epoch": 1.15, "learning_rate": 4.829778955837926e-05, "loss": 0.1533, "step": 11478 }, { "epoch": 1.15, "learning_rate": 4.829589301397284e-05, "loss": 0.3964, "step": 11479 }, { "epoch": 1.15, "learning_rate": 4.829399635313832e-05, "loss": 0.9354, "step": 11480 }, { "epoch": 1.15, "learning_rate": 4.8292099575887776e-05, "loss": 0.3763, "step": 11481 }, { "epoch": 1.15, "learning_rate": 4.829020268223326e-05, "loss": 0.6658, "step": 11482 }, { "epoch": 1.15, "learning_rate": 4.828830567218685e-05, "loss": 1.0599, "step": 11483 }, { "epoch": 1.15, "learning_rate": 4.828640854576062e-05, "loss": 0.2589, "step": 11484 }, { "epoch": 1.15, "learning_rate": 4.8284511302966655e-05, "loss": 1.1044, "step": 11485 }, { "epoch": 1.15, "learning_rate": 4.8282613943817e-05, "loss": 0.1996, "step": 11486 }, { "epoch": 1.15, "learning_rate": 4.828071646832375e-05, "loss": 0.429, "step": 11487 }, { "epoch": 1.15, "learning_rate": 4.827881887649898e-05, "loss": 0.4594, "step": 11488 }, { "epoch": 1.15, "learning_rate": 4.8276921168354753e-05, "loss": 1.0007, "step": 11489 }, { "epoch": 1.15, "learning_rate": 4.827502334390316e-05, "loss": 1.0558, "step": 11490 }, { "epoch": 1.15, "learning_rate": 4.827312540315627e-05, "loss": 0.4223, "step": 11491 }, { "epoch": 1.15, "learning_rate": 4.8271227346126157e-05, "loss": 0.8511, "step": 11492 }, { "epoch": 1.15, "learning_rate": 4.82693291728249e-05, "loss": 0.4455, "step": 11493 }, { "epoch": 1.15, "learning_rate": 4.826743088326459e-05, "loss": 0.317, "step": 11494 }, { "epoch": 1.15, "learning_rate": 4.8265532477457295e-05, "loss": 0.3478, "step": 11495 }, { "epoch": 1.15, "learning_rate": 4.82636339554151e-05, "loss": 0.7718, "step": 11496 }, { "epoch": 1.15, "learning_rate": 4.826173531715008e-05, "loss": 0.2378, "step": 11497 }, { "epoch": 1.15, "learning_rate": 4.8259836562674344e-05, "loss": 1.1812, "step": 11498 }, { "epoch": 1.15, "learning_rate": 4.8257937691999936e-05, "loss": 0.3063, "step": 11499 }, { "epoch": 1.15, "learning_rate": 4.825603870513896e-05, "loss": 1.3913, "step": 11500 }, { "epoch": 1.15, "learning_rate": 4.825413960210351e-05, "loss": 0.7309, "step": 11501 }, { "epoch": 1.15, "learning_rate": 4.825224038290565e-05, "loss": 1.1251, "step": 11502 }, { "epoch": 1.16, "learning_rate": 4.825034104755748e-05, "loss": 0.4443, "step": 11503 }, { "epoch": 1.16, "learning_rate": 4.8248441596071074e-05, "loss": 0.9789, "step": 11504 }, { "epoch": 1.16, "learning_rate": 4.824654202845854e-05, "loss": 0.1705, "step": 11505 }, { "epoch": 1.16, "learning_rate": 4.8244642344731944e-05, "loss": 0.6294, "step": 11506 }, { "epoch": 1.16, "learning_rate": 4.8242742544903396e-05, "loss": 0.7717, "step": 11507 }, { "epoch": 1.16, "learning_rate": 4.8240842628984965e-05, "loss": 0.3081, "step": 11508 }, { "epoch": 1.16, "learning_rate": 4.823894259698875e-05, "loss": 0.9288, "step": 11509 }, { "epoch": 1.16, "learning_rate": 4.8237042448926846e-05, "loss": 0.383, "step": 11510 }, { "epoch": 1.16, "learning_rate": 4.8235142184811344e-05, "loss": 0.9079, "step": 11511 }, { "epoch": 1.16, "learning_rate": 4.8233241804654334e-05, "loss": 0.1762, "step": 11512 }, { "epoch": 1.16, "learning_rate": 4.82313413084679e-05, "loss": 0.7783, "step": 11513 }, { "epoch": 1.16, "learning_rate": 4.822944069626416e-05, "loss": 0.8436, "step": 11514 }, { "epoch": 1.16, "learning_rate": 4.822753996805519e-05, "loss": 0.625, "step": 11515 }, { "epoch": 1.16, "learning_rate": 4.822563912385308e-05, "loss": 0.1413, "step": 11516 }, { "epoch": 1.16, "learning_rate": 4.8223738163669935e-05, "loss": 0.4337, "step": 11517 }, { "epoch": 1.16, "learning_rate": 4.8221837087517864e-05, "loss": 0.243, "step": 11518 }, { "epoch": 1.16, "learning_rate": 4.821993589540895e-05, "loss": 1.0202, "step": 11519 }, { "epoch": 1.16, "learning_rate": 4.821803458735529e-05, "loss": 0.5085, "step": 11520 }, { "epoch": 1.16, "learning_rate": 4.8216133163369e-05, "loss": 0.93, "step": 11521 }, { "epoch": 1.16, "learning_rate": 4.8214231623462167e-05, "loss": 0.6219, "step": 11522 }, { "epoch": 1.16, "learning_rate": 4.821232996764688e-05, "loss": 0.5527, "step": 11523 }, { "epoch": 1.16, "learning_rate": 4.821042819593527e-05, "loss": 0.2764, "step": 11524 }, { "epoch": 1.16, "learning_rate": 4.820852630833941e-05, "loss": 0.2195, "step": 11525 }, { "epoch": 1.16, "learning_rate": 4.8206624304871416e-05, "loss": 0.5787, "step": 11526 }, { "epoch": 1.16, "learning_rate": 4.82047221855434e-05, "loss": 0.431, "step": 11527 }, { "epoch": 1.16, "learning_rate": 4.820281995036745e-05, "loss": 1.1669, "step": 11528 }, { "epoch": 1.16, "learning_rate": 4.8200917599355676e-05, "loss": 0.8578, "step": 11529 }, { "epoch": 1.16, "learning_rate": 4.8199015132520187e-05, "loss": 0.2197, "step": 11530 }, { "epoch": 1.16, "learning_rate": 4.8197112549873094e-05, "loss": 0.7603, "step": 11531 }, { "epoch": 1.16, "learning_rate": 4.81952098514265e-05, "loss": 0.2816, "step": 11532 }, { "epoch": 1.16, "learning_rate": 4.819330703719251e-05, "loss": 0.1132, "step": 11533 }, { "epoch": 1.16, "learning_rate": 4.8191404107183236e-05, "loss": 0.7282, "step": 11534 }, { "epoch": 1.16, "learning_rate": 4.818950106141078e-05, "loss": 1.0349, "step": 11535 }, { "epoch": 1.16, "learning_rate": 4.818759789988727e-05, "loss": 0.2886, "step": 11536 }, { "epoch": 1.16, "learning_rate": 4.8185694622624794e-05, "loss": 0.1266, "step": 11537 }, { "epoch": 1.16, "learning_rate": 4.818379122963549e-05, "loss": 0.8667, "step": 11538 }, { "epoch": 1.16, "learning_rate": 4.818188772093145e-05, "loss": 1.1743, "step": 11539 }, { "epoch": 1.16, "learning_rate": 4.81799840965248e-05, "loss": 0.3446, "step": 11540 }, { "epoch": 1.16, "learning_rate": 4.8178080356427626e-05, "loss": 0.8348, "step": 11541 }, { "epoch": 1.16, "learning_rate": 4.817617650065208e-05, "loss": 0.7292, "step": 11542 }, { "epoch": 1.16, "learning_rate": 4.817427252921025e-05, "loss": 0.9917, "step": 11543 }, { "epoch": 1.16, "learning_rate": 4.817236844211428e-05, "loss": 0.3436, "step": 11544 }, { "epoch": 1.16, "learning_rate": 4.817046423937626e-05, "loss": 1.037, "step": 11545 }, { "epoch": 1.16, "learning_rate": 4.816855992100832e-05, "loss": 0.381, "step": 11546 }, { "epoch": 1.16, "learning_rate": 4.816665548702258e-05, "loss": 0.971, "step": 11547 }, { "epoch": 1.16, "learning_rate": 4.816475093743114e-05, "loss": 0.3491, "step": 11548 }, { "epoch": 1.16, "learning_rate": 4.816284627224615e-05, "loss": 0.8991, "step": 11549 }, { "epoch": 1.16, "learning_rate": 4.8160941491479714e-05, "loss": 0.5589, "step": 11550 }, { "epoch": 1.16, "learning_rate": 4.815903659514396e-05, "loss": 0.9376, "step": 11551 }, { "epoch": 1.16, "learning_rate": 4.8157131583250995e-05, "loss": 0.9989, "step": 11552 }, { "epoch": 1.16, "learning_rate": 4.815522645581296e-05, "loss": 1.0465, "step": 11553 }, { "epoch": 1.16, "learning_rate": 4.815332121284196e-05, "loss": 0.3444, "step": 11554 }, { "epoch": 1.16, "learning_rate": 4.8151415854350135e-05, "loss": 0.3068, "step": 11555 }, { "epoch": 1.16, "learning_rate": 4.8149510380349606e-05, "loss": 0.2458, "step": 11556 }, { "epoch": 1.16, "learning_rate": 4.81476047908525e-05, "loss": 0.1687, "step": 11557 }, { "epoch": 1.16, "learning_rate": 4.814569908587094e-05, "loss": 0.3508, "step": 11558 }, { "epoch": 1.16, "learning_rate": 4.814379326541705e-05, "loss": 0.2053, "step": 11559 }, { "epoch": 1.16, "learning_rate": 4.8141887329502964e-05, "loss": 0.6721, "step": 11560 }, { "epoch": 1.16, "learning_rate": 4.81399812781408e-05, "loss": 0.4152, "step": 11561 }, { "epoch": 1.16, "learning_rate": 4.8138075111342715e-05, "loss": 0.1721, "step": 11562 }, { "epoch": 1.16, "learning_rate": 4.813616882912081e-05, "loss": 1.0764, "step": 11563 }, { "epoch": 1.16, "learning_rate": 4.813426243148722e-05, "loss": 0.1659, "step": 11564 }, { "epoch": 1.16, "learning_rate": 4.8132355918454085e-05, "loss": 0.4457, "step": 11565 }, { "epoch": 1.16, "learning_rate": 4.8130449290033544e-05, "loss": 0.5814, "step": 11566 }, { "epoch": 1.16, "learning_rate": 4.812854254623771e-05, "loss": 0.3408, "step": 11567 }, { "epoch": 1.16, "learning_rate": 4.812663568707874e-05, "loss": 1.1188, "step": 11568 }, { "epoch": 1.16, "learning_rate": 4.8124728712568744e-05, "loss": 0.4922, "step": 11569 }, { "epoch": 1.16, "learning_rate": 4.8122821622719877e-05, "loss": 1.1996, "step": 11570 }, { "epoch": 1.16, "learning_rate": 4.812091441754426e-05, "loss": 0.6596, "step": 11571 }, { "epoch": 1.16, "learning_rate": 4.811900709705404e-05, "loss": 0.5573, "step": 11572 }, { "epoch": 1.16, "learning_rate": 4.811709966126136e-05, "loss": 0.2768, "step": 11573 }, { "epoch": 1.16, "learning_rate": 4.8115192110178334e-05, "loss": 1.1681, "step": 11574 }, { "epoch": 1.16, "learning_rate": 4.811328444381713e-05, "loss": 0.8278, "step": 11575 }, { "epoch": 1.16, "learning_rate": 4.8111376662189876e-05, "loss": 0.3968, "step": 11576 }, { "epoch": 1.16, "learning_rate": 4.8109468765308705e-05, "loss": 0.9234, "step": 11577 }, { "epoch": 1.16, "learning_rate": 4.8107560753185766e-05, "loss": 0.3934, "step": 11578 }, { "epoch": 1.16, "learning_rate": 4.810565262583319e-05, "loss": 0.3132, "step": 11579 }, { "epoch": 1.16, "learning_rate": 4.8103744383263136e-05, "loss": 0.5568, "step": 11580 }, { "epoch": 1.16, "learning_rate": 4.810183602548774e-05, "loss": 1.1603, "step": 11581 }, { "epoch": 1.16, "learning_rate": 4.8099927552519155e-05, "loss": 0.2463, "step": 11582 }, { "epoch": 1.16, "learning_rate": 4.8098018964369496e-05, "loss": 1.1873, "step": 11583 }, { "epoch": 1.16, "learning_rate": 4.8096110261050936e-05, "loss": 0.6183, "step": 11584 }, { "epoch": 1.16, "learning_rate": 4.809420144257563e-05, "loss": 0.6985, "step": 11585 }, { "epoch": 1.16, "learning_rate": 4.809229250895569e-05, "loss": 1.12, "step": 11586 }, { "epoch": 1.16, "learning_rate": 4.809038346020329e-05, "loss": 0.7787, "step": 11587 }, { "epoch": 1.16, "learning_rate": 4.8088474296330574e-05, "loss": 0.9033, "step": 11588 }, { "epoch": 1.16, "learning_rate": 4.808656501734969e-05, "loss": 0.0818, "step": 11589 }, { "epoch": 1.16, "learning_rate": 4.808465562327278e-05, "loss": 1.2613, "step": 11590 }, { "epoch": 1.16, "learning_rate": 4.8082746114112006e-05, "loss": 0.3404, "step": 11591 }, { "epoch": 1.16, "learning_rate": 4.808083648987952e-05, "loss": 0.54, "step": 11592 }, { "epoch": 1.16, "learning_rate": 4.8078926750587454e-05, "loss": 0.2232, "step": 11593 }, { "epoch": 1.16, "learning_rate": 4.807701689624799e-05, "loss": 0.8548, "step": 11594 }, { "epoch": 1.16, "learning_rate": 4.807510692687326e-05, "loss": 0.7906, "step": 11595 }, { "epoch": 1.16, "learning_rate": 4.807319684247543e-05, "loss": 0.9725, "step": 11596 }, { "epoch": 1.16, "learning_rate": 4.8071286643066644e-05, "loss": 0.3155, "step": 11597 }, { "epoch": 1.16, "learning_rate": 4.806937632865906e-05, "loss": 1.1856, "step": 11598 }, { "epoch": 1.16, "learning_rate": 4.806746589926486e-05, "loss": 1.1084, "step": 11599 }, { "epoch": 1.16, "learning_rate": 4.8065555354896166e-05, "loss": 0.4111, "step": 11600 }, { "epoch": 1.17, "learning_rate": 4.806364469556515e-05, "loss": 0.7584, "step": 11601 }, { "epoch": 1.17, "learning_rate": 4.8061733921283975e-05, "loss": 0.3317, "step": 11602 }, { "epoch": 1.17, "learning_rate": 4.80598230320648e-05, "loss": 0.6348, "step": 11603 }, { "epoch": 1.17, "learning_rate": 4.8057912027919763e-05, "loss": 0.1702, "step": 11604 }, { "epoch": 1.17, "learning_rate": 4.805600090886106e-05, "loss": 0.485, "step": 11605 }, { "epoch": 1.17, "learning_rate": 4.805408967490084e-05, "loss": 0.9046, "step": 11606 }, { "epoch": 1.17, "learning_rate": 4.8052178326051245e-05, "loss": 0.6153, "step": 11607 }, { "epoch": 1.17, "learning_rate": 4.805026686232447e-05, "loss": 0.9385, "step": 11608 }, { "epoch": 1.17, "learning_rate": 4.8048355283732655e-05, "loss": 0.8132, "step": 11609 }, { "epoch": 1.17, "learning_rate": 4.8046443590287976e-05, "loss": 0.6253, "step": 11610 }, { "epoch": 1.17, "learning_rate": 4.804453178200259e-05, "loss": 0.8497, "step": 11611 }, { "epoch": 1.17, "learning_rate": 4.8042619858888676e-05, "loss": 0.8479, "step": 11612 }, { "epoch": 1.17, "learning_rate": 4.804070782095839e-05, "loss": 0.2726, "step": 11613 }, { "epoch": 1.17, "learning_rate": 4.80387956682239e-05, "loss": 0.3014, "step": 11614 }, { "epoch": 1.17, "learning_rate": 4.803688340069738e-05, "loss": 0.2056, "step": 11615 }, { "epoch": 1.17, "learning_rate": 4.803497101839099e-05, "loss": 0.9884, "step": 11616 }, { "epoch": 1.17, "learning_rate": 4.803305852131691e-05, "loss": 0.3381, "step": 11617 }, { "epoch": 1.17, "learning_rate": 4.80311459094873e-05, "loss": 0.3525, "step": 11618 }, { "epoch": 1.17, "learning_rate": 4.802923318291435e-05, "loss": 0.3986, "step": 11619 }, { "epoch": 1.17, "learning_rate": 4.8027320341610205e-05, "loss": 0.5482, "step": 11620 }, { "epoch": 1.17, "learning_rate": 4.802540738558706e-05, "loss": 0.3292, "step": 11621 }, { "epoch": 1.17, "learning_rate": 4.802349431485708e-05, "loss": 0.7738, "step": 11622 }, { "epoch": 1.17, "learning_rate": 4.8021581129432424e-05, "loss": 1.0379, "step": 11623 }, { "epoch": 1.17, "learning_rate": 4.80196678293253e-05, "loss": 0.5044, "step": 11624 }, { "epoch": 1.17, "learning_rate": 4.801775441454785e-05, "loss": 0.9888, "step": 11625 }, { "epoch": 1.17, "learning_rate": 4.8015840885112274e-05, "loss": 0.5979, "step": 11626 }, { "epoch": 1.17, "learning_rate": 4.801392724103074e-05, "loss": 0.4666, "step": 11627 }, { "epoch": 1.17, "learning_rate": 4.801201348231543e-05, "loss": 1.056, "step": 11628 }, { "epoch": 1.17, "learning_rate": 4.8010099608978505e-05, "loss": 0.4083, "step": 11629 }, { "epoch": 1.17, "learning_rate": 4.800818562103217e-05, "loss": 1.1235, "step": 11630 }, { "epoch": 1.17, "learning_rate": 4.800627151848859e-05, "loss": 0.2233, "step": 11631 }, { "epoch": 1.17, "learning_rate": 4.800435730135994e-05, "loss": 0.3479, "step": 11632 }, { "epoch": 1.17, "learning_rate": 4.8002442969658417e-05, "loss": 0.3796, "step": 11633 }, { "epoch": 1.17, "learning_rate": 4.80005285233962e-05, "loss": 0.5032, "step": 11634 }, { "epoch": 1.17, "learning_rate": 4.799861396258545e-05, "loss": 1.1485, "step": 11635 }, { "epoch": 1.17, "learning_rate": 4.7996699287238384e-05, "loss": 0.6765, "step": 11636 }, { "epoch": 1.17, "learning_rate": 4.799478449736717e-05, "loss": 0.3958, "step": 11637 }, { "epoch": 1.17, "learning_rate": 4.799286959298398e-05, "loss": 0.452, "step": 11638 }, { "epoch": 1.17, "learning_rate": 4.7990954574101025e-05, "loss": 0.5083, "step": 11639 }, { "epoch": 1.17, "learning_rate": 4.798903944073047e-05, "loss": 0.9868, "step": 11640 }, { "epoch": 1.17, "learning_rate": 4.798712419288452e-05, "loss": 0.4243, "step": 11641 }, { "epoch": 1.17, "learning_rate": 4.798520883057535e-05, "loss": 0.1331, "step": 11642 }, { "epoch": 1.17, "learning_rate": 4.7983293353815155e-05, "loss": 0.1231, "step": 11643 }, { "epoch": 1.17, "learning_rate": 4.7981377762616124e-05, "loss": 0.7554, "step": 11644 }, { "epoch": 1.17, "learning_rate": 4.7979462056990436e-05, "loss": 1.0921, "step": 11645 }, { "epoch": 1.17, "learning_rate": 4.7977546236950296e-05, "loss": 0.905, "step": 11646 }, { "epoch": 1.17, "learning_rate": 4.7975630302507894e-05, "loss": 0.2457, "step": 11647 }, { "epoch": 1.17, "learning_rate": 4.797371425367542e-05, "loss": 1.6196, "step": 11648 }, { "epoch": 1.17, "learning_rate": 4.7971798090465065e-05, "loss": 0.6861, "step": 11649 }, { "epoch": 1.17, "learning_rate": 4.796988181288901e-05, "loss": 0.4753, "step": 11650 }, { "epoch": 1.17, "learning_rate": 4.796796542095948e-05, "loss": 0.7694, "step": 11651 }, { "epoch": 1.17, "learning_rate": 4.7966048914688646e-05, "loss": 0.7253, "step": 11652 }, { "epoch": 1.17, "learning_rate": 4.796413229408871e-05, "loss": 0.8736, "step": 11653 }, { "epoch": 1.17, "learning_rate": 4.796221555917188e-05, "loss": 0.3538, "step": 11654 }, { "epoch": 1.17, "learning_rate": 4.796029870995033e-05, "loss": 0.4274, "step": 11655 }, { "epoch": 1.17, "learning_rate": 4.7958381746436285e-05, "loss": 0.744, "step": 11656 }, { "epoch": 1.17, "learning_rate": 4.795646466864192e-05, "loss": 0.5205, "step": 11657 }, { "epoch": 1.17, "learning_rate": 4.7954547476579454e-05, "loss": 0.726, "step": 11658 }, { "epoch": 1.17, "learning_rate": 4.795263017026107e-05, "loss": 0.5357, "step": 11659 }, { "epoch": 1.17, "learning_rate": 4.795071274969898e-05, "loss": 0.6295, "step": 11660 }, { "epoch": 1.17, "learning_rate": 4.794879521490539e-05, "loss": 0.9844, "step": 11661 }, { "epoch": 1.17, "learning_rate": 4.7946877565892484e-05, "loss": 0.3808, "step": 11662 }, { "epoch": 1.17, "learning_rate": 4.794495980267248e-05, "loss": 0.7531, "step": 11663 }, { "epoch": 1.17, "learning_rate": 4.7943041925257575e-05, "loss": 1.2704, "step": 11664 }, { "epoch": 1.17, "learning_rate": 4.794112393365998e-05, "loss": 0.8241, "step": 11665 }, { "epoch": 1.17, "learning_rate": 4.79392058278919e-05, "loss": 0.5451, "step": 11666 }, { "epoch": 1.17, "learning_rate": 4.793728760796554e-05, "loss": 0.3599, "step": 11667 }, { "epoch": 1.17, "learning_rate": 4.7935369273893095e-05, "loss": 1.0755, "step": 11668 }, { "epoch": 1.17, "learning_rate": 4.7933450825686796e-05, "loss": 0.5637, "step": 11669 }, { "epoch": 1.17, "learning_rate": 4.7931532263358834e-05, "loss": 0.9113, "step": 11670 }, { "epoch": 1.17, "learning_rate": 4.792961358692142e-05, "loss": 1.2782, "step": 11671 }, { "epoch": 1.17, "learning_rate": 4.792769479638676e-05, "loss": 0.8429, "step": 11672 }, { "epoch": 1.17, "learning_rate": 4.792577589176708e-05, "loss": 0.2975, "step": 11673 }, { "epoch": 1.17, "learning_rate": 4.792385687307458e-05, "loss": 0.246, "step": 11674 }, { "epoch": 1.17, "learning_rate": 4.792193774032147e-05, "loss": 1.6067, "step": 11675 }, { "epoch": 1.17, "learning_rate": 4.792001849351997e-05, "loss": 0.4923, "step": 11676 }, { "epoch": 1.17, "learning_rate": 4.791809913268229e-05, "loss": 0.7289, "step": 11677 }, { "epoch": 1.17, "learning_rate": 4.791617965782064e-05, "loss": 0.6871, "step": 11678 }, { "epoch": 1.17, "learning_rate": 4.7914260068947235e-05, "loss": 0.393, "step": 11679 }, { "epoch": 1.17, "learning_rate": 4.79123403660743e-05, "loss": 0.7649, "step": 11680 }, { "epoch": 1.17, "learning_rate": 4.791042054921404e-05, "loss": 0.6916, "step": 11681 }, { "epoch": 1.17, "learning_rate": 4.7908500618378684e-05, "loss": 1.0905, "step": 11682 }, { "epoch": 1.17, "learning_rate": 4.790658057358044e-05, "loss": 0.4084, "step": 11683 }, { "epoch": 1.17, "learning_rate": 4.790466041483153e-05, "loss": 0.6166, "step": 11684 }, { "epoch": 1.17, "learning_rate": 4.7902740142144164e-05, "loss": 0.6745, "step": 11685 }, { "epoch": 1.17, "learning_rate": 4.790081975553058e-05, "loss": 0.6851, "step": 11686 }, { "epoch": 1.17, "learning_rate": 4.789889925500298e-05, "loss": 1.1651, "step": 11687 }, { "epoch": 1.17, "learning_rate": 4.7896978640573605e-05, "loss": 0.8454, "step": 11688 }, { "epoch": 1.17, "learning_rate": 4.7895057912254664e-05, "loss": 1.3057, "step": 11689 }, { "epoch": 1.17, "learning_rate": 4.789313707005838e-05, "loss": 0.2815, "step": 11690 }, { "epoch": 1.17, "learning_rate": 4.789121611399697e-05, "loss": 0.9114, "step": 11691 }, { "epoch": 1.17, "learning_rate": 4.788929504408268e-05, "loss": 0.6912, "step": 11692 }, { "epoch": 1.17, "learning_rate": 4.788737386032772e-05, "loss": 0.4553, "step": 11693 }, { "epoch": 1.17, "learning_rate": 4.788545256274431e-05, "loss": 0.5052, "step": 11694 }, { "epoch": 1.17, "learning_rate": 4.788353115134469e-05, "loss": 0.4168, "step": 11695 }, { "epoch": 1.17, "learning_rate": 4.788160962614108e-05, "loss": 1.2002, "step": 11696 }, { "epoch": 1.17, "learning_rate": 4.78796879871457e-05, "loss": 0.2916, "step": 11697 }, { "epoch": 1.17, "learning_rate": 4.78777662343708e-05, "loss": 0.6752, "step": 11698 }, { "epoch": 1.17, "learning_rate": 4.78758443678286e-05, "loss": 0.9137, "step": 11699 }, { "epoch": 1.18, "learning_rate": 4.787392238753131e-05, "loss": 0.7433, "step": 11700 }, { "epoch": 1.18, "learning_rate": 4.787200029349118e-05, "loss": 0.9021, "step": 11701 }, { "epoch": 1.18, "learning_rate": 4.787007808572045e-05, "loss": 0.2358, "step": 11702 }, { "epoch": 1.18, "learning_rate": 4.7868155764231344e-05, "loss": 1.0788, "step": 11703 }, { "epoch": 1.18, "learning_rate": 4.7866233329036084e-05, "loss": 0.3399, "step": 11704 }, { "epoch": 1.18, "learning_rate": 4.7864310780146905e-05, "loss": 0.6846, "step": 11705 }, { "epoch": 1.18, "learning_rate": 4.786238811757606e-05, "loss": 1.0143, "step": 11706 }, { "epoch": 1.18, "learning_rate": 4.786046534133576e-05, "loss": 0.2913, "step": 11707 }, { "epoch": 1.18, "learning_rate": 4.785854245143827e-05, "loss": 0.1321, "step": 11708 }, { "epoch": 1.18, "learning_rate": 4.78566194478958e-05, "loss": 0.5115, "step": 11709 }, { "epoch": 1.18, "learning_rate": 4.7854696330720596e-05, "loss": 0.2043, "step": 11710 }, { "epoch": 1.18, "learning_rate": 4.78527730999249e-05, "loss": 0.3778, "step": 11711 }, { "epoch": 1.18, "learning_rate": 4.7850849755520944e-05, "loss": 0.4896, "step": 11712 }, { "epoch": 1.18, "learning_rate": 4.784892629752097e-05, "loss": 0.4963, "step": 11713 }, { "epoch": 1.18, "learning_rate": 4.7847002725937224e-05, "loss": 0.8581, "step": 11714 }, { "epoch": 1.18, "learning_rate": 4.784507904078194e-05, "loss": 0.9499, "step": 11715 }, { "epoch": 1.18, "learning_rate": 4.784315524206736e-05, "loss": 0.6918, "step": 11716 }, { "epoch": 1.18, "learning_rate": 4.7841231329805726e-05, "loss": 0.7226, "step": 11717 }, { "epoch": 1.18, "learning_rate": 4.783930730400929e-05, "loss": 0.8515, "step": 11718 }, { "epoch": 1.18, "learning_rate": 4.783738316469029e-05, "loss": 0.1902, "step": 11719 }, { "epoch": 1.18, "learning_rate": 4.783545891186096e-05, "loss": 0.6407, "step": 11720 }, { "epoch": 1.18, "learning_rate": 4.783353454553355e-05, "loss": 0.4289, "step": 11721 }, { "epoch": 1.18, "learning_rate": 4.783161006572033e-05, "loss": 0.9431, "step": 11722 }, { "epoch": 1.18, "learning_rate": 4.7829685472433513e-05, "loss": 0.6161, "step": 11723 }, { "epoch": 1.18, "learning_rate": 4.782776076568536e-05, "loss": 0.6397, "step": 11724 }, { "epoch": 1.18, "learning_rate": 4.782583594548813e-05, "loss": 0.4638, "step": 11725 }, { "epoch": 1.18, "learning_rate": 4.782391101185405e-05, "loss": 1.1248, "step": 11726 }, { "epoch": 1.18, "learning_rate": 4.7821985964795385e-05, "loss": 0.5525, "step": 11727 }, { "epoch": 1.18, "learning_rate": 4.782006080432438e-05, "loss": 1.5129, "step": 11728 }, { "epoch": 1.18, "learning_rate": 4.781813553045328e-05, "loss": 0.3945, "step": 11729 }, { "epoch": 1.18, "learning_rate": 4.781621014319436e-05, "loss": 0.768, "step": 11730 }, { "epoch": 1.18, "learning_rate": 4.7814284642559855e-05, "loss": 0.8634, "step": 11731 }, { "epoch": 1.18, "learning_rate": 4.7812359028562e-05, "loss": 0.7927, "step": 11732 }, { "epoch": 1.18, "learning_rate": 4.7810433301213084e-05, "loss": 0.2897, "step": 11733 }, { "epoch": 1.18, "learning_rate": 4.780850746052534e-05, "loss": 0.5804, "step": 11734 }, { "epoch": 1.18, "learning_rate": 4.780658150651104e-05, "loss": 0.5069, "step": 11735 }, { "epoch": 1.18, "learning_rate": 4.780465543918242e-05, "loss": 0.3071, "step": 11736 }, { "epoch": 1.18, "learning_rate": 4.780272925855175e-05, "loss": 0.4358, "step": 11737 }, { "epoch": 1.18, "learning_rate": 4.780080296463128e-05, "loss": 0.2412, "step": 11738 }, { "epoch": 1.18, "learning_rate": 4.779887655743327e-05, "loss": 0.6931, "step": 11739 }, { "epoch": 1.18, "learning_rate": 4.7796950036969986e-05, "loss": 0.6464, "step": 11740 }, { "epoch": 1.18, "learning_rate": 4.779502340325367e-05, "loss": 0.8826, "step": 11741 }, { "epoch": 1.18, "learning_rate": 4.7793096656296615e-05, "loss": 0.6005, "step": 11742 }, { "epoch": 1.18, "learning_rate": 4.779116979611105e-05, "loss": 0.4402, "step": 11743 }, { "epoch": 1.18, "learning_rate": 4.778924282270925e-05, "loss": 0.3501, "step": 11744 }, { "epoch": 1.18, "learning_rate": 4.7787315736103476e-05, "loss": 1.1429, "step": 11745 }, { "epoch": 1.18, "learning_rate": 4.778538853630599e-05, "loss": 0.2226, "step": 11746 }, { "epoch": 1.18, "learning_rate": 4.778346122332906e-05, "loss": 0.1962, "step": 11747 }, { "epoch": 1.18, "learning_rate": 4.778153379718494e-05, "loss": 0.5758, "step": 11748 }, { "epoch": 1.18, "learning_rate": 4.777960625788591e-05, "loss": 0.7929, "step": 11749 }, { "epoch": 1.18, "learning_rate": 4.777767860544423e-05, "loss": 0.365, "step": 11750 }, { "epoch": 1.18, "learning_rate": 4.7775750839872165e-05, "loss": 0.9505, "step": 11751 }, { "epoch": 1.18, "learning_rate": 4.777382296118198e-05, "loss": 0.2902, "step": 11752 }, { "epoch": 1.18, "learning_rate": 4.7771894969385955e-05, "loss": 0.9338, "step": 11753 }, { "epoch": 1.18, "learning_rate": 4.776996686449635e-05, "loss": 0.7496, "step": 11754 }, { "epoch": 1.18, "learning_rate": 4.7768038646525434e-05, "loss": 1.2155, "step": 11755 }, { "epoch": 1.18, "learning_rate": 4.7766110315485475e-05, "loss": 0.9979, "step": 11756 }, { "epoch": 1.18, "learning_rate": 4.776418187138875e-05, "loss": 0.746, "step": 11757 }, { "epoch": 1.18, "learning_rate": 4.776225331424754e-05, "loss": 0.5216, "step": 11758 }, { "epoch": 1.18, "learning_rate": 4.776032464407409e-05, "loss": 0.796, "step": 11759 }, { "epoch": 1.18, "learning_rate": 4.775839586088071e-05, "loss": 0.4705, "step": 11760 }, { "epoch": 1.18, "learning_rate": 4.775646696467964e-05, "loss": 0.4531, "step": 11761 }, { "epoch": 1.18, "learning_rate": 4.775453795548317e-05, "loss": 1.162, "step": 11762 }, { "epoch": 1.18, "learning_rate": 4.7752608833303576e-05, "loss": 0.4412, "step": 11763 }, { "epoch": 1.18, "learning_rate": 4.775067959815314e-05, "loss": 0.428, "step": 11764 }, { "epoch": 1.18, "learning_rate": 4.774875025004412e-05, "loss": 0.3777, "step": 11765 }, { "epoch": 1.18, "learning_rate": 4.774682078898881e-05, "loss": 1.0163, "step": 11766 }, { "epoch": 1.18, "learning_rate": 4.7744891214999486e-05, "loss": 0.7698, "step": 11767 }, { "epoch": 1.18, "learning_rate": 4.7742961528088424e-05, "loss": 0.3915, "step": 11768 }, { "epoch": 1.18, "learning_rate": 4.7741031728267904e-05, "loss": 1.4668, "step": 11769 }, { "epoch": 1.18, "learning_rate": 4.773910181555022e-05, "loss": 0.4687, "step": 11770 }, { "epoch": 1.18, "learning_rate": 4.773717178994761e-05, "loss": 1.1915, "step": 11771 }, { "epoch": 1.18, "learning_rate": 4.773524165147241e-05, "loss": 0.4996, "step": 11772 }, { "epoch": 1.18, "learning_rate": 4.7733311400136876e-05, "loss": 1.0127, "step": 11773 }, { "epoch": 1.18, "learning_rate": 4.773138103595329e-05, "loss": 0.9366, "step": 11774 }, { "epoch": 1.18, "learning_rate": 4.772945055893394e-05, "loss": 0.2046, "step": 11775 }, { "epoch": 1.18, "learning_rate": 4.772751996909112e-05, "loss": 1.133, "step": 11776 }, { "epoch": 1.18, "learning_rate": 4.77255892664371e-05, "loss": 0.7033, "step": 11777 }, { "epoch": 1.18, "learning_rate": 4.772365845098416e-05, "loss": 0.2139, "step": 11778 }, { "epoch": 1.18, "learning_rate": 4.772172752274462e-05, "loss": 0.7888, "step": 11779 }, { "epoch": 1.18, "learning_rate": 4.771979648173074e-05, "loss": 0.5267, "step": 11780 }, { "epoch": 1.18, "learning_rate": 4.7717865327954816e-05, "loss": 0.5219, "step": 11781 }, { "epoch": 1.18, "learning_rate": 4.771593406142914e-05, "loss": 0.1198, "step": 11782 }, { "epoch": 1.18, "learning_rate": 4.7714002682166e-05, "loss": 0.3731, "step": 11783 }, { "epoch": 1.18, "learning_rate": 4.771207119017768e-05, "loss": 0.7528, "step": 11784 }, { "epoch": 1.18, "learning_rate": 4.771013958547648e-05, "loss": 0.4207, "step": 11785 }, { "epoch": 1.18, "learning_rate": 4.7708207868074686e-05, "loss": 0.5125, "step": 11786 }, { "epoch": 1.18, "learning_rate": 4.77062760379846e-05, "loss": 0.7533, "step": 11787 }, { "epoch": 1.18, "learning_rate": 4.77043440952185e-05, "loss": 0.4975, "step": 11788 }, { "epoch": 1.18, "learning_rate": 4.77024120397887e-05, "loss": 0.2931, "step": 11789 }, { "epoch": 1.18, "learning_rate": 4.770047987170748e-05, "loss": 0.6899, "step": 11790 }, { "epoch": 1.18, "learning_rate": 4.769854759098714e-05, "loss": 0.2124, "step": 11791 }, { "epoch": 1.18, "learning_rate": 4.7696615197639974e-05, "loss": 0.2966, "step": 11792 }, { "epoch": 1.18, "learning_rate": 4.769468269167829e-05, "loss": 0.3713, "step": 11793 }, { "epoch": 1.18, "learning_rate": 4.769275007311436e-05, "loss": 0.8389, "step": 11794 }, { "epoch": 1.18, "learning_rate": 4.769081734196051e-05, "loss": 0.5744, "step": 11795 }, { "epoch": 1.18, "learning_rate": 4.768888449822903e-05, "loss": 0.6496, "step": 11796 }, { "epoch": 1.18, "learning_rate": 4.768695154193221e-05, "loss": 0.4189, "step": 11797 }, { "epoch": 1.19, "learning_rate": 4.768501847308236e-05, "loss": 0.5492, "step": 11798 }, { "epoch": 1.19, "learning_rate": 4.7683085291691784e-05, "loss": 0.3261, "step": 11799 }, { "epoch": 1.19, "learning_rate": 4.768115199777279e-05, "loss": 0.3306, "step": 11800 }, { "epoch": 1.19, "learning_rate": 4.767921859133765e-05, "loss": 0.7277, "step": 11801 }, { "epoch": 1.19, "learning_rate": 4.76772850723987e-05, "loss": 1.0858, "step": 11802 }, { "epoch": 1.19, "learning_rate": 4.767535144096823e-05, "loss": 0.4618, "step": 11803 }, { "epoch": 1.19, "learning_rate": 4.767341769705855e-05, "loss": 0.261, "step": 11804 }, { "epoch": 1.19, "learning_rate": 4.767148384068196e-05, "loss": 0.4035, "step": 11805 }, { "epoch": 1.19, "learning_rate": 4.766954987185078e-05, "loss": 0.1983, "step": 11806 }, { "epoch": 1.19, "learning_rate": 4.76676157905773e-05, "loss": 0.471, "step": 11807 }, { "epoch": 1.19, "learning_rate": 4.766568159687383e-05, "loss": 0.7781, "step": 11808 }, { "epoch": 1.19, "learning_rate": 4.766374729075268e-05, "loss": 0.8935, "step": 11809 }, { "epoch": 1.19, "learning_rate": 4.7661812872226165e-05, "loss": 0.8686, "step": 11810 }, { "epoch": 1.19, "learning_rate": 4.7659878341306594e-05, "loss": 0.69, "step": 11811 }, { "epoch": 1.19, "learning_rate": 4.765794369800628e-05, "loss": 0.2542, "step": 11812 }, { "epoch": 1.19, "learning_rate": 4.7656008942337524e-05, "loss": 1.2952, "step": 11813 }, { "epoch": 1.19, "learning_rate": 4.7654074074312644e-05, "loss": 0.218, "step": 11814 }, { "epoch": 1.19, "learning_rate": 4.7652139093943954e-05, "loss": 0.1704, "step": 11815 }, { "epoch": 1.19, "learning_rate": 4.765020400124377e-05, "loss": 0.1564, "step": 11816 }, { "epoch": 1.19, "learning_rate": 4.764826879622439e-05, "loss": 0.1653, "step": 11817 }, { "epoch": 1.19, "learning_rate": 4.7646333478898146e-05, "loss": 0.2658, "step": 11818 }, { "epoch": 1.19, "learning_rate": 4.764439804927736e-05, "loss": 0.8307, "step": 11819 }, { "epoch": 1.19, "learning_rate": 4.764246250737432e-05, "loss": 0.6541, "step": 11820 }, { "epoch": 1.19, "learning_rate": 4.764052685320136e-05, "loss": 0.6523, "step": 11821 }, { "epoch": 1.19, "learning_rate": 4.763859108677081e-05, "loss": 1.1671, "step": 11822 }, { "epoch": 1.19, "learning_rate": 4.7636655208094976e-05, "loss": 0.2386, "step": 11823 }, { "epoch": 1.19, "learning_rate": 4.763471921718617e-05, "loss": 0.1432, "step": 11824 }, { "epoch": 1.19, "learning_rate": 4.763278311405673e-05, "loss": 0.5254, "step": 11825 }, { "epoch": 1.19, "learning_rate": 4.7630846898718956e-05, "loss": 0.1271, "step": 11826 }, { "epoch": 1.19, "learning_rate": 4.7628910571185184e-05, "loss": 0.7143, "step": 11827 }, { "epoch": 1.19, "learning_rate": 4.762697413146773e-05, "loss": 0.4677, "step": 11828 }, { "epoch": 1.19, "learning_rate": 4.762503757957892e-05, "loss": 0.6439, "step": 11829 }, { "epoch": 1.19, "learning_rate": 4.762310091553108e-05, "loss": 1.6026, "step": 11830 }, { "epoch": 1.19, "learning_rate": 4.762116413933652e-05, "loss": 1.6383, "step": 11831 }, { "epoch": 1.19, "learning_rate": 4.761922725100758e-05, "loss": 0.5893, "step": 11832 }, { "epoch": 1.19, "learning_rate": 4.761729025055658e-05, "loss": 0.6008, "step": 11833 }, { "epoch": 1.19, "learning_rate": 4.7615353137995846e-05, "loss": 0.2211, "step": 11834 }, { "epoch": 1.19, "learning_rate": 4.761341591333771e-05, "loss": 0.6234, "step": 11835 }, { "epoch": 1.19, "learning_rate": 4.761147857659449e-05, "loss": 0.4609, "step": 11836 }, { "epoch": 1.19, "learning_rate": 4.760954112777852e-05, "loss": 1.3748, "step": 11837 }, { "epoch": 1.19, "learning_rate": 4.760760356690214e-05, "loss": 0.438, "step": 11838 }, { "epoch": 1.19, "learning_rate": 4.760566589397765e-05, "loss": 1.1821, "step": 11839 }, { "epoch": 1.19, "learning_rate": 4.760372810901741e-05, "loss": 0.9313, "step": 11840 }, { "epoch": 1.19, "learning_rate": 4.760179021203374e-05, "loss": 0.2194, "step": 11841 }, { "epoch": 1.19, "learning_rate": 4.759985220303899e-05, "loss": 0.6594, "step": 11842 }, { "epoch": 1.19, "learning_rate": 4.759791408204546e-05, "loss": 0.6187, "step": 11843 }, { "epoch": 1.19, "learning_rate": 4.759597584906549e-05, "loss": 0.1821, "step": 11844 }, { "epoch": 1.19, "learning_rate": 4.759403750411144e-05, "loss": 0.9637, "step": 11845 }, { "epoch": 1.19, "learning_rate": 4.7592099047195624e-05, "loss": 0.6927, "step": 11846 }, { "epoch": 1.19, "learning_rate": 4.7590160478330376e-05, "loss": 0.6862, "step": 11847 }, { "epoch": 1.19, "learning_rate": 4.758822179752805e-05, "loss": 0.9733, "step": 11848 }, { "epoch": 1.19, "learning_rate": 4.758628300480096e-05, "loss": 0.7021, "step": 11849 }, { "epoch": 1.19, "learning_rate": 4.7584344100161454e-05, "loss": 0.6683, "step": 11850 }, { "epoch": 1.19, "learning_rate": 4.7582405083621885e-05, "loss": 0.6184, "step": 11851 }, { "epoch": 1.19, "learning_rate": 4.7580465955194556e-05, "loss": 0.395, "step": 11852 }, { "epoch": 1.19, "learning_rate": 4.7578526714891846e-05, "loss": 0.441, "step": 11853 }, { "epoch": 1.19, "learning_rate": 4.7576587362726083e-05, "loss": 0.2018, "step": 11854 }, { "epoch": 1.19, "learning_rate": 4.75746478987096e-05, "loss": 0.2375, "step": 11855 }, { "epoch": 1.19, "learning_rate": 4.757270832285474e-05, "loss": 0.5087, "step": 11856 }, { "epoch": 1.19, "learning_rate": 4.7570768635173854e-05, "loss": 0.6693, "step": 11857 }, { "epoch": 1.19, "learning_rate": 4.756882883567927e-05, "loss": 0.5604, "step": 11858 }, { "epoch": 1.19, "learning_rate": 4.756688892438335e-05, "loss": 0.4694, "step": 11859 }, { "epoch": 1.19, "learning_rate": 4.756494890129843e-05, "loss": 0.5676, "step": 11860 }, { "epoch": 1.19, "learning_rate": 4.756300876643687e-05, "loss": 0.8687, "step": 11861 }, { "epoch": 1.19, "learning_rate": 4.756106851981099e-05, "loss": 0.4755, "step": 11862 }, { "epoch": 1.19, "learning_rate": 4.755912816143315e-05, "loss": 0.4862, "step": 11863 }, { "epoch": 1.19, "learning_rate": 4.755718769131571e-05, "loss": 0.2733, "step": 11864 }, { "epoch": 1.19, "learning_rate": 4.7555247109470996e-05, "loss": 0.9601, "step": 11865 }, { "epoch": 1.19, "learning_rate": 4.755330641591137e-05, "loss": 0.9755, "step": 11866 }, { "epoch": 1.19, "learning_rate": 4.755136561064918e-05, "loss": 1.1043, "step": 11867 }, { "epoch": 1.19, "learning_rate": 4.7549424693696777e-05, "loss": 0.362, "step": 11868 }, { "epoch": 1.19, "learning_rate": 4.754748366506651e-05, "loss": 1.1441, "step": 11869 }, { "epoch": 1.19, "learning_rate": 4.754554252477073e-05, "loss": 0.6362, "step": 11870 }, { "epoch": 1.19, "learning_rate": 4.7543601272821804e-05, "loss": 1.4412, "step": 11871 }, { "epoch": 1.19, "learning_rate": 4.754165990923206e-05, "loss": 0.453, "step": 11872 }, { "epoch": 1.19, "learning_rate": 4.7539718434013866e-05, "loss": 0.3996, "step": 11873 }, { "epoch": 1.19, "learning_rate": 4.7537776847179584e-05, "loss": 0.5132, "step": 11874 }, { "epoch": 1.19, "learning_rate": 4.7535835148741556e-05, "loss": 1.1766, "step": 11875 }, { "epoch": 1.19, "learning_rate": 4.753389333871215e-05, "loss": 0.7362, "step": 11876 }, { "epoch": 1.19, "learning_rate": 4.753195141710371e-05, "loss": 1.3042, "step": 11877 }, { "epoch": 1.19, "learning_rate": 4.753000938392861e-05, "loss": 0.7447, "step": 11878 }, { "epoch": 1.19, "learning_rate": 4.752806723919918e-05, "loss": 1.5505, "step": 11879 }, { "epoch": 1.19, "learning_rate": 4.752612498292782e-05, "loss": 1.0147, "step": 11880 }, { "epoch": 1.19, "learning_rate": 4.7524182615126865e-05, "loss": 0.2915, "step": 11881 }, { "epoch": 1.19, "learning_rate": 4.752224013580867e-05, "loss": 0.4835, "step": 11882 }, { "epoch": 1.19, "learning_rate": 4.75202975449856e-05, "loss": 1.1835, "step": 11883 }, { "epoch": 1.19, "learning_rate": 4.751835484267003e-05, "loss": 0.7981, "step": 11884 }, { "epoch": 1.19, "learning_rate": 4.751641202887431e-05, "loss": 0.7314, "step": 11885 }, { "epoch": 1.19, "learning_rate": 4.7514469103610814e-05, "loss": 0.8907, "step": 11886 }, { "epoch": 1.19, "learning_rate": 4.751252606689189e-05, "loss": 0.4319, "step": 11887 }, { "epoch": 1.19, "learning_rate": 4.7510582918729925e-05, "loss": 0.6218, "step": 11888 }, { "epoch": 1.19, "learning_rate": 4.750863965913726e-05, "loss": 0.9927, "step": 11889 }, { "epoch": 1.19, "learning_rate": 4.750669628812628e-05, "loss": 0.2422, "step": 11890 }, { "epoch": 1.19, "learning_rate": 4.7504752805709335e-05, "loss": 0.9079, "step": 11891 }, { "epoch": 1.19, "learning_rate": 4.750280921189882e-05, "loss": 1.0735, "step": 11892 }, { "epoch": 1.19, "learning_rate": 4.750086550670707e-05, "loss": 0.3776, "step": 11893 }, { "epoch": 1.19, "learning_rate": 4.749892169014648e-05, "loss": 0.7556, "step": 11894 }, { "epoch": 1.19, "learning_rate": 4.7496977762229404e-05, "loss": 0.9383, "step": 11895 }, { "epoch": 1.19, "learning_rate": 4.74950337229682e-05, "loss": 1.0653, "step": 11896 }, { "epoch": 1.2, "learning_rate": 4.749308957237529e-05, "loss": 0.3631, "step": 11897 }, { "epoch": 1.2, "learning_rate": 4.7491145310462986e-05, "loss": 0.7217, "step": 11898 }, { "epoch": 1.2, "learning_rate": 4.74892009372437e-05, "loss": 0.546, "step": 11899 }, { "epoch": 1.2, "learning_rate": 4.7487256452729794e-05, "loss": 0.7112, "step": 11900 }, { "epoch": 1.2, "learning_rate": 4.748531185693364e-05, "loss": 0.5597, "step": 11901 }, { "epoch": 1.2, "learning_rate": 4.74833671498676e-05, "loss": 1.0084, "step": 11902 }, { "epoch": 1.2, "learning_rate": 4.748142233154409e-05, "loss": 0.8999, "step": 11903 }, { "epoch": 1.2, "learning_rate": 4.747947740197543e-05, "loss": 0.5019, "step": 11904 }, { "epoch": 1.2, "learning_rate": 4.747753236117404e-05, "loss": 1.1055, "step": 11905 }, { "epoch": 1.2, "learning_rate": 4.747558720915228e-05, "loss": 0.6516, "step": 11906 }, { "epoch": 1.2, "learning_rate": 4.747364194592253e-05, "loss": 0.1788, "step": 11907 }, { "epoch": 1.2, "learning_rate": 4.7471696571497165e-05, "loss": 0.4729, "step": 11908 }, { "epoch": 1.2, "learning_rate": 4.7469751085888575e-05, "loss": 0.3093, "step": 11909 }, { "epoch": 1.2, "learning_rate": 4.746780548910913e-05, "loss": 0.3963, "step": 11910 }, { "epoch": 1.2, "learning_rate": 4.746585978117122e-05, "loss": 0.2717, "step": 11911 }, { "epoch": 1.2, "learning_rate": 4.746391396208723e-05, "loss": 0.7875, "step": 11912 }, { "epoch": 1.2, "learning_rate": 4.746196803186953e-05, "loss": 0.442, "step": 11913 }, { "epoch": 1.2, "learning_rate": 4.74600219905305e-05, "loss": 1.0208, "step": 11914 }, { "epoch": 1.2, "learning_rate": 4.7458075838082544e-05, "loss": 0.2429, "step": 11915 }, { "epoch": 1.2, "learning_rate": 4.745612957453803e-05, "loss": 0.6685, "step": 11916 }, { "epoch": 1.2, "learning_rate": 4.745418319990935e-05, "loss": 0.4947, "step": 11917 }, { "epoch": 1.2, "learning_rate": 4.745223671420888e-05, "loss": 0.371, "step": 11918 }, { "epoch": 1.2, "learning_rate": 4.745029011744902e-05, "loss": 0.3448, "step": 11919 }, { "epoch": 1.2, "learning_rate": 4.7448343409642164e-05, "loss": 0.1818, "step": 11920 }, { "epoch": 1.2, "learning_rate": 4.7446396590800665e-05, "loss": 0.4652, "step": 11921 }, { "epoch": 1.2, "learning_rate": 4.744444966093695e-05, "loss": 0.6995, "step": 11922 }, { "epoch": 1.2, "learning_rate": 4.744250262006339e-05, "loss": 0.2956, "step": 11923 }, { "epoch": 1.2, "learning_rate": 4.7440555468192384e-05, "loss": 1.7681, "step": 11924 }, { "epoch": 1.2, "learning_rate": 4.743860820533631e-05, "loss": 0.555, "step": 11925 }, { "epoch": 1.2, "learning_rate": 4.7436660831507575e-05, "loss": 1.0342, "step": 11926 }, { "epoch": 1.2, "learning_rate": 4.7434713346718555e-05, "loss": 0.455, "step": 11927 }, { "epoch": 1.2, "learning_rate": 4.743276575098165e-05, "loss": 1.732, "step": 11928 }, { "epoch": 1.2, "learning_rate": 4.7430818044309265e-05, "loss": 0.4155, "step": 11929 }, { "epoch": 1.2, "learning_rate": 4.742887022671378e-05, "loss": 0.198, "step": 11930 }, { "epoch": 1.2, "learning_rate": 4.74269222982076e-05, "loss": 0.5924, "step": 11931 }, { "epoch": 1.2, "learning_rate": 4.7424974258803114e-05, "loss": 1.0097, "step": 11932 }, { "epoch": 1.2, "learning_rate": 4.742302610851272e-05, "loss": 1.1538, "step": 11933 }, { "epoch": 1.2, "learning_rate": 4.742107784734881e-05, "loss": 0.21, "step": 11934 }, { "epoch": 1.2, "learning_rate": 4.741912947532379e-05, "loss": 1.0963, "step": 11935 }, { "epoch": 1.2, "learning_rate": 4.741718099245007e-05, "loss": 0.6799, "step": 11936 }, { "epoch": 1.2, "learning_rate": 4.741523239874002e-05, "loss": 0.358, "step": 11937 }, { "epoch": 1.2, "learning_rate": 4.741328369420606e-05, "loss": 0.5212, "step": 11938 }, { "epoch": 1.2, "learning_rate": 4.7411334878860594e-05, "loss": 0.6827, "step": 11939 }, { "epoch": 1.2, "learning_rate": 4.740938595271601e-05, "loss": 1.3056, "step": 11940 }, { "epoch": 1.2, "learning_rate": 4.740743691578472e-05, "loss": 0.2461, "step": 11941 }, { "epoch": 1.2, "learning_rate": 4.7405487768079126e-05, "loss": 1.3038, "step": 11942 }, { "epoch": 1.2, "learning_rate": 4.740353850961162e-05, "loss": 0.2453, "step": 11943 }, { "epoch": 1.2, "learning_rate": 4.7401589140394613e-05, "loss": 0.6509, "step": 11944 }, { "epoch": 1.2, "learning_rate": 4.7399639660440526e-05, "loss": 0.9646, "step": 11945 }, { "epoch": 1.2, "learning_rate": 4.739769006976174e-05, "loss": 0.2129, "step": 11946 }, { "epoch": 1.2, "learning_rate": 4.739574036837068e-05, "loss": 0.7444, "step": 11947 }, { "epoch": 1.2, "learning_rate": 4.7393790556279746e-05, "loss": 0.4136, "step": 11948 }, { "epoch": 1.2, "learning_rate": 4.739184063350134e-05, "loss": 1.5101, "step": 11949 }, { "epoch": 1.2, "learning_rate": 4.738989060004788e-05, "loss": 0.274, "step": 11950 }, { "epoch": 1.2, "learning_rate": 4.738794045593178e-05, "loss": 1.0671, "step": 11951 }, { "epoch": 1.2, "learning_rate": 4.7385990201165435e-05, "loss": 0.7671, "step": 11952 }, { "epoch": 1.2, "learning_rate": 4.7384039835761255e-05, "loss": 0.8152, "step": 11953 }, { "epoch": 1.2, "learning_rate": 4.738208935973166e-05, "loss": 0.9067, "step": 11954 }, { "epoch": 1.2, "learning_rate": 4.738013877308907e-05, "loss": 0.7762, "step": 11955 }, { "epoch": 1.2, "learning_rate": 4.7378188075845885e-05, "loss": 0.8664, "step": 11956 }, { "epoch": 1.2, "learning_rate": 4.737623726801452e-05, "loss": 0.7913, "step": 11957 }, { "epoch": 1.2, "learning_rate": 4.7374286349607394e-05, "loss": 0.8153, "step": 11958 }, { "epoch": 1.2, "learning_rate": 4.737233532063692e-05, "loss": 0.8597, "step": 11959 }, { "epoch": 1.2, "learning_rate": 4.7370384181115505e-05, "loss": 0.3586, "step": 11960 }, { "epoch": 1.2, "learning_rate": 4.7368432931055586e-05, "loss": 0.3536, "step": 11961 }, { "epoch": 1.2, "learning_rate": 4.736648157046956e-05, "loss": 0.4472, "step": 11962 }, { "epoch": 1.2, "learning_rate": 4.7364530099369845e-05, "loss": 0.7796, "step": 11963 }, { "epoch": 1.2, "learning_rate": 4.736257851776887e-05, "loss": 0.1038, "step": 11964 }, { "epoch": 1.2, "learning_rate": 4.7360626825679057e-05, "loss": 0.1695, "step": 11965 }, { "epoch": 1.2, "learning_rate": 4.7358675023112815e-05, "loss": 0.3402, "step": 11966 }, { "epoch": 1.2, "learning_rate": 4.7356723110082567e-05, "loss": 0.3259, "step": 11967 }, { "epoch": 1.2, "learning_rate": 4.735477108660074e-05, "loss": 0.3506, "step": 11968 }, { "epoch": 1.2, "learning_rate": 4.735281895267975e-05, "loss": 0.2768, "step": 11969 }, { "epoch": 1.2, "learning_rate": 4.735086670833202e-05, "loss": 1.1854, "step": 11970 }, { "epoch": 1.2, "learning_rate": 4.734891435356997e-05, "loss": 0.3448, "step": 11971 }, { "epoch": 1.2, "learning_rate": 4.734696188840604e-05, "loss": 0.9301, "step": 11972 }, { "epoch": 1.2, "learning_rate": 4.734500931285264e-05, "loss": 0.5843, "step": 11973 }, { "epoch": 1.2, "learning_rate": 4.7343056626922204e-05, "loss": 1.1984, "step": 11974 }, { "epoch": 1.2, "learning_rate": 4.734110383062714e-05, "loss": 0.5227, "step": 11975 }, { "epoch": 1.2, "learning_rate": 4.733915092397989e-05, "loss": 0.2467, "step": 11976 }, { "epoch": 1.2, "learning_rate": 4.733719790699289e-05, "loss": 0.7747, "step": 11977 }, { "epoch": 1.2, "learning_rate": 4.7335244779678555e-05, "loss": 0.5782, "step": 11978 }, { "epoch": 1.2, "learning_rate": 4.7333291542049314e-05, "loss": 0.7882, "step": 11979 }, { "epoch": 1.2, "learning_rate": 4.733133819411759e-05, "loss": 0.6753, "step": 11980 }, { "epoch": 1.2, "learning_rate": 4.732938473589584e-05, "loss": 0.7433, "step": 11981 }, { "epoch": 1.2, "learning_rate": 4.7327431167396474e-05, "loss": 0.3858, "step": 11982 }, { "epoch": 1.2, "learning_rate": 4.732547748863192e-05, "loss": 0.5483, "step": 11983 }, { "epoch": 1.2, "learning_rate": 4.7323523699614626e-05, "loss": 0.2582, "step": 11984 }, { "epoch": 1.2, "learning_rate": 4.732156980035701e-05, "loss": 1.0386, "step": 11985 }, { "epoch": 1.2, "learning_rate": 4.7319615790871527e-05, "loss": 0.3339, "step": 11986 }, { "epoch": 1.2, "learning_rate": 4.731766167117059e-05, "loss": 0.9854, "step": 11987 }, { "epoch": 1.2, "learning_rate": 4.731570744126664e-05, "loss": 0.5399, "step": 11988 }, { "epoch": 1.2, "learning_rate": 4.731375310117211e-05, "loss": 0.4489, "step": 11989 }, { "epoch": 1.2, "learning_rate": 4.7311798650899444e-05, "loss": 0.319, "step": 11990 }, { "epoch": 1.2, "learning_rate": 4.7309844090461085e-05, "loss": 1.037, "step": 11991 }, { "epoch": 1.2, "learning_rate": 4.730788941986946e-05, "loss": 0.4454, "step": 11992 }, { "epoch": 1.2, "learning_rate": 4.7305934639137015e-05, "loss": 0.9118, "step": 11993 }, { "epoch": 1.2, "learning_rate": 4.730397974827618e-05, "loss": 1.4701, "step": 11994 }, { "epoch": 1.21, "learning_rate": 4.73020247472994e-05, "loss": 0.6216, "step": 11995 }, { "epoch": 1.21, "learning_rate": 4.730006963621911e-05, "loss": 0.6093, "step": 11996 }, { "epoch": 1.21, "learning_rate": 4.729811441504777e-05, "loss": 0.5334, "step": 11997 }, { "epoch": 1.21, "learning_rate": 4.72961590837978e-05, "loss": 0.9585, "step": 11998 }, { "epoch": 1.21, "learning_rate": 4.729420364248167e-05, "loss": 0.9908, "step": 11999 }, { "epoch": 1.21, "learning_rate": 4.729224809111179e-05, "loss": 1.1635, "step": 12000 }, { "epoch": 1.21, "learning_rate": 4.729029242970063e-05, "loss": 0.9516, "step": 12001 }, { "epoch": 1.21, "learning_rate": 4.728833665826062e-05, "loss": 1.0278, "step": 12002 }, { "epoch": 1.21, "learning_rate": 4.7286380776804226e-05, "loss": 0.7207, "step": 12003 }, { "epoch": 1.21, "learning_rate": 4.728442478534386e-05, "loss": 0.2695, "step": 12004 }, { "epoch": 1.21, "learning_rate": 4.728246868389201e-05, "loss": 0.3661, "step": 12005 }, { "epoch": 1.21, "learning_rate": 4.728051247246109e-05, "loss": 1.0071, "step": 12006 }, { "epoch": 1.21, "learning_rate": 4.7278556151063574e-05, "loss": 0.8107, "step": 12007 }, { "epoch": 1.21, "learning_rate": 4.727659971971189e-05, "loss": 0.7257, "step": 12008 }, { "epoch": 1.21, "learning_rate": 4.72746431784185e-05, "loss": 0.4656, "step": 12009 }, { "epoch": 1.21, "learning_rate": 4.7272686527195855e-05, "loss": 0.5484, "step": 12010 }, { "epoch": 1.21, "learning_rate": 4.72707297660564e-05, "loss": 0.5166, "step": 12011 }, { "epoch": 1.21, "learning_rate": 4.7268772895012594e-05, "loss": 0.2264, "step": 12012 }, { "epoch": 1.21, "learning_rate": 4.726681591407689e-05, "loss": 0.3403, "step": 12013 }, { "epoch": 1.21, "learning_rate": 4.726485882326174e-05, "loss": 0.6134, "step": 12014 }, { "epoch": 1.21, "learning_rate": 4.726290162257959e-05, "loss": 0.3508, "step": 12015 }, { "epoch": 1.21, "learning_rate": 4.726094431204289e-05, "loss": 2.2019, "step": 12016 }, { "epoch": 1.21, "learning_rate": 4.725898689166413e-05, "loss": 0.6198, "step": 12017 }, { "epoch": 1.21, "learning_rate": 4.725702936145572e-05, "loss": 0.7024, "step": 12018 }, { "epoch": 1.21, "learning_rate": 4.725507172143015e-05, "loss": 1.3843, "step": 12019 }, { "epoch": 1.21, "learning_rate": 4.7253113971599875e-05, "loss": 0.2434, "step": 12020 }, { "epoch": 1.21, "learning_rate": 4.725115611197735e-05, "loss": 0.3373, "step": 12021 }, { "epoch": 1.21, "learning_rate": 4.7249198142575014e-05, "loss": 0.5308, "step": 12022 }, { "epoch": 1.21, "learning_rate": 4.7247240063405356e-05, "loss": 0.2303, "step": 12023 }, { "epoch": 1.21, "learning_rate": 4.724528187448081e-05, "loss": 0.3501, "step": 12024 }, { "epoch": 1.21, "learning_rate": 4.724332357581386e-05, "loss": 0.4969, "step": 12025 }, { "epoch": 1.21, "learning_rate": 4.724136516741696e-05, "loss": 0.1517, "step": 12026 }, { "epoch": 1.21, "learning_rate": 4.723940664930257e-05, "loss": 1.0907, "step": 12027 }, { "epoch": 1.21, "learning_rate": 4.723744802148315e-05, "loss": 0.5523, "step": 12028 }, { "epoch": 1.21, "learning_rate": 4.7235489283971165e-05, "loss": 0.9181, "step": 12029 }, { "epoch": 1.21, "learning_rate": 4.723353043677909e-05, "loss": 0.2779, "step": 12030 }, { "epoch": 1.21, "learning_rate": 4.723157147991938e-05, "loss": 0.8566, "step": 12031 }, { "epoch": 1.21, "learning_rate": 4.7229612413404505e-05, "loss": 0.2273, "step": 12032 }, { "epoch": 1.21, "learning_rate": 4.7227653237246935e-05, "loss": 0.6528, "step": 12033 }, { "epoch": 1.21, "learning_rate": 4.7225693951459125e-05, "loss": 0.9535, "step": 12034 }, { "epoch": 1.21, "learning_rate": 4.722373455605356e-05, "loss": 0.237, "step": 12035 }, { "epoch": 1.21, "learning_rate": 4.7221775051042694e-05, "loss": 0.3931, "step": 12036 }, { "epoch": 1.21, "learning_rate": 4.721981543643901e-05, "loss": 0.6901, "step": 12037 }, { "epoch": 1.21, "learning_rate": 4.7217855712254966e-05, "loss": 0.4243, "step": 12038 }, { "epoch": 1.21, "learning_rate": 4.7215895878503036e-05, "loss": 0.2923, "step": 12039 }, { "epoch": 1.21, "learning_rate": 4.72139359351957e-05, "loss": 0.2675, "step": 12040 }, { "epoch": 1.21, "learning_rate": 4.721197588234542e-05, "loss": 0.9891, "step": 12041 }, { "epoch": 1.21, "learning_rate": 4.721001571996468e-05, "loss": 0.9767, "step": 12042 }, { "epoch": 1.21, "learning_rate": 4.720805544806594e-05, "loss": 0.2643, "step": 12043 }, { "epoch": 1.21, "learning_rate": 4.7206095066661684e-05, "loss": 0.6091, "step": 12044 }, { "epoch": 1.21, "learning_rate": 4.720413457576438e-05, "loss": 0.4747, "step": 12045 }, { "epoch": 1.21, "learning_rate": 4.720217397538652e-05, "loss": 0.2087, "step": 12046 }, { "epoch": 1.21, "learning_rate": 4.720021326554056e-05, "loss": 0.421, "step": 12047 }, { "epoch": 1.21, "learning_rate": 4.719825244623899e-05, "loss": 0.1706, "step": 12048 }, { "epoch": 1.21, "learning_rate": 4.7196291517494274e-05, "loss": 0.4614, "step": 12049 }, { "epoch": 1.21, "learning_rate": 4.719433047931891e-05, "loss": 1.9511, "step": 12050 }, { "epoch": 1.21, "learning_rate": 4.719236933172537e-05, "loss": 0.1828, "step": 12051 }, { "epoch": 1.21, "learning_rate": 4.719040807472613e-05, "loss": 0.3606, "step": 12052 }, { "epoch": 1.21, "learning_rate": 4.718844670833367e-05, "loss": 0.3872, "step": 12053 }, { "epoch": 1.21, "learning_rate": 4.718648523256046e-05, "loss": 0.4466, "step": 12054 }, { "epoch": 1.21, "learning_rate": 4.7184523647419027e-05, "loss": 0.9575, "step": 12055 }, { "epoch": 1.21, "learning_rate": 4.7182561952921794e-05, "loss": 0.1522, "step": 12056 }, { "epoch": 1.21, "learning_rate": 4.718060014908129e-05, "loss": 0.6596, "step": 12057 }, { "epoch": 1.21, "learning_rate": 4.717863823590998e-05, "loss": 0.3268, "step": 12058 }, { "epoch": 1.21, "learning_rate": 4.717667621342035e-05, "loss": 0.3356, "step": 12059 }, { "epoch": 1.21, "learning_rate": 4.7174714081624885e-05, "loss": 1.104, "step": 12060 }, { "epoch": 1.21, "learning_rate": 4.717275184053607e-05, "loss": 1.4319, "step": 12061 }, { "epoch": 1.21, "learning_rate": 4.71707894901664e-05, "loss": 0.5565, "step": 12062 }, { "epoch": 1.21, "learning_rate": 4.7168827030528364e-05, "loss": 0.1879, "step": 12063 }, { "epoch": 1.21, "learning_rate": 4.716686446163444e-05, "loss": 1.0142, "step": 12064 }, { "epoch": 1.21, "learning_rate": 4.716490178349712e-05, "loss": 0.2356, "step": 12065 }, { "epoch": 1.21, "learning_rate": 4.7162938996128885e-05, "loss": 0.735, "step": 12066 }, { "epoch": 1.21, "learning_rate": 4.716097609954225e-05, "loss": 0.4365, "step": 12067 }, { "epoch": 1.21, "learning_rate": 4.7159013093749684e-05, "loss": 0.4483, "step": 12068 }, { "epoch": 1.21, "learning_rate": 4.7157049978763684e-05, "loss": 0.7497, "step": 12069 }, { "epoch": 1.21, "learning_rate": 4.715508675459675e-05, "loss": 0.3514, "step": 12070 }, { "epoch": 1.21, "learning_rate": 4.715312342126137e-05, "loss": 0.6352, "step": 12071 }, { "epoch": 1.21, "learning_rate": 4.7151159978770035e-05, "loss": 0.5747, "step": 12072 }, { "epoch": 1.21, "learning_rate": 4.7149196427135244e-05, "loss": 0.8921, "step": 12073 }, { "epoch": 1.21, "learning_rate": 4.714723276636949e-05, "loss": 0.4589, "step": 12074 }, { "epoch": 1.21, "learning_rate": 4.714526899648527e-05, "loss": 0.6657, "step": 12075 }, { "epoch": 1.21, "learning_rate": 4.7143305117495084e-05, "loss": 0.1519, "step": 12076 }, { "epoch": 1.21, "learning_rate": 4.714134112941142e-05, "loss": 1.0795, "step": 12077 }, { "epoch": 1.21, "learning_rate": 4.713937703224679e-05, "loss": 0.4274, "step": 12078 }, { "epoch": 1.21, "learning_rate": 4.713741282601368e-05, "loss": 0.8757, "step": 12079 }, { "epoch": 1.21, "learning_rate": 4.713544851072459e-05, "loss": 0.7991, "step": 12080 }, { "epoch": 1.21, "learning_rate": 4.713348408639203e-05, "loss": 0.8272, "step": 12081 }, { "epoch": 1.21, "learning_rate": 4.7131519553028495e-05, "loss": 0.6772, "step": 12082 }, { "epoch": 1.21, "learning_rate": 4.712955491064649e-05, "loss": 0.2594, "step": 12083 }, { "epoch": 1.21, "learning_rate": 4.71275901592585e-05, "loss": 0.3294, "step": 12084 }, { "epoch": 1.21, "learning_rate": 4.712562529887705e-05, "loss": 0.6304, "step": 12085 }, { "epoch": 1.21, "learning_rate": 4.712366032951464e-05, "loss": 0.4404, "step": 12086 }, { "epoch": 1.21, "learning_rate": 4.712169525118376e-05, "loss": 0.7972, "step": 12087 }, { "epoch": 1.21, "learning_rate": 4.711973006389694e-05, "loss": 0.414, "step": 12088 }, { "epoch": 1.21, "learning_rate": 4.7117764767666664e-05, "loss": 0.1931, "step": 12089 }, { "epoch": 1.21, "learning_rate": 4.7115799362505433e-05, "loss": 0.4247, "step": 12090 }, { "epoch": 1.21, "learning_rate": 4.7113833848425774e-05, "loss": 0.3058, "step": 12091 }, { "epoch": 1.21, "learning_rate": 4.7111868225440195e-05, "loss": 1.357, "step": 12092 }, { "epoch": 1.21, "learning_rate": 4.710990249356119e-05, "loss": 1.2867, "step": 12093 }, { "epoch": 1.22, "learning_rate": 4.7107936652801274e-05, "loss": 0.6563, "step": 12094 }, { "epoch": 1.22, "learning_rate": 4.710597070317296e-05, "loss": 1.1219, "step": 12095 }, { "epoch": 1.22, "learning_rate": 4.710400464468875e-05, "loss": 0.6011, "step": 12096 }, { "epoch": 1.22, "learning_rate": 4.710203847736116e-05, "loss": 0.5874, "step": 12097 }, { "epoch": 1.22, "learning_rate": 4.710007220120271e-05, "loss": 0.7895, "step": 12098 }, { "epoch": 1.22, "learning_rate": 4.7098105816225914e-05, "loss": 0.4816, "step": 12099 }, { "epoch": 1.22, "learning_rate": 4.709613932244326e-05, "loss": 0.4062, "step": 12100 }, { "epoch": 1.22, "learning_rate": 4.7094172719867285e-05, "loss": 0.4097, "step": 12101 }, { "epoch": 1.22, "learning_rate": 4.70922060085105e-05, "loss": 0.4587, "step": 12102 }, { "epoch": 1.22, "learning_rate": 4.709023918838542e-05, "loss": 0.91, "step": 12103 }, { "epoch": 1.22, "learning_rate": 4.7088272259504554e-05, "loss": 0.1839, "step": 12104 }, { "epoch": 1.22, "learning_rate": 4.708630522188042e-05, "loss": 0.2911, "step": 12105 }, { "epoch": 1.22, "learning_rate": 4.708433807552555e-05, "loss": 0.3638, "step": 12106 }, { "epoch": 1.22, "learning_rate": 4.708237082045245e-05, "loss": 0.9462, "step": 12107 }, { "epoch": 1.22, "learning_rate": 4.708040345667364e-05, "loss": 0.4768, "step": 12108 }, { "epoch": 1.22, "learning_rate": 4.7078435984201634e-05, "loss": 0.2456, "step": 12109 }, { "epoch": 1.22, "learning_rate": 4.707646840304896e-05, "loss": 0.5533, "step": 12110 }, { "epoch": 1.22, "learning_rate": 4.707450071322814e-05, "loss": 0.2365, "step": 12111 }, { "epoch": 1.22, "learning_rate": 4.707253291475169e-05, "loss": 0.9219, "step": 12112 }, { "epoch": 1.22, "learning_rate": 4.707056500763213e-05, "loss": 0.6088, "step": 12113 }, { "epoch": 1.22, "learning_rate": 4.7068596991882e-05, "loss": 0.6898, "step": 12114 }, { "epoch": 1.22, "learning_rate": 4.706662886751381e-05, "loss": 0.8738, "step": 12115 }, { "epoch": 1.22, "learning_rate": 4.706466063454009e-05, "loss": 0.24, "step": 12116 }, { "epoch": 1.22, "learning_rate": 4.706269229297335e-05, "loss": 0.7103, "step": 12117 }, { "epoch": 1.22, "learning_rate": 4.706072384282613e-05, "loss": 0.3119, "step": 12118 }, { "epoch": 1.22, "learning_rate": 4.705875528411096e-05, "loss": 0.2124, "step": 12119 }, { "epoch": 1.22, "learning_rate": 4.705678661684036e-05, "loss": 1.5356, "step": 12120 }, { "epoch": 1.22, "learning_rate": 4.705481784102685e-05, "loss": 0.0892, "step": 12121 }, { "epoch": 1.22, "learning_rate": 4.7052848956682984e-05, "loss": 0.1967, "step": 12122 }, { "epoch": 1.22, "learning_rate": 4.705087996382126e-05, "loss": 1.1103, "step": 12123 }, { "epoch": 1.22, "learning_rate": 4.704891086245422e-05, "loss": 0.2169, "step": 12124 }, { "epoch": 1.22, "learning_rate": 4.704694165259441e-05, "loss": 0.2506, "step": 12125 }, { "epoch": 1.22, "learning_rate": 4.704497233425434e-05, "loss": 1.2778, "step": 12126 }, { "epoch": 1.22, "learning_rate": 4.704300290744655e-05, "loss": 0.7158, "step": 12127 }, { "epoch": 1.22, "learning_rate": 4.7041033372183576e-05, "loss": 0.4524, "step": 12128 }, { "epoch": 1.22, "learning_rate": 4.703906372847795e-05, "loss": 1.0536, "step": 12129 }, { "epoch": 1.22, "learning_rate": 4.70370939763422e-05, "loss": 0.4422, "step": 12130 }, { "epoch": 1.22, "learning_rate": 4.703512411578887e-05, "loss": 0.737, "step": 12131 }, { "epoch": 1.22, "learning_rate": 4.703315414683048e-05, "loss": 1.1883, "step": 12132 }, { "epoch": 1.22, "learning_rate": 4.7031184069479586e-05, "loss": 0.4904, "step": 12133 }, { "epoch": 1.22, "learning_rate": 4.702921388374872e-05, "loss": 0.5057, "step": 12134 }, { "epoch": 1.22, "learning_rate": 4.70272435896504e-05, "loss": 0.2286, "step": 12135 }, { "epoch": 1.22, "learning_rate": 4.702527318719719e-05, "loss": 0.4084, "step": 12136 }, { "epoch": 1.22, "learning_rate": 4.702330267640162e-05, "loss": 1.0506, "step": 12137 }, { "epoch": 1.22, "learning_rate": 4.702133205727622e-05, "loss": 0.3059, "step": 12138 }, { "epoch": 1.22, "learning_rate": 4.7019361329833545e-05, "loss": 0.2911, "step": 12139 }, { "epoch": 1.22, "learning_rate": 4.701739049408613e-05, "loss": 0.3077, "step": 12140 }, { "epoch": 1.22, "learning_rate": 4.70154195500465e-05, "loss": 0.9952, "step": 12141 }, { "epoch": 1.22, "learning_rate": 4.701344849772723e-05, "loss": 0.7528, "step": 12142 }, { "epoch": 1.22, "learning_rate": 4.701147733714084e-05, "loss": 0.2986, "step": 12143 }, { "epoch": 1.22, "learning_rate": 4.700950606829988e-05, "loss": 0.6845, "step": 12144 }, { "epoch": 1.22, "learning_rate": 4.700753469121689e-05, "loss": 0.5406, "step": 12145 }, { "epoch": 1.22, "learning_rate": 4.7005563205904434e-05, "loss": 0.6697, "step": 12146 }, { "epoch": 1.22, "learning_rate": 4.700359161237504e-05, "loss": 0.3476, "step": 12147 }, { "epoch": 1.22, "learning_rate": 4.700161991064124e-05, "loss": 0.9092, "step": 12148 }, { "epoch": 1.22, "learning_rate": 4.6999648100715613e-05, "loss": 0.7561, "step": 12149 }, { "epoch": 1.22, "learning_rate": 4.699767618261069e-05, "loss": 1.1506, "step": 12150 }, { "epoch": 1.22, "learning_rate": 4.699570415633901e-05, "loss": 0.4307, "step": 12151 }, { "epoch": 1.22, "learning_rate": 4.699373202191315e-05, "loss": 0.7426, "step": 12152 }, { "epoch": 1.22, "learning_rate": 4.699175977934564e-05, "loss": 0.1322, "step": 12153 }, { "epoch": 1.22, "learning_rate": 4.698978742864903e-05, "loss": 0.8014, "step": 12154 }, { "epoch": 1.22, "learning_rate": 4.698781496983589e-05, "loss": 0.7409, "step": 12155 }, { "epoch": 1.22, "learning_rate": 4.698584240291874e-05, "loss": 0.7661, "step": 12156 }, { "epoch": 1.22, "learning_rate": 4.698386972791016e-05, "loss": 0.6023, "step": 12157 }, { "epoch": 1.22, "learning_rate": 4.6981896944822685e-05, "loss": 0.4597, "step": 12158 }, { "epoch": 1.22, "learning_rate": 4.6979924053668895e-05, "loss": 0.6515, "step": 12159 }, { "epoch": 1.22, "learning_rate": 4.6977951054461314e-05, "loss": 0.8875, "step": 12160 }, { "epoch": 1.22, "learning_rate": 4.697597794721251e-05, "loss": 1.0537, "step": 12161 }, { "epoch": 1.22, "learning_rate": 4.697400473193505e-05, "loss": 0.1629, "step": 12162 }, { "epoch": 1.22, "learning_rate": 4.6972031408641486e-05, "loss": 1.0209, "step": 12163 }, { "epoch": 1.22, "learning_rate": 4.6970057977344353e-05, "loss": 0.7411, "step": 12164 }, { "epoch": 1.22, "learning_rate": 4.696808443805624e-05, "loss": 0.1732, "step": 12165 }, { "epoch": 1.22, "learning_rate": 4.696611079078969e-05, "loss": 0.3631, "step": 12166 }, { "epoch": 1.22, "learning_rate": 4.696413703555726e-05, "loss": 0.4992, "step": 12167 }, { "epoch": 1.22, "learning_rate": 4.696216317237151e-05, "loss": 1.1225, "step": 12168 }, { "epoch": 1.22, "learning_rate": 4.696018920124502e-05, "loss": 0.2318, "step": 12169 }, { "epoch": 1.22, "learning_rate": 4.695821512219034e-05, "loss": 0.9855, "step": 12170 }, { "epoch": 1.22, "learning_rate": 4.695624093522001e-05, "loss": 0.7639, "step": 12171 }, { "epoch": 1.22, "learning_rate": 4.695426664034664e-05, "loss": 0.6228, "step": 12172 }, { "epoch": 1.22, "learning_rate": 4.6952292237582746e-05, "loss": 0.8431, "step": 12173 }, { "epoch": 1.22, "learning_rate": 4.695031772694092e-05, "loss": 0.2533, "step": 12174 }, { "epoch": 1.22, "learning_rate": 4.6948343108433726e-05, "loss": 1.1751, "step": 12175 }, { "epoch": 1.22, "learning_rate": 4.6946368382073714e-05, "loss": 0.987, "step": 12176 }, { "epoch": 1.22, "learning_rate": 4.6944393547873465e-05, "loss": 0.2572, "step": 12177 }, { "epoch": 1.22, "learning_rate": 4.694241860584554e-05, "loss": 0.5842, "step": 12178 }, { "epoch": 1.22, "learning_rate": 4.6940443556002516e-05, "loss": 0.2973, "step": 12179 }, { "epoch": 1.22, "learning_rate": 4.693846839835694e-05, "loss": 0.7432, "step": 12180 }, { "epoch": 1.22, "learning_rate": 4.6936493132921405e-05, "loss": 0.2983, "step": 12181 }, { "epoch": 1.22, "learning_rate": 4.693451775970847e-05, "loss": 0.2568, "step": 12182 }, { "epoch": 1.22, "learning_rate": 4.693254227873069e-05, "loss": 0.5992, "step": 12183 }, { "epoch": 1.22, "learning_rate": 4.693056669000068e-05, "loss": 0.5305, "step": 12184 }, { "epoch": 1.22, "learning_rate": 4.692859099353097e-05, "loss": 1.3871, "step": 12185 }, { "epoch": 1.22, "learning_rate": 4.6926615189334144e-05, "loss": 0.8972, "step": 12186 }, { "epoch": 1.22, "learning_rate": 4.692463927742278e-05, "loss": 0.6342, "step": 12187 }, { "epoch": 1.22, "learning_rate": 4.692266325780945e-05, "loss": 1.2787, "step": 12188 }, { "epoch": 1.22, "learning_rate": 4.6920687130506734e-05, "loss": 1.0993, "step": 12189 }, { "epoch": 1.22, "learning_rate": 4.691871089552719e-05, "loss": 0.9466, "step": 12190 }, { "epoch": 1.22, "learning_rate": 4.691673455288342e-05, "loss": 0.7894, "step": 12191 }, { "epoch": 1.23, "learning_rate": 4.691475810258798e-05, "loss": 0.7049, "step": 12192 }, { "epoch": 1.23, "learning_rate": 4.691278154465346e-05, "loss": 1.0261, "step": 12193 }, { "epoch": 1.23, "learning_rate": 4.691080487909243e-05, "loss": 0.584, "step": 12194 }, { "epoch": 1.23, "learning_rate": 4.690882810591747e-05, "loss": 0.5619, "step": 12195 }, { "epoch": 1.23, "learning_rate": 4.690685122514115e-05, "loss": 0.4783, "step": 12196 }, { "epoch": 1.23, "learning_rate": 4.690487423677607e-05, "loss": 0.5664, "step": 12197 }, { "epoch": 1.23, "learning_rate": 4.6902897140834806e-05, "loss": 0.4016, "step": 12198 }, { "epoch": 1.23, "learning_rate": 4.6900919937329935e-05, "loss": 0.374, "step": 12199 }, { "epoch": 1.23, "learning_rate": 4.6898942626274024e-05, "loss": 0.4046, "step": 12200 }, { "epoch": 1.23, "learning_rate": 4.689696520767968e-05, "loss": 0.6892, "step": 12201 }, { "epoch": 1.23, "learning_rate": 4.689498768155948e-05, "loss": 0.4889, "step": 12202 }, { "epoch": 1.23, "learning_rate": 4.6893010047926e-05, "loss": 0.3867, "step": 12203 }, { "epoch": 1.23, "learning_rate": 4.689103230679184e-05, "loss": 0.5254, "step": 12204 }, { "epoch": 1.23, "learning_rate": 4.6889054458169566e-05, "loss": 0.4841, "step": 12205 }, { "epoch": 1.23, "learning_rate": 4.6887076502071774e-05, "loss": 0.3841, "step": 12206 }, { "epoch": 1.23, "learning_rate": 4.6885098438511055e-05, "loss": 0.3205, "step": 12207 }, { "epoch": 1.23, "learning_rate": 4.68831202675e-05, "loss": 0.2663, "step": 12208 }, { "epoch": 1.23, "learning_rate": 4.688114198905119e-05, "loss": 0.9797, "step": 12209 }, { "epoch": 1.23, "learning_rate": 4.687916360317719e-05, "loss": 1.1325, "step": 12210 }, { "epoch": 1.23, "learning_rate": 4.6877185109890644e-05, "loss": 1.1357, "step": 12211 }, { "epoch": 1.23, "learning_rate": 4.68752065092041e-05, "loss": 0.1505, "step": 12212 }, { "epoch": 1.23, "learning_rate": 4.687322780113016e-05, "loss": 0.2949, "step": 12213 }, { "epoch": 1.23, "learning_rate": 4.687124898568142e-05, "loss": 1.1292, "step": 12214 }, { "epoch": 1.23, "learning_rate": 4.686927006287047e-05, "loss": 0.9334, "step": 12215 }, { "epoch": 1.23, "learning_rate": 4.686729103270989e-05, "loss": 0.3892, "step": 12216 }, { "epoch": 1.23, "learning_rate": 4.686531189521231e-05, "loss": 0.5169, "step": 12217 }, { "epoch": 1.23, "learning_rate": 4.686333265039029e-05, "loss": 0.5531, "step": 12218 }, { "epoch": 1.23, "learning_rate": 4.6861353298256425e-05, "loss": 0.8587, "step": 12219 }, { "epoch": 1.23, "learning_rate": 4.6859373838823335e-05, "loss": 1.0352, "step": 12220 }, { "epoch": 1.23, "learning_rate": 4.68573942721036e-05, "loss": 0.7004, "step": 12221 }, { "epoch": 1.23, "learning_rate": 4.685541459810982e-05, "loss": 1.3161, "step": 12222 }, { "epoch": 1.23, "learning_rate": 4.6853434816854594e-05, "loss": 0.7401, "step": 12223 }, { "epoch": 1.23, "learning_rate": 4.685145492835053e-05, "loss": 0.0299, "step": 12224 }, { "epoch": 1.23, "learning_rate": 4.6849474932610205e-05, "loss": 0.9609, "step": 12225 }, { "epoch": 1.23, "learning_rate": 4.684749482964623e-05, "loss": 0.7517, "step": 12226 }, { "epoch": 1.23, "learning_rate": 4.684551461947121e-05, "loss": 0.5375, "step": 12227 }, { "epoch": 1.23, "learning_rate": 4.684353430209775e-05, "loss": 0.2802, "step": 12228 }, { "epoch": 1.23, "learning_rate": 4.684155387753844e-05, "loss": 0.8266, "step": 12229 }, { "epoch": 1.23, "learning_rate": 4.683957334580589e-05, "loss": 0.3022, "step": 12230 }, { "epoch": 1.23, "learning_rate": 4.68375927069127e-05, "loss": 1.0414, "step": 12231 }, { "epoch": 1.23, "learning_rate": 4.683561196087148e-05, "loss": 0.3336, "step": 12232 }, { "epoch": 1.23, "learning_rate": 4.6833631107694815e-05, "loss": 0.6168, "step": 12233 }, { "epoch": 1.23, "learning_rate": 4.683165014739534e-05, "loss": 0.4318, "step": 12234 }, { "epoch": 1.23, "learning_rate": 4.6829669079985645e-05, "loss": 0.5221, "step": 12235 }, { "epoch": 1.23, "learning_rate": 4.682768790547833e-05, "loss": 0.5865, "step": 12236 }, { "epoch": 1.23, "learning_rate": 4.6825706623886025e-05, "loss": 1.167, "step": 12237 }, { "epoch": 1.23, "learning_rate": 4.682372523522131e-05, "loss": 1.1171, "step": 12238 }, { "epoch": 1.23, "learning_rate": 4.682174373949681e-05, "loss": 0.2983, "step": 12239 }, { "epoch": 1.23, "learning_rate": 4.6819762136725136e-05, "loss": 0.478, "step": 12240 }, { "epoch": 1.23, "learning_rate": 4.68177804269189e-05, "loss": 0.4228, "step": 12241 }, { "epoch": 1.23, "learning_rate": 4.6815798610090694e-05, "loss": 0.2033, "step": 12242 }, { "epoch": 1.23, "learning_rate": 4.681381668625315e-05, "loss": 0.6903, "step": 12243 }, { "epoch": 1.23, "learning_rate": 4.6811834655418866e-05, "loss": 0.9471, "step": 12244 }, { "epoch": 1.23, "learning_rate": 4.680985251760047e-05, "loss": 0.852, "step": 12245 }, { "epoch": 1.23, "learning_rate": 4.680787027281056e-05, "loss": 0.8043, "step": 12246 }, { "epoch": 1.23, "learning_rate": 4.680588792106177e-05, "loss": 0.3495, "step": 12247 }, { "epoch": 1.23, "learning_rate": 4.680390546236669e-05, "loss": 1.0128, "step": 12248 }, { "epoch": 1.23, "learning_rate": 4.680192289673796e-05, "loss": 0.24, "step": 12249 }, { "epoch": 1.23, "learning_rate": 4.679994022418818e-05, "loss": 0.2087, "step": 12250 }, { "epoch": 1.23, "learning_rate": 4.679795744472997e-05, "loss": 0.1704, "step": 12251 }, { "epoch": 1.23, "learning_rate": 4.679597455837594e-05, "loss": 0.9056, "step": 12252 }, { "epoch": 1.23, "learning_rate": 4.679399156513873e-05, "loss": 0.0463, "step": 12253 }, { "epoch": 1.23, "learning_rate": 4.6792008465030945e-05, "loss": 0.6974, "step": 12254 }, { "epoch": 1.23, "learning_rate": 4.67900252580652e-05, "loss": 0.6086, "step": 12255 }, { "epoch": 1.23, "learning_rate": 4.678804194425413e-05, "loss": 1.0619, "step": 12256 }, { "epoch": 1.23, "learning_rate": 4.6786058523610344e-05, "loss": 0.8061, "step": 12257 }, { "epoch": 1.23, "learning_rate": 4.6784074996146464e-05, "loss": 1.3167, "step": 12258 }, { "epoch": 1.23, "learning_rate": 4.6782091361875124e-05, "loss": 0.5316, "step": 12259 }, { "epoch": 1.23, "learning_rate": 4.678010762080892e-05, "loss": 0.3955, "step": 12260 }, { "epoch": 1.23, "learning_rate": 4.6778123772960514e-05, "loss": 1.0792, "step": 12261 }, { "epoch": 1.23, "learning_rate": 4.67761398183425e-05, "loss": 0.4456, "step": 12262 }, { "epoch": 1.23, "learning_rate": 4.677415575696752e-05, "loss": 0.5339, "step": 12263 }, { "epoch": 1.23, "learning_rate": 4.677217158884819e-05, "loss": 0.477, "step": 12264 }, { "epoch": 1.23, "learning_rate": 4.6770187313997146e-05, "loss": 0.753, "step": 12265 }, { "epoch": 1.23, "learning_rate": 4.676820293242701e-05, "loss": 0.4467, "step": 12266 }, { "epoch": 1.23, "learning_rate": 4.67662184441504e-05, "loss": 1.132, "step": 12267 }, { "epoch": 1.23, "learning_rate": 4.676423384917996e-05, "loss": 0.9011, "step": 12268 }, { "epoch": 1.23, "learning_rate": 4.676224914752832e-05, "loss": 0.7662, "step": 12269 }, { "epoch": 1.23, "learning_rate": 4.676026433920809e-05, "loss": 0.6448, "step": 12270 }, { "epoch": 1.23, "learning_rate": 4.675827942423192e-05, "loss": 0.5375, "step": 12271 }, { "epoch": 1.23, "learning_rate": 4.675629440261244e-05, "loss": 0.8007, "step": 12272 }, { "epoch": 1.23, "learning_rate": 4.675430927436227e-05, "loss": 0.2058, "step": 12273 }, { "epoch": 1.23, "learning_rate": 4.6752324039494045e-05, "loss": 1.352, "step": 12274 }, { "epoch": 1.23, "learning_rate": 4.675033869802041e-05, "loss": 0.3524, "step": 12275 }, { "epoch": 1.23, "learning_rate": 4.6748353249953994e-05, "loss": 0.4078, "step": 12276 }, { "epoch": 1.23, "learning_rate": 4.6746367695307425e-05, "loss": 0.9967, "step": 12277 }, { "epoch": 1.23, "learning_rate": 4.6744382034093344e-05, "loss": 0.2423, "step": 12278 }, { "epoch": 1.23, "learning_rate": 4.674239626632439e-05, "loss": 0.7344, "step": 12279 }, { "epoch": 1.23, "learning_rate": 4.674041039201318e-05, "loss": 0.1692, "step": 12280 }, { "epoch": 1.23, "learning_rate": 4.673842441117238e-05, "loss": 0.6398, "step": 12281 }, { "epoch": 1.23, "learning_rate": 4.673643832381462e-05, "loss": 0.6902, "step": 12282 }, { "epoch": 1.23, "learning_rate": 4.673445212995252e-05, "loss": 0.9926, "step": 12283 }, { "epoch": 1.23, "learning_rate": 4.673246582959873e-05, "loss": 1.0891, "step": 12284 }, { "epoch": 1.23, "learning_rate": 4.673047942276591e-05, "loss": 1.3681, "step": 12285 }, { "epoch": 1.23, "learning_rate": 4.672849290946667e-05, "loss": 0.2754, "step": 12286 }, { "epoch": 1.23, "learning_rate": 4.672650628971367e-05, "loss": 0.8567, "step": 12287 }, { "epoch": 1.23, "learning_rate": 4.672451956351954e-05, "loss": 0.9488, "step": 12288 }, { "epoch": 1.23, "learning_rate": 4.6722532730896944e-05, "loss": 1.1103, "step": 12289 }, { "epoch": 1.23, "learning_rate": 4.6720545791858506e-05, "loss": 0.2961, "step": 12290 }, { "epoch": 1.24, "learning_rate": 4.671855874641687e-05, "loss": 0.3831, "step": 12291 }, { "epoch": 1.24, "learning_rate": 4.671657159458469e-05, "loss": 0.4594, "step": 12292 }, { "epoch": 1.24, "learning_rate": 4.671458433637461e-05, "loss": 0.2592, "step": 12293 }, { "epoch": 1.24, "learning_rate": 4.671259697179926e-05, "loss": 0.4926, "step": 12294 }, { "epoch": 1.24, "learning_rate": 4.671060950087132e-05, "loss": 0.4931, "step": 12295 }, { "epoch": 1.24, "learning_rate": 4.6708621923603404e-05, "loss": 0.6274, "step": 12296 }, { "epoch": 1.24, "learning_rate": 4.670663424000818e-05, "loss": 0.9651, "step": 12297 }, { "epoch": 1.24, "learning_rate": 4.670464645009829e-05, "loss": 0.1589, "step": 12298 }, { "epoch": 1.24, "learning_rate": 4.670265855388639e-05, "loss": 0.3266, "step": 12299 }, { "epoch": 1.24, "learning_rate": 4.670067055138512e-05, "loss": 1.1763, "step": 12300 }, { "epoch": 1.24, "learning_rate": 4.669868244260713e-05, "loss": 0.1739, "step": 12301 }, { "epoch": 1.24, "learning_rate": 4.669669422756509e-05, "loss": 0.4187, "step": 12302 }, { "epoch": 1.24, "learning_rate": 4.669470590627163e-05, "loss": 0.462, "step": 12303 }, { "epoch": 1.24, "learning_rate": 4.6692717478739424e-05, "loss": 0.9499, "step": 12304 }, { "epoch": 1.24, "learning_rate": 4.66907289449811e-05, "loss": 0.3623, "step": 12305 }, { "epoch": 1.24, "learning_rate": 4.6688740305009336e-05, "loss": 0.7634, "step": 12306 }, { "epoch": 1.24, "learning_rate": 4.668675155883677e-05, "loss": 0.7567, "step": 12307 }, { "epoch": 1.24, "learning_rate": 4.668476270647607e-05, "loss": 0.8038, "step": 12308 }, { "epoch": 1.24, "learning_rate": 4.668277374793989e-05, "loss": 1.1941, "step": 12309 }, { "epoch": 1.24, "learning_rate": 4.6680784683240883e-05, "loss": 1.3536, "step": 12310 }, { "epoch": 1.24, "learning_rate": 4.66787955123917e-05, "loss": 0.2761, "step": 12311 }, { "epoch": 1.24, "learning_rate": 4.6676806235405015e-05, "loss": 1.4459, "step": 12312 }, { "epoch": 1.24, "learning_rate": 4.6674816852293475e-05, "loss": 0.7049, "step": 12313 }, { "epoch": 1.24, "learning_rate": 4.6672827363069755e-05, "loss": 0.7825, "step": 12314 }, { "epoch": 1.24, "learning_rate": 4.667083776774649e-05, "loss": 0.4808, "step": 12315 }, { "epoch": 1.24, "learning_rate": 4.666884806633636e-05, "loss": 1.2034, "step": 12316 }, { "epoch": 1.24, "learning_rate": 4.666685825885202e-05, "loss": 0.6214, "step": 12317 }, { "epoch": 1.24, "learning_rate": 4.666486834530615e-05, "loss": 0.9084, "step": 12318 }, { "epoch": 1.24, "learning_rate": 4.666287832571138e-05, "loss": 0.3741, "step": 12319 }, { "epoch": 1.24, "learning_rate": 4.6660888200080396e-05, "loss": 0.2472, "step": 12320 }, { "epoch": 1.24, "learning_rate": 4.665889796842586e-05, "loss": 0.1175, "step": 12321 }, { "epoch": 1.24, "learning_rate": 4.6656907630760434e-05, "loss": 0.4518, "step": 12322 }, { "epoch": 1.24, "learning_rate": 4.665491718709678e-05, "loss": 0.7013, "step": 12323 }, { "epoch": 1.24, "learning_rate": 4.665292663744757e-05, "loss": 0.7034, "step": 12324 }, { "epoch": 1.24, "learning_rate": 4.665093598182548e-05, "loss": 0.4566, "step": 12325 }, { "epoch": 1.24, "learning_rate": 4.664894522024315e-05, "loss": 0.3646, "step": 12326 }, { "epoch": 1.24, "learning_rate": 4.6646954352713284e-05, "loss": 0.555, "step": 12327 }, { "epoch": 1.24, "learning_rate": 4.664496337924852e-05, "loss": 0.4868, "step": 12328 }, { "epoch": 1.24, "learning_rate": 4.664297229986155e-05, "loss": 1.1158, "step": 12329 }, { "epoch": 1.24, "learning_rate": 4.6640981114565025e-05, "loss": 0.8511, "step": 12330 }, { "epoch": 1.24, "learning_rate": 4.663898982337164e-05, "loss": 0.61, "step": 12331 }, { "epoch": 1.24, "learning_rate": 4.6636998426294046e-05, "loss": 0.1713, "step": 12332 }, { "epoch": 1.24, "learning_rate": 4.663500692334492e-05, "loss": 0.6778, "step": 12333 }, { "epoch": 1.24, "learning_rate": 4.6633015314536944e-05, "loss": 1.0395, "step": 12334 }, { "epoch": 1.24, "learning_rate": 4.6631023599882785e-05, "loss": 0.9941, "step": 12335 }, { "epoch": 1.24, "learning_rate": 4.662903177939512e-05, "loss": 0.3196, "step": 12336 }, { "epoch": 1.24, "learning_rate": 4.6627039853086616e-05, "loss": 0.3244, "step": 12337 }, { "epoch": 1.24, "learning_rate": 4.662504782096997e-05, "loss": 0.5977, "step": 12338 }, { "epoch": 1.24, "learning_rate": 4.662305568305783e-05, "loss": 0.3257, "step": 12339 }, { "epoch": 1.24, "learning_rate": 4.66210634393629e-05, "loss": 0.4389, "step": 12340 }, { "epoch": 1.24, "learning_rate": 4.661907108989783e-05, "loss": 0.2927, "step": 12341 }, { "epoch": 1.24, "learning_rate": 4.661707863467532e-05, "loss": 0.214, "step": 12342 }, { "epoch": 1.24, "learning_rate": 4.661508607370804e-05, "loss": 1.033, "step": 12343 }, { "epoch": 1.24, "learning_rate": 4.6613093407008686e-05, "loss": 0.3707, "step": 12344 }, { "epoch": 1.24, "learning_rate": 4.661110063458992e-05, "loss": 0.621, "step": 12345 }, { "epoch": 1.24, "learning_rate": 4.6609107756464424e-05, "loss": 1.3668, "step": 12346 }, { "epoch": 1.24, "learning_rate": 4.660711477264488e-05, "loss": 0.3909, "step": 12347 }, { "epoch": 1.24, "learning_rate": 4.660512168314399e-05, "loss": 0.1494, "step": 12348 }, { "epoch": 1.24, "learning_rate": 4.6603128487974405e-05, "loss": 0.6191, "step": 12349 }, { "epoch": 1.24, "learning_rate": 4.660113518714884e-05, "loss": 0.8682, "step": 12350 }, { "epoch": 1.24, "learning_rate": 4.659914178067996e-05, "loss": 0.2716, "step": 12351 }, { "epoch": 1.24, "learning_rate": 4.659714826858046e-05, "loss": 0.7453, "step": 12352 }, { "epoch": 1.24, "learning_rate": 4.6595154650863015e-05, "loss": 0.6717, "step": 12353 }, { "epoch": 1.24, "learning_rate": 4.659316092754033e-05, "loss": 0.9232, "step": 12354 }, { "epoch": 1.24, "learning_rate": 4.659116709862508e-05, "loss": 0.7625, "step": 12355 }, { "epoch": 1.24, "learning_rate": 4.658917316412994e-05, "loss": 0.5138, "step": 12356 }, { "epoch": 1.24, "learning_rate": 4.658717912406762e-05, "loss": 0.5486, "step": 12357 }, { "epoch": 1.24, "learning_rate": 4.6585184978450806e-05, "loss": 0.8946, "step": 12358 }, { "epoch": 1.24, "learning_rate": 4.658319072729218e-05, "loss": 0.7651, "step": 12359 }, { "epoch": 1.24, "learning_rate": 4.658119637060444e-05, "loss": 0.6459, "step": 12360 }, { "epoch": 1.24, "learning_rate": 4.657920190840027e-05, "loss": 0.1531, "step": 12361 }, { "epoch": 1.24, "learning_rate": 4.657720734069238e-05, "loss": 1.0742, "step": 12362 }, { "epoch": 1.24, "learning_rate": 4.657521266749344e-05, "loss": 0.7019, "step": 12363 }, { "epoch": 1.24, "learning_rate": 4.6573217888816144e-05, "loss": 0.575, "step": 12364 }, { "epoch": 1.24, "learning_rate": 4.6571223004673205e-05, "loss": 1.2329, "step": 12365 }, { "epoch": 1.24, "learning_rate": 4.65692280150773e-05, "loss": 0.4778, "step": 12366 }, { "epoch": 1.24, "learning_rate": 4.656723292004113e-05, "loss": 0.8374, "step": 12367 }, { "epoch": 1.24, "learning_rate": 4.65652377195774e-05, "loss": 0.5171, "step": 12368 }, { "epoch": 1.24, "learning_rate": 4.656324241369879e-05, "loss": 1.015, "step": 12369 }, { "epoch": 1.24, "learning_rate": 4.6561247002418014e-05, "loss": 0.4898, "step": 12370 }, { "epoch": 1.24, "learning_rate": 4.655925148574775e-05, "loss": 1.0573, "step": 12371 }, { "epoch": 1.24, "learning_rate": 4.6557255863700725e-05, "loss": 0.2013, "step": 12372 }, { "epoch": 1.24, "learning_rate": 4.655526013628962e-05, "loss": 0.6237, "step": 12373 }, { "epoch": 1.24, "learning_rate": 4.655326430352713e-05, "loss": 0.2897, "step": 12374 }, { "epoch": 1.24, "learning_rate": 4.6551268365425967e-05, "loss": 0.6358, "step": 12375 }, { "epoch": 1.24, "learning_rate": 4.654927232199883e-05, "loss": 0.2898, "step": 12376 }, { "epoch": 1.24, "learning_rate": 4.654727617325842e-05, "loss": 1.2447, "step": 12377 }, { "epoch": 1.24, "learning_rate": 4.654527991921743e-05, "loss": 1.2804, "step": 12378 }, { "epoch": 1.24, "learning_rate": 4.654328355988858e-05, "loss": 1.1412, "step": 12379 }, { "epoch": 1.24, "learning_rate": 4.654128709528457e-05, "loss": 0.2792, "step": 12380 }, { "epoch": 1.24, "learning_rate": 4.65392905254181e-05, "loss": 0.9149, "step": 12381 }, { "epoch": 1.24, "learning_rate": 4.6537293850301864e-05, "loss": 0.9897, "step": 12382 }, { "epoch": 1.24, "learning_rate": 4.65352970699486e-05, "loss": 0.4778, "step": 12383 }, { "epoch": 1.24, "learning_rate": 4.653330018437099e-05, "loss": 1.003, "step": 12384 }, { "epoch": 1.24, "learning_rate": 4.653130319358173e-05, "loss": 0.6977, "step": 12385 }, { "epoch": 1.24, "learning_rate": 4.652930609759357e-05, "loss": 0.4937, "step": 12386 }, { "epoch": 1.24, "learning_rate": 4.652730889641918e-05, "loss": 1.1994, "step": 12387 }, { "epoch": 1.24, "learning_rate": 4.652531159007129e-05, "loss": 0.7341, "step": 12388 }, { "epoch": 1.25, "learning_rate": 4.65233141785626e-05, "loss": 0.7125, "step": 12389 }, { "epoch": 1.25, "learning_rate": 4.652131666190583e-05, "loss": 0.3094, "step": 12390 }, { "epoch": 1.25, "learning_rate": 4.651931904011368e-05, "loss": 1.1531, "step": 12391 }, { "epoch": 1.25, "learning_rate": 4.651732131319887e-05, "loss": 0.8412, "step": 12392 }, { "epoch": 1.25, "learning_rate": 4.651532348117411e-05, "loss": 0.4753, "step": 12393 }, { "epoch": 1.25, "learning_rate": 4.651332554405211e-05, "loss": 1.0117, "step": 12394 }, { "epoch": 1.25, "learning_rate": 4.651132750184559e-05, "loss": 0.8223, "step": 12395 }, { "epoch": 1.25, "learning_rate": 4.650932935456727e-05, "loss": 0.7787, "step": 12396 }, { "epoch": 1.25, "learning_rate": 4.650733110222985e-05, "loss": 1.2352, "step": 12397 }, { "epoch": 1.25, "learning_rate": 4.6505332744846057e-05, "loss": 0.5998, "step": 12398 }, { "epoch": 1.25, "learning_rate": 4.650333428242861e-05, "loss": 0.4667, "step": 12399 }, { "epoch": 1.25, "learning_rate": 4.6501335714990225e-05, "loss": 0.6055, "step": 12400 }, { "epoch": 1.25, "learning_rate": 4.6499337042543604e-05, "loss": 0.9702, "step": 12401 }, { "epoch": 1.25, "learning_rate": 4.6497338265101495e-05, "loss": 1.1112, "step": 12402 }, { "epoch": 1.25, "learning_rate": 4.649533938267658e-05, "loss": 0.6241, "step": 12403 }, { "epoch": 1.25, "learning_rate": 4.649334039528161e-05, "loss": 0.5015, "step": 12404 }, { "epoch": 1.25, "learning_rate": 4.6491341302929304e-05, "loss": 0.4307, "step": 12405 }, { "epoch": 1.25, "learning_rate": 4.6489342105632365e-05, "loss": 1.1153, "step": 12406 }, { "epoch": 1.25, "learning_rate": 4.648734280340353e-05, "loss": 0.6699, "step": 12407 }, { "epoch": 1.25, "learning_rate": 4.6485343396255516e-05, "loss": 0.6763, "step": 12408 }, { "epoch": 1.25, "learning_rate": 4.648334388420105e-05, "loss": 0.6216, "step": 12409 }, { "epoch": 1.25, "learning_rate": 4.648134426725286e-05, "loss": 0.5122, "step": 12410 }, { "epoch": 1.25, "learning_rate": 4.647934454542365e-05, "loss": 0.4608, "step": 12411 }, { "epoch": 1.25, "learning_rate": 4.6477344718726175e-05, "loss": 0.5017, "step": 12412 }, { "epoch": 1.25, "learning_rate": 4.6475344787173135e-05, "loss": 0.3213, "step": 12413 }, { "epoch": 1.25, "learning_rate": 4.6473344750777277e-05, "loss": 1.0406, "step": 12414 }, { "epoch": 1.25, "learning_rate": 4.647134460955131e-05, "loss": 0.6115, "step": 12415 }, { "epoch": 1.25, "learning_rate": 4.646934436350799e-05, "loss": 0.5139, "step": 12416 }, { "epoch": 1.25, "learning_rate": 4.6467344012660014e-05, "loss": 0.3433, "step": 12417 }, { "epoch": 1.25, "learning_rate": 4.646534355702013e-05, "loss": 0.5173, "step": 12418 }, { "epoch": 1.25, "learning_rate": 4.646334299660106e-05, "loss": 0.7101, "step": 12419 }, { "epoch": 1.25, "learning_rate": 4.646134233141553e-05, "loss": 0.3343, "step": 12420 }, { "epoch": 1.25, "learning_rate": 4.64593415614763e-05, "loss": 1.0618, "step": 12421 }, { "epoch": 1.25, "learning_rate": 4.645734068679606e-05, "loss": 0.9228, "step": 12422 }, { "epoch": 1.25, "learning_rate": 4.6455339707387574e-05, "loss": 0.2156, "step": 12423 }, { "epoch": 1.25, "learning_rate": 4.645333862326357e-05, "loss": 0.8734, "step": 12424 }, { "epoch": 1.25, "learning_rate": 4.6451337434436776e-05, "loss": 0.7406, "step": 12425 }, { "epoch": 1.25, "learning_rate": 4.6449336140919924e-05, "loss": 0.1214, "step": 12426 }, { "epoch": 1.25, "learning_rate": 4.644733474272576e-05, "loss": 0.7507, "step": 12427 }, { "epoch": 1.25, "learning_rate": 4.6445333239867016e-05, "loss": 1.006, "step": 12428 }, { "epoch": 1.25, "learning_rate": 4.6443331632356426e-05, "loss": 0.3033, "step": 12429 }, { "epoch": 1.25, "learning_rate": 4.6441329920206726e-05, "loss": 0.6145, "step": 12430 }, { "epoch": 1.25, "learning_rate": 4.643932810343066e-05, "loss": 0.1565, "step": 12431 }, { "epoch": 1.25, "learning_rate": 4.643732618204096e-05, "loss": 1.0078, "step": 12432 }, { "epoch": 1.25, "learning_rate": 4.643532415605037e-05, "loss": 0.3551, "step": 12433 }, { "epoch": 1.25, "learning_rate": 4.643332202547163e-05, "loss": 0.6743, "step": 12434 }, { "epoch": 1.25, "learning_rate": 4.643131979031748e-05, "loss": 0.6956, "step": 12435 }, { "epoch": 1.25, "learning_rate": 4.642931745060066e-05, "loss": 0.9503, "step": 12436 }, { "epoch": 1.25, "learning_rate": 4.642731500633392e-05, "loss": 0.6425, "step": 12437 }, { "epoch": 1.25, "learning_rate": 4.642531245752999e-05, "loss": 0.3433, "step": 12438 }, { "epoch": 1.25, "learning_rate": 4.6423309804201626e-05, "loss": 0.6711, "step": 12439 }, { "epoch": 1.25, "learning_rate": 4.6421307046361564e-05, "loss": 0.6506, "step": 12440 }, { "epoch": 1.25, "learning_rate": 4.641930418402256e-05, "loss": 0.7032, "step": 12441 }, { "epoch": 1.25, "learning_rate": 4.641730121719733e-05, "loss": 0.5768, "step": 12442 }, { "epoch": 1.25, "learning_rate": 4.641529814589865e-05, "loss": 0.1807, "step": 12443 }, { "epoch": 1.25, "learning_rate": 4.641329497013926e-05, "loss": 0.3632, "step": 12444 }, { "epoch": 1.25, "learning_rate": 4.64112916899319e-05, "loss": 0.4742, "step": 12445 }, { "epoch": 1.25, "learning_rate": 4.6409288305289326e-05, "loss": 0.7755, "step": 12446 }, { "epoch": 1.25, "learning_rate": 4.640728481622429e-05, "loss": 0.682, "step": 12447 }, { "epoch": 1.25, "learning_rate": 4.640528122274953e-05, "loss": 0.2454, "step": 12448 }, { "epoch": 1.25, "learning_rate": 4.640327752487779e-05, "loss": 0.5216, "step": 12449 }, { "epoch": 1.25, "learning_rate": 4.6401273722621844e-05, "loss": 1.1276, "step": 12450 }, { "epoch": 1.25, "learning_rate": 4.6399269815994424e-05, "loss": 0.4703, "step": 12451 }, { "epoch": 1.25, "learning_rate": 4.63972658050083e-05, "loss": 0.7768, "step": 12452 }, { "epoch": 1.25, "learning_rate": 4.639526168967621e-05, "loss": 0.9414, "step": 12453 }, { "epoch": 1.25, "learning_rate": 4.639325747001091e-05, "loss": 0.5487, "step": 12454 }, { "epoch": 1.25, "learning_rate": 4.6391253146025155e-05, "loss": 0.6704, "step": 12455 }, { "epoch": 1.25, "learning_rate": 4.6389248717731705e-05, "loss": 0.362, "step": 12456 }, { "epoch": 1.25, "learning_rate": 4.6387244185143305e-05, "loss": 0.4247, "step": 12457 }, { "epoch": 1.25, "learning_rate": 4.638523954827272e-05, "loss": 0.5331, "step": 12458 }, { "epoch": 1.25, "learning_rate": 4.63832348071327e-05, "loss": 0.9263, "step": 12459 }, { "epoch": 1.25, "learning_rate": 4.6381229961736016e-05, "loss": 0.6291, "step": 12460 }, { "epoch": 1.25, "learning_rate": 4.6379225012095413e-05, "loss": 0.083, "step": 12461 }, { "epoch": 1.25, "learning_rate": 4.6377219958223646e-05, "loss": 0.6193, "step": 12462 }, { "epoch": 1.25, "learning_rate": 4.6375214800133486e-05, "loss": 0.3977, "step": 12463 }, { "epoch": 1.25, "learning_rate": 4.637320953783768e-05, "loss": 0.2889, "step": 12464 }, { "epoch": 1.25, "learning_rate": 4.637120417134901e-05, "loss": 0.7858, "step": 12465 }, { "epoch": 1.25, "learning_rate": 4.636919870068022e-05, "loss": 0.2354, "step": 12466 }, { "epoch": 1.25, "learning_rate": 4.636719312584408e-05, "loss": 0.4991, "step": 12467 }, { "epoch": 1.25, "learning_rate": 4.636518744685334e-05, "loss": 0.2894, "step": 12468 }, { "epoch": 1.25, "learning_rate": 4.636318166372078e-05, "loss": 0.1382, "step": 12469 }, { "epoch": 1.25, "learning_rate": 4.6361175776459156e-05, "loss": 0.9184, "step": 12470 }, { "epoch": 1.25, "learning_rate": 4.635916978508123e-05, "loss": 0.5451, "step": 12471 }, { "epoch": 1.25, "learning_rate": 4.6357163689599774e-05, "loss": 0.4322, "step": 12472 }, { "epoch": 1.25, "learning_rate": 4.6355157490027556e-05, "loss": 1.162, "step": 12473 }, { "epoch": 1.25, "learning_rate": 4.635315118637733e-05, "loss": 0.6452, "step": 12474 }, { "epoch": 1.25, "learning_rate": 4.635114477866186e-05, "loss": 0.5432, "step": 12475 }, { "epoch": 1.25, "learning_rate": 4.634913826689395e-05, "loss": 0.4652, "step": 12476 }, { "epoch": 1.25, "learning_rate": 4.634713165108633e-05, "loss": 0.9728, "step": 12477 }, { "epoch": 1.25, "learning_rate": 4.634512493125178e-05, "loss": 0.2934, "step": 12478 }, { "epoch": 1.25, "learning_rate": 4.6343118107403076e-05, "loss": 0.1755, "step": 12479 }, { "epoch": 1.25, "learning_rate": 4.6341111179553e-05, "loss": 0.2253, "step": 12480 }, { "epoch": 1.25, "learning_rate": 4.633910414771429e-05, "loss": 1.2963, "step": 12481 }, { "epoch": 1.25, "learning_rate": 4.6337097011899735e-05, "loss": 0.6424, "step": 12482 }, { "epoch": 1.25, "learning_rate": 4.6335089772122125e-05, "loss": 0.1052, "step": 12483 }, { "epoch": 1.25, "learning_rate": 4.6333082428394214e-05, "loss": 0.5898, "step": 12484 }, { "epoch": 1.25, "learning_rate": 4.6331074980728766e-05, "loss": 0.6787, "step": 12485 }, { "epoch": 1.25, "learning_rate": 4.632906742913859e-05, "loss": 0.2606, "step": 12486 }, { "epoch": 1.25, "learning_rate": 4.6327059773636424e-05, "loss": 0.2997, "step": 12487 }, { "epoch": 1.26, "learning_rate": 4.632505201423506e-05, "loss": 0.9254, "step": 12488 }, { "epoch": 1.26, "learning_rate": 4.632304415094729e-05, "loss": 0.9871, "step": 12489 }, { "epoch": 1.26, "learning_rate": 4.632103618378587e-05, "loss": 0.4342, "step": 12490 }, { "epoch": 1.26, "learning_rate": 4.631902811276358e-05, "loss": 0.583, "step": 12491 }, { "epoch": 1.26, "learning_rate": 4.63170199378932e-05, "loss": 0.4504, "step": 12492 }, { "epoch": 1.26, "learning_rate": 4.6315011659187524e-05, "loss": 0.3213, "step": 12493 }, { "epoch": 1.26, "learning_rate": 4.6313003276659314e-05, "loss": 0.436, "step": 12494 }, { "epoch": 1.26, "learning_rate": 4.631099479032136e-05, "loss": 0.977, "step": 12495 }, { "epoch": 1.26, "learning_rate": 4.630898620018644e-05, "loss": 1.6641, "step": 12496 }, { "epoch": 1.26, "learning_rate": 4.6306977506267336e-05, "loss": 0.1657, "step": 12497 }, { "epoch": 1.26, "learning_rate": 4.6304968708576824e-05, "loss": 0.5464, "step": 12498 }, { "epoch": 1.26, "learning_rate": 4.63029598071277e-05, "loss": 0.3789, "step": 12499 }, { "epoch": 1.26, "learning_rate": 4.630095080193274e-05, "loss": 0.7146, "step": 12500 }, { "epoch": 1.26, "learning_rate": 4.629894169300473e-05, "loss": 1.5818, "step": 12501 }, { "epoch": 1.26, "learning_rate": 4.6296932480356465e-05, "loss": 0.3516, "step": 12502 }, { "epoch": 1.26, "learning_rate": 4.629492316400071e-05, "loss": 0.8859, "step": 12503 }, { "epoch": 1.26, "learning_rate": 4.6292913743950265e-05, "loss": 0.2941, "step": 12504 }, { "epoch": 1.26, "learning_rate": 4.629090422021791e-05, "loss": 1.1469, "step": 12505 }, { "epoch": 1.26, "learning_rate": 4.628889459281645e-05, "loss": 1.0053, "step": 12506 }, { "epoch": 1.26, "learning_rate": 4.628688486175865e-05, "loss": 0.726, "step": 12507 }, { "epoch": 1.26, "learning_rate": 4.6284875027057315e-05, "loss": 1.0124, "step": 12508 }, { "epoch": 1.26, "learning_rate": 4.628286508872523e-05, "loss": 1.2471, "step": 12509 }, { "epoch": 1.26, "learning_rate": 4.628085504677519e-05, "loss": 0.9027, "step": 12510 }, { "epoch": 1.26, "learning_rate": 4.627884490121998e-05, "loss": 0.4686, "step": 12511 }, { "epoch": 1.26, "learning_rate": 4.62768346520724e-05, "loss": 0.5029, "step": 12512 }, { "epoch": 1.26, "learning_rate": 4.627482429934523e-05, "loss": 1.2748, "step": 12513 }, { "epoch": 1.26, "learning_rate": 4.6272813843051266e-05, "loss": 0.6338, "step": 12514 }, { "epoch": 1.26, "learning_rate": 4.627080328320331e-05, "loss": 0.2051, "step": 12515 }, { "epoch": 1.26, "learning_rate": 4.626879261981414e-05, "loss": 0.4, "step": 12516 }, { "epoch": 1.26, "learning_rate": 4.626678185289658e-05, "loss": 0.2634, "step": 12517 }, { "epoch": 1.26, "learning_rate": 4.62647709824634e-05, "loss": 0.7294, "step": 12518 }, { "epoch": 1.26, "learning_rate": 4.6262760008527403e-05, "loss": 1.4172, "step": 12519 }, { "epoch": 1.26, "learning_rate": 4.626074893110139e-05, "loss": 0.6796, "step": 12520 }, { "epoch": 1.26, "learning_rate": 4.6258737750198155e-05, "loss": 0.5824, "step": 12521 }, { "epoch": 1.26, "learning_rate": 4.6256726465830504e-05, "loss": 0.5, "step": 12522 }, { "epoch": 1.26, "learning_rate": 4.625471507801123e-05, "loss": 0.2452, "step": 12523 }, { "epoch": 1.26, "learning_rate": 4.625270358675312e-05, "loss": 0.7722, "step": 12524 }, { "epoch": 1.26, "learning_rate": 4.6250691992069006e-05, "loss": 0.5192, "step": 12525 }, { "epoch": 1.26, "learning_rate": 4.6248680293971656e-05, "loss": 1.1044, "step": 12526 }, { "epoch": 1.26, "learning_rate": 4.624666849247389e-05, "loss": 0.146, "step": 12527 }, { "epoch": 1.26, "learning_rate": 4.624465658758851e-05, "loss": 0.0583, "step": 12528 }, { "epoch": 1.26, "learning_rate": 4.624264457932831e-05, "loss": 0.6187, "step": 12529 }, { "epoch": 1.26, "learning_rate": 4.62406324677061e-05, "loss": 0.816, "step": 12530 }, { "epoch": 1.26, "learning_rate": 4.6238620252734685e-05, "loss": 0.2208, "step": 12531 }, { "epoch": 1.26, "learning_rate": 4.623660793442687e-05, "loss": 1.1293, "step": 12532 }, { "epoch": 1.26, "learning_rate": 4.623459551279546e-05, "loss": 0.6916, "step": 12533 }, { "epoch": 1.26, "learning_rate": 4.623258298785325e-05, "loss": 0.9291, "step": 12534 }, { "epoch": 1.26, "learning_rate": 4.623057035961307e-05, "loss": 0.4464, "step": 12535 }, { "epoch": 1.26, "learning_rate": 4.6228557628087706e-05, "loss": 0.8032, "step": 12536 }, { "epoch": 1.26, "learning_rate": 4.6226544793289976e-05, "loss": 0.6289, "step": 12537 }, { "epoch": 1.26, "learning_rate": 4.6224531855232696e-05, "loss": 1.2742, "step": 12538 }, { "epoch": 1.26, "learning_rate": 4.622251881392866e-05, "loss": 0.2889, "step": 12539 }, { "epoch": 1.26, "learning_rate": 4.622050566939068e-05, "loss": 1.6198, "step": 12540 }, { "epoch": 1.26, "learning_rate": 4.621849242163158e-05, "loss": 0.2631, "step": 12541 }, { "epoch": 1.26, "learning_rate": 4.621647907066417e-05, "loss": 1.171, "step": 12542 }, { "epoch": 1.26, "learning_rate": 4.6214465616501235e-05, "loss": 0.8949, "step": 12543 }, { "epoch": 1.26, "learning_rate": 4.621245205915563e-05, "loss": 0.4697, "step": 12544 }, { "epoch": 1.26, "learning_rate": 4.621043839864014e-05, "loss": 0.3666, "step": 12545 }, { "epoch": 1.26, "learning_rate": 4.620842463496759e-05, "loss": 0.7613, "step": 12546 }, { "epoch": 1.26, "learning_rate": 4.620641076815078e-05, "loss": 0.9717, "step": 12547 }, { "epoch": 1.26, "learning_rate": 4.620439679820256e-05, "loss": 0.7318, "step": 12548 }, { "epoch": 1.26, "learning_rate": 4.62023827251357e-05, "loss": 0.3668, "step": 12549 }, { "epoch": 1.26, "learning_rate": 4.620036854896305e-05, "loss": 0.5383, "step": 12550 }, { "epoch": 1.26, "learning_rate": 4.619835426969742e-05, "loss": 0.2575, "step": 12551 }, { "epoch": 1.26, "learning_rate": 4.6196339887351624e-05, "loss": 0.5987, "step": 12552 }, { "epoch": 1.26, "learning_rate": 4.619432540193848e-05, "loss": 0.2032, "step": 12553 }, { "epoch": 1.26, "learning_rate": 4.619231081347081e-05, "loss": 0.3783, "step": 12554 }, { "epoch": 1.26, "learning_rate": 4.6190296121961434e-05, "loss": 1.0912, "step": 12555 }, { "epoch": 1.26, "learning_rate": 4.6188281327423175e-05, "loss": 0.6929, "step": 12556 }, { "epoch": 1.26, "learning_rate": 4.618626642986885e-05, "loss": 0.9935, "step": 12557 }, { "epoch": 1.26, "learning_rate": 4.618425142931128e-05, "loss": 0.8519, "step": 12558 }, { "epoch": 1.26, "learning_rate": 4.61822363257633e-05, "loss": 1.1023, "step": 12559 }, { "epoch": 1.26, "learning_rate": 4.618022111923771e-05, "loss": 0.7586, "step": 12560 }, { "epoch": 1.26, "learning_rate": 4.6178205809747355e-05, "loss": 0.4861, "step": 12561 }, { "epoch": 1.26, "learning_rate": 4.617619039730506e-05, "loss": 1.3957, "step": 12562 }, { "epoch": 1.26, "learning_rate": 4.617417488192363e-05, "loss": 2.2313, "step": 12563 }, { "epoch": 1.26, "learning_rate": 4.617215926361591e-05, "loss": 0.98, "step": 12564 }, { "epoch": 1.26, "learning_rate": 4.617014354239472e-05, "loss": 0.88, "step": 12565 }, { "epoch": 1.26, "learning_rate": 4.61681277182729e-05, "loss": 0.857, "step": 12566 }, { "epoch": 1.26, "learning_rate": 4.616611179126325e-05, "loss": 0.7592, "step": 12567 }, { "epoch": 1.26, "learning_rate": 4.616409576137862e-05, "loss": 0.3377, "step": 12568 }, { "epoch": 1.26, "learning_rate": 4.6162079628631837e-05, "loss": 1.1361, "step": 12569 }, { "epoch": 1.26, "learning_rate": 4.616006339303572e-05, "loss": 0.5938, "step": 12570 }, { "epoch": 1.26, "learning_rate": 4.6158047054603114e-05, "loss": 1.1553, "step": 12571 }, { "epoch": 1.26, "learning_rate": 4.6156030613346835e-05, "loss": 0.303, "step": 12572 }, { "epoch": 1.26, "learning_rate": 4.615401406927974e-05, "loss": 0.7528, "step": 12573 }, { "epoch": 1.26, "learning_rate": 4.615199742241462e-05, "loss": 0.2931, "step": 12574 }, { "epoch": 1.26, "learning_rate": 4.614998067276436e-05, "loss": 1.1916, "step": 12575 }, { "epoch": 1.26, "learning_rate": 4.614796382034175e-05, "loss": 0.336, "step": 12576 }, { "epoch": 1.26, "learning_rate": 4.614594686515964e-05, "loss": 0.3638, "step": 12577 }, { "epoch": 1.26, "learning_rate": 4.614392980723088e-05, "loss": 0.5544, "step": 12578 }, { "epoch": 1.26, "learning_rate": 4.614191264656827e-05, "loss": 0.4822, "step": 12579 }, { "epoch": 1.26, "learning_rate": 4.613989538318469e-05, "loss": 0.3276, "step": 12580 }, { "epoch": 1.26, "learning_rate": 4.613787801709294e-05, "loss": 0.9168, "step": 12581 }, { "epoch": 1.26, "learning_rate": 4.6135860548305884e-05, "loss": 0.9102, "step": 12582 }, { "epoch": 1.26, "learning_rate": 4.613384297683635e-05, "loss": 0.7521, "step": 12583 }, { "epoch": 1.26, "learning_rate": 4.613182530269717e-05, "loss": 0.3703, "step": 12584 }, { "epoch": 1.26, "learning_rate": 4.61298075259012e-05, "loss": 0.7705, "step": 12585 }, { "epoch": 1.27, "learning_rate": 4.6127789646461263e-05, "loss": 0.7291, "step": 12586 }, { "epoch": 1.27, "learning_rate": 4.612577166439021e-05, "loss": 0.3712, "step": 12587 }, { "epoch": 1.27, "learning_rate": 4.612375357970088e-05, "loss": 0.2801, "step": 12588 }, { "epoch": 1.27, "learning_rate": 4.612173539240612e-05, "loss": 0.4744, "step": 12589 }, { "epoch": 1.27, "learning_rate": 4.6119717102518754e-05, "loss": 0.0943, "step": 12590 }, { "epoch": 1.27, "learning_rate": 4.611769871005166e-05, "loss": 0.3496, "step": 12591 }, { "epoch": 1.27, "learning_rate": 4.6115680215017655e-05, "loss": 1.0046, "step": 12592 }, { "epoch": 1.27, "learning_rate": 4.6113661617429595e-05, "loss": 0.3258, "step": 12593 }, { "epoch": 1.27, "learning_rate": 4.611164291730032e-05, "loss": 0.3139, "step": 12594 }, { "epoch": 1.27, "learning_rate": 4.610962411464268e-05, "loss": 0.5888, "step": 12595 }, { "epoch": 1.27, "learning_rate": 4.610760520946952e-05, "loss": 1.1205, "step": 12596 }, { "epoch": 1.27, "learning_rate": 4.610558620179368e-05, "loss": 1.0607, "step": 12597 }, { "epoch": 1.27, "learning_rate": 4.610356709162802e-05, "loss": 0.6308, "step": 12598 }, { "epoch": 1.27, "learning_rate": 4.61015478789854e-05, "loss": 0.5878, "step": 12599 }, { "epoch": 1.27, "learning_rate": 4.609952856387864e-05, "loss": 0.2669, "step": 12600 }, { "epoch": 1.27, "learning_rate": 4.6097509146320607e-05, "loss": 1.1494, "step": 12601 }, { "epoch": 1.27, "learning_rate": 4.609548962632414e-05, "loss": 0.2276, "step": 12602 }, { "epoch": 1.27, "learning_rate": 4.6093470003902125e-05, "loss": 0.4384, "step": 12603 }, { "epoch": 1.27, "learning_rate": 4.609145027906737e-05, "loss": 1.0719, "step": 12604 }, { "epoch": 1.27, "learning_rate": 4.6089430451832746e-05, "loss": 0.6611, "step": 12605 }, { "epoch": 1.27, "learning_rate": 4.608741052221112e-05, "loss": 0.6656, "step": 12606 }, { "epoch": 1.27, "learning_rate": 4.608539049021533e-05, "loss": 1.2172, "step": 12607 }, { "epoch": 1.27, "learning_rate": 4.608337035585823e-05, "loss": 0.3837, "step": 12608 }, { "epoch": 1.27, "learning_rate": 4.608135011915268e-05, "loss": 0.6939, "step": 12609 }, { "epoch": 1.27, "learning_rate": 4.6079329780111534e-05, "loss": 0.4015, "step": 12610 }, { "epoch": 1.27, "learning_rate": 4.6077309338747644e-05, "loss": 0.7591, "step": 12611 }, { "epoch": 1.27, "learning_rate": 4.607528879507388e-05, "loss": 1.0089, "step": 12612 }, { "epoch": 1.27, "learning_rate": 4.60732681491031e-05, "loss": 0.8008, "step": 12613 }, { "epoch": 1.27, "learning_rate": 4.607124740084815e-05, "loss": 0.6227, "step": 12614 }, { "epoch": 1.27, "learning_rate": 4.60692265503219e-05, "loss": 0.7271, "step": 12615 }, { "epoch": 1.27, "learning_rate": 4.6067205597537196e-05, "loss": 0.5321, "step": 12616 }, { "epoch": 1.27, "learning_rate": 4.6065184542506914e-05, "loss": 0.8763, "step": 12617 }, { "epoch": 1.27, "learning_rate": 4.60631633852439e-05, "loss": 0.7913, "step": 12618 }, { "epoch": 1.27, "learning_rate": 4.606114212576105e-05, "loss": 0.8248, "step": 12619 }, { "epoch": 1.27, "learning_rate": 4.605912076407118e-05, "loss": 1.0827, "step": 12620 }, { "epoch": 1.27, "learning_rate": 4.6057099300187176e-05, "loss": 0.2236, "step": 12621 }, { "epoch": 1.27, "learning_rate": 4.6055077734121906e-05, "loss": 0.8587, "step": 12622 }, { "epoch": 1.27, "learning_rate": 4.605305606588822e-05, "loss": 0.5876, "step": 12623 }, { "epoch": 1.27, "learning_rate": 4.605103429549901e-05, "loss": 0.6208, "step": 12624 }, { "epoch": 1.27, "learning_rate": 4.604901242296711e-05, "loss": 0.2781, "step": 12625 }, { "epoch": 1.27, "learning_rate": 4.60469904483054e-05, "loss": 1.3226, "step": 12626 }, { "epoch": 1.27, "learning_rate": 4.6044968371526746e-05, "loss": 1.0383, "step": 12627 }, { "epoch": 1.27, "learning_rate": 4.604294619264403e-05, "loss": 1.183, "step": 12628 }, { "epoch": 1.27, "learning_rate": 4.60409239116701e-05, "loss": 0.4364, "step": 12629 }, { "epoch": 1.27, "learning_rate": 4.603890152861783e-05, "loss": 0.6561, "step": 12630 }, { "epoch": 1.27, "learning_rate": 4.603687904350009e-05, "loss": 0.777, "step": 12631 }, { "epoch": 1.27, "learning_rate": 4.603485645632976e-05, "loss": 0.8632, "step": 12632 }, { "epoch": 1.27, "learning_rate": 4.60328337671197e-05, "loss": 0.0198, "step": 12633 }, { "epoch": 1.27, "learning_rate": 4.603081097588277e-05, "loss": 1.3553, "step": 12634 }, { "epoch": 1.27, "learning_rate": 4.602878808263189e-05, "loss": 1.1193, "step": 12635 }, { "epoch": 1.27, "learning_rate": 4.6026765087379874e-05, "loss": 1.195, "step": 12636 }, { "epoch": 1.27, "learning_rate": 4.602474199013963e-05, "loss": 0.7795, "step": 12637 }, { "epoch": 1.27, "learning_rate": 4.602271879092403e-05, "loss": 0.5797, "step": 12638 }, { "epoch": 1.27, "learning_rate": 4.602069548974594e-05, "loss": 1.3643, "step": 12639 }, { "epoch": 1.27, "learning_rate": 4.601867208661824e-05, "loss": 1.2137, "step": 12640 }, { "epoch": 1.27, "learning_rate": 4.6016648581553794e-05, "loss": 1.1531, "step": 12641 }, { "epoch": 1.27, "learning_rate": 4.60146249745655e-05, "loss": 0.8643, "step": 12642 }, { "epoch": 1.27, "learning_rate": 4.601260126566621e-05, "loss": 0.7712, "step": 12643 }, { "epoch": 1.27, "learning_rate": 4.601057745486883e-05, "loss": 0.6166, "step": 12644 }, { "epoch": 1.27, "learning_rate": 4.600855354218623e-05, "loss": 0.5402, "step": 12645 }, { "epoch": 1.27, "learning_rate": 4.600652952763128e-05, "loss": 0.1904, "step": 12646 }, { "epoch": 1.27, "learning_rate": 4.600450541121686e-05, "loss": 0.2984, "step": 12647 }, { "epoch": 1.27, "learning_rate": 4.600248119295587e-05, "loss": 0.7843, "step": 12648 }, { "epoch": 1.27, "learning_rate": 4.600045687286116e-05, "loss": 0.3124, "step": 12649 }, { "epoch": 1.27, "learning_rate": 4.5998432450945634e-05, "loss": 0.2333, "step": 12650 }, { "epoch": 1.27, "learning_rate": 4.599640792722218e-05, "loss": 0.3003, "step": 12651 }, { "epoch": 1.27, "learning_rate": 4.599438330170366e-05, "loss": 0.4781, "step": 12652 }, { "epoch": 1.27, "learning_rate": 4.599235857440297e-05, "loss": 1.0875, "step": 12653 }, { "epoch": 1.27, "learning_rate": 4.5990333745333e-05, "loss": 1.0041, "step": 12654 }, { "epoch": 1.27, "learning_rate": 4.598830881450663e-05, "loss": 0.8234, "step": 12655 }, { "epoch": 1.27, "learning_rate": 4.5986283781936735e-05, "loss": 0.7275, "step": 12656 }, { "epoch": 1.27, "learning_rate": 4.598425864763621e-05, "loss": 0.1692, "step": 12657 }, { "epoch": 1.27, "learning_rate": 4.598223341161796e-05, "loss": 0.3108, "step": 12658 }, { "epoch": 1.27, "learning_rate": 4.598020807389485e-05, "loss": 1.5473, "step": 12659 }, { "epoch": 1.27, "learning_rate": 4.597818263447977e-05, "loss": 0.3518, "step": 12660 }, { "epoch": 1.27, "learning_rate": 4.597615709338562e-05, "loss": 0.9845, "step": 12661 }, { "epoch": 1.27, "learning_rate": 4.597413145062528e-05, "loss": 1.9006, "step": 12662 }, { "epoch": 1.27, "learning_rate": 4.597210570621165e-05, "loss": 0.9076, "step": 12663 }, { "epoch": 1.27, "learning_rate": 4.5970079860157606e-05, "loss": 1.0349, "step": 12664 }, { "epoch": 1.27, "learning_rate": 4.596805391247606e-05, "loss": 0.7438, "step": 12665 }, { "epoch": 1.27, "learning_rate": 4.596602786317988e-05, "loss": 0.203, "step": 12666 }, { "epoch": 1.27, "learning_rate": 4.5964001712281984e-05, "loss": 0.5306, "step": 12667 }, { "epoch": 1.27, "learning_rate": 4.596197545979525e-05, "loss": 0.6434, "step": 12668 }, { "epoch": 1.27, "learning_rate": 4.5959949105732576e-05, "loss": 0.2498, "step": 12669 }, { "epoch": 1.27, "learning_rate": 4.5957922650106856e-05, "loss": 0.4042, "step": 12670 }, { "epoch": 1.27, "learning_rate": 4.5955896092930994e-05, "loss": 0.5931, "step": 12671 }, { "epoch": 1.27, "learning_rate": 4.595386943421787e-05, "loss": 1.3719, "step": 12672 }, { "epoch": 1.27, "learning_rate": 4.5951842673980394e-05, "loss": 0.312, "step": 12673 }, { "epoch": 1.27, "learning_rate": 4.594981581223146e-05, "loss": 0.388, "step": 12674 }, { "epoch": 1.27, "learning_rate": 4.5947788848983964e-05, "loss": 0.4217, "step": 12675 }, { "epoch": 1.27, "learning_rate": 4.594576178425081e-05, "loss": 0.9097, "step": 12676 }, { "epoch": 1.27, "learning_rate": 4.594373461804489e-05, "loss": 0.4955, "step": 12677 }, { "epoch": 1.27, "learning_rate": 4.5941707350379116e-05, "loss": 0.551, "step": 12678 }, { "epoch": 1.27, "learning_rate": 4.593967998126637e-05, "loss": 0.323, "step": 12679 }, { "epoch": 1.27, "learning_rate": 4.593765251071957e-05, "loss": 0.3164, "step": 12680 }, { "epoch": 1.27, "learning_rate": 4.593562493875161e-05, "loss": 0.7734, "step": 12681 }, { "epoch": 1.27, "learning_rate": 4.59335972653754e-05, "loss": 0.8677, "step": 12682 }, { "epoch": 1.27, "learning_rate": 4.5931569490603835e-05, "loss": 0.6156, "step": 12683 }, { "epoch": 1.27, "learning_rate": 4.592954161444982e-05, "loss": 0.3079, "step": 12684 }, { "epoch": 1.28, "learning_rate": 4.592751363692627e-05, "loss": 1.2546, "step": 12685 }, { "epoch": 1.28, "learning_rate": 4.592548555804607e-05, "loss": 0.5919, "step": 12686 }, { "epoch": 1.28, "learning_rate": 4.5923457377822154e-05, "loss": 0.3006, "step": 12687 }, { "epoch": 1.28, "learning_rate": 4.592142909626741e-05, "loss": 0.3515, "step": 12688 }, { "epoch": 1.28, "learning_rate": 4.591940071339473e-05, "loss": 0.718, "step": 12689 }, { "epoch": 1.28, "learning_rate": 4.591737222921706e-05, "loss": 0.4433, "step": 12690 }, { "epoch": 1.28, "learning_rate": 4.591534364374728e-05, "loss": 0.1256, "step": 12691 }, { "epoch": 1.28, "learning_rate": 4.5913314956998306e-05, "loss": 1.4077, "step": 12692 }, { "epoch": 1.28, "learning_rate": 4.591128616898305e-05, "loss": 1.1807, "step": 12693 }, { "epoch": 1.28, "learning_rate": 4.590925727971443e-05, "loss": 0.4224, "step": 12694 }, { "epoch": 1.28, "learning_rate": 4.590722828920534e-05, "loss": 0.4772, "step": 12695 }, { "epoch": 1.28, "learning_rate": 4.59051991974687e-05, "loss": 0.1024, "step": 12696 }, { "epoch": 1.28, "learning_rate": 4.590317000451744e-05, "loss": 0.7379, "step": 12697 }, { "epoch": 1.28, "learning_rate": 4.590114071036443e-05, "loss": 0.1791, "step": 12698 }, { "epoch": 1.28, "learning_rate": 4.589911131502263e-05, "loss": 0.6201, "step": 12699 }, { "epoch": 1.28, "learning_rate": 4.589708181850492e-05, "loss": 0.6948, "step": 12700 }, { "epoch": 1.28, "learning_rate": 4.5895052220824236e-05, "loss": 0.8954, "step": 12701 }, { "epoch": 1.28, "learning_rate": 4.589302252199349e-05, "loss": 1.1371, "step": 12702 }, { "epoch": 1.28, "learning_rate": 4.5890992722025585e-05, "loss": 0.6899, "step": 12703 }, { "epoch": 1.28, "learning_rate": 4.588896282093346e-05, "loss": 1.1085, "step": 12704 }, { "epoch": 1.28, "learning_rate": 4.588693281873e-05, "loss": 0.429, "step": 12705 }, { "epoch": 1.28, "learning_rate": 4.588490271542817e-05, "loss": 0.9631, "step": 12706 }, { "epoch": 1.28, "learning_rate": 4.588287251104083e-05, "loss": 0.8184, "step": 12707 }, { "epoch": 1.28, "learning_rate": 4.588084220558095e-05, "loss": 0.7119, "step": 12708 }, { "epoch": 1.28, "learning_rate": 4.587881179906143e-05, "loss": 0.3373, "step": 12709 }, { "epoch": 1.28, "learning_rate": 4.5876781291495204e-05, "loss": 0.1716, "step": 12710 }, { "epoch": 1.28, "learning_rate": 4.587475068289517e-05, "loss": 0.3633, "step": 12711 }, { "epoch": 1.28, "learning_rate": 4.5872719973274254e-05, "loss": 0.3063, "step": 12712 }, { "epoch": 1.28, "learning_rate": 4.58706891626454e-05, "loss": 0.3232, "step": 12713 }, { "epoch": 1.28, "learning_rate": 4.586865825102151e-05, "loss": 1.0039, "step": 12714 }, { "epoch": 1.28, "learning_rate": 4.586662723841552e-05, "loss": 0.8368, "step": 12715 }, { "epoch": 1.28, "learning_rate": 4.586459612484034e-05, "loss": 1.0522, "step": 12716 }, { "epoch": 1.28, "learning_rate": 4.5862564910308916e-05, "loss": 0.4292, "step": 12717 }, { "epoch": 1.28, "learning_rate": 4.586053359483417e-05, "loss": 0.6662, "step": 12718 }, { "epoch": 1.28, "learning_rate": 4.5858502178429015e-05, "loss": 1.0055, "step": 12719 }, { "epoch": 1.28, "learning_rate": 4.5856470661106385e-05, "loss": 0.4098, "step": 12720 }, { "epoch": 1.28, "learning_rate": 4.58544390428792e-05, "loss": 0.6579, "step": 12721 }, { "epoch": 1.28, "learning_rate": 4.58524073237604e-05, "loss": 0.3625, "step": 12722 }, { "epoch": 1.28, "learning_rate": 4.5850375503762914e-05, "loss": 0.5233, "step": 12723 }, { "epoch": 1.28, "learning_rate": 4.5848343582899665e-05, "loss": 0.2703, "step": 12724 }, { "epoch": 1.28, "learning_rate": 4.5846311561183585e-05, "loss": 0.3592, "step": 12725 }, { "epoch": 1.28, "learning_rate": 4.584427943862761e-05, "loss": 0.6478, "step": 12726 }, { "epoch": 1.28, "learning_rate": 4.584224721524468e-05, "loss": 0.1799, "step": 12727 }, { "epoch": 1.28, "learning_rate": 4.584021489104769e-05, "loss": 0.6039, "step": 12728 }, { "epoch": 1.28, "learning_rate": 4.583818246604962e-05, "loss": 0.3149, "step": 12729 }, { "epoch": 1.28, "learning_rate": 4.583614994026337e-05, "loss": 0.2658, "step": 12730 }, { "epoch": 1.28, "learning_rate": 4.5834117313701886e-05, "loss": 0.4159, "step": 12731 }, { "epoch": 1.28, "learning_rate": 4.58320845863781e-05, "loss": 0.4201, "step": 12732 }, { "epoch": 1.28, "learning_rate": 4.5830051758304965e-05, "loss": 0.5847, "step": 12733 }, { "epoch": 1.28, "learning_rate": 4.582801882949539e-05, "loss": 0.9287, "step": 12734 }, { "epoch": 1.28, "learning_rate": 4.5825985799962334e-05, "loss": 0.8898, "step": 12735 }, { "epoch": 1.28, "learning_rate": 4.5823952669718725e-05, "loss": 0.5008, "step": 12736 }, { "epoch": 1.28, "learning_rate": 4.582191943877748e-05, "loss": 0.9772, "step": 12737 }, { "epoch": 1.28, "learning_rate": 4.5819886107151585e-05, "loss": 0.3423, "step": 12738 }, { "epoch": 1.28, "learning_rate": 4.581785267485394e-05, "loss": 1.0028, "step": 12739 }, { "epoch": 1.28, "learning_rate": 4.58158191418975e-05, "loss": 0.17, "step": 12740 }, { "epoch": 1.28, "learning_rate": 4.5813785508295206e-05, "loss": 0.4051, "step": 12741 }, { "epoch": 1.28, "learning_rate": 4.581175177406e-05, "loss": 0.3832, "step": 12742 }, { "epoch": 1.28, "learning_rate": 4.580971793920482e-05, "loss": 0.5327, "step": 12743 }, { "epoch": 1.28, "learning_rate": 4.580768400374259e-05, "loss": 1.0608, "step": 12744 }, { "epoch": 1.28, "learning_rate": 4.58056499676863e-05, "loss": 0.6412, "step": 12745 }, { "epoch": 1.28, "learning_rate": 4.580361583104885e-05, "loss": 0.4256, "step": 12746 }, { "epoch": 1.28, "learning_rate": 4.5801581593843205e-05, "loss": 0.6289, "step": 12747 }, { "epoch": 1.28, "learning_rate": 4.57995472560823e-05, "loss": 0.578, "step": 12748 }, { "epoch": 1.28, "learning_rate": 4.57975128177791e-05, "loss": 0.6019, "step": 12749 }, { "epoch": 1.28, "learning_rate": 4.579547827894653e-05, "loss": 0.4174, "step": 12750 }, { "epoch": 1.28, "learning_rate": 4.579344363959754e-05, "loss": 0.7719, "step": 12751 }, { "epoch": 1.28, "learning_rate": 4.57914088997451e-05, "loss": 0.4843, "step": 12752 }, { "epoch": 1.28, "learning_rate": 4.5789374059402124e-05, "loss": 0.7893, "step": 12753 }, { "epoch": 1.28, "learning_rate": 4.5787339118581585e-05, "loss": 0.9573, "step": 12754 }, { "epoch": 1.28, "learning_rate": 4.5785304077296426e-05, "loss": 0.5482, "step": 12755 }, { "epoch": 1.28, "learning_rate": 4.57832689355596e-05, "loss": 0.259, "step": 12756 }, { "epoch": 1.28, "learning_rate": 4.578123369338404e-05, "loss": 0.3524, "step": 12757 }, { "epoch": 1.28, "learning_rate": 4.577919835078273e-05, "loss": 0.4491, "step": 12758 }, { "epoch": 1.28, "learning_rate": 4.57771629077686e-05, "loss": 0.4621, "step": 12759 }, { "epoch": 1.28, "learning_rate": 4.57751273643546e-05, "loss": 0.2819, "step": 12760 }, { "epoch": 1.28, "learning_rate": 4.577309172055369e-05, "loss": 0.8869, "step": 12761 }, { "epoch": 1.28, "learning_rate": 4.5771055976378825e-05, "loss": 0.2027, "step": 12762 }, { "epoch": 1.28, "learning_rate": 4.576902013184297e-05, "loss": 0.133, "step": 12763 }, { "epoch": 1.28, "learning_rate": 4.576698418695906e-05, "loss": 1.423, "step": 12764 }, { "epoch": 1.28, "learning_rate": 4.576494814174007e-05, "loss": 0.9428, "step": 12765 }, { "epoch": 1.28, "learning_rate": 4.576291199619894e-05, "loss": 0.5919, "step": 12766 }, { "epoch": 1.28, "learning_rate": 4.576087575034864e-05, "loss": 1.1716, "step": 12767 }, { "epoch": 1.28, "learning_rate": 4.5758839404202116e-05, "loss": 0.6726, "step": 12768 }, { "epoch": 1.28, "learning_rate": 4.5756802957772335e-05, "loss": 0.4972, "step": 12769 }, { "epoch": 1.28, "learning_rate": 4.575476641107226e-05, "loss": 0.7493, "step": 12770 }, { "epoch": 1.28, "learning_rate": 4.575272976411484e-05, "loss": 1.2234, "step": 12771 }, { "epoch": 1.28, "learning_rate": 4.575069301691304e-05, "loss": 1.0675, "step": 12772 }, { "epoch": 1.28, "learning_rate": 4.574865616947982e-05, "loss": 0.8256, "step": 12773 }, { "epoch": 1.28, "learning_rate": 4.5746619221828136e-05, "loss": 0.2329, "step": 12774 }, { "epoch": 1.28, "learning_rate": 4.574458217397098e-05, "loss": 1.3927, "step": 12775 }, { "epoch": 1.28, "learning_rate": 4.5742545025921276e-05, "loss": 1.331, "step": 12776 }, { "epoch": 1.28, "learning_rate": 4.5740507777692e-05, "loss": 0.7246, "step": 12777 }, { "epoch": 1.28, "learning_rate": 4.573847042929613e-05, "loss": 0.3338, "step": 12778 }, { "epoch": 1.28, "learning_rate": 4.573643298074662e-05, "loss": 0.3919, "step": 12779 }, { "epoch": 1.28, "learning_rate": 4.5734395432056436e-05, "loss": 0.4083, "step": 12780 }, { "epoch": 1.28, "learning_rate": 4.5732357783238544e-05, "loss": 1.0444, "step": 12781 }, { "epoch": 1.28, "learning_rate": 4.573032003430591e-05, "loss": 0.7239, "step": 12782 }, { "epoch": 1.29, "learning_rate": 4.572828218527151e-05, "loss": 0.3138, "step": 12783 }, { "epoch": 1.29, "learning_rate": 4.572624423614831e-05, "loss": 0.4166, "step": 12784 }, { "epoch": 1.29, "learning_rate": 4.572420618694926e-05, "loss": 0.2992, "step": 12785 }, { "epoch": 1.29, "learning_rate": 4.5722168037687355e-05, "loss": 0.225, "step": 12786 }, { "epoch": 1.29, "learning_rate": 4.5720129788375545e-05, "loss": 0.9812, "step": 12787 }, { "epoch": 1.29, "learning_rate": 4.571809143902682e-05, "loss": 0.1213, "step": 12788 }, { "epoch": 1.29, "learning_rate": 4.571605298965412e-05, "loss": 1.1832, "step": 12789 }, { "epoch": 1.29, "learning_rate": 4.5714014440270455e-05, "loss": 0.3359, "step": 12790 }, { "epoch": 1.29, "learning_rate": 4.571197579088878e-05, "loss": 1.1413, "step": 12791 }, { "epoch": 1.29, "learning_rate": 4.5709937041522065e-05, "loss": 0.6192, "step": 12792 }, { "epoch": 1.29, "learning_rate": 4.570789819218328e-05, "loss": 0.4288, "step": 12793 }, { "epoch": 1.29, "learning_rate": 4.570585924288541e-05, "loss": 0.2201, "step": 12794 }, { "epoch": 1.29, "learning_rate": 4.5703820193641445e-05, "loss": 0.3449, "step": 12795 }, { "epoch": 1.29, "learning_rate": 4.5701781044464315e-05, "loss": 0.431, "step": 12796 }, { "epoch": 1.29, "learning_rate": 4.5699741795367045e-05, "loss": 0.929, "step": 12797 }, { "epoch": 1.29, "learning_rate": 4.5697702446362565e-05, "loss": 0.1644, "step": 12798 }, { "epoch": 1.29, "learning_rate": 4.56956629974639e-05, "loss": 0.7965, "step": 12799 }, { "epoch": 1.29, "learning_rate": 4.569362344868401e-05, "loss": 1.0896, "step": 12800 }, { "epoch": 1.29, "learning_rate": 4.569158380003586e-05, "loss": 0.2449, "step": 12801 }, { "epoch": 1.29, "learning_rate": 4.568954405153244e-05, "loss": 0.8288, "step": 12802 }, { "epoch": 1.29, "learning_rate": 4.568750420318673e-05, "loss": 0.4254, "step": 12803 }, { "epoch": 1.29, "learning_rate": 4.568546425501171e-05, "loss": 0.3436, "step": 12804 }, { "epoch": 1.29, "learning_rate": 4.5683424207020364e-05, "loss": 1.0045, "step": 12805 }, { "epoch": 1.29, "learning_rate": 4.568138405922568e-05, "loss": 0.8379, "step": 12806 }, { "epoch": 1.29, "learning_rate": 4.567934381164063e-05, "loss": 0.3241, "step": 12807 }, { "epoch": 1.29, "learning_rate": 4.56773034642782e-05, "loss": 0.5656, "step": 12808 }, { "epoch": 1.29, "learning_rate": 4.5675263017151365e-05, "loss": 0.6578, "step": 12809 }, { "epoch": 1.29, "learning_rate": 4.567322247027313e-05, "loss": 0.9221, "step": 12810 }, { "epoch": 1.29, "learning_rate": 4.567118182365646e-05, "loss": 0.7232, "step": 12811 }, { "epoch": 1.29, "learning_rate": 4.566914107731437e-05, "loss": 0.4923, "step": 12812 }, { "epoch": 1.29, "learning_rate": 4.566710023125981e-05, "loss": 0.6546, "step": 12813 }, { "epoch": 1.29, "learning_rate": 4.566505928550578e-05, "loss": 0.6493, "step": 12814 }, { "epoch": 1.29, "learning_rate": 4.566301824006529e-05, "loss": 1.0268, "step": 12815 }, { "epoch": 1.29, "learning_rate": 4.56609770949513e-05, "loss": 0.437, "step": 12816 }, { "epoch": 1.29, "learning_rate": 4.5658935850176815e-05, "loss": 1.2114, "step": 12817 }, { "epoch": 1.29, "learning_rate": 4.565689450575481e-05, "loss": 1.0842, "step": 12818 }, { "epoch": 1.29, "learning_rate": 4.565485306169829e-05, "loss": 1.2849, "step": 12819 }, { "epoch": 1.29, "learning_rate": 4.5652811518020246e-05, "loss": 0.2648, "step": 12820 }, { "epoch": 1.29, "learning_rate": 4.565076987473365e-05, "loss": 0.2856, "step": 12821 }, { "epoch": 1.29, "learning_rate": 4.564872813185152e-05, "loss": 0.6217, "step": 12822 }, { "epoch": 1.29, "learning_rate": 4.5646686289386834e-05, "loss": 0.6685, "step": 12823 }, { "epoch": 1.29, "learning_rate": 4.564464434735259e-05, "loss": 0.5094, "step": 12824 }, { "epoch": 1.29, "learning_rate": 4.564260230576178e-05, "loss": 0.3605, "step": 12825 }, { "epoch": 1.29, "learning_rate": 4.56405601646274e-05, "loss": 0.5003, "step": 12826 }, { "epoch": 1.29, "learning_rate": 4.5638517923962445e-05, "loss": 0.3415, "step": 12827 }, { "epoch": 1.29, "learning_rate": 4.56364755837799e-05, "loss": 0.8725, "step": 12828 }, { "epoch": 1.29, "learning_rate": 4.563443314409278e-05, "loss": 0.1949, "step": 12829 }, { "epoch": 1.29, "learning_rate": 4.563239060491409e-05, "loss": 0.8701, "step": 12830 }, { "epoch": 1.29, "learning_rate": 4.5630347966256796e-05, "loss": 0.3695, "step": 12831 }, { "epoch": 1.29, "learning_rate": 4.5628305228133917e-05, "loss": 0.7546, "step": 12832 }, { "epoch": 1.29, "learning_rate": 4.562626239055845e-05, "loss": 0.7027, "step": 12833 }, { "epoch": 1.29, "learning_rate": 4.56242194535434e-05, "loss": 0.0961, "step": 12834 }, { "epoch": 1.29, "learning_rate": 4.562217641710175e-05, "loss": 0.1958, "step": 12835 }, { "epoch": 1.29, "learning_rate": 4.562013328124652e-05, "loss": 0.2771, "step": 12836 }, { "epoch": 1.29, "learning_rate": 4.5618090045990706e-05, "loss": 0.2084, "step": 12837 }, { "epoch": 1.29, "learning_rate": 4.56160467113473e-05, "loss": 1.188, "step": 12838 }, { "epoch": 1.29, "learning_rate": 4.561400327732932e-05, "loss": 0.275, "step": 12839 }, { "epoch": 1.29, "learning_rate": 4.561195974394976e-05, "loss": 0.4305, "step": 12840 }, { "epoch": 1.29, "learning_rate": 4.560991611122162e-05, "loss": 0.9263, "step": 12841 }, { "epoch": 1.29, "learning_rate": 4.560787237915792e-05, "loss": 0.0968, "step": 12842 }, { "epoch": 1.29, "learning_rate": 4.5605828547771666e-05, "loss": 0.5078, "step": 12843 }, { "epoch": 1.29, "learning_rate": 4.5603784617075836e-05, "loss": 0.9522, "step": 12844 }, { "epoch": 1.29, "learning_rate": 4.560174058708348e-05, "loss": 0.6001, "step": 12845 }, { "epoch": 1.29, "learning_rate": 4.5599696457807565e-05, "loss": 0.5964, "step": 12846 }, { "epoch": 1.29, "learning_rate": 4.559765222926113e-05, "loss": 1.1779, "step": 12847 }, { "epoch": 1.29, "learning_rate": 4.559560790145715e-05, "loss": 0.7634, "step": 12848 }, { "epoch": 1.29, "learning_rate": 4.5593563474408675e-05, "loss": 1.1861, "step": 12849 }, { "epoch": 1.29, "learning_rate": 4.5591518948128675e-05, "loss": 0.8959, "step": 12850 }, { "epoch": 1.29, "learning_rate": 4.55894743226302e-05, "loss": 0.4591, "step": 12851 }, { "epoch": 1.29, "learning_rate": 4.558742959792623e-05, "loss": 0.3483, "step": 12852 }, { "epoch": 1.29, "learning_rate": 4.558538477402978e-05, "loss": 1.1778, "step": 12853 }, { "epoch": 1.29, "learning_rate": 4.5583339850953884e-05, "loss": 0.6512, "step": 12854 }, { "epoch": 1.29, "learning_rate": 4.558129482871154e-05, "loss": 0.8654, "step": 12855 }, { "epoch": 1.29, "learning_rate": 4.557924970731576e-05, "loss": 0.779, "step": 12856 }, { "epoch": 1.29, "learning_rate": 4.557720448677956e-05, "loss": 0.3423, "step": 12857 }, { "epoch": 1.29, "learning_rate": 4.557515916711595e-05, "loss": 0.4597, "step": 12858 }, { "epoch": 1.29, "learning_rate": 4.557311374833796e-05, "loss": 0.451, "step": 12859 }, { "epoch": 1.29, "learning_rate": 4.55710682304586e-05, "loss": 0.5068, "step": 12860 }, { "epoch": 1.29, "learning_rate": 4.556902261349088e-05, "loss": 0.6139, "step": 12861 }, { "epoch": 1.29, "learning_rate": 4.556697689744783e-05, "loss": 0.9492, "step": 12862 }, { "epoch": 1.29, "learning_rate": 4.556493108234246e-05, "loss": 0.3116, "step": 12863 }, { "epoch": 1.29, "learning_rate": 4.556288516818778e-05, "loss": 0.4521, "step": 12864 }, { "epoch": 1.29, "learning_rate": 4.556083915499683e-05, "loss": 0.902, "step": 12865 }, { "epoch": 1.29, "learning_rate": 4.555879304278261e-05, "loss": 0.7213, "step": 12866 }, { "epoch": 1.29, "learning_rate": 4.5556746831558146e-05, "loss": 0.2439, "step": 12867 }, { "epoch": 1.29, "learning_rate": 4.5554700521336486e-05, "loss": 1.1088, "step": 12868 }, { "epoch": 1.29, "learning_rate": 4.555265411213062e-05, "loss": 0.5172, "step": 12869 }, { "epoch": 1.29, "learning_rate": 4.555060760395357e-05, "loss": 0.1704, "step": 12870 }, { "epoch": 1.29, "learning_rate": 4.554856099681837e-05, "loss": 0.2683, "step": 12871 }, { "epoch": 1.29, "learning_rate": 4.554651429073805e-05, "loss": 0.9243, "step": 12872 }, { "epoch": 1.29, "learning_rate": 4.554446748572562e-05, "loss": 0.7547, "step": 12873 }, { "epoch": 1.29, "learning_rate": 4.554242058179412e-05, "loss": 0.7096, "step": 12874 }, { "epoch": 1.29, "learning_rate": 4.554037357895656e-05, "loss": 1.4063, "step": 12875 }, { "epoch": 1.29, "learning_rate": 4.553832647722598e-05, "loss": 0.7886, "step": 12876 }, { "epoch": 1.29, "learning_rate": 4.55362792766154e-05, "loss": 0.8419, "step": 12877 }, { "epoch": 1.29, "learning_rate": 4.553423197713785e-05, "loss": 0.2184, "step": 12878 }, { "epoch": 1.29, "learning_rate": 4.553218457880636e-05, "loss": 0.1358, "step": 12879 }, { "epoch": 1.29, "learning_rate": 4.553013708163395e-05, "loss": 0.7226, "step": 12880 }, { "epoch": 1.29, "learning_rate": 4.552808948563366e-05, "loss": 0.9487, "step": 12881 }, { "epoch": 1.3, "learning_rate": 4.552604179081852e-05, "loss": 0.1392, "step": 12882 }, { "epoch": 1.3, "learning_rate": 4.5523993997201535e-05, "loss": 0.5889, "step": 12883 }, { "epoch": 1.3, "learning_rate": 4.552194610479578e-05, "loss": 0.158, "step": 12884 }, { "epoch": 1.3, "learning_rate": 4.5519898113614256e-05, "loss": 0.2594, "step": 12885 }, { "epoch": 1.3, "learning_rate": 4.5517850023670014e-05, "loss": 0.4862, "step": 12886 }, { "epoch": 1.3, "learning_rate": 4.551580183497607e-05, "loss": 0.6347, "step": 12887 }, { "epoch": 1.3, "learning_rate": 4.551375354754546e-05, "loss": 0.6151, "step": 12888 }, { "epoch": 1.3, "learning_rate": 4.551170516139124e-05, "loss": 0.7642, "step": 12889 }, { "epoch": 1.3, "learning_rate": 4.5509656676526416e-05, "loss": 0.5898, "step": 12890 }, { "epoch": 1.3, "learning_rate": 4.550760809296404e-05, "loss": 0.202, "step": 12891 }, { "epoch": 1.3, "learning_rate": 4.550555941071714e-05, "loss": 0.6974, "step": 12892 }, { "epoch": 1.3, "learning_rate": 4.550351062979877e-05, "loss": 0.3721, "step": 12893 }, { "epoch": 1.3, "learning_rate": 4.5501461750221945e-05, "loss": 0.7646, "step": 12894 }, { "epoch": 1.3, "learning_rate": 4.549941277199972e-05, "loss": 0.4176, "step": 12895 }, { "epoch": 1.3, "learning_rate": 4.5497363695145135e-05, "loss": 0.3032, "step": 12896 }, { "epoch": 1.3, "learning_rate": 4.549531451967121e-05, "loss": 0.2582, "step": 12897 }, { "epoch": 1.3, "learning_rate": 4.5493265245591014e-05, "loss": 0.729, "step": 12898 }, { "epoch": 1.3, "learning_rate": 4.549121587291755e-05, "loss": 0.8039, "step": 12899 }, { "epoch": 1.3, "learning_rate": 4.5489166401663896e-05, "loss": 1.0991, "step": 12900 }, { "epoch": 1.3, "learning_rate": 4.548711683184308e-05, "loss": 0.2608, "step": 12901 }, { "epoch": 1.3, "learning_rate": 4.548506716346814e-05, "loss": 0.822, "step": 12902 }, { "epoch": 1.3, "learning_rate": 4.5483017396552126e-05, "loss": 0.1116, "step": 12903 }, { "epoch": 1.3, "learning_rate": 4.5480967531108086e-05, "loss": 1.4947, "step": 12904 }, { "epoch": 1.3, "learning_rate": 4.547891756714905e-05, "loss": 0.3138, "step": 12905 }, { "epoch": 1.3, "learning_rate": 4.547686750468806e-05, "loss": 0.3623, "step": 12906 }, { "epoch": 1.3, "learning_rate": 4.5474817343738195e-05, "loss": 0.8335, "step": 12907 }, { "epoch": 1.3, "learning_rate": 4.547276708431246e-05, "loss": 0.5991, "step": 12908 }, { "epoch": 1.3, "learning_rate": 4.5470716726423935e-05, "loss": 0.864, "step": 12909 }, { "epoch": 1.3, "learning_rate": 4.5468666270085645e-05, "loss": 0.293, "step": 12910 }, { "epoch": 1.3, "learning_rate": 4.5466615715310654e-05, "loss": 0.1817, "step": 12911 }, { "epoch": 1.3, "learning_rate": 4.546456506211201e-05, "loss": 0.4523, "step": 12912 }, { "epoch": 1.3, "learning_rate": 4.5462514310502735e-05, "loss": 0.3304, "step": 12913 }, { "epoch": 1.3, "learning_rate": 4.546046346049592e-05, "loss": 0.277, "step": 12914 }, { "epoch": 1.3, "learning_rate": 4.5458412512104585e-05, "loss": 0.9161, "step": 12915 }, { "epoch": 1.3, "learning_rate": 4.545636146534181e-05, "loss": 0.8274, "step": 12916 }, { "epoch": 1.3, "learning_rate": 4.545431032022062e-05, "loss": 0.9711, "step": 12917 }, { "epoch": 1.3, "learning_rate": 4.545225907675408e-05, "loss": 0.5358, "step": 12918 }, { "epoch": 1.3, "learning_rate": 4.545020773495524e-05, "loss": 0.6739, "step": 12919 }, { "epoch": 1.3, "learning_rate": 4.5448156294837155e-05, "loss": 0.8847, "step": 12920 }, { "epoch": 1.3, "learning_rate": 4.5446104756412886e-05, "loss": 0.6135, "step": 12921 }, { "epoch": 1.3, "learning_rate": 4.5444053119695465e-05, "loss": 0.7242, "step": 12922 }, { "epoch": 1.3, "learning_rate": 4.544200138469798e-05, "loss": 1.2425, "step": 12923 }, { "epoch": 1.3, "learning_rate": 4.543994955143348e-05, "loss": 0.7829, "step": 12924 }, { "epoch": 1.3, "learning_rate": 4.5437897619914994e-05, "loss": 0.4671, "step": 12925 }, { "epoch": 1.3, "learning_rate": 4.5435845590155616e-05, "loss": 0.131, "step": 12926 }, { "epoch": 1.3, "learning_rate": 4.543379346216838e-05, "loss": 1.2604, "step": 12927 }, { "epoch": 1.3, "learning_rate": 4.5431741235966363e-05, "loss": 0.6289, "step": 12928 }, { "epoch": 1.3, "learning_rate": 4.542968891156261e-05, "loss": 0.9719, "step": 12929 }, { "epoch": 1.3, "learning_rate": 4.542763648897019e-05, "loss": 0.592, "step": 12930 }, { "epoch": 1.3, "learning_rate": 4.542558396820216e-05, "loss": 0.1684, "step": 12931 }, { "epoch": 1.3, "learning_rate": 4.542353134927157e-05, "loss": 0.171, "step": 12932 }, { "epoch": 1.3, "learning_rate": 4.542147863219151e-05, "loss": 0.582, "step": 12933 }, { "epoch": 1.3, "learning_rate": 4.541942581697502e-05, "loss": 0.7958, "step": 12934 }, { "epoch": 1.3, "learning_rate": 4.541737290363517e-05, "loss": 0.3152, "step": 12935 }, { "epoch": 1.3, "learning_rate": 4.5415319892185025e-05, "loss": 0.3319, "step": 12936 }, { "epoch": 1.3, "learning_rate": 4.541326678263764e-05, "loss": 0.2454, "step": 12937 }, { "epoch": 1.3, "learning_rate": 4.54112135750061e-05, "loss": 0.812, "step": 12938 }, { "epoch": 1.3, "learning_rate": 4.540916026930345e-05, "loss": 0.6736, "step": 12939 }, { "epoch": 1.3, "learning_rate": 4.5407106865542776e-05, "loss": 0.6279, "step": 12940 }, { "epoch": 1.3, "learning_rate": 4.5405053363737125e-05, "loss": 0.2532, "step": 12941 }, { "epoch": 1.3, "learning_rate": 4.540299976389959e-05, "loss": 1.1795, "step": 12942 }, { "epoch": 1.3, "learning_rate": 4.5400946066043206e-05, "loss": 0.8097, "step": 12943 }, { "epoch": 1.3, "learning_rate": 4.539889227018107e-05, "loss": 0.6393, "step": 12944 }, { "epoch": 1.3, "learning_rate": 4.539683837632624e-05, "loss": 0.2946, "step": 12945 }, { "epoch": 1.3, "learning_rate": 4.539478438449179e-05, "loss": 0.3038, "step": 12946 }, { "epoch": 1.3, "learning_rate": 4.5392730294690784e-05, "loss": 0.9119, "step": 12947 }, { "epoch": 1.3, "learning_rate": 4.5390676106936304e-05, "loss": 1.1953, "step": 12948 }, { "epoch": 1.3, "learning_rate": 4.538862182124141e-05, "loss": 0.5995, "step": 12949 }, { "epoch": 1.3, "learning_rate": 4.5386567437619186e-05, "loss": 1.007, "step": 12950 }, { "epoch": 1.3, "learning_rate": 4.53845129560827e-05, "loss": 0.994, "step": 12951 }, { "epoch": 1.3, "learning_rate": 4.538245837664501e-05, "loss": 0.2954, "step": 12952 }, { "epoch": 1.3, "learning_rate": 4.538040369931923e-05, "loss": 0.9952, "step": 12953 }, { "epoch": 1.3, "learning_rate": 4.53783489241184e-05, "loss": 0.5974, "step": 12954 }, { "epoch": 1.3, "learning_rate": 4.5376294051055614e-05, "loss": 0.6371, "step": 12955 }, { "epoch": 1.3, "learning_rate": 4.537423908014394e-05, "loss": 0.6899, "step": 12956 }, { "epoch": 1.3, "learning_rate": 4.5372184011396444e-05, "loss": 0.5217, "step": 12957 }, { "epoch": 1.3, "learning_rate": 4.5370128844826234e-05, "loss": 0.3204, "step": 12958 }, { "epoch": 1.3, "learning_rate": 4.5368073580446366e-05, "loss": 0.519, "step": 12959 }, { "epoch": 1.3, "learning_rate": 4.536601821826992e-05, "loss": 0.429, "step": 12960 }, { "epoch": 1.3, "learning_rate": 4.536396275830998e-05, "loss": 0.5514, "step": 12961 }, { "epoch": 1.3, "learning_rate": 4.536190720057964e-05, "loss": 0.5919, "step": 12962 }, { "epoch": 1.3, "learning_rate": 4.5359851545091944e-05, "loss": 0.3984, "step": 12963 }, { "epoch": 1.3, "learning_rate": 4.5357795791860014e-05, "loss": 0.3328, "step": 12964 }, { "epoch": 1.3, "learning_rate": 4.53557399408969e-05, "loss": 0.4497, "step": 12965 }, { "epoch": 1.3, "learning_rate": 4.535368399221571e-05, "loss": 0.6127, "step": 12966 }, { "epoch": 1.3, "learning_rate": 4.535162794582951e-05, "loss": 0.9431, "step": 12967 }, { "epoch": 1.3, "learning_rate": 4.5349571801751383e-05, "loss": 0.9896, "step": 12968 }, { "epoch": 1.3, "learning_rate": 4.534751555999444e-05, "loss": 0.3106, "step": 12969 }, { "epoch": 1.3, "learning_rate": 4.5345459220571726e-05, "loss": 0.8731, "step": 12970 }, { "epoch": 1.3, "learning_rate": 4.5343402783496365e-05, "loss": 0.4166, "step": 12971 }, { "epoch": 1.3, "learning_rate": 4.5341346248781415e-05, "loss": 0.8581, "step": 12972 }, { "epoch": 1.3, "learning_rate": 4.5339289616439975e-05, "loss": 0.614, "step": 12973 }, { "epoch": 1.3, "learning_rate": 4.533723288648513e-05, "loss": 0.6789, "step": 12974 }, { "epoch": 1.3, "learning_rate": 4.533517605892998e-05, "loss": 0.918, "step": 12975 }, { "epoch": 1.3, "learning_rate": 4.5333119133787594e-05, "loss": 0.625, "step": 12976 }, { "epoch": 1.3, "learning_rate": 4.5331062111071064e-05, "loss": 0.5701, "step": 12977 }, { "epoch": 1.3, "learning_rate": 4.53290049907935e-05, "loss": 0.6313, "step": 12978 }, { "epoch": 1.3, "learning_rate": 4.532694777296796e-05, "loss": 0.3338, "step": 12979 }, { "epoch": 1.31, "learning_rate": 4.5324890457607574e-05, "loss": 1.1523, "step": 12980 }, { "epoch": 1.31, "learning_rate": 4.532283304472542e-05, "loss": 0.572, "step": 12981 }, { "epoch": 1.31, "learning_rate": 4.532077553433457e-05, "loss": 0.8388, "step": 12982 }, { "epoch": 1.31, "learning_rate": 4.5318717926448144e-05, "loss": 0.4028, "step": 12983 }, { "epoch": 1.31, "learning_rate": 4.5316660221079214e-05, "loss": 0.4992, "step": 12984 }, { "epoch": 1.31, "learning_rate": 4.5314602418240895e-05, "loss": 0.3727, "step": 12985 }, { "epoch": 1.31, "learning_rate": 4.531254451794626e-05, "loss": 1.033, "step": 12986 }, { "epoch": 1.31, "learning_rate": 4.531048652020843e-05, "loss": 0.6855, "step": 12987 }, { "epoch": 1.31, "learning_rate": 4.530842842504048e-05, "loss": 1.2074, "step": 12988 }, { "epoch": 1.31, "learning_rate": 4.530637023245552e-05, "loss": 0.5564, "step": 12989 }, { "epoch": 1.31, "learning_rate": 4.530431194246664e-05, "loss": 0.4757, "step": 12990 }, { "epoch": 1.31, "learning_rate": 4.5302253555086934e-05, "loss": 0.5408, "step": 12991 }, { "epoch": 1.31, "learning_rate": 4.530019507032952e-05, "loss": 0.2651, "step": 12992 }, { "epoch": 1.31, "learning_rate": 4.529813648820748e-05, "loss": 0.6335, "step": 12993 }, { "epoch": 1.31, "learning_rate": 4.529607780873392e-05, "loss": 0.3219, "step": 12994 }, { "epoch": 1.31, "learning_rate": 4.529401903192194e-05, "loss": 1.3584, "step": 12995 }, { "epoch": 1.31, "learning_rate": 4.5291960157784643e-05, "loss": 0.9409, "step": 12996 }, { "epoch": 1.31, "learning_rate": 4.528990118633512e-05, "loss": 0.6931, "step": 12997 }, { "epoch": 1.31, "learning_rate": 4.5287842117586495e-05, "loss": 0.4976, "step": 12998 }, { "epoch": 1.31, "learning_rate": 4.528578295155185e-05, "loss": 1.2496, "step": 12999 }, { "epoch": 1.31, "learning_rate": 4.528372368824429e-05, "loss": 0.3269, "step": 13000 }, { "epoch": 1.31, "learning_rate": 4.5281664327676944e-05, "loss": 0.3557, "step": 13001 }, { "epoch": 1.31, "learning_rate": 4.5279604869862894e-05, "loss": 0.7521, "step": 13002 }, { "epoch": 1.31, "learning_rate": 4.527754531481524e-05, "loss": 1.4475, "step": 13003 }, { "epoch": 1.31, "learning_rate": 4.5275485662547114e-05, "loss": 0.4807, "step": 13004 }, { "epoch": 1.31, "learning_rate": 4.52734259130716e-05, "loss": 1.0526, "step": 13005 }, { "epoch": 1.31, "learning_rate": 4.527136606640182e-05, "loss": 0.4804, "step": 13006 }, { "epoch": 1.31, "learning_rate": 4.5269306122550875e-05, "loss": 0.945, "step": 13007 }, { "epoch": 1.31, "learning_rate": 4.526724608153187e-05, "loss": 0.6089, "step": 13008 }, { "epoch": 1.31, "learning_rate": 4.5265185943357914e-05, "loss": 0.4925, "step": 13009 }, { "epoch": 1.31, "learning_rate": 4.526312570804213e-05, "loss": 0.9312, "step": 13010 }, { "epoch": 1.31, "learning_rate": 4.526106537559762e-05, "loss": 0.5705, "step": 13011 }, { "epoch": 1.31, "learning_rate": 4.52590049460375e-05, "loss": 0.4833, "step": 13012 }, { "epoch": 1.31, "learning_rate": 4.525694441937487e-05, "loss": 0.8405, "step": 13013 }, { "epoch": 1.31, "learning_rate": 4.525488379562285e-05, "loss": 0.9125, "step": 13014 }, { "epoch": 1.31, "learning_rate": 4.525282307479456e-05, "loss": 0.594, "step": 13015 }, { "epoch": 1.31, "learning_rate": 4.5250762256903094e-05, "loss": 0.3915, "step": 13016 }, { "epoch": 1.31, "learning_rate": 4.524870134196159e-05, "loss": 0.3635, "step": 13017 }, { "epoch": 1.31, "learning_rate": 4.524664032998314e-05, "loss": 0.7109, "step": 13018 }, { "epoch": 1.31, "learning_rate": 4.524457922098089e-05, "loss": 0.9005, "step": 13019 }, { "epoch": 1.31, "learning_rate": 4.524251801496792e-05, "loss": 0.6621, "step": 13020 }, { "epoch": 1.31, "learning_rate": 4.524045671195737e-05, "loss": 0.285, "step": 13021 }, { "epoch": 1.31, "learning_rate": 4.523839531196235e-05, "loss": 0.8486, "step": 13022 }, { "epoch": 1.31, "learning_rate": 4.5236333814995975e-05, "loss": 0.8653, "step": 13023 }, { "epoch": 1.31, "learning_rate": 4.523427222107138e-05, "loss": 0.4305, "step": 13024 }, { "epoch": 1.31, "learning_rate": 4.523221053020166e-05, "loss": 0.8407, "step": 13025 }, { "epoch": 1.31, "learning_rate": 4.523014874239995e-05, "loss": 0.5626, "step": 13026 }, { "epoch": 1.31, "learning_rate": 4.522808685767937e-05, "loss": 0.6842, "step": 13027 }, { "epoch": 1.31, "learning_rate": 4.522602487605303e-05, "loss": 0.3773, "step": 13028 }, { "epoch": 1.31, "learning_rate": 4.522396279753408e-05, "loss": 0.5624, "step": 13029 }, { "epoch": 1.31, "learning_rate": 4.52219006221356e-05, "loss": 0.5274, "step": 13030 }, { "epoch": 1.31, "learning_rate": 4.521983834987075e-05, "loss": 0.1453, "step": 13031 }, { "epoch": 1.31, "learning_rate": 4.521777598075263e-05, "loss": 1.1078, "step": 13032 }, { "epoch": 1.31, "learning_rate": 4.5215713514794374e-05, "loss": 0.4528, "step": 13033 }, { "epoch": 1.31, "learning_rate": 4.521365095200911e-05, "loss": 0.1323, "step": 13034 }, { "epoch": 1.31, "learning_rate": 4.5211588292409956e-05, "loss": 0.3612, "step": 13035 }, { "epoch": 1.31, "learning_rate": 4.520952553601004e-05, "loss": 0.695, "step": 13036 }, { "epoch": 1.31, "learning_rate": 4.520746268282249e-05, "loss": 0.7936, "step": 13037 }, { "epoch": 1.31, "learning_rate": 4.520539973286044e-05, "loss": 0.7984, "step": 13038 }, { "epoch": 1.31, "learning_rate": 4.520333668613699e-05, "loss": 0.5071, "step": 13039 }, { "epoch": 1.31, "learning_rate": 4.520127354266531e-05, "loss": 0.4279, "step": 13040 }, { "epoch": 1.31, "learning_rate": 4.51992103024585e-05, "loss": 0.8395, "step": 13041 }, { "epoch": 1.31, "learning_rate": 4.519714696552969e-05, "loss": 0.4684, "step": 13042 }, { "epoch": 1.31, "learning_rate": 4.519508353189203e-05, "loss": 0.6116, "step": 13043 }, { "epoch": 1.31, "learning_rate": 4.5193020001558636e-05, "loss": 1.0894, "step": 13044 }, { "epoch": 1.31, "learning_rate": 4.519095637454265e-05, "loss": 0.4683, "step": 13045 }, { "epoch": 1.31, "learning_rate": 4.518889265085718e-05, "loss": 1.163, "step": 13046 }, { "epoch": 1.31, "learning_rate": 4.518682883051539e-05, "loss": 0.2141, "step": 13047 }, { "epoch": 1.31, "learning_rate": 4.518476491353039e-05, "loss": 1.2583, "step": 13048 }, { "epoch": 1.31, "learning_rate": 4.5182700899915325e-05, "loss": 1.1633, "step": 13049 }, { "epoch": 1.31, "learning_rate": 4.5180636789683325e-05, "loss": 0.5671, "step": 13050 }, { "epoch": 1.31, "learning_rate": 4.517857258284753e-05, "loss": 1.1533, "step": 13051 }, { "epoch": 1.31, "learning_rate": 4.517650827942107e-05, "loss": 0.9524, "step": 13052 }, { "epoch": 1.31, "learning_rate": 4.51744438794171e-05, "loss": 0.6373, "step": 13053 }, { "epoch": 1.31, "learning_rate": 4.5172379382848725e-05, "loss": 1.0508, "step": 13054 }, { "epoch": 1.31, "learning_rate": 4.51703147897291e-05, "loss": 0.7367, "step": 13055 }, { "epoch": 1.31, "learning_rate": 4.516825010007137e-05, "loss": 0.461, "step": 13056 }, { "epoch": 1.31, "learning_rate": 4.516618531388866e-05, "loss": 0.5332, "step": 13057 }, { "epoch": 1.31, "learning_rate": 4.516412043119412e-05, "loss": 0.2763, "step": 13058 }, { "epoch": 1.31, "learning_rate": 4.5162055452000887e-05, "loss": 0.3094, "step": 13059 }, { "epoch": 1.31, "learning_rate": 4.51599903763221e-05, "loss": 0.3937, "step": 13060 }, { "epoch": 1.31, "learning_rate": 4.515792520417091e-05, "loss": 0.2051, "step": 13061 }, { "epoch": 1.31, "learning_rate": 4.5155859935560433e-05, "loss": 0.725, "step": 13062 }, { "epoch": 1.31, "learning_rate": 4.5153794570503834e-05, "loss": 0.8852, "step": 13063 }, { "epoch": 1.31, "learning_rate": 4.515172910901426e-05, "loss": 0.265, "step": 13064 }, { "epoch": 1.31, "learning_rate": 4.514966355110484e-05, "loss": 1.0533, "step": 13065 }, { "epoch": 1.31, "learning_rate": 4.5147597896788726e-05, "loss": 0.8633, "step": 13066 }, { "epoch": 1.31, "learning_rate": 4.514553214607906e-05, "loss": 0.2481, "step": 13067 }, { "epoch": 1.31, "learning_rate": 4.514346629898899e-05, "loss": 0.3609, "step": 13068 }, { "epoch": 1.31, "learning_rate": 4.5141400355531666e-05, "loss": 1.0186, "step": 13069 }, { "epoch": 1.31, "learning_rate": 4.513933431572022e-05, "loss": 0.4395, "step": 13070 }, { "epoch": 1.31, "learning_rate": 4.513726817956781e-05, "loss": 0.5121, "step": 13071 }, { "epoch": 1.31, "learning_rate": 4.513520194708759e-05, "loss": 0.7523, "step": 13072 }, { "epoch": 1.31, "learning_rate": 4.513313561829271e-05, "loss": 0.2365, "step": 13073 }, { "epoch": 1.31, "learning_rate": 4.51310691931963e-05, "loss": 0.8084, "step": 13074 }, { "epoch": 1.31, "learning_rate": 4.5129002671811524e-05, "loss": 1.1863, "step": 13075 }, { "epoch": 1.31, "learning_rate": 4.512693605415153e-05, "loss": 0.3631, "step": 13076 }, { "epoch": 1.31, "learning_rate": 4.512486934022947e-05, "loss": 0.488, "step": 13077 }, { "epoch": 1.31, "learning_rate": 4.5122802530058493e-05, "loss": 0.9657, "step": 13078 }, { "epoch": 1.32, "learning_rate": 4.512073562365177e-05, "loss": 0.2158, "step": 13079 }, { "epoch": 1.32, "learning_rate": 4.511866862102242e-05, "loss": 1.1125, "step": 13080 }, { "epoch": 1.32, "learning_rate": 4.511660152218361e-05, "loss": 0.9628, "step": 13081 }, { "epoch": 1.32, "learning_rate": 4.5114534327148515e-05, "loss": 1.8686, "step": 13082 }, { "epoch": 1.32, "learning_rate": 4.5112467035930265e-05, "loss": 0.7257, "step": 13083 }, { "epoch": 1.32, "learning_rate": 4.511039964854202e-05, "loss": 0.3202, "step": 13084 }, { "epoch": 1.32, "learning_rate": 4.5108332164996953e-05, "loss": 0.2443, "step": 13085 }, { "epoch": 1.32, "learning_rate": 4.5106264585308197e-05, "loss": 0.7288, "step": 13086 }, { "epoch": 1.32, "learning_rate": 4.510419690948892e-05, "loss": 0.5089, "step": 13087 }, { "epoch": 1.32, "learning_rate": 4.5102129137552285e-05, "loss": 0.9868, "step": 13088 }, { "epoch": 1.32, "learning_rate": 4.510006126951144e-05, "loss": 0.9406, "step": 13089 }, { "epoch": 1.32, "learning_rate": 4.509799330537955e-05, "loss": 0.2875, "step": 13090 }, { "epoch": 1.32, "learning_rate": 4.5095925245169775e-05, "loss": 0.932, "step": 13091 }, { "epoch": 1.32, "learning_rate": 4.509385708889528e-05, "loss": 0.3471, "step": 13092 }, { "epoch": 1.32, "learning_rate": 4.509178883656922e-05, "loss": 0.4848, "step": 13093 }, { "epoch": 1.32, "learning_rate": 4.508972048820475e-05, "loss": 0.4932, "step": 13094 }, { "epoch": 1.32, "learning_rate": 4.508765204381504e-05, "loss": 1.0606, "step": 13095 }, { "epoch": 1.32, "learning_rate": 4.508558350341325e-05, "loss": 0.2197, "step": 13096 }, { "epoch": 1.32, "learning_rate": 4.508351486701256e-05, "loss": 0.696, "step": 13097 }, { "epoch": 1.32, "learning_rate": 4.5081446134626106e-05, "loss": 1.2457, "step": 13098 }, { "epoch": 1.32, "learning_rate": 4.507937730626707e-05, "loss": 0.6086, "step": 13099 }, { "epoch": 1.32, "learning_rate": 4.507730838194861e-05, "loss": 0.1831, "step": 13100 }, { "epoch": 1.32, "learning_rate": 4.5075239361683894e-05, "loss": 0.3298, "step": 13101 }, { "epoch": 1.32, "learning_rate": 4.50731702454861e-05, "loss": 0.3175, "step": 13102 }, { "epoch": 1.32, "learning_rate": 4.5071101033368374e-05, "loss": 1.1991, "step": 13103 }, { "epoch": 1.32, "learning_rate": 4.50690317253439e-05, "loss": 0.13, "step": 13104 }, { "epoch": 1.32, "learning_rate": 4.506696232142583e-05, "loss": 0.4904, "step": 13105 }, { "epoch": 1.32, "learning_rate": 4.5064892821627354e-05, "loss": 0.2732, "step": 13106 }, { "epoch": 1.32, "learning_rate": 4.506282322596163e-05, "loss": 0.6703, "step": 13107 }, { "epoch": 1.32, "learning_rate": 4.506075353444184e-05, "loss": 1.0453, "step": 13108 }, { "epoch": 1.32, "learning_rate": 4.505868374708114e-05, "loss": 0.5418, "step": 13109 }, { "epoch": 1.32, "learning_rate": 4.5056613863892685e-05, "loss": 0.7157, "step": 13110 }, { "epoch": 1.32, "learning_rate": 4.505454388488968e-05, "loss": 0.3299, "step": 13111 }, { "epoch": 1.32, "learning_rate": 4.5052473810085284e-05, "loss": 0.3466, "step": 13112 }, { "epoch": 1.32, "learning_rate": 4.505040363949268e-05, "loss": 1.2507, "step": 13113 }, { "epoch": 1.32, "learning_rate": 4.5048333373125024e-05, "loss": 0.421, "step": 13114 }, { "epoch": 1.32, "learning_rate": 4.50462630109955e-05, "loss": 0.9316, "step": 13115 }, { "epoch": 1.32, "learning_rate": 4.5044192553117274e-05, "loss": 0.7912, "step": 13116 }, { "epoch": 1.32, "learning_rate": 4.504212199950354e-05, "loss": 0.1287, "step": 13117 }, { "epoch": 1.32, "learning_rate": 4.504005135016746e-05, "loss": 0.5946, "step": 13118 }, { "epoch": 1.32, "learning_rate": 4.503798060512221e-05, "loss": 0.6563, "step": 13119 }, { "epoch": 1.32, "learning_rate": 4.503590976438098e-05, "loss": 0.987, "step": 13120 }, { "epoch": 1.32, "learning_rate": 4.503383882795694e-05, "loss": 0.3746, "step": 13121 }, { "epoch": 1.32, "learning_rate": 4.5031767795863275e-05, "loss": 0.8074, "step": 13122 }, { "epoch": 1.32, "learning_rate": 4.502969666811315e-05, "loss": 0.3438, "step": 13123 }, { "epoch": 1.32, "learning_rate": 4.502762544471975e-05, "loss": 1.0275, "step": 13124 }, { "epoch": 1.32, "learning_rate": 4.5025554125696266e-05, "loss": 0.9996, "step": 13125 }, { "epoch": 1.32, "learning_rate": 4.5023482711055866e-05, "loss": 0.44, "step": 13126 }, { "epoch": 1.32, "learning_rate": 4.502141120081174e-05, "loss": 0.9028, "step": 13127 }, { "epoch": 1.32, "learning_rate": 4.501933959497706e-05, "loss": 0.4014, "step": 13128 }, { "epoch": 1.32, "learning_rate": 4.501726789356503e-05, "loss": 0.5303, "step": 13129 }, { "epoch": 1.32, "learning_rate": 4.501519609658881e-05, "loss": 1.0278, "step": 13130 }, { "epoch": 1.32, "learning_rate": 4.5013124204061604e-05, "loss": 0.4013, "step": 13131 }, { "epoch": 1.32, "learning_rate": 4.5011052215996575e-05, "loss": 0.7918, "step": 13132 }, { "epoch": 1.32, "learning_rate": 4.500898013240692e-05, "loss": 0.4404, "step": 13133 }, { "epoch": 1.32, "learning_rate": 4.500690795330583e-05, "loss": 0.7088, "step": 13134 }, { "epoch": 1.32, "learning_rate": 4.500483567870649e-05, "loss": 0.3557, "step": 13135 }, { "epoch": 1.32, "learning_rate": 4.500276330862208e-05, "loss": 0.617, "step": 13136 }, { "epoch": 1.32, "learning_rate": 4.5000690843065794e-05, "loss": 0.4869, "step": 13137 }, { "epoch": 1.32, "learning_rate": 4.499861828205081e-05, "loss": 0.4695, "step": 13138 }, { "epoch": 1.32, "learning_rate": 4.499654562559034e-05, "loss": 0.6072, "step": 13139 }, { "epoch": 1.32, "learning_rate": 4.499447287369754e-05, "loss": 0.408, "step": 13140 }, { "epoch": 1.32, "learning_rate": 4.499240002638564e-05, "loss": 0.8957, "step": 13141 }, { "epoch": 1.32, "learning_rate": 4.4990327083667796e-05, "loss": 0.2642, "step": 13142 }, { "epoch": 1.32, "learning_rate": 4.4988254045557205e-05, "loss": 1.0117, "step": 13143 }, { "epoch": 1.32, "learning_rate": 4.4986180912067085e-05, "loss": 0.7815, "step": 13144 }, { "epoch": 1.32, "learning_rate": 4.49841076832106e-05, "loss": 0.2594, "step": 13145 }, { "epoch": 1.32, "learning_rate": 4.498203435900096e-05, "loss": 0.4488, "step": 13146 }, { "epoch": 1.32, "learning_rate": 4.4979960939451354e-05, "loss": 1.6906, "step": 13147 }, { "epoch": 1.32, "learning_rate": 4.4977887424574966e-05, "loss": 1.0401, "step": 13148 }, { "epoch": 1.32, "learning_rate": 4.4975813814385004e-05, "loss": 1.4298, "step": 13149 }, { "epoch": 1.32, "learning_rate": 4.4973740108894666e-05, "loss": 0.4597, "step": 13150 }, { "epoch": 1.32, "learning_rate": 4.4971666308117134e-05, "loss": 0.9732, "step": 13151 }, { "epoch": 1.32, "learning_rate": 4.496959241206562e-05, "loss": 0.321, "step": 13152 }, { "epoch": 1.32, "learning_rate": 4.496751842075331e-05, "loss": 0.583, "step": 13153 }, { "epoch": 1.32, "learning_rate": 4.496544433419341e-05, "loss": 0.1315, "step": 13154 }, { "epoch": 1.32, "learning_rate": 4.496337015239912e-05, "loss": 0.3164, "step": 13155 }, { "epoch": 1.32, "learning_rate": 4.496129587538362e-05, "loss": 0.4237, "step": 13156 }, { "epoch": 1.32, "learning_rate": 4.495922150316014e-05, "loss": 1.139, "step": 13157 }, { "epoch": 1.32, "learning_rate": 4.495714703574185e-05, "loss": 0.4852, "step": 13158 }, { "epoch": 1.32, "learning_rate": 4.4955072473141975e-05, "loss": 0.8326, "step": 13159 }, { "epoch": 1.32, "learning_rate": 4.495299781537371e-05, "loss": 1.1713, "step": 13160 }, { "epoch": 1.32, "learning_rate": 4.4950923062450255e-05, "loss": 0.5057, "step": 13161 }, { "epoch": 1.32, "learning_rate": 4.494884821438481e-05, "loss": 0.22, "step": 13162 }, { "epoch": 1.32, "learning_rate": 4.494677327119058e-05, "loss": 0.3897, "step": 13163 }, { "epoch": 1.32, "learning_rate": 4.494469823288078e-05, "loss": 0.5931, "step": 13164 }, { "epoch": 1.32, "learning_rate": 4.494262309946859e-05, "loss": 0.7297, "step": 13165 }, { "epoch": 1.32, "learning_rate": 4.494054787096725e-05, "loss": 0.78, "step": 13166 }, { "epoch": 1.32, "learning_rate": 4.493847254738993e-05, "loss": 0.3132, "step": 13167 }, { "epoch": 1.32, "learning_rate": 4.493639712874986e-05, "loss": 1.2582, "step": 13168 }, { "epoch": 1.32, "learning_rate": 4.493432161506025e-05, "loss": 1.1678, "step": 13169 }, { "epoch": 1.32, "learning_rate": 4.4932246006334284e-05, "loss": 0.3158, "step": 13170 }, { "epoch": 1.32, "learning_rate": 4.493017030258519e-05, "loss": 0.5787, "step": 13171 }, { "epoch": 1.32, "learning_rate": 4.492809450382617e-05, "loss": 0.9616, "step": 13172 }, { "epoch": 1.32, "learning_rate": 4.4926018610070445e-05, "loss": 0.8144, "step": 13173 }, { "epoch": 1.32, "learning_rate": 4.4923942621331207e-05, "loss": 1.7505, "step": 13174 }, { "epoch": 1.32, "learning_rate": 4.492186653762167e-05, "loss": 0.77, "step": 13175 }, { "epoch": 1.32, "learning_rate": 4.491979035895507e-05, "loss": 1.0107, "step": 13176 }, { "epoch": 1.33, "learning_rate": 4.4917714085344584e-05, "loss": 0.8721, "step": 13177 }, { "epoch": 1.33, "learning_rate": 4.491563771680345e-05, "loss": 1.4431, "step": 13178 }, { "epoch": 1.33, "learning_rate": 4.491356125334487e-05, "loss": 0.4823, "step": 13179 }, { "epoch": 1.33, "learning_rate": 4.491148469498206e-05, "loss": 0.666, "step": 13180 }, { "epoch": 1.33, "learning_rate": 4.490940804172823e-05, "loss": 0.5962, "step": 13181 }, { "epoch": 1.33, "learning_rate": 4.490733129359659e-05, "loss": 0.5362, "step": 13182 }, { "epoch": 1.33, "learning_rate": 4.490525445060037e-05, "loss": 0.1564, "step": 13183 }, { "epoch": 1.33, "learning_rate": 4.4903177512752795e-05, "loss": 0.1735, "step": 13184 }, { "epoch": 1.33, "learning_rate": 4.490110048006706e-05, "loss": 0.3902, "step": 13185 }, { "epoch": 1.33, "learning_rate": 4.4899023352556385e-05, "loss": 0.5735, "step": 13186 }, { "epoch": 1.33, "learning_rate": 4.4896946130234e-05, "loss": 0.1406, "step": 13187 }, { "epoch": 1.33, "learning_rate": 4.489486881311311e-05, "loss": 1.0394, "step": 13188 }, { "epoch": 1.33, "learning_rate": 4.489279140120695e-05, "loss": 0.7163, "step": 13189 }, { "epoch": 1.33, "learning_rate": 4.4890713894528716e-05, "loss": 0.502, "step": 13190 }, { "epoch": 1.33, "learning_rate": 4.4888636293091656e-05, "loss": 0.6974, "step": 13191 }, { "epoch": 1.33, "learning_rate": 4.488655859690897e-05, "loss": 0.3978, "step": 13192 }, { "epoch": 1.33, "learning_rate": 4.48844808059939e-05, "loss": 0.8948, "step": 13193 }, { "epoch": 1.33, "learning_rate": 4.488240292035965e-05, "loss": 0.5324, "step": 13194 }, { "epoch": 1.33, "learning_rate": 4.488032494001945e-05, "loss": 0.2719, "step": 13195 }, { "epoch": 1.33, "learning_rate": 4.487824686498653e-05, "loss": 0.4075, "step": 13196 }, { "epoch": 1.33, "learning_rate": 4.4876168695274094e-05, "loss": 0.9584, "step": 13197 }, { "epoch": 1.33, "learning_rate": 4.487409043089539e-05, "loss": 0.2274, "step": 13198 }, { "epoch": 1.33, "learning_rate": 4.487201207186362e-05, "loss": 0.677, "step": 13199 }, { "epoch": 1.33, "learning_rate": 4.4869933618192034e-05, "loss": 0.591, "step": 13200 }, { "epoch": 1.33, "learning_rate": 4.486785506989385e-05, "loss": 0.2658, "step": 13201 }, { "epoch": 1.33, "learning_rate": 4.486577642698228e-05, "loss": 0.8945, "step": 13202 }, { "epoch": 1.33, "learning_rate": 4.4863697689470576e-05, "loss": 0.6176, "step": 13203 }, { "epoch": 1.33, "learning_rate": 4.486161885737195e-05, "loss": 0.6654, "step": 13204 }, { "epoch": 1.33, "learning_rate": 4.4859539930699636e-05, "loss": 1.3011, "step": 13205 }, { "epoch": 1.33, "learning_rate": 4.485746090946686e-05, "loss": 0.578, "step": 13206 }, { "epoch": 1.33, "learning_rate": 4.485538179368685e-05, "loss": 0.7937, "step": 13207 }, { "epoch": 1.33, "learning_rate": 4.4853302583372846e-05, "loss": 0.6594, "step": 13208 }, { "epoch": 1.33, "learning_rate": 4.4851223278538075e-05, "loss": 0.8547, "step": 13209 }, { "epoch": 1.33, "learning_rate": 4.484914387919577e-05, "loss": 1.2342, "step": 13210 }, { "epoch": 1.33, "learning_rate": 4.4847064385359165e-05, "loss": 0.6085, "step": 13211 }, { "epoch": 1.33, "learning_rate": 4.4844984797041494e-05, "loss": 0.5613, "step": 13212 }, { "epoch": 1.33, "learning_rate": 4.484290511425597e-05, "loss": 0.3312, "step": 13213 }, { "epoch": 1.33, "learning_rate": 4.484082533701586e-05, "loss": 0.541, "step": 13214 }, { "epoch": 1.33, "learning_rate": 4.483874546533437e-05, "loss": 0.4651, "step": 13215 }, { "epoch": 1.33, "learning_rate": 4.483666549922476e-05, "loss": 0.1729, "step": 13216 }, { "epoch": 1.33, "learning_rate": 4.4834585438700256e-05, "loss": 1.5223, "step": 13217 }, { "epoch": 1.33, "learning_rate": 4.48325052837741e-05, "loss": 0.3963, "step": 13218 }, { "epoch": 1.33, "learning_rate": 4.48304250344595e-05, "loss": 0.8076, "step": 13219 }, { "epoch": 1.33, "learning_rate": 4.482834469076973e-05, "loss": 0.5837, "step": 13220 }, { "epoch": 1.33, "learning_rate": 4.4826264252718016e-05, "loss": 0.1445, "step": 13221 }, { "epoch": 1.33, "learning_rate": 4.4824183720317595e-05, "loss": 0.4546, "step": 13222 }, { "epoch": 1.33, "learning_rate": 4.4822103093581705e-05, "loss": 0.6757, "step": 13223 }, { "epoch": 1.33, "learning_rate": 4.48200223725236e-05, "loss": 0.6876, "step": 13224 }, { "epoch": 1.33, "learning_rate": 4.48179415571565e-05, "loss": 0.7546, "step": 13225 }, { "epoch": 1.33, "learning_rate": 4.481586064749366e-05, "loss": 0.5182, "step": 13226 }, { "epoch": 1.33, "learning_rate": 4.4813779643548315e-05, "loss": 1.0976, "step": 13227 }, { "epoch": 1.33, "learning_rate": 4.481169854533373e-05, "loss": 0.7202, "step": 13228 }, { "epoch": 1.33, "learning_rate": 4.48096173528631e-05, "loss": 0.221, "step": 13229 }, { "epoch": 1.33, "learning_rate": 4.480753606614973e-05, "loss": 0.651, "step": 13230 }, { "epoch": 1.33, "learning_rate": 4.480545468520681e-05, "loss": 0.3533, "step": 13231 }, { "epoch": 1.33, "learning_rate": 4.480337321004762e-05, "loss": 0.6038, "step": 13232 }, { "epoch": 1.33, "learning_rate": 4.480129164068539e-05, "loss": 0.7829, "step": 13233 }, { "epoch": 1.33, "learning_rate": 4.479920997713338e-05, "loss": 1.3303, "step": 13234 }, { "epoch": 1.33, "learning_rate": 4.479712821940482e-05, "loss": 0.5854, "step": 13235 }, { "epoch": 1.33, "learning_rate": 4.479504636751296e-05, "loss": 0.4267, "step": 13236 }, { "epoch": 1.33, "learning_rate": 4.479296442147106e-05, "loss": 1.1521, "step": 13237 }, { "epoch": 1.33, "learning_rate": 4.4790882381292365e-05, "loss": 0.6193, "step": 13238 }, { "epoch": 1.33, "learning_rate": 4.478880024699011e-05, "loss": 0.6375, "step": 13239 }, { "epoch": 1.33, "learning_rate": 4.4786718018577566e-05, "loss": 0.8333, "step": 13240 }, { "epoch": 1.33, "learning_rate": 4.478463569606797e-05, "loss": 0.8239, "step": 13241 }, { "epoch": 1.33, "learning_rate": 4.478255327947458e-05, "loss": 0.358, "step": 13242 }, { "epoch": 1.33, "learning_rate": 4.478047076881064e-05, "loss": 0.2882, "step": 13243 }, { "epoch": 1.33, "learning_rate": 4.4778388164089417e-05, "loss": 0.2565, "step": 13244 }, { "epoch": 1.33, "learning_rate": 4.4776305465324143e-05, "loss": 0.4366, "step": 13245 }, { "epoch": 1.33, "learning_rate": 4.477422267252808e-05, "loss": 0.7619, "step": 13246 }, { "epoch": 1.33, "learning_rate": 4.477213978571449e-05, "loss": 0.8468, "step": 13247 }, { "epoch": 1.33, "learning_rate": 4.477005680489662e-05, "loss": 0.5466, "step": 13248 }, { "epoch": 1.33, "learning_rate": 4.4767973730087725e-05, "loss": 0.8583, "step": 13249 }, { "epoch": 1.33, "learning_rate": 4.476589056130107e-05, "loss": 0.7653, "step": 13250 }, { "epoch": 1.33, "learning_rate": 4.47638072985499e-05, "loss": 0.4191, "step": 13251 }, { "epoch": 1.33, "learning_rate": 4.4761723941847476e-05, "loss": 0.9, "step": 13252 }, { "epoch": 1.33, "learning_rate": 4.4759640491207064e-05, "loss": 0.3157, "step": 13253 }, { "epoch": 1.33, "learning_rate": 4.475755694664191e-05, "loss": 0.3379, "step": 13254 }, { "epoch": 1.33, "learning_rate": 4.475547330816528e-05, "loss": 0.8741, "step": 13255 }, { "epoch": 1.33, "learning_rate": 4.475338957579042e-05, "loss": 0.6395, "step": 13256 }, { "epoch": 1.33, "learning_rate": 4.475130574953062e-05, "loss": 0.3294, "step": 13257 }, { "epoch": 1.33, "learning_rate": 4.474922182939911e-05, "loss": 0.3828, "step": 13258 }, { "epoch": 1.33, "learning_rate": 4.474713781540916e-05, "loss": 0.5459, "step": 13259 }, { "epoch": 1.33, "learning_rate": 4.4745053707574046e-05, "loss": 0.5278, "step": 13260 }, { "epoch": 1.33, "learning_rate": 4.4742969505907006e-05, "loss": 0.8327, "step": 13261 }, { "epoch": 1.33, "learning_rate": 4.474088521042133e-05, "loss": 0.5192, "step": 13262 }, { "epoch": 1.33, "learning_rate": 4.473880082113026e-05, "loss": 1.0943, "step": 13263 }, { "epoch": 1.33, "learning_rate": 4.4736716338047073e-05, "loss": 0.9227, "step": 13264 }, { "epoch": 1.33, "learning_rate": 4.473463176118502e-05, "loss": 0.3009, "step": 13265 }, { "epoch": 1.33, "learning_rate": 4.473254709055738e-05, "loss": 1.252, "step": 13266 }, { "epoch": 1.33, "learning_rate": 4.4730462326177423e-05, "loss": 0.249, "step": 13267 }, { "epoch": 1.33, "learning_rate": 4.47283774680584e-05, "loss": 0.2409, "step": 13268 }, { "epoch": 1.33, "learning_rate": 4.472629251621359e-05, "loss": 0.676, "step": 13269 }, { "epoch": 1.33, "learning_rate": 4.4724207470656256e-05, "loss": 0.5756, "step": 13270 }, { "epoch": 1.33, "learning_rate": 4.472212233139966e-05, "loss": 0.5733, "step": 13271 }, { "epoch": 1.33, "learning_rate": 4.4720037098457084e-05, "loss": 0.4182, "step": 13272 }, { "epoch": 1.33, "learning_rate": 4.471795177184179e-05, "loss": 1.019, "step": 13273 }, { "epoch": 1.33, "learning_rate": 4.471586635156706e-05, "loss": 0.5209, "step": 13274 }, { "epoch": 1.33, "learning_rate": 4.471378083764613e-05, "loss": 0.3295, "step": 13275 }, { "epoch": 1.34, "learning_rate": 4.471169523009232e-05, "loss": 1.1596, "step": 13276 }, { "epoch": 1.34, "learning_rate": 4.470960952891887e-05, "loss": 0.8418, "step": 13277 }, { "epoch": 1.34, "learning_rate": 4.4707523734139056e-05, "loss": 0.392, "step": 13278 }, { "epoch": 1.34, "learning_rate": 4.470543784576616e-05, "loss": 0.9639, "step": 13279 }, { "epoch": 1.34, "learning_rate": 4.470335186381346e-05, "loss": 0.1277, "step": 13280 }, { "epoch": 1.34, "learning_rate": 4.470126578829422e-05, "loss": 0.98, "step": 13281 }, { "epoch": 1.34, "learning_rate": 4.4699179619221705e-05, "loss": 0.6852, "step": 13282 }, { "epoch": 1.34, "learning_rate": 4.4697093356609216e-05, "loss": 0.292, "step": 13283 }, { "epoch": 1.34, "learning_rate": 4.4695007000470006e-05, "loss": 0.2366, "step": 13284 }, { "epoch": 1.34, "learning_rate": 4.469292055081736e-05, "loss": 0.8322, "step": 13285 }, { "epoch": 1.34, "learning_rate": 4.4690834007664566e-05, "loss": 0.5012, "step": 13286 }, { "epoch": 1.34, "learning_rate": 4.4688747371024886e-05, "loss": 0.4265, "step": 13287 }, { "epoch": 1.34, "learning_rate": 4.468666064091161e-05, "loss": 0.2263, "step": 13288 }, { "epoch": 1.34, "learning_rate": 4.4684573817338014e-05, "loss": 0.1745, "step": 13289 }, { "epoch": 1.34, "learning_rate": 4.4682486900317384e-05, "loss": 0.673, "step": 13290 }, { "epoch": 1.34, "learning_rate": 4.4680399889862976e-05, "loss": 0.5801, "step": 13291 }, { "epoch": 1.34, "learning_rate": 4.467831278598809e-05, "loss": 0.6404, "step": 13292 }, { "epoch": 1.34, "learning_rate": 4.467622558870601e-05, "loss": 0.17, "step": 13293 }, { "epoch": 1.34, "learning_rate": 4.467413829803002e-05, "loss": 0.2784, "step": 13294 }, { "epoch": 1.34, "learning_rate": 4.467205091397339e-05, "loss": 0.6248, "step": 13295 }, { "epoch": 1.34, "learning_rate": 4.4669963436549414e-05, "loss": 1.4162, "step": 13296 }, { "epoch": 1.34, "learning_rate": 4.466787586577136e-05, "loss": 0.8068, "step": 13297 }, { "epoch": 1.34, "learning_rate": 4.466578820165253e-05, "loss": 1.2289, "step": 13298 }, { "epoch": 1.34, "learning_rate": 4.4663700444206206e-05, "loss": 0.2599, "step": 13299 }, { "epoch": 1.34, "learning_rate": 4.466161259344567e-05, "loss": 1.0818, "step": 13300 }, { "epoch": 1.34, "learning_rate": 4.4659524649384206e-05, "loss": 1.351, "step": 13301 }, { "epoch": 1.34, "learning_rate": 4.46574366120351e-05, "loss": 0.9211, "step": 13302 }, { "epoch": 1.34, "learning_rate": 4.4655348481411654e-05, "loss": 0.9136, "step": 13303 }, { "epoch": 1.34, "learning_rate": 4.465326025752714e-05, "loss": 1.1214, "step": 13304 }, { "epoch": 1.34, "learning_rate": 4.465117194039485e-05, "loss": 0.6615, "step": 13305 }, { "epoch": 1.34, "learning_rate": 4.464908353002809e-05, "loss": 0.9613, "step": 13306 }, { "epoch": 1.34, "learning_rate": 4.4646995026440114e-05, "loss": 1.1381, "step": 13307 }, { "epoch": 1.34, "learning_rate": 4.464490642964424e-05, "loss": 0.9668, "step": 13308 }, { "epoch": 1.34, "learning_rate": 4.464281773965376e-05, "loss": 1.1161, "step": 13309 }, { "epoch": 1.34, "learning_rate": 4.464072895648195e-05, "loss": 0.4966, "step": 13310 }, { "epoch": 1.34, "learning_rate": 4.4638640080142116e-05, "loss": 0.3462, "step": 13311 }, { "epoch": 1.34, "learning_rate": 4.4636551110647546e-05, "loss": 0.2514, "step": 13312 }, { "epoch": 1.34, "learning_rate": 4.463446204801153e-05, "loss": 1.0384, "step": 13313 }, { "epoch": 1.34, "learning_rate": 4.463237289224736e-05, "loss": 0.4399, "step": 13314 }, { "epoch": 1.34, "learning_rate": 4.463028364336835e-05, "loss": 0.2226, "step": 13315 }, { "epoch": 1.34, "learning_rate": 4.462819430138777e-05, "loss": 0.937, "step": 13316 }, { "epoch": 1.34, "learning_rate": 4.4626104866318926e-05, "loss": 0.8401, "step": 13317 }, { "epoch": 1.34, "learning_rate": 4.462401533817512e-05, "loss": 0.2235, "step": 13318 }, { "epoch": 1.34, "learning_rate": 4.462192571696964e-05, "loss": 0.9099, "step": 13319 }, { "epoch": 1.34, "learning_rate": 4.4619836002715785e-05, "loss": 0.4876, "step": 13320 }, { "epoch": 1.34, "learning_rate": 4.461774619542685e-05, "loss": 0.9551, "step": 13321 }, { "epoch": 1.34, "learning_rate": 4.461565629511616e-05, "loss": 0.3488, "step": 13322 }, { "epoch": 1.34, "learning_rate": 4.4613566301796975e-05, "loss": 0.7119, "step": 13323 }, { "epoch": 1.34, "learning_rate": 4.461147621548261e-05, "loss": 1.1406, "step": 13324 }, { "epoch": 1.34, "learning_rate": 4.460938603618638e-05, "loss": 0.9321, "step": 13325 }, { "epoch": 1.34, "learning_rate": 4.4607295763921575e-05, "loss": 0.148, "step": 13326 }, { "epoch": 1.34, "learning_rate": 4.460520539870149e-05, "loss": 0.2544, "step": 13327 }, { "epoch": 1.34, "learning_rate": 4.460311494053945e-05, "loss": 0.7266, "step": 13328 }, { "epoch": 1.34, "learning_rate": 4.460102438944873e-05, "loss": 1.1484, "step": 13329 }, { "epoch": 1.34, "learning_rate": 4.459893374544263e-05, "loss": 0.8308, "step": 13330 }, { "epoch": 1.34, "learning_rate": 4.4596843008534484e-05, "loss": 0.2895, "step": 13331 }, { "epoch": 1.34, "learning_rate": 4.459475217873759e-05, "loss": 0.9489, "step": 13332 }, { "epoch": 1.34, "learning_rate": 4.4592661256065233e-05, "loss": 0.573, "step": 13333 }, { "epoch": 1.34, "learning_rate": 4.459057024053074e-05, "loss": 0.6888, "step": 13334 }, { "epoch": 1.34, "learning_rate": 4.4588479132147403e-05, "loss": 0.3712, "step": 13335 }, { "epoch": 1.34, "learning_rate": 4.458638793092853e-05, "loss": 0.8158, "step": 13336 }, { "epoch": 1.34, "learning_rate": 4.458429663688744e-05, "loss": 1.1597, "step": 13337 }, { "epoch": 1.34, "learning_rate": 4.458220525003744e-05, "loss": 1.5889, "step": 13338 }, { "epoch": 1.34, "learning_rate": 4.458011377039182e-05, "loss": 0.5248, "step": 13339 }, { "epoch": 1.34, "learning_rate": 4.4578022197963905e-05, "loss": 0.6546, "step": 13340 }, { "epoch": 1.34, "learning_rate": 4.457593053276701e-05, "loss": 0.2852, "step": 13341 }, { "epoch": 1.34, "learning_rate": 4.457383877481443e-05, "loss": 0.7727, "step": 13342 }, { "epoch": 1.34, "learning_rate": 4.4571746924119496e-05, "loss": 0.216, "step": 13343 }, { "epoch": 1.34, "learning_rate": 4.4569654980695494e-05, "loss": 0.178, "step": 13344 }, { "epoch": 1.34, "learning_rate": 4.4567562944555767e-05, "loss": 0.8828, "step": 13345 }, { "epoch": 1.34, "learning_rate": 4.456547081571359e-05, "loss": 0.2003, "step": 13346 }, { "epoch": 1.34, "learning_rate": 4.456337859418231e-05, "loss": 0.3144, "step": 13347 }, { "epoch": 1.34, "learning_rate": 4.4561286279975226e-05, "loss": 0.165, "step": 13348 }, { "epoch": 1.34, "learning_rate": 4.4559193873105664e-05, "loss": 0.8636, "step": 13349 }, { "epoch": 1.34, "learning_rate": 4.455710137358692e-05, "loss": 1.0548, "step": 13350 }, { "epoch": 1.34, "learning_rate": 4.455500878143233e-05, "loss": 0.9607, "step": 13351 }, { "epoch": 1.34, "learning_rate": 4.4552916096655196e-05, "loss": 1.0466, "step": 13352 }, { "epoch": 1.34, "learning_rate": 4.455082331926883e-05, "loss": 1.2744, "step": 13353 }, { "epoch": 1.34, "learning_rate": 4.454873044928658e-05, "loss": 0.3408, "step": 13354 }, { "epoch": 1.34, "learning_rate": 4.4546637486721726e-05, "loss": 1.1121, "step": 13355 }, { "epoch": 1.34, "learning_rate": 4.4544544431587614e-05, "loss": 0.822, "step": 13356 }, { "epoch": 1.34, "learning_rate": 4.4542451283897555e-05, "loss": 1.2062, "step": 13357 }, { "epoch": 1.34, "learning_rate": 4.454035804366487e-05, "loss": 1.3806, "step": 13358 }, { "epoch": 1.34, "learning_rate": 4.453826471090287e-05, "loss": 0.3143, "step": 13359 }, { "epoch": 1.34, "learning_rate": 4.453617128562488e-05, "loss": 0.7443, "step": 13360 }, { "epoch": 1.34, "learning_rate": 4.4534077767844245e-05, "loss": 1.3636, "step": 13361 }, { "epoch": 1.34, "learning_rate": 4.453198415757425e-05, "loss": 0.9544, "step": 13362 }, { "epoch": 1.34, "learning_rate": 4.452989045482825e-05, "loss": 0.4563, "step": 13363 }, { "epoch": 1.34, "learning_rate": 4.452779665961955e-05, "loss": 0.3828, "step": 13364 }, { "epoch": 1.34, "learning_rate": 4.452570277196147e-05, "loss": 0.3359, "step": 13365 }, { "epoch": 1.34, "learning_rate": 4.452360879186735e-05, "loss": 1.1155, "step": 13366 }, { "epoch": 1.34, "learning_rate": 4.452151471935051e-05, "loss": 0.6354, "step": 13367 }, { "epoch": 1.34, "learning_rate": 4.451942055442427e-05, "loss": 0.6103, "step": 13368 }, { "epoch": 1.34, "learning_rate": 4.451732629710196e-05, "loss": 0.2391, "step": 13369 }, { "epoch": 1.34, "learning_rate": 4.451523194739691e-05, "loss": 0.6012, "step": 13370 }, { "epoch": 1.34, "learning_rate": 4.4513137505322455e-05, "loss": 0.293, "step": 13371 }, { "epoch": 1.34, "learning_rate": 4.451104297089191e-05, "loss": 0.7374, "step": 13372 }, { "epoch": 1.34, "learning_rate": 4.4508948344118595e-05, "loss": 0.7003, "step": 13373 }, { "epoch": 1.35, "learning_rate": 4.450685362501587e-05, "loss": 0.25, "step": 13374 }, { "epoch": 1.35, "learning_rate": 4.450475881359704e-05, "loss": 0.646, "step": 13375 }, { "epoch": 1.35, "learning_rate": 4.4502663909875435e-05, "loss": 0.613, "step": 13376 }, { "epoch": 1.35, "learning_rate": 4.450056891386441e-05, "loss": 0.335, "step": 13377 }, { "epoch": 1.35, "learning_rate": 4.4498473825577264e-05, "loss": 1.0539, "step": 13378 }, { "epoch": 1.35, "learning_rate": 4.4496378645027356e-05, "loss": 0.9007, "step": 13379 }, { "epoch": 1.35, "learning_rate": 4.4494283372228004e-05, "loss": 0.5829, "step": 13380 }, { "epoch": 1.35, "learning_rate": 4.449218800719255e-05, "loss": 0.5259, "step": 13381 }, { "epoch": 1.35, "learning_rate": 4.4490092549934326e-05, "loss": 0.3738, "step": 13382 }, { "epoch": 1.35, "learning_rate": 4.4487997000466664e-05, "loss": 0.1883, "step": 13383 }, { "epoch": 1.35, "learning_rate": 4.44859013588029e-05, "loss": 0.6879, "step": 13384 }, { "epoch": 1.35, "learning_rate": 4.448380562495636e-05, "loss": 0.5935, "step": 13385 }, { "epoch": 1.35, "learning_rate": 4.448170979894041e-05, "loss": 0.498, "step": 13386 }, { "epoch": 1.35, "learning_rate": 4.4479613880768354e-05, "loss": 0.6158, "step": 13387 }, { "epoch": 1.35, "learning_rate": 4.447751787045355e-05, "loss": 0.6911, "step": 13388 }, { "epoch": 1.35, "learning_rate": 4.447542176800933e-05, "loss": 0.2114, "step": 13389 }, { "epoch": 1.35, "learning_rate": 4.447332557344903e-05, "loss": 0.4908, "step": 13390 }, { "epoch": 1.35, "learning_rate": 4.4471229286786e-05, "loss": 0.8707, "step": 13391 }, { "epoch": 1.35, "learning_rate": 4.446913290803356e-05, "loss": 0.1986, "step": 13392 }, { "epoch": 1.35, "learning_rate": 4.446703643720507e-05, "loss": 0.4089, "step": 13393 }, { "epoch": 1.35, "learning_rate": 4.4464939874313854e-05, "loss": 0.9488, "step": 13394 }, { "epoch": 1.35, "learning_rate": 4.446284321937328e-05, "loss": 0.3328, "step": 13395 }, { "epoch": 1.35, "learning_rate": 4.446074647239666e-05, "loss": 0.4864, "step": 13396 }, { "epoch": 1.35, "learning_rate": 4.4458649633397356e-05, "loss": 0.5735, "step": 13397 }, { "epoch": 1.35, "learning_rate": 4.44565527023887e-05, "loss": 0.8847, "step": 13398 }, { "epoch": 1.35, "learning_rate": 4.445445567938405e-05, "loss": 0.7507, "step": 13399 }, { "epoch": 1.35, "learning_rate": 4.445235856439674e-05, "loss": 0.527, "step": 13400 }, { "epoch": 1.35, "learning_rate": 4.445026135744012e-05, "loss": 1.4418, "step": 13401 }, { "epoch": 1.35, "learning_rate": 4.444816405852753e-05, "loss": 0.2273, "step": 13402 }, { "epoch": 1.35, "learning_rate": 4.4446066667672326e-05, "loss": 0.3255, "step": 13403 }, { "epoch": 1.35, "learning_rate": 4.444396918488784e-05, "loss": 1.1153, "step": 13404 }, { "epoch": 1.35, "learning_rate": 4.4441871610187444e-05, "loss": 0.3229, "step": 13405 }, { "epoch": 1.35, "learning_rate": 4.443977394358446e-05, "loss": 0.3681, "step": 13406 }, { "epoch": 1.35, "learning_rate": 4.443767618509224e-05, "loss": 0.1478, "step": 13407 }, { "epoch": 1.35, "learning_rate": 4.4435578334724156e-05, "loss": 0.9359, "step": 13408 }, { "epoch": 1.35, "learning_rate": 4.443348039249353e-05, "loss": 1.213, "step": 13409 }, { "epoch": 1.35, "learning_rate": 4.443138235841373e-05, "loss": 0.4142, "step": 13410 }, { "epoch": 1.35, "learning_rate": 4.442928423249812e-05, "loss": 0.1748, "step": 13411 }, { "epoch": 1.35, "learning_rate": 4.4427186014760005e-05, "loss": 0.7462, "step": 13412 }, { "epoch": 1.35, "learning_rate": 4.442508770521279e-05, "loss": 0.349, "step": 13413 }, { "epoch": 1.35, "learning_rate": 4.44229893038698e-05, "loss": 0.7596, "step": 13414 }, { "epoch": 1.35, "learning_rate": 4.442089081074438e-05, "loss": 1.0028, "step": 13415 }, { "epoch": 1.35, "learning_rate": 4.441879222584991e-05, "loss": 0.5316, "step": 13416 }, { "epoch": 1.35, "learning_rate": 4.441669354919972e-05, "loss": 0.2835, "step": 13417 }, { "epoch": 1.35, "learning_rate": 4.441459478080719e-05, "loss": 0.8724, "step": 13418 }, { "epoch": 1.35, "learning_rate": 4.4412495920685655e-05, "loss": 1.0293, "step": 13419 }, { "epoch": 1.35, "learning_rate": 4.441039696884848e-05, "loss": 0.218, "step": 13420 }, { "epoch": 1.35, "learning_rate": 4.440829792530902e-05, "loss": 0.5614, "step": 13421 }, { "epoch": 1.35, "learning_rate": 4.440619879008064e-05, "loss": 0.3402, "step": 13422 }, { "epoch": 1.35, "learning_rate": 4.4404099563176683e-05, "loss": 0.5021, "step": 13423 }, { "epoch": 1.35, "learning_rate": 4.440200024461051e-05, "loss": 0.4429, "step": 13424 }, { "epoch": 1.35, "learning_rate": 4.43999008343955e-05, "loss": 0.5231, "step": 13425 }, { "epoch": 1.35, "learning_rate": 4.439780133254499e-05, "loss": 0.2344, "step": 13426 }, { "epoch": 1.35, "learning_rate": 4.439570173907236e-05, "loss": 0.31, "step": 13427 }, { "epoch": 1.35, "learning_rate": 4.439360205399095e-05, "loss": 0.3076, "step": 13428 }, { "epoch": 1.35, "learning_rate": 4.439150227731414e-05, "loss": 0.6324, "step": 13429 }, { "epoch": 1.35, "learning_rate": 4.438940240905528e-05, "loss": 0.3738, "step": 13430 }, { "epoch": 1.35, "learning_rate": 4.438730244922773e-05, "loss": 0.1002, "step": 13431 }, { "epoch": 1.35, "learning_rate": 4.438520239784487e-05, "loss": 0.8316, "step": 13432 }, { "epoch": 1.35, "learning_rate": 4.4383102254920045e-05, "loss": 0.9708, "step": 13433 }, { "epoch": 1.35, "learning_rate": 4.438100202046665e-05, "loss": 0.8771, "step": 13434 }, { "epoch": 1.35, "learning_rate": 4.4378901694498006e-05, "loss": 1.1333, "step": 13435 }, { "epoch": 1.35, "learning_rate": 4.4376801277027514e-05, "loss": 0.6531, "step": 13436 }, { "epoch": 1.35, "learning_rate": 4.4374700768068525e-05, "loss": 0.6375, "step": 13437 }, { "epoch": 1.35, "learning_rate": 4.43726001676344e-05, "loss": 0.7414, "step": 13438 }, { "epoch": 1.35, "learning_rate": 4.437049947573853e-05, "loss": 1.2365, "step": 13439 }, { "epoch": 1.35, "learning_rate": 4.436839869239426e-05, "loss": 0.4672, "step": 13440 }, { "epoch": 1.35, "learning_rate": 4.4366297817614965e-05, "loss": 0.294, "step": 13441 }, { "epoch": 1.35, "learning_rate": 4.436419685141402e-05, "loss": 0.1698, "step": 13442 }, { "epoch": 1.35, "learning_rate": 4.436209579380479e-05, "loss": 1.0589, "step": 13443 }, { "epoch": 1.35, "learning_rate": 4.435999464480064e-05, "loss": 0.5283, "step": 13444 }, { "epoch": 1.35, "learning_rate": 4.435789340441496e-05, "loss": 0.5997, "step": 13445 }, { "epoch": 1.35, "learning_rate": 4.43557920726611e-05, "loss": 1.002, "step": 13446 }, { "epoch": 1.35, "learning_rate": 4.435369064955244e-05, "loss": 0.7707, "step": 13447 }, { "epoch": 1.35, "learning_rate": 4.435158913510235e-05, "loss": 0.7542, "step": 13448 }, { "epoch": 1.35, "learning_rate": 4.4349487529324215e-05, "loss": 0.9737, "step": 13449 }, { "epoch": 1.35, "learning_rate": 4.43473858322314e-05, "loss": 0.531, "step": 13450 }, { "epoch": 1.35, "learning_rate": 4.434528404383727e-05, "loss": 0.2933, "step": 13451 }, { "epoch": 1.35, "learning_rate": 4.434318216415521e-05, "loss": 0.3399, "step": 13452 }, { "epoch": 1.35, "learning_rate": 4.4341080193198607e-05, "loss": 0.2283, "step": 13453 }, { "epoch": 1.35, "learning_rate": 4.433897813098082e-05, "loss": 0.6094, "step": 13454 }, { "epoch": 1.35, "learning_rate": 4.433687597751523e-05, "loss": 0.4744, "step": 13455 }, { "epoch": 1.35, "learning_rate": 4.433477373281521e-05, "loss": 0.6473, "step": 13456 }, { "epoch": 1.35, "learning_rate": 4.433267139689415e-05, "loss": 0.2578, "step": 13457 }, { "epoch": 1.35, "learning_rate": 4.433056896976542e-05, "loss": 0.5901, "step": 13458 }, { "epoch": 1.35, "learning_rate": 4.43284664514424e-05, "loss": 0.4215, "step": 13459 }, { "epoch": 1.35, "learning_rate": 4.4326363841938466e-05, "loss": 1.1085, "step": 13460 }, { "epoch": 1.35, "learning_rate": 4.4324261141267015e-05, "loss": 1.1753, "step": 13461 }, { "epoch": 1.35, "learning_rate": 4.432215834944141e-05, "loss": 0.5856, "step": 13462 }, { "epoch": 1.35, "learning_rate": 4.432005546647503e-05, "loss": 0.959, "step": 13463 }, { "epoch": 1.35, "learning_rate": 4.4317952492381276e-05, "loss": 0.4691, "step": 13464 }, { "epoch": 1.35, "learning_rate": 4.431584942717351e-05, "loss": 0.1883, "step": 13465 }, { "epoch": 1.35, "learning_rate": 4.431374627086512e-05, "loss": 0.3367, "step": 13466 }, { "epoch": 1.35, "learning_rate": 4.43116430234695e-05, "loss": 0.1901, "step": 13467 }, { "epoch": 1.35, "learning_rate": 4.430953968500003e-05, "loss": 0.5637, "step": 13468 }, { "epoch": 1.35, "learning_rate": 4.43074362554701e-05, "loss": 0.6253, "step": 13469 }, { "epoch": 1.35, "learning_rate": 4.4305332734893075e-05, "loss": 0.4099, "step": 13470 }, { "epoch": 1.35, "learning_rate": 4.4303229123282365e-05, "loss": 0.5989, "step": 13471 }, { "epoch": 1.35, "learning_rate": 4.430112542065134e-05, "loss": 0.9405, "step": 13472 }, { "epoch": 1.36, "learning_rate": 4.429902162701339e-05, "loss": 0.638, "step": 13473 }, { "epoch": 1.36, "learning_rate": 4.429691774238192e-05, "loss": 0.3729, "step": 13474 }, { "epoch": 1.36, "learning_rate": 4.429481376677029e-05, "loss": 0.8166, "step": 13475 }, { "epoch": 1.36, "learning_rate": 4.429270970019191e-05, "loss": 1.2089, "step": 13476 }, { "epoch": 1.36, "learning_rate": 4.429060554266016e-05, "loss": 0.3759, "step": 13477 }, { "epoch": 1.36, "learning_rate": 4.428850129418844e-05, "loss": 0.5588, "step": 13478 }, { "epoch": 1.36, "learning_rate": 4.428639695479012e-05, "loss": 0.3049, "step": 13479 }, { "epoch": 1.36, "learning_rate": 4.428429252447861e-05, "loss": 1.3491, "step": 13480 }, { "epoch": 1.36, "learning_rate": 4.42821880032673e-05, "loss": 1.3291, "step": 13481 }, { "epoch": 1.36, "learning_rate": 4.4280083391169576e-05, "loss": 0.9249, "step": 13482 }, { "epoch": 1.36, "learning_rate": 4.427797868819883e-05, "loss": 0.1225, "step": 13483 }, { "epoch": 1.36, "learning_rate": 4.427587389436847e-05, "loss": 0.448, "step": 13484 }, { "epoch": 1.36, "learning_rate": 4.427376900969188e-05, "loss": 0.7959, "step": 13485 }, { "epoch": 1.36, "learning_rate": 4.4271664034182435e-05, "loss": 0.4864, "step": 13486 }, { "epoch": 1.36, "learning_rate": 4.426955896785357e-05, "loss": 0.3501, "step": 13487 }, { "epoch": 1.36, "learning_rate": 4.426745381071865e-05, "loss": 1.2972, "step": 13488 }, { "epoch": 1.36, "learning_rate": 4.426534856279108e-05, "loss": 0.4868, "step": 13489 }, { "epoch": 1.36, "learning_rate": 4.426324322408427e-05, "loss": 0.9756, "step": 13490 }, { "epoch": 1.36, "learning_rate": 4.42611377946116e-05, "loss": 0.641, "step": 13491 }, { "epoch": 1.36, "learning_rate": 4.425903227438647e-05, "loss": 0.2512, "step": 13492 }, { "epoch": 1.36, "learning_rate": 4.425692666342228e-05, "loss": 1.0528, "step": 13493 }, { "epoch": 1.36, "learning_rate": 4.425482096173244e-05, "loss": 0.3109, "step": 13494 }, { "epoch": 1.36, "learning_rate": 4.4252715169330335e-05, "loss": 0.5609, "step": 13495 }, { "epoch": 1.36, "learning_rate": 4.425060928622938e-05, "loss": 0.743, "step": 13496 }, { "epoch": 1.36, "learning_rate": 4.4248503312442966e-05, "loss": 0.8852, "step": 13497 }, { "epoch": 1.36, "learning_rate": 4.4246397247984497e-05, "loss": 1.1042, "step": 13498 }, { "epoch": 1.36, "learning_rate": 4.424429109286737e-05, "loss": 0.5404, "step": 13499 }, { "epoch": 1.36, "learning_rate": 4.424218484710502e-05, "loss": 0.1153, "step": 13500 }, { "epoch": 1.36, "learning_rate": 4.4240078510710794e-05, "loss": 0.1254, "step": 13501 }, { "epoch": 1.36, "learning_rate": 4.4237972083698127e-05, "loss": 0.3357, "step": 13502 }, { "epoch": 1.36, "learning_rate": 4.423586556608044e-05, "loss": 1.0923, "step": 13503 }, { "epoch": 1.36, "learning_rate": 4.42337589578711e-05, "loss": 0.6054, "step": 13504 }, { "epoch": 1.36, "learning_rate": 4.423165225908354e-05, "loss": 0.4306, "step": 13505 }, { "epoch": 1.36, "learning_rate": 4.4229545469731164e-05, "loss": 0.4841, "step": 13506 }, { "epoch": 1.36, "learning_rate": 4.4227438589827375e-05, "loss": 0.2392, "step": 13507 }, { "epoch": 1.36, "learning_rate": 4.4225331619385575e-05, "loss": 1.0198, "step": 13508 }, { "epoch": 1.36, "learning_rate": 4.422322455841918e-05, "loss": 0.5707, "step": 13509 }, { "epoch": 1.36, "learning_rate": 4.422111740694159e-05, "loss": 0.6072, "step": 13510 }, { "epoch": 1.36, "learning_rate": 4.421901016496622e-05, "loss": 1.311, "step": 13511 }, { "epoch": 1.36, "learning_rate": 4.421690283250648e-05, "loss": 0.799, "step": 13512 }, { "epoch": 1.36, "learning_rate": 4.421479540957578e-05, "loss": 0.1647, "step": 13513 }, { "epoch": 1.36, "learning_rate": 4.4212687896187525e-05, "loss": 0.7601, "step": 13514 }, { "epoch": 1.36, "learning_rate": 4.421058029235513e-05, "loss": 0.7691, "step": 13515 }, { "epoch": 1.36, "learning_rate": 4.4208472598092016e-05, "loss": 0.3763, "step": 13516 }, { "epoch": 1.36, "learning_rate": 4.420636481341158e-05, "loss": 0.6967, "step": 13517 }, { "epoch": 1.36, "learning_rate": 4.420425693832724e-05, "loss": 0.2653, "step": 13518 }, { "epoch": 1.36, "learning_rate": 4.420214897285242e-05, "loss": 0.6271, "step": 13519 }, { "epoch": 1.36, "learning_rate": 4.420004091700053e-05, "loss": 1.1053, "step": 13520 }, { "epoch": 1.36, "learning_rate": 4.419793277078497e-05, "loss": 0.5108, "step": 13521 }, { "epoch": 1.36, "learning_rate": 4.4195824534219174e-05, "loss": 0.4156, "step": 13522 }, { "epoch": 1.36, "learning_rate": 4.419371620731656e-05, "loss": 0.8518, "step": 13523 }, { "epoch": 1.36, "learning_rate": 4.419160779009051e-05, "loss": 0.6929, "step": 13524 }, { "epoch": 1.36, "learning_rate": 4.418949928255449e-05, "loss": 0.7558, "step": 13525 }, { "epoch": 1.36, "learning_rate": 4.41873906847219e-05, "loss": 0.8233, "step": 13526 }, { "epoch": 1.36, "learning_rate": 4.418528199660613e-05, "loss": 2.1217, "step": 13527 }, { "epoch": 1.36, "learning_rate": 4.418317321822063e-05, "loss": 0.5128, "step": 13528 }, { "epoch": 1.36, "learning_rate": 4.4181064349578816e-05, "loss": 0.1274, "step": 13529 }, { "epoch": 1.36, "learning_rate": 4.4178955390694105e-05, "loss": 0.9499, "step": 13530 }, { "epoch": 1.36, "learning_rate": 4.41768463415799e-05, "loss": 0.6624, "step": 13531 }, { "epoch": 1.36, "learning_rate": 4.417473720224966e-05, "loss": 0.8624, "step": 13532 }, { "epoch": 1.36, "learning_rate": 4.417262797271677e-05, "loss": 0.7374, "step": 13533 }, { "epoch": 1.36, "learning_rate": 4.417051865299466e-05, "loss": 0.6436, "step": 13534 }, { "epoch": 1.36, "learning_rate": 4.4168409243096775e-05, "loss": 1.4983, "step": 13535 }, { "epoch": 1.36, "learning_rate": 4.416629974303653e-05, "loss": 0.697, "step": 13536 }, { "epoch": 1.36, "learning_rate": 4.416419015282732e-05, "loss": 1.0769, "step": 13537 }, { "epoch": 1.36, "learning_rate": 4.41620804724826e-05, "loss": 0.9873, "step": 13538 }, { "epoch": 1.36, "learning_rate": 4.41599707020158e-05, "loss": 0.6421, "step": 13539 }, { "epoch": 1.36, "learning_rate": 4.4157860841440324e-05, "loss": 0.2284, "step": 13540 }, { "epoch": 1.36, "learning_rate": 4.4155750890769614e-05, "loss": 0.256, "step": 13541 }, { "epoch": 1.36, "learning_rate": 4.4153640850017085e-05, "loss": 0.6244, "step": 13542 }, { "epoch": 1.36, "learning_rate": 4.415153071919617e-05, "loss": 0.6926, "step": 13543 }, { "epoch": 1.36, "learning_rate": 4.4149420498320294e-05, "loss": 0.6872, "step": 13544 }, { "epoch": 1.36, "learning_rate": 4.414731018740289e-05, "loss": 0.1593, "step": 13545 }, { "epoch": 1.36, "learning_rate": 4.414519978645739e-05, "loss": 0.549, "step": 13546 }, { "epoch": 1.36, "learning_rate": 4.414308929549721e-05, "loss": 0.5374, "step": 13547 }, { "epoch": 1.36, "learning_rate": 4.41409787145358e-05, "loss": 0.5137, "step": 13548 }, { "epoch": 1.36, "learning_rate": 4.4138868043586585e-05, "loss": 0.8874, "step": 13549 }, { "epoch": 1.36, "learning_rate": 4.413675728266298e-05, "loss": 0.4014, "step": 13550 }, { "epoch": 1.36, "learning_rate": 4.4134646431778435e-05, "loss": 0.2895, "step": 13551 }, { "epoch": 1.36, "learning_rate": 4.4132535490946375e-05, "loss": 0.9716, "step": 13552 }, { "epoch": 1.36, "learning_rate": 4.413042446018024e-05, "loss": 0.6981, "step": 13553 }, { "epoch": 1.36, "learning_rate": 4.412831333949345e-05, "loss": 1.3437, "step": 13554 }, { "epoch": 1.36, "learning_rate": 4.412620212889947e-05, "loss": 0.5369, "step": 13555 }, { "epoch": 1.36, "learning_rate": 4.412409082841169e-05, "loss": 0.9619, "step": 13556 }, { "epoch": 1.36, "learning_rate": 4.412197943804358e-05, "loss": 1.0152, "step": 13557 }, { "epoch": 1.36, "learning_rate": 4.4119867957808574e-05, "loss": 0.7557, "step": 13558 }, { "epoch": 1.36, "learning_rate": 4.411775638772008e-05, "loss": 0.6762, "step": 13559 }, { "epoch": 1.36, "learning_rate": 4.411564472779157e-05, "loss": 0.316, "step": 13560 }, { "epoch": 1.36, "learning_rate": 4.411353297803645e-05, "loss": 0.3384, "step": 13561 }, { "epoch": 1.36, "learning_rate": 4.411142113846819e-05, "loss": 0.3974, "step": 13562 }, { "epoch": 1.36, "learning_rate": 4.4109309209100206e-05, "loss": 0.5611, "step": 13563 }, { "epoch": 1.36, "learning_rate": 4.4107197189945955e-05, "loss": 0.4443, "step": 13564 }, { "epoch": 1.36, "learning_rate": 4.410508508101887e-05, "loss": 0.8573, "step": 13565 }, { "epoch": 1.36, "learning_rate": 4.410297288233237e-05, "loss": 1.1925, "step": 13566 }, { "epoch": 1.36, "learning_rate": 4.410086059389993e-05, "loss": 0.5725, "step": 13567 }, { "epoch": 1.36, "learning_rate": 4.409874821573498e-05, "loss": 0.3156, "step": 13568 }, { "epoch": 1.36, "learning_rate": 4.409663574785095e-05, "loss": 0.6014, "step": 13569 }, { "epoch": 1.36, "learning_rate": 4.40945231902613e-05, "loss": 0.7292, "step": 13570 }, { "epoch": 1.37, "learning_rate": 4.4092410542979465e-05, "loss": 0.7587, "step": 13571 }, { "epoch": 1.37, "learning_rate": 4.4090297806018886e-05, "loss": 1.0297, "step": 13572 }, { "epoch": 1.37, "learning_rate": 4.408818497939301e-05, "loss": 0.3985, "step": 13573 }, { "epoch": 1.37, "learning_rate": 4.40860720631153e-05, "loss": 0.5083, "step": 13574 }, { "epoch": 1.37, "learning_rate": 4.4083959057199174e-05, "loss": 0.7963, "step": 13575 }, { "epoch": 1.37, "learning_rate": 4.4081845961658086e-05, "loss": 0.3573, "step": 13576 }, { "epoch": 1.37, "learning_rate": 4.407973277650549e-05, "loss": 0.1621, "step": 13577 }, { "epoch": 1.37, "learning_rate": 4.4077619501754845e-05, "loss": 0.7647, "step": 13578 }, { "epoch": 1.37, "learning_rate": 4.407550613741957e-05, "loss": 0.1312, "step": 13579 }, { "epoch": 1.37, "learning_rate": 4.407339268351313e-05, "loss": 0.2781, "step": 13580 }, { "epoch": 1.37, "learning_rate": 4.407127914004898e-05, "loss": 0.4881, "step": 13581 }, { "epoch": 1.37, "learning_rate": 4.406916550704055e-05, "loss": 0.8722, "step": 13582 }, { "epoch": 1.37, "learning_rate": 4.406705178450131e-05, "loss": 0.3406, "step": 13583 }, { "epoch": 1.37, "learning_rate": 4.4064937972444707e-05, "loss": 1.2383, "step": 13584 }, { "epoch": 1.37, "learning_rate": 4.406282407088419e-05, "loss": 0.9683, "step": 13585 }, { "epoch": 1.37, "learning_rate": 4.40607100798332e-05, "loss": 0.2101, "step": 13586 }, { "epoch": 1.37, "learning_rate": 4.4058595999305204e-05, "loss": 0.3433, "step": 13587 }, { "epoch": 1.37, "learning_rate": 4.405648182931366e-05, "loss": 0.9084, "step": 13588 }, { "epoch": 1.37, "learning_rate": 4.4054367569872e-05, "loss": 0.2441, "step": 13589 }, { "epoch": 1.37, "learning_rate": 4.4052253220993704e-05, "loss": 0.6864, "step": 13590 }, { "epoch": 1.37, "learning_rate": 4.4050138782692214e-05, "loss": 0.3977, "step": 13591 }, { "epoch": 1.37, "learning_rate": 4.4048024254980974e-05, "loss": 0.6759, "step": 13592 }, { "epoch": 1.37, "learning_rate": 4.404590963787346e-05, "loss": 0.2833, "step": 13593 }, { "epoch": 1.37, "learning_rate": 4.404379493138313e-05, "loss": 0.34, "step": 13594 }, { "epoch": 1.37, "learning_rate": 4.404168013552341e-05, "loss": 0.2184, "step": 13595 }, { "epoch": 1.37, "learning_rate": 4.4039565250307786e-05, "loss": 0.2011, "step": 13596 }, { "epoch": 1.37, "learning_rate": 4.403745027574972e-05, "loss": 1.4245, "step": 13597 }, { "epoch": 1.37, "learning_rate": 4.403533521186265e-05, "loss": 0.6127, "step": 13598 }, { "epoch": 1.37, "learning_rate": 4.4033220058660046e-05, "loss": 0.1885, "step": 13599 }, { "epoch": 1.37, "learning_rate": 4.4031104816155374e-05, "loss": 0.923, "step": 13600 }, { "epoch": 1.37, "learning_rate": 4.4028989484362094e-05, "loss": 1.0232, "step": 13601 }, { "epoch": 1.37, "learning_rate": 4.402687406329365e-05, "loss": 0.178, "step": 13602 }, { "epoch": 1.37, "learning_rate": 4.4024758552963516e-05, "loss": 0.9109, "step": 13603 }, { "epoch": 1.37, "learning_rate": 4.402264295338516e-05, "loss": 1.2683, "step": 13604 }, { "epoch": 1.37, "learning_rate": 4.402052726457204e-05, "loss": 0.3295, "step": 13605 }, { "epoch": 1.37, "learning_rate": 4.401841148653761e-05, "loss": 0.4202, "step": 13606 }, { "epoch": 1.37, "learning_rate": 4.401629561929535e-05, "loss": 0.3633, "step": 13607 }, { "epoch": 1.37, "learning_rate": 4.4014179662858717e-05, "loss": 0.8797, "step": 13608 }, { "epoch": 1.37, "learning_rate": 4.4012063617241166e-05, "loss": 0.9878, "step": 13609 }, { "epoch": 1.37, "learning_rate": 4.4009947482456186e-05, "loss": 0.351, "step": 13610 }, { "epoch": 1.37, "learning_rate": 4.4007831258517214e-05, "loss": 0.2602, "step": 13611 }, { "epoch": 1.37, "learning_rate": 4.400571494543774e-05, "loss": 0.3396, "step": 13612 }, { "epoch": 1.37, "learning_rate": 4.400359854323124e-05, "loss": 0.629, "step": 13613 }, { "epoch": 1.37, "learning_rate": 4.4001482051911146e-05, "loss": 1.4216, "step": 13614 }, { "epoch": 1.37, "learning_rate": 4.399936547149096e-05, "loss": 0.2804, "step": 13615 }, { "epoch": 1.37, "learning_rate": 4.399724880198412e-05, "loss": 0.5126, "step": 13616 }, { "epoch": 1.37, "learning_rate": 4.399513204340413e-05, "loss": 0.9625, "step": 13617 }, { "epoch": 1.37, "learning_rate": 4.399301519576443e-05, "loss": 0.7933, "step": 13618 }, { "epoch": 1.37, "learning_rate": 4.399089825907851e-05, "loss": 0.5044, "step": 13619 }, { "epoch": 1.37, "learning_rate": 4.398878123335984e-05, "loss": 0.448, "step": 13620 }, { "epoch": 1.37, "learning_rate": 4.398666411862189e-05, "loss": 0.3437, "step": 13621 }, { "epoch": 1.37, "learning_rate": 4.398454691487812e-05, "loss": 0.9344, "step": 13622 }, { "epoch": 1.37, "learning_rate": 4.398242962214203e-05, "loss": 0.6945, "step": 13623 }, { "epoch": 1.37, "learning_rate": 4.3980312240427066e-05, "loss": 1.09, "step": 13624 }, { "epoch": 1.37, "learning_rate": 4.397819476974671e-05, "loss": 0.7081, "step": 13625 }, { "epoch": 1.37, "learning_rate": 4.397607721011445e-05, "loss": 0.6214, "step": 13626 }, { "epoch": 1.37, "learning_rate": 4.3973959561543735e-05, "loss": 1.1842, "step": 13627 }, { "epoch": 1.37, "learning_rate": 4.397184182404807e-05, "loss": 0.3422, "step": 13628 }, { "epoch": 1.37, "learning_rate": 4.396972399764092e-05, "loss": 0.9146, "step": 13629 }, { "epoch": 1.37, "learning_rate": 4.396760608233576e-05, "loss": 0.1263, "step": 13630 }, { "epoch": 1.37, "learning_rate": 4.396548807814606e-05, "loss": 0.9178, "step": 13631 }, { "epoch": 1.37, "learning_rate": 4.396336998508531e-05, "loss": 0.5521, "step": 13632 }, { "epoch": 1.37, "learning_rate": 4.3961251803167e-05, "loss": 1.1855, "step": 13633 }, { "epoch": 1.37, "learning_rate": 4.395913353240458e-05, "loss": 0.1181, "step": 13634 }, { "epoch": 1.37, "learning_rate": 4.395701517281154e-05, "loss": 1.262, "step": 13635 }, { "epoch": 1.37, "learning_rate": 4.395489672440137e-05, "loss": 0.9201, "step": 13636 }, { "epoch": 1.37, "learning_rate": 4.395277818718755e-05, "loss": 1.1733, "step": 13637 }, { "epoch": 1.37, "learning_rate": 4.395065956118356e-05, "loss": 0.9146, "step": 13638 }, { "epoch": 1.37, "learning_rate": 4.3948540846402874e-05, "loss": 1.1734, "step": 13639 }, { "epoch": 1.37, "learning_rate": 4.394642204285898e-05, "loss": 0.2776, "step": 13640 }, { "epoch": 1.37, "learning_rate": 4.394430315056536e-05, "loss": 0.7112, "step": 13641 }, { "epoch": 1.37, "learning_rate": 4.39421841695355e-05, "loss": 1.2921, "step": 13642 }, { "epoch": 1.37, "learning_rate": 4.394006509978289e-05, "loss": 0.3783, "step": 13643 }, { "epoch": 1.37, "learning_rate": 4.3937945941321004e-05, "loss": 0.7714, "step": 13644 }, { "epoch": 1.37, "learning_rate": 4.393582669416334e-05, "loss": 0.1512, "step": 13645 }, { "epoch": 1.37, "learning_rate": 4.3933707358323365e-05, "loss": 0.2459, "step": 13646 }, { "epoch": 1.37, "learning_rate": 4.3931587933814585e-05, "loss": 0.8197, "step": 13647 }, { "epoch": 1.37, "learning_rate": 4.392946842065047e-05, "loss": 0.7078, "step": 13648 }, { "epoch": 1.37, "learning_rate": 4.392734881884453e-05, "loss": 0.8787, "step": 13649 }, { "epoch": 1.37, "learning_rate": 4.3925229128410225e-05, "loss": 0.7443, "step": 13650 }, { "epoch": 1.37, "learning_rate": 4.3923109349361074e-05, "loss": 0.2382, "step": 13651 }, { "epoch": 1.37, "learning_rate": 4.392098948171054e-05, "loss": 0.8304, "step": 13652 }, { "epoch": 1.37, "learning_rate": 4.3918869525472135e-05, "loss": 0.7953, "step": 13653 }, { "epoch": 1.37, "learning_rate": 4.3916749480659335e-05, "loss": 0.9011, "step": 13654 }, { "epoch": 1.37, "learning_rate": 4.391462934728563e-05, "loss": 0.6217, "step": 13655 }, { "epoch": 1.37, "learning_rate": 4.391250912536453e-05, "loss": 0.7548, "step": 13656 }, { "epoch": 1.37, "learning_rate": 4.39103888149095e-05, "loss": 0.4823, "step": 13657 }, { "epoch": 1.37, "learning_rate": 4.390826841593405e-05, "loss": 0.9192, "step": 13658 }, { "epoch": 1.37, "learning_rate": 4.390614792845168e-05, "loss": 1.4514, "step": 13659 }, { "epoch": 1.37, "learning_rate": 4.3904027352475874e-05, "loss": 0.9014, "step": 13660 }, { "epoch": 1.37, "learning_rate": 4.390190668802012e-05, "loss": 0.6323, "step": 13661 }, { "epoch": 1.37, "learning_rate": 4.389978593509793e-05, "loss": 0.9378, "step": 13662 }, { "epoch": 1.37, "learning_rate": 4.389766509372278e-05, "loss": 0.7092, "step": 13663 }, { "epoch": 1.37, "learning_rate": 4.389554416390817e-05, "loss": 0.0932, "step": 13664 }, { "epoch": 1.37, "learning_rate": 4.389342314566762e-05, "loss": 1.0194, "step": 13665 }, { "epoch": 1.37, "learning_rate": 4.38913020390146e-05, "loss": 0.2608, "step": 13666 }, { "epoch": 1.37, "learning_rate": 4.388918084396262e-05, "loss": 0.3775, "step": 13667 }, { "epoch": 1.37, "learning_rate": 4.388705956052518e-05, "loss": 0.4008, "step": 13668 }, { "epoch": 1.38, "learning_rate": 4.388493818871577e-05, "loss": 0.7444, "step": 13669 }, { "epoch": 1.38, "learning_rate": 4.38828167285479e-05, "loss": 0.5213, "step": 13670 }, { "epoch": 1.38, "learning_rate": 4.388069518003507e-05, "loss": 0.5154, "step": 13671 }, { "epoch": 1.38, "learning_rate": 4.3878573543190774e-05, "loss": 0.6923, "step": 13672 }, { "epoch": 1.38, "learning_rate": 4.387645181802851e-05, "loss": 0.1674, "step": 13673 }, { "epoch": 1.38, "learning_rate": 4.387433000456179e-05, "loss": 0.5395, "step": 13674 }, { "epoch": 1.38, "learning_rate": 4.387220810280411e-05, "loss": 0.1372, "step": 13675 }, { "epoch": 1.38, "learning_rate": 4.387008611276897e-05, "loss": 0.5183, "step": 13676 }, { "epoch": 1.38, "learning_rate": 4.386796403446989e-05, "loss": 0.7163, "step": 13677 }, { "epoch": 1.38, "learning_rate": 4.386584186792036e-05, "loss": 0.4444, "step": 13678 }, { "epoch": 1.38, "learning_rate": 4.386371961313388e-05, "loss": 0.4118, "step": 13679 }, { "epoch": 1.38, "learning_rate": 4.386159727012396e-05, "loss": 0.9389, "step": 13680 }, { "epoch": 1.38, "learning_rate": 4.3859474838904126e-05, "loss": 0.7736, "step": 13681 }, { "epoch": 1.38, "learning_rate": 4.385735231948785e-05, "loss": 0.7398, "step": 13682 }, { "epoch": 1.38, "learning_rate": 4.385522971188866e-05, "loss": 0.8025, "step": 13683 }, { "epoch": 1.38, "learning_rate": 4.385310701612006e-05, "loss": 0.709, "step": 13684 }, { "epoch": 1.38, "learning_rate": 4.385098423219555e-05, "loss": 1.6904, "step": 13685 }, { "epoch": 1.38, "learning_rate": 4.384886136012866e-05, "loss": 0.3103, "step": 13686 }, { "epoch": 1.38, "learning_rate": 4.384673839993287e-05, "loss": 0.2919, "step": 13687 }, { "epoch": 1.38, "learning_rate": 4.384461535162172e-05, "loss": 0.8828, "step": 13688 }, { "epoch": 1.38, "learning_rate": 4.384249221520869e-05, "loss": 0.8614, "step": 13689 }, { "epoch": 1.38, "learning_rate": 4.384036899070731e-05, "loss": 0.4917, "step": 13690 }, { "epoch": 1.38, "learning_rate": 4.383824567813109e-05, "loss": 0.3271, "step": 13691 }, { "epoch": 1.38, "learning_rate": 4.3836122277493535e-05, "loss": 0.3797, "step": 13692 }, { "epoch": 1.38, "learning_rate": 4.383399878880816e-05, "loss": 0.2426, "step": 13693 }, { "epoch": 1.38, "learning_rate": 4.3831875212088486e-05, "loss": 0.9791, "step": 13694 }, { "epoch": 1.38, "learning_rate": 4.382975154734802e-05, "loss": 0.3923, "step": 13695 }, { "epoch": 1.38, "learning_rate": 4.382762779460027e-05, "loss": 0.49, "step": 13696 }, { "epoch": 1.38, "learning_rate": 4.3825503953858754e-05, "loss": 0.4335, "step": 13697 }, { "epoch": 1.38, "learning_rate": 4.3823380025137e-05, "loss": 0.2611, "step": 13698 }, { "epoch": 1.38, "learning_rate": 4.382125600844851e-05, "loss": 0.4632, "step": 13699 }, { "epoch": 1.38, "learning_rate": 4.381913190380681e-05, "loss": 0.9795, "step": 13700 }, { "epoch": 1.38, "learning_rate": 4.38170077112254e-05, "loss": 0.6079, "step": 13701 }, { "epoch": 1.38, "learning_rate": 4.381488343071782e-05, "loss": 0.7471, "step": 13702 }, { "epoch": 1.38, "learning_rate": 4.381275906229757e-05, "loss": 0.8256, "step": 13703 }, { "epoch": 1.38, "learning_rate": 4.381063460597819e-05, "loss": 0.6204, "step": 13704 }, { "epoch": 1.38, "learning_rate": 4.3808510061773164e-05, "loss": 0.9431, "step": 13705 }, { "epoch": 1.38, "learning_rate": 4.380638542969604e-05, "loss": 0.4031, "step": 13706 }, { "epoch": 1.38, "learning_rate": 4.380426070976034e-05, "loss": 0.3693, "step": 13707 }, { "epoch": 1.38, "learning_rate": 4.380213590197957e-05, "loss": 0.4188, "step": 13708 }, { "epoch": 1.38, "learning_rate": 4.380001100636726e-05, "loss": 0.3025, "step": 13709 }, { "epoch": 1.38, "learning_rate": 4.379788602293692e-05, "loss": 0.5712, "step": 13710 }, { "epoch": 1.38, "learning_rate": 4.37957609517021e-05, "loss": 0.3218, "step": 13711 }, { "epoch": 1.38, "learning_rate": 4.3793635792676284e-05, "loss": 0.5749, "step": 13712 }, { "epoch": 1.38, "learning_rate": 4.379151054587304e-05, "loss": 0.6558, "step": 13713 }, { "epoch": 1.38, "learning_rate": 4.378938521130586e-05, "loss": 0.2555, "step": 13714 }, { "epoch": 1.38, "learning_rate": 4.3787259788988274e-05, "loss": 0.4535, "step": 13715 }, { "epoch": 1.38, "learning_rate": 4.378513427893381e-05, "loss": 0.3905, "step": 13716 }, { "epoch": 1.38, "learning_rate": 4.3783008681156004e-05, "loss": 0.3324, "step": 13717 }, { "epoch": 1.38, "learning_rate": 4.3780882995668375e-05, "loss": 0.5312, "step": 13718 }, { "epoch": 1.38, "learning_rate": 4.3778757222484436e-05, "loss": 0.4463, "step": 13719 }, { "epoch": 1.38, "learning_rate": 4.377663136161774e-05, "loss": 0.3003, "step": 13720 }, { "epoch": 1.38, "learning_rate": 4.377450541308179e-05, "loss": 0.5997, "step": 13721 }, { "epoch": 1.38, "learning_rate": 4.3772379376890146e-05, "loss": 0.6533, "step": 13722 }, { "epoch": 1.38, "learning_rate": 4.377025325305631e-05, "loss": 0.8757, "step": 13723 }, { "epoch": 1.38, "learning_rate": 4.376812704159382e-05, "loss": 0.2645, "step": 13724 }, { "epoch": 1.38, "learning_rate": 4.376600074251621e-05, "loss": 0.2761, "step": 13725 }, { "epoch": 1.38, "learning_rate": 4.3763874355836995e-05, "loss": 0.4702, "step": 13726 }, { "epoch": 1.38, "learning_rate": 4.376174788156974e-05, "loss": 0.2361, "step": 13727 }, { "epoch": 1.38, "learning_rate": 4.3759621319727945e-05, "loss": 1.134, "step": 13728 }, { "epoch": 1.38, "learning_rate": 4.375749467032515e-05, "loss": 0.6016, "step": 13729 }, { "epoch": 1.38, "learning_rate": 4.3755367933374906e-05, "loss": 0.7597, "step": 13730 }, { "epoch": 1.38, "learning_rate": 4.3753241108890726e-05, "loss": 0.738, "step": 13731 }, { "epoch": 1.38, "learning_rate": 4.3751114196886155e-05, "loss": 0.4475, "step": 13732 }, { "epoch": 1.38, "learning_rate": 4.374898719737472e-05, "loss": 0.2734, "step": 13733 }, { "epoch": 1.38, "learning_rate": 4.374686011036996e-05, "loss": 0.6029, "step": 13734 }, { "epoch": 1.38, "learning_rate": 4.374473293588541e-05, "loss": 0.4862, "step": 13735 }, { "epoch": 1.38, "learning_rate": 4.374260567393461e-05, "loss": 0.2728, "step": 13736 }, { "epoch": 1.38, "learning_rate": 4.3740478324531105e-05, "loss": 0.1776, "step": 13737 }, { "epoch": 1.38, "learning_rate": 4.373835088768842e-05, "loss": 0.5361, "step": 13738 }, { "epoch": 1.38, "learning_rate": 4.373622336342009e-05, "loss": 0.5606, "step": 13739 }, { "epoch": 1.38, "learning_rate": 4.3734095751739665e-05, "loss": 0.9437, "step": 13740 }, { "epoch": 1.38, "learning_rate": 4.3731968052660675e-05, "loss": 0.4372, "step": 13741 }, { "epoch": 1.38, "learning_rate": 4.372984026619666e-05, "loss": 0.4509, "step": 13742 }, { "epoch": 1.38, "learning_rate": 4.3727712392361176e-05, "loss": 0.1733, "step": 13743 }, { "epoch": 1.38, "learning_rate": 4.372558443116774e-05, "loss": 0.2354, "step": 13744 }, { "epoch": 1.38, "learning_rate": 4.372345638262992e-05, "loss": 0.8052, "step": 13745 }, { "epoch": 1.38, "learning_rate": 4.372132824676124e-05, "loss": 1.3384, "step": 13746 }, { "epoch": 1.38, "learning_rate": 4.371920002357524e-05, "loss": 0.5461, "step": 13747 }, { "epoch": 1.38, "learning_rate": 4.371707171308547e-05, "loss": 0.789, "step": 13748 }, { "epoch": 1.38, "learning_rate": 4.371494331530548e-05, "loss": 0.0852, "step": 13749 }, { "epoch": 1.38, "learning_rate": 4.371281483024881e-05, "loss": 0.4853, "step": 13750 }, { "epoch": 1.38, "learning_rate": 4.3710686257928996e-05, "loss": 0.4426, "step": 13751 }, { "epoch": 1.38, "learning_rate": 4.37085575983596e-05, "loss": 0.8126, "step": 13752 }, { "epoch": 1.38, "learning_rate": 4.370642885155415e-05, "loss": 0.4101, "step": 13753 }, { "epoch": 1.38, "learning_rate": 4.37043000175262e-05, "loss": 0.6845, "step": 13754 }, { "epoch": 1.38, "learning_rate": 4.3702171096289307e-05, "loss": 0.3154, "step": 13755 }, { "epoch": 1.38, "learning_rate": 4.3700042087856995e-05, "loss": 0.1885, "step": 13756 }, { "epoch": 1.38, "learning_rate": 4.369791299224285e-05, "loss": 0.5349, "step": 13757 }, { "epoch": 1.38, "learning_rate": 4.369578380946037e-05, "loss": 0.5998, "step": 13758 }, { "epoch": 1.38, "learning_rate": 4.369365453952315e-05, "loss": 0.5896, "step": 13759 }, { "epoch": 1.38, "learning_rate": 4.369152518244472e-05, "loss": 0.4498, "step": 13760 }, { "epoch": 1.38, "learning_rate": 4.368939573823862e-05, "loss": 1.1054, "step": 13761 }, { "epoch": 1.38, "learning_rate": 4.368726620691844e-05, "loss": 1.1094, "step": 13762 }, { "epoch": 1.38, "learning_rate": 4.368513658849768e-05, "loss": 0.8847, "step": 13763 }, { "epoch": 1.38, "learning_rate": 4.368300688298993e-05, "loss": 1.1605, "step": 13764 }, { "epoch": 1.38, "learning_rate": 4.368087709040871e-05, "loss": 0.8351, "step": 13765 }, { "epoch": 1.38, "learning_rate": 4.367874721076762e-05, "loss": 0.6637, "step": 13766 }, { "epoch": 1.38, "learning_rate": 4.367661724408017e-05, "loss": 0.9038, "step": 13767 }, { "epoch": 1.39, "learning_rate": 4.367448719035993e-05, "loss": 0.9003, "step": 13768 }, { "epoch": 1.39, "learning_rate": 4.367235704962046e-05, "loss": 0.9617, "step": 13769 }, { "epoch": 1.39, "learning_rate": 4.367022682187531e-05, "loss": 0.5684, "step": 13770 }, { "epoch": 1.39, "learning_rate": 4.3668096507138045e-05, "loss": 0.5113, "step": 13771 }, { "epoch": 1.39, "learning_rate": 4.36659661054222e-05, "loss": 0.5235, "step": 13772 }, { "epoch": 1.39, "learning_rate": 4.3663835616741354e-05, "loss": 0.2742, "step": 13773 }, { "epoch": 1.39, "learning_rate": 4.366170504110905e-05, "loss": 0.1869, "step": 13774 }, { "epoch": 1.39, "learning_rate": 4.365957437853886e-05, "loss": 0.8505, "step": 13775 }, { "epoch": 1.39, "learning_rate": 4.3657443629044326e-05, "loss": 0.4788, "step": 13776 }, { "epoch": 1.39, "learning_rate": 4.365531279263903e-05, "loss": 0.6299, "step": 13777 }, { "epoch": 1.39, "learning_rate": 4.36531818693365e-05, "loss": 0.6362, "step": 13778 }, { "epoch": 1.39, "learning_rate": 4.3651050859150326e-05, "loss": 0.9345, "step": 13779 }, { "epoch": 1.39, "learning_rate": 4.3648919762094056e-05, "loss": 0.5531, "step": 13780 }, { "epoch": 1.39, "learning_rate": 4.364678857818125e-05, "loss": 1.0335, "step": 13781 }, { "epoch": 1.39, "learning_rate": 4.364465730742548e-05, "loss": 0.4276, "step": 13782 }, { "epoch": 1.39, "learning_rate": 4.364252594984029e-05, "loss": 0.6439, "step": 13783 }, { "epoch": 1.39, "learning_rate": 4.3640394505439264e-05, "loss": 1.3263, "step": 13784 }, { "epoch": 1.39, "learning_rate": 4.363826297423595e-05, "loss": 0.5161, "step": 13785 }, { "epoch": 1.39, "learning_rate": 4.363613135624393e-05, "loss": 0.4075, "step": 13786 }, { "epoch": 1.39, "learning_rate": 4.363399965147675e-05, "loss": 0.3304, "step": 13787 }, { "epoch": 1.39, "learning_rate": 4.3631867859947985e-05, "loss": 1.5528, "step": 13788 }, { "epoch": 1.39, "learning_rate": 4.3629735981671195e-05, "loss": 0.2259, "step": 13789 }, { "epoch": 1.39, "learning_rate": 4.3627604016659945e-05, "loss": 0.6042, "step": 13790 }, { "epoch": 1.39, "learning_rate": 4.3625471964927816e-05, "loss": 1.1961, "step": 13791 }, { "epoch": 1.39, "learning_rate": 4.3623339826488366e-05, "loss": 0.5536, "step": 13792 }, { "epoch": 1.39, "learning_rate": 4.362120760135517e-05, "loss": 0.6158, "step": 13793 }, { "epoch": 1.39, "learning_rate": 4.3619075289541776e-05, "loss": 1.0652, "step": 13794 }, { "epoch": 1.39, "learning_rate": 4.361694289106178e-05, "loss": 0.2951, "step": 13795 }, { "epoch": 1.39, "learning_rate": 4.361481040592873e-05, "loss": 0.9286, "step": 13796 }, { "epoch": 1.39, "learning_rate": 4.361267783415621e-05, "loss": 0.4447, "step": 13797 }, { "epoch": 1.39, "learning_rate": 4.361054517575779e-05, "loss": 0.5487, "step": 13798 }, { "epoch": 1.39, "learning_rate": 4.360841243074704e-05, "loss": 0.8154, "step": 13799 }, { "epoch": 1.39, "learning_rate": 4.3606279599137526e-05, "loss": 0.516, "step": 13800 }, { "epoch": 1.39, "learning_rate": 4.360414668094282e-05, "loss": 0.5592, "step": 13801 }, { "epoch": 1.39, "learning_rate": 4.3602013676176504e-05, "loss": 0.2229, "step": 13802 }, { "epoch": 1.39, "learning_rate": 4.3599880584852155e-05, "loss": 0.4309, "step": 13803 }, { "epoch": 1.39, "learning_rate": 4.3597747406983325e-05, "loss": 0.5782, "step": 13804 }, { "epoch": 1.39, "learning_rate": 4.3595614142583616e-05, "loss": 0.3362, "step": 13805 }, { "epoch": 1.39, "learning_rate": 4.3593480791666585e-05, "loss": 1.0106, "step": 13806 }, { "epoch": 1.39, "learning_rate": 4.3591347354245807e-05, "loss": 1.1565, "step": 13807 }, { "epoch": 1.39, "learning_rate": 4.358921383033487e-05, "loss": 0.1723, "step": 13808 }, { "epoch": 1.39, "learning_rate": 4.358708021994736e-05, "loss": 0.7271, "step": 13809 }, { "epoch": 1.39, "learning_rate": 4.358494652309683e-05, "loss": 0.4993, "step": 13810 }, { "epoch": 1.39, "learning_rate": 4.358281273979686e-05, "loss": 0.1133, "step": 13811 }, { "epoch": 1.39, "learning_rate": 4.358067887006104e-05, "loss": 0.4531, "step": 13812 }, { "epoch": 1.39, "learning_rate": 4.357854491390295e-05, "loss": 1.0796, "step": 13813 }, { "epoch": 1.39, "learning_rate": 4.357641087133617e-05, "loss": 0.7125, "step": 13814 }, { "epoch": 1.39, "learning_rate": 4.357427674237427e-05, "loss": 0.3062, "step": 13815 }, { "epoch": 1.39, "learning_rate": 4.3572142527030836e-05, "loss": 1.2331, "step": 13816 }, { "epoch": 1.39, "learning_rate": 4.357000822531945e-05, "loss": 0.6009, "step": 13817 }, { "epoch": 1.39, "learning_rate": 4.3567873837253694e-05, "loss": 1.1948, "step": 13818 }, { "epoch": 1.39, "learning_rate": 4.356573936284716e-05, "loss": 0.4083, "step": 13819 }, { "epoch": 1.39, "learning_rate": 4.35636048021134e-05, "loss": 0.2449, "step": 13820 }, { "epoch": 1.39, "learning_rate": 4.356147015506605e-05, "loss": 0.057, "step": 13821 }, { "epoch": 1.39, "learning_rate": 4.355933542171863e-05, "loss": 1.0526, "step": 13822 }, { "epoch": 1.39, "learning_rate": 4.355720060208478e-05, "loss": 1.021, "step": 13823 }, { "epoch": 1.39, "learning_rate": 4.355506569617805e-05, "loss": 0.1646, "step": 13824 }, { "epoch": 1.39, "learning_rate": 4.355293070401205e-05, "loss": 0.6923, "step": 13825 }, { "epoch": 1.39, "learning_rate": 4.355079562560035e-05, "loss": 0.2883, "step": 13826 }, { "epoch": 1.39, "learning_rate": 4.354866046095654e-05, "loss": 0.6754, "step": 13827 }, { "epoch": 1.39, "learning_rate": 4.354652521009421e-05, "loss": 0.6162, "step": 13828 }, { "epoch": 1.39, "learning_rate": 4.3544389873026944e-05, "loss": 1.1033, "step": 13829 }, { "epoch": 1.39, "learning_rate": 4.3542254449768334e-05, "loss": 0.7895, "step": 13830 }, { "epoch": 1.39, "learning_rate": 4.354011894033197e-05, "loss": 0.5065, "step": 13831 }, { "epoch": 1.39, "learning_rate": 4.353798334473145e-05, "loss": 0.6342, "step": 13832 }, { "epoch": 1.39, "learning_rate": 4.353584766298034e-05, "loss": 1.19, "step": 13833 }, { "epoch": 1.39, "learning_rate": 4.3533711895092245e-05, "loss": 0.2465, "step": 13834 }, { "epoch": 1.39, "learning_rate": 4.353157604108076e-05, "loss": 0.2612, "step": 13835 }, { "epoch": 1.39, "learning_rate": 4.352944010095947e-05, "loss": 0.57, "step": 13836 }, { "epoch": 1.39, "learning_rate": 4.352730407474198e-05, "loss": 1.0307, "step": 13837 }, { "epoch": 1.39, "learning_rate": 4.352516796244186e-05, "loss": 0.745, "step": 13838 }, { "epoch": 1.39, "learning_rate": 4.3523031764072726e-05, "loss": 0.5358, "step": 13839 }, { "epoch": 1.39, "learning_rate": 4.352089547964816e-05, "loss": 0.6243, "step": 13840 }, { "epoch": 1.39, "learning_rate": 4.3518759109181756e-05, "loss": 0.3816, "step": 13841 }, { "epoch": 1.39, "learning_rate": 4.3516622652687115e-05, "loss": 0.3561, "step": 13842 }, { "epoch": 1.39, "learning_rate": 4.351448611017783e-05, "loss": 0.1962, "step": 13843 }, { "epoch": 1.39, "learning_rate": 4.351234948166749e-05, "loss": 0.8804, "step": 13844 }, { "epoch": 1.39, "learning_rate": 4.35102127671697e-05, "loss": 0.5386, "step": 13845 }, { "epoch": 1.39, "learning_rate": 4.350807596669806e-05, "loss": 1.2517, "step": 13846 }, { "epoch": 1.39, "learning_rate": 4.350593908026616e-05, "loss": 0.6884, "step": 13847 }, { "epoch": 1.39, "learning_rate": 4.3503802107887604e-05, "loss": 0.2414, "step": 13848 }, { "epoch": 1.39, "learning_rate": 4.3501665049575986e-05, "loss": 0.7472, "step": 13849 }, { "epoch": 1.39, "learning_rate": 4.349952790534491e-05, "loss": 2.3816, "step": 13850 }, { "epoch": 1.39, "learning_rate": 4.349739067520798e-05, "loss": 0.3902, "step": 13851 }, { "epoch": 1.39, "learning_rate": 4.349525335917877e-05, "loss": 0.4069, "step": 13852 }, { "epoch": 1.39, "learning_rate": 4.3493115957270924e-05, "loss": 0.2306, "step": 13853 }, { "epoch": 1.39, "learning_rate": 4.3490978469498003e-05, "loss": 0.7973, "step": 13854 }, { "epoch": 1.39, "learning_rate": 4.348884089587364e-05, "loss": 0.0819, "step": 13855 }, { "epoch": 1.39, "learning_rate": 4.348670323641142e-05, "loss": 1.2655, "step": 13856 }, { "epoch": 1.39, "learning_rate": 4.348456549112496e-05, "loss": 0.8582, "step": 13857 }, { "epoch": 1.39, "learning_rate": 4.348242766002785e-05, "loss": 0.7368, "step": 13858 }, { "epoch": 1.39, "learning_rate": 4.3480289743133685e-05, "loss": 1.0955, "step": 13859 }, { "epoch": 1.39, "learning_rate": 4.34781517404561e-05, "loss": 0.2848, "step": 13860 }, { "epoch": 1.39, "learning_rate": 4.347601365200868e-05, "loss": 0.8269, "step": 13861 }, { "epoch": 1.39, "learning_rate": 4.347387547780504e-05, "loss": 0.2061, "step": 13862 }, { "epoch": 1.39, "learning_rate": 4.347173721785878e-05, "loss": 0.2511, "step": 13863 }, { "epoch": 1.39, "learning_rate": 4.3469598872183506e-05, "loss": 1.0031, "step": 13864 }, { "epoch": 1.39, "learning_rate": 4.346746044079283e-05, "loss": 0.9688, "step": 13865 }, { "epoch": 1.4, "learning_rate": 4.346532192370036e-05, "loss": 0.3898, "step": 13866 }, { "epoch": 1.4, "learning_rate": 4.346318332091971e-05, "loss": 1.185, "step": 13867 }, { "epoch": 1.4, "learning_rate": 4.346104463246447e-05, "loss": 0.7314, "step": 13868 }, { "epoch": 1.4, "learning_rate": 4.3458905858348276e-05, "loss": 0.652, "step": 13869 }, { "epoch": 1.4, "learning_rate": 4.3456766998584706e-05, "loss": 0.2662, "step": 13870 }, { "epoch": 1.4, "learning_rate": 4.345462805318741e-05, "loss": 0.4857, "step": 13871 }, { "epoch": 1.4, "learning_rate": 4.345248902216997e-05, "loss": 1.0023, "step": 13872 }, { "epoch": 1.4, "learning_rate": 4.345034990554601e-05, "loss": 0.6845, "step": 13873 }, { "epoch": 1.4, "learning_rate": 4.344821070332914e-05, "loss": 0.5167, "step": 13874 }, { "epoch": 1.4, "learning_rate": 4.344607141553297e-05, "loss": 0.4848, "step": 13875 }, { "epoch": 1.4, "learning_rate": 4.344393204217113e-05, "loss": 0.9147, "step": 13876 }, { "epoch": 1.4, "learning_rate": 4.34417925832572e-05, "loss": 0.5694, "step": 13877 }, { "epoch": 1.4, "learning_rate": 4.3439653038804824e-05, "loss": 0.5706, "step": 13878 }, { "epoch": 1.4, "learning_rate": 4.343751340882761e-05, "loss": 0.6803, "step": 13879 }, { "epoch": 1.4, "learning_rate": 4.343537369333917e-05, "loss": 0.2671, "step": 13880 }, { "epoch": 1.4, "learning_rate": 4.343323389235313e-05, "loss": 0.5712, "step": 13881 }, { "epoch": 1.4, "learning_rate": 4.34310940058831e-05, "loss": 0.5135, "step": 13882 }, { "epoch": 1.4, "learning_rate": 4.342895403394269e-05, "loss": 0.2161, "step": 13883 }, { "epoch": 1.4, "learning_rate": 4.342681397654553e-05, "loss": 0.9184, "step": 13884 }, { "epoch": 1.4, "learning_rate": 4.342467383370523e-05, "loss": 0.171, "step": 13885 }, { "epoch": 1.4, "learning_rate": 4.342253360543541e-05, "loss": 0.5379, "step": 13886 }, { "epoch": 1.4, "learning_rate": 4.34203932917497e-05, "loss": 0.5704, "step": 13887 }, { "epoch": 1.4, "learning_rate": 4.341825289266171e-05, "loss": 1.0809, "step": 13888 }, { "epoch": 1.4, "learning_rate": 4.341611240818506e-05, "loss": 0.381, "step": 13889 }, { "epoch": 1.4, "learning_rate": 4.341397183833338e-05, "loss": 0.3532, "step": 13890 }, { "epoch": 1.4, "learning_rate": 4.341183118312027e-05, "loss": 0.9377, "step": 13891 }, { "epoch": 1.4, "learning_rate": 4.34096904425594e-05, "loss": 0.5586, "step": 13892 }, { "epoch": 1.4, "learning_rate": 4.340754961666433e-05, "loss": 0.6637, "step": 13893 }, { "epoch": 1.4, "learning_rate": 4.3405408705448725e-05, "loss": 1.4756, "step": 13894 }, { "epoch": 1.4, "learning_rate": 4.340326770892621e-05, "loss": 1.533, "step": 13895 }, { "epoch": 1.4, "learning_rate": 4.340112662711039e-05, "loss": 0.3829, "step": 13896 }, { "epoch": 1.4, "learning_rate": 4.3398985460014894e-05, "loss": 0.3753, "step": 13897 }, { "epoch": 1.4, "learning_rate": 4.339684420765335e-05, "loss": 0.5617, "step": 13898 }, { "epoch": 1.4, "learning_rate": 4.3394702870039404e-05, "loss": 0.3299, "step": 13899 }, { "epoch": 1.4, "learning_rate": 4.339256144718664e-05, "loss": 0.749, "step": 13900 }, { "epoch": 1.4, "learning_rate": 4.339041993910872e-05, "loss": 0.651, "step": 13901 }, { "epoch": 1.4, "learning_rate": 4.3388278345819265e-05, "loss": 0.7561, "step": 13902 }, { "epoch": 1.4, "learning_rate": 4.33861366673319e-05, "loss": 0.3269, "step": 13903 }, { "epoch": 1.4, "learning_rate": 4.338399490366025e-05, "loss": 0.2225, "step": 13904 }, { "epoch": 1.4, "learning_rate": 4.338185305481795e-05, "loss": 0.1649, "step": 13905 }, { "epoch": 1.4, "learning_rate": 4.337971112081863e-05, "loss": 0.3208, "step": 13906 }, { "epoch": 1.4, "learning_rate": 4.33775691016759e-05, "loss": 0.6703, "step": 13907 }, { "epoch": 1.4, "learning_rate": 4.3375426997403436e-05, "loss": 0.8598, "step": 13908 }, { "epoch": 1.4, "learning_rate": 4.3373284808014825e-05, "loss": 0.1325, "step": 13909 }, { "epoch": 1.4, "learning_rate": 4.337114253352373e-05, "loss": 0.6578, "step": 13910 }, { "epoch": 1.4, "learning_rate": 4.336900017394377e-05, "loss": 0.78, "step": 13911 }, { "epoch": 1.4, "learning_rate": 4.3366857729288566e-05, "loss": 0.7146, "step": 13912 }, { "epoch": 1.4, "learning_rate": 4.336471519957177e-05, "loss": 0.1861, "step": 13913 }, { "epoch": 1.4, "learning_rate": 4.336257258480701e-05, "loss": 0.3575, "step": 13914 }, { "epoch": 1.4, "learning_rate": 4.336042988500793e-05, "loss": 0.654, "step": 13915 }, { "epoch": 1.4, "learning_rate": 4.335828710018815e-05, "loss": 0.1844, "step": 13916 }, { "epoch": 1.4, "learning_rate": 4.3356144230361307e-05, "loss": 0.6145, "step": 13917 }, { "epoch": 1.4, "learning_rate": 4.3354001275541046e-05, "loss": 0.2037, "step": 13918 }, { "epoch": 1.4, "learning_rate": 4.335185823574101e-05, "loss": 0.6107, "step": 13919 }, { "epoch": 1.4, "learning_rate": 4.334971511097482e-05, "loss": 0.7787, "step": 13920 }, { "epoch": 1.4, "learning_rate": 4.334757190125612e-05, "loss": 0.7845, "step": 13921 }, { "epoch": 1.4, "learning_rate": 4.3345428606598556e-05, "loss": 0.528, "step": 13922 }, { "epoch": 1.4, "learning_rate": 4.3343285227015754e-05, "loss": 0.3466, "step": 13923 }, { "epoch": 1.4, "learning_rate": 4.334114176252137e-05, "loss": 1.1081, "step": 13924 }, { "epoch": 1.4, "learning_rate": 4.333899821312903e-05, "loss": 1.2132, "step": 13925 }, { "epoch": 1.4, "learning_rate": 4.333685457885237e-05, "loss": 0.2823, "step": 13926 }, { "epoch": 1.4, "learning_rate": 4.333471085970506e-05, "loss": 0.0565, "step": 13927 }, { "epoch": 1.4, "learning_rate": 4.333256705570071e-05, "loss": 0.5535, "step": 13928 }, { "epoch": 1.4, "learning_rate": 4.3330423166852986e-05, "loss": 0.2824, "step": 13929 }, { "epoch": 1.4, "learning_rate": 4.3328279193175505e-05, "loss": 0.2796, "step": 13930 }, { "epoch": 1.4, "learning_rate": 4.332613513468194e-05, "loss": 1.0099, "step": 13931 }, { "epoch": 1.4, "learning_rate": 4.33239909913859e-05, "loss": 0.6281, "step": 13932 }, { "epoch": 1.4, "learning_rate": 4.332184676330107e-05, "loss": 0.7006, "step": 13933 }, { "epoch": 1.4, "learning_rate": 4.331970245044107e-05, "loss": 0.3242, "step": 13934 }, { "epoch": 1.4, "learning_rate": 4.331755805281956e-05, "loss": 0.5061, "step": 13935 }, { "epoch": 1.4, "learning_rate": 4.331541357045016e-05, "loss": 0.3917, "step": 13936 }, { "epoch": 1.4, "learning_rate": 4.331326900334654e-05, "loss": 1.0704, "step": 13937 }, { "epoch": 1.4, "learning_rate": 4.331112435152234e-05, "loss": 0.4132, "step": 13938 }, { "epoch": 1.4, "learning_rate": 4.330897961499121e-05, "loss": 0.2862, "step": 13939 }, { "epoch": 1.4, "learning_rate": 4.3306834793766796e-05, "loss": 0.9229, "step": 13940 }, { "epoch": 1.4, "learning_rate": 4.330468988786274e-05, "loss": 0.659, "step": 13941 }, { "epoch": 1.4, "learning_rate": 4.330254489729271e-05, "loss": 0.6156, "step": 13942 }, { "epoch": 1.4, "learning_rate": 4.3300399822070344e-05, "loss": 1.1864, "step": 13943 }, { "epoch": 1.4, "learning_rate": 4.3298254662209294e-05, "loss": 0.4811, "step": 13944 }, { "epoch": 1.4, "learning_rate": 4.329610941772321e-05, "loss": 0.7479, "step": 13945 }, { "epoch": 1.4, "learning_rate": 4.3293964088625725e-05, "loss": 0.333, "step": 13946 }, { "epoch": 1.4, "learning_rate": 4.329181867493054e-05, "loss": 0.4914, "step": 13947 }, { "epoch": 1.4, "learning_rate": 4.328967317665126e-05, "loss": 0.4342, "step": 13948 }, { "epoch": 1.4, "learning_rate": 4.3287527593801554e-05, "loss": 0.9772, "step": 13949 }, { "epoch": 1.4, "learning_rate": 4.328538192639508e-05, "loss": 0.1121, "step": 13950 }, { "epoch": 1.4, "learning_rate": 4.32832361744455e-05, "loss": 0.1976, "step": 13951 }, { "epoch": 1.4, "learning_rate": 4.3281090337966445e-05, "loss": 0.9123, "step": 13952 }, { "epoch": 1.4, "learning_rate": 4.327894441697158e-05, "loss": 0.3931, "step": 13953 }, { "epoch": 1.4, "learning_rate": 4.327679841147458e-05, "loss": 0.5565, "step": 13954 }, { "epoch": 1.4, "learning_rate": 4.327465232148907e-05, "loss": 0.562, "step": 13955 }, { "epoch": 1.4, "learning_rate": 4.327250614702874e-05, "loss": 0.3067, "step": 13956 }, { "epoch": 1.4, "learning_rate": 4.3270359888107214e-05, "loss": 0.3224, "step": 13957 }, { "epoch": 1.4, "learning_rate": 4.326821354473818e-05, "loss": 0.3235, "step": 13958 }, { "epoch": 1.4, "learning_rate": 4.326606711693527e-05, "loss": 1.1048, "step": 13959 }, { "epoch": 1.4, "learning_rate": 4.326392060471216e-05, "loss": 0.4493, "step": 13960 }, { "epoch": 1.4, "learning_rate": 4.3261774008082506e-05, "loss": 0.8645, "step": 13961 }, { "epoch": 1.4, "learning_rate": 4.325962732705996e-05, "loss": 0.7866, "step": 13962 }, { "epoch": 1.4, "learning_rate": 4.3257480561658205e-05, "loss": 0.832, "step": 13963 }, { "epoch": 1.4, "learning_rate": 4.3255333711890876e-05, "loss": 0.1844, "step": 13964 }, { "epoch": 1.41, "learning_rate": 4.325318677777165e-05, "loss": 0.9968, "step": 13965 }, { "epoch": 1.41, "learning_rate": 4.3251039759314185e-05, "loss": 0.7785, "step": 13966 }, { "epoch": 1.41, "learning_rate": 4.324889265653214e-05, "loss": 0.4146, "step": 13967 }, { "epoch": 1.41, "learning_rate": 4.324674546943919e-05, "loss": 1.4186, "step": 13968 }, { "epoch": 1.41, "learning_rate": 4.324459819804898e-05, "loss": 0.8621, "step": 13969 }, { "epoch": 1.41, "learning_rate": 4.3242450842375196e-05, "loss": 1.0538, "step": 13970 }, { "epoch": 1.41, "learning_rate": 4.324030340243148e-05, "loss": 0.4084, "step": 13971 }, { "epoch": 1.41, "learning_rate": 4.323815587823152e-05, "loss": 0.3021, "step": 13972 }, { "epoch": 1.41, "learning_rate": 4.323600826978897e-05, "loss": 1.0192, "step": 13973 }, { "epoch": 1.41, "learning_rate": 4.3233860577117506e-05, "loss": 0.3004, "step": 13974 }, { "epoch": 1.41, "learning_rate": 4.323171280023077e-05, "loss": 0.5762, "step": 13975 }, { "epoch": 1.41, "learning_rate": 4.3229564939142466e-05, "loss": 0.283, "step": 13976 }, { "epoch": 1.41, "learning_rate": 4.322741699386623e-05, "loss": 0.5706, "step": 13977 }, { "epoch": 1.41, "learning_rate": 4.322526896441575e-05, "loss": 0.2514, "step": 13978 }, { "epoch": 1.41, "learning_rate": 4.3223120850804694e-05, "loss": 0.2906, "step": 13979 }, { "epoch": 1.41, "learning_rate": 4.322097265304672e-05, "loss": 0.3255, "step": 13980 }, { "epoch": 1.41, "learning_rate": 4.3218824371155505e-05, "loss": 0.6336, "step": 13981 }, { "epoch": 1.41, "learning_rate": 4.3216676005144726e-05, "loss": 0.0785, "step": 13982 }, { "epoch": 1.41, "learning_rate": 4.321452755502805e-05, "loss": 0.8454, "step": 13983 }, { "epoch": 1.41, "learning_rate": 4.3212379020819144e-05, "loss": 0.2768, "step": 13984 }, { "epoch": 1.41, "learning_rate": 4.321023040253167e-05, "loss": 0.1818, "step": 13985 }, { "epoch": 1.41, "learning_rate": 4.320808170017934e-05, "loss": 1.0491, "step": 13986 }, { "epoch": 1.41, "learning_rate": 4.320593291377578e-05, "loss": 1.044, "step": 13987 }, { "epoch": 1.41, "learning_rate": 4.3203784043334704e-05, "loss": 0.3981, "step": 13988 }, { "epoch": 1.41, "learning_rate": 4.320163508886977e-05, "loss": 0.8113, "step": 13989 }, { "epoch": 1.41, "learning_rate": 4.319948605039464e-05, "loss": 0.3672, "step": 13990 }, { "epoch": 1.41, "learning_rate": 4.319733692792301e-05, "loss": 1.18, "step": 13991 }, { "epoch": 1.41, "learning_rate": 4.3195187721468546e-05, "loss": 0.4549, "step": 13992 }, { "epoch": 1.41, "learning_rate": 4.3193038431044925e-05, "loss": 0.6647, "step": 13993 }, { "epoch": 1.41, "learning_rate": 4.319088905666583e-05, "loss": 0.3849, "step": 13994 }, { "epoch": 1.41, "learning_rate": 4.3188739598344937e-05, "loss": 0.1433, "step": 13995 }, { "epoch": 1.41, "learning_rate": 4.318659005609592e-05, "loss": 0.4868, "step": 13996 }, { "epoch": 1.41, "learning_rate": 4.318444042993246e-05, "loss": 0.768, "step": 13997 }, { "epoch": 1.41, "learning_rate": 4.318229071986823e-05, "loss": 0.8438, "step": 13998 }, { "epoch": 1.41, "learning_rate": 4.318014092591692e-05, "loss": 0.9415, "step": 13999 }, { "epoch": 1.41, "learning_rate": 4.317799104809222e-05, "loss": 0.9045, "step": 14000 }, { "epoch": 1.41, "learning_rate": 4.317584108640778e-05, "loss": 0.6711, "step": 14001 }, { "epoch": 1.41, "learning_rate": 4.3173691040877314e-05, "loss": 0.695, "step": 14002 }, { "epoch": 1.41, "learning_rate": 4.317154091151448e-05, "loss": 0.4904, "step": 14003 }, { "epoch": 1.41, "learning_rate": 4.316939069833297e-05, "loss": 0.8978, "step": 14004 }, { "epoch": 1.41, "learning_rate": 4.316724040134648e-05, "loss": 0.9473, "step": 14005 }, { "epoch": 1.41, "learning_rate": 4.316509002056867e-05, "loss": 0.9225, "step": 14006 }, { "epoch": 1.41, "learning_rate": 4.316293955601324e-05, "loss": 0.4995, "step": 14007 }, { "epoch": 1.41, "learning_rate": 4.316078900769386e-05, "loss": 0.5116, "step": 14008 }, { "epoch": 1.41, "learning_rate": 4.315863837562425e-05, "loss": 0.5874, "step": 14009 }, { "epoch": 1.41, "learning_rate": 4.315648765981804e-05, "loss": 1.1921, "step": 14010 }, { "epoch": 1.41, "learning_rate": 4.3154336860288965e-05, "loss": 0.275, "step": 14011 }, { "epoch": 1.41, "learning_rate": 4.31521859770507e-05, "loss": 0.3903, "step": 14012 }, { "epoch": 1.41, "learning_rate": 4.3150035010116915e-05, "loss": 0.3542, "step": 14013 }, { "epoch": 1.41, "learning_rate": 4.314788395950131e-05, "loss": 0.4174, "step": 14014 }, { "epoch": 1.41, "learning_rate": 4.314573282521758e-05, "loss": 0.2835, "step": 14015 }, { "epoch": 1.41, "learning_rate": 4.3143581607279404e-05, "loss": 0.6484, "step": 14016 }, { "epoch": 1.41, "learning_rate": 4.3141430305700474e-05, "loss": 0.6409, "step": 14017 }, { "epoch": 1.41, "learning_rate": 4.313927892049449e-05, "loss": 0.3199, "step": 14018 }, { "epoch": 1.41, "learning_rate": 4.313712745167511e-05, "loss": 0.7552, "step": 14019 }, { "epoch": 1.41, "learning_rate": 4.3134975899256063e-05, "loss": 0.7596, "step": 14020 }, { "epoch": 1.41, "learning_rate": 4.313282426325104e-05, "loss": 1.5346, "step": 14021 }, { "epoch": 1.41, "learning_rate": 4.3130672543673695e-05, "loss": 0.158, "step": 14022 }, { "epoch": 1.41, "learning_rate": 4.312852074053776e-05, "loss": 1.3761, "step": 14023 }, { "epoch": 1.41, "learning_rate": 4.31263688538569e-05, "loss": 0.2586, "step": 14024 }, { "epoch": 1.41, "learning_rate": 4.3124216883644835e-05, "loss": 0.8057, "step": 14025 }, { "epoch": 1.41, "learning_rate": 4.3122064829915245e-05, "loss": 0.3153, "step": 14026 }, { "epoch": 1.41, "learning_rate": 4.311991269268182e-05, "loss": 0.2826, "step": 14027 }, { "epoch": 1.41, "learning_rate": 4.311776047195827e-05, "loss": 0.4574, "step": 14028 }, { "epoch": 1.41, "learning_rate": 4.3115608167758276e-05, "loss": 0.5908, "step": 14029 }, { "epoch": 1.41, "learning_rate": 4.311345578009555e-05, "loss": 0.1952, "step": 14030 }, { "epoch": 1.41, "learning_rate": 4.311130330898377e-05, "loss": 0.1788, "step": 14031 }, { "epoch": 1.41, "learning_rate": 4.310915075443664e-05, "loss": 1.2117, "step": 14032 }, { "epoch": 1.41, "learning_rate": 4.310699811646787e-05, "loss": 0.3614, "step": 14033 }, { "epoch": 1.41, "learning_rate": 4.310484539509115e-05, "loss": 0.4818, "step": 14034 }, { "epoch": 1.41, "learning_rate": 4.310269259032018e-05, "loss": 1.0975, "step": 14035 }, { "epoch": 1.41, "learning_rate": 4.3100539702168655e-05, "loss": 0.9354, "step": 14036 }, { "epoch": 1.41, "learning_rate": 4.3098386730650284e-05, "loss": 1.4378, "step": 14037 }, { "epoch": 1.41, "learning_rate": 4.3096233675778755e-05, "loss": 0.4502, "step": 14038 }, { "epoch": 1.41, "learning_rate": 4.309408053756778e-05, "loss": 0.9215, "step": 14039 }, { "epoch": 1.41, "learning_rate": 4.309192731603106e-05, "loss": 0.4994, "step": 14040 }, { "epoch": 1.41, "learning_rate": 4.3089774011182296e-05, "loss": 1.0834, "step": 14041 }, { "epoch": 1.41, "learning_rate": 4.308762062303519e-05, "loss": 0.6399, "step": 14042 }, { "epoch": 1.41, "learning_rate": 4.308546715160344e-05, "loss": 0.1572, "step": 14043 }, { "epoch": 1.41, "learning_rate": 4.308331359690076e-05, "loss": 0.4081, "step": 14044 }, { "epoch": 1.41, "learning_rate": 4.3081159958940844e-05, "loss": 0.2926, "step": 14045 }, { "epoch": 1.41, "learning_rate": 4.307900623773741e-05, "loss": 0.5908, "step": 14046 }, { "epoch": 1.41, "learning_rate": 4.3076852433304146e-05, "loss": 0.3819, "step": 14047 }, { "epoch": 1.41, "learning_rate": 4.307469854565478e-05, "loss": 0.7597, "step": 14048 }, { "epoch": 1.41, "learning_rate": 4.3072544574803e-05, "loss": 0.9938, "step": 14049 }, { "epoch": 1.41, "learning_rate": 4.3070390520762517e-05, "loss": 0.4812, "step": 14050 }, { "epoch": 1.41, "learning_rate": 4.306823638354704e-05, "loss": 0.4623, "step": 14051 }, { "epoch": 1.41, "learning_rate": 4.306608216317028e-05, "loss": 0.451, "step": 14052 }, { "epoch": 1.41, "learning_rate": 4.306392785964595e-05, "loss": 0.5041, "step": 14053 }, { "epoch": 1.41, "learning_rate": 4.3061773472987746e-05, "loss": 0.7207, "step": 14054 }, { "epoch": 1.41, "learning_rate": 4.305961900320939e-05, "loss": 0.2322, "step": 14055 }, { "epoch": 1.41, "learning_rate": 4.305746445032457e-05, "loss": 1.0904, "step": 14056 }, { "epoch": 1.41, "learning_rate": 4.305530981434704e-05, "loss": 0.2005, "step": 14057 }, { "epoch": 1.41, "learning_rate": 4.305315509529047e-05, "loss": 0.5616, "step": 14058 }, { "epoch": 1.41, "learning_rate": 4.3051000293168584e-05, "loss": 0.6459, "step": 14059 }, { "epoch": 1.41, "learning_rate": 4.3048845407995106e-05, "loss": 0.7857, "step": 14060 }, { "epoch": 1.41, "learning_rate": 4.3046690439783736e-05, "loss": 0.4575, "step": 14061 }, { "epoch": 1.41, "learning_rate": 4.304453538854818e-05, "loss": 1.0569, "step": 14062 }, { "epoch": 1.42, "learning_rate": 4.304238025430218e-05, "loss": 0.3312, "step": 14063 }, { "epoch": 1.42, "learning_rate": 4.304022503705943e-05, "loss": 0.7889, "step": 14064 }, { "epoch": 1.42, "learning_rate": 4.303806973683364e-05, "loss": 0.8837, "step": 14065 }, { "epoch": 1.42, "learning_rate": 4.303591435363854e-05, "loss": 0.8787, "step": 14066 }, { "epoch": 1.42, "learning_rate": 4.3033758887487844e-05, "loss": 0.4339, "step": 14067 }, { "epoch": 1.42, "learning_rate": 4.303160333839526e-05, "loss": 0.2927, "step": 14068 }, { "epoch": 1.42, "learning_rate": 4.3029447706374514e-05, "loss": 0.6126, "step": 14069 }, { "epoch": 1.42, "learning_rate": 4.3027291991439314e-05, "loss": 0.5299, "step": 14070 }, { "epoch": 1.42, "learning_rate": 4.302513619360339e-05, "loss": 0.7035, "step": 14071 }, { "epoch": 1.42, "learning_rate": 4.3022980312880435e-05, "loss": 0.3125, "step": 14072 }, { "epoch": 1.42, "learning_rate": 4.302082434928421e-05, "loss": 1.0856, "step": 14073 }, { "epoch": 1.42, "learning_rate": 4.30186683028284e-05, "loss": 0.7371, "step": 14074 }, { "epoch": 1.42, "learning_rate": 4.3016512173526744e-05, "loss": 1.1233, "step": 14075 }, { "epoch": 1.42, "learning_rate": 4.3014355961392956e-05, "loss": 1.3255, "step": 14076 }, { "epoch": 1.42, "learning_rate": 4.3012199666440764e-05, "loss": 1.0223, "step": 14077 }, { "epoch": 1.42, "learning_rate": 4.301004328868387e-05, "loss": 0.2437, "step": 14078 }, { "epoch": 1.42, "learning_rate": 4.300788682813601e-05, "loss": 0.3724, "step": 14079 }, { "epoch": 1.42, "learning_rate": 4.300573028481092e-05, "loss": 0.4255, "step": 14080 }, { "epoch": 1.42, "learning_rate": 4.300357365872229e-05, "loss": 1.0786, "step": 14081 }, { "epoch": 1.42, "learning_rate": 4.300141694988388e-05, "loss": 0.6016, "step": 14082 }, { "epoch": 1.42, "learning_rate": 4.29992601583094e-05, "loss": 0.647, "step": 14083 }, { "epoch": 1.42, "learning_rate": 4.299710328401256e-05, "loss": 0.9965, "step": 14084 }, { "epoch": 1.42, "learning_rate": 4.2994946327007115e-05, "loss": 0.959, "step": 14085 }, { "epoch": 1.42, "learning_rate": 4.299278928730676e-05, "loss": 0.2893, "step": 14086 }, { "epoch": 1.42, "learning_rate": 4.299063216492525e-05, "loss": 1.2799, "step": 14087 }, { "epoch": 1.42, "learning_rate": 4.298847495987629e-05, "loss": 0.5962, "step": 14088 }, { "epoch": 1.42, "learning_rate": 4.298631767217362e-05, "loss": 0.6231, "step": 14089 }, { "epoch": 1.42, "learning_rate": 4.2984160301830966e-05, "loss": 0.905, "step": 14090 }, { "epoch": 1.42, "learning_rate": 4.298200284886205e-05, "loss": 0.7614, "step": 14091 }, { "epoch": 1.42, "learning_rate": 4.297984531328061e-05, "loss": 0.7956, "step": 14092 }, { "epoch": 1.42, "learning_rate": 4.297768769510037e-05, "loss": 1.0745, "step": 14093 }, { "epoch": 1.42, "learning_rate": 4.297552999433506e-05, "loss": 0.9378, "step": 14094 }, { "epoch": 1.42, "learning_rate": 4.297337221099842e-05, "loss": 0.2112, "step": 14095 }, { "epoch": 1.42, "learning_rate": 4.297121434510418e-05, "loss": 0.5835, "step": 14096 }, { "epoch": 1.42, "learning_rate": 4.2969056396666045e-05, "loss": 0.1574, "step": 14097 }, { "epoch": 1.42, "learning_rate": 4.296689836569779e-05, "loss": 0.634, "step": 14098 }, { "epoch": 1.42, "learning_rate": 4.296474025221312e-05, "loss": 0.3861, "step": 14099 }, { "epoch": 1.42, "learning_rate": 4.296258205622577e-05, "loss": 0.9662, "step": 14100 }, { "epoch": 1.42, "learning_rate": 4.296042377774949e-05, "loss": 0.279, "step": 14101 }, { "epoch": 1.42, "learning_rate": 4.295826541679799e-05, "loss": 0.1803, "step": 14102 }, { "epoch": 1.42, "learning_rate": 4.295610697338503e-05, "loss": 0.6219, "step": 14103 }, { "epoch": 1.42, "learning_rate": 4.295394844752432e-05, "loss": 0.917, "step": 14104 }, { "epoch": 1.42, "learning_rate": 4.295178983922962e-05, "loss": 0.9564, "step": 14105 }, { "epoch": 1.42, "learning_rate": 4.294963114851466e-05, "loss": 1.3209, "step": 14106 }, { "epoch": 1.42, "learning_rate": 4.294747237539317e-05, "loss": 0.1816, "step": 14107 }, { "epoch": 1.42, "learning_rate": 4.2945313519878885e-05, "loss": 0.5103, "step": 14108 }, { "epoch": 1.42, "learning_rate": 4.294315458198556e-05, "loss": 0.955, "step": 14109 }, { "epoch": 1.42, "learning_rate": 4.294099556172692e-05, "loss": 0.2731, "step": 14110 }, { "epoch": 1.42, "learning_rate": 4.2938836459116694e-05, "loss": 0.1379, "step": 14111 }, { "epoch": 1.42, "learning_rate": 4.2936677274168657e-05, "loss": 1.0371, "step": 14112 }, { "epoch": 1.42, "learning_rate": 4.293451800689651e-05, "loss": 1.1163, "step": 14113 }, { "epoch": 1.42, "learning_rate": 4.2932358657314014e-05, "loss": 1.3063, "step": 14114 }, { "epoch": 1.42, "learning_rate": 4.2930199225434914e-05, "loss": 0.3271, "step": 14115 }, { "epoch": 1.42, "learning_rate": 4.292803971127294e-05, "loss": 0.1483, "step": 14116 }, { "epoch": 1.42, "learning_rate": 4.2925880114841845e-05, "loss": 0.883, "step": 14117 }, { "epoch": 1.42, "learning_rate": 4.2923720436155356e-05, "loss": 0.2599, "step": 14118 }, { "epoch": 1.42, "learning_rate": 4.292156067522724e-05, "loss": 0.6064, "step": 14119 }, { "epoch": 1.42, "learning_rate": 4.291940083207121e-05, "loss": 1.2929, "step": 14120 }, { "epoch": 1.42, "learning_rate": 4.291724090670104e-05, "loss": 0.3011, "step": 14121 }, { "epoch": 1.42, "learning_rate": 4.291508089913046e-05, "loss": 0.1753, "step": 14122 }, { "epoch": 1.42, "learning_rate": 4.291292080937322e-05, "loss": 0.3823, "step": 14123 }, { "epoch": 1.42, "learning_rate": 4.291076063744306e-05, "loss": 0.5643, "step": 14124 }, { "epoch": 1.42, "learning_rate": 4.290860038335375e-05, "loss": 0.2387, "step": 14125 }, { "epoch": 1.42, "learning_rate": 4.2906440047118995e-05, "loss": 0.3152, "step": 14126 }, { "epoch": 1.42, "learning_rate": 4.290427962875257e-05, "loss": 0.3577, "step": 14127 }, { "epoch": 1.42, "learning_rate": 4.290211912826822e-05, "loss": 0.8047, "step": 14128 }, { "epoch": 1.42, "learning_rate": 4.289995854567969e-05, "loss": 0.4533, "step": 14129 }, { "epoch": 1.42, "learning_rate": 4.289779788100074e-05, "loss": 0.3995, "step": 14130 }, { "epoch": 1.42, "learning_rate": 4.28956371342451e-05, "loss": 0.9553, "step": 14131 }, { "epoch": 1.42, "learning_rate": 4.289347630542654e-05, "loss": 0.5026, "step": 14132 }, { "epoch": 1.42, "learning_rate": 4.28913153945588e-05, "loss": 1.1076, "step": 14133 }, { "epoch": 1.42, "learning_rate": 4.288915440165562e-05, "loss": 0.3895, "step": 14134 }, { "epoch": 1.42, "learning_rate": 4.288699332673078e-05, "loss": 0.9095, "step": 14135 }, { "epoch": 1.42, "learning_rate": 4.2884832169798014e-05, "loss": 1.1935, "step": 14136 }, { "epoch": 1.42, "learning_rate": 4.2882670930871076e-05, "loss": 0.5432, "step": 14137 }, { "epoch": 1.42, "learning_rate": 4.2880509609963714e-05, "loss": 0.8856, "step": 14138 }, { "epoch": 1.42, "learning_rate": 4.287834820708971e-05, "loss": 0.9555, "step": 14139 }, { "epoch": 1.42, "learning_rate": 4.287618672226278e-05, "loss": 0.6598, "step": 14140 }, { "epoch": 1.42, "learning_rate": 4.2874025155496694e-05, "loss": 1.2119, "step": 14141 }, { "epoch": 1.42, "learning_rate": 4.2871863506805214e-05, "loss": 0.4204, "step": 14142 }, { "epoch": 1.42, "learning_rate": 4.2869701776202084e-05, "loss": 0.3192, "step": 14143 }, { "epoch": 1.42, "learning_rate": 4.286753996370108e-05, "loss": 0.3374, "step": 14144 }, { "epoch": 1.42, "learning_rate": 4.286537806931594e-05, "loss": 0.473, "step": 14145 }, { "epoch": 1.42, "learning_rate": 4.286321609306044e-05, "loss": 0.8508, "step": 14146 }, { "epoch": 1.42, "learning_rate": 4.2861054034948304e-05, "loss": 0.8628, "step": 14147 }, { "epoch": 1.42, "learning_rate": 4.2858891894993334e-05, "loss": 0.504, "step": 14148 }, { "epoch": 1.42, "learning_rate": 4.285672967320926e-05, "loss": 0.2212, "step": 14149 }, { "epoch": 1.42, "learning_rate": 4.2854567369609844e-05, "loss": 0.6174, "step": 14150 }, { "epoch": 1.42, "learning_rate": 4.285240498420886e-05, "loss": 0.2639, "step": 14151 }, { "epoch": 1.42, "learning_rate": 4.2850242517020054e-05, "loss": 1.1369, "step": 14152 }, { "epoch": 1.42, "learning_rate": 4.2848079968057194e-05, "loss": 0.8871, "step": 14153 }, { "epoch": 1.42, "learning_rate": 4.2845917337334045e-05, "loss": 0.3805, "step": 14154 }, { "epoch": 1.42, "learning_rate": 4.284375462486436e-05, "loss": 0.496, "step": 14155 }, { "epoch": 1.42, "learning_rate": 4.284159183066191e-05, "loss": 0.238, "step": 14156 }, { "epoch": 1.42, "learning_rate": 4.283942895474045e-05, "loss": 1.2508, "step": 14157 }, { "epoch": 1.42, "learning_rate": 4.2837265997113757e-05, "loss": 0.4875, "step": 14158 }, { "epoch": 1.42, "learning_rate": 4.2835102957795575e-05, "loss": 1.208, "step": 14159 }, { "epoch": 1.42, "learning_rate": 4.283293983679969e-05, "loss": 0.9548, "step": 14160 }, { "epoch": 1.42, "learning_rate": 4.283077663413985e-05, "loss": 0.2209, "step": 14161 }, { "epoch": 1.43, "learning_rate": 4.282861334982984e-05, "loss": 0.3441, "step": 14162 }, { "epoch": 1.43, "learning_rate": 4.28264499838834e-05, "loss": 0.8475, "step": 14163 }, { "epoch": 1.43, "learning_rate": 4.282428653631433e-05, "loss": 0.4603, "step": 14164 }, { "epoch": 1.43, "learning_rate": 4.282212300713637e-05, "loss": 0.0626, "step": 14165 }, { "epoch": 1.43, "learning_rate": 4.281995939636329e-05, "loss": 0.7215, "step": 14166 }, { "epoch": 1.43, "learning_rate": 4.2817795704008874e-05, "loss": 0.2804, "step": 14167 }, { "epoch": 1.43, "learning_rate": 4.2815631930086874e-05, "loss": 0.5314, "step": 14168 }, { "epoch": 1.43, "learning_rate": 4.281346807461108e-05, "loss": 0.3871, "step": 14169 }, { "epoch": 1.43, "learning_rate": 4.2811304137595236e-05, "loss": 1.0105, "step": 14170 }, { "epoch": 1.43, "learning_rate": 4.2809140119053136e-05, "loss": 0.4655, "step": 14171 }, { "epoch": 1.43, "learning_rate": 4.2806976018998535e-05, "loss": 1.2261, "step": 14172 }, { "epoch": 1.43, "learning_rate": 4.280481183744521e-05, "loss": 0.5984, "step": 14173 }, { "epoch": 1.43, "learning_rate": 4.280264757440695e-05, "loss": 0.9212, "step": 14174 }, { "epoch": 1.43, "learning_rate": 4.2800483229897486e-05, "loss": 1.0238, "step": 14175 }, { "epoch": 1.43, "learning_rate": 4.2798318803930635e-05, "loss": 0.3409, "step": 14176 }, { "epoch": 1.43, "learning_rate": 4.279615429652014e-05, "loss": 1.1497, "step": 14177 }, { "epoch": 1.43, "learning_rate": 4.27939897076798e-05, "loss": 0.9067, "step": 14178 }, { "epoch": 1.43, "learning_rate": 4.279182503742337e-05, "loss": 0.5967, "step": 14179 }, { "epoch": 1.43, "learning_rate": 4.278966028576463e-05, "loss": 0.4681, "step": 14180 }, { "epoch": 1.43, "learning_rate": 4.2787495452717355e-05, "loss": 0.5422, "step": 14181 }, { "epoch": 1.43, "learning_rate": 4.278533053829533e-05, "loss": 0.2084, "step": 14182 }, { "epoch": 1.43, "learning_rate": 4.278316554251232e-05, "loss": 0.6852, "step": 14183 }, { "epoch": 1.43, "learning_rate": 4.278100046538211e-05, "loss": 1.1485, "step": 14184 }, { "epoch": 1.43, "learning_rate": 4.277883530691848e-05, "loss": 0.557, "step": 14185 }, { "epoch": 1.43, "learning_rate": 4.277667006713519e-05, "loss": 1.092, "step": 14186 }, { "epoch": 1.43, "learning_rate": 4.277450474604605e-05, "loss": 0.3107, "step": 14187 }, { "epoch": 1.43, "learning_rate": 4.277233934366481e-05, "loss": 0.1464, "step": 14188 }, { "epoch": 1.43, "learning_rate": 4.2770173860005256e-05, "loss": 0.1852, "step": 14189 }, { "epoch": 1.43, "learning_rate": 4.276800829508119e-05, "loss": 0.5992, "step": 14190 }, { "epoch": 1.43, "learning_rate": 4.276584264890636e-05, "loss": 0.9281, "step": 14191 }, { "epoch": 1.43, "learning_rate": 4.276367692149457e-05, "loss": 0.306, "step": 14192 }, { "epoch": 1.43, "learning_rate": 4.2761511112859595e-05, "loss": 0.676, "step": 14193 }, { "epoch": 1.43, "learning_rate": 4.275934522301522e-05, "loss": 0.616, "step": 14194 }, { "epoch": 1.43, "learning_rate": 4.2757179251975226e-05, "loss": 0.8702, "step": 14195 }, { "epoch": 1.43, "learning_rate": 4.27550131997534e-05, "loss": 0.4856, "step": 14196 }, { "epoch": 1.43, "learning_rate": 4.2752847066363524e-05, "loss": 0.3933, "step": 14197 }, { "epoch": 1.43, "learning_rate": 4.2750680851819364e-05, "loss": 1.365, "step": 14198 }, { "epoch": 1.43, "learning_rate": 4.274851455613474e-05, "loss": 0.7069, "step": 14199 }, { "epoch": 1.43, "learning_rate": 4.2746348179323406e-05, "loss": 0.3701, "step": 14200 }, { "epoch": 1.43, "learning_rate": 4.274418172139916e-05, "loss": 1.2719, "step": 14201 }, { "epoch": 1.43, "learning_rate": 4.27420151823758e-05, "loss": 0.4298, "step": 14202 }, { "epoch": 1.43, "learning_rate": 4.27398485622671e-05, "loss": 0.1915, "step": 14203 }, { "epoch": 1.43, "learning_rate": 4.2737681861086854e-05, "loss": 0.665, "step": 14204 }, { "epoch": 1.43, "learning_rate": 4.273551507884884e-05, "loss": 0.3451, "step": 14205 }, { "epoch": 1.43, "learning_rate": 4.273334821556686e-05, "loss": 1.1087, "step": 14206 }, { "epoch": 1.43, "learning_rate": 4.273118127125468e-05, "loss": 0.5893, "step": 14207 }, { "epoch": 1.43, "learning_rate": 4.2729014245926125e-05, "loss": 0.5691, "step": 14208 }, { "epoch": 1.43, "learning_rate": 4.2726847139594954e-05, "loss": 0.6789, "step": 14209 }, { "epoch": 1.43, "learning_rate": 4.272467995227497e-05, "loss": 1.2103, "step": 14210 }, { "epoch": 1.43, "learning_rate": 4.272251268397996e-05, "loss": 0.2738, "step": 14211 }, { "epoch": 1.43, "learning_rate": 4.272034533472373e-05, "loss": 0.7655, "step": 14212 }, { "epoch": 1.43, "learning_rate": 4.271817790452006e-05, "loss": 0.4761, "step": 14213 }, { "epoch": 1.43, "learning_rate": 4.271601039338273e-05, "loss": 0.2792, "step": 14214 }, { "epoch": 1.43, "learning_rate": 4.2713842801325565e-05, "loss": 0.4802, "step": 14215 }, { "epoch": 1.43, "learning_rate": 4.2711675128362325e-05, "loss": 0.3276, "step": 14216 }, { "epoch": 1.43, "learning_rate": 4.2709507374506836e-05, "loss": 0.9538, "step": 14217 }, { "epoch": 1.43, "learning_rate": 4.270733953977287e-05, "loss": 1.0315, "step": 14218 }, { "epoch": 1.43, "learning_rate": 4.2705171624174235e-05, "loss": 1.0499, "step": 14219 }, { "epoch": 1.43, "learning_rate": 4.2703003627724715e-05, "loss": 0.6917, "step": 14220 }, { "epoch": 1.43, "learning_rate": 4.2700835550438105e-05, "loss": 0.5605, "step": 14221 }, { "epoch": 1.43, "learning_rate": 4.269866739232822e-05, "loss": 0.5672, "step": 14222 }, { "epoch": 1.43, "learning_rate": 4.269649915340884e-05, "loss": 0.6619, "step": 14223 }, { "epoch": 1.43, "learning_rate": 4.2694330833693775e-05, "loss": 1.2205, "step": 14224 }, { "epoch": 1.43, "learning_rate": 4.269216243319681e-05, "loss": 0.4635, "step": 14225 }, { "epoch": 1.43, "learning_rate": 4.2689993951931764e-05, "loss": 0.6948, "step": 14226 }, { "epoch": 1.43, "learning_rate": 4.268782538991241e-05, "loss": 0.4784, "step": 14227 }, { "epoch": 1.43, "learning_rate": 4.2685656747152574e-05, "loss": 0.3576, "step": 14228 }, { "epoch": 1.43, "learning_rate": 4.268348802366604e-05, "loss": 1.0992, "step": 14229 }, { "epoch": 1.43, "learning_rate": 4.268131921946661e-05, "loss": 0.8899, "step": 14230 }, { "epoch": 1.43, "learning_rate": 4.267915033456809e-05, "loss": 0.7707, "step": 14231 }, { "epoch": 1.43, "learning_rate": 4.267698136898428e-05, "loss": 1.26, "step": 14232 }, { "epoch": 1.43, "learning_rate": 4.2674812322728985e-05, "loss": 0.9778, "step": 14233 }, { "epoch": 1.43, "learning_rate": 4.2672643195816006e-05, "loss": 0.6412, "step": 14234 }, { "epoch": 1.43, "learning_rate": 4.2670473988259154e-05, "loss": 0.2126, "step": 14235 }, { "epoch": 1.43, "learning_rate": 4.266830470007222e-05, "loss": 0.5024, "step": 14236 }, { "epoch": 1.43, "learning_rate": 4.2666135331269006e-05, "loss": 0.4833, "step": 14237 }, { "epoch": 1.43, "learning_rate": 4.266396588186333e-05, "loss": 1.4222, "step": 14238 }, { "epoch": 1.43, "learning_rate": 4.2661796351868994e-05, "loss": 0.6369, "step": 14239 }, { "epoch": 1.43, "learning_rate": 4.26596267412998e-05, "loss": 0.1309, "step": 14240 }, { "epoch": 1.43, "learning_rate": 4.2657457050169563e-05, "loss": 2.2539, "step": 14241 }, { "epoch": 1.43, "learning_rate": 4.265528727849209e-05, "loss": 0.3705, "step": 14242 }, { "epoch": 1.43, "learning_rate": 4.265311742628116e-05, "loss": 1.0303, "step": 14243 }, { "epoch": 1.43, "learning_rate": 4.2650947493550625e-05, "loss": 0.2516, "step": 14244 }, { "epoch": 1.43, "learning_rate": 4.2648777480314276e-05, "loss": 2.3516, "step": 14245 }, { "epoch": 1.43, "learning_rate": 4.26466073865859e-05, "loss": 0.9756, "step": 14246 }, { "epoch": 1.43, "learning_rate": 4.264443721237934e-05, "loss": 0.3533, "step": 14247 }, { "epoch": 1.43, "learning_rate": 4.264226695770839e-05, "loss": 0.2979, "step": 14248 }, { "epoch": 1.43, "learning_rate": 4.264009662258686e-05, "loss": 0.3104, "step": 14249 }, { "epoch": 1.43, "learning_rate": 4.263792620702855e-05, "loss": 0.7738, "step": 14250 }, { "epoch": 1.43, "learning_rate": 4.263575571104731e-05, "loss": 0.6414, "step": 14251 }, { "epoch": 1.43, "learning_rate": 4.2633585134656924e-05, "loss": 0.9925, "step": 14252 }, { "epoch": 1.43, "learning_rate": 4.263141447787119e-05, "loss": 0.6298, "step": 14253 }, { "epoch": 1.43, "learning_rate": 4.262924374070396e-05, "loss": 0.5692, "step": 14254 }, { "epoch": 1.43, "learning_rate": 4.262707292316901e-05, "loss": 1.6313, "step": 14255 }, { "epoch": 1.43, "learning_rate": 4.262490202528018e-05, "loss": 1.003, "step": 14256 }, { "epoch": 1.43, "learning_rate": 4.262273104705127e-05, "loss": 0.3854, "step": 14257 }, { "epoch": 1.43, "learning_rate": 4.2620559988496114e-05, "loss": 0.2779, "step": 14258 }, { "epoch": 1.43, "learning_rate": 4.2618388849628506e-05, "loss": 0.8626, "step": 14259 }, { "epoch": 1.44, "learning_rate": 4.2616217630462276e-05, "loss": 1.1224, "step": 14260 }, { "epoch": 1.44, "learning_rate": 4.261404633101124e-05, "loss": 0.2235, "step": 14261 }, { "epoch": 1.44, "learning_rate": 4.26118749512892e-05, "loss": 0.682, "step": 14262 }, { "epoch": 1.44, "learning_rate": 4.260970349130999e-05, "loss": 0.8114, "step": 14263 }, { "epoch": 1.44, "learning_rate": 4.260753195108742e-05, "loss": 0.5435, "step": 14264 }, { "epoch": 1.44, "learning_rate": 4.260536033063532e-05, "loss": 0.6552, "step": 14265 }, { "epoch": 1.44, "learning_rate": 4.2603188629967506e-05, "loss": 1.2775, "step": 14266 }, { "epoch": 1.44, "learning_rate": 4.260101684909778e-05, "loss": 1.0741, "step": 14267 }, { "epoch": 1.44, "learning_rate": 4.2598844988039996e-05, "loss": 0.689, "step": 14268 }, { "epoch": 1.44, "learning_rate": 4.259667304680794e-05, "loss": 0.9807, "step": 14269 }, { "epoch": 1.44, "learning_rate": 4.259450102541545e-05, "loss": 0.6079, "step": 14270 }, { "epoch": 1.44, "learning_rate": 4.259232892387635e-05, "loss": 0.3446, "step": 14271 }, { "epoch": 1.44, "learning_rate": 4.2590156742204465e-05, "loss": 0.5009, "step": 14272 }, { "epoch": 1.44, "learning_rate": 4.25879844804136e-05, "loss": 0.9101, "step": 14273 }, { "epoch": 1.44, "learning_rate": 4.25858121385176e-05, "loss": 0.9964, "step": 14274 }, { "epoch": 1.44, "learning_rate": 4.258363971653027e-05, "loss": 0.5108, "step": 14275 }, { "epoch": 1.44, "learning_rate": 4.258146721446545e-05, "loss": 0.578, "step": 14276 }, { "epoch": 1.44, "learning_rate": 4.257929463233697e-05, "loss": 0.4048, "step": 14277 }, { "epoch": 1.44, "learning_rate": 4.2577121970158626e-05, "loss": 0.2686, "step": 14278 }, { "epoch": 1.44, "learning_rate": 4.257494922794427e-05, "loss": 0.5516, "step": 14279 }, { "epoch": 1.44, "learning_rate": 4.2572776405707715e-05, "loss": 0.5973, "step": 14280 }, { "epoch": 1.44, "learning_rate": 4.257060350346281e-05, "loss": 0.2001, "step": 14281 }, { "epoch": 1.44, "learning_rate": 4.256843052122335e-05, "loss": 0.3694, "step": 14282 }, { "epoch": 1.44, "learning_rate": 4.256625745900318e-05, "loss": 0.6159, "step": 14283 }, { "epoch": 1.44, "learning_rate": 4.2564084316816144e-05, "loss": 0.419, "step": 14284 }, { "epoch": 1.44, "learning_rate": 4.256191109467604e-05, "loss": 1.2876, "step": 14285 }, { "epoch": 1.44, "learning_rate": 4.2559737792596714e-05, "loss": 1.1634, "step": 14286 }, { "epoch": 1.44, "learning_rate": 4.255756441059199e-05, "loss": 0.4654, "step": 14287 }, { "epoch": 1.44, "learning_rate": 4.255539094867572e-05, "loss": 0.9261, "step": 14288 }, { "epoch": 1.44, "learning_rate": 4.25532174068617e-05, "loss": 0.4165, "step": 14289 }, { "epoch": 1.44, "learning_rate": 4.25510437851638e-05, "loss": 0.4663, "step": 14290 }, { "epoch": 1.44, "learning_rate": 4.2548870083595816e-05, "loss": 1.3476, "step": 14291 }, { "epoch": 1.44, "learning_rate": 4.25466963021716e-05, "loss": 1.1754, "step": 14292 }, { "epoch": 1.44, "learning_rate": 4.254452244090498e-05, "loss": 0.8381, "step": 14293 }, { "epoch": 1.44, "learning_rate": 4.25423484998098e-05, "loss": 0.6592, "step": 14294 }, { "epoch": 1.44, "learning_rate": 4.254017447889988e-05, "loss": 0.3366, "step": 14295 }, { "epoch": 1.44, "learning_rate": 4.2538000378189056e-05, "loss": 0.7319, "step": 14296 }, { "epoch": 1.44, "learning_rate": 4.253582619769117e-05, "loss": 1.0199, "step": 14297 }, { "epoch": 1.44, "learning_rate": 4.253365193742006e-05, "loss": 0.3674, "step": 14298 }, { "epoch": 1.44, "learning_rate": 4.253147759738955e-05, "loss": 0.6548, "step": 14299 }, { "epoch": 1.44, "learning_rate": 4.252930317761349e-05, "loss": 0.3404, "step": 14300 }, { "epoch": 1.44, "learning_rate": 4.25271286781057e-05, "loss": 1.7323, "step": 14301 }, { "epoch": 1.44, "learning_rate": 4.252495409888004e-05, "loss": 0.7059, "step": 14302 }, { "epoch": 1.44, "learning_rate": 4.252277943995034e-05, "loss": 0.4615, "step": 14303 }, { "epoch": 1.44, "learning_rate": 4.252060470133043e-05, "loss": 0.6643, "step": 14304 }, { "epoch": 1.44, "learning_rate": 4.2518429883034154e-05, "loss": 0.8602, "step": 14305 }, { "epoch": 1.44, "learning_rate": 4.251625498507535e-05, "loss": 0.8145, "step": 14306 }, { "epoch": 1.44, "learning_rate": 4.251408000746786e-05, "loss": 0.9271, "step": 14307 }, { "epoch": 1.44, "learning_rate": 4.251190495022553e-05, "loss": 0.5312, "step": 14308 }, { "epoch": 1.44, "learning_rate": 4.25097298133622e-05, "loss": 0.929, "step": 14309 }, { "epoch": 1.44, "learning_rate": 4.25075545968917e-05, "loss": 0.2407, "step": 14310 }, { "epoch": 1.44, "learning_rate": 4.250537930082789e-05, "loss": 0.5134, "step": 14311 }, { "epoch": 1.44, "learning_rate": 4.250320392518459e-05, "loss": 1.2212, "step": 14312 }, { "epoch": 1.44, "learning_rate": 4.250102846997567e-05, "loss": 0.2435, "step": 14313 }, { "epoch": 1.44, "learning_rate": 4.249885293521495e-05, "loss": 1.0205, "step": 14314 }, { "epoch": 1.44, "learning_rate": 4.2496677320916294e-05, "loss": 0.4113, "step": 14315 }, { "epoch": 1.44, "learning_rate": 4.249450162709354e-05, "loss": 0.868, "step": 14316 }, { "epoch": 1.44, "learning_rate": 4.249232585376052e-05, "loss": 0.6087, "step": 14317 }, { "epoch": 1.44, "learning_rate": 4.2490150000931084e-05, "loss": 0.2612, "step": 14318 }, { "epoch": 1.44, "learning_rate": 4.2487974068619094e-05, "loss": 0.3888, "step": 14319 }, { "epoch": 1.44, "learning_rate": 4.24857980568384e-05, "loss": 0.833, "step": 14320 }, { "epoch": 1.44, "learning_rate": 4.248362196560282e-05, "loss": 1.2756, "step": 14321 }, { "epoch": 1.44, "learning_rate": 4.2481445794926224e-05, "loss": 0.3811, "step": 14322 }, { "epoch": 1.44, "learning_rate": 4.2479269544822454e-05, "loss": 0.8499, "step": 14323 }, { "epoch": 1.44, "learning_rate": 4.247709321530537e-05, "loss": 0.4215, "step": 14324 }, { "epoch": 1.44, "learning_rate": 4.24749168063888e-05, "loss": 0.971, "step": 14325 }, { "epoch": 1.44, "learning_rate": 4.247274031808661e-05, "loss": 0.1193, "step": 14326 }, { "epoch": 1.44, "learning_rate": 4.247056375041264e-05, "loss": 0.6195, "step": 14327 }, { "epoch": 1.44, "learning_rate": 4.246838710338075e-05, "loss": 0.5852, "step": 14328 }, { "epoch": 1.44, "learning_rate": 4.2466210377004794e-05, "loss": 0.933, "step": 14329 }, { "epoch": 1.44, "learning_rate": 4.2464033571298606e-05, "loss": 1.0638, "step": 14330 }, { "epoch": 1.44, "learning_rate": 4.246185668627606e-05, "loss": 0.6056, "step": 14331 }, { "epoch": 1.44, "learning_rate": 4.2459679721950994e-05, "loss": 0.1989, "step": 14332 }, { "epoch": 1.44, "learning_rate": 4.2457502678337266e-05, "loss": 1.335, "step": 14333 }, { "epoch": 1.44, "learning_rate": 4.245532555544873e-05, "loss": 0.2915, "step": 14334 }, { "epoch": 1.44, "learning_rate": 4.245314835329924e-05, "loss": 0.6519, "step": 14335 }, { "epoch": 1.44, "learning_rate": 4.245097107190266e-05, "loss": 1.6552, "step": 14336 }, { "epoch": 1.44, "learning_rate": 4.244879371127282e-05, "loss": 0.358, "step": 14337 }, { "epoch": 1.44, "learning_rate": 4.24466162714236e-05, "loss": 1.3355, "step": 14338 }, { "epoch": 1.44, "learning_rate": 4.244443875236885e-05, "loss": 0.4254, "step": 14339 }, { "epoch": 1.44, "learning_rate": 4.244226115412243e-05, "loss": 0.2911, "step": 14340 }, { "epoch": 1.44, "learning_rate": 4.244008347669819e-05, "loss": 0.5794, "step": 14341 }, { "epoch": 1.44, "learning_rate": 4.243790572010999e-05, "loss": 0.6404, "step": 14342 }, { "epoch": 1.44, "learning_rate": 4.2435727884371696e-05, "loss": 0.6878, "step": 14343 }, { "epoch": 1.44, "learning_rate": 4.243354996949715e-05, "loss": 1.0223, "step": 14344 }, { "epoch": 1.44, "learning_rate": 4.243137197550023e-05, "loss": 1.3655, "step": 14345 }, { "epoch": 1.44, "learning_rate": 4.242919390239478e-05, "loss": 0.334, "step": 14346 }, { "epoch": 1.44, "learning_rate": 4.242701575019467e-05, "loss": 0.9114, "step": 14347 }, { "epoch": 1.44, "learning_rate": 4.2424837518913765e-05, "loss": 0.9222, "step": 14348 }, { "epoch": 1.44, "learning_rate": 4.2422659208565916e-05, "loss": 0.4291, "step": 14349 }, { "epoch": 1.44, "learning_rate": 4.2420480819165e-05, "loss": 0.2505, "step": 14350 }, { "epoch": 1.44, "learning_rate": 4.2418302350724846e-05, "loss": 0.3916, "step": 14351 }, { "epoch": 1.44, "learning_rate": 4.241612380325937e-05, "loss": 0.6922, "step": 14352 }, { "epoch": 1.44, "learning_rate": 4.241394517678238e-05, "loss": 0.5161, "step": 14353 }, { "epoch": 1.44, "learning_rate": 4.241176647130778e-05, "loss": 0.6176, "step": 14354 }, { "epoch": 1.44, "learning_rate": 4.2409587686849414e-05, "loss": 0.3837, "step": 14355 }, { "epoch": 1.44, "learning_rate": 4.240740882342116e-05, "loss": 0.462, "step": 14356 }, { "epoch": 1.44, "learning_rate": 4.2405229881036866e-05, "loss": 0.8083, "step": 14357 }, { "epoch": 1.44, "learning_rate": 4.240305085971042e-05, "loss": 0.4998, "step": 14358 }, { "epoch": 1.45, "learning_rate": 4.2400871759455666e-05, "loss": 0.9548, "step": 14359 }, { "epoch": 1.45, "learning_rate": 4.239869258028647e-05, "loss": 0.8883, "step": 14360 }, { "epoch": 1.45, "learning_rate": 4.2396513322216736e-05, "loss": 0.986, "step": 14361 }, { "epoch": 1.45, "learning_rate": 4.239433398526029e-05, "loss": 0.1102, "step": 14362 }, { "epoch": 1.45, "learning_rate": 4.2392154569431036e-05, "loss": 1.0185, "step": 14363 }, { "epoch": 1.45, "learning_rate": 4.238997507474281e-05, "loss": 0.2632, "step": 14364 }, { "epoch": 1.45, "learning_rate": 4.23877955012095e-05, "loss": 0.6253, "step": 14365 }, { "epoch": 1.45, "learning_rate": 4.238561584884498e-05, "loss": 0.3864, "step": 14366 }, { "epoch": 1.45, "learning_rate": 4.238343611766309e-05, "loss": 0.7669, "step": 14367 }, { "epoch": 1.45, "learning_rate": 4.2381256307677745e-05, "loss": 0.6113, "step": 14368 }, { "epoch": 1.45, "learning_rate": 4.237907641890278e-05, "loss": 0.2068, "step": 14369 }, { "epoch": 1.45, "learning_rate": 4.23768964513521e-05, "loss": 0.4231, "step": 14370 }, { "epoch": 1.45, "learning_rate": 4.2374716405039544e-05, "loss": 1.0284, "step": 14371 }, { "epoch": 1.45, "learning_rate": 4.237253627997901e-05, "loss": 0.8592, "step": 14372 }, { "epoch": 1.45, "learning_rate": 4.237035607618435e-05, "loss": 1.2017, "step": 14373 }, { "epoch": 1.45, "learning_rate": 4.2368175793669455e-05, "loss": 0.7981, "step": 14374 }, { "epoch": 1.45, "learning_rate": 4.23659954324482e-05, "loss": 0.4903, "step": 14375 }, { "epoch": 1.45, "learning_rate": 4.236381499253445e-05, "loss": 1.0115, "step": 14376 }, { "epoch": 1.45, "learning_rate": 4.2361634473942084e-05, "loss": 0.1606, "step": 14377 }, { "epoch": 1.45, "learning_rate": 4.235945387668498e-05, "loss": 0.266, "step": 14378 }, { "epoch": 1.45, "learning_rate": 4.235727320077701e-05, "loss": 1.2207, "step": 14379 }, { "epoch": 1.45, "learning_rate": 4.235509244623206e-05, "loss": 0.9412, "step": 14380 }, { "epoch": 1.45, "learning_rate": 4.2352911613064005e-05, "loss": 1.0144, "step": 14381 }, { "epoch": 1.45, "learning_rate": 4.235073070128671e-05, "loss": 0.8474, "step": 14382 }, { "epoch": 1.45, "learning_rate": 4.2348549710914065e-05, "loss": 0.8285, "step": 14383 }, { "epoch": 1.45, "learning_rate": 4.234636864195996e-05, "loss": 0.4709, "step": 14384 }, { "epoch": 1.45, "learning_rate": 4.234418749443824e-05, "loss": 1.2988, "step": 14385 }, { "epoch": 1.45, "learning_rate": 4.2342006268362825e-05, "loss": 0.1959, "step": 14386 }, { "epoch": 1.45, "learning_rate": 4.233982496374757e-05, "loss": 0.252, "step": 14387 }, { "epoch": 1.45, "learning_rate": 4.233764358060636e-05, "loss": 0.7956, "step": 14388 }, { "epoch": 1.45, "learning_rate": 4.233546211895308e-05, "loss": 0.3308, "step": 14389 }, { "epoch": 1.45, "learning_rate": 4.2333280578801616e-05, "loss": 0.3627, "step": 14390 }, { "epoch": 1.45, "learning_rate": 4.233109896016585e-05, "loss": 1.8906, "step": 14391 }, { "epoch": 1.45, "learning_rate": 4.2328917263059655e-05, "loss": 0.9052, "step": 14392 }, { "epoch": 1.45, "learning_rate": 4.232673548749692e-05, "loss": 0.4976, "step": 14393 }, { "epoch": 1.45, "learning_rate": 4.232455363349153e-05, "loss": 0.3003, "step": 14394 }, { "epoch": 1.45, "learning_rate": 4.232237170105738e-05, "loss": 1.0355, "step": 14395 }, { "epoch": 1.45, "learning_rate": 4.232018969020833e-05, "loss": 0.6349, "step": 14396 }, { "epoch": 1.45, "learning_rate": 4.2318007600958285e-05, "loss": 0.5159, "step": 14397 }, { "epoch": 1.45, "learning_rate": 4.231582543332113e-05, "loss": 0.74, "step": 14398 }, { "epoch": 1.45, "learning_rate": 4.231364318731074e-05, "loss": 0.5414, "step": 14399 }, { "epoch": 1.45, "learning_rate": 4.2311460862941006e-05, "loss": 0.5891, "step": 14400 }, { "epoch": 1.45, "learning_rate": 4.230927846022583e-05, "loss": 0.2068, "step": 14401 }, { "epoch": 1.45, "learning_rate": 4.230709597917908e-05, "loss": 0.1323, "step": 14402 }, { "epoch": 1.45, "learning_rate": 4.230491341981466e-05, "loss": 0.6445, "step": 14403 }, { "epoch": 1.45, "learning_rate": 4.230273078214645e-05, "loss": 0.2445, "step": 14404 }, { "epoch": 1.45, "learning_rate": 4.230054806618834e-05, "loss": 0.3898, "step": 14405 }, { "epoch": 1.45, "learning_rate": 4.229836527195422e-05, "loss": 0.9194, "step": 14406 }, { "epoch": 1.45, "learning_rate": 4.229618239945798e-05, "loss": 0.7557, "step": 14407 }, { "epoch": 1.45, "learning_rate": 4.229399944871352e-05, "loss": 0.5318, "step": 14408 }, { "epoch": 1.45, "learning_rate": 4.229181641973472e-05, "loss": 1.1238, "step": 14409 }, { "epoch": 1.45, "learning_rate": 4.228963331253547e-05, "loss": 0.5374, "step": 14410 }, { "epoch": 1.45, "learning_rate": 4.228745012712968e-05, "loss": 0.4416, "step": 14411 }, { "epoch": 1.45, "learning_rate": 4.2285266863531225e-05, "loss": 0.5406, "step": 14412 }, { "epoch": 1.45, "learning_rate": 4.228308352175401e-05, "loss": 0.7671, "step": 14413 }, { "epoch": 1.45, "learning_rate": 4.228090010181192e-05, "loss": 0.4925, "step": 14414 }, { "epoch": 1.45, "learning_rate": 4.227871660371885e-05, "loss": 0.93, "step": 14415 }, { "epoch": 1.45, "learning_rate": 4.227653302748871e-05, "loss": 0.7175, "step": 14416 }, { "epoch": 1.45, "learning_rate": 4.227434937313538e-05, "loss": 0.6648, "step": 14417 }, { "epoch": 1.45, "learning_rate": 4.227216564067275e-05, "loss": 0.1732, "step": 14418 }, { "epoch": 1.45, "learning_rate": 4.226998183011474e-05, "loss": 1.0456, "step": 14419 }, { "epoch": 1.45, "learning_rate": 4.226779794147523e-05, "loss": 0.7115, "step": 14420 }, { "epoch": 1.45, "learning_rate": 4.226561397476812e-05, "loss": 0.6817, "step": 14421 }, { "epoch": 1.45, "learning_rate": 4.22634299300073e-05, "loss": 0.2791, "step": 14422 }, { "epoch": 1.45, "learning_rate": 4.226124580720669e-05, "loss": 0.3429, "step": 14423 }, { "epoch": 1.45, "learning_rate": 4.225906160638017e-05, "loss": 1.1953, "step": 14424 }, { "epoch": 1.45, "learning_rate": 4.225687732754164e-05, "loss": 0.8164, "step": 14425 }, { "epoch": 1.45, "learning_rate": 4.2254692970705015e-05, "loss": 0.597, "step": 14426 }, { "epoch": 1.45, "learning_rate": 4.225250853588419e-05, "loss": 0.168, "step": 14427 }, { "epoch": 1.45, "learning_rate": 4.225032402309305e-05, "loss": 0.2782, "step": 14428 }, { "epoch": 1.45, "learning_rate": 4.2248139432345504e-05, "loss": 0.2666, "step": 14429 }, { "epoch": 1.45, "learning_rate": 4.224595476365547e-05, "loss": 0.9992, "step": 14430 }, { "epoch": 1.45, "learning_rate": 4.224377001703683e-05, "loss": 0.9871, "step": 14431 }, { "epoch": 1.45, "learning_rate": 4.224158519250351e-05, "loss": 1.0655, "step": 14432 }, { "epoch": 1.45, "learning_rate": 4.223940029006938e-05, "loss": 0.3289, "step": 14433 }, { "epoch": 1.45, "learning_rate": 4.223721530974838e-05, "loss": 0.6316, "step": 14434 }, { "epoch": 1.45, "learning_rate": 4.2235030251554386e-05, "loss": 0.2161, "step": 14435 }, { "epoch": 1.45, "learning_rate": 4.223284511550132e-05, "loss": 0.2422, "step": 14436 }, { "epoch": 1.45, "learning_rate": 4.2230659901603075e-05, "loss": 0.1389, "step": 14437 }, { "epoch": 1.45, "learning_rate": 4.222847460987356e-05, "loss": 0.6236, "step": 14438 }, { "epoch": 1.45, "learning_rate": 4.22262892403267e-05, "loss": 1.1174, "step": 14439 }, { "epoch": 1.45, "learning_rate": 4.222410379297637e-05, "loss": 0.8251, "step": 14440 }, { "epoch": 1.45, "learning_rate": 4.2221918267836506e-05, "loss": 0.9139, "step": 14441 }, { "epoch": 1.45, "learning_rate": 4.221973266492099e-05, "loss": 1.0659, "step": 14442 }, { "epoch": 1.45, "learning_rate": 4.221754698424376e-05, "loss": 1.257, "step": 14443 }, { "epoch": 1.45, "learning_rate": 4.221536122581871e-05, "loss": 1.1537, "step": 14444 }, { "epoch": 1.45, "learning_rate": 4.221317538965972e-05, "loss": 0.3859, "step": 14445 }, { "epoch": 1.45, "learning_rate": 4.221098947578076e-05, "loss": 0.6698, "step": 14446 }, { "epoch": 1.45, "learning_rate": 4.220880348419569e-05, "loss": 1.0568, "step": 14447 }, { "epoch": 1.45, "learning_rate": 4.220661741491845e-05, "loss": 0.141, "step": 14448 }, { "epoch": 1.45, "learning_rate": 4.220443126796293e-05, "loss": 1.1063, "step": 14449 }, { "epoch": 1.45, "learning_rate": 4.220224504334306e-05, "loss": 0.8268, "step": 14450 }, { "epoch": 1.45, "learning_rate": 4.2200058741072745e-05, "loss": 0.7303, "step": 14451 }, { "epoch": 1.45, "learning_rate": 4.2197872361165885e-05, "loss": 0.398, "step": 14452 }, { "epoch": 1.45, "learning_rate": 4.219568590363642e-05, "loss": 0.6257, "step": 14453 }, { "epoch": 1.45, "learning_rate": 4.219349936849824e-05, "loss": 0.1047, "step": 14454 }, { "epoch": 1.45, "learning_rate": 4.219131275576527e-05, "loss": 1.331, "step": 14455 }, { "epoch": 1.45, "learning_rate": 4.218912606545143e-05, "loss": 0.8652, "step": 14456 }, { "epoch": 1.46, "learning_rate": 4.218693929757063e-05, "loss": 0.5341, "step": 14457 }, { "epoch": 1.46, "learning_rate": 4.2184752452136773e-05, "loss": 0.3395, "step": 14458 }, { "epoch": 1.46, "learning_rate": 4.21825655291638e-05, "loss": 0.296, "step": 14459 }, { "epoch": 1.46, "learning_rate": 4.2180378528665606e-05, "loss": 0.7306, "step": 14460 }, { "epoch": 1.46, "learning_rate": 4.217819145065611e-05, "loss": 0.509, "step": 14461 }, { "epoch": 1.46, "learning_rate": 4.2176004295149256e-05, "loss": 0.7254, "step": 14462 }, { "epoch": 1.46, "learning_rate": 4.217381706215892e-05, "loss": 1.0723, "step": 14463 }, { "epoch": 1.46, "learning_rate": 4.217162975169906e-05, "loss": 0.8287, "step": 14464 }, { "epoch": 1.46, "learning_rate": 4.216944236378357e-05, "loss": 0.6718, "step": 14465 }, { "epoch": 1.46, "learning_rate": 4.2167254898426386e-05, "loss": 0.4939, "step": 14466 }, { "epoch": 1.46, "learning_rate": 4.216506735564141e-05, "loss": 1.0587, "step": 14467 }, { "epoch": 1.46, "learning_rate": 4.216287973544258e-05, "loss": 1.1182, "step": 14468 }, { "epoch": 1.46, "learning_rate": 4.2160692037843815e-05, "loss": 0.4046, "step": 14469 }, { "epoch": 1.46, "learning_rate": 4.215850426285902e-05, "loss": 0.9869, "step": 14470 }, { "epoch": 1.46, "learning_rate": 4.215631641050214e-05, "loss": 0.6443, "step": 14471 }, { "epoch": 1.46, "learning_rate": 4.215412848078708e-05, "loss": 0.3821, "step": 14472 }, { "epoch": 1.46, "learning_rate": 4.2151940473727775e-05, "loss": 0.6589, "step": 14473 }, { "epoch": 1.46, "learning_rate": 4.214975238933814e-05, "loss": 0.5248, "step": 14474 }, { "epoch": 1.46, "learning_rate": 4.2147564227632106e-05, "loss": 1.0704, "step": 14475 }, { "epoch": 1.46, "learning_rate": 4.21453759886236e-05, "loss": 0.3708, "step": 14476 }, { "epoch": 1.46, "learning_rate": 4.2143187672326525e-05, "loss": 0.8544, "step": 14477 }, { "epoch": 1.46, "learning_rate": 4.2140999278754835e-05, "loss": 0.633, "step": 14478 }, { "epoch": 1.46, "learning_rate": 4.213881080792244e-05, "loss": 0.8745, "step": 14479 }, { "epoch": 1.46, "learning_rate": 4.2136622259843266e-05, "loss": 0.7397, "step": 14480 }, { "epoch": 1.46, "learning_rate": 4.213443363453125e-05, "loss": 0.8067, "step": 14481 }, { "epoch": 1.46, "learning_rate": 4.213224493200032e-05, "loss": 0.3483, "step": 14482 }, { "epoch": 1.46, "learning_rate": 4.2130056152264396e-05, "loss": 0.5251, "step": 14483 }, { "epoch": 1.46, "learning_rate": 4.212786729533741e-05, "loss": 0.4926, "step": 14484 }, { "epoch": 1.46, "learning_rate": 4.212567836123329e-05, "loss": 0.3742, "step": 14485 }, { "epoch": 1.46, "learning_rate": 4.2123489349965954e-05, "loss": 0.3828, "step": 14486 }, { "epoch": 1.46, "learning_rate": 4.2121300261549354e-05, "loss": 0.7395, "step": 14487 }, { "epoch": 1.46, "learning_rate": 4.211911109599741e-05, "loss": 0.8669, "step": 14488 }, { "epoch": 1.46, "learning_rate": 4.211692185332406e-05, "loss": 0.4313, "step": 14489 }, { "epoch": 1.46, "learning_rate": 4.211473253354322e-05, "loss": 0.8317, "step": 14490 }, { "epoch": 1.46, "learning_rate": 4.211254313666883e-05, "loss": 0.882, "step": 14491 }, { "epoch": 1.46, "learning_rate": 4.211035366271483e-05, "loss": 0.8312, "step": 14492 }, { "epoch": 1.46, "learning_rate": 4.210816411169514e-05, "loss": 0.7207, "step": 14493 }, { "epoch": 1.46, "learning_rate": 4.21059744836237e-05, "loss": 0.6218, "step": 14494 }, { "epoch": 1.46, "learning_rate": 4.2103784778514444e-05, "loss": 0.9248, "step": 14495 }, { "epoch": 1.46, "learning_rate": 4.210159499638131e-05, "loss": 0.328, "step": 14496 }, { "epoch": 1.46, "learning_rate": 4.209940513723823e-05, "loss": 0.9202, "step": 14497 }, { "epoch": 1.46, "learning_rate": 4.2097215201099135e-05, "loss": 1.0532, "step": 14498 }, { "epoch": 1.46, "learning_rate": 4.209502518797797e-05, "loss": 0.33, "step": 14499 }, { "epoch": 1.46, "learning_rate": 4.209283509788866e-05, "loss": 0.927, "step": 14500 }, { "epoch": 1.46, "learning_rate": 4.209064493084514e-05, "loss": 0.1841, "step": 14501 }, { "epoch": 1.46, "learning_rate": 4.2088454686861364e-05, "loss": 1.0283, "step": 14502 }, { "epoch": 1.46, "learning_rate": 4.2086264365951266e-05, "loss": 0.8707, "step": 14503 }, { "epoch": 1.46, "learning_rate": 4.2084073968128774e-05, "loss": 0.2364, "step": 14504 }, { "epoch": 1.46, "learning_rate": 4.2081883493407826e-05, "loss": 0.3643, "step": 14505 }, { "epoch": 1.46, "learning_rate": 4.207969294180238e-05, "loss": 0.6989, "step": 14506 }, { "epoch": 1.46, "learning_rate": 4.2077502313326354e-05, "loss": 0.3618, "step": 14507 }, { "epoch": 1.46, "learning_rate": 4.2075311607993714e-05, "loss": 0.5411, "step": 14508 }, { "epoch": 1.46, "learning_rate": 4.207312082581836e-05, "loss": 0.1642, "step": 14509 }, { "epoch": 1.46, "learning_rate": 4.207092996681427e-05, "loss": 0.6212, "step": 14510 }, { "epoch": 1.46, "learning_rate": 4.206873903099538e-05, "loss": 1.5109, "step": 14511 }, { "epoch": 1.46, "learning_rate": 4.206654801837562e-05, "loss": 0.7807, "step": 14512 }, { "epoch": 1.46, "learning_rate": 4.2064356928968934e-05, "loss": 0.4565, "step": 14513 }, { "epoch": 1.46, "learning_rate": 4.206216576278928e-05, "loss": 0.1912, "step": 14514 }, { "epoch": 1.46, "learning_rate": 4.205997451985059e-05, "loss": 0.9746, "step": 14515 }, { "epoch": 1.46, "learning_rate": 4.20577832001668e-05, "loss": 0.6294, "step": 14516 }, { "epoch": 1.46, "learning_rate": 4.2055591803751874e-05, "loss": 0.2779, "step": 14517 }, { "epoch": 1.46, "learning_rate": 4.205340033061974e-05, "loss": 0.8739, "step": 14518 }, { "epoch": 1.46, "learning_rate": 4.2051208780784356e-05, "loss": 0.3218, "step": 14519 }, { "epoch": 1.46, "learning_rate": 4.2049017154259665e-05, "loss": 0.2576, "step": 14520 }, { "epoch": 1.46, "learning_rate": 4.2046825451059614e-05, "loss": 0.5762, "step": 14521 }, { "epoch": 1.46, "learning_rate": 4.204463367119815e-05, "loss": 0.292, "step": 14522 }, { "epoch": 1.46, "learning_rate": 4.2042441814689206e-05, "loss": 0.5156, "step": 14523 }, { "epoch": 1.46, "learning_rate": 4.204024988154676e-05, "loss": 0.9533, "step": 14524 }, { "epoch": 1.46, "learning_rate": 4.203805787178473e-05, "loss": 0.3825, "step": 14525 }, { "epoch": 1.46, "learning_rate": 4.2035865785417093e-05, "loss": 1.0336, "step": 14526 }, { "epoch": 1.46, "learning_rate": 4.203367362245778e-05, "loss": 0.4201, "step": 14527 }, { "epoch": 1.46, "learning_rate": 4.203148138292074e-05, "loss": 0.6984, "step": 14528 }, { "epoch": 1.46, "learning_rate": 4.202928906681993e-05, "loss": 0.1929, "step": 14529 }, { "epoch": 1.46, "learning_rate": 4.202709667416931e-05, "loss": 1.0522, "step": 14530 }, { "epoch": 1.46, "learning_rate": 4.202490420498282e-05, "loss": 1.1225, "step": 14531 }, { "epoch": 1.46, "learning_rate": 4.20227116592744e-05, "loss": 0.3512, "step": 14532 }, { "epoch": 1.46, "learning_rate": 4.2020519037058044e-05, "loss": 0.9742, "step": 14533 }, { "epoch": 1.46, "learning_rate": 4.201832633834765e-05, "loss": 0.4479, "step": 14534 }, { "epoch": 1.46, "learning_rate": 4.201613356315721e-05, "loss": 0.1944, "step": 14535 }, { "epoch": 1.46, "learning_rate": 4.201394071150067e-05, "loss": 0.4127, "step": 14536 }, { "epoch": 1.46, "learning_rate": 4.2011747783391986e-05, "loss": 1.1164, "step": 14537 }, { "epoch": 1.46, "learning_rate": 4.20095547788451e-05, "loss": 0.6717, "step": 14538 }, { "epoch": 1.46, "learning_rate": 4.2007361697873976e-05, "loss": 0.8052, "step": 14539 }, { "epoch": 1.46, "learning_rate": 4.200516854049257e-05, "loss": 0.6467, "step": 14540 }, { "epoch": 1.46, "learning_rate": 4.200297530671484e-05, "loss": 1.0717, "step": 14541 }, { "epoch": 1.46, "learning_rate": 4.200078199655474e-05, "loss": 1.0334, "step": 14542 }, { "epoch": 1.46, "learning_rate": 4.199858861002623e-05, "loss": 0.7715, "step": 14543 }, { "epoch": 1.46, "learning_rate": 4.199639514714327e-05, "loss": 0.6095, "step": 14544 }, { "epoch": 1.46, "learning_rate": 4.199420160791981e-05, "loss": 1.1417, "step": 14545 }, { "epoch": 1.46, "learning_rate": 4.199200799236982e-05, "loss": 0.4649, "step": 14546 }, { "epoch": 1.46, "learning_rate": 4.198981430050725e-05, "loss": 0.7044, "step": 14547 }, { "epoch": 1.46, "learning_rate": 4.198762053234605e-05, "loss": 0.6178, "step": 14548 }, { "epoch": 1.46, "learning_rate": 4.1985426687900214e-05, "loss": 0.7085, "step": 14549 }, { "epoch": 1.46, "learning_rate": 4.198323276718367e-05, "loss": 0.9822, "step": 14550 }, { "epoch": 1.46, "learning_rate": 4.1981038770210394e-05, "loss": 0.8293, "step": 14551 }, { "epoch": 1.46, "learning_rate": 4.1978844696994355e-05, "loss": 0.3443, "step": 14552 }, { "epoch": 1.46, "learning_rate": 4.197665054754949e-05, "loss": 0.4529, "step": 14553 }, { "epoch": 1.46, "learning_rate": 4.197445632188978e-05, "loss": 0.948, "step": 14554 }, { "epoch": 1.46, "learning_rate": 4.1972262020029186e-05, "loss": 0.2793, "step": 14555 }, { "epoch": 1.47, "learning_rate": 4.197006764198168e-05, "loss": 0.5517, "step": 14556 }, { "epoch": 1.47, "learning_rate": 4.1967873187761205e-05, "loss": 0.5753, "step": 14557 }, { "epoch": 1.47, "learning_rate": 4.196567865738174e-05, "loss": 0.7292, "step": 14558 }, { "epoch": 1.47, "learning_rate": 4.1963484050857256e-05, "loss": 0.3045, "step": 14559 }, { "epoch": 1.47, "learning_rate": 4.19612893682017e-05, "loss": 0.5195, "step": 14560 }, { "epoch": 1.47, "learning_rate": 4.195909460942906e-05, "loss": 0.636, "step": 14561 }, { "epoch": 1.47, "learning_rate": 4.195689977455328e-05, "loss": 0.3044, "step": 14562 }, { "epoch": 1.47, "learning_rate": 4.195470486358835e-05, "loss": 1.1549, "step": 14563 }, { "epoch": 1.47, "learning_rate": 4.195250987654822e-05, "loss": 0.2839, "step": 14564 }, { "epoch": 1.47, "learning_rate": 4.195031481344686e-05, "loss": 0.9346, "step": 14565 }, { "epoch": 1.47, "learning_rate": 4.194811967429825e-05, "loss": 0.1948, "step": 14566 }, { "epoch": 1.47, "learning_rate": 4.194592445911635e-05, "loss": 0.3768, "step": 14567 }, { "epoch": 1.47, "learning_rate": 4.194372916791513e-05, "loss": 0.3116, "step": 14568 }, { "epoch": 1.47, "learning_rate": 4.194153380070857e-05, "loss": 0.6761, "step": 14569 }, { "epoch": 1.47, "learning_rate": 4.193933835751062e-05, "loss": 0.6235, "step": 14570 }, { "epoch": 1.47, "learning_rate": 4.193714283833526e-05, "loss": 0.4052, "step": 14571 }, { "epoch": 1.47, "learning_rate": 4.1934947243196474e-05, "loss": 0.9054, "step": 14572 }, { "epoch": 1.47, "learning_rate": 4.1932751572108216e-05, "loss": 0.2891, "step": 14573 }, { "epoch": 1.47, "learning_rate": 4.193055582508447e-05, "loss": 0.6084, "step": 14574 }, { "epoch": 1.47, "learning_rate": 4.1928360002139206e-05, "loss": 0.4943, "step": 14575 }, { "epoch": 1.47, "learning_rate": 4.19261641032864e-05, "loss": 1.1261, "step": 14576 }, { "epoch": 1.47, "learning_rate": 4.1923968128540014e-05, "loss": 1.2005, "step": 14577 }, { "epoch": 1.47, "learning_rate": 4.192177207791403e-05, "loss": 0.1504, "step": 14578 }, { "epoch": 1.47, "learning_rate": 4.191957595142244e-05, "loss": 1.3503, "step": 14579 }, { "epoch": 1.47, "learning_rate": 4.1917379749079186e-05, "loss": 0.2855, "step": 14580 }, { "epoch": 1.47, "learning_rate": 4.1915183470898265e-05, "loss": 0.4617, "step": 14581 }, { "epoch": 1.47, "learning_rate": 4.191298711689366e-05, "loss": 0.8514, "step": 14582 }, { "epoch": 1.47, "learning_rate": 4.1910790687079325e-05, "loss": 0.4697, "step": 14583 }, { "epoch": 1.47, "learning_rate": 4.190859418146926e-05, "loss": 1.4712, "step": 14584 }, { "epoch": 1.47, "learning_rate": 4.1906397600077416e-05, "loss": 1.2037, "step": 14585 }, { "epoch": 1.47, "learning_rate": 4.19042009429178e-05, "loss": 0.2665, "step": 14586 }, { "epoch": 1.47, "learning_rate": 4.190200421000437e-05, "loss": 0.8215, "step": 14587 }, { "epoch": 1.47, "learning_rate": 4.189980740135112e-05, "loss": 0.5173, "step": 14588 }, { "epoch": 1.47, "learning_rate": 4.189761051697202e-05, "loss": 0.5516, "step": 14589 }, { "epoch": 1.47, "learning_rate": 4.189541355688105e-05, "loss": 0.2895, "step": 14590 }, { "epoch": 1.47, "learning_rate": 4.1893216521092194e-05, "loss": 1.0415, "step": 14591 }, { "epoch": 1.47, "learning_rate": 4.1891019409619436e-05, "loss": 1.0226, "step": 14592 }, { "epoch": 1.47, "learning_rate": 4.188882222247675e-05, "loss": 0.7784, "step": 14593 }, { "epoch": 1.47, "learning_rate": 4.1886624959678116e-05, "loss": 0.3618, "step": 14594 }, { "epoch": 1.47, "learning_rate": 4.188442762123755e-05, "loss": 0.1169, "step": 14595 }, { "epoch": 1.47, "learning_rate": 4.188223020716898e-05, "loss": 1.0573, "step": 14596 }, { "epoch": 1.47, "learning_rate": 4.188003271748643e-05, "loss": 0.8696, "step": 14597 }, { "epoch": 1.47, "learning_rate": 4.1877835152203855e-05, "loss": 0.3858, "step": 14598 }, { "epoch": 1.47, "learning_rate": 4.1875637511335275e-05, "loss": 0.5095, "step": 14599 }, { "epoch": 1.47, "learning_rate": 4.187343979489464e-05, "loss": 0.8324, "step": 14600 }, { "epoch": 1.47, "learning_rate": 4.1871242002895966e-05, "loss": 0.8094, "step": 14601 }, { "epoch": 1.47, "learning_rate": 4.186904413535322e-05, "loss": 0.1502, "step": 14602 }, { "epoch": 1.47, "learning_rate": 4.186684619228038e-05, "loss": 1.0553, "step": 14603 }, { "epoch": 1.47, "learning_rate": 4.1864648173691455e-05, "loss": 0.4745, "step": 14604 }, { "epoch": 1.47, "learning_rate": 4.186245007960042e-05, "loss": 0.5793, "step": 14605 }, { "epoch": 1.47, "learning_rate": 4.186025191002127e-05, "loss": 0.7972, "step": 14606 }, { "epoch": 1.47, "learning_rate": 4.185805366496799e-05, "loss": 0.3276, "step": 14607 }, { "epoch": 1.47, "learning_rate": 4.1855855344454564e-05, "loss": 0.4251, "step": 14608 }, { "epoch": 1.47, "learning_rate": 4.185365694849499e-05, "loss": 1.4723, "step": 14609 }, { "epoch": 1.47, "learning_rate": 4.1851458477103236e-05, "loss": 0.47, "step": 14610 }, { "epoch": 1.47, "learning_rate": 4.1849259930293325e-05, "loss": 0.5211, "step": 14611 }, { "epoch": 1.47, "learning_rate": 4.184706130807922e-05, "loss": 0.4571, "step": 14612 }, { "epoch": 1.47, "learning_rate": 4.184486261047493e-05, "loss": 0.3526, "step": 14613 }, { "epoch": 1.47, "learning_rate": 4.1842663837494445e-05, "loss": 0.5257, "step": 14614 }, { "epoch": 1.47, "learning_rate": 4.184046498915175e-05, "loss": 0.8814, "step": 14615 }, { "epoch": 1.47, "learning_rate": 4.183826606546083e-05, "loss": 0.3172, "step": 14616 }, { "epoch": 1.47, "learning_rate": 4.1836067066435695e-05, "loss": 0.2897, "step": 14617 }, { "epoch": 1.47, "learning_rate": 4.183386799209034e-05, "loss": 0.4321, "step": 14618 }, { "epoch": 1.47, "learning_rate": 4.1831668842438736e-05, "loss": 1.3527, "step": 14619 }, { "epoch": 1.47, "learning_rate": 4.1829469617494905e-05, "loss": 0.9019, "step": 14620 }, { "epoch": 1.47, "learning_rate": 4.182727031727282e-05, "loss": 0.6648, "step": 14621 }, { "epoch": 1.47, "learning_rate": 4.1825070941786496e-05, "loss": 0.8203, "step": 14622 }, { "epoch": 1.47, "learning_rate": 4.182287149104992e-05, "loss": 0.4471, "step": 14623 }, { "epoch": 1.47, "learning_rate": 4.1820671965077077e-05, "loss": 0.5016, "step": 14624 }, { "epoch": 1.47, "learning_rate": 4.181847236388198e-05, "loss": 1.1408, "step": 14625 }, { "epoch": 1.47, "learning_rate": 4.1816272687478616e-05, "loss": 0.2295, "step": 14626 }, { "epoch": 1.47, "learning_rate": 4.181407293588099e-05, "loss": 0.3978, "step": 14627 }, { "epoch": 1.47, "learning_rate": 4.18118731091031e-05, "loss": 0.4413, "step": 14628 }, { "epoch": 1.47, "learning_rate": 4.180967320715894e-05, "loss": 0.7037, "step": 14629 }, { "epoch": 1.47, "learning_rate": 4.1807473230062526e-05, "loss": 1.0429, "step": 14630 }, { "epoch": 1.47, "learning_rate": 4.180527317782782e-05, "loss": 0.8284, "step": 14631 }, { "epoch": 1.47, "learning_rate": 4.1803073050468865e-05, "loss": 0.2502, "step": 14632 }, { "epoch": 1.47, "learning_rate": 4.180087284799964e-05, "loss": 1.0919, "step": 14633 }, { "epoch": 1.47, "learning_rate": 4.1798672570434156e-05, "loss": 0.789, "step": 14634 }, { "epoch": 1.47, "learning_rate": 4.179647221778639e-05, "loss": 0.7169, "step": 14635 }, { "epoch": 1.47, "learning_rate": 4.179427179007037e-05, "loss": 1.0794, "step": 14636 }, { "epoch": 1.47, "learning_rate": 4.17920712873001e-05, "loss": 1.2506, "step": 14637 }, { "epoch": 1.47, "learning_rate": 4.1789870709489565e-05, "loss": 0.574, "step": 14638 }, { "epoch": 1.47, "learning_rate": 4.178767005665278e-05, "loss": 0.5568, "step": 14639 }, { "epoch": 1.47, "learning_rate": 4.178546932880375e-05, "loss": 0.8566, "step": 14640 }, { "epoch": 1.47, "learning_rate": 4.178326852595647e-05, "loss": 0.6933, "step": 14641 }, { "epoch": 1.47, "learning_rate": 4.178106764812495e-05, "loss": 0.9574, "step": 14642 }, { "epoch": 1.47, "learning_rate": 4.177886669532321e-05, "loss": 0.2091, "step": 14643 }, { "epoch": 1.47, "learning_rate": 4.177666566756523e-05, "loss": 1.0882, "step": 14644 }, { "epoch": 1.47, "learning_rate": 4.177446456486504e-05, "loss": 1.0007, "step": 14645 }, { "epoch": 1.47, "learning_rate": 4.177226338723663e-05, "loss": 0.7334, "step": 14646 }, { "epoch": 1.47, "learning_rate": 4.177006213469402e-05, "loss": 0.4874, "step": 14647 }, { "epoch": 1.47, "learning_rate": 4.176786080725121e-05, "loss": 0.518, "step": 14648 }, { "epoch": 1.47, "learning_rate": 4.17656594049222e-05, "loss": 0.7101, "step": 14649 }, { "epoch": 1.47, "learning_rate": 4.1763457927721024e-05, "loss": 0.8066, "step": 14650 }, { "epoch": 1.47, "learning_rate": 4.176125637566167e-05, "loss": 0.1009, "step": 14651 }, { "epoch": 1.47, "learning_rate": 4.175905474875816e-05, "loss": 1.0367, "step": 14652 }, { "epoch": 1.47, "learning_rate": 4.1756853047024496e-05, "loss": 0.674, "step": 14653 }, { "epoch": 1.48, "learning_rate": 4.175465127047469e-05, "loss": 0.2966, "step": 14654 }, { "epoch": 1.48, "learning_rate": 4.1752449419122755e-05, "loss": 0.531, "step": 14655 }, { "epoch": 1.48, "learning_rate": 4.175024749298271e-05, "loss": 0.1556, "step": 14656 }, { "epoch": 1.48, "learning_rate": 4.1748045492068555e-05, "loss": 0.3006, "step": 14657 }, { "epoch": 1.48, "learning_rate": 4.1745843416394296e-05, "loss": 0.2019, "step": 14658 }, { "epoch": 1.48, "learning_rate": 4.1743641265973966e-05, "loss": 0.8857, "step": 14659 }, { "epoch": 1.48, "learning_rate": 4.174143904082158e-05, "loss": 0.4498, "step": 14660 }, { "epoch": 1.48, "learning_rate": 4.173923674095113e-05, "loss": 0.4866, "step": 14661 }, { "epoch": 1.48, "learning_rate": 4.173703436637665e-05, "loss": 0.1472, "step": 14662 }, { "epoch": 1.48, "learning_rate": 4.173483191711215e-05, "loss": 0.2354, "step": 14663 }, { "epoch": 1.48, "learning_rate": 4.173262939317164e-05, "loss": 0.2607, "step": 14664 }, { "epoch": 1.48, "learning_rate": 4.1730426794569137e-05, "loss": 0.0773, "step": 14665 }, { "epoch": 1.48, "learning_rate": 4.172822412131867e-05, "loss": 0.786, "step": 14666 }, { "epoch": 1.48, "learning_rate": 4.172602137343424e-05, "loss": 0.7734, "step": 14667 }, { "epoch": 1.48, "learning_rate": 4.1723818550929864e-05, "loss": 0.7994, "step": 14668 }, { "epoch": 1.48, "learning_rate": 4.172161565381957e-05, "loss": 0.2186, "step": 14669 }, { "epoch": 1.48, "learning_rate": 4.171941268211738e-05, "loss": 1.0779, "step": 14670 }, { "epoch": 1.48, "learning_rate": 4.1717209635837315e-05, "loss": 0.7735, "step": 14671 }, { "epoch": 1.48, "learning_rate": 4.171500651499336e-05, "loss": 0.5904, "step": 14672 }, { "epoch": 1.48, "learning_rate": 4.1712803319599585e-05, "loss": 0.6818, "step": 14673 }, { "epoch": 1.48, "learning_rate": 4.171060004966997e-05, "loss": 0.2374, "step": 14674 }, { "epoch": 1.48, "learning_rate": 4.1708396705218555e-05, "loss": 0.4094, "step": 14675 }, { "epoch": 1.48, "learning_rate": 4.1706193286259357e-05, "loss": 0.9204, "step": 14676 }, { "epoch": 1.48, "learning_rate": 4.1703989792806395e-05, "loss": 1.0716, "step": 14677 }, { "epoch": 1.48, "learning_rate": 4.17017862248737e-05, "loss": 0.5228, "step": 14678 }, { "epoch": 1.48, "learning_rate": 4.1699582582475285e-05, "loss": 0.8333, "step": 14679 }, { "epoch": 1.48, "learning_rate": 4.169737886562518e-05, "loss": 0.2011, "step": 14680 }, { "epoch": 1.48, "learning_rate": 4.16951750743374e-05, "loss": 0.4378, "step": 14681 }, { "epoch": 1.48, "learning_rate": 4.1692971208625986e-05, "loss": 1.0125, "step": 14682 }, { "epoch": 1.48, "learning_rate": 4.1690767268504945e-05, "loss": 0.304, "step": 14683 }, { "epoch": 1.48, "learning_rate": 4.168856325398831e-05, "loss": 0.8892, "step": 14684 }, { "epoch": 1.48, "learning_rate": 4.1686359165090105e-05, "loss": 0.6757, "step": 14685 }, { "epoch": 1.48, "learning_rate": 4.168415500182435e-05, "loss": 1.01, "step": 14686 }, { "epoch": 1.48, "learning_rate": 4.168195076420509e-05, "loss": 1.3559, "step": 14687 }, { "epoch": 1.48, "learning_rate": 4.167974645224632e-05, "loss": 1.0044, "step": 14688 }, { "epoch": 1.48, "learning_rate": 4.167754206596211e-05, "loss": 0.5653, "step": 14689 }, { "epoch": 1.48, "learning_rate": 4.167533760536645e-05, "loss": 0.4677, "step": 14690 }, { "epoch": 1.48, "learning_rate": 4.167313307047338e-05, "loss": 0.5978, "step": 14691 }, { "epoch": 1.48, "learning_rate": 4.1670928461296946e-05, "loss": 0.9812, "step": 14692 }, { "epoch": 1.48, "learning_rate": 4.166872377785115e-05, "loss": 0.6185, "step": 14693 }, { "epoch": 1.48, "learning_rate": 4.166651902015004e-05, "loss": 0.7046, "step": 14694 }, { "epoch": 1.48, "learning_rate": 4.166431418820764e-05, "loss": 1.681, "step": 14695 }, { "epoch": 1.48, "learning_rate": 4.166210928203799e-05, "loss": 0.1812, "step": 14696 }, { "epoch": 1.48, "learning_rate": 4.16599043016551e-05, "loss": 0.301, "step": 14697 }, { "epoch": 1.48, "learning_rate": 4.165769924707302e-05, "loss": 0.544, "step": 14698 }, { "epoch": 1.48, "learning_rate": 4.165549411830578e-05, "loss": 0.6757, "step": 14699 }, { "epoch": 1.48, "learning_rate": 4.165328891536741e-05, "loss": 0.2243, "step": 14700 }, { "epoch": 1.48, "learning_rate": 4.165108363827194e-05, "loss": 1.169, "step": 14701 }, { "epoch": 1.48, "learning_rate": 4.1648878287033404e-05, "loss": 1.0213, "step": 14702 }, { "epoch": 1.48, "learning_rate": 4.1646672861665833e-05, "loss": 2.0119, "step": 14703 }, { "epoch": 1.48, "learning_rate": 4.164446736218327e-05, "loss": 0.8281, "step": 14704 }, { "epoch": 1.48, "learning_rate": 4.164226178859976e-05, "loss": 0.249, "step": 14705 }, { "epoch": 1.48, "learning_rate": 4.164005614092931e-05, "loss": 0.3862, "step": 14706 }, { "epoch": 1.48, "learning_rate": 4.1637850419185973e-05, "loss": 1.0227, "step": 14707 }, { "epoch": 1.48, "learning_rate": 4.163564462338379e-05, "loss": 1.0286, "step": 14708 }, { "epoch": 1.48, "learning_rate": 4.163343875353679e-05, "loss": 0.9727, "step": 14709 }, { "epoch": 1.48, "learning_rate": 4.1631232809659004e-05, "loss": 1.3003, "step": 14710 }, { "epoch": 1.48, "learning_rate": 4.1629026791764476e-05, "loss": 0.4492, "step": 14711 }, { "epoch": 1.48, "learning_rate": 4.162682069986726e-05, "loss": 0.5183, "step": 14712 }, { "epoch": 1.48, "learning_rate": 4.162461453398136e-05, "loss": 1.0313, "step": 14713 }, { "epoch": 1.48, "learning_rate": 4.1622408294120846e-05, "loss": 0.0881, "step": 14714 }, { "epoch": 1.48, "learning_rate": 4.162020198029975e-05, "loss": 1.1232, "step": 14715 }, { "epoch": 1.48, "learning_rate": 4.1617995592532107e-05, "loss": 0.7911, "step": 14716 }, { "epoch": 1.48, "learning_rate": 4.161578913083196e-05, "loss": 0.3653, "step": 14717 }, { "epoch": 1.48, "learning_rate": 4.161358259521335e-05, "loss": 0.5244, "step": 14718 }, { "epoch": 1.48, "learning_rate": 4.1611375985690314e-05, "loss": 0.4478, "step": 14719 }, { "epoch": 1.48, "learning_rate": 4.16091693022769e-05, "loss": 0.8655, "step": 14720 }, { "epoch": 1.48, "learning_rate": 4.160696254498716e-05, "loss": 0.3477, "step": 14721 }, { "epoch": 1.48, "learning_rate": 4.160475571383511e-05, "loss": 0.9614, "step": 14722 }, { "epoch": 1.48, "learning_rate": 4.1602548808834816e-05, "loss": 0.517, "step": 14723 }, { "epoch": 1.48, "learning_rate": 4.1600341830000315e-05, "loss": 0.7488, "step": 14724 }, { "epoch": 1.48, "learning_rate": 4.1598134777345646e-05, "loss": 1.1058, "step": 14725 }, { "epoch": 1.48, "learning_rate": 4.159592765088487e-05, "loss": 0.4609, "step": 14726 }, { "epoch": 1.48, "learning_rate": 4.159372045063201e-05, "loss": 0.1583, "step": 14727 }, { "epoch": 1.48, "learning_rate": 4.159151317660114e-05, "loss": 0.9592, "step": 14728 }, { "epoch": 1.48, "learning_rate": 4.158930582880627e-05, "loss": 0.6664, "step": 14729 }, { "epoch": 1.48, "learning_rate": 4.1587098407261486e-05, "loss": 0.6208, "step": 14730 }, { "epoch": 1.48, "learning_rate": 4.15848909119808e-05, "loss": 0.5384, "step": 14731 }, { "epoch": 1.48, "learning_rate": 4.158268334297828e-05, "loss": 0.539, "step": 14732 }, { "epoch": 1.48, "learning_rate": 4.158047570026797e-05, "loss": 0.5995, "step": 14733 }, { "epoch": 1.48, "learning_rate": 4.157826798386393e-05, "loss": 0.2974, "step": 14734 }, { "epoch": 1.48, "learning_rate": 4.157606019378018e-05, "loss": 1.1918, "step": 14735 }, { "epoch": 1.48, "learning_rate": 4.157385233003078e-05, "loss": 1.2722, "step": 14736 }, { "epoch": 1.48, "learning_rate": 4.157164439262982e-05, "loss": 0.4632, "step": 14737 }, { "epoch": 1.48, "learning_rate": 4.156943638159129e-05, "loss": 0.7218, "step": 14738 }, { "epoch": 1.48, "learning_rate": 4.156722829692927e-05, "loss": 0.4018, "step": 14739 }, { "epoch": 1.48, "learning_rate": 4.156502013865782e-05, "loss": 0.6447, "step": 14740 }, { "epoch": 1.48, "learning_rate": 4.156281190679097e-05, "loss": 1.0119, "step": 14741 }, { "epoch": 1.48, "learning_rate": 4.156060360134279e-05, "loss": 0.5831, "step": 14742 }, { "epoch": 1.48, "learning_rate": 4.155839522232732e-05, "loss": 1.1126, "step": 14743 }, { "epoch": 1.48, "learning_rate": 4.155618676975864e-05, "loss": 0.5482, "step": 14744 }, { "epoch": 1.48, "learning_rate": 4.155397824365076e-05, "loss": 1.0493, "step": 14745 }, { "epoch": 1.48, "learning_rate": 4.155176964401777e-05, "loss": 1.0852, "step": 14746 }, { "epoch": 1.48, "learning_rate": 4.154956097087371e-05, "loss": 0.4902, "step": 14747 }, { "epoch": 1.48, "learning_rate": 4.154735222423264e-05, "loss": 0.2443, "step": 14748 }, { "epoch": 1.48, "learning_rate": 4.154514340410861e-05, "loss": 0.4193, "step": 14749 }, { "epoch": 1.48, "learning_rate": 4.154293451051568e-05, "loss": 0.3052, "step": 14750 }, { "epoch": 1.48, "learning_rate": 4.154072554346791e-05, "loss": 0.9084, "step": 14751 }, { "epoch": 1.48, "learning_rate": 4.1538516502979344e-05, "loss": 0.2626, "step": 14752 }, { "epoch": 1.49, "learning_rate": 4.153630738906405e-05, "loss": 1.0792, "step": 14753 }, { "epoch": 1.49, "learning_rate": 4.153409820173609e-05, "loss": 0.4675, "step": 14754 }, { "epoch": 1.49, "learning_rate": 4.153188894100951e-05, "loss": 0.6394, "step": 14755 }, { "epoch": 1.49, "learning_rate": 4.152967960689839e-05, "loss": 0.565, "step": 14756 }, { "epoch": 1.49, "learning_rate": 4.152747019941676e-05, "loss": 1.3994, "step": 14757 }, { "epoch": 1.49, "learning_rate": 4.1525260718578704e-05, "loss": 0.2456, "step": 14758 }, { "epoch": 1.49, "learning_rate": 4.152305116439826e-05, "loss": 0.9726, "step": 14759 }, { "epoch": 1.49, "learning_rate": 4.152084153688952e-05, "loss": 0.3071, "step": 14760 }, { "epoch": 1.49, "learning_rate": 4.151863183606651e-05, "loss": 0.5317, "step": 14761 }, { "epoch": 1.49, "learning_rate": 4.151642206194332e-05, "loss": 0.1761, "step": 14762 }, { "epoch": 1.49, "learning_rate": 4.151421221453401e-05, "loss": 0.6274, "step": 14763 }, { "epoch": 1.49, "learning_rate": 4.151200229385261e-05, "loss": 0.2639, "step": 14764 }, { "epoch": 1.49, "learning_rate": 4.150979229991322e-05, "loss": 0.3703, "step": 14765 }, { "epoch": 1.49, "learning_rate": 4.150758223272989e-05, "loss": 0.5489, "step": 14766 }, { "epoch": 1.49, "learning_rate": 4.150537209231669e-05, "loss": 0.604, "step": 14767 }, { "epoch": 1.49, "learning_rate": 4.1503161878687666e-05, "loss": 1.2976, "step": 14768 }, { "epoch": 1.49, "learning_rate": 4.150095159185691e-05, "loss": 0.439, "step": 14769 }, { "epoch": 1.49, "learning_rate": 4.149874123183846e-05, "loss": 0.5094, "step": 14770 }, { "epoch": 1.49, "learning_rate": 4.14965307986464e-05, "loss": 0.9275, "step": 14771 }, { "epoch": 1.49, "learning_rate": 4.1494320292294795e-05, "loss": 0.9117, "step": 14772 }, { "epoch": 1.49, "learning_rate": 4.149210971279771e-05, "loss": 0.5878, "step": 14773 }, { "epoch": 1.49, "learning_rate": 4.148989906016921e-05, "loss": 0.4094, "step": 14774 }, { "epoch": 1.49, "learning_rate": 4.1487688334423346e-05, "loss": 1.0185, "step": 14775 }, { "epoch": 1.49, "learning_rate": 4.148547753557423e-05, "loss": 0.6866, "step": 14776 }, { "epoch": 1.49, "learning_rate": 4.148326666363589e-05, "loss": 0.5686, "step": 14777 }, { "epoch": 1.49, "learning_rate": 4.14810557186224e-05, "loss": 0.6044, "step": 14778 }, { "epoch": 1.49, "learning_rate": 4.1478844700547856e-05, "loss": 0.9187, "step": 14779 }, { "epoch": 1.49, "learning_rate": 4.1476633609426306e-05, "loss": 0.2881, "step": 14780 }, { "epoch": 1.49, "learning_rate": 4.147442244527183e-05, "loss": 1.0427, "step": 14781 }, { "epoch": 1.49, "learning_rate": 4.147221120809848e-05, "loss": 0.5858, "step": 14782 }, { "epoch": 1.49, "learning_rate": 4.146999989792036e-05, "loss": 0.8037, "step": 14783 }, { "epoch": 1.49, "learning_rate": 4.146778851475151e-05, "loss": 0.2414, "step": 14784 }, { "epoch": 1.49, "learning_rate": 4.146557705860602e-05, "loss": 0.3881, "step": 14785 }, { "epoch": 1.49, "learning_rate": 4.1463365529497966e-05, "loss": 0.2304, "step": 14786 }, { "epoch": 1.49, "learning_rate": 4.146115392744141e-05, "loss": 0.1654, "step": 14787 }, { "epoch": 1.49, "learning_rate": 4.145894225245043e-05, "loss": 0.4034, "step": 14788 }, { "epoch": 1.49, "learning_rate": 4.1456730504539106e-05, "loss": 0.8062, "step": 14789 }, { "epoch": 1.49, "learning_rate": 4.1454518683721495e-05, "loss": 0.1553, "step": 14790 }, { "epoch": 1.49, "learning_rate": 4.145230679001169e-05, "loss": 0.6584, "step": 14791 }, { "epoch": 1.49, "learning_rate": 4.145009482342377e-05, "loss": 1.2394, "step": 14792 }, { "epoch": 1.49, "learning_rate": 4.1447882783971796e-05, "loss": 0.6255, "step": 14793 }, { "epoch": 1.49, "learning_rate": 4.144567067166985e-05, "loss": 0.2872, "step": 14794 }, { "epoch": 1.49, "learning_rate": 4.1443458486532014e-05, "loss": 0.7916, "step": 14795 }, { "epoch": 1.49, "learning_rate": 4.144124622857236e-05, "loss": 0.3884, "step": 14796 }, { "epoch": 1.49, "learning_rate": 4.143903389780496e-05, "loss": 0.3582, "step": 14797 }, { "epoch": 1.49, "learning_rate": 4.14368214942439e-05, "loss": 0.5958, "step": 14798 }, { "epoch": 1.49, "learning_rate": 4.143460901790328e-05, "loss": 0.7532, "step": 14799 }, { "epoch": 1.49, "learning_rate": 4.1432396468797136e-05, "loss": 0.2996, "step": 14800 }, { "epoch": 1.49, "learning_rate": 4.143018384693957e-05, "loss": 0.8599, "step": 14801 }, { "epoch": 1.49, "learning_rate": 4.1427971152344674e-05, "loss": 0.2378, "step": 14802 }, { "epoch": 1.49, "learning_rate": 4.1425758385026515e-05, "loss": 0.6154, "step": 14803 }, { "epoch": 1.49, "learning_rate": 4.1423545544999176e-05, "loss": 0.3365, "step": 14804 }, { "epoch": 1.49, "learning_rate": 4.142133263227674e-05, "loss": 0.4237, "step": 14805 }, { "epoch": 1.49, "learning_rate": 4.141911964687328e-05, "loss": 0.868, "step": 14806 }, { "epoch": 1.49, "learning_rate": 4.141690658880289e-05, "loss": 0.2329, "step": 14807 }, { "epoch": 1.49, "learning_rate": 4.141469345807965e-05, "loss": 1.1452, "step": 14808 }, { "epoch": 1.49, "learning_rate": 4.141248025471764e-05, "loss": 0.8868, "step": 14809 }, { "epoch": 1.49, "learning_rate": 4.141026697873095e-05, "loss": 1.2448, "step": 14810 }, { "epoch": 1.49, "learning_rate": 4.1408053630133667e-05, "loss": 1.1478, "step": 14811 }, { "epoch": 1.49, "learning_rate": 4.1405840208939865e-05, "loss": 0.3096, "step": 14812 }, { "epoch": 1.49, "learning_rate": 4.1403626715163635e-05, "loss": 0.6019, "step": 14813 }, { "epoch": 1.49, "learning_rate": 4.140141314881906e-05, "loss": 0.7645, "step": 14814 }, { "epoch": 1.49, "learning_rate": 4.1399199509920243e-05, "loss": 0.7074, "step": 14815 }, { "epoch": 1.49, "learning_rate": 4.139698579848124e-05, "loss": 0.2381, "step": 14816 }, { "epoch": 1.49, "learning_rate": 4.139477201451616e-05, "loss": 0.434, "step": 14817 }, { "epoch": 1.49, "learning_rate": 4.1392558158039095e-05, "loss": 0.2065, "step": 14818 }, { "epoch": 1.49, "learning_rate": 4.139034422906411e-05, "loss": 0.5413, "step": 14819 }, { "epoch": 1.49, "learning_rate": 4.1388130227605315e-05, "loss": 0.4049, "step": 14820 }, { "epoch": 1.49, "learning_rate": 4.1385916153676784e-05, "loss": 0.4223, "step": 14821 }, { "epoch": 1.49, "learning_rate": 4.138370200729263e-05, "loss": 0.22, "step": 14822 }, { "epoch": 1.49, "learning_rate": 4.138148778846691e-05, "loss": 0.6305, "step": 14823 }, { "epoch": 1.49, "learning_rate": 4.137927349721374e-05, "loss": 1.1749, "step": 14824 }, { "epoch": 1.49, "learning_rate": 4.13770591335472e-05, "loss": 0.556, "step": 14825 }, { "epoch": 1.49, "learning_rate": 4.137484469748139e-05, "loss": 0.0984, "step": 14826 }, { "epoch": 1.49, "learning_rate": 4.1372630189030384e-05, "loss": 0.0264, "step": 14827 }, { "epoch": 1.49, "learning_rate": 4.1370415608208295e-05, "loss": 1.1278, "step": 14828 }, { "epoch": 1.49, "learning_rate": 4.13682009550292e-05, "loss": 0.7656, "step": 14829 }, { "epoch": 1.49, "learning_rate": 4.13659862295072e-05, "loss": 0.3401, "step": 14830 }, { "epoch": 1.49, "learning_rate": 4.1363771431656405e-05, "loss": 0.4277, "step": 14831 }, { "epoch": 1.49, "learning_rate": 4.136155656149086e-05, "loss": 1.3128, "step": 14832 }, { "epoch": 1.49, "learning_rate": 4.1359341619024716e-05, "loss": 0.394, "step": 14833 }, { "epoch": 1.49, "learning_rate": 4.135712660427203e-05, "loss": 0.6968, "step": 14834 }, { "epoch": 1.49, "learning_rate": 4.1354911517246927e-05, "loss": 0.6051, "step": 14835 }, { "epoch": 1.49, "learning_rate": 4.1352696357963466e-05, "loss": 0.5851, "step": 14836 }, { "epoch": 1.49, "learning_rate": 4.135048112643577e-05, "loss": 0.4079, "step": 14837 }, { "epoch": 1.49, "learning_rate": 4.134826582267794e-05, "loss": 1.1291, "step": 14838 }, { "epoch": 1.49, "learning_rate": 4.134605044670406e-05, "loss": 0.3959, "step": 14839 }, { "epoch": 1.49, "learning_rate": 4.134383499852823e-05, "loss": 1.1378, "step": 14840 }, { "epoch": 1.49, "learning_rate": 4.134161947816454e-05, "loss": 1.3639, "step": 14841 }, { "epoch": 1.49, "learning_rate": 4.133940388562711e-05, "loss": 0.4714, "step": 14842 }, { "epoch": 1.49, "learning_rate": 4.1337188220930027e-05, "loss": 0.6227, "step": 14843 }, { "epoch": 1.49, "learning_rate": 4.1334972484087386e-05, "loss": 1.1995, "step": 14844 }, { "epoch": 1.49, "learning_rate": 4.133275667511329e-05, "loss": 0.2695, "step": 14845 }, { "epoch": 1.49, "learning_rate": 4.133054079402184e-05, "loss": 0.7362, "step": 14846 }, { "epoch": 1.49, "learning_rate": 4.132832484082715e-05, "loss": 0.9596, "step": 14847 }, { "epoch": 1.49, "learning_rate": 4.1326108815543304e-05, "loss": 1.3028, "step": 14848 }, { "epoch": 1.49, "learning_rate": 4.132389271818441e-05, "loss": 0.7733, "step": 14849 }, { "epoch": 1.49, "learning_rate": 4.132167654876457e-05, "loss": 0.6099, "step": 14850 }, { "epoch": 1.5, "learning_rate": 4.131946030729789e-05, "loss": 1.3594, "step": 14851 }, { "epoch": 1.5, "learning_rate": 4.131724399379847e-05, "loss": 0.62, "step": 14852 }, { "epoch": 1.5, "learning_rate": 4.131502760828041e-05, "loss": 0.8171, "step": 14853 }, { "epoch": 1.5, "learning_rate": 4.131281115075783e-05, "loss": 0.2576, "step": 14854 }, { "epoch": 1.5, "learning_rate": 4.131059462124481e-05, "loss": 0.9431, "step": 14855 }, { "epoch": 1.5, "learning_rate": 4.130837801975548e-05, "loss": 0.9395, "step": 14856 }, { "epoch": 1.5, "learning_rate": 4.1306161346303926e-05, "loss": 0.5692, "step": 14857 }, { "epoch": 1.5, "learning_rate": 4.1303944600904265e-05, "loss": 0.2454, "step": 14858 }, { "epoch": 1.5, "learning_rate": 4.1301727783570595e-05, "loss": 0.6526, "step": 14859 }, { "epoch": 1.5, "learning_rate": 4.129951089431704e-05, "loss": 1.0521, "step": 14860 }, { "epoch": 1.5, "learning_rate": 4.1297293933157695e-05, "loss": 0.4781, "step": 14861 }, { "epoch": 1.5, "learning_rate": 4.129507690010666e-05, "loss": 0.2857, "step": 14862 }, { "epoch": 1.5, "learning_rate": 4.129285979517806e-05, "loss": 0.1839, "step": 14863 }, { "epoch": 1.5, "learning_rate": 4.1290642618385994e-05, "loss": 0.2882, "step": 14864 }, { "epoch": 1.5, "learning_rate": 4.128842536974457e-05, "loss": 0.9096, "step": 14865 }, { "epoch": 1.5, "learning_rate": 4.12862080492679e-05, "loss": 0.6236, "step": 14866 }, { "epoch": 1.5, "learning_rate": 4.128399065697011e-05, "loss": 1.4616, "step": 14867 }, { "epoch": 1.5, "learning_rate": 4.128177319286528e-05, "loss": 0.698, "step": 14868 }, { "epoch": 1.5, "learning_rate": 4.127955565696754e-05, "loss": 1.1044, "step": 14869 }, { "epoch": 1.5, "learning_rate": 4.127733804929101e-05, "loss": 0.7483, "step": 14870 }, { "epoch": 1.5, "learning_rate": 4.1275120369849776e-05, "loss": 0.1933, "step": 14871 }, { "epoch": 1.5, "learning_rate": 4.127290261865797e-05, "loss": 1.341, "step": 14872 }, { "epoch": 1.5, "learning_rate": 4.127068479572971e-05, "loss": 0.2272, "step": 14873 }, { "epoch": 1.5, "learning_rate": 4.126846690107909e-05, "loss": 0.8335, "step": 14874 }, { "epoch": 1.5, "learning_rate": 4.126624893472024e-05, "loss": 0.3428, "step": 14875 }, { "epoch": 1.5, "learning_rate": 4.1264030896667254e-05, "loss": 0.8696, "step": 14876 }, { "epoch": 1.5, "learning_rate": 4.126181278693428e-05, "loss": 0.794, "step": 14877 }, { "epoch": 1.5, "learning_rate": 4.1259594605535395e-05, "loss": 1.1698, "step": 14878 }, { "epoch": 1.5, "learning_rate": 4.125737635248474e-05, "loss": 0.3403, "step": 14879 }, { "epoch": 1.5, "learning_rate": 4.1255158027796426e-05, "loss": 0.4917, "step": 14880 }, { "epoch": 1.5, "learning_rate": 4.125293963148457e-05, "loss": 0.8119, "step": 14881 }, { "epoch": 1.5, "learning_rate": 4.125072116356329e-05, "loss": 0.5533, "step": 14882 }, { "epoch": 1.5, "learning_rate": 4.1248502624046694e-05, "loss": 0.1817, "step": 14883 }, { "epoch": 1.5, "learning_rate": 4.1246284012948906e-05, "loss": 0.806, "step": 14884 }, { "epoch": 1.5, "learning_rate": 4.124406533028404e-05, "loss": 0.7066, "step": 14885 }, { "epoch": 1.5, "learning_rate": 4.124184657606623e-05, "loss": 0.3246, "step": 14886 }, { "epoch": 1.5, "learning_rate": 4.123962775030958e-05, "loss": 0.4557, "step": 14887 }, { "epoch": 1.5, "learning_rate": 4.1237408853028214e-05, "loss": 0.6396, "step": 14888 }, { "epoch": 1.5, "learning_rate": 4.123518988423625e-05, "loss": 0.8292, "step": 14889 }, { "epoch": 1.5, "learning_rate": 4.1232970843947824e-05, "loss": 0.3755, "step": 14890 }, { "epoch": 1.5, "learning_rate": 4.1230751732177046e-05, "loss": 1.1165, "step": 14891 }, { "epoch": 1.5, "learning_rate": 4.122853254893802e-05, "loss": 0.7857, "step": 14892 }, { "epoch": 1.5, "learning_rate": 4.12263132942449e-05, "loss": 0.7687, "step": 14893 }, { "epoch": 1.5, "learning_rate": 4.122409396811178e-05, "loss": 0.2022, "step": 14894 }, { "epoch": 1.5, "learning_rate": 4.12218745705528e-05, "loss": 0.7689, "step": 14895 }, { "epoch": 1.5, "learning_rate": 4.121965510158208e-05, "loss": 0.7444, "step": 14896 }, { "epoch": 1.5, "learning_rate": 4.121743556121374e-05, "loss": 0.9065, "step": 14897 }, { "epoch": 1.5, "learning_rate": 4.1215215949461917e-05, "loss": 1.2074, "step": 14898 }, { "epoch": 1.5, "learning_rate": 4.121299626634072e-05, "loss": 0.7443, "step": 14899 }, { "epoch": 1.5, "learning_rate": 4.121077651186428e-05, "loss": 1.0534, "step": 14900 }, { "epoch": 1.5, "learning_rate": 4.1208556686046716e-05, "loss": 0.323, "step": 14901 }, { "epoch": 1.5, "learning_rate": 4.120633678890217e-05, "loss": 0.7569, "step": 14902 }, { "epoch": 1.5, "learning_rate": 4.1204116820444764e-05, "loss": 0.5567, "step": 14903 }, { "epoch": 1.5, "learning_rate": 4.1201896780688616e-05, "loss": 0.5613, "step": 14904 }, { "epoch": 1.5, "learning_rate": 4.119967666964785e-05, "loss": 0.5333, "step": 14905 }, { "epoch": 1.5, "learning_rate": 4.119745648733662e-05, "loss": 0.4621, "step": 14906 }, { "epoch": 1.5, "learning_rate": 4.119523623376902e-05, "loss": 0.4062, "step": 14907 }, { "epoch": 1.5, "learning_rate": 4.11930159089592e-05, "loss": 0.2627, "step": 14908 }, { "epoch": 1.5, "learning_rate": 4.1190795512921294e-05, "loss": 1.1067, "step": 14909 }, { "epoch": 1.5, "learning_rate": 4.1188575045669407e-05, "loss": 0.9426, "step": 14910 }, { "epoch": 1.5, "learning_rate": 4.11863545072177e-05, "loss": 1.5787, "step": 14911 }, { "epoch": 1.5, "learning_rate": 4.1184133897580277e-05, "loss": 1.1027, "step": 14912 }, { "epoch": 1.5, "learning_rate": 4.1181913216771285e-05, "loss": 0.6021, "step": 14913 }, { "epoch": 1.5, "learning_rate": 4.117969246480485e-05, "loss": 0.5747, "step": 14914 }, { "epoch": 1.5, "learning_rate": 4.11774716416951e-05, "loss": 1.1058, "step": 14915 }, { "epoch": 1.5, "learning_rate": 4.117525074745619e-05, "loss": 1.5619, "step": 14916 }, { "epoch": 1.5, "learning_rate": 4.117302978210222e-05, "loss": 0.2727, "step": 14917 }, { "epoch": 1.5, "learning_rate": 4.117080874564734e-05, "loss": 0.6537, "step": 14918 }, { "epoch": 1.5, "learning_rate": 4.116858763810569e-05, "loss": 0.2762, "step": 14919 }, { "epoch": 1.5, "learning_rate": 4.116636645949139e-05, "loss": 0.2767, "step": 14920 }, { "epoch": 1.5, "learning_rate": 4.116414520981858e-05, "loss": 0.7232, "step": 14921 }, { "epoch": 1.5, "learning_rate": 4.1161923889101414e-05, "loss": 0.5909, "step": 14922 }, { "epoch": 1.5, "learning_rate": 4.1159702497354e-05, "loss": 0.7812, "step": 14923 }, { "epoch": 1.5, "learning_rate": 4.115748103459047e-05, "loss": 0.2651, "step": 14924 }, { "epoch": 1.5, "learning_rate": 4.1155259500825e-05, "loss": 0.3835, "step": 14925 }, { "epoch": 1.5, "learning_rate": 4.115303789607169e-05, "loss": 1.0424, "step": 14926 }, { "epoch": 1.5, "learning_rate": 4.115081622034469e-05, "loss": 0.3881, "step": 14927 }, { "epoch": 1.5, "learning_rate": 4.1148594473658125e-05, "loss": 0.215, "step": 14928 }, { "epoch": 1.5, "learning_rate": 4.114637265602617e-05, "loss": 0.6297, "step": 14929 }, { "epoch": 1.5, "learning_rate": 4.1144150767462924e-05, "loss": 0.3165, "step": 14930 }, { "epoch": 1.5, "learning_rate": 4.114192880798254e-05, "loss": 0.4104, "step": 14931 }, { "epoch": 1.5, "learning_rate": 4.1139706777599175e-05, "loss": 0.3602, "step": 14932 }, { "epoch": 1.5, "learning_rate": 4.113748467632694e-05, "loss": 0.2473, "step": 14933 }, { "epoch": 1.5, "learning_rate": 4.113526250418e-05, "loss": 1.0566, "step": 14934 }, { "epoch": 1.5, "learning_rate": 4.1133040261172476e-05, "loss": 0.3327, "step": 14935 }, { "epoch": 1.5, "learning_rate": 4.113081794731853e-05, "loss": 0.9343, "step": 14936 }, { "epoch": 1.5, "learning_rate": 4.1128595562632275e-05, "loss": 0.415, "step": 14937 }, { "epoch": 1.5, "learning_rate": 4.112637310712789e-05, "loss": 0.3688, "step": 14938 }, { "epoch": 1.5, "learning_rate": 4.112415058081949e-05, "loss": 0.433, "step": 14939 }, { "epoch": 1.5, "learning_rate": 4.112192798372122e-05, "loss": 0.4221, "step": 14940 }, { "epoch": 1.5, "learning_rate": 4.111970531584724e-05, "loss": 0.5702, "step": 14941 }, { "epoch": 1.5, "learning_rate": 4.111748257721169e-05, "loss": 0.4448, "step": 14942 }, { "epoch": 1.5, "learning_rate": 4.111525976782871e-05, "loss": 0.5221, "step": 14943 }, { "epoch": 1.5, "learning_rate": 4.111303688771243e-05, "loss": 0.2528, "step": 14944 }, { "epoch": 1.5, "learning_rate": 4.1110813936877036e-05, "loss": 0.2548, "step": 14945 }, { "epoch": 1.5, "learning_rate": 4.1108590915336635e-05, "loss": 0.158, "step": 14946 }, { "epoch": 1.5, "learning_rate": 4.110636782310538e-05, "loss": 1.2985, "step": 14947 }, { "epoch": 1.5, "learning_rate": 4.110414466019744e-05, "loss": 1.1934, "step": 14948 }, { "epoch": 1.5, "learning_rate": 4.110192142662693e-05, "loss": 1.0402, "step": 14949 }, { "epoch": 1.51, "learning_rate": 4.109969812240803e-05, "loss": 0.4567, "step": 14950 }, { "epoch": 1.51, "learning_rate": 4.109747474755486e-05, "loss": 0.5684, "step": 14951 }, { "epoch": 1.51, "learning_rate": 4.10952513020816e-05, "loss": 0.2448, "step": 14952 }, { "epoch": 1.51, "learning_rate": 4.109302778600237e-05, "loss": 0.5772, "step": 14953 }, { "epoch": 1.51, "learning_rate": 4.109080419933134e-05, "loss": 0.1289, "step": 14954 }, { "epoch": 1.51, "learning_rate": 4.1088580542082645e-05, "loss": 0.5652, "step": 14955 }, { "epoch": 1.51, "learning_rate": 4.108635681427044e-05, "loss": 0.8712, "step": 14956 }, { "epoch": 1.51, "learning_rate": 4.1084133015908884e-05, "loss": 0.6499, "step": 14957 }, { "epoch": 1.51, "learning_rate": 4.108190914701211e-05, "loss": 0.4662, "step": 14958 }, { "epoch": 1.51, "learning_rate": 4.107968520759429e-05, "loss": 0.8351, "step": 14959 }, { "epoch": 1.51, "learning_rate": 4.1077461197669566e-05, "loss": 1.2363, "step": 14960 }, { "epoch": 1.51, "learning_rate": 4.10752371172521e-05, "loss": 0.4473, "step": 14961 }, { "epoch": 1.51, "learning_rate": 4.1073012966356034e-05, "loss": 0.3138, "step": 14962 }, { "epoch": 1.51, "learning_rate": 4.1070788744995523e-05, "loss": 0.6053, "step": 14963 }, { "epoch": 1.51, "learning_rate": 4.1068564453184734e-05, "loss": 0.1855, "step": 14964 }, { "epoch": 1.51, "learning_rate": 4.10663400909378e-05, "loss": 0.3922, "step": 14965 }, { "epoch": 1.51, "learning_rate": 4.106411565826889e-05, "loss": 0.2036, "step": 14966 }, { "epoch": 1.51, "learning_rate": 4.106189115519216e-05, "loss": 0.6359, "step": 14967 }, { "epoch": 1.51, "learning_rate": 4.105966658172176e-05, "loss": 0.7403, "step": 14968 }, { "epoch": 1.51, "learning_rate": 4.105744193787185e-05, "loss": 0.2263, "step": 14969 }, { "epoch": 1.51, "learning_rate": 4.105521722365658e-05, "loss": 0.4074, "step": 14970 }, { "epoch": 1.51, "learning_rate": 4.105299243909013e-05, "loss": 1.0016, "step": 14971 }, { "epoch": 1.51, "learning_rate": 4.105076758418662e-05, "loss": 0.6938, "step": 14972 }, { "epoch": 1.51, "learning_rate": 4.1048542658960245e-05, "loss": 1.1298, "step": 14973 }, { "epoch": 1.51, "learning_rate": 