| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 1.0, | |
| "eval_steps": 500, | |
| "global_step": 1003, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 6.451612903225806e-08, | |
| "loss": 0.8746, | |
| "step": 1 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 1.2903225806451611e-07, | |
| "loss": 0.847, | |
| "step": 2 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 1.9354838709677418e-07, | |
| "loss": 0.865, | |
| "step": 3 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 2.5806451612903223e-07, | |
| "loss": 0.8424, | |
| "step": 4 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 3.225806451612903e-07, | |
| "loss": 0.8527, | |
| "step": 5 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 3.8709677419354837e-07, | |
| "loss": 0.8545, | |
| "step": 6 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 4.5161290322580644e-07, | |
| "loss": 0.8473, | |
| "step": 7 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 5.161290322580645e-07, | |
| "loss": 0.8568, | |
| "step": 8 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 5.806451612903226e-07, | |
| "loss": 0.8731, | |
| "step": 9 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 6.451612903225806e-07, | |
| "loss": 0.832, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 7.096774193548387e-07, | |
| "loss": 0.8493, | |
| "step": 11 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 7.741935483870967e-07, | |
| "loss": 0.8487, | |
| "step": 12 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 8.387096774193549e-07, | |
| "loss": 0.8505, | |
| "step": 13 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.032258064516129e-07, | |
| "loss": 0.8437, | |
| "step": 14 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.67741935483871e-07, | |
| "loss": 0.8517, | |
| "step": 15 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.032258064516129e-06, | |
| "loss": 0.8152, | |
| "step": 16 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.096774193548387e-06, | |
| "loss": 0.8417, | |
| "step": 17 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.1612903225806452e-06, | |
| "loss": 0.8355, | |
| "step": 18 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.2258064516129033e-06, | |
| "loss": 0.8312, | |
| "step": 19 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.2903225806451612e-06, | |
| "loss": 0.8336, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.354838709677419e-06, | |
| "loss": 0.8312, | |
| "step": 21 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.4193548387096774e-06, | |
| "loss": 0.8185, | |
| "step": 22 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.4838709677419353e-06, | |
| "loss": 0.8287, | |
| "step": 23 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.5483870967741935e-06, | |
| "loss": 0.8063, | |
| "step": 24 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.6129032258064514e-06, | |
| "loss": 0.8165, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.6774193548387097e-06, | |
| "loss": 0.831, | |
| "step": 26 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.7419354838709676e-06, | |
| "loss": 0.8, | |
| "step": 27 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.8064516129032258e-06, | |
| "loss": 0.822, | |
| "step": 28 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.8709677419354837e-06, | |
| "loss": 0.8127, | |
| "step": 29 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.935483870967742e-06, | |
| "loss": 0.8006, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2e-06, | |
| "loss": 0.791, | |
| "step": 31 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.99999477679776e-06, | |
| "loss": 0.7875, | |
| "step": 32 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.9999791072456057e-06, | |
| "loss": 0.7787, | |
| "step": 33 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.999952991507226e-06, | |
| "loss": 0.7887, | |
| "step": 34 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.999916429855437e-06, | |
| "loss": 0.7604, | |
| "step": 35 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.999869422672177e-06, | |
| "loss": 0.7571, | |
| "step": 36 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9998119704485013e-06, | |
| "loss": 0.7736, | |
| "step": 37 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9997440737845793e-06, | |
| "loss": 0.7372, | |
| "step": 38 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9996657333896875e-06, | |
| "loss": 0.7438, | |
| "step": 39 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9995769500822006e-06, | |
| "loss": 0.741, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9994777247895852e-06, | |
| "loss": 0.7279, | |
| "step": 41 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9993680585483893e-06, | |
| "loss": 0.7237, | |
| "step": 42 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.99924795250423e-06, | |
| "loss": 0.6997, | |
| "step": 43 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.999117407911784e-06, | |
| "loss": 0.7155, | |
| "step": 44 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9989764261347736e-06, | |
| "loss": 0.7062, | |
| "step": 45 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.99882500864595e-06, | |
| "loss": 0.6946, | |
| "step": 46 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.998663157027083e-06, | |
| "loss": 0.6844, | |
| "step": 47 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9984908729689396e-06, | |
| "loss": 0.6732, | |
| "step": 48 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9983081582712684e-06, | |
| "loss": 0.6798, | |
| "step": 49 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.998115014842781e-06, | |
| "loss": 0.672, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.997911444701132e-06, | |
| "loss": 0.6613, | |
| "step": 51 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9976974499728977e-06, | |
| "loss": 0.6648, | |
| "step": 52 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.997473032893553e-06, | |
| "loss": 0.6919, | |
| "step": 53 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.99723819580745e-06, | |
| "loss": 0.6494, | |
| "step": 54 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.996992941167792e-06, | |
| "loss": 0.6548, | |
| "step": 55 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9967372715366077e-06, | |
| "loss": 0.6446, | |
| "step": 56 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9964711895847258e-06, | |
| "loss": 0.6595, | |
| "step": 57 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9961946980917456e-06, | |
| "loss": 0.6321, | |
| "step": 58 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9959077999460094e-06, | |
| "loss": 0.6256, | |
| "step": 59 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.995610498144571e-06, | |
| "loss": 0.6111, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9953027957931657e-06, | |
| "loss": 0.6327, | |
| "step": 61 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9949846961061763e-06, | |
| "loss": 0.6345, | |
| "step": 62 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9946562024066015e-06, | |
| "loss": 0.6122, | |
| "step": 63 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9943173181260186e-06, | |
| "loss": 0.5981, | |
| "step": 64 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.99396804680455e-06, | |
| "loss": 0.5979, | |
| "step": 65 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.993608392090825e-06, | |
| "loss": 0.5808, | |
| "step": 66 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9932383577419428e-06, | |
| "loss": 0.6, | |
| "step": 67 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9928579476234315e-06, | |
| "loss": 0.5995, | |
| "step": 68 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9924671657092095e-06, | |
| "loss": 0.5863, | |
| "step": 69 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.992066016081542e-06, | |
| "loss": 0.5733, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.991654502931001e-06, | |
| "loss": 0.5904, | |
| "step": 71 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.991232630556419e-06, | |
| "loss": 0.5648, | |
| "step": 72 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.990800403364845e-06, | |
| "loss": 0.5636, | |
| "step": 73 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9903578258714997e-06, | |
| "loss": 0.5781, | |
| "step": 74 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.989904902699727e-06, | |
| "loss": 0.5541, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9894416385809443e-06, | |
| "loss": 0.5647, | |
| "step": 76 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.988968038354597e-06, | |
| "loss": 0.5569, | |
| "step": 77 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9884841069681043e-06, | |
| "loss": 0.5596, | |
| "step": 78 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.987989849476809e-06, | |
| "loss": 0.5249, | |
| "step": 79 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.987485271043925e-06, | |
| "loss": 0.5418, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9869703769404826e-06, | |
| "loss": 0.5254, | |
| "step": 81 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9864451725452738e-06, | |
| "loss": 0.5122, | |
| "step": 82 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.985909663344796e-06, | |
| "loss": 0.5142, | |
| "step": 83 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.985363854933196e-06, | |
| "loss": 0.5189, | |
| "step": 84 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.984807753012208e-06, | |
| "loss": 0.5067, | |
| "step": 85 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9842413633910978e-06, | |
| "loss": 0.5123, | |
| "step": 86 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9836646919866008e-06, | |
| "loss": 0.4988, | |
| "step": 87 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9830777448228598e-06, | |
| "loss": 0.4941, | |
| "step": 88 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.982480528031362e-06, | |
| "loss": 0.5129, | |
| "step": 89 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9818730478508757e-06, | |
| "loss": 0.474, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9812553106273845e-06, | |
| "loss": 0.4745, | |
| "step": 91 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9806273228140215e-06, | |
| "loss": 0.4951, | |
| "step": 92 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.979989090971001e-06, | |
| "loss": 0.4859, | |
| "step": 93 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9793406217655515e-06, | |
| "loss": 0.4586, | |
| "step": 94 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.978681921971844e-06, | |
| "loss": 0.466, | |
| "step": 95 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.978012998470924e-06, | |
| "loss": 0.4826, | |
| "step": 96 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9773338582506357e-06, | |
| "loss": 0.4611, | |
| "step": 97 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.976644508405553e-06, | |