4.104631766342514e-05, "loss": 0.8774, "step": 14974 }, { "epoch": 1.51, "learning_rate": 4.104409259759548e-05, "loss": 1.0409, "step": 14975 }, { "epoch": 1.51, "learning_rate": 4.1041867461485404e-05, "loss": 0.4816, "step": 14976 }, { "epoch": 1.51, "learning_rate": 4.103964225510909e-05, "loss": 0.6851, "step": 14977 }, { "epoch": 1.51, "learning_rate": 4.1037416978480695e-05, "loss": 0.9127, "step": 14978 }, { "epoch": 1.51, "learning_rate": 4.1035191631614375e-05, "loss": 1.0603, "step": 14979 }, { "epoch": 1.51, "learning_rate": 4.103296621452429e-05, "loss": 0.3695, "step": 14980 }, { "epoch": 1.51, "learning_rate": 4.103074072722461e-05, "loss": 0.1591, "step": 14981 }, { "epoch": 1.51, "learning_rate": 4.102851516972949e-05, "loss": 0.8773, "step": 14982 }, { "epoch": 1.51, "learning_rate": 4.102628954205309e-05, "loss": 0.2394, "step": 14983 }, { "epoch": 1.51, "learning_rate": 4.1024063844209595e-05, "loss": 1.0205, "step": 14984 }, { "epoch": 1.51, "learning_rate": 4.102183807621313e-05, "loss": 0.6433, "step": 14985 }, { "epoch": 1.51, "learning_rate": 4.10196122380779e-05, "loss": 0.2652, "step": 14986 }, { "epoch": 1.51, "learning_rate": 4.1017386329818034e-05, "loss": 0.7143, "step": 14987 }, { "epoch": 1.51, "learning_rate": 4.1015160351447716e-05, "loss": 1.3389, "step": 14988 }, { "epoch": 1.51, "learning_rate": 4.101293430298112e-05, "loss": 0.9032, "step": 14989 }, { "epoch": 1.51, "learning_rate": 4.1010708184432383e-05, "loss": 1.0925, "step": 14990 }, { "epoch": 1.51, "learning_rate": 4.1008481995815694e-05, "loss": 1.0332, "step": 14991 }, { "epoch": 1.51, "learning_rate": 4.1006255737145206e-05, "loss": 0.7531, "step": 14992 }, { "epoch": 1.51, "learning_rate": 4.1004029408435115e-05, "loss": 0.9313, "step": 14993 }, { "epoch": 1.51, "learning_rate": 4.1001803009699556e-05, "loss": 0.6681, "step": 14994 }, { "epoch": 1.51, "learning_rate": 4.099957654095269e-05, "loss": 0.3673, "step": 14995 }, { "epoch": 1.51, "learning_rate": 4.099735000220873e-05, "loss": 0.2334, "step": 14996 }, { "epoch": 1.51, "learning_rate": 4.09951233934818e-05, "loss": 0.1183, "step": 14997 }, { "epoch": 1.51, "learning_rate": 4.09928967147861e-05, "loss": 0.8949, "step": 14998 }, { "epoch": 1.51, "learning_rate": 4.099066996613578e-05, "loss": 0.9211, "step": 14999 }, { "epoch": 1.51, "learning_rate": 4.098844314754502e-05, "loss": 0.85, "step": 15000 }, { "epoch": 1.51, "learning_rate": 4.098621625902798e-05, "loss": 0.8498, "step": 15001 }, { "epoch": 1.51, "learning_rate": 4.098398930059885e-05, "loss": 0.3739, "step": 15002 }, { "epoch": 1.51, "learning_rate": 4.098176227227179e-05, "loss": 0.2722, "step": 15003 }, { "epoch": 1.51, "learning_rate": 4.097953517406097e-05, "loss": 0.2924, "step": 15004 }, { "epoch": 1.51, "learning_rate": 4.097730800598057e-05, "loss": 0.472, "step": 15005 }, { "epoch": 1.51, "learning_rate": 4.097508076804475e-05, "loss": 0.9315, "step": 15006 }, { "epoch": 1.51, "learning_rate": 4.0972853460267695e-05, "loss": 0.2102, "step": 15007 }, { "epoch": 1.51, "learning_rate": 4.097062608266357e-05, "loss": 0.3294, "step": 15008 }, { "epoch": 1.51, "learning_rate": 4.0968398635246555e-05, "loss": 0.4124, "step": 15009 }, { "epoch": 1.51, "learning_rate": 4.096617111803083e-05, "loss": 0.49, "step": 15010 }, { "epoch": 1.51, "learning_rate": 4.0963943531030556e-05, "loss": 1.0748, "step": 15011 }, { "epoch": 1.51, "learning_rate": 4.0961715874259926e-05, "loss": 0.4132, "step": 15012 }, { "epoch": 1.51, "learning_rate": 4.0959488147733104e-05, "loss": 0.682, "step": 15013 }, { "epoch": 1.51, "learning_rate": 4.095726035146426e-05, "loss": 0.744, "step": 15014 }, { "epoch": 1.51, "learning_rate": 4.095503248546758e-05, "loss": 0.7702, "step": 15015 }, { "epoch": 1.51, "learning_rate": 4.095280454975725e-05, "loss": 0.8905, "step": 15016 }, { "epoch": 1.51, "learning_rate": 4.095057654434743e-05, "loss": 0.225, "step": 15017 }, { "epoch": 1.51, "learning_rate": 4.0948348469252316e-05, "loss": 0.8455, "step": 15018 }, { "epoch": 1.51, "learning_rate": 4.0946120324486076e-05, "loss": 0.662, "step": 15019 }, { "epoch": 1.51, "learning_rate": 4.0943892110062876e-05, "loss": 0.6914, "step": 15020 }, { "epoch": 1.51, "learning_rate": 4.094166382599692e-05, "loss": 0.1868, "step": 15021 }, { "epoch": 1.51, "learning_rate": 4.093943547230238e-05, "loss": 0.8196, "step": 15022 }, { "epoch": 1.51, "learning_rate": 4.0937207048993433e-05, "loss": 0.8728, "step": 15023 }, { "epoch": 1.51, "learning_rate": 4.093497855608426e-05, "loss": 0.4912, "step": 15024 }, { "epoch": 1.51, "learning_rate": 4.0932749993589043e-05, "loss": 0.5555, "step": 15025 }, { "epoch": 1.51, "learning_rate": 4.093052136152197e-05, "loss": 0.8471, "step": 15026 }, { "epoch": 1.51, "learning_rate": 4.09282926598972e-05, "loss": 1.2049, "step": 15027 }, { "epoch": 1.51, "learning_rate": 4.092606388872895e-05, "loss": 0.2906, "step": 15028 }, { "epoch": 1.51, "learning_rate": 4.092383504803138e-05, "loss": 0.3948, "step": 15029 }, { "epoch": 1.51, "learning_rate": 4.0921606137818686e-05, "loss": 0.4816, "step": 15030 }, { "epoch": 1.51, "learning_rate": 4.091937715810503e-05, "loss": 0.9689, "step": 15031 }, { "epoch": 1.51, "learning_rate": 4.091714810890463e-05, "loss": 0.7077, "step": 15032 }, { "epoch": 1.51, "learning_rate": 4.091491899023164e-05, "loss": 1.2323, "step": 15033 }, { "epoch": 1.51, "learning_rate": 4.091268980210026e-05, "loss": 0.8624, "step": 15034 }, { "epoch": 1.51, "learning_rate": 4.091046054452468e-05, "loss": 1.1944, "step": 15035 }, { "epoch": 1.51, "learning_rate": 4.090823121751907e-05, "loss": 0.643, "step": 15036 }, { "epoch": 1.51, "learning_rate": 4.090600182109763e-05, "loss": 0.3863, "step": 15037 }, { "epoch": 1.51, "learning_rate": 4.090377235527454e-05, "loss": 0.4444, "step": 15038 }, { "epoch": 1.51, "learning_rate": 4.0901542820064004e-05, "loss": 1.0131, "step": 15039 }, { "epoch": 1.51, "learning_rate": 4.089931321548018e-05, "loss": 0.3517, "step": 15040 }, { "epoch": 1.51, "learning_rate": 4.089708354153728e-05, "loss": 0.8279, "step": 15041 }, { "epoch": 1.51, "learning_rate": 4.089485379824949e-05, "loss": 1.1284, "step": 15042 }, { "epoch": 1.51, "learning_rate": 4.0892623985630995e-05, "loss": 0.1663, "step": 15043 }, { "epoch": 1.51, "learning_rate": 4.089039410369598e-05, "loss": 0.3341, "step": 15044 }, { "epoch": 1.51, "learning_rate": 4.088816415245864e-05, "loss": 0.895, "step": 15045 }, { "epoch": 1.51, "learning_rate": 4.088593413193317e-05, "loss": 0.8833, "step": 15046 }, { "epoch": 1.51, "learning_rate": 4.088370404213375e-05, "loss": 0.1469, "step": 15047 }, { "epoch": 1.52, "learning_rate": 4.088147388307458e-05, "loss": 0.7245, "step": 15048 }, { "epoch": 1.52, "learning_rate": 4.087924365476985e-05, "loss": 0.1969, "step": 15049 }, { "epoch": 1.52, "learning_rate": 4.087701335723375e-05, "loss": 0.5103, "step": 15050 }, { "epoch": 1.52, "learning_rate": 4.087478299048047e-05, "loss": 0.9255, "step": 15051 }, { "epoch": 1.52, "learning_rate": 4.087255255452423e-05, "loss": 0.3572, "step": 15052 }, { "epoch": 1.52, "learning_rate": 4.087032204937918e-05, "loss": 0.6497, "step": 15053 }, { "epoch": 1.52, "learning_rate": 4.086809147505954e-05, "loss": 1.1601, "step": 15054 }, { "epoch": 1.52, "learning_rate": 4.086586083157951e-05, "loss": 0.5877, "step": 15055 }, { "epoch": 1.52, "learning_rate": 4.0863630118953257e-05, "loss": 0.7034, "step": 15056 }, { "epoch": 1.52, "learning_rate": 4.0861399337195006e-05, "loss": 0.6478, "step": 15057 }, { "epoch": 1.52, "learning_rate": 4.0859168486318955e-05, "loss": 0.6555, "step": 15058 }, { "epoch": 1.52, "learning_rate": 4.0856937566339266e-05, "loss": 0.7427, "step": 15059 }, { "epoch": 1.52, "learning_rate": 4.085470657727016e-05, "loss": 0.7375, "step": 15060 }, { "epoch": 1.52, "learning_rate": 4.085247551912584e-05, "loss": 0.5026, "step": 15061 }, { "epoch": 1.52, "learning_rate": 4.085024439192049e-05, "loss": 0.9106, "step": 15062 }, { "epoch": 1.52, "learning_rate": 4.0848013195668305e-05, "loss": 0.3013, "step": 15063 }, { "epoch": 1.52, "learning_rate": 4.084578193038349e-05, "loss": 1.111, "step": 15064 }, { "epoch": 1.52, "learning_rate": 4.084355059608025e-05, "loss": 0.2069, "step": 15065 }, { "epoch": 1.52, "learning_rate": 4.084131919277278e-05, "loss": 0.7331, "step": 15066 }, { "epoch": 1.52, "learning_rate": 4.083908772047529e-05, "loss": 0.511, "step": 15067 }, { "epoch": 1.52, "learning_rate": 4.083685617920196e-05, "loss": 1.3519, "step": 15068 }, { "epoch": 1.52, "learning_rate": 4.0834624568966995e-05, "loss": 0.1349, "step": 15069 }, { "epoch": 1.52, "learning_rate": 4.08323928897846e-05, "loss": 1.1035, "step": 15070 }, { "epoch": 1.52, "learning_rate": 4.083016114166899e-05, "loss": 0.5988, "step": 15071 }, { "epoch": 1.52, "learning_rate": 4.0827929324634344e-05, "loss": 0.8466, "step": 15072 }, { "epoch": 1.52, "learning_rate": 4.082569743869488e-05, "loss": 0.8058, "step": 15073 }, { "epoch": 1.52, "learning_rate": 4.08234654838648e-05, "loss": 0.3257, "step": 15074 }, { "epoch": 1.52, "learning_rate": 4.0821233460158284e-05, "loss": 0.3431, "step": 15075 }, { "epoch": 1.52, "learning_rate": 4.081900136758957e-05, "loss": 0.6136, "step": 15076 }, { "epoch": 1.52, "learning_rate": 4.081676920617284e-05, "loss": 0.8552, "step": 15077 }, { "epoch": 1.52, "learning_rate": 4.081453697592232e-05, "loss": 0.7712, "step": 15078 }, { "epoch": 1.52, "learning_rate": 4.081230467685218e-05, "loss": 0.866, "step": 15079 }, { "epoch": 1.52, "learning_rate": 4.0810072308976654e-05, "loss": 0.3221, "step": 15080 }, { "epoch": 1.52, "learning_rate": 4.080783987230995e-05, "loss": 1.0865, "step": 15081 }, { "epoch": 1.52, "learning_rate": 4.080560736686626e-05, "loss": 1.0406, "step": 15082 }, { "epoch": 1.52, "learning_rate": 4.0803374792659793e-05, "loss": 0.4498, "step": 15083 }, { "epoch": 1.52, "learning_rate": 4.080114214970476e-05, "loss": 0.3608, "step": 15084 }, { "epoch": 1.52, "learning_rate": 4.079890943801536e-05, "loss": 0.1797, "step": 15085 }, { "epoch": 1.52, "learning_rate": 4.079667665760581e-05, "loss": 1.0725, "step": 15086 }, { "epoch": 1.52, "learning_rate": 4.079444380849033e-05, "loss": 0.8208, "step": 15087 }, { "epoch": 1.52, "learning_rate": 4.07922108906831e-05, "loss": 0.5031, "step": 15088 }, { "epoch": 1.52, "learning_rate": 4.078997790419836e-05, "loss": 0.5309, "step": 15089 }, { "epoch": 1.52, "learning_rate": 4.07877448490503e-05, "loss": 0.3726, "step": 15090 }, { "epoch": 1.52, "learning_rate": 4.078551172525313e-05, "loss": 0.6187, "step": 15091 }, { "epoch": 1.52, "learning_rate": 4.078327853282107e-05, "loss": 0.395, "step": 15092 }, { "epoch": 1.52, "learning_rate": 4.078104527176832e-05, "loss": 0.8994, "step": 15093 }, { "epoch": 1.52, "learning_rate": 4.077881194210912e-05, "loss": 0.5983, "step": 15094 }, { "epoch": 1.52, "learning_rate": 4.077657854385764e-05, "loss": 1.1282, "step": 15095 }, { "epoch": 1.52, "learning_rate": 4.077434507702812e-05, "loss": 0.4845, "step": 15096 }, { "epoch": 1.52, "learning_rate": 4.077211154163477e-05, "loss": 0.3973, "step": 15097 }, { "epoch": 1.52, "learning_rate": 4.07698779376918e-05, "loss": 1.0463, "step": 15098 }, { "epoch": 1.52, "learning_rate": 4.076764426521343e-05, "loss": 0.7615, "step": 15099 }, { "epoch": 1.52, "learning_rate": 4.076541052421386e-05, "loss": 0.339, "step": 15100 }, { "epoch": 1.52, "learning_rate": 4.076317671470731e-05, "loss": 0.2704, "step": 15101 }, { "epoch": 1.52, "learning_rate": 4.0760942836707995e-05, "loss": 0.6052, "step": 15102 }, { "epoch": 1.52, "learning_rate": 4.0758708890230145e-05, "loss": 1.5864, "step": 15103 }, { "epoch": 1.52, "learning_rate": 4.075647487528796e-05, "loss": 0.7222, "step": 15104 }, { "epoch": 1.52, "learning_rate": 4.0754240791895656e-05, "loss": 1.0878, "step": 15105 }, { "epoch": 1.52, "learning_rate": 4.0752006640067455e-05, "loss": 1.2554, "step": 15106 }, { "epoch": 1.52, "learning_rate": 4.0749772419817587e-05, "loss": 0.6261, "step": 15107 }, { "epoch": 1.52, "learning_rate": 4.0747538131160246e-05, "loss": 0.2563, "step": 15108 }, { "epoch": 1.52, "learning_rate": 4.074530377410966e-05, "loss": 0.2846, "step": 15109 }, { "epoch": 1.52, "learning_rate": 4.074306934868006e-05, "loss": 0.72, "step": 15110 }, { "epoch": 1.52, "learning_rate": 4.074083485488564e-05, "loss": 0.9746, "step": 15111 }, { "epoch": 1.52, "learning_rate": 4.0738600292740634e-05, "loss": 0.8406, "step": 15112 }, { "epoch": 1.52, "learning_rate": 4.0736365662259265e-05, "loss": 0.4707, "step": 15113 }, { "epoch": 1.52, "learning_rate": 4.073413096345575e-05, "loss": 0.3327, "step": 15114 }, { "epoch": 1.52, "learning_rate": 4.073189619634431e-05, "loss": 0.3417, "step": 15115 }, { "epoch": 1.52, "learning_rate": 4.072966136093917e-05, "loss": 1.153, "step": 15116 }, { "epoch": 1.52, "learning_rate": 4.072742645725454e-05, "loss": 1.0435, "step": 15117 }, { "epoch": 1.52, "learning_rate": 4.072519148530464e-05, "loss": 0.1592, "step": 15118 }, { "epoch": 1.52, "learning_rate": 4.072295644510371e-05, "loss": 0.7109, "step": 15119 }, { "epoch": 1.52, "learning_rate": 4.0720721336665966e-05, "loss": 0.741, "step": 15120 }, { "epoch": 1.52, "learning_rate": 4.071848616000562e-05, "loss": 1.0084, "step": 15121 }, { "epoch": 1.52, "learning_rate": 4.071625091513692e-05, "loss": 0.298, "step": 15122 }, { "epoch": 1.52, "learning_rate": 4.071401560207407e-05, "loss": 0.761, "step": 15123 }, { "epoch": 1.52, "learning_rate": 4.0711780220831305e-05, "loss": 0.3779, "step": 15124 }, { "epoch": 1.52, "learning_rate": 4.070954477142284e-05, "loss": 0.749, "step": 15125 }, { "epoch": 1.52, "learning_rate": 4.070730925386291e-05, "loss": 0.2994, "step": 15126 }, { "epoch": 1.52, "learning_rate": 4.070507366816573e-05, "loss": 0.3313, "step": 15127 }, { "epoch": 1.52, "learning_rate": 4.0702838014345536e-05, "loss": 0.5921, "step": 15128 }, { "epoch": 1.52, "learning_rate": 4.070060229241656e-05, "loss": 0.6425, "step": 15129 }, { "epoch": 1.52, "learning_rate": 4.069836650239302e-05, "loss": 0.493, "step": 15130 }, { "epoch": 1.52, "learning_rate": 4.0696130644289134e-05, "loss": 0.3813, "step": 15131 }, { "epoch": 1.52, "learning_rate": 4.069389471811915e-05, "loss": 0.8365, "step": 15132 }, { "epoch": 1.52, "learning_rate": 4.069165872389729e-05, "loss": 0.2672, "step": 15133 }, { "epoch": 1.52, "learning_rate": 4.0689422661637775e-05, "loss": 0.1935, "step": 15134 }, { "epoch": 1.52, "learning_rate": 4.068718653135485e-05, "loss": 0.3512, "step": 15135 }, { "epoch": 1.52, "learning_rate": 4.0684950333062724e-05, "loss": 1.5996, "step": 15136 }, { "epoch": 1.52, "learning_rate": 4.068271406677565e-05, "loss": 1.0321, "step": 15137 }, { "epoch": 1.52, "learning_rate": 4.068047773250785e-05, "loss": 0.5479, "step": 15138 }, { "epoch": 1.52, "learning_rate": 4.0678241330273533e-05, "loss": 1.0869, "step": 15139 }, { "epoch": 1.52, "learning_rate": 4.067600486008697e-05, "loss": 0.2502, "step": 15140 }, { "epoch": 1.52, "learning_rate": 4.0673768321962355e-05, "loss": 0.449, "step": 15141 }, { "epoch": 1.52, "learning_rate": 4.067153171591396e-05, "loss": 0.2543, "step": 15142 }, { "epoch": 1.52, "learning_rate": 4.066929504195598e-05, "loss": 1.1222, "step": 15143 }, { "epoch": 1.52, "learning_rate": 4.066705830010267e-05, "loss": 0.8642, "step": 15144 }, { "epoch": 1.52, "learning_rate": 4.066482149036826e-05, "loss": 0.3837, "step": 15145 }, { "epoch": 1.52, "learning_rate": 4.066258461276699e-05, "loss": 0.4458, "step": 15146 }, { "epoch": 1.53, "learning_rate": 4.0660347667313084e-05, "loss": 0.7115, "step": 15147 }, { "epoch": 1.53, "learning_rate": 4.065811065402077e-05, "loss": 0.4186, "step": 15148 }, { "epoch": 1.53, "learning_rate": 4.065587357290431e-05, "loss": 0.3769, "step": 15149 }, { "epoch": 1.53, "learning_rate": 4.065363642397791e-05, "loss": 0.1451, "step": 15150 }, { "epoch": 1.53, "learning_rate": 4.0651399207255836e-05, "loss": 1.2595, "step": 15151 }, { "epoch": 1.53, "learning_rate": 4.06491619227523e-05, "loss": 1.1401, "step": 15152 }, { "epoch": 1.53, "learning_rate": 4.0646924570481546e-05, "loss": 0.3488, "step": 15153 }, { "epoch": 1.53, "learning_rate": 4.0644687150457824e-05, "loss": 1.3225, "step": 15154 }, { "epoch": 1.53, "learning_rate": 4.064244966269535e-05, "loss": 0.6573, "step": 15155 }, { "epoch": 1.53, "learning_rate": 4.064021210720839e-05, "loss": 0.4657, "step": 15156 }, { "epoch": 1.53, "learning_rate": 4.0637974484011155e-05, "loss": 0.7929, "step": 15157 }, { "epoch": 1.53, "learning_rate": 4.06357367931179e-05, "loss": 0.2164, "step": 15158 }, { "epoch": 1.53, "learning_rate": 4.0633499034542865e-05, "loss": 0.7743, "step": 15159 }, { "epoch": 1.53, "learning_rate": 4.063126120830029e-05, "loss": 0.2367, "step": 15160 }, { "epoch": 1.53, "learning_rate": 4.06290233144044e-05, "loss": 0.3127, "step": 15161 }, { "epoch": 1.53, "learning_rate": 4.062678535286947e-05, "loss": 1.1654, "step": 15162 }, { "epoch": 1.53, "learning_rate": 4.0624547323709706e-05, "loss": 0.2264, "step": 15163 }, { "epoch": 1.53, "learning_rate": 4.062230922693937e-05, "loss": 0.4406, "step": 15164 }, { "epoch": 1.53, "learning_rate": 4.06200710625727e-05, "loss": 0.2299, "step": 15165 }, { "epoch": 1.53, "learning_rate": 4.0617832830623924e-05, "loss": 0.5119, "step": 15166 }, { "epoch": 1.53, "learning_rate": 4.061559453110731e-05, "loss": 1.3632, "step": 15167 }, { "epoch": 1.53, "learning_rate": 4.0613356164037084e-05, "loss": 0.1576, "step": 15168 }, { "epoch": 1.53, "learning_rate": 4.061111772942751e-05, "loss": 0.7767, "step": 15169 }, { "epoch": 1.53, "learning_rate": 4.0608879227292814e-05, "loss": 0.3711, "step": 15170 }, { "epoch": 1.53, "learning_rate": 4.0606640657647244e-05, "loss": 0.9382, "step": 15171 }, { "epoch": 1.53, "learning_rate": 4.0604402020505054e-05, "loss": 0.9281, "step": 15172 }, { "epoch": 1.53, "learning_rate": 4.0602163315880474e-05, "loss": 0.753, "step": 15173 }, { "epoch": 1.53, "learning_rate": 4.059992454378776e-05, "loss": 0.7581, "step": 15174 }, { "epoch": 1.53, "learning_rate": 4.059768570424117e-05, "loss": 0.6525, "step": 15175 }, { "epoch": 1.53, "learning_rate": 4.059544679725493e-05, "loss": 1.2217, "step": 15176 }, { "epoch": 1.53, "learning_rate": 4.05932078228433e-05, "loss": 0.7834, "step": 15177 }, { "epoch": 1.53, "learning_rate": 4.059096878102053e-05, "loss": 1.003, "step": 15178 }, { "epoch": 1.53, "learning_rate": 4.058872967180086e-05, "loss": 0.9678, "step": 15179 }, { "epoch": 1.53, "learning_rate": 4.058649049519853e-05, "loss": 0.3568, "step": 15180 }, { "epoch": 1.53, "learning_rate": 4.058425125122783e-05, "loss": 0.346, "step": 15181 }, { "epoch": 1.53, "learning_rate": 4.0582011939902956e-05, "loss": 1.5671, "step": 15182 }, { "epoch": 1.53, "learning_rate": 4.0579772561238194e-05, "loss": 0.546, "step": 15183 }, { "epoch": 1.53, "learning_rate": 4.057753311524778e-05, "loss": 0.8092, "step": 15184 }, { "epoch": 1.53, "learning_rate": 4.057529360194597e-05, "loss": 0.8439, "step": 15185 }, { "epoch": 1.53, "learning_rate": 4.0573054021347026e-05, "loss": 0.7574, "step": 15186 }, { "epoch": 1.53, "learning_rate": 4.057081437346517e-05, "loss": 0.4083, "step": 15187 }, { "epoch": 1.53, "learning_rate": 4.056857465831469e-05, "loss": 0.6711, "step": 15188 }, { "epoch": 1.53, "learning_rate": 4.05663348759098e-05, "loss": 0.6795, "step": 15189 }, { "epoch": 1.53, "learning_rate": 4.056409502626479e-05, "loss": 0.2989, "step": 15190 }, { "epoch": 1.53, "learning_rate": 4.05618551093939e-05, "loss": 0.5815, "step": 15191 }, { "epoch": 1.53, "learning_rate": 4.055961512531138e-05, "loss": 0.2497, "step": 15192 }, { "epoch": 1.53, "learning_rate": 4.055737507403148e-05, "loss": 0.8936, "step": 15193 }, { "epoch": 1.53, "learning_rate": 4.055513495556847e-05, "loss": 0.8633, "step": 15194 }, { "epoch": 1.53, "learning_rate": 4.05528947699366e-05, "loss": 1.0105, "step": 15195 }, { "epoch": 1.53, "learning_rate": 4.05506545171501e-05, "loss": 0.1949, "step": 15196 }, { "epoch": 1.53, "learning_rate": 4.0548414197223264e-05, "loss": 0.877, "step": 15197 }, { "epoch": 1.53, "learning_rate": 4.0546173810170325e-05, "loss": 1.1401, "step": 15198 }, { "epoch": 1.53, "learning_rate": 4.0543933356005556e-05, "loss": 0.9092, "step": 15199 }, { "epoch": 1.53, "learning_rate": 4.054169283474321e-05, "loss": 0.2154, "step": 15200 }, { "epoch": 1.53, "learning_rate": 4.0539452246397536e-05, "loss": 0.2875, "step": 15201 }, { "epoch": 1.53, "learning_rate": 4.0537211590982794e-05, "loss": 0.6325, "step": 15202 }, { "epoch": 1.53, "learning_rate": 4.0534970868513244e-05, "loss": 0.9953, "step": 15203 }, { "epoch": 1.53, "learning_rate": 4.053273007900316e-05, "loss": 0.4153, "step": 15204 }, { "epoch": 1.53, "learning_rate": 4.053048922246678e-05, "loss": 1.4539, "step": 15205 }, { "epoch": 1.53, "learning_rate": 4.052824829891837e-05, "loss": 0.1661, "step": 15206 }, { "epoch": 1.53, "learning_rate": 4.0526007308372194e-05, "loss": 0.5526, "step": 15207 }, { "epoch": 1.53, "learning_rate": 4.052376625084252e-05, "loss": 0.8958, "step": 15208 }, { "epoch": 1.53, "learning_rate": 4.0521525126343594e-05, "loss": 0.4744, "step": 15209 }, { "epoch": 1.53, "learning_rate": 4.051928393488968e-05, "loss": 0.4708, "step": 15210 }, { "epoch": 1.53, "learning_rate": 4.0517042676495054e-05, "loss": 0.4009, "step": 15211 }, { "epoch": 1.53, "learning_rate": 4.051480135117396e-05, "loss": 1.1638, "step": 15212 }, { "epoch": 1.53, "learning_rate": 4.051255995894068e-05, "loss": 0.5245, "step": 15213 }, { "epoch": 1.53, "learning_rate": 4.051031849980946e-05, "loss": 0.2659, "step": 15214 }, { "epoch": 1.53, "learning_rate": 4.050807697379457e-05, "loss": 0.5599, "step": 15215 }, { "epoch": 1.53, "learning_rate": 4.0505835380910286e-05, "loss": 0.8331, "step": 15216 }, { "epoch": 1.53, "learning_rate": 4.0503593721170855e-05, "loss": 0.5841, "step": 15217 }, { "epoch": 1.53, "learning_rate": 4.050135199459055e-05, "loss": 1.1623, "step": 15218 }, { "epoch": 1.53, "learning_rate": 4.0499110201183634e-05, "loss": 0.3539, "step": 15219 }, { "epoch": 1.53, "learning_rate": 4.049686834096439e-05, "loss": 0.4866, "step": 15220 }, { "epoch": 1.53, "learning_rate": 4.049462641394705e-05, "loss": 0.8484, "step": 15221 }, { "epoch": 1.53, "learning_rate": 4.049238442014591e-05, "loss": 0.2865, "step": 15222 }, { "epoch": 1.53, "learning_rate": 4.049014235957522e-05, "loss": 1.1168, "step": 15223 }, { "epoch": 1.53, "learning_rate": 4.048790023224926e-05, "loss": 0.7729, "step": 15224 }, { "epoch": 1.53, "learning_rate": 4.048565803818229e-05, "loss": 0.8587, "step": 15225 }, { "epoch": 1.53, "learning_rate": 4.048341577738858e-05, "loss": 1.1004, "step": 15226 }, { "epoch": 1.53, "learning_rate": 4.04811734498824e-05, "loss": 0.8278, "step": 15227 }, { "epoch": 1.53, "learning_rate": 4.047893105567802e-05, "loss": 0.9667, "step": 15228 }, { "epoch": 1.53, "learning_rate": 4.047668859478971e-05, "loss": 0.5115, "step": 15229 }, { "epoch": 1.53, "learning_rate": 4.047444606723174e-05, "loss": 0.3664, "step": 15230 }, { "epoch": 1.53, "learning_rate": 4.047220347301839e-05, "loss": 0.9503, "step": 15231 }, { "epoch": 1.53, "learning_rate": 4.04699608121639e-05, "loss": 0.3546, "step": 15232 }, { "epoch": 1.53, "learning_rate": 4.046771808468258e-05, "loss": 0.5746, "step": 15233 }, { "epoch": 1.53, "learning_rate": 4.046547529058868e-05, "loss": 0.2201, "step": 15234 }, { "epoch": 1.53, "learning_rate": 4.0463232429896465e-05, "loss": 0.8011, "step": 15235 }, { "epoch": 1.53, "learning_rate": 4.046098950262024e-05, "loss": 0.5483, "step": 15236 }, { "epoch": 1.53, "learning_rate": 4.045874650877424e-05, "loss": 0.4338, "step": 15237 }, { "epoch": 1.53, "learning_rate": 4.045650344837276e-05, "loss": 0.3351, "step": 15238 }, { "epoch": 1.53, "learning_rate": 4.045426032143007e-05, "loss": 0.6617, "step": 15239 }, { "epoch": 1.53, "learning_rate": 4.0452017127960446e-05, "loss": 0.5969, "step": 15240 }, { "epoch": 1.53, "learning_rate": 4.044977386797817e-05, "loss": 0.2257, "step": 15241 }, { "epoch": 1.53, "learning_rate": 4.044753054149748e-05, "loss": 0.8154, "step": 15242 }, { "epoch": 1.53, "learning_rate": 4.044528714853272e-05, "loss": 1.0244, "step": 15243 }, { "epoch": 1.53, "learning_rate": 4.0443043689098094e-05, "loss": 0.7509, "step": 15244 }, { "epoch": 1.54, "learning_rate": 4.044080016320792e-05, "loss": 0.5972, "step": 15245 }, { "epoch": 1.54, "learning_rate": 4.043855657087647e-05, "loss": 0.8107, "step": 15246 }, { "epoch": 1.54, "learning_rate": 4.0436312912118014e-05, "loss": 0.7444, "step": 15247 }, { "epoch": 1.54, "learning_rate": 4.0434069186946834e-05, "loss": 0.7658, "step": 15248 }, { "epoch": 1.54, "learning_rate": 4.043182539537721e-05, "loss": 1.0621, "step": 15249 }, { "epoch": 1.54, "learning_rate": 4.0429581537423416e-05, "loss": 0.8488, "step": 15250 }, { "epoch": 1.54, "learning_rate": 4.0427337613099726e-05, "loss": 0.4485, "step": 15251 }, { "epoch": 1.54, "learning_rate": 4.042509362242044e-05, "loss": 0.4211, "step": 15252 }, { "epoch": 1.54, "learning_rate": 4.042284956539981e-05, "loss": 0.398, "step": 15253 }, { "epoch": 1.54, "learning_rate": 4.0420605442052134e-05, "loss": 0.9937, "step": 15254 }, { "epoch": 1.54, "learning_rate": 4.0418361252391694e-05, "loss": 0.6972, "step": 15255 }, { "epoch": 1.54, "learning_rate": 4.041611699643276e-05, "loss": 0.5052, "step": 15256 }, { "epoch": 1.54, "learning_rate": 4.0413872674189626e-05, "loss": 0.6947, "step": 15257 }, { "epoch": 1.54, "learning_rate": 4.041162828567656e-05, "loss": 0.8422, "step": 15258 }, { "epoch": 1.54, "learning_rate": 4.0409383830907866e-05, "loss": 0.827, "step": 15259 }, { "epoch": 1.54, "learning_rate": 4.04071393098978e-05, "loss": 0.1744, "step": 15260 }, { "epoch": 1.54, "learning_rate": 4.040489472266066e-05, "loss": 1.8571, "step": 15261 }, { "epoch": 1.54, "learning_rate": 4.040265006921074e-05, "loss": 0.3169, "step": 15262 }, { "epoch": 1.54, "learning_rate": 4.0400405349562305e-05, "loss": 0.3518, "step": 15263 }, { "epoch": 1.54, "learning_rate": 4.039816056372964e-05, "loss": 0.6325, "step": 15264 }, { "epoch": 1.54, "learning_rate": 4.039591571172704e-05, "loss": 0.2136, "step": 15265 }, { "epoch": 1.54, "learning_rate": 4.0393670793568794e-05, "loss": 0.7384, "step": 15266 }, { "epoch": 1.54, "learning_rate": 4.039142580926917e-05, "loss": 0.7572, "step": 15267 }, { "epoch": 1.54, "learning_rate": 4.0389180758842474e-05, "loss": 0.8624, "step": 15268 }, { "epoch": 1.54, "learning_rate": 4.0386935642302984e-05, "loss": 1.2896, "step": 15269 }, { "epoch": 1.54, "learning_rate": 4.0384690459664986e-05, "loss": 0.1583, "step": 15270 }, { "epoch": 1.54, "learning_rate": 4.0382445210942766e-05, "loss": 0.8072, "step": 15271 }, { "epoch": 1.54, "learning_rate": 4.038019989615061e-05, "loss": 1.0147, "step": 15272 }, { "epoch": 1.54, "learning_rate": 4.037795451530282e-05, "loss": 0.2761, "step": 15273 }, { "epoch": 1.54, "learning_rate": 4.037570906841368e-05, "loss": 0.9005, "step": 15274 }, { "epoch": 1.54, "learning_rate": 4.037346355549746e-05, "loss": 1.2001, "step": 15275 }, { "epoch": 1.54, "learning_rate": 4.037121797656847e-05, "loss": 1.188, "step": 15276 }, { "epoch": 1.54, "learning_rate": 4.0368972331640994e-05, "loss": 0.9093, "step": 15277 }, { "epoch": 1.54, "learning_rate": 4.0366726620729324e-05, "loss": 0.595, "step": 15278 }, { "epoch": 1.54, "learning_rate": 4.036448084384775e-05, "loss": 0.6163, "step": 15279 }, { "epoch": 1.54, "learning_rate": 4.0362235001010565e-05, "loss": 0.4717, "step": 15280 }, { "epoch": 1.54, "learning_rate": 4.035998909223205e-05, "loss": 0.5592, "step": 15281 }, { "epoch": 1.54, "learning_rate": 4.035774311752651e-05, "loss": 0.1024, "step": 15282 }, { "epoch": 1.54, "learning_rate": 4.0355497076908235e-05, "loss": 0.4645, "step": 15283 }, { "epoch": 1.54, "learning_rate": 4.0353250970391514e-05, "loss": 0.2564, "step": 15284 }, { "epoch": 1.54, "learning_rate": 4.0351004797990646e-05, "loss": 0.2613, "step": 15285 }, { "epoch": 1.54, "learning_rate": 4.034875855971992e-05, "loss": 0.3657, "step": 15286 }, { "epoch": 1.54, "learning_rate": 4.034651225559363e-05, "loss": 0.2329, "step": 15287 }, { "epoch": 1.54, "learning_rate": 4.034426588562607e-05, "loss": 1.275, "step": 15288 }, { "epoch": 1.54, "learning_rate": 4.034201944983154e-05, "loss": 0.9726, "step": 15289 }, { "epoch": 1.54, "learning_rate": 4.033977294822433e-05, "loss": 0.3256, "step": 15290 }, { "epoch": 1.54, "learning_rate": 4.0337526380818745e-05, "loss": 0.9619, "step": 15291 }, { "epoch": 1.54, "learning_rate": 4.0335279747629066e-05, "loss": 0.5255, "step": 15292 }, { "epoch": 1.54, "learning_rate": 4.0333033048669604e-05, "loss": 0.3342, "step": 15293 }, { "epoch": 1.54, "learning_rate": 4.033078628395465e-05, "loss": 0.6592, "step": 15294 }, { "epoch": 1.54, "learning_rate": 4.032853945349849e-05, "loss": 0.6397, "step": 15295 }, { "epoch": 1.54, "learning_rate": 4.032629255731546e-05, "loss": 0.6451, "step": 15296 }, { "epoch": 1.54, "learning_rate": 4.032404559541981e-05, "loss": 0.4865, "step": 15297 }, { "epoch": 1.54, "learning_rate": 4.032179856782587e-05, "loss": 0.5988, "step": 15298 }, { "epoch": 1.54, "learning_rate": 4.0319551474547924e-05, "loss": 0.2889, "step": 15299 }, { "epoch": 1.54, "learning_rate": 4.031730431560028e-05, "loss": 0.9159, "step": 15300 }, { "epoch": 1.54, "learning_rate": 4.031505709099725e-05, "loss": 1.1071, "step": 15301 }, { "epoch": 1.54, "learning_rate": 4.03128098007531e-05, "loss": 1.1109, "step": 15302 }, { "epoch": 1.54, "learning_rate": 4.031056244488217e-05, "loss": 0.8947, "step": 15303 }, { "epoch": 1.54, "learning_rate": 4.030831502339874e-05, "loss": 0.6744, "step": 15304 }, { "epoch": 1.54, "learning_rate": 4.0306067536317105e-05, "loss": 0.4268, "step": 15305 }, { "epoch": 1.54, "learning_rate": 4.030381998365158e-05, "loss": 0.8266, "step": 15306 }, { "epoch": 1.54, "learning_rate": 4.030157236541646e-05, "loss": 0.4244, "step": 15307 }, { "epoch": 1.54, "learning_rate": 4.0299324681626055e-05, "loss": 0.5233, "step": 15308 }, { "epoch": 1.54, "learning_rate": 4.029707693229467e-05, "loss": 0.6697, "step": 15309 }, { "epoch": 1.54, "learning_rate": 4.029482911743661e-05, "loss": 0.5795, "step": 15310 }, { "epoch": 1.54, "learning_rate": 4.029258123706617e-05, "loss": 0.2064, "step": 15311 }, { "epoch": 1.54, "learning_rate": 4.0290333291197655e-05, "loss": 0.4622, "step": 15312 }, { "epoch": 1.54, "learning_rate": 4.028808527984537e-05, "loss": 0.7071, "step": 15313 }, { "epoch": 1.54, "learning_rate": 4.0285837203023634e-05, "loss": 0.2155, "step": 15314 }, { "epoch": 1.54, "learning_rate": 4.0283589060746735e-05, "loss": 0.8418, "step": 15315 }, { "epoch": 1.54, "learning_rate": 4.028134085302899e-05, "loss": 0.4375, "step": 15316 }, { "epoch": 1.54, "learning_rate": 4.02790925798847e-05, "loss": 0.6007, "step": 15317 }, { "epoch": 1.54, "learning_rate": 4.027684424132818e-05, "loss": 0.145, "step": 15318 }, { "epoch": 1.54, "learning_rate": 4.027459583737374e-05, "loss": 0.235, "step": 15319 }, { "epoch": 1.54, "learning_rate": 4.027234736803566e-05, "loss": 0.9172, "step": 15320 }, { "epoch": 1.54, "learning_rate": 4.027009883332829e-05, "loss": 0.1951, "step": 15321 }, { "epoch": 1.54, "learning_rate": 4.026785023326589e-05, "loss": 0.9672, "step": 15322 }, { "epoch": 1.54, "learning_rate": 4.0265601567862827e-05, "loss": 0.9667, "step": 15323 }, { "epoch": 1.54, "learning_rate": 4.026335283713337e-05, "loss": 1.1784, "step": 15324 }, { "epoch": 1.54, "learning_rate": 4.0261104041091836e-05, "loss": 1.2082, "step": 15325 }, { "epoch": 1.54, "learning_rate": 4.025885517975255e-05, "loss": 0.3122, "step": 15326 }, { "epoch": 1.54, "learning_rate": 4.02566062531298e-05, "loss": 1.1087, "step": 15327 }, { "epoch": 1.54, "learning_rate": 4.025435726123792e-05, "loss": 1.2076, "step": 15328 }, { "epoch": 1.54, "learning_rate": 4.0252108204091195e-05, "loss": 1.1413, "step": 15329 }, { "epoch": 1.54, "learning_rate": 4.0249859081703975e-05, "loss": 0.716, "step": 15330 }, { "epoch": 1.54, "learning_rate": 4.0247609894090535e-05, "loss": 0.3301, "step": 15331 }, { "epoch": 1.54, "learning_rate": 4.024536064126521e-05, "loss": 0.6434, "step": 15332 }, { "epoch": 1.54, "learning_rate": 4.024311132324231e-05, "loss": 1.0152, "step": 15333 }, { "epoch": 1.54, "learning_rate": 4.0240861940036145e-05, "loss": 0.629, "step": 15334 }, { "epoch": 1.54, "learning_rate": 4.023861249166103e-05, "loss": 0.4033, "step": 15335 }, { "epoch": 1.54, "learning_rate": 4.023636297813127e-05, "loss": 1.3163, "step": 15336 }, { "epoch": 1.54, "learning_rate": 4.023411339946121e-05, "loss": 1.1858, "step": 15337 }, { "epoch": 1.54, "learning_rate": 4.0231863755665125e-05, "loss": 1.176, "step": 15338 }, { "epoch": 1.54, "learning_rate": 4.022961404675736e-05, "loss": 0.7275, "step": 15339 }, { "epoch": 1.54, "learning_rate": 4.022736427275223e-05, "loss": 0.4926, "step": 15340 }, { "epoch": 1.54, "learning_rate": 4.022511443366404e-05, "loss": 0.9458, "step": 15341 }, { "epoch": 1.54, "learning_rate": 4.022286452950711e-05, "loss": 0.458, "step": 15342 }, { "epoch": 1.54, "learning_rate": 4.022061456029576e-05, "loss": 0.644, "step": 15343 }, { "epoch": 1.55, "learning_rate": 4.0218364526044305e-05, "loss": 0.612, "step": 15344 }, { "epoch": 1.55, "learning_rate": 4.021611442676706e-05, "loss": 0.9102, "step": 15345 }, { "epoch": 1.55, "learning_rate": 4.0213864262478366e-05, "loss": 0.7672, "step": 15346 }, { "epoch": 1.55, "learning_rate": 4.021161403319251e-05, "loss": 0.3456, "step": 15347 }, { "epoch": 1.55, "learning_rate": 4.020936373892383e-05, "loss": 0.5458, "step": 15348 }, { "epoch": 1.55, "learning_rate": 4.020711337968665e-05, "loss": 0.3512, "step": 15349 }, { "epoch": 1.55, "learning_rate": 4.020486295549528e-05, "loss": 0.8964, "step": 15350 }, { "epoch": 1.55, "learning_rate": 4.020261246636404e-05, "loss": 0.8881, "step": 15351 }, { "epoch": 1.55, "learning_rate": 4.020036191230725e-05, "loss": 1.032, "step": 15352 }, { "epoch": 1.55, "learning_rate": 4.0198111293339254e-05, "loss": 1.1237, "step": 15353 }, { "epoch": 1.55, "learning_rate": 4.019586060947434e-05, "loss": 0.7504, "step": 15354 }, { "epoch": 1.55, "learning_rate": 4.019360986072686e-05, "loss": 0.8716, "step": 15355 }, { "epoch": 1.55, "learning_rate": 4.019135904711113e-05, "loss": 1.014, "step": 15356 }, { "epoch": 1.55, "learning_rate": 4.0189108168641454e-05, "loss": 0.3183, "step": 15357 }, { "epoch": 1.55, "learning_rate": 4.018685722533218e-05, "loss": 0.2504, "step": 15358 }, { "epoch": 1.55, "learning_rate": 4.018460621719761e-05, "loss": 0.81, "step": 15359 }, { "epoch": 1.55, "learning_rate": 4.018235514425209e-05, "loss": 0.5329, "step": 15360 }, { "epoch": 1.55, "learning_rate": 4.018010400650993e-05, "loss": 0.8105, "step": 15361 }, { "epoch": 1.55, "learning_rate": 4.017785280398546e-05, "loss": 0.215, "step": 15362 }, { "epoch": 1.55, "learning_rate": 4.0175601536693016e-05, "loss": 0.8066, "step": 15363 }, { "epoch": 1.55, "learning_rate": 4.017335020464691e-05, "loss": 0.64, "step": 15364 }, { "epoch": 1.55, "learning_rate": 4.0171098807861475e-05, "loss": 0.5242, "step": 15365 }, { "epoch": 1.55, "learning_rate": 4.016884734635103e-05, "loss": 0.751, "step": 15366 }, { "epoch": 1.55, "learning_rate": 4.0166595820129916e-05, "loss": 0.1536, "step": 15367 }, { "epoch": 1.55, "learning_rate": 4.016434422921244e-05, "loss": 0.6474, "step": 15368 }, { "epoch": 1.55, "learning_rate": 4.0162092573612966e-05, "loss": 0.3074, "step": 15369 }, { "epoch": 1.55, "learning_rate": 4.0159840853345786e-05, "loss": 0.1158, "step": 15370 }, { "epoch": 1.55, "learning_rate": 4.0157589068425246e-05, "loss": 0.1638, "step": 15371 }, { "epoch": 1.55, "learning_rate": 4.015533721886568e-05, "loss": 0.324, "step": 15372 }, { "epoch": 1.55, "learning_rate": 4.01530853046814e-05, "loss": 0.9046, "step": 15373 }, { "epoch": 1.55, "learning_rate": 4.015083332588676e-05, "loss": 0.4755, "step": 15374 }, { "epoch": 1.55, "learning_rate": 4.014858128249606e-05, "loss": 0.5434, "step": 15375 }, { "epoch": 1.55, "learning_rate": 4.014632917452367e-05, "loss": 1.3929, "step": 15376 }, { "epoch": 1.55, "learning_rate": 4.0144077001983884e-05, "loss": 0.8044, "step": 15377 }, { "epoch": 1.55, "learning_rate": 4.014182476489106e-05, "loss": 0.8053, "step": 15378 }, { "epoch": 1.55, "learning_rate": 4.0139572463259525e-05, "loss": 0.3557, "step": 15379 }, { "epoch": 1.55, "learning_rate": 4.013732009710361e-05, "loss": 0.4171, "step": 15380 }, { "epoch": 1.55, "learning_rate": 4.013506766643764e-05, "loss": 0.8749, "step": 15381 }, { "epoch": 1.55, "learning_rate": 4.0132815171275955e-05, "loss": 0.3141, "step": 15382 }, { "epoch": 1.55, "learning_rate": 4.0130562611632883e-05, "loss": 0.4314, "step": 15383 }, { "epoch": 1.55, "learning_rate": 4.012830998752277e-05, "loss": 0.901, "step": 15384 }, { "epoch": 1.55, "learning_rate": 4.012605729895995e-05, "loss": 0.9834, "step": 15385 }, { "epoch": 1.55, "learning_rate": 4.012380454595874e-05, "loss": 0.5055, "step": 15386 }, { "epoch": 1.55, "learning_rate": 4.01215517285335e-05, "loss": 1.0752, "step": 15387 }, { "epoch": 1.55, "learning_rate": 4.0119298846698554e-05, "loss": 0.4546, "step": 15388 }, { "epoch": 1.55, "learning_rate": 4.0117045900468236e-05, "loss": 0.5643, "step": 15389 }, { "epoch": 1.55, "learning_rate": 4.0114792889856896e-05, "loss": 0.6166, "step": 15390 }, { "epoch": 1.55, "learning_rate": 4.011253981487885e-05, "loss": 1.0512, "step": 15391 }, { "epoch": 1.55, "learning_rate": 4.011028667554846e-05, "loss": 0.4983, "step": 15392 }, { "epoch": 1.55, "learning_rate": 4.010803347188004e-05, "loss": 2.189, "step": 15393 }, { "epoch": 1.55, "learning_rate": 4.010578020388794e-05, "loss": 1.4909, "step": 15394 }, { "epoch": 1.55, "learning_rate": 4.0103526871586506e-05, "loss": 0.3682, "step": 15395 }, { "epoch": 1.55, "learning_rate": 4.010127347499008e-05, "loss": 0.7144, "step": 15396 }, { "epoch": 1.55, "learning_rate": 4.0099020014112976e-05, "loss": 0.8799, "step": 15397 }, { "epoch": 1.55, "learning_rate": 4.0096766488969555e-05, "loss": 0.9407, "step": 15398 }, { "epoch": 1.55, "learning_rate": 4.009451289957416e-05, "loss": 0.4173, "step": 15399 }, { "epoch": 1.55, "learning_rate": 4.009225924594111e-05, "loss": 0.3367, "step": 15400 }, { "epoch": 1.55, "learning_rate": 4.0090005528084776e-05, "loss": 0.7218, "step": 15401 }, { "epoch": 1.55, "learning_rate": 4.0087751746019476e-05, "loss": 0.8447, "step": 15402 }, { "epoch": 1.55, "learning_rate": 4.008549789975956e-05, "loss": 0.3769, "step": 15403 }, { "epoch": 1.55, "learning_rate": 4.0083243989319385e-05, "loss": 1.1191, "step": 15404 }, { "epoch": 1.55, "learning_rate": 4.008099001471327e-05, "loss": 0.8177, "step": 15405 }, { "epoch": 1.55, "learning_rate": 4.007873597595558e-05, "loss": 0.2308, "step": 15406 }, { "epoch": 1.55, "learning_rate": 4.007648187306063e-05, "loss": 0.8036, "step": 15407 }, { "epoch": 1.55, "learning_rate": 4.0074227706042806e-05, "loss": 0.3091, "step": 15408 }, { "epoch": 1.55, "learning_rate": 4.007197347491641e-05, "loss": 0.4597, "step": 15409 }, { "epoch": 1.55, "learning_rate": 4.0069719179695816e-05, "loss": 0.4932, "step": 15410 }, { "epoch": 1.55, "learning_rate": 4.006746482039535e-05, "loss": 0.5953, "step": 15411 }, { "epoch": 1.55, "learning_rate": 4.0065210397029385e-05, "loss": 0.2426, "step": 15412 }, { "epoch": 1.55, "learning_rate": 4.006295590961224e-05, "loss": 0.5798, "step": 15413 }, { "epoch": 1.55, "learning_rate": 4.006070135815827e-05, "loss": 0.5192, "step": 15414 }, { "epoch": 1.55, "learning_rate": 4.0058446742681824e-05, "loss": 1.0169, "step": 15415 }, { "epoch": 1.55, "learning_rate": 4.005619206319725e-05, "loss": 0.8129, "step": 15416 }, { "epoch": 1.55, "learning_rate": 4.0053937319718895e-05, "loss": 0.4752, "step": 15417 }, { "epoch": 1.55, "learning_rate": 4.005168251226112e-05, "loss": 0.36, "step": 15418 }, { "epoch": 1.55, "learning_rate": 4.004942764083824e-05, "loss": 0.331, "step": 15419 }, { "epoch": 1.55, "learning_rate": 4.004717270546464e-05, "loss": 0.6596, "step": 15420 }, { "epoch": 1.55, "learning_rate": 4.004491770615465e-05, "loss": 0.3063, "step": 15421 }, { "epoch": 1.55, "learning_rate": 4.004266264292263e-05, "loss": 0.3004, "step": 15422 }, { "epoch": 1.55, "learning_rate": 4.0040407515782915e-05, "loss": 0.6365, "step": 15423 }, { "epoch": 1.55, "learning_rate": 4.003815232474987e-05, "loss": 0.5584, "step": 15424 }, { "epoch": 1.55, "learning_rate": 4.0035897069837845e-05, "loss": 0.8582, "step": 15425 }, { "epoch": 1.55, "learning_rate": 4.0033641751061195e-05, "loss": 0.5495, "step": 15426 }, { "epoch": 1.55, "learning_rate": 4.003138636843425e-05, "loss": 0.616, "step": 15427 }, { "epoch": 1.55, "learning_rate": 4.002913092197139e-05, "loss": 1.2474, "step": 15428 }, { "epoch": 1.55, "learning_rate": 4.002687541168695e-05, "loss": 0.9856, "step": 15429 }, { "epoch": 1.55, "learning_rate": 4.0024619837595294e-05, "loss": 0.5009, "step": 15430 }, { "epoch": 1.55, "learning_rate": 4.002236419971078e-05, "loss": 0.597, "step": 15431 }, { "epoch": 1.55, "learning_rate": 4.002010849804774e-05, "loss": 0.292, "step": 15432 }, { "epoch": 1.55, "learning_rate": 4.001785273262054e-05, "loss": 0.9259, "step": 15433 }, { "epoch": 1.55, "learning_rate": 4.0015596903443545e-05, "loss": 0.3881, "step": 15434 }, { "epoch": 1.55, "learning_rate": 4.00133410105311e-05, "loss": 0.7286, "step": 15435 }, { "epoch": 1.55, "learning_rate": 4.001108505389755e-05, "loss": 0.9392, "step": 15436 }, { "epoch": 1.55, "learning_rate": 4.000882903355728e-05, "loss": 0.1551, "step": 15437 }, { "epoch": 1.55, "learning_rate": 4.000657294952462e-05, "loss": 1.41, "step": 15438 }, { "epoch": 1.55, "learning_rate": 4.0004316801813934e-05, "loss": 0.5391, "step": 15439 }, { "epoch": 1.55, "learning_rate": 4.000206059043959e-05, "loss": 0.8884, "step": 15440 }, { "epoch": 1.55, "learning_rate": 3.9999804315415934e-05, "loss": 1.4717, "step": 15441 }, { "epoch": 1.56, "learning_rate": 3.999754797675733e-05, "loss": 0.8201, "step": 15442 }, { "epoch": 1.56, "learning_rate": 3.999529157447813e-05, "loss": 0.4994, "step": 15443 }, { "epoch": 1.56, "learning_rate": 3.99930351085927e-05, "loss": 1.4137, "step": 15444 }, { "epoch": 1.56, "learning_rate": 3.999077857911539e-05, "loss": 0.2554, "step": 15445 }, { "epoch": 1.56, "learning_rate": 3.998852198606056e-05, "loss": 0.316, "step": 15446 }, { "epoch": 1.56, "learning_rate": 3.99862653294426e-05, "loss": 1.1825, "step": 15447 }, { "epoch": 1.56, "learning_rate": 3.998400860927582e-05, "loss": 0.9246, "step": 15448 }, { "epoch": 1.56, "learning_rate": 3.9981751825574624e-05, "loss": 0.8125, "step": 15449 }, { "epoch": 1.56, "learning_rate": 3.997949497835335e-05, "loss": 0.7412, "step": 15450 }, { "epoch": 1.56, "learning_rate": 3.997723806762638e-05, "loss": 0.6702, "step": 15451 }, { "epoch": 1.56, "learning_rate": 3.997498109340804e-05, "loss": 0.1483, "step": 15452 }, { "epoch": 1.56, "learning_rate": 3.997272405571273e-05, "loss": 1.0974, "step": 15453 }, { "epoch": 1.56, "learning_rate": 3.9970466954554794e-05, "loss": 0.6658, "step": 15454 }, { "epoch": 1.56, "learning_rate": 3.99682097899486e-05, "loss": 0.3785, "step": 15455 }, { "epoch": 1.56, "learning_rate": 3.99659525619085e-05, "loss": 1.0165, "step": 15456 }, { "epoch": 1.56, "learning_rate": 3.996369527044888e-05, "loss": 0.515, "step": 15457 }, { "epoch": 1.56, "learning_rate": 3.99614379155841e-05, "loss": 0.1307, "step": 15458 }, { "epoch": 1.56, "learning_rate": 3.9959180497328504e-05, "loss": 0.9276, "step": 15459 }, { "epoch": 1.56, "learning_rate": 3.995692301569648e-05, "loss": 0.1683, "step": 15460 }, { "epoch": 1.56, "learning_rate": 3.995466547070239e-05, "loss": 0.6473, "step": 15461 }, { "epoch": 1.56, "learning_rate": 3.995240786236059e-05, "loss": 0.7603, "step": 15462 }, { "epoch": 1.56, "learning_rate": 3.995015019068546e-05, "loss": 0.8702, "step": 15463 }, { "epoch": 1.56, "learning_rate": 3.994789245569134e-05, "loss": 0.2883, "step": 15464 }, { "epoch": 1.56, "learning_rate": 3.994563465739263e-05, "loss": 0.938, "step": 15465 }, { "epoch": 1.56, "learning_rate": 3.994337679580369e-05, "loss": 0.9655, "step": 15466 }, { "epoch": 1.56, "learning_rate": 3.994111887093888e-05, "loss": 0.1955, "step": 15467 }, { "epoch": 1.56, "learning_rate": 3.993886088281256e-05, "loss": 0.9586, "step": 15468 }, { "epoch": 1.56, "learning_rate": 3.993660283143912e-05, "loss": 0.4843, "step": 15469 }, { "epoch": 1.56, "learning_rate": 3.9934344716832926e-05, "loss": 0.4497, "step": 15470 }, { "epoch": 1.56, "learning_rate": 3.9932086539008323e-05, "loss": 0.6422, "step": 15471 }, { "epoch": 1.56, "learning_rate": 3.9929828297979714e-05, "loss": 0.118, "step": 15472 }, { "epoch": 1.56, "learning_rate": 3.992756999376145e-05, "loss": 0.7841, "step": 15473 }, { "epoch": 1.56, "learning_rate": 3.9925311626367914e-05, "loss": 0.815, "step": 15474 }, { "epoch": 1.56, "learning_rate": 3.992305319581346e-05, "loss": 0.36, "step": 15475 }, { "epoch": 1.56, "learning_rate": 3.992079470211248e-05, "loss": 0.4376, "step": 15476 }, { "epoch": 1.56, "learning_rate": 3.9918536145279336e-05, "loss": 0.4668, "step": 15477 }, { "epoch": 1.56, "learning_rate": 3.9916277525328384e-05, "loss": 0.4088, "step": 15478 }, { "epoch": 1.56, "learning_rate": 3.991401884227404e-05, "loss": 0.798, "step": 15479 }, { "epoch": 1.56, "learning_rate": 3.991176009613063e-05, "loss": 0.6832, "step": 15480 }, { "epoch": 1.56, "learning_rate": 3.990950128691256e-05, "loss": 0.9144, "step": 15481 }, { "epoch": 1.56, "learning_rate": 3.9907242414634186e-05, "loss": 0.508, "step": 15482 }, { "epoch": 1.56, "learning_rate": 3.99049834793099e-05, "loss": 1.3003, "step": 15483 }, { "epoch": 1.56, "learning_rate": 3.990272448095407e-05, "loss": 0.3855, "step": 15484 }, { "epoch": 1.56, "learning_rate": 3.990046541958105e-05, "loss": 0.7422, "step": 15485 }, { "epoch": 1.56, "learning_rate": 3.989820629520525e-05, "loss": 0.6031, "step": 15486 }, { "epoch": 1.56, "learning_rate": 3.989594710784103e-05, "loss": 1.0093, "step": 15487 }, { "epoch": 1.56, "learning_rate": 3.989368785750277e-05, "loss": 1.3158, "step": 15488 }, { "epoch": 1.56, "learning_rate": 3.989142854420483e-05, "loss": 0.3861, "step": 15489 }, { "epoch": 1.56, "learning_rate": 3.988916916796161e-05, "loss": 0.4943, "step": 15490 }, { "epoch": 1.56, "learning_rate": 3.9886909728787484e-05, "loss": 0.6319, "step": 15491 }, { "epoch": 1.56, "learning_rate": 3.9884650226696826e-05, "loss": 0.9108, "step": 15492 }, { "epoch": 1.56, "learning_rate": 3.9882390661704004e-05, "loss": 0.2991, "step": 15493 }, { "epoch": 1.56, "learning_rate": 3.988013103382341e-05, "loss": 0.7473, "step": 15494 }, { "epoch": 1.56, "learning_rate": 3.987787134306943e-05, "loss": 0.6497, "step": 15495 }, { "epoch": 1.56, "learning_rate": 3.9875611589456427e-05, "loss": 0.9124, "step": 15496 }, { "epoch": 1.56, "learning_rate": 3.987335177299878e-05, "loss": 1.1427, "step": 15497 }, { "epoch": 1.56, "learning_rate": 3.9871091893710894e-05, "loss": 0.6996, "step": 15498 }, { "epoch": 1.56, "learning_rate": 3.986883195160713e-05, "loss": 0.8625, "step": 15499 }, { "epoch": 1.56, "learning_rate": 3.9866571946701874e-05, "loss": 0.3805, "step": 15500 }, { "epoch": 1.56, "learning_rate": 3.98643118790095e-05, "loss": 0.7869, "step": 15501 }, { "epoch": 1.56, "learning_rate": 3.986205174854441e-05, "loss": 0.5645, "step": 15502 }, { "epoch": 1.56, "learning_rate": 3.9859791555320956e-05, "loss": 0.8317, "step": 15503 }, { "epoch": 1.56, "learning_rate": 3.985753129935355e-05, "loss": 0.3712, "step": 15504 }, { "epoch": 1.56, "learning_rate": 3.985527098065658e-05, "loss": 0.6044, "step": 15505 }, { "epoch": 1.56, "learning_rate": 3.985301059924439e-05, "loss": 0.7571, "step": 15506 }, { "epoch": 1.56, "learning_rate": 3.9850750155131406e-05, "loss": 0.9506, "step": 15507 }, { "epoch": 1.56, "learning_rate": 3.984848964833199e-05, "loss": 0.1913, "step": 15508 }, { "epoch": 1.56, "learning_rate": 3.9846229078860527e-05, "loss": 0.4331, "step": 15509 }, { "epoch": 1.56, "learning_rate": 3.984396844673141e-05, "loss": 1.3718, "step": 15510 }, { "epoch": 1.56, "learning_rate": 3.984170775195903e-05, "loss": 1.0395, "step": 15511 }, { "epoch": 1.56, "learning_rate": 3.983944699455776e-05, "loss": 1.0657, "step": 15512 }, { "epoch": 1.56, "learning_rate": 3.983718617454199e-05, "loss": 0.6866, "step": 15513 }, { "epoch": 1.56, "learning_rate": 3.983492529192612e-05, "loss": 0.5593, "step": 15514 }, { "epoch": 1.56, "learning_rate": 3.983266434672451e-05, "loss": 0.756, "step": 15515 }, { "epoch": 1.56, "learning_rate": 3.983040333895158e-05, "loss": 0.2785, "step": 15516 }, { "epoch": 1.56, "learning_rate": 3.982814226862169e-05, "loss": 0.5449, "step": 15517 }, { "epoch": 1.56, "learning_rate": 3.9825881135749257e-05, "loss": 0.7359, "step": 15518 }, { "epoch": 1.56, "learning_rate": 3.982361994034864e-05, "loss": 0.646, "step": 15519 }, { "epoch": 1.56, "learning_rate": 3.982135868243425e-05, "loss": 0.5528, "step": 15520 }, { "epoch": 1.56, "learning_rate": 3.9819097362020463e-05, "loss": 0.2619, "step": 15521 }, { "epoch": 1.56, "learning_rate": 3.981683597912169e-05, "loss": 0.9977, "step": 15522 }, { "epoch": 1.56, "learning_rate": 3.9814574533752295e-05, "loss": 0.4366, "step": 15523 }, { "epoch": 1.56, "learning_rate": 3.9812313025926676e-05, "loss": 0.6945, "step": 15524 }, { "epoch": 1.56, "learning_rate": 3.9810051455659246e-05, "loss": 0.8484, "step": 15525 }, { "epoch": 1.56, "learning_rate": 3.980778982296437e-05, "loss": 0.3743, "step": 15526 }, { "epoch": 1.56, "learning_rate": 3.980552812785645e-05, "loss": 1.0806, "step": 15527 }, { "epoch": 1.56, "learning_rate": 3.980326637034988e-05, "loss": 0.547, "step": 15528 }, { "epoch": 1.56, "learning_rate": 3.980100455045906e-05, "loss": 0.653, "step": 15529 }, { "epoch": 1.56, "learning_rate": 3.979874266819837e-05, "loss": 0.592, "step": 15530 }, { "epoch": 1.56, "learning_rate": 3.979648072358221e-05, "loss": 1.1331, "step": 15531 }, { "epoch": 1.56, "learning_rate": 3.979421871662497e-05, "loss": 0.309, "step": 15532 }, { "epoch": 1.56, "learning_rate": 3.9791956647341055e-05, "loss": 0.641, "step": 15533 }, { "epoch": 1.56, "learning_rate": 3.978969451574485e-05, "loss": 1.013, "step": 15534 }, { "epoch": 1.56, "learning_rate": 3.9787432321850747e-05, "loss": 1.8678, "step": 15535 }, { "epoch": 1.56, "learning_rate": 3.978517006567315e-05, "loss": 0.6856, "step": 15536 }, { "epoch": 1.56, "learning_rate": 3.978290774722646e-05, "loss": 0.3393, "step": 15537 }, { "epoch": 1.56, "learning_rate": 3.978064536652506e-05, "loss": 1.2729, "step": 15538 }, { "epoch": 1.56, "learning_rate": 3.977838292358336e-05, "loss": 0.8479, "step": 15539 }, { "epoch": 1.56, "learning_rate": 3.9776120418415744e-05, "loss": 0.4662, "step": 15540 }, { "epoch": 1.57, "learning_rate": 3.977385785103663e-05, "loss": 0.874, "step": 15541 }, { "epoch": 1.57, "learning_rate": 3.9771595221460386e-05, "loss": 0.9616, "step": 15542 }, { "epoch": 1.57, "learning_rate": 3.976933252970144e-05, "loss": 0.7209, "step": 15543 }, { "epoch": 1.57, "learning_rate": 3.976706977577417e-05, "loss": 0.8852, "step": 15544 }, { "epoch": 1.57, "learning_rate": 3.9764806959692983e-05, "loss": 0.5632, "step": 15545 }, { "epoch": 1.57, "learning_rate": 3.976254408147229e-05, "loss": 0.8573, "step": 15546 }, { "epoch": 1.57, "learning_rate": 3.976028114112647e-05, "loss": 0.5825, "step": 15547 }, { "epoch": 1.57, "learning_rate": 3.9758018138669936e-05, "loss": 0.2447, "step": 15548 }, { "epoch": 1.57, "learning_rate": 3.9755755074117084e-05, "loss": 0.3486, "step": 15549 }, { "epoch": 1.57, "learning_rate": 3.975349194748232e-05, "loss": 1.3924, "step": 15550 }, { "epoch": 1.57, "learning_rate": 3.9751228758780046e-05, "loss": 1.3142, "step": 15551 }, { "epoch": 1.57, "learning_rate": 3.974896550802466e-05, "loss": 0.7223, "step": 15552 }, { "epoch": 1.57, "learning_rate": 3.9746702195230566e-05, "loss": 0.4395, "step": 15553 }, { "epoch": 1.57, "learning_rate": 3.9744438820412175e-05, "loss": 0.208, "step": 15554 }, { "epoch": 1.57, "learning_rate": 3.9742175383583876e-05, "loss": 0.5582, "step": 15555 }, { "epoch": 1.57, "learning_rate": 3.973991188476007e-05, "loss": 0.7436, "step": 15556 }, { "epoch": 1.57, "learning_rate": 3.973764832395518e-05, "loss": 1.1418, "step": 15557 }, { "epoch": 1.57, "learning_rate": 3.97353847011836e-05, "loss": 1.2178, "step": 15558 }, { "epoch": 1.57, "learning_rate": 3.973312101645973e-05, "loss": 1.3591, "step": 15559 }, { "epoch": 1.57, "learning_rate": 3.973085726979798e-05, "loss": 1.1363, "step": 15560 }, { "epoch": 1.57, "learning_rate": 3.9728593461212765e-05, "loss": 0.3692, "step": 15561 }, { "epoch": 1.57, "learning_rate": 3.9726329590718486e-05, "loss": 0.233, "step": 15562 }, { "epoch": 1.57, "learning_rate": 3.972406565832954e-05, "loss": 1.1672, "step": 15563 }, { "epoch": 1.57, "learning_rate": 3.972180166406033e-05, "loss": 1.3852, "step": 15564 }, { "epoch": 1.57, "learning_rate": 3.9719537607925276e-05, "loss": 0.6741, "step": 15565 }, { "epoch": 1.57, "learning_rate": 3.9717273489938784e-05, "loss": 0.9651, "step": 15566 }, { "epoch": 1.57, "learning_rate": 3.971500931011526e-05, "loss": 1.3183, "step": 15567 }, { "epoch": 1.57, "learning_rate": 3.9712745068469115e-05, "loss": 0.365, "step": 15568 }, { "epoch": 1.57, "learning_rate": 3.971048076501476e-05, "loss": 0.6067, "step": 15569 }, { "epoch": 1.57, "learning_rate": 3.9708216399766596e-05, "loss": 0.4046, "step": 15570 }, { "epoch": 1.57, "learning_rate": 3.970595197273903e-05, "loss": 0.9076, "step": 15571 }, { "epoch": 1.57, "learning_rate": 3.970368748394648e-05, "loss": 0.3906, "step": 15572 }, { "epoch": 1.57, "learning_rate": 3.9701422933403365e-05, "loss": 0.1562, "step": 15573 }, { "epoch": 1.57, "learning_rate": 3.969915832112407e-05, "loss": 0.2329, "step": 15574 }, { "epoch": 1.57, "learning_rate": 3.9696893647123024e-05, "loss": 0.4623, "step": 15575 }, { "epoch": 1.57, "learning_rate": 3.969462891141464e-05, "loss": 0.5058, "step": 15576 }, { "epoch": 1.57, "learning_rate": 3.9692364114013326e-05, "loss": 1.182, "step": 15577 }, { "epoch": 1.57, "learning_rate": 3.96900992549335e-05, "loss": 0.2968, "step": 15578 }, { "epoch": 1.57, "learning_rate": 3.968783433418955e-05, "loss": 0.2012, "step": 15579 }, { "epoch": 1.57, "learning_rate": 3.968556935179593e-05, "loss": 0.453, "step": 15580 }, { "epoch": 1.57, "learning_rate": 3.9683304307767013e-05, "loss": 1.0156, "step": 15581 }, { "epoch": 1.57, "learning_rate": 3.968103920211724e-05, "loss": 0.6993, "step": 15582 }, { "epoch": 1.57, "learning_rate": 3.9678774034861016e-05, "loss": 0.4786, "step": 15583 }, { "epoch": 1.57, "learning_rate": 3.9676508806012755e-05, "loss": 0.8157, "step": 15584 }, { "epoch": 1.57, "learning_rate": 3.967424351558687e-05, "loss": 0.3542, "step": 15585 }, { "epoch": 1.57, "learning_rate": 3.967197816359779e-05, "loss": 0.2293, "step": 15586 }, { "epoch": 1.57, "learning_rate": 3.9669712750059916e-05, "loss": 0.6349, "step": 15587 }, { "epoch": 1.57, "learning_rate": 3.9667447274987655e-05, "loss": 0.2783, "step": 15588 }, { "epoch": 1.57, "learning_rate": 3.966518173839546e-05, "loss": 0.3245, "step": 15589 }, { "epoch": 1.57, "learning_rate": 3.96629161402977e-05, "loss": 0.3292, "step": 15590 }, { "epoch": 1.57, "learning_rate": 3.966065048070883e-05, "loss": 0.9499, "step": 15591 }, { "epoch": 1.57, "learning_rate": 3.965838475964325e-05, "loss": 0.3988, "step": 15592 }, { "epoch": 1.57, "learning_rate": 3.9656118977115406e-05, "loss": 0.571, "step": 15593 }, { "epoch": 1.57, "learning_rate": 3.9653853133139676e-05, "loss": 0.3635, "step": 15594 }, { "epoch": 1.57, "learning_rate": 3.965158722773049e-05, "loss": 0.4719, "step": 15595 }, { "epoch": 1.57, "learning_rate": 3.964932126090229e-05, "loss": 1.1534, "step": 15596 }, { "epoch": 1.57, "learning_rate": 3.9647055232669467e-05, "loss": 0.8236, "step": 15597 }, { "epoch": 1.57, "learning_rate": 3.964478914304647e-05, "loss": 0.8781, "step": 15598 }, { "epoch": 1.57, "learning_rate": 3.964252299204769e-05, "loss": 0.3473, "step": 15599 }, { "epoch": 1.57, "learning_rate": 3.9640256779687574e-05, "loss": 1.4696, "step": 15600 }, { "epoch": 1.57, "learning_rate": 3.963799050598052e-05, "loss": 0.5953, "step": 15601 }, { "epoch": 1.57, "learning_rate": 3.963572417094097e-05, "loss": 0.9855, "step": 15602 }, { "epoch": 1.57, "learning_rate": 3.9633457774583334e-05, "loss": 0.5555, "step": 15603 }, { "epoch": 1.57, "learning_rate": 3.963119131692202e-05, "loss": 0.4636, "step": 15604 }, { "epoch": 1.57, "learning_rate": 3.9628924797971486e-05, "loss": 0.3532, "step": 15605 }, { "epoch": 1.57, "learning_rate": 3.962665821774614e-05, "loss": 0.447, "step": 15606 }, { "epoch": 1.57, "learning_rate": 3.9624391576260395e-05, "loss": 0.2662, "step": 15607 }, { "epoch": 1.57, "learning_rate": 3.962212487352869e-05, "loss": 0.2814, "step": 15608 }, { "epoch": 1.57, "learning_rate": 3.961985810956545e-05, "loss": 0.8914, "step": 15609 }, { "epoch": 1.57, "learning_rate": 3.961759128438508e-05, "loss": 1.0463, "step": 15610 }, { "epoch": 1.57, "learning_rate": 3.961532439800201e-05, "loss": 0.8493, "step": 15611 }, { "epoch": 1.57, "learning_rate": 3.961305745043069e-05, "loss": 0.892, "step": 15612 }, { "epoch": 1.57, "learning_rate": 3.961079044168552e-05, "loss": 1.0474, "step": 15613 }, { "epoch": 1.57, "learning_rate": 3.960852337178094e-05, "loss": 0.8002, "step": 15614 }, { "epoch": 1.57, "learning_rate": 3.960625624073138e-05, "loss": 1.0919, "step": 15615 }, { "epoch": 1.57, "learning_rate": 3.960398904855125e-05, "loss": 0.6809, "step": 15616 }, { "epoch": 1.57, "learning_rate": 3.960172179525498e-05, "loss": 1.1364, "step": 15617 }, { "epoch": 1.57, "learning_rate": 3.959945448085701e-05, "loss": 1.2824, "step": 15618 }, { "epoch": 1.57, "learning_rate": 3.959718710537177e-05, "loss": 0.9043, "step": 15619 }, { "epoch": 1.57, "learning_rate": 3.959491966881368e-05, "loss": 0.2423, "step": 15620 }, { "epoch": 1.57, "learning_rate": 3.959265217119717e-05, "loss": 0.3185, "step": 15621 }, { "epoch": 1.57, "learning_rate": 3.959038461253667e-05, "loss": 1.0875, "step": 15622 }, { "epoch": 1.57, "learning_rate": 3.958811699284661e-05, "loss": 1.0095, "step": 15623 }, { "epoch": 1.57, "learning_rate": 3.958584931214142e-05, "loss": 0.3721, "step": 15624 }, { "epoch": 1.57, "learning_rate": 3.9583581570435544e-05, "loss": 0.7048, "step": 15625 }, { "epoch": 1.57, "learning_rate": 3.9581313767743395e-05, "loss": 0.7796, "step": 15626 }, { "epoch": 1.57, "learning_rate": 3.9579045904079395e-05, "loss": 0.7727, "step": 15627 }, { "epoch": 1.57, "learning_rate": 3.957677797945801e-05, "loss": 0.3734, "step": 15628 }, { "epoch": 1.57, "learning_rate": 3.9574509993893644e-05, "loss": 0.4862, "step": 15629 }, { "epoch": 1.57, "learning_rate": 3.957224194740074e-05, "loss": 0.4135, "step": 15630 }, { "epoch": 1.57, "learning_rate": 3.956997383999373e-05, "loss": 1.3156, "step": 15631 }, { "epoch": 1.57, "learning_rate": 3.956770567168705e-05, "loss": 0.291, "step": 15632 }, { "epoch": 1.57, "learning_rate": 3.956543744249513e-05, "loss": 0.7498, "step": 15633 }, { "epoch": 1.57, "learning_rate": 3.95631691524324e-05, "loss": 0.634, "step": 15634 }, { "epoch": 1.57, "learning_rate": 3.9560900801513304e-05, "loss": 1.0638, "step": 15635 }, { "epoch": 1.57, "learning_rate": 3.955863238975227e-05, "loss": 1.1389, "step": 15636 }, { "epoch": 1.57, "learning_rate": 3.955636391716373e-05, "loss": 0.5159, "step": 15637 }, { "epoch": 1.57, "learning_rate": 3.9554095383762135e-05, "loss": 0.1691, "step": 15638 }, { "epoch": 1.58, "learning_rate": 3.955182678956191e-05, "loss": 1.4088, "step": 15639 }, { "epoch": 1.58, "learning_rate": 3.9549558134577486e-05, "loss": 0.703, "step": 15640 }, { "epoch": 1.58, "learning_rate": 3.9547289418823325e-05, "loss": 0.1029, "step": 15641 }, { "epoch": 1.58, "learning_rate": 3.954502064231383e-05, "loss": 0.7981, "step": 15642 }, { "epoch": 1.58, "learning_rate": 3.954275180506345e-05, "loss": 0.833, "step": 15643 }, { "epoch": 1.58, "learning_rate": 3.954048290708665e-05, "loss": 0.5285, "step": 15644 }, { "epoch": 1.58, "learning_rate": 3.953821394839782e-05, "loss": 0.4299, "step": 15645 }, { "epoch": 1.58, "learning_rate": 3.953594492901144e-05, "loss": 0.227, "step": 15646 }, { "epoch": 1.58, "learning_rate": 3.953367584894193e-05, "loss": 1.0094, "step": 15647 }, { "epoch": 1.58, "learning_rate": 3.953140670820374e-05, "loss": 1.0543, "step": 15648 }, { "epoch": 1.58, "learning_rate": 3.95291375068113e-05, "loss": 1.0681, "step": 15649 }, { "epoch": 1.58, "learning_rate": 3.9526868244779046e-05, "loss": 0.2588, "step": 15650 }, { "epoch": 1.58, "learning_rate": 3.9524598922121434e-05, "loss": 0.1954, "step": 15651 }, { "epoch": 1.58, "learning_rate": 3.9522329538852897e-05, "loss": 0.2466, "step": 15652 }, { "epoch": 1.58, "learning_rate": 3.952006009498788e-05, "loss": 0.8918, "step": 15653 }, { "epoch": 1.58, "learning_rate": 3.9517790590540816e-05, "loss": 0.4034, "step": 15654 }, { "epoch": 1.58, "learning_rate": 3.951552102552616e-05, "loss": 0.8257, "step": 15655 }, { "epoch": 1.58, "learning_rate": 3.9513251399958345e-05, "loss": 0.4982, "step": 15656 }, { "epoch": 1.58, "learning_rate": 3.951098171385182e-05, "loss": 0.3452, "step": 15657 }, { "epoch": 1.58, "learning_rate": 3.950871196722103e-05, "loss": 0.2667, "step": 15658 }, { "epoch": 1.58, "learning_rate": 3.95064421600804e-05, "loss": 0.5534, "step": 15659 }, { "epoch": 1.58, "learning_rate": 3.950417229244439e-05, "loss": 0.4323, "step": 15660 }, { "epoch": 1.58, "learning_rate": 3.9501902364327445e-05, "loss": 0.5343, "step": 15661 }, { "epoch": 1.58, "learning_rate": 3.9499632375744017e-05, "loss": 0.2493, "step": 15662 }, { "epoch": 1.58, "learning_rate": 3.949736232670853e-05, "loss": 0.7939, "step": 15663 }, { "epoch": 1.58, "learning_rate": 3.949509221723546e-05, "loss": 0.7336, "step": 15664 }, { "epoch": 1.58, "learning_rate": 3.9492822047339224e-05, "loss": 1.3302, "step": 15665 }, { "epoch": 1.58, "learning_rate": 3.9490551817034276e-05, "loss": 0.5083, "step": 15666 }, { "epoch": 1.58, "learning_rate": 3.948828152633508e-05, "loss": 0.4926, "step": 15667 }, { "epoch": 1.58, "learning_rate": 3.948601117525605e-05, "loss": 0.6341, "step": 15668 }, { "epoch": 1.58, "learning_rate": 3.948374076381167e-05, "loss": 0.9775, "step": 15669 }, { "epoch": 1.58, "learning_rate": 3.9481470292016366e-05, "loss": 1.1543, "step": 15670 }, { "epoch": 1.58, "learning_rate": 3.947919975988459e-05, "loss": 0.5472, "step": 15671 }, { "epoch": 1.58, "learning_rate": 3.9476929167430794e-05, "loss": 0.9545, "step": 15672 }, { "epoch": 1.58, "learning_rate": 3.9474658514669425e-05, "loss": 0.492, "step": 15673 }, { "epoch": 1.58, "learning_rate": 3.947238780161494e-05, "loss": 0.9232, "step": 15674 }, { "epoch": 1.58, "learning_rate": 3.947011702828177e-05, "loss": 0.2782, "step": 15675 }, { "epoch": 1.58, "learning_rate": 3.946784619468439e-05, "loss": 0.6478, "step": 15676 }, { "epoch": 1.58, "learning_rate": 3.946557530083724e-05, "loss": 0.322, "step": 15677 }, { "epoch": 1.58, "learning_rate": 3.946330434675476e-05, "loss": 0.876, "step": 15678 }, { "epoch": 1.58, "learning_rate": 3.946103333245141e-05, "loss": 0.4425, "step": 15679 }, { "epoch": 1.58, "learning_rate": 3.9458762257941656e-05, "loss": 2.0379, "step": 15680 }, { "epoch": 1.58, "learning_rate": 3.9456491123239925e-05, "loss": 0.9326, "step": 15681 }, { "epoch": 1.58, "learning_rate": 3.9454219928360685e-05, "loss": 0.7625, "step": 15682 }, { "epoch": 1.58, "learning_rate": 3.9451948673318394e-05, "loss": 0.1327, "step": 15683 }, { "epoch": 1.58, "learning_rate": 3.944967735812749e-05, "loss": 0.7528, "step": 15684 }, { "epoch": 1.58, "learning_rate": 3.9447405982802433e-05, "loss": 0.2047, "step": 15685 }, { "epoch": 1.58, "learning_rate": 3.944513454735768e-05, "loss": 1.2446, "step": 15686 }, { "epoch": 1.58, "learning_rate": 3.9442863051807686e-05, "loss": 0.6171, "step": 15687 }, { "epoch": 1.58, "learning_rate": 3.9440591496166895e-05, "loss": 0.758, "step": 15688 }, { "epoch": 1.58, "learning_rate": 3.943831988044978e-05, "loss": 0.8096, "step": 15689 }, { "epoch": 1.58, "learning_rate": 3.943604820467079e-05, "loss": 0.2613, "step": 15690 }, { "epoch": 1.58, "learning_rate": 3.943377646884437e-05, "loss": 0.3362, "step": 15691 }, { "epoch": 1.58, "learning_rate": 3.943150467298499e-05, "loss": 0.5727, "step": 15692 }, { "epoch": 1.58, "learning_rate": 3.94292328171071e-05, "loss": 0.7865, "step": 15693 }, { "epoch": 1.58, "learning_rate": 3.9426960901225164e-05, "loss": 0.5121, "step": 15694 }, { "epoch": 1.58, "learning_rate": 3.942468892535363e-05, "loss": 0.5826, "step": 15695 }, { "epoch": 1.58, "learning_rate": 3.942241688950697e-05, "loss": 1.2373, "step": 15696 }, { "epoch": 1.58, "learning_rate": 3.9420144793699625e-05, "loss": 0.2703, "step": 15697 }, { "epoch": 1.58, "learning_rate": 3.941787263794606e-05, "loss": 0.3556, "step": 15698 }, { "epoch": 1.58, "learning_rate": 3.9415600422260746e-05, "loss": 0.2229, "step": 15699 }, { "epoch": 1.58, "learning_rate": 3.9413328146658125e-05, "loss": 1.021, "step": 15700 }, { "epoch": 1.58, "learning_rate": 3.941105581115267e-05, "loss": 0.6438, "step": 15701 }, { "epoch": 1.58, "learning_rate": 3.9408783415758825e-05, "loss": 0.7064, "step": 15702 }, { "epoch": 1.58, "learning_rate": 3.9406510960491076e-05, "loss": 0.5773, "step": 15703 }, { "epoch": 1.58, "learning_rate": 3.940423844536386e-05, "loss": 1.0265, "step": 15704 }, { "epoch": 1.58, "learning_rate": 3.9401965870391654e-05, "loss": 0.7484, "step": 15705 }, { "epoch": 1.58, "learning_rate": 3.9399693235588915e-05, "loss": 0.4579, "step": 15706 }, { "epoch": 1.58, "learning_rate": 3.939742054097009e-05, "loss": 0.2258, "step": 15707 }, { "epoch": 1.58, "learning_rate": 3.939514778654967e-05, "loss": 0.3133, "step": 15708 }, { "epoch": 1.58, "learning_rate": 3.9392874972342096e-05, "loss": 0.8003, "step": 15709 }, { "epoch": 1.58, "learning_rate": 3.939060209836185e-05, "loss": 0.5836, "step": 15710 }, { "epoch": 1.58, "learning_rate": 3.9388329164623364e-05, "loss": 0.4891, "step": 15711 }, { "epoch": 1.58, "learning_rate": 3.938605617114114e-05, "loss": 0.4695, "step": 15712 }, { "epoch": 1.58, "learning_rate": 3.9383783117929627e-05, "loss": 0.612, "step": 15713 }, { "epoch": 1.58, "learning_rate": 3.938151000500328e-05, "loss": 0.2597, "step": 15714 }, { "epoch": 1.58, "learning_rate": 3.937923683237657e-05, "loss": 0.8433, "step": 15715 }, { "epoch": 1.58, "learning_rate": 3.937696360006397e-05, "loss": 0.9478, "step": 15716 }, { "epoch": 1.58, "learning_rate": 3.937469030807994e-05, "loss": 0.8277, "step": 15717 }, { "epoch": 1.58, "learning_rate": 3.937241695643894e-05, "loss": 0.7461, "step": 15718 }, { "epoch": 1.58, "learning_rate": 3.937014354515546e-05, "loss": 0.3769, "step": 15719 }, { "epoch": 1.58, "learning_rate": 3.936787007424393e-05, "loss": 1.0964, "step": 15720 }, { "epoch": 1.58, "learning_rate": 3.9365596543718855e-05, "loss": 0.4308, "step": 15721 }, { "epoch": 1.58, "learning_rate": 3.9363322953594696e-05, "loss": 0.5964, "step": 15722 }, { "epoch": 1.58, "learning_rate": 3.936104930388589e-05, "loss": 0.2072, "step": 15723 }, { "epoch": 1.58, "learning_rate": 3.9358775594606936e-05, "loss": 0.4609, "step": 15724 }, { "epoch": 1.58, "learning_rate": 3.9356501825772296e-05, "loss": 0.2834, "step": 15725 }, { "epoch": 1.58, "learning_rate": 3.935422799739643e-05, "loss": 0.8136, "step": 15726 }, { "epoch": 1.58, "learning_rate": 3.9351954109493814e-05, "loss": 0.6147, "step": 15727 }, { "epoch": 1.58, "learning_rate": 3.934968016207893e-05, "loss": 0.5918, "step": 15728 }, { "epoch": 1.58, "learning_rate": 3.934740615516624e-05, "loss": 1.0924, "step": 15729 }, { "epoch": 1.58, "learning_rate": 3.93451320887702e-05, "loss": 0.7444, "step": 15730 }, { "epoch": 1.58, "learning_rate": 3.934285796290531e-05, "loss": 0.6527, "step": 15731 }, { "epoch": 1.58, "learning_rate": 3.934058377758602e-05, "loss": 0.237, "step": 15732 }, { "epoch": 1.58, "learning_rate": 3.93383095328268e-05, "loss": 0.5566, "step": 15733 }, { "epoch": 1.58, "learning_rate": 3.933603522864214e-05, "loss": 0.9561, "step": 15734 }, { "epoch": 1.58, "learning_rate": 3.933376086504651e-05, "loss": 0.2312, "step": 15735 }, { "epoch": 1.58, "learning_rate": 3.933148644205435e-05, "loss": 1.188, "step": 15736 }, { "epoch": 1.58, "learning_rate": 3.932921195968018e-05, "loss": 0.9592, "step": 15737 }, { "epoch": 1.59, "learning_rate": 3.932693741793846e-05, "loss": 0.9905, "step": 15738 }, { "epoch": 1.59, "learning_rate": 3.9324662816843645e-05, "loss": 0.3695, "step": 15739 }, { "epoch": 1.59, "learning_rate": 3.9322388156410226e-05, "loss": 1.072, "step": 15740 }, { "epoch": 1.59, "learning_rate": 3.932011343665267e-05, "loss": 0.6949, "step": 15741 }, { "epoch": 1.59, "learning_rate": 3.931783865758548e-05, "loss": 0.3943, "step": 15742 }, { "epoch": 1.59, "learning_rate": 3.9315563819223086e-05, "loss": 0.5591, "step": 15743 }, { "epoch": 1.59, "learning_rate": 3.9313288921579994e-05, "loss": 0.6357, "step": 15744 }, { "epoch": 1.59, "learning_rate": 3.9311013964670685e-05, "loss": 0.3562, "step": 15745 }, { "epoch": 1.59, "learning_rate": 3.930873894850961e-05, "loss": 0.1928, "step": 15746 }, { "epoch": 1.59, "learning_rate": 3.930646387311127e-05, "loss": 0.7185, "step": 15747 }, { "epoch": 1.59, "learning_rate": 3.930418873849013e-05, "loss": 0.2671, "step": 15748 }, { "epoch": 1.59, "learning_rate": 3.930191354466068e-05, "loss": 0.5603, "step": 15749 }, { "epoch": 1.59, "learning_rate": 3.9299638291637384e-05, "loss": 0.5252, "step": 15750 }, { "epoch": 1.59, "learning_rate": 3.929736297943473e-05, "loss": 0.519, "step": 15751 }, { "epoch": 1.59, "learning_rate": 3.9295087608067186e-05, "loss": 0.7275, "step": 15752 }, { "epoch": 1.59, "learning_rate": 3.929281217754925e-05, "loss": 0.8731, "step": 15753 }, { "epoch": 1.59, "learning_rate": 3.92905366878954e-05, "loss": 0.3913, "step": 15754 }, { "epoch": 1.59, "learning_rate": 3.9288261139120094e-05, "loss": 0.3695, "step": 15755 }, { "epoch": 1.59, "learning_rate": 3.9285985531237836e-05, "loss": 0.5035, "step": 15756 }, { "epoch": 1.59, "learning_rate": 3.928370986426309e-05, "loss": 0.7706, "step": 15757 }, { "epoch": 1.59, "learning_rate": 3.928143413821036e-05, "loss": 0.7161, "step": 15758 }, { "epoch": 1.59, "learning_rate": 3.9279158353094093e-05, "loss": 0.8377, "step": 15759 }, { "epoch": 1.59, "learning_rate": 3.927688250892881e-05, "loss": 0.8802, "step": 15760 }, { "epoch": 1.59, "learning_rate": 3.927460660572898e-05, "loss": 0.2906, "step": 15761 }, { "epoch": 1.59, "learning_rate": 3.9272330643509065e-05, "loss": 0.6427, "step": 15762 }, { "epoch": 1.59, "learning_rate": 3.927005462228357e-05, "loss": 0.4141, "step": 15763 }, { "epoch": 1.59, "learning_rate": 3.926777854206698e-05, "loss": 0.5823, "step": 15764 }, { "epoch": 1.59, "learning_rate": 3.926550240287377e-05, "loss": 0.8107, "step": 15765 }, { "epoch": 1.59, "learning_rate": 3.926322620471842e-05, "loss": 0.8692, "step": 15766 }, { "epoch": 1.59, "learning_rate": 3.9260949947615436e-05, "loss": 0.6298, "step": 15767 }, { "epoch": 1.59, "learning_rate": 3.9258673631579285e-05, "loss": 0.2921, "step": 15768 }, { "epoch": 1.59, "learning_rate": 3.925639725662446e-05, "loss": 1.0035, "step": 15769 }, { "epoch": 1.59, "learning_rate": 3.925412082276544e-05, "loss": 0.7428, "step": 15770 }, { "epoch": 1.59, "learning_rate": 3.925184433001671e-05, "loss": 0.7633, "step": 15771 }, { "epoch": 1.59, "learning_rate": 3.924956777839277e-05, "loss": 1.0009, "step": 15772 }, { "epoch": 1.59, "learning_rate": 3.9247291167908095e-05, "loss": 0.6956, "step": 15773 }, { "epoch": 1.59, "learning_rate": 3.924501449857718e-05, "loss": 0.2614, "step": 15774 }, { "epoch": 1.59, "learning_rate": 3.9242737770414505e-05, "loss": 0.6046, "step": 15775 }, { "epoch": 1.59, "learning_rate": 3.924046098343457e-05, "loss": 0.7444, "step": 15776 }, { "epoch": 1.59, "learning_rate": 3.9238184137651856e-05, "loss": 0.4361, "step": 15777 }, { "epoch": 1.59, "learning_rate": 3.923590723308085e-05, "loss": 0.5654, "step": 15778 }, { "epoch": 1.59, "learning_rate": 3.923363026973604e-05, "loss": 1.075, "step": 15779 }, { "epoch": 1.59, "learning_rate": 3.923135324763192e-05, "loss": 0.5588, "step": 15780 }, { "epoch": 1.59, "learning_rate": 3.922907616678299e-05, "loss": 1.1613, "step": 15781 }, { "epoch": 1.59, "learning_rate": 3.922679902720371e-05, "loss": 0.2461, "step": 15782 }, { "epoch": 1.59, "learning_rate": 3.922452182890861e-05, "loss": 1.0599, "step": 15783 }, { "epoch": 1.59, "learning_rate": 3.922224457191216e-05, "loss": 0.5475, "step": 15784 }, { "epoch": 1.59, "learning_rate": 3.921996725622885e-05, "loss": 0.9168, "step": 15785 }, { "epoch": 1.59, "learning_rate": 3.921768988187317e-05, "loss": 0.6335, "step": 15786 }, { "epoch": 1.59, "learning_rate": 3.921541244885963e-05, "loss": 0.439, "step": 15787 }, { "epoch": 1.59, "learning_rate": 3.9213134957202706e-05, "loss": 0.2354, "step": 15788 }, { "epoch": 1.59, "learning_rate": 3.9210857406916894e-05, "loss": 0.8617, "step": 15789 }, { "epoch": 1.59, "learning_rate": 3.920857979801669e-05, "loss": 0.5987, "step": 15790 }, { "epoch": 1.59, "learning_rate": 3.9206302130516586e-05, "loss": 0.4358, "step": 15791 }, { "epoch": 1.59, "learning_rate": 3.920402440443108e-05, "loss": 0.4898, "step": 15792 }, { "epoch": 1.59, "learning_rate": 3.9201746619774665e-05, "loss": 1.0666, "step": 15793 }, { "epoch": 1.59, "learning_rate": 3.919946877656183e-05, "loss": 0.5821, "step": 15794 }, { "epoch": 1.59, "learning_rate": 3.919719087480708e-05, "loss": 1.0651, "step": 15795 }, { "epoch": 1.59, "learning_rate": 3.91949129145249e-05, "loss": 0.5582, "step": 15796 }, { "epoch": 1.59, "learning_rate": 3.91926348957298e-05, "loss": 1.3459, "step": 15797 }, { "epoch": 1.59, "learning_rate": 3.9190356818436264e-05, "loss": 1.0093, "step": 15798 }, { "epoch": 1.59, "learning_rate": 3.9188078682658796e-05, "loss": 0.434, "step": 15799 }, { "epoch": 1.59, "learning_rate": 3.918580048841189e-05, "loss": 0.8479, "step": 15800 }, { "epoch": 1.59, "learning_rate": 3.918352223571005e-05, "loss": 1.1429, "step": 15801 }, { "epoch": 1.59, "learning_rate": 3.9181243924567756e-05, "loss": 0.8316, "step": 15802 }, { "epoch": 1.59, "learning_rate": 3.9178965554999524e-05, "loss": 1.0188, "step": 15803 }, { "epoch": 1.59, "learning_rate": 3.917668712701985e-05, "loss": 0.2941, "step": 15804 }, { "epoch": 1.59, "learning_rate": 3.917440864064322e-05, "loss": 1.1261, "step": 15805 }, { "epoch": 1.59, "learning_rate": 3.917213009588416e-05, "loss": 0.434, "step": 15806 }, { "epoch": 1.59, "learning_rate": 3.916985149275714e-05, "loss": 1.1216, "step": 15807 }, { "epoch": 1.59, "learning_rate": 3.9167572831276675e-05, "loss": 0.6453, "step": 15808 }, { "epoch": 1.59, "learning_rate": 3.9165294111457264e-05, "loss": 0.1316, "step": 15809 }, { "epoch": 1.59, "learning_rate": 3.916301533331342e-05, "loss": 0.9739, "step": 15810 }, { "epoch": 1.59, "learning_rate": 3.9160736496859624e-05, "loss": 0.8007, "step": 15811 }, { "epoch": 1.59, "learning_rate": 3.915845760211038e-05, "loss": 1.0611, "step": 15812 }, { "epoch": 1.59, "learning_rate": 3.91561786490802e-05, "loss": 0.8642, "step": 15813 }, { "epoch": 1.59, "learning_rate": 3.915389963778358e-05, "loss": 1.1065, "step": 15814 }, { "epoch": 1.59, "learning_rate": 3.915162056823502e-05, "loss": 1.0285, "step": 15815 }, { "epoch": 1.59, "learning_rate": 3.914934144044905e-05, "loss": 0.4048, "step": 15816 }, { "epoch": 1.59, "learning_rate": 3.914706225444014e-05, "loss": 0.2327, "step": 15817 }, { "epoch": 1.59, "learning_rate": 3.91447830102228e-05, "loss": 0.9576, "step": 15818 }, { "epoch": 1.59, "learning_rate": 3.9142503707811544e-05, "loss": 0.4022, "step": 15819 }, { "epoch": 1.59, "learning_rate": 3.914022434722087e-05, "loss": 0.1208, "step": 15820 }, { "epoch": 1.59, "learning_rate": 3.913794492846528e-05, "loss": 1.4149, "step": 15821 }, { "epoch": 1.59, "learning_rate": 3.91356654515593e-05, "loss": 0.8385, "step": 15822 }, { "epoch": 1.59, "learning_rate": 3.9133385916517404e-05, "loss": 0.7429, "step": 15823 }, { "epoch": 1.59, "learning_rate": 3.913110632335412e-05, "loss": 0.1908, "step": 15824 }, { "epoch": 1.59, "learning_rate": 3.912882667208395e-05, "loss": 1.5952, "step": 15825 }, { "epoch": 1.59, "learning_rate": 3.912654696272141e-05, "loss": 0.6092, "step": 15826 }, { "epoch": 1.59, "learning_rate": 3.912426719528099e-05, "loss": 0.3711, "step": 15827 }, { "epoch": 1.59, "learning_rate": 3.9121987369777194e-05, "loss": 0.628, "step": 15828 }, { "epoch": 1.59, "learning_rate": 3.911970748622456e-05, "loss": 0.2861, "step": 15829 }, { "epoch": 1.59, "learning_rate": 3.9117427544637563e-05, "loss": 0.2536, "step": 15830 }, { "epoch": 1.59, "learning_rate": 3.911514754503073e-05, "loss": 0.3284, "step": 15831 }, { "epoch": 1.59, "learning_rate": 3.911286748741856e-05, "loss": 1.127, "step": 15832 }, { "epoch": 1.59, "learning_rate": 3.911058737181558e-05, "loss": 0.3484, "step": 15833 }, { "epoch": 1.59, "learning_rate": 3.9108307198236276e-05, "loss": 1.0121, "step": 15834 }, { "epoch": 1.59, "learning_rate": 3.910602696669517e-05, "loss": 0.2294, "step": 15835 }, { "epoch": 1.6, "learning_rate": 3.910374667720678e-05, "loss": 0.6775, "step": 15836 }, { "epoch": 1.6, "learning_rate": 3.91014663297856e-05, "loss": 0.2567, "step": 15837 }, { "epoch": 1.6, "learning_rate": 3.909918592444616e-05, "loss": 0.948, "step": 15838 }, { "epoch": 1.6, "learning_rate": 3.9096905461202964e-05, "loss": 0.6546, "step": 15839 }, { "epoch": 1.6, "learning_rate": 3.909462494007052e-05, "loss": 0.7042, "step": 15840 }, { "epoch": 1.6, "learning_rate": 3.909234436106334e-05, "loss": 0.5479, "step": 15841 }, { "epoch": 1.6, "learning_rate": 3.909006372419593e-05, "loss": 0.689, "step": 15842 }, { "epoch": 1.6, "learning_rate": 3.9087783029482834e-05, "loss": 0.6335, "step": 15843 }, { "epoch": 1.6, "learning_rate": 3.908550227693852e-05, "loss": 0.3074, "step": 15844 }, { "epoch": 1.6, "learning_rate": 3.908322146657755e-05, "loss": 1.2262, "step": 15845 }, { "epoch": 1.6, "learning_rate": 3.9080940598414395e-05, "loss": 1.0847, "step": 15846 }, { "epoch": 1.6, "learning_rate": 3.9078659672463594e-05, "loss": 0.4221, "step": 15847 }, { "epoch": 1.6, "learning_rate": 3.907637868873966e-05, "loss": 0.4862, "step": 15848 }, { "epoch": 1.6, "learning_rate": 3.90740976472571e-05, "loss": 1.1558, "step": 15849 }, { "epoch": 1.6, "learning_rate": 3.907181654803044e-05, "loss": 0.4546, "step": 15850 }, { "epoch": 1.6, "learning_rate": 3.906953539107418e-05, "loss": 0.3153, "step": 15851 }, { "epoch": 1.6, "learning_rate": 3.906725417640287e-05, "loss": 0.9179, "step": 15852 }, { "epoch": 1.6, "learning_rate": 3.906497290403098e-05, "loss": 0.2481, "step": 15853 }, { "epoch": 1.6, "learning_rate": 3.9062691573973056e-05, "loss": 0.4589, "step": 15854 }, { "epoch": 1.6, "learning_rate": 3.9060410186243616e-05, "loss": 0.4352, "step": 15855 }, { "epoch": 1.6, "learning_rate": 3.905812874085717e-05, "loss": 0.6096, "step": 15856 }, { "epoch": 1.6, "learning_rate": 3.9055847237828246e-05, "loss": 0.3357, "step": 15857 }, { "epoch": 1.6, "learning_rate": 3.9053565677171344e-05, "loss": 0.2144, "step": 15858 }, { "epoch": 1.6, "learning_rate": 3.9051284058901e-05, "loss": 1.2508, "step": 15859 }, { "epoch": 1.6, "learning_rate": 3.9049002383031725e-05, "loss": 0.3938, "step": 15860 }, { "epoch": 1.6, "learning_rate": 3.904672064957804e-05, "loss": 0.9778, "step": 15861 }, { "epoch": 1.6, "learning_rate": 3.9044438858554465e-05, "loss": 0.1388, "step": 15862 }, { "epoch": 1.6, "learning_rate": 3.9042157009975533e-05, "loss": 0.6731, "step": 15863 }, { "epoch": 1.6, "learning_rate": 3.903987510385574e-05, "loss": 0.3386, "step": 15864 }, { "epoch": 1.6, "learning_rate": 3.903759314020963e-05, "loss": 0.2188, "step": 15865 }, { "epoch": 1.6, "learning_rate": 3.903531111905171e-05, "loss": 0.3574, "step": 15866 }, { "epoch": 1.6, "learning_rate": 3.903302904039651e-05, "loss": 0.706, "step": 15867 }, { "epoch": 1.6, "learning_rate": 3.903074690425855e-05, "loss": 0.9207, "step": 15868 }, { "epoch": 1.6, "learning_rate": 3.9028464710652344e-05, "loss": 0.6048, "step": 15869 }, { "epoch": 1.6, "learning_rate": 3.9026182459592433e-05, "loss": 0.1895, "step": 15870 }, { "epoch": 1.6, "learning_rate": 3.902390015109333e-05, "loss": 0.2765, "step": 15871 }, { "epoch": 1.6, "learning_rate": 3.9021617785169556e-05, "loss": 0.0426, "step": 15872 }, { "epoch": 1.6, "learning_rate": 3.901933536183564e-05, "loss": 1.0007, "step": 15873 }, { "epoch": 1.6, "learning_rate": 3.9017052881106106e-05, "loss": 0.7573, "step": 15874 }, { "epoch": 1.6, "learning_rate": 3.901477034299548e-05, "loss": 0.3066, "step": 15875 }, { "epoch": 1.6, "learning_rate": 3.9012487747518276e-05, "loss": 0.1092, "step": 15876 }, { "epoch": 1.6, "learning_rate": 3.9010205094689036e-05, "loss": 0.2679, "step": 15877 }, { "epoch": 1.6, "learning_rate": 3.900792238452227e-05, "loss": 0.8686, "step": 15878 }, { "epoch": 1.6, "learning_rate": 3.900563961703252e-05, "loss": 0.2904, "step": 15879 }, { "epoch": 1.6, "learning_rate": 3.9003356792234306e-05, "loss": 0.4961, "step": 15880 }, { "epoch": 1.6, "learning_rate": 3.9001073910142155e-05, "loss": 0.7587, "step": 15881 }, { "epoch": 1.6, "learning_rate": 3.899879097077059e-05, "loss": 0.3835, "step": 15882 }, { "epoch": 1.6, "learning_rate": 3.899650797413414e-05, "loss": 0.1548, "step": 15883 }, { "epoch": 1.6, "learning_rate": 3.8994224920247344e-05, "loss": 0.652, "step": 15884 }, { "epoch": 1.6, "learning_rate": 3.8991941809124713e-05, "loss": 0.6919, "step": 15885 }, { "epoch": 1.6, "learning_rate": 3.898965864078079e-05, "loss": 0.3377, "step": 15886 }, { "epoch": 1.6, "learning_rate": 3.89873754152301e-05, "loss": 0.9812, "step": 15887 }, { "epoch": 1.6, "learning_rate": 3.898509213248717e-05, "loss": 0.4394, "step": 15888 }, { "epoch": 1.6, "learning_rate": 3.898280879256654e-05, "loss": 0.2292, "step": 15889 }, { "epoch": 1.6, "learning_rate": 3.898052539548272e-05, "loss": 0.4845, "step": 15890 }, { "epoch": 1.6, "learning_rate": 3.8978241941250256e-05, "loss": 0.3871, "step": 15891 }, { "epoch": 1.6, "learning_rate": 3.8975958429883674e-05, "loss": 0.3423, "step": 15892 }, { "epoch": 1.6, "learning_rate": 3.897367486139751e-05, "loss": 0.815, "step": 15893 }, { "epoch": 1.6, "learning_rate": 3.8971391235806295e-05, "loss": 0.2558, "step": 15894 }, { "epoch": 1.6, "learning_rate": 3.896910755312456e-05, "loss": 0.3943, "step": 15895 }, { "epoch": 1.6, "learning_rate": 3.8966823813366834e-05, "loss": 0.609, "step": 15896 }, { "epoch": 1.6, "learning_rate": 3.8964540016547655e-05, "loss": 1.1737, "step": 15897 }, { "epoch": 1.6, "learning_rate": 3.8962256162681545e-05, "loss": 0.6633, "step": 15898 }, { "epoch": 1.6, "learning_rate": 3.895997225178305e-05, "loss": 0.4444, "step": 15899 }, { "epoch": 1.6, "learning_rate": 3.895768828386671e-05, "loss": 0.9331, "step": 15900 }, { "epoch": 1.6, "learning_rate": 3.895540425894704e-05, "loss": 1.0239, "step": 15901 }, { "epoch": 1.6, "learning_rate": 3.895312017703858e-05, "loss": 0.5702, "step": 15902 }, { "epoch": 1.6, "learning_rate": 3.895083603815588e-05, "loss": 1.7537, "step": 15903 }, { "epoch": 1.6, "learning_rate": 3.894855184231346e-05, "loss": 0.5347, "step": 15904 }, { "epoch": 1.6, "learning_rate": 3.894626758952586e-05, "loss": 0.5779, "step": 15905 }, { "epoch": 1.6, "learning_rate": 3.894398327980761e-05, "loss": 0.4837, "step": 15906 }, { "epoch": 1.6, "learning_rate": 3.8941698913173266e-05, "loss": 0.958, "step": 15907 }, { "epoch": 1.6, "learning_rate": 3.893941448963734e-05, "loss": 0.9506, "step": 15908 }, { "epoch": 1.6, "learning_rate": 3.893713000921439e-05, "loss": 0.3258, "step": 15909 }, { "epoch": 1.6, "learning_rate": 3.893484547191894e-05, "loss": 0.3801, "step": 15910 }, { "epoch": 1.6, "learning_rate": 3.893256087776553e-05, "loss": 0.4511, "step": 15911 }, { "epoch": 1.6, "learning_rate": 3.89302762267687e-05, "loss": 0.3267, "step": 15912 }, { "epoch": 1.6, "learning_rate": 3.8927991518943e-05, "loss": 0.395, "step": 15913 }, { "epoch": 1.6, "learning_rate": 3.8925706754302954e-05, "loss": 0.6262, "step": 15914 }, { "epoch": 1.6, "learning_rate": 3.892342193286309e-05, "loss": 0.2278, "step": 15915 }, { "epoch": 1.6, "learning_rate": 3.8921137054637986e-05, "loss": 0.7786, "step": 15916 }, { "epoch": 1.6, "learning_rate": 3.891885211964214e-05, "loss": 0.326, "step": 15917 }, { "epoch": 1.6, "learning_rate": 3.8916567127890126e-05, "loss": 0.3768, "step": 15918 }, { "epoch": 1.6, "learning_rate": 3.8914282079396476e-05, "loss": 0.5318, "step": 15919 }, { "epoch": 1.6, "learning_rate": 3.891199697417571e-05, "loss": 0.0959, "step": 15920 }, { "epoch": 1.6, "learning_rate": 3.890971181224239e-05, "loss": 0.5711, "step": 15921 }, { "epoch": 1.6, "learning_rate": 3.8907426593611054e-05, "loss": 0.5776, "step": 15922 }, { "epoch": 1.6, "learning_rate": 3.890514131829625e-05, "loss": 0.8955, "step": 15923 }, { "epoch": 1.6, "learning_rate": 3.8902855986312506e-05, "loss": 0.1984, "step": 15924 }, { "epoch": 1.6, "learning_rate": 3.890057059767438e-05, "loss": 1.5802, "step": 15925 }, { "epoch": 1.6, "learning_rate": 3.8898285152396405e-05, "loss": 0.4223, "step": 15926 }, { "epoch": 1.6, "learning_rate": 3.8895999650493133e-05, "loss": 1.134, "step": 15927 }, { "epoch": 1.6, "learning_rate": 3.889371409197909e-05, "loss": 1.0118, "step": 15928 }, { "epoch": 1.6, "learning_rate": 3.889142847686885e-05, "loss": 0.1722, "step": 15929 }, { "epoch": 1.6, "learning_rate": 3.888914280517694e-05, "loss": 0.5676, "step": 15930 }, { "epoch": 1.6, "learning_rate": 3.8886857076917894e-05, "loss": 0.7988, "step": 15931 }, { "epoch": 1.6, "learning_rate": 3.8884571292106276e-05, "loss": 0.5322, "step": 15932 }, { "epoch": 1.6, "learning_rate": 3.8882285450756624e-05, "loss": 0.4689, "step": 15933 }, { "epoch": 1.6, "learning_rate": 3.88799995528835e-05, "loss": 1.1263, "step": 15934 }, { "epoch": 1.61, "learning_rate": 3.887771359850142e-05, "loss": 1.1569, "step": 15935 }, { "epoch": 1.61, "learning_rate": 3.8875427587624965e-05, "loss": 0.5683, "step": 15936 }, { "epoch": 1.61, "learning_rate": 3.8873141520268654e-05, "loss": 0.4366, "step": 15937 }, { "epoch": 1.61, "learning_rate": 3.8870855396447047e-05, "loss": 1.2297, "step": 15938 }, { "epoch": 1.61, "learning_rate": 3.886856921617469e-05, "loss": 0.3697, "step": 15939 }, { "epoch": 1.61, "learning_rate": 3.886628297946613e-05, "loss": 0.9926, "step": 15940 }, { "epoch": 1.61, "learning_rate": 3.886399668633592e-05, "loss": 0.6417, "step": 15941 }, { "epoch": 1.61, "learning_rate": 3.886171033679861e-05, "loss": 0.3059, "step": 15942 }, { "epoch": 1.61, "learning_rate": 3.885942393086875e-05, "loss": 0.9479, "step": 15943 }, { "epoch": 1.61, "learning_rate": 3.8857137468560885e-05, "loss": 0.6494, "step": 15944 }, { "epoch": 1.61, "learning_rate": 3.885485094988956e-05, "loss": 0.5066, "step": 15945 }, { "epoch": 1.61, "learning_rate": 3.8852564374869335e-05, "loss": 1.0271, "step": 15946 }, { "epoch": 1.61, "learning_rate": 3.885027774351476e-05, "loss": 0.3331, "step": 15947 }, { "epoch": 1.61, "learning_rate": 3.8847991055840376e-05, "loss": 0.7682, "step": 15948 }, { "epoch": 1.61, "learning_rate": 3.884570431186075e-05, "loss": 0.8926, "step": 15949 }, { "epoch": 1.61, "learning_rate": 3.884341751159043e-05, "loss": 0.2788, "step": 15950 }, { "epoch": 1.61, "learning_rate": 3.884113065504396e-05, "loss": 0.8616, "step": 15951 }, { "epoch": 1.61, "learning_rate": 3.8838843742235906e-05, "loss": 0.8817, "step": 15952 }, { "epoch": 1.61, "learning_rate": 3.883655677318081e-05, "loss": 0.6764, "step": 15953 }, { "epoch": 1.61, "learning_rate": 3.883426974789321e-05, "loss": 0.4902, "step": 15954 }, { "epoch": 1.61, "learning_rate": 3.883198266638771e-05, "loss": 1.1843, "step": 15955 }, { "epoch": 1.61, "learning_rate": 3.8829695528678804e-05, "loss": 0.3581, "step": 15956 }, { "epoch": 1.61, "learning_rate": 3.882740833478109e-05, "loss": 1.3173, "step": 15957 }, { "epoch": 1.61, "learning_rate": 3.88251210847091e-05, "loss": 1.549, "step": 15958 }, { "epoch": 1.61, "learning_rate": 3.88228337784774e-05, "loss": 1.3935, "step": 15959 }, { "epoch": 1.61, "learning_rate": 3.882054641610054e-05, "loss": 0.3157, "step": 15960 }, { "epoch": 1.61, "learning_rate": 3.881825899759308e-05, "loss": 1.1127, "step": 15961 }, { "epoch": 1.61, "learning_rate": 3.881597152296958e-05, "loss": 0.726, "step": 15962 }, { "epoch": 1.61, "learning_rate": 3.881368399224458e-05, "loss": 0.2175, "step": 15963 }, { "epoch": 1.61, "learning_rate": 3.8811396405432654e-05, "loss": 0.3337, "step": 15964 }, { "epoch": 1.61, "learning_rate": 3.880910876254835e-05, "loss": 0.394, "step": 15965 }, { "epoch": 1.61, "learning_rate": 3.880682106360623e-05, "loss": 0.6034, "step": 15966 }, { "epoch": 1.61, "learning_rate": 3.880453330862086e-05, "loss": 0.9088, "step": 15967 }, { "epoch": 1.61, "learning_rate": 3.880224549760678e-05, "loss": 0.4151, "step": 15968 }, { "epoch": 1.61, "learning_rate": 3.879995763057857e-05, "loss": 0.1578, "step": 15969 }, { "epoch": 1.61, "learning_rate": 3.8797669707550756e-05, "loss": 2.4553, "step": 15970 }, { "epoch": 1.61, "learning_rate": 3.879538172853793e-05, "loss": 0.62, "step": 15971 }, { "epoch": 1.61, "learning_rate": 3.8793093693554643e-05, "loss": 0.2339, "step": 15972 }, { "epoch": 1.61, "learning_rate": 3.8790805602615446e-05, "loss": 0.7995, "step": 15973 }, { "epoch": 1.61, "learning_rate": 3.878851745573492e-05, "loss": 0.3446, "step": 15974 }, { "epoch": 1.61, "learning_rate": 3.87862292529276e-05, "loss": 0.9616, "step": 15975 }, { "epoch": 1.61, "learning_rate": 3.878394099420807e-05, "loss": 0.6461, "step": 15976 }, { "epoch": 1.61, "learning_rate": 3.878165267959087e-05, "loss": 0.6006, "step": 15977 }, { "epoch": 1.61, "learning_rate": 3.8779364309090574e-05, "loss": 0.1888, "step": 15978 }, { "epoch": 1.61, "learning_rate": 3.877707588272175e-05, "loss": 0.9262, "step": 15979 }, { "epoch": 1.61, "learning_rate": 3.877478740049894e-05, "loss": 0.3587, "step": 15980 }, { "epoch": 1.61, "learning_rate": 3.877249886243674e-05, "loss": 0.6871, "step": 15981 }, { "epoch": 1.61, "learning_rate": 3.8770210268549686e-05, "loss": 0.5939, "step": 15982 }, { "epoch": 1.61, "learning_rate": 3.876792161885236e-05, "loss": 0.6142, "step": 15983 }, { "epoch": 1.61, "learning_rate": 3.876563291335929e-05, "loss": 1.2346, "step": 15984 }, { "epoch": 1.61, "learning_rate": 3.876334415208509e-05, "loss": 0.608, "step": 15985 }, { "epoch": 1.61, "learning_rate": 3.8761055335044296e-05, "loss": 0.8233, "step": 15986 }, { "epoch": 1.61, "learning_rate": 3.875876646225147e-05, "loss": 1.172, "step": 15987 }, { "epoch": 1.61, "learning_rate": 3.8756477533721204e-05, "loss": 0.9845, "step": 15988 }, { "epoch": 1.61, "learning_rate": 3.875418854946803e-05, "loss": 0.4114, "step": 15989 }, { "epoch": 1.61, "learning_rate": 3.8751899509506535e-05, "loss": 0.621, "step": 15990 }, { "epoch": 1.61, "learning_rate": 3.8749610413851285e-05, "loss": 0.9596, "step": 15991 }, { "epoch": 1.61, "learning_rate": 3.874732126251685e-05, "loss": 0.8125, "step": 15992 }, { "epoch": 1.61, "learning_rate": 3.874503205551776e-05, "loss": 0.638, "step": 15993 }, { "epoch": 1.61, "learning_rate": 3.874274279286865e-05, "loss": 0.4763, "step": 15994 }, { "epoch": 1.61, "learning_rate": 3.8740453474584027e-05, "loss": 0.8415, "step": 15995 }, { "epoch": 1.61, "learning_rate": 3.873816410067849e-05, "loss": 0.5444, "step": 15996 }, { "epoch": 1.61, "learning_rate": 3.87358746711666e-05, "loss": 0.4143, "step": 15997 }, { "epoch": 1.61, "learning_rate": 3.8733585186062934e-05, "loss": 1.5963, "step": 15998 }, { "epoch": 1.61, "learning_rate": 3.873129564538205e-05, "loss": 1.3323, "step": 15999 }, { "epoch": 1.61, "learning_rate": 3.872900604913851e-05, "loss": 1.1721, "step": 16000 }, { "epoch": 1.61, "learning_rate": 3.872671639734691e-05, "loss": 1.1468, "step": 16001 }, { "epoch": 1.61, "learning_rate": 3.87244266900218e-05, "loss": 0.3768, "step": 16002 }, { "epoch": 1.61, "learning_rate": 3.872213692717776e-05, "loss": 0.3358, "step": 16003 }, { "epoch": 1.61, "learning_rate": 3.8719847108829365e-05, "loss": 1.3387, "step": 16004 }, { "epoch": 1.61, "learning_rate": 3.871755723499117e-05, "loss": 0.4074, "step": 16005 }, { "epoch": 1.61, "learning_rate": 3.871526730567776e-05, "loss": 0.8349, "step": 16006 }, { "epoch": 1.61, "learning_rate": 3.8712977320903715e-05, "loss": 0.3755, "step": 16007 }, { "epoch": 1.61, "learning_rate": 3.871068728068358e-05, "loss": 0.7224, "step": 16008 }, { "epoch": 1.61, "learning_rate": 3.870839718503195e-05, "loss": 0.268, "step": 16009 }, { "epoch": 1.61, "learning_rate": 3.87061070339634e-05, "loss": 0.6556, "step": 16010 }, { "epoch": 1.61, "learning_rate": 3.870381682749249e-05, "loss": 0.772, "step": 16011 }, { "epoch": 1.61, "learning_rate": 3.87015265656338e-05, "loss": 0.5479, "step": 16012 }, { "epoch": 1.61, "learning_rate": 3.869923624840191e-05, "loss": 0.2225, "step": 16013 }, { "epoch": 1.61, "learning_rate": 3.869694587581139e-05, "loss": 1.1792, "step": 16014 }, { "epoch": 1.61, "learning_rate": 3.8694655447876813e-05, "loss": 1.0698, "step": 16015 }, { "epoch": 1.61, "learning_rate": 3.869236496461275e-05, "loss": 0.5375, "step": 16016 }, { "epoch": 1.61, "learning_rate": 3.86900744260338e-05, "loss": 0.6988, "step": 16017 }, { "epoch": 1.61, "learning_rate": 3.868778383215451e-05, "loss": 0.7688, "step": 16018 }, { "epoch": 1.61, "learning_rate": 3.868549318298947e-05, "loss": 0.3513, "step": 16019 }, { "epoch": 1.61, "learning_rate": 3.8683202478553254e-05, "loss": 1.1194, "step": 16020 }, { "epoch": 1.61, "learning_rate": 3.8680911718860444e-05, "loss": 1.1857, "step": 16021 }, { "epoch": 1.61, "learning_rate": 3.867862090392561e-05, "loss": 0.5397, "step": 16022 }, { "epoch": 1.61, "learning_rate": 3.8676330033763346e-05, "loss": 0.4037, "step": 16023 }, { "epoch": 1.61, "learning_rate": 3.867403910838821e-05, "loss": 1.2354, "step": 16024 }, { "epoch": 1.61, "learning_rate": 3.8671748127814774e-05, "loss": 0.3644, "step": 16025 }, { "epoch": 1.61, "learning_rate": 3.866945709205765e-05, "loss": 0.2812, "step": 16026 }, { "epoch": 1.61, "learning_rate": 3.86671660011314e-05, "loss": 0.7449, "step": 16027 }, { "epoch": 1.61, "learning_rate": 3.8664874855050594e-05, "loss": 0.3305, "step": 16028 }, { "epoch": 1.61, "learning_rate": 3.8662583653829826e-05, "loss": 0.5842, "step": 16029 }, { "epoch": 1.61, "learning_rate": 3.866029239748367e-05, "loss": 0.5887, "step": 16030 }, { "epoch": 1.61, "learning_rate": 3.865800108602671e-05, "loss": 0.6834, "step": 16031 }, { "epoch": 1.61, "learning_rate": 3.865570971947352e-05, "loss": 0.4966, "step": 16032 }, { "epoch": 1.62, "learning_rate": 3.8653418297838686e-05, "loss": 0.5045, "step": 16033 }, { "epoch": 1.62, "learning_rate": 3.865112682113679e-05, "loss": 0.9384, "step": 16034 }, { "epoch": 1.62, "learning_rate": 3.8648835289382414e-05, "loss": 0.369, "step": 16035 }, { "epoch": 1.62, "learning_rate": 3.864654370259015e-05, "loss": 0.9241, "step": 16036 }, { "epoch": 1.62, "learning_rate": 3.864425206077456e-05, "loss": 0.3633, "step": 16037 }, { "epoch": 1.62, "learning_rate": 3.864196036395025e-05, "loss": 0.2392, "step": 16038 }, { "epoch": 1.62, "learning_rate": 3.863966861213177e-05, "loss": 1.3622, "step": 16039 }, { "epoch": 1.62, "learning_rate": 3.863737680533375e-05, "loss": 0.491, "step": 16040 }, { "epoch": 1.62, "learning_rate": 3.863508494357073e-05, "loss": 0.2822, "step": 16041 }, { "epoch": 1.62, "learning_rate": 3.863279302685732e-05, "loss": 0.1105, "step": 16042 }, { "epoch": 1.62, "learning_rate": 3.8630501055208115e-05, "loss": 0.908, "step": 16043 }, { "epoch": 1.62, "learning_rate": 3.862820902863767e-05, "loss": 0.9073, "step": 16044 }, { "epoch": 1.62, "learning_rate": 3.862591694716059e-05, "loss": 1.0354, "step": 16045 }, { "epoch": 1.62, "learning_rate": 3.8623624810791456e-05, "loss": 1.2772, "step": 16046 }, { "epoch": 1.62, "learning_rate": 3.862133261954485e-05, "loss": 0.5819, "step": 16047 }, { "epoch": 1.62, "learning_rate": 3.861904037343536e-05, "loss": 0.4894, "step": 16048 }, { "epoch": 1.62, "learning_rate": 3.861674807247759e-05, "loss": 0.7381, "step": 16049 }, { "epoch": 1.62, "learning_rate": 3.8614455716686094e-05, "loss": 0.6858, "step": 16050 }, { "epoch": 1.62, "learning_rate": 3.8612163306075486e-05, "loss": 0.8124, "step": 16051 }, { "epoch": 1.62, "learning_rate": 3.8609870840660355e-05, "loss": 0.2681, "step": 16052 }, { "epoch": 1.62, "learning_rate": 3.860757832045528e-05, "loss": 0.6006, "step": 16053 }, { "epoch": 1.62, "learning_rate": 3.860528574547485e-05, "loss": 0.3396, "step": 16054 }, { "epoch": 1.62, "learning_rate": 3.8602993115733646e-05, "loss": 0.1862, "step": 16055 }, { "epoch": 1.62, "learning_rate": 3.860070043124628e-05, "loss": 0.8955, "step": 16056 }, { "epoch": 1.62, "learning_rate": 3.859840769202731e-05, "loss": 0.8838, "step": 16057 }, { "epoch": 1.62, "learning_rate": 3.859611489809136e-05, "loss": 0.5273, "step": 16058 }, { "epoch": 1.62, "learning_rate": 3.8593822049453e-05, "loss": 0.7395, "step": 16059 }, { "epoch": 1.62, "learning_rate": 3.859152914612683e-05, "loss": 0.3403, "step": 16060 }, { "epoch": 1.62, "learning_rate": 3.858923618812743e-05, "loss": 0.6056, "step": 16061 }, { "epoch": 1.62, "learning_rate": 3.8586943175469406e-05, "loss": 0.3225, "step": 16062 }, { "epoch": 1.62, "learning_rate": 3.8584650108167334e-05, "loss": 0.8368, "step": 16063 }, { "epoch": 1.62, "learning_rate": 3.858235698623582e-05, "loss": 0.8167, "step": 16064 }, { "epoch": 1.62, "learning_rate": 3.858006380968944e-05, "loss": 0.6588, "step": 16065 }, { "epoch": 1.62, "learning_rate": 3.857777057854281e-05, "loss": 0.5013, "step": 16066 }, { "epoch": 1.62, "learning_rate": 3.857547729281051e-05, "loss": 0.6459, "step": 16067 }, { "epoch": 1.62, "learning_rate": 3.8573183952507135e-05, "loss": 0.4252, "step": 16068 }, { "epoch": 1.62, "learning_rate": 3.8570890557647285e-05, "loss": 0.2586, "step": 16069 }, { "epoch": 1.62, "learning_rate": 3.856859710824553e-05, "loss": 0.1293, "step": 16070 }, { "epoch": 1.62, "learning_rate": 3.856630360431649e-05, "loss": 0.2921, "step": 16071 }, { "epoch": 1.62, "learning_rate": 3.856401004587476e-05, "loss": 0.8287, "step": 16072 }, { "epoch": 1.62, "learning_rate": 3.856171643293492e-05, "loss": 0.6096, "step": 16073 }, { "epoch": 1.62, "learning_rate": 3.8559422765511585e-05, "loss": 0.3466, "step": 16074 }, { "epoch": 1.62, "learning_rate": 3.855712904361933e-05, "loss": 0.79, "step": 16075 }, { "epoch": 1.62, "learning_rate": 3.855483526727276e-05, "loss": 0.1808, "step": 16076 }, { "epoch": 1.62, "learning_rate": 3.855254143648648e-05, "loss": 0.8309, "step": 16077 }, { "epoch": 1.62, "learning_rate": 3.8550247551275074e-05, "loss": 0.1861, "step": 16078 }, { "epoch": 1.62, "learning_rate": 3.8547953611653146e-05, "loss": 0.6564, "step": 16079 }, { "epoch": 1.62, "learning_rate": 3.854565961763529e-05, "loss": 1.1336, "step": 16080 }, { "epoch": 1.62, "learning_rate": 3.854336556923611e-05, "loss": 0.6474, "step": 16081 }, { "epoch": 1.62, "learning_rate": 3.854107146647021e-05, "loss": 0.3304, "step": 16082 }, { "epoch": 1.62, "learning_rate": 3.853877730935217e-05, "loss": 0.2817, "step": 16083 }, { "epoch": 1.62, "learning_rate": 3.85364830978966e-05, "loss": 0.6564, "step": 16084 }, { "epoch": 1.62, "learning_rate": 3.85341888321181e-05, "loss": 0.4338, "step": 16085 }, { "epoch": 1.62, "learning_rate": 3.8531894512031274e-05, "loss": 0.506, "step": 16086 }, { "epoch": 1.62, "learning_rate": 3.852960013765071e-05, "loss": 0.7124, "step": 16087 }, { "epoch": 1.62, "learning_rate": 3.8527305708991026e-05, "loss": 0.5394, "step": 16088 }, { "epoch": 1.62, "learning_rate": 3.8525011226066796e-05, "loss": 0.2208, "step": 16089 }, { "epoch": 1.62, "learning_rate": 3.852271668889265e-05, "loss": 0.7381, "step": 16090 }, { "epoch": 1.62, "learning_rate": 3.852042209748318e-05, "loss": 0.706, "step": 16091 }, { "epoch": 1.62, "learning_rate": 3.851812745185297e-05, "loss": 0.4056, "step": 16092 }, { "epoch": 1.62, "learning_rate": 3.8515832752016656e-05, "loss": 0.2787, "step": 16093 }, { "epoch": 1.62, "learning_rate": 3.85135379979888e-05, "loss": 0.9753, "step": 16094 }, { "epoch": 1.62, "learning_rate": 3.851124318978405e-05, "loss": 0.4953, "step": 16095 }, { "epoch": 1.62, "learning_rate": 3.850894832741697e-05, "loss": 0.6076, "step": 16096 }, { "epoch": 1.62, "learning_rate": 3.850665341090218e-05, "loss": 0.2662, "step": 16097 }, { "epoch": 1.62, "learning_rate": 3.8504358440254294e-05, "loss": 0.848, "step": 16098 }, { "epoch": 1.62, "learning_rate": 3.85020634154879e-05, "loss": 0.9996, "step": 16099 }, { "epoch": 1.62, "learning_rate": 3.8499768336617615e-05, "loss": 1.2207, "step": 16100 }, { "epoch": 1.62, "learning_rate": 3.8497473203658035e-05, "loss": 1.189, "step": 16101 }, { "epoch": 1.62, "learning_rate": 3.8495178016623764e-05, "loss": 0.561, "step": 16102 }, { "epoch": 1.62, "learning_rate": 3.849288277552941e-05, "loss": 0.7021, "step": 16103 }, { "epoch": 1.62, "learning_rate": 3.8490587480389584e-05, "loss": 1.3528, "step": 16104 }, { "epoch": 1.62, "learning_rate": 3.848829213121889e-05, "loss": 0.9116, "step": 16105 }, { "epoch": 1.62, "learning_rate": 3.848599672803193e-05, "loss": 0.5951, "step": 16106 }, { "epoch": 1.62, "learning_rate": 3.848370127084332e-05, "loss": 0.5378, "step": 16107 }, { "epoch": 1.62, "learning_rate": 3.8481405759667666e-05, "loss": 0.5759, "step": 16108 }, { "epoch": 1.62, "learning_rate": 3.847911019451957e-05, "loss": 1.3856, "step": 16109 }, { "epoch": 1.62, "learning_rate": 3.8476814575413634e-05, "loss": 0.774, "step": 16110 }, { "epoch": 1.62, "learning_rate": 3.847451890236449e-05, "loss": 0.4209, "step": 16111 }, { "epoch": 1.62, "learning_rate": 3.8472223175386714e-05, "loss": 0.2356, "step": 16112 }, { "epoch": 1.62, "learning_rate": 3.846992739449494e-05, "loss": 0.7219, "step": 16113 }, { "epoch": 1.62, "learning_rate": 3.8467631559703775e-05, "loss": 0.398, "step": 16114 }, { "epoch": 1.62, "learning_rate": 3.8465335671027824e-05, "loss": 0.4961, "step": 16115 }, { "epoch": 1.62, "learning_rate": 3.846303972848169e-05, "loss": 0.7362, "step": 16116 }, { "epoch": 1.62, "learning_rate": 3.846074373208e-05, "loss": 0.4009, "step": 16117 }, { "epoch": 1.62, "learning_rate": 3.8458447681837344e-05, "loss": 0.3445, "step": 16118 }, { "epoch": 1.62, "learning_rate": 3.845615157776835e-05, "loss": 0.1174, "step": 16119 }, { "epoch": 1.62, "learning_rate": 3.845385541988762e-05, "loss": 0.605, "step": 16120 }, { "epoch": 1.62, "learning_rate": 3.8451559208209776e-05, "loss": 0.1312, "step": 16121 }, { "epoch": 1.62, "learning_rate": 3.8449262942749426e-05, "loss": 1.2596, "step": 16122 }, { "epoch": 1.62, "learning_rate": 3.844696662352117e-05, "loss": 0.9162, "step": 16123 }, { "epoch": 1.62, "learning_rate": 3.8444670250539646e-05, "loss": 0.2085, "step": 16124 }, { "epoch": 1.62, "learning_rate": 3.844237382381945e-05, "loss": 0.6118, "step": 16125 }, { "epoch": 1.62, "learning_rate": 3.8440077343375185e-05, "loss": 0.3137, "step": 16126 }, { "epoch": 1.62, "learning_rate": 3.84377808092215e-05, "loss": 0.3321, "step": 16127 }, { "epoch": 1.62, "learning_rate": 3.8435484221372967e-05, "loss": 1.1909, "step": 16128 }, { "epoch": 1.62, "learning_rate": 3.843318757984424e-05, "loss": 0.8892, "step": 16129 }, { "epoch": 1.62, "learning_rate": 3.84308908846499e-05, "loss": 0.8968, "step": 16130 }, { "epoch": 1.62, "learning_rate": 3.8428594135804586e-05, "loss": 0.2008, "step": 16131 }, { "epoch": 1.63, "learning_rate": 3.842629733332291e-05, "loss": 0.4046, "step": 16132 }, { "epoch": 1.63, "learning_rate": 3.842400047721947e-05, "loss": 0.458, "step": 16133 }, { "epoch": 1.63, "learning_rate": 3.84217035675089e-05, "loss": 0.5404, "step": 16134 }, { "epoch": 1.63, "learning_rate": 3.8419406604205816e-05, "loss": 1.0747, "step": 16135 }, { "epoch": 1.63, "learning_rate": 3.841710958732483e-05, "loss": 0.6059, "step": 16136 }, { "epoch": 1.63, "learning_rate": 3.841481251688056e-05, "loss": 0.5937, "step": 16137 }, { "epoch": 1.63, "learning_rate": 3.8412515392887626e-05, "loss": 0.4447, "step": 16138 }, { "epoch": 1.63, "learning_rate": 3.841021821536065e-05, "loss": 0.3268, "step": 16139 }, { "epoch": 1.63, "learning_rate": 3.840792098431424e-05, "loss": 1.5103, "step": 16140 }, { "epoch": 1.63, "learning_rate": 3.8405623699763015e-05, "loss": 0.2575, "step": 16141 }, { "epoch": 1.63, "learning_rate": 3.8403326361721595e-05, "loss": 0.6228, "step": 16142 }, { "epoch": 1.63, "learning_rate": 3.840102897020462e-05, "loss": 0.4454, "step": 16143 }, { "epoch": 1.63, "learning_rate": 3.839873152522667e-05, "loss": 0.4691, "step": 16144 }, { "epoch": 1.63, "learning_rate": 3.83964340268024e-05, "loss": 0.2775, "step": 16145 }, { "epoch": 1.63, "learning_rate": 3.839413647494642e-05, "loss": 0.3596, "step": 16146 }, { "epoch": 1.63, "learning_rate": 3.839183886967334e-05, "loss": 0.5567, "step": 16147 }, { "epoch": 1.63, "learning_rate": 3.83895412109978e-05, "loss": 0.2207, "step": 16148 }, { "epoch": 1.63, "learning_rate": 3.8387243498934405e-05, "loss": 0.7899, "step": 16149 }, { "epoch": 1.63, "learning_rate": 3.838494573349779e-05, "loss": 2.0756, "step": 16150 }, { "epoch": 1.63, "learning_rate": 3.838264791470256e-05, "loss": 0.3137, "step": 16151 }, { "epoch": 1.63, "learning_rate": 3.838035004256335e-05, "loss": 0.1985, "step": 16152 }, { "epoch": 1.63, "learning_rate": 3.837805211709478e-05, "loss": 1.1264, "step": 16153 }, { "epoch": 1.63, "learning_rate": 3.837575413831148e-05, "loss": 0.5454, "step": 16154 }, { "epoch": 1.63, "learning_rate": 3.837345610622807e-05, "loss": 0.9727, "step": 16155 }, { "epoch": 1.63, "learning_rate": 3.8371158020859163e-05, "loss": 0.6459, "step": 16156 }, { "epoch": 1.63, "learning_rate": 3.836885988221939e-05, "loss": 0.5005, "step": 16157 }, { "epoch": 1.63, "learning_rate": 3.8366561690323374e-05, "loss": 1.0298, "step": 16158 }, { "epoch": 1.63, "learning_rate": 3.836426344518576e-05, "loss": 0.6666, "step": 16159 }, { "epoch": 1.63, "learning_rate": 3.836196514682113e-05, "loss": 0.4497, "step": 16160 }, { "epoch": 1.63, "learning_rate": 3.8359666795244145e-05, "loss": 1.0176, "step": 16161 }, { "epoch": 1.63, "learning_rate": 3.835736839046943e-05, "loss": 0.8203, "step": 16162 }, { "epoch": 1.63, "learning_rate": 3.835506993251159e-05, "loss": 0.9535, "step": 16163 }, { "epoch": 1.63, "learning_rate": 3.835277142138527e-05, "loss": 1.0966, "step": 16164 }, { "epoch": 1.63, "learning_rate": 3.8350472857105085e-05, "loss": 0.4401, "step": 16165 }, { "epoch": 1.63, "learning_rate": 3.8348174239685674e-05, "loss": 0.0908, "step": 16166 }, { "epoch": 1.63, "learning_rate": 3.834587556914165e-05, "loss": 0.8676, "step": 16167 }, { "epoch": 1.63, "learning_rate": 3.834357684548765e-05, "loss": 0.485, "step": 16168 }, { "epoch": 1.63, "learning_rate": 3.834127806873831e-05, "loss": 1.2568, "step": 16169 }, { "epoch": 1.63, "learning_rate": 3.833897923890824e-05, "loss": 0.4774, "step": 16170 }, { "epoch": 1.63, "learning_rate": 3.833668035601209e-05, "loss": 0.6315, "step": 16171 }, { "epoch": 1.63, "learning_rate": 3.833438142006447e-05, "loss": 0.3419, "step": 16172 }, { "epoch": 1.63, "learning_rate": 3.833208243108002e-05, "loss": 0.6305, "step": 16173 }, { "epoch": 1.63, "learning_rate": 3.8329783389073356e-05, "loss": 1.2494, "step": 16174 }, { "epoch": 1.63, "learning_rate": 3.832748429405913e-05, "loss": 0.6861, "step": 16175 }, { "epoch": 1.63, "learning_rate": 3.8325185146051956e-05, "loss": 0.4527, "step": 16176 }, { "epoch": 1.63, "learning_rate": 3.832288594506647e-05, "loss": 0.6192, "step": 16177 }, { "epoch": 1.63, "learning_rate": 3.832058669111732e-05, "loss": 0.6012, "step": 16178 }, { "epoch": 1.63, "learning_rate": 3.8318287384219105e-05, "loss": 0.156, "step": 16179 }, { "epoch": 1.63, "learning_rate": 3.8315988024386476e-05, "loss": 1.2406, "step": 16180 }, { "epoch": 1.63, "learning_rate": 3.831368861163407e-05, "loss": 1.0505, "step": 16181 }, { "epoch": 1.63, "learning_rate": 3.831138914597651e-05, "loss": 0.2736, "step": 16182 }, { "epoch": 1.63, "learning_rate": 3.8309089627428426e-05, "loss": 0.7686, "step": 16183 }, { "epoch": 1.63, "learning_rate": 3.830679005600446e-05, "loss": 0.3556, "step": 16184 }, { "epoch": 1.63, "learning_rate": 3.830449043171924e-05, "loss": 0.9265, "step": 16185 }, { "epoch": 1.63, "learning_rate": 3.83021907545874e-05, "loss": 0.6054, "step": 16186 }, { "epoch": 1.63, "learning_rate": 3.8299891024623584e-05, "loss": 0.7869, "step": 16187 }, { "epoch": 1.63, "learning_rate": 3.829759124184241e-05, "loss": 0.6046, "step": 16188 }, { "epoch": 1.63, "learning_rate": 3.829529140625853e-05, "loss": 0.4034, "step": 16189 }, { "epoch": 1.63, "learning_rate": 3.8292991517886565e-05, "loss": 0.2778, "step": 16190 }, { "epoch": 1.63, "learning_rate": 3.829069157674116e-05, "loss": 0.8367, "step": 16191 }, { "epoch": 1.63, "learning_rate": 3.8288391582836953e-05, "loss": 0.7249, "step": 16192 }, { "epoch": 1.63, "learning_rate": 3.828609153618856e-05, "loss": 0.2372, "step": 16193 }, { "epoch": 1.63, "learning_rate": 3.828379143681065e-05, "loss": 0.79, "step": 16194 }, { "epoch": 1.63, "learning_rate": 3.828149128471784e-05, "loss": 0.789, "step": 16195 }, { "epoch": 1.63, "learning_rate": 3.8279191079924755e-05, "loss": 0.1836, "step": 16196 }, { "epoch": 1.63, "learning_rate": 3.827689082244606e-05, "loss": 0.3906, "step": 16197 }, { "epoch": 1.63, "learning_rate": 3.8274590512296376e-05, "loss": 1.0206, "step": 16198 }, { "epoch": 1.63, "learning_rate": 3.8272290149490347e-05, "loss": 0.9483, "step": 16199 }, { "epoch": 1.63, "learning_rate": 3.826998973404261e-05, "loss": 0.6408, "step": 16200 }, { "epoch": 1.63, "learning_rate": 3.826768926596781e-05, "loss": 0.4985, "step": 16201 }, { "epoch": 1.63, "learning_rate": 3.8265388745280575e-05, "loss": 0.1628, "step": 16202 }, { "epoch": 1.63, "learning_rate": 3.8263088171995546e-05, "loss": 0.9298, "step": 16203 }, { "epoch": 1.63, "learning_rate": 3.8260787546127375e-05, "loss": 0.5926, "step": 16204 }, { "epoch": 1.63, "learning_rate": 3.825848686769069e-05, "loss": 1.803, "step": 16205 }, { "epoch": 1.63, "learning_rate": 3.825618613670014e-05, "loss": 1.0488, "step": 16206 }, { "epoch": 1.63, "learning_rate": 3.825388535317036e-05, "loss": 0.5942, "step": 16207 }, { "epoch": 1.63, "learning_rate": 3.8251584517115994e-05, "loss": 0.1839, "step": 16208 }, { "epoch": 1.63, "learning_rate": 3.824928362855169e-05, "loss": 1.0767, "step": 16209 }, { "epoch": 1.63, "learning_rate": 3.824698268749207e-05, "loss": 1.0506, "step": 16210 }, { "epoch": 1.63, "learning_rate": 3.82446816939518e-05, "loss": 0.7769, "step": 16211 }, { "epoch": 1.63, "learning_rate": 3.824238064794551e-05, "loss": 0.7579, "step": 16212 }, { "epoch": 1.63, "learning_rate": 3.824007954948784e-05, "loss": 0.8681, "step": 16213 }, { "epoch": 1.63, "learning_rate": 3.823777839859344e-05, "loss": 0.2943, "step": 16214 }, { "epoch": 1.63, "learning_rate": 3.8235477195276955e-05, "loss": 1.0228, "step": 16215 }, { "epoch": 1.63, "learning_rate": 3.8233175939553025e-05, "loss": 0.6136, "step": 16216 }, { "epoch": 1.63, "learning_rate": 3.82308746314363e-05, "loss": 0.827, "step": 16217 }, { "epoch": 1.63, "learning_rate": 3.822857327094141e-05, "loss": 0.2715, "step": 16218 }, { "epoch": 1.63, "learning_rate": 3.822627185808302e-05, "loss": 0.0705, "step": 16219 }, { "epoch": 1.63, "learning_rate": 3.822397039287576e-05, "loss": 0.145, "step": 16220 }, { "epoch": 1.63, "learning_rate": 3.822166887533429e-05, "loss": 0.5323, "step": 16221 }, { "epoch": 1.63, "learning_rate": 3.821936730547323e-05, "loss": 0.5782, "step": 16222 }, { "epoch": 1.63, "learning_rate": 3.821706568330725e-05, "loss": 0.8395, "step": 16223 }, { "epoch": 1.63, "learning_rate": 3.821476400885099e-05, "loss": 0.7808, "step": 16224 }, { "epoch": 1.63, "learning_rate": 3.8212462282119105e-05, "loss": 0.8668, "step": 16225 }, { "epoch": 1.63, "learning_rate": 3.821016050312623e-05, "loss": 0.2162, "step": 16226 }, { "epoch": 1.63, "learning_rate": 3.8207858671887014e-05, "loss": 0.5288, "step": 16227 }, { "epoch": 1.63, "learning_rate": 3.820555678841611e-05, "loss": 0.615, "step": 16228 }, { "epoch": 1.63, "learning_rate": 3.820325485272816e-05, "loss": 0.8814, "step": 16229 }, { "epoch": 1.64, "learning_rate": 3.8200952864837815e-05, "loss": 0.943, "step": 16230 }, { "epoch": 1.64, "learning_rate": 3.819865082475973e-05, "loss": 0.9695, "step": 16231 }, { "epoch": 1.64, "learning_rate": 3.819634873250855e-05, "loss": 0.7149, "step": 16232 }, { "epoch": 1.64, "learning_rate": 3.819404658809893e-05, "loss": 0.3035, "step": 16233 }, { "epoch": 1.64, "learning_rate": 3.819174439154551e-05, "loss": 0.8983, "step": 16234 }, { "epoch": 1.64, "learning_rate": 3.818944214286295e-05, "loss": 0.5621, "step": 16235 }, { "epoch": 1.64, "learning_rate": 3.8187139842065884e-05, "loss": 0.9784, "step": 16236 }, { "epoch": 1.64, "learning_rate": 3.818483748916899e-05, "loss": 0.5534, "step": 16237 }, { "epoch": 1.64, "learning_rate": 3.818253508418689e-05, "loss": 0.4444, "step": 16238 }, { "epoch": 1.64, "learning_rate": 3.8180232627134255e-05, "loss": 0.3202, "step": 16239 }, { "epoch": 1.64, "learning_rate": 3.817793011802573e-05, "loss": 0.142, "step": 16240 }, { "epoch": 1.64, "learning_rate": 3.817562755687597e-05, "loss": 0.7182, "step": 16241 }, { "epoch": 1.64, "learning_rate": 3.817332494369963e-05, "loss": 0.3759, "step": 16242 }, { "epoch": 1.64, "learning_rate": 3.817102227851134e-05, "loss": 1.4447, "step": 16243 }, { "epoch": 1.64, "learning_rate": 3.81687195613258e-05, "loss": 0.5811, "step": 16244 }, { "epoch": 1.64, "learning_rate": 3.8166416792157624e-05, "loss": 0.4449, "step": 16245 }, { "epoch": 1.64, "learning_rate": 3.816411397102147e-05, "loss": 0.2541, "step": 16246 }, { "epoch": 1.64, "learning_rate": 3.816181109793201e-05, "loss": 0.8338, "step": 16247 }, { "epoch": 1.64, "learning_rate": 3.8159508172903895e-05, "loss": 0.453, "step": 16248 }, { "epoch": 1.64, "learning_rate": 3.815720519595176e-05, "loss": 0.158, "step": 16249 }, { "epoch": 1.64, "learning_rate": 3.815490216709029e-05, "loss": 0.305, "step": 16250 }, { "epoch": 1.64, "learning_rate": 3.815259908633412e-05, "loss": 0.6952, "step": 16251 }, { "epoch": 1.64, "learning_rate": 3.8150295953697895e-05, "loss": 1.1738, "step": 16252 }, { "epoch": 1.64, "learning_rate": 3.8147992769196304e-05, "loss": 0.671, "step": 16253 }, { "epoch": 1.64, "learning_rate": 3.8145689532843985e-05, "loss": 0.9789, "step": 16254 }, { "epoch": 1.64, "learning_rate": 3.8143386244655594e-05, "loss": 0.7387, "step": 16255 }, { "epoch": 1.64, "learning_rate": 3.8141082904645794e-05, "loss": 0.647, "step": 16256 }, { "epoch": 1.64, "learning_rate": 3.813877951282923e-05, "loss": 1.1709, "step": 16257 }, { "epoch": 1.64, "learning_rate": 3.8136476069220573e-05, "loss": 0.7616, "step": 16258 }, { "epoch": 1.64, "learning_rate": 3.813417257383448e-05, "loss": 0.6813, "step": 16259 }, { "epoch": 1.64, "learning_rate": 3.813186902668562e-05, "loss": 0.5043, "step": 16260 }, { "epoch": 1.64, "learning_rate": 3.8129565427788616e-05, "loss": 1.024, "step": 16261 }, { "epoch": 1.64, "learning_rate": 3.8127261777158153e-05, "loss": 0.6265, "step": 16262 }, { "epoch": 1.64, "learning_rate": 3.81249580748089e-05, "loss": 0.2606, "step": 16263 }, { "epoch": 1.64, "learning_rate": 3.81226543207555e-05, "loss": 0.6631, "step": 16264 }, { "epoch": 1.64, "learning_rate": 3.812035051501262e-05, "loss": 0.2862, "step": 16265 }, { "epoch": 1.64, "learning_rate": 3.811804665759492e-05, "loss": 0.5328, "step": 16266 }, { "epoch": 1.64, "learning_rate": 3.811574274851706e-05, "loss": 0.433, "step": 16267 }, { "epoch": 1.64, "learning_rate": 3.811343878779369e-05, "loss": 0.1711, "step": 16268 }, { "epoch": 1.64, "learning_rate": 3.811113477543948e-05, "loss": 0.1425, "step": 16269 }, { "epoch": 1.64, "learning_rate": 3.8108830711469104e-05, "loss": 0.8087, "step": 16270 }, { "epoch": 1.64, "learning_rate": 3.8106526595897216e-05, "loss": 1.0867, "step": 16271 }, { "epoch": 1.64, "learning_rate": 3.8104222428738466e-05, "loss": 0.2123, "step": 16272 }, { "epoch": 1.64, "learning_rate": 3.810191821000754e-05, "loss": 0.865, "step": 16273 }, { "epoch": 1.64, "learning_rate": 3.8099613939719086e-05, "loss": 0.2969, "step": 16274 }, { "epoch": 1.64, "learning_rate": 3.809730961788776e-05, "loss": 1.0737, "step": 16275 }, { "epoch": 1.64, "learning_rate": 3.809500524452825e-05, "loss": 0.5894, "step": 16276 }, { "epoch": 1.64, "learning_rate": 3.809270081965519e-05, "loss": 0.6781, "step": 16277 }, { "epoch": 1.64, "learning_rate": 3.8090396343283276e-05, "loss": 1.0461, "step": 16278 }, { "epoch": 1.64, "learning_rate": 3.8088091815427154e-05, "loss": 0.3085, "step": 16279 }, { "epoch": 1.64, "learning_rate": 3.808578723610149e-05, "loss": 0.639, "step": 16280 }, { "epoch": 1.64, "learning_rate": 3.808348260532095e-05, "loss": 1.1521, "step": 16281 }, { "epoch": 1.64, "learning_rate": 3.80811779231002e-05, "loss": 0.6614, "step": 16282 }, { "epoch": 1.64, "learning_rate": 3.807887318945391e-05, "loss": 0.563, "step": 16283 }, { "epoch": 1.64, "learning_rate": 3.807656840439675e-05, "loss": 1.0093, "step": 16284 }, { "epoch": 1.64, "learning_rate": 3.807426356794337e-05, "loss": 0.6107, "step": 16285 }, { "epoch": 1.64, "learning_rate": 3.807195868010846e-05, "loss": 0.5852, "step": 16286 }, { "epoch": 1.64, "learning_rate": 3.8069653740906675e-05, "loss": 0.2497, "step": 16287 }, { "epoch": 1.64, "learning_rate": 3.806734875035267e-05, "loss": 0.4283, "step": 16288 }, { "epoch": 1.64, "learning_rate": 3.806504370846114e-05, "loss": 0.7782, "step": 16289 }, { "epoch": 1.64, "learning_rate": 3.806273861524674e-05, "loss": 0.8143, "step": 16290 }, { "epoch": 1.64, "learning_rate": 3.806043347072413e-05, "loss": 0.3242, "step": 16291 }, { "epoch": 1.64, "learning_rate": 3.8058128274907996e-05, "loss": 1.3236, "step": 16292 }, { "epoch": 1.64, "learning_rate": 3.8055823027812996e-05, "loss": 0.2985, "step": 16293 }, { "epoch": 1.64, "learning_rate": 3.80535177294538e-05, "loss": 0.2264, "step": 16294 }, { "epoch": 1.64, "learning_rate": 3.8051212379845087e-05, "loss": 0.6869, "step": 16295 }, { "epoch": 1.64, "learning_rate": 3.804890697900151e-05, "loss": 0.867, "step": 16296 }, { "epoch": 1.64, "learning_rate": 3.804660152693776e-05, "loss": 0.2067, "step": 16297 }, { "epoch": 1.64, "learning_rate": 3.804429602366849e-05, "loss": 0.972, "step": 16298 }, { "epoch": 1.64, "learning_rate": 3.80419904692084e-05, "loss": 0.5932, "step": 16299 }, { "epoch": 1.64, "learning_rate": 3.803968486357212e-05, "loss": 1.0497, "step": 16300 }, { "epoch": 1.64, "learning_rate": 3.8037379206774355e-05, "loss": 0.2579, "step": 16301 }, { "epoch": 1.64, "learning_rate": 3.803507349882977e-05, "loss": 0.3131, "step": 16302 }, { "epoch": 1.64, "learning_rate": 3.803276773975303e-05, "loss": 0.9266, "step": 16303 }, { "epoch": 1.64, "learning_rate": 3.8030461929558806e-05, "loss": 0.9923, "step": 16304 }, { "epoch": 1.64, "learning_rate": 3.8028156068261785e-05, "loss": 0.6931, "step": 16305 }, { "epoch": 1.64, "learning_rate": 3.8025850155876635e-05, "loss": 0.9464, "step": 16306 }, { "epoch": 1.64, "learning_rate": 3.8023544192418015e-05, "loss": 0.6076, "step": 16307 }, { "epoch": 1.64, "learning_rate": 3.802123817790063e-05, "loss": 0.2997, "step": 16308 }, { "epoch": 1.64, "learning_rate": 3.801893211233911e-05, "loss": 0.1519, "step": 16309 }, { "epoch": 1.64, "learning_rate": 3.801662599574818e-05, "loss": 0.3797, "step": 16310 }, { "epoch": 1.64, "learning_rate": 3.801431982814247e-05, "loss": 0.5911, "step": 16311 }, { "epoch": 1.64, "learning_rate": 3.801201360953671e-05, "loss": 0.4722, "step": 16312 }, { "epoch": 1.64, "learning_rate": 3.800970733994552e-05, "loss": 0.6535, "step": 16313 }, { "epoch": 1.64, "learning_rate": 3.80074010193836e-05, "loss": 0.4185, "step": 16314 }, { "epoch": 1.64, "learning_rate": 3.800509464786564e-05, "loss": 1.0211, "step": 16315 }, { "epoch": 1.64, "learning_rate": 3.800278822540629e-05, "loss": 0.312, "step": 16316 }, { "epoch": 1.64, "learning_rate": 3.800048175202024e-05, "loss": 0.6942, "step": 16317 }, { "epoch": 1.64, "learning_rate": 3.7998175227722175e-05, "loss": 0.6044, "step": 16318 }, { "epoch": 1.64, "learning_rate": 3.799586865252677e-05, "loss": 0.8985, "step": 16319 }, { "epoch": 1.64, "learning_rate": 3.799356202644869e-05, "loss": 1.0119, "step": 16320 }, { "epoch": 1.64, "learning_rate": 3.799125534950262e-05, "loss": 1.044, "step": 16321 }, { "epoch": 1.64, "learning_rate": 3.7988948621703246e-05, "loss": 1.2345, "step": 16322 }, { "epoch": 1.64, "learning_rate": 3.7986641843065233e-05, "loss": 0.5579, "step": 16323 }, { "epoch": 1.64, "learning_rate": 3.798433501360328e-05, "loss": 0.4594, "step": 16324 }, { "epoch": 1.64, "learning_rate": 3.798202813333205e-05, "loss": 1.0487, "step": 16325 }, { "epoch": 1.64, "learning_rate": 3.797972120226623e-05, "loss": 1.2667, "step": 16326 }, { "epoch": 1.64, "learning_rate": 3.79774142204205e-05, "loss": 0.5416, "step": 16327 }, { "epoch": 1.65, "learning_rate": 3.797510718780956e-05, "loss": 0.3775, "step": 16328 }, { "epoch": 1.65, "learning_rate": 3.797280010444805e-05, "loss": 0.7215, "step": 16329 }, { "epoch": 1.65, "learning_rate": 3.7970492970350675e-05, "loss": 1.07, "step": 16330 }, { "epoch": 1.65, "learning_rate": 3.796818578553212e-05, "loss": 0.2297, "step": 16331 }, { "epoch": 1.65, "learning_rate": 3.796587855000706e-05, "loss": 0.7404, "step": 16332 }, { "epoch": 1.65, "learning_rate": 3.7963571263790174e-05, "loss": 1.0187, "step": 16333 }, { "epoch": 1.65, "learning_rate": 3.7961263926896164e-05, "loss": 0.4698, "step": 16334 }, { "epoch": 1.65, "learning_rate": 3.7958956539339684e-05, "loss": 0.3898, "step": 16335 }, { "epoch": 1.65, "learning_rate": 3.7956649101135445e-05, "loss": 0.8526, "step": 16336 }, { "epoch": 1.65, "learning_rate": 3.7954341612298104e-05, "loss": 0.5389, "step": 16337 }, { "epoch": 1.65, "learning_rate": 3.795203407284237e-05, "loss": 0.565, "step": 16338 }, { "epoch": 1.65, "learning_rate": 3.7949726482782906e-05, "loss": 0.6942, "step": 16339 }, { "epoch": 1.65, "learning_rate": 3.794741884213442e-05, "loss": 1.1793, "step": 16340 }, { "epoch": 1.65, "learning_rate": 3.794511115091158e-05, "loss": 0.2186, "step": 16341 }, { "epoch": 1.65, "learning_rate": 3.794280340912907e-05, "loss": 0.5693, "step": 16342 }, { "epoch": 1.65, "learning_rate": 3.794049561680158e-05, "loss": 0.4395, "step": 16343 }, { "epoch": 1.65, "learning_rate": 3.793818777394381e-05, "loss": 0.4244, "step": 16344 }, { "epoch": 1.65, "learning_rate": 3.793587988057042e-05, "loss": 0.8781, "step": 16345 }, { "epoch": 1.65, "learning_rate": 3.79335719366961e-05, "loss": 0.5574, "step": 16346 }, { "epoch": 1.65, "learning_rate": 3.793126394233557e-05, "loss": 0.8301, "step": 16347 }, { "epoch": 1.65, "learning_rate": 3.792895589750348e-05, "loss": 1.3069, "step": 16348 }, { "epoch": 1.65, "learning_rate": 3.792664780221453e-05, "loss": 0.6653, "step": 16349 }, { "epoch": 1.65, "learning_rate": 3.7924339656483405e-05, "loss": 0.388, "step": 16350 }, { "epoch": 1.65, "learning_rate": 3.792203146032481e-05, "loss": 0.6028, "step": 16351 }, { "epoch": 1.65, "learning_rate": 3.791972321375342e-05, "loss": 1.1577, "step": 16352 }, { "epoch": 1.65, "learning_rate": 3.791741491678391e-05, "loss": 0.5557, "step": 16353 }, { "epoch": 1.65, "learning_rate": 3.7915106569431e-05, "loss": 1.3979, "step": 16354 }, { "epoch": 1.65, "learning_rate": 3.791279817170935e-05, "loss": 0.8167, "step": 16355 }, { "epoch": 1.65, "learning_rate": 3.791048972363367e-05, "loss": 0.6539, "step": 16356 }, { "epoch": 1.65, "learning_rate": 3.790818122521864e-05, "loss": 1.35, "step": 16357 }, { "epoch": 1.65, "learning_rate": 3.790587267647895e-05, "loss": 0.1218, "step": 16358 }, { "epoch": 1.65, "learning_rate": 3.7903564077429297e-05, "loss": 0.5114, "step": 16359 }, { "epoch": 1.65, "learning_rate": 3.790125542808437e-05, "loss": 0.2457, "step": 16360 }, { "epoch": 1.65, "learning_rate": 3.789894672845887e-05, "loss": 0.8706, "step": 16361 }, { "epoch": 1.65, "learning_rate": 3.7896637978567454e-05, "loss": 1.2239, "step": 16362 }, { "epoch": 1.65, "learning_rate": 3.789432917842485e-05, "loss": 0.2418, "step": 16363 }, { "epoch": 1.65, "learning_rate": 3.789202032804574e-05, "loss": 0.209, "step": 16364 }, { "epoch": 1.65, "learning_rate": 3.788971142744481e-05, "loss": 0.5956, "step": 16365 }, { "epoch": 1.65, "learning_rate": 3.7887402476636754e-05, "loss": 0.3758, "step": 16366 }, { "epoch": 1.65, "learning_rate": 3.788509347563629e-05, "loss": 0.3362, "step": 16367 }, { "epoch": 1.65, "learning_rate": 3.7882784424458075e-05, "loss": 1.2798, "step": 16368 }, { "epoch": 1.65, "learning_rate": 3.788047532311682e-05, "loss": 0.9999, "step": 16369 }, { "epoch": 1.65, "learning_rate": 3.7878166171627225e-05, "loss": 0.3441, "step": 16370 }, { "epoch": 1.65, "learning_rate": 3.7875856970003965e-05, "loss": 0.6362, "step": 16371 }, { "epoch": 1.65, "learning_rate": 3.787354771826175e-05, "loss": 0.7869, "step": 16372 }, { "epoch": 1.65, "learning_rate": 3.7871238416415276e-05, "loss": 0.9169, "step": 16373 }, { "epoch": 1.65, "learning_rate": 3.786892906447924e-05, "loss": 0.9629, "step": 16374 }, { "epoch": 1.65, "learning_rate": 3.786661966246832e-05, "loss": 0.5867, "step": 16375 }, { "epoch": 1.65, "learning_rate": 3.786431021039724e-05, "loss": 0.7519, "step": 16376 }, { "epoch": 1.65, "learning_rate": 3.7862000708280674e-05, "loss": 0.9558, "step": 16377 }, { "epoch": 1.65, "learning_rate": 3.785969115613333e-05, "loss": 0.4197, "step": 16378 }, { "epoch": 1.65, "learning_rate": 3.78573815539699e-05, "loss": 0.3038, "step": 16379 }, { "epoch": 1.65, "learning_rate": 3.7855071901805076e-05, "loss": 0.5446, "step": 16380 }, { "epoch": 1.65, "learning_rate": 3.785276219965357e-05, "loss": 0.566, "step": 16381 }, { "epoch": 1.65, "learning_rate": 3.785045244753006e-05, "loss": 0.3252, "step": 16382 }, { "epoch": 1.65, "learning_rate": 3.784814264544928e-05, "loss": 0.7232, "step": 16383 }, { "epoch": 1.65, "learning_rate": 3.7845832793425896e-05, "loss": 0.8587, "step": 16384 }, { "epoch": 1.65, "learning_rate": 3.784352289147461e-05, "loss": 0.8756, "step": 16385 }, { "epoch": 1.65, "learning_rate": 3.784121293961014e-05, "loss": 1.3468, "step": 16386 }, { "epoch": 1.65, "learning_rate": 3.7838902937847165e-05, "loss": 0.5562, "step": 16387 }, { "epoch": 1.65, "learning_rate": 3.783659288620039e-05, "loss": 0.5118, "step": 16388 }, { "epoch": 1.65, "learning_rate": 3.783428278468453e-05, "loss": 0.2799, "step": 16389 }, { "epoch": 1.65, "learning_rate": 3.783197263331426e-05, "loss": 0.5987, "step": 16390 }, { "epoch": 1.65, "learning_rate": 3.78296624321043e-05, "loss": 0.3564, "step": 16391 }, { "epoch": 1.65, "learning_rate": 3.782735218106936e-05, "loss": 0.2412, "step": 16392 }, { "epoch": 1.65, "learning_rate": 3.782504188022412e-05, "loss": 0.5512, "step": 16393 }, { "epoch": 1.65, "learning_rate": 3.7822731529583285e-05, "loss": 0.4282, "step": 16394 }, { "epoch": 1.65, "learning_rate": 3.7820421129161566e-05, "loss": 0.3288, "step": 16395 }, { "epoch": 1.65, "learning_rate": 3.781811067897367e-05, "loss": 0.2814, "step": 16396 }, { "epoch": 1.65, "learning_rate": 3.781580017903428e-05, "loss": 1.2107, "step": 16397 }, { "epoch": 1.65, "learning_rate": 3.7813489629358114e-05, "loss": 0.897, "step": 16398 }, { "epoch": 1.65, "learning_rate": 3.7811179029959885e-05, "loss": 1.0238, "step": 16399 }, { "epoch": 1.65, "learning_rate": 3.780886838085427e-05, "loss": 0.8764, "step": 16400 }, { "epoch": 1.65, "learning_rate": 3.780655768205599e-05, "loss": 0.6704, "step": 16401 }, { "epoch": 1.65, "learning_rate": 3.780424693357975e-05, "loss": 0.32, "step": 16402 }, { "epoch": 1.65, "learning_rate": 3.780193613544025e-05, "loss": 1.3425, "step": 16403 }, { "epoch": 1.65, "learning_rate": 3.779962528765219e-05, "loss": 0.5475, "step": 16404 }, { "epoch": 1.65, "learning_rate": 3.7797314390230284e-05, "loss": 0.8053, "step": 16405 }, { "epoch": 1.65, "learning_rate": 3.779500344318925e-05, "loss": 0.4263, "step": 16406 }, { "epoch": 1.65, "learning_rate": 3.779269244654376e-05, "loss": 0.2853, "step": 16407 }, { "epoch": 1.65, "learning_rate": 3.779038140030854e-05, "loss": 0.5272, "step": 16408 }, { "epoch": 1.65, "learning_rate": 3.778807030449832e-05, "loss": 0.36, "step": 16409 }, { "epoch": 1.65, "learning_rate": 3.7785759159127764e-05, "loss": 1.0617, "step": 16410 }, { "epoch": 1.65, "learning_rate": 3.7783447964211604e-05, "loss": 0.9754, "step": 16411 }, { "epoch": 1.65, "learning_rate": 3.778113671976454e-05, "loss": 0.2747, "step": 16412 }, { "epoch": 1.65, "learning_rate": 3.777882542580128e-05, "loss": 0.9229, "step": 16413 }, { "epoch": 1.65, "learning_rate": 3.777651408233654e-05, "loss": 0.9022, "step": 16414 }, { "epoch": 1.65, "learning_rate": 3.777420268938502e-05, "loss": 0.4461, "step": 16415 }, { "epoch": 1.65, "learning_rate": 3.777189124696143e-05, "loss": 0.9298, "step": 16416 }, { "epoch": 1.65, "learning_rate": 3.776957975508048e-05, "loss": 0.3318, "step": 16417 }, { "epoch": 1.65, "learning_rate": 3.776726821375688e-05, "loss": 1.046, "step": 16418 }, { "epoch": 1.65, "learning_rate": 3.7764956623005345e-05, "loss": 0.3348, "step": 16419 }, { "epoch": 1.65, "learning_rate": 3.7762644982840576e-05, "loss": 0.3354, "step": 16420 }, { "epoch": 1.65, "learning_rate": 3.7760333293277286e-05, "loss": 0.1841, "step": 16421 }, { "epoch": 1.65, "learning_rate": 3.775802155433019e-05, "loss": 1.2023, "step": 16422 }, { "epoch": 1.65, "learning_rate": 3.7755709766013995e-05, "loss": 1.0246, "step": 16423 }, { "epoch": 1.65, "learning_rate": 3.775339792834341e-05, "loss": 0.3066, "step": 16424 }, { "epoch": 1.65, "learning_rate": 3.775108604133316e-05, "loss": 0.422, "step": 16425 }, { "epoch": 1.65, "learning_rate": 3.7748774104997935e-05, "loss": 1.2148, "step": 16426 }, { "epoch": 1.66, "learning_rate": 3.774646211935247e-05, "loss": 0.563, "step": 16427 }, { "epoch": 1.66, "learning_rate": 3.774415008441146e-05, "loss": 0.3982, "step": 16428 }, { "epoch": 1.66, "learning_rate": 3.7741838000189636e-05, "loss": 0.5766, "step": 16429 }, { "epoch": 1.66, "learning_rate": 3.773952586670168e-05, "loss": 0.8845, "step": 16430 }, { "epoch": 1.66, "learning_rate": 3.773721368396234e-05, "loss": 0.9051, "step": 16431 }, { "epoch": 1.66, "learning_rate": 3.773490145198632e-05, "loss": 0.4955, "step": 16432 }, { "epoch": 1.66, "learning_rate": 3.773258917078832e-05, "loss": 0.3422, "step": 16433 }, { "epoch": 1.66, "learning_rate": 3.773027684038306e-05, "loss": 0.8501, "step": 16434 }, { "epoch": 1.66, "learning_rate": 3.772796446078527e-05, "loss": 1.7295, "step": 16435 }, { "epoch": 1.66, "learning_rate": 3.7725652032009645e-05, "loss": 0.9269, "step": 16436 }, { "epoch": 1.66, "learning_rate": 3.772333955407091e-05, "loss": 0.7484, "step": 16437 }, { "epoch": 1.66, "learning_rate": 3.7721027026983794e-05, "loss": 1.3309, "step": 16438 }, { "epoch": 1.66, "learning_rate": 3.771871445076298e-05, "loss": 0.0608, "step": 16439 }, { "epoch": 1.66, "learning_rate": 3.771640182542322e-05, "loss": 1.2012, "step": 16440 }, { "epoch": 1.66, "learning_rate": 3.771408915097921e-05, "loss": 0.7465, "step": 16441 }, { "epoch": 1.66, "learning_rate": 3.7711776427445664e-05, "loss": 0.7193, "step": 16442 }, { "epoch": 1.66, "learning_rate": 3.770946365483731e-05, "loss": 0.4262, "step": 16443 }, { "epoch": 1.66, "learning_rate": 3.770715083316886e-05, "loss": 1.0619, "step": 16444 }, { "epoch": 1.66, "learning_rate": 3.7704837962455044e-05, "loss": 0.5991, "step": 16445 }, { "epoch": 1.66, "learning_rate": 3.7702525042710554e-05, "loss": 1.3874, "step": 16446 }, { "epoch": 1.66, "learning_rate": 3.770021207395014e-05, "loss": 0.2803, "step": 16447 }, { "epoch": 1.66, "learning_rate": 3.76978990561885e-05, "loss": 0.7135, "step": 16448 }, { "epoch": 1.66, "learning_rate": 3.769558598944036e-05, "loss": 0.6007, "step": 16449 }, { "epoch": 1.66, "learning_rate": 3.769327287372043e-05, "loss": 0.4532, "step": 16450 }, { "epoch": 1.66, "learning_rate": 3.769095970904345e-05, "loss": 0.5085, "step": 16451 }, { "epoch": 1.66, "learning_rate": 3.768864649542412e-05, "loss": 1.1119, "step": 16452 }, { "epoch": 1.66, "learning_rate": 3.768633323287717e-05, "loss": 0.7354, "step": 16453 }, { "epoch": 1.66, "learning_rate": 3.7684019921417326e-05, "loss": 0.5292, "step": 16454 }, { "epoch": 1.66, "learning_rate": 3.7681706561059295e-05, "loss": 0.1552, "step": 16455 }, { "epoch": 1.66, "learning_rate": 3.7679393151817806e-05, "loss": 0.617, "step": 16456 }, { "epoch": 1.66, "learning_rate": 3.767707969370759e-05, "loss": 0.1747, "step": 16457 }, { "epoch": 1.66, "learning_rate": 3.767476618674335e-05, "loss": 0.159, "step": 16458 }, { "epoch": 1.66, "learning_rate": 3.767245263093982e-05, "loss": 1.3226, "step": 16459 }, { "epoch": 1.66, "learning_rate": 3.7670139026311713e-05, "loss": 0.6244, "step": 16460 }, { "epoch": 1.66, "learning_rate": 3.766782537287377e-05, "loss": 0.2021, "step": 16461 }, { "epoch": 1.66, "learning_rate": 3.76655116706407e-05, "loss": 0.21, "step": 16462 }, { "epoch": 1.66, "learning_rate": 3.766319791962722e-05, "loss": 0.4962, "step": 16463 }, { "epoch": 1.66, "learning_rate": 3.766088411984809e-05, "loss": 0.4326, "step": 16464 }, { "epoch": 1.66, "learning_rate": 3.765857027131798e-05, "loss": 0.8812, "step": 16465 }, { "epoch": 1.66, "learning_rate": 3.7656256374051654e-05, "loss": 1.0168, "step": 16466 }, { "epoch": 1.66, "learning_rate": 3.7653942428063825e-05, "loss": 0.94, "step": 16467 }, { "epoch": 1.66, "learning_rate": 3.7651628433369215e-05, "loss": 1.2751, "step": 16468 }, { "epoch": 1.66, "learning_rate": 3.764931438998255e-05, "loss": 0.8855, "step": 16469 }, { "epoch": 1.66, "learning_rate": 3.764700029791858e-05, "loss": 0.2978, "step": 16470 }, { "epoch": 1.66, "learning_rate": 3.764468615719198e-05, "loss": 0.8311, "step": 16471 }, { "epoch": 1.66, "learning_rate": 3.764237196781751e-05, "loss": 0.5538, "step": 16472 }, { "epoch": 1.66, "learning_rate": 3.7640057729809904e-05, "loss": 0.1885, "step": 16473 }, { "epoch": 1.66, "learning_rate": 3.763774344318387e-05, "loss": 0.5499, "step": 16474 }, { "epoch": 1.66, "learning_rate": 3.763542910795415e-05, "loss": 0.7284, "step": 16475 }, { "epoch": 1.66, "learning_rate": 3.763311472413545e-05, "loss": 1.017, "step": 16476 }, { "epoch": 1.66, "learning_rate": 3.763080029174253e-05, "loss": 0.3035, "step": 16477 }, { "epoch": 1.66, "learning_rate": 3.762848581079008e-05, "loss": 1.051, "step": 16478 }, { "epoch": 1.66, "learning_rate": 3.762617128129286e-05, "loss": 0.2634, "step": 16479 }, { "epoch": 1.66, "learning_rate": 3.7623856703265595e-05, "loss": 0.33, "step": 16480 }, { "epoch": 1.66, "learning_rate": 3.762154207672299e-05, "loss": 0.2672, "step": 16481 }, { "epoch": 1.66, "learning_rate": 3.76192274016798e-05, "loss": 0.4817, "step": 16482 }, { "epoch": 1.66, "learning_rate": 3.761691267815074e-05, "loss": 0.4427, "step": 16483 }, { "epoch": 1.66, "learning_rate": 3.761459790615055e-05, "loss": 0.4209, "step": 16484 }, { "epoch": 1.66, "learning_rate": 3.761228308569395e-05, "loss": 0.1969, "step": 16485 }, { "epoch": 1.66, "learning_rate": 3.760996821679568e-05, "loss": 1.0766, "step": 16486 }, { "epoch": 1.66, "learning_rate": 3.760765329947047e-05, "loss": 0.466, "step": 16487 }, { "epoch": 1.66, "learning_rate": 3.760533833373305e-05, "loss": 1.1919, "step": 16488 }, { "epoch": 1.66, "learning_rate": 3.7603023319598144e-05, "loss": 0.5956, "step": 16489 }, { "epoch": 1.66, "learning_rate": 3.7600708257080486e-05, "loss": 0.5602, "step": 16490 }, { "epoch": 1.66, "learning_rate": 3.759839314619482e-05, "loss": 0.4595, "step": 16491 }, { "epoch": 1.66, "learning_rate": 3.7596077986955866e-05, "loss": 0.4241, "step": 16492 }, { "epoch": 1.66, "learning_rate": 3.7593762779378374e-05, "loss": 0.8481, "step": 16493 }, { "epoch": 1.66, "learning_rate": 3.759144752347705e-05, "loss": 0.6525, "step": 16494 }, { "epoch": 1.66, "learning_rate": 3.758913221926664e-05, "loss": 0.461, "step": 16495 }, { "epoch": 1.66, "learning_rate": 3.758681686676189e-05, "loss": 0.7358, "step": 16496 }, { "epoch": 1.66, "learning_rate": 3.758450146597752e-05, "loss": 0.6207, "step": 16497 }, { "epoch": 1.66, "learning_rate": 3.758218601692827e-05, "loss": 0.9634, "step": 16498 }, { "epoch": 1.66, "learning_rate": 3.757987051962886e-05, "loss": 0.4807, "step": 16499 }, { "epoch": 1.66, "learning_rate": 3.7577554974094054e-05, "loss": 0.6847, "step": 16500 }, { "epoch": 1.66, "learning_rate": 3.7575239380338555e-05, "loss": 0.0344, "step": 16501 }, { "epoch": 1.66, "learning_rate": 3.757292373837713e-05, "loss": 0.2108, "step": 16502 }, { "epoch": 1.66, "learning_rate": 3.757060804822449e-05, "loss": 0.4789, "step": 16503 }, { "epoch": 1.66, "learning_rate": 3.756829230989539e-05, "loss": 0.7217, "step": 16504 }, { "epoch": 1.66, "learning_rate": 3.756597652340455e-05, "loss": 0.9895, "step": 16505 }, { "epoch": 1.66, "learning_rate": 3.756366068876671e-05, "loss": 0.3449, "step": 16506 }, { "epoch": 1.66, "learning_rate": 3.7561344805996616e-05, "loss": 0.7644, "step": 16507 }, { "epoch": 1.66, "learning_rate": 3.755902887510899e-05, "loss": 0.4336, "step": 16508 }, { "epoch": 1.66, "learning_rate": 3.75567128961186e-05, "loss": 0.9413, "step": 16509 }, { "epoch": 1.66, "learning_rate": 3.755439686904015e-05, "loss": 1.2095, "step": 16510 }, { "epoch": 1.66, "learning_rate": 3.755208079388839e-05, "loss": 0.4561, "step": 16511 }, { "epoch": 1.66, "learning_rate": 3.7549764670678064e-05, "loss": 1.1271, "step": 16512 }, { "epoch": 1.66, "learning_rate": 3.754744849942391e-05, "loss": 0.9202, "step": 16513 }, { "epoch": 1.66, "learning_rate": 3.754513228014067e-05, "loss": 0.7462, "step": 16514 }, { "epoch": 1.66, "learning_rate": 3.7542816012843065e-05, "loss": 0.1309, "step": 16515 }, { "epoch": 1.66, "learning_rate": 3.754049969754586e-05, "loss": 0.0339, "step": 16516 }, { "epoch": 1.66, "learning_rate": 3.753818333426377e-05, "loss": 1.7444, "step": 16517 }, { "epoch": 1.66, "learning_rate": 3.753586692301156e-05, "loss": 1.0416, "step": 16518 }, { "epoch": 1.66, "learning_rate": 3.753355046380396e-05, "loss": 0.4612, "step": 16519 }, { "epoch": 1.66, "learning_rate": 3.7531233956655715e-05, "loss": 0.8544, "step": 16520 }, { "epoch": 1.66, "learning_rate": 3.752891740158155e-05, "loss": 0.5059, "step": 16521 }, { "epoch": 1.66, "learning_rate": 3.752660079859623e-05, "loss": 0.2525, "step": 16522 }, { "epoch": 1.66, "learning_rate": 3.7524284147714474e-05, "loss": 0.3391, "step": 16523 }, { "epoch": 1.66, "learning_rate": 3.7521967448951044e-05, "loss": 0.5027, "step": 16524 }, { "epoch": 1.67, "learning_rate": 3.751965070232068e-05, "loss": 0.8218, "step": 16525 }, { "epoch": 1.67, "learning_rate": 3.7517333907838106e-05, "loss": 0.6304, "step": 16526 }, { "epoch": 1.67, "learning_rate": 3.751501706551808e-05, "loss": 0.9583, "step": 16527 }, { "epoch": 1.67, "learning_rate": 3.751270017537536e-05, "loss": 0.303, "step": 16528 }, { "epoch": 1.67, "learning_rate": 3.751038323742466e-05, "loss": 0.1826, "step": 16529 }, { "epoch": 1.67, "learning_rate": 3.7508066251680744e-05, "loss": 0.7533, "step": 16530 }, { "epoch": 1.67, "learning_rate": 3.750574921815834e-05, "loss": 0.4223, "step": 16531 }, { "epoch": 1.67, "learning_rate": 3.750343213687222e-05, "loss": 0.2243, "step": 16532 }, { "epoch": 1.67, "learning_rate": 3.75011150078371e-05, "loss": 1.0361, "step": 16533 }, { "epoch": 1.67, "learning_rate": 3.749879783106774e-05, "loss": 0.8799, "step": 16534 }, { "epoch": 1.67, "learning_rate": 3.749648060657889e-05, "loss": 0.4453, "step": 16535 }, { "epoch": 1.67, "learning_rate": 3.749416333438528e-05, "loss": 0.181, "step": 16536 }, { "epoch": 1.67, "learning_rate": 3.749184601450167e-05, "loss": 0.678, "step": 16537 }, { "epoch": 1.67, "learning_rate": 3.7489528646942803e-05, "loss": 0.4251, "step": 16538 }, { "epoch": 1.67, "learning_rate": 3.7487211231723426e-05, "loss": 0.3088, "step": 16539 }, { "epoch": 1.67, "learning_rate": 3.7484893768858276e-05, "loss": 0.9262, "step": 16540 }, { "epoch": 1.67, "learning_rate": 3.748257625836212e-05, "loss": 0.2436, "step": 16541 }, { "epoch": 1.67, "learning_rate": 3.748025870024969e-05, "loss": 0.9657, "step": 16542 }, { "epoch": 1.67, "learning_rate": 3.7477941094535736e-05, "loss": 0.2023, "step": 16543 }, { "epoch": 1.67, "learning_rate": 3.747562344123502e-05, "loss": 0.5115, "step": 16544 }, { "epoch": 1.67, "learning_rate": 3.747330574036227e-05, "loss": 0.4546, "step": 16545 }, { "epoch": 1.67, "learning_rate": 3.7470987991932257e-05, "loss": 0.4373, "step": 16546 }, { "epoch": 1.67, "learning_rate": 3.74686701959597e-05, "loss": 0.785, "step": 16547 }, { "epoch": 1.67, "learning_rate": 3.746635235245939e-05, "loss": 0.2206, "step": 16548 }, { "epoch": 1.67, "learning_rate": 3.746403446144603e-05, "loss": 0.3453, "step": 16549 }, { "epoch": 1.67, "learning_rate": 3.746171652293441e-05, "loss": 0.3592, "step": 16550 }, { "epoch": 1.67, "learning_rate": 3.7459398536939264e-05, "loss": 0.535, "step": 16551 }, { "epoch": 1.67, "learning_rate": 3.7457080503475336e-05, "loss": 0.5156, "step": 16552 }, { "epoch": 1.67, "learning_rate": 3.745476242255739e-05, "loss": 0.4859, "step": 16553 }, { "epoch": 1.67, "learning_rate": 3.745244429420017e-05, "loss": 0.2939, "step": 16554 }, { "epoch": 1.67, "learning_rate": 3.745012611841843e-05, "loss": 0.4818, "step": 16555 }, { "epoch": 1.67, "learning_rate": 3.744780789522692e-05, "loss": 0.6775, "step": 16556 }, { "epoch": 1.67, "learning_rate": 3.74454896246404e-05, "loss": 0.2462, "step": 16557 }, { "epoch": 1.67, "learning_rate": 3.744317130667361e-05, "loss": 0.8268, "step": 16558 }, { "epoch": 1.67, "learning_rate": 3.7440852941341316e-05, "loss": 0.2794, "step": 16559 }, { "epoch": 1.67, "learning_rate": 3.7438534528658264e-05, "loss": 0.2119, "step": 16560 }, { "epoch": 1.67, "learning_rate": 3.74362160686392e-05, "loss": 0.6676, "step": 16561 }, { "epoch": 1.67, "learning_rate": 3.743389756129889e-05, "loss": 0.2463, "step": 16562 }, { "epoch": 1.67, "learning_rate": 3.743157900665208e-05, "loss": 0.6637, "step": 16563 }, { "epoch": 1.67, "learning_rate": 3.742926040471354e-05, "loss": 0.7645, "step": 16564 }, { "epoch": 1.67, "learning_rate": 3.7426941755498e-05, "loss": 0.5357, "step": 16565 }, { "epoch": 1.67, "learning_rate": 3.7424623059020236e-05, "loss": 0.9902, "step": 16566 }, { "epoch": 1.67, "learning_rate": 3.7422304315294984e-05, "loss": 0.4019, "step": 16567 }, { "epoch": 1.67, "learning_rate": 3.7419985524337026e-05, "loss": 0.3444, "step": 16568 }, { "epoch": 1.67, "learning_rate": 3.7417666686161095e-05, "loss": 0.8237, "step": 16569 }, { "epoch": 1.67, "learning_rate": 3.741534780078195e-05, "loss": 0.6288, "step": 16570 }, { "epoch": 1.67, "learning_rate": 3.741302886821437e-05, "loss": 0.4711, "step": 16571 }, { "epoch": 1.67, "learning_rate": 3.7410709888473074e-05, "loss": 0.6469, "step": 16572 }, { "epoch": 1.67, "learning_rate": 3.740839086157285e-05, "loss": 0.308, "step": 16573 }, { "epoch": 1.67, "learning_rate": 3.7406071787528444e-05, "loss": 0.8592, "step": 16574 }, { "epoch": 1.67, "learning_rate": 3.740375266635462e-05, "loss": 0.5063, "step": 16575 }, { "epoch": 1.67, "learning_rate": 3.740143349806612e-05, "loss": 1.1686, "step": 16576 }, { "epoch": 1.67, "learning_rate": 3.739911428267772e-05, "loss": 0.5073, "step": 16577 }, { "epoch": 1.67, "learning_rate": 3.739679502020417e-05, "loss": 1.2801, "step": 16578 }, { "epoch": 1.67, "learning_rate": 3.7394475710660223e-05, "loss": 0.1245, "step": 16579 }, { "epoch": 1.67, "learning_rate": 3.739215635406065e-05, "loss": 1.163, "step": 16580 }, { "epoch": 1.67, "learning_rate": 3.7389836950420214e-05, "loss": 1.0878, "step": 16581 }, { "epoch": 1.67, "learning_rate": 3.7387517499753654e-05, "loss": 0.6081, "step": 16582 }, { "epoch": 1.67, "learning_rate": 3.7385198002075756e-05, "loss": 0.201, "step": 16583 }, { "epoch": 1.67, "learning_rate": 3.7382878457401265e-05, "loss": 0.6405, "step": 16584 }, { "epoch": 1.67, "learning_rate": 3.738055886574493e-05, "loss": 0.4919, "step": 16585 }, { "epoch": 1.67, "learning_rate": 3.737823922712154e-05, "loss": 0.9842, "step": 16586 }, { "epoch": 1.67, "learning_rate": 3.737591954154585e-05, "loss": 1.7415, "step": 16587 }, { "epoch": 1.67, "learning_rate": 3.737359980903259e-05, "loss": 0.9512, "step": 16588 }, { "epoch": 1.67, "learning_rate": 3.737128002959656e-05, "loss": 0.3465, "step": 16589 }, { "epoch": 1.67, "learning_rate": 3.736896020325251e-05, "loss": 0.2532, "step": 16590 }, { "epoch": 1.67, "learning_rate": 3.73666403300152e-05, "loss": 0.2439, "step": 16591 }, { "epoch": 1.67, "learning_rate": 3.736432040989939e-05, "loss": 0.1258, "step": 16592 }, { "epoch": 1.67, "learning_rate": 3.7362000442919836e-05, "loss": 0.7251, "step": 16593 }, { "epoch": 1.67, "learning_rate": 3.735968042909133e-05, "loss": 0.687, "step": 16594 }, { "epoch": 1.67, "learning_rate": 3.735736036842861e-05, "loss": 0.8299, "step": 16595 }, { "epoch": 1.67, "learning_rate": 3.735504026094645e-05, "loss": 0.5299, "step": 16596 }, { "epoch": 1.67, "learning_rate": 3.735272010665961e-05, "loss": 0.7655, "step": 16597 }, { "epoch": 1.67, "learning_rate": 3.735039990558285e-05, "loss": 0.3658, "step": 16598 }, { "epoch": 1.67, "learning_rate": 3.734807965773095e-05, "loss": 0.5764, "step": 16599 }, { "epoch": 1.67, "learning_rate": 3.734575936311867e-05, "loss": 0.5016, "step": 16600 }, { "epoch": 1.67, "learning_rate": 3.734343902176076e-05, "loss": 1.0583, "step": 16601 }, { "epoch": 1.67, "learning_rate": 3.7341118633672005e-05, "loss": 0.4729, "step": 16602 }, { "epoch": 1.67, "learning_rate": 3.7338798198867165e-05, "loss": 0.8835, "step": 16603 }, { "epoch": 1.67, "learning_rate": 3.733647771736099e-05, "loss": 1.026, "step": 16604 }, { "epoch": 1.67, "learning_rate": 3.733415718916828e-05, "loss": 0.8544, "step": 16605 }, { "epoch": 1.67, "learning_rate": 3.733183661430378e-05, "loss": 0.8754, "step": 16606 }, { "epoch": 1.67, "learning_rate": 3.732951599278226e-05, "loss": 1.2381, "step": 16607 }, { "epoch": 1.67, "learning_rate": 3.732719532461849e-05, "loss": 0.6336, "step": 16608 }, { "epoch": 1.67, "learning_rate": 3.732487460982722e-05, "loss": 1.4762, "step": 16609 }, { "epoch": 1.67, "learning_rate": 3.732255384842326e-05, "loss": 0.7985, "step": 16610 }, { "epoch": 1.67, "learning_rate": 3.732023304042133e-05, "loss": 0.9155, "step": 16611 }, { "epoch": 1.67, "learning_rate": 3.731791218583624e-05, "loss": 0.4887, "step": 16612 }, { "epoch": 1.67, "learning_rate": 3.731559128468273e-05, "loss": 1.0906, "step": 16613 }, { "epoch": 1.67, "learning_rate": 3.7313270336975584e-05, "loss": 0.1867, "step": 16614 }, { "epoch": 1.67, "learning_rate": 3.731094934272957e-05, "loss": 0.4202, "step": 16615 }, { "epoch": 1.67, "learning_rate": 3.7308628301959453e-05, "loss": 0.7505, "step": 16616 }, { "epoch": 1.67, "learning_rate": 3.7306307214680004e-05, "loss": 0.8085, "step": 16617 }, { "epoch": 1.67, "learning_rate": 3.7303986080905994e-05, "loss": 0.8946, "step": 16618 }, { "epoch": 1.67, "learning_rate": 3.7301664900652203e-05, "loss": 0.9606, "step": 16619 }, { "epoch": 1.67, "learning_rate": 3.7299343673933384e-05, "loss": 0.6981, "step": 16620 }, { "epoch": 1.67, "learning_rate": 3.729702240076433e-05, "loss": 0.1973, "step": 16621 }, { "epoch": 1.67, "learning_rate": 3.7294701081159787e-05, "loss": 1.1176, "step": 16622 }, { "epoch": 1.67, "learning_rate": 3.7292379715134554e-05, "loss": 0.8007, "step": 16623 }, { "epoch": 1.68, "learning_rate": 3.729005830270339e-05, "loss": 0.7457, "step": 16624 }, { "epoch": 1.68, "learning_rate": 3.7287736843881054e-05, "loss": 0.4581, "step": 16625 }, { "epoch": 1.68, "learning_rate": 3.728541533868235e-05, "loss": 0.5578, "step": 16626 }, { "epoch": 1.68, "learning_rate": 3.7283093787122025e-05, "loss": 0.3233, "step": 16627 }, { "epoch": 1.68, "learning_rate": 3.728077218921486e-05, "loss": 0.3526, "step": 16628 }, { "epoch": 1.68, "learning_rate": 3.727845054497564e-05, "loss": 0.4676, "step": 16629 }, { "epoch": 1.68, "learning_rate": 3.7276128854419126e-05, "loss": 0.881, "step": 16630 }, { "epoch": 1.68, "learning_rate": 3.72738071175601e-05, "loss": 0.7963, "step": 16631 }, { "epoch": 1.68, "learning_rate": 3.727148533441333e-05, "loss": 0.8453, "step": 16632 }, { "epoch": 1.68, "learning_rate": 3.726916350499359e-05, "loss": 0.6614, "step": 16633 }, { "epoch": 1.68, "learning_rate": 3.726684162931566e-05, "loss": 1.0378, "step": 16634 }, { "epoch": 1.68, "learning_rate": 3.726451970739431e-05, "loss": 0.2838, "step": 16635 }, { "epoch": 1.68, "learning_rate": 3.726219773924432e-05, "loss": 0.8502, "step": 16636 }, { "epoch": 1.68, "learning_rate": 3.725987572488047e-05, "loss": 0.5824, "step": 16637 }, { "epoch": 1.68, "learning_rate": 3.725755366431753e-05, "loss": 1.2731, "step": 16638 }, { "epoch": 1.68, "learning_rate": 3.7255231557570286e-05, "loss": 0.8286, "step": 16639 }, { "epoch": 1.68, "learning_rate": 3.725290940465351e-05, "loss": 0.7659, "step": 16640 }, { "epoch": 1.68, "learning_rate": 3.725058720558197e-05, "loss": 0.9819, "step": 16641 }, { "epoch": 1.68, "learning_rate": 3.724826496037046e-05, "loss": 0.4462, "step": 16642 }, { "epoch": 1.68, "learning_rate": 3.7245942669033734e-05, "loss": 0.9617, "step": 16643 }, { "epoch": 1.68, "learning_rate": 3.72436203315866e-05, "loss": 0.8348, "step": 16644 }, { "epoch": 1.68, "learning_rate": 3.724129794804381e-05, "loss": 0.2313, "step": 16645 }, { "epoch": 1.68, "learning_rate": 3.7238975518420165e-05, "loss": 0.818, "step": 16646 }, { "epoch": 1.68, "learning_rate": 3.723665304273043e-05, "loss": 0.8678, "step": 16647 }, { "epoch": 1.68, "learning_rate": 3.723433052098938e-05, "loss": 1.1351, "step": 16648 }, { "epoch": 1.68, "learning_rate": 3.723200795321182e-05, "loss": 0.9446, "step": 16649 }, { "epoch": 1.68, "learning_rate": 3.722968533941249e-05, "loss": 0.5657, "step": 16650 }, { "epoch": 1.68, "learning_rate": 3.722736267960621e-05, "loss": 0.2381, "step": 16651 }, { "epoch": 1.68, "learning_rate": 3.722503997380774e-05, "loss": 0.4458, "step": 16652 }, { "epoch": 1.68, "learning_rate": 3.7222717222031864e-05, "loss": 0.4179, "step": 16653 }, { "epoch": 1.68, "learning_rate": 3.722039442429336e-05, "loss": 0.7253, "step": 16654 }, { "epoch": 1.68, "learning_rate": 3.7218071580607024e-05, "loss": 0.3453, "step": 16655 }, { "epoch": 1.68, "learning_rate": 3.721574869098761e-05, "loss": 0.9663, "step": 16656 }, { "epoch": 1.68, "learning_rate": 3.721342575544992e-05, "loss": 0.9852, "step": 16657 }, { "epoch": 1.68, "learning_rate": 3.7211102774008745e-05, "loss": 0.6496, "step": 16658 }, { "epoch": 1.68, "learning_rate": 3.720877974667884e-05, "loss": 0.4013, "step": 16659 }, { "epoch": 1.68, "learning_rate": 3.720645667347501e-05, "loss": 0.6304, "step": 16660 }, { "epoch": 1.68, "learning_rate": 3.720413355441203e-05, "loss": 0.292, "step": 16661 }, { "epoch": 1.68, "learning_rate": 3.7201810389504686e-05, "loss": 1.7129, "step": 16662 }, { "epoch": 1.68, "learning_rate": 3.719948717876776e-05, "loss": 0.3247, "step": 16663 }, { "epoch": 1.68, "learning_rate": 3.719716392221603e-05, "loss": 0.536, "step": 16664 }, { "epoch": 1.68, "learning_rate": 3.7194840619864296e-05, "loss": 1.1408, "step": 16665 }, { "epoch": 1.68, "learning_rate": 3.7192517271727336e-05, "loss": 0.6096, "step": 16666 }, { "epoch": 1.68, "learning_rate": 3.719019387781992e-05, "loss": 0.5641, "step": 16667 }, { "epoch": 1.68, "learning_rate": 3.718787043815686e-05, "loss": 0.3991, "step": 16668 }, { "epoch": 1.68, "learning_rate": 3.7185546952752916e-05, "loss": 0.107, "step": 16669 }, { "epoch": 1.68, "learning_rate": 3.7183223421622884e-05, "loss": 1.3426, "step": 16670 }, { "epoch": 1.68, "learning_rate": 3.718089984478156e-05, "loss": 0.3219, "step": 16671 }, { "epoch": 1.68, "learning_rate": 3.7178576222243716e-05, "loss": 0.1778, "step": 16672 }, { "epoch": 1.68, "learning_rate": 3.717625255402414e-05, "loss": 1.1782, "step": 16673 }, { "epoch": 1.68, "learning_rate": 3.717392884013763e-05, "loss": 0.7623, "step": 16674 }, { "epoch": 1.68, "learning_rate": 3.7171605080598955e-05, "loss": 0.7518, "step": 16675 }, { "epoch": 1.68, "learning_rate": 3.7169281275422925e-05, "loss": 0.3416, "step": 16676 }, { "epoch": 1.68, "learning_rate": 3.716695742462431e-05, "loss": 0.4983, "step": 16677 }, { "epoch": 1.68, "learning_rate": 3.71646335282179e-05, "loss": 0.9508, "step": 16678 }, { "epoch": 1.68, "learning_rate": 3.71623095862185e-05, "loss": 1.5456, "step": 16679 }, { "epoch": 1.68, "learning_rate": 3.715998559864087e-05, "loss": 0.4831, "step": 16680 }, { "epoch": 1.68, "learning_rate": 3.715766156549984e-05, "loss": 0.5186, "step": 16681 }, { "epoch": 1.68, "learning_rate": 3.715533748681016e-05, "loss": 0.294, "step": 16682 }, { "epoch": 1.68, "learning_rate": 3.715301336258663e-05, "loss": 0.5035, "step": 16683 }, { "epoch": 1.68, "learning_rate": 3.7150689192844045e-05, "loss": 1.0269, "step": 16684 }, { "epoch": 1.68, "learning_rate": 3.714836497759721e-05, "loss": 0.1652, "step": 16685 }, { "epoch": 1.68, "learning_rate": 3.7146040716860886e-05, "loss": 0.3796, "step": 16686 }, { "epoch": 1.68, "learning_rate": 3.714371641064988e-05, "loss": 0.845, "step": 16687 }, { "epoch": 1.68, "learning_rate": 3.714139205897897e-05, "loss": 1.0807, "step": 16688 }, { "epoch": 1.68, "learning_rate": 3.713906766186297e-05, "loss": 0.5535, "step": 16689 }, { "epoch": 1.68, "learning_rate": 3.7136743219316655e-05, "loss": 0.8038, "step": 16690 }, { "epoch": 1.68, "learning_rate": 3.713441873135483e-05, "loss": 0.6022, "step": 16691 }, { "epoch": 1.68, "learning_rate": 3.713209419799227e-05, "loss": 1.0806, "step": 16692 }, { "epoch": 1.68, "learning_rate": 3.712976961924377e-05, "loss": 0.6677, "step": 16693 }, { "epoch": 1.68, "learning_rate": 3.7127444995124136e-05, "loss": 0.7526, "step": 16694 }, { "epoch": 1.68, "learning_rate": 3.7125120325648155e-05, "loss": 0.4269, "step": 16695 }, { "epoch": 1.68, "learning_rate": 3.7122795610830616e-05, "loss": 0.2745, "step": 16696 }, { "epoch": 1.68, "learning_rate": 3.712047085068632e-05, "loss": 0.1534, "step": 16697 }, { "epoch": 1.68, "learning_rate": 3.711814604523005e-05, "loss": 0.2927, "step": 16698 }, { "epoch": 1.68, "learning_rate": 3.71158211944766e-05, "loss": 1.656, "step": 16699 }, { "epoch": 1.68, "learning_rate": 3.711349629844078e-05, "loss": 0.6714, "step": 16700 }, { "epoch": 1.68, "learning_rate": 3.711117135713737e-05, "loss": 1.2452, "step": 16701 }, { "epoch": 1.68, "learning_rate": 3.710884637058118e-05, "loss": 0.3746, "step": 16702 }, { "epoch": 1.68, "learning_rate": 3.7106521338786994e-05, "loss": 0.9986, "step": 16703 }, { "epoch": 1.68, "learning_rate": 3.7104196261769615e-05, "loss": 0.6912, "step": 16704 }, { "epoch": 1.68, "learning_rate": 3.710187113954382e-05, "loss": 0.6235, "step": 16705 }, { "epoch": 1.68, "learning_rate": 3.709954597212442e-05, "loss": 0.4453, "step": 16706 }, { "epoch": 1.68, "learning_rate": 3.709722075952622e-05, "loss": 0.4077, "step": 16707 }, { "epoch": 1.68, "learning_rate": 3.7094895501764004e-05, "loss": 0.4816, "step": 16708 }, { "epoch": 1.68, "learning_rate": 3.709257019885257e-05, "loss": 1.0241, "step": 16709 }, { "epoch": 1.68, "learning_rate": 3.7090244850806716e-05, "loss": 0.3859, "step": 16710 }, { "epoch": 1.68, "learning_rate": 3.708791945764124e-05, "loss": 0.2993, "step": 16711 }, { "epoch": 1.68, "learning_rate": 3.7085594019370945e-05, "loss": 1.0429, "step": 16712 }, { "epoch": 1.68, "learning_rate": 3.7083268536010625e-05, "loss": 0.677, "step": 16713 }, { "epoch": 1.68, "learning_rate": 3.7080943007575074e-05, "loss": 0.7507, "step": 16714 }, { "epoch": 1.68, "learning_rate": 3.70786174340791e-05, "loss": 0.4792, "step": 16715 }, { "epoch": 1.68, "learning_rate": 3.7076291815537496e-05, "loss": 0.6222, "step": 16716 }, { "epoch": 1.68, "learning_rate": 3.707396615196507e-05, "loss": 0.1968, "step": 16717 }, { "epoch": 1.68, "learning_rate": 3.70716404433766e-05, "loss": 1.4239, "step": 16718 }, { "epoch": 1.68, "learning_rate": 3.7069314689786904e-05, "loss": 0.8307, "step": 16719 }, { "epoch": 1.68, "learning_rate": 3.7066988891210794e-05, "loss": 1.1139, "step": 16720 }, { "epoch": 1.68, "learning_rate": 3.7064663047663035e-05, "loss": 0.3052, "step": 16721 }, { "epoch": 1.69, "learning_rate": 3.7062337159158453e-05, "loss": 0.5377, "step": 16722 }, { "epoch": 1.69, "learning_rate": 3.706001122571185e-05, "loss": 0.595, "step": 16723 }, { "epoch": 1.69, "learning_rate": 3.705768524733802e-05, "loss": 0.8007, "step": 16724 }, { "epoch": 1.69, "learning_rate": 3.705535922405177e-05, "loss": 1.0926, "step": 16725 }, { "epoch": 1.69, "learning_rate": 3.705303315586789e-05, "loss": 0.4541, "step": 16726 }, { "epoch": 1.69, "learning_rate": 3.70507070428012e-05, "loss": 0.9423, "step": 16727 }, { "epoch": 1.69, "learning_rate": 3.704838088486647e-05, "loss": 0.6061, "step": 16728 }, { "epoch": 1.69, "learning_rate": 3.7046054682078544e-05, "loss": 0.5655, "step": 16729 }, { "epoch": 1.69, "learning_rate": 3.70437284344522e-05, "loss": 0.0656, "step": 16730 }, { "epoch": 1.69, "learning_rate": 3.7041402142002254e-05, "loss": 0.8605, "step": 16731 }, { "epoch": 1.69, "learning_rate": 3.703907580474351e-05, "loss": 1.1887, "step": 16732 }, { "epoch": 1.69, "learning_rate": 3.703674942269075e-05, "loss": 0.9599, "step": 16733 }, { "epoch": 1.69, "learning_rate": 3.70344229958588e-05, "loss": 0.712, "step": 16734 }, { "epoch": 1.69, "learning_rate": 3.7032096524262454e-05, "loss": 0.3167, "step": 16735 }, { "epoch": 1.69, "learning_rate": 3.7029770007916526e-05, "loss": 0.868, "step": 16736 }, { "epoch": 1.69, "learning_rate": 3.7027443446835814e-05, "loss": 0.3952, "step": 16737 }, { "epoch": 1.69, "learning_rate": 3.702511684103513e-05, "loss": 0.4938, "step": 16738 }, { "epoch": 1.69, "learning_rate": 3.702279019052927e-05, "loss": 0.818, "step": 16739 }, { "epoch": 1.69, "learning_rate": 3.7020463495333044e-05, "loss": 1.166, "step": 16740 }, { "epoch": 1.69, "learning_rate": 3.701813675546126e-05, "loss": 0.1591, "step": 16741 }, { "epoch": 1.69, "learning_rate": 3.7015809970928724e-05, "loss": 0.3638, "step": 16742 }, { "epoch": 1.69, "learning_rate": 3.701348314175024e-05, "loss": 0.7098, "step": 16743 }, { "epoch": 1.69, "learning_rate": 3.701115626794061e-05, "loss": 0.3084, "step": 16744 }, { "epoch": 1.69, "learning_rate": 3.7008829349514664e-05, "loss": 0.5676, "step": 16745 }, { "epoch": 1.69, "learning_rate": 3.700650238648719e-05, "loss": 0.0587, "step": 16746 }, { "epoch": 1.69, "learning_rate": 3.7004175378873e-05, "loss": 0.8643, "step": 16747 }, { "epoch": 1.69, "learning_rate": 3.7001848326686906e-05, "loss": 0.2949, "step": 16748 }, { "epoch": 1.69, "learning_rate": 3.699952122994371e-05, "loss": 0.1615, "step": 16749 }, { "epoch": 1.69, "learning_rate": 3.699719408865822e-05, "loss": 0.3303, "step": 16750 }, { "epoch": 1.69, "learning_rate": 3.699486690284524e-05, "loss": 0.5659, "step": 16751 }, { "epoch": 1.69, "learning_rate": 3.699253967251961e-05, "loss": 0.5716, "step": 16752 }, { "epoch": 1.69, "learning_rate": 3.69902123976961e-05, "loss": 0.4621, "step": 16753 }, { "epoch": 1.69, "learning_rate": 3.698788507838954e-05, "loss": 0.7455, "step": 16754 }, { "epoch": 1.69, "learning_rate": 3.698555771461475e-05, "loss": 0.1342, "step": 16755 }, { "epoch": 1.69, "learning_rate": 3.698323030638652e-05, "loss": 0.8417, "step": 16756 }, { "epoch": 1.69, "learning_rate": 3.6980902853719665e-05, "loss": 0.3339, "step": 16757 }, { "epoch": 1.69, "learning_rate": 3.6978575356628996e-05, "loss": 0.5761, "step": 16758 }, { "epoch": 1.69, "learning_rate": 3.6976247815129345e-05, "loss": 0.8826, "step": 16759 }, { "epoch": 1.69, "learning_rate": 3.697392022923549e-05, "loss": 0.7391, "step": 16760 }, { "epoch": 1.69, "learning_rate": 3.6971592598962274e-05, "loss": 0.4995, "step": 16761 }, { "epoch": 1.69, "learning_rate": 3.696926492432449e-05, "loss": 0.562, "step": 16762 }, { "epoch": 1.69, "learning_rate": 3.696693720533695e-05, "loss": 0.8167, "step": 16763 }, { "epoch": 1.69, "learning_rate": 3.6964609442014474e-05, "loss": 0.1849, "step": 16764 }, { "epoch": 1.69, "learning_rate": 3.696228163437188e-05, "loss": 0.3447, "step": 16765 }, { "epoch": 1.69, "learning_rate": 3.695995378242396e-05, "loss": 0.4326, "step": 16766 }, { "epoch": 1.69, "learning_rate": 3.695762588618555e-05, "loss": 0.2786, "step": 16767 }, { "epoch": 1.69, "learning_rate": 3.6955297945671466e-05, "loss": 0.2599, "step": 16768 }, { "epoch": 1.69, "learning_rate": 3.695296996089649e-05, "loss": 0.1804, "step": 16769 }, { "epoch": 1.69, "learning_rate": 3.695064193187547e-05, "loss": 0.9229, "step": 16770 }, { "epoch": 1.69, "learning_rate": 3.694831385862321e-05, "loss": 0.3184, "step": 16771 }, { "epoch": 1.69, "learning_rate": 3.694598574115453e-05, "loss": 0.6352, "step": 16772 }, { "epoch": 1.69, "learning_rate": 3.694365757948423e-05, "loss": 0.8685, "step": 16773 }, { "epoch": 1.69, "learning_rate": 3.694132937362713e-05, "loss": 0.3975, "step": 16774 }, { "epoch": 1.69, "learning_rate": 3.693900112359806e-05, "loss": 0.4872, "step": 16775 }, { "epoch": 1.69, "learning_rate": 3.693667282941182e-05, "loss": 0.2136, "step": 16776 }, { "epoch": 1.69, "learning_rate": 3.693434449108324e-05, "loss": 0.6008, "step": 16777 }, { "epoch": 1.69, "learning_rate": 3.6932016108627126e-05, "loss": 0.297, "step": 16778 }, { "epoch": 1.69, "learning_rate": 3.6929687682058296e-05, "loss": 0.3925, "step": 16779 }, { "epoch": 1.69, "learning_rate": 3.6927359211391576e-05, "loss": 1.1865, "step": 16780 }, { "epoch": 1.69, "learning_rate": 3.692503069664177e-05, "loss": 0.264, "step": 16781 }, { "epoch": 1.69, "learning_rate": 3.692270213782371e-05, "loss": 0.4642, "step": 16782 }, { "epoch": 1.69, "learning_rate": 3.69203735349522e-05, "loss": 0.7167, "step": 16783 }, { "epoch": 1.69, "learning_rate": 3.691804488804207e-05, "loss": 0.6632, "step": 16784 }, { "epoch": 1.69, "learning_rate": 3.691571619710813e-05, "loss": 0.444, "step": 16785 }, { "epoch": 1.69, "learning_rate": 3.6913387462165214e-05, "loss": 0.8498, "step": 16786 }, { "epoch": 1.69, "learning_rate": 3.6911058683228124e-05, "loss": 0.7877, "step": 16787 }, { "epoch": 1.69, "learning_rate": 3.690872986031168e-05, "loss": 0.2922, "step": 16788 }, { "epoch": 1.69, "learning_rate": 3.6906400993430716e-05, "loss": 0.8407, "step": 16789 }, { "epoch": 1.69, "learning_rate": 3.690407208260003e-05, "loss": 1.2693, "step": 16790 }, { "epoch": 1.69, "learning_rate": 3.6901743127834474e-05, "loss": 0.1358, "step": 16791 }, { "epoch": 1.69, "learning_rate": 3.6899414129148834e-05, "loss": 0.2008, "step": 16792 }, { "epoch": 1.69, "learning_rate": 3.689708508655795e-05, "loss": 0.9523, "step": 16793 }, { "epoch": 1.69, "learning_rate": 3.689475600007665e-05, "loss": 0.4959, "step": 16794 }, { "epoch": 1.69, "learning_rate": 3.689242686971974e-05, "loss": 0.2978, "step": 16795 }, { "epoch": 1.69, "learning_rate": 3.689009769550205e-05, "loss": 0.0933, "step": 16796 }, { "epoch": 1.69, "learning_rate": 3.688776847743839e-05, "loss": 0.4109, "step": 16797 }, { "epoch": 1.69, "learning_rate": 3.6885439215543604e-05, "loss": 1.2707, "step": 16798 }, { "epoch": 1.69, "learning_rate": 3.688310990983249e-05, "loss": 0.7789, "step": 16799 }, { "epoch": 1.69, "learning_rate": 3.68807805603199e-05, "loss": 0.5476, "step": 16800 }, { "epoch": 1.69, "learning_rate": 3.687845116702063e-05, "loss": 0.7185, "step": 16801 }, { "epoch": 1.69, "learning_rate": 3.687612172994951e-05, "loss": 0.3879, "step": 16802 }, { "epoch": 1.69, "learning_rate": 3.687379224912137e-05, "loss": 0.8825, "step": 16803 }, { "epoch": 1.69, "learning_rate": 3.687146272455103e-05, "loss": 1.6675, "step": 16804 }, { "epoch": 1.69, "learning_rate": 3.686913315625332e-05, "loss": 0.645, "step": 16805 }, { "epoch": 1.69, "learning_rate": 3.686680354424305e-05, "loss": 0.3091, "step": 16806 }, { "epoch": 1.69, "learning_rate": 3.6864473888535073e-05, "loss": 0.3156, "step": 16807 }, { "epoch": 1.69, "learning_rate": 3.686214418914417e-05, "loss": 0.351, "step": 16808 }, { "epoch": 1.69, "learning_rate": 3.685981444608521e-05, "loss": 0.4687, "step": 16809 }, { "epoch": 1.69, "learning_rate": 3.6857484659372995e-05, "loss": 0.2702, "step": 16810 }, { "epoch": 1.69, "learning_rate": 3.6855154829022364e-05, "loss": 0.2808, "step": 16811 }, { "epoch": 1.69, "learning_rate": 3.6852824955048126e-05, "loss": 0.3484, "step": 16812 }, { "epoch": 1.69, "learning_rate": 3.685049503746511e-05, "loss": 0.3269, "step": 16813 }, { "epoch": 1.69, "learning_rate": 3.684816507628817e-05, "loss": 0.731, "step": 16814 }, { "epoch": 1.69, "learning_rate": 3.6845835071532096e-05, "loss": 0.1196, "step": 16815 }, { "epoch": 1.69, "learning_rate": 3.684350502321174e-05, "loss": 0.2937, "step": 16816 }, { "epoch": 1.69, "learning_rate": 3.684117493134192e-05, "loss": 0.2524, "step": 16817 }, { "epoch": 1.69, "learning_rate": 3.6838844795937464e-05, "loss": 0.8657, "step": 16818 }, { "epoch": 1.69, "learning_rate": 3.6836514617013206e-05, "loss": 0.4383, "step": 16819 }, { "epoch": 1.69, "learning_rate": 3.6834184394583966e-05, "loss": 0.6471, "step": 16820 }, { "epoch": 1.7, "learning_rate": 3.683185412866458e-05, "loss": 0.2652, "step": 16821 }, { "epoch": 1.7, "learning_rate": 3.682952381926986e-05, "loss": 0.8726, "step": 16822 }, { "epoch": 1.7, "learning_rate": 3.682719346641467e-05, "loss": 0.4134, "step": 16823 }, { "epoch": 1.7, "learning_rate": 3.6824863070113795e-05, "loss": 0.9493, "step": 16824 }, { "epoch": 1.7, "learning_rate": 3.6822532630382095e-05, "loss": 0.364, "step": 16825 }, { "epoch": 1.7, "learning_rate": 3.68202021472344e-05, "loss": 0.467, "step": 16826 }, { "epoch": 1.7, "learning_rate": 3.681787162068553e-05, "loss": 0.8782, "step": 16827 }, { "epoch": 1.7, "learning_rate": 3.681554105075032e-05, "loss": 0.2049, "step": 16828 }, { "epoch": 1.7, "learning_rate": 3.681321043744359e-05, "loss": 1.0524, "step": 16829 }, { "epoch": 1.7, "learning_rate": 3.68108797807802e-05, "loss": 0.6855, "step": 16830 }, { "epoch": 1.7, "learning_rate": 3.680854908077494e-05, "loss": 1.1547, "step": 16831 }, { "epoch": 1.7, "learning_rate": 3.6806218337442676e-05, "loss": 0.4911, "step": 16832 }, { "epoch": 1.7, "learning_rate": 3.680388755079823e-05, "loss": 0.0888, "step": 16833 }, { "epoch": 1.7, "learning_rate": 3.680155672085643e-05, "loss": 0.5077, "step": 16834 }, { "epoch": 1.7, "learning_rate": 3.67992258476321e-05, "loss": 0.9614, "step": 16835 }, { "epoch": 1.7, "learning_rate": 3.679689493114009e-05, "loss": 0.313, "step": 16836 }, { "epoch": 1.7, "learning_rate": 3.679456397139523e-05, "loss": 1.3317, "step": 16837 }, { "epoch": 1.7, "learning_rate": 3.6792232968412346e-05, "loss": 0.196, "step": 16838 }, { "epoch": 1.7, "learning_rate": 3.678990192220628e-05, "loss": 1.1008, "step": 16839 }, { "epoch": 1.7, "learning_rate": 3.6787570832791855e-05, "loss": 0.7196, "step": 16840 }, { "epoch": 1.7, "learning_rate": 3.678523970018392e-05, "loss": 0.2992, "step": 16841 }, { "epoch": 1.7, "learning_rate": 3.6782908524397296e-05, "loss": 0.0565, "step": 16842 }, { "epoch": 1.7, "learning_rate": 3.678057730544682e-05, "loss": 0.5457, "step": 16843 }, { "epoch": 1.7, "learning_rate": 3.677824604334733e-05, "loss": 0.7591, "step": 16844 }, { "epoch": 1.7, "learning_rate": 3.677591473811366e-05, "loss": 0.8424, "step": 16845 }, { "epoch": 1.7, "learning_rate": 3.6773583389760655e-05, "loss": 0.2037, "step": 16846 }, { "epoch": 1.7, "learning_rate": 3.677125199830313e-05, "loss": 0.8988, "step": 16847 }, { "epoch": 1.7, "learning_rate": 3.676892056375593e-05, "loss": 0.3828, "step": 16848 }, { "epoch": 1.7, "learning_rate": 3.676658908613391e-05, "loss": 0.3823, "step": 16849 }, { "epoch": 1.7, "learning_rate": 3.676425756545188e-05, "loss": 1.4837, "step": 16850 }, { "epoch": 1.7, "learning_rate": 3.67619260017247e-05, "loss": 0.4918, "step": 16851 }, { "epoch": 1.7, "learning_rate": 3.675959439496718e-05, "loss": 0.1639, "step": 16852 }, { "epoch": 1.7, "learning_rate": 3.675726274519419e-05, "loss": 0.2705, "step": 16853 }, { "epoch": 1.7, "learning_rate": 3.675493105242053e-05, "loss": 0.5527, "step": 16854 }, { "epoch": 1.7, "learning_rate": 3.675259931666107e-05, "loss": 0.1983, "step": 16855 }, { "epoch": 1.7, "learning_rate": 3.675026753793064e-05, "loss": 0.7609, "step": 16856 }, { "epoch": 1.7, "learning_rate": 3.674793571624407e-05, "loss": 0.4228, "step": 16857 }, { "epoch": 1.7, "learning_rate": 3.6745603851616206e-05, "loss": 0.4235, "step": 16858 }, { "epoch": 1.7, "learning_rate": 3.6743271944061884e-05, "loss": 0.2717, "step": 16859 }, { "epoch": 1.7, "learning_rate": 3.6740939993595945e-05, "loss": 0.456, "step": 16860 }, { "epoch": 1.7, "learning_rate": 3.673860800023322e-05, "loss": 0.6029, "step": 16861 }, { "epoch": 1.7, "learning_rate": 3.673627596398858e-05, "loss": 0.6074, "step": 16862 }, { "epoch": 1.7, "learning_rate": 3.673394388487682e-05, "loss": 0.2195, "step": 16863 }, { "epoch": 1.7, "learning_rate": 3.6731611762912814e-05, "loss": 0.4411, "step": 16864 }, { "epoch": 1.7, "learning_rate": 3.672927959811139e-05, "loss": 0.2027, "step": 16865 }, { "epoch": 1.7, "learning_rate": 3.6726947390487384e-05, "loss": 0.8116, "step": 16866 }, { "epoch": 1.7, "learning_rate": 3.6724615140055656e-05, "loss": 0.6437, "step": 16867 }, { "epoch": 1.7, "learning_rate": 3.6722282846831025e-05, "loss": 0.3859, "step": 16868 }, { "epoch": 1.7, "learning_rate": 3.6719950510828356e-05, "loss": 0.8632, "step": 16869 }, { "epoch": 1.7, "learning_rate": 3.671761813206246e-05, "loss": 0.912, "step": 16870 }, { "epoch": 1.7, "learning_rate": 3.671528571054821e-05, "loss": 0.6552, "step": 16871 }, { "epoch": 1.7, "learning_rate": 3.671295324630043e-05, "loss": 0.4198, "step": 16872 }, { "epoch": 1.7, "learning_rate": 3.671062073933397e-05, "loss": 0.1686, "step": 16873 }, { "epoch": 1.7, "learning_rate": 3.6708288189663677e-05, "loss": 0.0952, "step": 16874 }, { "epoch": 1.7, "learning_rate": 3.6705955597304386e-05, "loss": 0.3236, "step": 16875 }, { "epoch": 1.7, "learning_rate": 3.670362296227095e-05, "loss": 0.6919, "step": 16876 }, { "epoch": 1.7, "learning_rate": 3.6701290284578194e-05, "loss": 0.1959, "step": 16877 }, { "epoch": 1.7, "learning_rate": 3.669895756424099e-05, "loss": 0.6751, "step": 16878 }, { "epoch": 1.7, "learning_rate": 3.669662480127417e-05, "loss": 0.7061, "step": 16879 }, { "epoch": 1.7, "learning_rate": 3.669429199569257e-05, "loss": 0.7946, "step": 16880 }, { "epoch": 1.7, "learning_rate": 3.669195914751105e-05, "loss": 0.7274, "step": 16881 }, { "epoch": 1.7, "learning_rate": 3.668962625674444e-05, "loss": 0.8668, "step": 16882 }, { "epoch": 1.7, "learning_rate": 3.668729332340759e-05, "loss": 0.5714, "step": 16883 }, { "epoch": 1.7, "learning_rate": 3.6684960347515354e-05, "loss": 0.4151, "step": 16884 }, { "epoch": 1.7, "learning_rate": 3.668262732908258e-05, "loss": 1.0092, "step": 16885 }, { "epoch": 1.7, "learning_rate": 3.668029426812409e-05, "loss": 0.6018, "step": 16886 }, { "epoch": 1.7, "learning_rate": 3.6677961164654765e-05, "loss": 0.8659, "step": 16887 }, { "epoch": 1.7, "learning_rate": 3.667562801868943e-05, "loss": 0.5218, "step": 16888 }, { "epoch": 1.7, "learning_rate": 3.667329483024294e-05, "loss": 0.363, "step": 16889 }, { "epoch": 1.7, "learning_rate": 3.6670961599330144e-05, "loss": 1.0796, "step": 16890 }, { "epoch": 1.7, "learning_rate": 3.666862832596588e-05, "loss": 0.5559, "step": 16891 }, { "epoch": 1.7, "learning_rate": 3.666629501016501e-05, "loss": 0.6471, "step": 16892 }, { "epoch": 1.7, "learning_rate": 3.6663961651942354e-05, "loss": 0.9984, "step": 16893 }, { "epoch": 1.7, "learning_rate": 3.6661628251312796e-05, "loss": 0.4153, "step": 16894 }, { "epoch": 1.7, "learning_rate": 3.665929480829117e-05, "loss": 0.2739, "step": 16895 }, { "epoch": 1.7, "learning_rate": 3.665696132289233e-05, "loss": 0.9202, "step": 16896 }, { "epoch": 1.7, "learning_rate": 3.665462779513111e-05, "loss": 0.3121, "step": 16897 }, { "epoch": 1.7, "learning_rate": 3.665229422502238e-05, "loss": 1.6519, "step": 16898 }, { "epoch": 1.7, "learning_rate": 3.6649960612580984e-05, "loss": 0.4016, "step": 16899 }, { "epoch": 1.7, "learning_rate": 3.664762695782175e-05, "loss": 0.2969, "step": 16900 }, { "epoch": 1.7, "learning_rate": 3.664529326075957e-05, "loss": 0.2577, "step": 16901 }, { "epoch": 1.7, "learning_rate": 3.6642959521409246e-05, "loss": 0.4315, "step": 16902 }, { "epoch": 1.7, "learning_rate": 3.664062573978567e-05, "loss": 0.657, "step": 16903 }, { "epoch": 1.7, "learning_rate": 3.663829191590367e-05, "loss": 0.324, "step": 16904 }, { "epoch": 1.7, "learning_rate": 3.663595804977812e-05, "loss": 0.2781, "step": 16905 }, { "epoch": 1.7, "learning_rate": 3.6633624141423846e-05, "loss": 0.475, "step": 16906 }, { "epoch": 1.7, "learning_rate": 3.663129019085571e-05, "loss": 0.7561, "step": 16907 }, { "epoch": 1.7, "learning_rate": 3.662895619808857e-05, "loss": 0.8044, "step": 16908 }, { "epoch": 1.7, "learning_rate": 3.662662216313727e-05, "loss": 0.4115, "step": 16909 }, { "epoch": 1.7, "learning_rate": 3.662428808601667e-05, "loss": 0.62, "step": 16910 }, { "epoch": 1.7, "learning_rate": 3.662195396674163e-05, "loss": 0.6771, "step": 16911 }, { "epoch": 1.7, "learning_rate": 3.6619619805326986e-05, "loss": 0.3955, "step": 16912 }, { "epoch": 1.7, "learning_rate": 3.661728560178761e-05, "loss": 1.0228, "step": 16913 }, { "epoch": 1.7, "learning_rate": 3.661495135613832e-05, "loss": 0.4638, "step": 16914 }, { "epoch": 1.7, "learning_rate": 3.661261706839402e-05, "loss": 0.4552, "step": 16915 }, { "epoch": 1.7, "learning_rate": 3.6610282738569526e-05, "loss": 0.876, "step": 16916 }, { "epoch": 1.7, "learning_rate": 3.660794836667972e-05, "loss": 0.5643, "step": 16917 }, { "epoch": 1.7, "learning_rate": 3.6605613952739437e-05, "loss": 0.7679, "step": 16918 }, { "epoch": 1.71, "learning_rate": 3.660327949676354e-05, "loss": 0.2355, "step": 16919 }, { "epoch": 1.71, "learning_rate": 3.660094499876688e-05, "loss": 1.6229, "step": 16920 }, { "epoch": 1.71, "learning_rate": 3.659861045876433e-05, "loss": 0.5904, "step": 16921 }, { "epoch": 1.71, "learning_rate": 3.659627587677073e-05, "loss": 0.2214, "step": 16922 }, { "epoch": 1.71, "learning_rate": 3.659394125280093e-05, "loss": 1.0617, "step": 16923 }, { "epoch": 1.71, "learning_rate": 3.65916065868698e-05, "loss": 0.7304, "step": 16924 }, { "epoch": 1.71, "learning_rate": 3.658927187899219e-05, "loss": 0.2282, "step": 16925 }, { "epoch": 1.71, "learning_rate": 3.658693712918297e-05, "loss": 0.877, "step": 16926 }, { "epoch": 1.71, "learning_rate": 3.6584602337456985e-05, "loss": 0.917, "step": 16927 }, { "epoch": 1.71, "learning_rate": 3.6582267503829095e-05, "loss": 0.5271, "step": 16928 }, { "epoch": 1.71, "learning_rate": 3.657993262831416e-05, "loss": 0.5823, "step": 16929 }, { "epoch": 1.71, "learning_rate": 3.657759771092704e-05, "loss": 1.0291, "step": 16930 }, { "epoch": 1.71, "learning_rate": 3.657526275168258e-05, "loss": 1.0992, "step": 16931 }, { "epoch": 1.71, "learning_rate": 3.657292775059565e-05, "loss": 0.9573, "step": 16932 }, { "epoch": 1.71, "learning_rate": 3.657059270768111e-05, "loss": 0.7735, "step": 16933 }, { "epoch": 1.71, "learning_rate": 3.656825762295382e-05, "loss": 1.0857, "step": 16934 }, { "epoch": 1.71, "learning_rate": 3.656592249642864e-05, "loss": 0.7822, "step": 16935 }, { "epoch": 1.71, "learning_rate": 3.656358732812042e-05, "loss": 0.6054, "step": 16936 }, { "epoch": 1.71, "learning_rate": 3.656125211804403e-05, "loss": 0.2399, "step": 16937 }, { "epoch": 1.71, "learning_rate": 3.655891686621434e-05, "loss": 0.7558, "step": 16938 }, { "epoch": 1.71, "learning_rate": 3.655658157264618e-05, "loss": 0.2875, "step": 16939 }, { "epoch": 1.71, "learning_rate": 3.6554246237354436e-05, "loss": 0.7543, "step": 16940 }, { "epoch": 1.71, "learning_rate": 3.655191086035396e-05, "loss": 0.6606, "step": 16941 }, { "epoch": 1.71, "learning_rate": 3.654957544165962e-05, "loss": 0.2015, "step": 16942 }, { "epoch": 1.71, "learning_rate": 3.6547239981286275e-05, "loss": 0.7693, "step": 16943 }, { "epoch": 1.71, "learning_rate": 3.654490447924878e-05, "loss": 1.083, "step": 16944 }, { "epoch": 1.71, "learning_rate": 3.654256893556201e-05, "loss": 1.0651, "step": 16945 }, { "epoch": 1.71, "learning_rate": 3.65402333502408e-05, "loss": 0.5134, "step": 16946 }, { "epoch": 1.71, "learning_rate": 3.653789772330006e-05, "loss": 0.7046, "step": 16947 }, { "epoch": 1.71, "learning_rate": 3.6535562054754614e-05, "loss": 0.4092, "step": 16948 }, { "epoch": 1.71, "learning_rate": 3.6533226344619335e-05, "loss": 0.6465, "step": 16949 }, { "epoch": 1.71, "learning_rate": 3.6530890592909096e-05, "loss": 0.3325, "step": 16950 }, { "epoch": 1.71, "learning_rate": 3.6528554799638745e-05, "loss": 0.4113, "step": 16951 }, { "epoch": 1.71, "learning_rate": 3.6526218964823164e-05, "loss": 0.4088, "step": 16952 }, { "epoch": 1.71, "learning_rate": 3.65238830884772e-05, "loss": 0.8745, "step": 16953 }, { "epoch": 1.71, "learning_rate": 3.6521547170615734e-05, "loss": 0.2748, "step": 16954 }, { "epoch": 1.71, "learning_rate": 3.6519211211253606e-05, "loss": 0.5335, "step": 16955 }, { "epoch": 1.71, "learning_rate": 3.6516875210405716e-05, "loss": 0.3411, "step": 16956 }, { "epoch": 1.71, "learning_rate": 3.65145391680869e-05, "loss": 0.4385, "step": 16957 }, { "epoch": 1.71, "learning_rate": 3.651220308431204e-05, "loss": 0.2551, "step": 16958 }, { "epoch": 1.71, "learning_rate": 3.6509866959096004e-05, "loss": 1.0356, "step": 16959 }, { "epoch": 1.71, "learning_rate": 3.650753079245364e-05, "loss": 0.8746, "step": 16960 }, { "epoch": 1.71, "learning_rate": 3.650519458439983e-05, "loss": 0.2697, "step": 16961 }, { "epoch": 1.71, "learning_rate": 3.650285833494943e-05, "loss": 1.072, "step": 16962 }, { "epoch": 1.71, "learning_rate": 3.6500522044117326e-05, "loss": 1.0035, "step": 16963 }, { "epoch": 1.71, "learning_rate": 3.649818571191836e-05, "loss": 0.2592, "step": 16964 }, { "epoch": 1.71, "learning_rate": 3.649584933836742e-05, "loss": 0.487, "step": 16965 }, { "epoch": 1.71, "learning_rate": 3.649351292347937e-05, "loss": 1.1967, "step": 16966 }, { "epoch": 1.71, "learning_rate": 3.649117646726907e-05, "loss": 0.8722, "step": 16967 }, { "epoch": 1.71, "learning_rate": 3.648883996975139e-05, "loss": 0.9673, "step": 16968 }, { "epoch": 1.71, "learning_rate": 3.64865034309412e-05, "loss": 0.3401, "step": 16969 }, { "epoch": 1.71, "learning_rate": 3.648416685085337e-05, "loss": 0.2153, "step": 16970 }, { "epoch": 1.71, "learning_rate": 3.6481830229502766e-05, "loss": 0.9498, "step": 16971 }, { "epoch": 1.71, "learning_rate": 3.6479493566904274e-05, "loss": 0.3041, "step": 16972 }, { "epoch": 1.71, "learning_rate": 3.647715686307273e-05, "loss": 1.1607, "step": 16973 }, { "epoch": 1.71, "learning_rate": 3.647482011802304e-05, "loss": 0.894, "step": 16974 }, { "epoch": 1.71, "learning_rate": 3.6472483331770046e-05, "loss": 0.7741, "step": 16975 }, { "epoch": 1.71, "learning_rate": 3.647014650432864e-05, "loss": 0.4458, "step": 16976 }, { "epoch": 1.71, "learning_rate": 3.646780963571368e-05, "loss": 0.3944, "step": 16977 }, { "epoch": 1.71, "learning_rate": 3.6465472725940035e-05, "loss": 0.5933, "step": 16978 }, { "epoch": 1.71, "learning_rate": 3.646313577502259e-05, "loss": 1.1043, "step": 16979 }, { "epoch": 1.71, "learning_rate": 3.64607987829762e-05, "loss": 0.7011, "step": 16980 }, { "epoch": 1.71, "learning_rate": 3.645846174981575e-05, "loss": 0.9295, "step": 16981 }, { "epoch": 1.71, "learning_rate": 3.6456124675556105e-05, "loss": 0.4908, "step": 16982 }, { "epoch": 1.71, "learning_rate": 3.645378756021213e-05, "loss": 1.0326, "step": 16983 }, { "epoch": 1.71, "learning_rate": 3.6451450403798724e-05, "loss": 1.0538, "step": 16984 }, { "epoch": 1.71, "learning_rate": 3.6449113206330726e-05, "loss": 0.5331, "step": 16985 }, { "epoch": 1.71, "learning_rate": 3.6446775967823033e-05, "loss": 0.6614, "step": 16986 }, { "epoch": 1.71, "learning_rate": 3.6444438688290506e-05, "loss": 0.9539, "step": 16987 }, { "epoch": 1.71, "learning_rate": 3.644210136774802e-05, "loss": 0.095, "step": 16988 }, { "epoch": 1.71, "learning_rate": 3.643976400621046e-05, "loss": 0.8502, "step": 16989 }, { "epoch": 1.71, "learning_rate": 3.643742660369268e-05, "loss": 0.9377, "step": 16990 }, { "epoch": 1.71, "learning_rate": 3.643508916020958e-05, "loss": 0.4513, "step": 16991 }, { "epoch": 1.71, "learning_rate": 3.643275167577601e-05, "loss": 0.9351, "step": 16992 }, { "epoch": 1.71, "learning_rate": 3.643041415040686e-05, "loss": 0.9443, "step": 16993 }, { "epoch": 1.71, "learning_rate": 3.6428076584117e-05, "loss": 0.5311, "step": 16994 }, { "epoch": 1.71, "learning_rate": 3.642573897692131e-05, "loss": 1.4576, "step": 16995 }, { "epoch": 1.71, "learning_rate": 3.6423401328834653e-05, "loss": 0.9091, "step": 16996 }, { "epoch": 1.71, "learning_rate": 3.642106363987192e-05, "loss": 0.4347, "step": 16997 }, { "epoch": 1.71, "learning_rate": 3.6418725910047976e-05, "loss": 1.2993, "step": 16998 }, { "epoch": 1.71, "learning_rate": 3.641638813937771e-05, "loss": 0.6476, "step": 16999 }, { "epoch": 1.71, "learning_rate": 3.641405032787599e-05, "loss": 0.938, "step": 17000 }, { "epoch": 1.71, "learning_rate": 3.641171247555768e-05, "loss": 0.3415, "step": 17001 }, { "epoch": 1.71, "learning_rate": 3.640937458243769e-05, "loss": 0.6518, "step": 17002 }, { "epoch": 1.71, "learning_rate": 3.640703664853086e-05, "loss": 1.165, "step": 17003 }, { "epoch": 1.71, "learning_rate": 3.6404698673852096e-05, "loss": 0.2838, "step": 17004 }, { "epoch": 1.71, "learning_rate": 3.640236065841627e-05, "loss": 0.4091, "step": 17005 }, { "epoch": 1.71, "learning_rate": 3.640002260223825e-05, "loss": 1.0445, "step": 17006 }, { "epoch": 1.71, "learning_rate": 3.6397684505332926e-05, "loss": 0.6304, "step": 17007 }, { "epoch": 1.71, "learning_rate": 3.639534636771516e-05, "loss": 0.6868, "step": 17008 }, { "epoch": 1.71, "learning_rate": 3.6393008189399864e-05, "loss": 0.6894, "step": 17009 }, { "epoch": 1.71, "learning_rate": 3.639066997040187e-05, "loss": 0.1951, "step": 17010 }, { "epoch": 1.71, "learning_rate": 3.63883317107361e-05, "loss": 0.2165, "step": 17011 }, { "epoch": 1.71, "learning_rate": 3.638599341041741e-05, "loss": 0.9476, "step": 17012 }, { "epoch": 1.71, "learning_rate": 3.638365506946069e-05, "loss": 0.9171, "step": 17013 }, { "epoch": 1.71, "learning_rate": 3.6381316687880824e-05, "loss": 1.0969, "step": 17014 }, { "epoch": 1.71, "learning_rate": 3.637897826569268e-05, "loss": 0.9065, "step": 17015 }, { "epoch": 1.71, "learning_rate": 3.637663980291114e-05, "loss": 0.2744, "step": 17016 }, { "epoch": 1.71, "learning_rate": 3.637430129955109e-05, "loss": 0.7213, "step": 17017 }, { "epoch": 1.72, "learning_rate": 3.637196275562743e-05, "loss": 0.2717, "step": 17018 }, { "epoch": 1.72, "learning_rate": 3.6369624171155e-05, "loss": 0.2859, "step": 17019 }, { "epoch": 1.72, "learning_rate": 3.636728554614872e-05, "loss": 0.4472, "step": 17020 }, { "epoch": 1.72, "learning_rate": 3.636494688062345e-05, "loss": 0.6262, "step": 17021 }, { "epoch": 1.72, "learning_rate": 3.636260817459409e-05, "loss": 0.4481, "step": 17022 }, { "epoch": 1.72, "learning_rate": 3.6360269428075496e-05, "loss": 0.7821, "step": 17023 }, { "epoch": 1.72, "learning_rate": 3.635793064108258e-05, "loss": 0.374, "step": 17024 }, { "epoch": 1.72, "learning_rate": 3.63555918136302e-05, "loss": 0.4693, "step": 17025 }, { "epoch": 1.72, "learning_rate": 3.635325294573325e-05, "loss": 1.3603, "step": 17026 }, { "epoch": 1.72, "learning_rate": 3.635091403740663e-05, "loss": 0.9374, "step": 17027 }, { "epoch": 1.72, "learning_rate": 3.6348575088665204e-05, "loss": 0.4551, "step": 17028 }, { "epoch": 1.72, "learning_rate": 3.634623609952386e-05, "loss": 0.6478, "step": 17029 }, { "epoch": 1.72, "learning_rate": 3.634389706999748e-05, "loss": 0.4259, "step": 17030 }, { "epoch": 1.72, "learning_rate": 3.634155800010096e-05, "loss": 1.1711, "step": 17031 }, { "epoch": 1.72, "learning_rate": 3.633921888984917e-05, "loss": 0.0984, "step": 17032 }, { "epoch": 1.72, "learning_rate": 3.6336879739257e-05, "loss": 0.2844, "step": 17033 }, { "epoch": 1.72, "learning_rate": 3.6334540548339354e-05, "loss": 0.8158, "step": 17034 }, { "epoch": 1.72, "learning_rate": 3.6332201317111084e-05, "loss": 0.8659, "step": 17035 }, { "epoch": 1.72, "learning_rate": 3.6329862045587105e-05, "loss": 1.2557, "step": 17036 }, { "epoch": 1.72, "learning_rate": 3.6327522733782285e-05, "loss": 0.1998, "step": 17037 }, { "epoch": 1.72, "learning_rate": 3.632518338171152e-05, "loss": 0.8153, "step": 17038 }, { "epoch": 1.72, "learning_rate": 3.632284398938969e-05, "loss": 0.3074, "step": 17039 }, { "epoch": 1.72, "learning_rate": 3.632050455683169e-05, "loss": 0.5068, "step": 17040 }, { "epoch": 1.72, "learning_rate": 3.631816508405241e-05, "loss": 0.2776, "step": 17041 }, { "epoch": 1.72, "learning_rate": 3.6315825571066724e-05, "loss": 0.4855, "step": 17042 }, { "epoch": 1.72, "learning_rate": 3.6313486017889524e-05, "loss": 0.4964, "step": 17043 }, { "epoch": 1.72, "learning_rate": 3.631114642453571e-05, "loss": 0.6616, "step": 17044 }, { "epoch": 1.72, "learning_rate": 3.630880679102015e-05, "loss": 0.5961, "step": 17045 }, { "epoch": 1.72, "learning_rate": 3.6306467117357755e-05, "loss": 0.246, "step": 17046 }, { "epoch": 1.72, "learning_rate": 3.630412740356339e-05, "loss": 1.1034, "step": 17047 }, { "epoch": 1.72, "learning_rate": 3.6301787649651966e-05, "loss": 0.6657, "step": 17048 }, { "epoch": 1.72, "learning_rate": 3.629944785563836e-05, "loss": 0.8309, "step": 17049 }, { "epoch": 1.72, "learning_rate": 3.629710802153747e-05, "loss": 0.3481, "step": 17050 }, { "epoch": 1.72, "learning_rate": 3.6294768147364166e-05, "loss": 0.9388, "step": 17051 }, { "epoch": 1.72, "learning_rate": 3.629242823313336e-05, "loss": 0.2841, "step": 17052 }, { "epoch": 1.72, "learning_rate": 3.629008827885994e-05, "loss": 0.581, "step": 17053 }, { "epoch": 1.72, "learning_rate": 3.628774828455878e-05, "loss": 0.695, "step": 17054 }, { "epoch": 1.72, "learning_rate": 3.628540825024479e-05, "loss": 0.621, "step": 17055 }, { "epoch": 1.72, "learning_rate": 3.628306817593284e-05, "loss": 0.2096, "step": 17056 }, { "epoch": 1.72, "learning_rate": 3.628072806163785e-05, "loss": 0.6595, "step": 17057 }, { "epoch": 1.72, "learning_rate": 3.627838790737469e-05, "loss": 0.8098, "step": 17058 }, { "epoch": 1.72, "learning_rate": 3.627604771315826e-05, "loss": 0.3193, "step": 17059 }, { "epoch": 1.72, "learning_rate": 3.627370747900345e-05, "loss": 0.7151, "step": 17060 }, { "epoch": 1.72, "learning_rate": 3.627136720492515e-05, "loss": 0.264, "step": 17061 }, { "epoch": 1.72, "learning_rate": 3.626902689093826e-05, "loss": 0.1613, "step": 17062 }, { "epoch": 1.72, "learning_rate": 3.626668653705766e-05, "loss": 0.5753, "step": 17063 }, { "epoch": 1.72, "learning_rate": 3.626434614329825e-05, "loss": 1.0735, "step": 17064 }, { "epoch": 1.72, "learning_rate": 3.626200570967493e-05, "loss": 0.5211, "step": 17065 }, { "epoch": 1.72, "learning_rate": 3.625966523620259e-05, "loss": 0.4588, "step": 17066 }, { "epoch": 1.72, "learning_rate": 3.625732472289612e-05, "loss": 0.4164, "step": 17067 }, { "epoch": 1.72, "learning_rate": 3.625498416977041e-05, "loss": 0.4082, "step": 17068 }, { "epoch": 1.72, "learning_rate": 3.625264357684036e-05, "loss": 1.0726, "step": 17069 }, { "epoch": 1.72, "learning_rate": 3.625030294412087e-05, "loss": 0.3524, "step": 17070 }, { "epoch": 1.72, "learning_rate": 3.6247962271626836e-05, "loss": 0.1813, "step": 17071 }, { "epoch": 1.72, "learning_rate": 3.624562155937313e-05, "loss": 1.1973, "step": 17072 }, { "epoch": 1.72, "learning_rate": 3.624328080737468e-05, "loss": 0.6576, "step": 17073 }, { "epoch": 1.72, "learning_rate": 3.6240940015646355e-05, "loss": 0.3869, "step": 17074 }, { "epoch": 1.72, "learning_rate": 3.623859918420307e-05, "loss": 1.0096, "step": 17075 }, { "epoch": 1.72, "learning_rate": 3.623625831305971e-05, "loss": 0.5273, "step": 17076 }, { "epoch": 1.72, "learning_rate": 3.623391740223118e-05, "loss": 0.6505, "step": 17077 }, { "epoch": 1.72, "learning_rate": 3.623157645173236e-05, "loss": 0.6532, "step": 17078 }, { "epoch": 1.72, "learning_rate": 3.622923546157816e-05, "loss": 0.7412, "step": 17079 }, { "epoch": 1.72, "learning_rate": 3.622689443178349e-05, "loss": 1.0991, "step": 17080 }, { "epoch": 1.72, "learning_rate": 3.622455336236321e-05, "loss": 0.6067, "step": 17081 }, { "epoch": 1.72, "learning_rate": 3.6222212253332253e-05, "loss": 0.4531, "step": 17082 }, { "epoch": 1.72, "learning_rate": 3.6219871104705506e-05, "loss": 0.926, "step": 17083 }, { "epoch": 1.72, "learning_rate": 3.6217529916497865e-05, "loss": 1.0531, "step": 17084 }, { "epoch": 1.72, "learning_rate": 3.621518868872422e-05, "loss": 1.2086, "step": 17085 }, { "epoch": 1.72, "learning_rate": 3.621284742139949e-05, "loss": 0.4231, "step": 17086 }, { "epoch": 1.72, "learning_rate": 3.621050611453856e-05, "loss": 1.1561, "step": 17087 }, { "epoch": 1.72, "learning_rate": 3.620816476815632e-05, "loss": 0.6528, "step": 17088 }, { "epoch": 1.72, "learning_rate": 3.62058233822677e-05, "loss": 0.8136, "step": 17089 }, { "epoch": 1.72, "learning_rate": 3.6203481956887565e-05, "loss": 0.9744, "step": 17090 }, { "epoch": 1.72, "learning_rate": 3.620114049203084e-05, "loss": 0.2997, "step": 17091 }, { "epoch": 1.72, "learning_rate": 3.619879898771241e-05, "loss": 0.9862, "step": 17092 }, { "epoch": 1.72, "learning_rate": 3.619645744394719e-05, "loss": 0.1884, "step": 17093 }, { "epoch": 1.72, "learning_rate": 3.6194115860750064e-05, "loss": 0.9474, "step": 17094 }, { "epoch": 1.72, "learning_rate": 3.619177423813594e-05, "loss": 1.0553, "step": 17095 }, { "epoch": 1.72, "learning_rate": 3.618943257611973e-05, "loss": 0.5046, "step": 17096 }, { "epoch": 1.72, "learning_rate": 3.618709087471631e-05, "loss": 1.2213, "step": 17097 }, { "epoch": 1.72, "learning_rate": 3.6184749133940615e-05, "loss": 0.6627, "step": 17098 }, { "epoch": 1.72, "learning_rate": 3.618240735380752e-05, "loss": 0.6921, "step": 17099 }, { "epoch": 1.72, "learning_rate": 3.618006553433193e-05, "loss": 0.1586, "step": 17100 }, { "epoch": 1.72, "learning_rate": 3.617772367552876e-05, "loss": 0.3367, "step": 17101 }, { "epoch": 1.72, "learning_rate": 3.617538177741292e-05, "loss": 0.1447, "step": 17102 }, { "epoch": 1.72, "learning_rate": 3.617303983999929e-05, "loss": 1.0696, "step": 17103 }, { "epoch": 1.72, "learning_rate": 3.617069786330277e-05, "loss": 0.2291, "step": 17104 }, { "epoch": 1.72, "learning_rate": 3.6168355847338294e-05, "loss": 0.5253, "step": 17105 }, { "epoch": 1.72, "learning_rate": 3.616601379212073e-05, "loss": 0.7044, "step": 17106 }, { "epoch": 1.72, "learning_rate": 3.616367169766502e-05, "loss": 1.1673, "step": 17107 }, { "epoch": 1.72, "learning_rate": 3.616132956398603e-05, "loss": 0.0002, "step": 17108 }, { "epoch": 1.72, "learning_rate": 3.61589873910987e-05, "loss": 0.5392, "step": 17109 }, { "epoch": 1.72, "learning_rate": 3.615664517901791e-05, "loss": 0.2747, "step": 17110 }, { "epoch": 1.72, "learning_rate": 3.615430292775856e-05, "loss": 1.1089, "step": 17111 }, { "epoch": 1.72, "learning_rate": 3.615196063733559e-05, "loss": 0.5928, "step": 17112 }, { "epoch": 1.72, "learning_rate": 3.6149618307763864e-05, "loss": 0.5178, "step": 17113 }, { "epoch": 1.72, "learning_rate": 3.614727593905831e-05, "loss": 0.7061, "step": 17114 }, { "epoch": 1.72, "learning_rate": 3.614493353123384e-05, "loss": 0.6828, "step": 17115 }, { "epoch": 1.73, "learning_rate": 3.614259108430534e-05, "loss": 0.5046, "step": 17116 }, { "epoch": 1.73, "learning_rate": 3.6140248598287724e-05, "loss": 0.1799, "step": 17117 }, { "epoch": 1.73, "learning_rate": 3.613790607319592e-05, "loss": 0.8325, "step": 17118 }, { "epoch": 1.73, "learning_rate": 3.613556350904481e-05, "loss": 0.359, "step": 17119 }, { "epoch": 1.73, "learning_rate": 3.61332209058493e-05, "loss": 0.7681, "step": 17120 }, { "epoch": 1.73, "learning_rate": 3.613087826362431e-05, "loss": 0.5863, "step": 17121 }, { "epoch": 1.73, "learning_rate": 3.612853558238474e-05, "loss": 0.6436, "step": 17122 }, { "epoch": 1.73, "learning_rate": 3.6126192862145504e-05, "loss": 0.4252, "step": 17123 }, { "epoch": 1.73, "learning_rate": 3.61238501029215e-05, "loss": 0.6523, "step": 17124 }, { "epoch": 1.73, "learning_rate": 3.612150730472766e-05, "loss": 0.8689, "step": 17125 }, { "epoch": 1.73, "learning_rate": 3.611916446757886e-05, "loss": 0.8283, "step": 17126 }, { "epoch": 1.73, "learning_rate": 3.6116821591490036e-05, "loss": 0.3483, "step": 17127 }, { "epoch": 1.73, "learning_rate": 3.611447867647608e-05, "loss": 0.8897, "step": 17128 }, { "epoch": 1.73, "learning_rate": 3.611213572255191e-05, "loss": 0.1115, "step": 17129 }, { "epoch": 1.73, "learning_rate": 3.610979272973243e-05, "loss": 0.4373, "step": 17130 }, { "epoch": 1.73, "learning_rate": 3.610744969803256e-05, "loss": 0.9374, "step": 17131 }, { "epoch": 1.73, "learning_rate": 3.6105106627467205e-05, "loss": 0.3332, "step": 17132 }, { "epoch": 1.73, "learning_rate": 3.610276351805126e-05, "loss": 0.6424, "step": 17133 }, { "epoch": 1.73, "learning_rate": 3.610042036979966e-05, "loss": 0.3639, "step": 17134 }, { "epoch": 1.73, "learning_rate": 3.609807718272731e-05, "loss": 0.5284, "step": 17135 }, { "epoch": 1.73, "learning_rate": 3.6095733956849106e-05, "loss": 0.8757, "step": 17136 }, { "epoch": 1.73, "learning_rate": 3.609339069217997e-05, "loss": 0.3994, "step": 17137 }, { "epoch": 1.73, "learning_rate": 3.6091047388734827e-05, "loss": 1.1011, "step": 17138 }, { "epoch": 1.73, "learning_rate": 3.6088704046528567e-05, "loss": 0.803, "step": 17139 }, { "epoch": 1.73, "learning_rate": 3.60863606655761e-05, "loss": 0.9535, "step": 17140 }, { "epoch": 1.73, "learning_rate": 3.608401724589237e-05, "loss": 0.7158, "step": 17141 }, { "epoch": 1.73, "learning_rate": 3.608167378749226e-05, "loss": 1.1419, "step": 17142 }, { "epoch": 1.73, "learning_rate": 3.6079330290390684e-05, "loss": 0.475, "step": 17143 }, { "epoch": 1.73, "learning_rate": 3.607698675460257e-05, "loss": 0.8299, "step": 17144 }, { "epoch": 1.73, "learning_rate": 3.6074643180142826e-05, "loss": 0.2702, "step": 17145 }, { "epoch": 1.73, "learning_rate": 3.607229956702636e-05, "loss": 1.3484, "step": 17146 }, { "epoch": 1.73, "learning_rate": 3.606995591526809e-05, "loss": 1.253, "step": 17147 }, { "epoch": 1.73, "learning_rate": 3.606761222488293e-05, "loss": 0.3609, "step": 17148 }, { "epoch": 1.73, "learning_rate": 3.6065268495885784e-05, "loss": 1.0469, "step": 17149 }, { "epoch": 1.73, "learning_rate": 3.6062924728291595e-05, "loss": 0.8689, "step": 17150 }, { "epoch": 1.73, "learning_rate": 3.606058092211525e-05, "loss": 0.5728, "step": 17151 }, { "epoch": 1.73, "learning_rate": 3.605823707737168e-05, "loss": 0.5218, "step": 17152 }, { "epoch": 1.73, "learning_rate": 3.605589319407578e-05, "loss": 0.3746, "step": 17153 }, { "epoch": 1.73, "learning_rate": 3.605354927224249e-05, "loss": 0.2264, "step": 17154 }, { "epoch": 1.73, "learning_rate": 3.605120531188671e-05, "loss": 1.0048, "step": 17155 }, { "epoch": 1.73, "learning_rate": 3.604886131302336e-05, "loss": 0.33, "step": 17156 }, { "epoch": 1.73, "learning_rate": 3.6046517275667375e-05, "loss": 0.3207, "step": 17157 }, { "epoch": 1.73, "learning_rate": 3.604417319983364e-05, "loss": 0.6293, "step": 17158 }, { "epoch": 1.73, "learning_rate": 3.6041829085537084e-05, "loss": 0.1062, "step": 17159 }, { "epoch": 1.73, "learning_rate": 3.603948493279264e-05, "loss": 0.804, "step": 17160 }, { "epoch": 1.73, "learning_rate": 3.603714074161519e-05, "loss": 0.3618, "step": 17161 }, { "epoch": 1.73, "learning_rate": 3.603479651201969e-05, "loss": 1.0925, "step": 17162 }, { "epoch": 1.73, "learning_rate": 3.603245224402103e-05, "loss": 0.5483, "step": 17163 }, { "epoch": 1.73, "learning_rate": 3.6030107937634144e-05, "loss": 0.569, "step": 17164 }, { "epoch": 1.73, "learning_rate": 3.602776359287394e-05, "loss": 0.6452, "step": 17165 }, { "epoch": 1.73, "learning_rate": 3.6025419209755344e-05, "loss": 1.0458, "step": 17166 }, { "epoch": 1.73, "learning_rate": 3.6023074788293285e-05, "loss": 1.3649, "step": 17167 }, { "epoch": 1.73, "learning_rate": 3.602073032850266e-05, "loss": 0.3426, "step": 17168 }, { "epoch": 1.73, "learning_rate": 3.601838583039839e-05, "loss": 0.9874, "step": 17169 }, { "epoch": 1.73, "learning_rate": 3.601604129399541e-05, "loss": 0.6549, "step": 17170 }, { "epoch": 1.73, "learning_rate": 3.601369671930863e-05, "loss": 1.297, "step": 17171 }, { "epoch": 1.73, "learning_rate": 3.601135210635297e-05, "loss": 0.9966, "step": 17172 }, { "epoch": 1.73, "learning_rate": 3.600900745514336e-05, "loss": 1.0643, "step": 17173 }, { "epoch": 1.73, "learning_rate": 3.600666276569471e-05, "loss": 0.5576, "step": 17174 }, { "epoch": 1.73, "learning_rate": 3.600431803802194e-05, "loss": 1.0103, "step": 17175 }, { "epoch": 1.73, "learning_rate": 3.600197327213997e-05, "loss": 0.4518, "step": 17176 }, { "epoch": 1.73, "learning_rate": 3.599962846806373e-05, "loss": 0.208, "step": 17177 }, { "epoch": 1.73, "learning_rate": 3.599728362580814e-05, "loss": 0.858, "step": 17178 }, { "epoch": 1.73, "learning_rate": 3.599493874538811e-05, "loss": 0.9107, "step": 17179 }, { "epoch": 1.73, "learning_rate": 3.599259382681858e-05, "loss": 0.4342, "step": 17180 }, { "epoch": 1.73, "learning_rate": 3.599024887011445e-05, "loss": 0.7739, "step": 17181 }, { "epoch": 1.73, "learning_rate": 3.598790387529067e-05, "loss": 0.7197, "step": 17182 }, { "epoch": 1.73, "learning_rate": 3.5985558842362145e-05, "loss": 0.3273, "step": 17183 }, { "epoch": 1.73, "learning_rate": 3.598321377134379e-05, "loss": 0.1432, "step": 17184 }, { "epoch": 1.73, "learning_rate": 3.598086866225055e-05, "loss": 0.5366, "step": 17185 }, { "epoch": 1.73, "learning_rate": 3.597852351509733e-05, "loss": 0.5428, "step": 17186 }, { "epoch": 1.73, "learning_rate": 3.597617832989906e-05, "loss": 0.829, "step": 17187 }, { "epoch": 1.73, "learning_rate": 3.597383310667067e-05, "loss": 0.9543, "step": 17188 }, { "epoch": 1.73, "learning_rate": 3.597148784542707e-05, "loss": 0.3581, "step": 17189 }, { "epoch": 1.73, "learning_rate": 3.5969142546183205e-05, "loss": 0.3573, "step": 17190 }, { "epoch": 1.73, "learning_rate": 3.5966797208953976e-05, "loss": 0.4006, "step": 17191 }, { "epoch": 1.73, "learning_rate": 3.596445183375433e-05, "loss": 0.945, "step": 17192 }, { "epoch": 1.73, "learning_rate": 3.596210642059917e-05, "loss": 0.7738, "step": 17193 }, { "epoch": 1.73, "learning_rate": 3.595976096950344e-05, "loss": 0.4621, "step": 17194 }, { "epoch": 1.73, "learning_rate": 3.595741548048206e-05, "loss": 0.7152, "step": 17195 }, { "epoch": 1.73, "learning_rate": 3.5955069953549955e-05, "loss": 0.3373, "step": 17196 }, { "epoch": 1.73, "learning_rate": 3.5952724388722045e-05, "loss": 0.9664, "step": 17197 }, { "epoch": 1.73, "learning_rate": 3.595037878601326e-05, "loss": 0.5045, "step": 17198 }, { "epoch": 1.73, "learning_rate": 3.594803314543854e-05, "loss": 0.1855, "step": 17199 }, { "epoch": 1.73, "learning_rate": 3.594568746701279e-05, "loss": 0.872, "step": 17200 }, { "epoch": 1.73, "learning_rate": 3.594334175075095e-05, "loss": 0.2309, "step": 17201 }, { "epoch": 1.73, "learning_rate": 3.594099599666794e-05, "loss": 1.0473, "step": 17202 }, { "epoch": 1.73, "learning_rate": 3.593865020477869e-05, "loss": 0.2184, "step": 17203 }, { "epoch": 1.73, "learning_rate": 3.5936304375098133e-05, "loss": 0.5814, "step": 17204 }, { "epoch": 1.73, "learning_rate": 3.5933958507641194e-05, "loss": 0.2697, "step": 17205 }, { "epoch": 1.73, "learning_rate": 3.59316126024228e-05, "loss": 0.7827, "step": 17206 }, { "epoch": 1.73, "learning_rate": 3.592926665945788e-05, "loss": 0.7056, "step": 17207 }, { "epoch": 1.73, "learning_rate": 3.5926920678761366e-05, "loss": 0.4548, "step": 17208 }, { "epoch": 1.73, "learning_rate": 3.5924574660348174e-05, "loss": 1.1313, "step": 17209 }, { "epoch": 1.73, "learning_rate": 3.592222860423325e-05, "loss": 1.201, "step": 17210 }, { "epoch": 1.73, "learning_rate": 3.591988251043151e-05, "loss": 0.6917, "step": 17211 }, { "epoch": 1.73, "learning_rate": 3.591753637895791e-05, "loss": 1.2991, "step": 17212 }, { "epoch": 1.73, "learning_rate": 3.591519020982733e-05, "loss": 0.5254, "step": 17213 }, { "epoch": 1.73, "learning_rate": 3.5912844003054746e-05, "loss": 0.5628, "step": 17214 }, { "epoch": 1.74, "learning_rate": 3.5910497758655074e-05, "loss": 0.819, "step": 17215 }, { "epoch": 1.74, "learning_rate": 3.5908151476643236e-05, "loss": 0.4724, "step": 17216 }, { "epoch": 1.74, "learning_rate": 3.5905805157034174e-05, "loss": 0.5299, "step": 17217 }, { "epoch": 1.74, "learning_rate": 3.590345879984281e-05, "loss": 0.2025, "step": 17218 }, { "epoch": 1.74, "learning_rate": 3.590111240508408e-05, "loss": 0.9771, "step": 17219 }, { "epoch": 1.74, "learning_rate": 3.589876597277292e-05, "loss": 0.2277, "step": 17220 }, { "epoch": 1.74, "learning_rate": 3.589641950292426e-05, "loss": 1.1778, "step": 17221 }, { "epoch": 1.74, "learning_rate": 3.589407299555303e-05, "loss": 1.2417, "step": 17222 }, { "epoch": 1.74, "learning_rate": 3.5891726450674144e-05, "loss": 0.3294, "step": 17223 }, { "epoch": 1.74, "learning_rate": 3.588937986830257e-05, "loss": 0.1513, "step": 17224 }, { "epoch": 1.74, "learning_rate": 3.588703324845322e-05, "loss": 0.2557, "step": 17225 }, { "epoch": 1.74, "learning_rate": 3.588468659114103e-05, "loss": 0.8734, "step": 17226 }, { "epoch": 1.74, "learning_rate": 3.5882339896380924e-05, "loss": 1.4088, "step": 17227 }, { "epoch": 1.74, "learning_rate": 3.587999316418786e-05, "loss": 0.4089, "step": 17228 }, { "epoch": 1.74, "learning_rate": 3.587764639457674e-05, "loss": 1.0536, "step": 17229 }, { "epoch": 1.74, "learning_rate": 3.5875299587562526e-05, "loss": 1.0315, "step": 17230 }, { "epoch": 1.74, "learning_rate": 3.587295274316014e-05, "loss": 0.9537, "step": 17231 }, { "epoch": 1.74, "learning_rate": 3.587060586138451e-05, "loss": 0.0743, "step": 17232 }, { "epoch": 1.74, "learning_rate": 3.586825894225058e-05, "loss": 1.1738, "step": 17233 }, { "epoch": 1.74, "learning_rate": 3.586591198577328e-05, "loss": 0.727, "step": 17234 }, { "epoch": 1.74, "learning_rate": 3.586356499196756e-05, "loss": 0.8056, "step": 17235 }, { "epoch": 1.74, "learning_rate": 3.586121796084833e-05, "loss": 0.6845, "step": 17236 }, { "epoch": 1.74, "learning_rate": 3.5858870892430547e-05, "loss": 1.0261, "step": 17237 }, { "epoch": 1.74, "learning_rate": 3.5856523786729135e-05, "loss": 0.1454, "step": 17238 }, { "epoch": 1.74, "learning_rate": 3.5854176643759034e-05, "loss": 0.6087, "step": 17239 }, { "epoch": 1.74, "learning_rate": 3.585182946353518e-05, "loss": 0.5875, "step": 17240 }, { "epoch": 1.74, "learning_rate": 3.584948224607251e-05, "loss": 1.4079, "step": 17241 }, { "epoch": 1.74, "learning_rate": 3.584713499138596e-05, "loss": 1.0196, "step": 17242 }, { "epoch": 1.74, "learning_rate": 3.584478769949046e-05, "loss": 0.5831, "step": 17243 }, { "epoch": 1.74, "learning_rate": 3.584244037040096e-05, "loss": 0.3324, "step": 17244 }, { "epoch": 1.74, "learning_rate": 3.584009300413239e-05, "loss": 0.2599, "step": 17245 }, { "epoch": 1.74, "learning_rate": 3.58377456006997e-05, "loss": 0.1901, "step": 17246 }, { "epoch": 1.74, "learning_rate": 3.583539816011781e-05, "loss": 0.7307, "step": 17247 }, { "epoch": 1.74, "learning_rate": 3.583305068240167e-05, "loss": 0.4881, "step": 17248 }, { "epoch": 1.74, "learning_rate": 3.583070316756621e-05, "loss": 1.0808, "step": 17249 }, { "epoch": 1.74, "learning_rate": 3.5828355615626366e-05, "loss": 0.2357, "step": 17250 }, { "epoch": 1.74, "learning_rate": 3.58260080265971e-05, "loss": 0.5594, "step": 17251 }, { "epoch": 1.74, "learning_rate": 3.582366040049332e-05, "loss": 1.4326, "step": 17252 }, { "epoch": 1.74, "learning_rate": 3.582131273732999e-05, "loss": 0.2767, "step": 17253 }, { "epoch": 1.74, "learning_rate": 3.581896503712205e-05, "loss": 0.2195, "step": 17254 }, { "epoch": 1.74, "learning_rate": 3.581661729988441e-05, "loss": 0.2486, "step": 17255 }, { "epoch": 1.74, "learning_rate": 3.581426952563204e-05, "loss": 0.5618, "step": 17256 }, { "epoch": 1.74, "learning_rate": 3.581192171437986e-05, "loss": 0.298, "step": 17257 }, { "epoch": 1.74, "learning_rate": 3.580957386614284e-05, "loss": 1.3462, "step": 17258 }, { "epoch": 1.74, "learning_rate": 3.5807225980935886e-05, "loss": 0.8129, "step": 17259 }, { "epoch": 1.74, "learning_rate": 3.5804878058773956e-05, "loss": 0.1606, "step": 17260 }, { "epoch": 1.74, "learning_rate": 3.5802530099672e-05, "loss": 0.9302, "step": 17261 }, { "epoch": 1.74, "learning_rate": 3.580018210364495e-05, "loss": 1.0448, "step": 17262 }, { "epoch": 1.74, "learning_rate": 3.579783407070774e-05, "loss": 0.1249, "step": 17263 }, { "epoch": 1.74, "learning_rate": 3.5795486000875325e-05, "loss": 0.1679, "step": 17264 }, { "epoch": 1.74, "learning_rate": 3.579313789416264e-05, "loss": 0.3703, "step": 17265 }, { "epoch": 1.74, "learning_rate": 3.5790789750584625e-05, "loss": 0.3541, "step": 17266 }, { "epoch": 1.74, "learning_rate": 3.578844157015624e-05, "loss": 0.8882, "step": 17267 }, { "epoch": 1.74, "learning_rate": 3.57860933528924e-05, "loss": 0.9465, "step": 17268 }, { "epoch": 1.74, "learning_rate": 3.578374509880807e-05, "loss": 0.9455, "step": 17269 }, { "epoch": 1.74, "learning_rate": 3.578139680791819e-05, "loss": 0.2272, "step": 17270 }, { "epoch": 1.74, "learning_rate": 3.577904848023769e-05, "loss": 0.6419, "step": 17271 }, { "epoch": 1.74, "learning_rate": 3.577670011578153e-05, "loss": 1.2128, "step": 17272 }, { "epoch": 1.74, "learning_rate": 3.5774351714564645e-05, "loss": 0.8518, "step": 17273 }, { "epoch": 1.74, "learning_rate": 3.577200327660199e-05, "loss": 0.2353, "step": 17274 }, { "epoch": 1.74, "learning_rate": 3.5769654801908485e-05, "loss": 0.4306, "step": 17275 }, { "epoch": 1.74, "learning_rate": 3.5767306290499106e-05, "loss": 0.1403, "step": 17276 }, { "epoch": 1.74, "learning_rate": 3.576495774238878e-05, "loss": 0.4618, "step": 17277 }, { "epoch": 1.74, "learning_rate": 3.576260915759246e-05, "loss": 0.8651, "step": 17278 }, { "epoch": 1.74, "learning_rate": 3.576026053612509e-05, "loss": 0.7741, "step": 17279 }, { "epoch": 1.74, "learning_rate": 3.5757911878001606e-05, "loss": 0.6287, "step": 17280 }, { "epoch": 1.74, "learning_rate": 3.575556318323696e-05, "loss": 0.1317, "step": 17281 }, { "epoch": 1.74, "learning_rate": 3.57532144518461e-05, "loss": 0.4629, "step": 17282 }, { "epoch": 1.74, "learning_rate": 3.575086568384398e-05, "loss": 0.8989, "step": 17283 }, { "epoch": 1.74, "learning_rate": 3.574851687924553e-05, "loss": 0.1874, "step": 17284 }, { "epoch": 1.74, "learning_rate": 3.574616803806571e-05, "loss": 0.4475, "step": 17285 }, { "epoch": 1.74, "learning_rate": 3.574381916031946e-05, "loss": 0.3467, "step": 17286 }, { "epoch": 1.74, "learning_rate": 3.574147024602173e-05, "loss": 1.0119, "step": 17287 }, { "epoch": 1.74, "learning_rate": 3.573912129518747e-05, "loss": 0.2301, "step": 17288 }, { "epoch": 1.74, "learning_rate": 3.573677230783162e-05, "loss": 0.1527, "step": 17289 }, { "epoch": 1.74, "learning_rate": 3.573442328396914e-05, "loss": 0.8002, "step": 17290 }, { "epoch": 1.74, "learning_rate": 3.573207422361497e-05, "loss": 1.3521, "step": 17291 }, { "epoch": 1.74, "learning_rate": 3.5729725126784054e-05, "loss": 1.3884, "step": 17292 }, { "epoch": 1.74, "learning_rate": 3.572737599349136e-05, "loss": 0.0646, "step": 17293 }, { "epoch": 1.74, "learning_rate": 3.572502682375181e-05, "loss": 0.8583, "step": 17294 }, { "epoch": 1.74, "learning_rate": 3.5722677617580375e-05, "loss": 1.1947, "step": 17295 }, { "epoch": 1.74, "learning_rate": 3.5720328374991997e-05, "loss": 0.7075, "step": 17296 }, { "epoch": 1.74, "learning_rate": 3.5717979096001625e-05, "loss": 1.0898, "step": 17297 }, { "epoch": 1.74, "learning_rate": 3.57156297806242e-05, "loss": 1.3416, "step": 17298 }, { "epoch": 1.74, "learning_rate": 3.571328042887469e-05, "loss": 0.3095, "step": 17299 }, { "epoch": 1.74, "learning_rate": 3.5710931040768043e-05, "loss": 0.3808, "step": 17300 }, { "epoch": 1.74, "learning_rate": 3.570858161631919e-05, "loss": 1.3775, "step": 17301 }, { "epoch": 1.74, "learning_rate": 3.57062321555431e-05, "loss": 0.387, "step": 17302 }, { "epoch": 1.74, "learning_rate": 3.570388265845472e-05, "loss": 0.2532, "step": 17303 }, { "epoch": 1.74, "learning_rate": 3.5701533125069e-05, "loss": 0.4289, "step": 17304 }, { "epoch": 1.74, "learning_rate": 3.569918355540089e-05, "loss": 0.6744, "step": 17305 }, { "epoch": 1.74, "learning_rate": 3.569683394946536e-05, "loss": 1.0992, "step": 17306 }, { "epoch": 1.74, "learning_rate": 3.569448430727733e-05, "loss": 0.5268, "step": 17307 }, { "epoch": 1.74, "learning_rate": 3.5692134628851764e-05, "loss": 0.6053, "step": 17308 }, { "epoch": 1.74, "learning_rate": 3.568978491420363e-05, "loss": 0.7409, "step": 17309 }, { "epoch": 1.74, "learning_rate": 3.5687435163347874e-05, "loss": 0.2872, "step": 17310 }, { "epoch": 1.74, "learning_rate": 3.568508537629943e-05, "loss": 0.5978, "step": 17311 }, { "epoch": 1.74, "learning_rate": 3.5682735553073265e-05, "loss": 0.3191, "step": 17312 }, { "epoch": 1.75, "learning_rate": 3.5680385693684344e-05, "loss": 0.2528, "step": 17313 }, { "epoch": 1.75, "learning_rate": 3.56780357981476e-05, "loss": 0.6906, "step": 17314 }, { "epoch": 1.75, "learning_rate": 3.5675685866478e-05, "loss": 0.5816, "step": 17315 }, { "epoch": 1.75, "learning_rate": 3.56733358986905e-05, "loss": 1.5127, "step": 17316 }, { "epoch": 1.75, "learning_rate": 3.567098589480004e-05, "loss": 0.1692, "step": 17317 }, { "epoch": 1.75, "learning_rate": 3.566863585482159e-05, "loss": 0.5459, "step": 17318 }, { "epoch": 1.75, "learning_rate": 3.5666285778770094e-05, "loss": 0.2634, "step": 17319 }, { "epoch": 1.75, "learning_rate": 3.566393566666051e-05, "loss": 0.0038, "step": 17320 }, { "epoch": 1.75, "learning_rate": 3.566158551850779e-05, "loss": 0.8119, "step": 17321 }, { "epoch": 1.75, "learning_rate": 3.565923533432691e-05, "loss": 1.0106, "step": 17322 }, { "epoch": 1.75, "learning_rate": 3.565688511413279e-05, "loss": 1.435, "step": 17323 }, { "epoch": 1.75, "learning_rate": 3.5654534857940416e-05, "loss": 1.2487, "step": 17324 }, { "epoch": 1.75, "learning_rate": 3.565218456576473e-05, "loss": 0.8375, "step": 17325 }, { "epoch": 1.75, "learning_rate": 3.5649834237620685e-05, "loss": 0.2827, "step": 17326 }, { "epoch": 1.75, "learning_rate": 3.564748387352326e-05, "loss": 1.3181, "step": 17327 }, { "epoch": 1.75, "learning_rate": 3.5645133473487385e-05, "loss": 0.8065, "step": 17328 }, { "epoch": 1.75, "learning_rate": 3.564278303752803e-05, "loss": 0.2502, "step": 17329 }, { "epoch": 1.75, "learning_rate": 3.564043256566015e-05, "loss": 0.2804, "step": 17330 }, { "epoch": 1.75, "learning_rate": 3.5638082057898704e-05, "loss": 0.2049, "step": 17331 }, { "epoch": 1.75, "learning_rate": 3.5635731514258646e-05, "loss": 0.3794, "step": 17332 }, { "epoch": 1.75, "learning_rate": 3.563338093475494e-05, "loss": 0.5615, "step": 17333 }, { "epoch": 1.75, "learning_rate": 3.5631030319402545e-05, "loss": 1.0698, "step": 17334 }, { "epoch": 1.75, "learning_rate": 3.562867966821641e-05, "loss": 0.7515, "step": 17335 }, { "epoch": 1.75, "learning_rate": 3.5626328981211495e-05, "loss": 1.3533, "step": 17336 }, { "epoch": 1.75, "learning_rate": 3.5623978258402765e-05, "loss": 0.1506, "step": 17337 }, { "epoch": 1.75, "learning_rate": 3.562162749980518e-05, "loss": 0.3899, "step": 17338 }, { "epoch": 1.75, "learning_rate": 3.561927670543369e-05, "loss": 0.5583, "step": 17339 }, { "epoch": 1.75, "learning_rate": 3.561692587530327e-05, "loss": 0.5497, "step": 17340 }, { "epoch": 1.75, "learning_rate": 3.561457500942887e-05, "loss": 0.9442, "step": 17341 }, { "epoch": 1.75, "learning_rate": 3.561222410782544e-05, "loss": 1.1284, "step": 17342 }, { "epoch": 1.75, "learning_rate": 3.560987317050796e-05, "loss": 0.5219, "step": 17343 }, { "epoch": 1.75, "learning_rate": 3.560752219749137e-05, "loss": 0.7397, "step": 17344 }, { "epoch": 1.75, "learning_rate": 3.560517118879066e-05, "loss": 0.2812, "step": 17345 }, { "epoch": 1.75, "learning_rate": 3.5602820144420756e-05, "loss": 0.8427, "step": 17346 }, { "epoch": 1.75, "learning_rate": 3.560046906439664e-05, "loss": 0.2564, "step": 17347 }, { "epoch": 1.75, "learning_rate": 3.559811794873328e-05, "loss": 0.9666, "step": 17348 }, { "epoch": 1.75, "learning_rate": 3.5595766797445615e-05, "loss": 0.7277, "step": 17349 }, { "epoch": 1.75, "learning_rate": 3.5593415610548624e-05, "loss": 0.1462, "step": 17350 }, { "epoch": 1.75, "learning_rate": 3.559106438805726e-05, "loss": 0.8892, "step": 17351 }, { "epoch": 1.75, "learning_rate": 3.55887131299865e-05, "loss": 1.039, "step": 17352 }, { "epoch": 1.75, "learning_rate": 3.558636183635128e-05, "loss": 0.2201, "step": 17353 }, { "epoch": 1.75, "learning_rate": 3.558401050716659e-05, "loss": 0.7084, "step": 17354 }, { "epoch": 1.75, "learning_rate": 3.5581659142447374e-05, "loss": 0.6001, "step": 17355 }, { "epoch": 1.75, "learning_rate": 3.557930774220861e-05, "loss": 0.3543, "step": 17356 }, { "epoch": 1.75, "learning_rate": 3.557695630646525e-05, "loss": 1.116, "step": 17357 }, { "epoch": 1.75, "learning_rate": 3.557460483523227e-05, "loss": 1.2579, "step": 17358 }, { "epoch": 1.75, "learning_rate": 3.557225332852461e-05, "loss": 0.5884, "step": 17359 }, { "epoch": 1.75, "learning_rate": 3.5569901786357256e-05, "loss": 0.5588, "step": 17360 }, { "epoch": 1.75, "learning_rate": 3.556755020874517e-05, "loss": 0.1701, "step": 17361 }, { "epoch": 1.75, "learning_rate": 3.556519859570331e-05, "loss": 0.8452, "step": 17362 }, { "epoch": 1.75, "learning_rate": 3.5562846947246645e-05, "loss": 0.1865, "step": 17363 }, { "epoch": 1.75, "learning_rate": 3.556049526339013e-05, "loss": 0.6632, "step": 17364 }, { "epoch": 1.75, "learning_rate": 3.555814354414875e-05, "loss": 0.4023, "step": 17365 }, { "epoch": 1.75, "learning_rate": 3.555579178953745e-05, "loss": 0.9882, "step": 17366 }, { "epoch": 1.75, "learning_rate": 3.5553439999571205e-05, "loss": 1.0452, "step": 17367 }, { "epoch": 1.75, "learning_rate": 3.5551088174264994e-05, "loss": 0.7599, "step": 17368 }, { "epoch": 1.75, "learning_rate": 3.554873631363375e-05, "loss": 0.2619, "step": 17369 }, { "epoch": 1.75, "learning_rate": 3.554638441769247e-05, "loss": 0.7482, "step": 17370 }, { "epoch": 1.75, "learning_rate": 3.554403248645611e-05, "loss": 0.6441, "step": 17371 }, { "epoch": 1.75, "learning_rate": 3.554168051993963e-05, "loss": 0.2135, "step": 17372 }, { "epoch": 1.75, "learning_rate": 3.553932851815801e-05, "loss": 0.7641, "step": 17373 }, { "epoch": 1.75, "learning_rate": 3.553697648112621e-05, "loss": 1.2003, "step": 17374 }, { "epoch": 1.75, "learning_rate": 3.5534624408859194e-05, "loss": 1.0575, "step": 17375 }, { "epoch": 1.75, "learning_rate": 3.553227230137193e-05, "loss": 0.5804, "step": 17376 }, { "epoch": 1.75, "learning_rate": 3.55299201586794e-05, "loss": 1.0411, "step": 17377 }, { "epoch": 1.75, "learning_rate": 3.552756798079655e-05, "loss": 1.0224, "step": 17378 }, { "epoch": 1.75, "learning_rate": 3.5525215767738376e-05, "loss": 1.1963, "step": 17379 }, { "epoch": 1.75, "learning_rate": 3.552286351951982e-05, "loss": 0.1682, "step": 17380 }, { "epoch": 1.75, "learning_rate": 3.5520511236155865e-05, "loss": 1.1456, "step": 17381 }, { "epoch": 1.75, "learning_rate": 3.5518158917661475e-05, "loss": 0.7857, "step": 17382 }, { "epoch": 1.75, "learning_rate": 3.551580656405161e-05, "loss": 1.1527, "step": 17383 }, { "epoch": 1.75, "learning_rate": 3.5513454175341274e-05, "loss": 0.1564, "step": 17384 }, { "epoch": 1.75, "learning_rate": 3.5511101751545394e-05, "loss": 0.404, "step": 17385 }, { "epoch": 1.75, "learning_rate": 3.550874929267897e-05, "loss": 0.7417, "step": 17386 }, { "epoch": 1.75, "learning_rate": 3.5506396798756953e-05, "loss": 1.0817, "step": 17387 }, { "epoch": 1.75, "learning_rate": 3.550404426979433e-05, "loss": 0.1885, "step": 17388 }, { "epoch": 1.75, "learning_rate": 3.550169170580606e-05, "loss": 0.6133, "step": 17389 }, { "epoch": 1.75, "learning_rate": 3.5499339106807115e-05, "loss": 0.8453, "step": 17390 }, { "epoch": 1.75, "learning_rate": 3.549698647281247e-05, "loss": 0.4207, "step": 17391 }, { "epoch": 1.75, "learning_rate": 3.549463380383708e-05, "loss": 1.1136, "step": 17392 }, { "epoch": 1.75, "learning_rate": 3.549228109989595e-05, "loss": 0.2517, "step": 17393 }, { "epoch": 1.75, "learning_rate": 3.548992836100403e-05, "loss": 0.301, "step": 17394 }, { "epoch": 1.75, "learning_rate": 3.5487575587176284e-05, "loss": 0.8782, "step": 17395 }, { "epoch": 1.75, "learning_rate": 3.54852227784277e-05, "loss": 0.2707, "step": 17396 }, { "epoch": 1.75, "learning_rate": 3.548286993477325e-05, "loss": 0.1526, "step": 17397 }, { "epoch": 1.75, "learning_rate": 3.54805170562279e-05, "loss": 0.41, "step": 17398 }, { "epoch": 1.75, "learning_rate": 3.547816414280661e-05, "loss": 0.8654, "step": 17399 }, { "epoch": 1.75, "learning_rate": 3.547581119452439e-05, "loss": 0.1231, "step": 17400 }, { "epoch": 1.75, "learning_rate": 3.547345821139617e-05, "loss": 1.239, "step": 17401 }, { "epoch": 1.75, "learning_rate": 3.5471105193436955e-05, "loss": 0.5612, "step": 17402 }, { "epoch": 1.75, "learning_rate": 3.546875214066171e-05, "loss": 0.7131, "step": 17403 }, { "epoch": 1.75, "learning_rate": 3.54663990530854e-05, "loss": 0.517, "step": 17404 }, { "epoch": 1.75, "learning_rate": 3.546404593072301e-05, "loss": 0.5853, "step": 17405 }, { "epoch": 1.75, "learning_rate": 3.546169277358951e-05, "loss": 0.2141, "step": 17406 }, { "epoch": 1.75, "learning_rate": 3.545933958169988e-05, "loss": 1.0485, "step": 17407 }, { "epoch": 1.75, "learning_rate": 3.5456986355069074e-05, "loss": 0.4214, "step": 17408 }, { "epoch": 1.75, "learning_rate": 3.5454633093712095e-05, "loss": 0.9028, "step": 17409 }, { "epoch": 1.75, "learning_rate": 3.545227979764391e-05, "loss": 0.5576, "step": 17410 }, { "epoch": 1.75, "learning_rate": 3.5449926466879484e-05, "loss": 1.0728, "step": 17411 }, { "epoch": 1.76, "learning_rate": 3.54475731014338e-05, "loss": 0.2528, "step": 17412 }, { "epoch": 1.76, "learning_rate": 3.544521970132183e-05, "loss": 0.7179, "step": 17413 }, { "epoch": 1.76, "learning_rate": 3.544286626655856e-05, "loss": 0.4766, "step": 17414 }, { "epoch": 1.76, "learning_rate": 3.544051279715895e-05, "loss": 0.5455, "step": 17415 }, { "epoch": 1.76, "learning_rate": 3.5438159293137995e-05, "loss": 0.3786, "step": 17416 }, { "epoch": 1.76, "learning_rate": 3.5435805754510655e-05, "loss": 0.4837, "step": 17417 }, { "epoch": 1.76, "learning_rate": 3.5433452181291926e-05, "loss": 1.0117, "step": 17418 }, { "epoch": 1.76, "learning_rate": 3.543109857349677e-05, "loss": 1.0139, "step": 17419 }, { "epoch": 1.76, "learning_rate": 3.542874493114016e-05, "loss": 1.2007, "step": 17420 }, { "epoch": 1.76, "learning_rate": 3.5426391254237094e-05, "loss": 0.5499, "step": 17421 }, { "epoch": 1.76, "learning_rate": 3.542403754280253e-05, "loss": 0.6662, "step": 17422 }, { "epoch": 1.76, "learning_rate": 3.542168379685146e-05, "loss": 0.2454, "step": 17423 }, { "epoch": 1.76, "learning_rate": 3.541933001639885e-05, "loss": 0.0898, "step": 17424 }, { "epoch": 1.76, "learning_rate": 3.541697620145969e-05, "loss": 1.1101, "step": 17425 }, { "epoch": 1.76, "learning_rate": 3.541462235204896e-05, "loss": 1.0079, "step": 17426 }, { "epoch": 1.76, "learning_rate": 3.541226846818163e-05, "loss": 0.9624, "step": 17427 }, { "epoch": 1.76, "learning_rate": 3.540991454987268e-05, "loss": 0.9975, "step": 17428 }, { "epoch": 1.76, "learning_rate": 3.54075605971371e-05, "loss": 0.8167, "step": 17429 }, { "epoch": 1.76, "learning_rate": 3.540520660998985e-05, "loss": 0.6807, "step": 17430 }, { "epoch": 1.76, "learning_rate": 3.540285258844591e-05, "loss": 0.6851, "step": 17431 }, { "epoch": 1.76, "learning_rate": 3.5400498532520306e-05, "loss": 0.113, "step": 17432 }, { "epoch": 1.76, "learning_rate": 3.5398144442227956e-05, "loss": 0.6295, "step": 17433 }, { "epoch": 1.76, "learning_rate": 3.5395790317583874e-05, "loss": 0.1483, "step": 17434 }, { "epoch": 1.76, "learning_rate": 3.539343615860304e-05, "loss": 0.8435, "step": 17435 }, { "epoch": 1.76, "learning_rate": 3.5391081965300426e-05, "loss": 0.2737, "step": 17436 }, { "epoch": 1.76, "learning_rate": 3.5388727737691025e-05, "loss": 0.6248, "step": 17437 }, { "epoch": 1.76, "learning_rate": 3.5386373475789794e-05, "loss": 0.9252, "step": 17438 }, { "epoch": 1.76, "learning_rate": 3.5384019179611755e-05, "loss": 0.1872, "step": 17439 }, { "epoch": 1.76, "learning_rate": 3.538166484917184e-05, "loss": 1.0868, "step": 17440 }, { "epoch": 1.76, "learning_rate": 3.5379310484485074e-05, "loss": 0.2512, "step": 17441 }, { "epoch": 1.76, "learning_rate": 3.537695608556642e-05, "loss": 0.781, "step": 17442 }, { "epoch": 1.76, "learning_rate": 3.537460165243086e-05, "loss": 0.5576, "step": 17443 }, { "epoch": 1.76, "learning_rate": 3.537224718509339e-05, "loss": 0.4542, "step": 17444 }, { "epoch": 1.76, "learning_rate": 3.536989268356897e-05, "loss": 0.4039, "step": 17445 }, { "epoch": 1.76, "learning_rate": 3.53675381478726e-05, "loss": 0.3288, "step": 17446 }, { "epoch": 1.76, "learning_rate": 3.536518357801925e-05, "loss": 0.1613, "step": 17447 }, { "epoch": 1.76, "learning_rate": 3.536282897402393e-05, "loss": 0.5438, "step": 17448 }, { "epoch": 1.76, "learning_rate": 3.53604743359016e-05, "loss": 1.1223, "step": 17449 }, { "epoch": 1.76, "learning_rate": 3.535811966366724e-05, "loss": 0.4362, "step": 17450 }, { "epoch": 1.76, "learning_rate": 3.535576495733587e-05, "loss": 0.8156, "step": 17451 }, { "epoch": 1.76, "learning_rate": 3.535341021692243e-05, "loss": 0.8992, "step": 17452 }, { "epoch": 1.76, "learning_rate": 3.5351055442441925e-05, "loss": 0.9678, "step": 17453 }, { "epoch": 1.76, "learning_rate": 3.5348700633909337e-05, "loss": 0.7195, "step": 17454 }, { "epoch": 1.76, "learning_rate": 3.534634579133966e-05, "loss": 0.8955, "step": 17455 }, { "epoch": 1.76, "learning_rate": 3.534399091474786e-05, "loss": 0.3459, "step": 17456 }, { "epoch": 1.76, "learning_rate": 3.534163600414895e-05, "loss": 0.7387, "step": 17457 }, { "epoch": 1.76, "learning_rate": 3.533928105955789e-05, "loss": 0.8929, "step": 17458 }, { "epoch": 1.76, "learning_rate": 3.533692608098969e-05, "loss": 1.2841, "step": 17459 }, { "epoch": 1.76, "learning_rate": 3.533457106845931e-05, "loss": 0.2401, "step": 17460 }, { "epoch": 1.76, "learning_rate": 3.533221602198174e-05, "loss": 0.9553, "step": 17461 }, { "epoch": 1.76, "learning_rate": 3.5329860941572e-05, "loss": 1.1539, "step": 17462 }, { "epoch": 1.76, "learning_rate": 3.5327505827245035e-05, "loss": 0.9299, "step": 17463 }, { "epoch": 1.76, "learning_rate": 3.532515067901586e-05, "loss": 1.0037, "step": 17464 }, { "epoch": 1.76, "learning_rate": 3.5322795496899444e-05, "loss": 0.5043, "step": 17465 }, { "epoch": 1.76, "learning_rate": 3.5320440280910783e-05, "loss": 0.8537, "step": 17466 }, { "epoch": 1.76, "learning_rate": 3.531808503106487e-05, "loss": 1.1188, "step": 17467 }, { "epoch": 1.76, "learning_rate": 3.5315729747376686e-05, "loss": 0.5906, "step": 17468 }, { "epoch": 1.76, "learning_rate": 3.531337442986122e-05, "loss": 0.1937, "step": 17469 }, { "epoch": 1.76, "learning_rate": 3.5311019078533446e-05, "loss": 1.2644, "step": 17470 }, { "epoch": 1.76, "learning_rate": 3.530866369340839e-05, "loss": 0.1454, "step": 17471 }, { "epoch": 1.76, "learning_rate": 3.5306308274501e-05, "loss": 0.4684, "step": 17472 }, { "epoch": 1.76, "learning_rate": 3.53039528218263e-05, "loss": 0.8061, "step": 17473 }, { "epoch": 1.76, "learning_rate": 3.530159733539925e-05, "loss": 0.9335, "step": 17474 }, { "epoch": 1.76, "learning_rate": 3.5299241815234844e-05, "loss": 0.3289, "step": 17475 }, { "epoch": 1.76, "learning_rate": 3.529688626134809e-05, "loss": 0.9491, "step": 17476 }, { "epoch": 1.76, "learning_rate": 3.529453067375396e-05, "loss": 0.3766, "step": 17477 }, { "epoch": 1.76, "learning_rate": 3.529217505246746e-05, "loss": 0.6755, "step": 17478 }, { "epoch": 1.76, "learning_rate": 3.5289819397503564e-05, "loss": 0.4077, "step": 17479 }, { "epoch": 1.76, "learning_rate": 3.5287463708877275e-05, "loss": 0.4528, "step": 17480 }, { "epoch": 1.76, "learning_rate": 3.528510798660358e-05, "loss": 0.3741, "step": 17481 }, { "epoch": 1.76, "learning_rate": 3.528275223069746e-05, "loss": 0.943, "step": 17482 }, { "epoch": 1.76, "learning_rate": 3.528039644117392e-05, "loss": 0.2058, "step": 17483 }, { "epoch": 1.76, "learning_rate": 3.527804061804795e-05, "loss": 1.0111, "step": 17484 }, { "epoch": 1.76, "learning_rate": 3.527568476133453e-05, "loss": 0.5249, "step": 17485 }, { "epoch": 1.76, "learning_rate": 3.527332887104866e-05, "loss": 0.9574, "step": 17486 }, { "epoch": 1.76, "learning_rate": 3.527097294720533e-05, "loss": 0.5981, "step": 17487 }, { "epoch": 1.76, "learning_rate": 3.5268616989819534e-05, "loss": 0.2898, "step": 17488 }, { "epoch": 1.76, "learning_rate": 3.526626099890627e-05, "loss": 0.5052, "step": 17489 }, { "epoch": 1.76, "learning_rate": 3.5263904974480524e-05, "loss": 0.375, "step": 17490 }, { "epoch": 1.76, "learning_rate": 3.526154891655728e-05, "loss": 0.1957, "step": 17491 }, { "epoch": 1.76, "learning_rate": 3.5259192825151546e-05, "loss": 0.8659, "step": 17492 }, { "epoch": 1.76, "learning_rate": 3.5256836700278305e-05, "loss": 0.8901, "step": 17493 }, { "epoch": 1.76, "learning_rate": 3.525448054195256e-05, "loss": 1.4119, "step": 17494 }, { "epoch": 1.76, "learning_rate": 3.52521243501893e-05, "loss": 0.9963, "step": 17495 }, { "epoch": 1.76, "learning_rate": 3.524976812500352e-05, "loss": 1.0637, "step": 17496 }, { "epoch": 1.76, "learning_rate": 3.524741186641021e-05, "loss": 0.8709, "step": 17497 }, { "epoch": 1.76, "learning_rate": 3.524505557442437e-05, "loss": 0.487, "step": 17498 }, { "epoch": 1.76, "learning_rate": 3.524269924906099e-05, "loss": 0.113, "step": 17499 }, { "epoch": 1.76, "learning_rate": 3.5240342890335064e-05, "loss": 0.3235, "step": 17500 }, { "epoch": 1.76, "learning_rate": 3.523798649826159e-05, "loss": 0.5109, "step": 17501 }, { "epoch": 1.76, "learning_rate": 3.523563007285556e-05, "loss": 0.3825, "step": 17502 }, { "epoch": 1.76, "learning_rate": 3.5233273614131975e-05, "loss": 1.0929, "step": 17503 }, { "epoch": 1.76, "learning_rate": 3.523091712210583e-05, "loss": 0.9621, "step": 17504 }, { "epoch": 1.76, "learning_rate": 3.522856059679212e-05, "loss": 1.1935, "step": 17505 }, { "epoch": 1.76, "learning_rate": 3.522620403820583e-05, "loss": 0.5821, "step": 17506 }, { "epoch": 1.76, "learning_rate": 3.522384744636197e-05, "loss": 0.5458, "step": 17507 }, { "epoch": 1.76, "learning_rate": 3.522149082127554e-05, "loss": 0.3343, "step": 17508 }, { "epoch": 1.76, "learning_rate": 3.521913416296151e-05, "loss": 1.0383, "step": 17509 }, { "epoch": 1.77, "learning_rate": 3.521677747143492e-05, "loss": 0.9112, "step": 17510 }, { "epoch": 1.77, "learning_rate": 3.5214420746710725e-05, "loss": 1.0097, "step": 17511 }, { "epoch": 1.77, "learning_rate": 3.521206398880395e-05, "loss": 0.1281, "step": 17512 }, { "epoch": 1.77, "learning_rate": 3.520970719772957e-05, "loss": 0.7065, "step": 17513 }, { "epoch": 1.77, "learning_rate": 3.520735037350261e-05, "loss": 0.3, "step": 17514 }, { "epoch": 1.77, "learning_rate": 3.520499351613805e-05, "loss": 0.3033, "step": 17515 }, { "epoch": 1.77, "learning_rate": 3.520263662565088e-05, "loss": 0.2193, "step": 17516 }, { "epoch": 1.77, "learning_rate": 3.520027970205612e-05, "loss": 0.8734, "step": 17517 }, { "epoch": 1.77, "learning_rate": 3.519792274536875e-05, "loss": 0.1149, "step": 17518 }, { "epoch": 1.77, "learning_rate": 3.519556575560379e-05, "loss": 0.569, "step": 17519 }, { "epoch": 1.77, "learning_rate": 3.519320873277622e-05, "loss": 0.2419, "step": 17520 }, { "epoch": 1.77, "learning_rate": 3.519085167690104e-05, "loss": 0.2391, "step": 17521 }, { "epoch": 1.77, "learning_rate": 3.5188494587993264e-05, "loss": 0.9829, "step": 17522 }, { "epoch": 1.77, "learning_rate": 3.518613746606787e-05, "loss": 0.6602, "step": 17523 }, { "epoch": 1.77, "learning_rate": 3.518378031113988e-05, "loss": 0.2102, "step": 17524 }, { "epoch": 1.77, "learning_rate": 3.518142312322428e-05, "loss": 0.7762, "step": 17525 }, { "epoch": 1.77, "learning_rate": 3.517906590233608e-05, "loss": 0.2711, "step": 17526 }, { "epoch": 1.77, "learning_rate": 3.517670864849027e-05, "loss": 0.8218, "step": 17527 }, { "epoch": 1.77, "learning_rate": 3.517435136170185e-05, "loss": 0.564, "step": 17528 }, { "epoch": 1.77, "learning_rate": 3.5171994041985836e-05, "loss": 0.1574, "step": 17529 }, { "epoch": 1.77, "learning_rate": 3.5169636689357217e-05, "loss": 0.2248, "step": 17530 }, { "epoch": 1.77, "learning_rate": 3.5167279303830996e-05, "loss": 0.4214, "step": 17531 }, { "epoch": 1.77, "learning_rate": 3.5164921885422164e-05, "loss": 0.4398, "step": 17532 }, { "epoch": 1.77, "learning_rate": 3.516256443414576e-05, "loss": 0.806, "step": 17533 }, { "epoch": 1.77, "learning_rate": 3.5160206950016737e-05, "loss": 0.8299, "step": 17534 }, { "epoch": 1.77, "learning_rate": 3.5157849433050126e-05, "loss": 0.2654, "step": 17535 }, { "epoch": 1.77, "learning_rate": 3.515549188326092e-05, "loss": 1.2187, "step": 17536 }, { "epoch": 1.77, "learning_rate": 3.5153134300664137e-05, "loss": 0.4746, "step": 17537 }, { "epoch": 1.77, "learning_rate": 3.5150776685274764e-05, "loss": 0.8101, "step": 17538 }, { "epoch": 1.77, "learning_rate": 3.51484190371078e-05, "loss": 0.4293, "step": 17539 }, { "epoch": 1.77, "learning_rate": 3.514606135617826e-05, "loss": 0.4636, "step": 17540 }, { "epoch": 1.77, "learning_rate": 3.514370364250114e-05, "loss": 0.2122, "step": 17541 }, { "epoch": 1.77, "learning_rate": 3.514134589609145e-05, "loss": 0.5755, "step": 17542 }, { "epoch": 1.77, "learning_rate": 3.5138988116964195e-05, "loss": 0.6041, "step": 17543 }, { "epoch": 1.77, "learning_rate": 3.513663030513437e-05, "loss": 0.742, "step": 17544 }, { "epoch": 1.77, "learning_rate": 3.513427246061698e-05, "loss": 0.6142, "step": 17545 }, { "epoch": 1.77, "learning_rate": 3.513191458342704e-05, "loss": 1.0831, "step": 17546 }, { "epoch": 1.77, "learning_rate": 3.512955667357954e-05, "loss": 0.563, "step": 17547 }, { "epoch": 1.77, "learning_rate": 3.512719873108949e-05, "loss": 0.2675, "step": 17548 }, { "epoch": 1.77, "learning_rate": 3.512484075597192e-05, "loss": 0.6926, "step": 17549 }, { "epoch": 1.77, "learning_rate": 3.512248274824179e-05, "loss": 0.8463, "step": 17550 }, { "epoch": 1.77, "learning_rate": 3.512012470791414e-05, "loss": 0.2807, "step": 17551 }, { "epoch": 1.77, "learning_rate": 3.511776663500396e-05, "loss": 0.4217, "step": 17552 }, { "epoch": 1.77, "learning_rate": 3.511540852952627e-05, "loss": 0.2092, "step": 17553 }, { "epoch": 1.77, "learning_rate": 3.5113050391496044e-05, "loss": 0.5744, "step": 17554 }, { "epoch": 1.77, "learning_rate": 3.5110692220928325e-05, "loss": 0.6239, "step": 17555 }, { "epoch": 1.77, "learning_rate": 3.5108334017838104e-05, "loss": 0.8079, "step": 17556 }, { "epoch": 1.77, "learning_rate": 3.5105975782240386e-05, "loss": 0.4735, "step": 17557 }, { "epoch": 1.77, "learning_rate": 3.5103617514150185e-05, "loss": 0.7344, "step": 17558 }, { "epoch": 1.77, "learning_rate": 3.5101259213582505e-05, "loss": 0.7525, "step": 17559 }, { "epoch": 1.77, "learning_rate": 3.509890088055235e-05, "loss": 0.4056, "step": 17560 }, { "epoch": 1.77, "learning_rate": 3.5096542515074725e-05, "loss": 0.6494, "step": 17561 }, { "epoch": 1.77, "learning_rate": 3.509418411716465e-05, "loss": 0.1869, "step": 17562 }, { "epoch": 1.77, "learning_rate": 3.509182568683711e-05, "loss": 0.2485, "step": 17563 }, { "epoch": 1.77, "learning_rate": 3.5089467224107143e-05, "loss": 0.598, "step": 17564 }, { "epoch": 1.77, "learning_rate": 3.5087108728989745e-05, "loss": 0.7357, "step": 17565 }, { "epoch": 1.77, "learning_rate": 3.508475020149991e-05, "loss": 0.6507, "step": 17566 }, { "epoch": 1.77, "learning_rate": 3.508239164165266e-05, "loss": 0.8197, "step": 17567 }, { "epoch": 1.77, "learning_rate": 3.508003304946301e-05, "loss": 0.2793, "step": 17568 }, { "epoch": 1.77, "learning_rate": 3.507767442494597e-05, "loss": 0.5077, "step": 17569 }, { "epoch": 1.77, "learning_rate": 3.507531576811652e-05, "loss": 0.3033, "step": 17570 }, { "epoch": 1.77, "learning_rate": 3.50729570789897e-05, "loss": 0.6318, "step": 17571 }, { "epoch": 1.77, "learning_rate": 3.5070598357580526e-05, "loss": 0.8546, "step": 17572 }, { "epoch": 1.77, "learning_rate": 3.5068239603903975e-05, "loss": 0.4269, "step": 17573 }, { "epoch": 1.77, "learning_rate": 3.506588081797508e-05, "loss": 0.4963, "step": 17574 }, { "epoch": 1.77, "learning_rate": 3.506352199980885e-05, "loss": 0.3115, "step": 17575 }, { "epoch": 1.77, "learning_rate": 3.506116314942029e-05, "loss": 0.9121, "step": 17576 }, { "epoch": 1.77, "learning_rate": 3.5058804266824406e-05, "loss": 0.6165, "step": 17577 }, { "epoch": 1.77, "learning_rate": 3.5056445352036226e-05, "loss": 0.792, "step": 17578 }, { "epoch": 1.77, "learning_rate": 3.5054086405070754e-05, "loss": 0.2834, "step": 17579 }, { "epoch": 1.77, "learning_rate": 3.505172742594299e-05, "loss": 0.2647, "step": 17580 }, { "epoch": 1.77, "learning_rate": 3.5049368414667964e-05, "loss": 0.5848, "step": 17581 }, { "epoch": 1.77, "learning_rate": 3.504700937126066e-05, "loss": 1.2494, "step": 17582 }, { "epoch": 1.77, "learning_rate": 3.504465029573612e-05, "loss": 0.9285, "step": 17583 }, { "epoch": 1.77, "learning_rate": 3.504229118810935e-05, "loss": 1.0164, "step": 17584 }, { "epoch": 1.77, "learning_rate": 3.5039932048395353e-05, "loss": 0.5408, "step": 17585 }, { "epoch": 1.77, "learning_rate": 3.5037572876609144e-05, "loss": 0.5106, "step": 17586 }, { "epoch": 1.77, "learning_rate": 3.503521367276573e-05, "loss": 0.1771, "step": 17587 }, { "epoch": 1.77, "learning_rate": 3.503285443688015e-05, "loss": 0.7838, "step": 17588 }, { "epoch": 1.77, "learning_rate": 3.503049516896738e-05, "loss": 0.8896, "step": 17589 }, { "epoch": 1.77, "learning_rate": 3.502813586904246e-05, "loss": 0.4227, "step": 17590 }, { "epoch": 1.77, "learning_rate": 3.5025776537120404e-05, "loss": 1.105, "step": 17591 }, { "epoch": 1.77, "learning_rate": 3.502341717321621e-05, "loss": 0.5032, "step": 17592 }, { "epoch": 1.77, "learning_rate": 3.50210577773449e-05, "loss": 0.3844, "step": 17593 }, { "epoch": 1.77, "learning_rate": 3.50186983495215e-05, "loss": 1.1197, "step": 17594 }, { "epoch": 1.77, "learning_rate": 3.5016338889761e-05, "loss": 0.1429, "step": 17595 }, { "epoch": 1.77, "learning_rate": 3.5013979398078424e-05, "loss": 0.4301, "step": 17596 }, { "epoch": 1.77, "learning_rate": 3.50116198744888e-05, "loss": 0.641, "step": 17597 }, { "epoch": 1.77, "learning_rate": 3.500926031900713e-05, "loss": 0.257, "step": 17598 }, { "epoch": 1.77, "learning_rate": 3.500690073164844e-05, "loss": 1.0483, "step": 17599 }, { "epoch": 1.77, "learning_rate": 3.5004541112427725e-05, "loss": 0.2151, "step": 17600 }, { "epoch": 1.77, "learning_rate": 3.5002181461360025e-05, "loss": 0.1325, "step": 17601 }, { "epoch": 1.77, "learning_rate": 3.4999821778460345e-05, "loss": 0.3442, "step": 17602 }, { "epoch": 1.77, "learning_rate": 3.499746206374369e-05, "loss": 0.475, "step": 17603 }, { "epoch": 1.77, "learning_rate": 3.49951023172251e-05, "loss": 1.7067, "step": 17604 }, { "epoch": 1.77, "learning_rate": 3.4992742538919566e-05, "loss": 0.6231, "step": 17605 }, { "epoch": 1.77, "learning_rate": 3.499038272884213e-05, "loss": 0.5262, "step": 17606 }, { "epoch": 1.77, "learning_rate": 3.498802288700779e-05, "loss": 1.221, "step": 17607 }, { "epoch": 1.77, "learning_rate": 3.4985663013431576e-05, "loss": 0.7931, "step": 17608 }, { "epoch": 1.78, "learning_rate": 3.49833031081285e-05, "loss": 1.1824, "step": 17609 }, { "epoch": 1.78, "learning_rate": 3.498094317111357e-05, "loss": 0.9984, "step": 17610 }, { "epoch": 1.78, "learning_rate": 3.497858320240182e-05, "loss": 0.8121, "step": 17611 }, { "epoch": 1.78, "learning_rate": 3.497622320200825e-05, "loss": 0.4054, "step": 17612 }, { "epoch": 1.78, "learning_rate": 3.4973863169947904e-05, "loss": 0.3162, "step": 17613 }, { "epoch": 1.78, "learning_rate": 3.497150310623578e-05, "loss": 0.9448, "step": 17614 }, { "epoch": 1.78, "learning_rate": 3.4969143010886895e-05, "loss": 0.4968, "step": 17615 }, { "epoch": 1.78, "learning_rate": 3.496678288391628e-05, "loss": 0.2164, "step": 17616 }, { "epoch": 1.78, "learning_rate": 3.4964422725338945e-05, "loss": 0.7713, "step": 17617 }, { "epoch": 1.78, "learning_rate": 3.4962062535169915e-05, "loss": 0.9482, "step": 17618 }, { "epoch": 1.78, "learning_rate": 3.4959702313424195e-05, "loss": 0.193, "step": 17619 }, { "epoch": 1.78, "learning_rate": 3.495734206011684e-05, "loss": 0.8008, "step": 17620 }, { "epoch": 1.78, "learning_rate": 3.4954981775262824e-05, "loss": 0.23, "step": 17621 }, { "epoch": 1.78, "learning_rate": 3.49526214588772e-05, "loss": 1.0372, "step": 17622 }, { "epoch": 1.78, "learning_rate": 3.495026111097498e-05, "loss": 0.7797, "step": 17623 }, { "epoch": 1.78, "learning_rate": 3.4947900731571174e-05, "loss": 0.8014, "step": 17624 }, { "epoch": 1.78, "learning_rate": 3.494554032068081e-05, "loss": 0.6208, "step": 17625 }, { "epoch": 1.78, "learning_rate": 3.4943179878318906e-05, "loss": 0.4152, "step": 17626 }, { "epoch": 1.78, "learning_rate": 3.49408194045005e-05, "loss": 0.87, "step": 17627 }, { "epoch": 1.78, "learning_rate": 3.4938458899240585e-05, "loss": 1.1056, "step": 17628 }, { "epoch": 1.78, "learning_rate": 3.49360983625542e-05, "loss": 0.6009, "step": 17629 }, { "epoch": 1.78, "learning_rate": 3.493373779445636e-05, "loss": 0.1649, "step": 17630 }, { "epoch": 1.78, "learning_rate": 3.4931377194962095e-05, "loss": 0.5663, "step": 17631 }, { "epoch": 1.78, "learning_rate": 3.492901656408642e-05, "loss": 1.0068, "step": 17632 }, { "epoch": 1.78, "learning_rate": 3.4926655901844364e-05, "loss": 1.064, "step": 17633 }, { "epoch": 1.78, "learning_rate": 3.492429520825094e-05, "loss": 0.6934, "step": 17634 }, { "epoch": 1.78, "learning_rate": 3.492193448332116e-05, "loss": 0.4896, "step": 17635 }, { "epoch": 1.78, "learning_rate": 3.491957372707008e-05, "loss": 1.0217, "step": 17636 }, { "epoch": 1.78, "learning_rate": 3.4917212939512693e-05, "loss": 0.3604, "step": 17637 }, { "epoch": 1.78, "learning_rate": 3.491485212066404e-05, "loss": 0.3794, "step": 17638 }, { "epoch": 1.78, "learning_rate": 3.491249127053914e-05, "loss": 0.779, "step": 17639 }, { "epoch": 1.78, "learning_rate": 3.4910130389153016e-05, "loss": 0.3195, "step": 17640 }, { "epoch": 1.78, "learning_rate": 3.490776947652068e-05, "loss": 1.3233, "step": 17641 }, { "epoch": 1.78, "learning_rate": 3.490540853265717e-05, "loss": 1.3034, "step": 17642 }, { "epoch": 1.78, "learning_rate": 3.490304755757751e-05, "loss": 1.2224, "step": 17643 }, { "epoch": 1.78, "learning_rate": 3.490068655129672e-05, "loss": 0.7724, "step": 17644 }, { "epoch": 1.78, "learning_rate": 3.4898325513829816e-05, "loss": 0.2473, "step": 17645 }, { "epoch": 1.78, "learning_rate": 3.4895964445191844e-05, "loss": 0.5549, "step": 17646 }, { "epoch": 1.78, "learning_rate": 3.489360334539781e-05, "loss": 0.7871, "step": 17647 }, { "epoch": 1.78, "learning_rate": 3.489124221446275e-05, "loss": 0.8041, "step": 17648 }, { "epoch": 1.78, "learning_rate": 3.4888881052401685e-05, "loss": 0.7844, "step": 17649 }, { "epoch": 1.78, "learning_rate": 3.4886519859229636e-05, "loss": 1.2422, "step": 17650 }, { "epoch": 1.78, "learning_rate": 3.488415863496163e-05, "loss": 0.6942, "step": 17651 }, { "epoch": 1.78, "learning_rate": 3.488179737961271e-05, "loss": 1.394, "step": 17652 }, { "epoch": 1.78, "learning_rate": 3.487943609319788e-05, "loss": 0.5396, "step": 17653 }, { "epoch": 1.78, "learning_rate": 3.4877074775732174e-05, "loss": 0.4693, "step": 17654 }, { "epoch": 1.78, "learning_rate": 3.487471342723062e-05, "loss": 0.4564, "step": 17655 }, { "epoch": 1.78, "learning_rate": 3.487235204770825e-05, "loss": 0.4605, "step": 17656 }, { "epoch": 1.78, "learning_rate": 3.486999063718008e-05, "loss": 1.2734, "step": 17657 }, { "epoch": 1.78, "learning_rate": 3.4867629195661145e-05, "loss": 0.8844, "step": 17658 }, { "epoch": 1.78, "learning_rate": 3.486526772316648e-05, "loss": 1.3479, "step": 17659 }, { "epoch": 1.78, "learning_rate": 3.4862906219711074e-05, "loss": 0.9135, "step": 17660 }, { "epoch": 1.78, "learning_rate": 3.486054468531e-05, "loss": 1.3365, "step": 17661 }, { "epoch": 1.78, "learning_rate": 3.485818311997827e-05, "loss": 0.6214, "step": 17662 }, { "epoch": 1.78, "learning_rate": 3.485582152373091e-05, "loss": 0.3224, "step": 17663 }, { "epoch": 1.78, "learning_rate": 3.485345989658295e-05, "loss": 0.4376, "step": 17664 }, { "epoch": 1.78, "learning_rate": 3.485109823854941e-05, "loss": 0.6124, "step": 17665 }, { "epoch": 1.78, "learning_rate": 3.484873654964533e-05, "loss": 0.9527, "step": 17666 }, { "epoch": 1.78, "learning_rate": 3.4846374829885735e-05, "loss": 0.5639, "step": 17667 }, { "epoch": 1.78, "learning_rate": 3.4844013079285654e-05, "loss": 1.1562, "step": 17668 }, { "epoch": 1.78, "learning_rate": 3.484165129786012e-05, "loss": 0.8143, "step": 17669 }, { "epoch": 1.78, "learning_rate": 3.483928948562416e-05, "loss": 1.0595, "step": 17670 }, { "epoch": 1.78, "learning_rate": 3.4836927642592796e-05, "loss": 1.1369, "step": 17671 }, { "epoch": 1.78, "learning_rate": 3.483456576878107e-05, "loss": 0.4983, "step": 17672 }, { "epoch": 1.78, "learning_rate": 3.483220386420401e-05, "loss": 0.7705, "step": 17673 }, { "epoch": 1.78, "learning_rate": 3.4829841928876634e-05, "loss": 0.4412, "step": 17674 }, { "epoch": 1.78, "learning_rate": 3.482747996281399e-05, "loss": 0.8743, "step": 17675 }, { "epoch": 1.78, "learning_rate": 3.482511796603109e-05, "loss": 0.9253, "step": 17676 }, { "epoch": 1.78, "learning_rate": 3.482275593854298e-05, "loss": 1.0116, "step": 17677 }, { "epoch": 1.78, "learning_rate": 3.482039388036469e-05, "loss": 0.9432, "step": 17678 }, { "epoch": 1.78, "learning_rate": 3.481803179151124e-05, "loss": 0.4858, "step": 17679 }, { "epoch": 1.78, "learning_rate": 3.481566967199767e-05, "loss": 0.5907, "step": 17680 }, { "epoch": 1.78, "learning_rate": 3.4813307521839e-05, "loss": 0.7213, "step": 17681 }, { "epoch": 1.78, "learning_rate": 3.48109453410503e-05, "loss": 0.655, "step": 17682 }, { "epoch": 1.78, "learning_rate": 3.480858312964655e-05, "loss": 1.2024, "step": 17683 }, { "epoch": 1.78, "learning_rate": 3.480622088764282e-05, "loss": 0.5884, "step": 17684 }, { "epoch": 1.78, "learning_rate": 3.4803858615054115e-05, "loss": 0.2297, "step": 17685 }, { "epoch": 1.78, "learning_rate": 3.480149631189549e-05, "loss": 0.1242, "step": 17686 }, { "epoch": 1.78, "learning_rate": 3.4799133978181966e-05, "loss": 0.3032, "step": 17687 }, { "epoch": 1.78, "learning_rate": 3.4796771613928585e-05, "loss": 0.8472, "step": 17688 }, { "epoch": 1.78, "learning_rate": 3.479440921915037e-05, "loss": 0.5955, "step": 17689 }, { "epoch": 1.78, "learning_rate": 3.479204679386235e-05, "loss": 0.4601, "step": 17690 }, { "epoch": 1.78, "learning_rate": 3.478968433807958e-05, "loss": 0.5277, "step": 17691 }, { "epoch": 1.78, "learning_rate": 3.478732185181708e-05, "loss": 0.8628, "step": 17692 }, { "epoch": 1.78, "learning_rate": 3.478495933508988e-05, "loss": 0.4919, "step": 17693 }, { "epoch": 1.78, "learning_rate": 3.478259678791302e-05, "loss": 1.1153, "step": 17694 }, { "epoch": 1.78, "learning_rate": 3.478023421030153e-05, "loss": 1.0777, "step": 17695 }, { "epoch": 1.78, "learning_rate": 3.4777871602270456e-05, "loss": 0.4984, "step": 17696 }, { "epoch": 1.78, "learning_rate": 3.477550896383482e-05, "loss": 0.8828, "step": 17697 }, { "epoch": 1.78, "learning_rate": 3.477314629500967e-05, "loss": 0.117, "step": 17698 }, { "epoch": 1.78, "learning_rate": 3.477078359581002e-05, "loss": 0.5507, "step": 17699 }, { "epoch": 1.78, "learning_rate": 3.4768420866250924e-05, "loss": 0.4982, "step": 17700 }, { "epoch": 1.78, "learning_rate": 3.476605810634741e-05, "loss": 1.0102, "step": 17701 }, { "epoch": 1.78, "learning_rate": 3.476369531611452e-05, "loss": 0.3136, "step": 17702 }, { "epoch": 1.78, "learning_rate": 3.476133249556728e-05, "loss": 0.3728, "step": 17703 }, { "epoch": 1.78, "learning_rate": 3.4758969644720726e-05, "loss": 0.5805, "step": 17704 }, { "epoch": 1.78, "learning_rate": 3.475660676358992e-05, "loss": 0.5494, "step": 17705 }, { "epoch": 1.78, "learning_rate": 3.475424385218986e-05, "loss": 0.4645, "step": 17706 }, { "epoch": 1.79, "learning_rate": 3.47518809105356e-05, "loss": 1.128, "step": 17707 }, { "epoch": 1.79, "learning_rate": 3.4749517938642186e-05, "loss": 0.4268, "step": 17708 }, { "epoch": 1.79, "learning_rate": 3.474715493652464e-05, "loss": 0.9419, "step": 17709 }, { "epoch": 1.79, "learning_rate": 3.474479190419802e-05, "loss": 0.3831, "step": 17710 }, { "epoch": 1.79, "learning_rate": 3.474242884167733e-05, "loss": 1.0182, "step": 17711 }, { "epoch": 1.79, "learning_rate": 3.474006574897764e-05, "loss": 0.3185, "step": 17712 }, { "epoch": 1.79, "learning_rate": 3.4737702626113964e-05, "loss": 0.1947, "step": 17713 }, { "epoch": 1.79, "learning_rate": 3.473533947310136e-05, "loss": 0.4768, "step": 17714 }, { "epoch": 1.79, "learning_rate": 3.473297628995485e-05, "loss": 1.7282, "step": 17715 }, { "epoch": 1.79, "learning_rate": 3.473061307668949e-05, "loss": 0.6438, "step": 17716 }, { "epoch": 1.79, "learning_rate": 3.4728249833320295e-05, "loss": 0.3426, "step": 17717 }, { "epoch": 1.79, "learning_rate": 3.472588655986233e-05, "loss": 0.7005, "step": 17718 }, { "epoch": 1.79, "learning_rate": 3.472352325633062e-05, "loss": 0.2183, "step": 17719 }, { "epoch": 1.79, "learning_rate": 3.472115992274019e-05, "loss": 1.2386, "step": 17720 }, { "epoch": 1.79, "learning_rate": 3.471879655910611e-05, "loss": 1.044, "step": 17721 }, { "epoch": 1.79, "learning_rate": 3.4716433165443385e-05, "loss": 0.3204, "step": 17722 }, { "epoch": 1.79, "learning_rate": 3.471406974176709e-05, "loss": 0.6386, "step": 17723 }, { "epoch": 1.79, "learning_rate": 3.471170628809225e-05, "loss": 0.9585, "step": 17724 }, { "epoch": 1.79, "learning_rate": 3.4709342804433897e-05, "loss": 0.2914, "step": 17725 }, { "epoch": 1.79, "learning_rate": 3.470697929080708e-05, "loss": 0.9815, "step": 17726 }, { "epoch": 1.79, "learning_rate": 3.470461574722683e-05, "loss": 0.3593, "step": 17727 }, { "epoch": 1.79, "learning_rate": 3.470225217370821e-05, "loss": 0.3503, "step": 17728 }, { "epoch": 1.79, "learning_rate": 3.469988857026623e-05, "loss": 0.6855, "step": 17729 }, { "epoch": 1.79, "learning_rate": 3.469752493691596e-05, "loss": 1.1748, "step": 17730 }, { "epoch": 1.79, "learning_rate": 3.469516127367242e-05, "loss": 1.2157, "step": 17731 }, { "epoch": 1.79, "learning_rate": 3.469279758055066e-05, "loss": 0.3584, "step": 17732 }, { "epoch": 1.79, "learning_rate": 3.469043385756573e-05, "loss": 0.5909, "step": 17733 }, { "epoch": 1.79, "learning_rate": 3.468807010473266e-05, "loss": 0.7081, "step": 17734 }, { "epoch": 1.79, "learning_rate": 3.4685706322066486e-05, "loss": 0.6236, "step": 17735 }, { "epoch": 1.79, "learning_rate": 3.468334250958226e-05, "loss": 0.745, "step": 17736 }, { "epoch": 1.79, "learning_rate": 3.468097866729504e-05, "loss": 0.6885, "step": 17737 }, { "epoch": 1.79, "learning_rate": 3.4678614795219835e-05, "loss": 0.8959, "step": 17738 }, { "epoch": 1.79, "learning_rate": 3.4676250893371716e-05, "loss": 0.1853, "step": 17739 }, { "epoch": 1.79, "learning_rate": 3.467388696176571e-05, "loss": 1.1815, "step": 17740 }, { "epoch": 1.79, "learning_rate": 3.467152300041687e-05, "loss": 1.2496, "step": 17741 }, { "epoch": 1.79, "learning_rate": 3.466915900934024e-05, "loss": 1.238, "step": 17742 }, { "epoch": 1.79, "learning_rate": 3.466679498855084e-05, "loss": 0.9753, "step": 17743 }, { "epoch": 1.79, "learning_rate": 3.466443093806375e-05, "loss": 0.1859, "step": 17744 }, { "epoch": 1.79, "learning_rate": 3.466206685789398e-05, "loss": 1.5168, "step": 17745 }, { "epoch": 1.79, "learning_rate": 3.46597027480566e-05, "loss": 0.3457, "step": 17746 }, { "epoch": 1.79, "learning_rate": 3.465733860856664e-05, "loss": 1.5109, "step": 17747 }, { "epoch": 1.79, "learning_rate": 3.4654974439439156e-05, "loss": 0.5039, "step": 17748 }, { "epoch": 1.79, "learning_rate": 3.4652610240689184e-05, "loss": 0.6632, "step": 17749 }, { "epoch": 1.79, "learning_rate": 3.4650246012331765e-05, "loss": 1.1274, "step": 17750 }, { "epoch": 1.79, "learning_rate": 3.4647881754381955e-05, "loss": 0.2106, "step": 17751 }, { "epoch": 1.79, "learning_rate": 3.4645517466854784e-05, "loss": 1.2371, "step": 17752 }, { "epoch": 1.79, "learning_rate": 3.464315314976532e-05, "loss": 0.1375, "step": 17753 }, { "epoch": 1.79, "learning_rate": 3.464078880312858e-05, "loss": 0.5783, "step": 17754 }, { "epoch": 1.79, "learning_rate": 3.463842442695964e-05, "loss": 1.4744, "step": 17755 }, { "epoch": 1.79, "learning_rate": 3.4636060021273526e-05, "loss": 0.3148, "step": 17756 }, { "epoch": 1.79, "learning_rate": 3.463369558608529e-05, "loss": 0.6735, "step": 17757 }, { "epoch": 1.79, "learning_rate": 3.463133112140998e-05, "loss": 1.0927, "step": 17758 }, { "epoch": 1.79, "learning_rate": 3.462896662726263e-05, "loss": 0.3313, "step": 17759 }, { "epoch": 1.79, "learning_rate": 3.462660210365832e-05, "loss": 1.218, "step": 17760 }, { "epoch": 1.79, "learning_rate": 3.4624237550612054e-05, "loss": 0.5508, "step": 17761 }, { "epoch": 1.79, "learning_rate": 3.4621872968138904e-05, "loss": 0.3593, "step": 17762 }, { "epoch": 1.79, "learning_rate": 3.461950835625392e-05, "loss": 0.4363, "step": 17763 }, { "epoch": 1.79, "learning_rate": 3.461714371497214e-05, "loss": 0.4598, "step": 17764 }, { "epoch": 1.79, "learning_rate": 3.46147790443086e-05, "loss": 0.4449, "step": 17765 }, { "epoch": 1.79, "learning_rate": 3.461241434427838e-05, "loss": 0.4547, "step": 17766 }, { "epoch": 1.79, "learning_rate": 3.46100496148965e-05, "loss": 0.7423, "step": 17767 }, { "epoch": 1.79, "learning_rate": 3.460768485617802e-05, "loss": 0.5009, "step": 17768 }, { "epoch": 1.79, "learning_rate": 3.460532006813799e-05, "loss": 0.1255, "step": 17769 }, { "epoch": 1.79, "learning_rate": 3.4602955250791444e-05, "loss": 0.4058, "step": 17770 }, { "epoch": 1.79, "learning_rate": 3.4600590404153444e-05, "loss": 0.5307, "step": 17771 }, { "epoch": 1.79, "learning_rate": 3.4598225528239045e-05, "loss": 1.0293, "step": 17772 }, { "epoch": 1.79, "learning_rate": 3.459586062306329e-05, "loss": 0.2091, "step": 17773 }, { "epoch": 1.79, "learning_rate": 3.459349568864122e-05, "loss": 1.0566, "step": 17774 }, { "epoch": 1.79, "learning_rate": 3.459113072498789e-05, "loss": 0.1969, "step": 17775 }, { "epoch": 1.79, "learning_rate": 3.458876573211835e-05, "loss": 0.7671, "step": 17776 }, { "epoch": 1.79, "learning_rate": 3.458640071004765e-05, "loss": 1.314, "step": 17777 }, { "epoch": 1.79, "learning_rate": 3.458403565879085e-05, "loss": 0.3809, "step": 17778 }, { "epoch": 1.79, "learning_rate": 3.4581670578362986e-05, "loss": 0.2972, "step": 17779 }, { "epoch": 1.79, "learning_rate": 3.457930546877911e-05, "loss": 0.6437, "step": 17780 }, { "epoch": 1.79, "learning_rate": 3.4576940330054276e-05, "loss": 0.293, "step": 17781 }, { "epoch": 1.79, "learning_rate": 3.457457516220355e-05, "loss": 0.4388, "step": 17782 }, { "epoch": 1.79, "learning_rate": 3.457220996524195e-05, "loss": 0.2456, "step": 17783 }, { "epoch": 1.79, "learning_rate": 3.456984473918455e-05, "loss": 0.4211, "step": 17784 }, { "epoch": 1.79, "learning_rate": 3.456747948404639e-05, "loss": 0.4303, "step": 17785 }, { "epoch": 1.79, "learning_rate": 3.456511419984254e-05, "loss": 0.6141, "step": 17786 }, { "epoch": 1.79, "learning_rate": 3.4562748886588036e-05, "loss": 0.9173, "step": 17787 }, { "epoch": 1.79, "learning_rate": 3.4560383544297926e-05, "loss": 0.2671, "step": 17788 }, { "epoch": 1.79, "learning_rate": 3.455801817298728e-05, "loss": 0.2132, "step": 17789 }, { "epoch": 1.79, "learning_rate": 3.455565277267114e-05, "loss": 0.5787, "step": 17790 }, { "epoch": 1.79, "learning_rate": 3.4553287343364556e-05, "loss": 0.7939, "step": 17791 }, { "epoch": 1.79, "learning_rate": 3.455092188508259e-05, "loss": 0.3492, "step": 17792 }, { "epoch": 1.79, "learning_rate": 3.454855639784027e-05, "loss": 0.7908, "step": 17793 }, { "epoch": 1.79, "learning_rate": 3.4546190881652685e-05, "loss": 0.3011, "step": 17794 }, { "epoch": 1.79, "learning_rate": 3.4543825336534865e-05, "loss": 0.6931, "step": 17795 }, { "epoch": 1.79, "learning_rate": 3.454145976250187e-05, "loss": 0.3393, "step": 17796 }, { "epoch": 1.79, "learning_rate": 3.453909415956875e-05, "loss": 0.2606, "step": 17797 }, { "epoch": 1.79, "learning_rate": 3.453672852775056e-05, "loss": 0.551, "step": 17798 }, { "epoch": 1.79, "learning_rate": 3.4534362867062365e-05, "loss": 0.6279, "step": 17799 }, { "epoch": 1.79, "learning_rate": 3.45319971775192e-05, "loss": 0.461, "step": 17800 }, { "epoch": 1.79, "learning_rate": 3.452963145913613e-05, "loss": 0.3932, "step": 17801 }, { "epoch": 1.79, "learning_rate": 3.4527265711928205e-05, "loss": 0.457, "step": 17802 }, { "epoch": 1.79, "learning_rate": 3.45248999359105e-05, "loss": 0.4468, "step": 17803 }, { "epoch": 1.79, "learning_rate": 3.4522534131098034e-05, "loss": 0.92, "step": 17804 }, { "epoch": 1.79, "learning_rate": 3.45201682975059e-05, "loss": 0.5634, "step": 17805 }, { "epoch": 1.8, "learning_rate": 3.4517802435149125e-05, "loss": 0.6445, "step": 17806 }, { "epoch": 1.8, "learning_rate": 3.451543654404277e-05, "loss": 0.155, "step": 17807 }, { "epoch": 1.8, "learning_rate": 3.4513070624201894e-05, "loss": 0.4254, "step": 17808 }, { "epoch": 1.8, "learning_rate": 3.4510704675641555e-05, "loss": 0.2339, "step": 17809 }, { "epoch": 1.8, "learning_rate": 3.450833869837681e-05, "loss": 0.0913, "step": 17810 }, { "epoch": 1.8, "learning_rate": 3.450597269242272e-05, "loss": 0.4318, "step": 17811 }, { "epoch": 1.8, "learning_rate": 3.4503606657794324e-05, "loss": 1.0462, "step": 17812 }, { "epoch": 1.8, "learning_rate": 3.4501240594506686e-05, "loss": 0.8934, "step": 17813 }, { "epoch": 1.8, "learning_rate": 3.449887450257487e-05, "loss": 1.017, "step": 17814 }, { "epoch": 1.8, "learning_rate": 3.449650838201393e-05, "loss": 0.7043, "step": 17815 }, { "epoch": 1.8, "learning_rate": 3.4494142232838916e-05, "loss": 0.7773, "step": 17816 }, { "epoch": 1.8, "learning_rate": 3.4491776055064895e-05, "loss": 0.6392, "step": 17817 }, { "epoch": 1.8, "learning_rate": 3.448940984870692e-05, "loss": 0.5475, "step": 17818 }, { "epoch": 1.8, "learning_rate": 3.4487043613780045e-05, "loss": 1.0156, "step": 17819 }, { "epoch": 1.8, "learning_rate": 3.448467735029933e-05, "loss": 0.4506, "step": 17820 }, { "epoch": 1.8, "learning_rate": 3.448231105827985e-05, "loss": 0.6222, "step": 17821 }, { "epoch": 1.8, "learning_rate": 3.4479944737736636e-05, "loss": 0.4404, "step": 17822 }, { "epoch": 1.8, "learning_rate": 3.447757838868475e-05, "loss": 0.3647, "step": 17823 }, { "epoch": 1.8, "learning_rate": 3.4475212011139274e-05, "loss": 0.6273, "step": 17824 }, { "epoch": 1.8, "learning_rate": 3.447284560511524e-05, "loss": 0.8214, "step": 17825 }, { "epoch": 1.8, "learning_rate": 3.447047917062772e-05, "loss": 0.9619, "step": 17826 }, { "epoch": 1.8, "learning_rate": 3.446811270769176e-05, "loss": 0.3044, "step": 17827 }, { "epoch": 1.8, "learning_rate": 3.4465746216322454e-05, "loss": 0.7235, "step": 17828 }, { "epoch": 1.8, "learning_rate": 3.446337969653483e-05, "loss": 0.9801, "step": 17829 }, { "epoch": 1.8, "learning_rate": 3.4461013148343946e-05, "loss": 0.2668, "step": 17830 }, { "epoch": 1.8, "learning_rate": 3.4458646571764886e-05, "loss": 0.8037, "step": 17831 }, { "epoch": 1.8, "learning_rate": 3.445627996681268e-05, "loss": 0.693, "step": 17832 }, { "epoch": 1.8, "learning_rate": 3.445391333350241e-05, "loss": 0.8218, "step": 17833 }, { "epoch": 1.8, "learning_rate": 3.445154667184913e-05, "loss": 0.6417, "step": 17834 }, { "epoch": 1.8, "learning_rate": 3.4449179981867905e-05, "loss": 0.9832, "step": 17835 }, { "epoch": 1.8, "learning_rate": 3.444681326357378e-05, "loss": 0.2675, "step": 17836 }, { "epoch": 1.8, "learning_rate": 3.444444651698184e-05, "loss": 1.0333, "step": 17837 }, { "epoch": 1.8, "learning_rate": 3.444207974210713e-05, "loss": 0.8021, "step": 17838 }, { "epoch": 1.8, "learning_rate": 3.443971293896471e-05, "loss": 1.5117, "step": 17839 }, { "epoch": 1.8, "learning_rate": 3.443734610756964e-05, "loss": 0.4481, "step": 17840 }, { "epoch": 1.8, "learning_rate": 3.443497924793701e-05, "loss": 0.2139, "step": 17841 }, { "epoch": 1.8, "learning_rate": 3.4432612360081837e-05, "loss": 0.8811, "step": 17842 }, { "epoch": 1.8, "learning_rate": 3.443024544401921e-05, "loss": 0.5225, "step": 17843 }, { "epoch": 1.8, "learning_rate": 3.4427878499764193e-05, "loss": 0.8161, "step": 17844 }, { "epoch": 1.8, "learning_rate": 3.442551152733184e-05, "loss": 0.5911, "step": 17845 }, { "epoch": 1.8, "learning_rate": 3.44231445267372e-05, "loss": 0.2695, "step": 17846 }, { "epoch": 1.8, "learning_rate": 3.442077749799538e-05, "loss": 0.2845, "step": 17847 }, { "epoch": 1.8, "learning_rate": 3.441841044112139e-05, "loss": 0.618, "step": 17848 }, { "epoch": 1.8, "learning_rate": 3.441604335613032e-05, "loss": 0.5436, "step": 17849 }, { "epoch": 1.8, "learning_rate": 3.441367624303723e-05, "loss": 0.284, "step": 17850 }, { "epoch": 1.8, "learning_rate": 3.44113091018572e-05, "loss": 0.3901, "step": 17851 }, { "epoch": 1.8, "learning_rate": 3.440894193260526e-05, "loss": 0.4017, "step": 17852 }, { "epoch": 1.8, "learning_rate": 3.440657473529649e-05, "loss": 0.9998, "step": 17853 }, { "epoch": 1.8, "learning_rate": 3.440420750994597e-05, "loss": 0.8467, "step": 17854 }, { "epoch": 1.8, "learning_rate": 3.4401840256568735e-05, "loss": 0.7228, "step": 17855 }, { "epoch": 1.8, "learning_rate": 3.439947297517986e-05, "loss": 0.4037, "step": 17856 }, { "epoch": 1.8, "learning_rate": 3.439710566579442e-05, "loss": 0.3258, "step": 17857 }, { "epoch": 1.8, "learning_rate": 3.439473832842748e-05, "loss": 0.8045, "step": 17858 }, { "epoch": 1.8, "learning_rate": 3.439237096309408e-05, "loss": 0.764, "step": 17859 }, { "epoch": 1.8, "learning_rate": 3.439000356980932e-05, "loss": 0.446, "step": 17860 }, { "epoch": 1.8, "learning_rate": 3.4387636148588235e-05, "loss": 1.0702, "step": 17861 }, { "epoch": 1.8, "learning_rate": 3.43852686994459e-05, "loss": 0.8975, "step": 17862 }, { "epoch": 1.8, "learning_rate": 3.43829012223974e-05, "loss": 0.1562, "step": 17863 }, { "epoch": 1.8, "learning_rate": 3.4380533717457775e-05, "loss": 0.5173, "step": 17864 }, { "epoch": 1.8, "learning_rate": 3.437816618464209e-05, "loss": 0.1817, "step": 17865 }, { "epoch": 1.8, "learning_rate": 3.437579862396543e-05, "loss": 0.5754, "step": 17866 }, { "epoch": 1.8, "learning_rate": 3.437343103544286e-05, "loss": 1.2918, "step": 17867 }, { "epoch": 1.8, "learning_rate": 3.4371063419089424e-05, "loss": 0.3239, "step": 17868 }, { "epoch": 1.8, "learning_rate": 3.436869577492021e-05, "loss": 1.0416, "step": 17869 }, { "epoch": 1.8, "learning_rate": 3.436632810295028e-05, "loss": 0.3235, "step": 17870 }, { "epoch": 1.8, "learning_rate": 3.436396040319469e-05, "loss": 0.6869, "step": 17871 }, { "epoch": 1.8, "learning_rate": 3.436159267566853e-05, "loss": 0.8671, "step": 17872 }, { "epoch": 1.8, "learning_rate": 3.435922492038684e-05, "loss": 0.4973, "step": 17873 }, { "epoch": 1.8, "learning_rate": 3.4356857137364714e-05, "loss": 0.9733, "step": 17874 }, { "epoch": 1.8, "learning_rate": 3.435448932661719e-05, "loss": 0.613, "step": 17875 }, { "epoch": 1.8, "learning_rate": 3.4352121488159375e-05, "loss": 1.7579, "step": 17876 }, { "epoch": 1.8, "learning_rate": 3.4349753622006304e-05, "loss": 0.6541, "step": 17877 }, { "epoch": 1.8, "learning_rate": 3.4347385728173044e-05, "loss": 0.2371, "step": 17878 }, { "epoch": 1.8, "learning_rate": 3.434501780667469e-05, "loss": 0.1813, "step": 17879 }, { "epoch": 1.8, "learning_rate": 3.434264985752629e-05, "loss": 0.935, "step": 17880 }, { "epoch": 1.8, "learning_rate": 3.434028188074291e-05, "loss": 0.5612, "step": 17881 }, { "epoch": 1.8, "learning_rate": 3.4337913876339634e-05, "loss": 0.3699, "step": 17882 }, { "epoch": 1.8, "learning_rate": 3.433554584433153e-05, "loss": 0.5783, "step": 17883 }, { "epoch": 1.8, "learning_rate": 3.433317778473365e-05, "loss": 0.8301, "step": 17884 }, { "epoch": 1.8, "learning_rate": 3.4330809697561086e-05, "loss": 0.7832, "step": 17885 }, { "epoch": 1.8, "learning_rate": 3.43284415828289e-05, "loss": 0.5057, "step": 17886 }, { "epoch": 1.8, "learning_rate": 3.432607344055215e-05, "loss": 1.0673, "step": 17887 }, { "epoch": 1.8, "learning_rate": 3.4323705270745915e-05, "loss": 0.8728, "step": 17888 }, { "epoch": 1.8, "learning_rate": 3.432133707342526e-05, "loss": 0.3788, "step": 17889 }, { "epoch": 1.8, "learning_rate": 3.4318968848605264e-05, "loss": 0.2331, "step": 17890 }, { "epoch": 1.8, "learning_rate": 3.431660059630099e-05, "loss": 0.5356, "step": 17891 }, { "epoch": 1.8, "learning_rate": 3.431423231652753e-05, "loss": 0.1833, "step": 17892 }, { "epoch": 1.8, "learning_rate": 3.431186400929992e-05, "loss": 0.6072, "step": 17893 }, { "epoch": 1.8, "learning_rate": 3.430949567463324e-05, "loss": 0.6733, "step": 17894 }, { "epoch": 1.8, "learning_rate": 3.430712731254259e-05, "loss": 0.6579, "step": 17895 }, { "epoch": 1.8, "learning_rate": 3.430475892304301e-05, "loss": 0.1315, "step": 17896 }, { "epoch": 1.8, "learning_rate": 3.4302390506149584e-05, "loss": 0.8735, "step": 17897 }, { "epoch": 1.8, "learning_rate": 3.430002206187737e-05, "loss": 0.2631, "step": 17898 }, { "epoch": 1.8, "learning_rate": 3.429765359024147e-05, "loss": 0.2629, "step": 17899 }, { "epoch": 1.8, "learning_rate": 3.429528509125692e-05, "loss": 0.478, "step": 17900 }, { "epoch": 1.8, "learning_rate": 3.429291656493882e-05, "loss": 0.2787, "step": 17901 }, { "epoch": 1.8, "learning_rate": 3.4290548011302236e-05, "loss": 0.3903, "step": 17902 }, { "epoch": 1.8, "learning_rate": 3.4288179430362225e-05, "loss": 0.2138, "step": 17903 }, { "epoch": 1.81, "learning_rate": 3.4285810822133875e-05, "loss": 1.0327, "step": 17904 }, { "epoch": 1.81, "learning_rate": 3.428344218663226e-05, "loss": 0.647, "step": 17905 }, { "epoch": 1.81, "learning_rate": 3.428107352387245e-05, "loss": 1.019, "step": 17906 }, { "epoch": 1.81, "learning_rate": 3.4278704833869505e-05, "loss": 0.58, "step": 17907 }, { "epoch": 1.81, "learning_rate": 3.427633611663852e-05, "loss": 0.7786, "step": 17908 }, { "epoch": 1.81, "learning_rate": 3.427396737219456e-05, "loss": 1.3898, "step": 17909 }, { "epoch": 1.81, "learning_rate": 3.427159860055269e-05, "loss": 0.7652, "step": 17910 }, { "epoch": 1.81, "learning_rate": 3.4269229801727994e-05, "loss": 0.1927, "step": 17911 }, { "epoch": 1.81, "learning_rate": 3.426686097573555e-05, "loss": 0.3309, "step": 17912 }, { "epoch": 1.81, "learning_rate": 3.426449212259042e-05, "loss": 0.3417, "step": 17913 }, { "epoch": 1.81, "learning_rate": 3.4262123242307674e-05, "loss": 1.1543, "step": 17914 }, { "epoch": 1.81, "learning_rate": 3.425975433490242e-05, "loss": 0.3248, "step": 17915 }, { "epoch": 1.81, "learning_rate": 3.4257385400389697e-05, "loss": 0.5696, "step": 17916 }, { "epoch": 1.81, "learning_rate": 3.4255016438784595e-05, "loss": 0.2605, "step": 17917 }, { "epoch": 1.81, "learning_rate": 3.425264745010219e-05, "loss": 0.3027, "step": 17918 }, { "epoch": 1.81, "learning_rate": 3.425027843435754e-05, "loss": 1.0538, "step": 17919 }, { "epoch": 1.81, "learning_rate": 3.424790939156575e-05, "loss": 1.4559, "step": 17920 }, { "epoch": 1.81, "learning_rate": 3.424554032174187e-05, "loss": 0.6481, "step": 17921 }, { "epoch": 1.81, "learning_rate": 3.4243171224901e-05, "loss": 0.4233, "step": 17922 }, { "epoch": 1.81, "learning_rate": 3.424080210105819e-05, "loss": 0.8338, "step": 17923 }, { "epoch": 1.81, "learning_rate": 3.423843295022854e-05, "loss": 0.2992, "step": 17924 }, { "epoch": 1.81, "learning_rate": 3.423606377242711e-05, "loss": 0.5023, "step": 17925 }, { "epoch": 1.81, "learning_rate": 3.423369456766898e-05, "loss": 1.4432, "step": 17926 }, { "epoch": 1.81, "learning_rate": 3.423132533596923e-05, "loss": 0.5495, "step": 17927 }, { "epoch": 1.81, "learning_rate": 3.422895607734294e-05, "loss": 0.9701, "step": 17928 }, { "epoch": 1.81, "learning_rate": 3.4226586791805176e-05, "loss": 0.9123, "step": 17929 }, { "epoch": 1.81, "learning_rate": 3.4224217479371015e-05, "loss": 0.9893, "step": 17930 }, { "epoch": 1.81, "learning_rate": 3.422184814005555e-05, "loss": 0.8252, "step": 17931 }, { "epoch": 1.81, "learning_rate": 3.4219478773873844e-05, "loss": 0.6288, "step": 17932 }, { "epoch": 1.81, "learning_rate": 3.421710938084099e-05, "loss": 0.4034, "step": 17933 }, { "epoch": 1.81, "learning_rate": 3.421473996097205e-05, "loss": 0.5451, "step": 17934 }, { "epoch": 1.81, "learning_rate": 3.421237051428211e-05, "loss": 0.7544, "step": 17935 }, { "epoch": 1.81, "learning_rate": 3.421000104078624e-05, "loss": 0.244, "step": 17936 }, { "epoch": 1.81, "learning_rate": 3.420763154049953e-05, "loss": 0.6591, "step": 17937 }, { "epoch": 1.81, "learning_rate": 3.420526201343706e-05, "loss": 0.2123, "step": 17938 }, { "epoch": 1.81, "learning_rate": 3.420289245961388e-05, "loss": 0.5724, "step": 17939 }, { "epoch": 1.81, "learning_rate": 3.420052287904511e-05, "loss": 0.0968, "step": 17940 }, { "epoch": 1.81, "learning_rate": 3.419815327174582e-05, "loss": 0.5227, "step": 17941 }, { "epoch": 1.81, "learning_rate": 3.4195783637731054e-05, "loss": 0.3772, "step": 17942 }, { "epoch": 1.81, "learning_rate": 3.419341397701593e-05, "loss": 1.072, "step": 17943 }, { "epoch": 1.81, "learning_rate": 3.419104428961552e-05, "loss": 1.1896, "step": 17944 }, { "epoch": 1.81, "learning_rate": 3.41886745755449e-05, "loss": 0.448, "step": 17945 }, { "epoch": 1.81, "learning_rate": 3.418630483481913e-05, "loss": 0.1613, "step": 17946 }, { "epoch": 1.81, "learning_rate": 3.4183935067453325e-05, "loss": 0.4352, "step": 17947 }, { "epoch": 1.81, "learning_rate": 3.418156527346254e-05, "loss": 0.9389, "step": 17948 }, { "epoch": 1.81, "learning_rate": 3.417919545286187e-05, "loss": 0.2925, "step": 17949 }, { "epoch": 1.81, "learning_rate": 3.4176825605666385e-05, "loss": 1.0607, "step": 17950 }, { "epoch": 1.81, "learning_rate": 3.417445573189118e-05, "loss": 0.5643, "step": 17951 }, { "epoch": 1.81, "learning_rate": 3.417208583155132e-05, "loss": 1.1082, "step": 17952 }, { "epoch": 1.81, "learning_rate": 3.4169715904661884e-05, "loss": 0.3944, "step": 17953 }, { "epoch": 1.81, "learning_rate": 3.416734595123799e-05, "loss": 0.4168, "step": 17954 }, { "epoch": 1.81, "learning_rate": 3.416497597129466e-05, "loss": 0.5783, "step": 17955 }, { "epoch": 1.81, "learning_rate": 3.416260596484703e-05, "loss": 1.1867, "step": 17956 }, { "epoch": 1.81, "learning_rate": 3.416023593191015e-05, "loss": 0.4656, "step": 17957 }, { "epoch": 1.81, "learning_rate": 3.415786587249911e-05, "loss": 1.1356, "step": 17958 }, { "epoch": 1.81, "learning_rate": 3.4155495786628996e-05, "loss": 0.6488, "step": 17959 }, { "epoch": 1.81, "learning_rate": 3.4153125674314884e-05, "loss": 0.9506, "step": 17960 }, { "epoch": 1.81, "learning_rate": 3.415075553557186e-05, "loss": 0.9994, "step": 17961 }, { "epoch": 1.81, "learning_rate": 3.414838537041501e-05, "loss": 0.2961, "step": 17962 }, { "epoch": 1.81, "learning_rate": 3.414601517885941e-05, "loss": 0.7156, "step": 17963 }, { "epoch": 1.81, "learning_rate": 3.4143644960920156e-05, "loss": 0.5848, "step": 17964 }, { "epoch": 1.81, "learning_rate": 3.4141274716612316e-05, "loss": 0.5991, "step": 17965 }, { "epoch": 1.81, "learning_rate": 3.413890444595098e-05, "loss": 0.9717, "step": 17966 }, { "epoch": 1.81, "learning_rate": 3.4136534148951225e-05, "loss": 0.345, "step": 17967 }, { "epoch": 1.81, "learning_rate": 3.413416382562815e-05, "loss": 1.3003, "step": 17968 }, { "epoch": 1.81, "learning_rate": 3.413179347599681e-05, "loss": 0.541, "step": 17969 }, { "epoch": 1.81, "learning_rate": 3.412942310007234e-05, "loss": 0.1495, "step": 17970 }, { "epoch": 1.81, "learning_rate": 3.4127052697869764e-05, "loss": 0.7714, "step": 17971 }, { "epoch": 1.81, "learning_rate": 3.412468226940421e-05, "loss": 0.2097, "step": 17972 }, { "epoch": 1.81, "learning_rate": 3.412231181469074e-05, "loss": 0.7449, "step": 17973 }, { "epoch": 1.81, "learning_rate": 3.411994133374445e-05, "loss": 0.5459, "step": 17974 }, { "epoch": 1.81, "learning_rate": 3.411757082658041e-05, "loss": 0.7733, "step": 17975 }, { "epoch": 1.81, "learning_rate": 3.4115200293213724e-05, "loss": 1.3238, "step": 17976 }, { "epoch": 1.81, "learning_rate": 3.4112829733659474e-05, "loss": 0.4018, "step": 17977 }, { "epoch": 1.81, "learning_rate": 3.411045914793273e-05, "loss": 0.5689, "step": 17978 }, { "epoch": 1.81, "learning_rate": 3.4108088536048595e-05, "loss": 0.8521, "step": 17979 }, { "epoch": 1.81, "learning_rate": 3.410571789802214e-05, "loss": 0.6186, "step": 17980 }, { "epoch": 1.81, "learning_rate": 3.4103347233868464e-05, "loss": 1.141, "step": 17981 }, { "epoch": 1.81, "learning_rate": 3.410097654360265e-05, "loss": 0.6058, "step": 17982 }, { "epoch": 1.81, "learning_rate": 3.409860582723978e-05, "loss": 0.1738, "step": 17983 }, { "epoch": 1.81, "learning_rate": 3.409623508479494e-05, "loss": 0.3831, "step": 17984 }, { "epoch": 1.81, "learning_rate": 3.409386431628321e-05, "loss": 1.4643, "step": 17985 }, { "epoch": 1.81, "learning_rate": 3.40914935217197e-05, "loss": 0.377, "step": 17986 }, { "epoch": 1.81, "learning_rate": 3.408912270111947e-05, "loss": 0.5981, "step": 17987 }, { "epoch": 1.81, "learning_rate": 3.4086751854497625e-05, "loss": 1.177, "step": 17988 }, { "epoch": 1.81, "learning_rate": 3.408438098186924e-05, "loss": 0.4772, "step": 17989 }, { "epoch": 1.81, "learning_rate": 3.408201008324942e-05, "loss": 0.3671, "step": 17990 }, { "epoch": 1.81, "learning_rate": 3.4079639158653226e-05, "loss": 0.6405, "step": 17991 }, { "epoch": 1.81, "learning_rate": 3.407726820809577e-05, "loss": 0.359, "step": 17992 }, { "epoch": 1.81, "learning_rate": 3.4074897231592136e-05, "loss": 0.8078, "step": 17993 }, { "epoch": 1.81, "learning_rate": 3.4072526229157385e-05, "loss": 0.4841, "step": 17994 }, { "epoch": 1.81, "learning_rate": 3.407015520080664e-05, "loss": 0.6277, "step": 17995 }, { "epoch": 1.81, "learning_rate": 3.406778414655498e-05, "loss": 0.9113, "step": 17996 }, { "epoch": 1.81, "learning_rate": 3.406541306641748e-05, "loss": 0.3117, "step": 17997 }, { "epoch": 1.81, "learning_rate": 3.4063041960409244e-05, "loss": 0.2551, "step": 17998 }, { "epoch": 1.81, "learning_rate": 3.406067082854535e-05, "loss": 1.0233, "step": 17999 }, { "epoch": 1.81, "learning_rate": 3.4058299670840895e-05, "loss": 0.8486, "step": 18000 }, { "epoch": 1.81, "learning_rate": 3.405592848731096e-05, "loss": 0.2787, "step": 18001 }, { "epoch": 1.81, "learning_rate": 3.4053557277970644e-05, "loss": 0.4283, "step": 18002 }, { "epoch": 1.82, "learning_rate": 3.405118604283503e-05, "loss": 0.1776, "step": 18003 }, { "epoch": 1.82, "learning_rate": 3.404881478191921e-05, "loss": 0.2975, "step": 18004 }, { "epoch": 1.82, "learning_rate": 3.4046443495238274e-05, "loss": 0.4619, "step": 18005 }, { "epoch": 1.82, "learning_rate": 3.404407218280731e-05, "loss": 0.8351, "step": 18006 }, { "epoch": 1.82, "learning_rate": 3.404170084464141e-05, "loss": 0.8028, "step": 18007 }, { "epoch": 1.82, "learning_rate": 3.403932948075566e-05, "loss": 0.8719, "step": 18008 }, { "epoch": 1.82, "learning_rate": 3.403695809116516e-05, "loss": 0.5494, "step": 18009 }, { "epoch": 1.82, "learning_rate": 3.403458667588498e-05, "loss": 1.2085, "step": 18010 }, { "epoch": 1.82, "learning_rate": 3.4032215234930236e-05, "loss": 0.9666, "step": 18011 }, { "epoch": 1.82, "learning_rate": 3.402984376831601e-05, "loss": 0.7111, "step": 18012 }, { "epoch": 1.82, "learning_rate": 3.402747227605739e-05, "loss": 0.2166, "step": 18013 }, { "epoch": 1.82, "learning_rate": 3.402510075816947e-05, "loss": 0.4557, "step": 18014 }, { "epoch": 1.82, "learning_rate": 3.402272921466733e-05, "loss": 0.6348, "step": 18015 }, { "epoch": 1.82, "learning_rate": 3.4020357645566086e-05, "loss": 0.7498, "step": 18016 }, { "epoch": 1.82, "learning_rate": 3.40179860508808e-05, "loss": 0.6984, "step": 18017 }, { "epoch": 1.82, "learning_rate": 3.401561443062659e-05, "loss": 0.7025, "step": 18018 }, { "epoch": 1.82, "learning_rate": 3.4013242784818536e-05, "loss": 0.4229, "step": 18019 }, { "epoch": 1.82, "learning_rate": 3.401087111347173e-05, "loss": 1.04, "step": 18020 }, { "epoch": 1.82, "learning_rate": 3.4008499416601265e-05, "loss": 1.2361, "step": 18021 }, { "epoch": 1.82, "learning_rate": 3.400612769422224e-05, "loss": 0.1259, "step": 18022 }, { "epoch": 1.82, "learning_rate": 3.400375594634974e-05, "loss": 0.6748, "step": 18023 }, { "epoch": 1.82, "learning_rate": 3.4001384172998855e-05, "loss": 0.1165, "step": 18024 }, { "epoch": 1.82, "learning_rate": 3.399901237418469e-05, "loss": 0.5324, "step": 18025 }, { "epoch": 1.82, "learning_rate": 3.3996640549922326e-05, "loss": 0.1619, "step": 18026 }, { "epoch": 1.82, "learning_rate": 3.3994268700226864e-05, "loss": 1.2126, "step": 18027 }, { "epoch": 1.82, "learning_rate": 3.39918968251134e-05, "loss": 0.7412, "step": 18028 }, { "epoch": 1.82, "learning_rate": 3.398952492459702e-05, "loss": 0.5288, "step": 18029 }, { "epoch": 1.82, "learning_rate": 3.3987152998692815e-05, "loss": 0.5205, "step": 18030 }, { "epoch": 1.82, "learning_rate": 3.398478104741589e-05, "loss": 0.4315, "step": 18031 }, { "epoch": 1.82, "learning_rate": 3.3982409070781346e-05, "loss": 0.9216, "step": 18032 }, { "epoch": 1.82, "learning_rate": 3.398003706880424e-05, "loss": 0.7133, "step": 18033 }, { "epoch": 1.82, "learning_rate": 3.397766504149971e-05, "loss": 0.3234, "step": 18034 }, { "epoch": 1.82, "learning_rate": 3.397529298888283e-05, "loss": 0.9692, "step": 18035 }, { "epoch": 1.82, "learning_rate": 3.39729209109687e-05, "loss": 0.5308, "step": 18036 }, { "epoch": 1.82, "learning_rate": 3.397054880777241e-05, "loss": 0.735, "step": 18037 }, { "epoch": 1.82, "learning_rate": 3.396817667930906e-05, "loss": 0.23, "step": 18038 }, { "epoch": 1.82, "learning_rate": 3.396580452559374e-05, "loss": 1.2203, "step": 18039 }, { "epoch": 1.82, "learning_rate": 3.396343234664155e-05, "loss": 0.3997, "step": 18040 }, { "epoch": 1.82, "learning_rate": 3.396106014246758e-05, "loss": 0.5607, "step": 18041 }, { "epoch": 1.82, "learning_rate": 3.395868791308693e-05, "loss": 0.6174, "step": 18042 }, { "epoch": 1.82, "learning_rate": 3.3956315658514706e-05, "loss": 0.9532, "step": 18043 }, { "epoch": 1.82, "learning_rate": 3.3953943378765986e-05, "loss": 0.294, "step": 18044 }, { "epoch": 1.82, "learning_rate": 3.3951571073855875e-05, "loss": 0.45, "step": 18045 }, { "epoch": 1.82, "learning_rate": 3.394919874379948e-05, "loss": 0.3207, "step": 18046 }, { "epoch": 1.82, "learning_rate": 3.394682638861187e-05, "loss": 0.9323, "step": 18047 }, { "epoch": 1.82, "learning_rate": 3.394445400830817e-05, "loss": 0.9016, "step": 18048 }, { "epoch": 1.82, "learning_rate": 3.394208160290345e-05, "loss": 0.8224, "step": 18049 }, { "epoch": 1.82, "learning_rate": 3.393970917241284e-05, "loss": 0.2095, "step": 18050 }, { "epoch": 1.82, "learning_rate": 3.393733671685141e-05, "loss": 0.4691, "step": 18051 }, { "epoch": 1.82, "learning_rate": 3.393496423623427e-05, "loss": 0.2609, "step": 18052 }, { "epoch": 1.82, "learning_rate": 3.393259173057652e-05, "loss": 0.3313, "step": 18053 }, { "epoch": 1.82, "learning_rate": 3.3930219199893244e-05, "loss": 0.6821, "step": 18054 }, { "epoch": 1.82, "learning_rate": 3.392784664419955e-05, "loss": 0.6271, "step": 18055 }, { "epoch": 1.82, "learning_rate": 3.392547406351053e-05, "loss": 0.396, "step": 18056 }, { "epoch": 1.82, "learning_rate": 3.392310145784129e-05, "loss": 0.2304, "step": 18057 }, { "epoch": 1.82, "learning_rate": 3.392072882720693e-05, "loss": 0.3073, "step": 18058 }, { "epoch": 1.82, "learning_rate": 3.391835617162254e-05, "loss": 1.2134, "step": 18059 }, { "epoch": 1.82, "learning_rate": 3.3915983491103217e-05, "loss": 0.3582, "step": 18060 }, { "epoch": 1.82, "learning_rate": 3.3913610785664076e-05, "loss": 0.4118, "step": 18061 }, { "epoch": 1.82, "learning_rate": 3.39112380553202e-05, "loss": 0.5206, "step": 18062 }, { "epoch": 1.82, "learning_rate": 3.390886530008668e-05, "loss": 0.8247, "step": 18063 }, { "epoch": 1.82, "learning_rate": 3.390649251997865e-05, "loss": 0.7198, "step": 18064 }, { "epoch": 1.82, "learning_rate": 3.390411971501117e-05, "loss": 0.9356, "step": 18065 }, { "epoch": 1.82, "learning_rate": 3.3901746885199364e-05, "loss": 0.7229, "step": 18066 }, { "epoch": 1.82, "learning_rate": 3.3899374030558325e-05, "loss": 0.7825, "step": 18067 }, { "epoch": 1.82, "learning_rate": 3.389700115110315e-05, "loss": 0.431, "step": 18068 }, { "epoch": 1.82, "learning_rate": 3.389462824684895e-05, "loss": 0.2131, "step": 18069 }, { "epoch": 1.82, "learning_rate": 3.3892255317810805e-05, "loss": 0.4692, "step": 18070 }, { "epoch": 1.82, "learning_rate": 3.388988236400385e-05, "loss": 0.209, "step": 18071 }, { "epoch": 1.82, "learning_rate": 3.388750938544314e-05, "loss": 0.6597, "step": 18072 }, { "epoch": 1.82, "learning_rate": 3.388513638214381e-05, "loss": 0.2929, "step": 18073 }, { "epoch": 1.82, "learning_rate": 3.388276335412095e-05, "loss": 0.4959, "step": 18074 }, { "epoch": 1.82, "learning_rate": 3.388039030138965e-05, "loss": 0.2665, "step": 18075 }, { "epoch": 1.82, "learning_rate": 3.387801722396504e-05, "loss": 1.8311, "step": 18076 }, { "epoch": 1.82, "learning_rate": 3.3875644121862195e-05, "loss": 1.0672, "step": 18077 }, { "epoch": 1.82, "learning_rate": 3.387327099509622e-05, "loss": 0.8345, "step": 18078 }, { "epoch": 1.82, "learning_rate": 3.387089784368223e-05, "loss": 0.4611, "step": 18079 }, { "epoch": 1.82, "learning_rate": 3.386852466763532e-05, "loss": 0.4249, "step": 18080 }, { "epoch": 1.82, "learning_rate": 3.386615146697058e-05, "loss": 0.888, "step": 18081 }, { "epoch": 1.82, "learning_rate": 3.386377824170313e-05, "loss": 0.2666, "step": 18082 }, { "epoch": 1.82, "learning_rate": 3.386140499184807e-05, "loss": 0.28, "step": 18083 }, { "epoch": 1.82, "learning_rate": 3.3859031717420495e-05, "loss": 0.5933, "step": 18084 }, { "epoch": 1.82, "learning_rate": 3.3856658418435505e-05, "loss": 0.2283, "step": 18085 }, { "epoch": 1.82, "learning_rate": 3.38542850949082e-05, "loss": 0.4374, "step": 18086 }, { "epoch": 1.82, "learning_rate": 3.3851911746853714e-05, "loss": 0.9247, "step": 18087 }, { "epoch": 1.82, "learning_rate": 3.384953837428711e-05, "loss": 0.8365, "step": 18088 }, { "epoch": 1.82, "learning_rate": 3.384716497722351e-05, "loss": 0.9747, "step": 18089 }, { "epoch": 1.82, "learning_rate": 3.384479155567802e-05, "loss": 0.1406, "step": 18090 }, { "epoch": 1.82, "learning_rate": 3.384241810966573e-05, "loss": 0.2426, "step": 18091 }, { "epoch": 1.82, "learning_rate": 3.384004463920177e-05, "loss": 1.215, "step": 18092 }, { "epoch": 1.82, "learning_rate": 3.3837671144301214e-05, "loss": 1.1262, "step": 18093 }, { "epoch": 1.82, "learning_rate": 3.383529762497918e-05, "loss": 0.1455, "step": 18094 }, { "epoch": 1.82, "learning_rate": 3.383292408125077e-05, "loss": 0.3743, "step": 18095 }, { "epoch": 1.82, "learning_rate": 3.38305505131311e-05, "loss": 0.3507, "step": 18096 }, { "epoch": 1.82, "learning_rate": 3.382817692063525e-05, "loss": 0.4994, "step": 18097 }, { "epoch": 1.82, "learning_rate": 3.382580330377834e-05, "loss": 0.3189, "step": 18098 }, { "epoch": 1.82, "learning_rate": 3.382342966257548e-05, "loss": 0.4763, "step": 18099 }, { "epoch": 1.82, "learning_rate": 3.382105599704177e-05, "loss": 0.5857, "step": 18100 }, { "epoch": 1.83, "learning_rate": 3.381868230719231e-05, "loss": 1.3096, "step": 18101 }, { "epoch": 1.83, "learning_rate": 3.38163085930422e-05, "loss": 1.0995, "step": 18102 }, { "epoch": 1.83, "learning_rate": 3.3813934854606565e-05, "loss": 0.0603, "step": 18103 }, { "epoch": 1.83, "learning_rate": 3.3811561091900486e-05, "loss": 0.0875, "step": 18104 }, { "epoch": 1.83, "learning_rate": 3.38091873049391e-05, "loss": 0.8106, "step": 18105 }, { "epoch": 1.83, "learning_rate": 3.380681349373749e-05, "loss": 0.3046, "step": 18106 }, { "epoch": 1.83, "learning_rate": 3.380443965831076e-05, "loss": 0.8264, "step": 18107 }, { "epoch": 1.83, "learning_rate": 3.3802065798674024e-05, "loss": 0.3135, "step": 18108 }, { "epoch": 1.83, "learning_rate": 3.3799691914842397e-05, "loss": 0.6025, "step": 18109 }, { "epoch": 1.83, "learning_rate": 3.3797318006830973e-05, "loss": 0.3069, "step": 18110 }, { "epoch": 1.83, "learning_rate": 3.3794944074654846e-05, "loss": 0.9242, "step": 18111 }, { "epoch": 1.83, "learning_rate": 3.379257011832916e-05, "loss": 0.5763, "step": 18112 }, { "epoch": 1.83, "learning_rate": 3.379019613786899e-05, "loss": 0.5597, "step": 18113 }, { "epoch": 1.83, "learning_rate": 3.378782213328946e-05, "loss": 0.4286, "step": 18114 }, { "epoch": 1.83, "learning_rate": 3.378544810460566e-05, "loss": 0.5666, "step": 18115 }, { "epoch": 1.83, "learning_rate": 3.378307405183272e-05, "loss": 0.9552, "step": 18116 }, { "epoch": 1.83, "learning_rate": 3.3780699974985735e-05, "loss": 0.5523, "step": 18117 }, { "epoch": 1.83, "learning_rate": 3.37783258740798e-05, "loss": 0.9569, "step": 18118 }, { "epoch": 1.83, "learning_rate": 3.377595174913005e-05, "loss": 0.7315, "step": 18119 }, { "epoch": 1.83, "learning_rate": 3.377357760015157e-05, "loss": 0.7995, "step": 18120 }, { "epoch": 1.83, "learning_rate": 3.377120342715948e-05, "loss": 0.0978, "step": 18121 }, { "epoch": 1.83, "learning_rate": 3.376882923016889e-05, "loss": 1.0755, "step": 18122 }, { "epoch": 1.83, "learning_rate": 3.376645500919491e-05, "loss": 1.1113, "step": 18123 }, { "epoch": 1.83, "learning_rate": 3.376408076425263e-05, "loss": 0.3692, "step": 18124 }, { "epoch": 1.83, "learning_rate": 3.3761706495357175e-05, "loss": 0.2839, "step": 18125 }, { "epoch": 1.83, "learning_rate": 3.375933220252366e-05, "loss": 0.3288, "step": 18126 }, { "epoch": 1.83, "learning_rate": 3.375695788576718e-05, "loss": 0.9901, "step": 18127 }, { "epoch": 1.83, "learning_rate": 3.3754583545102846e-05, "loss": 0.8086, "step": 18128 }, { "epoch": 1.83, "learning_rate": 3.3752209180545774e-05, "loss": 1.0365, "step": 18129 }, { "epoch": 1.83, "learning_rate": 3.374983479211107e-05, "loss": 0.4173, "step": 18130 }, { "epoch": 1.83, "learning_rate": 3.374746037981385e-05, "loss": 0.5401, "step": 18131 }, { "epoch": 1.83, "learning_rate": 3.3745085943669204e-05, "loss": 1.0274, "step": 18132 }, { "epoch": 1.83, "learning_rate": 3.3742711483692264e-05, "loss": 0.7525, "step": 18133 }, { "epoch": 1.83, "learning_rate": 3.3740336999898126e-05, "loss": 0.8705, "step": 18134 }, { "epoch": 1.83, "learning_rate": 3.373796249230192e-05, "loss": 0.139, "step": 18135 }, { "epoch": 1.83, "learning_rate": 3.373558796091873e-05, "loss": 0.8021, "step": 18136 }, { "epoch": 1.83, "learning_rate": 3.3733213405763686e-05, "loss": 1.0453, "step": 18137 }, { "epoch": 1.83, "learning_rate": 3.373083882685188e-05, "loss": 0.6619, "step": 18138 }, { "epoch": 1.83, "learning_rate": 3.372846422419845e-05, "loss": 0.4516, "step": 18139 }, { "epoch": 1.83, "learning_rate": 3.3726089597818485e-05, "loss": 0.1888, "step": 18140 }, { "epoch": 1.83, "learning_rate": 3.37237149477271e-05, "loss": 1.188, "step": 18141 }, { "epoch": 1.83, "learning_rate": 3.372134027393942e-05, "loss": 0.7258, "step": 18142 }, { "epoch": 1.83, "learning_rate": 3.371896557647054e-05, "loss": 0.6978, "step": 18143 }, { "epoch": 1.83, "learning_rate": 3.371659085533558e-05, "loss": 0.4947, "step": 18144 }, { "epoch": 1.83, "learning_rate": 3.371421611054964e-05, "loss": 1.1396, "step": 18145 }, { "epoch": 1.83, "learning_rate": 3.371184134212785e-05, "loss": 0.525, "step": 18146 }, { "epoch": 1.83, "learning_rate": 3.370946655008532e-05, "loss": 0.1658, "step": 18147 }, { "epoch": 1.83, "learning_rate": 3.3707091734437145e-05, "loss": 0.8957, "step": 18148 }, { "epoch": 1.83, "learning_rate": 3.370471689519846e-05, "loss": 0.3484, "step": 18149 }, { "epoch": 1.83, "learning_rate": 3.3702342032384345e-05, "loss": 0.3435, "step": 18150 }, { "epoch": 1.83, "learning_rate": 3.369996714600995e-05, "loss": 1.2637, "step": 18151 }, { "epoch": 1.83, "learning_rate": 3.369759223609037e-05, "loss": 0.6717, "step": 18152 }, { "epoch": 1.83, "learning_rate": 3.369521730264072e-05, "loss": 0.3643, "step": 18153 }, { "epoch": 1.83, "learning_rate": 3.369284234567611e-05, "loss": 0.6217, "step": 18154 }, { "epoch": 1.83, "learning_rate": 3.369046736521166e-05, "loss": 1.9246, "step": 18155 }, { "epoch": 1.83, "learning_rate": 3.368809236126248e-05, "loss": 0.7515, "step": 18156 }, { "epoch": 1.83, "learning_rate": 3.368571733384367e-05, "loss": 1.0408, "step": 18157 }, { "epoch": 1.83, "learning_rate": 3.3683342282970375e-05, "loss": 1.5769, "step": 18158 }, { "epoch": 1.83, "learning_rate": 3.368096720865768e-05, "loss": 1.1642, "step": 18159 }, { "epoch": 1.83, "learning_rate": 3.367859211092072e-05, "loss": 0.1974, "step": 18160 }, { "epoch": 1.83, "learning_rate": 3.367621698977459e-05, "loss": 0.7153, "step": 18161 }, { "epoch": 1.83, "learning_rate": 3.367384184523441e-05, "loss": 0.2788, "step": 18162 }, { "epoch": 1.83, "learning_rate": 3.3671466677315305e-05, "loss": 0.0984, "step": 18163 }, { "epoch": 1.83, "learning_rate": 3.366909148603238e-05, "loss": 0.8421, "step": 18164 }, { "epoch": 1.83, "learning_rate": 3.366671627140076e-05, "loss": 1.0142, "step": 18165 }, { "epoch": 1.83, "learning_rate": 3.366434103343554e-05, "loss": 0.3962, "step": 18166 }, { "epoch": 1.83, "learning_rate": 3.366196577215186e-05, "loss": 0.2135, "step": 18167 }, { "epoch": 1.83, "learning_rate": 3.3659590487564814e-05, "loss": 0.6317, "step": 18168 }, { "epoch": 1.83, "learning_rate": 3.365721517968954e-05, "loss": 0.1855, "step": 18169 }, { "epoch": 1.83, "learning_rate": 3.365483984854112e-05, "loss": 0.7091, "step": 18170 }, { "epoch": 1.83, "learning_rate": 3.3652464494134706e-05, "loss": 1.1594, "step": 18171 }, { "epoch": 1.83, "learning_rate": 3.365008911648539e-05, "loss": 0.4801, "step": 18172 }, { "epoch": 1.83, "learning_rate": 3.364771371560829e-05, "loss": 0.3337, "step": 18173 }, { "epoch": 1.83, "learning_rate": 3.3645338291518544e-05, "loss": 0.8413, "step": 18174 }, { "epoch": 1.83, "learning_rate": 3.364296284423124e-05, "loss": 1.0459, "step": 18175 }, { "epoch": 1.83, "learning_rate": 3.36405873737615e-05, "loss": 0.645, "step": 18176 }, { "epoch": 1.83, "learning_rate": 3.363821188012446e-05, "loss": 0.1393, "step": 18177 }, { "epoch": 1.83, "learning_rate": 3.363583636333522e-05, "loss": 0.5258, "step": 18178 }, { "epoch": 1.83, "learning_rate": 3.36334608234089e-05, "loss": 0.4477, "step": 18179 }, { "epoch": 1.83, "learning_rate": 3.363108526036061e-05, "loss": 0.1763, "step": 18180 }, { "epoch": 1.83, "learning_rate": 3.3628709674205487e-05, "loss": 0.8725, "step": 18181 }, { "epoch": 1.83, "learning_rate": 3.3626334064958625e-05, "loss": 0.6242, "step": 18182 }, { "epoch": 1.83, "learning_rate": 3.3623958432635156e-05, "loss": 0.7154, "step": 18183 }, { "epoch": 1.83, "learning_rate": 3.36215827772502e-05, "loss": 1.2588, "step": 18184 }, { "epoch": 1.83, "learning_rate": 3.361920709881887e-05, "loss": 0.4495, "step": 18185 }, { "epoch": 1.83, "learning_rate": 3.361683139735628e-05, "loss": 0.4396, "step": 18186 }, { "epoch": 1.83, "learning_rate": 3.3614455672877544e-05, "loss": 1.213, "step": 18187 }, { "epoch": 1.83, "learning_rate": 3.36120799253978e-05, "loss": 0.7547, "step": 18188 }, { "epoch": 1.83, "learning_rate": 3.360970415493213e-05, "loss": 1.8248, "step": 18189 }, { "epoch": 1.83, "learning_rate": 3.36073283614957e-05, "loss": 0.8963, "step": 18190 }, { "epoch": 1.83, "learning_rate": 3.360495254510359e-05, "loss": 0.7866, "step": 18191 }, { "epoch": 1.83, "learning_rate": 3.360257670577094e-05, "loss": 0.3273, "step": 18192 }, { "epoch": 1.83, "learning_rate": 3.3600200843512865e-05, "loss": 0.1966, "step": 18193 }, { "epoch": 1.83, "learning_rate": 3.359782495834448e-05, "loss": 0.5798, "step": 18194 }, { "epoch": 1.83, "learning_rate": 3.359544905028091e-05, "loss": 0.8427, "step": 18195 }, { "epoch": 1.83, "learning_rate": 3.359307311933725e-05, "loss": 0.2928, "step": 18196 }, { "epoch": 1.83, "learning_rate": 3.359069716552866e-05, "loss": 0.2582, "step": 18197 }, { "epoch": 1.83, "learning_rate": 3.358832118887022e-05, "loss": 0.9851, "step": 18198 }, { "epoch": 1.83, "learning_rate": 3.3585945189377095e-05, "loss": 1.033, "step": 18199 }, { "epoch": 1.84, "learning_rate": 3.3583569167064364e-05, "loss": 0.3824, "step": 18200 }, { "epoch": 1.84, "learning_rate": 3.358119312194716e-05, "loss": 0.2249, "step": 18201 }, { "epoch": 1.84, "learning_rate": 3.357881705404061e-05, "loss": 0.4343, "step": 18202 }, { "epoch": 1.84, "learning_rate": 3.357644096335983e-05, "loss": 0.3164, "step": 18203 }, { "epoch": 1.84, "learning_rate": 3.357406484991994e-05, "loss": 0.191, "step": 18204 }, { "epoch": 1.84, "learning_rate": 3.357168871373605e-05, "loss": 0.9567, "step": 18205 }, { "epoch": 1.84, "learning_rate": 3.356931255482331e-05, "loss": 0.4787, "step": 18206 }, { "epoch": 1.84, "learning_rate": 3.356693637319681e-05, "loss": 0.1032, "step": 18207 }, { "epoch": 1.84, "learning_rate": 3.356456016887169e-05, "loss": 0.5194, "step": 18208 }, { "epoch": 1.84, "learning_rate": 3.356218394186307e-05, "loss": 1.9016, "step": 18209 }, { "epoch": 1.84, "learning_rate": 3.355980769218606e-05, "loss": 0.9987, "step": 18210 }, { "epoch": 1.84, "learning_rate": 3.355743141985579e-05, "loss": 0.912, "step": 18211 }, { "epoch": 1.84, "learning_rate": 3.355505512488737e-05, "loss": 0.7933, "step": 18212 }, { "epoch": 1.84, "learning_rate": 3.355267880729594e-05, "loss": 0.7861, "step": 18213 }, { "epoch": 1.84, "learning_rate": 3.355030246709661e-05, "loss": 0.5316, "step": 18214 }, { "epoch": 1.84, "learning_rate": 3.3547926104304514e-05, "loss": 0.2892, "step": 18215 }, { "epoch": 1.84, "learning_rate": 3.354554971893476e-05, "loss": 0.4333, "step": 18216 }, { "epoch": 1.84, "learning_rate": 3.354317331100247e-05, "loss": 0.2936, "step": 18217 }, { "epoch": 1.84, "learning_rate": 3.3540796880522785e-05, "loss": 1.0699, "step": 18218 }, { "epoch": 1.84, "learning_rate": 3.35384204275108e-05, "loss": 0.3021, "step": 18219 }, { "epoch": 1.84, "learning_rate": 3.353604395198168e-05, "loss": 1.1173, "step": 18220 }, { "epoch": 1.84, "learning_rate": 3.35336674539505e-05, "loss": 0.3234, "step": 18221 }, { "epoch": 1.84, "learning_rate": 3.353129093343241e-05, "loss": 1.0045, "step": 18222 }, { "epoch": 1.84, "learning_rate": 3.352891439044253e-05, "loss": 0.2046, "step": 18223 }, { "epoch": 1.84, "learning_rate": 3.3526537824995975e-05, "loss": 0.971, "step": 18224 }, { "epoch": 1.84, "learning_rate": 3.3524161237107886e-05, "loss": 0.1705, "step": 18225 }, { "epoch": 1.84, "learning_rate": 3.352178462679336e-05, "loss": 0.9635, "step": 18226 }, { "epoch": 1.84, "learning_rate": 3.351940799406755e-05, "loss": 0.1713, "step": 18227 }, { "epoch": 1.84, "learning_rate": 3.3517031338945556e-05, "loss": 0.3289, "step": 18228 }, { "epoch": 1.84, "learning_rate": 3.351465466144252e-05, "loss": 0.4913, "step": 18229 }, { "epoch": 1.84, "learning_rate": 3.351227796157356e-05, "loss": 0.9779, "step": 18230 }, { "epoch": 1.84, "learning_rate": 3.3509901239353796e-05, "loss": 0.4449, "step": 18231 }, { "epoch": 1.84, "learning_rate": 3.350752449479835e-05, "loss": 0.9497, "step": 18232 }, { "epoch": 1.84, "learning_rate": 3.350514772792236e-05, "loss": 0.2688, "step": 18233 }, { "epoch": 1.84, "learning_rate": 3.3502770938740945e-05, "loss": 0.3928, "step": 18234 }, { "epoch": 1.84, "learning_rate": 3.350039412726921e-05, "loss": 0.7341, "step": 18235 }, { "epoch": 1.84, "learning_rate": 3.349801729352233e-05, "loss": 0.5434, "step": 18236 }, { "epoch": 1.84, "learning_rate": 3.3495640437515375e-05, "loss": 0.1054, "step": 18237 }, { "epoch": 1.84, "learning_rate": 3.3493263559263504e-05, "loss": 0.5114, "step": 18238 }, { "epoch": 1.84, "learning_rate": 3.349088665878182e-05, "loss": 0.2287, "step": 18239 }, { "epoch": 1.84, "learning_rate": 3.3488509736085476e-05, "loss": 0.727, "step": 18240 }, { "epoch": 1.84, "learning_rate": 3.348613279118958e-05, "loss": 0.2427, "step": 18241 }, { "epoch": 1.84, "learning_rate": 3.3483755824109264e-05, "loss": 1.1034, "step": 18242 }, { "epoch": 1.84, "learning_rate": 3.348137883485965e-05, "loss": 0.5474, "step": 18243 }, { "epoch": 1.84, "learning_rate": 3.347900182345585e-05, "loss": 0.5173, "step": 18244 }, { "epoch": 1.84, "learning_rate": 3.3476624789913026e-05, "loss": 0.6435, "step": 18245 }, { "epoch": 1.84, "learning_rate": 3.347424773424627e-05, "loss": 0.5242, "step": 18246 }, { "epoch": 1.84, "learning_rate": 3.347187065647073e-05, "loss": 0.3725, "step": 18247 }, { "epoch": 1.84, "learning_rate": 3.346949355660152e-05, "loss": 0.3687, "step": 18248 }, { "epoch": 1.84, "learning_rate": 3.3467116434653785e-05, "loss": 0.2134, "step": 18249 }, { "epoch": 1.84, "learning_rate": 3.346473929064263e-05, "loss": 1.188, "step": 18250 }, { "epoch": 1.84, "learning_rate": 3.3462362124583194e-05, "loss": 0.2201, "step": 18251 }, { "epoch": 1.84, "learning_rate": 3.345998493649061e-05, "loss": 0.264, "step": 18252 }, { "epoch": 1.84, "learning_rate": 3.345760772637998e-05, "loss": 1.0856, "step": 18253 }, { "epoch": 1.84, "learning_rate": 3.3455230494266466e-05, "loss": 0.7342, "step": 18254 }, { "epoch": 1.84, "learning_rate": 3.345285324016518e-05, "loss": 0.8446, "step": 18255 }, { "epoch": 1.84, "learning_rate": 3.345047596409124e-05, "loss": 0.367, "step": 18256 }, { "epoch": 1.84, "learning_rate": 3.344809866605979e-05, "loss": 0.9637, "step": 18257 }, { "epoch": 1.84, "learning_rate": 3.3445721346085954e-05, "loss": 0.5309, "step": 18258 }, { "epoch": 1.84, "learning_rate": 3.344334400418485e-05, "loss": 0.9791, "step": 18259 }, { "epoch": 1.84, "learning_rate": 3.3440966640371613e-05, "loss": 0.2961, "step": 18260 }, { "epoch": 1.84, "learning_rate": 3.343858925466138e-05, "loss": 0.1536, "step": 18261 }, { "epoch": 1.84, "learning_rate": 3.3436211847069275e-05, "loss": 0.9317, "step": 18262 }, { "epoch": 1.84, "learning_rate": 3.343383441761042e-05, "loss": 0.2466, "step": 18263 }, { "epoch": 1.84, "learning_rate": 3.343145696629995e-05, "loss": 0.7896, "step": 18264 }, { "epoch": 1.84, "learning_rate": 3.3429079493152994e-05, "loss": 0.2219, "step": 18265 }, { "epoch": 1.84, "learning_rate": 3.342670199818469e-05, "loss": 0.3816, "step": 18266 }, { "epoch": 1.84, "learning_rate": 3.3424324481410136e-05, "loss": 1.1508, "step": 18267 }, { "epoch": 1.84, "learning_rate": 3.342194694284451e-05, "loss": 0.6758, "step": 18268 }, { "epoch": 1.84, "learning_rate": 3.34195693825029e-05, "loss": 0.775, "step": 18269 }, { "epoch": 1.84, "learning_rate": 3.341719180040045e-05, "loss": 0.4514, "step": 18270 }, { "epoch": 1.84, "learning_rate": 3.34148141965523e-05, "loss": 0.3059, "step": 18271 }, { "epoch": 1.84, "learning_rate": 3.3412436570973564e-05, "loss": 1.1743, "step": 18272 }, { "epoch": 1.84, "learning_rate": 3.341005892367938e-05, "loss": 1.2178, "step": 18273 }, { "epoch": 1.84, "learning_rate": 3.3407681254684874e-05, "loss": 1.012, "step": 18274 }, { "epoch": 1.84, "learning_rate": 3.3405303564005194e-05, "loss": 0.3378, "step": 18275 }, { "epoch": 1.84, "learning_rate": 3.340292585165545e-05, "loss": 1.0446, "step": 18276 }, { "epoch": 1.84, "learning_rate": 3.340054811765078e-05, "loss": 0.9461, "step": 18277 }, { "epoch": 1.84, "learning_rate": 3.3398170362006316e-05, "loss": 0.5709, "step": 18278 }, { "epoch": 1.84, "learning_rate": 3.339579258473719e-05, "loss": 0.3946, "step": 18279 }, { "epoch": 1.84, "learning_rate": 3.3393414785858537e-05, "loss": 0.506, "step": 18280 }, { "epoch": 1.84, "learning_rate": 3.339103696538547e-05, "loss": 0.9777, "step": 18281 }, { "epoch": 1.84, "learning_rate": 3.338865912333315e-05, "loss": 1.0682, "step": 18282 }, { "epoch": 1.84, "learning_rate": 3.338628125971667e-05, "loss": 0.446, "step": 18283 }, { "epoch": 1.84, "learning_rate": 3.3383903374551205e-05, "loss": 0.4749, "step": 18284 }, { "epoch": 1.84, "learning_rate": 3.338152546785186e-05, "loss": 0.0304, "step": 18285 }, { "epoch": 1.84, "learning_rate": 3.337914753963377e-05, "loss": 0.3955, "step": 18286 }, { "epoch": 1.84, "learning_rate": 3.3376769589912075e-05, "loss": 0.384, "step": 18287 }, { "epoch": 1.84, "learning_rate": 3.337439161870189e-05, "loss": 0.2474, "step": 18288 }, { "epoch": 1.84, "learning_rate": 3.337201362601838e-05, "loss": 0.2914, "step": 18289 }, { "epoch": 1.84, "learning_rate": 3.336963561187665e-05, "loss": 0.1897, "step": 18290 }, { "epoch": 1.84, "learning_rate": 3.3367257576291846e-05, "loss": 0.222, "step": 18291 }, { "epoch": 1.84, "learning_rate": 3.3364879519279083e-05, "loss": 0.6691, "step": 18292 }, { "epoch": 1.84, "learning_rate": 3.336250144085351e-05, "loss": 0.3074, "step": 18293 }, { "epoch": 1.84, "learning_rate": 3.3360123341030266e-05, "loss": 0.3179, "step": 18294 }, { "epoch": 1.84, "learning_rate": 3.335774521982447e-05, "loss": 0.9589, "step": 18295 }, { "epoch": 1.84, "learning_rate": 3.335536707725126e-05, "loss": 0.9139, "step": 18296 }, { "epoch": 1.84, "learning_rate": 3.3352988913325775e-05, "loss": 0.8001, "step": 18297 }, { "epoch": 1.85, "learning_rate": 3.335061072806314e-05, "loss": 0.3182, "step": 18298 }, { "epoch": 1.85, "learning_rate": 3.3348232521478485e-05, "loss": 1.1001, "step": 18299 }, { "epoch": 1.85, "learning_rate": 3.334585429358696e-05, "loss": 0.9765, "step": 18300 }, { "epoch": 1.85, "learning_rate": 3.3343476044403694e-05, "loss": 0.3186, "step": 18301 }, { "epoch": 1.85, "learning_rate": 3.334109777394381e-05, "loss": 0.5867, "step": 18302 }, { "epoch": 1.85, "learning_rate": 3.333871948222246e-05, "loss": 0.5172, "step": 18303 }, { "epoch": 1.85, "learning_rate": 3.333634116925477e-05, "loss": 0.5078, "step": 18304 }, { "epoch": 1.85, "learning_rate": 3.333396283505587e-05, "loss": 0.3213, "step": 18305 }, { "epoch": 1.85, "learning_rate": 3.333158447964089e-05, "loss": 0.6396, "step": 18306 }, { "epoch": 1.85, "learning_rate": 3.332920610302499e-05, "loss": 0.9781, "step": 18307 }, { "epoch": 1.85, "learning_rate": 3.3326827705223275e-05, "loss": 0.5644, "step": 18308 }, { "epoch": 1.85, "learning_rate": 3.3324449286250905e-05, "loss": 0.8927, "step": 18309 }, { "epoch": 1.85, "learning_rate": 3.3322070846123005e-05, "loss": 0.8438, "step": 18310 }, { "epoch": 1.85, "learning_rate": 3.33196923848547e-05, "loss": 0.6536, "step": 18311 }, { "epoch": 1.85, "learning_rate": 3.331731390246114e-05, "loss": 0.1632, "step": 18312 }, { "epoch": 1.85, "learning_rate": 3.3314935398957456e-05, "loss": 0.4364, "step": 18313 }, { "epoch": 1.85, "learning_rate": 3.331255687435879e-05, "loss": 0.9906, "step": 18314 }, { "epoch": 1.85, "learning_rate": 3.331017832868026e-05, "loss": 0.4868, "step": 18315 }, { "epoch": 1.85, "learning_rate": 3.330779976193702e-05, "loss": 0.2801, "step": 18316 }, { "epoch": 1.85, "learning_rate": 3.3305421174144206e-05, "loss": 0.2093, "step": 18317 }, { "epoch": 1.85, "learning_rate": 3.330304256531694e-05, "loss": 1.1935, "step": 18318 }, { "epoch": 1.85, "learning_rate": 3.330066393547038e-05, "loss": 1.444, "step": 18319 }, { "epoch": 1.85, "learning_rate": 3.3298285284619637e-05, "loss": 0.3033, "step": 18320 }, { "epoch": 1.85, "learning_rate": 3.329590661277987e-05, "loss": 0.9585, "step": 18321 }, { "epoch": 1.85, "learning_rate": 3.32935279199662e-05, "loss": 0.2622, "step": 18322 }, { "epoch": 1.85, "learning_rate": 3.3291149206193784e-05, "loss": 0.4089, "step": 18323 }, { "epoch": 1.85, "learning_rate": 3.328877047147773e-05, "loss": 0.8944, "step": 18324 }, { "epoch": 1.85, "learning_rate": 3.3286391715833204e-05, "loss": 0.2195, "step": 18325 }, { "epoch": 1.85, "learning_rate": 3.3284012939275324e-05, "loss": 0.1867, "step": 18326 }, { "epoch": 1.85, "learning_rate": 3.3281634141819244e-05, "loss": 0.5844, "step": 18327 }, { "epoch": 1.85, "learning_rate": 3.327925532348008e-05, "loss": 0.7529, "step": 18328 }, { "epoch": 1.85, "learning_rate": 3.327687648427299e-05, "loss": 1.0524, "step": 18329 }, { "epoch": 1.85, "learning_rate": 3.327449762421311e-05, "loss": 0.2592, "step": 18330 }, { "epoch": 1.85, "learning_rate": 3.327211874331556e-05, "loss": 1.243, "step": 18331 }, { "epoch": 1.85, "learning_rate": 3.32697398415955e-05, "loss": 0.4341, "step": 18332 }, { "epoch": 1.85, "learning_rate": 3.326736091906806e-05, "loss": 0.6365, "step": 18333 }, { "epoch": 1.85, "learning_rate": 3.3264981975748374e-05, "loss": 0.7861, "step": 18334 }, { "epoch": 1.85, "learning_rate": 3.3262603011651586e-05, "loss": 0.2212, "step": 18335 }, { "epoch": 1.85, "learning_rate": 3.326022402679284e-05, "loss": 0.3275, "step": 18336 }, { "epoch": 1.85, "learning_rate": 3.325784502118726e-05, "loss": 0.7109, "step": 18337 }, { "epoch": 1.85, "learning_rate": 3.3255465994849986e-05, "loss": 0.2191, "step": 18338 }, { "epoch": 1.85, "learning_rate": 3.325308694779619e-05, "loss": 0.3268, "step": 18339 }, { "epoch": 1.85, "learning_rate": 3.325070788004096e-05, "loss": 0.9091, "step": 18340 }, { "epoch": 1.85, "learning_rate": 3.3248328791599474e-05, "loss": 1.1292, "step": 18341 }, { "epoch": 1.85, "learning_rate": 3.3245949682486864e-05, "loss": 0.5285, "step": 18342 }, { "epoch": 1.85, "learning_rate": 3.324357055271826e-05, "loss": 0.3885, "step": 18343 }, { "epoch": 1.85, "learning_rate": 3.3241191402308806e-05, "loss": 1.1033, "step": 18344 }, { "epoch": 1.85, "learning_rate": 3.3238812231273636e-05, "loss": 0.7481, "step": 18345 }, { "epoch": 1.85, "learning_rate": 3.323643303962791e-05, "loss": 1.2798, "step": 18346 }, { "epoch": 1.85, "learning_rate": 3.323405382738675e-05, "loss": 0.8835, "step": 18347 }, { "epoch": 1.85, "learning_rate": 3.3231674594565295e-05, "loss": 0.9886, "step": 18348 }, { "epoch": 1.85, "learning_rate": 3.32292953411787e-05, "loss": 0.4085, "step": 18349 }, { "epoch": 1.85, "learning_rate": 3.322691606724209e-05, "loss": 0.4592, "step": 18350 }, { "epoch": 1.85, "learning_rate": 3.322453677277063e-05, "loss": 0.5255, "step": 18351 }, { "epoch": 1.85, "learning_rate": 3.3222157457779435e-05, "loss": 1.4367, "step": 18352 }, { "epoch": 1.85, "learning_rate": 3.3219778122283656e-05, "loss": 1.2368, "step": 18353 }, { "epoch": 1.85, "learning_rate": 3.321739876629843e-05, "loss": 0.7094, "step": 18354 }, { "epoch": 1.85, "learning_rate": 3.32150193898389e-05, "loss": 0.9218, "step": 18355 }, { "epoch": 1.85, "learning_rate": 3.321263999292022e-05, "loss": 0.3602, "step": 18356 }, { "epoch": 1.85, "learning_rate": 3.321026057555752e-05, "loss": 0.2257, "step": 18357 }, { "epoch": 1.85, "learning_rate": 3.320788113776594e-05, "loss": 0.3515, "step": 18358 }, { "epoch": 1.85, "learning_rate": 3.3205501679560624e-05, "loss": 0.99, "step": 18359 }, { "epoch": 1.85, "learning_rate": 3.320312220095672e-05, "loss": 1.0837, "step": 18360 }, { "epoch": 1.85, "learning_rate": 3.320074270196935e-05, "loss": 0.9716, "step": 18361 }, { "epoch": 1.85, "learning_rate": 3.3198363182613686e-05, "loss": 0.4569, "step": 18362 }, { "epoch": 1.85, "learning_rate": 3.319598364290484e-05, "loss": 0.0537, "step": 18363 }, { "epoch": 1.85, "learning_rate": 3.319360408285799e-05, "loss": 0.7125, "step": 18364 }, { "epoch": 1.85, "learning_rate": 3.319122450248825e-05, "loss": 0.3259, "step": 18365 }, { "epoch": 1.85, "learning_rate": 3.318884490181076e-05, "loss": 0.9844, "step": 18366 }, { "epoch": 1.85, "learning_rate": 3.3186465280840695e-05, "loss": 0.6874, "step": 18367 }, { "epoch": 1.85, "learning_rate": 3.318408563959316e-05, "loss": 1.0423, "step": 18368 }, { "epoch": 1.85, "learning_rate": 3.318170597808333e-05, "loss": 0.6106, "step": 18369 }, { "epoch": 1.85, "learning_rate": 3.317932629632632e-05, "loss": 0.2936, "step": 18370 }, { "epoch": 1.85, "learning_rate": 3.317694659433729e-05, "loss": 0.4169, "step": 18371 }, { "epoch": 1.85, "learning_rate": 3.317456687213138e-05, "loss": 0.4294, "step": 18372 }, { "epoch": 1.85, "learning_rate": 3.317218712972374e-05, "loss": 0.7, "step": 18373 }, { "epoch": 1.85, "learning_rate": 3.3169807367129505e-05, "loss": 0.2872, "step": 18374 }, { "epoch": 1.85, "learning_rate": 3.316742758436382e-05, "loss": 0.1356, "step": 18375 }, { "epoch": 1.85, "learning_rate": 3.316504778144183e-05, "loss": 1.1564, "step": 18376 }, { "epoch": 1.85, "learning_rate": 3.3162667958378684e-05, "loss": 0.2705, "step": 18377 }, { "epoch": 1.85, "learning_rate": 3.3160288115189524e-05, "loss": 1.0847, "step": 18378 }, { "epoch": 1.85, "learning_rate": 3.3157908251889484e-05, "loss": 0.2894, "step": 18379 }, { "epoch": 1.85, "learning_rate": 3.315552836849372e-05, "loss": 1.3179, "step": 18380 }, { "epoch": 1.85, "learning_rate": 3.315314846501738e-05, "loss": 0.5835, "step": 18381 }, { "epoch": 1.85, "learning_rate": 3.31507685414756e-05, "loss": 0.7865, "step": 18382 }, { "epoch": 1.85, "learning_rate": 3.314838859788353e-05, "loss": 0.7192, "step": 18383 }, { "epoch": 1.85, "learning_rate": 3.31460086342563e-05, "loss": 0.1984, "step": 18384 }, { "epoch": 1.85, "learning_rate": 3.3143628650609085e-05, "loss": 0.7902, "step": 18385 }, { "epoch": 1.85, "learning_rate": 3.3141248646957e-05, "loss": 0.3824, "step": 18386 }, { "epoch": 1.85, "learning_rate": 3.313886862331521e-05, "loss": 0.3461, "step": 18387 }, { "epoch": 1.85, "learning_rate": 3.313648857969886e-05, "loss": 0.7498, "step": 18388 }, { "epoch": 1.85, "learning_rate": 3.3134108516123074e-05, "loss": 1.231, "step": 18389 }, { "epoch": 1.85, "learning_rate": 3.3131728432603026e-05, "loss": 0.7918, "step": 18390 }, { "epoch": 1.85, "learning_rate": 3.3129348329153845e-05, "loss": 0.5747, "step": 18391 }, { "epoch": 1.85, "learning_rate": 3.312696820579068e-05, "loss": 0.9076, "step": 18392 }, { "epoch": 1.85, "learning_rate": 3.3124588062528675e-05, "loss": 0.989, "step": 18393 }, { "epoch": 1.85, "learning_rate": 3.3122207899383e-05, "loss": 0.5462, "step": 18394 }, { "epoch": 1.85, "learning_rate": 3.3119827716368765e-05, "loss": 0.3014, "step": 18395 }, { "epoch": 1.85, "learning_rate": 3.311744751350113e-05, "loss": 0.5824, "step": 18396 }, { "epoch": 1.86, "learning_rate": 3.3115067290795254e-05, "loss": 0.9105, "step": 18397 }, { "epoch": 1.86, "learning_rate": 3.311268704826628e-05, "loss": 0.4749, "step": 18398 }, { "epoch": 1.86, "learning_rate": 3.3110306785929336e-05, "loss": 0.4985, "step": 18399 }, { "epoch": 1.86, "learning_rate": 3.3107926503799584e-05, "loss": 1.0192, "step": 18400 }, { "epoch": 1.86, "learning_rate": 3.310554620189218e-05, "loss": 0.3347, "step": 18401 }, { "epoch": 1.86, "learning_rate": 3.3103165880222256e-05, "loss": 0.557, "step": 18402 }, { "epoch": 1.86, "learning_rate": 3.310078553880496e-05, "loss": 0.3278, "step": 18403 }, { "epoch": 1.86, "learning_rate": 3.309840517765545e-05, "loss": 1.1296, "step": 18404 }, { "epoch": 1.86, "learning_rate": 3.309602479678887e-05, "loss": 0.2929, "step": 18405 }, { "epoch": 1.86, "learning_rate": 3.309364439622037e-05, "loss": 1.1382, "step": 18406 }, { "epoch": 1.86, "learning_rate": 3.309126397596509e-05, "loss": 0.3219, "step": 18407 }, { "epoch": 1.86, "learning_rate": 3.308888353603818e-05, "loss": 0.6491, "step": 18408 }, { "epoch": 1.86, "learning_rate": 3.3086503076454783e-05, "loss": 1.7398, "step": 18409 }, { "epoch": 1.86, "learning_rate": 3.308412259723006e-05, "loss": 0.207, "step": 18410 }, { "epoch": 1.86, "learning_rate": 3.3081742098379164e-05, "loss": 0.2472, "step": 18411 }, { "epoch": 1.86, "learning_rate": 3.3079361579917225e-05, "loss": 0.3377, "step": 18412 }, { "epoch": 1.86, "learning_rate": 3.3076981041859396e-05, "loss": 0.9261, "step": 18413 }, { "epoch": 1.86, "learning_rate": 3.3074600484220835e-05, "loss": 1.2417, "step": 18414 }, { "epoch": 1.86, "learning_rate": 3.3072219907016694e-05, "loss": 0.1363, "step": 18415 }, { "epoch": 1.86, "learning_rate": 3.306983931026209e-05, "loss": 0.2716, "step": 18416 }, { "epoch": 1.86, "learning_rate": 3.306745869397223e-05, "loss": 0.8035, "step": 18417 }, { "epoch": 1.86, "learning_rate": 3.3065078058162205e-05, "loss": 0.7369, "step": 18418 }, { "epoch": 1.86, "learning_rate": 3.306269740284719e-05, "loss": 0.4459, "step": 18419 }, { "epoch": 1.86, "learning_rate": 3.306031672804235e-05, "loss": 0.5824, "step": 18420 }, { "epoch": 1.86, "learning_rate": 3.305793603376281e-05, "loss": 0.6377, "step": 18421 }, { "epoch": 1.86, "learning_rate": 3.3055555320023724e-05, "loss": 0.3727, "step": 18422 }, { "epoch": 1.86, "learning_rate": 3.305317458684024e-05, "loss": 0.5973, "step": 18423 }, { "epoch": 1.86, "learning_rate": 3.305079383422754e-05, "loss": 0.374, "step": 18424 }, { "epoch": 1.86, "learning_rate": 3.304841306220072e-05, "loss": 0.3945, "step": 18425 }, { "epoch": 1.86, "learning_rate": 3.3046032270774975e-05, "loss": 0.4672, "step": 18426 }, { "epoch": 1.86, "learning_rate": 3.3043651459965436e-05, "loss": 0.2697, "step": 18427 }, { "epoch": 1.86, "learning_rate": 3.304127062978726e-05, "loss": 0.1504, "step": 18428 }, { "epoch": 1.86, "learning_rate": 3.30388897802556e-05, "loss": 0.3588, "step": 18429 }, { "epoch": 1.86, "learning_rate": 3.3036508911385594e-05, "loss": 1.1035, "step": 18430 }, { "epoch": 1.86, "learning_rate": 3.303412802319239e-05, "loss": 0.215, "step": 18431 }, { "epoch": 1.86, "learning_rate": 3.303174711569116e-05, "loss": 0.5783, "step": 18432 }, { "epoch": 1.86, "learning_rate": 3.3029366188897054e-05, "loss": 0.3175, "step": 18433 }, { "epoch": 1.86, "learning_rate": 3.302698524282519e-05, "loss": 0.759, "step": 18434 }, { "epoch": 1.86, "learning_rate": 3.3024604277490765e-05, "loss": 0.2327, "step": 18435 }, { "epoch": 1.86, "learning_rate": 3.30222232929089e-05, "loss": 1.1182, "step": 18436 }, { "epoch": 1.86, "learning_rate": 3.301984228909476e-05, "loss": 0.224, "step": 18437 }, { "epoch": 1.86, "learning_rate": 3.3017461266063484e-05, "loss": 0.6925, "step": 18438 }, { "epoch": 1.86, "learning_rate": 3.301508022383024e-05, "loss": 0.8744, "step": 18439 }, { "epoch": 1.86, "learning_rate": 3.301269916241018e-05, "loss": 0.3519, "step": 18440 }, { "epoch": 1.86, "learning_rate": 3.301031808181843e-05, "loss": 0.6218, "step": 18441 }, { "epoch": 1.86, "learning_rate": 3.300793698207017e-05, "loss": 0.6692, "step": 18442 }, { "epoch": 1.86, "learning_rate": 3.3005555863180544e-05, "loss": 0.4528, "step": 18443 }, { "epoch": 1.86, "learning_rate": 3.3003174725164705e-05, "loss": 0.4621, "step": 18444 }, { "epoch": 1.86, "learning_rate": 3.30007935680378e-05, "loss": 0.6919, "step": 18445 }, { "epoch": 1.86, "learning_rate": 3.2998412391815e-05, "loss": 0.4519, "step": 18446 }, { "epoch": 1.86, "learning_rate": 3.2996031196511426e-05, "loss": 0.5744, "step": 18447 }, { "epoch": 1.86, "learning_rate": 3.2993649982142254e-05, "loss": 0.4332, "step": 18448 }, { "epoch": 1.86, "learning_rate": 3.2991268748722634e-05, "loss": 0.5143, "step": 18449 }, { "epoch": 1.86, "learning_rate": 3.298888749626772e-05, "loss": 0.5275, "step": 18450 }, { "epoch": 1.86, "learning_rate": 3.298650622479266e-05, "loss": 0.0804, "step": 18451 }, { "epoch": 1.86, "learning_rate": 3.298412493431261e-05, "loss": 0.3431, "step": 18452 }, { "epoch": 1.86, "learning_rate": 3.298174362484273e-05, "loss": 0.5515, "step": 18453 }, { "epoch": 1.86, "learning_rate": 3.297936229639817e-05, "loss": 0.4187, "step": 18454 }, { "epoch": 1.86, "learning_rate": 3.297698094899407e-05, "loss": 0.7124, "step": 18455 }, { "epoch": 1.86, "learning_rate": 3.2974599582645604e-05, "loss": 0.3911, "step": 18456 }, { "epoch": 1.86, "learning_rate": 3.2972218197367904e-05, "loss": 0.4147, "step": 18457 }, { "epoch": 1.86, "learning_rate": 3.2969836793176156e-05, "loss": 1.0407, "step": 18458 }, { "epoch": 1.86, "learning_rate": 3.2967455370085496e-05, "loss": 1.0045, "step": 18459 }, { "epoch": 1.86, "learning_rate": 3.296507392811107e-05, "loss": 0.6226, "step": 18460 }, { "epoch": 1.86, "learning_rate": 3.296269246726804e-05, "loss": 0.3614, "step": 18461 }, { "epoch": 1.86, "learning_rate": 3.296031098757156e-05, "loss": 0.3632, "step": 18462 }, { "epoch": 1.86, "learning_rate": 3.295792948903681e-05, "loss": 0.575, "step": 18463 }, { "epoch": 1.86, "learning_rate": 3.2955547971678896e-05, "loss": 0.3963, "step": 18464 }, { "epoch": 1.86, "learning_rate": 3.295316643551301e-05, "loss": 0.6723, "step": 18465 }, { "epoch": 1.86, "learning_rate": 3.295078488055429e-05, "loss": 0.4749, "step": 18466 }, { "epoch": 1.86, "learning_rate": 3.2948403306817905e-05, "loss": 0.5367, "step": 18467 }, { "epoch": 1.86, "learning_rate": 3.294602171431901e-05, "loss": 0.236, "step": 18468 }, { "epoch": 