| "loss": 0.4483, | |
| "step": 98 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9759449561369034e-06, | |
| "loss": 0.4522, | |
| "step": 99 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.975235208752493e-06, | |
| "loss": 0.4452, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.97451527366663e-06, | |
| "loss": 0.4399, | |
| "step": 101 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9737851584000467e-06, | |
| "loss": 0.4211, | |
| "step": 102 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9730448705798236e-06, | |
| "loss": 0.4176, | |
| "step": 103 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9722944179393063e-06, | |
| "loss": 0.4194, | |
| "step": 104 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9715338083180265e-06, | |
| "loss": 0.4156, | |
| "step": 105 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.97076304966162e-06, | |
| "loss": 0.4155, | |
| "step": 106 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9699821500217434e-06, | |
| "loss": 0.4094, | |
| "step": 107 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9691911175559903e-06, | |
| "loss": 0.4153, | |
| "step": 108 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.968389960527806e-06, | |
| "loss": 0.3849, | |
| "step": 109 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9675786873064e-06, | |
| "loss": 0.3875, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9667573063666622e-06, | |
| "loss": 0.391, | |
| "step": 111 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.965925826289068e-06, | |
| "loss": 0.3967, | |
| "step": 112 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9650842557595965e-06, | |
| "loss": 0.3996, | |
| "step": 113 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.964232603569633e-06, | |
| "loss": 0.3902, | |
| "step": 114 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9633708786158803e-06, | |
| "loss": 0.3744, | |
| "step": 115 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9624990899002667e-06, | |
| "loss": 0.3761, | |
| "step": 116 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9616172465298493e-06, | |
| "loss": 0.3675, | |
| "step": 117 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9607253577167204e-06, | |
| "loss": 0.3761, | |
| "step": 118 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.959823432777912e-06, | |
| "loss": 0.3512, | |
| "step": 119 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.958911481135296e-06, | |
| "loss": 0.3605, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9579895123154886e-06, | |
| "loss": 0.3422, | |
| "step": 121 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9570575359497493e-06, | |
| "loss": 0.3337, | |
| "step": 122 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9561155617738796e-06, | |
| "loss": 0.3447, | |
| "step": 123 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.955163599628123e-06, | |
| "loss": 0.3544, | |
| "step": 124 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.954201659457061e-06, | |
| "loss": 0.3338, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.95322975130951e-06, | |
| "loss": 0.3219, | |
| "step": 126 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.952247885338415e-06, | |
| "loss": 0.3149, | |
| "step": 127 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.951256071800746e-06, | |
| "loss": 0.3177, | |
| "step": 128 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9502543210573877e-06, | |
| "loss": 0.3099, | |
| "step": 129 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.949242643573034e-06, | |
| "loss": 0.3198, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.948221049916076e-06, | |
| "loss": 0.3091, | |
| "step": 131 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.947189550758496e-06, | |
| "loss": 0.2959, | |
| "step": 132 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9461481568757504e-06, | |
| "loss": 0.2997, | |
| "step": 133 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.945096879146661e-06, | |
| "loss": 0.2905, | |
| "step": 134 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9440357285532997e-06, | |
| "loss": 0.2781, | |
| "step": 135 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9429647161808756e-06, | |
| "loss": 0.3065, | |
| "step": 136 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9418838532176173e-06, | |
| "loss": 0.2977, | |
| "step": 137 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9407931509546557e-06, | |
| "loss": 0.2961, | |
| "step": 138 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9396926207859082e-06, | |
| "loss": 0.2791, | |
| "step": 139 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9385822742079583e-06, | |
| "loss": 0.2906, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.937462122819935e-06, | |
| "loss": 0.2812, | |
| "step": 141 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.936332178323393e-06, | |
| "loss": 0.2765, | |
| "step": 142 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9351924525221897e-06, | |
| "loss": 0.289, | |
| "step": 143 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9340429573223616e-06, | |
| "loss": 0.2643, | |
| "step": 144 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9328837047320007e-06, | |
| "loss": 0.2657, | |
| "step": 145 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.931714706861128e-06, | |
| "loss": 0.2618, | |
| "step": 146 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9305359759215686e-06, | |
| "loss": 0.2616, | |
| "step": 147 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.929347524226822e-06, | |
| "loss": 0.2531, | |
| "step": 148 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9281493641919365e-06, | |
| "loss": 0.248, | |
| "step": 149 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9269415083333756e-06, | |
| "loss": 0.2493, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9257239692688907e-06, | |
| "loss": 0.2655, | |
| "step": 151 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9244967597173862e-06, | |
| "loss": 0.2495, | |
| "step": 152 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9232598924987902e-06, | |
| "loss": 0.2632, | |
| "step": 153 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.922013380533918e-06, | |
| "loss": 0.2436, | |
| "step": 154 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9207572368443383e-06, | |
| "loss": 0.2606, | |
| "step": 155 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.919491474552235e-06, | |
| "loss": 0.2469, | |
| "step": 156 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9182161068802742e-06, | |
| "loss": 0.2406, | |
| "step": 157 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.916931147151461e-06, | |
| "loss": 0.2447, | |
| "step": 158 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.915636608789006e-06, | |
| "loss": 0.2518, | |
| "step": 159 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9143325053161793e-06, | |
| "loss": 0.2431, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.913018850356174e-06, | |
| "loss": 0.244, | |
| "step": 161 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9116956576319608e-06, | |
| "loss": 0.2356, | |
| "step": 162 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9103629409661467e-06, | |
| "loss": 0.2395, | |
| "step": 163 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.909020714280828e-06, | |
| "loss": 0.2333, | |
| "step": 164 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9076689915974486e-06, | |
| "loss": 0.2461, | |
| "step": 165 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.9063077870366499e-06, | |
| "loss": 0.2381, | |
| "step": 166 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.9049371148181253e-06, | |
| "loss": 0.2239, | |
| "step": 167 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.9035569892604713e-06, | |
| "loss": 0.2335, | |
| "step": 168 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.9021674247810378e-06, | |
| "loss": 0.2429, | |
| "step": 169 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.900768435895777e-06, | |
| "loss": 0.2399, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.899360037219093e-06, | |
| "loss": 0.2345, | |
| "step": 171 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.897942243463688e-06, | |
| "loss": 0.2343, | |
| "step": 172 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.8965150694404094e-06, | |
| "loss": 0.2282, | |
| "step": 173 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.8950785300580935e-06, | |
| "loss": 0.2219, | |
| "step": 174 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.8936326403234122e-06, | |
| "loss": 0.2314, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8921774153407143e-06, | |
| "loss": 0.2251, | |
| "step": 176 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8907128703118693e-06, | |
| "loss": 0.2274, | |
| "step": 177 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.889239020536106e-06, | |
| "loss": 0.2316, | |
| "step": 178 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.887755881409856e-06, | |
| "loss": 0.2282, | |
| "step": 179 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8862634684265898e-06, | |
| "loss": 0.2336, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8847617971766576e-06, | |
| "loss": 0.2215, | |
| "step": 181 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8832508833471247e-06, | |
| "loss": 0.2213, | |
| "step": 182 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8817307427216077e-06, | |
| "loss": 0.2147, | |
| "step": 183 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8802013911801108e-06, | |
| "loss": 0.2145, | |
| "step": 184 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8786628446988592e-06, | |
| "loss": 0.216, | |
| "step": 185 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8771151193501305e-06, | |
| "loss": 0.2072, | |
| "step": 186 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8755582313020908e-06, | |
| "loss": 0.2105, | |
| "step": 187 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.873992196818622e-06, | |
| "loss": 0.2129, | |
| "step": 188 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8724170322591536e-06, | |
| "loss": 0.2273, | |
| "step": 189 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.870832754078492e-06, | |
| "loss": 0.2145, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8692393788266476e-06, | |
| "loss": 0.218, | |
| "step": 191 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8676369231486633e-06, | |
| "loss": 0.2167, | |
| "step": 192 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8660254037844386e-06, | |
| "loss": 0.2038, | |
| "step": 193 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.864404837568557e-06, | |
| "loss": 0.2085, | |
| "step": 194 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8627752414301084e-06, | |
| "loss": 0.2174, | |
| "step": 195 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8611366323925135e-06, | |
| "loss": 0.2119, | |
| "step": 196 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.859489027573345e-06, | |
| "loss": 0.2116, | |
| "step": 197 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8578324441841494e-06, | |
| "loss": 0.2108, | |
| "step": 198 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8561668995302663e-06, | |
| "loss": 0.2193, | |
| "step": 199 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8544924110106494e-06, | |
| "loss": 0.2259, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8528089961176828e-06, | |
| "loss": 0.2217, | |
| "step": 201 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8511166724369996e-06, | |
| "loss": 0.2155, | |
| "step": 202 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8494154576472974e-06, | |
| "loss": 0.2082, | |
| "step": 203 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8477053695201536e-06, | |
| "loss": 0.2126, | |
| "step": 204 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.845986425919841e-06, | |
| "loss": 0.2109, | |
| "step": 205 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8442586448031395e-06, | |
| "loss": 0.2031, | |
| "step": 206 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8425220442191493e-06, | |
| "loss": 0.2181, | |
| "step": 207 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.840776642309103e-06, | |
| "loss": 0.2133, | |
| "step": 208 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8390224573061746e-06, | |
| "loss": 0.205, | |
| "step": 209 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.83725950753529e-06, | |
| "loss": 0.2161, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8354878114129364e-06, | |
| "loss": 0.2017, | |
| "step": 211 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8337073874469671e-06, | |
| "loss": 0.2055, | |
| "step": 212 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8319182542364115e-06, | |
| "loss": 0.2018, | |
| "step": 213 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8301204304712788e-06, | |
| "loss": 0.2023, | |
| "step": 214 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8283139349323631e-06, | |
| "loss": 0.2089, | |
| "step": 215 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8264987864910473e-06, | |
| "loss": 0.202, | |
| "step": 216 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8246750041091068e-06, | |
| "loss": 0.2112, | |
| "step": 217 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8228426068385092e-06, | |
| "loss": 0.2043, | |
| "step": 218 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8210016138212186e-06, | |
| "loss": 0.2004, | |
| "step": 219 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8191520442889917e-06, | |
| "loss": 0.2047, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8172939175631806e-06, | |
| "loss": 0.2149, | |
| "step": 221 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8154272530545283e-06, | |
| "loss": 0.2019, | |
| "step": 222 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8135520702629674e-06, | |
| "loss": 0.2047, | |
| "step": 223 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.811668388777416e-06, | |
| "loss": 0.21, | |
| "step": 224 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8097762282755724e-06, | |
| "loss": 0.1902, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.807875608523711e-06, | |
| "loss": 0.2032, | |
| "step": 226 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.8059665493764742e-06, | |
| "loss": 0.2124, | |
| "step": 227 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.8040490707766662e-06, | |
| "loss": 0.2053, | |
| "step": 228 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.8021231927550438e-06, | |
| "loss": 0.187, | |
| "step": 229 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.8001889354301078e-06, | |
| "loss": 0.2032, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.7982463190078926e-06, | |
| "loss": 0.1954, | |
| "step": 231 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.7962953637817556e-06, | |
| "loss": 0.2037, | |
| "step": 232 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.7943360901321638e-06, | |
| "loss": 0.2066, | |
| "step": 233 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.792368518526482e-06, | |
| "loss": 0.207, | |
| "step": 234 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.7903926695187592e-06, | |
| "loss": 0.1976, | |
| "step": 235 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7884085637495136e-06, | |
| "loss": 0.2051, | |
| "step": 236 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.786416221945516e-06, | |
| "loss": 0.2125, | |
| "step": 237 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7844156649195757e-06, | |
| "loss": 0.2028, | |
| "step": 238 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7824069135703196e-06, | |
| "loss": 0.2108, | |
| "step": 239 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7803899888819775e-06, | |
| "loss": 0.1977, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.77836491192416e-06, | |
| "loss": 0.2118, | |
| "step": 241 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7763317038516403e-06, | |
| "loss": 0.1985, | |
| "step": 242 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7742903859041323e-06, | |
| "loss": 0.1871, | |
| "step": 243 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7722409794060691e-06, | |
| "loss": 0.1987, | |
| "step": 244 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7701835057663797e-06, | |
| "loss": 0.2064, | |
| "step": 245 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7681179864782658e-06, | |
| "loss": 0.1912, | |
| "step": 246 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.766044443118978e-06, | |
| "loss": 0.2011, | |
| "step": 247 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7639628973495882e-06, | |
| "loss": 0.2219, | |
| "step": 248 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.761873370914766e-06, | |
| "loss": 0.2125, | |
| "step": 249 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7597758856425494e-06, | |
| "loss": 0.2159, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7576704634441178e-06, | |
| "loss": 0.1988, | |
| "step": 251 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7555571263135636e-06, | |
| "loss": 0.2084, | |
| "step": 252 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7534358963276605e-06, | |
| "loss": 0.2042, | |
| "step": 253 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7513067956456357e-06, | |
| "loss": 0.176, | |
| "step": 254 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.749169846508936e-06, | |
| "loss": 0.206, | |
| "step": 255 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7470250712409959e-06, | |
| "loss": 0.1951, | |
| "step": 256 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7448724922470058e-06, | |
| "loss": 0.1937, | |
| "step": 257 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7427121320136764e-06, | |
| "loss": 0.2041, | |
| "step": 258 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7405440131090047e-06, | |
| "loss": 0.1961, | |
| "step": 259 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7383681581820372e-06, | |
| "loss": 0.21, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7361845899626353e-06, | |
| "loss": 0.1984, | |
| "step": 261 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7339933312612352e-06, | |
| "loss": 0.2032, | |
| "step": 262 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7317944049686122e-06, | |
| "loss": 0.1925, | |
| "step": 263 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.729587834055639e-06, | |
| "loss": 0.1961, | |
| "step": 264 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7273736415730487e-06, | |
| "loss": 0.2011, | |
| "step": 265 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7251518506511911e-06, | |
| "loss": 0.2065, | |
| "step": 266 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7229224844997927e-06, | |
| "loss": 0.2086, | |
| "step": 267 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7206855664077145e-06, | |
| "loss": 0.1992, | |
| "step": 268 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7184411197427076e-06, | |
| "loss": 0.2237, | |
| "step": 269 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7161891679511693e-06, | |
| "loss": 0.204, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7139297345578992e-06, | |
| "loss": 0.2044, | |
| "step": 271 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7116628431658523e-06, | |
| "loss": 0.1936, | |
| "step": 272 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7093885174558929e-06, | |
| "loss": 0.2036, | |
| "step": 273 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7071067811865474e-06, | |
| "loss": 0.1989, | |
| "step": 274 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.704817658193756e-06, | |
| "loss": 0.189, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.7025211723906232e-06, | |
| "loss": 0.1934, | |
| "step": 276 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.7002173477671685e-06, | |
| "loss": 0.194, | |
| "step": 277 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.697906208390076e-06, | |
| "loss": 0.2012, | |
| "step": 278 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6955877784024417e-06, | |
| "loss": 0.2021, | |
| "step": 279 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6932620820235243e-06, | |
| "loss": 0.1906, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6909291435484878e-06, | |
| "loss": 0.2123, | |
| "step": 281 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6885889873481513e-06, | |
| "loss": 0.2063, | |
| "step": 282 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6862416378687337e-06, | |
| "loss": 0.1912, | |
| "step": 283 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.683887119631596e-06, | |
| "loss": 0.2059, | |
| "step": 284 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6815254572329894e-06, | |
| "loss": 0.2053, | |
| "step": 285 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6791566753437931e-06, | |
| "loss": 0.1989, | |
| "step": 286 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.676780798709262e-06, | |
| "loss": 0.1985, | |
| "step": 287 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6743978521487638e-06, | |
| "loss": 0.2003, | |
| "step": 288 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6720078605555223e-06, | |
| "loss": 0.2029, | |
| "step": 289 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6696108488963566e-06, | |
| "loss": 0.1846, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6672068422114194e-06, | |
| "loss": 0.2004, | |
| "step": 291 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6647958656139376e-06, | |
| "loss": 0.2042, | |
| "step": 292 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6623779442899476e-06, | |
| "loss": 0.2088, | |
| "step": 293 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6599531034980338e-06, | |