1.86, "learning_rate": 3.294364010307274e-05, "loss": 0.156, "step": 18469 }, { "epoch": 1.86, "learning_rate": 3.294125847309426e-05, "loss": 0.8298, "step": 18470 }, { "epoch": 1.86, "learning_rate": 3.293887682439874e-05, "loss": 0.5961, "step": 18471 }, { "epoch": 1.86, "learning_rate": 3.293649515700133e-05, "loss": 0.0812, "step": 18472 }, { "epoch": 1.86, "learning_rate": 3.2934113470917165e-05, "loss": 0.7115, "step": 18473 }, { "epoch": 1.86, "learning_rate": 3.293173176616144e-05, "loss": 0.4022, "step": 18474 }, { "epoch": 1.86, "learning_rate": 3.292935004274927e-05, "loss": 0.4149, "step": 18475 }, { "epoch": 1.86, "learning_rate": 3.292696830069585e-05, "loss": 0.8962, "step": 18476 }, { "epoch": 1.86, "learning_rate": 3.29245865400163e-05, "loss": 0.9729, "step": 18477 }, { "epoch": 1.86, "learning_rate": 3.292220476072579e-05, "loss": 0.6394, "step": 18478 }, { "epoch": 1.86, "learning_rate": 3.29198229628395e-05, "loss": 0.6524, "step": 18479 }, { "epoch": 1.86, "learning_rate": 3.2917441146372555e-05, "loss": 0.288, "step": 18480 }, { "epoch": 1.86, "learning_rate": 3.2915059311340127e-05, "loss": 1.1657, "step": 18481 }, { "epoch": 1.86, "learning_rate": 3.291267745775737e-05, "loss": 0.5115, "step": 18482 }, { "epoch": 1.86, "learning_rate": 3.291029558563944e-05, "loss": 0.8113, "step": 18483 }, { "epoch": 1.86, "learning_rate": 3.29079136950015e-05, "loss": 1.0026, "step": 18484 }, { "epoch": 1.86, "learning_rate": 3.290553178585871e-05, "loss": 1.2585, "step": 18485 }, { "epoch": 1.86, "learning_rate": 3.290314985822621e-05, "loss": 0.5265, "step": 18486 }, { "epoch": 1.86, "learning_rate": 3.290076791211917e-05, "loss": 0.5465, "step": 18487 }, { "epoch": 1.86, "learning_rate": 3.289838594755275e-05, "loss": 0.4016, "step": 18488 }, { "epoch": 1.86, "learning_rate": 3.2896003964542106e-05, "loss": 0.7756, "step": 18489 }, { "epoch": 1.86, "learning_rate": 3.289362196310239e-05, "loss": 0.3747, "step": 18490 }, { "epoch": 1.86, "learning_rate": 3.2891239943248756e-05, "loss": 0.6302, "step": 18491 }, { "epoch": 1.86, "learning_rate": 3.288885790499639e-05, "loss": 1.0033, "step": 18492 }, { "epoch": 1.86, "learning_rate": 3.288647584836043e-05, "loss": 0.7364, "step": 18493 }, { "epoch": 1.86, "learning_rate": 3.2884093773356016e-05, "loss": 0.9234, "step": 18494 }, { "epoch": 1.87, "learning_rate": 3.2881711679998346e-05, "loss": 1.3347, "step": 18495 }, { "epoch": 1.87, "learning_rate": 3.287932956830255e-05, "loss": 0.2366, "step": 18496 }, { "epoch": 1.87, "learning_rate": 3.2876947438283794e-05, "loss": 0.8877, "step": 18497 }, { "epoch": 1.87, "learning_rate": 3.287456528995724e-05, "loss": 0.6354, "step": 18498 }, { "epoch": 1.87, "learning_rate": 3.2872183123338053e-05, "loss": 0.9971, "step": 18499 }, { "epoch": 1.87, "learning_rate": 3.286980093844138e-05, "loss": 0.3385, "step": 18500 }, { "epoch": 1.87, "learning_rate": 3.2867418735282385e-05, "loss": 0.3215, "step": 18501 }, { "epoch": 1.87, "learning_rate": 3.286503651387623e-05, "loss": 0.3868, "step": 18502 }, { "epoch": 1.87, "learning_rate": 3.2862654274238064e-05, "loss": 0.0821, "step": 18503 }, { "epoch": 1.87, "learning_rate": 3.2860272016383065e-05, "loss": 1.3299, "step": 18504 }, { "epoch": 1.87, "learning_rate": 3.285788974032638e-05, "loss": 0.3347, "step": 18505 }, { "epoch": 1.87, "learning_rate": 3.2855507446083176e-05, "loss": 0.9258, "step": 18506 }, { "epoch": 1.87, "learning_rate": 3.28531251336686e-05, "loss": 1.1428, "step": 18507 }, { "epoch": 1.87, "learning_rate": 3.285074280309783e-05, "loss": 0.342, "step": 18508 }, { "epoch": 1.87, "learning_rate": 3.284836045438602e-05, "loss": 0.4816, "step": 18509 }, { "epoch": 1.87, "learning_rate": 3.284597808754831e-05, "loss": 0.3963, "step": 18510 }, { "epoch": 1.87, "learning_rate": 3.28435957025999e-05, "loss": 0.3709, "step": 18511 }, { "epoch": 1.87, "learning_rate": 3.2841213299555914e-05, "loss": 1.0957, "step": 18512 }, { "epoch": 1.87, "learning_rate": 3.2838830878431534e-05, "loss": 0.6113, "step": 18513 }, { "epoch": 1.87, "learning_rate": 3.2836448439241905e-05, "loss": 1.0425, "step": 18514 }, { "epoch": 1.87, "learning_rate": 3.283406598200221e-05, "loss": 0.186, "step": 18515 }, { "epoch": 1.87, "learning_rate": 3.2831683506727594e-05, "loss": 0.3725, "step": 18516 }, { "epoch": 1.87, "learning_rate": 3.282930101343321e-05, "loss": 0.5858, "step": 18517 }, { "epoch": 1.87, "learning_rate": 3.282691850213425e-05, "loss": 1.0477, "step": 18518 }, { "epoch": 1.87, "learning_rate": 3.2824535972845834e-05, "loss": 0.9831, "step": 18519 }, { "epoch": 1.87, "learning_rate": 3.282215342558316e-05, "loss": 0.9817, "step": 18520 }, { "epoch": 1.87, "learning_rate": 3.2819770860361375e-05, "loss": 0.2145, "step": 18521 }, { "epoch": 1.87, "learning_rate": 3.281738827719564e-05, "loss": 0.8448, "step": 18522 }, { "epoch": 1.87, "learning_rate": 3.2815005676101106e-05, "loss": 0.3015, "step": 18523 }, { "epoch": 1.87, "learning_rate": 3.281262305709297e-05, "loss": 0.2647, "step": 18524 }, { "epoch": 1.87, "learning_rate": 3.2810240420186356e-05, "loss": 1.0327, "step": 18525 }, { "epoch": 1.87, "learning_rate": 3.280785776539644e-05, "loss": 0.628, "step": 18526 }, { "epoch": 1.87, "learning_rate": 3.280547509273839e-05, "loss": 0.5031, "step": 18527 }, { "epoch": 1.87, "learning_rate": 3.280309240222735e-05, "loss": 0.4869, "step": 18528 }, { "epoch": 1.87, "learning_rate": 3.280070969387852e-05, "loss": 0.5231, "step": 18529 }, { "epoch": 1.87, "learning_rate": 3.279832696770701e-05, "loss": 0.707, "step": 18530 }, { "epoch": 1.87, "learning_rate": 3.279594422372805e-05, "loss": 0.2726, "step": 18531 }, { "epoch": 1.87, "learning_rate": 3.2793561461956744e-05, "loss": 0.6586, "step": 18532 }, { "epoch": 1.87, "learning_rate": 3.2791178682408264e-05, "loss": 0.1281, "step": 18533 }, { "epoch": 1.87, "learning_rate": 3.27887958850978e-05, "loss": 0.1624, "step": 18534 }, { "epoch": 1.87, "learning_rate": 3.2786413070040494e-05, "loss": 0.0722, "step": 18535 }, { "epoch": 1.87, "learning_rate": 3.2784030237251515e-05, "loss": 0.4722, "step": 18536 }, { "epoch": 1.87, "learning_rate": 3.2781647386746035e-05, "loss": 0.3578, "step": 18537 }, { "epoch": 1.87, "learning_rate": 3.27792645185392e-05, "loss": 0.8188, "step": 18538 }, { "epoch": 1.87, "learning_rate": 3.277688163264617e-05, "loss": 0.4517, "step": 18539 }, { "epoch": 1.87, "learning_rate": 3.2774498729082156e-05, "loss": 0.5902, "step": 18540 }, { "epoch": 1.87, "learning_rate": 3.2772115807862266e-05, "loss": 0.1345, "step": 18541 }, { "epoch": 1.87, "learning_rate": 3.276973286900168e-05, "loss": 0.5195, "step": 18542 }, { "epoch": 1.87, "learning_rate": 3.276734991251559e-05, "loss": 0.316, "step": 18543 }, { "epoch": 1.87, "learning_rate": 3.276496693841912e-05, "loss": 0.4018, "step": 18544 }, { "epoch": 1.87, "learning_rate": 3.276258394672745e-05, "loss": 0.2035, "step": 18545 }, { "epoch": 1.87, "learning_rate": 3.276020093745575e-05, "loss": 0.6567, "step": 18546 }, { "epoch": 1.87, "learning_rate": 3.275781791061919e-05, "loss": 1.1441, "step": 18547 }, { "epoch": 1.87, "learning_rate": 3.275543486623292e-05, "loss": 0.6425, "step": 18548 }, { "epoch": 1.87, "learning_rate": 3.2753051804312104e-05, "loss": 0.7188, "step": 18549 }, { "epoch": 1.87, "learning_rate": 3.275066872487193e-05, "loss": 1.0466, "step": 18550 }, { "epoch": 1.87, "learning_rate": 3.274828562792753e-05, "loss": 0.8523, "step": 18551 }, { "epoch": 1.87, "learning_rate": 3.27459025134941e-05, "loss": 0.7888, "step": 18552 }, { "epoch": 1.87, "learning_rate": 3.274351938158678e-05, "loss": 1.0077, "step": 18553 }, { "epoch": 1.87, "learning_rate": 3.274113623222075e-05, "loss": 0.5291, "step": 18554 }, { "epoch": 1.87, "learning_rate": 3.273875306541117e-05, "loss": 0.541, "step": 18555 }, { "epoch": 1.87, "learning_rate": 3.2736369881173216e-05, "loss": 0.3174, "step": 18556 }, { "epoch": 1.87, "learning_rate": 3.273398667952204e-05, "loss": 0.7135, "step": 18557 }, { "epoch": 1.87, "learning_rate": 3.2731603460472804e-05, "loss": 0.4697, "step": 18558 }, { "epoch": 1.87, "learning_rate": 3.272922022404069e-05, "loss": 0.5991, "step": 18559 }, { "epoch": 1.87, "learning_rate": 3.2726836970240854e-05, "loss": 0.3304, "step": 18560 }, { "epoch": 1.87, "learning_rate": 3.272445369908847e-05, "loss": 1.2169, "step": 18561 }, { "epoch": 1.87, "learning_rate": 3.2722070410598694e-05, "loss": 0.366, "step": 18562 }, { "epoch": 1.87, "learning_rate": 3.2719687104786706e-05, "loss": 0.3796, "step": 18563 }, { "epoch": 1.87, "learning_rate": 3.271730378166766e-05, "loss": 0.4856, "step": 18564 }, { "epoch": 1.87, "learning_rate": 3.2714920441256714e-05, "loss": 1.2408, "step": 18565 }, { "epoch": 1.87, "learning_rate": 3.271253708356907e-05, "loss": 0.1598, "step": 18566 }, { "epoch": 1.87, "learning_rate": 3.271015370861985e-05, "loss": 0.33, "step": 18567 }, { "epoch": 1.87, "learning_rate": 3.270777031642426e-05, "loss": 0.2872, "step": 18568 }, { "epoch": 1.87, "learning_rate": 3.2705386906997434e-05, "loss": 0.2613, "step": 18569 }, { "epoch": 1.87, "learning_rate": 3.270300348035457e-05, "loss": 0.7287, "step": 18570 }, { "epoch": 1.87, "learning_rate": 3.2700620036510804e-05, "loss": 0.374, "step": 18571 }, { "epoch": 1.87, "learning_rate": 3.269823657548133e-05, "loss": 0.1248, "step": 18572 }, { "epoch": 1.87, "learning_rate": 3.269585309728131e-05, "loss": 0.7269, "step": 18573 }, { "epoch": 1.87, "learning_rate": 3.2693469601925895e-05, "loss": 0.0764, "step": 18574 }, { "epoch": 1.87, "learning_rate": 3.269108608943027e-05, "loss": 0.5743, "step": 18575 }, { "epoch": 1.87, "learning_rate": 3.26887025598096e-05, "loss": 0.4094, "step": 18576 }, { "epoch": 1.87, "learning_rate": 3.268631901307905e-05, "loss": 0.5827, "step": 18577 }, { "epoch": 1.87, "learning_rate": 3.268393544925377e-05, "loss": 0.7487, "step": 18578 }, { "epoch": 1.87, "learning_rate": 3.2681551868348973e-05, "loss": 0.549, "step": 18579 }, { "epoch": 1.87, "learning_rate": 3.2679168270379784e-05, "loss": 0.2006, "step": 18580 }, { "epoch": 1.87, "learning_rate": 3.267678465536138e-05, "loss": 0.2483, "step": 18581 }, { "epoch": 1.87, "learning_rate": 3.2674401023308954e-05, "loss": 0.4657, "step": 18582 }, { "epoch": 1.87, "learning_rate": 3.267201737423765e-05, "loss": 0.6836, "step": 18583 }, { "epoch": 1.87, "learning_rate": 3.266963370816264e-05, "loss": 0.9555, "step": 18584 }, { "epoch": 1.87, "learning_rate": 3.266725002509909e-05, "loss": 1.0074, "step": 18585 }, { "epoch": 1.87, "learning_rate": 3.2664866325062195e-05, "loss": 0.6397, "step": 18586 }, { "epoch": 1.87, "learning_rate": 3.266248260806709e-05, "loss": 0.3978, "step": 18587 }, { "epoch": 1.87, "learning_rate": 3.2660098874128966e-05, "loss": 0.6076, "step": 18588 }, { "epoch": 1.87, "learning_rate": 3.265771512326299e-05, "loss": 1.1803, "step": 18589 }, { "epoch": 1.87, "learning_rate": 3.2655331355484316e-05, "loss": 0.2598, "step": 18590 }, { "epoch": 1.87, "learning_rate": 3.265294757080812e-05, "loss": 0.3389, "step": 18591 }, { "epoch": 1.87, "learning_rate": 3.265056376924959e-05, "loss": 0.8708, "step": 18592 }, { "epoch": 1.88, "learning_rate": 3.2648179950823865e-05, "loss": 0.8973, "step": 18593 }, { "epoch": 1.88, "learning_rate": 3.264579611554613e-05, "loss": 0.8656, "step": 18594 }, { "epoch": 1.88, "learning_rate": 3.264341226343157e-05, "loss": 1.0906, "step": 18595 }, { "epoch": 1.88, "learning_rate": 3.2641028394495334e-05, "loss": 0.9304, "step": 18596 }, { "epoch": 1.88, "learning_rate": 3.263864450875259e-05, "loss": 0.1424, "step": 18597 }, { "epoch": 1.88, "learning_rate": 3.263626060621852e-05, "loss": 0.8603, "step": 18598 }, { "epoch": 1.88, "learning_rate": 3.2633876686908294e-05, "loss": 0.384, "step": 18599 }, { "epoch": 1.88, "learning_rate": 3.2631492750837075e-05, "loss": 1.5322, "step": 18600 }, { "epoch": 1.88, "learning_rate": 3.262910879802004e-05, "loss": 0.2424, "step": 18601 }, { "epoch": 1.88, "learning_rate": 3.262672482847235e-05, "loss": 0.6632, "step": 18602 }, { "epoch": 1.88, "learning_rate": 3.262434084220919e-05, "loss": 0.7616, "step": 18603 }, { "epoch": 1.88, "learning_rate": 3.262195683924572e-05, "loss": 1.194, "step": 18604 }, { "epoch": 1.88, "learning_rate": 3.261957281959713e-05, "loss": 0.6503, "step": 18605 }, { "epoch": 1.88, "learning_rate": 3.261718878327855e-05, "loss": 0.8858, "step": 18606 }, { "epoch": 1.88, "learning_rate": 3.261480473030518e-05, "loss": 0.6496, "step": 18607 }, { "epoch": 1.88, "learning_rate": 3.26124206606922e-05, "loss": 0.8261, "step": 18608 }, { "epoch": 1.88, "learning_rate": 3.261003657445476e-05, "loss": 0.3146, "step": 18609 }, { "epoch": 1.88, "learning_rate": 3.260765247160804e-05, "loss": 0.0375, "step": 18610 }, { "epoch": 1.88, "learning_rate": 3.260526835216722e-05, "loss": 0.1204, "step": 18611 }, { "epoch": 1.88, "learning_rate": 3.260288421614746e-05, "loss": 0.4566, "step": 18612 }, { "epoch": 1.88, "learning_rate": 3.2600500063563935e-05, "loss": 1.2853, "step": 18613 }, { "epoch": 1.88, "learning_rate": 3.259811589443181e-05, "loss": 0.2614, "step": 18614 }, { "epoch": 1.88, "learning_rate": 3.259573170876627e-05, "loss": 0.9708, "step": 18615 }, { "epoch": 1.88, "learning_rate": 3.259334750658249e-05, "loss": 1.0198, "step": 18616 }, { "epoch": 1.88, "learning_rate": 3.259096328789561e-05, "loss": 0.0786, "step": 18617 }, { "epoch": 1.88, "learning_rate": 3.258857905272084e-05, "loss": 0.1742, "step": 18618 }, { "epoch": 1.88, "learning_rate": 3.258619480107333e-05, "loss": 0.614, "step": 18619 }, { "epoch": 1.88, "learning_rate": 3.2583810532968274e-05, "loss": 1.1984, "step": 18620 }, { "epoch": 1.88, "learning_rate": 3.258142624842083e-05, "loss": 0.3085, "step": 18621 }, { "epoch": 1.88, "learning_rate": 3.2579041947446155e-05, "loss": 0.1092, "step": 18622 }, { "epoch": 1.88, "learning_rate": 3.2576657630059446e-05, "loss": 0.7233, "step": 18623 }, { "epoch": 1.88, "learning_rate": 3.257427329627587e-05, "loss": 0.2031, "step": 18624 }, { "epoch": 1.88, "learning_rate": 3.25718889461106e-05, "loss": 0.8406, "step": 18625 }, { "epoch": 1.88, "learning_rate": 3.2569504579578796e-05, "loss": 0.4279, "step": 18626 }, { "epoch": 1.88, "learning_rate": 3.2567120196695657e-05, "loss": 0.126, "step": 18627 }, { "epoch": 1.88, "learning_rate": 3.256473579747634e-05, "loss": 0.8219, "step": 18628 }, { "epoch": 1.88, "learning_rate": 3.2562351381936003e-05, "loss": 0.4684, "step": 18629 }, { "epoch": 1.88, "learning_rate": 3.2559966950089854e-05, "loss": 0.3067, "step": 18630 }, { "epoch": 1.88, "learning_rate": 3.2557582501953045e-05, "loss": 1.2273, "step": 18631 }, { "epoch": 1.88, "learning_rate": 3.2555198037540755e-05, "loss": 0.902, "step": 18632 }, { "epoch": 1.88, "learning_rate": 3.2552813556868143e-05, "loss": 0.7297, "step": 18633 }, { "epoch": 1.88, "learning_rate": 3.255042905995042e-05, "loss": 0.8161, "step": 18634 }, { "epoch": 1.88, "learning_rate": 3.2548044546802717e-05, "loss": 0.5797, "step": 18635 }, { "epoch": 1.88, "learning_rate": 3.254566001744024e-05, "loss": 0.185, "step": 18636 }, { "epoch": 1.88, "learning_rate": 3.2543275471878154e-05, "loss": 0.1408, "step": 18637 }, { "epoch": 1.88, "learning_rate": 3.2540890910131614e-05, "loss": 0.7738, "step": 18638 }, { "epoch": 1.88, "learning_rate": 3.2538506332215826e-05, "loss": 0.2358, "step": 18639 }, { "epoch": 1.88, "learning_rate": 3.253612173814594e-05, "loss": 0.5669, "step": 18640 }, { "epoch": 1.88, "learning_rate": 3.253373712793714e-05, "loss": 0.2425, "step": 18641 }, { "epoch": 1.88, "learning_rate": 3.253135250160461e-05, "loss": 0.2437, "step": 18642 }, { "epoch": 1.88, "learning_rate": 3.252896785916351e-05, "loss": 0.4163, "step": 18643 }, { "epoch": 1.88, "learning_rate": 3.2526583200629025e-05, "loss": 0.655, "step": 18644 }, { "epoch": 1.88, "learning_rate": 3.2524198526016314e-05, "loss": 0.1292, "step": 18645 }, { "epoch": 1.88, "learning_rate": 3.252181383534057e-05, "loss": 0.3857, "step": 18646 }, { "epoch": 1.88, "learning_rate": 3.251942912861697e-05, "loss": 0.1331, "step": 18647 }, { "epoch": 1.88, "learning_rate": 3.251704440586067e-05, "loss": 1.0568, "step": 18648 }, { "epoch": 1.88, "learning_rate": 3.251465966708686e-05, "loss": 0.7951, "step": 18649 }, { "epoch": 1.88, "learning_rate": 3.251227491231072e-05, "loss": 0.603, "step": 18650 }, { "epoch": 1.88, "learning_rate": 3.250989014154741e-05, "loss": 0.3526, "step": 18651 }, { "epoch": 1.88, "learning_rate": 3.2507505354812114e-05, "loss": 0.9436, "step": 18652 }, { "epoch": 1.88, "learning_rate": 3.250512055212001e-05, "loss": 0.2111, "step": 18653 }, { "epoch": 1.88, "learning_rate": 3.2502735733486275e-05, "loss": 0.2633, "step": 18654 }, { "epoch": 1.88, "learning_rate": 3.250035089892608e-05, "loss": 0.8611, "step": 18655 }, { "epoch": 1.88, "learning_rate": 3.2497966048454594e-05, "loss": 0.7097, "step": 18656 }, { "epoch": 1.88, "learning_rate": 3.2495581182087014e-05, "loss": 0.6484, "step": 18657 }, { "epoch": 1.88, "learning_rate": 3.2493196299838494e-05, "loss": 0.6951, "step": 18658 }, { "epoch": 1.88, "learning_rate": 3.2490811401724234e-05, "loss": 1.7429, "step": 18659 }, { "epoch": 1.88, "learning_rate": 3.24884264877594e-05, "loss": 1.6683, "step": 18660 }, { "epoch": 1.88, "learning_rate": 3.2486041557959155e-05, "loss": 1.1477, "step": 18661 }, { "epoch": 1.88, "learning_rate": 3.2483656612338686e-05, "loss": 0.7041, "step": 18662 }, { "epoch": 1.88, "learning_rate": 3.248127165091318e-05, "loss": 0.6029, "step": 18663 }, { "epoch": 1.88, "learning_rate": 3.2478886673697805e-05, "loss": 0.9772, "step": 18664 }, { "epoch": 1.88, "learning_rate": 3.2476501680707724e-05, "loss": 0.2305, "step": 18665 }, { "epoch": 1.88, "learning_rate": 3.247411667195814e-05, "loss": 0.6326, "step": 18666 }, { "epoch": 1.88, "learning_rate": 3.2471731647464213e-05, "loss": 1.032, "step": 18667 }, { "epoch": 1.88, "learning_rate": 3.2469346607241135e-05, "loss": 0.9715, "step": 18668 }, { "epoch": 1.88, "learning_rate": 3.246696155130408e-05, "loss": 0.3221, "step": 18669 }, { "epoch": 1.88, "learning_rate": 3.246457647966821e-05, "loss": 0.2044, "step": 18670 }, { "epoch": 1.88, "learning_rate": 3.246219139234871e-05, "loss": 0.3413, "step": 18671 }, { "epoch": 1.88, "learning_rate": 3.245980628936075e-05, "loss": 0.9697, "step": 18672 }, { "epoch": 1.88, "learning_rate": 3.245742117071954e-05, "loss": 0.4139, "step": 18673 }, { "epoch": 1.88, "learning_rate": 3.245503603644023e-05, "loss": 0.8335, "step": 18674 }, { "epoch": 1.88, "learning_rate": 3.245265088653801e-05, "loss": 0.8782, "step": 18675 }, { "epoch": 1.88, "learning_rate": 3.245026572102804e-05, "loss": 0.5391, "step": 18676 }, { "epoch": 1.88, "learning_rate": 3.2447880539925516e-05, "loss": 0.4419, "step": 18677 }, { "epoch": 1.88, "learning_rate": 3.2445495343245613e-05, "loss": 0.2466, "step": 18678 }, { "epoch": 1.88, "learning_rate": 3.24431101310035e-05, "loss": 0.5729, "step": 18679 }, { "epoch": 1.88, "learning_rate": 3.244072490321438e-05, "loss": 0.4086, "step": 18680 }, { "epoch": 1.88, "learning_rate": 3.24383396598934e-05, "loss": 1.1317, "step": 18681 }, { "epoch": 1.88, "learning_rate": 3.243595440105576e-05, "loss": 1.0311, "step": 18682 }, { "epoch": 1.88, "learning_rate": 3.243356912671664e-05, "loss": 0.2817, "step": 18683 }, { "epoch": 1.88, "learning_rate": 3.24311838368912e-05, "loss": 0.5038, "step": 18684 }, { "epoch": 1.88, "learning_rate": 3.2428798531594644e-05, "loss": 0.093, "step": 18685 }, { "epoch": 1.88, "learning_rate": 3.242641321084213e-05, "loss": 1.2546, "step": 18686 }, { "epoch": 1.88, "learning_rate": 3.2424027874648844e-05, "loss": 0.8512, "step": 18687 }, { "epoch": 1.88, "learning_rate": 3.242164252302996e-05, "loss": 0.8133, "step": 18688 }, { "epoch": 1.88, "learning_rate": 3.241925715600068e-05, "loss": 0.6591, "step": 18689 }, { "epoch": 1.88, "learning_rate": 3.2416871773576155e-05, "loss": 0.3459, "step": 18690 }, { "epoch": 1.88, "learning_rate": 3.2414486375771586e-05, "loss": 0.2678, "step": 18691 }, { "epoch": 1.89, "learning_rate": 3.2412100962602144e-05, "loss": 0.5606, "step": 18692 }, { "epoch": 1.89, "learning_rate": 3.240971553408301e-05, "loss": 0.6568, "step": 18693 }, { "epoch": 1.89, "learning_rate": 3.240733009022935e-05, "loss": 1.1207, "step": 18694 }, { "epoch": 1.89, "learning_rate": 3.2404944631056363e-05, "loss": 0.2759, "step": 18695 }, { "epoch": 1.89, "learning_rate": 3.240255915657923e-05, "loss": 0.4788, "step": 18696 }, { "epoch": 1.89, "learning_rate": 3.240017366681312e-05, "loss": 0.3252, "step": 18697 }, { "epoch": 1.89, "learning_rate": 3.239778816177322e-05, "loss": 0.761, "step": 18698 }, { "epoch": 1.89, "learning_rate": 3.2395402641474704e-05, "loss": 1.0696, "step": 18699 }, { "epoch": 1.89, "learning_rate": 3.2393017105932766e-05, "loss": 0.2671, "step": 18700 }, { "epoch": 1.89, "learning_rate": 3.239063155516257e-05, "loss": 0.3435, "step": 18701 }, { "epoch": 1.89, "learning_rate": 3.238824598917931e-05, "loss": 0.8176, "step": 18702 }, { "epoch": 1.89, "learning_rate": 3.2385860407998154e-05, "loss": 0.4559, "step": 18703 }, { "epoch": 1.89, "learning_rate": 3.238347481163429e-05, "loss": 1.0874, "step": 18704 }, { "epoch": 1.89, "learning_rate": 3.238108920010291e-05, "loss": 0.3753, "step": 18705 }, { "epoch": 1.89, "learning_rate": 3.237870357341917e-05, "loss": 1.5641, "step": 18706 }, { "epoch": 1.89, "learning_rate": 3.237631793159827e-05, "loss": 0.7795, "step": 18707 }, { "epoch": 1.89, "learning_rate": 3.2373932274655386e-05, "loss": 0.5316, "step": 18708 }, { "epoch": 1.89, "learning_rate": 3.2371546602605706e-05, "loss": 0.1968, "step": 18709 }, { "epoch": 1.89, "learning_rate": 3.23691609154644e-05, "loss": 0.4434, "step": 18710 }, { "epoch": 1.89, "learning_rate": 3.236677521324666e-05, "loss": 0.0996, "step": 18711 }, { "epoch": 1.89, "learning_rate": 3.236438949596766e-05, "loss": 0.7473, "step": 18712 }, { "epoch": 1.89, "learning_rate": 3.2362003763642575e-05, "loss": 0.6368, "step": 18713 }, { "epoch": 1.89, "learning_rate": 3.235961801628661e-05, "loss": 0.6263, "step": 18714 }, { "epoch": 1.89, "learning_rate": 3.235723225391493e-05, "loss": 0.9418, "step": 18715 }, { "epoch": 1.89, "learning_rate": 3.235484647654272e-05, "loss": 0.7292, "step": 18716 }, { "epoch": 1.89, "learning_rate": 3.235246068418517e-05, "loss": 0.8387, "step": 18717 }, { "epoch": 1.89, "learning_rate": 3.235007487685744e-05, "loss": 0.2756, "step": 18718 }, { "epoch": 1.89, "learning_rate": 3.234768905457474e-05, "loss": 0.8071, "step": 18719 }, { "epoch": 1.89, "learning_rate": 3.234530321735223e-05, "loss": 0.8039, "step": 18720 }, { "epoch": 1.89, "learning_rate": 3.234291736520511e-05, "loss": 0.5933, "step": 18721 }, { "epoch": 1.89, "learning_rate": 3.234053149814856e-05, "loss": 0.3494, "step": 18722 }, { "epoch": 1.89, "learning_rate": 3.233814561619775e-05, "loss": 0.4816, "step": 18723 }, { "epoch": 1.89, "learning_rate": 3.2335759719367875e-05, "loss": 0.8102, "step": 18724 }, { "epoch": 1.89, "learning_rate": 3.23333738076741e-05, "loss": 0.7172, "step": 18725 }, { "epoch": 1.89, "learning_rate": 3.2330987881131645e-05, "loss": 0.5823, "step": 18726 }, { "epoch": 1.89, "learning_rate": 3.232860193975565e-05, "loss": 0.9825, "step": 18727 }, { "epoch": 1.89, "learning_rate": 3.232621598356133e-05, "loss": 0.1592, "step": 18728 }, { "epoch": 1.89, "learning_rate": 3.232383001256384e-05, "loss": 0.9571, "step": 18729 }, { "epoch": 1.89, "learning_rate": 3.23214440267784e-05, "loss": 0.8151, "step": 18730 }, { "epoch": 1.89, "learning_rate": 3.231905802622017e-05, "loss": 0.2506, "step": 18731 }, { "epoch": 1.89, "learning_rate": 3.231667201090433e-05, "loss": 1.3829, "step": 18732 }, { "epoch": 1.89, "learning_rate": 3.231428598084607e-05, "loss": 0.3564, "step": 18733 }, { "epoch": 1.89, "learning_rate": 3.231189993606057e-05, "loss": 1.1185, "step": 18734 }, { "epoch": 1.89, "learning_rate": 3.230951387656303e-05, "loss": 0.4229, "step": 18735 }, { "epoch": 1.89, "learning_rate": 3.2307127802368624e-05, "loss": 0.3403, "step": 18736 }, { "epoch": 1.89, "learning_rate": 3.2304741713492526e-05, "loss": 0.3752, "step": 18737 }, { "epoch": 1.89, "learning_rate": 3.2302355609949924e-05, "loss": 0.4412, "step": 18738 }, { "epoch": 1.89, "learning_rate": 3.229996949175602e-05, "loss": 1.0245, "step": 18739 }, { "epoch": 1.89, "learning_rate": 3.229758335892598e-05, "loss": 0.2597, "step": 18740 }, { "epoch": 1.89, "learning_rate": 3.229519721147499e-05, "loss": 1.0432, "step": 18741 }, { "epoch": 1.89, "learning_rate": 3.229281104941824e-05, "loss": 0.4875, "step": 18742 }, { "epoch": 1.89, "learning_rate": 3.229042487277091e-05, "loss": 0.2147, "step": 18743 }, { "epoch": 1.89, "learning_rate": 3.22880386815482e-05, "loss": 0.6742, "step": 18744 }, { "epoch": 1.89, "learning_rate": 3.2285652475765264e-05, "loss": 0.514, "step": 18745 }, { "epoch": 1.89, "learning_rate": 3.228326625543732e-05, "loss": 0.4265, "step": 18746 }, { "epoch": 1.89, "learning_rate": 3.228088002057952e-05, "loss": 0.5398, "step": 18747 }, { "epoch": 1.89, "learning_rate": 3.2278493771207087e-05, "loss": 0.9532, "step": 18748 }, { "epoch": 1.89, "learning_rate": 3.2276107507335175e-05, "loss": 0.468, "step": 18749 }, { "epoch": 1.89, "learning_rate": 3.2273721228978986e-05, "loss": 0.5474, "step": 18750 }, { "epoch": 1.89, "learning_rate": 3.2271334936153705e-05, "loss": 0.3731, "step": 18751 }, { "epoch": 1.89, "learning_rate": 3.22689486288745e-05, "loss": 0.4447, "step": 18752 }, { "epoch": 1.89, "learning_rate": 3.226656230715657e-05, "loss": 0.5544, "step": 18753 }, { "epoch": 1.89, "learning_rate": 3.226417597101511e-05, "loss": 0.7698, "step": 18754 }, { "epoch": 1.89, "learning_rate": 3.2261789620465295e-05, "loss": 0.6614, "step": 18755 }, { "epoch": 1.89, "learning_rate": 3.22594032555223e-05, "loss": 0.5746, "step": 18756 }, { "epoch": 1.89, "learning_rate": 3.225701687620133e-05, "loss": 0.1817, "step": 18757 }, { "epoch": 1.89, "learning_rate": 3.225463048251756e-05, "loss": 1.4998, "step": 18758 }, { "epoch": 1.89, "learning_rate": 3.2252244074486176e-05, "loss": 1.001, "step": 18759 }, { "epoch": 1.89, "learning_rate": 3.224985765212237e-05, "loss": 1.2521, "step": 18760 }, { "epoch": 1.89, "learning_rate": 3.224747121544133e-05, "loss": 0.9607, "step": 18761 }, { "epoch": 1.89, "learning_rate": 3.2245084764458236e-05, "loss": 0.1996, "step": 18762 }, { "epoch": 1.89, "learning_rate": 3.224269829918828e-05, "loss": 1.314, "step": 18763 }, { "epoch": 1.89, "learning_rate": 3.2240311819646645e-05, "loss": 0.7739, "step": 18764 }, { "epoch": 1.89, "learning_rate": 3.223792532584852e-05, "loss": 0.2575, "step": 18765 }, { "epoch": 1.89, "learning_rate": 3.223553881780908e-05, "loss": 0.3474, "step": 18766 }, { "epoch": 1.89, "learning_rate": 3.223315229554353e-05, "loss": 1.2254, "step": 18767 }, { "epoch": 1.89, "learning_rate": 3.223076575906704e-05, "loss": 0.5798, "step": 18768 }, { "epoch": 1.89, "learning_rate": 3.2228379208394805e-05, "loss": 0.522, "step": 18769 }, { "epoch": 1.89, "learning_rate": 3.222599264354202e-05, "loss": 0.3234, "step": 18770 }, { "epoch": 1.89, "learning_rate": 3.2223606064523864e-05, "loss": 1.1119, "step": 18771 }, { "epoch": 1.89, "learning_rate": 3.222121947135553e-05, "loss": 0.8004, "step": 18772 }, { "epoch": 1.89, "learning_rate": 3.221883286405219e-05, "loss": 0.7001, "step": 18773 }, { "epoch": 1.89, "learning_rate": 3.221644624262905e-05, "loss": 0.4279, "step": 18774 }, { "epoch": 1.89, "learning_rate": 3.221405960710127e-05, "loss": 0.7437, "step": 18775 }, { "epoch": 1.89, "learning_rate": 3.221167295748408e-05, "loss": 0.2352, "step": 18776 }, { "epoch": 1.89, "learning_rate": 3.220928629379263e-05, "loss": 0.4924, "step": 18777 }, { "epoch": 1.89, "learning_rate": 3.220689961604213e-05, "loss": 0.3723, "step": 18778 }, { "epoch": 1.89, "learning_rate": 3.220451292424776e-05, "loss": 0.3125, "step": 18779 }, { "epoch": 1.89, "learning_rate": 3.220212621842471e-05, "loss": 0.7469, "step": 18780 }, { "epoch": 1.89, "learning_rate": 3.2199739498588166e-05, "loss": 0.0293, "step": 18781 }, { "epoch": 1.89, "learning_rate": 3.21973527647533e-05, "loss": 1.2403, "step": 18782 }, { "epoch": 1.89, "learning_rate": 3.2194966016935346e-05, "loss": 0.6671, "step": 18783 }, { "epoch": 1.89, "learning_rate": 3.219257925514944e-05, "loss": 0.4891, "step": 18784 }, { "epoch": 1.89, "learning_rate": 3.21901924794108e-05, "loss": 0.4537, "step": 18785 }, { "epoch": 1.89, "learning_rate": 3.2187805689734617e-05, "loss": 0.8343, "step": 18786 }, { "epoch": 1.89, "learning_rate": 3.2185418886136055e-05, "loss": 0.5135, "step": 18787 }, { "epoch": 1.89, "learning_rate": 3.218303206863033e-05, "loss": 0.5406, "step": 18788 }, { "epoch": 1.89, "learning_rate": 3.2180645237232615e-05, "loss": 1.2596, "step": 18789 }, { "epoch": 1.9, "learning_rate": 3.217825839195811e-05, "loss": 1.0594, "step": 18790 }, { "epoch": 1.9, "learning_rate": 3.2175871532821995e-05, "loss": 1.085, "step": 18791 }, { "epoch": 1.9, "learning_rate": 3.2173484659839454e-05, "loss": 0.4787, "step": 18792 }, { "epoch": 1.9, "learning_rate": 3.217109777302569e-05, "loss": 0.4614, "step": 18793 }, { "epoch": 1.9, "learning_rate": 3.216871087239588e-05, "loss": 0.376, "step": 18794 }, { "epoch": 1.9, "learning_rate": 3.216632395796523e-05, "loss": 0.7922, "step": 18795 }, { "epoch": 1.9, "learning_rate": 3.216393702974891e-05, "loss": 1.1791, "step": 18796 }, { "epoch": 1.9, "learning_rate": 3.2161550087762124e-05, "loss": 0.2486, "step": 18797 }, { "epoch": 1.9, "learning_rate": 3.215916313202004e-05, "loss": 0.8929, "step": 18798 }, { "epoch": 1.9, "learning_rate": 3.215677616253788e-05, "loss": 0.7292, "step": 18799 }, { "epoch": 1.9, "learning_rate": 3.2154389179330815e-05, "loss": 0.5495, "step": 18800 }, { "epoch": 1.9, "learning_rate": 3.2152002182414026e-05, "loss": 0.028, "step": 18801 }, { "epoch": 1.9, "learning_rate": 3.214961517180272e-05, "loss": 1.019, "step": 18802 }, { "epoch": 1.9, "learning_rate": 3.214722814751209e-05, "loss": 0.1091, "step": 18803 }, { "epoch": 1.9, "learning_rate": 3.214484110955731e-05, "loss": 0.4638, "step": 18804 }, { "epoch": 1.9, "learning_rate": 3.214245405795357e-05, "loss": 0.7444, "step": 18805 }, { "epoch": 1.9, "learning_rate": 3.2140066992716076e-05, "loss": 0.177, "step": 18806 }, { "epoch": 1.9, "learning_rate": 3.2137679913859996e-05, "loss": 0.6146, "step": 18807 }, { "epoch": 1.9, "learning_rate": 3.213529282140055e-05, "loss": 0.3436, "step": 18808 }, { "epoch": 1.9, "learning_rate": 3.2132905715352906e-05, "loss": 0.8593, "step": 18809 }, { "epoch": 1.9, "learning_rate": 3.213051859573227e-05, "loss": 0.4525, "step": 18810 }, { "epoch": 1.9, "learning_rate": 3.2128131462553816e-05, "loss": 0.3065, "step": 18811 }, { "epoch": 1.9, "learning_rate": 3.2125744315832745e-05, "loss": 0.9615, "step": 18812 }, { "epoch": 1.9, "learning_rate": 3.2123357155584246e-05, "loss": 1.3374, "step": 18813 }, { "epoch": 1.9, "learning_rate": 3.2120969981823504e-05, "loss": 0.6878, "step": 18814 }, { "epoch": 1.9, "learning_rate": 3.211858279456572e-05, "loss": 0.2427, "step": 18815 }, { "epoch": 1.9, "learning_rate": 3.2116195593826084e-05, "loss": 0.3324, "step": 18816 }, { "epoch": 1.9, "learning_rate": 3.211380837961978e-05, "loss": 0.5097, "step": 18817 }, { "epoch": 1.9, "learning_rate": 3.211142115196201e-05, "loss": 0.3203, "step": 18818 }, { "epoch": 1.9, "learning_rate": 3.2109033910867956e-05, "loss": 0.8238, "step": 18819 }, { "epoch": 1.9, "learning_rate": 3.2106646656352815e-05, "loss": 1.0496, "step": 18820 }, { "epoch": 1.9, "learning_rate": 3.2104259388431764e-05, "loss": 0.4577, "step": 18821 }, { "epoch": 1.9, "learning_rate": 3.2101872107120024e-05, "loss": 1.2111, "step": 18822 }, { "epoch": 1.9, "learning_rate": 3.209948481243276e-05, "loss": 0.758, "step": 18823 }, { "epoch": 1.9, "learning_rate": 3.209709750438517e-05, "loss": 0.2509, "step": 18824 }, { "epoch": 1.9, "learning_rate": 3.209471018299245e-05, "loss": 0.6705, "step": 18825 }, { "epoch": 1.9, "learning_rate": 3.2092322848269796e-05, "loss": 0.5404, "step": 18826 }, { "epoch": 1.9, "learning_rate": 3.20899355002324e-05, "loss": 0.4279, "step": 18827 }, { "epoch": 1.9, "learning_rate": 3.208754813889544e-05, "loss": 0.7403, "step": 18828 }, { "epoch": 1.9, "learning_rate": 3.208516076427412e-05, "loss": 0.3658, "step": 18829 }, { "epoch": 1.9, "learning_rate": 3.2082773376383634e-05, "loss": 0.8655, "step": 18830 }, { "epoch": 1.9, "learning_rate": 3.208038597523916e-05, "loss": 1.1304, "step": 18831 }, { "epoch": 1.9, "learning_rate": 3.2077998560855915e-05, "loss": 1.0004, "step": 18832 }, { "epoch": 1.9, "learning_rate": 3.207561113324907e-05, "loss": 1.241, "step": 18833 }, { "epoch": 1.9, "learning_rate": 3.207322369243383e-05, "loss": 0.2431, "step": 18834 }, { "epoch": 1.9, "learning_rate": 3.207083623842538e-05, "loss": 0.8929, "step": 18835 }, { "epoch": 1.9, "learning_rate": 3.206844877123892e-05, "loss": 0.6108, "step": 18836 }, { "epoch": 1.9, "learning_rate": 3.206606129088963e-05, "loss": 0.6349, "step": 18837 }, { "epoch": 1.9, "learning_rate": 3.206367379739272e-05, "loss": 0.3467, "step": 18838 }, { "epoch": 1.9, "learning_rate": 3.206128629076337e-05, "loss": 0.1595, "step": 18839 }, { "epoch": 1.9, "learning_rate": 3.2058898771016775e-05, "loss": 1.1079, "step": 18840 }, { "epoch": 1.9, "learning_rate": 3.205651123816814e-05, "loss": 1.8629, "step": 18841 }, { "epoch": 1.9, "learning_rate": 3.205412369223265e-05, "loss": 0.6345, "step": 18842 }, { "epoch": 1.9, "learning_rate": 3.205173613322549e-05, "loss": 1.1869, "step": 18843 }, { "epoch": 1.9, "learning_rate": 3.2049348561161865e-05, "loss": 1.0098, "step": 18844 }, { "epoch": 1.9, "learning_rate": 3.2046960976056974e-05, "loss": 0.2643, "step": 18845 }, { "epoch": 1.9, "learning_rate": 3.204457337792599e-05, "loss": 1.3699, "step": 18846 }, { "epoch": 1.9, "learning_rate": 3.2042185766784126e-05, "loss": 1.0298, "step": 18847 }, { "epoch": 1.9, "learning_rate": 3.203979814264656e-05, "loss": 0.4017, "step": 18848 }, { "epoch": 1.9, "learning_rate": 3.203741050552851e-05, "loss": 0.6805, "step": 18849 }, { "epoch": 1.9, "learning_rate": 3.2035022855445144e-05, "loss": 0.3088, "step": 18850 }, { "epoch": 1.9, "learning_rate": 3.2032635192411663e-05, "loss": 0.2555, "step": 18851 }, { "epoch": 1.9, "learning_rate": 3.203024751644328e-05, "loss": 0.9447, "step": 18852 }, { "epoch": 1.9, "learning_rate": 3.2027859827555154e-05, "loss": 0.4423, "step": 18853 }, { "epoch": 1.9, "learning_rate": 3.2025472125762516e-05, "loss": 0.3485, "step": 18854 }, { "epoch": 1.9, "learning_rate": 3.202308441108054e-05, "loss": 0.3718, "step": 18855 }, { "epoch": 1.9, "learning_rate": 3.2020696683524417e-05, "loss": 0.2274, "step": 18856 }, { "epoch": 1.9, "learning_rate": 3.201830894310935e-05, "loss": 0.6273, "step": 18857 }, { "epoch": 1.9, "learning_rate": 3.2015921189850546e-05, "loss": 0.4266, "step": 18858 }, { "epoch": 1.9, "learning_rate": 3.2013533423763175e-05, "loss": 0.57, "step": 18859 }, { "epoch": 1.9, "learning_rate": 3.2011145644862445e-05, "loss": 0.5551, "step": 18860 }, { "epoch": 1.9, "learning_rate": 3.200875785316355e-05, "loss": 0.2108, "step": 18861 }, { "epoch": 1.9, "learning_rate": 3.2006370048681684e-05, "loss": 0.5359, "step": 18862 }, { "epoch": 1.9, "learning_rate": 3.2003982231432045e-05, "loss": 0.7636, "step": 18863 }, { "epoch": 1.9, "learning_rate": 3.200159440142982e-05, "loss": 0.8039, "step": 18864 }, { "epoch": 1.9, "learning_rate": 3.199920655869021e-05, "loss": 0.7355, "step": 18865 }, { "epoch": 1.9, "learning_rate": 3.199681870322841e-05, "loss": 0.3536, "step": 18866 }, { "epoch": 1.9, "learning_rate": 3.1994430835059623e-05, "loss": 0.3551, "step": 18867 }, { "epoch": 1.9, "learning_rate": 3.199204295419903e-05, "loss": 0.4389, "step": 18868 }, { "epoch": 1.9, "learning_rate": 3.1989655060661837e-05, "loss": 1.135, "step": 18869 }, { "epoch": 1.9, "learning_rate": 3.1987267154463235e-05, "loss": 0.4786, "step": 18870 }, { "epoch": 1.9, "learning_rate": 3.198487923561841e-05, "loss": 0.1686, "step": 18871 }, { "epoch": 1.9, "learning_rate": 3.1982491304142585e-05, "loss": 0.0471, "step": 18872 }, { "epoch": 1.9, "learning_rate": 3.1980103360050935e-05, "loss": 1.0534, "step": 18873 }, { "epoch": 1.9, "learning_rate": 3.197771540335866e-05, "loss": 0.4752, "step": 18874 }, { "epoch": 1.9, "learning_rate": 3.197532743408096e-05, "loss": 0.5904, "step": 18875 }, { "epoch": 1.9, "learning_rate": 3.197293945223301e-05, "loss": 0.5077, "step": 18876 }, { "epoch": 1.9, "learning_rate": 3.1970551457830044e-05, "loss": 1.4627, "step": 18877 }, { "epoch": 1.9, "learning_rate": 3.196816345088723e-05, "loss": 0.6403, "step": 18878 }, { "epoch": 1.9, "learning_rate": 3.1965775431419775e-05, "loss": 0.2155, "step": 18879 }, { "epoch": 1.9, "learning_rate": 3.1963387399442866e-05, "loss": 0.8385, "step": 18880 }, { "epoch": 1.9, "learning_rate": 3.196099935497171e-05, "loss": 1.2138, "step": 18881 }, { "epoch": 1.9, "learning_rate": 3.195861129802151e-05, "loss": 1.0431, "step": 18882 }, { "epoch": 1.9, "learning_rate": 3.195622322860744e-05, "loss": 0.9431, "step": 18883 }, { "epoch": 1.9, "learning_rate": 3.195383514674472e-05, "loss": 0.6439, "step": 18884 }, { "epoch": 1.9, "learning_rate": 3.1951447052448526e-05, "loss": 0.5584, "step": 18885 }, { "epoch": 1.9, "learning_rate": 3.194905894573407e-05, "loss": 0.1769, "step": 18886 }, { "epoch": 1.9, "learning_rate": 3.194667082661655e-05, "loss": 1.1217, "step": 18887 }, { "epoch": 1.9, "learning_rate": 3.1944282695111146e-05, "loss": 0.1804, "step": 18888 }, { "epoch": 1.91, "learning_rate": 3.1941894551233074e-05, "loss": 0.6742, "step": 18889 }, { "epoch": 1.91, "learning_rate": 3.193950639499752e-05, "loss": 0.3096, "step": 18890 }, { "epoch": 1.91, "learning_rate": 3.193711822641969e-05, "loss": 0.5175, "step": 18891 }, { "epoch": 1.91, "learning_rate": 3.1934730045514766e-05, "loss": 0.9479, "step": 18892 }, { "epoch": 1.91, "learning_rate": 3.193234185229798e-05, "loss": 0.5114, "step": 18893 }, { "epoch": 1.91, "learning_rate": 3.1929953646784484e-05, "loss": 0.9424, "step": 18894 }, { "epoch": 1.91, "learning_rate": 3.19275654289895e-05, "loss": 0.4951, "step": 18895 }, { "epoch": 1.91, "learning_rate": 3.192517719892822e-05, "loss": 0.096, "step": 18896 }, { "epoch": 1.91, "learning_rate": 3.1922788956615855e-05, "loss": 0.7037, "step": 18897 }, { "epoch": 1.91, "learning_rate": 3.192040070206759e-05, "loss": 0.9606, "step": 18898 }, { "epoch": 1.91, "learning_rate": 3.191801243529862e-05, "loss": 0.6973, "step": 18899 }, { "epoch": 1.91, "learning_rate": 3.1915624156324154e-05, "loss": 0.6452, "step": 18900 }, { "epoch": 1.91, "learning_rate": 3.191323586515938e-05, "loss": 0.679, "step": 18901 }, { "epoch": 1.91, "learning_rate": 3.1910847561819506e-05, "loss": 0.8223, "step": 18902 }, { "epoch": 1.91, "learning_rate": 3.1908459246319725e-05, "loss": 0.3578, "step": 18903 }, { "epoch": 1.91, "learning_rate": 3.190607091867523e-05, "loss": 0.6551, "step": 18904 }, { "epoch": 1.91, "learning_rate": 3.190368257890123e-05, "loss": 0.3471, "step": 18905 }, { "epoch": 1.91, "learning_rate": 3.190129422701291e-05, "loss": 1.1911, "step": 18906 }, { "epoch": 1.91, "learning_rate": 3.189890586302549e-05, "loss": 0.3023, "step": 18907 }, { "epoch": 1.91, "learning_rate": 3.189651748695414e-05, "loss": 0.4326, "step": 18908 }, { "epoch": 1.91, "learning_rate": 3.189412909881409e-05, "loss": 0.2526, "step": 18909 }, { "epoch": 1.91, "learning_rate": 3.189174069862051e-05, "loss": 0.844, "step": 18910 }, { "epoch": 1.91, "learning_rate": 3.188935228638861e-05, "loss": 0.4476, "step": 18911 }, { "epoch": 1.91, "learning_rate": 3.188696386213361e-05, "loss": 0.4993, "step": 18912 }, { "epoch": 1.91, "learning_rate": 3.188457542587067e-05, "loss": 0.853, "step": 18913 }, { "epoch": 1.91, "learning_rate": 3.1882186977615015e-05, "loss": 0.1906, "step": 18914 }, { "epoch": 1.91, "learning_rate": 3.187979851738184e-05, "loss": 0.9445, "step": 18915 }, { "epoch": 1.91, "learning_rate": 3.187741004518634e-05, "loss": 0.8789, "step": 18916 }, { "epoch": 1.91, "learning_rate": 3.187502156104372e-05, "loss": 0.9967, "step": 18917 }, { "epoch": 1.91, "learning_rate": 3.1872633064969176e-05, "loss": 1.1366, "step": 18918 }, { "epoch": 1.91, "learning_rate": 3.187024455697791e-05, "loss": 0.2479, "step": 18919 }, { "epoch": 1.91, "learning_rate": 3.1867856037085115e-05, "loss": 0.3675, "step": 18920 }, { "epoch": 1.91, "learning_rate": 3.1865467505305993e-05, "loss": 0.4619, "step": 18921 }, { "epoch": 1.91, "learning_rate": 3.186307896165575e-05, "loss": 0.9328, "step": 18922 }, { "epoch": 1.91, "learning_rate": 3.186069040614958e-05, "loss": 1.0139, "step": 18923 }, { "epoch": 1.91, "learning_rate": 3.1858301838802676e-05, "loss": 0.4459, "step": 18924 }, { "epoch": 1.91, "learning_rate": 3.185591325963026e-05, "loss": 0.8868, "step": 18925 }, { "epoch": 1.91, "learning_rate": 3.1853524668647505e-05, "loss": 0.0666, "step": 18926 }, { "epoch": 1.91, "learning_rate": 3.1851136065869635e-05, "loss": 0.9047, "step": 18927 }, { "epoch": 1.91, "learning_rate": 3.1848747451311833e-05, "loss": 0.56, "step": 18928 }, { "epoch": 1.91, "learning_rate": 3.184635882498931e-05, "loss": 0.2111, "step": 18929 }, { "epoch": 1.91, "learning_rate": 3.1843970186917254e-05, "loss": 0.8742, "step": 18930 }, { "epoch": 1.91, "learning_rate": 3.184158153711087e-05, "loss": 0.7359, "step": 18931 }, { "epoch": 1.91, "learning_rate": 3.183919287558538e-05, "loss": 1.2524, "step": 18932 }, { "epoch": 1.91, "learning_rate": 3.183680420235595e-05, "loss": 0.3367, "step": 18933 }, { "epoch": 1.91, "learning_rate": 3.18344155174378e-05, "loss": 0.3069, "step": 18934 }, { "epoch": 1.91, "learning_rate": 3.183202682084614e-05, "loss": 0.2345, "step": 18935 }, { "epoch": 1.91, "learning_rate": 3.1829638112596146e-05, "loss": 0.4856, "step": 18936 }, { "epoch": 1.91, "learning_rate": 3.1827249392703034e-05, "loss": 0.4704, "step": 18937 }, { "epoch": 1.91, "learning_rate": 3.1824860661182e-05, "loss": 0.5609, "step": 18938 }, { "epoch": 1.91, "learning_rate": 3.1822471918048254e-05, "loss": 0.2129, "step": 18939 }, { "epoch": 1.91, "learning_rate": 3.1820083163316975e-05, "loss": 1.0393, "step": 18940 }, { "epoch": 1.91, "learning_rate": 3.181769439700339e-05, "loss": 0.3428, "step": 18941 }, { "epoch": 1.91, "learning_rate": 3.181530561912269e-05, "loss": 0.8514, "step": 18942 }, { "epoch": 1.91, "learning_rate": 3.1812916829690075e-05, "loss": 0.3152, "step": 18943 }, { "epoch": 1.91, "learning_rate": 3.1810528028720746e-05, "loss": 0.2696, "step": 18944 }, { "epoch": 1.91, "learning_rate": 3.1808139216229905e-05, "loss": 0.306, "step": 18945 }, { "epoch": 1.91, "learning_rate": 3.180575039223275e-05, "loss": 0.5928, "step": 18946 }, { "epoch": 1.91, "learning_rate": 3.1803361556744485e-05, "loss": 0.4545, "step": 18947 }, { "epoch": 1.91, "learning_rate": 3.1800972709780324e-05, "loss": 0.8925, "step": 18948 }, { "epoch": 1.91, "learning_rate": 3.179858385135545e-05, "loss": 0.7614, "step": 18949 }, { "epoch": 1.91, "learning_rate": 3.179619498148507e-05, "loss": 0.4027, "step": 18950 }, { "epoch": 1.91, "learning_rate": 3.17938061001844e-05, "loss": 0.2652, "step": 18951 }, { "epoch": 1.91, "learning_rate": 3.179141720746862e-05, "loss": 0.2575, "step": 18952 }, { "epoch": 1.91, "learning_rate": 3.178902830335295e-05, "loss": 0.2801, "step": 18953 }, { "epoch": 1.91, "learning_rate": 3.1786639387852576e-05, "loss": 0.8501, "step": 18954 }, { "epoch": 1.91, "learning_rate": 3.1784250460982716e-05, "loss": 0.244, "step": 18955 }, { "epoch": 1.91, "learning_rate": 3.1781861522758555e-05, "loss": 0.3896, "step": 18956 }, { "epoch": 1.91, "learning_rate": 3.177947257319532e-05, "loss": 1.2311, "step": 18957 }, { "epoch": 1.91, "learning_rate": 3.177708361230819e-05, "loss": 0.4906, "step": 18958 }, { "epoch": 1.91, "learning_rate": 3.1774694640112376e-05, "loss": 0.9523, "step": 18959 }, { "epoch": 1.91, "learning_rate": 3.1772305656623086e-05, "loss": 0.2329, "step": 18960 }, { "epoch": 1.91, "learning_rate": 3.1769916661855505e-05, "loss": 0.985, "step": 18961 }, { "epoch": 1.91, "learning_rate": 3.176752765582486e-05, "loss": 0.1991, "step": 18962 }, { "epoch": 1.91, "learning_rate": 3.176513863854633e-05, "loss": 1.077, "step": 18963 }, { "epoch": 1.91, "learning_rate": 3.176274961003513e-05, "loss": 0.9642, "step": 18964 }, { "epoch": 1.91, "learning_rate": 3.1760360570306476e-05, "loss": 0.986, "step": 18965 }, { "epoch": 1.91, "learning_rate": 3.175797151937555e-05, "loss": 0.5675, "step": 18966 }, { "epoch": 1.91, "learning_rate": 3.175558245725756e-05, "loss": 0.6543, "step": 18967 }, { "epoch": 1.91, "learning_rate": 3.1753193383967706e-05, "loss": 0.4732, "step": 18968 }, { "epoch": 1.91, "learning_rate": 3.17508042995212e-05, "loss": 0.8008, "step": 18969 }, { "epoch": 1.91, "learning_rate": 3.174841520393324e-05, "loss": 0.5724, "step": 18970 }, { "epoch": 1.91, "learning_rate": 3.174602609721904e-05, "loss": 0.6715, "step": 18971 }, { "epoch": 1.91, "learning_rate": 3.174363697939379e-05, "loss": 0.9736, "step": 18972 }, { "epoch": 1.91, "learning_rate": 3.1741247850472696e-05, "loss": 0.3264, "step": 18973 }, { "epoch": 1.91, "learning_rate": 3.173885871047096e-05, "loss": 0.6299, "step": 18974 }, { "epoch": 1.91, "learning_rate": 3.1736469559403786e-05, "loss": 0.622, "step": 18975 }, { "epoch": 1.91, "learning_rate": 3.1734080397286394e-05, "loss": 1.0689, "step": 18976 }, { "epoch": 1.91, "learning_rate": 3.173169122413395e-05, "loss": 0.7686, "step": 18977 }, { "epoch": 1.91, "learning_rate": 3.1729302039961713e-05, "loss": 0.7102, "step": 18978 }, { "epoch": 1.91, "learning_rate": 3.172691284478483e-05, "loss": 0.7658, "step": 18979 }, { "epoch": 1.91, "learning_rate": 3.1724523638618546e-05, "loss": 0.27, "step": 18980 }, { "epoch": 1.91, "learning_rate": 3.172213442147804e-05, "loss": 0.7794, "step": 18981 }, { "epoch": 1.91, "learning_rate": 3.1719745193378534e-05, "loss": 0.6746, "step": 18982 }, { "epoch": 1.91, "learning_rate": 3.1717355954335216e-05, "loss": 0.9568, "step": 18983 }, { "epoch": 1.91, "learning_rate": 3.17149667043633e-05, "loss": 0.7136, "step": 18984 }, { "epoch": 1.91, "learning_rate": 3.1712577443478e-05, "loss": 1.0669, "step": 18985 }, { "epoch": 1.91, "learning_rate": 3.1710188171694494e-05, "loss": 0.5608, "step": 18986 }, { "epoch": 1.92, "learning_rate": 3.170779888902801e-05, "loss": 0.8476, "step": 18987 }, { "epoch": 1.92, "learning_rate": 3.1705409595493735e-05, "loss": 0.9477, "step": 18988 }, { "epoch": 1.92, "learning_rate": 3.170302029110689e-05, "loss": 0.3362, "step": 18989 }, { "epoch": 1.92, "learning_rate": 3.1700630975882666e-05, "loss": 0.467, "step": 18990 }, { "epoch": 1.92, "learning_rate": 3.169824164983628e-05, "loss": 0.6854, "step": 18991 }, { "epoch": 1.92, "learning_rate": 3.169585231298293e-05, "loss": 0.535, "step": 18992 }, { "epoch": 1.92, "learning_rate": 3.16934629653378e-05, "loss": 0.383, "step": 18993 }, { "epoch": 1.92, "learning_rate": 3.169107360691614e-05, "loss": 0.8642, "step": 18994 }, { "epoch": 1.92, "learning_rate": 3.168868423773312e-05, "loss": 0.4035, "step": 18995 }, { "epoch": 1.92, "learning_rate": 3.168629485780396e-05, "loss": 0.7156, "step": 18996 }, { "epoch": 1.92, "learning_rate": 3.168390546714386e-05, "loss": 0.478, "step": 18997 }, { "epoch": 1.92, "learning_rate": 3.168151606576803e-05, "loss": 0.3036, "step": 18998 }, { "epoch": 1.92, "learning_rate": 3.167912665369167e-05, "loss": 0.7586, "step": 18999 }, { "epoch": 1.92, "learning_rate": 3.167673723092998e-05, "loss": 0.2228, "step": 19000 }, { "epoch": 1.92, "learning_rate": 3.167434779749819e-05, "loss": 0.9223, "step": 19001 }, { "epoch": 1.92, "learning_rate": 3.1671958353411456e-05, "loss": 0.5786, "step": 19002 }, { "epoch": 1.92, "learning_rate": 3.166956889868504e-05, "loss": 0.4244, "step": 19003 }, { "epoch": 1.92, "learning_rate": 3.166717943333412e-05, "loss": 0.8286, "step": 19004 }, { "epoch": 1.92, "learning_rate": 3.16647899573739e-05, "loss": 0.8335, "step": 19005 }, { "epoch": 1.92, "learning_rate": 3.166240047081959e-05, "loss": 0.9263, "step": 19006 }, { "epoch": 1.92, "learning_rate": 3.1660010973686395e-05, "loss": 0.2164, "step": 19007 }, { "epoch": 1.92, "learning_rate": 3.165762146598953e-05, "loss": 1.3209, "step": 19008 }, { "epoch": 1.92, "learning_rate": 3.1655231947744175e-05, "loss": 0.7293, "step": 19009 }, { "epoch": 1.92, "learning_rate": 3.1652842418965573e-05, "loss": 1.0841, "step": 19010 }, { "epoch": 1.92, "learning_rate": 3.1650452879668896e-05, "loss": 0.1691, "step": 19011 }, { "epoch": 1.92, "learning_rate": 3.1648063329869377e-05, "loss": 1.2936, "step": 19012 }, { "epoch": 1.92, "learning_rate": 3.16456737695822e-05, "loss": 0.2657, "step": 19013 }, { "epoch": 1.92, "learning_rate": 3.164328419882258e-05, "loss": 0.8614, "step": 19014 }, { "epoch": 1.92, "learning_rate": 3.164089461760573e-05, "loss": 0.1978, "step": 19015 }, { "epoch": 1.92, "learning_rate": 3.1638505025946854e-05, "loss": 0.7981, "step": 19016 }, { "epoch": 1.92, "learning_rate": 3.1636115423861156e-05, "loss": 0.2295, "step": 19017 }, { "epoch": 1.92, "learning_rate": 3.163372581136383e-05, "loss": 0.4409, "step": 19018 }, { "epoch": 1.92, "learning_rate": 3.1631336188470104e-05, "loss": 1.0595, "step": 19019 }, { "epoch": 1.92, "learning_rate": 3.1628946555195176e-05, "loss": 0.4971, "step": 19020 }, { "epoch": 1.92, "learning_rate": 3.1626556911554245e-05, "loss": 0.7224, "step": 19021 }, { "epoch": 1.92, "learning_rate": 3.162416725756253e-05, "loss": 0.3434, "step": 19022 }, { "epoch": 1.92, "learning_rate": 3.162177759323523e-05, "loss": 0.7634, "step": 19023 }, { "epoch": 1.92, "learning_rate": 3.1619387918587555e-05, "loss": 0.6402, "step": 19024 }, { "epoch": 1.92, "learning_rate": 3.1616998233634706e-05, "loss": 0.151, "step": 19025 }, { "epoch": 1.92, "learning_rate": 3.1614608538391904e-05, "loss": 0.7025, "step": 19026 }, { "epoch": 1.92, "learning_rate": 3.161221883287434e-05, "loss": 0.455, "step": 19027 }, { "epoch": 1.92, "learning_rate": 3.160982911709723e-05, "loss": 0.3458, "step": 19028 }, { "epoch": 1.92, "learning_rate": 3.1607439391075794e-05, "loss": 0.679, "step": 19029 }, { "epoch": 1.92, "learning_rate": 3.160504965482521e-05, "loss": 0.5492, "step": 19030 }, { "epoch": 1.92, "learning_rate": 3.160265990836071e-05, "loss": 1.1884, "step": 19031 }, { "epoch": 1.92, "learning_rate": 3.1600270151697476e-05, "loss": 0.709, "step": 19032 }, { "epoch": 1.92, "learning_rate": 3.159788038485075e-05, "loss": 1.2297, "step": 19033 }, { "epoch": 1.92, "learning_rate": 3.159549060783571e-05, "loss": 1.1284, "step": 19034 }, { "epoch": 1.92, "learning_rate": 3.159310082066758e-05, "loss": 0.8924, "step": 19035 }, { "epoch": 1.92, "learning_rate": 3.159071102336156e-05, "loss": 0.7261, "step": 19036 }, { "epoch": 1.92, "learning_rate": 3.158832121593286e-05, "loss": 0.363, "step": 19037 }, { "epoch": 1.92, "learning_rate": 3.1585931398396685e-05, "loss": 0.8771, "step": 19038 }, { "epoch": 1.92, "learning_rate": 3.158354157076826e-05, "loss": 0.3338, "step": 19039 }, { "epoch": 1.92, "learning_rate": 3.158115173306277e-05, "loss": 0.9169, "step": 19040 }, { "epoch": 1.92, "learning_rate": 3.157876188529542e-05, "loss": 0.9957, "step": 19041 }, { "epoch": 1.92, "learning_rate": 3.157637202748145e-05, "loss": 0.2782, "step": 19042 }, { "epoch": 1.92, "learning_rate": 3.157398215963603e-05, "loss": 0.3677, "step": 19043 }, { "epoch": 1.92, "learning_rate": 3.157159228177439e-05, "loss": 0.2416, "step": 19044 }, { "epoch": 1.92, "learning_rate": 3.1569202393911745e-05, "loss": 0.5139, "step": 19045 }, { "epoch": 1.92, "learning_rate": 3.156681249606329e-05, "loss": 1.0944, "step": 19046 }, { "epoch": 1.92, "learning_rate": 3.1564422588244235e-05, "loss": 0.3016, "step": 19047 }, { "epoch": 1.92, "learning_rate": 3.156203267046978e-05, "loss": 0.0756, "step": 19048 }, { "epoch": 1.92, "learning_rate": 3.155964274275516e-05, "loss": 0.7743, "step": 19049 }, { "epoch": 1.92, "learning_rate": 3.155725280511555e-05, "loss": 0.6325, "step": 19050 }, { "epoch": 1.92, "learning_rate": 3.155486285756619e-05, "loss": 0.2058, "step": 19051 }, { "epoch": 1.92, "learning_rate": 3.1552472900122265e-05, "loss": 1.0788, "step": 19052 }, { "epoch": 1.92, "learning_rate": 3.1550082932798996e-05, "loss": 0.183, "step": 19053 }, { "epoch": 1.92, "learning_rate": 3.154769295561159e-05, "loss": 0.4399, "step": 19054 }, { "epoch": 1.92, "learning_rate": 3.154530296857526e-05, "loss": 0.8089, "step": 19055 }, { "epoch": 1.92, "learning_rate": 3.15429129717052e-05, "loss": 0.2769, "step": 19056 }, { "epoch": 1.92, "learning_rate": 3.154052296501663e-05, "loss": 1.5821, "step": 19057 }, { "epoch": 1.92, "learning_rate": 3.153813294852476e-05, "loss": 0.1559, "step": 19058 }, { "epoch": 1.92, "learning_rate": 3.1535742922244796e-05, "loss": 0.339, "step": 19059 }, { "epoch": 1.92, "learning_rate": 3.153335288619195e-05, "loss": 0.7658, "step": 19060 }, { "epoch": 1.92, "learning_rate": 3.153096284038143e-05, "loss": 0.3068, "step": 19061 }, { "epoch": 1.92, "learning_rate": 3.1528572784828445e-05, "loss": 0.2923, "step": 19062 }, { "epoch": 1.92, "learning_rate": 3.152618271954821e-05, "loss": 2.104, "step": 19063 }, { "epoch": 1.92, "learning_rate": 3.152379264455591e-05, "loss": 0.2973, "step": 19064 }, { "epoch": 1.92, "learning_rate": 3.1521402559866794e-05, "loss": 1.3757, "step": 19065 }, { "epoch": 1.92, "learning_rate": 3.151901246549604e-05, "loss": 0.912, "step": 19066 }, { "epoch": 1.92, "learning_rate": 3.151662236145887e-05, "loss": 0.4704, "step": 19067 }, { "epoch": 1.92, "learning_rate": 3.15142322477705e-05, "loss": 0.5638, "step": 19068 }, { "epoch": 1.92, "learning_rate": 3.151184212444612e-05, "loss": 0.1697, "step": 19069 }, { "epoch": 1.92, "learning_rate": 3.150945199150096e-05, "loss": 1.0086, "step": 19070 }, { "epoch": 1.92, "learning_rate": 3.150706184895022e-05, "loss": 0.2309, "step": 19071 }, { "epoch": 1.92, "learning_rate": 3.150467169680911e-05, "loss": 0.2654, "step": 19072 }, { "epoch": 1.92, "learning_rate": 3.150228153509284e-05, "loss": 1.049, "step": 19073 }, { "epoch": 1.92, "learning_rate": 3.149989136381662e-05, "loss": 0.6201, "step": 19074 }, { "epoch": 1.92, "learning_rate": 3.1497501182995665e-05, "loss": 0.3934, "step": 19075 }, { "epoch": 1.92, "learning_rate": 3.1495110992645185e-05, "loss": 0.1876, "step": 19076 }, { "epoch": 1.92, "learning_rate": 3.149272079278039e-05, "loss": 0.8147, "step": 19077 }, { "epoch": 1.92, "learning_rate": 3.149033058341648e-05, "loss": 1.4815, "step": 19078 }, { "epoch": 1.92, "learning_rate": 3.148794036456868e-05, "loss": 0.5823, "step": 19079 }, { "epoch": 1.92, "learning_rate": 3.148555013625218e-05, "loss": 0.7436, "step": 19080 }, { "epoch": 1.92, "learning_rate": 3.148315989848223e-05, "loss": 0.4416, "step": 19081 }, { "epoch": 1.92, "learning_rate": 3.148076965127399e-05, "loss": 1.0284, "step": 19082 }, { "epoch": 1.92, "learning_rate": 3.147837939464271e-05, "loss": 0.2352, "step": 19083 }, { "epoch": 1.92, "learning_rate": 3.1475989128603574e-05, "loss": 0.6896, "step": 19084 }, { "epoch": 1.92, "learning_rate": 3.147359885317181e-05, "loss": 0.4085, "step": 19085 }, { "epoch": 1.93, "learning_rate": 3.1471208568362624e-05, "loss": 0.2212, "step": 19086 }, { "epoch": 1.93, "learning_rate": 3.146881827419122e-05, "loss": 0.6661, "step": 19087 }, { "epoch": 1.93, "learning_rate": 3.146642797067283e-05, "loss": 0.6952, "step": 19088 }, { "epoch": 1.93, "learning_rate": 3.146403765782264e-05, "loss": 0.5851, "step": 19089 }, { "epoch": 1.93, "learning_rate": 3.146164733565587e-05, "loss": 0.3833, "step": 19090 }, { "epoch": 1.93, "learning_rate": 3.145925700418774e-05, "loss": 0.6799, "step": 19091 }, { "epoch": 1.93, "learning_rate": 3.1456866663433454e-05, "loss": 0.8368, "step": 19092 }, { "epoch": 1.93, "learning_rate": 3.145447631340822e-05, "loss": 0.5819, "step": 19093 }, { "epoch": 1.93, "learning_rate": 3.145208595412724e-05, "loss": 1.0211, "step": 19094 }, { "epoch": 1.93, "learning_rate": 3.144969558560575e-05, "loss": 0.4175, "step": 19095 }, { "epoch": 1.93, "learning_rate": 3.1447305207858944e-05, "loss": 0.6952, "step": 19096 }, { "epoch": 1.93, "learning_rate": 3.144491482090205e-05, "loss": 0.5337, "step": 19097 }, { "epoch": 1.93, "learning_rate": 3.144252442475025e-05, "loss": 1.2915, "step": 19098 }, { "epoch": 1.93, "learning_rate": 3.144013401941878e-05, "loss": 0.4233, "step": 19099 }, { "epoch": 1.93, "learning_rate": 3.143774360492285e-05, "loss": 0.3819, "step": 19100 }, { "epoch": 1.93, "learning_rate": 3.143535318127767e-05, "loss": 0.5463, "step": 19101 }, { "epoch": 1.93, "learning_rate": 3.1432962748498436e-05, "loss": 0.2236, "step": 19102 }, { "epoch": 1.93, "learning_rate": 3.1430572306600374e-05, "loss": 0.3733, "step": 19103 }, { "epoch": 1.93, "learning_rate": 3.1428181855598707e-05, "loss": 0.7871, "step": 19104 }, { "epoch": 1.93, "learning_rate": 3.142579139550861e-05, "loss": 1.1658, "step": 19105 }, { "epoch": 1.93, "learning_rate": 3.142340092634534e-05, "loss": 0.3188, "step": 19106 }, { "epoch": 1.93, "learning_rate": 3.1421010448124085e-05, "loss": 0.389, "step": 19107 }, { "epoch": 1.93, "learning_rate": 3.141861996086005e-05, "loss": 0.567, "step": 19108 }, { "epoch": 1.93, "learning_rate": 3.141622946456847e-05, "loss": 0.9651, "step": 19109 }, { "epoch": 1.93, "learning_rate": 3.141383895926454e-05, "loss": 0.8172, "step": 19110 }, { "epoch": 1.93, "learning_rate": 3.141144844496348e-05, "loss": 0.9751, "step": 19111 }, { "epoch": 1.93, "learning_rate": 3.140905792168048e-05, "loss": 0.7118, "step": 19112 }, { "epoch": 1.93, "learning_rate": 3.140666738943079e-05, "loss": 0.1645, "step": 19113 }, { "epoch": 1.93, "learning_rate": 3.14042768482296e-05, "loss": 0.9594, "step": 19114 }, { "epoch": 1.93, "learning_rate": 3.140188629809213e-05, "loss": 0.5225, "step": 19115 }, { "epoch": 1.93, "learning_rate": 3.1399495739033586e-05, "loss": 0.2192, "step": 19116 }, { "epoch": 1.93, "learning_rate": 3.139710517106918e-05, "loss": 0.536, "step": 19117 }, { "epoch": 1.93, "learning_rate": 3.139471459421413e-05, "loss": 1.0241, "step": 19118 }, { "epoch": 1.93, "learning_rate": 3.1392324008483645e-05, "loss": 0.5956, "step": 19119 }, { "epoch": 1.93, "learning_rate": 3.1389933413892945e-05, "loss": 0.5068, "step": 19120 }, { "epoch": 1.93, "learning_rate": 3.138754281045723e-05, "loss": 0.1478, "step": 19121 }, { "epoch": 1.93, "learning_rate": 3.1385152198191724e-05, "loss": 0.2207, "step": 19122 }, { "epoch": 1.93, "learning_rate": 3.138276157711164e-05, "loss": 0.4693, "step": 19123 }, { "epoch": 1.93, "learning_rate": 3.138037094723219e-05, "loss": 0.7628, "step": 19124 }, { "epoch": 1.93, "learning_rate": 3.1377980308568575e-05, "loss": 0.4652, "step": 19125 }, { "epoch": 1.93, "learning_rate": 3.137558966113602e-05, "loss": 1.0695, "step": 19126 }, { "epoch": 1.93, "learning_rate": 3.137319900494974e-05, "loss": 0.5671, "step": 19127 }, { "epoch": 1.93, "learning_rate": 3.137080834002494e-05, "loss": 0.2015, "step": 19128 }, { "epoch": 1.93, "learning_rate": 3.1368417666376835e-05, "loss": 0.5329, "step": 19129 }, { "epoch": 1.93, "learning_rate": 3.136602698402065e-05, "loss": 0.4513, "step": 19130 }, { "epoch": 1.93, "learning_rate": 3.1363636292971584e-05, "loss": 0.3447, "step": 19131 }, { "epoch": 1.93, "learning_rate": 3.136124559324485e-05, "loss": 1.107, "step": 19132 }, { "epoch": 1.93, "learning_rate": 3.1358854884855675e-05, "loss": 0.614, "step": 19133 }, { "epoch": 1.93, "learning_rate": 3.135646416781926e-05, "loss": 0.0864, "step": 19134 }, { "epoch": 1.93, "learning_rate": 3.135407344215082e-05, "loss": 0.4264, "step": 19135 }, { "epoch": 1.93, "learning_rate": 3.135168270786558e-05, "loss": 0.367, "step": 19136 }, { "epoch": 1.93, "learning_rate": 3.134929196497873e-05, "loss": 0.4329, "step": 19137 }, { "epoch": 1.93, "learning_rate": 3.1346901213505516e-05, "loss": 0.2859, "step": 19138 }, { "epoch": 1.93, "learning_rate": 3.134451045346113e-05, "loss": 0.2961, "step": 19139 }, { "epoch": 1.93, "learning_rate": 3.1342119684860796e-05, "loss": 0.3075, "step": 19140 }, { "epoch": 1.93, "learning_rate": 3.133972890771972e-05, "loss": 0.4919, "step": 19141 }, { "epoch": 1.93, "learning_rate": 3.1337338122053104e-05, "loss": 1.4173, "step": 19142 }, { "epoch": 1.93, "learning_rate": 3.13349473278762e-05, "loss": 1.207, "step": 19143 }, { "epoch": 1.93, "learning_rate": 3.133255652520417e-05, "loss": 1.5543, "step": 19144 }, { "epoch": 1.93, "learning_rate": 3.133016571405228e-05, "loss": 0.5492, "step": 19145 }, { "epoch": 1.93, "learning_rate": 3.132777489443572e-05, "loss": 1.0654, "step": 19146 }, { "epoch": 1.93, "learning_rate": 3.13253840663697e-05, "loss": 0.6501, "step": 19147 }, { "epoch": 1.93, "learning_rate": 3.132299322986945e-05, "loss": 0.6467, "step": 19148 }, { "epoch": 1.93, "learning_rate": 3.132060238495016e-05, "loss": 0.5939, "step": 19149 }, { "epoch": 1.93, "learning_rate": 3.131821153162707e-05, "loss": 0.8824, "step": 19150 }, { "epoch": 1.93, "learning_rate": 3.1315820669915374e-05, "loss": 0.7206, "step": 19151 }, { "epoch": 1.93, "learning_rate": 3.131342979983031e-05, "loss": 0.9648, "step": 19152 }, { "epoch": 1.93, "learning_rate": 3.131103892138706e-05, "loss": 0.4507, "step": 19153 }, { "epoch": 1.93, "learning_rate": 3.1308648034600874e-05, "loss": 0.5695, "step": 19154 }, { "epoch": 1.93, "learning_rate": 3.130625713948693e-05, "loss": 0.2096, "step": 19155 }, { "epoch": 1.93, "learning_rate": 3.130386623606048e-05, "loss": 0.7534, "step": 19156 }, { "epoch": 1.93, "learning_rate": 3.130147532433671e-05, "loss": 0.21, "step": 19157 }, { "epoch": 1.93, "learning_rate": 3.1299084404330855e-05, "loss": 0.9332, "step": 19158 }, { "epoch": 1.93, "learning_rate": 3.129669347605812e-05, "loss": 0.4752, "step": 19159 }, { "epoch": 1.93, "learning_rate": 3.129430253953372e-05, "loss": 0.3102, "step": 19160 }, { "epoch": 1.93, "learning_rate": 3.129191159477287e-05, "loss": 0.6678, "step": 19161 }, { "epoch": 1.93, "learning_rate": 3.1289520641790786e-05, "loss": 0.2688, "step": 19162 }, { "epoch": 1.93, "learning_rate": 3.128712968060268e-05, "loss": 0.4, "step": 19163 }, { "epoch": 1.93, "learning_rate": 3.128473871122378e-05, "loss": 0.7348, "step": 19164 }, { "epoch": 1.93, "learning_rate": 3.128234773366928e-05, "loss": 0.4633, "step": 19165 }, { "epoch": 1.93, "learning_rate": 3.127995674795442e-05, "loss": 0.9234, "step": 19166 }, { "epoch": 1.93, "learning_rate": 3.127756575409438e-05, "loss": 0.912, "step": 19167 }, { "epoch": 1.93, "learning_rate": 3.127517475210442e-05, "loss": 0.4181, "step": 19168 }, { "epoch": 1.93, "learning_rate": 3.127278374199972e-05, "loss": 0.6326, "step": 19169 }, { "epoch": 1.93, "learning_rate": 3.127039272379551e-05, "loss": 0.3409, "step": 19170 }, { "epoch": 1.93, "learning_rate": 3.126800169750701e-05, "loss": 0.9308, "step": 19171 }, { "epoch": 1.93, "learning_rate": 3.126561066314943e-05, "loss": 0.3832, "step": 19172 }, { "epoch": 1.93, "learning_rate": 3.126321962073798e-05, "loss": 1.1228, "step": 19173 }, { "epoch": 1.93, "learning_rate": 3.126082857028788e-05, "loss": 0.4962, "step": 19174 }, { "epoch": 1.93, "learning_rate": 3.125843751181435e-05, "loss": 0.4628, "step": 19175 }, { "epoch": 1.93, "learning_rate": 3.12560464453326e-05, "loss": 0.2232, "step": 19176 }, { "epoch": 1.93, "learning_rate": 3.125365537085785e-05, "loss": 0.783, "step": 19177 }, { "epoch": 1.93, "learning_rate": 3.125126428840532e-05, "loss": 0.4049, "step": 19178 }, { "epoch": 1.93, "learning_rate": 3.124887319799021e-05, "loss": 1.0291, "step": 19179 }, { "epoch": 1.93, "learning_rate": 3.124648209962775e-05, "loss": 0.9511, "step": 19180 }, { "epoch": 1.93, "learning_rate": 3.124409099333314e-05, "loss": 0.3227, "step": 19181 }, { "epoch": 1.93, "learning_rate": 3.124169987912163e-05, "loss": 2.3456, "step": 19182 }, { "epoch": 1.93, "learning_rate": 3.12393087570084e-05, "loss": 0.5454, "step": 19183 }, { "epoch": 1.94, "learning_rate": 3.123691762700869e-05, "loss": 1.0866, "step": 19184 }, { "epoch": 1.94, "learning_rate": 3.123452648913771e-05, "loss": 0.1551, "step": 19185 }, { "epoch": 1.94, "learning_rate": 3.1232135343410655e-05, "loss": 0.3906, "step": 19186 }, { "epoch": 1.94, "learning_rate": 3.1229744189842774e-05, "loss": 1.1967, "step": 19187 }, { "epoch": 1.94, "learning_rate": 3.1227353028449265e-05, "loss": 1.512, "step": 19188 }, { "epoch": 1.94, "learning_rate": 3.122496185924535e-05, "loss": 0.7982, "step": 19189 }, { "epoch": 1.94, "learning_rate": 3.122257068224623e-05, "loss": 0.9083, "step": 19190 }, { "epoch": 1.94, "learning_rate": 3.122017949746716e-05, "loss": 0.7891, "step": 19191 }, { "epoch": 1.94, "learning_rate": 3.1217788304923314e-05, "loss": 0.3918, "step": 19192 }, { "epoch": 1.94, "learning_rate": 3.1215397104629926e-05, "loss": 0.47, "step": 19193 }, { "epoch": 1.94, "learning_rate": 3.121300589660222e-05, "loss": 0.9853, "step": 19194 }, { "epoch": 1.94, "learning_rate": 3.121061468085541e-05, "loss": 0.7495, "step": 19195 }, { "epoch": 1.94, "learning_rate": 3.12082234574047e-05, "loss": 0.6133, "step": 19196 }, { "epoch": 1.94, "learning_rate": 3.120583222626531e-05, "loss": 0.433, "step": 19197 }, { "epoch": 1.94, "learning_rate": 3.120344098745248e-05, "loss": 0.8271, "step": 19198 }, { "epoch": 1.94, "learning_rate": 3.1201049740981396e-05, "loss": 0.6495, "step": 19199 }, { "epoch": 1.94, "learning_rate": 3.119865848686729e-05, "loss": 0.4007, "step": 19200 }, { "epoch": 1.94, "learning_rate": 3.119626722512538e-05, "loss": 0.5903, "step": 19201 }, { "epoch": 1.94, "learning_rate": 3.119387595577088e-05, "loss": 0.5437, "step": 19202 }, { "epoch": 1.94, "learning_rate": 3.119148467881902e-05, "loss": 0.8886, "step": 19203 }, { "epoch": 1.94, "learning_rate": 3.118909339428499e-05, "loss": 0.7272, "step": 19204 }, { "epoch": 1.94, "learning_rate": 3.1186702102184025e-05, "loss": 0.8574, "step": 19205 }, { "epoch": 1.94, "learning_rate": 3.1184310802531327e-05, "loss": 1.0486, "step": 19206 }, { "epoch": 1.94, "learning_rate": 3.118191949534214e-05, "loss": 0.7794, "step": 19207 }, { "epoch": 1.94, "learning_rate": 3.117952818063167e-05, "loss": 0.723, "step": 19208 }, { "epoch": 1.94, "learning_rate": 3.117713685841513e-05, "loss": 0.5778, "step": 19209 }, { "epoch": 1.94, "learning_rate": 3.117474552870773e-05, "loss": 0.4725, "step": 19210 }, { "epoch": 1.94, "learning_rate": 3.11723541915247e-05, "loss": 0.6184, "step": 19211 }, { "epoch": 1.94, "learning_rate": 3.116996284688126e-05, "loss": 0.2212, "step": 19212 }, { "epoch": 1.94, "learning_rate": 3.1167571494792606e-05, "loss": 0.5345, "step": 19213 }, { "epoch": 1.94, "learning_rate": 3.116518013527399e-05, "loss": 0.2091, "step": 19214 }, { "epoch": 1.94, "learning_rate": 3.11627887683406e-05, "loss": 0.3459, "step": 19215 }, { "epoch": 1.94, "learning_rate": 3.116039739400767e-05, "loss": 0.5442, "step": 19216 }, { "epoch": 1.94, "learning_rate": 3.1158006012290406e-05, "loss": 0.5778, "step": 19217 }, { "epoch": 1.94, "learning_rate": 3.115561462320403e-05, "loss": 0.7177, "step": 19218 }, { "epoch": 1.94, "learning_rate": 3.115322322676377e-05, "loss": 0.2652, "step": 19219 }, { "epoch": 1.94, "learning_rate": 3.115083182298484e-05, "loss": 0.9348, "step": 19220 }, { "epoch": 1.94, "learning_rate": 3.114844041188245e-05, "loss": 0.5787, "step": 19221 }, { "epoch": 1.94, "learning_rate": 3.114604899347181e-05, "loss": 0.8366, "step": 19222 }, { "epoch": 1.94, "learning_rate": 3.1143657567768164e-05, "loss": 0.9185, "step": 19223 }, { "epoch": 1.94, "learning_rate": 3.114126613478671e-05, "loss": 0.5166, "step": 19224 }, { "epoch": 1.94, "learning_rate": 3.113887469454268e-05, "loss": 0.2613, "step": 19225 }, { "epoch": 1.94, "learning_rate": 3.113648324705128e-05, "loss": 1.01, "step": 19226 }, { "epoch": 1.94, "learning_rate": 3.113409179232774e-05, "loss": 0.6936, "step": 19227 }, { "epoch": 1.94, "learning_rate": 3.113170033038726e-05, "loss": 0.6036, "step": 19228 }, { "epoch": 1.94, "learning_rate": 3.1129308861245065e-05, "loss": 0.5784, "step": 19229 }, { "epoch": 1.94, "learning_rate": 3.11269173849164e-05, "loss": 0.6396, "step": 19230 }, { "epoch": 1.94, "learning_rate": 3.112452590141644e-05, "loss": 0.9519, "step": 19231 }, { "epoch": 1.94, "learning_rate": 3.112213441076043e-05, "loss": 0.2105, "step": 19232 }, { "epoch": 1.94, "learning_rate": 3.1119742912963594e-05, "loss": 0.5209, "step": 19233 }, { "epoch": 1.94, "learning_rate": 3.111735140804113e-05, "loss": 0.1765, "step": 19234 }, { "epoch": 1.94, "learning_rate": 3.111495989600827e-05, "loss": 0.5946, "step": 19235 }, { "epoch": 1.94, "learning_rate": 3.111256837688022e-05, "loss": 0.6842, "step": 19236 }, { "epoch": 1.94, "learning_rate": 3.111017685067223e-05, "loss": 0.4329, "step": 19237 }, { "epoch": 1.94, "learning_rate": 3.1107785317399474e-05, "loss": 0.574, "step": 19238 }, { "epoch": 1.94, "learning_rate": 3.1105393777077204e-05, "loss": 0.1992, "step": 19239 }, { "epoch": 1.94, "learning_rate": 3.110300222972063e-05, "loss": 0.6666, "step": 19240 }, { "epoch": 1.94, "learning_rate": 3.1100610675344965e-05, "loss": 0.4169, "step": 19241 }, { "epoch": 1.94, "learning_rate": 3.1098219113965435e-05, "loss": 0.4048, "step": 19242 }, { "epoch": 1.94, "learning_rate": 3.109582754559726e-05, "loss": 0.3241, "step": 19243 }, { "epoch": 1.94, "learning_rate": 3.109343597025565e-05, "loss": 0.8739, "step": 19244 }, { "epoch": 1.94, "learning_rate": 3.1091044387955826e-05, "loss": 0.3228, "step": 19245 }, { "epoch": 1.94, "learning_rate": 3.1088652798713026e-05, "loss": 0.804, "step": 19246 }, { "epoch": 1.94, "learning_rate": 3.108626120254244e-05, "loss": 0.4311, "step": 19247 }, { "epoch": 1.94, "learning_rate": 3.10838695994593e-05, "loss": 0.3419, "step": 19248 }, { "epoch": 1.94, "learning_rate": 3.1081477989478824e-05, "loss": 0.1997, "step": 19249 }, { "epoch": 1.94, "learning_rate": 3.1079086372616245e-05, "loss": 0.4532, "step": 19250 }, { "epoch": 1.94, "learning_rate": 3.107669474888677e-05, "loss": 0.4286, "step": 19251 }, { "epoch": 1.94, "learning_rate": 3.10743031183056e-05, "loss": 0.3258, "step": 19252 }, { "epoch": 1.94, "learning_rate": 3.1071911480888e-05, "loss": 0.1235, "step": 19253 }, { "epoch": 1.94, "learning_rate": 3.1069519836649147e-05, "loss": 0.7382, "step": 19254 }, { "epoch": 1.94, "learning_rate": 3.106712818560429e-05, "loss": 0.3521, "step": 19255 }, { "epoch": 1.94, "learning_rate": 3.1064736527768624e-05, "loss": 0.2186, "step": 19256 }, { "epoch": 1.94, "learning_rate": 3.106234486315738e-05, "loss": 0.8636, "step": 19257 }, { "epoch": 1.94, "learning_rate": 3.105995319178578e-05, "loss": 0.443, "step": 19258 }, { "epoch": 1.94, "learning_rate": 3.105756151366905e-05, "loss": 0.5057, "step": 19259 }, { "epoch": 1.94, "learning_rate": 3.1055169828822385e-05, "loss": 0.8819, "step": 19260 }, { "epoch": 1.94, "learning_rate": 3.105277813726103e-05, "loss": 0.8063, "step": 19261 }, { "epoch": 1.94, "learning_rate": 3.105038643900019e-05, "loss": 0.6617, "step": 19262 }, { "epoch": 1.94, "learning_rate": 3.104799473405509e-05, "loss": 0.4957, "step": 19263 }, { "epoch": 1.94, "learning_rate": 3.104560302244095e-05, "loss": 0.8054, "step": 19264 }, { "epoch": 1.94, "learning_rate": 3.104321130417299e-05, "loss": 0.5401, "step": 19265 }, { "epoch": 1.94, "learning_rate": 3.104081957926644e-05, "loss": 0.6949, "step": 19266 }, { "epoch": 1.94, "learning_rate": 3.1038427847736504e-05, "loss": 0.4842, "step": 19267 }, { "epoch": 1.94, "learning_rate": 3.1036036109598404e-05, "loss": 0.673, "step": 19268 }, { "epoch": 1.94, "learning_rate": 3.103364436486738e-05, "loss": 0.6824, "step": 19269 }, { "epoch": 1.94, "learning_rate": 3.103125261355861e-05, "loss": 0.9321, "step": 19270 }, { "epoch": 1.94, "learning_rate": 3.102886085568736e-05, "loss": 0.5806, "step": 19271 }, { "epoch": 1.94, "learning_rate": 3.1026469091268824e-05, "loss": 0.8091, "step": 19272 }, { "epoch": 1.94, "learning_rate": 3.102407732031824e-05, "loss": 0.1838, "step": 19273 }, { "epoch": 1.94, "learning_rate": 3.10216855428508e-05, "loss": 0.3347, "step": 19274 }, { "epoch": 1.94, "learning_rate": 3.101929375888175e-05, "loss": 0.3737, "step": 19275 }, { "epoch": 1.94, "learning_rate": 3.101690196842631e-05, "loss": 0.9639, "step": 19276 }, { "epoch": 1.94, "learning_rate": 3.101451017149968e-05, "loss": 0.9084, "step": 19277 }, { "epoch": 1.94, "learning_rate": 3.10121183681171e-05, "loss": 0.2833, "step": 19278 }, { "epoch": 1.94, "learning_rate": 3.100972655829379e-05, "loss": 0.3164, "step": 19279 }, { "epoch": 1.94, "learning_rate": 3.100733474204495e-05, "loss": 0.9556, "step": 19280 }, { "epoch": 1.94, "learning_rate": 3.100494291938582e-05, "loss": 0.8468, "step": 19281 }, { "epoch": 1.94, "learning_rate": 3.100255109033162e-05, "loss": 0.2993, "step": 19282 }, { "epoch": 1.95, "learning_rate": 3.100015925489757e-05, "loss": 1.0089, "step": 19283 }, { "epoch": 1.95, "learning_rate": 3.099776741309887e-05, "loss": 1.1101, "step": 19284 }, { "epoch": 1.95, "learning_rate": 3.099537556495078e-05, "loss": 0.5672, "step": 19285 }, { "epoch": 1.95, "learning_rate": 3.0992983710468474e-05, "loss": 0.442, "step": 19286 }, { "epoch": 1.95, "learning_rate": 3.0990591849667216e-05, "loss": 1.2626, "step": 19287 }, { "epoch": 1.95, "learning_rate": 3.0988199982562196e-05, "loss": 1.0004, "step": 19288 }, { "epoch": 1.95, "learning_rate": 3.098580810916866e-05, "loss": 0.2911, "step": 19289 }, { "epoch": 1.95, "learning_rate": 3.098341622950181e-05, "loss": 0.5962, "step": 19290 }, { "epoch": 1.95, "learning_rate": 3.0981024343576864e-05, "loss": 0.5697, "step": 19291 }, { "epoch": 1.95, "learning_rate": 3.0978632451409075e-05, "loss": 1.0445, "step": 19292 }, { "epoch": 1.95, "learning_rate": 3.097624055301362e-05, "loss": 0.6063, "step": 19293 }, { "epoch": 1.95, "learning_rate": 3.097384864840575e-05, "loss": 0.988, "step": 19294 }, { "epoch": 1.95, "learning_rate": 3.097145673760068e-05, "loss": 0.3813, "step": 19295 }, { "epoch": 1.95, "learning_rate": 3.096906482061363e-05, "loss": 0.4785, "step": 19296 }, { "epoch": 1.95, "learning_rate": 3.0966672897459806e-05, "loss": 0.5517, "step": 19297 }, { "epoch": 1.95, "learning_rate": 3.096428096815446e-05, "loss": 0.2933, "step": 19298 }, { "epoch": 1.95, "learning_rate": 3.0961889032712794e-05, "loss": 0.1412, "step": 19299 }, { "epoch": 1.95, "learning_rate": 3.095949709115002e-05, "loss": 0.4444, "step": 19300 }, { "epoch": 1.95, "learning_rate": 3.095710514348139e-05, "loss": 0.5613, "step": 19301 }, { "epoch": 1.95, "learning_rate": 3.095471318972209e-05, "loss": 0.8276, "step": 19302 }, { "epoch": 1.95, "learning_rate": 3.095232122988737e-05, "loss": 0.7219, "step": 19303 }, { "epoch": 1.95, "learning_rate": 3.094992926399243e-05, "loss": 1.6411, "step": 19304 }, { "epoch": 1.95, "learning_rate": 3.094753729205251e-05, "loss": 0.7812, "step": 19305 }, { "epoch": 1.95, "learning_rate": 3.0945145314082825e-05, "loss": 0.3708, "step": 19306 }, { "epoch": 1.95, "learning_rate": 3.094275333009858e-05, "loss": 0.2303, "step": 19307 }, { "epoch": 1.95, "learning_rate": 3.094036134011503e-05, "loss": 0.5594, "step": 19308 }, { "epoch": 1.95, "learning_rate": 3.093796934414736e-05, "loss": 0.1765, "step": 19309 }, { "epoch": 1.95, "learning_rate": 3.0935577342210824e-05, "loss": 1.0654, "step": 19310 }, { "epoch": 1.95, "learning_rate": 3.093318533432063e-05, "loss": 0.1838, "step": 19311 }, { "epoch": 1.95, "learning_rate": 3.093079332049199e-05, "loss": 0.431, "step": 19312 }, { "epoch": 1.95, "learning_rate": 3.0928401300740135e-05, "loss": 0.2267, "step": 19313 }, { "epoch": 1.95, "learning_rate": 3.0926009275080305e-05, "loss": 0.7023, "step": 19314 }, { "epoch": 1.95, "learning_rate": 3.092361724352769e-05, "loss": 0.8793, "step": 19315 }, { "epoch": 1.95, "learning_rate": 3.0921225206097515e-05, "loss": 0.7466, "step": 19316 }, { "epoch": 1.95, "learning_rate": 3.0918833162805035e-05, "loss": 0.7762, "step": 19317 }, { "epoch": 1.95, "learning_rate": 3.091644111366544e-05, "loss": 0.5496, "step": 19318 }, { "epoch": 1.95, "learning_rate": 3.091404905869396e-05, "loss": 0.8666, "step": 19319 }, { "epoch": 1.95, "learning_rate": 3.091165699790582e-05, "loss": 0.3495, "step": 19320 }, { "epoch": 1.95, "learning_rate": 3.090926493131625e-05, "loss": 0.3391, "step": 19321 }, { "epoch": 1.95, "learning_rate": 3.090687285894046e-05, "loss": 0.786, "step": 19322 }, { "epoch": 1.95, "learning_rate": 3.090448078079367e-05, "loss": 0.9592, "step": 19323 }, { "epoch": 1.95, "learning_rate": 3.0902088696891117e-05, "loss": 0.451, "step": 19324 }, { "epoch": 1.95, "learning_rate": 3.0899696607248004e-05, "loss": 0.4621, "step": 19325 }, { "epoch": 1.95, "learning_rate": 3.089730451187958e-05, "loss": 0.2256, "step": 19326 }, { "epoch": 1.95, "learning_rate": 3.089491241080104e-05, "loss": 0.3025, "step": 19327 }, { "epoch": 1.95, "learning_rate": 3.089252030402763e-05, "loss": 0.3276, "step": 19328 }, { "epoch": 1.95, "learning_rate": 3.089012819157454e-05, "loss": 0.7745, "step": 19329 }, { "epoch": 1.95, "learning_rate": 3.0887736073457026e-05, "loss": 0.361, "step": 19330 }, { "epoch": 1.95, "learning_rate": 3.0885343949690296e-05, "loss": 0.1828, "step": 19331 }, { "epoch": 1.95, "learning_rate": 3.088295182028957e-05, "loss": 0.1843, "step": 19332 }, { "epoch": 1.95, "learning_rate": 3.088055968527008e-05, "loss": 1.3402, "step": 19333 }, { "epoch": 1.95, "learning_rate": 3.087816754464704e-05, "loss": 0.2767, "step": 19334 }, { "epoch": 1.95, "learning_rate": 3.087577539843568e-05, "loss": 0.8031, "step": 19335 }, { "epoch": 1.95, "learning_rate": 3.0873383246651205e-05, "loss": 0.5793, "step": 19336 }, { "epoch": 1.95, "learning_rate": 3.087099108930887e-05, "loss": 0.3432, "step": 19337 }, { "epoch": 1.95, "learning_rate": 3.086859892642387e-05, "loss": 0.904, "step": 19338 }, { "epoch": 1.95, "learning_rate": 3.0866206758011434e-05, "loss": 1.1291, "step": 19339 }, { "epoch": 1.95, "learning_rate": 3.0863814584086794e-05, "loss": 0.6654, "step": 19340 }, { "epoch": 1.95, "learning_rate": 3.086142240466516e-05, "loss": 0.8238, "step": 19341 }, { "epoch": 1.95, "learning_rate": 3.085903021976177e-05, "loss": 0.4402, "step": 19342 }, { "epoch": 1.95, "learning_rate": 3.085663802939182e-05, "loss": 0.3625, "step": 19343 }, { "epoch": 1.95, "learning_rate": 3.0854245833570575e-05, "loss": 0.259, "step": 19344 }, { "epoch": 1.95, "learning_rate": 3.085185363231322e-05, "loss": 1.2427, "step": 19345 }, { "epoch": 1.95, "learning_rate": 3.0849461425634994e-05, "loss": 1.2866, "step": 19346 }, { "epoch": 1.95, "learning_rate": 3.084706921355113e-05, "loss": 0.3539, "step": 19347 }, { "epoch": 1.95, "learning_rate": 3.0844676996076825e-05, "loss": 0.9376, "step": 19348 }, { "epoch": 1.95, "learning_rate": 3.084228477322732e-05, "loss": 0.4907, "step": 19349 }, { "epoch": 1.95, "learning_rate": 3.083989254501784e-05, "loss": 0.302, "step": 19350 }, { "epoch": 1.95, "learning_rate": 3.08375003114636e-05, "loss": 1.1266, "step": 19351 }, { "epoch": 1.95, "learning_rate": 3.083510807257983e-05, "loss": 0.6229, "step": 19352 }, { "epoch": 1.95, "learning_rate": 3.083271582838175e-05, "loss": 0.5886, "step": 19353 }, { "epoch": 1.95, "learning_rate": 3.083032357888458e-05, "loss": 0.4864, "step": 19354 }, { "epoch": 1.95, "learning_rate": 3.0827931324103545e-05, "loss": 0.905, "step": 19355 }, { "epoch": 1.95, "learning_rate": 3.082553906405387e-05, "loss": 0.1485, "step": 19356 }, { "epoch": 1.95, "learning_rate": 3.082314679875079e-05, "loss": 0.8198, "step": 19357 }, { "epoch": 1.95, "learning_rate": 3.0820754528209504e-05, "loss": 0.2148, "step": 19358 }, { "epoch": 1.95, "learning_rate": 3.081836225244525e-05, "loss": 0.4311, "step": 19359 }, { "epoch": 1.95, "learning_rate": 3.081596997147326e-05, "loss": 0.462, "step": 19360 }, { "epoch": 1.95, "learning_rate": 3.081357768530873e-05, "loss": 0.3797, "step": 19361 }, { "epoch": 1.95, "learning_rate": 3.0811185393966916e-05, "loss": 1.276, "step": 19362 }, { "epoch": 1.95, "learning_rate": 3.080879309746303e-05, "loss": 0.4, "step": 19363 }, { "epoch": 1.95, "learning_rate": 3.080640079581228e-05, "loss": 0.4969, "step": 19364 }, { "epoch": 1.95, "learning_rate": 3.0804008489029905e-05, "loss": 0.3773, "step": 19365 }, { "epoch": 1.95, "learning_rate": 3.080161617713113e-05, "loss": 0.3804, "step": 19366 }, { "epoch": 1.95, "learning_rate": 3.079922386013117e-05, "loss": 0.4151, "step": 19367 }, { "epoch": 1.95, "learning_rate": 3.079683153804525e-05, "loss": 0.2943, "step": 19368 }, { "epoch": 1.95, "learning_rate": 3.079443921088861e-05, "loss": 0.6005, "step": 19369 }, { "epoch": 1.95, "learning_rate": 3.079204687867646e-05, "loss": 0.5052, "step": 19370 }, { "epoch": 1.95, "learning_rate": 3.078965454142401e-05, "loss": 0.4714, "step": 19371 }, { "epoch": 1.95, "learning_rate": 3.0787262199146504e-05, "loss": 0.8167, "step": 19372 }, { "epoch": 1.95, "learning_rate": 3.078486985185917e-05, "loss": 0.7087, "step": 19373 }, { "epoch": 1.95, "learning_rate": 3.0782477499577216e-05, "loss": 0.1993, "step": 19374 }, { "epoch": 1.95, "learning_rate": 3.0780085142315866e-05, "loss": 0.2935, "step": 19375 }, { "epoch": 1.95, "learning_rate": 3.077769278009037e-05, "loss": 0.0589, "step": 19376 }, { "epoch": 1.95, "learning_rate": 3.0775300412915914e-05, "loss": 1.0589, "step": 19377 }, { "epoch": 1.95, "learning_rate": 3.077290804080775e-05, "loss": 0.8193, "step": 19378 }, { "epoch": 1.95, "learning_rate": 3.07705156637811e-05, "loss": 0.5328, "step": 19379 }, { "epoch": 1.95, "learning_rate": 3.076812328185117e-05, "loss": 0.35, "step": 19380 }, { "epoch": 1.96, "learning_rate": 3.07657308950332e-05, "loss": 0.2041, "step": 19381 }, { "epoch": 1.96, "learning_rate": 3.0763338503342405e-05, "loss": 0.784, "step": 19382 }, { "epoch": 1.96, "learning_rate": 3.076094610679402e-05, "loss": 0.2704, "step": 19383 }, { "epoch": 1.96, "learning_rate": 3.075855370540326e-05, "loss": 0.1193, "step": 19384 }, { "epoch": 1.96, "learning_rate": 3.0756161299185354e-05, "loss": 0.6088, "step": 19385 }, { "epoch": 1.96, "learning_rate": 3.075376888815554e-05, "loss": 0.3651, "step": 19386 }, { "epoch": 1.96, "learning_rate": 3.0751376472329e-05, "loss": 0.3003, "step": 19387 }, { "epoch": 1.96, "learning_rate": 3.0748984051721e-05, "loss": 0.1862, "step": 19388 }, { "epoch": 1.96, "learning_rate": 3.074659162634675e-05, "loss": 0.3292, "step": 19389 }, { "epoch": 1.96, "learning_rate": 3.074419919622148e-05, "loss": 0.2525, "step": 19390 }, { "epoch": 1.96, "learning_rate": 3.07418067613604e-05, "loss": 0.2226, "step": 19391 }, { "epoch": 1.96, "learning_rate": 3.0739414321778754e-05, "loss": 0.6522, "step": 19392 }, { "epoch": 1.96, "learning_rate": 3.073702187749174e-05, "loss": 0.4141, "step": 19393 }, { "epoch": 1.96, "learning_rate": 3.073462942851461e-05, "loss": 0.8387, "step": 19394 }, { "epoch": 1.96, "learning_rate": 3.073223697486259e-05, "loss": 1.3526, "step": 19395 }, { "epoch": 1.96, "learning_rate": 3.072984451655087e-05, "loss": 0.5438, "step": 19396 }, { "epoch": 1.96, "learning_rate": 3.072745205359471e-05, "loss": 0.5089, "step": 19397 }, { "epoch": 1.96, "learning_rate": 3.072505958600931e-05, "loss": 0.261, "step": 19398 }, { "epoch": 1.96, "learning_rate": 3.072266711380993e-05, "loss": 1.2523, "step": 19399 }, { "epoch": 1.96, "learning_rate": 3.0720274637011745e-05, "loss": 0.9039, "step": 19400 }, { "epoch": 1.96, "learning_rate": 3.071788215563002e-05, "loss": 1.3326, "step": 19401 }, { "epoch": 1.96, "learning_rate": 3.071548966967997e-05, "loss": 0.5729, "step": 19402 }, { "epoch": 1.96, "learning_rate": 3.071309717917681e-05, "loss": 0.6573, "step": 19403 }, { "epoch": 1.96, "learning_rate": 3.0710704684135774e-05, "loss": 0.4069, "step": 19404 }, { "epoch": 1.96, "learning_rate": 3.0708312184572076e-05, "loss": 0.6286, "step": 19405 }, { "epoch": 1.96, "learning_rate": 3.0705919680500956e-05, "loss": 0.6457, "step": 19406 }, { "epoch": 1.96, "learning_rate": 3.070352717193763e-05, "loss": 0.9062, "step": 19407 }, { "epoch": 1.96, "learning_rate": 3.0701134658897324e-05, "loss": 0.8129, "step": 19408 }, { "epoch": 1.96, "learning_rate": 3.0698742141395264e-05, "loss": 0.6249, "step": 19409 }, { "epoch": 1.96, "learning_rate": 3.0696349619446674e-05, "loss": 0.962, "step": 19410 }, { "epoch": 1.96, "learning_rate": 3.069395709306678e-05, "loss": 0.3616, "step": 19411 }, { "epoch": 1.96, "learning_rate": 3.06915645622708e-05, "loss": 0.8018, "step": 19412 }, { "epoch": 1.96, "learning_rate": 3.0689172027073986e-05, "loss": 0.6012, "step": 19413 }, { "epoch": 1.96, "learning_rate": 3.068677948749152e-05, "loss": 0.3009, "step": 19414 }, { "epoch": 1.96, "learning_rate": 3.068438694353867e-05, "loss": 0.328, "step": 19415 }, { "epoch": 1.96, "learning_rate": 3.068199439523062e-05, "loss": 0.5937, "step": 19416 }, { "epoch": 1.96, "learning_rate": 3.0679601842582636e-05, "loss": 0.1034, "step": 19417 }, { "epoch": 1.96, "learning_rate": 3.067720928560993e-05, "loss": 0.5553, "step": 19418 }, { "epoch": 1.96, "learning_rate": 3.06748167243277e-05, "loss": 0.1746, "step": 19419 }, { "epoch": 1.96, "learning_rate": 3.067242415875121e-05, "loss": 0.1993, "step": 19420 }, { "epoch": 1.96, "learning_rate": 3.067003158889566e-05, "loss": 0.2839, "step": 19421 }, { "epoch": 1.96, "learning_rate": 3.066763901477629e-05, "loss": 1.7962, "step": 19422 }, { "epoch": 1.96, "learning_rate": 3.0665246436408307e-05, "loss": 0.1925, "step": 19423 }, { "epoch": 1.96, "learning_rate": 3.066285385380696e-05, "loss": 0.1268, "step": 19424 }, { "epoch": 1.96, "learning_rate": 3.0660461266987464e-05, "loss": 1.0022, "step": 19425 }, { "epoch": 1.96, "learning_rate": 3.065806867596504e-05, "loss": 0.651, "step": 19426 }, { "epoch": 1.96, "learning_rate": 3.0655676080754915e-05, "loss": 0.2466, "step": 19427 }, { "epoch": 1.96, "learning_rate": 3.065328348137233e-05, "loss": 0.7267, "step": 19428 }, { "epoch": 1.96, "learning_rate": 3.065089087783248e-05, "loss": 0.6433, "step": 19429 }, { "epoch": 1.96, "learning_rate": 3.064849827015061e-05, "loss": 0.8167, "step": 19430 }, { "epoch": 1.96, "learning_rate": 3.064610565834196e-05, "loss": 0.5297, "step": 19431 }, { "epoch": 1.96, "learning_rate": 3.0643713042421716e-05, "loss": 0.4297, "step": 19432 }, { "epoch": 1.96, "learning_rate": 3.064132042240514e-05, "loss": 0.2495, "step": 19433 }, { "epoch": 1.96, "learning_rate": 3.0638927798307454e-05, "loss": 0.5661, "step": 19434 }, { "epoch": 1.96, "learning_rate": 3.063653517014386e-05, "loss": 0.472, "step": 19435 }, { "epoch": 1.96, "learning_rate": 3.0634142537929605e-05, "loss": 1.492, "step": 19436 }, { "epoch": 1.96, "learning_rate": 3.0631749901679895e-05, "loss": 0.7687, "step": 19437 }, { "epoch": 1.96, "learning_rate": 3.0629357261409984e-05, "loss": 0.9581, "step": 19438 }, { "epoch": 1.96, "learning_rate": 3.062696461713508e-05, "loss": 0.5312, "step": 19439 }, { "epoch": 1.96, "learning_rate": 3.062457196887041e-05, "loss": 0.969, "step": 19440 }, { "epoch": 1.96, "learning_rate": 3.0622179316631204e-05, "loss": 0.2104, "step": 19441 }, { "epoch": 1.96, "learning_rate": 3.061978666043268e-05, "loss": 0.2851, "step": 19442 }, { "epoch": 1.96, "learning_rate": 3.061739400029007e-05, "loss": 1.0936, "step": 19443 }, { "epoch": 1.96, "learning_rate": 3.061500133621861e-05, "loss": 0.346, "step": 19444 }, { "epoch": 1.96, "learning_rate": 3.061260866823351e-05, "loss": 0.2016, "step": 19445 }, { "epoch": 1.96, "learning_rate": 3.061021599634999e-05, "loss": 2.0884, "step": 19446 }, { "epoch": 1.96, "learning_rate": 3.060782332058331e-05, "loss": 1.6764, "step": 19447 }, { "epoch": 1.96, "learning_rate": 3.060543064094865e-05, "loss": 0.322, "step": 19448 }, { "epoch": 1.96, "learning_rate": 3.060303795746127e-05, "loss": 0.8698, "step": 19449 }, { "epoch": 1.96, "learning_rate": 3.0600645270136394e-05, "loss": 0.5394, "step": 19450 }, { "epoch": 1.96, "learning_rate": 3.0598252578989226e-05, "loss": 0.569, "step": 19451 }, { "epoch": 1.96, "learning_rate": 3.059585988403501e-05, "loss": 0.3809, "step": 19452 }, { "epoch": 1.96, "learning_rate": 3.059346718528897e-05, "loss": 0.9902, "step": 19453 }, { "epoch": 1.96, "learning_rate": 3.0591074482766335e-05, "loss": 0.4499, "step": 19454 }, { "epoch": 1.96, "learning_rate": 3.058868177648232e-05, "loss": 0.5539, "step": 19455 }, { "epoch": 1.96, "learning_rate": 3.0586289066452166e-05, "loss": 0.5142, "step": 19456 }, { "epoch": 1.96, "learning_rate": 3.058389635269108e-05, "loss": 0.7383, "step": 19457 }, { "epoch": 1.96, "learning_rate": 3.058150363521431e-05, "loss": 0.4493, "step": 19458 }, { "epoch": 1.96, "learning_rate": 3.057911091403707e-05, "loss": 0.5531, "step": 19459 }, { "epoch": 1.96, "learning_rate": 3.057671818917459e-05, "loss": 0.2753, "step": 19460 }, { "epoch": 1.96, "learning_rate": 3.057432546064209e-05, "loss": 0.9838, "step": 19461 }, { "epoch": 1.96, "learning_rate": 3.05719327284548e-05, "loss": 0.7382, "step": 19462 }, { "epoch": 1.96, "learning_rate": 3.0569539992627956e-05, "loss": 1.1073, "step": 19463 }, { "epoch": 1.96, "learning_rate": 3.056714725317676e-05, "loss": 0.7636, "step": 19464 }, { "epoch": 1.96, "learning_rate": 3.056475451011647e-05, "loss": 0.5418, "step": 19465 }, { "epoch": 1.96, "learning_rate": 3.056236176346229e-05, "loss": 0.4712, "step": 19466 }, { "epoch": 1.96, "learning_rate": 3.055996901322946e-05, "loss": 0.6967, "step": 19467 }, { "epoch": 1.96, "learning_rate": 3.0557576259433196e-05, "loss": 0.8529, "step": 19468 }, { "epoch": 1.96, "learning_rate": 3.055518350208872e-05, "loss": 1.0208, "step": 19469 }, { "epoch": 1.96, "learning_rate": 3.055279074121129e-05, "loss": 0.3118, "step": 19470 }, { "epoch": 1.96, "learning_rate": 3.055039797681609e-05, "loss": 0.97, "step": 19471 }, { "epoch": 1.96, "learning_rate": 3.054800520891838e-05, "loss": 0.8295, "step": 19472 }, { "epoch": 1.96, "learning_rate": 3.054561243753336e-05, "loss": 0.9495, "step": 19473 }, { "epoch": 1.96, "learning_rate": 3.054321966267628e-05, "loss": 0.5598, "step": 19474 }, { "epoch": 1.96, "learning_rate": 3.0540826884362355e-05, "loss": 0.3061, "step": 19475 }, { "epoch": 1.96, "learning_rate": 3.053843410260681e-05, "loss": 0.3769, "step": 19476 }, { "epoch": 1.96, "learning_rate": 3.053604131742488e-05, "loss": 0.4858, "step": 19477 }, { "epoch": 1.96, "learning_rate": 3.053364852883178e-05, "loss": 0.3982, "step": 19478 }, { "epoch": 1.96, "learning_rate": 3.053125573684274e-05, "loss": 0.3418, "step": 19479 }, { "epoch": 1.97, "learning_rate": 3.052886294147301e-05, "loss": 1.0534, "step": 19480 }, { "epoch": 1.97, "learning_rate": 3.052647014273778e-05, "loss": 0.6021, "step": 19481 }, { "epoch": 1.97, "learning_rate": 3.05240773406523e-05, "loss": 0.6532, "step": 19482 }, { "epoch": 1.97, "learning_rate": 3.05216845352318e-05, "loss": 0.3708, "step": 19483 }, { "epoch": 1.97, "learning_rate": 3.0519291726491485e-05, "loss": 0.2528, "step": 19484 }, { "epoch": 1.97, "learning_rate": 3.0516898914446596e-05, "loss": 0.7424, "step": 19485 }, { "epoch": 1.97, "learning_rate": 3.051450609911237e-05, "loss": 0.6392, "step": 19486 }, { "epoch": 1.97, "learning_rate": 3.0512113280504008e-05, "loss": 0.5113, "step": 19487 }, { "epoch": 1.97, "learning_rate": 3.0509720458636762e-05, "loss": 0.7291, "step": 19488 }, { "epoch": 1.97, "learning_rate": 3.050732763352584e-05, "loss": 1.026, "step": 19489 }, { "epoch": 1.97, "learning_rate": 3.050493480518649e-05, "loss": 1.079, "step": 19490 }, { "epoch": 1.97, "learning_rate": 3.0502541973633917e-05, "loss": 0.1216, "step": 19491 }, { "epoch": 1.97, "learning_rate": 3.050014913888336e-05, "loss": 1.278, "step": 19492 }, { "epoch": 1.97, "learning_rate": 3.049775630095005e-05, "loss": 0.8005, "step": 19493 }, { "epoch": 1.97, "learning_rate": 3.0495363459849205e-05, "loss": 0.9496, "step": 19494 }, { "epoch": 1.97, "learning_rate": 3.049297061559605e-05, "loss": 0.5711, "step": 19495 }, { "epoch": 1.97, "learning_rate": 3.0490577768205825e-05, "loss": 0.4641, "step": 19496 }, { "epoch": 1.97, "learning_rate": 3.048818491769375e-05, "loss": 0.9646, "step": 19497 }, { "epoch": 1.97, "learning_rate": 3.0485792064075048e-05, "loss": 1.2806, "step": 19498 }, { "epoch": 1.97, "learning_rate": 3.0483399207364952e-05, "loss": 0.2096, "step": 19499 }, { "epoch": 1.97, "learning_rate": 3.0481006347578684e-05, "loss": 0.2576, "step": 19500 }, { "epoch": 1.97, "learning_rate": 3.0478613484731476e-05, "loss": 0.8099, "step": 19501 }, { "epoch": 1.97, "learning_rate": 3.047622061883856e-05, "loss": 0.4215, "step": 19502 }, { "epoch": 1.97, "learning_rate": 3.0473827749915145e-05, "loss": 1.4956, "step": 19503 }, { "epoch": 1.97, "learning_rate": 3.0471434877976478e-05, "loss": 0.7264, "step": 19504 }, { "epoch": 1.97, "learning_rate": 3.0469042003037783e-05, "loss": 0.8679, "step": 19505 }, { "epoch": 1.97, "learning_rate": 3.0466649125114276e-05, "loss": 0.6185, "step": 19506 }, { "epoch": 1.97, "learning_rate": 3.0464256244221197e-05, "loss": 0.3627, "step": 19507 }, { "epoch": 1.97, "learning_rate": 3.0461863360373755e-05, "loss": 0.3034, "step": 19508 }, { "epoch": 1.97, "learning_rate": 3.0459470473587213e-05, "loss": 0.1328, "step": 19509 }, { "epoch": 1.97, "learning_rate": 3.0457077583876752e-05, "loss": 0.9382, "step": 19510 }, { "epoch": 1.97, "learning_rate": 3.0454684691257637e-05, "loss": 0.7704, "step": 19511 }, { "epoch": 1.97, "learning_rate": 3.0452291795745077e-05, "loss": 0.8226, "step": 19512 }, { "epoch": 1.97, "learning_rate": 3.044989889735431e-05, "loss": 0.8334, "step": 19513 }, { "epoch": 1.97, "learning_rate": 3.044750599610055e-05, "loss": 0.1407, "step": 19514 }, { "epoch": 1.97, "learning_rate": 3.0445113091999037e-05, "loss": 1.0359, "step": 19515 }, { "epoch": 1.97, "learning_rate": 3.0442720185064997e-05, "loss": 0.301, "step": 19516 }, { "epoch": 1.97, "learning_rate": 3.044032727531364e-05, "loss": 0.759, "step": 19517 }, { "epoch": 1.97, "learning_rate": 3.0437934362760222e-05, "loss": 0.3391, "step": 19518 }, { "epoch": 1.97, "learning_rate": 3.0435541447419946e-05, "loss": 0.71, "step": 19519 }, { "epoch": 1.97, "learning_rate": 3.043314852930806e-05, "loss": 1.2347, "step": 19520 }, { "epoch": 1.97, "learning_rate": 3.0430755608439772e-05, "loss": 0.6748, "step": 19521 }, { "epoch": 1.97, "learning_rate": 3.042836268483033e-05, "loss": 1.1104, "step": 19522 }, { "epoch": 1.97, "learning_rate": 3.042596975849494e-05, "loss": 0.2119, "step": 19523 }, { "epoch": 1.97, "learning_rate": 3.042357682944884e-05, "loss": 0.6069, "step": 19524 }, { "epoch": 1.97, "learning_rate": 3.042118389770727e-05, "loss": 1.0365, "step": 19525 }, { "epoch": 1.97, "learning_rate": 3.0418790963285438e-05, "loss": 1.862, "step": 19526 }, { "epoch": 1.97, "learning_rate": 3.0416398026198577e-05, "loss": 1.3069, "step": 19527 }, { "epoch": 1.97, "learning_rate": 3.041400508646192e-05, "loss": 0.3853, "step": 19528 }, { "epoch": 1.97, "learning_rate": 3.04116121440907e-05, "loss": 0.7031, "step": 19529 }, { "epoch": 1.97, "learning_rate": 3.0409219199100128e-05, "loss": 0.8719, "step": 19530 }, { "epoch": 1.97, "learning_rate": 3.0406826251505444e-05, "loss": 1.2592, "step": 19531 }, { "epoch": 1.97, "learning_rate": 3.040443330132187e-05, "loss": 1.1798, "step": 19532 }, { "epoch": 1.97, "learning_rate": 3.0402040348564633e-05, "loss": 0.7262, "step": 19533 }, { "epoch": 1.97, "learning_rate": 3.0399647393248966e-05, "loss": 1.1006, "step": 19534 }, { "epoch": 1.97, "learning_rate": 3.0397254435390102e-05, "loss": 0.4426, "step": 19535 }, { "epoch": 1.97, "learning_rate": 3.039486147500326e-05, "loss": 0.1534, "step": 19536 }, { "epoch": 1.97, "learning_rate": 3.0392468512103665e-05, "loss": 0.5428, "step": 19537 }, { "epoch": 1.97, "learning_rate": 3.0390075546706554e-05, "loss": 0.3947, "step": 19538 }, { "epoch": 1.97, "learning_rate": 3.0387682578827152e-05, "loss": 0.3914, "step": 19539 }, { "epoch": 1.97, "learning_rate": 3.0385289608480675e-05, "loss": 1.0548, "step": 19540 }, { "epoch": 1.97, "learning_rate": 3.0382896635682373e-05, "loss": 0.8065, "step": 19541 }, { "epoch": 1.97, "learning_rate": 3.0380503660447455e-05, "loss": 0.5519, "step": 19542 }, { "epoch": 1.97, "learning_rate": 3.037811068279116e-05, "loss": 0.318, "step": 19543 }, { "epoch": 1.97, "learning_rate": 3.0375717702728712e-05, "loss": 0.6288, "step": 19544 }, { "epoch": 1.97, "learning_rate": 3.0373324720275343e-05, "loss": 0.2888, "step": 19545 }, { "epoch": 1.97, "learning_rate": 3.037093173544627e-05, "loss": 0.3021, "step": 19546 }, { "epoch": 1.97, "learning_rate": 3.036853874825673e-05, "loss": 0.5717, "step": 19547 }, { "epoch": 1.97, "learning_rate": 3.036614575872196e-05, "loss": 1.0658, "step": 19548 }, { "epoch": 1.97, "learning_rate": 3.0363752766857163e-05, "loss": 0.1131, "step": 19549 }, { "epoch": 1.97, "learning_rate": 3.036135977267759e-05, "loss": 1.387, "step": 19550 }, { "epoch": 1.97, "learning_rate": 3.0358966776198457e-05, "loss": 1.6372, "step": 19551 }, { "epoch": 1.97, "learning_rate": 3.0356573777435007e-05, "loss": 1.2375, "step": 19552 }, { "epoch": 1.97, "learning_rate": 3.0354180776402448e-05, "loss": 0.7245, "step": 19553 }, { "epoch": 1.97, "learning_rate": 3.0351787773116015e-05, "loss": 1.2574, "step": 19554 }, { "epoch": 1.97, "learning_rate": 3.0349394767590945e-05, "loss": 1.1396, "step": 19555 }, { "epoch": 1.97, "learning_rate": 3.0347001759842447e-05, "loss": 0.488, "step": 19556 }, { "epoch": 1.97, "learning_rate": 3.0344608749885782e-05, "loss": 0.3778, "step": 19557 }, { "epoch": 1.97, "learning_rate": 3.034221573773614e-05, "loss": 0.1966, "step": 19558 }, { "epoch": 1.97, "learning_rate": 3.0339822723408777e-05, "loss": 0.3048, "step": 19559 }, { "epoch": 1.97, "learning_rate": 3.0337429706918903e-05, "loss": 0.297, "step": 19560 }, { "epoch": 1.97, "learning_rate": 3.0335036688281763e-05, "loss": 0.2447, "step": 19561 }, { "epoch": 1.97, "learning_rate": 3.0332643667512574e-05, "loss": 0.6183, "step": 19562 }, { "epoch": 1.97, "learning_rate": 3.0330250644626564e-05, "loss": 0.472, "step": 19563 }, { "epoch": 1.97, "learning_rate": 3.0327857619638977e-05, "loss": 0.6696, "step": 19564 }, { "epoch": 1.97, "learning_rate": 3.032546459256501e-05, "loss": 0.5283, "step": 19565 }, { "epoch": 1.97, "learning_rate": 3.032307156341992e-05, "loss": 0.726, "step": 19566 }, { "epoch": 1.97, "learning_rate": 3.032067853221893e-05, "loss": 0.1329, "step": 19567 }, { "epoch": 1.97, "learning_rate": 3.031828549897726e-05, "loss": 1.1272, "step": 19568 }, { "epoch": 1.97, "learning_rate": 3.031589246371014e-05, "loss": 0.9577, "step": 19569 }, { "epoch": 1.97, "learning_rate": 3.0313499426432802e-05, "loss": 0.2578, "step": 19570 }, { "epoch": 1.97, "learning_rate": 3.0311106387160473e-05, "loss": 0.5588, "step": 19571 }, { "epoch": 1.97, "learning_rate": 3.0308713345908376e-05, "loss": 0.4982, "step": 19572 }, { "epoch": 1.97, "learning_rate": 3.0306320302691757e-05, "loss": 0.621, "step": 19573 }, { "epoch": 1.97, "learning_rate": 3.0303927257525823e-05, "loss": 0.1411, "step": 19574 }, { "epoch": 1.97, "learning_rate": 3.0301534210425818e-05, "loss": 0.9547, "step": 19575 }, { "epoch": 1.97, "learning_rate": 3.0299141161406958e-05, "loss": 0.542, "step": 19576 }, { "epoch": 1.97, "learning_rate": 3.0296748110484476e-05, "loss": 0.4019, "step": 19577 }, { "epoch": 1.98, "learning_rate": 3.0294355057673612e-05, "loss": 0.457, "step": 19578 }, { "epoch": 1.98, "learning_rate": 3.029196200298957e-05, "loss": 0.5001, "step": 19579 }, { "epoch": 1.98, "learning_rate": 3.0289568946447603e-05, "loss": 0.1556, "step": 19580 }, { "epoch": 1.98, "learning_rate": 3.0287175888062924e-05, "loss": 0.1061, "step": 19581 }, { "epoch": 1.98, "learning_rate": 3.028478282785077e-05, "loss": 0.4887, "step": 19582 }, { "epoch": 1.98, "learning_rate": 3.0282389765826364e-05, "loss": 0.3822, "step": 19583 }, { "epoch": 1.98, "learning_rate": 3.027999670200494e-05, "loss": 0.2214, "step": 19584 }, { "epoch": 1.98, "learning_rate": 3.0277603636401726e-05, "loss": 0.7335, "step": 19585 }, { "epoch": 1.98, "learning_rate": 3.0275210569031935e-05, "loss": 0.2148, "step": 19586 }, { "epoch": 1.98, "learning_rate": 3.0272817499910828e-05, "loss": 0.6781, "step": 19587 }, { "epoch": 1.98, "learning_rate": 3.02704244290536e-05, "loss": 0.3742, "step": 19588 }, { "epoch": 1.98, "learning_rate": 3.02680313564755e-05, "loss": 1.2149, "step": 19589 }, { "epoch": 1.98, "learning_rate": 3.0265638282191745e-05, "loss": 1.0025, "step": 19590 }, { "epoch": 1.98, "learning_rate": 3.0263245206217576e-05, "loss": 1.1591, "step": 19591 }, { "epoch": 1.98, "learning_rate": 3.0260852128568213e-05, "loss": 1.3189, "step": 19592 }, { "epoch": 1.98, "learning_rate": 3.0258459049258886e-05, "loss": 0.8809, "step": 19593 }, { "epoch": 1.98, "learning_rate": 3.0256065968304817e-05, "loss": 0.3136, "step": 19594 }, { "epoch": 1.98, "learning_rate": 3.025367288572125e-05, "loss": 0.3579, "step": 19595 }, { "epoch": 1.98, "learning_rate": 3.0251279801523404e-05, "loss": 0.47, "step": 19596 }, { "epoch": 1.98, "learning_rate": 3.0248886715726506e-05, "loss": 0.9328, "step": 19597 }, { "epoch": 1.98, "learning_rate": 3.0246493628345784e-05, "loss": 0.2208, "step": 19598 }, { "epoch": 1.98, "learning_rate": 3.0244100539396477e-05, "loss": 0.3759, "step": 19599 }, { "epoch": 1.98, "learning_rate": 3.0241707448893806e-05, "loss": 0.9815, "step": 19600 }, { "epoch": 1.98, "learning_rate": 3.0239314356853004e-05, "loss": 1.1633, "step": 19601 }, { "epoch": 1.98, "learning_rate": 3.0236921263289284e-05, "loss": 0.3897, "step": 19602 }, { "epoch": 1.98, "learning_rate": 3.0234528168217904e-05, "loss": 0.0912, "step": 19603 }, { "epoch": 1.98, "learning_rate": 3.0232135071654065e-05, "loss": 0.8613, "step": 19604 }, { "epoch": 1.98, "learning_rate": 3.0229741973613004e-05, "loss": 0.4517, "step": 19605 }, { "epoch": 1.98, "learning_rate": 3.0227348874109964e-05, "loss": 0.9145, "step": 19606 }, { "epoch": 1.98, "learning_rate": 3.022495577316015e-05, "loss": 1.1703, "step": 19607 }, { "epoch": 1.98, "learning_rate": 3.022256267077881e-05, "loss": 0.5433, "step": 19608 }, { "epoch": 1.98, "learning_rate": 3.0220169566981166e-05, "loss": 0.5488, "step": 19609 }, { "epoch": 1.98, "learning_rate": 3.0217776461782446e-05, "loss": 0.8049, "step": 19610 }, { "epoch": 1.98, "learning_rate": 3.0215383355197875e-05, "loss": 0.3493, "step": 19611 }, { "epoch": 1.98, "learning_rate": 3.0212990247242703e-05, "loss": 1.1121, "step": 19612 }, { "epoch": 1.98, "learning_rate": 3.021059713793212e-05, "loss": 0.493, "step": 19613 }, { "epoch": 1.98, "learning_rate": 3.0208204027281387e-05, "loss": 0.7985, "step": 19614 }, { "epoch": 1.98, "learning_rate": 3.0205810915305723e-05, "loss": 1.1288, "step": 19615 }, { "epoch": 1.98, "learning_rate": 3.020341780202036e-05, "loss": 0.1089, "step": 19616 }, { "epoch": 1.98, "learning_rate": 3.020102468744052e-05, "loss": 0.1296, "step": 19617 }, { "epoch": 1.98, "learning_rate": 3.0198631571581425e-05, "loss": 0.8252, "step": 19618 }, { "epoch": 1.98, "learning_rate": 3.019623845445833e-05, "loss": 0.553, "step": 19619 }, { "epoch": 1.98, "learning_rate": 3.0193845336086436e-05, "loss": 0.2094, "step": 19620 }, { "epoch": 1.98, "learning_rate": 3.0191452216480987e-05, "loss": 0.5966, "step": 19621 }, { "epoch": 1.98, "learning_rate": 3.0189059095657213e-05, "loss": 0.7558, "step": 19622 }, { "epoch": 1.98, "learning_rate": 3.0186665973630344e-05, "loss": 1.4281, "step": 19623 }, { "epoch": 1.98, "learning_rate": 3.0184272850415595e-05, "loss": 0.4158, "step": 19624 }, { "epoch": 1.98, "learning_rate": 3.0181879726028205e-05, "loss": 1.2174, "step": 19625 }, { "epoch": 1.98, "learning_rate": 3.0179486600483408e-05, "loss": 0.1236, "step": 19626 }, { "epoch": 1.98, "learning_rate": 3.017709347379641e-05, "loss": 0.5578, "step": 19627 }, { "epoch": 1.98, "learning_rate": 3.0174700345982477e-05, "loss": 0.2525, "step": 19628 }, { "epoch": 1.98, "learning_rate": 3.0172307217056802e-05, "loss": 0.7673, "step": 19629 }, { "epoch": 1.98, "learning_rate": 3.016991408703464e-05, "loss": 0.5289, "step": 19630 }, { "epoch": 1.98, "learning_rate": 3.0167520955931203e-05, "loss": 0.3129, "step": 19631 }, { "epoch": 1.98, "learning_rate": 3.016512782376173e-05, "loss": 0.1038, "step": 19632 }, { "epoch": 1.98, "learning_rate": 3.0162734690541447e-05, "loss": 1.0775, "step": 19633 }, { "epoch": 1.98, "learning_rate": 3.016034155628557e-05, "loss": 0.2694, "step": 19634 }, { "epoch": 1.98, "learning_rate": 3.0157948421009364e-05, "loss": 0.5598, "step": 19635 }, { "epoch": 1.98, "learning_rate": 3.0155555284728014e-05, "loss": 0.1675, "step": 19636 }, { "epoch": 1.98, "learning_rate": 3.0153162147456778e-05, "loss": 0.181, "step": 19637 }, { "epoch": 1.98, "learning_rate": 3.0150769009210875e-05, "loss": 0.3051, "step": 19638 }, { "epoch": 1.98, "learning_rate": 3.014837587000553e-05, "loss": 0.9431, "step": 19639 }, { "epoch": 1.98, "learning_rate": 3.014598272985599e-05, "loss": 0.6951, "step": 19640 }, { "epoch": 1.98, "learning_rate": 3.014358958877745e-05, "loss": 0.9198, "step": 19641 }, { "epoch": 1.98, "learning_rate": 3.0141196446785185e-05, "loss": 1.0513, "step": 19642 }, { "epoch": 1.98, "learning_rate": 3.0138803303894383e-05, "loss": 0.7354, "step": 19643 }, { "epoch": 1.98, "learning_rate": 3.0136410160120292e-05, "loss": 0.3506, "step": 19644 }, { "epoch": 1.98, "learning_rate": 3.0134017015478144e-05, "loss": 0.6342, "step": 19645 }, { "epoch": 1.98, "learning_rate": 3.0131623869983167e-05, "loss": 0.8396, "step": 19646 }, { "epoch": 1.98, "learning_rate": 3.012923072365058e-05, "loss": 0.1259, "step": 19647 }, { "epoch": 1.98, "learning_rate": 3.0126837576495613e-05, "loss": 0.8094, "step": 19648 }, { "epoch": 1.98, "learning_rate": 3.012444442853351e-05, "loss": 0.9718, "step": 19649 }, { "epoch": 1.98, "learning_rate": 3.0122051279779475e-05, "loss": 0.7166, "step": 19650 }, { "epoch": 1.98, "learning_rate": 3.0119658130248772e-05, "loss": 0.1313, "step": 19651 }, { "epoch": 1.98, "learning_rate": 3.0117264979956593e-05, "loss": 0.7216, "step": 19652 }, { "epoch": 1.98, "learning_rate": 3.0114871828918188e-05, "loss": 0.4862, "step": 19653 }, { "epoch": 1.98, "learning_rate": 3.0112478677148787e-05, "loss": 0.234, "step": 19654 }, { "epoch": 1.98, "learning_rate": 3.0110085524663613e-05, "loss": 0.8076, "step": 19655 }, { "epoch": 1.98, "learning_rate": 3.01076923714779e-05, "loss": 0.2227, "step": 19656 }, { "epoch": 1.98, "learning_rate": 3.0105299217606862e-05, "loss": 0.4934, "step": 19657 }, { "epoch": 1.98, "learning_rate": 3.010290606306576e-05, "loss": 0.6598, "step": 19658 }, { "epoch": 1.98, "learning_rate": 3.0100512907869778e-05, "loss": 0.1621, "step": 19659 }, { "epoch": 1.98, "learning_rate": 3.0098119752034185e-05, "loss": 0.4702, "step": 19660 }, { "epoch": 1.98, "learning_rate": 3.009572659557419e-05, "loss": 0.7287, "step": 19661 }, { "epoch": 1.98, "learning_rate": 3.009333343850504e-05, "loss": 0.3499, "step": 19662 }, { "epoch": 1.98, "learning_rate": 3.009094028084194e-05, "loss": 0.7799, "step": 19663 }, { "epoch": 1.98, "learning_rate": 3.008854712260013e-05, "loss": 0.2108, "step": 19664 }, { "epoch": 1.98, "learning_rate": 3.0086153963794845e-05, "loss": 1.1116, "step": 19665 }, { "epoch": 1.98, "learning_rate": 3.0083760804441297e-05, "loss": 0.8538, "step": 19666 }, { "epoch": 1.98, "learning_rate": 3.0081367644554745e-05, "loss": 0.1915, "step": 19667 }, { "epoch": 1.98, "learning_rate": 3.0078974484150385e-05, "loss": 1.0459, "step": 19668 }, { "epoch": 1.98, "learning_rate": 3.0076581323243467e-05, "loss": 0.2033, "step": 19669 }, { "epoch": 1.98, "learning_rate": 3.007418816184922e-05, "loss": 0.3176, "step": 19670 }, { "epoch": 1.98, "learning_rate": 3.0071794999982865e-05, "loss": 1.071, "step": 19671 }, { "epoch": 1.98, "learning_rate": 3.0069401837659633e-05, "loss": 0.7713, "step": 19672 }, { "epoch": 1.98, "learning_rate": 3.0067008674894748e-05, "loss": 0.295, "step": 19673 }, { "epoch": 1.98, "learning_rate": 3.006461551170346e-05, "loss": 0.5979, "step": 19674 }, { "epoch": 1.98, "learning_rate": 3.0062222348100965e-05, "loss": 0.2222, "step": 19675 }, { "epoch": 1.98, "learning_rate": 3.0059829184102522e-05, "loss": 1.1451, "step": 19676 }, { "epoch": 1.99, "learning_rate": 3.0057436019723348e-05, "loss": 1.0509, "step": 19677 }, { "epoch": 1.99, "learning_rate": 3.0055042854978673e-05, "loss": 0.2453, "step": 19678 }, { "epoch": 1.99, "learning_rate": 3.0052649689883726e-05, "loss": 0.1193, "step": 19679 }, { "epoch": 1.99, "learning_rate": 3.0050256524453733e-05, "loss": 0.3809, "step": 19680 }, { "epoch": 1.99, "learning_rate": 3.004786335870393e-05, "loss": 1.0291, "step": 19681 }, { "epoch": 1.99, "learning_rate": 3.0045470192649536e-05, "loss": 0.6501, "step": 19682 }, { "epoch": 1.99, "learning_rate": 3.0043077026305795e-05, "loss": 0.5542, "step": 19683 }, { "epoch": 1.99, "learning_rate": 3.0040683859687926e-05, "loss": 0.4342, "step": 19684 }, { "epoch": 1.99, "learning_rate": 3.0038290692811157e-05, "loss": 0.1388, "step": 19685 }, { "epoch": 1.99, "learning_rate": 3.0035897525690724e-05, "loss": 0.3923, "step": 19686 }, { "epoch": 1.99, "learning_rate": 3.0033504358341856e-05, "loss": 0.2228, "step": 19687 }, { "epoch": 1.99, "learning_rate": 3.0031111190779773e-05, "loss": 0.1866, "step": 19688 }, { "epoch": 1.99, "learning_rate": 3.002871802301971e-05, "loss": 0.5166, "step": 19689 }, { "epoch": 1.99, "learning_rate": 3.002632485507691e-05, "loss": 0.7143, "step": 19690 }, { "epoch": 1.99, "learning_rate": 3.0023931686966567e-05, "loss": 0.6806, "step": 19691 }, { "epoch": 1.99, "learning_rate": 3.0021538518703947e-05, "loss": 0.0921, "step": 19692 }, { "epoch": 1.99, "learning_rate": 3.0019145350304257e-05, "loss": 0.5736, "step": 19693 }, { "epoch": 1.99, "learning_rate": 3.001675218178274e-05, "loss": 0.1502, "step": 19694 }, { "epoch": 1.99, "learning_rate": 3.0014359013154618e-05, "loss": 0.207, "step": 19695 }, { "epoch": 1.99, "learning_rate": 3.0011965844435108e-05, "loss": 0.9109, "step": 19696 }, { "epoch": 1.99, "learning_rate": 3.000957267563947e-05, "loss": 0.5631, "step": 19697 }, { "epoch": 1.99, "learning_rate": 3.0007179506782898e-05, "loss": 0.3944, "step": 19698 }, { "epoch": 1.99, "learning_rate": 3.000478633788065e-05, "loss": 0.3087, "step": 19699 }, { "epoch": 1.99, "learning_rate": 3.0002393168947938e-05, "loss": 0.4379, "step": 19700 }, { "epoch": 1.99, "learning_rate": 3e-05, "loss": 0.7004, "step": 19701 }, { "epoch": 1.99, "learning_rate": 2.9997606831052067e-05, "loss": 1.1201, "step": 19702 }, { "epoch": 1.99, "learning_rate": 2.999521366211935e-05, "loss": 0.142, "step": 19703 }, { "epoch": 1.99, "learning_rate": 2.9992820493217104e-05, "loss": 0.1334, "step": 19704 }, { "epoch": 1.99, "learning_rate": 2.999042732436054e-05, "loss": 0.3604, "step": 19705 }, { "epoch": 1.99, "learning_rate": 2.998803415556489e-05, "loss": 0.6859, "step": 19706 }, { "epoch": 1.99, "learning_rate": 2.998564098684539e-05, "loss": 0.2726, "step": 19707 }, { "epoch": 1.99, "learning_rate": 2.9983247818217272e-05, "loss": 1.1496, "step": 19708 }, { "epoch": 1.99, "learning_rate": 2.9980854649695745e-05, "loss": 0.4296, "step": 19709 }, { "epoch": 1.99, "learning_rate": 2.997846148129606e-05, "loss": 0.33, "step": 19710 }, { "epoch": 1.99, "learning_rate": 2.9976068313033435e-05, "loss": 0.4283, "step": 19711 }, { "epoch": 1.99, "learning_rate": 2.9973675144923098e-05, "loss": 0.7445, "step": 19712 }, { "epoch": 1.99, "learning_rate": 2.9971281976980294e-05, "loss": 0.319, "step": 19713 }, { "epoch": 1.99, "learning_rate": 2.9968888809220235e-05, "loss": 1.0265, "step": 19714 }, { "epoch": 1.99, "learning_rate": 2.9966495641658145e-05, "loss": 0.7572, "step": 19715 }, { "epoch": 1.99, "learning_rate": 2.9964102474309277e-05, "loss": 0.0881, "step": 19716 }, { "epoch": 1.99, "learning_rate": 2.996170930718884e-05, "loss": 0.0978, "step": 19717 }, { "epoch": 1.99, "learning_rate": 2.9959316140312082e-05, "loss": 1.1073, "step": 19718 }, { "epoch": 1.99, "learning_rate": 2.9956922973694213e-05, "loss": 0.6229, "step": 19719 }, { "epoch": 1.99, "learning_rate": 2.9954529807350465e-05, "loss": 0.3708, "step": 19720 }, { "epoch": 1.99, "learning_rate": 2.9952136641296074e-05, "loss": 0.4723, "step": 19721 }, { "epoch": 1.99, "learning_rate": 2.994974347554628e-05, "loss": 1.1126, "step": 19722 }, { "epoch": 1.99, "learning_rate": 2.9947350310116283e-05, "loss": 0.3115, "step": 19723 }, { "epoch": 1.99, "learning_rate": 2.9944957145021335e-05, "loss": 0.1029, "step": 19724 }, { "epoch": 1.99, "learning_rate": 2.994256398027666e-05, "loss": 0.4889, "step": 19725 }, { "epoch": 1.99, "learning_rate": 2.9940170815897476e-05, "loss": 0.6083, "step": 19726 }, { "epoch": 1.99, "learning_rate": 2.993777765189904e-05, "loss": 0.209, "step": 19727 }, { "epoch": 1.99, "learning_rate": 2.9935384488296555e-05, "loss": 1.0779, "step": 19728 }, { "epoch": 1.99, "learning_rate": 2.9932991325105253e-05, "loss": 0.3625, "step": 19729 }, { "epoch": 1.99, "learning_rate": 2.9930598162340372e-05, "loss": 0.5125, "step": 19730 }, { "epoch": 1.99, "learning_rate": 2.9928205000017147e-05, "loss": 0.8521, "step": 19731 }, { "epoch": 1.99, "learning_rate": 2.9925811838150784e-05, "loss": 0.4183, "step": 19732 }, { "epoch": 1.99, "learning_rate": 2.9923418676756534e-05, "loss": 0.7268, "step": 19733 }, { "epoch": 1.99, "learning_rate": 2.992102551584962e-05, "loss": 0.3358, "step": 19734 }, { "epoch": 1.99, "learning_rate": 2.991863235544526e-05, "loss": 1.2385, "step": 19735 }, { "epoch": 1.99, "learning_rate": 2.9916239195558708e-05, "loss": 0.3348, "step": 19736 }, { "epoch": 1.99, "learning_rate": 2.9913846036205167e-05, "loss": 0.4096, "step": 19737 }, { "epoch": 1.99, "learning_rate": 2.9911452877399877e-05, "loss": 1.357, "step": 19738 }, { "epoch": 1.99, "learning_rate": 2.9909059719158066e-05, "loss": 0.5683, "step": 19739 }, { "epoch": 1.99, "learning_rate": 2.9906666561494976e-05, "loss": 0.3708, "step": 19740 }, { "epoch": 1.99, "learning_rate": 2.9904273404425812e-05, "loss": 0.406, "step": 19741 }, { "epoch": 1.99, "learning_rate": 2.990188024796582e-05, "loss": 0.7561, "step": 19742 }, { "epoch": 1.99, "learning_rate": 2.989948709213022e-05, "loss": 0.3895, "step": 19743 }, { "epoch": 1.99, "learning_rate": 2.989709393693425e-05, "loss": 1.0559, "step": 19744 }, { "epoch": 1.99, "learning_rate": 2.9894700782393143e-05, "loss": 0.2214, "step": 19745 }, { "epoch": 1.99, "learning_rate": 2.9892307628522107e-05, "loss": 0.4425, "step": 19746 }, { "epoch": 1.99, "learning_rate": 2.9889914475336392e-05, "loss": 0.2998, "step": 19747 }, { "epoch": 1.99, "learning_rate": 2.9887521322851218e-05, "loss": 0.5588, "step": 19748 }, { "epoch": 1.99, "learning_rate": 2.9885128171081807e-05, "loss": 0.6433, "step": 19749 }, { "epoch": 1.99, "learning_rate": 2.9882735020043412e-05, "loss": 0.8831, "step": 19750 }, { "epoch": 1.99, "learning_rate": 2.9880341869751243e-05, "loss": 0.5492, "step": 19751 }, { "epoch": 1.99, "learning_rate": 2.9877948720220523e-05, "loss": 1.0609, "step": 19752 }, { "epoch": 1.99, "learning_rate": 2.9875555571466495e-05, "loss": 0.9946, "step": 19753 }, { "epoch": 1.99, "learning_rate": 2.9873162423504395e-05, "loss": 0.6651, "step": 19754 }, { "epoch": 1.99, "learning_rate": 2.9870769276349426e-05, "loss": 0.7438, "step": 19755 }, { "epoch": 1.99, "learning_rate": 2.986837613001684e-05, "loss": 0.3875, "step": 19756 }, { "epoch": 1.99, "learning_rate": 2.986598298452186e-05, "loss": 0.1761, "step": 19757 }, { "epoch": 1.99, "learning_rate": 2.9863589839879703e-05, "loss": 0.5507, "step": 19758 }, { "epoch": 1.99, "learning_rate": 2.9861196696105622e-05, "loss": 1.1398, "step": 19759 }, { "epoch": 1.99, "learning_rate": 2.9858803553214827e-05, "loss": 0.5606, "step": 19760 }, { "epoch": 1.99, "learning_rate": 2.9856410411222544e-05, "loss": 1.0855, "step": 19761 }, { "epoch": 1.99, "learning_rate": 2.9854017270144017e-05, "loss": 0.8499, "step": 19762 }, { "epoch": 1.99, "learning_rate": 2.9851624129994478e-05, "loss": 0.5429, "step": 19763 }, { "epoch": 1.99, "learning_rate": 2.9849230990789134e-05, "loss": 0.5493, "step": 19764 }, { "epoch": 1.99, "learning_rate": 2.984683785254323e-05, "loss": 0.5058, "step": 19765 }, { "epoch": 1.99, "learning_rate": 2.9844444715271994e-05, "loss": 1.0683, "step": 19766 }, { "epoch": 1.99, "learning_rate": 2.9842051578990644e-05, "loss": 0.1453, "step": 19767 }, { "epoch": 1.99, "learning_rate": 2.9839658443714432e-05, "loss": 0.5869, "step": 19768 }, { "epoch": 1.99, "learning_rate": 2.9837265309458558e-05, "loss": 0.1485, "step": 19769 }, { "epoch": 1.99, "learning_rate": 2.9834872176238274e-05, "loss": 0.5706, "step": 19770 }, { "epoch": 1.99, "learning_rate": 2.98324790440688e-05, "loss": 0.8645, "step": 19771 }, { "epoch": 1.99, "learning_rate": 2.9830085912965358e-05, "loss": 0.396, "step": 19772 }, { "epoch": 1.99, "learning_rate": 2.98276927829432e-05, "loss": 0.1956, "step": 19773 }, { "epoch": 1.99, "learning_rate": 2.982529965401753e-05, "loss": 0.3338, "step": 19774 }, { "epoch": 2.0, "learning_rate": 2.9822906526203584e-05, "loss": 0.2367, "step": 19775 }, { "epoch": 2.0, "learning_rate": 2.9820513399516597e-05, "loss": 0.3836, "step": 19776 }, { "epoch": 2.0, "learning_rate": 2.9818120273971804e-05, "loss": 0.9391, "step": 19777 }, { "epoch": 2.0, "learning_rate": 2.981572714958441e-05, "loss": 0.5751, "step": 19778 }, { "epoch": 2.0, "learning_rate": 2.981333402636966e-05, "loss": 0.7343, "step": 19779 }, { "epoch": 2.0, "learning_rate": 2.981094090434279e-05, "loss": 0.6018, "step": 19780 }, { "epoch": 2.0, "learning_rate": 2.9808547783519008e-05, "loss": 0.2139, "step": 19781 }, { "epoch": 2.0, "learning_rate": 2.980615466391357e-05, "loss": 1.2534, "step": 19782 }, { "epoch": 2.0, "learning_rate": 2.980376154554168e-05, "loss": 1.0876, "step": 19783 }, { "epoch": 2.0, "learning_rate": 2.9801368428418573e-05, "loss": 1.2762, "step": 19784 }, { "epoch": 2.0, "learning_rate": 2.979897531255949e-05, "loss": 0.5876, "step": 19785 }, { "epoch": 2.0, "learning_rate": 2.9796582197979656e-05, "loss": 0.5687, "step": 19786 }, { "epoch": 2.0, "learning_rate": 2.9794189084694282e-05, "loss": 1.4323, "step": 19787 }, { "epoch": 2.0, "learning_rate": 2.9791795972718618e-05, "loss": 1.0469, "step": 19788 }, { "epoch": 2.0, "learning_rate": 2.9789402862067886e-05, "loss": 1.023, "step": 19789 }, { "epoch": 2.0, "learning_rate": 2.9787009752757305e-05, "loss": 0.9073, "step": 19790 }, { "epoch": 2.0, "learning_rate": 2.978461664480213e-05, "loss": 0.9169, "step": 19791 }, { "epoch": 2.0, "learning_rate": 2.9782223538217563e-05, "loss": 0.3228, "step": 19792 }, { "epoch": 2.0, "learning_rate": 2.9779830433018836e-05, "loss": 1.0334, "step": 19793 }, { "epoch": 2.0, "learning_rate": 2.9777437329221193e-05, "loss": 1.2424, "step": 19794 }, { "epoch": 2.0, "learning_rate": 2.977504422683986e-05, "loss": 0.761, "step": 19795 }, { "epoch": 2.0, "learning_rate": 2.9772651125890044e-05, "loss": 0.4719, "step": 19796 }, { "epoch": 2.0, "learning_rate": 2.9770258026387e-05, "loss": 0.6229, "step": 19797 }, { "epoch": 2.0, "learning_rate": 2.9767864928345937e-05, "loss": 0.1792, "step": 19798 }, { "epoch": 2.0, "learning_rate": 2.97654718317821e-05, "loss": 0.7307, "step": 19799 }, { "epoch": 2.0, "learning_rate": 2.976307873671072e-05, "loss": 0.3157, "step": 19800 }, { "epoch": 2.0, "learning_rate": 2.9760685643147e-05, "loss": 0.6896, "step": 19801 }, { "epoch": 2.0, "learning_rate": 2.97582925511062e-05, "loss": 1.0113, "step": 19802 }, { "epoch": 2.0, "learning_rate": 2.9755899460603525e-05, "loss": 0.6655, "step": 19803 }, { "epoch": 2.0, "learning_rate": 2.9753506371654214e-05, "loss": 0.624, "step": 19804 }, { "epoch": 2.0, "learning_rate": 2.9751113284273503e-05, "loss": 0.3663, "step": 19805 }, { "epoch": 2.0, "learning_rate": 2.9748720198476608e-05, "loss": 1.0613, "step": 19806 }, { "epoch": 2.0, "learning_rate": 2.9746327114278752e-05, "loss": 0.6281, "step": 19807 }, { "epoch": 2.0, "learning_rate": 2.9743934031695185e-05, "loss": 0.5509, "step": 19808 }, { "epoch": 2.0, "learning_rate": 2.9741540950741126e-05, "loss": 0.5895, "step": 19809 }, { "epoch": 2.0, "learning_rate": 2.9739147871431792e-05, "loss": 0.5653, "step": 19810 }, { "epoch": 2.0, "learning_rate": 2.973675479378243e-05, "loss": 0.7002, "step": 19811 }, { "epoch": 2.0, "learning_rate": 2.973436171780826e-05, "loss": 1.2111, "step": 19812 }, { "epoch": 2.0, "learning_rate": 2.97319686435245e-05, "loss": 1.4914, "step": 19813 }, { "epoch": 2.0, "learning_rate": 2.9729575570946406e-05, "loss": 0.9135, "step": 19814 }, { "epoch": 2.0, "learning_rate": 2.972718250008918e-05, "loss": 0.5307, "step": 19815 }, { "epoch": 2.0, "learning_rate": 2.972478943096806e-05, "loss": 0.7211, "step": 19816 }, { "epoch": 2.0, "learning_rate": 2.9722396363598282e-05, "loss": 0.3394, "step": 19817 }, { "epoch": 2.0, "learning_rate": 2.972000329799507e-05, "loss": 1.0582, "step": 19818 }, { "epoch": 2.0, "learning_rate": 2.9717610234173638e-05, "loss": 0.4391, "step": 19819 }, { "epoch": 2.0, "learning_rate": 2.9715217172149238e-05, "loss": 0.9401, "step": 19820 }, { "epoch": 2.0, "learning_rate": 2.971282411193708e-05, "loss": 0.2127, "step": 19821 }, { "epoch": 2.0, "learning_rate": 2.97104310535524e-05, "loss": 0.2705, "step": 19822 }, { "epoch": 2.0, "learning_rate": 2.970803799701044e-05, "loss": 0.1156, "step": 19823 }, { "epoch": 2.0, "learning_rate": 2.970564494232639e-05, "loss": 1.2774, "step": 19824 }, { "epoch": 2.0, "learning_rate": 2.9703251889515526e-05, "loss": 0.7943, "step": 19825 }, { "epoch": 2.0, "learning_rate": 2.9700858838593047e-05, "loss": 1.1219, "step": 19826 }, { "epoch": 2.0, "learning_rate": 2.9698465789574184e-05, "loss": 0.489, "step": 19827 }, { "epoch": 2.0, "learning_rate": 2.9696072742474182e-05, "loss": 0.4348, "step": 19828 }, { "epoch": 2.0, "learning_rate": 2.9693679697308252e-05, "loss": 0.4158, "step": 19829 }, { "epoch": 2.0, "learning_rate": 2.969128665409162e-05, "loss": 0.2515, "step": 19830 }, { "epoch": 2.0, "learning_rate": 2.968889361283953e-05, "loss": 1.2987, "step": 19831 }, { "epoch": 2.0, "learning_rate": 2.9686500573567206e-05, "loss": 0.8552, "step": 19832 }, { "epoch": 2.0, "learning_rate": 2.9684107536289863e-05, "loss": 0.2089, "step": 19833 }, { "epoch": 2.0, "learning_rate": 2.9681714501022742e-05, "loss": 0.3828, "step": 19834 }, { "epoch": 2.0, "learning_rate": 2.9679321467781073e-05, "loss": 0.1768, "step": 19835 }, { "epoch": 2.0, "learning_rate": 2.9676928436580073e-05, "loss": 0.5552, "step": 19836 }, { "epoch": 2.0, "learning_rate": 2.967453540743499e-05, "loss": 1.0975, "step": 19837 }, { "epoch": 2.0, "learning_rate": 2.9672142380361034e-05, "loss": 0.5661, "step": 19838 }, { "epoch": 2.0, "learning_rate": 2.9669749355373434e-05, "loss": 1.2928, "step": 19839 }, { "epoch": 2.0, "learning_rate": 2.9667356332487428e-05, "loss": 0.6474, "step": 19840 }, { "epoch": 2.0, "learning_rate": 2.966496331171825e-05, "loss": 1.0023, "step": 19841 }, { "epoch": 2.0, "learning_rate": 2.96625702930811e-05, "loss": 0.5246, "step": 19842 }, { "epoch": 2.0, "learning_rate": 2.966017727659123e-05, "loss": 0.4613, "step": 19843 }, { "epoch": 2.0, "learning_rate": 2.9657784262263866e-05, "loss": 0.4892, "step": 19844 }, { "epoch": 2.0, "learning_rate": 2.9655391250114226e-05, "loss": 0.8391, "step": 19845 }, { "epoch": 2.0, "learning_rate": 2.9652998240157558e-05, "loss": 0.174, "step": 19846 }, { "epoch": 2.0, "learning_rate": 2.965060523240907e-05, "loss": 0.2365, "step": 19847 }, { "epoch": 2.0, "learning_rate": 2.964821222688399e-05, "loss": 0.3549, "step": 19848 }, { "epoch": 2.0, "learning_rate": 2.964581922359756e-05, "loss": 0.1381, "step": 19849 }, { "epoch": 2.0, "learning_rate": 2.9643426222564994e-05, "loss": 0.5464, "step": 19850 }, { "epoch": 2.0, "learning_rate": 2.9641033223801544e-05, "loss": 1.0063, "step": 19851 }, { "epoch": 2.0, "learning_rate": 2.9638640227322414e-05, "loss": 0.7857, "step": 19852 }, { "epoch": 2.0, "learning_rate": 2.9636247233142835e-05, "loss": 1.3938, "step": 19853 }, { "epoch": 2.0, "learning_rate": 2.9633854241278045e-05, "loss": 0.1327, "step": 19854 }, { "epoch": 2.0, "learning_rate": 2.9631461251743273e-05, "loss": 0.6544, "step": 19855 }, { "epoch": 2.0, "learning_rate": 2.9629068264553726e-05, "loss": 0.312, "step": 19856 }, { "epoch": 2.0, "learning_rate": 2.9626675279724662e-05, "loss": 0.2632, "step": 19857 }, { "epoch": 2.0, "learning_rate": 2.9624282297271293e-05, "loss": 0.8351, "step": 19858 }, { "epoch": 2.0, "learning_rate": 2.962188931720884e-05, "loss": 1.2085, "step": 19859 }, { "epoch": 2.0, "learning_rate": 2.961949633955255e-05, "loss": 0.5105, "step": 19860 }, { "epoch": 2.0, "learning_rate": 2.961710336431764e-05, "loss": 0.3524, "step": 19861 }, { "epoch": 2.0, "learning_rate": 2.9614710391519327e-05, "loss": 0.2436, "step": 19862 }, { "epoch": 2.0, "learning_rate": 2.9612317421172856e-05, "loss": 0.4169, "step": 19863 }, { "epoch": 2.0, "learning_rate": 2.9609924453293458e-05, "loss": 0.2531, "step": 19864 }, { "epoch": 2.0, "learning_rate": 2.9607531487896336e-05, "loss": 0.8456, "step": 19865 }, { "epoch": 2.0, "learning_rate": 2.9605138524996746e-05, "loss": 0.2801, "step": 19866 }, { "epoch": 2.0, "learning_rate": 2.9602745564609903e-05, "loss": 0.7938, "step": 19867 }, { "epoch": 2.0, "learning_rate": 2.960035260675103e-05, "loss": 0.4847, "step": 19868 }, { "epoch": 2.0, "learning_rate": 2.9597959651435372e-05, "loss": 0.5287, "step": 19869 }, { "epoch": 2.0, "learning_rate": 2.9595566698678144e-05, "loss": 0.6538, "step": 19870 }, { "epoch": 2.0, "learning_rate": 2.9593173748494565e-05, "loss": 0.454, "step": 19871 }, { "epoch": 2.0, "learning_rate": 2.9590780800899877e-05, "loss": 0.8857, "step": 19872 }, { "epoch": 2.0, "learning_rate": 2.9588387855909314e-05, "loss": 0.9218, "step": 19873 }, { "epoch": 2.01, "learning_rate": 2.9585994913538086e-05, "loss": 0.8058, "step": 19874 }, { "epoch": 2.01, "learning_rate": 2.9583601973801425e-05, "loss": 0.7572, "step": 19875 }, { "epoch": 2.01, "learning_rate": 2.958120903671457e-05, "loss": 0.6683, "step": 19876 }, { "epoch": 2.01, "learning_rate": 2.9578816102292736e-05, "loss": 0.642, "step": 19877 }, { "epoch": 2.01, "learning_rate": 2.9576423170551165e-05, "loss": 1.3407, "step": 19878 }, { "epoch": 2.01, "learning_rate": 2.957403024150506e-05, "loss": 0.8369, "step": 19879 }, { "epoch": 2.01, "learning_rate": 2.957163731516968e-05, "loss": 0.4708, "step": 19880 }, { "epoch": 2.01, "learning_rate": 2.956924439156023e-05, "loss": 0.9516, "step": 19881 }, { "epoch": 2.01, "learning_rate": 2.9566851470691942e-05, "loss": 0.5359, "step": 19882 }, { "epoch": 2.01, "learning_rate": 2.956445855258006e-05, "loss": 0.9826, "step": 19883 }, { "epoch": 2.01, "learning_rate": 2.956206563723979e-05, "loss": 0.6317, "step": 19884 }, { "epoch": 2.01, "learning_rate": 2.955967272468636e-05, "loss": 0.4869, "step": 19885 }, { "epoch": 2.01, "learning_rate": 2.955727981493501e-05, "loss": 0.4777, "step": 19886 }, { "epoch": 2.01, "learning_rate": 2.9554886908000975e-05, "loss": 0.4821, "step": 19887 }, { "epoch": 2.01, "learning_rate": 2.9552494003899447e-05, "loss": 0.8667, "step": 19888 }, { "epoch": 2.01, "learning_rate": 2.9550101102645695e-05, "loss": 0.2737, "step": 19889 }, { "epoch": 2.01, "learning_rate": 2.9547708204254928e-05, "loss": 0.4639, "step": 19890 }, { "epoch": 2.01, "learning_rate": 2.9545315308742365e-05, "loss": 0.8605, "step": 19891 }, { "epoch": 2.01, "learning_rate": 2.954292241612325e-05, "loss": 0.4625, "step": 19892 }, { "epoch": 2.01, "learning_rate": 2.95405295264128e-05, "loss": 0.9956, "step": 19893 }, { "epoch": 2.01, "learning_rate": 2.9538136639626243e-05, "loss": 0.6824, "step": 19894 }, { "epoch": 2.01, "learning_rate": 2.953574375577881e-05, "loss": 0.2843, "step": 19895 }, { "epoch": 2.01, "learning_rate": 2.9533350874885736e-05, "loss": 0.7038, "step": 19896 }, { "epoch": 2.01, "learning_rate": 2.9530957996962226e-05, "loss": 0.3307, "step": 19897 }, { "epoch": 2.01, "learning_rate": 2.9528565122023527e-05, "loss": 0.6508, "step": 19898 }, { "epoch": 2.01, "learning_rate": 2.952617225008486e-05, "loss": 0.1519, "step": 19899 }, { "epoch": 2.01, "learning_rate": 2.9523779381161444e-05, "loss": 0.1905, "step": 19900 }, { "epoch": 2.01, "learning_rate": 2.9521386515268533e-05, "loss": 0.504, "step": 19901 }, { "epoch": 2.01, "learning_rate": 2.9518993652421324e-05, "loss": 0.7027, "step": 19902 }, { "epoch": 2.01, "learning_rate": 2.9516600792635053e-05, "loss": 0.3939, "step": 19903 }, { "epoch": 2.01, "learning_rate": 2.9514207935924957e-05, "loss": 0.835, "step": 19904 }, { "epoch": 2.01, "learning_rate": 2.951181508230625e-05, "loss": 1.264, "step": 19905 }, { "epoch": 2.01, "learning_rate": 2.950942223179418e-05, "loss": 0.8103, "step": 19906 }, { "epoch": 2.01, "learning_rate": 2.950702938440395e-05, "loss": 0.4232, "step": 19907 }, { "epoch": 2.01, "learning_rate": 2.9504636540150797e-05, "loss": 0.483, "step": 19908 }, { "epoch": 2.01, "learning_rate": 2.9502243699049953e-05, "loss": 1.3305, "step": 19909 }, { "epoch": 2.01, "learning_rate": 2.9499850861116647e-05, "loss": 0.947, "step": 19910 }, { "epoch": 2.01, "learning_rate": 2.9497458026366078e-05, "loss": 0.1359, "step": 19911 }, { "epoch": 2.01, "learning_rate": 2.9495065194813518e-05, "loss": 0.2768, "step": 19912 }, { "epoch": 2.01, "learning_rate": 2.949267236647416e-05, "loss": 0.1635, "step": 19913 }, { "epoch": 2.01, "learning_rate": 2.949027954136324e-05, "loss": 0.7184, "step": 19914 }, { "epoch": 2.01, "learning_rate": 2.9487886719496e-05, "loss": 0.8536, "step": 19915 }, { "epoch": 2.01, "learning_rate": 2.9485493900887646e-05, "loss": 0.1208, "step": 19916 }, { "epoch": 2.01, "learning_rate": 2.9483101085553406e-05, "loss": 1.2402, "step": 19917 }, { "epoch": 2.01, "learning_rate": 2.948070827350852e-05, "loss": 0.4546, "step": 19918 }, { "epoch": 2.01, "learning_rate": 2.9478315464768218e-05, "loss": 0.6566, "step": 19919 }, { "epoch": 2.01, "learning_rate": 2.9475922659347695e-05, "loss": 0.3141, "step": 19920 }, { "epoch": 2.01, "learning_rate": 2.9473529857262226e-05, "loss": 0.4736, "step": 19921 }, { "epoch": 2.01, "learning_rate": 2.9471137058527e-05, "loss": 0.3435, "step": 19922 }, { "epoch": 2.01, "learning_rate": 2.9468744263157255e-05, "loss": 0.5908, "step": 19923 }, { "epoch": 2.01, "learning_rate": 2.946635147116823e-05, "loss": 0.4044, "step": 19924 }, { "epoch": 2.01, "learning_rate": 2.946395868257513e-05, "loss": 0.699, "step": 19925 }, { "epoch": 2.01, "learning_rate": 2.9461565897393193e-05, "loss": 0.5921, "step": 19926 }, { "epoch": 2.01, "learning_rate": 2.9459173115637653e-05, "loss": 1.1597, "step": 19927 }, { "epoch": 2.01, "learning_rate": 2.9456780337323733e-05, "loss": 0.3221, "step": 19928 }, { "epoch": 2.01, "learning_rate": 2.9454387562466643e-05, "loss": 0.6075, "step": 19929 }, { "epoch": 2.01, "learning_rate": 2.9451994791081632e-05, "loss": 0.6742, "step": 19930 }, { "epoch": 2.01, "learning_rate": 2.9449602023183917e-05, "loss": 0.2738, "step": 19931 }, { "epoch": 2.01, "learning_rate": 2.9447209258788714e-05, "loss": 0.2164, "step": 19932 }, { "epoch": 2.01, "learning_rate": 2.9444816497911283e-05, "loss": 0.4822, "step": 19933 }, { "epoch": 2.01, "learning_rate": 2.9442423740566802e-05, "loss": 0.1899, "step": 19934 }, { "epoch": 2.01, "learning_rate": 2.9440030986770546e-05, "loss": 0.6361, "step": 19935 }, { "epoch": 2.01, "learning_rate": 2.9437638236537713e-05, "loss": 0.2702, "step": 19936 }, { "epoch": 2.01, "learning_rate": 2.9435245489883527e-05, "loss": 0.0112, "step": 19937 }, { "epoch": 2.01, "learning_rate": 2.943285274682324e-05, "loss": 0.7421, "step": 19938 }, { "epoch": 2.01, "learning_rate": 2.943046000737206e-05, "loss": 0.3052, "step": 19939 }, { "epoch": 2.01, "learning_rate": 2.94280672715452e-05, "loss": 0.6627, "step": 19940 }, { "epoch": 2.01, "learning_rate": 2.9425674539357913e-05, "loss": 0.7184, "step": 19941 }, { "epoch": 2.01, "learning_rate": 2.9423281810825424e-05, "loss": 0.1291, "step": 19942 }, { "epoch": 2.01, "learning_rate": 2.942088908596293e-05, "loss": 0.4615, "step": 19943 }, { "epoch": 2.01, "learning_rate": 2.9418496364785695e-05, "loss": 0.332, "step": 19944 }, { "epoch": 2.01, "learning_rate": 2.941610364730892e-05, "loss": 0.4111, "step": 19945 }, { "epoch": 2.01, "learning_rate": 2.941371093354784e-05, "loss": 0.3467, "step": 19946 }, { "epoch": 2.01, "learning_rate": 2.9411318223517687e-05, "loss": 0.3205, "step": 19947 }, { "epoch": 2.01, "learning_rate": 2.9408925517233673e-05, "loss": 0.4282, "step": 19948 }, { "epoch": 2.0, "learning_rate": 2.9406532814711032e-05, "loss": 0.0754, "step": 19949 }, { "epoch": 2.0, "learning_rate": 2.9404140115964992e-05, "loss": 0.3332, "step": 19950 }, { "epoch": 2.0, "learning_rate": 2.9401747421010786e-05, "loss": 0.2341, "step": 19951 }, { "epoch": 2.0, "learning_rate": 2.939935472986361e-05, "loss": 0.3224, "step": 19952 }, { "epoch": 2.0, "learning_rate": 2.9396962042538733e-05, "loss": 0.3136, "step": 19953 }, { "epoch": 2.0, "learning_rate": 2.9394569359051357e-05, "loss": 0.1997, "step": 19954 }, { "epoch": 2.0, "learning_rate": 2.93921766794167e-05, "loss": 0.3335, "step": 19955 }, { "epoch": 2.0, "learning_rate": 2.9389784003650015e-05, "loss": 0.3716, "step": 19956 }, { "epoch": 2.0, "learning_rate": 2.9387391331766503e-05, "loss": 0.4339, "step": 19957 }, { "epoch": 2.0, "learning_rate": 2.9384998663781396e-05, "loss": 0.1928, "step": 19958 }, { "epoch": 2.0, "learning_rate": 2.9382605999709927e-05, "loss": 0.4754, "step": 19959 }, { "epoch": 2.0, "learning_rate": 2.938021333956732e-05, "loss": 0.3719, "step": 19960 }, { "epoch": 2.0, "learning_rate": 2.9377820683368804e-05, "loss": 0.2818, "step": 19961 }, { "epoch": 2.0, "learning_rate": 2.9375428031129596e-05, "loss": 0.4674, "step": 19962 }, { "epoch": 2.0, "learning_rate": 2.937303538286492e-05, "loss": 0.4003, "step": 19963 }, { "epoch": 2.0, "learning_rate": 2.9370642738590017e-05, "loss": 0.4706, "step": 19964 }, { "epoch": 2.0, "learning_rate": 2.9368250098320106e-05, "loss": 0.4157, "step": 19965 }, { "epoch": 2.0, "learning_rate": 2.9365857462070396e-05, "loss": 0.4127, "step": 19966 }, { "epoch": 2.0, "learning_rate": 2.9363464829856148e-05, "loss": 0.6744, "step": 19967 }, { "epoch": 2.0, "learning_rate": 2.936107220169256e-05, "loss": 0.2636, "step": 19968 }, { "epoch": 2.0, "learning_rate": 2.9358679577594857e-05, "loss": 0.6287, "step": 19969 }, { "epoch": 2.0, "learning_rate": 2.935628695757829e-05, "loss": 0.2121, "step": 19970 }, { "epoch": 2.0, "learning_rate": 2.9353894341658055e-05, "loss": 0.2002, "step": 19971 }, { "epoch": 2.0, "learning_rate": 2.935150172984939e-05, "loss": 0.0946, "step": 19972 }, { "epoch": 2.0, "learning_rate": 2.9349109122167526e-05, "loss": 0.4228, "step": 19973 }, { "epoch": 2.0, "learning_rate": 2.934671651862769e-05, "loss": 0.4778, "step": 19974 }, { "epoch": 2.0, "learning_rate": 2.9344323919245084e-05, "loss": 0.2897, "step": 19975 }, { "epoch": 2.0, "learning_rate": 2.9341931324034964e-05, "loss": 0.8145, "step": 19976 }, { "epoch": 2.0, "learning_rate": 2.9339538733012544e-05, "loss": 0.9056, "step": 19977 }, { "epoch": 2.0, "learning_rate": 2.9337146146193038e-05, "loss": 0.2064, "step": 19978 }, { "epoch": 2.0, "learning_rate": 2.93347535635917e-05, "loss": 1.0471, "step": 19979 }, { "epoch": 2.0, "learning_rate": 2.9332360985223724e-05, "loss": 0.3446, "step": 19980 }, { "epoch": 2.0, "learning_rate": 2.9329968411104342e-05, "loss": 0.146, "step": 19981 }, { "epoch": 2.0, "learning_rate": 2.9327575841248795e-05, "loss": 0.2277, "step": 19982 }, { "epoch": 2.0, "learning_rate": 2.9325183275672306e-05, "loss": 1.9836, "step": 19983 }, { "epoch": 2.0, "learning_rate": 2.9322790714390077e-05, "loss": 0.2476, "step": 19984 }, { "epoch": 2.0, "learning_rate": 2.9320398157417366e-05, "loss": 0.2559, "step": 19985 }, { "epoch": 2.0, "learning_rate": 2.931800560476937e-05, "loss": 0.3016, "step": 19986 }, { "epoch": 2.0, "learning_rate": 2.9315613056461336e-05, "loss": 0.2921, "step": 19987 }, { "epoch": 2.0, "learning_rate": 2.9313220512508486e-05, "loss": 0.2283, "step": 19988 }, { "epoch": 2.0, "learning_rate": 2.931082797292602e-05, "loss": 0.5678, "step": 19989 }, { "epoch": 2.0, "learning_rate": 2.93084354377292e-05, "loss": 0.0452, "step": 19990 }, { "epoch": 2.0, "learning_rate": 2.930604290693323e-05, "loss": 0.6168, "step": 19991 }, { "epoch": 2.0, "learning_rate": 2.9303650380553328e-05, "loss": 0.3723, "step": 19992 }, { "epoch": 2.0, "learning_rate": 2.9301257858604748e-05, "loss": 0.176, "step": 19993 }, { "epoch": 2.0, "learning_rate": 2.9298865341102684e-05, "loss": 0.1634, "step": 19994 }, { "epoch": 2.0, "learning_rate": 2.9296472828062374e-05, "loss": 0.1109, "step": 19995 }, { "epoch": 2.0, "learning_rate": 2.929408031949905e-05, "loss": 0.2097, "step": 19996 }, { "epoch": 2.0, "learning_rate": 2.9291687815427936e-05, "loss": 0.1472, "step": 19997 }, { "epoch": 2.01, "learning_rate": 2.9289295315864228e-05, "loss": 0.4224, "step": 19998 }, { "epoch": 2.01, "learning_rate": 2.9286902820823198e-05, "loss": 0.2154, "step": 19999 }, { "epoch": 2.01, "learning_rate": 2.928451033032004e-05, "loss": 0.4471, "step": 20000 }, { "epoch": 2.01, "learning_rate": 2.9282117844369976e-05, "loss": 0.2419, "step": 20001 }, { "epoch": 2.01, "learning_rate": 2.9279725362988257e-05, "loss": 0.8367, "step": 20002 }, { "epoch": 2.01, "learning_rate": 2.9277332886190087e-05, "loss": 0.2849, "step": 20003 }, { "epoch": 2.01, "learning_rate": 2.9274940413990683e-05, "loss": 0.2572, "step": 20004 }, { "epoch": 2.01, "learning_rate": 2.9272547946405297e-05, "loss": 0.2784, "step": 20005 }, { "epoch": 2.01, "learning_rate": 2.927015548344914e-05, "loss": 0.197, "step": 20006 }, { "epoch": 2.01, "learning_rate": 2.9267763025137417e-05, "loss": 0.3327, "step": 20007 }, { "epoch": 2.01, "learning_rate": 2.926537057148539e-05, "loss": 0.6762, "step": 20008 }, { "epoch": 2.01, "learning_rate": 2.9262978122508263e-05, "loss": 0.1831, "step": 20009 }, { "epoch": 2.01, "learning_rate": 2.9260585678221254e-05, "loss": 1.1088, "step": 20010 }, { "epoch": 2.01, "learning_rate": 2.9258193238639608e-05, "loss": 0.1715, "step": 20011 }, { "epoch": 2.01, "learning_rate": 2.9255800803778534e-05, "loss": 0.4766, "step": 20012 }, { "epoch": 2.01, "learning_rate": 2.9253408373653252e-05, "loss": 0.5574, "step": 20013 }, { "epoch": 2.01, "learning_rate": 2.9251015948279005e-05, "loss": 0.1296, "step": 20014 }, { "epoch": 2.01, "learning_rate": 2.9248623527671e-05, "loss": 0.6484, "step": 20015 }, { "epoch": 2.01, "learning_rate": 2.9246231111844473e-05, "loss": 0.6091, "step": 20016 }, { "epoch": 2.01, "learning_rate": 2.9243838700814654e-05, "loss": 0.3213, "step": 20017 }, { "epoch": 2.01, "learning_rate": 2.924144629459674e-05, "loss": 0.4285, "step": 20018 }, { "epoch": 2.01, "learning_rate": 2.923905389320598e-05, "loss": 0.2728, "step": 20019 }, { "epoch": 2.01, "learning_rate": 2.9236661496657607e-05, "loss": 0.4348, "step": 20020 }, { "epoch": 2.01, "learning_rate": 2.92342691049668e-05, "loss": 0.0886, "step": 20021 }, { "epoch": 2.01, "learning_rate": 2.923187671814884e-05, "loss": 0.3042, "step": 20022 }, { "epoch": 2.01, "learning_rate": 2.9229484336218915e-05, "loss": 0.2849, "step": 20023 }, { "epoch": 2.01, "learning_rate": 2.922709195919225e-05, "loss": 0.0976, "step": 20024 }, { "epoch": 2.01, "learning_rate": 2.922469958708409e-05, "loss": 0.1249, "step": 20025 }, { "epoch": 2.01, "learning_rate": 2.9222307219909644e-05, "loss": 0.2555, "step": 20026 }, { "epoch": 2.01, "learning_rate": 2.9219914857684132e-05, "loss": 0.65, "step": 20027 }, { "epoch": 2.01, "learning_rate": 2.9217522500422792e-05, "loss": 0.4853, "step": 20028 }, { "epoch": 2.01, "learning_rate": 2.9215130148140844e-05, "loss": 0.3203, "step": 20029 }, { "epoch": 2.01, "learning_rate": 2.9212737800853494e-05, "loss": 0.1854, "step": 20030 }, { "epoch": 2.01, "learning_rate": 2.9210345458576e-05, "loss": 0.8848, "step": 20031 }, { "epoch": 2.01, "learning_rate": 2.920795312132356e-05, "loss": 0.275, "step": 20032 }, { "epoch": 2.01, "learning_rate": 2.9205560789111395e-05, "loss": 0.3495, "step": 20033 }, { "epoch": 2.01, "learning_rate": 2.9203168461954755e-05, "loss": 0.4988, "step": 20034 }, { "epoch": 2.01, "learning_rate": 2.9200776139868844e-05, "loss": 0.1352, "step": 20035 }, { "epoch": 2.01, "learning_rate": 2.9198383822868878e-05, "loss": 0.4409, "step": 20036 }, { "epoch": 2.01, "learning_rate": 2.91959915109701e-05, "loss": 0.4061, "step": 20037 }, { "epoch": 2.01, "learning_rate": 2.9193599204187733e-05, "loss": 0.6009, "step": 20038 }, { "epoch": 2.01, "learning_rate": 2.9191206902536976e-05, "loss": 0.109, "step": 20039 }, { "epoch": 2.01, "learning_rate": 2.9188814606033093e-05, "loss": 0.7106, "step": 20040 }, { "epoch": 2.01, "learning_rate": 2.9186422314691265e-05, "loss": 0.5509, "step": 20041 }, { "epoch": 2.01, "learning_rate": 2.9184030028526747e-05, "loss": 0.142, "step": 20042 }, { "epoch": 2.01, "learning_rate": 2.918163774755476e-05, "loss": 0.2965, "step": 20043 }, { "epoch": 2.01, "learning_rate": 2.917924547179049e-05, "loss": 0.7682, "step": 20044 }, { "epoch": 2.01, "learning_rate": 2.917685320124922e-05, "loss": 0.1327, "step": 20045 }, { "epoch": 2.01, "learning_rate": 2.917446093594613e-05, "loss": 0.5627, "step": 20046 }, { "epoch": 2.01, "learning_rate": 2.9172068675896456e-05, "loss": 0.6324, "step": 20047 }, { "epoch": 2.01, "learning_rate": 2.916967642111542e-05, "loss": 0.3695, "step": 20048 }, { "epoch": 2.01, "learning_rate": 2.916728417161826e-05, "loss": 0.9522, "step": 20049 }, { "epoch": 2.01, "learning_rate": 2.9164891927420173e-05, "loss": 0.4595, "step": 20050 }, { "epoch": 2.01, "learning_rate": 2.91624996885364e-05, "loss": 0.5784, "step": 20051 }, { "epoch": 2.01, "learning_rate": 2.9160107454982174e-05, "loss": 0.406, "step": 20052 }, { "epoch": 2.01, "learning_rate": 2.9157715226772676e-05, "loss": 0.065, "step": 20053 }, { "epoch": 2.01, "learning_rate": 2.9155323003923184e-05, "loss": 0.1056, "step": 20054 }, { "epoch": 2.01, "learning_rate": 2.9152930786448884e-05, "loss": 0.3283, "step": 20055 }, { "epoch": 2.01, "learning_rate": 2.9150538574365007e-05, "loss": 0.2431, "step": 20056 }, { "epoch": 2.01, "learning_rate": 2.914814636768679e-05, "loss": 0.8502, "step": 20057 }, { "epoch": 2.01, "learning_rate": 2.914575416642944e-05, "loss": 0.4628, "step": 20058 }, { "epoch": 2.01, "learning_rate": 2.914336197060818e-05, "loss": 0.1553, "step": 20059 }, { "epoch": 2.01, "learning_rate": 2.9140969780238243e-05, "loss": 0.41, "step": 20060 }, { "epoch": 2.01, "learning_rate": 2.9138577595334857e-05, "loss": 0.4803, "step": 20061 }, { "epoch": 2.01, "learning_rate": 2.913618541591321e-05, "loss": 0.2013, "step": 20062 }, { "epoch": 2.01, "learning_rate": 2.9133793241988575e-05, "loss": 0.4683, "step": 20063 }, { "epoch": 2.01, "learning_rate": 2.9131401073576144e-05, "loss": 0.4993, "step": 20064 }, { "epoch": 2.01, "learning_rate": 2.912900891069114e-05, "loss": 0.7158, "step": 20065 }, { "epoch": 2.01, "learning_rate": 2.9126616753348803e-05, "loss": 1.4309, "step": 20066 }, { "epoch": 2.01, "learning_rate": 2.9124224601564324e-05, "loss": 0.7246, "step": 20067 }, { "epoch": 2.01, "learning_rate": 2.9121832455352966e-05, "loss": 0.4672, "step": 20068 }, { "epoch": 2.01, "learning_rate": 2.911944031472993e-05, "loss": 0.4343, "step": 20069 }, { "epoch": 2.01, "learning_rate": 2.911704817971043e-05, "loss": 0.2367, "step": 20070 }, { "epoch": 2.01, "learning_rate": 2.9114656050309705e-05, "loss": 0.3589, "step": 20071 }, { "epoch": 2.01, "learning_rate": 2.9112263926542983e-05, "loss": 0.4398, "step": 20072 }, { "epoch": 2.01, "learning_rate": 2.910987180842546e-05, "loss": 0.239, "step": 20073 }, { "epoch": 2.01, "learning_rate": 2.9107479695972385e-05, "loss": 0.5536, "step": 20074 }, { "epoch": 2.01, "learning_rate": 2.910508758919897e-05, "loss": 0.1985, "step": 20075 }, { "epoch": 2.01, "learning_rate": 2.910269548812042e-05, "loss": 0.2599, "step": 20076 }, { "epoch": 2.01, "learning_rate": 2.9100303392751994e-05, "loss": 0.2915, "step": 20077 }, { "epoch": 2.01, "learning_rate": 2.909791130310889e-05, "loss": 0.6467, "step": 20078 }, { "epoch": 2.01, "learning_rate": 2.9095519219206328e-05, "loss": 0.3638, "step": 20079 }, { "epoch": 2.01, "learning_rate": 2.9093127141059546e-05, "loss": 0.149, "step": 20080 }, { "epoch": 2.01, "learning_rate": 2.909073506868376e-05, "loss": 0.2986, "step": 20081 }, { "epoch": 2.01, "learning_rate": 2.908834300209418e-05, "loss": 0.6742, "step": 20082 }, { "epoch": 2.01, "learning_rate": 2.908595094130604e-05, "loss": 0.877, "step": 20083 }, { "epoch": 2.01, "learning_rate": 2.9083558886334574e-05, "loss": 0.2318, "step": 20084 }, { "epoch": 2.01, "learning_rate": 2.9081166837194967e-05, "loss": 0.1464, "step": 20085 }, { "epoch": 2.01, "learning_rate": 2.9078774793902486e-05, "loss": 0.2383, "step": 20086 }, { "epoch": 2.01, "learning_rate": 2.9076382756472327e-05, "loss": 2.3283, "step": 20087 }, { "epoch": 2.01, "learning_rate": 2.9073990724919706e-05, "loss": 0.4171, "step": 20088 }, { "epoch": 2.01, "learning_rate": 2.907159869925987e-05, "loss": 0.0474, "step": 20089 }, { "epoch": 2.01, "learning_rate": 2.906920667950802e-05, "loss": 0.5695, "step": 20090 }, { "epoch": 2.01, "learning_rate": 2.906681466567938e-05, "loss": 0.3548, "step": 20091 }, { "epoch": 2.01, "learning_rate": 2.906442265778918e-05, "loss": 0.3229, "step": 20092 }, { "epoch": 2.01, "learning_rate": 2.9062030655852647e-05, "loss": 0.4265, "step": 20093 }, { "epoch": 2.01, "learning_rate": 2.9059638659884974e-05, "loss": 0.3836, "step": 20094 }, { "epoch": 2.01, "learning_rate": 2.9057246669901423e-05, "loss": 0.3083, "step": 20095 }, { "epoch": 2.02, "learning_rate": 2.9054854685917183e-05, "loss": 0.5378, "step": 20096 }, { "epoch": 2.02, "learning_rate": 2.9052462707947497e-05, "loss": 0.9077, "step": 20097 }, { "epoch": 2.02, "learning_rate": 2.905007073600758e-05, "loss": 0.1759, "step": 20098 }, { "epoch": 2.02, "learning_rate": 2.904767877011263e-05, "loss": 0.834, "step": 20099 }, { "epoch": 2.02, "learning_rate": 2.9045286810277914e-05, "loss": 0.2524, "step": 20100 }, { "epoch": 2.02, "learning_rate": 2.904289485651862e-05, "loss": 0.3549, "step": 20101 }, { "epoch": 2.02, "learning_rate": 2.9040502908849977e-05, "loss": 0.3517, "step": 20102 }, { "epoch": 2.02, "learning_rate": 2.9038110967287214e-05, "loss": 0.2219, "step": 20103 }, { "epoch": 2.02, "learning_rate": 2.903571903184555e-05, "loss": 0.4687, "step": 20104 }, { "epoch": 2.02, "learning_rate": 2.903332710254019e-05, "loss": 0.3639, "step": 20105 }, { "epoch": 2.02, "learning_rate": 2.9030935179386378e-05, "loss": 0.1822, "step": 20106 }, { "epoch": 2.02, "learning_rate": 2.902854326239933e-05, "loss": 0.1844, "step": 20107 }, { "epoch": 2.02, "learning_rate": 2.9026151351594245e-05, "loss": 0.0966, "step": 20108 }, { "epoch": 2.02, "learning_rate": 2.9023759446986384e-05, "loss": 0.419, "step": 20109 }, { "epoch": 2.02, "learning_rate": 2.9021367548590937e-05, "loss": 0.3473, "step": 20110 }, { "epoch": 2.02, "learning_rate": 2.9018975656423127e-05, "loss": 0.2894, "step": 20111 }, { "epoch": 2.02, "learning_rate": 2.9016583770498194e-05, "loss": 0.2559, "step": 20112 }, { "epoch": 2.02, "learning_rate": 2.901419189083135e-05, "loss": 0.4704, "step": 20113 }, { "epoch": 2.02, "learning_rate": 2.9011800017437802e-05, "loss": 0.331, "step": 20114 }, { "epoch": 2.02, "learning_rate": 2.9009408150332792e-05, "loss": 0.4934, "step": 20115 }, { "epoch": 2.02, "learning_rate": 2.9007016289531534e-05, "loss": 0.1055, "step": 20116 }, { "epoch": 2.02, "learning_rate": 2.900462443504923e-05, "loss": 0.2462, "step": 20117 }, { "epoch": 2.02, "learning_rate": 2.9002232586901136e-05, "loss": 0.2412, "step": 20118 }, { "epoch": 2.02, "learning_rate": 2.8999840745102445e-05, "loss": 0.6663, "step": 20119 }, { "epoch": 2.02, "learning_rate": 2.8997448909668383e-05, "loss": 0.2853, "step": 20120 }, { "epoch": 2.02, "learning_rate": 2.899505708061419e-05, "loss": 0.1061, "step": 20121 }, { "epoch": 2.02, "learning_rate": 2.899266525795505e-05, "loss": 0.2182, "step": 20122 }, { "epoch": 2.02, "learning_rate": 2.8990273441706224e-05, "loss": 0.2595, "step": 20123 }, { "epoch": 2.02, "learning_rate": 2.8987881631882908e-05, "loss": 0.5957, "step": 20124 }, { "epoch": 2.02, "learning_rate": 2.898548982850032e-05, "loss": 0.6226, "step": 20125 }, { "epoch": 2.02, "learning_rate": 2.8983098031573693e-05, "loss": 0.3657, "step": 20126 }, { "epoch": 2.02, "learning_rate": 2.8980706241118257e-05, "loss": 0.2966, "step": 20127 }, { "epoch": 2.02, "learning_rate": 2.89783144571492e-05, "loss": 0.3288, "step": 20128 }, { "epoch": 2.02, "learning_rate": 2.8975922679681773e-05, "loss": 0.34, "step": 20129 }, { "epoch": 2.02, "learning_rate": 2.8973530908731185e-05, "loss": 0.3628, "step": 20130 }, { "epoch": 2.02, "learning_rate": 2.897113914431264e-05, "loss": 0.2348, "step": 20131 }, { "epoch": 2.02, "learning_rate": 2.8968747386441392e-05, "loss": 0.5583, "step": 20132 }, { "epoch": 2.02, "learning_rate": 2.8966355635132637e-05, "loss": 0.2997, "step": 20133 }, { "epoch": 2.02, "learning_rate": 2.8963963890401598e-05, "loss": 0.256, "step": 20134 }, { "epoch": 2.02, "learning_rate": 2.89615721522635e-05, "loss": 0.1461, "step": 20135 }, { "epoch": 2.02, "learning_rate": 2.895918042073357e-05, "loss": 1.2817, "step": 20136 }, { "epoch": 2.02, "learning_rate": 2.895678869582701e-05, "loss": 0.4504, "step": 20137 }, { "epoch": 2.02, "learning_rate": 2.895439697755905e-05, "loss": 0.756, "step": 20138 }, { "epoch": 2.02, "learning_rate": 2.8952005265944924e-05, "loss": 0.2916, "step": 20139 }, { "epoch": 2.02, "learning_rate": 2.8949613560999815e-05, "loss": 0.3903, "step": 20140 }, { "epoch": 2.02, "learning_rate": 2.8947221862738985e-05, "loss": 0.3512, "step": 20141 }, { "epoch": 2.02, "learning_rate": 2.8944830171177627e-05, "loss": 0.0536, "step": 20142 }, { "epoch": 2.02, "learning_rate": 2.8942438486330963e-05, "loss": 0.6072, "step": 20143 }, { "epoch": 2.02, "learning_rate": 2.8940046808214223e-05, "loss": 0.8617, "step": 20144 }, { "epoch": 2.02, "learning_rate": 2.893765513684263e-05, "loss": 0.3689, "step": 20145 }, { "epoch": 2.02, "learning_rate": 2.8935263472231384e-05, "loss": 0.5315, "step": 20146 }, { "epoch": 2.02, "learning_rate": 2.893287181439572e-05, "loss": 0.6601, "step": 20147 }, { "epoch": 2.02, "learning_rate": 2.8930480163350848e-05, "loss": 0.0628, "step": 20148 }, { "epoch": 2.02, "learning_rate": 2.8928088519112e-05, "loss": 0.5707, "step": 20149 }, { "epoch": 2.02, "learning_rate": 2.89256968816944e-05, "loss": 0.8864, "step": 20150 }, { "epoch": 2.02, "learning_rate": 2.8923305251113236e-05, "loss": 0.4764, "step": 20151 }, { "epoch": 2.02, "learning_rate": 2.8920913627383757e-05, "loss": 0.371, "step": 20152 }, { "epoch": 2.02, "learning_rate": 2.8918522010521184e-05, "loss": 0.5312, "step": 20153 }, { "epoch": 2.02, "learning_rate": 2.89161304005407e-05, "loss": 1.2085, "step": 20154 }, { "epoch": 2.02, "learning_rate": 2.891373879745757e-05, "loss": 0.5606, "step": 20155 }, { "epoch": 2.02, "learning_rate": 2.891134720128699e-05, "loss": 0.1806, "step": 20156 }, { "epoch": 2.02, "learning_rate": 2.8908955612044172e-05, "loss": 0.1379, "step": 20157 }, { "epoch": 2.02, "learning_rate": 2.8906564029744355e-05, "loss": 0.3043, "step": 20158 }, { "epoch": 2.02, "learning_rate": 2.8904172454402754e-05, "loss": 0.5285, "step": 20159 }, { "epoch": 2.02, "learning_rate": 2.8901780886034566e-05, "loss": 0.0842, "step": 20160 }, { "epoch": 2.02, "learning_rate": 2.8899389324655036e-05, "loss": 0.2501, "step": 20161 }, { "epoch": 2.02, "learning_rate": 2.8896997770279385e-05, "loss": 0.3945, "step": 20162 }, { "epoch": 2.02, "learning_rate": 2.8894606222922794e-05, "loss": 0.7193, "step": 20163 }, { "epoch": 2.02, "learning_rate": 2.8892214682600534e-05, "loss": 0.5778, "step": 20164 }, { "epoch": 2.02, "learning_rate": 2.8889823149327787e-05, "loss": 0.4481, "step": 20165 }, { "epoch": 2.02, "learning_rate": 2.8887431623119776e-05, "loss": 0.2295, "step": 20166 }, { "epoch": 2.02, "learning_rate": 2.8885040103991737e-05, "loss": 0.4898, "step": 20167 }, { "epoch": 2.02, "learning_rate": 2.8882648591958884e-05, "loss": 0.2204, "step": 20168 }, { "epoch": 2.02, "learning_rate": 2.8880257087036417e-05, "loss": 0.1899, "step": 20169 }, { "epoch": 2.02, "learning_rate": 2.8877865589239572e-05, "loss": 0.3181, "step": 20170 }, { "epoch": 2.02, "learning_rate": 2.8875474098583577e-05, "loss": 0.7321, "step": 20171 }, { "epoch": 2.02, "learning_rate": 2.887308261508361e-05, "loss": 0.5805, "step": 20172 }, { "epoch": 2.02, "learning_rate": 2.887069113875494e-05, "loss": 0.4879, "step": 20173 }, { "epoch": 2.02, "learning_rate": 2.8868299669612752e-05, "loss": 0.0916, "step": 20174 }, { "epoch": 2.02, "learning_rate": 2.8865908207672275e-05, "loss": 0.6262, "step": 20175 }, { "epoch": 2.02, "learning_rate": 2.8863516752948728e-05, "loss": 0.4019, "step": 20176 }, { "epoch": 2.02, "learning_rate": 2.886112530545732e-05, "loss": 0.183, "step": 20177 }, { "epoch": 2.02, "learning_rate": 2.8858733865213292e-05, "loss": 0.1162, "step": 20178 }, { "epoch": 2.02, "learning_rate": 2.8856342432231845e-05, "loss": 0.7525, "step": 20179 }, { "epoch": 2.02, "learning_rate": 2.885395100652819e-05, "loss": 0.166, "step": 20180 }, { "epoch": 2.02, "learning_rate": 2.885155958811756e-05, "loss": 0.2028, "step": 20181 }, { "epoch": 2.02, "learning_rate": 2.8849168177015175e-05, "loss": 0.2184, "step": 20182 }, { "epoch": 2.02, "learning_rate": 2.8846776773236232e-05, "loss": 0.0557, "step": 20183 }, { "epoch": 2.02, "learning_rate": 2.884438537679597e-05, "loss": 0.4873, "step": 20184 }, { "epoch": 2.02, "learning_rate": 2.884199398770961e-05, "loss": 0.3958, "step": 20185 }, { "epoch": 2.02, "learning_rate": 2.8839602605992334e-05, "loss": 0.1466, "step": 20186 }, { "epoch": 2.02, "learning_rate": 2.883721123165941e-05, "loss": 0.5493, "step": 20187 }, { "epoch": 2.02, "learning_rate": 2.883481986472602e-05, "loss": 0.1131, "step": 20188 }, { "epoch": 2.02, "learning_rate": 2.8832428505207392e-05, "loss": 0.1103, "step": 20189 }, { "epoch": 2.02, "learning_rate": 2.883003715311875e-05, "loss": 0.6855, "step": 20190 }, { "epoch": 2.02, "learning_rate": 2.8827645808475314e-05, "loss": 0.4192, "step": 20191 }, { "epoch": 2.02, "learning_rate": 2.8825254471292277e-05, "loss": 0.6248, "step": 20192 }, { "epoch": 2.02, "learning_rate": 2.8822863141584885e-05, "loss": 0.2863, "step": 20193 }, { "epoch": 2.02, "learning_rate": 2.8820471819368347e-05, "loss": 0.5779, "step": 20194 }, { "epoch": 2.03, "learning_rate": 2.881808050465786e-05, "loss": 0.1827, "step": 20195 }, { "epoch": 2.03, "learning_rate": 2.8815689197468675e-05, "loss": 0.4498, "step": 20196 }, { "epoch": 2.03, "learning_rate": 2.881329789781599e-05, "loss": 0.0813, "step": 20197 }, { "epoch": 2.03, "learning_rate": 2.881090660571502e-05, "loss": 0.7387, "step": 20198 }, { "epoch": 2.03, "learning_rate": 2.8808515321180994e-05, "loss": 0.3728, "step": 20199 }, { "epoch": 2.03, "learning_rate": 2.8806124044229127e-05, "loss": 0.8651, "step": 20200 }, { "epoch": 2.03, "learning_rate": 2.880373277487462e-05, "loss": 0.392, "step": 20201 }, { "epoch": 2.03, "learning_rate": 2.8801341513132712e-05, "loss": 0.4937, "step": 20202 }, { "epoch": 2.03, "learning_rate": 2.8798950259018603e-05, "loss": 0.203, "step": 20203 }, { "epoch": 2.03, "learning_rate": 2.8796559012547525e-05, "loss": 0.3092, "step": 20204 }, { "epoch": 2.03, "learning_rate": 2.8794167773734693e-05, "loss": 0.2419, "step": 20205 }, { "epoch": 2.03, "learning_rate": 2.8791776542595305e-05, "loss": 0.0341, "step": 20206 }, { "epoch": 2.03, "learning_rate": 2.8789385319144598e-05, "loss": 0.0694, "step": 20207 }, { "epoch": 2.03, "learning_rate": 2.8786994103397784e-05, "loss": 0.529, "step": 20208 }, { "epoch": 2.03, "learning_rate": 2.878460289537007e-05, "loss": 0.2054, "step": 20209 }, { "epoch": 2.03, "learning_rate": 2.8782211695076694e-05, "loss": 0.5887, "step": 20210 }, { "epoch": 2.03, "learning_rate": 2.8779820502532853e-05, "loss": 0.3574, "step": 20211 }, { "epoch": 2.03, "learning_rate": 2.8777429317753764e-05, "loss": 0.1926, "step": 20212 }, { "epoch": 2.03, "learning_rate": 2.877503814075466e-05, "loss": 0.3333, "step": 20213 }, { "epoch": 2.03, "learning_rate": 2.8772646971550747e-05, "loss": 0.1125, "step": 20214 }, { "epoch": 2.03, "learning_rate": 2.877025581015723e-05, "loss": 0.4398, "step": 20215 }, { "epoch": 2.03, "learning_rate": 2.8767864656589347e-05, "loss": 0.2095, "step": 20216 }, { "epoch": 2.03, "learning_rate": 2.876547351086231e-05, "loss": 0.3178, "step": 20217 }, { "epoch": 2.03, "learning_rate": 2.8763082372991307e-05, "loss": 0.2732, "step": 20218 }, { "epoch": 2.03, "learning_rate": 2.87606912429916e-05, "loss": 0.7156, "step": 20219 }, { "epoch": 2.03, "learning_rate": 2.8758300120878377e-05, "loss": 0.2388, "step": 20220 }, { "epoch": 2.03, "learning_rate": 2.875590900666685e-05, "loss": 0.9154, "step": 20221 }, { "epoch": 2.03, "learning_rate": 2.8753517900372253e-05, "loss": 0.2049, "step": 20222 }, { "epoch": 2.03, "learning_rate": 2.87511268020098e-05, "loss": 0.1993, "step": 20223 }, { "epoch": 2.03, "learning_rate": 2.874873571159469e-05, "loss": 0.1343, "step": 20224 }, { "epoch": 2.03, "learning_rate": 2.8746344629142155e-05, "loss": 0.8616, "step": 20225 }, { "epoch": 2.03, "learning_rate": 2.8743953554667412e-05, "loss": 0.5276, "step": 20226 }, { "epoch": 2.03, "learning_rate": 2.8741562488185652e-05, "loss": 0.1935, "step": 20227 }, { "epoch": 2.03, "learning_rate": 2.8739171429712128e-05, "loss": 0.3169, "step": 20228 }, { "epoch": 2.03, "learning_rate": 2.8736780379262032e-05, "loss": 0.3163, "step": 20229 }, { "epoch": 2.03, "learning_rate": 2.8734389336850577e-05, "loss": 0.1718, "step": 20230 }, { "epoch": 2.03, "learning_rate": 2.8731998302492996e-05, "loss": 0.2174, "step": 20231 }, { "epoch": 2.03, "learning_rate": 2.8729607276204487e-05, "loss": 0.2779, "step": 20232 }, { "epoch": 2.03, "learning_rate": 2.8727216258000285e-05, "loss": 0.4605, "step": 20233 }, { "epoch": 2.03, "learning_rate": 2.872482524789559e-05, "loss": 0.2953, "step": 20234 }, { "epoch": 2.03, "learning_rate": 2.8722434245905616e-05, "loss": 0.6608, "step": 20235 }, { "epoch": 2.03, "learning_rate": 2.872004325204559e-05, "loss": 0.2373, "step": 20236 }, { "epoch": 2.03, "learning_rate": 2.871765226633073e-05, "loss": 0.2908, "step": 20237 }, { "epoch": 2.03, "learning_rate": 2.871526128877623e-05, "loss": 0.4826, "step": 20238 }, { "epoch": 2.03, "learning_rate": 2.8712870319397322e-05, "loss": 0.2438, "step": 20239 }, { "epoch": 2.03, "learning_rate": 2.8710479358209222e-05, "loss": 0.7198, "step": 20240 }, { "epoch": 2.03, "learning_rate": 2.8708088405227133e-05, "loss": 0.3062, "step": 20241 }, { "epoch": 2.03, "learning_rate": 2.8705697460466287e-05, "loss": 0.1648, "step": 20242 }, { "epoch": 2.03, "learning_rate": 2.870330652394189e-05, "loss": 0.5152, "step": 20243 }, { "epoch": 2.03, "learning_rate": 2.8700915595669147e-05, "loss": 0.3968, "step": 20244 }, { "epoch": 2.03, "learning_rate": 2.8698524675663287e-05, "loss": 1.115, "step": 20245 }, { "epoch": 2.03, "learning_rate": 2.8696133763939528e-05, "loss": 0.123, "step": 20246 }, { "epoch": 2.03, "learning_rate": 2.8693742860513068e-05, "loss": 0.2052, "step": 20247 }, { "epoch": 2.03, "learning_rate": 2.8691351965399138e-05, "loss": 0.4847, "step": 20248 }, { "epoch": 2.03, "learning_rate": 2.868896107861295e-05, "loss": 0.4704, "step": 20249 }, { "epoch": 2.03, "learning_rate": 2.8686570200169697e-05, "loss": 0.5991, "step": 20250 }, { "epoch": 2.03, "learning_rate": 2.868417933008463e-05, "loss": 0.4621, "step": 20251 }, { "epoch": 2.03, "learning_rate": 2.868178846837294e-05, "loss": 0.2652, "step": 20252 }, { "epoch": 2.03, "learning_rate": 2.867939761504984e-05, "loss": 0.2476, "step": 20253 }, { "epoch": 2.03, "learning_rate": 2.8677006770130558e-05, "loss": 0.1019, "step": 20254 }, { "epoch": 2.03, "learning_rate": 2.8674615933630305e-05, "loss": 0.9286, "step": 20255 }, { "epoch": 2.03, "learning_rate": 2.867222510556428e-05, "loss": 0.1555, "step": 20256 }, { "epoch": 2.03, "learning_rate": 2.866983428594772e-05, "loss": 0.4854, "step": 20257 }, { "epoch": 2.03, "learning_rate": 2.866744347479582e-05, "loss": 0.2329, "step": 20258 }, { "epoch": 2.03, "learning_rate": 2.866505267212381e-05, "loss": 0.3633, "step": 20259 }, { "epoch": 2.03, "learning_rate": 2.86626618779469e-05, "loss": 0.3388, "step": 20260 }, { "epoch": 2.03, "learning_rate": 2.8660271092280294e-05, "loss": 0.5505, "step": 20261 }, { "epoch": 2.03, "learning_rate": 2.8657880315139212e-05, "loss": 0.1156, "step": 20262 }, { "epoch": 2.03, "learning_rate": 2.8655489546538875e-05, "loss": 0.1855, "step": 20263 }, { "epoch": 2.03, "learning_rate": 2.865309878649448e-05, "loss": 0.1812, "step": 20264 }, { "epoch": 2.03, "learning_rate": 2.865070803502127e-05, "loss": 0.1862, "step": 20265 }, { "epoch": 2.03, "learning_rate": 2.8648317292134428e-05, "loss": 0.1736, "step": 20266 }, { "epoch": 2.03, "learning_rate": 2.864592655784918e-05, "loss": 0.358, "step": 20267 }, { "epoch": 2.03, "learning_rate": 2.8643535832180746e-05, "loss": 0.5333, "step": 20268 }, { "epoch": 2.03, "learning_rate": 2.864114511514434e-05, "loss": 0.0776, "step": 20269 }, { "epoch": 2.03, "learning_rate": 2.8638754406755154e-05, "loss": 0.2179, "step": 20270 }, { "epoch": 2.03, "learning_rate": 2.8636363707028424e-05, "loss": 0.2195, "step": 20271 }, { "epoch": 2.03, "learning_rate": 2.863397301597936e-05, "loss": 0.8692, "step": 20272 }, { "epoch": 2.03, "learning_rate": 2.863158233362316e-05, "loss": 0.3215, "step": 20273 }, { "epoch": 2.03, "learning_rate": 2.8629191659975068e-05, "loss": 0.4044, "step": 20274 }, { "epoch": 2.03, "learning_rate": 2.862680099505027e-05, "loss": 0.3905, "step": 20275 }, { "epoch": 2.03, "learning_rate": 2.862441033886398e-05, "loss": 0.452, "step": 20276 }, { "epoch": 2.03, "learning_rate": 2.862201969143143e-05, "loss": 0.132, "step": 20277 }, { "epoch": 2.03, "learning_rate": 2.8619629052767824e-05, "loss": 0.0879, "step": 20278 }, { "epoch": 2.03, "learning_rate": 2.8617238422888366e-05, "loss": 0.4287, "step": 20279 }, { "epoch": 2.03, "learning_rate": 2.8614847801808277e-05, "loss": 0.2796, "step": 20280 }, { "epoch": 2.03, "learning_rate": 2.8612457189542782e-05, "loss": 0.0938, "step": 20281 }, { "epoch": 2.03, "learning_rate": 2.861006658610706e-05, "loss": 0.0965, "step": 20282 }, { "epoch": 2.03, "learning_rate": 2.860767599151636e-05, "loss": 0.69, "step": 20283 }, { "epoch": 2.03, "learning_rate": 2.8605285405785873e-05, "loss": 0.246, "step": 20284 }, { "epoch": 2.03, "learning_rate": 2.8602894828930824e-05, "loss": 0.3652, "step": 20285 }, { "epoch": 2.03, "learning_rate": 2.8600504260966422e-05, "loss": 0.0392, "step": 20286 }, { "epoch": 2.03, "learning_rate": 2.859811370190787e-05, "loss": 0.2343, "step": 20287 }, { "epoch": 2.03, "learning_rate": 2.8595723151770406e-05, "loss": 0.1845, "step": 20288 }, { "epoch": 2.03, "learning_rate": 2.8593332610569216e-05, "loss": 0.1825, "step": 20289 }, { "epoch": 2.03, "learning_rate": 2.8590942078319515e-05, "loss": 0.2812, "step": 20290 }, { "epoch": 2.03, "learning_rate": 2.8588551555036528e-05, "loss": 0.1903, "step": 20291 }, { "epoch": 2.03, "learning_rate": 2.8586161040735472e-05, "loss": 0.4411, "step": 20292 }, { "epoch": 2.04, "learning_rate": 2.8583770535431532e-05, "loss": 0.3095, "step": 20293 }, { "epoch": 2.04, "learning_rate": 2.8581380039139948e-05, "loss": 0.1299, "step": 20294 }, { "epoch": 2.04, "learning_rate": 2.8578989551875923e-05, "loss": 0.1719, "step": 20295 }, { "epoch": 2.04, "learning_rate": 2.857659907365466e-05, "loss": 0.4503, "step": 20296 }, { "epoch": 2.04, "learning_rate": 2.857420860449139e-05, "loss": 0.1028, "step": 20297 }, { "epoch": 2.04, "learning_rate": 2.857181814440131e-05, "loss": 0.4329, "step": 20298 }, { "epoch": 2.04, "learning_rate": 2.8569427693399628e-05, "loss": 0.2313, "step": 20299 }, { "epoch": 2.04, "learning_rate": 2.856703725150157e-05, "loss": 0.1111, "step": 20300 }, { "epoch": 2.04, "learning_rate": 2.8564646818722348e-05, "loss": 0.0356, "step": 20301 }, { "epoch": 2.04, "learning_rate": 2.8562256395077155e-05, "loss": 0.6556, "step": 20302 }, { "epoch": 2.04, "learning_rate": 2.8559865980581222e-05, "loss": 0.2925, "step": 20303 }, { "epoch": 2.04, "learning_rate": 2.8557475575249755e-05, "loss": 0.138, "step": 20304 }, { "epoch": 2.04, "learning_rate": 2.8555085179097957e-05, "loss": 0.3279, "step": 20305 }, { "epoch": 2.04, "learning_rate": 2.855269479214106e-05, "loss": 0.2874, "step": 20306 }, { "epoch": 2.04, "learning_rate": 2.855030441439426e-05, "loss": 0.5334, "step": 20307 }, { "epoch": 2.04, "learning_rate": 2.854791404587276e-05, "loss": 0.7397, "step": 20308 }, { "epoch": 2.04, "learning_rate": 2.854552368659179e-05, "loss": 0.2888, "step": 20309 }, { "epoch": 2.04, "learning_rate": 2.854313333656656e-05, "loss": 0.295, "step": 20310 }, { "epoch": 2.04, "learning_rate": 2.8540742995812266e-05, "loss": 0.3538, "step": 20311 }, { "epoch": 2.04, "learning_rate": 2.8538352664344132e-05, "loss": 0.8047, "step": 20312 }, { "epoch": 2.04, "learning_rate": 2.853596234217736e-05, "loss": 0.3304, "step": 20313 }, { "epoch": 2.04, "learning_rate": 2.853357202932717e-05, "loss": 0.8312, "step": 20314 }, { "epoch": 2.04, "learning_rate": 2.853118172580878e-05, "loss": 0.5027, "step": 20315 }, { "epoch": 2.04, "learning_rate": 2.8528791431637377e-05, "loss": 0.1117, "step": 20316 }, { "epoch": 2.04, "learning_rate": 2.8526401146828192e-05, "loss": 0.1609, "step": 20317 }, { "epoch": 2.04, "learning_rate": 2.852401087139643e-05, "loss": 0.2568, "step": 20318 }, { "epoch": 2.04, "learning_rate": 2.8521620605357296e-05, "loss": 0.1396, "step": 20319 }, { "epoch": 2.04, "learning_rate": 2.8519230348726017e-05, "loss": 0.142, "step": 20320 }, { "epoch": 2.04, "learning_rate": 2.8516840101517787e-05, "loss": 0.3358, "step": 20321 }, { "epoch": 2.04, "learning_rate": 2.8514449863747814e-05, "loss": 0.1837, "step": 20322 }, { "epoch": 2.04, "learning_rate": 2.8512059635431326e-05, "loss": 0.3718, "step": 20323 }, { "epoch": 2.04, "learning_rate": 2.850966941658353e-05, "loss": 0.0784, "step": 20324 }, { "epoch": 2.04, "learning_rate": 2.8507279207219617e-05, "loss": 0.4806, "step": 20325 }, { "epoch": 2.04, "learning_rate": 2.8504889007354816e-05, "loss": 0.4513, "step": 20326 }, { "epoch": 2.04, "learning_rate": 2.8502498817004336e-05, "loss": 0.5922, "step": 20327 }, { "epoch": 2.04, "learning_rate": 2.8500108636183377e-05, "loss": 0.5829, "step": 20328 }, { "epoch": 2.04, "learning_rate": 2.8497718464907168e-05, "loss": 0.3146, "step": 20329 }, { "epoch": 2.04, "learning_rate": 2.84953283031909e-05, "loss": 0.2789, "step": 20330 }, { "epoch": 2.04, "learning_rate": 2.8492938151049786e-05, "loss": 0.4438, "step": 20331 }, { "epoch": 2.04, "learning_rate": 2.8490548008499046e-05, "loss": 0.2466, "step": 20332 }, { "epoch": 2.04, "learning_rate": 2.848815787555389e-05, "loss": 0.1914, "step": 20333 }, { "epoch": 2.04, "learning_rate": 2.848576775222951e-05, "loss": 0.1167, "step": 20334 }, { "epoch": 2.04, "learning_rate": 2.8483377638541137e-05, "loss": 0.2328, "step": 20335 }, { "epoch": 2.04, "learning_rate": 2.8480987534503965e-05, "loss": 0.4268, "step": 20336 }, { "epoch": 2.04, "learning_rate": 2.847859744013321e-05, "loss": 0.0771, "step": 20337 }, { "epoch": 2.04, "learning_rate": 2.847620735544409e-05, "loss": 1.0514, "step": 20338 }, { "epoch": 2.04, "learning_rate": 2.8473817280451797e-05, "loss": 0.238, "step": 20339 }, { "epoch": 2.04, "learning_rate": 2.847142721517156e-05, "loss": 0.1981, "step": 20340 }, { "epoch": 2.04, "learning_rate": 2.846903715961857e-05, "loss": 0.4263, "step": 20341 }, { "epoch": 2.04, "learning_rate": 2.8466647113808046e-05, "loss": 0.3512, "step": 20342 }, { "epoch": 2.04, "learning_rate": 2.8464257077755206e-05, "loss": 0.3452, "step": 20343 }, { "epoch": 2.04, "learning_rate": 2.846186705147525e-05, "loss": 0.1263, "step": 20344 }, { "epoch": 2.04, "learning_rate": 2.8459477034983373e-05, "loss": 0.2278, "step": 20345 }, { "epoch": 2.04, "learning_rate": 2.8457087028294804e-05, "loss": 0.9874, "step": 20346 }, { "epoch": 2.04, "learning_rate": 2.8454697031424755e-05, "loss": 0.2213, "step": 20347 }, { "epoch": 2.04, "learning_rate": 2.8452307044388414e-05, "loss": 0.5043, "step": 20348 }, { "epoch": 2.04, "learning_rate": 2.8449917067201005e-05, "loss": 0.2782, "step": 20349 }, { "epoch": 2.04, "learning_rate": 2.844752709987774e-05, "loss": 0.3409, "step": 20350 }, { "epoch": 2.04, "learning_rate": 2.8445137142433812e-05, "loss": 0.1444, "step": 20351 }, { "epoch": 2.04, "learning_rate": 2.8442747194884453e-05, "loss": 0.6242, "step": 20352 }, { "epoch": 2.04, "learning_rate": 2.8440357257244853e-05, "loss": 0.2575, "step": 20353 }, { "epoch": 2.04, "learning_rate": 2.843796732953022e-05, "loss": 0.3792, "step": 20354 }, { "epoch": 2.04, "learning_rate": 2.8435577411755773e-05, "loss": 0.3044, "step": 20355 }, { "epoch": 2.04, "learning_rate": 2.8433187503936724e-05, "loss": 0.353, "step": 20356 }, { "epoch": 2.04, "learning_rate": 2.843079760608826e-05, "loss": 0.1594, "step": 20357 }, { "epoch": 2.04, "learning_rate": 2.842840771822561e-05, "loss": 0.0972, "step": 20358 }, { "epoch": 2.04, "learning_rate": 2.8426017840363977e-05, "loss": 0.9205, "step": 20359 }, { "epoch": 2.04, "learning_rate": 2.842362797251856e-05, "loss": 0.7966, "step": 20360 }, { "epoch": 2.04, "learning_rate": 2.8421238114704584e-05, "loss": 0.386, "step": 20361 }, { "epoch": 2.04, "learning_rate": 2.8418848266937247e-05, "loss": 0.1988, "step": 20362 }, { "epoch": 2.04, "learning_rate": 2.841645842923175e-05, "loss": 0.5704, "step": 20363 }, { "epoch": 2.04, "learning_rate": 2.8414068601603313e-05, "loss": 0.7088, "step": 20364 }, { "epoch": 2.04, "learning_rate": 2.8411678784067138e-05, "loss": 0.7405, "step": 20365 }, { "epoch": 2.04, "learning_rate": 2.8409288976638444e-05, "loss": 0.2263, "step": 20366 }, { "epoch": 2.04, "learning_rate": 2.8406899179332428e-05, "loss": 0.2537, "step": 20367 }, { "epoch": 2.04, "learning_rate": 2.840450939216429e-05, "loss": 0.4005, "step": 20368 }, { "epoch": 2.04, "learning_rate": 2.8402119615149254e-05, "loss": 0.198, "step": 20369 }, { "epoch": 2.04, "learning_rate": 2.839972984830253e-05, "loss": 0.135, "step": 20370 }, { "epoch": 2.04, "learning_rate": 2.8397340091639298e-05, "loss": 0.3149, "step": 20371 }, { "epoch": 2.04, "learning_rate": 2.8394950345174794e-05, "loss": 0.556, "step": 20372 }, { "epoch": 2.04, "learning_rate": 2.8392560608924214e-05, "loss": 0.6762, "step": 20373 }, { "epoch": 2.04, "learning_rate": 2.839017088290276e-05, "loss": 0.2159, "step": 20374 }, { "epoch": 2.04, "learning_rate": 2.838778116712566e-05, "loss": 0.3017, "step": 20375 }, { "epoch": 2.04, "learning_rate": 2.83853914616081e-05, "loss": 0.3479, "step": 20376 }, { "epoch": 2.04, "learning_rate": 2.8383001766365292e-05, "loss": 0.6201, "step": 20377 }, { "epoch": 2.04, "learning_rate": 2.8380612081412446e-05, "loss": 0.2448, "step": 20378 }, { "epoch": 2.04, "learning_rate": 2.837822240676478e-05, "loss": 0.6894, "step": 20379 }, { "epoch": 2.04, "learning_rate": 2.8375832742437478e-05, "loss": 0.294, "step": 20380 }, { "epoch": 2.04, "learning_rate": 2.837344308844576e-05, "loss": 0.2784, "step": 20381 }, { "epoch": 2.04, "learning_rate": 2.8371053444804836e-05, "loss": 0.2199, "step": 20382 }, { "epoch": 2.04, "learning_rate": 2.8368663811529897e-05, "loss": 0.3586, "step": 20383 }, { "epoch": 2.04, "learning_rate": 2.8366274188636177e-05, "loss": 0.8626, "step": 20384 }, { "epoch": 2.04, "learning_rate": 2.836388457613886e-05, "loss": 0.3821, "step": 20385 }, { "epoch": 2.04, "learning_rate": 2.836149497405315e-05, "loss": 0.4049, "step": 20386 }, { "epoch": 2.04, "learning_rate": 2.8359105382394272e-05, "loss": 0.6894, "step": 20387 }, { "epoch": 2.04, "learning_rate": 2.8356715801177426e-05, "loss": 0.525, "step": 20388 }, { "epoch": 2.04, "learning_rate": 2.8354326230417805e-05, "loss": 0.333, "step": 20389 }, { "epoch": 2.04, "learning_rate": 2.8351936670130635e-05, "loss": 0.2694, "step": 20390 }, { "epoch": 2.04, "learning_rate": 2.834954712033111e-05, "loss": 0.4691, "step": 20391 }, { "epoch": 2.05, "learning_rate": 2.834715758103443e-05, "loss": 0.0739, "step": 20392 }, { "epoch": 2.05, "learning_rate": 2.834476805225583e-05, "loss": 0.2714, "step": 20393 }, { "epoch": 2.05, "learning_rate": 2.834237853401048e-05, "loss": 0.5217, "step": 20394 }, { "epoch": 2.05, "learning_rate": 2.8339989026313607e-05, "loss": 0.5189, "step": 20395 }, { "epoch": 2.05, "learning_rate": 2.8337599529180414e-05, "loss": 0.5277, "step": 20396 }, { "epoch": 2.05, "learning_rate": 2.83352100426261e-05, "loss": 0.2825, "step": 20397 }, { "epoch": 2.05, "learning_rate": 2.8332820566665888e-05, "loss": 0.6167, "step": 20398 }, { "epoch": 2.05, "learning_rate": 2.8330431101314963e-05, "loss": 0.5091, "step": 20399 }, { "epoch": 2.05, "learning_rate": 2.8328041646588536e-05, "loss": 0.0865, "step": 20400 }, { "epoch": 2.05, "learning_rate": 2.8325652202501824e-05, "loss": 0.5927, "step": 20401 }, { "epoch": 2.05, "learning_rate": 2.832326276907003e-05, "loss": 0.2599, "step": 20402 }, { "epoch": 2.05, "learning_rate": 2.8320873346308338e-05, "loss": 0.6015, "step": 20403 }, { "epoch": 2.05, "learning_rate": 2.8318483934231977e-05, "loss": 0.4969, "step": 20404 }, { "epoch": 2.05, "learning_rate": 2.8316094532856144e-05, "loss": 0.277, "step": 20405 }, { "epoch": 2.05, "learning_rate": 2.8313705142196035e-05, "loss": 0.367, "step": 20406 }, { "epoch": 2.05, "learning_rate": 2.8311315762266883e-05, "loss": 0.6214, "step": 20407 }, { "epoch": 2.05, "learning_rate": 2.8308926393083868e-05, "loss": 0.574, "step": 20408 }, { "epoch": 2.05, "learning_rate": 2.8306537034662193e-05, "loss": 0.1812, "step": 20409 }, { "epoch": 2.05, "learning_rate": 2.8304147687017084e-05, "loss": 0.5123, "step": 20410 }, { "epoch": 2.05, "learning_rate": 2.8301758350163738e-05, "loss": 0.2302, "step": 20411 }, { "epoch": 2.05, "learning_rate": 2.8299369024117343e-05, "loss": 0.8609, "step": 20412 }, { "epoch": 2.05, "learning_rate": 2.829697970889312e-05, "loss": 0.7267, "step": 20413 }, { "epoch": 2.05, "learning_rate": 2.8294590404506274e-05, "loss": 0.1809, "step": 20414 }, { "epoch": 2.05, "learning_rate": 2.8292201110971995e-05, "loss": 0.2419, "step": 20415 }, { "epoch": 2.05, "learning_rate": 2.8289811828305518e-05, "loss": 0.3939, "step": 20416 }, { "epoch": 2.05, "learning_rate": 2.8287422556522015e-05, "loss": 0.3995, "step": 20417 }, { "epoch": 2.05, "learning_rate": 2.82850332956367e-05, "loss": 0.3345, "step": 20418 }, { "epoch": 2.05, "learning_rate": 2.8282644045664786e-05, "loss": 0.2069, "step": 20419 }, { "epoch": 2.05, "learning_rate": 2.8280254806621465e-05, "loss": 0.9418, "step": 20420 }, { "epoch": 2.05, "learning_rate": 2.827786557852196e-05, "loss": 0.1785, "step": 20421 }, { "epoch": 2.05, "learning_rate": 2.8275476361381462e-05, "loss": 0.1694, "step": 20422 }, { "epoch": 2.05, "learning_rate": 2.8273087155215166e-05, "loss": 0.1716, "step": 20423 }, { "epoch": 2.05, "learning_rate": 2.827069796003829e-05, "loss": 0.1589, "step": 20424 }, { "epoch": 2.05, "learning_rate": 2.826830877586605e-05, "loss": 0.1556, "step": 20425 }, { "epoch": 2.05, "learning_rate": 2.8265919602713614e-05, "loss": 0.2047, "step": 20426 }, { "epoch": 2.05, "learning_rate": 2.8263530440596215e-05, "loss": 0.3721, "step": 20427 }, { "epoch": 2.05, "learning_rate": 2.8261141289529046e-05, "loss": 0.1526, "step": 20428 }, { "epoch": 2.05, "learning_rate": 2.8258752149527305e-05, "loss": 0.1809, "step": 20429 }, { "epoch": 2.05, "learning_rate": 2.8256363020606223e-05, "loss": 0.3376, "step": 20430 }, { "epoch": 2.05, "learning_rate": 2.825397390278097e-05, "loss": 0.0691, "step": 20431 }, { "epoch": 2.05, "learning_rate": 2.825158479606676e-05, "loss": 0.3767, "step": 20432 }, { "epoch": 2.05, "learning_rate": 2.8249195700478802e-05, "loss": 0.3117, "step": 20433 }, { "epoch": 2.05, "learning_rate": 2.8246806616032306e-05, "loss": 0.6267, "step": 20434 }, { "epoch": 2.05, "learning_rate": 2.824441754274245e-05, "loss": 0.4312, "step": 20435 }, { "epoch": 2.05, "learning_rate": 2.824202848062446e-05, "loss": 0.747, "step": 20436 }, { "epoch": 2.05, "learning_rate": 2.8239639429693532e-05, "loss": 0.0992, "step": 20437 }, { "epoch": 2.05, "learning_rate": 2.8237250389964863e-05, "loss": 0.1756, "step": 20438 }, { "epoch": 2.05, "learning_rate": 2.8234861361453675e-05, "loss": 0.8631, "step": 20439 }, { "epoch": 2.05, "learning_rate": 2.8232472344175156e-05, "loss": 0.3914, "step": 20440 }, { "epoch": 2.05, "learning_rate": 2.8230083338144497e-05, "loss": 0.2436, "step": 20441 }, { "epoch": 2.05, "learning_rate": 2.8227694343376926e-05, "loss": 0.521, "step": 20442 }, { "epoch": 2.05, "learning_rate": 2.8225305359887636e-05, "loss": 0.3794, "step": 20443 }, { "epoch": 2.05, "learning_rate": 2.8222916387691816e-05, "loss": 0.5699, "step": 20444 }, { "epoch": 2.05, "learning_rate": 2.822052742680469e-05, "loss": 0.6655, "step": 20445 }, { "epoch": 2.05, "learning_rate": 2.8218138477241436e-05, "loss": 0.5334, "step": 20446 }, { "epoch": 2.05, "learning_rate": 2.8215749539017282e-05, "loss": 0.3122, "step": 20447 }, { "epoch": 2.05, "learning_rate": 2.8213360612147432e-05, "loss": 0.1586, "step": 20448 }, { "epoch": 2.05, "learning_rate": 2.8210971696647054e-05, "loss": 0.5684, "step": 20449 }, { "epoch": 2.05, "learning_rate": 2.8208582792531382e-05, "loss": 0.1404, "step": 20450 }, { "epoch": 2.05, "learning_rate": 2.8206193899815604e-05, "loss": 0.967, "step": 20451 }, { "epoch": 2.05, "learning_rate": 2.820380501851492e-05, "loss": 0.6881, "step": 20452 }, { "epoch": 2.05, "learning_rate": 2.8201416148644553e-05, "loss": 0.2793, "step": 20453 }, { "epoch": 2.05, "learning_rate": 2.8199027290219684e-05, "loss": 0.8437, "step": 20454 }, { "epoch": 2.05, "learning_rate": 2.8196638443255514e-05, "loss": 0.1867, "step": 20455 }, { "epoch": 2.05, "learning_rate": 2.819424960776725e-05, "loss": 0.5668, "step": 20456 }, { "epoch": 2.05, "learning_rate": 2.819186078377011e-05, "loss": 0.4708, "step": 20457 }, { "epoch": 2.05, "learning_rate": 2.8189471971279263e-05, "loss": 0.0821, "step": 20458 }, { "epoch": 2.05, "learning_rate": 2.8187083170309933e-05, "loss": 0.2988, "step": 20459 }, { "epoch": 2.05, "learning_rate": 2.8184694380877318e-05, "loss": 0.5734, "step": 20460 }, { "epoch": 2.05, "learning_rate": 2.818230560299661e-05, "loss": 0.2555, "step": 20461 }, { "epoch": 2.05, "learning_rate": 2.817991683668303e-05, "loss": 0.707, "step": 20462 }, { "epoch": 2.05, "learning_rate": 2.817752808195176e-05, "loss": 0.676, "step": 20463 }, { "epoch": 2.05, "learning_rate": 2.8175139338818e-05, "loss": 0.2047, "step": 20464 }, { "epoch": 2.05, "learning_rate": 2.8172750607296974e-05, "loss": 0.594, "step": 20465 }, { "epoch": 2.05, "learning_rate": 2.8170361887403866e-05, "loss": 0.3028, "step": 20466 }, { "epoch": 2.05, "learning_rate": 2.816797317915387e-05, "loss": 0.2383, "step": 20467 }, { "epoch": 2.05, "learning_rate": 2.8165584482562198e-05, "loss": 0.665, "step": 20468 }, { "epoch": 2.05, "learning_rate": 2.8163195797644053e-05, "loss": 0.1402, "step": 20469 }, { "epoch": 2.05, "learning_rate": 2.8160807124414624e-05, "loss": 0.3655, "step": 20470 }, { "epoch": 2.05, "learning_rate": 2.815841846288913e-05, "loss": 0.5641, "step": 20471 }, { "epoch": 2.05, "learning_rate": 2.8156029813082758e-05, "loss": 0.3571, "step": 20472 }, { "epoch": 2.05, "learning_rate": 2.8153641175010698e-05, "loss": 0.298, "step": 20473 }, { "epoch": 2.05, "learning_rate": 2.8151252548688175e-05, "loss": 0.3689, "step": 20474 }, { "epoch": 2.05, "learning_rate": 2.8148863934130367e-05, "loss": 0.4232, "step": 20475 }, { "epoch": 2.05, "learning_rate": 2.81464753313525e-05, "loss": 0.7484, "step": 20476 }, { "epoch": 2.05, "learning_rate": 2.8144086740369753e-05, "loss": 0.8431, "step": 20477 }, { "epoch": 2.05, "learning_rate": 2.8141698161197322e-05, "loss": 0.1156, "step": 20478 }, { "epoch": 2.05, "learning_rate": 2.8139309593850427e-05, "loss": 0.1221, "step": 20479 }, { "epoch": 2.05, "learning_rate": 2.8136921038344262e-05, "loss": 1.2046, "step": 20480 }, { "epoch": 2.05, "learning_rate": 2.813453249469401e-05, "loss": 0.1549, "step": 20481 }, { "epoch": 2.05, "learning_rate": 2.8132143962914893e-05, "loss": 0.9831, "step": 20482 }, { "epoch": 2.05, "learning_rate": 2.8129755443022097e-05, "loss": 0.6939, "step": 20483 }, { "epoch": 2.05, "learning_rate": 2.8127366935030822e-05, "loss": 0.2288, "step": 20484 }, { "epoch": 2.05, "learning_rate": 2.8124978438956284e-05, "loss": 0.2735, "step": 20485 }, { "epoch": 2.05, "learning_rate": 2.8122589954813662e-05, "loss": 0.1589, "step": 20486 }, { "epoch": 2.05, "learning_rate": 2.8120201482618162e-05, "loss": 0.4675, "step": 20487 }, { "epoch": 2.05, "learning_rate": 2.8117813022384983e-05, "loss": 0.0561, "step": 20488 }, { "epoch": 2.05, "learning_rate": 2.811542457412934e-05, "loss": 0.301, "step": 20489 }, { "epoch": 2.06, "learning_rate": 2.81130361378664e-05, "loss": 0.7217, "step": 20490 }, { "epoch": 2.06, "learning_rate": 2.8110647713611387e-05, "loss": 0.0361, "step": 20491 }, { "epoch": 2.06, "learning_rate": 2.8108259301379497e-05, "loss": 0.1738, "step": 20492 }, { "epoch": 2.06, "learning_rate": 2.8105870901185918e-05, "loss": 0.9028, "step": 20493 }, { "epoch": 2.06, "learning_rate": 2.8103482513045864e-05, "loss": 0.1449, "step": 20494 }, { "epoch": 2.06, "learning_rate": 2.8101094136974526e-05, "loss": 0.35, "step": 20495 }, { "epoch": 2.06, "learning_rate": 2.809870577298709e-05, "loss": 0.3813, "step": 20496 }, { "epoch": 2.06, "learning_rate": 2.8096317421098778e-05, "loss": 0.5604, "step": 20497 }, { "epoch": 2.06, "learning_rate": 2.8093929081324783e-05, "loss": 0.127, "step": 20498 }, { "epoch": 2.06, "learning_rate": 2.8091540753680284e-05, "loss": 0.1873, "step": 20499 }, { "epoch": 2.06, "learning_rate": 2.8089152438180502e-05, "loss": 0.2375, "step": 20500 }, { "epoch": 2.06, "learning_rate": 2.8086764134840617e-05, "loss": 0.0725, "step": 20501 }, { "epoch": 2.06, "learning_rate": 2.8084375843675847e-05, "loss": 0.6741, "step": 20502 }, { "epoch": 2.06, "learning_rate": 2.8081987564701386e-05, "loss": 0.311, "step": 20503 }, { "epoch": 2.06, "learning_rate": 2.8079599297932416e-05, "loss": 0.7103, "step": 20504 }, { "epoch": 2.06, "learning_rate": 2.8077211043384146e-05, "loss": 0.1122, "step": 20505 }, { "epoch": 2.06, "learning_rate": 2.807482280107178e-05, "loss": 0.4055, "step": 20506 }, { "epoch": 2.06, "learning_rate": 2.80724345710105e-05, "loss": 0.4076, "step": 20507 }, { "epoch": 2.06, "learning_rate": 2.8070046353215527e-05, "loss": 0.183, "step": 20508 }, { "epoch": 2.06, "learning_rate": 2.8067658147702035e-05, "loss": 0.5569, "step": 20509 }, { "epoch": 2.06, "learning_rate": 2.8065269954485225e-05, "loss": 0.2037, "step": 20510 }, { "epoch": 2.06, "learning_rate": 2.8062881773580313e-05, "loss": 0.3821, "step": 20511 }, { "epoch": 2.06, "learning_rate": 2.8060493605002486e-05, "loss": 0.2734, "step": 20512 }, { "epoch": 2.06, "learning_rate": 2.805810544876693e-05, "loss": 0.1556, "step": 20513 }, { "epoch": 2.06, "learning_rate": 2.805571730488886e-05, "loss": 0.3049, "step": 20514 }, { "epoch": 2.06, "learning_rate": 2.805332917338346e-05, "loss": 0.4035, "step": 20515 }, { "epoch": 2.06, "learning_rate": 2.805094105426593e-05, "loss": 0.4925, "step": 20516 }, { "epoch": 2.06, "learning_rate": 2.804855294755148e-05, "loss": 0.4706, "step": 20517 }, { "epoch": 2.06, "learning_rate": 2.8046164853255293e-05, "loss": 0.4952, "step": 20518 }, { "epoch": 2.06, "learning_rate": 2.804377677139256e-05, "loss": 0.1903, "step": 20519 }, { "epoch": 2.06, "learning_rate": 2.8041388701978493e-05, "loss": 0.3335, "step": 20520 }, { "epoch": 2.06, "learning_rate": 2.8039000645028295e-05, "loss": 0.3949, "step": 20521 }, { "epoch": 2.06, "learning_rate": 2.8036612600557135e-05, "loss": 0.1425, "step": 20522 }, { "epoch": 2.06, "learning_rate": 2.8034224568580233e-05, "loss": 0.4326, "step": 20523 }, { "epoch": 2.06, "learning_rate": 2.8031836549112778e-05, "loss": 0.2274, "step": 20524 }, { "epoch": 2.06, "learning_rate": 2.8029448542169958e-05, "loss": 0.3604, "step": 20525 }, { "epoch": 2.06, "learning_rate": 2.8027060547766993e-05, "loss": 0.7782, "step": 20526 }, { "epoch": 2.06, "learning_rate": 2.8024672565919056e-05, "loss": 0.177, "step": 20527 }, { "epoch": 2.06, "learning_rate": 2.8022284596641345e-05, "loss": 0.0991, "step": 20528 }, { "epoch": 2.06, "learning_rate": 2.801989663994907e-05, "loss": 0.8219, "step": 20529 }, { "epoch": 2.06, "learning_rate": 2.8017508695857413e-05, "loss": 0.1825, "step": 20530 }, { "epoch": 2.06, "learning_rate": 2.8015120764381587e-05, "loss": 0.1251, "step": 20531 }, { "epoch": 2.06, "learning_rate": 2.8012732845536773e-05, "loss": 0.5733, "step": 20532 }, { "epoch": 2.06, "learning_rate": 2.8010344939338165e-05, "loss": 0.293, "step": 20533 }, { "epoch": 2.06, "learning_rate": 2.800795704580097e-05, "loss": 0.3328, "step": 20534 }, { "epoch": 2.06, "learning_rate": 2.8005569164940388e-05, "loss": 0.4017, "step": 20535 }, { "epoch": 2.06, "learning_rate": 2.800318129677159e-05, "loss": 0.3902, "step": 20536 }, { "epoch": 2.06, "learning_rate": 2.8000793441309792e-05, "loss": 0.4001, "step": 20537 }, { "epoch": 2.06, "learning_rate": 2.7998405598570188e-05, "loss": 1.0787, "step": 20538 }, { "epoch": 2.06, "learning_rate": 2.799601776856796e-05, "loss": 1.4208, "step": 20539 }, { "epoch": 2.06, "learning_rate": 2.7993629951318324e-05, "loss": 0.1931, "step": 20540 }, { "epoch": 2.06, "learning_rate": 2.7991242146836457e-05, "loss": 0.3525, "step": 20541 }, { "epoch": 2.06, "learning_rate": 2.798885435513756e-05, "loss": 0.7337, "step": 20542 }, { "epoch": 2.06, "learning_rate": 2.798646657623683e-05, "loss": 0.4197, "step": 20543 }, { "epoch": 2.06, "learning_rate": 2.798407881014947e-05, "loss": 0.416, "step": 20544 }, { "epoch": 2.06, "learning_rate": 2.7981691056890648e-05, "loss": 0.5602, "step": 20545 }, { "epoch": 2.06, "learning_rate": 2.7979303316475585e-05, "loss": 1.1582, "step": 20546 }, { "epoch": 2.06, "learning_rate": 2.797691558891947e-05, "loss": 0.5705, "step": 20547 }, { "epoch": 2.06, "learning_rate": 2.7974527874237486e-05, "loss": 0.3547, "step": 20548 }, { "epoch": 2.06, "learning_rate": 2.7972140172444848e-05, "loss": 0.9218, "step": 20549 }, { "epoch": 2.06, "learning_rate": 2.7969752483556737e-05, "loss": 0.3257, "step": 20550 }, { "epoch": 2.06, "learning_rate": 2.7967364807588335e-05, "loss": 0.4645, "step": 20551 }, { "epoch": 2.06, "learning_rate": 2.796497714455486e-05, "loss": 0.5232, "step": 20552 }, { "epoch": 2.06, "learning_rate": 2.7962589494471504e-05, "loss": 0.2061, "step": 20553 }, { "epoch": 2.06, "learning_rate": 2.796020185735344e-05, "loss": 0.1384, "step": 20554 }, { "epoch": 2.06, "learning_rate": 2.795781423321588e-05, "loss": 0.3349, "step": 20555 }, { "epoch": 2.06, "learning_rate": 2.795542662207401e-05, "loss": 0.0651, "step": 20556 }, { "epoch": 2.06, "learning_rate": 2.795303902394303e-05, "loss": 0.0429, "step": 20557 }, { "epoch": 2.06, "learning_rate": 2.795065143883814e-05, "loss": 0.1681, "step": 20558 }, { "epoch": 2.06, "learning_rate": 2.794826386677451e-05, "loss": 0.4454, "step": 20559 }, { "epoch": 2.06, "learning_rate": 2.794587630776736e-05, "loss": 0.4369, "step": 20560 }, { "epoch": 2.06, "learning_rate": 2.7943488761831863e-05, "loss": 0.1789, "step": 20561 }, { "epoch": 2.06, "learning_rate": 2.794110122898322e-05, "loss": 0.4508, "step": 20562 }, { "epoch": 2.06, "learning_rate": 2.7938713709236638e-05, "loss": 1.1202, "step": 20563 }, { "epoch": 2.06, "learning_rate": 2.793632620260729e-05, "loss": 0.2828, "step": 20564 }, { "epoch": 2.06, "learning_rate": 2.793393870911037e-05, "loss": 0.2347, "step": 20565 }, { "epoch": 2.06, "learning_rate": 2.793155122876109e-05, "loss": 0.2363, "step": 20566 }, { "epoch": 2.06, "learning_rate": 2.792916376157463e-05, "loss": 0.4898, "step": 20567 }, { "epoch": 2.06, "learning_rate": 2.7926776307566177e-05, "loss": 0.5202, "step": 20568 }, { "epoch": 2.06, "learning_rate": 2.7924388866750933e-05, "loss": 0.3924, "step": 20569 }, { "epoch": 2.06, "learning_rate": 2.792200143914409e-05, "loss": 0.302, "step": 20570 }, { "epoch": 2.06, "learning_rate": 2.7919614024760834e-05, "loss": 0.4274, "step": 20571 }, { "epoch": 2.06, "learning_rate": 2.7917226623616378e-05, "loss": 0.406, "step": 20572 }, { "epoch": 2.06, "learning_rate": 2.7914839235725887e-05, "loss": 0.5169, "step": 20573 }, { "epoch": 2.06, "learning_rate": 2.7912451861104563e-05, "loss": 0.3773, "step": 20574 }, { "epoch": 2.06, "learning_rate": 2.7910064499767607e-05, "loss": 0.1185, "step": 20575 }, { "epoch": 2.06, "learning_rate": 2.7907677151730212e-05, "loss": 0.5987, "step": 20576 }, { "epoch": 2.06, "learning_rate": 2.7905289817007552e-05, "loss": 0.5189, "step": 20577 }, { "epoch": 2.06, "learning_rate": 2.7902902495614836e-05, "loss": 0.6223, "step": 20578 }, { "epoch": 2.06, "learning_rate": 2.7900515187567253e-05, "loss": 0.5456, "step": 20579 }, { "epoch": 2.06, "learning_rate": 2.789812789287998e-05, "loss": 0.1818, "step": 20580 }, { "epoch": 2.06, "learning_rate": 2.789574061156824e-05, "loss": 0.0471, "step": 20581 }, { "epoch": 2.06, "learning_rate": 2.7893353343647187e-05, "loss": 0.1603, "step": 20582 }, { "epoch": 2.06, "learning_rate": 2.789096608913205e-05, "loss": 0.3019, "step": 20583 }, { "epoch": 2.06, "learning_rate": 2.7888578848037995e-05, "loss": 0.5732, "step": 20584 }, { "epoch": 2.06, "learning_rate": 2.7886191620380218e-05, "loss": 0.0723, "step": 20585 }, { "epoch": 2.06, "learning_rate": 2.788380440617392e-05, "loss": 0.6776, "step": 20586 }, { "epoch": 2.06, "learning_rate": 2.7881417205434286e-05, "loss": 0.6656, "step": 20587 }, { "epoch": 2.06, "learning_rate": 2.7879030018176494e-05, "loss": 0.0845, "step": 20588 }, { "epoch": 2.07, "learning_rate": 2.787664284441576e-05, "loss": 0.2684, "step": 20589 }, { "epoch": 2.07, "learning_rate": 2.7874255684167267e-05, "loss": 0.1481, "step": 20590 }, { "epoch": 2.07, "learning_rate": 2.787186853744619e-05, "loss": 0.263, "step": 20591 }, { "epoch": 2.07, "learning_rate": 2.7869481404267737e-05, "loss": 0.2686, "step": 20592 }, { "epoch": 2.07, "learning_rate": 2.7867094284647095e-05, "loss": 0.599, "step": 20593 }, { "epoch": 2.07, "learning_rate": 2.786470717859945e-05, "loss": 0.1611, "step": 20594 }, { "epoch": 2.07, "learning_rate": 2.7862320086140002e-05, "loss": 0.3572, "step": 20595 }, { "epoch": 2.07, "learning_rate": 2.7859933007283935e-05, "loss": 0.3282, "step": 20596 }, { "epoch": 2.07, "learning_rate": 2.7857545942046434e-05, "loss": 0.1717, "step": 20597 }, { "epoch": 2.07, "learning_rate": 2.78551588904427e-05, "loss": 0.476, "step": 20598 }, { "epoch": 2.07, "learning_rate": 2.7852771852487924e-05, "loss": 0.1091, "step": 20599 }, { "epoch": 2.07, "learning_rate": 2.7850384828197282e-05, "loss": 0.1854, "step": 20600 }, { "epoch": 2.07, "learning_rate": 2.784799781758598e-05, "loss": 0.5808, "step": 20601 }, { "epoch": 2.07, "learning_rate": 2.7845610820669196e-05, "loss": 0.1507, "step": 20602 }, { "epoch": 2.07, "learning_rate": 2.784322383746212e-05, "loss": 0.5281, "step": 20603 }, { "epoch": 2.07, "learning_rate": 2.784083686797996e-05, "loss": 0.4735, "step": 20604 }, { "epoch": 2.07, "learning_rate": 2.783844991223789e-05, "loss": 0.2706, "step": 20605 }, { "epoch": 2.07, "learning_rate": 2.7836062970251093e-05, "loss": 0.224, "step": 20606 }, { "epoch": 2.07, "learning_rate": 2.7833676042034774e-05, "loss": 0.4217, "step": 20607 }, { "epoch": 2.07, "learning_rate": 2.7831289127604127e-05, "loss": 0.5269, "step": 20608 }, { "epoch": 2.07, "learning_rate": 2.7828902226974316e-05, "loss": 0.1512, "step": 20609 }, { "epoch": 2.07, "learning_rate": 2.782651534016055e-05, "loss": 0.3467, "step": 20610 }, { "epoch": 2.07, "learning_rate": 2.782412846717801e-05, "loss": 1.1714, "step": 20611 }, { "epoch": 2.07, "learning_rate": 2.782174160804189e-05, "loss": 0.103, "step": 20612 }, { "epoch": 2.07, "learning_rate": 2.781935476276739e-05, "loss": 0.1031, "step": 20613 }, { "epoch": 2.07, "learning_rate": 2.7816967931369664e-05, "loss": 0.5152, "step": 20614 }, { "epoch": 2.07, "learning_rate": 2.7814581113863943e-05, "loss": 0.1587, "step": 20615 }, { "epoch": 2.07, "learning_rate": 2.7812194310265392e-05, "loss": 0.1988, "step": 20616 }, { "epoch": 2.07, "learning_rate": 2.78098075205892e-05, "loss": 0.2985, "step": 20617 }, { "epoch": 2.07, "learning_rate": 2.7807420744850567e-05, "loss": 0.402, "step": 20618 }, { "epoch": 2.07, "learning_rate": 2.7805033983064673e-05, "loss": 0.886, "step": 20619 }, { "epoch": 2.07, "learning_rate": 2.7802647235246693e-05, "loss": 0.4674, "step": 20620 }, { "epoch": 2.07, "learning_rate": 2.7800260501411845e-05, "loss": 0.2017, "step": 20621 }, { "epoch": 2.07, "learning_rate": 2.7797873781575305e-05, "loss": 0.2302, "step": 20622 }, { "epoch": 2.07, "learning_rate": 2.7795487075752246e-05, "loss": 0.0422, "step": 20623 }, { "epoch": 2.07, "learning_rate": 2.7793100383957877e-05, "loss": 0.7294, "step": 20624 }, { "epoch": 2.07, "learning_rate": 2.7790713706207374e-05, "loss": 0.1151, "step": 20625 }, { "epoch": 2.07, "learning_rate": 2.7788327042515924e-05, "loss": 0.6106, "step": 20626 }, { "epoch": 2.07, "learning_rate": 2.7785940392898734e-05, "loss": 0.3492, "step": 20627 }, { "epoch": 2.07, "learning_rate": 2.7783553757370967e-05, "loss": 0.2598, "step": 20628 }, { "epoch": 2.07, "learning_rate": 2.7781167135947816e-05, "loss": 0.0997, "step": 20629 }, { "epoch": 2.07, "learning_rate": 2.7778780528644478e-05, "loss": 0.6574, "step": 20630 }, { "epoch": 2.07, "learning_rate": 2.7776393935476145e-05, "loss": 0.4761, "step": 20631 }, { "epoch": 2.07, "learning_rate": 2.777400735645798e-05, "loss": 0.1819, "step": 20632 }, { "epoch": 2.07, "learning_rate": 2.7771620791605193e-05, "loss": 0.3743, "step": 20633 }, { "epoch": 2.07, "learning_rate": 2.7769234240932967e-05, "loss": 0.4581, "step": 20634 }, { "epoch": 2.07, "learning_rate": 2.7766847704456474e-05, "loss": 0.3745, "step": 20635 }, { "epoch": 2.07, "learning_rate": 2.776446118219093e-05, "loss": 0.6003, "step": 20636 }, { "epoch": 2.07, "learning_rate": 2.7762074674151478e-05, "loss": 0.3857, "step": 20637 }, { "epoch": 2.07, "learning_rate": 2.7759688180353357e-05, "loss": 0.4414, "step": 20638 }, { "epoch": 2.07, "learning_rate": 2.7757301700811722e-05, "loss": 0.2793, "step": 20639 }, { "epoch": 2.07, "learning_rate": 2.7754915235541755e-05, "loss": 0.6568, "step": 20640 }, { "epoch": 2.07, "learning_rate": 2.7752528784558672e-05, "loss": 0.4765, "step": 20641 }, { "epoch": 2.07, "learning_rate": 2.775014234787763e-05, "loss": 0.2134, "step": 20642 }, { "epoch": 2.07, "learning_rate": 2.7747755925513822e-05, "loss": 0.116, "step": 20643 }, { "epoch": 2.07, "learning_rate": 2.7745369517482442e-05, "loss": 0.0874, "step": 20644 }, { "epoch": 2.07, "learning_rate": 2.7742983123798682e-05, "loss": 0.3781, "step": 20645 }, { "epoch": 2.07, "learning_rate": 2.7740596744477697e-05, "loss": 0.4492, "step": 20646 }, { "epoch": 2.07, "learning_rate": 2.7738210379534717e-05, "loss": 0.1918, "step": 20647 }, { "epoch": 2.07, "learning_rate": 2.7735824028984897e-05, "loss": 0.4943, "step": 20648 }, { "epoch": 2.07, "learning_rate": 2.7733437692843426e-05, "loss": 0.5669, "step": 20649 }, { "epoch": 2.07, "learning_rate": 2.7731051371125506e-05, "loss": 0.2508, "step": 20650 }, { "epoch": 2.07, "learning_rate": 2.772866506384631e-05, "loss": 0.7396, "step": 20651 }, { "epoch": 2.07, "learning_rate": 2.7726278771021016e-05, "loss": 0.1311, "step": 20652 }, { "epoch": 2.07, "learning_rate": 2.7723892492664827e-05, "loss": 0.8284, "step": 20653 }, { "epoch": 2.07, "learning_rate": 2.7721506228792925e-05, "loss": 0.446, "step": 20654 }, { "epoch": 2.07, "learning_rate": 2.771911997942048e-05, "loss": 0.212, "step": 20655 }, { "epoch": 2.07, "learning_rate": 2.771673374456269e-05, "loss": 0.5722, "step": 20656 }, { "epoch": 2.07, "learning_rate": 2.771434752423474e-05, "loss": 0.3462, "step": 20657 }, { "epoch": 2.07, "learning_rate": 2.771196131845181e-05, "loss": 0.3996, "step": 20658 }, { "epoch": 2.07, "learning_rate": 2.7709575127229093e-05, "loss": 0.7265, "step": 20659 }, { "epoch": 2.07, "learning_rate": 2.7707188950581768e-05, "loss": 0.5974, "step": 20660 }, { "epoch": 2.07, "learning_rate": 2.7704802788525012e-05, "loss": 0.3793, "step": 20661 }, { "epoch": 2.07, "learning_rate": 2.7702416641074025e-05, "loss": 0.5021, "step": 20662 }, { "epoch": 2.07, "learning_rate": 2.7700030508243978e-05, "loss": 0.5865, "step": 20663 }, { "epoch": 2.07, "learning_rate": 2.7697644390050077e-05, "loss": 0.3926, "step": 20664 }, { "epoch": 2.07, "learning_rate": 2.7695258286507482e-05, "loss": 0.6993, "step": 20665 }, { "epoch": 2.07, "learning_rate": 2.769287219763138e-05, "loss": 0.3165, "step": 20666 }, { "epoch": 2.07, "learning_rate": 2.7690486123436967e-05, "loss": 0.6372, "step": 20667 }, { "epoch": 2.07, "learning_rate": 2.768810006393943e-05, "loss": 0.3274, "step": 20668 }, { "epoch": 2.07, "learning_rate": 2.7685714019153925e-05, "loss": 0.3179, "step": 20669 }, { "epoch": 2.07, "learning_rate": 2.7683327989095677e-05, "loss": 0.6526, "step": 20670 }, { "epoch": 2.07, "learning_rate": 2.768094197377984e-05, "loss": 0.1601, "step": 20671 }, { "epoch": 2.07, "learning_rate": 2.7678555973221598e-05, "loss": 0.3328, "step": 20672 }, { "epoch": 2.07, "learning_rate": 2.7676169987436156e-05, "loss": 0.6628, "step": 20673 }, { "epoch": 2.07, "learning_rate": 2.7673784016438682e-05, "loss": 1.0632, "step": 20674 }, { "epoch": 2.07, "learning_rate": 2.767139806024435e-05, "loss": 0.3263, "step": 20675 }, { "epoch": 2.07, "learning_rate": 2.7669012118868364e-05, "loss": 0.5754, "step": 20676 }, { "epoch": 2.07, "learning_rate": 2.7666626192325903e-05, "loss": 0.3864, "step": 20677 }, { "epoch": 2.07, "learning_rate": 2.766424028063213e-05, "loss": 0.231, "step": 20678 }, { "epoch": 2.07, "learning_rate": 2.7661854383802258e-05, "loss": 0.3789, "step": 20679 }, { "epoch": 2.07, "learning_rate": 2.765946850185145e-05, "loss": 0.1657, "step": 20680 }, { "epoch": 2.07, "learning_rate": 2.7657082634794887e-05, "loss": 0.1268, "step": 20681 }, { "epoch": 2.07, "learning_rate": 2.765469678264777e-05, "loss": 0.1513, "step": 20682 }, { "epoch": 2.07, "learning_rate": 2.765231094542527e-05, "loss": 0.057, "step": 20683 }, { "epoch": 2.07, "learning_rate": 2.7649925123142557e-05, "loss": 0.3398, "step": 20684 }, { "epoch": 2.07, "learning_rate": 2.764753931581484e-05, "loss": 0.083, "step": 20685 }, { "epoch": 2.07, "learning_rate": 2.764515352345729e-05, "loss": 0.4143, "step": 20686 }, { "epoch": 2.08, "learning_rate": 2.7642767746085074e-05, "loss": 0.4223, "step": 20687 }, { "epoch": 2.08, "learning_rate": 2.7640381983713394e-05, "loss": 0.1101, "step": 20688 }, { "epoch": 2.08, "learning_rate": 2.7637996236357427e-05, "loss": 0.451, "step": 20689 }, { "epoch": 2.08, "learning_rate": 2.7635610504032347e-05, "loss": 0.4555, "step": 20690 }, { "epoch": 2.08, "learning_rate": 2.7633224786753354e-05, "loss": 0.6391, "step": 20691 }, { "epoch": 2.08, "learning_rate": 2.7630839084535596e-05, "loss": 0.782, "step": 20692 }, { "epoch": 2.08, "learning_rate": 2.7628453397394302e-05, "loss": 0.2177, "step": 20693 }, { "epoch": 2.08, "learning_rate": 2.7626067725344616e-05, "loss": 0.3321, "step": 20694 }, { "epoch": 2.08, "learning_rate": 2.7623682068401727e-05, "loss": 0.8146, "step": 20695 }, { "epoch": 2.08, "learning_rate": 2.7621296426580837e-05, "loss": 1.7095, "step": 20696 }, { "epoch": 2.08, "learning_rate": 2.7618910799897107e-05, "loss": 0.4799, "step": 20697 }, { "epoch": 2.08, "learning_rate": 2.7616525188365708e-05, "loss": 0.188, "step": 20698 }, { "epoch": 2.08, "learning_rate": 2.7614139592001847e-05, "loss": 0.1985, "step": 20699 }, { "epoch": 2.08, "learning_rate": 2.7611754010820703e-05, "loss": 0.3918, "step": 20700 }, { "epoch": 2.08, "learning_rate": 2.7609368444837427e-05, "loss": 0.14, "step": 20701 }, { "epoch": 2.08, "learning_rate": 2.760698289406724e-05, "loss": 0.889, "step": 20702 }, { "epoch": 2.08, "learning_rate": 2.7604597358525298e-05, "loss": 0.541, "step": 20703 }, { "epoch": 2.08, "learning_rate": 2.7602211838226778e-05, "loss": 0.668, "step": 20704 }, { "epoch": 2.08, "learning_rate": 2.7599826333186885e-05, "loss": 0.2945, "step": 20705 }, { "epoch": 2.08, "learning_rate": 2.7597440843420776e-05, "loss": 0.5152, "step": 20706 }, { "epoch": 2.08, "learning_rate": 2.759505536894363e-05, "loss": 0.1348, "step": 20707 }, { "epoch": 2.08, "learning_rate": 2.7592669909770648e-05, "loss": 0.0712, "step": 20708 }, { "epoch": 2.08, "learning_rate": 2.7590284465917007e-05, "loss": 0.3663, "step": 20709 }, { "epoch": 2.08, "learning_rate": 2.7587899037397858e-05, "loss": 0.2702, "step": 20710 }, { "epoch": 2.08, "learning_rate": 2.758551362422842e-05, "loss": 0.1875, "step": 20711 }, { "epoch": 2.08, "learning_rate": 2.758312822642385e-05, "loss": 0.6204, "step": 20712 }, { "epoch": 2.08, "learning_rate": 2.7580742843999322e-05, "loss": 0.1102, "step": 20713 }, { "epoch": 2.08, "learning_rate": 2.7578357476970042e-05, "loss": 0.2904, "step": 20714 }, { "epoch": 2.08, "learning_rate": 2.7575972125351167e-05, "loss": 0.3853, "step": 20715 }, { "epoch": 2.08, "learning_rate": 2.7573586789157878e-05, "loss": 0.2379, "step": 20716 }, { "epoch": 2.08, "learning_rate": 2.7571201468405365e-05, "loss": 0.2657, "step": 20717 }, { "epoch": 2.08, "learning_rate": 2.7568816163108794e-05, "loss": 0.2156, "step": 20718 }, { "epoch": 2.08, "learning_rate": 2.756643087328337e-05, "loss": 0.6623, "step": 20719 }, { "epoch": 2.08, "learning_rate": 2.756404559894424e-05, "loss": 0.1491, "step": 20720 }, { "epoch": 2.08, "learning_rate": 2.7561660340106596e-05, "loss": 0.0604, "step": 20721 }, { "epoch": 2.08, "learning_rate": 2.755927509678562e-05, "loss": 0.5146, "step": 20722 }, { "epoch": 2.08, "learning_rate": 2.75568898689965e-05, "loss": 0.2277, "step": 20723 }, { "epoch": 2.08, "learning_rate": 2.755450465675438e-05, "loss": 0.7133, "step": 20724 }, { "epoch": 2.08, "learning_rate": 2.7552119460074485e-05, "loss": 0.2033, "step": 20725 }, { "epoch": 2.08, "learning_rate": 2.7549734278971966e-05, "loss": 0.4496, "step": 20726 }, { "epoch": 2.08, "learning_rate": 2.7547349113461998e-05, "loss": 0.1925, "step": 20727 }, { "epoch": 2.08, "learning_rate": 2.7544963963559775e-05, "loss": 0.1221, "step": 20728 }, { "epoch": 2.08, "learning_rate": 2.7542578829280464e-05, "loss": 0.2339, "step": 20729 }, { "epoch": 2.08, "learning_rate": 2.7540193710639245e-05, "loss": 0.2791, "step": 20730 }, { "epoch": 2.08, "learning_rate": 2.7537808607651297e-05, "loss": 0.29, "step": 20731 }, { "epoch": 2.08, "learning_rate": 2.7535423520331808e-05, "loss": 0.166, "step": 20732 }, { "epoch": 2.08, "learning_rate": 2.7533038448695927e-05, "loss": 0.2055, "step": 20733 }, { "epoch": 2.08, "learning_rate": 2.7530653392758866e-05, "loss": 0.4138, "step": 20734 }, { "epoch": 2.08, "learning_rate": 2.7528268352535785e-05, "loss": 0.2822, "step": 20735 }, { "epoch": 2.08, "learning_rate": 2.7525883328041858e-05, "loss": 0.1724, "step": 20736 }, { "epoch": 2.08, "learning_rate": 2.752349831929228e-05, "loss": 0.5442, "step": 20737 }, { "epoch": 2.08, "learning_rate": 2.752111332630221e-05, "loss": 0.2117, "step": 20738 }, { "epoch": 2.08, "learning_rate": 2.7518728349086822e-05, "loss": 0.225, "step": 20739 }, { "epoch": 2.08, "learning_rate": 2.7516343387661315e-05, "loss": 0.4696, "step": 20740 }, { "epoch": 2.08, "learning_rate": 2.751395844204086e-05, "loss": 0.1524, "step": 20741 }, { "epoch": 2.08, "learning_rate": 2.7511573512240605e-05, "loss": 0.3536, "step": 20742 }, { "epoch": 2.08, "learning_rate": 2.750918859827577e-05, "loss": 0.578, "step": 20743 }, { "epoch": 2.08, "learning_rate": 2.7506803700161497e-05, "loss": 0.1721, "step": 20744 }, { "epoch": 2.08, "learning_rate": 2.7504418817912988e-05, "loss": 0.2442, "step": 20745 }, { "epoch": 2.08, "learning_rate": 2.750203395154541e-05, "loss": 0.329, "step": 20746 }, { "epoch": 2.08, "learning_rate": 2.749964910107392e-05, "loss": 0.2829, "step": 20747 }, { "epoch": 2.08, "learning_rate": 2.7497264266513733e-05, "loss": 0.3453, "step": 20748 }, { "epoch": 2.08, "learning_rate": 2.7494879447879993e-05, "loss": 0.5142, "step": 20749 }, { "epoch": 2.08, "learning_rate": 2.7492494645187884e-05, "loss": 0.0916, "step": 20750 }, { "epoch": 2.08, "learning_rate": 2.74901098584526e-05, "loss": 0.1637, "step": 20751 }, { "epoch": 2.08, "learning_rate": 2.7487725087689292e-05, "loss": 0.0856, "step": 20752 }, { "epoch": 2.08, "learning_rate": 2.748534033291314e-05, "loss": 0.3375, "step": 20753 }, { "epoch": 2.08, "learning_rate": 2.7482955594139332e-05, "loss": 0.4093, "step": 20754 }, { "epoch": 2.08, "learning_rate": 2.7480570871383044e-05, "loss": 0.0907, "step": 20755 }, { "epoch": 2.08, "learning_rate": 2.7478186164659426e-05, "loss": 0.1379, "step": 20756 }, { "epoch": 2.08, "learning_rate": 2.747580147398369e-05, "loss": 0.1263, "step": 20757 }, { "epoch": 2.08, "learning_rate": 2.7473416799370987e-05, "loss": 0.2652, "step": 20758 }, { "epoch": 2.08, "learning_rate": 2.7471032140836493e-05, "loss": 0.4147, "step": 20759 }, { "epoch": 2.08, "learning_rate": 2.74686474983954e-05, "loss": 0.4484, "step": 20760 }, { "epoch": 2.08, "learning_rate": 2.7466262872062863e-05, "loss": 0.2418, "step": 20761 }, { "epoch": 2.08, "learning_rate": 2.7463878261854062e-05, "loss": 0.1985, "step": 20762 }, { "epoch": 2.08, "learning_rate": 2.7461493667784183e-05, "loss": 0.6364, "step": 20763 }, { "epoch": 2.08, "learning_rate": 2.7459109089868397e-05, "loss": 0.6553, "step": 20764 }, { "epoch": 2.08, "learning_rate": 2.7456724528121854e-05, "loss": 0.1978, "step": 20765 }, { "epoch": 2.08, "learning_rate": 2.7454339982559766e-05, "loss": 0.2562, "step": 20766 }, { "epoch": 2.08, "learning_rate": 2.7451955453197285e-05, "loss": 0.224, "step": 20767 }, { "epoch": 2.08, "learning_rate": 2.7449570940049586e-05, "loss": 0.6083, "step": 20768 }, { "epoch": 2.08, "learning_rate": 2.7447186443131858e-05, "loss": 0.5947, "step": 20769 }, { "epoch": 2.08, "learning_rate": 2.744480196245926e-05, "loss": 0.3964, "step": 20770 }, { "epoch": 2.08, "learning_rate": 2.744241749804696e-05, "loss": 0.0901, "step": 20771 }, { "epoch": 2.08, "learning_rate": 2.744003304991015e-05, "loss": 0.0853, "step": 20772 }, { "epoch": 2.08, "learning_rate": 2.743764861806399e-05, "loss": 0.2426, "step": 20773 }, { "epoch": 2.08, "learning_rate": 2.743526420252367e-05, "loss": 0.3436, "step": 20774 }, { "epoch": 2.08, "learning_rate": 2.7432879803304355e-05, "loss": 0.8456, "step": 20775 }, { "epoch": 2.08, "learning_rate": 2.7430495420421202e-05, "loss": 0.6327, "step": 20776 }, { "epoch": 2.08, "learning_rate": 2.7428111053889404e-05, "loss": 0.1603, "step": 20777 }, { "epoch": 2.08, "learning_rate": 2.7425726703724143e-05, "loss": 0.2338, "step": 20778 }, { "epoch": 2.08, "learning_rate": 2.7423342369940553e-05, "loss": 0.8834, "step": 20779 }, { "epoch": 2.08, "learning_rate": 2.7420958052553853e-05, "loss": 0.3909, "step": 20780 }, { "epoch": 2.08, "learning_rate": 2.7418573751579186e-05, "loss": 0.2936, "step": 20781 }, { "epoch": 2.08, "learning_rate": 2.741618946703173e-05, "loss": 0.6931, "step": 20782 }, { "epoch": 2.08, "learning_rate": 2.741380519892667e-05, "loss": 1.3624, "step": 20783 }, { "epoch": 2.08, "learning_rate": 2.7411420947279166e-05, "loss": 0.2395, "step": 20784 }, { "epoch": 2.08, "learning_rate": 2.740903671210439e-05, "loss": 0.1951, "step": 20785 }, { "epoch": 2.09, "learning_rate": 2.7406652493417524e-05, "loss": 0.4219, "step": 20786 }, { "epoch": 2.09, "learning_rate": 2.740426829123374e-05, "loss": 0.2958, "step": 20787 }, { "epoch": 2.09, "learning_rate": 2.7401884105568185e-05, "loss": 0.4375, "step": 20788 }, { "epoch": 2.09, "learning_rate": 2.7399499936436073e-05, "loss": 0.164, "step": 20789 }, { "epoch": 2.09, "learning_rate": 2.7397115783852546e-05, "loss": 0.3878, "step": 20790 }, { "epoch": 2.09, "learning_rate": 2.739473164783278e-05, "loss": 0.3101, "step": 20791 }, { "epoch": 2.09, "learning_rate": 2.7392347528391964e-05, "loss": 0.0867, "step": 20792 }, { "epoch": 2.09, "learning_rate": 2.7389963425545245e-05, "loss": 0.1329, "step": 20793 }, { "epoch": 2.09, "learning_rate": 2.7387579339307803e-05, "loss": 0.651, "step": 20794 }, { "epoch": 2.09, "learning_rate": 2.738519526969482e-05, "loss": 0.3288, "step": 20795 }, { "epoch": 2.09, "learning_rate": 2.7382811216721462e-05, "loss": 0.2806, "step": 20796 }, { "epoch": 2.09, "learning_rate": 2.738042718040288e-05, "loss": 0.6096, "step": 20797 }, { "epoch": 2.09, "learning_rate": 2.7378043160754282e-05, "loss": 0.7298, "step": 20798 }, { "epoch": 2.09, "learning_rate": 2.737565915779081e-05, "loss": 0.4513, "step": 20799 }, { "epoch": 2.09, "learning_rate": 2.7373275171527647e-05, "loss": 0.5228, "step": 20800 }, { "epoch": 2.09, "learning_rate": 2.737089120197997e-05, "loss": 0.6021, "step": 20801 }, { "epoch": 2.09, "learning_rate": 2.736850724916292e-05, "loss": 0.3411, "step": 20802 }, { "epoch": 2.09, "learning_rate": 2.736612331309171e-05, "loss": 0.3559, "step": 20803 }, { "epoch": 2.09, "learning_rate": 2.7363739393781487e-05, "loss": 0.5159, "step": 20804 }, { "epoch": 2.09, "learning_rate": 2.7361355491247412e-05, "loss": 0.1856, "step": 20805 }, { "epoch": 2.09, "learning_rate": 2.7358971605504674e-05, "loss": 0.7364, "step": 20806 }, { "epoch": 2.09, "learning_rate": 2.7356587736568442e-05, "loss": 0.3101, "step": 20807 }, { "epoch": 2.09, "learning_rate": 2.7354203884453865e-05, "loss": 0.251, "step": 20808 }, { "epoch": 2.09, "learning_rate": 2.735182004917614e-05, "loss": 0.045, "step": 20809 }, { "epoch": 2.09, "learning_rate": 2.7349436230750428e-05, "loss": 0.224, "step": 20810 }, { "epoch": 2.09, "learning_rate": 2.7347052429191875e-05, "loss": 0.5529, "step": 20811 }, { "epoch": 2.09, "learning_rate": 2.7344668644515693e-05, "loss": 0.2823, "step": 20812 }, { "epoch": 2.09, "learning_rate": 2.7342284876737024e-05, "loss": 0.3962, "step": 20813 }, { "epoch": 2.09, "learning_rate": 2.7339901125871032e-05, "loss": 0.4054, "step": 20814 }, { "epoch": 2.09, "learning_rate": 2.7337517391932916e-05, "loss": 0.8411, "step": 20815 }, { "epoch": 2.09, "learning_rate": 2.7335133674937816e-05, "loss": 0.1146, "step": 20816 }, { "epoch": 2.09, "learning_rate": 2.7332749974900903e-05, "loss": 0.2768, "step": 20817 }, { "epoch": 2.09, "learning_rate": 2.7330366291837367e-05, "loss": 0.158, "step": 20818 }, { "epoch": 2.09, "learning_rate": 2.7327982625762365e-05, "loss": 0.1664, "step": 20819 }, { "epoch": 2.09, "learning_rate": 2.7325598976691048e-05, "loss": 0.3295, "step": 20820 }, { "epoch": 2.09, "learning_rate": 2.7323215344638625e-05, "loss": 0.2442, "step": 20821 }, { "epoch": 2.09, "learning_rate": 2.7320831729620228e-05, "loss": 0.189, "step": 20822 }, { "epoch": 2.09, "learning_rate": 2.7318448131651038e-05, "loss": 1.031, "step": 20823 }, { "epoch": 2.09, "learning_rate": 2.7316064550746236e-05, "loss": 0.241, "step": 20824 }, { "epoch": 2.09, "learning_rate": 2.7313680986920968e-05, "loss": 0.2229, "step": 20825 }, { "epoch": 2.09, "learning_rate": 2.7311297440190408e-05, "loss": 0.2034, "step": 20826 }, { "epoch": 2.09, "learning_rate": 2.7308913910569737e-05, "loss": 0.2494, "step": 20827 }, { "epoch": 2.09, "learning_rate": 2.7306530398074103e-05, "loss": 0.3566, "step": 20828 }, { "epoch": 2.09, "learning_rate": 2.7304146902718693e-05, "loss": 0.1455, "step": 20829 }, { "epoch": 2.09, "learning_rate": 2.730176342451867e-05, "loss": 0.532, "step": 20830 }, { "epoch": 2.09, "learning_rate": 2.7299379963489194e-05, "loss": 0.1207, "step": 20831 }, { "epoch": 2.09, "learning_rate": 2.729699651964544e-05, "loss": 0.1828, "step": 20832 }, { "epoch": 2.09, "learning_rate": 2.7294613093002574e-05, "loss": 0.1721, "step": 20833 }, { "epoch": 2.09, "learning_rate": 2.7292229683575743e-05, "loss": 0.4142, "step": 20834 }, { "epoch": 2.09, "learning_rate": 2.7289846291380154e-05, "loss": 0.7412, "step": 20835 }, { "epoch": 2.09, "learning_rate": 2.7287462916430945e-05, "loss": 0.3802, "step": 20836 }, { "epoch": 2.09, "learning_rate": 2.728507955874328e-05, "loss": 0.1579, "step": 20837 }, { "epoch": 2.09, "learning_rate": 2.7282696218332345e-05, "loss": 0.6648, "step": 20838 }, { "epoch": 2.09, "learning_rate": 2.7280312895213306e-05, "loss": 0.2135, "step": 20839 }, { "epoch": 2.09, "learning_rate": 2.7277929589401305e-05, "loss": 0.3956, "step": 20840 }, { "epoch": 2.09, "learning_rate": 2.7275546300911533e-05, "loss": 0.2555, "step": 20841 }, { "epoch": 2.09, "learning_rate": 2.7273163029759155e-05, "loss": 0.0922, "step": 20842 }, { "epoch": 2.09, "learning_rate": 2.727077977595931e-05, "loss": 0.3493, "step": 20843 }, { "epoch": 2.09, "learning_rate": 2.7268396539527204e-05, "loss": 0.2453, "step": 20844 }, { "epoch": 2.09, "learning_rate": 2.7266013320477977e-05, "loss": 0.5759, "step": 20845 }, { "epoch": 2.09, "learning_rate": 2.7263630118826793e-05, "loss": 0.2929, "step": 20846 }, { "epoch": 2.09, "learning_rate": 2.726124693458884e-05, "loss": 0.2972, "step": 20847 }, { "epoch": 2.09, "learning_rate": 2.7258863767779262e-05, "loss": 0.3488, "step": 20848 }, { "epoch": 2.09, "learning_rate": 2.7256480618413223e-05, "loss": 0.1956, "step": 20849 }, { "epoch": 2.09, "learning_rate": 2.725409748650591e-05, "loss": 0.2242, "step": 20850 }, { "epoch": 2.09, "learning_rate": 2.725171437207248e-05, "loss": 0.0893, "step": 20851 }, { "epoch": 2.09, "learning_rate": 2.7249331275128073e-05, "loss": 0.3412, "step": 20852 }, { "epoch": 2.09, "learning_rate": 2.7246948195687897e-05, "loss": 0.9725, "step": 20853 }, { "epoch": 2.09, "learning_rate": 2.7244565133767083e-05, "loss": 0.4751, "step": 20854 }, { "epoch": 2.09, "learning_rate": 2.7242182089380816e-05, "loss": 0.2633, "step": 20855 }, { "epoch": 2.09, "learning_rate": 2.7239799062544256e-05, "loss": 0.3523, "step": 20856 }, { "epoch": 2.09, "learning_rate": 2.7237416053272546e-05, "loss": 0.3045, "step": 20857 }, { "epoch": 2.09, "learning_rate": 2.723503306158089e-05, "loss": 0.4268, "step": 20858 }, { "epoch": 2.09, "learning_rate": 2.7232650087484423e-05, "loss": 0.8337, "step": 20859 }, { "epoch": 2.09, "learning_rate": 2.7230267130998313e-05, "loss": 0.0866, "step": 20860 }, { "epoch": 2.09, "learning_rate": 2.722788419213774e-05, "loss": 0.8184, "step": 20861 }, { "epoch": 2.09, "learning_rate": 2.722550127091786e-05, "loss": 0.2856, "step": 20862 }, { "epoch": 2.09, "learning_rate": 2.7223118367353818e-05, "loss": 0.5838, "step": 20863 }, { "epoch": 2.09, "learning_rate": 2.722073548146081e-05, "loss": 0.0385, "step": 20864 }, { "epoch": 2.09, "learning_rate": 2.7218352613253984e-05, "loss": 0.4287, "step": 20865 }, { "epoch": 2.09, "learning_rate": 2.721596976274848e-05, "loss": 0.3577, "step": 20866 }, { "epoch": 2.09, "learning_rate": 2.721358692995951e-05, "loss": 0.2048, "step": 20867 }, { "epoch": 2.09, "learning_rate": 2.7211204114902207e-05, "loss": 0.8767, "step": 20868 }, { "epoch": 2.09, "learning_rate": 2.7208821317591734e-05, "loss": 0.5819, "step": 20869 }, { "epoch": 2.09, "learning_rate": 2.7206438538043268e-05, "loss": 0.525, "step": 20870 }, { "epoch": 2.09, "learning_rate": 2.7204055776271967e-05, "loss": 0.3051, "step": 20871 }, { "epoch": 2.09, "learning_rate": 2.720167303229298e-05, "loss": 0.4486, "step": 20872 }, { "epoch": 2.09, "learning_rate": 2.7199290306121487e-05, "loss": 0.2623, "step": 20873 }, { "epoch": 2.09, "learning_rate": 2.7196907597772655e-05, "loss": 0.1436, "step": 20874 }, { "epoch": 2.09, "learning_rate": 2.7194524907261612e-05, "loss": 0.1164, "step": 20875 }, { "epoch": 2.09, "learning_rate": 2.7192142234603567e-05, "loss": 0.1479, "step": 20876 }, { "epoch": 2.09, "learning_rate": 2.7189759579813656e-05, "loss": 0.8288, "step": 20877 }, { "epoch": 2.09, "learning_rate": 2.718737694290704e-05, "loss": 0.4093, "step": 20878 }, { "epoch": 2.09, "learning_rate": 2.7184994323898896e-05, "loss": 0.1484, "step": 20879 }, { "epoch": 2.09, "learning_rate": 2.718261172280436e-05, "loss": 0.2659, "step": 20880 }, { "epoch": 2.09, "learning_rate": 2.718022913963863e-05, "loss": 0.4049, "step": 20881 }, { "epoch": 2.09, "learning_rate": 2.7177846574416845e-05, "loss": 0.2553, "step": 20882 }, { "epoch": 2.09, "learning_rate": 2.717546402715416e-05, "loss": 0.6329, "step": 20883 }, { "epoch": 2.1, "learning_rate": 2.7173081497865757e-05, "loss": 0.5434, "step": 20884 }, { "epoch": 2.1, "learning_rate": 2.7170698986566794e-05, "loss": 0.3614, "step": 20885 }, { "epoch": 2.1, "learning_rate": 2.7168316493272414e-05, "loss": 0.1159, "step": 20886 }, { "epoch": 2.1, "learning_rate": 2.7165934017997793e-05, "loss": 0.3952, "step": 20887 }, { "epoch": 2.1, "learning_rate": 2.7163551560758103e-05, "loss": 0.1587, "step": 20888 }, { "epoch": 2.1, "learning_rate": 2.7161169121568468e-05, "loss": 0.2893, "step": 20889 }, { "epoch": 2.1, "learning_rate": 2.7158786700444094e-05, "loss": 0.334, "step": 20890 }, { "epoch": 2.1, "learning_rate": 2.7156404297400112e-05, "loss": 0.2959, "step": 20891 }, { "epoch": 2.1, "learning_rate": 2.7154021912451686e-05, "loss": 0.6204, "step": 20892 }, { "epoch": 2.1, "learning_rate": 2.715163954561399e-05, "loss": 0.3253, "step": 20893 }, { "epoch": 2.1, "learning_rate": 2.7149257196902177e-05, "loss": 0.1678, "step": 20894 }, { "epoch": 2.1, "learning_rate": 2.7146874866331402e-05, "loss": 1.0886, "step": 20895 }, { "epoch": 2.1, "learning_rate": 2.714449255391683e-05, "loss": 0.2228, "step": 20896 }, { "epoch": 2.1, "learning_rate": 2.714211025967363e-05, "loss": 0.6234, "step": 20897 }, { "epoch": 2.1, "learning_rate": 2.7139727983616933e-05, "loss": 0.1964, "step": 20898 }, { "epoch": 2.1, "learning_rate": 2.713734572576194e-05, "loss": 0.2533, "step": 20899 }, { "epoch": 2.1, "learning_rate": 2.7134963486123785e-05, "loss": 0.2756, "step": 20900 }, { "epoch": 2.1, "learning_rate": 2.7132581264717623e-05, "loss": 0.1574, "step": 20901 }, { "epoch": 2.1, "learning_rate": 2.7130199061558628e-05, "loss": 0.1404, "step": 20902 }, { "epoch": 2.1, "learning_rate": 2.7127816876661965e-05, "loss": 0.1592, "step": 20903 }, { "epoch": 2.1, "learning_rate": 2.7125434710042765e-05, "loss": 0.3728, "step": 20904 }, { "epoch": 2.1, "learning_rate": 2.7123052561716214e-05, "loss": 0.3362, "step": 20905 }, { "epoch": 2.1, "learning_rate": 2.712067043169747e-05, "loss": 0.6086, "step": 20906 }, { "epoch": 2.1, "learning_rate": 2.7118288320001663e-05, "loss": 0.1127, "step": 20907 }, { "epoch": 2.1, "learning_rate": 2.7115906226643993e-05, "loss": 1.057, "step": 20908 }, { "epoch": 2.1, "learning_rate": 2.7113524151639585e-05, "loss": 0.3335, "step": 20909 }, { "epoch": 2.1, "learning_rate": 2.7111142095003615e-05, "loss": 0.213, "step": 20910 }, { "epoch": 2.1, "learning_rate": 2.710876005675125e-05, "loss": 0.4081, "step": 20911 }, { "epoch": 2.1, "learning_rate": 2.710637803689761e-05, "loss": 0.1572, "step": 20912 }, { "epoch": 2.1, "learning_rate": 2.7103996035457906e-05, "loss": 1.0631, "step": 20913 }, { "epoch": 2.1, "learning_rate": 2.7101614052447257e-05, "loss": 0.0611, "step": 20914 }, { "epoch": 2.1, "learning_rate": 2.709923208788083e-05, "loss": 0.5058, "step": 20915 }, { "epoch": 2.1, "learning_rate": 2.709685014177379e-05, "loss": 0.9522, "step": 20916 }, { "epoch": 2.1, "learning_rate": 2.70944682141413e-05, "loss": 0.1243, "step": 20917 }, { "epoch": 2.1, "learning_rate": 2.7092086304998498e-05, "loss": 0.6587, "step": 20918 }, { "epoch": 2.1, "learning_rate": 2.708970441436056e-05, "loss": 0.344, "step": 20919 }, { "epoch": 2.1, "learning_rate": 2.708732254224264e-05, "loss": 0.2162, "step": 20920 }, { "epoch": 2.1, "learning_rate": 2.7084940688659868e-05, "loss": 0.2015, "step": 20921 }, { "epoch": 2.1, "learning_rate": 2.708255885362745e-05, "loss": 0.1042, "step": 20922 }, { "epoch": 2.1, "learning_rate": 2.708017703716051e-05, "loss": 0.1731, "step": 20923 }, { "epoch": 2.1, "learning_rate": 2.7077795239274204e-05, "loss": 0.2913, "step": 20924 }, { "epoch": 2.1, "learning_rate": 2.7075413459983706e-05, "loss": 0.4082, "step": 20925 }, { "epoch": 2.1, "learning_rate": 2.707303169930417e-05, "loss": 0.4024, "step": 20926 }, { "epoch": 2.1, "learning_rate": 2.707064995725073e-05, "loss": 0.1948, "step": 20927 }, { "epoch": 2.1, "learning_rate": 2.7068268233838572e-05, "loss": 0.2914, "step": 20928 }, { "epoch": 2.1, "learning_rate": 2.7065886529082843e-05, "loss": 0.4208, "step": 20929 }, { "epoch": 2.1, "learning_rate": 2.7063504842998675e-05, "loss": 0.0745, "step": 20930 }, { "epoch": 2.1, "learning_rate": 2.7061123175601267e-05, "loss": 0.507, "step": 20931 }, { "epoch": 2.1, "learning_rate": 2.7058741526905747e-05, "loss": 0.7556, "step": 20932 }, { "epoch": 2.1, "learning_rate": 2.705635989692727e-05, "loss": 0.2387, "step": 20933 }, { "epoch": 2.1, "learning_rate": 2.7053978285681003e-05, "loss": 0.5061, "step": 20934 }, { "epoch": 2.1, "learning_rate": 2.7051596693182094e-05, "loss": 0.3086, "step": 20935 }, { "epoch": 2.1, "learning_rate": 2.7049215119445713e-05, "loss": 0.4459, "step": 20936 }, { "epoch": 2.1, "learning_rate": 2.7046833564487e-05, "loss": 0.3819, "step": 20937 }, { "epoch": 2.1, "learning_rate": 2.7044452028321103e-05, "loss": 0.2812, "step": 20938 }, { "epoch": 2.1, "learning_rate": 2.70420705109632e-05, "loss": 0.4645, "step": 20939 }, { "epoch": 2.1, "learning_rate": 2.703968901242844e-05, "loss": 0.499, "step": 20940 }, { "epoch": 2.1, "learning_rate": 2.7037307532731962e-05, "loss": 0.1847, "step": 20941 }, { "epoch": 2.1, "learning_rate": 2.703492607188894e-05, "loss": 0.1813, "step": 20942 }, { "epoch": 2.1, "learning_rate": 2.7032544629914523e-05, "loss": 0.1416, "step": 20943 }, { "epoch": 2.1, "learning_rate": 2.7030163206823843e-05, "loss": 0.2148, "step": 20944 }, { "epoch": 2.1, "learning_rate": 2.7027781802632094e-05, "loss": 0.6434, "step": 20945 }, { "epoch": 2.1, "learning_rate": 2.7025400417354408e-05, "loss": 0.5693, "step": 20946 }, { "epoch": 2.1, "learning_rate": 2.7023019051005936e-05, "loss": 0.3319, "step": 20947 }, { "epoch": 2.1, "learning_rate": 2.702063770360184e-05, "loss": 0.6323, "step": 20948 }, { "epoch": 2.1, "learning_rate": 2.701825637515728e-05, "loss": 0.2036, "step": 20949 }, { "epoch": 2.1, "learning_rate": 2.701587506568739e-05, "loss": 0.9483, "step": 20950 }, { "epoch": 2.1, "learning_rate": 2.7013493775207344e-05, "loss": 0.4173, "step": 20951 }, { "epoch": 2.1, "learning_rate": 2.7011112503732295e-05, "loss": 0.3039, "step": 20952 }, { "epoch": 2.1, "learning_rate": 2.7008731251277365e-05, "loss": 0.3161, "step": 20953 }, { "epoch": 2.1, "learning_rate": 2.7006350017857754e-05, "loss": 0.3613, "step": 20954 }, { "epoch": 2.1, "learning_rate": 2.7003968803488583e-05, "loss": 0.2063, "step": 20955 }, { "epoch": 2.1, "learning_rate": 2.7001587608185015e-05, "loss": 0.3236, "step": 20956 }, { "epoch": 2.1, "learning_rate": 2.6999206431962202e-05, "loss": 0.0418, "step": 20957 }, { "epoch": 2.1, "learning_rate": 2.6996825274835307e-05, "loss": 0.0982, "step": 20958 }, { "epoch": 2.1, "learning_rate": 2.699444413681946e-05, "loss": 0.772, "step": 20959 }, { "epoch": 2.1, "learning_rate": 2.699206301792984e-05, "loss": 0.2587, "step": 20960 }, { "epoch": 2.1, "learning_rate": 2.6989681918181568e-05, "loss": 0.1561, "step": 20961 }, { "epoch": 2.1, "learning_rate": 2.698730083758983e-05, "loss": 0.258, "step": 20962 }, { "epoch": 2.1, "learning_rate": 2.6984919776169766e-05, "loss": 0.7912, "step": 20963 }, { "epoch": 2.1, "learning_rate": 2.6982538733936514e-05, "loss": 0.3955, "step": 20964 }, { "epoch": 2.1, "learning_rate": 2.6980157710905244e-05, "loss": 0.1032, "step": 20965 }, { "epoch": 2.1, "learning_rate": 2.6977776707091106e-05, "loss": 0.124, "step": 20966 }, { "epoch": 2.1, "learning_rate": 2.6975395722509237e-05, "loss": 0.0814, "step": 20967 }, { "epoch": 2.1, "learning_rate": 2.697301475717481e-05, "loss": 0.39, "step": 20968 }, { "epoch": 2.1, "learning_rate": 2.6970633811102964e-05, "loss": 0.2054, "step": 20969 }, { "epoch": 2.1, "learning_rate": 2.6968252884308837e-05, "loss": 0.1255, "step": 20970 }, { "epoch": 2.1, "learning_rate": 2.6965871976807607e-05, "loss": 0.3144, "step": 20971 }, { "epoch": 2.1, "learning_rate": 2.696349108861442e-05, "loss": 0.29, "step": 20972 }, { "epoch": 2.1, "learning_rate": 2.696111021974441e-05, "loss": 0.0968, "step": 20973 }, { "epoch": 2.1, "learning_rate": 2.6958729370212744e-05, "loss": 0.6094, "step": 20974 }, { "epoch": 2.1, "learning_rate": 2.6956348540034566e-05, "loss": 0.1742, "step": 20975 }, { "epoch": 2.1, "learning_rate": 2.695396772922502e-05, "loss": 0.3211, "step": 20976 }, { "epoch": 2.1, "learning_rate": 2.695158693779928e-05, "loss": 0.5527, "step": 20977 }, { "epoch": 2.1, "learning_rate": 2.6949206165772475e-05, "loss": 0.9718, "step": 20978 }, { "epoch": 2.1, "learning_rate": 2.6946825413159754e-05, "loss": 0.5127, "step": 20979 }, { "epoch": 2.1, "learning_rate": 2.6944444679976277e-05, "loss": 0.2322, "step": 20980 }, { "epoch": 2.1, "learning_rate": 2.6942063966237205e-05, "loss": 0.1321, "step": 20981 }, { "epoch": 2.1, "learning_rate": 2.6939683271957657e-05, "loss": 0.2466, "step": 20982 }, { "epoch": 2.11, "learning_rate": 2.693730259715281e-05, "loss": 0.3717, "step": 20983 }, { "epoch": 2.11, "learning_rate": 2.6934921941837807e-05, "loss": 0.311, "step": 20984 }, { "epoch": 2.11, "learning_rate": 2.693254130602778e-05, "loss": 0.1263, "step": 20985 }, { "epoch": 2.11, "learning_rate": 2.6930160689737908e-05, "loss": 0.1458, "step": 20986 }, { "epoch": 2.11, "learning_rate": 2.692778009298332e-05, "loss": 0.6709, "step": 20987 }, { "epoch": 2.11, "learning_rate": 2.692539951577917e-05, "loss": 0.1588, "step": 20988 }, { "epoch": 2.11, "learning_rate": 2.6923018958140606e-05, "loss": 0.3119, "step": 20989 }, { "epoch": 2.11, "learning_rate": 2.692063842008278e-05, "loss": 0.2542, "step": 20990 }, { "epoch": 2.11, "learning_rate": 2.6918257901620848e-05, "loss": 0.2221, "step": 20991 }, { "epoch": 2.11, "learning_rate": 2.691587740276994e-05, "loss": 0.6005, "step": 20992 }, { "epoch": 2.11, "learning_rate": 2.6913496923545218e-05, "loss": 0.7863, "step": 20993 }, { "epoch": 2.11, "learning_rate": 2.6911116463961827e-05, "loss": 0.4541, "step": 20994 }, { "epoch": 2.11, "learning_rate": 2.6908736024034923e-05, "loss": 0.1923, "step": 20995 }, { "epoch": 2.11, "learning_rate": 2.6906355603779637e-05, "loss": 0.2378, "step": 20996 }, { "epoch": 2.11, "learning_rate": 2.690397520321113e-05, "loss": 0.1967, "step": 20997 }, { "epoch": 2.11, "learning_rate": 2.690159482234455e-05, "loss": 0.3834, "step": 20998 }, { "epoch": 2.11, "learning_rate": 2.6899214461195033e-05, "loss": 0.7427, "step": 20999 }, { "epoch": 2.11, "learning_rate": 2.6896834119777753e-05, "loss": 0.4199, "step": 21000 }, { "epoch": 2.11, "learning_rate": 2.689445379810783e-05, "loss": 0.4367, "step": 21001 }, { "epoch": 2.11, "learning_rate": 2.6892073496200414e-05, "loss": 0.094, "step": 21002 }, { "epoch": 2.11, "learning_rate": 2.6889693214070666e-05, "loss": 0.2551, "step": 21003 }, { "epoch": 2.11, "learning_rate": 2.6887312951733736e-05, "loss": 0.3725, "step": 21004 }, { "epoch": 2.11, "learning_rate": 2.6884932709204747e-05, "loss": 0.2315, "step": 21005 }, { "epoch": 2.11, "learning_rate": 2.688255248649887e-05, "loss": 0.4268, "step": 21006 }, { "epoch": 2.11, "learning_rate": 2.6880172283631246e-05, "loss": 0.4496, "step": 21007 }, { "epoch": 2.11, "learning_rate": 2.6877792100617004e-05, "loss": 0.8377, "step": 21008 }, { "epoch": 2.11, "learning_rate": 2.6875411937471326e-05, "loss": 0.2577, "step": 21009 }, { "epoch": 2.11, "learning_rate": 2.6873031794209326e-05, "loss": 0.3849, "step": 21010 }, { "epoch": 2.11, "learning_rate": 2.687065167084616e-05, "loss": 0.8105, "step": 21011 }, { "epoch": 2.11, "learning_rate": 2.6868271567396982e-05, "loss": 0.1307, "step": 21012 }, { "epoch": 2.11, "learning_rate": 2.6865891483876938e-05, "loss": 0.2639, "step": 21013 }, { "epoch": 2.11, "learning_rate": 2.6863511420301155e-05, "loss": 0.5078, "step": 21014 }, { "epoch": 2.11, "learning_rate": 2.68611313766848e-05, "loss": 0.3654, "step": 21015 }, { "epoch": 2.11, "learning_rate": 2.6858751353043e-05, "loss": 0.5618, "step": 21016 }, { "epoch": 2.11, "learning_rate": 2.6856371349390923e-05, "loss": 0.2154, "step": 21017 }, { "epoch": 2.11, "learning_rate": 2.6853991365743703e-05, "loss": 0.1938, "step": 21018 }, { "epoch": 2.11, "learning_rate": 2.685161140211648e-05, "loss": 0.6346, "step": 21019 }, { "epoch": 2.11, "learning_rate": 2.6849231458524405e-05, "loss": 0.1263, "step": 21020 }, { "epoch": 2.11, "learning_rate": 2.6846851534982626e-05, "loss": 0.7125, "step": 21021 }, { "epoch": 2.11, "learning_rate": 2.6844471631506275e-05, "loss": 0.2834, "step": 21022 }, { "epoch": 2.11, "learning_rate": 2.684209174811052e-05, "loss": 0.3453, "step": 21023 }, { "epoch": 2.11, "learning_rate": 2.6839711884810485e-05, "loss": 0.3639, "step": 21024 }, { "epoch": 2.11, "learning_rate": 2.6837332041621317e-05, "loss": 0.5459, "step": 21025 }, { "epoch": 2.11, "learning_rate": 2.6834952218558168e-05, "loss": 0.0816, "step": 21026 }, { "epoch": 2.11, "learning_rate": 2.6832572415636192e-05, "loss": 0.5134, "step": 21027 }, { "epoch": 2.11, "learning_rate": 2.6830192632870503e-05, "loss": 0.3561, "step": 21028 }, { "epoch": 2.11, "learning_rate": 2.6827812870276267e-05, "loss": 0.8579, "step": 21029 }, { "epoch": 2.11, "learning_rate": 2.682543312786862e-05, "loss": 0.1907, "step": 21030 }, { "epoch": 2.11, "learning_rate": 2.682305340566271e-05, "loss": 0.2006, "step": 21031 }, { "epoch": 2.11, "learning_rate": 2.682067370367369e-05, "loss": 0.336, "step": 21032 }, { "epoch": 2.11, "learning_rate": 2.6818294021916683e-05, "loss": 0.3809, "step": 21033 }, { "epoch": 2.11, "learning_rate": 2.681591436040684e-05, "loss": 0.3163, "step": 21034 }, { "epoch": 2.11, "learning_rate": 2.6813534719159314e-05, "loss": 0.1192, "step": 21035 }, { "epoch": 2.11, "learning_rate": 2.6811155098189244e-05, "loss": 0.2854, "step": 21036 }, { "epoch": 2.11, "learning_rate": 2.680877549751176e-05, "loss": 0.4881, "step": 21037 }, { "epoch": 2.11, "learning_rate": 2.6806395917142018e-05, "loss": 0.1504, "step": 21038 }, { "epoch": 2.11, "learning_rate": 2.680401635709516e-05, "loss": 0.1823, "step": 21039 }, { "epoch": 2.11, "learning_rate": 2.6801636817386316e-05, "loss": 0.5303, "step": 21040 }, { "epoch": 2.11, "learning_rate": 2.6799257298030655e-05, "loss": 0.3879, "step": 21041 }, { "epoch": 2.11, "learning_rate": 2.679687779904329e-05, "loss": 0.644, "step": 21042 }, { "epoch": 2.11, "learning_rate": 2.6794498320439384e-05, "loss": 0.1465, "step": 21043 }, { "epoch": 2.11, "learning_rate": 2.6792118862234068e-05, "loss": 0.3923, "step": 21044 }, { "epoch": 2.11, "learning_rate": 2.678973942444248e-05, "loss": 0.2948, "step": 21045 }, { "epoch": 2.11, "learning_rate": 2.6787360007079787e-05, "loss": 0.1554, "step": 21046 }, { "epoch": 2.11, "learning_rate": 2.6784980610161102e-05, "loss": 0.311, "step": 21047 }, { "epoch": 2.11, "learning_rate": 2.678260123370157e-05, "loss": 0.532, "step": 21048 }, { "epoch": 2.11, "learning_rate": 2.678022187771635e-05, "loss": 0.0408, "step": 21049 }, { "epoch": 2.11, "learning_rate": 2.6777842542220577e-05, "loss": 0.4919, "step": 21050 }, { "epoch": 2.11, "learning_rate": 2.677546322722938e-05, "loss": 0.8921, "step": 21051 }, { "epoch": 2.11, "learning_rate": 2.6773083932757906e-05, "loss": 0.1671, "step": 21052 }, { "epoch": 2.11, "learning_rate": 2.6770704658821303e-05, "loss": 0.5613, "step": 21053 }, { "epoch": 2.11, "learning_rate": 2.6768325405434703e-05, "loss": 0.7311, "step": 21054 }, { "epoch": 2.11, "learning_rate": 2.6765946172613263e-05, "loss": 0.5093, "step": 21055 }, { "epoch": 2.11, "learning_rate": 2.67635669603721e-05, "loss": 0.3565, "step": 21056 }, { "epoch": 2.11, "learning_rate": 2.6761187768726366e-05, "loss": 0.1209, "step": 21057 }, { "epoch": 2.11, "learning_rate": 2.6758808597691202e-05, "loss": 0.5031, "step": 21058 }, { "epoch": 2.11, "learning_rate": 2.6756429447281754e-05, "loss": 0.2291, "step": 21059 }, { "epoch": 2.11, "learning_rate": 2.6754050317513145e-05, "loss": 0.2087, "step": 21060 }, { "epoch": 2.11, "learning_rate": 2.6751671208400528e-05, "loss": 0.0762, "step": 21061 }, { "epoch": 2.11, "learning_rate": 2.6749292119959045e-05, "loss": 0.1756, "step": 21062 }, { "epoch": 2.11, "learning_rate": 2.6746913052203818e-05, "loss": 0.2366, "step": 21063 }, { "epoch": 2.11, "learning_rate": 2.6744534005150015e-05, "loss": 0.3872, "step": 21064 }, { "epoch": 2.11, "learning_rate": 2.674215497881275e-05, "loss": 0.3751, "step": 21065 }, { "epoch": 2.11, "learning_rate": 2.673977597320717e-05, "loss": 0.6656, "step": 21066 }, { "epoch": 2.11, "learning_rate": 2.673739698834842e-05, "loss": 0.305, "step": 21067 }, { "epoch": 2.11, "learning_rate": 2.6735018024251637e-05, "loss": 0.0969, "step": 21068 }, { "epoch": 2.11, "learning_rate": 2.6732639080931947e-05, "loss": 0.1569, "step": 21069 }, { "epoch": 2.11, "learning_rate": 2.6730260158404507e-05, "loss": 0.1386, "step": 21070 }, { "epoch": 2.11, "learning_rate": 2.6727881256684437e-05, "loss": 0.3172, "step": 21071 }, { "epoch": 2.11, "learning_rate": 2.6725502375786894e-05, "loss": 0.3775, "step": 21072 }, { "epoch": 2.11, "learning_rate": 2.672312351572702e-05, "loss": 0.5321, "step": 21073 }, { "epoch": 2.11, "learning_rate": 2.672074467651992e-05, "loss": 0.1433, "step": 21074 }, { "epoch": 2.11, "learning_rate": 2.6718365858180764e-05, "loss": 0.492, "step": 21075 }, { "epoch": 2.11, "learning_rate": 2.6715987060724677e-05, "loss": 0.0403, "step": 21076 }, { "epoch": 2.11, "learning_rate": 2.6713608284166797e-05, "loss": 0.4306, "step": 21077 }, { "epoch": 2.11, "learning_rate": 2.6711229528522276e-05, "loss": 0.5876, "step": 21078 }, { "epoch": 2.11, "learning_rate": 2.670885079380623e-05, "loss": 0.2807, "step": 21079 }, { "epoch": 2.11, "learning_rate": 2.6706472080033798e-05, "loss": 0.2809, "step": 21080 }, { "epoch": 2.12, "learning_rate": 2.6704093387220134e-05, "loss": 0.3041, "step": 21081 }, { "epoch": 2.12, "learning_rate": 2.670171471538037e-05, "loss": 0.4796, "step": 21082 }, { "epoch": 2.12, "learning_rate": 2.669933606452963e-05, "loss": 0.1707, "step": 21083 }, { "epoch": 2.12, "learning_rate": 2.669695743468306e-05, "loss": 1.0706, "step": 21084 }, { "epoch": 2.12, "learning_rate": 2.6694578825855802e-05, "loss": 0.801, "step": 21085 }, { "epoch": 2.12, "learning_rate": 2.6692200238062975e-05, "loss": 0.5393, "step": 21086 }, { "epoch": 2.12, "learning_rate": 2.6689821671319748e-05, "loss": 0.2475, "step": 21087 }, { "epoch": 2.12, "learning_rate": 2.6687443125641223e-05, "loss": 0.322, "step": 21088 }, { "epoch": 2.12, "learning_rate": 2.668506460104255e-05, "loss": 0.968, "step": 21089 }, { "epoch": 2.12, "learning_rate": 2.6682686097538866e-05, "loss": 0.095, "step": 21090 }, { "epoch": 2.12, "learning_rate": 2.6680307615145313e-05, "loss": 0.3492, "step": 21091 }, { "epoch": 2.12, "learning_rate": 2.6677929153877006e-05, "loss": 0.7944, "step": 21092 }, { "epoch": 2.12, "learning_rate": 2.66755507137491e-05, "loss": 0.2387, "step": 21093 }, { "epoch": 2.12, "learning_rate": 2.6673172294776726e-05, "loss": 0.1704, "step": 21094 }, { "epoch": 2.12, "learning_rate": 2.6670793896975016e-05, "loss": 0.3527, "step": 21095 }, { "epoch": 2.12, "learning_rate": 2.666841552035911e-05, "loss": 0.4424, "step": 21096 }, { "epoch": 2.12, "learning_rate": 2.666603716494414e-05, "loss": 0.4404, "step": 21097 }, { "epoch": 2.12, "learning_rate": 2.6663658830745234e-05, "loss": 0.5694, "step": 21098 }, { "epoch": 2.12, "learning_rate": 2.6661280517777543e-05, "loss": 0.2762, "step": 21099 }, { "epoch": 2.12, "learning_rate": 2.6658902226056185e-05, "loss": 0.2366, "step": 21100 }, { "epoch": 2.12, "learning_rate": 2.6656523955596314e-05, "loss": 0.345, "step": 21101 }, { "epoch": 2.12, "learning_rate": 2.6654145706413043e-05, "loss": 0.5675, "step": 21102 }, { "epoch": 2.12, "learning_rate": 2.6651767478521513e-05, "loss": 0.2135, "step": 21103 }, { "epoch": 2.12, "learning_rate": 2.6649389271936867e-05, "loss": 0.0762, "step": 21104 }, { "epoch": 2.12, "learning_rate": 2.664701108667424e-05, "loss": 0.2752, "step": 21105 }, { "epoch": 2.12, "learning_rate": 2.6644632922748745e-05, "loss": 0.2328, "step": 21106 }, { "epoch": 2.12, "learning_rate": 2.6642254780175537e-05, "loss": 0.1539, "step": 21107 }, { "epoch": 2.12, "learning_rate": 2.6639876658969742e-05, "loss": 0.3455, "step": 21108 }, { "epoch": 2.12, "learning_rate": 2.6637498559146484e-05, "loss": 0.6033, "step": 21109 }, { "epoch": 2.12, "learning_rate": 2.6635120480720925e-05, "loss": 0.3845, "step": 21110 }, { "epoch": 2.12, "learning_rate": 2.663274242370817e-05, "loss": 0.4266, "step": 21111 }, { "epoch": 2.12, "learning_rate": 2.6630364388123354e-05, "loss": 0.2936, "step": 21112 }, { "epoch": 2.12, "learning_rate": 2.6627986373981626e-05, "loss": 0.5342, "step": 21113 }, { "epoch": 2.12, "learning_rate": 2.6625608381298112e-05, "loss": 0.1971, "step": 21114 }, { "epoch": 2.12, "learning_rate": 2.6623230410087933e-05, "loss": 0.8048, "step": 21115 }, { "epoch": 2.12, "learning_rate": 2.6620852460366235e-05, "loss": 0.2818, "step": 21116 }, { "epoch": 2.12, "learning_rate": 2.661847453214815e-05, "loss": 0.051, "step": 21117 }, { "epoch": 2.12, "learning_rate": 2.6616096625448796e-05, "loss": 0.3862, "step": 21118 }, { "epoch": 2.12, "learning_rate": 2.6613718740283335e-05, "loss": 0.1872, "step": 21119 }, { "epoch": 2.12, "learning_rate": 2.661134087666687e-05, "loss": 0.1025, "step": 21120 }, { "epoch": 2.12, "learning_rate": 2.660896303461453e-05, "loss": 0.7818, "step": 21121 }, { "epoch": 2.12, "learning_rate": 2.6606585214141475e-05, "loss": 0.2652, "step": 21122 }, { "epoch": 2.12, "learning_rate": 2.660420741526281e-05, "loss": 0.4631, "step": 21123 }, { "epoch": 2.12, "learning_rate": 2.660182963799369e-05, "loss": 0.712, "step": 21124 }, { "epoch": 2.12, "learning_rate": 2.6599451882349228e-05, "loss": 0.1662, "step": 21125 }, { "epoch": 2.12, "learning_rate": 2.6597074148344555e-05, "loss": 0.2568, "step": 21126 }, { "epoch": 2.12, "learning_rate": 2.659469643599481e-05, "loss": 0.5039, "step": 21127 }, { "epoch": 2.12, "learning_rate": 2.659231874531513e-05, "loss": 0.7051, "step": 21128 }, { "epoch": 2.12, "learning_rate": 2.6589941076320623e-05, "loss": 0.4561, "step": 21129 }, { "epoch": 2.12, "learning_rate": 2.6587563429026445e-05, "loss": 0.5658, "step": 21130 }, { "epoch": 2.12, "learning_rate": 2.658518580344771e-05, "loss": 0.1211, "step": 21131 }, { "epoch": 2.12, "learning_rate": 2.658280819959955e-05, "loss": 0.2967, "step": 21132 }, { "epoch": 2.12, "learning_rate": 2.658043061749711e-05, "loss": 0.2001, "step": 21133 }, { "epoch": 2.12, "learning_rate": 2.6578053057155503e-05, "loss": 0.2224, "step": 21134 }, { "epoch": 2.12, "learning_rate": 2.657567551858986e-05, "loss": 0.6143, "step": 21135 }, { "epoch": 2.12, "learning_rate": 2.657329800181532e-05, "loss": 0.4957, "step": 21136 }, { "epoch": 2.12, "learning_rate": 2.6570920506847014e-05, "loss": 0.4367, "step": 21137 }, { "epoch": 2.12, "learning_rate": 2.6568543033700052e-05, "loss": 0.742, "step": 21138 }, { "epoch": 2.12, "learning_rate": 2.6566165582389586e-05, "loss": 0.2207, "step": 21139 }, { "epoch": 2.12, "learning_rate": 2.6563788152930733e-05, "loss": 0.5517, "step": 21140 }, { "epoch": 2.12, "learning_rate": 2.656141074533862e-05, "loss": 0.2683, "step": 21141 }, { "epoch": 2.12, "learning_rate": 2.655903335962839e-05, "loss": 0.7638, "step": 21142 }, { "epoch": 2.12, "learning_rate": 2.655665599581516e-05, "loss": 0.0407, "step": 21143 }, { "epoch": 2.12, "learning_rate": 2.6554278653914057e-05, "loss": 0.0561, "step": 21144 }, { "epoch": 2.12, "learning_rate": 2.6551901333940217e-05, "loss": 0.7997, "step": 21145 }, { "epoch": 2.12, "learning_rate": 2.6549524035908772e-05, "loss": 0.3647, "step": 21146 }, { "epoch": 2.12, "learning_rate": 2.654714675983483e-05, "loss": 0.2542, "step": 21147 }, { "epoch": 2.12, "learning_rate": 2.6544769505733543e-05, "loss": 0.4728, "step": 21148 }, { "epoch": 2.12, "learning_rate": 2.654239227362002e-05, "loss": 0.3517, "step": 21149 }, { "epoch": 2.12, "learning_rate": 2.65400150635094e-05, "loss": 0.7611, "step": 21150 }, { "epoch": 2.12, "learning_rate": 2.6537637875416817e-05, "loss": 0.2577, "step": 21151 }, { "epoch": 2.12, "learning_rate": 2.6535260709357375e-05, "loss": 0.1492, "step": 21152 }, { "epoch": 2.12, "learning_rate": 2.6532883565346223e-05, "loss": 0.5688, "step": 21153 }, { "epoch": 2.12, "learning_rate": 2.653050644339848e-05, "loss": 0.5669, "step": 21154 }, { "epoch": 2.12, "learning_rate": 2.652812934352927e-05, "loss": 0.5839, "step": 21155 }, { "epoch": 2.12, "learning_rate": 2.6525752265753736e-05, "loss": 0.0889, "step": 21156 }, { "epoch": 2.12, "learning_rate": 2.652337521008699e-05, "loss": 0.6452, "step": 21157 }, { "epoch": 2.12, "learning_rate": 2.6520998176544147e-05, "loss": 0.664, "step": 21158 }, { "epoch": 2.12, "learning_rate": 2.6518621165140364e-05, "loss": 0.5599, "step": 21159 }, { "epoch": 2.12, "learning_rate": 2.651624417589075e-05, "loss": 0.156, "step": 21160 }, { "epoch": 2.12, "learning_rate": 2.6513867208810423e-05, "loss": 0.431, "step": 21161 }, { "epoch": 2.12, "learning_rate": 2.651149026391453e-05, "loss": 0.984, "step": 21162 }, { "epoch": 2.12, "learning_rate": 2.6509113341218175e-05, "loss": 0.0585, "step": 21163 }, { "epoch": 2.12, "learning_rate": 2.6506736440736498e-05, "loss": 0.731, "step": 21164 }, { "epoch": 2.12, "learning_rate": 2.650435956248463e-05, "loss": 0.2976, "step": 21165 }, { "epoch": 2.12, "learning_rate": 2.650198270647768e-05, "loss": 0.693, "step": 21166 }, { "epoch": 2.12, "learning_rate": 2.6499605872730778e-05, "loss": 0.5502, "step": 21167 }, { "epoch": 2.12, "learning_rate": 2.649722906125906e-05, "loss": 0.4042, "step": 21168 }, { "epoch": 2.12, "learning_rate": 2.649485227207765e-05, "loss": 0.2542, "step": 21169 }, { "epoch": 2.12, "learning_rate": 2.649247550520165e-05, "loss": 0.2026, "step": 21170 }, { "epoch": 2.12, "learning_rate": 2.6490098760646212e-05, "loss": 0.2135, "step": 21171 }, { "epoch": 2.12, "learning_rate": 2.648772203842645e-05, "loss": 0.2629, "step": 21172 }, { "epoch": 2.12, "learning_rate": 2.648534533855748e-05, "loss": 0.6523, "step": 21173 }, { "epoch": 2.12, "learning_rate": 2.648296866105445e-05, "loss": 0.4059, "step": 21174 }, { "epoch": 2.12, "learning_rate": 2.6480592005932463e-05, "loss": 0.4809, "step": 21175 }, { "epoch": 2.12, "learning_rate": 2.647821537320664e-05, "loss": 0.4317, "step": 21176 }, { "epoch": 2.12, "learning_rate": 2.6475838762892125e-05, "loss": 0.4184, "step": 21177 }, { "epoch": 2.12, "learning_rate": 2.647346217500402e-05, "loss": 0.7252, "step": 21178 }, { "epoch": 2.12, "learning_rate": 2.647108560955748e-05, "loss": 0.6001, "step": 21179 }, { "epoch": 2.13, "learning_rate": 2.64687090665676e-05, "loss": 0.1994, "step": 21180 }, { "epoch": 2.13, "learning_rate": 2.6466332546049502e-05, "loss": 0.226, "step": 21181 }, { "epoch": 2.13, "learning_rate": 2.6463956048018327e-05, "loss": 0.1808, "step": 21182 }, { "epoch": 2.13, "learning_rate": 2.64615795724892e-05, "loss": 0.0979, "step": 21183 }, { "epoch": 2.13, "learning_rate": 2.645920311947722e-05, "loss": 0.6973, "step": 21184 }, { "epoch": 2.13, "learning_rate": 2.645682668899753e-05, "loss": 0.293, "step": 21185 }, { "epoch": 2.13, "learning_rate": 2.6454450281065246e-05, "loss": 0.1437, "step": 21186 }, { "epoch": 2.13, "learning_rate": 2.6452073895695488e-05, "loss": 0.3908, "step": 21187 }, { "epoch": 2.13, "learning_rate": 2.6449697532903393e-05, "loss": 0.6698, "step": 21188 }, { "epoch": 2.13, "learning_rate": 2.6447321192704065e-05, "loss": 0.2162, "step": 21189 }, { "epoch": 2.13, "learning_rate": 2.644494487511263e-05, "loss": 0.3038, "step": 21190 }, { "epoch": 2.13, "learning_rate": 2.6442568580144218e-05, "loss": 0.6995, "step": 21191 }, { "epoch": 2.13, "learning_rate": 2.6440192307813954e-05, "loss": 0.2276, "step": 21192 }, { "epoch": 2.13, "learning_rate": 2.643781605813694e-05, "loss": 0.3537, "step": 21193 }, { "epoch": 2.13, "learning_rate": 2.6435439831128312e-05, "loss": 0.4745, "step": 21194 }, { "epoch": 2.13, "learning_rate": 2.6433063626803193e-05, "loss": 0.2945, "step": 21195 }, { "epoch": 2.13, "learning_rate": 2.643068744517669e-05, "loss": 0.5525, "step": 21196 }, { "epoch": 2.13, "learning_rate": 2.642831128626395e-05, "loss": 0.397, "step": 21197 }, { "epoch": 2.13, "learning_rate": 2.6425935150080072e-05, "loss": 0.3495, "step": 21198 }, { "epoch": 2.13, "learning_rate": 2.6423559036640175e-05, "loss": 0.2748, "step": 21199 }, { "epoch": 2.13, "learning_rate": 2.6421182945959393e-05, "loss": 0.3564, "step": 21200 }, { "epoch": 2.13, "learning_rate": 2.6418806878052847e-05, "loss": 0.3215, "step": 21201 }, { "epoch": 2.13, "learning_rate": 2.6416430832935645e-05, "loss": 0.2955, "step": 21202 }, { "epoch": 2.13, "learning_rate": 2.6414054810622913e-05, "loss": 0.5764, "step": 21203 }, { "epoch": 2.13, "learning_rate": 2.6411678811129775e-05, "loss": 0.442, "step": 21204 }, { "epoch": 2.13, "learning_rate": 2.6409302834471344e-05, "loss": 0.2449, "step": 21205 }, { "epoch": 2.13, "learning_rate": 2.640692688066275e-05, "loss": 0.4277, "step": 21206 }, { "epoch": 2.13, "learning_rate": 2.64045509497191e-05, "loss": 0.7799, "step": 21207 }, { "epoch": 2.13, "learning_rate": 2.6402175041655526e-05, "loss": 0.3938, "step": 21208 }, { "epoch": 2.13, "learning_rate": 2.639979915648714e-05, "loss": 0.4979, "step": 21209 }, { "epoch": 2.13, "learning_rate": 2.6397423294229056e-05, "loss": 0.1969, "step": 21210 }, { "epoch": 2.13, "learning_rate": 2.6395047454896414e-05, "loss": 0.4429, "step": 21211 }, { "epoch": 2.13, "learning_rate": 2.639267163850431e-05, "loss": 0.2633, "step": 21212 }, { "epoch": 2.13, "learning_rate": 2.6390295845067865e-05, "loss": 0.1848, "step": 21213 }, { "epoch": 2.13, "learning_rate": 2.6387920074602213e-05, "loss": 0.1529, "step": 21214 }, { "epoch": 2.13, "learning_rate": 2.6385544327122468e-05, "loss": 0.4184, "step": 21215 }, { "epoch": 2.13, "learning_rate": 2.638316860264373e-05, "loss": 0.6246, "step": 21216 }, { "epoch": 2.13, "learning_rate": 2.638079290118114e-05, "loss": 0.51, "step": 21217 }, { "epoch": 2.13, "learning_rate": 2.6378417222749805e-05, "loss": 0.1353, "step": 21218 }, { "epoch": 2.13, "learning_rate": 2.637604156736484e-05, "loss": 0.53, "step": 21219 }, { "epoch": 2.13, "learning_rate": 2.6373665935041377e-05, "loss": 0.4445, "step": 21220 }, { "epoch": 2.13, "learning_rate": 2.6371290325794522e-05, "loss": 0.2772, "step": 21221 }, { "epoch": 2.13, "learning_rate": 2.636891473963939e-05, "loss": 0.2692, "step": 21222 }, { "epoch": 2.13, "learning_rate": 2.6366539176591105e-05, "loss": 0.4212, "step": 21223 }, { "epoch": 2.13, "learning_rate": 2.6364163636664793e-05, "loss": 0.0684, "step": 21224 }, { "epoch": 2.13, "learning_rate": 2.6361788119875548e-05, "loss": 0.128, "step": 21225 }, { "epoch": 2.13, "learning_rate": 2.6359412626238502e-05, "loss": 0.3038, "step": 21226 }, { "epoch": 2.13, "learning_rate": 2.635703715576877e-05, "loss": 0.2954, "step": 21227 }, { "epoch": 2.13, "learning_rate": 2.635466170848146e-05, "loss": 0.6363, "step": 21228 }, { "epoch": 2.13, "learning_rate": 2.6352286284391712e-05, "loss": 0.656, "step": 21229 }, { "epoch": 2.13, "learning_rate": 2.634991088351462e-05, "loss": 0.166, "step": 21230 }, { "epoch": 2.13, "learning_rate": 2.6347535505865302e-05, "loss": 0.463, "step": 21231 }, { "epoch": 2.13, "learning_rate": 2.6345160151458876e-05, "loss": 0.165, "step": 21232 }, { "epoch": 2.13, "learning_rate": 2.6342784820310465e-05, "loss": 0.3098, "step": 21233 }, { "epoch": 2.13, "learning_rate": 2.6340409512435188e-05, "loss": 0.3563, "step": 21234 }, { "epoch": 2.13, "learning_rate": 2.6338034227848147e-05, "loss": 0.3733, "step": 21235 }, { "epoch": 2.13, "learning_rate": 2.6335658966564452e-05, "loss": 0.8269, "step": 21236 }, { "epoch": 2.13, "learning_rate": 2.6333283728599243e-05, "loss": 0.2822, "step": 21237 }, { "epoch": 2.13, "learning_rate": 2.6330908513967626e-05, "loss": 0.1456, "step": 21238 }, { "epoch": 2.13, "learning_rate": 2.6328533322684693e-05, "loss": 0.1812, "step": 21239 }, { "epoch": 2.13, "learning_rate": 2.6326158154765588e-05, "loss": 0.5932, "step": 21240 }, { "epoch": 2.13, "learning_rate": 2.6323783010225416e-05, "loss": 0.3323, "step": 21241 }, { "epoch": 2.13, "learning_rate": 2.6321407889079284e-05, "loss": 1.237, "step": 21242 }, { "epoch": 2.13, "learning_rate": 2.6319032791342324e-05, "loss": 0.239, "step": 21243 }, { "epoch": 2.13, "learning_rate": 2.6316657717029633e-05, "loss": 0.9699, "step": 21244 }, { "epoch": 2.13, "learning_rate": 2.6314282666156326e-05, "loss": 0.4157, "step": 21245 }, { "epoch": 2.13, "learning_rate": 2.6311907638737527e-05, "loss": 0.1381, "step": 21246 }, { "epoch": 2.13, "learning_rate": 2.630953263478835e-05, "loss": 0.1394, "step": 21247 }, { "epoch": 2.13, "learning_rate": 2.6307157654323892e-05, "loss": 0.121, "step": 21248 }, { "epoch": 2.13, "learning_rate": 2.6304782697359288e-05, "loss": 0.4304, "step": 21249 }, { "epoch": 2.13, "learning_rate": 2.6302407763909637e-05, "loss": 0.2052, "step": 21250 }, { "epoch": 2.13, "learning_rate": 2.6300032853990045e-05, "loss": 0.2622, "step": 21251 }, { "epoch": 2.13, "learning_rate": 2.6297657967615656e-05, "loss": 0.2128, "step": 21252 }, { "epoch": 2.13, "learning_rate": 2.6295283104801556e-05, "loss": 0.4398, "step": 21253 }, { "epoch": 2.13, "learning_rate": 2.629290826556286e-05, "loss": 0.2351, "step": 21254 }, { "epoch": 2.13, "learning_rate": 2.6290533449914688e-05, "loss": 0.5559, "step": 21255 }, { "epoch": 2.13, "learning_rate": 2.628815865787216e-05, "loss": 0.3078, "step": 21256 }, { "epoch": 2.13, "learning_rate": 2.628578388945036e-05, "loss": 0.1816, "step": 21257 }, { "epoch": 2.13, "learning_rate": 2.6283409144664433e-05, "loss": 0.4419, "step": 21258 }, { "epoch": 2.13, "learning_rate": 2.6281034423529465e-05, "loss": 0.2429, "step": 21259 }, { "epoch": 2.13, "learning_rate": 2.6278659726060587e-05, "loss": 0.5999, "step": 21260 }, { "epoch": 2.13, "learning_rate": 2.6276285052272904e-05, "loss": 0.1885, "step": 21261 }, { "epoch": 2.13, "learning_rate": 2.627391040218152e-05, "loss": 0.1763, "step": 21262 }, { "epoch": 2.13, "learning_rate": 2.6271535775801556e-05, "loss": 0.312, "step": 21263 }, { "epoch": 2.13, "learning_rate": 2.626916117314812e-05, "loss": 1.5754, "step": 21264 }, { "epoch": 2.13, "learning_rate": 2.626678659423632e-05, "loss": 0.5064, "step": 21265 }, { "epoch": 2.13, "learning_rate": 2.626441203908128e-05, "loss": 0.2316, "step": 21266 }, { "epoch": 2.13, "learning_rate": 2.6262037507698092e-05, "loss": 0.3559, "step": 21267 }, { "epoch": 2.13, "learning_rate": 2.6259663000101872e-05, "loss": 0.811, "step": 21268 }, { "epoch": 2.13, "learning_rate": 2.6257288516307738e-05, "loss": 0.1236, "step": 21269 }, { "epoch": 2.13, "learning_rate": 2.6254914056330805e-05, "loss": 0.4536, "step": 21270 }, { "epoch": 2.13, "learning_rate": 2.6252539620186163e-05, "loss": 0.478, "step": 21271 }, { "epoch": 2.13, "learning_rate": 2.6250165207888935e-05, "loss": 0.3711, "step": 21272 }, { "epoch": 2.13, "learning_rate": 2.6247790819454228e-05, "loss": 0.3348, "step": 21273 }, { "epoch": 2.13, "learning_rate": 2.6245416454897152e-05, "loss": 0.5881, "step": 21274 }, { "epoch": 2.13, "learning_rate": 2.624304211423283e-05, "loss": 0.8378, "step": 21275 }, { "epoch": 2.13, "learning_rate": 2.624066779747635e-05, "loss": 0.3786, "step": 21276 }, { "epoch": 2.13, "learning_rate": 2.6238293504642823e-05, "loss": 0.3241, "step": 21277 }, { "epoch": 2.14, "learning_rate": 2.623591923574737e-05, "loss": 0.7402, "step": 21278 }, { "epoch": 2.14, "learning_rate": 2.6233544990805107e-05, "loss": 0.6979, "step": 21279 }, { "epoch": 2.14, "learning_rate": 2.6231170769831114e-05, "loss": 0.8364, "step": 21280 }, { "epoch": 2.14, "learning_rate": 2.6228796572840523e-05, "loss": 0.1419, "step": 21281 }, { "epoch": 2.14, "learning_rate": 2.6226422399848435e-05, "loss": 0.2784, "step": 21282 }, { "epoch": 2.14, "learning_rate": 2.6224048250869956e-05, "loss": 0.5729, "step": 21283 }, { "epoch": 2.14, "learning_rate": 2.6221674125920208e-05, "loss": 0.1145, "step": 21284 }, { "epoch": 2.14, "learning_rate": 2.6219300025014284e-05, "loss": 0.2862, "step": 21285 }, { "epoch": 2.14, "learning_rate": 2.621692594816729e-05, "loss": 0.2051, "step": 21286 }, { "epoch": 2.14, "learning_rate": 2.621455189539434e-05, "loss": 0.1404, "step": 21287 }, { "epoch": 2.14, "learning_rate": 2.6212177866710546e-05, "loss": 0.3593, "step": 21288 }, { "epoch": 2.14, "learning_rate": 2.6209803862131017e-05, "loss": 0.2277, "step": 21289 }, { "epoch": 2.14, "learning_rate": 2.620742988167085e-05, "loss": 0.4839, "step": 21290 }, { "epoch": 2.14, "learning_rate": 2.620505592534515e-05, "loss": 0.3923, "step": 21291 }, { "epoch": 2.14, "learning_rate": 2.6202681993169038e-05, "loss": 0.5644, "step": 21292 }, { "epoch": 2.14, "learning_rate": 2.620030808515762e-05, "loss": 0.0601, "step": 21293 }, { "epoch": 2.14, "learning_rate": 2.6197934201325977e-05, "loss": 0.1726, "step": 21294 }, { "epoch": 2.14, "learning_rate": 2.6195560341689246e-05, "loss": 0.0936, "step": 21295 }, { "epoch": 2.14, "learning_rate": 2.619318650626252e-05, "loss": 0.4981, "step": 21296 }, { "epoch": 2.14, "learning_rate": 2.61908126950609e-05, "loss": 0.2918, "step": 21297 }, { "epoch": 2.14, "learning_rate": 2.6188438908099512e-05, "loss": 0.3877, "step": 21298 }, { "epoch": 2.14, "learning_rate": 2.6186065145393443e-05, "loss": 0.2102, "step": 21299 }, { "epoch": 2.14, "learning_rate": 2.6183691406957803e-05, "loss": 0.5823, "step": 21300 }, { "epoch": 2.14, "learning_rate": 2.6181317692807695e-05, "loss": 0.2404, "step": 21301 }, { "epoch": 2.14, "learning_rate": 2.6178944002958244e-05, "loss": 0.3091, "step": 21302 }, { "epoch": 2.14, "learning_rate": 2.6176570337424523e-05, "loss": 0.7093, "step": 21303 }, { "epoch": 2.14, "learning_rate": 2.6174196696221658e-05, "loss": 0.2435, "step": 21304 }, { "epoch": 2.14, "learning_rate": 2.6171823079364756e-05, "loss": 1.2553, "step": 21305 }, { "epoch": 2.14, "learning_rate": 2.6169449486868905e-05, "loss": 0.4059, "step": 21306 }, { "epoch": 2.14, "learning_rate": 2.616707591874924e-05, "loss": 0.1153, "step": 21307 }, { "epoch": 2.14, "learning_rate": 2.6164702375020828e-05, "loss": 0.0633, "step": 21308 }, { "epoch": 2.14, "learning_rate": 2.616232885569879e-05, "loss": 0.4014, "step": 21309 }, { "epoch": 2.14, "learning_rate": 2.6159955360798238e-05, "loss": 0.6169, "step": 21310 }, { "epoch": 2.14, "learning_rate": 2.6157581890334274e-05, "loss": 0.2181, "step": 21311 }, { "epoch": 2.14, "learning_rate": 2.6155208444321986e-05, "loss": 0.2107, "step": 21312 }, { "epoch": 2.14, "learning_rate": 2.6152835022776496e-05, "loss": 0.5917, "step": 21313 }, { "epoch": 2.14, "learning_rate": 2.6150461625712894e-05, "loss": 0.0711, "step": 21314 }, { "epoch": 2.14, "learning_rate": 2.6148088253146295e-05, "loss": 0.6669, "step": 21315 }, { "epoch": 2.14, "learning_rate": 2.6145714905091802e-05, "loss": 0.2297, "step": 21316 }, { "epoch": 2.14, "learning_rate": 2.6143341581564497e-05, "loss": 0.2008, "step": 21317 }, { "epoch": 2.14, "learning_rate": 2.6140968282579517e-05, "loss": 0.1125, "step": 21318 }, { "epoch": 2.14, "learning_rate": 2.613859500815194e-05, "loss": 0.4532, "step": 21319 }, { "epoch": 2.14, "learning_rate": 2.6136221758296864e-05, "loss": 0.4599, "step": 21320 }, { "epoch": 2.14, "learning_rate": 2.6133848533029423e-05, "loss": 0.3397, "step": 21321 }, { "epoch": 2.14, "learning_rate": 2.613147533236469e-05, "loss": 0.5083, "step": 21322 }, { "epoch": 2.14, "learning_rate": 2.612910215631777e-05, "loss": 0.9131, "step": 21323 }, { "epoch": 2.14, "learning_rate": 2.6126729004903776e-05, "loss": 0.567, "step": 21324 }, { "epoch": 2.14, "learning_rate": 2.6124355878137817e-05, "loss": 0.6312, "step": 21325 }, { "epoch": 2.14, "learning_rate": 2.6121982776034967e-05, "loss": 0.1953, "step": 21326 }, { "epoch": 2.14, "learning_rate": 2.6119609698610347e-05, "loss": 0.7569, "step": 21327 }, { "epoch": 2.14, "learning_rate": 2.6117236645879058e-05, "loss": 0.2599, "step": 21328 }, { "epoch": 2.14, "learning_rate": 2.611486361785619e-05, "loss": 0.1727, "step": 21329 }, { "epoch": 2.14, "learning_rate": 2.6112490614556868e-05, "loss": 0.6173, "step": 21330 }, { "epoch": 2.14, "learning_rate": 2.6110117635996168e-05, "loss": 0.2365, "step": 21331 }, { "epoch": 2.14, "learning_rate": 2.610774468218919e-05, "loss": 0.4085, "step": 21332 }, { "epoch": 2.14, "learning_rate": 2.6105371753151054e-05, "loss": 0.2561, "step": 21333 }, { "epoch": 2.14, "learning_rate": 2.610299884889686e-05, "loss": 0.1927, "step": 21334 }, { "epoch": 2.14, "learning_rate": 2.6100625969441677e-05, "loss": 0.3392, "step": 21335 }, { "epoch": 2.14, "learning_rate": 2.6098253114800644e-05, "loss": 0.2075, "step": 21336 }, { "epoch": 2.14, "learning_rate": 2.6095880284988837e-05, "loss": 0.3527, "step": 21337 }, { "epoch": 2.14, "learning_rate": 2.6093507480021356e-05, "loss": 1.3785, "step": 21338 }, { "epoch": 2.14, "learning_rate": 2.609113469991332e-05, "loss": 0.9336, "step": 21339 }, { "epoch": 2.14, "learning_rate": 2.6088761944679803e-05, "loss": 0.3333, "step": 21340 }, { "epoch": 2.14, "learning_rate": 2.608638921433593e-05, "loss": 0.3983, "step": 21341 }, { "epoch": 2.14, "learning_rate": 2.608401650889678e-05, "loss": 0.3655, "step": 21342 }, { "epoch": 2.14, "learning_rate": 2.608164382837746e-05, "loss": 0.3454, "step": 21343 }, { "epoch": 2.14, "learning_rate": 2.6079271172793078e-05, "loss": 0.4739, "step": 21344 }, { "epoch": 2.14, "learning_rate": 2.607689854215871e-05, "loss": 0.4625, "step": 21345 }, { "epoch": 2.14, "learning_rate": 2.6074525936489466e-05, "loss": 0.1452, "step": 21346 }, { "epoch": 2.14, "learning_rate": 2.607215335580045e-05, "loss": 0.1768, "step": 21347 }, { "epoch": 2.14, "learning_rate": 2.6069780800106767e-05, "loss": 0.7179, "step": 21348 }, { "epoch": 2.14, "learning_rate": 2.6067408269423483e-05, "loss": 0.6391, "step": 21349 }, { "epoch": 2.14, "learning_rate": 2.6065035763765734e-05, "loss": 0.2479, "step": 21350 }, { "epoch": 2.14, "learning_rate": 2.6062663283148592e-05, "loss": 0.4085, "step": 21351 }, { "epoch": 2.14, "learning_rate": 2.6060290827587162e-05, "loss": 0.4142, "step": 21352 }, { "epoch": 2.14, "learning_rate": 2.605791839709655e-05, "loss": 0.8009, "step": 21353 }, { "epoch": 2.14, "learning_rate": 2.6055545991691844e-05, "loss": 0.0885, "step": 21354 }, { "epoch": 2.14, "learning_rate": 2.605317361138813e-05, "loss": 0.4332, "step": 21355 }, { "epoch": 2.14, "learning_rate": 2.605080125620053e-05, "loss": 0.5553, "step": 21356 }, { "epoch": 2.14, "learning_rate": 2.6048428926144133e-05, "loss": 0.2018, "step": 21357 }, { "epoch": 2.14, "learning_rate": 2.604605662123402e-05, "loss": 0.3116, "step": 21358 }, { "epoch": 2.14, "learning_rate": 2.6043684341485302e-05, "loss": 0.5093, "step": 21359 }, { "epoch": 2.14, "learning_rate": 2.6041312086913072e-05, "loss": 0.2185, "step": 21360 }, { "epoch": 2.14, "learning_rate": 2.603893985753242e-05, "loss": 0.1023, "step": 21361 }, { "epoch": 2.14, "learning_rate": 2.6036567653358465e-05, "loss": 0.1157, "step": 21362 }, { "epoch": 2.14, "learning_rate": 2.6034195474406273e-05, "loss": 0.1634, "step": 21363 }, { "epoch": 2.14, "learning_rate": 2.603182332069095e-05, "loss": 0.5507, "step": 21364 }, { "epoch": 2.14, "learning_rate": 2.6029451192227596e-05, "loss": 0.3847, "step": 21365 }, { "epoch": 2.14, "learning_rate": 2.6027079089031314e-05, "loss": 0.4681, "step": 21366 }, { "epoch": 2.14, "learning_rate": 2.6024707011117173e-05, "loss": 0.1361, "step": 21367 }, { "epoch": 2.14, "learning_rate": 2.6022334958500297e-05, "loss": 0.3142, "step": 21368 }, { "epoch": 2.14, "learning_rate": 2.6019962931195758e-05, "loss": 0.3463, "step": 21369 }, { "epoch": 2.14, "learning_rate": 2.6017590929218663e-05, "loss": 0.5569, "step": 21370 }, { "epoch": 2.14, "learning_rate": 2.6015218952584115e-05, "loss": 0.1586, "step": 21371 }, { "epoch": 2.14, "learning_rate": 2.601284700130718e-05, "loss": 0.3496, "step": 21372 }, { "epoch": 2.14, "learning_rate": 2.6010475075402988e-05, "loss": 0.2671, "step": 21373 }, { "epoch": 2.14, "learning_rate": 2.600810317488661e-05, "loss": 0.2839, "step": 21374 }, { "epoch": 2.14, "learning_rate": 2.600573129977313e-05, "loss": 0.3486, "step": 21375 }, { "epoch": 2.15, "learning_rate": 2.6003359450077682e-05, "loss": 0.627, "step": 21376 }, { "epoch": 2.15, "learning_rate": 2.600098762581532e-05, "loss": 0.3387, "step": 21377 }, { "epoch": 2.15, "learning_rate": 2.5998615827001146e-05, "loss": 0.3916, "step": 21378 }, { "epoch": 2.15, "learning_rate": 2.5996244053650264e-05, "loss": 0.9744, "step": 21379 }, { "epoch": 2.15, "learning_rate": 2.599387230577777e-05, "loss": 0.5206, "step": 21380 }, { "epoch": 2.15, "learning_rate": 2.599150058339873e-05, "loss": 0.1392, "step": 21381 }, { "epoch": 2.15, "learning_rate": 2.5989128886528272e-05, "loss": 0.5679, "step": 21382 }, { "epoch": 2.15, "learning_rate": 2.598675721518147e-05, "loss": 0.2861, "step": 21383 }, { "epoch": 2.15, "learning_rate": 2.598438556937341e-05, "loss": 0.1371, "step": 21384 }, { "epoch": 2.15, "learning_rate": 2.5982013949119203e-05, "loss": 0.6086, "step": 21385 }, { "epoch": 2.15, "learning_rate": 2.597964235443393e-05, "loss": 0.0623, "step": 21386 }, { "epoch": 2.15, "learning_rate": 2.5977270785332666e-05, "loss": 0.1559, "step": 21387 }, { "epoch": 2.15, "learning_rate": 2.5974899241830537e-05, "loss": 0.2371, "step": 21388 }, { "epoch": 2.15, "learning_rate": 2.5972527723942624e-05, "loss": 0.1565, "step": 21389 }, { "epoch": 2.15, "learning_rate": 2.5970156231683996e-05, "loss": 0.4139, "step": 21390 }, { "epoch": 2.15, "learning_rate": 2.5967784765069765e-05, "loss": 0.2222, "step": 21391 }, { "epoch": 2.15, "learning_rate": 2.5965413324115022e-05, "loss": 0.6988, "step": 21392 }, { "epoch": 2.15, "learning_rate": 2.5963041908834847e-05, "loss": 0.48, "step": 21393 }, { "epoch": 2.15, "learning_rate": 2.596067051924435e-05, "loss": 0.2991, "step": 21394 }, { "epoch": 2.15, "learning_rate": 2.5958299155358588e-05, "loss": 0.6752, "step": 21395 }, { "epoch": 2.15, "learning_rate": 2.5955927817192695e-05, "loss": 0.1547, "step": 21396 }, { "epoch": 2.15, "learning_rate": 2.595355650476173e-05, "loss": 0.2232, "step": 21397 }, { "epoch": 2.15, "learning_rate": 2.5951185218080788e-05, "loss": 0.1647, "step": 21398 }, { "epoch": 2.15, "learning_rate": 2.5948813957164974e-05, "loss": 0.5347, "step": 21399 }, { "epoch": 2.15, "learning_rate": 2.594644272202936e-05, "loss": 0.1392, "step": 21400 }, { "epoch": 2.15, "learning_rate": 2.5944071512689038e-05, "loss": 0.2794, "step": 21401 }, { "epoch": 2.15, "learning_rate": 2.5941700329159107e-05, "loss": 0.0733, "step": 21402 }, { "epoch": 2.15, "learning_rate": 2.5939329171454658e-05, "loss": 0.4104, "step": 21403 }, { "epoch": 2.15, "learning_rate": 2.5936958039590755e-05, "loss": 0.1701, "step": 21404 }, { "epoch": 2.15, "learning_rate": 2.5934586933582526e-05, "loss": 0.1912, "step": 21405 }, { "epoch": 2.15, "learning_rate": 2.5932215853445027e-05, "loss": 0.2109, "step": 21406 }, { "epoch": 2.15, "learning_rate": 2.5929844799193357e-05, "loss": 0.2491, "step": 21407 }, { "epoch": 2.15, "learning_rate": 2.5927473770842616e-05, "loss": 0.7172, "step": 21408 }, { "epoch": 2.15, "learning_rate": 2.592510276840788e-05, "loss": 1.3548, "step": 21409 }, { "epoch": 2.15, "learning_rate": 2.592273179190423e-05, "loss": 0.4074, "step": 21410 }, { "epoch": 2.15, "learning_rate": 2.592036084134677e-05, "loss": 0.4105, "step": 21411 }, { "epoch": 2.15, "learning_rate": 2.5917989916750594e-05, "loss": 0.2647, "step": 21412 }, { "epoch": 2.15, "learning_rate": 2.5915619018130752e-05, "loss": 0.1111, "step": 21413 }, { "epoch": 2.15, "learning_rate": 2.591324814550238e-05, "loss": 0.2681, "step": 21414 }, { "epoch": 2.15, "learning_rate": 2.5910877298880532e-05, "loss": 0.4784, "step": 21415 }, { "epoch": 2.15, "learning_rate": 2.5908506478280302e-05, "loss": 0.5444, "step": 21416 }, { "epoch": 2.15, "learning_rate": 2.5906135683716794e-05, "loss": 0.1811, "step": 21417 }, { "epoch": 2.15, "learning_rate": 2.5903764915205072e-05, "loss": 0.3342, "step": 21418 }, { "epoch": 2.15, "learning_rate": 2.5901394172760226e-05, "loss": 0.4511, "step": 21419 }, { "epoch": 2.15, "learning_rate": 2.5899023456397354e-05, "loss": 0.9014, "step": 21420 }, { "epoch": 2.15, "learning_rate": 2.5896652766131527e-05, "loss": 0.9484, "step": 21421 }, { "epoch": 2.15, "learning_rate": 2.5894282101977862e-05, "loss": 0.2732, "step": 21422 }, { "epoch": 2.15, "learning_rate": 2.589191146395141e-05, "loss": 0.5971, "step": 21423 }, { "epoch": 2.15, "learning_rate": 2.588954085206727e-05, "loss": 0.2901, "step": 21424 }, { "epoch": 2.15, "learning_rate": 2.588717026634053e-05, "loss": 0.2133, "step": 21425 }, { "epoch": 2.15, "learning_rate": 2.5884799706786284e-05, "loss": 0.7697, "step": 21426 }, { "epoch": 2.15, "learning_rate": 2.5882429173419585e-05, "loss": 0.5656, "step": 21427 }, { "epoch": 2.15, "learning_rate": 2.5880058666255562e-05, "loss": 0.1633, "step": 21428 }, { "epoch": 2.15, "learning_rate": 2.587768818530927e-05, "loss": 0.2239, "step": 21429 }, { "epoch": 2.15, "learning_rate": 2.5875317730595796e-05, "loss": 0.4158, "step": 21430 }, { "epoch": 2.15, "learning_rate": 2.587294730213024e-05, "loss": 0.5567, "step": 21431 }, { "epoch": 2.15, "learning_rate": 2.587057689992768e-05, "loss": 0.5015, "step": 21432 }, { "epoch": 2.15, "learning_rate": 2.586820652400318e-05, "loss": 0.1523, "step": 21433 }, { "epoch": 2.15, "learning_rate": 2.586583617437186e-05, "loss": 0.323, "step": 21434 }, { "epoch": 2.15, "learning_rate": 2.5863465851048786e-05, "loss": 0.4845, "step": 21435 }, { "epoch": 2.15, "learning_rate": 2.5861095554049023e-05, "loss": 0.2897, "step": 21436 }, { "epoch": 2.15, "learning_rate": 2.5858725283387692e-05, "loss": 0.1478, "step": 21437 }, { "epoch": 2.15, "learning_rate": 2.5856355039079853e-05, "loss": 0.459, "step": 21438 }, { "epoch": 2.15, "learning_rate": 2.5853984821140586e-05, "loss": 0.2252, "step": 21439 }, { "epoch": 2.15, "learning_rate": 2.5851614629584997e-05, "loss": 0.3911, "step": 21440 }, { "epoch": 2.15, "learning_rate": 2.5849244464428145e-05, "loss": 0.1979, "step": 21441 }, { "epoch": 2.15, "learning_rate": 2.584687432568512e-05, "loss": 0.4562, "step": 21442 }, { "epoch": 2.15, "learning_rate": 2.5844504213371012e-05, "loss": 0.7047, "step": 21443 }, { "epoch": 2.15, "learning_rate": 2.5842134127500902e-05, "loss": 0.1885, "step": 21444 }, { "epoch": 2.15, "learning_rate": 2.583976406808985e-05, "loss": 0.6435, "step": 21445 }, { "epoch": 2.15, "learning_rate": 2.583739403515298e-05, "loss": 0.0819, "step": 21446 }, { "epoch": 2.15, "learning_rate": 2.583502402870534e-05, "loss": 0.3122, "step": 21447 }, { "epoch": 2.15, "learning_rate": 2.583265404876202e-05, "loss": 0.4918, "step": 21448 }, { "epoch": 2.15, "learning_rate": 2.5830284095338114e-05, "loss": 0.2628, "step": 21449 }, { "epoch": 2.15, "learning_rate": 2.582791416844868e-05, "loss": 0.4094, "step": 21450 }, { "epoch": 2.15, "learning_rate": 2.5825544268108827e-05, "loss": 0.2063, "step": 21451 }, { "epoch": 2.15, "learning_rate": 2.5823174394333616e-05, "loss": 0.1904, "step": 21452 }, { "epoch": 2.15, "learning_rate": 2.5820804547138126e-05, "loss": 0.7192, "step": 21453 }, { "epoch": 2.15, "learning_rate": 2.5818434726537462e-05, "loss": 0.1846, "step": 21454 }, { "epoch": 2.15, "learning_rate": 2.5816064932546683e-05, "loss": 0.9362, "step": 21455 }, { "epoch": 2.15, "learning_rate": 2.5813695165180867e-05, "loss": 0.4172, "step": 21456 }, { "epoch": 2.15, "learning_rate": 2.581132542445511e-05, "loss": 0.1486, "step": 21457 }, { "epoch": 2.15, "learning_rate": 2.580895571038449e-05, "loss": 0.5403, "step": 21458 }, { "epoch": 2.15, "learning_rate": 2.5806586022984062e-05, "loss": 0.4722, "step": 21459 }, { "epoch": 2.15, "learning_rate": 2.5804216362268947e-05, "loss": 1.2268, "step": 21460 }, { "epoch": 2.15, "learning_rate": 2.5801846728254195e-05, "loss": 0.5333, "step": 21461 }, { "epoch": 2.15, "learning_rate": 2.5799477120954885e-05, "loss": 0.1877, "step": 21462 }, { "epoch": 2.15, "learning_rate": 2.579710754038612e-05, "loss": 0.2289, "step": 21463 }, { "epoch": 2.15, "learning_rate": 2.5794737986562957e-05, "loss": 0.1141, "step": 21464 }, { "epoch": 2.15, "learning_rate": 2.5792368459500476e-05, "loss": 0.1549, "step": 21465 }, { "epoch": 2.15, "learning_rate": 2.5789998959213764e-05, "loss": 0.3664, "step": 21466 }, { "epoch": 2.15, "learning_rate": 2.5787629485717907e-05, "loss": 0.1932, "step": 21467 }, { "epoch": 2.15, "learning_rate": 2.5785260039027954e-05, "loss": 0.5648, "step": 21468 }, { "epoch": 2.15, "learning_rate": 2.578289061915902e-05, "loss": 0.276, "step": 21469 }, { "epoch": 2.15, "learning_rate": 2.5780521226126158e-05, "loss": 0.0904, "step": 21470 }, { "epoch": 2.15, "learning_rate": 2.5778151859944453e-05, "loss": 0.6706, "step": 21471 }, { "epoch": 2.15, "learning_rate": 2.5775782520628993e-05, "loss": 0.1279, "step": 21472 }, { "epoch": 2.15, "learning_rate": 2.577341320819484e-05, "loss": 0.3953, "step": 21473 }, { "epoch": 2.15, "learning_rate": 2.577104392265707e-05, "loss": 0.2539, "step": 21474 }, { "epoch": 2.16, "learning_rate": 2.5768674664030776e-05, "loss": 0.3978, "step": 21475 }, { "epoch": 2.16, "learning_rate": 2.576630543233102e-05, "loss": 0.0784, "step": 21476 }, { "epoch": 2.16, "learning_rate": 2.5763936227572892e-05, "loss": 0.4323, "step": 21477 }, { "epoch": 2.16, "learning_rate": 2.5761567049771467e-05, "loss": 0.0387, "step": 21478 }, { "epoch": 2.16, "learning_rate": 2.5759197898941804e-05, "loss": 0.6861, "step": 21479 }, { "epoch": 2.16, "learning_rate": 2.5756828775099003e-05, "loss": 0.4183, "step": 21480 }, { "epoch": 2.16, "learning_rate": 2.5754459678258133e-05, "loss": 0.1033, "step": 21481 }, { "epoch": 2.16, "learning_rate": 2.575209060843425e-05, "loss": 0.596, "step": 21482 }, { "epoch": 2.16, "learning_rate": 2.574972156564246e-05, "loss": 0.1822, "step": 21483 }, { "epoch": 2.16, "learning_rate": 2.574735254989782e-05, "loss": 0.1501, "step": 21484 }, { "epoch": 2.16, "learning_rate": 2.5744983561215406e-05, "loss": 0.3084, "step": 21485 }, { "epoch": 2.16, "learning_rate": 2.5742614599610312e-05, "loss": 0.3955, "step": 21486 }, { "epoch": 2.16, "learning_rate": 2.5740245665097593e-05, "loss": 0.5458, "step": 21487 }, { "epoch": 2.16, "learning_rate": 2.5737876757692317e-05, "loss": 0.3671, "step": 21488 }, { "epoch": 2.16, "learning_rate": 2.5735507877409585e-05, "loss": 0.3446, "step": 21489 }, { "epoch": 2.16, "learning_rate": 2.5733139024264464e-05, "loss": 0.2931, "step": 21490 }, { "epoch": 2.16, "learning_rate": 2.5730770198272004e-05, "loss": 0.7002, "step": 21491 }, { "epoch": 2.16, "learning_rate": 2.5728401399447316e-05, "loss": 0.8995, "step": 21492 }, { "epoch": 2.16, "learning_rate": 2.572603262780545e-05, "loss": 0.2173, "step": 21493 }, { "epoch": 2.16, "learning_rate": 2.572366388336148e-05, "loss": 0.1656, "step": 21494 }, { "epoch": 2.16, "learning_rate": 2.5721295166130503e-05, "loss": 0.1739, "step": 21495 }, { "epoch": 2.16, "learning_rate": 2.5718926476127563e-05, "loss": 0.2119, "step": 21496 }, { "epoch": 2.16, "learning_rate": 2.5716557813367744e-05, "loss": 0.5195, "step": 21497 }, { "epoch": 2.16, "learning_rate": 2.5714189177866126e-05, "loss": 0.2254, "step": 21498 }, { "epoch": 2.16, "learning_rate": 2.5711820569637787e-05, "loss": 0.6809, "step": 21499 }, { "epoch": 2.16, "learning_rate": 2.570945198869777e-05, "loss": 0.2917, "step": 21500 }, { "epoch": 2.16, "learning_rate": 2.5707083435061187e-05, "loss": 0.1937, "step": 21501 }, { "epoch": 2.16, "learning_rate": 2.5704714908743085e-05, "loss": 0.6266, "step": 21502 }, { "epoch": 2.16, "learning_rate": 2.570234640975854e-05, "loss": 0.456, "step": 21503 }, { "epoch": 2.16, "learning_rate": 2.5699977938122635e-05, "loss": 0.4429, "step": 21504 }, { "epoch": 2.16, "learning_rate": 2.5697609493850418e-05, "loss": 0.4441, "step": 21505 }, { "epoch": 2.16, "learning_rate": 2.5695241076957e-05, "loss": 0.7045, "step": 21506 }, { "epoch": 2.16, "learning_rate": 2.5692872687457418e-05, "loss": 0.4164, "step": 21507 }, { "epoch": 2.16, "learning_rate": 2.569050432536675e-05, "loss": 0.8036, "step": 21508 }, { "epoch": 2.16, "learning_rate": 2.5688135990700084e-05, "loss": 0.5242, "step": 21509 }, { "epoch": 2.16, "learning_rate": 2.5685767683472485e-05, "loss": 0.5058, "step": 21510 }, { "epoch": 2.16, "learning_rate": 2.5683399403699002e-05, "loss": 0.1317, "step": 21511 }, { "epoch": 2.16, "learning_rate": 2.5681031151394734e-05, "loss": 0.4207, "step": 21512 }, { "epoch": 2.16, "learning_rate": 2.567866292657475e-05, "loss": 0.4377, "step": 21513 }, { "epoch": 2.16, "learning_rate": 2.5676294729254087e-05, "loss": 0.2875, "step": 21514 }, { "epoch": 2.16, "learning_rate": 2.5673926559447856e-05, "loss": 0.3181, "step": 21515 }, { "epoch": 2.16, "learning_rate": 2.567155841717111e-05, "loss": 0.2679, "step": 21516 }, { "epoch": 2.16, "learning_rate": 2.5669190302438912e-05, "loss": 0.2281, "step": 21517 }, { "epoch": 2.16, "learning_rate": 2.566682221526635e-05, "loss": 0.3027, "step": 21518 }, { "epoch": 2.16, "learning_rate": 2.566445415566848e-05, "loss": 0.3163, "step": 21519 }, { "epoch": 2.16, "learning_rate": 2.5662086123660364e-05, "loss": 0.935, "step": 21520 }, { "epoch": 2.16, "learning_rate": 2.5659718119257093e-05, "loss": 0.5022, "step": 21521 }, { "epoch": 2.16, "learning_rate": 2.565735014247373e-05, "loss": 0.2627, "step": 21522 }, { "epoch": 2.16, "learning_rate": 2.5654982193325314e-05, "loss": 0.6678, "step": 21523 }, { "epoch": 2.16, "learning_rate": 2.5652614271826965e-05, "loss": 0.3204, "step": 21524 }, { "epoch": 2.16, "learning_rate": 2.5650246377993714e-05, "loss": 0.7809, "step": 21525 }, { "epoch": 2.16, "learning_rate": 2.5647878511840633e-05, "loss": 0.2431, "step": 21526 }, { "epoch": 2.16, "learning_rate": 2.5645510673382815e-05, "loss": 0.3135, "step": 21527 }, { "epoch": 2.16, "learning_rate": 2.5643142862635298e-05, "loss": 0.1305, "step": 21528 }, { "epoch": 2.16, "learning_rate": 2.564077507961316e-05, "loss": 0.308, "step": 21529 }, { "epoch": 2.16, "learning_rate": 2.5638407324331478e-05, "loss": 0.6117, "step": 21530 }, { "epoch": 2.16, "learning_rate": 2.5636039596805303e-05, "loss": 0.4118, "step": 21531 }, { "epoch": 2.16, "learning_rate": 2.563367189704972e-05, "loss": 0.1065, "step": 21532 }, { "epoch": 2.16, "learning_rate": 2.5631304225079796e-05, "loss": 0.6975, "step": 21533 }, { "epoch": 2.16, "learning_rate": 2.5628936580910577e-05, "loss": 0.2003, "step": 21534 }, { "epoch": 2.16, "learning_rate": 2.5626568964557148e-05, "loss": 0.7421, "step": 21535 }, { "epoch": 2.16, "learning_rate": 2.562420137603458e-05, "loss": 0.1968, "step": 21536 }, { "epoch": 2.16, "learning_rate": 2.5621833815357902e-05, "loss": 0.4481, "step": 21537 }, { "epoch": 2.16, "learning_rate": 2.5619466282542234e-05, "loss": 0.6487, "step": 21538 }, { "epoch": 2.16, "learning_rate": 2.561709877760261e-05, "loss": 0.156, "step": 21539 }, { "epoch": 2.16, "learning_rate": 2.5614731300554092e-05, "loss": 0.1695, "step": 21540 }, { "epoch": 2.16, "learning_rate": 2.5612363851411763e-05, "loss": 0.1954, "step": 21541 }, { "epoch": 2.16, "learning_rate": 2.560999643019069e-05, "loss": 0.5142, "step": 21542 }, { "epoch": 2.16, "learning_rate": 2.5607629036905915e-05, "loss": 0.6521, "step": 21543 }, { "epoch": 2.16, "learning_rate": 2.5605261671572526e-05, "loss": 0.5087, "step": 21544 }, { "epoch": 2.16, "learning_rate": 2.5602894334205586e-05, "loss": 0.2767, "step": 21545 }, { "epoch": 2.16, "learning_rate": 2.5600527024820133e-05, "loss": 0.3341, "step": 21546 }, { "epoch": 2.16, "learning_rate": 2.5598159743431273e-05, "loss": 0.2778, "step": 21547 }, { "epoch": 2.16, "learning_rate": 2.5595792490054044e-05, "loss": 0.0908, "step": 21548 }, { "epoch": 2.16, "learning_rate": 2.5593425264703508e-05, "loss": 0.9921, "step": 21549 }, { "epoch": 2.16, "learning_rate": 2.559105806739475e-05, "loss": 0.3541, "step": 21550 }, { "epoch": 2.16, "learning_rate": 2.5588690898142818e-05, "loss": 0.2553, "step": 21551 }, { "epoch": 2.16, "learning_rate": 2.5586323756962766e-05, "loss": 0.6775, "step": 21552 }, { "epoch": 2.16, "learning_rate": 2.5583956643869683e-05, "loss": 0.1486, "step": 21553 }, { "epoch": 2.16, "learning_rate": 2.5581589558878623e-05, "loss": 0.1936, "step": 21554 }, { "epoch": 2.16, "learning_rate": 2.5579222502004628e-05, "loss": 0.3764, "step": 21555 }, { "epoch": 2.16, "learning_rate": 2.55768554732628e-05, "loss": 0.2877, "step": 21556 }, { "epoch": 2.16, "learning_rate": 2.5574488472668167e-05, "loss": 0.3303, "step": 21557 }, { "epoch": 2.16, "learning_rate": 2.557212150023581e-05, "loss": 0.2182, "step": 21558 }, { "epoch": 2.16, "learning_rate": 2.55697545559808e-05, "loss": 0.5332, "step": 21559 }, { "epoch": 2.16, "learning_rate": 2.556738763991816e-05, "loss": 0.4527, "step": 21560 }, { "epoch": 2.16, "learning_rate": 2.5565020752063005e-05, "loss": 0.2137, "step": 21561 }, { "epoch": 2.16, "learning_rate": 2.556265389243036e-05, "loss": 0.2401, "step": 21562 }, { "epoch": 2.16, "learning_rate": 2.556028706103529e-05, "loss": 0.4618, "step": 21563 }, { "epoch": 2.16, "learning_rate": 2.555792025789287e-05, "loss": 0.309, "step": 21564 }, { "epoch": 2.16, "learning_rate": 2.5555553483018168e-05, "loss": 0.379, "step": 21565 }, { "epoch": 2.16, "learning_rate": 2.555318673642622e-05, "loss": 0.3613, "step": 21566 }, { "epoch": 2.16, "learning_rate": 2.5550820018132103e-05, "loss": 0.3906, "step": 21567 }, { "epoch": 2.16, "learning_rate": 2.554845332815088e-05, "loss": 0.7485, "step": 21568 }, { "epoch": 2.16, "learning_rate": 2.5546086666497586e-05, "loss": 0.1582, "step": 21569 }, { "epoch": 2.16, "learning_rate": 2.5543720033187326e-05, "loss": 0.4349, "step": 21570 }, { "epoch": 2.16, "learning_rate": 2.5541353428235125e-05, "loss": 0.2172, "step": 21571 }, { "epoch": 2.16, "learning_rate": 2.5538986851656055e-05, "loss": 0.658, "step": 21572 }, { "epoch": 2.17, "learning_rate": 2.553662030346518e-05, "loss": 0.3053, "step": 21573 }, { "epoch": 2.17, "learning_rate": 2.5534253783677558e-05, "loss": 0.2335, "step": 21574 }, { "epoch": 2.17, "learning_rate": 2.5531887292308234e-05, "loss": 0.378, "step": 21575 }, { "epoch": 2.17, "learning_rate": 2.552952082937229e-05, "loss": 0.0916, "step": 21576 }, { "epoch": 2.17, "learning_rate": 2.5527154394884777e-05, "loss": 0.2523, "step": 21577 }, { "epoch": 2.17, "learning_rate": 2.5524787988860735e-05, "loss": 0.1035, "step": 21578 }, { "epoch": 2.17, "learning_rate": 2.5522421611315258e-05, "loss": 0.9099, "step": 21579 }, { "epoch": 2.17, "learning_rate": 2.552005526226338e-05, "loss": 0.3159, "step": 21580 }, { "epoch": 2.17, "learning_rate": 2.551768894172016e-05, "loss": 0.1805, "step": 21581 }, { "epoch": 2.17, "learning_rate": 2.5515322649700674e-05, "loss": 0.2975, "step": 21582 }, { "epoch": 2.17, "learning_rate": 2.5512956386219964e-05, "loss": 0.4102, "step": 21583 }, { "epoch": 2.17, "learning_rate": 2.5510590151293087e-05, "loss": 0.1758, "step": 21584 }, { "epoch": 2.17, "learning_rate": 2.5508223944935113e-05, "loss": 0.1586, "step": 21585 }, { "epoch": 2.17, "learning_rate": 2.5505857767161083e-05, "loss": 0.1995, "step": 21586 }, { "epoch": 2.17, "learning_rate": 2.5503491617986073e-05, "loss": 0.9391, "step": 21587 }, { "epoch": 2.17, "learning_rate": 2.5501125497425137e-05, "loss": 0.1113, "step": 21588 }, { "epoch": 2.17, "learning_rate": 2.5498759405493315e-05, "loss": 0.3419, "step": 21589 }, { "epoch": 2.17, "learning_rate": 2.5496393342205684e-05, "loss": 0.7217, "step": 21590 }, { "epoch": 2.17, "learning_rate": 2.5494027307577298e-05, "loss": 0.6445, "step": 21591 }, { "epoch": 2.17, "learning_rate": 2.5491661301623186e-05, "loss": 0.3318, "step": 21592 }, { "epoch": 2.17, "learning_rate": 2.548929532435845e-05, "loss": 0.3246, "step": 21593 }, { "epoch": 2.17, "learning_rate": 2.548692937579811e-05, "loss": 0.3193, "step": 21594 }, { "epoch": 2.17, "learning_rate": 2.5484563455957234e-05, "loss": 0.5145, "step": 21595 }, { "epoch": 2.17, "learning_rate": 2.5482197564850883e-05, "loss": 0.17, "step": 21596 }, { "epoch": 2.17, "learning_rate": 2.5479831702494115e-05, "loss": 0.6292, "step": 21597 }, { "epoch": 2.17, "learning_rate": 2.5477465868901965e-05, "loss": 0.518, "step": 21598 }, { "epoch": 2.17, "learning_rate": 2.5475100064089507e-05, "loss": 0.5228, "step": 21599 }, { "epoch": 2.17, "learning_rate": 2.54727342880718e-05, "loss": 0.4957, "step": 21600 }, { "epoch": 2.17, "learning_rate": 2.547036854086387e-05, "loss": 0.4118, "step": 21601 }, { "epoch": 2.17, "learning_rate": 2.546800282248081e-05, "loss": 0.3735, "step": 21602 }, { "epoch": 2.17, "learning_rate": 2.546563713293765e-05, "loss": 0.2861, "step": 21603 }, { "epoch": 2.17, "learning_rate": 2.546327147224944e-05, "loss": 0.1766, "step": 21604 }, { "epoch": 2.17, "learning_rate": 2.5460905840431256e-05, "loss": 0.6207, "step": 21605 }, { "epoch": 2.17, "learning_rate": 2.5458540237498145e-05, "loss": 0.4733, "step": 21606 }, { "epoch": 2.17, "learning_rate": 2.5456174663465143e-05, "loss": 0.3554, "step": 21607 }, { "epoch": 2.17, "learning_rate": 2.5453809118347323e-05, "loss": 0.0773, "step": 21608 }, { "epoch": 2.17, "learning_rate": 2.545144360215974e-05, "loss": 0.7002, "step": 21609 }, { "epoch": 2.17, "learning_rate": 2.544907811491742e-05, "loss": 0.1383, "step": 21610 }, { "epoch": 2.17, "learning_rate": 2.5446712656635453e-05, "loss": 0.2082, "step": 21611 }, { "epoch": 2.17, "learning_rate": 2.5444347227328866e-05, "loss": 0.4762, "step": 21612 }, { "epoch": 2.17, "learning_rate": 2.544198182701272e-05, "loss": 0.0879, "step": 21613 }, { "epoch": 2.17, "learning_rate": 2.5439616455702082e-05, "loss": 0.2698, "step": 21614 }, { "epoch": 2.17, "learning_rate": 2.5437251113411966e-05, "loss": 0.1723, "step": 21615 }, { "epoch": 2.17, "learning_rate": 2.5434885800157468e-05, "loss": 0.2778, "step": 21616 }, { "epoch": 2.17, "learning_rate": 2.5432520515953612e-05, "loss": 0.458, "step": 21617 }, { "epoch": 2.17, "learning_rate": 2.5430155260815454e-05, "loss": 0.2883, "step": 21618 }, { "epoch": 2.17, "learning_rate": 2.5427790034758055e-05, "loss": 0.4044, "step": 21619 }, { "epoch": 2.17, "learning_rate": 2.5425424837796468e-05, "loss": 0.516, "step": 21620 }, { "epoch": 2.17, "learning_rate": 2.5423059669945722e-05, "loss": 0.4234, "step": 21621 }, { "epoch": 2.17, "learning_rate": 2.5420694531220893e-05, "loss": 0.8642, "step": 21622 }, { "epoch": 2.17, "learning_rate": 2.541832942163703e-05, "loss": 0.2303, "step": 21623 }, { "epoch": 2.17, "learning_rate": 2.541596434120915e-05, "loss": 0.1479, "step": 21624 }, { "epoch": 2.17, "learning_rate": 2.5413599289952348e-05, "loss": 0.1595, "step": 21625 }, { "epoch": 2.17, "learning_rate": 2.5411234267881652e-05, "loss": 0.1264, "step": 21626 }, { "epoch": 2.17, "learning_rate": 2.540886927501211e-05, "loss": 0.3014, "step": 21627 }, { "epoch": 2.17, "learning_rate": 2.5406504311358784e-05, "loss": 0.229, "step": 21628 }, { "epoch": 2.17, "learning_rate": 2.5404139376936722e-05, "loss": 0.1323, "step": 21629 }, { "epoch": 2.17, "learning_rate": 2.5401774471760956e-05, "loss": 0.7602, "step": 21630 }, { "epoch": 2.17, "learning_rate": 2.5399409595846554e-05, "loss": 0.2786, "step": 21631 }, { "epoch": 2.17, "learning_rate": 2.5397044749208568e-05, "loss": 0.2475, "step": 21632 }, { "epoch": 2.17, "learning_rate": 2.5394679931862018e-05, "loss": 0.37, "step": 21633 }, { "epoch": 2.17, "learning_rate": 2.539231514382199e-05, "loss": 0.1967, "step": 21634 }, { "epoch": 2.17, "learning_rate": 2.538995038510351e-05, "loss": 0.1186, "step": 21635 }, { "epoch": 2.17, "learning_rate": 2.5387585655721628e-05, "loss": 0.6791, "step": 21636 }, { "epoch": 2.17, "learning_rate": 2.5385220955691398e-05, "loss": 0.3843, "step": 21637 }, { "epoch": 2.17, "learning_rate": 2.5382856285027868e-05, "loss": 0.1117, "step": 21638 }, { "epoch": 2.17, "learning_rate": 2.538049164374609e-05, "loss": 0.213, "step": 21639 }, { "epoch": 2.17, "learning_rate": 2.53781270318611e-05, "loss": 0.3736, "step": 21640 }, { "epoch": 2.17, "learning_rate": 2.5375762449387944e-05, "loss": 0.2701, "step": 21641 }, { "epoch": 2.17, "learning_rate": 2.5373397896341688e-05, "loss": 0.122, "step": 21642 }, { "epoch": 2.17, "learning_rate": 2.5371033372737372e-05, "loss": 0.5838, "step": 21643 }, { "epoch": 2.17, "learning_rate": 2.5368668878590023e-05, "loss": 0.759, "step": 21644 }, { "epoch": 2.17, "learning_rate": 2.5366304413914714e-05, "loss": 0.3846, "step": 21645 }, { "epoch": 2.17, "learning_rate": 2.5363939978726485e-05, "loss": 0.2882, "step": 21646 }, { "epoch": 2.17, "learning_rate": 2.536157557304036e-05, "loss": 0.616, "step": 21647 }, { "epoch": 2.17, "learning_rate": 2.535921119687142e-05, "loss": 0.6541, "step": 21648 }, { "epoch": 2.17, "learning_rate": 2.535684685023469e-05, "loss": 0.5576, "step": 21649 }, { "epoch": 2.17, "learning_rate": 2.5354482533145214e-05, "loss": 0.1906, "step": 21650 }, { "epoch": 2.17, "learning_rate": 2.535211824561805e-05, "loss": 0.0837, "step": 21651 }, { "epoch": 2.17, "learning_rate": 2.5349753987668247e-05, "loss": 0.4213, "step": 21652 }, { "epoch": 2.17, "learning_rate": 2.5347389759310824e-05, "loss": 0.7695, "step": 21653 }, { "epoch": 2.17, "learning_rate": 2.534502556056085e-05, "loss": 0.3238, "step": 21654 }, { "epoch": 2.17, "learning_rate": 2.534266139143337e-05, "loss": 0.6462, "step": 21655 }, { "epoch": 2.17, "learning_rate": 2.53402972519434e-05, "loss": 0.2434, "step": 21656 }, { "epoch": 2.17, "learning_rate": 2.5337933142106027e-05, "loss": 0.1212, "step": 21657 }, { "epoch": 2.17, "learning_rate": 2.5335569061936266e-05, "loss": 0.0306, "step": 21658 }, { "epoch": 2.17, "learning_rate": 2.5333205011449162e-05, "loss": 0.1671, "step": 21659 }, { "epoch": 2.17, "learning_rate": 2.5330840990659775e-05, "loss": 0.4629, "step": 21660 }, { "epoch": 2.17, "learning_rate": 2.532847699958314e-05, "loss": 0.4268, "step": 21661 }, { "epoch": 2.17, "learning_rate": 2.5326113038234295e-05, "loss": 0.122, "step": 21662 }, { "epoch": 2.17, "learning_rate": 2.5323749106628292e-05, "loss": 0.2115, "step": 21663 }, { "epoch": 2.17, "learning_rate": 2.5321385204780176e-05, "loss": 0.2383, "step": 21664 }, { "epoch": 2.17, "learning_rate": 2.5319021332704967e-05, "loss": 0.4893, "step": 21665 }, { "epoch": 2.17, "learning_rate": 2.5316657490417745e-05, "loss": 0.3711, "step": 21666 }, { "epoch": 2.17, "learning_rate": 2.5314293677933516e-05, "loss": 0.0707, "step": 21667 }, { "epoch": 2.17, "learning_rate": 2.531192989526735e-05, "loss": 0.1693, "step": 21668 }, { "epoch": 2.17, "learning_rate": 2.5309566142434278e-05, "loss": 0.508, "step": 21669 }, { "epoch": 2.17, "learning_rate": 2.5307202419449336e-05, "loss": 0.093, "step": 21670 }, { "epoch": 2.17, "learning_rate": 2.5304838726327587e-05, "loss": 0.5734, "step": 21671 }, { "epoch": 2.18, "learning_rate": 2.5302475063084048e-05, "loss": 0.54, "step": 21672 }, { "epoch": 2.18, "learning_rate": 2.530011142973377e-05, "loss": 0.3624, "step": 21673 }, { "epoch": 2.18, "learning_rate": 2.5297747826291796e-05, "loss": 0.2511, "step": 21674 }, { "epoch": 2.18, "learning_rate": 2.5295384252773176e-05, "loss": 0.1906, "step": 21675 }, { "epoch": 2.18, "learning_rate": 2.5293020709192926e-05, "loss": 0.3868, "step": 21676 }, { "epoch": 2.18, "learning_rate": 2.5290657195566112e-05, "loss": 0.3756, "step": 21677 }, { "epoch": 2.18, "learning_rate": 2.5288293711907764e-05, "loss": 0.9379, "step": 21678 }, { "epoch": 2.18, "learning_rate": 2.5285930258232905e-05, "loss": 0.1187, "step": 21679 }, { "epoch": 2.18, "learning_rate": 2.5283566834556613e-05, "loss": 0.3028, "step": 21680 }, { "epoch": 2.18, "learning_rate": 2.5281203440893903e-05, "loss": 0.4381, "step": 21681 }, { "epoch": 2.18, "learning_rate": 2.5278840077259812e-05, "loss": 0.7191, "step": 21682 }, { "epoch": 2.18, "learning_rate": 2.527647674366939e-05, "loss": 0.3188, "step": 21683 }, { "epoch": 2.18, "learning_rate": 2.527411344013768e-05, "loss": 0.5265, "step": 21684 }, { "epoch": 2.18, "learning_rate": 2.5271750166679703e-05, "loss": 0.2227, "step": 21685 }, { "epoch": 2.18, "learning_rate": 2.5269386923310518e-05, "loss": 0.3245, "step": 21686 }, { "epoch": 2.18, "learning_rate": 2.5267023710045158e-05, "loss": 0.7099, "step": 21687 }, { "epoch": 2.18, "learning_rate": 2.526466052689864e-05, "loss": 0.2363, "step": 21688 }, { "epoch": 2.18, "learning_rate": 2.526229737388604e-05, "loss": 0.1378, "step": 21689 }, { "epoch": 2.18, "learning_rate": 2.5259934251022373e-05, "loss": 0.588, "step": 21690 }, { "epoch": 2.18, "learning_rate": 2.5257571158322673e-05, "loss": 0.2225, "step": 21691 }, { "epoch": 2.18, "learning_rate": 2.5255208095801995e-05, "loss": 0.1528, "step": 21692 }, { "epoch": 2.18, "learning_rate": 2.5252845063475354e-05, "loss": 0.1213, "step": 21693 }, { "epoch": 2.18, "learning_rate": 2.5250482061357822e-05, "loss": 0.1996, "step": 21694 }, { "epoch": 2.18, "learning_rate": 2.5248119089464404e-05, "loss": 0.376, "step": 21695 }, { "epoch": 2.18, "learning_rate": 2.5245756147810144e-05, "loss": 0.24, "step": 21696 }, { "epoch": 2.18, "learning_rate": 2.524339323641009e-05, "loss": 0.2021, "step": 21697 }, { "epoch": 2.18, "learning_rate": 2.5241030355279276e-05, "loss": 0.1895, "step": 21698 }, { "epoch": 2.18, "learning_rate": 2.5238667504432727e-05, "loss": 0.7102, "step": 21699 }, { "epoch": 2.18, "learning_rate": 2.5236304683885486e-05, "loss": 0.4464, "step": 21700 }, { "epoch": 2.18, "learning_rate": 2.523394189365259e-05, "loss": 0.541, "step": 21701 }, { "epoch": 2.18, "learning_rate": 2.5231579133749077e-05, "loss": 0.4456, "step": 21702 }, { "epoch": 2.18, "learning_rate": 2.5229216404189987e-05, "loss": 0.159, "step": 21703 }, { "epoch": 2.18, "learning_rate": 2.5226853704990342e-05, "loss": 0.1664, "step": 21704 }, { "epoch": 2.18, "learning_rate": 2.522449103616518e-05, "loss": 0.1354, "step": 21705 }, { "epoch": 2.18, "learning_rate": 2.5222128397729545e-05, "loss": 0.2069, "step": 21706 }, { "epoch": 2.18, "learning_rate": 2.5219765789698476e-05, "loss": 0.1601, "step": 21707 }, { "epoch": 2.18, "learning_rate": 2.5217403212086982e-05, "loss": 0.578, "step": 21708 }, { "epoch": 2.18, "learning_rate": 2.5215040664910124e-05, "loss": 0.2158, "step": 21709 }, { "epoch": 2.18, "learning_rate": 2.5212678148182935e-05, "loss": 0.4998, "step": 21710 }, { "epoch": 2.18, "learning_rate": 2.5210315661920418e-05, "loss": 0.8449, "step": 21711 }, { "epoch": 2.18, "learning_rate": 2.520795320613765e-05, "loss": 0.7652, "step": 21712 }, { "epoch": 2.18, "learning_rate": 2.520559078084964e-05, "loss": 0.7048, "step": 21713 }, { "epoch": 2.18, "learning_rate": 2.520322838607142e-05, "loss": 0.176, "step": 21714 }, { "epoch": 2.18, "learning_rate": 2.5200866021818035e-05, "loss": 0.5375, "step": 21715 }, { "epoch": 2.18, "learning_rate": 2.5198503688104522e-05, "loss": 0.4429, "step": 21716 }, { "epoch": 2.18, "learning_rate": 2.5196141384945886e-05, "loss": 0.1451, "step": 21717 }, { "epoch": 2.18, "learning_rate": 2.5193779112357194e-05, "loss": 0.5269, "step": 21718 }, { "epoch": 2.18, "learning_rate": 2.519141687035345e-05, "loss": 0.0771, "step": 21719 }, { "epoch": 2.18, "learning_rate": 2.518905465894971e-05, "loss": 0.0898, "step": 21720 }, { "epoch": 2.18, "learning_rate": 2.5186692478161e-05, "loss": 0.3716, "step": 21721 }, { "epoch": 2.18, "learning_rate": 2.5184330328002336e-05, "loss": 0.5577, "step": 21722 }, { "epoch": 2.18, "learning_rate": 2.5181968208488767e-05, "loss": 0.1063, "step": 21723 }, { "epoch": 2.18, "learning_rate": 2.517960611963532e-05, "loss": 0.2482, "step": 21724 }, { "epoch": 2.18, "learning_rate": 2.5177244061457025e-05, "loss": 0.457, "step": 21725 }, { "epoch": 2.18, "learning_rate": 2.517488203396892e-05, "loss": 0.1364, "step": 21726 }, { "epoch": 2.18, "learning_rate": 2.517252003718602e-05, "loss": 0.3566, "step": 21727 }, { "epoch": 2.18, "learning_rate": 2.517015807112337e-05, "loss": 0.3162, "step": 21728 }, { "epoch": 2.18, "learning_rate": 2.5167796135796e-05, "loss": 0.1577, "step": 21729 }, { "epoch": 2.18, "learning_rate": 2.5165434231218942e-05, "loss": 0.67, "step": 21730 }, { "epoch": 2.18, "learning_rate": 2.5163072357407205e-05, "loss": 0.5901, "step": 21731 }, { "epoch": 2.18, "learning_rate": 2.516071051437585e-05, "loss": 0.4766, "step": 21732 }, { "epoch": 2.18, "learning_rate": 2.5158348702139885e-05, "loss": 0.1725, "step": 21733 }, { "epoch": 2.18, "learning_rate": 2.5155986920714344e-05, "loss": 0.2122, "step": 21734 }, { "epoch": 2.18, "learning_rate": 2.5153625170114273e-05, "loss": 0.0958, "step": 21735 }, { "epoch": 2.18, "learning_rate": 2.515126345035468e-05, "loss": 0.6277, "step": 21736 }, { "epoch": 2.18, "learning_rate": 2.5148901761450593e-05, "loss": 0.4636, "step": 21737 }, { "epoch": 2.18, "learning_rate": 2.5146540103417058e-05, "loss": 0.5853, "step": 21738 }, { "epoch": 2.18, "learning_rate": 2.51441784762691e-05, "loss": 0.2441, "step": 21739 }, { "epoch": 2.18, "learning_rate": 2.5141816880021734e-05, "loss": 1.4632, "step": 21740 }, { "epoch": 2.18, "learning_rate": 2.5139455314690005e-05, "loss": 0.3132, "step": 21741 }, { "epoch": 2.18, "learning_rate": 2.5137093780288937e-05, "loss": 0.7449, "step": 21742 }, { "epoch": 2.18, "learning_rate": 2.5134732276833534e-05, "loss": 0.4177, "step": 21743 }, { "epoch": 2.18, "learning_rate": 2.5132370804338863e-05, "loss": 0.2627, "step": 21744 }, { "epoch": 2.18, "learning_rate": 2.513000936281993e-05, "loss": 0.3527, "step": 21745 }, { "epoch": 2.18, "learning_rate": 2.512764795229175e-05, "loss": 0.6836, "step": 21746 }, { "epoch": 2.18, "learning_rate": 2.5125286572769378e-05, "loss": 0.4735, "step": 21747 }, { "epoch": 2.18, "learning_rate": 2.512292522426782e-05, "loss": 0.5545, "step": 21748 }, { "epoch": 2.18, "learning_rate": 2.5120563906802126e-05, "loss": 0.4745, "step": 21749 }, { "epoch": 2.18, "learning_rate": 2.51182026203873e-05, "loss": 0.8886, "step": 21750 }, { "epoch": 2.18, "learning_rate": 2.5115841365038366e-05, "loss": 0.5362, "step": 21751 }, { "epoch": 2.18, "learning_rate": 2.511348014077037e-05, "loss": 0.3125, "step": 21752 }, { "epoch": 2.18, "learning_rate": 2.5111118947598327e-05, "loss": 0.2033, "step": 21753 }, { "epoch": 2.18, "learning_rate": 2.5108757785537256e-05, "loss": 0.7628, "step": 21754 }, { "epoch": 2.18, "learning_rate": 2.5106396654602193e-05, "loss": 0.1662, "step": 21755 }, { "epoch": 2.18, "learning_rate": 2.510403555480816e-05, "loss": 0.7543, "step": 21756 }, { "epoch": 2.18, "learning_rate": 2.510167448617018e-05, "loss": 0.1038, "step": 21757 }, { "epoch": 2.18, "learning_rate": 2.509931344870329e-05, "loss": 0.599, "step": 21758 }, { "epoch": 2.18, "learning_rate": 2.50969524424225e-05, "loss": 0.6027, "step": 21759 }, { "epoch": 2.18, "learning_rate": 2.509459146734283e-05, "loss": 0.4849, "step": 21760 }, { "epoch": 2.18, "learning_rate": 2.5092230523479323e-05, "loss": 0.1073, "step": 21761 }, { "epoch": 2.18, "learning_rate": 2.5089869610847e-05, "loss": 0.5145, "step": 21762 }, { "epoch": 2.18, "learning_rate": 2.5087508729460866e-05, "loss": 0.4224, "step": 21763 }, { "epoch": 2.18, "learning_rate": 2.508514787933596e-05, "loss": 0.5616, "step": 21764 }, { "epoch": 2.18, "learning_rate": 2.508278706048731e-05, "loss": 0.0938, "step": 21765 }, { "epoch": 2.18, "learning_rate": 2.5080426272929925e-05, "loss": 0.3532, "step": 21766 }, { "epoch": 2.18, "learning_rate": 2.5078065516678846e-05, "loss": 0.1324, "step": 21767 }, { "epoch": 2.18, "learning_rate": 2.5075704791749077e-05, "loss": 0.3719, "step": 21768 }, { "epoch": 2.18, "learning_rate": 2.5073344098155645e-05, "loss": 0.3403, "step": 21769 }, { "epoch": 2.19, "learning_rate": 2.5070983435913585e-05, "loss": 0.2673, "step": 21770 }, { "epoch": 2.19, "learning_rate": 2.5068622805037917e-05, "loss": 0.4823, "step": 21771 }, { "epoch": 2.19, "learning_rate": 2.506626220554364e-05, "loss": 0.6024, "step": 21772 }, { "epoch": 2.19, "learning_rate": 2.506390163744581e-05, "loss": 0.6036, "step": 21773 }, { "epoch": 2.19, "learning_rate": 2.5061541100759417e-05, "loss": 0.6197, "step": 21774 }, { "epoch": 2.19, "learning_rate": 2.5059180595499508e-05, "loss": 0.2766, "step": 21775 }, { "epoch": 2.19, "learning_rate": 2.50568201216811e-05, "loss": 0.4477, "step": 21776 }, { "epoch": 2.19, "learning_rate": 2.5054459679319194e-05, "loss": 0.3523, "step": 21777 }, { "epoch": 2.19, "learning_rate": 2.5052099268428835e-05, "loss": 0.4485, "step": 21778 }, { "epoch": 2.19, "learning_rate": 2.504973888902503e-05, "loss": 0.2841, "step": 21779 }, { "epoch": 2.19, "learning_rate": 2.50473785411228e-05, "loss": 0.4464, "step": 21780 }, { "epoch": 2.19, "learning_rate": 2.5045018224737178e-05, "loss": 0.5821, "step": 21781 }, { "epoch": 2.19, "learning_rate": 2.5042657939883173e-05, "loss": 0.3276, "step": 21782 }, { "epoch": 2.19, "learning_rate": 2.5040297686575796e-05, "loss": 0.215, "step": 21783 }, { "epoch": 2.19, "learning_rate": 2.5037937464830093e-05, "loss": 0.4826, "step": 21784 }, { "epoch": 2.19, "learning_rate": 2.5035577274661067e-05, "loss": 0.111, "step": 21785 }, { "epoch": 2.19, "learning_rate": 2.5033217116083727e-05, "loss": 0.286, "step": 21786 }, { "epoch": 2.19, "learning_rate": 2.503085698911311e-05, "loss": 0.4996, "step": 21787 }, { "epoch": 2.19, "learning_rate": 2.502849689376423e-05, "loss": 0.2762, "step": 21788 }, { "epoch": 2.19, "learning_rate": 2.5026136830052098e-05, "loss": 0.55, "step": 21789 }, { "epoch": 2.19, "learning_rate": 2.502377679799175e-05, "loss": 0.647, "step": 21790 }, { "epoch": 2.19, "learning_rate": 2.502141679759819e-05, "loss": 0.1002, "step": 21791 }, { "epoch": 2.19, "learning_rate": 2.501905682888643e-05, "loss": 0.5448, "step": 21792 }, { "epoch": 2.19, "learning_rate": 2.5016696891871507e-05, "loss": 0.1654, "step": 21793 }, { "epoch": 2.19, "learning_rate": 2.5014336986568432e-05, "loss": 0.2006, "step": 21794 }, { "epoch": 2.19, "learning_rate": 2.501197711299221e-05, "loss": 0.1727, "step": 21795 }, { "epoch": 2.19, "learning_rate": 2.5009617271157874e-05, "loss": 0.2258, "step": 21796 }, { "epoch": 2.19, "learning_rate": 2.5007257461080436e-05, "loss": 0.2925, "step": 21797 }, { "epoch": 2.19, "learning_rate": 2.5004897682774905e-05, "loss": 0.4978, "step": 21798 }, { "epoch": 2.19, "learning_rate": 2.500253793625632e-05, "loss": 0.2549, "step": 21799 }, { "epoch": 2.19, "learning_rate": 2.500017822153967e-05, "loss": 0.2717, "step": 21800 }, { "epoch": 2.19, "learning_rate": 2.499781853863998e-05, "loss": 0.5439, "step": 21801 }, { "epoch": 2.19, "learning_rate": 2.4995458887572276e-05, "loss": 0.4072, "step": 21802 }, { "epoch": 2.19, "learning_rate": 2.4993099268351564e-05, "loss": 0.5313, "step": 21803 }, { "epoch": 2.19, "learning_rate": 2.4990739680992874e-05, "loss": 0.4134, "step": 21804 }, { "epoch": 2.19, "learning_rate": 2.498838012551121e-05, "loss": 0.1701, "step": 21805 }, { "epoch": 2.19, "learning_rate": 2.4986020601921574e-05, "loss": 0.8057, "step": 21806 }, { "epoch": 2.19, "learning_rate": 2.4983661110239006e-05, "loss": 0.1629, "step": 21807 }, { "epoch": 2.19, "learning_rate": 2.4981301650478515e-05, "loss": 0.6856, "step": 21808 }, { "epoch": 2.19, "learning_rate": 2.49789422226551e-05, "loss": 0.7655, "step": 21809 }, { "epoch": 2.19, "learning_rate": 2.4976582826783793e-05, "loss": 0.2513, "step": 21810 }, { "epoch": 2.19, "learning_rate": 2.49742234628796e-05, "loss": 1.4021, "step": 21811 }, { "epoch": 2.19, "learning_rate": 2.497186413095753e-05, "loss": 0.3089, "step": 21812 }, { "epoch": 2.19, "learning_rate": 2.496950483103262e-05, "loss": 0.3607, "step": 21813 }, { "epoch": 2.19, "learning_rate": 2.496714556311986e-05, "loss": 0.4855, "step": 21814 }, { "epoch": 2.19, "learning_rate": 2.4964786327234266e-05, "loss": 0.755, "step": 21815 }, { "epoch": 2.19, "learning_rate": 2.4962427123390858e-05, "loss": 0.1813, "step": 21816 }, { "epoch": 2.19, "learning_rate": 2.4960067951604658e-05, "loss": 0.5036, "step": 21817 }, { "epoch": 2.19, "learning_rate": 2.4957708811890653e-05, "loss": 0.421, "step": 21818 }, { "epoch": 2.19, "learning_rate": 2.495534970426388e-05, "loss": 0.4739, "step": 21819 }, { "epoch": 2.19, "learning_rate": 2.4952990628739342e-05, "loss": 0.4196, "step": 21820 }, { "epoch": 2.19, "learning_rate": 2.4950631585332045e-05, "loss": 0.1734, "step": 21821 }, { "epoch": 2.19, "learning_rate": 2.494827257405702e-05, "loss": 0.5703, "step": 21822 }, { "epoch": 2.19, "learning_rate": 2.494591359492926e-05, "loss": 0.1484, "step": 21823 }, { "epoch": 2.19, "learning_rate": 2.4943554647963776e-05, "loss": 0.1657, "step": 21824 }, { "epoch": 2.19, "learning_rate": 2.4941195733175592e-05, "loss": 0.1802, "step": 21825 }, { "epoch": 2.19, "learning_rate": 2.4938836850579726e-05, "loss": 0.1561, "step": 21826 }, { "epoch": 2.19, "learning_rate": 2.493647800019116e-05, "loss": 0.8695, "step": 21827 }, { "epoch": 2.19, "learning_rate": 2.493411918202493e-05, "loss": 0.534, "step": 21828 }, { "epoch": 2.19, "learning_rate": 2.493176039609603e-05, "loss": 0.3379, "step": 21829 }, { "epoch": 2.19, "learning_rate": 2.4929401642419482e-05, "loss": 0.2404, "step": 21830 }, { "epoch": 2.19, "learning_rate": 2.4927042921010303e-05, "loss": 0.3842, "step": 21831 }, { "epoch": 2.19, "learning_rate": 2.492468423188348e-05, "loss": 0.5358, "step": 21832 }, { "epoch": 2.19, "learning_rate": 2.4922325575054044e-05, "loss": 0.2474, "step": 21833 }, { "epoch": 2.19, "learning_rate": 2.4919966950536994e-05, "loss": 0.3445, "step": 21834 }, { "epoch": 2.19, "learning_rate": 2.4917608358347336e-05, "loss": 0.4001, "step": 21835 }, { "epoch": 2.19, "learning_rate": 2.49152497985001e-05, "loss": 0.7697, "step": 21836 }, { "epoch": 2.19, "learning_rate": 2.491289127101027e-05, "loss": 0.2824, "step": 21837 }, { "epoch": 2.19, "learning_rate": 2.4910532775892858e-05, "loss": 0.4843, "step": 21838 }, { "epoch": 2.19, "learning_rate": 2.4908174313162885e-05, "loss": 0.8972, "step": 21839 }, { "epoch": 2.19, "learning_rate": 2.4905815882835365e-05, "loss": 0.1781, "step": 21840 }, { "epoch": 2.19, "learning_rate": 2.4903457484925277e-05, "loss": 0.3208, "step": 21841 }, { "epoch": 2.19, "learning_rate": 2.490109911944766e-05, "loss": 0.1243, "step": 21842 }, { "epoch": 2.19, "learning_rate": 2.4898740786417503e-05, "loss": 0.3535, "step": 21843 }, { "epoch": 2.19, "learning_rate": 2.4896382485849813e-05, "loss": 0.0208, "step": 21844 }, { "epoch": 2.19, "learning_rate": 2.489402421775962e-05, "loss": 0.458, "step": 21845 }, { "epoch": 2.19, "learning_rate": 2.48916659821619e-05, "loss": 0.2123, "step": 21846 }, { "epoch": 2.19, "learning_rate": 2.4889307779071674e-05, "loss": 0.5356, "step": 21847 }, { "epoch": 2.19, "learning_rate": 2.4886949608503954e-05, "loss": 0.2519, "step": 21848 }, { "epoch": 2.19, "learning_rate": 2.488459147047375e-05, "loss": 0.4176, "step": 21849 }, { "epoch": 2.19, "learning_rate": 2.4882233364996045e-05, "loss": 1.0061, "step": 21850 }, { "epoch": 2.19, "learning_rate": 2.4879875292085865e-05, "loss": 0.4607, "step": 21851 }, { "epoch": 2.19, "learning_rate": 2.4877517251758212e-05, "loss": 0.0765, "step": 21852 }, { "epoch": 2.19, "learning_rate": 2.4875159244028084e-05, "loss": 0.2634, "step": 21853 }, { "epoch": 2.19, "learning_rate": 2.487280126891051e-05, "loss": 0.2229, "step": 21854 }, { "epoch": 2.19, "learning_rate": 2.487044332642046e-05, "loss": 0.8685, "step": 21855 }, { "epoch": 2.19, "learning_rate": 2.4868085416572967e-05, "loss": 0.3859, "step": 21856 }, { "epoch": 2.19, "learning_rate": 2.4865727539383024e-05, "loss": 0.5814, "step": 21857 }, { "epoch": 2.19, "learning_rate": 2.4863369694865635e-05, "loss": 0.8488, "step": 21858 }, { "epoch": 2.19, "learning_rate": 2.4861011883035816e-05, "loss": 0.0987, "step": 21859 }, { "epoch": 2.19, "learning_rate": 2.4858654103908558e-05, "loss": 0.0547, "step": 21860 }, { "epoch": 2.19, "learning_rate": 2.485629635749886e-05, "loss": 0.5547, "step": 21861 }, { "epoch": 2.19, "learning_rate": 2.4853938643821744e-05, "loss": 0.8066, "step": 21862 }, { "epoch": 2.19, "learning_rate": 2.485158096289221e-05, "loss": 0.6361, "step": 21863 }, { "epoch": 2.19, "learning_rate": 2.4849223314725248e-05, "loss": 0.5444, "step": 21864 }, { "epoch": 2.19, "learning_rate": 2.4846865699335872e-05, "loss": 0.365, "step": 21865 }, { "epoch": 2.19, "learning_rate": 2.484450811673908e-05, "loss": 0.544, "step": 21866 }, { "epoch": 2.19, "learning_rate": 2.4842150566949872e-05, "loss": 0.4745, "step": 21867 }, { "epoch": 2.19, "learning_rate": 2.483979304998327e-05, "loss": 0.3547, "step": 21868 }, { "epoch": 2.2, "learning_rate": 2.4837435565854258e-05, "loss": 0.8005, "step": 21869 }, { "epoch": 2.2, "learning_rate": 2.4835078114577834e-05, "loss": 0.2485, "step": 21870 }, { "epoch": 2.2, "learning_rate": 2.483272069616901e-05, "loss": 0.2491, "step": 21871 }, { "epoch": 2.2, "learning_rate": 2.48303633106428e-05, "loss": 0.3355, "step": 21872 }, { "epoch": 2.2, "learning_rate": 2.4828005958014172e-05, "loss": 0.1902, "step": 21873 }, { "epoch": 2.2, "learning_rate": 2.4825648638298153e-05, "loss": 0.2039, "step": 21874 }, { "epoch": 2.2, "learning_rate": 2.482329135150974e-05, "loss": 0.392, "step": 21875 }, { "epoch": 2.2, "learning_rate": 2.4820934097663923e-05, "loss": 0.0844, "step": 21876 }, { "epoch": 2.2, "learning_rate": 2.4818576876775727e-05, "loss": 0.3687, "step": 21877 }, { "epoch": 2.2, "learning_rate": 2.481621968886013e-05, "loss": 0.2562, "step": 21878 }, { "epoch": 2.2, "learning_rate": 2.481386253393213e-05, "loss": 0.3302, "step": 21879 }, { "epoch": 2.2, "learning_rate": 2.4811505412006744e-05, "loss": 0.6534, "step": 21880 }, { "epoch": 2.2, "learning_rate": 2.480914832309897e-05, "loss": 0.43, "step": 21881 }, { "epoch": 2.2, "learning_rate": 2.480679126722379e-05, "loss": 0.3716, "step": 21882 }, { "epoch": 2.2, "learning_rate": 2.4804434244396218e-05, "loss": 0.0828, "step": 21883 }, { "epoch": 2.2, "learning_rate": 2.4802077254631244e-05, "loss": 0.5723, "step": 21884 }, { "epoch": 2.2, "learning_rate": 2.4799720297943884e-05, "loss": 0.3208, "step": 21885 }, { "epoch": 2.2, "learning_rate": 2.4797363374349128e-05, "loss": 0.1592, "step": 21886 }, { "epoch": 2.2, "learning_rate": 2.4795006483861962e-05, "loss": 0.1906, "step": 21887 }, { "epoch": 2.2, "learning_rate": 2.4792649626497397e-05, "loss": 0.5667, "step": 21888 }, { "epoch": 2.2, "learning_rate": 2.479029280227043e-05, "loss": 0.2238, "step": 21889 }, { "epoch": 2.2, "learning_rate": 2.4787936011196053e-05, "loss": 0.2687, "step": 21890 }, { "epoch": 2.2, "learning_rate": 2.4785579253289283e-05, "loss": 0.3264, "step": 21891 }, { "epoch": 2.2, "learning_rate": 2.4783222528565094e-05, "loss": 0.4402, "step": 21892 }, { "epoch": 2.2, "learning_rate": 2.4780865837038486e-05, "loss": 0.3204, "step": 21893 }, { "epoch": 2.2, "learning_rate": 2.4778509178724468e-05, "loss": 0.0911, "step": 21894 }, { "epoch": 2.2, "learning_rate": 2.4776152553638038e-05, "loss": 0.8351, "step": 21895 }, { "epoch": 2.2, "learning_rate": 2.477379596179417e-05, "loss": 0.1777, "step": 21896 }, { "epoch": 2.2, "learning_rate": 2.477143940320789e-05, "loss": 0.7542, "step": 21897 }, { "epoch": 2.2, "learning_rate": 2.4769082877894176e-05, "loss": 0.0977, "step": 21898 }, { "epoch": 2.2, "learning_rate": 2.4766726385868027e-05, "loss": 0.3963, "step": 21899 }, { "epoch": 2.2, "learning_rate": 2.4764369927144447e-05, "loss": 0.4148, "step": 21900 }, { "epoch": 2.2, "learning_rate": 2.4762013501738418e-05, "loss": 0.1715, "step": 21901 }, { "epoch": 2.2, "learning_rate": 2.4759657109664944e-05, "loss": 0.4499, "step": 21902 }, { "epoch": 2.2, "learning_rate": 2.475730075093902e-05, "loss": 0.4579, "step": 21903 }, { "epoch": 2.2, "learning_rate": 2.4754944425575644e-05, "loss": 0.4612, "step": 21904 }, { "epoch": 2.2, "learning_rate": 2.4752588133589796e-05, "loss": 0.1813, "step": 21905 }, { "epoch": 2.2, "learning_rate": 2.4750231874996487e-05, "loss": 0.6059, "step": 21906 }, { "epoch": 2.2, "learning_rate": 2.4747875649810703e-05, "loss": 0.3603, "step": 21907 }, { "epoch": 2.2, "learning_rate": 2.4745519458047436e-05, "loss": 0.7155, "step": 21908 }, { "epoch": 2.2, "learning_rate": 2.47431632997217e-05, "loss": 0.3918, "step": 21909 }, { "epoch": 2.2, "learning_rate": 2.4740807174848455e-05, "loss": 0.7136, "step": 21910 }, { "epoch": 2.2, "learning_rate": 2.4738451083442723e-05, "loss": 0.1894, "step": 21911 }, { "epoch": 2.2, "learning_rate": 2.4736095025519485e-05, "loss": 0.1652, "step": 21912 }, { "epoch": 2.2, "learning_rate": 2.4733739001093726e-05, "loss": 0.4769, "step": 21913 }, { "epoch": 2.2, "learning_rate": 2.4731383010180467e-05, "loss": 0.3615, "step": 21914 }, { "epoch": 2.2, "learning_rate": 2.4729027052794673e-05, "loss": 0.3696, "step": 21915 }, { "epoch": 2.2, "learning_rate": 2.4726671128951343e-05, "loss": 0.1381, "step": 21916 }, { "epoch": 2.2, "learning_rate": 2.4724315238665472e-05, "loss": 0.0782, "step": 21917 }, { "epoch": 2.2, "learning_rate": 2.4721959381952065e-05, "loss": 0.2064, "step": 21918 }, { "epoch": 2.2, "learning_rate": 2.471960355882608e-05, "loss": 0.2445, "step": 21919 }, { "epoch": 2.2, "learning_rate": 2.4717247769302543e-05, "loss": 0.2136, "step": 21920 }, { "epoch": 2.2, "learning_rate": 2.4714892013396427e-05, "loss": 0.3, "step": 21921 }, { "epoch": 2.2, "learning_rate": 2.471253629112272e-05, "loss": 0.6634, "step": 21922 }, { "epoch": 2.2, "learning_rate": 2.4710180602496437e-05, "loss": 0.5466, "step": 21923 }, { "epoch": 2.2, "learning_rate": 2.4707824947532546e-05, "loss": 0.5969, "step": 21924 }, { "epoch": 2.2, "learning_rate": 2.4705469326246037e-05, "loss": 0.4159, "step": 21925 }, { "epoch": 2.2, "learning_rate": 2.470311373865191e-05, "loss": 0.2284, "step": 21926 }, { "epoch": 2.2, "learning_rate": 2.470075818476516e-05, "loss": 0.2602, "step": 21927 }, { "epoch": 2.2, "learning_rate": 2.4698402664600762e-05, "loss": 0.4413, "step": 21928 }, { "epoch": 2.2, "learning_rate": 2.4696047178173713e-05, "loss": 0.6731, "step": 21929 }, { "epoch": 2.2, "learning_rate": 2.4693691725499002e-05, "loss": 0.3221, "step": 21930 }, { "epoch": 2.2, "learning_rate": 2.4691336306591616e-05, "loss": 0.583, "step": 21931 }, { "epoch": 2.2, "learning_rate": 2.4688980921466553e-05, "loss": 0.4815, "step": 21932 }, { "epoch": 2.2, "learning_rate": 2.4686625570138792e-05, "loss": 0.3886, "step": 21933 }, { "epoch": 2.2, "learning_rate": 2.468427025262332e-05, "loss": 0.2506, "step": 21934 }, { "epoch": 2.2, "learning_rate": 2.4681914968935134e-05, "loss": 0.1904, "step": 21935 }, { "epoch": 2.2, "learning_rate": 2.467955971908921e-05, "loss": 0.634, "step": 21936 }, { "epoch": 2.2, "learning_rate": 2.467720450310056e-05, "loss": 0.4284, "step": 21937 }, { "epoch": 2.2, "learning_rate": 2.467484932098415e-05, "loss": 0.2328, "step": 21938 }, { "epoch": 2.2, "learning_rate": 2.467249417275496e-05, "loss": 0.6574, "step": 21939 }, { "epoch": 2.2, "learning_rate": 2.4670139058428004e-05, "loss": 0.0783, "step": 21940 }, { "epoch": 2.2, "learning_rate": 2.466778397801826e-05, "loss": 0.2914, "step": 21941 }, { "epoch": 2.2, "learning_rate": 2.4665428931540695e-05, "loss": 0.1111, "step": 21942 }, { "epoch": 2.2, "learning_rate": 2.466307391901032e-05, "loss": 0.4439, "step": 21943 }, { "epoch": 2.2, "learning_rate": 2.4660718940442113e-05, "loss": 0.1873, "step": 21944 }, { "epoch": 2.2, "learning_rate": 2.465836399585105e-05, "loss": 0.148, "step": 21945 }, { "epoch": 2.2, "learning_rate": 2.465600908525214e-05, "loss": 0.6499, "step": 21946 }, { "epoch": 2.2, "learning_rate": 2.465365420866035e-05, "loss": 0.5975, "step": 21947 }, { "epoch": 2.2, "learning_rate": 2.4651299366090665e-05, "loss": 0.1487, "step": 21948 }, { "epoch": 2.2, "learning_rate": 2.464894455755808e-05, "loss": 0.5414, "step": 21949 }, { "epoch": 2.2, "learning_rate": 2.4646589783077588e-05, "loss": 0.6485, "step": 21950 }, { "epoch": 2.2, "learning_rate": 2.4644235042664144e-05, "loss": 0.3255, "step": 21951 }, { "epoch": 2.2, "learning_rate": 2.4641880336332755e-05, "loss": 0.617, "step": 21952 }, { "epoch": 2.2, "learning_rate": 2.4639525664098407e-05, "loss": 0.2285, "step": 21953 }, { "epoch": 2.2, "learning_rate": 2.463717102597607e-05, "loss": 0.2982, "step": 