| "loss": 0.1955, | |
| "step": 294 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6575213685690636e-06, | |
| "loss": 0.1993, | |
| "step": 295 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6550827649059235e-06, | |
| "loss": 0.197, | |
| "step": 296 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6526373179832545e-06, | |
| "loss": 0.1948, | |
| "step": 297 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6501850533471835e-06, | |
| "loss": 0.1958, | |
| "step": 298 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6477259966150588e-06, | |
| "loss": 0.1996, | |
| "step": 299 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6452601734751823e-06, | |
| "loss": 0.1981, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6427876096865393e-06, | |
| "loss": 0.1897, | |
| "step": 301 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6403083310785314e-06, | |
| "loss": 0.2017, | |
| "step": 302 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6378223635507058e-06, | |
| "loss": 0.2075, | |
| "step": 303 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6353297330724848e-06, | |
| "loss": 0.1836, | |
| "step": 304 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.632830465682895e-06, | |
| "loss": 0.203, | |
| "step": 305 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6303245874902936e-06, | |
| "loss": 0.1866, | |
| "step": 306 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6278121246720987e-06, | |
| "loss": 0.1935, | |
| "step": 307 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.625293103474512e-06, | |
| "loss": 0.2037, | |
| "step": 308 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.622767550212249e-06, | |
| "loss": 0.1893, | |
| "step": 309 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6202354912682598e-06, | |
| "loss": 0.2012, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.617696953093457e-06, | |
| "loss": 0.2026, | |
| "step": 311 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6151519622064368e-06, | |
| "loss": 0.1988, | |
| "step": 312 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6126005451932027e-06, | |
| "loss": 0.1954, | |
| "step": 313 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.61004272870689e-06, | |
| "loss": 0.2026, | |
| "step": 314 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6074785394674835e-06, | |
| "loss": 0.2052, | |
| "step": 315 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.6049080042615414e-06, | |
| "loss": 0.1967, | |
| "step": 316 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.6023311499419147e-06, | |
| "loss": 0.2052, | |
| "step": 317 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.5997480034274648e-06, | |
| "loss": 0.1983, | |
| "step": 318 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.5971585917027862e-06, | |
| "loss": 0.1938, | |
| "step": 319 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.5945629418179203e-06, | |
| "loss": 0.2032, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.591961080888076e-06, | |
| "loss": 0.2117, | |
| "step": 321 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.5893530360933448e-06, | |
| "loss": 0.1971, | |
| "step": 322 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.586738834678418e-06, | |
| "loss": 0.2021, | |
| "step": 323 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.5841185039523002e-06, | |
| "loss": 0.194, | |
| "step": 324 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.5814920712880267e-06, | |
| "loss": 0.1952, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.578859564122375e-06, | |
| "loss": 0.2004, | |
| "step": 326 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5762210099555801e-06, | |
| "loss": 0.2066, | |
| "step": 327 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.573576436351046e-06, | |
| "loss": 0.1959, | |
| "step": 328 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5709258709350582e-06, | |
| "loss": 0.1893, | |
| "step": 329 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.568269341396495e-06, | |
| "loss": 0.1963, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5656068754865386e-06, | |
| "loss": 0.1963, | |
| "step": 331 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.562938501018385e-06, | |
| "loss": 0.1896, | |
| "step": 332 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5602642458669525e-06, | |
| "loss": 0.1858, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5575841379685928e-06, | |
| "loss": 0.1935, | |
| "step": 334 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5548982053207969e-06, | |
| "loss": 0.1938, | |
| "step": 335 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5522064759819035e-06, | |
| "loss": 0.1973, | |
| "step": 336 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.549508978070806e-06, | |
| "loss": 0.1845, | |
| "step": 337 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.546805739766659e-06, | |
| "loss": 0.1887, | |
| "step": 338 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5440967893085826e-06, | |
| "loss": 0.2031, | |
| "step": 339 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5413821549953696e-06, | |
| "loss": 0.186, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5386618651851877e-06, | |
| "loss": 0.201, | |
| "step": 341 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5359359482952847e-06, | |
| "loss": 0.1954, | |
| "step": 342 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5332044328016912e-06, | |
| "loss": 0.1792, | |
| "step": 343 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5304673472389224e-06, | |
| "loss": 0.1854, | |
| "step": 344 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5277247201996818e-06, | |
| "loss": 0.1885, | |
| "step": 345 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.52497658033456e-06, | |
| "loss": 0.1896, | |
| "step": 346 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5222229563517383e-06, | |
| "loss": 0.1866, | |
| "step": 347 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5194638770166863e-06, | |
| "loss": 0.1846, | |
| "step": 348 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.516699371151863e-06, | |
| "loss": 0.1807, | |
| "step": 349 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5139294676364145e-06, | |
| "loss": 0.1964, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.511154195405873e-06, | |
| "loss": 0.1993, | |
| "step": 351 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5083735834518555e-06, | |
| "loss": 0.1954, | |
| "step": 352 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5055876608217587e-06, | |
| "loss": 0.1934, | |
| "step": 353 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5027964566184576e-06, | |
| "loss": 0.1884, | |
| "step": 354 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5e-06, | |
| "loss": 0.1928, | |
| "step": 355 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.4971983201793028e-06, | |
| "loss": 0.195, | |
| "step": 356 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4943914464238467e-06, | |
| "loss": 0.1935, | |
| "step": 357 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4915794080553704e-06, | |
| "loss": 0.1785, | |
| "step": 358 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.488762234449564e-06, | |
| "loss": 0.1891, | |
| "step": 359 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4859399550357628e-06, | |
| "loss": 0.1966, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4831125992966385e-06, | |
| "loss": 0.192, | |
| "step": 361 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4802801967678932e-06, | |
| "loss": 0.1981, | |
| "step": 362 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4774427770379489e-06, | |
| "loss": 0.2036, | |
| "step": 363 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4746003697476404e-06, | |
| "loss": 0.195, | |
| "step": 364 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4717530045899036e-06, | |
| "loss": 0.1867, | |
| "step": 365 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4689007113094664e-06, | |
| "loss": 0.1828, | |
| "step": 366 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4660435197025389e-06, | |
| "loss": 0.1915, | |
| "step": 367 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4631814596164994e-06, | |
| "loss": 0.1976, | |
| "step": 368 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4603145609495857e-06, | |
| "loss": 0.1825, | |
| "step": 369 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4574428536505807e-06, | |
| "loss": 0.1944, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4545663677185004e-06, | |
| "loss": 0.1887, | |
| "step": 371 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4516851332022805e-06, | |
| "loss": 0.204, | |
| "step": 372 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4487991802004622e-06, | |
| "loss": 0.1889, | |
| "step": 373 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4459085388608775e-06, | |
| "loss": 0.1937, | |
| "step": 374 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4430132393803352e-06, | |
| "loss": 0.2044, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4401133120043048e-06, | |
| "loss": 0.1886, | |
| "step": 376 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4372087870266008e-06, | |
| "loss": 0.1957, | |
| "step": 377 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4342996947890655e-06, | |
| "loss": 0.1935, | |
| "step": 378 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4313860656812535e-06, | |
| "loss": 0.2017, | |
| "step": 379 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.428467930140113e-06, | |
| "loss": 0.1927, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4255453186496673e-06, | |
| "loss": 0.197, | |
| "step": 381 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4226182617406994e-06, | |
| "loss": 0.1954, | |
| "step": 382 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.419686789990429e-06, | |
| "loss": 0.1927, | |
| "step": 383 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4167509340221959e-06, | |
| "loss": 0.1957, | |
| "step": 384 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4138107245051392e-06, | |
| "loss": 0.1922, | |
| "step": 385 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4108661921538765e-06, | |
| "loss": 0.1879, | |
| "step": 386 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.4079173677281835e-06, | |
| "loss": 0.1833, | |
| "step": 387 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.4049642820326736e-06, | |
| "loss": 0.1859, | |
| "step": 388 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.402006965916474e-06, | |
| "loss": 0.1875, | |
| "step": 389 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.3990454502729047e-06, | |
| "loss": 0.1971, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.3960797660391568e-06, | |
| "loss": 0.1873, | |
| "step": 391 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.393109944195967e-06, | |
| "loss": 0.1864, | |
| "step": 392 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.3901360157672949e-06, | |
| "loss": 0.1909, | |
| "step": 393 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.3871580118200006e-06, | |
| "loss": 0.1892, | |
| "step": 394 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.3841759634635176e-06, | |
| "loss": 0.2082, | |
| "step": 395 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.381189901849529e-06, | |
| "loss": 0.1937, | |
| "step": 396 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3781998581716424e-06, | |
| "loss": 0.1777, | |
| "step": 397 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3752058636650636e-06, | |
| "loss": 0.1907, | |
| "step": 398 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3722079496062699e-06, | |
| "loss": 0.1956, | |
| "step": 399 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3692061473126845e-06, | |
| "loss": 0.1845, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.366200488142348e-06, | |
| "loss": 0.1925, | |
| "step": 401 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3631910034935921e-06, | |
| "loss": 0.1843, | |
| "step": 402 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3601777248047103e-06, | |
| "loss": 0.1898, | |
| "step": 403 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3571606835536314e-06, | |
| "loss": 0.175, | |
| "step": 404 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3541399112575875e-06, | |
| "loss": 0.1913, | |
| "step": 405 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3511154394727889e-06, | |
| "loss": 0.1908, | |
| "step": 406 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3480872997940905e-06, | |
| "loss": 0.1931, | |
| "step": 407 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3450555238546647e-06, | |
| "loss": 0.1876, | |
| "step": 408 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3420201433256689e-06, | |
| "loss": 0.1912, | |
| "step": 409 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3389811899159159e-06, | |
| "loss": 0.1811, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3359386953715422e-06, | |
| "loss": 0.1858, | |
| "step": 411 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3328926914756764e-06, | |
| "loss": 0.1956, | |
| "step": 412 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3298432100481077e-06, | |
| "loss": 0.183, | |
| "step": 413 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3267902829449521e-06, | |
| "loss": 0.1834, | |
| "step": 414 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3237339420583212e-06, | |
| "loss": 0.1915, | |
| "step": 415 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3206742193159876e-06, | |
| "loss": 0.1828, | |
| "step": 416 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3176111466810531e-06, | |
| "loss": 0.1935, | |
| "step": 417 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3145447561516136e-06, | |
| "loss": 0.1973, | |
| "step": 418 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3114750797604247e-06, | |
| "loss": 0.1914, | |
| "step": 419 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3084021495745671e-06, | |
| "loss": 0.1987, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3053259976951131e-06, | |
| "loss": 0.1877, | |
| "step": 421 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3022466562567893e-06, | |
| "loss": 0.1954, | |
| "step": 422 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.2991641574276418e-06, | |
| "loss": 0.1919, | |
| "step": 423 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.2960785334086999e-06, | |
| "loss": 0.1829, | |
| "step": 424 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.2929898164336405e-06, | |
| "loss": 0.1911, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.2898980387684507e-06, | |
| "loss": 0.1927, | |
| "step": 426 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2868032327110903e-06, | |
| "loss": 0.1858, | |
| "step": 427 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.283705430591155e-06, | |
| "loss": 0.1853, | |
| "step": 428 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2806046647695388e-06, | |
| "loss": 0.1892, | |
| "step": 429 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2775009676380956e-06, | |
| "loss": 0.1895, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2743943716193016e-06, | |
| "loss": 0.1981, | |
| "step": 431 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2712849091659145e-06, | |
| "loss": 0.1956, | |
| "step": 432 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2681726127606374e-06, | |
| "loss": 0.1856, | |
| "step": 433 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.265057514915777e-06, | |
| "loss": 0.1843, | |
| "step": 434 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2619396481729059e-06, | |
| "loss": 0.1898, | |
| "step": 435 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2588190451025207e-06, | |
| "loss": 0.1873, | |
| "step": 436 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2556957383037038e-06, | |
| "loss": 0.1793, | |
| "step": 437 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2525697604037807e-06, | |
| "loss": 0.1894, | |
| "step": 438 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2494411440579813e-06, | |
| "loss": 0.186, | |
| "step": 439 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2463099219490976e-06, | |
| "loss": 0.1839, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2431761267871416e-06, | |
| "loss": 0.1817, | |
| "step": 441 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2400397913090059e-06, | |
| "loss": 0.1967, | |
| "step": 442 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.236900948278119e-06, | |
| "loss": 0.1947, | |
| "step": 443 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2337596304841052e-06, | |
| "loss": 0.1833, | |
| "step": 444 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2306158707424401e-06, | |
| "loss": 0.1786, | |
| "step": 445 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.22746970189411e-06, | |
| "loss": 0.1896, | |
| "step": 446 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2243211568052676e-06, | |
| "loss": 0.1896, | |
| "step": 447 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2211702683668877e-06, | |
| "loss": 0.1836, | |
| "step": 448 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2180170694944258e-06, | |
| "loss": 0.203, | |
| "step": 449 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2148615931274729e-06, | |
| "loss": 0.1989, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2117038722294109e-06, | |
| "loss": 0.1896, | |
| "step": 451 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2085439397870698e-06, | |
| "loss": 0.1945, | |
| "step": 452 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.205381828810382e-06, | |
| "loss": 0.1869, | |
| "step": 453 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.202217572332038e-06, | |
| "loss": 0.1762, | |
| "step": 454 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.1990512034071406e-06, | |
| "loss": 0.1932, | |
| "step": 455 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.1958827551128602e-06, | |
| "loss": 0.1821, | |
| "step": 456 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1927122605480896e-06, | |
| "loss": 0.1788, | |
| "step": 457 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1895397528330973e-06, | |
| "loss": 0.202, | |
| "step": 458 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1863652651091822e-06, | |
| "loss": 0.1901, | |
| "step": 459 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1831888305383266e-06, | |
| "loss": 0.1857, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1800104823028514e-06, | |
| "loss": 0.1886, | |
| "step": 461 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1768302536050675e-06, | |
| "loss": 0.1948, | |
| "step": 462 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1736481776669305e-06, | |
| "loss": 0.1858, | |
| "step": 463 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1704642877296924e-06, | |
| "loss": 0.1784, | |
| "step": 464 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1672786170535552e-06, | |
| "loss": 0.1913, | |
| "step": 465 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.164091198917324e-06, | |
| "loss": 0.1733, | |
| "step": 466 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1609020666180573e-06, | |
| "loss": 0.1864, | |
| "step": 467 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1577112534707212e-06, | |
| "loss": 0.195, | |
| "step": 468 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1545187928078406e-06, | |
| "loss": 0.1924, | |
| "step": 469 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1513247179791503e-06, | |
| "loss": 0.189, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.148129062351249e-06, | |
| "loss": 0.1848, | |
| "step": 471 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1449318593072465e-06, | |
| "loss": 0.1885, | |
| "step": 472 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1417331422464205e-06, | |
| "loss": 0.1818, | |
| "step": 473 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.138532944583862e-06, | |
| "loss": 0.19, | |
| "step": 474 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1353312997501312e-06, | |
| "loss": 0.1876, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1321282411909043e-06, | |
| "loss": 0.2047, | |
| "step": 476 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1289238023666266e-06, | |
| "loss": 0.1931, | |
| "step": 477 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1257180167521627e-06, | |
| "loss": 0.1842, | |
| "step": 478 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1225109178364454e-06, | |
| "loss": 0.1869, | |
| "step": 479 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1193025391221274e-06, | |
| "loss": 0.1805, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1160929141252301e-06, | |
| "loss": 0.1938, | |
| "step": 481 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1128820763747949e-06, | |
| "loss": 0.1844, | |
| "step": 482 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1096700594125316e-06, | |
| "loss": 0.1883, | |
| "step": 483 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1064568967924682e-06, | |
| "loss": 0.1925, | |
| "step": 484 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1032426220806016e-06, | |
| "loss": 0.184, | |
| "step": 485 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.100027268854545e-06, | |
| "loss": 0.1911, | |
| "step": 486 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.096810870703179e-06, | |
| "loss": 0.1895, | |
| "step": 487 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0935934612263e-06, | |
| "loss": 0.1935, | |
| "step": 488 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0903750740342682e-06, | |
| "loss": 0.1889, | |
| "step": 489 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0871557427476583e-06, | |