21954 }, { "epoch": 2.2, "learning_rate": 2.463481642198075e-05, "loss": 0.2318, "step": 21955 }, { "epoch": 2.2, "learning_rate": 2.463246185212741e-05, "loss": 0.3232, "step": 21956 }, { "epoch": 2.2, "learning_rate": 2.4630107316431034e-05, "loss": 0.6949, "step": 21957 }, { "epoch": 2.2, "learning_rate": 2.4627752814906618e-05, "loss": 0.6565, "step": 21958 }, { "epoch": 2.2, "learning_rate": 2.4625398347569146e-05, "loss": 0.328, "step": 21959 }, { "epoch": 2.2, "learning_rate": 2.4623043914433584e-05, "loss": 0.4836, "step": 21960 }, { "epoch": 2.2, "learning_rate": 2.4620689515514927e-05, "loss": 0.0643, "step": 21961 }, { "epoch": 2.2, "learning_rate": 2.4618335150828162e-05, "loss": 0.9022, "step": 21962 }, { "epoch": 2.2, "learning_rate": 2.4615980820388253e-05, "loss": 0.4989, "step": 21963 }, { "epoch": 2.2, "learning_rate": 2.4613626524210208e-05, "loss": 0.2487, "step": 21964 }, { "epoch": 2.2, "learning_rate": 2.4611272262308983e-05, "loss": 0.3193, "step": 21965 }, { "epoch": 2.2, "learning_rate": 2.4608918034699575e-05, "loss": 0.4401, "step": 21966 }, { "epoch": 2.21, "learning_rate": 2.4606563841396964e-05, "loss": 0.4397, "step": 21967 }, { "epoch": 2.21, "learning_rate": 2.460420968241612e-05, "loss": 0.6078, "step": 21968 }, { "epoch": 2.21, "learning_rate": 2.460185555777205e-05, "loss": 0.3654, "step": 21969 }, { "epoch": 2.21, "learning_rate": 2.459950146747971e-05, "loss": 0.2398, "step": 21970 }, { "epoch": 2.21, "learning_rate": 2.459714741155408e-05, "loss": 0.6709, "step": 21971 }, { "epoch": 2.21, "learning_rate": 2.4594793390010155e-05, "loss": 0.1306, "step": 21972 }, { "epoch": 2.21, "learning_rate": 2.459243940286292e-05, "loss": 0.2964, "step": 21973 }, { "epoch": 2.21, "learning_rate": 2.4590085450127324e-05, "loss": 0.5178, "step": 21974 }, { "epoch": 2.21, "learning_rate": 2.4587731531818374e-05, "loss": 0.7513, "step": 21975 }, { "epoch": 2.21, "learning_rate": 2.4585377647951044e-05, "loss": 0.1498, "step": 21976 }, { "epoch": 2.21, "learning_rate": 2.4583023798540302e-05, "loss": 0.1329, "step": 21977 }, { "epoch": 2.21, "learning_rate": 2.4580669983601152e-05, "loss": 0.6512, "step": 21978 }, { "epoch": 2.21, "learning_rate": 2.457831620314855e-05, "loss": 0.1563, "step": 21979 }, { "epoch": 2.21, "learning_rate": 2.457596245719747e-05, "loss": 0.5572, "step": 21980 }, { "epoch": 2.21, "learning_rate": 2.457360874576291e-05, "loss": 0.0723, "step": 21981 }, { "epoch": 2.21, "learning_rate": 2.457125506885985e-05, "loss": 0.5461, "step": 21982 }, { "epoch": 2.21, "learning_rate": 2.456890142650324e-05, "loss": 0.8745, "step": 21983 }, { "epoch": 2.21, "learning_rate": 2.4566547818708082e-05, "loss": 0.2324, "step": 21984 }, { "epoch": 2.21, "learning_rate": 2.456419424548935e-05, "loss": 0.5152, "step": 21985 }, { "epoch": 2.21, "learning_rate": 2.4561840706862007e-05, "loss": 0.2331, "step": 21986 }, { "epoch": 2.21, "learning_rate": 2.455948720284106e-05, "loss": 0.389, "step": 21987 }, { "epoch": 2.21, "learning_rate": 2.4557133733441454e-05, "loss": 0.8527, "step": 21988 }, { "epoch": 2.21, "learning_rate": 2.4554780298678175e-05, "loss": 0.167, "step": 21989 }, { "epoch": 2.21, "learning_rate": 2.4552426898566208e-05, "loss": 0.6262, "step": 21990 }, { "epoch": 2.21, "learning_rate": 2.4550073533120518e-05, "loss": 0.1593, "step": 21991 }, { "epoch": 2.21, "learning_rate": 2.45477202023561e-05, "loss": 0.3548, "step": 21992 }, { "epoch": 2.21, "learning_rate": 2.454536690628791e-05, "loss": 0.1453, "step": 21993 }, { "epoch": 2.21, "learning_rate": 2.4543013644930924e-05, "loss": 0.1199, "step": 21994 }, { "epoch": 2.21, "learning_rate": 2.454066041830013e-05, "loss": 0.1829, "step": 21995 }, { "epoch": 2.21, "learning_rate": 2.4538307226410502e-05, "loss": 0.3937, "step": 21996 }, { "epoch": 2.21, "learning_rate": 2.4535954069276993e-05, "loss": 0.2851, "step": 21997 }, { "epoch": 2.21, "learning_rate": 2.4533600946914603e-05, "loss": 0.3781, "step": 21998 }, { "epoch": 2.21, "learning_rate": 2.4531247859338298e-05, "loss": 0.1019, "step": 21999 }, { "epoch": 2.21, "learning_rate": 2.4528894806563043e-05, "loss": 0.5969, "step": 22000 }, { "epoch": 2.21, "learning_rate": 2.4526541788603834e-05, "loss": 0.179, "step": 22001 }, { "epoch": 2.21, "learning_rate": 2.4524188805475626e-05, "loss": 0.3758, "step": 22002 }, { "epoch": 2.21, "learning_rate": 2.4521835857193387e-05, "loss": 0.2595, "step": 22003 }, { "epoch": 2.21, "learning_rate": 2.4519482943772108e-05, "loss": 0.2377, "step": 22004 }, { "epoch": 2.21, "learning_rate": 2.4517130065226763e-05, "loss": 0.1874, "step": 22005 }, { "epoch": 2.21, "learning_rate": 2.4514777221572304e-05, "loss": 0.1954, "step": 22006 }, { "epoch": 2.21, "learning_rate": 2.451242441282372e-05, "loss": 0.1235, "step": 22007 }, { "epoch": 2.21, "learning_rate": 2.4510071638995982e-05, "loss": 0.1531, "step": 22008 }, { "epoch": 2.21, "learning_rate": 2.4507718900104052e-05, "loss": 0.185, "step": 22009 }, { "epoch": 2.21, "learning_rate": 2.4505366196162918e-05, "loss": 0.2351, "step": 22010 }, { "epoch": 2.21, "learning_rate": 2.4503013527187544e-05, "loss": 0.7938, "step": 22011 }, { "epoch": 2.21, "learning_rate": 2.4500660893192893e-05, "loss": 0.4856, "step": 22012 }, { "epoch": 2.21, "learning_rate": 2.4498308294193947e-05, "loss": 0.3154, "step": 22013 }, { "epoch": 2.21, "learning_rate": 2.4495955730205685e-05, "loss": 0.4968, "step": 22014 }, { "epoch": 2.21, "learning_rate": 2.4493603201243048e-05, "loss": 0.4984, "step": 22015 }, { "epoch": 2.21, "learning_rate": 2.449125070732104e-05, "loss": 0.413, "step": 22016 }, { "epoch": 2.21, "learning_rate": 2.44888982484546e-05, "loss": 0.5078, "step": 22017 }, { "epoch": 2.21, "learning_rate": 2.4486545824658728e-05, "loss": 0.4045, "step": 22018 }, { "epoch": 2.21, "learning_rate": 2.448419343594839e-05, "loss": 0.3406, "step": 22019 }, { "epoch": 2.21, "learning_rate": 2.4481841082338523e-05, "loss": 0.3682, "step": 22020 }, { "epoch": 2.21, "learning_rate": 2.447948876384414e-05, "loss": 0.3471, "step": 22021 }, { "epoch": 2.21, "learning_rate": 2.4477136480480183e-05, "loss": 0.5491, "step": 22022 }, { "epoch": 2.21, "learning_rate": 2.4474784232261626e-05, "loss": 0.4924, "step": 22023 }, { "epoch": 2.21, "learning_rate": 2.4472432019203447e-05, "loss": 0.4805, "step": 22024 }, { "epoch": 2.21, "learning_rate": 2.447007984132061e-05, "loss": 0.0962, "step": 22025 }, { "epoch": 2.21, "learning_rate": 2.4467727698628065e-05, "loss": 0.5605, "step": 22026 }, { "epoch": 2.21, "learning_rate": 2.446537559114081e-05, "loss": 0.2105, "step": 22027 }, { "epoch": 2.21, "learning_rate": 2.4463023518873802e-05, "loss": 0.6356, "step": 22028 }, { "epoch": 2.21, "learning_rate": 2.4460671481841995e-05, "loss": 0.427, "step": 22029 }, { "epoch": 2.21, "learning_rate": 2.4458319480060373e-05, "loss": 0.8763, "step": 22030 }, { "epoch": 2.21, "learning_rate": 2.4455967513543896e-05, "loss": 0.3577, "step": 22031 }, { "epoch": 2.21, "learning_rate": 2.445361558230753e-05, "loss": 0.408, "step": 22032 }, { "epoch": 2.21, "learning_rate": 2.4451263686366253e-05, "loss": 0.1682, "step": 22033 }, { "epoch": 2.21, "learning_rate": 2.444891182573502e-05, "loss": 0.2037, "step": 22034 }, { "epoch": 2.21, "learning_rate": 2.4446560000428796e-05, "loss": 0.176, "step": 22035 }, { "epoch": 2.21, "learning_rate": 2.444420821046255e-05, "loss": 0.2977, "step": 22036 }, { "epoch": 2.21, "learning_rate": 2.4441856455851262e-05, "loss": 0.1829, "step": 22037 }, { "epoch": 2.21, "learning_rate": 2.4439504736609873e-05, "loss": 0.1368, "step": 22038 }, { "epoch": 2.21, "learning_rate": 2.4437153052753366e-05, "loss": 0.4789, "step": 22039 }, { "epoch": 2.21, "learning_rate": 2.4434801404296698e-05, "loss": 0.2192, "step": 22040 }, { "epoch": 2.21, "learning_rate": 2.4432449791254832e-05, "loss": 0.2969, "step": 22041 }, { "epoch": 2.21, "learning_rate": 2.443009821364275e-05, "loss": 0.2421, "step": 22042 }, { "epoch": 2.21, "learning_rate": 2.4427746671475397e-05, "loss": 0.097, "step": 22043 }, { "epoch": 2.21, "learning_rate": 2.442539516476774e-05, "loss": 0.0444, "step": 22044 }, { "epoch": 2.21, "learning_rate": 2.4423043693534752e-05, "loss": 0.1261, "step": 22045 }, { "epoch": 2.21, "learning_rate": 2.4420692257791387e-05, "loss": 0.1576, "step": 22046 }, { "epoch": 2.21, "learning_rate": 2.4418340857552628e-05, "loss": 0.1842, "step": 22047 }, { "epoch": 2.21, "learning_rate": 2.4415989492833416e-05, "loss": 0.2391, "step": 22048 }, { "epoch": 2.21, "learning_rate": 2.4413638163648717e-05, "loss": 0.705, "step": 22049 }, { "epoch": 2.21, "learning_rate": 2.4411286870013506e-05, "loss": 0.4689, "step": 22050 }, { "epoch": 2.21, "learning_rate": 2.4408935611942746e-05, "loss": 0.446, "step": 22051 }, { "epoch": 2.21, "learning_rate": 2.4406584389451374e-05, "loss": 0.7897, "step": 22052 }, { "epoch": 2.21, "learning_rate": 2.4404233202554386e-05, "loss": 0.1352, "step": 22053 }, { "epoch": 2.21, "learning_rate": 2.440188205126673e-05, "loss": 0.2474, "step": 22054 }, { "epoch": 2.21, "learning_rate": 2.4399530935603355e-05, "loss": 0.2831, "step": 22055 }, { "epoch": 2.21, "learning_rate": 2.439717985557925e-05, "loss": 0.2842, "step": 22056 }, { "epoch": 2.21, "learning_rate": 2.4394828811209355e-05, "loss": 0.0962, "step": 22057 }, { "epoch": 2.21, "learning_rate": 2.4392477802508628e-05, "loss": 0.3842, "step": 22058 }, { "epoch": 2.21, "learning_rate": 2.4390126829492047e-05, "loss": 0.3698, "step": 22059 }, { "epoch": 2.21, "learning_rate": 2.438777589217457e-05, "loss": 0.1986, "step": 22060 }, { "epoch": 2.21, "learning_rate": 2.438542499057114e-05, "loss": 0.2678, "step": 22061 }, { "epoch": 2.21, "learning_rate": 2.4383074124696737e-05, "loss": 0.4209, "step": 22062 }, { "epoch": 2.21, "learning_rate": 2.4380723294566314e-05, "loss": 0.7271, "step": 22063 }, { "epoch": 2.21, "learning_rate": 2.4378372500194823e-05, "loss": 0.957, "step": 22064 }, { "epoch": 2.21, "learning_rate": 2.4376021741597243e-05, "loss": 0.2298, "step": 22065 }, { "epoch": 2.22, "learning_rate": 2.4373671018788516e-05, "loss": 0.2757, "step": 22066 }, { "epoch": 2.22, "learning_rate": 2.43713203317836e-05, "loss": 0.3104, "step": 22067 }, { "epoch": 2.22, "learning_rate": 2.4368969680597463e-05, "loss": 0.4234, "step": 22068 }, { "epoch": 2.22, "learning_rate": 2.436661906524507e-05, "loss": 0.2425, "step": 22069 }, { "epoch": 2.22, "learning_rate": 2.436426848574136e-05, "loss": 0.3209, "step": 22070 }, { "epoch": 2.22, "learning_rate": 2.4361917942101304e-05, "loss": 0.3434, "step": 22071 }, { "epoch": 2.22, "learning_rate": 2.435956743433985e-05, "loss": 0.696, "step": 22072 }, { "epoch": 2.22, "learning_rate": 2.4357216962471973e-05, "loss": 0.1294, "step": 22073 }, { "epoch": 2.22, "learning_rate": 2.4354866526512624e-05, "loss": 0.4435, "step": 22074 }, { "epoch": 2.22, "learning_rate": 2.4352516126476742e-05, "loss": 0.5149, "step": 22075 }, { "epoch": 2.22, "learning_rate": 2.4350165762379313e-05, "loss": 0.4518, "step": 22076 }, { "epoch": 2.22, "learning_rate": 2.4347815434235274e-05, "loss": 0.2758, "step": 22077 }, { "epoch": 2.22, "learning_rate": 2.4345465142059585e-05, "loss": 0.1575, "step": 22078 }, { "epoch": 2.22, "learning_rate": 2.4343114885867218e-05, "loss": 0.114, "step": 22079 }, { "epoch": 2.22, "learning_rate": 2.4340764665673104e-05, "loss": 0.1108, "step": 22080 }, { "epoch": 2.22, "learning_rate": 2.4338414481492205e-05, "loss": 0.4949, "step": 22081 }, { "epoch": 2.22, "learning_rate": 2.4336064333339493e-05, "loss": 0.0718, "step": 22082 }, { "epoch": 2.22, "learning_rate": 2.4333714221229917e-05, "loss": 0.5341, "step": 22083 }, { "epoch": 2.22, "learning_rate": 2.433136414517841e-05, "loss": 0.3153, "step": 22084 }, { "epoch": 2.22, "learning_rate": 2.4329014105199963e-05, "loss": 0.295, "step": 22085 }, { "epoch": 2.22, "learning_rate": 2.4326664101309508e-05, "loss": 0.1615, "step": 22086 }, { "epoch": 2.22, "learning_rate": 2.4324314133521993e-05, "loss": 0.2183, "step": 22087 }, { "epoch": 2.22, "learning_rate": 2.4321964201852403e-05, "loss": 0.1643, "step": 22088 }, { "epoch": 2.22, "learning_rate": 2.4319614306315664e-05, "loss": 0.2019, "step": 22089 }, { "epoch": 2.22, "learning_rate": 2.4317264446926733e-05, "loss": 0.5605, "step": 22090 }, { "epoch": 2.22, "learning_rate": 2.4314914623700573e-05, "loss": 0.0703, "step": 22091 }, { "epoch": 2.22, "learning_rate": 2.4312564836652145e-05, "loss": 0.0814, "step": 22092 }, { "epoch": 2.22, "learning_rate": 2.4310215085796375e-05, "loss": 0.6323, "step": 22093 }, { "epoch": 2.22, "learning_rate": 2.4307865371148234e-05, "loss": 0.4228, "step": 22094 }, { "epoch": 2.22, "learning_rate": 2.430551569272268e-05, "loss": 0.3223, "step": 22095 }, { "epoch": 2.22, "learning_rate": 2.4303166050534648e-05, "loss": 0.3096, "step": 22096 }, { "epoch": 2.22, "learning_rate": 2.430081644459911e-05, "loss": 0.425, "step": 22097 }, { "epoch": 2.22, "learning_rate": 2.4298466874931006e-05, "loss": 0.6762, "step": 22098 }, { "epoch": 2.22, "learning_rate": 2.429611734154528e-05, "loss": 0.683, "step": 22099 }, { "epoch": 2.22, "learning_rate": 2.4293767844456904e-05, "loss": 0.3837, "step": 22100 }, { "epoch": 2.22, "learning_rate": 2.429141838368081e-05, "loss": 0.0899, "step": 22101 }, { "epoch": 2.22, "learning_rate": 2.4289068959231968e-05, "loss": 0.4541, "step": 22102 }, { "epoch": 2.22, "learning_rate": 2.4286719571125315e-05, "loss": 0.5962, "step": 22103 }, { "epoch": 2.22, "learning_rate": 2.42843702193758e-05, "loss": 0.4283, "step": 22104 }, { "epoch": 2.22, "learning_rate": 2.4282020903998383e-05, "loss": 0.1066, "step": 22105 }, { "epoch": 2.22, "learning_rate": 2.4279671625008012e-05, "loss": 0.1374, "step": 22106 }, { "epoch": 2.22, "learning_rate": 2.4277322382419623e-05, "loss": 0.4213, "step": 22107 }, { "epoch": 2.22, "learning_rate": 2.4274973176248193e-05, "loss": 0.374, "step": 22108 }, { "epoch": 2.22, "learning_rate": 2.427262400650865e-05, "loss": 0.3183, "step": 22109 }, { "epoch": 2.22, "learning_rate": 2.427027487321594e-05, "loss": 0.221, "step": 22110 }, { "epoch": 2.22, "learning_rate": 2.426792577638504e-05, "loss": 0.2435, "step": 22111 }, { "epoch": 2.22, "learning_rate": 2.4265576716030867e-05, "loss": 0.45, "step": 22112 }, { "epoch": 2.22, "learning_rate": 2.426322769216838e-05, "loss": 0.2014, "step": 22113 }, { "epoch": 2.22, "learning_rate": 2.4260878704812534e-05, "loss": 0.1108, "step": 22114 }, { "epoch": 2.22, "learning_rate": 2.4258529753978278e-05, "loss": 0.2952, "step": 22115 }, { "epoch": 2.22, "learning_rate": 2.4256180839680537e-05, "loss": 0.176, "step": 22116 }, { "epoch": 2.22, "learning_rate": 2.4253831961934298e-05, "loss": 0.263, "step": 22117 }, { "epoch": 2.22, "learning_rate": 2.4251483120754477e-05, "loss": 0.928, "step": 22118 }, { "epoch": 2.22, "learning_rate": 2.424913431615602e-05, "loss": 0.2598, "step": 22119 }, { "epoch": 2.22, "learning_rate": 2.4246785548153904e-05, "loss": 0.1302, "step": 22120 }, { "epoch": 2.22, "learning_rate": 2.424443681676305e-05, "loss": 0.2969, "step": 22121 }, { "epoch": 2.22, "learning_rate": 2.42420881219984e-05, "loss": 0.2987, "step": 22122 }, { "epoch": 2.22, "learning_rate": 2.4239739463874918e-05, "loss": 0.3104, "step": 22123 }, { "epoch": 2.22, "learning_rate": 2.423739084240755e-05, "loss": 0.0464, "step": 22124 }, { "epoch": 2.22, "learning_rate": 2.423504225761122e-05, "loss": 0.5548, "step": 22125 }, { "epoch": 2.22, "learning_rate": 2.4232693709500896e-05, "loss": 0.1906, "step": 22126 }, { "epoch": 2.22, "learning_rate": 2.4230345198091506e-05, "loss": 0.3955, "step": 22127 }, { "epoch": 2.22, "learning_rate": 2.4227996723398017e-05, "loss": 0.2627, "step": 22128 }, { "epoch": 2.22, "learning_rate": 2.4225648285435363e-05, "loss": 0.59, "step": 22129 }, { "epoch": 2.22, "learning_rate": 2.422329988421847e-05, "loss": 0.2186, "step": 22130 }, { "epoch": 2.22, "learning_rate": 2.4220951519762317e-05, "loss": 0.6562, "step": 22131 }, { "epoch": 2.22, "learning_rate": 2.421860319208182e-05, "loss": 0.6601, "step": 22132 }, { "epoch": 2.22, "learning_rate": 2.4216254901191934e-05, "loss": 0.4773, "step": 22133 }, { "epoch": 2.22, "learning_rate": 2.421390664710761e-05, "loss": 0.2883, "step": 22134 }, { "epoch": 2.22, "learning_rate": 2.421155842984377e-05, "loss": 0.5291, "step": 22135 }, { "epoch": 2.22, "learning_rate": 2.4209210249415373e-05, "loss": 0.6663, "step": 22136 }, { "epoch": 2.22, "learning_rate": 2.4206862105837363e-05, "loss": 0.0511, "step": 22137 }, { "epoch": 2.22, "learning_rate": 2.4204513999124686e-05, "loss": 0.655, "step": 22138 }, { "epoch": 2.22, "learning_rate": 2.4202165929292255e-05, "loss": 0.4053, "step": 22139 }, { "epoch": 2.22, "learning_rate": 2.4199817896355057e-05, "loss": 0.3232, "step": 22140 }, { "epoch": 2.22, "learning_rate": 2.4197469900328e-05, "loss": 0.7686, "step": 22141 }, { "epoch": 2.22, "learning_rate": 2.4195121941226035e-05, "loss": 0.0706, "step": 22142 }, { "epoch": 2.22, "learning_rate": 2.419277401906412e-05, "loss": 0.1738, "step": 22143 }, { "epoch": 2.22, "learning_rate": 2.4190426133857175e-05, "loss": 0.0824, "step": 22144 }, { "epoch": 2.22, "learning_rate": 2.4188078285620136e-05, "loss": 0.7035, "step": 22145 }, { "epoch": 2.22, "learning_rate": 2.4185730474367968e-05, "loss": 0.1895, "step": 22146 }, { "epoch": 2.22, "learning_rate": 2.4183382700115605e-05, "loss": 0.1408, "step": 22147 }, { "epoch": 2.22, "learning_rate": 2.418103496287796e-05, "loss": 0.1297, "step": 22148 }, { "epoch": 2.22, "learning_rate": 2.4178687262670017e-05, "loss": 0.6326, "step": 22149 }, { "epoch": 2.22, "learning_rate": 2.4176339599506685e-05, "loss": 0.2725, "step": 22150 }, { "epoch": 2.22, "learning_rate": 2.4173991973402905e-05, "loss": 0.8244, "step": 22151 }, { "epoch": 2.22, "learning_rate": 2.417164438437364e-05, "loss": 0.5078, "step": 22152 }, { "epoch": 2.22, "learning_rate": 2.416929683243379e-05, "loss": 0.3459, "step": 22153 }, { "epoch": 2.22, "learning_rate": 2.4166949317598342e-05, "loss": 0.7958, "step": 22154 }, { "epoch": 2.22, "learning_rate": 2.41646018398822e-05, "loss": 0.1612, "step": 22155 }, { "epoch": 2.22, "learning_rate": 2.4162254399300305e-05, "loss": 0.3679, "step": 22156 }, { "epoch": 2.22, "learning_rate": 2.4159906995867612e-05, "loss": 0.5357, "step": 22157 }, { "epoch": 2.22, "learning_rate": 2.4157559629599046e-05, "loss": 0.529, "step": 22158 }, { "epoch": 2.22, "learning_rate": 2.415521230050954e-05, "loss": 0.0256, "step": 22159 }, { "epoch": 2.22, "learning_rate": 2.4152865008614045e-05, "loss": 0.7181, "step": 22160 }, { "epoch": 2.22, "learning_rate": 2.4150517753927504e-05, "loss": 0.1138, "step": 22161 }, { "epoch": 2.22, "learning_rate": 2.414817053646482e-05, "loss": 0.2881, "step": 22162 }, { "epoch": 2.22, "learning_rate": 2.414582335624097e-05, "loss": 0.265, "step": 22163 }, { "epoch": 2.23, "learning_rate": 2.414347621327087e-05, "loss": 0.3253, "step": 22164 }, { "epoch": 2.23, "learning_rate": 2.4141129107569455e-05, "loss": 0.0975, "step": 22165 }, { "epoch": 2.23, "learning_rate": 2.4138782039151673e-05, "loss": 0.1941, "step": 22166 }, { "epoch": 2.23, "learning_rate": 2.4136435008032448e-05, "loss": 0.1594, "step": 22167 }, { "epoch": 2.23, "learning_rate": 2.4134088014226715e-05, "loss": 0.5369, "step": 22168 }, { "epoch": 2.23, "learning_rate": 2.413174105774942e-05, "loss": 0.2477, "step": 22169 }, { "epoch": 2.23, "learning_rate": 2.41293941386155e-05, "loss": 0.2342, "step": 22170 }, { "epoch": 2.23, "learning_rate": 2.412704725683986e-05, "loss": 0.5836, "step": 22171 }, { "epoch": 2.23, "learning_rate": 2.4124700412437475e-05, "loss": 0.1184, "step": 22172 }, { "epoch": 2.23, "learning_rate": 2.4122353605423263e-05, "loss": 0.3513, "step": 22173 }, { "epoch": 2.23, "learning_rate": 2.4120006835812145e-05, "loss": 0.4271, "step": 22174 }, { "epoch": 2.23, "learning_rate": 2.411766010361908e-05, "loss": 0.133, "step": 22175 }, { "epoch": 2.23, "learning_rate": 2.411531340885898e-05, "loss": 0.6044, "step": 22176 }, { "epoch": 2.23, "learning_rate": 2.4112966751546786e-05, "loss": 0.4228, "step": 22177 }, { "epoch": 2.23, "learning_rate": 2.4110620131697432e-05, "loss": 0.1215, "step": 22178 }, { "epoch": 2.23, "learning_rate": 2.410827354932586e-05, "loss": 0.155, "step": 22179 }, { "epoch": 2.23, "learning_rate": 2.4105927004446973e-05, "loss": 0.2941, "step": 22180 }, { "epoch": 2.23, "learning_rate": 2.4103580497075748e-05, "loss": 0.6154, "step": 22181 }, { "epoch": 2.23, "learning_rate": 2.410123402722708e-05, "loss": 0.4306, "step": 22182 }, { "epoch": 2.23, "learning_rate": 2.4098887594915917e-05, "loss": 0.2333, "step": 22183 }, { "epoch": 2.23, "learning_rate": 2.40965412001572e-05, "loss": 0.4378, "step": 22184 }, { "epoch": 2.23, "learning_rate": 2.4094194842965828e-05, "loss": 0.7075, "step": 22185 }, { "epoch": 2.23, "learning_rate": 2.409184852335677e-05, "loss": 0.3103, "step": 22186 }, { "epoch": 2.23, "learning_rate": 2.4089502241344934e-05, "loss": 0.1679, "step": 22187 }, { "epoch": 2.23, "learning_rate": 2.4087155996945255e-05, "loss": 0.2684, "step": 22188 }, { "epoch": 2.23, "learning_rate": 2.4084809790172674e-05, "loss": 0.1852, "step": 22189 }, { "epoch": 2.23, "learning_rate": 2.408246362104211e-05, "loss": 0.2919, "step": 22190 }, { "epoch": 2.23, "learning_rate": 2.4080117489568486e-05, "loss": 0.0485, "step": 22191 }, { "epoch": 2.23, "learning_rate": 2.4077771395766752e-05, "loss": 0.3257, "step": 22192 }, { "epoch": 2.23, "learning_rate": 2.4075425339651835e-05, "loss": 0.5685, "step": 22193 }, { "epoch": 2.23, "learning_rate": 2.407307932123864e-05, "loss": 0.4595, "step": 22194 }, { "epoch": 2.23, "learning_rate": 2.4070733340542126e-05, "loss": 0.2814, "step": 22195 }, { "epoch": 2.23, "learning_rate": 2.4068387397577204e-05, "loss": 0.7116, "step": 22196 }, { "epoch": 2.23, "learning_rate": 2.4066041492358807e-05, "loss": 0.1097, "step": 22197 }, { "epoch": 2.23, "learning_rate": 2.4063695624901875e-05, "loss": 0.706, "step": 22198 }, { "epoch": 2.23, "learning_rate": 2.4061349795221318e-05, "loss": 0.2875, "step": 22199 }, { "epoch": 2.23, "learning_rate": 2.4059004003332067e-05, "loss": 0.3782, "step": 22200 }, { "epoch": 2.23, "learning_rate": 2.405665824924906e-05, "loss": 0.3215, "step": 22201 }, { "epoch": 2.23, "learning_rate": 2.4054312532987226e-05, "loss": 0.212, "step": 22202 }, { "epoch": 2.23, "learning_rate": 2.4051966854561467e-05, "loss": 0.7442, "step": 22203 }, { "epoch": 2.23, "learning_rate": 2.4049621213986745e-05, "loss": 0.1586, "step": 22204 }, { "epoch": 2.23, "learning_rate": 2.4047275611277963e-05, "loss": 0.5407, "step": 22205 }, { "epoch": 2.23, "learning_rate": 2.404493004645005e-05, "loss": 0.3295, "step": 22206 }, { "epoch": 2.23, "learning_rate": 2.404258451951795e-05, "loss": 0.5004, "step": 22207 }, { "epoch": 2.23, "learning_rate": 2.4040239030496555e-05, "loss": 0.2304, "step": 22208 }, { "epoch": 2.23, "learning_rate": 2.4037893579400832e-05, "loss": 0.4563, "step": 22209 }, { "epoch": 2.23, "learning_rate": 2.403554816624568e-05, "loss": 0.2096, "step": 22210 }, { "epoch": 2.23, "learning_rate": 2.4033202791046022e-05, "loss": 0.6082, "step": 22211 }, { "epoch": 2.23, "learning_rate": 2.40308574538168e-05, "loss": 0.5364, "step": 22212 }, { "epoch": 2.23, "learning_rate": 2.4028512154572937e-05, "loss": 0.5591, "step": 22213 }, { "epoch": 2.23, "learning_rate": 2.4026166893329336e-05, "loss": 0.3577, "step": 22214 }, { "epoch": 2.23, "learning_rate": 2.4023821670100944e-05, "loss": 1.1394, "step": 22215 }, { "epoch": 2.23, "learning_rate": 2.4021476484902684e-05, "loss": 0.1805, "step": 22216 }, { "epoch": 2.23, "learning_rate": 2.401913133774945e-05, "loss": 0.3424, "step": 22217 }, { "epoch": 2.23, "learning_rate": 2.4016786228656217e-05, "loss": 0.8815, "step": 22218 }, { "epoch": 2.23, "learning_rate": 2.4014441157637867e-05, "loss": 0.149, "step": 22219 }, { "epoch": 2.23, "learning_rate": 2.4012096124709334e-05, "loss": 0.6955, "step": 22220 }, { "epoch": 2.23, "learning_rate": 2.400975112988555e-05, "loss": 0.1781, "step": 22221 }, { "epoch": 2.23, "learning_rate": 2.400740617318143e-05, "loss": 0.1212, "step": 22222 }, { "epoch": 2.23, "learning_rate": 2.4005061254611892e-05, "loss": 0.1469, "step": 22223 }, { "epoch": 2.23, "learning_rate": 2.4002716374191868e-05, "loss": 0.2173, "step": 22224 }, { "epoch": 2.23, "learning_rate": 2.4000371531936283e-05, "loss": 0.6648, "step": 22225 }, { "epoch": 2.23, "learning_rate": 2.3998026727860027e-05, "loss": 0.3822, "step": 22226 }, { "epoch": 2.23, "learning_rate": 2.3995681961978072e-05, "loss": 0.882, "step": 22227 }, { "epoch": 2.23, "learning_rate": 2.39933372343053e-05, "loss": 0.1094, "step": 22228 }, { "epoch": 2.23, "learning_rate": 2.399099254485664e-05, "loss": 0.2721, "step": 22229 }, { "epoch": 2.23, "learning_rate": 2.3988647893647035e-05, "loss": 0.078, "step": 22230 }, { "epoch": 2.23, "learning_rate": 2.3986303280691376e-05, "loss": 0.3306, "step": 22231 }, { "epoch": 2.23, "learning_rate": 2.3983958706004595e-05, "loss": 0.264, "step": 22232 }, { "epoch": 2.23, "learning_rate": 2.3981614169601613e-05, "loss": 0.1107, "step": 22233 }, { "epoch": 2.23, "learning_rate": 2.3979269671497344e-05, "loss": 0.3865, "step": 22234 }, { "epoch": 2.23, "learning_rate": 2.397692521170672e-05, "loss": 0.1667, "step": 22235 }, { "epoch": 2.23, "learning_rate": 2.3974580790244654e-05, "loss": 0.0902, "step": 22236 }, { "epoch": 2.23, "learning_rate": 2.3972236407126057e-05, "loss": 0.4623, "step": 22237 }, { "epoch": 2.23, "learning_rate": 2.3969892062365857e-05, "loss": 0.3276, "step": 22238 }, { "epoch": 2.23, "learning_rate": 2.396754775597898e-05, "loss": 0.7771, "step": 22239 }, { "epoch": 2.23, "learning_rate": 2.3965203487980313e-05, "loss": 0.3177, "step": 22240 }, { "epoch": 2.23, "learning_rate": 2.3962859258384814e-05, "loss": 0.4172, "step": 22241 }, { "epoch": 2.23, "learning_rate": 2.3960515067207375e-05, "loss": 0.3732, "step": 22242 }, { "epoch": 2.23, "learning_rate": 2.3958170914462914e-05, "loss": 0.1559, "step": 22243 }, { "epoch": 2.23, "learning_rate": 2.3955826800166363e-05, "loss": 0.5037, "step": 22244 }, { "epoch": 2.23, "learning_rate": 2.395348272433264e-05, "loss": 0.2367, "step": 22245 }, { "epoch": 2.23, "learning_rate": 2.395113868697664e-05, "loss": 0.3845, "step": 22246 }, { "epoch": 2.23, "learning_rate": 2.394879468811329e-05, "loss": 0.1018, "step": 22247 }, { "epoch": 2.23, "learning_rate": 2.3946450727757523e-05, "loss": 0.81, "step": 22248 }, { "epoch": 2.23, "learning_rate": 2.3944106805924215e-05, "loss": 0.1074, "step": 22249 }, { "epoch": 2.23, "learning_rate": 2.3941762922628335e-05, "loss": 0.6494, "step": 22250 }, { "epoch": 2.23, "learning_rate": 2.3939419077884756e-05, "loss": 0.7665, "step": 22251 }, { "epoch": 2.23, "learning_rate": 2.393707527170841e-05, "loss": 0.3789, "step": 22252 }, { "epoch": 2.23, "learning_rate": 2.3934731504114218e-05, "loss": 0.2514, "step": 22253 }, { "epoch": 2.23, "learning_rate": 2.393238777511708e-05, "loss": 0.1203, "step": 22254 }, { "epoch": 2.23, "learning_rate": 2.3930044084731918e-05, "loss": 0.8006, "step": 22255 }, { "epoch": 2.23, "learning_rate": 2.392770043297365e-05, "loss": 0.2581, "step": 22256 }, { "epoch": 2.23, "learning_rate": 2.3925356819857192e-05, "loss": 0.0424, "step": 22257 }, { "epoch": 2.23, "learning_rate": 2.392301324539743e-05, "loss": 0.1975, "step": 22258 }, { "epoch": 2.23, "learning_rate": 2.3920669709609324e-05, "loss": 0.3755, "step": 22259 }, { "epoch": 2.23, "learning_rate": 2.3918326212507755e-05, "loss": 0.1626, "step": 22260 }, { "epoch": 2.23, "learning_rate": 2.3915982754107636e-05, "loss": 0.3075, "step": 22261 }, { "epoch": 2.23, "learning_rate": 2.3913639334423903e-05, "loss": 0.0939, "step": 22262 }, { "epoch": 2.24, "learning_rate": 2.3911295953471435e-05, "loss": 0.3525, "step": 22263 }, { "epoch": 2.24, "learning_rate": 2.3908952611265182e-05, "loss": 0.536, "step": 22264 }, { "epoch": 2.24, "learning_rate": 2.390660930782003e-05, "loss": 0.1519, "step": 22265 }, { "epoch": 2.24, "learning_rate": 2.3904266043150895e-05, "loss": 0.2199, "step": 22266 }, { "epoch": 2.24, "learning_rate": 2.3901922817272698e-05, "loss": 0.56, "step": 22267 }, { "epoch": 2.24, "learning_rate": 2.3899579630200346e-05, "loss": 0.3749, "step": 22268 }, { "epoch": 2.24, "learning_rate": 2.389723648194874e-05, "loss": 0.2107, "step": 22269 }, { "epoch": 2.24, "learning_rate": 2.3894893372532803e-05, "loss": 0.6409, "step": 22270 }, { "epoch": 2.24, "learning_rate": 2.389255030196745e-05, "loss": 0.664, "step": 22271 }, { "epoch": 2.24, "learning_rate": 2.3890207270267566e-05, "loss": 0.1698, "step": 22272 }, { "epoch": 2.24, "learning_rate": 2.38878642774481e-05, "loss": 0.4501, "step": 22273 }, { "epoch": 2.24, "learning_rate": 2.3885521323523927e-05, "loss": 0.1655, "step": 22274 }, { "epoch": 2.24, "learning_rate": 2.3883178408509965e-05, "loss": 1.0484, "step": 22275 }, { "epoch": 2.24, "learning_rate": 2.388083553242114e-05, "loss": 0.1413, "step": 22276 }, { "epoch": 2.24, "learning_rate": 2.387849269527235e-05, "loss": 0.4758, "step": 22277 }, { "epoch": 2.24, "learning_rate": 2.3876149897078497e-05, "loss": 0.1683, "step": 22278 }, { "epoch": 2.24, "learning_rate": 2.38738071378545e-05, "loss": 0.3255, "step": 22279 }, { "epoch": 2.24, "learning_rate": 2.387146441761527e-05, "loss": 0.8134, "step": 22280 }, { "epoch": 2.24, "learning_rate": 2.3869121736375695e-05, "loss": 0.1435, "step": 22281 }, { "epoch": 2.24, "learning_rate": 2.386677909415071e-05, "loss": 0.1204, "step": 22282 }, { "epoch": 2.24, "learning_rate": 2.3864436490955204e-05, "loss": 0.4059, "step": 22283 }, { "epoch": 2.24, "learning_rate": 2.3862093926804085e-05, "loss": 0.1465, "step": 22284 }, { "epoch": 2.24, "learning_rate": 2.3859751401712274e-05, "loss": 0.5484, "step": 22285 }, { "epoch": 2.24, "learning_rate": 2.385740891569467e-05, "loss": 0.2253, "step": 22286 }, { "epoch": 2.24, "learning_rate": 2.3855066468766168e-05, "loss": 0.6534, "step": 22287 }, { "epoch": 2.24, "learning_rate": 2.385272406094169e-05, "loss": 0.3427, "step": 22288 }, { "epoch": 2.24, "learning_rate": 2.3850381692236137e-05, "loss": 0.1746, "step": 22289 }, { "epoch": 2.24, "learning_rate": 2.384803936266442e-05, "loss": 0.6775, "step": 22290 }, { "epoch": 2.24, "learning_rate": 2.384569707224144e-05, "loss": 0.361, "step": 22291 }, { "epoch": 2.24, "learning_rate": 2.3843354820982097e-05, "loss": 0.2208, "step": 22292 }, { "epoch": 2.24, "learning_rate": 2.384101260890131e-05, "loss": 0.4674, "step": 22293 }, { "epoch": 2.24, "learning_rate": 2.3838670436013974e-05, "loss": 0.0891, "step": 22294 }, { "epoch": 2.24, "learning_rate": 2.3836328302334984e-05, "loss": 0.0824, "step": 22295 }, { "epoch": 2.24, "learning_rate": 2.383398620787927e-05, "loss": 0.4284, "step": 22296 }, { "epoch": 2.24, "learning_rate": 2.3831644152661718e-05, "loss": 0.2326, "step": 22297 }, { "epoch": 2.24, "learning_rate": 2.3829302136697228e-05, "loss": 0.9035, "step": 22298 }, { "epoch": 2.24, "learning_rate": 2.382696016000072e-05, "loss": 0.5151, "step": 22299 }, { "epoch": 2.24, "learning_rate": 2.3824618222587094e-05, "loss": 0.3136, "step": 22300 }, { "epoch": 2.24, "learning_rate": 2.382227632447124e-05, "loss": 0.9785, "step": 22301 }, { "epoch": 2.24, "learning_rate": 2.381993446566807e-05, "loss": 0.1426, "step": 22302 }, { "epoch": 2.24, "learning_rate": 2.3817592646192496e-05, "loss": 0.4225, "step": 22303 }, { "epoch": 2.24, "learning_rate": 2.381525086605939e-05, "loss": 0.3459, "step": 22304 }, { "epoch": 2.24, "learning_rate": 2.381290912528369e-05, "loss": 0.3903, "step": 22305 }, { "epoch": 2.24, "learning_rate": 2.3810567423880283e-05, "loss": 0.344, "step": 22306 }, { "epoch": 2.24, "learning_rate": 2.380822576186406e-05, "loss": 0.4905, "step": 22307 }, { "epoch": 2.24, "learning_rate": 2.380588413924994e-05, "loss": 0.6035, "step": 22308 }, { "epoch": 2.24, "learning_rate": 2.3803542556052825e-05, "loss": 0.1136, "step": 22309 }, { "epoch": 2.24, "learning_rate": 2.3801201012287593e-05, "loss": 0.1936, "step": 22310 }, { "epoch": 2.24, "learning_rate": 2.379885950796917e-05, "loss": 0.1129, "step": 22311 }, { "epoch": 2.24, "learning_rate": 2.3796518043112446e-05, "loss": 0.1356, "step": 22312 }, { "epoch": 2.24, "learning_rate": 2.3794176617732303e-05, "loss": 0.2424, "step": 22313 }, { "epoch": 2.24, "learning_rate": 2.3791835231843682e-05, "loss": 0.3314, "step": 22314 }, { "epoch": 2.24, "learning_rate": 2.3789493885461447e-05, "loss": 0.2369, "step": 22315 }, { "epoch": 2.24, "learning_rate": 2.3787152578600512e-05, "loss": 0.235, "step": 22316 }, { "epoch": 2.24, "learning_rate": 2.378481131127579e-05, "loss": 0.4058, "step": 22317 }, { "epoch": 2.24, "learning_rate": 2.3782470083502137e-05, "loss": 0.5239, "step": 22318 }, { "epoch": 2.24, "learning_rate": 2.3780128895294502e-05, "loss": 0.1027, "step": 22319 }, { "epoch": 2.24, "learning_rate": 2.3777787746667748e-05, "loss": 0.8992, "step": 22320 }, { "epoch": 2.24, "learning_rate": 2.377544663763679e-05, "loss": 0.2755, "step": 22321 }, { "epoch": 2.24, "learning_rate": 2.3773105568216523e-05, "loss": 0.3327, "step": 22322 }, { "epoch": 2.24, "learning_rate": 2.3770764538421844e-05, "loss": 0.4814, "step": 22323 }, { "epoch": 2.24, "learning_rate": 2.3768423548267646e-05, "loss": 0.4733, "step": 22324 }, { "epoch": 2.24, "learning_rate": 2.3766082597768834e-05, "loss": 0.1472, "step": 22325 }, { "epoch": 2.24, "learning_rate": 2.37637416869403e-05, "loss": 0.4104, "step": 22326 }, { "epoch": 2.24, "learning_rate": 2.376140081579693e-05, "loss": 0.2651, "step": 22327 }, { "epoch": 2.24, "learning_rate": 2.375905998435365e-05, "loss": 0.3343, "step": 22328 }, { "epoch": 2.24, "learning_rate": 2.375671919262533e-05, "loss": 0.2781, "step": 22329 }, { "epoch": 2.24, "learning_rate": 2.375437844062687e-05, "loss": 0.2492, "step": 22330 }, { "epoch": 2.24, "learning_rate": 2.3752037728373176e-05, "loss": 0.3836, "step": 22331 }, { "epoch": 2.24, "learning_rate": 2.3749697055879142e-05, "loss": 0.7677, "step": 22332 }, { "epoch": 2.24, "learning_rate": 2.3747356423159644e-05, "loss": 0.1803, "step": 22333 }, { "epoch": 2.24, "learning_rate": 2.3745015830229597e-05, "loss": 0.1882, "step": 22334 }, { "epoch": 2.24, "learning_rate": 2.37426752771039e-05, "loss": 0.2949, "step": 22335 }, { "epoch": 2.24, "learning_rate": 2.3740334763797415e-05, "loss": 0.4388, "step": 22336 }, { "epoch": 2.24, "learning_rate": 2.373799429032508e-05, "loss": 0.7124, "step": 22337 }, { "epoch": 2.24, "learning_rate": 2.3735653856701756e-05, "loss": 0.3182, "step": 22338 }, { "epoch": 2.24, "learning_rate": 2.3733313462942345e-05, "loss": 0.0817, "step": 22339 }, { "epoch": 2.24, "learning_rate": 2.373097310906175e-05, "loss": 0.1935, "step": 22340 }, { "epoch": 2.24, "learning_rate": 2.372863279507486e-05, "loss": 0.509, "step": 22341 }, { "epoch": 2.24, "learning_rate": 2.3726292520996557e-05, "loss": 0.2016, "step": 22342 }, { "epoch": 2.24, "learning_rate": 2.3723952286841745e-05, "loss": 0.3001, "step": 22343 }, { "epoch": 2.24, "learning_rate": 2.372161209262531e-05, "loss": 0.2346, "step": 22344 }, { "epoch": 2.24, "learning_rate": 2.371927193836215e-05, "loss": 0.1676, "step": 22345 }, { "epoch": 2.24, "learning_rate": 2.371693182406716e-05, "loss": 0.0935, "step": 22346 }, { "epoch": 2.24, "learning_rate": 2.371459174975522e-05, "loss": 0.5092, "step": 22347 }, { "epoch": 2.24, "learning_rate": 2.3712251715441225e-05, "loss": 0.0931, "step": 22348 }, { "epoch": 2.24, "learning_rate": 2.370991172114008e-05, "loss": 0.4214, "step": 22349 }, { "epoch": 2.24, "learning_rate": 2.370757176686664e-05, "loss": 0.2399, "step": 22350 }, { "epoch": 2.24, "learning_rate": 2.3705231852635843e-05, "loss": 0.2462, "step": 22351 }, { "epoch": 2.24, "learning_rate": 2.3702891978462545e-05, "loss": 0.5741, "step": 22352 }, { "epoch": 2.24, "learning_rate": 2.3700552144361645e-05, "loss": 0.6836, "step": 22353 }, { "epoch": 2.24, "learning_rate": 2.369821235034804e-05, "loss": 0.4769, "step": 22354 }, { "epoch": 2.24, "learning_rate": 2.3695872596436615e-05, "loss": 0.2131, "step": 22355 }, { "epoch": 2.24, "learning_rate": 2.3693532882642253e-05, "loss": 0.7686, "step": 22356 }, { "epoch": 2.24, "learning_rate": 2.3691193208979852e-05, "loss": 0.5324, "step": 22357 }, { "epoch": 2.24, "learning_rate": 2.3688853575464305e-05, "loss": 0.1289, "step": 22358 }, { "epoch": 2.24, "learning_rate": 2.3686513982110474e-05, "loss": 0.597, "step": 22359 }, { "epoch": 2.24, "learning_rate": 2.3684174428933285e-05, "loss": 0.1929, "step": 22360 }, { "epoch": 2.25, "learning_rate": 2.3681834915947603e-05, "loss": 0.842, "step": 22361 }, { "epoch": 2.25, "learning_rate": 2.3679495443168308e-05, "loss": 0.3204, "step": 22362 }, { "epoch": 2.25, "learning_rate": 2.3677156010610308e-05, "loss": 0.5384, "step": 22363 }, { "epoch": 2.25, "learning_rate": 2.367481661828849e-05, "loss": 0.4675, "step": 22364 }, { "epoch": 2.25, "learning_rate": 2.367247726621772e-05, "loss": 0.0863, "step": 22365 }, { "epoch": 2.25, "learning_rate": 2.3670137954412907e-05, "loss": 0.2159, "step": 22366 }, { "epoch": 2.25, "learning_rate": 2.3667798682888928e-05, "loss": 0.3831, "step": 22367 }, { "epoch": 2.25, "learning_rate": 2.366545945166065e-05, "loss": 0.8284, "step": 22368 }, { "epoch": 2.25, "learning_rate": 2.3663120260743004e-05, "loss": 0.3306, "step": 22369 }, { "epoch": 2.25, "learning_rate": 2.3660781110150835e-05, "loss": 0.2929, "step": 22370 }, { "epoch": 2.25, "learning_rate": 2.365844199989905e-05, "loss": 0.1427, "step": 22371 }, { "epoch": 2.25, "learning_rate": 2.3656102930002522e-05, "loss": 0.374, "step": 22372 }, { "epoch": 2.25, "learning_rate": 2.365376390047614e-05, "loss": 0.0852, "step": 22373 }, { "epoch": 2.25, "learning_rate": 2.3651424911334804e-05, "loss": 0.595, "step": 22374 }, { "epoch": 2.25, "learning_rate": 2.3649085962593377e-05, "loss": 0.1842, "step": 22375 }, { "epoch": 2.25, "learning_rate": 2.364674705426674e-05, "loss": 0.4013, "step": 22376 }, { "epoch": 2.25, "learning_rate": 2.3644408186369802e-05, "loss": 0.3641, "step": 22377 }, { "epoch": 2.25, "learning_rate": 2.3642069358917435e-05, "loss": 0.7494, "step": 22378 }, { "epoch": 2.25, "learning_rate": 2.363973057192451e-05, "loss": 0.0335, "step": 22379 }, { "epoch": 2.25, "learning_rate": 2.3637391825405923e-05, "loss": 0.6768, "step": 22380 }, { "epoch": 2.25, "learning_rate": 2.363505311937656e-05, "loss": 0.5862, "step": 22381 }, { "epoch": 2.25, "learning_rate": 2.3632714453851278e-05, "loss": 0.1009, "step": 22382 }, { "epoch": 2.25, "learning_rate": 2.3630375828845e-05, "loss": 0.2532, "step": 22383 }, { "epoch": 2.25, "learning_rate": 2.3628037244372582e-05, "loss": 0.6914, "step": 22384 }, { "epoch": 2.25, "learning_rate": 2.3625698700448906e-05, "loss": 0.3499, "step": 22385 }, { "epoch": 2.25, "learning_rate": 2.3623360197088858e-05, "loss": 0.6165, "step": 22386 }, { "epoch": 2.25, "learning_rate": 2.362102173430733e-05, "loss": 0.4955, "step": 22387 }, { "epoch": 2.25, "learning_rate": 2.3618683312119184e-05, "loss": 0.3257, "step": 22388 }, { "epoch": 2.25, "learning_rate": 2.3616344930539313e-05, "loss": 0.5158, "step": 22389 }, { "epoch": 2.25, "learning_rate": 2.36140065895826e-05, "loss": 0.594, "step": 22390 }, { "epoch": 2.25, "learning_rate": 2.36116682892639e-05, "loss": 0.3764, "step": 22391 }, { "epoch": 2.25, "learning_rate": 2.3609330029598136e-05, "loss": 0.6655, "step": 22392 }, { "epoch": 2.25, "learning_rate": 2.3606991810600154e-05, "loss": 0.1803, "step": 22393 }, { "epoch": 2.25, "learning_rate": 2.360465363228484e-05, "loss": 0.8698, "step": 22394 }, { "epoch": 2.25, "learning_rate": 2.3602315494667082e-05, "loss": 0.7541, "step": 22395 }, { "epoch": 2.25, "learning_rate": 2.3599977397761747e-05, "loss": 1.4603, "step": 22396 }, { "epoch": 2.25, "learning_rate": 2.3597639341583738e-05, "loss": 0.3324, "step": 22397 }, { "epoch": 2.25, "learning_rate": 2.3595301326147906e-05, "loss": 0.7944, "step": 22398 }, { "epoch": 2.25, "learning_rate": 2.3592963351469137e-05, "loss": 0.333, "step": 22399 }, { "epoch": 2.25, "learning_rate": 2.3590625417562316e-05, "loss": 0.1963, "step": 22400 }, { "epoch": 2.25, "learning_rate": 2.3588287524442327e-05, "loss": 0.129, "step": 22401 }, { "epoch": 2.25, "learning_rate": 2.3585949672124016e-05, "loss": 0.2161, "step": 22402 }, { "epoch": 2.25, "learning_rate": 2.3583611860622296e-05, "loss": 0.7077, "step": 22403 }, { "epoch": 2.25, "learning_rate": 2.3581274089952025e-05, "loss": 0.5402, "step": 22404 }, { "epoch": 2.25, "learning_rate": 2.357893636012808e-05, "loss": 0.0776, "step": 22405 }, { "epoch": 2.25, "learning_rate": 2.357659867116535e-05, "loss": 0.3402, "step": 22406 }, { "epoch": 2.25, "learning_rate": 2.35742610230787e-05, "loss": 0.1173, "step": 22407 }, { "epoch": 2.25, "learning_rate": 2.3571923415883003e-05, "loss": 0.2143, "step": 22408 }, { "epoch": 2.25, "learning_rate": 2.356958584959314e-05, "loss": 0.2488, "step": 22409 }, { "epoch": 2.25, "learning_rate": 2.3567248324223997e-05, "loss": 0.1971, "step": 22410 }, { "epoch": 2.25, "learning_rate": 2.356491083979043e-05, "loss": 0.7693, "step": 22411 }, { "epoch": 2.25, "learning_rate": 2.356257339630732e-05, "loss": 0.6288, "step": 22412 }, { "epoch": 2.25, "learning_rate": 2.3560235993789555e-05, "loss": 0.4874, "step": 22413 }, { "epoch": 2.25, "learning_rate": 2.355789863225198e-05, "loss": 0.1334, "step": 22414 }, { "epoch": 2.25, "learning_rate": 2.3555561311709502e-05, "loss": 0.0413, "step": 22415 }, { "epoch": 2.25, "learning_rate": 2.355322403217698e-05, "loss": 0.1162, "step": 22416 }, { "epoch": 2.25, "learning_rate": 2.355088679366928e-05, "loss": 0.3069, "step": 22417 }, { "epoch": 2.25, "learning_rate": 2.3548549596201288e-05, "loss": 0.5796, "step": 22418 }, { "epoch": 2.25, "learning_rate": 2.3546212439787876e-05, "loss": 0.2977, "step": 22419 }, { "epoch": 2.25, "learning_rate": 2.3543875324443904e-05, "loss": 0.2524, "step": 22420 }, { "epoch": 2.25, "learning_rate": 2.3541538250184255e-05, "loss": 0.8389, "step": 22421 }, { "epoch": 2.25, "learning_rate": 2.3539201217023813e-05, "loss": 0.1238, "step": 22422 }, { "epoch": 2.25, "learning_rate": 2.3536864224977416e-05, "loss": 0.2419, "step": 22423 }, { "epoch": 2.25, "learning_rate": 2.353452727405997e-05, "loss": 0.3357, "step": 22424 }, { "epoch": 2.25, "learning_rate": 2.3532190364286323e-05, "loss": 0.1723, "step": 22425 }, { "epoch": 2.25, "learning_rate": 2.3529853495671366e-05, "loss": 0.6751, "step": 22426 }, { "epoch": 2.25, "learning_rate": 2.3527516668229952e-05, "loss": 0.4296, "step": 22427 }, { "epoch": 2.25, "learning_rate": 2.3525179881976962e-05, "loss": 0.268, "step": 22428 }, { "epoch": 2.25, "learning_rate": 2.3522843136927272e-05, "loss": 0.183, "step": 22429 }, { "epoch": 2.25, "learning_rate": 2.352050643309574e-05, "loss": 0.1936, "step": 22430 }, { "epoch": 2.25, "learning_rate": 2.3518169770497232e-05, "loss": 0.5189, "step": 22431 }, { "epoch": 2.25, "learning_rate": 2.351583314914663e-05, "loss": 0.4196, "step": 22432 }, { "epoch": 2.25, "learning_rate": 2.351349656905881e-05, "loss": 0.4491, "step": 22433 }, { "epoch": 2.25, "learning_rate": 2.3511160030248614e-05, "loss": 0.51, "step": 22434 }, { "epoch": 2.25, "learning_rate": 2.3508823532730937e-05, "loss": 0.1753, "step": 22435 }, { "epoch": 2.25, "learning_rate": 2.3506487076520635e-05, "loss": 0.3775, "step": 22436 }, { "epoch": 2.25, "learning_rate": 2.3504150661632578e-05, "loss": 0.1471, "step": 22437 }, { "epoch": 2.25, "learning_rate": 2.350181428808164e-05, "loss": 0.498, "step": 22438 }, { "epoch": 2.25, "learning_rate": 2.3499477955882683e-05, "loss": 0.309, "step": 22439 }, { "epoch": 2.25, "learning_rate": 2.349714166505057e-05, "loss": 0.4502, "step": 22440 }, { "epoch": 2.25, "learning_rate": 2.3494805415600173e-05, "loss": 0.567, "step": 22441 }, { "epoch": 2.25, "learning_rate": 2.3492469207546366e-05, "loss": 0.6662, "step": 22442 }, { "epoch": 2.25, "learning_rate": 2.3490133040904005e-05, "loss": 0.3568, "step": 22443 }, { "epoch": 2.25, "learning_rate": 2.348779691568796e-05, "loss": 0.2965, "step": 22444 }, { "epoch": 2.25, "learning_rate": 2.348546083191311e-05, "loss": 0.1413, "step": 22445 }, { "epoch": 2.25, "learning_rate": 2.3483124789594286e-05, "loss": 0.3518, "step": 22446 }, { "epoch": 2.25, "learning_rate": 2.3480788788746395e-05, "loss": 0.3137, "step": 22447 }, { "epoch": 2.25, "learning_rate": 2.347845282938428e-05, "loss": 0.0825, "step": 22448 }, { "epoch": 2.25, "learning_rate": 2.3476116911522804e-05, "loss": 0.185, "step": 22449 }, { "epoch": 2.25, "learning_rate": 2.3473781035176845e-05, "loss": 0.2318, "step": 22450 }, { "epoch": 2.25, "learning_rate": 2.3471445200361253e-05, "loss": 0.1999, "step": 22451 }, { "epoch": 2.25, "learning_rate": 2.3469109407090916e-05, "loss": 0.3825, "step": 22452 }, { "epoch": 2.25, "learning_rate": 2.346677365538067e-05, "loss": 0.3376, "step": 22453 }, { "epoch": 2.25, "learning_rate": 2.346443794524539e-05, "loss": 0.2625, "step": 22454 }, { "epoch": 2.25, "learning_rate": 2.3462102276699944e-05, "loss": 0.6419, "step": 22455 }, { "epoch": 2.25, "learning_rate": 2.3459766649759196e-05, "loss": 0.1894, "step": 22456 }, { "epoch": 2.25, "learning_rate": 2.3457431064438e-05, "loss": 0.0983, "step": 22457 }, { "epoch": 2.25, "learning_rate": 2.3455095520751222e-05, "loss": 0.1791, "step": 22458 }, { "epoch": 2.25, "learning_rate": 2.3452760018713733e-05, "loss": 0.4614, "step": 22459 }, { "epoch": 2.26, "learning_rate": 2.345042455834038e-05, "loss": 0.148, "step": 22460 }, { "epoch": 2.26, "learning_rate": 2.3448089139646043e-05, "loss": 0.3512, "step": 22461 }, { "epoch": 2.26, "learning_rate": 2.344575376264557e-05, "loss": 0.4434, "step": 22462 }, { "epoch": 2.26, "learning_rate": 2.3443418427353823e-05, "loss": 0.6606, "step": 22463 }, { "epoch": 2.26, "learning_rate": 2.3441083133785668e-05, "loss": 0.2522, "step": 22464 }, { "epoch": 2.26, "learning_rate": 2.3438747881955975e-05, "loss": 0.3477, "step": 22465 }, { "epoch": 2.26, "learning_rate": 2.3436412671879582e-05, "loss": 1.2348, "step": 22466 }, { "epoch": 2.26, "learning_rate": 2.3434077503571364e-05, "loss": 0.0255, "step": 22467 }, { "epoch": 2.26, "learning_rate": 2.3431742377046186e-05, "loss": 0.1787, "step": 22468 }, { "epoch": 2.26, "learning_rate": 2.342940729231889e-05, "loss": 0.1867, "step": 22469 }, { "epoch": 2.26, "learning_rate": 2.3427072249404357e-05, "loss": 0.5844, "step": 22470 }, { "epoch": 2.26, "learning_rate": 2.3424737248317433e-05, "loss": 0.3286, "step": 22471 }, { "epoch": 2.26, "learning_rate": 2.342240228907297e-05, "loss": 0.7244, "step": 22472 }, { "epoch": 2.26, "learning_rate": 2.342006737168585e-05, "loss": 0.2457, "step": 22473 }, { "epoch": 2.26, "learning_rate": 2.341773249617092e-05, "loss": 0.4281, "step": 22474 }, { "epoch": 2.26, "learning_rate": 2.341539766254302e-05, "loss": 0.3346, "step": 22475 }, { "epoch": 2.26, "learning_rate": 2.3413062870817037e-05, "loss": 0.2531, "step": 22476 }, { "epoch": 2.26, "learning_rate": 2.3410728121007817e-05, "loss": 0.3072, "step": 22477 }, { "epoch": 2.26, "learning_rate": 2.34083934131302e-05, "loss": 0.144, "step": 22478 }, { "epoch": 2.26, "learning_rate": 2.340605874719908e-05, "loss": 0.3572, "step": 22479 }, { "epoch": 2.26, "learning_rate": 2.340372412322928e-05, "loss": 0.4135, "step": 22480 }, { "epoch": 2.26, "learning_rate": 2.3401389541235675e-05, "loss": 0.123, "step": 22481 }, { "epoch": 2.26, "learning_rate": 2.339905500123312e-05, "loss": 0.8124, "step": 22482 }, { "epoch": 2.26, "learning_rate": 2.3396720503236457e-05, "loss": 0.5736, "step": 22483 }, { "epoch": 2.26, "learning_rate": 2.3394386047260572e-05, "loss": 0.2082, "step": 22484 }, { "epoch": 2.26, "learning_rate": 2.339205163332029e-05, "loss": 0.7098, "step": 22485 }, { "epoch": 2.26, "learning_rate": 2.3389717261430472e-05, "loss": 0.0603, "step": 22486 }, { "epoch": 2.26, "learning_rate": 2.3387382931605985e-05, "loss": 0.4026, "step": 22487 }, { "epoch": 2.26, "learning_rate": 2.338504864386168e-05, "loss": 0.5828, "step": 22488 }, { "epoch": 2.26, "learning_rate": 2.3382714398212405e-05, "loss": 0.4955, "step": 22489 }, { "epoch": 2.26, "learning_rate": 2.338038019467302e-05, "loss": 0.1841, "step": 22490 }, { "epoch": 2.26, "learning_rate": 2.3378046033258377e-05, "loss": 0.0845, "step": 22491 }, { "epoch": 2.26, "learning_rate": 2.3375711913983323e-05, "loss": 0.288, "step": 22492 }, { "epoch": 2.26, "learning_rate": 2.337337783686273e-05, "loss": 0.5683, "step": 22493 }, { "epoch": 2.26, "learning_rate": 2.3371043801911436e-05, "loss": 0.9337, "step": 22494 }, { "epoch": 2.26, "learning_rate": 2.3368709809144293e-05, "loss": 0.6622, "step": 22495 }, { "epoch": 2.26, "learning_rate": 2.3366375858576162e-05, "loss": 0.7308, "step": 22496 }, { "epoch": 2.26, "learning_rate": 2.3364041950221897e-05, "loss": 0.5106, "step": 22497 }, { "epoch": 2.26, "learning_rate": 2.336170808409633e-05, "loss": 0.5408, "step": 22498 }, { "epoch": 2.26, "learning_rate": 2.3359374260214336e-05, "loss": 0.6491, "step": 22499 }, { "epoch": 2.26, "learning_rate": 2.335704047859076e-05, "loss": 0.396, "step": 22500 }, { "epoch": 2.26, "learning_rate": 2.3354706739240444e-05, "loss": 0.4423, "step": 22501 }, { "epoch": 2.26, "learning_rate": 2.3352373042178256e-05, "loss": 0.1526, "step": 22502 }, { "epoch": 2.26, "learning_rate": 2.3350039387419035e-05, "loss": 0.4667, "step": 22503 }, { "epoch": 2.26, "learning_rate": 2.334770577497762e-05, "loss": 0.4179, "step": 22504 }, { "epoch": 2.26, "learning_rate": 2.3345372204868886e-05, "loss": 0.2379, "step": 22505 }, { "epoch": 2.26, "learning_rate": 2.3343038677107666e-05, "loss": 0.1243, "step": 22506 }, { "epoch": 2.26, "learning_rate": 2.334070519170883e-05, "loss": 0.4629, "step": 22507 }, { "epoch": 2.26, "learning_rate": 2.33383717486872e-05, "loss": 0.2596, "step": 22508 }, { "epoch": 2.26, "learning_rate": 2.3336038348057638e-05, "loss": 0.6691, "step": 22509 }, { "epoch": 2.26, "learning_rate": 2.3333704989835e-05, "loss": 0.433, "step": 22510 }, { "epoch": 2.26, "learning_rate": 2.3331371674034127e-05, "loss": 0.4174, "step": 22511 }, { "epoch": 2.26, "learning_rate": 2.332903840066986e-05, "loss": 0.3141, "step": 22512 }, { "epoch": 2.26, "learning_rate": 2.332670516975706e-05, "loss": 0.187, "step": 22513 }, { "epoch": 2.26, "learning_rate": 2.332437198131057e-05, "loss": 0.1534, "step": 22514 }, { "epoch": 2.26, "learning_rate": 2.332203883534523e-05, "loss": 0.4504, "step": 22515 }, { "epoch": 2.26, "learning_rate": 2.3319705731875908e-05, "loss": 0.262, "step": 22516 }, { "epoch": 2.26, "learning_rate": 2.3317372670917428e-05, "loss": 0.2323, "step": 22517 }, { "epoch": 2.26, "learning_rate": 2.3315039652484647e-05, "loss": 0.3569, "step": 22518 }, { "epoch": 2.26, "learning_rate": 2.3312706676592413e-05, "loss": 0.443, "step": 22519 }, { "epoch": 2.26, "learning_rate": 2.3310373743255574e-05, "loss": 0.4584, "step": 22520 }, { "epoch": 2.26, "learning_rate": 2.330804085248896e-05, "loss": 0.3705, "step": 22521 }, { "epoch": 2.26, "learning_rate": 2.3305708004307435e-05, "loss": 0.1445, "step": 22522 }, { "epoch": 2.26, "learning_rate": 2.3303375198725838e-05, "loss": 0.506, "step": 22523 }, { "epoch": 2.26, "learning_rate": 2.3301042435759004e-05, "loss": 0.3404, "step": 22524 }, { "epoch": 2.26, "learning_rate": 2.3298709715421804e-05, "loss": 0.1366, "step": 22525 }, { "epoch": 2.26, "learning_rate": 2.329637703772906e-05, "loss": 0.7134, "step": 22526 }, { "epoch": 2.26, "learning_rate": 2.3294044402695612e-05, "loss": 0.4697, "step": 22527 }, { "epoch": 2.26, "learning_rate": 2.329171181033633e-05, "loss": 0.4639, "step": 22528 }, { "epoch": 2.26, "learning_rate": 2.3289379260666038e-05, "loss": 0.3291, "step": 22529 }, { "epoch": 2.26, "learning_rate": 2.3287046753699574e-05, "loss": 0.7807, "step": 22530 }, { "epoch": 2.26, "learning_rate": 2.32847142894518e-05, "loss": 0.0792, "step": 22531 }, { "epoch": 2.26, "learning_rate": 2.328238186793754e-05, "loss": 0.343, "step": 22532 }, { "epoch": 2.26, "learning_rate": 2.3280049489171656e-05, "loss": 0.9086, "step": 22533 }, { "epoch": 2.26, "learning_rate": 2.3277717153168984e-05, "loss": 0.0847, "step": 22534 }, { "epoch": 2.26, "learning_rate": 2.3275384859944352e-05, "loss": 0.477, "step": 22535 }, { "epoch": 2.26, "learning_rate": 2.3273052609512615e-05, "loss": 0.5293, "step": 22536 }, { "epoch": 2.26, "learning_rate": 2.3270720401888614e-05, "loss": 0.1573, "step": 22537 }, { "epoch": 2.26, "learning_rate": 2.3268388237087185e-05, "loss": 0.1796, "step": 22538 }, { "epoch": 2.26, "learning_rate": 2.3266056115123186e-05, "loss": 0.7035, "step": 22539 }, { "epoch": 2.26, "learning_rate": 2.326372403601143e-05, "loss": 0.2087, "step": 22540 }, { "epoch": 2.26, "learning_rate": 2.3261391999766773e-05, "loss": 0.6772, "step": 22541 }, { "epoch": 2.26, "learning_rate": 2.3259060006404057e-05, "loss": 0.8202, "step": 22542 }, { "epoch": 2.26, "learning_rate": 2.3256728055938127e-05, "loss": 0.2499, "step": 22543 }, { "epoch": 2.26, "learning_rate": 2.3254396148383802e-05, "loss": 0.3957, "step": 22544 }, { "epoch": 2.26, "learning_rate": 2.3252064283755934e-05, "loss": 0.1543, "step": 22545 }, { "epoch": 2.26, "learning_rate": 2.3249732462069368e-05, "loss": 0.0273, "step": 22546 }, { "epoch": 2.26, "learning_rate": 2.3247400683338928e-05, "loss": 0.2299, "step": 22547 }, { "epoch": 2.26, "learning_rate": 2.3245068947579473e-05, "loss": 0.5101, "step": 22548 }, { "epoch": 2.26, "learning_rate": 2.3242737254805828e-05, "loss": 0.1502, "step": 22549 }, { "epoch": 2.26, "learning_rate": 2.324040560503282e-05, "loss": 0.3596, "step": 22550 }, { "epoch": 2.26, "learning_rate": 2.323807399827531e-05, "loss": 0.4521, "step": 22551 }, { "epoch": 2.26, "learning_rate": 2.323574243454813e-05, "loss": 0.276, "step": 22552 }, { "epoch": 2.26, "learning_rate": 2.3233410913866097e-05, "loss": 0.2857, "step": 22553 }, { "epoch": 2.26, "learning_rate": 2.323107943624407e-05, "loss": 0.2154, "step": 22554 }, { "epoch": 2.26, "learning_rate": 2.322874800169688e-05, "loss": 0.1806, "step": 22555 }, { "epoch": 2.26, "learning_rate": 2.3226416610239353e-05, "loss": 0.046, "step": 22556 }, { "epoch": 2.26, "learning_rate": 2.3224085261886347e-05, "loss": 0.1577, "step": 22557 }, { "epoch": 2.27, "learning_rate": 2.322175395665268e-05, "loss": 0.6505, "step": 22558 }, { "epoch": 2.27, "learning_rate": 2.3219422694553187e-05, "loss": 1.5781, "step": 22559 }, { "epoch": 2.27, "learning_rate": 2.3217091475602715e-05, "loss": 0.1732, "step": 22560 }, { "epoch": 2.27, "learning_rate": 2.3214760299816084e-05, "loss": 0.5345, "step": 22561 }, { "epoch": 2.27, "learning_rate": 2.321242916720815e-05, "loss": 0.2958, "step": 22562 }, { "epoch": 2.27, "learning_rate": 2.321009807779373e-05, "loss": 0.1919, "step": 22563 }, { "epoch": 2.27, "learning_rate": 2.3207767031587656e-05, "loss": 0.1471, "step": 22564 }, { "epoch": 2.27, "learning_rate": 2.3205436028604772e-05, "loss": 0.518, "step": 22565 }, { "epoch": 2.27, "learning_rate": 2.3203105068859915e-05, "loss": 0.6449, "step": 22566 }, { "epoch": 2.27, "learning_rate": 2.32007741523679e-05, "loss": 0.2831, "step": 22567 }, { "epoch": 2.27, "learning_rate": 2.3198443279143583e-05, "loss": 0.4765, "step": 22568 }, { "epoch": 2.27, "learning_rate": 2.319611244920178e-05, "loss": 0.6656, "step": 22569 }, { "epoch": 2.27, "learning_rate": 2.3193781662557326e-05, "loss": 0.4296, "step": 22570 }, { "epoch": 2.27, "learning_rate": 2.3191450919225064e-05, "loss": 0.461, "step": 22571 }, { "epoch": 2.27, "learning_rate": 2.3189120219219817e-05, "loss": 0.3429, "step": 22572 }, { "epoch": 2.27, "learning_rate": 2.3186789562556406e-05, "loss": 0.3169, "step": 22573 }, { "epoch": 2.27, "learning_rate": 2.3184458949249685e-05, "loss": 0.4083, "step": 22574 }, { "epoch": 2.27, "learning_rate": 2.3182128379314482e-05, "loss": 0.1446, "step": 22575 }, { "epoch": 2.27, "learning_rate": 2.317979785276561e-05, "loss": 0.4661, "step": 22576 }, { "epoch": 2.27, "learning_rate": 2.3177467369617906e-05, "loss": 0.3762, "step": 22577 }, { "epoch": 2.27, "learning_rate": 2.3175136929886213e-05, "loss": 0.4799, "step": 22578 }, { "epoch": 2.27, "learning_rate": 2.317280653358534e-05, "loss": 0.072, "step": 22579 }, { "epoch": 2.27, "learning_rate": 2.3170476180730148e-05, "loss": 0.2009, "step": 22580 }, { "epoch": 2.27, "learning_rate": 2.3168145871335436e-05, "loss": 0.0944, "step": 22581 }, { "epoch": 2.27, "learning_rate": 2.3165815605416042e-05, "loss": 0.2331, "step": 22582 }, { "epoch": 2.27, "learning_rate": 2.3163485382986803e-05, "loss": 0.7136, "step": 22583 }, { "epoch": 2.27, "learning_rate": 2.3161155204062547e-05, "loss": 0.2373, "step": 22584 }, { "epoch": 2.27, "learning_rate": 2.3158825068658086e-05, "loss": 0.0309, "step": 22585 }, { "epoch": 2.27, "learning_rate": 2.3156494976788265e-05, "loss": 0.9556, "step": 22586 }, { "epoch": 2.27, "learning_rate": 2.3154164928467902e-05, "loss": 0.693, "step": 22587 }, { "epoch": 2.27, "learning_rate": 2.3151834923711835e-05, "loss": 0.3428, "step": 22588 }, { "epoch": 2.27, "learning_rate": 2.3149504962534892e-05, "loss": 0.397, "step": 22589 }, { "epoch": 2.27, "learning_rate": 2.314717504495188e-05, "loss": 0.2174, "step": 22590 }, { "epoch": 2.27, "learning_rate": 2.3144845170977648e-05, "loss": 0.7566, "step": 22591 }, { "epoch": 2.27, "learning_rate": 2.3142515340627007e-05, "loss": 0.6195, "step": 22592 }, { "epoch": 2.27, "learning_rate": 2.314018555391479e-05, "loss": 0.2029, "step": 22593 }, { "epoch": 2.27, "learning_rate": 2.3137855810855832e-05, "loss": 0.2418, "step": 22594 }, { "epoch": 2.27, "learning_rate": 2.3135526111464945e-05, "loss": 0.7598, "step": 22595 }, { "epoch": 2.27, "learning_rate": 2.3133196455756946e-05, "loss": 0.5918, "step": 22596 }, { "epoch": 2.27, "learning_rate": 2.3130866843746686e-05, "loss": 0.6366, "step": 22597 }, { "epoch": 2.27, "learning_rate": 2.312853727544898e-05, "loss": 0.4632, "step": 22598 }, { "epoch": 2.27, "learning_rate": 2.3126207750878636e-05, "loss": 0.3642, "step": 22599 }, { "epoch": 2.27, "learning_rate": 2.31238782700505e-05, "loss": 0.5853, "step": 22600 }, { "epoch": 2.27, "learning_rate": 2.3121548832979383e-05, "loss": 0.5445, "step": 22601 }, { "epoch": 2.27, "learning_rate": 2.3119219439680104e-05, "loss": 0.3738, "step": 22602 }, { "epoch": 2.27, "learning_rate": 2.311689009016751e-05, "loss": 0.5082, "step": 22603 }, { "epoch": 2.27, "learning_rate": 2.3114560784456404e-05, "loss": 0.2079, "step": 22604 }, { "epoch": 2.27, "learning_rate": 2.311223152256161e-05, "loss": 0.1722, "step": 22605 }, { "epoch": 2.27, "learning_rate": 2.310990230449796e-05, "loss": 0.4915, "step": 22606 }, { "epoch": 2.27, "learning_rate": 2.3107573130280272e-05, "loss": 0.4991, "step": 22607 }, { "epoch": 2.27, "learning_rate": 2.3105243999923358e-05, "loss": 0.3753, "step": 22608 }, { "epoch": 2.27, "learning_rate": 2.3102914913442052e-05, "loss": 0.8965, "step": 22609 }, { "epoch": 2.27, "learning_rate": 2.310058587085117e-05, "loss": 0.7853, "step": 22610 }, { "epoch": 2.27, "learning_rate": 2.3098256872165535e-05, "loss": 0.1418, "step": 22611 }, { "epoch": 2.27, "learning_rate": 2.3095927917399976e-05, "loss": 0.5345, "step": 22612 }, { "epoch": 2.27, "learning_rate": 2.3093599006569292e-05, "loss": 0.4271, "step": 22613 }, { "epoch": 2.27, "learning_rate": 2.3091270139688325e-05, "loss": 0.2866, "step": 22614 }, { "epoch": 2.27, "learning_rate": 2.3088941316771885e-05, "loss": 0.2755, "step": 22615 }, { "epoch": 2.27, "learning_rate": 2.308661253783479e-05, "loss": 0.3005, "step": 22616 }, { "epoch": 2.27, "learning_rate": 2.3084283802891874e-05, "loss": 0.2781, "step": 22617 }, { "epoch": 2.27, "learning_rate": 2.3081955111957936e-05, "loss": 0.6766, "step": 22618 }, { "epoch": 2.27, "learning_rate": 2.30796264650478e-05, "loss": 0.1838, "step": 22619 }, { "epoch": 2.27, "learning_rate": 2.3077297862176296e-05, "loss": 0.2566, "step": 22620 }, { "epoch": 2.27, "learning_rate": 2.307496930335824e-05, "loss": 0.5317, "step": 22621 }, { "epoch": 2.27, "learning_rate": 2.3072640788608432e-05, "loss": 0.3373, "step": 22622 }, { "epoch": 2.27, "learning_rate": 2.3070312317941706e-05, "loss": 0.0959, "step": 22623 }, { "epoch": 2.27, "learning_rate": 2.306798389137288e-05, "loss": 0.0758, "step": 22624 }, { "epoch": 2.27, "learning_rate": 2.3065655508916763e-05, "loss": 0.16, "step": 22625 }, { "epoch": 2.27, "learning_rate": 2.3063327170588187e-05, "loss": 0.1432, "step": 22626 }, { "epoch": 2.27, "learning_rate": 2.306099887640195e-05, "loss": 0.9623, "step": 22627 }, { "epoch": 2.27, "learning_rate": 2.305867062637287e-05, "loss": 0.1593, "step": 22628 }, { "epoch": 2.27, "learning_rate": 2.3056342420515778e-05, "loss": 0.6544, "step": 22629 }, { "epoch": 2.27, "learning_rate": 2.3054014258845488e-05, "loss": 0.2676, "step": 22630 }, { "epoch": 2.27, "learning_rate": 2.3051686141376795e-05, "loss": 0.1334, "step": 22631 }, { "epoch": 2.27, "learning_rate": 2.304935806812453e-05, "loss": 0.1846, "step": 22632 }, { "epoch": 2.27, "learning_rate": 2.304703003910351e-05, "loss": 0.5484, "step": 22633 }, { "epoch": 2.27, "learning_rate": 2.3044702054328543e-05, "loss": 0.3658, "step": 22634 }, { "epoch": 2.27, "learning_rate": 2.3042374113814455e-05, "loss": 0.121, "step": 22635 }, { "epoch": 2.27, "learning_rate": 2.3040046217576048e-05, "loss": 0.486, "step": 22636 }, { "epoch": 2.27, "learning_rate": 2.3037718365628132e-05, "loss": 0.2036, "step": 22637 }, { "epoch": 2.27, "learning_rate": 2.303539055798553e-05, "loss": 0.2767, "step": 22638 }, { "epoch": 2.27, "learning_rate": 2.3033062794663066e-05, "loss": 0.3232, "step": 22639 }, { "epoch": 2.27, "learning_rate": 2.303073507567552e-05, "loss": 0.4328, "step": 22640 }, { "epoch": 2.27, "learning_rate": 2.3028407401037734e-05, "loss": 0.4116, "step": 22641 }, { "epoch": 2.27, "learning_rate": 2.3026079770764507e-05, "loss": 0.1039, "step": 22642 }, { "epoch": 2.27, "learning_rate": 2.302375218487066e-05, "loss": 0.6474, "step": 22643 }, { "epoch": 2.27, "learning_rate": 2.3021424643371006e-05, "loss": 0.5135, "step": 22644 }, { "epoch": 2.27, "learning_rate": 2.3019097146280337e-05, "loss": 0.7373, "step": 22645 }, { "epoch": 2.27, "learning_rate": 2.301676969361349e-05, "loss": 0.1681, "step": 22646 }, { "epoch": 2.27, "learning_rate": 2.3014442285385257e-05, "loss": 0.214, "step": 22647 }, { "epoch": 2.27, "learning_rate": 2.3012114921610452e-05, "loss": 0.2452, "step": 22648 }, { "epoch": 2.27, "learning_rate": 2.3009787602303904e-05, "loss": 0.7319, "step": 22649 }, { "epoch": 2.27, "learning_rate": 2.3007460327480403e-05, "loss": 0.5541, "step": 22650 }, { "epoch": 2.27, "learning_rate": 2.3005133097154755e-05, "loss": 0.9906, "step": 22651 }, { "epoch": 2.27, "learning_rate": 2.3002805911341786e-05, "loss": 1.1799, "step": 22652 }, { "epoch": 2.27, "learning_rate": 2.3000478770056305e-05, "loss": 0.3451, "step": 22653 }, { "epoch": 2.27, "learning_rate": 2.2998151673313102e-05, "loss": 0.6639, "step": 22654 }, { "epoch": 2.27, "learning_rate": 2.2995824621127e-05, "loss": 0.2852, "step": 22655 }, { "epoch": 2.27, "learning_rate": 2.2993497613512812e-05, "loss": 0.3243, "step": 22656 }, { "epoch": 2.28, "learning_rate": 2.299117065048533e-05, "loss": 0.6536, "step": 22657 }, { "epoch": 2.28, "learning_rate": 2.2988843732059385e-05, "loss": 0.4726, "step": 22658 }, { "epoch": 2.28, "learning_rate": 2.2986516858249765e-05, "loss": 0.5552, "step": 22659 }, { "epoch": 2.28, "learning_rate": 2.298419002907128e-05, "loss": 0.1538, "step": 22660 }, { "epoch": 2.28, "learning_rate": 2.2981863244538747e-05, "loss": 0.2436, "step": 22661 }, { "epoch": 2.28, "learning_rate": 2.2979536504666968e-05, "loss": 0.7801, "step": 22662 }, { "epoch": 2.28, "learning_rate": 2.2977209809470738e-05, "loss": 0.1581, "step": 22663 }, { "epoch": 2.28, "learning_rate": 2.2974883158964878e-05, "loss": 0.4655, "step": 22664 }, { "epoch": 2.28, "learning_rate": 2.297255655316419e-05, "loss": 0.424, "step": 22665 }, { "epoch": 2.28, "learning_rate": 2.2970229992083472e-05, "loss": 0.7755, "step": 22666 }, { "epoch": 2.28, "learning_rate": 2.296790347573755e-05, "loss": 0.2819, "step": 22667 }, { "epoch": 2.28, "learning_rate": 2.29655770041412e-05, "loss": 0.3515, "step": 22668 }, { "epoch": 2.28, "learning_rate": 2.2963250577309255e-05, "loss": 0.3219, "step": 22669 }, { "epoch": 2.28, "learning_rate": 2.29609241952565e-05, "loss": 0.1571, "step": 22670 }, { "epoch": 2.28, "learning_rate": 2.295859785799774e-05, "loss": 0.6356, "step": 22671 }, { "epoch": 2.28, "learning_rate": 2.2956271565547798e-05, "loss": 0.1543, "step": 22672 }, { "epoch": 2.28, "learning_rate": 2.2953945317921457e-05, "loss": 0.3447, "step": 22673 }, { "epoch": 2.28, "learning_rate": 2.2951619115133523e-05, "loss": 0.5856, "step": 22674 }, { "epoch": 2.28, "learning_rate": 2.294929295719881e-05, "loss": 0.5347, "step": 22675 }, { "epoch": 2.28, "learning_rate": 2.294696684413212e-05, "loss": 0.1151, "step": 22676 }, { "epoch": 2.28, "learning_rate": 2.2944640775948238e-05, "loss": 0.2441, "step": 22677 }, { "epoch": 2.28, "learning_rate": 2.294231475266198e-05, "loss": 0.3911, "step": 22678 }, { "epoch": 2.28, "learning_rate": 2.2939988774288153e-05, "loss": 0.6105, "step": 22679 }, { "epoch": 2.28, "learning_rate": 2.293766284084154e-05, "loss": 0.5744, "step": 22680 }, { "epoch": 2.28, "learning_rate": 2.2935336952336973e-05, "loss": 0.228, "step": 22681 }, { "epoch": 2.28, "learning_rate": 2.293301110878922e-05, "loss": 0.2154, "step": 22682 }, { "epoch": 2.28, "learning_rate": 2.2930685310213094e-05, "loss": 0.6477, "step": 22683 }, { "epoch": 2.28, "learning_rate": 2.2928359556623402e-05, "loss": 0.287, "step": 22684 }, { "epoch": 2.28, "learning_rate": 2.2926033848034947e-05, "loss": 0.4536, "step": 22685 }, { "epoch": 2.28, "learning_rate": 2.292370818446251e-05, "loss": 0.4644, "step": 22686 }, { "epoch": 2.28, "learning_rate": 2.2921382565920908e-05, "loss": 0.2175, "step": 22687 }, { "epoch": 2.28, "learning_rate": 2.291905699242493e-05, "loss": 0.4486, "step": 22688 }, { "epoch": 2.28, "learning_rate": 2.2916731463989377e-05, "loss": 0.7645, "step": 22689 }, { "epoch": 2.28, "learning_rate": 2.2914405980629063e-05, "loss": 0.2261, "step": 22690 }, { "epoch": 2.28, "learning_rate": 2.2912080542358766e-05, "loss": 0.1121, "step": 22691 }, { "epoch": 2.28, "learning_rate": 2.290975514919329e-05, "loss": 0.3221, "step": 22692 }, { "epoch": 2.28, "learning_rate": 2.2907429801147434e-05, "loss": 0.6499, "step": 22693 }, { "epoch": 2.28, "learning_rate": 2.2905104498235997e-05, "loss": 0.5327, "step": 22694 }, { "epoch": 2.28, "learning_rate": 2.2902779240473788e-05, "loss": 0.8494, "step": 22695 }, { "epoch": 2.28, "learning_rate": 2.2900454027875583e-05, "loss": 0.5368, "step": 22696 }, { "epoch": 2.28, "learning_rate": 2.2898128860456183e-05, "loss": 0.3339, "step": 22697 }, { "epoch": 2.28, "learning_rate": 2.2895803738230393e-05, "loss": 0.2593, "step": 22698 }, { "epoch": 2.28, "learning_rate": 2.2893478661213014e-05, "loss": 0.906, "step": 22699 }, { "epoch": 2.28, "learning_rate": 2.2891153629418816e-05, "loss": 0.2713, "step": 22700 }, { "epoch": 2.28, "learning_rate": 2.2888828642862625e-05, "loss": 0.1342, "step": 22701 }, { "epoch": 2.28, "learning_rate": 2.288650370155922e-05, "loss": 0.8307, "step": 22702 }, { "epoch": 2.28, "learning_rate": 2.2884178805523396e-05, "loss": 0.6019, "step": 22703 }, { "epoch": 2.28, "learning_rate": 2.288185395476996e-05, "loss": 0.2486, "step": 22704 }, { "epoch": 2.28, "learning_rate": 2.287952914931369e-05, "loss": 0.4978, "step": 22705 }, { "epoch": 2.28, "learning_rate": 2.287720438916939e-05, "loss": 0.5318, "step": 22706 }, { "epoch": 2.28, "learning_rate": 2.2874879674351846e-05, "loss": 0.0575, "step": 22707 }, { "epoch": 2.28, "learning_rate": 2.2872555004875873e-05, "loss": 0.1028, "step": 22708 }, { "epoch": 2.28, "learning_rate": 2.2870230380756233e-05, "loss": 0.5245, "step": 22709 }, { "epoch": 2.28, "learning_rate": 2.286790580200774e-05, "loss": 0.5574, "step": 22710 }, { "epoch": 2.28, "learning_rate": 2.286558126864518e-05, "loss": 0.2583, "step": 22711 }, { "epoch": 2.28, "learning_rate": 2.2863256780683343e-05, "loss": 0.177, "step": 22712 }, { "epoch": 2.28, "learning_rate": 2.2860932338137037e-05, "loss": 0.3608, "step": 22713 }, { "epoch": 2.28, "learning_rate": 2.2858607941021035e-05, "loss": 0.4728, "step": 22714 }, { "epoch": 2.28, "learning_rate": 2.2856283589350133e-05, "loss": 0.2388, "step": 22715 }, { "epoch": 2.28, "learning_rate": 2.2853959283139122e-05, "loss": 0.5041, "step": 22716 }, { "epoch": 2.28, "learning_rate": 2.285163502240281e-05, "loss": 0.1758, "step": 22717 }, { "epoch": 2.28, "learning_rate": 2.2849310807155953e-05, "loss": 0.3903, "step": 22718 }, { "epoch": 2.28, "learning_rate": 2.2846986637413375e-05, "loss": 0.2797, "step": 22719 }, { "epoch": 2.28, "learning_rate": 2.284466251318985e-05, "loss": 0.4021, "step": 22720 }, { "epoch": 2.28, "learning_rate": 2.2842338434500167e-05, "loss": 0.4127, "step": 22721 }, { "epoch": 2.28, "learning_rate": 2.284001440135913e-05, "loss": 0.4137, "step": 22722 }, { "epoch": 2.28, "learning_rate": 2.2837690413781503e-05, "loss": 0.8608, "step": 22723 }, { "epoch": 2.28, "learning_rate": 2.2835366471782097e-05, "loss": 0.3078, "step": 22724 }, { "epoch": 2.28, "learning_rate": 2.2833042575375696e-05, "loss": 0.1002, "step": 22725 }, { "epoch": 2.28, "learning_rate": 2.2830718724577076e-05, "loss": 0.5099, "step": 22726 }, { "epoch": 2.28, "learning_rate": 2.282839491940105e-05, "loss": 0.472, "step": 22727 }, { "epoch": 2.28, "learning_rate": 2.282607115986238e-05, "loss": 0.0816, "step": 22728 }, { "epoch": 2.28, "learning_rate": 2.2823747445975866e-05, "loss": 0.8321, "step": 22729 }, { "epoch": 2.28, "learning_rate": 2.2821423777756292e-05, "loss": 0.3286, "step": 22730 }, { "epoch": 2.28, "learning_rate": 2.2819100155218455e-05, "loss": 0.8678, "step": 22731 }, { "epoch": 2.28, "learning_rate": 2.2816776578377118e-05, "loss": 0.3121, "step": 22732 }, { "epoch": 2.28, "learning_rate": 2.281445304724709e-05, "loss": 0.3495, "step": 22733 }, { "epoch": 2.28, "learning_rate": 2.281212956184315e-05, "loss": 0.5005, "step": 22734 }, { "epoch": 2.28, "learning_rate": 2.2809806122180074e-05, "loss": 0.5062, "step": 22735 }, { "epoch": 2.28, "learning_rate": 2.2807482728272676e-05, "loss": 0.1128, "step": 22736 }, { "epoch": 2.28, "learning_rate": 2.280515938013571e-05, "loss": 0.5298, "step": 22737 }, { "epoch": 2.28, "learning_rate": 2.2802836077783967e-05, "loss": 0.277, "step": 22738 }, { "epoch": 2.28, "learning_rate": 2.2800512821232245e-05, "loss": 0.6382, "step": 22739 }, { "epoch": 2.28, "learning_rate": 2.2798189610495325e-05, "loss": 0.5607, "step": 22740 }, { "epoch": 2.28, "learning_rate": 2.279586644558797e-05, "loss": 0.1311, "step": 22741 }, { "epoch": 2.28, "learning_rate": 2.2793543326524996e-05, "loss": 0.3443, "step": 22742 }, { "epoch": 2.28, "learning_rate": 2.2791220253321165e-05, "loss": 0.5502, "step": 22743 }, { "epoch": 2.28, "learning_rate": 2.2788897225991263e-05, "loss": 0.178, "step": 22744 }, { "epoch": 2.28, "learning_rate": 2.2786574244550083e-05, "loss": 0.1184, "step": 22745 }, { "epoch": 2.28, "learning_rate": 2.27842513090124e-05, "loss": 0.4051, "step": 22746 }, { "epoch": 2.28, "learning_rate": 2.2781928419392988e-05, "loss": 0.1206, "step": 22747 }, { "epoch": 2.28, "learning_rate": 2.2779605575706642e-05, "loss": 0.1403, "step": 22748 }, { "epoch": 2.28, "learning_rate": 2.2777282777968134e-05, "loss": 0.1516, "step": 22749 }, { "epoch": 2.28, "learning_rate": 2.2774960026192266e-05, "loss": 0.415, "step": 22750 }, { "epoch": 2.28, "learning_rate": 2.2772637320393798e-05, "loss": 0.0897, "step": 22751 }, { "epoch": 2.28, "learning_rate": 2.2770314660587503e-05, "loss": 0.4098, "step": 22752 }, { "epoch": 2.28, "learning_rate": 2.276799204678819e-05, "loss": 1.2882, "step": 22753 }, { "epoch": 2.28, "learning_rate": 2.2765669479010626e-05, "loss": 0.0802, "step": 22754 }, { "epoch": 2.29, "learning_rate": 2.2763346957269578e-05, "loss": 0.1568, "step": 22755 }, { "epoch": 2.29, "learning_rate": 2.276102448157984e-05, "loss": 0.2176, "step": 22756 }, { "epoch": 2.29, "learning_rate": 2.2758702051956196e-05, "loss": 0.4557, "step": 22757 }, { "epoch": 2.29, "learning_rate": 2.2756379668413404e-05, "loss": 0.1696, "step": 22758 }, { "epoch": 2.29, "learning_rate": 2.2754057330966274e-05, "loss": 0.0415, "step": 22759 }, { "epoch": 2.29, "learning_rate": 2.2751735039629555e-05, "loss": 0.2706, "step": 22760 }, { "epoch": 2.29, "learning_rate": 2.2749412794418034e-05, "loss": 0.5374, "step": 22761 }, { "epoch": 2.29, "learning_rate": 2.27470905953465e-05, "loss": 0.1251, "step": 22762 }, { "epoch": 2.29, "learning_rate": 2.2744768442429722e-05, "loss": 0.4422, "step": 22763 }, { "epoch": 2.29, "learning_rate": 2.274244633568247e-05, "loss": 0.9774, "step": 22764 }, { "epoch": 2.29, "learning_rate": 2.2740124275119535e-05, "loss": 0.2684, "step": 22765 }, { "epoch": 2.29, "learning_rate": 2.2737802260755684e-05, "loss": 0.5054, "step": 22766 }, { "epoch": 2.29, "learning_rate": 2.2735480292605693e-05, "loss": 0.2635, "step": 22767 }, { "epoch": 2.29, "learning_rate": 2.2733158370684355e-05, "loss": 0.2905, "step": 22768 }, { "epoch": 2.29, "learning_rate": 2.2730836495006425e-05, "loss": 0.3375, "step": 22769 }, { "epoch": 2.29, "learning_rate": 2.272851466558668e-05, "loss": 0.2396, "step": 22770 }, { "epoch": 2.29, "learning_rate": 2.2726192882439908e-05, "loss": 0.1879, "step": 22771 }, { "epoch": 2.29, "learning_rate": 2.2723871145580886e-05, "loss": 0.4406, "step": 22772 }, { "epoch": 2.29, "learning_rate": 2.272154945502436e-05, "loss": 0.7281, "step": 22773 }, { "epoch": 2.29, "learning_rate": 2.271922781078514e-05, "loss": 0.3055, "step": 22774 }, { "epoch": 2.29, "learning_rate": 2.2716906212877983e-05, "loss": 0.1172, "step": 22775 }, { "epoch": 2.29, "learning_rate": 2.2714584661317654e-05, "loss": 0.3603, "step": 22776 }, { "epoch": 2.29, "learning_rate": 2.271226315611895e-05, "loss": 0.2131, "step": 22777 }, { "epoch": 2.29, "learning_rate": 2.2709941697296614e-05, "loss": 0.1198, "step": 22778 }, { "epoch": 2.29, "learning_rate": 2.2707620284865448e-05, "loss": 0.2769, "step": 22779 }, { "epoch": 2.29, "learning_rate": 2.2705298918840215e-05, "loss": 0.1084, "step": 22780 }, { "epoch": 2.29, "learning_rate": 2.2702977599235674e-05, "loss": 0.5488, "step": 22781 }, { "epoch": 2.29, "learning_rate": 2.270065632606662e-05, "loss": 0.2273, "step": 22782 }, { "epoch": 2.29, "learning_rate": 2.2698335099347808e-05, "loss": 0.2842, "step": 22783 }, { "epoch": 2.29, "learning_rate": 2.2696013919094007e-05, "loss": 0.2203, "step": 22784 }, { "epoch": 2.29, "learning_rate": 2.269369278532e-05, "loss": 0.0876, "step": 22785 }, { "epoch": 2.29, "learning_rate": 2.2691371698040558e-05, "loss": 0.6196, "step": 22786 }, { "epoch": 2.29, "learning_rate": 2.2689050657270434e-05, "loss": 0.6324, "step": 22787 }, { "epoch": 2.29, "learning_rate": 2.268672966302442e-05, "loss": 0.2156, "step": 22788 }, { "epoch": 2.29, "learning_rate": 2.2684408715317275e-05, "loss": 0.3269, "step": 22789 }, { "epoch": 2.29, "learning_rate": 2.268208781416376e-05, "loss": 0.1913, "step": 22790 }, { "epoch": 2.29, "learning_rate": 2.2679766959578672e-05, "loss": 0.1057, "step": 22791 }, { "epoch": 2.29, "learning_rate": 2.267744615157675e-05, "loss": 0.3268, "step": 22792 }, { "epoch": 2.29, "learning_rate": 2.2675125390172772e-05, "loss": 0.86, "step": 22793 }, { "epoch": 2.29, "learning_rate": 2.267280467538152e-05, "loss": 0.4143, "step": 22794 }, { "epoch": 2.29, "learning_rate": 2.2670484007217753e-05, "loss": 0.3484, "step": 22795 }, { "epoch": 2.29, "learning_rate": 2.2668163385696225e-05, "loss": 0.1954, "step": 22796 }, { "epoch": 2.29, "learning_rate": 2.2665842810831725e-05, "loss": 0.4898, "step": 22797 }, { "epoch": 2.29, "learning_rate": 2.2663522282639005e-05, "loss": 0.4102, "step": 22798 }, { "epoch": 2.29, "learning_rate": 2.266120180113284e-05, "loss": 0.2039, "step": 22799 }, { "epoch": 2.29, "learning_rate": 2.2658881366328003e-05, "loss": 0.2948, "step": 22800 }, { "epoch": 2.29, "learning_rate": 2.265656097823925e-05, "loss": 0.1618, "step": 22801 }, { "epoch": 2.29, "learning_rate": 2.265424063688134e-05, "loss": 0.1354, "step": 22802 }, { "epoch": 2.29, "learning_rate": 2.2651920342269053e-05, "loss": 0.6445, "step": 22803 }, { "epoch": 2.29, "learning_rate": 2.264960009441714e-05, "loss": 0.0665, "step": 22804 }, { "epoch": 2.29, "learning_rate": 2.2647279893340396e-05, "loss": 0.2663, "step": 22805 }, { "epoch": 2.29, "learning_rate": 2.264495973905356e-05, "loss": 0.2014, "step": 22806 }, { "epoch": 2.29, "learning_rate": 2.264263963157139e-05, "loss": 0.2727, "step": 22807 }, { "epoch": 2.29, "learning_rate": 2.264031957090867e-05, "loss": 0.3825, "step": 22808 }, { "epoch": 2.29, "learning_rate": 2.2637999557080166e-05, "loss": 0.2975, "step": 22809 }, { "epoch": 2.29, "learning_rate": 2.263567959010061e-05, "loss": 0.2628, "step": 22810 }, { "epoch": 2.29, "learning_rate": 2.2633359669984806e-05, "loss": 0.3375, "step": 22811 }, { "epoch": 2.29, "learning_rate": 2.2631039796747495e-05, "loss": 0.7192, "step": 22812 }, { "epoch": 2.29, "learning_rate": 2.2628719970403436e-05, "loss": 0.5043, "step": 22813 }, { "epoch": 2.29, "learning_rate": 2.262640019096741e-05, "loss": 0.3615, "step": 22814 }, { "epoch": 2.29, "learning_rate": 2.2624080458454165e-05, "loss": 0.0002, "step": 22815 }, { "epoch": 2.29, "learning_rate": 2.262176077287846e-05, "loss": 0.2405, "step": 22816 }, { "epoch": 2.29, "learning_rate": 2.2619441134255065e-05, "loss": 0.2765, "step": 22817 }, { "epoch": 2.29, "learning_rate": 2.261712154259875e-05, "loss": 0.1587, "step": 22818 }, { "epoch": 2.29, "learning_rate": 2.261480199792425e-05, "loss": 0.3138, "step": 22819 }, { "epoch": 2.29, "learning_rate": 2.2612482500246344e-05, "loss": 0.5632, "step": 22820 }, { "epoch": 2.29, "learning_rate": 2.2610163049579794e-05, "loss": 0.3457, "step": 22821 }, { "epoch": 2.29, "learning_rate": 2.2607843645939345e-05, "loss": 0.4036, "step": 22822 }, { "epoch": 2.29, "learning_rate": 2.260552428933978e-05, "loss": 0.5977, "step": 22823 }, { "epoch": 2.29, "learning_rate": 2.260320497979584e-05, "loss": 0.2543, "step": 22824 }, { "epoch": 2.29, "learning_rate": 2.2600885717322284e-05, "loss": 0.7089, "step": 22825 }, { "epoch": 2.29, "learning_rate": 2.2598566501933884e-05, "loss": 0.5596, "step": 22826 }, { "epoch": 2.29, "learning_rate": 2.2596247333645394e-05, "loss": 0.1207, "step": 22827 }, { "epoch": 2.29, "learning_rate": 2.2593928212471558e-05, "loss": 0.2522, "step": 22828 }, { "epoch": 2.29, "learning_rate": 2.2591609138427153e-05, "loss": 0.3196, "step": 22829 }, { "epoch": 2.29, "learning_rate": 2.2589290111526924e-05, "loss": 0.2715, "step": 22830 }, { "epoch": 2.29, "learning_rate": 2.2586971131785634e-05, "loss": 0.3301, "step": 22831 }, { "epoch": 2.29, "learning_rate": 2.2584652199218052e-05, "loss": 0.1087, "step": 22832 }, { "epoch": 2.29, "learning_rate": 2.25823333138389e-05, "loss": 0.5419, "step": 22833 }, { "epoch": 2.29, "learning_rate": 2.258001447566298e-05, "loss": 0.4366, "step": 22834 }, { "epoch": 2.29, "learning_rate": 2.2577695684705014e-05, "loss": 0.3253, "step": 22835 }, { "epoch": 2.29, "learning_rate": 2.2575376940979765e-05, "loss": 0.2207, "step": 22836 }, { "epoch": 2.29, "learning_rate": 2.2573058244502007e-05, "loss": 0.5374, "step": 22837 }, { "epoch": 2.29, "learning_rate": 2.2570739595286475e-05, "loss": 0.3453, "step": 22838 }, { "epoch": 2.29, "learning_rate": 2.2568420993347918e-05, "loss": 0.1951, "step": 22839 }, { "epoch": 2.29, "learning_rate": 2.2566102438701114e-05, "loss": 0.5277, "step": 22840 }, { "epoch": 2.29, "learning_rate": 2.256378393136081e-05, "loss": 0.2003, "step": 22841 }, { "epoch": 2.29, "learning_rate": 2.2561465471341738e-05, "loss": 0.1509, "step": 22842 }, { "epoch": 2.29, "learning_rate": 2.2559147058658693e-05, "loss": 0.4555, "step": 22843 }, { "epoch": 2.29, "learning_rate": 2.2556828693326392e-05, "loss": 0.0878, "step": 22844 }, { "epoch": 2.29, "learning_rate": 2.25545103753596e-05, "loss": 0.2519, "step": 22845 }, { "epoch": 2.29, "learning_rate": 2.2552192104773082e-05, "loss": 0.3122, "step": 22846 }, { "epoch": 2.29, "learning_rate": 2.2549873881581573e-05, "loss": 0.2441, "step": 22847 }, { "epoch": 2.29, "learning_rate": 2.2547555705799827e-05, "loss": 0.8622, "step": 22848 }, { "epoch": 2.29, "learning_rate": 2.254523757744261e-05, "loss": 0.2185, "step": 22849 }, { "epoch": 2.29, "learning_rate": 2.254291949652467e-05, "loss": 0.7301, "step": 22850 }, { "epoch": 2.29, "learning_rate": 2.254060146306074e-05, "loss": 0.6741, "step": 22851 }, { "epoch": 2.29, "learning_rate": 2.2538283477065596e-05, "loss": 0.2068, "step": 22852 }, { "epoch": 2.29, "learning_rate": 2.253596553855397e-05, "loss": 0.3159, "step": 22853 }, { "epoch": 2.3, "learning_rate": 2.2533647647540617e-05, "loss": 0.1449, "step": 22854 }, { "epoch": 2.3, "learning_rate": 2.25313298040403e-05, "loss": 0.4811, "step": 22855 }, { "epoch": 2.3, "learning_rate": 2.252901200806776e-05, "loss": 0.2486, "step": 22856 }, { "epoch": 2.3, "learning_rate": 2.252669425963773e-05, "loss": 1.1227, "step": 22857 }, { "epoch": 2.3, "learning_rate": 2.2524376558764987e-05, "loss": 0.27, "step": 22858 }, { "epoch": 2.3, "learning_rate": 2.252205890546426e-05, "loss": 0.5882, "step": 22859 }, { "epoch": 2.3, "learning_rate": 2.2519741299750316e-05, "loss": 0.5514, "step": 22860 }, { "epoch": 2.3, "learning_rate": 2.251742374163789e-05, "loss": 0.7919, "step": 22861 }, { "epoch": 2.3, "learning_rate": 2.2515106231141722e-05, "loss": 0.4229, "step": 22862 }, { "epoch": 2.3, "learning_rate": 2.251278876827658e-05, "loss": 0.9954, "step": 22863 }, { "epoch": 2.3, "learning_rate": 2.2510471353057208e-05, "loss": 0.3083, "step": 22864 }, { "epoch": 2.3, "learning_rate": 2.2508153985498325e-05, "loss": 1.3615, "step": 22865 }, { "epoch": 2.3, "learning_rate": 2.250583666561472e-05, "loss": 0.1188, "step": 22866 }, { "epoch": 2.3, "learning_rate": 2.250351939342112e-05, "loss": 0.6024, "step": 22867 }, { "epoch": 2.3, "learning_rate": 2.2501202168932257e-05, "loss": 0.9836, "step": 22868 }, { "epoch": 2.3, "learning_rate": 2.2498884992162907e-05, "loss": 0.4691, "step": 22869 }, { "epoch": 2.3, "learning_rate": 2.249656786312779e-05, "loss": 0.5792, "step": 22870 }, { "epoch": 2.3, "learning_rate": 2.249425078184166e-05, "loss": 0.605, "step": 22871 }, { "epoch": 2.3, "learning_rate": 2.2491933748319264e-05, "loss": 0.3118, "step": 22872 }, { "epoch": 2.3, "learning_rate": 2.2489616762575352e-05, "loss": 0.2192, "step": 22873 }, { "epoch": 2.3, "learning_rate": 2.2487299824624644e-05, "loss": 0.3491, "step": 22874 }, { "epoch": 2.3, "learning_rate": 2.248498293448192e-05, "loss": 0.5726, "step": 22875 }, { "epoch": 2.3, "learning_rate": 2.24826660921619e-05, "loss": 0.3905, "step": 22876 }, { "epoch": 2.3, "learning_rate": 2.2480349297679327e-05, "loss": 0.5092, "step": 22877 }, { "epoch": 2.3, "learning_rate": 2.2478032551048964e-05, "loss": 0.6868, "step": 22878 }, { "epoch": 2.3, "learning_rate": 2.247571585228553e-05, "loss": 0.2455, "step": 22879 }, { "epoch": 2.3, "learning_rate": 2.2473399201403775e-05, "loss": 0.7353, "step": 22880 }, { "epoch": 2.3, "learning_rate": 2.2471082598418453e-05, "loss": 0.5229, "step": 22881 }, { "epoch": 2.3, "learning_rate": 2.2468766043344303e-05, "loss": 0.109, "step": 22882 }, { "epoch": 2.3, "learning_rate": 2.2466449536196046e-05, "loss": 0.1294, "step": 22883 }, { "epoch": 2.3, "learning_rate": 2.246413307698844e-05, "loss": 0.7719, "step": 22884 }, { "epoch": 2.3, "learning_rate": 2.2461816665736224e-05, "loss": 0.3576, "step": 22885 }, { "epoch": 2.3, "learning_rate": 2.2459500302454145e-05, "loss": 0.2291, "step": 22886 }, { "epoch": 2.3, "learning_rate": 2.2457183987156943e-05, "loss": 0.4646, "step": 22887 }, { "epoch": 2.3, "learning_rate": 2.245486771985933e-05, "loss": 0.4914, "step": 22888 }, { "epoch": 2.3, "learning_rate": 2.2452551500576094e-05, "loss": 0.4321, "step": 22889 }, { "epoch": 2.3, "learning_rate": 2.2450235329321937e-05, "loss": 0.2158, "step": 22890 }, { "epoch": 2.3, "learning_rate": 2.2447919206111608e-05, "loss": 0.2227, "step": 22891 }, { "epoch": 2.3, "learning_rate": 2.244560313095986e-05, "loss": 0.1257, "step": 22892 }, { "epoch": 2.3, "learning_rate": 2.2443287103881412e-05, "loss": 0.1122, "step": 22893 }, { "epoch": 2.3, "learning_rate": 2.2440971124891005e-05, "loss": 0.4656, "step": 22894 }, { "epoch": 2.3, "learning_rate": 2.243865519400339e-05, "loss": 0.4932, "step": 22895 }, { "epoch": 2.3, "learning_rate": 2.24363393112333e-05, "loss": 0.5927, "step": 22896 }, { "epoch": 2.3, "learning_rate": 2.2434023476595453e-05, "loss": 0.0965, "step": 22897 }, { "epoch": 2.3, "learning_rate": 2.243170769010462e-05, "loss": 0.3115, "step": 22898 }, { "epoch": 2.3, "learning_rate": 2.2429391951775515e-05, "loss": 0.5391, "step": 22899 }, { "epoch": 2.3, "learning_rate": 2.2427076261622872e-05, "loss": 0.3378, "step": 22900 }, { "epoch": 2.3, "learning_rate": 2.2424760619661443e-05, "loss": 0.9471, "step": 22901 }, { "epoch": 2.3, "learning_rate": 2.2422445025905954e-05, "loss": 0.515, "step": 22902 }, { "epoch": 2.3, "learning_rate": 2.242012948037114e-05, "loss": 0.96, "step": 22903 }, { "epoch": 2.3, "learning_rate": 2.241781398307174e-05, "loss": 0.3035, "step": 22904 }, { "epoch": 2.3, "learning_rate": 2.2415498534022493e-05, "loss": 0.3921, "step": 22905 }, { "epoch": 2.3, "learning_rate": 2.241318313323811e-05, "loss": 0.8754, "step": 22906 }, { "epoch": 2.3, "learning_rate": 2.2410867780733367e-05, "loss": 0.2121, "step": 22907 }, { "epoch": 2.3, "learning_rate": 2.240855247652296e-05, "loss": 0.9001, "step": 22908 }, { "epoch": 2.3, "learning_rate": 2.2406237220621637e-05, "loss": 0.3594, "step": 22909 }, { "epoch": 2.3, "learning_rate": 2.2403922013044142e-05, "loss": 0.4514, "step": 22910 }, { "epoch": 2.3, "learning_rate": 2.2401606853805178e-05, "loss": 0.8313, "step": 22911 }, { "epoch": 2.3, "learning_rate": 2.2399291742919516e-05, "loss": 0.3006, "step": 22912 }, { "epoch": 2.3, "learning_rate": 2.2396976680401865e-05, "loss": 0.471, "step": 22913 }, { "epoch": 2.3, "learning_rate": 2.2394661666266956e-05, "loss": 0.7636, "step": 22914 }, { "epoch": 2.3, "learning_rate": 2.2392346700529537e-05, "loss": 0.2427, "step": 22915 }, { "epoch": 2.3, "learning_rate": 2.2390031783204324e-05, "loss": 0.4742, "step": 22916 }, { "epoch": 2.3, "learning_rate": 2.238771691430605e-05, "loss": 0.1679, "step": 22917 }, { "epoch": 2.3, "learning_rate": 2.2385402093849454e-05, "loss": 0.5853, "step": 22918 }, { "epoch": 2.3, "learning_rate": 2.2383087321849268e-05, "loss": 0.5343, "step": 22919 }, { "epoch": 2.3, "learning_rate": 2.2380772598320202e-05, "loss": 0.2574, "step": 22920 }, { "epoch": 2.3, "learning_rate": 2.237845792327701e-05, "loss": 0.255, "step": 22921 }, { "epoch": 2.3, "learning_rate": 2.2376143296734417e-05, "loss": 0.619, "step": 22922 }, { "epoch": 2.3, "learning_rate": 2.2373828718707136e-05, "loss": 0.4775, "step": 22923 }, { "epoch": 2.3, "learning_rate": 2.2371514189209917e-05, "loss": 0.1969, "step": 22924 }, { "epoch": 2.3, "learning_rate": 2.2369199708257478e-05, "loss": 0.6638, "step": 22925 }, { "epoch": 2.3, "learning_rate": 2.2366885275864543e-05, "loss": 0.4135, "step": 22926 }, { "epoch": 2.3, "learning_rate": 2.2364570892045854e-05, "loss": 0.3684, "step": 22927 }, { "epoch": 2.3, "learning_rate": 2.2362256556816136e-05, "loss": 1.099, "step": 22928 }, { "epoch": 2.3, "learning_rate": 2.2359942270190094e-05, "loss": 0.0326, "step": 22929 }, { "epoch": 2.3, "learning_rate": 2.235762803218249e-05, "loss": 0.4863, "step": 22930 }, { "epoch": 2.3, "learning_rate": 2.235531384280803e-05, "loss": 0.7367, "step": 22931 }, { "epoch": 2.3, "learning_rate": 2.2352999702081434e-05, "loss": 0.7359, "step": 22932 }, { "epoch": 2.3, "learning_rate": 2.2350685610017457e-05, "loss": 0.5501, "step": 22933 }, { "epoch": 2.3, "learning_rate": 2.2348371566630793e-05, "loss": 0.6377, "step": 22934 }, { "epoch": 2.3, "learning_rate": 2.2346057571936184e-05, "loss": 0.2814, "step": 22935 }, { "epoch": 2.3, "learning_rate": 2.234374362594835e-05, "loss": 0.4794, "step": 22936 }, { "epoch": 2.3, "learning_rate": 2.234142972868203e-05, "loss": 0.0769, "step": 22937 }, { "epoch": 2.3, "learning_rate": 2.233911588015192e-05, "loss": 0.4373, "step": 22938 }, { "epoch": 2.3, "learning_rate": 2.233680208037278e-05, "loss": 0.1812, "step": 22939 }, { "epoch": 2.3, "learning_rate": 2.23344883293593e-05, "loss": 0.4, "step": 22940 }, { "epoch": 2.3, "learning_rate": 2.233217462712623e-05, "loss": 0.2829, "step": 22941 }, { "epoch": 2.3, "learning_rate": 2.232986097368829e-05, "loss": 0.5202, "step": 22942 }, { "epoch": 2.3, "learning_rate": 2.232754736906018e-05, "loss": 0.6131, "step": 22943 }, { "epoch": 2.3, "learning_rate": 2.2325233813256656e-05, "loss": 0.2706, "step": 22944 }, { "epoch": 2.3, "learning_rate": 2.232292030629242e-05, "loss": 0.2997, "step": 22945 }, { "epoch": 2.3, "learning_rate": 2.2320606848182192e-05, "loss": 0.3969, "step": 22946 }, { "epoch": 2.3, "learning_rate": 2.2318293438940713e-05, "loss": 0.313, "step": 22947 }, { "epoch": 2.3, "learning_rate": 2.2315980078582685e-05, "loss": 0.2901, "step": 22948 }, { "epoch": 2.3, "learning_rate": 2.231366676712283e-05, "loss": 0.4508, "step": 22949 }, { "epoch": 2.3, "learning_rate": 2.231135350457588e-05, "loss": 0.4201, "step": 22950 }, { "epoch": 2.3, "learning_rate": 2.230904029095656e-05, "loss": 0.4432, "step": 22951 }, { "epoch": 2.31, "learning_rate": 2.2306727126279566e-05, "loss": 0.2039, "step": 22952 }, { "epoch": 2.31, "learning_rate": 2.230441401055965e-05, "loss": 0.0724, "step": 22953 }, { "epoch": 2.31, "learning_rate": 2.2302100943811508e-05, "loss": 0.4606, "step": 22954 }, { "epoch": 2.31, "learning_rate": 2.2299787926049863e-05, "loss": 0.4715, "step": 22955 }, { "epoch": 2.31, "learning_rate": 2.2297474957289447e-05, "loss": 0.3353, "step": 22956 }, { "epoch": 2.31, "learning_rate": 2.229516203754497e-05, "loss": 0.1806, "step": 22957 }, { "epoch": 2.31, "learning_rate": 2.229284916683114e-05, "loss": 0.251, "step": 22958 }, { "epoch": 2.31, "learning_rate": 2.229053634516269e-05, "loss": 0.175, "step": 22959 }, { "epoch": 2.31, "learning_rate": 2.2288223572554347e-05, "loss": 0.1859, "step": 22960 }, { "epoch": 2.31, "learning_rate": 2.2285910849020795e-05, "loss": 0.2392, "step": 22961 }, { "epoch": 2.31, "learning_rate": 2.228359817457679e-05, "loss": 0.1372, "step": 22962 }, { "epoch": 2.31, "learning_rate": 2.2281285549237022e-05, "loss": 0.586, "step": 22963 }, { "epoch": 2.31, "learning_rate": 2.2278972973016214e-05, "loss": 0.0692, "step": 22964 }, { "epoch": 2.31, "learning_rate": 2.2276660445929094e-05, "loss": 0.8541, "step": 22965 }, { "epoch": 2.31, "learning_rate": 2.227434796799035e-05, "loss": 0.5527, "step": 22966 }, { "epoch": 2.31, "learning_rate": 2.2272035539214735e-05, "loss": 0.9384, "step": 22967 }, { "epoch": 2.31, "learning_rate": 2.2269723159616943e-05, "loss": 0.1985, "step": 22968 }, { "epoch": 2.31, "learning_rate": 2.2267410829211683e-05, "loss": 0.6152, "step": 22969 }, { "epoch": 2.31, "learning_rate": 2.226509854801369e-05, "loss": 0.3387, "step": 22970 }, { "epoch": 2.31, "learning_rate": 2.2262786316037667e-05, "loss": 0.2933, "step": 22971 }, { "epoch": 2.31, "learning_rate": 2.2260474133298317e-05, "loss": 0.9987, "step": 22972 }, { "epoch": 2.31, "learning_rate": 2.2258161999810372e-05, "loss": 0.3452, "step": 22973 }, { "epoch": 2.31, "learning_rate": 2.225584991558855e-05, "loss": 0.0601, "step": 22974 }, { "epoch": 2.31, "learning_rate": 2.225353788064753e-05, "loss": 0.5291, "step": 22975 }, { "epoch": 2.31, "learning_rate": 2.2251225895002066e-05, "loss": 0.9112, "step": 22976 }, { "epoch": 2.31, "learning_rate": 2.224891395866685e-05, "loss": 0.9359, "step": 22977 }, { "epoch": 2.31, "learning_rate": 2.2246602071656585e-05, "loss": 0.0715, "step": 22978 }, { "epoch": 2.31, "learning_rate": 2.2244290233986014e-05, "loss": 0.2035, "step": 22979 }, { "epoch": 2.31, "learning_rate": 2.2241978445669822e-05, "loss": 0.2559, "step": 22980 }, { "epoch": 2.31, "learning_rate": 2.223966670672272e-05, "loss": 0.4497, "step": 22981 }, { "epoch": 2.31, "learning_rate": 2.223735501715943e-05, "loss": 0.7534, "step": 22982 }, { "epoch": 2.31, "learning_rate": 2.223504337699467e-05, "loss": 0.2197, "step": 22983 }, { "epoch": 2.31, "learning_rate": 2.223273178624312e-05, "loss": 0.6378, "step": 22984 }, { "epoch": 2.31, "learning_rate": 2.2230420244919527e-05, "loss": 0.4459, "step": 22985 }, { "epoch": 2.31, "learning_rate": 2.222810875303858e-05, "loss": 0.3067, "step": 22986 }, { "epoch": 2.31, "learning_rate": 2.2225797310614986e-05, "loss": 0.2489, "step": 22987 }, { "epoch": 2.31, "learning_rate": 2.222348591766347e-05, "loss": 0.3353, "step": 22988 }, { "epoch": 2.31, "learning_rate": 2.2221174574198728e-05, "loss": 0.2188, "step": 22989 }, { "epoch": 2.31, "learning_rate": 2.2218863280235467e-05, "loss": 0.2992, "step": 22990 }, { "epoch": 2.31, "learning_rate": 2.2216552035788405e-05, "loss": 0.6882, "step": 22991 }, { "epoch": 2.31, "learning_rate": 2.2214240840872237e-05, "loss": 0.0392, "step": 22992 }, { "epoch": 2.31, "learning_rate": 2.2211929695501683e-05, "loss": 0.5582, "step": 22993 }, { "epoch": 2.31, "learning_rate": 2.2209618599691457e-05, "loss": 0.39, "step": 22994 }, { "epoch": 2.31, "learning_rate": 2.220730755345624e-05, "loss": 0.2425, "step": 22995 }, { "epoch": 2.31, "learning_rate": 2.220499655681076e-05, "loss": 0.4879, "step": 22996 }, { "epoch": 2.31, "learning_rate": 2.220268560976972e-05, "loss": 0.2863, "step": 22997 }, { "epoch": 2.31, "learning_rate": 2.2200374712347804e-05, "loss": 0.2809, "step": 22998 }, { "epoch": 2.31, "learning_rate": 2.2198063864559762e-05, "loss": 0.2415, "step": 22999 }, { "epoch": 2.31, "learning_rate": 2.2195753066420258e-05, "loss": 0.4386, "step": 23000 }, { "epoch": 2.31, "learning_rate": 2.2193442317944013e-05, "loss": 0.3808, "step": 23001 }, { "epoch": 2.31, "learning_rate": 2.219113161914573e-05, "loss": 0.387, "step": 23002 }, { "epoch": 2.31, "learning_rate": 2.218882097004013e-05, "loss": 0.2713, "step": 23003 }, { "epoch": 2.31, "learning_rate": 2.2186510370641884e-05, "loss": 0.1476, "step": 23004 }, { "epoch": 2.31, "learning_rate": 2.2184199820965723e-05, "loss": 1.0201, "step": 23005 }, { "epoch": 2.31, "learning_rate": 2.2181889321026343e-05, "loss": 1.0506, "step": 23006 }, { "epoch": 2.31, "learning_rate": 2.217957887083843e-05, "loss": 0.4507, "step": 23007 }, { "epoch": 2.31, "learning_rate": 2.217726847041672e-05, "loss": 0.4618, "step": 23008 }, { "epoch": 2.31, "learning_rate": 2.217495811977589e-05, "loss": 0.0793, "step": 23009 }, { "epoch": 2.31, "learning_rate": 2.2172647818930644e-05, "loss": 0.1538, "step": 23010 }, { "epoch": 2.31, "learning_rate": 2.21703375678957e-05, "loss": 0.7527, "step": 23011 }, { "epoch": 2.31, "learning_rate": 2.216802736668575e-05, "loss": 0.0867, "step": 23012 }, { "epoch": 2.31, "learning_rate": 2.216571721531548e-05, "loss": 0.6527, "step": 23013 }, { "epoch": 2.31, "learning_rate": 2.2163407113799616e-05, "loss": 0.1316, "step": 23014 }, { "epoch": 2.31, "learning_rate": 2.2161097062152854e-05, "loss": 0.2417, "step": 23015 }, { "epoch": 2.31, "learning_rate": 2.2158787060389868e-05, "loss": 0.5178, "step": 23016 }, { "epoch": 2.31, "learning_rate": 2.21564771085254e-05, "loss": 0.1679, "step": 23017 }, { "epoch": 2.31, "learning_rate": 2.215416720657412e-05, "loss": 0.505, "step": 23018 }, { "epoch": 2.31, "learning_rate": 2.2151857354550725e-05, "loss": 0.2915, "step": 23019 }, { "epoch": 2.31, "learning_rate": 2.2149547552469942e-05, "loss": 0.4464, "step": 23020 }, { "epoch": 2.31, "learning_rate": 2.2147237800346427e-05, "loss": 0.3488, "step": 23021 }, { "epoch": 2.31, "learning_rate": 2.214492809819493e-05, "loss": 0.7743, "step": 23022 }, { "epoch": 2.31, "learning_rate": 2.214261844603011e-05, "loss": 0.2423, "step": 23023 }, { "epoch": 2.31, "learning_rate": 2.2140308843866674e-05, "loss": 0.6428, "step": 23024 }, { "epoch": 2.31, "learning_rate": 2.2137999291719327e-05, "loss": 0.1746, "step": 23025 }, { "epoch": 2.31, "learning_rate": 2.213568978960277e-05, "loss": 0.2996, "step": 23026 }, { "epoch": 2.31, "learning_rate": 2.2133380337531678e-05, "loss": 0.2231, "step": 23027 }, { "epoch": 2.31, "learning_rate": 2.2131070935520765e-05, "loss": 0.5395, "step": 23028 }, { "epoch": 2.31, "learning_rate": 2.2128761583584732e-05, "loss": 0.1165, "step": 23029 }, { "epoch": 2.31, "learning_rate": 2.2126452281738247e-05, "loss": 0.3003, "step": 23030 }, { "epoch": 2.31, "learning_rate": 2.2124143029996044e-05, "loss": 0.5705, "step": 23031 }, { "epoch": 2.31, "learning_rate": 2.212183382837279e-05, "loss": 0.2061, "step": 23032 }, { "epoch": 2.31, "learning_rate": 2.2119524676883184e-05, "loss": 0.5554, "step": 23033 }, { "epoch": 2.31, "learning_rate": 2.2117215575541927e-05, "loss": 0.1221, "step": 23034 }, { "epoch": 2.31, "learning_rate": 2.2114906524363725e-05, "loss": 0.0455, "step": 23035 }, { "epoch": 2.31, "learning_rate": 2.2112597523363238e-05, "loss": 0.6802, "step": 23036 }, { "epoch": 2.31, "learning_rate": 2.211028857255519e-05, "loss": 0.6368, "step": 23037 }, { "epoch": 2.31, "learning_rate": 2.210797967195427e-05, "loss": 0.2175, "step": 23038 }, { "epoch": 2.31, "learning_rate": 2.210567082157515e-05, "loss": 0.6726, "step": 23039 }, { "epoch": 2.31, "learning_rate": 2.2103362021432548e-05, "loss": 0.2858, "step": 23040 }, { "epoch": 2.31, "learning_rate": 2.210105327154115e-05, "loss": 0.5796, "step": 23041 }, { "epoch": 2.31, "learning_rate": 2.2098744571915632e-05, "loss": 0.4406, "step": 23042 }, { "epoch": 2.31, "learning_rate": 2.2096435922570712e-05, "loss": 0.3134, "step": 23043 }, { "epoch": 2.31, "learning_rate": 2.209412732352106e-05, "loss": 0.3627, "step": 23044 }, { "epoch": 2.31, "learning_rate": 2.2091818774781366e-05, "loss": 0.0921, "step": 23045 }, { "epoch": 2.31, "learning_rate": 2.2089510276366335e-05, "loss": 0.227, "step": 23046 }, { "epoch": 2.31, "learning_rate": 2.208720182829065e-05, "loss": 0.6403, "step": 23047 }, { "epoch": 2.31, "learning_rate": 2.2084893430569006e-05, "loss": 0.4123, "step": 23048 }, { "epoch": 2.31, "learning_rate": 2.2082585083216095e-05, "loss": 0.2257, "step": 23049 }, { "epoch": 2.31, "learning_rate": 2.2080276786246587e-05, "loss": 0.4617, "step": 23050 }, { "epoch": 2.32, "learning_rate": 2.207796853967519e-05, "loss": 0.199, "step": 23051 }, { "epoch": 2.32, "learning_rate": 2.2075660343516597e-05, "loss": 0.3434, "step": 23052 }, { "epoch": 2.32, "learning_rate": 2.2073352197785466e-05, "loss": 0.231, "step": 23053 }, { "epoch": 2.32, "learning_rate": 2.207104410249653e-05, "loss": 0.1162, "step": 23054 }, { "epoch": 2.32, "learning_rate": 2.206873605766444e-05, "loss": 0.0582, "step": 23055 }, { "epoch": 2.32, "learning_rate": 2.206642806330389e-05, "loss": 0.1458, "step": 23056 }, { "epoch": 2.32, "learning_rate": 2.2064120119429585e-05, "loss": 0.4631, "step": 23057 }, { "epoch": 2.32, "learning_rate": 2.206181222605621e-05, "loss": 0.1695, "step": 23058 }, { "epoch": 2.32, "learning_rate": 2.205950438319842e-05, "loss": 0.4282, "step": 23059 }, { "epoch": 2.32, "learning_rate": 2.2057196590870935e-05, "loss": 0.2079, "step": 23060 }, { "epoch": 2.32, "learning_rate": 2.2054888849088434e-05, "loss": 0.4006, "step": 23061 }, { "epoch": 2.32, "learning_rate": 2.205258115786558e-05, "loss": 0.1315, "step": 23062 }, { "epoch": 2.32, "learning_rate": 2.2050273517217095e-05, "loss": 0.2934, "step": 23063 }, { "epoch": 2.32, "learning_rate": 2.2047965927157636e-05, "loss": 0.2457, "step": 23064 }, { "epoch": 2.32, "learning_rate": 2.2045658387701894e-05, "loss": 0.5282, "step": 23065 }, { "epoch": 2.32, "learning_rate": 2.2043350898864563e-05, "loss": 0.388, "step": 23066 }, { "epoch": 2.32, "learning_rate": 2.2041043460660324e-05, "loss": 0.439, "step": 23067 }, { "epoch": 2.32, "learning_rate": 2.2038736073103847e-05, "loss": 0.2141, "step": 23068 }, { "epoch": 2.32, "learning_rate": 2.2036428736209828e-05, "loss": 0.2137, "step": 23069 }, { "epoch": 2.32, "learning_rate": 2.2034121449992955e-05, "loss": 0.3134, "step": 23070 }, { "epoch": 2.32, "learning_rate": 2.203181421446788e-05, "loss": 0.0955, "step": 23071 }, { "epoch": 2.32, "learning_rate": 2.202950702964933e-05, "loss": 0.3377, "step": 23072 }, { "epoch": 2.32, "learning_rate": 2.202719989555196e-05, "loss": 0.5003, "step": 23073 }, { "epoch": 2.32, "learning_rate": 2.202489281219045e-05, "loss": 0.4569, "step": 23074 }, { "epoch": 2.32, "learning_rate": 2.2022585779579503e-05, "loss": 0.5917, "step": 23075 }, { "epoch": 2.32, "learning_rate": 2.202027879773376e-05, "loss": 0.1248, "step": 23076 }, { "epoch": 2.32, "learning_rate": 2.201797186666795e-05, "loss": 0.3385, "step": 23077 }, { "epoch": 2.32, "learning_rate": 2.2015664986396726e-05, "loss": 0.2241, "step": 23078 }, { "epoch": 2.32, "learning_rate": 2.2013358156934765e-05, "loss": 0.5231, "step": 23079 }, { "epoch": 2.32, "learning_rate": 2.201105137829676e-05, "loss": 0.1361, "step": 23080 }, { "epoch": 2.32, "learning_rate": 2.2008744650497392e-05, "loss": 0.3051, "step": 23081 }, { "epoch": 2.32, "learning_rate": 2.200643797355132e-05, "loss": 0.1418, "step": 23082 }, { "epoch": 2.32, "learning_rate": 2.200413134747324e-05, "loss": 0.8671, "step": 23083 }, { "epoch": 2.32, "learning_rate": 2.2001824772277837e-05, "loss": 0.261, "step": 23084 }, { "epoch": 2.32, "learning_rate": 2.1999518247979752e-05, "loss": 0.5422, "step": 23085 }, { "epoch": 2.32, "learning_rate": 2.1997211774593717e-05, "loss": 0.3873, "step": 23086 }, { "epoch": 2.32, "learning_rate": 2.199490535213437e-05, "loss": 0.3219, "step": 23087 }, { "epoch": 2.32, "learning_rate": 2.1992598980616397e-05, "loss": 0.1739, "step": 23088 }, { "epoch": 2.32, "learning_rate": 2.199029266005448e-05, "loss": 0.7078, "step": 23089 }, { "epoch": 2.32, "learning_rate": 2.1987986390463305e-05, "loss": 0.1662, "step": 23090 }, { "epoch": 2.32, "learning_rate": 2.198568017185752e-05, "loss": 0.2607, "step": 23091 }, { "epoch": 2.32, "learning_rate": 2.1983374004251827e-05, "loss": 0.841, "step": 23092 }, { "epoch": 2.32, "learning_rate": 2.1981067887660898e-05, "loss": 0.2842, "step": 23093 }, { "epoch": 2.32, "learning_rate": 2.197876182209938e-05, "loss": 0.8333, "step": 23094 }, { "epoch": 2.32, "learning_rate": 2.1976455807581994e-05, "loss": 0.5045, "step": 23095 }, { "epoch": 2.32, "learning_rate": 2.197414984412338e-05, "loss": 0.098, "step": 23096 }, { "epoch": 2.32, "learning_rate": 2.1971843931738217e-05, "loss": 0.1818, "step": 23097 }, { "epoch": 2.32, "learning_rate": 2.1969538070441195e-05, "loss": 0.2167, "step": 23098 }, { "epoch": 2.32, "learning_rate": 2.1967232260246983e-05, "loss": 0.3758, "step": 23099 }, { "epoch": 2.32, "learning_rate": 2.1964926501170237e-05, "loss": 0.3713, "step": 23100 }, { "epoch": 2.32, "learning_rate": 2.1962620793225647e-05, "loss": 0.5553, "step": 23101 }, { "epoch": 2.32, "learning_rate": 2.1960315136427878e-05, "loss": 0.124, "step": 23102 }, { "epoch": 2.32, "learning_rate": 2.1958009530791607e-05, "loss": 0.6928, "step": 23103 }, { "epoch": 2.32, "learning_rate": 2.195570397633151e-05, "loss": 0.1472, "step": 23104 }, { "epoch": 2.32, "learning_rate": 2.1953398473062243e-05, "loss": 0.152, "step": 23105 }, { "epoch": 2.32, "learning_rate": 2.195109302099849e-05, "loss": 0.3515, "step": 23106 }, { "epoch": 2.32, "learning_rate": 2.1948787620154932e-05, "loss": 0.3773, "step": 23107 }, { "epoch": 2.32, "learning_rate": 2.19464822705462e-05, "loss": 0.5855, "step": 23108 }, { "epoch": 2.32, "learning_rate": 2.1944176972187012e-05, "loss": 0.0688, "step": 23109 }, { "epoch": 2.32, "learning_rate": 2.1941871725092012e-05, "loss": 0.2358, "step": 23110 }, { "epoch": 2.32, "learning_rate": 2.1939566529275872e-05, "loss": 0.643, "step": 23111 }, { "epoch": 2.32, "learning_rate": 2.1937261384753265e-05, "loss": 0.1892, "step": 23112 }, { "epoch": 2.32, "learning_rate": 2.193495629153887e-05, "loss": 0.4119, "step": 23113 }, { "epoch": 2.32, "learning_rate": 2.193265124964733e-05, "loss": 0.1096, "step": 23114 }, { "epoch": 2.32, "learning_rate": 2.1930346259093333e-05, "loss": 0.1931, "step": 23115 }, { "epoch": 2.32, "learning_rate": 2.1928041319891553e-05, "loss": 0.2922, "step": 23116 }, { "epoch": 2.32, "learning_rate": 2.1925736432056627e-05, "loss": 0.0462, "step": 23117 }, { "epoch": 2.32, "learning_rate": 2.192343159560326e-05, "loss": 0.6641, "step": 23118 }, { "epoch": 2.32, "learning_rate": 2.1921126810546095e-05, "loss": 0.4859, "step": 23119 }, { "epoch": 2.32, "learning_rate": 2.19188220768998e-05, "loss": 0.3216, "step": 23120 }, { "epoch": 2.32, "learning_rate": 2.191651739467906e-05, "loss": 0.3724, "step": 23121 }, { "epoch": 2.32, "learning_rate": 2.1914212763898524e-05, "loss": 0.886, "step": 23122 }, { "epoch": 2.32, "learning_rate": 2.1911908184572858e-05, "loss": 0.2838, "step": 23123 }, { "epoch": 2.32, "learning_rate": 2.1909603656716732e-05, "loss": 0.4599, "step": 23124 }, { "epoch": 2.32, "learning_rate": 2.1907299180344817e-05, "loss": 0.0076, "step": 23125 }, { "epoch": 2.32, "learning_rate": 2.1904994755471753e-05, "loss": 0.2106, "step": 23126 }, { "epoch": 2.32, "learning_rate": 2.1902690382112245e-05, "loss": 0.094, "step": 23127 }, { "epoch": 2.32, "learning_rate": 2.190038606028092e-05, "loss": 0.2946, "step": 23128 }, { "epoch": 2.32, "learning_rate": 2.1898081789992465e-05, "loss": 0.8332, "step": 23129 }, { "epoch": 2.32, "learning_rate": 2.1895777571261532e-05, "loss": 0.2009, "step": 23130 }, { "epoch": 2.32, "learning_rate": 2.1893473404102786e-05, "loss": 0.2535, "step": 23131 }, { "epoch": 2.32, "learning_rate": 2.18911692885309e-05, "loss": 0.2118, "step": 23132 }, { "epoch": 2.32, "learning_rate": 2.1888865224560524e-05, "loss": 0.3065, "step": 23133 }, { "epoch": 2.32, "learning_rate": 2.1886561212206315e-05, "loss": 0.5464, "step": 23134 }, { "epoch": 2.32, "learning_rate": 2.188425725148295e-05, "loss": 0.6918, "step": 23135 }, { "epoch": 2.32, "learning_rate": 2.1881953342405093e-05, "loss": 0.2332, "step": 23136 }, { "epoch": 2.32, "learning_rate": 2.1879649484987383e-05, "loss": 0.3283, "step": 23137 }, { "epoch": 2.32, "learning_rate": 2.1877345679244504e-05, "loss": 0.3196, "step": 23138 }, { "epoch": 2.32, "learning_rate": 2.187504192519111e-05, "loss": 0.1787, "step": 23139 }, { "epoch": 2.32, "learning_rate": 2.1872738222841838e-05, "loss": 0.4391, "step": 23140 }, { "epoch": 2.32, "learning_rate": 2.1870434572211393e-05, "loss": 0.9859, "step": 23141 }, { "epoch": 2.32, "learning_rate": 2.1868130973314396e-05, "loss": 0.0957, "step": 23142 }, { "epoch": 2.32, "learning_rate": 2.1865827426165522e-05, "loss": 0.1816, "step": 23143 }, { "epoch": 2.32, "learning_rate": 2.1863523930779425e-05, "loss": 0.2267, "step": 23144 }, { "epoch": 2.32, "learning_rate": 2.1861220487170777e-05, "loss": 0.1222, "step": 23145 }, { "epoch": 2.32, "learning_rate": 2.1858917095354217e-05, "loss": 0.0623, "step": 23146 }, { "epoch": 2.32, "learning_rate": 2.185661375534441e-05, "loss": 0.1776, "step": 23147 }, { "epoch": 2.32, "learning_rate": 2.185431046715603e-05, "loss": 0.0359, "step": 23148 }, { "epoch": 2.33, "learning_rate": 2.1852007230803694e-05, "loss": 0.1725, "step": 23149 }, { "epoch": 2.33, "learning_rate": 2.1849704046302107e-05, "loss": 0.4503, "step": 23150 }, { "epoch": 2.33, "learning_rate": 2.1847400913665895e-05, "loss": 1.0328, "step": 23151 }, { "epoch": 2.33, "learning_rate": 2.1845097832909713e-05, "loss": 0.4944, "step": 23152 }, { "epoch": 2.33, "learning_rate": 2.1842794804048235e-05, "loss": 0.174, "step": 23153 }, { "epoch": 2.33, "learning_rate": 2.1840491827096117e-05, "loss": 0.1678, "step": 23154 }, { "epoch": 2.33, "learning_rate": 2.183818890206799e-05, "loss": 0.2615, "step": 23155 }, { "epoch": 2.33, "learning_rate": 2.183588602897853e-05, "loss": 0.5046, "step": 23156 }, { "epoch": 2.33, "learning_rate": 2.183358320784238e-05, "loss": 0.1411, "step": 23157 }, { "epoch": 2.33, "learning_rate": 2.1831280438674204e-05, "loss": 0.8826, "step": 23158 }, { "epoch": 2.33, "learning_rate": 2.182897772148866e-05, "loss": 0.3361, "step": 23159 }, { "epoch": 2.33, "learning_rate": 2.182667505630038e-05, "loss": 0.3981, "step": 23160 }, { "epoch": 2.33, "learning_rate": 2.1824372443124037e-05, "loss": 0.1769, "step": 23161 }, { "epoch": 2.33, "learning_rate": 2.1822069881974277e-05, "loss": 0.1999, "step": 23162 }, { "epoch": 2.33, "learning_rate": 2.1819767372865747e-05, "loss": 0.3283, "step": 23163 }, { "epoch": 2.33, "learning_rate": 2.181746491581312e-05, "loss": 0.5977, "step": 23164 }, { "epoch": 2.33, "learning_rate": 2.1815162510831025e-05, "loss": 0.3738, "step": 23165 }, { "epoch": 2.33, "learning_rate": 2.1812860157934117e-05, "loss": 0.4177, "step": 23166 }, { "epoch": 2.33, "learning_rate": 2.1810557857137057e-05, "loss": 0.6011, "step": 23167 }, { "epoch": 2.33, "learning_rate": 2.18082556084545e-05, "loss": 0.4468, "step": 23168 }, { "epoch": 2.33, "learning_rate": 2.1805953411901076e-05, "loss": 0.272, "step": 23169 }, { "epoch": 2.33, "learning_rate": 2.180365126749145e-05, "loss": 0.2559, "step": 23170 }, { "epoch": 2.33, "learning_rate": 2.180134917524028e-05, "loss": 0.6752, "step": 23171 }, { "epoch": 2.33, "learning_rate": 2.1799047135162184e-05, "loss": 0.1394, "step": 23172 }, { "epoch": 2.33, "learning_rate": 2.179674514727185e-05, "loss": 0.1794, "step": 23173 }, { "epoch": 2.33, "learning_rate": 2.17944432115839e-05, "loss": 0.1884, "step": 23174 }, { "epoch": 2.33, "learning_rate": 2.179214132811299e-05, "loss": 0.1029, "step": 23175 }, { "epoch": 2.33, "learning_rate": 2.1789839496873775e-05, "loss": 0.4898, "step": 23176 }, { "epoch": 2.33, "learning_rate": 2.1787537717880906e-05, "loss": 0.2634, "step": 23177 }, { "epoch": 2.33, "learning_rate": 2.1785235991149012e-05, "loss": 0.3348, "step": 23178 }, { "epoch": 2.33, "learning_rate": 2.178293431669275e-05, "loss": 0.738, "step": 23179 }, { "epoch": 2.33, "learning_rate": 2.1780632694526783e-05, "loss": 0.306, "step": 23180 }, { "epoch": 2.33, "learning_rate": 2.177833112466572e-05, "loss": 0.2778, "step": 23181 }, { "epoch": 2.33, "learning_rate": 2.177602960712425e-05, "loss": 0.2624, "step": 23182 }, { "epoch": 2.33, "learning_rate": 2.1773728141916983e-05, "loss": 0.1521, "step": 23183 }, { "epoch": 2.33, "learning_rate": 2.177142672905859e-05, "loss": 0.6876, "step": 23184 }, { "epoch": 2.33, "learning_rate": 2.1769125368563703e-05, "loss": 0.4814, "step": 23185 }, { "epoch": 2.33, "learning_rate": 2.176682406044697e-05, "loss": 0.505, "step": 23186 }, { "epoch": 2.33, "learning_rate": 2.1764522804723047e-05, "loss": 0.1683, "step": 23187 }, { "epoch": 2.33, "learning_rate": 2.1762221601406563e-05, "loss": 0.2214, "step": 23188 }, { "epoch": 2.33, "learning_rate": 2.175992045051216e-05, "loss": 0.1046, "step": 23189 }, { "epoch": 2.33, "learning_rate": 2.175761935205449e-05, "loss": 0.4508, "step": 23190 }, { "epoch": 2.33, "learning_rate": 2.1755318306048207e-05, "loss": 0.5692, "step": 23191 }, { "epoch": 2.33, "learning_rate": 2.175301731250793e-05, "loss": 0.6249, "step": 23192 }, { "epoch": 2.33, "learning_rate": 2.1750716371448317e-05, "loss": 0.1483, "step": 23193 }, { "epoch": 2.33, "learning_rate": 2.1748415482884008e-05, "loss": 0.1523, "step": 23194 }, { "epoch": 2.33, "learning_rate": 2.1746114646829636e-05, "loss": 0.2198, "step": 23195 }, { "epoch": 2.33, "learning_rate": 2.1743813863299865e-05, "loss": 0.1671, "step": 23196 }, { "epoch": 2.33, "learning_rate": 2.1741513132309315e-05, "loss": 0.3187, "step": 23197 }, { "epoch": 2.33, "learning_rate": 2.1739212453872626e-05, "loss": 0.104, "step": 23198 }, { "epoch": 2.33, "learning_rate": 2.1736911828004452e-05, "loss": 0.5403, "step": 23199 }, { "epoch": 2.33, "learning_rate": 2.173461125471944e-05, "loss": 0.0741, "step": 23200 }, { "epoch": 2.33, "learning_rate": 2.1732310734032195e-05, "loss": 0.6234, "step": 23201 }, { "epoch": 2.33, "learning_rate": 2.1730010265957393e-05, "loss": 0.7689, "step": 23202 }, { "epoch": 2.33, "learning_rate": 2.1727709850509665e-05, "loss": 0.1863, "step": 23203 }, { "epoch": 2.33, "learning_rate": 2.1725409487703626e-05, "loss": 0.2122, "step": 23204 }, { "epoch": 2.33, "learning_rate": 2.172310917755395e-05, "loss": 0.4599, "step": 23205 }, { "epoch": 2.33, "learning_rate": 2.172080892007525e-05, "loss": 0.191, "step": 23206 }, { "epoch": 2.33, "learning_rate": 2.1718508715282174e-05, "loss": 0.3301, "step": 23207 }, { "epoch": 2.33, "learning_rate": 2.171620856318936e-05, "loss": 0.2027, "step": 23208 }, { "epoch": 2.33, "learning_rate": 2.171390846381143e-05, "loss": 0.4102, "step": 23209 }, { "epoch": 2.33, "learning_rate": 2.1711608417163058e-05, "loss": 0.3394, "step": 23210 }, { "epoch": 2.33, "learning_rate": 2.1709308423258848e-05, "loss": 0.3629, "step": 23211 }, { "epoch": 2.33, "learning_rate": 2.1707008482113433e-05, "loss": 0.1962, "step": 23212 }, { "epoch": 2.33, "learning_rate": 2.170470859374147e-05, "loss": 0.4833, "step": 23213 }, { "epoch": 2.33, "learning_rate": 2.1702408758157598e-05, "loss": 0.3713, "step": 23214 }, { "epoch": 2.33, "learning_rate": 2.1700108975376424e-05, "loss": 0.5782, "step": 23215 }, { "epoch": 2.33, "learning_rate": 2.1697809245412603e-05, "loss": 0.8351, "step": 23216 }, { "epoch": 2.33, "learning_rate": 2.1695509568280768e-05, "loss": 0.5151, "step": 23217 }, { "epoch": 2.33, "learning_rate": 2.1693209943995545e-05, "loss": 0.5214, "step": 23218 }, { "epoch": 2.33, "learning_rate": 2.1690910372571586e-05, "loss": 0.1348, "step": 23219 }, { "epoch": 2.33, "learning_rate": 2.1688610854023505e-05, "loss": 0.0891, "step": 23220 }, { "epoch": 2.33, "learning_rate": 2.1686311388365937e-05, "loss": 0.0595, "step": 23221 }, { "epoch": 2.33, "learning_rate": 2.1684011975613526e-05, "loss": 0.7676, "step": 23222 }, { "epoch": 2.33, "learning_rate": 2.1681712615780907e-05, "loss": 0.3288, "step": 23223 }, { "epoch": 2.33, "learning_rate": 2.167941330888269e-05, "loss": 0.1809, "step": 23224 }, { "epoch": 2.33, "learning_rate": 2.167711405493353e-05, "loss": 0.1729, "step": 23225 }, { "epoch": 2.33, "learning_rate": 2.167481485394805e-05, "loss": 0.3586, "step": 23226 }, { "epoch": 2.33, "learning_rate": 2.167251570594087e-05, "loss": 0.5493, "step": 23227 }, { "epoch": 2.33, "learning_rate": 2.1670216610926652e-05, "loss": 0.3809, "step": 23228 }, { "epoch": 2.33, "learning_rate": 2.1667917568919995e-05, "loss": 0.3085, "step": 23229 }, { "epoch": 2.33, "learning_rate": 2.1665618579935538e-05, "loss": 0.1718, "step": 23230 }, { "epoch": 2.33, "learning_rate": 2.1663319643987917e-05, "loss": 0.1626, "step": 23231 }, { "epoch": 2.33, "learning_rate": 2.1661020761091764e-05, "loss": 0.6481, "step": 23232 }, { "epoch": 2.33, "learning_rate": 2.1658721931261694e-05, "loss": 0.3016, "step": 23233 }, { "epoch": 2.33, "learning_rate": 2.165642315451235e-05, "loss": 0.2337, "step": 23234 }, { "epoch": 2.33, "learning_rate": 2.165412443085836e-05, "loss": 0.2394, "step": 23235 }, { "epoch": 2.33, "learning_rate": 2.1651825760314327e-05, "loss": 0.275, "step": 23236 }, { "epoch": 2.33, "learning_rate": 2.1649527142894923e-05, "loss": 0.6918, "step": 23237 }, { "epoch": 2.33, "learning_rate": 2.1647228578614733e-05, "loss": 0.3576, "step": 23238 }, { "epoch": 2.33, "learning_rate": 2.1644930067488413e-05, "loss": 0.4134, "step": 23239 }, { "epoch": 2.33, "learning_rate": 2.1642631609530577e-05, "loss": 0.2063, "step": 23240 }, { "epoch": 2.33, "learning_rate": 2.164033320475585e-05, "loss": 0.4814, "step": 23241 }, { "epoch": 2.33, "learning_rate": 2.163803485317887e-05, "loss": 0.2317, "step": 23242 }, { "epoch": 2.33, "learning_rate": 2.1635736554814257e-05, "loss": 0.1016, "step": 23243 }, { "epoch": 2.33, "learning_rate": 2.1633438309676624e-05, "loss": 0.2524, "step": 23244 }, { "epoch": 2.33, "learning_rate": 2.1631140117780615e-05, "loss": 0.292, "step": 23245 }, { "epoch": 2.33, "learning_rate": 2.1628841979140848e-05, "loss": 0.1173, "step": 23246 }, { "epoch": 2.33, "learning_rate": 2.162654389377194e-05, "loss": 0.2511, "step": 23247 }, { "epoch": 2.34, "learning_rate": 2.1624245861688523e-05, "loss": 0.5532, "step": 23248 }, { "epoch": 2.34, "learning_rate": 2.162194788290522e-05, "loss": 0.6551, "step": 23249 }, { "epoch": 2.34, "learning_rate": 2.1619649957436644e-05, "loss": 0.5314, "step": 23250 }, { "epoch": 2.34, "learning_rate": 2.1617352085297447e-05, "loss": 0.3675, "step": 23251 }, { "epoch": 2.34, "learning_rate": 2.1615054266502223e-05, "loss": 0.4391, "step": 23252 }, { "epoch": 2.34, "learning_rate": 2.1612756501065596e-05, "loss": 0.3572, "step": 23253 }, { "epoch": 2.34, "learning_rate": 2.1610458789002203e-05, "loss": 0.6551, "step": 23254 }, { "epoch": 2.34, "learning_rate": 2.1608161130326665e-05, "loss": 0.6123, "step": 23255 }, { "epoch": 2.34, "learning_rate": 2.1605863525053587e-05, "loss": 0.3428, "step": 23256 }, { "epoch": 2.34, "learning_rate": 2.1603565973197605e-05, "loss": 0.1058, "step": 23257 }, { "epoch": 2.34, "learning_rate": 2.1601268474773333e-05, "loss": 0.4701, "step": 23258 }, { "epoch": 2.34, "learning_rate": 2.1598971029795392e-05, "loss": 0.4074, "step": 23259 }, { "epoch": 2.34, "learning_rate": 2.159667363827841e-05, "loss": 0.288, "step": 23260 }, { "epoch": 2.34, "learning_rate": 2.1594376300236997e-05, "loss": 0.1866, "step": 23261 }, { "epoch": 2.34, "learning_rate": 2.1592079015685768e-05, "loss": 0.8424, "step": 23262 }, { "epoch": 2.34, "learning_rate": 2.1589781784639358e-05, "loss": 0.2481, "step": 23263 }, { "epoch": 2.34, "learning_rate": 2.1587484607112365e-05, "loss": 1.0681, "step": 23264 }, { "epoch": 2.34, "learning_rate": 2.158518748311944e-05, "loss": 0.8861, "step": 23265 }, { "epoch": 2.34, "learning_rate": 2.1582890412675173e-05, "loss": 0.1555, "step": 23266 }, { "epoch": 2.34, "learning_rate": 2.1580593395794182e-05, "loss": 0.5231, "step": 23267 }, { "epoch": 2.34, "learning_rate": 2.1578296432491095e-05, "loss": 0.1168, "step": 23268 }, { "epoch": 2.34, "learning_rate": 2.1575999522780538e-05, "loss": 0.2522, "step": 23269 }, { "epoch": 2.34, "learning_rate": 2.15737026666771e-05, "loss": 0.2971, "step": 23270 }, { "epoch": 2.34, "learning_rate": 2.1571405864195416e-05, "loss": 0.0836, "step": 23271 }, { "epoch": 2.34, "learning_rate": 2.1569109115350103e-05, "loss": 0.1543, "step": 23272 }, { "epoch": 2.34, "learning_rate": 2.1566812420155763e-05, "loss": 0.5751, "step": 23273 }, { "epoch": 2.34, "learning_rate": 2.156451577862703e-05, "loss": 0.3141, "step": 23274 }, { "epoch": 2.34, "learning_rate": 2.1562219190778512e-05, "loss": 0.1813, "step": 23275 }, { "epoch": 2.34, "learning_rate": 2.155992265662481e-05, "loss": 0.5949, "step": 23276 }, { "epoch": 2.34, "learning_rate": 2.1557626176180555e-05, "loss": 0.4438, "step": 23277 }, { "epoch": 2.34, "learning_rate": 2.1555329749460362e-05, "loss": 0.0966, "step": 23278 }, { "epoch": 2.34, "learning_rate": 2.1553033376478825e-05, "loss": 0.6231, "step": 23279 }, { "epoch": 2.34, "learning_rate": 2.155073705725058e-05, "loss": 0.0453, "step": 23280 }, { "epoch": 2.34, "learning_rate": 2.1548440791790225e-05, "loss": 0.7398, "step": 23281 }, { "epoch": 2.34, "learning_rate": 2.1546144580112378e-05, "loss": 0.1053, "step": 23282 }, { "epoch": 2.34, "learning_rate": 2.154384842223166e-05, "loss": 0.3366, "step": 23283 }, { "epoch": 2.34, "learning_rate": 2.1541552318162664e-05, "loss": 0.2231, "step": 23284 }, { "epoch": 2.34, "learning_rate": 2.153925626792001e-05, "loss": 0.1695, "step": 23285 }, { "epoch": 2.34, "learning_rate": 2.1536960271518317e-05, "loss": 0.1881, "step": 23286 }, { "epoch": 2.34, "learning_rate": 2.1534664328972195e-05, "loss": 0.5544, "step": 23287 }, { "epoch": 2.34, "learning_rate": 2.1532368440296233e-05, "loss": 0.3729, "step": 23288 }, { "epoch": 2.34, "learning_rate": 2.1530072605505062e-05, "loss": 0.671, "step": 23289 }, { "epoch": 2.34, "learning_rate": 2.1527776824613285e-05, "loss": 0.174, "step": 23290 }, { "epoch": 2.34, "learning_rate": 2.1525481097635516e-05, "loss": 0.1359, "step": 23291 }, { "epoch": 2.34, "learning_rate": 2.1523185424586374e-05, "loss": 0.6913, "step": 23292 }, { "epoch": 2.34, "learning_rate": 2.1520889805480437e-05, "loss": 0.5692, "step": 23293 }, { "epoch": 2.34, "learning_rate": 2.1518594240332335e-05, "loss": 0.1173, "step": 23294 }, { "epoch": 2.34, "learning_rate": 2.1516298729156682e-05, "loss": 0.6805, "step": 23295 }, { "epoch": 2.34, "learning_rate": 2.1514003271968065e-05, "loss": 0.4888, "step": 23296 }, { "epoch": 2.34, "learning_rate": 2.1511707868781113e-05, "loss": 0.3853, "step": 23297 }, { "epoch": 2.34, "learning_rate": 2.150941251961042e-05, "loss": 0.1194, "step": 23298 }, { "epoch": 2.34, "learning_rate": 2.150711722447059e-05, "loss": 0.4508, "step": 23299 }, { "epoch": 2.34, "learning_rate": 2.150482198337624e-05, "loss": 1.0413, "step": 23300 }, { "epoch": 2.34, "learning_rate": 2.150252679634198e-05, "loss": 0.1061, "step": 23301 }, { "epoch": 2.34, "learning_rate": 2.150023166338239e-05, "loss": 0.6147, "step": 23302 }, { "epoch": 2.34, "learning_rate": 2.1497936584512102e-05, "loss": 0.3786, "step": 23303 }, { "epoch": 2.34, "learning_rate": 2.149564155974571e-05, "loss": 0.1892, "step": 23304 }, { "epoch": 2.34, "learning_rate": 2.1493346589097814e-05, "loss": 0.6518, "step": 23305 }, { "epoch": 2.34, "learning_rate": 2.1491051672583036e-05, "loss": 0.5363, "step": 23306 }, { "epoch": 2.34, "learning_rate": 2.1488756810215962e-05, "loss": 0.7821, "step": 23307 }, { "epoch": 2.34, "learning_rate": 2.1486462002011196e-05, "loss": 0.3668, "step": 23308 }, { "epoch": 2.34, "learning_rate": 2.1484167247983356e-05, "loss": 0.46, "step": 23309 }, { "epoch": 2.34, "learning_rate": 2.1481872548147036e-05, "loss": 0.4657, "step": 23310 }, { "epoch": 2.34, "learning_rate": 2.1479577902516835e-05, "loss": 0.254, "step": 23311 }, { "epoch": 2.34, "learning_rate": 2.1477283311107353e-05, "loss": 0.1161, "step": 23312 }, { "epoch": 2.34, "learning_rate": 2.147498877393321e-05, "loss": 0.207, "step": 23313 }, { "epoch": 2.34, "learning_rate": 2.1472694291008983e-05, "loss": 0.2668, "step": 23314 }, { "epoch": 2.34, "learning_rate": 2.14703998623493e-05, "loss": 0.3617, "step": 23315 }, { "epoch": 2.34, "learning_rate": 2.1468105487968738e-05, "loss": 0.3157, "step": 23316 }, { "epoch": 2.34, "learning_rate": 2.1465811167881904e-05, "loss": 0.1216, "step": 23317 }, { "epoch": 2.34, "learning_rate": 2.1463516902103403e-05, "loss": 0.2413, "step": 23318 }, { "epoch": 2.34, "learning_rate": 2.146122269064783e-05, "loss": 0.1241, "step": 23319 }, { "epoch": 2.34, "learning_rate": 2.14589285335298e-05, "loss": 0.4302, "step": 23320 }, { "epoch": 2.34, "learning_rate": 2.1456634430763894e-05, "loss": 0.2178, "step": 23321 }, { "epoch": 2.34, "learning_rate": 2.145434038236471e-05, "loss": 0.4607, "step": 23322 }, { "epoch": 2.34, "learning_rate": 2.145204638834686e-05, "loss": 0.2885, "step": 23323 }, { "epoch": 2.34, "learning_rate": 2.1449752448724938e-05, "loss": 0.0706, "step": 23324 }, { "epoch": 2.34, "learning_rate": 2.1447458563513526e-05, "loss": 0.3978, "step": 23325 }, { "epoch": 2.34, "learning_rate": 2.1445164732727244e-05, "loss": 0.4069, "step": 23326 }, { "epoch": 2.34, "learning_rate": 2.1442870956380678e-05, "loss": 0.2596, "step": 23327 }, { "epoch": 2.34, "learning_rate": 2.144057723448842e-05, "loss": 0.1112, "step": 23328 }, { "epoch": 2.34, "learning_rate": 2.1438283567065083e-05, "loss": 0.9421, "step": 23329 }, { "epoch": 2.34, "learning_rate": 2.1435989954125247e-05, "loss": 0.5154, "step": 23330 }, { "epoch": 2.34, "learning_rate": 2.1433696395683506e-05, "loss": 0.1571, "step": 23331 }, { "epoch": 2.34, "learning_rate": 2.143140289175447e-05, "loss": 0.0735, "step": 23332 }, { "epoch": 2.34, "learning_rate": 2.1429109442352734e-05, "loss": 0.0702, "step": 23333 }, { "epoch": 2.34, "learning_rate": 2.142681604749287e-05, "loss": 0.2263, "step": 23334 }, { "epoch": 2.34, "learning_rate": 2.142452270718949e-05, "loss": 0.088, "step": 23335 }, { "epoch": 2.34, "learning_rate": 2.142222942145719e-05, "loss": 0.6682, "step": 23336 }, { "epoch": 2.34, "learning_rate": 2.1419936190310553e-05, "loss": 0.5286, "step": 23337 }, { "epoch": 2.34, "learning_rate": 2.141764301376419e-05, "loss": 0.2955, "step": 23338 }, { "epoch": 2.34, "learning_rate": 2.1415349891832674e-05, "loss": 0.7328, "step": 23339 }, { "epoch": 2.34, "learning_rate": 2.1413056824530602e-05, "loss": 0.4204, "step": 23340 }, { "epoch": 2.34, "learning_rate": 2.1410763811872576e-05, "loss": 1.1078, "step": 23341 }, { "epoch": 2.34, "learning_rate": 2.1408470853873187e-05, "loss": 0.3584, "step": 23342 }, { "epoch": 2.34, "learning_rate": 2.1406177950547006e-05, "loss": 0.2696, "step": 23343 }, { "epoch": 2.34, "learning_rate": 2.1403885101908647e-05, "loss": 0.4814, "step": 23344 }, { "epoch": 2.34, "learning_rate": 2.1401592307972686e-05, "loss": 0.7687, "step": 23345 }, { "epoch": 2.35, "learning_rate": 2.1399299568753728e-05, "loss": 0.3695, "step": 23346 }, { "epoch": 2.35, "learning_rate": 2.1397006884266366e-05, "loss": 0.2904, "step": 23347 }, { "epoch": 2.35, "learning_rate": 2.139471425452516e-05, "loss": 0.4496, "step": 23348 }, { "epoch": 2.35, "learning_rate": 2.1392421679544728e-05, "loss": 0.0533, "step": 23349 }, { "epoch": 2.35, "learning_rate": 2.139012915933965e-05, "loss": 0.1843, "step": 23350 }, { "epoch": 2.35, "learning_rate": 2.138783669392451e-05, "loss": 0.1732, "step": 23351 }, { "epoch": 2.35, "learning_rate": 2.138554428331391e-05, "loss": 0.3096, "step": 23352 }, { "epoch": 2.35, "learning_rate": 2.1383251927522428e-05, "loss": 0.3283, "step": 23353 }, { "epoch": 2.35, "learning_rate": 2.1380959626564642e-05, "loss": 0.1488, "step": 23354 }, { "epoch": 2.35, "learning_rate": 2.1378667380455157e-05, "loss": 0.3151, "step": 23355 }, { "epoch": 2.35, "learning_rate": 2.137637518920856e-05, "loss": 0.1747, "step": 23356 }, { "epoch": 2.35, "learning_rate": 2.1374083052839416e-05, "loss": 0.5527, "step": 23357 }, { "epoch": 2.35, "learning_rate": 2.1371790971362334e-05, "loss": 0.1804, "step": 23358 }, { "epoch": 2.35, "learning_rate": 2.1369498944791893e-05, "loss": 0.3766, "step": 23359 }, { "epoch": 2.35, "learning_rate": 2.136720697314267e-05, "loss": 0.2754, "step": 23360 }, { "epoch": 2.35, "learning_rate": 2.136491505642927e-05, "loss": 0.3532, "step": 23361 }, { "epoch": 2.35, "learning_rate": 2.136262319466626e-05, "loss": 0.1572, "step": 23362 }, { "epoch": 2.35, "learning_rate": 2.1360331387868226e-05, "loss": 0.1805, "step": 23363 }, { "epoch": 2.35, "learning_rate": 2.1358039636049764e-05, "loss": 0.8694, "step": 23364 }, { "epoch": 2.35, "learning_rate": 2.1355747939225448e-05, "loss": 0.1065, "step": 23365 }, { "epoch": 2.35, "learning_rate": 2.1353456297409857e-05, "loss": 0.1688, "step": 23366 }, { "epoch": 2.35, "learning_rate": 2.1351164710617584e-05, "loss": 0.751, "step": 23367 }, { "epoch": 2.35, "learning_rate": 2.1348873178863215e-05, "loss": 0.3764, "step": 23368 }, { "epoch": 2.35, "learning_rate": 2.1346581702161316e-05, "loss": 0.4408, "step": 23369 }, { "epoch": 2.35, "learning_rate": 2.134429028052649e-05, "loss": 0.8098, "step": 23370 }, { "epoch": 2.35, "learning_rate": 2.13419989139733e-05, "loss": 0.4355, "step": 23371 }, { "epoch": 2.35, "learning_rate": 2.1339707602516335e-05, "loss": 0.3276, "step": 23372 }, { "epoch": 2.35, "learning_rate": 2.133741634617018e-05, "loss": 0.3608, "step": 23373 }, { "epoch": 2.35, "learning_rate": 2.1335125144949404e-05, "loss": 0.195, "step": 23374 }, { "epoch": 2.35, "learning_rate": 2.1332833998868608e-05, "loss": 0.6245, "step": 23375 }, { "epoch": 2.35, "learning_rate": 2.133054290794235e-05, "loss": 0.221, "step": 23376 }, { "epoch": 2.35, "learning_rate": 2.1328251872185218e-05, "loss": 0.1561, "step": 23377 }, { "epoch": 2.35, "learning_rate": 2.1325960891611803e-05, "loss": 0.3589, "step": 23378 }, { "epoch": 2.35, "learning_rate": 2.1323669966236672e-05, "loss": 0.4624, "step": 23379 }, { "epoch": 2.35, "learning_rate": 2.1321379096074388e-05, "loss": 0.4131, "step": 23380 }, { "epoch": 2.35, "learning_rate": 2.1319088281139564e-05, "loss": 0.2947, "step": 23381 }, { "epoch": 2.35, "learning_rate": 2.131679752144675e-05, "loss": 0.1464, "step": 23382 }, { "epoch": 2.35, "learning_rate": 2.1314506817010532e-05, "loss": 0.2131, "step": 23383 }, { "epoch": 2.35, "learning_rate": 2.1312216167845497e-05, "loss": 0.505, "step": 23384 }, { "epoch": 2.35, "learning_rate": 2.130992557396621e-05, "loss": 0.1466, "step": 23385 }, { "epoch": 2.35, "learning_rate": 2.1307635035387246e-05, "loss": 0.3593, "step": 23386 }, { "epoch": 2.35, "learning_rate": 2.130534455212319e-05, "loss": 0.5969, "step": 23387 }, { "epoch": 2.35, "learning_rate": 2.130305412418862e-05, "loss": 0.127, "step": 23388 }, { "epoch": 2.35, "learning_rate": 2.130076375159809e-05, "loss": 0.1377, "step": 23389 }, { "epoch": 2.35, "learning_rate": 2.1298473434366202e-05, "loss": 0.4448, "step": 23390 }, { "epoch": 2.35, "learning_rate": 2.1296183172507514e-05, "loss": 0.1854, "step": 23391 }, { "epoch": 2.35, "learning_rate": 2.12938929660366e-05, "loss": 0.1976, "step": 23392 }, { "epoch": 2.35, "learning_rate": 2.1291602814968058e-05, "loss": 0.2201, "step": 23393 }, { "epoch": 2.35, "learning_rate": 2.128931271931643e-05, "loss": 0.1778, "step": 23394 }, { "epoch": 2.35, "learning_rate": 2.1287022679096294e-05, "loss": 0.3797, "step": 23395 }, { "epoch": 2.35, "learning_rate": 2.1284732694322242e-05, "loss": 0.2759, "step": 23396 }, { "epoch": 2.35, "learning_rate": 2.1282442765008838e-05, "loss": 0.7661, "step": 23397 }, { "epoch": 2.35, "learning_rate": 2.1280152891170647e-05, "loss": 0.5599, "step": 23398 }, { "epoch": 2.35, "learning_rate": 2.1277863072822245e-05, "loss": 0.2618, "step": 23399 }, { "epoch": 2.35, "learning_rate": 2.12755733099782e-05, "loss": 0.2056, "step": 23400 }, { "epoch": 2.35, "learning_rate": 2.1273283602653092e-05, "loss": 0.0972, "step": 23401 }, { "epoch": 2.35, "learning_rate": 2.1270993950861495e-05, "loss": 0.1301, "step": 23402 }, { "epoch": 2.35, "learning_rate": 2.126870435461796e-05, "loss": 0.1209, "step": 23403 }, { "epoch": 2.35, "learning_rate": 2.1266414813937075e-05, "loss": 0.2239, "step": 23404 }, { "epoch": 2.35, "learning_rate": 2.12641253288334e-05, "loss": 0.2872, "step": 23405 }, { "epoch": 2.35, "learning_rate": 2.1261835899321507e-05, "loss": 0.2525, "step": 23406 }, { "epoch": 2.35, "learning_rate": 2.1259546525415982e-05, "loss": 0.3211, "step": 23407 }, { "epoch": 2.35, "learning_rate": 2.1257257207131363e-05, "loss": 0.6796, "step": 23408 }, { "epoch": 2.35, "learning_rate": 2.125496794448223e-05, "loss": 0.575, "step": 23409 }, { "epoch": 2.35, "learning_rate": 2.125267873748316e-05, "loss": 0.4227, "step": 23410 }, { "epoch": 2.35, "learning_rate": 2.1250389586148726e-05, "loss": 0.335, "step": 23411 }, { "epoch": 2.35, "learning_rate": 2.1248100490493463e-05, "loss": 0.1887, "step": 23412 }, { "epoch": 2.35, "learning_rate": 2.1245811450531972e-05, "loss": 0.7309, "step": 23413 }, { "epoch": 2.35, "learning_rate": 2.1243522466278808e-05, "loss": 0.2459, "step": 23414 }, { "epoch": 2.35, "learning_rate": 2.1241233537748526e-05, "loss": 0.1228, "step": 23415 }, { "epoch": 2.35, "learning_rate": 2.1238944664955712e-05, "loss": 0.4606, "step": 23416 }, { "epoch": 2.35, "learning_rate": 2.1236655847914915e-05, "loss": 0.7477, "step": 23417 }, { "epoch": 2.35, "learning_rate": 2.1234367086640704e-05, "loss": 0.1776, "step": 23418 }, { "epoch": 2.35, "learning_rate": 2.123207838114765e-05, "loss": 0.1663, "step": 23419 }, { "epoch": 2.35, "learning_rate": 2.1229789731450326e-05, "loss": 0.4315, "step": 23420 }, { "epoch": 2.35, "learning_rate": 2.122750113756327e-05, "loss": 0.2255, "step": 23421 }, { "epoch": 2.35, "learning_rate": 2.1225212599501057e-05, "loss": 0.6103, "step": 23422 }, { "epoch": 2.35, "learning_rate": 2.1222924117278263e-05, "loss": 0.1725, "step": 23423 }, { "epoch": 2.35, "learning_rate": 2.1220635690909427e-05, "loss": 0.71, "step": 23424 }, { "epoch": 2.35, "learning_rate": 2.121834732040914e-05, "loss": 0.0761, "step": 23425 }, { "epoch": 2.35, "learning_rate": 2.121605900579194e-05, "loss": 0.0871, "step": 23426 }, { "epoch": 2.35, "learning_rate": 2.1213770747072402e-05, "loss": 0.3368, "step": 23427 }, { "epoch": 2.35, "learning_rate": 2.121148254426509e-05, "loss": 0.4283, "step": 23428 }, { "epoch": 2.35, "learning_rate": 2.1209194397384545e-05, "loss": 0.4929, "step": 23429 }, { "epoch": 2.35, "learning_rate": 2.120690630644536e-05, "loss": 0.6017, "step": 23430 }, { "epoch": 2.35, "learning_rate": 2.1204618271462075e-05, "loss": 0.0813, "step": 23431 }, { "epoch": 2.35, "learning_rate": 2.1202330292449242e-05, "loss": 0.198, "step": 23432 }, { "epoch": 2.35, "learning_rate": 2.120004236942144e-05, "loss": 0.128, "step": 23433 }, { "epoch": 2.35, "learning_rate": 2.119775450239323e-05, "loss": 0.4705, "step": 23434 }, { "epoch": 2.35, "learning_rate": 2.1195466691379145e-05, "loss": 0.3539, "step": 23435 }, { "epoch": 2.35, "learning_rate": 2.1193178936393767e-05, "loss": 0.1665, "step": 23436 }, { "epoch": 2.35, "learning_rate": 2.119089123745165e-05, "loss": 0.1303, "step": 23437 }, { "epoch": 2.35, "learning_rate": 2.1188603594567348e-05, "loss": 0.2021, "step": 23438 }, { "epoch": 2.35, "learning_rate": 2.1186316007755425e-05, "loss": 0.3023, "step": 23439 }, { "epoch": 2.35, "learning_rate": 2.1184028477030432e-05, "loss": 0.1833, "step": 23440 }, { "epoch": 2.35, "learning_rate": 2.118174100240692e-05, "loss": 0.1405, "step": 23441 }, { "epoch": 2.35, "learning_rate": 2.117945358389946e-05, "loss": 0.4583, "step": 23442 }, { "epoch": 2.35, "learning_rate": 2.1177166221522614e-05, "loss": 0.1086, "step": 23443 }, { "epoch": 2.35, "learning_rate": 2.1174878915290903e-05, "loss": 0.7688, "step": 23444 }, { "epoch": 2.36, "learning_rate": 2.1172591665218917e-05, "loss": 0.3048, "step": 23445 }, { "epoch": 2.36, "learning_rate": 2.11703044713212e-05, "loss": 0.7158, "step": 23446 }, { "epoch": 2.36, "learning_rate": 2.11680173336123e-05, "loss": 0.3459, "step": 23447 }, { "epoch": 2.36, "learning_rate": 2.1165730252106792e-05, "loss": 0.4914, "step": 23448 }, { "epoch": 2.36, "learning_rate": 2.1163443226819207e-05, "loss": 0.317, "step": 23449 }, { "epoch": 2.36, "learning_rate": 2.1161156257764102e-05, "loss": 0.1356, "step": 23450 }, { "epoch": 2.36, "learning_rate": 2.115886934495604e-05, "loss": 0.5006, "step": 23451 }, { "epoch": 2.36, "learning_rate": 2.1156582488409583e-05, "loss": 0.5633, "step": 23452 }, { "epoch": 2.36, "learning_rate": 2.1154295688139248e-05, "loss": 0.2888, "step": 23453 }, { "epoch": 2.36, "learning_rate": 2.1152008944159625e-05, "loss": 0.576, "step": 23454 }, { "epoch": 2.36, "learning_rate": 2.114972225648524e-05, "loss": 0.3407, "step": 23455 }, { "epoch": 2.36, "learning_rate": 2.1147435625130663e-05, "loss": 0.3681, "step": 23456 }, { "epoch": 2.36, "learning_rate": 2.114514905011045e-05, "loss": 0.2575, "step": 23457 }, { "epoch": 2.36, "learning_rate": 2.1142862531439123e-05, "loss": 0.87, "step": 23458 }, { "epoch": 2.36, "learning_rate": 2.1140576069131255e-05, "loss": 0.802, "step": 23459 }, { "epoch": 2.36, "learning_rate": 2.113828966320139e-05, "loss": 0.138, "step": 23460 }, { "epoch": 2.36, "learning_rate": 2.1136003313664073e-05, "loss": 0.2106, "step": 23461 }, { "epoch": 2.36, "learning_rate": 2.113371702053387e-05, "loss": 0.2812, "step": 23462 }, { "epoch": 2.36, "learning_rate": 2.1131430783825317e-05, "loss": 0.1656, "step": 23463 }, { "epoch": 2.36, "learning_rate": 2.112914460355296e-05, "loss": 0.1815, "step": 23464 }, { "epoch": 2.36, "learning_rate": 2.112685847973135e-05, "loss": 0.1562, "step": 23465 }, { "epoch": 2.36, "learning_rate": 2.112457241237505e-05, "loss": 0.6533, "step": 23466 }, { "epoch": 2.36, "learning_rate": 2.1122286401498578e-05, "loss": 0.4882, "step": 23467 }, { "epoch": 2.36, "learning_rate": 2.112000044711651e-05, "loss": 0.0804, "step": 23468 }, { "epoch": 2.36, "learning_rate": 2.1117714549243374e-05, "loss": 0.2161, "step": 23469 }, { "epoch": 2.36, "learning_rate": 2.1115428707893723e-05, "loss": 0.4312, "step": 23470 }, { "epoch": 2.36, "learning_rate": 2.1113142923082114e-05, "loss": 0.2949, "step": 23471 }, { "epoch": 2.36, "learning_rate": 2.1110857194823077e-05, "loss": 0.2003, "step": 23472 }, { "epoch": 2.36, "learning_rate": 2.1108571523131154e-05, "loss": 0.1081, "step": 23473 }, { "epoch": 2.36, "learning_rate": 2.110628590802091e-05, "loss": 0.1933, "step": 23474 }, { "epoch": 2.36, "learning_rate": 2.1104000349506885e-05, "loss": 0.1568, "step": 23475 }, { "epoch": 2.36, "learning_rate": 2.11017148476036e-05, "loss": 1.1114, "step": 23476 }, { "epoch": 2.36, "learning_rate": 2.1099429402325626e-05, "loss": 0.5314, "step": 23477 }, { "epoch": 2.36, "learning_rate": 2.1097144013687495e-05, "loss": 0.5719, "step": 23478 }, { "epoch": 2.36, "learning_rate": 2.1094858681703753e-05, "loss": 0.0971, "step": 23479 }, { "epoch": 2.36, "learning_rate": 2.1092573406388954e-05, "loss": 0.7188, "step": 23480 }, { "epoch": 2.36, "learning_rate": 2.1090288187757607e-05, "loss": 0.2651, "step": 23481 }, { "epoch": 2.36, "learning_rate": 2.1088003025824293e-05, "loss": 0.2495, "step": 23482 }, { "epoch": 2.36, "learning_rate": 2.1085717920603536e-05, "loss": 0.2017, "step": 23483 }, { "epoch": 2.36, "learning_rate": 2.108343287210987e-05, "loss": 0.4544, "step": 23484 }, { "epoch": 2.36, "learning_rate": 2.108114788035786e-05, "loss": 0.5423, "step": 23485 }, { "epoch": 2.36, "learning_rate": 2.1078862945362026e-05, "loss": 0.1579, "step": 23486 }, { "epoch": 2.36, "learning_rate": 2.1076578067136903e-05, "loss": 0.3431, "step": 23487 }, { "epoch": 2.36, "learning_rate": 2.1074293245697055e-05, "loss": 0.1054, "step": 23488 }, { "epoch": 2.36, "learning_rate": 2.1072008481057014e-05, "loss": 0.2042, "step": 23489 }, { "epoch": 2.36, "learning_rate": 2.10697237732313e-05, "loss": 0.095, "step": 23490 }, { "epoch": 2.36, "learning_rate": 2.1067439122234472e-05, "loss": 0.4012, "step": 23491 }, { "epoch": 2.36, "learning_rate": 2.1065154528081065e-05, "loss": 0.5595, "step": 23492 }, { "epoch": 2.36, "learning_rate": 2.1062869990785612e-05, "loss": 0.3466, "step": 23493 }, { "epoch": 2.36, "learning_rate": 2.1060585510362665e-05, "loss": 0.4707, "step": 23494 }, { "epoch": 2.36, "learning_rate": 2.1058301086826743e-05, "loss": 0.5034, "step": 23495 }, { "epoch": 2.36, "learning_rate": 2.1056016720192386e-05, "loss": 0.4045, "step": 23496 }, { "epoch": 2.36, "learning_rate": 2.1053732410474144e-05, "loss": 0.5791, "step": 23497 }, { "epoch": 2.36, "learning_rate": 2.105144815768655e-05, "loss": 0.2217, "step": 23498 }, { "epoch": 2.36, "learning_rate": 2.1049163961844127e-05, "loss": 0.1601, "step": 23499 }, { "epoch": 2.36, "learning_rate": 2.104687982296142e-05, "loss": 0.351, "step": 23500 }, { "epoch": 2.36, "learning_rate": 2.104459574105297e-05, "loss": 0.4079, "step": 23501 }, { "epoch": 2.36, "learning_rate": 2.1042311716133292e-05, "loss": 0.0895, "step": 23502 }, { "epoch": 2.36, "learning_rate": 2.104002774821695e-05, "loss": 0.6337, "step": 23503 }, { "epoch": 2.36, "learning_rate": 2.1037743837318463e-05, "loss": 0.2147, "step": 23504 }, { "epoch": 2.36, "learning_rate": 2.103545998345235e-05, "loss": 0.2274, "step": 23505 }, { "epoch": 2.36, "learning_rate": 2.103317618663317e-05, "loss": 0.1929, "step": 23506 }, { "epoch": 2.36, "learning_rate": 2.103089244687544e-05, "loss": 0.1565, "step": 23507 }, { "epoch": 2.36, "learning_rate": 2.102860876419371e-05, "loss": 0.1085, "step": 23508 }, { "epoch": 2.36, "learning_rate": 2.1026325138602492e-05, "loss": 0.3449, "step": 23509 }, { "epoch": 2.36, "learning_rate": 2.1024041570116324e-05, "loss": 0.1328, "step": 23510 }, { "epoch": 2.36, "learning_rate": 2.1021758058749745e-05, "loss": 0.4668, "step": 23511 }, { "epoch": 2.36, "learning_rate": 2.101947460451729e-05, "loss": 0.2062, "step": 23512 }, { "epoch": 2.36, "learning_rate": 2.1017191207433464e-05, "loss": 0.0466, "step": 23513 }, { "epoch": 2.36, "learning_rate": 2.1014907867512832e-05, "loss": 0.1514, "step": 23514 }, { "epoch": 2.36, "learning_rate": 2.1012624584769905e-05, "loss": 0.2191, "step": 23515 }, { "epoch": 2.36, "learning_rate": 2.101034135921921e-05, "loss": 1.0465, "step": 23516 }, { "epoch": 2.36, "learning_rate": 2.100805819087529e-05, "loss": 0.4028, "step": 23517 }, { "epoch": 2.36, "learning_rate": 2.1005775079752665e-05, "loss": 0.3364, "step": 23518 }, { "epoch": 2.36, "learning_rate": 2.1003492025865863e-05, "loss": 0.2091, "step": 23519 }, { "epoch": 2.36, "learning_rate": 2.1001209029229416e-05, "loss": 0.2121, "step": 23520 }, { "epoch": 2.36, "learning_rate": 2.099892608985786e-05, "loss": 0.9704, "step": 23521 }, { "epoch": 2.36, "learning_rate": 2.09966432077657e-05, "loss": 0.3684, "step": 23522 }, { "epoch": 2.36, "learning_rate": 2.0994360382967486e-05, "loss": 0.1971, "step": 23523 }, { "epoch": 2.36, "learning_rate": 2.0992077615477733e-05, "loss": 0.3099, "step": 23524 }, { "epoch": 2.36, "learning_rate": 2.098979490531097e-05, "loss": 0.2885, "step": 23525 }, { "epoch": 2.36, "learning_rate": 2.0987512252481732e-05, "loss": 0.2664, "step": 23526 }, { "epoch": 2.36, "learning_rate": 2.0985229657004534e-05, "loss": 0.3346, "step": 23527 }, { "epoch": 2.36, "learning_rate": 2.09829471188939e-05, "loss": 0.664, "step": 23528 }, { "epoch": 2.36, "learning_rate": 2.0980664638164362e-05, "loss": 0.7128, "step": 23529 }, { "epoch": 2.36, "learning_rate": 2.0978382214830455e-05, "loss": 0.1291, "step": 23530 }, { "epoch": 2.36, "learning_rate": 2.0976099848906677e-05, "loss": 0.3689, "step": 23531 }, { "epoch": 2.36, "learning_rate": 2.097381754040757e-05, "loss": 0.2246, "step": 23532 }, { "epoch": 2.36, "learning_rate": 2.0971535289347657e-05, "loss": 0.2681, "step": 23533 }, { "epoch": 2.36, "learning_rate": 2.0969253095741453e-05, "loss": 0.8257, "step": 23534 }, { "epoch": 2.36, "learning_rate": 2.09669709596035e-05, "loss": 0.3892, "step": 23535 }, { "epoch": 2.36, "learning_rate": 2.096468888094829e-05, "loss": 0.1303, "step": 23536 }, { "epoch": 2.36, "learning_rate": 2.0962406859790376e-05, "loss": 0.3394, "step": 23537 }, { "epoch": 2.36, "learning_rate": 2.0960124896144265e-05, "loss": 0.5283, "step": 23538 }, { "epoch": 2.36, "learning_rate": 2.0957842990024468e-05, "loss": 0.4652, "step": 23539 }, { "epoch": 2.36, "learning_rate": 2.0955561141445537e-05, "loss": 0.1247, "step": 23540 }, { "epoch": 2.36, "learning_rate": 2.0953279350421966e-05, "loss": 0.4408, "step": 23541 }, { "epoch": 2.36, "learning_rate": 2.095099761696828e-05, "loss": 0.3313, "step": 23542 }, { "epoch": 2.37, "learning_rate": 2.0948715941099005e-05, "loss": 0.4389, "step": 23543 }, { "epoch": 2.37, "learning_rate": 2.0946434322828667e-05, "loss": 0.3241, "step": 23544 }, { "epoch": 2.37, "learning_rate": 2.094415276217176e-05, "loss": 0.2418, "step": 23545 }, { "epoch": 2.37, "learning_rate": 2.094187125914283e-05, "loss": 0.3329, "step": 23546 }, { "epoch": 2.37, "learning_rate": 2.093958981375639e-05, "loss": 0.1724, "step": 23547 }, { "epoch": 2.37, "learning_rate": 2.0937308426026943e-05, "loss": 0.3228, "step": 23548 }, { "epoch": 2.37, "learning_rate": 2.093502709596903e-05, "loss": 0.5605, "step": 23549 }, { "epoch": 2.37, "learning_rate": 2.0932745823597148e-05, "loss": 0.251, "step": 23550 }, { "epoch": 2.37, "learning_rate": 2.0930464608925816e-05, "loss": 0.1986, "step": 23551 }, { "epoch": 2.37, "learning_rate": 2.092818345196956e-05, "loss": 0.3235, "step": 23552 }, { "epoch": 2.37, "learning_rate": 2.0925902352742908e-05, "loss": 0.1476, "step": 23553 }, { "epoch": 2.37, "learning_rate": 2.0923621311260347e-05, "loss": 0.0918, "step": 23554 }, { "epoch": 2.37, "learning_rate": 2.0921340327536414e-05, "loss": 0.2905, "step": 23555 }, { "epoch": 2.37, "learning_rate": 2.091905940158561e-05, "loss": 0.284, "step": 23556 }, { "epoch": 2.37, "learning_rate": 2.0916778533422457e-05, "loss": 0.65, "step": 23557 }, { "epoch": 2.37, "learning_rate": 2.0914497723061486e-05, "loss": 0.5814, "step": 23558 }, { "epoch": 2.37, "learning_rate": 2.091221697051718e-05, "loss": 0.4825, "step": 23559 }, { "epoch": 2.37, "learning_rate": 2.0909936275804068e-05, "loss": 0.1467, "step": 23560 }, { "epoch": 2.37, "learning_rate": 2.090765563893667e-05, "loss": 0.0782, "step": 23561 }, { "epoch": 2.37, "learning_rate": 2.0905375059929485e-05, "loss": 0.2419, "step": 23562 }, { "epoch": 2.37, "learning_rate": 2.0903094538797045e-05, "loss": 0.4692, "step": 23563 }, { "epoch": 2.37, "learning_rate": 2.0900814075553845e-05, "loss": 0.4344, "step": 23564 }, { "epoch": 2.37, "learning_rate": 2.0898533670214396e-05, "loss": 0.1979, "step": 23565 }, { "epoch": 2.37, "learning_rate": 2.0896253322793224e-05, "loss": 0.7412, "step": 23566 }, { "epoch": 2.37, "learning_rate": 2.0893973033304834e-05, "loss": 0.4099, "step": 23567 }, { "epoch": 2.37, "learning_rate": 2.0891692801763722e-05, "loss": 0.5054, "step": 23568 }, { "epoch": 2.37, "learning_rate": 2.088941262818443e-05, "loss": 0.3694, "step": 23569 }, { "epoch": 2.37, "learning_rate": 2.0887132512581445e-05, "loss": 0.0846, "step": 23570 }, { "epoch": 2.37, "learning_rate": 2.0884852454969273e-05, "loss": 0.1141, "step": 23571 }, { "epoch": 2.37, "learning_rate": 2.0882572455362445e-05, "loss": 0.1596, "step": 23572 }, { "epoch": 2.37, "learning_rate": 2.0880292513775453e-05, "loss": 0.3347, "step": 23573 }, { "epoch": 2.37, "learning_rate": 2.0878012630222805e-05, "loss": 0.6761, "step": 23574 }, { "epoch": 2.37, "learning_rate": 2.0875732804719018e-05, "loss": 0.7032, "step": 23575 }, { "epoch": 2.37, "learning_rate": 2.0873453037278602e-05, "loss": 0.6153, "step": 23576 }, { "epoch": 2.37, "learning_rate": 2.087117332791604e-05, "loss": 0.2237, "step": 23577 }, { "epoch": 2.37, "learning_rate": 2.086889367664588e-05, "loss": 0.4648, "step": 23578 }, { "epoch": 2.37, "learning_rate": 2.0866614083482598e-05, "loss": 0.1472, "step": 23579 }, { "epoch": 2.37, "learning_rate": 2.0864334548440703e-05, "loss": 0.6474, "step": 23580 }, { "epoch": 2.37, "learning_rate": 2.0862055071534722e-05, "loss": 0.2483, "step": 23581 }, { "epoch": 2.37, "learning_rate": 2.085977565277914e-05, "loss": 0.3928, "step": 23582 }, { "epoch": 2.37, "learning_rate": 2.0857496292188464e-05, "loss": 0.4261, "step": 23583 }, { "epoch": 2.37, "learning_rate": 2.085521698977721e-05, "loss": 0.5248, "step": 23584 }, { "epoch": 2.37, "learning_rate": 2.085293774555988e-05, "loss": 0.3156, "step": 23585 }, { "epoch": 2.37, "learning_rate": 2.0850658559550963e-05, "loss": 0.4899, "step": 23586 }, { "epoch": 2.37, "learning_rate": 2.0848379431764976e-05, "loss": 0.3498, "step": 23587 }, { "epoch": 2.37, "learning_rate": 2.084610036221642e-05, "loss": 0.7252, "step": 23588 }, { "epoch": 2.37, "learning_rate": 2.08438213509198e-05, "loss": 0.8664, "step": 23589 }, { "epoch": 2.37, "learning_rate": 2.084154239788963e-05, "loss": 0.1511, "step": 23590 }, { "epoch": 2.37, "learning_rate": 2.0839263503140377e-05, "loss": 0.5347, "step": 23591 }, { "epoch": 2.37, "learning_rate": 2.083698466668659e-05, "loss": 0.1628, "step": 23592 }, { "epoch": 2.37, "learning_rate": 2.0834705888542735e-05, "loss": 0.5629, "step": 23593 }, { "epoch": 2.37, "learning_rate": 2.0832427168723323e-05, "loss": 0.2176, "step": 23594 }, { "epoch": 2.37, "learning_rate": 2.083014850724287e-05, "loss": 0.772, "step": 23595 }, { "epoch": 2.37, "learning_rate": 2.0827869904115852e-05, "loss": 0.4963, "step": 23596 }, { "epoch": 2.37, "learning_rate": 2.0825591359356775e-05, "loss": 0.1771, "step": 23597 }, { "epoch": 2.37, "learning_rate": 2.0823312872980155e-05, "loss": 0.1455, "step": 23598 }, { "epoch": 2.37, "learning_rate": 2.0821034445000484e-05, "loss": 0.1735, "step": 23599 }, { "epoch": 2.37, "learning_rate": 2.0818756075432242e-05, "loss": 0.2495, "step": 23600 }, { "epoch": 2.37, "learning_rate": 2.081647776428996e-05, "loss": 0.3192, "step": 23601 }, { "epoch": 2.37, "learning_rate": 2.0814199511588115e-05, "loss": 0.1876, "step": 23602 }, { "epoch": 2.37, "learning_rate": 2.0811921317341206e-05, "loss": 0.2753, "step": 23603 }, { "epoch": 2.37, "learning_rate": 2.080964318156374e-05, "loss": 0.2217, "step": 23604 }, { "epoch": 2.37, "learning_rate": 2.0807365104270207e-05, "loss": 0.4187, "step": 23605 }, { "epoch": 2.37, "learning_rate": 2.0805087085475097e-05, "loss": 0.6932, "step": 23606 }, { "epoch": 2.37, "learning_rate": 2.0802809125192925e-05, "loss": 0.346, "step": 23607 }, { "epoch": 2.37, "learning_rate": 2.080053122343818e-05, "loss": 0.0724, "step": 23608 }, { "epoch": 2.37, "learning_rate": 2.0798253380225336e-05, "loss": 0.369, "step": 23609 }, { "epoch": 2.37, "learning_rate": 2.0795975595568927e-05, "loss": 0.1928, "step": 23610 }, { "epoch": 2.37, "learning_rate": 2.079369786948342e-05, "loss": 0.3179, "step": 23611 }, { "epoch": 2.37, "learning_rate": 2.079142020198331e-05, "loss": 0.8341, "step": 23612 }, { "epoch": 2.37, "learning_rate": 2.0789142593083117e-05, "loss": 0.2921, "step": 23613 }, { "epoch": 2.37, "learning_rate": 2.0786865042797305e-05, "loss": 0.0636, "step": 23614 }, { "epoch": 2.37, "learning_rate": 2.0784587551140377e-05, "loss": 0.1157, "step": 23615 }, { "epoch": 2.37, "learning_rate": 2.078231011812683e-05, "loss": 0.3804, "step": 23616 }, { "epoch": 2.37, "learning_rate": 2.0780032743771152e-05, "loss": 0.6518, "step": 23617 }, { "epoch": 2.37, "learning_rate": 2.077775542808785e-05, "loss": 0.1149, "step": 23618 }, { "epoch": 2.37, "learning_rate": 2.0775478171091393e-05, "loss": 0.5197, "step": 23619 }, { "epoch": 2.37, "learning_rate": 2.0773200972796282e-05, "loss": 0.1616, "step": 23620 }, { "epoch": 2.37, "learning_rate": 2.0770923833217017e-05, "loss": 0.1025, "step": 23621 }, { "epoch": 2.37, "learning_rate": 2.0768646752368086e-05, "loss": 0.607, "step": 23622 }, { "epoch": 2.37, "learning_rate": 2.0766369730263957e-05, "loss": 0.1522, "step": 23623 }, { "epoch": 2.37, "learning_rate": 2.076409276691916e-05, "loss": 0.1299, "step": 23624 }, { "epoch": 2.37, "learning_rate": 2.0761815862348153e-05, "loss": 0.4935, "step": 23625 }, { "epoch": 2.37, "learning_rate": 2.075953901656543e-05, "loss": 0.2312, "step": 23626 }, { "epoch": 2.37, "learning_rate": 2.07572622295855e-05, "loss": 0.3609, "step": 23627 }, { "epoch": 2.37, "learning_rate": 2.0754985501422826e-05, "loss": 0.8482, "step": 23628 }, { "epoch": 2.37, "learning_rate": 2.075270883209191e-05, "loss": 0.3769, "step": 23629 }, { "epoch": 2.37, "learning_rate": 2.0750432221607236e-05, "loss": 0.3224, "step": 23630 }, { "epoch": 2.37, "learning_rate": 2.0748155669983298e-05, "loss": 0.424, "step": 23631 }, { "epoch": 2.37, "learning_rate": 2.0745879177234563e-05, "loss": 0.3666, "step": 23632 }, { "epoch": 2.37, "learning_rate": 2.074360274337555e-05, "loss": 0.2165, "step": 23633 }, { "epoch": 2.37, "learning_rate": 2.074132636842072e-05, "loss": 0.3874, "step": 23634 }, { "epoch": 2.37, "learning_rate": 2.0739050052384562e-05, "loss": 0.2806, "step": 23635 }, { "epoch": 2.37, "learning_rate": 2.073677379528158e-05, "loss": 0.3257, "step": 23636 }, { "epoch": 2.37, "learning_rate": 2.073449759712624e-05, "loss": 0.5358, "step": 23637 }, { "epoch": 2.37, "learning_rate": 2.073222145793302e-05, "loss": 0.553, "step": 23638 }, { "epoch": 2.37, "learning_rate": 2.072994537771643e-05, "loss": 0.5328, "step": 23639 }, { "epoch": 2.37, "learning_rate": 2.0727669356490943e-05, "loss": 0.339, "step": 23640 }, { "epoch": 2.38, "learning_rate": 2.072539339427103e-05, "loss": 0.2117, "step": 23641 }, { "epoch": 2.38, "learning_rate": 2.0723117491071195e-05, "loss": 0.1638, "step": 23642 }, { "epoch": 2.38, "learning_rate": 2.0720841646905898e-05, "loss": 0.473, "step": 23643 }, { "epoch": 2.38, "learning_rate": 2.071856586178965e-05, "loss": 0.3316, "step": 23644 }, { "epoch": 2.38, "learning_rate": 2.071629013573692e-05, "loss": 0.4794, "step": 23645 }, { "epoch": 2.38, "learning_rate": 2.0714014468762165e-05, "loss": 0.5795, "step": 23646 }, { "epoch": 2.38, "learning_rate": 2.071173886087991e-05, "loss": 0.1751, "step": 23647 }, { "epoch": 2.38, "learning_rate": 2.070946331210461e-05, "loss": 0.4557, "step": 23648 }, { "epoch": 2.38, "learning_rate": 2.070718782245075e-05, "loss": 0.2768, "step": 23649 }, { "epoch": 2.38, "learning_rate": 2.070491239193282e-05, "loss": 0.2039, "step": 23650 }, { "epoch": 2.38, "learning_rate": 2.0702637020565284e-05, "loss": 0.3382, "step": 23651 }, { "epoch": 2.38, "learning_rate": 2.070036170836262e-05, "loss": 0.2892, "step": 23652 }, { "epoch": 2.38, "learning_rate": 2.0698086455339328e-05, "loss": 0.1026, "step": 23653 }, { "epoch": 2.38, "learning_rate": 2.069581126150988e-05, "loss": 0.1527, "step": 23654 }, { "epoch": 2.38, "learning_rate": 2.069353612688873e-05, "loss": 0.3286, "step": 23655 }, { "epoch": 2.38, "learning_rate": 2.0691261051490394e-05, "loss": 0.1485, "step": 23656 }, { "epoch": 2.38, "learning_rate": 2.0688986035329327e-05, "loss": 0.3667, "step": 23657 }, { "epoch": 2.38, "learning_rate": 2.0686711078420004e-05, "loss": 0.133, "step": 23658 }, { "epoch": 2.38, "learning_rate": 2.0684436180776922e-05, "loss": 0.3449, "step": 23659 }, { "epoch": 2.38, "learning_rate": 2.0682161342414536e-05, "loss": 0.064, "step": 23660 }, { "epoch": 2.38, "learning_rate": 2.0679886563347322e-05, "loss": 0.834, "step": 23661 }, { "epoch": 2.38, "learning_rate": 2.0677611843589775e-05, "loss": 0.8248, "step": 23662 }, { "epoch": 2.38, "learning_rate": 2.0675337183156363e-05, "loss": 0.4027, "step": 23663 }, { "epoch": 2.38, "learning_rate": 2.0673062582061546e-05, "loss": 0.2558, "step": 23664 }, { "epoch": 2.38, "learning_rate": 2.067078804031982e-05, "loss": 0.1017, "step": 23665 }, { "epoch": 2.38, "learning_rate": 2.0668513557945653e-05, "loss": 0.4217, "step": 23666 }, { "epoch": 2.38, "learning_rate": 2.0666239134953503e-05, "loss": 0.3101, "step": 23667 }, { "epoch": 2.38, "learning_rate": 2.0663964771357874e-05, "loss": 1.4766, "step": 23668 }, { "epoch": 2.38, "learning_rate": 2.0661690467173194e-05, "loss": 0.4594, "step": 23669 }, { "epoch": 2.38, "learning_rate": 2.065941622241399e-05, "loss": 0.0841, "step": 23670 }, { "epoch": 2.38, "learning_rate": 2.0657142037094697e-05, "loss": 0.1994, "step": 23671 }, { "epoch": 2.38, "learning_rate": 2.0654867911229795e-05, "loss": 0.1791, "step": 23672 }, { "epoch": 2.38, "learning_rate": 2.0652593844833765e-05, "loss": 0.7283, "step": 23673 }, { "epoch": 2.38, "learning_rate": 2.0650319837921075e-05, "loss": 0.2435, "step": 23674 }, { "epoch": 2.38, "learning_rate": 2.064804589050618e-05, "loss": 0.7141, "step": 23675 }, { "epoch": 2.38, "learning_rate": 2.0645772002603574e-05, "loss": 0.4155, "step": 23676 }, { "epoch": 2.38, "learning_rate": 2.0643498174227722e-05, "loss": 0.3001, "step": 23677 }, { "epoch": 2.38, "learning_rate": 2.0641224405393066e-05, "loss": 0.4471, "step": 23678 }, { "epoch": 2.38, "learning_rate": 2.063895069611412e-05, "loss": 0.3102, "step": 23679 }, { "epoch": 2.38, "learning_rate": 2.063667704640532e-05, "loss": 0.6887, "step": 23680 }, { "epoch": 2.38, "learning_rate": 2.063440345628114e-05, "loss": 0.2403, "step": 23681 }, { "epoch": 2.38, "learning_rate": 2.063212992575607e-05, "loss": 0.453, "step": 23682 }, { "epoch": 2.38, "learning_rate": 2.062985645484455e-05, "loss": 0.2044, "step": 23683 }, { "epoch": 2.38, "learning_rate": 2.0627583043561057e-05, "loss": 0.5697, "step": 23684 }, { "epoch": 2.38, "learning_rate": 2.062530969192006e-05, "loss": 0.5819, "step": 23685 }, { "epoch": 2.38, "learning_rate": 2.062303639993604e-05, "loss": 0.7935, "step": 23686 }, { "epoch": 2.38, "learning_rate": 2.062076316762343e-05, "loss": 0.6612, "step": 23687 }, { "epoch": 2.38, "learning_rate": 2.0618489994996727e-05, "loss": 0.5204, "step": 23688 }, { "epoch": 2.38, "learning_rate": 2.0616216882070385e-05, "loss": 0.4463, "step": 23689 }, { "epoch": 2.38, "learning_rate": 2.061394382885886e-05, "loss": 0.4959, "step": 23690 }, { "epoch": 2.38, "learning_rate": 2.0611670835376638e-05, "loss": 0.6904, "step": 23691 }, { "epoch": 2.38, "learning_rate": 2.0609397901638168e-05, "loss": 0.2287, "step": 23692 }, { "epoch": 2.38, "learning_rate": 2.0607125027657906e-05, "loss": 0.1407, "step": 23693 }, { "epoch": 2.38, "learning_rate": 2.060485221345034e-05, "loss": 0.3366, "step": 23694 }, { "epoch": 2.38, "learning_rate": 2.060257945902992e-05, "loss": 0.5239, "step": 23695 }, { "epoch": 2.38, "learning_rate": 2.0600306764411093e-05, "loss": 0.2722, "step": 23696 }, { "epoch": 2.38, "learning_rate": 2.0598034129608355e-05, "loss": 0.6917, "step": 23697 }, { "epoch": 2.38, "learning_rate": 2.0595761554636144e-05, "loss": 0.5217, "step": 23698 }, { "epoch": 2.38, "learning_rate": 2.059348903950893e-05, "loss": 0.459, "step": 23699 }, { "epoch": 2.38, "learning_rate": 2.059121658424118e-05, "loss": 0.2095, "step": 23700 }, { "epoch": 2.38, "learning_rate": 2.0588944188847333e-05, "loss": 0.3649, "step": 23701 }, { "epoch": 2.38, "learning_rate": 2.058667185334188e-05, "loss": 0.1057, "step": 23702 }, { "epoch": 2.38, "learning_rate": 2.0584399577739262e-05, "loss": 0.1993, "step": 23703 }, { "epoch": 2.38, "learning_rate": 2.0582127362053938e-05, "loss": 0.2636, "step": 23704 }, { "epoch": 2.38, "learning_rate": 2.0579855206300376e-05, "loss": 0.2902, "step": 23705 }, { "epoch": 2.38, "learning_rate": 2.0577583110493044e-05, "loss": 0.1782, "step": 23706 }, { "epoch": 2.38, "learning_rate": 2.0575311074646367e-05, "loss": 0.9746, "step": 23707 }, { "epoch": 2.38, "learning_rate": 2.0573039098774837e-05, "loss": 0.1392, "step": 23708 }, { "epoch": 2.38, "learning_rate": 2.0570767182892905e-05, "loss": 0.1291, "step": 23709 }, { "epoch": 2.38, "learning_rate": 2.0568495327015007e-05, "loss": 0.1611, "step": 23710 }, { "epoch": 2.38, "learning_rate": 2.056622353115563e-05, "loss": 0.4259, "step": 23711 }, { "epoch": 2.38, "learning_rate": 2.056395179532922e-05, "loss": 0.3441, "step": 23712 }, { "epoch": 2.38, "learning_rate": 2.056168011955022e-05, "loss": 0.6035, "step": 23713 }, { "epoch": 2.38, "learning_rate": 2.0559408503833107e-05, "loss": 0.427, "step": 23714 }, { "epoch": 2.38, "learning_rate": 2.055713694819233e-05, "loss": 0.7625, "step": 23715 }, { "epoch": 2.38, "learning_rate": 2.0554865452642328e-05, "loss": 0.4111, "step": 23716 }, { "epoch": 2.38, "learning_rate": 2.055259401719757e-05, "loss": 0.1563, "step": 23717 }, { "epoch": 2.38, "learning_rate": 2.0550322641872523e-05, "loss": 0.2641, "step": 23718 }, { "epoch": 2.38, "learning_rate": 2.0548051326681607e-05, "loss": 0.3357, "step": 23719 }, { "epoch": 2.38, "learning_rate": 2.054578007163932e-05, "loss": 0.9417, "step": 23720 }, { "epoch": 2.38, "learning_rate": 2.054350887676008e-05, "loss": 0.6, "step": 23721 }, { "epoch": 2.38, "learning_rate": 2.054123774205835e-05, "loss": 0.2345, "step": 23722 }, { "epoch": 2.38, "learning_rate": 2.0538966667548595e-05, "loss": 0.3325, "step": 23723 }, { "epoch": 2.38, "learning_rate": 2.053669565324524e-05, "loss": 0.5709, "step": 23724 }, { "epoch": 2.38, "learning_rate": 2.0534424699162767e-05, "loss": 0.5454, "step": 23725 }, { "epoch": 2.38, "learning_rate": 2.0532153805315614e-05, "loss": 0.5712, "step": 23726 }, { "epoch": 2.38, "learning_rate": 2.0529882971718225e-05, "loss": 0.4508, "step": 23727 }, { "epoch": 2.38, "learning_rate": 2.0527612198385063e-05, "loss": 0.2596, "step": 23728 }, { "epoch": 2.38, "learning_rate": 2.0525341485330577e-05, "loss": 0.4857, "step": 23729 }, { "epoch": 2.38, "learning_rate": 2.0523070832569207e-05, "loss": 0.0857, "step": 23730 }, { "epoch": 2.38, "learning_rate": 2.052080024011541e-05, "loss": 0.6914, "step": 23731 }, { "epoch": 2.38, "learning_rate": 2.0518529707983645e-05, "loss": 0.1198, "step": 23732 }, { "epoch": 2.38, "learning_rate": 2.0516259236188327e-05, "loss": 0.4246, "step": 23733 }, { "epoch": 2.38, "learning_rate": 2.0513988824743954e-05, "loss": 0.0975, "step": 23734 }, { "epoch": 2.38, "learning_rate": 2.0511718473664935e-05, "loss": 0.3903, "step": 23735 }, { "epoch": 2.38, "learning_rate": 2.0509448182965722e-05, "loss": 0.2115, "step": 23736 }, { "epoch": 2.38, "learning_rate": 2.050717795266078e-05, "loss": 0.2593, "step": 23737 }, { "epoch": 2.38, "learning_rate": 2.0504907782764553e-05, "loss": 0.4209, "step": 23738 }, { "epoch": 2.38, "learning_rate": 2.0502637673291464e-05, "loss": 0.1601, "step": 23739 }, { "epoch": 2.39, "learning_rate": 2.0500367624255985e-05, "loss": 0.6688, "step": 23740 }, { "epoch": 2.39, "learning_rate": 2.049809763567256e-05, "loss": 0.423, "step": 23741 }, { "epoch": 2.39, "learning_rate": 2.0495827707555608e-05, "loss": 0.4799, "step": 23742 }, { "epoch": 2.39, "learning_rate": 2.049355783991961e-05, "loss": 0.6927, "step": 23743 }, { "epoch": 2.39, "learning_rate": 2.049128803277899e-05, "loss": 0.2263, "step": 23744 }, { "epoch": 2.39, "learning_rate": 2.0489018286148186e-05, "loss": 0.1961, "step": 23745 }, { "epoch": 2.39, "learning_rate": 2.0486748600041666e-05, "loss": 0.1635, "step": 23746 }, { "epoch": 2.39, "learning_rate": 2.0484478974473852e-05, "loss": 0.7229, "step": 23747 }, { "epoch": 2.39, "learning_rate": 2.0482209409459186e-05, "loss": 0.36, "step": 23748 }, { "epoch": 2.39, "learning_rate": 2.0479939905012126e-05, "loss": 0.2778, "step": 23749 }, { "epoch": 2.39, "learning_rate": 2.0477670461147115e-05, "loss": 0.4822, "step": 23750 }, { "epoch": 2.39, "learning_rate": 2.0475401077878567e-05, "loss": 0.104, "step": 23751 }, { "epoch": 2.39, "learning_rate": 2.047313175522096e-05, "loss": 0.6912, "step": 23752 }, { "epoch": 2.39, "learning_rate": 2.0470862493188708e-05, "loss": 0.4373, "step": 23753 }, { "epoch": 2.39, "learning_rate": 2.0468593291796267e-05, "loss": 0.3824, "step": 23754 }, { "epoch": 2.39, "learning_rate": 2.046632415105808e-05, "loss": 0.2258, "step": 23755 }, { "epoch": 2.39, "learning_rate": 2.0464055070988557e-05, "loss": 0.0885, "step": 23756 }, { "epoch": 2.39, "learning_rate": 2.0461786051602182e-05, "loss": 0.1404, "step": 23757 }, { "epoch": 2.39, "learning_rate": 2.0459517092913368e-05, "loss": 0.1466, "step": 23758 }, { "epoch": 2.39, "learning_rate": 2.0457248194936548e-05, "loss": 0.1424, "step": 23759 }, { "epoch": 2.39, "learning_rate": 2.0454979357686173e-05, "loss": 0.0884, "step": 23760 }, { "epoch": 2.39, "learning_rate": 2.0452710581176694e-05, "loss": 0.6104, "step": 23761 }, { "epoch": 2.39, "learning_rate": 2.0450441865422512e-05, "loss": 0.5505, "step": 23762 }, { "epoch": 2.39, "learning_rate": 2.0448173210438093e-05, "loss": 0.643, "step": 23763 }, { "epoch": 2.39, "learning_rate": 2.0445904616237876e-05, "loss": 0.0798, "step": 23764 }, { "epoch": 2.39, "learning_rate": 2.0443636082836265e-05, "loss": 0.2135, "step": 23765 }, { "epoch": 2.39, "learning_rate": 2.044136761024774e-05, "loss": 0.06, "step": 23766 }, { "epoch": 2.39, "learning_rate": 2.0439099198486707e-05, "loss": 0.0909, "step": 23767 }, { "epoch": 2.39, "learning_rate": 2.0436830847567604e-05, "loss": 0.1879, "step": 23768 }, { "epoch": 2.39, "learning_rate": 2.0434562557504877e-05, "loss": 0.8264, "step": 23769 }, { "epoch": 2.39, "learning_rate": 2.0432294328312963e-05, "loss": 0.0846, "step": 23770 }, { "epoch": 2.39, "learning_rate": 2.0430026160006272e-05, "loss": 0.2981, "step": 23771 }, { "epoch": 2.39, "learning_rate": 2.0427758052599265e-05, "loss": 0.311, "step": 23772 }, { "epoch": 2.39, "learning_rate": 2.0425490006106368e-05, "loss": 0.5508, "step": 23773 }, { "epoch": 2.39, "learning_rate": 2.042322202054199e-05, "loss": 0.1925, "step": 23774 }, { "epoch": 2.39, "learning_rate": 2.0420954095920607e-05, "loss": 0.4392, "step": 23775 }, { "epoch": 2.39, "learning_rate": 2.0418686232256617e-05, "loss": 0.4541, "step": 23776 }, { "epoch": 2.39, "learning_rate": 2.041641842956446e-05, "loss": 0.2287, "step": 23777 }, { "epoch": 2.39, "learning_rate": 2.0414150687858582e-05, "loss": 0.1526, "step": 23778 }, { "epoch": 2.39, "learning_rate": 2.0411883007153383e-05, "loss": 0.4171, "step": 23779 }, { "epoch": 2.39, "learning_rate": 2.0409615387463332e-05, "loss": 0.2022, "step": 23780 }, { "epoch": 2.39, "learning_rate": 2.0407347828802835e-05, "loss": 0.1631, "step": 23781 }, { "epoch": 2.39, "learning_rate": 2.0405080331186322e-05, "loss": 0.8813, "step": 23782 }, { "epoch": 2.39, "learning_rate": 2.040281289462823e-05, "loss": 0.8254, "step": 23783 }, { "epoch": 2.39, "learning_rate": 2.0400545519142992e-05, "loss": 0.5077, "step": 23784 }, { "epoch": 2.39, "learning_rate": 2.039827820474502e-05, "loss": 0.3112, "step": 23785 }, { "epoch": 2.39, "learning_rate": 2.0396010951448762e-05, "loss": 0.3965, "step": 23786 }, { "epoch": 2.39, "learning_rate": 2.039374375926864e-05, "loss": 0.3579, "step": 23787 }, { "epoch": 2.39, "learning_rate": 2.039147662821906e-05, "loss": 0.3622, "step": 23788 }, { "epoch": 2.39, "learning_rate": 2.0389209558314485e-05, "loss": 0.4443, "step": 23789 }, { "epoch": 2.39, "learning_rate": 2.038694254956932e-05, "loss": 0.5677, "step": 23790 }, { "epoch": 2.39, "learning_rate": 2.0384675601997988e-05, "loss": 0.3339, "step": 23791 }, { "epoch": 2.39, "learning_rate": 2.0382408715614927e-05, "loss": 0.0852, "step": 23792 }, { "epoch": 2.39, "learning_rate": 2.038014189043457e-05, "loss": 0.1811, "step": 23793 }, { "epoch": 2.39, "learning_rate": 2.037787512647131e-05, "loss": 0.7044, "step": 23794 }, { "epoch": 2.39, "learning_rate": 2.0375608423739603e-05, "loss": 0.5789, "step": 23795 }, { "epoch": 2.39, "learning_rate": 2.037334178225387e-05, "loss": 0.8467, "step": 23796 }, { "epoch": 2.39, "learning_rate": 2.037107520202851e-05, "loss": 0.3392, "step": 23797 }, { "epoch": 2.39, "learning_rate": 2.036880868307798e-05, "loss": 0.4418, "step": 23798 }, { "epoch": 2.39, "learning_rate": 2.036654222541668e-05, "loss": 0.4052, "step": 23799 }, { "epoch": 2.39, "learning_rate": 2.036427582905904e-05, "loss": 0.4157, "step": 23800 }, { "epoch": 2.39, "learning_rate": 2.0362009494019485e-05, "loss": 0.4039, "step": 23801 }, { "epoch": 2.39, "learning_rate": 2.0359743220312445e-05, "loss": 0.1483, "step": 23802 }, { "epoch": 2.39, "learning_rate": 2.0357477007952315e-05, "loss": 0.2032, "step": 23803 }, { "epoch": 2.39, "learning_rate": 2.035521085695354e-05, "loss": 0.1934, "step": 23804 }, { "epoch": 2.39, "learning_rate": 2.0352944767330528e-05, "loss": 0.1509, "step": 23805 }, { "epoch": 2.39, "learning_rate": 2.0350678739097712e-05, "loss": 0.2662, "step": 23806 }, { "epoch": 2.39, "learning_rate": 2.0348412772269516e-05, "loss": 0.22, "step": 23807 }, { "epoch": 2.39, "learning_rate": 2.0346146866860332e-05, "loss": 0.1968, "step": 23808 }, { "epoch": 2.39, "learning_rate": 2.0343881022884606e-05, "loss": 0.6317, "step": 23809 }, { "epoch": 2.39, "learning_rate": 2.0341615240356752e-05, "loss": 0.2224, "step": 23810 }, { "epoch": 2.39, "learning_rate": 2.0339349519291168e-05, "loss": 0.1897, "step": 23811 }, { "epoch": 2.39, "learning_rate": 2.0337083859702303e-05, "loss": 0.4495, "step": 23812 }, { "epoch": 2.39, "learning_rate": 2.033481826160456e-05, "loss": 0.5538, "step": 23813 }, { "epoch": 2.39, "learning_rate": 2.0332552725012347e-05, "loss": 0.7859, "step": 23814 }, { "epoch": 2.39, "learning_rate": 2.0330287249940095e-05, "loss": 0.4124, "step": 23815 }, { "epoch": 2.39, "learning_rate": 2.0328021836402226e-05, "loss": 0.2219, "step": 23816 }, { "epoch": 2.39, "learning_rate": 2.0325756484413132e-05, "loss": 0.3807, "step": 23817 }, { "epoch": 2.39, "learning_rate": 2.032349119398725e-05, "loss": 0.3833, "step": 23818 }, { "epoch": 2.39, "learning_rate": 2.0321225965138993e-05, "loss": 0.5342, "step": 23819 }, { "epoch": 2.39, "learning_rate": 2.0318960797882763e-05, "loss": 0.4891, "step": 23820 }, { "epoch": 2.39, "learning_rate": 2.031669569223299e-05, "loss": 0.2803, "step": 23821 }, { "epoch": 2.39, "learning_rate": 2.0314430648204084e-05, "loss": 0.8281, "step": 23822 }, { "epoch": 2.39, "learning_rate": 2.0312165665810444e-05, "loss": 0.4256, "step": 23823 }, { "epoch": 2.39, "learning_rate": 2.030990074506651e-05, "loss": 0.5125, "step": 23824 }, { "epoch": 2.39, "learning_rate": 2.0307635885986682e-05, "loss": 0.3494, "step": 23825 }, { "epoch": 2.39, "learning_rate": 2.0305371088585363e-05, "loss": 0.108, "step": 23826 }, { "epoch": 2.39, "learning_rate": 2.0303106352876977e-05, "loss": 0.7759, "step": 23827 }, { "epoch": 2.39, "learning_rate": 2.030084167887594e-05, "loss": 0.1623, "step": 23828 }, { "epoch": 2.39, "learning_rate": 2.0298577066596644e-05, "loss": 0.2888, "step": 23829 }, { "epoch": 2.39, "learning_rate": 2.0296312516053526e-05, "loss": 0.379, "step": 23830 }, { "epoch": 2.39, "learning_rate": 2.029404802726098e-05, "loss": 0.3896, "step": 23831 }, { "epoch": 2.39, "learning_rate": 2.0291783600233413e-05, "loss": 0.3153, "step": 23832 }, { "epoch": 2.39, "learning_rate": 2.0289519234985244e-05, "loss": 0.2066, "step": 23833 }, { "epoch": 2.39, "learning_rate": 2.028725493153088e-05, "loss": 0.2862, "step": 23834 }, { "epoch": 2.39, "learning_rate": 2.028499068988474e-05, "loss": 0.0602, "step": 23835 }, { "epoch": 2.39, "learning_rate": 2.028272651006122e-05, "loss": 0.4328, "step": 23836 }, { "epoch": 2.39, "learning_rate": 2.0280462392074722e-05, "loss": 0.3218, "step": 23837 }, { "epoch": 2.4, "learning_rate": 2.027819833593967e-05, "loss": 0.1414, "step": 23838 }, { "epoch": 2.4, "learning_rate": 2.027593434167048e-05, "loss": 0.1417, "step": 23839 }, { "epoch": 2.4, "learning_rate": 2.0273670409281523e-05, "loss": 0.1322, "step": 23840 }, { "epoch": 2.4, "learning_rate": 2.0271406538787233e-05, "loss": 0.673, "step": 23841 }, { "epoch": 2.4, "learning_rate": 2.0269142730202025e-05, "loss": 0.6939, "step": 23842 }, { "epoch": 2.4, "learning_rate": 2.0266878983540266e-05, "loss": 0.3016, "step": 23843 }, { "epoch": 2.4, "learning_rate": 2.026461529881641e-05, "loss": 0.6741, "step": 23844 }, { "epoch": 2.4, "learning_rate": 2.0262351676044826e-05, "loss": 0.1714, "step": 23845 }, { "epoch": 2.4, "learning_rate": 2.0260088115239933e-05, "loss": 0.1109, "step": 23846 }, { "epoch": 2.4, "learning_rate": 2.0257824616416132e-05, "loss": 0.5595, "step": 23847 }, { "epoch": 2.4, "learning_rate": 2.025556117958784e-05, "loss": 0.7465, "step": 23848 }, { "epoch": 2.4, "learning_rate": 2.0253297804769435e-05, "loss": 0.4622, "step": 23849 }, { "epoch": 2.4, "learning_rate": 2.0251034491975343e-05, "loss": 0.0886, "step": 23850 }, { "epoch": 2.4, "learning_rate": 2.0248771241219963e-05, "loss": 0.5232, "step": 23851 }, { "epoch": 2.4, "learning_rate": 2.0246508052517678e-05, "loss": 0.5996, "step": 23852 }, { "epoch": 2.4, "learning_rate": 2.0244244925882925e-05, "loss": 0.1714, "step": 23853 }, { "epoch": 2.4, "learning_rate": 2.0241981861330076e-05, "loss": 1.2261, "step": 23854 }, { "epoch": 2.4, "learning_rate": 2.0239718858873537e-05, "loss": 0.1329, "step": 23855 }, { "epoch": 2.4, "learning_rate": 2.023745591852772e-05, "loss": 0.5511, "step": 23856 }, { "epoch": 2.4, "learning_rate": 2.0235193040307025e-05, "loss": 0.6081, "step": 23857 }, { "epoch": 2.4, "learning_rate": 2.023293022422584e-05, "loss": 0.1603, "step": 23858 }, { "epoch": 2.4, "learning_rate": 2.023066747029857e-05, "loss": 0.4184, "step": 23859 }, { "epoch": 2.4, "learning_rate": 2.0228404778539612e-05, "loss": 0.0756, "step": 23860 }, { "epoch": 2.4, "learning_rate": 2.022614214896338e-05, "loss": 0.0778, "step": 23861 }, { "epoch": 2.4, "learning_rate": 2.022387958158426e-05, "loss": 0.1587, "step": 23862 }, { "epoch": 2.4, "learning_rate": 2.0221617076416642e-05, "loss": 0.1533, "step": 23863 }, { "epoch": 2.4, "learning_rate": 2.0219354633474937e-05, "loss": 0.493, "step": 23864 }, { "epoch": 2.4, "learning_rate": 2.0217092252773545e-05, "loss": 1.0454, "step": 23865 }, { "epoch": 2.4, "learning_rate": 2.0214829934326846e-05, "loss": 0.2294, "step": 23866 }, { "epoch": 2.4, "learning_rate": 2.021256767814926e-05, "loss": 0.0794, "step": 23867 }, { "epoch": 2.4, "learning_rate": 2.021030548425516e-05, "loss": 0.3171, "step": 23868 }, { "epoch": 2.4, "learning_rate": 2.020804335265895e-05, "loss": 0.2985, "step": 23869 }, { "epoch": 2.4, "learning_rate": 2.020578128337503e-05, "loss": 0.8809, "step": 23870 }, { "epoch": 2.4, "learning_rate": 2.02035192764178e-05, "loss": 0.0574, "step": 23871 }, { "epoch": 2.4, "learning_rate": 2.0201257331801635e-05, "loss": 0.2254, "step": 23872 }, { "epoch": 2.4, "learning_rate": 2.0198995449540945e-05, "loss": 0.3004, "step": 23873 }, { "epoch": 2.4, "learning_rate": 2.0196733629650127e-05, "loss": 0.098, "step": 23874 }, { "epoch": 2.4, "learning_rate": 2.0194471872143546e-05, "loss": 0.3028, "step": 23875 }, { "epoch": 2.4, "learning_rate": 2.0192210177035636e-05, "loss": 0.2202, "step": 23876 }, { "epoch": 2.4, "learning_rate": 2.0189948544340766e-05, "loss": 0.3098, "step": 23877 }, { "epoch": 2.4, "learning_rate": 2.0187686974073322e-05, "loss": 0.4028, "step": 23878 }, { "epoch": 2.4, "learning_rate": 2.0185425466247714e-05, "loss": 0.066, "step": 23879 }, { "epoch": 2.4, "learning_rate": 2.0183164020878327e-05, "loss": 0.1857, "step": 23880 }, { "epoch": 2.4, "learning_rate": 2.0180902637979538e-05, "loss": 0.4453, "step": 23881 }, { "epoch": 2.4, "learning_rate": 2.0178641317565756e-05, "loss": 0.302, "step": 23882 }, { "epoch": 2.4, "learning_rate": 2.0176380059651375e-05, "loss": 0.3084, "step": 23883 }, { "epoch": 2.4, "learning_rate": 2.0174118864250752e-05, "loss": 0.2421, "step": 23884 }, { "epoch": 2.4, "learning_rate": 2.0171857731378315e-05, "loss": 0.427, "step": 23885 }, { "epoch": 2.4, "learning_rate": 2.0169596661048425e-05, "loss": 0.5653, "step": 23886 }, { "epoch": 2.4, "learning_rate": 2.016733565327549e-05, "loss": 0.1562, "step": 23887 }, { "epoch": 2.4, "learning_rate": 2.0165074708073892e-05, "loss": 0.6083, "step": 23888 }, { "epoch": 2.4, "learning_rate": 2.016281382545801e-05, "loss": 0.3076, "step": 23889 }, { "epoch": 2.4, "learning_rate": 2.016055300544225e-05, "loss": 0.1616, "step": 23890 }, { "epoch": 2.4, "learning_rate": 2.015829224804098e-05, "loss": 0.2259, "step": 23891 }, { "epoch": 2.4, "learning_rate": 2.0156031553268592e-05, "loss": 0.23, "step": 23892 }, { "epoch": 2.4, "learning_rate": 2.0153770921139475e-05, "loss": 0.9455, "step": 23893 }, { "epoch": 2.4, "learning_rate": 2.0151510351668024e-05, "loss": 0.69, "step": 23894 }, { "epoch": 2.4, "learning_rate": 2.01492498448686e-05, "loss": 0.7021, "step": 23895 }, { "epoch": 2.4, "learning_rate": 2.014698940075561e-05, "loss": 0.2658, "step": 23896 }, { "epoch": 2.4, "learning_rate": 2.0144729019343434e-05, "loss": 0.029, "step": 23897 }, { "epoch": 2.4, "learning_rate": 2.014246870064644e-05, "loss": 0.128, "step": 23898 }, { "epoch": 2.4, "learning_rate": 2.0140208444679046e-05, "loss": 0.1478, "step": 23899 }, { "epoch": 2.4, "learning_rate": 2.0137948251455605e-05, "loss": 0.397, "step": 23900 }, { "epoch": 2.4, "learning_rate": 2.0135688120990503e-05, "loss": 0.2431, "step": 23901 }, { "epoch": 2.4, "learning_rate": 2.0133428053298135e-05, "loss": 0.4298, "step": 23902 }, { "epoch": 2.4, "learning_rate": 2.0131168048392883e-05, "loss": 0.5762, "step": 23903 }, { "epoch": 2.4, "learning_rate": 2.012890810628911e-05, "loss": 0.1963, "step": 23904 }, { "epoch": 2.4, "learning_rate": 2.0126648227001217e-05, "loss": 0.8858, "step": 23905 }, { "epoch": 2.4, "learning_rate": 2.0124388410543592e-05, "loss": 0.9872, "step": 23906 }, { "epoch": 2.4, "learning_rate": 2.0122128656930576e-05, "loss": 0.2434, "step": 23907 }, { "epoch": 2.4, "learning_rate": 2.0119868966176598e-05, "loss": 0.3524, "step": 23908 }, { "epoch": 2.4, "learning_rate": 2.0117609338296004e-05, "loss": 0.8228, "step": 23909 }, { "epoch": 2.4, "learning_rate": 2.0115349773303186e-05, "loss": 0.1153, "step": 23910 }, { "epoch": 2.4, "learning_rate": 2.011309027121252e-05, "loss": 0.1206, "step": 23911 }, { "epoch": 2.4, "learning_rate": 2.0110830832038397e-05, "loss": 0.1469, "step": 23912 }, { "epoch": 2.4, "learning_rate": 2.010857145579517e-05, "loss": 0.4436, "step": 23913 }, { "epoch": 2.4, "learning_rate": 2.010631214249724e-05, "loss": 0.1493, "step": 23914 }, { "epoch": 2.4, "learning_rate": 2.010405289215897e-05, "loss": 0.3501, "step": 23915 }, { "epoch": 2.4, "learning_rate": 2.0101793704794745e-05, "loss": 0.5927, "step": 23916 }, { "epoch": 2.4, "learning_rate": 2.009953458041895e-05, "loss": 0.338, "step": 23917 }, { "epoch": 2.4, "learning_rate": 2.0097275519045937e-05, "loss": 0.1799, "step": 23918 }, { "epoch": 2.4, "learning_rate": 2.0095016520690103e-05, "loss": 0.175, "step": 23919 }, { "epoch": 2.4, "learning_rate": 2.0092757585365812e-05, "loss": 0.3901, "step": 23920 }, { "epoch": 2.4, "learning_rate": 2.009049871308744e-05, "loss": 0.6198, "step": 23921 }, { "epoch": 2.4, "learning_rate": 2.0088239903869375e-05, "loss": 0.4087, "step": 23922 }, { "epoch": 2.4, "learning_rate": 2.0085981157725973e-05, "loss": 0.5827, "step": 23923 }, { "epoch": 2.4, "learning_rate": 2.008372247467161e-05, "loss": 0.7977, "step": 23924 }, { "epoch": 2.4, "learning_rate": 2.0081463854720675e-05, "loss": 0.1424, "step": 23925 }, { "epoch": 2.4, "learning_rate": 2.0079205297887535e-05, "loss": 0.2377, "step": 23926 }, { "epoch": 2.4, "learning_rate": 2.0076946804186545e-05, "loss": 0.607, "step": 23927 }, { "epoch": 2.4, "learning_rate": 2.0074688373632098e-05, "loss": 0.2544, "step": 23928 }, { "epoch": 2.4, "learning_rate": 2.0072430006238553e-05, "loss": 0.1618, "step": 23929 }, { "epoch": 2.4, "learning_rate": 2.0070171702020287e-05, "loss": 0.0578, "step": 23930 }, { "epoch": 2.4, "learning_rate": 2.006791346099168e-05, "loss": 0.5491, "step": 23931 }, { "epoch": 2.4, "learning_rate": 2.006565528316709e-05, "loss": 0.3305, "step": 23932 }, { "epoch": 2.4, "learning_rate": 2.006339716856088e-05, "loss": 0.3882, "step": 23933 }, { "epoch": 2.4, "learning_rate": 2.006113911718744e-05, "loss": 0.2217, "step": 23934 }, { "epoch": 2.4, "learning_rate": 2.0058881129061136e-05, "loss": 0.4928, "step": 23935 }, { "epoch": 2.4, "learning_rate": 2.005662320419632e-05, "loss": 0.4854, "step": 23936 }, { "epoch": 2.41, "learning_rate": 2.005436534260737e-05, "loss": 0.263, "step": 23937 }, { "epoch": 2.41, "learning_rate": 2.0052107544308667e-05, "loss": 0.8245, "step": 23938 }, { "epoch": 2.41, "learning_rate": 2.004984980931455e-05, "loss": 0.1671, "step": 23939 }, { "epoch": 2.41, "learning_rate": 2.0047592137639417e-05, "loss": 0.0809, "step": 23940 }, { "epoch": 2.41, "learning_rate": 2.0045334529297613e-05, "loss": 0.4785, "step": 23941 }, { "epoch": 2.41, "learning_rate": 2.004307698430352e-05, "loss": 0.2968, "step": 23942 }, { "epoch": 2.41, "learning_rate": 2.0040819502671494e-05, "loss": 0.1642, "step": 23943 }, { "epoch": 2.41, "learning_rate": 2.0038562084415904e-05, "loss": 0.0961, "step": 23944 }, { "epoch": 2.41, "learning_rate": 2.003630472955112e-05, "loss": 0.5137, "step": 23945 }, { "epoch": 2.41, "learning_rate": 2.0034047438091502e-05, "loss": 0.2344, "step": 23946 }, { "epoch": 2.41, "learning_rate": 2.0031790210051407e-05, "loss": 0.2781, "step": 23947 }, { "epoch": 2.41, "learning_rate": 2.002953304544521e-05, "loss": 0.4418, "step": 23948 }, { "epoch": 2.41, "learning_rate": 2.0027275944287284e-05, "loss": 0.2724, "step": 23949 }, { "epoch": 2.41, "learning_rate": 2.0025018906591962e-05, "loss": 0.2201, "step": 23950 }, { "epoch": 2.41, "learning_rate": 2.0022761932373635e-05, "loss": 0.2746, "step": 23951 }, { "epoch": 2.41, "learning_rate": 2.002050502164665e-05, "loss": 0.3653, "step": 23952 }, { "epoch": 2.41, "learning_rate": 2.0018248174425377e-05, "loss": 0.2537, "step": 23953 }, { "epoch": 2.41, "learning_rate": 2.001599139072418e-05, "loss": 0.4298, "step": 23954 }, { "epoch": 2.41, "learning_rate": 2.0013734670557415e-05, "loss": 0.3137, "step": 23955 }, { "epoch": 2.41, "learning_rate": 2.001147801393943e-05, "loss": 0.4306, "step": 23956 }, { "epoch": 2.41, "learning_rate": 2.0009221420884616e-05, "loss": 0.7281, "step": 23957 }, { "epoch": 2.41, "learning_rate": 2.0006964891407314e-05, "loss": 0.3254, "step": 23958 }, { "epoch": 2.41, "learning_rate": 2.000470842552188e-05, "loss": 0.1008, "step": 23959 }, { "epoch": 2.41, "learning_rate": 2.0002452023242678e-05, "loss": 0.4252, "step": 23960 }, { "epoch": 2.41, "learning_rate": 2.000019568458407e-05, "loss": 0.2026, "step": 23961 }, { "epoch": 2.41, "learning_rate": 1.999793940956041e-05, "loss": 0.6523, "step": 23962 }, { "epoch": 2.41, "learning_rate": 1.9995683198186068e-05, "loss": 0.2391, "step": 23963 }, { "epoch": 2.41, "learning_rate": 1.999342705047539e-05, "loss": 0.5784, "step": 23964 }, { "epoch": 2.41, "learning_rate": 1.9991170966442726e-05, "loss": 0.685, "step": 23965 }, { "epoch": 2.41, "learning_rate": 1.9988914946102445e-05, "loss": 0.5577, "step": 23966 }, { "epoch": 2.41, "learning_rate": 1.99866589894689e-05, "loss": 0.2662, "step": 23967 }, { "epoch": 2.41, "learning_rate": 1.998440309655646e-05, "loss": 0.3132, "step": 23968 }, { "epoch": 2.41, "learning_rate": 1.9982147267379463e-05, "loss": 0.2829, "step": 23969 }, { "epoch": 2.41, "learning_rate": 1.997989150195226e-05, "loss": 0.7331, "step": 23970 }, { "epoch": 2.41, "learning_rate": 1.9977635800289223e-05, "loss": 0.4398, "step": 23971 }, { "epoch": 2.41, "learning_rate": 1.9975380162404707e-05, "loss": 0.1992, "step": 23972 }, { "epoch": 2.41, "learning_rate": 1.9973124588313045e-05, "loss": 0.5808, "step": 23973 }, { "epoch": 2.41, "learning_rate": 1.9970869078028608e-05, "loss": 0.7003, "step": 23974 }, { "epoch": 2.41, "learning_rate": 1.996861363156575e-05, "loss": 0.3127, "step": 23975 }, { "epoch": 2.41, "learning_rate": 1.9966358248938807e-05, "loss": 0.3433, "step": 23976 }, { "epoch": 2.41, "learning_rate": 1.996410293016216e-05, "loss": 0.4404, "step": 23977 }, { "epoch": 2.41, "learning_rate": 1.9961847675250133e-05, "loss": 0.1501, "step": 23978 }, { "epoch": 2.41, "learning_rate": 1.995959248421709e-05, "loss": 0.7187, "step": 23979 }, { "epoch": 2.41, "learning_rate": 1.995733735707738e-05, "loss": 0.2245, "step": 23980 }, { "epoch": 2.41, "learning_rate": 1.995508229384536e-05, "loss": 0.5115, "step": 23981 }, { "epoch": 2.41, "learning_rate": 1.9952827294535365e-05, "loss": 0.1503, "step": 23982 }, { "epoch": 2.41, "learning_rate": 1.9950572359161765e-05, "loss": 0.743, "step": 23983 }, { "epoch": 2.41, "learning_rate": 1.9948317487738894e-05, "loss": 0.7003, "step": 23984 }, { "epoch": 2.41, "learning_rate": 1.99460626802811e-05, "loss": 0.3006, "step": 23985 }, { "epoch": 2.41, "learning_rate": 1.9943807936802756e-05, "loss": 0.1666, "step": 23986 }, { "epoch": 2.41, "learning_rate": 1.9941553257318184e-05, "loss": 0.2906, "step": 23987 }, { "epoch": 2.41, "learning_rate": 1.9939298641841735e-05, "loss": 0.5304, "step": 23988 }, { "epoch": 2.41, "learning_rate": 1.9937044090387766e-05, "loss": 0.0774, "step": 23989 }, { "epoch": 2.41, "learning_rate": 1.993478960297063e-05, "loss": 0.2965, "step": 23990 }, { "epoch": 2.41, "learning_rate": 1.993253517960465e-05, "loss": 0.3759, "step": 23991 }, { "epoch": 2.41, "learning_rate": 1.9930280820304196e-05, "loss": 0.7803, "step": 23992 }, { "epoch": 2.41, "learning_rate": 1.9928026525083595e-05, "loss": 0.3241, "step": 23993 }, { "epoch": 2.41, "learning_rate": 1.9925772293957202e-05, "loss": 0.8965, "step": 23994 }, { "epoch": 2.41, "learning_rate": 1.9923518126939372e-05, "loss": 0.3479, "step": 23995 }, { "epoch": 2.41, "learning_rate": 1.992126402404443e-05, "loss": 0.1536, "step": 23996 }, { "epoch": 2.41, "learning_rate": 1.9919009985286734e-05, "loss": 0.4404, "step": 23997 }, { "epoch": 2.41, "learning_rate": 1.9916756010680623e-05, "loss": 0.0675, "step": 23998 }, { "epoch": 2.41, "learning_rate": 1.9914502100240433e-05, "loss": 0.6661, "step": 23999 }, { "epoch": 2.41, "learning_rate": 1.9912248253980532e-05, "loss": 0.2687, "step": 24000 }, { "epoch": 2.41, "learning_rate": 1.9909994471915236e-05, "loss": 0.3966, "step": 24001 }, { "epoch": 2.41, "learning_rate": 1.990774075405889e-05, "loss": 0.8188, "step": 24002 }, { "epoch": 2.41, "learning_rate": 1.9905487100425848e-05, "loss": 0.3751, "step": 24003 }, { "epoch": 2.41, "learning_rate": 1.990323351103045e-05, "loss": 0.2028, "step": 24004 }, { "epoch": 2.41, "learning_rate": 1.990097998588703e-05, "loss": 0.5294, "step": 24005 }, { "epoch": 2.41, "learning_rate": 1.989872652500993e-05, "loss": 0.2172, "step": 24006 }, { "epoch": 2.41, "learning_rate": 1.9896473128413496e-05, "loss": 0.2989, "step": 24007 }, { "epoch": 2.41, "learning_rate": 1.9894219796112053e-05, "loss": 0.1506, "step": 24008 }, { "epoch": 2.41, "learning_rate": 1.9891966528119966e-05, "loss": 0.6364, "step": 24009 }, { "epoch": 2.41, "learning_rate": 1.988971332445155e-05, "loss": 0.2679, "step": 24010 }, { "epoch": 2.41, "learning_rate": 1.988746018512115e-05, "loss": 0.3503, "step": 24011 }, { "epoch": 2.41, "learning_rate": 1.9885207110143113e-05, "loss": 0.5595, "step": 24012 }, { "epoch": 2.41, "learning_rate": 1.988295409953177e-05, "loss": 0.1526, "step": 24013 }, { "epoch": 2.41, "learning_rate": 1.9880701153301448e-05, "loss": 0.5467, "step": 24014 }, { "epoch": 2.41, "learning_rate": 1.98784482714665e-05, "loss": 0.2143, "step": 24015 }, { "epoch": 2.41, "learning_rate": 1.9876195454041262e-05, "loss": 0.4256, "step": 24016 }, { "epoch": 2.41, "learning_rate": 1.9873942701040057e-05, "loss": 0.7123, "step": 24017 }, { "epoch": 2.41, "learning_rate": 1.987169001247724e-05, "loss": 0.3564, "step": 24018 }, { "epoch": 2.41, "learning_rate": 1.9869437388367128e-05, "loss": 0.5332, "step": 24019 }, { "epoch": 2.41, "learning_rate": 1.9867184828724057e-05, "loss": 0.7808, "step": 24020 }, { "epoch": 2.41, "learning_rate": 1.9864932333562372e-05, "loss": 0.2989, "step": 24021 }, { "epoch": 2.41, "learning_rate": 1.9862679902896394e-05, "loss": 0.3915, "step": 24022 }, { "epoch": 2.41, "learning_rate": 1.9860427536740483e-05, "loss": 0.1106, "step": 24023 }, { "epoch": 2.41, "learning_rate": 1.985817523510894e-05, "loss": 0.6891, "step": 24024 }, { "epoch": 2.41, "learning_rate": 1.985592299801611e-05, "loss": 0.1226, "step": 24025 }, { "epoch": 2.41, "learning_rate": 1.9853670825476332e-05, "loss": 0.3431, "step": 24026 }, { "epoch": 2.41, "learning_rate": 1.985141871750394e-05, "loss": 0.4621, "step": 24027 }, { "epoch": 2.41, "learning_rate": 1.984916667411325e-05, "loss": 0.1599, "step": 24028 }, { "epoch": 2.41, "learning_rate": 1.9846914695318602e-05, "loss": 0.4572, "step": 24029 }, { "epoch": 2.41, "learning_rate": 1.984466278113433e-05, "loss": 0.6936, "step": 24030 }, { "epoch": 2.41, "learning_rate": 1.9842410931574752e-05, "loss": 0.2344, "step": 24031 }, { "epoch": 2.41, "learning_rate": 1.9840159146654222e-05, "loss": 0.0369, "step": 24032 }, { "epoch": 2.41, "learning_rate": 1.9837907426387046e-05, "loss": 0.1924, "step": 24033 }, { "epoch": 2.41, "learning_rate": 1.9835655770787552e-05, "loss": 0.2258, "step": 24034 }, { "epoch": 2.42, "learning_rate": 1.9833404179870086e-05, "loss": 1.2785, "step": 24035 }, { "epoch": 2.42, "learning_rate": 1.983115265364898e-05, "loss": 0.2955, "step": 24036 }, { "epoch": 2.42, "learning_rate": 1.9828901192138533e-05, "loss": 0.3928, "step": 24037 }, { "epoch": 2.42, "learning_rate": 1.9826649795353097e-05, "loss": 0.3974, "step": 24038 }, { "epoch": 2.42, "learning_rate": 1.9824398463306992e-05, "loss": 1.2477, "step": 24039 }, { "epoch": 2.42, "learning_rate": 1.9822147196014533e-05, "loss": 0.3703, "step": 24040 }, { "epoch": 2.42, "learning_rate": 1.9819895993490076e-05, "loss": 0.0605, "step": 24041 }, { "epoch": 2.42, "learning_rate": 1.981764485574792e-05, "loss": 0.1599, "step": 24042 }, { "epoch": 2.42, "learning_rate": 1.981539378280239e-05, "loss": 0.0565, "step": 24043 }, { "epoch": 2.42, "learning_rate": 1.981314277466783e-05, "loss": 0.3991, "step": 24044 }, { "epoch": 2.42, "learning_rate": 1.9810891831358558e-05, "loss": 0.1188, "step": 24045 }, { "epoch": 2.42, "learning_rate": 1.9808640952888882e-05, "loss": 0.383, "step": 24046 }, { "epoch": 2.42, "learning_rate": 1.9806390139273145e-05, "loss": 0.6873, "step": 24047 }, { "epoch": 2.42, "learning_rate": 1.980413939052566e-05, "loss": 0.233, "step": 24048 }, { "epoch": 2.42, "learning_rate": 1.980188870666075e-05, "loss": 0.7002, "step": 24049 }, { "epoch": 2.42, "learning_rate": 1.9799638087692752e-05, "loss": 0.2826, "step": 24050 }, { "epoch": 2.42, "learning_rate": 1.9797387533635967e-05, "loss": 0.1335, "step": 24051 }, { "epoch": 2.42, "learning_rate": 1.979513704450473e-05, "loss": 0.0779, "step": 24052 }, { "epoch": 2.42, "learning_rate": 1.979288662031336e-05, "loss": 0.5335, "step": 24053 }, { "epoch": 2.42, "learning_rate": 1.979063626107617e-05, "loss": 0.1374, "step": 24054 }, { "epoch": 2.42, "learning_rate": 1.9788385966807496e-05, "loss": 0.445, "step": 24055 }, { "epoch": 2.42, "learning_rate": 1.978613573752165e-05, "loss": 0.2281, "step": 24056 }, { "epoch": 2.42, "learning_rate": 1.978388557323294e-05, "loss": 0.3066, "step": 24057 }, { "epoch": 2.42, "learning_rate": 1.9781635473955696e-05, "loss": 0.0864, "step": 24058 }, { "epoch": 2.42, "learning_rate": 1.9779385439704254e-05, "loss": 0.7794, "step": 24059 }, { "epoch": 2.42, "learning_rate": 1.9777135470492897e-05, "loss": 0.5751, "step": 24060 }, { "epoch": 2.42, "learning_rate": 1.9774885566335968e-05, "loss": 0.4188, "step": 24061 }, { "epoch": 2.42, "learning_rate": 1.9772635727247774e-05, "loss": 0.5644, "step": 24062 }, { "epoch": 2.42, "learning_rate": 1.9770385953242634e-05, "loss": 0.5664, "step": 24063 }, { "epoch": 2.42, "learning_rate": 1.9768136244334876e-05, "loss": 0.1822, "step": 24064 }, { "epoch": 2.42, "learning_rate": 1.97658866005388e-05, "loss": 0.4245, "step": 24065 }, { "epoch": 2.42, "learning_rate": 1.9763637021868727e-05, "loss": 0.3225, "step": 24066 }, { "epoch": 2.42, "learning_rate": 1.9761387508338978e-05, "loss": 0.6072, "step": 24067 }, { "epoch": 2.42, "learning_rate": 1.9759138059963867e-05, "loss": 0.3116, "step": 24068 }, { "epoch": 2.42, "learning_rate": 1.9756888676757696e-05, "loss": 0.2718, "step": 24069 }, { "epoch": 2.42, "learning_rate": 1.9754639358734793e-05, "loss": 0.3929, "step": 24070 }, { "epoch": 2.42, "learning_rate": 1.975239010590947e-05, "loss": 0.7058, "step": 24071 }, { "epoch": 2.42, "learning_rate": 1.975014091829603e-05, "loss": 0.2142, "step": 24072 }, { "epoch": 2.42, "learning_rate": 1.9747891795908803e-05, "loss": 0.1617, "step": 24073 }, { "epoch": 2.42, "learning_rate": 1.974564273876209e-05, "loss": 0.9299, "step": 24074 }, { "epoch": 2.42, "learning_rate": 1.9743393746870202e-05, "loss": 0.237, "step": 24075 }, { "epoch": 2.42, "learning_rate": 1.9741144820247458e-05, "loss": 0.1902, "step": 24076 }, { "epoch": 2.42, "learning_rate": 1.973889595890816e-05, "loss": 0.3858, "step": 24077 }, { "epoch": 2.42, "learning_rate": 1.9736647162866633e-05, "loss": 0.2554, "step": 24078 }, { "epoch": 2.42, "learning_rate": 1.9734398432137182e-05, "loss": 0.6601, "step": 24079 }, { "epoch": 2.42, "learning_rate": 1.97321497667341e-05, "loss": 0.3041, "step": 24080 }, { "epoch": 2.42, "learning_rate": 1.9729901166671718e-05, "loss": 0.3923, "step": 24081 }, { "epoch": 2.42, "learning_rate": 1.9727652631964346e-05, "loss": 0.2964, "step": 24082 }, { "epoch": 2.42, "learning_rate": 1.9725404162626274e-05, "loss": 0.586, "step": 24083 }, { "epoch": 2.42, "learning_rate": 1.9723155758671827e-05, "loss": 0.6712, "step": 24084 }, { "epoch": 2.42, "learning_rate": 1.9720907420115303e-05, "loss": 0.3359, "step": 24085 }, { "epoch": 2.42, "learning_rate": 1.9718659146971014e-05, "loss": 0.3982, "step": 24086 }, { "epoch": 2.42, "learning_rate": 1.9716410939253274e-05, "loss": 0.1502, "step": 24087 }, { "epoch": 2.42, "learning_rate": 1.9714162796976378e-05, "loss": 0.1147, "step": 24088 }, { "epoch": 2.42, "learning_rate": 1.971191472015463e-05, "loss": 0.1485, "step": 24089 }, { "epoch": 2.42, "learning_rate": 1.970966670880235e-05, "loss": 0.6206, "step": 24090 }, { "epoch": 2.42, "learning_rate": 1.9707418762933844e-05, "loss": 0.4966, "step": 24091 }, { "epoch": 2.42, "learning_rate": 1.9705170882563395e-05, "loss": 0.1618, "step": 24092 }, { "epoch": 2.42, "learning_rate": 1.9702923067705326e-05, "loss": 0.1339, "step": 24093 }, { "epoch": 2.42, "learning_rate": 1.9700675318373943e-05, "loss": 0.1422, "step": 24094 }, { "epoch": 2.42, "learning_rate": 1.9698427634583536e-05, "loss": 0.506, "step": 24095 }, { "epoch": 2.42, "learning_rate": 1.9696180016348432e-05, "loss": 0.1882, "step": 24096 }, { "epoch": 2.42, "learning_rate": 1.9693932463682904e-05, "loss": 0.675, "step": 24097 }, { "epoch": 2.42, "learning_rate": 1.969168497660127e-05, "loss": 0.2717, "step": 24098 }, { "epoch": 2.42, "learning_rate": 1.9689437555117834e-05, "loss": 0.3408, "step": 24099 }, { "epoch": 2.42, "learning_rate": 1.9687190199246902e-05, "loss": 0.5645, "step": 24100 }, { "epoch": 2.42, "learning_rate": 1.9684942909002762e-05, "loss": 0.3232, "step": 24101 }, { "epoch": 2.42, "learning_rate": 1.9682695684399718e-05, "loss": 0.3094, "step": 24102 }, { "epoch": 2.42, "learning_rate": 1.9680448525452074e-05, "loss": 0.5222, "step": 24103 }, { "epoch": 2.42, "learning_rate": 1.9678201432174133e-05, "loss": 0.6274, "step": 24104 }, { "epoch": 2.42, "learning_rate": 1.96759544045802e-05, "loss": 0.2265, "step": 24105 }, { "epoch": 2.42, "learning_rate": 1.967370744268455e-05, "loss": 0.2509, "step": 24106 }, { "epoch": 2.42, "learning_rate": 1.9671460546501507e-05, "loss": 0.5845, "step": 24107 }, { "epoch": 2.42, "learning_rate": 1.966921371604536e-05, "loss": 0.7519, "step": 24108 }, { "epoch": 2.42, "learning_rate": 1.9666966951330398e-05, "loss": 0.3182, "step": 24109 }, { "epoch": 2.42, "learning_rate": 1.9664720252370942e-05, "loss": 0.3257, "step": 24110 }, { "epoch": 2.42, "learning_rate": 1.9662473619181267e-05, "loss": 0.0596, "step": 24111 }, { "epoch": 2.42, "learning_rate": 1.966022705177567e-05, "loss": 0.4206, "step": 24112 }, { "epoch": 2.42, "learning_rate": 1.965798055016846e-05, "loss": 0.7701, "step": 24113 }, { "epoch": 2.42, "learning_rate": 1.965573411437394e-05, "loss": 0.7798, "step": 24114 }, { "epoch": 2.42, "learning_rate": 1.9653487744406373e-05, "loss": 0.8164, "step": 24115 }, { "epoch": 2.42, "learning_rate": 1.9651241440280084e-05, "loss": 1.6996, "step": 24116 }, { "epoch": 2.42, "learning_rate": 1.9648995202009356e-05, "loss": 0.1413, "step": 24117 }, { "epoch": 2.42, "learning_rate": 1.9646749029608484e-05, "loss": 0.1742, "step": 24118 }, { "epoch": 2.42, "learning_rate": 1.9644502923091766e-05, "loss": 0.0565, "step": 24119 }, { "epoch": 2.42, "learning_rate": 1.9642256882473497e-05, "loss": 0.4496, "step": 24120 }, { "epoch": 2.42, "learning_rate": 1.9640010907767953e-05, "loss": 0.3469, "step": 24121 }, { "epoch": 2.42, "learning_rate": 1.9637764998989437e-05, "loss": 0.4631, "step": 24122 }, { "epoch": 2.42, "learning_rate": 1.9635519156152256e-05, "loss": 0.2835, "step": 24123 }, { "epoch": 2.42, "learning_rate": 1.9633273379270678e-05, "loss": 0.1132, "step": 24124 }, { "epoch": 2.42, "learning_rate": 1.9631027668359007e-05, "loss": 0.1674, "step": 24125 }, { "epoch": 2.42, "learning_rate": 1.9628782023431535e-05, "loss": 0.4772, "step": 24126 }, { "epoch": 2.42, "learning_rate": 1.962653644450254e-05, "loss": 0.2352, "step": 24127 }, { "epoch": 2.42, "learning_rate": 1.962429093158633e-05, "loss": 0.2571, "step": 24128 }, { "epoch": 2.42, "learning_rate": 1.9622045484697185e-05, "loss": 0.2191, "step": 24129 }, { "epoch": 2.42, "learning_rate": 1.9619800103849385e-05, "loss": 0.2223, "step": 24130 }, { "epoch": 2.42, "learning_rate": 1.961755478905724e-05, "loss": 0.4183, "step": 24131 }, { "epoch": 2.42, "learning_rate": 1.9615309540335012e-05, "loss": 0.2381, "step": 24132 }, { "epoch": 2.42, "learning_rate": 1.961306435769702e-05, "loss": 0.6157, "step": 24133 }, { "epoch": 2.43, "learning_rate": 1.961081924115753e-05, "loss": 1.5884, "step": 24134 }, { "epoch": 2.43, "learning_rate": 1.960857419073083e-05, "loss": 0.6242, "step": 24135 }, { "epoch": 2.43, "learning_rate": 1.9606329206431208e-05, "loss": 0.6006, "step": 24136 }, { "epoch": 2.43, "learning_rate": 1.9604084288272966e-05, "loss": 0.315, "step": 24137 }, { "epoch": 2.43, "learning_rate": 1.9601839436270363e-05, "loss": 0.8188, "step": 24138 }, { "epoch": 2.43, "learning_rate": 1.9599594650437703e-05, "loss": 0.5006, "step": 24139 }, { "epoch": 2.43, "learning_rate": 1.9597349930789273e-05, "loss": 0.2459, "step": 24140 }, { "epoch": 2.43, "learning_rate": 1.9595105277339334e-05, "loss": 0.2252, "step": 24141 }, { "epoch": 2.43, "learning_rate": 1.9592860690102205e-05, "loss": 0.522, "step": 24142 }, { "epoch": 2.43, "learning_rate": 1.9590616169092143e-05, "loss": 0.7705, "step": 24143 }, { "epoch": 2.43, "learning_rate": 1.9588371714323438e-05, "loss": 0.5128, "step": 24144 }, { "epoch": 2.43, "learning_rate": 1.9586127325810375e-05, "loss": 0.3912, "step": 24145 }, { "epoch": 2.43, "learning_rate": 1.9583883003567246e-05, "loss": 0.6828, "step": 24146 }, { "epoch": 2.43, "learning_rate": 1.958163874760831e-05, "loss": 0.3297, "step": 24147 }, { "epoch": 2.43, "learning_rate": 1.957939455794787e-05, "loss": 0.6529, "step": 24148 }, { "epoch": 2.43, "learning_rate": 1.9577150434600197e-05, "loss": 0.2217, "step": 24149 }, { "epoch": 2.43, "learning_rate": 1.9574906377579567e-05, "loss": 0.2436, "step": 24150 }, { "epoch": 2.43, "learning_rate": 1.957266238690028e-05, "loss": 0.8818, "step": 24151 }, { "epoch": 2.43, "learning_rate": 1.9570418462576592e-05, "loss": 0.4207, "step": 24152 }, { "epoch": 2.43, "learning_rate": 1.9568174604622793e-05, "loss": 0.1304, "step": 24153 }, { "epoch": 2.43, "learning_rate": 1.9565930813053164e-05, "loss": 0.3517, "step": 24154 }, { "epoch": 2.43, "learning_rate": 1.9563687087881994e-05, "loss": 0.3705, "step": 24155 }, { "epoch": 2.43, "learning_rate": 1.956144342912353e-05, "loss": 0.0796, "step": 24156 }, { "epoch": 2.43, "learning_rate": 1.9559199836792083e-05, "loss": 0.2828, "step": 24157 }, { "epoch": 2.43, "learning_rate": 1.9556956310901904e-05, "loss": 0.6469, "step": 24158 }, { "epoch": 2.43, "learning_rate": 1.955471285146729e-05, "loss": 0.5228, "step": 24159 }, { "epoch": 2.43, "learning_rate": 1.9552469458502517e-05, "loss": 0.2332, "step": 24160 }, { "epoch": 2.43, "learning_rate": 1.955022613202184e-05, "loss": 0.0911, "step": 24161 }, { "epoch": 2.43, "learning_rate": 1.9547982872039556e-05, "loss": 0.4412, "step": 24162 }, { "epoch": 2.43, "learning_rate": 1.9545739678569934e-05, "loss": 0.1731, "step": 24163 }, { "epoch": 2.43, "learning_rate": 1.9543496551627238e-05, "loss": 0.0328, "step": 24164 }, { "epoch": 2.43, "learning_rate": 1.954125349122577e-05, "loss": 0.297, "step": 24165 }, { "epoch": 2.43, "learning_rate": 1.9539010497379774e-05, "loss": 0.5212, "step": 24166 }, { "epoch": 2.43, "learning_rate": 1.953676757010353e-05, "loss": 0.1468, "step": 24167 }, { "epoch": 2.43, "learning_rate": 1.9534524709411326e-05, "loss": 0.5236, "step": 24168 }, { "epoch": 2.43, "learning_rate": 1.953228191531743e-05, "loss": 0.1668, "step": 24169 }, { "epoch": 2.43, "learning_rate": 1.9530039187836096e-05, "loss": 0.2907, "step": 24170 }, { "epoch": 2.43, "learning_rate": 1.952779652698162e-05, "loss": 0.3543, "step": 24171 }, { "epoch": 2.43, "learning_rate": 1.9525553932768263e-05, "loss": 0.4123, "step": 24172 }, { "epoch": 2.43, "learning_rate": 1.9523311405210286e-05, "loss": 0.6863, "step": 24173 }, { "epoch": 2.43, "learning_rate": 1.9521068944321984e-05, "loss": 0.1421, "step": 24174 }, { "epoch": 2.43, "learning_rate": 1.9518826550117604e-05, "loss": 0.6157, "step": 24175 }, { "epoch": 2.43, "learning_rate": 1.9516584222611424e-05, "loss": 0.6945, "step": 24176 }, { "epoch": 2.43, "learning_rate": 1.9514341961817715e-05, "loss": 0.3564, "step": 24177 }, { "epoch": 2.43, "learning_rate": 1.951209976775075e-05, "loss": 0.2298, "step": 24178 }, { "epoch": 2.43, "learning_rate": 1.9509857640424786e-05, "loss": 0.1459, "step": 24179 }, { "epoch": 2.43, "learning_rate": 1.95076155798541e-05, "loss": 0.4661, "step": 24180 }, { "epoch": 2.43, "learning_rate": 1.9505373586052956e-05, "loss": 0.2508, "step": 24181 }, { "epoch": 2.43, "learning_rate": 1.950313165903562e-05, "loss": 0.1721, "step": 24182 }, { "epoch": 2.43, "learning_rate": 1.950088979881637e-05, "loss": 0.2993, "step": 24183 }, { "epoch": 2.43, "learning_rate": 1.949864800540945e-05, "loss": 0.2343, "step": 24184 }, { "epoch": 2.43, "learning_rate": 1.9496406278829143e-05, "loss": 0.7358, "step": 24185 }, { "epoch": 2.43, "learning_rate": 1.949416461908972e-05, "loss": 0.4722, "step": 24186 }, { "epoch": 2.43, "learning_rate": 1.949192302620542e-05, "loss": 0.2257, "step": 24187 }, { "epoch": 2.43, "learning_rate": 1.9489681500190545e-05, "loss": 0.2888, "step": 24188 }, { "epoch": 2.43, "learning_rate": 1.948744004105933e-05, "loss": 0.7169, "step": 24189 }, { "epoch": 2.43, "learning_rate": 1.9485198648826037e-05, "loss": 0.5813, "step": 24190 }, { "epoch": 2.43, "learning_rate": 1.9482957323504948e-05, "loss": 0.1469, "step": 24191 }, { "epoch": 2.43, "learning_rate": 1.9480716065110327e-05, "loss": 0.1657, "step": 24192 }, { "epoch": 2.43, "learning_rate": 1.9478474873656415e-05, "loss": 0.1719, "step": 24193 }, { "epoch": 2.43, "learning_rate": 1.947623374915749e-05, "loss": 0.1155, "step": 24194 }, { "epoch": 2.43, "learning_rate": 1.947399269162781e-05, "loss": 0.2097, "step": 24195 }, { "epoch": 2.43, "learning_rate": 1.9471751701081627e-05, "loss": 0.0959, "step": 24196 }, { "epoch": 2.43, "learning_rate": 1.946951077753323e-05, "loss": 0.0743, "step": 24197 }, { "epoch": 2.43, "learning_rate": 1.946726992099685e-05, "loss": 0.5153, "step": 24198 }, { "epoch": 2.43, "learning_rate": 1.946502913148675e-05, "loss": 0.2751, "step": 24199 }, { "epoch": 2.43, "learning_rate": 1.9462788409017207e-05, "loss": 0.2631, "step": 24200 }, { "epoch": 2.43, "learning_rate": 1.946054775360248e-05, "loss": 0.6196, "step": 24201 }, { "epoch": 2.43, "learning_rate": 1.9458307165256794e-05, "loss": 0.3141, "step": 24202 }, { "epoch": 2.43, "learning_rate": 1.9456066643994446e-05, "loss": 0.4081, "step": 24203 }, { "epoch": 2.43, "learning_rate": 1.9453826189829673e-05, "loss": 0.7386, "step": 24204 }, { "epoch": 2.43, "learning_rate": 1.9451585802776737e-05, "loss": 0.2452, "step": 24205 }, { "epoch": 2.43, "learning_rate": 1.9449345482849905e-05, "loss": 0.2066, "step": 24206 }, { "epoch": 2.43, "learning_rate": 1.944710523006342e-05, "loss": 0.2411, "step": 24207 }, { "epoch": 2.43, "learning_rate": 1.9444865044431535e-05, "loss": 0.421, "step": 24208 }, { "epoch": 2.43, "learning_rate": 1.9442624925968525e-05, "loss": 0.2617, "step": 24209 }, { "epoch": 2.43, "learning_rate": 1.9440384874688635e-05, "loss": 0.3146, "step": 24210 }, { "epoch": 2.43, "learning_rate": 1.9438144890606104e-05, "loss": 0.3534, "step": 24211 }, { "epoch": 2.43, "learning_rate": 1.943590497373521e-05, "loss": 0.2684, "step": 24212 }, { "epoch": 2.43, "learning_rate": 1.9433665124090196e-05, "loss": 0.1387, "step": 24213 }, { "epoch": 2.43, "learning_rate": 1.943142534168532e-05, "loss": 0.1519, "step": 24214 }, { "epoch": 2.43, "learning_rate": 1.9429185626534834e-05, "loss": 0.4048, "step": 24215 }, { "epoch": 2.43, "learning_rate": 1.9426945978652975e-05, "loss": 0.2816, "step": 24216 }, { "epoch": 2.43, "learning_rate": 1.942470639805403e-05, "loss": 0.484, "step": 24217 }, { "epoch": 2.43, "learning_rate": 1.9422466884752225e-05, "loss": 0.1729, "step": 24218 }, { "epoch": 2.43, "learning_rate": 1.9420227438761808e-05, "loss": 0.2999, "step": 24219 }, { "epoch": 2.43, "learning_rate": 1.9417988060097052e-05, "loss": 0.2144, "step": 24220 }, { "epoch": 2.43, "learning_rate": 1.9415748748772185e-05, "loss": 0.364, "step": 24221 }, { "epoch": 2.43, "learning_rate": 1.9413509504801463e-05, "loss": 0.254, "step": 24222 }, { "epoch": 2.43, "learning_rate": 1.9411270328199148e-05, "loss": 0.3344, "step": 24223 }, { "epoch": 2.43, "learning_rate": 1.9409031218979482e-05, "loss": 0.5771, "step": 24224 }, { "epoch": 2.43, "learning_rate": 1.9406792177156704e-05, "loss": 0.5116, "step": 24225 }, { "epoch": 2.43, "learning_rate": 1.9404553202745073e-05, "loss": 0.2057, "step": 24226 }, { "epoch": 2.43, "learning_rate": 1.9402314295758837e-05, "loss": 0.5248, "step": 24227 }, { "epoch": 2.43, "learning_rate": 1.9400075456212232e-05, "loss": 0.2382, "step": 24228 }, { "epoch": 2.43, "learning_rate": 1.9397836684119534e-05, "loss": 0.4017, "step": 24229 }, { "epoch": 2.43, "learning_rate": 1.9395597979494958e-05, "loss": 0.2329, "step": 24230 }, { "epoch": 2.43, "learning_rate": 1.9393359342352758e-05, "loss": 0.0948, "step": 24231 }, { "epoch": 2.44, "learning_rate": 1.939112077270719e-05, "loss": 0.0753, "step": 24232 }, { "epoch": 2.44, "learning_rate": 1.93888822705725e-05, "loss": 0.7197, "step": 24233 }, { "epoch": 2.44, "learning_rate": 1.9386643835962914e-05, "loss": 0.2123, "step": 24234 }, { "epoch": 2.44, "learning_rate": 1.9384405468892697e-05, "loss": 0.3464, "step": 24235 }, { "epoch": 2.44, "learning_rate": 1.9382167169376078e-05, "loss": 0.3127, "step": 24236 }, { "epoch": 2.44, "learning_rate": 1.937992893742731e-05, "loss": 0.1339, "step": 24237 }, { "epoch": 2.44, "learning_rate": 1.9377690773060644e-05, "loss": 0.1095, "step": 24238 }, { "epoch": 2.44, "learning_rate": 1.9375452676290295e-05, "loss": 0.7313, "step": 24239 }, { "epoch": 2.44, "learning_rate": 1.937321464713054e-05, "loss": 0.213, "step": 24240 }, { "epoch": 2.44, "learning_rate": 1.93709766855956e-05, "loss": 0.2808, "step": 24241 }, { "epoch": 2.44, "learning_rate": 1.9368738791699714e-05, "loss": 0.1062, "step": 24242 }, { "epoch": 2.44, "learning_rate": 1.936650096545714e-05, "loss": 0.7079, "step": 24243 }, { "epoch": 2.44, "learning_rate": 1.9364263206882104e-05, "loss": 0.5231, "step": 24244 }, { "epoch": 2.44, "learning_rate": 1.936202551598885e-05, "loss": 1.1097, "step": 24245 }, { "epoch": 2.44, "learning_rate": 1.9359787892791617e-05, "loss": 0.5939, "step": 24246 }, { "epoch": 2.44, "learning_rate": 1.9357550337304656e-05, "loss": 0.1825, "step": 24247 }, { "epoch": 2.44, "learning_rate": 1.9355312849542178e-05, "loss": 0.7221, "step": 24248 }, { "epoch": 2.44, "learning_rate": 1.935307542951846e-05, "loss": 0.4606, "step": 24249 }, { "epoch": 2.44, "learning_rate": 1.935083807724771e-05, "loss": 0.1531, "step": 24250 }, { "epoch": 2.44, "learning_rate": 1.934860079274417e-05, "loss": 0.2767, "step": 24251 }, { "epoch": 2.44, "learning_rate": 1.934636357602209e-05, "loss": 0.417, "step": 24252 }, { "epoch": 2.44, "learning_rate": 1.93441264270957e-05, "loss": 0.3357, "step": 24253 }, { "epoch": 2.44, "learning_rate": 1.934188934597923e-05, "loss": 0.3738, "step": 24254 }, { "epoch": 2.44, "learning_rate": 1.9339652332686924e-05, "loss": 0.2617, "step": 24255 }, { "epoch": 2.44, "learning_rate": 1.933741538723302e-05, "loss": 0.2299, "step": 24256 }, { "epoch": 2.44, "learning_rate": 1.933517850963174e-05, "loss": 0.2238, "step": 24257 }, { "epoch": 2.44, "learning_rate": 1.9332941699897332e-05, "loss": 0.5559, "step": 24258 }, { "epoch": 2.44, "learning_rate": 1.9330704958044024e-05, "loss": 0.4953, "step": 24259 }, { "epoch": 2.44, "learning_rate": 1.9328468284086044e-05, "loss": 0.1672, "step": 24260 }, { "epoch": 2.44, "learning_rate": 1.9326231678037646e-05, "loss": 0.6571, "step": 24261 }, { "epoch": 2.44, "learning_rate": 1.9323995139913044e-05, "loss": 0.6962, "step": 24262 }, { "epoch": 2.44, "learning_rate": 1.9321758669726465e-05, "loss": 0.4285, "step": 24263 }, { "epoch": 2.44, "learning_rate": 1.9319522267492165e-05, "loss": 0.2427, "step": 24264 }, { "epoch": 2.44, "learning_rate": 1.9317285933224348e-05, "loss": 0.5345, "step": 24265 }, { "epoch": 2.44, "learning_rate": 1.9315049666937278e-05, "loss": 0.3605, "step": 24266 }, { "epoch": 2.44, "learning_rate": 1.931281346864516e-05, "loss": 0.2921, "step": 24267 }, { "epoch": 2.44, "learning_rate": 1.9310577338362223e-05, "loss": 0.1229, "step": 24268 }, { "epoch": 2.44, "learning_rate": 1.9308341276102713e-05, "loss": 0.3219, "step": 24269 }, { "epoch": 2.44, "learning_rate": 1.930610528188086e-05, "loss": 0.2046, "step": 24270 }, { "epoch": 2.44, "learning_rate": 1.9303869355710858e-05, "loss": 0.3333, "step": 24271 }, { "epoch": 2.44, "learning_rate": 1.9301633497606993e-05, "loss": 0.6798, "step": 24272 }, { "epoch": 2.44, "learning_rate": 1.9299397707583445e-05, "loss": 0.0754, "step": 24273 }, { "epoch": 2.44, "learning_rate": 1.929716198565446e-05, "loss": 0.3002, "step": 24274 }, { "epoch": 2.44, "learning_rate": 1.9294926331834276e-05, "loss": 0.6329, "step": 24275 }, { "epoch": 2.44, "learning_rate": 1.92926907461371e-05, "loss": 0.0953, "step": 24276 }, { "epoch": 2.44, "learning_rate": 1.929045522857716e-05, "loss": 0.0921, "step": 24277 }, { "epoch": 2.44, "learning_rate": 1.92882197791687e-05, "loss": 0.6046, "step": 24278 }, { "epoch": 2.44, "learning_rate": 1.9285984397925935e-05, "loss": 0.2258, "step": 24279 }, { "epoch": 2.44, "learning_rate": 1.9283749084863073e-05, "loss": 0.1469, "step": 24280 }, { "epoch": 2.44, "learning_rate": 1.9281513839994375e-05, "loss": 0.4252, "step": 24281 }, { "epoch": 2.44, "learning_rate": 1.927927866333404e-05, "loss": 0.2946, "step": 24282 }, { "epoch": 2.44, "learning_rate": 1.927704355489629e-05, "loss": 0.3866, "step": 24283 }, { "epoch": 2.44, "learning_rate": 1.9274808514695368e-05, "loss": 0.3758, "step": 24284 }, { "epoch": 2.44, "learning_rate": 1.9272573542745477e-05, "loss": 0.3171, "step": 24285 }, { "epoch": 2.44, "learning_rate": 1.9270338639060844e-05, "loss": 0.2317, "step": 24286 }, { "epoch": 2.44, "learning_rate": 1.9268103803655693e-05, "loss": 0.2626, "step": 24287 }, { "epoch": 2.44, "learning_rate": 1.926586903654426e-05, "loss": 0.5401, "step": 24288 }, { "epoch": 2.44, "learning_rate": 1.926363433774074e-05, "loss": 0.3952, "step": 24289 }, { "epoch": 2.44, "learning_rate": 1.9261399707259367e-05, "loss": 0.1647, "step": 24290 }, { "epoch": 2.44, "learning_rate": 1.9259165145114367e-05, "loss": 0.2317, "step": 24291 }, { "epoch": 2.44, "learning_rate": 1.9256930651319948e-05, "loss": 0.3775, "step": 24292 }, { "epoch": 2.44, "learning_rate": 1.9254696225890344e-05, "loss": 0.0223, "step": 24293 }, { "epoch": 2.44, "learning_rate": 1.9252461868839752e-05, "loss": 0.134, "step": 24294 }, { "epoch": 2.44, "learning_rate": 1.925022758018242e-05, "loss": 0.1139, "step": 24295 }, { "epoch": 2.44, "learning_rate": 1.9247993359932543e-05, "loss": 0.2327, "step": 24296 }, { "epoch": 2.44, "learning_rate": 1.924575920810434e-05, "loss": 0.8466, "step": 24297 }, { "epoch": 2.44, "learning_rate": 1.9243525124712047e-05, "loss": 0.2007, "step": 24298 }, { "epoch": 2.44, "learning_rate": 1.9241291109769863e-05, "loss": 0.5224, "step": 24299 }, { "epoch": 2.44, "learning_rate": 1.9239057163292e-05, "loss": 0.2939, "step": 24300 }, { "epoch": 2.44, "learning_rate": 1.9236823285292692e-05, "loss": 0.1144, "step": 24301 }, { "epoch": 2.44, "learning_rate": 1.923458947578616e-05, "loss": 0.3832, "step": 24302 }, { "epoch": 2.44, "learning_rate": 1.9232355734786576e-05, "loss": 0.2172, "step": 24303 }, { "epoch": 2.44, "learning_rate": 1.9230122062308202e-05, "loss": 0.3891, "step": 24304 }, { "epoch": 2.44, "learning_rate": 1.9227888458365232e-05, "loss": 0.6536, "step": 24305 }, { "epoch": 2.44, "learning_rate": 1.9225654922971876e-05, "loss": 0.1875, "step": 24306 }, { "epoch": 2.44, "learning_rate": 1.9223421456142367e-05, "loss": 0.3572, "step": 24307 }, { "epoch": 2.44, "learning_rate": 1.9221188057890896e-05, "loss": 0.4812, "step": 24308 }, { "epoch": 2.44, "learning_rate": 1.9218954728231675e-05, "loss": 0.316, "step": 24309 }, { "epoch": 2.44, "learning_rate": 1.9216721467178934e-05, "loss": 0.2883, "step": 24310 }, { "epoch": 2.44, "learning_rate": 1.9214488274746884e-05, "loss": 0.7636, "step": 24311 }, { "epoch": 2.44, "learning_rate": 1.9212255150949703e-05, "loss": 0.2814, "step": 24312 }, { "epoch": 2.44, "learning_rate": 1.921002209580165e-05, "loss": 0.4505, "step": 24313 }, { "epoch": 2.44, "learning_rate": 1.9207789109316902e-05, "loss": 0.408, "step": 24314 }, { "epoch": 2.44, "learning_rate": 1.9205556191509675e-05, "loss": 0.6182, "step": 24315 }, { "epoch": 2.44, "learning_rate": 1.9203323342394193e-05, "loss": 0.1505, "step": 24316 }, { "epoch": 2.44, "learning_rate": 1.920109056198465e-05, "loss": 0.8015, "step": 24317 }, { "epoch": 2.44, "learning_rate": 1.919885785029525e-05, "loss": 0.6039, "step": 24318 }, { "epoch": 2.44, "learning_rate": 1.9196625207340218e-05, "loss": 0.2807, "step": 24319 }, { "epoch": 2.44, "learning_rate": 1.9194392633133745e-05, "loss": 0.0837, "step": 24320 }, { "epoch": 2.44, "learning_rate": 1.919216012769006e-05, "loss": 0.4958, "step": 24321 }, { "epoch": 2.44, "learning_rate": 1.9189927691023347e-05, "loss": 0.5042, "step": 24322 }, { "epoch": 2.44, "learning_rate": 1.918769532314782e-05, "loss": 0.4859, "step": 24323 }, { "epoch": 2.44, "learning_rate": 1.918546302407769e-05, "loss": 0.1667, "step": 24324 }, { "epoch": 2.44, "learning_rate": 1.9183230793827167e-05, "loss": 0.143, "step": 24325 }, { "epoch": 2.44, "learning_rate": 1.918099863241043e-05, "loss": 0.1293, "step": 24326 }, { "epoch": 2.44, "learning_rate": 1.917876653984172e-05, "loss": 0.1418, "step": 24327 }, { "epoch": 2.44, "learning_rate": 1.9176534516135218e-05, "loss": 0.1656, "step": 24328 }, { "epoch": 2.44, "learning_rate": 1.9174302561305125e-05, "loss": 0.4771, "step": 24329 }, { "epoch": 2.44, "learning_rate": 1.9172070675365664e-05, "loss": 0.2095, "step": 24330 }, { "epoch": 2.45, "learning_rate": 1.916983885833102e-05, "loss": 0.4309, "step": 24331 }, { "epoch": 2.45, "learning_rate": 1.9167607110215397e-05, "loss": 0.3779, "step": 24332 }, { "epoch": 2.45, "learning_rate": 1.916537543103301e-05, "loss": 0.6222, "step": 24333 }, { "epoch": 2.45, "learning_rate": 1.9163143820798054e-05, "loss": 0.3402, "step": 24334 }, { "epoch": 2.45, "learning_rate": 1.9160912279524712e-05, "loss": 0.2862, "step": 24335 }, { "epoch": 2.45, "learning_rate": 1.9158680807227215e-05, "loss": 0.4682, "step": 24336 }, { "epoch": 2.45, "learning_rate": 1.9156449403919745e-05, "loss": 0.0231, "step": 24337 }, { "epoch": 2.45, "learning_rate": 1.9154218069616502e-05, "loss": 0.6855, "step": 24338 }, { "epoch": 2.45, "learning_rate": 1.9151986804331706e-05, "loss": 0.7933, "step": 24339 }, { "epoch": 2.45, "learning_rate": 1.9149755608079523e-05, "loss": 0.2865, "step": 24340 }, { "epoch": 2.45, "learning_rate": 1.914752448087417e-05, "loss": 0.0819, "step": 24341 }, { "epoch": 2.45, "learning_rate": 1.9145293422729843e-05, "loss": 0.5828, "step": 24342 }, { "epoch": 2.45, "learning_rate": 1.9143062433660746e-05, "loss": 0.6326, "step": 24343 }, { "epoch": 2.45, "learning_rate": 1.9140831513681054e-05, "loss": 0.2995, "step": 24344 }, { "epoch": 2.45, "learning_rate": 1.9138600662804996e-05, "loss": 0.2092, "step": 24345 }, { "epoch": 2.45, "learning_rate": 1.9136369881046745e-05, "loss": 0.1989, "step": 24346 }, { "epoch": 2.45, "learning_rate": 1.9134139168420498e-05, "loss": 0.193, "step": 24347 }, { "epoch": 2.45, "learning_rate": 1.9131908524940467e-05, "loss": 0.343, "step": 24348 }, { "epoch": 2.45, "learning_rate": 1.9129677950620822e-05, "loss": 0.5613, "step": 24349 }, { "epoch": 2.45, "learning_rate": 1.9127447445475784e-05, "loss": 0.2089, "step": 24350 }, { "epoch": 2.45, "learning_rate": 1.9125217009519526e-05, "loss": 0.2058, "step": 24351 }, { "epoch": 2.45, "learning_rate": 1.912298664276625e-05, "loss": 0.8269, "step": 24352 }, { "epoch": 2.45, "learning_rate": 1.912075634523016e-05, "loss": 0.1419, "step": 24353 }, { "epoch": 2.45, "learning_rate": 1.9118526116925432e-05, "loss": 0.4659, "step": 24354 }, { "epoch": 2.45, "learning_rate": 1.9116295957866255e-05, "loss": 0.1435, "step": 24355 }, { "epoch": 2.45, "learning_rate": 1.9114065868066837e-05, "loss": 0.365, "step": 24356 }, { "epoch": 2.45, "learning_rate": 1.911183584754137e-05, "loss": 0.7017, "step": 24357 }, { "epoch": 2.45, "learning_rate": 1.9109605896304025e-05, "loss": 0.0864, "step": 24358 }, { "epoch": 2.45, "learning_rate": 1.9107376014369017e-05, "loss": 0.6316, "step": 24359 }, { "epoch": 2.45, "learning_rate": 1.9105146201750516e-05, "loss": 0.2194, "step": 24360 }, { "epoch": 2.45, "learning_rate": 1.9102916458462715e-05, "loss": 0.3391, "step": 24361 }, { "epoch": 2.45, "learning_rate": 1.9100686784519822e-05, "loss": 0.3651, "step": 24362 }, { "epoch": 2.45, "learning_rate": 1.9098457179936008e-05, "loss": 0.3616, "step": 24363 }, { "epoch": 2.45, "learning_rate": 1.909622764472546e-05, "loss": 0.3425, "step": 24364 }, { "epoch": 2.45, "learning_rate": 1.909399817890237e-05, "loss": 0.7678, "step": 24365 }, { "epoch": 2.45, "learning_rate": 1.9091768782480937e-05, "loss": 0.5316, "step": 24366 }, { "epoch": 2.45, "learning_rate": 1.9089539455475322e-05, "loss": 0.2571, "step": 24367 }, { "epoch": 2.45, "learning_rate": 1.9087310197899744e-05, "loss": 0.4956, "step": 24368 }, { "epoch": 2.45, "learning_rate": 1.9085081009768366e-05, "loss": 0.5871, "step": 24369 }, { "epoch": 2.45, "learning_rate": 1.9082851891095377e-05, "loss": 0.2875, "step": 24370 }, { "epoch": 2.45, "learning_rate": 1.9080622841894973e-05, "loss": 0.2209, "step": 24371 }, { "epoch": 2.45, "learning_rate": 1.9078393862181326e-05, "loss": 0.4135, "step": 24372 }, { "epoch": 2.45, "learning_rate": 1.9076164951968624e-05, "loss": 0.3773, "step": 24373 }, { "epoch": 2.45, "learning_rate": 1.907393611127105e-05, "loss": 0.4999, "step": 24374 }, { "epoch": 2.45, "learning_rate": 1.907170734010279e-05, "loss": 0.445, "step": 24375 }, { "epoch": 2.45, "learning_rate": 1.9069478638478037e-05, "loss": 0.6752, "step": 24376 }, { "epoch": 2.45, "learning_rate": 1.9067250006410965e-05, "loss": 0.1048, "step": 24377 }, { "epoch": 2.45, "learning_rate": 1.9065021443915743e-05, "loss": 0.4402, "step": 24378 }, { "epoch": 2.45, "learning_rate": 1.9062792951006568e-05, "loss": 0.4271, "step": 24379 }, { "epoch": 2.45, "learning_rate": 1.906056452769763e-05, "loss": 0.2067, "step": 24380 }, { "epoch": 2.45, "learning_rate": 1.9058336174003073e-05, "loss": 0.1494, "step": 24381 }, { "epoch": 2.45, "learning_rate": 1.9056107889937126e-05, "loss": 0.109, "step": 24382 }, { "epoch": 2.45, "learning_rate": 1.905387967551394e-05, "loss": 0.4197, "step": 24383 }, { "epoch": 2.45, "learning_rate": 1.9051651530747686e-05, "loss": 0.2853, "step": 24384 }, { "epoch": 2.45, "learning_rate": 1.9049423455652576e-05, "loss": 0.4804, "step": 24385 }, { "epoch": 2.45, "learning_rate": 1.9047195450242758e-05, "loss": 0.5384, "step": 24386 }, { "epoch": 2.45, "learning_rate": 1.9044967514532418e-05, "loss": 0.4382, "step": 24387 }, { "epoch": 2.45, "learning_rate": 1.9042739648535743e-05, "loss": 0.3733, "step": 24388 }, { "epoch": 2.45, "learning_rate": 1.904051185226691e-05, "loss": 0.0428, "step": 24389 }, { "epoch": 2.45, "learning_rate": 1.903828412574008e-05, "loss": 0.5287, "step": 24390 }, { "epoch": 2.45, "learning_rate": 1.9036056468969446e-05, "loss": 0.128, "step": 24391 }, { "epoch": 2.45, "learning_rate": 1.9033828881969177e-05, "loss": 0.3845, "step": 24392 }, { "epoch": 2.45, "learning_rate": 1.9031601364753443e-05, "loss": 0.1757, "step": 24393 }, { "epoch": 2.45, "learning_rate": 1.9029373917336435e-05, "loss": 0.3799, "step": 24394 }, { "epoch": 2.45, "learning_rate": 1.902714653973232e-05, "loss": 0.6006, "step": 24395 }, { "epoch": 2.45, "learning_rate": 1.9024919231955253e-05, "loss": 0.2034, "step": 24396 }, { "epoch": 2.45, "learning_rate": 1.9022691994019437e-05, "loss": 0.1666, "step": 24397 }, { "epoch": 2.45, "learning_rate": 1.902046482593904e-05, "loss": 0.1864, "step": 24398 }, { "epoch": 2.45, "learning_rate": 1.9018237727728206e-05, "loss": 0.395, "step": 24399 }, { "epoch": 2.45, "learning_rate": 1.901601069940115e-05, "loss": 0.5128, "step": 24400 }, { "epoch": 2.45, "learning_rate": 1.9013783740972013e-05, "loss": 0.1055, "step": 24401 }, { "epoch": 2.45, "learning_rate": 1.901155685245498e-05, "loss": 0.1282, "step": 24402 }, { "epoch": 2.45, "learning_rate": 1.9009330033864228e-05, "loss": 0.5714, "step": 24403 }, { "epoch": 2.45, "learning_rate": 1.90071032852139e-05, "loss": 0.4181, "step": 24404 }, { "epoch": 2.45, "learning_rate": 1.90048766065182e-05, "loss": 0.1715, "step": 24405 }, { "epoch": 2.45, "learning_rate": 1.9002649997791277e-05, "loss": 0.0725, "step": 24406 }, { "epoch": 2.45, "learning_rate": 1.90004234590473e-05, "loss": 0.7826, "step": 24407 }, { "epoch": 2.45, "learning_rate": 1.8998196990300452e-05, "loss": 0.1242, "step": 24408 }, { "epoch": 2.45, "learning_rate": 1.89959705915649e-05, "loss": 0.4161, "step": 24409 }, { "epoch": 2.45, "learning_rate": 1.8993744262854786e-05, "loss": 0.0765, "step": 24410 }, { "epoch": 2.45, "learning_rate": 1.8991518004184307e-05, "loss": 0.2491, "step": 24411 }, { "epoch": 2.45, "learning_rate": 1.8989291815567628e-05, "loss": 0.567, "step": 24412 }, { "epoch": 2.45, "learning_rate": 1.8987065697018887e-05, "loss": 0.166, "step": 24413 }, { "epoch": 2.45, "learning_rate": 1.898483964855229e-05, "loss": 0.1234, "step": 24414 }, { "epoch": 2.45, "learning_rate": 1.898261367018197e-05, "loss": 0.1978, "step": 24415 }, { "epoch": 2.45, "learning_rate": 1.8980387761922108e-05, "loss": 0.2536, "step": 24416 }, { "epoch": 2.45, "learning_rate": 1.8978161923786876e-05, "loss": 0.2671, "step": 24417 }, { "epoch": 2.45, "learning_rate": 1.897593615579042e-05, "loss": 0.3919, "step": 24418 }, { "epoch": 2.45, "learning_rate": 1.8973710457946908e-05, "loss": 0.1616, "step": 24419 }, { "epoch": 2.45, "learning_rate": 1.8971484830270513e-05, "loss": 0.3884, "step": 24420 }, { "epoch": 2.45, "learning_rate": 1.89692592727754e-05, "loss": 0.1318, "step": 24421 }, { "epoch": 2.45, "learning_rate": 1.8967033785475705e-05, "loss": 0.1005, "step": 24422 }, { "epoch": 2.45, "learning_rate": 1.8964808368385634e-05, "loss": 0.587, "step": 24423 }, { "epoch": 2.45, "learning_rate": 1.8962583021519313e-05, "loss": 0.4656, "step": 24424 }, { "epoch": 2.45, "learning_rate": 1.896035774489091e-05, "loss": 0.1256, "step": 24425 }, { "epoch": 2.45, "learning_rate": 1.89581325385146e-05, "loss": 0.2311, "step": 24426 }, { "epoch": 2.45, "learning_rate": 1.895590740240453e-05, "loss": 0.1423, "step": 24427 }, { "epoch": 2.45, "learning_rate": 1.895368233657486e-05, "loss": 0.2794, "step": 24428 }, { "epoch": 2.46, "learning_rate": 1.8951457341039756e-05, "loss": 0.465, "step": 24429 }, { "epoch": 2.46, "learning_rate": 1.8949232415813372e-05, "loss": 0.578, "step": 24430 }, { "epoch": 2.46, "learning_rate": 1.8947007560909872e-05, "loss": 0.3678, "step": 24431 }, { "epoch": 2.46, "learning_rate": 1.8944782776343422e-05, "loss": 0.0655, "step": 24432 }, { "epoch": 2.46, "learning_rate": 1.894255806212815e-05, "loss": 0.1154, "step": 24433 }, { "epoch": 2.46, "learning_rate": 1.894033341827824e-05, "loss": 0.2175, "step": 24434 }, { "epoch": 2.46, "learning_rate": 1.8938108844807852e-05, "loss": 0.2517, "step": 24435 }, { "epoch": 2.46, "learning_rate": 1.8935884341731108e-05, "loss": 0.4091, "step": 24436 }, { "epoch": 2.46, "learning_rate": 1.893365990906221e-05, "loss": 0.4768, "step": 24437 }, { "epoch": 2.46, "learning_rate": 1.8931435546815277e-05, "loss": 0.1839, "step": 24438 }, { "epoch": 2.46, "learning_rate": 1.8929211255004478e-05, "loss": 0.1992, "step": 24439 }, { "epoch": 2.46, "learning_rate": 1.8926987033643967e-05, "loss": 0.7856, "step": 24440 }, { "epoch": 2.46, "learning_rate": 1.8924762882747907e-05, "loss": 1.4644, "step": 24441 }, { "epoch": 2.46, "learning_rate": 1.8922538802330432e-05, "loss": 0.2428, "step": 24442 }, { "epoch": 2.46, "learning_rate": 1.892031479240571e-05, "loss": 0.4288, "step": 24443 }, { "epoch": 2.46, "learning_rate": 1.8918090852987898e-05, "loss": 0.1218, "step": 24444 }, { "epoch": 2.46, "learning_rate": 1.891586698409112e-05, "loss": 0.4473, "step": 24445 }, { "epoch": 2.46, "learning_rate": 1.8913643185729567e-05, "loss": 0.6313, "step": 24446 }, { "epoch": 2.46, "learning_rate": 1.8911419457917367e-05, "loss": 0.1743, "step": 24447 }, { "epoch": 2.46, "learning_rate": 1.8909195800668666e-05, "loss": 0.258, "step": 24448 }, { "epoch": 2.46, "learning_rate": 1.8906972213997637e-05, "loss": 0.264, "step": 24449 }, { "epoch": 2.46, "learning_rate": 1.890474869791841e-05, "loss": 0.1649, "step": 24450 }, { "epoch": 2.46, "learning_rate": 1.8902525252445135e-05, "loss": 0.407, "step": 24451 }, { "epoch": 2.46, "learning_rate": 1.8900301877591974e-05, "loss": 0.523, "step": 24452 }, { "epoch": 2.46, "learning_rate": 1.889807857337308e-05, "loss": 0.6508, "step": 24453 }, { "epoch": 2.46, "learning_rate": 1.8895855339802568e-05, "loss": 0.0338, "step": 24454 }, { "epoch": 2.46, "learning_rate": 1.8893632176894627e-05, "loss": 1.1526, "step": 24455 }, { "epoch": 2.46, "learning_rate": 1.8891409084663377e-05, "loss": 0.7427, "step": 24456 }, { "epoch": 2.46, "learning_rate": 1.8889186063122973e-05, "loss": 0.7938, "step": 24457 }, { "epoch": 2.46, "learning_rate": 1.8886963112287573e-05, "loss": 0.082, "step": 24458 }, { "epoch": 2.46, "learning_rate": 1.8884740232171292e-05, "loss": 0.1457, "step": 24459 }, { "epoch": 2.46, "learning_rate": 1.8882517422788314e-05, "loss": 0.2499, "step": 24460 }, { "epoch": 2.46, "learning_rate": 1.888029468415276e-05, "loss": 0.1183, "step": 24461 }, { "epoch": 2.46, "learning_rate": 1.8878072016278772e-05, "loss": 0.1531, "step": 24462 }, { "epoch": 2.46, "learning_rate": 1.8875849419180512e-05, "loss": 0.182, "step": 24463 }, { "epoch": 2.46, "learning_rate": 1.8873626892872123e-05, "loss": 0.1466, "step": 24464 }, { "epoch": 2.46, "learning_rate": 1.8871404437367723e-05, "loss": 0.2929, "step": 24465 }, { "epoch": 2.46, "learning_rate": 1.886918205268148e-05, "loss": 0.6774, "step": 24466 }, { "epoch": 2.46, "learning_rate": 1.8866959738827536e-05, "loss": 0.4853, "step": 24467 }, { "epoch": 2.46, "learning_rate": 1.8864737495820003e-05, "loss": 0.1168, "step": 24468 }, { "epoch": 2.46, "learning_rate": 1.8862515323673063e-05, "loss": 0.1191, "step": 24469 }, { "epoch": 2.46, "learning_rate": 1.8860293222400834e-05, "loss": 0.2476, "step": 24470 }, { "epoch": 2.46, "learning_rate": 1.8858071192017455e-05, "loss": 0.3929, "step": 24471 }, { "epoch": 2.46, "learning_rate": 1.8855849232537077e-05, "loss": 0.3, "step": 24472 }, { "epoch": 2.46, "learning_rate": 1.8853627343973845e-05, "loss": 0.1993, "step": 24473 }, { "epoch": 2.46, "learning_rate": 1.885140552634187e-05, "loss": 0.2478, "step": 24474 }, { "epoch": 2.46, "learning_rate": 1.884918377965532e-05, "loss": 0.1958, "step": 24475 }, { "epoch": 2.46, "learning_rate": 1.884696210392833e-05, "loss": 0.5465, "step": 24476 }, { "epoch": 2.46, "learning_rate": 1.8844740499175004e-05, "loss": 0.3229, "step": 24477 }, { "epoch": 2.46, "learning_rate": 1.884251896540953e-05, "loss": 0.2136, "step": 24478 }, { "epoch": 2.46, "learning_rate": 1.8840297502646015e-05, "loss": 0.3988, "step": 24479 }, { "epoch": 2.46, "learning_rate": 1.8838076110898594e-05, "loss": 0.4461, "step": 24480 }, { "epoch": 2.46, "learning_rate": 1.8835854790181422e-05, "loss": 0.328, "step": 24481 }, { "epoch": 2.46, "learning_rate": 1.8833633540508606e-05, "loss": 0.3513, "step": 24482 }, { "epoch": 2.46, "learning_rate": 1.8831412361894315e-05, "loss": 0.2429, "step": 24483 }, { "epoch": 2.46, "learning_rate": 1.882919125435266e-05, "loss": 0.5197, "step": 24484 }, { "epoch": 2.46, "learning_rate": 1.882697021789778e-05, "loss": 0.3447, "step": 24485 }, { "epoch": 2.46, "learning_rate": 1.8824749252543812e-05, "loss": 0.4602, "step": 24486 }, { "epoch": 2.46, "learning_rate": 1.88225283583049e-05, "loss": 0.2184, "step": 24487 }, { "epoch": 2.46, "learning_rate": 1.8820307535195148e-05, "loss": 1.8296, "step": 24488 }, { "epoch": 2.46, "learning_rate": 1.8818086783228717e-05, "loss": 0.7772, "step": 24489 }, { "epoch": 2.46, "learning_rate": 1.8815866102419732e-05, "loss": 0.4024, "step": 24490 }, { "epoch": 2.46, "learning_rate": 1.8813645492782307e-05, "loss": 0.5904, "step": 24491 }, { "epoch": 2.46, "learning_rate": 1.88114249543306e-05, "loss": 0.6604, "step": 24492 }, { "epoch": 2.46, "learning_rate": 1.8809204487078717e-05, "loss": 0.1286, "step": 24493 }, { "epoch": 2.46, "learning_rate": 1.8806984091040796e-05, "loss": 0.094, "step": 24494 }, { "epoch": 2.46, "learning_rate": 1.880476376623098e-05, "loss": 0.1973, "step": 24495 }, { "epoch": 2.46, "learning_rate": 1.8802543512663397e-05, "loss": 0.126, "step": 24496 }, { "epoch": 2.46, "learning_rate": 1.880032333035215e-05, "loss": 1.3736, "step": 24497 }, { "epoch": 2.46, "learning_rate": 1.8798103219311393e-05, "loss": 0.2078, "step": 24498 }, { "epoch": 2.46, "learning_rate": 1.879588317955525e-05, "loss": 0.2602, "step": 24499 }, { "epoch": 2.46, "learning_rate": 1.879366321109783e-05, "loss": 0.2404, "step": 24500 }, { "epoch": 2.46, "learning_rate": 1.879144331395329e-05, "loss": 0.4397, "step": 24501 }, { "epoch": 2.46, "learning_rate": 1.8789223488135732e-05, "loss": 0.5022, "step": 24502 }, { "epoch": 2.46, "learning_rate": 1.8787003733659284e-05, "loss": 0.1917, "step": 24503 }, { "epoch": 2.46, "learning_rate": 1.8784784050538092e-05, "loss": 0.493, "step": 24504 }, { "epoch": 2.46, "learning_rate": 1.878256443878627e-05, "loss": 0.6994, "step": 24505 }, { "epoch": 2.46, "learning_rate": 1.8780344898417927e-05, "loss": 0.3802, "step": 24506 }, { "epoch": 2.46, "learning_rate": 1.8778125429447206e-05, "loss": 0.137, "step": 24507 }, { "epoch": 2.46, "learning_rate": 1.8775906031888233e-05, "loss": 0.4118, "step": 24508 }, { "epoch": 2.46, "learning_rate": 1.8773686705755107e-05, "loss": 0.5986, "step": 24509 }, { "epoch": 2.46, "learning_rate": 1.8771467451061986e-05, "loss": 0.6579, "step": 24510 }, { "epoch": 2.46, "learning_rate": 1.8769248267822966e-05, "loss": 0.2512, "step": 24511 }, { "epoch": 2.46, "learning_rate": 1.8767029156052178e-05, "loss": 0.3056, "step": 24512 }, { "epoch": 2.46, "learning_rate": 1.876481011576375e-05, "loss": 0.4763, "step": 24513 }, { "epoch": 2.46, "learning_rate": 1.8762591146971778e-05, "loss": 0.4242, "step": 24514 }, { "epoch": 2.46, "learning_rate": 1.8760372249690426e-05, "loss": 0.4752, "step": 24515 }, { "epoch": 2.46, "learning_rate": 1.8758153423933777e-05, "loss": 0.7003, "step": 24516 }, { "epoch": 2.46, "learning_rate": 1.875593466971596e-05, "loss": 0.4831, "step": 24517 }, { "epoch": 2.46, "learning_rate": 1.87537159870511e-05, "loss": 0.509, "step": 24518 }, { "epoch": 2.46, "learning_rate": 1.8751497375953317e-05, "loss": 0.143, "step": 24519 }, { "epoch": 2.46, "learning_rate": 1.874927883643672e-05, "loss": 0.3798, "step": 24520 }, { "epoch": 2.46, "learning_rate": 1.8747060368515435e-05, "loss": 0.6269, "step": 24521 }, { "epoch": 2.46, "learning_rate": 1.8744841972203583e-05, "loss": 0.3494, "step": 24522 }, { "epoch": 2.46, "learning_rate": 1.8742623647515257e-05, "loss": 0.2008, "step": 24523 }, { "epoch": 2.46, "learning_rate": 1.8740405394464614e-05, "loss": 0.6377, "step": 24524 }, { "epoch": 2.46, "learning_rate": 1.873818721306573e-05, "loss": 0.2281, "step": 24525 }, { "epoch": 2.46, "learning_rate": 1.8735969103332744e-05, "loss": 0.11, "step": 24526 }, { "epoch": 2.46, "learning_rate": 1.8733751065279768e-05, "loss": 0.3071, "step": 24527 }, { "epoch": 2.47, "learning_rate": 1.873153309892092e-05, "loss": 0.124, "step": 24528 }, { "epoch": 2.47, "learning_rate": 1.8729315204270294e-05, "loss": 0.1614, "step": 24529 }, { "epoch": 2.47, "learning_rate": 1.872709738134203e-05, "loss": 0.7233, "step": 24530 }, { "epoch": 2.47, "learning_rate": 1.8724879630150232e-05, "loss": 0.1938, "step": 24531 }, { "epoch": 2.47, "learning_rate": 1.8722661950708993e-05, "loss": 0.6839, "step": 24532 }, { "epoch": 2.47, "learning_rate": 1.872044434303246e-05, "loss": 0.5577, "step": 24533 }, { "epoch": 2.47, "learning_rate": 1.871822680713473e-05, "loss": 0.0926, "step": 24534 }, { "epoch": 2.47, "learning_rate": 1.8716009343029898e-05, "loss": 0.4602, "step": 24535 }, { "epoch": 2.47, "learning_rate": 1.8713791950732097e-05, "loss": 0.1469, "step": 24536 }, { "epoch": 2.47, "learning_rate": 1.8711574630255428e-05, "loss": 0.0927, "step": 24537 }, { "epoch": 2.47, "learning_rate": 1.8709357381614014e-05, "loss": 0.4549, "step": 24538 }, { "epoch": 2.47, "learning_rate": 1.870714020482195e-05, "loss": 0.3506, "step": 24539 }, { "epoch": 2.47, "learning_rate": 1.8704923099893344e-05, "loss": 0.3679, "step": 24540 }, { "epoch": 2.47, "learning_rate": 1.8702706066842314e-05, "loss": 0.3596, "step": 24541 }, { "epoch": 2.47, "learning_rate": 1.8700489105682967e-05, "loss": 0.1164, "step": 24542 }, { "epoch": 2.47, "learning_rate": 1.8698272216429404e-05, "loss": 0.1254, "step": 24543 }, { "epoch": 2.47, "learning_rate": 1.869605539909574e-05, "loss": 0.0822, "step": 24544 }, { "epoch": 2.47, "learning_rate": 1.8693838653696086e-05, "loss": 0.4449, "step": 24545 }, { "epoch": 2.47, "learning_rate": 1.869162198024452e-05, "loss": 0.4039, "step": 24546 }, { "epoch": 2.47, "learning_rate": 1.8689405378755195e-05, "loss": 0.0973, "step": 24547 }, { "epoch": 2.47, "learning_rate": 1.8687188849242182e-05, "loss": 0.3308, "step": 24548 }, { "epoch": 2.47, "learning_rate": 1.868497239171959e-05, "loss": 0.7979, "step": 24549 }, { "epoch": 2.47, "learning_rate": 1.8682756006201535e-05, "loss": 0.2735, "step": 24550 }, { "epoch": 2.47, "learning_rate": 1.868053969270212e-05, "loss": 0.1354, "step": 24551 }, { "epoch": 2.47, "learning_rate": 1.867832345123543e-05, "loss": 0.3182, "step": 24552 }, { "epoch": 2.47, "learning_rate": 1.8676107281815595e-05, "loss": 0.2248, "step": 24553 }, { "epoch": 2.47, "learning_rate": 1.8673891184456708e-05, "loss": 0.2313, "step": 24554 }, { "epoch": 2.47, "learning_rate": 1.867167515917285e-05, "loss": 0.2601, "step": 24555 }, { "epoch": 2.47, "learning_rate": 1.866945920597816e-05, "loss": 0.6244, "step": 24556 }, { "epoch": 2.47, "learning_rate": 1.8667243324886715e-05, "loss": 0.4711, "step": 24557 }, { "epoch": 2.47, "learning_rate": 1.866502751591262e-05, "loss": 0.3123, "step": 24558 }, { "epoch": 2.47, "learning_rate": 1.8662811779069982e-05, "loss": 0.1287, "step": 24559 }, { "epoch": 2.47, "learning_rate": 1.8660596114372902e-05, "loss": 0.0967, "step": 24560 }, { "epoch": 2.47, "learning_rate": 1.8658380521835463e-05, "loss": 0.2489, "step": 24561 }, { "epoch": 2.47, "learning_rate": 1.8656165001471784e-05, "loss": 0.458, "step": 24562 }, { "epoch": 2.47, "learning_rate": 1.8653949553295943e-05, "loss": 0.145, "step": 24563 }, { "epoch": 2.47, "learning_rate": 1.865173417732206e-05, "loss": 0.3607, "step": 24564 }, { "epoch": 2.47, "learning_rate": 1.8649518873564232e-05, "loss": 0.2363, "step": 24565 }, { "epoch": 2.47, "learning_rate": 1.8647303642036532e-05, "loss": 0.2493, "step": 24566 }, { "epoch": 2.47, "learning_rate": 1.864508848275308e-05, "loss": 0.2381, "step": 24567 }, { "epoch": 2.47, "learning_rate": 1.8642873395727972e-05, "loss": 0.0637, "step": 24568 }, { "epoch": 2.47, "learning_rate": 1.8640658380975285e-05, "loss": 0.1212, "step": 24569 }, { "epoch": 2.47, "learning_rate": 1.863844343850914e-05, "loss": 0.3159, "step": 24570 }, { "epoch": 2.47, "learning_rate": 1.8636228568343613e-05, "loss": 0.3677, "step": 24571 }, { "epoch": 2.47, "learning_rate": 1.8634013770492797e-05, "loss": 0.1104, "step": 24572 }, { "epoch": 2.47, "learning_rate": 1.8631799044970796e-05, "loss": 0.0562, "step": 24573 }, { "epoch": 2.47, "learning_rate": 1.8629584391791713e-05, "loss": 0.7584, "step": 24574 }, { "epoch": 2.47, "learning_rate": 1.8627369810969617e-05, "loss": 0.4929, "step": 24575 }, { "epoch": 2.47, "learning_rate": 1.862515530251862e-05, "loss": 0.3754, "step": 24576 }, { "epoch": 2.47, "learning_rate": 1.8622940866452808e-05, "loss": 0.5407, "step": 24577 }, { "epoch": 2.47, "learning_rate": 1.862072650278626e-05, "loss": 0.4016, "step": 24578 }, { "epoch": 2.47, "learning_rate": 1.8618512211533096e-05, "loss": 0.5263, "step": 24579 }, { "epoch": 2.47, "learning_rate": 1.8616297992707383e-05, "loss": 0.2584, "step": 24580 }, { "epoch": 2.47, "learning_rate": 1.8614083846323214e-05, "loss": 0.5009, "step": 24581 }, { "epoch": 2.47, "learning_rate": 1.861186977239469e-05, "loss": 0.2467, "step": 24582 }, { "epoch": 2.47, "learning_rate": 1.86096557709359e-05, "loss": 0.1969, "step": 24583 }, { "epoch": 2.47, "learning_rate": 1.8607441841960914e-05, "loss": 0.3889, "step": 24584 }, { "epoch": 2.47, "learning_rate": 1.860522798548384e-05, "loss": 0.189, "step": 24585 }, { "epoch": 2.47, "learning_rate": 1.860301420151877e-05, "loss": 0.1448, "step": 24586 }, { "epoch": 2.47, "learning_rate": 1.8600800490079765e-05, "loss": 0.4835, "step": 24587 }, { "epoch": 2.47, "learning_rate": 1.8598586851180946e-05, "loss": 0.4604, "step": 24588 }, { "epoch": 2.47, "learning_rate": 1.8596373284836373e-05, "loss": 0.6585, "step": 24589 }, { "epoch": 2.47, "learning_rate": 1.8594159791060143e-05, "loss": 0.3626, "step": 24590 }, { "epoch": 2.47, "learning_rate": 1.8591946369866338e-05, "loss": 0.6435, "step": 24591 }, { "epoch": 2.47, "learning_rate": 1.8589733021269044e-05, "loss": 0.2917, "step": 24592 }, { "epoch": 2.47, "learning_rate": 1.8587519745282365e-05, "loss": 0.4393, "step": 24593 }, { "epoch": 2.47, "learning_rate": 1.858530654192036e-05, "loss": 0.7158, "step": 24594 }, { "epoch": 2.47, "learning_rate": 1.8583093411197114e-05, "loss": 0.4761, "step": 24595 }, { "epoch": 2.47, "learning_rate": 1.8580880353126723e-05, "loss": 0.263, "step": 24596 }, { "epoch": 2.47, "learning_rate": 1.8578667367723276e-05, "loss": 0.1144, "step": 24597 }, { "epoch": 2.47, "learning_rate": 1.8576454455000832e-05, "loss": 0.5598, "step": 24598 }, { "epoch": 2.47, "learning_rate": 1.8574241614973487e-05, "loss": 1.0787, "step": 24599 }, { "epoch": 2.47, "learning_rate": 1.857202884765533e-05, "loss": 0.6655, "step": 24600 }, { "epoch": 2.47, "learning_rate": 1.8569816153060422e-05, "loss": 0.4042, "step": 24601 }, { "epoch": 2.47, "learning_rate": 1.856760353120287e-05, "loss": 0.573, "step": 24602 }, { "epoch": 2.47, "learning_rate": 1.856539098209673e-05, "loss": 0.1498, "step": 24603 }, { "epoch": 2.47, "learning_rate": 1.856317850575609e-05, "loss": 0.4044, "step": 24604 }, { "epoch": 2.47, "learning_rate": 1.8560966102195038e-05, "loss": 0.2068, "step": 24605 }, { "epoch": 2.47, "learning_rate": 1.8558753771427654e-05, "loss": 0.3206, "step": 24606 }, { "epoch": 2.47, "learning_rate": 1.855654151346799e-05, "loss": 0.0605, "step": 24607 }, { "epoch": 2.47, "learning_rate": 1.8554329328330156e-05, "loss": 0.2875, "step": 24608 }, { "epoch": 2.47, "learning_rate": 1.8552117216028216e-05, "loss": 0.0884, "step": 24609 }, { "epoch": 2.47, "learning_rate": 1.854990517657623e-05, "loss": 0.3695, "step": 24610 }, { "epoch": 2.47, "learning_rate": 1.854769320998831e-05, "loss": 0.5845, "step": 24611 }, { "epoch": 2.47, "learning_rate": 1.854548131627851e-05, "loss": 0.3664, "step": 24612 }, { "epoch": 2.47, "learning_rate": 1.8543269495460902e-05, "loss": 0.7529, "step": 24613 }, { "epoch": 2.47, "learning_rate": 1.854105774754957e-05, "loss": 0.3964, "step": 24614 }, { "epoch": 2.47, "learning_rate": 1.85388460725586e-05, "loss": 0.5842, "step": 24615 }, { "epoch": 2.47, "learning_rate": 1.853663447050204e-05, "loss": 0.7004, "step": 24616 }, { "epoch": 2.47, "learning_rate": 1.8534422941393983e-05, "loss": 0.3192, "step": 24617 }, { "epoch": 2.47, "learning_rate": 1.8532211485248488e-05, "loss": 1.0819, "step": 24618 }, { "epoch": 2.47, "learning_rate": 1.8530000102079644e-05, "loss": 0.197, "step": 24619 }, { "epoch": 2.47, "learning_rate": 1.852778879190152e-05, "loss": 0.4357, "step": 24620 }, { "epoch": 2.47, "learning_rate": 1.8525577554728177e-05, "loss": 0.0928, "step": 24621 }, { "epoch": 2.47, "learning_rate": 1.85233663905737e-05, "loss": 0.5565, "step": 24622 }, { "epoch": 2.47, "learning_rate": 1.8521155299452146e-05, "loss": 0.2816, "step": 24623 }, { "epoch": 2.47, "learning_rate": 1.851894428137759e-05, "loss": 0.2877, "step": 24624 }, { "epoch": 2.47, "learning_rate": 1.851673333636412e-05, "loss": 0.4219, "step": 24625 }, { "epoch": 2.48, "learning_rate": 1.8514522464425784e-05, "loss": 0.6624, "step": 24626 }, { "epoch": 2.48, "learning_rate": 1.8512311665576646e-05, "loss": 0.1884, "step": 24627 }, { "epoch": 2.48, "learning_rate": 1.85101009398308e-05, "loss": 0.0692, "step": 24628 }, { "epoch": 2.48, "learning_rate": 1.85078902872023e-05, "loss": 0.1208, "step": 24629 }, { "epoch": 2.48, "learning_rate": 1.8505679707705206e-05, "loss": 0.2458, "step": 24630 }, { "epoch": 2.48, "learning_rate": 1.85034692013536e-05, "loss": 0.6414, "step": 24631 }, { "epoch": 2.48, "learning_rate": 1.8501258768161542e-05, "loss": 0.4327, "step": 24632 }, { "epoch": 2.48, "learning_rate": 1.8499048408143095e-05, "loss": 0.6846, "step": 24633 }, { "epoch": 2.48, "learning_rate": 1.8496838121312336e-05, "loss": 0.4151, "step": 24634 }, { "epoch": 2.48, "learning_rate": 1.849462790768332e-05, "loss": 0.4999, "step": 24635 }, { "epoch": 2.48, "learning_rate": 1.849241776727011e-05, "loss": 0.3385, "step": 24636 }, { "epoch": 2.48, "learning_rate": 1.849020770008678e-05, "loss": 0.6496, "step": 24637 }, { "epoch": 2.48, "learning_rate": 1.8487997706147395e-05, "loss": 0.4998, "step": 24638 }, { "epoch": 2.48, "learning_rate": 1.8485787785466005e-05, "loss": 0.0001, "step": 24639 }, { "epoch": 2.48, "learning_rate": 1.848357793805668e-05, "loss": 0.7423, "step": 24640 }, { "epoch": 2.48, "learning_rate": 1.8481368163933498e-05, "loss": 0.2023, "step": 24641 }, { "epoch": 2.48, "learning_rate": 1.8479158463110483e-05, "loss": 0.285, "step": 24642 }, { "epoch": 2.48, "learning_rate": 1.8476948835601744e-05, "loss": 0.5394, "step": 24643 }, { "epoch": 2.48, "learning_rate": 1.847473928142131e-05, "loss": 0.6715, "step": 24644 }, { "epoch": 2.48, "learning_rate": 1.8472529800583242e-05, "loss": 0.4362, "step": 24645 }, { "epoch": 2.48, "learning_rate": 1.847032039310162e-05, "loss": 0.3129, "step": 24646 }, { "epoch": 2.48, "learning_rate": 1.8468111058990485e-05, "loss": 0.603, "step": 24647 }, { "epoch": 2.48, "learning_rate": 1.8465901798263916e-05, "loss": 0.5344, "step": 24648 }, { "epoch": 2.48, "learning_rate": 1.8463692610935956e-05, "loss": 0.186, "step": 24649 }, { "epoch": 2.48, "learning_rate": 1.846148349702066e-05, "loss": 0.3921, "step": 24650 }, { "epoch": 2.48, "learning_rate": 1.84592744565321e-05, "loss": 0.0832, "step": 24651 }, { "epoch": 2.48, "learning_rate": 1.8457065489484332e-05, "loss": 0.9396, "step": 24652 }, { "epoch": 2.48, "learning_rate": 1.84548565958914e-05, "loss": 0.5333, "step": 24653 }, { "epoch": 2.48, "learning_rate": 1.8452647775767365e-05, "loss": 0.1686, "step": 24654 }, { "epoch": 2.48, "learning_rate": 1.8450439029126297e-05, "loss": 0.4672, "step": 24655 }, { "epoch": 2.48, "learning_rate": 1.844823035598223e-05, "loss": 0.0861, "step": 24656 }, { "epoch": 2.48, "learning_rate": 1.8446021756349244e-05, "loss": 0.4738, "step": 24657 }, { "epoch": 2.48, "learning_rate": 1.8443813230241377e-05, "loss": 0.0654, "step": 24658 }, { "epoch": 2.48, "learning_rate": 1.8441604777672676e-05, "loss": 0.1547, "step": 24659 }, { "epoch": 2.48, "learning_rate": 1.8439396398657213e-05, "loss": 0.3129, "step": 24660 }, { "epoch": 2.48, "learning_rate": 1.8437188093209038e-05, "loss": 0.4561, "step": 24661 }, { "epoch": 2.48, "learning_rate": 1.8434979861342188e-05, "loss": 0.6531, "step": 24662 }, { "epoch": 2.48, "learning_rate": 1.8432771703070734e-05, "loss": 0.2236, "step": 24663 }, { "epoch": 2.48, "learning_rate": 1.843056361840872e-05, "loss": 0.5098, "step": 24664 }, { "epoch": 2.48, "learning_rate": 1.842835560737019e-05, "loss": 0.36, "step": 24665 }, { "epoch": 2.48, "learning_rate": 1.8426147669969215e-05, "loss": 0.3132, "step": 24666 }, { "epoch": 2.48, "learning_rate": 1.842393980621983e-05, "loss": 0.1664, "step": 24667 }, { "epoch": 2.48, "learning_rate": 1.8421732016136084e-05, "loss": 0.2485, "step": 24668 }, { "epoch": 2.48, "learning_rate": 1.841952429973203e-05, "loss": 0.4844, "step": 24669 }, { "epoch": 2.48, "learning_rate": 1.8417316657021727e-05, "loss": 0.5503, "step": 24670 }, { "epoch": 2.48, "learning_rate": 1.8415109088019206e-05, "loss": 0.888, "step": 24671 }, { "epoch": 2.48, "learning_rate": 1.8412901592738526e-05, "loss": 0.2457, "step": 24672 }, { "epoch": 2.48, "learning_rate": 1.8410694171193724e-05, "loss": 0.3474, "step": 24673 }, { "epoch": 2.48, "learning_rate": 1.8408486823398865e-05, "loss": 0.3349, "step": 24674 }, { "epoch": 2.48, "learning_rate": 1.840627954936799e-05, "loss": 0.2461, "step": 24675 }, { "epoch": 2.48, "learning_rate": 1.8404072349115138e-05, "loss": 0.4695, "step": 24676 }, { "epoch": 2.48, "learning_rate": 1.8401865222654355e-05, "loss": 0.3633, "step": 24677 }, { "epoch": 2.48, "learning_rate": 1.8399658169999693e-05, "loss": 0.289, "step": 24678 }, { "epoch": 2.48, "learning_rate": 1.8397451191165186e-05, "loss": 0.4009, "step": 24679 }, { "epoch": 2.48, "learning_rate": 1.8395244286164897e-05, "loss": 0.8786, "step": 24680 }, { "epoch": 2.48, "learning_rate": 1.8393037455012855e-05, "loss": 0.5093, "step": 24681 }, { "epoch": 2.48, "learning_rate": 1.8390830697723098e-05, "loss": 0.5535, "step": 24682 }, { "epoch": 2.48, "learning_rate": 1.8388624014309684e-05, "loss": 0.1899, "step": 24683 }, { "epoch": 2.48, "learning_rate": 1.838641740478666e-05, "loss": 0.3012, "step": 24684 }, { "epoch": 2.48, "learning_rate": 1.8384210869168046e-05, "loss": 0.3324, "step": 24685 }, { "epoch": 2.48, "learning_rate": 1.83820044074679e-05, "loss": 0.0518, "step": 24686 }, { "epoch": 2.48, "learning_rate": 1.8379798019700256e-05, "loss": 0.25, "step": 24687 }, { "epoch": 2.48, "learning_rate": 1.837759170587915e-05, "loss": 0.4156, "step": 24688 }, { "epoch": 2.48, "learning_rate": 1.837538546601864e-05, "loss": 0.2973, "step": 24689 }, { "epoch": 2.48, "learning_rate": 1.8373179300132753e-05, "loss": 0.2089, "step": 24690 }, { "epoch": 2.48, "learning_rate": 1.8370973208235522e-05, "loss": 0.1585, "step": 24691 }, { "epoch": 2.48, "learning_rate": 1.8368767190340998e-05, "loss": 0.2648, "step": 24692 }, { "epoch": 2.48, "learning_rate": 1.8366561246463228e-05, "loss": 0.3936, "step": 24693 }, { "epoch": 2.48, "learning_rate": 1.836435537661622e-05, "loss": 0.1711, "step": 24694 }, { "epoch": 2.48, "learning_rate": 1.836214958081403e-05, "loss": 0.3671, "step": 24695 }, { "epoch": 2.48, "learning_rate": 1.8359943859070696e-05, "loss": 0.1536, "step": 24696 }, { "epoch": 2.48, "learning_rate": 1.8357738211400246e-05, "loss": 0.2668, "step": 24697 }, { "epoch": 2.48, "learning_rate": 1.835553263781673e-05, "loss": 0.6869, "step": 24698 }, { "epoch": 2.48, "learning_rate": 1.8353327138334165e-05, "loss": 0.0053, "step": 24699 }, { "epoch": 2.48, "learning_rate": 1.8351121712966604e-05, "loss": 0.6722, "step": 24700 }, { "epoch": 2.48, "learning_rate": 1.834891636172807e-05, "loss": 0.4722, "step": 24701 }, { "epoch": 2.48, "learning_rate": 1.8346711084632593e-05, "loss": 0.4068, "step": 24702 }, { "epoch": 2.48, "learning_rate": 1.8344505881694227e-05, "loss": 0.1709, "step": 24703 }, { "epoch": 2.48, "learning_rate": 1.8342300752926983e-05, "loss": 1.1157, "step": 24704 }, { "epoch": 2.48, "learning_rate": 1.83400956983449e-05, "loss": 0.2475, "step": 24705 }, { "epoch": 2.48, "learning_rate": 1.8337890717962015e-05, "loss": 0.4372, "step": 24706 }, { "epoch": 2.48, "learning_rate": 1.8335685811792365e-05, "loss": 0.3136, "step": 24707 }, { "epoch": 2.48, "learning_rate": 1.833348097984996e-05, "loss": 0.3171, "step": 24708 }, { "epoch": 2.48, "learning_rate": 1.8331276222148854e-05, "loss": 0.533, "step": 24709 }, { "epoch": 2.48, "learning_rate": 1.8329071538703062e-05, "loss": 0.1495, "step": 24710 }, { "epoch": 2.48, "learning_rate": 1.8326866929526614e-05, "loss": 0.2753, "step": 24711 }, { "epoch": 2.48, "learning_rate": 1.832466239463356e-05, "loss": 0.2986, "step": 24712 }, { "epoch": 2.48, "learning_rate": 1.8322457934037903e-05, "loss": 0.2627, "step": 24713 }, { "epoch": 2.48, "learning_rate": 1.8320253547753672e-05, "loss": 1.1368, "step": 24714 }, { "epoch": 2.48, "learning_rate": 1.8318049235794915e-05, "loss": 0.3462, "step": 24715 }, { "epoch": 2.48, "learning_rate": 1.831584499817566e-05, "loss": 0.2966, "step": 24716 }, { "epoch": 2.48, "learning_rate": 1.83136408349099e-05, "loss": 0.5381, "step": 24717 }, { "epoch": 2.48, "learning_rate": 1.83114367460117e-05, "loss": 0.591, "step": 24718 }, { "epoch": 2.48, "learning_rate": 1.8309232731495063e-05, "loss": 0.3839, "step": 24719 }, { "epoch": 2.48, "learning_rate": 1.8307028791374015e-05, "loss": 0.2489, "step": 24720 }, { "epoch": 2.48, "learning_rate": 1.8304824925662603e-05, "loss": 0.5671, "step": 24721 }, { "epoch": 2.48, "learning_rate": 1.830262113437483e-05, "loss": 0.183, "step": 24722 }, { "epoch": 2.48, "learning_rate": 1.8300417417524717e-05, "loss": 0.8999, "step": 24723 }, { "epoch": 2.48, "learning_rate": 1.8298213775126304e-05, "loss": 0.2381, "step": 24724 }, { "epoch": 2.49, "learning_rate": 1.8296010207193613e-05, "loss": 0.6413, "step": 24725 }, { "epoch": 2.49, "learning_rate": 1.8293806713740652e-05, "loss": 0.4001, "step": 24726 }, { "epoch": 2.49, "learning_rate": 1.8291603294781456e-05, "loss": 0.1257, "step": 24727 }, { "epoch": 2.49, "learning_rate": 1.8289399950330033e-05, "loss": 0.5421, "step": 24728 }, { "epoch": 2.49, "learning_rate": 1.8287196680400424e-05, "loss": 0.6223, "step": 24729 }, { "epoch": 2.49, "learning_rate": 1.8284993485006644e-05, "loss": 0.2739, "step": 24730 }, { "epoch": 2.49, "learning_rate": 1.8282790364162696e-05, "loss": 0.5019, "step": 24731 }, { "epoch": 2.49, "learning_rate": 1.828058731788262e-05, "loss": 0.2103, "step": 24732 }, { "epoch": 2.49, "learning_rate": 1.8278384346180428e-05, "loss": 0.1784, "step": 24733 }, { "epoch": 2.49, "learning_rate": 1.827618144907013e-05, "loss": 0.765, "step": 24734 }, { "epoch": 2.49, "learning_rate": 1.827397862656577e-05, "loss": 0.4974, "step": 24735 }, { "epoch": 2.49, "learning_rate": 1.8271775878681342e-05, "loss": 0.5528, "step": 24736 }, { "epoch": 2.49, "learning_rate": 1.826957320543086e-05, "loss": 0.337, "step": 24737 }, { "epoch": 2.49, "learning_rate": 1.8267370606828364e-05, "loss": 0.3404, "step": 24738 }, { "epoch": 2.49, "learning_rate": 1.826516808288786e-05, "loss": 0.4936, "step": 24739 }, { "epoch": 2.49, "learning_rate": 1.8262965633623354e-05, "loss": 0.0508, "step": 24740 }, { "epoch": 2.49, "learning_rate": 1.826076325904887e-05, "loss": 0.2967, "step": 24741 }, { "epoch": 2.49, "learning_rate": 1.825856095917843e-05, "loss": 0.6392, "step": 24742 }, { "epoch": 2.49, "learning_rate": 1.825635873402603e-05, "loss": 0.4877, "step": 24743 }, { "epoch": 2.49, "learning_rate": 1.825415658360571e-05, "loss": 0.1382, "step": 24744 }, { "epoch": 2.49, "learning_rate": 1.825195450793146e-05, "loss": 0.1953, "step": 24745 }, { "epoch": 2.49, "learning_rate": 1.82497525070173e-05, "loss": 0.0374, "step": 24746 }, { "epoch": 2.49, "learning_rate": 1.8247550580877246e-05, "loss": 0.2437, "step": 24747 }, { "epoch": 2.49, "learning_rate": 1.824534872952532e-05, "loss": 0.3493, "step": 24748 }, { "epoch": 2.49, "learning_rate": 1.8243146952975513e-05, "loss": 0.1503, "step": 24749 }, { "epoch": 2.49, "learning_rate": 1.8240945251241847e-05, "loss": 0.1686, "step": 24750 }, { "epoch": 2.49, "learning_rate": 1.8238743624338332e-05, "loss": 0.3463, "step": 24751 }, { "epoch": 2.49, "learning_rate": 1.8236542072278974e-05, "loss": 0.5335, "step": 24752 }, { "epoch": 2.49, "learning_rate": 1.8234340595077802e-05, "loss": 1.0155, "step": 24753 }, { "epoch": 2.49, "learning_rate": 1.82321391927488e-05, "loss": 0.5185, "step": 24754 }, { "epoch": 2.49, "learning_rate": 1.8229937865305983e-05, "loss": 0.3398, "step": 24755 }, { "epoch": 2.49, "learning_rate": 1.822773661276337e-05, "loss": 0.0739, "step": 24756 }, { "epoch": 2.49, "learning_rate": 1.822553543513496e-05, "loss": 0.3717, "step": 24757 }, { "epoch": 2.49, "learning_rate": 1.8223334332434772e-05, "loss": 0.2619, "step": 24758 }, { "epoch": 2.49, "learning_rate": 1.82211333046768e-05, "loss": 0.3242, "step": 24759 }, { "epoch": 2.49, "learning_rate": 1.8218932351875047e-05, "loss": 0.1594, "step": 24760 }, { "epoch": 2.49, "learning_rate": 1.8216731474043533e-05, "loss": 0.0792, "step": 24761 }, { "epoch": 2.49, "learning_rate": 1.8214530671196265e-05, "loss": 0.6632, "step": 24762 }, { "epoch": 2.49, "learning_rate": 1.8212329943347225e-05, "loss": 0.5031, "step": 24763 }, { "epoch": 2.49, "learning_rate": 1.821012929051044e-05, "loss": 0.0926, "step": 24764 }, { "epoch": 2.49, "learning_rate": 1.820792871269991e-05, "loss": 0.1398, "step": 24765 }, { "epoch": 2.49, "learning_rate": 1.820572820992963e-05, "loss": 0.3476, "step": 24766 }, { "epoch": 2.49, "learning_rate": 1.8203527782213612e-05, "loss": 0.2215, "step": 24767 }, { "epoch": 2.49, "learning_rate": 1.8201327429565863e-05, "loss": 0.5814, "step": 24768 }, { "epoch": 2.49, "learning_rate": 1.8199127152000364e-05, "loss": 0.1758, "step": 24769 }, { "epoch": 2.49, "learning_rate": 1.8196926949531133e-05, "loss": 0.1446, "step": 24770 }, { "epoch": 2.49, "learning_rate": 1.8194726822172183e-05, "loss": 0.7428, "step": 24771 }, { "epoch": 2.49, "learning_rate": 1.8192526769937486e-05, "loss": 0.2331, "step": 24772 }, { "epoch": 2.49, "learning_rate": 1.819032679284106e-05, "loss": 0.3988, "step": 24773 }, { "epoch": 2.49, "learning_rate": 1.81881268908969e-05, "loss": 0.8889, "step": 24774 }, { "epoch": 2.49, "learning_rate": 1.8185927064119007e-05, "loss": 0.2956, "step": 24775 }, { "epoch": 2.49, "learning_rate": 1.8183727312521392e-05, "loss": 0.2731, "step": 24776 }, { "epoch": 2.49, "learning_rate": 1.818152763611803e-05, "loss": 0.2349, "step": 24777 }, { "epoch": 2.49, "learning_rate": 1.817932803492293e-05, "loss": 0.7858, "step": 24778 }, { "epoch": 2.49, "learning_rate": 1.817712850895009e-05, "loss": 0.1905, "step": 24779 }, { "epoch": 2.49, "learning_rate": 1.8174929058213502e-05, "loss": 0.5214, "step": 24780 }, { "epoch": 2.49, "learning_rate": 1.817272968272718e-05, "loss": 0.5259, "step": 24781 }, { "epoch": 2.49, "learning_rate": 1.81705303825051e-05, "loss": 0.1548, "step": 24782 }, { "epoch": 2.49, "learning_rate": 1.816833115756126e-05, "loss": 0.2192, "step": 24783 }, { "epoch": 2.49, "learning_rate": 1.816613200790966e-05, "loss": 0.2116, "step": 24784 }, { "epoch": 2.49, "learning_rate": 1.8163932933564307e-05, "loss": 0.107, "step": 24785 }, { "epoch": 2.49, "learning_rate": 1.816173393453917e-05, "loss": 0.153, "step": 24786 }, { "epoch": 2.49, "learning_rate": 1.8159535010848256e-05, "loss": 0.2483, "step": 24787 }, { "epoch": 2.49, "learning_rate": 1.815733616250556e-05, "loss": 0.1673, "step": 24788 }, { "epoch": 2.49, "learning_rate": 1.8155137389525065e-05, "loss": 0.5439, "step": 24789 }, { "epoch": 2.49, "learning_rate": 1.8152938691920782e-05, "loss": 0.5098, "step": 24790 }, { "epoch": 2.49, "learning_rate": 1.8150740069706683e-05, "loss": 0.267, "step": 24791 }, { "epoch": 2.49, "learning_rate": 1.8148541522896765e-05, "loss": 0.4772, "step": 24792 }, { "epoch": 2.49, "learning_rate": 1.8146343051505025e-05, "loss": 0.3405, "step": 24793 }, { "epoch": 2.49, "learning_rate": 1.814414465554545e-05, "loss": 0.4802, "step": 24794 }, { "epoch": 2.49, "learning_rate": 1.814194633503202e-05, "loss": 0.1399, "step": 24795 }, { "epoch": 2.49, "learning_rate": 1.813974808997873e-05, "loss": 0.2545, "step": 24796 }, { "epoch": 2.49, "learning_rate": 1.813754992039958e-05, "loss": 1.0059, "step": 24797 }, { "epoch": 2.49, "learning_rate": 1.8135351826308543e-05, "loss": 0.0421, "step": 24798 }, { "epoch": 2.49, "learning_rate": 1.8133153807719623e-05, "loss": 0.9728, "step": 24799 }, { "epoch": 2.49, "learning_rate": 1.8130955864646797e-05, "loss": 0.8122, "step": 24800 }, { "epoch": 2.49, "learning_rate": 1.812875799710404e-05, "loss": 0.4147, "step": 24801 }, { "epoch": 2.49, "learning_rate": 1.812656020510536e-05, "loss": 0.1535, "step": 24802 }, { "epoch": 2.49, "learning_rate": 1.812436248866474e-05, "loss": 0.173, "step": 24803 }, { "epoch": 2.49, "learning_rate": 1.8122164847796147e-05, "loss": 0.3971, "step": 24804 }, { "epoch": 2.49, "learning_rate": 1.8119967282513585e-05, "loss": 0.7189, "step": 24805 }, { "epoch": 2.49, "learning_rate": 1.811776979283103e-05, "loss": 0.2966, "step": 24806 }, { "epoch": 2.49, "learning_rate": 1.8115572378762465e-05, "loss": 0.1105, "step": 24807 }, { "epoch": 2.49, "learning_rate": 1.8113375040321882e-05, "loss": 0.3731, "step": 24808 }, { "epoch": 2.49, "learning_rate": 1.8111177777523253e-05, "loss": 0.2765, "step": 24809 }, { "epoch": 2.49, "learning_rate": 1.810898059038057e-05, "loss": 1.0518, "step": 24810 }, { "epoch": 2.49, "learning_rate": 1.8106783478907808e-05, "loss": 0.3566, "step": 24811 }, { "epoch": 2.49, "learning_rate": 1.8104586443118952e-05, "loss": 0.3377, "step": 24812 }, { "epoch": 2.49, "learning_rate": 1.8102389483027987e-05, "loss": 0.5969, "step": 24813 }, { "epoch": 2.49, "learning_rate": 1.810019259864889e-05, "loss": 0.1718, "step": 24814 }, { "epoch": 2.49, "learning_rate": 1.8097995789995627e-05, "loss": 0.0661, "step": 24815 }, { "epoch": 2.49, "learning_rate": 1.8095799057082205e-05, "loss": 0.5384, "step": 24816 }, { "epoch": 2.49, "learning_rate": 1.8093602399922592e-05, "loss": 0.276, "step": 24817 }, { "epoch": 2.49, "learning_rate": 1.809140581853075e-05, "loss": 0.1288, "step": 24818 }, { "epoch": 2.49, "learning_rate": 1.8089209312920677e-05, "loss": 0.1835, "step": 24819 }, { "epoch": 2.49, "learning_rate": 1.808701288310635e-05, "loss": 0.7685, "step": 24820 }, { "epoch": 2.49, "learning_rate": 1.808481652910173e-05, "loss": 0.7925, "step": 24821 }, { "epoch": 2.49, "learning_rate": 1.808262025092082e-05, "loss": 0.6394, "step": 24822 }, { "epoch": 2.5, "learning_rate": 1.8080424048577573e-05, "loss": 0.0276, "step": 24823 }, { "epoch": 2.5, "learning_rate": 1.8078227922085965e-05, "loss": 0.5251, "step": 24824 }, { "epoch": 2.5, "learning_rate": 1.8076031871459988e-05, "loss": 0.1357, "step": 24825 }, { "epoch": 2.5, "learning_rate": 1.8073835896713614e-05, "loss": 0.3169, "step": 24826 }, { "epoch": 2.5, "learning_rate": 1.8071639997860796e-05, "loss": 0.3947, "step": 24827 }, { "epoch": 2.5, "learning_rate": 1.8069444174915536e-05, "loss": 0.0754, "step": 24828 }, { "epoch": 2.5, "learning_rate": 1.8067248427891792e-05, "loss": 0.0932, "step": 24829 }, { "epoch": 2.5, "learning_rate": 1.8065052756803528e-05, "loss": 0.2749, "step": 24830 }, { "epoch": 2.5, "learning_rate": 1.8062857161664747e-05, "loss": 0.2084, "step": 24831 }, { "epoch": 2.5, "learning_rate": 1.8060661642489393e-05, "loss": 0.0993, "step": 24832 }, { "epoch": 2.5, "learning_rate": 1.805846619929144e-05, "loss": 0.1783, "step": 24833 }, { "epoch": 2.5, "learning_rate": 1.8056270832084872e-05, "loss": 0.5559, "step": 24834 }, { "epoch": 2.5, "learning_rate": 1.8054075540883645e-05, "loss": 0.415, "step": 24835 }, { "epoch": 2.5, "learning_rate": 1.8051880325701754e-05, "loss": 0.5329, "step": 24836 }, { "epoch": 2.5, "learning_rate": 1.804968518655314e-05, "loss": 0.0341, "step": 24837 }, { "epoch": 2.5, "learning_rate": 1.804749012345178e-05, "loss": 0.3217, "step": 24838 }, { "epoch": 2.5, "learning_rate": 1.8045295136411652e-05, "loss": 0.1164, "step": 24839 }, { "epoch": 2.5, "learning_rate": 1.804310022544672e-05, "loss": 0.2368, "step": 24840 }, { "epoch": 2.5, "learning_rate": 1.804090539057094e-05, "loss": 0.3423, "step": 24841 }, { "epoch": 2.5, "learning_rate": 1.80387106317983e-05, "loss": 0.5077, "step": 24842 }, { "epoch": 2.5, "learning_rate": 1.803651594914275e-05, "loss": 0.7014, "step": 24843 }, { "epoch": 2.5, "learning_rate": 1.8034321342618253e-05, "loss": 0.2294, "step": 24844 }, { "epoch": 2.5, "learning_rate": 1.80321268122388e-05, "loss": 0.1456, "step": 24845 }, { "epoch": 2.5, "learning_rate": 1.8029932358018332e-05, "loss": 0.126, "step": 24846 }, { "epoch": 2.5, "learning_rate": 1.8027737979970812e-05, "loss": 0.3556, "step": 24847 }, { "epoch": 2.5, "learning_rate": 1.802554367811022e-05, "loss": 0.1185, "step": 24848 }, { "epoch": 2.5, "learning_rate": 1.802334945245052e-05, "loss": 0.3276, "step": 24849 }, { "epoch": 2.5, "learning_rate": 1.8021155303005657e-05, "loss": 0.3406, "step": 24850 }, { "epoch": 2.5, "learning_rate": 1.8018961229789608e-05, "loss": 0.1034, "step": 24851 }, { "epoch": 2.5, "learning_rate": 1.8016767232816333e-05, "loss": 0.1363, "step": 24852 }, { "epoch": 2.5, "learning_rate": 1.801457331209979e-05, "loss": 0.3782, "step": 24853 }, { "epoch": 2.5, "learning_rate": 1.8012379467653947e-05, "loss": 0.6373, "step": 24854 }, { "epoch": 2.5, "learning_rate": 1.8010185699492762e-05, "loss": 0.2224, "step": 24855 }, { "epoch": 2.5, "learning_rate": 1.8007992007630184e-05, "loss": 0.9601, "step": 24856 }, { "epoch": 2.5, "learning_rate": 1.8005798392080192e-05, "loss": 0.4678, "step": 24857 }, { "epoch": 2.5, "learning_rate": 1.800360485285674e-05, "loss": 0.6644, "step": 24858 }, { "epoch": 2.5, "learning_rate": 1.800141138997377e-05, "loss": 0.4689, "step": 24859 }, { "epoch": 2.5, "learning_rate": 1.7999218003445264e-05, "loss": 0.7767, "step": 24860 }, { "epoch": 2.5, "learning_rate": 1.799702469328516e-05, "loss": 0.366, "step": 24861 }, { "epoch": 2.5, "learning_rate": 1.799483145950743e-05, "loss": 0.7447, "step": 24862 }, { "epoch": 2.5, "learning_rate": 1.7992638302126032e-05, "loss": 0.4857, "step": 24863 }, { "epoch": 2.5, "learning_rate": 1.7990445221154908e-05, "loss": 0.2833, "step": 24864 }, { "epoch": 2.5, "learning_rate": 1.7988252216608023e-05, "loss": 0.5218, "step": 24865 }, { "epoch": 2.5, "learning_rate": 1.7986059288499332e-05, "loss": 1.0644, "step": 24866 }, { "epoch": 2.5, "learning_rate": 1.7983866436842785e-05, "loss": 0.3973, "step": 24867 }, { "epoch": 2.5, "learning_rate": 1.7981673661652354e-05, "loss": 0.7118, "step": 24868 }, { "epoch": 2.5, "learning_rate": 1.7979480962941967e-05, "loss": 0.5099, "step": 24869 }, { "epoch": 2.5, "learning_rate": 1.797728834072559e-05, "loss": 0.6238, "step": 24870 }, { "epoch": 2.5, "learning_rate": 1.7975095795017186e-05, "loss": 0.3036, "step": 24871 }, { "epoch": 2.5, "learning_rate": 1.7972903325830695e-05, "loss": 0.7206, "step": 24872 }, { "epoch": 2.5, "learning_rate": 1.7970710933180065e-05, "loss": 0.4294, "step": 24873 }, { "epoch": 2.5, "learning_rate": 1.7968518617079262e-05, "loss": 0.309, "step": 24874 }, { "epoch": 2.5, "learning_rate": 1.7966326377542224e-05, "loss": 0.0577, "step": 24875 }, { "epoch": 2.5, "learning_rate": 1.7964134214582908e-05, "loss": 0.6185, "step": 24876 }, { "epoch": 2.5, "learning_rate": 1.796194212821527e-05, "loss": 0.199, "step": 24877 }, { "epoch": 2.5, "learning_rate": 1.7959750118453245e-05, "loss": 0.2804, "step": 24878 }, { "epoch": 2.5, "learning_rate": 1.795755818531079e-05, "loss": 0.2946, "step": 24879 }, { "epoch": 2.5, "learning_rate": 1.7955366328801857e-05, "loss": 0.5652, "step": 24880 }, { "epoch": 2.5, "learning_rate": 1.7953174548940398e-05, "loss": 0.4848, "step": 24881 }, { "epoch": 2.5, "learning_rate": 1.795098284574034e-05, "loss": 0.7211, "step": 24882 }, { "epoch": 2.5, "learning_rate": 1.794879121921565e-05, "loss": 0.1327, "step": 24883 }, { "epoch": 2.5, "learning_rate": 1.7946599669380264e-05, "loss": 0.5141, "step": 24884 }, { "epoch": 2.5, "learning_rate": 1.794440819624813e-05, "loss": 0.1497, "step": 24885 }, { "epoch": 2.5, "learning_rate": 1.7942216799833208e-05, "loss": 0.1891, "step": 24886 }, { "epoch": 2.5, "learning_rate": 1.7940025480149428e-05, "loss": 0.3509, "step": 24887 }, { "epoch": 2.5, "learning_rate": 1.793783423721073e-05, "loss": 0.8178, "step": 24888 }, { "epoch": 2.5, "learning_rate": 1.793564307103107e-05, "loss": 0.1408, "step": 24889 }, { "epoch": 2.5, "learning_rate": 1.793345198162438e-05, "loss": 0.6522, "step": 24890 }, { "epoch": 2.5, "learning_rate": 1.793126096900463e-05, "loss": 0.4225, "step": 24891 }, { "epoch": 2.5, "learning_rate": 1.7929070033185737e-05, "loss": 0.1005, "step": 24892 }, { "epoch": 2.5, "learning_rate": 1.7926879174181638e-05, "loss": 0.1008, "step": 24893 }, { "epoch": 2.5, "learning_rate": 1.7924688392006298e-05, "loss": 0.0535, "step": 24894 }, { "epoch": 2.5, "learning_rate": 1.792249768667365e-05, "loss": 0.1916, "step": 24895 }, { "epoch": 2.5, "learning_rate": 1.7920307058197622e-05, "loss": 0.673, "step": 24896 }, { "epoch": 2.5, "learning_rate": 1.7918116506592172e-05, "loss": 0.283, "step": 24897 }, { "epoch": 2.5, "learning_rate": 1.791592603187123e-05, "loss": 0.2867, "step": 24898 }, { "epoch": 2.5, "learning_rate": 1.7913735634048735e-05, "loss": 0.1224, "step": 24899 }, { "epoch": 2.5, "learning_rate": 1.7911545313138637e-05, "loss": 0.4651, "step": 24900 }, { "epoch": 2.5, "learning_rate": 1.7909355069154865e-05, "loss": 0.1805, "step": 24901 }, { "epoch": 2.5, "learning_rate": 1.7907164902111344e-05, "loss": 0.0582, "step": 24902 }, { "epoch": 2.5, "learning_rate": 1.7904974812022037e-05, "loss": 0.5016, "step": 24903 }, { "epoch": 2.5, "learning_rate": 1.7902784798900876e-05, "loss": 0.0002, "step": 24904 }, { "epoch": 2.5, "learning_rate": 1.7900594862761776e-05, "loss": 0.9537, "step": 24905 }, { "epoch": 2.5, "learning_rate": 1.7898405003618696e-05, "loss": 0.4813, "step": 24906 }, { "epoch": 2.5, "learning_rate": 1.7896215221485557e-05, "loss": 1.1006, "step": 24907 }, { "epoch": 2.5, "learning_rate": 1.78940255163763e-05, "loss": 0.6575, "step": 24908 }, { "epoch": 2.5, "learning_rate": 1.7891835888304868e-05, "loss": 0.2467, "step": 24909 }, { "epoch": 2.5, "learning_rate": 1.788964633728518e-05, "loss": 0.4551, "step": 24910 }, { "epoch": 2.5, "learning_rate": 1.7887456863331173e-05, "loss": 0.6355, "step": 24911 }, { "epoch": 2.5, "learning_rate": 1.7885267466456785e-05, "loss": 0.6313, "step": 24912 }, { "epoch": 2.5, "learning_rate": 1.7883078146675954e-05, "loss": 0.2957, "step": 24913 }, { "epoch": 2.5, "learning_rate": 1.788088890400259e-05, "loss": 0.4476, "step": 24914 }, { "epoch": 2.5, "learning_rate": 1.7878699738450645e-05, "loss": 0.1322, "step": 24915 }, { "epoch": 2.5, "learning_rate": 1.787651065003404e-05, "loss": 0.139, "step": 24916 }, { "epoch": 2.5, "learning_rate": 1.787432163876672e-05, "loss": 0.1816, "step": 24917 }, { "epoch": 2.5, "learning_rate": 1.78721327046626e-05, "loss": 0.4786, "step": 24918 }, { "epoch": 2.5, "learning_rate": 1.7869943847735606e-05, "loss": 0.6127, "step": 24919 }, { "epoch": 2.5, "learning_rate": 1.7867755067999682e-05, "loss": 0.4816, "step": 24920 }, { "epoch": 2.5, "learning_rate": 1.7865566365468748e-05, "loss": 0.2867, "step": 24921 }, { "epoch": 2.51, "learning_rate": 1.7863377740156728e-05, "loss": 0.5397, "step": 24922 }, { "epoch": 2.51, "learning_rate": 1.786118919207757e-05, "loss": 0.3142, "step": 24923 }, { "epoch": 2.51, "learning_rate": 1.7859000721245176e-05, "loss": 0.7052, "step": 24924 }, { "epoch": 2.51, "learning_rate": 1.7856812327673477e-05, "loss": 0.3712, "step": 24925 }, { "epoch": 2.51, "learning_rate": 1.785462401137641e-05, "loss": 0.0818, "step": 24926 }, { "epoch": 2.51, "learning_rate": 1.7852435772367902e-05, "loss": 0.5953, "step": 24927 }, { "epoch": 2.51, "learning_rate": 1.7850247610661863e-05, "loss": 0.1386, "step": 24928 }, { "epoch": 2.51, "learning_rate": 1.784805952627223e-05, "loss": 0.1545, "step": 24929 }, { "epoch": 2.51, "learning_rate": 1.7845871519212923e-05, "loss": 0.1455, "step": 24930 }, { "epoch": 2.51, "learning_rate": 1.7843683589497862e-05, "loss": 0.4799, "step": 24931 }, { "epoch": 2.51, "learning_rate": 1.7841495737140985e-05, "loss": 0.0975, "step": 24932 }, { "epoch": 2.51, "learning_rate": 1.7839307962156197e-05, "loss": 0.8098, "step": 24933 }, { "epoch": 2.51, "learning_rate": 1.783712026455742e-05, "loss": 0.2811, "step": 24934 }, { "epoch": 2.51, "learning_rate": 1.7834932644358592e-05, "loss": 0.2651, "step": 24935 }, { "epoch": 2.51, "learning_rate": 1.783274510157363e-05, "loss": 0.1015, "step": 24936 }, { "epoch": 2.51, "learning_rate": 1.7830557636216432e-05, "loss": 0.6225, "step": 24937 }, { "epoch": 2.51, "learning_rate": 1.7828370248300947e-05, "loss": 0.3754, "step": 24938 }, { "epoch": 2.51, "learning_rate": 1.7826182937841082e-05, "loss": 0.1539, "step": 24939 }, { "epoch": 2.51, "learning_rate": 1.7823995704850753e-05, "loss": 0.3713, "step": 24940 }, { "epoch": 2.51, "learning_rate": 1.7821808549343893e-05, "loss": 0.3462, "step": 24941 }, { "epoch": 2.51, "learning_rate": 1.7819621471334403e-05, "loss": 0.1896, "step": 24942 }, { "epoch": 2.51, "learning_rate": 1.781743447083621e-05, "loss": 0.4567, "step": 24943 }, { "epoch": 2.51, "learning_rate": 1.7815247547863228e-05, "loss": 0.4244, "step": 24944 }, { "epoch": 2.51, "learning_rate": 1.7813060702429374e-05, "loss": 0.2266, "step": 24945 }, { "epoch": 2.51, "learning_rate": 1.7810873934548573e-05, "loss": 0.1163, "step": 24946 }, { "epoch": 2.51, "learning_rate": 1.7808687244234732e-05, "loss": 0.3766, "step": 24947 }, { "epoch": 2.51, "learning_rate": 1.7806500631501758e-05, "loss": 0.5961, "step": 24948 }, { "epoch": 2.51, "learning_rate": 1.7804314096363584e-05, "loss": 0.2762, "step": 24949 }, { "epoch": 2.51, "learning_rate": 1.780212763883412e-05, "loss": 0.1613, "step": 24950 }, { "epoch": 2.51, "learning_rate": 1.7799941258927266e-05, "loss": 0.2192, "step": 24951 }, { "epoch": 2.51, "learning_rate": 1.779775495665695e-05, "loss": 0.4461, "step": 24952 }, { "epoch": 2.51, "learning_rate": 1.7795568732037073e-05, "loss": 0.1749, "step": 24953 }, { "epoch": 2.51, "learning_rate": 1.7793382585081556e-05, "loss": 1.1029, "step": 24954 }, { "epoch": 2.51, "learning_rate": 1.7791196515804314e-05, "loss": 0.259, "step": 24955 }, { "epoch": 2.51, "learning_rate": 1.7789010524219248e-05, "loss": 0.2616, "step": 24956 }, { "epoch": 2.51, "learning_rate": 1.7786824610340272e-05, "loss": 0.3005, "step": 24957 }, { "epoch": 2.51, "learning_rate": 1.7784638774181303e-05, "loss": 0.5691, "step": 24958 }, { "epoch": 2.51, "learning_rate": 1.7782453015756252e-05, "loss": 0.6258, "step": 24959 }, { "epoch": 2.51, "learning_rate": 1.7780267335079007e-05, "loss": 0.0965, "step": 24960 }, { "epoch": 2.51, "learning_rate": 1.7778081732163502e-05, "loss": 0.0569, "step": 24961 }, { "epoch": 2.51, "learning_rate": 1.7775896207023634e-05, "loss": 0.2718, "step": 24962 }, { "epoch": 2.51, "learning_rate": 1.7773710759673306e-05, "loss": 0.3539, "step": 24963 }, { "epoch": 2.51, "learning_rate": 1.7771525390126444e-05, "loss": 0.19, "step": 24964 }, { "epoch": 2.51, "learning_rate": 1.7769340098396933e-05, "loss": 0.1136, "step": 24965 }, { "epoch": 2.51, "learning_rate": 1.7767154884498687e-05, "loss": 0.2337, "step": 24966 }, { "epoch": 2.51, "learning_rate": 1.7764969748445622e-05, "loss": 0.5073, "step": 24967 }, { "epoch": 2.51, "learning_rate": 1.7762784690251633e-05, "loss": 0.357, "step": 24968 }, { "epoch": 2.51, "learning_rate": 1.7760599709930617e-05, "loss": 0.0976, "step": 24969 }, { "epoch": 2.51, "learning_rate": 1.7758414807496503e-05, "loss": 0.081, "step": 24970 }, { "epoch": 2.51, "learning_rate": 1.7756229982963162e-05, "loss": 0.1216, "step": 24971 }, { "epoch": 2.51, "learning_rate": 1.775404523634453e-05, "loss": 0.3858, "step": 24972 }, { "epoch": 2.51, "learning_rate": 1.77518605676545e-05, "loss": 0.9113, "step": 24973 }, { "epoch": 2.51, "learning_rate": 1.774967597690695e-05, "loss": 0.1813, "step": 24974 }, { "epoch": 2.51, "learning_rate": 1.774749146411582e-05, "loss": 0.2338, "step": 24975 }, { "epoch": 2.51, "learning_rate": 1.7745307029294994e-05, "loss": 0.4673, "step": 24976 }, { "epoch": 2.51, "learning_rate": 1.7743122672458355e-05, "loss": 0.3198, "step": 24977 }, { "epoch": 2.51, "learning_rate": 1.774093839361984e-05, "loss": 0.5546, "step": 24978 }, { "epoch": 2.51, "learning_rate": 1.7738754192793323e-05, "loss": 0.2239, "step": 24979 }, { "epoch": 2.51, "learning_rate": 1.77365700699927e-05, "loss": 0.3323, "step": 24980 }, { "epoch": 2.51, "learning_rate": 1.7734386025231884e-05, "loss": 0.2159, "step": 24981 }, { "epoch": 2.51, "learning_rate": 1.7732202058524784e-05, "loss": 0.2018, "step": 24982 }, { "epoch": 2.51, "learning_rate": 1.7730018169885265e-05, "loss": 0.3035, "step": 24983 }, { "epoch": 2.51, "learning_rate": 1.772783435932725e-05, "loss": 0.1368, "step": 24984 }, { "epoch": 2.51, "learning_rate": 1.772565062686463e-05, "loss": 0.3202, "step": 24985 }, { "epoch": 2.51, "learning_rate": 1.7723466972511294e-05, "loss": 0.2151, "step": 24986 }, { "epoch": 2.51, "learning_rate": 1.7721283396281152e-05, "loss": 0.5207, "step": 24987 }, { "epoch": 2.51, "learning_rate": 1.7719099898188088e-05, "loss": 0.2832, "step": 24988 }, { "epoch": 2.51, "learning_rate": 1.7716916478245993e-05, "loss": 0.5232, "step": 24989 }, { "epoch": 2.51, "learning_rate": 1.7714733136468777e-05, "loss": 0.0565, "step": 24990 }, { "epoch": 2.51, "learning_rate": 1.7712549872870333e-05, "loss": 0.2549, "step": 24991 }, { "epoch": 2.51, "learning_rate": 1.7710366687464533e-05, "loss": 0.2191, "step": 24992 }, { "epoch": 2.51, "learning_rate": 1.7708183580265292e-05, "loss": 0.2523, "step": 24993 }, { "epoch": 2.51, "learning_rate": 1.770600055128649e-05, "loss": 0.2483, "step": 24994 }, { "epoch": 2.51, "learning_rate": 1.770381760054202e-05, "loss": 0.3048, "step": 24995 }, { "epoch": 2.51, "learning_rate": 1.7701634728045794e-05, "loss": 0.0577, "step": 24996 }, { "epoch": 2.51, "learning_rate": 1.7699451933811664e-05, "loss": 0.5666, "step": 24997 }, { "epoch": 2.51, "learning_rate": 1.7697269217853556e-05, "loss": 0.3296, "step": 24998 }, { "epoch": 2.51, "learning_rate": 1.7695086580185345e-05, "loss": 0.2144, "step": 24999 }, { "epoch": 2.51, "learning_rate": 1.769290402082092e-05, "loss": 0.225, "step": 25000 }, { "epoch": 2.51, "learning_rate": 1.7690721539774176e-05, "loss": 1.0568, "step": 25001 }, { "epoch": 2.51, "learning_rate": 1.7688539137058996e-05, "loss": 0.2254, "step": 25002 }, { "epoch": 2.51, "learning_rate": 1.768635681268926e-05, "loss": 0.2955, "step": 25003 }, { "epoch": 2.51, "learning_rate": 1.7684174566678877e-05, "loss": 0.3701, "step": 25004 }, { "epoch": 2.51, "learning_rate": 1.7681992399041723e-05, "loss": 0.2221, "step": 25005 }, { "epoch": 2.51, "learning_rate": 1.767981030979167e-05, "loss": 0.2649, "step": 25006 }, { "epoch": 2.51, "learning_rate": 1.7677628298942628e-05, "loss": 0.13, "step": 25007 }, { "epoch": 2.51, "learning_rate": 1.7675446366508472e-05, "loss": 0.0297, "step": 25008 }, { "epoch": 2.51, "learning_rate": 1.767326451250308e-05, "loss": 0.6792, "step": 25009 }, { "epoch": 2.51, "learning_rate": 1.7671082736940353e-05, "loss": 0.1406, "step": 25010 }, { "epoch": 2.51, "learning_rate": 1.766890103983416e-05, "loss": 0.2236, "step": 25011 }, { "epoch": 2.51, "learning_rate": 1.7666719421198382e-05, "loss": 0.4511, "step": 25012 }, { "epoch": 2.51, "learning_rate": 1.766453788104692e-05, "loss": 0.3538, "step": 25013 }, { "epoch": 2.51, "learning_rate": 1.766235641939365e-05, "loss": 0.4733, "step": 25014 }, { "epoch": 2.51, "learning_rate": 1.766017503625244e-05, "loss": 0.4079, "step": 25015 }, { "epoch": 2.51, "learning_rate": 1.7657993731637187e-05, "loss": 0.2934, "step": 25016 }, { "epoch": 2.51, "learning_rate": 1.7655812505561763e-05, "loss": 0.3868, "step": 25017 }, { "epoch": 2.51, "learning_rate": 1.765363135804005e-05, "loss": 0.1468, "step": 25018 }, { "epoch": 2.51, "learning_rate": 1.765145028908594e-05, "loss": 0.1918, "step": 25019 }, { "epoch": 2.52, "learning_rate": 1.7649269298713297e-05, "loss": 0.3837, "step": 25020 }, { "epoch": 2.52, "learning_rate": 1.7647088386936004e-05, "loss": 0.3579, "step": 25021 }, { "epoch": 2.52, "learning_rate": 1.7644907553767942e-05, "loss": 0.0177, "step": 25022 }, { "epoch": 2.52, "learning_rate": 1.7642726799222997e-05, "loss": 0.1991, "step": 25023 }, { "epoch": 2.52, "learning_rate": 1.7640546123315028e-05, "loss": 0.6243, "step": 25024 }, { "epoch": 2.52, "learning_rate": 1.763836552605792e-05, "loss": 0.3153, "step": 25025 }, { "epoch": 2.52, "learning_rate": 1.763618500746555e-05, "loss": 0.1038, "step": 25026 }, { "epoch": 2.52, "learning_rate": 1.7634004567551802e-05, "loss": 0.6461, "step": 25027 }, { "epoch": 2.52, "learning_rate": 1.763182420633055e-05, "loss": 0.8201, "step": 25028 }, { "epoch": 2.52, "learning_rate": 1.7629643923815646e-05, "loss": 0.1322, "step": 25029 }, { "epoch": 2.52, "learning_rate": 1.7627463720021e-05, "loss": 0.2225, "step": 25030 }, { "epoch": 2.52, "learning_rate": 1.762528359496046e-05, "loss": 0.1365, "step": 25031 }, { "epoch": 2.52, "learning_rate": 1.7623103548647904e-05, "loss": 0.5172, "step": 25032 }, { "epoch": 2.52, "learning_rate": 1.7620923581097218e-05, "loss": 0.6995, "step": 25033 }, { "epoch": 2.52, "learning_rate": 1.7618743692322263e-05, "loss": 0.3343, "step": 25034 }, { "epoch": 2.52, "learning_rate": 1.7616563882336907e-05, "loss": 0.4781, "step": 25035 }, { "epoch": 2.52, "learning_rate": 1.7614384151155033e-05, "loss": 0.6594, "step": 25036 }, { "epoch": 2.52, "learning_rate": 1.7612204498790508e-05, "loss": 0.9136, "step": 25037 }, { "epoch": 2.52, "learning_rate": 1.7610024925257183e-05, "loss": 0.7502, "step": 25038 }, { "epoch": 2.52, "learning_rate": 1.7607845430568972e-05, "loss": 0.2441, "step": 25039 }, { "epoch": 2.52, "learning_rate": 1.7605666014739705e-05, "loss": 0.0634, "step": 25040 }, { "epoch": 2.52, "learning_rate": 1.760348667778326e-05, "loss": 0.7719, "step": 25041 }, { "epoch": 2.52, "learning_rate": 1.7601307419713526e-05, "loss": 0.3631, "step": 25042 }, { "epoch": 2.52, "learning_rate": 1.7599128240544345e-05, "loss": 0.4023, "step": 25043 }, { "epoch": 2.52, "learning_rate": 1.759694914028959e-05, "loss": 0.4976, "step": 25044 }, { "epoch": 2.52, "learning_rate": 1.7594770118963136e-05, "loss": 0.2452, "step": 25045 }, { "epoch": 2.52, "learning_rate": 1.7592591176578855e-05, "loss": 0.1034, "step": 25046 }, { "epoch": 2.52, "learning_rate": 1.759041231315059e-05, "loss": 0.7767, "step": 25047 }, { "epoch": 2.52, "learning_rate": 1.7588233528692225e-05, "loss": 0.3148, "step": 25048 }, { "epoch": 2.52, "learning_rate": 1.758605482321762e-05, "loss": 0.3369, "step": 25049 }, { "epoch": 2.52, "learning_rate": 1.7583876196740638e-05, "loss": 0.4823, "step": 25050 }, { "epoch": 2.52, "learning_rate": 1.7581697649275152e-05, "loss": 0.1634, "step": 25051 }, { "epoch": 2.52, "learning_rate": 1.7579519180835e-05, "loss": 0.5038, "step": 25052 }, { "epoch": 2.52, "learning_rate": 1.7577340791434082e-05, "loss": 0.4816, "step": 25053 }, { "epoch": 2.52, "learning_rate": 1.757516248108624e-05, "loss": 0.2562, "step": 25054 }, { "epoch": 2.52, "learning_rate": 1.7572984249805326e-05, "loss": 0.741, "step": 25055 }, { "epoch": 2.52, "learning_rate": 1.7570806097605224e-05, "loss": 0.0442, "step": 25056 }, { "epoch": 2.52, "learning_rate": 1.756862802449978e-05, "loss": 0.5482, "step": 25057 }, { "epoch": 2.52, "learning_rate": 1.7566450030502854e-05, "loss": 0.0919, "step": 25058 }, { "epoch": 2.52, "learning_rate": 1.7564272115628312e-05, "loss": 0.5435, "step": 25059 }, { "epoch": 2.52, "learning_rate": 1.756209427989002e-05, "loss": 0.5316, "step": 25060 }, { "epoch": 2.52, "learning_rate": 1.755991652330181e-05, "loss": 0.4416, "step": 25061 }, { "epoch": 2.52, "learning_rate": 1.7557738845877574e-05, "loss": 0.3057, "step": 25062 }, { "epoch": 2.52, "learning_rate": 1.755556124763115e-05, "loss": 0.8799, "step": 25063 }, { "epoch": 2.52, "learning_rate": 1.75533837285764e-05, "loss": 0.0715, "step": 25064 }, { "epoch": 2.52, "learning_rate": 1.7551206288727188e-05, "loss": 0.5914, "step": 25065 }, { "epoch": 2.52, "learning_rate": 1.7549028928097355e-05, "loss": 0.1398, "step": 25066 }, { "epoch": 2.52, "learning_rate": 1.754685164670076e-05, "loss": 0.3438, "step": 25067 }, { "epoch": 2.52, "learning_rate": 1.7544674444551274e-05, "loss": 0.181, "step": 25068 }, { "epoch": 2.52, "learning_rate": 1.7542497321662746e-05, "loss": 0.2366, "step": 25069 }, { "epoch": 2.52, "learning_rate": 1.7540320278049007e-05, "loss": 0.799, "step": 25070 }, { "epoch": 2.52, "learning_rate": 1.753814331372395e-05, "loss": 0.2762, "step": 25071 }, { "epoch": 2.52, "learning_rate": 1.75359664287014e-05, "loss": 0.1264, "step": 25072 }, { "epoch": 2.52, "learning_rate": 1.753378962299521e-05, "loss": 0.7658, "step": 25073 }, { "epoch": 2.52, "learning_rate": 1.7531612896619255e-05, "loss": 0.1426, "step": 25074 }, { "epoch": 2.52, "learning_rate": 1.7529436249587366e-05, "loss": 0.6288, "step": 25075 }, { "epoch": 2.52, "learning_rate": 1.7527259681913393e-05, "loss": 0.3549, "step": 25076 }, { "epoch": 2.52, "learning_rate": 1.7525083193611203e-05, "loss": 0.127, "step": 25077 }, { "epoch": 2.52, "learning_rate": 1.7522906784694634e-05, "loss": 0.1553, "step": 25078 }, { "epoch": 2.52, "learning_rate": 1.7520730455177547e-05, "loss": 0.5043, "step": 25079 }, { "epoch": 2.52, "learning_rate": 1.7518554205073777e-05, "loss": 0.2965, "step": 25080 }, { "epoch": 2.52, "learning_rate": 1.7516378034397178e-05, "loss": 0.5278, "step": 25081 }, { "epoch": 2.52, "learning_rate": 1.7514201943161614e-05, "loss": 0.1699, "step": 25082 }, { "epoch": 2.52, "learning_rate": 1.7512025931380908e-05, "loss": 0.6242, "step": 25083 }, { "epoch": 2.52, "learning_rate": 1.7509849999068907e-05, "loss": 0.2671, "step": 25084 }, { "epoch": 2.52, "learning_rate": 1.7507674146239483e-05, "loss": 0.3709, "step": 25085 }, { "epoch": 2.52, "learning_rate": 1.750549837290648e-05, "loss": 0.2093, "step": 25086 }, { "epoch": 2.52, "learning_rate": 1.7503322679083708e-05, "loss": 0.1131, "step": 25087 }, { "epoch": 2.52, "learning_rate": 1.7501147064785047e-05, "loss": 0.4729, "step": 25088 }, { "epoch": 2.52, "learning_rate": 1.7498971530024335e-05, "loss": 0.2099, "step": 25089 }, { "epoch": 2.52, "learning_rate": 1.749679607481541e-05, "loss": 0.2894, "step": 25090 }, { "epoch": 2.52, "learning_rate": 1.749462069917212e-05, "loss": 0.1976, "step": 25091 }, { "epoch": 2.52, "learning_rate": 1.74924454031083e-05, "loss": 0.4877, "step": 25092 }, { "epoch": 2.52, "learning_rate": 1.749027018663781e-05, "loss": 0.8494, "step": 25093 }, { "epoch": 2.52, "learning_rate": 1.7488095049774474e-05, "loss": 0.2547, "step": 25094 }, { "epoch": 2.52, "learning_rate": 1.7485919992532144e-05, "loss": 0.8891, "step": 25095 }, { "epoch": 2.52, "learning_rate": 1.748374501492466e-05, "loss": 0.2031, "step": 25096 }, { "epoch": 2.52, "learning_rate": 1.7481570116965858e-05, "loss": 0.1131, "step": 25097 }, { "epoch": 2.52, "learning_rate": 1.747939529866958e-05, "loss": 0.8163, "step": 25098 }, { "epoch": 2.52, "learning_rate": 1.7477220560049657e-05, "loss": 0.2193, "step": 25099 }, { "epoch": 2.52, "learning_rate": 1.7475045901119966e-05, "loss": 0.5064, "step": 25100 }, { "epoch": 2.52, "learning_rate": 1.7472871321894303e-05, "loss": 0.7665, "step": 25101 }, { "epoch": 2.52, "learning_rate": 1.7470696822386512e-05, "loss": 0.1095, "step": 25102 }, { "epoch": 2.52, "learning_rate": 1.7468522402610448e-05, "loss": 0.609, "step": 25103 }, { "epoch": 2.52, "learning_rate": 1.7466348062579953e-05, "loss": 0.2772, "step": 25104 }, { "epoch": 2.52, "learning_rate": 1.7464173802308826e-05, "loss": 0.6839, "step": 25105 }, { "epoch": 2.52, "learning_rate": 1.7461999621810946e-05, "loss": 0.4272, "step": 25106 }, { "epoch": 2.52, "learning_rate": 1.7459825521100128e-05, "loss": 0.2089, "step": 25107 }, { "epoch": 2.52, "learning_rate": 1.7457651500190208e-05, "loss": 0.3133, "step": 25108 }, { "epoch": 2.52, "learning_rate": 1.7455477559095023e-05, "loss": 0.6551, "step": 25109 }, { "epoch": 2.52, "learning_rate": 1.7453303697828405e-05, "loss": 0.3236, "step": 25110 }, { "epoch": 2.52, "learning_rate": 1.7451129916404192e-05, "loss": 0.4827, "step": 25111 }, { "epoch": 2.52, "learning_rate": 1.7448956214836214e-05, "loss": 1.1058, "step": 25112 }, { "epoch": 2.52, "learning_rate": 1.7446782593138292e-05, "loss": 0.5151, "step": 25113 }, { "epoch": 2.52, "learning_rate": 1.744460905132429e-05, "loss": 0.3907, "step": 25114 }, { "epoch": 2.52, "learning_rate": 1.7442435589408015e-05, "loss": 0.2415, "step": 25115 }, { "epoch": 2.52, "learning_rate": 1.7440262207403284e-05, "loss": 0.4303, "step": 25116 }, { "epoch": 2.52, "learning_rate": 1.743808890532396e-05, "loss": 0.6704, "step": 25117 }, { "epoch": 2.52, "learning_rate": 1.7435915683183878e-05, "loss": 0.1431, "step": 25118 }, { "epoch": 2.53, "learning_rate": 1.7433742540996817e-05, "loss": 0.2156, "step": 25119 }, { "epoch": 2.53, "learning_rate": 1.7431569478776655e-05, "loss": 0.1981, "step": 25120 }, { "epoch": 2.53, "learning_rate": 1.7429396496537197e-05, "loss": 0.6042, "step": 25121 }, { "epoch": 2.53, "learning_rate": 1.7427223594292283e-05, "loss": 0.2384, "step": 25122 }, { "epoch": 2.53, "learning_rate": 1.7425050772055732e-05, "loss": 0.2842, "step": 25123 }, { "epoch": 2.53, "learning_rate": 1.742287802984138e-05, "loss": 0.7558, "step": 25124 }, { "epoch": 2.53, "learning_rate": 1.742070536766304e-05, "loss": 0.2776, "step": 25125 }, { "epoch": 2.53, "learning_rate": 1.741853278553455e-05, "loss": 0.1893, "step": 25126 }, { "epoch": 2.53, "learning_rate": 1.7416360283469732e-05, "loss": 0.3078, "step": 25127 }, { "epoch": 2.53, "learning_rate": 1.7414187861482412e-05, "loss": 0.1223, "step": 25128 }, { "epoch": 2.53, "learning_rate": 1.7412015519586408e-05, "loss": 0.1619, "step": 25129 }, { "epoch": 2.53, "learning_rate": 1.7409843257795546e-05, "loss": 0.4112, "step": 25130 }, { "epoch": 2.53, "learning_rate": 1.7407671076123647e-05, "loss": 0.7338, "step": 25131 }, { "epoch": 2.53, "learning_rate": 1.7405498974584558e-05, "loss": 0.4475, "step": 25132 }, { "epoch": 2.53, "learning_rate": 1.7403326953192062e-05, "loss": 0.5849, "step": 25133 }, { "epoch": 2.53, "learning_rate": 1.740115501196001e-05, "loss": 0.4321, "step": 25134 }, { "epoch": 2.53, "learning_rate": 1.7398983150902217e-05, "loss": 1.1314, "step": 25135 }, { "epoch": 2.53, "learning_rate": 1.73968113700325e-05, "loss": 0.3158, "step": 25136 }, { "epoch": 2.53, "learning_rate": 1.739463966936468e-05, "loss": 0.3659, "step": 25137 }, { "epoch": 2.53, "learning_rate": 1.7392468048912577e-05, "loss": 0.2961, "step": 25138 }, { "epoch": 2.53, "learning_rate": 1.7390296508690015e-05, "loss": 0.1517, "step": 25139 }, { "epoch": 2.53, "learning_rate": 1.738812504871081e-05, "loss": 0.5328, "step": 25140 }, { "epoch": 2.53, "learning_rate": 1.7385953668988777e-05, "loss": 0.6444, "step": 25141 }, { "epoch": 2.53, "learning_rate": 1.7383782369537732e-05, "loss": 0.6264, "step": 25142 }, { "epoch": 2.53, "learning_rate": 1.7381611150371502e-05, "loss": 0.4823, "step": 25143 }, { "epoch": 2.53, "learning_rate": 1.7379440011503898e-05, "loss": 0.5921, "step": 25144 }, { "epoch": 2.53, "learning_rate": 1.7377268952948725e-05, "loss": 0.1987, "step": 25145 }, { "epoch": 2.53, "learning_rate": 1.7375097974719828e-05, "loss": 0.3443, "step": 25146 }, { "epoch": 2.53, "learning_rate": 1.7372927076831003e-05, "loss": 0.1775, "step": 25147 }, { "epoch": 2.53, "learning_rate": 1.7370756259296046e-05, "loss": 0.3455, "step": 25148 }, { "epoch": 2.53, "learning_rate": 1.736858552212881e-05, "loss": 0.678, "step": 25149 }, { "epoch": 2.53, "learning_rate": 1.7366414865343098e-05, "loss": 0.2493, "step": 25150 }, { "epoch": 2.53, "learning_rate": 1.7364244288952692e-05, "loss": 0.215, "step": 25151 }, { "epoch": 2.53, "learning_rate": 1.736207379297144e-05, "loss": 0.1463, "step": 25152 }, { "epoch": 2.53, "learning_rate": 1.7359903377413146e-05, "loss": 0.1129, "step": 25153 }, { "epoch": 2.53, "learning_rate": 1.7357733042291615e-05, "loss": 0.6346, "step": 25154 }, { "epoch": 2.53, "learning_rate": 1.7355562787620667e-05, "loss": 0.6875, "step": 25155 }, { "epoch": 2.53, "learning_rate": 1.73533926134141e-05, "loss": 0.2164, "step": 25156 }, { "epoch": 2.53, "learning_rate": 1.7351222519685736e-05, "loss": 1.1894, "step": 25157 }, { "epoch": 2.53, "learning_rate": 1.7349052506449376e-05, "loss": 0.4745, "step": 25158 }, { "epoch": 2.53, "learning_rate": 1.7346882573718826e-05, "loss": 0.1454, "step": 25159 }, { "epoch": 2.53, "learning_rate": 1.7344712721507926e-05, "loss": 0.1584, "step": 25160 }, { "epoch": 2.53, "learning_rate": 1.7342542949830445e-05, "loss": 0.3124, "step": 25161 }, { "epoch": 2.53, "learning_rate": 1.7340373258700193e-05, "loss": 0.1997, "step": 25162 }, { "epoch": 2.53, "learning_rate": 1.7338203648131005e-05, "loss": 0.4538, "step": 25163 }, { "epoch": 2.53, "learning_rate": 1.733603411813668e-05, "loss": 0.1569, "step": 25164 }, { "epoch": 2.53, "learning_rate": 1.7333864668730996e-05, "loss": 0.0961, "step": 25165 }, { "epoch": 2.53, "learning_rate": 1.733169529992779e-05, "loss": 0.4105, "step": 25166 }, { "epoch": 2.53, "learning_rate": 1.7329526011740854e-05, "loss": 0.2802, "step": 25167 }, { "epoch": 2.53, "learning_rate": 1.7327356804183996e-05, "loss": 0.4467, "step": 25168 }, { "epoch": 2.53, "learning_rate": 1.7325187677271016e-05, "loss": 0.4215, "step": 25169 }, { "epoch": 2.53, "learning_rate": 1.732301863101572e-05, "loss": 0.2609, "step": 25170 }, { "epoch": 2.53, "learning_rate": 1.732084966543192e-05, "loss": 0.1063, "step": 25171 }, { "epoch": 2.53, "learning_rate": 1.73186807805334e-05, "loss": 0.2236, "step": 25172 }, { "epoch": 2.53, "learning_rate": 1.731651197633397e-05, "loss": 0.205, "step": 25173 }, { "epoch": 2.53, "learning_rate": 1.7314343252847438e-05, "loss": 0.4316, "step": 25174 }, { "epoch": 2.53, "learning_rate": 1.7312174610087595e-05, "loss": 0.3279, "step": 25175 }, { "epoch": 2.53, "learning_rate": 1.7310006048068247e-05, "loss": 0.4531, "step": 25176 }, { "epoch": 2.53, "learning_rate": 1.7307837566803188e-05, "loss": 0.5025, "step": 25177 }, { "epoch": 2.53, "learning_rate": 1.730566916630624e-05, "loss": 0.0984, "step": 25178 }, { "epoch": 2.53, "learning_rate": 1.730350084659117e-05, "loss": 0.6677, "step": 25179 }, { "epoch": 2.53, "learning_rate": 1.7301332607671784e-05, "loss": 0.2555, "step": 25180 }, { "epoch": 2.53, "learning_rate": 1.7299164449561897e-05, "loss": 0.2668, "step": 25181 }, { "epoch": 2.53, "learning_rate": 1.7296996372275307e-05, "loss": 0.2559, "step": 25182 }, { "epoch": 2.53, "learning_rate": 1.7294828375825777e-05, "loss": 0.201, "step": 25183 }, { "epoch": 2.53, "learning_rate": 1.7292660460227134e-05, "loss": 0.739, "step": 25184 }, { "epoch": 2.53, "learning_rate": 1.7290492625493165e-05, "loss": 0.256, "step": 25185 }, { "epoch": 2.53, "learning_rate": 1.7288324871637673e-05, "loss": 0.116, "step": 25186 }, { "epoch": 2.53, "learning_rate": 1.7286157198674444e-05, "loss": 0.1451, "step": 25187 }, { "epoch": 2.53, "learning_rate": 1.7283989606617267e-05, "loss": 0.3175, "step": 25188 }, { "epoch": 2.53, "learning_rate": 1.7281822095479954e-05, "loss": 0.1155, "step": 25189 }, { "epoch": 2.53, "learning_rate": 1.727965466527628e-05, "loss": 0.4457, "step": 25190 }, { "epoch": 2.53, "learning_rate": 1.7277487316020034e-05, "loss": 0.2626, "step": 25191 }, { "epoch": 2.53, "learning_rate": 1.727532004772504e-05, "loss": 0.6302, "step": 25192 }, { "epoch": 2.53, "learning_rate": 1.7273152860405058e-05, "loss": 0.1801, "step": 25193 }, { "epoch": 2.53, "learning_rate": 1.7270985754073877e-05, "loss": 0.1554, "step": 25194 }, { "epoch": 2.53, "learning_rate": 1.7268818728745317e-05, "loss": 1.1352, "step": 25195 }, { "epoch": 2.53, "learning_rate": 1.7266651784433155e-05, "loss": 0.5878, "step": 25196 }, { "epoch": 2.53, "learning_rate": 1.7264484921151163e-05, "loss": 0.2004, "step": 25197 }, { "epoch": 2.53, "learning_rate": 1.726231813891315e-05, "loss": 0.65, "step": 25198 }, { "epoch": 2.53, "learning_rate": 1.72601514377329e-05, "loss": 0.2749, "step": 25199 }, { "epoch": 2.53, "learning_rate": 1.7257984817624203e-05, "loss": 0.5193, "step": 25200 }, { "epoch": 2.53, "learning_rate": 1.7255818278600837e-05, "loss": 0.6432, "step": 25201 }, { "epoch": 2.53, "learning_rate": 1.72536518206766e-05, "loss": 0.5101, "step": 25202 }, { "epoch": 2.53, "learning_rate": 1.7251485443865273e-05, "loss": 1.8228, "step": 25203 }, { "epoch": 2.53, "learning_rate": 1.7249319148180638e-05, "loss": 0.6649, "step": 25204 }, { "epoch": 2.53, "learning_rate": 1.724715293363649e-05, "loss": 0.1755, "step": 25205 }, { "epoch": 2.53, "learning_rate": 1.7244986800246608e-05, "loss": 0.2351, "step": 25206 }, { "epoch": 2.53, "learning_rate": 1.724282074802478e-05, "loss": 0.9231, "step": 25207 }, { "epoch": 2.53, "learning_rate": 1.7240654776984786e-05, "loss": 0.3203, "step": 25208 }, { "epoch": 2.53, "learning_rate": 1.72384888871404e-05, "loss": 0.9004, "step": 25209 }, { "epoch": 2.53, "learning_rate": 1.723632307850544e-05, "loss": 0.3298, "step": 25210 }, { "epoch": 2.53, "learning_rate": 1.723415735109365e-05, "loss": 0.4761, "step": 25211 }, { "epoch": 2.53, "learning_rate": 1.7231991704918816e-05, "loss": 0.1038, "step": 25212 }, { "epoch": 2.53, "learning_rate": 1.7229826139994743e-05, "loss": 0.1802, "step": 25213 }, { "epoch": 2.53, "learning_rate": 1.7227660656335194e-05, "loss": 0.5664, "step": 25214 }, { "epoch": 2.53, "learning_rate": 1.7225495253953958e-05, "loss": 0.5565, "step": 25215 }, { "epoch": 2.53, "learning_rate": 1.722332993286481e-05, "loss": 0.5441, "step": 25216 }, { "epoch": 2.54, "learning_rate": 1.722116469308153e-05, "loss": 0.5412, "step": 25217 }, { "epoch": 2.54, "learning_rate": 1.7218999534617895e-05, "loss": 0.367, "step": 25218 }, { "epoch": 2.54, "learning_rate": 1.7216834457487686e-05, "loss": 0.1499, "step": 25219 }, { "epoch": 2.54, "learning_rate": 1.7214669461704678e-05, "loss": 1.0908, "step": 25220 }, { "epoch": 2.54, "learning_rate": 1.721250454728265e-05, "loss": 0.7319, "step": 25221 }, { "epoch": 2.54, "learning_rate": 1.721033971423538e-05, "loss": 0.4355, "step": 25222 }, { "epoch": 2.54, "learning_rate": 1.720817496257663e-05, "loss": 0.4229, "step": 25223 }, { "epoch": 2.54, "learning_rate": 1.7206010292320213e-05, "loss": 0.2358, "step": 25224 }, { "epoch": 2.54, "learning_rate": 1.7203845703479863e-05, "loss": 0.0839, "step": 25225 }, { "epoch": 2.54, "learning_rate": 1.7201681196069363e-05, "loss": 0.5709, "step": 25226 }, { "epoch": 2.54, "learning_rate": 1.719951677010251e-05, "loss": 0.2338, "step": 25227 }, { "epoch": 2.54, "learning_rate": 1.7197352425593067e-05, "loss": 0.0873, "step": 25228 }, { "epoch": 2.54, "learning_rate": 1.7195188162554786e-05, "loss": 0.1209, "step": 25229 }, { "epoch": 2.54, "learning_rate": 1.7193023981001463e-05, "loss": 0.3972, "step": 25230 }, { "epoch": 2.54, "learning_rate": 1.7190859880946865e-05, "loss": 0.2098, "step": 25231 }, { "epoch": 2.54, "learning_rate": 1.7188695862404766e-05, "loss": 0.5841, "step": 25232 }, { "epoch": 2.54, "learning_rate": 1.718653192538893e-05, "loss": 0.1366, "step": 25233 }, { "epoch": 2.54, "learning_rate": 1.7184368069913127e-05, "loss": 0.1595, "step": 25234 }, { "epoch": 2.54, "learning_rate": 1.7182204295991137e-05, "loss": 0.2237, "step": 25235 }, { "epoch": 2.54, "learning_rate": 1.7180040603636716e-05, "loss": 0.0801, "step": 25236 }, { "epoch": 2.54, "learning_rate": 1.7177876992863643e-05, "loss": 0.3368, "step": 25237 }, { "epoch": 2.54, "learning_rate": 1.7175713463685684e-05, "loss": 0.4165, "step": 25238 }, { "epoch": 2.54, "learning_rate": 1.7173550016116605e-05, "loss": 0.2126, "step": 25239 }, { "epoch": 2.54, "learning_rate": 1.7171386650170162e-05, "loss": 0.756, "step": 25240 }, { "epoch": 2.54, "learning_rate": 1.716922336586014e-05, "loss": 0.3392, "step": 25241 }, { "epoch": 2.54, "learning_rate": 1.7167060163200322e-05, "loss": 0.2313, "step": 25242 }, { "epoch": 2.54, "learning_rate": 1.7164897042204424e-05, "loss": 0.1641, "step": 25243 }, { "epoch": 2.54, "learning_rate": 1.716273400288625e-05, "loss": 0.1632, "step": 25244 }, { "epoch": 2.54, "learning_rate": 1.716057104525955e-05, "loss": 0.1479, "step": 25245 }, { "epoch": 2.54, "learning_rate": 1.7158408169338092e-05, "loss": 0.2922, "step": 25246 }, { "epoch": 2.54, "learning_rate": 1.7156245375135642e-05, "loss": 0.0826, "step": 25247 }, { "epoch": 2.54, "learning_rate": 1.7154082662665963e-05, "loss": 0.4487, "step": 25248 }, { "epoch": 2.54, "learning_rate": 1.7151920031942807e-05, "loss": 0.6226, "step": 25249 }, { "epoch": 2.54, "learning_rate": 1.714975748297995e-05, "loss": 0.075, "step": 25250 }, { "epoch": 2.54, "learning_rate": 1.714759501579115e-05, "loss": 0.3386, "step": 25251 }, { "epoch": 2.54, "learning_rate": 1.714543263039016e-05, "loss": 0.1082, "step": 25252 }, { "epoch": 2.54, "learning_rate": 1.714327032679075e-05, "loss": 0.4776, "step": 25253 }, { "epoch": 2.54, "learning_rate": 1.7141108105006674e-05, "loss": 0.3931, "step": 25254 }, { "epoch": 2.54, "learning_rate": 1.7138945965051687e-05, "loss": 0.2836, "step": 25255 }, { "epoch": 2.54, "learning_rate": 1.713678390693957e-05, "loss": 0.4271, "step": 25256 }, { "epoch": 2.54, "learning_rate": 1.7134621930684065e-05, "loss": 0.247, "step": 25257 }, { "epoch": 2.54, "learning_rate": 1.7132460036298916e-05, "loss": 0.5257, "step": 25258 }, { "epoch": 2.54, "learning_rate": 1.7130298223797908e-05, "loss": 0.3781, "step": 25259 }, { "epoch": 2.54, "learning_rate": 1.7128136493194798e-05, "loss": 0.3718, "step": 25260 }, { "epoch": 2.54, "learning_rate": 1.7125974844503308e-05, "loss": 0.7479, "step": 25261 }, { "epoch": 2.54, "learning_rate": 1.7123813277737227e-05, "loss": 0.3565, "step": 25262 }, { "epoch": 2.54, "learning_rate": 1.7121651792910305e-05, "loss": 0.087, "step": 25263 }, { "epoch": 2.54, "learning_rate": 1.7119490390036284e-05, "loss": 0.5897, "step": 25264 }, { "epoch": 2.54, "learning_rate": 1.7117329069128933e-05, "loss": 0.7809, "step": 25265 }, { "epoch": 2.54, "learning_rate": 1.7115167830201994e-05, "loss": 0.2085, "step": 25266 }, { "epoch": 2.54, "learning_rate": 1.7113006673269224e-05, "loss": 0.7731, "step": 25267 }, { "epoch": 2.54, "learning_rate": 1.711084559834438e-05, "loss": 0.1712, "step": 25268 }, { "epoch": 2.54, "learning_rate": 1.71086846054412e-05, "loss": 0.4367, "step": 25269 }, { "epoch": 2.54, "learning_rate": 1.710652369457347e-05, "loss": 0.1909, "step": 25270 }, { "epoch": 2.54, "learning_rate": 1.7104362865754907e-05, "loss": 0.5444, "step": 25271 }, { "epoch": 2.54, "learning_rate": 1.710220211899926e-05, "loss": 0.4802, "step": 25272 }, { "epoch": 2.54, "learning_rate": 1.7100041454320305e-05, "loss": 0.4016, "step": 25273 }, { "epoch": 2.54, "learning_rate": 1.7097880871731794e-05, "loss": 0.4882, "step": 25274 }, { "epoch": 2.54, "learning_rate": 1.7095720371247434e-05, "loss": 0.1876, "step": 25275 }, { "epoch": 2.54, "learning_rate": 1.7093559952881013e-05, "loss": 0.3572, "step": 25276 }, { "epoch": 2.54, "learning_rate": 1.7091399616646265e-05, "loss": 0.6897, "step": 25277 }, { "epoch": 2.54, "learning_rate": 1.7089239362556938e-05, "loss": 0.5241, "step": 25278 }, { "epoch": 2.54, "learning_rate": 1.7087079190626784e-05, "loss": 0.4141, "step": 25279 }, { "epoch": 2.54, "learning_rate": 1.7084919100869544e-05, "loss": 0.3729, "step": 25280 }, { "epoch": 2.54, "learning_rate": 1.7082759093298964e-05, "loss": 0.5097, "step": 25281 }, { "epoch": 2.54, "learning_rate": 1.708059916792879e-05, "loss": 0.4229, "step": 25282 }, { "epoch": 2.54, "learning_rate": 1.7078439324772773e-05, "loss": 0.2613, "step": 25283 }, { "epoch": 2.54, "learning_rate": 1.7076279563844653e-05, "loss": 0.6781, "step": 25284 }, { "epoch": 2.54, "learning_rate": 1.7074119885158166e-05, "loss": 0.2423, "step": 25285 }, { "epoch": 2.54, "learning_rate": 1.707196028872707e-05, "loss": 0.6899, "step": 25286 }, { "epoch": 2.54, "learning_rate": 1.7069800774565085e-05, "loss": 0.3151, "step": 25287 }, { "epoch": 2.54, "learning_rate": 1.7067641342685994e-05, "loss": 0.7211, "step": 25288 }, { "epoch": 2.54, "learning_rate": 1.7065481993103504e-05, "loss": 0.5929, "step": 25289 }, { "epoch": 2.54, "learning_rate": 1.7063322725831352e-05, "loss": 0.2683, "step": 25290 }, { "epoch": 2.54, "learning_rate": 1.70611635408833e-05, "loss": 0.1467, "step": 25291 }, { "epoch": 2.54, "learning_rate": 1.7059004438273104e-05, "loss": 0.0883, "step": 25292 }, { "epoch": 2.54, "learning_rate": 1.7056845418014448e-05, "loss": 0.3683, "step": 25293 }, { "epoch": 2.54, "learning_rate": 1.7054686480121117e-05, "loss": 0.4432, "step": 25294 }, { "epoch": 2.54, "learning_rate": 1.705252762460684e-05, "loss": 1.0423, "step": 25295 }, { "epoch": 2.54, "learning_rate": 1.705036885148535e-05, "loss": 0.3218, "step": 25296 }, { "epoch": 2.54, "learning_rate": 1.7048210160770385e-05, "loss": 0.1585, "step": 25297 }, { "epoch": 2.54, "learning_rate": 1.7046051552475683e-05, "loss": 0.3775, "step": 25298 }, { "epoch": 2.54, "learning_rate": 1.7043893026614984e-05, "loss": 0.4579, "step": 25299 }, { "epoch": 2.54, "learning_rate": 1.7041734583202018e-05, "loss": 0.1065, "step": 25300 }, { "epoch": 2.54, "learning_rate": 1.7039576222250515e-05, "loss": 0.1063, "step": 25301 }, { "epoch": 2.54, "learning_rate": 1.7037417943774237e-05, "loss": 0.9116, "step": 25302 }, { "epoch": 2.54, "learning_rate": 1.7035259747786897e-05, "loss": 0.6666, "step": 25303 }, { "epoch": 2.54, "learning_rate": 1.703310163430221e-05, "loss": 0.1585, "step": 25304 }, { "epoch": 2.54, "learning_rate": 1.7030943603333946e-05, "loss": 0.4658, "step": 25305 }, { "epoch": 2.54, "learning_rate": 1.7028785654895837e-05, "loss": 0.2435, "step": 25306 }, { "epoch": 2.54, "learning_rate": 1.702662778900158e-05, "loss": 0.0835, "step": 25307 }, { "epoch": 2.54, "learning_rate": 1.7024470005664937e-05, "loss": 0.7892, "step": 25308 }, { "epoch": 2.54, "learning_rate": 1.702231230489963e-05, "loss": 0.1011, "step": 25309 }, { "epoch": 2.54, "learning_rate": 1.7020154686719396e-05, "loss": 0.3337, "step": 25310 }, { "epoch": 2.54, "learning_rate": 1.7017997151137956e-05, "loss": 0.2016, "step": 25311 }, { "epoch": 2.54, "learning_rate": 1.7015839698169042e-05, "loss": 0.3834, "step": 25312 }, { "epoch": 2.54, "learning_rate": 1.7013682327826386e-05, "loss": 0.5379, "step": 25313 }, { "epoch": 2.54, "learning_rate": 1.7011525040123717e-05, "loss": 0.1084, "step": 25314 }, { "epoch": 2.54, "learning_rate": 1.700936783507476e-05, "loss": 0.5723, "step": 25315 }, { "epoch": 2.55, "learning_rate": 1.7007210712693247e-05, "loss": 0.2817, "step": 25316 }, { "epoch": 2.55, "learning_rate": 1.7005053672992897e-05, "loss": 0.2658, "step": 25317 }, { "epoch": 2.55, "learning_rate": 1.7002896715987445e-05, "loss": 0.6548, "step": 25318 }, { "epoch": 2.55, "learning_rate": 1.7000739841690603e-05, "loss": 0.4688, "step": 25319 }, { "epoch": 2.55, "learning_rate": 1.6998583050116128e-05, "loss": 1.1785, "step": 25320 }, { "epoch": 2.55, "learning_rate": 1.6996426341277716e-05, "loss": 0.8019, "step": 25321 }, { "epoch": 2.55, "learning_rate": 1.6994269715189087e-05, "loss": 0.3891, "step": 25322 }, { "epoch": 2.55, "learning_rate": 1.6992113171863986e-05, "loss": 0.6606, "step": 25323 }, { "epoch": 2.55, "learning_rate": 1.6989956711316136e-05, "loss": 0.1097, "step": 25324 }, { "epoch": 2.55, "learning_rate": 1.6987800333559247e-05, "loss": 0.0703, "step": 25325 }, { "epoch": 2.55, "learning_rate": 1.6985644038607046e-05, "loss": 0.5909, "step": 25326 }, { "epoch": 2.55, "learning_rate": 1.6983487826473257e-05, "loss": 0.303, "step": 25327 }, { "epoch": 2.55, "learning_rate": 1.69813316971716e-05, "loss": 0.2041, "step": 25328 }, { "epoch": 2.55, "learning_rate": 1.6979175650715796e-05, "loss": 0.5319, "step": 25329 }, { "epoch": 2.55, "learning_rate": 1.6977019687119563e-05, "loss": 0.2693, "step": 25330 }, { "epoch": 2.55, "learning_rate": 1.6974863806396628e-05, "loss": 0.4616, "step": 25331 }, { "epoch": 2.55, "learning_rate": 1.6972708008560697e-05, "loss": 0.5001, "step": 25332 }, { "epoch": 2.55, "learning_rate": 1.697055229362549e-05, "loss": 0.8283, "step": 25333 }, { "epoch": 2.55, "learning_rate": 1.6968396661604754e-05, "loss": 0.0779, "step": 25334 }, { "epoch": 2.55, "learning_rate": 1.6966241112512168e-05, "loss": 0.0706, "step": 25335 }, { "epoch": 2.55, "learning_rate": 1.6964085646361462e-05, "loss": 0.2019, "step": 25336 }, { "epoch": 2.55, "learning_rate": 1.696193026316636e-05, "loss": 0.2684, "step": 25337 }, { "epoch": 2.55, "learning_rate": 1.695977496294059e-05, "loss": 0.363, "step": 25338 }, { "epoch": 2.55, "learning_rate": 1.6957619745697826e-05, "loss": 0.6075, "step": 25339 }, { "epoch": 2.55, "learning_rate": 1.6955464611451816e-05, "loss": 0.1905, "step": 25340 }, { "epoch": 2.55, "learning_rate": 1.6953309560216272e-05, "loss": 0.2544, "step": 25341 }, { "epoch": 2.55, "learning_rate": 1.6951154592004902e-05, "loss": 0.5966, "step": 25342 }, { "epoch": 2.55, "learning_rate": 1.694899970683142e-05, "loss": 0.1301, "step": 25343 }, { "epoch": 2.55, "learning_rate": 1.6946844904709537e-05, "loss": 0.6477, "step": 25344 }, { "epoch": 2.55, "learning_rate": 1.694469018565297e-05, "loss": 0.4356, "step": 25345 }, { "epoch": 2.55, "learning_rate": 1.6942535549675426e-05, "loss": 0.3021, "step": 25346 }, { "epoch": 2.55, "learning_rate": 1.6940380996790622e-05, "loss": 0.1988, "step": 25347 }, { "epoch": 2.55, "learning_rate": 1.6938226527012263e-05, "loss": 0.0983, "step": 25348 }, { "epoch": 2.55, "learning_rate": 1.6936072140354063e-05, "loss": 0.4087, "step": 25349 }, { "epoch": 2.55, "learning_rate": 1.6933917836829714e-05, "loss": 0.3796, "step": 25350 }, { "epoch": 2.55, "learning_rate": 1.693176361645296e-05, "loss": 0.3173, "step": 25351 }, { "epoch": 2.55, "learning_rate": 1.6929609479237495e-05, "loss": 0.4289, "step": 25352 }, { "epoch": 2.55, "learning_rate": 1.6927455425197004e-05, "loss": 0.4999, "step": 25353 }, { "epoch": 2.55, "learning_rate": 1.6925301454345225e-05, "loss": 0.1676, "step": 25354 }, { "epoch": 2.55, "learning_rate": 1.6923147566695853e-05, "loss": 0.1518, "step": 25355 }, { "epoch": 2.55, "learning_rate": 1.6920993762262595e-05, "loss": 0.2418, "step": 25356 }, { "epoch": 2.55, "learning_rate": 1.6918840041059157e-05, "loss": 0.0952, "step": 25357 }, { "epoch": 2.55, "learning_rate": 1.6916686403099244e-05, "loss": 0.3102, "step": 25358 }, { "epoch": 2.55, "learning_rate": 1.691453284839656e-05, "loss": 0.4472, "step": 25359 }, { "epoch": 2.55, "learning_rate": 1.691237937696482e-05, "loss": 0.1781, "step": 25360 }, { "epoch": 2.55, "learning_rate": 1.691022598881771e-05, "loss": 0.5142, "step": 25361 }, { "epoch": 2.55, "learning_rate": 1.6908072683968946e-05, "loss": 0.2283, "step": 25362 }, { "epoch": 2.55, "learning_rate": 1.6905919462432226e-05, "loss": 0.1111, "step": 25363 }, { "epoch": 2.55, "learning_rate": 1.6903766324221254e-05, "loss": 0.0494, "step": 25364 }, { "epoch": 2.55, "learning_rate": 1.6901613269349718e-05, "loss": 0.7574, "step": 25365 }, { "epoch": 2.55, "learning_rate": 1.6899460297831354e-05, "loss": 0.7667, "step": 25366 }, { "epoch": 2.55, "learning_rate": 1.6897307409679832e-05, "loss": 0.0432, "step": 25367 }, { "epoch": 2.55, "learning_rate": 1.689515460490885e-05, "loss": 0.1982, "step": 25368 }, { "epoch": 2.55, "learning_rate": 1.6893001883532127e-05, "loss": 0.5341, "step": 25369 }, { "epoch": 2.55, "learning_rate": 1.689084924556337e-05, "loss": 0.2829, "step": 25370 }, { "epoch": 2.55, "learning_rate": 1.6888696691016236e-05, "loss": 0.4392, "step": 25371 }, { "epoch": 2.55, "learning_rate": 1.688654421990446e-05, "loss": 0.28, "step": 25372 }, { "epoch": 2.55, "learning_rate": 1.6884391832241725e-05, "loss": 0.1464, "step": 25373 }, { "epoch": 2.55, "learning_rate": 1.6882239528041738e-05, "loss": 0.3638, "step": 25374 }, { "epoch": 2.55, "learning_rate": 1.6880087307318182e-05, "loss": 0.1628, "step": 25375 }, { "epoch": 2.55, "learning_rate": 1.687793517008475e-05, "loss": 0.1208, "step": 25376 }, { "epoch": 2.55, "learning_rate": 1.687578311635517e-05, "loss": 0.0854, "step": 25377 }, { "epoch": 2.55, "learning_rate": 1.6873631146143103e-05, "loss": 0.1866, "step": 25378 }, { "epoch": 2.55, "learning_rate": 1.6871479259462243e-05, "loss": 0.6214, "step": 25379 }, { "epoch": 2.55, "learning_rate": 1.6869327456326317e-05, "loss": 0.1727, "step": 25380 }, { "epoch": 2.55, "learning_rate": 1.686717573674898e-05, "loss": 0.8093, "step": 25381 }, { "epoch": 2.55, "learning_rate": 1.686502410074393e-05, "loss": 0.2416, "step": 25382 }, { "epoch": 2.55, "learning_rate": 1.686287254832489e-05, "loss": 0.6269, "step": 25383 }, { "epoch": 2.55, "learning_rate": 1.6860721079505535e-05, "loss": 0.1711, "step": 25384 }, { "epoch": 2.55, "learning_rate": 1.685856969429953e-05, "loss": 0.4539, "step": 25385 }, { "epoch": 2.55, "learning_rate": 1.6856418392720598e-05, "loss": 0.2712, "step": 25386 }, { "epoch": 2.55, "learning_rate": 1.685426717478243e-05, "loss": 0.5913, "step": 25387 }, { "epoch": 2.55, "learning_rate": 1.6852116040498693e-05, "loss": 0.3608, "step": 25388 }, { "epoch": 2.55, "learning_rate": 1.684996498988309e-05, "loss": 0.5366, "step": 25389 }, { "epoch": 2.55, "learning_rate": 1.684781402294931e-05, "loss": 0.8247, "step": 25390 }, { "epoch": 2.55, "learning_rate": 1.684566313971104e-05, "loss": 0.3056, "step": 25391 }, { "epoch": 2.55, "learning_rate": 1.6843512340181962e-05, "loss": 0.3347, "step": 25392 }, { "epoch": 2.55, "learning_rate": 1.684136162437577e-05, "loss": 0.4983, "step": 25393 }, { "epoch": 2.55, "learning_rate": 1.6839210992306142e-05, "loss": 0.6007, "step": 25394 }, { "epoch": 2.55, "learning_rate": 1.683706044398677e-05, "loss": 0.1253, "step": 25395 }, { "epoch": 2.55, "learning_rate": 1.6834909979431337e-05, "loss": 0.4001, "step": 25396 }, { "epoch": 2.55, "learning_rate": 1.6832759598653523e-05, "loss": 1.0118, "step": 25397 }, { "epoch": 2.55, "learning_rate": 1.6830609301667033e-05, "loss": 0.4053, "step": 25398 }, { "epoch": 2.55, "learning_rate": 1.682845908848553e-05, "loss": 0.6131, "step": 25399 }, { "epoch": 2.55, "learning_rate": 1.6826308959122688e-05, "loss": 0.388, "step": 25400 }, { "epoch": 2.55, "learning_rate": 1.682415891359222e-05, "loss": 0.5674, "step": 25401 }, { "epoch": 2.55, "learning_rate": 1.6822008951907797e-05, "loss": 0.5595, "step": 25402 }, { "epoch": 2.55, "learning_rate": 1.6819859074083076e-05, "loss": 0.6994, "step": 25403 }, { "epoch": 2.55, "learning_rate": 1.681770928013177e-05, "loss": 0.4222, "step": 25404 }, { "epoch": 2.55, "learning_rate": 1.681555957006755e-05, "loss": 0.623, "step": 25405 }, { "epoch": 2.55, "learning_rate": 1.681340994390409e-05, "loss": 0.4248, "step": 25406 }, { "epoch": 2.55, "learning_rate": 1.681126040165507e-05, "loss": 0.2726, "step": 25407 }, { "epoch": 2.55, "learning_rate": 1.6809110943334168e-05, "loss": 0.38, "step": 25408 }, { "epoch": 2.55, "learning_rate": 1.680696156895508e-05, "loss": 0.544, "step": 25409 }, { "epoch": 2.55, "learning_rate": 1.6804812278531465e-05, "loss": 0.0685, "step": 25410 }, { "epoch": 2.55, "learning_rate": 1.6802663072076988e-05, "loss": 0.3299, "step": 25411 }, { "epoch": 2.55, "learning_rate": 1.680051394960537e-05, "loss": 0.2348, "step": 25412 }, { "epoch": 2.55, "learning_rate": 1.6798364911130248e-05, "loss": 0.7948, "step": 25413 }, { "epoch": 2.56, "learning_rate": 1.6796215956665294e-05, "loss": 0.3515, "step": 25414 }, { "epoch": 2.56, "learning_rate": 1.6794067086224212e-05, "loss": 0.4455, "step": 25415 }, { "epoch": 2.56, "learning_rate": 1.6791918299820678e-05, "loss": 0.5887, "step": 25416 }, { "epoch": 2.56, "learning_rate": 1.6789769597468325e-05, "loss": 0.2171, "step": 25417 }, { "epoch": 2.56, "learning_rate": 1.6787620979180868e-05, "loss": 0.2426, "step": 25418 }, { "epoch": 2.56, "learning_rate": 1.678547244497196e-05, "loss": 0.2713, "step": 25419 }, { "epoch": 2.56, "learning_rate": 1.6783323994855283e-05, "loss": 0.6225, "step": 25420 }, { "epoch": 2.56, "learning_rate": 1.67811756288445e-05, "loss": 0.2991, "step": 25421 }, { "epoch": 2.56, "learning_rate": 1.6779027346953288e-05, "loss": 0.3707, "step": 25422 }, { "epoch": 2.56, "learning_rate": 1.6776879149195317e-05, "loss": 0.3915, "step": 25423 }, { "epoch": 2.56, "learning_rate": 1.677473103558426e-05, "loss": 0.5073, "step": 25424 }, { "epoch": 2.56, "learning_rate": 1.6772583006133775e-05, "loss": 0.4634, "step": 25425 }, { "epoch": 2.56, "learning_rate": 1.6770435060857546e-05, "loss": 0.351, "step": 25426 }, { "epoch": 2.56, "learning_rate": 1.6768287199769235e-05, "loss": 0.2292, "step": 25427 }, { "epoch": 2.56, "learning_rate": 1.6766139422882512e-05, "loss": 0.1545, "step": 25428 }, { "epoch": 2.56, "learning_rate": 1.676399173021103e-05, "loss": 0.3722, "step": 25429 }, { "epoch": 2.56, "learning_rate": 1.676184412176849e-05, "loss": 0.3676, "step": 25430 }, { "epoch": 2.56, "learning_rate": 1.6759696597568518e-05, "loss": 0.6767, "step": 25431 }, { "epoch": 2.56, "learning_rate": 1.675754915762481e-05, "loss": 0.2697, "step": 25432 }, { "epoch": 2.56, "learning_rate": 1.6755401801951026e-05, "loss": 0.1278, "step": 25433 }, { "epoch": 2.56, "learning_rate": 