| "loss": 0.1863, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0839355009969067e-06, | |
| "loss": 0.1838, | |
| "step": 491 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.080714382421962e-06, | |
| "loss": 0.181, | |
| "step": 492 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.077492420671931e-06, | |
| "loss": 0.189, | |
| "step": 493 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0742696494047294e-06, | |
| "loss": 0.1864, | |
| "step": 494 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0710461022867301e-06, | |
| "loss": 0.1947, | |
| "step": 495 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0678218129924093e-06, | |
| "loss": 0.177, | |
| "step": 496 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0645968152039979e-06, | |
| "loss": 0.1959, | |
| "step": 497 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0613711426111263e-06, | |
| "loss": 0.1771, | |
| "step": 498 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0581448289104758e-06, | |
| "loss": 0.1909, | |
| "step": 499 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0549179078054243e-06, | |
| "loss": 0.1757, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0516904130056945e-06, | |
| "loss": 0.1872, | |
| "step": 501 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.048462378227003e-06, | |
| "loss": 0.1836, | |
| "step": 502 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0452338371907063e-06, | |
| "loss": 0.1725, | |
| "step": 503 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0420048236234505e-06, | |
| "loss": 0.1727, | |
| "step": 504 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0387753712568168e-06, | |
| "loss": 0.1907, | |
| "step": 505 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0355455138269709e-06, | |
| "loss": 0.1824, | |
| "step": 506 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0323152850743107e-06, | |
| "loss": 0.1867, | |
| "step": 507 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0290847187431114e-06, | |
| "loss": 0.1891, | |
| "step": 508 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0258538485811763e-06, | |
| "loss": 0.1933, | |
| "step": 509 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.022622708339481e-06, | |
| "loss": 0.1812, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0193913317718244e-06, | |
| "loss": 0.1877, | |
| "step": 511 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0161597526344728e-06, | |
| "loss": 0.2032, | |
| "step": 512 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0129280046858085e-06, | |
| "loss": 0.1905, | |
| "step": 513 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0096961216859786e-06, | |
| "loss": 0.1782, | |
| "step": 514 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0064641373965392e-06, | |
| "loss": 0.1853, | |
| "step": 515 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0032320855801062e-06, | |
| "loss": 0.1861, | |
| "step": 516 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1e-06, | |
| "loss": 0.1873, | |
| "step": 517 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.96767914419894e-07, | |
| "loss": 0.1881, | |
| "step": 518 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.935358626034605e-07, | |
| "loss": 0.1864, | |
| "step": 519 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.903038783140215e-07, | |
| "loss": 0.1905, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.870719953141916e-07, | |
| "loss": 0.1844, | |
| "step": 521 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.838402473655275e-07, | |
| "loss": 0.1985, | |
| "step": 522 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.806086682281757e-07, | |
| "loss": 0.1915, | |
| "step": 523 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.773772916605189e-07, | |
| "loss": 0.1765, | |
| "step": 524 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.74146151418824e-07, | |
| "loss": 0.182, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.709152812568885e-07, | |
| "loss": 0.1841, | |
| "step": 526 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.676847149256894e-07, | |
| "loss": 0.1887, | |
| "step": 527 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.644544861730292e-07, | |
| "loss": 0.1789, | |
| "step": 528 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.61224628743183e-07, | |
| "loss": 0.1842, | |
| "step": 529 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.579951763765496e-07, | |
| "loss": 0.1736, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.547661628092936e-07, | |
| "loss": 0.168, | |
| "step": 531 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.515376217729971e-07, | |
| "loss": 0.1865, | |
| "step": 532 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.483095869943054e-07, | |
| "loss": 0.1907, | |
| "step": 533 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.450820921945757e-07, | |
| "loss": 0.1866, | |
| "step": 534 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.418551710895241e-07, | |
| "loss": 0.1782, | |
| "step": 535 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.386288573888735e-07, | |
| "loss": 0.1836, | |
| "step": 536 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.354031847960023e-07, | |
| "loss": 0.1885, | |
| "step": 537 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.321781870075908e-07, | |
| "loss": 0.1781, | |
| "step": 538 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.289538977132701e-07, | |
| "loss": 0.1811, | |
| "step": 539 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.257303505952704e-07, | |
| "loss": 0.1906, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.225075793280691e-07, | |
| "loss": 0.1783, | |
| "step": 541 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.192856175780381e-07, | |
| "loss": 0.1754, | |
| "step": 542 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.16064499003093e-07, | |
| "loss": 0.1954, | |
| "step": 543 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.128442572523417e-07, | |
| "loss": 0.1867, | |
| "step": 544 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.096249259657321e-07, | |
| "loss": 0.2066, | |
| "step": 545 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.064065387737003e-07, | |
| "loss": 0.1894, | |
| "step": 546 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 9.031891292968209e-07, | |
| "loss": 0.1939, | |
| "step": 547 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.99972731145455e-07, | |
| "loss": 0.187, | |
| "step": 548 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.967573779193987e-07, | |
| "loss": 0.191, | |
| "step": 549 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.935431032075317e-07, | |
| "loss": 0.1805, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.903299405874683e-07, | |
| "loss": 0.1908, | |
| "step": 551 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.871179236252052e-07, | |
| "loss": 0.1862, | |
| "step": 552 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.839070858747696e-07, | |
| "loss": 0.1863, | |
| "step": 553 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.806974608778727e-07, | |
| "loss": 0.1831, | |
| "step": 554 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.774890821635547e-07, | |
| "loss": 0.1948, | |
| "step": 555 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.742819832478374e-07, | |
| "loss": 0.1827, | |
| "step": 556 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.710761976333733e-07, | |
| "loss": 0.1838, | |
| "step": 557 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.678717588090959e-07, | |
| "loss": 0.1772, | |
| "step": 558 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.646687002498691e-07, | |
| "loss": 0.1797, | |
| "step": 559 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.614670554161376e-07, | |
| "loss": 0.1866, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.582668577535796e-07, | |
| "loss": 0.1909, | |
| "step": 561 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.550681406927533e-07, | |
| "loss": 0.188, | |
| "step": 562 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.518709376487514e-07, | |
| "loss": 0.1889, | |
| "step": 563 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.486752820208494e-07, | |
| "loss": 0.1884, | |
| "step": 564 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.454812071921595e-07, | |
| "loss": 0.1743, | |
| "step": 565 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.422887465292788e-07, | |
| "loss": 0.1935, | |
| "step": 566 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.390979333819426e-07, | |
| "loss": 0.1786, | |
| "step": 567 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.359088010826759e-07, | |
| "loss": 0.182, | |
| "step": 568 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.327213829464448e-07, | |
| "loss": 0.1866, | |
| "step": 569 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.295357122703079e-07, | |
| "loss": 0.1915, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.263518223330696e-07, | |
| "loss": 0.1816, | |
| "step": 571 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.231697463949326e-07, | |
| "loss": 0.178, | |
| "step": 572 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.199895176971487e-07, | |
| "loss": 0.1766, | |
| "step": 573 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.168111694616733e-07, | |
| "loss": 0.1844, | |
| "step": 574 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.136347348908179e-07, | |
| "loss": 0.193, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.104602471669029e-07, | |
| "loss": 0.1772, | |
| "step": 576 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 8.0728773945191e-07, | |
| "loss": 0.1823, | |
| "step": 577 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 8.041172448871397e-07, | |
| "loss": 0.1833, | |
| "step": 578 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 8.009487965928595e-07, | |
| "loss": 0.1758, | |
| "step": 579 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.977824276679622e-07, | |
| "loss": 0.188, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.946181711896179e-07, | |
| "loss": 0.1909, | |
| "step": 581 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.914560602129303e-07, | |
| "loss": 0.1799, | |
| "step": 582 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.882961277705896e-07, | |
| "loss": 0.1835, | |
| "step": 583 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.85138406872527e-07, | |
| "loss": 0.1869, | |
| "step": 584 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.819829305055742e-07, | |
| "loss": 0.1859, | |
| "step": 585 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.788297316331123e-07, | |
| "loss": 0.173, | |
| "step": 586 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.756788431947325e-07, | |
| "loss": 0.1865, | |
| "step": 587 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.725302981058897e-07, | |
| "loss": 0.1781, | |