1.675325453056082e-05, "loss": 0.1877, "step": 25434 }, { "epoch": 2.56, "learning_rate": 1.6751107343467864e-05, "loss": 0.4735, "step": 25435 }, { "epoch": 2.56, "learning_rate": 1.6748960240685824e-05, "loss": 0.2678, "step": 25436 }, { "epoch": 2.56, "learning_rate": 1.674681322222836e-05, "loss": 0.563, "step": 25437 }, { "epoch": 2.56, "learning_rate": 1.6744666288109132e-05, "loss": 0.5685, "step": 25438 }, { "epoch": 2.56, "learning_rate": 1.6742519438341803e-05, "loss": 0.3171, "step": 25439 }, { "epoch": 2.56, "learning_rate": 1.674037267294003e-05, "loss": 0.1225, "step": 25440 }, { "epoch": 2.56, "learning_rate": 1.6738225991917503e-05, "loss": 0.7092, "step": 25441 }, { "epoch": 2.56, "learning_rate": 1.673607939528785e-05, "loss": 0.2129, "step": 25442 }, { "epoch": 2.56, "learning_rate": 1.6733932883064726e-05, "loss": 0.1653, "step": 25443 }, { "epoch": 2.56, "learning_rate": 1.673178645526184e-05, "loss": 0.5074, "step": 25444 }, { "epoch": 2.56, "learning_rate": 1.6729640111892794e-05, "loss": 0.08, "step": 25445 }, { "epoch": 2.56, "learning_rate": 1.6727493852971264e-05, "loss": 1.0107, "step": 25446 }, { "epoch": 2.56, "learning_rate": 1.6725347678510924e-05, "loss": 0.3608, "step": 25447 }, { "epoch": 2.56, "learning_rate": 1.6723201588525436e-05, "loss": 0.4961, "step": 25448 }, { "epoch": 2.56, "learning_rate": 1.6721055583028416e-05, "loss": 0.5252, "step": 25449 }, { "epoch": 2.56, "learning_rate": 1.671890966203356e-05, "loss": 0.587, "step": 25450 }, { "epoch": 2.56, "learning_rate": 1.671676382555451e-05, "loss": 0.2574, "step": 25451 }, { "epoch": 2.56, "learning_rate": 1.671461807360492e-05, "loss": 0.4088, "step": 25452 }, { "epoch": 2.56, "learning_rate": 1.671247240619845e-05, "loss": 0.4435, "step": 25453 }, { "epoch": 2.56, "learning_rate": 1.671032682334875e-05, "loss": 0.6119, "step": 25454 }, { "epoch": 2.56, "learning_rate": 1.670818132506947e-05, "loss": 0.1718, "step": 25455 }, { "epoch": 2.56, "learning_rate": 1.6706035911374273e-05, "loss": 0.1154, "step": 25456 }, { "epoch": 2.56, "learning_rate": 1.670389058227679e-05, "loss": 0.3356, "step": 25457 }, { "epoch": 2.56, "learning_rate": 1.6701745337790705e-05, "loss": 0.3841, "step": 25458 }, { "epoch": 2.56, "learning_rate": 1.6699600177929664e-05, "loss": 0.5017, "step": 25459 }, { "epoch": 2.56, "learning_rate": 1.6697455102707286e-05, "loss": 0.1345, "step": 25460 }, { "epoch": 2.56, "learning_rate": 1.6695310112137252e-05, "loss": 0.2, "step": 25461 }, { "epoch": 2.56, "learning_rate": 1.6693165206233216e-05, "loss": 0.1364, "step": 25462 }, { "epoch": 2.56, "learning_rate": 1.6691020385008792e-05, "loss": 0.3651, "step": 25463 }, { "epoch": 2.56, "learning_rate": 1.668887564847766e-05, "loss": 0.4488, "step": 25464 }, { "epoch": 2.56, "learning_rate": 1.6686730996653463e-05, "loss": 0.6742, "step": 25465 }, { "epoch": 2.56, "learning_rate": 1.6684586429549843e-05, "loss": 0.2483, "step": 25466 }, { "epoch": 2.56, "learning_rate": 1.6682441947180455e-05, "loss": 0.4967, "step": 25467 }, { "epoch": 2.56, "learning_rate": 1.6680297549558932e-05, "loss": 0.9193, "step": 25468 }, { "epoch": 2.56, "learning_rate": 1.6678153236698933e-05, "loss": 0.4137, "step": 25469 }, { "epoch": 2.56, "learning_rate": 1.6676009008614097e-05, "loss": 0.4811, "step": 25470 }, { "epoch": 2.56, "learning_rate": 1.6673864865318072e-05, "loss": 0.5929, "step": 25471 }, { "epoch": 2.56, "learning_rate": 1.667172080682449e-05, "loss": 0.3862, "step": 25472 }, { "epoch": 2.56, "learning_rate": 1.666957683314703e-05, "loss": 0.7852, "step": 25473 }, { "epoch": 2.56, "learning_rate": 1.66674329442993e-05, "loss": 0.4314, "step": 25474 }, { "epoch": 2.56, "learning_rate": 1.666528914029494e-05, "loss": 0.4105, "step": 25475 }, { "epoch": 2.56, "learning_rate": 1.6663145421147634e-05, "loss": 0.3533, "step": 25476 }, { "epoch": 2.56, "learning_rate": 1.6661001786870987e-05, "loss": 0.2197, "step": 25477 }, { "epoch": 2.56, "learning_rate": 1.6658858237478632e-05, "loss": 0.5818, "step": 25478 }, { "epoch": 2.56, "learning_rate": 1.6656714772984244e-05, "loss": 0.2862, "step": 25479 }, { "epoch": 2.56, "learning_rate": 1.665457139340146e-05, "loss": 0.2035, "step": 25480 }, { "epoch": 2.56, "learning_rate": 1.665242809874388e-05, "loss": 0.8862, "step": 25481 }, { "epoch": 2.56, "learning_rate": 1.6650284889025183e-05, "loss": 0.3109, "step": 25482 }, { "epoch": 2.56, "learning_rate": 1.6648141764258997e-05, "loss": 0.3376, "step": 25483 }, { "epoch": 2.56, "learning_rate": 1.6645998724458953e-05, "loss": 0.5269, "step": 25484 }, { "epoch": 2.56, "learning_rate": 1.66438557696387e-05, "loss": 0.3104, "step": 25485 }, { "epoch": 2.56, "learning_rate": 1.664171289981185e-05, "loss": 0.1894, "step": 25486 }, { "epoch": 2.56, "learning_rate": 1.6639570114992082e-05, "loss": 0.1217, "step": 25487 }, { "epoch": 2.56, "learning_rate": 1.6637427415192994e-05, "loss": 0.3774, "step": 25488 }, { "epoch": 2.56, "learning_rate": 1.6635284800428228e-05, "loss": 0.5138, "step": 25489 }, { "epoch": 2.56, "learning_rate": 1.6633142270711432e-05, "loss": 0.1613, "step": 25490 }, { "epoch": 2.56, "learning_rate": 1.663099982605625e-05, "loss": 0.1902, "step": 25491 }, { "epoch": 2.56, "learning_rate": 1.6628857466476272e-05, "loss": 0.3289, "step": 25492 }, { "epoch": 2.56, "learning_rate": 1.662671519198517e-05, "loss": 0.2969, "step": 25493 }, { "epoch": 2.56, "learning_rate": 1.662457300259658e-05, "loss": 0.3924, "step": 25494 }, { "epoch": 2.56, "learning_rate": 1.662243089832409e-05, "loss": 0.419, "step": 25495 }, { "epoch": 2.56, "learning_rate": 1.662028887918138e-05, "loss": 0.4255, "step": 25496 }, { "epoch": 2.56, "learning_rate": 1.6618146945182055e-05, "loss": 0.3732, "step": 25497 }, { "epoch": 2.56, "learning_rate": 1.6616005096339754e-05, "loss": 0.4258, "step": 25498 }, { "epoch": 2.56, "learning_rate": 1.6613863332668103e-05, "loss": 0.4001, "step": 25499 }, { "epoch": 2.56, "learning_rate": 1.661172165418074e-05, "loss": 0.4674, "step": 25500 }, { "epoch": 2.56, "learning_rate": 1.6609580060891284e-05, "loss": 0.3494, "step": 25501 }, { "epoch": 2.56, "learning_rate": 1.6607438552813363e-05, "loss": 0.7889, "step": 25502 }, { "epoch": 2.56, "learning_rate": 1.6605297129960614e-05, "loss": 0.4208, "step": 25503 }, { "epoch": 2.56, "learning_rate": 1.660315579234664e-05, "loss": 0.0753, "step": 25504 }, { "epoch": 2.56, "learning_rate": 1.6601014539985117e-05, "loss": 0.6054, "step": 25505 }, { "epoch": 2.56, "learning_rate": 1.6598873372889623e-05, "loss": 0.2038, "step": 25506 }, { "epoch": 2.56, "learning_rate": 1.659673229107379e-05, "loss": 0.2226, "step": 25507 }, { "epoch": 2.56, "learning_rate": 1.659459129455128e-05, "loss": 0.3494, "step": 25508 }, { "epoch": 2.56, "learning_rate": 1.6592450383335678e-05, "loss": 0.3053, "step": 25509 }, { "epoch": 2.56, "learning_rate": 1.659030955744061e-05, "loss": 0.2446, "step": 25510 }, { "epoch": 2.56, "learning_rate": 1.658816881687972e-05, "loss": 0.0244, "step": 25511 }, { "epoch": 2.56, "learning_rate": 1.6586028161666623e-05, "loss": 0.605, "step": 25512 }, { "epoch": 2.57, "learning_rate": 1.6583887591814944e-05, "loss": 0.5908, "step": 25513 }, { "epoch": 2.57, "learning_rate": 1.6581747107338295e-05, "loss": 0.1193, "step": 25514 }, { "epoch": 2.57, "learning_rate": 1.6579606708250305e-05, "loss": 0.3449, "step": 25515 }, { "epoch": 2.57, "learning_rate": 1.657746639456459e-05, "loss": 0.1766, "step": 25516 }, { "epoch": 2.57, "learning_rate": 1.6575326166294778e-05, "loss": 0.3134, "step": 25517 }, { "epoch": 2.57, "learning_rate": 1.657318602345447e-05, "loss": 0.1244, "step": 25518 }, { "epoch": 2.57, "learning_rate": 1.6571045966057317e-05, "loss": 0.2135, "step": 25519 }, { "epoch": 2.57, "learning_rate": 1.656890599411691e-05, "loss": 0.6132, "step": 25520 }, { "epoch": 2.57, "learning_rate": 1.6566766107646865e-05, "loss": 0.2464, "step": 25521 }, { "epoch": 2.57, "learning_rate": 1.6564626306660823e-05, "loss": 0.7402, "step": 25522 }, { "epoch": 2.57, "learning_rate": 1.6562486591172396e-05, "loss": 0.1109, "step": 25523 }, { "epoch": 2.57, "learning_rate": 1.6560346961195174e-05, "loss": 0.3809, "step": 25524 }, { "epoch": 2.57, "learning_rate": 1.65582074167428e-05, "loss": 0.7911, "step": 25525 }, { "epoch": 2.57, "learning_rate": 1.6556067957828894e-05, "loss": 0.1107, "step": 25526 }, { "epoch": 2.57, "learning_rate": 1.655392858446703e-05, "loss": 0.5021, "step": 25527 }, { "epoch": 2.57, "learning_rate": 1.6551789296670863e-05, "loss": 0.5164, "step": 25528 }, { "epoch": 2.57, "learning_rate": 1.654965009445399e-05, "loss": 0.0783, "step": 25529 }, { "epoch": 2.57, "learning_rate": 1.654751097783003e-05, "loss": 0.5956, "step": 25530 }, { "epoch": 2.57, "learning_rate": 1.6545371946812595e-05, "loss": 0.1169, "step": 25531 }, { "epoch": 2.57, "learning_rate": 1.6543233001415292e-05, "loss": 0.5321, "step": 25532 }, { "epoch": 2.57, "learning_rate": 1.654109414165174e-05, "loss": 0.2934, "step": 25533 }, { "epoch": 2.57, "learning_rate": 1.6538955367535537e-05, "loss": 0.118, "step": 25534 }, { "epoch": 2.57, "learning_rate": 1.6536816679080304e-05, "loss": 0.1105, "step": 25535 }, { "epoch": 2.57, "learning_rate": 1.6534678076299633e-05, "loss": 0.4308, "step": 25536 }, { "epoch": 2.57, "learning_rate": 1.653253955920718e-05, "loss": 0.2652, "step": 25537 }, { "epoch": 2.57, "learning_rate": 1.6530401127816495e-05, "loss": 0.7499, "step": 25538 }, { "epoch": 2.57, "learning_rate": 1.652826278214122e-05, "loss": 0.1612, "step": 25539 }, { "epoch": 2.57, "learning_rate": 1.6526124522194975e-05, "loss": 0.905, "step": 25540 }, { "epoch": 2.57, "learning_rate": 1.6523986347991314e-05, "loss": 0.4384, "step": 25541 }, { "epoch": 2.57, "learning_rate": 1.6521848259543898e-05, "loss": 0.8057, "step": 25542 }, { "epoch": 2.57, "learning_rate": 1.651971025686631e-05, "loss": 0.1786, "step": 25543 }, { "epoch": 2.57, "learning_rate": 1.651757233997216e-05, "loss": 0.23, "step": 25544 }, { "epoch": 2.57, "learning_rate": 1.6515434508875047e-05, "loss": 0.3752, "step": 25545 }, { "epoch": 2.57, "learning_rate": 1.651329676358858e-05, "loss": 0.5757, "step": 25546 }, { "epoch": 2.57, "learning_rate": 1.651115910412636e-05, "loss": 0.1781, "step": 25547 }, { "epoch": 2.57, "learning_rate": 1.6509021530501995e-05, "loss": 0.384, "step": 25548 }, { "epoch": 2.57, "learning_rate": 1.6506884042729084e-05, "loss": 0.6293, "step": 25549 }, { "epoch": 2.57, "learning_rate": 1.6504746640821218e-05, "loss": 0.609, "step": 25550 }, { "epoch": 2.57, "learning_rate": 1.6502609324792037e-05, "loss": 0.2244, "step": 25551 }, { "epoch": 2.57, "learning_rate": 1.6500472094655097e-05, "loss": 0.5221, "step": 25552 }, { "epoch": 2.57, "learning_rate": 1.6498334950424012e-05, "loss": 0.1833, "step": 25553 }, { "epoch": 2.57, "learning_rate": 1.6496197892112395e-05, "loss": 0.2303, "step": 25554 }, { "epoch": 2.57, "learning_rate": 1.649406091973385e-05, "loss": 0.3811, "step": 25555 }, { "epoch": 2.57, "learning_rate": 1.649192403330194e-05, "loss": 0.1813, "step": 25556 }, { "epoch": 2.57, "learning_rate": 1.64897872328303e-05, "loss": 0.5545, "step": 25557 }, { "epoch": 2.57, "learning_rate": 1.6487650518332523e-05, "loss": 0.2175, "step": 25558 }, { "epoch": 2.57, "learning_rate": 1.6485513889822177e-05, "loss": 0.2023, "step": 25559 }, { "epoch": 2.57, "learning_rate": 1.648337734731289e-05, "loss": 0.6447, "step": 25560 }, { "epoch": 2.57, "learning_rate": 1.648124089081825e-05, "loss": 0.5309, "step": 25561 }, { "epoch": 2.57, "learning_rate": 1.6479104520351845e-05, "loss": 0.1678, "step": 25562 }, { "epoch": 2.57, "learning_rate": 1.647696823592728e-05, "loss": 0.436, "step": 25563 }, { "epoch": 2.57, "learning_rate": 1.6474832037558143e-05, "loss": 0.0814, "step": 25564 }, { "epoch": 2.57, "learning_rate": 1.6472695925258028e-05, "loss": 0.5626, "step": 25565 }, { "epoch": 2.57, "learning_rate": 1.6470559899040532e-05, "loss": 0.282, "step": 25566 }, { "epoch": 2.57, "learning_rate": 1.6468423958919236e-05, "loss": 0.3113, "step": 25567 }, { "epoch": 2.57, "learning_rate": 1.646628810490775e-05, "loss": 0.1561, "step": 25568 }, { "epoch": 2.57, "learning_rate": 1.646415233701967e-05, "loss": 0.3754, "step": 25569 }, { "epoch": 2.57, "learning_rate": 1.6462016655268557e-05, "loss": 0.2712, "step": 25570 }, { "epoch": 2.57, "learning_rate": 1.645988105966803e-05, "loss": 0.1733, "step": 25571 }, { "epoch": 2.57, "learning_rate": 1.6457745550231677e-05, "loss": 0.2043, "step": 25572 }, { "epoch": 2.57, "learning_rate": 1.6455610126973057e-05, "loss": 0.2128, "step": 25573 }, { "epoch": 2.57, "learning_rate": 1.6453474789905796e-05, "loss": 0.9831, "step": 25574 }, { "epoch": 2.57, "learning_rate": 1.6451339539043466e-05, "loss": 0.5119, "step": 25575 }, { "epoch": 2.57, "learning_rate": 1.6449204374399658e-05, "loss": 0.4659, "step": 25576 }, { "epoch": 2.57, "learning_rate": 1.6447069295987954e-05, "loss": 0.2689, "step": 25577 }, { "epoch": 2.57, "learning_rate": 1.6444934303821952e-05, "loss": 0.1668, "step": 25578 }, { "epoch": 2.57, "learning_rate": 1.6442799397915226e-05, "loss": 0.8182, "step": 25579 }, { "epoch": 2.57, "learning_rate": 1.644066457828137e-05, "loss": 0.1535, "step": 25580 }, { "epoch": 2.57, "learning_rate": 1.643852984493397e-05, "loss": 0.5181, "step": 25581 }, { "epoch": 2.57, "learning_rate": 1.643639519788659e-05, "loss": 0.2458, "step": 25582 }, { "epoch": 2.57, "learning_rate": 1.6434260637152857e-05, "loss": 0.5137, "step": 25583 }, { "epoch": 2.57, "learning_rate": 1.643212616274631e-05, "loss": 0.4402, "step": 25584 }, { "epoch": 2.57, "learning_rate": 1.6429991774680547e-05, "loss": 0.792, "step": 25585 }, { "epoch": 2.57, "learning_rate": 1.6427857472969166e-05, "loss": 0.6027, "step": 25586 }, { "epoch": 2.57, "learning_rate": 1.6425723257625746e-05, "loss": 0.4934, "step": 25587 }, { "epoch": 2.57, "learning_rate": 1.6423589128663837e-05, "loss": 0.1556, "step": 25588 }, { "epoch": 2.57, "learning_rate": 1.642145508609705e-05, "loss": 0.6677, "step": 25589 }, { "epoch": 2.57, "learning_rate": 1.6419321129938972e-05, "loss": 0.4048, "step": 25590 }, { "epoch": 2.57, "learning_rate": 1.641718726020314e-05, "loss": 0.6485, "step": 25591 }, { "epoch": 2.57, "learning_rate": 1.641505347690318e-05, "loss": 0.2243, "step": 25592 }, { "epoch": 2.57, "learning_rate": 1.641291978005265e-05, "loss": 0.2099, "step": 25593 }, { "epoch": 2.57, "learning_rate": 1.6410786169665125e-05, "loss": 0.2326, "step": 25594 }, { "epoch": 2.57, "learning_rate": 1.640865264575419e-05, "loss": 0.2961, "step": 25595 }, { "epoch": 2.57, "learning_rate": 1.6406519208333413e-05, "loss": 0.3793, "step": 25596 }, { "epoch": 2.57, "learning_rate": 1.6404385857416393e-05, "loss": 0.398, "step": 25597 }, { "epoch": 2.57, "learning_rate": 1.6402252593016683e-05, "loss": 0.1565, "step": 25598 }, { "epoch": 2.57, "learning_rate": 1.640011941514785e-05, "loss": 0.163, "step": 25599 }, { "epoch": 2.57, "learning_rate": 1.639798632382349e-05, "loss": 0.7723, "step": 25600 }, { "epoch": 2.57, "learning_rate": 1.6395853319057187e-05, "loss": 0.3274, "step": 25601 }, { "epoch": 2.57, "learning_rate": 1.639372040086248e-05, "loss": 0.537, "step": 25602 }, { "epoch": 2.57, "learning_rate": 1.6391587569252966e-05, "loss": 0.7476, "step": 25603 }, { "epoch": 2.57, "learning_rate": 1.6389454824242222e-05, "loss": 0.6357, "step": 25604 }, { "epoch": 2.57, "learning_rate": 1.638732216584379e-05, "loss": 0.5642, "step": 25605 }, { "epoch": 2.57, "learning_rate": 1.638518959407127e-05, "loss": 0.1522, "step": 25606 }, { "epoch": 2.57, "learning_rate": 1.6383057108938227e-05, "loss": 0.0582, "step": 25607 }, { "epoch": 2.57, "learning_rate": 1.6380924710458226e-05, "loss": 0.2186, "step": 25608 }, { "epoch": 2.57, "learning_rate": 1.6378792398644843e-05, "loss": 0.3729, "step": 25609 }, { "epoch": 2.57, "learning_rate": 1.637666017351164e-05, "loss": 0.1461, "step": 25610 }, { "epoch": 2.58, "learning_rate": 1.637452803507219e-05, "loss": 0.2968, "step": 25611 }, { "epoch": 2.58, "learning_rate": 1.637239598334006e-05, "loss": 0.4839, "step": 25612 }, { "epoch": 2.58, "learning_rate": 1.6370264018328817e-05, "loss": 0.2378, "step": 25613 }, { "epoch": 2.58, "learning_rate": 1.6368132140052016e-05, "loss": 0.0935, "step": 25614 }, { "epoch": 2.58, "learning_rate": 1.636600034852326e-05, "loss": 0.1791, "step": 25615 }, { "epoch": 2.58, "learning_rate": 1.6363868643756082e-05, "loss": 0.215, "step": 25616 }, { "epoch": 2.58, "learning_rate": 1.6361737025764043e-05, "loss": 0.161, "step": 25617 }, { "epoch": 2.58, "learning_rate": 1.6359605494560734e-05, "loss": 0.2103, "step": 25618 }, { "epoch": 2.58, "learning_rate": 1.635747405015972e-05, "loss": 0.2269, "step": 25619 }, { "epoch": 2.58, "learning_rate": 1.6355342692574525e-05, "loss": 0.7207, "step": 25620 }, { "epoch": 2.58, "learning_rate": 1.635321142181875e-05, "loss": 0.0579, "step": 25621 }, { "epoch": 2.58, "learning_rate": 1.6351080237905946e-05, "loss": 0.2251, "step": 25622 }, { "epoch": 2.58, "learning_rate": 1.6348949140849676e-05, "loss": 0.1896, "step": 25623 }, { "epoch": 2.58, "learning_rate": 1.6346818130663502e-05, "loss": 0.2823, "step": 25624 }, { "epoch": 2.58, "learning_rate": 1.634468720736098e-05, "loss": 0.1662, "step": 25625 }, { "epoch": 2.58, "learning_rate": 1.634255637095568e-05, "loss": 0.4753, "step": 25626 }, { "epoch": 2.58, "learning_rate": 1.6340425621461147e-05, "loss": 0.5598, "step": 25627 }, { "epoch": 2.58, "learning_rate": 1.6338294958890945e-05, "loss": 0.1934, "step": 25628 }, { "epoch": 2.58, "learning_rate": 1.6336164383258657e-05, "loss": 0.6322, "step": 25629 }, { "epoch": 2.58, "learning_rate": 1.633403389457781e-05, "loss": 0.6354, "step": 25630 }, { "epoch": 2.58, "learning_rate": 1.633190349286196e-05, "loss": 0.4066, "step": 25631 }, { "epoch": 2.58, "learning_rate": 1.6329773178124686e-05, "loss": 0.1411, "step": 25632 }, { "epoch": 2.58, "learning_rate": 1.632764295037955e-05, "loss": 0.3275, "step": 25633 }, { "epoch": 2.58, "learning_rate": 1.6325512809640066e-05, "loss": 0.1548, "step": 25634 }, { "epoch": 2.58, "learning_rate": 1.6323382755919833e-05, "loss": 0.0929, "step": 25635 }, { "epoch": 2.58, "learning_rate": 1.63212527892324e-05, "loss": 0.191, "step": 25636 }, { "epoch": 2.58, "learning_rate": 1.6319122909591285e-05, "loss": 0.2657, "step": 25637 }, { "epoch": 2.58, "learning_rate": 1.6316993117010077e-05, "loss": 0.4109, "step": 25638 }, { "epoch": 2.58, "learning_rate": 1.631486341150232e-05, "loss": 0.3155, "step": 25639 }, { "epoch": 2.58, "learning_rate": 1.6312733793081574e-05, "loss": 0.314, "step": 25640 }, { "epoch": 2.58, "learning_rate": 1.6310604261761377e-05, "loss": 0.1907, "step": 25641 }, { "epoch": 2.58, "learning_rate": 1.630847481755529e-05, "loss": 0.2086, "step": 25642 }, { "epoch": 2.58, "learning_rate": 1.6306345460476854e-05, "loss": 0.3347, "step": 25643 }, { "epoch": 2.58, "learning_rate": 1.6304216190539634e-05, "loss": 0.1895, "step": 25644 }, { "epoch": 2.58, "learning_rate": 1.630208700775717e-05, "loss": 0.8175, "step": 25645 }, { "epoch": 2.58, "learning_rate": 1.6299957912142997e-05, "loss": 0.1142, "step": 25646 }, { "epoch": 2.58, "learning_rate": 1.629782890371071e-05, "loss": 0.5794, "step": 25647 }, { "epoch": 2.58, "learning_rate": 1.6295699982473794e-05, "loss": 0.2084, "step": 25648 }, { "epoch": 2.58, "learning_rate": 1.6293571148445848e-05, "loss": 0.4231, "step": 25649 }, { "epoch": 2.58, "learning_rate": 1.6291442401640407e-05, "loss": 0.3915, "step": 25650 }, { "epoch": 2.58, "learning_rate": 1.6289313742071002e-05, "loss": 0.0436, "step": 25651 }, { "epoch": 2.58, "learning_rate": 1.6287185169751194e-05, "loss": 0.8111, "step": 25652 }, { "epoch": 2.58, "learning_rate": 1.628505668469452e-05, "loss": 0.0918, "step": 25653 }, { "epoch": 2.58, "learning_rate": 1.6282928286914528e-05, "loss": 0.7026, "step": 25654 }, { "epoch": 2.58, "learning_rate": 1.6280799976424767e-05, "loss": 0.4857, "step": 25655 }, { "epoch": 2.58, "learning_rate": 1.627867175323877e-05, "loss": 0.3544, "step": 25656 }, { "epoch": 2.58, "learning_rate": 1.6276543617370087e-05, "loss": 0.153, "step": 25657 }, { "epoch": 2.58, "learning_rate": 1.6274415568832263e-05, "loss": 0.3782, "step": 25658 }, { "epoch": 2.58, "learning_rate": 1.6272287607638835e-05, "loss": 0.5196, "step": 25659 }, { "epoch": 2.58, "learning_rate": 1.627015973380333e-05, "loss": 0.0964, "step": 25660 }, { "epoch": 2.58, "learning_rate": 1.6268031947339337e-05, "loss": 0.2343, "step": 25661 }, { "epoch": 2.58, "learning_rate": 1.626590424826035e-05, "loss": 0.1037, "step": 25662 }, { "epoch": 2.58, "learning_rate": 1.626377663657991e-05, "loss": 0.7292, "step": 25663 }, { "epoch": 2.58, "learning_rate": 1.6261649112311582e-05, "loss": 0.7048, "step": 25664 }, { "epoch": 2.58, "learning_rate": 1.6259521675468904e-05, "loss": 0.8973, "step": 25665 }, { "epoch": 2.58, "learning_rate": 1.6257394326065382e-05, "loss": 0.2115, "step": 25666 }, { "epoch": 2.58, "learning_rate": 1.6255267064114587e-05, "loss": 0.3, "step": 25667 }, { "epoch": 2.58, "learning_rate": 1.6253139889630057e-05, "loss": 0.3113, "step": 25668 }, { "epoch": 2.58, "learning_rate": 1.6251012802625284e-05, "loss": 0.1225, "step": 25669 }, { "epoch": 2.58, "learning_rate": 1.6248885803113854e-05, "loss": 0.2774, "step": 25670 }, { "epoch": 2.58, "learning_rate": 1.6246758891109276e-05, "loss": 0.3859, "step": 25671 }, { "epoch": 2.58, "learning_rate": 1.62446320666251e-05, "loss": 0.4074, "step": 25672 }, { "epoch": 2.58, "learning_rate": 1.6242505329674848e-05, "loss": 0.332, "step": 25673 }, { "epoch": 2.58, "learning_rate": 1.624037868027205e-05, "loss": 0.3193, "step": 25674 }, { "epoch": 2.58, "learning_rate": 1.6238252118430272e-05, "loss": 0.1187, "step": 25675 }, { "epoch": 2.58, "learning_rate": 1.6236125644163004e-05, "loss": 0.3904, "step": 25676 }, { "epoch": 2.58, "learning_rate": 1.623399925748379e-05, "loss": 0.2696, "step": 25677 }, { "epoch": 2.58, "learning_rate": 1.623187295840618e-05, "loss": 0.1006, "step": 25678 }, { "epoch": 2.58, "learning_rate": 1.62297467469437e-05, "loss": 0.4335, "step": 25679 }, { "epoch": 2.58, "learning_rate": 1.6227620623109855e-05, "loss": 0.6327, "step": 25680 }, { "epoch": 2.58, "learning_rate": 1.62254945869182e-05, "loss": 0.1899, "step": 25681 }, { "epoch": 2.58, "learning_rate": 1.6223368638382258e-05, "loss": 0.087, "step": 25682 }, { "epoch": 2.58, "learning_rate": 1.6221242777515562e-05, "loss": 0.3859, "step": 25683 }, { "epoch": 2.58, "learning_rate": 1.621911700433163e-05, "loss": 0.1895, "step": 25684 }, { "epoch": 2.58, "learning_rate": 1.6216991318844e-05, "loss": 0.2665, "step": 25685 }, { "epoch": 2.58, "learning_rate": 1.621486572106619e-05, "loss": 0.2344, "step": 25686 }, { "epoch": 2.58, "learning_rate": 1.621274021101173e-05, "loss": 0.183, "step": 25687 }, { "epoch": 2.58, "learning_rate": 1.6210614788694148e-05, "loss": 0.1248, "step": 25688 }, { "epoch": 2.58, "learning_rate": 1.620848945412697e-05, "loss": 0.2455, "step": 25689 }, { "epoch": 2.58, "learning_rate": 1.6206364207323714e-05, "loss": 0.5687, "step": 25690 }, { "epoch": 2.58, "learning_rate": 1.620423904829791e-05, "loss": 0.1484, "step": 25691 }, { "epoch": 2.58, "learning_rate": 1.620211397706307e-05, "loss": 0.2971, "step": 25692 }, { "epoch": 2.58, "learning_rate": 1.619998899363275e-05, "loss": 0.5944, "step": 25693 }, { "epoch": 2.58, "learning_rate": 1.6197864098020443e-05, "loss": 0.4755, "step": 25694 }, { "epoch": 2.58, "learning_rate": 1.6195739290239663e-05, "loss": 0.1991, "step": 25695 }, { "epoch": 2.58, "learning_rate": 1.6193614570303955e-05, "loss": 0.7469, "step": 25696 }, { "epoch": 2.58, "learning_rate": 1.6191489938226848e-05, "loss": 0.5145, "step": 25697 }, { "epoch": 2.58, "learning_rate": 1.6189365394021823e-05, "loss": 0.1935, "step": 25698 }, { "epoch": 2.58, "learning_rate": 1.6187240937702433e-05, "loss": 0.4091, "step": 25699 }, { "epoch": 2.58, "learning_rate": 1.6185116569282196e-05, "loss": 0.1398, "step": 25700 }, { "epoch": 2.58, "learning_rate": 1.61829922887746e-05, "loss": 0.4182, "step": 25701 }, { "epoch": 2.58, "learning_rate": 1.61808680961932e-05, "loss": 0.1793, "step": 25702 }, { "epoch": 2.58, "learning_rate": 1.6178743991551492e-05, "loss": 0.3648, "step": 25703 }, { "epoch": 2.58, "learning_rate": 1.6176619974863005e-05, "loss": 0.1773, "step": 25704 }, { "epoch": 2.58, "learning_rate": 1.6174496046141247e-05, "loss": 0.3629, "step": 25705 }, { "epoch": 2.58, "learning_rate": 1.6172372205399724e-05, "loss": 0.0733, "step": 25706 }, { "epoch": 2.58, "learning_rate": 1.617024845265199e-05, "loss": 0.1368, "step": 25707 }, { "epoch": 2.58, "learning_rate": 1.6168124787911522e-05, "loss": 0.2529, "step": 25708 }, { "epoch": 2.58, "learning_rate": 1.6166001211191836e-05, "loss": 0.0608, "step": 25709 }, { "epoch": 2.59, "learning_rate": 1.6163877722506463e-05, "loss": 0.3178, "step": 25710 }, { "epoch": 2.59, "learning_rate": 1.6161754321868924e-05, "loss": 0.2023, "step": 25711 }, { "epoch": 2.59, "learning_rate": 1.6159631009292692e-05, "loss": 0.2541, "step": 25712 }, { "epoch": 2.59, "learning_rate": 1.615750778479131e-05, "loss": 0.3176, "step": 25713 }, { "epoch": 2.59, "learning_rate": 1.6155384648378287e-05, "loss": 0.1517, "step": 25714 }, { "epoch": 2.59, "learning_rate": 1.615326160006713e-05, "loss": 0.5583, "step": 25715 }, { "epoch": 2.59, "learning_rate": 1.6151138639871347e-05, "loss": 0.098, "step": 25716 }, { "epoch": 2.59, "learning_rate": 1.614901576780445e-05, "loss": 0.1259, "step": 25717 }, { "epoch": 2.59, "learning_rate": 1.6146892983879944e-05, "loss": 0.3938, "step": 25718 }, { "epoch": 2.59, "learning_rate": 1.6144770288111345e-05, "loss": 0.3716, "step": 25719 }, { "epoch": 2.59, "learning_rate": 1.6142647680512157e-05, "loss": 0.268, "step": 25720 }, { "epoch": 2.59, "learning_rate": 1.6140525161095886e-05, "loss": 0.478, "step": 25721 }, { "epoch": 2.59, "learning_rate": 1.6138402729876043e-05, "loss": 0.3216, "step": 25722 }, { "epoch": 2.59, "learning_rate": 1.6136280386866126e-05, "loss": 0.3075, "step": 25723 }, { "epoch": 2.59, "learning_rate": 1.613415813207964e-05, "loss": 0.5167, "step": 25724 }, { "epoch": 2.59, "learning_rate": 1.613203596553012e-05, "loss": 0.3663, "step": 25725 }, { "epoch": 2.59, "learning_rate": 1.6129913887231032e-05, "loss": 0.332, "step": 25726 }, { "epoch": 2.59, "learning_rate": 1.612779189719589e-05, "loss": 0.6857, "step": 25727 }, { "epoch": 2.59, "learning_rate": 1.612566999543821e-05, "loss": 0.3446, "step": 25728 }, { "epoch": 2.59, "learning_rate": 1.612354818197149e-05, "loss": 1.041, "step": 25729 }, { "epoch": 2.59, "learning_rate": 1.6121426456809228e-05, "loss": 0.1034, "step": 25730 }, { "epoch": 2.59, "learning_rate": 1.6119304819964933e-05, "loss": 0.2624, "step": 25731 }, { "epoch": 2.59, "learning_rate": 1.61171832714521e-05, "loss": 0.2872, "step": 25732 }, { "epoch": 2.59, "learning_rate": 1.6115061811284228e-05, "loss": 0.3451, "step": 25733 }, { "epoch": 2.59, "learning_rate": 1.6112940439474827e-05, "loss": 0.8331, "step": 25734 }, { "epoch": 2.59, "learning_rate": 1.6110819156037384e-05, "loss": 0.3904, "step": 25735 }, { "epoch": 2.59, "learning_rate": 1.6108697960985407e-05, "loss": 0.559, "step": 25736 }, { "epoch": 2.59, "learning_rate": 1.6106576854332386e-05, "loss": 0.2576, "step": 25737 }, { "epoch": 2.59, "learning_rate": 1.6104455836091818e-05, "loss": 0.3585, "step": 25738 }, { "epoch": 2.59, "learning_rate": 1.6102334906277233e-05, "loss": 0.2473, "step": 25739 }, { "epoch": 2.59, "learning_rate": 1.6100214064902087e-05, "loss": 0.1987, "step": 25740 }, { "epoch": 2.59, "learning_rate": 1.6098093311979878e-05, "loss": 0.7187, "step": 25741 }, { "epoch": 2.59, "learning_rate": 1.6095972647524128e-05, "loss": 0.2813, "step": 25742 }, { "epoch": 2.59, "learning_rate": 1.609385207154833e-05, "loss": 0.2452, "step": 25743 }, { "epoch": 2.59, "learning_rate": 1.6091731584065943e-05, "loss": 0.3125, "step": 25744 }, { "epoch": 2.59, "learning_rate": 1.6089611185090498e-05, "loss": 0.3745, "step": 25745 }, { "epoch": 2.59, "learning_rate": 1.6087490874635475e-05, "loss": 0.4161, "step": 25746 }, { "epoch": 2.59, "learning_rate": 1.6085370652714368e-05, "loss": 0.1435, "step": 25747 }, { "epoch": 2.59, "learning_rate": 1.608325051934067e-05, "loss": 1.0532, "step": 25748 }, { "epoch": 2.59, "learning_rate": 1.608113047452787e-05, "loss": 0.4197, "step": 25749 }, { "epoch": 2.59, "learning_rate": 1.6079010518289462e-05, "loss": 0.1297, "step": 25750 }, { "epoch": 2.59, "learning_rate": 1.6076890650638937e-05, "loss": 0.0642, "step": 25751 }, { "epoch": 2.59, "learning_rate": 1.6074770871589776e-05, "loss": 0.6859, "step": 25752 }, { "epoch": 2.59, "learning_rate": 1.607265118115548e-05, "loss": 0.2993, "step": 25753 }, { "epoch": 2.59, "learning_rate": 1.6070531579349536e-05, "loss": 0.3428, "step": 25754 }, { "epoch": 2.59, "learning_rate": 1.6068412066185416e-05, "loss": 0.4508, "step": 25755 }, { "epoch": 2.59, "learning_rate": 1.6066292641676633e-05, "loss": 0.3411, "step": 25756 }, { "epoch": 2.59, "learning_rate": 1.6064173305836676e-05, "loss": 0.6712, "step": 25757 }, { "epoch": 2.59, "learning_rate": 1.6062054058678994e-05, "loss": 0.3878, "step": 25758 }, { "epoch": 2.59, "learning_rate": 1.605993490021711e-05, "loss": 0.2089, "step": 25759 }, { "epoch": 2.59, "learning_rate": 1.6057815830464497e-05, "loss": 0.2992, "step": 25760 }, { "epoch": 2.59, "learning_rate": 1.605569684943464e-05, "loss": 0.0587, "step": 25761 }, { "epoch": 2.59, "learning_rate": 1.6053577957141022e-05, "loss": 0.5113, "step": 25762 }, { "epoch": 2.59, "learning_rate": 1.605145915359713e-05, "loss": 0.2995, "step": 25763 }, { "epoch": 2.59, "learning_rate": 1.604934043881645e-05, "loss": 0.3407, "step": 25764 }, { "epoch": 2.59, "learning_rate": 1.6047221812812452e-05, "loss": 0.4682, "step": 25765 }, { "epoch": 2.59, "learning_rate": 1.604510327559863e-05, "loss": 0.63, "step": 25766 }, { "epoch": 2.59, "learning_rate": 1.6042984827188464e-05, "loss": 0.4138, "step": 25767 }, { "epoch": 2.59, "learning_rate": 1.6040866467595432e-05, "loss": 0.0236, "step": 25768 }, { "epoch": 2.59, "learning_rate": 1.6038748196833016e-05, "loss": 0.1217, "step": 25769 }, { "epoch": 2.59, "learning_rate": 1.6036630014914685e-05, "loss": 0.491, "step": 25770 }, { "epoch": 2.59, "learning_rate": 1.6034511921853946e-05, "loss": 0.2579, "step": 25771 }, { "epoch": 2.59, "learning_rate": 1.6032393917664253e-05, "loss": 0.4262, "step": 25772 }, { "epoch": 2.59, "learning_rate": 1.603027600235908e-05, "loss": 0.1445, "step": 25773 }, { "epoch": 2.59, "learning_rate": 1.6028158175951927e-05, "loss": 0.7618, "step": 25774 }, { "epoch": 2.59, "learning_rate": 1.602604043845627e-05, "loss": 0.2537, "step": 25775 }, { "epoch": 2.59, "learning_rate": 1.602392278988556e-05, "loss": 0.3025, "step": 25776 }, { "epoch": 2.59, "learning_rate": 1.602180523025329e-05, "loss": 0.6147, "step": 25777 }, { "epoch": 2.59, "learning_rate": 1.601968775957294e-05, "loss": 0.3811, "step": 25778 }, { "epoch": 2.59, "learning_rate": 1.601757037785798e-05, "loss": 0.2318, "step": 25779 }, { "epoch": 2.59, "learning_rate": 1.6015453085121876e-05, "loss": 0.8984, "step": 25780 }, { "epoch": 2.59, "learning_rate": 1.6013335881378114e-05, "loss": 0.25, "step": 25781 }, { "epoch": 2.59, "learning_rate": 1.601121876664016e-05, "loss": 0.767, "step": 25782 }, { "epoch": 2.59, "learning_rate": 1.600910174092149e-05, "loss": 0.2439, "step": 25783 }, { "epoch": 2.59, "learning_rate": 1.600698480423556e-05, "loss": 0.5747, "step": 25784 }, { "epoch": 2.59, "learning_rate": 1.600486795659588e-05, "loss": 0.3487, "step": 25785 }, { "epoch": 2.59, "learning_rate": 1.6002751198015887e-05, "loss": 0.3187, "step": 25786 }, { "epoch": 2.59, "learning_rate": 1.6000634528509047e-05, "loss": 0.2868, "step": 25787 }, { "epoch": 2.59, "learning_rate": 1.5998517948088853e-05, "loss": 0.1411, "step": 25788 }, { "epoch": 2.59, "learning_rate": 1.599640145676878e-05, "loss": 0.4663, "step": 25789 }, { "epoch": 2.59, "learning_rate": 1.5994285054562253e-05, "loss": 0.3175, "step": 25790 }, { "epoch": 2.59, "learning_rate": 1.599216874148278e-05, "loss": 0.1694, "step": 25791 }, { "epoch": 2.59, "learning_rate": 1.599005251754382e-05, "loss": 0.3279, "step": 25792 }, { "epoch": 2.59, "learning_rate": 1.5987936382758832e-05, "loss": 0.3458, "step": 25793 }, { "epoch": 2.59, "learning_rate": 1.598582033714129e-05, "loss": 0.4457, "step": 25794 }, { "epoch": 2.59, "learning_rate": 1.5983704380704655e-05, "loss": 0.0765, "step": 25795 }, { "epoch": 2.59, "learning_rate": 1.5981588513462387e-05, "loss": 0.0749, "step": 25796 }, { "epoch": 2.59, "learning_rate": 1.5979472735427966e-05, "loss": 0.6453, "step": 25797 }, { "epoch": 2.59, "learning_rate": 1.597735704661484e-05, "loss": 1.1258, "step": 25798 }, { "epoch": 2.59, "learning_rate": 1.5975241447036486e-05, "loss": 0.459, "step": 25799 }, { "epoch": 2.59, "learning_rate": 1.5973125936706357e-05, "loss": 0.2697, "step": 25800 }, { "epoch": 2.59, "learning_rate": 1.597101051563792e-05, "loss": 0.3147, "step": 25801 }, { "epoch": 2.59, "learning_rate": 1.5968895183844617e-05, "loss": 0.3338, "step": 25802 }, { "epoch": 2.59, "learning_rate": 1.596677994133996e-05, "loss": 0.2276, "step": 25803 }, { "epoch": 2.59, "learning_rate": 1.596466478813736e-05, "loss": 0.1948, "step": 25804 }, { "epoch": 2.59, "learning_rate": 1.5962549724250283e-05, "loss": 0.0548, "step": 25805 }, { "epoch": 2.59, "learning_rate": 1.596043474969221e-05, "loss": 0.6795, "step": 25806 }, { "epoch": 2.59, "learning_rate": 1.59583198644766e-05, "loss": 0.5019, "step": 25807 }, { "epoch": 2.6, "learning_rate": 1.5956205068616884e-05, "loss": 0.2975, "step": 25808 }, { "epoch": 2.6, "learning_rate": 1.5954090362126544e-05, "loss": 0.2713, "step": 25809 }, { "epoch": 2.6, "learning_rate": 1.5951975745019024e-05, "loss": 0.5007, "step": 25810 }, { "epoch": 2.6, "learning_rate": 1.5949861217307795e-05, "loss": 0.0415, "step": 25811 }, { "epoch": 2.6, "learning_rate": 1.59477467790063e-05, "loss": 0.7641, "step": 25812 }, { "epoch": 2.6, "learning_rate": 1.5945632430127996e-05, "loss": 0.1719, "step": 25813 }, { "epoch": 2.6, "learning_rate": 1.5943518170686344e-05, "loss": 0.4447, "step": 25814 }, { "epoch": 2.6, "learning_rate": 1.5941404000694795e-05, "loss": 0.1132, "step": 25815 }, { "epoch": 2.6, "learning_rate": 1.593928992016679e-05, "loss": 0.1825, "step": 25816 }, { "epoch": 2.6, "learning_rate": 1.5937175929115822e-05, "loss": 0.515, "step": 25817 }, { "epoch": 2.6, "learning_rate": 1.5935062027555302e-05, "loss": 0.0963, "step": 25818 }, { "epoch": 2.6, "learning_rate": 1.5932948215498683e-05, "loss": 0.1658, "step": 25819 }, { "epoch": 2.6, "learning_rate": 1.5930834492959446e-05, "loss": 0.8151, "step": 25820 }, { "epoch": 2.6, "learning_rate": 1.5928720859951037e-05, "loss": 0.3501, "step": 25821 }, { "epoch": 2.6, "learning_rate": 1.592660731648687e-05, "loss": 0.2347, "step": 25822 }, { "epoch": 2.6, "learning_rate": 1.5924493862580438e-05, "loss": 0.2732, "step": 25823 }, { "epoch": 2.6, "learning_rate": 1.5922380498245163e-05, "loss": 0.1312, "step": 25824 }, { "epoch": 2.6, "learning_rate": 1.592026722349451e-05, "loss": 0.331, "step": 25825 }, { "epoch": 2.6, "learning_rate": 1.5918154038341916e-05, "loss": 0.3195, "step": 25826 }, { "epoch": 2.6, "learning_rate": 1.5916040942800834e-05, "loss": 0.3322, "step": 25827 }, { "epoch": 2.6, "learning_rate": 1.5913927936884712e-05, "loss": 0.2002, "step": 25828 }, { "epoch": 2.6, "learning_rate": 1.591181502060699e-05, "loss": 0.8519, "step": 25829 }, { "epoch": 2.6, "learning_rate": 1.5909702193981123e-05, "loss": 0.0897, "step": 25830 }, { "epoch": 2.6, "learning_rate": 1.5907589457020543e-05, "loss": 0.607, "step": 25831 }, { "epoch": 2.6, "learning_rate": 1.590547680973871e-05, "loss": 0.2464, "step": 25832 }, { "epoch": 2.6, "learning_rate": 1.5903364252149056e-05, "loss": 0.5096, "step": 25833 }, { "epoch": 2.6, "learning_rate": 1.5901251784265023e-05, "loss": 0.4493, "step": 25834 }, { "epoch": 2.6, "learning_rate": 1.589913940610008e-05, "loss": 0.3515, "step": 25835 }, { "epoch": 2.6, "learning_rate": 1.5897027117667626e-05, "loss": 0.165, "step": 25836 }, { "epoch": 2.6, "learning_rate": 1.589491491898114e-05, "loss": 0.1893, "step": 25837 }, { "epoch": 2.6, "learning_rate": 1.5892802810054047e-05, "loss": 0.6858, "step": 25838 }, { "epoch": 2.6, "learning_rate": 1.5890690790899792e-05, "loss": 0.1326, "step": 25839 }, { "epoch": 2.6, "learning_rate": 1.5888578861531812e-05, "loss": 0.8924, "step": 25840 }, { "epoch": 2.6, "learning_rate": 1.588646702196355e-05, "loss": 0.1844, "step": 25841 }, { "epoch": 2.6, "learning_rate": 1.588435527220844e-05, "loss": 0.6832, "step": 25842 }, { "epoch": 2.6, "learning_rate": 1.5882243612279924e-05, "loss": 0.1811, "step": 25843 }, { "epoch": 2.6, "learning_rate": 1.588013204219144e-05, "loss": 0.2191, "step": 25844 }, { "epoch": 2.6, "learning_rate": 1.5878020561956427e-05, "loss": 0.631, "step": 25845 }, { "epoch": 2.6, "learning_rate": 1.5875909171588312e-05, "loss": 0.6783, "step": 25846 }, { "epoch": 2.6, "learning_rate": 1.5873797871100547e-05, "loss": 0.2902, "step": 25847 }, { "epoch": 2.6, "learning_rate": 1.587168666050654e-05, "loss": 0.3977, "step": 25848 }, { "epoch": 2.6, "learning_rate": 1.586957553981977e-05, "loss": 0.5214, "step": 25849 }, { "epoch": 2.6, "learning_rate": 1.5867464509053633e-05, "loss": 0.4253, "step": 25850 }, { "epoch": 2.6, "learning_rate": 1.5865353568221563e-05, "loss": 0.2971, "step": 25851 }, { "epoch": 2.6, "learning_rate": 1.586324271733702e-05, "loss": 0.3563, "step": 25852 }, { "epoch": 2.6, "learning_rate": 1.586113195641343e-05, "loss": 0.8317, "step": 25853 }, { "epoch": 2.6, "learning_rate": 1.5859021285464197e-05, "loss": 0.2335, "step": 25854 }, { "epoch": 2.6, "learning_rate": 1.5856910704502786e-05, "loss": 0.3909, "step": 25855 }, { "epoch": 2.6, "learning_rate": 1.5854800213542616e-05, "loss": 0.7999, "step": 25856 }, { "epoch": 2.6, "learning_rate": 1.5852689812597112e-05, "loss": 0.5819, "step": 25857 }, { "epoch": 2.6, "learning_rate": 1.5850579501679708e-05, "loss": 0.2116, "step": 25858 }, { "epoch": 2.6, "learning_rate": 1.5848469280803837e-05, "loss": 0.5502, "step": 25859 }, { "epoch": 2.6, "learning_rate": 1.5846359149982924e-05, "loss": 0.4548, "step": 25860 }, { "epoch": 2.6, "learning_rate": 1.5844249109230398e-05, "loss": 0.1148, "step": 25861 }, { "epoch": 2.6, "learning_rate": 1.584213915855968e-05, "loss": 0.0595, "step": 25862 }, { "epoch": 2.6, "learning_rate": 1.5840029297984206e-05, "loss": 0.7052, "step": 25863 }, { "epoch": 2.6, "learning_rate": 1.5837919527517396e-05, "loss": 0.0612, "step": 25864 }, { "epoch": 2.6, "learning_rate": 1.583580984717267e-05, "loss": 0.3957, "step": 25865 }, { "epoch": 2.6, "learning_rate": 1.5833700256963478e-05, "loss": 0.1503, "step": 25866 }, { "epoch": 2.6, "learning_rate": 1.5831590756903233e-05, "loss": 0.6549, "step": 25867 }, { "epoch": 2.6, "learning_rate": 1.582948134700533e-05, "loss": 0.4165, "step": 25868 }, { "epoch": 2.6, "learning_rate": 1.5827372027283237e-05, "loss": 0.6765, "step": 25869 }, { "epoch": 2.6, "learning_rate": 1.582526279775035e-05, "loss": 0.7154, "step": 25870 }, { "epoch": 2.6, "learning_rate": 1.58231536584201e-05, "loss": 0.2174, "step": 25871 }, { "epoch": 2.6, "learning_rate": 1.5821044609305907e-05, "loss": 0.4767, "step": 25872 }, { "epoch": 2.6, "learning_rate": 1.5818935650421192e-05, "loss": 0.4434, "step": 25873 }, { "epoch": 2.6, "learning_rate": 1.5816826781779374e-05, "loss": 0.1952, "step": 25874 }, { "epoch": 2.6, "learning_rate": 1.5814718003393877e-05, "loss": 0.4334, "step": 25875 }, { "epoch": 2.6, "learning_rate": 1.5812609315278117e-05, "loss": 0.1716, "step": 25876 }, { "epoch": 2.6, "learning_rate": 1.5810500717445513e-05, "loss": 0.2809, "step": 25877 }, { "epoch": 2.6, "learning_rate": 1.5808392209909487e-05, "loss": 0.3132, "step": 25878 }, { "epoch": 2.6, "learning_rate": 1.5806283792683456e-05, "loss": 0.8894, "step": 25879 }, { "epoch": 2.6, "learning_rate": 1.580417546578082e-05, "loss": 0.0659, "step": 25880 }, { "epoch": 2.6, "learning_rate": 1.5802067229215032e-05, "loss": 0.0814, "step": 25881 }, { "epoch": 2.6, "learning_rate": 1.5799959082999484e-05, "loss": 0.7199, "step": 25882 }, { "epoch": 2.6, "learning_rate": 1.5797851027147578e-05, "loss": 0.6201, "step": 25883 }, { "epoch": 2.6, "learning_rate": 1.5795743061672756e-05, "loss": 0.763, "step": 25884 }, { "epoch": 2.6, "learning_rate": 1.5793635186588432e-05, "loss": 0.4788, "step": 25885 }, { "epoch": 2.6, "learning_rate": 1.579152740190799e-05, "loss": 0.6227, "step": 25886 }, { "epoch": 2.6, "learning_rate": 1.5789419707644874e-05, "loss": 0.3185, "step": 25887 }, { "epoch": 2.6, "learning_rate": 1.578731210381248e-05, "loss": 0.6866, "step": 25888 }, { "epoch": 2.6, "learning_rate": 1.5785204590424228e-05, "loss": 0.404, "step": 25889 }, { "epoch": 2.6, "learning_rate": 1.5783097167493525e-05, "loss": 0.1474, "step": 25890 }, { "epoch": 2.6, "learning_rate": 1.5780989835033787e-05, "loss": 0.5859, "step": 25891 }, { "epoch": 2.6, "learning_rate": 1.577888259305842e-05, "loss": 0.4143, "step": 25892 }, { "epoch": 2.6, "learning_rate": 1.577677544158083e-05, "loss": 0.2812, "step": 25893 }, { "epoch": 2.6, "learning_rate": 1.577466838061442e-05, "loss": 0.3958, "step": 25894 }, { "epoch": 2.6, "learning_rate": 1.5772561410172634e-05, "loss": 0.392, "step": 25895 }, { "epoch": 2.6, "learning_rate": 1.577045453026884e-05, "loss": 0.2501, "step": 25896 }, { "epoch": 2.6, "learning_rate": 1.576834774091645e-05, "loss": 0.1784, "step": 25897 }, { "epoch": 2.6, "learning_rate": 1.5766241042128895e-05, "loss": 0.3856, "step": 25898 }, { "epoch": 2.6, "learning_rate": 1.576413443391958e-05, "loss": 0.4075, "step": 25899 }, { "epoch": 2.6, "learning_rate": 1.5762027916301868e-05, "loss": 0.0683, "step": 25900 }, { "epoch": 2.6, "learning_rate": 1.575992148928921e-05, "loss": 0.6734, "step": 25901 }, { "epoch": 2.6, "learning_rate": 1.5757815152894995e-05, "loss": 0.227, "step": 25902 }, { "epoch": 2.6, "learning_rate": 1.5755708907132626e-05, "loss": 0.4302, "step": 25903 }, { "epoch": 2.6, "learning_rate": 1.5753602752015505e-05, "loss": 0.2195, "step": 25904 }, { "epoch": 2.6, "learning_rate": 1.575149668755704e-05, "loss": 0.8267, "step": 25905 }, { "epoch": 2.6, "learning_rate": 1.5749390713770624e-05, "loss": 0.0926, "step": 25906 }, { "epoch": 2.61, "learning_rate": 1.5747284830669667e-05, "loss": 0.4553, "step": 25907 }, { "epoch": 2.61, "learning_rate": 1.574517903826757e-05, "loss": 0.2262, "step": 25908 }, { "epoch": 2.61, "learning_rate": 1.5743073336577728e-05, "loss": 0.0675, "step": 25909 }, { "epoch": 2.61, "learning_rate": 1.5740967725613543e-05, "loss": 0.11, "step": 25910 }, { "epoch": 2.61, "learning_rate": 1.573886220538842e-05, "loss": 0.7839, "step": 25911 }, { "epoch": 2.61, "learning_rate": 1.5736756775915734e-05, "loss": 0.4305, "step": 25912 }, { "epoch": 2.61, "learning_rate": 1.573465143720893e-05, "loss": 0.374, "step": 25913 }, { "epoch": 2.61, "learning_rate": 1.5732546189281362e-05, "loss": 0.0681, "step": 25914 }, { "epoch": 2.61, "learning_rate": 1.573044103214643e-05, "loss": 0.1718, "step": 25915 }, { "epoch": 2.61, "learning_rate": 1.572833596581756e-05, "loss": 0.4287, "step": 25916 }, { "epoch": 2.61, "learning_rate": 1.572623099030814e-05, "loss": 0.7195, "step": 25917 }, { "epoch": 2.61, "learning_rate": 1.5724126105631533e-05, "loss": 0.6147, "step": 25918 }, { "epoch": 2.61, "learning_rate": 1.5722021311801163e-05, "loss": 0.4613, "step": 25919 }, { "epoch": 2.61, "learning_rate": 1.5719916608830425e-05, "loss": 0.5069, "step": 25920 }, { "epoch": 2.61, "learning_rate": 1.5717811996732703e-05, "loss": 0.3445, "step": 25921 }, { "epoch": 2.61, "learning_rate": 1.5715707475521395e-05, "loss": 0.4513, "step": 25922 }, { "epoch": 2.61, "learning_rate": 1.5713603045209882e-05, "loss": 0.2319, "step": 25923 }, { "epoch": 2.61, "learning_rate": 1.5711498705811577e-05, "loss": 0.0165, "step": 25924 }, { "epoch": 2.61, "learning_rate": 1.5709394457339847e-05, "loss": 0.2066, "step": 25925 }, { "epoch": 2.61, "learning_rate": 1.5707290299808088e-05, "loss": 0.7036, "step": 25926 }, { "epoch": 2.61, "learning_rate": 1.570518623322972e-05, "loss": 0.4573, "step": 25927 }, { "epoch": 2.61, "learning_rate": 1.5703082257618094e-05, "loss": 0.2146, "step": 25928 }, { "epoch": 2.61, "learning_rate": 1.5700978372986602e-05, "loss": 0.2545, "step": 25929 }, { "epoch": 2.61, "learning_rate": 1.5698874579348657e-05, "loss": 0.494, "step": 25930 }, { "epoch": 2.61, "learning_rate": 1.569677087671765e-05, "loss": 0.1525, "step": 25931 }, { "epoch": 2.61, "learning_rate": 1.5694667265106927e-05, "loss": 0.316, "step": 25932 }, { "epoch": 2.61, "learning_rate": 1.5692563744529904e-05, "loss": 0.4482, "step": 25933 }, { "epoch": 2.61, "learning_rate": 1.5690460314999968e-05, "loss": 0.1608, "step": 25934 }, { "epoch": 2.61, "learning_rate": 1.56883569765305e-05, "loss": 0.487, "step": 25935 }, { "epoch": 2.61, "learning_rate": 1.568625372913488e-05, "loss": 0.1614, "step": 25936 }, { "epoch": 2.61, "learning_rate": 1.56841505728265e-05, "loss": 0.0628, "step": 25937 }, { "epoch": 2.61, "learning_rate": 1.5682047507618736e-05, "loss": 0.5309, "step": 25938 }, { "epoch": 2.61, "learning_rate": 1.567994453352498e-05, "loss": 0.4831, "step": 25939 }, { "epoch": 2.61, "learning_rate": 1.5677841650558603e-05, "loss": 0.9697, "step": 25940 }, { "epoch": 2.61, "learning_rate": 1.5675738858732997e-05, "loss": 0.1821, "step": 25941 }, { "epoch": 2.61, "learning_rate": 1.567363615806154e-05, "loss": 0.1315, "step": 25942 }, { "epoch": 2.61, "learning_rate": 1.567153354855761e-05, "loss": 0.0603, "step": 25943 }, { "epoch": 2.61, "learning_rate": 1.5669431030234578e-05, "loss": 0.3512, "step": 25944 }, { "epoch": 2.61, "learning_rate": 1.566732860310586e-05, "loss": 0.2691, "step": 25945 }, { "epoch": 2.61, "learning_rate": 1.5665226267184788e-05, "loss": 0.5486, "step": 25946 }, { "epoch": 2.61, "learning_rate": 1.5663124022484776e-05, "loss": 0.1685, "step": 25947 }, { "epoch": 2.61, "learning_rate": 1.5661021869019184e-05, "loss": 0.6699, "step": 25948 }, { "epoch": 2.61, "learning_rate": 1.5658919806801395e-05, "loss": 0.3626, "step": 25949 }, { "epoch": 2.61, "learning_rate": 1.565681783584479e-05, "loss": 0.1373, "step": 25950 }, { "epoch": 2.61, "learning_rate": 1.565471595616273e-05, "loss": 0.6684, "step": 25951 }, { "epoch": 2.61, "learning_rate": 1.5652614167768612e-05, "loss": 0.3604, "step": 25952 }, { "epoch": 2.61, "learning_rate": 1.565051247067579e-05, "loss": 1.4113, "step": 25953 }, { "epoch": 2.61, "learning_rate": 1.5648410864897653e-05, "loss": 0.1853, "step": 25954 }, { "epoch": 2.61, "learning_rate": 1.5646309350447567e-05, "loss": 0.2811, "step": 25955 }, { "epoch": 2.61, "learning_rate": 1.564420792733891e-05, "loss": 0.134, "step": 25956 }, { "epoch": 2.61, "learning_rate": 1.5642106595585047e-05, "loss": 0.113, "step": 25957 }, { "epoch": 2.61, "learning_rate": 1.5640005355199353e-05, "loss": 0.0518, "step": 25958 }, { "epoch": 2.61, "learning_rate": 1.563790420619522e-05, "loss": 0.4379, "step": 25959 }, { "epoch": 2.61, "learning_rate": 1.563580314858599e-05, "loss": 0.7948, "step": 25960 }, { "epoch": 2.61, "learning_rate": 1.5633702182385036e-05, "loss": 0.7055, "step": 25961 }, { "epoch": 2.61, "learning_rate": 1.563160130760574e-05, "loss": 0.0328, "step": 25962 }, { "epoch": 2.61, "learning_rate": 1.5629500524261485e-05, "loss": 0.7388, "step": 25963 }, { "epoch": 2.61, "learning_rate": 1.56273998323656e-05, "loss": 0.349, "step": 25964 }, { "epoch": 2.61, "learning_rate": 1.562529923193148e-05, "loss": 0.7036, "step": 25965 }, { "epoch": 2.61, "learning_rate": 1.562319872297249e-05, "loss": 0.5742, "step": 25966 }, { "epoch": 2.61, "learning_rate": 1.5621098305502e-05, "loss": 0.2673, "step": 25967 }, { "epoch": 2.61, "learning_rate": 1.561899797953336e-05, "loss": 0.1588, "step": 25968 }, { "epoch": 2.61, "learning_rate": 1.5616897745079953e-05, "loss": 0.3012, "step": 25969 }, { "epoch": 2.61, "learning_rate": 1.5614797602155136e-05, "loss": 0.0811, "step": 25970 }, { "epoch": 2.61, "learning_rate": 1.5612697550772273e-05, "loss": 0.4105, "step": 25971 }, { "epoch": 2.61, "learning_rate": 1.561059759094472e-05, "loss": 0.4239, "step": 25972 }, { "epoch": 2.61, "learning_rate": 1.560849772268587e-05, "loss": 0.4501, "step": 25973 }, { "epoch": 2.61, "learning_rate": 1.5606397946009057e-05, "loss": 0.1682, "step": 25974 }, { "epoch": 2.61, "learning_rate": 1.5604298260927642e-05, "loss": 0.5424, "step": 25975 }, { "epoch": 2.61, "learning_rate": 1.5602198667455006e-05, "loss": 0.1801, "step": 25976 }, { "epoch": 2.61, "learning_rate": 1.560009916560451e-05, "loss": 0.6276, "step": 25977 }, { "epoch": 2.61, "learning_rate": 1.5597999755389484e-05, "loss": 0.7708, "step": 25978 }, { "epoch": 2.61, "learning_rate": 1.5595900436823318e-05, "loss": 0.6078, "step": 25979 }, { "epoch": 2.61, "learning_rate": 1.5593801209919365e-05, "loss": 0.223, "step": 25980 }, { "epoch": 2.61, "learning_rate": 1.559170207469098e-05, "loss": 0.3447, "step": 25981 }, { "epoch": 2.61, "learning_rate": 1.5589603031151523e-05, "loss": 0.188, "step": 25982 }, { "epoch": 2.61, "learning_rate": 1.5587504079314346e-05, "loss": 0.9666, "step": 25983 }, { "epoch": 2.61, "learning_rate": 1.5585405219192816e-05, "loss": 0.4003, "step": 25984 }, { "epoch": 2.61, "learning_rate": 1.558330645080028e-05, "loss": 0.1964, "step": 25985 }, { "epoch": 2.61, "learning_rate": 1.5581207774150094e-05, "loss": 0.1247, "step": 25986 }, { "epoch": 2.61, "learning_rate": 1.5579109189255623e-05, "loss": 0.5053, "step": 25987 }, { "epoch": 2.61, "learning_rate": 1.557701069613021e-05, "loss": 0.5821, "step": 25988 }, { "epoch": 2.61, "learning_rate": 1.5574912294787217e-05, "loss": 0.4841, "step": 25989 }, { "epoch": 2.61, "learning_rate": 1.5572813985239986e-05, "loss": 0.3427, "step": 25990 }, { "epoch": 2.61, "learning_rate": 1.55707157675019e-05, "loss": 0.4147, "step": 25991 }, { "epoch": 2.61, "learning_rate": 1.5568617641586273e-05, "loss": 0.3095, "step": 25992 }, { "epoch": 2.61, "learning_rate": 1.5566519607506466e-05, "loss": 0.1486, "step": 25993 }, { "epoch": 2.61, "learning_rate": 1.556442166527585e-05, "loss": 0.592, "step": 25994 }, { "epoch": 2.61, "learning_rate": 1.5562323814907774e-05, "loss": 0.0947, "step": 25995 }, { "epoch": 2.61, "learning_rate": 1.5560226056415547e-05, "loss": 0.6584, "step": 25996 }, { "epoch": 2.61, "learning_rate": 1.5558128389812568e-05, "loss": 0.1679, "step": 25997 }, { "epoch": 2.61, "learning_rate": 1.5556030815112162e-05, "loss": 0.281, "step": 25998 }, { "epoch": 2.61, "learning_rate": 1.5553933332327682e-05, "loss": 0.2383, "step": 25999 }, { "epoch": 2.61, "learning_rate": 1.5551835941472477e-05, "loss": 0.5757, "step": 26000 }, { "epoch": 2.61, "learning_rate": 1.554973864255989e-05, "loss": 0.4003, "step": 26001 }, { "epoch": 2.61, "learning_rate": 1.5547641435603267e-05, "loss": 0.1319, "step": 26002 }, { "epoch": 2.61, "learning_rate": 1.5545544320615956e-05, "loss": 0.3841, "step": 26003 }, { "epoch": 2.61, "learning_rate": 1.5543447297611294e-05, "loss": 0.3811, "step": 26004 }, { "epoch": 2.62, "learning_rate": 1.5541350366602656e-05, "loss": 0.254, "step": 26005 }, { "epoch": 2.62, "learning_rate": 1.5539253527603352e-05, "loss": 0.2175, "step": 26006 }, { "epoch": 2.62, "learning_rate": 1.5537156780626723e-05, "loss": 0.088, "step": 26007 }, { "epoch": 2.62, "learning_rate": 1.553506012568614e-05, "loss": 0.2243, "step": 26008 }, { "epoch": 2.62, "learning_rate": 1.5532963562794945e-05, "loss": 0.348, "step": 26009 }, { "epoch": 2.62, "learning_rate": 1.5530867091966443e-05, "loss": 0.3092, "step": 26010 }, { "epoch": 2.62, "learning_rate": 1.5528770713214007e-05, "loss": 0.2783, "step": 26011 }, { "epoch": 2.62, "learning_rate": 1.552667442655097e-05, "loss": 0.4253, "step": 26012 }, { "epoch": 2.62, "learning_rate": 1.5524578231990673e-05, "loss": 0.4422, "step": 26013 }, { "epoch": 2.62, "learning_rate": 1.5522482129546452e-05, "loss": 0.332, "step": 26014 }, { "epoch": 2.62, "learning_rate": 1.5520386119231647e-05, "loss": 0.2015, "step": 26015 }, { "epoch": 2.62, "learning_rate": 1.55182902010596e-05, "loss": 0.236, "step": 26016 }, { "epoch": 2.62, "learning_rate": 1.5516194375043637e-05, "loss": 0.5943, "step": 26017 }, { "epoch": 2.62, "learning_rate": 1.551409864119711e-05, "loss": 0.5765, "step": 26018 }, { "epoch": 2.62, "learning_rate": 1.5512002999533347e-05, "loss": 0.2487, "step": 26019 }, { "epoch": 2.62, "learning_rate": 1.5509907450065686e-05, "loss": 1.0312, "step": 26020 }, { "epoch": 2.62, "learning_rate": 1.5507811992807457e-05, "loss": 0.3485, "step": 26021 }, { "epoch": 2.62, "learning_rate": 1.5505716627771994e-05, "loss": 0.1137, "step": 26022 }, { "epoch": 2.62, "learning_rate": 1.550362135497266e-05, "loss": 0.4668, "step": 26023 }, { "epoch": 2.62, "learning_rate": 1.5501526174422747e-05, "loss": 0.4496, "step": 26024 }, { "epoch": 2.62, "learning_rate": 1.5499431086135598e-05, "loss": 0.2556, "step": 26025 }, { "epoch": 2.62, "learning_rate": 1.5497336090124563e-05, "loss": 0.3998, "step": 26026 }, { "epoch": 2.62, "learning_rate": 1.549524118640297e-05, "loss": 0.1026, "step": 26027 }, { "epoch": 2.62, "learning_rate": 1.5493146374984137e-05, "loss": 0.5805, "step": 26028 }, { "epoch": 2.62, "learning_rate": 1.5491051655881407e-05, "loss": 0.3318, "step": 26029 }, { "epoch": 2.62, "learning_rate": 1.54889570291081e-05, "loss": 0.2428, "step": 26030 }, { "epoch": 2.62, "learning_rate": 1.5486862494677553e-05, "loss": 0.2995, "step": 26031 }, { "epoch": 2.62, "learning_rate": 1.548476805260309e-05, "loss": 0.4696, "step": 26032 }, { "epoch": 2.62, "learning_rate": 1.5482673702898043e-05, "loss": 0.5429, "step": 26033 }, { "epoch": 2.62, "learning_rate": 1.5480579445575737e-05, "loss": 0.2389, "step": 26034 }, { "epoch": 2.62, "learning_rate": 1.5478485280649503e-05, "loss": 0.3509, "step": 26035 }, { "epoch": 2.62, "learning_rate": 1.547639120813265e-05, "loss": 0.4511, "step": 26036 }, { "epoch": 2.62, "learning_rate": 1.5474297228038542e-05, "loss": 0.2933, "step": 26037 }, { "epoch": 2.62, "learning_rate": 1.547220334038047e-05, "loss": 0.354, "step": 26038 }, { "epoch": 2.62, "learning_rate": 1.5470109545171754e-05, "loss": 0.148, "step": 26039 }, { "epoch": 2.62, "learning_rate": 1.5468015842425748e-05, "loss": 0.5768, "step": 26040 }, { "epoch": 2.62, "learning_rate": 1.546592223215577e-05, "loss": 0.5348, "step": 26041 }, { "epoch": 2.62, "learning_rate": 1.5463828714375115e-05, "loss": 0.2235, "step": 26042 }, { "epoch": 2.62, "learning_rate": 1.5461735289097133e-05, "loss": 0.4383, "step": 26043 }, { "epoch": 2.62, "learning_rate": 1.5459641956335138e-05, "loss": 0.5159, "step": 26044 }, { "epoch": 2.62, "learning_rate": 1.5457548716102454e-05, "loss": 0.3192, "step": 26045 }, { "epoch": 2.62, "learning_rate": 1.545545556841239e-05, "loss": 0.3209, "step": 26046 }, { "epoch": 2.62, "learning_rate": 1.5453362513278276e-05, "loss": 1.1743, "step": 26047 }, { "epoch": 2.62, "learning_rate": 1.545126955071343e-05, "loss": 0.4275, "step": 26048 }, { "epoch": 2.62, "learning_rate": 1.544917668073117e-05, "loss": 0.3651, "step": 26049 }, { "epoch": 2.62, "learning_rate": 1.5447083903344816e-05, "loss": 0.1741, "step": 26050 }, { "epoch": 2.62, "learning_rate": 1.5444991218567684e-05, "loss": 0.3608, "step": 26051 }, { "epoch": 2.62, "learning_rate": 1.5442898626413088e-05, "loss": 0.1566, "step": 26052 }, { "epoch": 2.62, "learning_rate": 1.5440806126894337e-05, "loss": 0.8626, "step": 26053 }, { "epoch": 2.62, "learning_rate": 1.5438713720024765e-05, "loss": 0.1183, "step": 26054 }, { "epoch": 2.62, "learning_rate": 1.5436621405817698e-05, "loss": 0.1028, "step": 26055 }, { "epoch": 2.62, "learning_rate": 1.5434529184286406e-05, "loss": 0.2084, "step": 26056 }, { "epoch": 2.62, "learning_rate": 1.5432437055444242e-05, "loss": 0.208, "step": 26057 }, { "epoch": 2.62, "learning_rate": 1.5430345019304504e-05, "loss": 0.2766, "step": 26058 }, { "epoch": 2.62, "learning_rate": 1.5428253075880512e-05, "loss": 0.7602, "step": 26059 }, { "epoch": 2.62, "learning_rate": 1.5426161225185567e-05, "loss": 0.0823, "step": 26060 }, { "epoch": 2.62, "learning_rate": 1.542406946723299e-05, "loss": 0.5697, "step": 26061 }, { "epoch": 2.62, "learning_rate": 1.5421977802036093e-05, "loss": 0.1975, "step": 26062 }, { "epoch": 2.62, "learning_rate": 1.5419886229608185e-05, "loss": 0.5496, "step": 26063 }, { "epoch": 2.62, "learning_rate": 1.5417794749962572e-05, "loss": 0.3437, "step": 26064 }, { "epoch": 2.62, "learning_rate": 1.541570336311257e-05, "loss": 0.2491, "step": 26065 }, { "epoch": 2.62, "learning_rate": 1.5413612069071475e-05, "loss": 0.4398, "step": 26066 }, { "epoch": 2.62, "learning_rate": 1.541152086785261e-05, "loss": 0.7444, "step": 26067 }, { "epoch": 2.62, "learning_rate": 1.540942975946926e-05, "loss": 0.3845, "step": 26068 }, { "epoch": 2.62, "learning_rate": 1.5407338743934778e-05, "loss": 0.2705, "step": 26069 }, { "epoch": 2.62, "learning_rate": 1.540524782126243e-05, "loss": 0.106, "step": 26070 }, { "epoch": 2.62, "learning_rate": 1.540315699146551e-05, "loss": 1.0605, "step": 26071 }, { "epoch": 2.62, "learning_rate": 1.5401066254557366e-05, "loss": 0.1536, "step": 26072 }, { "epoch": 2.62, "learning_rate": 1.5398975610551297e-05, "loss": 0.348, "step": 26073 }, { "epoch": 2.62, "learning_rate": 1.5396885059460565e-05, "loss": 0.6259, "step": 26074 }, { "epoch": 2.62, "learning_rate": 1.539479460129851e-05, "loss": 0.6375, "step": 26075 }, { "epoch": 2.62, "learning_rate": 1.539270423607843e-05, "loss": 0.0733, "step": 26076 }, { "epoch": 2.62, "learning_rate": 1.5390613963813624e-05, "loss": 0.6975, "step": 26077 }, { "epoch": 2.62, "learning_rate": 1.538852378451739e-05, "loss": 0.2408, "step": 26078 }, { "epoch": 2.62, "learning_rate": 1.5386433698203034e-05, "loss": 0.1531, "step": 26079 }, { "epoch": 2.62, "learning_rate": 1.5384343704883856e-05, "loss": 0.1179, "step": 26080 }, { "epoch": 2.62, "learning_rate": 1.5382253804573152e-05, "loss": 0.3584, "step": 26081 }, { "epoch": 2.62, "learning_rate": 1.5380163997284213e-05, "loss": 0.3724, "step": 26082 }, { "epoch": 2.62, "learning_rate": 1.537807428303037e-05, "loss": 0.7689, "step": 26083 }, { "epoch": 2.62, "learning_rate": 1.5375984661824897e-05, "loss": 0.2338, "step": 26084 }, { "epoch": 2.62, "learning_rate": 1.5373895133681072e-05, "loss": 0.1646, "step": 26085 }, { "epoch": 2.62, "learning_rate": 1.5371805698612234e-05, "loss": 0.4352, "step": 26086 }, { "epoch": 2.62, "learning_rate": 1.5369716356631666e-05, "loss": 0.1854, "step": 26087 }, { "epoch": 2.62, "learning_rate": 1.5367627107752632e-05, "loss": 0.2452, "step": 26088 }, { "epoch": 2.62, "learning_rate": 1.5365537951988473e-05, "loss": 0.1528, "step": 26089 }, { "epoch": 2.62, "learning_rate": 1.5363448889352455e-05, "loss": 0.2092, "step": 26090 }, { "epoch": 2.62, "learning_rate": 1.5361359919857885e-05, "loss": 0.1117, "step": 26091 }, { "epoch": 2.62, "learning_rate": 1.535927104351805e-05, "loss": 0.1044, "step": 26092 }, { "epoch": 2.62, "learning_rate": 1.5357182260346242e-05, "loss": 0.1053, "step": 26093 }, { "epoch": 2.62, "learning_rate": 1.5355093570355763e-05, "loss": 0.6214, "step": 26094 }, { "epoch": 2.62, "learning_rate": 1.535300497355989e-05, "loss": 0.2072, "step": 26095 }, { "epoch": 2.62, "learning_rate": 1.5350916469971927e-05, "loss": 0.4831, "step": 26096 }, { "epoch": 2.62, "learning_rate": 1.5348828059605154e-05, "loss": 0.2319, "step": 26097 }, { "epoch": 2.62, "learning_rate": 1.5346739742472866e-05, "loss": 0.5783, "step": 26098 }, { "epoch": 2.62, "learning_rate": 1.5344651518588355e-05, "loss": 0.2778, "step": 26099 }, { "epoch": 2.62, "learning_rate": 1.534256338796489e-05, "loss": 0.435, "step": 26100 }, { "epoch": 2.62, "learning_rate": 1.5340475350615805e-05, "loss": 0.5712, "step": 26101 }, { "epoch": 2.62, "learning_rate": 1.5338387406554344e-05, "loss": 0.5642, "step": 26102 }, { "epoch": 2.62, "learning_rate": 1.5336299555793792e-05, "loss": 0.198, "step": 26103 }, { "epoch": 2.63, "learning_rate": 1.5334211798347472e-05, "loss": 0.2821, "step": 26104 }, { "epoch": 2.63, "learning_rate": 1.5332124134228655e-05, "loss": 0.1768, "step": 26105 }, { "epoch": 2.63, "learning_rate": 1.5330036563450598e-05, "loss": 0.3437, "step": 26106 }, { "epoch": 2.63, "learning_rate": 1.5327949086026614e-05, "loss": 0.6672, "step": 26107 }, { "epoch": 2.63, "learning_rate": 1.5325861701969985e-05, "loss": 0.509, "step": 26108 }, { "epoch": 2.63, "learning_rate": 1.532377441129399e-05, "loss": 0.3343, "step": 26109 }, { "epoch": 2.63, "learning_rate": 1.5321687214011914e-05, "loss": 0.142, "step": 26110 }, { "epoch": 2.63, "learning_rate": 1.5319600110137022e-05, "loss": 0.5171, "step": 26111 }, { "epoch": 2.63, "learning_rate": 1.5317513099682635e-05, "loss": 0.0957, "step": 26112 }, { "epoch": 2.63, "learning_rate": 1.5315426182661995e-05, "loss": 0.1114, "step": 26113 }, { "epoch": 2.63, "learning_rate": 1.5313339359088384e-05, "loss": 0.4537, "step": 26114 }, { "epoch": 2.63, "learning_rate": 1.531125262897512e-05, "loss": 0.6893, "step": 26115 }, { "epoch": 2.63, "learning_rate": 1.5309165992335442e-05, "loss": 0.0944, "step": 26116 }, { "epoch": 2.63, "learning_rate": 1.5307079449182636e-05, "loss": 0.2505, "step": 26117 }, { "epoch": 2.63, "learning_rate": 1.5304992999529996e-05, "loss": 0.6726, "step": 26118 }, { "epoch": 2.63, "learning_rate": 1.5302906643390802e-05, "loss": 1.2466, "step": 26119 }, { "epoch": 2.63, "learning_rate": 1.5300820380778297e-05, "loss": 0.8972, "step": 26120 }, { "epoch": 2.63, "learning_rate": 1.5298734211705788e-05, "loss": 0.4547, "step": 26121 }, { "epoch": 2.63, "learning_rate": 1.5296648136186547e-05, "loss": 0.7259, "step": 26122 }, { "epoch": 2.63, "learning_rate": 1.5294562154233838e-05, "loss": 0.1519, "step": 26123 }, { "epoch": 2.63, "learning_rate": 1.5292476265860942e-05, "loss": 0.7269, "step": 26124 }, { "epoch": 2.63, "learning_rate": 1.5290390471081136e-05, "loss": 0.1162, "step": 26125 }, { "epoch": 2.63, "learning_rate": 1.5288304769907685e-05, "loss": 0.925, "step": 26126 }, { "epoch": 2.63, "learning_rate": 1.5286219162353866e-05, "loss": 0.2343, "step": 26127 }, { "epoch": 2.63, "learning_rate": 1.5284133648432956e-05, "loss": 0.2039, "step": 26128 }, { "epoch": 2.63, "learning_rate": 1.5282048228158216e-05, "loss": 0.5907, "step": 26129 }, { "epoch": 2.63, "learning_rate": 1.5279962901542925e-05, "loss": 0.2452, "step": 26130 }, { "epoch": 2.63, "learning_rate": 1.527787766860035e-05, "loss": 0.0143, "step": 26131 }, { "epoch": 2.63, "learning_rate": 1.527579252934375e-05, "loss": 0.387, "step": 26132 }, { "epoch": 2.63, "learning_rate": 1.5273707483786427e-05, "loss": 0.4491, "step": 26133 }, { "epoch": 2.63, "learning_rate": 1.5271622531941602e-05, "loss": 0.2072, "step": 26134 }, { "epoch": 2.63, "learning_rate": 1.526953767382258e-05, "loss": 1.1962, "step": 26135 }, { "epoch": 2.63, "learning_rate": 1.526745290944262e-05, "loss": 0.4466, "step": 26136 }, { "epoch": 2.63, "learning_rate": 1.526536823881498e-05, "loss": 0.1145, "step": 26137 }, { "epoch": 2.63, "learning_rate": 1.5263283661952938e-05, "loss": 0.0789, "step": 26138 }, { "epoch": 2.63, "learning_rate": 1.526119917886975e-05, "loss": 0.2545, "step": 26139 }, { "epoch": 2.63, "learning_rate": 1.525911478957868e-05, "loss": 0.4899, "step": 26140 }, { "epoch": 2.63, "learning_rate": 1.5257030494092999e-05, "loss": 0.2754, "step": 26141 }, { "epoch": 2.63, "learning_rate": 1.5254946292425968e-05, "loss": 0.3815, "step": 26142 }, { "epoch": 2.63, "learning_rate": 1.5252862184590834e-05, "loss": 0.1908, "step": 26143 }, { "epoch": 2.63, "learning_rate": 1.5250778170600904e-05, "loss": 0.0976, "step": 26144 }, { "epoch": 2.63, "learning_rate": 1.5248694250469393e-05, "loss": 0.4294, "step": 26145 }, { "epoch": 2.63, "learning_rate": 1.5246610424209573e-05, "loss": 0.4828, "step": 26146 }, { "epoch": 2.63, "learning_rate": 1.5244526691834733e-05, "loss": 0.3851, "step": 26147 }, { "epoch": 2.63, "learning_rate": 1.5242443053358101e-05, "loss": 0.1642, "step": 26148 }, { "epoch": 2.63, "learning_rate": 1.5240359508792935e-05, "loss": 0.2711, "step": 26149 }, { "epoch": 2.63, "learning_rate": 1.5238276058152519e-05, "loss": 0.0954, "step": 26150 }, { "epoch": 2.63, "learning_rate": 1.5236192701450111e-05, "loss": 0.6588, "step": 26151 }, { "epoch": 2.63, "learning_rate": 1.5234109438698931e-05, "loss": 0.7007, "step": 26152 }, { "epoch": 2.63, "learning_rate": 1.5232026269912273e-05, "loss": 0.5146, "step": 26153 }, { "epoch": 2.63, "learning_rate": 1.5229943195103383e-05, "loss": 0.3761, "step": 26154 }, { "epoch": 2.63, "learning_rate": 1.5227860214285515e-05, "loss": 0.5633, "step": 26155 }, { "epoch": 2.63, "learning_rate": 1.5225777327471926e-05, "loss": 0.1125, "step": 26156 }, { "epoch": 2.63, "learning_rate": 1.5223694534675867e-05, "loss": 0.207, "step": 26157 }, { "epoch": 2.63, "learning_rate": 1.5221611835910598e-05, "loss": 0.1855, "step": 26158 }, { "epoch": 2.63, "learning_rate": 1.5219529231189369e-05, "loss": 0.2304, "step": 26159 }, { "epoch": 2.63, "learning_rate": 1.5217446720525428e-05, "loss": 0.0877, "step": 26160 }, { "epoch": 2.63, "learning_rate": 1.5215364303932038e-05, "loss": 1.017, "step": 26161 }, { "epoch": 2.63, "learning_rate": 1.5213281981422443e-05, "loss": 0.643, "step": 26162 }, { "epoch": 2.63, "learning_rate": 1.5211199753009883e-05, "loss": 0.4491, "step": 26163 }, { "epoch": 2.63, "learning_rate": 1.5209117618707637e-05, "loss": 0.5851, "step": 26164 }, { "epoch": 2.63, "learning_rate": 1.5207035578528954e-05, "loss": 0.6903, "step": 26165 }, { "epoch": 2.63, "learning_rate": 1.520495363248704e-05, "loss": 0.1669, "step": 26166 }, { "epoch": 2.63, "learning_rate": 1.5202871780595185e-05, "loss": 0.5172, "step": 26167 }, { "epoch": 2.63, "learning_rate": 1.5200790022866628e-05, "loss": 0.4857, "step": 26168 }, { "epoch": 2.63, "learning_rate": 1.5198708359314611e-05, "loss": 0.2049, "step": 26169 }, { "epoch": 2.63, "learning_rate": 1.5196626789952382e-05, "loss": 0.4133, "step": 26170 }, { "epoch": 2.63, "learning_rate": 1.5194545314793193e-05, "loss": 0.4204, "step": 26171 }, { "epoch": 2.63, "learning_rate": 1.5192463933850283e-05, "loss": 0.2261, "step": 26172 }, { "epoch": 2.63, "learning_rate": 1.5190382647136895e-05, "loss": 0.7015, "step": 26173 }, { "epoch": 2.63, "learning_rate": 1.5188301454666286e-05, "loss": 0.3687, "step": 26174 }, { "epoch": 2.63, "learning_rate": 1.5186220356451676e-05, "loss": 0.0956, "step": 26175 }, { "epoch": 2.63, "learning_rate": 1.5184139352506348e-05, "loss": 0.2247, "step": 26176 }, { "epoch": 2.63, "learning_rate": 1.5182058442843509e-05, "loss": 0.483, "step": 26177 }, { "epoch": 2.63, "learning_rate": 1.51799776274764e-05, "loss": 0.3655, "step": 26178 }, { "epoch": 2.63, "learning_rate": 1.5177896906418303e-05, "loss": 0.3944, "step": 26179 }, { "epoch": 2.63, "learning_rate": 1.5175816279682415e-05, "loss": 0.4299, "step": 26180 }, { "epoch": 2.63, "learning_rate": 1.5173735747281988e-05, "loss": 0.1144, "step": 26181 }, { "epoch": 2.63, "learning_rate": 1.5171655309230268e-05, "loss": 0.7086, "step": 26182 }, { "epoch": 2.63, "learning_rate": 1.5169574965540514e-05, "loss": 0.2129, "step": 26183 }, { "epoch": 2.63, "learning_rate": 1.5167494716225914e-05, "loss": 0.7927, "step": 26184 }, { "epoch": 2.63, "learning_rate": 1.5165414561299747e-05, "loss": 0.2012, "step": 26185 }, { "epoch": 2.63, "learning_rate": 1.516333450077524e-05, "loss": 0.4926, "step": 26186 }, { "epoch": 2.63, "learning_rate": 1.5161254534665627e-05, "loss": 0.5214, "step": 26187 }, { "epoch": 2.63, "learning_rate": 1.5159174662984146e-05, "loss": 0.7319, "step": 26188 }, { "epoch": 2.63, "learning_rate": 1.5157094885744021e-05, "loss": 0.3723, "step": 26189 }, { "epoch": 2.63, "learning_rate": 1.5155015202958522e-05, "loss": 0.3833, "step": 26190 }, { "epoch": 2.63, "learning_rate": 1.5152935614640847e-05, "loss": 0.1243, "step": 26191 }, { "epoch": 2.63, "learning_rate": 1.5150856120804226e-05, "loss": 0.214, "step": 26192 }, { "epoch": 2.63, "learning_rate": 1.5148776721461931e-05, "loss": 0.5493, "step": 26193 }, { "epoch": 2.63, "learning_rate": 1.5146697416627164e-05, "loss": 0.0999, "step": 26194 }, { "epoch": 2.63, "learning_rate": 1.5144618206313151e-05, "loss": 0.1675, "step": 26195 }, { "epoch": 2.63, "learning_rate": 1.5142539090533146e-05, "loss": 0.185, "step": 26196 }, { "epoch": 2.63, "learning_rate": 1.5140460069300384e-05, "loss": 0.0862, "step": 26197 }, { "epoch": 2.63, "learning_rate": 1.5138381142628056e-05, "loss": 0.7386, "step": 26198 }, { "epoch": 2.63, "learning_rate": 1.5136302310529432e-05, "loss": 0.4235, "step": 26199 }, { "epoch": 2.63, "learning_rate": 1.5134223573017726e-05, "loss": 0.3206, "step": 26200 }, { "epoch": 2.63, "learning_rate": 1.5132144930106162e-05, "loss": 0.4997, "step": 26201 }, { "epoch": 2.64, "learning_rate": 1.513006638180797e-05, "loss": 0.2336, "step": 26202 }, { "epoch": 2.64, "learning_rate": 1.5127987928136383e-05, "loss": 0.4217, "step": 26203 }, { "epoch": 2.64, "learning_rate": 1.5125909569104622e-05, "loss": 0.2272, "step": 26204 }, { "epoch": 2.64, "learning_rate": 1.5123831304725914e-05, "loss": 0.3104, "step": 26205 }, { "epoch": 2.64, "learning_rate": 1.5121753135013484e-05, "loss": 0.1367, "step": 26206 }, { "epoch": 2.64, "learning_rate": 1.5119675059980546e-05, "loss": 0.5602, "step": 26207 }, { "epoch": 2.64, "learning_rate": 1.5117597079640358e-05, "loss": 0.1579, "step": 26208 }, { "epoch": 2.64, "learning_rate": 1.5115519194006108e-05, "loss": 0.4378, "step": 26209 }, { "epoch": 2.64, "learning_rate": 1.5113441403091023e-05, "loss": 0.7704, "step": 26210 }, { "epoch": 2.64, "learning_rate": 1.5111363706908352e-05, "loss": 0.2615, "step": 26211 }, { "epoch": 2.64, "learning_rate": 1.5109286105471294e-05, "loss": 0.1158, "step": 26212 }, { "epoch": 2.64, "learning_rate": 1.5107208598793057e-05, "loss": 0.2841, "step": 26213 }, { "epoch": 2.64, "learning_rate": 1.5105131186886892e-05, "loss": 0.2334, "step": 26214 }, { "epoch": 2.64, "learning_rate": 1.5103053869766016e-05, "loss": 0.3591, "step": 26215 }, { "epoch": 2.64, "learning_rate": 1.510097664744362e-05, "loss": 0.2419, "step": 26216 }, { "epoch": 2.64, "learning_rate": 1.509889951993295e-05, "loss": 0.3169, "step": 26217 }, { "epoch": 2.64, "learning_rate": 1.5096822487247213e-05, "loss": 0.1734, "step": 26218 }, { "epoch": 2.64, "learning_rate": 1.5094745549399626e-05, "loss": 0.1589, "step": 26219 }, { "epoch": 2.64, "learning_rate": 1.5092668706403414e-05, "loss": 0.4918, "step": 26220 }, { "epoch": 2.64, "learning_rate": 1.5090591958271771e-05, "loss": 0.2675, "step": 26221 }, { "epoch": 2.64, "learning_rate": 1.5088515305017954e-05, "loss": 0.5814, "step": 26222 }, { "epoch": 2.64, "learning_rate": 1.5086438746655144e-05, "loss": 0.4151, "step": 26223 }, { "epoch": 2.64, "learning_rate": 1.508436228319655e-05, "loss": 0.7853, "step": 26224 }, { "epoch": 2.64, "learning_rate": 1.5082285914655423e-05, "loss": 0.0829, "step": 26225 }, { "epoch": 2.64, "learning_rate": 1.5080209641044942e-05, "loss": 0.5817, "step": 26226 }, { "epoch": 2.64, "learning_rate": 1.507813346237832e-05, "loss": 0.228, "step": 26227 }, { "epoch": 2.64, "learning_rate": 1.5076057378668795e-05, "loss": 0.2907, "step": 26228 }, { "epoch": 2.64, "learning_rate": 1.5073981389929567e-05, "loss": 0.4843, "step": 26229 }, { "epoch": 2.64, "learning_rate": 1.5071905496173827e-05, "loss": 0.3802, "step": 26230 }, { "epoch": 2.64, "learning_rate": 1.5069829697414811e-05, "loss": 0.7652, "step": 26231 }, { "epoch": 2.64, "learning_rate": 1.5067753993665719e-05, "loss": 0.0438, "step": 26232 }, { "epoch": 2.64, "learning_rate": 1.5065678384939763e-05, "loss": 0.2152, "step": 26233 }, { "epoch": 2.64, "learning_rate": 1.5063602871250142e-05, "loss": 0.3469, "step": 26234 }, { "epoch": 2.64, "learning_rate": 1.5061527452610075e-05, "loss": 0.381, "step": 26235 }, { "epoch": 2.64, "learning_rate": 1.5059452129032763e-05, "loss": 0.088, "step": 26236 }, { "epoch": 2.64, "learning_rate": 1.5057376900531413e-05, "loss": 0.0999, "step": 26237 }, { "epoch": 2.64, "learning_rate": 1.5055301767119233e-05, "loss": 0.1278, "step": 26238 }, { "epoch": 2.64, "learning_rate": 1.5053226728809414e-05, "loss": 0.1741, "step": 26239 }, { "epoch": 2.64, "learning_rate": 1.5051151785615202e-05, "loss": 1.1312, "step": 26240 }, { "epoch": 2.64, "learning_rate": 1.504907693754976e-05, "loss": 0.5194, "step": 26241 }, { "epoch": 2.64, "learning_rate": 1.5047002184626288e-05, "loss": 0.2148, "step": 26242 }, { "epoch": 2.64, "learning_rate": 1.5044927526858035e-05, "loss": 0.4522, "step": 26243 }, { "epoch": 2.64, "learning_rate": 1.5042852964258146e-05, "loss": 0.4379, "step": 26244 }, { "epoch": 2.64, "learning_rate": 1.5040778496839869e-05, "loss": 0.3766, "step": 26245 }, { "epoch": 2.64, "learning_rate": 1.5038704124616383e-05, "loss": 0.7201, "step": 26246 }, { "epoch": 2.64, "learning_rate": 1.5036629847600892e-05, "loss": 0.1144, "step": 26247 }, { "epoch": 2.64, "learning_rate": 1.5034555665806594e-05, "loss": 0.8437, "step": 26248 }, { "epoch": 2.64, "learning_rate": 1.5032481579246695e-05, "loss": 0.4543, "step": 26249 }, { "epoch": 2.64, "learning_rate": 1.5030407587934386e-05, "loss": 0.0801, "step": 26250 }, { "epoch": 2.64, "learning_rate": 1.502833369188287e-05, "loss": 0.3281, "step": 26251 }, { "epoch": 2.64, "learning_rate": 1.502625989110534e-05, "loss": 0.195, "step": 26252 }, { "epoch": 2.64, "learning_rate": 1.502418618561499e-05, "loss": 0.5707, "step": 26253 }, { "epoch": 2.64, "learning_rate": 1.502211257542504e-05, "loss": 0.5456, "step": 26254 }, { "epoch": 2.64, "learning_rate": 1.502003906054866e-05, "loss": 0.5572, "step": 26255 }, { "epoch": 2.64, "learning_rate": 1.501796564099904e-05, "loss": 0.4705, "step": 26256 }, { "epoch": 2.64, "learning_rate": 1.5015892316789407e-05, "loss": 0.1325, "step": 26257 }, { "epoch": 2.64, "learning_rate": 1.5013819087932925e-05, "loss": 0.6863, "step": 26258 }, { "epoch": 2.64, "learning_rate": 1.5011745954442788e-05, "loss": 0.5818, "step": 26259 }, { "epoch": 2.64, "learning_rate": 1.5009672916332209e-05, "loss": 0.6373, "step": 26260 }, { "epoch": 2.64, "learning_rate": 1.500759997361438e-05, "loss": 0.0949, "step": 26261 }, { "epoch": 2.64, "learning_rate": 1.5005527126302458e-05, "loss": 0.1375, "step": 26262 }, { "epoch": 2.64, "learning_rate": 1.5003454374409668e-05, "loss": 0.2014, "step": 26263 }, { "epoch": 2.64, "learning_rate": 1.500138171794919e-05, "loss": 0.1457, "step": 26264 }, { "epoch": 2.64, "learning_rate": 1.4999309156934214e-05, "loss": 0.3921, "step": 26265 }, { "epoch": 2.64, "learning_rate": 1.4997236691377929e-05, "loss": 0.1927, "step": 26266 }, { "epoch": 2.64, "learning_rate": 1.4995164321293517e-05, "loss": 0.2338, "step": 26267 }, { "epoch": 2.64, "learning_rate": 1.4993092046694174e-05, "loss": 0.3012, "step": 26268 }, { "epoch": 2.64, "learning_rate": 1.4991019867593086e-05, "loss": 0.7652, "step": 26269 }, { "epoch": 2.64, "learning_rate": 1.4988947784003425e-05, "loss": 0.607, "step": 26270 }, { "epoch": 2.64, "learning_rate": 1.4986875795938402e-05, "loss": 0.1361, "step": 26271 }, { "epoch": 2.64, "learning_rate": 1.4984803903411202e-05, "loss": 0.2402, "step": 26272 }, { "epoch": 2.64, "learning_rate": 1.4982732106434974e-05, "loss": 1.2232, "step": 26273 }, { "epoch": 2.64, "learning_rate": 1.4980660405022937e-05, "loss": 0.1836, "step": 26274 }, { "epoch": 2.64, "learning_rate": 1.4978588799188276e-05, "loss": 0.4452, "step": 26275 }, { "epoch": 2.64, "learning_rate": 1.4976517288944136e-05, "loss": 0.331, "step": 26276 }, { "epoch": 2.64, "learning_rate": 1.4974445874303739e-05, "loss": 0.4372, "step": 26277 }, { "epoch": 2.64, "learning_rate": 1.4972374555280251e-05, "loss": 0.2418, "step": 26278 }, { "epoch": 2.64, "learning_rate": 1.4970303331886856e-05, "loss": 0.2435, "step": 26279 }, { "epoch": 2.64, "learning_rate": 1.4968232204136732e-05, "loss": 0.5435, "step": 26280 }, { "epoch": 2.64, "learning_rate": 1.496616117204306e-05, "loss": 0.2144, "step": 26281 }, { "epoch": 2.64, "learning_rate": 1.4964090235619022e-05, "loss": 0.4559, "step": 26282 }, { "epoch": 2.64, "learning_rate": 1.4962019394877789e-05, "loss": 0.4722, "step": 26283 }, { "epoch": 2.64, "learning_rate": 1.4959948649832547e-05, "loss": 0.2144, "step": 26284 }, { "epoch": 2.64, "learning_rate": 1.4957878000496456e-05, "loss": 0.2451, "step": 26285 }, { "epoch": 2.64, "learning_rate": 1.495580744688273e-05, "loss": 0.4019, "step": 26286 }, { "epoch": 2.64, "learning_rate": 1.4953736989004516e-05, "loss": 0.1314, "step": 26287 }, { "epoch": 2.64, "learning_rate": 1.4951666626874981e-05, "loss": 0.1934, "step": 26288 }, { "epoch": 2.64, "learning_rate": 1.4949596360507337e-05, "loss": 0.6304, "step": 26289 }, { "epoch": 2.64, "learning_rate": 1.4947526189914726e-05, "loss": 0.5529, "step": 26290 }, { "epoch": 2.64, "learning_rate": 1.4945456115110321e-05, "loss": 0.2377, "step": 26291 }, { "epoch": 2.64, "learning_rate": 1.4943386136107315e-05, "loss": 0.0691, "step": 26292 }, { "epoch": 2.64, "learning_rate": 1.4941316252918886e-05, "loss": 0.5267, "step": 26293 }, { "epoch": 2.64, "learning_rate": 1.4939246465558166e-05, "loss": 0.271, "step": 26294 }, { "epoch": 2.64, "learning_rate": 1.4937176774038368e-05, "loss": 0.5769, "step": 26295 }, { "epoch": 2.64, "learning_rate": 1.4935107178372644e-05, "loss": 0.444, "step": 26296 }, { "epoch": 2.64, "learning_rate": 1.4933037678574168e-05, "loss": 0.5024, "step": 26297 }, { "epoch": 2.64, "learning_rate": 1.4930968274656106e-05, "loss": 0.2864, "step": 26298 }, { "epoch": 2.64, "learning_rate": 1.492889896663162e-05, "loss": 0.2442, "step": 26299 }, { "epoch": 2.65, "learning_rate": 1.492682975451391e-05, "loss": 0.8658, "step": 26300 }, { "epoch": 2.65, "learning_rate": 1.4924760638316111e-05, "loss": 0.608, "step": 26301 }, { "epoch": 2.65, "learning_rate": 1.4922691618051389e-05, "loss": 0.7816, "step": 26302 }, { "epoch": 2.65, "learning_rate": 1.4920622693732932e-05, "loss": 0.1586, "step": 26303 }, { "epoch": 2.65, "learning_rate": 1.4918553865373907e-05, "loss": 0.4982, "step": 26304 }, { "epoch": 2.65, "learning_rate": 1.4916485132987445e-05, "loss": 0.5249, "step": 26305 }, { "epoch": 2.65, "learning_rate": 1.4914416496586746e-05, "loss": 0.3495, "step": 26306 }, { "epoch": 2.65, "learning_rate": 1.4912347956184972e-05, "loss": 0.5849, "step": 26307 }, { "epoch": 2.65, "learning_rate": 1.4910279511795253e-05, "loss": 0.7541, "step": 26308 }, { "epoch": 2.65, "learning_rate": 1.4908211163430788e-05, "loss": 0.3067, "step": 26309 }, { "epoch": 2.65, "learning_rate": 1.4906142911104724e-05, "loss": 0.5199, "step": 26310 }, { "epoch": 2.65, "learning_rate": 1.4904074754830227e-05, "loss": 0.5296, "step": 26311 }, { "epoch": 2.65, "learning_rate": 1.4902006694620453e-05, "loss": 0.2655, "step": 26312 }, { "epoch": 2.65, "learning_rate": 1.4899938730488564e-05, "loss": 0.3954, "step": 26313 }, { "epoch": 2.65, "learning_rate": 1.4897870862447723e-05, "loss": 1.1259, "step": 26314 }, { "epoch": 2.65, "learning_rate": 1.4895803090511088e-05, "loss": 0.3246, "step": 26315 }, { "epoch": 2.65, "learning_rate": 1.4893735414691812e-05, "loss": 0.3463, "step": 26316 }, { "epoch": 2.65, "learning_rate": 1.4891667835003048e-05, "loss": 0.4881, "step": 26317 }, { "epoch": 2.65, "learning_rate": 1.4889600351457985e-05, "loss": 0.1894, "step": 26318 }, { "epoch": 2.65, "learning_rate": 1.4887532964069746e-05, "loss": 0.5362, "step": 26319 }, { "epoch": 2.65, "learning_rate": 1.4885465672851487e-05, "loss": 0.2799, "step": 26320 }, { "epoch": 2.65, "learning_rate": 1.4883398477816397e-05, "loss": 0.9047, "step": 26321 }, { "epoch": 2.65, "learning_rate": 1.4881331378977595e-05, "loss": 0.1617, "step": 26322 }, { "epoch": 2.65, "learning_rate": 1.4879264376348241e-05, "loss": 0.5761, "step": 26323 }, { "epoch": 2.65, "learning_rate": 1.4877197469941503e-05, "loss": 0.1073, "step": 26324 }, { "epoch": 2.65, "learning_rate": 1.487513065977053e-05, "loss": 0.2984, "step": 26325 }, { "epoch": 2.65, "learning_rate": 1.487306394584847e-05, "loss": 0.2454, "step": 26326 }, { "epoch": 2.65, "learning_rate": 1.487099732818848e-05, "loss": 0.3386, "step": 26327 }, { "epoch": 2.65, "learning_rate": 1.4868930806803705e-05, "loss": 0.2367, "step": 26328 }, { "epoch": 2.65, "learning_rate": 1.48668643817073e-05, "loss": 0.1292, "step": 26329 }, { "epoch": 2.65, "learning_rate": 1.4864798052912413e-05, "loss": 0.1024, "step": 26330 }, { "epoch": 2.65, "learning_rate": 1.486273182043218e-05, "loss": 0.4768, "step": 26331 }, { "epoch": 2.65, "learning_rate": 1.4860665684279789e-05, "loss": 0.4823, "step": 26332 }, { "epoch": 2.65, "learning_rate": 1.485859964446835e-05, "loss": 0.1323, "step": 26333 }, { "epoch": 2.65, "learning_rate": 1.4856533701011007e-05, "loss": 0.4247, "step": 26334 }, { "epoch": 2.65, "learning_rate": 1.4854467853920937e-05, "loss": 0.2584, "step": 26335 }, { "epoch": 2.65, "learning_rate": 1.4852402103211284e-05, "loss": 0.1049, "step": 26336 }, { "epoch": 2.65, "learning_rate": 1.485033644889516e-05, "loss": 0.5682, "step": 26337 }, { "epoch": 2.65, "learning_rate": 1.484827089098574e-05, "loss": 0.2672, "step": 26338 }, { "epoch": 2.65, "learning_rate": 1.4846205429496174e-05, "loss": 0.514, "step": 26339 }, { "epoch": 2.65, "learning_rate": 1.4844140064439568e-05, "loss": 0.8983, "step": 26340 }, { "epoch": 2.65, "learning_rate": 1.4842074795829101e-05, "loss": 0.0786, "step": 26341 }, { "epoch": 2.65, "learning_rate": 1.4840009623677903e-05, "loss": 0.5479, "step": 26342 }, { "epoch": 2.65, "learning_rate": 1.4837944547999117e-05, "loss": 0.735, "step": 26343 }, { "epoch": 2.65, "learning_rate": 1.4835879568805884e-05, "loss": 0.8715, "step": 26344 }, { "epoch": 2.65, "learning_rate": 1.4833814686111344e-05, "loss": 0.1615, "step": 26345 }, { "epoch": 2.65, "learning_rate": 1.483174989992864e-05, "loss": 0.4104, "step": 26346 }, { "epoch": 2.65, "learning_rate": 1.4829685210270907e-05, "loss": 0.1874, "step": 26347 }, { "epoch": 2.65, "learning_rate": 1.4827620617151287e-05, "loss": 0.1247, "step": 26348 }, { "epoch": 2.65, "learning_rate": 1.4825556120582904e-05, "loss": 0.7616, "step": 26349 }, { "epoch": 2.65, "learning_rate": 1.4823491720578934e-05, "loss": 0.3116, "step": 26350 }, { "epoch": 2.65, "learning_rate": 1.4821427417152468e-05, "loss": 0.4553, "step": 26351 }, { "epoch": 2.65, "learning_rate": 1.4819363210316673e-05, "loss": 0.3921, "step": 26352 }, { "epoch": 2.65, "learning_rate": 1.4817299100084686e-05, "loss": 0.0792, "step": 26353 }, { "epoch": 2.65, "learning_rate": 1.4815235086469612e-05, "loss": 0.2603, "step": 26354 }, { "epoch": 2.65, "learning_rate": 1.4813171169484614e-05, "loss": 0.1258, "step": 26355 }, { "epoch": 2.65, "learning_rate": 1.4811107349142819e-05, "loss": 0.1094, "step": 26356 }, { "epoch": 2.65, "learning_rate": 1.4809043625457357e-05, "loss": 0.2657, "step": 26357 }, { "epoch": 2.65, "learning_rate": 1.4806979998441366e-05, "loss": 0.0918, "step": 26358 }, { "epoch": 2.65, "learning_rate": 1.4804916468107972e-05, "loss": 0.71, "step": 26359 }, { "epoch": 2.65, "learning_rate": 1.4802853034470309e-05, "loss": 0.2793, "step": 26360 }, { "epoch": 2.65, "learning_rate": 1.4800789697541507e-05, "loss": 0.4101, "step": 26361 }, { "epoch": 2.65, "learning_rate": 1.4798726457334697e-05, "loss": 0.3421, "step": 26362 }, { "epoch": 2.65, "learning_rate": 1.4796663313862999e-05, "loss": 0.3026, "step": 26363 }, { "epoch": 2.65, "learning_rate": 1.4794600267139577e-05, "loss": 0.5428, "step": 26364 }, { "epoch": 2.65, "learning_rate": 1.4792537317177522e-05, "loss": 0.4636, "step": 26365 }, { "epoch": 2.65, "learning_rate": 1.479047446398996e-05, "loss": 0.5932, "step": 26366 }, { "epoch": 2.65, "learning_rate": 1.4788411707590047e-05, "loss": 0.3176, "step": 26367 }, { "epoch": 2.65, "learning_rate": 1.4786349047990905e-05, "loss": 0.5189, "step": 26368 }, { "epoch": 2.65, "learning_rate": 1.4784286485205626e-05, "loss": 0.1198, "step": 26369 }, { "epoch": 2.65, "learning_rate": 1.4782224019247373e-05, "loss": 0.4977, "step": 26370 }, { "epoch": 2.65, "learning_rate": 1.4780161650129267e-05, "loss": 0.3078, "step": 26371 }, { "epoch": 2.65, "learning_rate": 1.4778099377864402e-05, "loss": 0.0675, "step": 26372 }, { "epoch": 2.65, "learning_rate": 1.477603720246593e-05, "loss": 0.2182, "step": 26373 }, { "epoch": 2.65, "learning_rate": 1.4773975123946968e-05, "loss": 0.3313, "step": 26374 }, { "epoch": 2.65, "learning_rate": 1.4771913142320636e-05, "loss": 0.6634, "step": 26375 }, { "epoch": 2.65, "learning_rate": 1.4769851257600056e-05, "loss": 0.0573, "step": 26376 }, { "epoch": 2.65, "learning_rate": 1.4767789469798348e-05, "loss": 0.127, "step": 26377 }, { "epoch": 2.65, "learning_rate": 1.4765727778928632e-05, "loss": 0.1968, "step": 26378 }, { "epoch": 2.65, "learning_rate": 1.4763666185004031e-05, "loss": 0.1134, "step": 26379 }, { "epoch": 2.65, "learning_rate": 1.4761604688037648e-05, "loss": 0.9154, "step": 26380 }, { "epoch": 2.65, "learning_rate": 1.4759543288042631e-05, "loss": 0.1129, "step": 26381 }, { "epoch": 2.65, "learning_rate": 1.4757481985032092e-05, "loss": 0.115, "step": 26382 }, { "epoch": 2.65, "learning_rate": 1.4755420779019116e-05, "loss": 0.364, "step": 26383 }, { "epoch": 2.65, "learning_rate": 1.4753359670016857e-05, "loss": 0.3551, "step": 26384 }, { "epoch": 2.65, "learning_rate": 1.4751298658038424e-05, "loss": 0.8249, "step": 26385 }, { "epoch": 2.65, "learning_rate": 1.4749237743096908e-05, "loss": 0.3405, "step": 26386 }, { "epoch": 2.65, "learning_rate": 1.4747176925205448e-05, "loss": 0.4343, "step": 26387 }, { "epoch": 2.65, "learning_rate": 1.4745116204377152e-05, "loss": 0.5235, "step": 26388 }, { "epoch": 2.65, "learning_rate": 1.4743055580625134e-05, "loss": 0.4287, "step": 26389 }, { "epoch": 2.65, "learning_rate": 1.4740995053962505e-05, "loss": 0.3277, "step": 26390 }, { "epoch": 2.65, "learning_rate": 1.4738934624402384e-05, "loss": 0.0954, "step": 26391 }, { "epoch": 2.65, "learning_rate": 1.4736874291957874e-05, "loss": 0.1518, "step": 26392 }, { "epoch": 2.65, "learning_rate": 1.473481405664209e-05, "loss": 0.6771, "step": 26393 }, { "epoch": 2.65, "learning_rate": 1.4732753918468139e-05, "loss": 0.5372, "step": 26394 }, { "epoch": 2.65, "learning_rate": 1.4730693877449125e-05, "loss": 0.1726, "step": 26395 }, { "epoch": 2.65, "learning_rate": 1.4728633933598188e-05, "loss": 0.4538, "step": 26396 }, { "epoch": 2.65, "learning_rate": 1.4726574086928406e-05, "loss": 0.3029, "step": 26397 }, { "epoch": 2.65, "learning_rate": 1.4724514337452886e-05, "loss": 0.2331, "step": 26398 }, { "epoch": 2.66, "learning_rate": 1.4722454685184754e-05, "loss": 0.105, "step": 26399 }, { "epoch": 2.66, "learning_rate": 1.472039513013712e-05, "loss": 0.1371, "step": 26400 }, { "epoch": 2.66, "learning_rate": 1.4718335672323056e-05, "loss": 0.4792, "step": 26401 }, { "epoch": 2.66, "learning_rate": 1.4716276311755703e-05, "loss": 0.6278, "step": 26402 }, { "epoch": 2.66, "learning_rate": 1.4714217048448166e-05, "loss": 0.4049, "step": 26403 }, { "epoch": 2.66, "learning_rate": 1.4712157882413512e-05, "loss": 0.8194, "step": 26404 }, { "epoch": 2.66, "learning_rate": 1.471009881366488e-05, "loss": 0.0404, "step": 26405 }, { "epoch": 2.66, "learning_rate": 1.4708039842215363e-05, "loss": 0.1458, "step": 26406 }, { "epoch": 2.66, "learning_rate": 1.4705980968078065e-05, "loss": 0.1927, "step": 26407 }, { "epoch": 2.66, "learning_rate": 1.4703922191266084e-05, "loss": 0.3863, "step": 26408 }, { "epoch": 2.66, "learning_rate": 1.4701863511792514e-05, "loss": 0.204, "step": 26409 }, { "epoch": 2.66, "learning_rate": 1.4699804929670483e-05, "loss": 0.162, "step": 26410 }, { "epoch": 2.66, "learning_rate": 1.4697746444913067e-05, "loss": 0.2645, "step": 26411 }, { "epoch": 2.66, "learning_rate": 1.4695688057533357e-05, "loss": 0.344, "step": 26412 }, { "epoch": 2.66, "learning_rate": 1.4693629767544477e-05, "loss": 0.3006, "step": 26413 }, { "epoch": 2.66, "learning_rate": 1.4691571574959526e-05, "loss": 0.2303, "step": 26414 }, { "epoch": 2.66, "learning_rate": 1.468951347979157e-05, "loss": 0.2191, "step": 26415 }, { "epoch": 2.66, "learning_rate": 1.4687455482053736e-05, "loss": 0.1067, "step": 26416 }, { "epoch": 2.66, "learning_rate": 1.468539758175912e-05, "loss": 0.1937, "step": 26417 }, { "epoch": 2.66, "learning_rate": 1.4683339778920787e-05, "loss": 0.6069, "step": 26418 }, { "epoch": 2.66, "learning_rate": 1.4681282073551862e-05, "loss": 0.1255, "step": 26419 }, { "epoch": 2.66, "learning_rate": 1.467922446566543e-05, "loss": 0.2733, "step": 26420 }, { "epoch": 2.66, "learning_rate": 1.4677166955274587e-05, "loss": 0.259, "step": 26421 }, { "epoch": 2.66, "learning_rate": 1.4675109542392426e-05, "loss": 0.4668, "step": 26422 }, { "epoch": 2.66, "learning_rate": 1.4673052227032035e-05, "loss": 0.2752, "step": 26423 }, { "epoch": 2.66, "learning_rate": 1.4670995009206512e-05, "loss": 0.3214, "step": 26424 }, { "epoch": 2.66, "learning_rate": 1.466893788892894e-05, "loss": 0.6755, "step": 26425 }, { "epoch": 2.66, "learning_rate": 1.4666880866212418e-05, "loss": 0.2487, "step": 26426 }, { "epoch": 2.66, "learning_rate": 1.4664823941070022e-05, "loss": 0.4996, "step": 26427 }, { "epoch": 2.66, "learning_rate": 1.4662767113514878e-05, "loss": 0.3126, "step": 26428 }, { "epoch": 2.66, "learning_rate": 1.4660710383560035e-05, "loss": 0.5302, "step": 26429 }, { "epoch": 2.66, "learning_rate": 1.4658653751218585e-05, "loss": 0.1332, "step": 26430 }, { "epoch": 2.66, "learning_rate": 1.4656597216503639e-05, "loss": 0.9804, "step": 26431 }, { "epoch": 2.66, "learning_rate": 1.4654540779428269e-05, "loss": 0.8455, "step": 26432 }, { "epoch": 2.66, "learning_rate": 1.4652484440005566e-05, "loss": 0.3935, "step": 26433 }, { "epoch": 2.66, "learning_rate": 1.4650428198248611e-05, "loss": 0.2336, "step": 26434 }, { "epoch": 2.66, "learning_rate": 1.4648372054170495e-05, "loss": 0.2443, "step": 26435 }, { "epoch": 2.66, "learning_rate": 1.4646316007784297e-05, "loss": 0.2309, "step": 26436 }, { "epoch": 2.66, "learning_rate": 1.4644260059103101e-05, "loss": 0.1606, "step": 26437 }, { "epoch": 2.66, "learning_rate": 1.4642204208139994e-05, "loss": 0.0993, "step": 26438 }, { "epoch": 2.66, "learning_rate": 1.464014845490806e-05, "loss": 0.3102, "step": 26439 }, { "epoch": 2.66, "learning_rate": 1.4638092799420374e-05, "loss": 0.4602, "step": 26440 }, { "epoch": 2.66, "learning_rate": 1.4636037241690013e-05, "loss": 0.6032, "step": 26441 }, { "epoch": 2.66, "learning_rate": 1.4633981781730087e-05, "loss": 0.2973, "step": 26442 }, { "epoch": 2.66, "learning_rate": 1.4631926419553644e-05, "loss": 0.142, "step": 26443 }, { "epoch": 2.66, "learning_rate": 1.4629871155173767e-05, "loss": 0.309, "step": 26444 }, { "epoch": 2.66, "learning_rate": 1.462781598860355e-05, "loss": 0.6905, "step": 26445 }, { "epoch": 2.66, "learning_rate": 1.462576091985608e-05, "loss": 0.4265, "step": 26446 }, { "epoch": 2.66, "learning_rate": 1.462370594894439e-05, "loss": 0.5009, "step": 26447 }, { "epoch": 2.66, "learning_rate": 1.4621651075881604e-05, "loss": 0.4534, "step": 26448 }, { "epoch": 2.66, "learning_rate": 1.4619596300680786e-05, "loss": 0.0848, "step": 26449 }, { "epoch": 2.66, "learning_rate": 1.4617541623354986e-05, "loss": 0.2731, "step": 26450 }, { "epoch": 2.66, "learning_rate": 1.461548704391731e-05, "loss": 0.5254, "step": 26451 }, { "epoch": 2.66, "learning_rate": 1.4613432562380824e-05, "loss": 0.8647, "step": 26452 }, { "epoch": 2.66, "learning_rate": 1.4611378178758596e-05, "loss": 0.3577, "step": 26453 }, { "epoch": 2.66, "learning_rate": 1.4609323893063706e-05, "loss": 0.6992, "step": 26454 }, { "epoch": 2.66, "learning_rate": 1.4607269705309223e-05, "loss": 0.0955, "step": 26455 }, { "epoch": 2.66, "learning_rate": 1.4605215615508222e-05, "loss": 0.6037, "step": 26456 }, { "epoch": 2.66, "learning_rate": 1.4603161623673769e-05, "loss": 0.5616, "step": 26457 }, { "epoch": 2.66, "learning_rate": 1.460110772981894e-05, "loss": 0.3568, "step": 26458 }, { "epoch": 2.66, "learning_rate": 1.4599053933956792e-05, "loss": 0.4947, "step": 26459 }, { "epoch": 2.66, "learning_rate": 1.4597000236100427e-05, "loss": 0.2429, "step": 26460 }, { "epoch": 2.66, "learning_rate": 1.4594946636262873e-05, "loss": 0.4839, "step": 26461 }, { "epoch": 2.66, "learning_rate": 1.4592893134457227e-05, "loss": 0.1265, "step": 26462 }, { "epoch": 2.66, "learning_rate": 1.4590839730696547e-05, "loss": 0.1572, "step": 26463 }, { "epoch": 2.66, "learning_rate": 1.4588786424993906e-05, "loss": 0.6289, "step": 26464 }, { "epoch": 2.66, "learning_rate": 1.4586733217362362e-05, "loss": 0.1825, "step": 26465 }, { "epoch": 2.66, "learning_rate": 1.4584680107814981e-05, "loss": 0.4377, "step": 26466 }, { "epoch": 2.66, "learning_rate": 1.4582627096364835e-05, "loss": 0.631, "step": 26467 }, { "epoch": 2.66, "learning_rate": 1.4580574183024986e-05, "loss": 0.6499, "step": 26468 }, { "epoch": 2.66, "learning_rate": 1.4578521367808496e-05, "loss": 0.2712, "step": 26469 }, { "epoch": 2.66, "learning_rate": 1.4576468650728431e-05, "loss": 0.1403, "step": 26470 }, { "epoch": 2.66, "learning_rate": 1.457441603179785e-05, "loss": 0.6558, "step": 26471 }, { "epoch": 2.66, "learning_rate": 1.4572363511029819e-05, "loss": 0.2385, "step": 26472 }, { "epoch": 2.66, "learning_rate": 1.4570311088437385e-05, "loss": 0.333, "step": 26473 }, { "epoch": 2.66, "learning_rate": 1.4568258764033645e-05, "loss": 0.0437, "step": 26474 }, { "epoch": 2.66, "learning_rate": 1.4566206537831623e-05, "loss": 0.139, "step": 26475 }, { "epoch": 2.66, "learning_rate": 1.4564154409844382e-05, "loss": 0.1906, "step": 26476 }, { "epoch": 2.66, "learning_rate": 1.4562102380084998e-05, "loss": 0.6348, "step": 26477 }, { "epoch": 2.66, "learning_rate": 1.4560050448566536e-05, "loss": 0.3811, "step": 26478 }, { "epoch": 2.66, "learning_rate": 1.4557998615302015e-05, "loss": 0.4178, "step": 26479 }, { "epoch": 2.66, "learning_rate": 1.455594688030453e-05, "loss": 0.2808, "step": 26480 }, { "epoch": 2.66, "learning_rate": 1.4553895243587132e-05, "loss": 0.3542, "step": 26481 }, { "epoch": 2.66, "learning_rate": 1.4551843705162851e-05, "loss": 0.2258, "step": 26482 }, { "epoch": 2.66, "learning_rate": 1.4549792265044765e-05, "loss": 0.5053, "step": 26483 }, { "epoch": 2.66, "learning_rate": 1.4547740923245927e-05, "loss": 0.4577, "step": 26484 }, { "epoch": 2.66, "learning_rate": 1.4545689679779387e-05, "loss": 0.7566, "step": 26485 }, { "epoch": 2.66, "learning_rate": 1.4543638534658199e-05, "loss": 0.0296, "step": 26486 }, { "epoch": 2.66, "learning_rate": 1.4541587487895402e-05, "loss": 0.6568, "step": 26487 }, { "epoch": 2.66, "learning_rate": 1.4539536539504087e-05, "loss": 0.2372, "step": 26488 }, { "epoch": 2.66, "learning_rate": 1.4537485689497269e-05, "loss": 0.2458, "step": 26489 }, { "epoch": 2.66, "learning_rate": 1.4535434937887998e-05, "loss": 0.3062, "step": 26490 }, { "epoch": 2.66, "learning_rate": 1.4533384284689346e-05, "loss": 0.2362, "step": 26491 }, { "epoch": 2.66, "learning_rate": 1.4531333729914362e-05, "loss": 0.1489, "step": 26492 }, { "epoch": 2.66, "learning_rate": 1.4529283273576065e-05, "loss": 0.5497, "step": 26493 }, { "epoch": 2.66, "learning_rate": 1.4527232915687535e-05, "loss": 0.3167, "step": 26494 }, { "epoch": 2.66, "learning_rate": 1.4525182656261812e-05, "loss": 0.2909, "step": 26495 }, { "epoch": 2.66, "learning_rate": 1.4523132495311936e-05, "loss": 0.6176, "step": 26496 }, { "epoch": 2.67, "learning_rate": 1.4521082432850958e-05, "loss": 0.1667, "step": 26497 }, { "epoch": 2.67, "learning_rate": 1.4519032468891922e-05, "loss": 0.1644, "step": 26498 }, { "epoch": 2.67, "learning_rate": 1.4516982603447875e-05, "loss": 0.151, "step": 26499 }, { "epoch": 2.67, "learning_rate": 1.451493283653186e-05, "loss": 0.7784, "step": 26500 }, { "epoch": 2.67, "learning_rate": 1.4512883168156924e-05, "loss": 0.1025, "step": 26501 }, { "epoch": 2.67, "learning_rate": 1.4510833598336104e-05, "loss": 0.1518, "step": 26502 }, { "epoch": 2.67, "learning_rate": 1.4508784127082449e-05, "loss": 0.2208, "step": 26503 }, { "epoch": 2.67, "learning_rate": 1.4506734754409001e-05, "loss": 0.3414, "step": 26504 }, { "epoch": 2.67, "learning_rate": 1.450468548032878e-05, "loss": 0.452, "step": 26505 }, { "epoch": 2.67, "learning_rate": 1.4502636304854875e-05, "loss": 0.3057, "step": 26506 }, { "epoch": 2.67, "learning_rate": 1.4500587228000284e-05, "loss": 0.4086, "step": 26507 }, { "epoch": 2.67, "learning_rate": 1.449853824977805e-05, "loss": 0.3144, "step": 26508 }, { "epoch": 2.67, "learning_rate": 1.449648937020123e-05, "loss": 0.161, "step": 26509 }, { "epoch": 2.67, "learning_rate": 1.4494440589282867e-05, "loss": 0.1494, "step": 26510 }, { "epoch": 2.67, "learning_rate": 1.4492391907035964e-05, "loss": 0.1058, "step": 26511 }, { "epoch": 2.67, "learning_rate": 1.4490343323473587e-05, "loss": 0.2481, "step": 26512 }, { "epoch": 2.67, "learning_rate": 1.4488294838608768e-05, "loss": 0.6748, "step": 26513 }, { "epoch": 2.67, "learning_rate": 1.448624645245454e-05, "loss": 0.0725, "step": 26514 }, { "epoch": 2.67, "learning_rate": 1.4484198165023938e-05, "loss": 0.3263, "step": 26515 }, { "epoch": 2.67, "learning_rate": 1.4482149976329993e-05, "loss": 0.8722, "step": 26516 }, { "epoch": 2.67, "learning_rate": 1.4480101886385746e-05, "loss": 0.1185, "step": 26517 }, { "epoch": 2.67, "learning_rate": 1.4478053895204224e-05, "loss": 0.3538, "step": 26518 }, { "epoch": 2.67, "learning_rate": 1.447600600279845e-05, "loss": 0.2071, "step": 26519 }, { "epoch": 2.67, "learning_rate": 1.4473958209181492e-05, "loss": 0.2178, "step": 26520 }, { "epoch": 2.67, "learning_rate": 1.447191051436635e-05, "loss": 0.1425, "step": 26521 }, { "epoch": 2.67, "learning_rate": 1.4469862918366045e-05, "loss": 0.5954, "step": 26522 }, { "epoch": 2.67, "learning_rate": 1.446781542119364e-05, "loss": 0.1409, "step": 26523 }, { "epoch": 2.67, "learning_rate": 1.4465768022862157e-05, "loss": 0.2335, "step": 26524 }, { "epoch": 2.67, "learning_rate": 1.4463720723384595e-05, "loss": 0.5836, "step": 26525 }, { "epoch": 2.67, "learning_rate": 1.4461673522774019e-05, "loss": 0.5282, "step": 26526 }, { "epoch": 2.67, "learning_rate": 1.4459626421043436e-05, "loss": 1.4731, "step": 26527 }, { "epoch": 2.67, "learning_rate": 1.4457579418205881e-05, "loss": 0.4047, "step": 26528 }, { "epoch": 2.67, "learning_rate": 1.445553251427438e-05, "loss": 0.2177, "step": 26529 }, { "epoch": 2.67, "learning_rate": 1.4453485709261955e-05, "loss": 0.5557, "step": 26530 }, { "epoch": 2.67, "learning_rate": 1.4451439003181634e-05, "loss": 0.1197, "step": 26531 }, { "epoch": 2.67, "learning_rate": 1.4449392396046437e-05, "loss": 0.3165, "step": 26532 }, { "epoch": 2.67, "learning_rate": 1.4447345887869394e-05, "loss": 0.6111, "step": 26533 }, { "epoch": 2.67, "learning_rate": 1.4445299478663524e-05, "loss": 0.2001, "step": 26534 }, { "epoch": 2.67, "learning_rate": 1.4443253168441852e-05, "loss": 0.1711, "step": 26535 }, { "epoch": 2.67, "learning_rate": 1.4441206957217398e-05, "loss": 0.2413, "step": 26536 }, { "epoch": 2.67, "learning_rate": 1.4439160845003171e-05, "loss": 0.1089, "step": 26537 }, { "epoch": 2.67, "learning_rate": 1.4437114831812229e-05, "loss": 0.8006, "step": 26538 }, { "epoch": 2.67, "learning_rate": 1.4435068917657557e-05, "loss": 0.627, "step": 26539 }, { "epoch": 2.67, "learning_rate": 1.4433023102552175e-05, "loss": 0.5331, "step": 26540 }, { "epoch": 2.67, "learning_rate": 1.443097738650912e-05, "loss": 0.5923, "step": 26541 }, { "epoch": 2.67, "learning_rate": 1.4428931769541404e-05, "loss": 0.6507, "step": 26542 }, { "epoch": 2.67, "learning_rate": 1.4426886251662041e-05, "loss": 0.2763, "step": 26543 }, { "epoch": 2.67, "learning_rate": 1.4424840832884054e-05, "loss": 0.1481, "step": 26544 }, { "epoch": 2.67, "learning_rate": 1.4422795513220447e-05, "loss": 0.0718, "step": 26545 }, { "epoch": 2.67, "learning_rate": 1.442075029268425e-05, "loss": 0.293, "step": 26546 }, { "epoch": 2.67, "learning_rate": 1.441870517128847e-05, "loss": 0.4337, "step": 26547 }, { "epoch": 2.67, "learning_rate": 1.4416660149046122e-05, "loss": 0.8037, "step": 26548 }, { "epoch": 2.67, "learning_rate": 1.4414615225970223e-05, "loss": 0.3176, "step": 26549 }, { "epoch": 2.67, "learning_rate": 1.4412570402073783e-05, "loss": 0.1196, "step": 26550 }, { "epoch": 2.67, "learning_rate": 1.4410525677369802e-05, "loss": 0.2803, "step": 26551 }, { "epoch": 2.67, "learning_rate": 1.4408481051871326e-05, "loss": 0.1699, "step": 26552 }, { "epoch": 2.67, "learning_rate": 1.4406436525591337e-05, "loss": 0.3052, "step": 26553 }, { "epoch": 2.67, "learning_rate": 1.4404392098542843e-05, "loss": 0.1944, "step": 26554 }, { "epoch": 2.67, "learning_rate": 1.4402347770738877e-05, "loss": 0.3147, "step": 26555 }, { "epoch": 2.67, "learning_rate": 1.4400303542192444e-05, "loss": 0.3206, "step": 26556 }, { "epoch": 2.67, "learning_rate": 1.4398259412916524e-05, "loss": 0.6846, "step": 26557 }, { "epoch": 2.67, "learning_rate": 1.4396215382924157e-05, "loss": 0.3259, "step": 26558 }, { "epoch": 2.67, "learning_rate": 1.4394171452228338e-05, "loss": 0.1679, "step": 26559 }, { "epoch": 2.67, "learning_rate": 1.4392127620842078e-05, "loss": 0.2923, "step": 26560 }, { "epoch": 2.67, "learning_rate": 1.4390083888778378e-05, "loss": 0.2255, "step": 26561 }, { "epoch": 2.67, "learning_rate": 1.4388040256050248e-05, "loss": 0.0594, "step": 26562 }, { "epoch": 2.67, "learning_rate": 1.4385996722670687e-05, "loss": 0.6647, "step": 26563 }, { "epoch": 2.67, "learning_rate": 1.4383953288652706e-05, "loss": 0.2032, "step": 26564 }, { "epoch": 2.67, "learning_rate": 1.4381909954009305e-05, "loss": 0.4868, "step": 26565 }, { "epoch": 2.67, "learning_rate": 1.4379866718753487e-05, "loss": 0.5737, "step": 26566 }, { "epoch": 2.67, "learning_rate": 1.4377823582898255e-05, "loss": 0.541, "step": 26567 }, { "epoch": 2.67, "learning_rate": 1.4375780546456602e-05, "loss": 0.3356, "step": 26568 }, { "epoch": 2.67, "learning_rate": 1.4373737609441544e-05, "loss": 0.4431, "step": 26569 }, { "epoch": 2.67, "learning_rate": 1.4371694771866092e-05, "loss": 0.2252, "step": 26570 }, { "epoch": 2.67, "learning_rate": 1.4369652033743205e-05, "loss": 0.3491, "step": 26571 }, { "epoch": 2.67, "learning_rate": 1.4367609395085917e-05, "loss": 0.4921, "step": 26572 }, { "epoch": 2.67, "learning_rate": 1.4365566855907214e-05, "loss": 0.2367, "step": 26573 }, { "epoch": 2.67, "learning_rate": 1.43635244162201e-05, "loss": 1.1427, "step": 26574 }, { "epoch": 2.67, "learning_rate": 1.4361482076037564e-05, "loss": 0.3926, "step": 26575 }, { "epoch": 2.67, "learning_rate": 1.4359439835372607e-05, "loss": 0.3317, "step": 26576 }, { "epoch": 2.67, "learning_rate": 1.4357397694238228e-05, "loss": 0.3576, "step": 26577 }, { "epoch": 2.67, "learning_rate": 1.4355355652647417e-05, "loss": 0.1339, "step": 26578 }, { "epoch": 2.67, "learning_rate": 1.4353313710613176e-05, "loss": 0.1665, "step": 26579 }, { "epoch": 2.67, "learning_rate": 1.435127186814849e-05, "loss": 0.6261, "step": 26580 }, { "epoch": 2.67, "learning_rate": 1.4349230125266354e-05, "loss": 0.2198, "step": 26581 }, { "epoch": 2.67, "learning_rate": 1.434718848197977e-05, "loss": 0.3142, "step": 26582 }, { "epoch": 2.67, "learning_rate": 1.4345146938301708e-05, "loss": 0.3894, "step": 26583 }, { "epoch": 2.67, "learning_rate": 1.4343105494245199e-05, "loss": 0.1247, "step": 26584 }, { "epoch": 2.67, "learning_rate": 1.43410641498232e-05, "loss": 0.1831, "step": 26585 }, { "epoch": 2.67, "learning_rate": 1.4339022905048702e-05, "loss": 0.4578, "step": 26586 }, { "epoch": 2.67, "learning_rate": 1.4336981759934712e-05, "loss": 0.5023, "step": 26587 }, { "epoch": 2.67, "learning_rate": 1.4334940714494222e-05, "loss": 0.1945, "step": 26588 }, { "epoch": 2.67, "learning_rate": 1.4332899768740191e-05, "loss": 0.6445, "step": 26589 }, { "epoch": 2.67, "learning_rate": 1.4330858922685636e-05, "loss": 0.4152, "step": 26590 }, { "epoch": 2.67, "learning_rate": 1.4328818176343537e-05, "loss": 0.6787, "step": 26591 }, { "epoch": 2.67, "learning_rate": 1.4326777529726874e-05, "loss": 0.1962, "step": 26592 }, { "epoch": 2.67, "learning_rate": 1.4324736982848636e-05, "loss": 1.2631, "step": 26593 }, { "epoch": 2.67, "learning_rate": 1.4322696535721808e-05, "loss": 0.4906, "step": 26594 }, { "epoch": 2.67, "learning_rate": 1.4320656188359378e-05, "loss": 0.1685, "step": 26595 }, { "epoch": 2.68, "learning_rate": 1.4318615940774327e-05, "loss": 0.6225, "step": 26596 }, { "epoch": 2.68, "learning_rate": 1.4316575792979629e-05, "loss": 0.4399, "step": 26597 }, { "epoch": 2.68, "learning_rate": 1.4314535744988295e-05, "loss": 0.1895, "step": 26598 }, { "epoch": 2.68, "learning_rate": 1.4312495796813277e-05, "loss": 0.2111, "step": 26599 }, { "epoch": 2.68, "learning_rate": 1.4310455948467562e-05, "loss": 0.1621, "step": 26600 }, { "epoch": 2.68, "learning_rate": 1.4308416199964142e-05, "loss": 0.1267, "step": 26601 }, { "epoch": 2.68, "learning_rate": 1.4306376551316005e-05, "loss": 0.5396, "step": 26602 }, { "epoch": 2.68, "learning_rate": 1.4304337002536097e-05, "loss": 0.1338, "step": 26603 }, { "epoch": 2.68, "learning_rate": 1.4302297553637426e-05, "loss": 0.4133, "step": 26604 }, { "epoch": 2.68, "learning_rate": 1.4300258204632963e-05, "loss": 0.2573, "step": 26605 }, { "epoch": 2.68, "learning_rate": 1.4298218955535683e-05, "loss": 0.6029, "step": 26606 }, { "epoch": 2.68, "learning_rate": 1.4296179806358567e-05, "loss": 0.3905, "step": 26607 }, { "epoch": 2.68, "learning_rate": 1.429414075711459e-05, "loss": 0.5192, "step": 26608 }, { "epoch": 2.68, "learning_rate": 1.4292101807816722e-05, "loss": 0.3166, "step": 26609 }, { "epoch": 2.68, "learning_rate": 1.4290062958477943e-05, "loss": 0.4734, "step": 26610 }, { "epoch": 2.68, "learning_rate": 1.4288024209111228e-05, "loss": 0.267, "step": 26611 }, { "epoch": 2.68, "learning_rate": 1.428598555972955e-05, "loss": 0.5518, "step": 26612 }, { "epoch": 2.68, "learning_rate": 1.4283947010345881e-05, "loss": 0.366, "step": 26613 }, { "epoch": 2.68, "learning_rate": 1.4281908560973195e-05, "loss": 0.0849, "step": 26614 }, { "epoch": 2.68, "learning_rate": 1.4279870211624453e-05, "loss": 0.2049, "step": 26615 }, { "epoch": 2.68, "learning_rate": 1.4277831962312658e-05, "loss": 0.3024, "step": 26616 }, { "epoch": 2.68, "learning_rate": 1.427579381305075e-05, "loss": 0.5931, "step": 26617 }, { "epoch": 2.68, "learning_rate": 1.4273755763851695e-05, "loss": 0.0858, "step": 26618 }, { "epoch": 2.68, "learning_rate": 1.4271717814728487e-05, "loss": 0.1135, "step": 26619 }, { "epoch": 2.68, "learning_rate": 1.4269679965694099e-05, "loss": 0.1056, "step": 26620 }, { "epoch": 2.68, "learning_rate": 1.4267642216761459e-05, "loss": 0.3163, "step": 26621 }, { "epoch": 2.68, "learning_rate": 1.4265604567943565e-05, "loss": 0.4157, "step": 26622 }, { "epoch": 2.68, "learning_rate": 1.4263567019253386e-05, "loss": 0.0936, "step": 26623 }, { "epoch": 2.68, "learning_rate": 1.4261529570703874e-05, "loss": 0.9641, "step": 26624 }, { "epoch": 2.68, "learning_rate": 1.4259492222308001e-05, "loss": 0.1247, "step": 26625 }, { "epoch": 2.68, "learning_rate": 1.4257454974078734e-05, "loss": 0.1562, "step": 26626 }, { "epoch": 2.68, "learning_rate": 1.4255417826029035e-05, "loss": 1.142, "step": 26627 }, { "epoch": 2.68, "learning_rate": 1.4253380778171866e-05, "loss": 0.2684, "step": 26628 }, { "epoch": 2.68, "learning_rate": 1.4251343830520182e-05, "loss": 0.0265, "step": 26629 }, { "epoch": 2.68, "learning_rate": 1.4249306983086973e-05, "loss": 0.4879, "step": 26630 }, { "epoch": 2.68, "learning_rate": 1.4247270235885175e-05, "loss": 0.2872, "step": 26631 }, { "epoch": 2.68, "learning_rate": 1.4245233588927745e-05, "loss": 0.3753, "step": 26632 }, { "epoch": 2.68, "learning_rate": 1.4243197042227665e-05, "loss": 0.1909, "step": 26633 }, { "epoch": 2.68, "learning_rate": 1.4241160595797897e-05, "loss": 0.2697, "step": 26634 }, { "epoch": 2.68, "learning_rate": 1.4239124249651365e-05, "loss": 0.1158, "step": 26635 }, { "epoch": 2.68, "learning_rate": 1.423708800380106e-05, "loss": 0.535, "step": 26636 }, { "epoch": 2.68, "learning_rate": 1.4235051858259935e-05, "loss": 0.4539, "step": 26637 }, { "epoch": 2.68, "learning_rate": 1.4233015813040938e-05, "loss": 0.1579, "step": 26638 }, { "epoch": 2.68, "learning_rate": 1.4230979868157033e-05, "loss": 0.2471, "step": 26639 }, { "epoch": 2.68, "learning_rate": 1.4228944023621173e-05, "loss": 0.1423, "step": 26640 }, { "epoch": 2.68, "learning_rate": 1.4226908279446313e-05, "loss": 0.2881, "step": 26641 }, { "epoch": 2.68, "learning_rate": 1.422487263564541e-05, "loss": 0.4093, "step": 26642 }, { "epoch": 2.68, "learning_rate": 1.4222837092231415e-05, "loss": 0.3559, "step": 26643 }, { "epoch": 2.68, "learning_rate": 1.4220801649217282e-05, "loss": 0.1613, "step": 26644 }, { "epoch": 2.68, "learning_rate": 1.4218766306615966e-05, "loss": 0.3613, "step": 26645 }, { "epoch": 2.68, "learning_rate": 1.4216731064440415e-05, "loss": 0.4208, "step": 26646 }, { "epoch": 2.68, "learning_rate": 1.4214695922703575e-05, "loss": 0.5693, "step": 26647 }, { "epoch": 2.68, "learning_rate": 1.4212660881418429e-05, "loss": 0.7772, "step": 26648 }, { "epoch": 2.68, "learning_rate": 1.4210625940597874e-05, "loss": 0.3354, "step": 26649 }, { "epoch": 2.68, "learning_rate": 1.4208591100254908e-05, "loss": 0.5113, "step": 26650 }, { "epoch": 2.68, "learning_rate": 1.4206556360402456e-05, "loss": 0.5705, "step": 26651 }, { "epoch": 2.68, "learning_rate": 1.4204521721053474e-05, "loss": 0.3152, "step": 26652 }, { "epoch": 2.68, "learning_rate": 1.4202487182220903e-05, "loss": 0.0234, "step": 26653 }, { "epoch": 2.68, "learning_rate": 1.4200452743917698e-05, "loss": 0.1461, "step": 26654 }, { "epoch": 2.68, "learning_rate": 1.41984184061568e-05, "loss": 0.4, "step": 26655 }, { "epoch": 2.68, "learning_rate": 1.4196384168951156e-05, "loss": 0.1707, "step": 26656 }, { "epoch": 2.68, "learning_rate": 1.419435003231371e-05, "loss": 0.5199, "step": 26657 }, { "epoch": 2.68, "learning_rate": 1.419231599625741e-05, "loss": 0.1998, "step": 26658 }, { "epoch": 2.68, "learning_rate": 1.4190282060795196e-05, "loss": 0.6223, "step": 26659 }, { "epoch": 2.68, "learning_rate": 1.4188248225940013e-05, "loss": 0.1578, "step": 26660 }, { "epoch": 2.68, "learning_rate": 1.418621449170479e-05, "loss": 0.0975, "step": 26661 }, { "epoch": 2.68, "learning_rate": 1.4184180858102508e-05, "loss": 0.354, "step": 26662 }, { "epoch": 2.68, "learning_rate": 1.418214732514607e-05, "loss": 0.1485, "step": 26663 }, { "epoch": 2.68, "learning_rate": 1.4180113892848416e-05, "loss": 0.1735, "step": 26664 }, { "epoch": 2.68, "learning_rate": 1.417808056122251e-05, "loss": 0.5555, "step": 26665 }, { "epoch": 2.68, "learning_rate": 1.4176047330281292e-05, "loss": 0.3971, "step": 26666 }, { "epoch": 2.68, "learning_rate": 1.4174014200037671e-05, "loss": 0.3782, "step": 26667 }, { "epoch": 2.68, "learning_rate": 1.4171981170504609e-05, "loss": 0.4788, "step": 26668 }, { "epoch": 2.68, "learning_rate": 1.416994824169504e-05, "loss": 0.0931, "step": 26669 }, { "epoch": 2.68, "learning_rate": 1.4167915413621897e-05, "loss": 0.4155, "step": 26670 }, { "epoch": 2.68, "learning_rate": 1.416588268629812e-05, "loss": 0.3045, "step": 26671 }, { "epoch": 2.68, "learning_rate": 1.4163850059736638e-05, "loss": 0.3117, "step": 26672 }, { "epoch": 2.68, "learning_rate": 1.4161817533950393e-05, "loss": 0.2225, "step": 26673 }, { "epoch": 2.68, "learning_rate": 1.4159785108952313e-05, "loss": 0.6385, "step": 26674 }, { "epoch": 2.68, "learning_rate": 1.4157752784755338e-05, "loss": 0.2805, "step": 26675 }, { "epoch": 2.68, "learning_rate": 1.4155720561372396e-05, "loss": 0.2564, "step": 26676 }, { "epoch": 2.68, "learning_rate": 1.4153688438816421e-05, "loss": 0.152, "step": 26677 }, { "epoch": 2.68, "learning_rate": 1.4151656417100333e-05, "loss": 0.4362, "step": 26678 }, { "epoch": 2.68, "learning_rate": 1.4149624496237086e-05, "loss": 0.1317, "step": 26679 }, { "epoch": 2.68, "learning_rate": 1.414759267623961e-05, "loss": 0.0963, "step": 26680 }, { "epoch": 2.68, "learning_rate": 1.4145560957120802e-05, "loss": 0.1308, "step": 26681 }, { "epoch": 2.68, "learning_rate": 1.4143529338893625e-05, "loss": 0.2684, "step": 26682 }, { "epoch": 2.68, "learning_rate": 1.4141497821570995e-05, "loss": 0.5644, "step": 26683 }, { "epoch": 2.68, "learning_rate": 1.4139466405165838e-05, "loss": 0.1178, "step": 26684 }, { "epoch": 2.68, "learning_rate": 1.4137435089691084e-05, "loss": 0.3847, "step": 26685 }, { "epoch": 2.68, "learning_rate": 1.4135403875159659e-05, "loss": 0.172, "step": 26686 }, { "epoch": 2.68, "learning_rate": 1.4133372761584485e-05, "loss": 0.3598, "step": 26687 }, { "epoch": 2.68, "learning_rate": 1.4131341748978496e-05, "loss": 0.4881, "step": 26688 }, { "epoch": 2.68, "learning_rate": 1.412931083735461e-05, "loss": 0.3729, "step": 26689 }, { "epoch": 2.68, "learning_rate": 1.4127280026725751e-05, "loss": 0.3354, "step": 26690 }, { "epoch": 2.68, "learning_rate": 1.4125249317104845e-05, "loss": 0.3625, "step": 26691 }, { "epoch": 2.68, "learning_rate": 1.412321870850481e-05, "loss": 0.1716, "step": 26692 }, { "epoch": 2.68, "learning_rate": 1.4121188200938563e-05, "loss": 0.2806, "step": 26693 }, { "epoch": 2.69, "learning_rate": 1.4119157794419055e-05, "loss": 0.1066, "step": 26694 }, { "epoch": 2.69, "learning_rate": 1.4117127488959174e-05, "loss": 0.5069, "step": 26695 }, { "epoch": 2.69, "learning_rate": 1.4115097284571843e-05, "loss": 0.2689, "step": 26696 }, { "epoch": 2.69, "learning_rate": 1.411306718127e-05, "loss": 0.3809, "step": 26697 }, { "epoch": 2.69, "learning_rate": 1.4111037179066564e-05, "loss": 0.2213, "step": 26698 }, { "epoch": 2.69, "learning_rate": 1.4109007277974418e-05, "loss": 0.3056, "step": 26699 }, { "epoch": 2.69, "learning_rate": 1.410697747800652e-05, "loss": 0.0205, "step": 26700 }, { "epoch": 2.69, "learning_rate": 1.410494777917577e-05, "loss": 0.7234, "step": 26701 }, { "epoch": 2.69, "learning_rate": 1.4102918181495086e-05, "loss": 0.1236, "step": 26702 }, { "epoch": 2.69, "learning_rate": 1.410088868497738e-05, "loss": 0.3704, "step": 26703 }, { "epoch": 2.69, "learning_rate": 1.4098859289635574e-05, "loss": 0.7219, "step": 26704 }, { "epoch": 2.69, "learning_rate": 1.4096829995482576e-05, "loss": 0.2723, "step": 26705 }, { "epoch": 2.69, "learning_rate": 1.4094800802531304e-05, "loss": 0.0588, "step": 26706 }, { "epoch": 2.69, "learning_rate": 1.4092771710794658e-05, "loss": 0.3257, "step": 26707 }, { "epoch": 2.69, "learning_rate": 1.4090742720285582e-05, "loss": 0.3122, "step": 26708 }, { "epoch": 2.69, "learning_rate": 1.4088713831016954e-05, "loss": 0.1733, "step": 26709 }, { "epoch": 2.69, "learning_rate": 1.4086685043001688e-05, "loss": 0.4709, "step": 26710 }, { "epoch": 2.69, "learning_rate": 1.408465635625272e-05, "loss": 0.5007, "step": 26711 }, { "epoch": 2.69, "learning_rate": 1.4082627770782952e-05, "loss": 0.1054, "step": 26712 }, { "epoch": 2.69, "learning_rate": 1.4080599286605266e-05, "loss": 0.2953, "step": 26713 }, { "epoch": 2.69, "learning_rate": 1.4078570903732596e-05, "loss": 0.4053, "step": 26714 }, { "epoch": 2.69, "learning_rate": 1.407654262217785e-05, "loss": 0.2457, "step": 26715 }, { "epoch": 2.69, "learning_rate": 1.4074514441953925e-05, "loss": 0.1686, "step": 26716 }, { "epoch": 2.69, "learning_rate": 1.4072486363073735e-05, "loss": 0.393, "step": 26717 }, { "epoch": 2.69, "learning_rate": 1.407045838555018e-05, "loss": 0.1771, "step": 26718 }, { "epoch": 2.69, "learning_rate": 1.4068430509396168e-05, "loss": 0.1312, "step": 26719 }, { "epoch": 2.69, "learning_rate": 1.4066402734624606e-05, "loss": 0.2897, "step": 26720 }, { "epoch": 2.69, "learning_rate": 1.4064375061248392e-05, "loss": 0.4041, "step": 26721 }, { "epoch": 2.69, "learning_rate": 1.4062347489280436e-05, "loss": 0.4339, "step": 26722 }, { "epoch": 2.69, "learning_rate": 1.4060320018733638e-05, "loss": 0.1349, "step": 26723 }, { "epoch": 2.69, "learning_rate": 1.4058292649620898e-05, "loss": 0.8937, "step": 26724 }, { "epoch": 2.69, "learning_rate": 1.4056265381955105e-05, "loss": 0.2842, "step": 26725 }, { "epoch": 2.69, "learning_rate": 1.4054238215749201e-05, "loss": 0.317, "step": 26726 }, { "epoch": 2.69, "learning_rate": 1.4052211151016046e-05, "loss": 0.3036, "step": 26727 }, { "epoch": 2.69, "learning_rate": 1.4050184187768538e-05, "loss": 0.1544, "step": 26728 }, { "epoch": 2.69, "learning_rate": 1.4048157326019608e-05, "loss": 0.0575, "step": 26729 }, { "epoch": 2.69, "learning_rate": 1.4046130565782131e-05, "loss": 0.1285, "step": 26730 }, { "epoch": 2.69, "learning_rate": 1.4044103907069011e-05, "loss": 0.5511, "step": 26731 }, { "epoch": 2.69, "learning_rate": 1.4042077349893147e-05, "loss": 0.6999, "step": 26732 }, { "epoch": 2.69, "learning_rate": 1.404005089426743e-05, "loss": 0.5826, "step": 26733 }, { "epoch": 2.69, "learning_rate": 1.4038024540204755e-05, "loss": 0.4844, "step": 26734 }, { "epoch": 2.69, "learning_rate": 1.4035998287718024e-05, "loss": 0.4116, "step": 26735 }, { "epoch": 2.69, "learning_rate": 1.4033972136820124e-05, "loss": 0.6659, "step": 26736 }, { "epoch": 2.69, "learning_rate": 1.403194608752395e-05, "loss": 0.7018, "step": 26737 }, { "epoch": 2.69, "learning_rate": 1.4029920139842401e-05, "loss": 0.4617, "step": 26738 }, { "epoch": 2.69, "learning_rate": 1.402789429378835e-05, "loss": 0.3628, "step": 26739 }, { "epoch": 2.69, "learning_rate": 1.4025868549374726e-05, "loss": 0.7753, "step": 26740 }, { "epoch": 2.69, "learning_rate": 1.4023842906614391e-05, "loss": 0.6091, "step": 26741 }, { "epoch": 2.69, "learning_rate": 1.4021817365520226e-05, "loss": 0.1659, "step": 26742 }, { "epoch": 2.69, "learning_rate": 1.4019791926105151e-05, "loss": 0.1085, "step": 26743 }, { "epoch": 2.69, "learning_rate": 1.4017766588382052e-05, "loss": 0.5784, "step": 26744 }, { "epoch": 2.69, "learning_rate": 1.4015741352363784e-05, "loss": 0.2604, "step": 26745 }, { "epoch": 2.69, "learning_rate": 1.4013716218063265e-05, "loss": 0.2686, "step": 26746 }, { "epoch": 2.69, "learning_rate": 1.4011691185493375e-05, "loss": 0.6266, "step": 26747 }, { "epoch": 2.69, "learning_rate": 1.4009666254667006e-05, "loss": 0.201, "step": 26748 }, { "epoch": 2.69, "learning_rate": 1.4007641425597034e-05, "loss": 0.5042, "step": 26749 }, { "epoch": 2.69, "learning_rate": 1.4005616698296346e-05, "loss": 0.1043, "step": 26750 }, { "epoch": 2.69, "learning_rate": 1.4003592072777832e-05, "loss": 0.4973, "step": 26751 }, { "epoch": 2.69, "learning_rate": 1.4001567549054374e-05, "loss": 0.5267, "step": 26752 }, { "epoch": 2.69, "learning_rate": 1.399954312713885e-05, "loss": 0.1421, "step": 26753 }, { "epoch": 2.69, "learning_rate": 1.399751880704415e-05, "loss": 0.0671, "step": 26754 }, { "epoch": 2.69, "learning_rate": 1.399549458878315e-05, "loss": 0.5108, "step": 26755 }, { "epoch": 2.69, "learning_rate": 1.3993470472368735e-05, "loss": 0.3051, "step": 26756 }, { "epoch": 2.69, "learning_rate": 1.3991446457813773e-05, "loss": 0.1322, "step": 26757 }, { "epoch": 2.69, "learning_rate": 1.398942254513118e-05, "loss": 0.2447, "step": 26758 }, { "epoch": 2.69, "learning_rate": 1.3987398734333786e-05, "loss": 0.6582, "step": 26759 }, { "epoch": 2.69, "learning_rate": 1.3985375025434509e-05, "loss": 0.4404, "step": 26760 }, { "epoch": 2.69, "learning_rate": 1.3983351418446212e-05, "loss": 0.137, "step": 26761 }, { "epoch": 2.69, "learning_rate": 1.398132791338177e-05, "loss": 0.5437, "step": 26762 }, { "epoch": 2.69, "learning_rate": 1.3979304510254068e-05, "loss": 0.2464, "step": 26763 }, { "epoch": 2.69, "learning_rate": 1.3977281209075979e-05, "loss": 0.3349, "step": 26764 }, { "epoch": 2.69, "learning_rate": 1.3975258009860373e-05, "loss": 0.8781, "step": 26765 }, { "epoch": 2.69, "learning_rate": 1.397323491262013e-05, "loss": 0.3246, "step": 26766 }, { "epoch": 2.69, "learning_rate": 1.3971211917368125e-05, "loss": 0.0681, "step": 26767 }, { "epoch": 2.69, "learning_rate": 1.3969189024117225e-05, "loss": 0.5135, "step": 26768 }, { "epoch": 2.69, "learning_rate": 1.3967166232880312e-05, "loss": 0.0592, "step": 26769 }, { "epoch": 2.69, "learning_rate": 1.396514354367025e-05, "loss": 0.3392, "step": 26770 }, { "epoch": 2.69, "learning_rate": 1.3963120956499907e-05, "loss": 0.3648, "step": 26771 }, { "epoch": 2.69, "learning_rate": 1.3961098471382184e-05, "loss": 0.2162, "step": 26772 }, { "epoch": 2.69, "learning_rate": 1.3959076088329915e-05, "loss": 0.157, "step": 26773 }, { "epoch": 2.69, "learning_rate": 1.3957053807355973e-05, "loss": 0.2689, "step": 26774 }, { "epoch": 2.69, "learning_rate": 1.3955031628473248e-05, "loss": 0.2486, "step": 26775 }, { "epoch": 2.69, "learning_rate": 1.3953009551694612e-05, "loss": 0.5194, "step": 26776 }, { "epoch": 2.69, "learning_rate": 1.3950987577032895e-05, "loss": 0.2211, "step": 26777 }, { "epoch": 2.69, "learning_rate": 1.3948965704501e-05, "loss": 0.226, "step": 26778 }, { "epoch": 2.69, "learning_rate": 1.3946943934111778e-05, "loss": 0.7543, "step": 26779 }, { "epoch": 2.69, "learning_rate": 1.39449222658781e-05, "loss": 0.3425, "step": 26780 }, { "epoch": 2.69, "learning_rate": 1.3942900699812827e-05, "loss": 0.3449, "step": 26781 }, { "epoch": 2.69, "learning_rate": 1.3940879235928827e-05, "loss": 0.2166, "step": 26782 }, { "epoch": 2.69, "learning_rate": 1.3938857874238963e-05, "loss": 0.3251, "step": 26783 }, { "epoch": 2.69, "learning_rate": 1.3936836614756097e-05, "loss": 0.1478, "step": 26784 }, { "epoch": 2.69, "learning_rate": 1.3934815457493081e-05, "loss": 0.6144, "step": 26785 }, { "epoch": 2.69, "learning_rate": 1.3932794402462809e-05, "loss": 0.6702, "step": 26786 }, { "epoch": 2.69, "learning_rate": 1.3930773449678113e-05, "loss": 0.2744, "step": 26787 }, { "epoch": 2.69, "learning_rate": 1.3928752599151847e-05, "loss": 0.543, "step": 26788 }, { "epoch": 2.69, "learning_rate": 1.39267318508969e-05, "loss": 0.1134, "step": 26789 }, { "epoch": 2.69, "learning_rate": 1.3924711204926127e-05, "loss": 0.425, "step": 26790 }, { "epoch": 2.69, "learning_rate": 1.3922690661252353e-05, "loss": 0.7023, "step": 26791 }, { "epoch": 2.69, "learning_rate": 1.392067021988847e-05, "loss": 1.0567, "step": 26792 }, { "epoch": 2.7, "learning_rate": 1.3918649880847326e-05, "loss": 0.2904, "step": 26793 }, { "epoch": 2.7, "learning_rate": 1.391662964414178e-05, "loss": 0.5081, "step": 26794 }, { "epoch": 2.7, "learning_rate": 1.391460950978468e-05, "loss": 0.1889, "step": 26795 }, { "epoch": 2.7, "learning_rate": 1.3912589477788884e-05, "loss": 0.1447, "step": 26796 }, { "epoch": 2.7, "learning_rate": 1.3910569548167255e-05, "loss": 0.5121, "step": 26797 }, { "epoch": 2.7, "learning_rate": 1.3908549720932635e-05, "loss": 0.2205, "step": 26798 }, { "epoch": 2.7, "learning_rate": 1.3906529996097887e-05, "loss": 0.5827, "step": 26799 }, { "epoch": 2.7, "learning_rate": 1.390451037367586e-05, "loss": 0.4313, "step": 26800 }, { "epoch": 2.7, "learning_rate": 1.3902490853679403e-05, "loss": 0.1902, "step": 26801 }, { "epoch": 2.7, "learning_rate": 1.3900471436121373e-05, "loss": 0.5242, "step": 26802 }, { "epoch": 2.7, "learning_rate": 1.3898452121014606e-05, "loss": 0.3094, "step": 26803 }, { "epoch": 2.7, "learning_rate": 1.3896432908371986e-05, "loss": 0.1798, "step": 26804 }, { "epoch": 2.7, "learning_rate": 1.3894413798206329e-05, "loss": 0.6239, "step": 26805 }, { "epoch": 2.7, "learning_rate": 1.3892394790530486e-05, "loss": 0.4329, "step": 26806 }, { "epoch": 2.7, "learning_rate": 1.3890375885357326e-05, "loss": 0.25, "step": 26807 }, { "epoch": 2.7, "learning_rate": 1.3888357082699696e-05, "loss": 0.4207, "step": 26808 }, { "epoch": 2.7, "learning_rate": 1.388633838257041e-05, "loss": 0.4085, "step": 26809 }, { "epoch": 2.7, "learning_rate": 1.3884319784982348e-05, "loss": 0.0603, "step": 26810 }, { "epoch": 2.7, "learning_rate": 1.3882301289948343e-05, "loss": 0.12, "step": 26811 }, { "epoch": 2.7, "learning_rate": 1.3880282897481241e-05, "loss": 0.359, "step": 26812 }, { "epoch": 2.7, "learning_rate": 1.387826460759389e-05, "loss": 0.3265, "step": 26813 }, { "epoch": 2.7, "learning_rate": 1.3876246420299117e-05, "loss": 0.5043, "step": 26814 }, { "epoch": 2.7, "learning_rate": 1.3874228335609797e-05, "loss": 0.2316, "step": 26815 }, { "epoch": 2.7, "learning_rate": 1.3872210353538745e-05, "loss": 0.1453, "step": 26816 }, { "epoch": 2.7, "learning_rate": 1.38701924740988e-05, "loss": 0.5212, "step": 26817 }, { "epoch": 2.7, "learning_rate": 1.386817469730284e-05, "loss": 0.1551, "step": 26818 }, { "epoch": 2.7, "learning_rate": 1.3866157023163663e-05, "loss": 0.0627, "step": 26819 }, { "epoch": 2.7, "learning_rate": 1.3864139451694114e-05, "loss": 0.7316, "step": 26820 }, { "epoch": 2.7, "learning_rate": 1.3862121982907057e-05, "loss": 0.3027, "step": 26821 }, { "epoch": 2.7, "learning_rate": 1.3860104616815327e-05, "loss": 0.5161, "step": 26822 }, { "epoch": 2.7, "learning_rate": 1.3858087353431727e-05, "loss": 0.2302, "step": 26823 }, { "epoch": 2.7, "learning_rate": 1.3856070192769131e-05, "loss": 0.1996, "step": 26824 }, { "epoch": 2.7, "learning_rate": 1.385405313484036e-05, "loss": 0.5376, "step": 26825 }, { "epoch": 2.7, "learning_rate": 1.3852036179658257e-05, "loss": 0.787, "step": 26826 }, { "epoch": 2.7, "learning_rate": 1.3850019327235653e-05, "loss": 0.1146, "step": 26827 }, { "epoch": 2.7, "learning_rate": 1.3848002577585379e-05, "loss": 0.872, "step": 26828 }, { "epoch": 2.7, "learning_rate": 1.3845985930720273e-05, "loss": 0.3944, "step": 26829 }, { "epoch": 2.7, "learning_rate": 1.3843969386653166e-05, "loss": 0.3582, "step": 26830 }, { "epoch": 2.7, "learning_rate": 1.3841952945396894e-05, "loss": 0.3372, "step": 26831 }, { "epoch": 2.7, "learning_rate": 1.3839936606964287e-05, "loss": 0.5234, "step": 26832 }, { "epoch": 2.7, "learning_rate": 1.3837920371368173e-05, "loss": 0.0665, "step": 26833 }, { "epoch": 2.7, "learning_rate": 1.3835904238621389e-05, "loss": 0.14, "step": 26834 }, { "epoch": 2.7, "learning_rate": 1.3833888208736749e-05, "loss": 0.411, "step": 26835 }, { "epoch": 2.7, "learning_rate": 1.3831872281727114e-05, "loss": 0.5402, "step": 26836 }, { "epoch": 2.7, "learning_rate": 1.3829856457605285e-05, "loss": 0.549, "step": 26837 }, { "epoch": 2.7, "learning_rate": 1.3827840736384087e-05, "loss": 0.4788, "step": 26838 }, { "epoch": 2.7, "learning_rate": 1.3825825118076366e-05, "loss": 0.4539, "step": 26839 }, { "epoch": 2.7, "learning_rate": 1.3823809602694944e-05, "loss": 1.2558, "step": 26840 }, { "epoch": 2.7, "learning_rate": 1.3821794190252645e-05, "loss": 0.3386, "step": 26841 }, { "epoch": 2.7, "learning_rate": 1.381977888076229e-05, "loss": 0.2593, "step": 26842 }, { "epoch": 2.7, "learning_rate": 1.3817763674236709e-05, "loss": 0.4469, "step": 26843 }, { "epoch": 2.7, "learning_rate": 1.3815748570688723e-05, "loss": 0.0844, "step": 26844 }, { "epoch": 2.7, "learning_rate": 1.3813733570131158e-05, "loss": 0.3461, "step": 26845 }, { "epoch": 2.7, "learning_rate": 1.3811718672576825e-05, "loss": 0.5282, "step": 26846 }, { "epoch": 2.7, "learning_rate": 1.3809703878038574e-05, "loss": 0.1714, "step": 26847 }, { "epoch": 2.7, "learning_rate": 1.3807689186529201e-05, "loss": 0.4158, "step": 26848 }, { "epoch": 2.7, "learning_rate": 1.380567459806152e-05, "loss": 0.1708, "step": 26849 }, { "epoch": 2.7, "learning_rate": 1.3803660112648389e-05, "loss": 0.1795, "step": 26850 }, { "epoch": 2.7, "learning_rate": 1.3801645730302591e-05, "loss": 0.4439, "step": 26851 }, { "epoch": 2.7, "learning_rate": 1.3799631451036947e-05, "loss": 0.2732, "step": 26852 }, { "epoch": 2.7, "learning_rate": 1.3797617274864298e-05, "loss": 0.4837, "step": 26853 }, { "epoch": 2.7, "learning_rate": 1.379560320179746e-05, "loss": 0.3939, "step": 26854 }, { "epoch": 2.7, "learning_rate": 1.3793589231849216e-05, "loss": 0.1143, "step": 26855 }, { "epoch": 2.7, "learning_rate": 1.3791575365032415e-05, "loss": 0.2703, "step": 26856 }, { "epoch": 2.7, "learning_rate": 1.3789561601359863e-05, "loss": 0.7078, "step": 26857 }, { "epoch": 2.7, "learning_rate": 1.3787547940844375e-05, "loss": 0.0368, "step": 26858 }, { "epoch": 2.7, "learning_rate": 1.3785534383498766e-05, "loss": 0.082, "step": 26859 }, { "epoch": 2.7, "learning_rate": 1.3783520929335846e-05, "loss": 0.4262, "step": 26860 }, { "epoch": 2.7, "learning_rate": 1.3781507578368427e-05, "loss": 0.3343, "step": 26861 }, { "epoch": 2.7, "learning_rate": 1.3779494330609329e-05, "loss": 0.2711, "step": 26862 }, { "epoch": 2.7, "learning_rate": 1.3777481186071354e-05, "loss": 0.938, "step": 26863 }, { "epoch": 2.7, "learning_rate": 1.377546814476732e-05, "loss": 0.777, "step": 26864 }, { "epoch": 2.7, "learning_rate": 1.3773455206710032e-05, "loss": 0.2602, "step": 26865 }, { "epoch": 2.7, "learning_rate": 1.3771442371912294e-05, "loss": 0.0225, "step": 26866 }, { "epoch": 2.7, "learning_rate": 1.3769429640386934e-05, "loss": 0.2131, "step": 26867 }, { "epoch": 2.7, "learning_rate": 1.3767417012146757e-05, "loss": 0.3113, "step": 26868 }, { "epoch": 2.7, "learning_rate": 1.3765404487204543e-05, "loss": 0.3201, "step": 26869 }, { "epoch": 2.7, "learning_rate": 1.3763392065573131e-05, "loss": 0.6685, "step": 26870 }, { "epoch": 2.7, "learning_rate": 1.3761379747265313e-05, "loss": 0.3834, "step": 26871 }, { "epoch": 2.7, "learning_rate": 1.3759367532293899e-05, "loss": 0.2764, "step": 26872 }, { "epoch": 2.7, "learning_rate": 1.3757355420671691e-05, "loss": 0.7281, "step": 26873 }, { "epoch": 2.7, "learning_rate": 1.3755343412411495e-05, "loss": 0.272, "step": 26874 }, { "epoch": 2.7, "learning_rate": 1.3753331507526115e-05, "loss": 0.8068, "step": 26875 }, { "epoch": 2.7, "learning_rate": 1.3751319706028349e-05, "loss": 0.1097, "step": 26876 }, { "epoch": 2.7, "learning_rate": 1.3749308007931006e-05, "loss": 0.4975, "step": 26877 }, { "epoch": 2.7, "learning_rate": 1.3747296413246869e-05, "loss": 0.0778, "step": 26878 }, { "epoch": 2.7, "learning_rate": 1.3745284921988782e-05, "loss": 0.2238, "step": 26879 }, { "epoch": 2.7, "learning_rate": 1.3743273534169504e-05, "loss": 0.5705, "step": 26880 }, { "epoch": 2.7, "learning_rate": 1.374126224980184e-05, "loss": 0.4308, "step": 26881 }, { "epoch": 2.7, "learning_rate": 1.373925106889862e-05, "loss": 0.3412, "step": 26882 }, { "epoch": 2.7, "learning_rate": 1.3737239991472608e-05, "loss": 0.2148, "step": 26883 }, { "epoch": 2.7, "learning_rate": 1.3735229017536602e-05, "loss": 0.1601, "step": 26884 }, { "epoch": 2.7, "learning_rate": 1.3733218147103424e-05, "loss": 0.6184, "step": 26885 }, { "epoch": 2.7, "learning_rate": 1.3731207380185867e-05, "loss": 0.851, "step": 26886 }, { "epoch": 2.7, "learning_rate": 1.37291967167967e-05, "loss": 0.281, "step": 26887 }, { "epoch": 2.7, "learning_rate": 1.372718615694874e-05, "loss": 0.3748, "step": 26888 }, { "epoch": 2.7, "learning_rate": 1.372517570065478e-05, "loss": 0.4704, "step": 26889 }, { "epoch": 2.7, "learning_rate": 1.3723165347927611e-05, "loss": 0.3702, "step": 26890 }, { "epoch": 2.71, "learning_rate": 1.3721155098780024e-05, "loss": 0.2513, "step": 26891 }, { "epoch": 2.71, "learning_rate": 1.3719144953224815e-05, "loss": 0.3044, "step": 26892 }, { "epoch": 2.71, "learning_rate": 1.371713491127477e-05, "loss": 0.2455, "step": 26893 }, { "epoch": 2.71, "learning_rate": 1.371512497294269e-05, "loss": 0.0788, "step": 26894 }, { "epoch": 2.71, "learning_rate": 1.3713115138241344e-05, "loss": 0.5495, "step": 26895 }, { "epoch": 2.71, "learning_rate": 1.3711105407183561e-05, "loss": 0.4079, "step": 26896 }, { "epoch": 2.71, "learning_rate": 1.3709095779782094e-05, "loss": 0.3955, "step": 26897 }, { "epoch": 2.71, "learning_rate": 1.3707086256049735e-05, "loss": 0.1432, "step": 26898 }, { "epoch": 2.71, "learning_rate": 1.3705076835999292e-05, "loss": 0.7032, "step": 26899 }, { "epoch": 2.71, "learning_rate": 1.3703067519643553e-05, "loss": 0.7471, "step": 26900 }, { "epoch": 2.71, "learning_rate": 1.3701058306995269e-05, "loss": 0.1381, "step": 26901 }, { "epoch": 2.71, "learning_rate": 1.369904919806726e-05, "loss": 0.4217, "step": 26902 }, { "epoch": 2.71, "learning_rate": 1.3697040192872301e-05, "loss": 0.1587, "step": 26903 }, { "epoch": 2.71, "learning_rate": 1.3695031291423176e-05, "loss": 0.5591, "step": 26904 }, { "epoch": 2.71, "learning_rate": 1.369302249373267e-05, "loss": 0.2411, "step": 26905 }, { "epoch": 2.71, "learning_rate": 1.3691013799813566e-05, "loss": 0.4461, "step": 26906 }, { "epoch": 2.71, "learning_rate": 1.3689005209678646e-05, "loss": 0.1951, "step": 26907 }, { "epoch": 2.71, "learning_rate": 1.368699672334069e-05, "loss": 0.6413, "step": 26908 }, { "epoch": 2.71, "learning_rate": 1.3684988340812482e-05, "loss": 0.2011, "step": 26909 }, { "epoch": 2.71, "learning_rate": 1.368298006210679e-05, "loss": 0.2598, "step": 26910 }, { "epoch": 2.71, "learning_rate": 1.3680971887236428e-05, "loss": 0.6688, "step": 26911 }, { "epoch": 2.71, "learning_rate": 1.3678963816214142e-05, "loss": 0.4549, "step": 26912 }, { "epoch": 2.71, "learning_rate": 1.367695584905271e-05, "loss": 0.1942, "step": 26913 }, { "epoch": 2.71, "learning_rate": 1.3674947985764947e-05, "loss": 0.116, "step": 26914 }, { "epoch": 2.71, "learning_rate": 1.3672940226363589e-05, "loss": 0.1481, "step": 26915 }, { "epoch": 2.71, "learning_rate": 1.3670932570861418e-05, "loss": 0.1844, "step": 26916 }, { "epoch": 2.71, "learning_rate": 1.3668925019271235e-05, "loss": 0.6522, "step": 26917 }, { "epoch": 2.71, "learning_rate": 1.3666917571605806e-05, "loss": 0.5811, "step": 26918 }, { "epoch": 2.71, "learning_rate": 1.3664910227877881e-05, "loss": 0.7579, "step": 26919 }, { "epoch": 2.71, "learning_rate": 1.3662902988100264e-05, "loss": 0.3101, "step": 26920 }, { "epoch": 2.71, "learning_rate": 1.3660895852285718e-05, "loss": 0.72, "step": 26921 }, { "epoch": 2.71, "learning_rate": 1.3658888820447013e-05, "loss": 0.3579, "step": 26922 }, { "epoch": 2.71, "learning_rate": 1.3656881892596924e-05, "loss": 0.4737, "step": 26923 }, { "epoch": 2.71, "learning_rate": 1.3654875068748211e-05, "loss": 0.0437, "step": 26924 }, { "epoch": 2.71, "learning_rate": 1.3652868348913676e-05, "loss": 0.2352, "step": 26925 }, { "epoch": 2.71, "learning_rate": 1.3650861733106062e-05, "loss": 0.3867, "step": 26926 }, { "epoch": 2.71, "learning_rate": 1.3648855221338126e-05, "loss": 0.2224, "step": 26927 }, { "epoch": 2.71, "learning_rate": 1.3646848813622682e-05, "loss": 0.0808, "step": 26928 }, { "epoch": 2.71, "learning_rate": 1.3644842509972457e-05, "loss": 0.6481, "step": 26929 }, { "epoch": 2.71, "learning_rate": 1.3642836310400225e-05, "loss": 0.2015, "step": 26930 }, { "epoch": 2.71, "learning_rate": 1.364083021491877e-05, "loss": 0.5283, "step": 26931 }, { "epoch": 2.71, "learning_rate": 1.3638824223540856e-05, "loss": 0.6558, "step": 26932 }, { "epoch": 2.71, "learning_rate": 1.3636818336279221e-05, "loss": 0.4148, "step": 26933 }, { "epoch": 2.71, "learning_rate": 1.3634812553146662e-05, "loss": 0.2223, "step": 26934 }, { "epoch": 2.71, "learning_rate": 1.3632806874155927e-05, "loss": 0.4513, "step": 26935 }, { "epoch": 2.71, "learning_rate": 1.3630801299319781e-05, "loss": 0.3021, "step": 26936 }, { "epoch": 2.71, "learning_rate": 1.3628795828650994e-05, "loss": 0.0592, "step": 26937 }, { "epoch": 2.71, "learning_rate": 1.362679046216232e-05, "loss": 0.3988, "step": 26938 }, { "epoch": 2.71, "learning_rate": 1.3624785199866522e-05, "loss": 0.4388, "step": 26939 }, { "epoch": 2.71, "learning_rate": 1.3622780041776364e-05, "loss": 0.5273, "step": 26940 }, { "epoch": 2.71, "learning_rate": 1.36207749879046e-05, "loss": 0.2182, "step": 26941 }, { "epoch": 2.71, "learning_rate": 1.3618770038263982e-05, "loss": 0.5032, "step": 26942 }, { "epoch": 2.71, "learning_rate": 1.3616765192867306e-05, "loss": 0.1683, "step": 26943 }, { "epoch": 2.71, "learning_rate": 1.3614760451727288e-05, "loss": 0.4693, "step": 26944 }, { "epoch": 2.71, "learning_rate": 1.3612755814856695e-05, "loss": 0.2108, "step": 26945 }, { "epoch": 2.71, "learning_rate": 1.3610751282268306e-05, "loss": 0.2497, "step": 26946 }, { "epoch": 2.71, "learning_rate": 1.3608746853974845e-05, "loss": 0.2378, "step": 26947 }, { "epoch": 2.71, "learning_rate": 1.3606742529989092e-05, "loss": 0.029, "step": 26948 }, { "epoch": 2.71, "learning_rate": 1.3604738310323793e-05, "loss": 0.5068, "step": 26949 }, { "epoch": 2.71, "learning_rate": 1.36027341949917e-05, "loss": 0.5005, "step": 26950 }, { "epoch": 2.71, "learning_rate": 1.3600730184005574e-05, "loss": 0.6079, "step": 26951 }, { "epoch": 2.71, "learning_rate": 1.3598726277378159e-05, "loss": 0.1841, "step": 26952 }, { "epoch": 2.71, "learning_rate": 1.3596722475122213e-05, "loss": 0.317, "step": 26953 }, { "epoch": 2.71, "learning_rate": 1.3594718777250481e-05, "loss": 0.4563, "step": 26954 }, { "epoch": 2.71, "learning_rate": 1.359271518377572e-05, "loss": 0.6064, "step": 26955 }, { "epoch": 2.71, "learning_rate": 1.3590711694710667e-05, "loss": 0.2526, "step": 26956 }, { "epoch": 2.71, "learning_rate": 1.3588708310068104e-05, "loss": 0.4066, "step": 26957 }, { "epoch": 2.71, "learning_rate": 1.358670502986075e-05, "loss": 0.4136, "step": 26958 }, { "epoch": 2.71, "learning_rate": 1.3584701854101347e-05, "loss": 0.6264, "step": 26959 }, { "epoch": 2.71, "learning_rate": 1.3582698782802679e-05, "loss": 0.0534, "step": 26960 }, { "epoch": 2.71, "learning_rate": 1.358069581597746e-05, "loss": 0.1547, "step": 26961 }, { "epoch": 2.71, "learning_rate": 1.3578692953638438e-05, "loss": 0.9097, "step": 26962 }, { "epoch": 2.71, "learning_rate": 1.3576690195798375e-05, "loss": 0.1279, "step": 26963 }, { "epoch": 2.71, "learning_rate": 1.3574687542470018e-05, "loss": 0.1189, "step": 26964 }, { "epoch": 2.71, "learning_rate": 1.357268499366608e-05, "loss": 0.6277, "step": 26965 }, { "epoch": 2.71, "learning_rate": 1.3570682549399338e-05, "loss": 0.7502, "step": 26966 }, { "epoch": 2.71, "learning_rate": 1.3568680209682521e-05, "loss": 0.7801, "step": 26967 }, { "epoch": 2.71, "learning_rate": 1.3566677974528372e-05, "loss": 0.3256, "step": 26968 }, { "epoch": 2.71, "learning_rate": 1.3564675843949633e-05, "loss": 0.3052, "step": 26969 }, { "epoch": 2.71, "learning_rate": 1.3562673817959049e-05, "loss": 0.3565, "step": 26970 }, { "epoch": 2.71, "learning_rate": 1.3560671896569348e-05, "loss": 0.2647, "step": 26971 }, { "epoch": 2.71, "learning_rate": 1.3558670079793281e-05, "loss": 0.1457, "step": 26972 }, { "epoch": 2.71, "learning_rate": 1.3556668367643583e-05, "loss": 0.3084, "step": 26973 }, { "epoch": 2.71, "learning_rate": 1.3554666760132982e-05, "loss": 0.2947, "step": 26974 }, { "epoch": 2.71, "learning_rate": 1.3552665257274248e-05, "loss": 0.3838, "step": 26975 }, { "epoch": 2.71, "learning_rate": 1.3550663859080071e-05, "loss": 0.8555, "step": 26976 }, { "epoch": 2.71, "learning_rate": 1.3548662565563225e-05, "loss": 0.6722, "step": 26977 }, { "epoch": 2.71, "learning_rate": 1.3546661376736442e-05, "loss": 0.3699, "step": 26978 }, { "epoch": 2.71, "learning_rate": 1.3544660292612424e-05, "loss": 0.7943, "step": 26979 }, { "epoch": 2.71, "learning_rate": 1.354265931320394e-05, "loss": 0.6806, "step": 26980 }, { "epoch": 2.71, "learning_rate": 1.354065843852371e-05, "loss": 0.1301, "step": 26981 }, { "epoch": 2.71, "learning_rate": 1.353865766858447e-05, "loss": 0.0703, "step": 26982 }, { "epoch": 2.71, "learning_rate": 1.3536657003398946e-05, "loss": 0.4943, "step": 26983 }, { "epoch": 2.71, "learning_rate": 1.3534656442979879e-05, "loss": 0.5416, "step": 26984 }, { "epoch": 2.71, "learning_rate": 1.3532655987339993e-05, "loss": 0.3662, "step": 26985 }, { "epoch": 2.71, "learning_rate": 1.353065563649202e-05, "loss": 0.3868, "step": 26986 }, { "epoch": 2.71, "learning_rate": 1.352865539044869e-05, "loss": 0.1903, "step": 26987 }, { "epoch": 2.71, "learning_rate": 1.352665524922272e-05, "loss": 0.2304, "step": 26988 }, { "epoch": 2.71, "learning_rate": 1.352465521282687e-05, "loss": 0.221, "step": 26989 }, { "epoch": 2.72, "learning_rate": 1.3522655281273836e-05, "loss": 0.1286, "step": 26990 }, { "epoch": 2.72, "learning_rate": 1.3520655454576344e-05, "loss": 0.3618, "step": 26991 }, { "epoch": 2.72, "learning_rate": 1.3518655732747156e-05, "loss": 0.6836, "step": 26992 }, { "epoch": 2.72, "learning_rate": 1.3516656115798959e-05, "loss": 0.0406, "step": 26993 }, { "epoch": 2.72, "learning_rate": 1.3514656603744482e-05, "loss": 0.2924, "step": 26994 }, { "epoch": 2.72, "learning_rate": 1.351265719659647e-05, "loss": 0.3411, "step": 26995 }, { "epoch": 2.72, "learning_rate": 1.3510657894367649e-05, "loss": 0.7199, "step": 26996 }, { "epoch": 2.72, "learning_rate": 1.3508658697070702e-05, "loss": 0.1605, "step": 26997 }, { "epoch": 2.72, "learning_rate": 1.350665960471839e-05, "loss": 0.3211, "step": 26998 }, { "epoch": 2.72, "learning_rate": 1.3504660617323423e-05, "loss": 0.0477, "step": 26999 }, { "epoch": 2.72, "learning_rate": 1.350266173489852e-05, "loss": 0.4149, "step": 27000 }, { "epoch": 2.72, "learning_rate": 1.3500662957456401e-05, "loss": 0.5734, "step": 27001 }, { "epoch": 2.72, "learning_rate": 1.3498664285009775e-05, "loss": 0.3124, "step": 27002 }, { "epoch": 2.72, "learning_rate": 1.3496665717571398e-05, "loss": 0.0486, "step": 27003 }, { "epoch": 2.72, "learning_rate": 1.3494667255153947e-05, "loss": 0.5538, "step": 27004 }, { "epoch": 2.72, "learning_rate": 1.3492668897770146e-05, "loss": 0.4541, "step": 27005 }, { "epoch": 2.72, "learning_rate": 1.3490670645432732e-05, "loss": 0.2994, "step": 27006 }, { "epoch": 2.72, "learning_rate": 1.3488672498154417e-05, "loss": 0.1205, "step": 27007 }, { "epoch": 2.72, "learning_rate": 1.3486674455947888e-05, "loss": 0.1548, "step": 27008 }, { "epoch": 2.72, "learning_rate": 1.3484676518825894e-05, "loss": 0.3679, "step": 27009 }, { "epoch": 2.72, "learning_rate": 1.3482678686801146e-05, "loss": 0.3826, "step": 27010 }, { "epoch": 2.72, "learning_rate": 1.3480680959886323e-05, "loss": 0.6756, "step": 27011 }, { "epoch": 2.72, "learning_rate": 1.3478683338094174e-05, "loss": 0.4235, "step": 27012 }, { "epoch": 2.72, "learning_rate": 1.3476685821437402e-05, "loss": 0.4516, "step": 27013 }, { "epoch": 2.72, "learning_rate": 1.3474688409928714e-05, "loss": 0.4672, "step": 27014 }, { "epoch": 2.72, "learning_rate": 1.3472691103580823e-05, "loss": 0.2942, "step": 27015 }, { "epoch": 2.72, "learning_rate": 1.3470693902406437e-05, "loss": 0.2572, "step": 27016 }, { "epoch": 2.72, "learning_rate": 1.3468696806418267e-05, "loss": 0.2052, "step": 27017 }, { "epoch": 2.72, "learning_rate": 1.3466699815629023e-05, "loss": 0.294, "step": 27018 }, { "epoch": 2.72, "learning_rate": 1.3464702930051413e-05, "loss": 0.0407, "step": 27019 }, { "epoch": 2.72, "learning_rate": 1.3462706149698129e-05, "loss": 0.5001, "step": 27020 }, { "epoch": 2.72, "learning_rate": 1.3460709474581914e-05, "loss": 0.1373, "step": 27021 }, { "epoch": 2.72, "learning_rate": 1.3458712904715444e-05, "loss": 0.2328, "step": 27022 }, { "epoch": 2.72, "learning_rate": 1.3456716440111421e-05, "loss": 0.4845, "step": 27023 }, { "epoch": 2.72, "learning_rate": 1.3454720080782582e-05, "loss": 0.3355, "step": 27024 }, { "epoch": 2.72, "learning_rate": 1.3452723826741599e-05, "loss": 0.242, "step": 27025 }, { "epoch": 2.72, "learning_rate": 1.3450727678001174e-05, "loss": 0.2344, "step": 27026 }, { "epoch": 2.72, "learning_rate": 1.3448731634574037e-05, "loss": 0.1443, "step": 27027 }, { "epoch": 2.72, "learning_rate": 1.3446735696472873e-05, "loss": 0.2895, "step": 27028 }, { "epoch": 2.72, "learning_rate": 1.3444739863710386e-05, "loss": 0.3876, "step": 27029 }, { "epoch": 2.72, "learning_rate": 1.3442744136299278e-05, "loss": 0.8621, "step": 27030 }, { "epoch": 2.72, "learning_rate": 1.3440748514252245e-05, "loss": 0.4782, "step": 27031 }, { "epoch": 2.72, "learning_rate": 1.3438752997581993e-05, "loss": 0.6431, "step": 27032 }, { "epoch": 2.72, "learning_rate": 1.3436757586301214e-05, "loss": 0.1233, "step": 27033 }, { "epoch": 2.72, "learning_rate": 1.3434762280422597e-05, "loss": 0.8486, "step": 27034 }, { "epoch": 2.72, "learning_rate": 1.3432767079958876e-05, "loss": 0.8348, "step": 27035 }, { "epoch": 2.72, "learning_rate": 1.343077198492271e-05, "loss": 0.2755, "step": 27036 }, { "epoch": 2.72, "learning_rate": 1.3428776995326795e-05, "loss": 0.4483, "step": 27037 }, { "epoch": 2.72, "learning_rate": 1.3426782111183854e-05, "loss": 0.4243, "step": 27038 }, { "epoch": 2.72, "learning_rate": 1.3424787332506575e-05, "loss": 0.1808, "step": 27039 }, { "epoch": 2.72, "learning_rate": 1.3422792659307625e-05, "loss": 0.79, "step": 27040 }, { "epoch": 2.72, "learning_rate": 1.3420798091599725e-05, "loss": 0.3181, "step": 27041 }, { "epoch": 2.72, "learning_rate": 1.341880362939557e-05, "loss": 0.2472, "step": 27042 }, { "epoch": 2.72, "learning_rate": 1.3416809272707819e-05, "loss": 0.1154, "step": 27043 }, { "epoch": 2.72, "learning_rate": 1.3414815021549194e-05, "loss": 0.0924, "step": 27044 }, { "epoch": 2.72, "learning_rate": 1.341282087593238e-05, "loss": 0.6994, "step": 27045 }, { "epoch": 2.72, "learning_rate": 1.3410826835870064e-05, "loss": 0.2772, "step": 27046 }, { "epoch": 2.72, "learning_rate": 1.3408832901374932e-05, "loss": 0.4385, "step": 27047 }, { "epoch": 2.72, "learning_rate": 1.340683907245968e-05, "loss": 0.0916, "step": 27048 }, { "epoch": 2.72, "learning_rate": 1.340484534913699e-05, "loss": 0.2158, "step": 27049 }, { "epoch": 2.72, "learning_rate": 1.340285173141955e-05, "loss": 0.2547, "step": 27050 }, { "epoch": 2.72, "learning_rate": 1.3400858219320046e-05, "loss": 0.4916, "step": 27051 }, { "epoch": 2.72, "learning_rate": 1.3398864812851158e-05, "loss": 0.3322, "step": 27052 }, { "epoch": 2.72, "learning_rate": 1.33968715120256e-05, "loss": 0.4342, "step": 27053 }, { "epoch": 2.72, "learning_rate": 1.3394878316856026e-05, "loss": 0.1789, "step": 27054 }, { "epoch": 2.72, "learning_rate": 1.3392885227355116e-05, "loss": 0.1186, "step": 27055 }, { "epoch": 2.72, "learning_rate": 1.339089224353559e-05, "loss": 0.1024, "step": 27056 }, { "epoch": 2.72, "learning_rate": 1.3388899365410087e-05, "loss": 0.2098, "step": 27057 }, { "epoch": 2.72, "learning_rate": 1.3386906592991317e-05, "loss": 0.4299, "step": 27058 }, { "epoch": 2.72, "learning_rate": 1.3384913926291955e-05, "loss": 0.2398, "step": 27059 }, { "epoch": 2.72, "learning_rate": 1.3382921365324681e-05, "loss": 0.2813, "step": 27060 }, { "epoch": 2.72, "learning_rate": 1.3380928910102174e-05, "loss": 0.7266, "step": 27061 }, { "epoch": 2.72, "learning_rate": 1.337893656063711e-05, "loss": 0.0714, "step": 27062 }, { "epoch": 2.72, "learning_rate": 1.3376944316942174e-05, "loss": 0.7279, "step": 27063 }, { "epoch": 2.72, "learning_rate": 1.337495217903004e-05, "loss": 0.3305, "step": 27064 }, { "epoch": 2.72, "learning_rate": 1.3372960146913385e-05, "loss": 0.5952, "step": 27065 }, { "epoch": 2.72, "learning_rate": 1.3370968220604875e-05, "loss": 0.1383, "step": 27066 }, { "epoch": 2.72, "learning_rate": 1.3368976400117219e-05, "loss": 0.075, "step": 27067 }, { "epoch": 2.72, "learning_rate": 1.3366984685463061e-05, "loss": 0.7504, "step": 27068 }, { "epoch": 2.72, "learning_rate": 1.3364993076655073e-05, "loss": 0.371, "step": 27069 }, { "epoch": 2.72, "learning_rate": 1.3363001573705952e-05, "loss": 0.0607, "step": 27070 }, { "epoch": 2.72, "learning_rate": 1.336101017662837e-05, "loss": 0.8139, "step": 27071 }, { "epoch": 2.72, "learning_rate": 1.3359018885434971e-05, "loss": 0.3206, "step": 27072 }, { "epoch": 2.72, "learning_rate": 1.3357027700138454e-05, "loss": 0.2599, "step": 27073 }, { "epoch": 2.72, "learning_rate": 1.3355036620751492e-05, "loss": 0.3541, "step": 27074 }, { "epoch": 2.72, "learning_rate": 1.3353045647286722e-05, "loss": 0.5464, "step": 27075 }, { "epoch": 2.72, "learning_rate": 1.335105477975685e-05, "loss": 0.4017, "step": 27076 }, { "epoch": 2.72, "learning_rate": 1.334906401817453e-05, "loss": 0.5064, "step": 27077 }, { "epoch": 2.72, "learning_rate": 1.3347073362552432e-05, "loss": 0.1416, "step": 27078 }, { "epoch": 2.72, "learning_rate": 1.3345082812903225e-05, "loss": 0.1658, "step": 27079 }, { "epoch": 2.72, "learning_rate": 1.3343092369239577e-05, "loss": 0.3009, "step": 27080 }, { "epoch": 2.72, "learning_rate": 1.334110203157415e-05, "loss": 0.1574, "step": 27081 }, { "epoch": 2.72, "learning_rate": 1.3339111799919612e-05, "loss": 0.1324, "step": 27082 }, { "epoch": 2.72, "learning_rate": 1.3337121674288618e-05, "loss": 0.373, "step": 27083 }, { "epoch": 2.72, "learning_rate": 1.3335131654693854e-05, "loss": 0.2713, "step": 27084 }, { "epoch": 2.72, "learning_rate": 1.3333141741147983e-05, "loss": 0.3398, "step": 27085 }, { "epoch": 2.72, "learning_rate": 1.3331151933663638e-05, "loss": 0.2044, "step": 27086 }, { "epoch": 2.72, "learning_rate": 1.3329162232253509e-05, "loss": 0.5497, "step": 27087 }, { "epoch": 2.73, "learning_rate": 1.3327172636930262e-05, "loss": 0.244, "step": 27088 }, { "epoch": 2.73, "learning_rate": 1.3325183147706523e-05, "loss": 0.1823, "step": 27089 }, { "epoch": 2.73, "learning_rate": 1.3323193764594983e-05, "loss": 0.2424, "step": 27090 }, { "epoch": 2.73, "learning_rate": 1.3321204487608299e-05, "loss": 0.464, "step": 27091 }, { "epoch": 2.73, "learning_rate": 1.3319215316759125e-05, "loss": 0.3458, "step": 27092 }, { "epoch": 2.73, "learning_rate": 1.3317226252060117e-05, "loss": 0.2322, "step": 27093 }, { "epoch": 2.73, "learning_rate": 1.3315237293523934e-05, "loss": 0.269, "step": 27094 }, { "epoch": 2.73, "learning_rate": 1.3313248441163232e-05, "loss": 0.3833, "step": 27095 }, { "epoch": 2.73, "learning_rate": 1.3311259694990672e-05, "loss": 0.2826, "step": 27096 }, { "epoch": 2.73, "learning_rate": 1.3309271055018907e-05, "loss": 0.0914, "step": 27097 }, { "epoch": 2.73, "learning_rate": 1.3307282521260577e-05, "loss": 0.3201, "step": 27098 }, { "epoch": 2.73, "learning_rate": 1.3305294093728375e-05, "loss": 0.2107, "step": 27099 }, { "epoch": 2.73, "learning_rate": 1.3303305772434919e-05, "loss": 0.1077, "step": 27100 }, { "epoch": 2.73, "learning_rate": 1.3301317557392862e-05, "loss": 0.4678, "step": 27101 }, { "epoch": 2.73, "learning_rate": 1.3299329448614879e-05, "loss": 0.1622, "step": 27102 }, { "epoch": 2.73, "learning_rate": 1.3297341446113625e-05, "loss": 0.2876, "step": 27103 }, { "epoch": 2.73, "learning_rate": 1.329535354990171e-05, "loss": 0.4968, "step": 27104 }, { "epoch": 2.73, "learning_rate": 1.329336575999182e-05, "loss": 0.4235, "step": 27105 }, { "epoch": 2.73, "learning_rate": 1.3291378076396608e-05, "loss": 0.6654, "step": 27106 }, { "epoch": 2.73, "learning_rate": 1.3289390499128685e-05, "loss": 0.655, "step": 27107 }, { "epoch": 2.73, "learning_rate": 1.3287403028200738e-05, "loss": 0.7711, "step": 27108 }, { "epoch": 2.73, "learning_rate": 1.3285415663625399e-05, "loss": 0.1444, "step": 27109 }, { "epoch": 2.73, "learning_rate": 1.3283428405415316e-05, "loss": 0.0714, "step": 27110 }, { "epoch": 2.73, "learning_rate": 1.3281441253583135e-05, "loss": 0.3364, "step": 27111 }, { "epoch": 2.73, "learning_rate": 1.3279454208141492e-05, "loss": 0.1254, "step": 27112 }, { "epoch": 2.73, "learning_rate": 1.3277467269103065e-05, "loss": 0.0607, "step": 27113 }, { "epoch": 2.73, "learning_rate": 1.3275480436480459e-05, "loss": 0.228, "step": 27114 }, { "epoch": 2.73, "learning_rate": 1.3273493710286327e-05, "loss": 0.3907, "step": 27115 }, { "epoch": 2.73, "learning_rate": 1.3271507090533325e-05, "loss": 0.1908, "step": 27116 }, { "epoch": 2.73, "learning_rate": 1.32695205772341e-05, "loss": 0.2599, "step": 27117 }, { "epoch": 2.73, "learning_rate": 1.3267534170401263e-05, "loss": 0.1069, "step": 27118 }, { "epoch": 2.73, "learning_rate": 1.326554787004748e-05, "loss": 0.0371, "step": 27119 }, { "epoch": 2.73, "learning_rate": 1.3263561676185397e-05, "loss": 0.379, "step": 27120 }, { "epoch": 2.73, "learning_rate": 1.3261575588827618e-05, "loss": 0.054, "step": 27121 }, { "epoch": 2.73, "learning_rate": 1.3259589607986816e-05, "loss": 0.8595, "step": 27122 }, { "epoch": 2.73, "learning_rate": 1.3257603733675619e-05, "loss": 0.4484, "step": 27123 }, { "epoch": 2.73, "learning_rate": 1.325561796590666e-05, "loss": 0.1516, "step": 27124 }, { "epoch": 2.73, "learning_rate": 1.3253632304692578e-05, "loss": 0.2159, "step": 27125 }, { "epoch": 2.73, "learning_rate": 1.325164675004601e-05, "loss": 0.8871, "step": 27126 }, { "epoch": 2.73, "learning_rate": 1.3249661301979593e-05, "loss": 0.3626, "step": 27127 }, { "epoch": 2.73, "learning_rate": 1.3247675960505957e-05, "loss": 0.315, "step": 27128 }, { "epoch": 2.73, "learning_rate": 1.324569072563774e-05, "loss": 0.4907, "step": 27129 }, { "epoch": 2.73, "learning_rate": 1.3243705597387562e-05, "loss": 0.5108, "step": 27130 }, { "epoch": 2.73, "learning_rate": 1.3241720575768088e-05, "loss": 0.2582, "step": 27131 }, { "epoch": 2.73, "learning_rate": 1.3239735660791919e-05, "loss": 0.1297, "step": 27132 }, { "epoch": 2.73, "learning_rate": 1.3237750852471684e-05, "loss": 0.7617, "step": 27133 }, { "epoch": 2.73, "learning_rate": 1.3235766150820039e-05, "loss": 0.2878, "step": 27134 }, { "epoch": 2.73, "learning_rate": 1.323378155584961e-05, "loss": 0.3219, "step": 27135 }, { "epoch": 2.73, "learning_rate": 1.3231797067572996e-05, "loss": 0.4138, "step": 27136 }, { "epoch": 2.73, "learning_rate": 1.3229812686002856e-05, "loss": 0.3075, "step": 27137 }, { "epoch": 2.73, "learning_rate": 1.3227828411151809e-05, "loss": 0.9419, "step": 27138 }, { "epoch": 2.73, "learning_rate": 1.3225844243032482e-05, "loss": 0.4592, "step": 27139 }, { "epoch": 2.73, "learning_rate": 1.3223860181657503e-05, "loss": 0.3482, "step": 27140 }, { "epoch": 2.73, "learning_rate": 1.3221876227039493e-05, "loss": 0.5496, "step": 27141 }, { "epoch": 2.73, "learning_rate": 1.3219892379191079e-05, "loss": 0.4083, "step": 27142 }, { "epoch": 2.73, "learning_rate": 1.3217908638124888e-05, "loss": 0.2373, "step": 27143 }, { "epoch": 2.73, "learning_rate": 1.3215925003853534e-05, "loss": 0.3685, "step": 27144 }, { "epoch": 2.73, "learning_rate": 1.3213941476389666e-05, "loss": 0.6283, "step": 27145 }, { "epoch": 2.73, "learning_rate": 1.3211958055745881e-05, "loss": 0.6962, "step": 27146 }, { "epoch": 2.73, "learning_rate": 1.3209974741934794e-05, "loss": 0.1179, "step": 27147 }, { "epoch": 2.73, "learning_rate": 1.3207991534969055e-05, "loss": 0.0924, "step": 27148 }, { "epoch": 2.73, "learning_rate": 1.3206008434861278e-05, "loss": 0.2169, "step": 27149 }, { "epoch": 2.73, "learning_rate": 1.3204025441624052e-05, "loss": 0.4272, "step": 27150 }, { "epoch": 2.73, "learning_rate": 1.3202042555270033e-05, "loss": 0.5061, "step": 27151 }, { "epoch": 2.73, "learning_rate": 1.3200059775811836e-05, "loss": 0.0882, "step": 27152 }, { "epoch": 2.73, "learning_rate": 1.3198077103262041e-05, "loss": 0.4927, "step": 27153 }, { "epoch": 2.73, "learning_rate": 1.3196094537633307e-05, "loss": 0.2966, "step": 27154 }, { "epoch": 2.73, "learning_rate": 1.3194112078938233e-05, "loss": 0.214, "step": 27155 }, { "epoch": 2.73, "learning_rate": 1.3192129727189437e-05, "loss": 0.3735, "step": 27156 }, { "epoch": 2.73, "learning_rate": 1.3190147482399532e-05, "loss": 0.2057, "step": 27157 }, { "epoch": 2.73, "learning_rate": 1.3188165344581134e-05, "loss": 0.4298, "step": 27158 }, { "epoch": 2.73, "learning_rate": 1.318618331374686e-05, "loss": 0.1871, "step": 27159 }, { "epoch": 2.73, "learning_rate": 1.3184201389909312e-05, "loss": 0.1204, "step": 27160 }, { "epoch": 2.73, "learning_rate": 1.3182219573081114e-05, "loss": 0.3701, "step": 27161 }, { "epoch": 2.73, "learning_rate": 1.318023786327486e-05, "loss": 0.3804, "step": 27162 }, { "epoch": 2.73, "learning_rate": 1.31782562605032e-05, "loss": 0.467, "step": 27163 }, { "epoch": 2.73, "learning_rate": 1.317627476477869e-05, "loss": 0.9151, "step": 27164 }, { "epoch": 2.73, "learning_rate": 1.317429337611398e-05, "loss": 0.3732, "step": 27165 }, { "epoch": 2.73, "learning_rate": 1.3172312094521667e-05, "loss": 0.3746, "step": 27166 }, { "epoch": 2.73, "learning_rate": 1.317033092001436e-05, "loss": 0.0333, "step": 27167 }, { "epoch": 2.73, "learning_rate": 1.3168349852604662e-05, "loss": 0.4121, "step": 27168 }, { "epoch": 2.73, "learning_rate": 1.3166368892305185e-05, "loss": 0.1565, "step": 27169 }, { "epoch": 2.73, "learning_rate": 1.3164388039128532e-05, "loss": 0.3846, "step": 27170 }, { "epoch": 2.73, "learning_rate": 1.3162407293087307e-05, "loss": 0.3276, "step": 27171 }, { "epoch": 2.73, "learning_rate": 1.3160426654194118e-05, "loss": 0.5748, "step": 27172 }, { "epoch": 2.73, "learning_rate": 1.3158446122461566e-05, "loss": 0.4577, "step": 27173 }, { "epoch": 2.73, "learning_rate": 1.3156465697902258e-05, "loss": 0.5316, "step": 27174 }, { "epoch": 2.73, "learning_rate": 1.3154485380528794e-05, "loss": 0.252, "step": 27175 }, { "epoch": 2.73, "learning_rate": 1.3152505170353766e-05, "loss": 0.2617, "step": 27176 }, { "epoch": 2.73, "learning_rate": 1.3150525067389805e-05, "loss": 0.2354, "step": 27177 }, { "epoch": 2.73, "learning_rate": 1.3148545071649483e-05, "loss": 0.4471, "step": 27178 }, { "epoch": 2.73, "learning_rate": 1.3146565183145403e-05, "loss": 0.2371, "step": 27179 }, { "epoch": 2.73, "learning_rate": 1.3144585401890178e-05, "loss": 0.3811, "step": 27180 }, { "epoch": 2.73, "learning_rate": 1.314260572789641e-05, "loss": 0.798, "step": 27181 }, { "epoch": 2.73, "learning_rate": 1.3140626161176663e-05, "loss": 0.0924, "step": 27182 }, { "epoch": 2.73, "learning_rate": 1.3138646701743572e-05, "loss": 0.2307, "step": 27183 }, { "epoch": 2.73, "learning_rate": 1.3136667349609729e-05, "loss": 0.7985, "step": 27184 }, { "epoch": 2.73, "learning_rate": 1.3134688104787698e-05, "loss": 0.1593, "step": 27185 }, { "epoch": 2.73, "learning_rate": 1.3132708967290108e-05, "loss": 0.0617, "step": 27186 }, { "epoch": 2.74, "learning_rate": 1.313072993712954e-05, "loss": 0.6586, "step": 27187 }, { "epoch": 2.74, "learning_rate": 1.3128751014318587e-05, "loss": 0.2015, "step": 27188 }, { "epoch": 2.74, "learning_rate": 1.3126772198869847e-05, "loss": 0.5291, "step": 27189 }, { "epoch": 2.74, "learning_rate": 1.312479349079591e-05, "loss": 0.4651, "step": 27190 }, { "epoch": 2.74, "learning_rate": 1.3122814890109368e-05, "loss": 0.6013, "step": 27191 }, { "epoch": 2.74, "learning_rate": 1.3120836396822808e-05, "loss": 0.0801, "step": 27192 }, { "epoch": 2.74, "learning_rate": 1.3118858010948818e-05, "loss": 0.1079, "step": 27193 }, { "epoch": 2.74, "learning_rate": 1.3116879732500002e-05, "loss": 0.3462, "step": 27194 }, { "epoch": 2.74, "learning_rate": 1.3114901561488952e-05, "loss": 0.8326, "step": 27195 }, { "epoch": 2.74, "learning_rate": 1.311292349792822e-05, "loss": 0.208, "step": 27196 }, { "epoch": 2.74, "learning_rate": 1.3110945541830435e-05, "loss": 0.2488, "step": 27197 }, { "epoch": 2.74, "learning_rate": 1.3108967693208167e-05, "loss": 0.7676, "step": 27198 }, { "epoch": 2.74, "learning_rate": 1.3106989952074001e-05, "loss": 0.3695, "step": 27199 }, { "epoch": 2.74, "learning_rate": 1.3105012318440524e-05, "loss": 0.2945, "step": 27200 }, { "epoch": 2.74, "learning_rate": 1.310303479232032e-05, "loss": 0.0769, "step": 27201 }, { "epoch": 2.74, "learning_rate": 1.310105737372598e-05, "loss": 0.0781, "step": 27202 }, { "epoch": 2.74, "learning_rate": 1.3099080062670077e-05, "loss": 0.1493, "step": 27203 }, { "epoch": 2.74, "learning_rate": 1.3097102859165203e-05, "loss": 0.4268, "step": 27204 }, { "epoch": 2.74, "learning_rate": 1.3095125763223936e-05, "loss": 0.6942, "step": 27205 }, { "epoch": 2.74, "learning_rate": 1.3093148774858853e-05, "loss": 0.6507, "step": 27206 }, { "epoch": 2.74, "learning_rate": 1.3091171894082543e-05, "loss": 0.1281, "step": 27207 }, { "epoch": 2.74, "learning_rate": 1.3089195120907573e-05, "loss": 0.1148, "step": 27208 }, { "epoch": 2.74, "learning_rate": 1.308721845534655e-05, "loss": 0.4193, "step": 27209 }, { "epoch": 2.74, "learning_rate": 1.3085241897412026e-05, "loss": 0.4932, "step": 27210 }, { "epoch": 2.74, "learning_rate": 1.3083265447116578e-05, "loss": 0.1548, "step": 27211 }, { "epoch": 2.74, "learning_rate": 1.3081289104472803e-05, "loss": 0.4541, "step": 27212 }, { "epoch": 2.74, "learning_rate": 1.307931286949328e-05, "loss": 0.4445, "step": 27213 }, { "epoch": 2.74, "learning_rate": 1.3077336742190547e-05, "loss": 0.1266, "step": 27214 }, { "epoch": 2.74, "learning_rate": 1.3075360722577221e-05, "loss": 0.5442, "step": 27215 }, { "epoch": 2.74, "learning_rate": 1.3073384810665871e-05, "loss": 0.5284, "step": 27216 }, { "epoch": 2.74, "learning_rate": 1.3071409006469036e-05, "loss": 0.4367, "step": 27217 }, { "epoch": 2.74, "learning_rate": 1.3069433309999328e-05, "loss": 0.2036, "step": 27218 }, { "epoch": 2.74, "learning_rate": 1.3067457721269307e-05, "loss": 0.2112, "step": 27219 }, { "epoch": 2.74, "learning_rate": 1.3065482240291539e-05, "loss": 0.4178, "step": 27220 }, { "epoch": 2.74, "learning_rate": 1.3063506867078603e-05, "loss": 0.2764, "step": 27221 }, { "epoch": 2.74, "learning_rate": 1.3061531601643056e-05, "loss": 0.3694, "step": 27222 }, { "epoch": 2.74, "learning_rate": 1.3059556443997497e-05, "loss": 0.602, "step": 27223 }, { "epoch": 2.74, "learning_rate": 1.3057581394154468e-05, "loss": 0.5425, "step": 27224 }, { "epoch": 2.74, "learning_rate": 1.3055606452126535e-05, "loss": 0.1089, "step": 27225 }, { "epoch": 2.74, "learning_rate": 1.3053631617926286e-05, "loss": 0.6666, "step": 27226 }, { "epoch": 2.74, "learning_rate": 1.3051656891566291e-05, "loss": 0.5991, "step": 27227 }, { "epoch": 2.74, "learning_rate": 1.3049682273059077e-05, "loss": 0.1512, "step": 27228 }, { "epoch": 2.74, "learning_rate": 1.3047707762417252e-05, "loss": 0.2062, "step": 27229 }, { "epoch": 2.74, "learning_rate": 1.3045733359653367e-05, "loss": 0.1778, "step": 27230 }, { "epoch": 2.74, "learning_rate": 1.3043759064779984e-05, "loss": 0.451, "step": 27231 }, { "epoch": 2.74, "learning_rate": 1.3041784877809668e-05, "loss": 0.157, "step": 27232 }, { "epoch": 2.74, "learning_rate": 1.3039810798754982e-05, "loss": 0.4353, "step": 27233 }, { "epoch": 2.74, "learning_rate": 1.3037836827628489e-05, "loss": 0.4327, "step": 27234 }, { "epoch": 2.74, "learning_rate": 1.3035862964442748e-05, "loss": 0.6099, "step": 27235 }, { "epoch": 2.74, "learning_rate": 1.3033889209210322e-05, "loss": 0.3991, "step": 27236 }, { "epoch": 2.74, "learning_rate": 1.303191556194377e-05, "loss": 0.5348, "step": 27237 }, { "epoch": 2.74, "learning_rate": 1.3029942022655652e-05, "loss": 0.675, "step": 27238 }, { "epoch": 2.74, "learning_rate": 1.302796859135853e-05, "loss": 0.1411, "step": 27239 }, { "epoch": 2.74, "learning_rate": 1.3025995268064943e-05, "loss": 0.3714, "step": 27240 }, { "epoch": 2.74, "learning_rate": 1.3024022052787494e-05, "loss": 0.0705, "step": 27241 }, { "epoch": 2.74, "learning_rate": 1.3022048945538698e-05, "loss": 0.1867, "step": 27242 }, { "epoch": 2.74, "learning_rate": 1.3020075946331108e-05, "loss": 0.3349, "step": 27243 }, { "epoch": 2.74, "learning_rate": 1.301810305517731e-05, "loss": 0.211, "step": 27244 }, { "epoch": 2.74, "learning_rate": 1.3016130272089845e-05, "loss": 0.4853, "step": 27245 }, { "epoch": 2.74, "learning_rate": 1.301415759708126e-05, "loss": 0.0847, "step": 27246 }, { "epoch": 2.74, "learning_rate": 1.301218503016412e-05, "loss": 0.4225, "step": 27247 }, { "epoch": 2.74, "learning_rate": 1.3010212571350971e-05, "loss": 0.317, "step": 27248 }, { "epoch": 2.74, "learning_rate": 1.3008240220654364e-05, "loss": 0.4134, "step": 27249 }, { "epoch": 2.74, "learning_rate": 1.3006267978086855e-05, "loss": 0.4099, "step": 27250 }, { "epoch": 2.74, "learning_rate": 1.3004295843660989e-05, "loss": 0.5514, "step": 27251 }, { "epoch": 2.74, "learning_rate": 1.300232381738932e-05, "loss": 0.1456, "step": 27252 }, { "epoch": 2.74, "learning_rate": 1.3000351899284397e-05, "loss": 0.3574, "step": 27253 }, { "epoch": 2.74, "learning_rate": 1.2998380089358755e-05, "loss": 0.3669, "step": 27254 }, { "epoch": 2.74, "learning_rate": 1.2996408387624979e-05, "loss": 1.1451, "step": 27255 }, { "epoch": 2.74, "learning_rate": 1.2994436794095577e-05, "loss": 0.4195, "step": 27256 }, { "epoch": 2.74, "learning_rate": 1.2992465308783102e-05, "loss": 0.5616, "step": 27257 }, { "epoch": 2.74, "learning_rate": 1.2990493931700116e-05, "loss": 0.1319, "step": 27258 }, { "epoch": 2.74, "learning_rate": 1.2988522662859168e-05, "loss": 0.4458, "step": 27259 }, { "epoch": 2.74, "learning_rate": 1.2986551502272768e-05, "loss": 0.1884, "step": 27260 }, { "epoch": 2.74, "learning_rate": 1.2984580449953495e-05, "loss": 0.3212, "step": 27261 }, { "epoch": 2.74, "learning_rate": 1.2982609505913877e-05, "loss": 0.3759, "step": 27262 }, { "epoch": 2.74, "learning_rate": 1.2980638670166459e-05, "loss": 0.4508, "step": 27263 }, { "epoch": 2.74, "learning_rate": 1.297866794272378e-05, "loss": 0.254, "step": 27264 }, { "epoch": 2.74, "learning_rate": 1.2976697323598388e-05, "loss": 0.5736, "step": 27265 }, { "epoch": 2.74, "learning_rate": 1.2974726812802813e-05, "loss": 0.1637, "step": 27266 }, { "epoch": 2.74, "learning_rate": 1.2972756410349602e-05, "loss": 0.2114, "step": 27267 }, { "epoch": 2.74, "learning_rate": 1.2970786116251292e-05, "loss": 0.2402, "step": 27268 }, { "epoch": 2.74, "learning_rate": 1.2968815930520419e-05, "loss": 0.1709, "step": 27269 }, { "epoch": 2.74, "learning_rate": 1.2966845853169524e-05, "loss": 0.6013, "step": 27270 }, { "epoch": 2.74, "learning_rate": 1.2964875884211144e-05, "loss": 0.3937, "step": 27271 }, { "epoch": 2.74, "learning_rate": 1.29629060236578e-05, "loss": 0.648, "step": 27272 }, { "epoch": 2.74, "learning_rate": 1.2960936271522064e-05, "loss": 0.6283, "step": 27273 }, { "epoch": 2.74, "learning_rate": 1.2958966627816426e-05, "loss": 0.4976, "step": 27274 }, { "epoch": 2.74, "learning_rate": 1.2956997092553449e-05, "loss": 0.334, "step": 27275 }, { "epoch": 2.74, "learning_rate": 1.2955027665745663e-05, "loss": 1.0604, "step": 27276 }, { "epoch": 2.74, "learning_rate": 1.2953058347405596e-05, "loss": 0.2963, "step": 27277 }, { "epoch": 2.74, "learning_rate": 1.2951089137545777e-05, "loss": 0.0525, "step": 27278 }, { "epoch": 2.74, "learning_rate": 1.2949120036178746e-05, "loss": 0.3419, "step": 27279 }, { "epoch": 2.74, "learning_rate": 1.2947151043317027e-05, "loss": 0.1911, "step": 27280 }, { "epoch": 2.74, "learning_rate": 1.294518215897315e-05, "loss": 0.3156, "step": 27281 }, { "epoch": 2.74, "learning_rate": 1.2943213383159649e-05, "loss": 0.3197, "step": 27282 }, { "epoch": 2.74, "learning_rate": 1.2941244715889046e-05, "loss": 0.3507, "step": 27283 }, { "epoch": 2.74, "learning_rate": 1.2939276157173873e-05, "loss": 0.3152, "step": 27284 }, { "epoch": 2.75, "learning_rate": 1.2937307707026656e-05, "loss": 0.3904, "step": 27285 }, { "epoch": 2.75, "learning_rate": 1.2935339365459912e-05, "loss": 0.5555, "step": 27286 }, { "epoch": 2.75, "learning_rate": 1.2933371132486198e-05, "loss": 0.1731, "step": 27287 }, { "epoch": 2.75, "learning_rate": 1.2931403008118006e-05, "loss": 0.4489, "step": 27288 }, { "epoch": 2.75, "learning_rate": 1.2929434992367861e-05, "loss": 0.2192, "step": 27289 }, { "epoch": 2.75, "learning_rate": 1.292746708524831e-05, "loss": 0.1961, "step": 27290 }, { "epoch": 2.75, "learning_rate": 1.2925499286771875e-05, "loss": 0.2824, "step": 27291 }, { "epoch": 2.75, "learning_rate": 1.2923531596951044e-05, "loss": 0.8775, "step": 27292 }, { "epoch": 2.75, "learning_rate": 1.2921564015798372e-05, "loss": 0.725, "step": 27293 }, { "epoch": 2.75, "learning_rate": 1.291959654332637e-05, "loss": 0.285, "step": 27294 }, { "epoch": 2.75, "learning_rate": 1.2917629179547558e-05, "loss": 0.7935, "step": 27295 }, { "epoch": 2.75, "learning_rate": 1.2915661924474456e-05, "loss": 0.6832, "step": 27296 }, { "epoch": 2.75, "learning_rate": 1.2913694778119583e-05, "loss": 0.1267, "step": 27297 }, { "epoch": 2.75, "learning_rate": 1.2911727740495456e-05, "loss": 0.2965, "step": 27298 }, { "epoch": 2.75, "learning_rate": 1.2909760811614593e-05, "loss": 0.3039, "step": 27299 }, { "epoch": 2.75, "learning_rate": 1.2907793991489496e-05, "loss": 0.0915, "step": 27300 }, { "epoch": 2.75, "learning_rate": 1.2905827280132723e-05, "loss": 0.1723, "step": 27301 }, { "epoch": 2.75, "learning_rate": 1.290386067755675e-05, "loss": 0.7788, "step": 27302 }, { "epoch": 2.75, "learning_rate": 1.2901894183774094e-05, "loss": 0.442, "step": 27303 }, { "epoch": 2.75, "learning_rate": 1.2899927798797286e-05, "loss": 0.1239, "step": 27304 }, { "epoch": 2.75, "learning_rate": 1.2897961522638845e-05, "loss": 0.4153, "step": 27305 }, { "epoch": 2.75, "learning_rate": 1.2895995355311249e-05, "loss": 0.0755, "step": 27306 }, { "epoch": 2.75, "learning_rate": 1.2894029296827043e-05, "loss": 0.259, "step": 27307 }, { "epoch": 2.75, "learning_rate": 1.2892063347198729e-05, "loss": 0.2021, "step": 27308 }, { "epoch": 2.75, "learning_rate": 1.2890097506438813e-05, "loss": 0.1298, "step": 27309 }, { "epoch": 2.75, "learning_rate": 1.288813177455981e-05, "loss": 0.3247, "step": 27310 }, { "epoch": 2.75, "learning_rate": 1.2886166151574226e-05, "loss": 0.4437, "step": 27311 }, { "epoch": 2.75, "learning_rate": 1.2884200637494568e-05, "loss": 0.1806, "step": 27312 }, { "epoch": 2.75, "learning_rate": 1.2882235232333346e-05, "loss": 1.0747, "step": 27313 }, { "epoch": 2.75, "learning_rate": 1.288026993610307e-05, "loss": 0.24, "step": 27314 }, { "epoch": 2.75, "learning_rate": 1.2878304748816242e-05, "loss": 0.5221, "step": 27315 }, { "epoch": 2.75, "learning_rate": 1.287633967048537e-05, "loss": 0.3432, "step": 27316 }, { "epoch": 2.75, "learning_rate": 1.2874374701122957e-05, "loss": 0.3352, "step": 27317 }, { "epoch": 2.75, "learning_rate": 1.2872409840741495e-05, "loss": 0.1665, "step": 27318 }, { "epoch": 2.75, "learning_rate": 1.2870445089353529e-05, "loss": 0.2233, "step": 27319 }, { "epoch": 2.75, "learning_rate": 1.2868480446971518e-05, "loss": 0.3381, "step": 27320 }, { "epoch": 2.75, "learning_rate": 1.286651591360797e-05, "loss": 0.2886, "step": 27321 }, { "epoch": 2.75, "learning_rate": 1.2864551489275409e-05, "loss": 0.1107, "step": 27322 }, { "epoch": 2.75, "learning_rate": 1.2862587173986336e-05, "loss": 0.0576, "step": 27323 }, { "epoch": 2.75, "learning_rate": 1.2860622967753217e-05, "loss": 0.3725, "step": 27324 }, { "epoch": 2.75, "learning_rate": 1.285865887058858e-05, "loss": 0.8799, "step": 27325 }, { "epoch": 2.75, "learning_rate": 1.285669488250492e-05, "loss": 0.1763, "step": 27326 }, { "epoch": 2.75, "learning_rate": 1.2854731003514733e-05, "loss": 0.3249, "step": 27327 }, { "epoch": 2.75, "learning_rate": 1.2852767233630514e-05, "loss": 0.1528, "step": 27328 }, { "epoch": 2.75, "learning_rate": 1.285080357286476e-05, "loss": 0.4942, "step": 27329 }, { "epoch": 2.75, "learning_rate": 1.2848840021229974e-05, "loss": 0.1651, "step": 27330 }, { "epoch": 2.75, "learning_rate": 1.2846876578738639e-05, "loss": 0.4112, "step": 27331 }, { "epoch": 2.75, "learning_rate": 1.2844913245403248e-05, "loss": 0.4504, "step": 27332 }, { "epoch": 2.75, "learning_rate": 1.2842950021236324e-05, "loss": 0.3399, "step": 27333 }, { "epoch": 2.75, "learning_rate": 1.2840986906250326e-05, "loss": 0.1009, "step": 27334 }, { "epoch": 2.75, "learning_rate": 1.2839023900457751e-05, "loss": 0.839, "step": 27335 }, { "epoch": 2.75, "learning_rate": 1.2837061003871113e-05, "loss": 0.1373, "step": 27336 }, { "epoch": 2.75, "learning_rate": 1.2835098216502897e-05, "loss": 0.158, "step": 27337 }, { "epoch": 2.75, "learning_rate": 1.2833135538365568e-05, "loss": 0.3426, "step": 27338 }, { "epoch": 2.75, "learning_rate": 1.2831172969471641e-05, "loss": 0.8195, "step": 27339 }, { "epoch": 2.75, "learning_rate": 1.2829210509833598e-05, "loss": 0.0795, "step": 27340 }, { "epoch": 2.75, "learning_rate": 1.2827248159463928e-05, "loss": 0.2124, "step": 27341 }, { "epoch": 2.75, "learning_rate": 1.2825285918375118e-05, "loss": 0.5185, "step": 27342 }, { "epoch": 2.75, "learning_rate": 1.2823323786579655e-05, "loss": 0.6339, "step": 27343 }, { "epoch": 2.75, "learning_rate": 1.2821361764090025e-05, "loss": 0.3217, "step": 27344 }, { "epoch": 2.75, "learning_rate": 1.2819399850918716e-05, "loss": 0.3479, "step": 27345 }, { "epoch": 2.75, "learning_rate": 1.2817438047078209e-05, "loss": 0.0469, "step": 27346 }, { "epoch": 2.75, "learning_rate": 1.2815476352580989e-05, "loss": 0.2982, "step": 27347 }, { "epoch": 2.75, "learning_rate": 1.281351476743954e-05, "loss": 0.1388, "step": 27348 }, { "epoch": 2.75, "learning_rate": 1.2811553291666342e-05, "loss": 0.6226, "step": 27349 }, { "epoch": 2.75, "learning_rate": 1.2809591925273874e-05, "loss": 0.5302, "step": 27350 }, { "epoch": 2.75, "learning_rate": 1.2807630668274643e-05, "loss": 0.4618, "step": 27351 }, { "epoch": 2.75, "learning_rate": 1.2805669520681099e-05, "loss": 0.484, "step": 27352 }, { "epoch": 2.75, "learning_rate": 1.2803708482505721e-05, "loss": 0.1785, "step": 27353 }, { "epoch": 2.75, "learning_rate": 1.2801747553761017e-05, "loss": 0.6325, "step": 27354 }, { "epoch": 2.75, "learning_rate": 1.2799786734459442e-05, "loss": 0.5361, "step": 27355 }, { "epoch": 2.75, "learning_rate": 1.2797826024613486e-05, "loss": 0.1446, "step": 27356 }, { "epoch": 2.75, "learning_rate": 1.2795865424235618e-05, "loss": 0.3085, "step": 27357 }, { "epoch": 2.75, "learning_rate": 1.279390493333832e-05, "loss": 0.6095, "step": 27358 }, { "epoch": 2.75, "learning_rate": 1.2791944551934065e-05, "loss": 0.3278, "step": 27359 }, { "epoch": 2.75, "learning_rate": 1.2789984280035327e-05, "loss": 0.8232, "step": 27360 }, { "epoch": 2.75, "learning_rate": 1.2788024117654581e-05, "loss": 0.2224, "step": 27361 }, { "epoch": 2.75, "learning_rate": 1.2786064064804306e-05, "loss": 0.2522, "step": 27362 }, { "epoch": 2.75, "learning_rate": 1.2784104121496969e-05, "loss": 0.553, "step": 27363 }, { "epoch": 2.75, "learning_rate": 1.278214428774503e-05, "loss": 0.8492, "step": 27364 }, { "epoch": 2.75, "learning_rate": 1.2780184563561e-05, "loss": 0.5112, "step": 27365 }, { "epoch": 2.75, "learning_rate": 1.2778224948957313e-05, "loss": 0.2463, "step": 27366 }, { "epoch": 2.75, "learning_rate": 1.2776265443946442e-05, "loss": 0.7852, "step": 27367 }, { "epoch": 2.75, "learning_rate": 1.2774306048540872e-05, "loss": 0.4082, "step": 27368 }, { "epoch": 2.75, "learning_rate": 1.2772346762753081e-05, "loss": 0.2694, "step": 27369 }, { "epoch": 2.75, "learning_rate": 1.2770387586595495e-05, "loss": 0.9185, "step": 27370 }, { "epoch": 2.75, "learning_rate": 1.2768428520080623e-05, "loss": 0.581, "step": 27371 }, { "epoch": 2.75, "learning_rate": 1.2766469563220913e-05, "loss": 0.2388, "step": 27372 }, { "epoch": 2.75, "learning_rate": 1.2764510716028838e-05, "loss": 0.2234, "step": 27373 }, { "epoch": 2.75, "learning_rate": 1.2762551978516858e-05, "loss": 0.3966, "step": 27374 }, { "epoch": 2.75, "learning_rate": 1.276059335069744e-05, "loss": 0.41, "step": 27375 }, { "epoch": 2.75, "learning_rate": 1.275863483258305e-05, "loss": 0.11, "step": 27376 }, { "epoch": 2.75, "learning_rate": 1.2756676424186145e-05, "loss": 0.7642, "step": 27377 }, { "epoch": 2.75, "learning_rate": 1.2754718125519194e-05, "loss": 0.2837, "step": 27378 }, { "epoch": 2.75, "learning_rate": 1.2752759936594657e-05, "loss": 0.1905, "step": 27379 }, { "epoch": 2.75, "learning_rate": 1.2750801857424994e-05, "loss": 0.1425, "step": 27380 }, { "epoch": 2.75, "learning_rate": 1.2748843888022654e-05, "loss": 0.3701, "step": 27381 }, { "epoch": 2.75, "learning_rate": 1.274688602840012e-05, "loss": 0.7354, "step": 27382 }, { "epoch": 2.75, "learning_rate": 1.274492827856985e-05, "loss": 0.5146, "step": 27383 }, { "epoch": 2.76, "learning_rate": 1.2742970638544272e-05, "loss": 0.1969, "step": 27384 }, { "epoch": 2.76, "learning_rate": 1.2741013108335876e-05, "loss": 0.1581, "step": 27385 }, { "epoch": 2.76, "learning_rate": 1.2739055687957105e-05, "loss": 0.1917, "step": 27386 }, { "epoch": 2.76, "learning_rate": 1.2737098377420417e-05, "loss": 0.0551, "step": 27387 }, { "epoch": 2.76, "learning_rate": 1.273514117673827e-05, "loss": 0.5859, "step": 27388 }, { "epoch": 2.76, "learning_rate": 1.2733184085923115e-05, "loss": 0.5823, "step": 27389 }, { "epoch": 2.76, "learning_rate": 1.2731227104987406e-05, "loss": 0.903, "step": 27390 }, { "epoch": 2.76, "learning_rate": 1.27292702339436e-05, "loss": 0.199, "step": 27391 }, { "epoch": 2.76, "learning_rate": 1.2727313472804148e-05, "loss": 0.1559, "step": 27392 }, { "epoch": 2.76, "learning_rate": 1.2725356821581503e-05, "loss": 0.0462, "step": 27393 }, { "epoch": 2.76, "learning_rate": 1.2723400280288116e-05, "loss": 0.0792, "step": 27394 }, { "epoch": 2.76, "learning_rate": 1.2721443848936436e-05, "loss": 0.6666, "step": 27395 }, { "epoch": 2.76, "learning_rate": 1.2719487527538904e-05, "loss": 0.0637, "step": 27396 }, { "epoch": 2.76, "learning_rate": 1.2717531316108001e-05, "loss": 0.6555, "step": 27397 }, { "epoch": 2.76, "learning_rate": 1.2715575214656142e-05, "loss": 0.5366, "step": 27398 }, { "epoch": 2.76, "learning_rate": 1.2713619223195779e-05, "loss": 0.4417, "step": 27399 }, { "epoch": 2.76, "learning_rate": 1.2711663341739377e-05, "loss": 0.1581, "step": 27400 }, { "epoch": 2.76, "learning_rate": 1.2709707570299381e-05, "loss": 0.1575, "step": 27401 }, { "epoch": 2.76, "learning_rate": 1.2707751908888212e-05, "loss": 0.1908, "step": 27402 }, { "epoch": 2.76, "learning_rate": 1.2705796357518338e-05, "loss": 0.6517, "step": 27403 }, { "epoch": 2.76, "learning_rate": 1.2703840916202198e-05, "loss": 0.1234, "step": 27404 }, { "epoch": 2.76, "learning_rate": 1.2701885584952234e-05, "loss": 0.7665, "step": 27405 }, { "epoch": 2.76, "learning_rate": 1.269993036378089e-05, "loss": 0.4545, "step": 27406 }, { "epoch": 2.76, "learning_rate": 1.2697975252700607e-05, "loss": 0.0981, "step": 27407 }, { "epoch": 2.76, "learning_rate": 1.2696020251723833e-05, "loss": 0.4655, "step": 27408 }, { "epoch": 2.76, "learning_rate": 1.2694065360862997e-05, "loss": 0.4494, "step": 27409 }, { "epoch": 2.76, "learning_rate": 1.2692110580130537e-05, "loss": 0.5084, "step": 27410 }, { "epoch": 2.76, "learning_rate": 1.2690155909538921e-05, "loss": 0.6816, "step": 27411 }, { "epoch": 2.76, "learning_rate": 1.2688201349100558e-05, "loss": 0.451, "step": 27412 }, { "epoch": 2.76, "learning_rate": 1.2686246898827889e-05, "loss": 0.4269, "step": 27413 }, { "epoch": 2.76, "learning_rate": 1.2684292558733366e-05, "loss": 0.1513, "step": 27414 }, { "epoch": 2.76, "learning_rate": 1.2682338328829425e-05, "loss": 0.2597, "step": 27415 }, { "epoch": 2.76, "learning_rate": 1.2680384209128478e-05, "loss": 0.5342, "step": 27416 }, { "epoch": 2.76, "learning_rate": 1.2678430199642985e-05, "loss": 0.4952, "step": 27417 }, { "epoch": 2.76, "learning_rate": 1.2676476300385375e-05, "loss": 0.2614, "step": 27418 }, { "epoch": 2.76, "learning_rate": 1.267452251136808e-05, "loss": 0.4846, "step": 27419 }, { "epoch": 2.76, "learning_rate": 1.2672568832603531e-05, "loss": 0.2323, "step": 27420 }, { "epoch": 2.76, "learning_rate": 1.2670615264104161e-05, "loss": 0.0443, "step": 27421 }, { "epoch": 2.76, "learning_rate": 1.2668661805882407e-05, "loss": 0.6427, "step": 27422 }, { "epoch": 2.76, "learning_rate": 1.2666708457950693e-05, "loss": 0.5553, "step": 27423 }, { "epoch": 2.76, "learning_rate": 1.2664755220321455e-05, "loss": 0.0186, "step": 27424 }, { "epoch": 2.76, "learning_rate": 1.2662802093007117e-05, "loss": 0.1171, "step": 27425 }, { "epoch": 2.76, "learning_rate": 1.266084907602011e-05, "loss": 0.2814, "step": 27426 }, { "epoch": 2.76, "learning_rate": 1.2658896169372868e-05, "loss": 0.499, "step": 27427 }, { "epoch": 2.76, "learning_rate": 1.2656943373077801e-05, "loss": 0.4489, "step": 27428 }, { "epoch": 2.76, "learning_rate": 1.265499068714737e-05, "loss": 0.202, "step": 27429 }, { "epoch": 2.76, "learning_rate": 1.265303811159397e-05, "loss": 0.3386, "step": 27430 }, { "epoch": 2.76, "learning_rate": 1.2651085646430025e-05, "loss": 0.2888, "step": 27431 }, { "epoch": 2.76, "learning_rate": 1.2649133291667983e-05, "loss": 0.054, "step": 27432 }, { "epoch": 2.76, "learning_rate": 1.2647181047320266e-05, "loss": 0.5869, "step": 27433 }, { "epoch": 2.76, "learning_rate": 1.2645228913399267e-05, "loss": 0.3665, "step": 27434 }, { "epoch": 2.76, "learning_rate": 1.2643276889917435e-05, "loss": 0.7655, "step": 27435 }, { "epoch": 2.76, "learning_rate": 1.264132497688719e-05, "loss": 0.1752, "step": 27436 }, { "epoch": 2.76, "learning_rate": 1.2639373174320948e-05, "loss": 0.7531, "step": 27437 }, { "epoch": 2.76, "learning_rate": 1.2637421482231133e-05, "loss": 0.1204, "step": 27438 }, { "epoch": 2.76, "learning_rate": 1.2635469900630161e-05, "loss": 0.2596, "step": 27439 }, { "epoch": 2.76, "learning_rate": 1.2633518429530453e-05, "loss": 0.129, "step": 27440 }, { "epoch": 2.76, "learning_rate": 1.2631567068944426e-05, "loss": 0.3371, "step": 27441 }, { "epoch": 2.76, "learning_rate": 1.2629615818884492e-05, "loss": 0.2045, "step": 27442 }, { "epoch": 2.76, "learning_rate": 1.2627664679363095e-05, "loss": 0.1504, "step": 27443 }, { "epoch": 2.76, "learning_rate": 1.2625713650392616e-05, "loss": 0.9084, "step": 27444 }, { "epoch": 2.76, "learning_rate": 1.262376273198548e-05, "loss": 0.2155, "step": 27445 }, { "epoch": 2.76, "learning_rate": 1.2621811924154115e-05, "loss": 0.1054, "step": 27446 }, { "epoch": 2.76, "learning_rate": 1.2619861226910942e-05, "loss": 0.6763, "step": 27447 }, { "epoch": 2.76, "learning_rate": 1.2617910640268335e-05, "loss": 0.9025, "step": 27448 }, { "epoch": 2.76, "learning_rate": 1.2615960164238746e-05, "loss": 0.3022, "step": 27449 }, { "epoch": 2.76, "learning_rate": 1.2614009798834573e-05, "loss": 0.5614, "step": 27450 }, { "epoch": 2.76, "learning_rate": 1.2612059544068226e-05, "loss": 0.1451, "step": 27451 }, { "epoch": 2.76, "learning_rate": 1.2610109399952121e-05, "loss": 0.3102, "step": 27452 }, { "epoch": 2.76, "learning_rate": 1.2608159366498661e-05, "loss": 0.3624, "step": 27453 }, { "epoch": 2.76, "learning_rate": 1.2606209443720262e-05, "loss": 0.0428, "step": 27454 }, { "epoch": 2.76, "learning_rate": 1.2604259631629323e-05, "loss": 0.6344, "step": 27455 }, { "epoch": 2.76, "learning_rate": 1.2602309930238262e-05, "loss": 0.6554, "step": 27456 }, { "epoch": 2.76, "learning_rate": 1.2600360339559482e-05, "loss": 0.3096, "step": 27457 }, { "epoch": 2.76, "learning_rate": 1.259841085960539e-05, "loss": 0.2083, "step": 27458 }, { "epoch": 2.76, "learning_rate": 1.2596461490388392e-05, "loss": 0.3502, "step": 27459 }, { "epoch": 2.76, "learning_rate": 1.2594512231920878e-05, "loss": 0.6166, "step": 27460 }, { "epoch": 2.76, "learning_rate": 1.2592563084215294e-05, "loss": 0.1885, "step": 27461 }, { "epoch": 2.76, "learning_rate": 1.2590614047283989e-05, "loss": 0.3927, "step": 27462 }, { "epoch": 2.76, "learning_rate": 1.2588665121139408e-05, "loss": 0.2859, "step": 27463 }, { "epoch": 2.76, "learning_rate": 1.2586716305793935e-05, "loss": 0.3213, "step": 27464 }, { "epoch": 2.76, "learning_rate": 1.2584767601259981e-05, "loss": 0.2268, "step": 27465 }, { "epoch": 2.76, "learning_rate": 1.2582819007549936e-05, "loss": 0.6201, "step": 27466 }, { "epoch": 2.76, "learning_rate": 1.2580870524676207e-05, "loss": 0.3676, "step": 27467 }, { "epoch": 2.76, "learning_rate": 1.257892215265119e-05, "loss": 0.1844, "step": 27468 }, { "epoch": 2.76, "learning_rate": 1.2576973891487286e-05, "loss": 0.4887, "step": 27469 }, { "epoch": 2.76, "learning_rate": 1.2575025741196891e-05, "loss": 0.2763, "step": 27470 }, { "epoch": 2.76, "learning_rate": 1.2573077701792405e-05, "loss": 0.2067, "step": 27471 }, { "epoch": 2.76, "learning_rate": 1.2571129773286222e-05, "loss": 0.1095, "step": 27472 }, { "epoch": 2.76, "learning_rate": 1.256918195569074e-05, "loss": 0.592, "step": 27473 }, { "epoch": 2.76, "learning_rate": 1.256723424901834e-05, "loss": 0.2653, "step": 27474 }, { "epoch": 2.76, "learning_rate": 1.2565286653281453e-05, "loss": 0.1722, "step": 27475 }, { "epoch": 2.76, "learning_rate": 1.256333916849244e-05, "loss": 0.5781, "step": 27476 }, { "epoch": 2.76, "learning_rate": 1.2561391794663689e-05, "loss": 0.3168, "step": 27477 }, { "epoch": 2.76, "learning_rate": 1.255944453180762e-05, "loss": 0.1124, "step": 27478 }, { "epoch": 2.76, "learning_rate": 1.255749737993662e-05, "loss": 0.5711, "step": 27479 }, { "epoch": 2.76, "learning_rate": 1.255555033906305e-05, "loss": 0.1872, "step": 27480 }, { "epoch": 2.76, "learning_rate": 1.2553603409199333e-05, "loss": 0.5188, "step": 27481 }, { "epoch": 2.77, "learning_rate": 1.2551656590357848e-05, "loss": 1.0248, "step": 27482 }, { "epoch": 2.77, "learning_rate": 1.254970988255098e-05, "loss": 0.2567, "step": 27483 }, { "epoch": 2.77, "learning_rate": 1.2547763285791123e-05, "loss": 0.49, "step": 27484 }, { "epoch": 2.77, "learning_rate": 1.2545816800090659e-05, "loss": 0.8947, "step": 27485 }, { "epoch": 2.77, "learning_rate": 1.2543870425461974e-05, "loss": 0.307, "step": 27486 }, { "epoch": 2.77, "learning_rate": 1.2541924161917465e-05, "loss": 0.1484, "step": 27487 }, { "epoch": 2.77, "learning_rate": 1.2539978009469501e-05, "loss": 0.5218, "step": 27488 }, { "epoch": 2.77, "learning_rate": 1.253803196813048e-05, "loss": 0.1644, "step": 27489 }, { "epoch": 2.77, "learning_rate": 1.253608603791278e-05, "loss": 0.3631, "step": 27490 }, { "epoch": 2.77, "learning_rate": 1.2534140218828773e-05, "loss": 0.4701, "step": 27491 }, { "epoch": 2.77, "learning_rate": 1.2532194510890863e-05, "loss": 0.2296, "step": 27492 }, { "epoch": 2.77, "learning_rate": 1.2530248914111435e-05, "loss": 0.3673, "step": 27493 }, { "epoch": 2.77, "learning_rate": 1.2528303428502836e-05, "loss": 0.1724, "step": 27494 }, { "epoch": 2.77, "learning_rate": 1.2526358054077476e-05, "loss": 0.749, "step": 27495 }, { "epoch": 2.77, "learning_rate": 1.2524412790847727e-05, "loss": 0.143, "step": 27496 }, { "epoch": 2.77, "learning_rate": 1.2522467638825966e-05, "loss": 0.5459, "step": 27497 }, { "epoch": 2.77, "learning_rate": 1.2520522598024575e-05, "loss": 0.4142, "step": 27498 }, { "epoch": 2.77, "learning_rate": 1.2518577668455928e-05, "loss": 0.7778, "step": 27499 }, { "epoch": 2.77, "learning_rate": 1.2516632850132399e-05, "loss": 0.1606, "step": 27500 }, { "epoch": 2.77, "learning_rate": 1.2514688143066371e-05, "loss": 0.3244, "step": 27501 }, { "epoch": 2.77, "learning_rate": 1.2512743547270212e-05, "loss": 0.3895, "step": 27502 }, { "epoch": 2.77, "learning_rate": 1.2510799062756305e-05, "loss": 0.4588, "step": 27503 }, { "epoch": 2.77, "learning_rate": 1.2508854689537016e-05, "loss": 0.344, "step": 27504 }, { "epoch": 2.77, "learning_rate": 1.2506910427624725e-05, "loss": 0.6696, "step": 27505 }, { "epoch": 2.77, "learning_rate": 1.2504966277031787e-05, "loss": 0.2657, "step": 27506 }, { "epoch": 2.77, "learning_rate": 1.2503022237770614e-05, "loss": 0.2571, "step": 27507 }, { "epoch": 2.77, "learning_rate": 1.2501078309853537e-05, "loss": 0.6692, "step": 27508 }, { "epoch": 2.77, "learning_rate": 1.2499134493292932e-05, "loss": 0.4649, "step": 27509 }, { "epoch": 2.77, "learning_rate": 1.2497190788101188e-05, "loss": 0.4789, "step": 27510 }, { "epoch": 2.77, "learning_rate": 1.2495247194290673e-05, "loss": 0.1407, "step": 27511 }, { "epoch": 2.77, "learning_rate": 1.2493303711873724e-05, "loss": 0.5141, "step": 27512 }, { "epoch": 2.77, "learning_rate": 1.2491360340862742e-05, "loss": 0.1365, "step": 27513 }, { "epoch": 2.77, "learning_rate": 1.2489417081270081e-05, "loss": 0.1328, "step": 27514 }, { "epoch": 2.77, "learning_rate": 1.2487473933108112e-05, "loss": 0.3704, "step": 27515 }, { "epoch": 2.77, "learning_rate": 1.2485530896389191e-05, "loss": 0.4389, "step": 27516 }, { "epoch": 2.77, "learning_rate": 1.2483587971125681e-05, "loss": 0.0953, "step": 27517 }, { "epoch": 2.77, "learning_rate": 1.2481645157329977e-05, "loss": 0.2947, "step": 27518 }, { "epoch": 2.77, "learning_rate": 1.2479702455014404e-05, "loss": 0.1576, "step": 27519 }, { "epoch": 2.77, "learning_rate": 1.2477759864191333e-05, "loss": 0.3763, "step": 27520 }, { "epoch": 2.77, "learning_rate": 1.2475817384873152e-05, "loss": 0.2063, "step": 27521 }, { "epoch": 2.77, "learning_rate": 1.2473875017072192e-05, "loss": 0.2786, "step": 27522 }, { "epoch": 2.77, "learning_rate": 1.247193276080081e-05, "loss": 0.3017, "step": 27523 }, { "epoch": 2.77, "learning_rate": 1.2469990616071395e-05, "loss": 0.6014, "step": 27524 }, { "epoch": 2.77, "learning_rate": 1.2468048582896302e-05, "loss": 0.3589, "step": 27525 }, { "epoch": 2.77, "learning_rate": 1.2466106661287855e-05, "loss": 0.5683, "step": 27526 }, { "epoch": 2.77, "learning_rate": 1.2464164851258444e-05, "loss": 0.6396, "step": 27527 }, { "epoch": 2.77, "learning_rate": 1.246222315282042e-05, "loss": 0.6787, "step": 27528 }, { "epoch": 2.77, "learning_rate": 1.2460281565986134e-05, "loss": 0.6176, "step": 27529 }, { "epoch": 2.77, "learning_rate": 1.2458340090767944e-05, "loss": 0.3518, "step": 27530 }, { "epoch": 2.77, "learning_rate": 1.2456398727178205e-05, "loss": 0.3277, "step": 27531 }, { "epoch": 2.77, "learning_rate": 1.2454457475229273e-05, "loss": 0.5628, "step": 27532 }, { "epoch": 2.77, "learning_rate": 1.2452516334933494e-05, "loss": 0.8203, "step": 27533 }, { "epoch": 2.77, "learning_rate": 1.2450575306303228e-05, "loss": 0.7658, "step": 27534 }, { "epoch": 2.77, "learning_rate": 1.2448634389350826e-05, "loss": 0.1653, "step": 27535 }, { "epoch": 2.77, "learning_rate": 1.2446693584088638e-05, "loss": 0.6963, "step": 27536 }, { "epoch": 2.77, "learning_rate": 1.2444752890529013e-05, "loss": 0.2781, "step": 27537 }, { "epoch": 2.77, "learning_rate": 1.244281230868429e-05, "loss": 0.0611, "step": 27538 }, { "epoch": 2.77, "learning_rate": 1.2440871838566853e-05, "loss": 0.1919, "step": 27539 }, { "epoch": 2.77, "learning_rate": 1.2438931480189017e-05, "loss": 0.3311, "step": 27540 }, { "epoch": 2.77, "learning_rate": 1.2436991233563133e-05, "loss": 0.4246, "step": 27541 }, { "epoch": 2.77, "learning_rate": 1.2435051098701563e-05, "loss": 0.2397, "step": 27542 }, { "epoch": 2.77, "learning_rate": 1.2433111075616645e-05, "loss": 0.5339, "step": 27543 }, { "epoch": 2.77, "learning_rate": 1.2431171164320727e-05, "loss": 0.1918, "step": 27544 }, { "epoch": 2.77, "learning_rate": 1.2429231364826151e-05, "loss": 0.6914, "step": 27545 }, { "epoch": 2.77, "learning_rate": 1.2427291677145264e-05, "loss": 0.2285, "step": 27546 }, { "epoch": 2.77, "learning_rate": 1.2425352101290409e-05, "loss": 0.1789, "step": 27547 }, { "epoch": 2.77, "learning_rate": 1.2423412637273925e-05, "loss": 0.154, "step": 27548 }, { "epoch": 2.77, "learning_rate": 1.2421473285108147e-05, "loss": 0.1772, "step": 27549 }, { "epoch": 2.77, "learning_rate": 1.2419534044805442e-05, "loss": 0.2934, "step": 27550 }, { "epoch": 2.77, "learning_rate": 1.2417594916378128e-05, "loss": 0.1359, "step": 27551 }, { "epoch": 2.77, "learning_rate": 1.2415655899838544e-05, "loss": 0.9302, "step": 27552 }, { "epoch": 2.77, "learning_rate": 1.2413716995199055e-05, "loss": 0.5434, "step": 27553 }, { "epoch": 2.77, "learning_rate": 1.2411778202471969e-05, "loss": 0.16, "step": 27554 }, { "epoch": 2.77, "learning_rate": 1.2409839521669624e-05, "loss": 0.2766, "step": 27555 }, { "epoch": 2.77, "learning_rate": 1.2407900952804381e-05, "loss": 0.491, "step": 27556 }, { "epoch": 2.77, "learning_rate": 1.2405962495888577e-05, "loss": 0.3277, "step": 27557 }, { "epoch": 2.77, "learning_rate": 1.240402415093451e-05, "loss": 0.2054, "step": 27558 }, { "epoch": 2.77, "learning_rate": 1.240208591795455e-05, "loss": 0.2845, "step": 27559 }, { "epoch": 2.77, "learning_rate": 1.240014779696102e-05, "loss": 0.6506, "step": 27560 }, { "epoch": 2.77, "learning_rate": 1.2398209787966256e-05, "loss": 0.2858, "step": 27561 }, { "epoch": 2.77, "learning_rate": 1.239627189098259e-05, "loss": 0.3718, "step": 27562 }, { "epoch": 2.77, "learning_rate": 1.2394334106022349e-05, "loss": 0.0642, "step": 27563 }, { "epoch": 2.77, "learning_rate": 1.2392396433097874e-05, "loss": 0.0945, "step": 27564 }, { "epoch": 2.77, "learning_rate": 1.2390458872221486e-05, "loss": 0.2284, "step": 27565 }, { "epoch": 2.77, "learning_rate": 1.2388521423405518e-05, "loss": 0.2663, "step": 27566 }, { "epoch": 2.77, "learning_rate": 1.2386584086662301e-05, "loss": 0.0521, "step": 27567 }, { "epoch": 2.77, "learning_rate": 1.2384646862004162e-05, "loss": 0.5509, "step": 27568 }, { "epoch": 2.77, "learning_rate": 1.2382709749443429e-05, "loss": 0.1517, "step": 27569 }, { "epoch": 2.77, "learning_rate": 1.2380772748992421e-05, "loss": 0.2285, "step": 27570 }, { "epoch": 2.77, "learning_rate": 1.237883586066349e-05, "loss": 0.745, "step": 27571 }, { "epoch": 2.77, "learning_rate": 1.2376899084468925e-05, "loss": 0.2497, "step": 27572 }, { "epoch": 2.77, "learning_rate": 1.237496242042108e-05, "loss": 0.189, "step": 27573 }, { "epoch": 2.77, "learning_rate": 1.237302586853227e-05, "loss": 0.2245, "step": 27574 }, { "epoch": 2.77, "learning_rate": 1.2371089428814816e-05, "loss": 0.5295, "step": 27575 }, { "epoch": 2.77, "learning_rate": 1.2369153101281047e-05, "loss": 0.1834, "step": 27576 }, { "epoch": 2.77, "learning_rate": 1.2367216885943275e-05, "loss": 0.4422, "step": 27577 }, { "epoch": 2.77, "learning_rate": 1.236528078281383e-05, "loss": 0.3632, "step": 27578 }, { "epoch": 2.77, "learning_rate": 1.2363344791905029e-05, "loss": 0.0953, "step": 27579 }, { "epoch": 2.77, "learning_rate": 1.2361408913229192e-05, "loss": 0.5476, "step": 27580 }, { "epoch": 2.78, "learning_rate": 1.2359473146798628e-05, "loss": 0.3113, "step": 27581 }, { "epoch": 2.78, "learning_rate": 1.235753749262569e-05, "loss": 0.3827, "step": 27582 }, { "epoch": 2.78, "learning_rate": 1.2355601950722655e-05, "loss": 0.2959, "step": 27583 }, { "epoch": 2.78, "learning_rate": 1.2353666521101852e-05, "loss": 0.647, "step": 27584 }, { "epoch": 2.78, "learning_rate": 1.2351731203775621e-05, "loss": 0.7174, "step": 27585 }, { "epoch": 2.78, "learning_rate": 1.2349795998756248e-05, "loss": 0.3539, "step": 27586 }, { "epoch": 2.78, "learning_rate": 1.234786090605605e-05, "loss": 0.1907, "step": 27587 }, { "epoch": 2.78, "learning_rate": 1.234592592568736e-05, "loss": 0.0925, "step": 27588 }, { "epoch": 2.78, "learning_rate": 1.2343991057662492e-05, "loss": 0.2788, "step": 27589 }, { "epoch": 2.78, "learning_rate": 1.2342056301993727e-05, "loss": 0.4962, "step": 27590 }, { "epoch": 2.78, "learning_rate": 1.234012165869341e-05, "loss": 0.2484, "step": 27591 }, { "epoch": 2.78, "learning_rate": 1.2338187127773837e-05, "loss": 0.0177, "step": 27592 }, { "epoch": 2.78, "learning_rate": 1.2336252709247325e-05, "loss": 0.5612, "step": 27593 }, { "epoch": 2.78, "learning_rate": 1.2334318403126179e-05, "loss": 0.4429, "step": 27594 }, { "epoch": 2.78, "learning_rate": 1.2332384209422715e-05, "loss": 0.3673, "step": 27595 }, { "epoch": 2.78, "learning_rate": 1.2330450128149232e-05, "loss": 0.2, "step": 27596 }, { "epoch": 2.78, "learning_rate": 1.2328516159318045e-05, "loss": 0.9355, "step": 27597 }, { "epoch": 2.78, "learning_rate": 1.2326582302941448e-05, "loss": 0.0302, "step": 27598 }, { "epoch": 2.78, "learning_rate": 1.2324648559031774e-05, "loss": 0.4334, "step": 27599 }, { "epoch": 2.78, "learning_rate": 1.2322714927601308e-05, "loss": 0.1263, "step": 27600 }, { "epoch": 2.78, "learning_rate": 1.2320781408662349e-05, "loss": 0.736, "step": 27601 }, { "epoch": 2.78, "learning_rate": 1.2318848002227218e-05, "loss": 0.1083, "step": 27602 }, { "epoch": 2.78, "learning_rate": 1.2316914708308226e-05, "loss": 0.124, "step": 27603 }, { "epoch": 2.78, "learning_rate": 1.2314981526917639e-05, "loss": 0.3925, "step": 27604 }, { "epoch": 2.78, "learning_rate": 1.2313048458067791e-05, "loss": 0.3846, "step": 27605 }, { "epoch": 2.78, "learning_rate": 1.2311115501770977e-05, "loss": 0.4595, "step": 27606 }, { "epoch": 2.78, "learning_rate": 1.2309182658039493e-05, "loss": 0.1788, "step": 27607 }, { "epoch": 2.78, "learning_rate": 1.2307249926885644e-05, "loss": 0.0842, "step": 27608 }, { "epoch": 2.78, "learning_rate": 1.2305317308321721e-05, "loss": 0.178, "step": 27609 }, { "epoch": 2.78, "learning_rate": 1.2303384802360031e-05, "loss": 0.2848, "step": 27610 }, { "epoch": 2.78, "learning_rate": 1.2301452409012867e-05, "loss": 0.7083, "step": 27611 }, { "epoch": 2.78, "learning_rate": 1.2299520128292528e-05, "loss": 0.2949, "step": 27612 }, { "epoch": 2.78, "learning_rate": 1.2297587960211296e-05, "loss": 0.0269, "step": 27613 }, { "epoch": 2.78, "learning_rate": 1.22956559047815e-05, "loss": 0.2389, "step": 27614 }, { "epoch": 2.78, "learning_rate": 1.2293723962015408e-05, "loss": 0.2784, "step": 27615 }, { "epoch": 2.78, "learning_rate": 1.2291792131925308e-05, "loss": 0.1664, "step": 27616 }, { "epoch": 2.78, "learning_rate": 1.2289860414523528e-05, "loss": 0.2608, "step": 27617 }, { "epoch": 2.78, "learning_rate": 1.228792880982233e-05, "loss": 0.2254, "step": 27618 }, { "epoch": 2.78, "learning_rate": 1.2285997317834003e-05, "loss": 0.4106, "step": 27619 }, { "epoch": 2.78, "learning_rate": 1.228406593857086e-05, "loss": 0.1334, "step": 27620 }, { "epoch": 2.78, "learning_rate": 1.2282134672045195e-05, "loss": 0.2888, "step": 27621 }, { "epoch": 2.78, "learning_rate": 1.228020351826926e-05, "loss": 0.7407, "step": 27622 }, { "epoch": 2.78, "learning_rate": 1.2278272477255382e-05, "loss": 0.3187, "step": 27623 }, { "epoch": 2.78, "learning_rate": 1.2276341549015836e-05, "loss": 0.0732, "step": 27624 }, { "epoch": 2.78, "learning_rate": 1.2274410733562909e-05, "loss": 0.1644, "step": 27625 }, { "epoch": 2.78, "learning_rate": 1.227248003090889e-05, "loss": 0.5573, "step": 27626 }, { "epoch": 2.78, "learning_rate": 1.2270549441066052e-05, "loss": 0.3936, "step": 27627 }, { "epoch": 2.78, "learning_rate": 1.2268618964046718e-05, "loss": 0.1819, "step": 27628 }, { "epoch": 2.78, "learning_rate": 1.2266688599863132e-05, "loss": 0.5884, "step": 27629 }, { "epoch": 2.78, "learning_rate": 1.2264758348527586e-05, "loss": 0.5986, "step": 27630 }, { "epoch": 2.78, "learning_rate": 1.2262828210052391e-05, "loss": 0.5107, "step": 27631 }, { "epoch": 2.78, "learning_rate": 1.22608981844498e-05, "loss": 0.1315, "step": 27632 }, { "epoch": 2.78, "learning_rate": 1.2258968271732093e-05, "loss": 0.2609, "step": 27633 }, { "epoch": 2.78, "learning_rate": 1.2257038471911576e-05, "loss": 0.1172, "step": 27634 }, { "epoch": 2.78, "learning_rate": 1.2255108785000524e-05, "loss": 0.097, "step": 27635 }, { "epoch": 2.78, "learning_rate": 1.2253179211011187e-05, "loss": 0.7972, "step": 27636 }, { "epoch": 2.78, "learning_rate": 1.225124974995588e-05, "loss": 0.1592, "step": 27637 }, { "epoch": 2.78, "learning_rate": 1.2249320401846868e-05, "loss": 0.5087, "step": 27638 }, { "epoch": 2.78, "learning_rate": 1.2247391166696424e-05, "loss": 0.242, "step": 27639 }, { "epoch": 2.78, "learning_rate": 1.2245462044516834e-05, "loss": 0.3056, "step": 27640 }, { "epoch": 2.78, "learning_rate": 1.2243533035320367e-05, "loss": 0.2488, "step": 27641 }, { "epoch": 2.78, "learning_rate": 1.2241604139119303e-05, "loss": 0.6758, "step": 27642 }, { "epoch": 2.78, "learning_rate": 1.2239675355925912e-05, "loss": 0.6285, "step": 27643 }, { "epoch": 2.78, "learning_rate": 1.223774668575247e-05, "loss": 0.2694, "step": 27644 }, { "epoch": 2.78, "learning_rate": 1.2235818128611243e-05, "loss": 0.7645, "step": 27645 }, { "epoch": 2.78, "learning_rate": 1.2233889684514533e-05, "loss": 1.2007, "step": 27646 }, { "epoch": 2.78, "learning_rate": 1.2231961353474576e-05, "loss": 0.1648, "step": 27647 }, { "epoch": 2.78, "learning_rate": 1.223003313550365e-05, "loss": 0.1243, "step": 27648 }, { "epoch": 2.78, "learning_rate": 1.2228105030614055e-05, "loss": 0.1498, "step": 27649 }, { "epoch": 2.78, "learning_rate": 1.2226177038818024e-05, "loss": 0.2472, "step": 27650 }, { "epoch": 2.78, "learning_rate": 1.2224249160127836e-05, "loss": 0.5244, "step": 27651 }, { "epoch": 2.78, "learning_rate": 1.2222321394555772e-05, "loss": 0.1684, "step": 27652 }, { "epoch": 2.78, "learning_rate": 1.222039374211409e-05, "loss": 0.1527, "step": 27653 }, { "epoch": 2.78, "learning_rate": 1.221846620281506e-05, "loss": 0.1212, "step": 27654 }, { "epoch": 2.78, "learning_rate": 1.2216538776670948e-05, "loss": 0.4086, "step": 27655 }, { "epoch": 2.78, "learning_rate": 1.2214611463694016e-05, "loss": 0.0472, "step": 27656 }, { "epoch": 2.78, "learning_rate": 1.2212684263896532e-05, "loss": 0.4664, "step": 27657 }, { "epoch": 2.78, "learning_rate": 1.2210757177290758e-05, "loss": 0.4627, "step": 27658 }, { "epoch": 2.78, "learning_rate": 1.2208830203888946e-05, "loss": 0.1017, "step": 27659 }, { "epoch": 2.78, "learning_rate": 1.2206903343703395e-05, "loss": 0.6579, "step": 27660 }, { "epoch": 2.78, "learning_rate": 1.220497659674633e-05, "loss": 0.2398, "step": 27661 }, { "epoch": 2.78, "learning_rate": 1.2203049963030015e-05, "loss": 0.8434, "step": 27662 }, { "epoch": 2.78, "learning_rate": 1.2201123442566738e-05, "loss": 0.7195, "step": 27663 }, { "epoch": 2.78, "learning_rate": 1.2199197035368733e-05, "loss": 0.2661, "step": 27664 }, { "epoch": 2.78, "learning_rate": 1.2197270741448253e-05, "loss": 0.3336, "step": 27665 }, { "epoch": 2.78, "learning_rate": 1.219534456081758e-05, "loss": 0.2192, "step": 27666 }, { "epoch": 2.78, "learning_rate": 1.2193418493488976e-05, "loss": 0.2194, "step": 27667 }, { "epoch": 2.78, "learning_rate": 1.2191492539474655e-05, "loss": 0.6351, "step": 27668 }, { "epoch": 2.78, "learning_rate": 1.2189566698786918e-05, "loss": 0.6582, "step": 27669 }, { "epoch": 2.78, "learning_rate": 1.2187640971438e-05, "loss": 0.1091, "step": 27670 }, { "epoch": 2.78, "learning_rate": 1.2185715357440159e-05, "loss": 0.2006, "step": 27671 }, { "epoch": 2.78, "learning_rate": 1.2183789856805644e-05, "loss": 0.3748, "step": 27672 }, { "epoch": 2.78, "learning_rate": 1.2181864469546718e-05, "loss": 0.4614, "step": 27673 }, { "epoch": 2.78, "learning_rate": 1.2179939195675629e-05, "loss": 0.2373, "step": 27674 }, { "epoch": 2.78, "learning_rate": 1.2178014035204623e-05, "loss": 1.0126, "step": 27675 }, { "epoch": 2.78, "learning_rate": 1.217608898814596e-05, "loss": 0.5654, "step": 27676 }, { "epoch": 2.78, "learning_rate": 1.2174164054511874e-05, "loss": 0.0007, "step": 27677 }, { "epoch": 2.78, "learning_rate": 1.2172239234314646e-05, "loss": 0.5181, "step": 27678 }, { "epoch": 2.79, "learning_rate": 1.2170314527566488e-05, "loss": 0.367, "step": 27679 }, { "epoch": 2.79, "learning_rate": 1.2168389934279674e-05, "loss": 0.2965, "step": 27680 }, { "epoch": 2.79, "learning_rate": 1.2166465454466455e-05, "loss": 0.186, "step": 27681 }, { "epoch": 2.79, "learning_rate": 1.2164541088139044e-05, "loss": 0.5284, "step": 27682 }, { "epoch": 2.79, "learning_rate": 1.216261683530972e-05, "loss": 0.9743, "step": 27683 }, { "epoch": 2.79, "learning_rate": 1.2160692695990713e-05, "loss": 0.3898, "step": 27684 }, { "epoch": 2.79, "learning_rate": 1.2158768670194276e-05, "loss": 0.2557, "step": 27685 }, { "epoch": 2.79, "learning_rate": 1.2156844757932646e-05, "loss": 0.6546, "step": 27686 }, { "epoch": 2.79, "learning_rate": 1.2154920959218068e-05, "loss": 0.2654, "step": 27687 }, { "epoch": 2.79, "learning_rate": 1.215299727406278e-05, "loss": 0.8559, "step": 27688 }, { "epoch": 2.79, "learning_rate": 1.2151073702479034e-05, "loss": 0.2967, "step": 27689 }, { "epoch": 2.79, "learning_rate": 1.2149150244479062e-05, "loss": 0.533, "step": 27690 }, { "epoch": 2.79, "learning_rate": 1.21472269000751e-05, "loss": 0.2916, "step": 27691 }, { "epoch": 2.79, "learning_rate": 1.2145303669279414e-05, "loss": 0.1688, "step": 27692 }, { "epoch": 2.79, "learning_rate": 1.2143380552104212e-05, "loss": 0.1222, "step": 27693 }, { "epoch": 2.79, "learning_rate": 1.2141457548561731e-05, "loss": 0.3484, "step": 27694 }, { "epoch": 2.79, "learning_rate": 1.2139534658664244e-05, "loss": 0.1838, "step": 27695 }, { "epoch": 2.79, "learning_rate": 1.2137611882423951e-05, "loss": 0.0517, "step": 27696 }, { "epoch": 2.79, "learning_rate": 1.2135689219853091e-05, "loss": 0.1869, "step": 27697 }, { "epoch": 2.79, "learning_rate": 1.2133766670963921e-05, "loss": 0.3824, "step": 27698 }, { "epoch": 2.79, "learning_rate": 1.2131844235768673e-05, "loss": 0.1112, "step": 27699 }, { "epoch": 2.79, "learning_rate": 1.212992191427955e-05, "loss": 0.1214, "step": 27700 }, { "epoch": 2.79, "learning_rate": 1.2127999706508816e-05, "loss": 0.1234, "step": 27701 }, { "epoch": 2.79, "learning_rate": 1.2126077612468692e-05, "loss": 0.6239, "step": 27702 }, { "epoch": 2.79, "learning_rate": 1.212415563217141e-05, "loss": 0.5223, "step": 27703 }, { "epoch": 2.79, "learning_rate": 1.2122233765629204e-05, "loss": 0.5624, "step": 27704 }, { "epoch": 2.79, "learning_rate": 1.2120312012854288e-05, "loss": 0.7073, "step": 27705 }, { "epoch": 2.79, "learning_rate": 1.211839037385893e-05, "loss": 0.2307, "step": 27706 }, { "epoch": 2.79, "learning_rate": 1.2116468848655318e-05, "loss": 0.3492, "step": 27707 }, { "epoch": 2.79, "learning_rate": 1.2114547437255686e-05, "loss": 0.1177, "step": 27708 }, { "epoch": 2.79, "learning_rate": 1.211262613967228e-05, "loss": 0.263, "step": 27709 }, { "epoch": 2.79, "learning_rate": 1.211070495591733e-05, "loss": 0.6795, "step": 27710 }, { "epoch": 2.79, "learning_rate": 1.2108783886003024e-05, "loss": 0.2728, "step": 27711 }, { "epoch": 2.79, "learning_rate": 1.2106862929941622e-05, "loss": 0.6519, "step": 27712 }, { "epoch": 2.79, "learning_rate": 1.2104942087745347e-05, "loss": 0.2659, "step": 27713 }, { "epoch": 2.79, "learning_rate": 1.2103021359426393e-05, "loss": 0.32, "step": 27714 }, { "epoch": 2.79, "learning_rate": 1.2101100744997014e-05, "loss": 0.3434, "step": 27715 }, { "epoch": 2.79, "learning_rate": 1.2099180244469424e-05, "loss": 0.6946, "step": 27716 }, { "epoch": 2.79, "learning_rate": 1.2097259857855838e-05, "loss": 0.5303, "step": 27717 }, { "epoch": 2.79, "learning_rate": 1.2095339585168479e-05, "loss": 0.2058, "step": 27718 }, { "epoch": 2.79, "learning_rate": 1.2093419426419566e-05, "loss": 1.0751, "step": 27719 }, { "epoch": 2.79, "learning_rate": 1.2091499381621323e-05, "loss": 0.0886, "step": 27720 }, { "epoch": 2.79, "learning_rate": 1.2089579450785965e-05, "loss": 0.4739, "step": 27721 }, { "epoch": 2.79, "learning_rate": 1.2087659633925706e-05, "loss": 0.5578, "step": 27722 }, { "epoch": 2.79, "learning_rate": 1.2085739931052761e-05, "loss": 0.3264, "step": 27723 }, { "epoch": 2.79, "learning_rate": 1.2083820342179372e-05, "loss": 0.3273, "step": 27724 }, { "epoch": 2.79, "learning_rate": 1.2081900867317723e-05, "loss": 0.2314, "step": 27725 }, { "epoch": 2.79, "learning_rate": 1.207998150648003e-05, "loss": 0.3834, "step": 27726 }, { "epoch": 2.79, "learning_rate": 1.2078062259678538e-05, "loss": 0.3766, "step": 27727 }, { "epoch": 2.79, "learning_rate": 1.2076143126925431e-05, "loss": 0.572, "step": 27728 }, { "epoch": 2.79, "learning_rate": 1.207422410823292e-05, "loss": 0.2047, "step": 27729 }, { "epoch": 2.79, "learning_rate": 1.2072305203613239e-05, "loss": 0.3637, "step": 27730 }, { "epoch": 2.79, "learning_rate": 1.2070386413078594e-05, "loss": 0.6242, "step": 27731 }, { "epoch": 2.79, "learning_rate": 1.2068467736641172e-05, "loss": 0.5881, "step": 27732 }, { "epoch": 2.79, "learning_rate": 1.2066549174313209e-05, "loss": 0.724, "step": 27733 }, { "epoch": 2.79, "learning_rate": 1.2064630726106904e-05, "loss": 0.1089, "step": 27734 }, { "epoch": 2.79, "learning_rate": 1.2062712392034468e-05, "loss": 0.221, "step": 27735 }, { "epoch": 2.79, "learning_rate": 1.2060794172108105e-05, "loss": 0.1901, "step": 27736 }, { "epoch": 2.79, "learning_rate": 1.2058876066340013e-05, "loss": 0.5954, "step": 27737 }, { "epoch": 2.79, "learning_rate": 1.2056958074742433e-05, "loss": 0.1493, "step": 27738 }, { "epoch": 2.79, "learning_rate": 1.2055040197327531e-05, "loss": 0.356, "step": 27739 }, { "epoch": 2.79, "learning_rate": 1.2053122434107518e-05, "loss": 0.4431, "step": 27740 }, { "epoch": 2.79, "learning_rate": 1.2051204785094617e-05, "loss": 0.1328, "step": 27741 }, { "epoch": 2.79, "learning_rate": 1.204928725030103e-05, "loss": 0.3455, "step": 27742 }, { "epoch": 2.79, "learning_rate": 1.204736982973893e-05, "loss": 0.4709, "step": 27743 }, { "epoch": 2.79, "learning_rate": 1.2045452523420548e-05, "loss": 0.2673, "step": 27744 }, { "epoch": 2.79, "learning_rate": 1.204353533135809e-05, "loss": 0.3668, "step": 27745 }, { "epoch": 2.79, "learning_rate": 1.2041618253563718e-05, "loss": 0.0708, "step": 27746 }, { "epoch": 2.79, "learning_rate": 1.2039701290049669e-05, "loss": 0.3057, "step": 27747 }, { "epoch": 2.79, "learning_rate": 1.2037784440828124e-05, "loss": 0.4077, "step": 27748 }, { "epoch": 2.79, "learning_rate": 1.203586770591129e-05, "loss": 0.1962, "step": 27749 }, { "epoch": 2.79, "learning_rate": 1.2033951085311357e-05, "loss": 0.5881, "step": 27750 }, { "epoch": 2.79, "learning_rate": 1.2032034579040526e-05, "loss": 0.2322, "step": 27751 }, { "epoch": 2.79, "learning_rate": 1.2030118187110992e-05, "loss": 0.7558, "step": 27752 }, { "epoch": 2.79, "learning_rate": 1.202820190953495e-05, "loss": 0.396, "step": 27753 }, { "epoch": 2.79, "learning_rate": 1.2026285746324593e-05, "loss": 0.4401, "step": 27754 }, { "epoch": 2.79, "learning_rate": 1.2024369697492104e-05, "loss": 0.5382, "step": 27755 }, { "epoch": 2.79, "learning_rate": 1.2022453763049714e-05, "loss": 0.243, "step": 27756 }, { "epoch": 2.79, "learning_rate": 1.2020537943009574e-05, "loss": 0.2965, "step": 27757 }, { "epoch": 2.79, "learning_rate": 1.2018622237383881e-05, "loss": 0.3233, "step": 27758 }, { "epoch": 2.79, "learning_rate": 1.201670664618486e-05, "loss": 0.7299, "step": 27759 }, { "epoch": 2.79, "learning_rate": 1.2014791169424654e-05, "loss": 0.199, "step": 27760 }, { "epoch": 2.79, "learning_rate": 1.2012875807115483e-05, "loss": 0.6623, "step": 27761 }, { "epoch": 2.79, "learning_rate": 1.2010960559269529e-05, "loss": 0.2195, "step": 27762 }, { "epoch": 2.79, "learning_rate": 1.2009045425898979e-05, "loss": 0.2463, "step": 27763 }, { "epoch": 2.79, "learning_rate": 1.2007130407016019e-05, "loss": 0.5192, "step": 27764 }, { "epoch": 2.79, "learning_rate": 1.200521550263284e-05, "loss": 0.3491, "step": 27765 }, { "epoch": 2.79, "learning_rate": 1.2003300712761619e-05, "loss": 0.2643, "step": 27766 }, { "epoch": 2.79, "learning_rate": 1.2001386037414548e-05, "loss": 0.1786, "step": 27767 }, { "epoch": 2.79, "learning_rate": 1.1999471476603813e-05, "loss": 0.4107, "step": 27768 }, { "epoch": 2.79, "learning_rate": 1.1997557030341577e-05, "loss": 0.4424, "step": 27769 }, { "epoch": 2.79, "learning_rate": 1.1995642698640065e-05, "loss": 0.4213, "step": 27770 }, { "epoch": 2.79, "learning_rate": 1.1993728481511422e-05, "loss": 0.3843, "step": 27771 }, { "epoch": 2.79, "learning_rate": 1.1991814378967827e-05, "loss": 0.0546, "step": 27772 }, { "epoch": 2.79, "learning_rate": 1.198990039102149e-05, "loss": 0.3739, "step": 27773 }, { "epoch": 2.79, "learning_rate": 1.1987986517684585e-05, "loss": 0.3382, "step": 27774 }, { "epoch": 2.79, "learning_rate": 1.1986072758969258e-05, "loss": 0.1638, "step": 27775 }, { "epoch": 2.79, "learning_rate": 1.1984159114887725e-05, "loss": 0.1252, "step": 27776 }, { "epoch": 2.79, "learning_rate": 1.1982245585452158e-05, "loss": 0.7717, "step": 27777 }, { "epoch": 2.8, "learning_rate": 1.1980332170674707e-05, "loss": 0.5845, "step": 27778 }, { "epoch": 2.8, "learning_rate": 1.1978418870567574e-05, "loss": 0.7146, "step": 27779 }, { "epoch": 2.8, "learning_rate": 1.197650568514293e-05, "loss": 0.0478, "step": 27780 }, { "epoch": 2.8, "learning_rate": 1.1974592614412947e-05, "loss": 0.1664, "step": 27781 }, { "epoch": 2.8, "learning_rate": 1.19726796583898e-05, "loss": 0.5484, "step": 27782 }, { "epoch": 2.8, "learning_rate": 1.197076681708566e-05, "loss": 0.1597, "step": 27783 }, { "epoch": 2.8, "learning_rate": 1.1968854090512701e-05, "loss": 0.5382, "step": 27784 }, { "epoch": 2.8, "learning_rate": 1.1966941478683099e-05, "loss": 0.853, "step": 27785 }, { "epoch": 2.8, "learning_rate": 1.1965028981609008e-05, "loss": 0.1754, "step": 27786 }, { "epoch": 2.8, "learning_rate": 1.1963116599302622e-05, "loss": 0.6124, "step": 27787 }, { "epoch": 2.8, "learning_rate": 1.1961204331776111e-05, "loss": 0.1574, "step": 27788 }, { "epoch": 2.8, "learning_rate": 1.1959292179041614e-05, "loss": 0.1786, "step": 27789 }, { "epoch": 2.8, "learning_rate": 1.1957380141111329e-05, "loss": 0.8405, "step": 27790 }, { "epoch": 2.8, "learning_rate": 1.1955468217997419e-05, "loss": 0.2627, "step": 27791 }, { "epoch": 2.8, "learning_rate": 1.1953556409712027e-05, "loss": 0.4657, "step": 27792 }, { "epoch": 2.8, "learning_rate": 1.195164471626735e-05, "loss": 0.2346, "step": 27793 }, { "epoch": 2.8, "learning_rate": 1.1949733137675535e-05, "loss": 0.101, "step": 27794 }, { "epoch": 2.8, "learning_rate": 1.1947821673948754e-05, "loss": 0.2951, "step": 27795 }, { "epoch": 2.8, "learning_rate": 1.194591032509917e-05, "loss": 0.5278, "step": 27796 }, { "epoch": 2.8, "learning_rate": 1.1943999091138944e-05, "loss": 0.5133, "step": 27797 }, { "epoch": 2.8, "learning_rate": 1.1942087972080238e-05, "loss": 0.2956, "step": 27798 }, { "epoch": 2.8, "learning_rate": 1.1940176967935214e-05, "loss": 0.2848, "step": 27799 }, { "epoch": 2.8, "learning_rate": 1.1938266078716035e-05, "loss": 0.4847, "step": 27800 }, { "epoch": 2.8, "learning_rate": 1.1936355304434846e-05, "loss": 0.1515, "step": 27801 }, { "epoch": 2.8, "learning_rate": 1.1934444645103844e-05, "loss": 0.2086, "step": 27802 }, { "epoch": 2.8, "learning_rate": 1.1932534100735151e-05, "loss": 0.3762, "step": 27803 }, { "epoch": 2.8, "learning_rate": 1.193062367134093e-05, "loss": 0.1259, "step": 27804 }, { "epoch": 2.8, "learning_rate": 1.1928713356933357e-05, "loss": 0.3387, "step": 27805 }, { "epoch": 2.8, "learning_rate": 1.1926803157524586e-05, "loss": 0.1782, "step": 27806 }, { "epoch": 2.8, "learning_rate": 1.1924893073126744e-05, "loss": 0.1292, "step": 27807 }, { "epoch": 2.8, "learning_rate": 1.1922983103752015e-05, "loss": 0.4632, "step": 27808 }, { "epoch": 2.8, "learning_rate": 1.1921073249412556e-05, "loss": 0.1465, "step": 27809 }, { "epoch": 2.8, "learning_rate": 1.191916351012049e-05, "loss": 0.0706, "step": 27810 }, { "epoch": 2.8, "learning_rate": 1.1917253885888e-05, "loss": 0.5431, "step": 27811 }, { "epoch": 2.8, "learning_rate": 1.1915344376727225e-05, "loss": 0.6041, "step": 27812 }, { "epoch": 2.8, "learning_rate": 1.191343498265032e-05, "loss": 0.3127, "step": 27813 }, { "epoch": 2.8, "learning_rate": 1.1911525703669434e-05, "loss": 0.2513, "step": 27814 }, { "epoch": 2.8, "learning_rate": 1.1909616539796703e-05, "loss": 0.5138, "step": 27815 }, { "epoch": 2.8, "learning_rate": 1.1907707491044317e-05, "loss": 0.224, "step": 27816 }, { "epoch": 2.8, "learning_rate": 1.1905798557424385e-05, "loss": 0.2015, "step": 27817 }, { "epoch": 2.8, "learning_rate": 1.190388973894906e-05, "loss": 0.2832, "step": 27818 }, { "epoch": 2.8, "learning_rate": 1.19019810356305e-05, "loss": 0.3617, "step": 27819 }, { "epoch": 2.8, "learning_rate": 1.1900072447480863e-05, "loss": 0.284, "step": 27820 }, { "epoch": 2.8, "learning_rate": 1.189816397451226e-05, "loss": 0.1228, "step": 27821 }, { "epoch": 2.8, "learning_rate": 1.1896255616736863e-05, "loss": 0.1539, "step": 27822 }, { "epoch": 2.8, "learning_rate": 1.189434737416682e-05, "loss": 0.6099, "step": 27823 }, { "epoch": 2.8, "learning_rate": 1.189243924681424e-05, "loss": 0.0827, "step": 27824 }, { "epoch": 2.8, "learning_rate": 1.18905312346913e-05, "loss": 0.2056, "step": 27825 }, { "epoch": 2.8, "learning_rate": 1.1888623337810132e-05, "loss": 0.3833, "step": 27826 }, { "epoch": 2.8, "learning_rate": 1.1886715556182874e-05, "loss": 0.1544, "step": 27827 }, { "epoch": 2.8, "learning_rate": 1.1884807889821665e-05, "loss": 0.6059, "step": 27828 }, { "epoch": 2.8, "learning_rate": 1.1882900338738648e-05, "loss": 0.6726, "step": 27829 }, { "epoch": 2.8, "learning_rate": 1.1880992902945963e-05, "loss": 0.3848, "step": 27830 }, { "epoch": 2.8, "learning_rate": 1.1879085582455743e-05, "loss": 0.226, "step": 27831 }, { "epoch": 2.8, "learning_rate": 1.1877178377280132e-05, "loss": 0.4241, "step": 27832 }, { "epoch": 2.8, "learning_rate": 1.1875271287431254e-05, "loss": 0.1478, "step": 27833 }, { "epoch": 2.8, "learning_rate": 1.1873364312921273e-05, "loss": 0.089, "step": 27834 }, { "epoch": 2.8, "learning_rate": 1.1871457453762296e-05, "loss": 0.938, "step": 27835 }, { "epoch": 2.8, "learning_rate": 1.1869550709966454e-05, "loss": 0.4018, "step": 27836 }, { "epoch": 2.8, "learning_rate": 1.186764408154591e-05, "loss": 0.2495, "step": 27837 }, { "epoch": 2.8, "learning_rate": 1.1865737568512788e-05, "loss": 0.8624, "step": 27838 }, { "epoch": 2.8, "learning_rate": 1.1863831170879193e-05, "loss": 0.705, "step": 27839 }, { "epoch": 2.8, "learning_rate": 1.1861924888657288e-05, "loss": 0.2351, "step": 27840 }, { "epoch": 2.8, "learning_rate": 1.1860018721859193e-05, "loss": 0.2976, "step": 27841 }, { "epoch": 2.8, "learning_rate": 1.185811267049704e-05, "loss": 0.3154, "step": 27842 }, { "epoch": 2.8, "learning_rate": 1.1856206734582952e-05, "loss": 0.1713, "step": 27843 }, { "epoch": 2.8, "learning_rate": 1.1854300914129064e-05, "loss": 0.7148, "step": 27844 }, { "epoch": 2.8, "learning_rate": 1.1852395209147505e-05, "loss": 0.8599, "step": 27845 }, { "epoch": 2.8, "learning_rate": 1.1850489619650398e-05, "loss": 0.1332, "step": 27846 }, { "epoch": 2.8, "learning_rate": 1.1848584145649856e-05, "loss": 0.6894, "step": 27847 }, { "epoch": 2.8, "learning_rate": 1.1846678787158043e-05, "loss": 0.9023, "step": 27848 }, { "epoch": 2.8, "learning_rate": 1.1844773544187052e-05, "loss": 0.3448, "step": 27849 }, { "epoch": 2.8, "learning_rate": 1.1842868416749003e-05, "loss": 0.0342, "step": 27850 }, { "epoch": 2.8, "learning_rate": 1.1840963404856044e-05, "loss": 0.3819, "step": 27851 }, { "epoch": 2.8, "learning_rate": 1.1839058508520294e-05, "loss": 0.3723, "step": 27852 }, { "epoch": 2.8, "learning_rate": 1.1837153727753846e-05, "loss": 0.3991, "step": 27853 }, { "epoch": 2.8, "learning_rate": 1.1835249062568856e-05, "loss": 0.3318, "step": 27854 }, { "epoch": 2.8, "learning_rate": 1.1833344512977435e-05, "loss": 0.0632, "step": 27855 }, { "epoch": 2.8, "learning_rate": 1.1831440078991685e-05, "loss": 0.4408, "step": 27856 }, { "epoch": 2.8, "learning_rate": 1.1829535760623742e-05, "loss": 0.2228, "step": 27857 }, { "epoch": 2.8, "learning_rate": 1.1827631557885725e-05, "loss": 0.2315, "step": 27858 }, { "epoch": 2.8, "learning_rate": 1.1825727470789748e-05, "loss": 0.4692, "step": 27859 }, { "epoch": 2.8, "learning_rate": 1.1823823499347926e-05, "loss": 0.2412, "step": 27860 }, { "epoch": 2.8, "learning_rate": 1.1821919643572379e-05, "loss": 0.3492, "step": 27861 }, { "epoch": 2.8, "learning_rate": 1.1820015903475218e-05, "loss": 0.3067, "step": 27862 }, { "epoch": 2.8, "learning_rate": 1.1818112279068561e-05, "loss": 0.1509, "step": 27863 }, { "epoch": 2.8, "learning_rate": 1.1816208770364521e-05, "loss": 0.3684, "step": 27864 }, { "epoch": 2.8, "learning_rate": 1.18143053773752e-05, "loss": 0.1945, "step": 27865 }, { "epoch": 2.8, "learning_rate": 1.1812402100112741e-05, "loss": 0.1205, "step": 27866 }, { "epoch": 2.8, "learning_rate": 1.1810498938589226e-05, "loss": 0.6253, "step": 27867 }, { "epoch": 2.8, "learning_rate": 1.1808595892816766e-05, "loss": 0.049, "step": 27868 }, { "epoch": 2.8, "learning_rate": 1.1806692962807492e-05, "loss": 0.2077, "step": 27869 }, { "epoch": 2.8, "learning_rate": 1.1804790148573503e-05, "loss": 0.466, "step": 27870 }, { "epoch": 2.8, "learning_rate": 1.1802887450126908e-05, "loss": 0.1256, "step": 27871 }, { "epoch": 2.8, "learning_rate": 1.1800984867479813e-05, "loss": 0.5281, "step": 27872 }, { "epoch": 2.8, "learning_rate": 1.1799082400644327e-05, "loss": 0.4777, "step": 27873 }, { "epoch": 2.8, "learning_rate": 1.179718004963256e-05, "loss": 0.2513, "step": 27874 }, { "epoch": 2.8, "learning_rate": 1.179527781445661e-05, "loss": 0.2252, "step": 27875 }, { "epoch": 2.81, "learning_rate": 1.179337569512859e-05, "loss": 0.1335, "step": 27876 }, { "epoch": 2.81, "learning_rate": 1.1791473691660601e-05, "loss": 0.1849, "step": 27877 }, { "epoch": 2.81, "learning_rate": 1.1789571804064745e-05, "loss": 0.2024, "step": 27878 }, { "epoch": 2.81, "learning_rate": 1.1787670032353115e-05, "loss": 0.2451, "step": 27879 }, { "epoch": 2.81, "learning_rate": 1.1785768376537848e-05, "loss": 0.444, "step": 27880 }, { "epoch": 2.81, "learning_rate": 1.178386683663101e-05, "loss": 0.5277, "step": 27881 }, { "epoch": 2.81, "learning_rate": 1.1781965412644705e-05, "loss": 0.2118, "step": 27882 }, { "epoch": 2.81, "learning_rate": 1.1780064104591051e-05, "loss": 0.1739, "step": 27883 }, { "epoch": 2.81, "learning_rate": 1.1778162912482146e-05, "loss": 0.1781, "step": 27884 }, { "epoch": 2.81, "learning_rate": 1.1776261836330064e-05, "loss": 0.1492, "step": 27885 }, { "epoch": 2.81, "learning_rate": 1.1774360876146924e-05, "loss": 0.2201, "step": 27886 }, { "epoch": 2.81, "learning_rate": 1.1772460031944831e-05, "loss": 0.2739, "step": 27887 }, { "epoch": 2.81, "learning_rate": 1.1770559303735849e-05, "loss": 0.2434, "step": 27888 }, { "epoch": 2.81, "learning_rate": 1.1768658691532101e-05, "loss": 0.4543, "step": 27889 }, { "epoch": 2.81, "learning_rate": 1.1766758195345673e-05, "loss": 0.4526, "step": 27890 }, { "epoch": 2.81, "learning_rate": 1.1764857815188663e-05, "loss": 0.309, "step": 27891 }, { "epoch": 2.81, "learning_rate": 1.1762957551073157e-05, "loss": 0.416, "step": 27892 }, { "epoch": 2.81, "learning_rate": 1.1761057403011255e-05, "loss": 0.0777, "step": 27893 }, { "epoch": 2.81, "learning_rate": 1.1759157371015044e-05, "loss": 0.5489, "step": 27894 }, { "epoch": 2.81, "learning_rate": 1.1757257455096616e-05, "loss": 0.7506, "step": 27895 }, { "epoch": 2.81, "learning_rate": 1.1755357655268049e-05, "loss": 0.0657, "step": 27896 }, { "epoch": 2.81, "learning_rate": 1.1753457971541459e-05, "loss": 0.5295, "step": 27897 }, { "epoch": 2.81, "learning_rate": 1.1751558403928931e-05, "loss": 0.1247, "step": 27898 }, { "epoch": 2.81, "learning_rate": 1.174965895244252e-05, "loss": 0.3206, "step": 27899 }, { "epoch": 2.81, "learning_rate": 1.1747759617094348e-05, "loss": 1.5493, "step": 27900 }, { "epoch": 2.81, "learning_rate": 1.1745860397896491e-05, "loss": 0.2471, "step": 27901 }, { "epoch": 2.81, "learning_rate": 1.1743961294861035e-05, "loss": 0.4765, "step": 27902 }, { "epoch": 2.81, "learning_rate": 1.1742062308000065e-05, "loss": 0.4681, "step": 27903 }, { "epoch": 2.81, "learning_rate": 1.1740163437325662e-05, "loss": 0.3709, "step": 27904 }, { "epoch": 2.81, "learning_rate": 1.1738264682849914e-05, "loss": 0.1928, "step": 27905 }, { "epoch": 2.81, "learning_rate": 1.1736366044584903e-05, "loss": 0.0618, "step": 27906 }, { "epoch": 2.81, "learning_rate": 1.173446752254271e-05, "loss": 0.5304, "step": 27907 }, { "epoch": 2.81, "learning_rate": 1.1732569116735415e-05, "loss": 0.4089, "step": 27908 }, { "epoch": 2.81, "learning_rate": 1.1730670827175102e-05, "loss": 0.5239, "step": 27909 }, { "epoch": 2.81, "learning_rate": 1.1728772653873853e-05, "loss": 0.159, "step": 27910 }, { "epoch": 2.81, "learning_rate": 1.1726874596843731e-05, "loss": 0.1733, "step": 27911 }, { "epoch": 2.81, "learning_rate": 1.172497665609685e-05, "loss": 0.5696, "step": 27912 }, { "epoch": 2.81, "learning_rate": 1.1723078831645253e-05, "loss": 0.146, "step": 27913 }, { "epoch": 2.81, "learning_rate": 1.1721181123501019e-05, "loss": 0.774, "step": 27914 }, { "epoch": 2.81, "learning_rate": 1.1719283531676245e-05, "loss": 0.5174, "step": 27915 }, { "epoch": 2.81, "learning_rate": 1.171738605618301e-05, "loss": 0.5658, "step": 27916 }, { "epoch": 2.81, "learning_rate": 1.1715488697033353e-05, "loss": 0.1098, "step": 27917 }, { "epoch": 2.81, "learning_rate": 1.1713591454239378e-05, "loss": 0.1562, "step": 27918 }, { "epoch": 2.81, "learning_rate": 1.1711694327813164e-05, "loss": 0.0632, "step": 27919 }, { "epoch": 2.81, "learning_rate": 1.1709797317766746e-05, "loss": 0.4498, "step": 27920 }, { "epoch": 2.81, "learning_rate": 1.1707900424112231e-05, "loss": 0.1202, "step": 27921 }, { "epoch": 2.81, "learning_rate": 1.1706003646861683e-05, "loss": 0.0521, "step": 27922 }, { "epoch": 2.81, "learning_rate": 1.1704106986027164e-05, "loss": 0.2211, "step": 27923 }, { "epoch": 2.81, "learning_rate": 1.170221044162075e-05, "loss": 0.4155, "step": 27924 }, { "epoch": 2.81, "learning_rate": 1.1700314013654497e-05, "loss": 0.1512, "step": 27925 }, { "epoch": 2.81, "learning_rate": 1.1698417702140504e-05, "loss": 0.9191, "step": 27926 }, { "epoch": 2.81, "learning_rate": 1.1696521507090809e-05, "loss": 0.7947, "step": 27927 }, { "epoch": 2.81, "learning_rate": 1.1694625428517476e-05, "loss": 0.3662, "step": 27928 }, { "epoch": 2.81, "learning_rate": 1.1692729466432597e-05, "loss": 0.1582, "step": 27929 }, { "epoch": 2.81, "learning_rate": 1.1690833620848232e-05, "loss": 0.3127, "step": 27930 }, { "epoch": 2.81, "learning_rate": 1.1688937891776414e-05, "loss": 0.1095, "step": 27931 }, { "epoch": 2.81, "learning_rate": 1.1687042279229243e-05, "loss": 0.4502, "step": 27932 }, { "epoch": 2.81, "learning_rate": 1.1685146783218766e-05, "loss": 0.2665, "step": 27933 }, { "epoch": 2.81, "learning_rate": 1.1683251403757048e-05, "loss": 0.2965, "step": 27934 }, { "epoch": 2.81, "learning_rate": 1.1681356140856149e-05, "loss": 0.1365, "step": 27935 }, { "epoch": 2.81, "learning_rate": 1.1679460994528134e-05, "loss": 0.2853, "step": 27936 }, { "epoch": 2.81, "learning_rate": 1.1677565964785058e-05, "loss": 0.4684, "step": 27937 }, { "epoch": 2.81, "learning_rate": 1.167567105163898e-05, "loss": 0.1556, "step": 27938 }, { "epoch": 2.81, "learning_rate": 1.167377625510196e-05, "loss": 0.5831, "step": 27939 }, { "epoch": 2.81, "learning_rate": 1.1671881575186062e-05, "loss": 0.0351, "step": 27940 }, { "epoch": 2.81, "learning_rate": 1.1669987011903333e-05, "loss": 0.1517, "step": 27941 }, { "epoch": 2.81, "learning_rate": 1.1668092565265834e-05, "loss": 0.4515, "step": 27942 }, { "epoch": 2.81, "learning_rate": 1.1666198235285612e-05, "loss": 0.4561, "step": 27943 }, { "epoch": 2.81, "learning_rate": 1.1664304021974747e-05, "loss": 0.7767, "step": 27944 }, { "epoch": 2.81, "learning_rate": 1.1662409925345267e-05, "loss": 0.2437, "step": 27945 }, { "epoch": 2.81, "learning_rate": 1.1660515945409225e-05, "loss": 0.4966, "step": 27946 }, { "epoch": 2.81, "learning_rate": 1.1658622082178692e-05, "loss": 0.5496, "step": 27947 }, { "epoch": 2.81, "learning_rate": 1.1656728335665721e-05, "loss": 0.4248, "step": 27948 }, { "epoch": 2.81, "learning_rate": 1.1654834705882332e-05, "loss": 0.6509, "step": 27949 }, { "epoch": 2.81, "learning_rate": 1.1652941192840607e-05, "loss": 0.3042, "step": 27950 }, { "epoch": 2.81, "learning_rate": 1.1651047796552585e-05, "loss": 0.0247, "step": 27951 }, { "epoch": 2.81, "learning_rate": 1.1649154517030313e-05, "loss": 0.1628, "step": 27952 }, { "epoch": 2.81, "learning_rate": 1.1647261354285843e-05, "loss": 0.6216, "step": 27953 }, { "epoch": 2.81, "learning_rate": 1.1645368308331217e-05, "loss": 0.6731, "step": 27954 }, { "epoch": 2.81, "learning_rate": 1.1643475379178486e-05, "loss": 0.4536, "step": 27955 }, { "epoch": 2.81, "learning_rate": 1.1641582566839696e-05, "loss": 0.2633, "step": 27956 }, { "epoch": 2.81, "learning_rate": 1.1639689871326878e-05, "loss": 0.7384, "step": 27957 }, { "epoch": 2.81, "learning_rate": 1.163779729265211e-05, "loss": 0.1783, "step": 27958 }, { "epoch": 2.81, "learning_rate": 1.1635904830827405e-05, "loss": 0.2283, "step": 27959 }, { "epoch": 2.81, "learning_rate": 1.1634012485864807e-05, "loss": 0.5042, "step": 27960 }, { "epoch": 2.81, "learning_rate": 1.1632120257776374e-05, "loss": 0.6257, "step": 27961 }, { "epoch": 2.81, "learning_rate": 1.1630228146574155e-05, "loss": 0.4342, "step": 27962 }, { "epoch": 2.81, "learning_rate": 1.1628336152270151e-05, "loss": 1.1222, "step": 27963 }, { "epoch": 2.81, "learning_rate": 1.1626444274876436e-05, "loss": 0.103, "step": 27964 }, { "epoch": 2.81, "learning_rate": 1.1624552514405044e-05, "loss": 1.3272, "step": 27965 }, { "epoch": 2.81, "learning_rate": 1.1622660870868009e-05, "loss": 0.7818, "step": 27966 }, { "epoch": 2.81, "learning_rate": 1.1620769344277367e-05, "loss": 0.623, "step": 27967 }, { "epoch": 2.81, "learning_rate": 1.1618877934645158e-05, "loss": 0.6927, "step": 27968 }, { "epoch": 2.81, "learning_rate": 1.1616986641983417e-05, "loss": 0.1333, "step": 27969 }, { "epoch": 2.81, "learning_rate": 1.1615095466304178e-05, "loss": 0.2454, "step": 27970 }, { "epoch": 2.81, "learning_rate": 1.1613204407619481e-05, "loss": 0.1826, "step": 27971 }, { "epoch": 2.81, "learning_rate": 1.1611313465941354e-05, "loss": 0.4416, "step": 27972 }, { "epoch": 2.81, "learning_rate": 1.1609422641281835e-05, "loss": 0.7007, "step": 27973 }, { "epoch": 2.81, "learning_rate": 1.1607531933652952e-05, "loss": 0.3192, "step": 27974 }, { "epoch": 2.82, "learning_rate": 1.160564134306673e-05, "loss": 0.2722, "step": 27975 }, { "epoch": 2.82, "learning_rate": 1.160375086953523e-05, "loss": 0.2991, "step": 27976 }, { "epoch": 2.82, "learning_rate": 1.1601860513070438e-05, "loss": 0.7043, "step": 27977 }, { "epoch": 2.82, "learning_rate": 1.159997027368442e-05, "loss": 0.2884, "step": 27978 }, { "epoch": 2.82, "learning_rate": 1.1598080151389188e-05, "loss": 0.1, "step": 27979 }, { "epoch": 2.82, "learning_rate": 1.1596190146196777e-05, "loss": 0.4415, "step": 27980 }, { "epoch": 2.82, "learning_rate": 1.1594300258119206e-05, "loss": 0.0987, "step": 27981 }, { "epoch": 2.82, "learning_rate": 1.1592410487168513e-05, "loss": 0.5005, "step": 27982 }, { "epoch": 2.82, "learning_rate": 1.1590520833356713e-05, "loss": 0.4613, "step": 27983 }, { "epoch": 2.82, "learning_rate": 1.1588631296695835e-05, "loss": 0.3783, "step": 27984 }, { "epoch": 2.82, "learning_rate": 1.1586741877197904e-05, "loss": 0.7836, "step": 27985 }, { "epoch": 2.82, "learning_rate": 1.1584852574874943e-05, "loss": 0.2508, "step": 27986 }, { "epoch": 2.82, "learning_rate": 1.1582963389738974e-05, "loss": 0.4594, "step": 27987 }, { "epoch": 2.82, "learning_rate": 1.158107432180202e-05, "loss": 0.0992, "step": 27988 }, { "epoch": 2.82, "learning_rate": 1.1579185371076093e-05, "loss": 0.3325, "step": 27989 }, { "epoch": 2.82, "learning_rate": 1.157729653757324e-05, "loss": 0.585, "step": 27990 }, { "epoch": 2.82, "learning_rate": 1.1575407821305457e-05, "loss": 0.5998, "step": 27991 }, { "epoch": 2.82, "learning_rate": 1.1573519222284758e-05, "loss": 0.2711, "step": 27992 }, { "epoch": 2.82, "learning_rate": 1.1571630740523182e-05, "loss": 0.3621, "step": 27993 }, { "epoch": 2.82, "learning_rate": 1.1569742376032749e-05, "loss": 0.1514, "step": 27994 }, { "epoch": 2.82, "learning_rate": 1.1567854128825447e-05, "loss": 0.0749, "step": 27995 }, { "epoch": 2.82, "learning_rate": 1.1565965998913312e-05, "loss": 0.9995, "step": 27996 }, { "epoch": 2.82, "learning_rate": 1.1564077986308363e-05, "loss": 0.2919, "step": 27997 }, { "epoch": 2.82, "learning_rate": 1.1562190091022605e-05, "loss": 0.5561, "step": 27998 }, { "epoch": 2.82, "learning_rate": 1.1560302313068055e-05, "loss": 0.2655, "step": 27999 }, { "epoch": 2.82, "learning_rate": 1.155841465245673e-05, "loss": 0.4811, "step": 28000 }, { "epoch": 2.82, "learning_rate": 1.1556527109200634e-05, "loss": 0.5192, "step": 28001 }, { "epoch": 2.82, "learning_rate": 1.1554639683311788e-05, "loss": 0.2058, "step": 28002 }, { "epoch": 2.82, "learning_rate": 1.1552752374802183e-05, "loss": 0.1821, "step": 28003 }, { "epoch": 2.82, "learning_rate": 1.155086518368387e-05, "loss": 0.0793, "step": 28004 }, { "epoch": 2.82, "learning_rate": 1.1548978109968816e-05, "loss": 0.1213, "step": 28005 }, { "epoch": 2.82, "learning_rate": 1.1547091153669041e-05, "loss": 0.6435, "step": 28006 }, { "epoch": 2.82, "learning_rate": 1.1545204314796566e-05, "loss": 0.1767, "step": 28007 }, { "epoch": 2.82, "learning_rate": 1.1543317593363403e-05, "loss": 0.1253, "step": 28008 }, { "epoch": 2.82, "learning_rate": 1.1541430989381521e-05, "loss": 0.6796, "step": 28009 }, { "epoch": 2.82, "learning_rate": 1.1539544502862963e-05, "loss": 0.3262, "step": 28010 }, { "epoch": 2.82, "learning_rate": 1.1537658133819721e-05, "loss": 0.3793, "step": 28011 }, { "epoch": 2.82, "learning_rate": 1.1535771882263798e-05, "loss": 0.6947, "step": 28012 }, { "epoch": 2.82, "learning_rate": 1.15338857482072e-05, "loss": 0.1672, "step": 28013 }, { "epoch": 2.82, "learning_rate": 1.1531999731661928e-05, "loss": 0.6518, "step": 28014 }, { "epoch": 2.82, "learning_rate": 1.1530113832639984e-05, "loss": 0.2343, "step": 28015 }, { "epoch": 2.82, "learning_rate": 1.1528228051153367e-05, "loss": 0.7181, "step": 28016 }, { "epoch": 2.82, "learning_rate": 1.152634238721408e-05, "loss": 0.2425, "step": 28017 }, { "epoch": 2.82, "learning_rate": 1.1524456840834126e-05, "loss": 0.4496, "step": 28018 }, { "epoch": 2.82, "learning_rate": 1.1522571412025497e-05, "loss": 0.0307, "step": 28019 }, { "epoch": 2.82, "learning_rate": 1.1520686100800197e-05, "loss": 0.0604, "step": 28020 }, { "epoch": 2.82, "learning_rate": 1.1518800907170208e-05, "loss": 0.21, "step": 28021 }, { "epoch": 2.82, "learning_rate": 1.151691583114756e-05, "loss": 0.4264, "step": 28022 }, { "epoch": 2.82, "learning_rate": 1.1515030872744218e-05, "loss": 0.3775, "step": 28023 }, { "epoch": 2.82, "learning_rate": 1.151314603197218e-05, "loss": 0.3741, "step": 28024 }, { "epoch": 2.82, "learning_rate": 1.1511261308843455e-05, "loss": 0.7709, "step": 28025 }, { "epoch": 2.82, "learning_rate": 1.1509376703370042e-05, "loss": 0.0471, "step": 28026 }, { "epoch": 2.82, "learning_rate": 1.15074922155639e-05, "loss": 0.7751, "step": 28027 }, { "epoch": 2.82, "learning_rate": 1.1505607845437052e-05, "loss": 0.3105, "step": 28028 }, { "epoch": 2.82, "learning_rate": 1.1503723593001482e-05, "loss": 0.6381, "step": 28029 }, { "epoch": 2.82, "learning_rate": 1.150183945826918e-05, "loss": 0.3256, "step": 28030 }, { "epoch": 2.82, "learning_rate": 1.149995544125213e-05, "loss": 0.1589, "step": 28031 }, { "epoch": 2.82, "learning_rate": 1.1498071541962327e-05, "loss": 0.1727, "step": 28032 }, { "epoch": 2.82, "learning_rate": 1.1496187760411761e-05, "loss": 0.1306, "step": 28033 }, { "epoch": 2.82, "learning_rate": 1.1494304096612414e-05, "loss": 0.4327, "step": 28034 }, { "epoch": 2.82, "learning_rate": 1.1492420550576268e-05, "loss": 0.8426, "step": 28035 }, { "epoch": 2.82, "learning_rate": 1.1490537122315336e-05, "loss": 0.2428, "step": 28036 }, { "epoch": 2.82, "learning_rate": 1.1488653811841575e-05, "loss": 0.155, "step": 28037 }, { "epoch": 2.82, "learning_rate": 1.1486770619166969e-05, "loss": 0.2183, "step": 28038 }, { "epoch": 2.82, "learning_rate": 1.148488754430352e-05, "loss": 0.635, "step": 28039 }, { "epoch": 2.82, "learning_rate": 1.1483004587263217e-05, "loss": 0.6346, "step": 28040 }, { "epoch": 2.82, "learning_rate": 1.1481121748058003e-05, "loss": 0.5691, "step": 28041 }, { "epoch": 2.82, "learning_rate": 1.1479239026699902e-05, "loss": 0.4688, "step": 28042 }, { "epoch": 2.82, "learning_rate": 1.1477356423200872e-05, "loss": 0.2368, "step": 28043 }, { "epoch": 2.82, "learning_rate": 1.1475473937572902e-05, "loss": 0.1518, "step": 28044 }, { "epoch": 2.82, "learning_rate": 1.147359156982797e-05, "loss": 0.1794, "step": 28045 }, { "epoch": 2.82, "learning_rate": 1.1471709319978053e-05, "loss": 0.8104, "step": 28046 }, { "epoch": 2.82, "learning_rate": 1.1469827188035128e-05, "loss": 0.4271, "step": 28047 }, { "epoch": 2.82, "learning_rate": 1.1467945174011175e-05, "loss": 0.2707, "step": 28048 }, { "epoch": 2.82, "learning_rate": 1.1466063277918165e-05, "loss": 1.0495, "step": 28049 }, { "epoch": 2.82, "learning_rate": 1.1464181499768081e-05, "loss": 0.1295, "step": 28050 }, { "epoch": 2.82, "learning_rate": 1.1462299839572894e-05, "loss": 0.6948, "step": 28051 }, { "epoch": 2.82, "learning_rate": 1.1460418297344579e-05, "loss": 0.7622, "step": 28052 }, { "epoch": 2.82, "learning_rate": 1.1458536873095101e-05, "loss": 0.1754, "step": 28053 }, { "epoch": 2.82, "learning_rate": 1.1456655566836458e-05, "loss": 0.5247, "step": 28054 }, { "epoch": 2.82, "learning_rate": 1.1454774378580596e-05, "loss": 0.1917, "step": 28055 }, { "epoch": 2.82, "learning_rate": 1.1452893308339486e-05, "loss": 0.59, "step": 28056 }, { "epoch": 2.82, "learning_rate": 1.1451012356125115e-05, "loss": 0.0793, "step": 28057 }, { "epoch": 2.82, "learning_rate": 1.1449131521949446e-05, "loss": 0.5361, "step": 28058 }, { "epoch": 2.82, "learning_rate": 1.1447250805824443e-05, "loss": 0.4576, "step": 28059 }, { "epoch": 2.82, "learning_rate": 1.1445370207762085e-05, "loss": 0.3736, "step": 28060 }, { "epoch": 2.82, "learning_rate": 1.1443489727774328e-05, "loss": 0.3117, "step": 28061 }, { "epoch": 2.82, "learning_rate": 1.1441609365873144e-05, "loss": 0.1272, "step": 28062 }, { "epoch": 2.82, "learning_rate": 1.14397291220705e-05, "loss": 0.0493, "step": 28063 }, { "epoch": 2.82, "learning_rate": 1.1437848996378356e-05, "loss": 0.863, "step": 28064 }, { "epoch": 2.82, "learning_rate": 1.1435968988808681e-05, "loss": 0.4174, "step": 28065 }, { "epoch": 2.82, "learning_rate": 1.1434089099373439e-05, "loss": 0.1969, "step": 28066 }, { "epoch": 2.82, "learning_rate": 1.143220932808458e-05, "loss": 0.3949, "step": 28067 }, { "epoch": 2.82, "learning_rate": 1.1430329674954097e-05, "loss": 0.6114, "step": 28068 }, { "epoch": 2.82, "learning_rate": 1.142845013999392e-05, "loss": 0.1852, "step": 28069 }, { "epoch": 2.82, "learning_rate": 1.1426570723216012e-05, "loss": 0.5591, "step": 28070 }, { "epoch": 2.82, "learning_rate": 1.1424691424632353e-05, "loss": 0.362, "step": 28071 }, { "epoch": 2.82, "learning_rate": 1.1422812244254903e-05, "loss": 0.4524, "step": 28072 }, { "epoch": 2.83, "learning_rate": 1.1420933182095585e-05, "loss": 0.1141, "step": 28073 }, { "epoch": 2.83, "learning_rate": 1.1419054238166393e-05, "loss": 0.5952, "step": 28074 }, { "epoch": 2.83, "learning_rate": 1.141717541247927e-05, "loss": 0.6953, "step": 28075 }, { "epoch": 2.83, "learning_rate": 1.1415296705046174e-05, "loss": 0.3606, "step": 28076 }, { "epoch": 2.83, "learning_rate": 1.1413418115879061e-05, "loss": 0.4602, "step": 28077 }, { "epoch": 2.83, "learning_rate": 1.1411539644989883e-05, "loss": 0.6982, "step": 28078 }, { "epoch": 2.83, "learning_rate": 1.1409661292390597e-05, "loss": 0.0912, "step": 28079 }, { "epoch": 2.83, "learning_rate": 1.1407783058093151e-05, "loss": 0.1688, "step": 28080 }, { "epoch": 2.83, "learning_rate": 1.1405904942109503e-05, "loss": 0.3179, "step": 28081 }, { "epoch": 2.83, "learning_rate": 1.1404026944451603e-05, "loss": 0.1425, "step": 28082 }, { "epoch": 2.83, "learning_rate": 1.1402149065131402e-05, "loss": 0.2541, "step": 28083 }, { "epoch": 2.83, "learning_rate": 1.1400271304160838e-05, "loss": 0.2233, "step": 28084 }, { "epoch": 2.83, "learning_rate": 1.139839366155188e-05, "loss": 0.4039, "step": 28085 }, { "epoch": 2.83, "learning_rate": 1.139651613731648e-05, "loss": 0.4771, "step": 28086 }, { "epoch": 2.83, "learning_rate": 1.1394638731466552e-05, "loss": 0.1028, "step": 28087 }, { "epoch": 2.83, "learning_rate": 1.1392761444014078e-05, "loss": 0.0755, "step": 28088 }, { "epoch": 2.83, "learning_rate": 1.1390884274970992e-05, "loss": 0.5216, "step": 28089 }, { "epoch": 2.83, "learning_rate": 1.1389007224349237e-05, "loss": 0.3799, "step": 28090 }, { "epoch": 2.83, "learning_rate": 1.1387130292160763e-05, "loss": 0.7247, "step": 28091 }, { "epoch": 2.83, "learning_rate": 1.1385253478417508e-05, "loss": 0.742, "step": 28092 }, { "epoch": 2.83, "learning_rate": 1.1383376783131421e-05, "loss": 0.273, "step": 28093 }, { "epoch": 2.83, "learning_rate": 1.1381500206314442e-05, "loss": 0.0639, "step": 28094 }, { "epoch": 2.83, "learning_rate": 1.137962374797851e-05, "loss": 0.3046, "step": 28095 }, { "epoch": 2.83, "learning_rate": 1.1377747408135571e-05, "loss": 0.2873, "step": 28096 }, { "epoch": 2.83, "learning_rate": 1.1375871186797561e-05, "loss": 0.5146, "step": 28097 }, { "epoch": 2.83, "learning_rate": 1.1373995083976425e-05, "loss": 0.0641, "step": 28098 }, { "epoch": 2.83, "learning_rate": 1.1372119099684088e-05, "loss": 0.5863, "step": 28099 }, { "epoch": 2.83, "learning_rate": 1.1370243233932515e-05, "loss": 0.039, "step": 28100 }, { "epoch": 2.83, "learning_rate": 1.1368367486733621e-05, "loss": 0.3674, "step": 28101 }, { "epoch": 2.83, "learning_rate": 1.1366491858099337e-05, "loss": 0.1217, "step": 28102 }, { "epoch": 2.83, "learning_rate": 1.1364616348041622e-05, "loss": 0.4491, "step": 28103 }, { "epoch": 2.83, "learning_rate": 1.1362740956572407e-05, "loss": 0.3045, "step": 28104 }, { "epoch": 2.83, "learning_rate": 1.13608656837036e-05, "loss": 0.1213, "step": 28105 }, { "epoch": 2.83, "learning_rate": 1.1358990529447163e-05, "loss": 0.1832, "step": 28106 }, { "epoch": 2.83, "learning_rate": 1.1357115493815017e-05, "loss": 0.5745, "step": 28107 }, { "epoch": 2.83, "learning_rate": 1.1355240576819099e-05, "loss": 0.2823, "step": 28108 }, { "epoch": 2.83, "learning_rate": 1.1353365778471334e-05, "loss": 0.1471, "step": 28109 }, { "epoch": 2.83, "learning_rate": 1.1351491098783659e-05, "loss": 0.1727, "step": 28110 }, { "epoch": 2.83, "learning_rate": 1.1349616537767995e-05, "loss": 0.1937, "step": 28111 }, { "epoch": 2.83, "learning_rate": 1.1347742095436281e-05, "loss": 0.4293, "step": 28112 }, { "epoch": 2.83, "learning_rate": 1.134586777180043e-05, "loss": 0.1139, "step": 28113 }, { "epoch": 2.83, "learning_rate": 1.13439935668724e-05, "loss": 0.1978, "step": 28114 }, { "epoch": 2.83, "learning_rate": 1.1342119480664085e-05, "loss": 0.6109, "step": 28115 }, { "epoch": 2.83, "learning_rate": 1.1340245513187417e-05, "loss": 0.4972, "step": 28116 }, { "epoch": 2.83, "learning_rate": 1.1338371664454337e-05, "loss": 0.3925, "step": 28117 }, { "epoch": 2.83, "learning_rate": 1.1336497934476772e-05, "loss": 0.2675, "step": 28118 }, { "epoch": 2.83, "learning_rate": 1.1334624323266613e-05, "loss": 0.2979, "step": 28119 }, { "epoch": 2.83, "learning_rate": 1.1332750830835813e-05, "loss": 0.3079, "step": 28120 }, { "epoch": 2.83, "learning_rate": 1.1330877457196286e-05, "loss": 0.7197, "step": 28121 }, { "epoch": 2.83, "learning_rate": 1.1329004202359954e-05, "loss": 0.3404, "step": 28122 }, { "epoch": 2.83, "learning_rate": 1.1327131066338735e-05, "loss": 0.3639, "step": 28123 }, { "epoch": 2.83, "learning_rate": 1.1325258049144552e-05, "loss": 0.3045, "step": 28124 }, { "epoch": 2.83, "learning_rate": 1.132338515078932e-05, "loss": 0.5457, "step": 28125 }, { "epoch": 2.83, "learning_rate": 1.1321512371284962e-05, "loss": 0.5239, "step": 28126 }, { "epoch": 2.83, "learning_rate": 1.1319639710643393e-05, "loss": 0.1059, "step": 28127 }, { "epoch": 2.83, "learning_rate": 1.1317767168876531e-05, "loss": 0.6718, "step": 28128 }, { "epoch": 2.83, "learning_rate": 1.131589474599629e-05, "loss": 0.0542, "step": 28129 }, { "epoch": 2.83, "learning_rate": 1.131402244201459e-05, "loss": 0.1259, "step": 28130 }, { "epoch": 2.83, "learning_rate": 1.1312150256943328e-05, "loss": 0.1733, "step": 28131 }, { "epoch": 2.83, "learning_rate": 1.1310278190794458e-05, "loss": 0.3912, "step": 28132 }, { "epoch": 2.83, "learning_rate": 1.1308406243579853e-05, "loss": 0.2965, "step": 28133 }, { "epoch": 2.83, "learning_rate": 1.1306534415311432e-05, "loss": 0.7575, "step": 28134 }, { "epoch": 2.83, "learning_rate": 1.1304662706001126e-05, "loss": 0.3615, "step": 28135 }, { "epoch": 2.83, "learning_rate": 1.1302791115660843e-05, "loss": 0.6139, "step": 28136 }, { "epoch": 2.83, "learning_rate": 1.1300919644302465e-05, "loss": 0.2108, "step": 28137 }, { "epoch": 2.83, "learning_rate": 1.129904829193793e-05, "loss": 0.0454, "step": 28138 }, { "epoch": 2.83, "learning_rate": 1.1297177058579136e-05, "loss": 0.3633, "step": 28139 }, { "epoch": 2.83, "learning_rate": 1.1295305944237996e-05, "loss": 0.3712, "step": 28140 }, { "epoch": 2.83, "learning_rate": 1.129343494892641e-05, "loss": 0.3989, "step": 28141 }, { "epoch": 2.83, "learning_rate": 1.1291564072656293e-05, "loss": 0.2902, "step": 28142 }, { "epoch": 2.83, "learning_rate": 1.1289693315439539e-05, "loss": 0.2158, "step": 28143 }, { "epoch": 2.83, "learning_rate": 1.1287822677288062e-05, "loss": 0.3606, "step": 28144 }, { "epoch": 2.83, "learning_rate": 1.1285952158213754e-05, "loss": 0.1529, "step": 28145 }, { "epoch": 2.83, "learning_rate": 1.1284081758228543e-05, "loss": 1.013, "step": 28146 }, { "epoch": 2.83, "learning_rate": 1.128221147734431e-05, "loss": 0.672, "step": 28147 }, { "epoch": 2.83, "learning_rate": 1.1280341315572951e-05, "loss": 0.3667, "step": 28148 }, { "epoch": 2.83, "learning_rate": 1.127847127292639e-05, "loss": 0.4382, "step": 28149 }, { "epoch": 2.83, "learning_rate": 1.1276601349416523e-05, "loss": 0.0657, "step": 28150 }, { "epoch": 2.83, "learning_rate": 1.1274731545055224e-05, "loss": 0.4313, "step": 28151 }, { "epoch": 2.83, "learning_rate": 1.1272861859854419e-05, "loss": 0.5549, "step": 28152 }, { "epoch": 2.83, "learning_rate": 1.1270992293826e-05, "loss": 0.3984, "step": 28153 }, { "epoch": 2.83, "learning_rate": 1.1269122846981858e-05, "loss": 0.2872, "step": 28154 }, { "epoch": 2.83, "learning_rate": 1.1267253519333894e-05, "loss": 0.4176, "step": 28155 }, { "epoch": 2.83, "learning_rate": 1.1265384310894e-05, "loss": 0.3728, "step": 28156 }, { "epoch": 2.83, "learning_rate": 1.1263515221674076e-05, "loss": 0.3367, "step": 28157 }, { "epoch": 2.83, "learning_rate": 1.1261646251686011e-05, "loss": 0.2864, "step": 28158 }, { "epoch": 2.83, "learning_rate": 1.1259777400941703e-05, "loss": 0.4277, "step": 28159 }, { "epoch": 2.83, "learning_rate": 1.125790866945304e-05, "loss": 0.2264, "step": 28160 }, { "epoch": 2.83, "learning_rate": 1.125604005723192e-05, "loss": 0.1223, "step": 28161 }, { "epoch": 2.83, "learning_rate": 1.1254171564290228e-05, "loss": 0.5654, "step": 28162 }, { "epoch": 2.83, "learning_rate": 1.1252303190639848e-05, "loss": 0.6969, "step": 28163 }, { "epoch": 2.83, "learning_rate": 1.1250434936292698e-05, "loss": 0.6513, "step": 28164 }, { "epoch": 2.83, "learning_rate": 1.1248566801260637e-05, "loss": 0.2967, "step": 28165 }, { "epoch": 2.83, "learning_rate": 1.1246698785555553e-05, "loss": 0.3642, "step": 28166 }, { "epoch": 2.83, "learning_rate": 1.1244830889189355e-05, "loss": 0.0876, "step": 28167 }, { "epoch": 2.83, "learning_rate": 1.1242963112173918e-05, "loss": 0.1513, "step": 28168 }, { "epoch": 2.83, "learning_rate": 1.1241095454521128e-05, "loss": 0.7466, "step": 28169 }, { "epoch": 2.83, "learning_rate": 1.123922791624287e-05, "loss": 0.2323, "step": 28170 }, { "epoch": 2.83, "learning_rate": 1.1237360497351029e-05, "loss": 0.4355, "step": 28171 }, { "epoch": 2.84, "learning_rate": 1.123549319785749e-05, "loss": 0.6713, "step": 28172 }, { "epoch": 2.84, "learning_rate": 1.123362601777413e-05, "loss": 0.6287, "step": 28173 }, { "epoch": 2.84, "learning_rate": 1.1231758957112839e-05, "loss": 0.1301, "step": 28174 }, { "epoch": 2.84, "learning_rate": 1.1229892015885492e-05, "loss": 0.0816, "step": 28175 }, { "epoch": 2.84, "learning_rate": 1.1228025194103973e-05, "loss": 0.37, "step": 28176 }, { "epoch": 2.84, "learning_rate": 1.1226158491780146e-05, "loss": 0.2867, "step": 28177 }, { "epoch": 2.84, "learning_rate": 1.1224291908925932e-05, "loss": 0.2584, "step": 28178 }, { "epoch": 2.84, "learning_rate": 1.1222425445553164e-05, "loss": 0.329, "step": 28179 }, { "epoch": 2.84, "learning_rate": 1.1220559101673732e-05, "loss": 0.4417, "step": 28180 }, { "epoch": 2.84, "learning_rate": 1.1218692877299526e-05, "loss": 0.2833, "step": 28181 }, { "epoch": 2.84, "learning_rate": 1.1216826772442425e-05, "loss": 0.209, "step": 28182 }, { "epoch": 2.84, "learning_rate": 1.1214960787114273e-05, "loss": 0.2665, "step": 28183 }, { "epoch": 2.84, "learning_rate": 1.1213094921326972e-05, "loss": 0.2015, "step": 28184 }, { "epoch": 2.84, "learning_rate": 1.1211229175092388e-05, "loss": 0.3032, "step": 28185 }, { "epoch": 2.84, "learning_rate": 1.1209363548422398e-05, "loss": 0.7641, "step": 28186 }, { "epoch": 2.84, "learning_rate": 1.1207498041328867e-05, "loss": 0.2879, "step": 28187 }, { "epoch": 2.84, "learning_rate": 1.1205632653823667e-05, "loss": 0.6869, "step": 28188 }, { "epoch": 2.84, "learning_rate": 1.1203767385918676e-05, "loss": 0.1193, "step": 28189 }, { "epoch": 2.84, "learning_rate": 1.1201902237625756e-05, "loss": 0.158, "step": 28190 }, { "epoch": 2.84, "learning_rate": 1.1200037208956781e-05, "loss": 0.0832, "step": 28191 }, { "epoch": 2.84, "learning_rate": 1.1198172299923615e-05, "loss": 0.2259, "step": 28192 }, { "epoch": 2.84, "learning_rate": 1.119630751053813e-05, "loss": 0.179, "step": 28193 }, { "epoch": 2.84, "learning_rate": 1.1194442840812177e-05, "loss": 0.2809, "step": 28194 }, { "epoch": 2.84, "learning_rate": 1.119257829075765e-05, "loss": 1.5545, "step": 28195 }, { "epoch": 2.84, "learning_rate": 1.119071386038641e-05, "loss": 0.217, "step": 28196 }, { "epoch": 2.84, "learning_rate": 1.1188849549710288e-05, "loss": 0.4103, "step": 28197 }, { "epoch": 2.84, "learning_rate": 1.1186985358741185e-05, "loss": 0.3828, "step": 28198 }, { "epoch": 2.84, "learning_rate": 1.1185121287490948e-05, "loss": 0.9581, "step": 28199 }, { "epoch": 2.84, "learning_rate": 1.1183257335971444e-05, "loss": 0.3999, "step": 28200 }, { "epoch": 2.84, "learning_rate": 1.1181393504194527e-05, "loss": 0.1819, "step": 28201 }, { "epoch": 2.84, "learning_rate": 1.1179529792172068e-05, "loss": 0.0537, "step": 28202 }, { "epoch": 2.84, "learning_rate": 1.1177666199915918e-05, "loss": 0.1992, "step": 28203 }, { "epoch": 2.84, "learning_rate": 1.1175802727437942e-05, "loss": 0.5459, "step": 28204 }, { "epoch": 2.84, "learning_rate": 1.1173939374749998e-05, "loss": 0.3931, "step": 28205 }, { "epoch": 2.84, "learning_rate": 1.117207614186394e-05, "loss": 0.899, "step": 28206 }, { "epoch": 2.84, "learning_rate": 1.1170213028791626e-05, "loss": 0.7247, "step": 28207 }, { "epoch": 2.84, "learning_rate": 1.1168350035544916e-05, "loss": 0.5463, "step": 28208 }, { "epoch": 2.84, "learning_rate": 1.1166487162135648e-05, "loss": 0.622, "step": 28209 }, { "epoch": 2.84, "learning_rate": 1.1164624408575714e-05, "loss": 0.2841, "step": 28210 }, { "epoch": 2.84, "learning_rate": 1.1162761774876935e-05, "loss": 0.6754, "step": 28211 }, { "epoch": 2.84, "learning_rate": 1.1160899261051164e-05, "loss": 0.2223, "step": 28212 }, { "epoch": 2.84, "learning_rate": 1.1159036867110275e-05, "loss": 0.4625, "step": 28213 }, { "epoch": 2.84, "learning_rate": 1.1157174593066114e-05, "loss": 0.3273, "step": 28214 }, { "epoch": 2.84, "learning_rate": 1.1155312438930509e-05, "loss": 0.5333, "step": 28215 }, { "epoch": 2.84, "learning_rate": 1.1153450404715335e-05, "loss": 0.4951, "step": 28216 }, { "epoch": 2.84, "learning_rate": 1.1151588490432433e-05, "loss": 0.7573, "step": 28217 }, { "epoch": 2.84, "learning_rate": 1.1149726696093652e-05, "loss": 0.249, "step": 28218 }, { "epoch": 2.84, "learning_rate": 1.114786502171084e-05, "loss": 0.0746, "step": 28219 }, { "epoch": 2.84, "learning_rate": 1.1146003467295832e-05, "loss": 0.4594, "step": 28220 }, { "epoch": 2.84, "learning_rate": 1.1144142032860505e-05, "loss": 0.5128, "step": 28221 }, { "epoch": 2.84, "learning_rate": 1.1142280718416678e-05, "loss": 0.4221, "step": 28222 }, { "epoch": 2.84, "learning_rate": 1.1140419523976187e-05, "loss": 0.7487, "step": 28223 }, { "epoch": 2.84, "learning_rate": 1.1138558449550918e-05, "loss": 0.448, "step": 28224 }, { "epoch": 2.84, "learning_rate": 1.1136697495152674e-05, "loss": 0.8366, "step": 28225 }, { "epoch": 2.84, "learning_rate": 1.1134836660793303e-05, "loss": 0.3178, "step": 28226 }, { "epoch": 2.84, "learning_rate": 1.1132975946484664e-05, "loss": 0.3896, "step": 28227 }, { "epoch": 2.84, "learning_rate": 1.1131115352238597e-05, "loss": 0.2108, "step": 28228 }, { "epoch": 2.84, "learning_rate": 1.1129254878066916e-05, "loss": 0.2391, "step": 28229 }, { "epoch": 2.84, "learning_rate": 1.1127394523981488e-05, "loss": 0.4493, "step": 28230 }, { "epoch": 2.84, "learning_rate": 1.1125534289994142e-05, "loss": 0.0132, "step": 28231 }, { "epoch": 2.84, "learning_rate": 1.1123674176116714e-05, "loss": 0.6505, "step": 28232 }, { "epoch": 2.84, "learning_rate": 1.1121814182361046e-05, "loss": 0.5355, "step": 28233 }, { "epoch": 2.84, "learning_rate": 1.1119954308738967e-05, "loss": 0.118, "step": 28234 }, { "epoch": 2.84, "learning_rate": 1.1118094555262318e-05, "loss": 0.3146, "step": 28235 }, { "epoch": 2.84, "learning_rate": 1.111623492194293e-05, "loss": 0.4127, "step": 28236 }, { "epoch": 2.84, "learning_rate": 1.1114375408792644e-05, "loss": 0.5061, "step": 28237 }, { "epoch": 2.84, "learning_rate": 1.1112516015823286e-05, "loss": 0.2133, "step": 28238 }, { "epoch": 2.84, "learning_rate": 1.1110656743046692e-05, "loss": 0.6995, "step": 28239 }, { "epoch": 2.84, "learning_rate": 1.1108797590474688e-05, "loss": 0.4536, "step": 28240 }, { "epoch": 2.84, "learning_rate": 1.1106938558119104e-05, "loss": 0.3993, "step": 28241 }, { "epoch": 2.84, "learning_rate": 1.1105079645991795e-05, "loss": 0.5647, "step": 28242 }, { "epoch": 2.84, "learning_rate": 1.1103220854104561e-05, "loss": 0.3671, "step": 28243 }, { "epoch": 2.84, "learning_rate": 1.110136218246923e-05, "loss": 0.1265, "step": 28244 }, { "epoch": 2.84, "learning_rate": 1.109950363109765e-05, "loss": 0.3597, "step": 28245 }, { "epoch": 2.84, "learning_rate": 1.109764520000165e-05, "loss": 0.4958, "step": 28246 }, { "epoch": 2.84, "learning_rate": 1.1095786889193026e-05, "loss": 0.3928, "step": 28247 }, { "epoch": 2.84, "learning_rate": 1.1093928698683633e-05, "loss": 0.5009, "step": 28248 }, { "epoch": 2.84, "learning_rate": 1.1092070628485282e-05, "loss": 0.1246, "step": 28249 }, { "epoch": 2.84, "learning_rate": 1.1090212678609804e-05, "loss": 0.36, "step": 28250 }, { "epoch": 2.84, "learning_rate": 1.1088354849069018e-05, "loss": 0.2672, "step": 28251 }, { "epoch": 2.84, "learning_rate": 1.1086497139874735e-05, "loss": 0.3365, "step": 28252 }, { "epoch": 2.84, "learning_rate": 1.1084639551038812e-05, "loss": 0.3114, "step": 28253 }, { "epoch": 2.84, "learning_rate": 1.1082782082573038e-05, "loss": 0.1829, "step": 28254 }, { "epoch": 2.84, "learning_rate": 1.1080924734489228e-05, "loss": 0.1553, "step": 28255 }, { "epoch": 2.84, "learning_rate": 1.107906750679924e-05, "loss": 0.9798, "step": 28256 }, { "epoch": 2.84, "learning_rate": 1.1077210399514857e-05, "loss": 0.5469, "step": 28257 }, { "epoch": 2.84, "learning_rate": 1.10753534126479e-05, "loss": 0.5821, "step": 28258 }, { "epoch": 2.84, "learning_rate": 1.1073496546210203e-05, "loss": 0.2312, "step": 28259 }, { "epoch": 2.84, "learning_rate": 1.1071639800213581e-05, "loss": 0.388, "step": 28260 }, { "epoch": 2.84, "learning_rate": 1.1069783174669826e-05, "loss": 0.8445, "step": 28261 }, { "epoch": 2.84, "learning_rate": 1.1067926669590782e-05, "loss": 0.2091, "step": 28262 }, { "epoch": 2.84, "learning_rate": 1.1066070284988247e-05, "loss": 0.3591, "step": 28263 }, { "epoch": 2.84, "learning_rate": 1.1064214020874038e-05, "loss": 0.8862, "step": 28264 }, { "epoch": 2.84, "learning_rate": 1.1062357877259967e-05, "loss": 0.5989, "step": 28265 }, { "epoch": 2.84, "learning_rate": 1.106050185415785e-05, "loss": 0.2656, "step": 28266 }, { "epoch": 2.84, "learning_rate": 1.105864595157949e-05, "loss": 0.0573, "step": 28267 }, { "epoch": 2.84, "learning_rate": 1.1056790169536704e-05, "loss": 0.659, "step": 28268 }, { "epoch": 2.84, "learning_rate": 1.1054934508041296e-05, "loss": 0.3881, "step": 28269 }, { "epoch": 2.85, "learning_rate": 1.1053078967105081e-05, "loss": 0.3715, "step": 28270 }, { "epoch": 2.85, "learning_rate": 1.1051223546739862e-05, "loss": 0.3058, "step": 28271 }, { "epoch": 2.85, "learning_rate": 1.1049368246957447e-05, "loss": 0.3224, "step": 28272 }, { "epoch": 2.85, "learning_rate": 1.1047513067769635e-05, "loss": 0.2364, "step": 28273 }, { "epoch": 2.85, "learning_rate": 1.104565800918826e-05, "loss": 0.16, "step": 28274 }, { "epoch": 2.85, "learning_rate": 1.1043803071225085e-05, "loss": 0.371, "step": 28275 }, { "epoch": 2.85, "learning_rate": 1.1041948253891947e-05, "loss": 0.9517, "step": 28276 }, { "epoch": 2.85, "learning_rate": 1.1040093557200636e-05, "loss": 0.0576, "step": 28277 }, { "epoch": 2.85, "learning_rate": 1.1038238981162962e-05, "loss": 0.3631, "step": 28278 }, { "epoch": 2.85, "learning_rate": 1.103638452579072e-05, "loss": 0.2724, "step": 28279 }, { "epoch": 2.85, "learning_rate": 1.1034530191095711e-05, "loss": 0.1474, "step": 28280 }, { "epoch": 2.85, "learning_rate": 1.1032675977089737e-05, "loss": 0.307, "step": 28281 }, { "epoch": 2.85, "learning_rate": 1.1030821883784598e-05, "loss": 0.3593, "step": 28282 }, { "epoch": 2.85, "learning_rate": 1.1028967911192093e-05, "loss": 0.3788, "step": 28283 }, { "epoch": 2.85, "learning_rate": 1.1027114059324007e-05, "loss": 0.6026, "step": 28284 }, { "epoch": 2.85, "learning_rate": 1.1025260328192175e-05, "loss": 0.545, "step": 28285 }, { "epoch": 2.85, "learning_rate": 1.1023406717808354e-05, "loss": 0.2868, "step": 28286 }, { "epoch": 2.85, "learning_rate": 1.1021553228184343e-05, "loss": 0.5769, "step": 28287 }, { "epoch": 2.85, "learning_rate": 1.1019699859331967e-05, "loss": 0.3095, "step": 28288 }, { "epoch": 2.85, "learning_rate": 1.1017846611262994e-05, "loss": 0.4965, "step": 28289 }, { "epoch": 2.85, "learning_rate": 1.1015993483989212e-05, "loss": 0.2358, "step": 28290 }, { "epoch": 2.85, "learning_rate": 1.1014140477522433e-05, "loss": 0.1648, "step": 28291 }, { "epoch": 2.85, "learning_rate": 1.1012287591874454e-05, "loss": 0.6448, "step": 28292 }, { "epoch": 2.85, "learning_rate": 1.1010434827057034e-05, "loss": 0.9039, "step": 28293 }, { "epoch": 2.85, "learning_rate": 1.1008582183081993e-05, "loss": 0.3674, "step": 28294 }, { "epoch": 2.85, "learning_rate": 1.1006729659961108e-05, "loss": 0.3737, "step": 28295 }, { "epoch": 2.85, "learning_rate": 1.1004877257706172e-05, "loss": 0.7327, "step": 28296 }, { "epoch": 2.85, "learning_rate": 1.1003024976328973e-05, "loss": 0.1945, "step": 28297 }, { "epoch": 2.85, "learning_rate": 1.1001172815841292e-05, "loss": 0.3551, "step": 28298 }, { "epoch": 2.85, "learning_rate": 1.0999320776254923e-05, "loss": 0.969, "step": 28299 }, { "epoch": 2.85, "learning_rate": 1.0997468857581643e-05, "loss": 0.2568, "step": 28300 }, { "epoch": 2.85, "learning_rate": 1.0995617059833239e-05, "loss": 0.7104, "step": 28301 }, { "epoch": 2.85, "learning_rate": 1.0993765383021513e-05, "loss": 0.4542, "step": 28302 }, { "epoch": 2.85, "learning_rate": 1.0991913827158227e-05, "loss": 0.5087, "step": 28303 }, { "epoch": 2.85, "learning_rate": 1.0990062392255162e-05, "loss": 0.4627, "step": 28304 }, { "epoch": 2.85, "learning_rate": 1.0988211078324114e-05, "loss": 0.1293, "step": 28305 }, { "epoch": 2.85, "learning_rate": 1.0986359885376868e-05, "loss": 0.2754, "step": 28306 }, { "epoch": 2.85, "learning_rate": 1.0984508813425178e-05, "loss": 0.2303, "step": 28307 }, { "epoch": 2.85, "learning_rate": 1.0982657862480851e-05, "loss": 0.3531, "step": 28308 }, { "epoch": 2.85, "learning_rate": 1.098080703255565e-05, "loss": 0.1614, "step": 28309 }, { "epoch": 2.85, "learning_rate": 1.0978956323661361e-05, "loss": 0.2566, "step": 28310 }, { "epoch": 2.85, "learning_rate": 1.0977105735809758e-05, "loss": 0.0156, "step": 28311 }, { "epoch": 2.85, "learning_rate": 1.0975255269012616e-05, "loss": 0.5605, "step": 28312 }, { "epoch": 2.85, "learning_rate": 1.0973404923281713e-05, "loss": 0.3035, "step": 28313 }, { "epoch": 2.85, "learning_rate": 1.097155469862882e-05, "loss": 0.1934, "step": 28314 }, { "epoch": 2.85, "learning_rate": 1.0969704595065718e-05, "loss": 0.2317, "step": 28315 }, { "epoch": 2.85, "learning_rate": 1.0967854612604166e-05, "loss": 0.5119, "step": 28316 }, { "epoch": 2.85, "learning_rate": 1.0966004751255966e-05, "loss": 0.071, "step": 28317 }, { "epoch": 2.85, "learning_rate": 1.0964155011032861e-05, "loss": 0.6133, "step": 28318 }, { "epoch": 2.85, "learning_rate": 1.096230539194662e-05, "loss": 0.6927, "step": 28319 }, { "epoch": 2.85, "learning_rate": 1.0960455894009049e-05, "loss": 0.181, "step": 28320 }, { "epoch": 2.85, "learning_rate": 1.095860651723188e-05, "loss": 0.0946, "step": 28321 }, { "epoch": 2.85, "learning_rate": 1.0956757261626887e-05, "loss": 0.5702, "step": 28322 }, { "epoch": 2.85, "learning_rate": 1.0954908127205857e-05, "loss": 0.4546, "step": 28323 }, { "epoch": 2.85, "learning_rate": 1.0953059113980556e-05, "loss": 0.4262, "step": 28324 }, { "epoch": 2.85, "learning_rate": 1.0951210221962724e-05, "loss": 0.3935, "step": 28325 }, { "epoch": 2.85, "learning_rate": 1.094936145116415e-05, "loss": 0.1523, "step": 28326 }, { "epoch": 2.85, "learning_rate": 1.0947512801596593e-05, "loss": 1.9169, "step": 28327 }, { "epoch": 2.85, "learning_rate": 1.0945664273271817e-05, "loss": 0.4327, "step": 28328 }, { "epoch": 2.85, "learning_rate": 1.0943815866201585e-05, "loss": 0.305, "step": 28329 }, { "epoch": 2.85, "learning_rate": 1.0941967580397647e-05, "loss": 0.1603, "step": 28330 }, { "epoch": 2.85, "learning_rate": 1.0940119415871802e-05, "loss": 0.2527, "step": 28331 }, { "epoch": 2.85, "learning_rate": 1.0938271372635774e-05, "loss": 0.2778, "step": 28332 }, { "epoch": 2.85, "learning_rate": 1.0936423450701326e-05, "loss": 0.1398, "step": 28333 }, { "epoch": 2.85, "learning_rate": 1.0934575650080245e-05, "loss": 0.2248, "step": 28334 }, { "epoch": 2.85, "learning_rate": 1.0932727970784263e-05, "loss": 0.2336, "step": 28335 }, { "epoch": 2.85, "learning_rate": 1.0930880412825137e-05, "loss": 0.4755, "step": 28336 }, { "epoch": 2.85, "learning_rate": 1.092903297621464e-05, "loss": 0.6419, "step": 28337 }, { "epoch": 2.85, "learning_rate": 1.0927185660964539e-05, "loss": 0.1294, "step": 28338 }, { "epoch": 2.85, "learning_rate": 1.0925338467086548e-05, "loss": 0.7113, "step": 28339 }, { "epoch": 2.85, "learning_rate": 1.092349139459246e-05, "loss": 0.8086, "step": 28340 }, { "epoch": 2.85, "learning_rate": 1.0921644443494013e-05, "loss": 0.4109, "step": 28341 }, { "epoch": 2.85, "learning_rate": 1.0919797613802962e-05, "loss": 0.7132, "step": 28342 }, { "epoch": 2.85, "learning_rate": 1.0917950905531063e-05, "loss": 0.1441, "step": 28343 }, { "epoch": 2.85, "learning_rate": 1.0916104318690063e-05, "loss": 0.0835, "step": 28344 }, { "epoch": 2.85, "learning_rate": 1.091425785329172e-05, "loss": 0.3706, "step": 28345 }, { "epoch": 2.85, "learning_rate": 1.0912411509347775e-05, "loss": 0.2917, "step": 28346 }, { "epoch": 2.85, "learning_rate": 1.091056528686998e-05, "loss": 0.1118, "step": 28347 }, { "epoch": 2.85, "learning_rate": 1.0908719185870082e-05, "loss": 0.7093, "step": 28348 }, { "epoch": 2.85, "learning_rate": 1.0906873206359847e-05, "loss": 0.1606, "step": 28349 }, { "epoch": 2.85, "learning_rate": 1.0905027348351002e-05, "loss": 0.299, "step": 28350 }, { "epoch": 2.85, "learning_rate": 1.0903181611855283e-05, "loss": 0.4654, "step": 28351 }, { "epoch": 2.85, "learning_rate": 1.0901335996884478e-05, "loss": 0.1024, "step": 28352 }, { "epoch": 2.85, "learning_rate": 1.0899490503450289e-05, "loss": 0.6424, "step": 28353 }, { "epoch": 2.85, "learning_rate": 1.0897645131564473e-05, "loss": 0.2135, "step": 28354 }, { "epoch": 2.85, "learning_rate": 1.0895799881238781e-05, "loss": 0.5942, "step": 28355 }, { "epoch": 2.85, "learning_rate": 1.0893954752484951e-05, "loss": 0.5033, "step": 28356 }, { "epoch": 2.85, "learning_rate": 1.0892109745314726e-05, "loss": 0.0821, "step": 28357 }, { "epoch": 2.85, "learning_rate": 1.0890264859739847e-05, "loss": 0.1453, "step": 28358 }, { "epoch": 2.85, "learning_rate": 1.0888420095772052e-05, "loss": 0.0395, "step": 28359 }, { "epoch": 2.85, "learning_rate": 1.088657545342308e-05, "loss": 0.375, "step": 28360 }, { "epoch": 2.85, "learning_rate": 1.088473093270467e-05, "loss": 0.2728, "step": 28361 }, { "epoch": 2.85, "learning_rate": 1.0882886533628552e-05, "loss": 0.0808, "step": 28362 }, { "epoch": 2.85, "learning_rate": 1.0881042256206492e-05, "loss": 0.3549, "step": 28363 }, { "epoch": 2.85, "learning_rate": 1.0879198100450195e-05, "loss": 0.282, "step": 28364 }, { "epoch": 2.85, "learning_rate": 1.0877354066371397e-05, "loss": 0.1225, "step": 28365 }, { "epoch": 2.85, "learning_rate": 1.0875510153981864e-05, "loss": 0.8373, "step": 28366 }, { "epoch": 2.85, "learning_rate": 1.08736663632933e-05, "loss": 0.0879, "step": 28367 }, { "epoch": 2.85, "learning_rate": 1.0871822694317435e-05, "loss": 0.289, "step": 28368 }, { "epoch": 2.86, "learning_rate": 1.0869979147066026e-05, "loss": 0.4215, "step": 28369 }, { "epoch": 2.86, "learning_rate": 1.08681357215508e-05, "loss": 0.5646, "step": 28370 }, { "epoch": 2.86, "learning_rate": 1.086629241778346e-05, "loss": 0.2783, "step": 28371 }, { "epoch": 2.86, "learning_rate": 1.086444923577577e-05, "loss": 0.7893, "step": 28372 }, { "epoch": 2.86, "learning_rate": 1.086260617553944e-05, "loss": 0.5483, "step": 28373 }, { "epoch": 2.86, "learning_rate": 1.0860763237086209e-05, "loss": 0.3796, "step": 28374 }, { "epoch": 2.86, "learning_rate": 1.0858920420427799e-05, "loss": 0.0831, "step": 28375 }, { "epoch": 2.86, "learning_rate": 1.0857077725575936e-05, "loss": 0.8085, "step": 28376 }, { "epoch": 2.86, "learning_rate": 1.085523515254235e-05, "loss": 0.251, "step": 28377 }, { "epoch": 2.86, "learning_rate": 1.085339270133876e-05, "loss": 0.3747, "step": 28378 }, { "epoch": 2.86, "learning_rate": 1.0851550371976897e-05, "loss": 0.2898, "step": 28379 }, { "epoch": 2.86, "learning_rate": 1.0849708164468474e-05, "loss": 0.8975, "step": 28380 }, { "epoch": 2.86, "learning_rate": 1.0847866078825245e-05, "loss": 0.3519, "step": 28381 }, { "epoch": 2.86, "learning_rate": 1.0846024115058889e-05, "loss": 0.135, "step": 28382 }, { "epoch": 2.86, "learning_rate": 1.0844182273181157e-05, "loss": 0.1965, "step": 28383 }, { "epoch": 2.86, "learning_rate": 1.0842340553203776e-05, "loss": 0.1708, "step": 28384 }, { "epoch": 2.86, "learning_rate": 1.0840498955138428e-05, "loss": 0.4204, "step": 28385 }, { "epoch": 2.86, "learning_rate": 1.0838657478996865e-05, "loss": 0.3279, "step": 28386 }, { "epoch": 2.86, "learning_rate": 1.08368161247908e-05, "loss": 0.3052, "step": 28387 }, { "epoch": 2.86, "learning_rate": 1.0834974892531945e-05, "loss": 0.0581, "step": 28388 }, { "epoch": 2.86, "learning_rate": 1.0833133782232018e-05, "loss": 0.0844, "step": 28389 }, { "epoch": 2.86, "learning_rate": 1.083129279390274e-05, "loss": 0.1237, "step": 28390 }, { "epoch": 2.86, "learning_rate": 1.082945192755582e-05, "loss": 0.2523, "step": 28391 }, { "epoch": 2.86, "learning_rate": 1.0827611183202975e-05, "loss": 0.1666, "step": 28392 }, { "epoch": 2.86, "learning_rate": 1.082577056085592e-05, "loss": 0.0835, "step": 28393 }, { "epoch": 2.86, "learning_rate": 1.0823930060526357e-05, "loss": 0.3826, "step": 28394 }, { "epoch": 2.86, "learning_rate": 1.0822089682226025e-05, "loss": 0.4803, "step": 28395 }, { "epoch": 2.86, "learning_rate": 1.0820249425966612e-05, "loss": 0.2553, "step": 28396 }, { "epoch": 2.86, "learning_rate": 1.0818409291759825e-05, "loss": 0.2552, "step": 28397 }, { "epoch": 2.86, "learning_rate": 1.0816569279617405e-05, "loss": 0.2716, "step": 28398 }, { "epoch": 2.86, "learning_rate": 1.081472938955103e-05, "loss": 0.6508, "step": 28399 }, { "epoch": 2.86, "learning_rate": 1.0812889621572403e-05, "loss": 0.628, "step": 28400 }, { "epoch": 2.86, "learning_rate": 1.0811049975693265e-05, "loss": 0.0332, "step": 28401 }, { "epoch": 2.86, "learning_rate": 1.0809210451925312e-05, "loss": 0.5859, "step": 28402 }, { "epoch": 2.86, "learning_rate": 1.0807371050280223e-05, "loss": 0.0982, "step": 28403 }, { "epoch": 2.86, "learning_rate": 1.0805531770769734e-05, "loss": 0.2314, "step": 28404 }, { "epoch": 2.86, "learning_rate": 1.0803692613405539e-05, "loss": 0.2512, "step": 28405 }, { "epoch": 2.86, "learning_rate": 1.0801853578199339e-05, "loss": 0.1305, "step": 28406 }, { "epoch": 2.86, "learning_rate": 1.080001466516284e-05, "loss": 0.6643, "step": 28407 }, { "epoch": 2.86, "learning_rate": 1.0798175874307748e-05, "loss": 0.4267, "step": 28408 }, { "epoch": 2.86, "learning_rate": 1.0796337205645757e-05, "loss": 0.1064, "step": 28409 }, { "epoch": 2.86, "learning_rate": 1.0794498659188569e-05, "loss": 0.1854, "step": 28410 }, { "epoch": 2.86, "learning_rate": 1.0792660234947878e-05, "loss": 0.2831, "step": 28411 }, { "epoch": 2.86, "learning_rate": 1.0790821932935398e-05, "loss": 0.1921, "step": 28412 }, { "epoch": 2.86, "learning_rate": 1.0788983753162833e-05, "loss": 0.4509, "step": 28413 }, { "epoch": 2.86, "learning_rate": 1.0787145695641846e-05, "loss": 0.6913, "step": 28414 }, { "epoch": 2.86, "learning_rate": 1.0785307760384163e-05, "loss": 0.1542, "step": 28415 }, { "epoch": 2.86, "learning_rate": 1.0783469947401482e-05, "loss": 0.2127, "step": 28416 }, { "epoch": 2.86, "learning_rate": 1.0781632256705469e-05, "loss": 0.226, "step": 28417 }, { "epoch": 2.86, "learning_rate": 1.0779794688307848e-05, "loss": 0.6352, "step": 28418 }, { "epoch": 2.86, "learning_rate": 1.07779572422203e-05, "loss": 0.216, "step": 28419 }, { "epoch": 2.86, "learning_rate": 1.0776119918454522e-05, "loss": 0.5325, "step": 28420 }, { "epoch": 2.86, "learning_rate": 1.0774282717022203e-05, "loss": 0.3463, "step": 28421 }, { "epoch": 2.86, "learning_rate": 1.0772445637935033e-05, "loss": 0.5532, "step": 28422 }, { "epoch": 2.86, "learning_rate": 1.0770608681204706e-05, "loss": 0.1656, "step": 28423 }, { "epoch": 2.86, "learning_rate": 1.076877184684291e-05, "loss": 0.2387, "step": 28424 }, { "epoch": 2.86, "learning_rate": 1.0766935134861333e-05, "loss": 0.2632, "step": 28425 }, { "epoch": 2.86, "learning_rate": 1.0765098545271653e-05, "loss": 0.4373, "step": 28426 }, { "epoch": 2.86, "learning_rate": 1.0763262078085594e-05, "loss": 0.3617, "step": 28427 }, { "epoch": 2.86, "learning_rate": 1.0761425733314803e-05, "loss": 0.0597, "step": 28428 }, { "epoch": 2.86, "learning_rate": 1.075958951097097e-05, "loss": 0.2392, "step": 28429 }, { "epoch": 2.86, "learning_rate": 1.075775341106581e-05, "loss": 0.2391, "step": 28430 }, { "epoch": 2.86, "learning_rate": 1.0755917433610978e-05, "loss": 0.7311, "step": 28431 }, { "epoch": 2.86, "learning_rate": 1.0754081578618156e-05, "loss": 0.3534, "step": 28432 }, { "epoch": 2.86, "learning_rate": 1.0752245846099048e-05, "loss": 0.4548, "step": 28433 }, { "epoch": 2.86, "learning_rate": 1.0750410236065333e-05, "loss": 0.2827, "step": 28434 }, { "epoch": 2.86, "learning_rate": 1.0748574748528668e-05, "loss": 0.2223, "step": 28435 }, { "epoch": 2.86, "learning_rate": 1.0746739383500758e-05, "loss": 0.2739, "step": 28436 }, { "epoch": 2.86, "learning_rate": 1.0744904140993273e-05, "loss": 0.0847, "step": 28437 }, { "epoch": 2.86, "learning_rate": 1.0743069021017893e-05, "loss": 0.1957, "step": 28438 }, { "epoch": 2.86, "learning_rate": 1.07412340235863e-05, "loss": 0.2998, "step": 28439 }, { "epoch": 2.86, "learning_rate": 1.0739399148710154e-05, "loss": 0.5018, "step": 28440 }, { "epoch": 2.86, "learning_rate": 1.0737564396401165e-05, "loss": 0.5772, "step": 28441 }, { "epoch": 2.86, "learning_rate": 1.0735729766670976e-05, "loss": 0.2551, "step": 28442 }, { "epoch": 2.86, "learning_rate": 1.073389525953127e-05, "loss": 0.3892, "step": 28443 }, { "epoch": 2.86, "learning_rate": 1.073206087499373e-05, "loss": 0.4049, "step": 28444 }, { "epoch": 2.86, "learning_rate": 1.0730226613070037e-05, "loss": 0.1356, "step": 28445 }, { "epoch": 2.86, "learning_rate": 1.072839247377183e-05, "loss": 0.1406, "step": 28446 }, { "epoch": 2.86, "learning_rate": 1.0726558457110813e-05, "loss": 0.6997, "step": 28447 }, { "epoch": 2.86, "learning_rate": 1.0724724563098658e-05, "loss": 0.7543, "step": 28448 }, { "epoch": 2.86, "learning_rate": 1.0722890791747001e-05, "loss": 0.431, "step": 28449 }, { "epoch": 2.86, "learning_rate": 1.0721057143067543e-05, "loss": 0.1376, "step": 28450 }, { "epoch": 2.86, "learning_rate": 1.0719223617071946e-05, "loss": 0.0938, "step": 28451 }, { "epoch": 2.86, "learning_rate": 1.071739021377187e-05, "loss": 0.3708, "step": 28452 }, { "epoch": 2.86, "learning_rate": 1.0715556933178994e-05, "loss": 0.1389, "step": 28453 }, { "epoch": 2.86, "learning_rate": 1.0713723775304972e-05, "loss": 0.296, "step": 28454 }, { "epoch": 2.86, "learning_rate": 1.0711890740161476e-05, "loss": 0.1916, "step": 28455 }, { "epoch": 2.86, "learning_rate": 1.0710057827760173e-05, "loss": 0.733, "step": 28456 }, { "epoch": 2.86, "learning_rate": 1.070822503811272e-05, "loss": 0.325, "step": 28457 }, { "epoch": 2.86, "learning_rate": 1.0706392371230776e-05, "loss": 0.4291, "step": 28458 }, { "epoch": 2.86, "learning_rate": 1.070455982712603e-05, "loss": 0.1623, "step": 28459 }, { "epoch": 2.86, "learning_rate": 1.0702727405810116e-05, "loss": 0.2629, "step": 28460 }, { "epoch": 2.86, "learning_rate": 1.0700895107294693e-05, "loss": 0.1921, "step": 28461 }, { "epoch": 2.86, "learning_rate": 1.0699062931591452e-05, "loss": 0.4228, "step": 28462 }, { "epoch": 2.86, "learning_rate": 1.0697230878712024e-05, "loss": 0.6132, "step": 28463 }, { "epoch": 2.86, "learning_rate": 1.0695398948668062e-05, "loss": 0.2384, "step": 28464 }, { "epoch": 2.86, "learning_rate": 1.0693567141471248e-05, "loss": 0.5882, "step": 28465 }, { "epoch": 2.86, "learning_rate": 1.069173545713323e-05, "loss": 0.1762, "step": 28466 }, { "epoch": 2.87, "learning_rate": 1.0689903895665661e-05, "loss": 0.0851, "step": 28467 }, { "epoch": 2.87, "learning_rate": 1.0688072457080198e-05, "loss": 0.4182, "step": 28468 }, { "epoch": 2.87, "learning_rate": 1.0686241141388496e-05, "loss": 0.5767, "step": 28469 }, { "epoch": 2.87, "learning_rate": 1.0684409948602208e-05, "loss": 0.3654, "step": 28470 }, { "epoch": 2.87, "learning_rate": 1.0682578878732984e-05, "loss": 0.1085, "step": 28471 }, { "epoch": 2.87, "learning_rate": 1.0680747931792473e-05, "loss": 0.4138, "step": 28472 }, { "epoch": 2.87, "learning_rate": 1.0678917107792353e-05, "loss": 0.4867, "step": 28473 }, { "epoch": 2.87, "learning_rate": 1.067708640674424e-05, "loss": 0.1993, "step": 28474 }, { "epoch": 2.87, "learning_rate": 1.067525582865979e-05, "loss": 0.1974, "step": 28475 }, { "epoch": 2.87, "learning_rate": 1.0673425373550675e-05, "loss": 0.1532, "step": 28476 }, { "epoch": 2.87, "learning_rate": 1.0671595041428535e-05, "loss": 0.2778, "step": 28477 }, { "epoch": 2.87, "learning_rate": 1.066976483230499e-05, "loss": 0.1401, "step": 28478 }, { "epoch": 2.87, "learning_rate": 1.0667934746191718e-05, "loss": 0.319, "step": 28479 }, { "epoch": 2.87, "learning_rate": 1.0666104783100368e-05, "loss": 0.2835, "step": 28480 }, { "epoch": 2.87, "learning_rate": 1.066427494304255e-05, "loss": 0.2453, "step": 28481 }, { "epoch": 2.87, "learning_rate": 1.0662445226029942e-05, "loss": 0.8261, "step": 28482 }, { "epoch": 2.87, "learning_rate": 1.0660615632074175e-05, "loss": 0.4367, "step": 28483 }, { "epoch": 2.87, "learning_rate": 1.0658786161186892e-05, "loss": 0.0457, "step": 28484 }, { "epoch": 2.87, "learning_rate": 1.065695681337974e-05, "loss": 0.5047, "step": 28485 }, { "epoch": 2.87, "learning_rate": 1.0655127588664352e-05, "loss": 0.4434, "step": 28486 }, { "epoch": 2.87, "learning_rate": 1.0653298487052374e-05, "loss": 0.3055, "step": 28487 }, { "epoch": 2.87, "learning_rate": 1.0651469508555447e-05, "loss": 0.2443, "step": 28488 }, { "epoch": 2.87, "learning_rate": 1.0649640653185205e-05, "loss": 0.4585, "step": 28489 }, { "epoch": 2.87, "learning_rate": 1.064781192095328e-05, "loss": 0.4414, "step": 28490 }, { "epoch": 2.87, "learning_rate": 1.0645983311871337e-05, "loss": 0.3161, "step": 28491 }, { "epoch": 2.87, "learning_rate": 1.0644154825950973e-05, "loss": 0.1392, "step": 28492 }, { "epoch": 2.87, "learning_rate": 1.0642326463203858e-05, "loss": 0.1424, "step": 28493 }, { "epoch": 2.87, "learning_rate": 1.064049822364162e-05, "loss": 0.0691, "step": 28494 }, { "epoch": 2.87, "learning_rate": 1.0638670107275865e-05, "loss": 0.5252, "step": 28495 }, { "epoch": 2.87, "learning_rate": 1.0636842114118263e-05, "loss": 0.5187, "step": 28496 }, { "epoch": 2.87, "learning_rate": 1.0635014244180426e-05, "loss": 0.307, "step": 28497 }, { "epoch": 2.87, "learning_rate": 1.0633186497473995e-05, "loss": 0.263, "step": 28498 }, { "epoch": 2.87, "learning_rate": 1.0631358874010597e-05, "loss": 0.1184, "step": 28499 }, { "epoch": 2.87, "learning_rate": 1.0629531373801866e-05, "loss": 0.54, "step": 28500 }, { "epoch": 2.87, "learning_rate": 1.0627703996859423e-05, "loss": 0.6953, "step": 28501 }, { "epoch": 2.87, "learning_rate": 1.0625876743194906e-05, "loss": 0.5698, "step": 28502 }, { "epoch": 2.87, "learning_rate": 1.0624049612819937e-05, "loss": 0.1795, "step": 28503 }, { "epoch": 2.87, "learning_rate": 1.0622222605746136e-05, "loss": 0.2984, "step": 28504 }, { "epoch": 2.87, "learning_rate": 1.062039572198516e-05, "loss": 0.5115, "step": 28505 }, { "epoch": 2.87, "learning_rate": 1.0618568961548601e-05, "loss": 0.1541, "step": 28506 }, { "epoch": 2.87, "learning_rate": 1.0616742324448087e-05, "loss": 0.4886, "step": 28507 }, { "epoch": 2.87, "learning_rate": 1.061491581069526e-05, "loss": 0.3329, "step": 28508 }, { "epoch": 2.87, "learning_rate": 1.0613089420301746e-05, "loss": 0.4205, "step": 28509 }, { "epoch": 2.87, "learning_rate": 1.0611263153279135e-05, "loss": 0.3397, "step": 28510 }, { "epoch": 2.87, "learning_rate": 1.060943700963908e-05, "loss": 0.7455, "step": 28511 }, { "epoch": 2.87, "learning_rate": 1.0607610989393201e-05, "loss": 0.4066, "step": 28512 }, { "epoch": 2.87, "learning_rate": 1.060578509255309e-05, "loss": 0.1769, "step": 28513 }, { "epoch": 2.87, "learning_rate": 1.0603959319130391e-05, "loss": 0.2173, "step": 28514 }, { "epoch": 2.87, "learning_rate": 1.0602133669136718e-05, "loss": 0.1773, "step": 28515 }, { "epoch": 2.87, "learning_rate": 1.0600308142583688e-05, "loss": 0.2215, "step": 28516 }, { "epoch": 2.87, "learning_rate": 1.0598482739482917e-05, "loss": 0.3944, "step": 28517 }, { "epoch": 2.87, "learning_rate": 1.059665745984601e-05, "loss": 0.1191, "step": 28518 }, { "epoch": 2.87, "learning_rate": 1.0594832303684612e-05, "loss": 0.8677, "step": 28519 }, { "epoch": 2.87, "learning_rate": 1.0593007271010313e-05, "loss": 0.7781, "step": 28520 }, { "epoch": 2.87, "learning_rate": 1.059118236183472e-05, "loss": 0.347, "step": 28521 }, { "epoch": 2.87, "learning_rate": 1.0589357576169471e-05, "loss": 0.0864, "step": 28522 }, { "epoch": 2.87, "learning_rate": 1.0587532914026175e-05, "loss": 0.8197, "step": 28523 }, { "epoch": 2.87, "learning_rate": 1.0585708375416415e-05, "loss": 0.4116, "step": 28524 }, { "epoch": 2.87, "learning_rate": 1.058388396035183e-05, "loss": 0.2019, "step": 28525 }, { "epoch": 2.87, "learning_rate": 1.0582059668844033e-05, "loss": 0.2274, "step": 28526 }, { "epoch": 2.87, "learning_rate": 1.0580235500904602e-05, "loss": 0.7462, "step": 28527 }, { "epoch": 2.87, "learning_rate": 1.0578411456545174e-05, "loss": 0.1167, "step": 28528 }, { "epoch": 2.87, "learning_rate": 1.0576587535777344e-05, "loss": 0.4855, "step": 28529 }, { "epoch": 2.87, "learning_rate": 1.0574763738612724e-05, "loss": 0.189, "step": 28530 }, { "epoch": 2.87, "learning_rate": 1.0572940065062918e-05, "loss": 0.069, "step": 28531 }, { "epoch": 2.87, "learning_rate": 1.0571116515139531e-05, "loss": 0.3192, "step": 28532 }, { "epoch": 2.87, "learning_rate": 1.0569293088854168e-05, "loss": 0.2239, "step": 28533 }, { "epoch": 2.87, "learning_rate": 1.0567469786218432e-05, "loss": 0.5888, "step": 28534 }, { "epoch": 2.87, "learning_rate": 1.0565646607243924e-05, "loss": 0.2774, "step": 28535 }, { "epoch": 2.87, "learning_rate": 1.0563823551942237e-05, "loss": 0.3905, "step": 28536 }, { "epoch": 2.87, "learning_rate": 1.0562000620325007e-05, "loss": 0.7811, "step": 28537 }, { "epoch": 2.87, "learning_rate": 1.0560177812403797e-05, "loss": 0.233, "step": 28538 }, { "epoch": 2.87, "learning_rate": 1.0558355128190214e-05, "loss": 0.463, "step": 28539 }, { "epoch": 2.87, "learning_rate": 1.055653256769587e-05, "loss": 0.6037, "step": 28540 }, { "epoch": 2.87, "learning_rate": 1.0554710130932367e-05, "loss": 0.6948, "step": 28541 }, { "epoch": 2.87, "learning_rate": 1.0552887817911268e-05, "loss": 1.1287, "step": 28542 }, { "epoch": 2.87, "learning_rate": 1.0551065628644207e-05, "loss": 0.1347, "step": 28543 }, { "epoch": 2.87, "learning_rate": 1.0549243563142775e-05, "loss": 0.1292, "step": 28544 }, { "epoch": 2.87, "learning_rate": 1.0547421621418536e-05, "loss": 0.6561, "step": 28545 }, { "epoch": 2.87, "learning_rate": 1.054559980348312e-05, "loss": 0.8695, "step": 28546 }, { "epoch": 2.87, "learning_rate": 1.0543778109348098e-05, "loss": 0.3046, "step": 28547 }, { "epoch": 2.87, "learning_rate": 1.0541956539025079e-05, "loss": 0.3392, "step": 28548 }, { "epoch": 2.87, "learning_rate": 1.054013509252564e-05, "loss": 0.1208, "step": 28549 }, { "epoch": 2.87, "learning_rate": 1.0538313769861372e-05, "loss": 0.2158, "step": 28550 }, { "epoch": 2.87, "learning_rate": 1.0536492571043892e-05, "loss": 0.0737, "step": 28551 }, { "epoch": 2.87, "learning_rate": 1.0534671496084756e-05, "loss": 1.4083, "step": 28552 }, { "epoch": 2.87, "learning_rate": 1.0532850544995557e-05, "loss": 0.4566, "step": 28553 }, { "epoch": 2.87, "learning_rate": 1.0531029717787904e-05, "loss": 0.7948, "step": 28554 }, { "epoch": 2.87, "learning_rate": 1.0529209014473376e-05, "loss": 0.2978, "step": 28555 }, { "epoch": 2.87, "learning_rate": 1.0527388435063538e-05, "loss": 0.2173, "step": 28556 }, { "epoch": 2.87, "learning_rate": 1.0525567979570003e-05, "loss": 0.3284, "step": 28557 }, { "epoch": 2.87, "learning_rate": 1.0523747648004354e-05, "loss": 0.4442, "step": 28558 }, { "epoch": 2.87, "learning_rate": 1.0521927440378146e-05, "loss": 0.6171, "step": 28559 }, { "epoch": 2.87, "learning_rate": 1.0520107356702991e-05, "loss": 0.8295, "step": 28560 }, { "epoch": 2.87, "learning_rate": 1.0518287396990463e-05, "loss": 0.423, "step": 28561 }, { "epoch": 2.87, "learning_rate": 1.0516467561252143e-05, "loss": 0.2917, "step": 28562 }, { "epoch": 2.87, "learning_rate": 1.0514647849499612e-05, "loss": 0.2544, "step": 28563 }, { "epoch": 2.87, "learning_rate": 1.0512828261744448e-05, "loss": 0.3455, "step": 28564 }, { "epoch": 2.88, "learning_rate": 1.0511008797998232e-05, "loss": 0.2094, "step": 28565 }, { "epoch": 2.88, "learning_rate": 1.0509189458272542e-05, "loss": 0.531, "step": 28566 }, { "epoch": 2.88, "learning_rate": 1.0507370242578957e-05, "loss": 0.6599, "step": 28567 }, { "epoch": 2.88, "learning_rate": 1.0505551150929041e-05, "loss": 0.524, "step": 28568 }, { "epoch": 2.88, "learning_rate": 1.0503732183334401e-05, "loss": 0.1469, "step": 28569 }, { "epoch": 2.88, "learning_rate": 1.0501913339806585e-05, "loss": 0.2494, "step": 28570 }, { "epoch": 2.88, "learning_rate": 1.0500094620357162e-05, "loss": 0.335, "step": 28571 }, { "epoch": 2.88, "learning_rate": 1.0498276024997731e-05, "loss": 0.1787, "step": 28572 }, { "epoch": 2.88, "learning_rate": 1.0496457553739854e-05, "loss": 0.6507, "step": 28573 }, { "epoch": 2.88, "learning_rate": 1.0494639206595098e-05, "loss": 0.5756, "step": 28574 }, { "epoch": 2.88, "learning_rate": 1.049282098357504e-05, "loss": 0.055, "step": 28575 }, { "epoch": 2.88, "learning_rate": 1.0491002884691247e-05, "loss": 0.4465, "step": 28576 }, { "epoch": 2.88, "learning_rate": 1.0489184909955289e-05, "loss": 0.0841, "step": 28577 }, { "epoch": 2.88, "learning_rate": 1.048736705937874e-05, "loss": 0.1079, "step": 28578 }, { "epoch": 2.88, "learning_rate": 1.0485549332973162e-05, "loss": 0.2772, "step": 28579 }, { "epoch": 2.88, "learning_rate": 1.048373173075012e-05, "loss": 0.2684, "step": 28580 }, { "epoch": 2.88, "learning_rate": 1.0481914252721191e-05, "loss": 0.0976, "step": 28581 }, { "epoch": 2.88, "learning_rate": 1.048009689889792e-05, "loss": 0.3353, "step": 28582 }, { "epoch": 2.88, "learning_rate": 1.047827966929191e-05, "loss": 0.0349, "step": 28583 }, { "epoch": 2.88, "learning_rate": 1.047646256391469e-05, "loss": 0.7698, "step": 28584 }, { "epoch": 2.88, "learning_rate": 1.0474645582777829e-05, "loss": 0.2692, "step": 28585 }, { "epoch": 2.88, "learning_rate": 1.0472828725892903e-05, "loss": 0.1773, "step": 28586 }, { "epoch": 2.88, "learning_rate": 1.0471011993271475e-05, "loss": 0.2928, "step": 28587 }, { "epoch": 2.88, "learning_rate": 1.046919538492508e-05, "loss": 0.2965, "step": 28588 }, { "epoch": 2.88, "learning_rate": 1.0467378900865303e-05, "loss": 0.2862, "step": 28589 }, { "epoch": 2.88, "learning_rate": 1.0465562541103706e-05, "loss": 0.1569, "step": 28590 }, { "epoch": 2.88, "learning_rate": 1.046374630565182e-05, "loss": 0.5088, "step": 28591 }, { "epoch": 2.88, "learning_rate": 1.0461930194521229e-05, "loss": 0.469, "step": 28592 }, { "epoch": 2.88, "learning_rate": 1.0460114207723482e-05, "loss": 0.0969, "step": 28593 }, { "epoch": 2.88, "learning_rate": 1.0458298345270137e-05, "loss": 0.6064, "step": 28594 }, { "epoch": 2.88, "learning_rate": 1.0456482607172742e-05, "loss": 0.3003, "step": 28595 }, { "epoch": 2.88, "learning_rate": 1.0454666993442862e-05, "loss": 0.2889, "step": 28596 }, { "epoch": 2.88, "learning_rate": 1.0452851504092045e-05, "loss": 0.2316, "step": 28597 }, { "epoch": 2.88, "learning_rate": 1.0451036139131842e-05, "loss": 0.1097, "step": 28598 }, { "epoch": 2.88, "learning_rate": 1.0449220898573802e-05, "loss": 0.2914, "step": 28599 }, { "epoch": 2.88, "learning_rate": 1.044740578242949e-05, "loss": 0.69, "step": 28600 }, { "epoch": 2.88, "learning_rate": 1.0445590790710458e-05, "loss": 0.0721, "step": 28601 }, { "epoch": 2.88, "learning_rate": 1.0443775923428231e-05, "loss": 0.08, "step": 28602 }, { "epoch": 2.88, "learning_rate": 1.0441961180594383e-05, "loss": 0.2485, "step": 28603 }, { "epoch": 2.88, "learning_rate": 1.0440146562220456e-05, "loss": 0.3914, "step": 28604 }, { "epoch": 2.88, "learning_rate": 1.0438332068317993e-05, "loss": 0.344, "step": 28605 }, { "epoch": 2.88, "learning_rate": 1.0436517698898544e-05, "loss": 0.6177, "step": 28606 }, { "epoch": 2.88, "learning_rate": 1.0434703453973652e-05, "loss": 0.0772, "step": 28607 }, { "epoch": 2.88, "learning_rate": 1.0432889333554866e-05, "loss": 0.4801, "step": 28608 }, { "epoch": 2.88, "learning_rate": 1.0431075337653725e-05, "loss": 0.5863, "step": 28609 }, { "epoch": 2.88, "learning_rate": 1.042926146628178e-05, "loss": 0.2696, "step": 28610 }, { "epoch": 2.88, "learning_rate": 1.0427447719450571e-05, "loss": 0.2269, "step": 28611 }, { "epoch": 2.88, "learning_rate": 1.0425634097171639e-05, "loss": 0.5776, "step": 28612 }, { "epoch": 2.88, "learning_rate": 1.0423820599456521e-05, "loss": 0.8117, "step": 28613 }, { "epoch": 2.88, "learning_rate": 1.0422007226316753e-05, "loss": 0.0611, "step": 28614 }, { "epoch": 2.88, "learning_rate": 1.0420193977763904e-05, "loss": 0.3348, "step": 28615 }, { "epoch": 2.88, "learning_rate": 1.0418380853809482e-05, "loss": 0.0714, "step": 28616 }, { "epoch": 2.88, "learning_rate": 1.0416567854465025e-05, "loss": 0.8273, "step": 28617 }, { "epoch": 2.88, "learning_rate": 1.041475497974209e-05, "loss": 0.7153, "step": 28618 }, { "epoch": 2.88, "learning_rate": 1.0412942229652214e-05, "loss": 0.2085, "step": 28619 }, { "epoch": 2.88, "learning_rate": 1.0411129604206903e-05, "loss": 0.5189, "step": 28620 }, { "epoch": 2.88, "learning_rate": 1.040931710341772e-05, "loss": 0.0734, "step": 28621 }, { "epoch": 2.88, "learning_rate": 1.0407504727296201e-05, "loss": 0.1092, "step": 28622 }, { "epoch": 2.88, "learning_rate": 1.0405692475853849e-05, "loss": 0.3038, "step": 28623 }, { "epoch": 2.88, "learning_rate": 1.0403880349102224e-05, "loss": 0.2837, "step": 28624 }, { "epoch": 2.88, "learning_rate": 1.0402068347052852e-05, "loss": 0.143, "step": 28625 }, { "epoch": 2.88, "learning_rate": 1.0400256469717261e-05, "loss": 0.6389, "step": 28626 }, { "epoch": 2.88, "learning_rate": 1.039844471710698e-05, "loss": 0.3976, "step": 28627 }, { "epoch": 2.88, "learning_rate": 1.0396633089233532e-05, "loss": 0.2501, "step": 28628 }, { "epoch": 2.88, "learning_rate": 1.0394821586108475e-05, "loss": 0.401, "step": 28629 }, { "epoch": 2.88, "learning_rate": 1.0393010207743302e-05, "loss": 0.5644, "step": 28630 }, { "epoch": 2.88, "learning_rate": 1.0391198954149543e-05, "loss": 0.5457, "step": 28631 }, { "epoch": 2.88, "learning_rate": 1.0389387825338745e-05, "loss": 0.7271, "step": 28632 }, { "epoch": 2.88, "learning_rate": 1.0387576821322432e-05, "loss": 0.1688, "step": 28633 }, { "epoch": 2.88, "learning_rate": 1.0385765942112101e-05, "loss": 0.2795, "step": 28634 }, { "epoch": 2.88, "learning_rate": 1.0383955187719302e-05, "loss": 0.1424, "step": 28635 }, { "epoch": 2.88, "learning_rate": 1.0382144558155553e-05, "loss": 0.4095, "step": 28636 }, { "epoch": 2.88, "learning_rate": 1.0380334053432372e-05, "loss": 0.067, "step": 28637 }, { "epoch": 2.88, "learning_rate": 1.0378523673561281e-05, "loss": 0.2809, "step": 28638 }, { "epoch": 2.88, "learning_rate": 1.03767134185538e-05, "loss": 0.0975, "step": 28639 }, { "epoch": 2.88, "learning_rate": 1.0374903288421448e-05, "loss": 0.2584, "step": 28640 }, { "epoch": 2.88, "learning_rate": 1.0373093283175747e-05, "loss": 0.46, "step": 28641 }, { "epoch": 2.88, "learning_rate": 1.0371283402828216e-05, "loss": 0.3624, "step": 28642 }, { "epoch": 2.88, "learning_rate": 1.0369473647390369e-05, "loss": 0.1342, "step": 28643 }, { "epoch": 2.88, "learning_rate": 1.0367664016873724e-05, "loss": 0.3335, "step": 28644 }, { "epoch": 2.88, "learning_rate": 1.0365854511289798e-05, "loss": 0.1354, "step": 28645 }, { "epoch": 2.88, "learning_rate": 1.0364045130650091e-05, "loss": 0.4093, "step": 28646 }, { "epoch": 2.88, "learning_rate": 1.0362235874966154e-05, "loss": 0.3808, "step": 28647 }, { "epoch": 2.88, "learning_rate": 1.0360426744249466e-05, "loss": 1.3756, "step": 28648 }, { "epoch": 2.88, "learning_rate": 1.0358617738511543e-05, "loss": 0.4899, "step": 28649 }, { "epoch": 2.88, "learning_rate": 1.0356808857763914e-05, "loss": 0.2448, "step": 28650 }, { "epoch": 2.88, "learning_rate": 1.035500010201809e-05, "loss": 0.0661, "step": 28651 }, { "epoch": 2.88, "learning_rate": 1.0353191471285549e-05, "loss": 0.6287, "step": 28652 }, { "epoch": 2.88, "learning_rate": 1.0351382965577839e-05, "loss": 0.1368, "step": 28653 }, { "epoch": 2.88, "learning_rate": 1.034957458490645e-05, "loss": 0.0369, "step": 28654 }, { "epoch": 2.88, "learning_rate": 1.0347766329282894e-05, "loss": 0.0729, "step": 28655 }, { "epoch": 2.88, "learning_rate": 1.0345958198718673e-05, "loss": 0.1593, "step": 28656 }, { "epoch": 2.88, "learning_rate": 1.0344150193225303e-05, "loss": 0.0439, "step": 28657 }, { "epoch": 2.88, "learning_rate": 1.0342342312814281e-05, "loss": 0.4495, "step": 28658 }, { "epoch": 2.88, "learning_rate": 1.0340534557497115e-05, "loss": 0.126, "step": 28659 }, { "epoch": 2.88, "learning_rate": 1.0338726927285297e-05, "loss": 0.3338, "step": 28660 }, { "epoch": 2.88, "learning_rate": 1.0336919422190363e-05, "loss": 0.0786, "step": 28661 }, { "epoch": 2.88, "learning_rate": 1.0335112042223783e-05, "loss": 0.3858, "step": 28662 }, { "epoch": 2.88, "learning_rate": 1.0333304787397058e-05, "loss": 0.2805, "step": 28663 }, { "epoch": 2.89, "learning_rate": 1.033149765772171e-05, "loss": 0.3999, "step": 28664 }, { "epoch": 2.89, "learning_rate": 1.0329690653209242e-05, "loss": 0.611, "step": 28665 }, { "epoch": 2.89, "learning_rate": 1.0327883773871113e-05, "loss": 0.337, "step": 28666 }, { "epoch": 2.89, "learning_rate": 1.0326077019718864e-05, "loss": 0.3001, "step": 28667 }, { "epoch": 2.89, "learning_rate": 1.0324270390763975e-05, "loss": 0.1757, "step": 28668 }, { "epoch": 2.89, "learning_rate": 1.0322463887017944e-05, "loss": 0.2981, "step": 28669 }, { "epoch": 2.89, "learning_rate": 1.0320657508492266e-05, "loss": 0.7333, "step": 28670 }, { "epoch": 2.89, "learning_rate": 1.0318851255198436e-05, "loss": 0.4499, "step": 28671 }, { "epoch": 2.89, "learning_rate": 1.0317045127147954e-05, "loss": 0.2625, "step": 28672 }, { "epoch": 2.89, "learning_rate": 1.0315239124352306e-05, "loss": 0.8067, "step": 28673 }, { "epoch": 2.89, "learning_rate": 1.0313433246822987e-05, "loss": 0.2312, "step": 28674 }, { "epoch": 2.89, "learning_rate": 1.0311627494571494e-05, "loss": 0.3108, "step": 28675 }, { "epoch": 2.89, "learning_rate": 1.0309821867609309e-05, "loss": 0.0917, "step": 28676 }, { "epoch": 2.89, "learning_rate": 1.0308016365947927e-05, "loss": 0.4123, "step": 28677 }, { "epoch": 2.89, "learning_rate": 1.030621098959883e-05, "loss": 0.1194, "step": 28678 }, { "epoch": 2.89, "learning_rate": 1.0304405738573534e-05, "loss": 0.1388, "step": 28679 }, { "epoch": 2.89, "learning_rate": 1.030260061288349e-05, "loss": 0.2881, "step": 28680 }, { "epoch": 2.89, "learning_rate": 1.0300795612540208e-05, "loss": 0.1529, "step": 28681 }, { "epoch": 2.89, "learning_rate": 1.029899073755517e-05, "loss": 0.171, "step": 28682 }, { "epoch": 2.89, "learning_rate": 1.029718598793986e-05, "loss": 0.0476, "step": 28683 }, { "epoch": 2.89, "learning_rate": 1.0295381363705764e-05, "loss": 0.0492, "step": 28684 }, { "epoch": 2.89, "learning_rate": 1.0293576864864367e-05, "loss": 0.2173, "step": 28685 }, { "epoch": 2.89, "learning_rate": 1.0291772491427146e-05, "loss": 0.2086, "step": 28686 }, { "epoch": 2.89, "learning_rate": 1.0289968243405589e-05, "loss": 0.5228, "step": 28687 }, { "epoch": 2.89, "learning_rate": 1.0288164120811178e-05, "loss": 0.2666, "step": 28688 }, { "epoch": 2.89, "learning_rate": 1.028636012365539e-05, "loss": 0.1134, "step": 28689 }, { "epoch": 2.89, "learning_rate": 1.0284556251949706e-05, "loss": 0.6186, "step": 28690 }, { "epoch": 2.89, "learning_rate": 1.0282752505705605e-05, "loss": 0.1816, "step": 28691 }, { "epoch": 2.89, "learning_rate": 1.0280948884934556e-05, "loss": 0.4725, "step": 28692 }, { "epoch": 2.89, "learning_rate": 1.027914538964807e-05, "loss": 0.339, "step": 28693 }, { "epoch": 2.89, "learning_rate": 1.027734201985759e-05, "loss": 0.4876, "step": 28694 }, { "epoch": 2.89, "learning_rate": 1.0275538775574592e-05, "loss": 0.1343, "step": 28695 }, { "epoch": 2.89, "learning_rate": 1.027373565681057e-05, "loss": 0.2008, "step": 28696 }, { "epoch": 2.89, "learning_rate": 1.0271932663577003e-05, "loss": 0.6761, "step": 28697 }, { "epoch": 2.89, "learning_rate": 1.0270129795885328e-05, "loss": 0.5057, "step": 28698 }, { "epoch": 2.89, "learning_rate": 1.0268327053747053e-05, "loss": 0.4453, "step": 28699 }, { "epoch": 2.89, "learning_rate": 1.026652443717364e-05, "loss": 0.8588, "step": 28700 }, { "epoch": 2.89, "learning_rate": 1.0264721946176555e-05, "loss": 0.4239, "step": 28701 }, { "epoch": 2.89, "learning_rate": 1.0262919580767276e-05, "loss": 0.2184, "step": 28702 }, { "epoch": 2.89, "learning_rate": 1.0261117340957267e-05, "loss": 0.0421, "step": 28703 }, { "epoch": 2.89, "learning_rate": 1.0259315226757996e-05, "loss": 0.2296, "step": 28704 }, { "epoch": 2.89, "learning_rate": 1.0257513238180935e-05, "loss": 1.039, "step": 28705 }, { "epoch": 2.89, "learning_rate": 1.025571137523755e-05, "loss": 0.7968, "step": 28706 }, { "epoch": 2.89, "learning_rate": 1.0253909637939306e-05, "loss": 0.7016, "step": 28707 }, { "epoch": 2.89, "learning_rate": 1.0252108026297667e-05, "loss": 0.3697, "step": 28708 }, { "epoch": 2.89, "learning_rate": 1.0250306540324095e-05, "loss": 0.2276, "step": 28709 }, { "epoch": 2.89, "learning_rate": 1.0248505180030065e-05, "loss": 0.0387, "step": 28710 }, { "epoch": 2.89, "learning_rate": 1.0246703945427043e-05, "loss": 0.338, "step": 28711 }, { "epoch": 2.89, "learning_rate": 1.0244902836526467e-05, "loss": 0.0897, "step": 28712 }, { "epoch": 2.89, "learning_rate": 1.024310185333982e-05, "loss": 0.192, "step": 28713 }, { "epoch": 2.89, "learning_rate": 1.024130099587856e-05, "loss": 0.2194, "step": 28714 }, { "epoch": 2.89, "learning_rate": 1.0239500264154144e-05, "loss": 0.4039, "step": 28715 }, { "epoch": 2.89, "learning_rate": 1.023769965817803e-05, "loss": 0.2371, "step": 28716 }, { "epoch": 2.89, "learning_rate": 1.0235899177961677e-05, "loss": 0.3894, "step": 28717 }, { "epoch": 2.89, "learning_rate": 1.0234098823516541e-05, "loss": 0.614, "step": 28718 }, { "epoch": 2.89, "learning_rate": 1.0232298594854085e-05, "loss": 0.1085, "step": 28719 }, { "epoch": 2.89, "learning_rate": 1.023049849198576e-05, "loss": 0.4227, "step": 28720 }, { "epoch": 2.89, "learning_rate": 1.022869851492302e-05, "loss": 0.4217, "step": 28721 }, { "epoch": 2.89, "learning_rate": 1.0226898663677319e-05, "loss": 0.5552, "step": 28722 }, { "epoch": 2.89, "learning_rate": 1.0225098938260118e-05, "loss": 0.4733, "step": 28723 }, { "epoch": 2.89, "learning_rate": 1.022329933868285e-05, "loss": 0.6128, "step": 28724 }, { "epoch": 2.89, "learning_rate": 1.0221499864957004e-05, "loss": 0.0775, "step": 28725 }, { "epoch": 2.89, "learning_rate": 1.0219700517094e-05, "loss": 0.6506, "step": 28726 }, { "epoch": 2.89, "learning_rate": 1.0217901295105283e-05, "loss": 0.3633, "step": 28727 }, { "epoch": 2.89, "learning_rate": 1.0216102199002331e-05, "loss": 0.3426, "step": 28728 }, { "epoch": 2.89, "learning_rate": 1.0214303228796587e-05, "loss": 0.3351, "step": 28729 }, { "epoch": 2.89, "learning_rate": 1.0212504384499468e-05, "loss": 0.5225, "step": 28730 }, { "epoch": 2.89, "learning_rate": 1.0210705666122457e-05, "loss": 0.1839, "step": 28731 }, { "epoch": 2.89, "learning_rate": 1.0208907073676985e-05, "loss": 0.3953, "step": 28732 }, { "epoch": 2.89, "learning_rate": 1.0207108607174499e-05, "loss": 0.3024, "step": 28733 }, { "epoch": 2.89, "learning_rate": 1.0205310266626447e-05, "loss": 0.2771, "step": 28734 }, { "epoch": 2.89, "learning_rate": 1.0203512052044268e-05, "loss": 0.1961, "step": 28735 }, { "epoch": 2.89, "learning_rate": 1.020171396343941e-05, "loss": 0.1273, "step": 28736 }, { "epoch": 2.89, "learning_rate": 1.0199916000823308e-05, "loss": 0.2097, "step": 28737 }, { "epoch": 2.89, "learning_rate": 1.0198118164207402e-05, "loss": 0.4904, "step": 28738 }, { "epoch": 2.89, "learning_rate": 1.0196320453603156e-05, "loss": 0.0611, "step": 28739 }, { "epoch": 2.89, "learning_rate": 1.0194522869021986e-05, "loss": 0.6134, "step": 28740 }, { "epoch": 2.89, "learning_rate": 1.0192725410475326e-05, "loss": 0.3147, "step": 28741 }, { "epoch": 2.89, "learning_rate": 1.0190928077974635e-05, "loss": 0.5103, "step": 28742 }, { "epoch": 2.89, "learning_rate": 1.0189130871531355e-05, "loss": 0.9151, "step": 28743 }, { "epoch": 2.89, "learning_rate": 1.018733379115689e-05, "loss": 0.1393, "step": 28744 }, { "epoch": 2.89, "learning_rate": 1.0185536836862705e-05, "loss": 0.4533, "step": 28745 }, { "epoch": 2.89, "learning_rate": 1.0183740008660224e-05, "loss": 0.503, "step": 28746 }, { "epoch": 2.89, "learning_rate": 1.0181943306560884e-05, "loss": 0.5419, "step": 28747 }, { "epoch": 2.89, "learning_rate": 1.0180146730576117e-05, "loss": 0.4586, "step": 28748 }, { "epoch": 2.89, "learning_rate": 1.0178350280717356e-05, "loss": 0.3121, "step": 28749 }, { "epoch": 2.89, "learning_rate": 1.0176553956996037e-05, "loss": 0.1191, "step": 28750 }, { "epoch": 2.89, "learning_rate": 1.0174757759423584e-05, "loss": 0.2138, "step": 28751 }, { "epoch": 2.89, "learning_rate": 1.0172961688011431e-05, "loss": 0.1469, "step": 28752 }, { "epoch": 2.89, "learning_rate": 1.0171165742771007e-05, "loss": 0.2752, "step": 28753 }, { "epoch": 2.89, "learning_rate": 1.0169369923713744e-05, "loss": 0.2588, "step": 28754 }, { "epoch": 2.89, "learning_rate": 1.0167574230851062e-05, "loss": 0.0943, "step": 28755 }, { "epoch": 2.89, "learning_rate": 1.0165778664194385e-05, "loss": 0.7227, "step": 28756 }, { "epoch": 2.89, "learning_rate": 1.016398322375517e-05, "loss": 0.583, "step": 28757 }, { "epoch": 2.89, "learning_rate": 1.0162187909544807e-05, "loss": 0.2175, "step": 28758 }, { "epoch": 2.89, "learning_rate": 1.0160392721574723e-05, "loss": 0.3521, "step": 28759 }, { "epoch": 2.89, "learning_rate": 1.0158597659856362e-05, "loss": 0.2917, "step": 28760 }, { "epoch": 2.89, "learning_rate": 1.0156802724401151e-05, "loss": 0.2465, "step": 28761 }, { "epoch": 2.9, "learning_rate": 1.0155007915220476e-05, "loss": 0.2701, "step": 28762 }, { "epoch": 2.9, "learning_rate": 1.0153213232325791e-05, "loss": 0.2447, "step": 28763 }, { "epoch": 2.9, "learning_rate": 1.0151418675728508e-05, "loss": 0.6651, "step": 28764 }, { "epoch": 2.9, "learning_rate": 1.0149624245440044e-05, "loss": 0.2714, "step": 28765 }, { "epoch": 2.9, "learning_rate": 1.0147829941471822e-05, "loss": 0.8772, "step": 28766 }, { "epoch": 2.9, "learning_rate": 1.0146035763835259e-05, "loss": 0.6432, "step": 28767 }, { "epoch": 2.9, "learning_rate": 1.0144241712541768e-05, "loss": 0.4409, "step": 28768 }, { "epoch": 2.9, "learning_rate": 1.0142447787602773e-05, "loss": 0.6785, "step": 28769 }, { "epoch": 2.9, "learning_rate": 1.0140653989029673e-05, "loss": 0.3197, "step": 28770 }, { "epoch": 2.9, "learning_rate": 1.0138860316833915e-05, "loss": 0.2599, "step": 28771 }, { "epoch": 2.9, "learning_rate": 1.0137066771026887e-05, "loss": 0.6219, "step": 28772 }, { "epoch": 2.9, "learning_rate": 1.0135273351620002e-05, "loss": 0.5031, "step": 28773 }, { "epoch": 2.9, "learning_rate": 1.0133480058624686e-05, "loss": 0.2457, "step": 28774 }, { "epoch": 2.9, "learning_rate": 1.0131686892052357e-05, "loss": 0.3173, "step": 28775 }, { "epoch": 2.9, "learning_rate": 1.0129893851914395e-05, "loss": 0.0839, "step": 28776 }, { "epoch": 2.9, "learning_rate": 1.0128100938222241e-05, "loss": 0.4193, "step": 28777 }, { "epoch": 2.9, "learning_rate": 1.012630815098729e-05, "loss": 0.6315, "step": 28778 }, { "epoch": 2.9, "learning_rate": 1.0124515490220954e-05, "loss": 0.7097, "step": 28779 }, { "epoch": 2.9, "learning_rate": 1.0122722955934642e-05, "loss": 0.3923, "step": 28780 }, { "epoch": 2.9, "learning_rate": 1.0120930548139759e-05, "loss": 0.6169, "step": 28781 }, { "epoch": 2.9, "learning_rate": 1.011913826684771e-05, "loss": 0.1461, "step": 28782 }, { "epoch": 2.9, "learning_rate": 1.0117346112069902e-05, "loss": 0.1389, "step": 28783 }, { "epoch": 2.9, "learning_rate": 1.0115554083817743e-05, "loss": 0.3218, "step": 28784 }, { "epoch": 2.9, "learning_rate": 1.011376218210263e-05, "loss": 0.6543, "step": 28785 }, { "epoch": 2.9, "learning_rate": 1.011197040693597e-05, "loss": 0.5333, "step": 28786 }, { "epoch": 2.9, "learning_rate": 1.0110178758329165e-05, "loss": 0.7285, "step": 28787 }, { "epoch": 2.9, "learning_rate": 1.0108387236293607e-05, "loss": 0.1261, "step": 28788 }, { "epoch": 2.9, "learning_rate": 1.0106595840840723e-05, "loss": 0.0842, "step": 28789 }, { "epoch": 2.9, "learning_rate": 1.0104804571981879e-05, "loss": 0.2593, "step": 28790 }, { "epoch": 2.9, "learning_rate": 1.0103013429728497e-05, "loss": 0.5306, "step": 28791 }, { "epoch": 2.9, "learning_rate": 1.0101222414091968e-05, "loss": 0.2927, "step": 28792 }, { "epoch": 2.9, "learning_rate": 1.009943152508369e-05, "loss": 0.1717, "step": 28793 }, { "epoch": 2.9, "learning_rate": 1.0097640762715062e-05, "loss": 0.668, "step": 28794 }, { "epoch": 2.9, "learning_rate": 1.0095850126997472e-05, "loss": 0.5871, "step": 28795 }, { "epoch": 2.9, "learning_rate": 1.0094059617942324e-05, "loss": 0.2508, "step": 28796 }, { "epoch": 2.9, "learning_rate": 1.0092269235561006e-05, "loss": 0.232, "step": 28797 }, { "epoch": 2.9, "learning_rate": 1.0090478979864916e-05, "loss": 0.047, "step": 28798 }, { "epoch": 2.9, "learning_rate": 1.008868885086544e-05, "loss": 0.1769, "step": 28799 }, { "epoch": 2.9, "learning_rate": 1.0086898848573976e-05, "loss": 0.3952, "step": 28800 }, { "epoch": 2.9, "learning_rate": 1.0085108973001911e-05, "loss": 0.2645, "step": 28801 }, { "epoch": 2.9, "learning_rate": 1.0083319224160628e-05, "loss": 0.2135, "step": 28802 }, { "epoch": 2.9, "learning_rate": 1.0081529602061545e-05, "loss": 0.2951, "step": 28803 }, { "epoch": 2.9, "learning_rate": 1.007974010671602e-05, "loss": 0.3079, "step": 28804 }, { "epoch": 2.9, "learning_rate": 1.0077950738135438e-05, "loss": 0.2638, "step": 28805 }, { "epoch": 2.9, "learning_rate": 1.007616149633121e-05, "loss": 0.3658, "step": 28806 }, { "epoch": 2.9, "learning_rate": 1.0074372381314724e-05, "loss": 0.6091, "step": 28807 }, { "epoch": 2.9, "learning_rate": 1.0072583393097328e-05, "loss": 0.1963, "step": 28808 }, { "epoch": 2.9, "learning_rate": 1.0070794531690444e-05, "loss": 0.1788, "step": 28809 }, { "epoch": 2.9, "learning_rate": 1.0069005797105439e-05, "loss": 0.2522, "step": 28810 }, { "epoch": 2.9, "learning_rate": 1.00672171893537e-05, "loss": 0.2715, "step": 28811 }, { "epoch": 2.9, "learning_rate": 1.006542870844661e-05, "loss": 0.19, "step": 28812 }, { "epoch": 2.9, "learning_rate": 1.0063640354395547e-05, "loss": 0.5511, "step": 28813 }, { "epoch": 2.9, "learning_rate": 1.006185212721189e-05, "loss": 0.271, "step": 28814 }, { "epoch": 2.9, "learning_rate": 1.0060064026907025e-05, "loss": 0.2987, "step": 28815 }, { "epoch": 2.9, "learning_rate": 1.0058276053492314e-05, "loss": 0.3547, "step": 28816 }, { "epoch": 2.9, "learning_rate": 1.0056488206979168e-05, "loss": 0.2457, "step": 28817 }, { "epoch": 2.9, "learning_rate": 1.0054700487378937e-05, "loss": 0.2027, "step": 28818 }, { "epoch": 2.9, "learning_rate": 1.0052912894702994e-05, "loss": 0.1434, "step": 28819 }, { "epoch": 2.9, "learning_rate": 1.0051125428962732e-05, "loss": 0.3924, "step": 28820 }, { "epoch": 2.9, "learning_rate": 1.0049338090169533e-05, "loss": 0.5213, "step": 28821 }, { "epoch": 2.9, "learning_rate": 1.0047550878334735e-05, "loss": 0.1199, "step": 28822 }, { "epoch": 2.9, "learning_rate": 1.0045763793469741e-05, "loss": 0.224, "step": 28823 }, { "epoch": 2.9, "learning_rate": 1.0043976835585917e-05, "loss": 0.3687, "step": 28824 }, { "epoch": 2.9, "learning_rate": 1.0042190004694634e-05, "loss": 0.0627, "step": 28825 }, { "epoch": 2.9, "learning_rate": 1.004040330080726e-05, "loss": 0.2581, "step": 28826 }, { "epoch": 2.9, "learning_rate": 1.0038616723935165e-05, "loss": 0.5617, "step": 28827 }, { "epoch": 2.9, "learning_rate": 1.0036830274089722e-05, "loss": 0.2608, "step": 28828 }, { "epoch": 2.9, "learning_rate": 1.0035043951282295e-05, "loss": 0.5929, "step": 28829 }, { "epoch": 2.9, "learning_rate": 1.0033257755524256e-05, "loss": 0.3947, "step": 28830 }, { "epoch": 2.9, "learning_rate": 1.0031471686826965e-05, "loss": 0.513, "step": 28831 }, { "epoch": 2.9, "learning_rate": 1.0029685745201795e-05, "loss": 0.4577, "step": 28832 }, { "epoch": 2.9, "learning_rate": 1.0027899930660103e-05, "loss": 0.6339, "step": 28833 }, { "epoch": 2.9, "learning_rate": 1.0026114243213255e-05, "loss": 0.1059, "step": 28834 }, { "epoch": 2.9, "learning_rate": 1.002432868287263e-05, "loss": 0.7914, "step": 28835 }, { "epoch": 2.9, "learning_rate": 1.0022543249649573e-05, "loss": 0.251, "step": 28836 }, { "epoch": 2.9, "learning_rate": 1.0020757943555441e-05, "loss": 0.1406, "step": 28837 }, { "epoch": 2.9, "learning_rate": 1.0018972764601615e-05, "loss": 0.4838, "step": 28838 }, { "epoch": 2.9, "learning_rate": 1.0017187712799454e-05, "loss": 0.5153, "step": 28839 }, { "epoch": 2.9, "learning_rate": 1.0015402788160288e-05, "loss": 0.4964, "step": 28840 }, { "epoch": 2.9, "learning_rate": 1.0013617990695507e-05, "loss": 0.1689, "step": 28841 }, { "epoch": 2.9, "learning_rate": 1.0011833320416458e-05, "loss": 0.6427, "step": 28842 }, { "epoch": 2.9, "learning_rate": 1.0010048777334498e-05, "loss": 0.3738, "step": 28843 }, { "epoch": 2.9, "learning_rate": 1.0008264361460982e-05, "loss": 0.1767, "step": 28844 }, { "epoch": 2.9, "learning_rate": 1.0006480072807268e-05, "loss": 0.2545, "step": 28845 }, { "epoch": 2.9, "learning_rate": 1.0004695911384709e-05, "loss": 0.2836, "step": 28846 }, { "epoch": 2.9, "learning_rate": 1.0002911877204656e-05, "loss": 0.3287, "step": 28847 }, { "epoch": 2.9, "learning_rate": 1.0001127970278456e-05, "loss": 0.6039, "step": 28848 }, { "epoch": 2.9, "learning_rate": 9.999344190617488e-06, "loss": 0.3413, "step": 28849 }, { "epoch": 2.9, "learning_rate": 9.997560538233078e-06, "loss": 0.0545, "step": 28850 }, { "epoch": 2.9, "learning_rate": 9.995777013136572e-06, "loss": 0.3591, "step": 28851 }, { "epoch": 2.9, "learning_rate": 9.993993615339341e-06, "loss": 0.1218, "step": 28852 }, { "epoch": 2.9, "learning_rate": 9.992210344852735e-06, "loss": 0.1445, "step": 28853 }, { "epoch": 2.9, "learning_rate": 9.990427201688071e-06, "loss": 0.3109, "step": 28854 }, { "epoch": 2.9, "learning_rate": 9.988644185856723e-06, "loss": 0.4579, "step": 28855 }, { "epoch": 2.9, "learning_rate": 9.986861297370035e-06, "loss": 0.3288, "step": 28856 }, { "epoch": 2.9, "learning_rate": 9.985078536239346e-06, "loss": 0.3062, "step": 28857 }, { "epoch": 2.9, "learning_rate": 9.983295902476002e-06, "loss": 0.2398, "step": 28858 }, { "epoch": 2.9, "learning_rate": 9.981513396091349e-06, "loss": 0.0757, "step": 28859 }, { "epoch": 2.9, "learning_rate": 9.979731017096733e-06, "loss": 0.4012, "step": 28860 }, { "epoch": 2.91, "learning_rate": 9.977948765503488e-06, "loss": 0.9285, "step": 28861 }, { "epoch": 2.91, "learning_rate": 9.97616664132296e-06, "loss": 0.5022, "step": 28862 }, { "epoch": 2.91, "learning_rate": 9.974384644566496e-06, "loss": 0.6813, "step": 28863 }, { "epoch": 2.91, "learning_rate": 9.972602775245425e-06, "loss": 0.7614, "step": 28864 }, { "epoch": 2.91, "learning_rate": 9.970821033371091e-06, "loss": 0.8736, "step": 28865 }, { "epoch": 2.91, "learning_rate": 9.969039418954826e-06, "loss": 0.0497, "step": 28866 }, { "epoch": 2.91, "learning_rate": 9.96725793200799e-06, "loss": 0.719, "step": 28867 }, { "epoch": 2.91, "learning_rate": 9.965476572541894e-06, "loss": 0.0919, "step": 28868 }, { "epoch": 2.91, "learning_rate": 9.963695340567873e-06, "loss": 0.6641, "step": 28869 }, { "epoch": 2.91, "learning_rate": 9.961914236097287e-06, "loss": 0.5032, "step": 28870 }, { "epoch": 2.91, "learning_rate": 9.960133259141455e-06, "loss": 0.2187, "step": 28871 }, { "epoch": 2.91, "learning_rate": 9.958352409711708e-06, "loss": 0.3298, "step": 28872 }, { "epoch": 2.91, "learning_rate": 9.956571687819384e-06, "loss": 0.1138, "step": 28873 }, { "epoch": 2.91, "learning_rate": 9.95479109347581e-06, "loss": 0.2862, "step": 28874 }, { "epoch": 2.91, "learning_rate": 9.953010626692322e-06, "loss": 0.2109, "step": 28875 }, { "epoch": 2.91, "learning_rate": 9.951230287480252e-06, "loss": 0.8113, "step": 28876 }, { "epoch": 2.91, "learning_rate": 9.949450075850923e-06, "loss": 0.3132, "step": 28877 }, { "epoch": 2.91, "learning_rate": 9.947669991815664e-06, "loss": 0.1818, "step": 28878 }, { "epoch": 2.91, "learning_rate": 9.945890035385806e-06, "loss": 0.526, "step": 28879 }, { "epoch": 2.91, "learning_rate": 9.944110206572665e-06, "loss": 0.1298, "step": 28880 }, { "epoch": 2.91, "learning_rate": 9.942330505387595e-06, "loss": 1.1236, "step": 28881 }, { "epoch": 2.91, "learning_rate": 9.940550931841895e-06, "loss": 0.2765, "step": 28882 }, { "epoch": 2.91, "learning_rate": 9.938771485946887e-06, "loss": 0.1765, "step": 28883 }, { "epoch": 2.91, "learning_rate": 9.936992167713919e-06, "loss": 0.3611, "step": 28884 }, { "epoch": 2.91, "learning_rate": 9.935212977154308e-06, "loss": 0.3217, "step": 28885 }, { "epoch": 2.91, "learning_rate": 9.93343391427935e-06, "loss": 0.2324, "step": 28886 }, { "epoch": 2.91, "learning_rate": 9.931654979100392e-06, "loss": 0.3294, "step": 28887 }, { "epoch": 2.91, "learning_rate": 9.929876171628749e-06, "loss": 0.3534, "step": 28888 }, { "epoch": 2.91, "learning_rate": 9.928097491875738e-06, "loss": 0.1561, "step": 28889 }, { "epoch": 2.91, "learning_rate": 9.92631893985268e-06, "loss": 0.1651, "step": 28890 }, { "epoch": 2.91, "learning_rate": 9.924540515570895e-06, "loss": 0.1687, "step": 28891 }, { "epoch": 2.91, "learning_rate": 9.92276221904169e-06, "loss": 0.3404, "step": 28892 }, { "epoch": 2.91, "learning_rate": 9.920984050276394e-06, "loss": 1.7602, "step": 28893 }, { "epoch": 2.91, "learning_rate": 9.919206009286313e-06, "loss": 0.255, "step": 28894 }, { "epoch": 2.91, "learning_rate": 9.91742809608277e-06, "loss": 0.4825, "step": 28895 }, { "epoch": 2.91, "learning_rate": 9.915650310677073e-06, "loss": 0.6703, "step": 28896 }, { "epoch": 2.91, "learning_rate": 9.913872653080527e-06, "loss": 0.3674, "step": 28897 }, { "epoch": 2.91, "learning_rate": 9.912095123304464e-06, "loss": 0.4222, "step": 28898 }, { "epoch": 2.91, "learning_rate": 9.910317721360196e-06, "loss": 0.0831, "step": 28899 }, { "epoch": 2.91, "learning_rate": 9.908540447259003e-06, "loss": 0.1053, "step": 28900 }, { "epoch": 2.91, "learning_rate": 9.906763301012225e-06, "loss": 0.2447, "step": 28901 }, { "epoch": 2.91, "learning_rate": 9.904986282631165e-06, "loss": 0.726, "step": 28902 }, { "epoch": 2.91, "learning_rate": 9.903209392127123e-06, "loss": 0.9096, "step": 28903 }, { "epoch": 2.91, "learning_rate": 9.901432629511414e-06, "loss": 0.9008, "step": 28904 }, { "epoch": 2.91, "learning_rate": 9.89965599479534e-06, "loss": 0.2544, "step": 28905 }, { "epoch": 2.91, "learning_rate": 9.89787948799021e-06, "loss": 0.7393, "step": 28906 }, { "epoch": 2.91, "learning_rate": 9.896103109107329e-06, "loss": 0.4053, "step": 28907 }, { "epoch": 2.91, "learning_rate": 9.894326858157995e-06, "loss": 0.282, "step": 28908 }, { "epoch": 2.91, "learning_rate": 9.892550735153521e-06, "loss": 0.8061, "step": 28909 }, { "epoch": 2.91, "learning_rate": 9.890774740105203e-06, "loss": 0.1807, "step": 28910 }, { "epoch": 2.91, "learning_rate": 9.888998873024347e-06, "loss": 0.1328, "step": 28911 }, { "epoch": 2.91, "learning_rate": 9.88722313392224e-06, "loss": 0.3508, "step": 28912 }, { "epoch": 2.91, "learning_rate": 9.885447522810213e-06, "loss": 0.1238, "step": 28913 }, { "epoch": 2.91, "learning_rate": 9.883672039699538e-06, "loss": 0.501, "step": 28914 }, { "epoch": 2.91, "learning_rate": 9.881896684601508e-06, "loss": 0.3021, "step": 28915 }, { "epoch": 2.91, "learning_rate": 9.880121457527446e-06, "loss": 0.3447, "step": 28916 }, { "epoch": 2.91, "learning_rate": 9.878346358488649e-06, "loss": 0.196, "step": 28917 }, { "epoch": 2.91, "learning_rate": 9.876571387496376e-06, "loss": 0.1424, "step": 28918 }, { "epoch": 2.91, "learning_rate": 9.87479654456196e-06, "loss": 0.2645, "step": 28919 }, { "epoch": 2.91, "learning_rate": 9.873021829696686e-06, "loss": 0.2422, "step": 28920 }, { "epoch": 2.91, "learning_rate": 9.871247242911839e-06, "loss": 0.3055, "step": 28921 }, { "epoch": 2.91, "learning_rate": 9.86947278421872e-06, "loss": 0.1331, "step": 28922 }, { "epoch": 2.91, "learning_rate": 9.86769845362862e-06, "loss": 0.0741, "step": 28923 }, { "epoch": 2.91, "learning_rate": 9.865924251152825e-06, "loss": 0.3764, "step": 28924 }, { "epoch": 2.91, "learning_rate": 9.86415017680263e-06, "loss": 1.03, "step": 28925 }, { "epoch": 2.91, "learning_rate": 9.862376230589312e-06, "loss": 0.1108, "step": 28926 }, { "epoch": 2.91, "learning_rate": 9.860602412524194e-06, "loss": 0.3129, "step": 28927 }, { "epoch": 2.91, "learning_rate": 9.85882872261853e-06, "loss": 0.2203, "step": 28928 }, { "epoch": 2.91, "learning_rate": 9.857055160883608e-06, "loss": 0.5717, "step": 28929 }, { "epoch": 2.91, "learning_rate": 9.855281727330732e-06, "loss": 0.7244, "step": 28930 }, { "epoch": 2.91, "learning_rate": 9.85350842197119e-06, "loss": 0.3759, "step": 28931 }, { "epoch": 2.91, "learning_rate": 9.851735244816239e-06, "loss": 0.1488, "step": 28932 }, { "epoch": 2.91, "learning_rate": 9.849962195877186e-06, "loss": 0.7631, "step": 28933 }, { "epoch": 2.91, "learning_rate": 9.848189275165312e-06, "loss": 0.281, "step": 28934 }, { "epoch": 2.91, "learning_rate": 9.846416482691893e-06, "loss": 0.3176, "step": 28935 }, { "epoch": 2.91, "learning_rate": 9.844643818468213e-06, "loss": 0.2961, "step": 28936 }, { "epoch": 2.91, "learning_rate": 9.842871282505552e-06, "loss": 0.141, "step": 28937 }, { "epoch": 2.91, "learning_rate": 9.84109887481519e-06, "loss": 0.2996, "step": 28938 }, { "epoch": 2.91, "learning_rate": 9.839326595408407e-06, "loss": 0.8775, "step": 28939 }, { "epoch": 2.91, "learning_rate": 9.83755444429648e-06, "loss": 0.1582, "step": 28940 }, { "epoch": 2.91, "learning_rate": 9.835782421490687e-06, "loss": 0.2213, "step": 28941 }, { "epoch": 2.91, "learning_rate": 9.834010527002299e-06, "loss": 0.3289, "step": 28942 }, { "epoch": 2.91, "learning_rate": 9.832238760842603e-06, "loss": 0.7253, "step": 28943 }, { "epoch": 2.91, "learning_rate": 9.830467123022853e-06, "loss": 0.2413, "step": 28944 }, { "epoch": 2.91, "learning_rate": 9.828695613554359e-06, "loss": 0.0376, "step": 28945 }, { "epoch": 2.91, "learning_rate": 9.826924232448361e-06, "loss": 0.3301, "step": 28946 }, { "epoch": 2.91, "learning_rate": 9.825152979716136e-06, "loss": 0.2558, "step": 28947 }, { "epoch": 2.91, "learning_rate": 9.823381855368969e-06, "loss": 0.1564, "step": 28948 }, { "epoch": 2.91, "learning_rate": 9.821610859418136e-06, "loss": 0.1371, "step": 28949 }, { "epoch": 2.91, "learning_rate": 9.819839991874879e-06, "loss": 0.3177, "step": 28950 }, { "epoch": 2.91, "learning_rate": 9.81806925275049e-06, "loss": 0.3184, "step": 28951 }, { "epoch": 2.91, "learning_rate": 9.81629864205623e-06, "loss": 0.1618, "step": 28952 }, { "epoch": 2.91, "learning_rate": 9.814528159803364e-06, "loss": 0.5495, "step": 28953 }, { "epoch": 2.91, "learning_rate": 9.812757806003167e-06, "loss": 0.1383, "step": 28954 }, { "epoch": 2.91, "learning_rate": 9.810987580666887e-06, "loss": 0.1192, "step": 28955 }, { "epoch": 2.91, "learning_rate": 9.809217483805823e-06, "loss": 0.147, "step": 28956 }, { "epoch": 2.91, "learning_rate": 9.807447515431208e-06, "loss": 0.1609, "step": 28957 }, { "epoch": 2.91, "learning_rate": 9.805677675554306e-06, "loss": 0.5066, "step": 28958 }, { "epoch": 2.92, "learning_rate": 9.80390796418641e-06, "loss": 0.4883, "step": 28959 }, { "epoch": 2.92, "learning_rate": 9.802138381338747e-06, "loss": 0.5378, "step": 28960 }, { "epoch": 2.92, "learning_rate": 9.800368927022584e-06, "loss": 0.4601, "step": 28961 }, { "epoch": 2.92, "learning_rate": 9.798599601249199e-06, "loss": 0.1879, "step": 28962 }, { "epoch": 2.92, "learning_rate": 9.79683040402985e-06, "loss": 0.3958, "step": 28963 }, { "epoch": 2.92, "learning_rate": 9.795061335375765e-06, "loss": 0.6656, "step": 28964 }, { "epoch": 2.92, "learning_rate": 9.793292395298238e-06, "loss": 0.8532, "step": 28965 }, { "epoch": 2.92, "learning_rate": 9.791523583808506e-06, "loss": 0.6175, "step": 28966 }, { "epoch": 2.92, "learning_rate": 9.789754900917834e-06, "loss": 0.3224, "step": 28967 }, { "epoch": 2.92, "learning_rate": 9.787986346637475e-06, "loss": 0.4976, "step": 28968 }, { "epoch": 2.92, "learning_rate": 9.786217920978677e-06, "loss": 0.2507, "step": 28969 }, { "epoch": 2.92, "learning_rate": 9.784449623952702e-06, "loss": 0.1634, "step": 28970 }, { "epoch": 2.92, "learning_rate": 9.782681455570794e-06, "loss": 0.1565, "step": 28971 }, { "epoch": 2.92, "learning_rate": 9.780913415844217e-06, "loss": 0.9065, "step": 28972 }, { "epoch": 2.92, "learning_rate": 9.77914550478421e-06, "loss": 0.4881, "step": 28973 }, { "epoch": 2.92, "learning_rate": 9.777377722402032e-06, "loss": 0.4284, "step": 28974 }, { "epoch": 2.92, "learning_rate": 9.775610068708926e-06, "loss": 0.2211, "step": 28975 }, { "epoch": 2.92, "learning_rate": 9.773842543716139e-06, "loss": 0.1307, "step": 28976 }, { "epoch": 2.92, "learning_rate": 9.772075147434938e-06, "loss": 0.5071, "step": 28977 }, { "epoch": 2.92, "learning_rate": 9.77030787987654e-06, "loss": 0.1775, "step": 28978 }, { "epoch": 2.92, "learning_rate": 9.768540741052213e-06, "loss": 0.8838, "step": 28979 }, { "epoch": 2.92, "learning_rate": 9.766773730973198e-06, "loss": 0.7755, "step": 28980 }, { "epoch": 2.92, "learning_rate": 9.765006849650737e-06, "loss": 0.7288, "step": 28981 }, { "epoch": 2.92, "learning_rate": 9.763240097096076e-06, "loss": 0.5064, "step": 28982 }, { "epoch": 2.92, "learning_rate": 9.761473473320458e-06, "loss": 0.0619, "step": 28983 }, { "epoch": 2.92, "learning_rate": 9.759706978335123e-06, "loss": 0.7074, "step": 28984 }, { "epoch": 2.92, "learning_rate": 9.757940612151309e-06, "loss": 0.0365, "step": 28985 }, { "epoch": 2.92, "learning_rate": 9.756174374780263e-06, "loss": 0.5456, "step": 28986 }, { "epoch": 2.92, "learning_rate": 9.754408266233212e-06, "loss": 0.1666, "step": 28987 }, { "epoch": 2.92, "learning_rate": 9.752642286521427e-06, "loss": 0.6848, "step": 28988 }, { "epoch": 2.92, "learning_rate": 9.750876435656112e-06, "loss": 0.2752, "step": 28989 }, { "epoch": 2.92, "learning_rate": 9.749110713648508e-06, "loss": 0.019, "step": 28990 }, { "epoch": 2.92, "learning_rate": 9.747345120509877e-06, "loss": 0.4695, "step": 28991 }, { "epoch": 2.92, "learning_rate": 9.745579656251428e-06, "loss": 0.4128, "step": 28992 }, { "epoch": 2.92, "learning_rate": 9.743814320884396e-06, "loss": 0.3209, "step": 28993 }, { "epoch": 2.92, "learning_rate": 9.742049114420034e-06, "loss": 0.2393, "step": 28994 }, { "epoch": 2.92, "learning_rate": 9.740284036869574e-06, "loss": 0.5724, "step": 28995 }, { "epoch": 2.92, "learning_rate": 9.738519088244218e-06, "loss": 0.1973, "step": 28996 }, { "epoch": 2.92, "learning_rate": 9.736754268555232e-06, "loss": 0.3696, "step": 28997 }, { "epoch": 2.92, "learning_rate": 9.734989577813829e-06, "loss": 0.1298, "step": 28998 }, { "epoch": 2.92, "learning_rate": 9.733225016031246e-06, "loss": 0.8525, "step": 28999 }, { "epoch": 2.92, "learning_rate": 9.731460583218709e-06, "loss": 0.3495, "step": 29000 }, { "epoch": 2.92, "learning_rate": 9.729696279387445e-06, "loss": 0.6153, "step": 29001 }, { "epoch": 2.92, "learning_rate": 9.727932104548683e-06, "loss": 0.5678, "step": 29002 }, { "epoch": 2.92, "learning_rate": 9.72616805871365e-06, "loss": 0.691, "step": 29003 }, { "epoch": 2.92, "learning_rate": 9.724404141893569e-06, "loss": 0.0599, "step": 29004 }, { "epoch": 2.92, "learning_rate": 9.722640354099667e-06, "loss": 0.2453, "step": 29005 }, { "epoch": 2.92, "learning_rate": 9.72087669534317e-06, "loss": 0.0081, "step": 29006 }, { "epoch": 2.92, "learning_rate": 9.719113165635287e-06, "loss": 0.2923, "step": 29007 }, { "epoch": 2.92, "learning_rate": 9.717349764987266e-06, "loss": 0.1295, "step": 29008 }, { "epoch": 2.92, "learning_rate": 9.715586493410323e-06, "loss": 0.0943, "step": 29009 }, { "epoch": 2.92, "learning_rate": 9.713823350915653e-06, "loss": 0.6881, "step": 29010 }, { "epoch": 2.92, "learning_rate": 9.712060337514502e-06, "loss": 0.4542, "step": 29011 }, { "epoch": 2.92, "learning_rate": 9.71029745321808e-06, "loss": 0.6475, "step": 29012 }, { "epoch": 2.92, "learning_rate": 9.708534698037607e-06, "loss": 0.2513, "step": 29013 }, { "epoch": 2.92, "learning_rate": 9.706772071984299e-06, "loss": 0.3031, "step": 29014 }, { "epoch": 2.92, "learning_rate": 9.705009575069378e-06, "loss": 0.0648, "step": 29015 }, { "epoch": 2.92, "learning_rate": 9.70324720730405e-06, "loss": 0.2908, "step": 29016 }, { "epoch": 2.92, "learning_rate": 9.70148496869954e-06, "loss": 0.3773, "step": 29017 }, { "epoch": 2.92, "learning_rate": 9.699722859267056e-06, "loss": 0.104, "step": 29018 }, { "epoch": 2.92, "learning_rate": 9.6979608790178e-06, "loss": 0.24, "step": 29019 }, { "epoch": 2.92, "learning_rate": 9.696199027963022e-06, "loss": 0.8093, "step": 29020 }, { "epoch": 2.92, "learning_rate": 9.694437306113897e-06, "loss": 0.111, "step": 29021 }, { "epoch": 2.92, "learning_rate": 9.69267571348164e-06, "loss": 0.2463, "step": 29022 }, { "epoch": 2.92, "learning_rate": 9.69091425007749e-06, "loss": 0.6285, "step": 29023 }, { "epoch": 2.92, "learning_rate": 9.689152915912622e-06, "loss": 0.2859, "step": 29024 }, { "epoch": 2.92, "learning_rate": 9.68739171099825e-06, "loss": 0.4982, "step": 29025 }, { "epoch": 2.92, "learning_rate": 9.6856306353456e-06, "loss": 0.674, "step": 29026 }, { "epoch": 2.92, "learning_rate": 9.68386968896588e-06, "loss": 0.1007, "step": 29027 }, { "epoch": 2.92, "learning_rate": 9.682108871870266e-06, "loss": 0.1354, "step": 29028 }, { "epoch": 2.92, "learning_rate": 9.68034818406999e-06, "loss": 0.3404, "step": 29029 }, { "epoch": 2.92, "learning_rate": 9.678587625576244e-06, "loss": 0.5523, "step": 29030 }, { "epoch": 2.92, "learning_rate": 9.676827196400241e-06, "loss": 0.5992, "step": 29031 }, { "epoch": 2.92, "learning_rate": 9.675066896553177e-06, "loss": 0.0769, "step": 29032 }, { "epoch": 2.92, "learning_rate": 9.673306726046247e-06, "loss": 0.3317, "step": 29033 }, { "epoch": 2.92, "learning_rate": 9.671546684890678e-06, "loss": 0.4309, "step": 29034 }, { "epoch": 2.92, "learning_rate": 9.66978677309764e-06, "loss": 0.6365, "step": 29035 }, { "epoch": 2.92, "learning_rate": 9.66802699067834e-06, "loss": 0.6003, "step": 29036 }, { "epoch": 2.92, "learning_rate": 9.666267337643996e-06, "loss": 0.5095, "step": 29037 }, { "epoch": 2.92, "learning_rate": 9.664507814005787e-06, "loss": 0.0614, "step": 29038 }, { "epoch": 2.92, "learning_rate": 9.662748419774903e-06, "loss": 0.185, "step": 29039 }, { "epoch": 2.92, "learning_rate": 9.660989154962561e-06, "loss": 0.1622, "step": 29040 }, { "epoch": 2.92, "learning_rate": 9.659230019579955e-06, "loss": 0.5856, "step": 29041 }, { "epoch": 2.92, "learning_rate": 9.657471013638251e-06, "loss": 0.5822, "step": 29042 }, { "epoch": 2.92, "learning_rate": 9.655712137148674e-06, "loss": 0.55, "step": 29043 }, { "epoch": 2.92, "learning_rate": 9.653953390122406e-06, "loss": 0.9057, "step": 29044 }, { "epoch": 2.92, "learning_rate": 9.652194772570639e-06, "loss": 0.3002, "step": 29045 }, { "epoch": 2.92, "learning_rate": 9.650436284504564e-06, "loss": 0.5557, "step": 29046 }, { "epoch": 2.92, "learning_rate": 9.648677925935369e-06, "loss": 0.5568, "step": 29047 }, { "epoch": 2.92, "learning_rate": 9.646919696874247e-06, "loss": 0.3218, "step": 29048 }, { "epoch": 2.92, "learning_rate": 9.645161597332387e-06, "loss": 0.5758, "step": 29049 }, { "epoch": 2.92, "learning_rate": 9.643403627320971e-06, "loss": 0.9011, "step": 29050 }, { "epoch": 2.92, "learning_rate": 9.641645786851183e-06, "loss": 0.2203, "step": 29051 }, { "epoch": 2.92, "learning_rate": 9.639888075934235e-06, "loss": 0.0951, "step": 29052 }, { "epoch": 2.92, "learning_rate": 9.638130494581285e-06, "loss": 0.3095, "step": 29053 }, { "epoch": 2.92, "learning_rate": 9.636373042803515e-06, "loss": 0.3104, "step": 29054 }, { "epoch": 2.92, "learning_rate": 9.634615720612145e-06, "loss": 0.1476, "step": 29055 }, { "epoch": 2.92, "learning_rate": 9.632858528018319e-06, "loss": 0.1415, "step": 29056 }, { "epoch": 2.92, "learning_rate": 9.631101465033224e-06, "loss": 0.365, "step": 29057 }, { "epoch": 2.93, "learning_rate": 9.629344531668061e-06, "loss": 0.3954, "step": 29058 }, { "epoch": 2.93, "learning_rate": 9.627587727934e-06, "loss": 0.5586, "step": 29059 }, { "epoch": 2.93, "learning_rate": 9.625831053842219e-06, "loss": 0.2048, "step": 29060 }, { "epoch": 2.93, "learning_rate": 9.624074509403899e-06, "loss": 0.1318, "step": 29061 }, { "epoch": 2.93, "learning_rate": 9.62231809463022e-06, "loss": 0.6912, "step": 29062 }, { "epoch": 2.93, "learning_rate": 9.620561809532351e-06, "loss": 0.0355, "step": 29063 }, { "epoch": 2.93, "learning_rate": 9.61880565412148e-06, "loss": 0.5984, "step": 29064 }, { "epoch": 2.93, "learning_rate": 9.617049628408763e-06, "loss": 0.3535, "step": 29065 }, { "epoch": 2.93, "learning_rate": 9.615293732405408e-06, "loss": 0.1401, "step": 29066 }, { "epoch": 2.93, "learning_rate": 9.613537966122563e-06, "loss": 0.2453, "step": 29067 }, { "epoch": 2.93, "learning_rate": 9.611782329571393e-06, "loss": 0.8437, "step": 29068 }, { "epoch": 2.93, "learning_rate": 9.610026822763104e-06, "loss": 0.4247, "step": 29069 }, { "epoch": 2.93, "learning_rate": 9.608271445708838e-06, "loss": 0.1102, "step": 29070 }, { "epoch": 2.93, "learning_rate": 9.606516198419767e-06, "loss": 0.2445, "step": 29071 }, { "epoch": 2.93, "learning_rate": 9.60476108090708e-06, "loss": 0.5251, "step": 29072 }, { "epoch": 2.93, "learning_rate": 9.603006093181941e-06, "loss": 0.2366, "step": 29073 }, { "epoch": 2.93, "learning_rate": 9.601251235255495e-06, "loss": 0.1456, "step": 29074 }, { "epoch": 2.93, "learning_rate": 9.599496507138931e-06, "loss": 0.2991, "step": 29075 }, { "epoch": 2.93, "learning_rate": 9.597741908843417e-06, "loss": 0.4076, "step": 29076 }, { "epoch": 2.93, "learning_rate": 9.595987440380109e-06, "loss": 0.5025, "step": 29077 }, { "epoch": 2.93, "learning_rate": 9.594233101760174e-06, "loss": 0.5162, "step": 29078 }, { "epoch": 2.93, "learning_rate": 9.592478892994776e-06, "loss": 0.2102, "step": 29079 }, { "epoch": 2.93, "learning_rate": 9.59072481409508e-06, "loss": 0.4818, "step": 29080 }, { "epoch": 2.93, "learning_rate": 9.588970865072249e-06, "loss": 0.296, "step": 29081 }, { "epoch": 2.93, "learning_rate": 9.58721704593744e-06, "loss": 0.3837, "step": 29082 }, { "epoch": 2.93, "learning_rate": 9.585463356701809e-06, "loss": 0.2137, "step": 29083 }, { "epoch": 2.93, "learning_rate": 9.583709797376543e-06, "loss": 0.0728, "step": 29084 }, { "epoch": 2.93, "learning_rate": 9.58195636797277e-06, "loss": 0.6449, "step": 29085 }, { "epoch": 2.93, "learning_rate": 9.580203068501648e-06, "loss": 1.0797, "step": 29086 }, { "epoch": 2.93, "learning_rate": 9.578449898974367e-06, "loss": 0.4657, "step": 29087 }, { "epoch": 2.93, "learning_rate": 9.57669685940204e-06, "loss": 0.12, "step": 29088 }, { "epoch": 2.93, "learning_rate": 9.574943949795857e-06, "loss": 0.5817, "step": 29089 }, { "epoch": 2.93, "learning_rate": 9.573191170166961e-06, "loss": 0.1168, "step": 29090 }, { "epoch": 2.93, "learning_rate": 9.571438520526503e-06, "loss": 0.3098, "step": 29091 }, { "epoch": 2.93, "learning_rate": 9.569686000885642e-06, "loss": 0.5796, "step": 29092 }, { "epoch": 2.93, "learning_rate": 9.567933611255523e-06, "loss": 0.1619, "step": 29093 }, { "epoch": 2.93, "learning_rate": 9.566181351647305e-06, "loss": 0.6202, "step": 29094 }, { "epoch": 2.93, "learning_rate": 9.564429222072137e-06, "loss": 0.4656, "step": 29095 }, { "epoch": 2.93, "learning_rate": 9.562677222541166e-06, "loss": 0.0878, "step": 29096 }, { "epoch": 2.93, "learning_rate": 9.56092535306553e-06, "loss": 0.1801, "step": 29097 }, { "epoch": 2.93, "learning_rate": 9.559173613656411e-06, "loss": 0.0772, "step": 29098 }, { "epoch": 2.93, "learning_rate": 9.557422004324924e-06, "loss": 0.3809, "step": 29099 }, { "epoch": 2.93, "learning_rate": 9.555670525082221e-06, "loss": 0.1678, "step": 29100 }, { "epoch": 2.93, "learning_rate": 9.553919175939469e-06, "loss": 0.2114, "step": 29101 }, { "epoch": 2.93, "learning_rate": 9.552167956907789e-06, "loss": 0.6967, "step": 29102 }, { "epoch": 2.93, "learning_rate": 9.550416867998325e-06, "loss": 0.3536, "step": 29103 }, { "epoch": 2.93, "learning_rate": 9.548665909222235e-06, "loss": 0.1936, "step": 29104 }, { "epoch": 2.93, "learning_rate": 9.546915080590666e-06, "loss": 0.6787, "step": 29105 }, { "epoch": 2.93, "learning_rate": 9.545164382114731e-06, "loss": 0.2023, "step": 29106 }, { "epoch": 2.93, "learning_rate": 9.543413813805598e-06, "loss": 0.8078, "step": 29107 }, { "epoch": 2.93, "learning_rate": 9.541663375674394e-06, "loss": 0.8341, "step": 29108 }, { "epoch": 2.93, "learning_rate": 9.539913067732265e-06, "loss": 0.1217, "step": 29109 }, { "epoch": 2.93, "learning_rate": 9.538162889990346e-06, "loss": 0.5909, "step": 29110 }, { "epoch": 2.93, "learning_rate": 9.53641284245977e-06, "loss": 0.4049, "step": 29111 }, { "epoch": 2.93, "learning_rate": 9.534662925151682e-06, "loss": 0.1753, "step": 29112 }, { "epoch": 2.93, "learning_rate": 9.532913138077211e-06, "loss": 0.1558, "step": 29113 }, { "epoch": 2.93, "learning_rate": 9.531163481247486e-06, "loss": 0.198, "step": 29114 }, { "epoch": 2.93, "learning_rate": 9.529413954673663e-06, "loss": 0.2386, "step": 29115 }, { "epoch": 2.93, "learning_rate": 9.527664558366867e-06, "loss": 0.2746, "step": 29116 }, { "epoch": 2.93, "learning_rate": 9.525915292338208e-06, "loss": 0.3614, "step": 29117 }, { "epoch": 2.93, "learning_rate": 9.524166156598846e-06, "loss": 0.3113, "step": 29118 }, { "epoch": 2.93, "learning_rate": 9.52241715115991e-06, "loss": 0.5267, "step": 29119 }, { "epoch": 2.93, "learning_rate": 9.520668276032502e-06, "loss": 0.4844, "step": 29120 }, { "epoch": 2.93, "learning_rate": 9.518919531227778e-06, "loss": 0.5561, "step": 29121 }, { "epoch": 2.93, "learning_rate": 9.517170916756861e-06, "loss": 0.8903, "step": 29122 }, { "epoch": 2.93, "learning_rate": 9.515422432630877e-06, "loss": 0.1981, "step": 29123 }, { "epoch": 2.93, "learning_rate": 9.513674078860952e-06, "loss": 0.3949, "step": 29124 }, { "epoch": 2.93, "learning_rate": 9.511925855458209e-06, "loss": 0.0909, "step": 29125 }, { "epoch": 2.93, "learning_rate": 9.510177762433778e-06, "loss": 0.0736, "step": 29126 }, { "epoch": 2.93, "learning_rate": 9.50842979979878e-06, "loss": 0.2271, "step": 29127 }, { "epoch": 2.93, "learning_rate": 9.50668196756434e-06, "loss": 0.1315, "step": 29128 }, { "epoch": 2.93, "learning_rate": 9.50493426574157e-06, "loss": 0.3217, "step": 29129 }, { "epoch": 2.93, "learning_rate": 9.503186694341622e-06, "loss": 0.6046, "step": 29130 }, { "epoch": 2.93, "learning_rate": 9.501439253375586e-06, "loss": 0.1217, "step": 29131 }, { "epoch": 2.93, "learning_rate": 9.49969194285458e-06, "loss": 0.5442, "step": 29132 }, { "epoch": 2.93, "learning_rate": 9.49794476278976e-06, "loss": 0.2254, "step": 29133 }, { "epoch": 2.93, "learning_rate": 9.496197713192208e-06, "loss": 0.6965, "step": 29134 }, { "epoch": 2.93, "learning_rate": 9.494450794073044e-06, "loss": 0.6255, "step": 29135 }, { "epoch": 2.93, "learning_rate": 9.492704005443405e-06, "loss": 0.0824, "step": 29136 }, { "epoch": 2.93, "learning_rate": 9.490957347314403e-06, "loss": 0.2417, "step": 29137 }, { "epoch": 2.93, "learning_rate": 9.48921081969713e-06, "loss": 0.2097, "step": 29138 }, { "epoch": 2.93, "learning_rate": 9.487464422602727e-06, "loss": 0.3767, "step": 29139 }, { "epoch": 2.93, "learning_rate": 9.485718156042296e-06, "loss": 0.293, "step": 29140 }, { "epoch": 2.93, "learning_rate": 9.48397202002695e-06, "loss": 0.1357, "step": 29141 }, { "epoch": 2.93, "learning_rate": 9.482226014567803e-06, "loss": 0.5283, "step": 29142 }, { "epoch": 2.93, "learning_rate": 9.480480139675954e-06, "loss": 0.6024, "step": 29143 }, { "epoch": 2.93, "learning_rate": 9.478734395362538e-06, "loss": 0.3516, "step": 29144 }, { "epoch": 2.93, "learning_rate": 9.476988781638646e-06, "loss": 0.1747, "step": 29145 }, { "epoch": 2.93, "learning_rate": 9.475243298515378e-06, "loss": 0.0712, "step": 29146 }, { "epoch": 2.93, "learning_rate": 9.473497946003863e-06, "loss": 0.131, "step": 29147 }, { "epoch": 2.93, "learning_rate": 9.47175272411521e-06, "loss": 0.2964, "step": 29148 }, { "epoch": 2.93, "learning_rate": 9.470007632860494e-06, "loss": 0.1626, "step": 29149 }, { "epoch": 2.93, "learning_rate": 9.468262672250848e-06, "loss": 0.2858, "step": 29150 }, { "epoch": 2.93, "learning_rate": 9.466517842297377e-06, "loss": 0.0402, "step": 29151 }, { "epoch": 2.93, "learning_rate": 9.464773143011153e-06, "loss": 0.5598, "step": 29152 }, { "epoch": 2.93, "learning_rate": 9.463028574403313e-06, "loss": 0.3801, "step": 29153 }, { "epoch": 2.93, "learning_rate": 9.461284136484948e-06, "loss": 0.1182, "step": 29154 }, { "epoch": 2.93, "learning_rate": 9.459539829267156e-06, "loss": 0.2193, "step": 29155 }, { "epoch": 2.94, "learning_rate": 9.45779565276104e-06, "loss": 0.4282, "step": 29156 }, { "epoch": 2.94, "learning_rate": 9.456051606977696e-06, "loss": 0.3666, "step": 29157 }, { "epoch": 2.94, "learning_rate": 9.454307691928224e-06, "loss": 0.0041, "step": 29158 }, { "epoch": 2.94, "learning_rate": 9.452563907623725e-06, "loss": 0.1836, "step": 29159 }, { "epoch": 2.94, "learning_rate": 9.450820254075287e-06, "loss": 0.1726, "step": 29160 }, { "epoch": 2.94, "learning_rate": 9.449076731294008e-06, "loss": 0.792, "step": 29161 }, { "epoch": 2.94, "learning_rate": 9.447333339291002e-06, "loss": 0.1984, "step": 29162 }, { "epoch": 2.94, "learning_rate": 9.44559007807734e-06, "loss": 0.129, "step": 29163 }, { "epoch": 2.94, "learning_rate": 9.443846947664114e-06, "loss": 0.7509, "step": 29164 }, { "epoch": 2.94, "learning_rate": 9.442103948062443e-06, "loss": 0.2949, "step": 29165 }, { "epoch": 2.94, "learning_rate": 9.440361079283395e-06, "loss": 0.444, "step": 29166 }, { "epoch": 2.94, "learning_rate": 9.43861834133806e-06, "loss": 0.4461, "step": 29167 }, { "epoch": 2.94, "learning_rate": 9.43687573423754e-06, "loss": 0.2295, "step": 29168 }, { "epoch": 2.94, "learning_rate": 9.435133257992925e-06, "loss": 0.4796, "step": 29169 }, { "epoch": 2.94, "learning_rate": 9.433390912615294e-06, "loss": 0.606, "step": 29170 }, { "epoch": 2.94, "learning_rate": 9.431648698115745e-06, "loss": 0.2339, "step": 29171 }, { "epoch": 2.94, "learning_rate": 9.429906614505353e-06, "loss": 0.0775, "step": 29172 }, { "epoch": 2.94, "learning_rate": 9.428164661795213e-06, "loss": 0.3631, "step": 29173 }, { "epoch": 2.94, "learning_rate": 9.426422839996408e-06, "loss": 0.0715, "step": 29174 }, { "epoch": 2.94, "learning_rate": 9.424681149120013e-06, "loss": 0.5564, "step": 29175 }, { "epoch": 2.94, "learning_rate": 9.422939589177134e-06, "loss": 0.6351, "step": 29176 }, { "epoch": 2.94, "learning_rate": 9.421198160178835e-06, "loss": 0.1782, "step": 29177 }, { "epoch": 2.94, "learning_rate": 9.419456862136192e-06, "loss": 0.1426, "step": 29178 }, { "epoch": 2.94, "learning_rate": 9.417715695060306e-06, "loss": 0.2986, "step": 29179 }, { "epoch": 2.94, "learning_rate": 9.415974658962254e-06, "loss": 0.8474, "step": 29180 }, { "epoch": 2.94, "learning_rate": 9.41423375385309e-06, "loss": 0.2526, "step": 29181 }, { "epoch": 2.94, "learning_rate": 9.412492979743923e-06, "loss": 0.1781, "step": 29182 }, { "epoch": 2.94, "learning_rate": 9.410752336645828e-06, "loss": 0.8389, "step": 29183 }, { "epoch": 2.94, "learning_rate": 9.409011824569853e-06, "loss": 0.5173, "step": 29184 }, { "epoch": 2.94, "learning_rate": 9.407271443527103e-06, "loss": 0.1689, "step": 29185 }, { "epoch": 2.94, "learning_rate": 9.405531193528644e-06, "loss": 0.5614, "step": 29186 }, { "epoch": 2.94, "learning_rate": 9.403791074585549e-06, "loss": 0.4303, "step": 29187 }, { "epoch": 2.94, "learning_rate": 9.402051086708891e-06, "loss": 0.5522, "step": 29188 }, { "epoch": 2.94, "learning_rate": 9.400311229909749e-06, "loss": 0.111, "step": 29189 }, { "epoch": 2.94, "learning_rate": 9.398571504199187e-06, "loss": 0.171, "step": 29190 }, { "epoch": 2.94, "learning_rate": 9.396831909588275e-06, "loss": 0.5321, "step": 29191 }, { "epoch": 2.94, "learning_rate": 9.395092446088093e-06, "loss": 0.5945, "step": 29192 }, { "epoch": 2.94, "learning_rate": 9.393353113709691e-06, "loss": 0.3682, "step": 29193 }, { "epoch": 2.94, "learning_rate": 9.39161391246417e-06, "loss": 0.7152, "step": 29194 }, { "epoch": 2.94, "learning_rate": 9.389874842362562e-06, "loss": 0.1004, "step": 29195 }, { "epoch": 2.94, "learning_rate": 9.388135903415956e-06, "loss": 0.3144, "step": 29196 }, { "epoch": 2.94, "learning_rate": 9.386397095635423e-06, "loss": 0.4831, "step": 29197 }, { "epoch": 2.94, "learning_rate": 9.384658419031999e-06, "loss": 0.5196, "step": 29198 }, { "epoch": 2.94, "learning_rate": 9.382919873616775e-06, "loss": 0.371, "step": 29199 }, { "epoch": 2.94, "learning_rate": 9.381181459400805e-06, "loss": 0.5295, "step": 29200 }, { "epoch": 2.94, "learning_rate": 9.379443176395151e-06, "loss": 0.2511, "step": 29201 }, { "epoch": 2.94, "learning_rate": 9.377705024610878e-06, "loss": 0.5436, "step": 29202 }, { "epoch": 2.94, "learning_rate": 9.375967004059047e-06, "loss": 0.1259, "step": 29203 }, { "epoch": 2.94, "learning_rate": 9.374229114750715e-06, "loss": 0.1112, "step": 29204 }, { "epoch": 2.94, "learning_rate": 9.37249135669694e-06, "loss": 0.7169, "step": 29205 }, { "epoch": 2.94, "learning_rate": 9.370753729908784e-06, "loss": 0.3133, "step": 29206 }, { "epoch": 2.94, "learning_rate": 9.369016234397298e-06, "loss": 0.4089, "step": 29207 }, { "epoch": 2.94, "learning_rate": 9.367278870173558e-06, "loss": 0.7319, "step": 29208 }, { "epoch": 2.94, "learning_rate": 9.3655416372486e-06, "loss": 0.2163, "step": 29209 }, { "epoch": 2.94, "learning_rate": 9.363804535633472e-06, "loss": 0.1639, "step": 29210 }, { "epoch": 2.94, "learning_rate": 9.362067565339255e-06, "loss": 0.14, "step": 29211 }, { "epoch": 2.94, "learning_rate": 9.360330726376996e-06, "loss": 0.0954, "step": 29212 }, { "epoch": 2.94, "learning_rate": 9.358594018757725e-06, "loss": 0.0982, "step": 29213 }, { "epoch": 2.94, "learning_rate": 9.356857442492516e-06, "loss": 0.2524, "step": 29214 }, { "epoch": 2.94, "learning_rate": 9.355120997592421e-06, "loss": 0.2645, "step": 29215 }, { "epoch": 2.94, "learning_rate": 9.353384684068466e-06, "loss": 0.2498, "step": 29216 }, { "epoch": 2.94, "learning_rate": 9.351648501931724e-06, "loss": 0.4825, "step": 29217 }, { "epoch": 2.94, "learning_rate": 9.349912451193239e-06, "loss": 0.1152, "step": 29218 }, { "epoch": 2.94, "learning_rate": 9.348176531864053e-06, "loss": 0.1912, "step": 29219 }, { "epoch": 2.94, "learning_rate": 9.346440743955216e-06, "loss": 0.6651, "step": 29220 }, { "epoch": 2.94, "learning_rate": 9.344705087477774e-06, "loss": 0.171, "step": 29221 }, { "epoch": 2.94, "learning_rate": 9.34296956244277e-06, "loss": 0.2257, "step": 29222 }, { "epoch": 2.94, "learning_rate": 9.341234168861249e-06, "loss": 0.0214, "step": 29223 }, { "epoch": 2.94, "learning_rate": 9.339498906744246e-06, "loss": 0.8508, "step": 29224 }, { "epoch": 2.94, "learning_rate": 9.337763776102821e-06, "loss": 0.0819, "step": 29225 }, { "epoch": 2.94, "learning_rate": 9.336028776948016e-06, "loss": 0.0795, "step": 29226 }, { "epoch": 2.94, "learning_rate": 9.334293909290847e-06, "loss": 0.0646, "step": 29227 }, { "epoch": 2.94, "learning_rate": 9.332559173142378e-06, "loss": 0.1806, "step": 29228 }, { "epoch": 2.94, "learning_rate": 9.330824568513647e-06, "loss": 1.0867, "step": 29229 }, { "epoch": 2.94, "learning_rate": 9.32909009541567e-06, "loss": 0.0204, "step": 29230 }, { "epoch": 2.94, "learning_rate": 9.327355753859506e-06, "loss": 0.5477, "step": 29231 }, { "epoch": 2.94, "learning_rate": 9.32562154385619e-06, "loss": 0.1165, "step": 29232 }, { "epoch": 2.94, "learning_rate": 9.32388746541675e-06, "loss": 0.1357, "step": 29233 }, { "epoch": 2.94, "learning_rate": 9.322153518552224e-06, "loss": 1.0463, "step": 29234 }, { "epoch": 2.94, "learning_rate": 9.320419703273648e-06, "loss": 0.7198, "step": 29235 }, { "epoch": 2.94, "learning_rate": 9.318686019592056e-06, "loss": 0.1625, "step": 29236 }, { "epoch": 2.94, "learning_rate": 9.316952467518476e-06, "loss": 0.6512, "step": 29237 }, { "epoch": 2.94, "learning_rate": 9.315219047063945e-06, "loss": 0.4693, "step": 29238 }, { "epoch": 2.94, "learning_rate": 9.31348575823948e-06, "loss": 0.5598, "step": 29239 }, { "epoch": 2.94, "learning_rate": 9.31175260105614e-06, "loss": 0.4003, "step": 29240 }, { "epoch": 2.94, "learning_rate": 9.310019575524928e-06, "loss": 0.4463, "step": 29241 }, { "epoch": 2.94, "learning_rate": 9.308286681656871e-06, "loss": 0.2223, "step": 29242 }, { "epoch": 2.94, "learning_rate": 9.306553919463016e-06, "loss": 0.5929, "step": 29243 }, { "epoch": 2.94, "learning_rate": 9.304821288954389e-06, "loss": 0.6283, "step": 29244 }, { "epoch": 2.94, "learning_rate": 9.303088790141987e-06, "loss": 0.3468, "step": 29245 }, { "epoch": 2.94, "learning_rate": 9.301356423036863e-06, "loss": 0.1466, "step": 29246 }, { "epoch": 2.94, "learning_rate": 9.299624187650045e-06, "loss": 0.6194, "step": 29247 }, { "epoch": 2.94, "learning_rate": 9.297892083992522e-06, "loss": 0.8032, "step": 29248 }, { "epoch": 2.94, "learning_rate": 9.296160112075347e-06, "loss": 0.2602, "step": 29249 }, { "epoch": 2.94, "learning_rate": 9.294428271909535e-06, "loss": 0.2287, "step": 29250 }, { "epoch": 2.94, "learning_rate": 9.292696563506104e-06, "loss": 0.1751, "step": 29251 }, { "epoch": 2.94, "learning_rate": 9.290964986876075e-06, "loss": 0.1786, "step": 29252 }, { "epoch": 2.94, "learning_rate": 9.289233542030458e-06, "loss": 0.3607, "step": 29253 }, { "epoch": 2.94, "learning_rate": 9.2875022289803e-06, "loss": 0.4572, "step": 29254 }, { "epoch": 2.95, "learning_rate": 9.285771047736586e-06, "loss": 0.3317, "step": 29255 }, { "epoch": 2.95, "learning_rate": 9.284039998310334e-06, "loss": 0.2112, "step": 29256 }, { "epoch": 2.95, "learning_rate": 9.282309080712584e-06, "loss": 0.1564, "step": 29257 }, { "epoch": 2.95, "learning_rate": 9.280578294954347e-06, "loss": 0.4753, "step": 29258 }, { "epoch": 2.95, "learning_rate": 9.278847641046607e-06, "loss": 0.3218, "step": 29259 }, { "epoch": 2.95, "learning_rate": 9.277117119000405e-06, "loss": 0.3369, "step": 29260 }, { "epoch": 2.95, "learning_rate": 9.27538672882676e-06, "loss": 0.3067, "step": 29261 }, { "epoch": 2.95, "learning_rate": 9.273656470536648e-06, "loss": 0.4829, "step": 29262 }, { "epoch": 2.95, "learning_rate": 9.271926344141112e-06, "loss": 0.2423, "step": 29263 }, { "epoch": 2.95, "learning_rate": 9.270196349651154e-06, "loss": 0.3581, "step": 29264 }, { "epoch": 2.95, "learning_rate": 9.268466487077778e-06, "loss": 0.2856, "step": 29265 }, { "epoch": 2.95, "learning_rate": 9.266736756431995e-06, "loss": 0.3224, "step": 29266 }, { "epoch": 2.95, "learning_rate": 9.265007157724812e-06, "loss": 0.1028, "step": 29267 }, { "epoch": 2.95, "learning_rate": 9.263277690967234e-06, "loss": 0.2866, "step": 29268 }, { "epoch": 2.95, "learning_rate": 9.26154835617027e-06, "loss": 0.2805, "step": 29269 }, { "epoch": 2.95, "learning_rate": 9.259819153344921e-06, "loss": 0.4967, "step": 29270 }, { "epoch": 2.95, "learning_rate": 9.258090082502184e-06, "loss": 0.1961, "step": 29271 }, { "epoch": 2.95, "learning_rate": 9.256361143653093e-06, "loss": 0.0986, "step": 29272 }, { "epoch": 2.95, "learning_rate": 9.254632336808615e-06, "loss": 0.275, "step": 29273 }, { "epoch": 2.95, "learning_rate": 9.25290366197976e-06, "loss": 0.2512, "step": 29274 }, { "epoch": 2.95, "learning_rate": 9.25117511917754e-06, "loss": 0.1572, "step": 29275 }, { "epoch": 2.95, "learning_rate": 9.249446708412947e-06, "loss": 0.1361, "step": 29276 }, { "epoch": 2.95, "learning_rate": 9.24771842969698e-06, "loss": 0.0918, "step": 29277 }, { "epoch": 2.95, "learning_rate": 9.245990283040644e-06, "loss": 0.042, "step": 29278 }, { "epoch": 2.95, "learning_rate": 9.244262268454924e-06, "loss": 0.1927, "step": 29279 }, { "epoch": 2.95, "learning_rate": 9.242534385950828e-06, "loss": 0.2456, "step": 29280 }, { "epoch": 2.95, "learning_rate": 9.240806635539345e-06, "loss": 0.604, "step": 29281 }, { "epoch": 2.95, "learning_rate": 9.239079017231473e-06, "loss": 0.2397, "step": 29282 }, { "epoch": 2.95, "learning_rate": 9.2373515310382e-06, "loss": 0.1756, "step": 29283 }, { "epoch": 2.95, "learning_rate": 9.235624176970528e-06, "loss": 0.2654, "step": 29284 }, { "epoch": 2.95, "learning_rate": 9.233896955039435e-06, "loss": 0.1231, "step": 29285 }, { "epoch": 2.95, "learning_rate": 9.232169865255938e-06, "loss": 0.7422, "step": 29286 }, { "epoch": 2.95, "learning_rate": 9.230442907631004e-06, "loss": 0.4119, "step": 29287 }, { "epoch": 2.95, "learning_rate": 9.228716082175619e-06, "loss": 0.5022, "step": 29288 }, { "epoch": 2.95, "learning_rate": 9.226989388900793e-06, "loss": 0.227, "step": 29289 }, { "epoch": 2.95, "learning_rate": 9.225262827817513e-06, "loss": 0.3999, "step": 29290 }, { "epoch": 2.95, "learning_rate": 9.223536398936739e-06, "loss": 0.1907, "step": 29291 }, { "epoch": 2.95, "learning_rate": 9.221810102269487e-06, "loss": 0.6021, "step": 29292 }, { "epoch": 2.95, "learning_rate": 9.220083937826736e-06, "loss": 0.2415, "step": 29293 }, { "epoch": 2.95, "learning_rate": 9.218357905619452e-06, "loss": 0.5654, "step": 29294 }, { "epoch": 2.95, "learning_rate": 9.21663200565864e-06, "loss": 0.285, "step": 29295 }, { "epoch": 2.95, "learning_rate": 9.214906237955272e-06, "loss": 0.2736, "step": 29296 }, { "epoch": 2.95, "learning_rate": 9.21318060252034e-06, "loss": 0.6197, "step": 29297 }, { "epoch": 2.95, "learning_rate": 9.211455099364815e-06, "loss": 0.012, "step": 29298 }, { "epoch": 2.95, "learning_rate": 9.209729728499682e-06, "loss": 0.9095, "step": 29299 }, { "epoch": 2.95, "learning_rate": 9.208004489935922e-06, "loss": 0.7481, "step": 29300 }, { "epoch": 2.95, "learning_rate": 9.20627938368451e-06, "loss": 0.3838, "step": 29301 }, { "epoch": 2.95, "learning_rate": 9.204554409756422e-06, "loss": 0.2807, "step": 29302 }, { "epoch": 2.95, "learning_rate": 9.202829568162634e-06, "loss": 0.6589, "step": 29303 }, { "epoch": 2.95, "learning_rate": 9.201104858914145e-06, "loss": 0.3301, "step": 29304 }, { "epoch": 2.95, "learning_rate": 9.199380282021892e-06, "loss": 0.0702, "step": 29305 }, { "epoch": 2.95, "learning_rate": 9.19765583749688e-06, "loss": 0.2727, "step": 29306 }, { "epoch": 2.95, "learning_rate": 9.195931525350074e-06, "loss": 0.4902, "step": 29307 }, { "epoch": 2.95, "learning_rate": 9.194207345592446e-06, "loss": 0.0863, "step": 29308 }, { "epoch": 2.95, "learning_rate": 9.192483298234967e-06, "loss": 0.0783, "step": 29309 }, { "epoch": 2.95, "learning_rate": 9.190759383288605e-06, "loss": 0.1252, "step": 29310 }, { "epoch": 2.95, "learning_rate": 9.189035600764337e-06, "loss": 0.5082, "step": 29311 }, { "epoch": 2.95, "learning_rate": 9.18731195067313e-06, "loss": 0.1725, "step": 29312 }, { "epoch": 2.95, "learning_rate": 9.185588433025953e-06, "loss": 0.1268, "step": 29313 }, { "epoch": 2.95, "learning_rate": 9.183865047833773e-06, "loss": 0.7132, "step": 29314 }, { "epoch": 2.95, "learning_rate": 9.182141795107554e-06, "loss": 0.259, "step": 29315 }, { "epoch": 2.95, "learning_rate": 9.180418674858268e-06, "loss": 0.3153, "step": 29316 }, { "epoch": 2.95, "learning_rate": 9.178695687096866e-06, "loss": 0.4125, "step": 29317 }, { "epoch": 2.95, "learning_rate": 9.17697283183434e-06, "loss": 1.0957, "step": 29318 }, { "epoch": 2.95, "learning_rate": 9.175250109081635e-06, "loss": 1.6287, "step": 29319 }, { "epoch": 2.95, "learning_rate": 9.173527518849702e-06, "loss": 0.3694, "step": 29320 }, { "epoch": 2.95, "learning_rate": 9.171805061149527e-06, "loss": 0.3972, "step": 29321 }, { "epoch": 2.95, "learning_rate": 9.170082735992074e-06, "loss": 0.0324, "step": 29322 }, { "epoch": 2.95, "learning_rate": 9.168360543388267e-06, "loss": 0.1338, "step": 29323 }, { "epoch": 2.95, "learning_rate": 9.166638483349097e-06, "loss": 0.6257, "step": 29324 }, { "epoch": 2.95, "learning_rate": 9.164916555885529e-06, "loss": 0.318, "step": 29325 }, { "epoch": 2.95, "learning_rate": 9.163194761008484e-06, "loss": 0.2248, "step": 29326 }, { "epoch": 2.95, "learning_rate": 9.161473098728952e-06, "loss": 0.1434, "step": 29327 }, { "epoch": 2.95, "learning_rate": 9.159751569057875e-06, "loss": 0.1744, "step": 29328 }, { "epoch": 2.95, "learning_rate": 9.158030172006215e-06, "loss": 0.2921, "step": 29329 }, { "epoch": 2.95, "learning_rate": 9.15630890758492e-06, "loss": 0.3646, "step": 29330 }, { "epoch": 2.95, "learning_rate": 9.154587775804935e-06, "loss": 0.2376, "step": 29331 }, { "epoch": 2.95, "learning_rate": 9.152866776677244e-06, "loss": 0.064, "step": 29332 }, { "epoch": 2.95, "learning_rate": 9.151145910212766e-06, "loss": 0.6589, "step": 29333 }, { "epoch": 2.95, "learning_rate": 9.149425176422458e-06, "loss": 0.255, "step": 29334 }, { "epoch": 2.95, "learning_rate": 9.147704575317285e-06, "loss": 0.2883, "step": 29335 }, { "epoch": 2.95, "learning_rate": 9.145984106908195e-06, "loss": 0.0697, "step": 29336 }, { "epoch": 2.95, "learning_rate": 9.144263771206113e-06, "loss": 0.4954, "step": 29337 }, { "epoch": 2.95, "learning_rate": 9.142543568222012e-06, "loss": 0.2403, "step": 29338 }, { "epoch": 2.95, "learning_rate": 9.140823497966828e-06, "loss": 0.5692, "step": 29339 }, { "epoch": 2.95, "learning_rate": 9.139103560451508e-06, "loss": 0.0568, "step": 29340 }, { "epoch": 2.95, "learning_rate": 9.137383755686995e-06, "loss": 0.3937, "step": 29341 }, { "epoch": 2.95, "learning_rate": 9.135664083684236e-06, "loss": 0.5439, "step": 29342 }, { "epoch": 2.95, "learning_rate": 9.133944544454175e-06, "loss": 0.5193, "step": 29343 }, { "epoch": 2.95, "learning_rate": 9.132225138007755e-06, "loss": 0.0888, "step": 29344 }, { "epoch": 2.95, "learning_rate": 9.130505864355912e-06, "loss": 0.0886, "step": 29345 }, { "epoch": 2.95, "learning_rate": 9.128786723509592e-06, "loss": 0.9125, "step": 29346 }, { "epoch": 2.95, "learning_rate": 9.12706771547973e-06, "loss": 0.3786, "step": 29347 }, { "epoch": 2.95, "learning_rate": 9.125348840277275e-06, "loss": 0.9229, "step": 29348 }, { "epoch": 2.95, "learning_rate": 9.123630097913143e-06, "loss": 0.1431, "step": 29349 }, { "epoch": 2.95, "learning_rate": 9.12191148839831e-06, "loss": 0.0929, "step": 29350 }, { "epoch": 2.95, "learning_rate": 9.120193011743682e-06, "loss": 0.3168, "step": 29351 }, { "epoch": 2.95, "learning_rate": 9.118474667960189e-06, "loss": 0.0398, "step": 29352 }, { "epoch": 2.96, "learning_rate": 9.116756457058792e-06, "loss": 0.2051, "step": 29353 }, { "epoch": 2.96, "learning_rate": 9.11503837905042e-06, "loss": 0.2076, "step": 29354 }, { "epoch": 2.96, "learning_rate": 9.113320433945983e-06, "loss": 0.7482, "step": 29355 }, { "epoch": 2.96, "learning_rate": 9.111602621756435e-06, "loss": 0.294, "step": 29356 }, { "epoch": 2.96, "learning_rate": 9.109884942492702e-06, "loss": 0.2963, "step": 29357 }, { "epoch": 2.96, "learning_rate": 9.108167396165715e-06, "loss": 0.1615, "step": 29358 }, { "epoch": 2.96, "learning_rate": 9.106449982786401e-06, "loss": 0.3386, "step": 29359 }, { "epoch": 2.96, "learning_rate": 9.104732702365697e-06, "loss": 0.3177, "step": 29360 }, { "epoch": 2.96, "learning_rate": 9.10301555491452e-06, "loss": 0.3824, "step": 29361 }, { "epoch": 2.96, "learning_rate": 9.101298540443804e-06, "loss": 0.249, "step": 29362 }, { "epoch": 2.96, "learning_rate": 9.099581658964465e-06, "loss": 0.1492, "step": 29363 }, { "epoch": 2.96, "learning_rate": 9.097864910487454e-06, "loss": 0.2981, "step": 29364 }, { "epoch": 2.96, "learning_rate": 9.096148295023668e-06, "loss": 0.1655, "step": 29365 }, { "epoch": 2.96, "learning_rate": 9.094431812584038e-06, "loss": 0.4568, "step": 29366 }, { "epoch": 2.96, "learning_rate": 9.092715463179496e-06, "loss": 0.3368, "step": 29367 }, { "epoch": 2.96, "learning_rate": 9.090999246820973e-06, "loss": 0.4507, "step": 29368 }, { "epoch": 2.96, "learning_rate": 9.089283163519354e-06, "loss": 0.2959, "step": 29369 }, { "epoch": 2.96, "learning_rate": 9.087567213285592e-06, "loss": 0.3445, "step": 29370 }, { "epoch": 2.96, "learning_rate": 9.085851396130599e-06, "loss": 0.2932, "step": 29371 }, { "epoch": 2.96, "learning_rate": 9.084135712065292e-06, "loss": 0.4324, "step": 29372 }, { "epoch": 2.96, "learning_rate": 9.082420161100584e-06, "loss": 0.5202, "step": 29373 }, { "epoch": 2.96, "learning_rate": 9.080704743247398e-06, "loss": 0.2047, "step": 29374 }, { "epoch": 2.96, "learning_rate": 9.078989458516653e-06, "loss": 0.0929, "step": 29375 }, { "epoch": 2.96, "learning_rate": 9.077274306919255e-06, "loss": 0.4782, "step": 29376 }, { "epoch": 2.96, "learning_rate": 9.075559288466126e-06, "loss": 0.2665, "step": 29377 }, { "epoch": 2.96, "learning_rate": 9.07384440316818e-06, "loss": 1.2662, "step": 29378 }, { "epoch": 2.96, "learning_rate": 9.072129651036327e-06, "loss": 0.8375, "step": 29379 }, { "epoch": 2.96, "learning_rate": 9.070415032081476e-06, "loss": 0.3286, "step": 29380 }, { "epoch": 2.96, "learning_rate": 9.068700546314535e-06, "loss": 0.2542, "step": 29381 }, { "epoch": 2.96, "learning_rate": 9.06698619374644e-06, "loss": 0.7729, "step": 29382 }, { "epoch": 2.96, "learning_rate": 9.065271974388071e-06, "loss": 0.0685, "step": 29383 }, { "epoch": 2.96, "learning_rate": 9.063557888250338e-06, "loss": 0.3242, "step": 29384 }, { "epoch": 2.96, "learning_rate": 9.061843935344166e-06, "loss": 0.2793, "step": 29385 }, { "epoch": 2.96, "learning_rate": 9.060130115680459e-06, "loss": 0.096, "step": 29386 }, { "epoch": 2.96, "learning_rate": 9.058416429270113e-06, "loss": 0.8662, "step": 29387 }, { "epoch": 2.96, "learning_rate": 9.05670287612404e-06, "loss": 0.821, "step": 29388 }, { "epoch": 2.96, "learning_rate": 9.054989456253144e-06, "loss": 0.4373, "step": 29389 }, { "epoch": 2.96, "learning_rate": 9.053276169668324e-06, "loss": 0.1583, "step": 29390 }, { "epoch": 2.96, "learning_rate": 9.051563016380486e-06, "loss": 0.2866, "step": 29391 }, { "epoch": 2.96, "learning_rate": 9.049849996400535e-06, "loss": 0.7403, "step": 29392 }, { "epoch": 2.96, "learning_rate": 9.048137109739368e-06, "loss": 0.1728, "step": 29393 }, { "epoch": 2.96, "learning_rate": 9.046424356407884e-06, "loss": 0.3713, "step": 29394 }, { "epoch": 2.96, "learning_rate": 9.044711736416977e-06, "loss": 0.4492, "step": 29395 }, { "epoch": 2.96, "learning_rate": 9.04299924977757e-06, "loss": 0.4865, "step": 29396 }, { "epoch": 2.96, "learning_rate": 9.041286896500533e-06, "loss": 0.2076, "step": 29397 }, { "epoch": 2.96, "learning_rate": 9.039574676596767e-06, "loss": 0.099, "step": 29398 }, { "epoch": 2.96, "learning_rate": 9.037862590077181e-06, "loss": 0.3133, "step": 29399 }, { "epoch": 2.96, "learning_rate": 9.036150636952672e-06, "loss": 0.1998, "step": 29400 }, { "epoch": 2.96, "learning_rate": 9.034438817234108e-06, "loss": 0.8407, "step": 29401 }, { "epoch": 2.96, "learning_rate": 9.032727130932406e-06, "loss": 0.6325, "step": 29402 }, { "epoch": 2.96, "learning_rate": 9.031015578058455e-06, "loss": 0.456, "step": 29403 }, { "epoch": 2.96, "learning_rate": 9.029304158623143e-06, "loss": 0.7755, "step": 29404 }, { "epoch": 2.96, "learning_rate": 9.027592872637362e-06, "loss": 0.1158, "step": 29405 }, { "epoch": 2.96, "learning_rate": 9.025881720112005e-06, "loss": 0.2858, "step": 29406 }, { "epoch": 2.96, "learning_rate": 9.024170701057954e-06, "loss": 0.2747, "step": 29407 }, { "epoch": 2.96, "learning_rate": 9.0224598154861e-06, "loss": 0.5599, "step": 29408 }, { "epoch": 2.96, "learning_rate": 9.020749063407334e-06, "loss": 0.1349, "step": 29409 }, { "epoch": 2.96, "learning_rate": 9.01903844483254e-06, "loss": 0.4674, "step": 29410 }, { "epoch": 2.96, "learning_rate": 9.017327959772602e-06, "loss": 0.6413, "step": 29411 }, { "epoch": 2.96, "learning_rate": 9.015617608238398e-06, "loss": 0.2153, "step": 29412 }, { "epoch": 2.96, "learning_rate": 9.01390739024083e-06, "loss": 0.1507, "step": 29413 }, { "epoch": 2.96, "learning_rate": 9.012197305790783e-06, "loss": 0.5442, "step": 29414 }, { "epoch": 2.96, "learning_rate": 9.010487354899108e-06, "loss": 0.5988, "step": 29415 }, { "epoch": 2.96, "learning_rate": 9.008777537576715e-06, "loss": 0.6886, "step": 29416 }, { "epoch": 2.96, "learning_rate": 9.007067853834475e-06, "loss": 0.2392, "step": 29417 }, { "epoch": 2.96, "learning_rate": 9.00535830368327e-06, "loss": 0.3104, "step": 29418 }, { "epoch": 2.96, "learning_rate": 9.003648887133983e-06, "loss": 0.5372, "step": 29419 }, { "epoch": 2.96, "learning_rate": 9.001939604197483e-06, "loss": 0.5722, "step": 29420 }, { "epoch": 2.96, "learning_rate": 9.000230454884651e-06, "loss": 0.1573, "step": 29421 }, { "epoch": 2.96, "learning_rate": 8.998521439206366e-06, "loss": 0.4313, "step": 29422 }, { "epoch": 2.96, "learning_rate": 8.996812557173498e-06, "loss": 0.7117, "step": 29423 }, { "epoch": 2.96, "learning_rate": 8.995103808796924e-06, "loss": 0.3788, "step": 29424 }, { "epoch": 2.96, "learning_rate": 8.993395194087524e-06, "loss": 0.4561, "step": 29425 }, { "epoch": 2.96, "learning_rate": 8.991686713056163e-06, "loss": 0.2741, "step": 29426 }, { "epoch": 2.96, "learning_rate": 8.989978365713704e-06, "loss": 0.3151, "step": 29427 }, { "epoch": 2.96, "learning_rate": 8.98827015207105e-06, "loss": 0.1366, "step": 29428 }, { "epoch": 2.96, "learning_rate": 8.986562072139046e-06, "loss": 0.2296, "step": 29429 }, { "epoch": 2.96, "learning_rate": 8.984854125928551e-06, "loss": 0.0841, "step": 29430 }, { "epoch": 2.96, "learning_rate": 8.983146313450466e-06, "loss": 0.0308, "step": 29431 }, { "epoch": 2.96, "learning_rate": 8.981438634715646e-06, "loss": 0.4412, "step": 29432 }, { "epoch": 2.96, "learning_rate": 8.979731089734941e-06, "loss": 0.4327, "step": 29433 }, { "epoch": 2.96, "learning_rate": 8.978023678519236e-06, "loss": 0.8488, "step": 29434 }, { "epoch": 2.96, "learning_rate": 8.976316401079394e-06, "loss": 0.2976, "step": 29435 }, { "epoch": 2.96, "learning_rate": 8.974609257426277e-06, "loss": 0.1727, "step": 29436 }, { "epoch": 2.96, "learning_rate": 8.972902247570746e-06, "loss": 0.3643, "step": 29437 }, { "epoch": 2.96, "learning_rate": 8.971195371523668e-06, "loss": 0.6326, "step": 29438 }, { "epoch": 2.96, "learning_rate": 8.9694886292959e-06, "loss": 0.4146, "step": 29439 }, { "epoch": 2.96, "learning_rate": 8.967782020898307e-06, "loss": 0.1845, "step": 29440 }, { "epoch": 2.96, "learning_rate": 8.966075546341742e-06, "loss": 0.213, "step": 29441 }, { "epoch": 2.96, "learning_rate": 8.964369205637086e-06, "loss": 0.2316, "step": 29442 }, { "epoch": 2.96, "learning_rate": 8.962662998795174e-06, "loss": 0.1559, "step": 29443 }, { "epoch": 2.96, "learning_rate": 8.960956925826861e-06, "loss": 0.1269, "step": 29444 }, { "epoch": 2.96, "learning_rate": 8.959250986743025e-06, "loss": 0.1022, "step": 29445 }, { "epoch": 2.96, "learning_rate": 8.957545181554517e-06, "loss": 0.167, "step": 29446 }, { "epoch": 2.96, "learning_rate": 8.955839510272173e-06, "loss": 0.2178, "step": 29447 }, { "epoch": 2.96, "learning_rate": 8.954133972906868e-06, "loss": 0.5741, "step": 29448 }, { "epoch": 2.96, "learning_rate": 8.95242856946945e-06, "loss": 0.2983, "step": 29449 }, { "epoch": 2.96, "learning_rate": 8.950723299970767e-06, "loss": 0.6939, "step": 29450 }, { "epoch": 2.96, "learning_rate": 8.949018164421675e-06, "loss": 0.3717, "step": 29451 }, { "epoch": 2.97, "learning_rate": 8.94731316283302e-06, "loss": 0.6932, "step": 29452 }, { "epoch": 2.97, "learning_rate": 8.94560829521566e-06, "loss": 0.2507, "step": 29453 }, { "epoch": 2.97, "learning_rate": 8.943903561580434e-06, "loss": 0.4229, "step": 29454 }, { "epoch": 2.97, "learning_rate": 8.9421989619382e-06, "loss": 0.4034, "step": 29455 }, { "epoch": 2.97, "learning_rate": 8.9404944962998e-06, "loss": 0.0745, "step": 29456 }, { "epoch": 2.97, "learning_rate": 8.938790164676082e-06, "loss": 0.5829, "step": 29457 }, { "epoch": 2.97, "learning_rate": 8.937085967077892e-06, "loss": 0.159, "step": 29458 }, { "epoch": 2.97, "learning_rate": 8.93538190351606e-06, "loss": 0.2938, "step": 29459 }, { "epoch": 2.97, "learning_rate": 8.93367797400147e-06, "loss": 0.6728, "step": 29460 }, { "epoch": 2.97, "learning_rate": 8.931974178544925e-06, "loss": 0.2036, "step": 29461 }, { "epoch": 2.97, "learning_rate": 8.930270517157272e-06, "loss": 0.0769, "step": 29462 }, { "epoch": 2.97, "learning_rate": 8.928566989849373e-06, "loss": 0.3596, "step": 29463 }, { "epoch": 2.97, "learning_rate": 8.926863596632066e-06, "loss": 0.5958, "step": 29464 }, { "epoch": 2.97, "learning_rate": 8.925160337516168e-06, "loss": 0.3784, "step": 29465 }, { "epoch": 2.97, "learning_rate": 8.923457212512538e-06, "loss": 0.3646, "step": 29466 }, { "epoch": 2.97, "learning_rate": 8.921754221632006e-06, "loss": 0.1628, "step": 29467 }, { "epoch": 2.97, "learning_rate": 8.920051364885416e-06, "loss": 0.3956, "step": 29468 }, { "epoch": 2.97, "learning_rate": 8.918348642283597e-06, "loss": 0.108, "step": 29469 }, { "epoch": 2.97, "learning_rate": 8.916646053837388e-06, "loss": 0.4704, "step": 29470 }, { "epoch": 2.97, "learning_rate": 8.914943599557624e-06, "loss": 0.2234, "step": 29471 }, { "epoch": 2.97, "learning_rate": 8.913241279455136e-06, "loss": 0.3932, "step": 29472 }, { "epoch": 2.97, "learning_rate": 8.911539093540753e-06, "loss": 0.3514, "step": 29473 }, { "epoch": 2.97, "learning_rate": 8.909837041825327e-06, "loss": 0.4825, "step": 29474 }, { "epoch": 2.97, "learning_rate": 8.908135124319669e-06, "loss": 0.1451, "step": 29475 }, { "epoch": 2.97, "learning_rate": 8.906433341034604e-06, "loss": 0.4582, "step": 29476 }, { "epoch": 2.97, "learning_rate": 8.904731691980982e-06, "loss": 0.4007, "step": 29477 }, { "epoch": 2.97, "learning_rate": 8.903030177169636e-06, "loss": 0.6327, "step": 29478 }, { "epoch": 2.97, "learning_rate": 8.901328796611357e-06, "loss": 0.0893, "step": 29479 }, { "epoch": 2.97, "learning_rate": 8.899627550317005e-06, "loss": 0.2363, "step": 29480 }, { "epoch": 2.97, "learning_rate": 8.897926438297396e-06, "loss": 0.7557, "step": 29481 }, { "epoch": 2.97, "learning_rate": 8.896225460563358e-06, "loss": 0.4536, "step": 29482 }, { "epoch": 2.97, "learning_rate": 8.894524617125712e-06, "loss": 0.4998, "step": 29483 }, { "epoch": 2.97, "learning_rate": 8.892823907995283e-06, "loss": 0.0268, "step": 29484 }, { "epoch": 2.97, "learning_rate": 8.89112333318289e-06, "loss": 0.1355, "step": 29485 }, { "epoch": 2.97, "learning_rate": 8.88942289269936e-06, "loss": 0.5543, "step": 29486 }, { "epoch": 2.97, "learning_rate": 8.887722586555511e-06, "loss": 0.1469, "step": 29487 }, { "epoch": 2.97, "learning_rate": 8.886022414762164e-06, "loss": 0.3843, "step": 29488 }, { "epoch": 2.97, "learning_rate": 8.884322377330138e-06, "loss": 0.0869, "step": 29489 }, { "epoch": 2.97, "learning_rate": 8.882622474270253e-06, "loss": 0.146, "step": 29490 }, { "epoch": 2.97, "learning_rate": 8.880922705593316e-06, "loss": 0.0603, "step": 29491 }, { "epoch": 2.97, "learning_rate": 8.879223071310172e-06, "loss": 0.0829, "step": 29492 }, { "epoch": 2.97, "learning_rate": 8.877523571431595e-06, "loss": 0.5655, "step": 29493 }, { "epoch": 2.97, "learning_rate": 8.875824205968438e-06, "loss": 0.134, "step": 29494 }, { "epoch": 2.97, "learning_rate": 8.874124974931493e-06, "loss": 0.2232, "step": 29495 }, { "epoch": 2.97, "learning_rate": 8.872425878331582e-06, "loss": 0.7294, "step": 29496 }, { "epoch": 2.97, "learning_rate": 8.870726916179519e-06, "loss": 0.2364, "step": 29497 }, { "epoch": 2.97, "learning_rate": 8.86902808848611e-06, "loss": 0.2972, "step": 29498 }, { "epoch": 2.97, "learning_rate": 8.867329395262167e-06, "loss": 0.2435, "step": 29499 }, { "epoch": 2.97, "learning_rate": 8.8656308365185e-06, "loss": 0.1328, "step": 29500 }, { "epoch": 2.97, "learning_rate": 8.863932412265917e-06, "loss": 0.6747, "step": 29501 }, { "epoch": 2.97, "learning_rate": 8.862234122515231e-06, "loss": 0.0992, "step": 29502 }, { "epoch": 2.97, "learning_rate": 8.860535967277243e-06, "loss": 0.0815, "step": 29503 }, { "epoch": 2.97, "learning_rate": 8.858837946562764e-06, "loss": 0.1662, "step": 29504 }, { "epoch": 2.97, "learning_rate": 8.85714006038259e-06, "loss": 0.1888, "step": 29505 }, { "epoch": 2.97, "learning_rate": 8.85544230874755e-06, "loss": 0.2282, "step": 29506 }, { "epoch": 2.97, "learning_rate": 8.853744691668423e-06, "loss": 0.1622, "step": 29507 }, { "epoch": 2.97, "learning_rate": 8.852047209156009e-06, "loss": 0.1112, "step": 29508 }, { "epoch": 2.97, "learning_rate": 8.850349861221132e-06, "loss": 0.4903, "step": 29509 }, { "epoch": 2.97, "learning_rate": 8.848652647874592e-06, "loss": 0.4132, "step": 29510 }, { "epoch": 2.97, "learning_rate": 8.84695556912716e-06, "loss": 0.4766, "step": 29511 }, { "epoch": 2.97, "learning_rate": 8.845258624989665e-06, "loss": 0.1546, "step": 29512 }, { "epoch": 2.97, "learning_rate": 8.843561815472899e-06, "loss": 0.7386, "step": 29513 }, { "epoch": 2.97, "learning_rate": 8.841865140587659e-06, "loss": 0.6948, "step": 29514 }, { "epoch": 2.97, "learning_rate": 8.840168600344735e-06, "loss": 0.344, "step": 29515 }, { "epoch": 2.97, "learning_rate": 8.838472194754933e-06, "loss": 0.1721, "step": 29516 }, { "epoch": 2.97, "learning_rate": 8.836775923829042e-06, "loss": 0.4179, "step": 29517 }, { "epoch": 2.97, "learning_rate": 8.835079787577859e-06, "loss": 0.0984, "step": 29518 }, { "epoch": 2.97, "learning_rate": 8.833383786012173e-06, "loss": 0.797, "step": 29519 }, { "epoch": 2.97, "learning_rate": 8.831687919142783e-06, "loss": 0.1631, "step": 29520 }, { "epoch": 2.97, "learning_rate": 8.829992186980478e-06, "loss": 0.8004, "step": 29521 }, { "epoch": 2.97, "learning_rate": 8.828296589536041e-06, "loss": 0.1713, "step": 29522 }, { "epoch": 2.97, "learning_rate": 8.826601126820279e-06, "loss": 0.7124, "step": 29523 }, { "epoch": 2.97, "learning_rate": 8.824905798843982e-06, "loss": 0.246, "step": 29524 }, { "epoch": 2.97, "learning_rate": 8.82321060561791e-06, "loss": 0.5019, "step": 29525 }, { "epoch": 2.97, "learning_rate": 8.821515547152877e-06, "loss": 0.25, "step": 29526 }, { "epoch": 2.97, "learning_rate": 8.819820623459668e-06, "loss": 0.1795, "step": 29527 }, { "epoch": 2.97, "learning_rate": 8.818125834549055e-06, "loss": 0.2685, "step": 29528 }, { "epoch": 2.97, "learning_rate": 8.816431180431837e-06, "loss": 0.2707, "step": 29529 }, { "epoch": 2.97, "learning_rate": 8.81473666111879e-06, "loss": 0.1743, "step": 29530 }, { "epoch": 2.97, "learning_rate": 8.813042276620703e-06, "loss": 0.1971, "step": 29531 }, { "epoch": 2.97, "learning_rate": 8.811348026948353e-06, "loss": 0.1743, "step": 29532 }, { "epoch": 2.97, "learning_rate": 8.809653912112524e-06, "loss": 0.6397, "step": 29533 }, { "epoch": 2.97, "learning_rate": 8.807959932123996e-06, "loss": 0.6834, "step": 29534 }, { "epoch": 2.97, "learning_rate": 8.80626608699355e-06, "loss": 0.2458, "step": 29535 }, { "epoch": 2.97, "learning_rate": 8.804572376731961e-06, "loss": 0.5817, "step": 29536 }, { "epoch": 2.97, "learning_rate": 8.802878801350004e-06, "loss": 0.5728, "step": 29537 }, { "epoch": 2.97, "learning_rate": 8.801185360858479e-06, "loss": 0.3501, "step": 29538 }, { "epoch": 2.97, "learning_rate": 8.799492055268138e-06, "loss": 0.149, "step": 29539 }, { "epoch": 2.97, "learning_rate": 8.797798884589755e-06, "loss": 0.1041, "step": 29540 }, { "epoch": 2.97, "learning_rate": 8.796105848834121e-06, "loss": 0.2216, "step": 29541 }, { "epoch": 2.97, "learning_rate": 8.794412948012017e-06, "loss": 0.7132, "step": 29542 }, { "epoch": 2.97, "learning_rate": 8.79272018213418e-06, "loss": 0.2813, "step": 29543 }, { "epoch": 2.97, "learning_rate": 8.791027551211413e-06, "loss": 0.3341, "step": 29544 }, { "epoch": 2.97, "learning_rate": 8.78933505525448e-06, "loss": 0.1636, "step": 29545 }, { "epoch": 2.97, "learning_rate": 8.787642694274152e-06, "loss": 0.4869, "step": 29546 }, { "epoch": 2.97, "learning_rate": 8.785950468281193e-06, "loss": 0.7781, "step": 29547 }, { "epoch": 2.97, "learning_rate": 8.784258377286374e-06, "loss": 0.4004, "step": 29548 }, { "epoch": 2.97, "learning_rate": 8.782566421300466e-06, "loss": 0.6625, "step": 29549 }, { "epoch": 2.98, "learning_rate": 8.780874600334235e-06, "loss": 0.1549, "step": 29550 }, { "epoch": 2.98, "learning_rate": 8.779182914398434e-06, "loss": 0.0918, "step": 29551 }, { "epoch": 2.98, "learning_rate": 8.777491363503858e-06, "loss": 0.1162, "step": 29552 }, { "epoch": 2.98, "learning_rate": 8.775799947661248e-06, "loss": 0.7233, "step": 29553 }, { "epoch": 2.98, "learning_rate": 8.77410866688136e-06, "loss": 0.5093, "step": 29554 }, { "epoch": 2.98, "learning_rate": 8.77241752117498e-06, "loss": 0.509, "step": 29555 }, { "epoch": 2.98, "learning_rate": 8.770726510552872e-06, "loss": 0.5175, "step": 29556 }, { "epoch": 2.98, "learning_rate": 8.76903563502576e-06, "loss": 0.419, "step": 29557 }, { "epoch": 2.98, "learning_rate": 8.767344894604444e-06, "loss": 0.355, "step": 29558 }, { "epoch": 2.98, "learning_rate": 8.765654289299664e-06, "loss": 0.1444, "step": 29559 }, { "epoch": 2.98, "learning_rate": 8.763963819122183e-06, "loss": 0.2558, "step": 29560 }, { "epoch": 2.98, "learning_rate": 8.762273484082758e-06, "loss": 0.3758, "step": 29561 }, { "epoch": 2.98, "learning_rate": 8.760583284192145e-06, "loss": 0.1905, "step": 29562 }, { "epoch": 2.98, "learning_rate": 8.758893219461101e-06, "loss": 0.3193, "step": 29563 }, { "epoch": 2.98, "learning_rate": 8.757203289900378e-06, "loss": 0.4122, "step": 29564 }, { "epoch": 2.98, "learning_rate": 8.755513495520732e-06, "loss": 0.9171, "step": 29565 }, { "epoch": 2.98, "learning_rate": 8.753823836332917e-06, "loss": 0.5518, "step": 29566 }, { "epoch": 2.98, "learning_rate": 8.752134312347685e-06, "loss": 0.0679, "step": 29567 }, { "epoch": 2.98, "learning_rate": 8.750444923575788e-06, "loss": 0.1492, "step": 29568 }, { "epoch": 2.98, "learning_rate": 8.748755670027964e-06, "loss": 0.7737, "step": 29569 }, { "epoch": 2.98, "learning_rate": 8.747066551714995e-06, "loss": 0.3175, "step": 29570 }, { "epoch": 2.98, "learning_rate": 8.7453775686476e-06, "loss": 0.5785, "step": 29571 }, { "epoch": 2.98, "learning_rate": 8.743688720836528e-06, "loss": 0.4046, "step": 29572 }, { "epoch": 2.98, "learning_rate": 8.742000008292543e-06, "loss": 0.4213, "step": 29573 }, { "epoch": 2.98, "learning_rate": 8.740311431026385e-06, "loss": 0.6967, "step": 29574 }, { "epoch": 2.98, "learning_rate": 8.738622989048795e-06, "loss": 0.2241, "step": 29575 }, { "epoch": 2.98, "learning_rate": 8.736934682370519e-06, "loss": 0.2709, "step": 29576 }, { "epoch": 2.98, "learning_rate": 8.735246511002305e-06, "loss": 0.5956, "step": 29577 }, { "epoch": 2.98, "learning_rate": 8.73355847495489e-06, "loss": 0.3217, "step": 29578 }, { "epoch": 2.98, "learning_rate": 8.731870574239019e-06, "loss": 0.602, "step": 29579 }, { "epoch": 2.98, "learning_rate": 8.730182808865433e-06, "loss": 0.2563, "step": 29580 }, { "epoch": 2.98, "learning_rate": 8.728495178844877e-06, "loss": 0.1334, "step": 29581 }, { "epoch": 2.98, "learning_rate": 8.72680768418808e-06, "loss": 0.29, "step": 29582 }, { "epoch": 2.98, "learning_rate": 8.725120324905777e-06, "loss": 0.1903, "step": 29583 }, { "epoch": 2.98, "learning_rate": 8.723433101008735e-06, "loss": 0.5969, "step": 29584 }, { "epoch": 2.98, "learning_rate": 8.72174601250766e-06, "loss": 0.2875, "step": 29585 }, { "epoch": 2.98, "learning_rate": 8.720059059413292e-06, "loss": 0.3856, "step": 29586 }, { "epoch": 2.98, "learning_rate": 8.71837224173638e-06, "loss": 0.2038, "step": 29587 }, { "epoch": 2.98, "learning_rate": 8.716685559487662e-06, "loss": 0.3456, "step": 29588 }, { "epoch": 2.98, "learning_rate": 8.71499901267784e-06, "loss": 1.095, "step": 29589 }, { "epoch": 2.98, "learning_rate": 8.713312601317678e-06, "loss": 0.2011, "step": 29590 }, { "epoch": 2.98, "learning_rate": 8.711626325417896e-06, "loss": 0.0942, "step": 29591 }, { "epoch": 2.98, "learning_rate": 8.709940184989226e-06, "loss": 0.6365, "step": 29592 }, { "epoch": 2.98, "learning_rate": 8.708254180042398e-06, "loss": 0.1055, "step": 29593 }, { "epoch": 2.98, "learning_rate": 8.70656831058814e-06, "loss": 0.4177, "step": 29594 }, { "epoch": 2.98, "learning_rate": 8.704882576637177e-06, "loss": 0.6054, "step": 29595 }, { "epoch": 2.98, "learning_rate": 8.703196978200245e-06, "loss": 0.2273, "step": 29596 }, { "epoch": 2.98, "learning_rate": 8.701511515288065e-06, "loss": 0.1878, "step": 29597 }, { "epoch": 2.98, "learning_rate": 8.699826187911363e-06, "loss": 0.2537, "step": 29598 }, { "epoch": 2.98, "learning_rate": 8.698140996080866e-06, "loss": 0.5269, "step": 29599 }, { "epoch": 2.98, "learning_rate": 8.696455939807294e-06, "loss": 0.3164, "step": 29600 }, { "epoch": 2.98, "learning_rate": 8.694771019101363e-06, "loss": 0.407, "step": 29601 }, { "epoch": 2.98, "learning_rate": 8.69308623397382e-06, "loss": 0.5863, "step": 29602 }, { "epoch": 2.98, "learning_rate": 8.691401584435355e-06, "loss": 0.3031, "step": 29603 }, { "epoch": 2.98, "learning_rate": 8.689717070496713e-06, "loss": 0.0995, "step": 29604 }, { "epoch": 2.98, "learning_rate": 8.688032692168605e-06, "loss": 0.1003, "step": 29605 }, { "epoch": 2.98, "learning_rate": 8.68634844946175e-06, "loss": 0.344, "step": 29606 }, { "epoch": 2.98, "learning_rate": 8.684664342386865e-06, "loss": 0.4065, "step": 29607 }, { "epoch": 2.98, "learning_rate": 8.682980370954667e-06, "loss": 0.1886, "step": 29608 }, { "epoch": 2.98, "learning_rate": 8.681296535175874e-06, "loss": 0.5746, "step": 29609 }, { "epoch": 2.98, "learning_rate": 8.679612835061197e-06, "loss": 0.1924, "step": 29610 }, { "epoch": 2.98, "learning_rate": 8.677929270621356e-06, "loss": 0.517, "step": 29611 }, { "epoch": 2.98, "learning_rate": 8.67624584186706e-06, "loss": 0.1115, "step": 29612 }, { "epoch": 2.98, "learning_rate": 8.674562548809022e-06, "loss": 0.0723, "step": 29613 }, { "epoch": 2.98, "learning_rate": 8.672879391457961e-06, "loss": 0.6112, "step": 29614 }, { "epoch": 2.98, "learning_rate": 8.67119636982457e-06, "loss": 0.6001, "step": 29615 }, { "epoch": 2.98, "learning_rate": 8.669513483919585e-06, "loss": 0.2763, "step": 29616 }, { "epoch": 2.98, "learning_rate": 8.667830733753699e-06, "loss": 0.424, "step": 29617 }, { "epoch": 2.98, "learning_rate": 8.666148119337611e-06, "loss": 0.2237, "step": 29618 }, { "epoch": 2.98, "learning_rate": 8.664465640682051e-06, "loss": 0.2428, "step": 29619 }, { "epoch": 2.98, "learning_rate": 8.662783297797722e-06, "loss": 0.2626, "step": 29620 }, { "epoch": 2.98, "learning_rate": 8.661101090695308e-06, "loss": 0.2587, "step": 29621 }, { "epoch": 2.98, "learning_rate": 8.659419019385536e-06, "loss": 0.385, "step": 29622 }, { "epoch": 2.98, "learning_rate": 8.657737083879101e-06, "loss": 0.2531, "step": 29623 }, { "epoch": 2.98, "learning_rate": 8.656055284186711e-06, "loss": 0.0979, "step": 29624 }, { "epoch": 2.98, "learning_rate": 8.654373620319066e-06, "loss": 0.582, "step": 29625 }, { "epoch": 2.98, "learning_rate": 8.652692092286866e-06, "loss": 0.5031, "step": 29626 }, { "epoch": 2.98, "learning_rate": 8.651010700100811e-06, "loss": 0.1157, "step": 29627 }, { "epoch": 2.98, "learning_rate": 8.649329443771606e-06, "loss": 0.5545, "step": 29628 }, { "epoch": 2.98, "learning_rate": 8.647648323309935e-06, "loss": 0.0669, "step": 29629 }, { "epoch": 2.98, "learning_rate": 8.645967338726525e-06, "loss": 0.3393, "step": 29630 }, { "epoch": 2.98, "learning_rate": 8.644286490032048e-06, "loss": 0.6793, "step": 29631 }, { "epoch": 2.98, "learning_rate": 8.642605777237196e-06, "loss": 0.0774, "step": 29632 }, { "epoch": 2.98, "learning_rate": 8.640925200352688e-06, "loss": 0.2455, "step": 29633 }, { "epoch": 2.98, "learning_rate": 8.639244759389214e-06, "loss": 0.3865, "step": 29634 }, { "epoch": 2.98, "learning_rate": 8.637564454357443e-06, "loss": 0.4289, "step": 29635 }, { "epoch": 2.98, "learning_rate": 8.635884285268092e-06, "loss": 0.0707, "step": 29636 }, { "epoch": 2.98, "learning_rate": 8.634204252131846e-06, "loss": 0.5839, "step": 29637 }, { "epoch": 2.98, "learning_rate": 8.632524354959402e-06, "loss": 0.532, "step": 29638 }, { "epoch": 2.98, "learning_rate": 8.630844593761438e-06, "loss": 0.114, "step": 29639 }, { "epoch": 2.98, "learning_rate": 8.629164968548651e-06, "loss": 0.1638, "step": 29640 }, { "epoch": 2.98, "learning_rate": 8.62748547933173e-06, "loss": 0.422, "step": 29641 }, { "epoch": 2.98, "learning_rate": 8.625806126121362e-06, "loss": 0.6087, "step": 29642 }, { "epoch": 2.98, "learning_rate": 8.624126908928229e-06, "loss": 0.4224, "step": 29643 }, { "epoch": 2.98, "learning_rate": 8.622447827763023e-06, "loss": 0.1785, "step": 29644 }, { "epoch": 2.98, "learning_rate": 8.620768882636427e-06, "loss": 0.5095, "step": 29645 }, { "epoch": 2.98, "learning_rate": 8.619090073559122e-06, "loss": 0.1202, "step": 29646 }, { "epoch": 2.98, "learning_rate": 8.617411400541787e-06, "loss": 0.32, "step": 29647 }, { "epoch": 2.98, "learning_rate": 8.61573286359513e-06, "loss": 0.1586, "step": 29648 }, { "epoch": 2.99, "learning_rate": 8.614054462729803e-06, "loss": 0.1619, "step": 29649 }, { "epoch": 2.99, "learning_rate": 8.61237619795649e-06, "loss": 0.1645, "step": 29650 }, { "epoch": 2.99, "learning_rate": 8.610698069285887e-06, "loss": 0.3199, "step": 29651 }, { "epoch": 2.99, "learning_rate": 8.609020076728674e-06, "loss": 0.2028, "step": 29652 }, { "epoch": 2.99, "learning_rate": 8.607342220295504e-06, "loss": 0.4353, "step": 29653 }, { "epoch": 2.99, "learning_rate": 8.605664499997075e-06, "loss": 0.5314, "step": 29654 }, { "epoch": 2.99, "learning_rate": 8.60398691584406e-06, "loss": 0.1365, "step": 29655 }, { "epoch": 2.99, "learning_rate": 8.602309467847134e-06, "loss": 0.2289, "step": 29656 }, { "epoch": 2.99, "learning_rate": 8.600632156016968e-06, "loss": 0.324, "step": 29657 }, { "epoch": 2.99, "learning_rate": 8.59895498036424e-06, "loss": 0.5132, "step": 29658 }, { "epoch": 2.99, "learning_rate": 8.59727794089962e-06, "loss": 0.5742, "step": 29659 }, { "epoch": 2.99, "learning_rate": 8.595601037633783e-06, "loss": 0.31, "step": 29660 }, { "epoch": 2.99, "learning_rate": 8.59392427057739e-06, "loss": 0.0483, "step": 29661 }, { "epoch": 2.99, "learning_rate": 8.592247639741136e-06, "loss": 0.7189, "step": 29662 }, { "epoch": 2.99, "learning_rate": 8.590571145135663e-06, "loss": 0.6104, "step": 29663 }, { "epoch": 2.99, "learning_rate": 8.588894786771644e-06, "loss": 0.0646, "step": 29664 }, { "epoch": 2.99, "learning_rate": 8.587218564659761e-06, "loss": 0.2109, "step": 29665 }, { "epoch": 2.99, "learning_rate": 8.585542478810685e-06, "loss": 0.4553, "step": 29666 }, { "epoch": 2.99, "learning_rate": 8.58386652923505e-06, "loss": 0.451, "step": 29667 }, { "epoch": 2.99, "learning_rate": 8.582190715943552e-06, "loss": 0.2226, "step": 29668 }, { "epoch": 2.99, "learning_rate": 8.580515038946842e-06, "loss": 0.0679, "step": 29669 }, { "epoch": 2.99, "learning_rate": 8.578839498255586e-06, "loss": 0.7435, "step": 29670 }, { "epoch": 2.99, "learning_rate": 8.577164093880449e-06, "loss": 0.1675, "step": 29671 }, { "epoch": 2.99, "learning_rate": 8.575488825832087e-06, "loss": 0.6305, "step": 29672 }, { "epoch": 2.99, "learning_rate": 8.573813694121167e-06, "loss": 0.2489, "step": 29673 }, { "epoch": 2.99, "learning_rate": 8.572138698758344e-06, "loss": 0.314, "step": 29674 }, { "epoch": 2.99, "learning_rate": 8.570463839754277e-06, "loss": 0.2993, "step": 29675 }, { "epoch": 2.99, "learning_rate": 8.56878911711963e-06, "loss": 0.2627, "step": 29676 }, { "epoch": 2.99, "learning_rate": 8.567114530865055e-06, "loss": 0.1822, "step": 29677 }, { "epoch": 2.99, "learning_rate": 8.565440081001207e-06, "loss": 0.8363, "step": 29678 }, { "epoch": 2.99, "learning_rate": 8.56376576753874e-06, "loss": 0.1013, "step": 29679 }, { "epoch": 2.99, "learning_rate": 8.562091590488327e-06, "loss": 0.531, "step": 29680 }, { "epoch": 2.99, "learning_rate": 8.560417549860598e-06, "loss": 0.1373, "step": 29681 }, { "epoch": 2.99, "learning_rate": 8.55874364566621e-06, "loss": 0.4189, "step": 29682 }, { "epoch": 2.99, "learning_rate": 8.557069877915828e-06, "loss": 0.4754, "step": 29683 }, { "epoch": 2.99, "learning_rate": 8.555396246620094e-06, "loss": 0.5627, "step": 29684 }, { "epoch": 2.99, "learning_rate": 8.553722751789662e-06, "loss": 0.7413, "step": 29685 }, { "epoch": 2.99, "learning_rate": 8.552049393435179e-06, "loss": 0.764, "step": 29686 }, { "epoch": 2.99, "learning_rate": 8.550376171567297e-06, "loss": 0.1738, "step": 29687 }, { "epoch": 2.99, "learning_rate": 8.548703086196661e-06, "loss": 0.354, "step": 29688 }, { "epoch": 2.99, "learning_rate": 8.547030137333917e-06, "loss": 0.4591, "step": 29689 }, { "epoch": 2.99, "learning_rate": 8.545357324989712e-06, "loss": 0.0371, "step": 29690 }, { "epoch": 2.99, "learning_rate": 8.54368464917469e-06, "loss": 0.5682, "step": 29691 }, { "epoch": 2.99, "learning_rate": 8.5420121098995e-06, "loss": 0.2704, "step": 29692 }, { "epoch": 2.99, "learning_rate": 8.540339707174769e-06, "loss": 0.2738, "step": 29693 }, { "epoch": 2.99, "learning_rate": 8.538667441011172e-06, "loss": 0.4579, "step": 29694 }, { "epoch": 2.99, "learning_rate": 8.536995311419323e-06, "loss": 0.3053, "step": 29695 }, { "epoch": 2.99, "learning_rate": 8.535323318409858e-06, "loss": 0.215, "step": 29696 }, { "epoch": 2.99, "learning_rate": 8.533651461993444e-06, "loss": 0.2418, "step": 29697 }, { "epoch": 2.99, "learning_rate": 8.53197974218071e-06, "loss": 0.6258, "step": 29698 }, { "epoch": 2.99, "learning_rate": 8.530308158982276e-06, "loss": 0.3892, "step": 29699 }, { "epoch": 2.99, "learning_rate": 8.5286367124088e-06, "loss": 0.2609, "step": 29700 }, { "epoch": 2.99, "learning_rate": 8.526965402470913e-06, "loss": 0.1857, "step": 29701 }, { "epoch": 2.99, "learning_rate": 8.525294229179251e-06, "loss": 0.2554, "step": 29702 }, { "epoch": 2.99, "learning_rate": 8.523623192544443e-06, "loss": 0.1721, "step": 29703 }, { "epoch": 2.99, "learning_rate": 8.521952292577133e-06, "loss": 1.0493, "step": 29704 }, { "epoch": 2.99, "learning_rate": 8.520281529287943e-06, "loss": 0.5802, "step": 29705 }, { "epoch": 2.99, "learning_rate": 8.518610902687512e-06, "loss": 0.4552, "step": 29706 }, { "epoch": 2.99, "learning_rate": 8.516940412786468e-06, "loss": 0.1474, "step": 29707 }, { "epoch": 2.99, "learning_rate": 8.515270059595442e-06, "loss": 0.5196, "step": 29708 }, { "epoch": 2.99, "learning_rate": 8.513599843125066e-06, "loss": 0.056, "step": 29709 }, { "epoch": 2.99, "learning_rate": 8.511929763385957e-06, "loss": 0.7947, "step": 29710 }, { "epoch": 2.99, "learning_rate": 8.510259820388762e-06, "loss": 0.1473, "step": 29711 }, { "epoch": 2.99, "learning_rate": 8.508590014144108e-06, "loss": 0.7195, "step": 29712 }, { "epoch": 2.99, "learning_rate": 8.506920344662593e-06, "loss": 0.2988, "step": 29713 }, { "epoch": 2.99, "learning_rate": 8.505250811954868e-06, "loss": 0.112, "step": 29714 }, { "epoch": 2.99, "learning_rate": 8.503581416031549e-06, "loss": 0.2318, "step": 29715 }, { "epoch": 2.99, "learning_rate": 8.501912156903264e-06, "loss": 0.147, "step": 29716 }, { "epoch": 2.99, "learning_rate": 8.50024303458063e-06, "loss": 0.5339, "step": 29717 }, { "epoch": 2.99, "learning_rate": 8.498574049074269e-06, "loss": 0.2605, "step": 29718 }, { "epoch": 2.99, "learning_rate": 8.496905200394804e-06, "loss": 0.2668, "step": 29719 }, { "epoch": 2.99, "learning_rate": 8.495236488552854e-06, "loss": 0.2976, "step": 29720 }, { "epoch": 2.99, "learning_rate": 8.493567913559041e-06, "loss": 0.5965, "step": 29721 }, { "epoch": 2.99, "learning_rate": 8.491899475423977e-06, "loss": 0.4427, "step": 29722 }, { "epoch": 2.99, "learning_rate": 8.490231174158284e-06, "loss": 0.3945, "step": 29723 }, { "epoch": 2.99, "learning_rate": 8.488563009772577e-06, "loss": 0.6963, "step": 29724 }, { "epoch": 2.99, "learning_rate": 8.486894982277463e-06, "loss": 0.2299, "step": 29725 }, { "epoch": 2.99, "learning_rate": 8.485227091683583e-06, "loss": 0.2535, "step": 29726 }, { "epoch": 2.99, "learning_rate": 8.483559338001524e-06, "loss": 0.3637, "step": 29727 }, { "epoch": 2.99, "learning_rate": 8.481891721241898e-06, "loss": 0.226, "step": 29728 }, { "epoch": 2.99, "learning_rate": 8.480224241415335e-06, "loss": 0.0511, "step": 29729 }, { "epoch": 2.99, "learning_rate": 8.47855689853245e-06, "loss": 0.1712, "step": 29730 }, { "epoch": 2.99, "learning_rate": 8.476889692603825e-06, "loss": 0.1088, "step": 29731 }, { "epoch": 2.99, "learning_rate": 8.475222623640093e-06, "loss": 0.525, "step": 29732 }, { "epoch": 2.99, "learning_rate": 8.473555691651856e-06, "loss": 0.6096, "step": 29733 }, { "epoch": 2.99, "learning_rate": 8.471888896649723e-06, "loss": 0.2098, "step": 29734 }, { "epoch": 2.99, "learning_rate": 8.470222238644297e-06, "loss": 0.4444, "step": 29735 }, { "epoch": 2.99, "learning_rate": 8.468555717646187e-06, "loss": 0.227, "step": 29736 }, { "epoch": 2.99, "learning_rate": 8.466889333665997e-06, "loss": 0.0972, "step": 29737 }, { "epoch": 2.99, "learning_rate": 8.465223086714332e-06, "loss": 0.153, "step": 29738 }, { "epoch": 2.99, "learning_rate": 8.463556976801787e-06, "loss": 0.1643, "step": 29739 }, { "epoch": 2.99, "learning_rate": 8.461891003938991e-06, "loss": 0.163, "step": 29740 }, { "epoch": 2.99, "learning_rate": 8.460225168136517e-06, "loss": 0.4523, "step": 29741 }, { "epoch": 2.99, "learning_rate": 8.458559469404968e-06, "loss": 0.0789, "step": 29742 }, { "epoch": 2.99, "learning_rate": 8.456893907754958e-06, "loss": 0.3713, "step": 29743 }, { "epoch": 2.99, "learning_rate": 8.45522848319709e-06, "loss": 0.344, "step": 29744 }, { "epoch": 2.99, "learning_rate": 8.453563195741935e-06, "loss": 0.1775, "step": 29745 }, { "epoch": 2.99, "learning_rate": 8.451898045400116e-06, "loss": 0.0773, "step": 29746 }, { "epoch": 3.0, "learning_rate": 8.45023303218222e-06, "loss": 0.8461, "step": 29747 }, { "epoch": 3.0, "learning_rate": 8.448568156098844e-06, "loss": 0.2903, "step": 29748 }, { "epoch": 3.0, "learning_rate": 8.44690341716058e-06, "loss": 0.1537, "step": 29749 }, { "epoch": 3.0, "learning_rate": 8.445238815378026e-06, "loss": 0.2829, "step": 29750 }, { "epoch": 3.0, "learning_rate": 8.44357435076177e-06, "loss": 0.5768, "step": 29751 }, { "epoch": 3.0, "learning_rate": 8.441910023322404e-06, "loss": 0.3327, "step": 29752 }, { "epoch": 3.0, "learning_rate": 8.440245833070526e-06, "loss": 0.3139, "step": 29753 }, { "epoch": 3.0, "learning_rate": 8.438581780016723e-06, "loss": 0.3285, "step": 29754 }, { "epoch": 3.0, "learning_rate": 8.436917864171578e-06, "loss": 0.2423, "step": 29755 }, { "epoch": 3.0, "learning_rate": 8.435254085545687e-06, "loss": 0.5212, "step": 29756 }, { "epoch": 3.0, "learning_rate": 8.433590444149626e-06, "loss": 0.7536, "step": 29757 }, { "epoch": 3.0, "learning_rate": 8.43192693999401e-06, "loss": 0.2506, "step": 29758 }, { "epoch": 3.0, "learning_rate": 8.430263573089395e-06, "loss": 0.0985, "step": 29759 }, { "epoch": 3.0, "learning_rate": 8.42860034344637e-06, "loss": 0.1858, "step": 29760 }, { "epoch": 3.0, "learning_rate": 8.426937251075534e-06, "loss": 0.4735, "step": 29761 }, { "epoch": 3.0, "learning_rate": 8.425274295987475e-06, "loss": 0.1529, "step": 29762 }, { "epoch": 3.0, "learning_rate": 8.423611478192742e-06, "loss": 0.0953, "step": 29763 }, { "epoch": 3.0, "learning_rate": 8.421948797701948e-06, "loss": 0.0337, "step": 29764 }, { "epoch": 3.0, "learning_rate": 8.420286254525658e-06, "loss": 0.1858, "step": 29765 }, { "epoch": 3.0, "learning_rate": 8.418623848674462e-06, "loss": 0.2805, "step": 29766 }, { "epoch": 3.0, "learning_rate": 8.416961580158935e-06, "loss": 0.6454, "step": 29767 }, { "epoch": 3.0, "learning_rate": 8.41529944898964e-06, "loss": 0.9615, "step": 29768 }, { "epoch": 3.0, "learning_rate": 8.41363745517719e-06, "loss": 0.4673, "step": 29769 }, { "epoch": 3.0, "learning_rate": 8.41197559873213e-06, "loss": 0.1288, "step": 29770 }, { "epoch": 3.0, "learning_rate": 8.410313879665032e-06, "loss": 0.2722, "step": 29771 }, { "epoch": 3.0, "learning_rate": 8.408652297986506e-06, "loss": 0.2232, "step": 29772 }, { "epoch": 3.0, "learning_rate": 8.406990853707095e-06, "loss": 0.5513, "step": 29773 }, { "epoch": 3.0, "learning_rate": 8.405329546837366e-06, "loss": 0.1171, "step": 29774 }, { "epoch": 3.0, "learning_rate": 8.403668377387916e-06, "loss": 0.5252, "step": 29775 }, { "epoch": 3.0, "learning_rate": 8.402007345369313e-06, "loss": 0.5151, "step": 29776 }, { "epoch": 3.0, "learning_rate": 8.400346450792099e-06, "loss": 0.2499, "step": 29777 }, { "epoch": 3.0, "learning_rate": 8.398685693666877e-06, "loss": 0.3028, "step": 29778 }, { "epoch": 3.0, "learning_rate": 8.397025074004197e-06, "loss": 0.3432, "step": 29779 }, { "epoch": 3.0, "learning_rate": 8.395364591814637e-06, "loss": 0.6326, "step": 29780 }, { "epoch": 3.0, "learning_rate": 8.393704247108754e-06, "loss": 0.2955, "step": 29781 }, { "epoch": 3.0, "learning_rate": 8.392044039897118e-06, "loss": 0.5242, "step": 29782 }, { "epoch": 3.0, "learning_rate": 8.390383970190291e-06, "loss": 0.6235, "step": 29783 }, { "epoch": 3.0, "learning_rate": 8.388724037998845e-06, "loss": 0.9118, "step": 29784 }, { "epoch": 3.0, "learning_rate": 8.387064243333336e-06, "loss": 0.3712, "step": 29785 }, { "epoch": 3.0, "learning_rate": 8.385404586204329e-06, "loss": 0.7687, "step": 29786 }, { "epoch": 3.0, "learning_rate": 8.38374506662238e-06, "loss": 0.6373, "step": 29787 }, { "epoch": 3.0, "learning_rate": 8.38208568459806e-06, "loss": 0.1369, "step": 29788 }, { "epoch": 3.0, "learning_rate": 8.380426440141909e-06, "loss": 0.5892, "step": 29789 }, { "epoch": 3.0, "learning_rate": 8.378767333264522e-06, "loss": 0.0655, "step": 29790 }, { "epoch": 3.0, "learning_rate": 8.377108363976412e-06, "loss": 0.4759, "step": 29791 }, { "epoch": 3.0, "learning_rate": 8.37544953228817e-06, "loss": 0.6334, "step": 29792 }, { "epoch": 3.0, "learning_rate": 8.373790838210342e-06, "loss": 0.3589, "step": 29793 }, { "epoch": 3.0, "learning_rate": 8.372132281753482e-06, "loss": 0.3301, "step": 29794 }, { "epoch": 3.0, "learning_rate": 8.37047386292814e-06, "loss": 0.4206, "step": 29795 }, { "epoch": 3.0, "learning_rate": 8.36881558174488e-06, "loss": 0.7144, "step": 29796 }, { "epoch": 3.0, "learning_rate": 8.367157438214246e-06, "loss": 0.7446, "step": 29797 }, { "epoch": 3.0, "learning_rate": 8.365499432346794e-06, "loss": 0.3326, "step": 29798 }, { "epoch": 3.0, "learning_rate": 8.363841564153069e-06, "loss": 0.1876, "step": 29799 }, { "epoch": 3.0, "learning_rate": 8.362183833643619e-06, "loss": 0.6851, "step": 29800 }, { "epoch": 3.0, "learning_rate": 8.36052624082902e-06, "loss": 0.0467, "step": 29801 }, { "epoch": 3.0, "learning_rate": 8.358868785719788e-06, "loss": 0.1772, "step": 29802 }, { "epoch": 3.0, "learning_rate": 8.357211468326475e-06, "loss": 0.4728, "step": 29803 }, { "epoch": 3.0, "learning_rate": 8.355554288659654e-06, "loss": 0.6619, "step": 29804 }, { "epoch": 3.0, "learning_rate": 8.35389724672984e-06, "loss": 0.2291, "step": 29805 }, { "epoch": 3.0, "learning_rate": 8.35224034254758e-06, "loss": 0.2819, "step": 29806 }, { "epoch": 3.0, "learning_rate": 8.350583576123438e-06, "loss": 0.3809, "step": 29807 }, { "epoch": 3.0, "learning_rate": 8.348926947467956e-06, "loss": 0.4243, "step": 29808 }, { "epoch": 3.0, "learning_rate": 8.347270456591645e-06, "loss": 0.1236, "step": 29809 }, { "epoch": 3.0, "learning_rate": 8.345614103505078e-06, "loss": 0.1133, "step": 29810 }, { "epoch": 3.0, "learning_rate": 8.343957888218788e-06, "loss": 0.3528, "step": 29811 }, { "epoch": 3.0, "learning_rate": 8.342301810743309e-06, "loss": 0.5094, "step": 29812 }, { "epoch": 3.0, "learning_rate": 8.340645871089182e-06, "loss": 0.5144, "step": 29813 }, { "epoch": 3.0, "learning_rate": 8.338990069266945e-06, "loss": 0.458, "step": 29814 }, { "epoch": 3.0, "learning_rate": 8.337334405287136e-06, "loss": 0.2126, "step": 29815 }, { "epoch": 3.0, "learning_rate": 8.335678879160292e-06, "loss": 0.5663, "step": 29816 }, { "epoch": 3.0, "learning_rate": 8.334023490896943e-06, "loss": 0.3328, "step": 29817 }, { "epoch": 3.0, "learning_rate": 8.332368240507625e-06, "loss": 0.2251, "step": 29818 }, { "epoch": 3.0, "learning_rate": 8.330713128002876e-06, "loss": 0.2932, "step": 29819 }, { "epoch": 3.0, "learning_rate": 8.329058153393216e-06, "loss": 0.5232, "step": 29820 }, { "epoch": 3.0, "learning_rate": 8.327403316689192e-06, "loss": 0.141, "step": 29821 }, { "epoch": 3.0, "learning_rate": 8.325748617901342e-06, "loss": 0.7669, "step": 29822 }, { "epoch": 3.0, "learning_rate": 8.32409405704016e-06, "loss": 0.1591, "step": 29823 }, { "epoch": 3.0, "learning_rate": 8.322439634116208e-06, "loss": 0.9693, "step": 29824 }, { "epoch": 3.0, "learning_rate": 8.320785349140007e-06, "loss": 0.1264, "step": 29825 }, { "epoch": 3.0, "learning_rate": 8.319131202122076e-06, "loss": 0.2839, "step": 29826 }, { "epoch": 3.0, "learning_rate": 8.31747719307295e-06, "loss": 0.2677, "step": 29827 }, { "epoch": 3.0, "learning_rate": 8.315823322003148e-06, "loss": 0.2793, "step": 29828 }, { "epoch": 3.0, "learning_rate": 8.314169588923198e-06, "loss": 0.4379, "step": 29829 }, { "epoch": 3.0, "learning_rate": 8.312515993843623e-06, "loss": 0.3701, "step": 29830 }, { "epoch": 3.0, "learning_rate": 8.310862536774944e-06, "loss": 0.2593, "step": 29831 }, { "epoch": 3.0, "learning_rate": 8.30920921772768e-06, "loss": 0.2233, "step": 29832 }, { "epoch": 3.0, "learning_rate": 8.307556036712371e-06, "loss": 0.086, "step": 29833 }, { "epoch": 3.0, "learning_rate": 8.305902993739516e-06, "loss": 0.4606, "step": 29834 }, { "epoch": 3.0, "learning_rate": 8.304250088819635e-06, "loss": 0.4182, "step": 29835 }, { "epoch": 3.0, "learning_rate": 8.302597321963267e-06, "loss": 0.3847, "step": 29836 }, { "epoch": 3.0, "learning_rate": 8.300944693180907e-06, "loss": 0.1862, "step": 29837 }, { "epoch": 3.0, "learning_rate": 8.299292202483072e-06, "loss": 0.6363, "step": 29838 }, { "epoch": 3.0, "learning_rate": 8.297639849880296e-06, "loss": 0.593, "step": 29839 }, { "epoch": 3.0, "learning_rate": 8.295987635383093e-06, "loss": 0.2636, "step": 29840 }, { "epoch": 3.0, "learning_rate": 8.294335559001948e-06, "loss": 0.6546, "step": 29841 }, { "epoch": 3.0, "learning_rate": 8.292683620747407e-06, "loss": 0.4377, "step": 29842 }, { "epoch": 3.0, "learning_rate": 8.291031820629967e-06, "loss": 0.32, "step": 29843 }, { "epoch": 3.0, "learning_rate": 8.289380158660141e-06, "loss": 0.303, "step": 29844 }, { "epoch": 3.0, "learning_rate": 8.28772863484844e-06, "loss": 0.5285, "step": 29845 }, { "epoch": 3.01, "learning_rate": 8.28607724920537e-06, "loss": 0.4848, "step": 29846 }, { "epoch": 3.01, "learning_rate": 8.284426001741457e-06, "loss": 0.5834, "step": 29847 }, { "epoch": 3.01, "learning_rate": 8.28277489246719e-06, "loss": 0.1651, "step": 29848 }, { "epoch": 3.01, "learning_rate": 8.28112392139307e-06, "loss": 0.1987, "step": 29849 }, { "epoch": 3.01, "learning_rate": 8.279473088529636e-06, "loss": 0.8257, "step": 29850 }, { "epoch": 3.01, "learning_rate": 8.27782239388736e-06, "loss": 0.1791, "step": 29851 }, { "epoch": 3.01, "learning_rate": 8.276171837476753e-06, "loss": 0.5545, "step": 29852 }, { "epoch": 3.01, "learning_rate": 8.274521419308332e-06, "loss": 0.2481, "step": 29853 }, { "epoch": 3.01, "learning_rate": 8.272871139392601e-06, "loss": 0.4668, "step": 29854 }, { "epoch": 3.01, "learning_rate": 8.271220997740034e-06, "loss": 0.1346, "step": 29855 }, { "epoch": 3.01, "learning_rate": 8.269570994361161e-06, "loss": 0.2476, "step": 29856 }, { "epoch": 3.01, "learning_rate": 8.26792112926647e-06, "loss": 0.7154, "step": 29857 }, { "epoch": 3.01, "learning_rate": 8.266271402466463e-06, "loss": 0.3582, "step": 29858 }, { "epoch": 3.01, "learning_rate": 8.264621813971636e-06, "loss": 0.2579, "step": 29859 }, { "epoch": 3.01, "learning_rate": 8.262972363792488e-06, "loss": 0.0303, "step": 29860 }, { "epoch": 3.01, "learning_rate": 8.261323051939513e-06, "loss": 0.158, "step": 29861 }, { "epoch": 3.01, "learning_rate": 8.25967387842321e-06, "loss": 0.1015, "step": 29862 }, { "epoch": 3.01, "learning_rate": 8.258024843254071e-06, "loss": 0.0534, "step": 29863 }, { "epoch": 3.01, "learning_rate": 8.25637594644258e-06, "loss": 0.3691, "step": 29864 }, { "epoch": 3.01, "learning_rate": 8.254727187999264e-06, "loss": 0.5372, "step": 29865 }, { "epoch": 3.01, "learning_rate": 8.253078567934579e-06, "loss": 0.3291, "step": 29866 }, { "epoch": 3.01, "learning_rate": 8.25143008625902e-06, "loss": 0.5108, "step": 29867 }, { "epoch": 3.01, "learning_rate": 8.249781742983105e-06, "loss": 0.21, "step": 29868 }, { "epoch": 3.01, "learning_rate": 8.248133538117296e-06, "loss": 0.7707, "step": 29869 }, { "epoch": 3.01, "learning_rate": 8.246485471672081e-06, "loss": 0.2302, "step": 29870 }, { "epoch": 3.01, "learning_rate": 8.244837543657966e-06, "loss": 0.172, "step": 29871 }, { "epoch": 3.01, "learning_rate": 8.24318975408543e-06, "loss": 0.3593, "step": 29872 }, { "epoch": 3.01, "learning_rate": 8.241542102964956e-06, "loss": 0.0955, "step": 29873 }, { "epoch": 3.01, "learning_rate": 8.239894590307031e-06, "loss": 0.2495, "step": 29874 }, { "epoch": 3.01, "learning_rate": 8.238247216122139e-06, "loss": 0.1712, "step": 29875 }, { "epoch": 3.01, "learning_rate": 8.236599980420764e-06, "loss": 0.048, "step": 29876 }, { "epoch": 3.01, "learning_rate": 8.234952883213387e-06, "loss": 0.1245, "step": 29877 }, { "epoch": 3.01, "learning_rate": 8.23330592451048e-06, "loss": 0.2012, "step": 29878 }, { "epoch": 3.01, "learning_rate": 8.231659104322554e-06, "loss": 0.6323, "step": 29879 }, { "epoch": 3.01, "learning_rate": 8.230012422660058e-06, "loss": 0.0885, "step": 29880 }, { "epoch": 3.01, "learning_rate": 8.228365879533471e-06, "loss": 0.3662, "step": 29881 }, { "epoch": 3.01, "learning_rate": 8.226719474953302e-06, "loss": 0.4823, "step": 29882 }, { "epoch": 3.01, "learning_rate": 8.225073208929995e-06, "loss": 0.5243, "step": 29883 }, { "epoch": 3.01, "learning_rate": 8.223427081474033e-06, "loss": 0.138, "step": 29884 }, { "epoch": 3.01, "learning_rate": 8.221781092595903e-06, "loss": 0.3853, "step": 29885 }, { "epoch": 3.01, "learning_rate": 8.220135242306082e-06, "loss": 0.2837, "step": 29886 }, { "epoch": 3.01, "learning_rate": 8.218489530615019e-06, "loss": 0.3595, "step": 29887 }, { "epoch": 3.01, "learning_rate": 8.21684395753321e-06, "loss": 1.0154, "step": 29888 }, { "epoch": 3.01, "learning_rate": 8.21519852307112e-06, "loss": 0.3438, "step": 29889 }, { "epoch": 3.01, "learning_rate": 8.213553227239218e-06, "loss": 0.2399, "step": 29890 }, { "epoch": 3.01, "learning_rate": 8.211908070047978e-06, "loss": 0.6174, "step": 29891 }, { "epoch": 3.01, "learning_rate": 8.210263051507864e-06, "loss": 0.7038, "step": 29892 }, { "epoch": 3.01, "learning_rate": 8.208618171629347e-06, "loss": 0.2136, "step": 29893 }, { "epoch": 3.01, "learning_rate": 8.206973430422896e-06, "loss": 0.3337, "step": 29894 }, { "epoch": 3.01, "learning_rate": 8.205328827898972e-06, "loss": 0.1035, "step": 29895 }, { "epoch": 3.01, "learning_rate": 8.203684364068037e-06, "loss": 0.029, "step": 29896 }, { "epoch": 3.01, "learning_rate": 8.202040038940583e-06, "loss": 0.155, "step": 29897 }, { "epoch": 3.01, "learning_rate": 8.200395852527043e-06, "loss": 0.0264, "step": 29898 }, { "epoch": 3.01, "learning_rate": 8.198751804837884e-06, "loss": 1.1022, "step": 29899 }, { "epoch": 3.01, "learning_rate": 8.197107895883591e-06, "loss": 0.1034, "step": 29900 }, { "epoch": 3.01, "learning_rate": 8.19546412567459e-06, "loss": 0.4237, "step": 29901 }, { "epoch": 3.01, "learning_rate": 8.193820494221369e-06, "loss": 0.2502, "step": 29902 }, { "epoch": 3.01, "learning_rate": 8.19217700153438e-06, "loss": 0.1171, "step": 29903 }, { "epoch": 3.01, "learning_rate": 8.19053364762408e-06, "loss": 0.4026, "step": 29904 }, { "epoch": 3.01, "learning_rate": 8.18889043250093e-06, "loss": 0.5393, "step": 29905 }, { "epoch": 3.01, "learning_rate": 8.18724735617538e-06, "loss": 0.1646, "step": 29906 }, { "epoch": 3.01, "learning_rate": 8.185604418657895e-06, "loss": 0.1407, "step": 29907 }, { "epoch": 3.01, "learning_rate": 8.18396161995892e-06, "loss": 0.196, "step": 29908 }, { "epoch": 3.0, "learning_rate": 8.182318960088915e-06, "loss": 0.1068, "step": 29909 }, { "epoch": 3.0, "learning_rate": 8.180676439058326e-06, "loss": 0.0831, "step": 29910 }, { "epoch": 3.0, "learning_rate": 8.179034056877626e-06, "loss": 0.0631, "step": 29911 }, { "epoch": 3.0, "learning_rate": 8.177391813557244e-06, "loss": 0.165, "step": 29912 }, { "epoch": 3.0, "learning_rate": 8.175749709107631e-06, "loss": 0.2994, "step": 29913 }, { "epoch": 3.0, "learning_rate": 8.17410774353926e-06, "loss": 0.2369, "step": 29914 }, { "epoch": 3.0, "learning_rate": 8.172465916862556e-06, "loss": 0.2198, "step": 29915 }, { "epoch": 3.0, "learning_rate": 8.17082422908797e-06, "loss": 0.0441, "step": 29916 }, { "epoch": 3.0, "learning_rate": 8.16918268022596e-06, "loss": 0.5609, "step": 29917 }, { "epoch": 3.0, "learning_rate": 8.167541270286978e-06, "loss": 0.1044, "step": 29918 }, { "epoch": 3.0, "learning_rate": 8.165899999281437e-06, "loss": 0.0508, "step": 29919 }, { "epoch": 3.0, "learning_rate": 8.164258867219813e-06, "loss": 0.1576, "step": 29920 }, { "epoch": 3.0, "learning_rate": 8.162617874112538e-06, "loss": 0.1228, "step": 29921 }, { "epoch": 3.0, "learning_rate": 8.160977019970054e-06, "loss": 0.2489, "step": 29922 }, { "epoch": 3.0, "learning_rate": 8.159336304802807e-06, "loss": 0.4035, "step": 29923 }, { "epoch": 3.0, "learning_rate": 8.157695728621236e-06, "loss": 0.0297, "step": 29924 }, { "epoch": 3.0, "learning_rate": 8.156055291435778e-06, "loss": 0.2569, "step": 29925 }, { "epoch": 3.0, "learning_rate": 8.154414993256873e-06, "loss": 0.3714, "step": 29926 }, { "epoch": 3.0, "learning_rate": 8.152774834094954e-06, "loss": 0.2041, "step": 29927 }, { "epoch": 3.0, "learning_rate": 8.151134813960473e-06, "loss": 0.0643, "step": 29928 }, { "epoch": 3.0, "learning_rate": 8.149494932863867e-06, "loss": 0.3204, "step": 29929 }, { "epoch": 3.0, "learning_rate": 8.147855190815546e-06, "loss": 0.372, "step": 29930 }, { "epoch": 3.0, "learning_rate": 8.146215587825967e-06, "loss": 0.0478, "step": 29931 }, { "epoch": 3.0, "learning_rate": 8.144576123905571e-06, "loss": 0.2334, "step": 29932 }, { "epoch": 3.0, "learning_rate": 8.142936799064761e-06, "loss": 0.5406, "step": 29933 }, { "epoch": 3.0, "learning_rate": 8.141297613313993e-06, "loss": 0.0308, "step": 29934 }, { "epoch": 3.0, "learning_rate": 8.139658566663693e-06, "loss": 0.0716, "step": 29935 }, { "epoch": 3.0, "learning_rate": 8.138019659124288e-06, "loss": 0.3982, "step": 29936 }, { "epoch": 3.0, "learning_rate": 8.136380890706209e-06, "loss": 0.3097, "step": 29937 }, { "epoch": 3.0, "learning_rate": 8.134742261419889e-06, "loss": 0.241, "step": 29938 }, { "epoch": 3.0, "learning_rate": 8.133103771275747e-06, "loss": 0.1116, "step": 29939 }, { "epoch": 3.0, "learning_rate": 8.131465420284214e-06, "loss": 0.35, "step": 29940 }, { "epoch": 3.0, "learning_rate": 8.129827208455718e-06, "loss": 0.0907, "step": 29941 }, { "epoch": 3.0, "learning_rate": 8.128189135800672e-06, "loss": 0.7136, "step": 29942 }, { "epoch": 3.0, "learning_rate": 8.126551202329527e-06, "loss": 0.5071, "step": 29943 }, { "epoch": 3.0, "learning_rate": 8.124913408052682e-06, "loss": 0.1485, "step": 29944 }, { "epoch": 3.0, "learning_rate": 8.123275752980554e-06, "loss": 0.1483, "step": 29945 }, { "epoch": 3.0, "learning_rate": 8.121638237123596e-06, "loss": 0.1916, "step": 29946 }, { "epoch": 3.0, "learning_rate": 8.120000860492203e-06, "loss": 0.284, "step": 29947 }, { "epoch": 3.0, "learning_rate": 8.118363623096791e-06, "loss": 0.1085, "step": 29948 }, { "epoch": 3.0, "learning_rate": 8.116726524947796e-06, "loss": 0.1383, "step": 29949 }, { "epoch": 3.0, "learning_rate": 8.115089566055638e-06, "loss": 0.288, "step": 29950 }, { "epoch": 3.0, "learning_rate": 8.113452746430708e-06, "loss": 0.4437, "step": 29951 }, { "epoch": 3.0, "learning_rate": 8.111816066083445e-06, "loss": 0.1295, "step": 29952 }, { "epoch": 3.0, "learning_rate": 8.11017952502426e-06, "loss": 0.4996, "step": 29953 }, { "epoch": 3.0, "learning_rate": 8.108543123263566e-06, "loss": 0.2102, "step": 29954 }, { "epoch": 3.0, "learning_rate": 8.106906860811775e-06, "loss": 0.2037, "step": 29955 }, { "epoch": 3.0, "learning_rate": 8.10527073767929e-06, "loss": 0.2176, "step": 29956 }, { "epoch": 3.0, "learning_rate": 8.103634753876551e-06, "loss": 0.1041, "step": 29957 }, { "epoch": 3.01, "learning_rate": 8.101998909413943e-06, "loss": 0.0744, "step": 29958 }, { "epoch": 3.01, "learning_rate": 8.100363204301872e-06, "loss": 0.4613, "step": 29959 }, { "epoch": 3.01, "learning_rate": 8.098727638550771e-06, "loss": 0.0604, "step": 29960 }, { "epoch": 3.01, "learning_rate": 8.09709221217104e-06, "loss": 0.1804, "step": 29961 }, { "epoch": 3.01, "learning_rate": 8.095456925173065e-06, "loss": 0.0787, "step": 29962 }, { "epoch": 3.01, "learning_rate": 8.093821777567283e-06, "loss": 0.1732, "step": 29963 }, { "epoch": 3.01, "learning_rate": 8.092186769364089e-06, "loss": 0.2503, "step": 29964 }, { "epoch": 3.01, "learning_rate": 8.09055190057387e-06, "loss": 0.2041, "step": 29965 }, { "epoch": 3.01, "learning_rate": 8.08891717120705e-06, "loss": 0.4062, "step": 29966 }, { "epoch": 3.01, "learning_rate": 8.08728258127403e-06, "loss": 0.0954, "step": 29967 }, { "epoch": 3.01, "learning_rate": 8.085648130785203e-06, "loss": 0.1583, "step": 29968 }, { "epoch": 3.01, "learning_rate": 8.08401381975098e-06, "loss": 0.3124, "step": 29969 }, { "epoch": 3.01, "learning_rate": 8.082379648181751e-06, "loss": 0.0233, "step": 29970 }, { "epoch": 3.01, "learning_rate": 8.080745616087922e-06, "loss": 0.1557, "step": 29971 }, { "epoch": 3.01, "learning_rate": 8.079111723479887e-06, "loss": 0.2059, "step": 29972 }, { "epoch": 3.01, "learning_rate": 8.077477970368048e-06, "loss": 0.4096, "step": 29973 }, { "epoch": 3.01, "learning_rate": 8.075844356762792e-06, "loss": 0.4408, "step": 29974 }, { "epoch": 3.01, "learning_rate": 8.074210882674537e-06, "loss": 0.0697, "step": 29975 }, { "epoch": 3.01, "learning_rate": 8.072577548113655e-06, "loss": 0.05, "step": 29976 }, { "epoch": 3.01, "learning_rate": 8.070944353090537e-06, "loss": 0.0438, "step": 29977 }, { "epoch": 3.01, "learning_rate": 8.06931129761561e-06, "loss": 0.1463, "step": 29978 }, { "epoch": 3.01, "learning_rate": 8.06767838169923e-06, "loss": 0.1152, "step": 29979 }, { "epoch": 3.01, "learning_rate": 8.066045605351793e-06, "loss": 0.03, "step": 29980 }, { "epoch": 3.01, "learning_rate": 8.064412968583708e-06, "loss": 0.0691, "step": 29981 }, { "epoch": 3.01, "learning_rate": 8.062780471405352e-06, "loss": 0.111, "step": 29982 }, { "epoch": 3.01, "learning_rate": 8.061148113827114e-06, "loss": 0.5476, "step": 29983 }, { "epoch": 3.01, "learning_rate": 8.059515895859388e-06, "loss": 0.1325, "step": 29984 }, { "epoch": 3.01, "learning_rate": 8.057883817512554e-06, "loss": 0.1462, "step": 29985 }, { "epoch": 3.01, "learning_rate": 8.056251878797002e-06, "loss": 0.4056, "step": 29986 }, { "epoch": 3.01, "learning_rate": 8.054620079723114e-06, "loss": 0.247, "step": 29987 }, { "epoch": 3.01, "learning_rate": 8.052988420301266e-06, "loss": 0.1482, "step": 29988 }, { "epoch": 3.01, "learning_rate": 8.05135690054187e-06, "loss": 0.1004, "step": 29989 }, { "epoch": 3.01, "learning_rate": 8.049725520455276e-06, "loss": 0.0452, "step": 29990 }, { "epoch": 3.01, "learning_rate": 8.048094280051874e-06, "loss": 0.2054, "step": 29991 }, { "epoch": 3.01, "learning_rate": 8.046463179342058e-06, "loss": 0.6622, "step": 29992 }, { "epoch": 3.01, "learning_rate": 8.044832218336204e-06, "loss": 0.3181, "step": 29993 }, { "epoch": 3.01, "learning_rate": 8.043201397044672e-06, "loss": 0.0688, "step": 29994 }, { "epoch": 3.01, "learning_rate": 8.041570715477857e-06, "loss": 0.2325, "step": 29995 }, { "epoch": 3.01, "learning_rate": 8.039940173646147e-06, "loss": 0.5188, "step": 29996 }, { "epoch": 3.01, "learning_rate": 8.038309771559884e-06, "loss": 0.1845, "step": 29997 }, { "epoch": 3.01, "learning_rate": 8.036679509229472e-06, "loss": 0.2509, "step": 29998 }, { "epoch": 3.01, "learning_rate": 8.035049386665274e-06, "loss": 0.0726, "step": 29999 }, { "epoch": 3.01, "learning_rate": 8.03341940387767e-06, "loss": 0.2181, "step": 30000 }, { "epoch": 3.01, "learning_rate": 8.031789560877025e-06, "loss": 0.0355, "step": 30001 }, { "epoch": 3.01, "learning_rate": 8.030159857673713e-06, "loss": 0.2505, "step": 30002 }, { "epoch": 3.01, "learning_rate": 8.028530294278107e-06, "loss": 0.2352, "step": 30003 }, { "epoch": 3.01, "learning_rate": 8.02690087070058e-06, "loss": 0.1748, "step": 30004 }, { "epoch": 3.01, "learning_rate": 8.025271586951494e-06, "loss": 0.0395, "step": 30005 }, { "epoch": 3.01, "learning_rate": 8.023642443041212e-06, "loss": 0.1818, "step": 30006 }, { "epoch": 3.01, "learning_rate": 8.022013438980125e-06, "loss": 0.387, "step": 30007 }, { "epoch": 3.01, "learning_rate": 8.020384574778565e-06, "loss": 0.1469, "step": 30008 }, { "epoch": 3.01, "learning_rate": 8.018755850446928e-06, "loss": 0.4798, "step": 30009 }, { "epoch": 3.01, "learning_rate": 8.017127265995575e-06, "loss": 0.0893, "step": 30010 }, { "epoch": 3.01, "learning_rate": 8.015498821434842e-06, "loss": 0.061, "step": 30011 }, { "epoch": 3.01, "learning_rate": 8.013870516775122e-06, "loss": 0.2764, "step": 30012 }, { "epoch": 3.01, "learning_rate": 8.012242352026763e-06, "loss": 0.1352, "step": 30013 }, { "epoch": 3.01, "learning_rate": 8.010614327200131e-06, "loss": 0.2485, "step": 30014 }, { "epoch": 3.01, "learning_rate": 8.008986442305583e-06, "loss": 0.3858, "step": 30015 }, { "epoch": 3.01, "learning_rate": 8.007358697353485e-06, "loss": 0.0631, "step": 30016 }, { "epoch": 3.01, "learning_rate": 8.005731092354184e-06, "loss": 0.1754, "step": 30017 }, { "epoch": 3.01, "learning_rate": 8.004103627318044e-06, "loss": 0.9104, "step": 30018 }, { "epoch": 3.01, "learning_rate": 8.002476302255421e-06, "loss": 0.0936, "step": 30019 }, { "epoch": 3.01, "learning_rate": 8.000849117176663e-06, "loss": 0.0958, "step": 30020 }, { "epoch": 3.01, "learning_rate": 7.999222072092151e-06, "loss": 0.1053, "step": 30021 }, { "epoch": 3.01, "learning_rate": 7.99759516701221e-06, "loss": 0.032, "step": 30022 }, { "epoch": 3.01, "learning_rate": 7.995968401947195e-06, "loss": 0.2968, "step": 30023 }, { "epoch": 3.01, "learning_rate": 7.994341776907472e-06, "loss": 0.1731, "step": 30024 }, { "epoch": 3.01, "learning_rate": 7.992715291903398e-06, "loss": 0.1387, "step": 30025 }, { "epoch": 3.01, "learning_rate": 7.991088946945295e-06, "loss": 0.2442, "step": 30026 }, { "epoch": 3.01, "learning_rate": 7.989462742043537e-06, "loss": 0.4055, "step": 30027 }, { "epoch": 3.01, "learning_rate": 7.987836677208473e-06, "loss": 0.4877, "step": 30028 }, { "epoch": 3.01, "learning_rate": 7.986210752450425e-06, "loss": 0.2543, "step": 30029 }, { "epoch": 3.01, "learning_rate": 7.984584967779768e-06, "loss": 0.1083, "step": 30030 }, { "epoch": 3.01, "learning_rate": 7.982959323206837e-06, "loss": 0.0762, "step": 30031 }, { "epoch": 3.01, "learning_rate": 7.981333818741975e-06, "loss": 0.2358, "step": 30032 }, { "epoch": 3.01, "learning_rate": 7.979708454395528e-06, "loss": 0.019, "step": 30033 }, { "epoch": 3.01, "learning_rate": 7.97808323017784e-06, "loss": 0.115, "step": 30034 }, { "epoch": 3.01, "learning_rate": 7.976458146099256e-06, "loss": 0.1438, "step": 30035 }, { "epoch": 3.01, "learning_rate": 7.974833202170112e-06, "loss": 0.1366, "step": 30036 }, { "epoch": 3.01, "learning_rate": 7.973208398400741e-06, "loss": 0.6074, "step": 30037 }, { "epoch": 3.01, "learning_rate": 7.9715837348015e-06, "loss": 0.079, "step": 30038 }, { "epoch": 3.01, "learning_rate": 7.969959211382731e-06, "loss": 0.0452, "step": 30039 }, { "epoch": 3.01, "learning_rate": 7.968334828154744e-06, "loss": 0.1384, "step": 30040 }, { "epoch": 3.01, "learning_rate": 7.9667105851279e-06, "loss": 0.2447, "step": 30041 }, { "epoch": 3.01, "learning_rate": 7.96508648231254e-06, "loss": 0.116, "step": 30042 }, { "epoch": 3.01, "learning_rate": 7.963462519718968e-06, "loss": 0.0339, "step": 30043 }, { "epoch": 3.01, "learning_rate": 7.961838697357543e-06, "loss": 0.1292, "step": 30044 }, { "epoch": 3.01, "learning_rate": 7.960215015238593e-06, "loss": 0.4009, "step": 30045 }, { "epoch": 3.01, "learning_rate": 7.958591473372452e-06, "loss": 0.0951, "step": 30046 }, { "epoch": 3.01, "learning_rate": 7.956968071769453e-06, "loss": 0.3598, "step": 30047 }, { "epoch": 3.01, "learning_rate": 7.955344810439921e-06, "loss": 0.0798, "step": 30048 }, { "epoch": 3.01, "learning_rate": 7.953721689394185e-06, "loss": 0.2573, "step": 30049 }, { "epoch": 3.01, "learning_rate": 7.952098708642584e-06, "loss": 0.0571, "step": 30050 }, { "epoch": 3.01, "learning_rate": 7.950475868195433e-06, "loss": 0.4335, "step": 30051 }, { "epoch": 3.01, "learning_rate": 7.948853168063062e-06, "loss": 0.1365, "step": 30052 }, { "epoch": 3.01, "learning_rate": 7.947230608255816e-06, "loss": 0.5312, "step": 30053 }, { "epoch": 3.01, "learning_rate": 7.945608188783995e-06, "loss": 0.2996, "step": 30054 }, { "epoch": 3.01, "learning_rate": 7.943985909657924e-06, "loss": 0.1069, "step": 30055 }, { "epoch": 3.02, "learning_rate": 7.942363770887947e-06, "loss": 0.1843, "step": 30056 }, { "epoch": 3.02, "learning_rate": 7.940741772484384e-06, "loss": 0.04, "step": 30057 }, { "epoch": 3.02, "learning_rate": 7.939119914457532e-06, "loss": 0.5107, "step": 30058 }, { "epoch": 3.02, "learning_rate": 7.937498196817739e-06, "loss": 0.5315, "step": 30059 }, { "epoch": 3.02, "learning_rate": 7.935876619575321e-06, "loss": 0.0656, "step": 30060 }, { "epoch": 3.02, "learning_rate": 7.93425518274057e-06, "loss": 0.1331, "step": 30061 }, { "epoch": 3.02, "learning_rate": 7.93263388632384e-06, "loss": 0.1218, "step": 30062 }, { "epoch": 3.02, "learning_rate": 7.931012730335428e-06, "loss": 0.0328, "step": 30063 }, { "epoch": 3.02, "learning_rate": 7.929391714785657e-06, "loss": 0.3092, "step": 30064 }, { "epoch": 3.02, "learning_rate": 7.927770839684842e-06, "loss": 0.1836, "step": 30065 }, { "epoch": 3.02, "learning_rate": 7.926150105043285e-06, "loss": 0.2238, "step": 30066 }, { "epoch": 3.02, "learning_rate": 7.924529510871333e-06, "loss": 0.0777, "step": 30067 }, { "epoch": 3.02, "learning_rate": 7.922909057179265e-06, "loss": 0.2, "step": 30068 }, { "epoch": 3.02, "learning_rate": 7.921288743977398e-06, "loss": 0.0664, "step": 30069 }, { "epoch": 3.02, "learning_rate": 7.91966857127606e-06, "loss": 0.6984, "step": 30070 }, { "epoch": 3.02, "learning_rate": 7.918048539085558e-06, "loss": 0.5045, "step": 30071 }, { "epoch": 3.02, "learning_rate": 7.91642864741618e-06, "loss": 0.1868, "step": 30072 }, { "epoch": 3.02, "learning_rate": 7.914808896278258e-06, "loss": 0.2942, "step": 30073 }, { "epoch": 3.02, "learning_rate": 7.913189285682096e-06, "loss": 0.1495, "step": 30074 }, { "epoch": 3.02, "learning_rate": 7.911569815637979e-06, "loss": 0.1841, "step": 30075 }, { "epoch": 3.02, "learning_rate": 7.909950486156236e-06, "loss": 0.299, "step": 30076 }, { "epoch": 3.02, "learning_rate": 7.908331297247168e-06, "loss": 0.1203, "step": 30077 }, { "epoch": 3.02, "learning_rate": 7.906712248921073e-06, "loss": 0.2503, "step": 30078 }, { "epoch": 3.02, "learning_rate": 7.905093341188258e-06, "loss": 0.2336, "step": 30079 }, { "epoch": 3.02, "learning_rate": 7.903474574059023e-06, "loss": 0.1617, "step": 30080 }, { "epoch": 3.02, "learning_rate": 7.901855947543669e-06, "loss": 0.1765, "step": 30081 }, { "epoch": 3.02, "learning_rate": 7.900237461652498e-06, "loss": 0.0537, "step": 30082 }, { "epoch": 3.02, "learning_rate": 7.89861911639581e-06, "loss": 0.2471, "step": 30083 }, { "epoch": 3.02, "learning_rate": 7.897000911783892e-06, "loss": 0.4459, "step": 30084 }, { "epoch": 3.02, "learning_rate": 7.895382847827071e-06, "loss": 0.0873, "step": 30085 }, { "epoch": 3.02, "learning_rate": 7.893764924535612e-06, "loss": 0.2403, "step": 30086 }, { "epoch": 3.02, "learning_rate": 7.892147141919815e-06, "loss": 0.0706, "step": 30087 }, { "epoch": 3.02, "learning_rate": 7.890529499989994e-06, "loss": 0.1716, "step": 30088 }, { "epoch": 3.02, "learning_rate": 7.888911998756435e-06, "loss": 0.3049, "step": 30089 }, { "epoch": 3.02, "learning_rate": 7.887294638229426e-06, "loss": 0.1092, "step": 30090 }, { "epoch": 3.02, "learning_rate": 7.88567741841926e-06, "loss": 0.1273, "step": 30091 }, { "epoch": 3.02, "learning_rate": 7.884060339336232e-06, "loss": 0.0697, "step": 30092 }, { "epoch": 3.02, "learning_rate": 7.882443400990632e-06, "loss": 0.1988, "step": 30093 }, { "epoch": 3.02, "learning_rate": 7.880826603392746e-06, "loss": 0.3447, "step": 30094 }, { "epoch": 3.02, "learning_rate": 7.879209946552865e-06, "loss": 0.239, "step": 30095 }, { "epoch": 3.02, "learning_rate": 7.877593430481276e-06, "loss": 0.0813, "step": 30096 }, { "epoch": 3.02, "learning_rate": 7.875977055188268e-06, "loss": 0.1836, "step": 30097 }, { "epoch": 3.02, "learning_rate": 7.874360820684116e-06, "loss": 0.0607, "step": 30098 }, { "epoch": 3.02, "learning_rate": 7.872744726979134e-06, "loss": 0.0987, "step": 30099 }, { "epoch": 3.02, "learning_rate": 7.87112877408358e-06, "loss": 0.1702, "step": 30100 }, { "epoch": 3.02, "learning_rate": 7.86951296200773e-06, "loss": 0.1009, "step": 30101 }, { "epoch": 3.02, "learning_rate": 7.867897290761892e-06, "loss": 0.1033, "step": 30102 }, { "epoch": 3.02, "learning_rate": 7.866281760356346e-06, "loss": 0.1612, "step": 30103 }, { "epoch": 3.02, "learning_rate": 7.864666370801346e-06, "loss": 0.3744, "step": 30104 }, { "epoch": 3.02, "learning_rate": 7.863051122107194e-06, "loss": 0.1403, "step": 30105 }, { "epoch": 3.02, "learning_rate": 7.861436014284175e-06, "loss": 0.4063, "step": 30106 }, { "epoch": 3.02, "learning_rate": 7.859821047342535e-06, "loss": 0.2414, "step": 30107 }, { "epoch": 3.02, "learning_rate": 7.858206221292584e-06, "loss": 0.167, "step": 30108 }, { "epoch": 3.02, "learning_rate": 7.856591536144582e-06, "loss": 0.3889, "step": 30109 }, { "epoch": 3.02, "learning_rate": 7.854976991908808e-06, "loss": 0.0755, "step": 30110 }, { "epoch": 3.02, "learning_rate": 7.853362588595536e-06, "loss": 0.2352, "step": 30111 }, { "epoch": 3.02, "learning_rate": 7.851748326215039e-06, "loss": 0.2296, "step": 30112 }, { "epoch": 3.02, "learning_rate": 7.850134204777592e-06, "loss": 0.0956, "step": 30113 }, { "epoch": 3.02, "learning_rate": 7.848520224293465e-06, "loss": 0.1516, "step": 30114 }, { "epoch": 3.02, "learning_rate": 7.846906384772925e-06, "loss": 0.1577, "step": 30115 }, { "epoch": 3.02, "learning_rate": 7.845292686226242e-06, "loss": 0.4195, "step": 30116 }, { "epoch": 3.02, "learning_rate": 7.843679128663702e-06, "loss": 0.0352, "step": 30117 }, { "epoch": 3.02, "learning_rate": 7.84206571209554e-06, "loss": 0.1124, "step": 30118 }, { "epoch": 3.02, "learning_rate": 7.840452436532056e-06, "loss": 0.3581, "step": 30119 }, { "epoch": 3.02, "learning_rate": 7.838839301983499e-06, "loss": 0.1719, "step": 30120 }, { "epoch": 3.02, "learning_rate": 7.837226308460137e-06, "loss": 0.1871, "step": 30121 }, { "epoch": 3.02, "learning_rate": 7.83561345597224e-06, "loss": 0.0659, "step": 30122 }, { "epoch": 3.02, "learning_rate": 7.834000744530065e-06, "loss": 0.0623, "step": 30123 }, { "epoch": 3.02, "learning_rate": 7.832388174143879e-06, "loss": 0.2392, "step": 30124 }, { "epoch": 3.02, "learning_rate": 7.830775744823937e-06, "loss": 0.3723, "step": 30125 }, { "epoch": 3.02, "learning_rate": 7.829163456580508e-06, "loss": 0.1065, "step": 30126 }, { "epoch": 3.02, "learning_rate": 7.82755130942385e-06, "loss": 0.0119, "step": 30127 }, { "epoch": 3.02, "learning_rate": 7.825939303364217e-06, "loss": 0.5019, "step": 30128 }, { "epoch": 3.02, "learning_rate": 7.824327438411873e-06, "loss": 0.1191, "step": 30129 }, { "epoch": 3.02, "learning_rate": 7.822715714577062e-06, "loss": 0.1823, "step": 30130 }, { "epoch": 3.02, "learning_rate": 7.821104131870072e-06, "loss": 0.2801, "step": 30131 }, { "epoch": 3.02, "learning_rate": 7.819492690301128e-06, "loss": 0.6678, "step": 30132 }, { "epoch": 3.02, "learning_rate": 7.817881389880485e-06, "loss": 0.0938, "step": 30133 }, { "epoch": 3.02, "learning_rate": 7.816270230618418e-06, "loss": 0.1774, "step": 30134 }, { "epoch": 3.02, "learning_rate": 7.814659212525175e-06, "loss": 0.0578, "step": 30135 }, { "epoch": 3.02, "learning_rate": 7.813048335610984e-06, "loss": 0.1505, "step": 30136 }, { "epoch": 3.02, "learning_rate": 7.81143759988612e-06, "loss": 0.0788, "step": 30137 }, { "epoch": 3.02, "learning_rate": 7.809827005360836e-06, "loss": 0.263, "step": 30138 }, { "epoch": 3.02, "learning_rate": 7.808216552045356e-06, "loss": 0.3459, "step": 30139 }, { "epoch": 3.02, "learning_rate": 7.80660623994995e-06, "loss": 0.1272, "step": 30140 }, { "epoch": 3.02, "learning_rate": 7.80499606908486e-06, "loss": 0.1354, "step": 30141 }, { "epoch": 3.02, "learning_rate": 7.803386039460331e-06, "loss": 0.0731, "step": 30142 }, { "epoch": 3.02, "learning_rate": 7.801776151086608e-06, "loss": 0.0756, "step": 30143 }, { "epoch": 3.02, "learning_rate": 7.80016640397393e-06, "loss": 0.1092, "step": 30144 }, { "epoch": 3.02, "learning_rate": 7.798556798132564e-06, "loss": 0.1772, "step": 30145 }, { "epoch": 3.02, "learning_rate": 7.796947333572732e-06, "loss": 0.4578, "step": 30146 }, { "epoch": 3.02, "learning_rate": 7.795338010304666e-06, "loss": 0.1791, "step": 30147 }, { "epoch": 3.02, "learning_rate": 7.793728828338631e-06, "loss": 0.3789, "step": 30148 }, { "epoch": 3.02, "learning_rate": 7.792119787684872e-06, "loss": 0.2928, "step": 30149 }, { "epoch": 3.02, "learning_rate": 7.790510888353593e-06, "loss": 0.0894, "step": 30150 }, { "epoch": 3.02, "learning_rate": 7.788902130355063e-06, "loss": 0.0472, "step": 30151 }, { "epoch": 3.02, "learning_rate": 7.78729351369951e-06, "loss": 0.2313, "step": 30152 }, { "epoch": 3.02, "learning_rate": 7.785685038397174e-06, "loss": 0.3022, "step": 30153 }, { "epoch": 3.02, "learning_rate": 7.784076704458285e-06, "loss": 0.07, "step": 30154 }, { "epoch": 3.03, "learning_rate": 7.78246851189308e-06, "loss": 0.4566, "step": 30155 }, { "epoch": 3.03, "learning_rate": 7.780860460711795e-06, "loss": 0.202, "step": 30156 }, { "epoch": 3.03, "learning_rate": 7.779252550924663e-06, "loss": 0.2119, "step": 30157 }, { "epoch": 3.03, "learning_rate": 7.777644782541912e-06, "loss": 0.1733, "step": 30158 }, { "epoch": 3.03, "learning_rate": 7.776037155573775e-06, "loss": 0.1466, "step": 30159 }, { "epoch": 3.03, "learning_rate": 7.774429670030485e-06, "loss": 0.4831, "step": 30160 }, { "epoch": 3.03, "learning_rate": 7.772822325922267e-06, "loss": 0.1723, "step": 30161 }, { "epoch": 3.03, "learning_rate": 7.771215123259347e-06, "loss": 0.1722, "step": 30162 }, { "epoch": 3.03, "learning_rate": 7.769608062051971e-06, "loss": 0.1349, "step": 30163 }, { "epoch": 3.03, "learning_rate": 7.768001142310346e-06, "loss": 0.0791, "step": 30164 }, { "epoch": 3.03, "learning_rate": 7.766394364044694e-06, "loss": 0.4621, "step": 30165 }, { "epoch": 3.03, "learning_rate": 7.764787727265262e-06, "loss": 0.058, "step": 30166 }, { "epoch": 3.03, "learning_rate": 7.76318123198227e-06, "loss": 0.3606, "step": 30167 }, { "epoch": 3.03, "learning_rate": 7.761574878205914e-06, "loss": 0.1453, "step": 30168 }, { "epoch": 3.03, "learning_rate": 7.759968665946447e-06, "loss": 0.1579, "step": 30169 }, { "epoch": 3.03, "learning_rate": 7.758362595214075e-06, "loss": 0.1833, "step": 30170 }, { "epoch": 3.03, "learning_rate": 7.756756666019025e-06, "loss": 0.2039, "step": 30171 }, { "epoch": 3.03, "learning_rate": 7.755150878371516e-06, "loss": 0.2561, "step": 30172 }, { "epoch": 3.03, "learning_rate": 7.75354523228176e-06, "loss": 0.6562, "step": 30173 }, { "epoch": 3.03, "learning_rate": 7.751939727759981e-06, "loss": 0.0635, "step": 30174 }, { "epoch": 3.03, "learning_rate": 7.750334364816393e-06, "loss": 0.2723, "step": 30175 }, { "epoch": 3.03, "learning_rate": 7.748729143461205e-06, "loss": 0.3527, "step": 30176 }, { "epoch": 3.03, "learning_rate": 7.74712406370466e-06, "loss": 0.2199, "step": 30177 }, { "epoch": 3.03, "learning_rate": 7.745519125556938e-06, "loss": 0.0916, "step": 30178 }, { "epoch": 3.03, "learning_rate": 7.74391432902826e-06, "loss": 0.1741, "step": 30179 }, { "epoch": 3.03, "learning_rate": 7.742309674128854e-06, "loss": 0.1379, "step": 30180 }, { "epoch": 3.03, "learning_rate": 7.74070516086893e-06, "loss": 0.2467, "step": 30181 }, { "epoch": 3.03, "learning_rate": 7.739100789258673e-06, "loss": 0.1056, "step": 30182 }, { "epoch": 3.03, "learning_rate": 7.737496559308319e-06, "loss": 0.3804, "step": 30183 }, { "epoch": 3.03, "learning_rate": 7.735892471028068e-06, "loss": 0.388, "step": 30184 }, { "epoch": 3.03, "learning_rate": 7.734288524428127e-06, "loss": 0.2208, "step": 30185 }, { "epoch": 3.03, "learning_rate": 7.732684719518706e-06, "loss": 0.2513, "step": 30186 }, { "epoch": 3.03, "learning_rate": 7.731081056310006e-06, "loss": 0.1474, "step": 30187 }, { "epoch": 3.03, "learning_rate": 7.729477534812235e-06, "loss": 0.3087, "step": 30188 }, { "epoch": 3.03, "learning_rate": 7.727874155035596e-06, "loss": 0.0687, "step": 30189 }, { "epoch": 3.03, "learning_rate": 7.726270916990297e-06, "loss": 0.1302, "step": 30190 }, { "epoch": 3.03, "learning_rate": 7.724667820686533e-06, "loss": 0.4655, "step": 30191 }, { "epoch": 3.03, "learning_rate": 7.72306486613451e-06, "loss": 0.3455, "step": 30192 }, { "epoch": 3.03, "learning_rate": 7.721462053344424e-06, "loss": 0.0759, "step": 30193 }, { "epoch": 3.03, "learning_rate": 7.719859382326474e-06, "loss": 0.2216, "step": 30194 }, { "epoch": 3.03, "learning_rate": 7.71825685309088e-06, "loss": 0.1568, "step": 30195 }, { "epoch": 3.03, "learning_rate": 7.716654465647816e-06, "loss": 0.4016, "step": 30196 }, { "epoch": 3.03, "learning_rate": 7.715052220007474e-06, "loss": 0.6969, "step": 30197 }, { "epoch": 3.03, "learning_rate": 7.713450116180072e-06, "loss": 0.1322, "step": 30198 }, { "epoch": 3.03, "learning_rate": 7.711848154175793e-06, "loss": 0.177, "step": 30199 }, { "epoch": 3.03, "learning_rate": 7.710246334004834e-06, "loss": 0.1377, "step": 30200 }, { "epoch": 3.03, "learning_rate": 7.708644655677391e-06, "loss": 0.249, "step": 30201 }, { "epoch": 3.03, "learning_rate": 7.707043119203652e-06, "loss": 0.1552, "step": 30202 }, { "epoch": 3.03, "learning_rate": 7.705441724593811e-06, "loss": 0.1255, "step": 30203 }, { "epoch": 3.03, "learning_rate": 7.703840471858055e-06, "loss": 0.1495, "step": 30204 }, { "epoch": 3.03, "learning_rate": 7.702239361006576e-06, "loss": 0.1421, "step": 30205 }, { "epoch": 3.03, "learning_rate": 7.700638392049565e-06, "loss": 0.0996, "step": 30206 }, { "epoch": 3.03, "learning_rate": 7.699037564997208e-06, "loss": 0.0828, "step": 30207 }, { "epoch": 3.03, "learning_rate": 7.697436879859684e-06, "loss": 0.2352, "step": 30208 }, { "epoch": 3.03, "learning_rate": 7.695836336647206e-06, "loss": 0.0772, "step": 30209 }, { "epoch": 3.03, "learning_rate": 7.69423593536993e-06, "loss": 0.0227, "step": 30210 }, { "epoch": 3.03, "learning_rate": 7.69263567603804e-06, "loss": 0.3771, "step": 30211 }, { "epoch": 3.03, "learning_rate": 7.691035558661741e-06, "loss": 0.3857, "step": 30212 }, { "epoch": 3.03, "learning_rate": 7.689435583251214e-06, "loss": 0.2121, "step": 30213 }, { "epoch": 3.03, "learning_rate": 7.687835749816615e-06, "loss": 0.4986, "step": 30214 }, { "epoch": 3.03, "learning_rate": 7.686236058368149e-06, "loss": 0.196, "step": 30215 }, { "epoch": 3.03, "learning_rate": 7.684636508915988e-06, "loss": 0.122, "step": 30216 }, { "epoch": 3.03, "learning_rate": 7.683037101470312e-06, "loss": 0.1003, "step": 30217 }, { "epoch": 3.03, "learning_rate": 7.681437836041299e-06, "loss": 0.2663, "step": 30218 }, { "epoch": 3.03, "learning_rate": 7.679838712639123e-06, "loss": 0.0616, "step": 30219 }, { "epoch": 3.03, "learning_rate": 7.678239731273962e-06, "loss": 0.1782, "step": 30220 }, { "epoch": 3.03, "learning_rate": 7.676640891955993e-06, "loss": 0.076, "step": 30221 }, { "epoch": 3.03, "learning_rate": 7.675042194695388e-06, "loss": 0.2091, "step": 30222 }, { "epoch": 3.03, "learning_rate": 7.673443639502326e-06, "loss": 0.0525, "step": 30223 }, { "epoch": 3.03, "learning_rate": 7.671845226386968e-06, "loss": 0.2785, "step": 30224 }, { "epoch": 3.03, "learning_rate": 7.67024695535949e-06, "loss": 0.2579, "step": 30225 }, { "epoch": 3.03, "learning_rate": 7.668648826430071e-06, "loss": 0.0688, "step": 30226 }, { "epoch": 3.03, "learning_rate": 7.667050839608888e-06, "loss": 0.1512, "step": 30227 }, { "epoch": 3.03, "learning_rate": 7.665452994906079e-06, "loss": 0.0548, "step": 30228 }, { "epoch": 3.03, "learning_rate": 7.663855292331837e-06, "loss": 0.2146, "step": 30229 }, { "epoch": 3.03, "learning_rate": 7.662257731896323e-06, "loss": 0.1228, "step": 30230 }, { "epoch": 3.03, "learning_rate": 7.660660313609703e-06, "loss": 0.1597, "step": 30231 }, { "epoch": 3.03, "learning_rate": 7.659063037482143e-06, "loss": 0.0519, "step": 30232 }, { "epoch": 3.03, "learning_rate": 7.65746590352381e-06, "loss": 0.249, "step": 30233 }, { "epoch": 3.03, "learning_rate": 7.655868911744862e-06, "loss": 0.212, "step": 30234 }, { "epoch": 3.03, "learning_rate": 7.654272062155465e-06, "loss": 0.0543, "step": 30235 }, { "epoch": 3.03, "learning_rate": 7.652675354765778e-06, "loss": 0.2079, "step": 30236 }, { "epoch": 3.03, "learning_rate": 7.651078789585962e-06, "loss": 0.1795, "step": 30237 }, { "epoch": 3.03, "learning_rate": 7.649482366626183e-06, "loss": 0.1307, "step": 30238 }, { "epoch": 3.03, "learning_rate": 7.647886085896593e-06, "loss": 0.023, "step": 30239 }, { "epoch": 3.03, "learning_rate": 7.646289947407344e-06, "loss": 0.2399, "step": 30240 }, { "epoch": 3.03, "learning_rate": 7.644693951168624e-06, "loss": 0.4962, "step": 30241 }, { "epoch": 3.03, "learning_rate": 7.643098097190552e-06, "loss": 0.7388, "step": 30242 }, { "epoch": 3.03, "learning_rate": 7.641502385483295e-06, "loss": 0.3312, "step": 30243 }, { "epoch": 3.03, "learning_rate": 7.639906816057016e-06, "loss": 0.205, "step": 30244 }, { "epoch": 3.03, "learning_rate": 7.638311388921876e-06, "loss": 0.1502, "step": 30245 }, { "epoch": 3.03, "learning_rate": 7.636716104087996e-06, "loss": 0.3175, "step": 30246 }, { "epoch": 3.03, "learning_rate": 7.635120961565556e-06, "loss": 0.1262, "step": 30247 }, { "epoch": 3.03, "learning_rate": 7.633525961364696e-06, "loss": 0.3724, "step": 30248 }, { "epoch": 3.03, "learning_rate": 7.631931103495575e-06, "loss": 0.2123, "step": 30249 }, { "epoch": 3.03, "learning_rate": 7.63033638796833e-06, "loss": 0.3516, "step": 30250 }, { "epoch": 3.03, "learning_rate": 7.628741814793116e-06, "loss": 0.2547, "step": 30251 }, { "epoch": 3.03, "learning_rate": 7.627147383980076e-06, "loss": 0.2269, "step": 30252 }, { "epoch": 3.04, "learning_rate": 7.625553095539362e-06, "loss": 0.0172, "step": 30253 }, { "epoch": 3.04, "learning_rate": 7.623958949481106e-06, "loss": 0.2236, "step": 30254 }, { "epoch": 3.04, "learning_rate": 7.622364945815481e-06, "loss": 0.2565, "step": 30255 }, { "epoch": 3.04, "learning_rate": 7.620771084552602e-06, "loss": 0.2092, "step": 30256 }, { "epoch": 3.04, "learning_rate": 7.619177365702619e-06, "loss": 0.3637, "step": 30257 }, { "epoch": 3.04, "learning_rate": 7.617583789275681e-06, "loss": 0.0941, "step": 30258 }, { "epoch": 3.04, "learning_rate": 7.6159903552819335e-06, "loss": 0.0444, "step": 30259 }, { "epoch": 3.04, "learning_rate": 7.614397063731492e-06, "loss": 0.3358, "step": 30260 }, { "epoch": 3.04, "learning_rate": 7.612803914634523e-06, "loss": 0.6542, "step": 30261 }, { "epoch": 3.04, "learning_rate": 7.611210908001151e-06, "loss": 0.108, "step": 30262 }, { "epoch": 3.04, "learning_rate": 7.6096180438415165e-06, "loss": 0.3922, "step": 30263 }, { "epoch": 3.04, "learning_rate": 7.6080253221657525e-06, "loss": 0.0793, "step": 30264 }, { "epoch": 3.04, "learning_rate": 7.606432742983998e-06, "loss": 0.0695, "step": 30265 }, { "epoch": 3.04, "learning_rate": 7.604840306306386e-06, "loss": 0.2305, "step": 30266 }, { "epoch": 3.04, "learning_rate": 7.603248012143055e-06, "loss": 0.3017, "step": 30267 }, { "epoch": 3.04, "learning_rate": 7.6016558605041286e-06, "loss": 0.108, "step": 30268 }, { "epoch": 3.04, "learning_rate": 7.600063851399747e-06, "loss": 0.2466, "step": 30269 }, { "epoch": 3.04, "learning_rate": 7.598471984840034e-06, "loss": 0.132, "step": 30270 }, { "epoch": 3.04, "learning_rate": 7.596880260835126e-06, "loss": 0.153, "step": 30271 }, { "epoch": 3.04, "learning_rate": 7.595288679395142e-06, "loss": 0.4297, "step": 30272 }, { "epoch": 3.04, "learning_rate": 7.593697240530232e-06, "loss": 0.0508, "step": 30273 }, { "epoch": 3.04, "learning_rate": 7.5921059442505025e-06, "loss": 0.2245, "step": 30274 }, { "epoch": 3.04, "learning_rate": 7.590514790566076e-06, "loss": 0.1065, "step": 30275 }, { "epoch": 3.04, "learning_rate": 7.588923779487099e-06, "loss": 0.221, "step": 30276 }, { "epoch": 3.04, "learning_rate": 7.587332911023692e-06, "loss": 0.1547, "step": 30277 }, { "epoch": 3.04, "learning_rate": 7.585742185185957e-06, "loss": 0.0918, "step": 30278 }, { "epoch": 3.04, "learning_rate": 7.584151601984037e-06, "loss": 0.1492, "step": 30279 }, { "epoch": 3.04, "learning_rate": 7.58256116142805e-06, "loss": 0.158, "step": 30280 }, { "epoch": 3.04, "learning_rate": 7.58097086352812e-06, "loss": 0.1482, "step": 30281 }, { "epoch": 3.04, "learning_rate": 7.579380708294358e-06, "loss": 0.0799, "step": 30282 }, { "epoch": 3.04, "learning_rate": 7.57779069573689e-06, "loss": 0.3947, "step": 30283 }, { "epoch": 3.04, "learning_rate": 7.576200825865832e-06, "loss": 0.1486, "step": 30284 }, { "epoch": 3.04, "learning_rate": 7.574611098691299e-06, "loss": 0.1622, "step": 30285 }, { "epoch": 3.04, "learning_rate": 7.573021514223405e-06, "loss": 0.459, "step": 30286 }, { "epoch": 3.04, "learning_rate": 7.5714320724722864e-06, "loss": 0.1092, "step": 30287 }, { "epoch": 3.04, "learning_rate": 7.569842773448031e-06, "loss": 0.1017, "step": 30288 }, { "epoch": 3.04, "learning_rate": 7.568253617160758e-06, "loss": 0.103, "step": 30289 }, { "epoch": 3.04, "learning_rate": 7.5666646036205934e-06, "loss": 0.0579, "step": 30290 }, { "epoch": 3.04, "learning_rate": 7.565075732837647e-06, "loss": 0.4829, "step": 30291 }, { "epoch": 3.04, "learning_rate": 7.5634870048220094e-06, "loss": 0.1468, "step": 30292 }, { "epoch": 3.04, "learning_rate": 7.561898419583815e-06, "loss": 0.3432, "step": 30293 }, { "epoch": 3.04, "learning_rate": 7.560309977133158e-06, "loss": 0.3023, "step": 30294 }, { "epoch": 3.04, "learning_rate": 7.558721677480156e-06, "loss": 0.9734, "step": 30295 }, { "epoch": 3.04, "learning_rate": 7.557133520634913e-06, "loss": 0.0438, "step": 30296 }, { "epoch": 3.04, "learning_rate": 7.555545506607533e-06, "loss": 0.1146, "step": 30297 }, { "epoch": 3.04, "learning_rate": 7.553957635408123e-06, "loss": 0.279, "step": 30298 }, { "epoch": 3.04, "learning_rate": 7.552369907046788e-06, "loss": 0.0983, "step": 30299 }, { "epoch": 3.04, "learning_rate": 7.550782321533633e-06, "loss": 0.2609, "step": 30300 }, { "epoch": 3.04, "learning_rate": 7.549194878878756e-06, "loss": 0.088, "step": 30301 }, { "epoch": 3.04, "learning_rate": 7.547607579092263e-06, "loss": 0.0979, "step": 30302 }, { "epoch": 3.04, "learning_rate": 7.546020422184255e-06, "loss": 0.3028, "step": 30303 }, { "epoch": 3.04, "learning_rate": 7.544433408164823e-06, "loss": 0.1417, "step": 30304 }, { "epoch": 3.04, "learning_rate": 7.5428465370440895e-06, "loss": 0.1755, "step": 30305 }, { "epoch": 3.04, "learning_rate": 7.54125980883212e-06, "loss": 0.0995, "step": 30306 }, { "epoch": 3.04, "learning_rate": 7.539673223539039e-06, "loss": 0.1925, "step": 30307 }, { "epoch": 3.04, "learning_rate": 7.538086781174932e-06, "loss": 0.3214, "step": 30308 }, { "epoch": 3.04, "learning_rate": 7.5365004817498996e-06, "loss": 0.2195, "step": 30309 }, { "epoch": 3.04, "learning_rate": 7.534914325274029e-06, "loss": 0.1254, "step": 30310 }, { "epoch": 3.04, "learning_rate": 7.5333283117574205e-06, "loss": 0.2272, "step": 30311 }, { "epoch": 3.04, "learning_rate": 7.531742441210161e-06, "loss": 0.1172, "step": 30312 }, { "epoch": 3.04, "learning_rate": 7.530156713642347e-06, "loss": 0.2065, "step": 30313 }, { "epoch": 3.04, "learning_rate": 7.528571129064065e-06, "loss": 0.1264, "step": 30314 }, { "epoch": 3.04, "learning_rate": 7.52698568748541e-06, "loss": 0.132, "step": 30315 }, { "epoch": 3.04, "learning_rate": 7.5254003889164694e-06, "loss": 0.4, "step": 30316 }, { "epoch": 3.04, "learning_rate": 7.523815233367331e-06, "loss": 0.0641, "step": 30317 }, { "epoch": 3.04, "learning_rate": 7.522230220848073e-06, "loss": 0.4042, "step": 30318 }, { "epoch": 3.04, "learning_rate": 7.520645351368805e-06, "loss": 0.1839, "step": 30319 }, { "epoch": 3.04, "learning_rate": 7.519060624939593e-06, "loss": 0.1936, "step": 30320 }, { "epoch": 3.04, "learning_rate": 7.517476041570518e-06, "loss": 0.0816, "step": 30321 }, { "epoch": 3.04, "learning_rate": 7.515891601271681e-06, "loss": 0.2716, "step": 30322 }, { "epoch": 3.04, "learning_rate": 7.514307304053166e-06, "loss": 0.1051, "step": 30323 }, { "epoch": 3.04, "learning_rate": 7.512723149925027e-06, "loss": 0.197, "step": 30324 }, { "epoch": 3.04, "learning_rate": 7.511139138897372e-06, "loss": 0.3846, "step": 30325 }, { "epoch": 3.04, "learning_rate": 7.509555270980275e-06, "loss": 0.1209, "step": 30326 }, { "epoch": 3.04, "learning_rate": 7.507971546183809e-06, "loss": 0.2633, "step": 30327 }, { "epoch": 3.04, "learning_rate": 7.506387964518058e-06, "loss": 0.071, "step": 30328 }, { "epoch": 3.04, "learning_rate": 7.504804525993095e-06, "loss": 0.1993, "step": 30329 }, { "epoch": 3.04, "learning_rate": 7.503221230618997e-06, "loss": 0.4845, "step": 30330 }, { "epoch": 3.04, "learning_rate": 7.501638078405845e-06, "loss": 0.148, "step": 30331 }, { "epoch": 3.04, "learning_rate": 7.500055069363699e-06, "loss": 0.1682, "step": 30332 }, { "epoch": 3.04, "learning_rate": 7.49847220350266e-06, "loss": 0.0616, "step": 30333 }, { "epoch": 3.04, "learning_rate": 7.496889480832776e-06, "loss": 0.0609, "step": 30334 }, { "epoch": 3.04, "learning_rate": 7.495306901364116e-06, "loss": 0.1294, "step": 30335 }, { "epoch": 3.04, "learning_rate": 7.493724465106773e-06, "loss": 0.192, "step": 30336 }, { "epoch": 3.04, "learning_rate": 7.4921421720708135e-06, "loss": 0.046, "step": 30337 }, { "epoch": 3.04, "learning_rate": 7.490560022266278e-06, "loss": 0.1676, "step": 30338 }, { "epoch": 3.04, "learning_rate": 7.488978015703268e-06, "loss": 0.1785, "step": 30339 }, { "epoch": 3.04, "learning_rate": 7.487396152391839e-06, "loss": 0.1028, "step": 30340 }, { "epoch": 3.04, "learning_rate": 7.485814432342055e-06, "loss": 0.0814, "step": 30341 }, { "epoch": 3.04, "learning_rate": 7.484232855563983e-06, "loss": 0.0967, "step": 30342 }, { "epoch": 3.04, "learning_rate": 7.48265142206769e-06, "loss": 0.4112, "step": 30343 }, { "epoch": 3.04, "learning_rate": 7.481070131863232e-06, "loss": 0.2646, "step": 30344 }, { "epoch": 3.04, "learning_rate": 7.479488984960682e-06, "loss": 0.1806, "step": 30345 }, { "epoch": 3.04, "learning_rate": 7.477907981370097e-06, "loss": 0.0931, "step": 30346 }, { "epoch": 3.04, "learning_rate": 7.476327121101536e-06, "loss": 0.141, "step": 30347 }, { "epoch": 3.04, "learning_rate": 7.474746404165058e-06, "loss": 0.7593, "step": 30348 }, { "epoch": 3.04, "learning_rate": 7.473165830570728e-06, "loss": 0.1414, "step": 30349 }, { "epoch": 3.04, "learning_rate": 7.471585400328592e-06, "loss": 0.2534, "step": 30350 }, { "epoch": 3.04, "learning_rate": 7.470005113448734e-06, "loss": 0.1317, "step": 30351 }, { "epoch": 3.05, "learning_rate": 7.468424969941185e-06, "loss": 0.2573, "step": 30352 }, { "epoch": 3.05, "learning_rate": 7.466844969815997e-06, "loss": 0.1133, "step": 30353 }, { "epoch": 3.05, "learning_rate": 7.465265113083244e-06, "loss": 0.4319, "step": 30354 }, { "epoch": 3.05, "learning_rate": 7.46368539975298e-06, "loss": 0.1895, "step": 30355 }, { "epoch": 3.05, "learning_rate": 7.462105829835233e-06, "loss": 0.3965, "step": 30356 }, { "epoch": 3.05, "learning_rate": 7.460526403340079e-06, "loss": 0.2185, "step": 30357 }, { "epoch": 3.05, "learning_rate": 7.4589471202775605e-06, "loss": 0.3692, "step": 30358 }, { "epoch": 3.05, "learning_rate": 7.4573679806577264e-06, "loss": 0.5675, "step": 30359 }, { "epoch": 3.05, "learning_rate": 7.455788984490629e-06, "loss": 0.1417, "step": 30360 }, { "epoch": 3.05, "learning_rate": 7.4542101317863135e-06, "loss": 0.1812, "step": 30361 }, { "epoch": 3.05, "learning_rate": 7.452631422554827e-06, "loss": 0.3813, "step": 30362 }, { "epoch": 3.05, "learning_rate": 7.451052856806218e-06, "loss": 0.0285, "step": 30363 }, { "epoch": 3.05, "learning_rate": 7.449474434550519e-06, "loss": 0.128, "step": 30364 }, { "epoch": 3.05, "learning_rate": 7.447896155797805e-06, "loss": 0.334, "step": 30365 }, { "epoch": 3.05, "learning_rate": 7.446318020558094e-06, "loss": 0.1363, "step": 30366 }, { "epoch": 3.05, "learning_rate": 7.444740028841422e-06, "loss": 0.0925, "step": 30367 }, { "epoch": 3.05, "learning_rate": 7.443162180657856e-06, "loss": 0.5698, "step": 30368 }, { "epoch": 3.05, "learning_rate": 7.4415844760174345e-06, "loss": 0.1666, "step": 30369 }, { "epoch": 3.05, "learning_rate": 7.440006914930169e-06, "loss": 0.2217, "step": 30370 }, { "epoch": 3.05, "learning_rate": 7.438429497406124e-06, "loss": 0.3535, "step": 30371 }, { "epoch": 3.05, "learning_rate": 7.436852223455333e-06, "loss": 0.4316, "step": 30372 }, { "epoch": 3.05, "learning_rate": 7.43527509308783e-06, "loss": 0.0385, "step": 30373 }, { "epoch": 3.05, "learning_rate": 7.433698106313655e-06, "loss": 0.2706, "step": 30374 }, { "epoch": 3.05, "learning_rate": 7.432121263142838e-06, "loss": 0.0546, "step": 30375 }, { "epoch": 3.05, "learning_rate": 7.430544563585416e-06, "loss": 0.1655, "step": 30376 }, { "epoch": 3.05, "learning_rate": 7.428968007651421e-06, "loss": 0.204, "step": 30377 }, { "epoch": 3.05, "learning_rate": 7.4273915953508905e-06, "loss": 0.2169, "step": 30378 }, { "epoch": 3.05, "learning_rate": 7.425815326693851e-06, "loss": 0.0901, "step": 30379 }, { "epoch": 3.05, "learning_rate": 7.424239201690334e-06, "loss": 0.3483, "step": 30380 }, { "epoch": 3.05, "learning_rate": 7.422663220350374e-06, "loss": 0.0591, "step": 30381 }, { "epoch": 3.05, "learning_rate": 7.421087382683984e-06, "loss": 0.0964, "step": 30382 }, { "epoch": 3.05, "learning_rate": 7.419511688701226e-06, "loss": 0.1276, "step": 30383 }, { "epoch": 3.05, "learning_rate": 7.417936138412091e-06, "loss": 0.1916, "step": 30384 }, { "epoch": 3.05, "learning_rate": 7.416360731826616e-06, "loss": 0.0809, "step": 30385 }, { "epoch": 3.05, "learning_rate": 7.414785468954838e-06, "loss": 0.1527, "step": 30386 }, { "epoch": 3.05, "learning_rate": 7.413210349806772e-06, "loss": 0.3988, "step": 30387 }, { "epoch": 3.05, "learning_rate": 7.411635374392442e-06, "loss": 0.223, "step": 30388 }, { "epoch": 3.05, "learning_rate": 7.410060542721873e-06, "loss": 0.4388, "step": 30389 }, { "epoch": 3.05, "learning_rate": 7.4084858548050845e-06, "loss": 0.0958, "step": 30390 }, { "epoch": 3.05, "learning_rate": 7.406911310652098e-06, "loss": 0.1851, "step": 30391 }, { "epoch": 3.05, "learning_rate": 7.405336910272934e-06, "loss": 0.256, "step": 30392 }, { "epoch": 3.05, "learning_rate": 7.403762653677611e-06, "loss": 0.1502, "step": 30393 }, { "epoch": 3.05, "learning_rate": 7.402188540876146e-06, "loss": 0.1963, "step": 30394 }, { "epoch": 3.05, "learning_rate": 7.400614571878556e-06, "loss": 0.0452, "step": 30395 }, { "epoch": 3.05, "learning_rate": 7.3990407466948465e-06, "loss": 0.2353, "step": 30396 }, { "epoch": 3.05, "learning_rate": 7.397467065335063e-06, "loss": 0.0302, "step": 30397 }, { "epoch": 3.05, "learning_rate": 7.395893527809194e-06, "loss": 0.1419, "step": 30398 }, { "epoch": 3.05, "learning_rate": 7.394320134127249e-06, "loss": 0.1769, "step": 30399 }, { "epoch": 3.05, "learning_rate": 7.39274688429926e-06, "loss": 0.519, "step": 30400 }, { "epoch": 3.05, "learning_rate": 7.391173778335237e-06, "loss": 0.1345, "step": 30401 }, { "epoch": 3.05, "learning_rate": 7.389600816245168e-06, "loss": 0.3923, "step": 30402 }, { "epoch": 3.05, "learning_rate": 7.388027998039083e-06, "loss": 0.0802, "step": 30403 }, { "epoch": 3.05, "learning_rate": 7.386455323726988e-06, "loss": 0.2498, "step": 30404 }, { "epoch": 3.05, "learning_rate": 7.384882793318893e-06, "loss": 0.1357, "step": 30405 }, { "epoch": 3.05, "learning_rate": 7.383310406824794e-06, "loss": 0.0843, "step": 30406 }, { "epoch": 3.05, "learning_rate": 7.381738164254708e-06, "loss": 0.3367, "step": 30407 }, { "epoch": 3.05, "learning_rate": 7.380166065618633e-06, "loss": 0.1516, "step": 30408 }, { "epoch": 3.05, "learning_rate": 7.378594110926579e-06, "loss": 0.1542, "step": 30409 }, { "epoch": 3.05, "learning_rate": 7.377022300188543e-06, "loss": 0.1079, "step": 30410 }, { "epoch": 3.05, "learning_rate": 7.3754506334145344e-06, "loss": 0.0607, "step": 30411 }, { "epoch": 3.05, "learning_rate": 7.373879110614548e-06, "loss": 0.264, "step": 30412 }, { "epoch": 3.05, "learning_rate": 7.372307731798591e-06, "loss": 0.6635, "step": 30413 }, { "epoch": 3.05, "learning_rate": 7.370736496976647e-06, "loss": 0.1528, "step": 30414 }, { "epoch": 3.05, "learning_rate": 7.369165406158747e-06, "loss": 0.1705, "step": 30415 }, { "epoch": 3.05, "learning_rate": 7.3675944593548525e-06, "loss": 0.1107, "step": 30416 }, { "epoch": 3.05, "learning_rate": 7.366023656574984e-06, "loss": 0.1369, "step": 30417 }, { "epoch": 3.05, "learning_rate": 7.36445299782913e-06, "loss": 0.0606, "step": 30418 }, { "epoch": 3.05, "learning_rate": 7.362882483127289e-06, "loss": 0.1138, "step": 30419 }, { "epoch": 3.05, "learning_rate": 7.36131211247945e-06, "loss": 0.0611, "step": 30420 }, { "epoch": 3.05, "learning_rate": 7.359741885895609e-06, "loss": 0.1873, "step": 30421 }, { "epoch": 3.05, "learning_rate": 7.358171803385757e-06, "loss": 0.1728, "step": 30422 }, { "epoch": 3.05, "learning_rate": 7.35660186495989e-06, "loss": 0.2721, "step": 30423 }, { "epoch": 3.05, "learning_rate": 7.35503207062799e-06, "loss": 0.5321, "step": 30424 }, { "epoch": 3.05, "learning_rate": 7.353462420400054e-06, "loss": 0.1799, "step": 30425 }, { "epoch": 3.05, "learning_rate": 7.351892914286067e-06, "loss": 0.4006, "step": 30426 }, { "epoch": 3.05, "learning_rate": 7.350323552296019e-06, "loss": 0.492, "step": 30427 }, { "epoch": 3.05, "learning_rate": 7.348754334439886e-06, "loss": 0.3238, "step": 30428 }, { "epoch": 3.05, "learning_rate": 7.347185260727679e-06, "loss": 0.0887, "step": 30429 }, { "epoch": 3.05, "learning_rate": 7.345616331169361e-06, "loss": 0.3366, "step": 30430 }, { "epoch": 3.05, "learning_rate": 7.344047545774914e-06, "loss": 0.4079, "step": 30431 }, { "epoch": 3.05, "learning_rate": 7.3424789045543375e-06, "loss": 0.202, "step": 30432 }, { "epoch": 3.05, "learning_rate": 7.340910407517615e-06, "loss": 0.4654, "step": 30433 }, { "epoch": 3.05, "learning_rate": 7.3393420546746995e-06, "loss": 0.2065, "step": 30434 }, { "epoch": 3.05, "learning_rate": 7.337773846035602e-06, "loss": 0.3103, "step": 30435 }, { "epoch": 3.05, "learning_rate": 7.3362057816102905e-06, "loss": 0.4512, "step": 30436 }, { "epoch": 3.05, "learning_rate": 7.334637861408741e-06, "loss": 0.3099, "step": 30437 }, { "epoch": 3.05, "learning_rate": 7.333070085440936e-06, "loss": 0.532, "step": 30438 }, { "epoch": 3.05, "learning_rate": 7.3315024537168486e-06, "loss": 0.0702, "step": 30439 }, { "epoch": 3.05, "learning_rate": 7.329934966246457e-06, "loss": 0.0952, "step": 30440 }, { "epoch": 3.05, "learning_rate": 7.328367623039736e-06, "loss": 0.2911, "step": 30441 }, { "epoch": 3.05, "learning_rate": 7.326800424106648e-06, "loss": 0.147, "step": 30442 }, { "epoch": 3.05, "learning_rate": 7.3252333694571945e-06, "loss": 0.0958, "step": 30443 }, { "epoch": 3.05, "learning_rate": 7.323666459101322e-06, "loss": 0.1118, "step": 30444 }, { "epoch": 3.05, "learning_rate": 7.322099693048999e-06, "loss": 0.1767, "step": 30445 }, { "epoch": 3.05, "learning_rate": 7.320533071310217e-06, "loss": 0.3027, "step": 30446 }, { "epoch": 3.05, "learning_rate": 7.3189665938949445e-06, "loss": 0.2371, "step": 30447 }, { "epoch": 3.05, "learning_rate": 7.31740026081312e-06, "loss": 0.1442, "step": 30448 }, { "epoch": 3.05, "learning_rate": 7.315834072074744e-06, "loss": 0.2562, "step": 30449 }, { "epoch": 3.06, "learning_rate": 7.31426802768977e-06, "loss": 0.1432, "step": 30450 }, { "epoch": 3.06, "learning_rate": 7.312702127668161e-06, "loss": 0.1919, "step": 30451 }, { "epoch": 3.06, "learning_rate": 7.311136372019888e-06, "loss": 0.6515, "step": 30452 }, { "epoch": 3.06, "learning_rate": 7.30957076075491e-06, "loss": 0.2149, "step": 30453 }, { "epoch": 3.06, "learning_rate": 7.30800529388319e-06, "loss": 0.3346, "step": 30454 }, { "epoch": 3.06, "learning_rate": 7.306439971414696e-06, "loss": 0.2226, "step": 30455 }, { "epoch": 3.06, "learning_rate": 7.304874793359383e-06, "loss": 0.1375, "step": 30456 }, { "epoch": 3.06, "learning_rate": 7.30330975972721e-06, "loss": 0.1026, "step": 30457 }, { "epoch": 3.06, "learning_rate": 7.301744870528143e-06, "loss": 0.0372, "step": 30458 }, { "epoch": 3.06, "learning_rate": 7.300180125772136e-06, "loss": 0.1113, "step": 30459 }, { "epoch": 3.06, "learning_rate": 7.29861552546914e-06, "loss": 0.0695, "step": 30460 }, { "epoch": 3.06, "learning_rate": 7.297051069629135e-06, "loss": 0.6552, "step": 30461 }, { "epoch": 3.06, "learning_rate": 7.295486758262053e-06, "loss": 0.1122, "step": 30462 }, { "epoch": 3.06, "learning_rate": 7.293922591377844e-06, "loss": 0.026, "step": 30463 }, { "epoch": 3.06, "learning_rate": 7.292358568986485e-06, "loss": 0.0946, "step": 30464 }, { "epoch": 3.06, "learning_rate": 7.290794691097927e-06, "loss": 0.2148, "step": 30465 }, { "epoch": 3.06, "learning_rate": 7.289230957722094e-06, "loss": 0.3338, "step": 30466 }, { "epoch": 3.06, "learning_rate": 7.287667368868964e-06, "loss": 0.0876, "step": 30467 }, { "epoch": 3.06, "learning_rate": 7.286103924548477e-06, "loss": 0.1683, "step": 30468 }, { "epoch": 3.06, "learning_rate": 7.2845406247705885e-06, "loss": 0.1977, "step": 30469 }, { "epoch": 3.06, "learning_rate": 7.282977469545239e-06, "loss": 0.1545, "step": 30470 }, { "epoch": 3.06, "learning_rate": 7.281414458882369e-06, "loss": 0.0922, "step": 30471 }, { "epoch": 3.06, "learning_rate": 7.279851592791951e-06, "loss": 0.059, "step": 30472 }, { "epoch": 3.06, "learning_rate": 7.278288871283911e-06, "loss": 0.9017, "step": 30473 }, { "epoch": 3.06, "learning_rate": 7.276726294368186e-06, "loss": 0.1679, "step": 30474 }, { "epoch": 3.06, "learning_rate": 7.275163862054747e-06, "loss": 0.0887, "step": 30475 }, { "epoch": 3.06, "learning_rate": 7.27360157435351e-06, "loss": 0.3631, "step": 30476 }, { "epoch": 3.06, "learning_rate": 7.272039431274421e-06, "loss": 0.4643, "step": 30477 }, { "epoch": 3.06, "learning_rate": 7.270477432827435e-06, "loss": 0.3779, "step": 30478 }, { "epoch": 3.06, "learning_rate": 7.26891557902249e-06, "loss": 0.0969, "step": 30479 }, { "epoch": 3.06, "learning_rate": 7.267353869869506e-06, "loss": 0.1502, "step": 30480 }, { "epoch": 3.06, "learning_rate": 7.265792305378441e-06, "loss": 0.1151, "step": 30481 }, { "epoch": 3.06, "learning_rate": 7.264230885559225e-06, "loss": 0.1041, "step": 30482 }, { "epoch": 3.06, "learning_rate": 7.262669610421794e-06, "loss": 0.1466, "step": 30483 }, { "epoch": 3.06, "learning_rate": 7.2611084799760816e-06, "loss": 0.3313, "step": 30484 }, { "epoch": 3.06, "learning_rate": 7.259547494232027e-06, "loss": 0.1722, "step": 30485 }, { "epoch": 3.06, "learning_rate": 7.257986653199562e-06, "loss": 0.6601, "step": 30486 }, { "epoch": 3.06, "learning_rate": 7.256425956888618e-06, "loss": 0.3006, "step": 30487 }, { "epoch": 3.06, "learning_rate": 7.254865405309128e-06, "loss": 0.0927, "step": 30488 }, { "epoch": 3.06, "learning_rate": 7.25330499847102e-06, "loss": 0.0504, "step": 30489 }, { "epoch": 3.06, "learning_rate": 7.251744736384226e-06, "loss": 0.3405, "step": 30490 }, { "epoch": 3.06, "learning_rate": 7.2501846190586755e-06, "loss": 0.069, "step": 30491 }, { "epoch": 3.06, "learning_rate": 7.248624646504286e-06, "loss": 0.0202, "step": 30492 }, { "epoch": 3.06, "learning_rate": 7.24706481873101e-06, "loss": 0.508, "step": 30493 }, { "epoch": 3.06, "learning_rate": 7.245505135748751e-06, "loss": 0.1109, "step": 30494 }, { "epoch": 3.06, "learning_rate": 7.243945597567433e-06, "loss": 0.1152, "step": 30495 }, { "epoch": 3.06, "learning_rate": 7.242386204196998e-06, "loss": 0.541, "step": 30496 }, { "epoch": 3.06, "learning_rate": 7.240826955647358e-06, "loss": 0.1948, "step": 30497 }, { "epoch": 3.06, "learning_rate": 7.239267851928436e-06, "loss": 0.0733, "step": 30498 }, { "epoch": 3.06, "learning_rate": 7.237708893050156e-06, "loss": 0.1416, "step": 30499 }, { "epoch": 3.06, "learning_rate": 7.236150079022439e-06, "loss": 0.1746, "step": 30500 }, { "epoch": 3.06, "learning_rate": 7.2345914098552015e-06, "loss": 0.3581, "step": 30501 }, { "epoch": 3.06, "learning_rate": 7.2330328855583664e-06, "loss": 0.91, "step": 30502 }, { "epoch": 3.06, "learning_rate": 7.2314745061418354e-06, "loss": 0.0999, "step": 30503 }, { "epoch": 3.06, "learning_rate": 7.229916271615561e-06, "loss": 0.0921, "step": 30504 }, { "epoch": 3.06, "learning_rate": 7.228358181989425e-06, "loss": 0.1383, "step": 30505 }, { "epoch": 3.06, "learning_rate": 7.226800237273347e-06, "loss": 0.5308, "step": 30506 }, { "epoch": 3.06, "learning_rate": 7.225242437477267e-06, "loss": 0.2912, "step": 30507 }, { "epoch": 3.06, "learning_rate": 7.223684782611073e-06, "loss": 0.2506, "step": 30508 }, { "epoch": 3.06, "learning_rate": 7.222127272684675e-06, "loss": 0.0973, "step": 30509 }, { "epoch": 3.06, "learning_rate": 7.220569907708e-06, "loss": 0.0512, "step": 30510 }, { "epoch": 3.06, "learning_rate": 7.219012687690963e-06, "loss": 0.1423, "step": 30511 }, { "epoch": 3.06, "learning_rate": 7.217455612643448e-06, "loss": 0.1271, "step": 30512 }, { "epoch": 3.06, "learning_rate": 7.215898682575382e-06, "loss": 0.1577, "step": 30513 }, { "epoch": 3.06, "learning_rate": 7.2143418974966725e-06, "loss": 0.1938, "step": 30514 }, { "epoch": 3.06, "learning_rate": 7.212785257417223e-06, "loss": 0.2745, "step": 30515 }, { "epoch": 3.06, "learning_rate": 7.2112287623469375e-06, "loss": 0.3832, "step": 30516 }, { "epoch": 3.06, "learning_rate": 7.2096724122957255e-06, "loss": 0.016, "step": 30517 }, { "epoch": 3.06, "learning_rate": 7.208116207273486e-06, "loss": 0.0991, "step": 30518 }, { "epoch": 3.06, "learning_rate": 7.2065601472901275e-06, "loss": 0.1734, "step": 30519 }, { "epoch": 3.06, "learning_rate": 7.205004232355545e-06, "loss": 0.3968, "step": 30520 }, { "epoch": 3.06, "learning_rate": 7.203448462479648e-06, "loss": 0.3205, "step": 30521 }, { "epoch": 3.06, "learning_rate": 7.201892837672331e-06, "loss": 0.2317, "step": 30522 }, { "epoch": 3.06, "learning_rate": 7.200337357943487e-06, "loss": 0.3349, "step": 30523 }, { "epoch": 3.06, "learning_rate": 7.198782023303031e-06, "loss": 0.0983, "step": 30524 }, { "epoch": 3.06, "learning_rate": 7.197226833760858e-06, "loss": 0.5163, "step": 30525 }, { "epoch": 3.06, "learning_rate": 7.195671789326848e-06, "loss": 0.0592, "step": 30526 }, { "epoch": 3.06, "learning_rate": 7.194116890010911e-06, "loss": 1.0047, "step": 30527 }, { "epoch": 3.06, "learning_rate": 7.192562135822937e-06, "loss": 0.0475, "step": 30528 }, { "epoch": 3.06, "learning_rate": 7.191007526772826e-06, "loss": 0.0832, "step": 30529 }, { "epoch": 3.06, "learning_rate": 7.189453062870465e-06, "loss": 0.2895, "step": 30530 }, { "epoch": 3.06, "learning_rate": 7.187898744125743e-06, "loss": 0.1181, "step": 30531 }, { "epoch": 3.06, "learning_rate": 7.186344570548559e-06, "loss": 0.3754, "step": 30532 }, { "epoch": 3.06, "learning_rate": 7.184790542148798e-06, "loss": 0.0565, "step": 30533 }, { "epoch": 3.06, "learning_rate": 7.183236658936353e-06, "loss": 0.2601, "step": 30534 }, { "epoch": 3.06, "learning_rate": 7.1816829209210984e-06, "loss": 0.423, "step": 30535 }, { "epoch": 3.06, "learning_rate": 7.180129328112951e-06, "loss": 0.2392, "step": 30536 }, { "epoch": 3.06, "learning_rate": 7.178575880521771e-06, "loss": 0.4529, "step": 30537 }, { "epoch": 3.06, "learning_rate": 7.177022578157446e-06, "loss": 0.368, "step": 30538 }, { "epoch": 3.06, "learning_rate": 7.175469421029883e-06, "loss": 0.1529, "step": 30539 }, { "epoch": 3.06, "learning_rate": 7.173916409148939e-06, "loss": 0.1016, "step": 30540 }, { "epoch": 3.06, "learning_rate": 7.172363542524501e-06, "loss": 0.06, "step": 30541 }, { "epoch": 3.06, "learning_rate": 7.170810821166465e-06, "loss": 0.0745, "step": 30542 }, { "epoch": 3.06, "learning_rate": 7.169258245084714e-06, "loss": 0.0519, "step": 30543 }, { "epoch": 3.06, "learning_rate": 7.167705814289098e-06, "loss": 0.4692, "step": 30544 }, { "epoch": 3.06, "learning_rate": 7.166153528789529e-06, "loss": 0.3093, "step": 30545 }, { "epoch": 3.06, "learning_rate": 7.164601388595869e-06, "loss": 0.7367, "step": 30546 }, { "epoch": 3.06, "learning_rate": 7.163049393718003e-06, "loss": 0.1245, "step": 30547 }, { "epoch": 3.06, "learning_rate": 7.161497544165798e-06, "loss": 0.1283, "step": 30548 }, { "epoch": 3.07, "learning_rate": 7.15994583994913e-06, "loss": 0.0774, "step": 30549 }, { "epoch": 3.07, "learning_rate": 7.158394281077895e-06, "loss": 0.1995, "step": 30550 }, { "epoch": 3.07, "learning_rate": 7.1568428675619365e-06, "loss": 0.0656, "step": 30551 }, { "epoch": 3.07, "learning_rate": 7.155291599411136e-06, "loss": 0.3455, "step": 30552 }, { "epoch": 3.07, "learning_rate": 7.153740476635383e-06, "loss": 0.0732, "step": 30553 }, { "epoch": 3.07, "learning_rate": 7.152189499244528e-06, "loss": 0.0265, "step": 30554 }, { "epoch": 3.07, "learning_rate": 7.150638667248439e-06, "loss": 0.2705, "step": 30555 }, { "epoch": 3.07, "learning_rate": 7.1490879806569985e-06, "loss": 0.2799, "step": 30556 }, { "epoch": 3.07, "learning_rate": 7.147537439480082e-06, "loss": 1.1636, "step": 30557 }, { "epoch": 3.07, "learning_rate": 7.145987043727525e-06, "loss": 0.0855, "step": 30558 }, { "epoch": 3.07, "learning_rate": 7.144436793409218e-06, "loss": 0.0662, "step": 30559 }, { "epoch": 3.07, "learning_rate": 7.142886688535026e-06, "loss": 0.0518, "step": 30560 }, { "epoch": 3.07, "learning_rate": 7.141336729114801e-06, "loss": 0.0798, "step": 30561 }, { "epoch": 3.07, "learning_rate": 7.139786915158418e-06, "loss": 0.3392, "step": 30562 }, { "epoch": 3.07, "learning_rate": 7.138237246675735e-06, "loss": 0.0991, "step": 30563 }, { "epoch": 3.07, "learning_rate": 7.13668772367661e-06, "loss": 0.1673, "step": 30564 }, { "epoch": 3.07, "learning_rate": 7.13513834617091e-06, "loss": 0.0158, "step": 30565 }, { "epoch": 3.07, "learning_rate": 7.133589114168488e-06, "loss": 0.4598, "step": 30566 }, { "epoch": 3.07, "learning_rate": 7.132040027679202e-06, "loss": 0.2107, "step": 30567 }, { "epoch": 3.07, "learning_rate": 7.130491086712925e-06, "loss": 0.2031, "step": 30568 }, { "epoch": 3.07, "learning_rate": 7.128942291279499e-06, "loss": 0.3021, "step": 30569 }, { "epoch": 3.07, "learning_rate": 7.127393641388773e-06, "loss": 0.1233, "step": 30570 }, { "epoch": 3.07, "learning_rate": 7.125845137050631e-06, "loss": 0.0702, "step": 30571 }, { "epoch": 3.07, "learning_rate": 7.1242967782749e-06, "loss": 0.0172, "step": 30572 }, { "epoch": 3.07, "learning_rate": 7.122748565071431e-06, "loss": 0.0582, "step": 30573 }, { "epoch": 3.07, "learning_rate": 7.121200497450097e-06, "loss": 0.1978, "step": 30574 }, { "epoch": 3.07, "learning_rate": 7.1196525754207494e-06, "loss": 0.1538, "step": 30575 }, { "epoch": 3.07, "learning_rate": 7.1181047989932125e-06, "loss": 0.1002, "step": 30576 }, { "epoch": 3.07, "learning_rate": 7.116557168177358e-06, "loss": 0.0857, "step": 30577 }, { "epoch": 3.07, "learning_rate": 7.115009682983028e-06, "loss": 0.1791, "step": 30578 }, { "epoch": 3.07, "learning_rate": 7.113462343420072e-06, "loss": 0.0794, "step": 30579 }, { "epoch": 3.07, "learning_rate": 7.111915149498335e-06, "loss": 0.3615, "step": 30580 }, { "epoch": 3.07, "learning_rate": 7.110368101227655e-06, "loss": 0.3801, "step": 30581 }, { "epoch": 3.07, "learning_rate": 7.1088211986179e-06, "loss": 0.194, "step": 30582 }, { "epoch": 3.07, "learning_rate": 7.107274441678892e-06, "loss": 0.683, "step": 30583 }, { "epoch": 3.07, "learning_rate": 7.10572783042047e-06, "loss": 0.0809, "step": 30584 }, { "epoch": 3.07, "learning_rate": 7.104181364852507e-06, "loss": 0.4367, "step": 30585 }, { "epoch": 3.07, "learning_rate": 7.102635044984812e-06, "loss": 0.0672, "step": 30586 }, { "epoch": 3.07, "learning_rate": 7.101088870827231e-06, "loss": 0.0406, "step": 30587 }, { "epoch": 3.07, "learning_rate": 7.099542842389617e-06, "loss": 0.1406, "step": 30588 }, { "epoch": 3.07, "learning_rate": 7.0979969596818075e-06, "loss": 0.2057, "step": 30589 }, { "epoch": 3.07, "learning_rate": 7.096451222713614e-06, "loss": 0.4636, "step": 30590 }, { "epoch": 3.07, "learning_rate": 7.094905631494905e-06, "loss": 0.1623, "step": 30591 }, { "epoch": 3.07, "learning_rate": 7.0933601860355025e-06, "loss": 0.1565, "step": 30592 }, { "epoch": 3.07, "learning_rate": 7.09181488634524e-06, "loss": 0.6505, "step": 30593 }, { "epoch": 3.07, "learning_rate": 7.090269732433954e-06, "loss": 0.0548, "step": 30594 }, { "epoch": 3.07, "learning_rate": 7.088724724311475e-06, "loss": 0.224, "step": 30595 }, { "epoch": 3.07, "learning_rate": 7.087179861987636e-06, "loss": 0.0711, "step": 30596 }, { "epoch": 3.07, "learning_rate": 7.085635145472269e-06, "loss": 0.1797, "step": 30597 }, { "epoch": 3.07, "learning_rate": 7.084090574775202e-06, "loss": 0.0269, "step": 30598 }, { "epoch": 3.07, "learning_rate": 7.082546149906256e-06, "loss": 0.0465, "step": 30599 }, { "epoch": 3.07, "learning_rate": 7.081001870875288e-06, "loss": 0.0954, "step": 30600 }, { "epoch": 3.07, "learning_rate": 7.079457737692093e-06, "loss": 0.0353, "step": 30601 }, { "epoch": 3.07, "learning_rate": 7.0779137503665016e-06, "loss": 0.2655, "step": 30602 }, { "epoch": 3.07, "learning_rate": 7.076369908908364e-06, "loss": 0.2782, "step": 30603 }, { "epoch": 3.07, "learning_rate": 7.0748262133274745e-06, "loss": 0.4857, "step": 30604 }, { "epoch": 3.07, "learning_rate": 7.073282663633673e-06, "loss": 0.0952, "step": 30605 }, { "epoch": 3.07, "learning_rate": 7.071739259836784e-06, "loss": 0.0524, "step": 30606 }, { "epoch": 3.07, "learning_rate": 7.07019600194662e-06, "loss": 0.4419, "step": 30607 }, { "epoch": 3.07, "learning_rate": 7.0686528899730065e-06, "loss": 0.1396, "step": 30608 }, { "epoch": 3.07, "learning_rate": 7.067109923925763e-06, "loss": 0.1255, "step": 30609 }, { "epoch": 3.07, "learning_rate": 7.06556710381471e-06, "loss": 0.2744, "step": 30610 }, { "epoch": 3.07, "learning_rate": 7.064024429649658e-06, "loss": 0.2697, "step": 30611 }, { "epoch": 3.07, "learning_rate": 7.062481901440434e-06, "loss": 0.1233, "step": 30612 }, { "epoch": 3.07, "learning_rate": 7.06093951919684e-06, "loss": 0.2311, "step": 30613 }, { "epoch": 3.07, "learning_rate": 7.059397282928716e-06, "loss": 0.3073, "step": 30614 }, { "epoch": 3.07, "learning_rate": 7.0578551926458515e-06, "loss": 0.2015, "step": 30615 }, { "epoch": 3.07, "learning_rate": 7.0563132483580615e-06, "loss": 0.0445, "step": 30616 }, { "epoch": 3.07, "learning_rate": 7.054771450075184e-06, "loss": 0.0829, "step": 30617 }, { "epoch": 3.07, "learning_rate": 7.053229797807002e-06, "loss": 0.4169, "step": 30618 }, { "epoch": 3.07, "learning_rate": 7.0516882915633275e-06, "loss": 0.2825, "step": 30619 }, { "epoch": 3.07, "learning_rate": 7.05014693135399e-06, "loss": 0.064, "step": 30620 }, { "epoch": 3.07, "learning_rate": 7.048605717188793e-06, "loss": 0.1655, "step": 30621 }, { "epoch": 3.07, "learning_rate": 7.047064649077521e-06, "loss": 0.0478, "step": 30622 }, { "epoch": 3.07, "learning_rate": 7.045523727030008e-06, "loss": 0.1285, "step": 30623 }, { "epoch": 3.07, "learning_rate": 7.043982951056047e-06, "loss": 0.3284, "step": 30624 }, { "epoch": 3.07, "learning_rate": 7.042442321165445e-06, "loss": 0.6627, "step": 30625 }, { "epoch": 3.07, "learning_rate": 7.040901837368011e-06, "loss": 0.2057, "step": 30626 }, { "epoch": 3.07, "learning_rate": 7.03936149967354e-06, "loss": 0.0431, "step": 30627 }, { "epoch": 3.07, "learning_rate": 7.037821308091841e-06, "loss": 0.0615, "step": 30628 }, { "epoch": 3.07, "learning_rate": 7.036281262632709e-06, "loss": 0.1372, "step": 30629 }, { "epoch": 3.07, "learning_rate": 7.03474136330594e-06, "loss": 0.5542, "step": 30630 }, { "epoch": 3.07, "learning_rate": 7.0332016101213455e-06, "loss": 0.0996, "step": 30631 }, { "epoch": 3.07, "learning_rate": 7.031662003088732e-06, "loss": 0.2034, "step": 30632 }, { "epoch": 3.07, "learning_rate": 7.030122542217865e-06, "loss": 0.2828, "step": 30633 }, { "epoch": 3.07, "learning_rate": 7.028583227518569e-06, "loss": 0.0866, "step": 30634 }, { "epoch": 3.07, "learning_rate": 7.02704405900064e-06, "loss": 0.2925, "step": 30635 }, { "epoch": 3.07, "learning_rate": 7.0255050366738435e-06, "loss": 0.0006, "step": 30636 }, { "epoch": 3.07, "learning_rate": 7.023966160548006e-06, "loss": 0.0562, "step": 30637 }, { "epoch": 3.07, "learning_rate": 7.0224274306329035e-06, "loss": 0.0688, "step": 30638 }, { "epoch": 3.07, "learning_rate": 7.020888846938332e-06, "loss": 0.0515, "step": 30639 }, { "epoch": 3.07, "learning_rate": 7.0193504094740836e-06, "loss": 0.3876, "step": 30640 }, { "epoch": 3.07, "learning_rate": 7.017812118249945e-06, "loss": 0.1707, "step": 30641 }, { "epoch": 3.07, "learning_rate": 7.01627397327571e-06, "loss": 0.0917, "step": 30642 }, { "epoch": 3.07, "learning_rate": 7.014735974561162e-06, "loss": 0.0948, "step": 30643 }, { "epoch": 3.07, "learning_rate": 7.01319812211609e-06, "loss": 0.2929, "step": 30644 }, { "epoch": 3.07, "learning_rate": 7.011660415950274e-06, "loss": 0.0905, "step": 30645 }, { "epoch": 3.07, "learning_rate": 7.010122856073522e-06, "loss": 0.0441, "step": 30646 }, { "epoch": 3.08, "learning_rate": 7.008585442495593e-06, "loss": 0.0749, "step": 30647 }, { "epoch": 3.08, "learning_rate": 7.0070481752262696e-06, "loss": 0.4853, "step": 30648 }, { "epoch": 3.08, "learning_rate": 7.005511054275363e-06, "loss": 0.0839, "step": 30649 }, { "epoch": 3.08, "learning_rate": 7.0039740796526294e-06, "loss": 0.2231, "step": 30650 }, { "epoch": 3.08, "learning_rate": 7.002437251367848e-06, "loss": 0.1032, "step": 30651 }, { "epoch": 3.08, "learning_rate": 7.000900569430817e-06, "loss": 0.0321, "step": 30652 }, { "epoch": 3.08, "learning_rate": 6.999364033851312e-06, "loss": 0.0491, "step": 30653 }, { "epoch": 3.08, "learning_rate": 6.9978276446390856e-06, "loss": 0.1665, "step": 30654 }, { "epoch": 3.08, "learning_rate": 6.996291401803946e-06, "loss": 0.3042, "step": 30655 }, { "epoch": 3.08, "learning_rate": 6.9947553053556525e-06, "loss": 0.0796, "step": 30656 }, { "epoch": 3.08, "learning_rate": 6.993219355303987e-06, "loss": 0.3558, "step": 30657 }, { "epoch": 3.08, "learning_rate": 6.991683551658722e-06, "loss": 0.0848, "step": 30658 }, { "epoch": 3.08, "learning_rate": 6.990147894429622e-06, "loss": 0.0797, "step": 30659 }, { "epoch": 3.08, "learning_rate": 6.988612383626482e-06, "loss": 0.3961, "step": 30660 }, { "epoch": 3.08, "learning_rate": 6.987077019259054e-06, "loss": 0.2692, "step": 30661 }, { "epoch": 3.08, "learning_rate": 6.985541801337102e-06, "loss": 0.0475, "step": 30662 }, { "epoch": 3.08, "learning_rate": 6.984006729870415e-06, "loss": 0.1213, "step": 30663 }, { "epoch": 3.08, "learning_rate": 6.982471804868764e-06, "loss": 0.2596, "step": 30664 }, { "epoch": 3.08, "learning_rate": 6.980937026341886e-06, "loss": 0.3756, "step": 30665 }, { "epoch": 3.08, "learning_rate": 6.979402394299579e-06, "loss": 0.1038, "step": 30666 }, { "epoch": 3.08, "learning_rate": 6.977867908751605e-06, "loss": 0.1114, "step": 30667 }, { "epoch": 3.08, "learning_rate": 6.976333569707703e-06, "loss": 0.1292, "step": 30668 }, { "epoch": 3.08, "learning_rate": 6.974799377177666e-06, "loss": 0.2728, "step": 30669 }, { "epoch": 3.08, "learning_rate": 6.973265331171249e-06, "loss": 0.2115, "step": 30670 }, { "epoch": 3.08, "learning_rate": 6.9717314316982074e-06, "loss": 0.1704, "step": 30671 }, { "epoch": 3.08, "learning_rate": 6.97019767876831e-06, "loss": 0.5583, "step": 30672 }, { "epoch": 3.08, "learning_rate": 6.968664072391312e-06, "loss": 0.1749, "step": 30673 }, { "epoch": 3.08, "learning_rate": 6.967130612576973e-06, "loss": 0.0561, "step": 30674 }, { "epoch": 3.08, "learning_rate": 6.965597299335054e-06, "loss": 0.0872, "step": 30675 }, { "epoch": 3.08, "learning_rate": 6.96406413267531e-06, "loss": 0.2583, "step": 30676 }, { "epoch": 3.08, "learning_rate": 6.9625311126074895e-06, "loss": 0.0633, "step": 30677 }, { "epoch": 3.08, "learning_rate": 6.960998239141377e-06, "loss": 0.0825, "step": 30678 }, { "epoch": 3.08, "learning_rate": 6.959465512286698e-06, "loss": 0.27, "step": 30679 }, { "epoch": 3.08, "learning_rate": 6.957932932053203e-06, "loss": 0.215, "step": 30680 }, { "epoch": 3.08, "learning_rate": 6.95640049845068e-06, "loss": 0.0915, "step": 30681 }, { "epoch": 3.08, "learning_rate": 6.9548682114888465e-06, "loss": 0.0527, "step": 30682 }, { "epoch": 3.08, "learning_rate": 6.953336071177458e-06, "loss": 0.0731, "step": 30683 }, { "epoch": 3.08, "learning_rate": 6.9518040775262755e-06, "loss": 0.0549, "step": 30684 }, { "epoch": 3.08, "learning_rate": 6.9502722305450475e-06, "loss": 0.4171, "step": 30685 }, { "epoch": 3.08, "learning_rate": 6.948740530243517e-06, "loss": 0.2108, "step": 30686 }, { "epoch": 3.08, "learning_rate": 6.947208976631434e-06, "loss": 0.2519, "step": 30687 }, { "epoch": 3.08, "learning_rate": 6.945677569718544e-06, "loss": 0.0829, "step": 30688 }, { "epoch": 3.08, "learning_rate": 6.944146309514589e-06, "loss": 0.4101, "step": 30689 }, { "epoch": 3.08, "learning_rate": 6.942615196029317e-06, "loss": 0.1768, "step": 30690 }, { "epoch": 3.08, "learning_rate": 6.9410842292724615e-06, "loss": 0.1503, "step": 30691 }, { "epoch": 3.08, "learning_rate": 6.939553409253795e-06, "loss": 0.2921, "step": 30692 }, { "epoch": 3.08, "learning_rate": 6.938022735983025e-06, "loss": 0.2769, "step": 30693 }, { "epoch": 3.08, "learning_rate": 6.936492209469898e-06, "loss": 0.174, "step": 30694 }, { "epoch": 3.08, "learning_rate": 6.934961829724169e-06, "loss": 0.0847, "step": 30695 }, { "epoch": 3.08, "learning_rate": 6.933431596755577e-06, "loss": 0.0816, "step": 30696 }, { "epoch": 3.08, "learning_rate": 6.931901510573835e-06, "loss": 0.2336, "step": 30697 }, { "epoch": 3.08, "learning_rate": 6.930371571188704e-06, "loss": 0.2529, "step": 30698 }, { "epoch": 3.08, "learning_rate": 6.92884177860992e-06, "loss": 0.087, "step": 30699 }, { "epoch": 3.08, "learning_rate": 6.927312132847192e-06, "loss": 0.1241, "step": 30700 }, { "epoch": 3.08, "learning_rate": 6.925782633910282e-06, "loss": 0.3685, "step": 30701 }, { "epoch": 3.08, "learning_rate": 6.924253281808917e-06, "loss": 0.1457, "step": 30702 }, { "epoch": 3.08, "learning_rate": 6.922724076552823e-06, "loss": 0.1064, "step": 30703 }, { "epoch": 3.08, "learning_rate": 6.921195018151732e-06, "loss": 0.1285, "step": 30704 }, { "epoch": 3.08, "learning_rate": 6.91966610661538e-06, "loss": 0.0945, "step": 30705 }, { "epoch": 3.08, "learning_rate": 6.918137341953491e-06, "loss": 0.0967, "step": 30706 }, { "epoch": 3.08, "learning_rate": 6.916608724175794e-06, "loss": 0.407, "step": 30707 }, { "epoch": 3.08, "learning_rate": 6.91508025329202e-06, "loss": 0.2458, "step": 30708 }, { "epoch": 3.08, "learning_rate": 6.913551929311882e-06, "loss": 0.0645, "step": 30709 }, { "epoch": 3.08, "learning_rate": 6.912023752245135e-06, "loss": 0.0645, "step": 30710 }, { "epoch": 3.08, "learning_rate": 6.910495722101477e-06, "loss": 0.49, "step": 30711 }, { "epoch": 3.08, "learning_rate": 6.908967838890633e-06, "loss": 0.1362, "step": 30712 }, { "epoch": 3.08, "learning_rate": 6.907440102622351e-06, "loss": 0.2208, "step": 30713 }, { "epoch": 3.08, "learning_rate": 6.905912513306316e-06, "loss": 0.2692, "step": 30714 }, { "epoch": 3.08, "learning_rate": 6.904385070952281e-06, "loss": 0.1744, "step": 30715 }, { "epoch": 3.08, "learning_rate": 6.902857775569949e-06, "loss": 0.1269, "step": 30716 }, { "epoch": 3.08, "learning_rate": 6.9013306271690455e-06, "loss": 0.2267, "step": 30717 }, { "epoch": 3.08, "learning_rate": 6.899803625759284e-06, "loss": 0.0616, "step": 30718 }, { "epoch": 3.08, "learning_rate": 6.898276771350388e-06, "loss": 0.0674, "step": 30719 }, { "epoch": 3.08, "learning_rate": 6.896750063952069e-06, "loss": 0.3341, "step": 30720 }, { "epoch": 3.08, "learning_rate": 6.895223503574045e-06, "loss": 0.2501, "step": 30721 }, { "epoch": 3.08, "learning_rate": 6.8936970902260265e-06, "loss": 0.2665, "step": 30722 }, { "epoch": 3.08, "learning_rate": 6.892170823917722e-06, "loss": 0.036, "step": 30723 }, { "epoch": 3.08, "learning_rate": 6.89064470465887e-06, "loss": 0.0556, "step": 30724 }, { "epoch": 3.08, "learning_rate": 6.889118732459155e-06, "loss": 0.0571, "step": 30725 }, { "epoch": 3.08, "learning_rate": 6.8875929073282885e-06, "loss": 0.1666, "step": 30726 }, { "epoch": 3.08, "learning_rate": 6.886067229275994e-06, "loss": 0.2176, "step": 30727 }, { "epoch": 3.08, "learning_rate": 6.8845416983119844e-06, "loss": 0.0663, "step": 30728 }, { "epoch": 3.08, "learning_rate": 6.883016314445942e-06, "loss": 0.2397, "step": 30729 }, { "epoch": 3.08, "learning_rate": 6.881491077687599e-06, "loss": 0.0926, "step": 30730 }, { "epoch": 3.08, "learning_rate": 6.879965988046658e-06, "loss": 0.0578, "step": 30731 }, { "epoch": 3.08, "learning_rate": 6.878441045532804e-06, "loss": 0.1664, "step": 30732 }, { "epoch": 3.08, "learning_rate": 6.876916250155762e-06, "loss": 0.1971, "step": 30733 }, { "epoch": 3.08, "learning_rate": 6.875391601925229e-06, "loss": 0.3251, "step": 30734 }, { "epoch": 3.08, "learning_rate": 6.8738671008509065e-06, "loss": 0.4062, "step": 30735 }, { "epoch": 3.08, "learning_rate": 6.872342746942497e-06, "loss": 0.1059, "step": 30736 }, { "epoch": 3.08, "learning_rate": 6.870818540209699e-06, "loss": 0.4266, "step": 30737 }, { "epoch": 3.08, "learning_rate": 6.869294480662212e-06, "loss": 0.2664, "step": 30738 }, { "epoch": 3.08, "learning_rate": 6.867770568309737e-06, "loss": 0.1428, "step": 30739 }, { "epoch": 3.08, "learning_rate": 6.86624680316196e-06, "loss": 0.1354, "step": 30740 }, { "epoch": 3.08, "learning_rate": 6.864723185228598e-06, "loss": 0.212, "step": 30741 }, { "epoch": 3.08, "learning_rate": 6.863199714519346e-06, "loss": 0.0568, "step": 30742 }, { "epoch": 3.08, "learning_rate": 6.8616763910438715e-06, "loss": 0.3239, "step": 30743 }, { "epoch": 3.08, "learning_rate": 6.860153214811895e-06, "loss": 0.5051, "step": 30744 }, { "epoch": 3.08, "learning_rate": 6.858630185833108e-06, "loss": 0.1255, "step": 30745 }, { "epoch": 3.09, "learning_rate": 6.857107304117176e-06, "loss": 0.1586, "step": 30746 }, { "epoch": 3.09, "learning_rate": 6.855584569673822e-06, "loss": 0.3609, "step": 30747 }, { "epoch": 3.09, "learning_rate": 6.854061982512721e-06, "loss": 0.1665, "step": 30748 }, { "epoch": 3.09, "learning_rate": 6.852539542643564e-06, "loss": 0.1693, "step": 30749 }, { "epoch": 3.09, "learning_rate": 6.85101725007604e-06, "loss": 0.1649, "step": 30750 }, { "epoch": 3.09, "learning_rate": 6.8494951048198375e-06, "loss": 0.4621, "step": 30751 }, { "epoch": 3.09, "learning_rate": 6.847973106884638e-06, "loss": 0.2151, "step": 30752 }, { "epoch": 3.09, "learning_rate": 6.846451256280131e-06, "loss": 0.1347, "step": 30753 }, { "epoch": 3.09, "learning_rate": 6.844929553016002e-06, "loss": 0.2608, "step": 30754 }, { "epoch": 3.09, "learning_rate": 6.843407997101926e-06, "loss": 0.026, "step": 30755 }, { "epoch": 3.09, "learning_rate": 6.8418865885476065e-06, "loss": 0.4502, "step": 30756 }, { "epoch": 3.09, "learning_rate": 6.840365327362701e-06, "loss": 0.4677, "step": 30757 }, { "epoch": 3.09, "learning_rate": 6.838844213556894e-06, "loss": 0.4704, "step": 30758 }, { "epoch": 3.09, "learning_rate": 6.837323247139878e-06, "loss": 0.1622, "step": 30759 }, { "epoch": 3.09, "learning_rate": 6.835802428121336e-06, "loss": 0.4217, "step": 30760 }, { "epoch": 3.09, "learning_rate": 6.834281756510916e-06, "loss": 0.374, "step": 30761 }, { "epoch": 3.09, "learning_rate": 6.832761232318324e-06, "loss": 0.0643, "step": 30762 }, { "epoch": 3.09, "learning_rate": 6.831240855553236e-06, "loss": 0.1394, "step": 30763 }, { "epoch": 3.09, "learning_rate": 6.8297206262253e-06, "loss": 0.0294, "step": 30764 }, { "epoch": 3.09, "learning_rate": 6.828200544344215e-06, "loss": 0.2019, "step": 30765 }, { "epoch": 3.09, "learning_rate": 6.826680609919648e-06, "loss": 0.3078, "step": 30766 }, { "epoch": 3.09, "learning_rate": 6.825160822961266e-06, "loss": 0.1893, "step": 30767 }, { "epoch": 3.09, "learning_rate": 6.823641183478749e-06, "loss": 0.4404, "step": 30768 }, { "epoch": 3.09, "learning_rate": 6.822121691481753e-06, "loss": 0.6384, "step": 30769 }, { "epoch": 3.09, "learning_rate": 6.8206023469799724e-06, "loss": 1.4152, "step": 30770 }, { "epoch": 3.09, "learning_rate": 6.819083149983054e-06, "loss": 0.1601, "step": 30771 }, { "epoch": 3.09, "learning_rate": 6.817564100500663e-06, "loss": 0.0508, "step": 30772 }, { "epoch": 3.09, "learning_rate": 6.816045198542484e-06, "loss": 0.0775, "step": 30773 }, { "epoch": 3.09, "learning_rate": 6.814526444118181e-06, "loss": 0.1244, "step": 30774 }, { "epoch": 3.09, "learning_rate": 6.813007837237395e-06, "loss": 0.2191, "step": 30775 }, { "epoch": 3.09, "learning_rate": 6.811489377909816e-06, "loss": 0.1395, "step": 30776 }, { "epoch": 3.09, "learning_rate": 6.809971066145104e-06, "loss": 0.294, "step": 30777 }, { "epoch": 3.09, "learning_rate": 6.808452901952897e-06, "loss": 0.5794, "step": 30778 }, { "epoch": 3.09, "learning_rate": 6.806934885342883e-06, "loss": 0.3064, "step": 30779 }, { "epoch": 3.09, "learning_rate": 6.805417016324712e-06, "loss": 0.1405, "step": 30780 }, { "epoch": 3.09, "learning_rate": 6.803899294908045e-06, "loss": 0.3788, "step": 30781 }, { "epoch": 3.09, "learning_rate": 6.802381721102534e-06, "loss": 0.2942, "step": 30782 }, { "epoch": 3.09, "learning_rate": 6.800864294917846e-06, "loss": 0.2006, "step": 30783 }, { "epoch": 3.09, "learning_rate": 6.7993470163636284e-06, "loss": 0.3828, "step": 30784 }, { "epoch": 3.09, "learning_rate": 6.797829885449544e-06, "loss": 0.0883, "step": 30785 }, { "epoch": 3.09, "learning_rate": 6.79631290218524e-06, "loss": 0.2671, "step": 30786 }, { "epoch": 3.09, "learning_rate": 6.794796066580366e-06, "loss": 0.1493, "step": 30787 }, { "epoch": 3.09, "learning_rate": 6.793279378644603e-06, "loss": 0.5117, "step": 30788 }, { "epoch": 3.09, "learning_rate": 6.791762838387569e-06, "loss": 0.0692, "step": 30789 }, { "epoch": 3.09, "learning_rate": 6.79024644581892e-06, "loss": 0.3053, "step": 30790 }, { "epoch": 3.09, "learning_rate": 6.788730200948322e-06, "loss": 0.1269, "step": 30791 }, { "epoch": 3.09, "learning_rate": 6.787214103785423e-06, "loss": 0.0417, "step": 30792 }, { "epoch": 3.09, "learning_rate": 6.785698154339846e-06, "loss": 0.3334, "step": 30793 }, { "epoch": 3.09, "learning_rate": 6.784182352621265e-06, "loss": 0.1273, "step": 30794 }, { "epoch": 3.09, "learning_rate": 6.782666698639315e-06, "loss": 0.0566, "step": 30795 }, { "epoch": 3.09, "learning_rate": 6.7811511924036395e-06, "loss": 0.0928, "step": 30796 }, { "epoch": 3.09, "learning_rate": 6.779635833923885e-06, "loss": 0.1017, "step": 30797 }, { "epoch": 3.09, "learning_rate": 6.778120623209697e-06, "loss": 0.0881, "step": 30798 }, { "epoch": 3.09, "learning_rate": 6.776605560270715e-06, "loss": 0.293, "step": 30799 }, { "epoch": 3.09, "learning_rate": 6.775090645116579e-06, "loss": 0.1343, "step": 30800 }, { "epoch": 3.09, "learning_rate": 6.773575877756924e-06, "loss": 0.3146, "step": 30801 }, { "epoch": 3.09, "learning_rate": 6.7720612582014115e-06, "loss": 0.1556, "step": 30802 }, { "epoch": 3.09, "learning_rate": 6.770546786459657e-06, "loss": 0.2879, "step": 30803 }, { "epoch": 3.09, "learning_rate": 6.769032462541299e-06, "loss": 0.1605, "step": 30804 }, { "epoch": 3.09, "learning_rate": 6.767518286455987e-06, "loss": 0.3106, "step": 30805 }, { "epoch": 3.09, "learning_rate": 6.76600425821336e-06, "loss": 0.1475, "step": 30806 }, { "epoch": 3.09, "learning_rate": 6.764490377823026e-06, "loss": 0.1073, "step": 30807 }, { "epoch": 3.09, "learning_rate": 6.762976645294645e-06, "loss": 0.0331, "step": 30808 }, { "epoch": 3.09, "learning_rate": 6.7614630606378495e-06, "loss": 0.2855, "step": 30809 }, { "epoch": 3.09, "learning_rate": 6.759949623862247e-06, "loss": 0.0605, "step": 30810 }, { "epoch": 3.09, "learning_rate": 6.7584363349774915e-06, "loss": 0.0558, "step": 30811 }, { "epoch": 3.09, "learning_rate": 6.756923193993204e-06, "loss": 0.1163, "step": 30812 }, { "epoch": 3.09, "learning_rate": 6.7554102009190175e-06, "loss": 0.057, "step": 30813 }, { "epoch": 3.09, "learning_rate": 6.7538973557645545e-06, "loss": 0.1103, "step": 30814 }, { "epoch": 3.09, "learning_rate": 6.75238465853945e-06, "loss": 0.1075, "step": 30815 }, { "epoch": 3.09, "learning_rate": 6.750872109253323e-06, "loss": 0.0376, "step": 30816 }, { "epoch": 3.09, "learning_rate": 6.7493597079158e-06, "loss": 0.0928, "step": 30817 }, { "epoch": 3.09, "learning_rate": 6.7478474545365085e-06, "loss": 0.1359, "step": 30818 }, { "epoch": 3.09, "learning_rate": 6.746335349125062e-06, "loss": 0.0259, "step": 30819 }, { "epoch": 3.09, "learning_rate": 6.7448233916911095e-06, "loss": 0.2271, "step": 30820 }, { "epoch": 3.09, "learning_rate": 6.743311582244236e-06, "loss": 0.1771, "step": 30821 }, { "epoch": 3.09, "learning_rate": 6.7417999207940876e-06, "loss": 0.0952, "step": 30822 }, { "epoch": 3.09, "learning_rate": 6.740288407350276e-06, "loss": 0.2519, "step": 30823 }, { "epoch": 3.09, "learning_rate": 6.738777041922424e-06, "loss": 0.1979, "step": 30824 }, { "epoch": 3.09, "learning_rate": 6.737265824520144e-06, "loss": 0.5129, "step": 30825 }, { "epoch": 3.09, "learning_rate": 6.735754755153054e-06, "loss": 0.0826, "step": 30826 }, { "epoch": 3.09, "learning_rate": 6.734243833830771e-06, "loss": 0.0137, "step": 30827 }, { "epoch": 3.09, "learning_rate": 6.73273306056291e-06, "loss": 0.398, "step": 30828 }, { "epoch": 3.09, "learning_rate": 6.731222435359084e-06, "loss": 0.0448, "step": 30829 }, { "epoch": 3.09, "learning_rate": 6.729711958228907e-06, "loss": 0.3557, "step": 30830 }, { "epoch": 3.09, "learning_rate": 6.728201629181989e-06, "loss": 0.1278, "step": 30831 }, { "epoch": 3.09, "learning_rate": 6.7266914482279465e-06, "loss": 0.3263, "step": 30832 }, { "epoch": 3.09, "learning_rate": 6.725181415376374e-06, "loss": 0.1443, "step": 30833 }, { "epoch": 3.09, "learning_rate": 6.72367153063691e-06, "loss": 0.2107, "step": 30834 }, { "epoch": 3.09, "learning_rate": 6.722161794019139e-06, "loss": 0.2777, "step": 30835 }, { "epoch": 3.09, "learning_rate": 6.7206522055326635e-06, "loss": 0.4321, "step": 30836 }, { "epoch": 3.09, "learning_rate": 6.719142765187115e-06, "loss": 0.0751, "step": 30837 }, { "epoch": 3.09, "learning_rate": 6.717633472992088e-06, "loss": 0.2574, "step": 30838 }, { "epoch": 3.09, "learning_rate": 6.716124328957173e-06, "loss": 0.2946, "step": 30839 }, { "epoch": 3.09, "learning_rate": 6.71461533309199e-06, "loss": 0.3182, "step": 30840 }, { "epoch": 3.09, "learning_rate": 6.713106485406146e-06, "loss": 0.033, "step": 30841 }, { "epoch": 3.09, "learning_rate": 6.711597785909217e-06, "loss": 0.5247, "step": 30842 }, { "epoch": 3.09, "learning_rate": 6.710089234610828e-06, "loss": 0.2397, "step": 30843 }, { "epoch": 3.1, "learning_rate": 6.708580831520575e-06, "loss": 0.3913, "step": 30844 }, { "epoch": 3.1, "learning_rate": 6.707072576648049e-06, "loss": 0.3817, "step": 30845 }, { "epoch": 3.1, "learning_rate": 6.705564470002851e-06, "loss": 0.0538, "step": 30846 }, { "epoch": 3.1, "learning_rate": 6.704056511594575e-06, "loss": 0.0352, "step": 30847 }, { "epoch": 3.1, "learning_rate": 6.702548701432835e-06, "loss": 0.3751, "step": 30848 }, { "epoch": 3.1, "learning_rate": 6.7010410395272045e-06, "loss": 0.1022, "step": 30849 }, { "epoch": 3.1, "learning_rate": 6.699533525887275e-06, "loss": 0.3964, "step": 30850 }, { "epoch": 3.1, "learning_rate": 6.6980261605226635e-06, "loss": 0.3436, "step": 30851 }, { "epoch": 3.1, "learning_rate": 6.696518943442953e-06, "loss": 0.5571, "step": 30852 }, { "epoch": 3.1, "learning_rate": 6.695011874657716e-06, "loss": 0.2259, "step": 30853 }, { "epoch": 3.1, "learning_rate": 6.693504954176564e-06, "loss": 0.1071, "step": 30854 }, { "epoch": 3.1, "learning_rate": 6.691998182009081e-06, "loss": 0.1246, "step": 30855 }, { "epoch": 3.1, "learning_rate": 6.690491558164859e-06, "loss": 0.0759, "step": 30856 }, { "epoch": 3.1, "learning_rate": 6.688985082653478e-06, "loss": 0.0882, "step": 30857 }, { "epoch": 3.1, "learning_rate": 6.687478755484527e-06, "loss": 0.5127, "step": 30858 }, { "epoch": 3.1, "learning_rate": 6.685972576667592e-06, "loss": 0.1468, "step": 30859 }, { "epoch": 3.1, "learning_rate": 6.68446654621226e-06, "loss": 0.251, "step": 30860 }, { "epoch": 3.1, "learning_rate": 6.6829606641281136e-06, "loss": 0.1306, "step": 30861 }, { "epoch": 3.1, "learning_rate": 6.6814549304247354e-06, "loss": 0.0279, "step": 30862 }, { "epoch": 3.1, "learning_rate": 6.679949345111707e-06, "loss": 0.1451, "step": 30863 }, { "epoch": 3.1, "learning_rate": 6.6784439081986105e-06, "loss": 0.7692, "step": 30864 }, { "epoch": 3.1, "learning_rate": 6.676938619695016e-06, "loss": 0.1116, "step": 30865 }, { "epoch": 3.1, "learning_rate": 6.6754334796105274e-06, "loss": 0.0978, "step": 30866 }, { "epoch": 3.1, "learning_rate": 6.673928487954699e-06, "loss": 0.17, "step": 30867 }, { "epoch": 3.1, "learning_rate": 6.6724236447371055e-06, "loss": 0.1248, "step": 30868 }, { "epoch": 3.1, "learning_rate": 6.670918949967346e-06, "loss": 0.1709, "step": 30869 }, { "epoch": 3.1, "learning_rate": 6.66941440365499e-06, "loss": 0.5439, "step": 30870 }, { "epoch": 3.1, "learning_rate": 6.667910005809589e-06, "loss": 0.1192, "step": 30871 }, { "epoch": 3.1, "learning_rate": 6.666405756440739e-06, "loss": 0.1188, "step": 30872 }, { "epoch": 3.1, "learning_rate": 6.664901655558015e-06, "loss": 0.1929, "step": 30873 }, { "epoch": 3.1, "learning_rate": 6.6633977031709645e-06, "loss": 0.0527, "step": 30874 }, { "epoch": 3.1, "learning_rate": 6.661893899289182e-06, "loss": 0.238, "step": 30875 }, { "epoch": 3.1, "learning_rate": 6.66039024392223e-06, "loss": 0.0659, "step": 30876 }, { "epoch": 3.1, "learning_rate": 6.658886737079673e-06, "loss": 0.5601, "step": 30877 }, { "epoch": 3.1, "learning_rate": 6.6573833787710825e-06, "loss": 0.1562, "step": 30878 }, { "epoch": 3.1, "learning_rate": 6.655880169006015e-06, "loss": 0.2856, "step": 30879 }, { "epoch": 3.1, "learning_rate": 6.6543771077940615e-06, "loss": 0.2922, "step": 30880 }, { "epoch": 3.1, "learning_rate": 6.652874195144764e-06, "loss": 0.0397, "step": 30881 }, { "epoch": 3.1, "learning_rate": 6.651371431067683e-06, "loss": 0.129, "step": 30882 }, { "epoch": 3.1, "learning_rate": 6.649868815572403e-06, "loss": 0.2026, "step": 30883 }, { "epoch": 3.1, "learning_rate": 6.648366348668482e-06, "loss": 0.0066, "step": 30884 }, { "epoch": 3.1, "learning_rate": 6.646864030365459e-06, "loss": 0.4522, "step": 30885 }, { "epoch": 3.1, "learning_rate": 6.645361860672917e-06, "loss": 0.7764, "step": 30886 }, { "epoch": 3.1, "learning_rate": 6.643859839600408e-06, "loss": 0.109, "step": 30887 }, { "epoch": 3.1, "learning_rate": 6.642357967157488e-06, "loss": 0.1781, "step": 30888 }, { "epoch": 3.1, "learning_rate": 6.6408562433537176e-06, "loss": 0.3191, "step": 30889 }, { "epoch": 3.1, "learning_rate": 6.639354668198652e-06, "loss": 1.5571, "step": 30890 }, { "epoch": 3.1, "learning_rate": 6.637853241701844e-06, "loss": 0.0485, "step": 30891 }, { "epoch": 3.1, "learning_rate": 6.636351963872852e-06, "loss": 0.2343, "step": 30892 }, { "epoch": 3.1, "learning_rate": 6.63485083472123e-06, "loss": 0.0249, "step": 30893 }, { "epoch": 3.1, "learning_rate": 6.633349854256528e-06, "loss": 0.0607, "step": 30894 }, { "epoch": 3.1, "learning_rate": 6.631849022488297e-06, "loss": 0.1861, "step": 30895 }, { "epoch": 3.1, "learning_rate": 6.630348339426086e-06, "loss": 0.1073, "step": 30896 }, { "epoch": 3.1, "learning_rate": 6.628847805079446e-06, "loss": 0.1809, "step": 30897 }, { "epoch": 3.1, "learning_rate": 6.627347419457942e-06, "loss": 0.4609, "step": 30898 }, { "epoch": 3.1, "learning_rate": 6.625847182571099e-06, "loss": 0.1395, "step": 30899 }, { "epoch": 3.1, "learning_rate": 6.6243470944284625e-06, "loss": 0.1407, "step": 30900 }, { "epoch": 3.1, "learning_rate": 6.622847155039598e-06, "loss": 0.182, "step": 30901 }, { "epoch": 3.1, "learning_rate": 6.621347364414043e-06, "loss": 0.3445, "step": 30902 }, { "epoch": 3.1, "learning_rate": 6.619847722561339e-06, "loss": 0.1545, "step": 30903 }, { "epoch": 3.1, "learning_rate": 6.618348229491027e-06, "loss": 0.0877, "step": 30904 }, { "epoch": 3.1, "learning_rate": 6.6168488852126545e-06, "loss": 0.1875, "step": 30905 }, { "epoch": 3.1, "learning_rate": 6.615349689735759e-06, "loss": 0.676, "step": 30906 }, { "epoch": 3.1, "learning_rate": 6.6138506430698846e-06, "loss": 0.1676, "step": 30907 }, { "epoch": 3.1, "learning_rate": 6.6123517452245656e-06, "loss": 0.3193, "step": 30908 }, { "epoch": 3.1, "learning_rate": 6.610852996209345e-06, "loss": 0.1677, "step": 30909 }, { "epoch": 3.1, "learning_rate": 6.609354396033758e-06, "loss": 0.4133, "step": 30910 }, { "epoch": 3.1, "learning_rate": 6.60785594470733e-06, "loss": 0.1453, "step": 30911 }, { "epoch": 3.1, "learning_rate": 6.6063576422396284e-06, "loss": 0.0716, "step": 30912 }, { "epoch": 3.1, "learning_rate": 6.604859488640158e-06, "loss": 0.1568, "step": 30913 }, { "epoch": 3.1, "learning_rate": 6.603361483918451e-06, "loss": 0.053, "step": 30914 }, { "epoch": 3.1, "learning_rate": 6.601863628084063e-06, "loss": 0.1187, "step": 30915 }, { "epoch": 3.1, "learning_rate": 6.60036592114652e-06, "loss": 0.2336, "step": 30916 }, { "epoch": 3.1, "learning_rate": 6.5988683631153305e-06, "loss": 0.0734, "step": 30917 }, { "epoch": 3.1, "learning_rate": 6.5973709540000475e-06, "loss": 0.1871, "step": 30918 }, { "epoch": 3.1, "learning_rate": 6.595873693810196e-06, "loss": 0.0176, "step": 30919 }, { "epoch": 3.1, "learning_rate": 6.5943765825552994e-06, "loss": 0.3514, "step": 30920 }, { "epoch": 3.1, "learning_rate": 6.5928796202448825e-06, "loss": 0.0891, "step": 30921 }, { "epoch": 3.1, "learning_rate": 6.5913828068884814e-06, "loss": 0.3882, "step": 30922 }, { "epoch": 3.1, "learning_rate": 6.589886142495613e-06, "loss": 0.0956, "step": 30923 }, { "epoch": 3.1, "learning_rate": 6.5883896270758e-06, "loss": 0.0637, "step": 30924 }, { "epoch": 3.1, "learning_rate": 6.5868932606385715e-06, "loss": 0.0361, "step": 30925 }, { "epoch": 3.1, "learning_rate": 6.585397043193447e-06, "loss": 0.388, "step": 30926 }, { "epoch": 3.1, "learning_rate": 6.583900974749951e-06, "loss": 0.3802, "step": 30927 }, { "epoch": 3.1, "learning_rate": 6.582405055317592e-06, "loss": 0.2692, "step": 30928 }, { "epoch": 3.1, "learning_rate": 6.580909284905905e-06, "loss": 0.1621, "step": 30929 }, { "epoch": 3.1, "learning_rate": 6.579413663524413e-06, "loss": 0.5778, "step": 30930 }, { "epoch": 3.1, "learning_rate": 6.577918191182605e-06, "loss": 0.1303, "step": 30931 }, { "epoch": 3.1, "learning_rate": 6.576422867890023e-06, "loss": 0.1384, "step": 30932 }, { "epoch": 3.1, "learning_rate": 6.574927693656176e-06, "loss": 0.3358, "step": 30933 }, { "epoch": 3.1, "learning_rate": 6.573432668490578e-06, "loss": 0.1304, "step": 30934 }, { "epoch": 3.1, "learning_rate": 6.5719377924027426e-06, "loss": 0.0959, "step": 30935 }, { "epoch": 3.1, "learning_rate": 6.570443065402179e-06, "loss": 0.1738, "step": 30936 }, { "epoch": 3.1, "learning_rate": 6.568948487498409e-06, "loss": 0.3412, "step": 30937 }, { "epoch": 3.1, "learning_rate": 6.567454058700931e-06, "loss": 0.108, "step": 30938 }, { "epoch": 3.1, "learning_rate": 6.565959779019267e-06, "loss": 0.2011, "step": 30939 }, { "epoch": 3.1, "learning_rate": 6.564465648462915e-06, "loss": 0.1193, "step": 30940 }, { "epoch": 3.1, "learning_rate": 6.562971667041391e-06, "loss": 0.1945, "step": 30941 }, { "epoch": 3.1, "learning_rate": 6.5614778347642e-06, "loss": 0.35, "step": 30942 }, { "epoch": 3.11, "learning_rate": 6.5599841516408385e-06, "loss": 0.1919, "step": 30943 }, { "epoch": 3.11, "learning_rate": 6.558490617680836e-06, "loss": 0.2191, "step": 30944 }, { "epoch": 3.11, "learning_rate": 6.556997232893677e-06, "loss": 0.1012, "step": 30945 }, { "epoch": 3.11, "learning_rate": 6.555503997288858e-06, "loss": 0.3354, "step": 30946 }, { "epoch": 3.11, "learning_rate": 6.554010910875901e-06, "loss": 0.1328, "step": 30947 }, { "epoch": 3.11, "learning_rate": 6.552517973664308e-06, "loss": 0.24, "step": 30948 }, { "epoch": 3.11, "learning_rate": 6.551025185663556e-06, "loss": 0.0436, "step": 30949 }, { "epoch": 3.11, "learning_rate": 6.549532546883164e-06, "loss": 0.0513, "step": 30950 }, { "epoch": 3.11, "learning_rate": 6.54804005733263e-06, "loss": 0.3428, "step": 30951 }, { "epoch": 3.11, "learning_rate": 6.546547717021443e-06, "loss": 0.5283, "step": 30952 }, { "epoch": 3.11, "learning_rate": 6.545055525959106e-06, "loss": 0.333, "step": 30953 }, { "epoch": 3.11, "learning_rate": 6.543563484155115e-06, "loss": 0.1058, "step": 30954 }, { "epoch": 3.11, "learning_rate": 6.542071591618962e-06, "loss": 0.1083, "step": 30955 }, { "epoch": 3.11, "learning_rate": 6.54057984836014e-06, "loss": 0.1646, "step": 30956 }, { "epoch": 3.11, "learning_rate": 6.539088254388137e-06, "loss": 0.0795, "step": 30957 }, { "epoch": 3.11, "learning_rate": 6.537596809712466e-06, "loss": 0.2522, "step": 30958 }, { "epoch": 3.11, "learning_rate": 6.536105514342596e-06, "loss": 0.2025, "step": 30959 }, { "epoch": 3.11, "learning_rate": 6.534614368288016e-06, "loss": 0.3414, "step": 30960 }, { "epoch": 3.11, "learning_rate": 6.533123371558233e-06, "loss": 0.2134, "step": 30961 }, { "epoch": 3.11, "learning_rate": 6.53163252416273e-06, "loss": 0.135, "step": 30962 }, { "epoch": 3.11, "learning_rate": 6.530141826110976e-06, "loss": 0.105, "step": 30963 }, { "epoch": 3.11, "learning_rate": 6.5286512774124805e-06, "loss": 0.2126, "step": 30964 }, { "epoch": 3.11, "learning_rate": 6.527160878076718e-06, "loss": 0.0489, "step": 30965 }, { "epoch": 3.11, "learning_rate": 6.525670628113173e-06, "loss": 0.0813, "step": 30966 }, { "epoch": 3.11, "learning_rate": 6.524180527531331e-06, "loss": 0.5293, "step": 30967 }, { "epoch": 3.11, "learning_rate": 6.522690576340677e-06, "loss": 0.0938, "step": 30968 }, { "epoch": 3.11, "learning_rate": 6.521200774550683e-06, "loss": 0.4036, "step": 30969 }, { "epoch": 3.11, "learning_rate": 6.519711122170837e-06, "loss": 0.4879, "step": 30970 }, { "epoch": 3.11, "learning_rate": 6.5182216192106184e-06, "loss": 0.0354, "step": 30971 }, { "epoch": 3.11, "learning_rate": 6.5167322656795045e-06, "loss": 0.5343, "step": 30972 }, { "epoch": 3.11, "learning_rate": 6.515243061586973e-06, "loss": 0.1224, "step": 30973 }, { "epoch": 3.11, "learning_rate": 6.513754006942502e-06, "loss": 0.1698, "step": 30974 }, { "epoch": 3.11, "learning_rate": 6.512265101755553e-06, "loss": 0.3135, "step": 30975 }, { "epoch": 3.11, "learning_rate": 6.510776346035634e-06, "loss": 0.1928, "step": 30976 }, { "epoch": 3.11, "learning_rate": 6.5092877397921894e-06, "loss": 0.3733, "step": 30977 }, { "epoch": 3.11, "learning_rate": 6.507799283034694e-06, "loss": 0.1521, "step": 30978 }, { "epoch": 3.11, "learning_rate": 6.506310975772636e-06, "loss": 0.3992, "step": 30979 }, { "epoch": 3.11, "learning_rate": 6.5048228180154824e-06, "loss": 0.7035, "step": 30980 }, { "epoch": 3.11, "learning_rate": 6.503334809772683e-06, "loss": 0.0827, "step": 30981 }, { "epoch": 3.11, "learning_rate": 6.501846951053731e-06, "loss": 0.0413, "step": 30982 }, { "epoch": 3.11, "learning_rate": 6.500359241868087e-06, "loss": 0.1359, "step": 30983 }, { "epoch": 3.11, "learning_rate": 6.498871682225215e-06, "loss": 0.2962, "step": 30984 }, { "epoch": 3.11, "learning_rate": 6.497384272134586e-06, "loss": 0.2148, "step": 30985 }, { "epoch": 3.11, "learning_rate": 6.495897011605662e-06, "loss": 0.303, "step": 30986 }, { "epoch": 3.11, "learning_rate": 6.494409900647906e-06, "loss": 0.1981, "step": 30987 }, { "epoch": 3.11, "learning_rate": 6.492922939270783e-06, "loss": 0.0666, "step": 30988 }, { "epoch": 3.11, "learning_rate": 6.491436127483749e-06, "loss": 0.1549, "step": 30989 }, { "epoch": 3.11, "learning_rate": 6.48994946529629e-06, "loss": 0.0412, "step": 30990 }, { "epoch": 3.11, "learning_rate": 6.488462952717839e-06, "loss": 0.1834, "step": 30991 }, { "epoch": 3.11, "learning_rate": 6.486976589757857e-06, "loss": 0.0874, "step": 30992 }, { "epoch": 3.11, "learning_rate": 6.485490376425821e-06, "loss": 0.0464, "step": 30993 }, { "epoch": 3.11, "learning_rate": 6.484004312731187e-06, "loss": 0.1763, "step": 30994 }, { "epoch": 3.11, "learning_rate": 6.4825183986833856e-06, "loss": 0.0688, "step": 30995 }, { "epoch": 3.11, "learning_rate": 6.481032634291901e-06, "loss": 0.0452, "step": 30996 }, { "epoch": 3.11, "learning_rate": 6.479547019566176e-06, "loss": 0.1966, "step": 30997 }, { "epoch": 3.11, "learning_rate": 6.478061554515668e-06, "loss": 0.2838, "step": 30998 }, { "epoch": 3.11, "learning_rate": 6.476576239149826e-06, "loss": 0.2491, "step": 30999 }, { "epoch": 3.11, "learning_rate": 6.475091073478102e-06, "loss": 0.1961, "step": 31000 }, { "epoch": 3.11, "learning_rate": 6.473606057509954e-06, "loss": 0.0699, "step": 31001 }, { "epoch": 3.11, "learning_rate": 6.472121191254825e-06, "loss": 0.1537, "step": 31002 }, { "epoch": 3.11, "learning_rate": 6.470636474722164e-06, "loss": 0.3564, "step": 31003 }, { "epoch": 3.11, "learning_rate": 6.469151907921425e-06, "loss": 0.1056, "step": 31004 }, { "epoch": 3.11, "learning_rate": 6.46766749086205e-06, "loss": 0.4392, "step": 31005 }, { "epoch": 3.11, "learning_rate": 6.466183223553487e-06, "loss": 0.4874, "step": 31006 }, { "epoch": 3.11, "learning_rate": 6.464699106005174e-06, "loss": 0.0997, "step": 31007 }, { "epoch": 3.11, "learning_rate": 6.463215138226578e-06, "loss": 0.2358, "step": 31008 }, { "epoch": 3.11, "learning_rate": 6.461731320227118e-06, "loss": 0.2304, "step": 31009 }, { "epoch": 3.11, "learning_rate": 6.460247652016239e-06, "loss": 0.5555, "step": 31010 }, { "epoch": 3.11, "learning_rate": 6.458764133603398e-06, "loss": 0.1971, "step": 31011 }, { "epoch": 3.11, "learning_rate": 6.457280764998023e-06, "loss": 0.204, "step": 31012 }, { "epoch": 3.11, "learning_rate": 6.45579754620956e-06, "loss": 0.2936, "step": 31013 }, { "epoch": 3.11, "learning_rate": 6.454314477247445e-06, "loss": 0.174, "step": 31014 }, { "epoch": 3.11, "learning_rate": 6.452831558121114e-06, "loss": 0.0772, "step": 31015 }, { "epoch": 3.11, "learning_rate": 6.4513487888400025e-06, "loss": 0.065, "step": 31016 }, { "epoch": 3.11, "learning_rate": 6.449866169413553e-06, "loss": 0.3321, "step": 31017 }, { "epoch": 3.11, "learning_rate": 6.448383699851195e-06, "loss": 0.5411, "step": 31018 }, { "epoch": 3.11, "learning_rate": 6.446901380162363e-06, "loss": 0.2222, "step": 31019 }, { "epoch": 3.11, "learning_rate": 6.445419210356491e-06, "loss": 0.4038, "step": 31020 }, { "epoch": 3.11, "learning_rate": 6.443937190443005e-06, "loss": 0.0446, "step": 31021 }, { "epoch": 3.11, "learning_rate": 6.442455320431355e-06, "loss": 0.1725, "step": 31022 }, { "epoch": 3.11, "learning_rate": 6.440973600330949e-06, "loss": 0.036, "step": 31023 }, { "epoch": 3.11, "learning_rate": 6.439492030151219e-06, "loss": 0.8881, "step": 31024 }, { "epoch": 3.11, "learning_rate": 6.438010609901605e-06, "loss": 0.1579, "step": 31025 }, { "epoch": 3.11, "learning_rate": 6.436529339591537e-06, "loss": 0.1504, "step": 31026 }, { "epoch": 3.11, "learning_rate": 6.435048219230416e-06, "loss": 0.0415, "step": 31027 }, { "epoch": 3.11, "learning_rate": 6.433567248827691e-06, "loss": 0.2854, "step": 31028 }, { "epoch": 3.11, "learning_rate": 6.43208642839278e-06, "loss": 0.2898, "step": 31029 }, { "epoch": 3.11, "learning_rate": 6.430605757935104e-06, "loss": 0.2701, "step": 31030 }, { "epoch": 3.11, "learning_rate": 6.429125237464088e-06, "loss": 0.0622, "step": 31031 }, { "epoch": 3.11, "learning_rate": 6.42764486698915e-06, "loss": 0.1982, "step": 31032 }, { "epoch": 3.11, "learning_rate": 6.4261646465197134e-06, "loss": 0.0247, "step": 31033 }, { "epoch": 3.11, "learning_rate": 6.4246845760651974e-06, "loss": 0.122, "step": 31034 }, { "epoch": 3.11, "learning_rate": 6.423204655635021e-06, "loss": 0.1305, "step": 31035 }, { "epoch": 3.11, "learning_rate": 6.4217248852386e-06, "loss": 0.4035, "step": 31036 }, { "epoch": 3.11, "learning_rate": 6.420245264885353e-06, "loss": 0.287, "step": 31037 }, { "epoch": 3.11, "learning_rate": 6.418765794584688e-06, "loss": 0.1317, "step": 31038 }, { "epoch": 3.11, "learning_rate": 6.417286474346034e-06, "loss": 0.0276, "step": 31039 }, { "epoch": 3.11, "learning_rate": 6.415807304178807e-06, "loss": 0.1757, "step": 31040 }, { "epoch": 3.12, "learning_rate": 6.41432828409239e-06, "loss": 0.0902, "step": 31041 }, { "epoch": 3.12, "learning_rate": 6.412849414096227e-06, "loss": 0.2262, "step": 31042 }, { "epoch": 3.12, "learning_rate": 6.411370694199717e-06, "loss": 0.2714, "step": 31043 }, { "epoch": 3.12, "learning_rate": 6.409892124412268e-06, "loss": 0.0967, "step": 31044 }, { "epoch": 3.12, "learning_rate": 6.4084137047432935e-06, "loss": 0.3596, "step": 31045 }, { "epoch": 3.12, "learning_rate": 6.406935435202199e-06, "loss": 0.049, "step": 31046 }, { "epoch": 3.12, "learning_rate": 6.4054573157983906e-06, "loss": 0.0971, "step": 31047 }, { "epoch": 3.12, "learning_rate": 6.403979346541278e-06, "loss": 0.1573, "step": 31048 }, { "epoch": 3.12, "learning_rate": 6.402501527440265e-06, "loss": 0.1132, "step": 31049 }, { "epoch": 3.12, "learning_rate": 6.401023858504753e-06, "loss": 0.2984, "step": 31050 }, { "epoch": 3.12, "learning_rate": 6.39954633974415e-06, "loss": 0.0225, "step": 31051 }, { "epoch": 3.12, "learning_rate": 6.398068971167856e-06, "loss": 0.1322, "step": 31052 }, { "epoch": 3.12, "learning_rate": 6.396591752785261e-06, "loss": 0.0596, "step": 31053 }, { "epoch": 3.12, "learning_rate": 6.395114684605797e-06, "loss": 0.136, "step": 31054 }, { "epoch": 3.12, "learning_rate": 6.393637766638832e-06, "loss": 0.0415, "step": 31055 }, { "epoch": 3.12, "learning_rate": 6.392160998893769e-06, "loss": 0.2165, "step": 31056 }, { "epoch": 3.12, "learning_rate": 6.3906843813800235e-06, "loss": 0.4996, "step": 31057 }, { "epoch": 3.12, "learning_rate": 6.389207914106987e-06, "loss": 0.1939, "step": 31058 }, { "epoch": 3.12, "learning_rate": 6.387731597084033e-06, "loss": 0.1353, "step": 31059 }, { "epoch": 3.12, "learning_rate": 6.386255430320581e-06, "loss": 0.2017, "step": 31060 }, { "epoch": 3.12, "learning_rate": 6.384779413826016e-06, "loss": 0.2008, "step": 31061 }, { "epoch": 3.12, "learning_rate": 6.383303547609731e-06, "loss": 0.1308, "step": 31062 }, { "epoch": 3.12, "learning_rate": 6.381827831681115e-06, "loss": 0.9435, "step": 31063 }, { "epoch": 3.12, "learning_rate": 6.380352266049567e-06, "loss": 0.5451, "step": 31064 }, { "epoch": 3.12, "learning_rate": 6.3788768507244695e-06, "loss": 0.0902, "step": 31065 }, { "epoch": 3.12, "learning_rate": 6.377401585715212e-06, "loss": 0.3156, "step": 31066 }, { "epoch": 3.12, "learning_rate": 6.37592647103118e-06, "loss": 0.1378, "step": 31067 }, { "epoch": 3.12, "learning_rate": 6.37445150668178e-06, "loss": 0.3639, "step": 31068 }, { "epoch": 3.12, "learning_rate": 6.372976692676373e-06, "loss": 0.1312, "step": 31069 }, { "epoch": 3.12, "learning_rate": 6.371502029024347e-06, "loss": 0.2404, "step": 31070 }, { "epoch": 3.12, "learning_rate": 6.370027515735102e-06, "loss": 0.3451, "step": 31071 }, { "epoch": 3.12, "learning_rate": 6.368553152818022e-06, "loss": 0.17, "step": 31072 }, { "epoch": 3.12, "learning_rate": 6.3670789402824624e-06, "loss": 0.024, "step": 31073 }, { "epoch": 3.12, "learning_rate": 6.36560487813783e-06, "loss": 0.1454, "step": 31074 }, { "epoch": 3.12, "learning_rate": 6.364130966393496e-06, "loss": 0.2643, "step": 31075 }, { "epoch": 3.12, "learning_rate": 6.362657205058843e-06, "loss": 0.2334, "step": 31076 }, { "epoch": 3.12, "learning_rate": 6.361183594143244e-06, "loss": 0.0187, "step": 31077 }, { "epoch": 3.12, "learning_rate": 6.359710133656082e-06, "loss": 0.2278, "step": 31078 }, { "epoch": 3.12, "learning_rate": 6.358236823606731e-06, "loss": 0.2009, "step": 31079 }, { "epoch": 3.12, "learning_rate": 6.356763664004569e-06, "loss": 0.4308, "step": 31080 }, { "epoch": 3.12, "learning_rate": 6.355290654858967e-06, "loss": 0.0633, "step": 31081 }, { "epoch": 3.12, "learning_rate": 6.353817796179302e-06, "loss": 0.1132, "step": 31082 }, { "epoch": 3.12, "learning_rate": 6.352345087974946e-06, "loss": 0.1241, "step": 31083 }, { "epoch": 3.12, "learning_rate": 6.3508725302552675e-06, "loss": 0.2044, "step": 31084 }, { "epoch": 3.12, "learning_rate": 6.3494001230296295e-06, "loss": 0.1331, "step": 31085 }, { "epoch": 3.12, "learning_rate": 6.347927866307431e-06, "loss": 0.0876, "step": 31086 }, { "epoch": 3.12, "learning_rate": 6.3464557600980125e-06, "loss": 0.0852, "step": 31087 }, { "epoch": 3.12, "learning_rate": 6.344983804410744e-06, "loss": 0.2577, "step": 31088 }, { "epoch": 3.12, "learning_rate": 6.343511999255007e-06, "loss": 0.1332, "step": 31089 }, { "epoch": 3.12, "learning_rate": 6.34204034464017e-06, "loss": 0.1216, "step": 31090 }, { "epoch": 3.12, "learning_rate": 6.340568840575571e-06, "loss": 0.1465, "step": 31091 }, { "epoch": 3.12, "learning_rate": 6.339097487070601e-06, "loss": 0.0714, "step": 31092 }, { "epoch": 3.12, "learning_rate": 6.3376262841346115e-06, "loss": 0.1341, "step": 31093 }, { "epoch": 3.12, "learning_rate": 6.3361552317769676e-06, "loss": 0.0857, "step": 31094 }, { "epoch": 3.12, "learning_rate": 6.33468433000703e-06, "loss": 0.193, "step": 31095 }, { "epoch": 3.12, "learning_rate": 6.333213578834157e-06, "loss": 0.2037, "step": 31096 }, { "epoch": 3.12, "learning_rate": 6.331742978267711e-06, "loss": 0.2729, "step": 31097 }, { "epoch": 3.12, "learning_rate": 6.330272528317049e-06, "loss": 0.1037, "step": 31098 }, { "epoch": 3.12, "learning_rate": 6.328802228991519e-06, "loss": 0.049, "step": 31099 }, { "epoch": 3.12, "learning_rate": 6.327332080300504e-06, "loss": 0.2545, "step": 31100 }, { "epoch": 3.12, "learning_rate": 6.325862082253333e-06, "loss": 0.4179, "step": 31101 }, { "epoch": 3.12, "learning_rate": 6.324392234859362e-06, "loss": 0.6509, "step": 31102 }, { "epoch": 3.12, "learning_rate": 6.32292253812796e-06, "loss": 0.2154, "step": 31103 }, { "epoch": 3.12, "learning_rate": 6.321452992068479e-06, "loss": 0.046, "step": 31104 }, { "epoch": 3.12, "learning_rate": 6.3199835966902495e-06, "loss": 0.374, "step": 31105 }, { "epoch": 3.12, "learning_rate": 6.31851435200264e-06, "loss": 0.4551, "step": 31106 }, { "epoch": 3.12, "learning_rate": 6.3170452580149996e-06, "loss": 0.4276, "step": 31107 }, { "epoch": 3.12, "learning_rate": 6.315576314736674e-06, "loss": 0.3773, "step": 31108 }, { "epoch": 3.12, "learning_rate": 6.31410752217701e-06, "loss": 0.2186, "step": 31109 }, { "epoch": 3.12, "learning_rate": 6.312638880345355e-06, "loss": 0.0414, "step": 31110 }, { "epoch": 3.12, "learning_rate": 6.311170389251056e-06, "loss": 0.0657, "step": 31111 }, { "epoch": 3.12, "learning_rate": 6.309702048903456e-06, "loss": 0.0561, "step": 31112 }, { "epoch": 3.12, "learning_rate": 6.3082338593119005e-06, "loss": 0.4158, "step": 31113 }, { "epoch": 3.12, "learning_rate": 6.306765820485728e-06, "loss": 0.0583, "step": 31114 }, { "epoch": 3.12, "learning_rate": 6.305297932434288e-06, "loss": 0.1082, "step": 31115 }, { "epoch": 3.12, "learning_rate": 6.303830195166917e-06, "loss": 0.0627, "step": 31116 }, { "epoch": 3.12, "learning_rate": 6.302362608692951e-06, "loss": 0.7224, "step": 31117 }, { "epoch": 3.12, "learning_rate": 6.300895173021748e-06, "loss": 0.1906, "step": 31118 }, { "epoch": 3.12, "learning_rate": 6.299427888162617e-06, "loss": 0.5436, "step": 31119 }, { "epoch": 3.12, "learning_rate": 6.29796075412492e-06, "loss": 0.2893, "step": 31120 }, { "epoch": 3.12, "learning_rate": 6.296493770917979e-06, "loss": 0.1908, "step": 31121 }, { "epoch": 3.12, "learning_rate": 6.29502693855114e-06, "loss": 0.2448, "step": 31122 }, { "epoch": 3.12, "learning_rate": 6.293560257033729e-06, "loss": 0.1713, "step": 31123 }, { "epoch": 3.12, "learning_rate": 6.292093726375085e-06, "loss": 0.0506, "step": 31124 }, { "epoch": 3.12, "learning_rate": 6.290627346584534e-06, "loss": 0.0443, "step": 31125 }, { "epoch": 3.12, "learning_rate": 6.289161117671412e-06, "loss": 0.5837, "step": 31126 }, { "epoch": 3.12, "learning_rate": 6.287695039645048e-06, "loss": 0.2325, "step": 31127 }, { "epoch": 3.12, "learning_rate": 6.286229112514774e-06, "loss": 0.051, "step": 31128 }, { "epoch": 3.12, "learning_rate": 6.284763336289915e-06, "loss": 0.4928, "step": 31129 }, { "epoch": 3.12, "learning_rate": 6.283297710979803e-06, "loss": 0.4062, "step": 31130 }, { "epoch": 3.12, "learning_rate": 6.281832236593751e-06, "loss": 0.0103, "step": 31131 }, { "epoch": 3.12, "learning_rate": 6.280366913141115e-06, "loss": 0.2825, "step": 31132 }, { "epoch": 3.12, "learning_rate": 6.2789017406311896e-06, "loss": 0.0723, "step": 31133 }, { "epoch": 3.12, "learning_rate": 6.277436719073303e-06, "loss": 0.4169, "step": 31134 }, { "epoch": 3.12, "learning_rate": 6.275971848476792e-06, "loss": 0.1866, "step": 31135 }, { "epoch": 3.12, "learning_rate": 6.274507128850978e-06, "loss": 0.4656, "step": 31136 }, { "epoch": 3.12, "learning_rate": 6.2730425602051625e-06, "loss": 0.1079, "step": 31137 }, { "epoch": 3.12, "learning_rate": 6.271578142548684e-06, "loss": 0.2698, "step": 31138 }, { "epoch": 3.12, "learning_rate": 6.2701138758908524e-06, "loss": 0.1043, "step": 31139 }, { "epoch": 3.13, "learning_rate": 6.268649760240994e-06, "loss": 0.1839, "step": 31140 }, { "epoch": 3.13, "learning_rate": 6.267185795608418e-06, "loss": 0.2605, "step": 31141 }, { "epoch": 3.13, "learning_rate": 6.265721982002442e-06, "loss": 0.4957, "step": 31142 }, { "epoch": 3.13, "learning_rate": 6.2642583194323845e-06, "loss": 0.1703, "step": 31143 }, { "epoch": 3.13, "learning_rate": 6.262794807907556e-06, "loss": 0.5174, "step": 31144 }, { "epoch": 3.13, "learning_rate": 6.261331447437266e-06, "loss": 0.129, "step": 31145 }, { "epoch": 3.13, "learning_rate": 6.259868238030845e-06, "loss": 0.2296, "step": 31146 }, { "epoch": 3.13, "learning_rate": 6.258405179697584e-06, "loss": 0.3033, "step": 31147 }, { "epoch": 3.13, "learning_rate": 6.256942272446793e-06, "loss": 0.3731, "step": 31148 }, { "epoch": 3.13, "learning_rate": 6.2554795162877975e-06, "loss": 0.2169, "step": 31149 }, { "epoch": 3.13, "learning_rate": 6.254016911229904e-06, "loss": 0.4464, "step": 31150 }, { "epoch": 3.13, "learning_rate": 6.252554457282398e-06, "loss": 0.2646, "step": 31151 }, { "epoch": 3.13, "learning_rate": 6.251092154454612e-06, "loss": 0.0689, "step": 31152 }, { "epoch": 3.13, "learning_rate": 6.2496300027558385e-06, "loss": 0.2719, "step": 31153 }, { "epoch": 3.13, "learning_rate": 6.248168002195384e-06, "loss": 0.5584, "step": 31154 }, { "epoch": 3.13, "learning_rate": 6.246706152782553e-06, "loss": 0.1945, "step": 31155 }, { "epoch": 3.13, "learning_rate": 6.24524445452665e-06, "loss": 0.0464, "step": 31156 }, { "epoch": 3.13, "learning_rate": 6.243782907436973e-06, "loss": 0.1244, "step": 31157 }, { "epoch": 3.13, "learning_rate": 6.242321511522824e-06, "loss": 0.2315, "step": 31158 }, { "epoch": 3.13, "learning_rate": 6.240860266793503e-06, "loss": 0.1173, "step": 31159 }, { "epoch": 3.13, "learning_rate": 6.23939917325831e-06, "loss": 0.1249, "step": 31160 }, { "epoch": 3.13, "learning_rate": 6.23793823092654e-06, "loss": 0.1044, "step": 31161 }, { "epoch": 3.13, "learning_rate": 6.236477439807492e-06, "loss": 0.1011, "step": 31162 }, { "epoch": 3.13, "learning_rate": 6.2350167999104555e-06, "loss": 0.1705, "step": 31163 }, { "epoch": 3.13, "learning_rate": 6.233556311244744e-06, "loss": 0.1398, "step": 31164 }, { "epoch": 3.13, "learning_rate": 6.2320959738196345e-06, "loss": 0.2592, "step": 31165 }, { "epoch": 3.13, "learning_rate": 6.230635787644412e-06, "loss": 0.0574, "step": 31166 }, { "epoch": 3.13, "learning_rate": 6.229175752728395e-06, "loss": 0.603, "step": 31167 }, { "epoch": 3.13, "learning_rate": 6.227715869080864e-06, "loss": 0.4798, "step": 31168 }, { "epoch": 3.13, "learning_rate": 6.226256136711092e-06, "loss": 0.1384, "step": 31169 }, { "epoch": 3.13, "learning_rate": 6.224796555628388e-06, "loss": 0.1937, "step": 31170 }, { "epoch": 3.13, "learning_rate": 6.223337125842034e-06, "loss": 0.3087, "step": 31171 }, { "epoch": 3.13, "learning_rate": 6.22187784736132e-06, "loss": 0.2396, "step": 31172 }, { "epoch": 3.13, "learning_rate": 6.220418720195527e-06, "loss": 0.1094, "step": 31173 }, { "epoch": 3.13, "learning_rate": 6.218959744353938e-06, "loss": 0.316, "step": 31174 }, { "epoch": 3.13, "learning_rate": 6.2175009198458545e-06, "loss": 0.1106, "step": 31175 }, { "epoch": 3.13, "learning_rate": 6.216042246680542e-06, "loss": 0.0931, "step": 31176 }, { "epoch": 3.13, "learning_rate": 6.214583724867281e-06, "loss": 0.0702, "step": 31177 }, { "epoch": 3.13, "learning_rate": 6.213125354415377e-06, "loss": 0.2737, "step": 31178 }, { "epoch": 3.13, "learning_rate": 6.211667135334085e-06, "loss": 0.1319, "step": 31179 }, { "epoch": 3.13, "learning_rate": 6.210209067632688e-06, "loss": 0.2211, "step": 31180 }, { "epoch": 3.13, "learning_rate": 6.208751151320477e-06, "loss": 0.122, "step": 31181 }, { "epoch": 3.13, "learning_rate": 6.207293386406732e-06, "loss": 0.1936, "step": 31182 }, { "epoch": 3.13, "learning_rate": 6.205835772900703e-06, "loss": 0.0169, "step": 31183 }, { "epoch": 3.13, "learning_rate": 6.20437831081169e-06, "loss": 0.2706, "step": 31184 }, { "epoch": 3.13, "learning_rate": 6.2029210001489654e-06, "loss": 0.1098, "step": 31185 }, { "epoch": 3.13, "learning_rate": 6.2014638409217946e-06, "loss": 0.3655, "step": 31186 }, { "epoch": 3.13, "learning_rate": 6.200006833139458e-06, "loss": 0.1012, "step": 31187 }, { "epoch": 3.13, "learning_rate": 6.198549976811223e-06, "loss": 0.0992, "step": 31188 }, { "epoch": 3.13, "learning_rate": 6.1970932719463605e-06, "loss": 0.1458, "step": 31189 }, { "epoch": 3.13, "learning_rate": 6.195636718554142e-06, "loss": 0.1216, "step": 31190 }, { "epoch": 3.13, "learning_rate": 6.194180316643834e-06, "loss": 0.1342, "step": 31191 }, { "epoch": 3.13, "learning_rate": 6.192724066224708e-06, "loss": 0.1525, "step": 31192 }, { "epoch": 3.13, "learning_rate": 6.1912679673060295e-06, "loss": 0.3274, "step": 31193 }, { "epoch": 3.13, "learning_rate": 6.189812019897064e-06, "loss": 0.1363, "step": 31194 }, { "epoch": 3.13, "learning_rate": 6.188356224007067e-06, "loss": 0.0924, "step": 31195 }, { "epoch": 3.13, "learning_rate": 6.1869005796453295e-06, "loss": 0.1302, "step": 31196 }, { "epoch": 3.13, "learning_rate": 6.185445086821091e-06, "loss": 0.4958, "step": 31197 }, { "epoch": 3.13, "learning_rate": 6.183989745543611e-06, "loss": 0.1875, "step": 31198 }, { "epoch": 3.13, "learning_rate": 6.182534555822172e-06, "loss": 0.2617, "step": 31199 }, { "epoch": 3.13, "learning_rate": 6.181079517666017e-06, "loss": 0.099, "step": 31200 }, { "epoch": 3.13, "learning_rate": 6.1796246310844155e-06, "loss": 0.2605, "step": 31201 }, { "epoch": 3.13, "learning_rate": 6.178169896086618e-06, "loss": 0.0868, "step": 31202 }, { "epoch": 3.13, "learning_rate": 6.176715312681886e-06, "loss": 0.2353, "step": 31203 }, { "epoch": 3.13, "learning_rate": 6.1752608808794755e-06, "loss": 0.1092, "step": 31204 }, { "epoch": 3.13, "learning_rate": 6.173806600688641e-06, "loss": 0.1638, "step": 31205 }, { "epoch": 3.13, "learning_rate": 6.172352472118629e-06, "loss": 0.8593, "step": 31206 }, { "epoch": 3.13, "learning_rate": 6.170898495178719e-06, "loss": 0.1588, "step": 31207 }, { "epoch": 3.13, "learning_rate": 6.169444669878136e-06, "loss": 0.0187, "step": 31208 }, { "epoch": 3.13, "learning_rate": 6.167990996226134e-06, "loss": 0.151, "step": 31209 }, { "epoch": 3.13, "learning_rate": 6.166537474231985e-06, "loss": 0.3407, "step": 31210 }, { "epoch": 3.13, "learning_rate": 6.165084103904918e-06, "loss": 0.1096, "step": 31211 }, { "epoch": 3.13, "learning_rate": 6.163630885254183e-06, "loss": 0.4502, "step": 31212 }, { "epoch": 3.13, "learning_rate": 6.162177818289038e-06, "loss": 0.1582, "step": 31213 }, { "epoch": 3.13, "learning_rate": 6.160724903018736e-06, "loss": 0.2791, "step": 31214 }, { "epoch": 3.13, "learning_rate": 6.159272139452495e-06, "loss": 0.3194, "step": 31215 }, { "epoch": 3.13, "learning_rate": 6.157819527599584e-06, "loss": 0.0831, "step": 31216 }, { "epoch": 3.13, "learning_rate": 6.156367067469239e-06, "loss": 0.0502, "step": 31217 }, { "epoch": 3.13, "learning_rate": 6.1549147590707035e-06, "loss": 0.2163, "step": 31218 }, { "epoch": 3.13, "learning_rate": 6.153462602413221e-06, "loss": 0.2082, "step": 31219 }, { "epoch": 3.13, "learning_rate": 6.1520105975060284e-06, "loss": 0.1258, "step": 31220 }, { "epoch": 3.13, "learning_rate": 6.15055874435837e-06, "loss": 0.2507, "step": 31221 }, { "epoch": 3.13, "learning_rate": 6.149107042979483e-06, "loss": 0.0866, "step": 31222 }, { "epoch": 3.13, "learning_rate": 6.147655493378602e-06, "loss": 0.133, "step": 31223 }, { "epoch": 3.13, "learning_rate": 6.146204095564971e-06, "loss": 0.4189, "step": 31224 }, { "epoch": 3.13, "learning_rate": 6.144752849547822e-06, "loss": 0.0182, "step": 31225 }, { "epoch": 3.13, "learning_rate": 6.143301755336381e-06, "loss": 0.854, "step": 31226 }, { "epoch": 3.13, "learning_rate": 6.1418508129399045e-06, "loss": 0.5343, "step": 31227 }, { "epoch": 3.13, "learning_rate": 6.140400022367617e-06, "loss": 0.0883, "step": 31228 }, { "epoch": 3.13, "learning_rate": 6.138949383628732e-06, "loss": 0.4845, "step": 31229 }, { "epoch": 3.13, "learning_rate": 6.137498896732506e-06, "loss": 0.0956, "step": 31230 }, { "epoch": 3.13, "learning_rate": 6.136048561688158e-06, "loss": 0.7402, "step": 31231 }, { "epoch": 3.13, "learning_rate": 6.13459837850492e-06, "loss": 0.01, "step": 31232 }, { "epoch": 3.13, "learning_rate": 6.133148347192015e-06, "loss": 0.176, "step": 31233 }, { "epoch": 3.13, "learning_rate": 6.131698467758679e-06, "loss": 0.0584, "step": 31234 }, { "epoch": 3.13, "learning_rate": 6.13024874021413e-06, "loss": 0.1674, "step": 31235 }, { "epoch": 3.13, "learning_rate": 6.1287991645676e-06, "loss": 0.0826, "step": 31236 }, { "epoch": 3.13, "learning_rate": 6.127349740828313e-06, "loss": 0.0364, "step": 31237 }, { "epoch": 3.14, "learning_rate": 6.125900469005479e-06, "loss": 0.2433, "step": 31238 }, { "epoch": 3.14, "learning_rate": 6.12445134910835e-06, "loss": 0.0477, "step": 31239 }, { "epoch": 3.14, "learning_rate": 6.123002381146122e-06, "loss": 0.1761, "step": 31240 }, { "epoch": 3.14, "learning_rate": 6.1215535651280155e-06, "loss": 0.2886, "step": 31241 }, { "epoch": 3.14, "learning_rate": 6.120104901063272e-06, "loss": 0.1906, "step": 31242 }, { "epoch": 3.14, "learning_rate": 6.11865638896109e-06, "loss": 0.2545, "step": 31243 }, { "epoch": 3.14, "learning_rate": 6.117208028830687e-06, "loss": 0.1612, "step": 31244 }, { "epoch": 3.14, "learning_rate": 6.115759820681293e-06, "loss": 0.401, "step": 31245 }, { "epoch": 3.14, "learning_rate": 6.114311764522128e-06, "loss": 0.2813, "step": 31246 }, { "epoch": 3.14, "learning_rate": 6.112863860362377e-06, "loss": 0.2337, "step": 31247 }, { "epoch": 3.14, "learning_rate": 6.111416108211284e-06, "loss": 0.6654, "step": 31248 }, { "epoch": 3.14, "learning_rate": 6.1099685080780495e-06, "loss": 0.168, "step": 31249 }, { "epoch": 3.14, "learning_rate": 6.108521059971887e-06, "loss": 0.395, "step": 31250 }, { "epoch": 3.14, "learning_rate": 6.107073763902007e-06, "loss": 0.1698, "step": 31251 }, { "epoch": 3.14, "learning_rate": 6.105626619877623e-06, "loss": 0.2185, "step": 31252 }, { "epoch": 3.14, "learning_rate": 6.104179627907939e-06, "loss": 0.3235, "step": 31253 }, { "epoch": 3.14, "learning_rate": 6.102732788002169e-06, "loss": 0.3206, "step": 31254 }, { "epoch": 3.14, "learning_rate": 6.1012861001695035e-06, "loss": 0.2422, "step": 31255 }, { "epoch": 3.14, "learning_rate": 6.099839564419181e-06, "loss": 0.295, "step": 31256 }, { "epoch": 3.14, "learning_rate": 6.098393180760377e-06, "loss": 0.1645, "step": 31257 }, { "epoch": 3.14, "learning_rate": 6.0969469492023e-06, "loss": 0.3478, "step": 31258 }, { "epoch": 3.14, "learning_rate": 6.095500869754167e-06, "loss": 0.1026, "step": 31259 }, { "epoch": 3.14, "learning_rate": 6.094054942425181e-06, "loss": 0.2439, "step": 31260 }, { "epoch": 3.14, "learning_rate": 6.092609167224519e-06, "loss": 0.1895, "step": 31261 }, { "epoch": 3.14, "learning_rate": 6.091163544161406e-06, "loss": 0.0572, "step": 31262 }, { "epoch": 3.14, "learning_rate": 6.089718073245034e-06, "loss": 0.159, "step": 31263 }, { "epoch": 3.14, "learning_rate": 6.0882727544845976e-06, "loss": 0.3321, "step": 31264 }, { "epoch": 3.14, "learning_rate": 6.0868275878892975e-06, "loss": 0.1589, "step": 31265 }, { "epoch": 3.14, "learning_rate": 6.085382573468331e-06, "loss": 0.0448, "step": 31266 }, { "epoch": 3.14, "learning_rate": 6.083937711230891e-06, "loss": 0.2409, "step": 31267 }, { "epoch": 3.14, "learning_rate": 6.082493001186171e-06, "loss": 0.1716, "step": 31268 }, { "epoch": 3.14, "learning_rate": 6.08104844334337e-06, "loss": 0.2166, "step": 31269 }, { "epoch": 3.14, "learning_rate": 6.079604037711669e-06, "loss": 0.0548, "step": 31270 }, { "epoch": 3.14, "learning_rate": 6.0781597843002845e-06, "loss": 0.0884, "step": 31271 }, { "epoch": 3.14, "learning_rate": 6.076715683118382e-06, "loss": 0.2648, "step": 31272 }, { "epoch": 3.14, "learning_rate": 6.0752717341751514e-06, "loss": 0.0559, "step": 31273 }, { "epoch": 3.14, "learning_rate": 6.073827937479805e-06, "loss": 0.1302, "step": 31274 }, { "epoch": 3.14, "learning_rate": 6.07238429304151e-06, "loss": 0.1867, "step": 31275 }, { "epoch": 3.14, "learning_rate": 6.07094080086945e-06, "loss": 0.0797, "step": 31276 }, { "epoch": 3.14, "learning_rate": 6.069497460972829e-06, "loss": 0.4194, "step": 31277 }, { "epoch": 3.14, "learning_rate": 6.06805427336083e-06, "loss": 0.7896, "step": 31278 }, { "epoch": 3.14, "learning_rate": 6.066611238042612e-06, "loss": 0.3695, "step": 31279 }, { "epoch": 3.14, "learning_rate": 6.065168355027384e-06, "loss": 0.3801, "step": 31280 }, { "epoch": 3.14, "learning_rate": 6.063725624324319e-06, "loss": 0.2059, "step": 31281 }, { "epoch": 3.14, "learning_rate": 6.062283045942599e-06, "loss": 0.5611, "step": 31282 }, { "epoch": 3.14, "learning_rate": 6.0608406198914045e-06, "loss": 0.1934, "step": 31283 }, { "epoch": 3.14, "learning_rate": 6.059398346179904e-06, "loss": 0.2046, "step": 31284 }, { "epoch": 3.14, "learning_rate": 6.0579562248173e-06, "loss": 0.7572, "step": 31285 }, { "epoch": 3.14, "learning_rate": 6.056514255812746e-06, "loss": 0.4047, "step": 31286 }, { "epoch": 3.14, "learning_rate": 6.055072439175421e-06, "loss": 0.0869, "step": 31287 }, { "epoch": 3.14, "learning_rate": 6.05363077491452e-06, "loss": 0.1207, "step": 31288 }, { "epoch": 3.14, "learning_rate": 6.052189263039197e-06, "loss": 0.109, "step": 31289 }, { "epoch": 3.14, "learning_rate": 6.050747903558624e-06, "loss": 0.026, "step": 31290 }, { "epoch": 3.14, "learning_rate": 6.0493066964819876e-06, "loss": 0.0959, "step": 31291 }, { "epoch": 3.14, "learning_rate": 6.0478656418184595e-06, "loss": 0.0686, "step": 31292 }, { "epoch": 3.14, "learning_rate": 6.046424739577186e-06, "loss": 0.2396, "step": 31293 }, { "epoch": 3.14, "learning_rate": 6.044983989767361e-06, "loss": 0.4092, "step": 31294 }, { "epoch": 3.14, "learning_rate": 6.043543392398145e-06, "loss": 0.2697, "step": 31295 }, { "epoch": 3.14, "learning_rate": 6.042102947478706e-06, "loss": 0.1632, "step": 31296 }, { "epoch": 3.14, "learning_rate": 6.040662655018209e-06, "loss": 0.1866, "step": 31297 }, { "epoch": 3.14, "learning_rate": 6.039222515025823e-06, "loss": 0.1009, "step": 31298 }, { "epoch": 3.14, "learning_rate": 6.037782527510708e-06, "loss": 0.2551, "step": 31299 }, { "epoch": 3.14, "learning_rate": 6.0363426924820265e-06, "loss": 0.3826, "step": 31300 }, { "epoch": 3.14, "learning_rate": 6.034903009948948e-06, "loss": 0.2972, "step": 31301 }, { "epoch": 3.14, "learning_rate": 6.033463479920622e-06, "loss": 0.1294, "step": 31302 }, { "epoch": 3.14, "learning_rate": 6.03202410240623e-06, "loss": 0.1298, "step": 31303 }, { "epoch": 3.14, "learning_rate": 6.030584877414912e-06, "loss": 0.0422, "step": 31304 }, { "epoch": 3.14, "learning_rate": 6.029145804955824e-06, "loss": 0.2412, "step": 31305 }, { "epoch": 3.14, "learning_rate": 6.027706885038153e-06, "loss": 0.0566, "step": 31306 }, { "epoch": 3.14, "learning_rate": 6.026268117671024e-06, "loss": 0.3338, "step": 31307 }, { "epoch": 3.14, "learning_rate": 6.0248295028636e-06, "loss": 0.0705, "step": 31308 }, { "epoch": 3.14, "learning_rate": 6.023391040625047e-06, "loss": 0.2928, "step": 31309 }, { "epoch": 3.14, "learning_rate": 6.021952730964511e-06, "loss": 0.0871, "step": 31310 }, { "epoch": 3.14, "learning_rate": 6.020514573891147e-06, "loss": 0.2665, "step": 31311 }, { "epoch": 3.14, "learning_rate": 6.019076569414106e-06, "loss": 0.2752, "step": 31312 }, { "epoch": 3.14, "learning_rate": 6.017638717542539e-06, "loss": 0.3644, "step": 31313 }, { "epoch": 3.14, "learning_rate": 6.016201018285594e-06, "loss": 0.0691, "step": 31314 }, { "epoch": 3.14, "learning_rate": 6.014763471652426e-06, "loss": 0.4996, "step": 31315 }, { "epoch": 3.14, "learning_rate": 6.013326077652167e-06, "loss": 0.0626, "step": 31316 }, { "epoch": 3.14, "learning_rate": 6.011888836293994e-06, "loss": 0.1934, "step": 31317 }, { "epoch": 3.14, "learning_rate": 6.010451747587026e-06, "loss": 0.0095, "step": 31318 }, { "epoch": 3.14, "learning_rate": 6.009014811540409e-06, "loss": 0.1452, "step": 31319 }, { "epoch": 3.14, "learning_rate": 6.0075780281633116e-06, "loss": 0.1305, "step": 31320 }, { "epoch": 3.14, "learning_rate": 6.006141397464852e-06, "loss": 0.3252, "step": 31321 }, { "epoch": 3.14, "learning_rate": 6.004704919454171e-06, "loss": 0.1548, "step": 31322 }, { "epoch": 3.14, "learning_rate": 6.00326859414043e-06, "loss": 0.469, "step": 31323 }, { "epoch": 3.14, "learning_rate": 6.0018324215327656e-06, "loss": 0.0441, "step": 31324 }, { "epoch": 3.14, "learning_rate": 6.000396401640294e-06, "loss": 0.2507, "step": 31325 }, { "epoch": 3.14, "learning_rate": 5.99896053447218e-06, "loss": 0.5284, "step": 31326 }, { "epoch": 3.14, "learning_rate": 5.997524820037551e-06, "loss": 0.1094, "step": 31327 }, { "epoch": 3.14, "learning_rate": 5.996089258345541e-06, "loss": 0.2497, "step": 31328 }, { "epoch": 3.14, "learning_rate": 5.9946538494052885e-06, "loss": 0.1394, "step": 31329 }, { "epoch": 3.14, "learning_rate": 5.993218593225926e-06, "loss": 0.2605, "step": 31330 }, { "epoch": 3.14, "learning_rate": 5.991783489816588e-06, "loss": 0.0154, "step": 31331 }, { "epoch": 3.14, "learning_rate": 5.990348539186406e-06, "loss": 0.0615, "step": 31332 }, { "epoch": 3.14, "learning_rate": 5.988913741344511e-06, "loss": 0.2486, "step": 31333 }, { "epoch": 3.14, "learning_rate": 5.9874790963000295e-06, "loss": 0.2798, "step": 31334 }, { "epoch": 3.14, "learning_rate": 5.986044604062111e-06, "loss": 0.4191, "step": 31335 }, { "epoch": 3.15, "learning_rate": 5.984610264639856e-06, "loss": 0.6886, "step": 31336 }, { "epoch": 3.15, "learning_rate": 5.9831760780424105e-06, "loss": 0.2116, "step": 31337 }, { "epoch": 3.15, "learning_rate": 5.981742044278904e-06, "loss": 0.3292, "step": 31338 }, { "epoch": 3.15, "learning_rate": 5.980308163358439e-06, "loss": 0.1037, "step": 31339 }, { "epoch": 3.15, "learning_rate": 5.9788744352901615e-06, "loss": 0.154, "step": 31340 }, { "epoch": 3.15, "learning_rate": 5.977440860083191e-06, "loss": 0.1323, "step": 31341 }, { "epoch": 3.15, "learning_rate": 5.976007437746647e-06, "loss": 0.2557, "step": 31342 }, { "epoch": 3.15, "learning_rate": 5.974574168289651e-06, "loss": 0.3108, "step": 31343 }, { "epoch": 3.15, "learning_rate": 5.973141051721328e-06, "loss": 0.3688, "step": 31344 }, { "epoch": 3.15, "learning_rate": 5.971708088050795e-06, "loss": 0.2349, "step": 31345 }, { "epoch": 3.15, "learning_rate": 5.970275277287171e-06, "loss": 0.1193, "step": 31346 }, { "epoch": 3.15, "learning_rate": 5.968842619439572e-06, "loss": 0.118, "step": 31347 }, { "epoch": 3.15, "learning_rate": 5.967410114517109e-06, "loss": 0.0866, "step": 31348 }, { "epoch": 3.15, "learning_rate": 5.965977762528921e-06, "loss": 0.2351, "step": 31349 }, { "epoch": 3.15, "learning_rate": 5.964545563484102e-06, "loss": 0.3227, "step": 31350 }, { "epoch": 3.15, "learning_rate": 5.963113517391762e-06, "loss": 0.0915, "step": 31351 }, { "epoch": 3.15, "learning_rate": 5.961681624261037e-06, "loss": 0.3644, "step": 31352 }, { "epoch": 3.15, "learning_rate": 5.960249884101021e-06, "loss": 0.5646, "step": 31353 }, { "epoch": 3.15, "learning_rate": 5.9588182969208204e-06, "loss": 0.0957, "step": 31354 }, { "epoch": 3.15, "learning_rate": 5.957386862729565e-06, "loss": 0.1038, "step": 31355 }, { "epoch": 3.15, "learning_rate": 5.9559555815363564e-06, "loss": 0.1111, "step": 31356 }, { "epoch": 3.15, "learning_rate": 5.954524453350289e-06, "loss": 0.419, "step": 31357 }, { "epoch": 3.15, "learning_rate": 5.953093478180483e-06, "loss": 0.1244, "step": 31358 }, { "epoch": 3.15, "learning_rate": 5.951662656036044e-06, "loss": 0.2644, "step": 31359 }, { "epoch": 3.15, "learning_rate": 5.950231986926077e-06, "loss": 0.4205, "step": 31360 }, { "epoch": 3.15, "learning_rate": 5.9488014708596835e-06, "loss": 0.0825, "step": 31361 }, { "epoch": 3.15, "learning_rate": 5.947371107845962e-06, "loss": 0.4067, "step": 31362 }, { "epoch": 3.15, "learning_rate": 5.945940897894035e-06, "loss": 0.0437, "step": 31363 }, { "epoch": 3.15, "learning_rate": 5.944510841012982e-06, "loss": 0.0372, "step": 31364 }, { "epoch": 3.15, "learning_rate": 5.943080937211901e-06, "loss": 0.108, "step": 31365 }, { "epoch": 3.15, "learning_rate": 5.941651186499911e-06, "loss": 0.2057, "step": 31366 }, { "epoch": 3.15, "learning_rate": 5.9402215888861085e-06, "loss": 0.4126, "step": 31367 }, { "epoch": 3.15, "learning_rate": 5.9387921443795666e-06, "loss": 0.0667, "step": 31368 }, { "epoch": 3.15, "learning_rate": 5.937362852989408e-06, "loss": 0.4866, "step": 31369 }, { "epoch": 3.15, "learning_rate": 5.935933714724724e-06, "loss": 0.4968, "step": 31370 }, { "epoch": 3.15, "learning_rate": 5.9345047295945885e-06, "loss": 0.3077, "step": 31371 }, { "epoch": 3.15, "learning_rate": 5.933075897608115e-06, "loss": 0.2663, "step": 31372 }, { "epoch": 3.15, "learning_rate": 5.931647218774394e-06, "loss": 0.226, "step": 31373 }, { "epoch": 3.15, "learning_rate": 5.930218693102513e-06, "loss": 0.2161, "step": 31374 }, { "epoch": 3.15, "learning_rate": 5.928790320601566e-06, "loss": 0.4547, "step": 31375 }, { "epoch": 3.15, "learning_rate": 5.927362101280637e-06, "loss": 0.1033, "step": 31376 }, { "epoch": 3.15, "learning_rate": 5.925934035148819e-06, "loss": 0.1314, "step": 31377 }, { "epoch": 3.15, "learning_rate": 5.924506122215199e-06, "loss": 0.2895, "step": 31378 }, { "epoch": 3.15, "learning_rate": 5.923078362488862e-06, "loss": 0.1329, "step": 31379 }, { "epoch": 3.15, "learning_rate": 5.921650755978888e-06, "loss": 0.1343, "step": 31380 }, { "epoch": 3.15, "learning_rate": 5.920223302694386e-06, "loss": 0.128, "step": 31381 }, { "epoch": 3.15, "learning_rate": 5.918796002644413e-06, "loss": 0.1018, "step": 31382 }, { "epoch": 3.15, "learning_rate": 5.9173688558380515e-06, "loss": 0.188, "step": 31383 }, { "epoch": 3.15, "learning_rate": 5.915941862284413e-06, "loss": 0.134, "step": 31384 }, { "epoch": 3.15, "learning_rate": 5.914515021992548e-06, "loss": 0.2882, "step": 31385 }, { "epoch": 3.15, "learning_rate": 5.91308833497154e-06, "loss": 0.0316, "step": 31386 }, { "epoch": 3.15, "learning_rate": 5.911661801230479e-06, "loss": 0.0974, "step": 31387 }, { "epoch": 3.15, "learning_rate": 5.91023542077845e-06, "loss": 0.2536, "step": 31388 }, { "epoch": 3.15, "learning_rate": 5.908809193624503e-06, "loss": 0.1374, "step": 31389 }, { "epoch": 3.15, "learning_rate": 5.907383119777739e-06, "loss": 0.0669, "step": 31390 }, { "epoch": 3.15, "learning_rate": 5.90595719924722e-06, "loss": 0.1357, "step": 31391 }, { "epoch": 3.15, "learning_rate": 5.904531432042025e-06, "loss": 0.0182, "step": 31392 }, { "epoch": 3.15, "learning_rate": 5.903105818171225e-06, "loss": 0.7431, "step": 31393 }, { "epoch": 3.15, "learning_rate": 5.901680357643883e-06, "loss": 0.2764, "step": 31394 }, { "epoch": 3.15, "learning_rate": 5.900255050469098e-06, "loss": 0.4345, "step": 31395 }, { "epoch": 3.15, "learning_rate": 5.898829896655913e-06, "loss": 0.0708, "step": 31396 }, { "epoch": 3.15, "learning_rate": 5.897404896213401e-06, "loss": 0.1902, "step": 31397 }, { "epoch": 3.15, "learning_rate": 5.895980049150638e-06, "loss": 0.0629, "step": 31398 }, { "epoch": 3.15, "learning_rate": 5.894555355476699e-06, "loss": 0.0704, "step": 31399 }, { "epoch": 3.15, "learning_rate": 5.893130815200626e-06, "loss": 0.0476, "step": 31400 }, { "epoch": 3.15, "learning_rate": 5.891706428331502e-06, "loss": 0.0603, "step": 31401 }, { "epoch": 3.15, "learning_rate": 5.890282194878396e-06, "loss": 0.2978, "step": 31402 }, { "epoch": 3.15, "learning_rate": 5.888858114850348e-06, "loss": 0.0823, "step": 31403 }, { "epoch": 3.15, "learning_rate": 5.88743418825644e-06, "loss": 0.2389, "step": 31404 }, { "epoch": 3.15, "learning_rate": 5.8860104151057294e-06, "loss": 0.1147, "step": 31405 }, { "epoch": 3.15, "learning_rate": 5.884586795407276e-06, "loss": 0.0152, "step": 31406 }, { "epoch": 3.15, "learning_rate": 5.883163329170137e-06, "loss": 0.5496, "step": 31407 }, { "epoch": 3.15, "learning_rate": 5.881740016403371e-06, "loss": 0.2282, "step": 31408 }, { "epoch": 3.15, "learning_rate": 5.8803168571160375e-06, "loss": 0.3545, "step": 31409 }, { "epoch": 3.15, "learning_rate": 5.878893851317192e-06, "loss": 0.2312, "step": 31410 }, { "epoch": 3.15, "learning_rate": 5.877470999015892e-06, "loss": 0.515, "step": 31411 }, { "epoch": 3.15, "learning_rate": 5.8760483002211785e-06, "loss": 0.099, "step": 31412 }, { "epoch": 3.15, "learning_rate": 5.874625754942132e-06, "loss": 0.075, "step": 31413 }, { "epoch": 3.15, "learning_rate": 5.873203363187786e-06, "loss": 0.1711, "step": 31414 }, { "epoch": 3.15, "learning_rate": 5.8717811249671845e-06, "loss": 0.1065, "step": 31415 }, { "epoch": 3.15, "learning_rate": 5.870359040289403e-06, "loss": 0.1352, "step": 31416 }, { "epoch": 3.15, "learning_rate": 5.868937109163464e-06, "loss": 0.2629, "step": 31417 }, { "epoch": 3.15, "learning_rate": 5.867515331598437e-06, "loss": 0.1964, "step": 31418 }, { "epoch": 3.15, "learning_rate": 5.866093707603361e-06, "loss": 0.5769, "step": 31419 }, { "epoch": 3.15, "learning_rate": 5.8646722371872865e-06, "loss": 0.133, "step": 31420 }, { "epoch": 3.15, "learning_rate": 5.863250920359255e-06, "loss": 0.1227, "step": 31421 }, { "epoch": 3.15, "learning_rate": 5.861829757128311e-06, "loss": 0.0803, "step": 31422 }, { "epoch": 3.15, "learning_rate": 5.860408747503499e-06, "loss": 0.29, "step": 31423 }, { "epoch": 3.15, "learning_rate": 5.858987891493863e-06, "loss": 0.1048, "step": 31424 }, { "epoch": 3.15, "learning_rate": 5.857567189108448e-06, "loss": 0.0551, "step": 31425 }, { "epoch": 3.15, "learning_rate": 5.856146640356279e-06, "loss": 0.2654, "step": 31426 }, { "epoch": 3.15, "learning_rate": 5.854726245246426e-06, "loss": 0.1507, "step": 31427 }, { "epoch": 3.15, "learning_rate": 5.8533060037879e-06, "loss": 0.0511, "step": 31428 }, { "epoch": 3.15, "learning_rate": 5.851885915989743e-06, "loss": 0.089, "step": 31429 }, { "epoch": 3.15, "learning_rate": 5.850465981861006e-06, "loss": 0.1486, "step": 31430 }, { "epoch": 3.15, "learning_rate": 5.8490462014107225e-06, "loss": 0.1278, "step": 31431 }, { "epoch": 3.15, "learning_rate": 5.8476265746479054e-06, "loss": 0.2114, "step": 31432 }, { "epoch": 3.15, "learning_rate": 5.846207101581616e-06, "loss": 0.2799, "step": 31433 }, { "epoch": 3.15, "learning_rate": 5.844787782220879e-06, "loss": 0.084, "step": 31434 }, { "epoch": 3.16, "learning_rate": 5.8433686165747105e-06, "loss": 0.2065, "step": 31435 }, { "epoch": 3.16, "learning_rate": 5.841949604652164e-06, "loss": 0.0822, "step": 31436 }, { "epoch": 3.16, "learning_rate": 5.840530746462256e-06, "loss": 0.2158, "step": 31437 }, { "epoch": 3.16, "learning_rate": 5.8391120420140235e-06, "loss": 0.3147, "step": 31438 }, { "epoch": 3.16, "learning_rate": 5.837693491316489e-06, "loss": 0.3169, "step": 31439 }, { "epoch": 3.16, "learning_rate": 5.8362750943786805e-06, "loss": 0.123, "step": 31440 }, { "epoch": 3.16, "learning_rate": 5.8348568512096255e-06, "loss": 0.1416, "step": 31441 }, { "epoch": 3.16, "learning_rate": 5.833438761818348e-06, "loss": 0.1467, "step": 31442 }, { "epoch": 3.16, "learning_rate": 5.8320208262138675e-06, "loss": 0.1596, "step": 31443 }, { "epoch": 3.16, "learning_rate": 5.830603044405219e-06, "loss": 0.2166, "step": 31444 }, { "epoch": 3.16, "learning_rate": 5.829185416401424e-06, "loss": 0.0369, "step": 31445 }, { "epoch": 3.16, "learning_rate": 5.827767942211482e-06, "loss": 0.4024, "step": 31446 }, { "epoch": 3.16, "learning_rate": 5.826350621844443e-06, "loss": 0.3704, "step": 31447 }, { "epoch": 3.16, "learning_rate": 5.824933455309313e-06, "loss": 0.1002, "step": 31448 }, { "epoch": 3.16, "learning_rate": 5.823516442615098e-06, "loss": 0.4518, "step": 31449 }, { "epoch": 3.16, "learning_rate": 5.822099583770834e-06, "loss": 0.2607, "step": 31450 }, { "epoch": 3.16, "learning_rate": 5.820682878785534e-06, "loss": 0.312, "step": 31451 }, { "epoch": 3.16, "learning_rate": 5.819266327668205e-06, "loss": 0.0967, "step": 31452 }, { "epoch": 3.16, "learning_rate": 5.8178499304278695e-06, "loss": 0.8281, "step": 31453 }, { "epoch": 3.16, "learning_rate": 5.816433687073536e-06, "loss": 0.0126, "step": 31454 }, { "epoch": 3.16, "learning_rate": 5.815017597614219e-06, "loss": 0.6009, "step": 31455 }, { "epoch": 3.16, "learning_rate": 5.813601662058927e-06, "loss": 0.7452, "step": 31456 }, { "epoch": 3.16, "learning_rate": 5.812185880416678e-06, "loss": 0.0965, "step": 31457 }, { "epoch": 3.16, "learning_rate": 5.810770252696465e-06, "loss": 0.2653, "step": 31458 }, { "epoch": 3.16, "learning_rate": 5.8093547789073255e-06, "loss": 0.2938, "step": 31459 }, { "epoch": 3.16, "learning_rate": 5.80793945905824e-06, "loss": 0.4488, "step": 31460 }, { "epoch": 3.16, "learning_rate": 5.806524293158218e-06, "loss": 0.2087, "step": 31461 }, { "epoch": 3.16, "learning_rate": 5.8051092812162795e-06, "loss": 0.1137, "step": 31462 }, { "epoch": 3.16, "learning_rate": 5.8036944232414304e-06, "loss": 0.2358, "step": 31463 }, { "epoch": 3.16, "learning_rate": 5.802279719242649e-06, "loss": 0.1048, "step": 31464 }, { "epoch": 3.16, "learning_rate": 5.80086516922896e-06, "loss": 0.0602, "step": 31465 }, { "epoch": 3.16, "learning_rate": 5.799450773209371e-06, "loss": 0.123, "step": 31466 }, { "epoch": 3.16, "learning_rate": 5.7980365311928515e-06, "loss": 0.3475, "step": 31467 }, { "epoch": 3.16, "learning_rate": 5.79662244318843e-06, "loss": 0.2416, "step": 31468 }, { "epoch": 3.16, "learning_rate": 5.795208509205093e-06, "loss": 0.0339, "step": 31469 }, { "epoch": 3.16, "learning_rate": 5.793794729251846e-06, "loss": 0.1802, "step": 31470 }, { "epoch": 3.16, "learning_rate": 5.792381103337679e-06, "loss": 0.096, "step": 31471 }, { "epoch": 3.16, "learning_rate": 5.790967631471581e-06, "loss": 0.0064, "step": 31472 }, { "epoch": 3.16, "learning_rate": 5.78955431366257e-06, "loss": 0.2063, "step": 31473 }, { "epoch": 3.16, "learning_rate": 5.788141149919618e-06, "loss": 0.2143, "step": 31474 }, { "epoch": 3.16, "learning_rate": 5.786728140251719e-06, "loss": 0.3842, "step": 31475 }, { "epoch": 3.16, "learning_rate": 5.785315284667879e-06, "loss": 0.5196, "step": 31476 }, { "epoch": 3.16, "learning_rate": 5.783902583177088e-06, "loss": 0.0869, "step": 31477 }, { "epoch": 3.16, "learning_rate": 5.78249003578831e-06, "loss": 0.1367, "step": 31478 }, { "epoch": 3.16, "learning_rate": 5.781077642510564e-06, "loss": 0.0603, "step": 31479 }, { "epoch": 3.16, "learning_rate": 5.779665403352831e-06, "loss": 0.0689, "step": 31480 }, { "epoch": 3.16, "learning_rate": 5.77825331832408e-06, "loss": 0.0486, "step": 31481 }, { "epoch": 3.16, "learning_rate": 5.776841387433319e-06, "loss": 0.2981, "step": 31482 }, { "epoch": 3.16, "learning_rate": 5.775429610689523e-06, "loss": 0.3338, "step": 31483 }, { "epoch": 3.16, "learning_rate": 5.774017988101678e-06, "loss": 0.1296, "step": 31484 }, { "epoch": 3.16, "learning_rate": 5.772606519678768e-06, "loss": 0.5647, "step": 31485 }, { "epoch": 3.16, "learning_rate": 5.771195205429771e-06, "loss": 0.1807, "step": 31486 }, { "epoch": 3.16, "learning_rate": 5.769784045363674e-06, "loss": 0.3601, "step": 31487 }, { "epoch": 3.16, "learning_rate": 5.768373039489449e-06, "loss": 0.1274, "step": 31488 }, { "epoch": 3.16, "learning_rate": 5.766962187816084e-06, "loss": 0.0828, "step": 31489 }, { "epoch": 3.16, "learning_rate": 5.765551490352543e-06, "loss": 0.0857, "step": 31490 }, { "epoch": 3.16, "learning_rate": 5.76414094710783e-06, "loss": 0.2704, "step": 31491 }, { "epoch": 3.16, "learning_rate": 5.762730558090894e-06, "loss": 0.5434, "step": 31492 }, { "epoch": 3.16, "learning_rate": 5.761320323310715e-06, "loss": 0.2802, "step": 31493 }, { "epoch": 3.16, "learning_rate": 5.759910242776279e-06, "loss": 0.1834, "step": 31494 }, { "epoch": 3.16, "learning_rate": 5.758500316496561e-06, "loss": 0.3116, "step": 31495 }, { "epoch": 3.16, "learning_rate": 5.757090544480509e-06, "loss": 0.0731, "step": 31496 }, { "epoch": 3.16, "learning_rate": 5.755680926737118e-06, "loss": 0.1882, "step": 31497 }, { "epoch": 3.16, "learning_rate": 5.754271463275351e-06, "loss": 0.2357, "step": 31498 }, { "epoch": 3.16, "learning_rate": 5.7528621541041784e-06, "loss": 0.6264, "step": 31499 }, { "epoch": 3.16, "learning_rate": 5.751452999232565e-06, "loss": 0.3893, "step": 31500 }, { "epoch": 3.16, "learning_rate": 5.750043998669481e-06, "loss": 0.1501, "step": 31501 }, { "epoch": 3.16, "learning_rate": 5.7486351524238925e-06, "loss": 0.1229, "step": 31502 }, { "epoch": 3.16, "learning_rate": 5.747226460504762e-06, "loss": 0.2452, "step": 31503 }, { "epoch": 3.16, "learning_rate": 5.74581792292105e-06, "loss": 0.1943, "step": 31504 }, { "epoch": 3.16, "learning_rate": 5.744409539681741e-06, "loss": 0.27, "step": 31505 }, { "epoch": 3.16, "learning_rate": 5.743001310795775e-06, "loss": 0.8778, "step": 31506 }, { "epoch": 3.16, "learning_rate": 5.741593236272113e-06, "loss": 0.5545, "step": 31507 }, { "epoch": 3.16, "learning_rate": 5.74018531611973e-06, "loss": 0.0791, "step": 31508 }, { "epoch": 3.16, "learning_rate": 5.738777550347589e-06, "loss": 0.335, "step": 31509 }, { "epoch": 3.16, "learning_rate": 5.737369938964619e-06, "loss": 0.037, "step": 31510 }, { "epoch": 3.16, "learning_rate": 5.7359624819798074e-06, "loss": 0.0976, "step": 31511 }, { "epoch": 3.16, "learning_rate": 5.7345551794021054e-06, "loss": 0.0731, "step": 31512 }, { "epoch": 3.16, "learning_rate": 5.73314803124045e-06, "loss": 0.1506, "step": 31513 }, { "epoch": 3.16, "learning_rate": 5.7317410375038146e-06, "loss": 0.011, "step": 31514 }, { "epoch": 3.16, "learning_rate": 5.7303341982011514e-06, "loss": 0.0532, "step": 31515 }, { "epoch": 3.16, "learning_rate": 5.728927513341403e-06, "loss": 0.2917, "step": 31516 }, { "epoch": 3.16, "learning_rate": 5.727520982933529e-06, "loss": 0.4985, "step": 31517 }, { "epoch": 3.16, "learning_rate": 5.72611460698648e-06, "loss": 0.1114, "step": 31518 }, { "epoch": 3.16, "learning_rate": 5.7247083855092005e-06, "loss": 0.1494, "step": 31519 }, { "epoch": 3.16, "learning_rate": 5.723302318510645e-06, "loss": 0.2564, "step": 31520 }, { "epoch": 3.16, "learning_rate": 5.7218964059997545e-06, "loss": 0.1645, "step": 31521 }, { "epoch": 3.16, "learning_rate": 5.720490647985474e-06, "loss": 0.1641, "step": 31522 }, { "epoch": 3.16, "learning_rate": 5.719085044476771e-06, "loss": 0.135, "step": 31523 }, { "epoch": 3.16, "learning_rate": 5.717679595482556e-06, "loss": 0.4081, "step": 31524 }, { "epoch": 3.16, "learning_rate": 5.716274301011798e-06, "loss": 0.0634, "step": 31525 }, { "epoch": 3.16, "learning_rate": 5.714869161073434e-06, "loss": 0.1225, "step": 31526 }, { "epoch": 3.16, "learning_rate": 5.713464175676403e-06, "loss": 0.3642, "step": 31527 }, { "epoch": 3.16, "learning_rate": 5.712059344829647e-06, "loss": 0.1493, "step": 31528 }, { "epoch": 3.16, "learning_rate": 5.7106546685421076e-06, "loss": 0.3093, "step": 31529 }, { "epoch": 3.16, "learning_rate": 5.70925014682272e-06, "loss": 0.0562, "step": 31530 }, { "epoch": 3.16, "learning_rate": 5.707845779680427e-06, "loss": 0.1134, "step": 31531 }, { "epoch": 3.16, "learning_rate": 5.706441567124156e-06, "loss": 0.0271, "step": 31532 }, { "epoch": 3.17, "learning_rate": 5.705037509162854e-06, "loss": 0.2662, "step": 31533 }, { "epoch": 3.17, "learning_rate": 5.703633605805451e-06, "loss": 0.3248, "step": 31534 }, { "epoch": 3.17, "learning_rate": 5.70222985706088e-06, "loss": 0.1195, "step": 31535 }, { "epoch": 3.17, "learning_rate": 5.700826262938065e-06, "loss": 0.139, "step": 31536 }, { "epoch": 3.17, "learning_rate": 5.699422823445964e-06, "loss": 0.6223, "step": 31537 }, { "epoch": 3.17, "learning_rate": 5.698019538593489e-06, "loss": 0.2106, "step": 31538 }, { "epoch": 3.17, "learning_rate": 5.69661640838956e-06, "loss": 0.2467, "step": 31539 }, { "epoch": 3.17, "learning_rate": 5.695213432843125e-06, "loss": 0.2477, "step": 31540 }, { "epoch": 3.17, "learning_rate": 5.693810611963116e-06, "loss": 0.1937, "step": 31541 }, { "epoch": 3.17, "learning_rate": 5.692407945758437e-06, "loss": 0.1423, "step": 31542 }, { "epoch": 3.17, "learning_rate": 5.691005434238029e-06, "loss": 0.0615, "step": 31543 }, { "epoch": 3.17, "learning_rate": 5.689603077410827e-06, "loss": 0.0727, "step": 31544 }, { "epoch": 3.17, "learning_rate": 5.688200875285725e-06, "loss": 0.5863, "step": 31545 }, { "epoch": 3.17, "learning_rate": 5.686798827871671e-06, "loss": 0.1277, "step": 31546 }, { "epoch": 3.17, "learning_rate": 5.685396935177581e-06, "loss": 0.2698, "step": 31547 }, { "epoch": 3.17, "learning_rate": 5.683995197212376e-06, "loss": 0.0915, "step": 31548 }, { "epoch": 3.17, "learning_rate": 5.682593613984974e-06, "loss": 0.1815, "step": 31549 }, { "epoch": 3.17, "learning_rate": 5.681192185504298e-06, "loss": 0.1316, "step": 31550 }, { "epoch": 3.17, "learning_rate": 5.67979091177926e-06, "loss": 0.0925, "step": 31551 }, { "epoch": 3.17, "learning_rate": 5.6783897928187835e-06, "loss": 0.0324, "step": 31552 }, { "epoch": 3.17, "learning_rate": 5.676988828631774e-06, "loss": 0.0048, "step": 31553 }, { "epoch": 3.17, "learning_rate": 5.675588019227162e-06, "loss": 0.3182, "step": 31554 }, { "epoch": 3.17, "learning_rate": 5.6741873646138585e-06, "loss": 0.1972, "step": 31555 }, { "epoch": 3.17, "learning_rate": 5.672786864800762e-06, "loss": 0.097, "step": 31556 }, { "epoch": 3.17, "learning_rate": 5.671386519796801e-06, "loss": 0.3326, "step": 31557 }, { "epoch": 3.17, "learning_rate": 5.669986329610881e-06, "loss": 0.1228, "step": 31558 }, { "epoch": 3.17, "learning_rate": 5.6685862942519125e-06, "loss": 0.1749, "step": 31559 }, { "epoch": 3.17, "learning_rate": 5.667186413728805e-06, "loss": 0.3028, "step": 31560 }, { "epoch": 3.17, "learning_rate": 5.665786688050468e-06, "loss": 0.0337, "step": 31561 }, { "epoch": 3.17, "learning_rate": 5.664387117225805e-06, "loss": 0.1369, "step": 31562 }, { "epoch": 3.17, "learning_rate": 5.662987701263726e-06, "loss": 0.0887, "step": 31563 }, { "epoch": 3.17, "learning_rate": 5.661588440173134e-06, "loss": 0.1794, "step": 31564 }, { "epoch": 3.17, "learning_rate": 5.660189333962934e-06, "loss": 0.2136, "step": 31565 }, { "epoch": 3.17, "learning_rate": 5.6587903826420315e-06, "loss": 0.2319, "step": 31566 }, { "epoch": 3.17, "learning_rate": 5.657391586219327e-06, "loss": 0.1888, "step": 31567 }, { "epoch": 3.17, "learning_rate": 5.655992944703713e-06, "loss": 0.1353, "step": 31568 }, { "epoch": 3.17, "learning_rate": 5.654594458104116e-06, "loss": 0.1274, "step": 31569 }, { "epoch": 3.17, "learning_rate": 5.653196126429408e-06, "loss": 0.4133, "step": 31570 }, { "epoch": 3.17, "learning_rate": 5.651797949688492e-06, "loss": 0.139, "step": 31571 }, { "epoch": 3.17, "learning_rate": 5.650399927890279e-06, "loss": 0.1551, "step": 31572 }, { "epoch": 3.17, "learning_rate": 5.6490020610436665e-06, "loss": 0.1166, "step": 31573 }, { "epoch": 3.17, "learning_rate": 5.647604349157522e-06, "loss": 0.1392, "step": 31574 }, { "epoch": 3.17, "learning_rate": 5.646206792240772e-06, "loss": 0.2154, "step": 31575 }, { "epoch": 3.17, "learning_rate": 5.644809390302303e-06, "loss": 0.5961, "step": 31576 }, { "epoch": 3.17, "learning_rate": 5.643412143350986e-06, "loss": 0.0715, "step": 31577 }, { "epoch": 3.17, "learning_rate": 5.642015051395735e-06, "loss": 0.0184, "step": 31578 }, { "epoch": 3.17, "learning_rate": 5.640618114445436e-06, "loss": 0.3063, "step": 31579 }, { "epoch": 3.17, "learning_rate": 5.639221332508976e-06, "loss": 0.5536, "step": 31580 }, { "epoch": 3.17, "learning_rate": 5.637824705595244e-06, "loss": 0.084, "step": 31581 }, { "epoch": 3.17, "learning_rate": 5.636428233713119e-06, "loss": 1.5094, "step": 31582 }, { "epoch": 3.17, "learning_rate": 5.635031916871512e-06, "loss": 0.1202, "step": 31583 }, { "epoch": 3.17, "learning_rate": 5.633635755079287e-06, "loss": 0.166, "step": 31584 }, { "epoch": 3.17, "learning_rate": 5.632239748345326e-06, "loss": 0.3206, "step": 31585 }, { "epoch": 3.17, "learning_rate": 5.630843896678527e-06, "loss": 0.0603, "step": 31586 }, { "epoch": 3.17, "learning_rate": 5.6294482000877765e-06, "loss": 0.3515, "step": 31587 }, { "epoch": 3.17, "learning_rate": 5.6280526585819315e-06, "loss": 0.1629, "step": 31588 }, { "epoch": 3.17, "learning_rate": 5.626657272169897e-06, "loss": 0.1329, "step": 31589 }, { "epoch": 3.17, "learning_rate": 5.62526204086054e-06, "loss": 0.2568, "step": 31590 }, { "epoch": 3.17, "learning_rate": 5.623866964662744e-06, "loss": 0.2169, "step": 31591 }, { "epoch": 3.17, "learning_rate": 5.622472043585387e-06, "loss": 0.0378, "step": 31592 }, { "epoch": 3.17, "learning_rate": 5.621077277637343e-06, "loss": 0.1965, "step": 31593 }, { "epoch": 3.17, "learning_rate": 5.619682666827488e-06, "loss": 0.1313, "step": 31594 }, { "epoch": 3.17, "learning_rate": 5.618288211164702e-06, "loss": 0.4498, "step": 31595 }, { "epoch": 3.17, "learning_rate": 5.6168939106578495e-06, "loss": 0.4323, "step": 31596 }, { "epoch": 3.17, "learning_rate": 5.615499765315809e-06, "loss": 0.3281, "step": 31597 }, { "epoch": 3.17, "learning_rate": 5.614105775147455e-06, "loss": 0.2167, "step": 31598 }, { "epoch": 3.17, "learning_rate": 5.612711940161652e-06, "loss": 0.2894, "step": 31599 }, { "epoch": 3.17, "learning_rate": 5.6113182603672656e-06, "loss": 0.1255, "step": 31600 }, { "epoch": 3.17, "learning_rate": 5.609924735773188e-06, "loss": 0.1869, "step": 31601 }, { "epoch": 3.17, "learning_rate": 5.608531366388262e-06, "loss": 0.1771, "step": 31602 }, { "epoch": 3.17, "learning_rate": 5.607138152221358e-06, "loss": 0.1031, "step": 31603 }, { "epoch": 3.17, "learning_rate": 5.6057450932813525e-06, "loss": 0.1243, "step": 31604 }, { "epoch": 3.17, "learning_rate": 5.6043521895771074e-06, "loss": 0.1115, "step": 31605 }, { "epoch": 3.17, "learning_rate": 5.602959441117484e-06, "loss": 0.5932, "step": 31606 }, { "epoch": 3.17, "learning_rate": 5.601566847911343e-06, "loss": 0.1913, "step": 31607 }, { "epoch": 3.17, "learning_rate": 5.600174409967551e-06, "loss": 0.3426, "step": 31608 }, { "epoch": 3.17, "learning_rate": 5.598782127294969e-06, "loss": 0.3909, "step": 31609 }, { "epoch": 3.17, "learning_rate": 5.597389999902452e-06, "loss": 0.5354, "step": 31610 }, { "epoch": 3.17, "learning_rate": 5.595998027798861e-06, "loss": 0.5259, "step": 31611 }, { "epoch": 3.17, "learning_rate": 5.594606210993057e-06, "loss": 0.0421, "step": 31612 }, { "epoch": 3.17, "learning_rate": 5.593214549493895e-06, "loss": 0.3687, "step": 31613 }, { "epoch": 3.17, "learning_rate": 5.591823043310219e-06, "loss": 0.0825, "step": 31614 }, { "epoch": 3.17, "learning_rate": 5.590431692450917e-06, "loss": 0.547, "step": 31615 }, { "epoch": 3.17, "learning_rate": 5.589040496924811e-06, "loss": 0.6666, "step": 31616 }, { "epoch": 3.17, "learning_rate": 5.587649456740758e-06, "loss": 0.2761, "step": 31617 }, { "epoch": 3.17, "learning_rate": 5.586258571907623e-06, "loss": 0.3895, "step": 31618 }, { "epoch": 3.17, "learning_rate": 5.58486784243426e-06, "loss": 0.3124, "step": 31619 }, { "epoch": 3.17, "learning_rate": 5.583477268329495e-06, "loss": 0.0778, "step": 31620 }, { "epoch": 3.17, "learning_rate": 5.5820868496021984e-06, "loss": 0.3751, "step": 31621 }, { "epoch": 3.17, "learning_rate": 5.580696586261215e-06, "loss": 0.3005, "step": 31622 }, { "epoch": 3.17, "learning_rate": 5.579306478315384e-06, "loss": 0.2019, "step": 31623 }, { "epoch": 3.17, "learning_rate": 5.577916525773561e-06, "loss": 0.0782, "step": 31624 }, { "epoch": 3.17, "learning_rate": 5.576526728644583e-06, "loss": 0.0458, "step": 31625 }, { "epoch": 3.17, "learning_rate": 5.575137086937301e-06, "loss": 0.5032, "step": 31626 }, { "epoch": 3.17, "learning_rate": 5.573747600660554e-06, "loss": 0.1783, "step": 31627 }, { "epoch": 3.17, "learning_rate": 5.572358269823184e-06, "loss": 0.1278, "step": 31628 }, { "epoch": 3.17, "learning_rate": 5.570969094434033e-06, "loss": 0.1569, "step": 31629 }, { "epoch": 3.17, "learning_rate": 5.569580074501945e-06, "loss": 0.0678, "step": 31630 }, { "epoch": 3.17, "learning_rate": 5.568191210035751e-06, "loss": 0.4858, "step": 31631 }, { "epoch": 3.18, "learning_rate": 5.56680250104429e-06, "loss": 0.0472, "step": 31632 }, { "epoch": 3.18, "learning_rate": 5.565413947536415e-06, "loss": 0.0261, "step": 31633 }, { "epoch": 3.18, "learning_rate": 5.564025549520936e-06, "loss": 0.3441, "step": 31634 }, { "epoch": 3.18, "learning_rate": 5.562637307006712e-06, "loss": 0.2315, "step": 31635 }, { "epoch": 3.18, "learning_rate": 5.561249220002565e-06, "loss": 0.3283, "step": 31636 }, { "epoch": 3.18, "learning_rate": 5.559861288517334e-06, "loss": 0.1228, "step": 31637 }, { "epoch": 3.18, "learning_rate": 5.558473512559846e-06, "loss": 0.2768, "step": 31638 }, { "epoch": 3.18, "learning_rate": 5.557085892138933e-06, "loss": 0.0711, "step": 31639 }, { "epoch": 3.18, "learning_rate": 5.5556984272634284e-06, "loss": 0.1647, "step": 31640 }, { "epoch": 3.18, "learning_rate": 5.5543111179421606e-06, "loss": 0.3667, "step": 31641 }, { "epoch": 3.18, "learning_rate": 5.552923964183958e-06, "loss": 0.3778, "step": 31642 }, { "epoch": 3.18, "learning_rate": 5.551536965997644e-06, "loss": 0.1386, "step": 31643 }, { "epoch": 3.18, "learning_rate": 5.550150123392053e-06, "loss": 0.11, "step": 31644 }, { "epoch": 3.18, "learning_rate": 5.548763436376002e-06, "loss": 0.5099, "step": 31645 }, { "epoch": 3.18, "learning_rate": 5.5473769049583124e-06, "loss": 0.3214, "step": 31646 }, { "epoch": 3.18, "learning_rate": 5.5459905291478265e-06, "loss": 0.3129, "step": 31647 }, { "epoch": 3.18, "learning_rate": 5.54460430895335e-06, "loss": 0.3372, "step": 31648 }, { "epoch": 3.18, "learning_rate": 5.543218244383698e-06, "loss": 0.1118, "step": 31649 }, { "epoch": 3.18, "learning_rate": 5.54183233544771e-06, "loss": 0.0312, "step": 31650 }, { "epoch": 3.18, "learning_rate": 5.540446582154206e-06, "loss": 0.2836, "step": 31651 }, { "epoch": 3.18, "learning_rate": 5.539060984511976e-06, "loss": 0.0376, "step": 31652 }, { "epoch": 3.18, "learning_rate": 5.537675542529869e-06, "loss": 0.2599, "step": 31653 }, { "epoch": 3.18, "learning_rate": 5.536290256216685e-06, "loss": 0.2976, "step": 31654 }, { "epoch": 3.18, "learning_rate": 5.534905125581246e-06, "loss": 0.1755, "step": 31655 }, { "epoch": 3.18, "learning_rate": 5.533520150632359e-06, "loss": 0.2769, "step": 31656 }, { "epoch": 3.18, "learning_rate": 5.532135331378847e-06, "loss": 0.1002, "step": 31657 }, { "epoch": 3.18, "learning_rate": 5.5307506678295175e-06, "loss": 0.3908, "step": 31658 }, { "epoch": 3.18, "learning_rate": 5.52936615999318e-06, "loss": 0.2564, "step": 31659 }, { "epoch": 3.18, "learning_rate": 5.527981807878642e-06, "loss": 0.1989, "step": 31660 }, { "epoch": 3.18, "learning_rate": 5.526597611494728e-06, "loss": 0.2573, "step": 31661 }, { "epoch": 3.18, "learning_rate": 5.525213570850236e-06, "loss": 0.0614, "step": 31662 }, { "epoch": 3.18, "learning_rate": 5.523829685953962e-06, "loss": 0.1038, "step": 31663 }, { "epoch": 3.18, "learning_rate": 5.522445956814735e-06, "loss": 0.4735, "step": 31664 }, { "epoch": 3.18, "learning_rate": 5.521062383441354e-06, "loss": 0.1469, "step": 31665 }, { "epoch": 3.18, "learning_rate": 5.519678965842607e-06, "loss": 0.2505, "step": 31666 }, { "epoch": 3.18, "learning_rate": 5.51829570402732e-06, "loss": 0.4536, "step": 31667 }, { "epoch": 3.18, "learning_rate": 5.5169125980042816e-06, "loss": 0.2227, "step": 31668 }, { "epoch": 3.18, "learning_rate": 5.5155296477823015e-06, "loss": 0.4425, "step": 31669 }, { "epoch": 3.18, "learning_rate": 5.514146853370174e-06, "loss": 0.1714, "step": 31670 }, { "epoch": 3.18, "learning_rate": 5.5127642147766995e-06, "loss": 0.3355, "step": 31671 }, { "epoch": 3.18, "learning_rate": 5.511381732010681e-06, "loss": 0.5148, "step": 31672 }, { "epoch": 3.18, "learning_rate": 5.509999405080912e-06, "loss": 0.1991, "step": 31673 }, { "epoch": 3.18, "learning_rate": 5.508617233996191e-06, "loss": 0.7304, "step": 31674 }, { "epoch": 3.18, "learning_rate": 5.507235218765312e-06, "loss": 0.1588, "step": 31675 }, { "epoch": 3.18, "learning_rate": 5.505853359397073e-06, "loss": 0.1027, "step": 31676 }, { "epoch": 3.18, "learning_rate": 5.504471655900263e-06, "loss": 0.1305, "step": 31677 }, { "epoch": 3.18, "learning_rate": 5.503090108283668e-06, "loss": 0.1286, "step": 31678 }, { "epoch": 3.18, "learning_rate": 5.501708716556105e-06, "loss": 0.2457, "step": 31679 }, { "epoch": 3.18, "learning_rate": 5.500327480726344e-06, "loss": 0.114, "step": 31680 }, { "epoch": 3.18, "learning_rate": 5.4989464008031665e-06, "loss": 0.1604, "step": 31681 }, { "epoch": 3.18, "learning_rate": 5.497565476795383e-06, "loss": 0.2281, "step": 31682 }, { "epoch": 3.18, "learning_rate": 5.496184708711777e-06, "loss": 0.1352, "step": 31683 }, { "epoch": 3.18, "learning_rate": 5.494804096561117e-06, "loss": 0.3402, "step": 31684 }, { "epoch": 3.18, "learning_rate": 5.4934236403522065e-06, "loss": 0.1709, "step": 31685 }, { "epoch": 3.18, "learning_rate": 5.4920433400938286e-06, "loss": 0.1443, "step": 31686 }, { "epoch": 3.18, "learning_rate": 5.490663195794763e-06, "loss": 0.0373, "step": 31687 }, { "epoch": 3.18, "learning_rate": 5.4892832074637885e-06, "loss": 0.5889, "step": 31688 }, { "epoch": 3.18, "learning_rate": 5.487903375109696e-06, "loss": 0.0626, "step": 31689 }, { "epoch": 3.18, "learning_rate": 5.4865236987412615e-06, "loss": 0.3994, "step": 31690 }, { "epoch": 3.18, "learning_rate": 5.485144178367264e-06, "loss": 0.256, "step": 31691 }, { "epoch": 3.18, "learning_rate": 5.483764813996473e-06, "loss": 0.0335, "step": 31692 }, { "epoch": 3.18, "learning_rate": 5.482385605637693e-06, "loss": 0.3167, "step": 31693 }, { "epoch": 3.18, "learning_rate": 5.481006553299679e-06, "loss": 0.2053, "step": 31694 }, { "epoch": 3.18, "learning_rate": 5.479627656991201e-06, "loss": 0.4418, "step": 31695 }, { "epoch": 3.18, "learning_rate": 5.478248916721055e-06, "loss": 0.2094, "step": 31696 }, { "epoch": 3.18, "learning_rate": 5.4768703324980075e-06, "loss": 0.25, "step": 31697 }, { "epoch": 3.18, "learning_rate": 5.475491904330815e-06, "loss": 0.0377, "step": 31698 }, { "epoch": 3.18, "learning_rate": 5.474113632228274e-06, "loss": 0.6227, "step": 31699 }, { "epoch": 3.18, "learning_rate": 5.47273551619914e-06, "loss": 0.1323, "step": 31700 }, { "epoch": 3.18, "learning_rate": 5.471357556252187e-06, "loss": 0.1592, "step": 31701 }, { "epoch": 3.18, "learning_rate": 5.4699797523961806e-06, "loss": 0.3663, "step": 31702 }, { "epoch": 3.18, "learning_rate": 5.468602104639894e-06, "loss": 0.3, "step": 31703 }, { "epoch": 3.18, "learning_rate": 5.467224612992093e-06, "loss": 0.1787, "step": 31704 }, { "epoch": 3.18, "learning_rate": 5.465847277461539e-06, "loss": 0.0962, "step": 31705 }, { "epoch": 3.18, "learning_rate": 5.464470098056999e-06, "loss": 0.3862, "step": 31706 }, { "epoch": 3.18, "learning_rate": 5.463093074787242e-06, "loss": 0.4631, "step": 31707 }, { "epoch": 3.18, "learning_rate": 5.461716207661023e-06, "loss": 0.0649, "step": 31708 }, { "epoch": 3.18, "learning_rate": 5.460339496687108e-06, "loss": 0.2055, "step": 31709 }, { "epoch": 3.18, "learning_rate": 5.4589629418742505e-06, "loss": 0.1955, "step": 31710 }, { "epoch": 3.18, "learning_rate": 5.45758654323123e-06, "loss": 0.1385, "step": 31711 }, { "epoch": 3.18, "learning_rate": 5.456210300766786e-06, "loss": 0.0335, "step": 31712 }, { "epoch": 3.18, "learning_rate": 5.454834214489675e-06, "loss": 0.0447, "step": 31713 }, { "epoch": 3.18, "learning_rate": 5.453458284408667e-06, "loss": 0.2003, "step": 31714 }, { "epoch": 3.18, "learning_rate": 5.452082510532515e-06, "loss": 0.0934, "step": 31715 }, { "epoch": 3.18, "learning_rate": 5.4507068928699725e-06, "loss": 0.0881, "step": 31716 }, { "epoch": 3.18, "learning_rate": 5.44933143142979e-06, "loss": 0.3418, "step": 31717 }, { "epoch": 3.18, "learning_rate": 5.447956126220724e-06, "loss": 0.4013, "step": 31718 }, { "epoch": 3.18, "learning_rate": 5.446580977251524e-06, "loss": 0.2481, "step": 31719 }, { "epoch": 3.18, "learning_rate": 5.445205984530942e-06, "loss": 0.179, "step": 31720 }, { "epoch": 3.18, "learning_rate": 5.443831148067731e-06, "loss": 0.096, "step": 31721 }, { "epoch": 3.18, "learning_rate": 5.442456467870633e-06, "loss": 0.0849, "step": 31722 }, { "epoch": 3.18, "learning_rate": 5.4410819439484005e-06, "loss": 0.1224, "step": 31723 }, { "epoch": 3.18, "learning_rate": 5.4397075763097745e-06, "loss": 0.1738, "step": 31724 }, { "epoch": 3.18, "learning_rate": 5.438333364963518e-06, "loss": 0.3176, "step": 31725 }, { "epoch": 3.18, "learning_rate": 5.43695930991836e-06, "loss": 0.0717, "step": 31726 }, { "epoch": 3.18, "learning_rate": 5.43558541118304e-06, "loss": 0.3359, "step": 31727 }, { "epoch": 3.18, "learning_rate": 5.434211668766316e-06, "loss": 0.4324, "step": 31728 }, { "epoch": 3.18, "learning_rate": 5.4328380826769315e-06, "loss": 0.2411, "step": 31729 }, { "epoch": 3.19, "learning_rate": 5.431464652923605e-06, "loss": 0.2322, "step": 31730 }, { "epoch": 3.19, "learning_rate": 5.4300913795150994e-06, "loss": 0.5526, "step": 31731 }, { "epoch": 3.19, "learning_rate": 5.428718262460147e-06, "loss": 0.1923, "step": 31732 }, { "epoch": 3.19, "learning_rate": 5.427345301767481e-06, "loss": 0.3176, "step": 31733 }, { "epoch": 3.19, "learning_rate": 5.425972497445845e-06, "loss": 0.1505, "step": 31734 }, { "epoch": 3.19, "learning_rate": 5.424599849503967e-06, "loss": 0.0766, "step": 31735 }, { "epoch": 3.19, "learning_rate": 5.423227357950589e-06, "loss": 0.0235, "step": 31736 }, { "epoch": 3.19, "learning_rate": 5.421855022794442e-06, "loss": 0.1037, "step": 31737 }, { "epoch": 3.19, "learning_rate": 5.42048284404426e-06, "loss": 0.0679, "step": 31738 }, { "epoch": 3.19, "learning_rate": 5.419110821708776e-06, "loss": 0.0531, "step": 31739 }, { "epoch": 3.19, "learning_rate": 5.41773895579672e-06, "loss": 0.3034, "step": 31740 }, { "epoch": 3.19, "learning_rate": 5.41636724631681e-06, "loss": 0.4382, "step": 31741 }, { "epoch": 3.19, "learning_rate": 5.414995693277797e-06, "loss": 0.0134, "step": 31742 }, { "epoch": 3.19, "learning_rate": 5.4136242966884084e-06, "loss": 0.352, "step": 31743 }, { "epoch": 3.19, "learning_rate": 5.412253056557342e-06, "loss": 0.476, "step": 31744 }, { "epoch": 3.19, "learning_rate": 5.410881972893356e-06, "loss": 0.1243, "step": 31745 }, { "epoch": 3.19, "learning_rate": 5.4095110457051615e-06, "loss": 0.1891, "step": 31746 }, { "epoch": 3.19, "learning_rate": 5.408140275001482e-06, "loss": 0.1279, "step": 31747 }, { "epoch": 3.19, "learning_rate": 5.406769660791041e-06, "loss": 0.3708, "step": 31748 }, { "epoch": 3.19, "learning_rate": 5.405399203082565e-06, "loss": 0.0362, "step": 31749 }, { "epoch": 3.19, "learning_rate": 5.404028901884773e-06, "loss": 0.0572, "step": 31750 }, { "epoch": 3.19, "learning_rate": 5.402658757206382e-06, "loss": 0.1457, "step": 31751 }, { "epoch": 3.19, "learning_rate": 5.401288769056116e-06, "loss": 0.1945, "step": 31752 }, { "epoch": 3.19, "learning_rate": 5.399918937442688e-06, "loss": 0.3347, "step": 31753 }, { "epoch": 3.19, "learning_rate": 5.398549262374817e-06, "loss": 0.1143, "step": 31754 }, { "epoch": 3.19, "learning_rate": 5.397179743861224e-06, "loss": 0.2701, "step": 31755 }, { "epoch": 3.19, "learning_rate": 5.395810381910608e-06, "loss": 0.249, "step": 31756 }, { "epoch": 3.19, "learning_rate": 5.394441176531713e-06, "loss": 0.1357, "step": 31757 }, { "epoch": 3.19, "learning_rate": 5.393072127733222e-06, "loss": 0.1705, "step": 31758 }, { "epoch": 3.19, "learning_rate": 5.391703235523856e-06, "loss": 0.4635, "step": 31759 }, { "epoch": 3.19, "learning_rate": 5.390334499912335e-06, "loss": 0.0631, "step": 31760 }, { "epoch": 3.19, "learning_rate": 5.388965920907368e-06, "loss": 0.148, "step": 31761 }, { "epoch": 3.19, "learning_rate": 5.387597498517649e-06, "loss": 0.5449, "step": 31762 }, { "epoch": 3.19, "learning_rate": 5.3862292327519005e-06, "loss": 0.3504, "step": 31763 }, { "epoch": 3.19, "learning_rate": 5.384861123618828e-06, "loss": 0.1523, "step": 31764 }, { "epoch": 3.19, "learning_rate": 5.383493171127133e-06, "loss": 0.0826, "step": 31765 }, { "epoch": 3.19, "learning_rate": 5.382125375285524e-06, "loss": 0.11, "step": 31766 }, { "epoch": 3.19, "learning_rate": 5.380757736102704e-06, "loss": 0.2387, "step": 31767 }, { "epoch": 3.19, "learning_rate": 5.3793902535873765e-06, "loss": 0.2385, "step": 31768 }, { "epoch": 3.19, "learning_rate": 5.378022927748244e-06, "loss": 0.2682, "step": 31769 }, { "epoch": 3.19, "learning_rate": 5.376655758593997e-06, "loss": 1.0983, "step": 31770 }, { "epoch": 3.19, "learning_rate": 5.375288746133361e-06, "loss": 0.1906, "step": 31771 }, { "epoch": 3.19, "learning_rate": 5.373921890375013e-06, "loss": 0.424, "step": 31772 }, { "epoch": 3.19, "learning_rate": 5.372555191327649e-06, "loss": 0.1875, "step": 31773 }, { "epoch": 3.19, "learning_rate": 5.371188648999987e-06, "loss": 0.0679, "step": 31774 }, { "epoch": 3.19, "learning_rate": 5.369822263400713e-06, "loss": 0.2822, "step": 31775 }, { "epoch": 3.19, "learning_rate": 5.36845603453851e-06, "loss": 0.6332, "step": 31776 }, { "epoch": 3.19, "learning_rate": 5.367089962422087e-06, "loss": 0.0924, "step": 31777 }, { "epoch": 3.19, "learning_rate": 5.365724047060133e-06, "loss": 0.3248, "step": 31778 }, { "epoch": 3.19, "learning_rate": 5.364358288461342e-06, "loss": 0.23, "step": 31779 }, { "epoch": 3.19, "learning_rate": 5.362992686634399e-06, "loss": 0.1767, "step": 31780 }, { "epoch": 3.19, "learning_rate": 5.361627241588002e-06, "loss": 0.2068, "step": 31781 }, { "epoch": 3.19, "learning_rate": 5.360261953330833e-06, "loss": 0.0722, "step": 31782 }, { "epoch": 3.19, "learning_rate": 5.358896821871585e-06, "loss": 0.523, "step": 31783 }, { "epoch": 3.19, "learning_rate": 5.357531847218945e-06, "loss": 0.6243, "step": 31784 }, { "epoch": 3.19, "learning_rate": 5.356167029381595e-06, "loss": 0.0251, "step": 31785 }, { "epoch": 3.19, "learning_rate": 5.3548023683682265e-06, "loss": 0.0564, "step": 31786 }, { "epoch": 3.19, "learning_rate": 5.353437864187518e-06, "loss": 0.1893, "step": 31787 }, { "epoch": 3.19, "learning_rate": 5.352073516848149e-06, "loss": 0.1082, "step": 31788 }, { "epoch": 3.19, "learning_rate": 5.3507093263588205e-06, "loss": 0.1085, "step": 31789 }, { "epoch": 3.19, "learning_rate": 5.3493452927281935e-06, "loss": 0.2158, "step": 31790 }, { "epoch": 3.19, "learning_rate": 5.3479814159649496e-06, "loss": 0.089, "step": 31791 }, { "epoch": 3.19, "learning_rate": 5.346617696077781e-06, "loss": 0.0526, "step": 31792 }, { "epoch": 3.19, "learning_rate": 5.345254133075365e-06, "loss": 0.5057, "step": 31793 }, { "epoch": 3.19, "learning_rate": 5.34389072696636e-06, "loss": 0.1214, "step": 31794 }, { "epoch": 3.19, "learning_rate": 5.342527477759459e-06, "loss": 0.074, "step": 31795 }, { "epoch": 3.19, "learning_rate": 5.341164385463336e-06, "loss": 0.1117, "step": 31796 }, { "epoch": 3.19, "learning_rate": 5.339801450086661e-06, "loss": 0.0779, "step": 31797 }, { "epoch": 3.19, "learning_rate": 5.338438671638106e-06, "loss": 0.0272, "step": 31798 }, { "epoch": 3.19, "learning_rate": 5.337076050126348e-06, "loss": 0.2127, "step": 31799 }, { "epoch": 3.19, "learning_rate": 5.3357135855600565e-06, "loss": 0.2148, "step": 31800 }, { "epoch": 3.19, "learning_rate": 5.334351277947899e-06, "loss": 0.1598, "step": 31801 }, { "epoch": 3.19, "learning_rate": 5.33298912729854e-06, "loss": 0.3517, "step": 31802 }, { "epoch": 3.19, "learning_rate": 5.33162713362067e-06, "loss": 0.4109, "step": 31803 }, { "epoch": 3.19, "learning_rate": 5.330265296922933e-06, "loss": 0.2166, "step": 31804 }, { "epoch": 3.19, "learning_rate": 5.328903617213993e-06, "loss": 0.4465, "step": 31805 }, { "epoch": 3.19, "learning_rate": 5.3275420945025335e-06, "loss": 0.1416, "step": 31806 }, { "epoch": 3.19, "learning_rate": 5.326180728797221e-06, "loss": 0.1555, "step": 31807 }, { "epoch": 3.19, "learning_rate": 5.32481952010669e-06, "loss": 0.0901, "step": 31808 }, { "epoch": 3.19, "learning_rate": 5.323458468439626e-06, "loss": 0.2136, "step": 31809 }, { "epoch": 3.19, "learning_rate": 5.32209757380469e-06, "loss": 0.3668, "step": 31810 }, { "epoch": 3.19, "learning_rate": 5.320736836210533e-06, "loss": 0.1344, "step": 31811 }, { "epoch": 3.19, "learning_rate": 5.319376255665821e-06, "loss": 0.0497, "step": 31812 }, { "epoch": 3.19, "learning_rate": 5.318015832179208e-06, "loss": 0.176, "step": 31813 }, { "epoch": 3.19, "learning_rate": 5.316655565759352e-06, "loss": 0.2797, "step": 31814 }, { "epoch": 3.19, "learning_rate": 5.315295456414912e-06, "loss": 0.1467, "step": 31815 }, { "epoch": 3.19, "learning_rate": 5.31393550415454e-06, "loss": 0.0993, "step": 31816 }, { "epoch": 3.19, "learning_rate": 5.312575708986891e-06, "loss": 0.2453, "step": 31817 }, { "epoch": 3.19, "learning_rate": 5.31121607092062e-06, "loss": 0.1818, "step": 31818 }, { "epoch": 3.19, "learning_rate": 5.309856589964378e-06, "loss": 0.1152, "step": 31819 }, { "epoch": 3.19, "learning_rate": 5.308497266126807e-06, "loss": 0.3116, "step": 31820 }, { "epoch": 3.19, "learning_rate": 5.307138099416581e-06, "loss": 0.242, "step": 31821 }, { "epoch": 3.19, "learning_rate": 5.305779089842322e-06, "loss": 0.2462, "step": 31822 }, { "epoch": 3.19, "learning_rate": 5.304420237412694e-06, "loss": 0.2538, "step": 31823 }, { "epoch": 3.19, "learning_rate": 5.303061542136346e-06, "loss": 0.0803, "step": 31824 }, { "epoch": 3.19, "learning_rate": 5.301703004021915e-06, "loss": 0.2292, "step": 31825 }, { "epoch": 3.19, "learning_rate": 5.300344623078054e-06, "loss": 0.0933, "step": 31826 }, { "epoch": 3.19, "learning_rate": 5.298986399313399e-06, "loss": 0.105, "step": 31827 }, { "epoch": 3.19, "learning_rate": 5.297628332736601e-06, "loss": 0.0235, "step": 31828 }, { "epoch": 3.2, "learning_rate": 5.296270423356297e-06, "loss": 0.3622, "step": 31829 }, { "epoch": 3.2, "learning_rate": 5.2949126711811336e-06, "loss": 0.1761, "step": 31830 }, { "epoch": 3.2, "learning_rate": 5.293555076219745e-06, "loss": 0.1072, "step": 31831 }, { "epoch": 3.2, "learning_rate": 5.292197638480774e-06, "loss": 0.326, "step": 31832 }, { "epoch": 3.2, "learning_rate": 5.29084035797286e-06, "loss": 0.1136, "step": 31833 }, { "epoch": 3.2, "learning_rate": 5.289483234704625e-06, "loss": 0.1251, "step": 31834 }, { "epoch": 3.2, "learning_rate": 5.288126268684738e-06, "loss": 0.0377, "step": 31835 }, { "epoch": 3.2, "learning_rate": 5.286769459921807e-06, "loss": 0.5614, "step": 31836 }, { "epoch": 3.2, "learning_rate": 5.2854128084244644e-06, "loss": 0.1187, "step": 31837 }, { "epoch": 3.2, "learning_rate": 5.28405631420136e-06, "loss": 0.1587, "step": 31838 }, { "epoch": 3.2, "learning_rate": 5.282699977261128e-06, "loss": 0.2895, "step": 31839 }, { "epoch": 3.2, "learning_rate": 5.281343797612374e-06, "loss": 0.3134, "step": 31840 }, { "epoch": 3.2, "learning_rate": 5.279987775263751e-06, "loss": 0.2256, "step": 31841 }, { "epoch": 3.2, "learning_rate": 5.278631910223886e-06, "loss": 0.2206, "step": 31842 }, { "epoch": 3.2, "learning_rate": 5.277276202501399e-06, "loss": 0.5771, "step": 31843 }, { "epoch": 3.2, "learning_rate": 5.275920652104924e-06, "loss": 0.1805, "step": 31844 }, { "epoch": 3.2, "learning_rate": 5.2745652590430815e-06, "loss": 0.1087, "step": 31845 }, { "epoch": 3.2, "learning_rate": 5.273210023324504e-06, "loss": 0.5022, "step": 31846 }, { "epoch": 3.2, "learning_rate": 5.271854944957808e-06, "loss": 0.2623, "step": 31847 }, { "epoch": 3.2, "learning_rate": 5.27050002395162e-06, "loss": 0.4749, "step": 31848 }, { "epoch": 3.2, "learning_rate": 5.269145260314563e-06, "loss": 0.1696, "step": 31849 }, { "epoch": 3.2, "learning_rate": 5.267790654055257e-06, "loss": 0.0319, "step": 31850 }, { "epoch": 3.2, "learning_rate": 5.266436205182318e-06, "loss": 0.6369, "step": 31851 }, { "epoch": 3.2, "learning_rate": 5.265081913704373e-06, "loss": 0.1782, "step": 31852 }, { "epoch": 3.2, "learning_rate": 5.263727779630048e-06, "loss": 0.2182, "step": 31853 }, { "epoch": 3.2, "learning_rate": 5.262373802967933e-06, "loss": 0.2121, "step": 31854 }, { "epoch": 3.2, "learning_rate": 5.261019983726667e-06, "loss": 0.1791, "step": 31855 }, { "epoch": 3.2, "learning_rate": 5.259666321914861e-06, "loss": 0.1666, "step": 31856 }, { "epoch": 3.2, "learning_rate": 5.258312817541124e-06, "loss": 0.1868, "step": 31857 }, { "epoch": 3.2, "learning_rate": 5.256959470614073e-06, "loss": 0.1099, "step": 31858 }, { "epoch": 3.2, "learning_rate": 5.255606281142321e-06, "loss": 0.1403, "step": 31859 }, { "epoch": 3.2, "learning_rate": 5.254253249134474e-06, "loss": 0.1558, "step": 31860 }, { "epoch": 3.2, "learning_rate": 5.252900374599149e-06, "loss": 0.0333, "step": 31861 }, { "epoch": 3.2, "learning_rate": 5.251547657544948e-06, "loss": 0.0924, "step": 31862 }, { "epoch": 3.2, "learning_rate": 5.250195097980486e-06, "loss": 0.1308, "step": 31863 }, { "epoch": 3.2, "learning_rate": 5.248842695914368e-06, "loss": 0.1996, "step": 31864 }, { "epoch": 3.2, "learning_rate": 5.247490451355197e-06, "loss": 0.3529, "step": 31865 }, { "epoch": 3.2, "learning_rate": 5.246138364311574e-06, "loss": 0.4015, "step": 31866 }, { "epoch": 3.2, "learning_rate": 5.244786434792124e-06, "loss": 0.0806, "step": 31867 }, { "epoch": 3.2, "learning_rate": 5.243434662805428e-06, "loss": 0.2966, "step": 31868 }, { "epoch": 3.2, "learning_rate": 5.242083048360088e-06, "loss": 1.0775, "step": 31869 }, { "epoch": 3.2, "learning_rate": 5.240731591464722e-06, "loss": 0.0297, "step": 31870 }, { "epoch": 3.2, "learning_rate": 5.239380292127928e-06, "loss": 0.0452, "step": 31871 }, { "epoch": 3.2, "learning_rate": 5.23802915035828e-06, "loss": 0.3465, "step": 31872 }, { "epoch": 3.2, "learning_rate": 5.236678166164408e-06, "loss": 0.4749, "step": 31873 }, { "epoch": 3.2, "learning_rate": 5.235327339554892e-06, "loss": 0.2304, "step": 31874 }, { "epoch": 3.2, "learning_rate": 5.233976670538334e-06, "loss": 0.3188, "step": 31875 }, { "epoch": 3.2, "learning_rate": 5.232626159123325e-06, "loss": 0.0878, "step": 31876 }, { "epoch": 3.2, "learning_rate": 5.231275805318454e-06, "loss": 0.3322, "step": 31877 }, { "epoch": 3.2, "learning_rate": 5.229925609132338e-06, "loss": 0.3785, "step": 31878 }, { "epoch": 3.2, "learning_rate": 5.2285755705735424e-06, "loss": 0.1385, "step": 31879 }, { "epoch": 3.2, "learning_rate": 5.227225689650662e-06, "loss": 0.0493, "step": 31880 }, { "epoch": 3.2, "learning_rate": 5.225875966372309e-06, "loss": 0.1626, "step": 31881 }, { "epoch": 3.2, "learning_rate": 5.22452640074705e-06, "loss": 0.2571, "step": 31882 }, { "epoch": 3.2, "learning_rate": 5.2231769927834705e-06, "loss": 0.0828, "step": 31883 }, { "epoch": 3.2, "learning_rate": 5.221827742490175e-06, "loss": 0.2019, "step": 31884 }, { "epoch": 3.2, "learning_rate": 5.220478649875749e-06, "loss": 0.2836, "step": 31885 }, { "epoch": 3.2, "learning_rate": 5.219129714948757e-06, "loss": 0.1181, "step": 31886 }, { "epoch": 3.2, "learning_rate": 5.217780937717801e-06, "loss": 0.1163, "step": 31887 }, { "epoch": 3.2, "learning_rate": 5.216432318191461e-06, "loss": 0.0529, "step": 31888 }, { "epoch": 3.2, "learning_rate": 5.21508385637832e-06, "loss": 0.4394, "step": 31889 }, { "epoch": 3.2, "learning_rate": 5.213735552286951e-06, "loss": 0.15, "step": 31890 }, { "epoch": 3.2, "learning_rate": 5.2123874059259444e-06, "loss": 0.1241, "step": 31891 }, { "epoch": 3.2, "learning_rate": 5.211039417303873e-06, "loss": 0.091, "step": 31892 }, { "epoch": 3.2, "learning_rate": 5.209691586429317e-06, "loss": 0.2249, "step": 31893 }, { "epoch": 3.2, "learning_rate": 5.208343913310853e-06, "loss": 0.0852, "step": 31894 }, { "epoch": 3.2, "learning_rate": 5.2069963979570556e-06, "loss": 0.2421, "step": 31895 }, { "epoch": 3.2, "learning_rate": 5.205649040376503e-06, "loss": 0.1198, "step": 31896 }, { "epoch": 3.2, "learning_rate": 5.204301840577769e-06, "loss": 0.1576, "step": 31897 }, { "epoch": 3.2, "learning_rate": 5.202954798569415e-06, "loss": 0.3104, "step": 31898 }, { "epoch": 3.2, "learning_rate": 5.2016079143600415e-06, "loss": 0.0587, "step": 31899 }, { "epoch": 3.2, "learning_rate": 5.2002611879581915e-06, "loss": 0.297, "step": 31900 }, { "epoch": 3.2, "learning_rate": 5.198914619372435e-06, "loss": 0.3317, "step": 31901 }, { "epoch": 3.2, "learning_rate": 5.197568208611363e-06, "loss": 0.1827, "step": 31902 }, { "epoch": 3.2, "learning_rate": 5.19622195568353e-06, "loss": 0.0806, "step": 31903 }, { "epoch": 3.2, "learning_rate": 5.1948758605975065e-06, "loss": 0.3421, "step": 31904 }, { "epoch": 3.2, "learning_rate": 5.193529923361853e-06, "loss": 0.1773, "step": 31905 }, { "epoch": 3.2, "learning_rate": 5.1921841439851415e-06, "loss": 0.7648, "step": 31906 }, { "epoch": 3.2, "learning_rate": 5.190838522475933e-06, "loss": 0.2385, "step": 31907 }, { "epoch": 3.2, "learning_rate": 5.189493058842786e-06, "loss": 0.5468, "step": 31908 }, { "epoch": 3.2, "learning_rate": 5.1881477530942615e-06, "loss": 0.1285, "step": 31909 }, { "epoch": 3.2, "learning_rate": 5.186802605238941e-06, "loss": 0.6248, "step": 31910 }, { "epoch": 3.2, "learning_rate": 5.1854576152853635e-06, "loss": 0.286, "step": 31911 }, { "epoch": 3.2, "learning_rate": 5.184112783242084e-06, "loss": 0.2089, "step": 31912 }, { "epoch": 3.2, "learning_rate": 5.182768109117686e-06, "loss": 0.538, "step": 31913 }, { "epoch": 3.2, "learning_rate": 5.181423592920705e-06, "loss": 0.1847, "step": 31914 }, { "epoch": 3.2, "learning_rate": 5.1800792346596926e-06, "loss": 0.2361, "step": 31915 }, { "epoch": 3.2, "learning_rate": 5.1787350343432236e-06, "loss": 0.1446, "step": 31916 }, { "epoch": 3.2, "learning_rate": 5.177390991979849e-06, "loss": 0.0978, "step": 31917 }, { "epoch": 3.2, "learning_rate": 5.176047107578098e-06, "loss": 0.0916, "step": 31918 }, { "epoch": 3.2, "learning_rate": 5.17470338114655e-06, "loss": 0.279, "step": 31919 }, { "epoch": 3.2, "learning_rate": 5.173359812693747e-06, "loss": 0.0488, "step": 31920 }, { "epoch": 3.2, "learning_rate": 5.172016402228232e-06, "loss": 0.4123, "step": 31921 }, { "epoch": 3.2, "learning_rate": 5.170673149758565e-06, "loss": 0.1127, "step": 31922 }, { "epoch": 3.2, "learning_rate": 5.169330055293287e-06, "loss": 0.1907, "step": 31923 }, { "epoch": 3.2, "learning_rate": 5.167987118840944e-06, "loss": 0.113, "step": 31924 }, { "epoch": 3.2, "learning_rate": 5.166644340410089e-06, "loss": 0.1692, "step": 31925 }, { "epoch": 3.2, "learning_rate": 5.165301720009258e-06, "loss": 0.6695, "step": 31926 }, { "epoch": 3.21, "learning_rate": 5.163959257646999e-06, "loss": 0.0481, "step": 31927 }, { "epoch": 3.21, "learning_rate": 5.16261695333186e-06, "loss": 0.075, "step": 31928 }, { "epoch": 3.21, "learning_rate": 5.161274807072373e-06, "loss": 0.516, "step": 31929 }, { "epoch": 3.21, "learning_rate": 5.15993281887708e-06, "loss": 0.0194, "step": 31930 }, { "epoch": 3.21, "learning_rate": 5.1585909887545395e-06, "loss": 0.1296, "step": 31931 }, { "epoch": 3.21, "learning_rate": 5.1572493167132594e-06, "loss": 0.4753, "step": 31932 }, { "epoch": 3.21, "learning_rate": 5.155907802761801e-06, "loss": 0.131, "step": 31933 }, { "epoch": 3.21, "learning_rate": 5.154566446908694e-06, "loss": 0.0712, "step": 31934 }, { "epoch": 3.21, "learning_rate": 5.1532252491624775e-06, "loss": 0.1942, "step": 31935 }, { "epoch": 3.21, "learning_rate": 5.151884209531679e-06, "loss": 0.2017, "step": 31936 }, { "epoch": 3.21, "learning_rate": 5.150543328024841e-06, "loss": 0.5611, "step": 31937 }, { "epoch": 3.21, "learning_rate": 5.1492026046504894e-06, "loss": 0.1363, "step": 31938 }, { "epoch": 3.21, "learning_rate": 5.147862039417158e-06, "loss": 0.261, "step": 31939 }, { "epoch": 3.21, "learning_rate": 5.146521632333381e-06, "loss": 0.0877, "step": 31940 }, { "epoch": 3.21, "learning_rate": 5.145181383407676e-06, "loss": 0.5029, "step": 31941 }, { "epoch": 3.21, "learning_rate": 5.143841292648599e-06, "loss": 0.0748, "step": 31942 }, { "epoch": 3.21, "learning_rate": 5.142501360064651e-06, "loss": 0.2698, "step": 31943 }, { "epoch": 3.21, "learning_rate": 5.141161585664359e-06, "loss": 0.1701, "step": 31944 }, { "epoch": 3.21, "learning_rate": 5.139821969456276e-06, "loss": 0.1053, "step": 31945 }, { "epoch": 3.21, "learning_rate": 5.1384825114489e-06, "loss": 0.139, "step": 31946 }, { "epoch": 3.21, "learning_rate": 5.1371432116507556e-06, "loss": 0.0888, "step": 31947 }, { "epoch": 3.21, "learning_rate": 5.13580407007038e-06, "loss": 0.5321, "step": 31948 }, { "epoch": 3.21, "learning_rate": 5.134465086716297e-06, "loss": 0.2761, "step": 31949 }, { "epoch": 3.21, "learning_rate": 5.133126261597002e-06, "loss": 0.1813, "step": 31950 }, { "epoch": 3.21, "learning_rate": 5.131787594721042e-06, "loss": 0.0514, "step": 31951 }, { "epoch": 3.21, "learning_rate": 5.130449086096925e-06, "loss": 0.0275, "step": 31952 }, { "epoch": 3.21, "learning_rate": 5.129110735733168e-06, "loss": 0.446, "step": 31953 }, { "epoch": 3.21, "learning_rate": 5.12777254363829e-06, "loss": 0.0201, "step": 31954 }, { "epoch": 3.21, "learning_rate": 5.126434509820806e-06, "loss": 0.0197, "step": 31955 }, { "epoch": 3.21, "learning_rate": 5.125096634289229e-06, "loss": 0.2469, "step": 31956 }, { "epoch": 3.21, "learning_rate": 5.123758917052074e-06, "loss": 0.2252, "step": 31957 }, { "epoch": 3.21, "learning_rate": 5.122421358117846e-06, "loss": 0.3505, "step": 31958 }, { "epoch": 3.21, "learning_rate": 5.1210839574950816e-06, "loss": 0.1545, "step": 31959 }, { "epoch": 3.21, "learning_rate": 5.119746715192265e-06, "loss": 0.2672, "step": 31960 }, { "epoch": 3.21, "learning_rate": 5.118409631217908e-06, "loss": 0.1555, "step": 31961 }, { "epoch": 3.21, "learning_rate": 5.117072705580535e-06, "loss": 0.4121, "step": 31962 }, { "epoch": 3.21, "learning_rate": 5.1157359382886524e-06, "loss": 0.5358, "step": 31963 }, { "epoch": 3.21, "learning_rate": 5.114399329350746e-06, "loss": 0.2906, "step": 31964 }, { "epoch": 3.21, "learning_rate": 5.113062878775343e-06, "loss": 0.0456, "step": 31965 }, { "epoch": 3.21, "learning_rate": 5.111726586570939e-06, "loss": 0.1541, "step": 31966 }, { "epoch": 3.21, "learning_rate": 5.1103904527460386e-06, "loss": 0.1134, "step": 31967 }, { "epoch": 3.21, "learning_rate": 5.109054477309147e-06, "loss": 0.3243, "step": 31968 }, { "epoch": 3.21, "learning_rate": 5.107718660268762e-06, "loss": 0.1201, "step": 31969 }, { "epoch": 3.21, "learning_rate": 5.106383001633385e-06, "loss": 0.6414, "step": 31970 }, { "epoch": 3.21, "learning_rate": 5.105047501411518e-06, "loss": 0.1664, "step": 31971 }, { "epoch": 3.21, "learning_rate": 5.10371215961166e-06, "loss": 0.2174, "step": 31972 }, { "epoch": 3.21, "learning_rate": 5.102376976242297e-06, "loss": 0.462, "step": 31973 }, { "epoch": 3.21, "learning_rate": 5.101041951311949e-06, "loss": 0.0758, "step": 31974 }, { "epoch": 3.21, "learning_rate": 5.099707084829093e-06, "loss": 0.155, "step": 31975 }, { "epoch": 3.21, "learning_rate": 5.0983723768022214e-06, "loss": 0.2449, "step": 31976 }, { "epoch": 3.21, "learning_rate": 5.097037827239849e-06, "loss": 0.6676, "step": 31977 }, { "epoch": 3.21, "learning_rate": 5.095703436150444e-06, "loss": 0.1781, "step": 31978 }, { "epoch": 3.21, "learning_rate": 5.0943692035425035e-06, "loss": 0.1273, "step": 31979 }, { "epoch": 3.21, "learning_rate": 5.093035129424531e-06, "loss": 0.1308, "step": 31980 }, { "epoch": 3.21, "learning_rate": 5.091701213805013e-06, "loss": 0.1019, "step": 31981 }, { "epoch": 3.21, "learning_rate": 5.09036745669242e-06, "loss": 0.336, "step": 31982 }, { "epoch": 3.21, "learning_rate": 5.08903385809526e-06, "loss": 0.5107, "step": 31983 }, { "epoch": 3.21, "learning_rate": 5.087700418022011e-06, "loss": 0.1223, "step": 31984 }, { "epoch": 3.21, "learning_rate": 5.086367136481159e-06, "loss": 0.0397, "step": 31985 }, { "epoch": 3.21, "learning_rate": 5.085034013481191e-06, "loss": 0.2423, "step": 31986 }, { "epoch": 3.21, "learning_rate": 5.083701049030577e-06, "loss": 0.1037, "step": 31987 }, { "epoch": 3.21, "learning_rate": 5.082368243137829e-06, "loss": 0.1638, "step": 31988 }, { "epoch": 3.21, "learning_rate": 5.081035595811402e-06, "loss": 0.3031, "step": 31989 }, { "epoch": 3.21, "learning_rate": 5.079703107059778e-06, "loss": 0.0771, "step": 31990 }, { "epoch": 3.21, "learning_rate": 5.078370776891457e-06, "loss": 0.4768, "step": 31991 }, { "epoch": 3.21, "learning_rate": 5.077038605314899e-06, "loss": 0.1327, "step": 31992 }, { "epoch": 3.21, "learning_rate": 5.075706592338574e-06, "loss": 0.0997, "step": 31993 }, { "epoch": 3.21, "learning_rate": 5.074374737970981e-06, "loss": 0.3731, "step": 31994 }, { "epoch": 3.21, "learning_rate": 5.073043042220592e-06, "loss": 0.0631, "step": 31995 }, { "epoch": 3.21, "learning_rate": 5.071711505095862e-06, "loss": 0.4873, "step": 31996 }, { "epoch": 3.21, "learning_rate": 5.07038012660528e-06, "loss": 0.1612, "step": 31997 }, { "epoch": 3.21, "learning_rate": 5.0690489067573155e-06, "loss": 0.2467, "step": 31998 }, { "epoch": 3.21, "learning_rate": 5.067717845560442e-06, "loss": 0.2379, "step": 31999 }, { "epoch": 3.21, "learning_rate": 5.066386943023125e-06, "loss": 0.0519, "step": 32000 }, { "epoch": 3.21, "learning_rate": 5.065056199153839e-06, "loss": 0.07, "step": 32001 }, { "epoch": 3.21, "learning_rate": 5.063725613961049e-06, "loss": 0.1543, "step": 32002 }, { "epoch": 3.21, "learning_rate": 5.062395187453221e-06, "loss": 0.2373, "step": 32003 }, { "epoch": 3.21, "learning_rate": 5.061064919638824e-06, "loss": 0.1114, "step": 32004 }, { "epoch": 3.21, "learning_rate": 5.059734810526316e-06, "loss": 0.0595, "step": 32005 }, { "epoch": 3.21, "learning_rate": 5.058404860124182e-06, "loss": 0.0745, "step": 32006 }, { "epoch": 3.21, "learning_rate": 5.057075068440863e-06, "loss": 0.1845, "step": 32007 }, { "epoch": 3.21, "learning_rate": 5.055745435484825e-06, "loss": 0.4437, "step": 32008 }, { "epoch": 3.21, "learning_rate": 5.054415961264545e-06, "loss": 0.0804, "step": 32009 }, { "epoch": 3.21, "learning_rate": 5.053086645788467e-06, "loss": 0.0365, "step": 32010 }, { "epoch": 3.21, "learning_rate": 5.051757489065046e-06, "loss": 0.4827, "step": 32011 }, { "epoch": 3.21, "learning_rate": 5.0504284911027596e-06, "loss": 0.2979, "step": 32012 }, { "epoch": 3.21, "learning_rate": 5.049099651910054e-06, "loss": 0.0573, "step": 32013 }, { "epoch": 3.21, "learning_rate": 5.0477709714953855e-06, "loss": 0.0027, "step": 32014 }, { "epoch": 3.21, "learning_rate": 5.046442449867212e-06, "loss": 0.4638, "step": 32015 }, { "epoch": 3.21, "learning_rate": 5.045114087033984e-06, "loss": 0.0156, "step": 32016 }, { "epoch": 3.21, "learning_rate": 5.043785883004157e-06, "loss": 0.6581, "step": 32017 }, { "epoch": 3.21, "learning_rate": 5.0424578377861845e-06, "loss": 0.0721, "step": 32018 }, { "epoch": 3.21, "learning_rate": 5.041129951388504e-06, "loss": 0.5497, "step": 32019 }, { "epoch": 3.21, "learning_rate": 5.039802223819598e-06, "loss": 0.0824, "step": 32020 }, { "epoch": 3.21, "learning_rate": 5.038474655087886e-06, "loss": 0.2371, "step": 32021 }, { "epoch": 3.21, "learning_rate": 5.037147245201816e-06, "loss": 0.1141, "step": 32022 }, { "epoch": 3.21, "learning_rate": 5.03581999416986e-06, "loss": 0.3829, "step": 32023 }, { "epoch": 3.21, "learning_rate": 5.0344929020004435e-06, "loss": 0.075, "step": 32024 }, { "epoch": 3.21, "learning_rate": 5.033165968702007e-06, "loss": 0.0987, "step": 32025 }, { "epoch": 3.22, "learning_rate": 5.0318391942830135e-06, "loss": 0.2473, "step": 32026 }, { "epoch": 3.22, "learning_rate": 5.030512578751906e-06, "loss": 0.0791, "step": 32027 }, { "epoch": 3.22, "learning_rate": 5.029186122117101e-06, "loss": 0.3254, "step": 32028 }, { "epoch": 3.22, "learning_rate": 5.027859824387062e-06, "loss": 0.1077, "step": 32029 }, { "epoch": 3.22, "learning_rate": 5.026533685570223e-06, "loss": 0.0632, "step": 32030 }, { "epoch": 3.22, "learning_rate": 5.025207705675026e-06, "loss": 0.1312, "step": 32031 }, { "epoch": 3.22, "learning_rate": 5.0238818847099045e-06, "loss": 0.4689, "step": 32032 }, { "epoch": 3.22, "learning_rate": 5.022556222683299e-06, "loss": 0.0688, "step": 32033 }, { "epoch": 3.22, "learning_rate": 5.021230719603642e-06, "loss": 0.1716, "step": 32034 }, { "epoch": 3.22, "learning_rate": 5.019905375479371e-06, "loss": 0.0949, "step": 32035 }, { "epoch": 3.22, "learning_rate": 5.018580190318919e-06, "loss": 0.1696, "step": 32036 }, { "epoch": 3.22, "learning_rate": 5.017255164130712e-06, "loss": 0.054, "step": 32037 }, { "epoch": 3.22, "learning_rate": 5.0159302969232036e-06, "loss": 0.1401, "step": 32038 }, { "epoch": 3.22, "learning_rate": 5.014605588704794e-06, "loss": 0.1347, "step": 32039 }, { "epoch": 3.22, "learning_rate": 5.013281039483939e-06, "loss": 0.1829, "step": 32040 }, { "epoch": 3.22, "learning_rate": 5.011956649269066e-06, "loss": 0.2343, "step": 32041 }, { "epoch": 3.22, "learning_rate": 5.010632418068578e-06, "loss": 0.3792, "step": 32042 }, { "epoch": 3.22, "learning_rate": 5.0093083458909285e-06, "loss": 0.4287, "step": 32043 }, { "epoch": 3.22, "learning_rate": 5.007984432744533e-06, "loss": 0.0761, "step": 32044 }, { "epoch": 3.22, "learning_rate": 5.006660678637816e-06, "loss": 0.1531, "step": 32045 }, { "epoch": 3.22, "learning_rate": 5.005337083579205e-06, "loss": 0.179, "step": 32046 }, { "epoch": 3.22, "learning_rate": 5.004013647577117e-06, "loss": 0.1893, "step": 32047 }, { "epoch": 3.22, "learning_rate": 5.002690370639981e-06, "loss": 0.2114, "step": 32048 }, { "epoch": 3.22, "learning_rate": 5.001367252776212e-06, "loss": 0.0573, "step": 32049 }, { "epoch": 3.22, "learning_rate": 5.000044293994231e-06, "loss": 0.0505, "step": 32050 }, { "epoch": 3.22, "learning_rate": 4.998721494302452e-06, "loss": 0.0597, "step": 32051 }, { "epoch": 3.22, "learning_rate": 4.997398853709309e-06, "loss": 0.1199, "step": 32052 }, { "epoch": 3.22, "learning_rate": 4.996076372223205e-06, "loss": 0.3461, "step": 32053 }, { "epoch": 3.22, "learning_rate": 4.994754049852551e-06, "loss": 0.1315, "step": 32054 }, { "epoch": 3.22, "learning_rate": 4.993431886605785e-06, "loss": 0.0965, "step": 32055 }, { "epoch": 3.22, "learning_rate": 4.992109882491296e-06, "loss": 0.0781, "step": 32056 }, { "epoch": 3.22, "learning_rate": 4.990788037517501e-06, "loss": 0.1074, "step": 32057 }, { "epoch": 3.22, "learning_rate": 4.989466351692823e-06, "loss": 0.229, "step": 32058 }, { "epoch": 3.22, "learning_rate": 4.988144825025675e-06, "loss": 0.4833, "step": 32059 }, { "epoch": 3.22, "learning_rate": 4.986823457524442e-06, "loss": 0.2371, "step": 32060 }, { "epoch": 3.22, "learning_rate": 4.9855022491975585e-06, "loss": 0.0757, "step": 32061 }, { "epoch": 3.22, "learning_rate": 4.9841812000534224e-06, "loss": 0.1336, "step": 32062 }, { "epoch": 3.22, "learning_rate": 4.98286031010044e-06, "loss": 0.0373, "step": 32063 }, { "epoch": 3.22, "learning_rate": 4.981539579347018e-06, "loss": 0.1438, "step": 32064 }, { "epoch": 3.22, "learning_rate": 4.9802190078015585e-06, "loss": 0.2621, "step": 32065 }, { "epoch": 3.22, "learning_rate": 4.97889859547247e-06, "loss": 0.1098, "step": 32066 }, { "epoch": 3.22, "learning_rate": 4.977578342368151e-06, "loss": 0.1395, "step": 32067 }, { "epoch": 3.22, "learning_rate": 4.976258248496995e-06, "loss": 0.0372, "step": 32068 }, { "epoch": 3.22, "learning_rate": 4.9749383138674235e-06, "loss": 0.2412, "step": 32069 }, { "epoch": 3.22, "learning_rate": 4.973618538487831e-06, "loss": 0.1352, "step": 32070 }, { "epoch": 3.22, "learning_rate": 4.9722989223665915e-06, "loss": 0.1037, "step": 32071 }, { "epoch": 3.22, "learning_rate": 4.970979465512132e-06, "loss": 0.1377, "step": 32072 }, { "epoch": 3.22, "learning_rate": 4.9696601679328425e-06, "loss": 0.1872, "step": 32073 }, { "epoch": 3.22, "learning_rate": 4.968341029637099e-06, "loss": 0.1703, "step": 32074 }, { "epoch": 3.22, "learning_rate": 4.967022050633318e-06, "loss": 0.0833, "step": 32075 }, { "epoch": 3.22, "learning_rate": 4.965703230929888e-06, "loss": 0.1641, "step": 32076 }, { "epoch": 3.22, "learning_rate": 4.964384570535196e-06, "loss": 0.1984, "step": 32077 }, { "epoch": 3.22, "learning_rate": 4.963066069457638e-06, "loss": 0.1268, "step": 32078 }, { "epoch": 3.22, "learning_rate": 4.961747727705606e-06, "loss": 0.6636, "step": 32079 }, { "epoch": 3.22, "learning_rate": 4.960429545287483e-06, "loss": 0.0551, "step": 32080 }, { "epoch": 3.22, "learning_rate": 4.959111522211659e-06, "loss": 0.1971, "step": 32081 }, { "epoch": 3.22, "learning_rate": 4.957793658486528e-06, "loss": 0.1381, "step": 32082 }, { "epoch": 3.22, "learning_rate": 4.95647595412046e-06, "loss": 0.1017, "step": 32083 }, { "epoch": 3.22, "learning_rate": 4.95515840912187e-06, "loss": 0.0856, "step": 32084 }, { "epoch": 3.22, "learning_rate": 4.9538410234991135e-06, "loss": 0.136, "step": 32085 }, { "epoch": 3.22, "learning_rate": 4.95252379726058e-06, "loss": 0.0899, "step": 32086 }, { "epoch": 3.22, "learning_rate": 4.9512067304146715e-06, "loss": 0.158, "step": 32087 }, { "epoch": 3.22, "learning_rate": 4.9498898229697455e-06, "loss": 0.4387, "step": 32088 }, { "epoch": 3.22, "learning_rate": 4.948573074934185e-06, "loss": 0.0734, "step": 32089 }, { "epoch": 3.22, "learning_rate": 4.94725648631638e-06, "loss": 0.4548, "step": 32090 }, { "epoch": 3.22, "learning_rate": 4.945940057124715e-06, "loss": 0.0355, "step": 32091 }, { "epoch": 3.22, "learning_rate": 4.944623787367541e-06, "loss": 0.5063, "step": 32092 }, { "epoch": 3.22, "learning_rate": 4.943307677053257e-06, "loss": 0.0738, "step": 32093 }, { "epoch": 3.22, "learning_rate": 4.9419917261902285e-06, "loss": 0.2379, "step": 32094 }, { "epoch": 3.22, "learning_rate": 4.9406759347868344e-06, "loss": 0.3375, "step": 32095 }, { "epoch": 3.22, "learning_rate": 4.939360302851443e-06, "loss": 0.0517, "step": 32096 }, { "epoch": 3.22, "learning_rate": 4.938044830392424e-06, "loss": 0.4127, "step": 32097 }, { "epoch": 3.22, "learning_rate": 4.936729517418171e-06, "loss": 0.2306, "step": 32098 }, { "epoch": 3.22, "learning_rate": 4.935414363937025e-06, "loss": 0.0955, "step": 32099 }, { "epoch": 3.22, "learning_rate": 4.934099369957362e-06, "loss": 0.2579, "step": 32100 }, { "epoch": 3.22, "learning_rate": 4.932784535487561e-06, "loss": 0.1101, "step": 32101 }, { "epoch": 3.22, "learning_rate": 4.931469860535992e-06, "loss": 0.1846, "step": 32102 }, { "epoch": 3.22, "learning_rate": 4.930155345110995e-06, "loss": 0.462, "step": 32103 }, { "epoch": 3.22, "learning_rate": 4.928840989220963e-06, "loss": 0.2995, "step": 32104 }, { "epoch": 3.22, "learning_rate": 4.927526792874253e-06, "loss": 0.1963, "step": 32105 }, { "epoch": 3.22, "learning_rate": 4.926212756079212e-06, "loss": 0.1116, "step": 32106 }, { "epoch": 3.22, "learning_rate": 4.924898878844222e-06, "loss": 0.1641, "step": 32107 }, { "epoch": 3.22, "learning_rate": 4.923585161177634e-06, "loss": 0.1372, "step": 32108 }, { "epoch": 3.22, "learning_rate": 4.922271603087814e-06, "loss": 0.3336, "step": 32109 }, { "epoch": 3.22, "learning_rate": 4.920958204583117e-06, "loss": 0.0133, "step": 32110 }, { "epoch": 3.22, "learning_rate": 4.919644965671898e-06, "loss": 0.0236, "step": 32111 }, { "epoch": 3.22, "learning_rate": 4.918331886362521e-06, "loss": 0.0634, "step": 32112 }, { "epoch": 3.22, "learning_rate": 4.917018966663338e-06, "loss": 0.3664, "step": 32113 }, { "epoch": 3.22, "learning_rate": 4.915706206582701e-06, "loss": 0.1885, "step": 32114 }, { "epoch": 3.22, "learning_rate": 4.914393606128961e-06, "loss": 0.2384, "step": 32115 }, { "epoch": 3.22, "learning_rate": 4.913081165310495e-06, "loss": 0.192, "step": 32116 }, { "epoch": 3.22, "learning_rate": 4.911768884135629e-06, "loss": 0.0715, "step": 32117 }, { "epoch": 3.22, "learning_rate": 4.910456762612712e-06, "loss": 0.5038, "step": 32118 }, { "epoch": 3.22, "learning_rate": 4.90914480075012e-06, "loss": 0.1546, "step": 32119 }, { "epoch": 3.22, "learning_rate": 4.907832998556169e-06, "loss": 0.1936, "step": 32120 }, { "epoch": 3.22, "learning_rate": 4.90652135603923e-06, "loss": 0.1893, "step": 32121 }, { "epoch": 3.22, "learning_rate": 4.905209873207646e-06, "loss": 0.2785, "step": 32122 }, { "epoch": 3.22, "learning_rate": 4.903898550069756e-06, "loss": 0.1745, "step": 32123 }, { "epoch": 3.23, "learning_rate": 4.902587386633912e-06, "loss": 0.4649, "step": 32124 }, { "epoch": 3.23, "learning_rate": 4.901276382908453e-06, "loss": 0.2075, "step": 32125 }, { "epoch": 3.23, "learning_rate": 4.899965538901725e-06, "loss": 0.4528, "step": 32126 }, { "epoch": 3.23, "learning_rate": 4.898654854622065e-06, "loss": 0.0641, "step": 32127 }, { "epoch": 3.23, "learning_rate": 4.897344330077818e-06, "loss": 0.2194, "step": 32128 }, { "epoch": 3.23, "learning_rate": 4.896033965277315e-06, "loss": 0.1618, "step": 32129 }, { "epoch": 3.23, "learning_rate": 4.894723760228916e-06, "loss": 0.2962, "step": 32130 }, { "epoch": 3.23, "learning_rate": 4.893413714940937e-06, "loss": 0.1519, "step": 32131 }, { "epoch": 3.23, "learning_rate": 4.892103829421714e-06, "loss": 0.279, "step": 32132 }, { "epoch": 3.23, "learning_rate": 4.8907941036795975e-06, "loss": 0.252, "step": 32133 }, { "epoch": 3.23, "learning_rate": 4.889484537722926e-06, "loss": 0.3194, "step": 32134 }, { "epoch": 3.23, "learning_rate": 4.888175131560005e-06, "loss": 0.5047, "step": 32135 }, { "epoch": 3.23, "learning_rate": 4.886865885199193e-06, "loss": 0.1264, "step": 32136 }, { "epoch": 3.23, "learning_rate": 4.885556798648819e-06, "loss": 0.3316, "step": 32137 }, { "epoch": 3.23, "learning_rate": 4.884247871917194e-06, "loss": 0.3286, "step": 32138 }, { "epoch": 3.23, "learning_rate": 4.882939105012666e-06, "loss": 0.0753, "step": 32139 }, { "epoch": 3.23, "learning_rate": 4.88163049794356e-06, "loss": 0.1454, "step": 32140 }, { "epoch": 3.23, "learning_rate": 4.8803220507182025e-06, "loss": 0.1262, "step": 32141 }, { "epoch": 3.23, "learning_rate": 4.87901376334492e-06, "loss": 0.0872, "step": 32142 }, { "epoch": 3.23, "learning_rate": 4.877705635832038e-06, "loss": 0.2124, "step": 32143 }, { "epoch": 3.23, "learning_rate": 4.876397668187879e-06, "loss": 0.087, "step": 32144 }, { "epoch": 3.23, "learning_rate": 4.875089860420767e-06, "loss": 0.5009, "step": 32145 }, { "epoch": 3.23, "learning_rate": 4.873782212539024e-06, "loss": 0.1848, "step": 32146 }, { "epoch": 3.23, "learning_rate": 4.872474724550969e-06, "loss": 0.2225, "step": 32147 }, { "epoch": 3.23, "learning_rate": 4.87116739646494e-06, "loss": 0.265, "step": 32148 }, { "epoch": 3.23, "learning_rate": 4.869860228289224e-06, "loss": 0.2534, "step": 32149 }, { "epoch": 3.23, "learning_rate": 4.868553220032168e-06, "loss": 0.1757, "step": 32150 }, { "epoch": 3.23, "learning_rate": 4.867246371702083e-06, "loss": 0.1297, "step": 32151 }, { "epoch": 3.23, "learning_rate": 4.86593968330727e-06, "loss": 0.8174, "step": 32152 }, { "epoch": 3.23, "learning_rate": 4.864633154856062e-06, "loss": 0.5488, "step": 32153 }, { "epoch": 3.23, "learning_rate": 4.863326786356765e-06, "loss": 0.1326, "step": 32154 }, { "epoch": 3.23, "learning_rate": 4.862020577817694e-06, "loss": 0.0665, "step": 32155 }, { "epoch": 3.23, "learning_rate": 4.860714529247161e-06, "loss": 0.104, "step": 32156 }, { "epoch": 3.23, "learning_rate": 4.85940864065348e-06, "loss": 0.2021, "step": 32157 }, { "epoch": 3.23, "learning_rate": 4.858102912044957e-06, "loss": 0.167, "step": 32158 }, { "epoch": 3.23, "learning_rate": 4.8567973434299e-06, "loss": 0.2485, "step": 32159 }, { "epoch": 3.23, "learning_rate": 4.855491934816624e-06, "loss": 0.1273, "step": 32160 }, { "epoch": 3.23, "learning_rate": 4.8541866862134195e-06, "loss": 0.3351, "step": 32161 }, { "epoch": 3.23, "learning_rate": 4.8528815976286244e-06, "loss": 0.1411, "step": 32162 }, { "epoch": 3.23, "learning_rate": 4.851576669070514e-06, "loss": 0.1013, "step": 32163 }, { "epoch": 3.23, "learning_rate": 4.850271900547396e-06, "loss": 0.3328, "step": 32164 }, { "epoch": 3.23, "learning_rate": 4.84896729206759e-06, "loss": 0.0871, "step": 32165 }, { "epoch": 3.23, "learning_rate": 4.847662843639392e-06, "loss": 0.0715, "step": 32166 }, { "epoch": 3.23, "learning_rate": 4.8463585552710865e-06, "loss": 0.1713, "step": 32167 }, { "epoch": 3.23, "learning_rate": 4.845054426970993e-06, "loss": 0.2696, "step": 32168 }, { "epoch": 3.23, "learning_rate": 4.8437504587474115e-06, "loss": 0.0278, "step": 32169 }, { "epoch": 3.23, "learning_rate": 4.842446650608619e-06, "loss": 0.4576, "step": 32170 }, { "epoch": 3.23, "learning_rate": 4.841143002562931e-06, "loss": 0.0762, "step": 32171 }, { "epoch": 3.23, "learning_rate": 4.839839514618636e-06, "loss": 0.1156, "step": 32172 }, { "epoch": 3.23, "learning_rate": 4.8385361867840324e-06, "loss": 0.1726, "step": 32173 }, { "epoch": 3.23, "learning_rate": 4.837233019067415e-06, "loss": 0.4766, "step": 32174 }, { "epoch": 3.23, "learning_rate": 4.835930011477064e-06, "loss": 0.4319, "step": 32175 }, { "epoch": 3.23, "learning_rate": 4.834627164021297e-06, "loss": 0.324, "step": 32176 }, { "epoch": 3.23, "learning_rate": 4.833324476708383e-06, "loss": 0.2181, "step": 32177 }, { "epoch": 3.23, "learning_rate": 4.832021949546611e-06, "loss": 0.0967, "step": 32178 }, { "epoch": 3.23, "learning_rate": 4.830719582544282e-06, "loss": 0.0981, "step": 32179 }, { "epoch": 3.23, "learning_rate": 4.829417375709688e-06, "loss": 0.3077, "step": 32180 }, { "epoch": 3.23, "learning_rate": 4.82811532905109e-06, "loss": 0.2961, "step": 32181 }, { "epoch": 3.23, "learning_rate": 4.826813442576802e-06, "loss": 0.1639, "step": 32182 }, { "epoch": 3.23, "learning_rate": 4.825511716295099e-06, "loss": 0.2791, "step": 32183 }, { "epoch": 3.23, "learning_rate": 4.824210150214249e-06, "loss": 0.2182, "step": 32184 }, { "epoch": 3.23, "learning_rate": 4.822908744342558e-06, "loss": 0.3491, "step": 32185 }, { "epoch": 3.23, "learning_rate": 4.821607498688296e-06, "loss": 0.0311, "step": 32186 }, { "epoch": 3.23, "learning_rate": 4.820306413259744e-06, "loss": 0.1242, "step": 32187 }, { "epoch": 3.23, "learning_rate": 4.819005488065188e-06, "loss": 0.1037, "step": 32188 }, { "epoch": 3.23, "learning_rate": 4.8177047231128976e-06, "loss": 0.0784, "step": 32189 }, { "epoch": 3.23, "learning_rate": 4.816404118411154e-06, "loss": 0.0749, "step": 32190 }, { "epoch": 3.23, "learning_rate": 4.8151036739682365e-06, "loss": 0.1133, "step": 32191 }, { "epoch": 3.23, "learning_rate": 4.813803389792419e-06, "loss": 0.1192, "step": 32192 }, { "epoch": 3.23, "learning_rate": 4.812503265891967e-06, "loss": 0.1179, "step": 32193 }, { "epoch": 3.23, "learning_rate": 4.811203302275177e-06, "loss": 0.1751, "step": 32194 }, { "epoch": 3.23, "learning_rate": 4.809903498950299e-06, "loss": 0.1857, "step": 32195 }, { "epoch": 3.23, "learning_rate": 4.808603855925606e-06, "loss": 0.1081, "step": 32196 }, { "epoch": 3.23, "learning_rate": 4.80730437320938e-06, "loss": 0.0716, "step": 32197 }, { "epoch": 3.23, "learning_rate": 4.806005050809894e-06, "loss": 0.2843, "step": 32198 }, { "epoch": 3.23, "learning_rate": 4.804705888735394e-06, "loss": 0.069, "step": 32199 }, { "epoch": 3.23, "learning_rate": 4.803406886994168e-06, "loss": 0.1854, "step": 32200 }, { "epoch": 3.23, "learning_rate": 4.802108045594472e-06, "loss": 0.2621, "step": 32201 }, { "epoch": 3.23, "learning_rate": 4.800809364544579e-06, "loss": 0.2429, "step": 32202 }, { "epoch": 3.23, "learning_rate": 4.799510843852747e-06, "loss": 0.3843, "step": 32203 }, { "epoch": 3.23, "learning_rate": 4.798212483527241e-06, "loss": 0.1601, "step": 32204 }, { "epoch": 3.23, "learning_rate": 4.796914283576322e-06, "loss": 0.1267, "step": 32205 }, { "epoch": 3.23, "learning_rate": 4.795616244008257e-06, "loss": 0.2796, "step": 32206 }, { "epoch": 3.23, "learning_rate": 4.794318364831293e-06, "loss": 0.1489, "step": 32207 }, { "epoch": 3.23, "learning_rate": 4.7930206460537115e-06, "loss": 0.3704, "step": 32208 }, { "epoch": 3.23, "learning_rate": 4.791723087683751e-06, "loss": 0.264, "step": 32209 }, { "epoch": 3.23, "learning_rate": 4.790425689729665e-06, "loss": 0.0489, "step": 32210 }, { "epoch": 3.23, "learning_rate": 4.789128452199733e-06, "loss": 0.4626, "step": 32211 }, { "epoch": 3.23, "learning_rate": 4.787831375102197e-06, "loss": 0.1493, "step": 32212 }, { "epoch": 3.23, "learning_rate": 4.7865344584453e-06, "loss": 0.3673, "step": 32213 }, { "epoch": 3.23, "learning_rate": 4.7852377022373125e-06, "loss": 0.2958, "step": 32214 }, { "epoch": 3.23, "learning_rate": 4.78394110648649e-06, "loss": 0.3138, "step": 32215 }, { "epoch": 3.23, "learning_rate": 4.782644671201058e-06, "loss": 0.4881, "step": 32216 }, { "epoch": 3.23, "learning_rate": 4.781348396389288e-06, "loss": 0.1129, "step": 32217 }, { "epoch": 3.23, "learning_rate": 4.780052282059422e-06, "loss": 0.0556, "step": 32218 }, { "epoch": 3.23, "learning_rate": 4.7787563282197134e-06, "loss": 0.0161, "step": 32219 }, { "epoch": 3.23, "learning_rate": 4.777460534878399e-06, "loss": 0.2181, "step": 32220 }, { "epoch": 3.23, "learning_rate": 4.776164902043737e-06, "loss": 0.2782, "step": 32221 }, { "epoch": 3.23, "learning_rate": 4.774869429723962e-06, "loss": 0.1997, "step": 32222 }, { "epoch": 3.24, "learning_rate": 4.773574117927323e-06, "loss": 0.2636, "step": 32223 }, { "epoch": 3.24, "learning_rate": 4.7722789666620615e-06, "loss": 0.058, "step": 32224 }, { "epoch": 3.24, "learning_rate": 4.770983975936412e-06, "loss": 0.2478, "step": 32225 }, { "epoch": 3.24, "learning_rate": 4.769689145758637e-06, "loss": 0.1345, "step": 32226 }, { "epoch": 3.24, "learning_rate": 4.768394476136953e-06, "loss": 0.305, "step": 32227 }, { "epoch": 3.24, "learning_rate": 4.767099967079606e-06, "loss": 0.1073, "step": 32228 }, { "epoch": 3.24, "learning_rate": 4.76580561859484e-06, "loss": 0.3077, "step": 32229 }, { "epoch": 3.24, "learning_rate": 4.764511430690886e-06, "loss": 0.1836, "step": 32230 }, { "epoch": 3.24, "learning_rate": 4.763217403375984e-06, "loss": 0.1859, "step": 32231 }, { "epoch": 3.24, "learning_rate": 4.761923536658364e-06, "loss": 0.0843, "step": 32232 }, { "epoch": 3.24, "learning_rate": 4.760629830546262e-06, "loss": 0.043, "step": 32233 }, { "epoch": 3.24, "learning_rate": 4.7593362850479104e-06, "loss": 0.3172, "step": 32234 }, { "epoch": 3.24, "learning_rate": 4.7580429001715405e-06, "loss": 0.1486, "step": 32235 }, { "epoch": 3.24, "learning_rate": 4.756749675925384e-06, "loss": 0.1443, "step": 32236 }, { "epoch": 3.24, "learning_rate": 4.755456612317669e-06, "loss": 0.4354, "step": 32237 }, { "epoch": 3.24, "learning_rate": 4.7541637093566234e-06, "loss": 0.2827, "step": 32238 }, { "epoch": 3.24, "learning_rate": 4.752870967050468e-06, "loss": 0.1415, "step": 32239 }, { "epoch": 3.24, "learning_rate": 4.751578385407454e-06, "loss": 0.1545, "step": 32240 }, { "epoch": 3.24, "learning_rate": 4.7502859644357845e-06, "loss": 0.1503, "step": 32241 }, { "epoch": 3.24, "learning_rate": 4.748993704143677e-06, "loss": 0.2145, "step": 32242 }, { "epoch": 3.24, "learning_rate": 4.747701604539379e-06, "loss": 0.1191, "step": 32243 }, { "epoch": 3.24, "learning_rate": 4.746409665631109e-06, "loss": 0.0522, "step": 32244 }, { "epoch": 3.24, "learning_rate": 4.745117887427064e-06, "loss": 0.1879, "step": 32245 }, { "epoch": 3.24, "learning_rate": 4.743826269935489e-06, "loss": 0.583, "step": 32246 }, { "epoch": 3.24, "learning_rate": 4.7425348131646055e-06, "loss": 0.1, "step": 32247 }, { "epoch": 3.24, "learning_rate": 4.7412435171226044e-06, "loss": 0.0987, "step": 32248 }, { "epoch": 3.24, "learning_rate": 4.73995238181773e-06, "loss": 0.0619, "step": 32249 }, { "epoch": 3.24, "learning_rate": 4.738661407258187e-06, "loss": 0.1658, "step": 32250 }, { "epoch": 3.24, "learning_rate": 4.737370593452197e-06, "loss": 0.1099, "step": 32251 }, { "epoch": 3.24, "learning_rate": 4.736079940407965e-06, "loss": 0.0269, "step": 32252 }, { "epoch": 3.24, "learning_rate": 4.7347894481337124e-06, "loss": 0.0997, "step": 32253 }, { "epoch": 3.24, "learning_rate": 4.733499116637648e-06, "loss": 0.0652, "step": 32254 }, { "epoch": 3.24, "learning_rate": 4.732208945927985e-06, "loss": 0.2488, "step": 32255 }, { "epoch": 3.24, "learning_rate": 4.730918936012922e-06, "loss": 0.414, "step": 32256 }, { "epoch": 3.24, "learning_rate": 4.729629086900684e-06, "loss": 0.4615, "step": 32257 }, { "epoch": 3.24, "learning_rate": 4.7283393985994834e-06, "loss": 0.3043, "step": 32258 }, { "epoch": 3.24, "learning_rate": 4.727049871117501e-06, "loss": 0.1648, "step": 32259 }, { "epoch": 3.24, "learning_rate": 4.725760504462967e-06, "loss": 0.1838, "step": 32260 }, { "epoch": 3.24, "learning_rate": 4.7244712986440806e-06, "loss": 0.3936, "step": 32261 }, { "epoch": 3.24, "learning_rate": 4.72318225366904e-06, "loss": 0.5177, "step": 32262 }, { "epoch": 3.24, "learning_rate": 4.721893369546053e-06, "loss": 0.2746, "step": 32263 }, { "epoch": 3.24, "learning_rate": 4.720604646283318e-06, "loss": 0.4541, "step": 32264 }, { "epoch": 3.24, "learning_rate": 4.719316083889042e-06, "loss": 0.188, "step": 32265 }, { "epoch": 3.24, "learning_rate": 4.718027682371419e-06, "loss": 0.2565, "step": 32266 }, { "epoch": 3.24, "learning_rate": 4.71673944173865e-06, "loss": 0.0547, "step": 32267 }, { "epoch": 3.24, "learning_rate": 4.715451361998934e-06, "loss": 0.149, "step": 32268 }, { "epoch": 3.24, "learning_rate": 4.714163443160466e-06, "loss": 0.5744, "step": 32269 }, { "epoch": 3.24, "learning_rate": 4.712875685231445e-06, "loss": 0.1744, "step": 32270 }, { "epoch": 3.24, "learning_rate": 4.711588088220052e-06, "loss": 0.19, "step": 32271 }, { "epoch": 3.24, "learning_rate": 4.7103006521345095e-06, "loss": 0.257, "step": 32272 }, { "epoch": 3.24, "learning_rate": 4.7090133769829855e-06, "loss": 0.1553, "step": 32273 }, { "epoch": 3.24, "learning_rate": 4.707726262773672e-06, "loss": 0.3002, "step": 32274 }, { "epoch": 3.24, "learning_rate": 4.706439309514774e-06, "loss": 0.2642, "step": 32275 }, { "epoch": 3.24, "learning_rate": 4.705152517214483e-06, "loss": 0.0532, "step": 32276 }, { "epoch": 3.24, "learning_rate": 4.7038658858809615e-06, "loss": 0.0907, "step": 32277 }, { "epoch": 3.24, "learning_rate": 4.70257941552242e-06, "loss": 0.2269, "step": 32278 }, { "epoch": 3.24, "learning_rate": 4.701293106147049e-06, "loss": 0.3752, "step": 32279 }, { "epoch": 3.24, "learning_rate": 4.700006957763011e-06, "loss": 0.2354, "step": 32280 }, { "epoch": 3.24, "learning_rate": 4.69872097037851e-06, "loss": 0.2186, "step": 32281 }, { "epoch": 3.24, "learning_rate": 4.697435144001723e-06, "loss": 0.2277, "step": 32282 }, { "epoch": 3.24, "learning_rate": 4.696149478640838e-06, "loss": 0.3124, "step": 32283 }, { "epoch": 3.24, "learning_rate": 4.694863974304025e-06, "loss": 0.0793, "step": 32284 }, { "epoch": 3.24, "learning_rate": 4.6935786309994676e-06, "loss": 0.2567, "step": 32285 }, { "epoch": 3.24, "learning_rate": 4.692293448735364e-06, "loss": 0.3749, "step": 32286 }, { "epoch": 3.24, "learning_rate": 4.691008427519868e-06, "loss": 0.4263, "step": 32287 }, { "epoch": 3.24, "learning_rate": 4.68972356736116e-06, "loss": 0.2427, "step": 32288 }, { "epoch": 3.24, "learning_rate": 4.688438868267428e-06, "loss": 0.3359, "step": 32289 }, { "epoch": 3.24, "learning_rate": 4.6871543302468524e-06, "loss": 0.4072, "step": 32290 }, { "epoch": 3.24, "learning_rate": 4.685869953307579e-06, "loss": 0.3626, "step": 32291 }, { "epoch": 3.24, "learning_rate": 4.6845857374578106e-06, "loss": 0.3795, "step": 32292 }, { "epoch": 3.24, "learning_rate": 4.683301682705704e-06, "loss": 0.0996, "step": 32293 }, { "epoch": 3.24, "learning_rate": 4.682017789059437e-06, "loss": 0.233, "step": 32294 }, { "epoch": 3.24, "learning_rate": 4.680734056527177e-06, "loss": 0.0935, "step": 32295 }, { "epoch": 3.24, "learning_rate": 4.6794504851170935e-06, "loss": 0.2112, "step": 32296 }, { "epoch": 3.24, "learning_rate": 4.678167074837356e-06, "loss": 0.0299, "step": 32297 }, { "epoch": 3.24, "learning_rate": 4.676883825696127e-06, "loss": 0.1406, "step": 32298 }, { "epoch": 3.24, "learning_rate": 4.675600737701579e-06, "loss": 0.1946, "step": 32299 }, { "epoch": 3.24, "learning_rate": 4.6743178108618724e-06, "loss": 0.0631, "step": 32300 }, { "epoch": 3.24, "learning_rate": 4.673035045185176e-06, "loss": 0.2799, "step": 32301 }, { "epoch": 3.24, "learning_rate": 4.671752440679645e-06, "loss": 0.1672, "step": 32302 }, { "epoch": 3.24, "learning_rate": 4.670469997353441e-06, "loss": 0.347, "step": 32303 }, { "epoch": 3.24, "learning_rate": 4.669187715214745e-06, "loss": 0.4414, "step": 32304 }, { "epoch": 3.24, "learning_rate": 4.667905594271694e-06, "loss": 0.4707, "step": 32305 }, { "epoch": 3.24, "learning_rate": 4.666623634532448e-06, "loss": 0.1471, "step": 32306 }, { "epoch": 3.24, "learning_rate": 4.665341836005178e-06, "loss": 0.1459, "step": 32307 }, { "epoch": 3.24, "learning_rate": 4.664060198698042e-06, "loss": 0.147, "step": 32308 }, { "epoch": 3.24, "learning_rate": 4.662778722619178e-06, "loss": 0.0564, "step": 32309 }, { "epoch": 3.24, "learning_rate": 4.661497407776757e-06, "loss": 0.1959, "step": 32310 }, { "epoch": 3.24, "learning_rate": 4.660216254178924e-06, "loss": 0.3093, "step": 32311 }, { "epoch": 3.24, "learning_rate": 4.658935261833839e-06, "loss": 0.1326, "step": 32312 }, { "epoch": 3.24, "learning_rate": 4.657654430749648e-06, "loss": 0.3605, "step": 32313 }, { "epoch": 3.24, "learning_rate": 4.656373760934502e-06, "loss": 0.1618, "step": 32314 }, { "epoch": 3.24, "learning_rate": 4.655093252396555e-06, "loss": 0.1937, "step": 32315 }, { "epoch": 3.24, "learning_rate": 4.653812905143949e-06, "loss": 0.1387, "step": 32316 }, { "epoch": 3.24, "learning_rate": 4.6525327191848325e-06, "loss": 0.2938, "step": 32317 }, { "epoch": 3.24, "learning_rate": 4.651252694527367e-06, "loss": 0.8607, "step": 32318 }, { "epoch": 3.24, "learning_rate": 4.649972831179681e-06, "loss": 0.2771, "step": 32319 }, { "epoch": 3.24, "learning_rate": 4.648693129149914e-06, "loss": 0.2554, "step": 32320 }, { "epoch": 3.25, "learning_rate": 4.647413588446229e-06, "loss": 0.0212, "step": 32321 }, { "epoch": 3.25, "learning_rate": 4.646134209076767e-06, "loss": 0.2082, "step": 32322 }, { "epoch": 3.25, "learning_rate": 4.644854991049647e-06, "loss": 1.3222, "step": 32323 }, { "epoch": 3.25, "learning_rate": 4.643575934373031e-06, "loss": 0.2049, "step": 32324 }, { "epoch": 3.25, "learning_rate": 4.642297039055052e-06, "loss": 0.2126, "step": 32325 }, { "epoch": 3.25, "learning_rate": 4.641018305103848e-06, "loss": 0.263, "step": 32326 }, { "epoch": 3.25, "learning_rate": 4.6397397325275545e-06, "loss": 0.161, "step": 32327 }, { "epoch": 3.25, "learning_rate": 4.6384613213343115e-06, "loss": 0.2142, "step": 32328 }, { "epoch": 3.25, "learning_rate": 4.637183071532253e-06, "loss": 0.1575, "step": 32329 }, { "epoch": 3.25, "learning_rate": 4.635904983129514e-06, "loss": 0.2059, "step": 32330 }, { "epoch": 3.25, "learning_rate": 4.634627056134226e-06, "loss": 0.107, "step": 32331 }, { "epoch": 3.25, "learning_rate": 4.6333492905545195e-06, "loss": 0.0989, "step": 32332 }, { "epoch": 3.25, "learning_rate": 4.63207168639853e-06, "loss": 0.0645, "step": 32333 }, { "epoch": 3.25, "learning_rate": 4.630794243674389e-06, "loss": 0.0641, "step": 32334 }, { "epoch": 3.25, "learning_rate": 4.629516962390209e-06, "loss": 0.2196, "step": 32335 }, { "epoch": 3.25, "learning_rate": 4.6282398425541505e-06, "loss": 0.1068, "step": 32336 }, { "epoch": 3.25, "learning_rate": 4.626962884174307e-06, "loss": 0.1763, "step": 32337 }, { "epoch": 3.25, "learning_rate": 4.625686087258826e-06, "loss": 0.5538, "step": 32338 }, { "epoch": 3.25, "learning_rate": 4.624409451815823e-06, "loss": 0.285, "step": 32339 }, { "epoch": 3.25, "learning_rate": 4.623132977853429e-06, "loss": 0.1624, "step": 32340 }, { "epoch": 3.25, "learning_rate": 4.621856665379761e-06, "loss": 0.5244, "step": 32341 }, { "epoch": 3.25, "learning_rate": 4.620580514402943e-06, "loss": 0.121, "step": 32342 }, { "epoch": 3.25, "learning_rate": 4.619304524931096e-06, "loss": 0.1171, "step": 32343 }, { "epoch": 3.25, "learning_rate": 4.618028696972339e-06, "loss": 0.198, "step": 32344 }, { "epoch": 3.25, "learning_rate": 4.6167530305347925e-06, "loss": 0.1852, "step": 32345 }, { "epoch": 3.25, "learning_rate": 4.615477525626572e-06, "loss": 0.0643, "step": 32346 }, { "epoch": 3.25, "learning_rate": 4.614202182255796e-06, "loss": 0.143, "step": 32347 }, { "epoch": 3.25, "learning_rate": 4.6129270004305824e-06, "loss": 0.0334, "step": 32348 }, { "epoch": 3.25, "learning_rate": 4.611651980159033e-06, "loss": 0.1685, "step": 32349 }, { "epoch": 3.25, "learning_rate": 4.610377121449288e-06, "loss": 0.1447, "step": 32350 }, { "epoch": 3.25, "learning_rate": 4.609102424309439e-06, "loss": 0.1587, "step": 32351 }, { "epoch": 3.25, "learning_rate": 4.607827888747592e-06, "loss": 0.1853, "step": 32352 }, { "epoch": 3.25, "learning_rate": 4.60655351477188e-06, "loss": 0.192, "step": 32353 }, { "epoch": 3.25, "learning_rate": 4.6052793023904045e-06, "loss": 0.1284, "step": 32354 }, { "epoch": 3.25, "learning_rate": 4.604005251611262e-06, "loss": 0.0539, "step": 32355 }, { "epoch": 3.25, "learning_rate": 4.6027313624425705e-06, "loss": 0.389, "step": 32356 }, { "epoch": 3.25, "learning_rate": 4.6014576348924375e-06, "loss": 0.0614, "step": 32357 }, { "epoch": 3.25, "learning_rate": 4.6001840689689666e-06, "loss": 0.1597, "step": 32358 }, { "epoch": 3.25, "learning_rate": 4.598910664680261e-06, "loss": 0.0915, "step": 32359 }, { "epoch": 3.25, "learning_rate": 4.597637422034422e-06, "loss": 0.1068, "step": 32360 }, { "epoch": 3.25, "learning_rate": 4.596364341039558e-06, "loss": 0.0383, "step": 32361 }, { "epoch": 3.25, "learning_rate": 4.595091421703768e-06, "loss": 0.1086, "step": 32362 }, { "epoch": 3.25, "learning_rate": 4.59381866403515e-06, "loss": 0.2599, "step": 32363 }, { "epoch": 3.25, "learning_rate": 4.592546068041806e-06, "loss": 0.0977, "step": 32364 }, { "epoch": 3.25, "learning_rate": 4.591273633731834e-06, "loss": 0.052, "step": 32365 }, { "epoch": 3.25, "learning_rate": 4.59000136111332e-06, "loss": 0.1611, "step": 32366 }, { "epoch": 3.25, "learning_rate": 4.588729250194382e-06, "loss": 0.116, "step": 32367 }, { "epoch": 3.25, "learning_rate": 4.587457300983109e-06, "loss": 0.0624, "step": 32368 }, { "epoch": 3.25, "learning_rate": 4.586185513487576e-06, "loss": 0.1012, "step": 32369 }, { "epoch": 3.25, "learning_rate": 4.584913887715898e-06, "loss": 0.1812, "step": 32370 }, { "epoch": 3.25, "learning_rate": 4.583642423676161e-06, "loss": 0.385, "step": 32371 }, { "epoch": 3.25, "learning_rate": 4.582371121376452e-06, "loss": 0.2347, "step": 32372 }, { "epoch": 3.25, "learning_rate": 4.581099980824867e-06, "loss": 0.3729, "step": 32373 }, { "epoch": 3.25, "learning_rate": 4.579829002029488e-06, "loss": 0.3481, "step": 32374 }, { "epoch": 3.25, "learning_rate": 4.578558184998412e-06, "loss": 0.1431, "step": 32375 }, { "epoch": 3.25, "learning_rate": 4.577287529739717e-06, "loss": 0.3898, "step": 32376 }, { "epoch": 3.25, "learning_rate": 4.576017036261494e-06, "loss": 0.2361, "step": 32377 }, { "epoch": 3.25, "learning_rate": 4.574746704571827e-06, "loss": 0.0834, "step": 32378 }, { "epoch": 3.25, "learning_rate": 4.5734765346788e-06, "loss": 0.583, "step": 32379 }, { "epoch": 3.25, "learning_rate": 4.572206526590496e-06, "loss": 0.2768, "step": 32380 }, { "epoch": 3.25, "learning_rate": 4.570936680314988e-06, "loss": 0.0824, "step": 32381 }, { "epoch": 3.25, "learning_rate": 4.56966699586038e-06, "loss": 0.1303, "step": 32382 }, { "epoch": 3.25, "learning_rate": 4.568397473234729e-06, "loss": 0.0982, "step": 32383 }, { "epoch": 3.25, "learning_rate": 4.567128112446115e-06, "loss": 0.4657, "step": 32384 }, { "epoch": 3.25, "learning_rate": 4.565858913502632e-06, "loss": 0.2445, "step": 32385 }, { "epoch": 3.25, "learning_rate": 4.564589876412353e-06, "loss": 0.4124, "step": 32386 }, { "epoch": 3.25, "learning_rate": 4.563321001183331e-06, "loss": 0.1489, "step": 32387 }, { "epoch": 3.25, "learning_rate": 4.5620522878236685e-06, "loss": 0.1549, "step": 32388 }, { "epoch": 3.25, "learning_rate": 4.560783736341426e-06, "loss": 0.2247, "step": 32389 }, { "epoch": 3.25, "learning_rate": 4.559515346744679e-06, "loss": 0.0888, "step": 32390 }, { "epoch": 3.25, "learning_rate": 4.558247119041498e-06, "loss": 0.3524, "step": 32391 }, { "epoch": 3.25, "learning_rate": 4.556979053239952e-06, "loss": 0.079, "step": 32392 }, { "epoch": 3.25, "learning_rate": 4.555711149348114e-06, "loss": 0.5196, "step": 32393 }, { "epoch": 3.25, "learning_rate": 4.554443407374053e-06, "loss": 0.3362, "step": 32394 }, { "epoch": 3.25, "learning_rate": 4.553175827325824e-06, "loss": 0.1356, "step": 32395 }, { "epoch": 3.25, "learning_rate": 4.551908409211518e-06, "loss": 0.5093, "step": 32396 }, { "epoch": 3.25, "learning_rate": 4.550641153039179e-06, "loss": 0.3512, "step": 32397 }, { "epoch": 3.25, "learning_rate": 4.549374058816873e-06, "loss": 0.1191, "step": 32398 }, { "epoch": 3.25, "learning_rate": 4.548107126552674e-06, "loss": 0.1362, "step": 32399 }, { "epoch": 3.25, "learning_rate": 4.546840356254646e-06, "loss": 0.138, "step": 32400 }, { "epoch": 3.25, "learning_rate": 4.545573747930828e-06, "loss": 0.161, "step": 32401 }, { "epoch": 3.25, "learning_rate": 4.5443073015893045e-06, "loss": 0.1389, "step": 32402 }, { "epoch": 3.25, "learning_rate": 4.543041017238125e-06, "loss": 0.0966, "step": 32403 }, { "epoch": 3.25, "learning_rate": 4.541774894885348e-06, "loss": 0.0571, "step": 32404 }, { "epoch": 3.25, "learning_rate": 4.540508934539032e-06, "loss": 0.0419, "step": 32405 }, { "epoch": 3.25, "learning_rate": 4.539243136207227e-06, "loss": 0.0843, "step": 32406 }, { "epoch": 3.25, "learning_rate": 4.537977499897998e-06, "loss": 0.3352, "step": 32407 }, { "epoch": 3.25, "learning_rate": 4.536712025619392e-06, "loss": 0.0175, "step": 32408 }, { "epoch": 3.25, "learning_rate": 4.535446713379464e-06, "loss": 0.2486, "step": 32409 }, { "epoch": 3.25, "learning_rate": 4.534181563186266e-06, "loss": 0.0615, "step": 32410 }, { "epoch": 3.25, "learning_rate": 4.532916575047847e-06, "loss": 0.2688, "step": 32411 }, { "epoch": 3.25, "learning_rate": 4.531651748972262e-06, "loss": 0.155, "step": 32412 }, { "epoch": 3.25, "learning_rate": 4.53038708496755e-06, "loss": 0.2865, "step": 32413 }, { "epoch": 3.25, "learning_rate": 4.529122583041775e-06, "loss": 0.1756, "step": 32414 }, { "epoch": 3.25, "learning_rate": 4.52785824320297e-06, "loss": 0.0858, "step": 32415 }, { "epoch": 3.25, "learning_rate": 4.52659406545918e-06, "loss": 