| "step": 588 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.693841292575598e-07, | |
| "loss": 0.1778, | |
| "step": 589 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.662403695158952e-07, | |
| "loss": 0.2005, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.630990517218808e-07, | |
| "loss": 0.1844, | |
| "step": 591 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.599602086909942e-07, | |
| "loss": 0.1816, | |
| "step": 592 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.568238732128584e-07, | |
| "loss": 0.1845, | |
| "step": 593 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.536900780509027e-07, | |
| "loss": 0.1772, | |
| "step": 594 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.505588559420187e-07, | |
| "loss": 0.1864, | |
| "step": 595 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.474302395962194e-07, | |
| "loss": 0.1829, | |
| "step": 596 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.443042616962965e-07, | |
| "loss": 0.1893, | |
| "step": 597 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.411809548974791e-07, | |
| "loss": 0.1738, | |
| "step": 598 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.38060351827094e-07, | |
| "loss": 0.1801, | |
| "step": 599 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.34942485084223e-07, | |
| "loss": 0.1853, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.318273872393625e-07, | |
| "loss": 0.1724, | |
| "step": 601 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.287150908340855e-07, | |
| "loss": 0.1924, | |
| "step": 602 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.256056283806986e-07, | |
| "loss": 0.1844, | |
| "step": 603 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.224990323619043e-07, | |
| "loss": 0.1874, | |
| "step": 604 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.193953352304611e-07, | |
| "loss": 0.1754, | |
| "step": 605 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.16294569408845e-07, | |
| "loss": 0.1876, | |
| "step": 606 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.1319676728891e-07, | |
| "loss": 0.1835, | |
| "step": 607 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.101019612315491e-07, | |
| "loss": 0.1734, | |
| "step": 608 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.070101835663593e-07, | |
| "loss": 0.1856, | |
| "step": 609 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.039214665913002e-07, | |
| "loss": 0.1945, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.008358425723585e-07, | |
| "loss": 0.1807, | |
| "step": 611 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 6.977533437432108e-07, | |
| "loss": 0.1768, | |
| "step": 612 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 6.946740023048869e-07, | |
| "loss": 0.1964, | |
| "step": 613 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 6.915978504254331e-07, | |
| "loss": 0.1883, | |
| "step": 614 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 6.885249202395754e-07, | |
| "loss": 0.1847, | |
| "step": 615 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 6.854552438483865e-07, | |
| "loss": 0.1813, | |
| "step": 616 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.823888533189468e-07, | |
| "loss": 0.1785, | |
| "step": 617 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.793257806840123e-07, | |
| "loss": 0.1894, | |
| "step": 618 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.76266057941679e-07, | |
| "loss": 0.1747, | |
| "step": 619 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.732097170550479e-07, | |
| "loss": 0.1782, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.701567899518923e-07, | |
| "loss": 0.1811, | |
| "step": 621 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.671073085243233e-07, | |
| "loss": 0.182, | |
| "step": 622 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.640613046284579e-07, | |
| "loss": 0.1781, | |
| "step": 623 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.610188100840844e-07, | |
| "loss": 0.1833, | |
| "step": 624 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.579798566743313e-07, | |
| "loss": 0.1897, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.549444761453354e-07, | |
| "loss": 0.1875, | |
| "step": 626 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.519127002059095e-07, | |
| "loss": 0.1768, | |
| "step": 627 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.488845605272112e-07, | |
| "loss": 0.1802, | |
| "step": 628 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.458600887424124e-07, | |
| "loss": 0.1881, | |
| "step": 629 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.428393164463688e-07, | |
| "loss": 0.1821, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.398222751952897e-07, | |
| "loss": 0.1856, | |
| "step": 631 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.368089965064079e-07, | |
| "loss": 0.1829, | |
| "step": 632 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.33799511857652e-07, | |
| "loss": 0.1876, | |
| "step": 633 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.307938526873157e-07, | |
| "loss": 0.1836, | |
| "step": 634 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.277920503937301e-07, | |
| "loss": 0.1748, | |
| "step": 635 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.247941363349364e-07, | |
| "loss": 0.171, | |
| "step": 636 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.218001418283576e-07, | |
| "loss": 0.1954, | |
| "step": 637 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.188100981504712e-07, | |
| "loss": 0.1943, | |
| "step": 638 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.158240365364823e-07, | |
| "loss": 0.1915, | |
| "step": 639 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.128419881799995e-07, | |
| "loss": 0.1845, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.098639842327051e-07, | |
| "loss": 0.1808, | |
| "step": 641 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.068900558040332e-07, | |
| "loss": 0.1814, | |
| "step": 642 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.039202339608431e-07, | |
| "loss": 0.1845, | |
| "step": 643 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.009545497270951e-07, | |
| "loss": 0.1786, | |
| "step": 644 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 5.979930340835264e-07, | |
| "loss": 0.1937, | |
| "step": 645 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 5.950357179673263e-07, | |
| "loss": 0.1846, | |
| "step": 646 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.920826322718164e-07, | |
| "loss": 0.1913, | |
| "step": 647 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.891338078461238e-07, | |
| "loss": 0.1796, | |
| "step": 648 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.861892754948608e-07, | |
| "loss": 0.2015, | |
| "step": 649 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.83249065977804e-07, | |
| "loss": 0.1937, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.803132100095709e-07, | |
| "loss": 0.1752, | |
| "step": 651 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.773817382593007e-07, | |
| "loss": 0.1756, | |
| "step": 652 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.744546813503327e-07, | |
| "loss": 0.1839, | |
| "step": 653 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.715320698598874e-07, | |
| "loss": 0.1823, | |
| "step": 654 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.686139343187467e-07, | |
| "loss": 0.1778, | |
| "step": 655 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.657003052109344e-07, | |
| "loss": 0.1723, | |
| "step": 656 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.627912129733992e-07, | |
| "loss": 0.1743, | |
| "step": 657 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.598866879956954e-07, | |
| "loss": 0.1765, | |
| "step": 658 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.569867606196651e-07, | |
| "loss": 0.1845, | |
| "step": 659 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.540914611391226e-07, | |
| "loss": 0.1864, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.512008197995378e-07, | |
| "loss": 0.1829, | |
| "step": 661 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.483148667977194e-07, | |
| "loss": 0.1872, | |
| "step": 662 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.454336322814995e-07, | |
| "loss": 0.1853, | |
| "step": 663 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.425571463494193e-07, | |
| "loss": 0.1927, | |
| "step": 664 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.396854390504145e-07, | |
| "loss": 0.1938, | |
| "step": 665 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.368185403835004e-07, | |
| "loss": 0.1822, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.339564802974614e-07, | |
| "loss": 0.1803, | |
| "step": 667 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.310992886905337e-07, | |
| "loss": 0.1779, | |
| "step": 668 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.282469954100968e-07, | |
| "loss": 0.1815, | |
| "step": 669 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.253996302523596e-07, | |
| "loss": 0.197, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.225572229620509e-07, | |
| "loss": 0.1777, | |
| "step": 671 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.197198032321069e-07, | |
| "loss": 0.1803, | |
| "step": 672 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.168874007033615e-07, | |
| "loss": 0.1901, | |
| "step": 673 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.140600449642373e-07, | |
| "loss": 0.1878, | |
| "step": 674 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.112377655504358e-07, | |
| "loss": 0.1866, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.084205919446295e-07, | |
| "loss": 0.1778, | |
| "step": 676 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.056085535761532e-07, | |
| "loss": 0.181, | |
| "step": 677 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 5.028016798206973e-07, | |
| "loss": 0.1676, | |
| "step": 678 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 5.000000000000002e-07, | |
| "loss": 0.1891, | |
| "step": 679 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.972035433815421e-07, | |
| "loss": 0.1739, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.94412339178241e-07, | |
| "loss": 0.1855, | |
| "step": 681 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.916264165481448e-07, | |
| "loss": 0.1832, | |
| "step": 682 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.888458045941268e-07, | |
| "loss": 0.1776, | |
| "step": 683 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.860705323635856e-07, | |
| "loss": 0.1954, | |
| "step": 684 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.83300628848137e-07, | |
| "loss": 0.1851, | |
| "step": 685 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.805361229833136e-07, | |
| "loss": 0.187, | |