0.0104, "step": 32416 }, { "epoch": 3.25, "learning_rate": 4.5253300498184606e-06, "loss": 0.0652, "step": 32417 }, { "epoch": 3.25, "learning_rate": 4.524066196288849e-06, "loss": 0.2013, "step": 32418 }, { "epoch": 3.25, "learning_rate": 4.522802504878388e-06, "loss": 0.3076, "step": 32419 }, { "epoch": 3.26, "learning_rate": 4.52153897559512e-06, "loss": 0.0275, "step": 32420 }, { "epoch": 3.26, "learning_rate": 4.520275608447086e-06, "loss": 0.3032, "step": 32421 }, { "epoch": 3.26, "learning_rate": 4.519012403442325e-06, "loss": 0.1427, "step": 32422 }, { "epoch": 3.26, "learning_rate": 4.517749360588873e-06, "loss": 0.069, "step": 32423 }, { "epoch": 3.26, "learning_rate": 4.516486479894774e-06, "loss": 0.3341, "step": 32424 }, { "epoch": 3.26, "learning_rate": 4.51522376136806e-06, "loss": 0.0793, "step": 32425 }, { "epoch": 3.26, "learning_rate": 4.513961205016766e-06, "loss": 0.3495, "step": 32426 }, { "epoch": 3.26, "learning_rate": 4.51269881084892e-06, "loss": 0.2199, "step": 32427 }, { "epoch": 3.26, "learning_rate": 4.511436578872576e-06, "loss": 0.1329, "step": 32428 }, { "epoch": 3.26, "learning_rate": 4.5101745090957505e-06, "loss": 0.2031, "step": 32429 }, { "epoch": 3.26, "learning_rate": 4.508912601526468e-06, "loss": 0.1848, "step": 32430 }, { "epoch": 3.26, "learning_rate": 4.507650856172773e-06, "loss": 0.0409, "step": 32431 }, { "epoch": 3.26, "learning_rate": 4.506389273042701e-06, "loss": 0.063, "step": 32432 }, { "epoch": 3.26, "learning_rate": 4.505127852144255e-06, "loss": 0.0324, "step": 32433 }, { "epoch": 3.26, "learning_rate": 4.503866593485482e-06, "loss": 0.0768, "step": 32434 }, { "epoch": 3.26, "learning_rate": 4.502605497074403e-06, "loss": 0.0558, "step": 32435 }, { "epoch": 3.26, "learning_rate": 4.501344562919044e-06, "loss": 0.1029, "step": 32436 }, { "epoch": 3.26, "learning_rate": 4.5000837910274274e-06, "loss": 0.3482, "step": 32437 }, { "epoch": 3.26, "learning_rate": 4.498823181407576e-06, "loss": 0.3223, "step": 32438 }, { "epoch": 3.26, "learning_rate": 4.4975627340675116e-06, "loss": 0.1039, "step": 32439 }, { "epoch": 3.26, "learning_rate": 4.4963024490152605e-06, "loss": 0.2198, "step": 32440 }, { "epoch": 3.26, "learning_rate": 4.495042326258836e-06, "loss": 0.0661, "step": 32441 }, { "epoch": 3.26, "learning_rate": 4.493782365806258e-06, "loss": 0.5537, "step": 32442 }, { "epoch": 3.26, "learning_rate": 4.492522567665548e-06, "loss": 0.0419, "step": 32443 }, { "epoch": 3.26, "learning_rate": 4.491262931844722e-06, "loss": 0.0521, "step": 32444 }, { "epoch": 3.26, "learning_rate": 4.490003458351787e-06, "loss": 0.0547, "step": 32445 }, { "epoch": 3.26, "learning_rate": 4.488744147194776e-06, "loss": 0.1519, "step": 32446 }, { "epoch": 3.26, "learning_rate": 4.487484998381685e-06, "loss": 0.0246, "step": 32447 }, { "epoch": 3.26, "learning_rate": 4.486226011920534e-06, "loss": 0.2566, "step": 32448 }, { "epoch": 3.26, "learning_rate": 4.48496718781934e-06, "loss": 0.1209, "step": 32449 }, { "epoch": 3.26, "learning_rate": 4.483708526086104e-06, "loss": 0.2156, "step": 32450 }, { "epoch": 3.26, "learning_rate": 4.482450026728845e-06, "loss": 0.244, "step": 32451 }, { "epoch": 3.26, "learning_rate": 4.481191689755561e-06, "loss": 0.4258, "step": 32452 }, { "epoch": 3.26, "learning_rate": 4.479933515174268e-06, "loss": 0.0176, "step": 32453 }, { "epoch": 3.26, "learning_rate": 4.478675502992972e-06, "loss": 0.2599, "step": 32454 }, { "epoch": 3.26, "learning_rate": 4.477417653219672e-06, "loss": 0.0544, "step": 32455 }, { "epoch": 3.26, "learning_rate": 4.4761599658623794e-06, "loss": 0.4343, "step": 32456 }, { "epoch": 3.26, "learning_rate": 4.474902440929095e-06, "loss": 0.5131, "step": 32457 }, { "epoch": 3.26, "learning_rate": 4.473645078427821e-06, "loss": 0.1546, "step": 32458 }, { "epoch": 3.26, "learning_rate": 4.472387878366553e-06, "loss": 0.2521, "step": 32459 }, { "epoch": 3.26, "learning_rate": 4.471130840753309e-06, "loss": 0.1485, "step": 32460 }, { "epoch": 3.26, "learning_rate": 4.469873965596071e-06, "loss": 0.338, "step": 32461 }, { "epoch": 3.26, "learning_rate": 4.468617252902835e-06, "loss": 0.2501, "step": 32462 }, { "epoch": 3.26, "learning_rate": 4.467360702681614e-06, "loss": 0.5969, "step": 32463 }, { "epoch": 3.26, "learning_rate": 4.466104314940404e-06, "loss": 0.0799, "step": 32464 }, { "epoch": 3.26, "learning_rate": 4.4648480896871795e-06, "loss": 0.5563, "step": 32465 }, { "epoch": 3.26, "learning_rate": 4.463592026929951e-06, "loss": 0.234, "step": 32466 }, { "epoch": 3.26, "learning_rate": 4.46233612667671e-06, "loss": 0.3159, "step": 32467 }, { "epoch": 3.26, "learning_rate": 4.461080388935447e-06, "loss": 0.0655, "step": 32468 }, { "epoch": 3.26, "learning_rate": 4.459824813714154e-06, "loss": 0.2998, "step": 32469 }, { "epoch": 3.26, "learning_rate": 4.458569401020819e-06, "loss": 0.2096, "step": 32470 }, { "epoch": 3.26, "learning_rate": 4.457314150863432e-06, "loss": 0.3813, "step": 32471 }, { "epoch": 3.26, "learning_rate": 4.45605906324998e-06, "loss": 0.1168, "step": 32472 }, { "epoch": 3.26, "learning_rate": 4.454804138188447e-06, "loss": 0.2037, "step": 32473 }, { "epoch": 3.26, "learning_rate": 4.453549375686832e-06, "loss": 0.0472, "step": 32474 }, { "epoch": 3.26, "learning_rate": 4.452294775753106e-06, "loss": 0.3171, "step": 32475 }, { "epoch": 3.26, "learning_rate": 4.45104033839525e-06, "loss": 0.2593, "step": 32476 }, { "epoch": 3.26, "learning_rate": 4.449786063621261e-06, "loss": 0.0944, "step": 32477 }, { "epoch": 3.26, "learning_rate": 4.448531951439117e-06, "loss": 0.0872, "step": 32478 }, { "epoch": 3.26, "learning_rate": 4.447278001856785e-06, "loss": 0.0482, "step": 32479 }, { "epoch": 3.26, "learning_rate": 4.446024214882263e-06, "loss": 0.059, "step": 32480 }, { "epoch": 3.26, "learning_rate": 4.444770590523521e-06, "loss": 0.1773, "step": 32481 }, { "epoch": 3.26, "learning_rate": 4.4435171287885345e-06, "loss": 0.0551, "step": 32482 }, { "epoch": 3.26, "learning_rate": 4.442263829685286e-06, "loss": 0.1639, "step": 32483 }, { "epoch": 3.26, "learning_rate": 4.441010693221746e-06, "loss": 0.2848, "step": 32484 }, { "epoch": 3.26, "learning_rate": 4.439757719405894e-06, "loss": 0.1891, "step": 32485 }, { "epoch": 3.26, "learning_rate": 4.438504908245694e-06, "loss": 0.209, "step": 32486 }, { "epoch": 3.26, "learning_rate": 4.437252259749131e-06, "loss": 0.6885, "step": 32487 }, { "epoch": 3.26, "learning_rate": 4.435999773924167e-06, "loss": 0.0664, "step": 32488 }, { "epoch": 3.26, "learning_rate": 4.434747450778776e-06, "loss": 0.059, "step": 32489 }, { "epoch": 3.26, "learning_rate": 4.433495290320924e-06, "loss": 0.0549, "step": 32490 }, { "epoch": 3.26, "learning_rate": 4.4322432925585806e-06, "loss": 0.2159, "step": 32491 }, { "epoch": 3.26, "learning_rate": 4.430991457499724e-06, "loss": 0.0981, "step": 32492 }, { "epoch": 3.26, "learning_rate": 4.429739785152306e-06, "loss": 0.3789, "step": 32493 }, { "epoch": 3.26, "learning_rate": 4.428488275524288e-06, "loss": 0.2557, "step": 32494 }, { "epoch": 3.26, "learning_rate": 4.427236928623652e-06, "loss": 0.1605, "step": 32495 }, { "epoch": 3.26, "learning_rate": 4.425985744458355e-06, "loss": 0.0356, "step": 32496 }, { "epoch": 3.26, "learning_rate": 4.424734723036346e-06, "loss": 0.259, "step": 32497 }, { "epoch": 3.26, "learning_rate": 4.423483864365601e-06, "loss": 0.3382, "step": 32498 }, { "epoch": 3.26, "learning_rate": 4.422233168454074e-06, "loss": 0.0793, "step": 32499 }, { "epoch": 3.26, "learning_rate": 4.420982635309727e-06, "loss": 0.1578, "step": 32500 }, { "epoch": 3.26, "learning_rate": 4.4197322649405136e-06, "loss": 0.0354, "step": 32501 }, { "epoch": 3.26, "learning_rate": 4.418482057354394e-06, "loss": 0.3619, "step": 32502 }, { "epoch": 3.26, "learning_rate": 4.417232012559322e-06, "loss": 0.0344, "step": 32503 }, { "epoch": 3.26, "learning_rate": 4.415982130563253e-06, "loss": 0.0875, "step": 32504 }, { "epoch": 3.26, "learning_rate": 4.414732411374136e-06, "loss": 0.0399, "step": 32505 }, { "epoch": 3.26, "learning_rate": 4.413482854999944e-06, "loss": 0.5146, "step": 32506 }, { "epoch": 3.26, "learning_rate": 4.412233461448604e-06, "loss": 0.3687, "step": 32507 }, { "epoch": 3.26, "learning_rate": 4.410984230728069e-06, "loss": 0.1044, "step": 32508 }, { "epoch": 3.26, "learning_rate": 4.409735162846303e-06, "loss": 0.0447, "step": 32509 }, { "epoch": 3.26, "learning_rate": 4.408486257811254e-06, "loss": 0.3554, "step": 32510 }, { "epoch": 3.26, "learning_rate": 4.407237515630851e-06, "loss": 0.492, "step": 32511 }, { "epoch": 3.26, "learning_rate": 4.405988936313058e-06, "loss": 0.3678, "step": 32512 }, { "epoch": 3.26, "learning_rate": 4.4047405198658155e-06, "loss": 0.0321, "step": 32513 }, { "epoch": 3.26, "learning_rate": 4.403492266297066e-06, "loss": 0.0922, "step": 32514 }, { "epoch": 3.26, "learning_rate": 4.402244175614756e-06, "loss": 0.1099, "step": 32515 }, { "epoch": 3.26, "learning_rate": 4.400996247826823e-06, "loss": 0.1571, "step": 32516 }, { "epoch": 3.26, "learning_rate": 4.399748482941215e-06, "loss": 0.3851, "step": 32517 }, { "epoch": 3.27, "learning_rate": 4.398500880965869e-06, "loss": 0.0847, "step": 32518 }, { "epoch": 3.27, "learning_rate": 4.397253441908722e-06, "loss": 0.1706, "step": 32519 }, { "epoch": 3.27, "learning_rate": 4.396006165777714e-06, "loss": 0.2466, "step": 32520 }, { "epoch": 3.27, "learning_rate": 4.394759052580786e-06, "loss": 0.235, "step": 32521 }, { "epoch": 3.27, "learning_rate": 4.3935121023258676e-06, "loss": 0.0332, "step": 32522 }, { "epoch": 3.27, "learning_rate": 4.3922653150208895e-06, "loss": 0.3735, "step": 32523 }, { "epoch": 3.27, "learning_rate": 4.391018690673808e-06, "loss": 0.1569, "step": 32524 }, { "epoch": 3.27, "learning_rate": 4.3897722292925345e-06, "loss": 0.4035, "step": 32525 }, { "epoch": 3.27, "learning_rate": 4.388525930885002e-06, "loss": 0.2438, "step": 32526 }, { "epoch": 3.27, "learning_rate": 4.3872797954591545e-06, "loss": 0.1054, "step": 32527 }, { "epoch": 3.27, "learning_rate": 4.386033823022913e-06, "loss": 0.0836, "step": 32528 }, { "epoch": 3.27, "learning_rate": 4.384788013584209e-06, "loss": 0.1577, "step": 32529 }, { "epoch": 3.27, "learning_rate": 4.3835423671509714e-06, "loss": 0.0555, "step": 32530 }, { "epoch": 3.27, "learning_rate": 4.382296883731123e-06, "loss": 0.6113, "step": 32531 }, { "epoch": 3.27, "learning_rate": 4.381051563332591e-06, "loss": 0.0802, "step": 32532 }, { "epoch": 3.27, "learning_rate": 4.379806405963303e-06, "loss": 0.4263, "step": 32533 }, { "epoch": 3.27, "learning_rate": 4.378561411631178e-06, "loss": 0.1408, "step": 32534 }, { "epoch": 3.27, "learning_rate": 4.3773165803441436e-06, "loss": 0.0385, "step": 32535 }, { "epoch": 3.27, "learning_rate": 4.376071912110118e-06, "loss": 0.071, "step": 32536 }, { "epoch": 3.27, "learning_rate": 4.374827406937016e-06, "loss": 0.2003, "step": 32537 }, { "epoch": 3.27, "learning_rate": 4.37358306483278e-06, "loss": 0.1221, "step": 32538 }, { "epoch": 3.27, "learning_rate": 4.372338885805303e-06, "loss": 0.1439, "step": 32539 }, { "epoch": 3.27, "learning_rate": 4.371094869862508e-06, "loss": 0.1773, "step": 32540 }, { "epoch": 3.27, "learning_rate": 4.369851017012318e-06, "loss": 0.5675, "step": 32541 }, { "epoch": 3.27, "learning_rate": 4.368607327262656e-06, "loss": 0.5869, "step": 32542 }, { "epoch": 3.27, "learning_rate": 4.367363800621414e-06, "loss": 0.081, "step": 32543 }, { "epoch": 3.27, "learning_rate": 4.366120437096521e-06, "loss": 0.021, "step": 32544 }, { "epoch": 3.27, "learning_rate": 4.364877236695887e-06, "loss": 0.152, "step": 32545 }, { "epoch": 3.27, "learning_rate": 4.363634199427423e-06, "loss": 0.113, "step": 32546 }, { "epoch": 3.27, "learning_rate": 4.362391325299038e-06, "loss": 0.5303, "step": 32547 }, { "epoch": 3.27, "learning_rate": 4.361148614318641e-06, "loss": 1.093, "step": 32548 }, { "epoch": 3.27, "learning_rate": 4.359906066494142e-06, "loss": 0.1456, "step": 32549 }, { "epoch": 3.27, "learning_rate": 4.358663681833447e-06, "loss": 0.2729, "step": 32550 }, { "epoch": 3.27, "learning_rate": 4.357421460344461e-06, "loss": 0.0307, "step": 32551 }, { "epoch": 3.27, "learning_rate": 4.356179402035091e-06, "loss": 0.0478, "step": 32552 }, { "epoch": 3.27, "learning_rate": 4.354937506913238e-06, "loss": 0.5094, "step": 32553 }, { "epoch": 3.27, "learning_rate": 4.353695774986799e-06, "loss": 0.181, "step": 32554 }, { "epoch": 3.27, "learning_rate": 4.3524542062636944e-06, "loss": 0.1086, "step": 32555 }, { "epoch": 3.27, "learning_rate": 4.351212800751817e-06, "loss": 0.122, "step": 32556 }, { "epoch": 3.27, "learning_rate": 4.3499715584590514e-06, "loss": 0.2409, "step": 32557 }, { "epoch": 3.27, "learning_rate": 4.348730479393318e-06, "loss": 0.3792, "step": 32558 }, { "epoch": 3.27, "learning_rate": 4.347489563562503e-06, "loss": 0.3358, "step": 32559 }, { "epoch": 3.27, "learning_rate": 4.346248810974508e-06, "loss": 0.4635, "step": 32560 }, { "epoch": 3.27, "learning_rate": 4.345008221637225e-06, "loss": 0.1179, "step": 32561 }, { "epoch": 3.27, "learning_rate": 4.343767795558553e-06, "loss": 0.0463, "step": 32562 }, { "epoch": 3.27, "learning_rate": 4.342527532746381e-06, "loss": 0.3235, "step": 32563 }, { "epoch": 3.27, "learning_rate": 4.3412874332085995e-06, "loss": 0.1686, "step": 32564 }, { "epoch": 3.27, "learning_rate": 4.340047496953109e-06, "loss": 0.0723, "step": 32565 }, { "epoch": 3.27, "learning_rate": 4.3388077239877895e-06, "loss": 0.0908, "step": 32566 }, { "epoch": 3.27, "learning_rate": 4.337568114320539e-06, "loss": 0.1405, "step": 32567 }, { "epoch": 3.27, "learning_rate": 4.336328667959244e-06, "loss": 0.0567, "step": 32568 }, { "epoch": 3.27, "learning_rate": 4.335089384911781e-06, "loss": 0.146, "step": 32569 }, { "epoch": 3.27, "learning_rate": 4.333850265186058e-06, "loss": 0.5336, "step": 32570 }, { "epoch": 3.27, "learning_rate": 4.332611308789944e-06, "loss": 0.145, "step": 32571 }, { "epoch": 3.27, "learning_rate": 4.331372515731321e-06, "loss": 0.1371, "step": 32572 }, { "epoch": 3.27, "learning_rate": 4.330133886018081e-06, "loss": 0.2051, "step": 32573 }, { "epoch": 3.27, "learning_rate": 4.328895419658113e-06, "loss": 0.1644, "step": 32574 }, { "epoch": 3.27, "learning_rate": 4.327657116659276e-06, "loss": 0.0923, "step": 32575 }, { "epoch": 3.27, "learning_rate": 4.32641897702947e-06, "loss": 0.2956, "step": 32576 }, { "epoch": 3.27, "learning_rate": 4.325181000776567e-06, "loss": 0.1828, "step": 32577 }, { "epoch": 3.27, "learning_rate": 4.323943187908442e-06, "loss": 0.5088, "step": 32578 }, { "epoch": 3.27, "learning_rate": 4.3227055384329775e-06, "loss": 0.1894, "step": 32579 }, { "epoch": 3.27, "learning_rate": 4.321468052358045e-06, "loss": 0.0585, "step": 32580 }, { "epoch": 3.27, "learning_rate": 4.320230729691523e-06, "loss": 0.2663, "step": 32581 }, { "epoch": 3.27, "learning_rate": 4.318993570441282e-06, "loss": 0.2861, "step": 32582 }, { "epoch": 3.27, "learning_rate": 4.317756574615191e-06, "loss": 0.1991, "step": 32583 }, { "epoch": 3.27, "learning_rate": 4.3165197422211424e-06, "loss": 0.1045, "step": 32584 }, { "epoch": 3.27, "learning_rate": 4.315283073266981e-06, "loss": 0.1524, "step": 32585 }, { "epoch": 3.27, "learning_rate": 4.314046567760582e-06, "loss": 0.1298, "step": 32586 }, { "epoch": 3.27, "learning_rate": 4.312810225709826e-06, "loss": 0.1286, "step": 32587 }, { "epoch": 3.27, "learning_rate": 4.311574047122578e-06, "loss": 0.0822, "step": 32588 }, { "epoch": 3.27, "learning_rate": 4.310338032006691e-06, "loss": 0.1559, "step": 32589 }, { "epoch": 3.27, "learning_rate": 4.309102180370044e-06, "loss": 0.1079, "step": 32590 }, { "epoch": 3.27, "learning_rate": 4.3078664922205e-06, "loss": 0.2462, "step": 32591 }, { "epoch": 3.27, "learning_rate": 4.306630967565918e-06, "loss": 0.5773, "step": 32592 }, { "epoch": 3.27, "learning_rate": 4.305395606414163e-06, "loss": 0.1996, "step": 32593 }, { "epoch": 3.27, "learning_rate": 4.3041604087730935e-06, "loss": 0.3139, "step": 32594 }, { "epoch": 3.27, "learning_rate": 4.302925374650576e-06, "loss": 0.1519, "step": 32595 }, { "epoch": 3.27, "learning_rate": 4.301690504054461e-06, "loss": 0.3611, "step": 32596 }, { "epoch": 3.27, "learning_rate": 4.300455796992615e-06, "loss": 0.4463, "step": 32597 }, { "epoch": 3.27, "learning_rate": 4.299221253472889e-06, "loss": 0.0298, "step": 32598 }, { "epoch": 3.27, "learning_rate": 4.297986873503147e-06, "loss": 0.1347, "step": 32599 }, { "epoch": 3.27, "learning_rate": 4.296752657091235e-06, "loss": 0.174, "step": 32600 }, { "epoch": 3.27, "learning_rate": 4.295518604245004e-06, "loss": 0.2318, "step": 32601 }, { "epoch": 3.27, "learning_rate": 4.2942847149723285e-06, "loss": 0.1138, "step": 32602 }, { "epoch": 3.27, "learning_rate": 4.2930509892810375e-06, "loss": 0.1762, "step": 32603 }, { "epoch": 3.27, "learning_rate": 4.291817427178986e-06, "loss": 0.1978, "step": 32604 }, { "epoch": 3.27, "learning_rate": 4.290584028674036e-06, "loss": 1.3455, "step": 32605 }, { "epoch": 3.27, "learning_rate": 4.289350793774033e-06, "loss": 0.1348, "step": 32606 }, { "epoch": 3.27, "learning_rate": 4.2881177224868095e-06, "loss": 0.1089, "step": 32607 }, { "epoch": 3.27, "learning_rate": 4.28688481482023e-06, "loss": 0.4617, "step": 32608 }, { "epoch": 3.27, "learning_rate": 4.285652070782131e-06, "loss": 0.1523, "step": 32609 }, { "epoch": 3.27, "learning_rate": 4.284419490380364e-06, "loss": 0.0831, "step": 32610 }, { "epoch": 3.27, "learning_rate": 4.283187073622767e-06, "loss": 0.1597, "step": 32611 }, { "epoch": 3.27, "learning_rate": 4.281954820517177e-06, "loss": 0.1009, "step": 32612 }, { "epoch": 3.27, "learning_rate": 4.280722731071458e-06, "loss": 0.0595, "step": 32613 }, { "epoch": 3.27, "learning_rate": 4.279490805293426e-06, "loss": 0.0965, "step": 32614 }, { "epoch": 3.27, "learning_rate": 4.278259043190926e-06, "loss": 0.0327, "step": 32615 }, { "epoch": 3.27, "learning_rate": 4.27702744477181e-06, "loss": 0.1487, "step": 32616 }, { "epoch": 3.28, "learning_rate": 4.275796010043902e-06, "loss": 0.1947, "step": 32617 }, { "epoch": 3.28, "learning_rate": 4.274564739015038e-06, "loss": 0.0367, "step": 32618 }, { "epoch": 3.28, "learning_rate": 4.273333631693058e-06, "loss": 0.1103, "step": 32619 }, { "epoch": 3.28, "learning_rate": 4.2721026880858076e-06, "loss": 0.2752, "step": 32620 }, { "epoch": 3.28, "learning_rate": 4.270871908201092e-06, "loss": 0.0805, "step": 32621 }, { "epoch": 3.28, "learning_rate": 4.269641292046769e-06, "loss": 0.4125, "step": 32622 }, { "epoch": 3.28, "learning_rate": 4.268410839630659e-06, "loss": 0.1281, "step": 32623 }, { "epoch": 3.28, "learning_rate": 4.2671805509605925e-06, "loss": 0.1364, "step": 32624 }, { "epoch": 3.28, "learning_rate": 4.265950426044403e-06, "loss": 0.2772, "step": 32625 }, { "epoch": 3.28, "learning_rate": 4.264720464889911e-06, "loss": 0.1496, "step": 32626 }, { "epoch": 3.28, "learning_rate": 4.26349066750495e-06, "loss": 0.1503, "step": 32627 }, { "epoch": 3.28, "learning_rate": 4.262261033897344e-06, "loss": 0.1933, "step": 32628 }, { "epoch": 3.28, "learning_rate": 4.261031564074918e-06, "loss": 0.2193, "step": 32629 }, { "epoch": 3.28, "learning_rate": 4.2598022580454965e-06, "loss": 0.1739, "step": 32630 }, { "epoch": 3.28, "learning_rate": 4.258573115816898e-06, "loss": 0.4001, "step": 32631 }, { "epoch": 3.28, "learning_rate": 4.25734413739695e-06, "loss": 0.0578, "step": 32632 }, { "epoch": 3.28, "learning_rate": 4.2561153227934624e-06, "loss": 0.2861, "step": 32633 }, { "epoch": 3.28, "learning_rate": 4.254886672014279e-06, "loss": 0.5368, "step": 32634 }, { "epoch": 3.28, "learning_rate": 4.25365818506719e-06, "loss": 1.4684, "step": 32635 }, { "epoch": 3.28, "learning_rate": 4.252429861960029e-06, "loss": 0.0141, "step": 32636 }, { "epoch": 3.28, "learning_rate": 4.25120170270061e-06, "loss": 0.2302, "step": 32637 }, { "epoch": 3.28, "learning_rate": 4.249973707296751e-06, "loss": 0.0125, "step": 32638 }, { "epoch": 3.28, "learning_rate": 4.248745875756263e-06, "loss": 0.1682, "step": 32639 }, { "epoch": 3.28, "learning_rate": 4.247518208086957e-06, "loss": 0.2442, "step": 32640 }, { "epoch": 3.28, "learning_rate": 4.24629070429665e-06, "loss": 0.0895, "step": 32641 }, { "epoch": 3.28, "learning_rate": 4.2450633643931545e-06, "loss": 0.0786, "step": 32642 }, { "epoch": 3.28, "learning_rate": 4.243836188384274e-06, "loss": 0.5452, "step": 32643 }, { "epoch": 3.28, "learning_rate": 4.2426091762778176e-06, "loss": 0.156, "step": 32644 }, { "epoch": 3.28, "learning_rate": 4.24138232808161e-06, "loss": 0.2717, "step": 32645 }, { "epoch": 3.28, "learning_rate": 4.240155643803442e-06, "loss": 0.0766, "step": 32646 }, { "epoch": 3.28, "learning_rate": 4.238929123451113e-06, "loss": 0.2205, "step": 32647 }, { "epoch": 3.28, "learning_rate": 4.2377027670324546e-06, "loss": 0.7854, "step": 32648 }, { "epoch": 3.28, "learning_rate": 4.2364765745552495e-06, "loss": 0.1399, "step": 32649 }, { "epoch": 3.28, "learning_rate": 4.235250546027298e-06, "loss": 0.1343, "step": 32650 }, { "epoch": 3.28, "learning_rate": 4.234024681456419e-06, "loss": 0.3121, "step": 32651 }, { "epoch": 3.28, "learning_rate": 4.2327989808504105e-06, "loss": 0.2904, "step": 32652 }, { "epoch": 3.28, "learning_rate": 4.231573444217055e-06, "loss": 0.4149, "step": 32653 }, { "epoch": 3.28, "learning_rate": 4.230348071564168e-06, "loss": 0.0554, "step": 32654 }, { "epoch": 3.28, "learning_rate": 4.229122862899545e-06, "loss": 0.0978, "step": 32655 }, { "epoch": 3.28, "learning_rate": 4.227897818230978e-06, "loss": 0.0494, "step": 32656 }, { "epoch": 3.28, "learning_rate": 4.226672937566265e-06, "loss": 0.0467, "step": 32657 }, { "epoch": 3.28, "learning_rate": 4.2254482209132026e-06, "loss": 0.0676, "step": 32658 }, { "epoch": 3.28, "learning_rate": 4.224223668279582e-06, "loss": 0.2852, "step": 32659 }, { "epoch": 3.28, "learning_rate": 4.222999279673196e-06, "loss": 0.0455, "step": 32660 }, { "epoch": 3.28, "learning_rate": 4.221775055101836e-06, "loss": 0.0324, "step": 32661 }, { "epoch": 3.28, "learning_rate": 4.220550994573295e-06, "loss": 0.2036, "step": 32662 }, { "epoch": 3.28, "learning_rate": 4.219327098095358e-06, "loss": 0.1929, "step": 32663 }, { "epoch": 3.28, "learning_rate": 4.218103365675811e-06, "loss": 0.1394, "step": 32664 }, { "epoch": 3.28, "learning_rate": 4.216879797322454e-06, "loss": 0.1775, "step": 32665 }, { "epoch": 3.28, "learning_rate": 4.215656393043068e-06, "loss": 0.0948, "step": 32666 }, { "epoch": 3.28, "learning_rate": 4.214433152845427e-06, "loss": 0.3323, "step": 32667 }, { "epoch": 3.28, "learning_rate": 4.213210076737328e-06, "loss": 0.0499, "step": 32668 }, { "epoch": 3.28, "learning_rate": 4.211987164726552e-06, "loss": 0.2305, "step": 32669 }, { "epoch": 3.28, "learning_rate": 4.210764416820878e-06, "loss": 0.1558, "step": 32670 }, { "epoch": 3.28, "learning_rate": 4.209541833028087e-06, "loss": 0.4645, "step": 32671 }, { "epoch": 3.28, "learning_rate": 4.208319413355965e-06, "loss": 0.1975, "step": 32672 }, { "epoch": 3.28, "learning_rate": 4.207097157812286e-06, "loss": 0.3206, "step": 32673 }, { "epoch": 3.28, "learning_rate": 4.205875066404824e-06, "loss": 0.1795, "step": 32674 }, { "epoch": 3.28, "learning_rate": 4.204653139141365e-06, "loss": 0.14, "step": 32675 }, { "epoch": 3.28, "learning_rate": 4.2034313760296705e-06, "loss": 0.287, "step": 32676 }, { "epoch": 3.28, "learning_rate": 4.2022097770775395e-06, "loss": 0.2436, "step": 32677 }, { "epoch": 3.28, "learning_rate": 4.200988342292727e-06, "loss": 0.3065, "step": 32678 }, { "epoch": 3.28, "learning_rate": 4.199767071683e-06, "loss": 0.0846, "step": 32679 }, { "epoch": 3.28, "learning_rate": 4.198545965256155e-06, "loss": 0.2223, "step": 32680 }, { "epoch": 3.28, "learning_rate": 4.197325023019945e-06, "loss": 0.2463, "step": 32681 }, { "epoch": 3.28, "learning_rate": 4.1961042449821306e-06, "loss": 0.1037, "step": 32682 }, { "epoch": 3.28, "learning_rate": 4.194883631150505e-06, "loss": 0.2315, "step": 32683 }, { "epoch": 3.28, "learning_rate": 4.193663181532824e-06, "loss": 0.0847, "step": 32684 }, { "epoch": 3.28, "learning_rate": 4.192442896136843e-06, "loss": 0.4703, "step": 32685 }, { "epoch": 3.28, "learning_rate": 4.191222774970346e-06, "loss": 0.2506, "step": 32686 }, { "epoch": 3.28, "learning_rate": 4.190002818041086e-06, "loss": 0.451, "step": 32687 }, { "epoch": 3.28, "learning_rate": 4.18878302535683e-06, "loss": 0.1166, "step": 32688 }, { "epoch": 3.28, "learning_rate": 4.187563396925341e-06, "loss": 0.3323, "step": 32689 }, { "epoch": 3.28, "learning_rate": 4.1863439327543715e-06, "loss": 0.4514, "step": 32690 }, { "epoch": 3.28, "learning_rate": 4.185124632851701e-06, "loss": 0.2361, "step": 32691 }, { "epoch": 3.28, "learning_rate": 4.183905497225073e-06, "loss": 0.5379, "step": 32692 }, { "epoch": 3.28, "learning_rate": 4.182686525882242e-06, "loss": 0.1308, "step": 32693 }, { "epoch": 3.28, "learning_rate": 4.181467718830985e-06, "loss": 0.111, "step": 32694 }, { "epoch": 3.28, "learning_rate": 4.180249076079038e-06, "loss": 0.0522, "step": 32695 }, { "epoch": 3.28, "learning_rate": 4.179030597634159e-06, "loss": 0.1553, "step": 32696 }, { "epoch": 3.28, "learning_rate": 4.177812283504112e-06, "loss": 0.1066, "step": 32697 }, { "epoch": 3.28, "learning_rate": 4.1765941336966505e-06, "loss": 0.0887, "step": 32698 }, { "epoch": 3.28, "learning_rate": 4.175376148219509e-06, "loss": 0.3367, "step": 32699 }, { "epoch": 3.28, "learning_rate": 4.174158327080453e-06, "loss": 0.2222, "step": 32700 }, { "epoch": 3.28, "learning_rate": 4.17294067028723e-06, "loss": 0.1887, "step": 32701 }, { "epoch": 3.28, "learning_rate": 4.171723177847589e-06, "loss": 0.1513, "step": 32702 }, { "epoch": 3.28, "learning_rate": 4.170505849769276e-06, "loss": 0.2261, "step": 32703 }, { "epoch": 3.28, "learning_rate": 4.1692886860600376e-06, "loss": 0.29, "step": 32704 }, { "epoch": 3.28, "learning_rate": 4.168071686727618e-06, "loss": 0.2442, "step": 32705 }, { "epoch": 3.28, "learning_rate": 4.166854851779765e-06, "loss": 0.2318, "step": 32706 }, { "epoch": 3.28, "learning_rate": 4.1656381812242205e-06, "loss": 0.2582, "step": 32707 }, { "epoch": 3.28, "learning_rate": 4.164421675068717e-06, "loss": 0.2585, "step": 32708 }, { "epoch": 3.28, "learning_rate": 4.16320533332102e-06, "loss": 0.2908, "step": 32709 }, { "epoch": 3.28, "learning_rate": 4.161989155988852e-06, "loss": 0.0561, "step": 32710 }, { "epoch": 3.28, "learning_rate": 4.160773143079945e-06, "loss": 0.6144, "step": 32711 }, { "epoch": 3.28, "learning_rate": 4.1595572946020645e-06, "loss": 0.1907, "step": 32712 }, { "epoch": 3.28, "learning_rate": 4.1583416105629224e-06, "loss": 0.1498, "step": 32713 }, { "epoch": 3.28, "learning_rate": 4.157126090970257e-06, "loss": 0.1545, "step": 32714 }, { "epoch": 3.29, "learning_rate": 4.1559107358318185e-06, "loss": 0.2886, "step": 32715 }, { "epoch": 3.29, "learning_rate": 4.15469554515533e-06, "loss": 0.1115, "step": 32716 }, { "epoch": 3.29, "learning_rate": 4.153480518948527e-06, "loss": 0.4627, "step": 32717 }, { "epoch": 3.29, "learning_rate": 4.1522656572191455e-06, "loss": 0.106, "step": 32718 }, { "epoch": 3.29, "learning_rate": 4.151050959974909e-06, "loss": 0.125, "step": 32719 }, { "epoch": 3.29, "learning_rate": 4.149836427223552e-06, "loss": 0.4708, "step": 32720 }, { "epoch": 3.29, "learning_rate": 4.1486220589727984e-06, "loss": 0.0311, "step": 32721 }, { "epoch": 3.29, "learning_rate": 4.147407855230378e-06, "loss": 0.0367, "step": 32722 }, { "epoch": 3.29, "learning_rate": 4.146193816004029e-06, "loss": 0.0741, "step": 32723 }, { "epoch": 3.29, "learning_rate": 4.144979941301464e-06, "loss": 0.1421, "step": 32724 }, { "epoch": 3.29, "learning_rate": 4.143766231130402e-06, "loss": 0.1267, "step": 32725 }, { "epoch": 3.29, "learning_rate": 4.1425526854985875e-06, "loss": 0.08, "step": 32726 }, { "epoch": 3.29, "learning_rate": 4.141339304413726e-06, "loss": 0.2744, "step": 32727 }, { "epoch": 3.29, "learning_rate": 4.140126087883538e-06, "loss": 0.5468, "step": 32728 }, { "epoch": 3.29, "learning_rate": 4.138913035915756e-06, "loss": 0.2851, "step": 32729 }, { "epoch": 3.29, "learning_rate": 4.1377001485181e-06, "loss": 0.2319, "step": 32730 }, { "epoch": 3.29, "learning_rate": 4.136487425698267e-06, "loss": 0.504, "step": 32731 }, { "epoch": 3.29, "learning_rate": 4.135274867464001e-06, "loss": 0.0548, "step": 32732 }, { "epoch": 3.29, "learning_rate": 4.134062473822999e-06, "loss": 0.2253, "step": 32733 }, { "epoch": 3.29, "learning_rate": 4.132850244782987e-06, "loss": 0.3698, "step": 32734 }, { "epoch": 3.29, "learning_rate": 4.131638180351676e-06, "loss": 0.2599, "step": 32735 }, { "epoch": 3.29, "learning_rate": 4.130426280536781e-06, "loss": 0.2724, "step": 32736 }, { "epoch": 3.29, "learning_rate": 4.129214545346007e-06, "loss": 0.1456, "step": 32737 }, { "epoch": 3.29, "learning_rate": 4.128002974787072e-06, "loss": 0.5983, "step": 32738 }, { "epoch": 3.29, "learning_rate": 4.126791568867685e-06, "loss": 0.5763, "step": 32739 }, { "epoch": 3.29, "learning_rate": 4.1255803275955474e-06, "loss": 0.0809, "step": 32740 }, { "epoch": 3.29, "learning_rate": 4.124369250978386e-06, "loss": 0.186, "step": 32741 }, { "epoch": 3.29, "learning_rate": 4.123158339023888e-06, "loss": 0.3006, "step": 32742 }, { "epoch": 3.29, "learning_rate": 4.121947591739758e-06, "loss": 0.0881, "step": 32743 }, { "epoch": 3.29, "learning_rate": 4.1207370091337246e-06, "loss": 0.1523, "step": 32744 }, { "epoch": 3.29, "learning_rate": 4.1195265912134605e-06, "loss": 0.1912, "step": 32745 }, { "epoch": 3.29, "learning_rate": 4.118316337986692e-06, "loss": 0.149, "step": 32746 }, { "epoch": 3.29, "learning_rate": 4.117106249461112e-06, "loss": 0.1136, "step": 32747 }, { "epoch": 3.29, "learning_rate": 4.115896325644423e-06, "loss": 0.0236, "step": 32748 }, { "epoch": 3.29, "learning_rate": 4.11468656654432e-06, "loss": 0.0811, "step": 32749 }, { "epoch": 3.29, "learning_rate": 4.113476972168508e-06, "loss": 0.0619, "step": 32750 }, { "epoch": 3.29, "learning_rate": 4.112267542524678e-06, "loss": 0.6983, "step": 32751 }, { "epoch": 3.29, "learning_rate": 4.11105827762053e-06, "loss": 0.0649, "step": 32752 }, { "epoch": 3.29, "learning_rate": 4.109849177463761e-06, "loss": 0.1028, "step": 32753 }, { "epoch": 3.29, "learning_rate": 4.10864024206205e-06, "loss": 0.0821, "step": 32754 }, { "epoch": 3.29, "learning_rate": 4.107431471423121e-06, "loss": 0.0691, "step": 32755 }, { "epoch": 3.29, "learning_rate": 4.106222865554642e-06, "loss": 0.1531, "step": 32756 }, { "epoch": 3.29, "learning_rate": 4.105014424464301e-06, "loss": 0.4354, "step": 32757 }, { "epoch": 3.29, "learning_rate": 4.103806148159811e-06, "loss": 0.1144, "step": 32758 }, { "epoch": 3.29, "learning_rate": 4.10259803664884e-06, "loss": 0.0386, "step": 32759 }, { "epoch": 3.29, "learning_rate": 4.1013900899390785e-06, "loss": 0.0876, "step": 32760 }, { "epoch": 3.29, "learning_rate": 4.100182308038224e-06, "loss": 0.5729, "step": 32761 }, { "epoch": 3.29, "learning_rate": 4.098974690953963e-06, "loss": 0.2486, "step": 32762 }, { "epoch": 3.29, "learning_rate": 4.097767238693959e-06, "loss": 0.3796, "step": 32763 }, { "epoch": 3.29, "learning_rate": 4.09655995126592e-06, "loss": 0.0979, "step": 32764 }, { "epoch": 3.29, "learning_rate": 4.0953528286775164e-06, "loss": 0.2026, "step": 32765 }, { "epoch": 3.29, "learning_rate": 4.094145870936436e-06, "loss": 0.0747, "step": 32766 }, { "epoch": 3.29, "learning_rate": 4.092939078050354e-06, "loss": 0.1697, "step": 32767 }, { "epoch": 3.29, "learning_rate": 4.0917324500269525e-06, "loss": 0.1684, "step": 32768 }, { "epoch": 3.29, "learning_rate": 4.0905259868739085e-06, "loss": 0.0755, "step": 32769 }, { "epoch": 3.29, "learning_rate": 4.0893196885989025e-06, "loss": 0.1851, "step": 32770 }, { "epoch": 3.29, "learning_rate": 4.088113555209602e-06, "loss": 0.1779, "step": 32771 }, { "epoch": 3.29, "learning_rate": 4.086907586713696e-06, "loss": 0.0916, "step": 32772 }, { "epoch": 3.29, "learning_rate": 4.085701783118861e-06, "loss": 0.3085, "step": 32773 }, { "epoch": 3.29, "learning_rate": 4.084496144432744e-06, "loss": 0.0525, "step": 32774 }, { "epoch": 3.29, "learning_rate": 4.083290670663046e-06, "loss": 0.0257, "step": 32775 }, { "epoch": 3.29, "learning_rate": 4.082085361817432e-06, "loss": 0.2421, "step": 32776 }, { "epoch": 3.29, "learning_rate": 4.080880217903552e-06, "loss": 0.249, "step": 32777 }, { "epoch": 3.29, "learning_rate": 4.079675238929098e-06, "loss": 0.269, "step": 32778 }, { "epoch": 3.29, "learning_rate": 4.07847042490173e-06, "loss": 1.9447, "step": 32779 }, { "epoch": 3.29, "learning_rate": 4.077265775829116e-06, "loss": 0.0957, "step": 32780 }, { "epoch": 3.29, "learning_rate": 4.076061291718918e-06, "loss": 0.346, "step": 32781 }, { "epoch": 3.29, "learning_rate": 4.074856972578808e-06, "loss": 0.0129, "step": 32782 }, { "epoch": 3.29, "learning_rate": 4.073652818416442e-06, "loss": 0.1114, "step": 32783 }, { "epoch": 3.29, "learning_rate": 4.072448829239489e-06, "loss": 0.0542, "step": 32784 }, { "epoch": 3.29, "learning_rate": 4.071245005055608e-06, "loss": 0.2819, "step": 32785 }, { "epoch": 3.29, "learning_rate": 4.070041345872454e-06, "loss": 0.2369, "step": 32786 }, { "epoch": 3.29, "learning_rate": 4.068837851697703e-06, "loss": 0.2125, "step": 32787 }, { "epoch": 3.29, "learning_rate": 4.0676345225389996e-06, "loss": 0.4736, "step": 32788 }, { "epoch": 3.29, "learning_rate": 4.066431358403998e-06, "loss": 0.5965, "step": 32789 }, { "epoch": 3.29, "learning_rate": 4.0652283593003745e-06, "loss": 0.1285, "step": 32790 }, { "epoch": 3.29, "learning_rate": 4.064025525235764e-06, "loss": 0.1653, "step": 32791 }, { "epoch": 3.29, "learning_rate": 4.062822856217822e-06, "loss": 0.5936, "step": 32792 }, { "epoch": 3.29, "learning_rate": 4.061620352254215e-06, "loss": 0.0795, "step": 32793 }, { "epoch": 3.29, "learning_rate": 4.060418013352598e-06, "loss": 0.11, "step": 32794 }, { "epoch": 3.29, "learning_rate": 4.059215839520596e-06, "loss": 0.1005, "step": 32795 }, { "epoch": 3.29, "learning_rate": 4.058013830765885e-06, "loss": 0.0556, "step": 32796 }, { "epoch": 3.29, "learning_rate": 4.056811987096104e-06, "loss": 0.1756, "step": 32797 }, { "epoch": 3.29, "learning_rate": 4.055610308518902e-06, "loss": 0.0566, "step": 32798 }, { "epoch": 3.29, "learning_rate": 4.054408795041928e-06, "loss": 0.5059, "step": 32799 }, { "epoch": 3.29, "learning_rate": 4.053207446672817e-06, "loss": 0.1457, "step": 32800 }, { "epoch": 3.29, "learning_rate": 4.052006263419237e-06, "loss": 0.2791, "step": 32801 }, { "epoch": 3.29, "learning_rate": 4.0508052452888125e-06, "loss": 0.163, "step": 32802 }, { "epoch": 3.29, "learning_rate": 4.049604392289186e-06, "loss": 0.0998, "step": 32803 }, { "epoch": 3.29, "learning_rate": 4.048403704428011e-06, "loss": 0.031, "step": 32804 }, { "epoch": 3.29, "learning_rate": 4.0472031817129285e-06, "loss": 0.0301, "step": 32805 }, { "epoch": 3.29, "learning_rate": 4.046002824151557e-06, "loss": 0.1661, "step": 32806 }, { "epoch": 3.29, "learning_rate": 4.044802631751559e-06, "loss": 0.1277, "step": 32807 }, { "epoch": 3.29, "learning_rate": 4.043602604520571e-06, "loss": 0.3921, "step": 32808 }, { "epoch": 3.29, "learning_rate": 4.042402742466205e-06, "loss": 0.146, "step": 32809 }, { "epoch": 3.29, "learning_rate": 4.041203045596121e-06, "loss": 0.0929, "step": 32810 }, { "epoch": 3.29, "learning_rate": 4.040003513917948e-06, "loss": 0.5132, "step": 32811 }, { "epoch": 3.29, "learning_rate": 4.0388041474393154e-06, "loss": 0.169, "step": 32812 }, { "epoch": 3.29, "learning_rate": 4.037604946167854e-06, "loss": 0.145, "step": 32813 }, { "epoch": 3.3, "learning_rate": 4.0364059101112015e-06, "loss": 0.0389, "step": 32814 }, { "epoch": 3.3, "learning_rate": 4.035207039276985e-06, "loss": 0.0947, "step": 32815 }, { "epoch": 3.3, "learning_rate": 4.034008333672834e-06, "loss": 0.2134, "step": 32816 }, { "epoch": 3.3, "learning_rate": 4.0328097933063735e-06, "loss": 0.0906, "step": 32817 }, { "epoch": 3.3, "learning_rate": 4.031611418185227e-06, "loss": 0.4339, "step": 32818 }, { "epoch": 3.3, "learning_rate": 4.0304132083170385e-06, "loss": 0.0741, "step": 32819 }, { "epoch": 3.3, "learning_rate": 4.029215163709418e-06, "loss": 0.2184, "step": 32820 }, { "epoch": 3.3, "learning_rate": 4.028017284369985e-06, "loss": 0.3168, "step": 32821 }, { "epoch": 3.3, "learning_rate": 4.02681957030638e-06, "loss": 0.0689, "step": 32822 }, { "epoch": 3.3, "learning_rate": 4.025622021526209e-06, "loss": 0.3192, "step": 32823 }, { "epoch": 3.3, "learning_rate": 4.024424638037093e-06, "loss": 0.3754, "step": 32824 }, { "epoch": 3.3, "learning_rate": 4.023227419846665e-06, "loss": 0.2531, "step": 32825 }, { "epoch": 3.3, "learning_rate": 4.022030366962531e-06, "loss": 0.2061, "step": 32826 }, { "epoch": 3.3, "learning_rate": 4.020833479392317e-06, "loss": 0.123, "step": 32827 }, { "epoch": 3.3, "learning_rate": 4.019636757143636e-06, "loss": 0.1647, "step": 32828 }, { "epoch": 3.3, "learning_rate": 4.018440200224099e-06, "loss": 0.5654, "step": 32829 }, { "epoch": 3.3, "learning_rate": 4.01724380864133e-06, "loss": 0.145, "step": 32830 }, { "epoch": 3.3, "learning_rate": 4.0160475824029355e-06, "loss": 0.1213, "step": 32831 }, { "epoch": 3.3, "learning_rate": 4.014851521516519e-06, "loss": 0.1118, "step": 32832 }, { "epoch": 3.3, "learning_rate": 4.0136556259897195e-06, "loss": 0.4689, "step": 32833 }, { "epoch": 3.3, "learning_rate": 4.012459895830123e-06, "loss": 0.3876, "step": 32834 }, { "epoch": 3.3, "learning_rate": 4.011264331045337e-06, "loss": 0.237, "step": 32835 }, { "epoch": 3.3, "learning_rate": 4.010068931642985e-06, "loss": 0.1091, "step": 32836 }, { "epoch": 3.3, "learning_rate": 4.008873697630675e-06, "loss": 0.2438, "step": 32837 }, { "epoch": 3.3, "learning_rate": 4.007678629015994e-06, "loss": 0.0745, "step": 32838 }, { "epoch": 3.3, "learning_rate": 4.006483725806562e-06, "loss": 0.145, "step": 32839 }, { "epoch": 3.3, "learning_rate": 4.005288988009984e-06, "loss": 0.0602, "step": 32840 }, { "epoch": 3.3, "learning_rate": 4.00409441563385e-06, "loss": 0.0101, "step": 32841 }, { "epoch": 3.3, "learning_rate": 4.002900008685773e-06, "loss": 0.1747, "step": 32842 }, { "epoch": 3.3, "learning_rate": 4.001705767173349e-06, "loss": 0.3226, "step": 32843 }, { "epoch": 3.3, "learning_rate": 4.000511691104184e-06, "loss": 0.0695, "step": 32844 }, { "epoch": 3.3, "learning_rate": 3.999317780485868e-06, "loss": 0.0965, "step": 32845 }, { "epoch": 3.3, "learning_rate": 3.998124035326003e-06, "loss": 0.0641, "step": 32846 }, { "epoch": 3.3, "learning_rate": 3.996930455632185e-06, "loss": 0.196, "step": 32847 }, { "epoch": 3.3, "learning_rate": 3.995737041412012e-06, "loss": 0.217, "step": 32848 }, { "epoch": 3.3, "learning_rate": 3.9945437926730745e-06, "loss": 0.0188, "step": 32849 }, { "epoch": 3.3, "learning_rate": 3.9933507094229595e-06, "loss": 0.1425, "step": 32850 }, { "epoch": 3.3, "learning_rate": 3.992157791669281e-06, "loss": 0.1028, "step": 32851 }, { "epoch": 3.3, "learning_rate": 3.990965039419603e-06, "loss": 0.1604, "step": 32852 }, { "epoch": 3.3, "learning_rate": 3.989772452681536e-06, "loss": 0.2939, "step": 32853 }, { "epoch": 3.3, "learning_rate": 3.9885800314626675e-06, "loss": 0.3123, "step": 32854 }, { "epoch": 3.3, "learning_rate": 3.987387775770571e-06, "loss": 0.1447, "step": 32855 }, { "epoch": 3.3, "learning_rate": 3.986195685612848e-06, "loss": 0.0739, "step": 32856 }, { "epoch": 3.3, "learning_rate": 3.985003760997078e-06, "loss": 0.0747, "step": 32857 }, { "epoch": 3.3, "learning_rate": 3.983812001930849e-06, "loss": 0.1591, "step": 32858 }, { "epoch": 3.3, "learning_rate": 3.982620408421741e-06, "loss": 0.6678, "step": 32859 }, { "epoch": 3.3, "learning_rate": 3.981428980477342e-06, "loss": 0.0305, "step": 32860 }, { "epoch": 3.3, "learning_rate": 3.980237718105229e-06, "loss": 0.6258, "step": 32861 }, { "epoch": 3.3, "learning_rate": 3.979046621312985e-06, "loss": 0.157, "step": 32862 }, { "epoch": 3.3, "learning_rate": 3.977855690108192e-06, "loss": 0.3288, "step": 32863 }, { "epoch": 3.3, "learning_rate": 3.976664924498417e-06, "loss": 0.0843, "step": 32864 }, { "epoch": 3.3, "learning_rate": 3.9754743244912595e-06, "loss": 0.2267, "step": 32865 }, { "epoch": 3.3, "learning_rate": 3.974283890094278e-06, "loss": 0.062, "step": 32866 }, { "epoch": 3.3, "learning_rate": 3.973093621315046e-06, "loss": 0.2592, "step": 32867 }, { "epoch": 3.3, "learning_rate": 3.971903518161151e-06, "loss": 0.083, "step": 32868 }, { "epoch": 3.3, "learning_rate": 3.970713580640166e-06, "loss": 0.5813, "step": 32869 }, { "epoch": 3.3, "learning_rate": 3.969523808759646e-06, "loss": 0.231, "step": 32870 }, { "epoch": 3.3, "learning_rate": 3.968334202527181e-06, "loss": 0.4834, "step": 32871 }, { "epoch": 3.3, "learning_rate": 3.967144761950343e-06, "loss": 0.4133, "step": 32872 }, { "epoch": 3.3, "learning_rate": 3.965955487036675e-06, "loss": 0.1889, "step": 32873 }, { "epoch": 3.3, "learning_rate": 3.964766377793772e-06, "loss": 0.5011, "step": 32874 }, { "epoch": 3.3, "learning_rate": 3.963577434229188e-06, "loss": 0.0849, "step": 32875 }, { "epoch": 3.3, "learning_rate": 3.962388656350496e-06, "loss": 0.0892, "step": 32876 }, { "epoch": 3.3, "learning_rate": 3.961200044165256e-06, "loss": 0.2449, "step": 32877 }, { "epoch": 3.3, "learning_rate": 3.960011597681026e-06, "loss": 0.0707, "step": 32878 }, { "epoch": 3.3, "learning_rate": 3.958823316905389e-06, "loss": 0.0796, "step": 32879 }, { "epoch": 3.3, "learning_rate": 3.95763520184589e-06, "loss": 0.2367, "step": 32880 }, { "epoch": 3.3, "learning_rate": 3.956447252510085e-06, "loss": 0.0785, "step": 32881 }, { "epoch": 3.3, "learning_rate": 3.9552594689055485e-06, "loss": 0.1397, "step": 32882 }, { "epoch": 3.3, "learning_rate": 3.954071851039839e-06, "loss": 0.0399, "step": 32883 }, { "epoch": 3.3, "learning_rate": 3.9528843989204955e-06, "loss": 0.6112, "step": 32884 }, { "epoch": 3.3, "learning_rate": 3.951697112555092e-06, "loss": 0.1692, "step": 32885 }, { "epoch": 3.3, "learning_rate": 3.9505099919511876e-06, "loss": 0.2537, "step": 32886 }, { "epoch": 3.3, "learning_rate": 3.949323037116315e-06, "loss": 0.4456, "step": 32887 }, { "epoch": 3.3, "learning_rate": 3.948136248058044e-06, "loss": 0.2642, "step": 32888 }, { "epoch": 3.3, "learning_rate": 3.946949624783926e-06, "loss": 0.2602, "step": 32889 }, { "epoch": 3.3, "learning_rate": 3.945763167301509e-06, "loss": 0.121, "step": 32890 }, { "epoch": 3.3, "learning_rate": 3.944576875618343e-06, "loss": 0.0946, "step": 32891 }, { "epoch": 3.3, "learning_rate": 3.9433907497419784e-06, "loss": 0.1677, "step": 32892 }, { "epoch": 3.3, "learning_rate": 3.942204789679963e-06, "loss": 0.3688, "step": 32893 }, { "epoch": 3.3, "learning_rate": 3.9410189954398435e-06, "loss": 0.1204, "step": 32894 }, { "epoch": 3.3, "learning_rate": 3.939833367029165e-06, "loss": 0.1245, "step": 32895 }, { "epoch": 3.3, "learning_rate": 3.938647904455466e-06, "loss": 0.8006, "step": 32896 }, { "epoch": 3.3, "learning_rate": 3.937462607726312e-06, "loss": 0.2945, "step": 32897 }, { "epoch": 3.3, "learning_rate": 3.936277476849224e-06, "loss": 0.3017, "step": 32898 }, { "epoch": 3.3, "learning_rate": 3.935092511831742e-06, "loss": 0.2161, "step": 32899 }, { "epoch": 3.3, "learning_rate": 3.933907712681426e-06, "loss": 0.0975, "step": 32900 }, { "epoch": 3.3, "learning_rate": 3.932723079405811e-06, "loss": 0.1199, "step": 32901 }, { "epoch": 3.3, "learning_rate": 3.931538612012414e-06, "loss": 0.14, "step": 32902 }, { "epoch": 3.3, "learning_rate": 3.930354310508799e-06, "loss": 0.0388, "step": 32903 }, { "epoch": 3.3, "learning_rate": 3.929170174902497e-06, "loss": 0.4018, "step": 32904 }, { "epoch": 3.3, "learning_rate": 3.927986205201024e-06, "loss": 0.0631, "step": 32905 }, { "epoch": 3.3, "learning_rate": 3.926802401411935e-06, "loss": 0.2906, "step": 32906 }, { "epoch": 3.3, "learning_rate": 3.925618763542757e-06, "loss": 0.0898, "step": 32907 }, { "epoch": 3.3, "learning_rate": 3.924435291601022e-06, "loss": 0.1332, "step": 32908 }, { "epoch": 3.3, "learning_rate": 3.923251985594261e-06, "loss": 0.0777, "step": 32909 }, { "epoch": 3.3, "learning_rate": 3.922068845529998e-06, "loss": 0.091, "step": 32910 }, { "epoch": 3.3, "learning_rate": 3.9208858714157815e-06, "loss": 0.4394, "step": 32911 }, { "epoch": 3.31, "learning_rate": 3.919703063259119e-06, "loss": 0.0383, "step": 32912 }, { "epoch": 3.31, "learning_rate": 3.918520421067538e-06, "loss": 0.2158, "step": 32913 }, { "epoch": 3.31, "learning_rate": 3.91733794484858e-06, "loss": 0.1459, "step": 32914 }, { "epoch": 3.31, "learning_rate": 3.916155634609765e-06, "loss": 0.0423, "step": 32915 }, { "epoch": 3.31, "learning_rate": 3.914973490358603e-06, "loss": 0.0348, "step": 32916 }, { "epoch": 3.31, "learning_rate": 3.913791512102629e-06, "loss": 0.102, "step": 32917 }, { "epoch": 3.31, "learning_rate": 3.91260969984937e-06, "loss": 0.1663, "step": 32918 }, { "epoch": 3.31, "learning_rate": 3.911428053606326e-06, "loss": 0.0726, "step": 32919 }, { "epoch": 3.31, "learning_rate": 3.910246573381036e-06, "loss": 0.0933, "step": 32920 }, { "epoch": 3.31, "learning_rate": 3.909065259181012e-06, "loss": 0.1263, "step": 32921 }, { "epoch": 3.31, "learning_rate": 3.907884111013772e-06, "loss": 0.233, "step": 32922 }, { "epoch": 3.31, "learning_rate": 3.9067031288868305e-06, "loss": 0.1555, "step": 32923 }, { "epoch": 3.31, "learning_rate": 3.905522312807704e-06, "loss": 0.2614, "step": 32924 }, { "epoch": 3.31, "learning_rate": 3.904341662783904e-06, "loss": 0.5021, "step": 32925 }, { "epoch": 3.31, "learning_rate": 3.9031611788229515e-06, "loss": 0.1454, "step": 32926 }, { "epoch": 3.31, "learning_rate": 3.901980860932349e-06, "loss": 0.3004, "step": 32927 }, { "epoch": 3.31, "learning_rate": 3.900800709119609e-06, "loss": 0.3049, "step": 32928 }, { "epoch": 3.31, "learning_rate": 3.8996207233922565e-06, "loss": 0.3882, "step": 32929 }, { "epoch": 3.31, "learning_rate": 3.898440903757781e-06, "loss": 0.5563, "step": 32930 }, { "epoch": 3.31, "learning_rate": 3.897261250223692e-06, "loss": 0.2147, "step": 32931 }, { "epoch": 3.31, "learning_rate": 3.89608176279751e-06, "loss": 0.1398, "step": 32932 }, { "epoch": 3.31, "learning_rate": 3.894902441486729e-06, "loss": 0.0799, "step": 32933 }, { "epoch": 3.31, "learning_rate": 3.893723286298862e-06, "loss": 0.1978, "step": 32934 }, { "epoch": 3.31, "learning_rate": 3.892544297241409e-06, "loss": 0.2187, "step": 32935 }, { "epoch": 3.31, "learning_rate": 3.8913654743218704e-06, "loss": 0.079, "step": 32936 }, { "epoch": 3.31, "learning_rate": 3.890186817547753e-06, "loss": 0.1062, "step": 32937 }, { "epoch": 3.31, "learning_rate": 3.889008326926549e-06, "loss": 0.2058, "step": 32938 }, { "epoch": 3.31, "learning_rate": 3.887830002465765e-06, "loss": 0.0593, "step": 32939 }, { "epoch": 3.31, "learning_rate": 3.8866518441729e-06, "loss": 0.1567, "step": 32940 }, { "epoch": 3.31, "learning_rate": 3.885473852055446e-06, "loss": 0.1186, "step": 32941 }, { "epoch": 3.31, "learning_rate": 3.8842960261208945e-06, "loss": 0.382, "step": 32942 }, { "epoch": 3.31, "learning_rate": 3.883118366376762e-06, "loss": 0.1037, "step": 32943 }, { "epoch": 3.31, "learning_rate": 3.881940872830525e-06, "loss": 0.1107, "step": 32944 }, { "epoch": 3.31, "learning_rate": 3.880763545489671e-06, "loss": 0.2684, "step": 32945 }, { "epoch": 3.31, "learning_rate": 3.87958638436171e-06, "loss": 0.0193, "step": 32946 }, { "epoch": 3.31, "learning_rate": 3.878409389454129e-06, "loss": 0.2759, "step": 32947 }, { "epoch": 3.31, "learning_rate": 3.877232560774401e-06, "loss": 0.3533, "step": 32948 }, { "epoch": 3.31, "learning_rate": 3.8760558983300345e-06, "loss": 0.1454, "step": 32949 }, { "epoch": 3.31, "learning_rate": 3.874879402128517e-06, "loss": 0.2374, "step": 32950 }, { "epoch": 3.31, "learning_rate": 3.873703072177314e-06, "loss": 0.068, "step": 32951 }, { "epoch": 3.31, "learning_rate": 3.872526908483934e-06, "loss": 0.279, "step": 32952 }, { "epoch": 3.31, "learning_rate": 3.871350911055853e-06, "loss": 0.1187, "step": 32953 }, { "epoch": 3.31, "learning_rate": 3.870175079900554e-06, "loss": 0.3088, "step": 32954 }, { "epoch": 3.31, "learning_rate": 3.868999415025522e-06, "loss": 0.2227, "step": 32955 }, { "epoch": 3.31, "learning_rate": 3.867823916438236e-06, "loss": 0.0325, "step": 32956 }, { "epoch": 3.31, "learning_rate": 3.866648584146174e-06, "loss": 0.1243, "step": 32957 }, { "epoch": 3.31, "learning_rate": 3.865473418156824e-06, "loss": 0.0479, "step": 32958 }, { "epoch": 3.31, "learning_rate": 3.864298418477657e-06, "loss": 0.5358, "step": 32959 }, { "epoch": 3.31, "learning_rate": 3.863123585116145e-06, "loss": 0.2917, "step": 32960 }, { "epoch": 3.31, "learning_rate": 3.861948918079787e-06, "loss": 0.1558, "step": 32961 }, { "epoch": 3.31, "learning_rate": 3.860774417376027e-06, "loss": 0.2083, "step": 32962 }, { "epoch": 3.31, "learning_rate": 3.859600083012362e-06, "loss": 0.2527, "step": 32963 }, { "epoch": 3.31, "learning_rate": 3.85842591499626e-06, "loss": 0.0854, "step": 32964 }, { "epoch": 3.31, "learning_rate": 3.857251913335187e-06, "loss": 0.0945, "step": 32965 }, { "epoch": 3.31, "learning_rate": 3.856078078036621e-06, "loss": 0.1926, "step": 32966 }, { "epoch": 3.31, "learning_rate": 3.854904409108027e-06, "loss": 0.2161, "step": 32967 }, { "epoch": 3.31, "learning_rate": 3.8537309065568735e-06, "loss": 0.0537, "step": 32968 }, { "epoch": 3.31, "learning_rate": 3.852557570390633e-06, "loss": 0.0738, "step": 32969 }, { "epoch": 3.31, "learning_rate": 3.8513844006167664e-06, "loss": 0.1158, "step": 32970 }, { "epoch": 3.31, "learning_rate": 3.850211397242743e-06, "loss": 0.9675, "step": 32971 }, { "epoch": 3.31, "learning_rate": 3.849038560276028e-06, "loss": 0.0619, "step": 32972 }, { "epoch": 3.31, "learning_rate": 3.847865889724081e-06, "loss": 0.2438, "step": 32973 }, { "epoch": 3.31, "learning_rate": 3.846693385594364e-06, "loss": 0.231, "step": 32974 }, { "epoch": 3.31, "learning_rate": 3.84552104789435e-06, "loss": 0.0505, "step": 32975 }, { "epoch": 3.31, "learning_rate": 3.844348876631487e-06, "loss": 0.3165, "step": 32976 }, { "epoch": 3.31, "learning_rate": 3.843176871813229e-06, "loss": 0.0644, "step": 32977 }, { "epoch": 3.31, "learning_rate": 3.842005033447049e-06, "loss": 0.363, "step": 32978 }, { "epoch": 3.31, "learning_rate": 3.840833361540405e-06, "loss": 0.2007, "step": 32979 }, { "epoch": 3.31, "learning_rate": 3.839661856100732e-06, "loss": 0.0231, "step": 32980 }, { "epoch": 3.31, "learning_rate": 3.838490517135507e-06, "loss": 0.2512, "step": 32981 }, { "epoch": 3.31, "learning_rate": 3.837319344652183e-06, "loss": 0.3717, "step": 32982 }, { "epoch": 3.31, "learning_rate": 3.8361483386581906e-06, "loss": 0.0279, "step": 32983 }, { "epoch": 3.31, "learning_rate": 3.834977499161005e-06, "loss": 0.1832, "step": 32984 }, { "epoch": 3.31, "learning_rate": 3.83380682616807e-06, "loss": 0.1311, "step": 32985 }, { "epoch": 3.31, "learning_rate": 3.832636319686833e-06, "loss": 0.1674, "step": 32986 }, { "epoch": 3.31, "learning_rate": 3.831465979724744e-06, "loss": 0.0709, "step": 32987 }, { "epoch": 3.31, "learning_rate": 3.830295806289245e-06, "loss": 0.1828, "step": 32988 }, { "epoch": 3.31, "learning_rate": 3.829125799387798e-06, "loss": 0.3498, "step": 32989 }, { "epoch": 3.31, "learning_rate": 3.827955959027836e-06, "loss": 0.1368, "step": 32990 }, { "epoch": 3.31, "learning_rate": 3.826786285216798e-06, "loss": 0.1868, "step": 32991 }, { "epoch": 3.31, "learning_rate": 3.825616777962138e-06, "loss": 0.15, "step": 32992 }, { "epoch": 3.31, "learning_rate": 3.824447437271306e-06, "loss": 0.511, "step": 32993 }, { "epoch": 3.31, "learning_rate": 3.82327826315172e-06, "loss": 0.0609, "step": 32994 }, { "epoch": 3.31, "learning_rate": 3.822109255610841e-06, "loss": 0.1134, "step": 32995 }, { "epoch": 3.31, "learning_rate": 3.820940414656097e-06, "loss": 0.1111, "step": 32996 }, { "epoch": 3.31, "learning_rate": 3.819771740294928e-06, "loss": 0.2083, "step": 32997 }, { "epoch": 3.31, "learning_rate": 3.8186032325347734e-06, "loss": 0.1878, "step": 32998 }, { "epoch": 3.31, "learning_rate": 3.817434891383071e-06, "loss": 0.0285, "step": 32999 }, { "epoch": 3.31, "learning_rate": 3.816266716847247e-06, "loss": 0.1106, "step": 33000 }, { "epoch": 3.31, "learning_rate": 3.815098708934746e-06, "loss": 0.1869, "step": 33001 }, { "epoch": 3.31, "learning_rate": 3.813930867652993e-06, "loss": 0.1266, "step": 33002 }, { "epoch": 3.31, "learning_rate": 3.812763193009423e-06, "loss": 0.1263, "step": 33003 }, { "epoch": 3.31, "learning_rate": 3.8115956850114675e-06, "loss": 0.1018, "step": 33004 }, { "epoch": 3.31, "learning_rate": 3.81042834366655e-06, "loss": 0.023, "step": 33005 }, { "epoch": 3.31, "learning_rate": 3.809261168982101e-06, "loss": 0.1618, "step": 33006 }, { "epoch": 3.31, "learning_rate": 3.8080941609655617e-06, "loss": 0.0962, "step": 33007 }, { "epoch": 3.31, "learning_rate": 3.8069273196243425e-06, "loss": 0.0924, "step": 33008 }, { "epoch": 3.31, "learning_rate": 3.8057606449658645e-06, "loss": 0.1668, "step": 33009 }, { "epoch": 3.31, "learning_rate": 3.804594136997568e-06, "loss": 0.2953, "step": 33010 }, { "epoch": 3.32, "learning_rate": 3.803427795726877e-06, "loss": 0.3641, "step": 33011 }, { "epoch": 3.32, "learning_rate": 3.8022616211611925e-06, "loss": 0.1911, "step": 33012 }, { "epoch": 3.32, "learning_rate": 3.8010956133079553e-06, "loss": 0.2033, "step": 33013 }, { "epoch": 3.32, "learning_rate": 3.7999297721745797e-06, "loss": 0.075, "step": 33014 }, { "epoch": 3.32, "learning_rate": 3.798764097768482e-06, "loss": 0.3677, "step": 33015 }, { "epoch": 3.32, "learning_rate": 3.7975985900970844e-06, "loss": 0.1208, "step": 33016 }, { "epoch": 3.32, "learning_rate": 3.7964332491678e-06, "loss": 0.3807, "step": 33017 }, { "epoch": 3.32, "learning_rate": 3.7952680749880465e-06, "loss": 0.0985, "step": 33018 }, { "epoch": 3.32, "learning_rate": 3.794103067565238e-06, "loss": 0.0407, "step": 33019 }, { "epoch": 3.32, "learning_rate": 3.7929382269067815e-06, "loss": 0.4826, "step": 33020 }, { "epoch": 3.32, "learning_rate": 3.791773553020108e-06, "loss": 0.037, "step": 33021 }, { "epoch": 3.32, "learning_rate": 3.790609045912615e-06, "loss": 0.3479, "step": 33022 }, { "epoch": 3.32, "learning_rate": 3.789444705591707e-06, "loss": 0.2831, "step": 33023 }, { "epoch": 3.32, "learning_rate": 3.7882805320648072e-06, "loss": 0.1841, "step": 33024 }, { "epoch": 3.32, "learning_rate": 3.787116525339327e-06, "loss": 0.2627, "step": 33025 }, { "epoch": 3.32, "learning_rate": 3.78595268542265e-06, "loss": 0.2007, "step": 33026 }, { "epoch": 3.32, "learning_rate": 3.7847890123222072e-06, "loss": 0.1614, "step": 33027 }, { "epoch": 3.32, "learning_rate": 3.783625506045393e-06, "loss": 0.1185, "step": 33028 }, { "epoch": 3.32, "learning_rate": 3.7824621665996108e-06, "loss": 0.1532, "step": 33029 }, { "epoch": 3.32, "learning_rate": 3.7812989939922684e-06, "loss": 0.3746, "step": 33030 }, { "epoch": 3.32, "learning_rate": 3.780135988230766e-06, "loss": 0.0515, "step": 33031 }, { "epoch": 3.32, "learning_rate": 3.7789731493225023e-06, "loss": 0.0478, "step": 33032 }, { "epoch": 3.32, "learning_rate": 3.777810477274877e-06, "loss": 0.2731, "step": 33033 }, { "epoch": 3.32, "learning_rate": 3.7766479720952907e-06, "loss": 0.2864, "step": 33034 }, { "epoch": 3.32, "learning_rate": 3.7754856337911415e-06, "loss": 0.1239, "step": 33035 }, { "epoch": 3.32, "learning_rate": 3.7743234623698262e-06, "loss": 0.1684, "step": 33036 }, { "epoch": 3.32, "learning_rate": 3.7731614578387396e-06, "loss": 0.2594, "step": 33037 }, { "epoch": 3.32, "learning_rate": 3.7719996202052686e-06, "loss": 0.1123, "step": 33038 }, { "epoch": 3.32, "learning_rate": 3.7708379494768273e-06, "loss": 0.1609, "step": 33039 }, { "epoch": 3.32, "learning_rate": 3.76967644566079e-06, "loss": 0.1602, "step": 33040 }, { "epoch": 3.32, "learning_rate": 3.7685151087645474e-06, "loss": 0.2953, "step": 33041 }, { "epoch": 3.32, "learning_rate": 3.7673539387955004e-06, "loss": 0.1411, "step": 33042 }, { "epoch": 3.32, "learning_rate": 3.7661929357610324e-06, "loss": 0.042, "step": 33043 }, { "epoch": 3.32, "learning_rate": 3.765032099668535e-06, "loss": 0.0949, "step": 33044 }, { "epoch": 3.32, "learning_rate": 3.7638714305253917e-06, "loss": 0.1161, "step": 33045 }, { "epoch": 3.32, "learning_rate": 3.7627109283389905e-06, "loss": 0.6414, "step": 33046 }, { "epoch": 3.32, "learning_rate": 3.761550593116718e-06, "loss": 0.0801, "step": 33047 }, { "epoch": 3.32, "learning_rate": 3.7603904248659527e-06, "loss": 0.3412, "step": 33048 }, { "epoch": 3.32, "learning_rate": 3.7592304235940812e-06, "loss": 0.2527, "step": 33049 }, { "epoch": 3.32, "learning_rate": 3.7580705893084877e-06, "loss": 0.05, "step": 33050 }, { "epoch": 3.32, "learning_rate": 3.7569109220165464e-06, "loss": 0.0585, "step": 33051 }, { "epoch": 3.32, "learning_rate": 3.755751421725635e-06, "loss": 0.3689, "step": 33052 }, { "epoch": 3.32, "learning_rate": 3.754592088443151e-06, "loss": 0.171, "step": 33053 }, { "epoch": 3.32, "learning_rate": 3.753432922176454e-06, "loss": 0.1316, "step": 33054 }, { "epoch": 3.32, "learning_rate": 3.7522739229329163e-06, "loss": 0.4403, "step": 33055 }, { "epoch": 3.32, "learning_rate": 3.7511150907199275e-06, "loss": 0.079, "step": 33056 }, { "epoch": 3.32, "learning_rate": 3.749956425544866e-06, "loss": 0.6797, "step": 33057 }, { "epoch": 3.32, "learning_rate": 3.748797927415082e-06, "loss": 0.1004, "step": 33058 }, { "epoch": 3.32, "learning_rate": 3.7476395963379695e-06, "loss": 0.0927, "step": 33059 }, { "epoch": 3.32, "learning_rate": 3.7464814323208896e-06, "loss": 0.2211, "step": 33060 }, { "epoch": 3.32, "learning_rate": 3.745323435371213e-06, "loss": 0.1515, "step": 33061 }, { "epoch": 3.32, "learning_rate": 3.7441656054963136e-06, "loss": 0.1023, "step": 33062 }, { "epoch": 3.32, "learning_rate": 3.7430079427035555e-06, "loss": 0.1884, "step": 33063 }, { "epoch": 3.32, "learning_rate": 3.7418504470003033e-06, "loss": 0.2122, "step": 33064 }, { "epoch": 3.32, "learning_rate": 3.7406931183939273e-06, "loss": 0.1216, "step": 33065 }, { "epoch": 3.32, "learning_rate": 3.739535956891792e-06, "loss": 0.0414, "step": 33066 }, { "epoch": 3.32, "learning_rate": 3.738378962501261e-06, "loss": 0.0949, "step": 33067 }, { "epoch": 3.32, "learning_rate": 3.737222135229692e-06, "loss": 0.1957, "step": 33068 }, { "epoch": 3.32, "learning_rate": 3.7360654750844457e-06, "loss": 0.4708, "step": 33069 }, { "epoch": 3.32, "learning_rate": 3.7349089820728967e-06, "loss": 0.0926, "step": 33070 }, { "epoch": 3.32, "learning_rate": 3.7337526562023985e-06, "loss": 0.2405, "step": 33071 }, { "epoch": 3.32, "learning_rate": 3.732596497480295e-06, "loss": 0.1111, "step": 33072 }, { "epoch": 3.32, "learning_rate": 3.7314405059139577e-06, "loss": 0.4695, "step": 33073 }, { "epoch": 3.32, "learning_rate": 3.730284681510744e-06, "loss": 0.3736, "step": 33074 }, { "epoch": 3.32, "learning_rate": 3.729129024278004e-06, "loss": 0.1352, "step": 33075 }, { "epoch": 3.32, "learning_rate": 3.7279735342230924e-06, "loss": 0.3761, "step": 33076 }, { "epoch": 3.32, "learning_rate": 3.7268182113533632e-06, "loss": 0.268, "step": 33077 }, { "epoch": 3.32, "learning_rate": 3.7256630556761674e-06, "loss": 0.2863, "step": 33078 }, { "epoch": 3.32, "learning_rate": 3.724508067198855e-06, "loss": 0.1011, "step": 33079 }, { "epoch": 3.32, "learning_rate": 3.7233532459287813e-06, "loss": 0.2699, "step": 33080 }, { "epoch": 3.32, "learning_rate": 3.722198591873289e-06, "loss": 0.0486, "step": 33081 }, { "epoch": 3.32, "learning_rate": 3.721044105039727e-06, "loss": 0.0749, "step": 33082 }, { "epoch": 3.32, "learning_rate": 3.719889785435445e-06, "loss": 0.0346, "step": 33083 }, { "epoch": 3.32, "learning_rate": 3.7187356330677813e-06, "loss": 0.1512, "step": 33084 }, { "epoch": 3.32, "learning_rate": 3.7175816479440993e-06, "loss": 0.1378, "step": 33085 }, { "epoch": 3.32, "learning_rate": 3.7164278300717205e-06, "loss": 0.0279, "step": 33086 }, { "epoch": 3.32, "learning_rate": 3.7152741794579914e-06, "loss": 0.2087, "step": 33087 }, { "epoch": 3.32, "learning_rate": 3.714120696110267e-06, "loss": 0.0761, "step": 33088 }, { "epoch": 3.32, "learning_rate": 3.712967380035881e-06, "loss": 0.3233, "step": 33089 }, { "epoch": 3.32, "learning_rate": 3.711814231242161e-06, "loss": 0.3857, "step": 33090 }, { "epoch": 3.32, "learning_rate": 3.7106612497364577e-06, "loss": 0.0846, "step": 33091 }, { "epoch": 3.32, "learning_rate": 3.7095084355261087e-06, "loss": 0.2014, "step": 33092 }, { "epoch": 3.32, "learning_rate": 3.708355788618445e-06, "loss": 0.0537, "step": 33093 }, { "epoch": 3.32, "learning_rate": 3.7072033090208034e-06, "loss": 0.0396, "step": 33094 }, { "epoch": 3.32, "learning_rate": 3.7060509967405153e-06, "loss": 0.6735, "step": 33095 }, { "epoch": 3.32, "learning_rate": 3.704898851784918e-06, "loss": 0.3073, "step": 33096 }, { "epoch": 3.32, "learning_rate": 3.703746874161342e-06, "loss": 0.2236, "step": 33097 }, { "epoch": 3.32, "learning_rate": 3.702595063877109e-06, "loss": 0.2743, "step": 33098 }, { "epoch": 3.32, "learning_rate": 3.701443420939572e-06, "loss": 0.1036, "step": 33099 }, { "epoch": 3.32, "learning_rate": 3.7002919453560358e-06, "loss": 0.172, "step": 33100 }, { "epoch": 3.32, "learning_rate": 3.6991406371338344e-06, "loss": 0.2597, "step": 33101 }, { "epoch": 3.32, "learning_rate": 3.6979894962802985e-06, "loss": 0.2147, "step": 33102 }, { "epoch": 3.32, "learning_rate": 3.696838522802759e-06, "loss": 0.252, "step": 33103 }, { "epoch": 3.32, "learning_rate": 3.6956877167085235e-06, "loss": 0.0721, "step": 33104 }, { "epoch": 3.32, "learning_rate": 3.694537078004926e-06, "loss": 0.1067, "step": 33105 }, { "epoch": 3.32, "learning_rate": 3.6933866066992905e-06, "loss": 0.414, "step": 33106 }, { "epoch": 3.32, "learning_rate": 3.6922363027989348e-06, "loss": 0.5757, "step": 33107 }, { "epoch": 3.32, "learning_rate": 3.6910861663111793e-06, "loss": 0.2432, "step": 33108 }, { "epoch": 3.33, "learning_rate": 3.689936197243342e-06, "loss": 0.4918, "step": 33109 }, { "epoch": 3.33, "learning_rate": 3.6887863956027436e-06, "loss": 0.4419, "step": 33110 }, { "epoch": 3.33, "learning_rate": 3.6876367613966976e-06, "loss": 0.0699, "step": 33111 }, { "epoch": 3.33, "learning_rate": 3.6864872946325223e-06, "loss": 0.4611, "step": 33112 }, { "epoch": 3.33, "learning_rate": 3.6853379953175314e-06, "loss": 0.1762, "step": 33113 }, { "epoch": 3.33, "learning_rate": 3.684188863459039e-06, "loss": 0.2123, "step": 33114 }, { "epoch": 3.33, "learning_rate": 3.68303989906436e-06, "loss": 0.1683, "step": 33115 }, { "epoch": 3.33, "learning_rate": 3.681891102140794e-06, "loss": 0.4264, "step": 33116 }, { "epoch": 3.33, "learning_rate": 3.680742472695673e-06, "loss": 0.0593, "step": 33117 }, { "epoch": 3.33, "learning_rate": 3.67959401073629e-06, "loss": 0.0796, "step": 33118 }, { "epoch": 3.33, "learning_rate": 3.6784457162699537e-06, "loss": 0.6094, "step": 33119 }, { "epoch": 3.33, "learning_rate": 3.6772975893039772e-06, "loss": 0.1435, "step": 33120 }, { "epoch": 3.33, "learning_rate": 3.6761496298456753e-06, "loss": 0.2016, "step": 33121 }, { "epoch": 3.33, "learning_rate": 3.6750018379023287e-06, "loss": 0.4691, "step": 33122 }, { "epoch": 3.33, "learning_rate": 3.6738542134812646e-06, "loss": 0.5641, "step": 33123 }, { "epoch": 3.33, "learning_rate": 3.672706756589774e-06, "loss": 0.1671, "step": 33124 }, { "epoch": 3.33, "learning_rate": 3.671559467235165e-06, "loss": 0.1385, "step": 33125 }, { "epoch": 3.33, "learning_rate": 3.670412345424734e-06, "loss": 0.275, "step": 33126 }, { "epoch": 3.33, "learning_rate": 3.6692653911657827e-06, "loss": 0.0503, "step": 33127 }, { "epoch": 3.33, "learning_rate": 3.668118604465608e-06, "loss": 0.1299, "step": 33128 }, { "epoch": 3.33, "learning_rate": 3.666971985331511e-06, "loss": 0.3155, "step": 33129 }, { "epoch": 3.33, "learning_rate": 3.6658255337707825e-06, "loss": 0.4292, "step": 33130 }, { "epoch": 3.33, "learning_rate": 3.6646792497907335e-06, "loss": 0.0678, "step": 33131 }, { "epoch": 3.33, "learning_rate": 3.663533133398641e-06, "loss": 0.1127, "step": 33132 }, { "epoch": 3.33, "learning_rate": 3.6623871846017965e-06, "loss": 0.444, "step": 33133 }, { "epoch": 3.33, "learning_rate": 3.661241403407507e-06, "loss": 0.0982, "step": 33134 }, { "epoch": 3.33, "learning_rate": 3.6600957898230668e-06, "loss": 0.1505, "step": 33135 }, { "epoch": 3.33, "learning_rate": 3.6589503438557405e-06, "loss": 0.3061, "step": 33136 }, { "epoch": 3.33, "learning_rate": 3.6578050655128416e-06, "loss": 0.0856, "step": 33137 }, { "epoch": 3.33, "learning_rate": 3.656659954801651e-06, "loss": 0.0537, "step": 33138 }, { "epoch": 3.33, "learning_rate": 3.6555150117294566e-06, "loss": 0.0334, "step": 33139 }, { "epoch": 3.33, "learning_rate": 3.654370236303539e-06, "loss": 0.0748, "step": 33140 }, { "epoch": 3.33, "learning_rate": 3.6532256285311894e-06, "loss": 0.1116, "step": 33141 }, { "epoch": 3.33, "learning_rate": 3.652081188419688e-06, "loss": 0.259, "step": 33142 }, { "epoch": 3.33, "learning_rate": 3.6509369159763163e-06, "loss": 0.08, "step": 33143 }, { "epoch": 3.33, "learning_rate": 3.6497928112083613e-06, "loss": 0.1543, "step": 33144 }, { "epoch": 3.33, "learning_rate": 3.6486488741231007e-06, "loss": 0.1895, "step": 33145 }, { "epoch": 3.33, "learning_rate": 3.647505104727812e-06, "loss": 0.0824, "step": 33146 }, { "epoch": 3.33, "learning_rate": 3.6463615030297793e-06, "loss": 0.1307, "step": 33147 }, { "epoch": 3.33, "learning_rate": 3.645218069036267e-06, "loss": 0.0381, "step": 33148 }, { "epoch": 3.33, "learning_rate": 3.6440748027545757e-06, "loss": 0.2258, "step": 33149 }, { "epoch": 3.33, "learning_rate": 3.642931704191953e-06, "loss": 0.4048, "step": 33150 }, { "epoch": 3.33, "learning_rate": 3.6417887733556897e-06, "loss": 0.0955, "step": 33151 }, { "epoch": 3.33, "learning_rate": 3.64064601025306e-06, "loss": 0.097, "step": 33152 }, { "epoch": 3.33, "learning_rate": 3.6395034148913285e-06, "loss": 0.2085, "step": 33153 }, { "epoch": 3.33, "learning_rate": 3.638360987277769e-06, "loss": 0.205, "step": 33154 }, { "epoch": 3.33, "learning_rate": 3.6372187274196523e-06, "loss": 0.0939, "step": 33155 }, { "epoch": 3.33, "learning_rate": 3.636076635324249e-06, "loss": 0.3083, "step": 33156 }, { "epoch": 3.33, "learning_rate": 3.634934710998824e-06, "loss": 0.056, "step": 33157 }, { "epoch": 3.33, "learning_rate": 3.633792954450641e-06, "loss": 0.0285, "step": 33158 }, { "epoch": 3.33, "learning_rate": 3.6326513656869743e-06, "loss": 0.213, "step": 33159 }, { "epoch": 3.33, "learning_rate": 3.6315099447150813e-06, "loss": 0.495, "step": 33160 }, { "epoch": 3.33, "learning_rate": 3.6303686915422265e-06, "loss": 0.5988, "step": 33161 }, { "epoch": 3.33, "learning_rate": 3.6292276061756703e-06, "loss": 0.1969, "step": 33162 }, { "epoch": 3.33, "learning_rate": 3.628086688622687e-06, "loss": 0.1746, "step": 33163 }, { "epoch": 3.33, "learning_rate": 3.6269459388905246e-06, "loss": 0.5781, "step": 33164 }, { "epoch": 3.33, "learning_rate": 3.625805356986437e-06, "loss": 0.161, "step": 33165 }, { "epoch": 3.33, "learning_rate": 3.6246649429176982e-06, "loss": 0.442, "step": 33166 }, { "epoch": 3.33, "learning_rate": 3.6235246966915625e-06, "loss": 0.382, "step": 33167 }, { "epoch": 3.33, "learning_rate": 3.622384618315271e-06, "loss": 0.0405, "step": 33168 }, { "epoch": 3.33, "learning_rate": 3.6212447077960943e-06, "loss": 0.3317, "step": 33169 }, { "epoch": 3.33, "learning_rate": 3.6201049651412834e-06, "loss": 0.3557, "step": 33170 }, { "epoch": 3.33, "learning_rate": 3.6189653903580887e-06, "loss": 0.1008, "step": 33171 }, { "epoch": 3.33, "learning_rate": 3.6178259834537588e-06, "loss": 1.008, "step": 33172 }, { "epoch": 3.33, "learning_rate": 3.6166867444355535e-06, "loss": 0.3862, "step": 33173 }, { "epoch": 3.33, "learning_rate": 3.6155476733107107e-06, "loss": 0.368, "step": 33174 }, { "epoch": 3.33, "learning_rate": 3.6144087700864913e-06, "loss": 0.1223, "step": 33175 }, { "epoch": 3.33, "learning_rate": 3.6132700347701263e-06, "loss": 0.1535, "step": 33176 }, { "epoch": 3.33, "learning_rate": 3.6121314673688867e-06, "loss": 0.0556, "step": 33177 }, { "epoch": 3.33, "learning_rate": 3.6109930678899993e-06, "loss": 0.5244, "step": 33178 }, { "epoch": 3.33, "learning_rate": 3.609854836340706e-06, "loss": 0.5954, "step": 33179 }, { "epoch": 3.33, "learning_rate": 3.6087167727282634e-06, "loss": 0.15, "step": 33180 }, { "epoch": 3.33, "learning_rate": 3.6075788770599127e-06, "loss": 0.1763, "step": 33181 }, { "epoch": 3.33, "learning_rate": 3.6064411493428785e-06, "loss": 0.0617, "step": 33182 }, { "epoch": 3.33, "learning_rate": 3.605303589584418e-06, "loss": 0.057, "step": 33183 }, { "epoch": 3.33, "learning_rate": 3.604166197791765e-06, "loss": 0.1587, "step": 33184 }, { "epoch": 3.33, "learning_rate": 3.6030289739721546e-06, "loss": 0.2112, "step": 33185 }, { "epoch": 3.33, "learning_rate": 3.601891918132827e-06, "loss": 0.2732, "step": 33186 }, { "epoch": 3.33, "learning_rate": 3.6007550302810197e-06, "loss": 0.7663, "step": 33187 }, { "epoch": 3.33, "learning_rate": 3.599618310423961e-06, "loss": 0.1819, "step": 33188 }, { "epoch": 3.33, "learning_rate": 3.598481758568891e-06, "loss": 0.1656, "step": 33189 }, { "epoch": 3.33, "learning_rate": 3.597345374723038e-06, "loss": 0.0482, "step": 33190 }, { "epoch": 3.33, "learning_rate": 3.5962091588936354e-06, "loss": 0.2892, "step": 33191 }, { "epoch": 3.33, "learning_rate": 3.595073111087911e-06, "loss": 0.2528, "step": 33192 }, { "epoch": 3.33, "learning_rate": 3.593937231313099e-06, "loss": 0.0388, "step": 33193 }, { "epoch": 3.33, "learning_rate": 3.5928015195764176e-06, "loss": 0.1855, "step": 33194 }, { "epoch": 3.33, "learning_rate": 3.591665975885113e-06, "loss": 0.1969, "step": 33195 }, { "epoch": 3.33, "learning_rate": 3.5905306002463945e-06, "loss": 0.5438, "step": 33196 }, { "epoch": 3.33, "learning_rate": 3.5893953926674883e-06, "loss": 0.0455, "step": 33197 }, { "epoch": 3.33, "learning_rate": 3.588260353155626e-06, "loss": 0.1051, "step": 33198 }, { "epoch": 3.33, "learning_rate": 3.587125481718038e-06, "loss": 0.2565, "step": 33199 }, { "epoch": 3.33, "learning_rate": 3.5859907783619184e-06, "loss": 0.328, "step": 33200 }, { "epoch": 3.33, "learning_rate": 3.584856243094512e-06, "loss": 0.1939, "step": 33201 }, { "epoch": 3.33, "learning_rate": 3.583721875923033e-06, "loss": 0.2888, "step": 33202 }, { "epoch": 3.33, "learning_rate": 3.582587676854698e-06, "loss": 0.1641, "step": 33203 }, { "epoch": 3.33, "learning_rate": 3.581453645896726e-06, "loss": 0.0286, "step": 33204 }, { "epoch": 3.33, "learning_rate": 3.5803197830563338e-06, "loss": 0.1091, "step": 33205 }, { "epoch": 3.33, "learning_rate": 3.5791860883407325e-06, "loss": 0.1587, "step": 33206 }, { "epoch": 3.33, "learning_rate": 3.5780525617571423e-06, "loss": 0.4146, "step": 33207 }, { "epoch": 3.34, "learning_rate": 3.5769192033127685e-06, "loss": 0.4279, "step": 33208 }, { "epoch": 3.34, "learning_rate": 3.5757860130148414e-06, "loss": 0.1512, "step": 33209 }, { "epoch": 3.34, "learning_rate": 3.5746529908705514e-06, "loss": 0.281, "step": 33210 }, { "epoch": 3.34, "learning_rate": 3.5735201368871107e-06, "loss": 0.0523, "step": 33211 }, { "epoch": 3.34, "learning_rate": 3.5723874510717426e-06, "loss": 0.2517, "step": 33212 }, { "epoch": 3.34, "learning_rate": 3.5712549334316547e-06, "loss": 0.3152, "step": 33213 }, { "epoch": 3.34, "learning_rate": 3.5701225839740313e-06, "loss": 0.5206, "step": 33214 }, { "epoch": 3.34, "learning_rate": 3.5689904027061004e-06, "loss": 0.1462, "step": 33215 }, { "epoch": 3.34, "learning_rate": 3.5678583896350624e-06, "loss": 0.0678, "step": 33216 }, { "epoch": 3.34, "learning_rate": 3.566726544768115e-06, "loss": 0.186, "step": 33217 }, { "epoch": 3.34, "learning_rate": 3.5655948681124663e-06, "loss": 0.3815, "step": 33218 }, { "epoch": 3.34, "learning_rate": 3.5644633596753128e-06, "loss": 0.3446, "step": 33219 }, { "epoch": 3.34, "learning_rate": 3.56333201946386e-06, "loss": 0.0229, "step": 33220 }, { "epoch": 3.34, "learning_rate": 3.562200847485304e-06, "loss": 0.1247, "step": 33221 }, { "epoch": 3.34, "learning_rate": 3.561069843746847e-06, "loss": 0.4634, "step": 33222 }, { "epoch": 3.34, "learning_rate": 3.5599390082556824e-06, "loss": 0.2185, "step": 33223 }, { "epoch": 3.34, "learning_rate": 3.5588083410190055e-06, "loss": 0.0848, "step": 33224 }, { "epoch": 3.34, "learning_rate": 3.5576778420440158e-06, "loss": 0.1721, "step": 33225 }, { "epoch": 3.34, "learning_rate": 3.5565475113378986e-06, "loss": 0.5129, "step": 33226 }, { "epoch": 3.34, "learning_rate": 3.5554173489078644e-06, "loss": 0.2267, "step": 33227 }, { "epoch": 3.34, "learning_rate": 3.554287354761091e-06, "loss": 0.2444, "step": 33228 }, { "epoch": 3.34, "learning_rate": 3.5531575289047623e-06, "loss": 0.2547, "step": 33229 }, { "epoch": 3.34, "learning_rate": 3.5520278713460863e-06, "loss": 0.0541, "step": 33230 }, { "epoch": 3.34, "learning_rate": 3.5508983820922436e-06, "loss": 0.356, "step": 33231 }, { "epoch": 3.34, "learning_rate": 3.549769061150422e-06, "loss": 0.033, "step": 33232 }, { "epoch": 3.34, "learning_rate": 3.548639908527809e-06, "loss": 0.108, "step": 33233 }, { "epoch": 3.34, "learning_rate": 3.5475109242315883e-06, "loss": 0.1131, "step": 33234 }, { "epoch": 3.34, "learning_rate": 3.546382108268945e-06, "loss": 0.5069, "step": 33235 }, { "epoch": 3.34, "learning_rate": 3.54525346064706e-06, "loss": 0.1291, "step": 33236 }, { "epoch": 3.34, "learning_rate": 3.54412498137312e-06, "loss": 0.2247, "step": 33237 }, { "epoch": 3.34, "learning_rate": 3.5429966704543038e-06, "loss": 0.0386, "step": 33238 }, { "epoch": 3.34, "learning_rate": 3.5418685278977946e-06, "loss": 0.2781, "step": 33239 }, { "epoch": 3.34, "learning_rate": 3.5407405537107605e-06, "loss": 0.0505, "step": 33240 }, { "epoch": 3.34, "learning_rate": 3.539612747900399e-06, "loss": 0.0167, "step": 33241 }, { "epoch": 3.34, "learning_rate": 3.5384851104738714e-06, "loss": 0.1653, "step": 33242 }, { "epoch": 3.34, "learning_rate": 3.5373576414383546e-06, "loss": 0.044, "step": 33243 }, { "epoch": 3.34, "learning_rate": 3.53623034080103e-06, "loss": 0.41, "step": 33244 }, { "epoch": 3.34, "learning_rate": 3.535103208569075e-06, "loss": 0.1148, "step": 33245 }, { "epoch": 3.34, "learning_rate": 3.5339762447496404e-06, "loss": 0.4697, "step": 33246 }, { "epoch": 3.34, "learning_rate": 3.5328494493499207e-06, "loss": 0.2305, "step": 33247 }, { "epoch": 3.34, "learning_rate": 3.5317228223770804e-06, "loss": 0.3822, "step": 33248 }, { "epoch": 3.34, "learning_rate": 3.5305963638382867e-06, "loss": 0.2328, "step": 33249 }, { "epoch": 3.34, "learning_rate": 3.529470073740707e-06, "loss": 0.1381, "step": 33250 }, { "epoch": 3.34, "learning_rate": 3.5283439520915096e-06, "loss": 0.093, "step": 33251 }, { "epoch": 3.34, "learning_rate": 3.5272179988978614e-06, "loss": 0.3987, "step": 33252 }, { "epoch": 3.34, "learning_rate": 3.5260922141669272e-06, "loss": 0.1519, "step": 33253 }, { "epoch": 3.34, "learning_rate": 3.524966597905871e-06, "loss": 0.4856, "step": 33254 }, { "epoch": 3.34, "learning_rate": 3.5238411501218538e-06, "loss": 0.0464, "step": 33255 }, { "epoch": 3.34, "learning_rate": 3.522715870822043e-06, "loss": 0.1186, "step": 33256 }, { "epoch": 3.34, "learning_rate": 3.5215907600135934e-06, "loss": 0.1972, "step": 33257 }, { "epoch": 3.34, "learning_rate": 3.520465817703662e-06, "loss": 0.3094, "step": 33258 }, { "epoch": 3.34, "learning_rate": 3.5193410438994235e-06, "loss": 0.1193, "step": 33259 }, { "epoch": 3.34, "learning_rate": 3.5182164386080152e-06, "loss": 0.4564, "step": 33260 }, { "epoch": 3.34, "learning_rate": 3.5170920018366084e-06, "loss": 0.0815, "step": 33261 }, { "epoch": 3.34, "learning_rate": 3.5159677335923535e-06, "loss": 0.1597, "step": 33262 }, { "epoch": 3.34, "learning_rate": 3.5148436338824055e-06, "loss": 0.1154, "step": 33263 }, { "epoch": 3.34, "learning_rate": 3.5137197027139145e-06, "loss": 0.1086, "step": 33264 }, { "epoch": 3.34, "learning_rate": 3.512595940094039e-06, "loss": 0.4274, "step": 33265 }, { "epoch": 3.34, "learning_rate": 3.5114723460299257e-06, "loss": 0.0901, "step": 33266 }, { "epoch": 3.34, "learning_rate": 3.510348920528723e-06, "loss": 0.1467, "step": 33267 }, { "epoch": 3.34, "learning_rate": 3.509225663597585e-06, "loss": 0.3482, "step": 33268 }, { "epoch": 3.34, "learning_rate": 3.508102575243656e-06, "loss": 0.0584, "step": 33269 }, { "epoch": 3.34, "learning_rate": 3.506979655474086e-06, "loss": 0.4402, "step": 33270 }, { "epoch": 3.34, "learning_rate": 3.5058569042960174e-06, "loss": 0.0831, "step": 33271 }, { "epoch": 3.34, "learning_rate": 3.5047343217165907e-06, "loss": 0.1896, "step": 33272 }, { "epoch": 3.34, "learning_rate": 3.5036119077429698e-06, "loss": 0.36, "step": 33273 }, { "epoch": 3.34, "learning_rate": 3.5024896623822756e-06, "loss": 0.2675, "step": 33274 }, { "epoch": 3.34, "learning_rate": 3.5013675856416496e-06, "loss": 0.0993, "step": 33275 }, { "epoch": 3.34, "learning_rate": 3.5002456775282453e-06, "loss": 0.1837, "step": 33276 }, { "epoch": 3.34, "learning_rate": 3.4991239380492045e-06, "loss": 0.2347, "step": 33277 }, { "epoch": 3.34, "learning_rate": 3.4980023672116448e-06, "loss": 0.1696, "step": 33278 }, { "epoch": 3.34, "learning_rate": 3.4968809650227227e-06, "loss": 0.3857, "step": 33279 }, { "epoch": 3.34, "learning_rate": 3.4957597314895675e-06, "loss": 0.2188, "step": 33280 }, { "epoch": 3.34, "learning_rate": 3.494638666619315e-06, "loss": 0.1914, "step": 33281 }, { "epoch": 3.34, "learning_rate": 3.493517770419098e-06, "loss": 0.0506, "step": 33282 }, { "epoch": 3.34, "learning_rate": 3.4923970428960493e-06, "loss": 0.2356, "step": 33283 }, { "epoch": 3.34, "learning_rate": 3.491276484057304e-06, "loss": 0.2143, "step": 33284 }, { "epoch": 3.34, "learning_rate": 3.49015609390999e-06, "loss": 0.2125, "step": 33285 }, { "epoch": 3.34, "learning_rate": 3.4890358724612303e-06, "loss": 0.1833, "step": 33286 }, { "epoch": 3.34, "learning_rate": 3.487915819718174e-06, "loss": 0.0133, "step": 33287 }, { "epoch": 3.34, "learning_rate": 3.486795935687931e-06, "loss": 0.3208, "step": 33288 }, { "epoch": 3.34, "learning_rate": 3.485676220377626e-06, "loss": 0.0598, "step": 33289 }, { "epoch": 3.34, "learning_rate": 3.484556673794397e-06, "loss": 0.1221, "step": 33290 }, { "epoch": 3.34, "learning_rate": 3.4834372959453674e-06, "loss": 0.2366, "step": 33291 }, { "epoch": 3.34, "learning_rate": 3.4823180868376426e-06, "loss": 0.2898, "step": 33292 }, { "epoch": 3.34, "learning_rate": 3.481199046478366e-06, "loss": 0.2041, "step": 33293 }, { "epoch": 3.34, "learning_rate": 3.480080174874649e-06, "loss": 0.1013, "step": 33294 }, { "epoch": 3.34, "learning_rate": 3.478961472033613e-06, "loss": 0.1395, "step": 33295 }, { "epoch": 3.34, "learning_rate": 3.477842937962378e-06, "loss": 0.3414, "step": 33296 }, { "epoch": 3.34, "learning_rate": 3.476724572668062e-06, "loss": 0.0423, "step": 33297 }, { "epoch": 3.34, "learning_rate": 3.4756063761577794e-06, "loss": 0.0855, "step": 33298 }, { "epoch": 3.34, "learning_rate": 3.4744883484386445e-06, "loss": 0.1159, "step": 33299 }, { "epoch": 3.34, "learning_rate": 3.4733704895177786e-06, "loss": 0.0566, "step": 33300 }, { "epoch": 3.34, "learning_rate": 3.4722527994022926e-06, "loss": 0.2199, "step": 33301 }, { "epoch": 3.34, "learning_rate": 3.4711352780992968e-06, "loss": 0.0891, "step": 33302 }, { "epoch": 3.34, "learning_rate": 3.4700179256159026e-06, "loss": 0.0713, "step": 33303 }, { "epoch": 3.34, "learning_rate": 3.4689007419592178e-06, "loss": 0.3304, "step": 33304 }, { "epoch": 3.34, "learning_rate": 3.4677837271363664e-06, "loss": 0.4079, "step": 33305 }, { "epoch": 3.35, "learning_rate": 3.46666688115444e-06, "loss": 0.027, "step": 33306 }, { "epoch": 3.35, "learning_rate": 3.4655502040205457e-06, "loss": 0.1378, "step": 33307 }, { "epoch": 3.35, "learning_rate": 3.464433695741801e-06, "loss": 0.206, "step": 33308 }, { "epoch": 3.35, "learning_rate": 3.463317356325314e-06, "loss": 0.3518, "step": 33309 }, { "epoch": 3.35, "learning_rate": 3.4622011857781654e-06, "loss": 0.1721, "step": 33310 }, { "epoch": 3.35, "learning_rate": 3.46108518410748e-06, "loss": 0.2217, "step": 33311 }, { "epoch": 3.35, "learning_rate": 3.4599693513203544e-06, "loss": 0.2759, "step": 33312 }, { "epoch": 3.35, "learning_rate": 3.4588536874238873e-06, "loss": 0.3518, "step": 33313 }, { "epoch": 3.35, "learning_rate": 3.4577381924251762e-06, "loss": 0.2822, "step": 33314 }, { "epoch": 3.35, "learning_rate": 3.4566228663313248e-06, "loss": 0.2315, "step": 33315 }, { "epoch": 3.35, "learning_rate": 3.4555077091494248e-06, "loss": 0.1858, "step": 33316 }, { "epoch": 3.35, "learning_rate": 3.4543927208865764e-06, "loss": 0.2162, "step": 33317 }, { "epoch": 3.35, "learning_rate": 3.4532779015498716e-06, "loss": 0.2726, "step": 33318 }, { "epoch": 3.35, "learning_rate": 3.4521632511464143e-06, "loss": 0.0684, "step": 33319 }, { "epoch": 3.35, "learning_rate": 3.451048769683288e-06, "loss": 0.0714, "step": 33320 }, { "epoch": 3.35, "learning_rate": 3.449934457167582e-06, "loss": 0.2152, "step": 33321 }, { "epoch": 3.35, "learning_rate": 3.4488203136063958e-06, "loss": 0.2422, "step": 33322 }, { "epoch": 3.35, "learning_rate": 3.4477063390068274e-06, "loss": 0.095, "step": 33323 }, { "epoch": 3.35, "learning_rate": 3.4465925333759383e-06, "loss": 0.4583, "step": 33324 }, { "epoch": 3.35, "learning_rate": 3.4454788967208394e-06, "loss": 0.0593, "step": 33325 }, { "epoch": 3.35, "learning_rate": 3.444365429048615e-06, "loss": 0.0848, "step": 33326 }, { "epoch": 3.35, "learning_rate": 3.443252130366342e-06, "loss": 0.1772, "step": 33327 }, { "epoch": 3.35, "learning_rate": 3.442139000681113e-06, "loss": 0.2123, "step": 33328 }, { "epoch": 3.35, "learning_rate": 3.441026040000007e-06, "loss": 0.0814, "step": 33329 }, { "epoch": 3.35, "learning_rate": 3.4399132483301066e-06, "loss": 0.2241, "step": 33330 }, { "epoch": 3.35, "learning_rate": 3.4388006256784955e-06, "loss": 0.1462, "step": 33331 }, { "epoch": 3.35, "learning_rate": 3.4376881720522514e-06, "loss": 0.2538, "step": 33332 }, { "epoch": 3.35, "learning_rate": 3.4365758874584584e-06, "loss": 0.1264, "step": 33333 }, { "epoch": 3.35, "learning_rate": 3.435463771904188e-06, "loss": 0.1537, "step": 33334 }, { "epoch": 3.35, "learning_rate": 3.4343518253965213e-06, "loss": 0.119, "step": 33335 }, { "epoch": 3.35, "learning_rate": 3.4332400479425284e-06, "loss": 0.3378, "step": 33336 }, { "epoch": 3.35, "learning_rate": 3.4321284395493014e-06, "loss": 0.2314, "step": 33337 }, { "epoch": 3.35, "learning_rate": 3.4310170002238938e-06, "loss": 0.0754, "step": 33338 }, { "epoch": 3.35, "learning_rate": 3.429905729973384e-06, "loss": 0.0636, "step": 33339 }, { "epoch": 3.35, "learning_rate": 3.4287946288048487e-06, "loss": 0.3053, "step": 33340 }, { "epoch": 3.35, "learning_rate": 3.42768369672536e-06, "loss": 0.4648, "step": 33341 }, { "epoch": 3.35, "learning_rate": 3.4265729337419817e-06, "loss": 0.0567, "step": 33342 }, { "epoch": 3.35, "learning_rate": 3.4254623398617814e-06, "loss": 0.1521, "step": 33343 }, { "epoch": 3.35, "learning_rate": 3.4243519150918336e-06, "loss": 0.0846, "step": 33344 }, { "epoch": 3.35, "learning_rate": 3.423241659439199e-06, "loss": 0.0419, "step": 33345 }, { "epoch": 3.35, "learning_rate": 3.4221315729109424e-06, "loss": 0.1899, "step": 33346 }, { "epoch": 3.35, "learning_rate": 3.4210216555141315e-06, "loss": 0.4428, "step": 33347 }, { "epoch": 3.35, "learning_rate": 3.419911907255827e-06, "loss": 0.3365, "step": 33348 }, { "epoch": 3.35, "learning_rate": 3.4188023281430937e-06, "loss": 0.7355, "step": 33349 }, { "epoch": 3.35, "learning_rate": 3.417692918182982e-06, "loss": 0.1529, "step": 33350 }, { "epoch": 3.35, "learning_rate": 3.416583677382573e-06, "loss": 0.175, "step": 33351 }, { "epoch": 3.35, "learning_rate": 3.415474605748905e-06, "loss": 0.1154, "step": 33352 }, { "epoch": 3.35, "learning_rate": 3.4143657032890386e-06, "loss": 0.3345, "step": 33353 }, { "epoch": 3.35, "learning_rate": 3.413256970010041e-06, "loss": 0.3797, "step": 33354 }, { "epoch": 3.35, "learning_rate": 3.412148405918971e-06, "loss": 0.1217, "step": 33355 }, { "epoch": 3.35, "learning_rate": 3.411040011022859e-06, "loss": 0.1806, "step": 33356 }, { "epoch": 3.35, "learning_rate": 3.4099317853287794e-06, "loss": 0.104, "step": 33357 }, { "epoch": 3.35, "learning_rate": 3.408823728843776e-06, "loss": 0.1395, "step": 33358 }, { "epoch": 3.35, "learning_rate": 3.4077158415749077e-06, "loss": 0.4592, "step": 33359 }, { "epoch": 3.35, "learning_rate": 3.4066081235292178e-06, "loss": 0.0849, "step": 33360 }, { "epoch": 3.35, "learning_rate": 3.4055005747137544e-06, "loss": 0.3668, "step": 33361 }, { "epoch": 3.35, "learning_rate": 3.4043931951355722e-06, "loss": 0.259, "step": 33362 }, { "epoch": 3.35, "learning_rate": 3.4032859848017117e-06, "loss": 0.0636, "step": 33363 }, { "epoch": 3.35, "learning_rate": 3.402178943719221e-06, "loss": 0.4728, "step": 33364 }, { "epoch": 3.35, "learning_rate": 3.4010720718951473e-06, "loss": 0.153, "step": 33365 }, { "epoch": 3.35, "learning_rate": 3.399965369336532e-06, "loss": 0.0583, "step": 33366 }, { "epoch": 3.35, "learning_rate": 3.398858836050409e-06, "loss": 0.1006, "step": 33367 }, { "epoch": 3.35, "learning_rate": 3.3977524720438368e-06, "loss": 0.2459, "step": 33368 }, { "epoch": 3.35, "learning_rate": 3.396646277323856e-06, "loss": 0.3013, "step": 33369 }, { "epoch": 3.35, "learning_rate": 3.3955402518974843e-06, "loss": 0.1945, "step": 33370 }, { "epoch": 3.35, "learning_rate": 3.394434395771776e-06, "loss": 0.4265, "step": 33371 }, { "epoch": 3.35, "learning_rate": 3.393328708953769e-06, "loss": 0.2, "step": 33372 }, { "epoch": 3.35, "learning_rate": 3.3922231914504977e-06, "loss": 0.1422, "step": 33373 }, { "epoch": 3.35, "learning_rate": 3.391117843268996e-06, "loss": 0.1029, "step": 33374 }, { "epoch": 3.35, "learning_rate": 3.3900126644162954e-06, "loss": 0.5312, "step": 33375 }, { "epoch": 3.35, "learning_rate": 3.3889076548994303e-06, "loss": 0.1178, "step": 33376 }, { "epoch": 3.35, "learning_rate": 3.3878028147254382e-06, "loss": 0.1815, "step": 33377 }, { "epoch": 3.35, "learning_rate": 3.38669814390134e-06, "loss": 0.0542, "step": 33378 }, { "epoch": 3.35, "learning_rate": 3.3855936424341737e-06, "loss": 0.1029, "step": 33379 }, { "epoch": 3.35, "learning_rate": 3.3844893103309637e-06, "loss": 0.1175, "step": 33380 }, { "epoch": 3.35, "learning_rate": 3.383385147598741e-06, "loss": 0.2568, "step": 33381 }, { "epoch": 3.35, "learning_rate": 3.3822811542445196e-06, "loss": 0.4319, "step": 33382 }, { "epoch": 3.35, "learning_rate": 3.3811773302753514e-06, "loss": 0.0244, "step": 33383 }, { "epoch": 3.35, "learning_rate": 3.3800736756982363e-06, "loss": 0.4577, "step": 33384 }, { "epoch": 3.35, "learning_rate": 3.3789701905201996e-06, "loss": 0.1776, "step": 33385 }, { "epoch": 3.35, "learning_rate": 3.3778668747482755e-06, "loss": 0.2992, "step": 33386 }, { "epoch": 3.35, "learning_rate": 3.376763728389485e-06, "loss": 0.119, "step": 33387 }, { "epoch": 3.35, "learning_rate": 3.375660751450829e-06, "loss": 0.2795, "step": 33388 }, { "epoch": 3.35, "learning_rate": 3.374557943939345e-06, "loss": 0.5652, "step": 33389 }, { "epoch": 3.35, "learning_rate": 3.373455305862048e-06, "loss": 0.2246, "step": 33390 }, { "epoch": 3.35, "learning_rate": 3.3723528372259516e-06, "loss": 0.1387, "step": 33391 }, { "epoch": 3.35, "learning_rate": 3.371250538038071e-06, "loss": 0.2321, "step": 33392 }, { "epoch": 3.35, "learning_rate": 3.3701484083054168e-06, "loss": 0.1957, "step": 33393 }, { "epoch": 3.35, "learning_rate": 3.3690464480350204e-06, "loss": 0.201, "step": 33394 }, { "epoch": 3.35, "learning_rate": 3.3679446572338722e-06, "loss": 0.0775, "step": 33395 }, { "epoch": 3.35, "learning_rate": 3.366843035908991e-06, "loss": 0.0863, "step": 33396 }, { "epoch": 3.35, "learning_rate": 3.3657415840673966e-06, "loss": 0.2922, "step": 33397 }, { "epoch": 3.35, "learning_rate": 3.3646403017160875e-06, "loss": 0.088, "step": 33398 }, { "epoch": 3.35, "learning_rate": 3.363539188862068e-06, "loss": 0.1266, "step": 33399 }, { "epoch": 3.35, "learning_rate": 3.362438245512359e-06, "loss": 0.1056, "step": 33400 }, { "epoch": 3.35, "learning_rate": 3.3613374716739655e-06, "loss": 0.1534, "step": 33401 }, { "epoch": 3.35, "learning_rate": 3.3602368673538743e-06, "loss": 0.1834, "step": 33402 }, { "epoch": 3.35, "learning_rate": 3.3591364325591067e-06, "loss": 0.0756, "step": 33403 }, { "epoch": 3.35, "learning_rate": 3.358036167296661e-06, "loss": 0.0758, "step": 33404 }, { "epoch": 3.36, "learning_rate": 3.3569360715735375e-06, "loss": 0.3925, "step": 33405 }, { "epoch": 3.36, "learning_rate": 3.355836145396738e-06, "loss": 0.0469, "step": 33406 }, { "epoch": 3.36, "learning_rate": 3.3547363887732595e-06, "loss": 0.2391, "step": 33407 }, { "epoch": 3.36, "learning_rate": 3.3536368017101037e-06, "loss": 0.0972, "step": 33408 }, { "epoch": 3.36, "learning_rate": 3.3525373842142648e-06, "loss": 0.2337, "step": 33409 }, { "epoch": 3.36, "learning_rate": 3.351438136292744e-06, "loss": 0.1295, "step": 33410 }, { "epoch": 3.36, "learning_rate": 3.350339057952532e-06, "loss": 0.1021, "step": 33411 }, { "epoch": 3.36, "learning_rate": 3.349240149200623e-06, "loss": 0.0685, "step": 33412 }, { "epoch": 3.36, "learning_rate": 3.348141410044012e-06, "loss": 0.2796, "step": 33413 }, { "epoch": 3.36, "learning_rate": 3.3470428404896836e-06, "loss": 0.0985, "step": 33414 }, { "epoch": 3.36, "learning_rate": 3.3459444405446484e-06, "loss": 0.1218, "step": 33415 }, { "epoch": 3.36, "learning_rate": 3.344846210215874e-06, "loss": 0.0784, "step": 33416 }, { "epoch": 3.36, "learning_rate": 3.343748149510355e-06, "loss": 0.0675, "step": 33417 }, { "epoch": 3.36, "learning_rate": 3.3426502584350895e-06, "loss": 0.5742, "step": 33418 }, { "epoch": 3.36, "learning_rate": 3.3415525369970612e-06, "loss": 0.3221, "step": 33419 }, { "epoch": 3.36, "learning_rate": 3.3404549852032385e-06, "loss": 0.1518, "step": 33420 }, { "epoch": 3.36, "learning_rate": 3.3393576030606253e-06, "loss": 0.3172, "step": 33421 }, { "epoch": 3.36, "learning_rate": 3.3382603905761967e-06, "loss": 0.1549, "step": 33422 }, { "epoch": 3.36, "learning_rate": 3.3371633477569363e-06, "loss": 0.1006, "step": 33423 }, { "epoch": 3.36, "learning_rate": 3.336066474609826e-06, "loss": 0.1518, "step": 33424 }, { "epoch": 3.36, "learning_rate": 3.334969771141839e-06, "loss": 0.1913, "step": 33425 }, { "epoch": 3.36, "learning_rate": 3.3338732373599744e-06, "loss": 0.1816, "step": 33426 }, { "epoch": 3.36, "learning_rate": 3.3327768732711894e-06, "loss": 0.0919, "step": 33427 }, { "epoch": 3.36, "learning_rate": 3.3316806788824616e-06, "loss": 0.2122, "step": 33428 }, { "epoch": 3.36, "learning_rate": 3.3305846542007823e-06, "loss": 0.5159, "step": 33429 }, { "epoch": 3.36, "learning_rate": 3.3294887992331125e-06, "loss": 0.4142, "step": 33430 }, { "epoch": 3.36, "learning_rate": 3.3283931139864267e-06, "loss": 0.0998, "step": 33431 }, { "epoch": 3.36, "learning_rate": 3.3272975984677022e-06, "loss": 0.1384, "step": 33432 }, { "epoch": 3.36, "learning_rate": 3.3262022526839174e-06, "loss": 0.1858, "step": 33433 }, { "epoch": 3.36, "learning_rate": 3.3251070766420234e-06, "loss": 0.8629, "step": 33434 }, { "epoch": 3.36, "learning_rate": 3.324012070349004e-06, "loss": 0.1283, "step": 33435 }, { "epoch": 3.36, "learning_rate": 3.322917233811824e-06, "loss": 0.1281, "step": 33436 }, { "epoch": 3.36, "learning_rate": 3.3218225670374515e-06, "loss": 0.3095, "step": 33437 }, { "epoch": 3.36, "learning_rate": 3.3207280700328503e-06, "loss": 0.4854, "step": 33438 }, { "epoch": 3.36, "learning_rate": 3.319633742804985e-06, "loss": 0.1642, "step": 33439 }, { "epoch": 3.36, "learning_rate": 3.3185395853608237e-06, "loss": 0.1703, "step": 33440 }, { "epoch": 3.36, "learning_rate": 3.3174455977073237e-06, "loss": 0.0745, "step": 33441 }, { "epoch": 3.36, "learning_rate": 3.3163517798514496e-06, "loss": 0.0496, "step": 33442 }, { "epoch": 3.36, "learning_rate": 3.315258131800163e-06, "loss": 0.2756, "step": 33443 }, { "epoch": 3.36, "learning_rate": 3.314164653560424e-06, "loss": 0.4809, "step": 33444 }, { "epoch": 3.36, "learning_rate": 3.313071345139188e-06, "loss": 0.6274, "step": 33445 }, { "epoch": 3.36, "learning_rate": 3.3119782065434057e-06, "loss": 0.4302, "step": 33446 }, { "epoch": 3.36, "learning_rate": 3.3108852377800547e-06, "loss": 0.0379, "step": 33447 }, { "epoch": 3.36, "learning_rate": 3.3097924388560665e-06, "loss": 0.1297, "step": 33448 }, { "epoch": 3.36, "learning_rate": 3.3086998097784083e-06, "loss": 0.1411, "step": 33449 }, { "epoch": 3.36, "learning_rate": 3.307607350554035e-06, "loss": 0.1117, "step": 33450 }, { "epoch": 3.36, "learning_rate": 3.3065150611898907e-06, "loss": 0.5792, "step": 33451 }, { "epoch": 3.36, "learning_rate": 3.3054229416929304e-06, "loss": 0.1961, "step": 33452 }, { "epoch": 3.36, "learning_rate": 3.3043309920701047e-06, "loss": 0.0497, "step": 33453 }, { "epoch": 3.36, "learning_rate": 3.3032392123283616e-06, "loss": 0.2147, "step": 33454 }, { "epoch": 3.36, "learning_rate": 3.3021476024746457e-06, "loss": 0.3637, "step": 33455 }, { "epoch": 3.36, "learning_rate": 3.3010561625159075e-06, "loss": 0.0819, "step": 33456 }, { "epoch": 3.36, "learning_rate": 3.2999648924590854e-06, "loss": 0.121, "step": 33457 }, { "epoch": 3.36, "learning_rate": 3.29887379231114e-06, "loss": 0.1736, "step": 33458 }, { "epoch": 3.36, "learning_rate": 3.2977828620789997e-06, "loss": 0.2829, "step": 33459 }, { "epoch": 3.36, "learning_rate": 3.296692101769605e-06, "loss": 0.1223, "step": 33460 }, { "epoch": 3.36, "learning_rate": 3.295601511389914e-06, "loss": 0.2535, "step": 33461 }, { "epoch": 3.36, "learning_rate": 3.2945110909468513e-06, "loss": 0.4745, "step": 33462 }, { "epoch": 3.36, "learning_rate": 3.2934208404473544e-06, "loss": 0.4202, "step": 33463 }, { "epoch": 3.36, "learning_rate": 3.2923307598983743e-06, "loss": 0.0043, "step": 33464 }, { "epoch": 3.36, "learning_rate": 3.291240849306846e-06, "loss": 0.1179, "step": 33465 }, { "epoch": 3.36, "learning_rate": 3.2901511086796897e-06, "loss": 0.1654, "step": 33466 }, { "epoch": 3.36, "learning_rate": 3.289061538023854e-06, "loss": 0.4182, "step": 33467 }, { "epoch": 3.36, "learning_rate": 3.2879721373462736e-06, "loss": 0.0327, "step": 33468 }, { "epoch": 3.36, "learning_rate": 3.2868829066538754e-06, "loss": 0.0659, "step": 33469 }, { "epoch": 3.36, "learning_rate": 3.285793845953591e-06, "loss": 0.0663, "step": 33470 }, { "epoch": 3.36, "learning_rate": 3.284704955252351e-06, "loss": 0.205, "step": 33471 }, { "epoch": 3.36, "learning_rate": 3.2836162345570877e-06, "loss": 0.4408, "step": 33472 }, { "epoch": 3.36, "learning_rate": 3.2825276838747242e-06, "loss": 0.2133, "step": 33473 }, { "epoch": 3.36, "learning_rate": 3.2814393032121894e-06, "loss": 0.1543, "step": 33474 }, { "epoch": 3.36, "learning_rate": 3.2803510925764203e-06, "loss": 0.1642, "step": 33475 }, { "epoch": 3.36, "learning_rate": 3.279263051974325e-06, "loss": 0.5191, "step": 33476 }, { "epoch": 3.36, "learning_rate": 3.278175181412828e-06, "loss": 0.2126, "step": 33477 }, { "epoch": 3.36, "learning_rate": 3.277087480898867e-06, "loss": 0.153, "step": 33478 }, { "epoch": 3.36, "learning_rate": 3.27599995043936e-06, "loss": 0.1554, "step": 33479 }, { "epoch": 3.36, "learning_rate": 3.2749125900412114e-06, "loss": 0.2725, "step": 33480 }, { "epoch": 3.36, "learning_rate": 3.273825399711359e-06, "loss": 0.124, "step": 33481 }, { "epoch": 3.36, "learning_rate": 3.2727383794567134e-06, "loss": 0.1325, "step": 33482 }, { "epoch": 3.36, "learning_rate": 3.2716515292841963e-06, "loss": 0.13, "step": 33483 }, { "epoch": 3.36, "learning_rate": 3.270564849200716e-06, "loss": 0.1332, "step": 33484 }, { "epoch": 3.36, "learning_rate": 3.2694783392131953e-06, "loss": 0.1876, "step": 33485 }, { "epoch": 3.36, "learning_rate": 3.2683919993285474e-06, "loss": 0.5184, "step": 33486 }, { "epoch": 3.36, "learning_rate": 3.2673058295536816e-06, "loss": 0.0001, "step": 33487 }, { "epoch": 3.36, "learning_rate": 3.2662198298955103e-06, "loss": 0.156, "step": 33488 }, { "epoch": 3.36, "learning_rate": 3.265134000360941e-06, "loss": 0.0965, "step": 33489 }, { "epoch": 3.36, "learning_rate": 3.264048340956901e-06, "loss": 0.5412, "step": 33490 }, { "epoch": 3.36, "learning_rate": 3.2629628516902823e-06, "loss": 0.6105, "step": 33491 }, { "epoch": 3.36, "learning_rate": 3.2618775325679884e-06, "loss": 0.0304, "step": 33492 }, { "epoch": 3.36, "learning_rate": 3.2607923835969445e-06, "loss": 0.7795, "step": 33493 }, { "epoch": 3.36, "learning_rate": 3.2597074047840447e-06, "loss": 0.163, "step": 33494 }, { "epoch": 3.36, "learning_rate": 3.2586225961361837e-06, "loss": 0.1959, "step": 33495 }, { "epoch": 3.36, "learning_rate": 3.2575379576602826e-06, "loss": 0.3159, "step": 33496 }, { "epoch": 3.36, "learning_rate": 3.2564534893632457e-06, "loss": 0.1402, "step": 33497 }, { "epoch": 3.36, "learning_rate": 3.255369191251951e-06, "loss": 0.0353, "step": 33498 }, { "epoch": 3.36, "learning_rate": 3.2542850633333164e-06, "loss": 0.2395, "step": 33499 }, { "epoch": 3.36, "learning_rate": 3.2532011056142396e-06, "loss": 0.0549, "step": 33500 }, { "epoch": 3.36, "learning_rate": 3.2521173181016153e-06, "loss": 0.4081, "step": 33501 }, { "epoch": 3.36, "learning_rate": 3.2510337008023407e-06, "loss": 0.1329, "step": 33502 }, { "epoch": 3.37, "learning_rate": 3.2499502537233073e-06, "loss": 0.0216, "step": 33503 }, { "epoch": 3.37, "learning_rate": 3.2488669768714262e-06, "loss": 0.0913, "step": 33504 }, { "epoch": 3.37, "learning_rate": 3.2477838702535722e-06, "loss": 0.2456, "step": 33505 }, { "epoch": 3.37, "learning_rate": 3.2467009338766364e-06, "loss": 0.0218, "step": 33506 }, { "epoch": 3.37, "learning_rate": 3.2456181677475327e-06, "loss": 0.1642, "step": 33507 }, { "epoch": 3.37, "learning_rate": 3.2445355718731296e-06, "loss": 0.4919, "step": 33508 }, { "epoch": 3.37, "learning_rate": 3.243453146260321e-06, "loss": 0.1275, "step": 33509 }, { "epoch": 3.37, "learning_rate": 3.2423708909160023e-06, "loss": 0.3237, "step": 33510 }, { "epoch": 3.37, "learning_rate": 3.24128880584706e-06, "loss": 0.1827, "step": 33511 }, { "epoch": 3.37, "learning_rate": 3.240206891060367e-06, "loss": 0.2837, "step": 33512 }, { "epoch": 3.37, "learning_rate": 3.2391251465628225e-06, "loss": 0.077, "step": 33513 }, { "epoch": 3.37, "learning_rate": 3.238043572361302e-06, "loss": 0.3792, "step": 33514 }, { "epoch": 3.37, "learning_rate": 3.236962168462697e-06, "loss": 0.1167, "step": 33515 }, { "epoch": 3.37, "learning_rate": 3.2358809348738817e-06, "loss": 0.0714, "step": 33516 }, { "epoch": 3.37, "learning_rate": 3.234799871601737e-06, "loss": 0.1039, "step": 33517 }, { "epoch": 3.37, "learning_rate": 3.2337189786531442e-06, "loss": 0.5001, "step": 33518 }, { "epoch": 3.37, "learning_rate": 3.232638256034981e-06, "loss": 0.0561, "step": 33519 }, { "epoch": 3.37, "learning_rate": 3.2315577037541254e-06, "loss": 0.1856, "step": 33520 }, { "epoch": 3.37, "learning_rate": 3.2304773218174487e-06, "loss": 0.5088, "step": 33521 }, { "epoch": 3.37, "learning_rate": 3.229397110231842e-06, "loss": 0.2399, "step": 33522 }, { "epoch": 3.37, "learning_rate": 3.2283170690041632e-06, "loss": 0.2036, "step": 33523 }, { "epoch": 3.37, "learning_rate": 3.227237198141283e-06, "loss": 0.6364, "step": 33524 }, { "epoch": 3.37, "learning_rate": 3.226157497650093e-06, "loss": 0.1036, "step": 33525 }, { "epoch": 3.37, "learning_rate": 3.225077967537444e-06, "loss": 0.1999, "step": 33526 }, { "epoch": 3.37, "learning_rate": 3.223998607810208e-06, "loss": 0.1694, "step": 33527 }, { "epoch": 3.37, "learning_rate": 3.222919418475262e-06, "loss": 0.5941, "step": 33528 }, { "epoch": 3.37, "learning_rate": 3.221840399539474e-06, "loss": 0.1072, "step": 33529 }, { "epoch": 3.37, "learning_rate": 3.2207615510097054e-06, "loss": 0.1377, "step": 33530 }, { "epoch": 3.37, "learning_rate": 3.219682872892824e-06, "loss": 0.2386, "step": 33531 }, { "epoch": 3.37, "learning_rate": 3.218604365195691e-06, "loss": 0.2155, "step": 33532 }, { "epoch": 3.37, "learning_rate": 3.217526027925174e-06, "loss": 0.1725, "step": 33533 }, { "epoch": 3.37, "learning_rate": 3.216447861088131e-06, "loss": 0.0651, "step": 33534 }, { "epoch": 3.37, "learning_rate": 3.21536986469142e-06, "loss": 0.1185, "step": 33535 }, { "epoch": 3.37, "learning_rate": 3.2142920387419183e-06, "loss": 0.2035, "step": 33536 }, { "epoch": 3.37, "learning_rate": 3.2132143832464643e-06, "loss": 0.1559, "step": 33537 }, { "epoch": 3.37, "learning_rate": 3.212136898211919e-06, "loss": 0.0785, "step": 33538 }, { "epoch": 3.37, "learning_rate": 3.2110595836451537e-06, "loss": 0.1126, "step": 33539 }, { "epoch": 3.37, "learning_rate": 3.2099824395530097e-06, "loss": 0.2083, "step": 33540 }, { "epoch": 3.37, "learning_rate": 3.2089054659423412e-06, "loss": 0.1556, "step": 33541 }, { "epoch": 3.37, "learning_rate": 3.2078286628200126e-06, "loss": 0.1951, "step": 33542 }, { "epoch": 3.37, "learning_rate": 3.2067520301928755e-06, "loss": 0.2304, "step": 33543 }, { "epoch": 3.37, "learning_rate": 3.205675568067761e-06, "loss": 0.2133, "step": 33544 }, { "epoch": 3.37, "learning_rate": 3.2045992764515432e-06, "loss": 0.2016, "step": 33545 }, { "epoch": 3.37, "learning_rate": 3.20352315535106e-06, "loss": 0.0937, "step": 33546 }, { "epoch": 3.37, "learning_rate": 3.2024472047731634e-06, "loss": 0.3685, "step": 33547 }, { "epoch": 3.37, "learning_rate": 3.2013714247247007e-06, "loss": 0.1557, "step": 33548 }, { "epoch": 3.37, "learning_rate": 3.2002958152125128e-06, "loss": 0.2147, "step": 33549 }, { "epoch": 3.37, "learning_rate": 3.1992203762434445e-06, "loss": 0.1369, "step": 33550 }, { "epoch": 3.37, "learning_rate": 3.1981451078243475e-06, "loss": 0.0918, "step": 33551 }, { "epoch": 3.37, "learning_rate": 3.197070009962055e-06, "loss": 0.1355, "step": 33552 }, { "epoch": 3.37, "learning_rate": 3.19599508266341e-06, "loss": 0.5808, "step": 33553 }, { "epoch": 3.37, "learning_rate": 3.1949203259352654e-06, "loss": 0.3087, "step": 33554 }, { "epoch": 3.37, "learning_rate": 3.1938457397844434e-06, "loss": 0.4021, "step": 33555 }, { "epoch": 3.37, "learning_rate": 3.1927713242177848e-06, "loss": 0.012, "step": 33556 }, { "epoch": 3.37, "learning_rate": 3.1916970792421442e-06, "loss": 0.3219, "step": 33557 }, { "epoch": 3.37, "learning_rate": 3.1906230048643328e-06, "loss": 0.258, "step": 33558 }, { "epoch": 3.37, "learning_rate": 3.189549101091205e-06, "loss": 0.3035, "step": 33559 }, { "epoch": 3.37, "learning_rate": 3.1884753679295854e-06, "loss": 0.3008, "step": 33560 }, { "epoch": 3.37, "learning_rate": 3.1874018053863087e-06, "loss": 0.1038, "step": 33561 }, { "epoch": 3.37, "learning_rate": 3.186328413468206e-06, "loss": 0.2012, "step": 33562 }, { "epoch": 3.37, "learning_rate": 3.1852551921821116e-06, "loss": 0.3417, "step": 33563 }, { "epoch": 3.37, "learning_rate": 3.1841821415348503e-06, "loss": 0.3304, "step": 33564 }, { "epoch": 3.37, "learning_rate": 3.1831092615332535e-06, "loss": 0.1039, "step": 33565 }, { "epoch": 3.37, "learning_rate": 3.182036552184149e-06, "loss": 0.053, "step": 33566 }, { "epoch": 3.37, "learning_rate": 3.180964013494354e-06, "loss": 0.1191, "step": 33567 }, { "epoch": 3.37, "learning_rate": 3.179891645470714e-06, "loss": 0.6781, "step": 33568 }, { "epoch": 3.37, "learning_rate": 3.178819448120036e-06, "loss": 0.2935, "step": 33569 }, { "epoch": 3.37, "learning_rate": 3.177747421449139e-06, "loss": 0.2647, "step": 33570 }, { "epoch": 3.37, "learning_rate": 3.1766755654648693e-06, "loss": 0.1113, "step": 33571 }, { "epoch": 3.37, "learning_rate": 3.1756038801740227e-06, "loss": 0.1395, "step": 33572 }, { "epoch": 3.37, "learning_rate": 3.1745323655834234e-06, "loss": 0.2509, "step": 33573 }, { "epoch": 3.37, "learning_rate": 3.1734610216999026e-06, "loss": 0.1558, "step": 33574 }, { "epoch": 3.37, "learning_rate": 3.1723898485302747e-06, "loss": 0.1137, "step": 33575 }, { "epoch": 3.37, "learning_rate": 3.1713188460813414e-06, "loss": 0.0817, "step": 33576 }, { "epoch": 3.37, "learning_rate": 3.1702480143599365e-06, "loss": 0.181, "step": 33577 }, { "epoch": 3.37, "learning_rate": 3.1691773533728618e-06, "loss": 0.2297, "step": 33578 }, { "epoch": 3.37, "learning_rate": 3.1681068631269383e-06, "loss": 0.435, "step": 33579 }, { "epoch": 3.37, "learning_rate": 3.167036543628977e-06, "loss": 0.3211, "step": 33580 }, { "epoch": 3.37, "learning_rate": 3.1659663948857865e-06, "loss": 0.1277, "step": 33581 }, { "epoch": 3.37, "learning_rate": 3.164896416904174e-06, "loss": 0.3315, "step": 33582 }, { "epoch": 3.37, "learning_rate": 3.1638266096909573e-06, "loss": 0.0684, "step": 33583 }, { "epoch": 3.37, "learning_rate": 3.1627569732529284e-06, "loss": 0.0745, "step": 33584 }, { "epoch": 3.37, "learning_rate": 3.161687507596914e-06, "loss": 0.4401, "step": 33585 }, { "epoch": 3.37, "learning_rate": 3.1606182127297166e-06, "loss": 0.3545, "step": 33586 }, { "epoch": 3.37, "learning_rate": 3.1595490886581202e-06, "loss": 0.7141, "step": 33587 }, { "epoch": 3.37, "learning_rate": 3.158480135388949e-06, "loss": 0.2113, "step": 33588 }, { "epoch": 3.37, "learning_rate": 3.157411352929008e-06, "loss": 0.0847, "step": 33589 }, { "epoch": 3.37, "learning_rate": 3.1563427412850754e-06, "loss": 0.6131, "step": 33590 }, { "epoch": 3.37, "learning_rate": 3.1552743004639717e-06, "loss": 0.4621, "step": 33591 }, { "epoch": 3.37, "learning_rate": 3.1542060304724918e-06, "loss": 0.5047, "step": 33592 }, { "epoch": 3.37, "learning_rate": 3.1531379313174302e-06, "loss": 0.0367, "step": 33593 }, { "epoch": 3.37, "learning_rate": 3.1520700030055883e-06, "loss": 0.3316, "step": 33594 }, { "epoch": 3.37, "learning_rate": 3.151002245543757e-06, "loss": 0.4154, "step": 33595 }, { "epoch": 3.37, "learning_rate": 3.1499346589387347e-06, "loss": 0.1203, "step": 33596 }, { "epoch": 3.37, "learning_rate": 3.148867243197312e-06, "loss": 0.1909, "step": 33597 }, { "epoch": 3.37, "learning_rate": 3.147799998326284e-06, "loss": 0.5504, "step": 33598 }, { "epoch": 3.37, "learning_rate": 3.1467329243324383e-06, "loss": 0.2605, "step": 33599 }, { "epoch": 3.37, "learning_rate": 3.1456660212225796e-06, "loss": 0.2712, "step": 33600 }, { "epoch": 3.38, "learning_rate": 3.144599289003479e-06, "loss": 0.1289, "step": 33601 }, { "epoch": 3.38, "learning_rate": 3.1435327276819283e-06, "loss": 0.2059, "step": 33602 }, { "epoch": 3.38, "learning_rate": 3.1424663372647276e-06, "loss": 0.1398, "step": 33603 }, { "epoch": 3.38, "learning_rate": 3.1414001177586493e-06, "loss": 0.2215, "step": 33604 }, { "epoch": 3.38, "learning_rate": 3.1403340691704773e-06, "loss": 0.0579, "step": 33605 }, { "epoch": 3.38, "learning_rate": 3.13926819150701e-06, "loss": 0.2081, "step": 33606 }, { "epoch": 3.38, "learning_rate": 3.1382024847750245e-06, "loss": 0.1182, "step": 33607 }, { "epoch": 3.38, "learning_rate": 3.1371369489812896e-06, "loss": 0.0583, "step": 33608 }, { "epoch": 3.38, "learning_rate": 3.1360715841325994e-06, "loss": 0.0837, "step": 33609 }, { "epoch": 3.38, "learning_rate": 3.135006390235732e-06, "loss": 0.1066, "step": 33610 }, { "epoch": 3.38, "learning_rate": 3.1339413672974615e-06, "loss": 0.6151, "step": 33611 }, { "epoch": 3.38, "learning_rate": 3.1328765153245696e-06, "loss": 0.474, "step": 33612 }, { "epoch": 3.38, "learning_rate": 3.1318118343238276e-06, "loss": 0.3032, "step": 33613 }, { "epoch": 3.38, "learning_rate": 3.1307473243020203e-06, "loss": 0.1659, "step": 33614 }, { "epoch": 3.38, "learning_rate": 3.129682985265915e-06, "loss": 0.1096, "step": 33615 }, { "epoch": 3.38, "learning_rate": 3.1286188172222762e-06, "loss": 0.112, "step": 33616 }, { "epoch": 3.38, "learning_rate": 3.127554820177889e-06, "loss": 0.4673, "step": 33617 }, { "epoch": 3.38, "learning_rate": 3.126490994139528e-06, "loss": 0.5389, "step": 33618 }, { "epoch": 3.38, "learning_rate": 3.1254273391139445e-06, "loss": 0.5048, "step": 33619 }, { "epoch": 3.38, "learning_rate": 3.124363855107922e-06, "loss": 0.0999, "step": 33620 }, { "epoch": 3.38, "learning_rate": 3.12330054212823e-06, "loss": 0.0887, "step": 33621 }, { "epoch": 3.38, "learning_rate": 3.1222374001816188e-06, "loss": 0.2755, "step": 33622 }, { "epoch": 3.38, "learning_rate": 3.12117442927487e-06, "loss": 0.4191, "step": 33623 }, { "epoch": 3.38, "learning_rate": 3.1201116294147414e-06, "loss": 0.2478, "step": 33624 }, { "epoch": 3.38, "learning_rate": 3.1190490006079943e-06, "loss": 0.046, "step": 33625 }, { "epoch": 3.38, "learning_rate": 3.1179865428613962e-06, "loss": 0.0185, "step": 33626 }, { "epoch": 3.38, "learning_rate": 3.116924256181702e-06, "loss": 0.2253, "step": 33627 }, { "epoch": 3.38, "learning_rate": 3.115862140575677e-06, "loss": 0.2987, "step": 33628 }, { "epoch": 3.38, "learning_rate": 3.1148001960500784e-06, "loss": 0.1164, "step": 33629 }, { "epoch": 3.38, "learning_rate": 3.113738422611664e-06, "loss": 0.0901, "step": 33630 }, { "epoch": 3.38, "learning_rate": 3.112676820267185e-06, "loss": 0.1918, "step": 33631 }, { "epoch": 3.38, "learning_rate": 3.1116153890234135e-06, "loss": 0.1593, "step": 33632 }, { "epoch": 3.38, "learning_rate": 3.110554128887083e-06, "loss": 0.0964, "step": 33633 }, { "epoch": 3.38, "learning_rate": 3.1094930398649556e-06, "loss": 0.6435, "step": 33634 }, { "epoch": 3.38, "learning_rate": 3.1084321219637957e-06, "loss": 0.1837, "step": 33635 }, { "epoch": 3.38, "learning_rate": 3.1073713751903377e-06, "loss": 0.2111, "step": 33636 }, { "epoch": 3.38, "learning_rate": 3.1063107995513296e-06, "loss": 0.0808, "step": 33637 }, { "epoch": 3.38, "learning_rate": 3.105250395053536e-06, "loss": 0.0352, "step": 33638 }, { "epoch": 3.38, "learning_rate": 3.104190161703698e-06, "loss": 0.1001, "step": 33639 }, { "epoch": 3.38, "learning_rate": 3.103130099508564e-06, "loss": 0.6285, "step": 33640 }, { "epoch": 3.38, "learning_rate": 3.1020702084748753e-06, "loss": 0.1148, "step": 33641 }, { "epoch": 3.38, "learning_rate": 3.101010488609379e-06, "loss": 0.2057, "step": 33642 }, { "epoch": 3.38, "learning_rate": 3.0999509399188207e-06, "loss": 0.4094, "step": 33643 }, { "epoch": 3.38, "learning_rate": 3.0988915624099413e-06, "loss": 0.3224, "step": 33644 }, { "epoch": 3.38, "learning_rate": 3.097832356089475e-06, "loss": 0.5075, "step": 33645 }, { "epoch": 3.38, "learning_rate": 3.0967733209641835e-06, "loss": 0.4333, "step": 33646 }, { "epoch": 3.38, "learning_rate": 3.0957144570407847e-06, "loss": 0.7064, "step": 33647 }, { "epoch": 3.38, "learning_rate": 3.0946557643260165e-06, "loss": 0.1591, "step": 33648 }, { "epoch": 3.38, "learning_rate": 3.0935972428266333e-06, "loss": 0.081, "step": 33649 }, { "epoch": 3.38, "learning_rate": 3.0925388925493636e-06, "loss": 0.0528, "step": 33650 }, { "epoch": 3.38, "learning_rate": 3.0914807135009317e-06, "loss": 0.5794, "step": 33651 }, { "epoch": 3.38, "learning_rate": 3.0904227056880853e-06, "loss": 0.3754, "step": 33652 }, { "epoch": 3.38, "learning_rate": 3.089364869117556e-06, "loss": 0.4429, "step": 33653 }, { "epoch": 3.38, "learning_rate": 3.088307203796061e-06, "loss": 0.0329, "step": 33654 }, { "epoch": 3.38, "learning_rate": 3.0872497097303466e-06, "loss": 0.2055, "step": 33655 }, { "epoch": 3.38, "learning_rate": 3.0861923869271393e-06, "loss": 0.0978, "step": 33656 }, { "epoch": 3.38, "learning_rate": 3.0851352353931604e-06, "loss": 0.2315, "step": 33657 }, { "epoch": 3.38, "learning_rate": 3.084078255135145e-06, "loss": 0.279, "step": 33658 }, { "epoch": 3.38, "learning_rate": 3.083021446159815e-06, "loss": 0.3811, "step": 33659 }, { "epoch": 3.38, "learning_rate": 3.0819648084738973e-06, "loss": 0.3198, "step": 33660 }, { "epoch": 3.38, "learning_rate": 3.0809083420841167e-06, "loss": 0.2153, "step": 33661 }, { "epoch": 3.38, "learning_rate": 3.0798520469971948e-06, "loss": 0.2061, "step": 33662 }, { "epoch": 3.38, "learning_rate": 3.078795923219846e-06, "loss": 0.0429, "step": 33663 }, { "epoch": 3.38, "learning_rate": 3.0777399707588116e-06, "loss": 0.4271, "step": 33664 }, { "epoch": 3.38, "learning_rate": 3.0766841896207865e-06, "loss": 0.0635, "step": 33665 }, { "epoch": 3.38, "learning_rate": 3.075628579812505e-06, "loss": 0.1183, "step": 33666 }, { "epoch": 3.38, "learning_rate": 3.0745731413406818e-06, "loss": 0.206, "step": 33667 }, { "epoch": 3.38, "learning_rate": 3.073517874212032e-06, "loss": 0.1094, "step": 33668 }, { "epoch": 3.38, "learning_rate": 3.0724627784332727e-06, "loss": 0.2686, "step": 33669 }, { "epoch": 3.38, "learning_rate": 3.071407854011116e-06, "loss": 0.0342, "step": 33670 }, { "epoch": 3.38, "learning_rate": 3.070353100952272e-06, "loss": 0.2433, "step": 33671 }, { "epoch": 3.38, "learning_rate": 3.0692985192634605e-06, "loss": 0.0882, "step": 33672 }, { "epoch": 3.38, "learning_rate": 3.068244108951388e-06, "loss": 0.222, "step": 33673 }, { "epoch": 3.38, "learning_rate": 3.067189870022763e-06, "loss": 0.114, "step": 33674 }, { "epoch": 3.38, "learning_rate": 3.066135802484297e-06, "loss": 0.4137, "step": 33675 }, { "epoch": 3.38, "learning_rate": 3.0650819063426938e-06, "loss": 0.0661, "step": 33676 }, { "epoch": 3.38, "learning_rate": 3.064028181604659e-06, "loss": 0.1595, "step": 33677 }, { "epoch": 3.38, "learning_rate": 3.0629746282769133e-06, "loss": 0.0974, "step": 33678 }, { "epoch": 3.38, "learning_rate": 3.061921246366145e-06, "loss": 0.0659, "step": 33679 }, { "epoch": 3.38, "learning_rate": 3.0608680358790554e-06, "loss": 0.5295, "step": 33680 }, { "epoch": 3.38, "learning_rate": 3.059814996822359e-06, "loss": 0.4486, "step": 33681 }, { "epoch": 3.38, "learning_rate": 3.05876212920276e-06, "loss": 0.3319, "step": 33682 }, { "epoch": 3.38, "learning_rate": 3.057709433026934e-06, "loss": 0.3337, "step": 33683 }, { "epoch": 3.38, "learning_rate": 3.056656908301605e-06, "loss": 0.0639, "step": 33684 }, { "epoch": 3.38, "learning_rate": 3.0556045550334643e-06, "loss": 0.0094, "step": 33685 }, { "epoch": 3.38, "learning_rate": 3.054552373229197e-06, "loss": 0.5369, "step": 33686 }, { "epoch": 3.38, "learning_rate": 3.0535003628955136e-06, "loss": 0.1553, "step": 33687 }, { "epoch": 3.38, "learning_rate": 3.0524485240391e-06, "loss": 0.0414, "step": 33688 }, { "epoch": 3.38, "learning_rate": 3.051396856666656e-06, "loss": 0.0649, "step": 33689 }, { "epoch": 3.38, "learning_rate": 3.0503453607848676e-06, "loss": 0.2205, "step": 33690 }, { "epoch": 3.38, "learning_rate": 3.0492940364004252e-06, "loss": 0.0782, "step": 33691 }, { "epoch": 3.38, "learning_rate": 3.048242883520034e-06, "loss": 0.1569, "step": 33692 }, { "epoch": 3.38, "learning_rate": 3.047191902150368e-06, "loss": 0.1266, "step": 33693 }, { "epoch": 3.38, "learning_rate": 3.0461410922981127e-06, "loss": 0.5573, "step": 33694 }, { "epoch": 3.38, "learning_rate": 3.0450904539699652e-06, "loss": 0.1753, "step": 33695 }, { "epoch": 3.38, "learning_rate": 3.0440399871726176e-06, "loss": 0.3012, "step": 33696 }, { "epoch": 3.38, "learning_rate": 3.042989691912731e-06, "loss": 0.3912, "step": 33697 }, { "epoch": 3.38, "learning_rate": 3.0419395681970097e-06, "loss": 0.1681, "step": 33698 }, { "epoch": 3.38, "learning_rate": 3.040889616032129e-06, "loss": 0.0567, "step": 33699 }, { "epoch": 3.39, "learning_rate": 3.039839835424769e-06, "loss": 0.4573, "step": 33700 }, { "epoch": 3.39, "learning_rate": 3.038790226381616e-06, "loss": 0.1012, "step": 33701 }, { "epoch": 3.39, "learning_rate": 3.0377407889093435e-06, "loss": 0.2948, "step": 33702 }, { "epoch": 3.39, "learning_rate": 3.03669152301463e-06, "loss": 0.0302, "step": 33703 }, { "epoch": 3.39, "learning_rate": 3.0356424287041562e-06, "loss": 0.1688, "step": 33704 }, { "epoch": 3.39, "learning_rate": 3.0345935059845976e-06, "loss": 0.1566, "step": 33705 }, { "epoch": 3.39, "learning_rate": 3.033544754862625e-06, "loss": 0.1486, "step": 33706 }, { "epoch": 3.39, "learning_rate": 3.0324961753449164e-06, "loss": 0.3133, "step": 33707 }, { "epoch": 3.39, "learning_rate": 3.0314477674381435e-06, "loss": 0.1391, "step": 33708 }, { "epoch": 3.39, "learning_rate": 3.0303995311489706e-06, "loss": 0.3621, "step": 33709 }, { "epoch": 3.39, "learning_rate": 3.029351466484089e-06, "loss": 0.1746, "step": 33710 }, { "epoch": 3.39, "learning_rate": 3.0283035734501472e-06, "loss": 0.2123, "step": 33711 }, { "epoch": 3.39, "learning_rate": 3.0272558520538164e-06, "loss": 0.1851, "step": 33712 }, { "epoch": 3.39, "learning_rate": 3.0262083023017705e-06, "loss": 0.0615, "step": 33713 }, { "epoch": 3.39, "learning_rate": 3.0251609242006816e-06, "loss": 0.3284, "step": 33714 }, { "epoch": 3.39, "learning_rate": 3.0241137177571976e-06, "loss": 0.0589, "step": 33715 }, { "epoch": 3.39, "learning_rate": 3.023066682977996e-06, "loss": 0.3725, "step": 33716 }, { "epoch": 3.39, "learning_rate": 3.022019819869739e-06, "loss": 0.0523, "step": 33717 }, { "epoch": 3.39, "learning_rate": 3.020973128439077e-06, "loss": 0.2095, "step": 33718 }, { "epoch": 3.39, "learning_rate": 3.0199266086926824e-06, "loss": 0.2641, "step": 33719 }, { "epoch": 3.39, "learning_rate": 3.018880260637209e-06, "loss": 0.1414, "step": 33720 }, { "epoch": 3.39, "learning_rate": 3.017834084279322e-06, "loss": 0.4815, "step": 33721 }, { "epoch": 3.39, "learning_rate": 3.016788079625672e-06, "loss": 0.3022, "step": 33722 }, { "epoch": 3.39, "learning_rate": 3.015742246682911e-06, "loss": 0.0459, "step": 33723 }, { "epoch": 3.39, "learning_rate": 3.0146965854577136e-06, "loss": 0.1578, "step": 33724 }, { "epoch": 3.39, "learning_rate": 3.0136510959567176e-06, "loss": 0.0945, "step": 33725 }, { "epoch": 3.39, "learning_rate": 3.012605778186571e-06, "loss": 0.3377, "step": 33726 }, { "epoch": 3.39, "learning_rate": 3.0115606321539423e-06, "loss": 0.0516, "step": 33727 }, { "epoch": 3.39, "learning_rate": 3.010515657865479e-06, "loss": 0.0655, "step": 33728 }, { "epoch": 3.39, "learning_rate": 3.009470855327816e-06, "loss": 0.116, "step": 33729 }, { "epoch": 3.39, "learning_rate": 3.008426224547618e-06, "loss": 0.2025, "step": 33730 }, { "epoch": 3.39, "learning_rate": 3.0073817655315293e-06, "loss": 0.1579, "step": 33731 }, { "epoch": 3.39, "learning_rate": 3.006337478286192e-06, "loss": 0.2738, "step": 33732 }, { "epoch": 3.39, "learning_rate": 3.005293362818257e-06, "loss": 0.1176, "step": 33733 }, { "epoch": 3.39, "learning_rate": 3.004249419134365e-06, "loss": 0.0118, "step": 33734 }, { "epoch": 3.39, "learning_rate": 3.0032056472411585e-06, "loss": 0.1584, "step": 33735 }, { "epoch": 3.39, "learning_rate": 3.0021620471452817e-06, "loss": 0.0434, "step": 33736 }, { "epoch": 3.39, "learning_rate": 3.001118618853376e-06, "loss": 0.2166, "step": 33737 }, { "epoch": 3.39, "learning_rate": 3.0000753623720788e-06, "loss": 0.07, "step": 33738 }, { "epoch": 3.39, "learning_rate": 2.9990322777080324e-06, "loss": 0.1045, "step": 33739 }, { "epoch": 3.39, "learning_rate": 2.9979893648678715e-06, "loss": 0.1145, "step": 33740 }, { "epoch": 3.39, "learning_rate": 2.996946623858227e-06, "loss": 0.1311, "step": 33741 }, { "epoch": 3.39, "learning_rate": 2.9959040546857563e-06, "loss": 0.3747, "step": 33742 }, { "epoch": 3.39, "learning_rate": 2.994861657357072e-06, "loss": 0.1382, "step": 33743 }, { "epoch": 3.39, "learning_rate": 2.9938194318788114e-06, "loss": 0.0607, "step": 33744 }, { "epoch": 3.39, "learning_rate": 2.992777378257613e-06, "loss": 0.2365, "step": 33745 }, { "epoch": 3.39, "learning_rate": 2.9917354965001074e-06, "loss": 0.1185, "step": 33746 }, { "epoch": 3.39, "learning_rate": 2.9906937866129204e-06, "loss": 0.2118, "step": 33747 }, { "epoch": 3.39, "learning_rate": 2.9896522486026855e-06, "loss": 0.7578, "step": 33748 }, { "epoch": 3.39, "learning_rate": 2.9886108824760282e-06, "loss": 0.2298, "step": 33749 }, { "epoch": 3.39, "learning_rate": 2.9875696882395733e-06, "loss": 0.2832, "step": 33750 }, { "epoch": 3.39, "learning_rate": 2.986528665899951e-06, "loss": 0.2837, "step": 33751 }, { "epoch": 3.39, "learning_rate": 2.9854878154637844e-06, "loss": 0.2865, "step": 33752 }, { "epoch": 3.39, "learning_rate": 2.9844471369376935e-06, "loss": 0.0724, "step": 33753 }, { "epoch": 3.39, "learning_rate": 2.983406630328307e-06, "loss": 0.6804, "step": 33754 }, { "epoch": 3.39, "learning_rate": 2.982366295642236e-06, "loss": 0.0308, "step": 33755 }, { "epoch": 3.39, "learning_rate": 2.9813261328861185e-06, "loss": 0.0383, "step": 33756 }, { "epoch": 3.39, "learning_rate": 2.9802861420665594e-06, "loss": 0.1102, "step": 33757 }, { "epoch": 3.39, "learning_rate": 2.9792463231901735e-06, "loss": 0.3314, "step": 33758 }, { "epoch": 3.39, "learning_rate": 2.9782066762635884e-06, "loss": 0.0355, "step": 33759 }, { "epoch": 3.39, "learning_rate": 2.9771672012934257e-06, "loss": 0.0745, "step": 33760 }, { "epoch": 3.39, "learning_rate": 2.976127898286277e-06, "loss": 0.6886, "step": 33761 }, { "epoch": 3.39, "learning_rate": 2.975088767248777e-06, "loss": 0.1512, "step": 33762 }, { "epoch": 3.39, "learning_rate": 2.97404980818753e-06, "loss": 0.1633, "step": 33763 }, { "epoch": 3.39, "learning_rate": 2.973011021109151e-06, "loss": 0.1313, "step": 33764 }, { "epoch": 3.39, "learning_rate": 2.9719724060202447e-06, "loss": 0.0792, "step": 33765 }, { "epoch": 3.39, "learning_rate": 2.970933962927429e-06, "loss": 0.1289, "step": 33766 }, { "epoch": 3.39, "learning_rate": 2.969895691837302e-06, "loss": 0.2275, "step": 33767 }, { "epoch": 3.39, "learning_rate": 2.9688575927564787e-06, "loss": 0.0873, "step": 33768 }, { "epoch": 3.39, "learning_rate": 2.9678196656915633e-06, "loss": 0.2815, "step": 33769 }, { "epoch": 3.39, "learning_rate": 2.966781910649158e-06, "loss": 0.2405, "step": 33770 }, { "epoch": 3.39, "learning_rate": 2.96574432763587e-06, "loss": 0.0495, "step": 33771 }, { "epoch": 3.39, "learning_rate": 2.964706916658294e-06, "loss": 0.1423, "step": 33772 }, { "epoch": 3.39, "learning_rate": 2.9636696777230452e-06, "loss": 0.692, "step": 33773 }, { "epoch": 3.39, "learning_rate": 2.962632610836725e-06, "loss": 0.4979, "step": 33774 }, { "epoch": 3.39, "learning_rate": 2.9615957160059113e-06, "loss": 0.1709, "step": 33775 }, { "epoch": 3.39, "learning_rate": 2.9605589932372225e-06, "loss": 0.1993, "step": 33776 }, { "epoch": 3.39, "learning_rate": 2.959522442537249e-06, "loss": 0.0913, "step": 33777 }, { "epoch": 3.39, "learning_rate": 2.95848606391259e-06, "loss": 0.5362, "step": 33778 }, { "epoch": 3.39, "learning_rate": 2.957449857369836e-06, "loss": 0.2275, "step": 33779 }, { "epoch": 3.39, "learning_rate": 2.9564138229155856e-06, "loss": 0.0059, "step": 33780 }, { "epoch": 3.39, "learning_rate": 2.95537796055643e-06, "loss": 0.4357, "step": 33781 }, { "epoch": 3.39, "learning_rate": 2.9543422702989608e-06, "loss": 0.1834, "step": 33782 }, { "epoch": 3.39, "learning_rate": 2.9533067521497657e-06, "loss": 0.2326, "step": 33783 }, { "epoch": 3.39, "learning_rate": 2.95227140611544e-06, "loss": 0.1388, "step": 33784 }, { "epoch": 3.39, "learning_rate": 2.951236232202571e-06, "loss": 0.1954, "step": 33785 }, { "epoch": 3.39, "learning_rate": 2.9502012304177407e-06, "loss": 0.2488, "step": 33786 }, { "epoch": 3.39, "learning_rate": 2.949166400767537e-06, "loss": 0.3738, "step": 33787 }, { "epoch": 3.39, "learning_rate": 2.9481317432585575e-06, "loss": 0.0433, "step": 33788 }, { "epoch": 3.39, "learning_rate": 2.947097257897371e-06, "loss": 0.5382, "step": 33789 }, { "epoch": 3.39, "learning_rate": 2.9460629446905617e-06, "loss": 0.0062, "step": 33790 }, { "epoch": 3.39, "learning_rate": 2.9450288036447214e-06, "loss": 0.0637, "step": 33791 }, { "epoch": 3.39, "learning_rate": 2.943994834766431e-06, "loss": 0.0991, "step": 33792 }, { "epoch": 3.39, "learning_rate": 2.942961038062253e-06, "loss": 0.3952, "step": 33793 }, { "epoch": 3.39, "learning_rate": 2.941927413538784e-06, "loss": 0.4098, "step": 33794 }, { "epoch": 3.39, "learning_rate": 2.9408939612025964e-06, "loss": 0.3078, "step": 33795 }, { "epoch": 3.39, "learning_rate": 2.9398606810602647e-06, "loss": 0.5823, "step": 33796 }, { "epoch": 3.39, "learning_rate": 2.9388275731183632e-06, "loss": 0.3147, "step": 33797 }, { "epoch": 3.4, "learning_rate": 2.937794637383471e-06, "loss": 0.0505, "step": 33798 }, { "epoch": 3.4, "learning_rate": 2.936761873862158e-06, "loss": 0.2243, "step": 33799 }, { "epoch": 3.4, "learning_rate": 2.935729282560997e-06, "loss": 0.2712, "step": 33800 }, { "epoch": 3.4, "learning_rate": 2.934696863486556e-06, "loss": 0.1398, "step": 33801 }, { "epoch": 3.4, "learning_rate": 2.933664616645415e-06, "loss": 0.048, "step": 33802 }, { "epoch": 3.4, "learning_rate": 2.9326325420441336e-06, "loss": 0.0715, "step": 33803 }, { "epoch": 3.4, "learning_rate": 2.931600639689276e-06, "loss": 0.0606, "step": 33804 }, { "epoch": 3.4, "learning_rate": 2.9305689095874166e-06, "loss": 0.0157, "step": 33805 }, { "epoch": 3.4, "learning_rate": 2.9295373517451274e-06, "loss": 0.0757, "step": 33806 }, { "epoch": 3.4, "learning_rate": 2.928505966168953e-06, "loss": 0.4694, "step": 33807 }, { "epoch": 3.4, "learning_rate": 2.9274747528654744e-06, "loss": 0.0147, "step": 33808 }, { "epoch": 3.4, "learning_rate": 2.9264437118412468e-06, "loss": 0.2493, "step": 33809 }, { "epoch": 3.4, "learning_rate": 2.9254128431028316e-06, "loss": 0.1454, "step": 33810 }, { "epoch": 3.4, "learning_rate": 2.92438214665679e-06, "loss": 0.1807, "step": 33811 }, { "epoch": 3.4, "learning_rate": 2.92335162250968e-06, "loss": 0.0725, "step": 33812 }, { "epoch": 3.4, "learning_rate": 2.9223212706680602e-06, "loss": 0.054, "step": 33813 }, { "epoch": 3.4, "learning_rate": 2.921291091138485e-06, "loss": 0.1022, "step": 33814 }, { "epoch": 3.4, "learning_rate": 2.9202610839275123e-06, "loss": 0.1177, "step": 33815 }, { "epoch": 3.4, "learning_rate": 2.9192312490416974e-06, "loss": 0.134, "step": 33816 }, { "epoch": 3.4, "learning_rate": 2.9182015864875944e-06, "loss": 0.1073, "step": 33817 }, { "epoch": 3.4, "learning_rate": 2.9171720962717517e-06, "loss": 0.1293, "step": 33818 }, { "epoch": 3.4, "learning_rate": 2.9161427784007176e-06, "loss": 0.3475, "step": 33819 }, { "epoch": 3.4, "learning_rate": 2.91511363288106e-06, "loss": 0.6462, "step": 33820 }, { "epoch": 3.4, "learning_rate": 2.9140846597193107e-06, "loss": 0.2684, "step": 33821 }, { "epoch": 3.4, "learning_rate": 2.913055858922017e-06, "loss": 0.4637, "step": 33822 }, { "epoch": 3.4, "learning_rate": 2.912027230495734e-06, "loss": 0.2501, "step": 33823 }, { "epoch": 3.4, "learning_rate": 2.9109987744470133e-06, "loss": 0.1849, "step": 33824 }, { "epoch": 3.4, "learning_rate": 2.9099704907823798e-06, "loss": 0.0872, "step": 33825 }, { "epoch": 3.4, "learning_rate": 2.9089423795083946e-06, "loss": 0.13, "step": 33826 }, { "epoch": 3.4, "learning_rate": 2.9079144406315927e-06, "loss": 0.153, "step": 33827 }, { "epoch": 3.4, "learning_rate": 2.9068866741585152e-06, "loss": 0.2278, "step": 33828 }, { "epoch": 3.4, "learning_rate": 2.9058590800957073e-06, "loss": 0.1483, "step": 33829 }, { "epoch": 3.4, "learning_rate": 2.9048316584497037e-06, "loss": 0.5302, "step": 33830 }, { "epoch": 3.4, "learning_rate": 2.9038044092270453e-06, "loss": 0.0365, "step": 33831 }, { "epoch": 3.4, "learning_rate": 2.9027773324342644e-06, "loss": 0.1878, "step": 33832 }, { "epoch": 3.4, "learning_rate": 2.9017504280778983e-06, "loss": 0.0567, "step": 33833 }, { "epoch": 3.4, "learning_rate": 2.900723696164492e-06, "loss": 0.3374, "step": 33834 }, { "epoch": 3.4, "learning_rate": 2.8996971367005675e-06, "loss": 0.3352, "step": 33835 }, { "epoch": 3.4, "learning_rate": 2.8986707496926557e-06, "loss": 0.2062, "step": 33836 }, { "epoch": 3.4, "learning_rate": 2.897644535147298e-06, "loss": 0.046, "step": 33837 }, { "epoch": 3.4, "learning_rate": 2.896618493071026e-06, "loss": 0.113, "step": 33838 }, { "epoch": 3.4, "learning_rate": 2.8955926234703513e-06, "loss": 0.4477, "step": 33839 }, { "epoch": 3.4, "learning_rate": 2.8945669263518214e-06, "loss": 0.1718, "step": 33840 }, { "epoch": 3.4, "learning_rate": 2.893541401721955e-06, "loss": 0.3386, "step": 33841 }, { "epoch": 3.4, "learning_rate": 2.892516049587277e-06, "loss": 0.1782, "step": 33842 }, { "epoch": 3.4, "learning_rate": 2.8914908699543186e-06, "loss": 0.2687, "step": 33843 }, { "epoch": 3.4, "learning_rate": 2.8904658628295978e-06, "loss": 0.1442, "step": 33844 }, { "epoch": 3.4, "learning_rate": 2.8894410282196394e-06, "loss": 0.1839, "step": 33845 }, { "epoch": 3.4, "learning_rate": 2.888416366130965e-06, "loss": 0.1521, "step": 33846 }, { "epoch": 3.4, "learning_rate": 2.887391876570096e-06, "loss": 0.2775, "step": 33847 }, { "epoch": 3.4, "learning_rate": 2.88636755954355e-06, "loss": 0.4866, "step": 33848 }, { "epoch": 3.4, "learning_rate": 2.885343415057846e-06, "loss": 0.3611, "step": 33849 }, { "epoch": 3.4, "learning_rate": 2.884319443119502e-06, "loss": 0.324, "step": 33850 }, { "epoch": 3.4, "learning_rate": 2.883295643735029e-06, "loss": 0.3494, "step": 33851 }, { "epoch": 3.4, "learning_rate": 2.882272016910955e-06, "loss": 0.2771, "step": 33852 }, { "epoch": 3.4, "learning_rate": 2.881248562653782e-06, "loss": 0.076, "step": 33853 }, { "epoch": 3.4, "learning_rate": 2.8802252809700215e-06, "loss": 0.3711, "step": 33854 }, { "epoch": 3.4, "learning_rate": 2.8792021718661977e-06, "loss": 0.0424, "step": 33855 }, { "epoch": 3.4, "learning_rate": 2.8781792353488124e-06, "loss": 0.1555, "step": 33856 }, { "epoch": 3.4, "learning_rate": 2.877156471424377e-06, "loss": 0.2809, "step": 33857 }, { "epoch": 3.4, "learning_rate": 2.8761338800993963e-06, "loss": 0.1078, "step": 33858 }, { "epoch": 3.4, "learning_rate": 2.8751114613803853e-06, "loss": 0.2601, "step": 33859 }, { "epoch": 3.4, "learning_rate": 2.874089215273845e-06, "loss": 0.3256, "step": 33860 }, { "epoch": 3.4, "learning_rate": 2.873067141786281e-06, "loss": 0.1105, "step": 33861 }, { "epoch": 3.4, "learning_rate": 2.8720452409242006e-06, "loss": 0.1742, "step": 33862 }, { "epoch": 3.4, "learning_rate": 2.871023512694102e-06, "loss": 0.1716, "step": 33863 }, { "epoch": 3.4, "learning_rate": 2.8700019571024904e-06, "loss": 0.0831, "step": 33864 }, { "epoch": 3.4, "learning_rate": 2.868980574155857e-06, "loss": 0.3141, "step": 33865 }, { "epoch": 3.4, "learning_rate": 2.867959363860724e-06, "loss": 0.444, "step": 33866 }, { "epoch": 3.4, "learning_rate": 2.8669383262235717e-06, "loss": 0.0144, "step": 33867 }, { "epoch": 3.4, "learning_rate": 2.8659174612508955e-06, "loss": 0.1442, "step": 33868 }, { "epoch": 3.4, "learning_rate": 2.8648967689492e-06, "loss": 0.051, "step": 33869 }, { "epoch": 3.4, "learning_rate": 2.8638762493249904e-06, "loss": 0.1358, "step": 33870 }, { "epoch": 3.4, "learning_rate": 2.8628559023847343e-06, "loss": 0.1842, "step": 33871 }, { "epoch": 3.4, "learning_rate": 2.861835728134944e-06, "loss": 0.2132, "step": 33872 }, { "epoch": 3.4, "learning_rate": 2.86081572658211e-06, "loss": 0.1397, "step": 33873 }, { "epoch": 3.4, "learning_rate": 2.859795897732721e-06, "loss": 0.1207, "step": 33874 }, { "epoch": 3.4, "learning_rate": 2.858776241593265e-06, "loss": 0.273, "step": 33875 }, { "epoch": 3.4, "learning_rate": 2.8577567581702304e-06, "loss": 0.2005, "step": 33876 }, { "epoch": 3.4, "learning_rate": 2.856737447470108e-06, "loss": 0.0914, "step": 33877 }, { "epoch": 3.4, "learning_rate": 2.8557183094993834e-06, "loss": 0.2792, "step": 33878 }, { "epoch": 3.4, "learning_rate": 2.854699344264541e-06, "loss": 0.4013, "step": 33879 }, { "epoch": 3.4, "learning_rate": 2.8536805517720656e-06, "loss": 0.1956, "step": 33880 }, { "epoch": 3.4, "learning_rate": 2.8526619320284385e-06, "loss": 0.015, "step": 33881 }, { "epoch": 3.4, "learning_rate": 2.8516434850401384e-06, "loss": 0.1518, "step": 33882 }, { "epoch": 3.4, "learning_rate": 2.8506252108136565e-06, "loss": 0.2199, "step": 33883 }, { "epoch": 3.4, "learning_rate": 2.8496071093554744e-06, "loss": 0.2672, "step": 33884 }, { "epoch": 3.4, "learning_rate": 2.848589180672053e-06, "loss": 0.0561, "step": 33885 }, { "epoch": 3.4, "learning_rate": 2.847571424769885e-06, "loss": 0.1181, "step": 33886 }, { "epoch": 3.4, "learning_rate": 2.846553841655447e-06, "loss": 0.1707, "step": 33887 }, { "epoch": 3.4, "learning_rate": 2.8455364313352084e-06, "loss": 0.1957, "step": 33888 }, { "epoch": 3.4, "learning_rate": 2.8445191938156433e-06, "loss": 0.2899, "step": 33889 }, { "epoch": 3.4, "learning_rate": 2.8435021291032304e-06, "loss": 0.4443, "step": 33890 }, { "epoch": 3.4, "learning_rate": 2.8424852372044373e-06, "loss": 0.1385, "step": 33891 }, { "epoch": 3.4, "learning_rate": 2.8414685181257393e-06, "loss": 0.0527, "step": 33892 }, { "epoch": 3.4, "learning_rate": 2.8404519718736045e-06, "loss": 0.0474, "step": 33893 }, { "epoch": 3.4, "learning_rate": 2.839435598454497e-06, "loss": 0.2953, "step": 33894 }, { "epoch": 3.4, "learning_rate": 2.838419397874893e-06, "loss": 0.1158, "step": 33895 }, { "epoch": 3.4, "learning_rate": 2.8374033701412528e-06, "loss": 0.2466, "step": 33896 }, { "epoch": 3.41, "learning_rate": 2.8363875152600414e-06, "loss": 0.2809, "step": 33897 }, { "epoch": 3.41, "learning_rate": 2.8353718332377344e-06, "loss": 0.3095, "step": 33898 }, { "epoch": 3.41, "learning_rate": 2.8343563240807823e-06, "loss": 0.1542, "step": 33899 }, { "epoch": 3.41, "learning_rate": 2.8333409877956474e-06, "loss": 0.1058, "step": 33900 }, { "epoch": 3.41, "learning_rate": 2.832325824388804e-06, "loss": 0.0192, "step": 33901 }, { "epoch": 3.41, "learning_rate": 2.8313108338667037e-06, "loss": 0.0623, "step": 33902 }, { "epoch": 3.41, "learning_rate": 2.830296016235798e-06, "loss": 0.5421, "step": 33903 }, { "epoch": 3.41, "learning_rate": 2.829281371502559e-06, "loss": 0.2297, "step": 33904 }, { "epoch": 3.41, "learning_rate": 2.8282668996734373e-06, "loss": 0.2347, "step": 33905 }, { "epoch": 3.41, "learning_rate": 2.827252600754885e-06, "loss": 0.0777, "step": 33906 }, { "epoch": 3.41, "learning_rate": 2.826238474753363e-06, "loss": 0.1605, "step": 33907 }, { "epoch": 3.41, "learning_rate": 2.8252245216753203e-06, "loss": 0.4152, "step": 33908 }, { "epoch": 3.41, "learning_rate": 2.824210741527211e-06, "loss": 0.3064, "step": 33909 }, { "epoch": 3.41, "learning_rate": 2.823197134315487e-06, "loss": 0.1913, "step": 33910 }, { "epoch": 3.41, "learning_rate": 2.822183700046593e-06, "loss": 0.2382, "step": 33911 }, { "epoch": 3.41, "learning_rate": 2.8211704387269943e-06, "loss": 0.5684, "step": 33912 }, { "epoch": 3.41, "learning_rate": 2.8201573503631186e-06, "loss": 0.0741, "step": 33913 }, { "epoch": 3.41, "learning_rate": 2.8191444349614204e-06, "loss": 0.0907, "step": 33914 }, { "epoch": 3.41, "learning_rate": 2.818131692528352e-06, "loss": 0.4718, "step": 33915 }, { "epoch": 3.41, "learning_rate": 2.8171191230703543e-06, "loss": 0.0827, "step": 33916 }, { "epoch": 3.41, "learning_rate": 2.816106726593862e-06, "loss": 0.0268, "step": 33917 }, { "epoch": 3.41, "learning_rate": 2.815094503105331e-06, "loss": 0.2626, "step": 33918 }, { "epoch": 3.41, "learning_rate": 2.8140824526111954e-06, "loss": 0.0696, "step": 33919 }, { "epoch": 3.41, "learning_rate": 2.8130705751178997e-06, "loss": 0.444, "step": 33920 }, { "epoch": 3.41, "learning_rate": 2.8120588706318762e-06, "loss": 0.1524, "step": 33921 }, { "epoch": 3.41, "learning_rate": 2.8110473391595693e-06, "loss": 0.2544, "step": 33922 }, { "epoch": 3.41, "learning_rate": 2.810035980707414e-06, "loss": 0.0103, "step": 33923 }, { "epoch": 3.41, "learning_rate": 2.809024795281848e-06, "loss": 0.1818, "step": 33924 }, { "epoch": 3.41, "learning_rate": 2.8080137828893006e-06, "loss": 0.1088, "step": 33925 }, { "epoch": 3.41, "learning_rate": 2.8070029435362123e-06, "loss": 0.0749, "step": 33926 }, { "epoch": 3.41, "learning_rate": 2.8059922772290116e-06, "loss": 0.0583, "step": 33927 }, { "epoch": 3.41, "learning_rate": 2.8049817839741297e-06, "loss": 0.0834, "step": 33928 }, { "epoch": 3.41, "learning_rate": 2.8039714637779922e-06, "loss": 0.0503, "step": 33929 }, { "epoch": 3.41, "learning_rate": 2.80296131664705e-06, "loss": 0.0108, "step": 33930 }, { "epoch": 3.41, "learning_rate": 2.8019513425877053e-06, "loss": 0.2906, "step": 33931 }, { "epoch": 3.41, "learning_rate": 2.800941541606392e-06, "loss": 0.0651, "step": 33932 }, { "epoch": 3.41, "learning_rate": 2.7999319137095456e-06, "loss": 0.0124, "step": 33933 }, { "epoch": 3.41, "learning_rate": 2.7989224589035876e-06, "loss": 0.1735, "step": 33934 }, { "epoch": 3.41, "learning_rate": 2.797913177194933e-06, "loss": 0.4873, "step": 33935 }, { "epoch": 3.41, "learning_rate": 2.796904068590013e-06, "loss": 0.2303, "step": 33936 }, { "epoch": 3.41, "learning_rate": 2.7958951330952487e-06, "loss": 0.109, "step": 33937 }, { "epoch": 3.41, "learning_rate": 2.7948863707170558e-06, "loss": 0.2831, "step": 33938 }, { "epoch": 3.41, "learning_rate": 2.793877781461859e-06, "loss": 0.1933, "step": 33939 }, { "epoch": 3.41, "learning_rate": 2.792869365336076e-06, "loss": 0.4314, "step": 33940 }, { "epoch": 3.41, "learning_rate": 2.791861122346119e-06, "loss": 0.2343, "step": 33941 }, { "epoch": 3.41, "learning_rate": 2.7908530524984088e-06, "loss": 0.0467, "step": 33942 }, { "epoch": 3.41, "learning_rate": 2.789845155799351e-06, "loss": 0.2948, "step": 33943 }, { "epoch": 3.41, "learning_rate": 2.7888374322553835e-06, "loss": 0.0265, "step": 33944 }, { "epoch": 3.41, "learning_rate": 2.7878298818728942e-06, "loss": 0.3529, "step": 33945 }, { "epoch": 3.41, "learning_rate": 2.786822504658295e-06, "loss": 0.6645, "step": 33946 }, { "epoch": 3.41, "learning_rate": 2.7858153006180144e-06, "loss": 0.4902, "step": 33947 }, { "epoch": 3.41, "learning_rate": 2.7848082697584564e-06, "loss": 0.0866, "step": 33948 }, { "epoch": 3.41, "learning_rate": 2.7838014120860166e-06, "loss": 0.3574, "step": 33949 }, { "epoch": 3.41, "learning_rate": 2.7827947276071163e-06, "loss": 0.0825, "step": 33950 }, { "epoch": 3.41, "learning_rate": 2.781788216328154e-06, "loss": 0.4322, "step": 33951 }, { "epoch": 3.41, "learning_rate": 2.7807818782555374e-06, "loss": 0.2796, "step": 33952 }, { "epoch": 3.41, "learning_rate": 2.7797757133956715e-06, "loss": 0.2436, "step": 33953 }, { "epoch": 3.41, "learning_rate": 2.778769721754958e-06, "loss": 0.1533, "step": 33954 }, { "epoch": 3.41, "learning_rate": 2.777763903339798e-06, "loss": 0.0726, "step": 33955 }, { "epoch": 3.41, "learning_rate": 2.776758258156594e-06, "loss": 0.6247, "step": 33956 }, { "epoch": 3.41, "learning_rate": 2.7757527862117402e-06, "loss": 0.0239, "step": 33957 }, { "epoch": 3.41, "learning_rate": 2.774747487511642e-06, "loss": 0.3539, "step": 33958 }, { "epoch": 3.41, "learning_rate": 2.7737423620626935e-06, "loss": 0.3835, "step": 33959 }, { "epoch": 3.41, "learning_rate": 2.772737409871293e-06, "loss": 0.0676, "step": 33960 }, { "epoch": 3.41, "learning_rate": 2.7717326309438264e-06, "loss": 0.5002, "step": 33961 }, { "epoch": 3.41, "learning_rate": 2.770728025286704e-06, "loss": 0.4939, "step": 33962 }, { "epoch": 3.41, "learning_rate": 2.7697235929063015e-06, "loss": 0.0179, "step": 33963 }, { "epoch": 3.41, "learning_rate": 2.7687193338090266e-06, "loss": 0.1043, "step": 33964 }, { "epoch": 3.41, "learning_rate": 2.767715248001258e-06, "loss": 0.224, "step": 33965 }, { "epoch": 3.41, "learning_rate": 2.7667113354893935e-06, "loss": 0.052, "step": 33966 }, { "epoch": 3.41, "learning_rate": 2.765707596279815e-06, "loss": 0.3386, "step": 33967 }, { "epoch": 3.41, "learning_rate": 2.764704030378917e-06, "loss": 0.2457, "step": 33968 }, { "epoch": 3.41, "learning_rate": 2.7637006377930786e-06, "loss": 0.3176, "step": 33969 }, { "epoch": 3.41, "learning_rate": 2.762697418528687e-06, "loss": 0.2426, "step": 33970 }, { "epoch": 3.41, "learning_rate": 2.7616943725921307e-06, "loss": 0.1322, "step": 33971 }, { "epoch": 3.41, "learning_rate": 2.760691499989788e-06, "loss": 0.2348, "step": 33972 }, { "epoch": 3.41, "learning_rate": 2.7596888007280405e-06, "loss": 0.4309, "step": 33973 }, { "epoch": 3.41, "learning_rate": 2.7586862748132735e-06, "loss": 0.1418, "step": 33974 }, { "epoch": 3.41, "learning_rate": 2.757683922251858e-06, "loss": 0.286, "step": 33975 }, { "epoch": 3.41, "learning_rate": 2.756681743050189e-06, "loss": 0.2007, "step": 33976 }, { "epoch": 3.41, "learning_rate": 2.755679737214628e-06, "loss": 0.0554, "step": 33977 }, { "epoch": 3.41, "learning_rate": 2.7546779047515503e-06, "loss": 0.1513, "step": 33978 }, { "epoch": 3.41, "learning_rate": 2.7536762456673435e-06, "loss": 0.364, "step": 33979 }, { "epoch": 3.41, "learning_rate": 2.752674759968383e-06, "loss": 0.0271, "step": 33980 }, { "epoch": 3.41, "learning_rate": 2.751673447661024e-06, "loss": 0.2918, "step": 33981 }, { "epoch": 3.41, "learning_rate": 2.7506723087516538e-06, "loss": 0.2871, "step": 33982 }, { "epoch": 3.41, "learning_rate": 2.7496713432466382e-06, "loss": 0.0435, "step": 33983 }, { "epoch": 3.41, "learning_rate": 2.748670551152348e-06, "loss": 0.3061, "step": 33984 }, { "epoch": 3.41, "learning_rate": 2.7476699324751518e-06, "loss": 0.0977, "step": 33985 }, { "epoch": 3.41, "learning_rate": 2.746669487221418e-06, "loss": 0.0173, "step": 33986 }, { "epoch": 3.41, "learning_rate": 2.7456692153975118e-06, "loss": 0.2281, "step": 33987 }, { "epoch": 3.41, "learning_rate": 2.744669117009797e-06, "loss": 0.2087, "step": 33988 }, { "epoch": 3.41, "learning_rate": 2.743669192064633e-06, "loss": 0.0239, "step": 33989 }, { "epoch": 3.41, "learning_rate": 2.7426694405684037e-06, "loss": 0.2831, "step": 33990 }, { "epoch": 3.41, "learning_rate": 2.7416698625274484e-06, "loss": 0.2452, "step": 33991 }, { "epoch": 3.41, "learning_rate": 2.740670457948131e-06, "loss": 0.0895, "step": 33992 }, { "epoch": 3.41, "learning_rate": 2.739671226836824e-06, "loss": 0.2737, "step": 33993 }, { "epoch": 3.41, "learning_rate": 2.7386721691998852e-06, "loss": 0.1174, "step": 33994 }, { "epoch": 3.42, "learning_rate": 2.7376732850436527e-06, "loss": 0.482, "step": 33995 }, { "epoch": 3.42, "learning_rate": 2.736674574374505e-06, "loss": 0.2785, "step": 33996 }, { "epoch": 3.42, "learning_rate": 2.7356760371987865e-06, "loss": 0.2369, "step": 33997 }, { "epoch": 3.42, "learning_rate": 2.7346776735228527e-06, "loss": 0.2542, "step": 33998 }, { "epoch": 3.42, "learning_rate": 2.733679483353062e-06, "loss": 0.9305, "step": 33999 }, { "epoch": 3.42, "learning_rate": 2.732681466695759e-06, "loss": 0.0076, "step": 34000 }, { "epoch": 3.42, "learning_rate": 2.7316836235572986e-06, "loss": 0.1311, "step": 34001 }, { "epoch": 3.42, "learning_rate": 2.730685953944029e-06, "loss": 0.177, "step": 34002 }, { "epoch": 3.42, "learning_rate": 2.7296884578623014e-06, "loss": 0.1914, "step": 34003 }, { "epoch": 3.42, "learning_rate": 2.7286911353184653e-06, "loss": 0.1255, "step": 34004 }, { "epoch": 3.42, "learning_rate": 2.727693986318861e-06, "loss": 0.1481, "step": 34005 }, { "epoch": 3.42, "learning_rate": 2.7266970108698377e-06, "loss": 0.17, "step": 34006 }, { "epoch": 3.42, "learning_rate": 2.7257002089777326e-06, "loss": 0.1366, "step": 34007 }, { "epoch": 3.42, "learning_rate": 2.7247035806489083e-06, "loss": 0.0236, "step": 34008 }, { "epoch": 3.42, "learning_rate": 2.7237071258896897e-06, "loss": 0.1993, "step": 34009 }, { "epoch": 3.42, "learning_rate": 2.722710844706414e-06, "loss": 0.2745, "step": 34010 }, { "epoch": 3.42, "learning_rate": 2.7217147371054375e-06, "loss": 0.1296, "step": 34011 }, { "epoch": 3.42, "learning_rate": 2.7207188030930976e-06, "loss": 0.1552, "step": 34012 }, { "epoch": 3.42, "learning_rate": 2.7197230426757124e-06, "loss": 0.1321, "step": 34013 }, { "epoch": 3.42, "learning_rate": 2.7187274558596407e-06, "loss": 0.0361, "step": 34014 }, { "epoch": 3.42, "learning_rate": 2.7177320426512042e-06, "loss": 0.0911, "step": 34015 }, { "epoch": 3.42, "learning_rate": 2.7167368030567474e-06, "loss": 0.3357, "step": 34016 }, { "epoch": 3.42, "learning_rate": 2.7157417370825952e-06, "loss": 0.1806, "step": 34017 }, { "epoch": 3.42, "learning_rate": 2.7147468447350833e-06, "loss": 0.126, "step": 34018 }, { "epoch": 3.42, "learning_rate": 2.713752126020542e-06, "loss": 0.1861, "step": 34019 }, { "epoch": 3.42, "learning_rate": 2.7127575809453044e-06, "loss": 0.1947, "step": 34020 }, { "epoch": 3.42, "learning_rate": 2.711763209515691e-06, "loss": 0.2738, "step": 34021 }, { "epoch": 3.42, "learning_rate": 2.7107690117380434e-06, "loss": 0.2255, "step": 34022 }, { "epoch": 3.42, "learning_rate": 2.709774987618681e-06, "loss": 0.1434, "step": 34023 }, { "epoch": 3.42, "learning_rate": 2.7087811371639173e-06, "loss": 0.0336, "step": 34024 }, { "epoch": 3.42, "learning_rate": 2.707787460380098e-06, "loss": 0.3583, "step": 34025 }, { "epoch": 3.42, "learning_rate": 2.706793957273538e-06, "loss": 0.0493, "step": 34026 }, { "epoch": 3.42, "learning_rate": 2.705800627850552e-06, "loss": 0.0191, "step": 34027 }, { "epoch": 3.42, "learning_rate": 2.7048074721174686e-06, "loss": 0.0789, "step": 34028 }, { "epoch": 3.42, "learning_rate": 2.7038144900806093e-06, "loss": 0.2791, "step": 34029 }, { "epoch": 3.42, "learning_rate": 2.7028216817462923e-06, "loss": 0.1615, "step": 34030 }, { "epoch": 3.42, "learning_rate": 2.7018290471208328e-06, "loss": 0.0312, "step": 34031 }, { "epoch": 3.42, "learning_rate": 2.7008365862105457e-06, "loss": 0.1064, "step": 34032 }, { "epoch": 3.42, "learning_rate": 2.699844299021752e-06, "loss": 0.1912, "step": 34033 }, { "epoch": 3.42, "learning_rate": 2.698852185560764e-06, "loss": 0.2629, "step": 34034 }, { "epoch": 3.42, "learning_rate": 2.697860245833893e-06, "loss": 0.0655, "step": 34035 }, { "epoch": 3.42, "learning_rate": 2.696868479847454e-06, "loss": 0.6089, "step": 34036 }, { "epoch": 3.42, "learning_rate": 2.6958768876077555e-06, "loss": 0.0629, "step": 34037 }, { "epoch": 3.42, "learning_rate": 2.694885469121112e-06, "loss": 0.414, "step": 34038 }, { "epoch": 3.42, "learning_rate": 2.693894224393826e-06, "loss": 0.3183, "step": 34039 }, { "epoch": 3.42, "learning_rate": 2.692903153432218e-06, "loss": 0.2815, "step": 34040 }, { "epoch": 3.42, "learning_rate": 2.69191225624258e-06, "loss": 0.6108, "step": 34041 }, { "epoch": 3.42, "learning_rate": 2.6909215328312177e-06, "loss": 0.4997, "step": 34042 }, { "epoch": 3.42, "learning_rate": 2.6899309832044514e-06, "loss": 0.1309, "step": 34043 }, { "epoch": 3.42, "learning_rate": 2.6889406073685705e-06, "loss": 0.2216, "step": 34044 }, { "epoch": 3.42, "learning_rate": 2.6879504053298832e-06, "loss": 0.1159, "step": 34045 }, { "epoch": 3.42, "learning_rate": 2.6869603770946902e-06, "loss": 0.1465, "step": 34046 }, { "epoch": 3.42, "learning_rate": 2.685970522669291e-06, "loss": 0.214, "step": 34047 }, { "epoch": 3.42, "learning_rate": 2.6849808420599863e-06, "loss": 0.2378, "step": 34048 }, { "epoch": 3.42, "learning_rate": 2.6839913352730716e-06, "loss": 0.1489, "step": 34049 }, { "epoch": 3.42, "learning_rate": 2.6830020023148416e-06, "loss": 0.0467, "step": 34050 }, { "epoch": 3.42, "learning_rate": 2.682012843191598e-06, "loss": 0.0283, "step": 34051 }, { "epoch": 3.42, "learning_rate": 2.681023857909629e-06, "loss": 0.0334, "step": 34052 }, { "epoch": 3.42, "learning_rate": 2.68003504647523e-06, "loss": 0.1813, "step": 34053 }, { "epoch": 3.42, "learning_rate": 2.679046408894705e-06, "loss": 0.3372, "step": 34054 }, { "epoch": 3.42, "learning_rate": 2.678057945174327e-06, "loss": 0.1327, "step": 34055 }, { "epoch": 3.42, "learning_rate": 2.67706965532039e-06, "loss": 0.2248, "step": 34056 }, { "epoch": 3.42, "learning_rate": 2.6760815393391958e-06, "loss": 0.1959, "step": 34057 }, { "epoch": 3.42, "learning_rate": 2.675093597237026e-06, "loss": 0.168, "step": 34058 }, { "epoch": 3.42, "learning_rate": 2.674105829020159e-06, "loss": 0.0249, "step": 34059 }, { "epoch": 3.42, "learning_rate": 2.6731182346948903e-06, "loss": 0.1636, "step": 34060 }, { "epoch": 3.42, "learning_rate": 2.6721308142675006e-06, "loss": 0.14, "step": 34061 }, { "epoch": 3.42, "learning_rate": 2.6711435677442755e-06, "loss": 0.1064, "step": 34062 }, { "epoch": 3.42, "learning_rate": 2.6701564951314927e-06, "loss": 0.1659, "step": 34063 }, { "epoch": 3.42, "learning_rate": 2.669169596435441e-06, "loss": 0.1138, "step": 34064 }, { "epoch": 3.42, "learning_rate": 2.668182871662396e-06, "loss": 0.0957, "step": 34065 }, { "epoch": 3.42, "learning_rate": 2.6671963208186377e-06, "loss": 0.0647, "step": 34066 }, { "epoch": 3.42, "learning_rate": 2.666209943910446e-06, "loss": 0.3075, "step": 34067 }, { "epoch": 3.42, "learning_rate": 2.6652237409440915e-06, "loss": 0.1993, "step": 34068 }, { "epoch": 3.42, "learning_rate": 2.6642377119258565e-06, "loss": 0.161, "step": 34069 }, { "epoch": 3.42, "learning_rate": 2.663251856862009e-06, "loss": 0.0622, "step": 34070 }, { "epoch": 3.42, "learning_rate": 2.662266175758834e-06, "loss": 0.1452, "step": 34071 }, { "epoch": 3.42, "learning_rate": 2.6612806686225997e-06, "loss": 0.237, "step": 34072 }, { "epoch": 3.42, "learning_rate": 2.660295335459565e-06, "loss": 0.2754, "step": 34073 }, { "epoch": 3.42, "learning_rate": 2.6593101762760173e-06, "loss": 0.1125, "step": 34074 }, { "epoch": 3.42, "learning_rate": 2.658325191078216e-06, "loss": 0.208, "step": 34075 }, { "epoch": 3.42, "learning_rate": 2.657340379872435e-06, "loss": 0.1107, "step": 34076 }, { "epoch": 3.42, "learning_rate": 2.6563557426649364e-06, "loss": 0.1233, "step": 34077 }, { "epoch": 3.42, "learning_rate": 2.655371279461989e-06, "loss": 0.1717, "step": 34078 }, { "epoch": 3.42, "learning_rate": 2.6543869902698537e-06, "loss": 0.1133, "step": 34079 }, { "epoch": 3.42, "learning_rate": 2.6534028750947993e-06, "loss": 0.4536, "step": 34080 }, { "epoch": 3.42, "learning_rate": 2.6524189339430837e-06, "loss": 0.091, "step": 34081 }, { "epoch": 3.42, "learning_rate": 2.6514351668209723e-06, "loss": 0.0646, "step": 34082 }, { "epoch": 3.42, "learning_rate": 2.6504515737347234e-06, "loss": 0.2697, "step": 34083 }, { "epoch": 3.42, "learning_rate": 2.6494681546905954e-06, "loss": 0.0708, "step": 34084 }, { "epoch": 3.42, "learning_rate": 2.6484849096948426e-06, "loss": 0.0246, "step": 34085 }, { "epoch": 3.42, "learning_rate": 2.6475018387537375e-06, "loss": 0.5176, "step": 34086 }, { "epoch": 3.42, "learning_rate": 2.646518941873518e-06, "loss": 0.1128, "step": 34087 }, { "epoch": 3.42, "learning_rate": 2.6455362190604427e-06, "loss": 0.0059, "step": 34088 }, { "epoch": 3.42, "learning_rate": 2.6445536703207764e-06, "loss": 0.1091, "step": 34089 }, { "epoch": 3.42, "learning_rate": 2.643571295660767e-06, "loss": 0.4511, "step": 34090 }, { "epoch": 3.42, "learning_rate": 2.6425890950866505e-06, "loss": 0.2596, "step": 34091 }, { "epoch": 3.42, "learning_rate": 2.6416070686046977e-06, "loss": 0.3084, "step": 34092 }, { "epoch": 3.42, "learning_rate": 2.6406252162211473e-06, "loss": 0.3607, "step": 34093 }, { "epoch": 3.43, "learning_rate": 2.6396435379422547e-06, "loss": 0.1038, "step": 34094 }, { "epoch": 3.43, "learning_rate": 2.638662033774257e-06, "loss": 0.1848, "step": 34095 }, { "epoch": 3.43, "learning_rate": 2.6376807037234074e-06, "loss": 0.1906, "step": 34096 }, { "epoch": 3.43, "learning_rate": 2.63669954779595e-06, "loss": 0.1647, "step": 34097 }, { "epoch": 3.43, "learning_rate": 2.6357185659981266e-06, "loss": 0.0854, "step": 34098 }, { "epoch": 3.43, "learning_rate": 2.634737758336172e-06, "loss": 0.0691, "step": 34099 }, { "epoch": 3.43, "learning_rate": 2.6337571248163517e-06, "loss": 0.1297, "step": 34100 }, { "epoch": 3.43, "learning_rate": 2.6327766654448803e-06, "loss": 0.1103, "step": 34101 }, { "epoch": 3.43, "learning_rate": 2.631796380228003e-06, "loss": 0.0287, "step": 34102 }, { "epoch": 3.43, "learning_rate": 2.6308162691719683e-06, "loss": 0.0999, "step": 34103 }, { "epoch": 3.43, "learning_rate": 2.6298363322830143e-06, "loss": 0.2861, "step": 34104 }, { "epoch": 3.43, "learning_rate": 2.628856569567356e-06, "loss": 0.2801, "step": 34105 }, { "epoch": 3.43, "learning_rate": 2.6278769810312484e-06, "loss": 0.2262, "step": 34106 }, { "epoch": 3.43, "learning_rate": 2.62689756668092e-06, "loss": 0.0334, "step": 34107 }, { "epoch": 3.43, "learning_rate": 2.6259183265225993e-06, "loss": 0.0674, "step": 34108 }, { "epoch": 3.43, "learning_rate": 2.6249392605625245e-06, "loss": 0.4855, "step": 34109 }, { "epoch": 3.43, "learning_rate": 2.6239603688069204e-06, "loss": 0.0299, "step": 34110 }, { "epoch": 3.43, "learning_rate": 2.622981651262019e-06, "loss": 0.1324, "step": 34111 }, { "epoch": 3.43, "learning_rate": 2.6220031079340444e-06, "loss": 0.0671, "step": 34112 }, { "epoch": 3.43, "learning_rate": 2.6210247388292297e-06, "loss": 0.4302, "step": 34113 }, { "epoch": 3.43, "learning_rate": 2.620046543953799e-06, "loss": 0.4463, "step": 34114 }, { "epoch": 3.43, "learning_rate": 2.6190685233139743e-06, "loss": 0.0251, "step": 34115 }, { "epoch": 3.43, "learning_rate": 2.6180906769159806e-06, "loss": 0.5414, "step": 34116 }, { "epoch": 3.43, "learning_rate": 2.6171130047660397e-06, "loss": 0.3175, "step": 34117 }, { "epoch": 3.43, "learning_rate": 2.6161355068703828e-06, "loss": 0.2576, "step": 34118 }, { "epoch": 3.43, "learning_rate": 2.6151581832352155e-06, "loss": 0.3843, "step": 34119 }, { "epoch": 3.43, "learning_rate": 2.6141810338667595e-06, "loss": 0.103, "step": 34120 }, { "epoch": 3.43, "learning_rate": 2.6132040587712425e-06, "loss": 0.0819, "step": 34121 }, { "epoch": 3.43, "learning_rate": 2.6122272579548834e-06, "loss": 0.4342, "step": 34122 }, { "epoch": 3.43, "learning_rate": 2.611250631423877e-06, "loss": 0.2006, "step": 34123 }, { "epoch": 3.43, "learning_rate": 2.610274179184462e-06, "loss": 0.123, "step": 34124 }, { "epoch": 3.43, "learning_rate": 2.6092979012428397e-06, "loss": 0.0591, "step": 34125 }, { "epoch": 3.43, "learning_rate": 2.6083217976052252e-06, "loss": 0.4609, "step": 34126 }, { "epoch": 3.43, "learning_rate": 2.607345868277834e-06, "loss": 0.0126, "step": 34127 }, { "epoch": 3.43, "learning_rate": 2.6063701132668637e-06, "loss": 0.319, "step": 34128 }, { "epoch": 3.43, "learning_rate": 2.6053945325785465e-06, "loss": 0.2216, "step": 34129 }, { "epoch": 3.43, "learning_rate": 2.604419126219071e-06, "loss": 0.388, "step": 34130 }, { "epoch": 3.43, "learning_rate": 2.6034438941946415e-06, "loss": 0.5031, "step": 34131 }, { "epoch": 3.43, "learning_rate": 2.602468836511487e-06, "loss": 0.0924, "step": 34132 }, { "epoch": 3.43, "learning_rate": 2.601493953175792e-06, "loss": 0.1882, "step": 34133 }, { "epoch": 3.43, "learning_rate": 2.600519244193762e-06, "loss": 0.5284, "step": 34134 }, { "epoch": 3.43, "learning_rate": 2.5995447095716085e-06, "loss": 0.0473, "step": 34135 }, { "epoch": 3.43, "learning_rate": 2.5985703493155366e-06, "loss": 0.2291, "step": 34136 }, { "epoch": 3.43, "learning_rate": 2.5975961634317247e-06, "loss": 0.2568, "step": 34137 }, { "epoch": 3.43, "learning_rate": 2.5966221519263943e-06, "loss": 0.0628, "step": 34138 }, { "epoch": 3.43, "learning_rate": 2.595648314805734e-06, "loss": 0.0863, "step": 34139 }, { "epoch": 3.43, "learning_rate": 2.5946746520759457e-06, "loss": 0.3239, "step": 34140 }, { "epoch": 3.43, "learning_rate": 2.5937011637432205e-06, "loss": 0.1018, "step": 34141 }, { "epoch": 3.43, "learning_rate": 2.592727849813754e-06, "loss": 0.0979, "step": 34142 }, { "epoch": 3.43, "learning_rate": 2.591754710293741e-06, "loss": 0.0302, "step": 34143 }, { "epoch": 3.43, "learning_rate": 2.5907817451893767e-06, "loss": 0.4769, "step": 34144 }, { "epoch": 3.43, "learning_rate": 2.589808954506849e-06, "loss": 0.0418, "step": 34145 }, { "epoch": 3.43, "learning_rate": 2.5888363382523506e-06, "loss": 0.3835, "step": 34146 }, { "epoch": 3.43, "learning_rate": 2.5878638964320687e-06, "loss": 0.2829, "step": 34147 }, { "epoch": 3.43, "learning_rate": 2.586891629052196e-06, "loss": 0.2522, "step": 34148 }, { "epoch": 3.43, "learning_rate": 2.585919536118907e-06, "loss": 0.0971, "step": 34149 }, { "epoch": 3.43, "learning_rate": 2.5849476176384103e-06, "loss": 0.1267, "step": 34150 }, { "epoch": 3.43, "learning_rate": 2.583975873616867e-06, "loss": 0.0911, "step": 34151 }, { "epoch": 3.43, "learning_rate": 2.583004304060477e-06, "loss": 0.1197, "step": 34152 }, { "epoch": 3.43, "learning_rate": 2.582032908975417e-06, "loss": 0.2035, "step": 34153 }, { "epoch": 3.43, "learning_rate": 2.5810616883678695e-06, "loss": 0.1436, "step": 34154 }, { "epoch": 3.43, "learning_rate": 2.580090642244013e-06, "loss": 0.2768, "step": 34155 }, { "epoch": 3.43, "learning_rate": 2.5791197706100322e-06, "loss": 0.5629, "step": 34156 }, { "epoch": 3.43, "learning_rate": 2.5781490734720993e-06, "loss": 0.8051, "step": 34157 }, { "epoch": 3.43, "learning_rate": 2.577178550836392e-06, "loss": 0.2119, "step": 34158 }, { "epoch": 3.43, "learning_rate": 2.5762082027090927e-06, "loss": 0.1034, "step": 34159 }, { "epoch": 3.43, "learning_rate": 2.575238029096363e-06, "loss": 0.2125, "step": 34160 }, { "epoch": 3.43, "learning_rate": 2.5742680300043973e-06, "loss": 0.5668, "step": 34161 }, { "epoch": 3.43, "learning_rate": 2.5732982054393515e-06, "loss": 0.1379, "step": 34162 }, { "epoch": 3.43, "learning_rate": 2.5723285554073973e-06, "loss": 0.0243, "step": 34163 }, { "epoch": 3.43, "learning_rate": 2.5713590799147223e-06, "loss": 0.3689, "step": 34164 }, { "epoch": 3.43, "learning_rate": 2.570389778967476e-06, "loss": 0.1136, "step": 34165 }, { "epoch": 3.43, "learning_rate": 2.569420652571829e-06, "loss": 0.2081, "step": 34166 }, { "epoch": 3.43, "learning_rate": 2.56845170073396e-06, "loss": 0.0426, "step": 34167 }, { "epoch": 3.43, "learning_rate": 2.567482923460035e-06, "loss": 0.2064, "step": 34168 }, { "epoch": 3.43, "learning_rate": 2.5665143207562016e-06, "loss": 0.086, "step": 34169 }, { "epoch": 3.43, "learning_rate": 2.565545892628638e-06, "loss": 0.17, "step": 34170 }, { "epoch": 3.43, "learning_rate": 2.5645776390835063e-06, "loss": 0.0676, "step": 34171 }, { "epoch": 3.43, "learning_rate": 2.563609560126965e-06, "loss": 0.1065, "step": 34172 }, { "epoch": 3.43, "learning_rate": 2.562641655765172e-06, "loss": 0.1404, "step": 34173 }, { "epoch": 3.43, "learning_rate": 2.5616739260042932e-06, "loss": 0.5975, "step": 34174 }, { "epoch": 3.43, "learning_rate": 2.560706370850483e-06, "loss": 0.2086, "step": 34175 }, { "epoch": 3.43, "learning_rate": 2.5597389903099e-06, "loss": 0.076, "step": 34176 }, { "epoch": 3.43, "learning_rate": 2.5587717843886993e-06, "loss": 0.0119, "step": 34177 }, { "epoch": 3.43, "learning_rate": 2.5578047530930327e-06, "loss": 0.0888, "step": 34178 }, { "epoch": 3.43, "learning_rate": 2.5568378964290582e-06, "loss": 0.086, "step": 34179 }, { "epoch": 3.43, "learning_rate": 2.555871214402925e-06, "loss": 0.1269, "step": 34180 }, { "epoch": 3.43, "learning_rate": 2.554904707020791e-06, "loss": 0.2536, "step": 34181 }, { "epoch": 3.43, "learning_rate": 2.5539383742888076e-06, "loss": 0.2411, "step": 34182 }, { "epoch": 3.43, "learning_rate": 2.5529722162131075e-06, "loss": 0.0755, "step": 34183 }, { "epoch": 3.43, "learning_rate": 2.5520062327998583e-06, "loss": 0.261, "step": 34184 }, { "epoch": 3.43, "learning_rate": 2.551040424055199e-06, "loss": 0.1957, "step": 34185 }, { "epoch": 3.43, "learning_rate": 2.5500747899852737e-06, "loss": 0.0711, "step": 34186 }, { "epoch": 3.43, "learning_rate": 2.549109330596232e-06, "loss": 0.4332, "step": 34187 }, { "epoch": 3.43, "learning_rate": 2.548144045894215e-06, "loss": 0.0911, "step": 34188 }, { "epoch": 3.43, "learning_rate": 2.5471789358853683e-06, "loss": 0.0426, "step": 34189 }, { "epoch": 3.43, "learning_rate": 2.5462140005758293e-06, "loss": 0.2968, "step": 34190 }, { "epoch": 3.43, "learning_rate": 2.545249239971741e-06, "loss": 0.102, "step": 34191 }, { "epoch": 3.44, "learning_rate": 2.5442846540792373e-06, "loss": 0.0614, "step": 34192 }, { "epoch": 3.44, "learning_rate": 2.543320242904471e-06, "loss": 0.0394, "step": 34193 }, { "epoch": 3.44, "learning_rate": 2.5423560064535665e-06, "loss": 0.0109, "step": 34194 }, { "epoch": 3.44, "learning_rate": 2.5413919447326563e-06, "loss": 0.1734, "step": 34195 }, { "epoch": 3.44, "learning_rate": 2.540428057747891e-06, "loss": 0.0945, "step": 34196 }, { "epoch": 3.44, "learning_rate": 2.5394643455053934e-06, "loss": 0.4775, "step": 34197 }, { "epoch": 3.44, "learning_rate": 2.5385008080112917e-06, "loss": 0.1831, "step": 34198 }, { "epoch": 3.44, "learning_rate": 2.537537445271727e-06, "loss": 0.1408, "step": 34199 }, { "epoch": 3.44, "learning_rate": 2.536574257292835e-06, "loss": 0.3635, "step": 34200 }, { "epoch": 3.44, "learning_rate": 2.5356112440807244e-06, "loss": 0.2893, "step": 34201 }, { "epoch": 3.44, "learning_rate": 2.5346484056415398e-06, "loss": 0.0571, "step": 34202 }, { "epoch": 3.44, "learning_rate": 2.5336857419814065e-06, "loss": 0.0503, "step": 34203 }, { "epoch": 3.44, "learning_rate": 2.532723253106449e-06, "loss": 0.0747, "step": 34204 }, { "epoch": 3.44, "learning_rate": 2.53176093902279e-06, "loss": 0.1102, "step": 34205 }, { "epoch": 3.44, "learning_rate": 2.530798799736551e-06, "loss": 0.095, "step": 34206 }, { "epoch": 3.44, "learning_rate": 2.529836835253867e-06, "loss": 0.2302, "step": 34207 }, { "epoch": 3.44, "learning_rate": 2.5288750455808495e-06, "loss": 0.0813, "step": 34208 }, { "epoch": 3.44, "learning_rate": 2.5279134307236107e-06, "loss": 0.0749, "step": 34209 }, { "epoch": 3.44, "learning_rate": 2.5269519906882954e-06, "loss": 0.0542, "step": 34210 }, { "epoch": 3.44, "learning_rate": 2.525990725480999e-06, "loss": 0.2551, "step": 34211 }, { "epoch": 3.44, "learning_rate": 2.5250296351078396e-06, "loss": 0.0323, "step": 34212 }, { "epoch": 3.44, "learning_rate": 2.524068719574946e-06, "loss": 0.2169, "step": 34213 }, { "epoch": 3.44, "learning_rate": 2.5231079788884325e-06, "loss": 0.186, "step": 34214 }, { "epoch": 3.44, "learning_rate": 2.5221474130543955e-06, "loss": 0.0767, "step": 34215 }, { "epoch": 3.44, "learning_rate": 2.5211870220789624e-06, "loss": 0.1936, "step": 34216 }, { "epoch": 3.44, "learning_rate": 2.5202268059682452e-06, "loss": 0.2456, "step": 34217 }, { "epoch": 3.44, "learning_rate": 2.5192667647283495e-06, "loss": 0.3326, "step": 34218 }, { "epoch": 3.44, "learning_rate": 2.5183068983653866e-06, "loss": 0.2021, "step": 34219 }, { "epoch": 3.44, "learning_rate": 2.517347206885462e-06, "loss": 0.2808, "step": 34220 }, { "epoch": 3.44, "learning_rate": 2.5163876902946868e-06, "loss": 0.4092, "step": 34221 }, { "epoch": 3.44, "learning_rate": 2.5154283485991636e-06, "loss": 0.2397, "step": 34222 }, { "epoch": 3.44, "learning_rate": 2.5144691818050003e-06, "loss": 0.0768, "step": 34223 }, { "epoch": 3.44, "learning_rate": 2.5135101899182923e-06, "loss": 0.3215, "step": 34224 }, { "epoch": 3.44, "learning_rate": 2.512551372945161e-06, "loss": 0.1279, "step": 34225 }, { "epoch": 3.44, "learning_rate": 2.511592730891692e-06, "loss": 0.1707, "step": 34226 }, { "epoch": 3.44, "learning_rate": 2.51063426376398e-06, "loss": 0.0989, "step": 34227 }, { "epoch": 3.44, "learning_rate": 2.5096759715681474e-06, "loss": 0.0562, "step": 34228 }, { "epoch": 3.44, "learning_rate": 2.508717854310275e-06, "loss": 0.5282, "step": 34229 }, { "epoch": 3.44, "learning_rate": 2.507759911996456e-06, "loss": 0.2069, "step": 34230 }, { "epoch": 3.44, "learning_rate": 2.5068021446328005e-06, "loss": 0.2786, "step": 34231 }, { "epoch": 3.44, "learning_rate": 2.505844552225405e-06, "loss": 0.1907, "step": 34232 }, { "epoch": 3.44, "learning_rate": 2.50488713478034e-06, "loss": 0.4055, "step": 34233 }, { "epoch": 3.44, "learning_rate": 2.503929892303722e-06, "loss": 0.0481, "step": 34234 }, { "epoch": 3.44, "learning_rate": 2.502972824801636e-06, "loss": 0.436, "step": 34235 }, { "epoch": 3.44, "learning_rate": 2.502015932280166e-06, "loss": 0.1952, "step": 34236 }, { "epoch": 3.44, "learning_rate": 2.501059214745408e-06, "loss": 0.4138, "step": 34237 }, { "epoch": 3.44, "learning_rate": 2.5001026722034437e-06, "loss": 0.3558, "step": 34238 }, { "epoch": 3.44, "learning_rate": 2.4991463046603746e-06, "loss": 0.1331, "step": 34239 }, { "epoch": 3.44, "learning_rate": 2.498190112122273e-06, "loss": 0.1311, "step": 34240 }, { "epoch": 3.44, "learning_rate": 2.49723409459522e-06, "loss": 0.1331, "step": 34241 }, { "epoch": 3.44, "learning_rate": 2.4962782520853145e-06, "loss": 0.632, "step": 34242 }, { "epoch": 3.44, "learning_rate": 2.495322584598628e-06, "loss": 0.212, "step": 34243 }, { "epoch": 3.44, "learning_rate": 2.494367092141243e-06, "loss": 0.2205, "step": 34244 }, { "epoch": 3.44, "learning_rate": 2.4934117747192443e-06, "loss": 0.104, "step": 34245 }, { "epoch": 3.44, "learning_rate": 2.4924566323387133e-06, "loss": 0.1616, "step": 34246 }, { "epoch": 3.44, "learning_rate": 2.4915016650057156e-06, "loss": 0.0698, "step": 34247 }, { "epoch": 3.44, "learning_rate": 2.4905468727263425e-06, "loss": 0.3849, "step": 34248 }, { "epoch": 3.44, "learning_rate": 2.489592255506663e-06, "loss": 0.0699, "step": 34249 }, { "epoch": 3.44, "learning_rate": 2.488637813352752e-06, "loss": 0.0505, "step": 34250 }, { "epoch": 3.44, "learning_rate": 2.4876835462706848e-06, "loss": 0.2432, "step": 34251 }, { "epoch": 3.44, "learning_rate": 2.486729454266533e-06, "loss": 0.4159, "step": 34252 }, { "epoch": 3.44, "learning_rate": 2.485775537346369e-06, "loss": 0.0708, "step": 34253 }, { "epoch": 3.44, "learning_rate": 2.4848217955162635e-06, "loss": 0.3522, "step": 34254 }, { "epoch": 3.44, "learning_rate": 2.4838682287822824e-06, "loss": 0.1701, "step": 34255 }, { "epoch": 3.44, "learning_rate": 2.4829148371504906e-06, "loss": 0.386, "step": 34256 }, { "epoch": 3.44, "learning_rate": 2.4819616206269736e-06, "loss": 0.2666, "step": 34257 }, { "epoch": 3.44, "learning_rate": 2.4810085792177793e-06, "loss": 0.1599, "step": 34258 }, { "epoch": 3.44, "learning_rate": 2.4800557129289703e-06, "loss": 0.2279, "step": 34259 }, { "epoch": 3.44, "learning_rate": 2.479103021766628e-06, "loss": 0.2918, "step": 34260 }, { "epoch": 3.44, "learning_rate": 2.478150505736797e-06, "loss": 0.2156, "step": 34261 }, { "epoch": 3.44, "learning_rate": 2.47719816484555e-06, "loss": 0.0897, "step": 34262 }, { "epoch": 3.44, "learning_rate": 2.4762459990989417e-06, "loss": 0.358, "step": 34263 }, { "epoch": 3.44, "learning_rate": 2.475294008503034e-06, "loss": 0.0275, "step": 34264 }, { "epoch": 3.44, "learning_rate": 2.4743421930638856e-06, "loss": 0.5373, "step": 34265 }, { "epoch": 3.44, "learning_rate": 2.4733905527875542e-06, "loss": 0.1248, "step": 34266 }, { "epoch": 3.44, "learning_rate": 2.4724390876800894e-06, "loss": 0.1391, "step": 34267 }, { "epoch": 3.44, "learning_rate": 2.4714877977475524e-06, "loss": 0.0552, "step": 34268 }, { "epoch": 3.44, "learning_rate": 2.4705366829959953e-06, "loss": 0.0447, "step": 34269 }, { "epoch": 3.44, "learning_rate": 2.469585743431463e-06, "loss": 0.1748, "step": 34270 }, { "epoch": 3.44, "learning_rate": 2.468634979060027e-06, "loss": 0.0152, "step": 34271 }, { "epoch": 3.44, "learning_rate": 2.4676843898877165e-06, "loss": 0.1976, "step": 34272 }, { "epoch": 3.44, "learning_rate": 2.4667339759205865e-06, "loss": 0.2236, "step": 34273 }, { "epoch": 3.44, "learning_rate": 2.4657837371646953e-06, "loss": 0.1371, "step": 34274 }, { "epoch": 3.44, "learning_rate": 2.464833673626078e-06, "loss": 0.1739, "step": 34275 }, { "epoch": 3.44, "learning_rate": 2.4638837853107766e-06, "loss": 0.2234, "step": 34276 }, { "epoch": 3.44, "learning_rate": 2.4629340722248497e-06, "loss": 0.3083, "step": 34277 }, { "epoch": 3.44, "learning_rate": 2.461984534374342e-06, "loss": 0.1025, "step": 34278 }, { "epoch": 3.44, "learning_rate": 2.4610351717652757e-06, "loss": 0.3168, "step": 34279 }, { "epoch": 3.44, "learning_rate": 2.4600859844037128e-06, "loss": 0.14, "step": 34280 }, { "epoch": 3.44, "learning_rate": 2.4591369722956845e-06, "loss": 0.1805, "step": 34281 }, { "epoch": 3.44, "learning_rate": 2.4581881354472303e-06, "loss": 0.1062, "step": 34282 }, { "epoch": 3.44, "learning_rate": 2.457239473864391e-06, "loss": 0.1883, "step": 34283 }, { "epoch": 3.44, "learning_rate": 2.4562909875532024e-06, "loss": 0.0883, "step": 34284 }, { "epoch": 3.44, "learning_rate": 2.4553426765196994e-06, "loss": 0.0289, "step": 34285 }, { "epoch": 3.44, "learning_rate": 2.4543945407699176e-06, "loss": 0.0808, "step": 34286 }, { "epoch": 3.44, "learning_rate": 2.4534465803098816e-06, "loss": 0.1729, "step": 34287 }, { "epoch": 3.44, "learning_rate": 2.4524987951456403e-06, "loss": 0.2008, "step": 34288 }, { "epoch": 3.44, "learning_rate": 2.4515511852832218e-06, "loss": 0.2829, "step": 34289 }, { "epoch": 3.44, "learning_rate": 2.4506037507286416e-06, "loss": 0.2262, "step": 34290 }, { "epoch": 3.45, "learning_rate": 2.449656491487945e-06, "loss": 0.2828, "step": 34291 }, { "epoch": 3.45, "learning_rate": 2.4487094075671566e-06, "loss": 0.5589, "step": 34292 }, { "epoch": 3.45, "learning_rate": 2.4477624989722924e-06, "loss": 0.1294, "step": 34293 }, { "epoch": 3.45, "learning_rate": 2.4468157657093904e-06, "loss": 0.3781, "step": 34294 }, { "epoch": 3.45, "learning_rate": 2.445869207784469e-06, "loss": 0.066, "step": 34295 }, { "epoch": 3.45, "learning_rate": 2.444922825203557e-06, "loss": 0.2424, "step": 34296 }, { "epoch": 3.45, "learning_rate": 2.443976617972673e-06, "loss": 0.2667, "step": 34297 }, { "epoch": 3.45, "learning_rate": 2.4430305860978386e-06, "loss": 0.5545, "step": 34298 }, { "epoch": 3.45, "learning_rate": 2.442084729585072e-06, "loss": 0.2752, "step": 34299 }, { "epoch": 3.45, "learning_rate": 2.4411390484403995e-06, "loss": 0.2799, "step": 34300 }, { "epoch": 3.45, "learning_rate": 2.440193542669832e-06, "loss": 0.2992, "step": 34301 }, { "epoch": 3.45, "learning_rate": 2.4392482122793814e-06, "loss": 0.3088, "step": 34302 }, { "epoch": 3.45, "learning_rate": 2.4383030572750833e-06, "loss": 0.2207, "step": 34303 }, { "epoch": 3.45, "learning_rate": 2.4373580776629325e-06, "loss": 0.0625, "step": 34304 }, { "epoch": 3.45, "learning_rate": 2.4364132734489442e-06, "loss": 0.6698, "step": 34305 }, { "epoch": 3.45, "learning_rate": 2.4354686446391473e-06, "loss": 0.2386, "step": 34306 }, { "epoch": 3.45, "learning_rate": 2.4345241912395366e-06, "loss": 0.2022, "step": 34307 }, { "epoch": 3.45, "learning_rate": 2.4335799132561243e-06, "loss": 0.6794, "step": 34308 }, { "epoch": 3.45, "learning_rate": 2.432635810694925e-06, "loss": 0.1402, "step": 34309 }, { "epoch": 3.45, "learning_rate": 2.4316918835619516e-06, "loss": 0.2178, "step": 34310 }, { "epoch": 3.45, "learning_rate": 2.430748131863192e-06, "loss": 0.5289, "step": 34311 }, { "epoch": 3.45, "learning_rate": 2.4298045556046677e-06, "loss": 0.0938, "step": 34312 }, { "epoch": 3.45, "learning_rate": 2.428861154792381e-06, "loss": 0.21, "step": 34313 }, { "epoch": 3.45, "learning_rate": 2.4279179294323305e-06, "loss": 0.062, "step": 34314 }, { "epoch": 3.45, "learning_rate": 2.4269748795305245e-06, "loss": 0.7736, "step": 34315 }, { "epoch": 3.45, "learning_rate": 2.426032005092952e-06, "loss": 0.0186, "step": 34316 }, { "epoch": 3.45, "learning_rate": 2.4250893061256373e-06, "loss": 0.1818, "step": 34317 }, { "epoch": 3.45, "learning_rate": 2.4241467826345563e-06, "loss": 0.4942, "step": 34318 }, { "epoch": 3.45, "learning_rate": 2.4232044346257076e-06, "loss": 0.2445, "step": 34319 }, { "epoch": 3.45, "learning_rate": 2.4222622621051026e-06, "loss": 0.0796, "step": 34320 }, { "epoch": 3.45, "learning_rate": 2.421320265078737e-06, "loss": 0.1312, "step": 34321 }, { "epoch": 3.45, "learning_rate": 2.4203784435525856e-06, "loss": 0.1069, "step": 34322 }, { "epoch": 3.45, "learning_rate": 2.419436797532657e-06, "loss": 0.2451, "step": 34323 }, { "epoch": 3.45, "learning_rate": 2.4184953270249467e-06, "loss": 0.2016, "step": 34324 }, { "epoch": 3.45, "learning_rate": 2.4175540320354293e-06, "loss": 0.0978, "step": 34325 }, { "epoch": 3.45, "learning_rate": 2.416612912570111e-06, "loss": 0.393, "step": 34326 }, { "epoch": 3.45, "learning_rate": 2.415671968634976e-06, "loss": 0.3879, "step": 34327 }, { "epoch": 3.45, "learning_rate": 2.41473120023601e-06, "loss": 0.5247, "step": 34328 }, { "epoch": 3.45, "learning_rate": 2.4137906073792014e-06, "loss": 0.1291, "step": 34329 }, { "epoch": 3.45, "learning_rate": 2.4128501900705326e-06, "loss": 0.0681, "step": 34330 }, { "epoch": 3.45, "learning_rate": 2.4119099483159914e-06, "loss": 0.2684, "step": 34331 }, { "epoch": 3.45, "learning_rate": 2.41096988212156e-06, "loss": 0.4182, "step": 34332 }, { "epoch": 3.45, "learning_rate": 2.4100299914932233e-06, "loss": 0.0607, "step": 34333 }, { "epoch": 3.45, "learning_rate": 2.4090902764369536e-06, "loss": 0.2384, "step": 34334 }, { "epoch": 3.45, "learning_rate": 2.4081507369587465e-06, "loss": 0.0991, "step": 34335 }, { "epoch": 3.45, "learning_rate": 2.407211373064566e-06, "loss": 0.1497, "step": 34336 }, { "epoch": 3.45, "learning_rate": 2.406272184760392e-06, "loss": 0.3542, "step": 34337 }, { "epoch": 3.45, "learning_rate": 2.4053331720522155e-06, "loss": 0.3214, "step": 34338 }, { "epoch": 3.45, "learning_rate": 2.404394334945995e-06, "loss": 0.1451, "step": 34339 }, { "epoch": 3.45, "learning_rate": 2.403455673447703e-06, "loss": 0.1038, "step": 34340 }, { "epoch": 3.45, "learning_rate": 2.402517187563331e-06, "loss": 0.065, "step": 34341 }, { "epoch": 3.45, "learning_rate": 2.401578877298838e-06, "loss": 0.2069, "step": 34342 }, { "epoch": 3.45, "learning_rate": 2.4006407426601984e-06, "loss": 0.0243, "step": 34343 }, { "epoch": 3.45, "learning_rate": 2.3997027836533814e-06, "loss": 0.1678, "step": 34344 }, { "epoch": 3.45, "learning_rate": 2.3987650002843586e-06, "loss": 0.1202, "step": 34345 }, { "epoch": 3.45, "learning_rate": 2.397827392559092e-06, "loss": 0.0266, "step": 34346 }, { "epoch": 3.45, "learning_rate": 2.3968899604835536e-06, "loss": 0.2209, "step": 34347 }, { "epoch": 3.45, "learning_rate": 2.395952704063702e-06, "loss": 0.1516, "step": 34348 }, { "epoch": 3.45, "learning_rate": 2.3950156233055154e-06, "loss": 0.2908, "step": 34349 }, { "epoch": 3.45, "learning_rate": 2.3940787182149424e-06, "loss": 0.0399, "step": 34350 }, { "epoch": 3.45, "learning_rate": 2.3931419887979455e-06, "loss": 0.1959, "step": 34351 }, { "epoch": 3.45, "learning_rate": 2.392205435060496e-06, "loss": 0.1084, "step": 34352 }, { "epoch": 3.45, "learning_rate": 2.391269057008556e-06, "loss": 0.083, "step": 34353 }, { "epoch": 3.45, "learning_rate": 2.3903328546480642e-06, "loss": 0.281, "step": 34354 }, { "epoch": 3.45, "learning_rate": 2.3893968279849955e-06, "loss": 0.2159, "step": 34355 }, { "epoch": 3.45, "learning_rate": 2.3884609770253084e-06, "loss": 0.1401, "step": 34356 }, { "epoch": 3.45, "learning_rate": 2.387525301774942e-06, "loss": 0.1889, "step": 34357 }, { "epoch": 3.45, "learning_rate": 2.3865898022398647e-06, "loss": 0.3321, "step": 34358 }, { "epoch": 3.45, "learning_rate": 2.3856544784260247e-06, "loss": 0.443, "step": 34359 }, { "epoch": 3.45, "learning_rate": 2.3847193303393746e-06, "loss": 0.0344, "step": 34360 }, { "epoch": 3.45, "learning_rate": 2.3837843579858653e-06, "loss": 0.0888, "step": 34361 }, { "epoch": 3.45, "learning_rate": 2.3828495613714464e-06, "loss": 0.366, "step": 34362 }, { "epoch": 3.45, "learning_rate": 2.3819149405020656e-06, "loss": 0.1731, "step": 34363 }, { "epoch": 3.45, "learning_rate": 2.3809804953836757e-06, "loss": 0.5599, "step": 34364 }, { "epoch": 3.45, "learning_rate": 2.3800462260222146e-06, "loss": 0.307, "step": 34365 }, { "epoch": 3.45, "learning_rate": 2.379112132423631e-06, "loss": 0.4149, "step": 34366 }, { "epoch": 3.45, "learning_rate": 2.3781782145938767e-06, "loss": 0.6417, "step": 34367 }, { "epoch": 3.45, "learning_rate": 2.377244472538881e-06, "loss": 0.049, "step": 34368 }, { "epoch": 3.45, "learning_rate": 2.3763109062645982e-06, "loss": 0.034, "step": 34369 }, { "epoch": 3.45, "learning_rate": 2.3753775157769677e-06, "loss": 0.2563, "step": 34370 }, { "epoch": 3.45, "learning_rate": 2.374444301081917e-06, "loss": 0.3652, "step": 34371 }, { "epoch": 3.45, "learning_rate": 2.3735112621853962e-06, "loss": 0.0593, "step": 34372 }, { "epoch": 3.45, "learning_rate": 2.3725783990933427e-06, "loss": 0.1416, "step": 34373 }, { "epoch": 3.45, "learning_rate": 2.3716457118116886e-06, "loss": 0.3719, "step": 34374 }, { "epoch": 3.45, "learning_rate": 2.3707132003463726e-06, "loss": 0.1723, "step": 34375 }, { "epoch": 3.45, "learning_rate": 2.369780864703324e-06, "loss": 0.1958, "step": 34376 }, { "epoch": 3.45, "learning_rate": 2.368848704888483e-06, "loss": 0.1265, "step": 34377 }, { "epoch": 3.45, "learning_rate": 2.3679167209077736e-06, "loss": 0.1225, "step": 34378 }, { "epoch": 3.45, "learning_rate": 2.3669849127671326e-06, "loss": 0.213, "step": 34379 }, { "epoch": 3.45, "learning_rate": 2.366053280472483e-06, "loss": 0.0328, "step": 34380 }, { "epoch": 3.45, "learning_rate": 2.365121824029767e-06, "loss": 0.3855, "step": 34381 }, { "epoch": 3.45, "learning_rate": 2.364190543444895e-06, "loss": 0.1358, "step": 34382 }, { "epoch": 3.45, "learning_rate": 2.3632594387238006e-06, "loss": 0.0786, "step": 34383 }, { "epoch": 3.45, "learning_rate": 2.362328509872411e-06, "loss": 0.5977, "step": 34384 }, { "epoch": 3.45, "learning_rate": 2.3613977568966562e-06, "loss": 0.1149, "step": 34385 }, { "epoch": 3.45, "learning_rate": 2.3604671798024436e-06, "loss": 0.2661, "step": 34386 }, { "epoch": 3.45, "learning_rate": 2.359536778595706e-06, "loss": 0.1008, "step": 34387 }, { "epoch": 3.45, "learning_rate": 2.358606553282364e-06, "loss": 0.177, "step": 34388 }, { "epoch": 3.46, "learning_rate": 2.357676503868328e-06, "loss": 0.0456, "step": 34389 }, { "epoch": 3.46, "learning_rate": 2.356746630359529e-06, "loss": 0.3987, "step": 34390 }, { "epoch": 3.46, "learning_rate": 2.355816932761875e-06, "loss": 0.3027, "step": 34391 }, { "epoch": 3.46, "learning_rate": 2.354887411081286e-06, "loss": 0.57, "step": 34392 }, { "epoch": 3.46, "learning_rate": 2.3539580653236793e-06, "loss": 0.1031, "step": 34393 }, { "epoch": 3.46, "learning_rate": 2.353028895494964e-06, "loss": 0.3297, "step": 34394 }, { "epoch": 3.46, "learning_rate": 2.3520999016010557e-06, "loss": 0.5455, "step": 34395 }, { "epoch": 3.46, "learning_rate": 2.351171083647866e-06, "loss": 0.5764, "step": 34396 }, { "epoch": 3.46, "learning_rate": 2.3502424416413006e-06, "loss": 0.3281, "step": 34397 }, { "epoch": 3.46, "learning_rate": 2.3493139755872806e-06, "loss": 0.0452, "step": 34398 }, { "epoch": 3.46, "learning_rate": 2.3483856854917084e-06, "loss": 0.0368, "step": 34399 }, { "epoch": 3.46, "learning_rate": 2.3474575713604827e-06, "loss": 0.0382, "step": 34400 }, { "epoch": 3.46, "learning_rate": 2.3465296331995255e-06, "loss": 0.3775, "step": 34401 }, { "epoch": 3.46, "learning_rate": 2.345601871014735e-06, "loss": 0.1426, "step": 34402 }, { "epoch": 3.46, "learning_rate": 2.344674284812004e-06, "loss": 0.4927, "step": 34403 }, { "epoch": 3.46, "learning_rate": 2.3437468745972536e-06, "loss": 0.2524, "step": 34404 }, { "epoch": 3.46, "learning_rate": 2.3428196403763722e-06, "loss": 0.7695, "step": 34405 }, { "epoch": 3.46, "learning_rate": 2.3418925821552696e-06, "loss": 0.5932, "step": 34406 }, { "epoch": 3.46, "learning_rate": 2.3409656999398397e-06, "loss": 0.0858, "step": 34407 }, { "epoch": 3.46, "learning_rate": 2.340038993735979e-06, "loss": 0.3236, "step": 34408 }, { "epoch": 3.46, "learning_rate": 2.339112463549592e-06, "loss": 0.13, "step": 34409 }, { "epoch": 3.46, "learning_rate": 2.3381861093865674e-06, "loss": 0.1596, "step": 34410 }, { "epoch": 3.46, "learning_rate": 2.3372599312528043e-06, "loss": 0.2748, "step": 34411 }, { "epoch": 3.46, "learning_rate": 2.3363339291541906e-06, "loss": 0.1842, "step": 34412 }, { "epoch": 3.46, "learning_rate": 2.335408103096636e-06, "loss": 0.3259, "step": 34413 }, { "epoch": 3.46, "learning_rate": 2.334482453086014e-06, "loss": 0.0457, "step": 34414 }, { "epoch": 3.46, "learning_rate": 2.3335569791282153e-06, "loss": 0.2368, "step": 34415 }, { "epoch": 3.46, "learning_rate": 2.3326316812291406e-06, "loss": 0.1464, "step": 34416 }, { "epoch": 3.46, "learning_rate": 2.3317065593946786e-06, "loss": 0.1144, "step": 34417 }, { "epoch": 3.46, "learning_rate": 2.330781613630698e-06, "loss": 0.192, "step": 34418 }, { "epoch": 3.46, "learning_rate": 2.3298568439431044e-06, "loss": 0.0914, "step": 34419 }, { "epoch": 3.46, "learning_rate": 2.3289322503377795e-06, "loss": 0.3547, "step": 34420 }, { "epoch": 3.46, "learning_rate": 2.3280078328205924e-06, "loss": 0.4595, "step": 34421 }, { "epoch": 3.46, "learning_rate": 2.3270835913974444e-06, "loss": 0.0439, "step": 34422 }, { "epoch": 3.46, "learning_rate": 2.3261595260742075e-06, "loss": 0.1756, "step": 34423 }, { "epoch": 3.46, "learning_rate": 2.3252356368567606e-06, "loss": 0.1251, "step": 34424 }, { "epoch": 3.46, "learning_rate": 2.3243119237509892e-06, "loss": 0.4614, "step": 34425 }, { "epoch": 3.46, "learning_rate": 2.3233883867627647e-06, "loss": 0.288, "step": 34426 }, { "epoch": 3.46, "learning_rate": 2.3224650258979763e-06, "loss": 0.1504, "step": 34427 }, { "epoch": 3.46, "learning_rate": 2.3215418411624857e-06, "loss": 0.1187, "step": 34428 }, { "epoch": 3.46, "learning_rate": 2.320618832562168e-06, "loss": 0.0562, "step": 34429 }, { "epoch": 3.46, "learning_rate": 2.319696000102909e-06, "loss": 0.0993, "step": 34430 }, { "epoch": 3.46, "learning_rate": 2.31877334379058e-06, "loss": 0.064, "step": 34431 }, { "epoch": 3.46, "learning_rate": 2.3178508636310336e-06, "loss": 0.9213, "step": 34432 }, { "epoch": 3.46, "learning_rate": 2.3169285596301613e-06, "loss": 0.1596, "step": 34433 }, { "epoch": 3.46, "learning_rate": 2.316006431793829e-06, "loss": 0.2597, "step": 34434 }, { "epoch": 3.46, "learning_rate": 2.3150844801278915e-06, "loss": 0.1783, "step": 34435 }, { "epoch": 3.46, "learning_rate": 2.3141627046382275e-06, "loss": 0.5134, "step": 34436 }, { "epoch": 3.46, "learning_rate": 2.313241105330699e-06, "loss": 0.2434, "step": 34437 }, { "epoch": 3.46, "learning_rate": 2.3123196822111745e-06, "loss": 0.3324, "step": 34438 }, { "epoch": 3.46, "learning_rate": 2.311398435285513e-06, "loss": 0.4138, "step": 34439 }, { "epoch": 3.46, "learning_rate": 2.310477364559577e-06, "loss": 0.1793, "step": 34440 }, { "epoch": 3.46, "learning_rate": 2.3095564700392335e-06, "loss": 0.3019, "step": 34441 }, { "epoch": 3.46, "learning_rate": 2.308635751730336e-06, "loss": 0.1557, "step": 34442 }, { "epoch": 3.46, "learning_rate": 2.3077152096387454e-06, "loss": 0.1513, "step": 34443 }, { "epoch": 3.46, "learning_rate": 2.3067948437703148e-06, "loss": 0.109, "step": 34444 }, { "epoch": 3.46, "learning_rate": 2.3058746541309216e-06, "loss": 0.0278, "step": 34445 }, { "epoch": 3.46, "learning_rate": 2.3049546407263957e-06, "loss": 0.0392, "step": 34446 }, { "epoch": 3.46, "learning_rate": 2.3040348035626014e-06, "loss": 0.3196, "step": 34447 }, { "epoch": 3.46, "learning_rate": 2.3031151426453977e-06, "loss": 0.0981, "step": 34448 }, { "epoch": 3.46, "learning_rate": 2.3021956579806337e-06, "loss": 0.2039, "step": 34449 }, { "epoch": 3.46, "learning_rate": 2.3012763495741574e-06, "loss": 0.3681, "step": 34450 }, { "epoch": 3.46, "learning_rate": 2.3003572174318243e-06, "loss": 0.3032, "step": 34451 }, { "epoch": 3.46, "learning_rate": 2.29943826155948e-06, "loss": 0.2244, "step": 34452 }, { "epoch": 3.46, "learning_rate": 2.2985194819629697e-06, "loss": 1.0437, "step": 34453 }, { "epoch": 3.46, "learning_rate": 2.297600878648145e-06, "loss": 0.0509, "step": 34454 }, { "epoch": 3.46, "learning_rate": 2.296682451620852e-06, "loss": 0.1198, "step": 34455 }, { "epoch": 3.46, "learning_rate": 2.2957642008869285e-06, "loss": 0.0706, "step": 34456 }, { "epoch": 3.46, "learning_rate": 2.294846126452227e-06, "loss": 0.1325, "step": 34457 }, { "epoch": 3.46, "learning_rate": 2.293928228322576e-06, "loss": 0.0551, "step": 34458 }, { "epoch": 3.46, "learning_rate": 2.2930105065038377e-06, "loss": 0.3863, "step": 34459 }, { "epoch": 3.46, "learning_rate": 2.292092961001837e-06, "loss": 0.0802, "step": 34460 }, { "epoch": 3.46, "learning_rate": 2.29117559182241e-06, "loss": 0.01, "step": 34461 }, { "epoch": 3.46, "learning_rate": 2.2902583989714045e-06, "loss": 0.4262, "step": 34462 }, { "epoch": 3.46, "learning_rate": 2.2893413824546596e-06, "loss": 0.3696, "step": 34463 }, { "epoch": 3.46, "learning_rate": 2.2884245422779935e-06, "loss": 0.1218, "step": 34464 }, { "epoch": 3.46, "learning_rate": 2.287507878447256e-06, "loss": 0.1476, "step": 34465 }, { "epoch": 3.46, "learning_rate": 2.2865913909682812e-06, "loss": 0.6063, "step": 34466 }, { "epoch": 3.46, "learning_rate": 2.2856750798468884e-06, "loss": 0.3132, "step": 34467 }, { "epoch": 3.46, "learning_rate": 2.2847589450889196e-06, "loss": 0.3047, "step": 34468 }, { "epoch": 3.46, "learning_rate": 2.2838429867001997e-06, "loss": 0.1714, "step": 34469 }, { "epoch": 3.46, "learning_rate": 2.282927204686561e-06, "loss": 0.2148, "step": 34470 }, { "epoch": 3.46, "learning_rate": 2.282011599053829e-06, "loss": 0.1601, "step": 34471 }, { "epoch": 3.46, "learning_rate": 2.2810961698078327e-06, "loss": 0.0731, "step": 34472 }, { "epoch": 3.46, "learning_rate": 2.280180916954393e-06, "loss": 0.1078, "step": 34473 }, { "epoch": 3.46, "learning_rate": 2.279265840499336e-06, "loss": 0.301, "step": 34474 }, { "epoch": 3.46, "learning_rate": 2.2783509404484904e-06, "loss": 0.371, "step": 34475 }, { "epoch": 3.46, "learning_rate": 2.2774362168076646e-06, "loss": 0.1436, "step": 34476 }, { "epoch": 3.46, "learning_rate": 2.2765216695826975e-06, "loss": 0.1346, "step": 34477 }, { "epoch": 3.46, "learning_rate": 2.2756072987793943e-06, "loss": 0.1646, "step": 34478 }, { "epoch": 3.46, "learning_rate": 2.2746931044035803e-06, "loss": 0.1802, "step": 34479 }, { "epoch": 3.46, "learning_rate": 2.2737790864610742e-06, "loss": 0.1834, "step": 34480 }, { "epoch": 3.46, "learning_rate": 2.2728652449576915e-06, "loss": 0.1398, "step": 34481 }, { "epoch": 3.46, "learning_rate": 2.271951579899244e-06, "loss": 0.2937, "step": 34482 }, { "epoch": 3.46, "learning_rate": 2.2710380912915506e-06, "loss": 0.2474, "step": 34483 }, { "epoch": 3.46, "learning_rate": 2.27012477914042e-06, "loss": 0.4033, "step": 34484 }, { "epoch": 3.46, "learning_rate": 2.2692116434516674e-06, "loss": 0.0498, "step": 34485 }, { "epoch": 3.46, "learning_rate": 2.2682986842311013e-06, "loss": 0.0389, "step": 34486 }, { "epoch": 3.46, "learning_rate": 2.2673859014845343e-06, "loss": 0.7015, "step": 34487 }, { "epoch": 3.47, "learning_rate": 2.2664732952177714e-06, "loss": 0.0889, "step": 34488 }, { "epoch": 3.47, "learning_rate": 2.2655608654366245e-06, "loss": 0.6094, "step": 34489 }, { "epoch": 3.47, "learning_rate": 2.2646486121468925e-06, "loss": 0.1248, "step": 34490 }, { "epoch": 3.47, "learning_rate": 2.2637365353543936e-06, "loss": 0.1687, "step": 34491 }, { "epoch": 3.47, "learning_rate": 2.2628246350649207e-06, "loss": 0.115, "step": 34492 }, { "epoch": 3.47, "learning_rate": 2.261912911284275e-06, "loss": 0.2213, "step": 34493 }, { "epoch": 3.47, "learning_rate": 2.261001364018269e-06, "loss": 0.1595, "step": 34494 }, { "epoch": 3.47, "learning_rate": 2.2600899932727016e-06, "loss": 0.1053, "step": 34495 }, { "epoch": 3.47, "learning_rate": 2.2591787990533573e-06, "loss": 0.1, "step": 34496 }, { "epoch": 3.47, "learning_rate": 2.2582677813660524e-06, "loss": 0.2359, "step": 34497 }, { "epoch": 3.47, "learning_rate": 2.257356940216585e-06, "loss": 0.0513, "step": 34498 }, { "epoch": 3.47, "learning_rate": 2.2564462756107306e-06, "loss": 0.3444, "step": 34499 }, { "epoch": 3.47, "learning_rate": 2.255535787554305e-06, "loss": 0.1778, "step": 34500 }, { "epoch": 3.47, "learning_rate": 2.2546254760530925e-06, "loss": 0.0578, "step": 34501 }, { "epoch": 3.47, "learning_rate": 2.253715341112893e-06, "loss": 0.3809, "step": 34502 }, { "epoch": 3.47, "learning_rate": 2.252805382739491e-06, "loss": 0.1995, "step": 34503 }, { "epoch": 3.47, "learning_rate": 2.251895600938676e-06, "loss": 0.0619, "step": 34504 }, { "epoch": 3.47, "learning_rate": 2.250985995716249e-06, "loss": 0.0974, "step": 34505 }, { "epoch": 3.47, "learning_rate": 2.25007656707799e-06, "loss": 0.9665, "step": 34506 }, { "epoch": 3.47, "learning_rate": 2.2491673150296797e-06, "loss": 0.2621, "step": 34507 }, { "epoch": 3.47, "learning_rate": 2.248258239577118e-06, "loss": 0.3599, "step": 34508 }, { "epoch": 3.47, "learning_rate": 2.247349340726086e-06, "loss": 0.3698, "step": 34509 }, { "epoch": 3.47, "learning_rate": 2.246440618482356e-06, "loss": 0.1597, "step": 34510 }, { "epoch": 3.47, "learning_rate": 2.245532072851727e-06, "loss": 0.3012, "step": 34511 }, { "epoch": 3.47, "learning_rate": 2.244623703839971e-06, "loss": 0.0482, "step": 34512 }, { "epoch": 3.47, "learning_rate": 2.2437155114528728e-06, "loss": 0.0908, "step": 34513 }, { "epoch": 3.47, "learning_rate": 2.2428074956962087e-06, "loss": 0.251, "step": 34514 }, { "epoch": 3.47, "learning_rate": 2.2418996565757597e-06, "loss": 0.0739, "step": 34515 }, { "epoch": 3.47, "learning_rate": 2.240991994097299e-06, "loss": 0.3125, "step": 34516 }, { "epoch": 3.47, "learning_rate": 2.2400845082666075e-06, "loss": 0.228, "step": 34517 }, { "epoch": 3.47, "learning_rate": 2.239177199089458e-06, "loss": 0.2793, "step": 34518 }, { "epoch": 3.47, "learning_rate": 2.2382700665716226e-06, "loss": 0.459, "step": 34519 }, { "epoch": 3.47, "learning_rate": 2.237363110718876e-06, "loss": 0.1502, "step": 34520 }, { "epoch": 3.47, "learning_rate": 2.236456331536991e-06, "loss": 0.0871, "step": 34521 }, { "epoch": 3.47, "learning_rate": 2.2355497290317293e-06, "loss": 0.2326, "step": 34522 }, { "epoch": 3.47, "learning_rate": 2.2346433032088764e-06, "loss": 0.105, "step": 34523 }, { "epoch": 3.47, "learning_rate": 2.2337370540741876e-06, "loss": 0.1899, "step": 34524 }, { "epoch": 3.47, "learning_rate": 2.232830981633428e-06, "loss": 0.1599, "step": 34525 }, { "epoch": 3.47, "learning_rate": 2.2319250858923734e-06, "loss": 0.2841, "step": 34526 }, { "epoch": 3.47, "learning_rate": 2.231019366856789e-06, "loss": 0.3614, "step": 34527 }, { "epoch": 3.47, "learning_rate": 2.230113824532424e-06, "loss": 0.4868, "step": 34528 }, { "epoch": 3.47, "learning_rate": 2.229208458925053e-06, "loss": 0.2255, "step": 34529 }, { "epoch": 3.47, "learning_rate": 2.2283032700404417e-06, "loss": 0.1684, "step": 34530 }, { "epoch": 3.47, "learning_rate": 2.227398257884332e-06, "loss": 0.0588, "step": 34531 }, { "epoch": 3.47, "learning_rate": 2.226493422462503e-06, "loss": 0.2183, "step": 34532 }, { "epoch": 3.47, "learning_rate": 2.2255887637807004e-06, "loss": 0.1254, "step": 34533 }, { "epoch": 3.47, "learning_rate": 2.224684281844689e-06, "loss": 0.4222, "step": 34534 }, { "epoch": 3.47, "learning_rate": 2.2237799766602173e-06, "loss": 0.0239, "step": 34535 }, { "epoch": 3.47, "learning_rate": 2.222875848233041e-06, "loss": 0.6094, "step": 34536 }, { "epoch": 3.47, "learning_rate": 2.221971896568923e-06, "loss": 0.1314, "step": 34537 }, { "epoch": 3.47, "learning_rate": 2.221068121673604e-06, "loss": 0.1435, "step": 34538 }, { "epoch": 3.47, "learning_rate": 2.220164523552837e-06, "loss": 0.2419, "step": 34539 }, { "epoch": 3.47, "learning_rate": 2.2192611022123798e-06, "loss": 0.1377, "step": 34540 }, { "epoch": 3.47, "learning_rate": 2.2183578576579787e-06, "loss": 0.1007, "step": 34541 }, { "epoch": 3.47, "learning_rate": 2.2174547898953724e-06, "loss": 0.0614, "step": 34542 }, { "epoch": 3.47, "learning_rate": 2.216551898930319e-06, "loss": 0.7979, "step": 34543 }, { "epoch": 3.47, "learning_rate": 2.215649184768558e-06, "loss": 1.438, "step": 34544 }, { "epoch": 3.47, "learning_rate": 2.2147466474158373e-06, "loss": 0.0723, "step": 34545 }, { "epoch": 3.47, "learning_rate": 2.2138442868778995e-06, "loss": 0.072, "step": 34546 }, { "epoch": 3.47, "learning_rate": 2.212942103160487e-06, "loss": 0.426, "step": 34547 }, { "epoch": 3.47, "learning_rate": 2.2120400962693378e-06, "loss": 0.1087, "step": 34548 }, { "epoch": 3.47, "learning_rate": 2.211138266210194e-06, "loss": 0.3795, "step": 34549 }, { "epoch": 3.47, "learning_rate": 2.2102366129887986e-06, "loss": 0.1106, "step": 34550 }, { "epoch": 3.47, "learning_rate": 2.209335136610883e-06, "loss": 0.261, "step": 34551 }, { "epoch": 3.47, "learning_rate": 2.2084338370821856e-06, "loss": 0.257, "step": 34552 }, { "epoch": 3.47, "learning_rate": 2.2075327144084456e-06, "loss": 0.0453, "step": 34553 }, { "epoch": 3.47, "learning_rate": 2.2066317685953886e-06, "loss": 0.0152, "step": 34554 }, { "epoch": 3.47, "learning_rate": 2.205730999648763e-06, "loss": 0.506, "step": 34555 }, { "epoch": 3.47, "learning_rate": 2.2048304075742874e-06, "loss": 0.098, "step": 34556 }, { "epoch": 3.47, "learning_rate": 2.2039299923776913e-06, "loss": 0.7673, "step": 34557 }, { "epoch": 3.47, "learning_rate": 2.203029754064716e-06, "loss": 0.259, "step": 34558 }, { "epoch": 3.47, "learning_rate": 2.2021296926410872e-06, "loss": 0.1514, "step": 34559 }, { "epoch": 3.47, "learning_rate": 2.201229808112527e-06, "loss": 0.039, "step": 34560 }, { "epoch": 3.47, "learning_rate": 2.2003301004847674e-06, "loss": 0.5183, "step": 34561 }, { "epoch": 3.47, "learning_rate": 2.199430569763531e-06, "loss": 0.1907, "step": 34562 }, { "epoch": 3.47, "learning_rate": 2.198531215954539e-06, "loss": 0.2314, "step": 34563 }, { "epoch": 3.47, "learning_rate": 2.197632039063521e-06, "loss": 0.4294, "step": 34564 }, { "epoch": 3.47, "learning_rate": 2.196733039096198e-06, "loss": 0.7765, "step": 34565 }, { "epoch": 3.47, "learning_rate": 2.1958342160582867e-06, "loss": 0.3331, "step": 34566 }, { "epoch": 3.47, "learning_rate": 2.1949355699555083e-06, "loss": 0.1697, "step": 34567 }, { "epoch": 3.47, "learning_rate": 2.194037100793582e-06, "loss": 0.5327, "step": 34568 }, { "epoch": 3.47, "learning_rate": 2.19313880857823e-06, "loss": 0.7445, "step": 34569 }, { "epoch": 3.47, "learning_rate": 2.192240693315164e-06, "loss": 0.1674, "step": 34570 }, { "epoch": 3.47, "learning_rate": 2.1913427550100927e-06, "loss": 0.05, "step": 34571 }, { "epoch": 3.47, "learning_rate": 2.190444993668742e-06, "loss": 0.4151, "step": 34572 }, { "epoch": 3.47, "learning_rate": 2.189547409296824e-06, "loss": 0.5574, "step": 34573 }, { "epoch": 3.47, "learning_rate": 2.1886500019000366e-06, "loss": 0.114, "step": 34574 }, { "epoch": 3.47, "learning_rate": 2.187752771484106e-06, "loss": 0.1913, "step": 34575 }, { "epoch": 3.47, "learning_rate": 2.1868557180547377e-06, "loss": 0.0503, "step": 34576 }, { "epoch": 3.47, "learning_rate": 2.1859588416176403e-06, "loss": 0.2045, "step": 34577 }, { "epoch": 3.47, "learning_rate": 2.1850621421785153e-06, "loss": 0.114, "step": 34578 }, { "epoch": 3.47, "learning_rate": 2.184165619743076e-06, "loss": 0.124, "step": 34579 }, { "epoch": 3.47, "learning_rate": 2.1832692743170267e-06, "loss": 0.0409, "step": 34580 }, { "epoch": 3.47, "learning_rate": 2.182373105906067e-06, "loss": 0.029, "step": 34581 }, { "epoch": 3.47, "learning_rate": 2.1814771145159053e-06, "loss": 0.2331, "step": 34582 }, { "epoch": 3.47, "learning_rate": 2.1805813001522367e-06, "loss": 0.0097, "step": 34583 }, { "epoch": 3.47, "learning_rate": 2.1796856628207706e-06, "loss": 0.2075, "step": 34584 }, { "epoch": 3.47, "learning_rate": 2.178790202527192e-06, "loss": 0.4884, "step": 34585 }, { "epoch": 3.48, "learning_rate": 2.1778949192772168e-06, "loss": 0.1101, "step": 34586 }, { "epoch": 3.48, "learning_rate": 2.1769998130765368e-06, "loss": 0.5265, "step": 34587 }, { "epoch": 3.48, "learning_rate": 2.1761048839308374e-06, "loss": 0.106, "step": 34588 }, { "epoch": 3.48, "learning_rate": 2.175210131845827e-06, "loss": 0.1725, "step": 34589 }, { "epoch": 3.48, "learning_rate": 2.1743155568271922e-06, "loss": 0.2768, "step": 34590 }, { "epoch": 3.48, "learning_rate": 2.1734211588806272e-06, "loss": 0.1073, "step": 34591 }, { "epoch": 3.48, "learning_rate": 2.1725269380118284e-06, "loss": 0.0925, "step": 34592 }, { "epoch": 3.48, "learning_rate": 2.1716328942264773e-06, "loss": 0.0233, "step": 34593 }, { "epoch": 3.48, "learning_rate": 2.170739027530273e-06, "loss": 0.2651, "step": 34594 }, { "epoch": 3.48, "learning_rate": 2.169845337928894e-06, "loss": 0.4118, "step": 34595 }, { "epoch": 3.48, "learning_rate": 2.168951825428036e-06, "loss": 0.0456, "step": 34596 }, { "epoch": 3.48, "learning_rate": 2.1680584900333778e-06, "loss": 0.5019, "step": 34597 }, { "epoch": 3.48, "learning_rate": 2.167165331750608e-06, "loss": 0.3139, "step": 34598 }, { "epoch": 3.48, "learning_rate": 2.166272350585412e-06, "loss": 0.0901, "step": 34599 }, { "epoch": 3.48, "learning_rate": 2.1653795465434654e-06, "loss": 0.2693, "step": 34600 }, { "epoch": 3.48, "learning_rate": 2.1644869196304605e-06, "loss": 0.1945, "step": 34601 }, { "epoch": 3.48, "learning_rate": 2.1635944698520724e-06, "loss": 0.4171, "step": 34602 }, { "epoch": 3.48, "learning_rate": 2.1627021972139706e-06, "loss": 0.3092, "step": 34603 }, { "epoch": 3.48, "learning_rate": 2.1618101017218494e-06, "loss": 0.1699, "step": 34604 }, { "epoch": 3.48, "learning_rate": 2.160918183381382e-06, "loss": 0.0433, "step": 34605 }, { "epoch": 3.48, "learning_rate": 2.1600264421982295e-06, "loss": 0.2045, "step": 34606 }, { "epoch": 3.48, "learning_rate": 2.159134878178085e-06, "loss": 0.0776, "step": 34607 }, { "epoch": 3.48, "learning_rate": 2.1582434913266136e-06, "loss": 0.2648, "step": 34608 }, { "epoch": 3.48, "learning_rate": 2.1573522816494908e-06, "loss": 0.3013, "step": 34609 }, { "epoch": 3.48, "learning_rate": 2.156461249152385e-06, "loss": 0.1786, "step": 34610 }, { "epoch": 3.48, "learning_rate": 2.155570393840969e-06, "loss": 0.0979, "step": 34611 }, { "epoch": 3.48, "learning_rate": 2.1546797157209073e-06, "loss": 0.0514, "step": 34612 }, { "epoch": 3.48, "learning_rate": 2.1537892147978734e-06, "loss": 0.2062, "step": 34613 }, { "epoch": 3.48, "learning_rate": 2.152898891077528e-06, "loss": 0.1459, "step": 34614 }, { "epoch": 3.48, "learning_rate": 2.1520087445655477e-06, "loss": 0.3098, "step": 34615 }, { "epoch": 3.48, "learning_rate": 2.151118775267584e-06, "loss": 0.1457, "step": 34616 }, { "epoch": 3.48, "learning_rate": 2.1502289831893063e-06, "loss": 0.0336, "step": 34617 }, { "epoch": 3.48, "learning_rate": 2.149339368336376e-06, "loss": 0.1249, "step": 34618 }, { "epoch": 3.48, "learning_rate": 2.1484499307144622e-06, "loss": 0.0777, "step": 34619 }, { "epoch": 3.48, "learning_rate": 2.1475606703292106e-06, "loss": 0.0793, "step": 34620 }, { "epoch": 3.48, "learning_rate": 2.1466715871862897e-06, "loss": 0.3054, "step": 34621 }, { "epoch": 3.48, "learning_rate": 2.1457826812913552e-06, "loss": 0.5188, "step": 34622 }, { "epoch": 3.48, "learning_rate": 2.1448939526500623e-06, "loss": 0.4748, "step": 34623 }, { "epoch": 3.48, "learning_rate": 2.14400540126807e-06, "loss": 0.0709, "step": 34624 }, { "epoch": 3.48, "learning_rate": 2.143117027151027e-06, "loss": 0.0381, "step": 34625 }, { "epoch": 3.48, "learning_rate": 2.1422288303045924e-06, "loss": 0.127, "step": 34626 }, { "epoch": 3.48, "learning_rate": 2.1413408107344146e-06, "loss": 0.1129, "step": 34627 }, { "epoch": 3.48, "learning_rate": 2.140452968446146e-06, "loss": 0.1098, "step": 34628 }, { "epoch": 3.48, "learning_rate": 2.139565303445439e-06, "loss": 0.0349, "step": 34629 }, { "epoch": 3.48, "learning_rate": 2.138677815737938e-06, "loss": 0.6212, "step": 34630 }, { "epoch": 3.48, "learning_rate": 2.1377905053292903e-06, "loss": 0.1217, "step": 34631 }, { "epoch": 3.48, "learning_rate": 2.136903372225143e-06, "loss": 0.2977, "step": 34632 }, { "epoch": 3.48, "learning_rate": 2.136016416431149e-06, "loss": 0.1694, "step": 34633 }, { "epoch": 3.48, "learning_rate": 2.1351296379529438e-06, "loss": 0.1976, "step": 34634 }, { "epoch": 3.48, "learning_rate": 2.1342430367961696e-06, "loss": 0.2099, "step": 34635 }, { "epoch": 3.48, "learning_rate": 2.133356612966475e-06, "loss": 0.1424, "step": 34636 }, { "epoch": 3.48, "learning_rate": 2.1324703664695055e-06, "loss": 0.4446, "step": 34637 }, { "epoch": 3.48, "learning_rate": 2.1315842973108803e-06, "loss": 0.3159, "step": 34638 }, { "epoch": 3.48, "learning_rate": 2.1306984054962574e-06, "loss": 0.0773, "step": 34639 }, { "epoch": 3.48, "learning_rate": 2.1298126910312664e-06, "loss": 0.165, "step": 34640 }, { "epoch": 3.48, "learning_rate": 2.1289271539215462e-06, "loss": 0.2032, "step": 34641 }, { "epoch": 3.48, "learning_rate": 2.1280417941727316e-06, "loss": 0.0687, "step": 34642 }, { "epoch": 3.48, "learning_rate": 2.127156611790455e-06, "loss": 0.4641, "step": 34643 }, { "epoch": 3.48, "learning_rate": 2.1262716067803524e-06, "loss": 0.2142, "step": 34644 }, { "epoch": 3.48, "learning_rate": 2.125386779148052e-06, "loss": 0.0996, "step": 34645 }, { "epoch": 3.48, "learning_rate": 2.1245021288991795e-06, "loss": 0.143, "step": 34646 }, { "epoch": 3.48, "learning_rate": 2.1236176560393807e-06, "loss": 0.656, "step": 34647 }, { "epoch": 3.48, "learning_rate": 2.122733360574274e-06, "loss": 0.7493, "step": 34648 }, { "epoch": 3.48, "learning_rate": 2.1218492425094783e-06, "loss": 0.0868, "step": 34649 }, { "epoch": 3.48, "learning_rate": 2.1209653018506323e-06, "loss": 0.3568, "step": 34650 }, { "epoch": 3.48, "learning_rate": 2.1200815386033657e-06, "loss": 0.2664, "step": 34651 }, { "epoch": 3.48, "learning_rate": 2.119197952773283e-06, "loss": 0.2774, "step": 34652 }, { "epoch": 3.48, "learning_rate": 2.1183145443660236e-06, "loss": 0.207, "step": 34653 }, { "epoch": 3.48, "learning_rate": 2.1174313133872027e-06, "loss": 0.2479, "step": 34654 }, { "epoch": 3.48, "learning_rate": 2.1165482598424423e-06, "loss": 0.0578, "step": 34655 }, { "epoch": 3.48, "learning_rate": 2.115665383737362e-06, "loss": 0.0968, "step": 34656 }, { "epoch": 3.48, "learning_rate": 2.1147826850775796e-06, "loss": 0.576, "step": 34657 }, { "epoch": 3.48, "learning_rate": 2.1139001638687115e-06, "loss": 0.1665, "step": 34658 }, { "epoch": 3.48, "learning_rate": 2.113017820116376e-06, "loss": 0.0533, "step": 34659 }, { "epoch": 3.48, "learning_rate": 2.1121356538261825e-06, "loss": 0.0952, "step": 34660 }, { "epoch": 3.48, "learning_rate": 2.1112536650037527e-06, "loss": 0.1697, "step": 34661 }, { "epoch": 3.48, "learning_rate": 2.110371853654692e-06, "loss": 0.2409, "step": 34662 }, { "epoch": 3.48, "learning_rate": 2.1094902197846164e-06, "loss": 0.1924, "step": 34663 }, { "epoch": 3.48, "learning_rate": 2.1086087633991314e-06, "loss": 0.3022, "step": 34664 }, { "epoch": 3.48, "learning_rate": 2.1077274845038587e-06, "loss": 0.2184, "step": 34665 }, { "epoch": 3.48, "learning_rate": 2.106846383104387e-06, "loss": 0.0767, "step": 34666 }, { "epoch": 3.48, "learning_rate": 2.1059654592063393e-06, "loss": 0.0375, "step": 34667 }, { "epoch": 3.48, "learning_rate": 2.105084712815314e-06, "loss": 0.0385, "step": 34668 }, { "epoch": 3.48, "learning_rate": 2.1042041439369197e-06, "loss": 0.0819, "step": 34669 }, { "epoch": 3.48, "learning_rate": 2.103323752576759e-06, "loss": 0.2073, "step": 34670 }, { "epoch": 3.48, "learning_rate": 2.1024435387404306e-06, "loss": 0.0765, "step": 34671 }, { "epoch": 3.48, "learning_rate": 2.10156350243354e-06, "loss": 0.0495, "step": 34672 }, { "epoch": 3.48, "learning_rate": 2.1006836436616896e-06, "loss": 0.1705, "step": 34673 }, { "epoch": 3.48, "learning_rate": 2.099803962430471e-06, "loss": 0.1408, "step": 34674 }, { "epoch": 3.48, "learning_rate": 2.098924458745487e-06, "loss": 0.1429, "step": 34675 }, { "epoch": 3.48, "learning_rate": 2.0980451326123327e-06, "loss": 0.4622, "step": 34676 }, { "epoch": 3.48, "learning_rate": 2.0971659840366074e-06, "loss": 0.0357, "step": 34677 }, { "epoch": 3.48, "learning_rate": 2.0962870130238964e-06, "loss": 0.0468, "step": 34678 }, { "epoch": 3.48, "learning_rate": 2.0954082195798086e-06, "loss": 0.6075, "step": 34679 }, { "epoch": 3.48, "learning_rate": 2.0945296037099258e-06, "loss": 0.4127, "step": 34680 }, { "epoch": 3.48, "learning_rate": 2.093651165419834e-06, "loss": 0.4127, "step": 34681 }, { "epoch": 3.48, "learning_rate": 2.092772904715132e-06, "loss": 0.1095, "step": 34682 }, { "epoch": 3.48, "learning_rate": 2.0918948216014154e-06, "loss": 0.3491, "step": 34683 }, { "epoch": 3.48, "learning_rate": 2.091016916084253e-06, "loss": 0.0926, "step": 34684 }, { "epoch": 3.49, "learning_rate": 2.0901391881692465e-06, "loss": 0.7312, "step": 34685 }, { "epoch": 3.49, "learning_rate": 2.089261637861979e-06, "loss": 0.0662, "step": 34686 }, { "epoch": 3.49, "learning_rate": 2.0883842651680287e-06, "loss": 0.4707, "step": 34687 }, { "epoch": 3.49, "learning_rate": 2.0875070700929843e-06, "loss": 0.8186, "step": 34688 }, { "epoch": 3.49, "learning_rate": 2.086630052642425e-06, "loss": 0.3576, "step": 34689 }, { "epoch": 3.49, "learning_rate": 2.085753212821937e-06, "loss": 0.2294, "step": 34690 }, { "epoch": 3.49, "learning_rate": 2.084876550637095e-06, "loss": 0.2425, "step": 34691 }, { "epoch": 3.49, "learning_rate": 2.0840000660934776e-06, "loss": 0.9148, "step": 34692 }, { "epoch": 3.49, "learning_rate": 2.0831237591966644e-06, "loss": 0.4527, "step": 34693 }, { "epoch": 3.49, "learning_rate": 2.0822476299522307e-06, "loss": 0.1719, "step": 34694 }, { "epoch": 3.49, "learning_rate": 2.081371678365749e-06, "loss": 0.1179, "step": 34695 }, { "epoch": 3.49, "learning_rate": 2.080495904442804e-06, "loss": 0.0334, "step": 34696 }, { "epoch": 3.49, "learning_rate": 2.0796203081889654e-06, "loss": 0.2617, "step": 34697 }, { "epoch": 3.49, "learning_rate": 2.0787448896097915e-06, "loss": 0.0613, "step": 34698 }, { "epoch": 3.49, "learning_rate": 2.0778696487108682e-06, "loss": 0.3814, "step": 34699 }, { "epoch": 3.49, "learning_rate": 2.076994585497758e-06, "loss": 0.3025, "step": 34700 }, { "epoch": 3.49, "learning_rate": 2.0761196999760355e-06, "loss": 0.3731, "step": 34701 }, { "epoch": 3.49, "learning_rate": 2.0752449921512638e-06, "loss": 0.4292, "step": 34702 }, { "epoch": 3.49, "learning_rate": 2.074370462029008e-06, "loss": 0.0986, "step": 34703 }, { "epoch": 3.49, "learning_rate": 2.073496109614834e-06, "loss": 0.4996, "step": 34704 }, { "epoch": 3.49, "learning_rate": 2.0726219349143107e-06, "loss": 0.0797, "step": 34705 }, { "epoch": 3.49, "learning_rate": 2.0717479379329963e-06, "loss": 0.3075, "step": 34706 }, { "epoch": 3.49, "learning_rate": 2.0708741186764535e-06, "loss": 0.2849, "step": 34707 }, { "epoch": 3.49, "learning_rate": 2.0700004771502412e-06, "loss": 0.4752, "step": 34708 }, { "epoch": 3.49, "learning_rate": 2.0691270133599215e-06, "loss": 0.1678, "step": 34709 }, { "epoch": 3.49, "learning_rate": 2.0682537273110467e-06, "loss": 0.3913, "step": 34710 }, { "epoch": 3.49, "learning_rate": 2.0673806190091924e-06, "loss": 0.7836, "step": 34711 }, { "epoch": 3.49, "learning_rate": 2.0665076884598944e-06, "loss": 0.028, "step": 34712 }, { "epoch": 3.49, "learning_rate": 2.065634935668711e-06, "loss": 0.0072, "step": 34713 }, { "epoch": 3.49, "learning_rate": 2.0647623606412047e-06, "loss": 0.4073, "step": 34714 }, { "epoch": 3.49, "learning_rate": 2.063889963382928e-06, "loss": 0.2871, "step": 34715 }, { "epoch": 3.49, "learning_rate": 2.063017743899419e-06, "loss": 0.234, "step": 34716 }, { "epoch": 3.49, "learning_rate": 2.0621457021962407e-06, "loss": 0.0454, "step": 34717 }, { "epoch": 3.49, "learning_rate": 2.0612738382789388e-06, "loss": 0.3042, "step": 34718 }, { "epoch": 3.49, "learning_rate": 2.0604021521530648e-06, "loss": 0.5481, "step": 34719 }, { "epoch": 3.49, "learning_rate": 2.059530643824161e-06, "loss": 0.1954, "step": 34720 }, { "epoch": 3.49, "learning_rate": 2.058659313297774e-06, "loss": 0.1161, "step": 34721 }, { "epoch": 3.49, "learning_rate": 2.0577881605794513e-06, "loss": 0.331, "step": 34722 }, { "epoch": 3.49, "learning_rate": 2.0569171856747326e-06, "loss": 0.6525, "step": 34723 }, { "epoch": 3.49, "learning_rate": 2.05604638858916e-06, "loss": 0.1545, "step": 34724 }, { "epoch": 3.49, "learning_rate": 2.055175769328289e-06, "loss": 0.3896, "step": 34725 }, { "epoch": 3.49, "learning_rate": 2.0543053278976422e-06, "loss": 0.1321, "step": 34726 }, { "epoch": 3.49, "learning_rate": 2.0534350643027578e-06, "loss": 0.2445, "step": 34727 }, { "epoch": 3.49, "learning_rate": 2.0525649785491885e-06, "loss": 0.0209, "step": 34728 }, { "epoch": 3.49, "learning_rate": 2.0516950706424697e-06, "loss": 0.3889, "step": 34729 }, { "epoch": 3.49, "learning_rate": 2.0508253405881207e-06, "loss": 0.1938, "step": 34730 }, { "epoch": 3.49, "learning_rate": 2.049955788391693e-06, "loss": 0.3616, "step": 34731 }, { "epoch": 3.49, "learning_rate": 2.049086414058713e-06, "loss": 0.7083, "step": 34732 }, { "epoch": 3.49, "learning_rate": 2.0482172175947122e-06, "loss": 0.0211, "step": 34733 }, { "epoch": 3.49, "learning_rate": 2.047348199005227e-06, "loss": 0.0574, "step": 34734 }, { "epoch": 3.49, "learning_rate": 2.046479358295782e-06, "loss": 0.1873, "step": 34735 }, { "epoch": 3.49, "learning_rate": 2.0456106954719065e-06, "loss": 0.1738, "step": 34736 }, { "epoch": 3.49, "learning_rate": 2.044742210539133e-06, "loss": 0.1892, "step": 34737 }, { "epoch": 3.49, "learning_rate": 2.043873903502984e-06, "loss": 0.144, "step": 34738 }, { "epoch": 3.49, "learning_rate": 2.043005774368988e-06, "loss": 0.2699, "step": 34739 }, { "epoch": 3.49, "learning_rate": 2.042137823142667e-06, "loss": 0.0621, "step": 34740 }, { "epoch": 3.49, "learning_rate": 2.0412700498295433e-06, "loss": 0.3271, "step": 34741 }, { "epoch": 3.49, "learning_rate": 2.0404024544351397e-06, "loss": 0.1465, "step": 34742 }, { "epoch": 3.49, "learning_rate": 2.0395350369649846e-06, "loss": 0.2497, "step": 34743 }, { "epoch": 3.49, "learning_rate": 2.0386677974245905e-06, "loss": 0.1041, "step": 34744 }, { "epoch": 3.49, "learning_rate": 2.0378007358194697e-06, "loss": 0.4765, "step": 34745 }, { "epoch": 3.49, "learning_rate": 2.036933852155154e-06, "loss": 0.3929, "step": 34746 }, { "epoch": 3.49, "learning_rate": 2.0360671464371528e-06, "loss": 0.1075, "step": 34747 }, { "epoch": 3.49, "learning_rate": 2.0352006186709813e-06, "loss": 0.4051, "step": 34748 }, { "epoch": 3.49, "learning_rate": 2.034334268862156e-06, "loss": 0.2882, "step": 34749 }, { "epoch": 3.49, "learning_rate": 2.0334680970161913e-06, "loss": 0.244, "step": 34750 }, { "epoch": 3.49, "learning_rate": 2.032602103138593e-06, "loss": 0.006, "step": 34751 }, { "epoch": 3.49, "learning_rate": 2.0317362872348776e-06, "loss": 0.1095, "step": 34752 }, { "epoch": 3.49, "learning_rate": 2.03087064931055e-06, "loss": 0.2469, "step": 34753 }, { "epoch": 3.49, "learning_rate": 2.030005189371126e-06, "loss": 0.2193, "step": 34754 }, { "epoch": 3.49, "learning_rate": 2.029139907422104e-06, "loss": 0.1106, "step": 34755 }, { "epoch": 3.49, "learning_rate": 2.0282748034689936e-06, "loss": 0.0298, "step": 34756 }, { "epoch": 3.49, "learning_rate": 2.02740987751731e-06, "loss": 0.207, "step": 34757 }, { "epoch": 3.49, "learning_rate": 2.0265451295725423e-06, "loss": 0.4482, "step": 34758 }, { "epoch": 3.49, "learning_rate": 2.0256805596401962e-06, "loss": 0.0724, "step": 34759 }, { "epoch": 3.49, "learning_rate": 2.0248161677257806e-06, "loss": 0.2932, "step": 34760 }, { "epoch": 3.49, "learning_rate": 2.0239519538348006e-06, "loss": 0.7187, "step": 34761 }, { "epoch": 3.49, "learning_rate": 2.023087917972736e-06, "loss": 0.1537, "step": 34762 }, { "epoch": 3.49, "learning_rate": 2.0222240601451013e-06, "loss": 0.2045, "step": 34763 }, { "epoch": 3.49, "learning_rate": 2.02136038035739e-06, "loss": 0.1181, "step": 34764 }, { "epoch": 3.49, "learning_rate": 2.020496878615097e-06, "loss": 0.506, "step": 34765 }, { "epoch": 3.49, "learning_rate": 2.019633554923718e-06, "loss": 0.1558, "step": 34766 }, { "epoch": 3.49, "learning_rate": 2.018770409288748e-06, "loss": 0.7627, "step": 34767 }, { "epoch": 3.49, "learning_rate": 2.017907441715677e-06, "loss": 0.2704, "step": 34768 }, { "epoch": 3.49, "learning_rate": 2.0170446522100008e-06, "loss": 0.236, "step": 34769 }, { "epoch": 3.49, "learning_rate": 2.0161820407772035e-06, "loss": 0.1017, "step": 34770 }, { "epoch": 3.49, "learning_rate": 2.0153196074227788e-06, "loss": 0.2506, "step": 34771 }, { "epoch": 3.49, "learning_rate": 2.0144573521522146e-06, "loss": 0.2443, "step": 34772 }, { "epoch": 3.49, "learning_rate": 2.0135952749710005e-06, "loss": 0.1672, "step": 34773 }, { "epoch": 3.49, "learning_rate": 2.0127333758846123e-06, "loss": 0.4509, "step": 34774 }, { "epoch": 3.49, "learning_rate": 2.0118716548985517e-06, "loss": 0.5852, "step": 34775 }, { "epoch": 3.49, "learning_rate": 2.0110101120182812e-06, "loss": 0.8209, "step": 34776 }, { "epoch": 3.49, "learning_rate": 2.010148747249303e-06, "loss": 0.6238, "step": 34777 }, { "epoch": 3.49, "learning_rate": 2.0092875605970894e-06, "loss": 0.3597, "step": 34778 }, { "epoch": 3.49, "learning_rate": 2.0084265520671196e-06, "loss": 0.203, "step": 34779 }, { "epoch": 3.49, "learning_rate": 2.0075657216648757e-06, "loss": 0.0509, "step": 34780 }, { "epoch": 3.49, "learning_rate": 2.0067050693958366e-06, "loss": 0.3211, "step": 34781 }, { "epoch": 3.49, "learning_rate": 2.0058445952654745e-06, "loss": 0.0932, "step": 34782 }, { "epoch": 3.5, "learning_rate": 2.0049842992792686e-06, "loss": 0.6262, "step": 34783 }, { "epoch": 3.5, "learning_rate": 2.0041241814426945e-06, "loss": 0.2841, "step": 34784 }, { "epoch": 3.5, "learning_rate": 2.0032642417612215e-06, "loss": 0.254, "step": 34785 }, { "epoch": 3.5, "learning_rate": 2.0024044802403242e-06, "loss": 0.1812, "step": 34786 }, { "epoch": 3.5, "learning_rate": 2.0015448968854756e-06, "loss": 0.2882, "step": 34787 }, { "epoch": 3.5, "learning_rate": 2.000685491702141e-06, "loss": 0.269, "step": 34788 }, { "epoch": 3.5, "learning_rate": 1.9998262646957997e-06, "loss": 0.1097, "step": 34789 }, { "epoch": 3.5, "learning_rate": 1.9989672158719074e-06, "loss": 0.1199, "step": 34790 }, { "epoch": 3.5, "learning_rate": 1.9981083452359292e-06, "loss": 0.2922, "step": 34791 }, { "epoch": 3.5, "learning_rate": 1.997249652793345e-06, "loss": 0.2243, "step": 34792 }, { "epoch": 3.5, "learning_rate": 1.996391138549616e-06, "loss": 0.1098, "step": 34793 }, { "epoch": 3.5, "learning_rate": 1.9955328025101882e-06, "loss": 0.2731, "step": 34794 }, { "epoch": 3.5, "learning_rate": 1.9946746446805406e-06, "loss": 0.0914, "step": 34795 }, { "epoch": 3.5, "learning_rate": 1.9938166650661326e-06, "loss": 0.1231, "step": 34796 }, { "epoch": 3.5, "learning_rate": 1.992958863672423e-06, "loss": 0.2394, "step": 34797 }, { "epoch": 3.5, "learning_rate": 1.992101240504863e-06, "loss": 0.087, "step": 34798 }, { "epoch": 3.5, "learning_rate": 1.99124379556892e-06, "loss": 0.2717, "step": 34799 }, { "epoch": 3.5, "learning_rate": 1.990386528870045e-06, "loss": 0.1455, "step": 34800 }, { "epoch": 3.5, "learning_rate": 1.9895294404136975e-06, "loss": 0.1541, "step": 34801 }, { "epoch": 3.5, "learning_rate": 1.9886725302053197e-06, "loss": 0.1637, "step": 34802 }, { "epoch": 3.5, "learning_rate": 1.9878157982503874e-06, "loss": 0.152, "step": 34803 }, { "epoch": 3.5, "learning_rate": 1.986959244554333e-06, "loss": 0.3557, "step": 34804 }, { "epoch": 3.5, "learning_rate": 1.986102869122608e-06, "loss": 0.5667, "step": 34805 }, { "epoch": 3.5, "learning_rate": 1.9852466719606722e-06, "loss": 0.4116, "step": 34806 }, { "epoch": 3.5, "learning_rate": 1.9843906530739742e-06, "loss": 0.1267, "step": 34807 }, { "epoch": 3.5, "learning_rate": 1.9835348124679465e-06, "loss": 0.2413, "step": 34808 }, { "epoch": 3.5, "learning_rate": 1.982679150148048e-06, "loss": 0.3857, "step": 34809 }, { "epoch": 3.5, "learning_rate": 1.981823666119724e-06, "loss": 0.1108, "step": 34810 }, { "epoch": 3.5, "learning_rate": 1.9809683603884143e-06, "loss": 0.3867, "step": 34811 }, { "epoch": 3.5, "learning_rate": 1.980113232959564e-06, "loss": 0.2784, "step": 34812 }, { "epoch": 3.5, "learning_rate": 1.979258283838612e-06, "loss": 0.0931, "step": 34813 }, { "epoch": 3.5, "learning_rate": 1.978403513031001e-06, "loss": 0.1523, "step": 34814 }, { "epoch": 3.5, "learning_rate": 1.977548920542173e-06, "loss": 0.3483, "step": 34815 }, { "epoch": 3.5, "learning_rate": 1.9766945063775597e-06, "loss": 0.4475, "step": 34816 }, { "epoch": 3.5, "learning_rate": 1.975840270542605e-06, "loss": 0.0803, "step": 34817 }, { "epoch": 3.5, "learning_rate": 1.97498621304274e-06, "loss": 0.3822, "step": 34818 }, { "epoch": 3.5, "learning_rate": 1.9741323338834005e-06, "loss": 0.3299, "step": 34819 }, { "epoch": 3.5, "learning_rate": 1.9732786330700193e-06, "loss": 0.3387, "step": 34820 }, { "epoch": 3.5, "learning_rate": 1.972425110608038e-06, "loss": 0.3792, "step": 34821 }, { "epoch": 3.5, "learning_rate": 1.9715717665028764e-06, "loss": 0.0892, "step": 34822 }, { "epoch": 3.5, "learning_rate": 1.9707186007599666e-06, "loss": 0.2831, "step": 34823 }, { "epoch": 3.5, "learning_rate": 1.9698656133847436e-06, "loss": 0.1993, "step": 34824 }, { "epoch": 3.5, "learning_rate": 1.9690128043826374e-06, "loss": 0.3688, "step": 34825 }, { "epoch": 3.5, "learning_rate": 1.9681601737590625e-06, "loss": 0.2626, "step": 34826 }, { "epoch": 3.5, "learning_rate": 1.967307721519456e-06, "loss": 0.1949, "step": 34827 }, { "epoch": 3.5, "learning_rate": 1.9664554476692386e-06, "loss": 0.1043, "step": 34828 }, { "epoch": 3.5, "learning_rate": 1.965603352213834e-06, "loss": 0.0686, "step": 34829 }, { "epoch": 3.5, "learning_rate": 1.964751435158667e-06, "loss": 0.2158, "step": 34830 }, { "epoch": 3.5, "learning_rate": 1.9638996965091504e-06, "loss": 0.3391, "step": 34831 }, { "epoch": 3.5, "learning_rate": 1.9630481362707196e-06, "loss": 0.1914, "step": 34832 }, { "epoch": 3.5, "learning_rate": 1.962196754448781e-06, "loss": 0.0626, "step": 34833 }, { "epoch": 3.5, "learning_rate": 1.961345551048752e-06, "loss": 0.1942, "step": 34834 }, { "epoch": 3.5, "learning_rate": 1.960494526076063e-06, "loss": 0.0177, "step": 34835 }, { "epoch": 3.5, "learning_rate": 1.959643679536113e-06, "loss": 0.2136, "step": 34836 }, { "epoch": 3.5, "learning_rate": 1.95879301143432e-06, "loss": 0.7673, "step": 34837 }, { "epoch": 3.5, "learning_rate": 1.957942521776107e-06, "loss": 0.1975, "step": 34838 }, { "epoch": 3.5, "learning_rate": 1.9570922105668833e-06, "loss": 0.025, "step": 34839 }, { "epoch": 3.5, "learning_rate": 1.956242077812047e-06, "loss": 0.4008, "step": 34840 }, { "epoch": 3.5, "learning_rate": 1.955392123517021e-06, "loss": 0.0519, "step": 34841 }, { "epoch": 3.5, "learning_rate": 1.954542347687215e-06, "loss": 0.433, "step": 34842 }, { "epoch": 3.5, "learning_rate": 1.95369275032803e-06, "loss": 0.079, "step": 34843 }, { "epoch": 3.5, "learning_rate": 1.9528433314448756e-06, "loss": 0.1868, "step": 34844 }, { "epoch": 3.5, "learning_rate": 1.9519940910431546e-06, "loss": 0.0719, "step": 34845 }, { "epoch": 3.5, "learning_rate": 1.951145029128275e-06, "loss": 0.1582, "step": 34846 }, { "epoch": 3.5, "learning_rate": 1.9502961457056366e-06, "loss": 0.161, "step": 34847 }, { "epoch": 3.5, "learning_rate": 1.949447440780645e-06, "loss": 0.0576, "step": 34848 }, { "epoch": 3.5, "learning_rate": 1.948598914358699e-06, "loss": 0.146, "step": 34849 }, { "epoch": 3.5, "learning_rate": 1.9477505664451944e-06, "loss": 0.0696, "step": 34850 }, { "epoch": 3.5, "learning_rate": 1.946902397045537e-06, "loss": 0.0793, "step": 34851 }, { "epoch": 3.5, "learning_rate": 1.9460544061651154e-06, "loss": 0.0891, "step": 34852 }, { "epoch": 3.5, "learning_rate": 1.9452065938093422e-06, "loss": 0.1192, "step": 34853 }, { "epoch": 3.5, "learning_rate": 1.9443589599835965e-06, "loss": 0.4983, "step": 34854 }, { "epoch": 3.5, "learning_rate": 1.9435115046932705e-06, "loss": 0.0352, "step": 34855 }, { "epoch": 3.5, "learning_rate": 1.94266422794377e-06, "loss": 0.1904, "step": 34856 }, { "epoch": 3.5, "learning_rate": 1.941817129740481e-06, "loss": 0.2503, "step": 34857 }, { "epoch": 3.5, "learning_rate": 1.9409702100887918e-06, "loss": 0.4731, "step": 34858 }, { "epoch": 3.5, "learning_rate": 1.9401234689940985e-06, "loss": 0.0324, "step": 34859 }, { "epoch": 3.5, "learning_rate": 1.93927690646178e-06, "loss": 0.0463, "step": 34860 }, { "epoch": 3.5, "learning_rate": 1.938430522497232e-06, "loss": 0.1776, "step": 34861 }, { "epoch": 3.5, "learning_rate": 1.937584317105834e-06, "loss": 0.1902, "step": 34862 }, { "epoch": 3.5, "learning_rate": 1.9367382902929707e-06, "loss": 0.468, "step": 34863 }, { "epoch": 3.5, "learning_rate": 1.935892442064039e-06, "loss": 0.0976, "step": 34864 }, { "epoch": 3.5, "learning_rate": 1.935046772424407e-06, "loss": 0.3986, "step": 34865 }, { "epoch": 3.5, "learning_rate": 1.9342012813794576e-06, "loss": 0.1614, "step": 34866 }, { "epoch": 3.5, "learning_rate": 1.9333559689345826e-06, "loss": 0.0406, "step": 34867 }, { "epoch": 3.5, "learning_rate": 1.9325108350951482e-06, "loss": 0.1928, "step": 34868 }, { "epoch": 3.5, "learning_rate": 1.93166587986653e-06, "loss": 0.0489, "step": 34869 }, { "epoch": 3.5, "learning_rate": 1.9308211032541203e-06, "loss": 0.1409, "step": 34870 }, { "epoch": 3.5, "learning_rate": 1.9299765052632912e-06, "loss": 0.1004, "step": 34871 }, { "epoch": 3.5, "learning_rate": 1.9291320858994055e-06, "loss": 0.6546, "step": 34872 }, { "epoch": 3.5, "learning_rate": 1.9282878451678456e-06, "loss": 0.1736, "step": 34873 }, { "epoch": 3.5, "learning_rate": 1.9274437830739867e-06, "loss": 0.0316, "step": 34874 }, { "epoch": 3.5, "learning_rate": 1.9265998996231914e-06, "loss": 0.1778, "step": 34875 }, { "epoch": 3.5, "learning_rate": 1.9257561948208393e-06, "loss": 0.0646, "step": 34876 }, { "epoch": 3.5, "learning_rate": 1.924912668672292e-06, "loss": 0.2796, "step": 34877 }, { "epoch": 3.5, "learning_rate": 1.9240693211829218e-06, "loss": 0.4324, "step": 34878 }, { "epoch": 3.5, "learning_rate": 1.9232261523580917e-06, "loss": 0.1986, "step": 34879 }, { "epoch": 3.5, "learning_rate": 1.9223831622031707e-06, "loss": 0.0433, "step": 34880 }, { "epoch": 3.5, "learning_rate": 1.921540350723524e-06, "loss": 0.1518, "step": 34881 }, { "epoch": 3.51, "learning_rate": 1.9206977179245112e-06, "loss": 0.0689, "step": 34882 }, { "epoch": 3.51, "learning_rate": 1.9198552638114907e-06, "loss": 0.1414, "step": 34883 }, { "epoch": 3.51, "learning_rate": 1.9190129883898324e-06, "loss": 0.2081, "step": 34884 }, { "epoch": 3.51, "learning_rate": 1.918170891664901e-06, "loss": 0.1218, "step": 34885 }, { "epoch": 3.51, "learning_rate": 1.9173289736420364e-06, "loss": 0.3878, "step": 34886 }, { "epoch": 3.51, "learning_rate": 1.9164872343266102e-06, "loss": 0.0899, "step": 34887 }, { "epoch": 3.51, "learning_rate": 1.9156456737239793e-06, "loss": 0.3292, "step": 34888 }, { "epoch": 3.51, "learning_rate": 1.9148042918394913e-06, "loss": 0.0751, "step": 34889 }, { "epoch": 3.51, "learning_rate": 1.9139630886785066e-06, "loss": 0.0618, "step": 34890 }, { "epoch": 3.51, "learning_rate": 1.9131220642463764e-06, "loss": 0.1388, "step": 34891 }, { "epoch": 3.51, "learning_rate": 1.912281218548454e-06, "loss": 0.1424, "step": 34892 }, { "epoch": 3.51, "learning_rate": 1.911440551590088e-06, "loss": 0.449, "step": 34893 }, { "epoch": 3.51, "learning_rate": 1.9106000633766276e-06, "loss": 0.159, "step": 34894 }, { "epoch": 3.51, "learning_rate": 1.9097597539134182e-06, "loss": 0.122, "step": 34895 }, { "epoch": 3.51, "learning_rate": 1.9089196232058224e-06, "loss": 0.0522, "step": 34896 }, { "epoch": 3.51, "learning_rate": 1.9080796712591697e-06, "loss": 0.0679, "step": 34897 }, { "epoch": 3.51, "learning_rate": 1.9072398980788052e-06, "loss": 0.0661, "step": 34898 }, { "epoch": 3.51, "learning_rate": 1.9064003036700916e-06, "loss": 0.4925, "step": 34899 }, { "epoch": 3.51, "learning_rate": 1.9055608880383545e-06, "loss": 0.1185, "step": 34900 }, { "epoch": 3.51, "learning_rate": 1.904721651188933e-06, "loss": 0.0961, "step": 34901 }, { "epoch": 3.51, "learning_rate": 1.9038825931271796e-06, "loss": 0.2829, "step": 34902 }, { "epoch": 3.51, "learning_rate": 1.9030437138584367e-06, "loss": 0.1287, "step": 34903 }, { "epoch": 3.51, "learning_rate": 1.902205013388023e-06, "loss": 0.1482, "step": 34904 }, { "epoch": 3.51, "learning_rate": 1.9013664917212913e-06, "loss": 0.0825, "step": 34905 }, { "epoch": 3.51, "learning_rate": 1.9005281488635773e-06, "loss": 0.4204, "step": 34906 }, { "epoch": 3.51, "learning_rate": 1.8996899848202098e-06, "loss": 0.4338, "step": 34907 }, { "epoch": 3.51, "learning_rate": 1.898851999596528e-06, "loss": 0.1208, "step": 34908 }, { "epoch": 3.51, "learning_rate": 1.8980141931978578e-06, "loss": 0.0021, "step": 34909 }, { "epoch": 3.51, "learning_rate": 1.8971765656295382e-06, "loss": 1.1042, "step": 34910 }, { "epoch": 3.51, "learning_rate": 1.8963391168968946e-06, "loss": 0.1489, "step": 34911 }, { "epoch": 3.51, "learning_rate": 1.8955018470052532e-06, "loss": 0.3377, "step": 34912 }, { "epoch": 3.51, "learning_rate": 1.8946647559599594e-06, "loss": 0.2406, "step": 34913 }, { "epoch": 3.51, "learning_rate": 1.8938278437663193e-06, "loss": 0.5161, "step": 34914 }, { "epoch": 3.51, "learning_rate": 1.8929911104296616e-06, "loss": 0.281, "step": 34915 }, { "epoch": 3.51, "learning_rate": 1.8921545559553188e-06, "loss": 0.0475, "step": 34916 }, { "epoch": 3.51, "learning_rate": 1.8913181803486202e-06, "loss": 0.1739, "step": 34917 }, { "epoch": 3.51, "learning_rate": 1.890481983614868e-06, "loss": 0.1169, "step": 34918 }, { "epoch": 3.51, "learning_rate": 1.889645965759401e-06, "loss": 0.3792, "step": 34919 }, { "epoch": 3.51, "learning_rate": 1.8888101267875324e-06, "loss": 0.1303, "step": 34920 }, { "epoch": 3.51, "learning_rate": 1.887974466704584e-06, "loss": 0.3485, "step": 34921 }, { "epoch": 3.51, "learning_rate": 1.8871389855158716e-06, "loss": 0.084, "step": 34922 }, { "epoch": 3.51, "learning_rate": 1.8863036832267077e-06, "loss": 0.1667, "step": 34923 }, { "epoch": 3.51, "learning_rate": 1.885468559842418e-06, "loss": 0.3566, "step": 34924 }, { "epoch": 3.51, "learning_rate": 1.8846336153683086e-06, "loss": 0.1049, "step": 34925 }, { "epoch": 3.51, "learning_rate": 1.8837988498096947e-06, "loss": 0.1945, "step": 34926 }, { "epoch": 3.51, "learning_rate": 1.8829642631718823e-06, "loss": 0.2202, "step": 34927 }, { "epoch": 3.51, "learning_rate": 1.8821298554602008e-06, "loss": 0.2682, "step": 34928 }, { "epoch": 3.51, "learning_rate": 1.8812956266799452e-06, "loss": 0.3685, "step": 34929 }, { "epoch": 3.51, "learning_rate": 1.8804615768364186e-06, "loss": 0.1208, "step": 34930 }, { "epoch": 3.51, "learning_rate": 1.8796277059349499e-06, "loss": 0.0855, "step": 34931 }, { "epoch": 3.51, "learning_rate": 1.878794013980828e-06, "loss": 0.1661, "step": 34932 }, { "epoch": 3.51, "learning_rate": 1.8779605009793587e-06, "loss": 0.2501, "step": 34933 }, { "epoch": 3.51, "learning_rate": 1.8771271669358548e-06, "loss": 0.1947, "step": 34934 }, { "epoch": 3.51, "learning_rate": 1.8762940118556182e-06, "loss": 0.0704, "step": 34935 }, { "epoch": 3.51, "learning_rate": 1.8754610357439383e-06, "loss": 0.2354, "step": 34936 }, { "epoch": 3.51, "learning_rate": 1.8746282386061308e-06, "loss": 0.1978, "step": 34937 }, { "epoch": 3.51, "learning_rate": 1.8737956204474915e-06, "loss": 0.1525, "step": 34938 }, { "epoch": 3.51, "learning_rate": 1.8729631812733162e-06, "loss": 0.0626, "step": 34939 }, { "epoch": 3.51, "learning_rate": 1.8721309210889004e-06, "loss": 0.1288, "step": 34940 }, { "epoch": 3.51, "learning_rate": 1.8712988398995434e-06, "loss": 0.5621, "step": 34941 }, { "epoch": 3.51, "learning_rate": 1.8704669377105443e-06, "loss": 0.0659, "step": 34942 }, { "epoch": 3.51, "learning_rate": 1.8696352145271921e-06, "loss": 0.3686, "step": 34943 }, { "epoch": 3.51, "learning_rate": 1.8688036703547727e-06, "loss": 0.1128, "step": 34944 }, { "epoch": 3.51, "learning_rate": 1.8679723051985953e-06, "loss": 0.0329, "step": 34945 }, { "epoch": 3.51, "learning_rate": 1.8671411190639354e-06, "loss": 0.0242, "step": 34946 }, { "epoch": 3.51, "learning_rate": 1.866310111956082e-06, "loss": 0.3399, "step": 34947 }, { "epoch": 3.51, "learning_rate": 1.8654792838803314e-06, "loss": 0.4667, "step": 34948 }, { "epoch": 3.51, "learning_rate": 1.8646486348419757e-06, "loss": 0.1049, "step": 34949 }, { "epoch": 3.51, "learning_rate": 1.8638181648462838e-06, "loss": 0.0727, "step": 34950 }, { "epoch": 3.51, "learning_rate": 1.8629878738985518e-06, "loss": 0.1126, "step": 34951 }, { "epoch": 3.51, "learning_rate": 1.862157762004062e-06, "loss": 0.0964, "step": 34952 }, { "epoch": 3.51, "learning_rate": 1.8613278291680968e-06, "loss": 0.1142, "step": 34953 }, { "epoch": 3.51, "learning_rate": 1.8604980753959355e-06, "loss": 0.0551, "step": 34954 }, { "epoch": 3.51, "learning_rate": 1.8596685006928605e-06, "loss": 0.5276, "step": 34955 }, { "epoch": 3.51, "learning_rate": 1.8588391050641473e-06, "loss": 0.0706, "step": 34956 }, { "epoch": 3.51, "learning_rate": 1.8580098885150821e-06, "loss": 0.0146, "step": 34957 }, { "epoch": 3.51, "learning_rate": 1.8571808510509303e-06, "loss": 0.2038, "step": 34958 }, { "epoch": 3.51, "learning_rate": 1.8563519926769712e-06, "loss": 0.2146, "step": 34959 }, { "epoch": 3.51, "learning_rate": 1.8555233133984939e-06, "loss": 0.0896, "step": 34960 }, { "epoch": 3.51, "learning_rate": 1.854694813220751e-06, "loss": 0.0802, "step": 34961 }, { "epoch": 3.51, "learning_rate": 1.8538664921490212e-06, "loss": 0.0391, "step": 34962 }, { "epoch": 3.51, "learning_rate": 1.853038350188584e-06, "loss": 0.1426, "step": 34963 }, { "epoch": 3.51, "learning_rate": 1.8522103873446949e-06, "loss": 0.2562, "step": 34964 }, { "epoch": 3.51, "learning_rate": 1.8513826036226366e-06, "loss": 0.349, "step": 34965 }, { "epoch": 3.51, "learning_rate": 1.8505549990276715e-06, "loss": 0.3636, "step": 34966 }, { "epoch": 3.51, "learning_rate": 1.8497275735650654e-06, "loss": 0.1762, "step": 34967 }, { "epoch": 3.51, "learning_rate": 1.8489003272400839e-06, "loss": 0.1948, "step": 34968 }, { "epoch": 3.51, "learning_rate": 1.8480732600579897e-06, "loss": 0.0571, "step": 34969 }, { "epoch": 3.51, "learning_rate": 1.8472463720240519e-06, "loss": 0.27, "step": 34970 }, { "epoch": 3.51, "learning_rate": 1.846419663143526e-06, "loss": 0.1913, "step": 34971 }, { "epoch": 3.51, "learning_rate": 1.8455931334216747e-06, "loss": 0.0671, "step": 34972 }, { "epoch": 3.51, "learning_rate": 1.8447667828637537e-06, "loss": 0.0399, "step": 34973 }, { "epoch": 3.51, "learning_rate": 1.8439406114750356e-06, "loss": 0.2769, "step": 34974 }, { "epoch": 3.51, "learning_rate": 1.8431146192607629e-06, "loss": 0.1889, "step": 34975 }, { "epoch": 3.51, "learning_rate": 1.8422888062261944e-06, "loss": 0.1031, "step": 34976 }, { "epoch": 3.51, "learning_rate": 1.8414631723765962e-06, "loss": 0.1434, "step": 34977 }, { "epoch": 3.51, "learning_rate": 1.8406377177172107e-06, "loss": 0.2244, "step": 34978 }, { "epoch": 3.51, "learning_rate": 1.839812442253287e-06, "loss": 0.6867, "step": 34979 }, { "epoch": 3.52, "learning_rate": 1.8389873459900907e-06, "loss": 0.1287, "step": 34980 }, { "epoch": 3.52, "learning_rate": 1.8381624289328714e-06, "loss": 0.268, "step": 34981 }, { "epoch": 3.52, "learning_rate": 1.8373376910868645e-06, "loss": 0.5053, "step": 34982 }, { "epoch": 3.52, "learning_rate": 1.8365131324573293e-06, "loss": 0.0883, "step": 34983 }, { "epoch": 3.52, "learning_rate": 1.8356887530495114e-06, "loss": 0.2839, "step": 34984 }, { "epoch": 3.52, "learning_rate": 1.834864552868657e-06, "loss": 0.0522, "step": 34985 }, { "epoch": 3.52, "learning_rate": 1.8340405319200081e-06, "loss": 0.1082, "step": 34986 }, { "epoch": 3.52, "learning_rate": 1.833216690208811e-06, "loss": 0.1278, "step": 34987 }, { "epoch": 3.52, "learning_rate": 1.8323930277403079e-06, "loss": 0.0889, "step": 34988 }, { "epoch": 3.52, "learning_rate": 1.8315695445197411e-06, "loss": 0.0665, "step": 34989 }, { "epoch": 3.52, "learning_rate": 1.83074624055235e-06, "loss": 0.1763, "step": 34990 }, { "epoch": 3.52, "learning_rate": 1.8299231158433671e-06, "loss": 0.1694, "step": 34991 }, { "epoch": 3.52, "learning_rate": 1.829100170398048e-06, "loss": 0.2065, "step": 34992 }, { "epoch": 3.52, "learning_rate": 1.828277404221609e-06, "loss": 0.0808, "step": 34993 }, { "epoch": 3.52, "learning_rate": 1.8274548173192984e-06, "loss": 0.223, "step": 34994 }, { "epoch": 3.52, "learning_rate": 1.826632409696356e-06, "loss": 0.3856, "step": 34995 }, { "epoch": 3.52, "learning_rate": 1.8258101813579942e-06, "loss": 0.1455, "step": 34996 }, { "epoch": 3.52, "learning_rate": 1.8249881323094652e-06, "loss": 0.0693, "step": 34997 }, { "epoch": 3.52, "learning_rate": 1.8241662625559952e-06, "loss": 0.0967, "step": 34998 }, { "epoch": 3.52, "learning_rate": 1.8233445721028098e-06, "loss": 0.1301, "step": 34999 }, { "epoch": 3.52, "learning_rate": 1.8225230609551414e-06, "loss": 0.1877, "step": 35000 }, { "epoch": 3.52, "learning_rate": 1.8217017291182158e-06, "loss": 0.0435, "step": 35001 }, { "epoch": 3.52, "learning_rate": 1.8208805765972625e-06, "loss": 0.0851, "step": 35002 }, { "epoch": 3.52, "learning_rate": 1.8200596033975037e-06, "loss": 0.0224, "step": 35003 }, { "epoch": 3.52, "learning_rate": 1.8192388095241651e-06, "loss": 0.1816, "step": 35004 }, { "epoch": 3.52, "learning_rate": 1.8184181949824697e-06, "loss": 0.0338, "step": 35005 }, { "epoch": 3.52, "learning_rate": 1.8175977597776461e-06, "loss": 0.4826, "step": 35006 }, { "epoch": 3.52, "learning_rate": 1.8167775039149037e-06, "loss": 0.0982, "step": 35007 }, { "epoch": 3.52, "learning_rate": 1.815957427399465e-06, "loss": 0.1696, "step": 35008 }, { "epoch": 3.52, "learning_rate": 1.8151375302365591e-06, "loss": 0.2268, "step": 35009 }, { "epoch": 3.52, "learning_rate": 1.814317812431392e-06, "loss": 0.1521, "step": 35010 }, { "epoch": 3.52, "learning_rate": 1.8134982739891825e-06, "loss": 0.212, "step": 35011 }, { "epoch": 3.52, "learning_rate": 1.8126789149151469e-06, "loss": 0.3437, "step": 35012 }, { "epoch": 3.52, "learning_rate": 1.8118597352145072e-06, "loss": 0.4006, "step": 35013 }, { "epoch": 3.52, "learning_rate": 1.8110407348924596e-06, "loss": 0.1697, "step": 35014 }, { "epoch": 3.52, "learning_rate": 1.8102219139542299e-06, "loss": 0.4059, "step": 35015 }, { "epoch": 3.52, "learning_rate": 1.8094032724050202e-06, "loss": 0.1067, "step": 35016 }, { "epoch": 3.52, "learning_rate": 1.8085848102500468e-06, "loss": 0.1178, "step": 35017 }, { "epoch": 3.52, "learning_rate": 1.8077665274945155e-06, "loss": 0.1286, "step": 35018 }, { "epoch": 3.52, "learning_rate": 1.8069484241436285e-06, "loss": 0.0538, "step": 35019 }, { "epoch": 3.52, "learning_rate": 1.8061305002026053e-06, "loss": 0.0511, "step": 35020 }, { "epoch": 3.52, "learning_rate": 1.805312755676638e-06, "loss": 0.3678, "step": 35021 }, { "epoch": 3.52, "learning_rate": 1.8044951905709294e-06, "loss": 0.082, "step": 35022 }, { "epoch": 3.52, "learning_rate": 1.8036778048906921e-06, "loss": 0.1673, "step": 35023 }, { "epoch": 3.52, "learning_rate": 1.8028605986411285e-06, "loss": 0.353, "step": 35024 }, { "epoch": 3.52, "learning_rate": 1.8020435718274208e-06, "loss": 0.0487, "step": 35025 }, { "epoch": 3.52, "learning_rate": 1.8012267244547887e-06, "loss": 0.3287, "step": 35026 }, { "epoch": 3.52, "learning_rate": 1.8004100565284243e-06, "loss": 0.2724, "step": 35027 }, { "epoch": 3.52, "learning_rate": 1.7995935680535169e-06, "loss": 0.0646, "step": 35028 }, { "epoch": 3.52, "learning_rate": 1.7987772590352692e-06, "loss": 0.2167, "step": 35029 }, { "epoch": 3.52, "learning_rate": 1.7979611294788768e-06, "loss": 0.0799, "step": 35030 }, { "epoch": 3.52, "learning_rate": 1.7971451793895289e-06, "loss": 0.2246, "step": 35031 }, { "epoch": 3.52, "learning_rate": 1.7963294087724214e-06, "loss": 0.1465, "step": 35032 }, { "epoch": 3.52, "learning_rate": 1.7955138176327468e-06, "loss": 0.0903, "step": 35033 }, { "epoch": 3.52, "learning_rate": 1.794698405975691e-06, "loss": 0.4487, "step": 35034 }, { "epoch": 3.52, "learning_rate": 1.7938831738064433e-06, "loss": 0.2307, "step": 35035 }, { "epoch": 3.52, "learning_rate": 1.7930681211301925e-06, "loss": 0.0487, "step": 35036 }, { "epoch": 3.52, "learning_rate": 1.7922532479521249e-06, "loss": 0.0899, "step": 35037 }, { "epoch": 3.52, "learning_rate": 1.7914385542774325e-06, "loss": 0.1874, "step": 35038 }, { "epoch": 3.52, "learning_rate": 1.7906240401112917e-06, "loss": 0.2396, "step": 35039 }, { "epoch": 3.52, "learning_rate": 1.7898097054588813e-06, "loss": 0.0893, "step": 35040 }, { "epoch": 3.52, "learning_rate": 1.7889955503254007e-06, "loss": 0.0645, "step": 35041 }, { "epoch": 3.52, "learning_rate": 1.7881815747160158e-06, "loss": 0.1552, "step": 35042 }, { "epoch": 3.52, "learning_rate": 1.7873677786359056e-06, "loss": 0.1658, "step": 35043 }, { "epoch": 3.52, "learning_rate": 1.7865541620902593e-06, "loss": 0.2056, "step": 35044 }, { "epoch": 3.52, "learning_rate": 1.7857407250842494e-06, "loss": 0.222, "step": 35045 }, { "epoch": 3.52, "learning_rate": 1.784927467623052e-06, "loss": 0.1983, "step": 35046 }, { "epoch": 3.52, "learning_rate": 1.7841143897118429e-06, "loss": 0.0705, "step": 35047 }, { "epoch": 3.52, "learning_rate": 1.7833014913557976e-06, "loss": 0.4051, "step": 35048 }, { "epoch": 3.52, "learning_rate": 1.7824887725600857e-06, "loss": 0.0291, "step": 35049 }, { "epoch": 3.52, "learning_rate": 1.7816762333298796e-06, "loss": 0.155, "step": 35050 }, { "epoch": 3.52, "learning_rate": 1.7808638736703485e-06, "loss": 0.1881, "step": 35051 }, { "epoch": 3.52, "learning_rate": 1.780051693586675e-06, "loss": 0.5185, "step": 35052 }, { "epoch": 3.52, "learning_rate": 1.7792396930840116e-06, "loss": 0.0287, "step": 35053 }, { "epoch": 3.52, "learning_rate": 1.778427872167524e-06, "loss": 0.0117, "step": 35054 }, { "epoch": 3.52, "learning_rate": 1.7776162308423916e-06, "loss": 0.179, "step": 35055 }, { "epoch": 3.52, "learning_rate": 1.7768047691137768e-06, "loss": 0.2296, "step": 35056 }, { "epoch": 3.52, "learning_rate": 1.7759934869868322e-06, "loss": 0.0866, "step": 35057 }, { "epoch": 3.52, "learning_rate": 1.7751823844667336e-06, "loss": 0.3157, "step": 35058 }, { "epoch": 3.52, "learning_rate": 1.7743714615586403e-06, "loss": 0.056, "step": 35059 }, { "epoch": 3.52, "learning_rate": 1.7735607182676983e-06, "loss": 0.1144, "step": 35060 }, { "epoch": 3.52, "learning_rate": 1.7727501545990865e-06, "loss": 0.1237, "step": 35061 }, { "epoch": 3.52, "learning_rate": 1.771939770557951e-06, "loss": 0.0592, "step": 35062 }, { "epoch": 3.52, "learning_rate": 1.7711295661494543e-06, "loss": 0.1817, "step": 35063 }, { "epoch": 3.52, "learning_rate": 1.7703195413787487e-06, "loss": 0.0968, "step": 35064 }, { "epoch": 3.52, "learning_rate": 1.7695096962509905e-06, "loss": 0.6307, "step": 35065 }, { "epoch": 3.52, "learning_rate": 1.7687000307713353e-06, "loss": 0.1234, "step": 35066 }, { "epoch": 3.52, "learning_rate": 1.7678905449449322e-06, "loss": 0.3124, "step": 35067 }, { "epoch": 3.52, "learning_rate": 1.767081238776934e-06, "loss": 0.1539, "step": 35068 }, { "epoch": 3.52, "learning_rate": 1.7662721122724835e-06, "loss": 0.3525, "step": 35069 }, { "epoch": 3.52, "learning_rate": 1.7654631654367492e-06, "loss": 0.4129, "step": 35070 }, { "epoch": 3.52, "learning_rate": 1.764654398274861e-06, "loss": 0.4734, "step": 35071 }, { "epoch": 3.52, "learning_rate": 1.7638458107919643e-06, "loss": 0.1764, "step": 35072 }, { "epoch": 3.52, "learning_rate": 1.7630374029932218e-06, "loss": 0.0344, "step": 35073 }, { "epoch": 3.52, "learning_rate": 1.7622291748837594e-06, "loss": 0.2298, "step": 35074 }, { "epoch": 3.52, "learning_rate": 1.7614211264687296e-06, "loss": 0.0514, "step": 35075 }, { "epoch": 3.52, "learning_rate": 1.7606132577532751e-06, "loss": 0.4199, "step": 35076 }, { "epoch": 3.52, "learning_rate": 1.759805568742535e-06, "loss": 0.0349, "step": 35077 }, { "epoch": 3.52, "learning_rate": 1.7589980594416488e-06, "loss": 0.0403, "step": 35078 }, { "epoch": 3.53, "learning_rate": 1.7581907298557588e-06, "loss": 0.2425, "step": 35079 }, { "epoch": 3.53, "learning_rate": 1.757383579989994e-06, "loss": 0.228, "step": 35080 }, { "epoch": 3.53, "learning_rate": 1.7565766098495006e-06, "loss": 0.2036, "step": 35081 }, { "epoch": 3.53, "learning_rate": 1.7557698194394077e-06, "loss": 0.1902, "step": 35082 }, { "epoch": 3.53, "learning_rate": 1.7549632087648448e-06, "loss": 0.1124, "step": 35083 }, { "epoch": 3.53, "learning_rate": 1.7541567778309608e-06, "loss": 0.2937, "step": 35084 }, { "epoch": 3.53, "learning_rate": 1.753350526642875e-06, "loss": 0.6446, "step": 35085 }, { "epoch": 3.53, "learning_rate": 1.7525444552057136e-06, "loss": 0.4036, "step": 35086 }, { "epoch": 3.53, "learning_rate": 1.7517385635246218e-06, "loss": 0.1819, "step": 35087 }, { "epoch": 3.53, "learning_rate": 1.7509328516047196e-06, "loss": 0.2205, "step": 35088 }, { "epoch": 3.53, "learning_rate": 1.7501273194511292e-06, "loss": 0.5697, "step": 35089 }, { "epoch": 3.53, "learning_rate": 1.749321967068983e-06, "loss": 0.0397, "step": 35090 }, { "epoch": 3.53, "learning_rate": 1.7485167944634107e-06, "loss": 0.28, "step": 35091 }, { "epoch": 3.53, "learning_rate": 1.747711801639521e-06, "loss": 0.244, "step": 35092 }, { "epoch": 3.53, "learning_rate": 1.7469069886024502e-06, "loss": 0.0665, "step": 35093 }, { "epoch": 3.53, "learning_rate": 1.7461023553573175e-06, "loss": 0.0077, "step": 35094 }, { "epoch": 3.53, "learning_rate": 1.7452979019092386e-06, "loss": 0.1503, "step": 35095 }, { "epoch": 3.53, "learning_rate": 1.7444936282633361e-06, "loss": 0.2987, "step": 35096 }, { "epoch": 3.53, "learning_rate": 1.7436895344247261e-06, "loss": 0.1506, "step": 35097 }, { "epoch": 3.53, "learning_rate": 1.7428856203985277e-06, "loss": 0.1187, "step": 35098 }, { "epoch": 3.53, "learning_rate": 1.7420818861898535e-06, "loss": 0.0843, "step": 35099 }, { "epoch": 3.53, "learning_rate": 1.7412783318038195e-06, "loss": 0.066, "step": 35100 }, { "epoch": 3.53, "learning_rate": 1.7404749572455413e-06, "loss": 0.0496, "step": 35101 }, { "epoch": 3.53, "learning_rate": 1.7396717625201353e-06, "loss": 0.3205, "step": 35102 }, { "epoch": 3.53, "learning_rate": 1.7388687476327004e-06, "loss": 0.2923, "step": 35103 }, { "epoch": 3.53, "learning_rate": 1.7380659125883558e-06, "loss": 0.2968, "step": 35104 }, { "epoch": 3.53, "learning_rate": 1.7372632573922143e-06, "loss": 0.1027, "step": 35105 }, { "epoch": 3.53, "learning_rate": 1.7364607820493683e-06, "loss": 0.3773, "step": 35106 }, { "epoch": 3.53, "learning_rate": 1.7356584865649372e-06, "loss": 0.0394, "step": 35107 }, { "epoch": 3.53, "learning_rate": 1.7348563709440235e-06, "loss": 0.0644, "step": 35108 }, { "epoch": 3.53, "learning_rate": 1.7340544351917331e-06, "loss": 0.4633, "step": 35109 }, { "epoch": 3.53, "learning_rate": 1.7332526793131654e-06, "loss": 0.4505, "step": 35110 }, { "epoch": 3.53, "learning_rate": 1.732451103313426e-06, "loss": 0.0984, "step": 35111 }, { "epoch": 3.53, "learning_rate": 1.7316497071976146e-06, "loss": 0.2153, "step": 35112 }, { "epoch": 3.53, "learning_rate": 1.7308484909708268e-06, "loss": 0.0596, "step": 35113 }, { "epoch": 3.53, "learning_rate": 1.7300474546381683e-06, "loss": 0.0778, "step": 35114 }, { "epoch": 3.53, "learning_rate": 1.7292465982047289e-06, "loss": 0.2261, "step": 35115 }, { "epoch": 3.53, "learning_rate": 1.7284459216756177e-06, "loss": 0.2804, "step": 35116 }, { "epoch": 3.53, "learning_rate": 1.727645425055917e-06, "loss": 0.0725, "step": 35117 }, { "epoch": 3.53, "learning_rate": 1.726845108350723e-06, "loss": 0.0833, "step": 35118 }, { "epoch": 3.53, "learning_rate": 1.7260449715651315e-06, "loss": 0.1167, "step": 35119 }, { "epoch": 3.53, "learning_rate": 1.725245014704242e-06, "loss": 0.2836, "step": 35120 }, { "epoch": 3.53, "learning_rate": 1.7244452377731268e-06, "loss": 0.1827, "step": 35121 }, { "epoch": 3.53, "learning_rate": 1.7236456407768885e-06, "loss": 0.0432, "step": 35122 }, { "epoch": 3.53, "learning_rate": 1.72284622372062e-06, "loss": 0.0725, "step": 35123 }, { "epoch": 3.53, "learning_rate": 1.7220469866093902e-06, "loss": 0.1944, "step": 35124 }, { "epoch": 3.53, "learning_rate": 1.7212479294482986e-06, "loss": 0.9906, "step": 35125 }, { "epoch": 3.53, "learning_rate": 1.720449052242431e-06, "loss": 0.2218, "step": 35126 }, { "epoch": 3.53, "learning_rate": 1.7196503549968667e-06, "loss": 0.0165, "step": 35127 }, { "epoch": 3.53, "learning_rate": 1.7188518377166884e-06, "loss": 0.2011, "step": 35128 }, { "epoch": 3.53, "learning_rate": 1.718053500406972e-06, "loss": 0.2812, "step": 35129 }, { "epoch": 3.53, "learning_rate": 1.7172553430728132e-06, "loss": 0.1652, "step": 35130 }, { "epoch": 3.53, "learning_rate": 1.7164573657192818e-06, "loss": 0.2143, "step": 35131 }, { "epoch": 3.53, "learning_rate": 1.7156595683514465e-06, "loss": 0.1869, "step": 35132 }, { "epoch": 3.53, "learning_rate": 1.7148619509744035e-06, "loss": 0.2405, "step": 35133 }, { "epoch": 3.53, "learning_rate": 1.7140645135932186e-06, "loss": 0.1432, "step": 35134 }, { "epoch": 3.53, "learning_rate": 1.7132672562129614e-06, "loss": 0.5818, "step": 35135 }, { "epoch": 3.53, "learning_rate": 1.712470178838711e-06, "loss": 0.1709, "step": 35136 }, { "epoch": 3.53, "learning_rate": 1.7116732814755465e-06, "loss": 0.0448, "step": 35137 }, { "epoch": 3.53, "learning_rate": 1.710876564128524e-06, "loss": 0.0421, "step": 35138 }, { "epoch": 3.53, "learning_rate": 1.7100800268027229e-06, "loss": 0.5682, "step": 35139 }, { "epoch": 3.53, "learning_rate": 1.7092836695032121e-06, "loss": 0.0419, "step": 35140 }, { "epoch": 3.53, "learning_rate": 1.708487492235058e-06, "loss": 0.0413, "step": 35141 }, { "epoch": 3.53, "learning_rate": 1.7076914950033262e-06, "loss": 0.3023, "step": 35142 }, { "epoch": 3.53, "learning_rate": 1.7068956778130828e-06, "loss": 0.0596, "step": 35143 }, { "epoch": 3.53, "learning_rate": 1.7061000406693904e-06, "loss": 0.0374, "step": 35144 }, { "epoch": 3.53, "learning_rate": 1.7053045835773152e-06, "loss": 0.206, "step": 35145 }, { "epoch": 3.53, "learning_rate": 1.7045093065419192e-06, "loss": 0.1358, "step": 35146 }, { "epoch": 3.53, "learning_rate": 1.7037142095682556e-06, "loss": 0.5012, "step": 35147 }, { "epoch": 3.53, "learning_rate": 1.7029192926614e-06, "loss": 0.1059, "step": 35148 }, { "epoch": 3.53, "learning_rate": 1.7021245558263954e-06, "loss": 0.3559, "step": 35149 }, { "epoch": 3.53, "learning_rate": 1.701329999068304e-06, "loss": 0.1063, "step": 35150 }, { "epoch": 3.53, "learning_rate": 1.7005356223921853e-06, "loss": 0.1127, "step": 35151 }, { "epoch": 3.53, "learning_rate": 1.6997414258030953e-06, "loss": 0.3796, "step": 35152 }, { "epoch": 3.53, "learning_rate": 1.6989474093060765e-06, "loss": 0.0431, "step": 35153 }, { "epoch": 3.53, "learning_rate": 1.6981535729061948e-06, "loss": 0.1691, "step": 35154 }, { "epoch": 3.53, "learning_rate": 1.6973599166084997e-06, "loss": 0.2497, "step": 35155 }, { "epoch": 3.53, "learning_rate": 1.696566440418037e-06, "loss": 0.1502, "step": 35156 }, { "epoch": 3.53, "learning_rate": 1.6957731443398594e-06, "loss": 0.1049, "step": 35157 }, { "epoch": 3.53, "learning_rate": 1.694980028379013e-06, "loss": 0.1704, "step": 35158 }, { "epoch": 3.53, "learning_rate": 1.6941870925405468e-06, "loss": 0.0763, "step": 35159 }, { "epoch": 3.53, "learning_rate": 1.6933943368295036e-06, "loss": 0.4685, "step": 35160 }, { "epoch": 3.53, "learning_rate": 1.6926017612509293e-06, "loss": 0.2992, "step": 35161 }, { "epoch": 3.53, "learning_rate": 1.6918093658098766e-06, "loss": 0.2982, "step": 35162 }, { "epoch": 3.53, "learning_rate": 1.6910171505113747e-06, "loss": 0.0474, "step": 35163 }, { "epoch": 3.53, "learning_rate": 1.6902251153604664e-06, "loss": 0.0547, "step": 35164 }, { "epoch": 3.53, "learning_rate": 1.6894332603621977e-06, "loss": 0.3144, "step": 35165 }, { "epoch": 3.53, "learning_rate": 1.688641585521614e-06, "loss": 0.0914, "step": 35166 }, { "epoch": 3.53, "learning_rate": 1.6878500908437356e-06, "loss": 0.2302, "step": 35167 }, { "epoch": 3.53, "learning_rate": 1.687058776333611e-06, "loss": 0.1892, "step": 35168 }, { "epoch": 3.53, "learning_rate": 1.6862676419962796e-06, "loss": 0.2334, "step": 35169 }, { "epoch": 3.53, "learning_rate": 1.6854766878367612e-06, "loss": 0.1954, "step": 35170 }, { "epoch": 3.53, "learning_rate": 1.6846859138601012e-06, "loss": 0.1204, "step": 35171 }, { "epoch": 3.53, "learning_rate": 1.6838953200713291e-06, "loss": 0.2662, "step": 35172 }, { "epoch": 3.53, "learning_rate": 1.6831049064754777e-06, "loss": 0.2046, "step": 35173 }, { "epoch": 3.53, "learning_rate": 1.6823146730775695e-06, "loss": 0.0767, "step": 35174 }, { "epoch": 3.53, "learning_rate": 1.6815246198826407e-06, "loss": 0.0395, "step": 35175 }, { "epoch": 3.53, "learning_rate": 1.680734746895717e-06, "loss": 0.176, "step": 35176 }, { "epoch": 3.54, "learning_rate": 1.6799450541218243e-06, "loss": 0.0463, "step": 35177 }, { "epoch": 3.54, "learning_rate": 1.6791555415659854e-06, "loss": 0.1454, "step": 35178 }, { "epoch": 3.54, "learning_rate": 1.6783662092332231e-06, "loss": 0.1855, "step": 35179 }, { "epoch": 3.54, "learning_rate": 1.6775770571285732e-06, "loss": 0.4007, "step": 35180 }, { "epoch": 3.54, "learning_rate": 1.6767880852570417e-06, "loss": 0.0325, "step": 35181 }, { "epoch": 3.54, "learning_rate": 1.6759992936236579e-06, "loss": 0.2579, "step": 35182 }, { "epoch": 3.54, "learning_rate": 1.6752106822334413e-06, "loss": 0.1753, "step": 35183 }, { "epoch": 3.54, "learning_rate": 1.6744222510914075e-06, "loss": 0.3079, "step": 35184 }, { "epoch": 3.54, "learning_rate": 1.6736340002025762e-06, "loss": 0.1265, "step": 35185 }, { "epoch": 3.54, "learning_rate": 1.67284592957196e-06, "loss": 0.3523, "step": 35186 }, { "epoch": 3.54, "learning_rate": 1.6720580392045781e-06, "loss": 0.8175, "step": 35187 }, { "epoch": 3.54, "learning_rate": 1.67127032910544e-06, "loss": 0.0831, "step": 35188 }, { "epoch": 3.54, "learning_rate": 1.6704827992795646e-06, "loss": 0.5073, "step": 35189 }, { "epoch": 3.54, "learning_rate": 1.669695449731955e-06, "loss": 0.0957, "step": 35190 }, { "epoch": 3.54, "learning_rate": 1.6689082804676304e-06, "loss": 0.2534, "step": 35191 }, { "epoch": 3.54, "learning_rate": 1.6681212914915934e-06, "loss": 0.0822, "step": 35192 }, { "epoch": 3.54, "learning_rate": 1.66733448280885e-06, "loss": 0.1051, "step": 35193 }, { "epoch": 3.54, "learning_rate": 1.6665478544244228e-06, "loss": 0.3609, "step": 35194 }, { "epoch": 3.54, "learning_rate": 1.665761406343298e-06, "loss": 0.1295, "step": 35195 }, { "epoch": 3.54, "learning_rate": 1.664975138570488e-06, "loss": 0.1669, "step": 35196 }, { "epoch": 3.54, "learning_rate": 1.664189051110999e-06, "loss": 0.208, "step": 35197 }, { "epoch": 3.54, "learning_rate": 1.6634031439698372e-06, "loss": 0.0496, "step": 35198 }, { "epoch": 3.54, "learning_rate": 1.6626174171519881e-06, "loss": 0.2088, "step": 35199 }, { "epoch": 3.54, "learning_rate": 1.661831870662468e-06, "loss": 0.4775, "step": 35200 }, { "epoch": 3.54, "learning_rate": 1.6610465045062728e-06, "loss": 0.0316, "step": 35201 }, { "epoch": 3.54, "learning_rate": 1.6602613186883886e-06, "loss": 0.0694, "step": 35202 }, { "epoch": 3.54, "learning_rate": 1.6594763132138247e-06, "loss": 0.056, "step": 35203 }, { "epoch": 3.54, "learning_rate": 1.6586914880875702e-06, "loss": 0.0987, "step": 35204 }, { "epoch": 3.54, "learning_rate": 1.6579068433146249e-06, "loss": 0.1028, "step": 35205 }, { "epoch": 3.54, "learning_rate": 1.6571223788999745e-06, "loss": 0.1744, "step": 35206 }, { "epoch": 3.54, "learning_rate": 1.6563380948486184e-06, "loss": 0.4852, "step": 35207 }, { "epoch": 3.54, "learning_rate": 1.6555539911655426e-06, "loss": 0.1333, "step": 35208 }, { "epoch": 3.54, "learning_rate": 1.6547700678557398e-06, "loss": 0.2269, "step": 35209 }, { "epoch": 3.54, "learning_rate": 1.6539863249241927e-06, "loss": 0.0617, "step": 35210 }, { "epoch": 3.54, "learning_rate": 1.653202762375897e-06, "loss": 0.0718, "step": 35211 }, { "epoch": 3.54, "learning_rate": 1.6524193802158393e-06, "loss": 0.5564, "step": 35212 }, { "epoch": 3.54, "learning_rate": 1.651636178448992e-06, "loss": 0.4123, "step": 35213 }, { "epoch": 3.54, "learning_rate": 1.650853157080351e-06, "loss": 0.3567, "step": 35214 }, { "epoch": 3.54, "learning_rate": 1.650070316114899e-06, "loss": 0.4002, "step": 35215 }, { "epoch": 3.54, "learning_rate": 1.649287655557612e-06, "loss": 0.054, "step": 35216 }, { "epoch": 3.54, "learning_rate": 1.6485051754134728e-06, "loss": 0.186, "step": 35217 }, { "epoch": 3.54, "learning_rate": 1.647722875687464e-06, "loss": 0.2273, "step": 35218 }, { "epoch": 3.54, "learning_rate": 1.6469407563845583e-06, "loss": 0.1576, "step": 35219 }, { "epoch": 3.54, "learning_rate": 1.6461588175097387e-06, "loss": 0.1713, "step": 35220 }, { "epoch": 3.54, "learning_rate": 1.645377059067974e-06, "loss": 0.0519, "step": 35221 }, { "epoch": 3.54, "learning_rate": 1.6445954810642471e-06, "loss": 0.2501, "step": 35222 }, { "epoch": 3.54, "learning_rate": 1.6438140835035242e-06, "loss": 0.3909, "step": 35223 }, { "epoch": 3.54, "learning_rate": 1.6430328663907846e-06, "loss": 0.4205, "step": 35224 }, { "epoch": 3.54, "learning_rate": 1.6422518297309907e-06, "loss": 0.266, "step": 35225 }, { "epoch": 3.54, "learning_rate": 1.6414709735291255e-06, "loss": 0.1613, "step": 35226 }, { "epoch": 3.54, "learning_rate": 1.6406902977901484e-06, "loss": 0.1505, "step": 35227 }, { "epoch": 3.54, "learning_rate": 1.6399098025190217e-06, "loss": 0.2096, "step": 35228 }, { "epoch": 3.54, "learning_rate": 1.6391294877207285e-06, "loss": 0.2176, "step": 35229 }, { "epoch": 3.54, "learning_rate": 1.6383493534002314e-06, "loss": 0.4742, "step": 35230 }, { "epoch": 3.54, "learning_rate": 1.6375693995624763e-06, "loss": 0.1201, "step": 35231 }, { "epoch": 3.54, "learning_rate": 1.6367896262124494e-06, "loss": 0.1333, "step": 35232 }, { "epoch": 3.54, "learning_rate": 1.636010033355103e-06, "loss": 0.5564, "step": 35233 }, { "epoch": 3.54, "learning_rate": 1.6352306209953939e-06, "loss": 0.1676, "step": 35234 }, { "epoch": 3.54, "learning_rate": 1.6344513891382906e-06, "loss": 0.3685, "step": 35235 }, { "epoch": 3.54, "learning_rate": 1.6336723377887463e-06, "loss": 0.1221, "step": 35236 }, { "epoch": 3.54, "learning_rate": 1.6328934669517204e-06, "loss": 0.4741, "step": 35237 }, { "epoch": 3.54, "learning_rate": 1.6321147766321687e-06, "loss": 0.1352, "step": 35238 }, { "epoch": 3.54, "learning_rate": 1.631336266835044e-06, "loss": 0.4843, "step": 35239 }, { "epoch": 3.54, "learning_rate": 1.630557937565309e-06, "loss": 0.1485, "step": 35240 }, { "epoch": 3.54, "learning_rate": 1.6297797888279097e-06, "loss": 0.0502, "step": 35241 }, { "epoch": 3.54, "learning_rate": 1.6290018206277923e-06, "loss": 0.0985, "step": 35242 }, { "epoch": 3.54, "learning_rate": 1.6282240329699227e-06, "loss": 0.2655, "step": 35243 }, { "epoch": 3.54, "learning_rate": 1.6274464258592436e-06, "loss": 0.5191, "step": 35244 }, { "epoch": 3.54, "learning_rate": 1.626668999300691e-06, "loss": 0.3123, "step": 35245 }, { "epoch": 3.54, "learning_rate": 1.6258917532992311e-06, "loss": 0.1209, "step": 35246 }, { "epoch": 3.54, "learning_rate": 1.6251146878598034e-06, "loss": 0.1014, "step": 35247 }, { "epoch": 3.54, "learning_rate": 1.6243378029873469e-06, "loss": 0.2165, "step": 35248 }, { "epoch": 3.54, "learning_rate": 1.6235610986868143e-06, "loss": 0.1081, "step": 35249 }, { "epoch": 3.54, "learning_rate": 1.622784574963142e-06, "loss": 0.6612, "step": 35250 }, { "epoch": 3.54, "learning_rate": 1.6220082318212725e-06, "loss": 0.7616, "step": 35251 }, { "epoch": 3.54, "learning_rate": 1.6212320692661486e-06, "loss": 0.0533, "step": 35252 }, { "epoch": 3.54, "learning_rate": 1.6204560873027063e-06, "loss": 0.2553, "step": 35253 }, { "epoch": 3.54, "learning_rate": 1.619680285935885e-06, "loss": 0.0734, "step": 35254 }, { "epoch": 3.54, "learning_rate": 1.6189046651706242e-06, "loss": 0.1051, "step": 35255 }, { "epoch": 3.54, "learning_rate": 1.6181292250118562e-06, "loss": 0.2662, "step": 35256 }, { "epoch": 3.54, "learning_rate": 1.6173539654645143e-06, "loss": 0.2024, "step": 35257 }, { "epoch": 3.54, "learning_rate": 1.6165788865335407e-06, "loss": 0.068, "step": 35258 }, { "epoch": 3.54, "learning_rate": 1.6158039882238584e-06, "loss": 0.6986, "step": 35259 }, { "epoch": 3.54, "learning_rate": 1.615029270540397e-06, "loss": 0.098, "step": 35260 }, { "epoch": 3.54, "learning_rate": 1.6142547334880953e-06, "loss": 0.0808, "step": 35261 }, { "epoch": 3.54, "learning_rate": 1.6134803770718764e-06, "loss": 0.1123, "step": 35262 }, { "epoch": 3.54, "learning_rate": 1.612706201296673e-06, "loss": 0.0985, "step": 35263 }, { "epoch": 3.54, "learning_rate": 1.6119322061674047e-06, "loss": 0.5274, "step": 35264 }, { "epoch": 3.54, "learning_rate": 1.6111583916890038e-06, "loss": 0.0542, "step": 35265 }, { "epoch": 3.54, "learning_rate": 1.61038475786639e-06, "loss": 0.1118, "step": 35266 }, { "epoch": 3.54, "learning_rate": 1.6096113047044857e-06, "loss": 0.198, "step": 35267 }, { "epoch": 3.54, "learning_rate": 1.6088380322082175e-06, "loss": 0.2441, "step": 35268 }, { "epoch": 3.54, "learning_rate": 1.608064940382501e-06, "loss": 0.0154, "step": 35269 }, { "epoch": 3.54, "learning_rate": 1.607292029232259e-06, "loss": 0.0937, "step": 35270 }, { "epoch": 3.54, "learning_rate": 1.6065192987624077e-06, "loss": 0.189, "step": 35271 }, { "epoch": 3.54, "learning_rate": 1.6057467489778732e-06, "loss": 0.0879, "step": 35272 }, { "epoch": 3.54, "learning_rate": 1.604974379883558e-06, "loss": 0.1102, "step": 35273 }, { "epoch": 3.54, "learning_rate": 1.6042021914843818e-06, "loss": 0.1243, "step": 35274 }, { "epoch": 3.54, "learning_rate": 1.6034301837852638e-06, "loss": 0.0793, "step": 35275 }, { "epoch": 3.55, "learning_rate": 1.6026583567911201e-06, "loss": 0.2008, "step": 35276 }, { "epoch": 3.55, "learning_rate": 1.6018867105068469e-06, "loss": 0.0508, "step": 35277 }, { "epoch": 3.55, "learning_rate": 1.6011152449373667e-06, "loss": 0.1116, "step": 35278 }, { "epoch": 3.55, "learning_rate": 1.6003439600875857e-06, "loss": 0.1244, "step": 35279 }, { "epoch": 3.55, "learning_rate": 1.59957285596241e-06, "loss": 0.3141, "step": 35280 }, { "epoch": 3.55, "learning_rate": 1.5988019325667492e-06, "loss": 0.5846, "step": 35281 }, { "epoch": 3.55, "learning_rate": 1.5980311899055123e-06, "loss": 0.2357, "step": 35282 }, { "epoch": 3.55, "learning_rate": 1.5972606279835956e-06, "loss": 0.5267, "step": 35283 }, { "epoch": 3.55, "learning_rate": 1.5964902468059083e-06, "loss": 0.3385, "step": 35284 }, { "epoch": 3.55, "learning_rate": 1.5957200463773535e-06, "loss": 0.4004, "step": 35285 }, { "epoch": 3.55, "learning_rate": 1.5949500267028304e-06, "loss": 0.4669, "step": 35286 }, { "epoch": 3.55, "learning_rate": 1.5941801877872387e-06, "loss": 0.1328, "step": 35287 }, { "epoch": 3.55, "learning_rate": 1.5934105296354806e-06, "loss": 0.0837, "step": 35288 }, { "epoch": 3.55, "learning_rate": 1.5926410522524426e-06, "loss": 0.3934, "step": 35289 }, { "epoch": 3.55, "learning_rate": 1.591871755643044e-06, "loss": 0.0352, "step": 35290 }, { "epoch": 3.55, "learning_rate": 1.5911026398121543e-06, "loss": 0.0541, "step": 35291 }, { "epoch": 3.55, "learning_rate": 1.590333704764686e-06, "loss": 0.1871, "step": 35292 }, { "epoch": 3.55, "learning_rate": 1.5895649505055255e-06, "loss": 0.1985, "step": 35293 }, { "epoch": 3.55, "learning_rate": 1.5887963770395654e-06, "loss": 0.3161, "step": 35294 }, { "epoch": 3.55, "learning_rate": 1.5880279843716983e-06, "loss": 0.0998, "step": 35295 }, { "epoch": 3.55, "learning_rate": 1.5872597725068105e-06, "loss": 0.1213, "step": 35296 }, { "epoch": 3.55, "learning_rate": 1.5864917414497948e-06, "loss": 0.0744, "step": 35297 }, { "epoch": 3.55, "learning_rate": 1.5857238912055373e-06, "loss": 0.0979, "step": 35298 }, { "epoch": 3.55, "learning_rate": 1.5849562217789205e-06, "loss": 0.1232, "step": 35299 }, { "epoch": 3.55, "learning_rate": 1.5841887331748372e-06, "loss": 0.3421, "step": 35300 }, { "epoch": 3.55, "learning_rate": 1.5834214253981638e-06, "loss": 0.403, "step": 35301 }, { "epoch": 3.55, "learning_rate": 1.5826542984537862e-06, "loss": 0.0399, "step": 35302 }, { "epoch": 3.55, "learning_rate": 1.5818873523465804e-06, "loss": 0.3918, "step": 35303 }, { "epoch": 3.55, "learning_rate": 1.581120587081446e-06, "loss": 0.3494, "step": 35304 }, { "epoch": 3.55, "learning_rate": 1.580354002663239e-06, "loss": 0.0476, "step": 35305 }, { "epoch": 3.55, "learning_rate": 1.579587599096849e-06, "loss": 0.0704, "step": 35306 }, { "epoch": 3.55, "learning_rate": 1.5788213763871517e-06, "loss": 0.3006, "step": 35307 }, { "epoch": 3.55, "learning_rate": 1.5780553345390303e-06, "loss": 0.2085, "step": 35308 }, { "epoch": 3.55, "learning_rate": 1.577289473557344e-06, "loss": 0.1237, "step": 35309 }, { "epoch": 3.55, "learning_rate": 1.5765237934469756e-06, "loss": 0.375, "step": 35310 }, { "epoch": 3.55, "learning_rate": 1.5757582942128013e-06, "loss": 0.749, "step": 35311 }, { "epoch": 3.55, "learning_rate": 1.574992975859687e-06, "loss": 0.256, "step": 35312 }, { "epoch": 3.55, "learning_rate": 1.5742278383925024e-06, "loss": 0.1326, "step": 35313 }, { "epoch": 3.55, "learning_rate": 1.57346288181612e-06, "loss": 0.1078, "step": 35314 }, { "epoch": 3.55, "learning_rate": 1.5726981061354028e-06, "loss": 0.3037, "step": 35315 }, { "epoch": 3.55, "learning_rate": 1.5719335113552236e-06, "loss": 0.157, "step": 35316 }, { "epoch": 3.55, "learning_rate": 1.5711690974804383e-06, "loss": 0.1524, "step": 35317 }, { "epoch": 3.55, "learning_rate": 1.5704048645159298e-06, "loss": 0.0329, "step": 35318 }, { "epoch": 3.55, "learning_rate": 1.569640812466544e-06, "loss": 0.0335, "step": 35319 }, { "epoch": 3.55, "learning_rate": 1.568876941337144e-06, "loss": 0.0527, "step": 35320 }, { "epoch": 3.55, "learning_rate": 1.5681132511325992e-06, "loss": 0.0897, "step": 35321 }, { "epoch": 3.55, "learning_rate": 1.5673497418577686e-06, "loss": 0.1812, "step": 35322 }, { "epoch": 3.55, "learning_rate": 1.5665864135175023e-06, "loss": 0.1307, "step": 35323 }, { "epoch": 3.55, "learning_rate": 1.565823266116666e-06, "loss": 0.1105, "step": 35324 }, { "epoch": 3.55, "learning_rate": 1.5650602996601126e-06, "loss": 0.1128, "step": 35325 }, { "epoch": 3.55, "learning_rate": 1.5642975141526983e-06, "loss": 0.0314, "step": 35326 }, { "epoch": 3.55, "learning_rate": 1.563534909599279e-06, "loss": 0.0448, "step": 35327 }, { "epoch": 3.55, "learning_rate": 1.5627724860047044e-06, "loss": 0.2723, "step": 35328 }, { "epoch": 3.55, "learning_rate": 1.5620102433738304e-06, "loss": 0.2322, "step": 35329 }, { "epoch": 3.55, "learning_rate": 1.5612481817114998e-06, "loss": 0.0679, "step": 35330 }, { "epoch": 3.55, "learning_rate": 1.5604863010225723e-06, "loss": 0.1134, "step": 35331 }, { "epoch": 3.55, "learning_rate": 1.5597246013118871e-06, "loss": 0.0801, "step": 35332 }, { "epoch": 3.55, "learning_rate": 1.558963082584297e-06, "loss": 0.508, "step": 35333 }, { "epoch": 3.55, "learning_rate": 1.5582017448446485e-06, "loss": 0.1392, "step": 35334 }, { "epoch": 3.55, "learning_rate": 1.5574405880977772e-06, "loss": 0.4514, "step": 35335 }, { "epoch": 3.55, "learning_rate": 1.5566796123485426e-06, "loss": 0.4124, "step": 35336 }, { "epoch": 3.55, "learning_rate": 1.5559188176017747e-06, "loss": 0.1107, "step": 35337 }, { "epoch": 3.55, "learning_rate": 1.5551582038623124e-06, "loss": 0.2295, "step": 35338 }, { "epoch": 3.55, "learning_rate": 1.5543977711350088e-06, "loss": 0.2242, "step": 35339 }, { "epoch": 3.55, "learning_rate": 1.5536375194247033e-06, "loss": 0.1098, "step": 35340 }, { "epoch": 3.55, "learning_rate": 1.5528774487362151e-06, "loss": 0.3396, "step": 35341 }, { "epoch": 3.55, "learning_rate": 1.5521175590743974e-06, "loss": 0.1145, "step": 35342 }, { "epoch": 3.55, "learning_rate": 1.551357850444083e-06, "loss": 0.2274, "step": 35343 }, { "epoch": 3.55, "learning_rate": 1.5505983228501042e-06, "loss": 0.3214, "step": 35344 }, { "epoch": 3.55, "learning_rate": 1.5498389762972943e-06, "loss": 0.4031, "step": 35345 }, { "epoch": 3.55, "learning_rate": 1.5490798107904858e-06, "loss": 0.1269, "step": 35346 }, { "epoch": 3.55, "learning_rate": 1.5483208263345117e-06, "loss": 0.0881, "step": 35347 }, { "epoch": 3.55, "learning_rate": 1.547562022934198e-06, "loss": 0.0543, "step": 35348 }, { "epoch": 3.55, "learning_rate": 1.5468034005943742e-06, "loss": 0.0084, "step": 35349 }, { "epoch": 3.55, "learning_rate": 1.5460449593198767e-06, "loss": 0.2654, "step": 35350 }, { "epoch": 3.55, "learning_rate": 1.545286699115518e-06, "loss": 0.4754, "step": 35351 }, { "epoch": 3.55, "learning_rate": 1.5445286199861307e-06, "loss": 0.3917, "step": 35352 }, { "epoch": 3.55, "learning_rate": 1.5437707219365382e-06, "loss": 0.0737, "step": 35353 }, { "epoch": 3.55, "learning_rate": 1.5430130049715697e-06, "loss": 0.1168, "step": 35354 }, { "epoch": 3.55, "learning_rate": 1.542255469096031e-06, "loss": 0.1614, "step": 35355 }, { "epoch": 3.55, "learning_rate": 1.5414981143147588e-06, "loss": 0.2823, "step": 35356 }, { "epoch": 3.55, "learning_rate": 1.5407409406325657e-06, "loss": 0.0312, "step": 35357 }, { "epoch": 3.55, "learning_rate": 1.539983948054271e-06, "loss": 0.3429, "step": 35358 }, { "epoch": 3.55, "learning_rate": 1.539227136584691e-06, "loss": 0.0285, "step": 35359 }, { "epoch": 3.55, "learning_rate": 1.5384705062286418e-06, "loss": 0.1618, "step": 35360 }, { "epoch": 3.55, "learning_rate": 1.5377140569909399e-06, "loss": 0.4215, "step": 35361 }, { "epoch": 3.55, "learning_rate": 1.5369577888764009e-06, "loss": 0.1687, "step": 35362 }, { "epoch": 3.55, "learning_rate": 1.5362017018898311e-06, "loss": 0.301, "step": 35363 }, { "epoch": 3.55, "learning_rate": 1.5354457960360468e-06, "loss": 0.0491, "step": 35364 }, { "epoch": 3.55, "learning_rate": 1.5346900713198542e-06, "loss": 0.2814, "step": 35365 }, { "epoch": 3.55, "learning_rate": 1.5339345277460692e-06, "loss": 0.2388, "step": 35366 }, { "epoch": 3.55, "learning_rate": 1.5331791653194882e-06, "loss": 0.1731, "step": 35367 }, { "epoch": 3.55, "learning_rate": 1.5324239840449371e-06, "loss": 0.0979, "step": 35368 }, { "epoch": 3.55, "learning_rate": 1.5316689839272058e-06, "loss": 0.5378, "step": 35369 }, { "epoch": 3.55, "learning_rate": 1.5309141649710968e-06, "loss": 0.1123, "step": 35370 }, { "epoch": 3.55, "learning_rate": 1.5301595271814262e-06, "loss": 0.0191, "step": 35371 }, { "epoch": 3.55, "learning_rate": 1.529405070562987e-06, "loss": 0.3715, "step": 35372 }, { "epoch": 3.55, "learning_rate": 1.5286507951205853e-06, "loss": 0.5386, "step": 35373 }, { "epoch": 3.56, "learning_rate": 1.5278967008590205e-06, "loss": 0.0872, "step": 35374 }, { "epoch": 3.56, "learning_rate": 1.5271427877830856e-06, "loss": 0.17, "step": 35375 }, { "epoch": 3.56, "learning_rate": 1.5263890558975867e-06, "loss": 0.206, "step": 35376 }, { "epoch": 3.56, "learning_rate": 1.5256355052073134e-06, "loss": 0.4107, "step": 35377 }, { "epoch": 3.56, "learning_rate": 1.5248821357170618e-06, "loss": 0.3712, "step": 35378 }, { "epoch": 3.56, "learning_rate": 1.524128947431631e-06, "loss": 0.4087, "step": 35379 }, { "epoch": 3.56, "learning_rate": 1.523375940355811e-06, "loss": 0.1558, "step": 35380 }, { "epoch": 3.56, "learning_rate": 1.522623114494388e-06, "loss": 0.1578, "step": 35381 }, { "epoch": 3.56, "learning_rate": 1.5218704698521646e-06, "loss": 0.431, "step": 35382 }, { "epoch": 3.56, "learning_rate": 1.5211180064339236e-06, "loss": 0.091, "step": 35383 }, { "epoch": 3.56, "learning_rate": 1.520365724244448e-06, "loss": 0.1262, "step": 35384 }, { "epoch": 3.56, "learning_rate": 1.5196136232885337e-06, "loss": 0.0517, "step": 35385 }, { "epoch": 3.56, "learning_rate": 1.5188617035709674e-06, "loss": 0.6892, "step": 35386 }, { "epoch": 3.56, "learning_rate": 1.5181099650965247e-06, "loss": 0.3748, "step": 35387 }, { "epoch": 3.56, "learning_rate": 1.5173584078699986e-06, "loss": 0.3715, "step": 35388 }, { "epoch": 3.56, "learning_rate": 1.5166070318961655e-06, "loss": 0.153, "step": 35389 }, { "epoch": 3.56, "learning_rate": 1.5158558371798149e-06, "loss": 0.224, "step": 35390 }, { "epoch": 3.56, "learning_rate": 1.515104823725716e-06, "loss": 0.1809, "step": 35391 }, { "epoch": 3.56, "learning_rate": 1.5143539915386585e-06, "loss": 0.5323, "step": 35392 }, { "epoch": 3.56, "learning_rate": 1.5136033406234118e-06, "loss": 0.1576, "step": 35393 }, { "epoch": 3.56, "learning_rate": 1.512852870984759e-06, "loss": 0.2748, "step": 35394 }, { "epoch": 3.56, "learning_rate": 1.5121025826274726e-06, "loss": 0.2571, "step": 35395 }, { "epoch": 3.56, "learning_rate": 1.511352475556329e-06, "loss": 0.1207, "step": 35396 }, { "epoch": 3.56, "learning_rate": 1.5106025497761012e-06, "loss": 0.0771, "step": 35397 }, { "epoch": 3.56, "learning_rate": 1.509852805291555e-06, "loss": 0.1025, "step": 35398 }, { "epoch": 3.56, "learning_rate": 1.5091032421074734e-06, "loss": 0.1706, "step": 35399 }, { "epoch": 3.56, "learning_rate": 1.5083538602286228e-06, "loss": 0.2703, "step": 35400 }, { "epoch": 3.56, "learning_rate": 1.5076046596597593e-06, "loss": 0.1684, "step": 35401 }, { "epoch": 3.56, "learning_rate": 1.5068556404056688e-06, "loss": 0.6338, "step": 35402 }, { "epoch": 3.56, "learning_rate": 1.5061068024711078e-06, "loss": 0.2293, "step": 35403 }, { "epoch": 3.56, "learning_rate": 1.5053581458608423e-06, "loss": 0.2449, "step": 35404 }, { "epoch": 3.56, "learning_rate": 1.5046096705796385e-06, "loss": 0.0831, "step": 35405 }, { "epoch": 3.56, "learning_rate": 1.5038613766322595e-06, "loss": 0.1058, "step": 35406 }, { "epoch": 3.56, "learning_rate": 1.5031132640234646e-06, "loss": 0.0434, "step": 35407 }, { "epoch": 3.56, "learning_rate": 1.50236533275802e-06, "loss": 0.0374, "step": 35408 }, { "epoch": 3.56, "learning_rate": 1.5016175828406787e-06, "loss": 0.0721, "step": 35409 }, { "epoch": 3.56, "learning_rate": 1.5008700142762e-06, "loss": 0.0424, "step": 35410 }, { "epoch": 3.56, "learning_rate": 1.5001226270693435e-06, "loss": 0.0805, "step": 35411 }, { "epoch": 3.56, "learning_rate": 1.4993754212248655e-06, "loss": 0.2348, "step": 35412 }, { "epoch": 3.56, "learning_rate": 1.4986283967475157e-06, "loss": 0.0948, "step": 35413 }, { "epoch": 3.56, "learning_rate": 1.4978815536420632e-06, "loss": 0.1701, "step": 35414 }, { "epoch": 3.56, "learning_rate": 1.4971348919132443e-06, "loss": 0.074, "step": 35415 }, { "epoch": 3.56, "learning_rate": 1.4963884115658089e-06, "loss": 0.1932, "step": 35416 }, { "epoch": 3.56, "learning_rate": 1.4956421126045194e-06, "loss": 0.0579, "step": 35417 }, { "epoch": 3.56, "learning_rate": 1.4948959950341257e-06, "loss": 0.1928, "step": 35418 }, { "epoch": 3.56, "learning_rate": 1.4941500588593638e-06, "loss": 0.7644, "step": 35419 }, { "epoch": 3.56, "learning_rate": 1.4934043040849865e-06, "loss": 0.7419, "step": 35420 }, { "epoch": 3.56, "learning_rate": 1.4926587307157402e-06, "loss": 0.8513, "step": 35421 }, { "epoch": 3.56, "learning_rate": 1.491913338756371e-06, "loss": 0.274, "step": 35422 }, { "epoch": 3.56, "learning_rate": 1.4911681282116185e-06, "loss": 0.2397, "step": 35423 }, { "epoch": 3.56, "learning_rate": 1.4904230990862288e-06, "loss": 0.3181, "step": 35424 }, { "epoch": 3.56, "learning_rate": 1.4896782513849416e-06, "loss": 0.377, "step": 35425 }, { "epoch": 3.56, "learning_rate": 1.488933585112493e-06, "loss": 0.2458, "step": 35426 }, { "epoch": 3.56, "learning_rate": 1.4881891002736226e-06, "loss": 0.2076, "step": 35427 }, { "epoch": 3.56, "learning_rate": 1.4874447968730765e-06, "loss": 0.2536, "step": 35428 }, { "epoch": 3.56, "learning_rate": 1.4867006749155842e-06, "loss": 0.5729, "step": 35429 }, { "epoch": 3.56, "learning_rate": 1.4859567344058755e-06, "loss": 0.1087, "step": 35430 }, { "epoch": 3.56, "learning_rate": 1.4852129753486965e-06, "loss": 0.0833, "step": 35431 }, { "epoch": 3.56, "learning_rate": 1.4844693977487767e-06, "loss": 0.1357, "step": 35432 }, { "epoch": 3.56, "learning_rate": 1.483726001610839e-06, "loss": 0.06, "step": 35433 }, { "epoch": 3.56, "learning_rate": 1.4829827869396228e-06, "loss": 0.2963, "step": 35434 }, { "epoch": 3.56, "learning_rate": 1.482239753739858e-06, "loss": 0.4167, "step": 35435 }, { "epoch": 3.56, "learning_rate": 1.4814969020162706e-06, "loss": 0.3404, "step": 35436 }, { "epoch": 3.56, "learning_rate": 1.4807542317735867e-06, "loss": 0.0653, "step": 35437 }, { "epoch": 3.56, "learning_rate": 1.4800117430165328e-06, "loss": 0.4775, "step": 35438 }, { "epoch": 3.56, "learning_rate": 1.4792694357498382e-06, "loss": 0.05, "step": 35439 }, { "epoch": 3.56, "learning_rate": 1.4785273099782193e-06, "loss": 0.4553, "step": 35440 }, { "epoch": 3.56, "learning_rate": 1.4777853657064056e-06, "loss": 0.1216, "step": 35441 }, { "epoch": 3.56, "learning_rate": 1.4770436029391133e-06, "loss": 0.0474, "step": 35442 }, { "epoch": 3.56, "learning_rate": 1.4763020216810653e-06, "loss": 0.2756, "step": 35443 }, { "epoch": 3.56, "learning_rate": 1.475560621936981e-06, "loss": 0.0572, "step": 35444 }, { "epoch": 3.56, "learning_rate": 1.4748194037115736e-06, "loss": 0.1735, "step": 35445 }, { "epoch": 3.56, "learning_rate": 1.4740783670095691e-06, "loss": 0.4383, "step": 35446 }, { "epoch": 3.56, "learning_rate": 1.4733375118356774e-06, "loss": 0.366, "step": 35447 }, { "epoch": 3.56, "learning_rate": 1.4725968381946076e-06, "loss": 0.097, "step": 35448 }, { "epoch": 3.56, "learning_rate": 1.4718563460910862e-06, "loss": 0.4963, "step": 35449 }, { "epoch": 3.56, "learning_rate": 1.4711160355298191e-06, "loss": 0.238, "step": 35450 }, { "epoch": 3.56, "learning_rate": 1.4703759065155097e-06, "loss": 0.2635, "step": 35451 }, { "epoch": 3.56, "learning_rate": 1.4696359590528807e-06, "loss": 0.0928, "step": 35452 }, { "epoch": 3.56, "learning_rate": 1.4688961931466317e-06, "loss": 0.0607, "step": 35453 }, { "epoch": 3.56, "learning_rate": 1.4681566088014752e-06, "loss": 0.0614, "step": 35454 }, { "epoch": 3.56, "learning_rate": 1.467417206022118e-06, "loss": 0.2699, "step": 35455 }, { "epoch": 3.56, "learning_rate": 1.4666779848132595e-06, "loss": 0.0039, "step": 35456 }, { "epoch": 3.56, "learning_rate": 1.465938945179609e-06, "loss": 0.2686, "step": 35457 }, { "epoch": 3.56, "learning_rate": 1.4652000871258697e-06, "loss": 0.1124, "step": 35458 }, { "epoch": 3.56, "learning_rate": 1.4644614106567344e-06, "loss": 0.0915, "step": 35459 }, { "epoch": 3.56, "learning_rate": 1.4637229157769228e-06, "loss": 0.6662, "step": 35460 }, { "epoch": 3.56, "learning_rate": 1.4629846024911176e-06, "loss": 0.1221, "step": 35461 }, { "epoch": 3.56, "learning_rate": 1.4622464708040185e-06, "loss": 0.2717, "step": 35462 }, { "epoch": 3.56, "learning_rate": 1.4615085207203315e-06, "loss": 0.0215, "step": 35463 }, { "epoch": 3.56, "learning_rate": 1.46077075224475e-06, "loss": 0.1617, "step": 35464 }, { "epoch": 3.56, "learning_rate": 1.4600331653819565e-06, "loss": 0.0778, "step": 35465 }, { "epoch": 3.56, "learning_rate": 1.4592957601366641e-06, "loss": 0.0946, "step": 35466 }, { "epoch": 3.56, "learning_rate": 1.4585585365135523e-06, "loss": 0.3993, "step": 35467 }, { "epoch": 3.56, "learning_rate": 1.4578214945173174e-06, "loss": 0.2788, "step": 35468 }, { "epoch": 3.56, "learning_rate": 1.4570846341526488e-06, "loss": 0.4339, "step": 35469 }, { "epoch": 3.56, "learning_rate": 1.4563479554242332e-06, "loss": 0.4216, "step": 35470 }, { "epoch": 3.56, "learning_rate": 1.455611458336763e-06, "loss": 0.1626, "step": 35471 }, { "epoch": 3.56, "learning_rate": 1.4548751428949215e-06, "loss": 0.192, "step": 35472 }, { "epoch": 3.57, "learning_rate": 1.4541390091033947e-06, "loss": 0.1951, "step": 35473 }, { "epoch": 3.57, "learning_rate": 1.4534030569668688e-06, "loss": 0.1163, "step": 35474 }, { "epoch": 3.57, "learning_rate": 1.452667286490027e-06, "loss": 0.3843, "step": 35475 }, { "epoch": 3.57, "learning_rate": 1.4519316976775522e-06, "loss": 0.0586, "step": 35476 }, { "epoch": 3.57, "learning_rate": 1.451196290534117e-06, "loss": 0.0231, "step": 35477 }, { "epoch": 3.57, "learning_rate": 1.4504610650644146e-06, "loss": 0.0704, "step": 35478 }, { "epoch": 3.57, "learning_rate": 1.449726021273111e-06, "loss": 0.1323, "step": 35479 }, { "epoch": 3.57, "learning_rate": 1.4489911591648964e-06, "loss": 0.2157, "step": 35480 }, { "epoch": 3.57, "learning_rate": 1.4482564787444364e-06, "loss": 0.1094, "step": 35481 }, { "epoch": 3.57, "learning_rate": 1.447521980016414e-06, "loss": 0.6079, "step": 35482 }, { "epoch": 3.57, "learning_rate": 1.4467876629854993e-06, "loss": 0.0809, "step": 35483 }, { "epoch": 3.57, "learning_rate": 1.4460535276563647e-06, "loss": 0.4113, "step": 35484 }, { "epoch": 3.57, "learning_rate": 1.4453195740336833e-06, "loss": 0.2748, "step": 35485 }, { "epoch": 3.57, "learning_rate": 1.4445858021221249e-06, "loss": 0.1136, "step": 35486 }, { "epoch": 3.57, "learning_rate": 1.4438522119263587e-06, "loss": 0.4378, "step": 35487 }, { "epoch": 3.57, "learning_rate": 1.4431188034510545e-06, "loss": 0.3605, "step": 35488 }, { "epoch": 3.57, "learning_rate": 1.4423855767008786e-06, "loss": 0.2028, "step": 35489 }, { "epoch": 3.57, "learning_rate": 1.4416525316804973e-06, "loss": 0.1154, "step": 35490 }, { "epoch": 3.57, "learning_rate": 1.4409196683945736e-06, "loss": 0.1188, "step": 35491 }, { "epoch": 3.57, "learning_rate": 1.4401869868477769e-06, "loss": 0.0444, "step": 35492 }, { "epoch": 3.57, "learning_rate": 1.4394544870447635e-06, "loss": 0.3262, "step": 35493 }, { "epoch": 3.57, "learning_rate": 1.438722168990193e-06, "loss": 0.3366, "step": 35494 }, { "epoch": 3.57, "learning_rate": 1.437990032688732e-06, "loss": 0.1459, "step": 35495 }, { "epoch": 3.57, "learning_rate": 1.4372580781450395e-06, "loss": 0.128, "step": 35496 }, { "epoch": 3.57, "learning_rate": 1.4365263053637657e-06, "loss": 0.2586, "step": 35497 }, { "epoch": 3.57, "learning_rate": 1.4357947143495764e-06, "loss": 0.2458, "step": 35498 }, { "epoch": 3.57, "learning_rate": 1.4350633051071215e-06, "loss": 0.1527, "step": 35499 }, { "epoch": 3.57, "learning_rate": 1.4343320776410574e-06, "loss": 0.1254, "step": 35500 }, { "epoch": 3.57, "learning_rate": 1.4336010319560367e-06, "loss": 0.6741, "step": 35501 }, { "epoch": 3.57, "learning_rate": 1.4328701680567125e-06, "loss": 0.0448, "step": 35502 }, { "epoch": 3.57, "learning_rate": 1.4321394859477344e-06, "loss": 0.5558, "step": 35503 }, { "epoch": 3.57, "learning_rate": 1.4314089856337554e-06, "loss": 0.2074, "step": 35504 }, { "epoch": 3.57, "learning_rate": 1.4306786671194182e-06, "loss": 0.0725, "step": 35505 }, { "epoch": 3.57, "learning_rate": 1.4299485304093762e-06, "loss": 0.1354, "step": 35506 }, { "epoch": 3.57, "learning_rate": 1.429218575508272e-06, "loss": 0.1128, "step": 35507 }, { "epoch": 3.57, "learning_rate": 1.4284888024207454e-06, "loss": 0.4204, "step": 35508 }, { "epoch": 3.57, "learning_rate": 1.427759211151456e-06, "loss": 0.0936, "step": 35509 }, { "epoch": 3.57, "learning_rate": 1.4270298017050364e-06, "loss": 0.1094, "step": 35510 }, { "epoch": 3.57, "learning_rate": 1.4263005740861235e-06, "loss": 0.0992, "step": 35511 }, { "epoch": 3.57, "learning_rate": 1.4255715282993698e-06, "loss": 0.3009, "step": 35512 }, { "epoch": 3.57, "learning_rate": 1.424842664349405e-06, "loss": 0.1698, "step": 35513 }, { "epoch": 3.57, "learning_rate": 1.4241139822408756e-06, "loss": 0.0996, "step": 35514 }, { "epoch": 3.57, "learning_rate": 1.4233854819784109e-06, "loss": 0.1244, "step": 35515 }, { "epoch": 3.57, "learning_rate": 1.4226571635666507e-06, "loss": 0.2294, "step": 35516 }, { "epoch": 3.57, "learning_rate": 1.4219290270102282e-06, "loss": 0.0934, "step": 35517 }, { "epoch": 3.57, "learning_rate": 1.4212010723137759e-06, "loss": 0.0897, "step": 35518 }, { "epoch": 3.57, "learning_rate": 1.4204732994819302e-06, "loss": 0.1079, "step": 35519 }, { "epoch": 3.57, "learning_rate": 1.4197457085193211e-06, "loss": 0.1513, "step": 35520 }, { "epoch": 3.57, "learning_rate": 1.4190182994305779e-06, "loss": 0.565, "step": 35521 }, { "epoch": 3.57, "learning_rate": 1.4182910722203269e-06, "loss": 0.0826, "step": 35522 }, { "epoch": 3.57, "learning_rate": 1.4175640268931977e-06, "loss": 0.0971, "step": 35523 }, { "epoch": 3.57, "learning_rate": 1.4168371634538235e-06, "loss": 0.3918, "step": 35524 }, { "epoch": 3.57, "learning_rate": 1.4161104819068204e-06, "loss": 0.1722, "step": 35525 }, { "epoch": 3.57, "learning_rate": 1.4153839822568114e-06, "loss": 0.2168, "step": 35526 }, { "epoch": 3.57, "learning_rate": 1.414657664508433e-06, "loss": 0.5049, "step": 35527 }, { "epoch": 3.57, "learning_rate": 1.4139315286662979e-06, "loss": 0.1668, "step": 35528 }, { "epoch": 3.57, "learning_rate": 1.4132055747350226e-06, "loss": 0.0379, "step": 35529 }, { "epoch": 3.57, "learning_rate": 1.412479802719233e-06, "loss": 0.1712, "step": 35530 }, { "epoch": 3.57, "learning_rate": 1.4117542126235493e-06, "loss": 0.0722, "step": 35531 }, { "epoch": 3.57, "learning_rate": 1.4110288044525843e-06, "loss": 0.8616, "step": 35532 }, { "epoch": 3.57, "learning_rate": 1.4103035782109575e-06, "loss": 0.2999, "step": 35533 }, { "epoch": 3.57, "learning_rate": 1.4095785339032818e-06, "loss": 0.0854, "step": 35534 }, { "epoch": 3.57, "learning_rate": 1.4088536715341739e-06, "loss": 0.5767, "step": 35535 }, { "epoch": 3.57, "learning_rate": 1.4081289911082463e-06, "loss": 0.4419, "step": 35536 }, { "epoch": 3.57, "learning_rate": 1.4074044926301022e-06, "loss": 0.0528, "step": 35537 }, { "epoch": 3.57, "learning_rate": 1.4066801761043645e-06, "loss": 0.5208, "step": 35538 }, { "epoch": 3.57, "learning_rate": 1.4059560415356366e-06, "loss": 0.1167, "step": 35539 }, { "epoch": 3.57, "learning_rate": 1.4052320889285207e-06, "loss": 0.1786, "step": 35540 }, { "epoch": 3.57, "learning_rate": 1.404508318287634e-06, "loss": 0.6003, "step": 35541 }, { "epoch": 3.57, "learning_rate": 1.403784729617582e-06, "loss": 0.1326, "step": 35542 }, { "epoch": 3.57, "learning_rate": 1.4030613229229583e-06, "loss": 0.152, "step": 35543 }, { "epoch": 3.57, "learning_rate": 1.4023380982083756e-06, "loss": 0.2571, "step": 35544 }, { "epoch": 3.57, "learning_rate": 1.4016150554784336e-06, "loss": 0.2494, "step": 35545 }, { "epoch": 3.57, "learning_rate": 1.4008921947377384e-06, "loss": 0.1418, "step": 35546 }, { "epoch": 3.57, "learning_rate": 1.4001695159908801e-06, "loss": 0.1021, "step": 35547 }, { "epoch": 3.57, "learning_rate": 1.399447019242468e-06, "loss": 0.0916, "step": 35548 }, { "epoch": 3.57, "learning_rate": 1.398724704497092e-06, "loss": 0.1008, "step": 35549 }, { "epoch": 3.57, "learning_rate": 1.3980025717593514e-06, "loss": 0.2245, "step": 35550 }, { "epoch": 3.57, "learning_rate": 1.3972806210338428e-06, "loss": 0.2038, "step": 35551 }, { "epoch": 3.57, "learning_rate": 1.3965588523251593e-06, "loss": 0.2576, "step": 35552 }, { "epoch": 3.57, "learning_rate": 1.3958372656378938e-06, "loss": 0.3239, "step": 35553 }, { "epoch": 3.57, "learning_rate": 1.395115860976639e-06, "loss": 0.0542, "step": 35554 }, { "epoch": 3.57, "learning_rate": 1.3943946383459782e-06, "loss": 0.1177, "step": 35555 }, { "epoch": 3.57, "learning_rate": 1.3936735977505177e-06, "loss": 0.1762, "step": 35556 }, { "epoch": 3.57, "learning_rate": 1.3929527391948337e-06, "loss": 0.1033, "step": 35557 }, { "epoch": 3.57, "learning_rate": 1.3922320626835095e-06, "loss": 0.1417, "step": 35558 }, { "epoch": 3.57, "learning_rate": 1.3915115682211444e-06, "loss": 0.2164, "step": 35559 }, { "epoch": 3.57, "learning_rate": 1.3907912558123114e-06, "loss": 0.189, "step": 35560 }, { "epoch": 3.57, "learning_rate": 1.390071125461604e-06, "loss": 0.1514, "step": 35561 }, { "epoch": 3.57, "learning_rate": 1.3893511771735977e-06, "loss": 0.3389, "step": 35562 }, { "epoch": 3.57, "learning_rate": 1.3886314109528764e-06, "loss": 0.1822, "step": 35563 }, { "epoch": 3.57, "learning_rate": 1.3879118268040225e-06, "loss": 0.1217, "step": 35564 }, { "epoch": 3.57, "learning_rate": 1.3871924247316125e-06, "loss": 0.012, "step": 35565 }, { "epoch": 3.57, "learning_rate": 1.3864732047402195e-06, "loss": 0.1626, "step": 35566 }, { "epoch": 3.57, "learning_rate": 1.3857541668344364e-06, "loss": 0.133, "step": 35567 }, { "epoch": 3.57, "learning_rate": 1.385035311018823e-06, "loss": 0.1891, "step": 35568 }, { "epoch": 3.57, "learning_rate": 1.384316637297952e-06, "loss": 0.0971, "step": 35569 }, { "epoch": 3.57, "learning_rate": 1.383598145676417e-06, "loss": 0.3515, "step": 35570 }, { "epoch": 3.58, "learning_rate": 1.3828798361587702e-06, "loss": 0.386, "step": 35571 }, { "epoch": 3.58, "learning_rate": 1.3821617087495887e-06, "loss": 0.0412, "step": 35572 }, { "epoch": 3.58, "learning_rate": 1.381443763453445e-06, "loss": 0.1147, "step": 35573 }, { "epoch": 3.58, "learning_rate": 1.3807260002749088e-06, "loss": 0.1369, "step": 35574 }, { "epoch": 3.58, "learning_rate": 1.380008419218537e-06, "loss": 0.5024, "step": 35575 }, { "epoch": 3.58, "learning_rate": 1.379291020288912e-06, "loss": 0.4407, "step": 35576 }, { "epoch": 3.58, "learning_rate": 1.378573803490587e-06, "loss": 0.0718, "step": 35577 }, { "epoch": 3.58, "learning_rate": 1.3778567688281319e-06, "loss": 0.3113, "step": 35578 }, { "epoch": 3.58, "learning_rate": 1.3771399163061093e-06, "loss": 0.0511, "step": 35579 }, { "epoch": 3.58, "learning_rate": 1.376423245929076e-06, "loss": 0.3272, "step": 35580 }, { "epoch": 3.58, "learning_rate": 1.375706757701598e-06, "loss": 0.1746, "step": 35581 }, { "epoch": 3.58, "learning_rate": 1.3749904516282352e-06, "loss": 0.1037, "step": 35582 }, { "epoch": 3.58, "learning_rate": 1.3742743277135406e-06, "loss": 0.137, "step": 35583 }, { "epoch": 3.58, "learning_rate": 1.373558385962077e-06, "loss": 0.6576, "step": 35584 }, { "epoch": 3.58, "learning_rate": 1.3728426263783944e-06, "loss": 0.2293, "step": 35585 }, { "epoch": 3.58, "learning_rate": 1.3721270489670523e-06, "loss": 0.0223, "step": 35586 }, { "epoch": 3.58, "learning_rate": 1.3714116537326004e-06, "loss": 0.2381, "step": 35587 }, { "epoch": 3.58, "learning_rate": 1.370696440679602e-06, "loss": 0.5234, "step": 35588 }, { "epoch": 3.58, "learning_rate": 1.3699814098125896e-06, "loss": 0.2026, "step": 35589 }, { "epoch": 3.58, "learning_rate": 1.36926656113613e-06, "loss": 0.1985, "step": 35590 }, { "epoch": 3.58, "learning_rate": 1.3685518946547659e-06, "loss": 0.1789, "step": 35591 }, { "epoch": 3.58, "learning_rate": 1.3678374103730473e-06, "loss": 0.4216, "step": 35592 }, { "epoch": 3.58, "learning_rate": 1.367123108295517e-06, "loss": 0.0379, "step": 35593 }, { "epoch": 3.58, "learning_rate": 1.3664089884267216e-06, "loss": 0.4166, "step": 35594 }, { "epoch": 3.58, "learning_rate": 1.3656950507712107e-06, "loss": 0.198, "step": 35595 }, { "epoch": 3.58, "learning_rate": 1.3649812953335206e-06, "loss": 0.0709, "step": 35596 }, { "epoch": 3.58, "learning_rate": 1.3642677221181975e-06, "loss": 0.047, "step": 35597 }, { "epoch": 3.58, "learning_rate": 1.363554331129775e-06, "loss": 0.1893, "step": 35598 }, { "epoch": 3.58, "learning_rate": 1.3628411223728088e-06, "loss": 0.2276, "step": 35599 }, { "epoch": 3.58, "learning_rate": 1.3621280958518223e-06, "loss": 0.1022, "step": 35600 }, { "epoch": 3.58, "learning_rate": 1.3614152515713552e-06, "loss": 0.1562, "step": 35601 }, { "epoch": 3.58, "learning_rate": 1.3607025895359537e-06, "loss": 0.1418, "step": 35602 }, { "epoch": 3.58, "learning_rate": 1.3599901097501444e-06, "loss": 0.3858, "step": 35603 }, { "epoch": 3.58, "learning_rate": 1.3592778122184568e-06, "loss": 0.0471, "step": 35604 }, { "epoch": 3.58, "learning_rate": 1.3585656969454341e-06, "loss": 0.013, "step": 35605 }, { "epoch": 3.58, "learning_rate": 1.357853763935606e-06, "loss": 0.2133, "step": 35606 }, { "epoch": 3.58, "learning_rate": 1.3571420131934953e-06, "loss": 0.145, "step": 35607 }, { "epoch": 3.58, "learning_rate": 1.3564304447236386e-06, "loss": 0.6108, "step": 35608 }, { "epoch": 3.58, "learning_rate": 1.3557190585305623e-06, "loss": 0.9909, "step": 35609 }, { "epoch": 3.58, "learning_rate": 1.355007854618796e-06, "loss": 0.034, "step": 35610 }, { "epoch": 3.58, "learning_rate": 1.3542968329928595e-06, "loss": 0.2222, "step": 35611 }, { "epoch": 3.58, "learning_rate": 1.353585993657279e-06, "loss": 0.3344, "step": 35612 }, { "epoch": 3.58, "learning_rate": 1.3528753366165814e-06, "loss": 0.0292, "step": 35613 }, { "epoch": 3.58, "learning_rate": 1.3521648618752892e-06, "loss": 0.0316, "step": 35614 }, { "epoch": 3.58, "learning_rate": 1.3514545694379156e-06, "loss": 0.0715, "step": 35615 }, { "epoch": 3.58, "learning_rate": 1.3507444593089968e-06, "loss": 0.0384, "step": 35616 }, { "epoch": 3.58, "learning_rate": 1.3500345314930329e-06, "loss": 0.0529, "step": 35617 }, { "epoch": 3.58, "learning_rate": 1.34932478599455e-06, "loss": 0.2384, "step": 35618 }, { "epoch": 3.58, "learning_rate": 1.348615222818068e-06, "loss": 0.1166, "step": 35619 }, { "epoch": 3.58, "learning_rate": 1.3479058419681033e-06, "loss": 0.2625, "step": 35620 }, { "epoch": 3.58, "learning_rate": 1.347196643449159e-06, "loss": 0.1833, "step": 35621 }, { "epoch": 3.58, "learning_rate": 1.3464876272657577e-06, "loss": 0.5697, "step": 35622 }, { "epoch": 3.58, "learning_rate": 1.3457787934224098e-06, "loss": 0.0084, "step": 35623 }, { "epoch": 3.58, "learning_rate": 1.345070141923621e-06, "loss": 0.4296, "step": 35624 }, { "epoch": 3.58, "learning_rate": 1.3443616727739083e-06, "loss": 0.1662, "step": 35625 }, { "epoch": 3.58, "learning_rate": 1.3436533859777777e-06, "loss": 0.2047, "step": 35626 }, { "epoch": 3.58, "learning_rate": 1.3429452815397359e-06, "loss": 0.1031, "step": 35627 }, { "epoch": 3.58, "learning_rate": 1.3422373594642856e-06, "loss": 0.3891, "step": 35628 }, { "epoch": 3.58, "learning_rate": 1.3415296197559368e-06, "loss": 0.3063, "step": 35629 }, { "epoch": 3.58, "learning_rate": 1.3408220624191857e-06, "loss": 0.0908, "step": 35630 }, { "epoch": 3.58, "learning_rate": 1.3401146874585524e-06, "loss": 0.32, "step": 35631 }, { "epoch": 3.58, "learning_rate": 1.3394074948785196e-06, "loss": 0.1521, "step": 35632 }, { "epoch": 3.58, "learning_rate": 1.3387004846835904e-06, "loss": 0.2688, "step": 35633 }, { "epoch": 3.58, "learning_rate": 1.3379936568782747e-06, "loss": 0.0508, "step": 35634 }, { "epoch": 3.58, "learning_rate": 1.3372870114670621e-06, "loss": 0.2359, "step": 35635 }, { "epoch": 3.58, "learning_rate": 1.3365805484544457e-06, "loss": 0.2302, "step": 35636 }, { "epoch": 3.58, "learning_rate": 1.335874267844932e-06, "loss": 0.1902, "step": 35637 }, { "epoch": 3.58, "learning_rate": 1.335168169643014e-06, "loss": 0.0376, "step": 35638 }, { "epoch": 3.58, "learning_rate": 1.3344622538531715e-06, "loss": 0.1643, "step": 35639 }, { "epoch": 3.58, "learning_rate": 1.3337565204799107e-06, "loss": 0.0567, "step": 35640 }, { "epoch": 3.58, "learning_rate": 1.3330509695277182e-06, "loss": 0.1197, "step": 35641 }, { "epoch": 3.58, "learning_rate": 1.3323456010010836e-06, "loss": 0.1556, "step": 35642 }, { "epoch": 3.58, "learning_rate": 1.3316404149044969e-06, "loss": 0.5918, "step": 35643 }, { "epoch": 3.58, "learning_rate": 1.330935411242441e-06, "loss": 0.0605, "step": 35644 }, { "epoch": 3.58, "learning_rate": 1.3302305900194123e-06, "loss": 0.0774, "step": 35645 }, { "epoch": 3.58, "learning_rate": 1.329525951239887e-06, "loss": 0.2388, "step": 35646 }, { "epoch": 3.58, "learning_rate": 1.3288214949083456e-06, "loss": 0.1751, "step": 35647 }, { "epoch": 3.58, "learning_rate": 1.328117221029287e-06, "loss": 0.1273, "step": 35648 }, { "epoch": 3.58, "learning_rate": 1.3274131296071812e-06, "loss": 0.0696, "step": 35649 }, { "epoch": 3.58, "learning_rate": 1.3267092206465014e-06, "loss": 0.2882, "step": 35650 }, { "epoch": 3.58, "learning_rate": 1.3260054941517442e-06, "loss": 0.2126, "step": 35651 }, { "epoch": 3.58, "learning_rate": 1.3253019501273822e-06, "loss": 0.194, "step": 35652 }, { "epoch": 3.58, "learning_rate": 1.3245985885778822e-06, "loss": 0.1105, "step": 35653 }, { "epoch": 3.58, "learning_rate": 1.323895409507734e-06, "loss": 0.1101, "step": 35654 }, { "epoch": 3.58, "learning_rate": 1.3231924129214034e-06, "loss": 0.0279, "step": 35655 }, { "epoch": 3.58, "learning_rate": 1.3224895988233709e-06, "loss": 0.6326, "step": 35656 }, { "epoch": 3.58, "learning_rate": 1.3217869672181026e-06, "loss": 0.2583, "step": 35657 }, { "epoch": 3.58, "learning_rate": 1.3210845181100716e-06, "loss": 0.3626, "step": 35658 }, { "epoch": 3.58, "learning_rate": 1.3203822515037512e-06, "loss": 0.136, "step": 35659 }, { "epoch": 3.58, "learning_rate": 1.319680167403604e-06, "loss": 0.2768, "step": 35660 }, { "epoch": 3.58, "learning_rate": 1.3189782658141037e-06, "loss": 0.2764, "step": 35661 }, { "epoch": 3.58, "learning_rate": 1.3182765467397128e-06, "loss": 0.057, "step": 35662 }, { "epoch": 3.58, "learning_rate": 1.3175750101849048e-06, "loss": 0.0584, "step": 35663 }, { "epoch": 3.58, "learning_rate": 1.316873656154136e-06, "loss": 0.0516, "step": 35664 }, { "epoch": 3.58, "learning_rate": 1.3161724846518663e-06, "loss": 0.0679, "step": 35665 }, { "epoch": 3.58, "learning_rate": 1.3154714956825686e-06, "loss": 0.1868, "step": 35666 }, { "epoch": 3.58, "learning_rate": 1.3147706892506995e-06, "loss": 0.2058, "step": 35667 }, { "epoch": 3.58, "learning_rate": 1.3140700653607085e-06, "loss": 0.1124, "step": 35668 }, { "epoch": 3.58, "learning_rate": 1.313369624017069e-06, "loss": 0.0694, "step": 35669 }, { "epoch": 3.59, "learning_rate": 1.3126693652242305e-06, "loss": 0.2327, "step": 35670 }, { "epoch": 3.59, "learning_rate": 1.311969288986653e-06, "loss": 0.2059, "step": 35671 }, { "epoch": 3.59, "learning_rate": 1.3112693953087896e-06, "loss": 0.1246, "step": 35672 }, { "epoch": 3.59, "learning_rate": 1.3105696841950932e-06, "loss": 0.5362, "step": 35673 }, { "epoch": 3.59, "learning_rate": 1.309870155650017e-06, "loss": 0.3149, "step": 35674 }, { "epoch": 3.59, "learning_rate": 1.309170809678014e-06, "loss": 0.5032, "step": 35675 }, { "epoch": 3.59, "learning_rate": 1.3084716462835279e-06, "loss": 0.2815, "step": 35676 }, { "epoch": 3.59, "learning_rate": 1.3077726654710243e-06, "loss": 0.2269, "step": 35677 }, { "epoch": 3.59, "learning_rate": 1.3070738672449335e-06, "loss": 0.1004, "step": 35678 }, { "epoch": 3.59, "learning_rate": 1.306375251609705e-06, "loss": 0.2789, "step": 35679 }, { "epoch": 3.59, "learning_rate": 1.3056768185697987e-06, "loss": 0.0944, "step": 35680 }, { "epoch": 3.59, "learning_rate": 1.3049785681296445e-06, "loss": 0.033, "step": 35681 }, { "epoch": 3.59, "learning_rate": 1.3042805002936853e-06, "loss": 0.1805, "step": 35682 }, { "epoch": 3.59, "learning_rate": 1.3035826150663744e-06, "loss": 0.2897, "step": 35683 }, { "epoch": 3.59, "learning_rate": 1.3028849124521514e-06, "loss": 0.6247, "step": 35684 }, { "epoch": 3.59, "learning_rate": 1.3021873924554428e-06, "loss": 0.0635, "step": 35685 }, { "epoch": 3.59, "learning_rate": 1.3014900550806985e-06, "loss": 0.5662, "step": 35686 }, { "epoch": 3.59, "learning_rate": 1.3007929003323581e-06, "loss": 0.1705, "step": 35687 }, { "epoch": 3.59, "learning_rate": 1.3000959282148517e-06, "loss": 0.4698, "step": 35688 }, { "epoch": 3.59, "learning_rate": 1.2993991387326154e-06, "loss": 0.0321, "step": 35689 }, { "epoch": 3.59, "learning_rate": 1.2987025318900859e-06, "loss": 0.5095, "step": 35690 }, { "epoch": 3.59, "learning_rate": 1.2980061076916961e-06, "loss": 0.2456, "step": 35691 }, { "epoch": 3.59, "learning_rate": 1.297309866141876e-06, "loss": 0.1015, "step": 35692 }, { "epoch": 3.59, "learning_rate": 1.2966138072450551e-06, "loss": 0.1735, "step": 35693 }, { "epoch": 3.59, "learning_rate": 1.2959179310056636e-06, "loss": 0.1703, "step": 35694 }, { "epoch": 3.59, "learning_rate": 1.2952222374281375e-06, "loss": 0.5653, "step": 35695 }, { "epoch": 3.59, "learning_rate": 1.2945267265168902e-06, "loss": 0.0558, "step": 35696 }, { "epoch": 3.59, "learning_rate": 1.293831398276355e-06, "loss": 0.1334, "step": 35697 }, { "epoch": 3.59, "learning_rate": 1.2931362527109647e-06, "loss": 0.1317, "step": 35698 }, { "epoch": 3.59, "learning_rate": 1.2924412898251258e-06, "loss": 0.1491, "step": 35699 }, { "epoch": 3.59, "learning_rate": 1.2917465096232717e-06, "loss": 0.0525, "step": 35700 }, { "epoch": 3.59, "learning_rate": 1.291051912109822e-06, "loss": 0.4676, "step": 35701 }, { "epoch": 3.59, "learning_rate": 1.2903574972891964e-06, "loss": 0.263, "step": 35702 }, { "epoch": 3.59, "learning_rate": 1.2896632651658147e-06, "loss": 0.4557, "step": 35703 }, { "epoch": 3.59, "learning_rate": 1.288969215744097e-06, "loss": 0.331, "step": 35704 }, { "epoch": 3.59, "learning_rate": 1.2882753490284527e-06, "loss": 0.5665, "step": 35705 }, { "epoch": 3.59, "learning_rate": 1.287581665023302e-06, "loss": 0.4567, "step": 35706 }, { "epoch": 3.59, "learning_rate": 1.286888163733061e-06, "loss": 0.4479, "step": 35707 }, { "epoch": 3.59, "learning_rate": 1.2861948451621364e-06, "loss": 0.2689, "step": 35708 }, { "epoch": 3.59, "learning_rate": 1.2855017093149512e-06, "loss": 0.0605, "step": 35709 }, { "epoch": 3.59, "learning_rate": 1.2848087561959053e-06, "loss": 0.216, "step": 35710 }, { "epoch": 3.59, "learning_rate": 1.2841159858094088e-06, "loss": 0.0684, "step": 35711 }, { "epoch": 3.59, "learning_rate": 1.2834233981598843e-06, "loss": 0.441, "step": 35712 }, { "epoch": 3.59, "learning_rate": 1.2827309932517218e-06, "loss": 0.7068, "step": 35713 }, { "epoch": 3.59, "learning_rate": 1.2820387710893312e-06, "loss": 0.0883, "step": 35714 }, { "epoch": 3.59, "learning_rate": 1.2813467316771255e-06, "loss": 0.091, "step": 35715 }, { "epoch": 3.59, "learning_rate": 1.2806548750195114e-06, "loss": 0.1786, "step": 35716 }, { "epoch": 3.59, "learning_rate": 1.2799632011208717e-06, "loss": 0.448, "step": 35717 }, { "epoch": 3.59, "learning_rate": 1.2792717099856265e-06, "loss": 0.1419, "step": 35718 }, { "epoch": 3.59, "learning_rate": 1.278580401618169e-06, "loss": 0.254, "step": 35719 }, { "epoch": 3.59, "learning_rate": 1.2778892760229022e-06, "loss": 0.1516, "step": 35720 }, { "epoch": 3.59, "learning_rate": 1.2771983332042192e-06, "loss": 0.1961, "step": 35721 }, { "epoch": 3.59, "learning_rate": 1.2765075731665165e-06, "loss": 0.117, "step": 35722 }, { "epoch": 3.59, "learning_rate": 1.2758169959142008e-06, "loss": 0.0681, "step": 35723 }, { "epoch": 3.59, "learning_rate": 1.2751266014516516e-06, "loss": 0.1235, "step": 35724 }, { "epoch": 3.59, "learning_rate": 1.274436389783269e-06, "loss": 0.2586, "step": 35725 }, { "epoch": 3.59, "learning_rate": 1.2737463609134458e-06, "loss": 0.1674, "step": 35726 }, { "epoch": 3.59, "learning_rate": 1.2730565148465789e-06, "loss": 0.1889, "step": 35727 }, { "epoch": 3.59, "learning_rate": 1.2723668515870447e-06, "loss": 0.025, "step": 35728 }, { "epoch": 3.59, "learning_rate": 1.2716773711392427e-06, "loss": 0.1994, "step": 35729 }, { "epoch": 3.59, "learning_rate": 1.2709880735075595e-06, "loss": 0.232, "step": 35730 }, { "epoch": 3.59, "learning_rate": 1.2702989586963754e-06, "loss": 0.3715, "step": 35731 }, { "epoch": 3.59, "learning_rate": 1.2696100267100796e-06, "loss": 0.1206, "step": 35732 }, { "epoch": 3.59, "learning_rate": 1.2689212775530556e-06, "loss": 0.1764, "step": 35733 }, { "epoch": 3.59, "learning_rate": 1.2682327112296899e-06, "loss": 0.0696, "step": 35734 }, { "epoch": 3.59, "learning_rate": 1.267544327744359e-06, "loss": 0.0596, "step": 35735 }, { "epoch": 3.59, "learning_rate": 1.2668561271014456e-06, "loss": 0.124, "step": 35736 }, { "epoch": 3.59, "learning_rate": 1.26616810930533e-06, "loss": 0.1914, "step": 35737 }, { "epoch": 3.59, "learning_rate": 1.2654802743603888e-06, "loss": 0.1155, "step": 35738 }, { "epoch": 3.59, "learning_rate": 1.2647926222710015e-06, "loss": 0.1353, "step": 35739 }, { "epoch": 3.59, "learning_rate": 1.2641051530415347e-06, "loss": 0.0802, "step": 35740 }, { "epoch": 3.59, "learning_rate": 1.263417866676382e-06, "loss": 0.1843, "step": 35741 }, { "epoch": 3.59, "learning_rate": 1.2627307631799023e-06, "loss": 0.1104, "step": 35742 }, { "epoch": 3.59, "learning_rate": 1.2620438425564663e-06, "loss": 0.0552, "step": 35743 }, { "epoch": 3.59, "learning_rate": 1.26135710481046e-06, "loss": 0.3597, "step": 35744 }, { "epoch": 3.59, "learning_rate": 1.26067054994624e-06, "loss": 0.4713, "step": 35745 }, { "epoch": 3.59, "learning_rate": 1.2599841779681765e-06, "loss": 0.1079, "step": 35746 }, { "epoch": 3.59, "learning_rate": 1.2592979888806422e-06, "loss": 0.1441, "step": 35747 }, { "epoch": 3.59, "learning_rate": 1.2586119826880071e-06, "loss": 0.0236, "step": 35748 }, { "epoch": 3.59, "learning_rate": 1.2579261593946246e-06, "loss": 0.049, "step": 35749 }, { "epoch": 3.59, "learning_rate": 1.2572405190048708e-06, "loss": 0.1855, "step": 35750 }, { "epoch": 3.59, "learning_rate": 1.2565550615231025e-06, "loss": 0.2683, "step": 35751 }, { "epoch": 3.59, "learning_rate": 1.2558697869536829e-06, "loss": 0.1911, "step": 35752 }, { "epoch": 3.59, "learning_rate": 1.2551846953009715e-06, "loss": 0.9417, "step": 35753 }, { "epoch": 3.59, "learning_rate": 1.2544997865693286e-06, "loss": 0.1914, "step": 35754 }, { "epoch": 3.59, "learning_rate": 1.2538150607631205e-06, "loss": 0.5207, "step": 35755 }, { "epoch": 3.59, "learning_rate": 1.2531305178866935e-06, "loss": 0.0779, "step": 35756 }, { "epoch": 3.59, "learning_rate": 1.2524461579444046e-06, "loss": 0.6564, "step": 35757 }, { "epoch": 3.59, "learning_rate": 1.2517619809406167e-06, "loss": 0.0494, "step": 35758 }, { "epoch": 3.59, "learning_rate": 1.2510779868796828e-06, "loss": 0.0822, "step": 35759 }, { "epoch": 3.59, "learning_rate": 1.2503941757659432e-06, "loss": 0.1798, "step": 35760 }, { "epoch": 3.59, "learning_rate": 1.249710547603764e-06, "loss": 0.0459, "step": 35761 }, { "epoch": 3.59, "learning_rate": 1.2490271023974919e-06, "loss": 0.1264, "step": 35762 }, { "epoch": 3.59, "learning_rate": 1.2483438401514634e-06, "loss": 0.1075, "step": 35763 }, { "epoch": 3.59, "learning_rate": 1.2476607608700452e-06, "loss": 0.2385, "step": 35764 }, { "epoch": 3.59, "learning_rate": 1.2469778645575735e-06, "loss": 0.0845, "step": 35765 }, { "epoch": 3.59, "learning_rate": 1.2462951512183985e-06, "loss": 0.3254, "step": 35766 }, { "epoch": 3.59, "learning_rate": 1.245612620856863e-06, "loss": 0.3646, "step": 35767 }, { "epoch": 3.6, "learning_rate": 1.2449302734773073e-06, "loss": 0.0345, "step": 35768 }, { "epoch": 3.6, "learning_rate": 1.2442481090840773e-06, "loss": 0.2069, "step": 35769 }, { "epoch": 3.6, "learning_rate": 1.2435661276815135e-06, "loss": 0.3713, "step": 35770 }, { "epoch": 3.6, "learning_rate": 1.2428843292739557e-06, "loss": 0.3007, "step": 35771 }, { "epoch": 3.6, "learning_rate": 1.2422027138657398e-06, "loss": 0.1332, "step": 35772 }, { "epoch": 3.6, "learning_rate": 1.2415212814612131e-06, "loss": 0.2098, "step": 35773 }, { "epoch": 3.6, "learning_rate": 1.2408400320646985e-06, "loss": 0.0909, "step": 35774 }, { "epoch": 3.6, "learning_rate": 1.2401589656805356e-06, "loss": 0.4479, "step": 35775 }, { "epoch": 3.6, "learning_rate": 1.2394780823130647e-06, "loss": 0.0734, "step": 35776 }, { "epoch": 3.6, "learning_rate": 1.2387973819666088e-06, "loss": 0.035, "step": 35777 }, { "epoch": 3.6, "learning_rate": 1.238116864645511e-06, "loss": 0.2828, "step": 35778 }, { "epoch": 3.6, "learning_rate": 1.2374365303540947e-06, "loss": 0.1441, "step": 35779 }, { "epoch": 3.6, "learning_rate": 1.2367563790966897e-06, "loss": 0.248, "step": 35780 }, { "epoch": 3.6, "learning_rate": 1.2360764108776224e-06, "loss": 0.1959, "step": 35781 }, { "epoch": 3.6, "learning_rate": 1.2353966257012262e-06, "loss": 0.1981, "step": 35782 }, { "epoch": 3.6, "learning_rate": 1.2347170235718208e-06, "loss": 0.2489, "step": 35783 }, { "epoch": 3.6, "learning_rate": 1.2340376044937363e-06, "loss": 0.0976, "step": 35784 }, { "epoch": 3.6, "learning_rate": 1.2333583684712924e-06, "loss": 0.1662, "step": 35785 }, { "epoch": 3.6, "learning_rate": 1.2326793155088056e-06, "loss": 0.3169, "step": 35786 }, { "epoch": 3.6, "learning_rate": 1.232000445610616e-06, "loss": 0.4074, "step": 35787 }, { "epoch": 3.6, "learning_rate": 1.2313217587810234e-06, "loss": 0.3157, "step": 35788 }, { "epoch": 3.6, "learning_rate": 1.2306432550243541e-06, "loss": 0.1146, "step": 35789 }, { "epoch": 3.6, "learning_rate": 1.2299649343449314e-06, "loss": 0.0547, "step": 35790 }, { "epoch": 3.6, "learning_rate": 1.2292867967470689e-06, "loss": 0.1909, "step": 35791 }, { "epoch": 3.6, "learning_rate": 1.2286088422350727e-06, "loss": 0.2713, "step": 35792 }, { "epoch": 3.6, "learning_rate": 1.2279310708132696e-06, "loss": 0.2865, "step": 35793 }, { "epoch": 3.6, "learning_rate": 1.2272534824859728e-06, "loss": 0.1864, "step": 35794 }, { "epoch": 3.6, "learning_rate": 1.2265760772574785e-06, "loss": 0.4661, "step": 35795 }, { "epoch": 3.6, "learning_rate": 1.2258988551321138e-06, "loss": 0.0675, "step": 35796 }, { "epoch": 3.6, "learning_rate": 1.2252218161141814e-06, "loss": 0.0792, "step": 35797 }, { "epoch": 3.6, "learning_rate": 1.2245449602079916e-06, "loss": 0.172, "step": 35798 }, { "epoch": 3.6, "learning_rate": 1.2238682874178508e-06, "loss": 0.226, "step": 35799 }, { "epoch": 3.6, "learning_rate": 1.2231917977480622e-06, "loss": 0.3143, "step": 35800 }, { "epoch": 3.6, "learning_rate": 1.222515491202939e-06, "loss": 0.5201, "step": 35801 }, { "epoch": 3.6, "learning_rate": 1.2218393677867779e-06, "loss": 0.325, "step": 35802 }, { "epoch": 3.6, "learning_rate": 1.221163427503882e-06, "loss": 0.3078, "step": 35803 }, { "epoch": 3.6, "learning_rate": 1.2204876703585511e-06, "loss": 0.0411, "step": 35804 }, { "epoch": 3.6, "learning_rate": 1.2198120963550953e-06, "loss": 0.0713, "step": 35805 }, { "epoch": 3.6, "learning_rate": 1.2191367054978008e-06, "loss": 0.0852, "step": 35806 }, { "epoch": 3.6, "learning_rate": 1.2184614977909713e-06, "loss": 0.0894, "step": 35807 }, { "epoch": 3.6, "learning_rate": 1.2177864732389132e-06, "loss": 0.2312, "step": 35808 }, { "epoch": 3.6, "learning_rate": 1.2171116318458997e-06, "loss": 0.1899, "step": 35809 }, { "epoch": 3.6, "learning_rate": 1.2164369736162473e-06, "loss": 0.045, "step": 35810 }, { "epoch": 3.6, "learning_rate": 1.2157624985542359e-06, "loss": 0.03, "step": 35811 }, { "epoch": 3.6, "learning_rate": 1.2150882066641621e-06, "loss": 0.0915, "step": 35812 }, { "epoch": 3.6, "learning_rate": 1.2144140979503193e-06, "loss": 0.1835, "step": 35813 }, { "epoch": 3.6, "learning_rate": 1.2137401724169938e-06, "loss": 0.367, "step": 35814 }, { "epoch": 3.6, "learning_rate": 1.2130664300684757e-06, "loss": 0.4191, "step": 35815 }, { "epoch": 3.6, "learning_rate": 1.2123928709090483e-06, "loss": 0.0486, "step": 35816 }, { "epoch": 3.6, "learning_rate": 1.2117194949430043e-06, "loss": 0.6445, "step": 35817 }, { "epoch": 3.6, "learning_rate": 1.2110463021746209e-06, "loss": 0.0246, "step": 35818 }, { "epoch": 3.6, "learning_rate": 1.2103732926081944e-06, "loss": 0.5743, "step": 35819 }, { "epoch": 3.6, "learning_rate": 1.2097004662479982e-06, "loss": 0.5937, "step": 35820 }, { "epoch": 3.6, "learning_rate": 1.2090278230983088e-06, "loss": 0.0881, "step": 35821 }, { "epoch": 3.6, "learning_rate": 1.2083553631634192e-06, "loss": 0.4852, "step": 35822 }, { "epoch": 3.6, "learning_rate": 1.2076830864476096e-06, "loss": 0.1248, "step": 35823 }, { "epoch": 3.6, "learning_rate": 1.2070109929551399e-06, "loss": 0.1665, "step": 35824 }, { "epoch": 3.6, "learning_rate": 1.2063390826903032e-06, "loss": 0.1321, "step": 35825 }, { "epoch": 3.6, "learning_rate": 1.2056673556573761e-06, "loss": 0.0858, "step": 35826 }, { "epoch": 3.6, "learning_rate": 1.204995811860622e-06, "loss": 0.3223, "step": 35827 }, { "epoch": 3.6, "learning_rate": 1.2043244513043206e-06, "loss": 0.0353, "step": 35828 }, { "epoch": 3.6, "learning_rate": 1.2036532739927453e-06, "loss": 0.1862, "step": 35829 }, { "epoch": 3.6, "learning_rate": 1.202982279930166e-06, "loss": 0.0781, "step": 35830 }, { "epoch": 3.6, "learning_rate": 1.2023114691208524e-06, "loss": 0.1153, "step": 35831 }, { "epoch": 3.6, "learning_rate": 1.2016408415690716e-06, "loss": 0.191, "step": 35832 }, { "epoch": 3.6, "learning_rate": 1.2009703972790998e-06, "loss": 0.2675, "step": 35833 }, { "epoch": 3.6, "learning_rate": 1.20030013625519e-06, "loss": 0.2314, "step": 35834 }, { "epoch": 3.6, "learning_rate": 1.1996300585016128e-06, "loss": 0.4776, "step": 35835 }, { "epoch": 3.6, "learning_rate": 1.1989601640226378e-06, "loss": 0.0892, "step": 35836 }, { "epoch": 3.6, "learning_rate": 1.198290452822528e-06, "loss": 0.1955, "step": 35837 }, { "epoch": 3.6, "learning_rate": 1.1976209249055304e-06, "loss": 0.0639, "step": 35838 }, { "epoch": 3.6, "learning_rate": 1.1969515802759246e-06, "loss": 0.2685, "step": 35839 }, { "epoch": 3.6, "learning_rate": 1.1962824189379607e-06, "loss": 0.4651, "step": 35840 }, { "epoch": 3.6, "learning_rate": 1.1956134408958952e-06, "loss": 0.3532, "step": 35841 }, { "epoch": 3.6, "learning_rate": 1.1949446461539882e-06, "loss": 0.2591, "step": 35842 }, { "epoch": 3.6, "learning_rate": 1.194276034716496e-06, "loss": 0.1928, "step": 35843 }, { "epoch": 3.6, "learning_rate": 1.1936076065876723e-06, "loss": 0.2402, "step": 35844 }, { "epoch": 3.6, "learning_rate": 1.1929393617717732e-06, "loss": 0.5721, "step": 35845 }, { "epoch": 3.6, "learning_rate": 1.192271300273049e-06, "loss": 0.1288, "step": 35846 }, { "epoch": 3.6, "learning_rate": 1.1916034220957495e-06, "loss": 0.4582, "step": 35847 }, { "epoch": 3.6, "learning_rate": 1.1909357272441278e-06, "loss": 0.652, "step": 35848 }, { "epoch": 3.6, "learning_rate": 1.1902682157224309e-06, "loss": 0.0635, "step": 35849 }, { "epoch": 3.6, "learning_rate": 1.189600887534905e-06, "loss": 0.3286, "step": 35850 }, { "epoch": 3.6, "learning_rate": 1.1889337426858072e-06, "loss": 0.1597, "step": 35851 }, { "epoch": 3.6, "learning_rate": 1.1882667811793669e-06, "loss": 0.1266, "step": 35852 }, { "epoch": 3.6, "learning_rate": 1.1876000030198375e-06, "loss": 0.4833, "step": 35853 }, { "epoch": 3.6, "learning_rate": 1.1869334082114625e-06, "loss": 0.0825, "step": 35854 }, { "epoch": 3.6, "learning_rate": 1.1862669967584849e-06, "loss": 0.2597, "step": 35855 }, { "epoch": 3.6, "learning_rate": 1.185600768665135e-06, "loss": 0.1184, "step": 35856 }, { "epoch": 3.6, "learning_rate": 1.1849347239356656e-06, "loss": 0.5389, "step": 35857 }, { "epoch": 3.6, "learning_rate": 1.1842688625743103e-06, "loss": 0.5196, "step": 35858 }, { "epoch": 3.6, "learning_rate": 1.1836031845853024e-06, "loss": 0.1839, "step": 35859 }, { "epoch": 3.6, "learning_rate": 1.182937689972885e-06, "loss": 0.0355, "step": 35860 }, { "epoch": 3.6, "learning_rate": 1.1822723787412881e-06, "loss": 0.0708, "step": 35861 }, { "epoch": 3.6, "learning_rate": 1.1816072508947485e-06, "loss": 0.68, "step": 35862 }, { "epoch": 3.6, "learning_rate": 1.1809423064374958e-06, "loss": 0.5273, "step": 35863 }, { "epoch": 3.6, "learning_rate": 1.1802775453737602e-06, "loss": 0.0458, "step": 35864 }, { "epoch": 3.6, "learning_rate": 1.1796129677077816e-06, "loss": 0.2233, "step": 35865 }, { "epoch": 3.6, "learning_rate": 1.1789485734437766e-06, "loss": 0.3734, "step": 35866 }, { "epoch": 3.61, "learning_rate": 1.1782843625859785e-06, "loss": 0.0674, "step": 35867 }, { "epoch": 3.61, "learning_rate": 1.1776203351386173e-06, "loss": 0.0986, "step": 35868 }, { "epoch": 3.61, "learning_rate": 1.1769564911059195e-06, "loss": 0.0917, "step": 35869 }, { "epoch": 3.61, "learning_rate": 1.1762928304920951e-06, "loss": 0.3524, "step": 35870 }, { "epoch": 3.61, "learning_rate": 1.175629353301384e-06, "loss": 0.0585, "step": 35871 }, { "epoch": 3.61, "learning_rate": 1.1749660595380063e-06, "loss": 0.0504, "step": 35872 }, { "epoch": 3.61, "learning_rate": 1.1743029492061718e-06, "loss": 0.5671, "step": 35873 }, { "epoch": 3.61, "learning_rate": 1.1736400223101106e-06, "loss": 0.2895, "step": 35874 }, { "epoch": 3.61, "learning_rate": 1.1729772788540394e-06, "loss": 0.1362, "step": 35875 }, { "epoch": 3.61, "learning_rate": 1.1723147188421746e-06, "loss": 0.0741, "step": 35876 }, { "epoch": 3.61, "learning_rate": 1.1716523422787296e-06, "loss": 0.0806, "step": 35877 }, { "epoch": 3.61, "learning_rate": 1.1709901491679243e-06, "loss": 0.2107, "step": 35878 }, { "epoch": 3.61, "learning_rate": 1.1703281395139687e-06, "loss": 0.4932, "step": 35879 }, { "epoch": 3.61, "learning_rate": 1.1696663133210794e-06, "loss": 0.239, "step": 35880 }, { "epoch": 3.61, "learning_rate": 1.1690046705934666e-06, "loss": 0.1348, "step": 35881 }, { "epoch": 3.61, "learning_rate": 1.1683432113353332e-06, "loss": 0.1789, "step": 35882 }, { "epoch": 3.61, "learning_rate": 1.167681935550906e-06, "loss": 0.4624, "step": 35883 }, { "epoch": 3.61, "learning_rate": 1.1670208432443784e-06, "loss": 0.0824, "step": 35884 }, { "epoch": 3.61, "learning_rate": 1.1663599344199571e-06, "loss": 0.1534, "step": 35885 }, { "epoch": 3.61, "learning_rate": 1.1656992090818553e-06, "loss": 0.0649, "step": 35886 }, { "epoch": 3.61, "learning_rate": 1.165038667234276e-06, "loss": 0.1082, "step": 35887 }, { "epoch": 3.61, "learning_rate": 1.164378308881423e-06, "loss": 0.0981, "step": 35888 }, { "epoch": 3.61, "learning_rate": 1.1637181340274927e-06, "loss": 0.2213, "step": 35889 }, { "epoch": 3.61, "learning_rate": 1.1630581426766918e-06, "loss": 0.1457, "step": 35890 }, { "epoch": 3.61, "learning_rate": 1.1623983348332201e-06, "loss": 0.4428, "step": 35891 }, { "epoch": 3.61, "learning_rate": 1.1617387105012744e-06, "loss": 0.2062, "step": 35892 }, { "epoch": 3.61, "learning_rate": 1.1610792696850513e-06, "loss": 0.4216, "step": 35893 }, { "epoch": 3.61, "learning_rate": 1.160420012388751e-06, "loss": 0.1072, "step": 35894 }, { "epoch": 3.61, "learning_rate": 1.1597609386165664e-06, "loss": 0.1232, "step": 35895 }, { "epoch": 3.61, "learning_rate": 1.1591020483726878e-06, "loss": 0.24, "step": 35896 }, { "epoch": 3.61, "learning_rate": 1.1584433416613182e-06, "loss": 0.4601, "step": 35897 }, { "epoch": 3.61, "learning_rate": 1.1577848184866413e-06, "loss": 0.1435, "step": 35898 }, { "epoch": 3.61, "learning_rate": 1.1571264788528468e-06, "loss": 0.0504, "step": 35899 }, { "epoch": 3.61, "learning_rate": 1.1564683227641283e-06, "loss": 0.6079, "step": 35900 }, { "epoch": 3.61, "learning_rate": 1.1558103502246754e-06, "loss": 0.0192, "step": 35901 }, { "epoch": 3.61, "learning_rate": 1.1551525612386682e-06, "loss": 0.1783, "step": 35902 }, { "epoch": 3.61, "learning_rate": 1.1544949558103002e-06, "loss": 0.4039, "step": 35903 }, { "epoch": 3.61, "learning_rate": 1.1538375339437545e-06, "loss": 0.0555, "step": 35904 }, { "epoch": 3.61, "learning_rate": 1.153180295643208e-06, "loss": 0.2181, "step": 35905 }, { "epoch": 3.61, "learning_rate": 1.1525232409128506e-06, "loss": 0.0727, "step": 35906 }, { "epoch": 3.61, "learning_rate": 1.151866369756862e-06, "loss": 0.2708, "step": 35907 }, { "epoch": 3.61, "learning_rate": 1.1512096821794226e-06, "loss": 0.2067, "step": 35908 }, { "epoch": 3.61, "learning_rate": 1.150553178184709e-06, "loss": 0.3649, "step": 35909 }, { "epoch": 3.61, "learning_rate": 1.1498968577768975e-06, "loss": 0.1329, "step": 35910 }, { "epoch": 3.61, "learning_rate": 1.1492407209601684e-06, "loss": 0.0798, "step": 35911 }, { "epoch": 3.61, "learning_rate": 1.1485847677386985e-06, "loss": 0.2232, "step": 35912 }, { "epoch": 3.61, "learning_rate": 1.147928998116654e-06, "loss": 0.1438, "step": 35913 }, { "epoch": 3.61, "learning_rate": 1.1472734120982187e-06, "loss": 0.3418, "step": 35914 }, { "epoch": 3.61, "learning_rate": 1.1466180096875622e-06, "loss": 0.1879, "step": 35915 }, { "epoch": 3.61, "learning_rate": 1.1459627908888415e-06, "loss": 0.0994, "step": 35916 }, { "epoch": 3.61, "learning_rate": 1.1453077557062463e-06, "loss": 0.1062, "step": 35917 }, { "epoch": 3.61, "learning_rate": 1.1446529041439302e-06, "loss": 0.216, "step": 35918 }, { "epoch": 3.61, "learning_rate": 1.1439982362060697e-06, "loss": 0.1607, "step": 35919 }, { "epoch": 3.61, "learning_rate": 1.1433437518968248e-06, "loss": 0.0462, "step": 35920 }, { "epoch": 3.61, "learning_rate": 1.1426894512203656e-06, "loss": 0.3097, "step": 35921 }, { "epoch": 3.61, "learning_rate": 1.142035334180852e-06, "loss": 0.3441, "step": 35922 }, { "epoch": 3.61, "learning_rate": 1.1413814007824442e-06, "loss": 0.2479, "step": 35923 }, { "epoch": 3.61, "learning_rate": 1.1407276510293086e-06, "loss": 0.1679, "step": 35924 }, { "epoch": 3.61, "learning_rate": 1.1400740849256053e-06, "loss": 0.315, "step": 35925 }, { "epoch": 3.61, "learning_rate": 1.139420702475491e-06, "loss": 0.174, "step": 35926 }, { "epoch": 3.61, "learning_rate": 1.1387675036831225e-06, "loss": 0.0576, "step": 35927 }, { "epoch": 3.61, "learning_rate": 1.1381144885526562e-06, "loss": 0.1691, "step": 35928 }, { "epoch": 3.61, "learning_rate": 1.1374616570882556e-06, "loss": 0.3175, "step": 35929 }, { "epoch": 3.61, "learning_rate": 1.1368090092940674e-06, "loss": 0.0446, "step": 35930 }, { "epoch": 3.61, "learning_rate": 1.1361565451742417e-06, "loss": 0.0176, "step": 35931 }, { "epoch": 3.61, "learning_rate": 1.1355042647329384e-06, "loss": 0.3833, "step": 35932 }, { "epoch": 3.61, "learning_rate": 1.1348521679743106e-06, "loss": 0.322, "step": 35933 }, { "epoch": 3.61, "learning_rate": 1.1342002549024956e-06, "loss": 0.0633, "step": 35934 }, { "epoch": 3.61, "learning_rate": 1.1335485255216492e-06, "loss": 0.0804, "step": 35935 }, { "epoch": 3.61, "learning_rate": 1.1328969798359257e-06, "loss": 0.1662, "step": 35936 }, { "epoch": 3.61, "learning_rate": 1.1322456178494545e-06, "loss": 0.4262, "step": 35937 }, { "epoch": 3.61, "learning_rate": 1.1315944395663958e-06, "loss": 0.3494, "step": 35938 }, { "epoch": 3.61, "learning_rate": 1.1309434449908895e-06, "loss": 0.1615, "step": 35939 }, { "epoch": 3.61, "learning_rate": 1.1302926341270725e-06, "loss": 0.2806, "step": 35940 }, { "epoch": 3.61, "learning_rate": 1.1296420069790948e-06, "loss": 0.0687, "step": 35941 }, { "epoch": 3.61, "learning_rate": 1.128991563551086e-06, "loss": 0.1769, "step": 35942 }, { "epoch": 3.61, "learning_rate": 1.1283413038472e-06, "loss": 0.2933, "step": 35943 }, { "epoch": 3.61, "learning_rate": 1.1276912278715635e-06, "loss": 0.176, "step": 35944 }, { "epoch": 3.61, "learning_rate": 1.1270413356283126e-06, "loss": 0.5178, "step": 35945 }, { "epoch": 3.61, "learning_rate": 1.1263916271215913e-06, "loss": 0.1188, "step": 35946 }, { "epoch": 3.61, "learning_rate": 1.1257421023555325e-06, "loss": 0.0841, "step": 35947 }, { "epoch": 3.61, "learning_rate": 1.1250927613342565e-06, "loss": 0.0764, "step": 35948 }, { "epoch": 3.61, "learning_rate": 1.1244436040619132e-06, "loss": 0.106, "step": 35949 }, { "epoch": 3.61, "learning_rate": 1.1237946305426262e-06, "loss": 0.3296, "step": 35950 }, { "epoch": 3.61, "learning_rate": 1.1231458407805217e-06, "loss": 0.1786, "step": 35951 }, { "epoch": 3.61, "learning_rate": 1.1224972347797336e-06, "loss": 0.0679, "step": 35952 }, { "epoch": 3.61, "learning_rate": 1.1218488125443881e-06, "loss": 0.0459, "step": 35953 }, { "epoch": 3.61, "learning_rate": 1.121200574078609e-06, "loss": 0.108, "step": 35954 }, { "epoch": 3.61, "learning_rate": 1.1205525193865263e-06, "loss": 0.078, "step": 35955 }, { "epoch": 3.61, "learning_rate": 1.1199046484722596e-06, "loss": 0.2065, "step": 35956 }, { "epoch": 3.61, "learning_rate": 1.1192569613399328e-06, "loss": 0.1378, "step": 35957 }, { "epoch": 3.61, "learning_rate": 1.118609457993669e-06, "loss": 0.6164, "step": 35958 }, { "epoch": 3.61, "learning_rate": 1.1179621384375849e-06, "loss": 0.1232, "step": 35959 }, { "epoch": 3.61, "learning_rate": 1.1173150026758007e-06, "loss": 0.6872, "step": 35960 }, { "epoch": 3.61, "learning_rate": 1.1166680507124428e-06, "loss": 0.4526, "step": 35961 }, { "epoch": 3.61, "learning_rate": 1.116021282551618e-06, "loss": 0.0548, "step": 35962 }, { "epoch": 3.61, "learning_rate": 1.1153746981974432e-06, "loss": 0.097, "step": 35963 }, { "epoch": 3.61, "learning_rate": 1.1147282976540385e-06, "loss": 0.2142, "step": 35964 }, { "epoch": 3.62, "learning_rate": 1.1140820809255204e-06, "loss": 0.0888, "step": 35965 }, { "epoch": 3.62, "learning_rate": 1.1134360480159855e-06, "loss": 0.1638, "step": 35966 }, { "epoch": 3.62, "learning_rate": 1.1127901989295574e-06, "loss": 0.1122, "step": 35967 }, { "epoch": 3.62, "learning_rate": 1.112144533670346e-06, "loss": 0.1792, "step": 35968 }, { "epoch": 3.62, "learning_rate": 1.1114990522424552e-06, "loss": 0.1038, "step": 35969 }, { "epoch": 3.62, "learning_rate": 1.1108537546499942e-06, "loss": 0.1833, "step": 35970 }, { "epoch": 3.62, "learning_rate": 1.1102086408970701e-06, "loss": 0.1124, "step": 35971 }, { "epoch": 3.62, "learning_rate": 1.10956371098779e-06, "loss": 0.1048, "step": 35972 }, { "epoch": 3.62, "learning_rate": 1.1089189649262532e-06, "loss": 0.3676, "step": 35973 }, { "epoch": 3.62, "learning_rate": 1.1082744027165638e-06, "loss": 0.1695, "step": 35974 }, { "epoch": 3.62, "learning_rate": 1.1076300243628313e-06, "loss": 0.2181, "step": 35975 }, { "epoch": 3.62, "learning_rate": 1.106985829869146e-06, "loss": 0.0461, "step": 35976 }, { "epoch": 3.62, "learning_rate": 1.1063418192396113e-06, "loss": 0.213, "step": 35977 }, { "epoch": 3.62, "learning_rate": 1.105697992478324e-06, "loss": 0.3936, "step": 35978 }, { "epoch": 3.62, "learning_rate": 1.1050543495893906e-06, "loss": 0.1697, "step": 35979 }, { "epoch": 3.62, "learning_rate": 1.1044108905768914e-06, "loss": 0.2289, "step": 35980 }, { "epoch": 3.62, "learning_rate": 1.1037676154449295e-06, "loss": 0.1715, "step": 35981 }, { "epoch": 3.62, "learning_rate": 1.1031245241975985e-06, "loss": 0.031, "step": 35982 }, { "epoch": 3.62, "learning_rate": 1.1024816168389917e-06, "loss": 0.0565, "step": 35983 }, { "epoch": 3.62, "learning_rate": 1.1018388933731993e-06, "loss": 0.2949, "step": 35984 }, { "epoch": 3.62, "learning_rate": 1.1011963538043113e-06, "loss": 0.1285, "step": 35985 }, { "epoch": 3.62, "learning_rate": 1.1005539981364143e-06, "loss": 0.0905, "step": 35986 }, { "epoch": 3.62, "learning_rate": 1.0999118263735953e-06, "loss": 0.2236, "step": 35987 }, { "epoch": 3.62, "learning_rate": 1.0992698385199474e-06, "loss": 0.0496, "step": 35988 }, { "epoch": 3.62, "learning_rate": 1.0986280345795507e-06, "loss": 0.1422, "step": 35989 }, { "epoch": 3.62, "learning_rate": 1.0979864145564888e-06, "loss": 0.5363, "step": 35990 }, { "epoch": 3.62, "learning_rate": 1.0973449784548484e-06, "loss": 0.0552, "step": 35991 }, { "epoch": 3.62, "learning_rate": 1.096703726278706e-06, "loss": 0.279, "step": 35992 }, { "epoch": 3.62, "learning_rate": 1.096062658032152e-06, "loss": 0.1089, "step": 35993 }, { "epoch": 3.62, "learning_rate": 1.095421773719253e-06, "loss": 0.3376, "step": 35994 }, { "epoch": 3.62, "learning_rate": 1.0947810733440954e-06, "loss": 0.0702, "step": 35995 }, { "epoch": 3.62, "learning_rate": 1.0941405569107565e-06, "loss": 0.1275, "step": 35996 }, { "epoch": 3.62, "learning_rate": 1.0935002244233127e-06, "loss": 0.0958, "step": 35997 }, { "epoch": 3.62, "learning_rate": 1.0928600758858343e-06, "loss": 0.0713, "step": 35998 }, { "epoch": 3.62, "learning_rate": 1.0922201113023976e-06, "loss": 0.1809, "step": 35999 }, { "epoch": 3.62, "learning_rate": 1.0915803306770767e-06, "loss": 0.0861, "step": 36000 }, { "epoch": 3.62, "learning_rate": 1.090940734013941e-06, "loss": 0.5308, "step": 36001 }, { "epoch": 3.62, "learning_rate": 1.0903013213170576e-06, "loss": 0.0836, "step": 36002 }, { "epoch": 3.62, "learning_rate": 1.0896620925905031e-06, "loss": 0.1471, "step": 36003 }, { "epoch": 3.62, "learning_rate": 1.0890230478383379e-06, "loss": 0.3459, "step": 36004 }, { "epoch": 3.62, "learning_rate": 1.088384187064635e-06, "loss": 0.0988, "step": 36005 }, { "epoch": 3.62, "learning_rate": 1.0877455102734514e-06, "loss": 0.1312, "step": 36006 }, { "epoch": 3.62, "learning_rate": 1.0871070174688601e-06, "loss": 0.209, "step": 36007 }, { "epoch": 3.62, "learning_rate": 1.086468708654922e-06, "loss": 0.0424, "step": 36008 }, { "epoch": 3.62, "learning_rate": 1.085830583835693e-06, "loss": 0.0756, "step": 36009 }, { "epoch": 3.62, "learning_rate": 1.0851926430152403e-06, "loss": 0.5488, "step": 36010 }, { "epoch": 3.62, "learning_rate": 1.0845548861976274e-06, "loss": 0.1308, "step": 36011 }, { "epoch": 3.62, "learning_rate": 1.0839173133869006e-06, "loss": 0.1206, "step": 36012 }, { "epoch": 3.62, "learning_rate": 1.0832799245871271e-06, "loss": 0.1142, "step": 36013 }, { "epoch": 3.62, "learning_rate": 1.082642719802357e-06, "loss": 0.2932, "step": 36014 }, { "epoch": 3.62, "learning_rate": 1.08200569903665e-06, "loss": 0.5302, "step": 36015 }, { "epoch": 3.62, "learning_rate": 1.0813688622940598e-06, "loss": 0.0733, "step": 36016 }, { "epoch": 3.62, "learning_rate": 1.0807322095786331e-06, "loss": 0.0416, "step": 36017 }, { "epoch": 3.62, "learning_rate": 1.0800957408944268e-06, "loss": 0.0458, "step": 36018 }, { "epoch": 3.62, "learning_rate": 1.0794594562454908e-06, "loss": 0.1826, "step": 36019 }, { "epoch": 3.62, "learning_rate": 1.0788233556358683e-06, "loss": 0.2229, "step": 36020 }, { "epoch": 3.62, "learning_rate": 1.07818743906962e-06, "loss": 0.0872, "step": 36021 }, { "epoch": 3.62, "learning_rate": 1.077551706550779e-06, "loss": 0.2132, "step": 36022 }, { "epoch": 3.62, "learning_rate": 1.0769161580833918e-06, "loss": 0.8773, "step": 36023 }, { "epoch": 3.62, "learning_rate": 1.0762807936715125e-06, "loss": 0.3779, "step": 36024 }, { "epoch": 3.62, "learning_rate": 1.075645613319184e-06, "loss": 0.1034, "step": 36025 }, { "epoch": 3.62, "learning_rate": 1.075010617030433e-06, "loss": 0.0855, "step": 36026 }, { "epoch": 3.62, "learning_rate": 1.0743758048093166e-06, "loss": 0.1526, "step": 36027 }, { "epoch": 3.62, "learning_rate": 1.0737411766598682e-06, "loss": 0.2913, "step": 36028 }, { "epoch": 3.62, "learning_rate": 1.0731067325861278e-06, "loss": 0.1876, "step": 36029 }, { "epoch": 3.62, "learning_rate": 1.072472472592132e-06, "loss": 0.214, "step": 36030 }, { "epoch": 3.62, "learning_rate": 1.0718383966819145e-06, "loss": 0.1005, "step": 36031 }, { "epoch": 3.62, "learning_rate": 1.0712045048595121e-06, "loss": 0.1359, "step": 36032 }, { "epoch": 3.62, "learning_rate": 1.0705707971289614e-06, "loss": 0.1819, "step": 36033 }, { "epoch": 3.62, "learning_rate": 1.0699372734942926e-06, "loss": 0.1283, "step": 36034 }, { "epoch": 3.62, "learning_rate": 1.0693039339595389e-06, "loss": 0.0421, "step": 36035 }, { "epoch": 3.62, "learning_rate": 1.0686707785287276e-06, "loss": 0.1836, "step": 36036 }, { "epoch": 3.62, "learning_rate": 1.0680378072058883e-06, "loss": 0.1538, "step": 36037 }, { "epoch": 3.62, "learning_rate": 1.067405019995048e-06, "loss": 0.0753, "step": 36038 }, { "epoch": 3.62, "learning_rate": 1.0667724169002402e-06, "loss": 0.0433, "step": 36039 }, { "epoch": 3.62, "learning_rate": 1.066139997925485e-06, "loss": 0.3931, "step": 36040 }, { "epoch": 3.62, "learning_rate": 1.0655077630748056e-06, "loss": 0.1241, "step": 36041 }, { "epoch": 3.62, "learning_rate": 1.064875712352229e-06, "loss": 0.3142, "step": 36042 }, { "epoch": 3.62, "learning_rate": 1.0642438457617787e-06, "loss": 0.1276, "step": 36043 }, { "epoch": 3.62, "learning_rate": 1.063612163307468e-06, "loss": 0.1375, "step": 36044 }, { "epoch": 3.62, "learning_rate": 1.0629806649933238e-06, "loss": 0.1421, "step": 36045 }, { "epoch": 3.62, "learning_rate": 1.0623493508233595e-06, "loss": 0.1808, "step": 36046 }, { "epoch": 3.62, "learning_rate": 1.0617182208015985e-06, "loss": 0.5107, "step": 36047 }, { "epoch": 3.62, "learning_rate": 1.0610872749320544e-06, "loss": 0.0554, "step": 36048 }, { "epoch": 3.62, "learning_rate": 1.0604565132187439e-06, "loss": 0.1485, "step": 36049 }, { "epoch": 3.62, "learning_rate": 1.0598259356656736e-06, "loss": 0.2338, "step": 36050 }, { "epoch": 3.62, "learning_rate": 1.0591955422768674e-06, "loss": 0.2273, "step": 36051 }, { "epoch": 3.62, "learning_rate": 1.058565333056325e-06, "loss": 0.0771, "step": 36052 }, { "epoch": 3.62, "learning_rate": 1.05793530800807e-06, "loss": 0.2218, "step": 36053 }, { "epoch": 3.62, "learning_rate": 1.0573054671360994e-06, "loss": 0.1674, "step": 36054 }, { "epoch": 3.62, "learning_rate": 1.056675810444423e-06, "loss": 0.0355, "step": 36055 }, { "epoch": 3.62, "learning_rate": 1.0560463379370577e-06, "loss": 0.0026, "step": 36056 }, { "epoch": 3.62, "learning_rate": 1.0554170496180037e-06, "loss": 0.3817, "step": 36057 }, { "epoch": 3.62, "learning_rate": 1.054787945491258e-06, "loss": 0.0925, "step": 36058 }, { "epoch": 3.62, "learning_rate": 1.054159025560837e-06, "loss": 0.172, "step": 36059 }, { "epoch": 3.62, "learning_rate": 1.0535302898307308e-06, "loss": 0.2967, "step": 36060 }, { "epoch": 3.62, "learning_rate": 1.05290173830495e-06, "loss": 0.6004, "step": 36061 }, { "epoch": 3.62, "learning_rate": 1.052273370987491e-06, "loss": 0.1586, "step": 36062 }, { "epoch": 3.62, "learning_rate": 1.0516451878823508e-06, "loss": 0.4666, "step": 36063 }, { "epoch": 3.63, "learning_rate": 1.051017188993526e-06, "loss": 0.2651, "step": 36064 }, { "epoch": 3.63, "learning_rate": 1.050389374325017e-06, "loss": 0.7272, "step": 36065 }, { "epoch": 3.63, "learning_rate": 1.0497617438808172e-06, "loss": 0.1448, "step": 36066 }, { "epoch": 3.63, "learning_rate": 1.0491342976649199e-06, "loss": 0.462, "step": 36067 }, { "epoch": 3.63, "learning_rate": 1.0485070356813186e-06, "loss": 0.5688, "step": 36068 }, { "epoch": 3.63, "learning_rate": 1.0478799579340037e-06, "loss": 0.0502, "step": 36069 }, { "epoch": 3.63, "learning_rate": 1.047253064426965e-06, "loss": 0.3747, "step": 36070 }, { "epoch": 3.63, "learning_rate": 1.0466263551641997e-06, "loss": 0.1326, "step": 36071 }, { "epoch": 3.63, "learning_rate": 1.0459998301496875e-06, "loss": 0.1863, "step": 36072 }, { "epoch": 3.63, "learning_rate": 1.0453734893874124e-06, "loss": 0.0706, "step": 36073 }, { "epoch": 3.63, "learning_rate": 1.0447473328813705e-06, "loss": 0.093, "step": 36074 }, { "epoch": 3.63, "learning_rate": 1.0441213606355427e-06, "loss": 0.1588, "step": 36075 }, { "epoch": 3.63, "learning_rate": 1.0434955726539087e-06, "loss": 0.3629, "step": 36076 }, { "epoch": 3.63, "learning_rate": 1.0428699689404553e-06, "loss": 0.1448, "step": 36077 }, { "epoch": 3.63, "learning_rate": 1.0422445494991596e-06, "loss": 0.0894, "step": 36078 }, { "epoch": 3.63, "learning_rate": 1.0416193143340047e-06, "loss": 0.1642, "step": 36079 }, { "epoch": 3.63, "learning_rate": 1.0409942634489678e-06, "loss": 0.2343, "step": 36080 }, { "epoch": 3.63, "learning_rate": 1.0403693968480254e-06, "loss": 0.2926, "step": 36081 }, { "epoch": 3.63, "learning_rate": 1.0397447145351545e-06, "loss": 0.2912, "step": 36082 }, { "epoch": 3.63, "learning_rate": 1.0391202165143353e-06, "loss": 0.0987, "step": 36083 }, { "epoch": 3.63, "learning_rate": 1.0384959027895313e-06, "loss": 0.4183, "step": 36084 }, { "epoch": 3.63, "learning_rate": 1.0378717733647293e-06, "loss": 0.0471, "step": 36085 }, { "epoch": 3.63, "learning_rate": 1.0372478282438924e-06, "loss": 0.1584, "step": 36086 }, { "epoch": 3.63, "learning_rate": 1.0366240674309878e-06, "loss": 0.0966, "step": 36087 }, { "epoch": 3.63, "learning_rate": 1.0360004909299892e-06, "loss": 0.1587, "step": 36088 }, { "epoch": 3.63, "learning_rate": 1.035377098744873e-06, "loss": 0.222, "step": 36089 }, { "epoch": 3.63, "learning_rate": 1.0347538908795896e-06, "loss": 0.2179, "step": 36090 }, { "epoch": 3.63, "learning_rate": 1.0341308673381155e-06, "loss": 0.2258, "step": 36091 }, { "epoch": 3.63, "learning_rate": 1.033508028124418e-06, "loss": 0.1135, "step": 36092 }, { "epoch": 3.63, "learning_rate": 1.0328853732424503e-06, "loss": 0.2336, "step": 36093 }, { "epoch": 3.63, "learning_rate": 1.0322629026961827e-06, "loss": 0.1516, "step": 36094 }, { "epoch": 3.63, "learning_rate": 1.0316406164895753e-06, "loss": 0.4673, "step": 36095 }, { "epoch": 3.63, "learning_rate": 1.0310185146265882e-06, "loss": 0.037, "step": 36096 }, { "epoch": 3.63, "learning_rate": 1.0303965971111751e-06, "loss": 0.1931, "step": 36097 }, { "epoch": 3.63, "learning_rate": 1.0297748639472993e-06, "loss": 0.515, "step": 36098 }, { "epoch": 3.63, "learning_rate": 1.0291533151389142e-06, "loss": 0.4872, "step": 36099 }, { "epoch": 3.63, "learning_rate": 1.028531950689977e-06, "loss": 0.5103, "step": 36100 }, { "epoch": 3.63, "learning_rate": 1.0279107706044444e-06, "loss": 0.2747, "step": 36101 }, { "epoch": 3.63, "learning_rate": 1.0272897748862597e-06, "loss": 0.1235, "step": 36102 }, { "epoch": 3.63, "learning_rate": 1.0266689635393866e-06, "loss": 0.2374, "step": 36103 }, { "epoch": 3.63, "learning_rate": 1.0260483365677653e-06, "loss": 0.1256, "step": 36104 }, { "epoch": 3.63, "learning_rate": 1.0254278939753525e-06, "loss": 0.2466, "step": 36105 }, { "epoch": 3.63, "learning_rate": 1.0248076357660951e-06, "loss": 0.2882, "step": 36106 }, { "epoch": 3.63, "learning_rate": 1.0241875619439366e-06, "loss": 0.3349, "step": 36107 }, { "epoch": 3.63, "learning_rate": 1.0235676725128307e-06, "loss": 0.1005, "step": 36108 }, { "epoch": 3.63, "learning_rate": 1.022947967476714e-06, "loss": 0.0838, "step": 36109 }, { "epoch": 3.63, "learning_rate": 1.02232844683953e-06, "loss": 0.0742, "step": 36110 }, { "epoch": 3.63, "learning_rate": 1.021709110605229e-06, "loss": 0.1004, "step": 36111 }, { "epoch": 3.63, "learning_rate": 1.0210899587777444e-06, "loss": 0.0952, "step": 36112 }, { "epoch": 3.63, "learning_rate": 1.02047099136102e-06, "loss": 0.0989, "step": 36113 }, { "epoch": 3.63, "learning_rate": 1.0198522083589957e-06, "loss": 0.439, "step": 36114 }, { "epoch": 3.63, "learning_rate": 1.019233609775605e-06, "loss": 0.2529, "step": 36115 }, { "epoch": 3.63, "learning_rate": 1.0186151956147816e-06, "loss": 0.3919, "step": 36116 }, { "epoch": 3.63, "learning_rate": 1.0179969658804754e-06, "loss": 0.0391, "step": 36117 }, { "epoch": 3.63, "learning_rate": 1.017378920576607e-06, "loss": 0.0581, "step": 36118 }, { "epoch": 3.63, "learning_rate": 1.016761059707113e-06, "loss": 0.2331, "step": 36119 }, { "epoch": 3.63, "learning_rate": 1.016143383275927e-06, "loss": 0.1901, "step": 36120 }, { "epoch": 3.63, "learning_rate": 1.015525891286979e-06, "loss": 0.6988, "step": 36121 }, { "epoch": 3.63, "learning_rate": 1.0149085837441963e-06, "loss": 0.099, "step": 36122 }, { "epoch": 3.63, "learning_rate": 1.0142914606515086e-06, "loss": 0.2003, "step": 36123 }, { "epoch": 3.63, "learning_rate": 1.0136745220128464e-06, "loss": 0.0746, "step": 36124 }, { "epoch": 3.63, "learning_rate": 1.0130577678321295e-06, "loss": 0.1943, "step": 36125 }, { "epoch": 3.63, "learning_rate": 1.0124411981132853e-06, "loss": 0.1614, "step": 36126 }, { "epoch": 3.63, "learning_rate": 1.0118248128602403e-06, "loss": 0.5038, "step": 36127 }, { "epoch": 3.63, "learning_rate": 1.0112086120769149e-06, "loss": 0.133, "step": 36128 }, { "epoch": 3.63, "learning_rate": 1.0105925957672291e-06, "loss": 0.3598, "step": 36129 }, { "epoch": 3.63, "learning_rate": 1.0099767639350998e-06, "loss": 0.5462, "step": 36130 }, { "epoch": 3.63, "learning_rate": 1.0093611165844574e-06, "loss": 0.1971, "step": 36131 }, { "epoch": 3.63, "learning_rate": 1.0087456537192086e-06, "loss": 0.2178, "step": 36132 }, { "epoch": 3.63, "learning_rate": 1.0081303753432702e-06, "loss": 0.5701, "step": 36133 }, { "epoch": 3.63, "learning_rate": 1.0075152814605658e-06, "loss": 0.3422, "step": 36134 }, { "epoch": 3.63, "learning_rate": 1.0069003720750057e-06, "loss": 0.1675, "step": 36135 }, { "epoch": 3.63, "learning_rate": 1.0062856471904968e-06, "loss": 0.0106, "step": 36136 }, { "epoch": 3.63, "learning_rate": 1.0056711068109593e-06, "loss": 0.3514, "step": 36137 }, { "epoch": 3.63, "learning_rate": 1.005056750940303e-06, "loss": 0.2938, "step": 36138 }, { "epoch": 3.63, "learning_rate": 1.004442579582432e-06, "loss": 0.0078, "step": 36139 }, { "epoch": 3.63, "learning_rate": 1.0038285927412595e-06, "loss": 0.3601, "step": 36140 }, { "epoch": 3.63, "learning_rate": 1.0032147904206923e-06, "loss": 0.0983, "step": 36141 }, { "epoch": 3.63, "learning_rate": 1.0026011726246342e-06, "loss": 0.0632, "step": 36142 }, { "epoch": 3.63, "learning_rate": 1.0019877393569888e-06, "loss": 0.1036, "step": 36143 }, { "epoch": 3.63, "learning_rate": 1.001374490621666e-06, "loss": 0.2386, "step": 36144 }, { "epoch": 3.63, "learning_rate": 1.0007614264225628e-06, "loss": 0.1668, "step": 36145 }, { "epoch": 3.63, "learning_rate": 1.0001485467635795e-06, "loss": 0.1688, "step": 36146 }, { "epoch": 3.63, "learning_rate": 9.99535851648623e-07, "loss": 0.1318, "step": 36147 }, { "epoch": 3.63, "learning_rate": 9.9892334108158e-07, "loss": 0.0664, "step": 36148 }, { "epoch": 3.63, "learning_rate": 9.983110150663677e-07, "loss": 0.0525, "step": 36149 }, { "epoch": 3.63, "learning_rate": 9.976988736068627e-07, "loss": 0.0765, "step": 36150 }, { "epoch": 3.63, "learning_rate": 9.970869167069686e-07, "loss": 0.0659, "step": 36151 }, { "epoch": 3.63, "learning_rate": 9.964751443705855e-07, "loss": 0.0703, "step": 36152 }, { "epoch": 3.63, "learning_rate": 9.958635566016005e-07, "loss": 0.7509, "step": 36153 }, { "epoch": 3.63, "learning_rate": 9.952521534039038e-07, "loss": 0.1297, "step": 36154 }, { "epoch": 3.63, "learning_rate": 9.94640934781389e-07, "loss": 0.5886, "step": 36155 }, { "epoch": 3.63, "learning_rate": 9.94029900737946e-07, "loss": 0.2267, "step": 36156 }, { "epoch": 3.63, "learning_rate": 9.934190512774588e-07, "loss": 0.1413, "step": 36157 }, { "epoch": 3.63, "learning_rate": 9.928083864038206e-07, "loss": 0.0431, "step": 36158 }, { "epoch": 3.63, "learning_rate": 9.921979061209153e-07, "loss": 0.313, "step": 36159 }, { "epoch": 3.63, "learning_rate": 9.915876104326294e-07, "loss": 0.0435, "step": 36160 }, { "epoch": 3.63, "learning_rate": 9.909774993428433e-07, "loss": 0.1894, "step": 36161 }, { "epoch": 3.64, "learning_rate": 9.90367572855434e-07, "loss": 0.3611, "step": 36162 }, { "epoch": 3.64, "learning_rate": 9.897578309743017e-07, "loss": 0.0772, "step": 36163 }, { "epoch": 3.64, "learning_rate": 9.891482737033064e-07, "loss": 0.1678, "step": 36164 }, { "epoch": 3.64, "learning_rate": 9.885389010463354e-07, "loss": 0.1214, "step": 36165 }, { "epoch": 3.64, "learning_rate": 9.879297130072684e-07, "loss": 0.1562, "step": 36166 }, { "epoch": 3.64, "learning_rate": 9.873207095899827e-07, "loss": 0.2983, "step": 36167 }, { "epoch": 3.64, "learning_rate": 9.867118907983452e-07, "loss": 0.2629, "step": 36168 }, { "epoch": 3.64, "learning_rate": 9.86103256636236e-07, "loss": 0.3135, "step": 36169 }, { "epoch": 3.64, "learning_rate": 9.854948071075321e-07, "loss": 0.0576, "step": 36170 }, { "epoch": 3.64, "learning_rate": 9.848865422161e-07, "loss": 0.2774, "step": 36171 }, { "epoch": 3.64, "learning_rate": 9.84278461965814e-07, "loss": 0.2546, "step": 36172 }, { "epoch": 3.64, "learning_rate": 9.836705663605406e-07, "loss": 0.0999, "step": 36173 }, { "epoch": 3.64, "learning_rate": 9.8306285540415e-07, "loss": 0.2273, "step": 36174 }, { "epoch": 3.64, "learning_rate": 9.824553291005057e-07, "loss": 0.0655, "step": 36175 }, { "epoch": 3.64, "learning_rate": 9.818479874534781e-07, "loss": 0.0674, "step": 36176 }, { "epoch": 3.64, "learning_rate": 9.81240830466934e-07, "loss": 0.2139, "step": 36177 }, { "epoch": 3.64, "learning_rate": 9.806338581447306e-07, "loss": 0.195, "step": 36178 }, { "epoch": 3.64, "learning_rate": 9.800270704907343e-07, "loss": 0.3277, "step": 36179 }, { "epoch": 3.64, "learning_rate": 9.794204675088024e-07, "loss": 0.1277, "step": 36180 }, { "epoch": 3.64, "learning_rate": 9.788140492028086e-07, "loss": 0.3411, "step": 36181 }, { "epoch": 3.64, "learning_rate": 9.782078155765962e-07, "loss": 0.2898, "step": 36182 }, { "epoch": 3.64, "learning_rate": 9.776017666340254e-07, "loss": 0.24, "step": 36183 }, { "epoch": 3.64, "learning_rate": 9.769959023789565e-07, "loss": 0.2525, "step": 36184 }, { "epoch": 3.64, "learning_rate": 9.7639022281525e-07, "loss": 0.3172, "step": 36185 }, { "epoch": 3.64, "learning_rate": 9.757847279467525e-07, "loss": 0.0936, "step": 36186 }, { "epoch": 3.64, "learning_rate": 9.75179417777321e-07, "loss": 0.0353, "step": 36187 }, { "epoch": 3.64, "learning_rate": 9.745742923108025e-07, "loss": 0.1138, "step": 36188 }, { "epoch": 3.64, "learning_rate": 9.73969351551054e-07, "loss": 0.1413, "step": 36189 }, { "epoch": 3.64, "learning_rate": 9.733645955019222e-07, "loss": 0.1187, "step": 36190 }, { "epoch": 3.64, "learning_rate": 9.727600241672573e-07, "loss": 0.031, "step": 36191 }, { "epoch": 3.64, "learning_rate": 9.721556375509033e-07, "loss": 0.1508, "step": 36192 }, { "epoch": 3.64, "learning_rate": 9.715514356567102e-07, "loss": 0.1691, "step": 36193 }, { "epoch": 3.64, "learning_rate": 9.709474184885148e-07, "loss": 0.0707, "step": 36194 }, { "epoch": 3.64, "learning_rate": 9.703435860501742e-07, "loss": 0.0777, "step": 36195 }, { "epoch": 3.64, "learning_rate": 9.69739938345522e-07, "loss": 0.2085, "step": 36196 }, { "epoch": 3.64, "learning_rate": 9.691364753783948e-07, "loss": 0.0557, "step": 36197 }, { "epoch": 3.64, "learning_rate": 9.685331971526433e-07, "loss": 0.1177, "step": 36198 }, { "epoch": 3.64, "learning_rate": 9.679301036721078e-07, "loss": 0.5592, "step": 36199 }, { "epoch": 3.64, "learning_rate": 9.673271949406115e-07, "loss": 0.2091, "step": 36200 }, { "epoch": 3.64, "learning_rate": 9.66724470962005e-07, "loss": 0.0283, "step": 36201 }, { "epoch": 3.64, "learning_rate": 9.661219317401181e-07, "loss": 0.0608, "step": 36202 }, { "epoch": 3.64, "learning_rate": 9.655195772787883e-07, "loss": 0.1521, "step": 36203 }, { "epoch": 3.64, "learning_rate": 9.649174075818456e-07, "loss": 0.1637, "step": 36204 }, { "epoch": 3.64, "learning_rate": 9.643154226531237e-07, "loss": 0.14, "step": 36205 }, { "epoch": 3.64, "learning_rate": 9.637136224964526e-07, "loss": 0.0306, "step": 36206 }, { "epoch": 3.64, "learning_rate": 9.631120071156596e-07, "loss": 0.0704, "step": 36207 }, { "epoch": 3.64, "learning_rate": 9.625105765145782e-07, "loss": 0.3234, "step": 36208 }, { "epoch": 3.64, "learning_rate": 9.619093306970317e-07, "loss": 0.0492, "step": 36209 }, { "epoch": 3.64, "learning_rate": 9.613082696668473e-07, "loss": 0.0415, "step": 36210 }, { "epoch": 3.64, "learning_rate": 9.607073934278488e-07, "loss": 0.1673, "step": 36211 }, { "epoch": 3.64, "learning_rate": 9.601067019838628e-07, "loss": 0.5241, "step": 36212 }, { "epoch": 3.64, "learning_rate": 9.59506195338713e-07, "loss": 0.3354, "step": 36213 }, { "epoch": 3.64, "learning_rate": 9.589058734962131e-07, "loss": 0.1612, "step": 36214 }, { "epoch": 3.64, "learning_rate": 9.583057364601933e-07, "loss": 0.1395, "step": 36215 }, { "epoch": 3.64, "learning_rate": 9.577057842344672e-07, "loss": 0.1492, "step": 36216 }, { "epoch": 3.64, "learning_rate": 9.571060168228518e-07, "loss": 0.1613, "step": 36217 }, { "epoch": 3.64, "learning_rate": 9.56506434229164e-07, "loss": 0.0481, "step": 36218 }, { "epoch": 3.64, "learning_rate": 9.55907036457221e-07, "loss": 0.1063, "step": 36219 }, { "epoch": 3.64, "learning_rate": 9.553078235108393e-07, "loss": 0.522, "step": 36220 }, { "epoch": 3.64, "learning_rate": 9.54708795393826e-07, "loss": 0.0295, "step": 36221 }, { "epoch": 3.64, "learning_rate": 9.541099521100016e-07, "loss": 0.3029, "step": 36222 }, { "epoch": 3.64, "learning_rate": 9.535112936631663e-07, "loss": 0.1781, "step": 36223 }, { "epoch": 3.64, "learning_rate": 9.529128200571369e-07, "loss": 0.2136, "step": 36224 }, { "epoch": 3.64, "learning_rate": 9.523145312957237e-07, "loss": 0.1371, "step": 36225 }, { "epoch": 3.64, "learning_rate": 9.517164273827239e-07, "loss": 0.3206, "step": 36226 }, { "epoch": 3.64, "learning_rate": 9.511185083219575e-07, "loss": 0.2895, "step": 36227 }, { "epoch": 3.64, "learning_rate": 9.505207741172184e-07, "loss": 0.2904, "step": 36228 }, { "epoch": 3.64, "learning_rate": 9.4992322477231e-07, "loss": 0.5074, "step": 36229 }, { "epoch": 3.64, "learning_rate": 9.49325860291046e-07, "loss": 0.1092, "step": 36230 }, { "epoch": 3.64, "learning_rate": 9.4872868067722e-07, "loss": 0.0919, "step": 36231 }, { "epoch": 3.64, "learning_rate": 9.48131685934629e-07, "loss": 0.0694, "step": 36232 }, { "epoch": 3.64, "learning_rate": 9.475348760670766e-07, "loss": 0.0978, "step": 36233 }, { "epoch": 3.64, "learning_rate": 9.469382510783631e-07, "loss": 0.1269, "step": 36234 }, { "epoch": 3.64, "learning_rate": 9.463418109722788e-07, "loss": 0.0925, "step": 36235 }, { "epoch": 3.64, "learning_rate": 9.457455557526273e-07, "loss": 0.1124, "step": 36236 }, { "epoch": 3.64, "learning_rate": 9.451494854231924e-07, "loss": 0.162, "step": 36237 }, { "epoch": 3.64, "learning_rate": 9.445535999877775e-07, "loss": 0.1152, "step": 36238 }, { "epoch": 3.64, "learning_rate": 9.439578994501697e-07, "loss": 0.1668, "step": 36239 }, { "epoch": 3.64, "learning_rate": 9.433623838141559e-07, "loss": 0.2398, "step": 36240 }, { "epoch": 3.64, "learning_rate": 9.427670530835397e-07, "loss": 0.1058, "step": 36241 }, { "epoch": 3.64, "learning_rate": 9.421719072620949e-07, "loss": 0.2779, "step": 36242 }, { "epoch": 3.64, "learning_rate": 9.415769463536083e-07, "loss": 0.8767, "step": 36243 }, { "epoch": 3.64, "learning_rate": 9.40982170361877e-07, "loss": 0.0888, "step": 36244 }, { "epoch": 3.64, "learning_rate": 9.403875792906846e-07, "loss": 0.1862, "step": 36245 }, { "epoch": 3.64, "learning_rate": 9.397931731438048e-07, "loss": 0.0876, "step": 36246 }, { "epoch": 3.64, "learning_rate": 9.391989519250277e-07, "loss": 0.1854, "step": 36247 }, { "epoch": 3.64, "learning_rate": 9.386049156381304e-07, "loss": 0.1055, "step": 36248 }, { "epoch": 3.64, "learning_rate": 9.380110642869e-07, "loss": 0.7402, "step": 36249 }, { "epoch": 3.64, "learning_rate": 9.374173978751099e-07, "loss": 0.6249, "step": 36250 }, { "epoch": 3.64, "learning_rate": 9.368239164065407e-07, "loss": 0.4199, "step": 36251 }, { "epoch": 3.64, "learning_rate": 9.36230619884969e-07, "loss": 0.0574, "step": 36252 }, { "epoch": 3.64, "learning_rate": 9.356375083141655e-07, "loss": 0.3849, "step": 36253 }, { "epoch": 3.64, "learning_rate": 9.350445816979103e-07, "loss": 0.0494, "step": 36254 }, { "epoch": 3.64, "learning_rate": 9.344518400399737e-07, "loss": 0.0562, "step": 36255 }, { "epoch": 3.64, "learning_rate": 9.338592833441295e-07, "loss": 0.1217, "step": 36256 }, { "epoch": 3.64, "learning_rate": 9.332669116141479e-07, "loss": 0.5732, "step": 36257 }, { "epoch": 3.64, "learning_rate": 9.326747248537926e-07, "loss": 0.0576, "step": 36258 }, { "epoch": 3.64, "learning_rate": 9.320827230668471e-07, "loss": 0.6245, "step": 36259 }, { "epoch": 3.65, "learning_rate": 9.314909062570653e-07, "loss": 0.2446, "step": 36260 }, { "epoch": 3.65, "learning_rate": 9.308992744282108e-07, "loss": 0.0988, "step": 36261 }, { "epoch": 3.65, "learning_rate": 9.303078275840604e-07, "loss": 0.3553, "step": 36262 }, { "epoch": 3.65, "learning_rate": 9.29716565728378e-07, "loss": 0.0853, "step": 36263 }, { "epoch": 3.65, "learning_rate": 9.291254888649104e-07, "loss": 0.451, "step": 36264 }, { "epoch": 3.65, "learning_rate": 9.285345969974346e-07, "loss": 0.1243, "step": 36265 }, { "epoch": 3.65, "learning_rate": 9.279438901297011e-07, "loss": 0.0487, "step": 36266 }, { "epoch": 3.65, "learning_rate": 9.273533682654767e-07, "loss": 0.2577, "step": 36267 }, { "epoch": 3.65, "learning_rate": 9.267630314085151e-07, "loss": 0.3443, "step": 36268 }, { "epoch": 3.65, "learning_rate": 9.261728795625702e-07, "loss": 0.1713, "step": 36269 }, { "epoch": 3.65, "learning_rate": 9.255829127314086e-07, "loss": 0.1151, "step": 36270 }, { "epoch": 3.65, "learning_rate": 9.249931309187709e-07, "loss": 0.039, "step": 36271 }, { "epoch": 3.65, "learning_rate": 9.24403534128414e-07, "loss": 0.1361, "step": 36272 }, { "epoch": 3.65, "learning_rate": 9.238141223641017e-07, "loss": 0.2203, "step": 36273 }, { "epoch": 3.65, "learning_rate": 9.232248956295675e-07, "loss": 0.1985, "step": 36274 }, { "epoch": 3.65, "learning_rate": 9.226358539285684e-07, "loss": 0.4929, "step": 36275 }, { "epoch": 3.65, "learning_rate": 9.220469972648549e-07, "loss": 0.1073, "step": 36276 }, { "epoch": 3.65, "learning_rate": 9.214583256421772e-07, "loss": 0.2701, "step": 36277 }, { "epoch": 3.65, "learning_rate": 9.20869839064269e-07, "loss": 0.0774, "step": 36278 }, { "epoch": 3.65, "learning_rate": 9.202815375348839e-07, "loss": 0.4591, "step": 36279 }, { "epoch": 3.65, "learning_rate": 9.196934210577657e-07, "loss": 0.0565, "step": 36280 }, { "epoch": 3.65, "learning_rate": 9.191054896366579e-07, "loss": 0.1313, "step": 36281 }, { "epoch": 3.65, "learning_rate": 9.185177432753011e-07, "loss": 0.1421, "step": 36282 }, { "epoch": 3.65, "learning_rate": 9.179301819774288e-07, "loss": 0.0344, "step": 36283 }, { "epoch": 3.65, "learning_rate": 9.173428057467914e-07, "loss": 0.175, "step": 36284 }, { "epoch": 3.65, "learning_rate": 9.167556145871159e-07, "loss": 0.2838, "step": 36285 }, { "epoch": 3.65, "learning_rate": 9.161686085021459e-07, "loss": 0.2637, "step": 36286 }, { "epoch": 3.65, "learning_rate": 9.155817874956151e-07, "loss": 0.1025, "step": 36287 }, { "epoch": 3.65, "learning_rate": 9.149951515712574e-07, "loss": 0.6028, "step": 36288 }, { "epoch": 3.65, "learning_rate": 9.144087007328061e-07, "loss": 0.1396, "step": 36289 }, { "epoch": 3.65, "learning_rate": 9.138224349839918e-07, "loss": 0.0564, "step": 36290 }, { "epoch": 3.65, "learning_rate": 9.132363543285483e-07, "loss": 0.0223, "step": 36291 }, { "epoch": 3.65, "learning_rate": 9.126504587702022e-07, "loss": 0.1836, "step": 36292 }, { "epoch": 3.65, "learning_rate": 9.120647483126843e-07, "loss": 0.1114, "step": 36293 }, { "epoch": 3.65, "learning_rate": 9.11479222959718e-07, "loss": 0.1254, "step": 36294 }, { "epoch": 3.65, "learning_rate": 9.108938827150337e-07, "loss": 0.1068, "step": 36295 }, { "epoch": 3.65, "learning_rate": 9.103087275823552e-07, "loss": 0.4056, "step": 36296 }, { "epoch": 3.65, "learning_rate": 9.097237575654061e-07, "loss": 0.2038, "step": 36297 }, { "epoch": 3.65, "learning_rate": 9.0913897266791e-07, "loss": 0.0106, "step": 36298 }, { "epoch": 3.65, "learning_rate": 9.085543728935841e-07, "loss": 0.1117, "step": 36299 }, { "epoch": 3.65, "learning_rate": 9.07969958246152e-07, "loss": 0.377, "step": 36300 }, { "epoch": 3.65, "learning_rate": 9.073857287293308e-07, "loss": 0.5683, "step": 36301 }, { "epoch": 3.65, "learning_rate": 9.068016843468441e-07, "loss": 0.3654, "step": 36302 }, { "epoch": 3.65, "learning_rate": 9.06217825102399e-07, "loss": 0.0202, "step": 36303 }, { "epoch": 3.65, "learning_rate": 9.056341509997124e-07, "loss": 0.1831, "step": 36304 }, { "epoch": 3.65, "learning_rate": 9.050506620425114e-07, "loss": 0.2115, "step": 36305 }, { "epoch": 3.65, "learning_rate": 9.044673582344965e-07, "loss": 0.0225, "step": 36306 }, { "epoch": 3.65, "learning_rate": 9.038842395793778e-07, "loss": 0.0418, "step": 36307 }, { "epoch": 3.65, "learning_rate": 9.033013060808726e-07, "loss": 0.3224, "step": 36308 }, { "epoch": 3.65, "learning_rate": 9.027185577426944e-07, "loss": 0.1097, "step": 36309 }, { "epoch": 3.65, "learning_rate": 9.021359945685403e-07, "loss": 0.2239, "step": 36310 }, { "epoch": 3.65, "learning_rate": 9.015536165621274e-07, "loss": 0.557, "step": 36311 }, { "epoch": 3.65, "learning_rate": 9.00971423727156e-07, "loss": 0.1064, "step": 36312 }, { "epoch": 3.65, "learning_rate": 9.003894160673331e-07, "loss": 0.086, "step": 36313 }, { "epoch": 3.65, "learning_rate": 8.998075935863593e-07, "loss": 0.047, "step": 36314 }, { "epoch": 3.65, "learning_rate": 8.992259562879413e-07, "loss": 0.1357, "step": 36315 }, { "epoch": 3.65, "learning_rate": 8.986445041757796e-07, "loss": 0.1413, "step": 36316 }, { "epoch": 3.65, "learning_rate": 8.980632372535746e-07, "loss": 0.2155, "step": 36317 }, { "epoch": 3.65, "learning_rate": 8.974821555250201e-07, "loss": 0.2859, "step": 36318 }, { "epoch": 3.65, "learning_rate": 8.969012589938231e-07, "loss": 0.4047, "step": 36319 }, { "epoch": 3.65, "learning_rate": 8.963205476636738e-07, "loss": 0.0258, "step": 36320 }, { "epoch": 3.65, "learning_rate": 8.957400215382661e-07, "loss": 0.0968, "step": 36321 }, { "epoch": 3.65, "learning_rate": 8.95159680621297e-07, "loss": 0.133, "step": 36322 }, { "epoch": 3.65, "learning_rate": 8.945795249164668e-07, "loss": 0.2185, "step": 36323 }, { "epoch": 3.65, "learning_rate": 8.939995544274493e-07, "loss": 0.3079, "step": 36324 }, { "epoch": 3.65, "learning_rate": 8.934197691579548e-07, "loss": 0.5538, "step": 36325 }, { "epoch": 3.65, "learning_rate": 8.928401691116605e-07, "loss": 0.2684, "step": 36326 }, { "epoch": 3.65, "learning_rate": 8.922607542922601e-07, "loss": 0.2928, "step": 36327 }, { "epoch": 3.65, "learning_rate": 8.916815247034371e-07, "loss": 0.2974, "step": 36328 }, { "epoch": 3.65, "learning_rate": 8.911024803488787e-07, "loss": 0.2767, "step": 36329 }, { "epoch": 3.65, "learning_rate": 8.905236212322754e-07, "loss": 0.4668, "step": 36330 }, { "epoch": 3.65, "learning_rate": 8.899449473573007e-07, "loss": 0.4192, "step": 36331 }, { "epoch": 3.65, "learning_rate": 8.893664587276451e-07, "loss": 0.3154, "step": 36332 }, { "epoch": 3.65, "learning_rate": 8.887881553469857e-07, "loss": 0.1677, "step": 36333 }, { "epoch": 3.65, "learning_rate": 8.882100372190061e-07, "loss": 0.083, "step": 36334 }, { "epoch": 3.65, "learning_rate": 8.876321043473834e-07, "loss": 0.3036, "step": 36335 }, { "epoch": 3.65, "learning_rate": 8.87054356735788e-07, "loss": 0.2287, "step": 36336 }, { "epoch": 3.65, "learning_rate": 8.864767943879138e-07, "loss": 0.052, "step": 36337 }, { "epoch": 3.65, "learning_rate": 8.858994173074208e-07, "loss": 0.1416, "step": 36338 }, { "epoch": 3.65, "learning_rate": 8.853222254979865e-07, "loss": 0.1315, "step": 36339 }, { "epoch": 3.65, "learning_rate": 8.847452189632877e-07, "loss": 0.0575, "step": 36340 }, { "epoch": 3.65, "learning_rate": 8.84168397706998e-07, "loss": 0.297, "step": 36341 }, { "epoch": 3.65, "learning_rate": 8.835917617327782e-07, "loss": 0.487, "step": 36342 }, { "epoch": 3.65, "learning_rate": 8.830153110443084e-07, "loss": 0.1127, "step": 36343 }, { "epoch": 3.65, "learning_rate": 8.824390456452491e-07, "loss": 0.0993, "step": 36344 }, { "epoch": 3.65, "learning_rate": 8.81862965539274e-07, "loss": 0.1643, "step": 36345 }, { "epoch": 3.65, "learning_rate": 8.812870707300435e-07, "loss": 0.0696, "step": 36346 }, { "epoch": 3.65, "learning_rate": 8.807113612212248e-07, "loss": 0.1591, "step": 36347 }, { "epoch": 3.65, "learning_rate": 8.801358370164847e-07, "loss": 0.1135, "step": 36348 }, { "epoch": 3.65, "learning_rate": 8.795604981194804e-07, "loss": 0.2697, "step": 36349 }, { "epoch": 3.65, "learning_rate": 8.789853445338724e-07, "loss": 0.1556, "step": 36350 }, { "epoch": 3.65, "learning_rate": 8.784103762633278e-07, "loss": 0.3389, "step": 36351 }, { "epoch": 3.65, "learning_rate": 8.778355933115001e-07, "loss": 0.3925, "step": 36352 }, { "epoch": 3.65, "learning_rate": 8.772609956820432e-07, "loss": 0.1467, "step": 36353 }, { "epoch": 3.65, "learning_rate": 8.766865833786242e-07, "loss": 0.1413, "step": 36354 }, { "epoch": 3.65, "learning_rate": 8.761123564048934e-07, "loss": 0.0934, "step": 36355 }, { "epoch": 3.65, "learning_rate": 8.755383147644979e-07, "loss": 0.2553, "step": 36356 }, { "epoch": 3.65, "learning_rate": 8.749644584611017e-07, "loss": 0.0799, "step": 36357 }, { "epoch": 3.65, "learning_rate": 8.743907874983514e-07, "loss": 0.1054, "step": 36358 }, { "epoch": 3.66, "learning_rate": 8.738173018798978e-07, "loss": 0.1113, "step": 36359 }, { "epoch": 3.66, "learning_rate": 8.732440016093912e-07, "loss": 0.1832, "step": 36360 }, { "epoch": 3.66, "learning_rate": 8.726708866904787e-07, "loss": 0.1126, "step": 36361 }, { "epoch": 3.66, "learning_rate": 8.720979571268107e-07, "loss": 0.3027, "step": 36362 }, { "epoch": 3.66, "learning_rate": 8.715252129220275e-07, "loss": 0.1596, "step": 36363 }, { "epoch": 3.66, "learning_rate": 8.709526540797763e-07, "loss": 0.2086, "step": 36364 }, { "epoch": 3.66, "learning_rate": 8.703802806037009e-07, "loss": 0.1963, "step": 36365 }, { "epoch": 3.66, "learning_rate": 8.698080924974483e-07, "loss": 0.4766, "step": 36366 }, { "epoch": 3.66, "learning_rate": 8.692360897646523e-07, "loss": 0.1617, "step": 36367 }, { "epoch": 3.66, "learning_rate": 8.686642724089533e-07, "loss": 0.0745, "step": 36368 }, { "epoch": 3.66, "learning_rate": 8.680926404339984e-07, "loss": 0.0621, "step": 36369 }, { "epoch": 3.66, "learning_rate": 8.67521193843418e-07, "loss": 0.1008, "step": 36370 }, { "epoch": 3.66, "learning_rate": 8.669499326408458e-07, "loss": 0.1723, "step": 36371 }, { "epoch": 3.66, "learning_rate": 8.663788568299258e-07, "loss": 0.536, "step": 36372 }, { "epoch": 3.66, "learning_rate": 8.658079664142848e-07, "loss": 0.2148, "step": 36373 }, { "epoch": 3.66, "learning_rate": 8.6523726139756e-07, "loss": 0.3536, "step": 36374 }, { "epoch": 3.66, "learning_rate": 8.64666741783382e-07, "loss": 0.2792, "step": 36375 }, { "epoch": 3.66, "learning_rate": 8.64096407575381e-07, "loss": 0.5428, "step": 36376 }, { "epoch": 3.66, "learning_rate": 8.635262587771875e-07, "loss": 0.1961, "step": 36377 }, { "epoch": 3.66, "learning_rate": 8.629562953924286e-07, "loss": 0.117, "step": 36378 }, { "epoch": 3.66, "learning_rate": 8.62386517424728e-07, "loss": 0.1189, "step": 36379 }, { "epoch": 3.66, "learning_rate": 8.61816924877723e-07, "loss": 0.1919, "step": 36380 }, { "epoch": 3.66, "learning_rate": 8.612475177550272e-07, "loss": 0.3055, "step": 36381 }, { "epoch": 3.66, "learning_rate": 8.60678296060261e-07, "loss": 0.1297, "step": 36382 }, { "epoch": 3.66, "learning_rate": 8.601092597970617e-07, "loss": 0.0867, "step": 36383 }, { "epoch": 3.66, "learning_rate": 8.595404089690362e-07, "loss": 0.2062, "step": 36384 }, { "epoch": 3.66, "learning_rate": 8.589717435798084e-07, "loss": 0.0435, "step": 36385 }, { "epoch": 3.66, "learning_rate": 8.584032636330019e-07, "loss": 0.1131, "step": 36386 }, { "epoch": 3.66, "learning_rate": 8.57834969132234e-07, "loss": 0.0531, "step": 36387 }, { "epoch": 3.66, "learning_rate": 8.572668600811118e-07, "loss": 0.1346, "step": 36388 }, { "epoch": 3.66, "learning_rate": 8.566989364832622e-07, "loss": 0.2152, "step": 36389 }, { "epoch": 3.66, "learning_rate": 8.561311983422926e-07, "loss": 0.0989, "step": 36390 }, { "epoch": 3.66, "learning_rate": 8.555636456618133e-07, "loss": 0.27, "step": 36391 }, { "epoch": 3.66, "learning_rate": 8.549962784454446e-07, "loss": 0.1059, "step": 36392 }, { "epoch": 3.66, "learning_rate": 8.544290966967905e-07, "loss": 0.3105, "step": 36393 }, { "epoch": 3.66, "learning_rate": 8.538621004194613e-07, "loss": 0.1138, "step": 36394 }, { "epoch": 3.66, "learning_rate": 8.532952896170676e-07, "loss": 0.1775, "step": 36395 }, { "epoch": 3.66, "learning_rate": 8.527286642932131e-07, "loss": 0.5668, "step": 36396 }, { "epoch": 3.66, "learning_rate": 8.521622244515015e-07, "loss": 0.2129, "step": 36397 }, { "epoch": 3.66, "learning_rate": 8.5159597009555e-07, "loss": 0.1547, "step": 36398 }, { "epoch": 3.66, "learning_rate": 8.510299012289491e-07, "loss": 0.0943, "step": 36399 }, { "epoch": 3.66, "learning_rate": 8.504640178553025e-07, "loss": 0.1819, "step": 36400 }, { "epoch": 3.66, "learning_rate": 8.498983199782173e-07, "loss": 0.0555, "step": 36401 }, { "epoch": 3.66, "learning_rate": 8.493328076012874e-07, "loss": 0.0093, "step": 36402 }, { "epoch": 3.66, "learning_rate": 8.487674807281131e-07, "loss": 0.1432, "step": 36403 }, { "epoch": 3.66, "learning_rate": 8.48202339362295e-07, "loss": 0.1827, "step": 36404 }, { "epoch": 3.66, "learning_rate": 8.47637383507427e-07, "loss": 0.1936, "step": 36405 }, { "epoch": 3.66, "learning_rate": 8.470726131671058e-07, "loss": 0.1756, "step": 36406 }, { "epoch": 3.66, "learning_rate": 8.465080283449223e-07, "loss": 0.3902, "step": 36407 }, { "epoch": 3.66, "learning_rate": 8.459436290444734e-07, "loss": 0.3499, "step": 36408 }, { "epoch": 3.66, "learning_rate": 8.453794152693462e-07, "loss": 0.2355, "step": 36409 }, { "epoch": 3.66, "learning_rate": 8.448153870231345e-07, "loss": 0.0564, "step": 36410 }, { "epoch": 3.66, "learning_rate": 8.442515443094223e-07, "loss": 0.1363, "step": 36411 }, { "epoch": 3.66, "learning_rate": 8.436878871318099e-07, "loss": 0.1381, "step": 36412 }, { "epoch": 3.66, "learning_rate": 8.431244154938711e-07, "loss": 0.2411, "step": 36413 }, { "epoch": 3.66, "learning_rate": 8.425611293991964e-07, "loss": 0.2265, "step": 36414 }, { "epoch": 3.66, "learning_rate": 8.419980288513729e-07, "loss": 0.0618, "step": 36415 }, { "epoch": 3.66, "learning_rate": 8.414351138539811e-07, "loss": 0.3317, "step": 36416 }, { "epoch": 3.66, "learning_rate": 8.408723844106014e-07, "loss": 0.3603, "step": 36417 }, { "epoch": 3.66, "learning_rate": 8.403098405248177e-07, "loss": 0.0419, "step": 36418 }, { "epoch": 3.66, "learning_rate": 8.397474822002138e-07, "loss": 0.1357, "step": 36419 }, { "epoch": 3.66, "learning_rate": 8.391853094403601e-07, "loss": 0.2333, "step": 36420 }, { "epoch": 3.66, "learning_rate": 8.386233222488371e-07, "loss": 0.1772, "step": 36421 }, { "epoch": 3.66, "learning_rate": 8.38061520629222e-07, "loss": 0.2041, "step": 36422 }, { "epoch": 3.66, "learning_rate": 8.374999045850917e-07, "loss": 0.0737, "step": 36423 }, { "epoch": 3.66, "learning_rate": 8.36938474120017e-07, "loss": 0.0806, "step": 36424 }, { "epoch": 3.66, "learning_rate": 8.363772292375715e-07, "loss": 0.0848, "step": 36425 }, { "epoch": 3.66, "learning_rate": 8.358161699413258e-07, "loss": 0.0804, "step": 36426 }, { "epoch": 3.66, "learning_rate": 8.352552962348537e-07, "loss": 0.4779, "step": 36427 }, { "epoch": 3.66, "learning_rate": 8.346946081217155e-07, "loss": 0.0301, "step": 36428 }, { "epoch": 3.66, "learning_rate": 8.341341056054919e-07, "loss": 0.303, "step": 36429 }, { "epoch": 3.66, "learning_rate": 8.335737886897465e-07, "loss": 0.4227, "step": 36430 }, { "epoch": 3.66, "learning_rate": 8.330136573780367e-07, "loss": 0.1459, "step": 36431 }, { "epoch": 3.66, "learning_rate": 8.324537116739329e-07, "loss": 0.4612, "step": 36432 }, { "epoch": 3.66, "learning_rate": 8.31893951581002e-07, "loss": 0.0901, "step": 36433 }, { "epoch": 3.66, "learning_rate": 8.313343771027948e-07, "loss": 0.0622, "step": 36434 }, { "epoch": 3.66, "learning_rate": 8.307749882428817e-07, "loss": 0.5232, "step": 36435 }, { "epoch": 3.66, "learning_rate": 8.302157850048231e-07, "loss": 0.0581, "step": 36436 }, { "epoch": 3.66, "learning_rate": 8.296567673921729e-07, "loss": 0.4756, "step": 36437 }, { "epoch": 3.66, "learning_rate": 8.290979354084915e-07, "loss": 0.1727, "step": 36438 }, { "epoch": 3.66, "learning_rate": 8.285392890573295e-07, "loss": 0.3819, "step": 36439 }, { "epoch": 3.66, "learning_rate": 8.279808283422507e-07, "loss": 0.0747, "step": 36440 }, { "epoch": 3.66, "learning_rate": 8.274225532668022e-07, "loss": 0.3402, "step": 36441 }, { "epoch": 3.66, "learning_rate": 8.26864463834538e-07, "loss": 0.0908, "step": 36442 }, { "epoch": 3.66, "learning_rate": 8.263065600490083e-07, "loss": 0.0321, "step": 36443 }, { "epoch": 3.66, "learning_rate": 8.257488419137704e-07, "loss": 0.207, "step": 36444 }, { "epoch": 3.66, "learning_rate": 8.251913094323682e-07, "loss": 0.1766, "step": 36445 }, { "epoch": 3.66, "learning_rate": 8.246339626083455e-07, "loss": 0.168, "step": 36446 }, { "epoch": 3.66, "learning_rate": 8.240768014452593e-07, "loss": 0.202, "step": 36447 }, { "epoch": 3.66, "learning_rate": 8.235198259466437e-07, "loss": 0.4398, "step": 36448 }, { "epoch": 3.66, "learning_rate": 8.22963036116049e-07, "loss": 0.0961, "step": 36449 }, { "epoch": 3.66, "learning_rate": 8.22406431957019e-07, "loss": 0.0686, "step": 36450 }, { "epoch": 3.66, "learning_rate": 8.218500134730977e-07, "loss": 0.1238, "step": 36451 }, { "epoch": 3.66, "learning_rate": 8.21293780667819e-07, "loss": 0.4242, "step": 36452 }, { "epoch": 3.66, "learning_rate": 8.207377335447297e-07, "loss": 0.4658, "step": 36453 }, { "epoch": 3.66, "learning_rate": 8.20181872107364e-07, "loss": 0.2126, "step": 36454 }, { "epoch": 3.66, "learning_rate": 8.196261963592589e-07, "loss": 0.6152, "step": 36455 }, { "epoch": 3.66, "learning_rate": 8.190707063039515e-07, "loss": 0.1398, "step": 36456 }, { "epoch": 3.67, "learning_rate": 8.185154019449758e-07, "loss": 0.034, "step": 36457 }, { "epoch": 3.67, "learning_rate": 8.179602832858757e-07, "loss": 1.6729, "step": 36458 }, { "epoch": 3.67, "learning_rate": 8.174053503301682e-07, "loss": 0.1087, "step": 36459 }, { "epoch": 3.67, "learning_rate": 8.168506030813872e-07, "loss": 0.1452, "step": 36460 }, { "epoch": 3.67, "learning_rate": 8.162960415430731e-07, "loss": 0.0267, "step": 36461 }, { "epoch": 3.67, "learning_rate": 8.157416657187499e-07, "loss": 0.2211, "step": 36462 }, { "epoch": 3.67, "learning_rate": 8.151874756119415e-07, "loss": 0.0553, "step": 36463 }, { "epoch": 3.67, "learning_rate": 8.146334712261783e-07, "loss": 0.1922, "step": 36464 }, { "epoch": 3.67, "learning_rate": 8.140796525649874e-07, "loss": 0.1703, "step": 36465 }, { "epoch": 3.67, "learning_rate": 8.135260196318861e-07, "loss": 0.0681, "step": 36466 }, { "epoch": 3.67, "learning_rate": 8.129725724304049e-07, "loss": 0.187, "step": 36467 }, { "epoch": 3.67, "learning_rate": 8.124193109640643e-07, "loss": 0.1519, "step": 36468 }, { "epoch": 3.67, "learning_rate": 8.118662352363815e-07, "loss": 0.2239, "step": 36469 }, { "epoch": 3.67, "learning_rate": 8.11313345250877e-07, "loss": 0.1243, "step": 36470 }, { "epoch": 3.67, "learning_rate": 8.107606410110746e-07, "loss": 0.2213, "step": 36471 }, { "epoch": 3.67, "learning_rate": 8.102081225204849e-07, "loss": 0.2523, "step": 36472 }, { "epoch": 3.67, "learning_rate": 8.096557897826251e-07, "loss": 0.1795, "step": 36473 }, { "epoch": 3.67, "learning_rate": 8.091036428010124e-07, "loss": 0.3487, "step": 36474 }, { "epoch": 3.67, "learning_rate": 8.085516815791571e-07, "loss": 0.2322, "step": 36475 }, { "epoch": 3.67, "learning_rate": 8.0799990612058e-07, "loss": 0.5285, "step": 36476 }, { "epoch": 3.67, "learning_rate": 8.074483164287816e-07, "loss": 0.2386, "step": 36477 }, { "epoch": 3.67, "learning_rate": 8.068969125072756e-07, "loss": 0.1019, "step": 36478 }, { "epoch": 3.67, "learning_rate": 8.063456943595759e-07, "loss": 0.0679, "step": 36479 }, { "epoch": 3.67, "learning_rate": 8.057946619891864e-07, "loss": 0.1533, "step": 36480 }, { "epoch": 3.67, "learning_rate": 8.052438153996077e-07, "loss": 0.0414, "step": 36481 }, { "epoch": 3.67, "learning_rate": 8.046931545943536e-07, "loss": 0.5262, "step": 36482 }, { "epoch": 3.67, "learning_rate": 8.04142679576928e-07, "loss": 0.0779, "step": 36483 }, { "epoch": 3.67, "learning_rate": 8.035923903508314e-07, "loss": 0.1516, "step": 36484 }, { "epoch": 3.67, "learning_rate": 8.030422869195675e-07, "loss": 0.5495, "step": 36485 }, { "epoch": 3.67, "learning_rate": 8.024923692866337e-07, "loss": 0.3126, "step": 36486 }, { "epoch": 3.67, "learning_rate": 8.019426374555305e-07, "loss": 0.1863, "step": 36487 }, { "epoch": 3.67, "learning_rate": 8.013930914297552e-07, "loss": 0.2384, "step": 36488 }, { "epoch": 3.67, "learning_rate": 8.008437312128081e-07, "loss": 0.0475, "step": 36489 }, { "epoch": 3.67, "learning_rate": 8.002945568081866e-07, "loss": 0.1988, "step": 36490 }, { "epoch": 3.67, "learning_rate": 7.997455682193811e-07, "loss": 0.1482, "step": 36491 }, { "epoch": 3.67, "learning_rate": 7.991967654498822e-07, "loss": 0.1225, "step": 36492 }, { "epoch": 3.67, "learning_rate": 7.986481485031871e-07, "loss": 0.1441, "step": 36493 }, { "epoch": 3.67, "learning_rate": 7.980997173827931e-07, "loss": 0.1584, "step": 36494 }, { "epoch": 3.67, "learning_rate": 7.975514720921772e-07, "loss": 0.0578, "step": 36495 }, { "epoch": 3.67, "learning_rate": 7.970034126348368e-07, "loss": 0.4104, "step": 36496 }, { "epoch": 3.67, "learning_rate": 7.964555390142592e-07, "loss": 0.0841, "step": 36497 }, { "epoch": 3.67, "learning_rate": 7.959078512339246e-07, "loss": 0.5761, "step": 36498 }, { "epoch": 3.67, "learning_rate": 7.953603492973238e-07, "loss": 0.0544, "step": 36499 }, { "epoch": 3.67, "learning_rate": 7.94813033207944e-07, "loss": 0.0856, "step": 36500 }, { "epoch": 3.67, "learning_rate": 7.942659029692589e-07, "loss": 0.4082, "step": 36501 }, { "epoch": 3.67, "learning_rate": 7.937189585847593e-07, "loss": 0.2511, "step": 36502 }, { "epoch": 3.67, "learning_rate": 7.931722000579189e-07, "loss": 0.3693, "step": 36503 }, { "epoch": 3.67, "learning_rate": 7.926256273922183e-07, "loss": 0.0825, "step": 36504 }, { "epoch": 3.67, "learning_rate": 7.920792405911414e-07, "loss": 0.0194, "step": 36505 }, { "epoch": 3.67, "learning_rate": 7.915330396581588e-07, "loss": 0.1787, "step": 36506 }, { "epoch": 3.67, "learning_rate": 7.909870245967443e-07, "loss": 0.4627, "step": 36507 }, { "epoch": 3.67, "learning_rate": 7.904411954103819e-07, "loss": 0.2231, "step": 36508 }, { "epoch": 3.67, "learning_rate": 7.898955521025386e-07, "loss": 0.4158, "step": 36509 }, { "epoch": 3.67, "learning_rate": 7.893500946766851e-07, "loss": 0.3235, "step": 36510 }, { "epoch": 3.67, "learning_rate": 7.888048231363021e-07, "loss": 0.1377, "step": 36511 }, { "epoch": 3.67, "learning_rate": 7.882597374848466e-07, "loss": 0.1873, "step": 36512 }, { "epoch": 3.67, "learning_rate": 7.877148377257959e-07, "loss": 0.1558, "step": 36513 }, { "epoch": 3.67, "learning_rate": 7.871701238626139e-07, "loss": 0.1506, "step": 36514 }, { "epoch": 3.67, "learning_rate": 7.866255958987711e-07, "loss": 0.6073, "step": 36515 }, { "epoch": 3.67, "learning_rate": 7.860812538377249e-07, "loss": 0.5537, "step": 36516 }, { "epoch": 3.67, "learning_rate": 7.85537097682949e-07, "loss": 0.2516, "step": 36517 }, { "epoch": 3.67, "learning_rate": 7.849931274378975e-07, "loss": 0.377, "step": 36518 }, { "epoch": 3.67, "learning_rate": 7.844493431060373e-07, "loss": 0.2238, "step": 36519 }, { "epoch": 3.67, "learning_rate": 7.839057446908293e-07, "loss": 0.2748, "step": 36520 }, { "epoch": 3.67, "learning_rate": 7.833623321957239e-07, "loss": 0.2953, "step": 36521 }, { "epoch": 3.67, "learning_rate": 7.828191056241951e-07, "loss": 0.3566, "step": 36522 }, { "epoch": 3.67, "learning_rate": 7.822760649796868e-07, "loss": 0.314, "step": 36523 }, { "epoch": 3.67, "learning_rate": 7.817332102656527e-07, "loss": 0.3871, "step": 36524 }, { "epoch": 3.67, "learning_rate": 7.811905414855603e-07, "loss": 0.1935, "step": 36525 }, { "epoch": 3.67, "learning_rate": 7.806480586428566e-07, "loss": 0.2752, "step": 36526 }, { "epoch": 3.67, "learning_rate": 7.801057617409891e-07, "loss": 0.1112, "step": 36527 }, { "epoch": 3.67, "learning_rate": 7.795636507834147e-07, "loss": 0.0184, "step": 36528 }, { "epoch": 3.67, "learning_rate": 7.790217257735843e-07, "loss": 0.407, "step": 36529 }, { "epoch": 3.67, "learning_rate": 7.784799867149351e-07, "loss": 0.1781, "step": 36530 }, { "epoch": 3.67, "learning_rate": 7.779384336109307e-07, "loss": 0.2364, "step": 36531 }, { "epoch": 3.67, "learning_rate": 7.773970664650054e-07, "loss": 0.4193, "step": 36532 }, { "epoch": 3.67, "learning_rate": 7.768558852806096e-07, "loss": 0.3099, "step": 36533 }, { "epoch": 3.67, "learning_rate": 7.763148900611871e-07, "loss": 0.5296, "step": 36534 }, { "epoch": 3.67, "learning_rate": 7.757740808101754e-07, "loss": 0.1614, "step": 36535 }, { "epoch": 3.67, "learning_rate": 7.752334575310249e-07, "loss": 0.0803, "step": 36536 }, { "epoch": 3.67, "learning_rate": 7.746930202271696e-07, "loss": 0.143, "step": 36537 }, { "epoch": 3.67, "learning_rate": 7.741527689020433e-07, "loss": 0.0266, "step": 36538 }, { "epoch": 3.67, "learning_rate": 7.736127035590967e-07, "loss": 0.0644, "step": 36539 }, { "epoch": 3.67, "learning_rate": 7.730728242017637e-07, "loss": 0.13, "step": 36540 }, { "epoch": 3.67, "learning_rate": 7.725331308334715e-07, "loss": 0.166, "step": 36541 }, { "epoch": 3.67, "learning_rate": 7.719936234576641e-07, "loss": 0.1826, "step": 36542 }, { "epoch": 3.67, "learning_rate": 7.714543020777687e-07, "loss": 0.0778, "step": 36543 }, { "epoch": 3.67, "learning_rate": 7.709151666972159e-07, "loss": 0.1166, "step": 36544 }, { "epoch": 3.67, "learning_rate": 7.703762173194429e-07, "loss": 0.1546, "step": 36545 }, { "epoch": 3.67, "learning_rate": 7.698374539478769e-07, "loss": 0.3348, "step": 36546 }, { "epoch": 3.67, "learning_rate": 7.692988765859421e-07, "loss": 0.0705, "step": 36547 }, { "epoch": 3.67, "learning_rate": 7.687604852370722e-07, "loss": 0.12, "step": 36548 }, { "epoch": 3.67, "learning_rate": 7.682222799046879e-07, "loss": 0.409, "step": 36549 }, { "epoch": 3.67, "learning_rate": 7.676842605922197e-07, "loss": 0.3168, "step": 36550 }, { "epoch": 3.67, "learning_rate": 7.671464273030882e-07, "loss": 0.053, "step": 36551 }, { "epoch": 3.67, "learning_rate": 7.66608780040714e-07, "loss": 0.281, "step": 36552 }, { "epoch": 3.67, "learning_rate": 7.660713188085178e-07, "loss": 0.1103, "step": 36553 }, { "epoch": 3.67, "learning_rate": 7.655340436099301e-07, "loss": 0.1454, "step": 36554 }, { "epoch": 3.67, "learning_rate": 7.649969544483582e-07, "loss": 0.2033, "step": 36555 }, { "epoch": 3.68, "learning_rate": 7.644600513272227e-07, "loss": 0.2143, "step": 36556 }, { "epoch": 3.68, "learning_rate": 7.639233342499441e-07, "loss": 0.2111, "step": 36557 }, { "epoch": 3.68, "learning_rate": 7.633868032199365e-07, "loss": 0.3868, "step": 36558 }, { "epoch": 3.68, "learning_rate": 7.62850458240607e-07, "loss": 0.1183, "step": 36559 }, { "epoch": 3.68, "learning_rate": 7.623142993153798e-07, "loss": 0.1122, "step": 36560 }, { "epoch": 3.68, "learning_rate": 7.617783264476652e-07, "loss": 0.2411, "step": 36561 }, { "epoch": 3.68, "learning_rate": 7.61242539640864e-07, "loss": 0.0984, "step": 36562 }, { "epoch": 3.68, "learning_rate": 7.607069388983934e-07, "loss": 0.0813, "step": 36563 }, { "epoch": 3.68, "learning_rate": 7.60171524223664e-07, "loss": 0.337, "step": 36564 }, { "epoch": 3.68, "learning_rate": 7.596362956200764e-07, "loss": 0.3889, "step": 36565 }, { "epoch": 3.68, "learning_rate": 7.591012530910413e-07, "loss": 0.0739, "step": 36566 }, { "epoch": 3.68, "learning_rate": 7.585663966399592e-07, "loss": 0.3233, "step": 36567 }, { "epoch": 3.68, "learning_rate": 7.580317262702408e-07, "loss": 0.1741, "step": 36568 }, { "epoch": 3.68, "learning_rate": 7.574972419852832e-07, "loss": 0.133, "step": 36569 }, { "epoch": 3.68, "learning_rate": 7.569629437884873e-07, "loss": 0.1957, "step": 36570 }, { "epoch": 3.68, "learning_rate": 7.564288316832535e-07, "loss": 0.0995, "step": 36571 }, { "epoch": 3.68, "learning_rate": 7.558949056729858e-07, "loss": 0.633, "step": 36572 }, { "epoch": 3.68, "learning_rate": 7.553611657610748e-07, "loss": 0.3661, "step": 36573 }, { "epoch": 3.68, "learning_rate": 7.548276119509212e-07, "loss": 0.2569, "step": 36574 }, { "epoch": 3.68, "learning_rate": 7.542942442459221e-07, "loss": 0.2596, "step": 36575 }, { "epoch": 3.68, "learning_rate": 7.537610626494618e-07, "loss": 0.055, "step": 36576 }, { "epoch": 3.68, "learning_rate": 7.532280671649439e-07, "loss": 0.2352, "step": 36577 }, { "epoch": 3.68, "learning_rate": 7.526952577957591e-07, "loss": 0.1272, "step": 36578 }, { "epoch": 3.68, "learning_rate": 7.521626345452914e-07, "loss": 0.1981, "step": 36579 }, { "epoch": 3.68, "learning_rate": 7.516301974169382e-07, "loss": 0.3555, "step": 36580 }, { "epoch": 3.68, "learning_rate": 7.5109794641408e-07, "loss": 0.4752, "step": 36581 }, { "epoch": 3.68, "learning_rate": 7.505658815401106e-07, "loss": 0.2858, "step": 36582 }, { "epoch": 3.68, "learning_rate": 7.500340027984109e-07, "loss": 0.0743, "step": 36583 }, { "epoch": 3.68, "learning_rate": 7.495023101923682e-07, "loss": 0.0634, "step": 36584 }, { "epoch": 3.68, "learning_rate": 7.489708037253628e-07, "loss": 0.0913, "step": 36585 }, { "epoch": 3.68, "learning_rate": 7.484394834007857e-07, "loss": 0.1341, "step": 36586 }, { "epoch": 3.68, "learning_rate": 7.479083492220073e-07, "loss": 0.2057, "step": 36587 }, { "epoch": 3.68, "learning_rate": 7.473774011924117e-07, "loss": 0.2019, "step": 36588 }, { "epoch": 3.68, "learning_rate": 7.468466393153795e-07, "loss": 0.282, "step": 36589 }, { "epoch": 3.68, "learning_rate": 7.463160635942878e-07, "loss": 0.1394, "step": 36590 }, { "epoch": 3.68, "learning_rate": 7.457856740325109e-07, "loss": 0.0332, "step": 36591 }, { "epoch": 3.68, "learning_rate": 7.452554706334225e-07, "loss": 0.0893, "step": 36592 }, { "epoch": 3.68, "learning_rate": 7.447254534004033e-07, "loss": 0.2022, "step": 36593 }, { "epoch": 3.68, "learning_rate": 7.441956223368208e-07, "loss": 0.1483, "step": 36594 }, { "epoch": 3.68, "learning_rate": 7.436659774460486e-07, "loss": 0.0717, "step": 36595 }, { "epoch": 3.68, "learning_rate": 7.431365187314543e-07, "loss": 0.5024, "step": 36596 }, { "epoch": 3.68, "learning_rate": 7.426072461964117e-07, "loss": 0.087, "step": 36597 }, { "epoch": 3.68, "learning_rate": 7.42078159844285e-07, "loss": 0.0914, "step": 36598 }, { "epoch": 3.68, "learning_rate": 7.415492596784412e-07, "loss": 0.1264, "step": 36599 }, { "epoch": 3.68, "learning_rate": 7.410205457022512e-07, "loss": 0.0748, "step": 36600 }, { "epoch": 3.68, "learning_rate": 7.404920179190755e-07, "loss": 0.261, "step": 36601 }, { "epoch": 3.68, "learning_rate": 7.399636763322715e-07, "loss": 0.0757, "step": 36602 }, { "epoch": 3.68, "learning_rate": 7.39435520945213e-07, "loss": 0.0585, "step": 36603 }, { "epoch": 3.68, "learning_rate": 7.389075517612575e-07, "loss": 0.518, "step": 36604 }, { "epoch": 3.68, "learning_rate": 7.383797687837556e-07, "loss": 0.0938, "step": 36605 }, { "epoch": 3.68, "learning_rate": 7.378521720160781e-07, "loss": 0.258, "step": 36606 }, { "epoch": 3.68, "learning_rate": 7.373247614615819e-07, "loss": 0.0905, "step": 36607 }, { "epoch": 3.68, "learning_rate": 7.36797537123608e-07, "loss": 0.7316, "step": 36608 }, { "epoch": 3.68, "learning_rate": 7.362704990055302e-07, "loss": 0.0932, "step": 36609 }, { "epoch": 3.68, "learning_rate": 7.357436471106926e-07, "loss": 0.0746, "step": 36610 }, { "epoch": 3.68, "learning_rate": 7.352169814424492e-07, "loss": 0.4528, "step": 36611 }, { "epoch": 3.68, "learning_rate": 7.346905020041506e-07, "loss": 0.1113, "step": 36612 }, { "epoch": 3.68, "learning_rate": 7.341642087991507e-07, "loss": 0.1281, "step": 36613 }, { "epoch": 3.68, "learning_rate": 7.336381018307969e-07, "loss": 0.1798, "step": 36614 }, { "epoch": 3.68, "learning_rate": 7.331121811024366e-07, "loss": 0.0929, "step": 36615 }, { "epoch": 3.68, "learning_rate": 7.325864466174137e-07, "loss": 0.076, "step": 36616 }, { "epoch": 3.68, "learning_rate": 7.320608983790789e-07, "loss": 0.3126, "step": 36617 }, { "epoch": 3.68, "learning_rate": 7.315355363907761e-07, "loss": 0.2283, "step": 36618 }, { "epoch": 3.68, "learning_rate": 7.310103606558427e-07, "loss": 0.1342, "step": 36619 }, { "epoch": 3.68, "learning_rate": 7.304853711776294e-07, "loss": 0.3402, "step": 36620 }, { "epoch": 3.68, "learning_rate": 7.299605679594701e-07, "loss": 0.1182, "step": 36621 }, { "epoch": 3.68, "learning_rate": 7.294359510047055e-07, "loss": 0.1682, "step": 36622 }, { "epoch": 3.68, "learning_rate": 7.289115203166797e-07, "loss": 0.1589, "step": 36623 }, { "epoch": 3.68, "learning_rate": 7.283872758987231e-07, "loss": 0.3923, "step": 36624 }, { "epoch": 3.68, "learning_rate": 7.278632177541733e-07, "loss": 0.2353, "step": 36625 }, { "epoch": 3.68, "learning_rate": 7.273393458863676e-07, "loss": 0.1707, "step": 36626 }, { "epoch": 3.68, "learning_rate": 7.268156602986398e-07, "loss": 0.2784, "step": 36627 }, { "epoch": 3.68, "learning_rate": 7.262921609943207e-07, "loss": 0.4957, "step": 36628 }, { "epoch": 3.68, "learning_rate": 7.257688479767411e-07, "loss": 0.1552, "step": 36629 }, { "epoch": 3.68, "learning_rate": 7.252457212492314e-07, "loss": 0.1737, "step": 36630 }, { "epoch": 3.68, "learning_rate": 7.247227808151191e-07, "loss": 0.041, "step": 36631 }, { "epoch": 3.68, "learning_rate": 7.242000266777382e-07, "loss": 0.0952, "step": 36632 }, { "epoch": 3.68, "learning_rate": 7.236774588404094e-07, "loss": 0.2605, "step": 36633 }, { "epoch": 3.68, "learning_rate": 7.231550773064599e-07, "loss": 0.0778, "step": 36634 }, { "epoch": 3.68, "learning_rate": 7.226328820792138e-07, "loss": 0.0831, "step": 36635 }, { "epoch": 3.68, "learning_rate": 7.221108731619985e-07, "loss": 0.1329, "step": 36636 }, { "epoch": 3.68, "learning_rate": 7.215890505581246e-07, "loss": 0.2153, "step": 36637 }, { "epoch": 3.68, "learning_rate": 7.210674142709228e-07, "loss": 0.3039, "step": 36638 }, { "epoch": 3.68, "learning_rate": 7.205459643037138e-07, "loss": 0.1461, "step": 36639 }, { "epoch": 3.68, "learning_rate": 7.200247006598049e-07, "loss": 0.2848, "step": 36640 }, { "epoch": 3.68, "learning_rate": 7.195036233425234e-07, "loss": 0.0649, "step": 36641 }, { "epoch": 3.68, "learning_rate": 7.189827323551834e-07, "loss": 0.129, "step": 36642 }, { "epoch": 3.68, "learning_rate": 7.184620277010989e-07, "loss": 0.6103, "step": 36643 }, { "epoch": 3.68, "learning_rate": 7.179415093835806e-07, "loss": 0.0942, "step": 36644 }, { "epoch": 3.68, "learning_rate": 7.174211774059391e-07, "loss": 0.3679, "step": 36645 }, { "epoch": 3.68, "learning_rate": 7.169010317714986e-07, "loss": 0.0166, "step": 36646 }, { "epoch": 3.68, "learning_rate": 7.163810724835563e-07, "loss": 0.2449, "step": 36647 }, { "epoch": 3.68, "learning_rate": 7.158612995454194e-07, "loss": 0.1823, "step": 36648 }, { "epoch": 3.68, "learning_rate": 7.153417129604056e-07, "loss": 0.0717, "step": 36649 }, { "epoch": 3.68, "learning_rate": 7.148223127318187e-07, "loss": 0.1959, "step": 36650 }, { "epoch": 3.68, "learning_rate": 7.143030988629595e-07, "loss": 0.1941, "step": 36651 }, { "epoch": 3.68, "learning_rate": 7.137840713571319e-07, "loss": 0.1344, "step": 36652 }, { "epoch": 3.68, "learning_rate": 7.132652302176468e-07, "loss": 0.3605, "step": 36653 }, { "epoch": 3.69, "learning_rate": 7.12746575447798e-07, "loss": 0.3256, "step": 36654 }, { "epoch": 3.69, "learning_rate": 7.122281070508896e-07, "loss": 0.3019, "step": 36655 }, { "epoch": 3.69, "learning_rate": 7.11709825030219e-07, "loss": 0.1586, "step": 36656 }, { "epoch": 3.69, "learning_rate": 7.111917293890834e-07, "loss": 0.07, "step": 36657 }, { "epoch": 3.69, "learning_rate": 7.106738201307838e-07, "loss": 0.0744, "step": 36658 }, { "epoch": 3.69, "learning_rate": 7.101560972586141e-07, "loss": 0.1652, "step": 36659 }, { "epoch": 3.69, "learning_rate": 7.096385607758682e-07, "loss": 0.4227, "step": 36660 }, { "epoch": 3.69, "learning_rate": 7.09121210685837e-07, "loss": 0.113, "step": 36661 }, { "epoch": 3.69, "learning_rate": 7.086040469918209e-07, "loss": 0.0838, "step": 36662 }, { "epoch": 3.69, "learning_rate": 7.080870696970976e-07, "loss": 0.154, "step": 36663 }, { "epoch": 3.69, "learning_rate": 7.075702788049742e-07, "loss": 0.0929, "step": 36664 }, { "epoch": 3.69, "learning_rate": 7.07053674318725e-07, "loss": 0.3395, "step": 36665 }, { "epoch": 3.69, "learning_rate": 7.065372562416406e-07, "loss": 0.0396, "step": 36666 }, { "epoch": 3.69, "learning_rate": 7.060210245770149e-07, "loss": 0.0353, "step": 36667 }, { "epoch": 3.69, "learning_rate": 7.055049793281254e-07, "loss": 0.2388, "step": 36668 }, { "epoch": 3.69, "learning_rate": 7.049891204982562e-07, "loss": 0.2846, "step": 36669 }, { "epoch": 3.69, "learning_rate": 7.044734480906945e-07, "loss": 0.1092, "step": 36670 }, { "epoch": 3.69, "learning_rate": 7.039579621087178e-07, "loss": 0.1707, "step": 36671 }, { "epoch": 3.69, "learning_rate": 7.034426625556068e-07, "loss": 0.1193, "step": 36672 }, { "epoch": 3.69, "learning_rate": 7.029275494346421e-07, "loss": 0.4322, "step": 36673 }, { "epoch": 3.69, "learning_rate": 7.024126227491046e-07, "loss": 0.1638, "step": 36674 }, { "epoch": 3.69, "learning_rate": 7.018978825022649e-07, "loss": 0.189, "step": 36675 }, { "epoch": 3.69, "learning_rate": 7.013833286974003e-07, "loss": 0.0432, "step": 36676 }, { "epoch": 3.69, "learning_rate": 7.00868961337785e-07, "loss": 0.2385, "step": 36677 }, { "epoch": 3.69, "learning_rate": 7.003547804266997e-07, "loss": 0.0391, "step": 36678 }, { "epoch": 3.69, "learning_rate": 6.998407859674083e-07, "loss": 0.3784, "step": 36679 }, { "epoch": 3.69, "learning_rate": 6.993269779631784e-07, "loss": 0.0729, "step": 36680 }, { "epoch": 3.69, "learning_rate": 6.988133564172905e-07, "loss": 0.441, "step": 36681 }, { "epoch": 3.69, "learning_rate": 6.982999213330088e-07, "loss": 0.2425, "step": 36682 }, { "epoch": 3.69, "learning_rate": 6.977866727135973e-07, "loss": 0.3727, "step": 36683 }, { "epoch": 3.69, "learning_rate": 6.972736105623234e-07, "loss": 0.0357, "step": 36684 }, { "epoch": 3.69, "learning_rate": 6.967607348824545e-07, "loss": 0.1464, "step": 36685 }, { "epoch": 3.69, "learning_rate": 6.962480456772513e-07, "loss": 0.0504, "step": 36686 }, { "epoch": 3.69, "learning_rate": 6.957355429499779e-07, "loss": 0.4599, "step": 36687 }, { "epoch": 3.69, "learning_rate": 6.952232267038983e-07, "loss": 0.2744, "step": 36688 }, { "epoch": 3.69, "learning_rate": 6.947110969422666e-07, "loss": 0.1167, "step": 36689 }, { "epoch": 3.69, "learning_rate": 6.941991536683467e-07, "loss": 0.1185, "step": 36690 }, { "epoch": 3.69, "learning_rate": 6.936873968853963e-07, "loss": 0.0355, "step": 36691 }, { "epoch": 3.69, "learning_rate": 6.931758265966692e-07, "loss": 0.1504, "step": 36692 }, { "epoch": 3.69, "learning_rate": 6.926644428054196e-07, "loss": 0.2279, "step": 36693 }, { "epoch": 3.69, "learning_rate": 6.921532455149082e-07, "loss": 0.0526, "step": 36694 }, { "epoch": 3.69, "learning_rate": 6.916422347283791e-07, "loss": 0.1604, "step": 36695 }, { "epoch": 3.69, "learning_rate": 6.911314104490962e-07, "loss": 0.0732, "step": 36696 }, { "epoch": 3.69, "learning_rate": 6.906207726803004e-07, "loss": 0.5102, "step": 36697 }, { "epoch": 3.69, "learning_rate": 6.901103214252424e-07, "loss": 0.1119, "step": 36698 }, { "epoch": 3.69, "learning_rate": 6.89600056687173e-07, "loss": 0.1318, "step": 36699 }, { "epoch": 3.69, "learning_rate": 6.890899784693394e-07, "loss": 0.0314, "step": 36700 }, { "epoch": 3.69, "learning_rate": 6.885800867749892e-07, "loss": 0.0738, "step": 36701 }, { "epoch": 3.69, "learning_rate": 6.88070381607363e-07, "loss": 0.0258, "step": 36702 }, { "epoch": 3.69, "learning_rate": 6.87560862969705e-07, "loss": 0.0678, "step": 36703 }, { "epoch": 3.69, "learning_rate": 6.870515308652592e-07, "loss": 0.1849, "step": 36704 }, { "epoch": 3.69, "learning_rate": 6.865423852972663e-07, "loss": 0.1151, "step": 36705 }, { "epoch": 3.69, "learning_rate": 6.860334262689638e-07, "loss": 0.1906, "step": 36706 }, { "epoch": 3.69, "learning_rate": 6.855246537835958e-07, "loss": 0.1799, "step": 36707 }, { "epoch": 3.69, "learning_rate": 6.850160678443995e-07, "loss": 0.2613, "step": 36708 }, { "epoch": 3.69, "learning_rate": 6.845076684546025e-07, "loss": 0.3761, "step": 36709 }, { "epoch": 3.69, "learning_rate": 6.839994556174523e-07, "loss": 0.1162, "step": 36710 }, { "epoch": 3.69, "learning_rate": 6.834914293361761e-07, "loss": 0.2777, "step": 36711 }, { "epoch": 3.69, "learning_rate": 6.829835896140013e-07, "loss": 0.3449, "step": 36712 }, { "epoch": 3.69, "learning_rate": 6.824759364541722e-07, "loss": 0.1408, "step": 36713 }, { "epoch": 3.69, "learning_rate": 6.81968469859916e-07, "loss": 0.2259, "step": 36714 }, { "epoch": 3.69, "learning_rate": 6.814611898344536e-07, "loss": 0.1277, "step": 36715 }, { "epoch": 3.69, "learning_rate": 6.809540963810224e-07, "loss": 0.284, "step": 36716 }, { "epoch": 3.69, "learning_rate": 6.80447189502843e-07, "loss": 0.4001, "step": 36717 }, { "epoch": 3.69, "learning_rate": 6.799404692031463e-07, "loss": 0.0759, "step": 36718 }, { "epoch": 3.69, "learning_rate": 6.794339354851531e-07, "loss": 0.0742, "step": 36719 }, { "epoch": 3.69, "learning_rate": 6.789275883520907e-07, "loss": 0.1795, "step": 36720 }, { "epoch": 3.69, "learning_rate": 6.784214278071765e-07, "loss": 0.181, "step": 36721 }, { "epoch": 3.69, "learning_rate": 6.779154538536348e-07, "loss": 0.0651, "step": 36722 }, { "epoch": 3.69, "learning_rate": 6.774096664946827e-07, "loss": 0.2838, "step": 36723 }, { "epoch": 3.69, "learning_rate": 6.769040657335446e-07, "loss": 0.037, "step": 36724 }, { "epoch": 3.69, "learning_rate": 6.76398651573431e-07, "loss": 0.3708, "step": 36725 }, { "epoch": 3.69, "learning_rate": 6.758934240175563e-07, "loss": 0.0736, "step": 36726 }, { "epoch": 3.69, "learning_rate": 6.753883830691477e-07, "loss": 0.356, "step": 36727 }, { "epoch": 3.69, "learning_rate": 6.748835287314126e-07, "loss": 0.2469, "step": 36728 }, { "epoch": 3.69, "learning_rate": 6.743788610075551e-07, "loss": 0.3267, "step": 36729 }, { "epoch": 3.69, "learning_rate": 6.738743799007996e-07, "loss": 0.0871, "step": 36730 }, { "epoch": 3.69, "learning_rate": 6.733700854143532e-07, "loss": 0.0798, "step": 36731 }, { "epoch": 3.69, "learning_rate": 6.728659775514201e-07, "loss": 0.2125, "step": 36732 }, { "epoch": 3.69, "learning_rate": 6.72362056315211e-07, "loss": 0.122, "step": 36733 }, { "epoch": 3.69, "learning_rate": 6.718583217089336e-07, "loss": 0.1951, "step": 36734 }, { "epoch": 3.69, "learning_rate": 6.713547737357951e-07, "loss": 0.239, "step": 36735 }, { "epoch": 3.69, "learning_rate": 6.708514123989962e-07, "loss": 0.0619, "step": 36736 }, { "epoch": 3.69, "learning_rate": 6.703482377017411e-07, "loss": 0.049, "step": 36737 }, { "epoch": 3.69, "learning_rate": 6.698452496472307e-07, "loss": 0.0404, "step": 36738 }, { "epoch": 3.69, "learning_rate": 6.693424482386689e-07, "loss": 0.0951, "step": 36739 }, { "epoch": 3.69, "learning_rate": 6.688398334792567e-07, "loss": 0.0838, "step": 36740 }, { "epoch": 3.69, "learning_rate": 6.683374053721814e-07, "loss": 0.3909, "step": 36741 }, { "epoch": 3.69, "learning_rate": 6.67835163920657e-07, "loss": 0.2204, "step": 36742 }, { "epoch": 3.69, "learning_rate": 6.673331091278711e-07, "loss": 0.0889, "step": 36743 }, { "epoch": 3.69, "learning_rate": 6.668312409970112e-07, "loss": 0.0837, "step": 36744 }, { "epoch": 3.69, "learning_rate": 6.663295595312812e-07, "loss": 0.0903, "step": 36745 }, { "epoch": 3.69, "learning_rate": 6.658280647338755e-07, "loss": 0.4058, "step": 36746 }, { "epoch": 3.69, "learning_rate": 6.653267566079746e-07, "loss": 0.2595, "step": 36747 }, { "epoch": 3.69, "learning_rate": 6.648256351567794e-07, "loss": 0.0978, "step": 36748 }, { "epoch": 3.69, "learning_rate": 6.64324700383474e-07, "loss": 0.2538, "step": 36749 }, { "epoch": 3.69, "learning_rate": 6.638239522912426e-07, "loss": 0.0723, "step": 36750 }, { "epoch": 3.69, "learning_rate": 6.633233908832792e-07, "loss": 0.1394, "step": 36751 }, { "epoch": 3.69, "learning_rate": 6.628230161627647e-07, "loss": 0.2174, "step": 36752 }, { "epoch": 3.7, "learning_rate": 6.623228281328831e-07, "loss": 0.1177, "step": 36753 }, { "epoch": 3.7, "learning_rate": 6.618228267968185e-07, "loss": 0.3873, "step": 36754 }, { "epoch": 3.7, "learning_rate": 6.613230121577485e-07, "loss": 0.2353, "step": 36755 }, { "epoch": 3.7, "learning_rate": 6.608233842188671e-07, "loss": 0.1314, "step": 36756 }, { "epoch": 3.7, "learning_rate": 6.603239429833385e-07, "loss": 0.121, "step": 36757 }, { "epoch": 3.7, "learning_rate": 6.598246884543469e-07, "loss": 0.2419, "step": 36758 }, { "epoch": 3.7, "learning_rate": 6.593256206350695e-07, "loss": 0.1824, "step": 36759 }, { "epoch": 3.7, "learning_rate": 6.588267395286873e-07, "loss": 0.0466, "step": 36760 }, { "epoch": 3.7, "learning_rate": 6.583280451383611e-07, "loss": 0.105, "step": 36761 }, { "epoch": 3.7, "learning_rate": 6.578295374672783e-07, "loss": 0.0134, "step": 36762 }, { "epoch": 3.7, "learning_rate": 6.573312165186063e-07, "loss": 0.1557, "step": 36763 }, { "epoch": 3.7, "learning_rate": 6.568330822955127e-07, "loss": 0.0582, "step": 36764 }, { "epoch": 3.7, "learning_rate": 6.563351348011748e-07, "loss": 0.2658, "step": 36765 }, { "epoch": 3.7, "learning_rate": 6.558373740387535e-07, "loss": 0.3019, "step": 36766 }, { "epoch": 3.7, "learning_rate": 6.55339800011423e-07, "loss": 0.2996, "step": 36767 }, { "epoch": 3.7, "learning_rate": 6.548424127223441e-07, "loss": 0.1839, "step": 36768 }, { "epoch": 3.7, "learning_rate": 6.543452121746873e-07, "loss": 0.5943, "step": 36769 }, { "epoch": 3.7, "learning_rate": 6.538481983716138e-07, "loss": 0.5037, "step": 36770 }, { "epoch": 3.7, "learning_rate": 6.533513713162842e-07, "loss": 0.1736, "step": 36771 }, { "epoch": 3.7, "learning_rate": 6.528547310118626e-07, "loss": 0.0942, "step": 36772 }, { "epoch": 3.7, "learning_rate": 6.5235827746151e-07, "loss": 0.1872, "step": 36773 }, { "epoch": 3.7, "learning_rate": 6.51862010668387e-07, "loss": 0.5186, "step": 36774 }, { "epoch": 3.7, "learning_rate": 6.513659306356478e-07, "loss": 0.2631, "step": 36775 }, { "epoch": 3.7, "learning_rate": 6.508700373664467e-07, "loss": 0.035, "step": 36776 }, { "epoch": 3.7, "learning_rate": 6.50374330863951e-07, "loss": 0.0461, "step": 36777 }, { "epoch": 3.7, "learning_rate": 6.498788111313047e-07, "loss": 0.3341, "step": 36778 }, { "epoch": 3.7, "learning_rate": 6.493834781716623e-07, "loss": 0.4361, "step": 36779 }, { "epoch": 3.7, "learning_rate": 6.488883319881811e-07, "loss": 0.0873, "step": 36780 }, { "epoch": 3.7, "learning_rate": 6.483933725840086e-07, "loss": 0.1336, "step": 36781 }, { "epoch": 3.7, "learning_rate": 6.478985999622922e-07, "loss": 0.1808, "step": 36782 }, { "epoch": 3.7, "learning_rate": 6.474040141261861e-07, "loss": 0.3053, "step": 36783 }, { "epoch": 3.7, "learning_rate": 6.469096150788312e-07, "loss": 0.2851, "step": 36784 }, { "epoch": 3.7, "learning_rate": 6.464154028233815e-07, "loss": 0.0197, "step": 36785 }, { "epoch": 3.7, "learning_rate": 6.459213773629747e-07, "loss": 0.0751, "step": 36786 }, { "epoch": 3.7, "learning_rate": 6.454275387007546e-07, "loss": 0.037, "step": 36787 }, { "epoch": 3.7, "learning_rate": 6.449338868398724e-07, "loss": 0.1299, "step": 36788 }, { "epoch": 3.7, "learning_rate": 6.444404217834621e-07, "loss": 0.0812, "step": 36789 }, { "epoch": 3.7, "learning_rate": 6.439471435346611e-07, "loss": 0.8847, "step": 36790 }, { "epoch": 3.7, "learning_rate": 6.43454052096617e-07, "loss": 0.4875, "step": 36791 }, { "epoch": 3.7, "learning_rate": 6.42961147472464e-07, "loss": 0.041, "step": 36792 }, { "epoch": 3.7, "learning_rate": 6.42468429665336e-07, "loss": 0.2438, "step": 36793 }, { "epoch": 3.7, "learning_rate": 6.419758986783708e-07, "loss": 0.0955, "step": 36794 }, { "epoch": 3.7, "learning_rate": 6.414835545147024e-07, "loss": 0.2192, "step": 36795 }, { "epoch": 3.7, "learning_rate": 6.409913971774651e-07, "loss": 0.2283, "step": 36796 }, { "epoch": 3.7, "learning_rate": 6.404994266697895e-07, "loss": 0.2491, "step": 36797 }, { "epoch": 3.7, "learning_rate": 6.400076429948098e-07, "loss": 0.2796, "step": 36798 }, { "epoch": 3.7, "learning_rate": 6.395160461556471e-07, "loss": 0.041, "step": 36799 }, { "epoch": 3.7, "learning_rate": 6.390246361554386e-07, "loss": 0.1145, "step": 36800 }, { "epoch": 3.7, "learning_rate": 6.385334129973086e-07, "loss": 0.1171, "step": 36801 }, { "epoch": 3.7, "learning_rate": 6.380423766843779e-07, "loss": 0.0648, "step": 36802 }, { "epoch": 3.7, "learning_rate": 6.375515272197807e-07, "loss": 0.348, "step": 36803 }, { "epoch": 3.7, "learning_rate": 6.370608646066312e-07, "loss": 0.4446, "step": 36804 }, { "epoch": 3.7, "learning_rate": 6.365703888480567e-07, "loss": 0.1565, "step": 36805 }, { "epoch": 3.7, "learning_rate": 6.360800999471816e-07, "loss": 0.1365, "step": 36806 }, { "epoch": 3.7, "learning_rate": 6.355899979071167e-07, "loss": 0.0436, "step": 36807 }, { "epoch": 3.7, "learning_rate": 6.351000827309894e-07, "loss": 0.3922, "step": 36808 }, { "epoch": 3.7, "learning_rate": 6.34610354421914e-07, "loss": 0.0469, "step": 36809 }, { "epoch": 3.7, "learning_rate": 6.341208129830079e-07, "loss": 0.1763, "step": 36810 }, { "epoch": 3.7, "learning_rate": 6.336314584173853e-07, "loss": 0.3253, "step": 36811 }, { "epoch": 3.7, "learning_rate": 6.33142290728157e-07, "loss": 0.1956, "step": 36812 }, { "epoch": 3.7, "learning_rate": 6.32653309918444e-07, "loss": 0.1191, "step": 36813 }, { "epoch": 3.7, "learning_rate": 6.321645159913503e-07, "loss": 0.2217, "step": 36814 }, { "epoch": 3.7, "learning_rate": 6.316759089499935e-07, "loss": 0.0223, "step": 36815 }, { "epoch": 3.7, "learning_rate": 6.311874887974744e-07, "loss": 0.1374, "step": 36816 }, { "epoch": 3.7, "learning_rate": 6.306992555369107e-07, "loss": 0.1056, "step": 36817 }, { "epoch": 3.7, "learning_rate": 6.302112091713997e-07, "loss": 0.1023, "step": 36818 }, { "epoch": 3.7, "learning_rate": 6.297233497040522e-07, "loss": 0.0809, "step": 36819 }, { "epoch": 3.7, "learning_rate": 6.292356771379759e-07, "loss": 0.1037, "step": 36820 }, { "epoch": 3.7, "learning_rate": 6.287481914762683e-07, "loss": 0.1234, "step": 36821 }, { "epoch": 3.7, "learning_rate": 6.282608927220301e-07, "loss": 0.5069, "step": 36822 }, { "epoch": 3.7, "learning_rate": 6.277737808783691e-07, "loss": 0.0774, "step": 36823 }, { "epoch": 3.7, "learning_rate": 6.272868559483857e-07, "loss": 0.1731, "step": 36824 }, { "epoch": 3.7, "learning_rate": 6.268001179351679e-07, "loss": 0.9642, "step": 36825 }, { "epoch": 3.7, "learning_rate": 6.263135668418263e-07, "loss": 0.0142, "step": 36826 }, { "epoch": 3.7, "learning_rate": 6.258272026714484e-07, "loss": 0.1195, "step": 36827 }, { "epoch": 3.7, "learning_rate": 6.253410254271286e-07, "loss": 0.2782, "step": 36828 }, { "epoch": 3.7, "learning_rate": 6.248550351119675e-07, "loss": 0.1379, "step": 36829 }, { "epoch": 3.7, "learning_rate": 6.243692317290528e-07, "loss": 0.1057, "step": 36830 }, { "epoch": 3.7, "learning_rate": 6.238836152814786e-07, "loss": 0.1513, "step": 36831 }, { "epoch": 3.7, "learning_rate": 6.233981857723325e-07, "loss": 0.1353, "step": 36832 }, { "epoch": 3.7, "learning_rate": 6.229129432047054e-07, "loss": 0.1519, "step": 36833 }, { "epoch": 3.7, "learning_rate": 6.224278875816846e-07, "loss": 0.3203, "step": 36834 }, { "epoch": 3.7, "learning_rate": 6.21943018906358e-07, "loss": 0.1065, "step": 36835 }, { "epoch": 3.7, "learning_rate": 6.214583371818062e-07, "loss": 0.0555, "step": 36836 }, { "epoch": 3.7, "learning_rate": 6.2097384241112e-07, "loss": 0.0376, "step": 36837 }, { "epoch": 3.7, "learning_rate": 6.204895345973838e-07, "loss": 0.5264, "step": 36838 }, { "epoch": 3.7, "learning_rate": 6.200054137436684e-07, "loss": 0.0118, "step": 36839 }, { "epoch": 3.7, "learning_rate": 6.19521479853068e-07, "loss": 0.3296, "step": 36840 }, { "epoch": 3.7, "learning_rate": 6.190377329286534e-07, "loss": 0.1026, "step": 36841 }, { "epoch": 3.7, "learning_rate": 6.185541729735055e-07, "loss": 0.5187, "step": 36842 }, { "epoch": 3.7, "learning_rate": 6.180707999906987e-07, "loss": 0.1118, "step": 36843 }, { "epoch": 3.7, "learning_rate": 6.175876139833169e-07, "loss": 0.0236, "step": 36844 }, { "epoch": 3.7, "learning_rate": 6.171046149544279e-07, "loss": 0.1364, "step": 36845 }, { "epoch": 3.7, "learning_rate": 6.166218029071057e-07, "loss": 0.1423, "step": 36846 }, { "epoch": 3.7, "learning_rate": 6.16139177844428e-07, "loss": 0.2778, "step": 36847 }, { "epoch": 3.7, "learning_rate": 6.15656739769459e-07, "loss": 0.1322, "step": 36848 }, { "epoch": 3.7, "learning_rate": 6.151744886852728e-07, "loss": 0.2745, "step": 36849 }, { "epoch": 3.7, "learning_rate": 6.146924245949403e-07, "loss": 0.1794, "step": 36850 }, { "epoch": 3.71, "learning_rate": 6.142105475015191e-07, "loss": 0.1345, "step": 36851 }, { "epoch": 3.71, "learning_rate": 6.137288574080935e-07, "loss": 0.1135, "step": 36852 }, { "epoch": 3.71, "learning_rate": 6.132473543177109e-07, "loss": 0.0564, "step": 36853 }, { "epoch": 3.71, "learning_rate": 6.127660382334455e-07, "loss": 0.0933, "step": 36854 }, { "epoch": 3.71, "learning_rate": 6.122849091583549e-07, "loss": 0.0674, "step": 36855 }, { "epoch": 3.71, "learning_rate": 6.1180396709551e-07, "loss": 0.1021, "step": 36856 }, { "epoch": 3.71, "learning_rate": 6.113232120479584e-07, "loss": 0.1666, "step": 36857 }, { "epoch": 3.71, "learning_rate": 6.108426440187675e-07, "loss": 0.0326, "step": 36858 }, { "epoch": 3.71, "learning_rate": 6.10362263010995e-07, "loss": 0.5246, "step": 36859 }, { "epoch": 3.71, "learning_rate": 6.098820690276952e-07, "loss": 0.2233, "step": 36860 }, { "epoch": 3.71, "learning_rate": 6.094020620719254e-07, "loss": 0.094, "step": 36861 }, { "epoch": 3.71, "learning_rate": 6.0892224214674e-07, "loss": 0.1613, "step": 36862 }, { "epoch": 3.71, "learning_rate": 6.084426092551931e-07, "loss": 0.223, "step": 36863 }, { "epoch": 3.71, "learning_rate": 6.07963163400339e-07, "loss": 0.1563, "step": 36864 }, { "epoch": 3.71, "learning_rate": 6.074839045852187e-07, "loss": 0.0926, "step": 36865 }, { "epoch": 3.71, "learning_rate": 6.070048328128997e-07, "loss": 0.2798, "step": 36866 }, { "epoch": 3.71, "learning_rate": 6.065259480864128e-07, "loss": 0.1391, "step": 36867 }, { "epoch": 3.71, "learning_rate": 6.060472504088122e-07, "loss": 0.2243, "step": 36868 }, { "epoch": 3.71, "learning_rate": 6.055687397831488e-07, "loss": 0.2252, "step": 36869 }, { "epoch": 3.71, "learning_rate": 6.050904162124637e-07, "loss": 0.0528, "step": 36870 }, { "epoch": 3.71, "learning_rate": 6.046122796997977e-07, "loss": 0.1121, "step": 36871 }, { "epoch": 3.71, "learning_rate": 6.041343302481983e-07, "loss": 0.0557, "step": 36872 }, { "epoch": 3.71, "learning_rate": 6.036565678607064e-07, "loss": 0.6692, "step": 36873 }, { "epoch": 3.71, "learning_rate": 6.031789925403597e-07, "loss": 0.1169, "step": 36874 }, { "epoch": 3.71, "learning_rate": 6.02701604290199e-07, "loss": 0.093, "step": 36875 }, { "epoch": 3.71, "learning_rate": 6.022244031132618e-07, "loss": 0.0827, "step": 36876 }, { "epoch": 3.71, "learning_rate": 6.017473890125857e-07, "loss": 0.1726, "step": 36877 }, { "epoch": 3.71, "learning_rate": 6.012705619912052e-07, "loss": 0.1788, "step": 36878 }, { "epoch": 3.71, "learning_rate": 6.007939220521574e-07, "loss": 0.2009, "step": 36879 }, { "epoch": 3.71, "learning_rate": 6.003174691984703e-07, "loss": 0.1753, "step": 36880 }, { "epoch": 3.71, "learning_rate": 5.99841203433178e-07, "loss": 0.1766, "step": 36881 }, { "epoch": 3.71, "learning_rate": 5.993651247593146e-07, "loss": 0.1009, "step": 36882 }, { "epoch": 3.71, "learning_rate": 5.988892331799012e-07, "loss": 0.1674, "step": 36883 }, { "epoch": 3.71, "learning_rate": 5.984135286979786e-07, "loss": 0.3672, "step": 36884 }, { "epoch": 3.71, "learning_rate": 5.979380113165644e-07, "loss": 0.1059, "step": 36885 }, { "epoch": 3.71, "learning_rate": 5.974626810386863e-07, "loss": 0.1461, "step": 36886 }, { "epoch": 3.71, "learning_rate": 5.969875378673717e-07, "loss": 0.6072, "step": 36887 }, { "epoch": 3.71, "learning_rate": 5.965125818056449e-07, "loss": 0.2201, "step": 36888 }, { "epoch": 3.71, "learning_rate": 5.960378128565236e-07, "loss": 0.2091, "step": 36889 }, { "epoch": 3.71, "learning_rate": 5.955632310230353e-07, "loss": 0.0296, "step": 36890 }, { "epoch": 3.71, "learning_rate": 5.950888363081941e-07, "loss": 0.0527, "step": 36891 }, { "epoch": 3.71, "learning_rate": 5.946146287150212e-07, "loss": 0.2761, "step": 36892 }, { "epoch": 3.71, "learning_rate": 5.941406082465339e-07, "loss": 0.1639, "step": 36893 }, { "epoch": 3.71, "learning_rate": 5.9366677490575e-07, "loss": 0.1145, "step": 36894 }, { "epoch": 3.71, "learning_rate": 5.931931286956838e-07, "loss": 0.3354, "step": 36895 }, { "epoch": 3.71, "learning_rate": 5.927196696193493e-07, "loss": 0.2317, "step": 36896 }, { "epoch": 3.71, "learning_rate": 5.922463976797577e-07, "loss": 0.2146, "step": 36897 }, { "epoch": 3.71, "learning_rate": 5.917733128799297e-07, "loss": 0.122, "step": 36898 }, { "epoch": 3.71, "learning_rate": 5.913004152228629e-07, "loss": 0.1143, "step": 36899 }, { "epoch": 3.71, "learning_rate": 5.908277047115718e-07, "loss": 0.1046, "step": 36900 }, { "epoch": 3.71, "learning_rate": 5.903551813490704e-07, "loss": 0.0757, "step": 36901 }, { "epoch": 3.71, "learning_rate": 5.898828451383597e-07, "loss": 0.0244, "step": 36902 }, { "epoch": 3.71, "learning_rate": 5.894106960824408e-07, "loss": 0.1727, "step": 36903 }, { "epoch": 3.71, "learning_rate": 5.889387341843277e-07, "loss": 0.2069, "step": 36904 }, { "epoch": 3.71, "learning_rate": 5.884669594470216e-07, "loss": 0.1554, "step": 36905 }, { "epoch": 3.71, "learning_rate": 5.879953718735231e-07, "loss": 0.0932, "step": 36906 }, { "epoch": 3.71, "learning_rate": 5.875239714668302e-07, "loss": 0.2814, "step": 36907 }, { "epoch": 3.71, "learning_rate": 5.870527582299468e-07, "loss": 0.3235, "step": 36908 }, { "epoch": 3.71, "learning_rate": 5.865817321658706e-07, "loss": 0.0962, "step": 36909 }, { "epoch": 3.71, "learning_rate": 5.861108932775994e-07, "loss": 0.5629, "step": 36910 }, { "epoch": 3.71, "learning_rate": 5.856402415681273e-07, "loss": 0.083, "step": 36911 }, { "epoch": 3.71, "learning_rate": 5.851697770404518e-07, "loss": 0.3162, "step": 36912 }, { "epoch": 3.71, "learning_rate": 5.846994996975674e-07, "loss": 0.196, "step": 36913 }, { "epoch": 3.71, "learning_rate": 5.842294095424616e-07, "loss": 0.3443, "step": 36914 }, { "epoch": 3.71, "learning_rate": 5.837595065781321e-07, "loss": 0.3467, "step": 36915 }, { "epoch": 3.71, "learning_rate": 5.832897908075696e-07, "loss": 0.0767, "step": 36916 }, { "epoch": 3.71, "learning_rate": 5.82820262233752e-07, "loss": 0.1444, "step": 36917 }, { "epoch": 3.71, "learning_rate": 5.8235092085968e-07, "loss": 0.542, "step": 36918 }, { "epoch": 3.71, "learning_rate": 5.818817666883381e-07, "loss": 0.7194, "step": 36919 }, { "epoch": 3.71, "learning_rate": 5.81412799722707e-07, "loss": 0.3144, "step": 36920 }, { "epoch": 3.71, "learning_rate": 5.809440199657745e-07, "loss": 0.0727, "step": 36921 }, { "epoch": 3.71, "learning_rate": 5.804754274205215e-07, "loss": 0.028, "step": 36922 }, { "epoch": 3.71, "learning_rate": 5.800070220899323e-07, "loss": 0.3765, "step": 36923 }, { "epoch": 3.71, "learning_rate": 5.795388039769845e-07, "loss": 0.319, "step": 36924 }, { "epoch": 3.71, "learning_rate": 5.79070773084659e-07, "loss": 0.0208, "step": 36925 }, { "epoch": 3.71, "learning_rate": 5.786029294159367e-07, "loss": 0.5159, "step": 36926 }, { "epoch": 3.71, "learning_rate": 5.781352729737954e-07, "loss": 0.1724, "step": 36927 }, { "epoch": 3.71, "learning_rate": 5.776678037612059e-07, "loss": 0.3996, "step": 36928 }, { "epoch": 3.71, "learning_rate": 5.772005217811427e-07, "loss": 0.1254, "step": 36929 }, { "epoch": 3.71, "learning_rate": 5.767334270365865e-07, "loss": 0.3779, "step": 36930 }, { "epoch": 3.71, "learning_rate": 5.76266519530505e-07, "loss": 0.318, "step": 36931 }, { "epoch": 3.71, "learning_rate": 5.757997992658659e-07, "loss": 0.2262, "step": 36932 }, { "epoch": 3.71, "learning_rate": 5.753332662456467e-07, "loss": 0.5826, "step": 36933 }, { "epoch": 3.71, "learning_rate": 5.748669204728185e-07, "loss": 0.1455, "step": 36934 }, { "epoch": 3.71, "learning_rate": 5.744007619503356e-07, "loss": 0.1073, "step": 36935 }, { "epoch": 3.71, "learning_rate": 5.739347906811721e-07, "loss": 0.1462, "step": 36936 }, { "epoch": 3.71, "learning_rate": 5.734690066682958e-07, "loss": 0.2181, "step": 36937 }, { "epoch": 3.71, "learning_rate": 5.730034099146708e-07, "loss": 0.0789, "step": 36938 }, { "epoch": 3.71, "learning_rate": 5.72538000423255e-07, "loss": 0.3543, "step": 36939 }, { "epoch": 3.71, "learning_rate": 5.720727781970125e-07, "loss": 0.3762, "step": 36940 }, { "epoch": 3.71, "learning_rate": 5.716077432389011e-07, "loss": 0.0754, "step": 36941 }, { "epoch": 3.71, "learning_rate": 5.711428955518882e-07, "loss": 0.1603, "step": 36942 }, { "epoch": 3.71, "learning_rate": 5.706782351389217e-07, "loss": 0.0696, "step": 36943 }, { "epoch": 3.71, "learning_rate": 5.702137620029657e-07, "loss": 0.3276, "step": 36944 }, { "epoch": 3.71, "learning_rate": 5.697494761469745e-07, "loss": 0.1346, "step": 36945 }, { "epoch": 3.71, "learning_rate": 5.692853775738993e-07, "loss": 0.0338, "step": 36946 }, { "epoch": 3.71, "learning_rate": 5.688214662866975e-07, "loss": 0.2785, "step": 36947 }, { "epoch": 3.71, "learning_rate": 5.683577422883235e-07, "loss": 0.1717, "step": 36948 }, { "epoch": 3.71, "learning_rate": 5.678942055817182e-07, "loss": 0.4594, "step": 36949 }, { "epoch": 3.72, "learning_rate": 5.674308561698394e-07, "loss": 0.4073, "step": 36950 }, { "epoch": 3.72, "learning_rate": 5.669676940556346e-07, "loss": 0.0727, "step": 36951 }, { "epoch": 3.72, "learning_rate": 5.665047192420514e-07, "loss": 0.1127, "step": 36952 }, { "epoch": 3.72, "learning_rate": 5.660419317320343e-07, "loss": 0.3887, "step": 36953 }, { "epoch": 3.72, "learning_rate": 5.655793315285307e-07, "loss": 0.2966, "step": 36954 }, { "epoch": 3.72, "learning_rate": 5.651169186344817e-07, "loss": 0.0811, "step": 36955 }, { "epoch": 3.72, "learning_rate": 5.646546930528284e-07, "loss": 0.0885, "step": 36956 }, { "epoch": 3.72, "learning_rate": 5.641926547865183e-07, "loss": 0.3571, "step": 36957 }, { "epoch": 3.72, "learning_rate": 5.637308038384859e-07, "loss": 0.3749, "step": 36958 }, { "epoch": 3.72, "learning_rate": 5.632691402116752e-07, "loss": 0.1027, "step": 36959 }, { "epoch": 3.72, "learning_rate": 5.628076639090207e-07, "loss": 0.316, "step": 36960 }, { "epoch": 3.72, "learning_rate": 5.623463749334535e-07, "loss": 0.065, "step": 36961 }, { "epoch": 3.72, "learning_rate": 5.618852732879243e-07, "loss": 0.2765, "step": 36962 }, { "epoch": 3.72, "learning_rate": 5.614243589753543e-07, "loss": 0.0564, "step": 36963 }, { "epoch": 3.72, "learning_rate": 5.609636319986811e-07, "loss": 0.0546, "step": 36964 }, { "epoch": 3.72, "learning_rate": 5.605030923608357e-07, "loss": 0.0646, "step": 36965 }, { "epoch": 3.72, "learning_rate": 5.600427400647524e-07, "loss": 0.1351, "step": 36966 }, { "epoch": 3.72, "learning_rate": 5.595825751133521e-07, "loss": 0.131, "step": 36967 }, { "epoch": 3.72, "learning_rate": 5.591225975095693e-07, "loss": 0.1378, "step": 36968 }, { "epoch": 3.72, "learning_rate": 5.586628072563348e-07, "loss": 0.137, "step": 36969 }, { "epoch": 3.72, "learning_rate": 5.582032043565666e-07, "loss": 0.0994, "step": 36970 }, { "epoch": 3.72, "learning_rate": 5.577437888131954e-07, "loss": 0.1765, "step": 36971 }, { "epoch": 3.72, "learning_rate": 5.572845606291387e-07, "loss": 0.1433, "step": 36972 }, { "epoch": 3.72, "learning_rate": 5.56825519807328e-07, "loss": 0.2227, "step": 36973 }, { "epoch": 3.72, "learning_rate": 5.563666663506739e-07, "loss": 0.1438, "step": 36974 }, { "epoch": 3.72, "learning_rate": 5.559080002621008e-07, "loss": 0.0737, "step": 36975 }, { "epoch": 3.72, "learning_rate": 5.554495215445332e-07, "loss": 0.1014, "step": 36976 }, { "epoch": 3.72, "learning_rate": 5.549912302008819e-07, "loss": 0.4617, "step": 36977 }, { "epoch": 3.72, "learning_rate": 5.545331262340614e-07, "loss": 0.7602, "step": 36978 }, { "epoch": 3.72, "learning_rate": 5.540752096469925e-07, "loss": 0.1714, "step": 36979 }, { "epoch": 3.72, "learning_rate": 5.536174804425897e-07, "loss": 0.2002, "step": 36980 }, { "epoch": 3.72, "learning_rate": 5.531599386237607e-07, "loss": 0.167, "step": 36981 }, { "epoch": 3.72, "learning_rate": 5.527025841934197e-07, "loss": 0.4709, "step": 36982 }, { "epoch": 3.72, "learning_rate": 5.522454171544777e-07, "loss": 0.343, "step": 36983 }, { "epoch": 3.72, "learning_rate": 5.517884375098459e-07, "loss": 0.2545, "step": 36984 }, { "epoch": 3.72, "learning_rate": 5.51331645262425e-07, "loss": 0.1835, "step": 36985 }, { "epoch": 3.72, "learning_rate": 5.508750404151297e-07, "loss": 0.2216, "step": 36986 }, { "epoch": 3.72, "learning_rate": 5.504186229708608e-07, "loss": 0.3015, "step": 36987 }, { "epoch": 3.72, "learning_rate": 5.49962392932526e-07, "loss": 0.1211, "step": 36988 }, { "epoch": 3.72, "learning_rate": 5.495063503030296e-07, "loss": 0.2224, "step": 36989 }, { "epoch": 3.72, "learning_rate": 5.49050495085266e-07, "loss": 0.1159, "step": 36990 }, { "epoch": 3.72, "learning_rate": 5.485948272821461e-07, "loss": 0.1646, "step": 36991 }, { "epoch": 3.72, "learning_rate": 5.481393468965612e-07, "loss": 0.311, "step": 36992 }, { "epoch": 3.72, "learning_rate": 5.476840539314121e-07, "loss": 0.1513, "step": 36993 }, { "epoch": 3.72, "learning_rate": 5.472289483896032e-07, "loss": 0.4936, "step": 36994 }, { "epoch": 3.72, "learning_rate": 5.467740302740188e-07, "loss": 0.62, "step": 36995 }, { "epoch": 3.72, "learning_rate": 5.463192995875632e-07, "loss": 0.3761, "step": 36996 }, { "epoch": 3.72, "learning_rate": 5.458647563331276e-07, "loss": 0.6497, "step": 36997 }, { "epoch": 3.72, "learning_rate": 5.454104005136029e-07, "loss": 0.2505, "step": 36998 }, { "epoch": 3.72, "learning_rate": 5.449562321318802e-07, "loss": 0.4071, "step": 36999 }, { "epoch": 3.72, "learning_rate": 5.445022511908537e-07, "loss": 0.3955, "step": 37000 }, { "epoch": 3.72, "learning_rate": 5.440484576934046e-07, "loss": 0.2605, "step": 37001 }, { "epoch": 3.72, "learning_rate": 5.435948516424305e-07, "loss": 0.2942, "step": 37002 }, { "epoch": 3.72, "learning_rate": 5.431414330408091e-07, "loss": 0.2618, "step": 37003 }, { "epoch": 3.72, "learning_rate": 5.426882018914314e-07, "loss": 0.089, "step": 37004 }, { "epoch": 3.72, "learning_rate": 5.422351581971819e-07, "loss": 0.0808, "step": 37005 }, { "epoch": 3.72, "learning_rate": 5.41782301960938e-07, "loss": 0.348, "step": 37006 }, { "epoch": 3.72, "learning_rate": 5.413296331855843e-07, "loss": 0.2667, "step": 37007 }, { "epoch": 3.72, "learning_rate": 5.408771518740085e-07, "loss": 0.2262, "step": 37008 }, { "epoch": 3.72, "learning_rate": 5.404248580290782e-07, "loss": 0.1049, "step": 37009 }, { "epoch": 3.72, "learning_rate": 5.399727516536778e-07, "loss": 0.3053, "step": 37010 }, { "epoch": 3.72, "learning_rate": 5.395208327506817e-07, "loss": 0.1154, "step": 37011 }, { "epoch": 3.72, "learning_rate": 5.390691013229743e-07, "loss": 0.5716, "step": 37012 }, { "epoch": 3.72, "learning_rate": 5.386175573734165e-07, "loss": 0.2644, "step": 37013 }, { "epoch": 3.72, "learning_rate": 5.381662009048927e-07, "loss": 0.6717, "step": 37014 }, { "epoch": 3.72, "learning_rate": 5.377150319202706e-07, "loss": 0.4176, "step": 37015 }, { "epoch": 3.72, "learning_rate": 5.372640504224213e-07, "loss": 0.023, "step": 37016 }, { "epoch": 3.72, "learning_rate": 5.368132564142158e-07, "loss": 0.1116, "step": 37017 }, { "epoch": 3.72, "learning_rate": 5.363626498985252e-07, "loss": 0.0638, "step": 37018 }, { "epoch": 3.72, "learning_rate": 5.359122308782105e-07, "loss": 0.3278, "step": 37019 }, { "epoch": 3.72, "learning_rate": 5.354619993561427e-07, "loss": 0.1452, "step": 37020 }, { "epoch": 3.72, "learning_rate": 5.350119553351862e-07, "loss": 0.1908, "step": 37021 }, { "epoch": 3.72, "learning_rate": 5.345620988182054e-07, "loss": 0.2623, "step": 37022 }, { "epoch": 3.72, "learning_rate": 5.341124298080613e-07, "loss": 0.3814, "step": 37023 }, { "epoch": 3.72, "learning_rate": 5.336629483076117e-07, "loss": 0.0743, "step": 37024 }, { "epoch": 3.72, "learning_rate": 5.332136543197274e-07, "loss": 0.084, "step": 37025 }, { "epoch": 3.72, "learning_rate": 5.327645478472632e-07, "loss": 0.1, "step": 37026 }, { "epoch": 3.72, "learning_rate": 5.323156288930697e-07, "loss": 0.1718, "step": 37027 }, { "epoch": 3.72, "learning_rate": 5.318668974600116e-07, "loss": 0.181, "step": 37028 }, { "epoch": 3.72, "learning_rate": 5.314183535509431e-07, "loss": 0.593, "step": 37029 }, { "epoch": 3.72, "learning_rate": 5.309699971687188e-07, "loss": 0.2263, "step": 37030 }, { "epoch": 3.72, "learning_rate": 5.305218283161895e-07, "loss": 0.0762, "step": 37031 }, { "epoch": 3.72, "learning_rate": 5.300738469962096e-07, "loss": 0.1907, "step": 37032 }, { "epoch": 3.72, "learning_rate": 5.296260532116271e-07, "loss": 0.0802, "step": 37033 }, { "epoch": 3.72, "learning_rate": 5.291784469652927e-07, "loss": 0.3881, "step": 37034 }, { "epoch": 3.72, "learning_rate": 5.287310282600577e-07, "loss": 0.2791, "step": 37035 }, { "epoch": 3.72, "learning_rate": 5.282837970987631e-07, "loss": 0.035, "step": 37036 }, { "epoch": 3.72, "learning_rate": 5.278367534842665e-07, "loss": 0.4257, "step": 37037 }, { "epoch": 3.72, "learning_rate": 5.273898974193992e-07, "loss": 0.1472, "step": 37038 }, { "epoch": 3.72, "learning_rate": 5.269432289070086e-07, "loss": 0.2065, "step": 37039 }, { "epoch": 3.72, "learning_rate": 5.264967479499461e-07, "loss": 0.5238, "step": 37040 }, { "epoch": 3.72, "learning_rate": 5.260504545510425e-07, "loss": 0.3306, "step": 37041 }, { "epoch": 3.72, "learning_rate": 5.25604348713139e-07, "loss": 0.0643, "step": 37042 }, { "epoch": 3.72, "learning_rate": 5.251584304390766e-07, "loss": 0.2546, "step": 37043 }, { "epoch": 3.72, "learning_rate": 5.247126997316998e-07, "loss": 0.0959, "step": 37044 }, { "epoch": 3.72, "learning_rate": 5.242671565938295e-07, "loss": 0.6208, "step": 37045 }, { "epoch": 3.72, "learning_rate": 5.238218010283169e-07, "loss": 0.6293, "step": 37046 }, { "epoch": 3.72, "learning_rate": 5.233766330379863e-07, "loss": 0.2824, "step": 37047 }, { "epoch": 3.73, "learning_rate": 5.229316526256723e-07, "loss": 0.0681, "step": 37048 }, { "epoch": 3.73, "learning_rate": 5.22486859794209e-07, "loss": 0.1517, "step": 37049 }, { "epoch": 3.73, "learning_rate": 5.220422545464242e-07, "loss": 0.2905, "step": 37050 }, { "epoch": 3.73, "learning_rate": 5.215978368851526e-07, "loss": 0.1668, "step": 37051 }, { "epoch": 3.73, "learning_rate": 5.21153606813215e-07, "loss": 0.1651, "step": 37052 }, { "epoch": 3.73, "learning_rate": 5.207095643334392e-07, "loss": 0.0337, "step": 37053 }, { "epoch": 3.73, "learning_rate": 5.202657094486562e-07, "loss": 0.0614, "step": 37054 }, { "epoch": 3.73, "learning_rate": 5.198220421616873e-07, "loss": 0.3542, "step": 37055 }, { "epoch": 3.73, "learning_rate": 5.193785624753566e-07, "loss": 0.1201, "step": 37056 }, { "epoch": 3.73, "learning_rate": 5.189352703924855e-07, "loss": 0.0532, "step": 37057 }, { "epoch": 3.73, "learning_rate": 5.184921659158948e-07, "loss": 0.1148, "step": 37058 }, { "epoch": 3.73, "learning_rate": 5.180492490484023e-07, "loss": 0.07, "step": 37059 }, { "epoch": 3.73, "learning_rate": 5.176065197928325e-07, "loss": 0.1447, "step": 37060 }, { "epoch": 3.73, "learning_rate": 5.171639781519966e-07, "loss": 0.6098, "step": 37061 }, { "epoch": 3.73, "learning_rate": 5.167216241287154e-07, "loss": 0.1634, "step": 37062 }, { "epoch": 3.73, "learning_rate": 5.162794577258001e-07, "loss": 0.0311, "step": 37063 }, { "epoch": 3.73, "learning_rate": 5.158374789460685e-07, "loss": 0.1707, "step": 37064 }, { "epoch": 3.73, "learning_rate": 5.153956877923283e-07, "loss": 0.2056, "step": 37065 }, { "epoch": 3.73, "learning_rate": 5.149540842673938e-07, "loss": 0.0389, "step": 37066 }, { "epoch": 3.73, "learning_rate": 5.145126683740764e-07, "loss": 0.1619, "step": 37067 }, { "epoch": 3.73, "learning_rate": 5.140714401151802e-07, "loss": 0.1172, "step": 37068 }, { "epoch": 3.73, "learning_rate": 5.136303994935198e-07, "loss": 0.1036, "step": 37069 }, { "epoch": 3.73, "learning_rate": 5.131895465118996e-07, "loss": 0.2085, "step": 37070 }, { "epoch": 3.73, "learning_rate": 5.127488811731174e-07, "loss": 0.2818, "step": 37071 }, { "epoch": 3.73, "learning_rate": 5.123084034799907e-07, "loss": 0.1162, "step": 37072 }, { "epoch": 3.73, "learning_rate": 5.118681134353142e-07, "loss": 0.1373, "step": 37073 }, { "epoch": 3.73, "learning_rate": 5.114280110418856e-07, "loss": 0.3251, "step": 37074 }, { "epoch": 3.73, "learning_rate": 5.10988096302516e-07, "loss": 0.0523, "step": 37075 }, { "epoch": 3.73, "learning_rate": 5.10548369220003e-07, "loss": 0.1715, "step": 37076 }, { "epoch": 3.73, "learning_rate": 5.101088297971346e-07, "loss": 0.1493, "step": 37077 }, { "epoch": 3.73, "learning_rate": 5.096694780367183e-07, "loss": 0.2789, "step": 37078 }, { "epoch": 3.73, "learning_rate": 5.092303139415455e-07, "loss": 0.108, "step": 37079 }, { "epoch": 3.73, "learning_rate": 5.087913375144137e-07, "loss": 0.1536, "step": 37080 }, { "epoch": 3.73, "learning_rate": 5.083525487581142e-07, "loss": 0.3503, "step": 37081 }, { "epoch": 3.73, "learning_rate": 5.07913947675438e-07, "loss": 0.1025, "step": 37082 }, { "epoch": 3.73, "learning_rate": 5.074755342691828e-07, "loss": 0.3456, "step": 37083 }, { "epoch": 3.73, "learning_rate": 5.070373085421332e-07, "loss": 0.2023, "step": 37084 }, { "epoch": 3.73, "learning_rate": 5.065992704970734e-07, "loss": 0.0846, "step": 37085 }, { "epoch": 3.73, "learning_rate": 5.061614201368015e-07, "loss": 0.0695, "step": 37086 }, { "epoch": 3.73, "learning_rate": 5.057237574640982e-07, "loss": 0.0446, "step": 37087 }, { "epoch": 3.73, "learning_rate": 5.052862824817417e-07, "loss": 0.1122, "step": 37088 }, { "epoch": 3.73, "learning_rate": 5.048489951925295e-07, "loss": 0.1681, "step": 37089 }, { "epoch": 3.73, "learning_rate": 5.044118955992394e-07, "loss": 0.131, "step": 37090 }, { "epoch": 3.73, "learning_rate": 5.039749837046425e-07, "loss": 0.1854, "step": 37091 }, { "epoch": 3.73, "learning_rate": 5.035382595115335e-07, "loss": 0.2153, "step": 37092 }, { "epoch": 3.73, "learning_rate": 5.031017230226864e-07, "loss": 0.5448, "step": 37093 }, { "epoch": 3.73, "learning_rate": 5.026653742408794e-07, "loss": 0.0416, "step": 37094 }, { "epoch": 3.73, "learning_rate": 5.022292131688866e-07, "loss": 0.106, "step": 37095 }, { "epoch": 3.73, "learning_rate": 5.01793239809486e-07, "loss": 0.0863, "step": 37096 }, { "epoch": 3.73, "learning_rate": 5.01357454165452e-07, "loss": 0.0328, "step": 37097 }, { "epoch": 3.73, "learning_rate": 5.00921856239559e-07, "loss": 0.6451, "step": 37098 }, { "epoch": 3.73, "learning_rate": 5.004864460345748e-07, "loss": 0.138, "step": 37099 }, { "epoch": 3.73, "learning_rate": 5.000512235532706e-07, "loss": 0.1118, "step": 37100 }, { "epoch": 3.73, "learning_rate": 4.996161887984241e-07, "loss": 0.1782, "step": 37101 }, { "epoch": 3.73, "learning_rate": 4.99181341772793e-07, "loss": 0.1463, "step": 37102 }, { "epoch": 3.73, "learning_rate": 4.987466824791487e-07, "loss": 0.3313, "step": 37103 }, { "epoch": 3.73, "learning_rate": 4.98312210920262e-07, "loss": 0.1791, "step": 37104 }, { "epoch": 3.73, "learning_rate": 4.978779270988876e-07, "loss": 0.0639, "step": 37105 }, { "epoch": 3.73, "learning_rate": 4.974438310177998e-07, "loss": 0.114, "step": 37106 }, { "epoch": 3.73, "learning_rate": 4.970099226797531e-07, "loss": 0.1336, "step": 37107 }, { "epoch": 3.73, "learning_rate": 4.965762020875153e-07, "loss": 0.0559, "step": 37108 }, { "epoch": 3.73, "learning_rate": 4.961426692438409e-07, "loss": 0.0598, "step": 37109 }, { "epoch": 3.73, "learning_rate": 4.957093241514876e-07, "loss": 0.1601, "step": 37110 }, { "epoch": 3.73, "learning_rate": 4.9527616681322e-07, "loss": 0.035, "step": 37111 }, { "epoch": 3.73, "learning_rate": 4.948431972317891e-07, "loss": 0.1593, "step": 37112 }, { "epoch": 3.73, "learning_rate": 4.944104154099527e-07, "loss": 0.3817, "step": 37113 }, { "epoch": 3.73, "learning_rate": 4.93977821350462e-07, "loss": 0.1303, "step": 37114 }, { "epoch": 3.73, "learning_rate": 4.935454150560748e-07, "loss": 0.2038, "step": 37115 }, { "epoch": 3.73, "learning_rate": 4.931131965295387e-07, "loss": 0.1121, "step": 37116 }, { "epoch": 3.73, "learning_rate": 4.926811657736019e-07, "loss": 0.066, "step": 37117 }, { "epoch": 3.73, "learning_rate": 4.922493227910252e-07, "loss": 0.0127, "step": 37118 }, { "epoch": 3.73, "learning_rate": 4.918176675845432e-07, "loss": 0.0632, "step": 37119 }, { "epoch": 3.73, "learning_rate": 4.913862001569036e-07, "loss": 0.2838, "step": 37120 }, { "epoch": 3.73, "learning_rate": 4.90954920510861e-07, "loss": 0.0802, "step": 37121 }, { "epoch": 3.73, "learning_rate": 4.905238286491597e-07, "loss": 0.1046, "step": 37122 }, { "epoch": 3.73, "learning_rate": 4.900929245745312e-07, "loss": 0.0983, "step": 37123 }, { "epoch": 3.73, "learning_rate": 4.896622082897295e-07, "loss": 0.0547, "step": 37124 }, { "epoch": 3.73, "learning_rate": 4.892316797974894e-07, "loss": 0.0992, "step": 37125 }, { "epoch": 3.73, "learning_rate": 4.888013391005553e-07, "loss": 0.2316, "step": 37126 }, { "epoch": 3.73, "learning_rate": 4.883711862016583e-07, "loss": 0.1171, "step": 37127 }, { "epoch": 3.73, "learning_rate": 4.879412211035427e-07, "loss": 0.2696, "step": 37128 }, { "epoch": 3.73, "learning_rate": 4.875114438089435e-07, "loss": 0.1717, "step": 37129 }, { "epoch": 3.73, "learning_rate": 4.870818543205946e-07, "loss": 0.1171, "step": 37130 }, { "epoch": 3.73, "learning_rate": 4.866524526412242e-07, "loss": 0.0451, "step": 37131 }, { "epoch": 3.73, "learning_rate": 4.862232387735732e-07, "loss": 0.0442, "step": 37132 }, { "epoch": 3.73, "learning_rate": 4.85794212720373e-07, "loss": 0.0877, "step": 37133 }, { "epoch": 3.73, "learning_rate": 4.853653744843479e-07, "loss": 0.2228, "step": 37134 }, { "epoch": 3.73, "learning_rate": 4.84936724068229e-07, "loss": 0.0552, "step": 37135 }, { "epoch": 3.73, "learning_rate": 4.845082614747475e-07, "loss": 0.3663, "step": 37136 }, { "epoch": 3.73, "learning_rate": 4.840799867066248e-07, "loss": 0.252, "step": 37137 }, { "epoch": 3.73, "learning_rate": 4.836518997665885e-07, "loss": 0.0776, "step": 37138 }, { "epoch": 3.73, "learning_rate": 4.832240006573629e-07, "loss": 0.655, "step": 37139 }, { "epoch": 3.73, "learning_rate": 4.827962893816729e-07, "loss": 0.1169, "step": 37140 }, { "epoch": 3.73, "learning_rate": 4.823687659422394e-07, "loss": 0.3091, "step": 37141 }, { "epoch": 3.73, "learning_rate": 4.819414303417769e-07, "loss": 0.1039, "step": 37142 }, { "epoch": 3.73, "learning_rate": 4.815142825830132e-07, "loss": 0.5375, "step": 37143 }, { "epoch": 3.73, "learning_rate": 4.81087322668663e-07, "loss": 0.1381, "step": 37144 }, { "epoch": 3.73, "learning_rate": 4.806605506014438e-07, "loss": 0.0957, "step": 37145 }, { "epoch": 3.73, "learning_rate": 4.802339663840705e-07, "loss": 0.1405, "step": 37146 }, { "epoch": 3.74, "learning_rate": 4.798075700192606e-07, "loss": 0.4871, "step": 37147 }, { "epoch": 3.74, "learning_rate": 4.79381361509722e-07, "loss": 0.2277, "step": 37148 }, { "epoch": 3.74, "learning_rate": 4.789553408581693e-07, "loss": 0.3666, "step": 37149 }, { "epoch": 3.74, "learning_rate": 4.785295080673202e-07, "loss": 0.0846, "step": 37150 }, { "epoch": 3.74, "learning_rate": 4.781038631398726e-07, "loss": 0.0759, "step": 37151 }, { "epoch": 3.74, "learning_rate": 4.776784060785444e-07, "loss": 0.0688, "step": 37152 }, { "epoch": 3.74, "learning_rate": 4.772531368860367e-07, "loss": 0.1679, "step": 37153 }, { "epoch": 3.74, "learning_rate": 4.7682805556506394e-07, "loss": 0.0182, "step": 37154 }, { "epoch": 3.74, "learning_rate": 4.7640316211832403e-07, "loss": 0.492, "step": 37155 }, { "epoch": 3.74, "learning_rate": 4.7597845654852145e-07, "loss": 0.2013, "step": 37156 }, { "epoch": 3.74, "learning_rate": 4.7555393885836406e-07, "loss": 0.2128, "step": 37157 }, { "epoch": 3.74, "learning_rate": 4.751296090505464e-07, "loss": 0.1691, "step": 37158 }, { "epoch": 3.74, "learning_rate": 4.7470546712777287e-07, "loss": 0.3315, "step": 37159 }, { "epoch": 3.74, "learning_rate": 4.742815130927414e-07, "loss": 0.6177, "step": 37160 }, { "epoch": 3.74, "learning_rate": 4.738577469481531e-07, "loss": 0.2808, "step": 37161 }, { "epoch": 3.74, "learning_rate": 4.734341686966992e-07, "loss": 0.4487, "step": 37162 }, { "epoch": 3.74, "learning_rate": 4.7301077834107755e-07, "loss": 0.3402, "step": 37163 }, { "epoch": 3.74, "learning_rate": 4.725875758839826e-07, "loss": 0.0752, "step": 37164 }, { "epoch": 3.74, "learning_rate": 4.7216456132810895e-07, "loss": 0.4997, "step": 37165 }, { "epoch": 3.74, "learning_rate": 4.717417346761443e-07, "loss": 0.4547, "step": 37166 }, { "epoch": 3.74, "learning_rate": 4.713190959307867e-07, "loss": 0.1368, "step": 37167 }, { "epoch": 3.74, "learning_rate": 4.7089664509472054e-07, "loss": 0.5021, "step": 37168 }, { "epoch": 3.74, "learning_rate": 4.704743821706303e-07, "loss": 0.0413, "step": 37169 }, { "epoch": 3.74, "learning_rate": 4.700523071612106e-07, "loss": 0.0348, "step": 37170 }, { "epoch": 3.74, "learning_rate": 4.696304200691426e-07, "loss": 0.0423, "step": 37171 }, { "epoch": 3.74, "learning_rate": 4.692087208971141e-07, "loss": 0.0853, "step": 37172 }, { "epoch": 3.74, "learning_rate": 4.687872096478063e-07, "loss": 0.1903, "step": 37173 }, { "epoch": 3.74, "learning_rate": 4.6836588632390377e-07, "loss": 0.2818, "step": 37174 }, { "epoch": 3.74, "learning_rate": 4.6794475092808433e-07, "loss": 0.2158, "step": 37175 }, { "epoch": 3.74, "learning_rate": 4.675238034630325e-07, "loss": 0.2421, "step": 37176 }, { "epoch": 3.74, "learning_rate": 4.6710304393142275e-07, "loss": 0.6264, "step": 37177 }, { "epoch": 3.74, "learning_rate": 4.6668247233593307e-07, "loss": 0.5268, "step": 37178 }, { "epoch": 3.74, "learning_rate": 4.6626208867924456e-07, "loss": 0.4121, "step": 37179 }, { "epoch": 3.74, "learning_rate": 4.6584189296402845e-07, "loss": 0.1274, "step": 37180 }, { "epoch": 3.74, "learning_rate": 4.654218851929559e-07, "loss": 0.3101, "step": 37181 }, { "epoch": 3.74, "learning_rate": 4.6500206536870483e-07, "loss": 0.0533, "step": 37182 }, { "epoch": 3.74, "learning_rate": 4.6458243349394634e-07, "loss": 0.5294, "step": 37183 }, { "epoch": 3.74, "learning_rate": 4.6416298957134504e-07, "loss": 0.1077, "step": 37184 }, { "epoch": 3.74, "learning_rate": 4.637437336035788e-07, "loss": 0.2338, "step": 37185 }, { "epoch": 3.74, "learning_rate": 4.6332466559330876e-07, "loss": 0.2132, "step": 37186 }, { "epoch": 3.74, "learning_rate": 4.6290578554320616e-07, "loss": 0.1289, "step": 37187 }, { "epoch": 3.74, "learning_rate": 4.624870934559322e-07, "loss": 0.039, "step": 37188 }, { "epoch": 3.74, "learning_rate": 4.6206858933415807e-07, "loss": 0.0279, "step": 37189 }, { "epoch": 3.74, "learning_rate": 4.6165027318053834e-07, "loss": 0.154, "step": 37190 }, { "epoch": 3.74, "learning_rate": 4.612321449977408e-07, "loss": 0.1996, "step": 37191 }, { "epoch": 3.74, "learning_rate": 4.608142047884234e-07, "loss": 0.3033, "step": 37192 }, { "epoch": 3.74, "learning_rate": 4.6039645255525065e-07, "loss": 0.5303, "step": 37193 }, { "epoch": 3.74, "learning_rate": 4.5997888830087707e-07, "loss": 0.1855, "step": 37194 }, { "epoch": 3.74, "learning_rate": 4.5956151202795394e-07, "loss": 0.1748, "step": 37195 }, { "epoch": 3.74, "learning_rate": 4.5914432373914904e-07, "loss": 0.1299, "step": 37196 }, { "epoch": 3.74, "learning_rate": 4.5872732343711364e-07, "loss": 0.1162, "step": 37197 }, { "epoch": 3.74, "learning_rate": 4.583105111244956e-07, "loss": 0.2292, "step": 37198 }, { "epoch": 3.74, "learning_rate": 4.578938868039495e-07, "loss": 0.0672, "step": 37199 }, { "epoch": 3.74, "learning_rate": 4.574774504781365e-07, "loss": 0.1122, "step": 37200 }, { "epoch": 3.74, "learning_rate": 4.5706120214968783e-07, "loss": 0.3452, "step": 37201 }, { "epoch": 3.74, "learning_rate": 4.566451418212714e-07, "loss": 0.0169, "step": 37202 }, { "epoch": 3.74, "learning_rate": 4.562292694955217e-07, "loss": 0.3393, "step": 37203 }, { "epoch": 3.74, "learning_rate": 4.558135851750933e-07, "loss": 0.0744, "step": 37204 }, { "epoch": 3.74, "learning_rate": 4.5539808886262414e-07, "loss": 0.11, "step": 37205 }, { "epoch": 3.74, "learning_rate": 4.5498278056076534e-07, "loss": 0.255, "step": 37206 }, { "epoch": 3.74, "learning_rate": 4.545676602721549e-07, "loss": 0.0999, "step": 37207 }, { "epoch": 3.74, "learning_rate": 4.541527279994373e-07, "loss": 0.1335, "step": 37208 }, { "epoch": 3.74, "learning_rate": 4.537379837452504e-07, "loss": 0.4021, "step": 37209 }, { "epoch": 3.74, "learning_rate": 4.5332342751223553e-07, "loss": 0.2387, "step": 37210 }, { "epoch": 3.74, "learning_rate": 4.5290905930303384e-07, "loss": 0.1264, "step": 37211 }, { "epoch": 3.74, "learning_rate": 4.524948791202732e-07, "loss": 0.2257, "step": 37212 }, { "epoch": 3.74, "learning_rate": 4.520808869665982e-07, "loss": 0.2815, "step": 37213 }, { "epoch": 3.74, "learning_rate": 4.516670828446401e-07, "loss": 0.087, "step": 37214 }, { "epoch": 3.74, "learning_rate": 4.5125346675702674e-07, "loss": 0.3793, "step": 37215 }, { "epoch": 3.74, "learning_rate": 4.5084003870639934e-07, "loss": 0.0307, "step": 37216 }, { "epoch": 3.74, "learning_rate": 4.5042679869538586e-07, "loss": 0.3451, "step": 37217 }, { "epoch": 3.74, "learning_rate": 4.500137467266141e-07, "loss": 0.2454, "step": 37218 }, { "epoch": 3.74, "learning_rate": 4.496008828027121e-07, "loss": 0.0216, "step": 37219 }, { "epoch": 3.74, "learning_rate": 4.491882069263109e-07, "loss": 0.0391, "step": 37220 }, { "epoch": 3.74, "learning_rate": 4.487757191000319e-07, "loss": 0.0717, "step": 37221 }, { "epoch": 3.74, "learning_rate": 4.4836341932650625e-07, "loss": 0.5235, "step": 37222 }, { "epoch": 3.74, "learning_rate": 4.4795130760834855e-07, "loss": 0.2468, "step": 37223 }, { "epoch": 3.74, "learning_rate": 4.475393839481867e-07, "loss": 0.3415, "step": 37224 }, { "epoch": 3.74, "learning_rate": 4.471276483486486e-07, "loss": 0.0664, "step": 37225 }, { "epoch": 3.74, "learning_rate": 4.4671610081234214e-07, "loss": 0.0769, "step": 37226 }, { "epoch": 3.74, "learning_rate": 4.4630474134189194e-07, "loss": 0.5915, "step": 37227 }, { "epoch": 3.74, "learning_rate": 4.4589356993991583e-07, "loss": 0.0901, "step": 37228 }, { "epoch": 3.74, "learning_rate": 4.454825866090317e-07, "loss": 0.1562, "step": 37229 }, { "epoch": 3.74, "learning_rate": 4.4507179135185096e-07, "loss": 0.1012, "step": 37230 }, { "epoch": 3.74, "learning_rate": 4.4466118417099135e-07, "loss": 0.0372, "step": 37231 }, { "epoch": 3.74, "learning_rate": 4.442507650690675e-07, "loss": 0.171, "step": 37232 }, { "epoch": 3.74, "learning_rate": 4.43840534048684e-07, "loss": 0.2067, "step": 37233 }, { "epoch": 3.74, "learning_rate": 4.4343049111245535e-07, "loss": 0.1979, "step": 37234 }, { "epoch": 3.74, "learning_rate": 4.430206362629929e-07, "loss": 0.0431, "step": 37235 }, { "epoch": 3.74, "learning_rate": 4.4261096950290457e-07, "loss": 0.2113, "step": 37236 }, { "epoch": 3.74, "learning_rate": 4.422014908347949e-07, "loss": 0.0966, "step": 37237 }, { "epoch": 3.74, "learning_rate": 4.4179220026126845e-07, "loss": 0.4536, "step": 37238 }, { "epoch": 3.74, "learning_rate": 4.413830977849331e-07, "loss": 0.1335, "step": 37239 }, { "epoch": 3.74, "learning_rate": 4.409741834083936e-07, "loss": 0.1867, "step": 37240 }, { "epoch": 3.74, "learning_rate": 4.4056545713424436e-07, "loss": 0.1301, "step": 37241 }, { "epoch": 3.74, "learning_rate": 4.4015691896509333e-07, "loss": 0.2041, "step": 37242 }, { "epoch": 3.74, "learning_rate": 4.3974856890354185e-07, "loss": 0.1499, "step": 37243 }, { "epoch": 3.74, "learning_rate": 4.3934040695218113e-07, "loss": 0.415, "step": 37244 }, { "epoch": 3.75, "learning_rate": 4.3893243311361577e-07, "loss": 0.1608, "step": 37245 }, { "epoch": 3.75, "learning_rate": 4.3852464739044027e-07, "loss": 0.183, "step": 37246 }, { "epoch": 3.75, "learning_rate": 4.381170497852427e-07, "loss": 0.168, "step": 37247 }, { "epoch": 3.75, "learning_rate": 4.377096403006242e-07, "loss": 0.3394, "step": 37248 }, { "epoch": 3.75, "learning_rate": 4.373024189391761e-07, "loss": 0.376, "step": 37249 }, { "epoch": 3.75, "learning_rate": 4.3689538570348965e-07, "loss": 0.6219, "step": 37250 }, { "epoch": 3.75, "learning_rate": 4.3648854059615604e-07, "loss": 0.2604, "step": 37251 }, { "epoch": 3.75, "learning_rate": 4.3608188361976e-07, "loss": 0.1338, "step": 37252 }, { "epoch": 3.75, "learning_rate": 4.3567541477689267e-07, "loss": 0.1414, "step": 37253 }, { "epoch": 3.75, "learning_rate": 4.35269134070142e-07, "loss": 0.062, "step": 37254 }, { "epoch": 3.75, "learning_rate": 4.3486304150208935e-07, "loss": 0.0445, "step": 37255 }, { "epoch": 3.75, "learning_rate": 4.344571370753192e-07, "loss": 0.3826, "step": 37256 }, { "epoch": 3.75, "learning_rate": 4.340514207924229e-07, "loss": 0.4274, "step": 37257 }, { "epoch": 3.75, "learning_rate": 4.3364589265597164e-07, "loss": 0.5447, "step": 37258 }, { "epoch": 3.75, "learning_rate": 4.332405526685501e-07, "loss": 0.3057, "step": 37259 }, { "epoch": 3.75, "learning_rate": 4.328354008327395e-07, "loss": 0.1427, "step": 37260 }, { "epoch": 3.75, "learning_rate": 4.3243043715111784e-07, "loss": 0.2867, "step": 37261 }, { "epoch": 3.75, "learning_rate": 4.320256616262563e-07, "loss": 0.1944, "step": 37262 }, { "epoch": 3.75, "learning_rate": 4.3162107426073625e-07, "loss": 0.2117, "step": 37263 }, { "epoch": 3.75, "learning_rate": 4.312166750571356e-07, "loss": 0.4405, "step": 37264 }, { "epoch": 3.75, "learning_rate": 4.308124640180189e-07, "loss": 0.6439, "step": 37265 }, { "epoch": 3.75, "learning_rate": 4.304084411459641e-07, "loss": 0.0637, "step": 37266 }, { "epoch": 3.75, "learning_rate": 4.3000460644354255e-07, "loss": 0.1804, "step": 37267 }, { "epoch": 3.75, "learning_rate": 4.2960095991332214e-07, "loss": 0.1652, "step": 37268 }, { "epoch": 3.75, "learning_rate": 4.291975015578709e-07, "loss": 0.3793, "step": 37269 }, { "epoch": 3.75, "learning_rate": 4.287942313797566e-07, "loss": 0.1653, "step": 37270 }, { "epoch": 3.75, "learning_rate": 4.283911493815507e-07, "loss": 0.3878, "step": 37271 }, { "epoch": 3.75, "learning_rate": 4.27988255565811e-07, "loss": 0.0651, "step": 37272 }, { "epoch": 3.75, "learning_rate": 4.275855499351022e-07, "loss": 0.4964, "step": 37273 }, { "epoch": 3.75, "learning_rate": 4.271830324919923e-07, "loss": 0.3032, "step": 37274 }, { "epoch": 3.75, "learning_rate": 4.267807032390425e-07, "loss": 0.1467, "step": 37275 }, { "epoch": 3.75, "learning_rate": 4.263785621788041e-07, "loss": 0.2847, "step": 37276 }, { "epoch": 3.75, "learning_rate": 4.2597660931384174e-07, "loss": 0.1656, "step": 37277 }, { "epoch": 3.75, "learning_rate": 4.2557484464672005e-07, "loss": 0.391, "step": 37278 }, { "epoch": 3.75, "learning_rate": 4.251732681799836e-07, "loss": 0.1404, "step": 37279 }, { "epoch": 3.75, "learning_rate": 4.2477187991619705e-07, "loss": 0.1391, "step": 37280 }, { "epoch": 3.75, "learning_rate": 4.2437067985790833e-07, "loss": 0.2892, "step": 37281 }, { "epoch": 3.75, "learning_rate": 4.2396966800767546e-07, "loss": 0.319, "step": 37282 }, { "epoch": 3.75, "learning_rate": 4.235688443680463e-07, "loss": 0.0739, "step": 37283 }, { "epoch": 3.75, "learning_rate": 4.231682089415756e-07, "loss": 0.0667, "step": 37284 }, { "epoch": 3.75, "learning_rate": 4.2276776173081124e-07, "loss": 0.4378, "step": 37285 }, { "epoch": 3.75, "learning_rate": 4.2236750273830116e-07, "loss": 0.1548, "step": 37286 }, { "epoch": 3.75, "learning_rate": 4.2196743196659007e-07, "loss": 0.0579, "step": 37287 }, { "epoch": 3.75, "learning_rate": 4.2156754941822584e-07, "loss": 0.1622, "step": 37288 }, { "epoch": 3.75, "learning_rate": 4.211678550957565e-07, "loss": 0.0829, "step": 37289 }, { "epoch": 3.75, "learning_rate": 4.2076834900172336e-07, "loss": 0.0811, "step": 37290 }, { "epoch": 3.75, "learning_rate": 4.2036903113866096e-07, "loss": 0.1817, "step": 37291 }, { "epoch": 3.75, "learning_rate": 4.1996990150912406e-07, "loss": 0.0763, "step": 37292 }, { "epoch": 3.75, "learning_rate": 4.195709601156439e-07, "loss": 0.0729, "step": 37293 }, { "epoch": 3.75, "learning_rate": 4.1917220696076175e-07, "loss": 0.3462, "step": 37294 }, { "epoch": 3.75, "learning_rate": 4.1877364204701564e-07, "loss": 0.147, "step": 37295 }, { "epoch": 3.75, "learning_rate": 4.183752653769368e-07, "loss": 0.2332, "step": 37296 }, { "epoch": 3.75, "learning_rate": 4.179770769530666e-07, "loss": 0.0245, "step": 37297 }, { "epoch": 3.75, "learning_rate": 4.175790767779397e-07, "loss": 0.0659, "step": 37298 }, { "epoch": 3.75, "learning_rate": 4.171812648540807e-07, "loss": 0.0903, "step": 37299 }, { "epoch": 3.75, "learning_rate": 4.167836411840309e-07, "loss": 0.2516, "step": 37300 }, { "epoch": 3.75, "learning_rate": 4.163862057703116e-07, "loss": 0.1677, "step": 37301 }, { "epoch": 3.75, "learning_rate": 4.159889586154575e-07, "loss": 0.1016, "step": 37302 }, { "epoch": 3.75, "learning_rate": 4.1559189972199986e-07, "loss": 0.0088, "step": 37303 }, { "epoch": 3.75, "learning_rate": 4.1519502909245664e-07, "loss": 0.2414, "step": 37304 }, { "epoch": 3.75, "learning_rate": 4.147983467293559e-07, "loss": 0.1738, "step": 37305 }, { "epoch": 3.75, "learning_rate": 4.144018526352256e-07, "loss": 0.4142, "step": 37306 }, { "epoch": 3.75, "learning_rate": 4.140055468125903e-07, "loss": 0.077, "step": 37307 }, { "epoch": 3.75, "learning_rate": 4.136094292639614e-07, "loss": 0.4102, "step": 37308 }, { "epoch": 3.75, "learning_rate": 4.1321349999187354e-07, "loss": 0.0685, "step": 37309 }, { "epoch": 3.75, "learning_rate": 4.12817758998838e-07, "loss": 0.176, "step": 37310 }, { "epoch": 3.75, "learning_rate": 4.124222062873728e-07, "loss": 0.0517, "step": 37311 }, { "epoch": 3.75, "learning_rate": 4.1202684185999927e-07, "loss": 0.3642, "step": 37312 }, { "epoch": 3.75, "learning_rate": 4.1163166571922874e-07, "loss": 0.1799, "step": 37313 }, { "epoch": 3.75, "learning_rate": 4.1123667786757915e-07, "loss": 0.373, "step": 37314 }, { "epoch": 3.75, "learning_rate": 4.1084187830756516e-07, "loss": 0.0368, "step": 37315 }, { "epoch": 3.75, "learning_rate": 4.104472670416948e-07, "loss": 0.029, "step": 37316 }, { "epoch": 3.75, "learning_rate": 4.1005284407248266e-07, "loss": 0.3961, "step": 37317 }, { "epoch": 3.75, "learning_rate": 4.096586094024368e-07, "loss": 0.3386, "step": 37318 }, { "epoch": 3.75, "learning_rate": 4.0926456303406855e-07, "loss": 0.2807, "step": 37319 }, { "epoch": 3.75, "learning_rate": 4.0887070496987924e-07, "loss": 0.0825, "step": 37320 }, { "epoch": 3.75, "learning_rate": 4.0847703521238346e-07, "loss": 0.0526, "step": 37321 }, { "epoch": 3.75, "learning_rate": 4.080835537640826e-07, "loss": 0.0122, "step": 37322 }, { "epoch": 3.75, "learning_rate": 4.0769026062747794e-07, "loss": 0.4425, "step": 37323 }, { "epoch": 3.75, "learning_rate": 4.0729715580507756e-07, "loss": 0.2406, "step": 37324 }, { "epoch": 3.75, "learning_rate": 4.069042392993827e-07, "loss": 0.038, "step": 37325 }, { "epoch": 3.75, "learning_rate": 4.065115111128881e-07, "loss": 0.3195, "step": 37326 }, { "epoch": 3.75, "learning_rate": 4.0611897124809506e-07, "loss": 0.2944, "step": 37327 }, { "epoch": 3.75, "learning_rate": 4.057266197075049e-07, "loss": 0.2459, "step": 37328 }, { "epoch": 3.75, "learning_rate": 4.053344564936123e-07, "loss": 0.2828, "step": 37329 }, { "epoch": 3.75, "learning_rate": 4.0494248160891533e-07, "loss": 0.1827, "step": 37330 }, { "epoch": 3.75, "learning_rate": 4.0455069505590194e-07, "loss": 0.2018, "step": 37331 }, { "epoch": 3.75, "learning_rate": 4.041590968370701e-07, "loss": 0.883, "step": 37332 }, { "epoch": 3.75, "learning_rate": 4.0376768695491116e-07, "loss": 0.2349, "step": 37333 }, { "epoch": 3.75, "learning_rate": 4.033764654119165e-07, "loss": 0.0514, "step": 37334 }, { "epoch": 3.75, "learning_rate": 4.029854322105742e-07, "loss": 0.3029, "step": 37335 }, { "epoch": 3.75, "learning_rate": 4.025945873533754e-07, "loss": 0.1238, "step": 37336 }, { "epoch": 3.75, "learning_rate": 4.022039308428016e-07, "loss": 0.505, "step": 37337 }, { "epoch": 3.75, "learning_rate": 4.018134626813441e-07, "loss": 0.1512, "step": 37338 }, { "epoch": 3.75, "learning_rate": 4.014231828714876e-07, "loss": 0.23, "step": 37339 }, { "epoch": 3.75, "learning_rate": 4.010330914157101e-07, "loss": 0.1523, "step": 37340 }, { "epoch": 3.75, "learning_rate": 4.006431883164996e-07, "loss": 0.1157, "step": 37341 }, { "epoch": 3.75, "learning_rate": 4.0025347357633747e-07, "loss": 0.2537, "step": 37342 }, { "epoch": 3.75, "learning_rate": 3.998639471977017e-07, "loss": 0.1264, "step": 37343 }, { "epoch": 3.76, "learning_rate": 3.9947460918307035e-07, "loss": 0.4057, "step": 37344 }, { "epoch": 3.76, "learning_rate": 3.9908545953492137e-07, "loss": 0.1651, "step": 37345 }, { "epoch": 3.76, "learning_rate": 3.9869649825573284e-07, "loss": 0.2898, "step": 37346 }, { "epoch": 3.76, "learning_rate": 3.9830772534797943e-07, "loss": 0.1107, "step": 37347 }, { "epoch": 3.76, "learning_rate": 3.9791914081412917e-07, "loss": 0.1171, "step": 37348 }, { "epoch": 3.76, "learning_rate": 3.975307446566667e-07, "loss": 0.1166, "step": 37349 }, { "epoch": 3.76, "learning_rate": 3.971425368780568e-07, "loss": 0.1022, "step": 37350 }, { "epoch": 3.76, "learning_rate": 3.967545174807674e-07, "loss": 0.2927, "step": 37351 }, { "epoch": 3.76, "learning_rate": 3.963666864672699e-07, "loss": 0.157, "step": 37352 }, { "epoch": 3.76, "learning_rate": 3.9597904384003905e-07, "loss": 1.2718, "step": 37353 }, { "epoch": 3.76, "learning_rate": 3.955915896015294e-07, "loss": 0.0687, "step": 37354 }, { "epoch": 3.76, "learning_rate": 3.9520432375421577e-07, "loss": 0.185, "step": 37355 }, { "epoch": 3.76, "learning_rate": 3.948172463005595e-07, "loss": 0.3049, "step": 37356 }, { "epoch": 3.76, "learning_rate": 3.944303572430219e-07, "loss": 0.6362, "step": 37357 }, { "epoch": 3.76, "learning_rate": 3.9404365658406773e-07, "loss": 0.064, "step": 37358 }, { "epoch": 3.76, "learning_rate": 3.936571443261583e-07, "loss": 0.0723, "step": 37359 }, { "epoch": 3.76, "learning_rate": 3.9327082047175165e-07, "loss": 0.1062, "step": 37360 }, { "epoch": 3.76, "learning_rate": 3.9288468502330586e-07, "loss": 0.1067, "step": 37361 }, { "epoch": 3.76, "learning_rate": 3.9249873798327896e-07, "loss": 0.1914, "step": 37362 }, { "epoch": 3.76, "learning_rate": 3.9211297935412895e-07, "loss": 0.097, "step": 37363 }, { "epoch": 3.76, "learning_rate": 3.917274091383072e-07, "loss": 0.316, "step": 37364 }, { "epoch": 3.76, "learning_rate": 3.913420273382684e-07, "loss": 0.2978, "step": 37365 }, { "epoch": 3.76, "learning_rate": 3.9095683395646066e-07, "loss": 0.2539, "step": 37366 }, { "epoch": 3.76, "learning_rate": 3.905718289953453e-07, "loss": 0.1878, "step": 37367 }, { "epoch": 3.76, "learning_rate": 3.9018701245736697e-07, "loss": 0.2652, "step": 37368 }, { "epoch": 3.76, "learning_rate": 3.898023843449705e-07, "loss": 0.0942, "step": 37369 }, { "epoch": 3.76, "learning_rate": 3.8941794466061053e-07, "loss": 0.1319, "step": 37370 }, { "epoch": 3.76, "learning_rate": 3.8903369340673176e-07, "loss": 0.204, "step": 37371 }, { "epoch": 3.76, "learning_rate": 3.886496305857723e-07, "loss": 0.0946, "step": 37372 }, { "epoch": 3.76, "learning_rate": 3.882657562001868e-07, "loss": 0.1055, "step": 37373 }, { "epoch": 3.76, "learning_rate": 3.8788207025241327e-07, "loss": 0.2403, "step": 37374 }, { "epoch": 3.76, "learning_rate": 3.874985727448899e-07, "loss": 0.0926, "step": 37375 }, { "epoch": 3.76, "learning_rate": 3.871152636800612e-07, "loss": 0.242, "step": 37376 }, { "epoch": 3.76, "learning_rate": 3.8673214306036874e-07, "loss": 0.148, "step": 37377 }, { "epoch": 3.76, "learning_rate": 3.863492108882438e-07, "loss": 0.1743, "step": 37378 }, { "epoch": 3.76, "learning_rate": 3.8596646716612783e-07, "loss": 0.0967, "step": 37379 }, { "epoch": 3.76, "learning_rate": 3.8558391189645546e-07, "loss": 0.2447, "step": 37380 }, { "epoch": 3.76, "learning_rate": 3.8520154508166486e-07, "loss": 0.0454, "step": 37381 }, { "epoch": 3.76, "learning_rate": 3.848193667241806e-07, "loss": 0.1573, "step": 37382 }, { "epoch": 3.76, "learning_rate": 3.8443737682644087e-07, "loss": 0.1376, "step": 37383 }, { "epoch": 3.76, "learning_rate": 3.84055575390877e-07, "loss": 0.1217, "step": 37384 }, { "epoch": 3.76, "learning_rate": 3.83673962419917e-07, "loss": 0.1725, "step": 37385 }, { "epoch": 3.76, "learning_rate": 3.8329253791598575e-07, "loss": 0.1324, "step": 37386 }, { "epoch": 3.76, "learning_rate": 3.829113018815178e-07, "loss": 0.4367, "step": 37387 }, { "epoch": 3.76, "learning_rate": 3.8253025431893463e-07, "loss": 0.3136, "step": 37388 }, { "epoch": 3.76, "learning_rate": 3.8214939523066093e-07, "loss": 0.2071, "step": 37389 }, { "epoch": 3.76, "learning_rate": 3.817687246191215e-07, "loss": 0.4092, "step": 37390 }, { "epoch": 3.76, "learning_rate": 3.8138824248673767e-07, "loss": 0.122, "step": 37391 }, { "epoch": 3.76, "learning_rate": 3.8100794883593413e-07, "loss": 0.3088, "step": 37392 }, { "epoch": 3.76, "learning_rate": 3.8062784366912575e-07, "loss": 0.103, "step": 37393 }, { "epoch": 3.76, "learning_rate": 3.8024792698873714e-07, "loss": 0.1408, "step": 37394 }, { "epoch": 3.76, "learning_rate": 3.798681987971797e-07, "loss": 0.4851, "step": 37395 }, { "epoch": 3.76, "learning_rate": 3.794886590968749e-07, "loss": 0.105, "step": 37396 }, { "epoch": 3.76, "learning_rate": 3.791093078902341e-07, "loss": 0.2208, "step": 37397 }, { "epoch": 3.76, "learning_rate": 3.787301451796721e-07, "loss": 0.0226, "step": 37398 }, { "epoch": 3.76, "learning_rate": 3.7835117096761016e-07, "loss": 0.1089, "step": 37399 }, { "epoch": 3.76, "learning_rate": 3.779723852564465e-07, "loss": 0.3315, "step": 37400 }, { "epoch": 3.76, "learning_rate": 3.775937880485958e-07, "loss": 0.0462, "step": 37401 }, { "epoch": 3.76, "learning_rate": 3.7721537934647274e-07, "loss": 0.2984, "step": 37402 }, { "epoch": 3.76, "learning_rate": 3.768371591524822e-07, "loss": 0.0707, "step": 37403 }, { "epoch": 3.76, "learning_rate": 3.764591274690321e-07, "loss": 0.4708, "step": 37404 }, { "epoch": 3.76, "learning_rate": 3.7608128429852397e-07, "loss": 0.5034, "step": 37405 }, { "epoch": 3.76, "learning_rate": 3.7570362964336577e-07, "loss": 0.2268, "step": 37406 }, { "epoch": 3.76, "learning_rate": 3.753261635059624e-07, "loss": 0.0784, "step": 37407 }, { "epoch": 3.76, "learning_rate": 3.7494888588871177e-07, "loss": 0.0248, "step": 37408 }, { "epoch": 3.76, "learning_rate": 3.745717967940154e-07, "loss": 1.0947, "step": 37409 }, { "epoch": 3.76, "learning_rate": 3.741948962242747e-07, "loss": 0.0776, "step": 37410 }, { "epoch": 3.76, "learning_rate": 3.7381818418188775e-07, "loss": 0.1712, "step": 37411 }, { "epoch": 3.76, "learning_rate": 3.7344166066924923e-07, "loss": 0.1533, "step": 37412 }, { "epoch": 3.76, "learning_rate": 3.7306532568876063e-07, "loss": 0.0996, "step": 37413 }, { "epoch": 3.76, "learning_rate": 3.726891792428133e-07, "loss": 0.2982, "step": 37414 }, { "epoch": 3.76, "learning_rate": 3.7231322133379876e-07, "loss": 0.4663, "step": 37415 }, { "epoch": 3.76, "learning_rate": 3.71937451964115e-07, "loss": 0.1717, "step": 37416 }, { "epoch": 3.76, "learning_rate": 3.7156187113615015e-07, "loss": 0.2049, "step": 37417 }, { "epoch": 3.76, "learning_rate": 3.7118647885228896e-07, "loss": 0.3418, "step": 37418 }, { "epoch": 3.76, "learning_rate": 3.7081127511493283e-07, "loss": 0.1503, "step": 37419 }, { "epoch": 3.76, "learning_rate": 3.704362599264566e-07, "loss": 0.322, "step": 37420 }, { "epoch": 3.76, "learning_rate": 3.700614332892549e-07, "loss": 0.2057, "step": 37421 }, { "epoch": 3.76, "learning_rate": 3.696867952057126e-07, "loss": 0.1935, "step": 37422 }, { "epoch": 3.76, "learning_rate": 3.693123456782077e-07, "loss": 0.0164, "step": 37423 }, { "epoch": 3.76, "learning_rate": 3.6893808470912837e-07, "loss": 0.0441, "step": 37424 }, { "epoch": 3.76, "learning_rate": 3.68564012300856e-07, "loss": 0.3712, "step": 37425 }, { "epoch": 3.76, "learning_rate": 3.6819012845576874e-07, "loss": 0.0955, "step": 37426 }, { "epoch": 3.76, "learning_rate": 3.678164331762479e-07, "loss": 0.3493, "step": 37427 }, { "epoch": 3.76, "learning_rate": 3.674429264646684e-07, "loss": 0.0794, "step": 37428 }, { "epoch": 3.76, "learning_rate": 3.670696083234115e-07, "loss": 0.147, "step": 37429 }, { "epoch": 3.76, "learning_rate": 3.666964787548488e-07, "loss": 0.0487, "step": 37430 }, { "epoch": 3.76, "learning_rate": 3.663235377613583e-07, "loss": 0.2072, "step": 37431 }, { "epoch": 3.76, "learning_rate": 3.6595078534531146e-07, "loss": 0.2268, "step": 37432 }, { "epoch": 3.76, "learning_rate": 3.655782215090797e-07, "loss": 0.0393, "step": 37433 }, { "epoch": 3.76, "learning_rate": 3.652058462550345e-07, "loss": 0.0239, "step": 37434 }, { "epoch": 3.76, "learning_rate": 3.6483365958554726e-07, "loss": 0.6248, "step": 37435 }, { "epoch": 3.76, "learning_rate": 3.644616615029828e-07, "loss": 0.2502, "step": 37436 }, { "epoch": 3.76, "learning_rate": 3.6408985200971247e-07, "loss": 0.3894, "step": 37437 }, { "epoch": 3.76, "learning_rate": 3.637182311080978e-07, "loss": 0.108, "step": 37438 }, { "epoch": 3.76, "learning_rate": 3.633467988005068e-07, "loss": 0.1443, "step": 37439 }, { "epoch": 3.76, "learning_rate": 3.6297555508930433e-07, "loss": 0.1163, "step": 37440 }, { "epoch": 3.76, "learning_rate": 3.626044999768485e-07, "loss": 0.1822, "step": 37441 }, { "epoch": 3.77, "learning_rate": 3.622336334655041e-07, "loss": 0.2735, "step": 37442 }, { "epoch": 3.77, "learning_rate": 3.618629555576325e-07, "loss": 0.3112, "step": 37443 }, { "epoch": 3.77, "learning_rate": 3.614924662555852e-07, "loss": 0.3266, "step": 37444 }, { "epoch": 3.77, "learning_rate": 3.6112216556172696e-07, "loss": 0.3016, "step": 37445 }, { "epoch": 3.77, "learning_rate": 3.6075205347841255e-07, "loss": 0.094, "step": 37446 }, { "epoch": 3.77, "learning_rate": 3.603821300079968e-07, "loss": 0.3742, "step": 37447 }, { "epoch": 3.77, "learning_rate": 3.600123951528311e-07, "loss": 0.4126, "step": 37448 }, { "epoch": 3.77, "learning_rate": 3.59642848915277e-07, "loss": 0.2137, "step": 37449 }, { "epoch": 3.77, "learning_rate": 3.592734912976725e-07, "loss": 0.3686, "step": 37450 }, { "epoch": 3.77, "learning_rate": 3.5890432230237914e-07, "loss": 0.102, "step": 37451 }, { "epoch": 3.77, "learning_rate": 3.58535341931745e-07, "loss": 0.3026, "step": 37452 }, { "epoch": 3.77, "learning_rate": 3.5816655018811153e-07, "loss": 0.3719, "step": 37453 }, { "epoch": 3.77, "learning_rate": 3.577979470738335e-07, "loss": 0.2797, "step": 37454 }, { "epoch": 3.77, "learning_rate": 3.5742953259125244e-07, "loss": 0.1045, "step": 37455 }, { "epoch": 3.77, "learning_rate": 3.5706130674270974e-07, "loss": 0.3301, "step": 37456 }, { "epoch": 3.77, "learning_rate": 3.5669326953055693e-07, "loss": 0.1552, "step": 37457 }, { "epoch": 3.77, "learning_rate": 3.563254209571254e-07, "loss": 0.0054, "step": 37458 }, { "epoch": 3.77, "learning_rate": 3.559577610247666e-07, "loss": 0.3187, "step": 37459 }, { "epoch": 3.77, "learning_rate": 3.555902897358154e-07, "loss": 0.1067, "step": 37460 }, { "epoch": 3.77, "learning_rate": 3.5522300709260656e-07, "loss": 0.1725, "step": 37461 }, { "epoch": 3.77, "learning_rate": 3.5485591309748157e-07, "loss": 0.0896, "step": 37462 }, { "epoch": 3.77, "learning_rate": 3.5448900775278184e-07, "loss": 0.3379, "step": 37463 }, { "epoch": 3.77, "learning_rate": 3.541222910608288e-07, "loss": 0.0632, "step": 37464 }, { "epoch": 3.77, "learning_rate": 3.537557630239674e-07, "loss": 0.2395, "step": 37465 }, { "epoch": 3.77, "learning_rate": 3.533894236445256e-07, "loss": 0.0383, "step": 37466 }, { "epoch": 3.77, "learning_rate": 3.5302327292483835e-07, "loss": 0.2675, "step": 37467 }, { "epoch": 3.77, "learning_rate": 3.526573108672304e-07, "loss": 0.2262, "step": 37468 }, { "epoch": 3.77, "learning_rate": 3.5229153747402987e-07, "loss": 0.6458, "step": 37469 }, { "epoch": 3.77, "learning_rate": 3.5192595274757157e-07, "loss": 0.1246, "step": 37470 }, { "epoch": 3.77, "learning_rate": 3.5156055669017696e-07, "loss": 0.1731, "step": 37471 }, { "epoch": 3.77, "learning_rate": 3.5119534930417084e-07, "loss": 0.0969, "step": 37472 }, { "epoch": 3.77, "learning_rate": 3.5083033059188143e-07, "loss": 0.779, "step": 37473 }, { "epoch": 3.77, "learning_rate": 3.5046550055562674e-07, "loss": 0.1617, "step": 37474 }, { "epoch": 3.77, "learning_rate": 3.501008591977284e-07, "loss": 0.2115, "step": 37475 }, { "epoch": 3.77, "learning_rate": 3.497364065205078e-07, "loss": 0.5287, "step": 37476 }, { "epoch": 3.77, "learning_rate": 3.493721425262897e-07, "loss": 0.0925, "step": 37477 }, { "epoch": 3.77, "learning_rate": 3.4900806721738566e-07, "loss": 0.2195, "step": 37478 }, { "epoch": 3.77, "learning_rate": 3.486441805961138e-07, "loss": 0.2076, "step": 37479 }, { "epoch": 3.77, "learning_rate": 3.4828048266479227e-07, "loss": 0.0016, "step": 37480 }, { "epoch": 3.77, "learning_rate": 3.4791697342573256e-07, "loss": 0.1593, "step": 37481 }, { "epoch": 3.77, "learning_rate": 3.475536528812495e-07, "loss": 0.3179, "step": 37482 }, { "epoch": 3.77, "learning_rate": 3.4719052103365454e-07, "loss": 0.3072, "step": 37483 }, { "epoch": 3.77, "learning_rate": 3.468275778852559e-07, "loss": 0.1383, "step": 37484 }, { "epoch": 3.77, "learning_rate": 3.4646482343836826e-07, "loss": 0.1266, "step": 37485 }, { "epoch": 3.77, "learning_rate": 3.461022576952999e-07, "loss": 0.0458, "step": 37486 }, { "epoch": 3.77, "learning_rate": 3.457398806583523e-07, "loss": 0.3166, "step": 37487 }, { "epoch": 3.77, "learning_rate": 3.4537769232983683e-07, "loss": 0.2265, "step": 37488 }, { "epoch": 3.77, "learning_rate": 3.450156927120551e-07, "loss": 0.0514, "step": 37489 }, { "epoch": 3.77, "learning_rate": 3.446538818073119e-07, "loss": 0.6745, "step": 37490 }, { "epoch": 3.77, "learning_rate": 3.442922596179154e-07, "loss": 0.0468, "step": 37491 }, { "epoch": 3.77, "learning_rate": 3.4393082614615714e-07, "loss": 0.054, "step": 37492 }, { "epoch": 3.77, "learning_rate": 3.435695813943385e-07, "loss": 0.3277, "step": 37493 }, { "epoch": 3.77, "learning_rate": 3.432085253647643e-07, "loss": 0.1124, "step": 37494 }, { "epoch": 3.77, "learning_rate": 3.428476580597295e-07, "loss": 0.3101, "step": 37495 }, { "epoch": 3.77, "learning_rate": 3.4248697948152885e-07, "loss": 0.2849, "step": 37496 }, { "epoch": 3.77, "learning_rate": 3.421264896324572e-07, "loss": 0.1384, "step": 37497 }, { "epoch": 3.77, "learning_rate": 3.4176618851481266e-07, "loss": 0.2068, "step": 37498 }, { "epoch": 3.77, "learning_rate": 3.414060761308868e-07, "loss": 0.1168, "step": 37499 }, { "epoch": 3.77, "learning_rate": 3.410461524829678e-07, "loss": 0.1029, "step": 37500 }, { "epoch": 3.77, "learning_rate": 3.4068641757334706e-07, "loss": 0.0953, "step": 37501 }, { "epoch": 3.77, "learning_rate": 3.403268714043195e-07, "loss": 0.0549, "step": 37502 }, { "epoch": 3.77, "learning_rate": 3.399675139781633e-07, "loss": 0.4122, "step": 37503 }, { "epoch": 3.77, "learning_rate": 3.3960834529717324e-07, "loss": 0.1653, "step": 37504 }, { "epoch": 3.77, "learning_rate": 3.3924936536363415e-07, "loss": 0.2806, "step": 37505 }, { "epoch": 3.77, "learning_rate": 3.3889057417982425e-07, "loss": 0.0249, "step": 37506 }, { "epoch": 3.77, "learning_rate": 3.385319717480351e-07, "loss": 0.1271, "step": 37507 }, { "epoch": 3.77, "learning_rate": 3.381735580705414e-07, "loss": 0.1893, "step": 37508 }, { "epoch": 3.77, "learning_rate": 3.378153331496281e-07, "loss": 0.3711, "step": 37509 }, { "epoch": 3.77, "learning_rate": 3.3745729698757333e-07, "loss": 0.2523, "step": 37510 }, { "epoch": 3.77, "learning_rate": 3.370994495866586e-07, "loss": 0.0306, "step": 37511 }, { "epoch": 3.77, "learning_rate": 3.367417909491555e-07, "loss": 0.4769, "step": 37512 }, { "epoch": 3.77, "learning_rate": 3.363843210773454e-07, "loss": 0.042, "step": 37513 }, { "epoch": 3.77, "learning_rate": 3.360270399734999e-07, "loss": 0.096, "step": 37514 }, { "epoch": 3.77, "learning_rate": 3.356699476398939e-07, "loss": 0.2261, "step": 37515 }, { "epoch": 3.77, "learning_rate": 3.3531304407880216e-07, "loss": 0.4331, "step": 37516 }, { "epoch": 3.77, "learning_rate": 3.3495632929248953e-07, "loss": 0.1742, "step": 37517 }, { "epoch": 3.77, "learning_rate": 3.345998032832309e-07, "loss": 0.1854, "step": 37518 }, { "epoch": 3.77, "learning_rate": 3.342434660532945e-07, "loss": 0.1702, "step": 37519 }, { "epoch": 3.77, "learning_rate": 3.3388731760494504e-07, "loss": 0.3019, "step": 37520 }, { "epoch": 3.77, "learning_rate": 3.335313579404542e-07, "loss": 0.0758, "step": 37521 }, { "epoch": 3.77, "learning_rate": 3.3317558706208005e-07, "loss": 0.3852, "step": 37522 }, { "epoch": 3.77, "learning_rate": 3.328200049720942e-07, "loss": 0.0706, "step": 37523 }, { "epoch": 3.77, "learning_rate": 3.3246461167275475e-07, "loss": 0.1224, "step": 37524 }, { "epoch": 3.77, "learning_rate": 3.3210940716632333e-07, "loss": 0.0278, "step": 37525 }, { "epoch": 3.77, "learning_rate": 3.317543914550614e-07, "loss": 0.0792, "step": 37526 }, { "epoch": 3.77, "learning_rate": 3.3139956454123375e-07, "loss": 0.4404, "step": 37527 }, { "epoch": 3.77, "learning_rate": 3.310449264270854e-07, "loss": 0.0929, "step": 37528 }, { "epoch": 3.77, "learning_rate": 3.3069047711488444e-07, "loss": 0.1856, "step": 37529 }, { "epoch": 3.77, "learning_rate": 3.3033621660687906e-07, "loss": 0.1483, "step": 37530 }, { "epoch": 3.77, "learning_rate": 3.299821449053308e-07, "loss": 0.0887, "step": 37531 }, { "epoch": 3.77, "learning_rate": 3.29628262012488e-07, "loss": 0.3252, "step": 37532 }, { "epoch": 3.77, "learning_rate": 3.2927456793060194e-07, "loss": 0.1272, "step": 37533 }, { "epoch": 3.77, "learning_rate": 3.2892106266192766e-07, "loss": 0.1096, "step": 37534 }, { "epoch": 3.77, "learning_rate": 3.2856774620871e-07, "loss": 0.3707, "step": 37535 }, { "epoch": 3.77, "learning_rate": 3.282146185732038e-07, "loss": 0.1549, "step": 37536 }, { "epoch": 3.77, "learning_rate": 3.2786167975764723e-07, "loss": 0.2336, "step": 37537 }, { "epoch": 3.77, "learning_rate": 3.275089297642953e-07, "loss": 0.2126, "step": 37538 }, { "epoch": 3.77, "learning_rate": 3.2715636859538267e-07, "loss": 0.2945, "step": 37539 }, { "epoch": 3.77, "learning_rate": 3.2680399625316103e-07, "loss": 0.1105, "step": 37540 }, { "epoch": 3.78, "learning_rate": 3.2645181273987526e-07, "loss": 0.1149, "step": 37541 }, { "epoch": 3.78, "learning_rate": 3.260998180577568e-07, "loss": 0.1447, "step": 37542 }, { "epoch": 3.78, "learning_rate": 3.257480122090506e-07, "loss": 0.1048, "step": 37543 }, { "epoch": 3.78, "learning_rate": 3.2539639519599817e-07, "loss": 0.4815, "step": 37544 }, { "epoch": 3.78, "learning_rate": 3.2504496702083444e-07, "loss": 0.0362, "step": 37545 }, { "epoch": 3.78, "learning_rate": 3.246937276857942e-07, "loss": 0.0808, "step": 37546 }, { "epoch": 3.78, "learning_rate": 3.2434267719311237e-07, "loss": 0.2146, "step": 37547 }, { "epoch": 3.78, "learning_rate": 3.239918155450272e-07, "loss": 0.3653, "step": 37548 }, { "epoch": 3.78, "learning_rate": 3.236411427437669e-07, "loss": 0.369, "step": 37549 }, { "epoch": 3.78, "learning_rate": 3.2329065879156625e-07, "loss": 0.2181, "step": 37550 }, { "epoch": 3.78, "learning_rate": 3.229403636906536e-07, "loss": 0.1243, "step": 37551 }, { "epoch": 3.78, "learning_rate": 3.2259025744326045e-07, "loss": 0.1506, "step": 37552 }, { "epoch": 3.78, "learning_rate": 3.2224034005161164e-07, "loss": 0.1753, "step": 37553 }, { "epoch": 3.78, "learning_rate": 3.218906115179321e-07, "loss": 0.2097, "step": 37554 }, { "epoch": 3.78, "learning_rate": 3.2154107184445667e-07, "loss": 0.286, "step": 37555 }, { "epoch": 3.78, "learning_rate": 3.2119172103340034e-07, "loss": 1.0154, "step": 37556 }, { "epoch": 3.78, "learning_rate": 3.2084255908698787e-07, "loss": 0.1384, "step": 37557 }, { "epoch": 3.78, "learning_rate": 3.2049358600744427e-07, "loss": 0.1601, "step": 37558 }, { "epoch": 3.78, "learning_rate": 3.2014480179699104e-07, "loss": 0.0865, "step": 37559 }, { "epoch": 3.78, "learning_rate": 3.19796206457843e-07, "loss": 0.074, "step": 37560 }, { "epoch": 3.78, "learning_rate": 3.1944779999222183e-07, "loss": 0.1225, "step": 37561 }, { "epoch": 3.78, "learning_rate": 3.190995824023424e-07, "loss": 0.13, "step": 37562 }, { "epoch": 3.78, "learning_rate": 3.1875155369042284e-07, "loss": 0.1988, "step": 37563 }, { "epoch": 3.78, "learning_rate": 3.184037138586782e-07, "loss": 0.1248, "step": 37564 }, { "epoch": 3.78, "learning_rate": 3.1805606290931987e-07, "loss": 0.3053, "step": 37565 }, { "epoch": 3.78, "learning_rate": 3.177086008445629e-07, "loss": 0.2116, "step": 37566 }, { "epoch": 3.78, "learning_rate": 3.1736132766661543e-07, "loss": 0.1338, "step": 37567 }, { "epoch": 3.78, "learning_rate": 3.170142433776857e-07, "loss": 0.1396, "step": 37568 }, { "epoch": 3.78, "learning_rate": 3.1666734797998864e-07, "loss": 0.2131, "step": 37569 }, { "epoch": 3.78, "learning_rate": 3.1632064147572913e-07, "loss": 0.0994, "step": 37570 }, { "epoch": 3.78, "learning_rate": 3.1597412386710876e-07, "loss": 0.1862, "step": 37571 }, { "epoch": 3.78, "learning_rate": 3.156277951563391e-07, "loss": 0.1986, "step": 37572 }, { "epoch": 3.78, "learning_rate": 3.1528165534562504e-07, "loss": 0.1791, "step": 37573 }, { "epoch": 3.78, "learning_rate": 3.1493570443716145e-07, "loss": 0.1804, "step": 37574 }, { "epoch": 3.78, "learning_rate": 3.145899424331566e-07, "loss": 0.1522, "step": 37575 }, { "epoch": 3.78, "learning_rate": 3.142443693358055e-07, "loss": 0.202, "step": 37576 }, { "epoch": 3.78, "learning_rate": 3.1389898514731284e-07, "loss": 0.1544, "step": 37577 }, { "epoch": 3.78, "learning_rate": 3.135537898698737e-07, "loss": 0.0206, "step": 37578 }, { "epoch": 3.78, "learning_rate": 3.1320878350568626e-07, "loss": 0.6175, "step": 37579 }, { "epoch": 3.78, "learning_rate": 3.128639660569421e-07, "loss": 0.1038, "step": 37580 }, { "epoch": 3.78, "learning_rate": 3.125193375258395e-07, "loss": 0.136, "step": 37581 }, { "epoch": 3.78, "learning_rate": 3.1217489791457e-07, "loss": 0.2274, "step": 37582 }, { "epoch": 3.78, "learning_rate": 3.118306472253252e-07, "loss": 0.069, "step": 37583 }, { "epoch": 3.78, "learning_rate": 3.1148658546029665e-07, "loss": 0.2475, "step": 37584 }, { "epoch": 3.78, "learning_rate": 3.111427126216726e-07, "loss": 0.1246, "step": 37585 }, { "epoch": 3.78, "learning_rate": 3.1079902871163804e-07, "loss": 1.0064, "step": 37586 }, { "epoch": 3.78, "learning_rate": 3.104555337323911e-07, "loss": 0.273, "step": 37587 }, { "epoch": 3.78, "learning_rate": 3.1011222768611013e-07, "loss": 0.0768, "step": 37588 }, { "epoch": 3.78, "learning_rate": 3.097691105749767e-07, "loss": 0.4331, "step": 37589 }, { "epoch": 3.78, "learning_rate": 3.0942618240117906e-07, "loss": 0.2993, "step": 37590 }, { "epoch": 3.78, "learning_rate": 3.0908344316689874e-07, "loss": 0.1443, "step": 37591 }, { "epoch": 3.78, "learning_rate": 3.087408928743174e-07, "loss": 0.1106, "step": 37592 }, { "epoch": 3.78, "learning_rate": 3.0839853152560994e-07, "loss": 0.0969, "step": 37593 }, { "epoch": 3.78, "learning_rate": 3.080563591229646e-07, "loss": 0.0759, "step": 37594 }, { "epoch": 3.78, "learning_rate": 3.0771437566854964e-07, "loss": 0.318, "step": 37595 }, { "epoch": 3.78, "learning_rate": 3.073725811645467e-07, "loss": 0.2825, "step": 37596 }, { "epoch": 3.78, "learning_rate": 3.070309756131273e-07, "loss": 0.1515, "step": 37597 }, { "epoch": 3.78, "learning_rate": 3.0668955901646647e-07, "loss": 0.1319, "step": 37598 }, { "epoch": 3.78, "learning_rate": 3.06348331376739e-07, "loss": 0.0508, "step": 37599 }, { "epoch": 3.78, "learning_rate": 3.0600729269611327e-07, "loss": 0.6035, "step": 37600 }, { "epoch": 3.78, "learning_rate": 3.0566644297676416e-07, "loss": 0.2711, "step": 37601 }, { "epoch": 3.78, "learning_rate": 3.053257822208566e-07, "loss": 0.2585, "step": 37602 }, { "epoch": 3.78, "learning_rate": 3.0498531043055554e-07, "loss": 0.074, "step": 37603 }, { "epoch": 3.78, "learning_rate": 3.046450276080326e-07, "loss": 0.1397, "step": 37604 }, { "epoch": 3.78, "learning_rate": 3.0430493375545597e-07, "loss": 0.2683, "step": 37605 }, { "epoch": 3.78, "learning_rate": 3.039650288749807e-07, "loss": 0.1005, "step": 37606 }, { "epoch": 3.78, "learning_rate": 3.036253129687783e-07, "loss": 0.1505, "step": 37607 }, { "epoch": 3.78, "learning_rate": 3.03285786039007e-07, "loss": 0.4035, "step": 37608 }, { "epoch": 3.78, "learning_rate": 3.0294644808782856e-07, "loss": 0.1429, "step": 37609 }, { "epoch": 3.78, "learning_rate": 3.0260729911739784e-07, "loss": 0.3436, "step": 37610 }, { "epoch": 3.78, "learning_rate": 3.022683391298797e-07, "loss": 0.2691, "step": 37611 }, { "epoch": 3.78, "learning_rate": 3.019295681274292e-07, "loss": 0.1346, "step": 37612 }, { "epoch": 3.78, "learning_rate": 3.015909861122013e-07, "loss": 0.0843, "step": 37613 }, { "epoch": 3.78, "learning_rate": 3.0125259308634745e-07, "loss": 0.1037, "step": 37614 }, { "epoch": 3.78, "learning_rate": 3.009143890520261e-07, "loss": 0.0431, "step": 37615 }, { "epoch": 3.78, "learning_rate": 3.005763740113887e-07, "loss": 0.3516, "step": 37616 }, { "epoch": 3.78, "learning_rate": 3.0023854796658367e-07, "loss": 0.2825, "step": 37617 }, { "epoch": 3.78, "learning_rate": 2.9990091091975925e-07, "loss": 0.3557, "step": 37618 }, { "epoch": 3.78, "learning_rate": 2.9956346287307035e-07, "loss": 0.0847, "step": 37619 }, { "epoch": 3.78, "learning_rate": 2.9922620382865864e-07, "loss": 0.2868, "step": 37620 }, { "epoch": 3.78, "learning_rate": 2.9888913378867564e-07, "loss": 0.0924, "step": 37621 }, { "epoch": 3.78, "learning_rate": 2.985522527552631e-07, "loss": 0.0837, "step": 37622 }, { "epoch": 3.78, "learning_rate": 2.9821556073056255e-07, "loss": 0.0821, "step": 37623 }, { "epoch": 3.78, "learning_rate": 2.978790577167223e-07, "loss": 0.2019, "step": 37624 }, { "epoch": 3.78, "learning_rate": 2.975427437158773e-07, "loss": 0.0727, "step": 37625 }, { "epoch": 3.78, "learning_rate": 2.9720661873017254e-07, "loss": 0.1315, "step": 37626 }, { "epoch": 3.78, "learning_rate": 2.968706827617429e-07, "loss": 0.2677, "step": 37627 }, { "epoch": 3.78, "learning_rate": 2.965349358127334e-07, "loss": 0.0872, "step": 37628 }, { "epoch": 3.78, "learning_rate": 2.961993778852723e-07, "loss": 0.0977, "step": 37629 }, { "epoch": 3.78, "learning_rate": 2.958640089814979e-07, "loss": 0.1601, "step": 37630 }, { "epoch": 3.78, "learning_rate": 2.955288291035485e-07, "loss": 0.056, "step": 37631 }, { "epoch": 3.78, "learning_rate": 2.951938382535491e-07, "loss": 0.351, "step": 37632 }, { "epoch": 3.78, "learning_rate": 2.9485903643363786e-07, "loss": 0.1272, "step": 37633 }, { "epoch": 3.78, "learning_rate": 2.945244236459432e-07, "loss": 0.2041, "step": 37634 }, { "epoch": 3.78, "learning_rate": 2.9418999989259343e-07, "loss": 0.5308, "step": 37635 }, { "epoch": 3.78, "learning_rate": 2.938557651757201e-07, "loss": 0.0738, "step": 37636 }, { "epoch": 3.78, "learning_rate": 2.935217194974482e-07, "loss": 0.11, "step": 37637 }, { "epoch": 3.78, "learning_rate": 2.931878628598994e-07, "loss": 0.2208, "step": 37638 }, { "epoch": 3.79, "learning_rate": 2.928541952652053e-07, "loss": 0.1021, "step": 37639 }, { "epoch": 3.79, "learning_rate": 2.925207167154842e-07, "loss": 0.066, "step": 37640 }, { "epoch": 3.79, "learning_rate": 2.921874272128611e-07, "loss": 0.0941, "step": 37641 }, { "epoch": 3.79, "learning_rate": 2.918543267594542e-07, "loss": 0.3646, "step": 37642 }, { "epoch": 3.79, "learning_rate": 2.915214153573886e-07, "loss": 0.009, "step": 37643 }, { "epoch": 3.79, "learning_rate": 2.911886930087759e-07, "loss": 0.1937, "step": 37644 }, { "epoch": 3.79, "learning_rate": 2.9085615971573775e-07, "loss": 0.1385, "step": 37645 }, { "epoch": 3.79, "learning_rate": 2.9052381548038574e-07, "loss": 0.053, "step": 37646 }, { "epoch": 3.79, "learning_rate": 2.9019166030484157e-07, "loss": 0.0313, "step": 37647 }, { "epoch": 3.79, "learning_rate": 2.898596941912168e-07, "loss": 0.1685, "step": 37648 }, { "epoch": 3.79, "learning_rate": 2.8952791714161987e-07, "loss": 0.1696, "step": 37649 }, { "epoch": 3.79, "learning_rate": 2.8919632915816227e-07, "loss": 0.0479, "step": 37650 }, { "epoch": 3.79, "learning_rate": 2.8886493024296245e-07, "loss": 0.251, "step": 37651 }, { "epoch": 3.79, "learning_rate": 2.8853372039811867e-07, "loss": 0.1279, "step": 37652 }, { "epoch": 3.79, "learning_rate": 2.8820269962574585e-07, "loss": 0.2441, "step": 37653 }, { "epoch": 3.79, "learning_rate": 2.8787186792794904e-07, "loss": 0.2619, "step": 37654 }, { "epoch": 3.79, "learning_rate": 2.8754122530683323e-07, "loss": 0.1365, "step": 37655 }, { "epoch": 3.79, "learning_rate": 2.8721077176450005e-07, "loss": 0.0978, "step": 37656 }, { "epoch": 3.79, "learning_rate": 2.868805073030545e-07, "loss": 0.4086, "step": 37657 }, { "epoch": 3.79, "learning_rate": 2.865504319245982e-07, "loss": 0.158, "step": 37658 }, { "epoch": 3.79, "learning_rate": 2.8622054563122944e-07, "loss": 0.2449, "step": 37659 }, { "epoch": 3.79, "learning_rate": 2.858908484250533e-07, "loss": 0.2421, "step": 37660 }, { "epoch": 3.79, "learning_rate": 2.855613403081614e-07, "loss": 0.212, "step": 37661 }, { "epoch": 3.79, "learning_rate": 2.8523202128265203e-07, "loss": 0.1517, "step": 37662 }, { "epoch": 3.79, "learning_rate": 2.8490289135062354e-07, "loss": 0.4146, "step": 37663 }, { "epoch": 3.79, "learning_rate": 2.845739505141676e-07, "loss": 0.4113, "step": 37664 }, { "epoch": 3.79, "learning_rate": 2.8424519877538245e-07, "loss": 0.0778, "step": 37665 }, { "epoch": 3.79, "learning_rate": 2.8391663613635324e-07, "loss": 0.1165, "step": 37666 }, { "epoch": 3.79, "learning_rate": 2.835882625991715e-07, "loss": 0.0434, "step": 37667 }, { "epoch": 3.79, "learning_rate": 2.8326007816593225e-07, "loss": 0.1712, "step": 37668 }, { "epoch": 3.79, "learning_rate": 2.8293208283872386e-07, "loss": 0.3014, "step": 37669 }, { "epoch": 3.79, "learning_rate": 2.8260427661962795e-07, "loss": 0.0679, "step": 37670 }, { "epoch": 3.79, "learning_rate": 2.822766595107329e-07, "loss": 0.201, "step": 37671 }, { "epoch": 3.79, "learning_rate": 2.8194923151412366e-07, "loss": 0.0777, "step": 37672 }, { "epoch": 3.79, "learning_rate": 2.816219926318853e-07, "loss": 0.4361, "step": 37673 }, { "epoch": 3.79, "learning_rate": 2.812949428660994e-07, "loss": 0.1447, "step": 37674 }, { "epoch": 3.79, "learning_rate": 2.809680822188476e-07, "loss": 0.149, "step": 37675 }, { "epoch": 3.79, "learning_rate": 2.8064141069220505e-07, "loss": 0.0844, "step": 37676 }, { "epoch": 3.79, "learning_rate": 2.8031492828826e-07, "loss": 0.3707, "step": 37677 }, { "epoch": 3.79, "learning_rate": 2.799886350090808e-07, "loss": 0.3385, "step": 37678 }, { "epoch": 3.79, "learning_rate": 2.796625308567491e-07, "loss": 0.1641, "step": 37679 }, { "epoch": 3.79, "learning_rate": 2.7933661583333993e-07, "loss": 0.2633, "step": 37680 }, { "epoch": 3.79, "learning_rate": 2.7901088994092497e-07, "loss": 0.2986, "step": 37681 }, { "epoch": 3.79, "learning_rate": 2.786853531815792e-07, "loss": 0.3151, "step": 37682 }, { "epoch": 3.79, "learning_rate": 2.783600055573743e-07, "loss": 0.0327, "step": 37683 }, { "epoch": 3.79, "learning_rate": 2.780348470703786e-07, "loss": 0.2188, "step": 37684 }, { "epoch": 3.79, "learning_rate": 2.7770987772266055e-07, "loss": 0.1074, "step": 37685 }, { "epoch": 3.79, "learning_rate": 2.773850975162917e-07, "loss": 0.1409, "step": 37686 }, { "epoch": 3.79, "learning_rate": 2.770605064533371e-07, "loss": 0.0843, "step": 37687 }, { "epoch": 3.79, "learning_rate": 2.767361045358618e-07, "loss": 0.0996, "step": 37688 }, { "epoch": 3.79, "learning_rate": 2.7641189176593087e-07, "loss": 0.0577, "step": 37689 }, { "epoch": 3.79, "learning_rate": 2.7608786814560584e-07, "loss": 0.5022, "step": 37690 }, { "epoch": 3.79, "learning_rate": 2.757640336769518e-07, "loss": 0.037, "step": 37691 }, { "epoch": 3.79, "learning_rate": 2.7544038836202714e-07, "loss": 0.0657, "step": 37692 }, { "epoch": 3.79, "learning_rate": 2.751169322028935e-07, "loss": 0.0511, "step": 37693 }, { "epoch": 3.79, "learning_rate": 2.7479366520160585e-07, "loss": 0.1328, "step": 37694 }, { "epoch": 3.79, "learning_rate": 2.744705873602227e-07, "loss": 0.0517, "step": 37695 }, { "epoch": 3.79, "learning_rate": 2.7414769868080227e-07, "loss": 0.2199, "step": 37696 }, { "epoch": 3.79, "learning_rate": 2.738249991653996e-07, "loss": 0.1184, "step": 37697 }, { "epoch": 3.79, "learning_rate": 2.735024888160631e-07, "loss": 0.0639, "step": 37698 }, { "epoch": 3.79, "learning_rate": 2.7318016763484446e-07, "loss": 0.0562, "step": 37699 }, { "epoch": 3.79, "learning_rate": 2.728580356238053e-07, "loss": 0.2221, "step": 37700 }, { "epoch": 3.79, "learning_rate": 2.7253609278498735e-07, "loss": 0.299, "step": 37701 }, { "epoch": 3.79, "learning_rate": 2.72214339120439e-07, "loss": 0.0924, "step": 37702 }, { "epoch": 3.79, "learning_rate": 2.7189277463220865e-07, "loss": 0.2941, "step": 37703 }, { "epoch": 3.79, "learning_rate": 2.7157139932234785e-07, "loss": 0.1631, "step": 37704 }, { "epoch": 3.79, "learning_rate": 2.712502131928951e-07, "loss": 0.2608, "step": 37705 }, { "epoch": 3.79, "learning_rate": 2.709292162458954e-07, "loss": 0.1293, "step": 37706 }, { "epoch": 3.79, "learning_rate": 2.706084084833937e-07, "loss": 0.5392, "step": 37707 }, { "epoch": 3.79, "learning_rate": 2.702877899074285e-07, "loss": 0.1662, "step": 37708 }, { "epoch": 3.79, "learning_rate": 2.699673605200448e-07, "loss": 0.1347, "step": 37709 }, { "epoch": 3.79, "learning_rate": 2.6964712032327755e-07, "loss": 0.2555, "step": 37710 }, { "epoch": 3.79, "learning_rate": 2.693270693191685e-07, "loss": 0.238, "step": 37711 }, { "epoch": 3.79, "learning_rate": 2.690072075097494e-07, "loss": 0.2636, "step": 37712 }, { "epoch": 3.79, "learning_rate": 2.6868753489705856e-07, "loss": 0.342, "step": 37713 }, { "epoch": 3.79, "learning_rate": 2.683680514831277e-07, "loss": 0.2343, "step": 37714 }, { "epoch": 3.79, "learning_rate": 2.6804875726999854e-07, "loss": 0.4347, "step": 37715 }, { "epoch": 3.79, "learning_rate": 2.677296522596895e-07, "loss": 0.3379, "step": 37716 }, { "epoch": 3.79, "learning_rate": 2.674107364542422e-07, "loss": 0.0703, "step": 37717 }, { "epoch": 3.79, "learning_rate": 2.670920098556784e-07, "loss": 0.0221, "step": 37718 }, { "epoch": 3.79, "learning_rate": 2.6677347246603313e-07, "loss": 0.0074, "step": 37719 }, { "epoch": 3.79, "learning_rate": 2.6645512428733143e-07, "loss": 0.5209, "step": 37720 }, { "epoch": 3.79, "learning_rate": 2.6613696532159504e-07, "loss": 0.1866, "step": 37721 }, { "epoch": 3.79, "learning_rate": 2.6581899557085234e-07, "loss": 0.1036, "step": 37722 }, { "epoch": 3.79, "learning_rate": 2.655012150371283e-07, "loss": 0.973, "step": 37723 }, { "epoch": 3.79, "learning_rate": 2.6518362372244145e-07, "loss": 0.3225, "step": 37724 }, { "epoch": 3.79, "learning_rate": 2.6486622162881335e-07, "loss": 0.2196, "step": 37725 }, { "epoch": 3.79, "learning_rate": 2.645490087582658e-07, "loss": 0.4678, "step": 37726 }, { "epoch": 3.79, "learning_rate": 2.642319851128172e-07, "loss": 0.0601, "step": 37727 }, { "epoch": 3.79, "learning_rate": 2.6391515069448256e-07, "loss": 0.083, "step": 37728 }, { "epoch": 3.79, "learning_rate": 2.635985055052836e-07, "loss": 0.1321, "step": 37729 }, { "epoch": 3.79, "learning_rate": 2.6328204954722546e-07, "loss": 0.326, "step": 37730 }, { "epoch": 3.79, "learning_rate": 2.6296578282232974e-07, "loss": 0.1225, "step": 37731 }, { "epoch": 3.79, "learning_rate": 2.626497053326082e-07, "loss": 0.0144, "step": 37732 }, { "epoch": 3.79, "learning_rate": 2.6233381708007264e-07, "loss": 0.0244, "step": 37733 }, { "epoch": 3.79, "learning_rate": 2.620181180667314e-07, "loss": 0.0273, "step": 37734 }, { "epoch": 3.79, "learning_rate": 2.617026082945928e-07, "loss": 0.2448, "step": 37735 }, { "epoch": 3.79, "learning_rate": 2.613872877656653e-07, "loss": 0.5432, "step": 37736 }, { "epoch": 3.79, "learning_rate": 2.610721564819574e-07, "loss": 0.1302, "step": 37737 }, { "epoch": 3.8, "learning_rate": 2.6075721444547397e-07, "loss": 0.0954, "step": 37738 }, { "epoch": 3.8, "learning_rate": 2.604424616582168e-07, "loss": 0.1881, "step": 37739 }, { "epoch": 3.8, "learning_rate": 2.6012789812219105e-07, "loss": 0.0518, "step": 37740 }, { "epoch": 3.8, "learning_rate": 2.5981352383939506e-07, "loss": 0.1345, "step": 37741 }, { "epoch": 3.8, "learning_rate": 2.5949933881183385e-07, "loss": 0.8338, "step": 37742 }, { "epoch": 3.8, "learning_rate": 2.5918534304150586e-07, "loss": 0.0761, "step": 37743 }, { "epoch": 3.8, "learning_rate": 2.588715365304095e-07, "loss": 0.4291, "step": 37744 }, { "epoch": 3.8, "learning_rate": 2.585579192805365e-07, "loss": 0.0897, "step": 37745 }, { "epoch": 3.8, "learning_rate": 2.582444912938886e-07, "loss": 0.2624, "step": 37746 }, { "epoch": 3.8, "learning_rate": 2.579312525724609e-07, "loss": 0.2036, "step": 37747 }, { "epoch": 3.8, "learning_rate": 2.576182031182384e-07, "loss": 0.3236, "step": 37748 }, { "epoch": 3.8, "learning_rate": 2.5730534293322285e-07, "loss": 0.1119, "step": 37749 }, { "epoch": 3.8, "learning_rate": 2.569926720194027e-07, "loss": 0.1125, "step": 37750 }, { "epoch": 3.8, "learning_rate": 2.56680190378763e-07, "loss": 0.1215, "step": 37751 }, { "epoch": 3.8, "learning_rate": 2.5636789801329885e-07, "loss": 0.0773, "step": 37752 }, { "epoch": 3.8, "learning_rate": 2.56055794924992e-07, "loss": 0.1192, "step": 37753 }, { "epoch": 3.8, "learning_rate": 2.557438811158341e-07, "loss": 0.2309, "step": 37754 }, { "epoch": 3.8, "learning_rate": 2.5543215658780375e-07, "loss": 0.0854, "step": 37755 }, { "epoch": 3.8, "learning_rate": 2.551206213428858e-07, "loss": 0.56, "step": 37756 }, { "epoch": 3.8, "learning_rate": 2.548092753830722e-07, "loss": 0.1719, "step": 37757 }, { "epoch": 3.8, "learning_rate": 2.5449811871033126e-07, "loss": 0.1359, "step": 37758 }, { "epoch": 3.8, "learning_rate": 2.5418715132664806e-07, "loss": 0.3849, "step": 37759 }, { "epoch": 3.8, "learning_rate": 2.53876373234001e-07, "loss": 0.0566, "step": 37760 }, { "epoch": 3.8, "learning_rate": 2.535657844343753e-07, "loss": 0.5837, "step": 37761 }, { "epoch": 3.8, "learning_rate": 2.5325538492973255e-07, "loss": 0.0823, "step": 37762 }, { "epoch": 3.8, "learning_rate": 2.5294517472206125e-07, "loss": 0.0747, "step": 37763 }, { "epoch": 3.8, "learning_rate": 2.526351538133298e-07, "loss": 0.1712, "step": 37764 }, { "epoch": 3.8, "learning_rate": 2.5232532220550995e-07, "loss": 0.281, "step": 37765 }, { "epoch": 3.8, "learning_rate": 2.520156799005768e-07, "loss": 0.0912, "step": 37766 }, { "epoch": 3.8, "learning_rate": 2.517062269004988e-07, "loss": 0.8004, "step": 37767 }, { "epoch": 3.8, "learning_rate": 2.513969632072477e-07, "loss": 0.2664, "step": 37768 }, { "epoch": 3.8, "learning_rate": 2.510878888227852e-07, "loss": 0.037, "step": 37769 }, { "epoch": 3.8, "learning_rate": 2.5077900374908646e-07, "loss": 0.3861, "step": 37770 }, { "epoch": 3.8, "learning_rate": 2.504703079881099e-07, "loss": 0.1489, "step": 37771 }, { "epoch": 3.8, "learning_rate": 2.501618015418239e-07, "loss": 0.1705, "step": 37772 }, { "epoch": 3.8, "learning_rate": 2.4985348441219356e-07, "loss": 0.2159, "step": 37773 }, { "epoch": 3.8, "learning_rate": 2.495453566011707e-07, "loss": 0.1392, "step": 37774 }, { "epoch": 3.8, "learning_rate": 2.492374181107304e-07, "loss": 0.1632, "step": 37775 }, { "epoch": 3.8, "learning_rate": 2.4892966894282444e-07, "loss": 0.284, "step": 37776 }, { "epoch": 3.8, "learning_rate": 2.486221090994112e-07, "loss": 0.1057, "step": 37777 }, { "epoch": 3.8, "learning_rate": 2.483147385824458e-07, "loss": 0.145, "step": 37778 }, { "epoch": 3.8, "learning_rate": 2.4800755739389334e-07, "loss": 0.3547, "step": 37779 }, { "epoch": 3.8, "learning_rate": 2.4770056553569896e-07, "loss": 0.1651, "step": 37780 }, { "epoch": 3.8, "learning_rate": 2.4739376300981773e-07, "loss": 0.2219, "step": 37781 }, { "epoch": 3.8, "learning_rate": 2.470871498182081e-07, "loss": 0.2965, "step": 37782 }, { "epoch": 3.8, "learning_rate": 2.467807259628152e-07, "loss": 0.2252, "step": 37783 }, { "epoch": 3.8, "learning_rate": 2.4647449144559406e-07, "loss": 0.0564, "step": 37784 }, { "epoch": 3.8, "learning_rate": 2.4616844626848654e-07, "loss": 0.0276, "step": 37785 }, { "epoch": 3.8, "learning_rate": 2.4586259043344437e-07, "loss": 0.1154, "step": 37786 }, { "epoch": 3.8, "learning_rate": 2.4555692394241605e-07, "loss": 0.0913, "step": 37787 }, { "epoch": 3.8, "learning_rate": 2.4525144679734325e-07, "loss": 0.0975, "step": 37788 }, { "epoch": 3.8, "learning_rate": 2.449461590001711e-07, "loss": 0.2803, "step": 37789 }, { "epoch": 3.8, "learning_rate": 2.446410605528415e-07, "loss": 0.0199, "step": 37790 }, { "epoch": 3.8, "learning_rate": 2.4433615145729614e-07, "loss": 0.2168, "step": 37791 }, { "epoch": 3.8, "learning_rate": 2.440314317154768e-07, "loss": 0.2702, "step": 37792 }, { "epoch": 3.8, "learning_rate": 2.437269013293253e-07, "loss": 0.1245, "step": 37793 }, { "epoch": 3.8, "learning_rate": 2.4342256030077003e-07, "loss": 0.1759, "step": 37794 }, { "epoch": 3.8, "learning_rate": 2.4311840863175615e-07, "loss": 0.0345, "step": 37795 }, { "epoch": 3.8, "learning_rate": 2.4281444632421544e-07, "loss": 0.0986, "step": 37796 }, { "epoch": 3.8, "learning_rate": 2.4251067338008303e-07, "loss": 0.1019, "step": 37797 }, { "epoch": 3.8, "learning_rate": 2.4220708980129404e-07, "loss": 0.066, "step": 37798 }, { "epoch": 3.8, "learning_rate": 2.419036955897802e-07, "loss": 0.2495, "step": 37799 }, { "epoch": 3.8, "learning_rate": 2.416004907474667e-07, "loss": 0.3438, "step": 37800 }, { "epoch": 3.8, "learning_rate": 2.4129747527629195e-07, "loss": 0.0913, "step": 37801 }, { "epoch": 3.8, "learning_rate": 2.409946491781745e-07, "loss": 0.4813, "step": 37802 }, { "epoch": 3.8, "learning_rate": 2.4069201245504935e-07, "loss": 0.3849, "step": 37803 }, { "epoch": 3.8, "learning_rate": 2.4038956510883834e-07, "loss": 0.6555, "step": 37804 }, { "epoch": 3.8, "learning_rate": 2.400873071414666e-07, "loss": 0.2342, "step": 37805 }, { "epoch": 3.8, "learning_rate": 2.39785238554856e-07, "loss": 0.0261, "step": 37806 }, { "epoch": 3.8, "learning_rate": 2.394833593509349e-07, "loss": 0.1906, "step": 37807 }, { "epoch": 3.8, "learning_rate": 2.3918166953161516e-07, "loss": 0.0993, "step": 37808 }, { "epoch": 3.8, "learning_rate": 2.3888016909882514e-07, "loss": 0.0681, "step": 37809 }, { "epoch": 3.8, "learning_rate": 2.385788580544801e-07, "loss": 0.6116, "step": 37810 }, { "epoch": 3.8, "learning_rate": 2.3827773640049844e-07, "loss": 0.118, "step": 37811 }, { "epoch": 3.8, "learning_rate": 2.3797680413879198e-07, "loss": 0.3152, "step": 37812 }, { "epoch": 3.8, "learning_rate": 2.376760612712825e-07, "loss": 0.2201, "step": 37813 }, { "epoch": 3.8, "learning_rate": 2.373755077998785e-07, "loss": 0.708, "step": 37814 }, { "epoch": 3.8, "learning_rate": 2.370751437264951e-07, "loss": 0.4635, "step": 37815 }, { "epoch": 3.8, "learning_rate": 2.3677496905304075e-07, "loss": 0.4935, "step": 37816 }, { "epoch": 3.8, "learning_rate": 2.364749837814306e-07, "loss": 0.456, "step": 37817 }, { "epoch": 3.8, "learning_rate": 2.3617518791356985e-07, "loss": 0.1342, "step": 37818 }, { "epoch": 3.8, "learning_rate": 2.3587558145137023e-07, "loss": 0.032, "step": 37819 }, { "epoch": 3.8, "learning_rate": 2.3557616439673025e-07, "loss": 0.1954, "step": 37820 }, { "epoch": 3.8, "learning_rate": 2.3527693675156503e-07, "loss": 0.0861, "step": 37821 }, { "epoch": 3.8, "learning_rate": 2.3497789851777306e-07, "loss": 0.3094, "step": 37822 }, { "epoch": 3.8, "learning_rate": 2.3467904969725619e-07, "loss": 0.2292, "step": 37823 }, { "epoch": 3.8, "learning_rate": 2.343803902919228e-07, "loss": 0.4932, "step": 37824 }, { "epoch": 3.8, "learning_rate": 2.3408192030366816e-07, "loss": 0.0948, "step": 37825 }, { "epoch": 3.8, "learning_rate": 2.3378363973439065e-07, "loss": 0.0718, "step": 37826 }, { "epoch": 3.8, "learning_rate": 2.3348554858599214e-07, "loss": 0.3044, "step": 37827 }, { "epoch": 3.8, "learning_rate": 2.3318764686036442e-07, "loss": 0.0692, "step": 37828 }, { "epoch": 3.8, "learning_rate": 2.3288993455940932e-07, "loss": 0.3231, "step": 37829 }, { "epoch": 3.8, "learning_rate": 2.3259241168501866e-07, "loss": 0.0564, "step": 37830 }, { "epoch": 3.8, "learning_rate": 2.3229507823908756e-07, "loss": 0.013, "step": 37831 }, { "epoch": 3.8, "learning_rate": 2.3199793422350458e-07, "loss": 0.2819, "step": 37832 }, { "epoch": 3.8, "learning_rate": 2.3170097964016145e-07, "loss": 0.1188, "step": 37833 }, { "epoch": 3.8, "learning_rate": 2.3140421449095007e-07, "loss": 0.1678, "step": 37834 }, { "epoch": 3.8, "learning_rate": 2.3110763877775888e-07, "loss": 0.072, "step": 37835 }, { "epoch": 3.81, "learning_rate": 2.3081125250247302e-07, "loss": 0.5311, "step": 37836 }, { "epoch": 3.81, "learning_rate": 2.3051505566697774e-07, "loss": 0.2441, "step": 37837 }, { "epoch": 3.81, "learning_rate": 2.3021904827316142e-07, "loss": 0.4159, "step": 37838 }, { "epoch": 3.81, "learning_rate": 2.2992323032290595e-07, "loss": 0.4021, "step": 37839 }, { "epoch": 3.81, "learning_rate": 2.2962760181809317e-07, "loss": 0.0871, "step": 37840 }, { "epoch": 3.81, "learning_rate": 2.2933216276060487e-07, "loss": 0.1213, "step": 37841 }, { "epoch": 3.81, "learning_rate": 2.2903691315232288e-07, "loss": 0.3515, "step": 37842 }, { "epoch": 3.81, "learning_rate": 2.287418529951224e-07, "loss": 0.0629, "step": 37843 }, { "epoch": 3.81, "learning_rate": 2.2844698229088524e-07, "loss": 0.3512, "step": 37844 }, { "epoch": 3.81, "learning_rate": 2.2815230104148653e-07, "loss": 0.4121, "step": 37845 }, { "epoch": 3.81, "learning_rate": 2.2785780924879816e-07, "loss": 0.0573, "step": 37846 }, { "epoch": 3.81, "learning_rate": 2.2756350691469862e-07, "loss": 0.2766, "step": 37847 }, { "epoch": 3.81, "learning_rate": 2.2726939404105973e-07, "loss": 0.5807, "step": 37848 }, { "epoch": 3.81, "learning_rate": 2.2697547062974998e-07, "loss": 0.0416, "step": 37849 }, { "epoch": 3.81, "learning_rate": 2.266817366826446e-07, "loss": 0.1116, "step": 37850 }, { "epoch": 3.81, "learning_rate": 2.2638819220160867e-07, "loss": 0.0272, "step": 37851 }, { "epoch": 3.81, "learning_rate": 2.2609483718851076e-07, "loss": 0.1212, "step": 37852 }, { "epoch": 3.81, "learning_rate": 2.2580167164521938e-07, "loss": 0.1198, "step": 37853 }, { "epoch": 3.81, "learning_rate": 2.2550869557360297e-07, "loss": 0.4251, "step": 37854 }, { "epoch": 3.81, "learning_rate": 2.2521590897551681e-07, "loss": 0.0207, "step": 37855 }, { "epoch": 3.81, "learning_rate": 2.2492331185283266e-07, "loss": 0.0911, "step": 37856 }, { "epoch": 3.81, "learning_rate": 2.2463090420740906e-07, "loss": 0.0646, "step": 37857 }, { "epoch": 3.81, "learning_rate": 2.2433868604110787e-07, "loss": 0.1487, "step": 37858 }, { "epoch": 3.81, "learning_rate": 2.2404665735578755e-07, "loss": 0.1079, "step": 37859 }, { "epoch": 3.81, "learning_rate": 2.2375481815330667e-07, "loss": 0.1053, "step": 37860 }, { "epoch": 3.81, "learning_rate": 2.2346316843552372e-07, "loss": 0.0246, "step": 37861 }, { "epoch": 3.81, "learning_rate": 2.2317170820429056e-07, "loss": 0.3658, "step": 37862 }, { "epoch": 3.81, "learning_rate": 2.2288043746146569e-07, "loss": 0.1576, "step": 37863 }, { "epoch": 3.81, "learning_rate": 2.2258935620890763e-07, "loss": 0.1125, "step": 37864 }, { "epoch": 3.81, "learning_rate": 2.2229846444845824e-07, "loss": 0.177, "step": 37865 }, { "epoch": 3.81, "learning_rate": 2.2200776218197272e-07, "loss": 0.1635, "step": 37866 }, { "epoch": 3.81, "learning_rate": 2.217172494113029e-07, "loss": 0.2483, "step": 37867 }, { "epoch": 3.81, "learning_rate": 2.214269261382973e-07, "loss": 0.1636, "step": 37868 }, { "epoch": 3.81, "learning_rate": 2.2113679236480112e-07, "loss": 0.2201, "step": 37869 }, { "epoch": 3.81, "learning_rate": 2.2084684809266286e-07, "loss": 0.0803, "step": 37870 }, { "epoch": 3.81, "learning_rate": 2.2055709332372776e-07, "loss": 0.7391, "step": 37871 }, { "epoch": 3.81, "learning_rate": 2.202675280598343e-07, "loss": 0.3739, "step": 37872 }, { "epoch": 3.81, "learning_rate": 2.1997815230283102e-07, "loss": 0.0566, "step": 37873 }, { "epoch": 3.81, "learning_rate": 2.1968896605455979e-07, "loss": 0.1589, "step": 37874 }, { "epoch": 3.81, "learning_rate": 2.1939996931685913e-07, "loss": 0.3288, "step": 37875 }, { "epoch": 3.81, "learning_rate": 2.1911116209156756e-07, "loss": 0.2233, "step": 37876 }, { "epoch": 3.81, "learning_rate": 2.1882254438052364e-07, "loss": 0.2044, "step": 37877 }, { "epoch": 3.81, "learning_rate": 2.185341161855625e-07, "loss": 0.1709, "step": 37878 }, { "epoch": 3.81, "learning_rate": 2.1824587750851944e-07, "loss": 0.7224, "step": 37879 }, { "epoch": 3.81, "learning_rate": 2.179578283512329e-07, "loss": 0.041, "step": 37880 }, { "epoch": 3.81, "learning_rate": 2.1766996871553146e-07, "loss": 0.4568, "step": 37881 }, { "epoch": 3.81, "learning_rate": 2.1738229860325033e-07, "loss": 0.0758, "step": 37882 }, { "epoch": 3.81, "learning_rate": 2.1709481801621467e-07, "loss": 0.6376, "step": 37883 }, { "epoch": 3.81, "learning_rate": 2.1680752695625972e-07, "loss": 0.2634, "step": 37884 }, { "epoch": 3.81, "learning_rate": 2.16520425425214e-07, "loss": 0.1371, "step": 37885 }, { "epoch": 3.81, "learning_rate": 2.162335134248994e-07, "loss": 0.3403, "step": 37886 }, { "epoch": 3.81, "learning_rate": 2.1594679095714443e-07, "loss": 0.1942, "step": 37887 }, { "epoch": 3.81, "learning_rate": 2.156602580237743e-07, "loss": 0.031, "step": 37888 }, { "epoch": 3.81, "learning_rate": 2.1537391462661093e-07, "loss": 0.1196, "step": 37889 }, { "epoch": 3.81, "learning_rate": 2.1508776076747948e-07, "loss": 0.1263, "step": 37890 }, { "epoch": 3.81, "learning_rate": 2.148017964481952e-07, "loss": 0.1421, "step": 37891 }, { "epoch": 3.81, "learning_rate": 2.1451602167058326e-07, "loss": 0.1009, "step": 37892 }, { "epoch": 3.81, "learning_rate": 2.1423043643646223e-07, "loss": 0.1512, "step": 37893 }, { "epoch": 3.81, "learning_rate": 2.13945040747644e-07, "loss": 0.0364, "step": 37894 }, { "epoch": 3.81, "learning_rate": 2.136598346059504e-07, "loss": 0.3917, "step": 37895 }, { "epoch": 3.81, "learning_rate": 2.1337481801319669e-07, "loss": 0.1761, "step": 37896 }, { "epoch": 3.81, "learning_rate": 2.130899909711914e-07, "loss": 0.0614, "step": 37897 }, { "epoch": 3.81, "learning_rate": 2.128053534817498e-07, "loss": 0.4421, "step": 37898 }, { "epoch": 3.81, "learning_rate": 2.125209055466837e-07, "loss": 0.0662, "step": 37899 }, { "epoch": 3.81, "learning_rate": 2.1223664716780166e-07, "loss": 0.2378, "step": 37900 }, { "epoch": 3.81, "learning_rate": 2.1195257834691563e-07, "loss": 0.0495, "step": 37901 }, { "epoch": 3.81, "learning_rate": 2.1166869908583076e-07, "loss": 0.2894, "step": 37902 }, { "epoch": 3.81, "learning_rate": 2.1138500938635562e-07, "loss": 0.1122, "step": 37903 }, { "epoch": 3.81, "learning_rate": 2.1110150925029215e-07, "loss": 0.1798, "step": 37904 }, { "epoch": 3.81, "learning_rate": 2.1081819867944553e-07, "loss": 0.55, "step": 37905 }, { "epoch": 3.81, "learning_rate": 2.1053507767562097e-07, "loss": 0.3658, "step": 37906 }, { "epoch": 3.81, "learning_rate": 2.102521462406204e-07, "loss": 0.2229, "step": 37907 }, { "epoch": 3.81, "learning_rate": 2.0996940437623901e-07, "loss": 0.3124, "step": 37908 }, { "epoch": 3.81, "learning_rate": 2.0968685208428206e-07, "loss": 0.3178, "step": 37909 }, { "epoch": 3.81, "learning_rate": 2.0940448936654476e-07, "loss": 0.1014, "step": 37910 }, { "epoch": 3.81, "learning_rate": 2.091223162248257e-07, "loss": 0.0481, "step": 37911 }, { "epoch": 3.81, "learning_rate": 2.0884033266091672e-07, "loss": 0.1809, "step": 37912 }, { "epoch": 3.81, "learning_rate": 2.0855853867661312e-07, "loss": 0.3876, "step": 37913 }, { "epoch": 3.81, "learning_rate": 2.082769342737134e-07, "loss": 0.201, "step": 37914 }, { "epoch": 3.81, "learning_rate": 2.079955194540062e-07, "loss": 0.1162, "step": 37915 }, { "epoch": 3.81, "learning_rate": 2.0771429421928e-07, "loss": 0.4449, "step": 37916 }, { "epoch": 3.81, "learning_rate": 2.0743325857132678e-07, "loss": 0.2062, "step": 37917 }, { "epoch": 3.81, "learning_rate": 2.0715241251193505e-07, "loss": 0.0491, "step": 37918 }, { "epoch": 3.81, "learning_rate": 2.068717560428901e-07, "loss": 0.4533, "step": 37919 }, { "epoch": 3.81, "learning_rate": 2.0659128916598046e-07, "loss": 0.0686, "step": 37920 }, { "epoch": 3.81, "learning_rate": 2.063110118829914e-07, "loss": 0.1739, "step": 37921 }, { "epoch": 3.81, "learning_rate": 2.060309241957048e-07, "loss": 0.2325, "step": 37922 }, { "epoch": 3.81, "learning_rate": 2.0575102610590259e-07, "loss": 0.0767, "step": 37923 }, { "epoch": 3.81, "learning_rate": 2.0547131761536664e-07, "loss": 0.1217, "step": 37924 }, { "epoch": 3.81, "learning_rate": 2.0519179872587556e-07, "loss": 0.2133, "step": 37925 }, { "epoch": 3.81, "learning_rate": 2.0491246943920793e-07, "loss": 0.0978, "step": 37926 }, { "epoch": 3.81, "learning_rate": 2.0463332975714566e-07, "loss": 0.5242, "step": 37927 }, { "epoch": 3.81, "learning_rate": 2.0435437968146064e-07, "loss": 0.1199, "step": 37928 }, { "epoch": 3.81, "learning_rate": 2.0407561921393149e-07, "loss": 0.2558, "step": 37929 }, { "epoch": 3.81, "learning_rate": 2.0379704835632672e-07, "loss": 0.125, "step": 37930 }, { "epoch": 3.81, "learning_rate": 2.0351866711042831e-07, "loss": 0.0585, "step": 37931 }, { "epoch": 3.81, "learning_rate": 2.0324047547799818e-07, "loss": 0.0103, "step": 37932 }, { "epoch": 3.81, "learning_rate": 2.029624734608082e-07, "loss": 0.0912, "step": 37933 }, { "epoch": 3.81, "learning_rate": 2.0268466106063033e-07, "loss": 0.1444, "step": 37934 }, { "epoch": 3.82, "learning_rate": 2.0240703827923645e-07, "loss": 0.0838, "step": 37935 }, { "epoch": 3.82, "learning_rate": 2.0212960511838518e-07, "loss": 0.1971, "step": 37936 }, { "epoch": 3.82, "learning_rate": 2.018523615798451e-07, "loss": 0.1777, "step": 37937 }, { "epoch": 3.82, "learning_rate": 2.0157530766538147e-07, "loss": 0.1897, "step": 37938 }, { "epoch": 3.82, "learning_rate": 2.0129844337675618e-07, "loss": 0.5156, "step": 37939 }, { "epoch": 3.82, "learning_rate": 2.0102176871573118e-07, "loss": 0.212, "step": 37940 }, { "epoch": 3.82, "learning_rate": 2.0074528368407176e-07, "loss": 0.0362, "step": 37941 }, { "epoch": 3.82, "learning_rate": 2.0046898828352978e-07, "loss": 0.1205, "step": 37942 }, { "epoch": 3.82, "learning_rate": 2.0019288251586386e-07, "loss": 0.273, "step": 37943 }, { "epoch": 3.82, "learning_rate": 1.9991696638283595e-07, "loss": 0.1044, "step": 37944 }, { "epoch": 3.82, "learning_rate": 1.9964123988620131e-07, "loss": 0.0719, "step": 37945 }, { "epoch": 3.82, "learning_rate": 1.9936570302771518e-07, "loss": 0.1065, "step": 37946 }, { "epoch": 3.82, "learning_rate": 1.9909035580912617e-07, "loss": 0.2667, "step": 37947 }, { "epoch": 3.82, "learning_rate": 1.9881519823218952e-07, "loss": 0.0416, "step": 37948 }, { "epoch": 3.82, "learning_rate": 1.985402302986572e-07, "loss": 0.11, "step": 37949 }, { "epoch": 3.82, "learning_rate": 1.9826545201027446e-07, "loss": 0.2574, "step": 37950 }, { "epoch": 3.82, "learning_rate": 1.9799086336879658e-07, "loss": 0.2159, "step": 37951 }, { "epoch": 3.82, "learning_rate": 1.977164643759688e-07, "loss": 0.2642, "step": 37952 }, { "epoch": 3.82, "learning_rate": 1.9744225503353308e-07, "loss": 0.0168, "step": 37953 }, { "epoch": 3.82, "learning_rate": 1.9716823534323801e-07, "loss": 0.0496, "step": 37954 }, { "epoch": 3.82, "learning_rate": 1.9689440530682556e-07, "loss": 0.1783, "step": 37955 }, { "epoch": 3.82, "learning_rate": 1.9662076492604097e-07, "loss": 0.3281, "step": 37956 }, { "epoch": 3.82, "learning_rate": 1.963473142026262e-07, "loss": 0.4145, "step": 37957 }, { "epoch": 3.82, "learning_rate": 1.960740531383165e-07, "loss": 0.1452, "step": 37958 }, { "epoch": 3.82, "learning_rate": 1.958009817348505e-07, "loss": 0.2861, "step": 37959 }, { "epoch": 3.82, "learning_rate": 1.955280999939768e-07, "loss": 0.1164, "step": 37960 }, { "epoch": 3.82, "learning_rate": 1.9525540791741737e-07, "loss": 0.1705, "step": 37961 }, { "epoch": 3.82, "learning_rate": 1.9498290550691745e-07, "loss": 0.0869, "step": 37962 }, { "epoch": 3.82, "learning_rate": 1.9471059276420568e-07, "loss": 0.0749, "step": 37963 }, { "epoch": 3.82, "learning_rate": 1.9443846969102064e-07, "loss": 0.1639, "step": 37964 }, { "epoch": 3.82, "learning_rate": 1.9416653628908764e-07, "loss": 0.2372, "step": 37965 }, { "epoch": 3.82, "learning_rate": 1.9389479256013864e-07, "loss": 0.0798, "step": 37966 }, { "epoch": 3.82, "learning_rate": 1.936232385059089e-07, "loss": 0.168, "step": 37967 }, { "epoch": 3.82, "learning_rate": 1.9335187412811706e-07, "loss": 0.4106, "step": 37968 }, { "epoch": 3.82, "learning_rate": 1.9308069942849838e-07, "loss": 0.1131, "step": 37969 }, { "epoch": 3.82, "learning_rate": 1.9280971440877483e-07, "loss": 0.4326, "step": 37970 }, { "epoch": 3.82, "learning_rate": 1.9253891907067166e-07, "loss": 0.0282, "step": 37971 }, { "epoch": 3.82, "learning_rate": 1.9226831341590756e-07, "loss": 0.0971, "step": 37972 }, { "epoch": 3.82, "learning_rate": 1.919978974462111e-07, "loss": 0.2143, "step": 37973 }, { "epoch": 3.82, "learning_rate": 1.9172767116330092e-07, "loss": 0.2766, "step": 37974 }, { "epoch": 3.82, "learning_rate": 1.9145763456889564e-07, "loss": 0.113, "step": 37975 }, { "epoch": 3.82, "learning_rate": 1.911877876647139e-07, "loss": 0.2179, "step": 37976 }, { "epoch": 3.82, "learning_rate": 1.909181304524743e-07, "loss": 0.2784, "step": 37977 }, { "epoch": 3.82, "learning_rate": 1.9064866293389217e-07, "loss": 0.0748, "step": 37978 }, { "epoch": 3.82, "learning_rate": 1.9037938511067942e-07, "loss": 0.4326, "step": 37979 }, { "epoch": 3.82, "learning_rate": 1.901102969845514e-07, "loss": 0.0725, "step": 37980 }, { "epoch": 3.82, "learning_rate": 1.898413985572267e-07, "loss": 0.0534, "step": 37981 }, { "epoch": 3.82, "learning_rate": 1.8957268983040398e-07, "loss": 0.2177, "step": 37982 }, { "epoch": 3.82, "learning_rate": 1.893041708058052e-07, "loss": 0.7163, "step": 37983 }, { "epoch": 3.82, "learning_rate": 1.890358414851323e-07, "loss": 0.1634, "step": 37984 }, { "epoch": 3.82, "learning_rate": 1.8876770187009396e-07, "loss": 0.4269, "step": 37985 }, { "epoch": 3.82, "learning_rate": 1.8849975196239878e-07, "loss": 0.0917, "step": 37986 }, { "epoch": 3.82, "learning_rate": 1.8823199176374873e-07, "loss": 0.1111, "step": 37987 }, { "epoch": 3.82, "learning_rate": 1.8796442127585245e-07, "loss": 0.4066, "step": 37988 }, { "epoch": 3.82, "learning_rate": 1.8769704050040526e-07, "loss": 0.0634, "step": 37989 }, { "epoch": 3.82, "learning_rate": 1.874298494391158e-07, "loss": 0.4146, "step": 37990 }, { "epoch": 3.82, "learning_rate": 1.8716284809367933e-07, "loss": 0.5264, "step": 37991 }, { "epoch": 3.82, "learning_rate": 1.868960364658012e-07, "loss": 0.0401, "step": 37992 }, { "epoch": 3.82, "learning_rate": 1.8662941455717342e-07, "loss": 0.0721, "step": 37993 }, { "epoch": 3.82, "learning_rate": 1.8636298236949123e-07, "loss": 0.3681, "step": 37994 }, { "epoch": 3.82, "learning_rate": 1.8609673990445663e-07, "loss": 0.2774, "step": 37995 }, { "epoch": 3.82, "learning_rate": 1.8583068716375828e-07, "loss": 0.1015, "step": 37996 }, { "epoch": 3.82, "learning_rate": 1.855648241490915e-07, "loss": 0.0565, "step": 37997 }, { "epoch": 3.82, "learning_rate": 1.8529915086215154e-07, "loss": 0.203, "step": 37998 }, { "epoch": 3.82, "learning_rate": 1.8503366730462378e-07, "loss": 0.342, "step": 37999 }, { "epoch": 3.82, "learning_rate": 1.8476837347820019e-07, "loss": 0.7659, "step": 38000 }, { "epoch": 3.82, "learning_rate": 1.8450326938456608e-07, "loss": 0.5026, "step": 38001 }, { "epoch": 3.82, "learning_rate": 1.8423835502541343e-07, "loss": 0.0681, "step": 38002 }, { "epoch": 3.82, "learning_rate": 1.839736304024242e-07, "loss": 0.0561, "step": 38003 }, { "epoch": 3.82, "learning_rate": 1.8370909551728376e-07, "loss": 0.6633, "step": 38004 }, { "epoch": 3.82, "learning_rate": 1.8344475037167407e-07, "loss": 0.1866, "step": 38005 }, { "epoch": 3.82, "learning_rate": 1.831805949672838e-07, "loss": 0.1461, "step": 38006 }, { "epoch": 3.82, "learning_rate": 1.8291662930578824e-07, "loss": 0.2118, "step": 38007 }, { "epoch": 3.82, "learning_rate": 1.8265285338886606e-07, "loss": 0.2696, "step": 38008 }, { "epoch": 3.82, "learning_rate": 1.8238926721819926e-07, "loss": 0.2167, "step": 38009 }, { "epoch": 3.82, "learning_rate": 1.8212587079546318e-07, "loss": 0.1615, "step": 38010 }, { "epoch": 3.82, "learning_rate": 1.8186266412233643e-07, "loss": 0.103, "step": 38011 }, { "epoch": 3.82, "learning_rate": 1.8159964720049106e-07, "loss": 0.3127, "step": 38012 }, { "epoch": 3.82, "learning_rate": 1.8133682003160234e-07, "loss": 0.2892, "step": 38013 }, { "epoch": 3.82, "learning_rate": 1.8107418261734233e-07, "loss": 0.0825, "step": 38014 }, { "epoch": 3.82, "learning_rate": 1.8081173495938298e-07, "loss": 0.2184, "step": 38015 }, { "epoch": 3.82, "learning_rate": 1.8054947705939297e-07, "loss": 0.1127, "step": 38016 }, { "epoch": 3.82, "learning_rate": 1.8028740891904094e-07, "loss": 0.0593, "step": 38017 }, { "epoch": 3.82, "learning_rate": 1.8002553053999893e-07, "loss": 0.3755, "step": 38018 }, { "epoch": 3.82, "learning_rate": 1.7976384192392893e-07, "loss": 0.1947, "step": 38019 }, { "epoch": 3.82, "learning_rate": 1.7950234307249624e-07, "loss": 0.444, "step": 38020 }, { "epoch": 3.82, "learning_rate": 1.7924103398736958e-07, "loss": 0.2295, "step": 38021 }, { "epoch": 3.82, "learning_rate": 1.7897991467020758e-07, "loss": 0.6647, "step": 38022 }, { "epoch": 3.82, "learning_rate": 1.7871898512266893e-07, "loss": 0.0549, "step": 38023 }, { "epoch": 3.82, "learning_rate": 1.7845824534642563e-07, "loss": 0.1709, "step": 38024 }, { "epoch": 3.82, "learning_rate": 1.78197695343123e-07, "loss": 0.3972, "step": 38025 }, { "epoch": 3.82, "learning_rate": 1.7793733511442646e-07, "loss": 0.1557, "step": 38026 }, { "epoch": 3.82, "learning_rate": 1.7767716466199457e-07, "loss": 0.1551, "step": 38027 }, { "epoch": 3.82, "learning_rate": 1.7741718398747607e-07, "loss": 0.4878, "step": 38028 }, { "epoch": 3.82, "learning_rate": 1.7715739309253298e-07, "loss": 0.142, "step": 38029 }, { "epoch": 3.82, "learning_rate": 1.7689779197881395e-07, "loss": 0.0802, "step": 38030 }, { "epoch": 3.82, "learning_rate": 1.76638380647971e-07, "loss": 0.1023, "step": 38031 }, { "epoch": 3.82, "learning_rate": 1.7637915910165948e-07, "loss": 0.0811, "step": 38032 }, { "epoch": 3.83, "learning_rate": 1.7612012734152138e-07, "loss": 0.0793, "step": 38033 }, { "epoch": 3.83, "learning_rate": 1.758612853692121e-07, "loss": 0.2475, "step": 38034 }, { "epoch": 3.83, "learning_rate": 1.756026331863736e-07, "loss": 0.2145, "step": 38035 }, { "epoch": 3.83, "learning_rate": 1.7534417079465793e-07, "loss": 0.1514, "step": 38036 }, { "epoch": 3.83, "learning_rate": 1.7508589819570044e-07, "loss": 0.3216, "step": 38037 }, { "epoch": 3.83, "learning_rate": 1.7482781539115645e-07, "loss": 0.2443, "step": 38038 }, { "epoch": 3.83, "learning_rate": 1.74569922382658e-07, "loss": 0.0292, "step": 38039 }, { "epoch": 3.83, "learning_rate": 1.7431221917185047e-07, "loss": 0.0746, "step": 38040 }, { "epoch": 3.83, "learning_rate": 1.7405470576037586e-07, "loss": 0.0897, "step": 38041 }, { "epoch": 3.83, "learning_rate": 1.7379738214986952e-07, "loss": 0.2876, "step": 38042 }, { "epoch": 3.83, "learning_rate": 1.7354024834197013e-07, "loss": 0.2314, "step": 38043 }, { "epoch": 3.83, "learning_rate": 1.7328330433831308e-07, "loss": 0.093, "step": 38044 }, { "epoch": 3.83, "learning_rate": 1.7302655014053702e-07, "loss": 0.582, "step": 38045 }, { "epoch": 3.83, "learning_rate": 1.7276998575027069e-07, "loss": 0.215, "step": 38046 }, { "epoch": 3.83, "learning_rate": 1.7251361116914943e-07, "loss": 0.0818, "step": 38047 }, { "epoch": 3.83, "learning_rate": 1.722574263988086e-07, "loss": 0.1626, "step": 38048 }, { "epoch": 3.83, "learning_rate": 1.7200143144087023e-07, "loss": 0.1329, "step": 38049 }, { "epoch": 3.83, "learning_rate": 1.7174562629696965e-07, "loss": 0.0564, "step": 38050 }, { "epoch": 3.83, "learning_rate": 1.714900109687323e-07, "loss": 0.2196, "step": 38051 }, { "epoch": 3.83, "learning_rate": 1.7123458545778347e-07, "loss": 0.1047, "step": 38052 }, { "epoch": 3.83, "learning_rate": 1.709793497657519e-07, "loss": 0.2687, "step": 38053 }, { "epoch": 3.83, "learning_rate": 1.707243038942563e-07, "loss": 0.0894, "step": 38054 }, { "epoch": 3.83, "learning_rate": 1.7046944784492868e-07, "loss": 0.1265, "step": 38055 }, { "epoch": 3.83, "learning_rate": 1.7021478161938442e-07, "loss": 0.47, "step": 38056 }, { "epoch": 3.83, "learning_rate": 1.6996030521924555e-07, "loss": 0.0183, "step": 38057 }, { "epoch": 3.83, "learning_rate": 1.6970601864613077e-07, "loss": 0.189, "step": 38058 }, { "epoch": 3.83, "learning_rate": 1.6945192190165882e-07, "loss": 0.319, "step": 38059 }, { "epoch": 3.83, "learning_rate": 1.6919801498744835e-07, "loss": 0.3379, "step": 38060 }, { "epoch": 3.83, "learning_rate": 1.6894429790511145e-07, "loss": 0.4953, "step": 38061 }, { "epoch": 3.83, "learning_rate": 1.6869077065626347e-07, "loss": 0.1284, "step": 38062 }, { "epoch": 3.83, "learning_rate": 1.6843743324252313e-07, "loss": 0.1689, "step": 38063 }, { "epoch": 3.83, "learning_rate": 1.681842856654958e-07, "loss": 0.338, "step": 38064 }, { "epoch": 3.83, "learning_rate": 1.679313279267969e-07, "loss": 0.2273, "step": 38065 }, { "epoch": 3.83, "learning_rate": 1.6767856002803173e-07, "loss": 0.1478, "step": 38066 }, { "epoch": 3.83, "learning_rate": 1.6742598197081238e-07, "loss": 0.3629, "step": 38067 }, { "epoch": 3.83, "learning_rate": 1.6717359375674756e-07, "loss": 0.0809, "step": 38068 }, { "epoch": 3.83, "learning_rate": 1.66921395387436e-07, "loss": 0.0276, "step": 38069 }, { "epoch": 3.83, "learning_rate": 1.6666938686449307e-07, "loss": 0.0189, "step": 38070 }, { "epoch": 3.83, "learning_rate": 1.6641756818951414e-07, "loss": 0.0563, "step": 38071 }, { "epoch": 3.83, "learning_rate": 1.661659393641013e-07, "loss": 0.2775, "step": 38072 }, { "epoch": 3.83, "learning_rate": 1.6591450038986322e-07, "loss": 0.1143, "step": 38073 }, { "epoch": 3.83, "learning_rate": 1.6566325126839864e-07, "loss": 0.1171, "step": 38074 }, { "epoch": 3.83, "learning_rate": 1.654121920012963e-07, "loss": 0.2131, "step": 38075 }, { "epoch": 3.83, "learning_rate": 1.6516132259016492e-07, "loss": 0.2031, "step": 38076 }, { "epoch": 3.83, "learning_rate": 1.6491064303659986e-07, "loss": 0.0818, "step": 38077 }, { "epoch": 3.83, "learning_rate": 1.6466015334218987e-07, "loss": 0.039, "step": 38078 }, { "epoch": 3.83, "learning_rate": 1.6440985350853367e-07, "loss": 0.1806, "step": 38079 }, { "epoch": 3.83, "learning_rate": 1.641597435372233e-07, "loss": 0.099, "step": 38080 }, { "epoch": 3.83, "learning_rate": 1.6390982342985083e-07, "loss": 0.3814, "step": 38081 }, { "epoch": 3.83, "learning_rate": 1.6366009318800502e-07, "loss": 0.1054, "step": 38082 }, { "epoch": 3.83, "learning_rate": 1.6341055281327455e-07, "loss": 0.2731, "step": 38083 }, { "epoch": 3.83, "learning_rate": 1.6316120230725152e-07, "loss": 0.1974, "step": 38084 }, { "epoch": 3.83, "learning_rate": 1.629120416715213e-07, "loss": 0.2589, "step": 38085 }, { "epoch": 3.83, "learning_rate": 1.6266307090766263e-07, "loss": 0.109, "step": 38086 }, { "epoch": 3.83, "learning_rate": 1.6241429001727093e-07, "loss": 0.3099, "step": 38087 }, { "epoch": 3.83, "learning_rate": 1.6216569900192156e-07, "loss": 0.34, "step": 38088 }, { "epoch": 3.83, "learning_rate": 1.6191729786319997e-07, "loss": 0.4607, "step": 38089 }, { "epoch": 3.83, "learning_rate": 1.6166908660268486e-07, "loss": 0.2596, "step": 38090 }, { "epoch": 3.83, "learning_rate": 1.6142106522195498e-07, "loss": 0.2663, "step": 38091 }, { "epoch": 3.83, "learning_rate": 1.611732337225924e-07, "loss": 0.1732, "step": 38092 }, { "epoch": 3.83, "learning_rate": 1.6092559210617253e-07, "loss": 0.1451, "step": 38093 }, { "epoch": 3.83, "learning_rate": 1.6067814037426742e-07, "loss": 0.1053, "step": 38094 }, { "epoch": 3.83, "learning_rate": 1.6043087852845917e-07, "loss": 0.0772, "step": 38095 }, { "epoch": 3.83, "learning_rate": 1.601838065703165e-07, "loss": 0.0711, "step": 38096 }, { "epoch": 3.83, "learning_rate": 1.599369245014115e-07, "loss": 0.2353, "step": 38097 }, { "epoch": 3.83, "learning_rate": 1.5969023232331625e-07, "loss": 0.3773, "step": 38098 }, { "epoch": 3.83, "learning_rate": 1.5944373003759948e-07, "loss": 0.2133, "step": 38099 }, { "epoch": 3.83, "learning_rate": 1.591974176458333e-07, "loss": 0.3013, "step": 38100 }, { "epoch": 3.83, "learning_rate": 1.5895129514957974e-07, "loss": 0.2831, "step": 38101 }, { "epoch": 3.83, "learning_rate": 1.587053625504109e-07, "loss": 0.3708, "step": 38102 }, { "epoch": 3.83, "learning_rate": 1.5845961984988888e-07, "loss": 0.4293, "step": 38103 }, { "epoch": 3.83, "learning_rate": 1.5821406704957575e-07, "loss": 0.4112, "step": 38104 }, { "epoch": 3.83, "learning_rate": 1.5796870415103692e-07, "loss": 0.2917, "step": 38105 }, { "epoch": 3.83, "learning_rate": 1.5772353115583115e-07, "loss": 0.2388, "step": 38106 }, { "epoch": 3.83, "learning_rate": 1.5747854806552054e-07, "loss": 0.3894, "step": 38107 }, { "epoch": 3.83, "learning_rate": 1.5723375488166714e-07, "loss": 0.1991, "step": 38108 }, { "epoch": 3.83, "learning_rate": 1.5698915160582305e-07, "loss": 0.4047, "step": 38109 }, { "epoch": 3.83, "learning_rate": 1.5674473823954705e-07, "loss": 0.2184, "step": 38110 }, { "epoch": 3.83, "learning_rate": 1.5650051478439453e-07, "loss": 0.0624, "step": 38111 }, { "epoch": 3.83, "learning_rate": 1.562564812419176e-07, "loss": 0.1543, "step": 38112 }, { "epoch": 3.83, "learning_rate": 1.5601263761367502e-07, "loss": 0.0801, "step": 38113 }, { "epoch": 3.83, "learning_rate": 1.5576898390121218e-07, "loss": 0.0973, "step": 38114 }, { "epoch": 3.83, "learning_rate": 1.5552552010608123e-07, "loss": 0.4966, "step": 38115 }, { "epoch": 3.83, "learning_rate": 1.5528224622983757e-07, "loss": 0.3822, "step": 38116 }, { "epoch": 3.83, "learning_rate": 1.5503916227401992e-07, "loss": 0.3374, "step": 38117 }, { "epoch": 3.83, "learning_rate": 1.5479626824018045e-07, "loss": 0.4357, "step": 38118 }, { "epoch": 3.83, "learning_rate": 1.5455356412986455e-07, "loss": 0.0494, "step": 38119 }, { "epoch": 3.83, "learning_rate": 1.5431104994461432e-07, "loss": 0.1224, "step": 38120 }, { "epoch": 3.83, "learning_rate": 1.5406872568597517e-07, "loss": 0.3744, "step": 38121 }, { "epoch": 3.83, "learning_rate": 1.5382659135548926e-07, "loss": 0.5848, "step": 38122 }, { "epoch": 3.83, "learning_rate": 1.5358464695469864e-07, "loss": 0.239, "step": 38123 }, { "epoch": 3.83, "learning_rate": 1.5334289248513879e-07, "loss": 0.0983, "step": 38124 }, { "epoch": 3.83, "learning_rate": 1.5310132794835176e-07, "loss": 0.0148, "step": 38125 }, { "epoch": 3.83, "learning_rate": 1.5285995334587298e-07, "loss": 0.6424, "step": 38126 }, { "epoch": 3.83, "learning_rate": 1.5261876867923797e-07, "loss": 0.1336, "step": 38127 }, { "epoch": 3.83, "learning_rate": 1.5237777394998542e-07, "loss": 0.1091, "step": 38128 }, { "epoch": 3.83, "learning_rate": 1.5213696915964414e-07, "loss": 0.0697, "step": 38129 }, { "epoch": 3.83, "learning_rate": 1.5189635430974625e-07, "loss": 0.177, "step": 38130 }, { "epoch": 3.83, "learning_rate": 1.5165592940182716e-07, "loss": 0.2253, "step": 38131 }, { "epoch": 3.84, "learning_rate": 1.5141569443741566e-07, "loss": 0.1744, "step": 38132 }, { "epoch": 3.84, "learning_rate": 1.511756494180405e-07, "loss": 0.3283, "step": 38133 }, { "epoch": 3.84, "learning_rate": 1.509357943452272e-07, "loss": 0.2384, "step": 38134 }, { "epoch": 3.84, "learning_rate": 1.5069612922050113e-07, "loss": 0.0931, "step": 38135 }, { "epoch": 3.84, "learning_rate": 1.5045665404539112e-07, "loss": 0.0587, "step": 38136 }, { "epoch": 3.84, "learning_rate": 1.5021736882142257e-07, "loss": 0.2403, "step": 38137 }, { "epoch": 3.84, "learning_rate": 1.4997827355011095e-07, "loss": 0.1463, "step": 38138 }, { "epoch": 3.84, "learning_rate": 1.4973936823298507e-07, "loss": 0.1607, "step": 38139 }, { "epoch": 3.84, "learning_rate": 1.4950065287156034e-07, "loss": 0.0731, "step": 38140 }, { "epoch": 3.84, "learning_rate": 1.4926212746735889e-07, "loss": 0.3398, "step": 38141 }, { "epoch": 3.84, "learning_rate": 1.4902379202189952e-07, "loss": 0.0111, "step": 38142 }, { "epoch": 3.84, "learning_rate": 1.4878564653669435e-07, "loss": 0.3994, "step": 38143 }, { "epoch": 3.84, "learning_rate": 1.4854769101326216e-07, "loss": 0.41, "step": 38144 }, { "epoch": 3.84, "learning_rate": 1.4830992545311506e-07, "loss": 0.5022, "step": 38145 }, { "epoch": 3.84, "learning_rate": 1.480723498577685e-07, "loss": 0.1552, "step": 38146 }, { "epoch": 3.84, "learning_rate": 1.4783496422873132e-07, "loss": 0.1301, "step": 38147 }, { "epoch": 3.84, "learning_rate": 1.4759776856751893e-07, "loss": 0.18, "step": 38148 }, { "epoch": 3.84, "learning_rate": 1.4736076287563683e-07, "loss": 0.1633, "step": 38149 }, { "epoch": 3.84, "learning_rate": 1.4712394715459376e-07, "loss": 0.4259, "step": 38150 }, { "epoch": 3.84, "learning_rate": 1.4688732140589523e-07, "loss": 0.0878, "step": 38151 }, { "epoch": 3.84, "learning_rate": 1.4665088563105332e-07, "loss": 0.1188, "step": 38152 }, { "epoch": 3.84, "learning_rate": 1.4641463983156689e-07, "loss": 0.1558, "step": 38153 }, { "epoch": 3.84, "learning_rate": 1.4617858400893803e-07, "loss": 0.0847, "step": 38154 }, { "epoch": 3.84, "learning_rate": 1.4594271816467552e-07, "loss": 0.0886, "step": 38155 }, { "epoch": 3.84, "learning_rate": 1.4570704230027486e-07, "loss": 0.2397, "step": 38156 }, { "epoch": 3.84, "learning_rate": 1.4547155641723818e-07, "loss": 0.2979, "step": 38157 }, { "epoch": 3.84, "learning_rate": 1.4523626051706095e-07, "loss": 0.1563, "step": 38158 }, { "epoch": 3.84, "learning_rate": 1.450011546012453e-07, "loss": 0.0132, "step": 38159 }, { "epoch": 3.84, "learning_rate": 1.447662386712867e-07, "loss": 0.2217, "step": 38160 }, { "epoch": 3.84, "learning_rate": 1.4453151272867393e-07, "loss": 0.1358, "step": 38161 }, { "epoch": 3.84, "learning_rate": 1.4429697677490582e-07, "loss": 0.0795, "step": 38162 }, { "epoch": 3.84, "learning_rate": 1.4406263081147786e-07, "loss": 0.3455, "step": 38163 }, { "epoch": 3.84, "learning_rate": 1.4382847483987217e-07, "loss": 0.1466, "step": 38164 }, { "epoch": 3.84, "learning_rate": 1.4359450886158754e-07, "loss": 0.3089, "step": 38165 }, { "epoch": 3.84, "learning_rate": 1.4336073287810946e-07, "loss": 0.1107, "step": 38166 }, { "epoch": 3.84, "learning_rate": 1.4312714689092342e-07, "loss": 0.8256, "step": 38167 }, { "epoch": 3.84, "learning_rate": 1.4289375090151824e-07, "loss": 0.3308, "step": 38168 }, { "epoch": 3.84, "learning_rate": 1.426605449113827e-07, "loss": 0.1571, "step": 38169 }, { "epoch": 3.84, "learning_rate": 1.4242752892199562e-07, "loss": 0.0845, "step": 38170 }, { "epoch": 3.84, "learning_rate": 1.4219470293483915e-07, "loss": 0.1461, "step": 38171 }, { "epoch": 3.84, "learning_rate": 1.419620669513988e-07, "loss": 0.0647, "step": 38172 }, { "epoch": 3.84, "learning_rate": 1.4172962097315333e-07, "loss": 0.1363, "step": 38173 }, { "epoch": 3.84, "learning_rate": 1.414973650015816e-07, "loss": 0.0692, "step": 38174 }, { "epoch": 3.84, "learning_rate": 1.4126529903816242e-07, "loss": 0.3657, "step": 38175 }, { "epoch": 3.84, "learning_rate": 1.4103342308437127e-07, "loss": 0.3151, "step": 38176 }, { "epoch": 3.84, "learning_rate": 1.4080173714168364e-07, "loss": 0.3953, "step": 38177 }, { "epoch": 3.84, "learning_rate": 1.4057024121157503e-07, "loss": 0.0792, "step": 38178 }, { "epoch": 3.84, "learning_rate": 1.4033893529551757e-07, "loss": 0.0813, "step": 38179 }, { "epoch": 3.84, "learning_rate": 1.4010781939498674e-07, "loss": 0.2067, "step": 38180 }, { "epoch": 3.84, "learning_rate": 1.3987689351144806e-07, "loss": 0.0968, "step": 38181 }, { "epoch": 3.84, "learning_rate": 1.3964615764637367e-07, "loss": 0.1862, "step": 38182 }, { "epoch": 3.84, "learning_rate": 1.3941561180123242e-07, "loss": 0.1665, "step": 38183 }, { "epoch": 3.84, "learning_rate": 1.391852559774931e-07, "loss": 0.2934, "step": 38184 }, { "epoch": 3.84, "learning_rate": 1.3895509017661457e-07, "loss": 0.3203, "step": 38185 }, { "epoch": 3.84, "learning_rate": 1.3872511440006897e-07, "loss": 0.0294, "step": 38186 }, { "epoch": 3.84, "learning_rate": 1.3849532864931512e-07, "loss": 0.1498, "step": 38187 }, { "epoch": 3.84, "learning_rate": 1.3826573292581856e-07, "loss": 0.124, "step": 38188 }, { "epoch": 3.84, "learning_rate": 1.3803632723104142e-07, "loss": 0.7904, "step": 38189 }, { "epoch": 3.84, "learning_rate": 1.3780711156643588e-07, "loss": 0.2481, "step": 38190 }, { "epoch": 3.84, "learning_rate": 1.3757808593346743e-07, "loss": 0.061, "step": 38191 }, { "epoch": 3.84, "learning_rate": 1.373492503335949e-07, "loss": 0.357, "step": 38192 }, { "epoch": 3.84, "learning_rate": 1.3712060476826716e-07, "loss": 0.159, "step": 38193 }, { "epoch": 3.84, "learning_rate": 1.3689214923894632e-07, "loss": 0.1242, "step": 38194 }, { "epoch": 3.84, "learning_rate": 1.3666388374708127e-07, "loss": 0.23, "step": 38195 }, { "epoch": 3.84, "learning_rate": 1.364358082941275e-07, "loss": 0.2083, "step": 38196 }, { "epoch": 3.84, "learning_rate": 1.3620792288153382e-07, "loss": 0.2409, "step": 38197 }, { "epoch": 3.84, "learning_rate": 1.359802275107558e-07, "loss": 0.2622, "step": 38198 }, { "epoch": 3.84, "learning_rate": 1.3575272218323552e-07, "loss": 0.1323, "step": 38199 }, { "epoch": 3.84, "learning_rate": 1.3552540690042525e-07, "loss": 0.3774, "step": 38200 }, { "epoch": 3.84, "learning_rate": 1.3529828166377378e-07, "loss": 0.1258, "step": 38201 }, { "epoch": 3.84, "learning_rate": 1.3507134647471664e-07, "loss": 0.2021, "step": 38202 }, { "epoch": 3.84, "learning_rate": 1.3484460133470933e-07, "loss": 0.0848, "step": 38203 }, { "epoch": 3.84, "learning_rate": 1.3461804624518737e-07, "loss": 0.0729, "step": 38204 }, { "epoch": 3.84, "learning_rate": 1.3439168120759627e-07, "loss": 0.0796, "step": 38205 }, { "epoch": 3.84, "learning_rate": 1.3416550622337153e-07, "loss": 0.0526, "step": 38206 }, { "epoch": 3.84, "learning_rate": 1.339395212939587e-07, "loss": 0.2753, "step": 38207 }, { "epoch": 3.84, "learning_rate": 1.3371372642079328e-07, "loss": 0.1779, "step": 38208 }, { "epoch": 3.84, "learning_rate": 1.3348812160531076e-07, "loss": 0.0521, "step": 38209 }, { "epoch": 3.84, "learning_rate": 1.3326270684895003e-07, "loss": 0.3124, "step": 38210 }, { "epoch": 3.84, "learning_rate": 1.3303748215313993e-07, "loss": 0.3476, "step": 38211 }, { "epoch": 3.84, "learning_rate": 1.328124475193193e-07, "loss": 0.2495, "step": 38212 }, { "epoch": 3.84, "learning_rate": 1.3258760294892035e-07, "loss": 0.2697, "step": 38213 }, { "epoch": 3.84, "learning_rate": 1.323629484433686e-07, "loss": 0.0824, "step": 38214 }, { "epoch": 3.84, "learning_rate": 1.3213848400409622e-07, "loss": 0.326, "step": 38215 }, { "epoch": 3.84, "learning_rate": 1.3191420963253542e-07, "loss": 0.3726, "step": 38216 }, { "epoch": 3.84, "learning_rate": 1.3169012533010837e-07, "loss": 0.0317, "step": 38217 }, { "epoch": 3.84, "learning_rate": 1.3146623109824395e-07, "loss": 0.3436, "step": 38218 }, { "epoch": 3.84, "learning_rate": 1.3124252693836436e-07, "loss": 0.4199, "step": 38219 }, { "epoch": 3.84, "learning_rate": 1.310190128518951e-07, "loss": 0.4635, "step": 38220 }, { "epoch": 3.84, "learning_rate": 1.3079568884025839e-07, "loss": 0.2323, "step": 38221 }, { "epoch": 3.84, "learning_rate": 1.3057255490487641e-07, "loss": 0.1122, "step": 38222 }, { "epoch": 3.84, "learning_rate": 1.3034961104716803e-07, "loss": 0.3184, "step": 38223 }, { "epoch": 3.84, "learning_rate": 1.3012685726854877e-07, "loss": 0.193, "step": 38224 }, { "epoch": 3.84, "learning_rate": 1.2990429357044088e-07, "loss": 0.1149, "step": 38225 }, { "epoch": 3.84, "learning_rate": 1.296819199542598e-07, "loss": 0.0068, "step": 38226 }, { "epoch": 3.84, "learning_rate": 1.2945973642141783e-07, "loss": 0.4767, "step": 38227 }, { "epoch": 3.84, "learning_rate": 1.2923774297333046e-07, "loss": 0.2691, "step": 38228 }, { "epoch": 3.84, "learning_rate": 1.2901593961141322e-07, "loss": 0.1169, "step": 38229 }, { "epoch": 3.85, "learning_rate": 1.28794326337075e-07, "loss": 0.3211, "step": 38230 }, { "epoch": 3.85, "learning_rate": 1.2857290315172132e-07, "loss": 0.4017, "step": 38231 }, { "epoch": 3.85, "learning_rate": 1.2835167005677106e-07, "loss": 0.5358, "step": 38232 }, { "epoch": 3.85, "learning_rate": 1.281306270536231e-07, "loss": 0.1862, "step": 38233 }, { "epoch": 3.85, "learning_rate": 1.2790977414369298e-07, "loss": 0.1187, "step": 38234 }, { "epoch": 3.85, "learning_rate": 1.2768911132837623e-07, "loss": 0.0258, "step": 38235 }, { "epoch": 3.85, "learning_rate": 1.2746863860908509e-07, "loss": 0.1002, "step": 38236 }, { "epoch": 3.85, "learning_rate": 1.272483559872184e-07, "loss": 0.1733, "step": 38237 }, { "epoch": 3.85, "learning_rate": 1.270282634641784e-07, "loss": 0.2141, "step": 38238 }, { "epoch": 3.85, "learning_rate": 1.2680836104136728e-07, "loss": 0.273, "step": 38239 }, { "epoch": 3.85, "learning_rate": 1.2658864872018393e-07, "loss": 0.1428, "step": 38240 }, { "epoch": 3.85, "learning_rate": 1.2636912650202393e-07, "loss": 0.1594, "step": 38241 }, { "epoch": 3.85, "learning_rate": 1.2614979438828612e-07, "loss": 0.1316, "step": 38242 }, { "epoch": 3.85, "learning_rate": 1.2593065238036938e-07, "loss": 0.1382, "step": 38243 }, { "epoch": 3.85, "learning_rate": 1.2571170047966264e-07, "loss": 0.0335, "step": 38244 }, { "epoch": 3.85, "learning_rate": 1.2549293868756474e-07, "loss": 0.4291, "step": 38245 }, { "epoch": 3.85, "learning_rate": 1.2527436700546124e-07, "loss": 0.2054, "step": 38246 }, { "epoch": 3.85, "learning_rate": 1.250559854347477e-07, "loss": 0.4099, "step": 38247 }, { "epoch": 3.85, "learning_rate": 1.2483779397681638e-07, "loss": 0.0899, "step": 38248 }, { "epoch": 3.85, "learning_rate": 1.2461979263304945e-07, "loss": 0.1461, "step": 38249 }, { "epoch": 3.85, "learning_rate": 1.2440198140483584e-07, "loss": 0.1524, "step": 38250 }, { "epoch": 3.85, "learning_rate": 1.241843602935644e-07, "loss": 0.27, "step": 38251 }, { "epoch": 3.85, "learning_rate": 1.2396692930061736e-07, "loss": 0.3841, "step": 38252 }, { "epoch": 3.85, "learning_rate": 1.2374968842738033e-07, "loss": 0.1466, "step": 38253 }, { "epoch": 3.85, "learning_rate": 1.2353263767523548e-07, "loss": 0.1454, "step": 38254 }, { "epoch": 3.85, "learning_rate": 1.2331577704556175e-07, "loss": 0.0743, "step": 38255 }, { "epoch": 3.85, "learning_rate": 1.2309910653974133e-07, "loss": 0.027, "step": 38256 }, { "epoch": 3.85, "learning_rate": 1.2288262615915315e-07, "loss": 0.7031, "step": 38257 }, { "epoch": 3.85, "learning_rate": 1.2266633590517273e-07, "loss": 0.4208, "step": 38258 }, { "epoch": 3.85, "learning_rate": 1.22450235779179e-07, "loss": 0.0421, "step": 38259 }, { "epoch": 3.85, "learning_rate": 1.2223432578254422e-07, "loss": 0.0799, "step": 38260 }, { "epoch": 3.85, "learning_rate": 1.2201860591664394e-07, "loss": 0.1125, "step": 38261 }, { "epoch": 3.85, "learning_rate": 1.218030761828537e-07, "loss": 0.2371, "step": 38262 }, { "epoch": 3.85, "learning_rate": 1.2158773658254242e-07, "loss": 0.0522, "step": 38263 }, { "epoch": 3.85, "learning_rate": 1.2137258711707898e-07, "loss": 0.1325, "step": 38264 }, { "epoch": 3.85, "learning_rate": 1.2115762778783236e-07, "loss": 0.0226, "step": 38265 }, { "epoch": 3.85, "learning_rate": 1.209428585961747e-07, "loss": 0.9231, "step": 38266 }, { "epoch": 3.85, "learning_rate": 1.2072827954347167e-07, "loss": 0.0387, "step": 38267 }, { "epoch": 3.85, "learning_rate": 1.205138906310821e-07, "loss": 0.2211, "step": 38268 }, { "epoch": 3.85, "learning_rate": 1.2029969186037825e-07, "loss": 0.2455, "step": 38269 }, { "epoch": 3.85, "learning_rate": 1.200856832327224e-07, "loss": 0.1396, "step": 38270 }, { "epoch": 3.85, "learning_rate": 1.198718647494701e-07, "loss": 0.1044, "step": 38271 }, { "epoch": 3.85, "learning_rate": 1.1965823641199024e-07, "loss": 0.3354, "step": 38272 }, { "epoch": 3.85, "learning_rate": 1.1944479822163512e-07, "loss": 0.151, "step": 38273 }, { "epoch": 3.85, "learning_rate": 1.1923155017976694e-07, "loss": 0.3117, "step": 38274 }, { "epoch": 3.85, "learning_rate": 1.1901849228774463e-07, "loss": 0.1925, "step": 38275 }, { "epoch": 3.85, "learning_rate": 1.1880562454692045e-07, "loss": 0.2019, "step": 38276 }, { "epoch": 3.85, "learning_rate": 1.1859294695864664e-07, "loss": 0.124, "step": 38277 }, { "epoch": 3.85, "learning_rate": 1.1838045952428211e-07, "loss": 0.2018, "step": 38278 }, { "epoch": 3.85, "learning_rate": 1.1816816224517579e-07, "loss": 0.1187, "step": 38279 }, { "epoch": 3.85, "learning_rate": 1.1795605512267993e-07, "loss": 0.1074, "step": 38280 }, { "epoch": 3.85, "learning_rate": 1.1774413815814678e-07, "loss": 0.1612, "step": 38281 }, { "epoch": 3.85, "learning_rate": 1.1753241135291859e-07, "loss": 0.2819, "step": 38282 }, { "epoch": 3.85, "learning_rate": 1.1732087470834762e-07, "loss": 0.0258, "step": 38283 }, { "epoch": 3.85, "learning_rate": 1.1710952822577947e-07, "loss": 0.0934, "step": 38284 }, { "epoch": 3.85, "learning_rate": 1.1689837190655639e-07, "loss": 0.2371, "step": 38285 }, { "epoch": 3.85, "learning_rate": 1.166874057520273e-07, "loss": 0.1464, "step": 38286 }, { "epoch": 3.85, "learning_rate": 1.164766297635278e-07, "loss": 0.0431, "step": 38287 }, { "epoch": 3.85, "learning_rate": 1.1626604394240347e-07, "loss": 0.0933, "step": 38288 }, { "epoch": 3.85, "learning_rate": 1.1605564828999327e-07, "loss": 0.2153, "step": 38289 }, { "epoch": 3.85, "learning_rate": 1.1584544280763942e-07, "loss": 0.1341, "step": 38290 }, { "epoch": 3.85, "learning_rate": 1.1563542749667422e-07, "loss": 0.2438, "step": 38291 }, { "epoch": 3.85, "learning_rate": 1.1542560235843658e-07, "loss": 0.1614, "step": 38292 }, { "epoch": 3.85, "learning_rate": 1.152159673942621e-07, "loss": 0.037, "step": 38293 }, { "epoch": 3.85, "learning_rate": 1.1500652260548638e-07, "loss": 0.1191, "step": 38294 }, { "epoch": 3.85, "learning_rate": 1.1479726799343504e-07, "loss": 0.2067, "step": 38295 }, { "epoch": 3.85, "learning_rate": 1.1458820355945032e-07, "loss": 0.1497, "step": 38296 }, { "epoch": 3.85, "learning_rate": 1.1437932930485451e-07, "loss": 0.1929, "step": 38297 }, { "epoch": 3.85, "learning_rate": 1.1417064523097987e-07, "loss": 0.4063, "step": 38298 }, { "epoch": 3.85, "learning_rate": 1.1396215133915533e-07, "loss": 0.3856, "step": 38299 }, { "epoch": 3.85, "learning_rate": 1.137538476307065e-07, "loss": 0.2093, "step": 38300 }, { "epoch": 3.85, "learning_rate": 1.1354573410695901e-07, "loss": 0.2558, "step": 38301 }, { "epoch": 3.85, "learning_rate": 1.133378107692351e-07, "loss": 0.0263, "step": 38302 }, { "epoch": 3.85, "learning_rate": 1.1313007761886041e-07, "loss": 0.1581, "step": 38303 }, { "epoch": 3.85, "learning_rate": 1.1292253465715719e-07, "loss": 0.2408, "step": 38304 }, { "epoch": 3.85, "learning_rate": 1.1271518188544439e-07, "loss": 0.2058, "step": 38305 }, { "epoch": 3.85, "learning_rate": 1.1250801930504428e-07, "loss": 0.0967, "step": 38306 }, { "epoch": 3.85, "learning_rate": 1.123010469172725e-07, "loss": 0.1902, "step": 38307 }, { "epoch": 3.85, "learning_rate": 1.1209426472344797e-07, "loss": 0.1358, "step": 38308 }, { "epoch": 3.85, "learning_rate": 1.1188767272488298e-07, "loss": 0.5663, "step": 38309 }, { "epoch": 3.85, "learning_rate": 1.1168127092289648e-07, "loss": 0.1164, "step": 38310 }, { "epoch": 3.85, "learning_rate": 1.1147505931879742e-07, "loss": 0.2615, "step": 38311 }, { "epoch": 3.85, "learning_rate": 1.1126903791390475e-07, "loss": 0.1706, "step": 38312 }, { "epoch": 3.85, "learning_rate": 1.1106320670952408e-07, "loss": 0.4956, "step": 38313 }, { "epoch": 3.85, "learning_rate": 1.108575657069677e-07, "loss": 0.0955, "step": 38314 }, { "epoch": 3.85, "learning_rate": 1.1065211490754123e-07, "loss": 0.3357, "step": 38315 }, { "epoch": 3.85, "learning_rate": 1.1044685431255697e-07, "loss": 0.3171, "step": 38316 }, { "epoch": 3.85, "learning_rate": 1.1024178392331718e-07, "loss": 0.2736, "step": 38317 }, { "epoch": 3.85, "learning_rate": 1.1003690374112752e-07, "loss": 0.0628, "step": 38318 }, { "epoch": 3.85, "learning_rate": 1.0983221376729358e-07, "loss": 0.2326, "step": 38319 }, { "epoch": 3.85, "learning_rate": 1.0962771400311433e-07, "loss": 0.1633, "step": 38320 }, { "epoch": 3.85, "learning_rate": 1.094234044498954e-07, "loss": 0.149, "step": 38321 }, { "epoch": 3.85, "learning_rate": 1.0921928510893575e-07, "loss": 0.105, "step": 38322 }, { "epoch": 3.85, "learning_rate": 1.0901535598153101e-07, "loss": 0.2155, "step": 38323 }, { "epoch": 3.85, "learning_rate": 1.0881161706898346e-07, "loss": 0.328, "step": 38324 }, { "epoch": 3.85, "learning_rate": 1.0860806837258541e-07, "loss": 0.279, "step": 38325 }, { "epoch": 3.85, "learning_rate": 1.0840470989363583e-07, "loss": 0.0608, "step": 38326 }, { "epoch": 3.85, "learning_rate": 1.08201541633427e-07, "loss": 0.1131, "step": 38327 }, { "epoch": 3.85, "learning_rate": 1.0799856359325455e-07, "loss": 0.1447, "step": 38328 }, { "epoch": 3.86, "learning_rate": 1.0779577577440413e-07, "loss": 0.257, "step": 38329 }, { "epoch": 3.86, "learning_rate": 1.0759317817817138e-07, "loss": 0.2625, "step": 38330 }, { "epoch": 3.86, "learning_rate": 1.0739077080584525e-07, "loss": 0.0804, "step": 38331 }, { "epoch": 3.86, "learning_rate": 1.0718855365871139e-07, "loss": 0.2612, "step": 38332 }, { "epoch": 3.86, "learning_rate": 1.0698652673805543e-07, "loss": 0.0366, "step": 38333 }, { "epoch": 3.86, "learning_rate": 1.0678469004516966e-07, "loss": 0.0844, "step": 38334 }, { "epoch": 3.86, "learning_rate": 1.0658304358132976e-07, "loss": 0.291, "step": 38335 }, { "epoch": 3.86, "learning_rate": 1.06381587347828e-07, "loss": 0.0926, "step": 38336 }, { "epoch": 3.86, "learning_rate": 1.0618032134593669e-07, "loss": 0.1476, "step": 38337 }, { "epoch": 3.86, "learning_rate": 1.0597924557694483e-07, "loss": 0.1216, "step": 38338 }, { "epoch": 3.86, "learning_rate": 1.0577836004212804e-07, "loss": 0.2449, "step": 38339 }, { "epoch": 3.86, "learning_rate": 1.0557766474276532e-07, "loss": 0.1178, "step": 38340 }, { "epoch": 3.86, "learning_rate": 1.0537715968013561e-07, "loss": 0.2194, "step": 38341 }, { "epoch": 3.86, "learning_rate": 1.0517684485551126e-07, "loss": 0.0483, "step": 38342 }, { "epoch": 3.86, "learning_rate": 1.0497672027016791e-07, "loss": 0.253, "step": 38343 }, { "epoch": 3.86, "learning_rate": 1.047767859253812e-07, "loss": 0.2643, "step": 38344 }, { "epoch": 3.86, "learning_rate": 1.0457704182242345e-07, "loss": 0.1055, "step": 38345 }, { "epoch": 3.86, "learning_rate": 1.0437748796256364e-07, "loss": 0.1097, "step": 38346 }, { "epoch": 3.86, "learning_rate": 1.0417812434707407e-07, "loss": 0.1004, "step": 38347 }, { "epoch": 3.86, "learning_rate": 1.0397895097722043e-07, "loss": 0.2906, "step": 38348 }, { "epoch": 3.86, "learning_rate": 1.0377996785427168e-07, "loss": 0.1721, "step": 38349 }, { "epoch": 3.86, "learning_rate": 1.0358117497949348e-07, "loss": 0.0527, "step": 38350 }, { "epoch": 3.86, "learning_rate": 1.0338257235415149e-07, "loss": 0.1909, "step": 38351 }, { "epoch": 3.86, "learning_rate": 1.0318415997951136e-07, "loss": 0.1333, "step": 38352 }, { "epoch": 3.86, "learning_rate": 1.0298593785683208e-07, "loss": 0.0782, "step": 38353 }, { "epoch": 3.86, "learning_rate": 1.027879059873793e-07, "loss": 0.7606, "step": 38354 }, { "epoch": 3.86, "learning_rate": 1.025900643724087e-07, "loss": 0.1809, "step": 38355 }, { "epoch": 3.86, "learning_rate": 1.0239241301317925e-07, "loss": 0.3504, "step": 38356 }, { "epoch": 3.86, "learning_rate": 1.0219495191095662e-07, "loss": 0.1502, "step": 38357 }, { "epoch": 3.86, "learning_rate": 1.0199768106698648e-07, "loss": 0.0523, "step": 38358 }, { "epoch": 3.86, "learning_rate": 1.0180060048253114e-07, "loss": 0.1085, "step": 38359 }, { "epoch": 3.86, "learning_rate": 1.0160371015884296e-07, "loss": 0.1517, "step": 38360 }, { "epoch": 3.86, "learning_rate": 1.0140701009717756e-07, "loss": 0.0927, "step": 38361 }, { "epoch": 3.86, "learning_rate": 1.0121050029878065e-07, "loss": 0.0536, "step": 38362 }, { "epoch": 3.86, "learning_rate": 1.0101418076490787e-07, "loss": 0.1166, "step": 38363 }, { "epoch": 3.86, "learning_rate": 1.008180514968049e-07, "loss": 0.2507, "step": 38364 }, { "epoch": 3.86, "learning_rate": 1.0062211249572073e-07, "loss": 0.219, "step": 38365 }, { "epoch": 3.86, "learning_rate": 1.0042636376290437e-07, "loss": 0.0344, "step": 38366 }, { "epoch": 3.86, "learning_rate": 1.002308052996015e-07, "loss": 0.1051, "step": 38367 }, { "epoch": 3.86, "learning_rate": 1.0003543710705442e-07, "loss": 0.1191, "step": 38368 }, { "epoch": 3.86, "learning_rate": 9.984025918650886e-08, "loss": 0.1656, "step": 38369 }, { "epoch": 3.86, "learning_rate": 9.964527153920044e-08, "loss": 0.1328, "step": 38370 }, { "epoch": 3.86, "learning_rate": 9.94504741663782e-08, "loss": 0.0677, "step": 38371 }, { "epoch": 3.86, "learning_rate": 9.925586706928113e-08, "loss": 0.1056, "step": 38372 }, { "epoch": 3.86, "learning_rate": 9.906145024914159e-08, "loss": 0.123, "step": 38373 }, { "epoch": 3.86, "learning_rate": 9.886722370719859e-08, "loss": 0.1877, "step": 38374 }, { "epoch": 3.86, "learning_rate": 9.867318744469111e-08, "loss": 0.1507, "step": 38375 }, { "epoch": 3.86, "learning_rate": 9.847934146285486e-08, "loss": 0.1534, "step": 38376 }, { "epoch": 3.86, "learning_rate": 9.828568576291885e-08, "loss": 0.4585, "step": 38377 }, { "epoch": 3.86, "learning_rate": 9.809222034611875e-08, "loss": 0.147, "step": 38378 }, { "epoch": 3.86, "learning_rate": 9.789894521368025e-08, "loss": 0.0001, "step": 38379 }, { "epoch": 3.86, "learning_rate": 9.770586036684237e-08, "loss": 0.0546, "step": 38380 }, { "epoch": 3.86, "learning_rate": 9.751296580682744e-08, "loss": 0.2032, "step": 38381 }, { "epoch": 3.86, "learning_rate": 9.732026153486451e-08, "loss": 0.5527, "step": 38382 }, { "epoch": 3.86, "learning_rate": 9.712774755217923e-08, "loss": 0.1214, "step": 38383 }, { "epoch": 3.86, "learning_rate": 9.693542385999732e-08, "loss": 0.092, "step": 38384 }, { "epoch": 3.86, "learning_rate": 9.674329045954444e-08, "loss": 0.1862, "step": 38385 }, { "epoch": 3.86, "learning_rate": 9.655134735203963e-08, "loss": 0.245, "step": 38386 }, { "epoch": 3.86, "learning_rate": 9.635959453870525e-08, "loss": 0.2644, "step": 38387 }, { "epoch": 3.86, "learning_rate": 9.616803202076362e-08, "loss": 0.096, "step": 38388 }, { "epoch": 3.86, "learning_rate": 9.597665979943382e-08, "loss": 0.1511, "step": 38389 }, { "epoch": 3.86, "learning_rate": 9.57854778759315e-08, "loss": 0.1719, "step": 38390 }, { "epoch": 3.86, "learning_rate": 9.559448625147571e-08, "loss": 0.1789, "step": 38391 }, { "epoch": 3.86, "learning_rate": 9.54036849272788e-08, "loss": 0.0733, "step": 38392 }, { "epoch": 3.86, "learning_rate": 9.521307390455647e-08, "loss": 0.0619, "step": 38393 }, { "epoch": 3.86, "learning_rate": 9.502265318452107e-08, "loss": 0.1099, "step": 38394 }, { "epoch": 3.86, "learning_rate": 9.483242276838833e-08, "loss": 0.1226, "step": 38395 }, { "epoch": 3.86, "learning_rate": 9.464238265736392e-08, "loss": 0.0698, "step": 38396 }, { "epoch": 3.86, "learning_rate": 9.445253285265686e-08, "loss": 0.4033, "step": 38397 }, { "epoch": 3.86, "learning_rate": 9.426287335547956e-08, "loss": 0.2631, "step": 38398 }, { "epoch": 3.86, "learning_rate": 9.407340416703436e-08, "loss": 0.019, "step": 38399 }, { "epoch": 3.86, "learning_rate": 9.38841252885303e-08, "loss": 0.0626, "step": 38400 }, { "epoch": 3.86, "learning_rate": 9.369503672117308e-08, "loss": 0.1048, "step": 38401 }, { "epoch": 3.86, "learning_rate": 9.35061384661584e-08, "loss": 0.3668, "step": 38402 }, { "epoch": 3.86, "learning_rate": 9.331743052469866e-08, "loss": 0.3519, "step": 38403 }, { "epoch": 3.86, "learning_rate": 9.312891289798952e-08, "loss": 0.5287, "step": 38404 }, { "epoch": 3.86, "learning_rate": 9.294058558722673e-08, "loss": 0.093, "step": 38405 }, { "epoch": 3.86, "learning_rate": 9.275244859361598e-08, "loss": 0.2053, "step": 38406 }, { "epoch": 3.86, "learning_rate": 9.256450191834964e-08, "loss": 0.2636, "step": 38407 }, { "epoch": 3.86, "learning_rate": 9.237674556262677e-08, "loss": 0.1385, "step": 38408 }, { "epoch": 3.86, "learning_rate": 9.218917952764305e-08, "loss": 0.0902, "step": 38409 }, { "epoch": 3.86, "learning_rate": 9.200180381458755e-08, "loss": 0.1963, "step": 38410 }, { "epoch": 3.86, "learning_rate": 9.181461842465266e-08, "loss": 0.3323, "step": 38411 }, { "epoch": 3.86, "learning_rate": 9.162762335903408e-08, "loss": 0.2406, "step": 38412 }, { "epoch": 3.86, "learning_rate": 9.144081861892084e-08, "loss": 0.069, "step": 38413 }, { "epoch": 3.86, "learning_rate": 9.125420420549868e-08, "loss": 0.0304, "step": 38414 }, { "epoch": 3.86, "learning_rate": 9.106778011995665e-08, "loss": 0.1122, "step": 38415 }, { "epoch": 3.86, "learning_rate": 9.088154636348378e-08, "loss": 0.1565, "step": 38416 }, { "epoch": 3.86, "learning_rate": 9.069550293725915e-08, "loss": 0.2673, "step": 38417 }, { "epoch": 3.86, "learning_rate": 9.050964984247512e-08, "loss": 0.0598, "step": 38418 }, { "epoch": 3.86, "learning_rate": 9.03239870803041e-08, "loss": 0.0235, "step": 38419 }, { "epoch": 3.86, "learning_rate": 9.013851465193513e-08, "loss": 0.2168, "step": 38420 }, { "epoch": 3.86, "learning_rate": 8.995323255854726e-08, "loss": 0.151, "step": 38421 }, { "epoch": 3.86, "learning_rate": 8.976814080131956e-08, "loss": 0.7275, "step": 38422 }, { "epoch": 3.86, "learning_rate": 8.958323938142776e-08, "loss": 0.0163, "step": 38423 }, { "epoch": 3.86, "learning_rate": 8.939852830004758e-08, "loss": 0.1865, "step": 38424 }, { "epoch": 3.86, "learning_rate": 8.921400755835806e-08, "loss": 0.2332, "step": 38425 }, { "epoch": 3.86, "learning_rate": 8.902967715753496e-08, "loss": 0.1432, "step": 38426 }, { "epoch": 3.87, "learning_rate": 8.884553709874399e-08, "loss": 0.4992, "step": 38427 }, { "epoch": 3.87, "learning_rate": 8.866158738316421e-08, "loss": 0.2133, "step": 38428 }, { "epoch": 3.87, "learning_rate": 8.847782801196135e-08, "loss": 0.1908, "step": 38429 }, { "epoch": 3.87, "learning_rate": 8.829425898630783e-08, "loss": 0.1032, "step": 38430 }, { "epoch": 3.87, "learning_rate": 8.811088030736936e-08, "loss": 0.1412, "step": 38431 }, { "epoch": 3.87, "learning_rate": 8.7927691976315e-08, "loss": 0.081, "step": 38432 }, { "epoch": 3.87, "learning_rate": 8.774469399430718e-08, "loss": 0.1441, "step": 38433 }, { "epoch": 3.87, "learning_rate": 8.756188636251828e-08, "loss": 0.0385, "step": 38434 }, { "epoch": 3.87, "learning_rate": 8.73792690821007e-08, "loss": 0.148, "step": 38435 }, { "epoch": 3.87, "learning_rate": 8.719684215422685e-08, "loss": 0.6251, "step": 38436 }, { "epoch": 3.87, "learning_rate": 8.701460558004914e-08, "loss": 0.254, "step": 38437 }, { "epoch": 3.87, "learning_rate": 8.683255936073332e-08, "loss": 0.2203, "step": 38438 }, { "epoch": 3.87, "learning_rate": 8.665070349743509e-08, "loss": 0.3105, "step": 38439 }, { "epoch": 3.87, "learning_rate": 8.646903799131023e-08, "loss": 0.0877, "step": 38440 }, { "epoch": 3.87, "learning_rate": 8.628756284352113e-08, "loss": 0.0833, "step": 38441 }, { "epoch": 3.87, "learning_rate": 8.610627805521354e-08, "loss": 0.2826, "step": 38442 }, { "epoch": 3.87, "learning_rate": 8.592518362754986e-08, "loss": 0.2088, "step": 38443 }, { "epoch": 3.87, "learning_rate": 8.574427956167586e-08, "loss": 0.1003, "step": 38444 }, { "epoch": 3.87, "learning_rate": 8.556356585874391e-08, "loss": 0.0893, "step": 38445 }, { "epoch": 3.87, "learning_rate": 8.53830425199098e-08, "loss": 0.239, "step": 38446 }, { "epoch": 3.87, "learning_rate": 8.52027095463126e-08, "loss": 0.187, "step": 38447 }, { "epoch": 3.87, "learning_rate": 8.502256693910804e-08, "loss": 0.1593, "step": 38448 }, { "epoch": 3.87, "learning_rate": 8.484261469943854e-08, "loss": 0.1261, "step": 38449 }, { "epoch": 3.87, "learning_rate": 8.466285282845321e-08, "loss": 0.0697, "step": 38450 }, { "epoch": 3.87, "learning_rate": 8.448328132729111e-08, "loss": 0.2411, "step": 38451 }, { "epoch": 3.87, "learning_rate": 8.4303900197098e-08, "loss": 0.4944, "step": 38452 }, { "epoch": 3.87, "learning_rate": 8.412470943901297e-08, "loss": 0.3185, "step": 38453 }, { "epoch": 3.87, "learning_rate": 8.394570905417842e-08, "loss": 0.1061, "step": 38454 }, { "epoch": 3.87, "learning_rate": 8.376689904373014e-08, "loss": 0.2268, "step": 38455 }, { "epoch": 3.87, "learning_rate": 8.358827940881053e-08, "loss": 0.3996, "step": 38456 }, { "epoch": 3.87, "learning_rate": 8.340985015055536e-08, "loss": 0.504, "step": 38457 }, { "epoch": 3.87, "learning_rate": 8.323161127010038e-08, "loss": 0.225, "step": 38458 }, { "epoch": 3.87, "learning_rate": 8.305356276857468e-08, "loss": 0.3548, "step": 38459 }, { "epoch": 3.87, "learning_rate": 8.287570464711736e-08, "loss": 0.1239, "step": 38460 }, { "epoch": 3.87, "learning_rate": 8.269803690685751e-08, "loss": 0.049, "step": 38461 }, { "epoch": 3.87, "learning_rate": 8.252055954892757e-08, "loss": 0.1372, "step": 38462 }, { "epoch": 3.87, "learning_rate": 8.234327257445329e-08, "loss": 0.2911, "step": 38463 }, { "epoch": 3.87, "learning_rate": 8.21661759845671e-08, "loss": 0.4766, "step": 38464 }, { "epoch": 3.87, "learning_rate": 8.198926978039146e-08, "loss": 0.2415, "step": 38465 }, { "epoch": 3.87, "learning_rate": 8.181255396305543e-08, "loss": 0.2383, "step": 38466 }, { "epoch": 3.87, "learning_rate": 8.16360285336848e-08, "loss": 0.0647, "step": 38467 }, { "epoch": 3.87, "learning_rate": 8.145969349340199e-08, "loss": 0.0554, "step": 38468 }, { "epoch": 3.87, "learning_rate": 8.128354884332612e-08, "loss": 0.0657, "step": 38469 }, { "epoch": 3.87, "learning_rate": 8.110759458457961e-08, "loss": 0.0826, "step": 38470 }, { "epoch": 3.87, "learning_rate": 8.093183071828159e-08, "loss": 0.5752, "step": 38471 }, { "epoch": 3.87, "learning_rate": 8.075625724555447e-08, "loss": 0.3448, "step": 38472 }, { "epoch": 3.87, "learning_rate": 8.05808741675107e-08, "loss": 0.207, "step": 38473 }, { "epoch": 3.87, "learning_rate": 8.040568148526939e-08, "loss": 0.128, "step": 38474 }, { "epoch": 3.87, "learning_rate": 8.023067919994299e-08, "loss": 0.1628, "step": 38475 }, { "epoch": 3.87, "learning_rate": 8.005586731264725e-08, "loss": 0.0948, "step": 38476 }, { "epoch": 3.87, "learning_rate": 7.988124582449463e-08, "loss": 0.1406, "step": 38477 }, { "epoch": 3.87, "learning_rate": 7.970681473659425e-08, "loss": 0.0936, "step": 38478 }, { "epoch": 3.87, "learning_rate": 7.953257405005854e-08, "loss": 0.1923, "step": 38479 }, { "epoch": 3.87, "learning_rate": 7.935852376599329e-08, "loss": 0.2615, "step": 38480 }, { "epoch": 3.87, "learning_rate": 7.918466388551093e-08, "loss": 0.1657, "step": 38481 }, { "epoch": 3.87, "learning_rate": 7.901099440971394e-08, "loss": 0.272, "step": 38482 }, { "epoch": 3.87, "learning_rate": 7.883751533970473e-08, "loss": 0.3328, "step": 38483 }, { "epoch": 3.87, "learning_rate": 7.866422667659579e-08, "loss": 0.1644, "step": 38484 }, { "epoch": 3.87, "learning_rate": 7.849112842148287e-08, "loss": 0.1552, "step": 38485 }, { "epoch": 3.87, "learning_rate": 7.831822057547178e-08, "loss": 0.188, "step": 38486 }, { "epoch": 3.87, "learning_rate": 7.814550313965829e-08, "loss": 0.2248, "step": 38487 }, { "epoch": 3.87, "learning_rate": 7.797297611514486e-08, "loss": 0.1445, "step": 38488 }, { "epoch": 3.87, "learning_rate": 7.78006395030273e-08, "loss": 0.079, "step": 38489 }, { "epoch": 3.87, "learning_rate": 7.762849330440469e-08, "loss": 0.0446, "step": 38490 }, { "epoch": 3.87, "learning_rate": 7.745653752037286e-08, "loss": 0.1711, "step": 38491 }, { "epoch": 3.87, "learning_rate": 7.728477215202423e-08, "loss": 0.1723, "step": 38492 }, { "epoch": 3.87, "learning_rate": 7.711319720044796e-08, "loss": 0.2141, "step": 38493 }, { "epoch": 3.87, "learning_rate": 7.694181266674316e-08, "loss": 0.1467, "step": 38494 }, { "epoch": 3.87, "learning_rate": 7.677061855199897e-08, "loss": 0.0503, "step": 38495 }, { "epoch": 3.87, "learning_rate": 7.659961485730116e-08, "loss": 0.0604, "step": 38496 }, { "epoch": 3.87, "learning_rate": 7.642880158373888e-08, "loss": 0.061, "step": 38497 }, { "epoch": 3.87, "learning_rate": 7.625817873240126e-08, "loss": 0.1283, "step": 38498 }, { "epoch": 3.87, "learning_rate": 7.60877463043741e-08, "loss": 0.2045, "step": 38499 }, { "epoch": 3.87, "learning_rate": 7.591750430073985e-08, "loss": 0.3902, "step": 38500 }, { "epoch": 3.87, "learning_rate": 7.574745272258099e-08, "loss": 0.1324, "step": 38501 }, { "epoch": 3.87, "learning_rate": 7.557759157098665e-08, "loss": 0.2169, "step": 38502 }, { "epoch": 3.87, "learning_rate": 7.54079208470293e-08, "loss": 0.1419, "step": 38503 }, { "epoch": 3.87, "learning_rate": 7.52384405517914e-08, "loss": 0.2017, "step": 38504 }, { "epoch": 3.87, "learning_rate": 7.506915068635212e-08, "loss": 0.0876, "step": 38505 }, { "epoch": 3.87, "learning_rate": 7.490005125179055e-08, "loss": 0.1248, "step": 38506 }, { "epoch": 3.87, "learning_rate": 7.473114224917921e-08, "loss": 0.3163, "step": 38507 }, { "epoch": 3.87, "learning_rate": 7.456242367959721e-08, "loss": 0.1432, "step": 38508 }, { "epoch": 3.87, "learning_rate": 7.439389554411369e-08, "loss": 0.0368, "step": 38509 }, { "epoch": 3.87, "learning_rate": 7.422555784380447e-08, "loss": 0.2237, "step": 38510 }, { "epoch": 3.87, "learning_rate": 7.40574105797387e-08, "loss": 0.1533, "step": 38511 }, { "epoch": 3.87, "learning_rate": 7.38894537529855e-08, "loss": 0.2554, "step": 38512 }, { "epoch": 3.87, "learning_rate": 7.372168736461737e-08, "loss": 0.0835, "step": 38513 }, { "epoch": 3.87, "learning_rate": 7.355411141570012e-08, "loss": 0.396, "step": 38514 }, { "epoch": 3.87, "learning_rate": 7.338672590729956e-08, "loss": 0.4138, "step": 38515 }, { "epoch": 3.87, "learning_rate": 7.321953084047816e-08, "loss": 0.2989, "step": 38516 }, { "epoch": 3.87, "learning_rate": 7.305252621630842e-08, "loss": 0.1892, "step": 38517 }, { "epoch": 3.87, "learning_rate": 7.288571203584282e-08, "loss": 0.3299, "step": 38518 }, { "epoch": 3.87, "learning_rate": 7.27190883001505e-08, "loss": 0.0614, "step": 38519 }, { "epoch": 3.87, "learning_rate": 7.255265501028729e-08, "loss": 0.2745, "step": 38520 }, { "epoch": 3.87, "learning_rate": 7.238641216731567e-08, "loss": 0.0831, "step": 38521 }, { "epoch": 3.87, "learning_rate": 7.222035977229146e-08, "loss": 0.0852, "step": 38522 }, { "epoch": 3.87, "learning_rate": 7.20544978262705e-08, "loss": 0.1557, "step": 38523 }, { "epoch": 3.87, "learning_rate": 7.188882633031191e-08, "loss": 0.1863, "step": 38524 }, { "epoch": 3.88, "learning_rate": 7.172334528546487e-08, "loss": 0.2066, "step": 38525 }, { "epoch": 3.88, "learning_rate": 7.15580546927852e-08, "loss": 0.1086, "step": 38526 }, { "epoch": 3.88, "learning_rate": 7.13929545533254e-08, "loss": 0.6437, "step": 38527 }, { "epoch": 3.88, "learning_rate": 7.122804486813795e-08, "loss": 0.0653, "step": 38528 }, { "epoch": 3.88, "learning_rate": 7.106332563826867e-08, "loss": 0.6271, "step": 38529 }, { "epoch": 3.88, "learning_rate": 7.089879686476342e-08, "loss": 0.2347, "step": 38530 }, { "epoch": 3.88, "learning_rate": 7.073445854867467e-08, "loss": 0.2063, "step": 38531 }, { "epoch": 3.88, "learning_rate": 7.057031069104825e-08, "loss": 0.0639, "step": 38532 }, { "epoch": 3.88, "learning_rate": 7.040635329292666e-08, "loss": 0.0417, "step": 38533 }, { "epoch": 3.88, "learning_rate": 7.024258635534908e-08, "loss": 0.1131, "step": 38534 }, { "epoch": 3.88, "learning_rate": 7.007900987936467e-08, "loss": 0.217, "step": 38535 }, { "epoch": 3.88, "learning_rate": 6.991562386601258e-08, "loss": 0.2448, "step": 38536 }, { "epoch": 3.88, "learning_rate": 6.975242831632867e-08, "loss": 0.0328, "step": 38537 }, { "epoch": 3.88, "learning_rate": 6.958942323135542e-08, "loss": 0.0476, "step": 38538 }, { "epoch": 3.88, "learning_rate": 6.942660861212869e-08, "loss": 0.1531, "step": 38539 }, { "epoch": 3.88, "learning_rate": 6.926398445968429e-08, "loss": 0.2269, "step": 38540 }, { "epoch": 3.88, "learning_rate": 6.910155077506142e-08, "loss": 0.1815, "step": 38541 }, { "epoch": 3.88, "learning_rate": 6.89393075592859e-08, "loss": 0.519, "step": 38542 }, { "epoch": 3.88, "learning_rate": 6.877725481339691e-08, "loss": 0.0948, "step": 38543 }, { "epoch": 3.88, "learning_rate": 6.86153925384203e-08, "loss": 0.2313, "step": 38544 }, { "epoch": 3.88, "learning_rate": 6.845372073539192e-08, "loss": 0.1841, "step": 38545 }, { "epoch": 3.88, "learning_rate": 6.829223940533424e-08, "loss": 0.1522, "step": 38546 }, { "epoch": 3.88, "learning_rate": 6.813094854927981e-08, "loss": 0.1032, "step": 38547 }, { "epoch": 3.88, "learning_rate": 6.796984816825446e-08, "loss": 0.2561, "step": 38548 }, { "epoch": 3.88, "learning_rate": 6.780893826328072e-08, "loss": 0.304, "step": 38549 }, { "epoch": 3.88, "learning_rate": 6.764821883538441e-08, "loss": 0.1425, "step": 38550 }, { "epoch": 3.88, "learning_rate": 6.748768988558807e-08, "loss": 0.0601, "step": 38551 }, { "epoch": 3.88, "learning_rate": 6.732735141491086e-08, "loss": 0.1687, "step": 38552 }, { "epoch": 3.88, "learning_rate": 6.716720342437865e-08, "loss": 0.0715, "step": 38553 }, { "epoch": 3.88, "learning_rate": 6.700724591500729e-08, "loss": 0.0403, "step": 38554 }, { "epoch": 3.88, "learning_rate": 6.684747888781262e-08, "loss": 0.3769, "step": 38555 }, { "epoch": 3.88, "learning_rate": 6.668790234381717e-08, "loss": 0.1628, "step": 38556 }, { "epoch": 3.88, "learning_rate": 6.652851628403012e-08, "loss": 0.2012, "step": 38557 }, { "epoch": 3.88, "learning_rate": 6.636932070946733e-08, "loss": 0.0644, "step": 38558 }, { "epoch": 3.88, "learning_rate": 6.621031562114465e-08, "loss": 0.1864, "step": 38559 }, { "epoch": 3.88, "learning_rate": 6.60515010200713e-08, "loss": 0.1306, "step": 38560 }, { "epoch": 3.88, "learning_rate": 6.589287690725976e-08, "loss": 0.2217, "step": 38561 }, { "epoch": 3.88, "learning_rate": 6.573444328371591e-08, "loss": 0.1029, "step": 38562 }, { "epoch": 3.88, "learning_rate": 6.55762001504523e-08, "loss": 0.032, "step": 38563 }, { "epoch": 3.88, "learning_rate": 6.541814750847474e-08, "loss": 0.0532, "step": 38564 }, { "epoch": 3.88, "learning_rate": 6.526028535878581e-08, "loss": 0.4922, "step": 38565 }, { "epoch": 3.88, "learning_rate": 6.510261370239467e-08, "loss": 0.3917, "step": 38566 }, { "epoch": 3.88, "learning_rate": 6.494513254030387e-08, "loss": 0.2306, "step": 38567 }, { "epoch": 3.88, "learning_rate": 6.478784187351262e-08, "loss": 0.2016, "step": 38568 }, { "epoch": 3.88, "learning_rate": 6.463074170302341e-08, "loss": 0.1924, "step": 38569 }, { "epoch": 3.88, "learning_rate": 6.447383202983881e-08, "loss": 0.2568, "step": 38570 }, { "epoch": 3.88, "learning_rate": 6.431711285495134e-08, "loss": 0.4148, "step": 38571 }, { "epoch": 3.88, "learning_rate": 6.416058417936022e-08, "loss": 0.0338, "step": 38572 }, { "epoch": 3.88, "learning_rate": 6.400424600406796e-08, "loss": 0.1772, "step": 38573 }, { "epoch": 3.88, "learning_rate": 6.384809833006378e-08, "loss": 0.245, "step": 38574 }, { "epoch": 3.88, "learning_rate": 6.369214115833688e-08, "loss": 0.2275, "step": 38575 }, { "epoch": 3.88, "learning_rate": 6.35363744898898e-08, "loss": 0.7108, "step": 38576 }, { "epoch": 3.88, "learning_rate": 6.338079832570843e-08, "loss": 0.1295, "step": 38577 }, { "epoch": 3.88, "learning_rate": 6.322541266678195e-08, "loss": 0.1082, "step": 38578 }, { "epoch": 3.88, "learning_rate": 6.307021751409958e-08, "loss": 0.0759, "step": 38579 }, { "epoch": 3.88, "learning_rate": 6.291521286865054e-08, "loss": 0.1776, "step": 38580 }, { "epoch": 3.88, "learning_rate": 6.27603987314207e-08, "loss": 0.2434, "step": 38581 }, { "epoch": 3.88, "learning_rate": 6.260577510339594e-08, "loss": 0.3041, "step": 38582 }, { "epoch": 3.88, "learning_rate": 6.245134198555547e-08, "loss": 0.6324, "step": 38583 }, { "epoch": 3.88, "learning_rate": 6.22970993788885e-08, "loss": 0.2868, "step": 38584 }, { "epoch": 3.88, "learning_rate": 6.214304728437425e-08, "loss": 0.1055, "step": 38585 }, { "epoch": 3.88, "learning_rate": 6.198918570299528e-08, "loss": 0.3273, "step": 38586 }, { "epoch": 3.88, "learning_rate": 6.183551463572412e-08, "loss": 0.1107, "step": 38587 }, { "epoch": 3.88, "learning_rate": 6.168203408354334e-08, "loss": 0.3256, "step": 38588 }, { "epoch": 3.88, "learning_rate": 6.152874404742881e-08, "loss": 0.3535, "step": 38589 }, { "epoch": 3.88, "learning_rate": 6.137564452835643e-08, "loss": 0.0661, "step": 38590 }, { "epoch": 3.88, "learning_rate": 6.122273552730207e-08, "loss": 0.1524, "step": 38591 }, { "epoch": 3.88, "learning_rate": 6.107001704523496e-08, "loss": 0.3685, "step": 38592 }, { "epoch": 3.88, "learning_rate": 6.091748908312767e-08, "loss": 0.2561, "step": 38593 }, { "epoch": 3.88, "learning_rate": 6.076515164195606e-08, "loss": 0.0216, "step": 38594 }, { "epoch": 3.88, "learning_rate": 6.061300472268272e-08, "loss": 0.0599, "step": 38595 }, { "epoch": 3.88, "learning_rate": 6.046104832628019e-08, "loss": 0.068, "step": 38596 }, { "epoch": 3.88, "learning_rate": 6.030928245371103e-08, "loss": 0.3555, "step": 38597 }, { "epoch": 3.88, "learning_rate": 6.015770710594448e-08, "loss": 0.1049, "step": 38598 }, { "epoch": 3.88, "learning_rate": 6.000632228394976e-08, "loss": 0.064, "step": 38599 }, { "epoch": 3.88, "learning_rate": 5.985512798867942e-08, "loss": 0.1168, "step": 38600 }, { "epoch": 3.88, "learning_rate": 5.970412422110272e-08, "loss": 0.0496, "step": 38601 }, { "epoch": 3.88, "learning_rate": 5.95533109821822e-08, "loss": 0.057, "step": 38602 }, { "epoch": 3.88, "learning_rate": 5.940268827287043e-08, "loss": 0.1326, "step": 38603 }, { "epoch": 3.88, "learning_rate": 5.9252256094129986e-08, "loss": 0.0584, "step": 38604 }, { "epoch": 3.88, "learning_rate": 5.910201444692009e-08, "loss": 0.1647, "step": 38605 }, { "epoch": 3.88, "learning_rate": 5.8951963332196656e-08, "loss": 0.2452, "step": 38606 }, { "epoch": 3.88, "learning_rate": 5.8802102750908914e-08, "loss": 0.1787, "step": 38607 }, { "epoch": 3.88, "learning_rate": 5.86524327040161e-08, "loss": 0.169, "step": 38608 }, { "epoch": 3.88, "learning_rate": 5.850295319246746e-08, "loss": 0.0956, "step": 38609 }, { "epoch": 3.88, "learning_rate": 5.8353664217218886e-08, "loss": 0.0949, "step": 38610 }, { "epoch": 3.88, "learning_rate": 5.820456577921296e-08, "loss": 0.1743, "step": 38611 }, { "epoch": 3.88, "learning_rate": 5.805565787940559e-08, "loss": 0.3225, "step": 38612 }, { "epoch": 3.88, "learning_rate": 5.790694051874268e-08, "loss": 0.3028, "step": 38613 }, { "epoch": 3.88, "learning_rate": 5.775841369816681e-08, "loss": 0.521, "step": 38614 }, { "epoch": 3.88, "learning_rate": 5.7610077418627227e-08, "loss": 0.2227, "step": 38615 }, { "epoch": 3.88, "learning_rate": 5.7461931681066506e-08, "loss": 0.3041, "step": 38616 }, { "epoch": 3.88, "learning_rate": 5.7313976486427224e-08, "loss": 0.2126, "step": 38617 }, { "epoch": 3.88, "learning_rate": 5.716621183564863e-08, "loss": 0.1795, "step": 38618 }, { "epoch": 3.88, "learning_rate": 5.7018637729676634e-08, "loss": 0.1492, "step": 38619 }, { "epoch": 3.88, "learning_rate": 5.6871254169443834e-08, "loss": 0.1442, "step": 38620 }, { "epoch": 3.88, "learning_rate": 5.672406115589279e-08, "loss": 0.0291, "step": 38621 }, { "epoch": 3.88, "learning_rate": 5.657705868995944e-08, "loss": 0.1182, "step": 38622 }, { "epoch": 3.88, "learning_rate": 5.6430246772579686e-08, "loss": 0.1103, "step": 38623 }, { "epoch": 3.89, "learning_rate": 5.628362540468612e-08, "loss": 0.2986, "step": 38624 }, { "epoch": 3.89, "learning_rate": 5.613719458721134e-08, "loss": 0.0946, "step": 38625 }, { "epoch": 3.89, "learning_rate": 5.5990954321091245e-08, "loss": 0.447, "step": 38626 }, { "epoch": 3.89, "learning_rate": 5.5844904607248446e-08, "loss": 0.3268, "step": 38627 }, { "epoch": 3.89, "learning_rate": 5.569904544662219e-08, "loss": 0.5125, "step": 38628 }, { "epoch": 3.89, "learning_rate": 5.555337684013173e-08, "loss": 0.0987, "step": 38629 }, { "epoch": 3.89, "learning_rate": 5.540789878870966e-08, "loss": 0.1321, "step": 38630 }, { "epoch": 3.89, "learning_rate": 5.526261129328192e-08, "loss": 0.2745, "step": 38631 }, { "epoch": 3.89, "learning_rate": 5.511751435477108e-08, "loss": 0.4996, "step": 38632 }, { "epoch": 3.89, "learning_rate": 5.497260797409642e-08, "loss": 0.1927, "step": 38633 }, { "epoch": 3.89, "learning_rate": 5.482789215219053e-08, "loss": 0.0701, "step": 38634 }, { "epoch": 3.89, "learning_rate": 5.4683366889966e-08, "loss": 0.476, "step": 38635 }, { "epoch": 3.89, "learning_rate": 5.453903218834211e-08, "loss": 0.1991, "step": 38636 }, { "epoch": 3.89, "learning_rate": 5.439488804824144e-08, "loss": 0.2103, "step": 38637 }, { "epoch": 3.89, "learning_rate": 5.425093447057994e-08, "loss": 0.4298, "step": 38638 }, { "epoch": 3.89, "learning_rate": 5.410717145627353e-08, "loss": 0.3527, "step": 38639 }, { "epoch": 3.89, "learning_rate": 5.3963599006238153e-08, "loss": 0.3968, "step": 38640 }, { "epoch": 3.89, "learning_rate": 5.382021712138641e-08, "loss": 0.3452, "step": 38641 }, { "epoch": 3.89, "learning_rate": 5.3677025802630894e-08, "loss": 0.167, "step": 38642 }, { "epoch": 3.89, "learning_rate": 5.35340250508809e-08, "loss": 0.0976, "step": 38643 }, { "epoch": 3.89, "learning_rate": 5.339121486704901e-08, "loss": 0.3544, "step": 38644 }, { "epoch": 3.89, "learning_rate": 5.3248595252041174e-08, "loss": 0.1785, "step": 38645 }, { "epoch": 3.89, "learning_rate": 5.310616620677e-08, "loss": 0.0197, "step": 38646 }, { "epoch": 3.89, "learning_rate": 5.296392773213809e-08, "loss": 0.0724, "step": 38647 }, { "epoch": 3.89, "learning_rate": 5.28218798290514e-08, "loss": 0.354, "step": 38648 }, { "epoch": 3.89, "learning_rate": 5.268002249841253e-08, "loss": 0.2337, "step": 38649 }, { "epoch": 3.89, "learning_rate": 5.253835574112742e-08, "loss": 0.2344, "step": 38650 }, { "epoch": 3.89, "learning_rate": 5.239687955809202e-08, "loss": 0.1367, "step": 38651 }, { "epoch": 3.89, "learning_rate": 5.2255593950215617e-08, "loss": 0.369, "step": 38652 }, { "epoch": 3.89, "learning_rate": 5.2114498918387485e-08, "loss": 0.0989, "step": 38653 }, { "epoch": 3.89, "learning_rate": 5.197359446351024e-08, "loss": 0.1694, "step": 38654 }, { "epoch": 3.89, "learning_rate": 5.1832880586483164e-08, "loss": 0.6095, "step": 38655 }, { "epoch": 3.89, "learning_rate": 5.1692357288195544e-08, "loss": 0.0907, "step": 38656 }, { "epoch": 3.89, "learning_rate": 5.1552024569546666e-08, "loss": 0.0561, "step": 38657 }, { "epoch": 3.89, "learning_rate": 5.14118824314258e-08, "loss": 0.285, "step": 38658 }, { "epoch": 3.89, "learning_rate": 5.127193087472892e-08, "loss": 0.2102, "step": 38659 }, { "epoch": 3.89, "learning_rate": 5.11321699003453e-08, "loss": 0.1078, "step": 38660 }, { "epoch": 3.89, "learning_rate": 5.099259950916091e-08, "loss": 0.2371, "step": 38661 }, { "epoch": 3.89, "learning_rate": 5.0853219702065024e-08, "loss": 0.2227, "step": 38662 }, { "epoch": 3.89, "learning_rate": 5.071403047995027e-08, "loss": 0.129, "step": 38663 }, { "epoch": 3.89, "learning_rate": 5.057503184369594e-08, "loss": 0.0488, "step": 38664 }, { "epoch": 3.89, "learning_rate": 5.0436223794187995e-08, "loss": 0.0499, "step": 38665 }, { "epoch": 3.89, "learning_rate": 5.02976063323124e-08, "loss": 0.4684, "step": 38666 }, { "epoch": 3.89, "learning_rate": 5.015917945894843e-08, "loss": 0.2888, "step": 38667 }, { "epoch": 3.89, "learning_rate": 5.0020943174978736e-08, "loss": 0.4082, "step": 38668 }, { "epoch": 3.89, "learning_rate": 4.988289748127928e-08, "loss": 0.2174, "step": 38669 }, { "epoch": 3.89, "learning_rate": 4.9745042378736006e-08, "loss": 0.0459, "step": 38670 }, { "epoch": 3.89, "learning_rate": 4.960737786821823e-08, "loss": 0.1139, "step": 38671 }, { "epoch": 3.89, "learning_rate": 4.946990395060524e-08, "loss": 0.4443, "step": 38672 }, { "epoch": 3.89, "learning_rate": 4.9332620626773015e-08, "loss": 0.1779, "step": 38673 }, { "epoch": 3.89, "learning_rate": 4.9195527897597516e-08, "loss": 0.5184, "step": 38674 }, { "epoch": 3.89, "learning_rate": 4.905862576394471e-08, "loss": 0.1592, "step": 38675 }, { "epoch": 3.89, "learning_rate": 4.892191422668724e-08, "loss": 0.3499, "step": 38676 }, { "epoch": 3.89, "learning_rate": 4.8785393286697736e-08, "loss": 0.3147, "step": 38677 }, { "epoch": 3.89, "learning_rate": 4.86490629448455e-08, "loss": 0.1902, "step": 38678 }, { "epoch": 3.89, "learning_rate": 4.851292320199319e-08, "loss": 0.074, "step": 38679 }, { "epoch": 3.89, "learning_rate": 4.837697405901342e-08, "loss": 0.4419, "step": 38680 }, { "epoch": 3.89, "learning_rate": 4.824121551676885e-08, "loss": 0.1046, "step": 38681 }, { "epoch": 3.89, "learning_rate": 4.810564757612212e-08, "loss": 0.4097, "step": 38682 }, { "epoch": 3.89, "learning_rate": 4.7970270237935875e-08, "loss": 0.1926, "step": 38683 }, { "epoch": 3.89, "learning_rate": 4.783508350307608e-08, "loss": 0.1937, "step": 38684 }, { "epoch": 3.89, "learning_rate": 4.7700087372395396e-08, "loss": 0.13, "step": 38685 }, { "epoch": 3.89, "learning_rate": 4.756528184675646e-08, "loss": 0.1832, "step": 38686 }, { "epoch": 3.89, "learning_rate": 4.743066692702192e-08, "loss": 0.2354, "step": 38687 }, { "epoch": 3.89, "learning_rate": 4.729624261404108e-08, "loss": 0.1075, "step": 38688 }, { "epoch": 3.89, "learning_rate": 4.716200890867328e-08, "loss": 0.6052, "step": 38689 }, { "epoch": 3.89, "learning_rate": 4.702796581177116e-08, "loss": 0.13, "step": 38690 }, { "epoch": 3.89, "learning_rate": 4.689411332418736e-08, "loss": 0.099, "step": 38691 }, { "epoch": 3.89, "learning_rate": 4.676045144677454e-08, "loss": 0.2316, "step": 38692 }, { "epoch": 3.89, "learning_rate": 4.662698018038536e-08, "loss": 0.2918, "step": 38693 }, { "epoch": 3.89, "learning_rate": 4.649369952586579e-08, "loss": 0.2627, "step": 38694 }, { "epoch": 3.89, "learning_rate": 4.636060948406851e-08, "loss": 0.4928, "step": 38695 }, { "epoch": 3.89, "learning_rate": 4.6227710055832814e-08, "loss": 0.1452, "step": 38696 }, { "epoch": 3.89, "learning_rate": 4.609500124201138e-08, "loss": 0.1023, "step": 38697 }, { "epoch": 3.89, "learning_rate": 4.596248304344686e-08, "loss": 0.2509, "step": 38698 }, { "epoch": 3.89, "learning_rate": 4.5830155460981906e-08, "loss": 0.2209, "step": 38699 }, { "epoch": 3.89, "learning_rate": 4.569801849545585e-08, "loss": 0.1795, "step": 38700 }, { "epoch": 3.89, "learning_rate": 4.556607214771469e-08, "loss": 0.0969, "step": 38701 }, { "epoch": 3.89, "learning_rate": 4.543431641859774e-08, "loss": 0.2026, "step": 38702 }, { "epoch": 3.89, "learning_rate": 4.530275130893769e-08, "loss": 0.3657, "step": 38703 }, { "epoch": 3.89, "learning_rate": 4.517137681957717e-08, "loss": 0.1724, "step": 38704 }, { "epoch": 3.89, "learning_rate": 4.50401929513522e-08, "loss": 0.266, "step": 38705 }, { "epoch": 3.89, "learning_rate": 4.4909199705092105e-08, "loss": 0.2068, "step": 38706 }, { "epoch": 3.89, "learning_rate": 4.477839708163955e-08, "loss": 0.0715, "step": 38707 }, { "epoch": 3.89, "learning_rate": 4.4647785081817213e-08, "loss": 0.2749, "step": 38708 }, { "epoch": 3.89, "learning_rate": 4.451736370646442e-08, "loss": 0.0769, "step": 38709 }, { "epoch": 3.89, "learning_rate": 4.438713295640384e-08, "loss": 0.2355, "step": 38710 }, { "epoch": 3.89, "learning_rate": 4.425709283246815e-08, "loss": 0.3354, "step": 38711 }, { "epoch": 3.89, "learning_rate": 4.412724333548668e-08, "loss": 0.2397, "step": 38712 }, { "epoch": 3.89, "learning_rate": 4.399758446628543e-08, "loss": 1.3266, "step": 38713 }, { "epoch": 3.89, "learning_rate": 4.3868116225683766e-08, "loss": 0.1541, "step": 38714 }, { "epoch": 3.89, "learning_rate": 4.3738838614511e-08, "loss": 0.0913, "step": 38715 }, { "epoch": 3.89, "learning_rate": 4.3609751633589824e-08, "loss": 0.367, "step": 38716 }, { "epoch": 3.89, "learning_rate": 4.3480855283736245e-08, "loss": 0.0833, "step": 38717 }, { "epoch": 3.89, "learning_rate": 4.3352149565779596e-08, "loss": 0.1436, "step": 38718 }, { "epoch": 3.89, "learning_rate": 4.322363448052924e-08, "loss": 0.3188, "step": 38719 }, { "epoch": 3.89, "learning_rate": 4.309531002881117e-08, "loss": 0.0865, "step": 38720 }, { "epoch": 3.89, "learning_rate": 4.2967176211434756e-08, "loss": 0.4966, "step": 38721 }, { "epoch": 3.9, "learning_rate": 4.283923302922266e-08, "loss": 0.6166, "step": 38722 }, { "epoch": 3.9, "learning_rate": 4.2711480482980905e-08, "loss": 0.3487, "step": 38723 }, { "epoch": 3.9, "learning_rate": 4.258391857353216e-08, "loss": 0.2973, "step": 38724 }, { "epoch": 3.9, "learning_rate": 4.245654730167914e-08, "loss": 0.2388, "step": 38725 }, { "epoch": 3.9, "learning_rate": 4.232936666823784e-08, "loss": 0.4377, "step": 38726 }, { "epoch": 3.9, "learning_rate": 4.2202376674014276e-08, "loss": 0.6977, "step": 38727 }, { "epoch": 3.9, "learning_rate": 4.2075577319821145e-08, "loss": 0.2945, "step": 38728 }, { "epoch": 3.9, "learning_rate": 4.1948968606461135e-08, "loss": 0.0649, "step": 38729 }, { "epoch": 3.9, "learning_rate": 4.182255053474027e-08, "loss": 0.2109, "step": 38730 }, { "epoch": 3.9, "learning_rate": 4.169632310546456e-08, "loss": 0.0616, "step": 38731 }, { "epoch": 3.9, "learning_rate": 4.157028631943671e-08, "loss": 0.4418, "step": 38732 }, { "epoch": 3.9, "learning_rate": 4.144444017745941e-08, "loss": 0.2169, "step": 38733 }, { "epoch": 3.9, "learning_rate": 4.131878468033201e-08, "loss": 0.1625, "step": 38734 }, { "epoch": 3.9, "learning_rate": 4.119331982885721e-08, "loss": 0.1195, "step": 38735 }, { "epoch": 3.9, "learning_rate": 4.106804562382771e-08, "loss": 0.2098, "step": 38736 }, { "epoch": 3.9, "learning_rate": 4.0942962066046196e-08, "loss": 0.584, "step": 38737 }, { "epoch": 3.9, "learning_rate": 4.08180691563087e-08, "loss": 0.2894, "step": 38738 }, { "epoch": 3.9, "learning_rate": 4.0693366895404593e-08, "loss": 0.146, "step": 38739 }, { "epoch": 3.9, "learning_rate": 4.0568855284129904e-08, "loss": 0.2056, "step": 38740 }, { "epoch": 3.9, "learning_rate": 4.044453432328066e-08, "loss": 0.3417, "step": 38741 }, { "epoch": 3.9, "learning_rate": 4.0320404013646226e-08, "loss": 0.3496, "step": 38742 }, { "epoch": 3.9, "learning_rate": 4.019646435601265e-08, "loss": 0.1443, "step": 38743 }, { "epoch": 3.9, "learning_rate": 4.007271535117263e-08, "loss": 0.3862, "step": 38744 }, { "epoch": 3.9, "learning_rate": 3.9949156999912196e-08, "loss": 0.2213, "step": 38745 }, { "epoch": 3.9, "learning_rate": 3.982578930301739e-08, "loss": 0.3776, "step": 38746 }, { "epoch": 3.9, "learning_rate": 3.970261226127425e-08, "loss": 0.0377, "step": 38747 }, { "epoch": 3.9, "learning_rate": 3.957962587546882e-08, "loss": 0.0883, "step": 38748 }, { "epoch": 3.9, "learning_rate": 3.945683014637713e-08, "loss": 0.0444, "step": 38749 }, { "epoch": 3.9, "learning_rate": 3.9334225074788565e-08, "loss": 0.6376, "step": 38750 }, { "epoch": 3.9, "learning_rate": 3.9211810661475835e-08, "loss": 0.0709, "step": 38751 }, { "epoch": 3.9, "learning_rate": 3.908958690722164e-08, "loss": 0.2341, "step": 38752 }, { "epoch": 3.9, "learning_rate": 3.8967553812805367e-08, "loss": 0.1232, "step": 38753 }, { "epoch": 3.9, "learning_rate": 3.884571137900306e-08, "loss": 0.1552, "step": 38754 }, { "epoch": 3.9, "learning_rate": 3.8724059606584095e-08, "loss": 0.8982, "step": 38755 }, { "epoch": 3.9, "learning_rate": 3.860259849633119e-08, "loss": 0.1121, "step": 38756 }, { "epoch": 3.9, "learning_rate": 3.848132804901372e-08, "loss": 0.408, "step": 38757 }, { "epoch": 3.9, "learning_rate": 3.836024826540108e-08, "loss": 0.108, "step": 38758 }, { "epoch": 3.9, "learning_rate": 3.8239359146265975e-08, "loss": 0.5325, "step": 38759 }, { "epoch": 3.9, "learning_rate": 3.81186606923778e-08, "loss": 0.0796, "step": 38760 }, { "epoch": 3.9, "learning_rate": 3.799815290450592e-08, "loss": 0.1261, "step": 38761 }, { "epoch": 3.9, "learning_rate": 3.787783578341641e-08, "loss": 0.2068, "step": 38762 }, { "epoch": 3.9, "learning_rate": 3.7757709329871995e-08, "loss": 0.1079, "step": 38763 }, { "epoch": 3.9, "learning_rate": 3.763777354463871e-08, "loss": 0.2264, "step": 38764 }, { "epoch": 3.9, "learning_rate": 3.751802842847929e-08, "loss": 0.3206, "step": 38765 }, { "epoch": 3.9, "learning_rate": 3.7398473982159786e-08, "loss": 0.066, "step": 38766 }, { "epoch": 3.9, "learning_rate": 3.7279110206436264e-08, "loss": 0.4736, "step": 38767 }, { "epoch": 3.9, "learning_rate": 3.715993710207144e-08, "loss": 0.0294, "step": 38768 }, { "epoch": 3.9, "learning_rate": 3.704095466982138e-08, "loss": 0.2152, "step": 38769 }, { "epoch": 3.9, "learning_rate": 3.6922162910445477e-08, "loss": 0.4161, "step": 38770 }, { "epoch": 3.9, "learning_rate": 3.680356182469979e-08, "loss": 0.1804, "step": 38771 }, { "epoch": 3.9, "learning_rate": 3.668515141333373e-08, "loss": 0.2317, "step": 38772 }, { "epoch": 3.9, "learning_rate": 3.656693167710668e-08, "loss": 0.5719, "step": 38773 }, { "epoch": 3.9, "learning_rate": 3.644890261676803e-08, "loss": 0.1082, "step": 38774 }, { "epoch": 3.9, "learning_rate": 3.6331064233070535e-08, "loss": 0.2681, "step": 38775 }, { "epoch": 3.9, "learning_rate": 3.621341652676358e-08, "loss": 0.131, "step": 38776 }, { "epoch": 3.9, "learning_rate": 3.6095959498596565e-08, "loss": 0.2369, "step": 38777 }, { "epoch": 3.9, "learning_rate": 3.5978693149312234e-08, "loss": 0.1894, "step": 38778 }, { "epoch": 3.9, "learning_rate": 3.5861617479663324e-08, "loss": 0.6544, "step": 38779 }, { "epoch": 3.9, "learning_rate": 3.5744732490392564e-08, "loss": 0.0682, "step": 38780 }, { "epoch": 3.9, "learning_rate": 3.56280381822427e-08, "loss": 0.0307, "step": 38781 }, { "epoch": 3.9, "learning_rate": 3.5511534555959794e-08, "loss": 0.0302, "step": 38782 }, { "epoch": 3.9, "learning_rate": 3.5395221612279924e-08, "loss": 0.1179, "step": 38783 }, { "epoch": 3.9, "learning_rate": 3.527909935194251e-08, "loss": 0.0867, "step": 38784 }, { "epoch": 3.9, "learning_rate": 3.5163167775693615e-08, "loss": 0.1596, "step": 38785 }, { "epoch": 3.9, "learning_rate": 3.504742688426599e-08, "loss": 0.1191, "step": 38786 }, { "epoch": 3.9, "learning_rate": 3.4931876678399035e-08, "loss": 0.1599, "step": 38787 }, { "epoch": 3.9, "learning_rate": 3.48165171588255e-08, "loss": 0.0312, "step": 38788 }, { "epoch": 3.9, "learning_rate": 3.4701348326278136e-08, "loss": 0.2935, "step": 38789 }, { "epoch": 3.9, "learning_rate": 3.458637018149302e-08, "loss": 0.0689, "step": 38790 }, { "epoch": 3.9, "learning_rate": 3.44715827252029e-08, "loss": 0.0498, "step": 38791 }, { "epoch": 3.9, "learning_rate": 3.435698595813386e-08, "loss": 0.0289, "step": 38792 }, { "epoch": 3.9, "learning_rate": 3.424257988101864e-08, "loss": 0.2768, "step": 38793 }, { "epoch": 3.9, "learning_rate": 3.412836449458334e-08, "loss": 0.291, "step": 38794 }, { "epoch": 3.9, "learning_rate": 3.401433979955404e-08, "loss": 0.6936, "step": 38795 }, { "epoch": 3.9, "learning_rate": 3.3900505796660154e-08, "loss": 0.0906, "step": 38796 }, { "epoch": 3.9, "learning_rate": 3.378686248662444e-08, "loss": 0.2964, "step": 38797 }, { "epoch": 3.9, "learning_rate": 3.367340987016965e-08, "loss": 0.1952, "step": 38798 }, { "epoch": 3.9, "learning_rate": 3.356014794801521e-08, "loss": 0.0825, "step": 38799 }, { "epoch": 3.9, "learning_rate": 3.344707672088387e-08, "loss": 0.0665, "step": 38800 }, { "epoch": 3.9, "learning_rate": 3.33341961894984e-08, "loss": 0.057, "step": 38801 }, { "epoch": 3.9, "learning_rate": 3.322150635457155e-08, "loss": 0.0746, "step": 38802 }, { "epoch": 3.9, "learning_rate": 3.3109007216822754e-08, "loss": 0.2877, "step": 38803 }, { "epoch": 3.9, "learning_rate": 3.299669877697142e-08, "loss": 0.3612, "step": 38804 }, { "epoch": 3.9, "learning_rate": 3.2884581035727004e-08, "loss": 0.1782, "step": 38805 }, { "epoch": 3.9, "learning_rate": 3.2772653993805576e-08, "loss": 0.2406, "step": 38806 }, { "epoch": 3.9, "learning_rate": 3.2660917651916586e-08, "loss": 0.1599, "step": 38807 }, { "epoch": 3.9, "learning_rate": 3.254937201077613e-08, "loss": 0.538, "step": 38808 }, { "epoch": 3.9, "learning_rate": 3.243801707109029e-08, "loss": 0.0571, "step": 38809 }, { "epoch": 3.9, "learning_rate": 3.232685283356851e-08, "loss": 0.0482, "step": 38810 }, { "epoch": 3.9, "learning_rate": 3.2215879298916895e-08, "loss": 0.3639, "step": 38811 }, { "epoch": 3.9, "learning_rate": 3.210509646784488e-08, "loss": 0.2905, "step": 38812 }, { "epoch": 3.9, "learning_rate": 3.199450434105522e-08, "loss": 0.0674, "step": 38813 }, { "epoch": 3.9, "learning_rate": 3.188410291925403e-08, "loss": 0.1478, "step": 38814 }, { "epoch": 3.9, "learning_rate": 3.177389220314075e-08, "loss": 0.1992, "step": 38815 }, { "epoch": 3.9, "learning_rate": 3.166387219341815e-08, "loss": 0.6129, "step": 38816 }, { "epoch": 3.9, "learning_rate": 3.155404289078567e-08, "loss": 0.9251, "step": 38817 }, { "epoch": 3.9, "learning_rate": 3.1444404295939425e-08, "loss": 0.2484, "step": 38818 }, { "epoch": 3.9, "learning_rate": 3.13349564095855e-08, "loss": 0.0961, "step": 38819 }, { "epoch": 3.9, "learning_rate": 3.1225699232413365e-08, "loss": 0.1033, "step": 38820 }, { "epoch": 3.91, "learning_rate": 3.111663276511911e-08, "loss": 0.4375, "step": 38821 }, { "epoch": 3.91, "learning_rate": 3.100775700839886e-08, "loss": 0.0931, "step": 38822 }, { "epoch": 3.91, "learning_rate": 3.089907196294539e-08, "loss": 0.0474, "step": 38823 }, { "epoch": 3.91, "learning_rate": 3.0790577629448145e-08, "loss": 0.2475, "step": 38824 }, { "epoch": 3.91, "learning_rate": 3.068227400859991e-08, "loss": 0.1833, "step": 38825 }, { "epoch": 3.91, "learning_rate": 3.057416110108679e-08, "loss": 0.0964, "step": 38826 }, { "epoch": 3.91, "learning_rate": 3.046623890759825e-08, "loss": 0.0579, "step": 38827 }, { "epoch": 3.91, "learning_rate": 3.0358507428823736e-08, "loss": 0.3711, "step": 38828 }, { "epoch": 3.91, "learning_rate": 3.0250966665446024e-08, "loss": 0.1732, "step": 38829 }, { "epoch": 3.91, "learning_rate": 3.0143616618151236e-08, "loss": 0.0374, "step": 38830 }, { "epoch": 3.91, "learning_rate": 3.0036457287618834e-08, "loss": 0.3424, "step": 38831 }, { "epoch": 3.91, "learning_rate": 2.992948867453493e-08, "loss": 0.4234, "step": 38832 }, { "epoch": 3.91, "learning_rate": 2.982271077957899e-08, "loss": 0.1451, "step": 38833 }, { "epoch": 3.91, "learning_rate": 2.9716123603430456e-08, "loss": 0.1082, "step": 38834 }, { "epoch": 3.91, "learning_rate": 2.9609727146765462e-08, "loss": 0.2573, "step": 38835 }, { "epoch": 3.91, "learning_rate": 2.950352141026347e-08, "loss": 0.492, "step": 38836 }, { "epoch": 3.91, "learning_rate": 2.9397506394600593e-08, "loss": 0.1029, "step": 38837 }, { "epoch": 3.91, "learning_rate": 2.9291682100449634e-08, "loss": 0.2343, "step": 38838 }, { "epoch": 3.91, "learning_rate": 2.918604852848339e-08, "loss": 0.0475, "step": 38839 }, { "epoch": 3.91, "learning_rate": 2.908060567937798e-08, "loss": 0.1118, "step": 38840 }, { "epoch": 3.91, "learning_rate": 2.8975353553802876e-08, "loss": 0.1167, "step": 38841 }, { "epoch": 3.91, "learning_rate": 2.8870292152424204e-08, "loss": 0.1063, "step": 38842 }, { "epoch": 3.91, "learning_rate": 2.8765421475914767e-08, "loss": 0.057, "step": 38843 }, { "epoch": 3.91, "learning_rate": 2.8660741524940693e-08, "loss": 0.141, "step": 38844 }, { "epoch": 3.91, "learning_rate": 2.8556252300168117e-08, "loss": 0.3166, "step": 38845 }, { "epoch": 3.91, "learning_rate": 2.8451953802263177e-08, "loss": 0.2269, "step": 38846 }, { "epoch": 3.91, "learning_rate": 2.834784603188534e-08, "loss": 0.1914, "step": 38847 }, { "epoch": 3.91, "learning_rate": 2.824392898970074e-08, "loss": 0.4786, "step": 38848 }, { "epoch": 3.91, "learning_rate": 2.8140202676372184e-08, "loss": 0.2158, "step": 38849 }, { "epoch": 3.91, "learning_rate": 2.803666709255581e-08, "loss": 0.1636, "step": 38850 }, { "epoch": 3.91, "learning_rate": 2.7933322238911095e-08, "loss": 0.0317, "step": 38851 }, { "epoch": 3.91, "learning_rate": 2.783016811610084e-08, "loss": 0.1243, "step": 38852 }, { "epoch": 3.91, "learning_rate": 2.7727204724771192e-08, "loss": 0.0582, "step": 38853 }, { "epoch": 3.91, "learning_rate": 2.762443206558829e-08, "loss": 0.1387, "step": 38854 }, { "epoch": 3.91, "learning_rate": 2.752185013920161e-08, "loss": 0.2564, "step": 38855 }, { "epoch": 3.91, "learning_rate": 2.7419458946260633e-08, "loss": 0.2599, "step": 38856 }, { "epoch": 3.91, "learning_rate": 2.7317258487421505e-08, "loss": 0.5793, "step": 38857 }, { "epoch": 3.91, "learning_rate": 2.72152487633337e-08, "loss": 0.7228, "step": 38858 }, { "epoch": 3.91, "learning_rate": 2.711342977464337e-08, "loss": 0.0737, "step": 38859 }, { "epoch": 3.91, "learning_rate": 2.7011801522003334e-08, "loss": 0.1392, "step": 38860 }, { "epoch": 3.91, "learning_rate": 2.6910364006059728e-08, "loss": 0.1804, "step": 38861 }, { "epoch": 3.91, "learning_rate": 2.680911722745205e-08, "loss": 0.1587, "step": 38862 }, { "epoch": 3.91, "learning_rate": 2.670806118683311e-08, "loss": 0.2183, "step": 38863 }, { "epoch": 3.91, "learning_rate": 2.6607195884835734e-08, "loss": 0.1481, "step": 38864 }, { "epoch": 3.91, "learning_rate": 2.650652132211273e-08, "loss": 0.2166, "step": 38865 }, { "epoch": 3.91, "learning_rate": 2.640603749930026e-08, "loss": 0.3685, "step": 38866 }, { "epoch": 3.91, "learning_rate": 2.6305744417034483e-08, "loss": 0.155, "step": 38867 }, { "epoch": 3.91, "learning_rate": 2.6205642075954882e-08, "loss": 0.2284, "step": 38868 }, { "epoch": 3.91, "learning_rate": 2.6105730476704282e-08, "loss": 0.1825, "step": 38869 }, { "epoch": 3.91, "learning_rate": 2.6006009619912174e-08, "loss": 0.0553, "step": 38870 }, { "epoch": 3.91, "learning_rate": 2.590647950621472e-08, "loss": 0.2802, "step": 38871 }, { "epoch": 3.91, "learning_rate": 2.5807140136244746e-08, "loss": 0.0615, "step": 38872 }, { "epoch": 3.91, "learning_rate": 2.5707991510635078e-08, "loss": 0.0597, "step": 38873 }, { "epoch": 3.91, "learning_rate": 2.5609033630018542e-08, "loss": 0.0536, "step": 38874 }, { "epoch": 3.91, "learning_rate": 2.5510266495024636e-08, "loss": 0.0963, "step": 38875 }, { "epoch": 3.91, "learning_rate": 2.5411690106279527e-08, "loss": 0.204, "step": 38876 }, { "epoch": 3.91, "learning_rate": 2.5313304464409382e-08, "loss": 0.5009, "step": 38877 }, { "epoch": 3.91, "learning_rate": 2.5215109570043693e-08, "loss": 0.1386, "step": 38878 }, { "epoch": 3.91, "learning_rate": 2.5117105423808627e-08, "loss": 0.3469, "step": 38879 }, { "epoch": 3.91, "learning_rate": 2.5019292026323694e-08, "loss": 0.0125, "step": 38880 }, { "epoch": 3.91, "learning_rate": 2.492166937821172e-08, "loss": 0.0265, "step": 38881 }, { "epoch": 3.91, "learning_rate": 2.4824237480095545e-08, "loss": 0.0545, "step": 38882 }, { "epoch": 3.91, "learning_rate": 2.4726996332594676e-08, "loss": 0.0137, "step": 38883 }, { "epoch": 3.91, "learning_rate": 2.462994593632861e-08, "loss": 0.403, "step": 38884 }, { "epoch": 3.91, "learning_rate": 2.453308629191353e-08, "loss": 0.0314, "step": 38885 }, { "epoch": 3.91, "learning_rate": 2.4436417399965604e-08, "loss": 0.3007, "step": 38886 }, { "epoch": 3.91, "learning_rate": 2.4339939261104337e-08, "loss": 0.2211, "step": 38887 }, { "epoch": 3.91, "learning_rate": 2.424365187593591e-08, "loss": 0.5121, "step": 38888 }, { "epoch": 3.91, "learning_rate": 2.4147555245083163e-08, "loss": 0.1439, "step": 38889 }, { "epoch": 3.91, "learning_rate": 2.4051649369148942e-08, "loss": 0.3212, "step": 38890 }, { "epoch": 3.91, "learning_rate": 2.3955934248746093e-08, "loss": 0.3467, "step": 38891 }, { "epoch": 3.91, "learning_rate": 2.3860409884484123e-08, "loss": 0.0788, "step": 38892 }, { "epoch": 3.91, "learning_rate": 2.3765076276972553e-08, "loss": 0.261, "step": 38893 }, { "epoch": 3.91, "learning_rate": 2.366993342681423e-08, "loss": 0.2198, "step": 38894 }, { "epoch": 3.91, "learning_rate": 2.357498133461866e-08, "loss": 0.1566, "step": 38895 }, { "epoch": 3.91, "learning_rate": 2.3480220000988706e-08, "loss": 0.4214, "step": 38896 }, { "epoch": 3.91, "learning_rate": 2.338564942652388e-08, "loss": 0.3368, "step": 38897 }, { "epoch": 3.91, "learning_rate": 2.32912696118337e-08, "loss": 0.0963, "step": 38898 }, { "epoch": 3.91, "learning_rate": 2.319708055751102e-08, "loss": 0.2671, "step": 38899 }, { "epoch": 3.91, "learning_rate": 2.3103082264158694e-08, "loss": 0.2784, "step": 38900 }, { "epoch": 3.91, "learning_rate": 2.300927473237624e-08, "loss": 0.4322, "step": 38901 }, { "epoch": 3.91, "learning_rate": 2.291565796275652e-08, "loss": 0.1771, "step": 38902 }, { "epoch": 3.91, "learning_rate": 2.2822231955899053e-08, "loss": 0.2227, "step": 38903 }, { "epoch": 3.91, "learning_rate": 2.27289967123967e-08, "loss": 0.0562, "step": 38904 }, { "epoch": 3.91, "learning_rate": 2.2635952232842316e-08, "loss": 0.1382, "step": 38905 }, { "epoch": 3.91, "learning_rate": 2.2543098517832094e-08, "loss": 0.2812, "step": 38906 }, { "epoch": 3.91, "learning_rate": 2.24504355679489e-08, "loss": 0.152, "step": 38907 }, { "epoch": 3.91, "learning_rate": 2.2357963383788927e-08, "loss": 0.08, "step": 38908 }, { "epoch": 3.91, "learning_rate": 2.22656819659417e-08, "loss": 0.0877, "step": 38909 }, { "epoch": 3.91, "learning_rate": 2.2173591314986753e-08, "loss": 0.0352, "step": 38910 }, { "epoch": 3.91, "learning_rate": 2.2081691431516948e-08, "loss": 0.4446, "step": 38911 }, { "epoch": 3.91, "learning_rate": 2.1989982316115152e-08, "loss": 0.1627, "step": 38912 }, { "epoch": 3.91, "learning_rate": 2.1898463969367567e-08, "loss": 0.1637, "step": 38913 }, { "epoch": 3.91, "learning_rate": 2.180713639185039e-08, "loss": 0.2801, "step": 38914 }, { "epoch": 3.91, "learning_rate": 2.1715999584149827e-08, "loss": 0.1585, "step": 38915 }, { "epoch": 3.91, "learning_rate": 2.162505354684208e-08, "loss": 0.2101, "step": 38916 }, { "epoch": 3.91, "learning_rate": 2.153429828051001e-08, "loss": 0.2509, "step": 38917 }, { "epoch": 3.91, "learning_rate": 2.1443733785729835e-08, "loss": 0.3066, "step": 38918 }, { "epoch": 3.92, "learning_rate": 2.1353360063077754e-08, "loss": 0.0176, "step": 38919 }, { "epoch": 3.92, "learning_rate": 2.126317711312664e-08, "loss": 0.1922, "step": 38920 }, { "epoch": 3.92, "learning_rate": 2.1173184936452706e-08, "loss": 0.1279, "step": 38921 }, { "epoch": 3.92, "learning_rate": 2.108338353362882e-08, "loss": 0.1225, "step": 38922 }, { "epoch": 3.92, "learning_rate": 2.0993772905224527e-08, "loss": 0.3181, "step": 38923 }, { "epoch": 3.92, "learning_rate": 2.0904353051812707e-08, "loss": 0.1753, "step": 38924 }, { "epoch": 3.92, "learning_rate": 2.081512397395957e-08, "loss": 0.1256, "step": 38925 }, { "epoch": 3.92, "learning_rate": 2.0726085672234663e-08, "loss": 0.0672, "step": 38926 }, { "epoch": 3.92, "learning_rate": 2.0637238147204196e-08, "loss": 0.1001, "step": 38927 }, { "epoch": 3.92, "learning_rate": 2.0548581399434384e-08, "loss": 0.2961, "step": 38928 }, { "epoch": 3.92, "learning_rate": 2.0460115429488112e-08, "loss": 0.5053, "step": 38929 }, { "epoch": 3.92, "learning_rate": 2.0371840237928263e-08, "loss": 0.2038, "step": 38930 }, { "epoch": 3.92, "learning_rate": 2.028375582531772e-08, "loss": 0.4021, "step": 38931 }, { "epoch": 3.92, "learning_rate": 2.019586219221603e-08, "loss": 0.3434, "step": 38932 }, { "epoch": 3.92, "learning_rate": 2.010815933918275e-08, "loss": 0.273, "step": 38933 }, { "epoch": 3.92, "learning_rate": 2.0020647266777435e-08, "loss": 0.1761, "step": 38934 }, { "epoch": 3.92, "learning_rate": 1.9933325975556305e-08, "loss": 0.4323, "step": 38935 }, { "epoch": 3.92, "learning_rate": 1.9846195466072246e-08, "loss": 0.0159, "step": 38936 }, { "epoch": 3.92, "learning_rate": 1.975925573888482e-08, "loss": 0.2998, "step": 38937 }, { "epoch": 3.92, "learning_rate": 1.9672506794543578e-08, "loss": 0.1627, "step": 38938 }, { "epoch": 3.92, "learning_rate": 1.9585948633598082e-08, "loss": 0.6403, "step": 38939 }, { "epoch": 3.92, "learning_rate": 1.949958125660456e-08, "loss": 0.1477, "step": 38940 }, { "epoch": 3.92, "learning_rate": 1.9413404664112565e-08, "loss": 0.2167, "step": 38941 }, { "epoch": 3.92, "learning_rate": 1.9327418856668333e-08, "loss": 0.4829, "step": 38942 }, { "epoch": 3.92, "learning_rate": 1.924162383481476e-08, "loss": 0.1414, "step": 38943 }, { "epoch": 3.92, "learning_rate": 1.9156019599108067e-08, "loss": 0.2979, "step": 38944 }, { "epoch": 3.92, "learning_rate": 1.907060615008449e-08, "loss": 0.4319, "step": 38945 }, { "epoch": 3.92, "learning_rate": 1.8985383488286933e-08, "loss": 0.117, "step": 38946 }, { "epoch": 3.92, "learning_rate": 1.8900351614264955e-08, "loss": 0.1749, "step": 38947 }, { "epoch": 3.92, "learning_rate": 1.881551052855479e-08, "loss": 0.0443, "step": 38948 }, { "epoch": 3.92, "learning_rate": 1.8730860231699344e-08, "loss": 0.0482, "step": 38949 }, { "epoch": 3.92, "learning_rate": 1.8646400724231516e-08, "loss": 0.053, "step": 38950 }, { "epoch": 3.92, "learning_rate": 1.856213200669421e-08, "loss": 0.4498, "step": 38951 }, { "epoch": 3.92, "learning_rate": 1.8478054079623664e-08, "loss": 0.2309, "step": 38952 }, { "epoch": 3.92, "learning_rate": 1.8394166943549452e-08, "loss": 0.244, "step": 38953 }, { "epoch": 3.92, "learning_rate": 1.8310470599011143e-08, "loss": 0.1825, "step": 38954 }, { "epoch": 3.92, "learning_rate": 1.8226965046541645e-08, "loss": 0.2127, "step": 38955 }, { "epoch": 3.92, "learning_rate": 1.81436502866672e-08, "loss": 0.04, "step": 38956 }, { "epoch": 3.92, "learning_rate": 1.8060526319920724e-08, "loss": 0.0634, "step": 38957 }, { "epoch": 3.92, "learning_rate": 1.7977593146835113e-08, "loss": 0.9013, "step": 38958 }, { "epoch": 3.92, "learning_rate": 1.789485076792996e-08, "loss": 0.064, "step": 38959 }, { "epoch": 3.92, "learning_rate": 1.781229918373817e-08, "loss": 0.367, "step": 38960 }, { "epoch": 3.92, "learning_rate": 1.7729938394782652e-08, "loss": 0.3124, "step": 38961 }, { "epoch": 3.92, "learning_rate": 1.764776840158966e-08, "loss": 0.0269, "step": 38962 }, { "epoch": 3.92, "learning_rate": 1.7565789204675442e-08, "loss": 0.084, "step": 38963 }, { "epoch": 3.92, "learning_rate": 1.748400080457291e-08, "loss": 0.1976, "step": 38964 }, { "epoch": 3.92, "learning_rate": 1.740240320179165e-08, "loss": 0.1674, "step": 38965 }, { "epoch": 3.92, "learning_rate": 1.7320996396857914e-08, "loss": 0.285, "step": 38966 }, { "epoch": 3.92, "learning_rate": 1.723978039028795e-08, "loss": 0.1459, "step": 38967 }, { "epoch": 3.92, "learning_rate": 1.7158755182594687e-08, "loss": 0.0437, "step": 38968 }, { "epoch": 3.92, "learning_rate": 1.7077920774301037e-08, "loss": 0.4854, "step": 38969 }, { "epoch": 3.92, "learning_rate": 1.6997277165913262e-08, "loss": 0.0273, "step": 38970 }, { "epoch": 3.92, "learning_rate": 1.6916824357950944e-08, "loss": 0.7127, "step": 38971 }, { "epoch": 3.92, "learning_rate": 1.683656235092368e-08, "loss": 0.1783, "step": 38972 }, { "epoch": 3.92, "learning_rate": 1.6756491145344387e-08, "loss": 0.125, "step": 38973 }, { "epoch": 3.92, "learning_rate": 1.6676610741716003e-08, "loss": 0.0315, "step": 38974 }, { "epoch": 3.92, "learning_rate": 1.659692114055478e-08, "loss": 0.0929, "step": 38975 }, { "epoch": 3.92, "learning_rate": 1.651742234236364e-08, "loss": 0.2195, "step": 38976 }, { "epoch": 3.92, "learning_rate": 1.6438114347652187e-08, "loss": 0.2879, "step": 38977 }, { "epoch": 3.92, "learning_rate": 1.635899715691669e-08, "loss": 0.2333, "step": 38978 }, { "epoch": 3.92, "learning_rate": 1.6280070770670065e-08, "loss": 0.1088, "step": 38979 }, { "epoch": 3.92, "learning_rate": 1.6201335189408582e-08, "loss": 0.058, "step": 38980 }, { "epoch": 3.92, "learning_rate": 1.6122790413638512e-08, "loss": 0.1097, "step": 38981 }, { "epoch": 3.92, "learning_rate": 1.6044436443852783e-08, "loss": 0.1526, "step": 38982 }, { "epoch": 3.92, "learning_rate": 1.5966273280557664e-08, "loss": 0.0713, "step": 38983 }, { "epoch": 3.92, "learning_rate": 1.588830092424276e-08, "loss": 0.0594, "step": 38984 }, { "epoch": 3.92, "learning_rate": 1.5810519375411005e-08, "loss": 0.1775, "step": 38985 }, { "epoch": 3.92, "learning_rate": 1.5732928634552002e-08, "loss": 0.0967, "step": 38986 }, { "epoch": 3.92, "learning_rate": 1.5655528702165355e-08, "loss": 0.3705, "step": 38987 }, { "epoch": 3.92, "learning_rate": 1.5578319578737344e-08, "loss": 0.1521, "step": 38988 }, { "epoch": 3.92, "learning_rate": 1.5501301264764233e-08, "loss": 0.1847, "step": 38989 }, { "epoch": 3.92, "learning_rate": 1.5424473760732305e-08, "loss": 0.1641, "step": 38990 }, { "epoch": 3.92, "learning_rate": 1.5347837067134495e-08, "loss": 0.3609, "step": 38991 }, { "epoch": 3.92, "learning_rate": 1.5271391184453753e-08, "loss": 0.1342, "step": 38992 }, { "epoch": 3.92, "learning_rate": 1.5195136113179687e-08, "loss": 0.0556, "step": 38993 }, { "epoch": 3.92, "learning_rate": 1.511907185379857e-08, "loss": 0.0169, "step": 38994 }, { "epoch": 3.92, "learning_rate": 1.5043198406790026e-08, "loss": 0.0324, "step": 38995 }, { "epoch": 3.92, "learning_rate": 1.4967515772643658e-08, "loss": 0.13, "step": 38996 }, { "epoch": 3.92, "learning_rate": 1.4892023951835754e-08, "loss": 0.5921, "step": 38997 }, { "epoch": 3.92, "learning_rate": 1.4816722944845928e-08, "loss": 0.1358, "step": 38998 }, { "epoch": 3.92, "learning_rate": 1.474161275215713e-08, "loss": 0.2518, "step": 38999 }, { "epoch": 3.92, "learning_rate": 1.4666693374245644e-08, "loss": 0.4468, "step": 39000 }, { "epoch": 3.92, "learning_rate": 1.4591964811587754e-08, "loss": 0.1829, "step": 39001 }, { "epoch": 3.92, "learning_rate": 1.451742706465975e-08, "loss": 0.1014, "step": 39002 }, { "epoch": 3.92, "learning_rate": 1.4443080133937913e-08, "loss": 0.0653, "step": 39003 }, { "epoch": 3.92, "learning_rate": 1.436892401989187e-08, "loss": 0.1344, "step": 39004 }, { "epoch": 3.92, "learning_rate": 1.4294958722994577e-08, "loss": 0.1304, "step": 39005 }, { "epoch": 3.92, "learning_rate": 1.4221184243715656e-08, "loss": 0.1241, "step": 39006 }, { "epoch": 3.92, "learning_rate": 1.4147600582528065e-08, "loss": 0.4116, "step": 39007 }, { "epoch": 3.92, "learning_rate": 1.4074207739898093e-08, "loss": 0.0686, "step": 39008 }, { "epoch": 3.92, "learning_rate": 1.4001005716288706e-08, "loss": 0.0372, "step": 39009 }, { "epoch": 3.92, "learning_rate": 1.3927994512172858e-08, "loss": 0.118, "step": 39010 }, { "epoch": 3.92, "learning_rate": 1.3855174128013515e-08, "loss": 0.1606, "step": 39011 }, { "epoch": 3.92, "learning_rate": 1.3782544564270305e-08, "loss": 0.0537, "step": 39012 }, { "epoch": 3.92, "learning_rate": 1.3710105821406194e-08, "loss": 0.1107, "step": 39013 }, { "epoch": 3.92, "learning_rate": 1.3637857899887474e-08, "loss": 0.3627, "step": 39014 }, { "epoch": 3.92, "learning_rate": 1.3565800800167117e-08, "loss": 0.199, "step": 39015 }, { "epoch": 3.92, "learning_rate": 1.3493934522708085e-08, "loss": 0.1156, "step": 39016 }, { "epoch": 3.92, "learning_rate": 1.342225906796668e-08, "loss": 0.0441, "step": 39017 }, { "epoch": 3.93, "learning_rate": 1.3350774436395873e-08, "loss": 0.3373, "step": 39018 }, { "epoch": 3.93, "learning_rate": 1.3279480628455298e-08, "loss": 0.1348, "step": 39019 }, { "epoch": 3.93, "learning_rate": 1.3208377644594593e-08, "loss": 0.305, "step": 39020 }, { "epoch": 3.93, "learning_rate": 1.3137465485270062e-08, "loss": 0.1279, "step": 39021 }, { "epoch": 3.93, "learning_rate": 1.3066744150931342e-08, "loss": 0.1309, "step": 39022 }, { "epoch": 3.93, "learning_rate": 1.2996213642028077e-08, "loss": 0.1041, "step": 39023 }, { "epoch": 3.93, "learning_rate": 1.2925873959009904e-08, "loss": 0.0851, "step": 39024 }, { "epoch": 3.93, "learning_rate": 1.2855725102323135e-08, "loss": 0.3009, "step": 39025 }, { "epoch": 3.93, "learning_rate": 1.2785767072414079e-08, "loss": 0.432, "step": 39026 }, { "epoch": 3.93, "learning_rate": 1.2715999869729045e-08, "loss": 0.2091, "step": 39027 }, { "epoch": 3.93, "learning_rate": 1.2646423494711013e-08, "loss": 0.243, "step": 39028 }, { "epoch": 3.93, "learning_rate": 1.257703794780296e-08, "loss": 0.2731, "step": 39029 }, { "epoch": 3.93, "learning_rate": 1.2507843229447868e-08, "loss": 0.1099, "step": 39030 }, { "epoch": 3.93, "learning_rate": 1.2438839340085384e-08, "loss": 0.5632, "step": 39031 }, { "epoch": 3.93, "learning_rate": 1.2370026280155156e-08, "loss": 0.2368, "step": 39032 }, { "epoch": 3.93, "learning_rate": 1.23014040500935e-08, "loss": 0.2276, "step": 39033 }, { "epoch": 3.93, "learning_rate": 1.2232972650336738e-08, "loss": 0.1264, "step": 39034 }, { "epoch": 3.93, "learning_rate": 1.2164732081321184e-08, "loss": 0.1181, "step": 39035 }, { "epoch": 3.93, "learning_rate": 1.2096682343483156e-08, "loss": 0.2419, "step": 39036 }, { "epoch": 3.93, "learning_rate": 1.2028823437252312e-08, "loss": 0.1188, "step": 39037 }, { "epoch": 3.93, "learning_rate": 1.1961155363061638e-08, "loss": 0.0503, "step": 39038 }, { "epoch": 3.93, "learning_rate": 1.189367812134412e-08, "loss": 0.3024, "step": 39039 }, { "epoch": 3.93, "learning_rate": 1.1826391712526086e-08, "loss": 0.3064, "step": 39040 }, { "epoch": 3.93, "learning_rate": 1.175929613703719e-08, "loss": 0.0518, "step": 39041 }, { "epoch": 3.93, "learning_rate": 1.1692391395303759e-08, "loss": 0.0878, "step": 39042 }, { "epoch": 3.93, "learning_rate": 1.1625677487752118e-08, "loss": 0.2013, "step": 39043 }, { "epoch": 3.93, "learning_rate": 1.1559154414805261e-08, "loss": 0.036, "step": 39044 }, { "epoch": 3.93, "learning_rate": 1.1492822176889517e-08, "loss": 0.196, "step": 39045 }, { "epoch": 3.93, "learning_rate": 1.1426680774421216e-08, "loss": 0.1341, "step": 39046 }, { "epoch": 3.93, "learning_rate": 1.1360730207830017e-08, "loss": 0.0714, "step": 39047 }, { "epoch": 3.93, "learning_rate": 1.129497047752892e-08, "loss": 0.2229, "step": 39048 }, { "epoch": 3.93, "learning_rate": 1.1229401583937593e-08, "loss": 0.2266, "step": 39049 }, { "epoch": 3.93, "learning_rate": 1.1164023527475698e-08, "loss": 0.3535, "step": 39050 }, { "epoch": 3.93, "learning_rate": 1.1098836308559567e-08, "loss": 0.1586, "step": 39051 }, { "epoch": 3.93, "learning_rate": 1.1033839927598876e-08, "loss": 0.3995, "step": 39052 }, { "epoch": 3.93, "learning_rate": 1.0969034385009957e-08, "loss": 0.5227, "step": 39053 }, { "epoch": 3.93, "learning_rate": 1.0904419681205812e-08, "loss": 0.3952, "step": 39054 }, { "epoch": 3.93, "learning_rate": 1.0839995816599445e-08, "loss": 0.0505, "step": 39055 }, { "epoch": 3.93, "learning_rate": 1.0775762791597199e-08, "loss": 0.2621, "step": 39056 }, { "epoch": 3.93, "learning_rate": 1.0711720606612074e-08, "loss": 0.3448, "step": 39057 }, { "epoch": 3.93, "learning_rate": 1.0647869262047083e-08, "loss": 0.094, "step": 39058 }, { "epoch": 3.93, "learning_rate": 1.0584208758308566e-08, "loss": 0.3446, "step": 39059 }, { "epoch": 3.93, "learning_rate": 1.0520739095806198e-08, "loss": 0.6692, "step": 39060 }, { "epoch": 3.93, "learning_rate": 1.0457460274942987e-08, "loss": 0.0775, "step": 39061 }, { "epoch": 3.93, "learning_rate": 1.0394372296115285e-08, "loss": 0.2551, "step": 39062 }, { "epoch": 3.93, "learning_rate": 1.0331475159732761e-08, "loss": 0.059, "step": 39063 }, { "epoch": 3.93, "learning_rate": 1.0268768866191768e-08, "loss": 0.1148, "step": 39064 }, { "epoch": 3.93, "learning_rate": 1.0206253415888656e-08, "loss": 0.0986, "step": 39065 }, { "epoch": 3.93, "learning_rate": 1.0143928809226432e-08, "loss": 0.1712, "step": 39066 }, { "epoch": 3.93, "learning_rate": 1.008179504660145e-08, "loss": 0.1919, "step": 39067 }, { "epoch": 3.93, "learning_rate": 1.0019852128403394e-08, "loss": 0.0877, "step": 39068 }, { "epoch": 3.93, "learning_rate": 9.958100055031948e-09, "loss": 0.7281, "step": 39069 }, { "epoch": 3.93, "learning_rate": 9.89653882688013e-09, "loss": 0.2817, "step": 39070 }, { "epoch": 3.93, "learning_rate": 9.835168444334298e-09, "loss": 0.1253, "step": 39071 }, { "epoch": 3.93, "learning_rate": 9.773988907790799e-09, "loss": 0.08, "step": 39072 }, { "epoch": 3.93, "learning_rate": 9.713000217635993e-09, "loss": 0.1712, "step": 39073 }, { "epoch": 3.93, "learning_rate": 9.652202374259566e-09, "loss": 0.1116, "step": 39074 }, { "epoch": 3.93, "learning_rate": 9.59159537804788e-09, "loss": 0.3446, "step": 39075 }, { "epoch": 3.93, "learning_rate": 9.531179229387287e-09, "loss": 0.0933, "step": 39076 }, { "epoch": 3.93, "learning_rate": 9.470953928657488e-09, "loss": 0.3127, "step": 39077 }, { "epoch": 3.93, "learning_rate": 9.41091947624817e-09, "loss": 0.067, "step": 39078 }, { "epoch": 3.93, "learning_rate": 9.351075872539028e-09, "loss": 0.0749, "step": 39079 }, { "epoch": 3.93, "learning_rate": 9.291423117909758e-09, "loss": 0.0479, "step": 39080 }, { "epoch": 3.93, "learning_rate": 9.23196121274339e-09, "loss": 0.0798, "step": 39081 }, { "epoch": 3.93, "learning_rate": 9.172690157412955e-09, "loss": 0.4671, "step": 39082 }, { "epoch": 3.93, "learning_rate": 9.113609952298151e-09, "loss": 0.1474, "step": 39083 }, { "epoch": 3.93, "learning_rate": 9.054720597775345e-09, "loss": 0.1593, "step": 39084 }, { "epoch": 3.93, "learning_rate": 8.996022094220901e-09, "loss": 0.2869, "step": 39085 }, { "epoch": 3.93, "learning_rate": 8.937514442004524e-09, "loss": 0.1005, "step": 39086 }, { "epoch": 3.93, "learning_rate": 8.87919764150258e-09, "loss": 0.4772, "step": 39087 }, { "epoch": 3.93, "learning_rate": 8.82107169308144e-09, "loss": 0.346, "step": 39088 }, { "epoch": 3.93, "learning_rate": 8.763136597117472e-09, "loss": 0.1287, "step": 39089 }, { "epoch": 3.93, "learning_rate": 8.705392353973717e-09, "loss": 0.0324, "step": 39090 }, { "epoch": 3.93, "learning_rate": 8.647838964019882e-09, "loss": 0.2278, "step": 39091 }, { "epoch": 3.93, "learning_rate": 8.59047642762234e-09, "loss": 0.1844, "step": 39092 }, { "epoch": 3.93, "learning_rate": 8.533304745144132e-09, "loss": 0.0654, "step": 39093 }, { "epoch": 3.93, "learning_rate": 8.476323916951634e-09, "loss": 0.0202, "step": 39094 }, { "epoch": 3.93, "learning_rate": 8.419533943404556e-09, "loss": 0.3494, "step": 39095 }, { "epoch": 3.93, "learning_rate": 8.362934824869272e-09, "loss": 0.3217, "step": 39096 }, { "epoch": 3.93, "learning_rate": 8.306526561702165e-09, "loss": 0.049, "step": 39097 }, { "epoch": 3.93, "learning_rate": 8.250309154262948e-09, "loss": 0.1313, "step": 39098 }, { "epoch": 3.93, "learning_rate": 8.194282602908e-09, "loss": 0.0805, "step": 39099 }, { "epoch": 3.93, "learning_rate": 8.138446907993701e-09, "loss": 0.3701, "step": 39100 }, { "epoch": 3.93, "learning_rate": 8.082802069879768e-09, "loss": 0.3071, "step": 39101 }, { "epoch": 3.93, "learning_rate": 8.02734808891592e-09, "loss": 0.0119, "step": 39102 }, { "epoch": 3.93, "learning_rate": 7.972084965458537e-09, "loss": 0.1982, "step": 39103 }, { "epoch": 3.93, "learning_rate": 7.91701269985734e-09, "loss": 0.2023, "step": 39104 }, { "epoch": 3.93, "learning_rate": 7.862131292458718e-09, "loss": 0.319, "step": 39105 }, { "epoch": 3.93, "learning_rate": 7.807440743619055e-09, "loss": 0.1794, "step": 39106 }, { "epoch": 3.93, "learning_rate": 7.752941053684737e-09, "loss": 0.2584, "step": 39107 }, { "epoch": 3.93, "learning_rate": 7.698632222995494e-09, "loss": 0.2224, "step": 39108 }, { "epoch": 3.93, "learning_rate": 7.644514251907708e-09, "loss": 0.0395, "step": 39109 }, { "epoch": 3.93, "learning_rate": 7.590587140757777e-09, "loss": 0.2713, "step": 39110 }, { "epoch": 3.93, "learning_rate": 7.536850889892088e-09, "loss": 0.179, "step": 39111 }, { "epoch": 3.93, "learning_rate": 7.483305499653704e-09, "loss": 0.2026, "step": 39112 }, { "epoch": 3.93, "learning_rate": 7.4299509703823485e-09, "loss": 0.0757, "step": 39113 }, { "epoch": 3.93, "learning_rate": 7.3767873024144224e-09, "loss": 0.6808, "step": 39114 }, { "epoch": 3.93, "learning_rate": 7.323814496092984e-09, "loss": 0.2432, "step": 39115 }, { "epoch": 3.94, "learning_rate": 7.2710325517511e-09, "loss": 0.5168, "step": 39116 }, { "epoch": 3.94, "learning_rate": 7.218441469728498e-09, "loss": 0.2529, "step": 39117 }, { "epoch": 3.94, "learning_rate": 7.166041250354915e-09, "loss": 0.1547, "step": 39118 }, { "epoch": 3.94, "learning_rate": 7.113831893970079e-09, "loss": 0.09, "step": 39119 }, { "epoch": 3.94, "learning_rate": 7.061813400900397e-09, "loss": 0.1014, "step": 39120 }, { "epoch": 3.94, "learning_rate": 7.0099857714822635e-09, "loss": 0.5295, "step": 39121 }, { "epoch": 3.94, "learning_rate": 6.958349006038756e-09, "loss": 0.0659, "step": 39122 }, { "epoch": 3.94, "learning_rate": 6.90690310490627e-09, "loss": 0.0092, "step": 39123 }, { "epoch": 3.94, "learning_rate": 6.855648068404552e-09, "loss": 0.8217, "step": 39124 }, { "epoch": 3.94, "learning_rate": 6.804583896866667e-09, "loss": 0.1651, "step": 39125 }, { "epoch": 3.94, "learning_rate": 6.753710590612361e-09, "loss": 0.6013, "step": 39126 }, { "epoch": 3.94, "learning_rate": 6.703028149968038e-09, "loss": 0.0697, "step": 39127 }, { "epoch": 3.94, "learning_rate": 6.6525365752567734e-09, "loss": 0.1793, "step": 39128 }, { "epoch": 3.94, "learning_rate": 6.602235866798312e-09, "loss": 0.1895, "step": 39129 }, { "epoch": 3.94, "learning_rate": 6.552126024912397e-09, "loss": 0.1097, "step": 39130 }, { "epoch": 3.94, "learning_rate": 6.502207049918773e-09, "loss": 0.044, "step": 39131 }, { "epoch": 3.94, "learning_rate": 6.452478942137185e-09, "loss": 0.0843, "step": 39132 }, { "epoch": 3.94, "learning_rate": 6.402941701880716e-09, "loss": 0.0448, "step": 39133 }, { "epoch": 3.94, "learning_rate": 6.353595329462447e-09, "loss": 0.2327, "step": 39134 }, { "epoch": 3.94, "learning_rate": 6.304439825202124e-09, "loss": 0.4613, "step": 39135 }, { "epoch": 3.94, "learning_rate": 6.255475189412829e-09, "loss": 0.2228, "step": 39136 }, { "epoch": 3.94, "learning_rate": 6.2067014224009846e-09, "loss": 0.1381, "step": 39137 }, { "epoch": 3.94, "learning_rate": 6.1581185244763415e-09, "loss": 0.055, "step": 39138 }, { "epoch": 3.94, "learning_rate": 6.1097264959553146e-09, "loss": 0.3002, "step": 39139 }, { "epoch": 3.94, "learning_rate": 6.061525337140994e-09, "loss": 0.1453, "step": 39140 }, { "epoch": 3.94, "learning_rate": 6.0135150483398016e-09, "loss": 0.1522, "step": 39141 }, { "epoch": 3.94, "learning_rate": 5.96569562986149e-09, "loss": 0.4419, "step": 39142 }, { "epoch": 3.94, "learning_rate": 5.9180670820024875e-09, "loss": 0.1786, "step": 39143 }, { "epoch": 3.94, "learning_rate": 5.870629405072547e-09, "loss": 0.3412, "step": 39144 }, { "epoch": 3.94, "learning_rate": 5.8233825993747604e-09, "loss": 0.1476, "step": 39145 }, { "epoch": 3.94, "learning_rate": 5.776326665202225e-09, "loss": 0.2347, "step": 39146 }, { "epoch": 3.94, "learning_rate": 5.729461602861364e-09, "loss": 0.136, "step": 39147 }, { "epoch": 3.94, "learning_rate": 5.682787412645274e-09, "loss": 0.0686, "step": 39148 }, { "epoch": 3.94, "learning_rate": 5.636304094853717e-09, "loss": 0.0557, "step": 39149 }, { "epoch": 3.94, "learning_rate": 5.590011649786453e-09, "loss": 0.0234, "step": 39150 }, { "epoch": 3.94, "learning_rate": 5.543910077729919e-09, "loss": 0.0297, "step": 39151 }, { "epoch": 3.94, "learning_rate": 5.497999378980545e-09, "loss": 0.0329, "step": 39152 }, { "epoch": 3.94, "learning_rate": 5.45227955383476e-09, "loss": 0.1087, "step": 39153 }, { "epoch": 3.94, "learning_rate": 5.406750602579003e-09, "loss": 0.2172, "step": 39154 }, { "epoch": 3.94, "learning_rate": 5.36141252550304e-09, "loss": 0.1102, "step": 39155 }, { "epoch": 3.94, "learning_rate": 5.316265322896641e-09, "loss": 0.1546, "step": 39156 }, { "epoch": 3.94, "learning_rate": 5.271308995049573e-09, "loss": 0.1638, "step": 39157 }, { "epoch": 3.94, "learning_rate": 5.226543542241613e-09, "loss": 0.0879, "step": 39158 }, { "epoch": 3.94, "learning_rate": 5.181968964765859e-09, "loss": 0.0885, "step": 39159 }, { "epoch": 3.94, "learning_rate": 5.137585262898759e-09, "loss": 0.1779, "step": 39160 }, { "epoch": 3.94, "learning_rate": 5.093392436923417e-09, "loss": 0.0599, "step": 39161 }, { "epoch": 3.94, "learning_rate": 5.049390487126271e-09, "loss": 0.0615, "step": 39162 }, { "epoch": 3.94, "learning_rate": 5.005579413783767e-09, "loss": 0.256, "step": 39163 }, { "epoch": 3.94, "learning_rate": 4.961959217172352e-09, "loss": 0.5488, "step": 39164 }, { "epoch": 3.94, "learning_rate": 4.91852989757513e-09, "loss": 1.0488, "step": 39165 }, { "epoch": 3.94, "learning_rate": 4.875291455265218e-09, "loss": 0.4221, "step": 39166 }, { "epoch": 3.94, "learning_rate": 4.8322438905157306e-09, "loss": 0.1326, "step": 39167 }, { "epoch": 3.94, "learning_rate": 4.789387203606443e-09, "loss": 0.1051, "step": 39168 }, { "epoch": 3.94, "learning_rate": 4.74672139480381e-09, "loss": 0.1702, "step": 39169 }, { "epoch": 3.94, "learning_rate": 4.7042464643842766e-09, "loss": 0.4077, "step": 39170 }, { "epoch": 3.94, "learning_rate": 4.661962412617626e-09, "loss": 0.121, "step": 39171 }, { "epoch": 3.94, "learning_rate": 4.6198692397703135e-09, "loss": 0.2441, "step": 39172 }, { "epoch": 3.94, "learning_rate": 4.577966946108792e-09, "loss": 0.05, "step": 39173 }, { "epoch": 3.94, "learning_rate": 4.5362555319028446e-09, "loss": 0.218, "step": 39174 }, { "epoch": 3.94, "learning_rate": 4.494734997418926e-09, "loss": 0.1629, "step": 39175 }, { "epoch": 3.94, "learning_rate": 4.453405342920158e-09, "loss": 0.0874, "step": 39176 }, { "epoch": 3.94, "learning_rate": 4.412266568669665e-09, "loss": 0.3833, "step": 39177 }, { "epoch": 3.94, "learning_rate": 4.371318674927238e-09, "loss": 0.1506, "step": 39178 }, { "epoch": 3.94, "learning_rate": 4.33056166195267e-09, "loss": 0.081, "step": 39179 }, { "epoch": 3.94, "learning_rate": 4.289995530009083e-09, "loss": 0.2276, "step": 39180 }, { "epoch": 3.94, "learning_rate": 4.24962027935294e-09, "loss": 0.4228, "step": 39181 }, { "epoch": 3.94, "learning_rate": 4.20943591024403e-09, "loss": 0.0365, "step": 39182 }, { "epoch": 3.94, "learning_rate": 4.169442422932157e-09, "loss": 0.1043, "step": 39183 }, { "epoch": 3.94, "learning_rate": 4.129639817673781e-09, "loss": 0.25, "step": 39184 }, { "epoch": 3.94, "learning_rate": 4.0900280947253625e-09, "loss": 0.1014, "step": 39185 }, { "epoch": 3.94, "learning_rate": 4.050607254336703e-09, "loss": 0.1926, "step": 39186 }, { "epoch": 3.94, "learning_rate": 4.011377296760932e-09, "loss": 0.5008, "step": 39187 }, { "epoch": 3.94, "learning_rate": 3.9723382222445206e-09, "loss": 0.2043, "step": 39188 }, { "epoch": 3.94, "learning_rate": 3.9334900310339375e-09, "loss": 0.0933, "step": 39189 }, { "epoch": 3.94, "learning_rate": 3.894832723382313e-09, "loss": 0.2653, "step": 39190 }, { "epoch": 3.94, "learning_rate": 3.856366299532788e-09, "loss": 0.267, "step": 39191 }, { "epoch": 3.94, "learning_rate": 3.818090759731829e-09, "loss": 0.1991, "step": 39192 }, { "epoch": 3.94, "learning_rate": 3.7800061042192465e-09, "loss": 0.3725, "step": 39193 }, { "epoch": 3.94, "learning_rate": 3.742112333238179e-09, "loss": 0.3215, "step": 39194 }, { "epoch": 3.94, "learning_rate": 3.7044094470350954e-09, "loss": 0.4366, "step": 39195 }, { "epoch": 3.94, "learning_rate": 3.6668974458431426e-09, "loss": 0.1973, "step": 39196 }, { "epoch": 3.94, "learning_rate": 3.6295763299054596e-09, "loss": 0.0606, "step": 39197 }, { "epoch": 3.94, "learning_rate": 3.592446099455193e-09, "loss": 0.3951, "step": 39198 }, { "epoch": 3.94, "learning_rate": 3.5555067547354825e-09, "loss": 0.0025, "step": 39199 }, { "epoch": 3.94, "learning_rate": 3.5187582959728127e-09, "loss": 0.2203, "step": 39200 }, { "epoch": 3.94, "learning_rate": 3.482200723406992e-09, "loss": 0.3158, "step": 39201 }, { "epoch": 3.94, "learning_rate": 3.4458340372711672e-09, "loss": 0.2458, "step": 39202 }, { "epoch": 3.94, "learning_rate": 3.409658237791824e-09, "loss": 0.0689, "step": 39203 }, { "epoch": 3.94, "learning_rate": 3.3736733252054396e-09, "loss": 0.3569, "step": 39204 }, { "epoch": 3.94, "learning_rate": 3.3378792997351693e-09, "loss": 0.3713, "step": 39205 }, { "epoch": 3.94, "learning_rate": 3.3022761616141596e-09, "loss": 0.2357, "step": 39206 }, { "epoch": 3.94, "learning_rate": 3.2668639110622346e-09, "loss": 0.1834, "step": 39207 }, { "epoch": 3.94, "learning_rate": 3.231642548312541e-09, "loss": 0.2494, "step": 39208 }, { "epoch": 3.94, "learning_rate": 3.196612073581573e-09, "loss": 0.1303, "step": 39209 }, { "epoch": 3.94, "learning_rate": 3.161772487099146e-09, "loss": 0.0639, "step": 39210 }, { "epoch": 3.94, "learning_rate": 3.127123789081754e-09, "loss": 0.1427, "step": 39211 }, { "epoch": 3.94, "learning_rate": 3.0926659797525515e-09, "loss": 0.34, "step": 39212 }, { "epoch": 3.94, "learning_rate": 3.0583990593313627e-09, "loss": 0.3673, "step": 39213 }, { "epoch": 3.94, "learning_rate": 3.024323028034681e-09, "loss": 0.2636, "step": 39214 }, { "epoch": 3.95, "learning_rate": 2.9904378860790003e-09, "loss": 0.6422, "step": 39215 }, { "epoch": 3.95, "learning_rate": 2.956743633680814e-09, "loss": 0.3946, "step": 39216 }, { "epoch": 3.95, "learning_rate": 2.9232402710532847e-09, "loss": 0.2142, "step": 39217 }, { "epoch": 3.95, "learning_rate": 2.889927798412906e-09, "loss": 0.1897, "step": 39218 }, { "epoch": 3.95, "learning_rate": 2.85680621596951e-09, "loss": 0.0907, "step": 39219 }, { "epoch": 3.95, "learning_rate": 2.8238755239329285e-09, "loss": 0.4198, "step": 39220 }, { "epoch": 3.95, "learning_rate": 2.791135722512994e-09, "loss": 0.2294, "step": 39221 }, { "epoch": 3.95, "learning_rate": 2.758586811919539e-09, "loss": 0.101, "step": 39222 }, { "epoch": 3.95, "learning_rate": 2.7262287923590647e-09, "loss": 0.1946, "step": 39223 }, { "epoch": 3.95, "learning_rate": 2.694061664038072e-09, "loss": 0.1291, "step": 39224 }, { "epoch": 3.95, "learning_rate": 2.6620854271597326e-09, "loss": 0.1554, "step": 39225 }, { "epoch": 3.95, "learning_rate": 2.6303000819272173e-09, "loss": 0.0956, "step": 39226 }, { "epoch": 3.95, "learning_rate": 2.598705628543696e-09, "loss": 0.2604, "step": 39227 }, { "epoch": 3.95, "learning_rate": 2.5673020672123405e-09, "loss": 0.0227, "step": 39228 }, { "epoch": 3.95, "learning_rate": 2.5360893981296597e-09, "loss": 0.3482, "step": 39229 }, { "epoch": 3.95, "learning_rate": 2.5050676214988244e-09, "loss": 0.7139, "step": 39230 }, { "epoch": 3.95, "learning_rate": 2.474236737509683e-09, "loss": 0.9467, "step": 39231 }, { "epoch": 3.95, "learning_rate": 2.4435967463654064e-09, "loss": 0.0901, "step": 39232 }, { "epoch": 3.95, "learning_rate": 2.413147648259173e-09, "loss": 0.1076, "step": 39233 }, { "epoch": 3.95, "learning_rate": 2.3828894433808313e-09, "loss": 0.0424, "step": 39234 }, { "epoch": 3.95, "learning_rate": 2.3528221319302213e-09, "loss": 0.5165, "step": 39235 }, { "epoch": 3.95, "learning_rate": 2.3229457140905296e-09, "loss": 0.1938, "step": 39236 }, { "epoch": 3.95, "learning_rate": 2.293260190058266e-09, "loss": 0.1024, "step": 39237 }, { "epoch": 3.95, "learning_rate": 2.263765560019948e-09, "loss": 0.3863, "step": 39238 }, { "epoch": 3.95, "learning_rate": 2.234461824162093e-09, "loss": 0.0341, "step": 39239 }, { "epoch": 3.95, "learning_rate": 2.2053489826745488e-09, "loss": 0.1247, "step": 39240 }, { "epoch": 3.95, "learning_rate": 2.176427035737172e-09, "loss": 0.1399, "step": 39241 }, { "epoch": 3.95, "learning_rate": 2.14769598353981e-09, "loss": 0.1701, "step": 39242 }, { "epoch": 3.95, "learning_rate": 2.11915582626232e-09, "loss": 0.3387, "step": 39243 }, { "epoch": 3.95, "learning_rate": 2.0908065640845576e-09, "loss": 0.0551, "step": 39244 }, { "epoch": 3.95, "learning_rate": 2.06264819719304e-09, "loss": 0.3681, "step": 39245 }, { "epoch": 3.95, "learning_rate": 2.0346807257609624e-09, "loss": 0.2445, "step": 39246 }, { "epoch": 3.95, "learning_rate": 2.0069041499681807e-09, "loss": 0.2139, "step": 39247 }, { "epoch": 3.95, "learning_rate": 1.9793184699912205e-09, "loss": 0.4502, "step": 39248 }, { "epoch": 3.95, "learning_rate": 1.951923686006607e-09, "loss": 0.0943, "step": 39249 }, { "epoch": 3.95, "learning_rate": 1.9247197981875355e-09, "loss": 0.1975, "step": 39250 }, { "epoch": 3.95, "learning_rate": 1.8977068067072e-09, "loss": 0.3745, "step": 39251 }, { "epoch": 3.95, "learning_rate": 1.8708847117387963e-09, "loss": 0.1139, "step": 39252 }, { "epoch": 3.95, "learning_rate": 1.8442535134555183e-09, "loss": 0.0601, "step": 39253 }, { "epoch": 3.95, "learning_rate": 1.8178132120205694e-09, "loss": 0.1458, "step": 39254 }, { "epoch": 3.95, "learning_rate": 1.7915638076038133e-09, "loss": 0.3773, "step": 39255 }, { "epoch": 3.95, "learning_rate": 1.7655053003751142e-09, "loss": 0.2218, "step": 39256 }, { "epoch": 3.95, "learning_rate": 1.7396376904976751e-09, "loss": 0.1357, "step": 39257 }, { "epoch": 3.95, "learning_rate": 1.7139609781380294e-09, "loss": 0.0246, "step": 39258 }, { "epoch": 3.95, "learning_rate": 1.6884751634593798e-09, "loss": 0.0501, "step": 39259 }, { "epoch": 3.95, "learning_rate": 1.663180246624929e-09, "loss": 0.1002, "step": 39260 }, { "epoch": 3.95, "learning_rate": 1.6380762277912186e-09, "loss": 0.1064, "step": 39261 }, { "epoch": 3.95, "learning_rate": 1.6131631071214514e-09, "loss": 0.1532, "step": 39262 }, { "epoch": 3.95, "learning_rate": 1.5884408847721687e-09, "loss": 0.4817, "step": 39263 }, { "epoch": 3.95, "learning_rate": 1.5639095609032427e-09, "loss": 0.2285, "step": 39264 }, { "epoch": 3.95, "learning_rate": 1.539569135671215e-09, "loss": 0.0262, "step": 39265 }, { "epoch": 3.95, "learning_rate": 1.515419609229296e-09, "loss": 0.2839, "step": 39266 }, { "epoch": 3.95, "learning_rate": 1.4914609817306967e-09, "loss": 0.1265, "step": 39267 }, { "epoch": 3.95, "learning_rate": 1.4676932533252974e-09, "loss": 0.1812, "step": 39268 }, { "epoch": 3.95, "learning_rate": 1.4441164241729698e-09, "loss": 0.015, "step": 39269 }, { "epoch": 3.95, "learning_rate": 1.4207304944136023e-09, "loss": 0.621, "step": 39270 }, { "epoch": 3.95, "learning_rate": 1.3975354642037364e-09, "loss": 0.0982, "step": 39271 }, { "epoch": 3.95, "learning_rate": 1.3745313336865905e-09, "loss": 0.2769, "step": 39272 }, { "epoch": 3.95, "learning_rate": 1.3517181030087144e-09, "loss": 0.0512, "step": 39273 }, { "epoch": 3.95, "learning_rate": 1.3290957723166575e-09, "loss": 0.3248, "step": 39274 }, { "epoch": 3.95, "learning_rate": 1.306664341756969e-09, "loss": 0.1623, "step": 39275 }, { "epoch": 3.95, "learning_rate": 1.2844238114662066e-09, "loss": 0.1764, "step": 39276 }, { "epoch": 3.95, "learning_rate": 1.2623741815909196e-09, "loss": 0.0656, "step": 39277 }, { "epoch": 3.95, "learning_rate": 1.240515452270996e-09, "loss": 0.5139, "step": 39278 }, { "epoch": 3.95, "learning_rate": 1.2188476236429935e-09, "loss": 0.0649, "step": 39279 }, { "epoch": 3.95, "learning_rate": 1.1973706958434693e-09, "loss": 0.2015, "step": 39280 }, { "epoch": 3.95, "learning_rate": 1.1760846690156423e-09, "loss": 0.2305, "step": 39281 }, { "epoch": 3.95, "learning_rate": 1.1549895432894086e-09, "loss": 0.1664, "step": 39282 }, { "epoch": 3.95, "learning_rate": 1.1340853188013255e-09, "loss": 0.3548, "step": 39283 }, { "epoch": 3.95, "learning_rate": 1.11337199568462e-09, "loss": 0.0682, "step": 39284 }, { "epoch": 3.95, "learning_rate": 1.092849574069188e-09, "loss": 0.4427, "step": 39285 }, { "epoch": 3.95, "learning_rate": 1.0725180540849255e-09, "loss": 0.0987, "step": 39286 }, { "epoch": 3.95, "learning_rate": 1.0523774358650596e-09, "loss": 0.1766, "step": 39287 }, { "epoch": 3.95, "learning_rate": 1.0324277195361553e-09, "loss": 0.4394, "step": 39288 }, { "epoch": 3.95, "learning_rate": 1.0126689052214478e-09, "loss": 0.1884, "step": 39289 }, { "epoch": 3.95, "learning_rate": 9.931009930541635e-10, "loss": 0.0634, "step": 39290 }, { "epoch": 3.95, "learning_rate": 9.737239831508759e-10, "loss": 0.0539, "step": 39291 }, { "epoch": 3.95, "learning_rate": 9.545378756414813e-10, "loss": 0.026, "step": 39292 }, { "epoch": 3.95, "learning_rate": 9.355426706458837e-10, "loss": 0.0499, "step": 39293 }, { "epoch": 3.95, "learning_rate": 9.167383682806562e-10, "loss": 0.2094, "step": 39294 }, { "epoch": 3.95, "learning_rate": 8.981249686723647e-10, "loss": 0.3039, "step": 39295 }, { "epoch": 3.95, "learning_rate": 8.797024719342517e-10, "loss": 0.1268, "step": 39296 }, { "epoch": 3.95, "learning_rate": 8.614708781862212e-10, "loss": 0.0523, "step": 39297 }, { "epoch": 3.95, "learning_rate": 8.434301875415162e-10, "loss": 0.2325, "step": 39298 }, { "epoch": 3.95, "learning_rate": 8.255804001200406e-10, "loss": 0.0113, "step": 39299 }, { "epoch": 3.95, "learning_rate": 8.079215160317066e-10, "loss": 0.2136, "step": 39300 }, { "epoch": 3.95, "learning_rate": 7.904535353864261e-10, "loss": 0.0926, "step": 39301 }, { "epoch": 3.95, "learning_rate": 7.731764583007728e-10, "loss": 0.1455, "step": 39302 }, { "epoch": 3.95, "learning_rate": 7.560902848846585e-10, "loss": 0.283, "step": 39303 }, { "epoch": 3.95, "learning_rate": 7.39195015241334e-10, "loss": 0.4983, "step": 39304 }, { "epoch": 3.95, "learning_rate": 7.224906494807116e-10, "loss": 0.3362, "step": 39305 }, { "epoch": 3.95, "learning_rate": 7.05977187712703e-10, "loss": 0.4351, "step": 39306 }, { "epoch": 3.95, "learning_rate": 6.896546300372286e-10, "loss": 0.2367, "step": 39307 }, { "epoch": 3.95, "learning_rate": 6.735229765642004e-10, "loss": 0.4804, "step": 39308 }, { "epoch": 3.95, "learning_rate": 6.57582227386877e-10, "loss": 0.2714, "step": 39309 }, { "epoch": 3.95, "learning_rate": 6.418323826151706e-10, "loss": 0.2188, "step": 39310 }, { "epoch": 3.95, "learning_rate": 6.262734423490013e-10, "loss": 0.1595, "step": 39311 }, { "epoch": 3.95, "learning_rate": 6.109054066816278e-10, "loss": 0.1497, "step": 39312 }, { "epoch": 3.96, "learning_rate": 5.957282757163007e-10, "loss": 0.0975, "step": 39313 }, { "epoch": 3.96, "learning_rate": 5.80742049546279e-10, "loss": 0.1279, "step": 39314 }, { "epoch": 3.96, "learning_rate": 5.659467282648212e-10, "loss": 0.085, "step": 39315 }, { "epoch": 3.96, "learning_rate": 5.513423119751782e-10, "loss": 0.0298, "step": 39316 }, { "epoch": 3.96, "learning_rate": 5.369288007606165e-10, "loss": 0.2898, "step": 39317 }, { "epoch": 3.96, "learning_rate": 5.227061947177258e-10, "loss": 0.126, "step": 39318 }, { "epoch": 3.96, "learning_rate": 5.08674493936434e-10, "loss": 0.5641, "step": 39319 }, { "epoch": 3.96, "learning_rate": 4.948336985033386e-10, "loss": 0.2184, "step": 39320 }, { "epoch": 3.96, "learning_rate": 4.811838085116982e-10, "loss": 0.1174, "step": 39321 }, { "epoch": 3.96, "learning_rate": 4.677248240447796e-10, "loss": 0.1572, "step": 39322 }, { "epoch": 3.96, "learning_rate": 4.5445674518584947e-10, "loss": 0.0783, "step": 39323 }, { "epoch": 3.96, "learning_rate": 4.413795720248359e-10, "loss": 0.1332, "step": 39324 }, { "epoch": 3.96, "learning_rate": 4.28493304641675e-10, "loss": 0.4851, "step": 39325 }, { "epoch": 3.96, "learning_rate": 4.1579794312296413e-10, "loss": 0.1959, "step": 39326 }, { "epoch": 3.96, "learning_rate": 4.0329348754197806e-10, "loss": 0.1627, "step": 39327 }, { "epoch": 3.96, "learning_rate": 3.909799379819834e-10, "loss": 0.0363, "step": 39328 }, { "epoch": 3.96, "learning_rate": 3.788572945229163e-10, "loss": 0.1143, "step": 39329 }, { "epoch": 3.96, "learning_rate": 3.6692555723805145e-10, "loss": 0.1518, "step": 39330 }, { "epoch": 3.96, "learning_rate": 3.551847262073249e-10, "loss": 0.2288, "step": 39331 }, { "epoch": 3.96, "learning_rate": 3.436348015040114e-10, "loss": 0.0785, "step": 39332 }, { "epoch": 3.96, "learning_rate": 3.3227578319805497e-10, "loss": 0.0363, "step": 39333 }, { "epoch": 3.96, "learning_rate": 3.2110767136939166e-10, "loss": 0.0861, "step": 39334 }, { "epoch": 3.96, "learning_rate": 3.1013046608130425e-10, "loss": 0.28, "step": 39335 }, { "epoch": 3.96, "learning_rate": 2.9934416741039807e-10, "loss": 0.089, "step": 39336 }, { "epoch": 3.96, "learning_rate": 2.887487754199558e-10, "loss": 0.1141, "step": 39337 }, { "epoch": 3.96, "learning_rate": 2.783442901799216e-10, "loss": 0.4284, "step": 39338 }, { "epoch": 3.96, "learning_rate": 2.681307117569087e-10, "loss": 0.094, "step": 39339 }, { "epoch": 3.96, "learning_rate": 2.581080402142e-10, "loss": 0.3034, "step": 39340 }, { "epoch": 3.96, "learning_rate": 2.4827627561507806e-10, "loss": 0.1035, "step": 39341 }, { "epoch": 3.96, "learning_rate": 2.3863541802615627e-10, "loss": 0.1382, "step": 39342 }, { "epoch": 3.96, "learning_rate": 2.2918546750405612e-10, "loss": 0.2922, "step": 39343 }, { "epoch": 3.96, "learning_rate": 2.1992642411206021e-10, "loss": 0.0113, "step": 39344 }, { "epoch": 3.96, "learning_rate": 2.1085828790678996e-10, "loss": 0.1617, "step": 39345 }, { "epoch": 3.96, "learning_rate": 2.019810589481974e-10, "loss": 0.4356, "step": 39346 }, { "epoch": 3.96, "learning_rate": 1.9329473728957326e-10, "loss": 0.0925, "step": 39347 }, { "epoch": 3.96, "learning_rate": 1.8479932299086956e-10, "loss": 0.2866, "step": 39348 }, { "epoch": 3.96, "learning_rate": 1.7649481610204632e-10, "loss": 0.155, "step": 39349 }, { "epoch": 3.96, "learning_rate": 1.6838121667639427e-10, "loss": 0.0408, "step": 39350 }, { "epoch": 3.96, "learning_rate": 1.604585247705348e-10, "loss": 0.2871, "step": 39351 }, { "epoch": 3.96, "learning_rate": 1.5272674042776658e-10, "loss": 0.1581, "step": 39352 }, { "epoch": 3.96, "learning_rate": 1.451858637013803e-10, "loss": 0.1125, "step": 39353 }, { "epoch": 3.96, "learning_rate": 1.378358946380054e-10, "loss": 0.4301, "step": 39354 }, { "epoch": 3.96, "learning_rate": 1.3067683328760182e-10, "loss": 0.4192, "step": 39355 }, { "epoch": 3.96, "learning_rate": 1.2370867969013767e-10, "loss": 0.1227, "step": 39356 }, { "epoch": 3.96, "learning_rate": 1.1693143389224227e-10, "loss": 0.1107, "step": 39357 }, { "epoch": 3.96, "learning_rate": 1.1034509594054498e-10, "loss": 0.0679, "step": 39358 }, { "epoch": 3.96, "learning_rate": 1.039496658716832e-10, "loss": 0.1406, "step": 39359 }, { "epoch": 3.96, "learning_rate": 9.77451437289556e-11, "loss": 0.1093, "step": 39360 }, { "epoch": 3.96, "learning_rate": 9.173152955566089e-11, "loss": 0.7555, "step": 39361 }, { "epoch": 3.96, "learning_rate": 8.590882338177508e-11, "loss": 0.409, "step": 39362 }, { "epoch": 3.96, "learning_rate": 8.027702524726621e-11, "loss": 0.0757, "step": 39363 }, { "epoch": 3.96, "learning_rate": 7.48361351921023e-11, "loss": 0.1243, "step": 39364 }, { "epoch": 3.96, "learning_rate": 6.958615324959005e-11, "loss": 0.2944, "step": 39365 }, { "epoch": 3.96, "learning_rate": 6.452707944970549e-11, "loss": 0.0379, "step": 39366 }, { "epoch": 3.96, "learning_rate": 5.965891382575528e-11, "loss": 0.5705, "step": 39367 }, { "epoch": 3.96, "learning_rate": 5.498165641104613e-11, "loss": 0.3106, "step": 39368 }, { "epoch": 3.96, "learning_rate": 5.0495307232223395e-11, "loss": 0.2157, "step": 39369 }, { "epoch": 3.96, "learning_rate": 4.619986632259376e-11, "loss": 0.1217, "step": 39370 }, { "epoch": 3.96, "learning_rate": 4.209533370214125e-11, "loss": 0.2093, "step": 39371 }, { "epoch": 3.96, "learning_rate": 3.818170940417254e-11, "loss": 0.1472, "step": 39372 }, { "epoch": 3.96, "learning_rate": 3.445899344867165e-11, "loss": 0.0422, "step": 39373 }, { "epoch": 3.96, "learning_rate": 3.0927185865614604e-11, "loss": 0.4155, "step": 39374 }, { "epoch": 3.96, "learning_rate": 2.758628666832408e-11, "loss": 0.2287, "step": 39375 }, { "epoch": 3.96, "learning_rate": 2.4436295883445427e-11, "loss": 0.1104, "step": 39376 }, { "epoch": 3.96, "learning_rate": 2.147721353429333e-11, "loss": 0.0774, "step": 39377 }, { "epoch": 3.96, "learning_rate": 1.8709039630859792e-11, "loss": 0.3785, "step": 39378 }, { "epoch": 3.96, "learning_rate": 1.6131774199790174e-11, "loss": 0.305, "step": 39379 }, { "epoch": 3.96, "learning_rate": 1.3745417251076476e-11, "loss": 0.5966, "step": 39380 }, { "epoch": 3.96, "learning_rate": 1.1549968804702715e-11, "loss": 0.3024, "step": 39381 }, { "epoch": 3.96, "learning_rate": 9.545428870660899e-12, "loss": 0.348, "step": 39382 }, { "epoch": 3.96, "learning_rate": 7.731797462273703e-12, "loss": 0.0965, "step": 39383 }, { "epoch": 3.96, "learning_rate": 6.1090745928638056e-12, "loss": 0.179, "step": 39384 }, { "epoch": 3.96, "learning_rate": 4.6772602724232116e-12, "loss": 0.0431, "step": 39385 }, { "epoch": 3.96, "learning_rate": 3.436354510943929e-12, "loss": 0.085, "step": 39386 }, { "epoch": 3.96, "learning_rate": 2.3863573117566264e-12, "loss": 0.082, "step": 39387 }, { "epoch": 3.96, "learning_rate": 1.5272686881839803e-12, "loss": 0.0299, "step": 39388 }, { "epoch": 3.96, "learning_rate": 8.590886402259912e-13, "loss": 0.0758, "step": 39389 }, { "epoch": 3.96, "learning_rate": 3.818171745439969e-13, "loss": 0.5835, "step": 39390 }, { "epoch": 3.96, "learning_rate": 9.545429446866649e-14, "loss": 0.0763, "step": 39391 }, { "epoch": 3.96, "learning_rate": 0.0, "loss": 0.1937, "step": 39392 } ], "logging_steps": 1, "max_steps": 39392, "num_train_epochs": 4, "save_steps": 500, "total_flos": 2.9875508351729664e+18, "trial_name": null, "trial_params": null }