| "step": 686 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.777770436482617e-07, | |
| "loss": 0.1719, | |
| "step": 687 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.750234196654399e-07, | |
| "loss": 0.1817, | |
| "step": 688 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.7227527980031836e-07, | |
| "loss": 0.1791, | |
| "step": 689 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.6953265276107724e-07, | |
| "loss": 0.1758, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.667955671983089e-07, | |
| "loss": 0.172, | |
| "step": 691 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.640640517047153e-07, | |
| "loss": 0.1856, | |
| "step": 692 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.6133813481481243e-07, | |
| "loss": 0.1809, | |
| "step": 693 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.586178450046303e-07, | |
| "loss": 0.176, | |
| "step": 694 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.559032106914172e-07, | |
| "loss": 0.1847, | |
| "step": 695 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.5319426023334153e-07, | |
| "loss": 0.1813, | |
| "step": 696 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.50491021929194e-07, | |
| "loss": 0.1796, | |
| "step": 697 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.477935240180966e-07, | |
| "loss": 0.1813, | |
| "step": 698 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.451017946792032e-07, | |
| "loss": 0.1788, | |
| "step": 699 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.424158620314072e-07, | |
| "loss": 0.1751, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.3973575413304754e-07, | |
| "loss": 0.1882, | |
| "step": 701 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.3706149898161537e-07, | |
| "loss": 0.2013, | |
| "step": 702 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.3439312451346154e-07, | |
| "loss": 0.1808, | |
| "step": 703 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.3173065860350475e-07, | |
| "loss": 0.1931, | |
| "step": 704 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.2907412906494166e-07, | |
| "loss": 0.1831, | |
| "step": 705 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.2642356364895417e-07, | |
| "loss": 0.1855, | |
| "step": 706 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.237789900444196e-07, | |
| "loss": 0.1852, | |
| "step": 707 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.2114043587762483e-07, | |
| "loss": 0.1801, | |
| "step": 708 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.185079287119733e-07, | |
| "loss": 0.1862, | |
| "step": 709 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.1588149604769983e-07, | |
| "loss": 0.1861, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.1326116532158217e-07, | |
| "loss": 0.1888, | |
| "step": 711 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.106469639066552e-07, | |
| "loss": 0.178, | |
| "step": 712 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.080389191119241e-07, | |
| "loss": 0.1815, | |
| "step": 713 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.054370581820795e-07, | |
| "loss": 0.1797, | |
| "step": 714 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.02841408297214e-07, | |
| "loss": 0.185, | |
| "step": 715 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.002519965725352e-07, | |
| "loss": 0.1868, | |
| "step": 716 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 3.976688500580856e-07, | |
| "loss": 0.1801, | |
| "step": 717 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.950919957384582e-07, | |
| "loss": 0.1875, | |
| "step": 718 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.9252146053251635e-07, | |
| "loss": 0.1806, | |
| "step": 719 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.899572712931103e-07, | |
| "loss": 0.1747, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.8739945480679715e-07, | |
| "loss": 0.1939, | |
| "step": 721 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.8484803779356344e-07, | |
| "loss": 0.1822, | |
| "step": 722 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.82303046906543e-07, | |
| "loss": 0.1776, | |
| "step": 723 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.7976450873174007e-07, | |
| "loss": 0.1822, | |
| "step": 724 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.7723244978775106e-07, | |
| "loss": 0.1768, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.747068965254879e-07, | |
| "loss": 0.1858, | |
| "step": 726 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.721878753279016e-07, | |
| "loss": 0.1903, | |
| "step": 727 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.6967541250970604e-07, | |
| "loss": 0.1924, | |
| "step": 728 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.671695343171052e-07, | |
| "loss": 0.175, | |
| "step": 729 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.646702669275151e-07, | |
| "loss": 0.1922, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.621776364492939e-07, | |
| "loss": 0.178, | |
| "step": 731 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.5969166892146843e-07, | |
| "loss": 0.1707, | |
| "step": 732 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.5721239031346063e-07, | |
| "loss": 0.1869, | |
| "step": 733 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.547398265248177e-07, | |
| "loss": 0.1763, | |
| "step": 734 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.522740033849411e-07, | |
| "loss": 0.1773, | |
| "step": 735 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.4981494665281663e-07, | |
| "loss": 0.1756, | |
| "step": 736 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.473626820167457e-07, | |
| "loss": 0.1724, | |
| "step": 737 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.44917235094076e-07, | |
| "loss": 0.1759, | |
| "step": 738 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.4247863143093647e-07, | |
| "loss": 0.1819, | |
| "step": 739 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.400468965019663e-07, | |
| "loss": 0.1804, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.3762205571005233e-07, | |
| "loss": 0.1933, | |
| "step": 741 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.352041343860621e-07, | |
| "loss": 0.1888, | |
| "step": 742 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.327931577885803e-07, | |
| "loss": 0.1871, | |
| "step": 743 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.303891511036437e-07, | |
| "loss": 0.185, | |
| "step": 744 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.2799213944447757e-07, | |
| "loss": 0.1947, | |
| "step": 745 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.256021478512362e-07, | |
| "loss": 0.1871, | |
| "step": 746 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.232192012907381e-07, | |
| "loss": 0.1892, | |
| "step": 747 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.208433246562069e-07, | |
| "loss": 0.1874, | |
| "step": 748 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.184745427670109e-07, | |
| "loss": 0.1755, | |
| "step": 749 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.161128803684039e-07, | |
| "loss": 0.1881, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.137583621312665e-07, | |
| "loss": 0.1842, | |
| "step": 751 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.114110126518483e-07, | |
| "loss": 0.1795, | |
| "step": 752 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.0907085645151243e-07, | |
| "loss": 0.1789, | |
| "step": 753 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.0673791797647596e-07, | |
| "loss": 0.1829, | |
| "step": 754 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.04412221597558e-07, | |
| "loss": 0.1856, | |
| "step": 755 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.020937916099241e-07, | |
| "loss": 0.1844, | |
| "step": 756 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 2.997826522328315e-07, | |
| "loss": 0.1988, | |
| "step": 757 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.974788276093768e-07, | |
| "loss": 0.1946, | |
| "step": 758 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.951823418062439e-07, | |
| "loss": 0.1891, | |
| "step": 759 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.9289321881345254e-07, | |
| "loss": 0.1859, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.906114825441072e-07, | |
| "loss": 0.1838, | |
| "step": 761 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.8833715683414794e-07, | |
| "loss": 0.1852, | |
| "step": 762 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.860702654421011e-07, | |
| "loss": 0.1722, | |
| "step": 763 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.838108320488309e-07, | |
| "loss": 0.1746, | |
| "step": 764 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.8155888025729267e-07, | |
| "loss": 0.1934, | |
| "step": 765 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.7931443359228537e-07, | |
| "loss": 0.1847, | |
| "step": 766 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.770775155002071e-07, | |
| "loss": 0.1873, | |
| "step": 767 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.748481493488092e-07, | |
| "loss": 0.1865, | |
| "step": 768 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.7262635842695124e-07, | |
| "loss": 0.1872, | |
| "step": 769 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.704121659443609e-07, | |
| "loss": 0.184, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.6820559503138797e-07, | |
| "loss": 0.1811, | |
| "step": 771 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.6600666873876474e-07, | |
| "loss": 0.1898, | |
| "step": 772 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.6381541003736486e-07, | |
| "loss": 0.1843, | |
| "step": 773 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.6163184181796283e-07, | |
| "loss": 0.1743, | |
| "step": 774 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.5945598689099556e-07, | |
| "loss": 0.1894, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.5728786798632343e-07, | |
| "loss": 0.186, | |
| "step": 776 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.551275077529943e-07, | |
| "loss": 0.1885, | |
| "step": 777 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.5297492875900415e-07, | |
| "loss": 0.1884, | |
| "step": 778 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.5083015349106397e-07, | |
| "loss": 0.1819, | |
| "step": 779 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.4869320435436403e-07, | |
| "loss": 0.186, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.4656410367233924e-07, | |
| "loss": 0.1782, | |
| "step": 781 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.444428736864368e-07, | |
| "loss": 0.1723, | |
| "step": 782 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.423295365558821e-07, | |
| "loss": 0.1899, | |
| "step": 783 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.402241143574507e-07, | |
| "loss": 0.1761, | |
| "step": 784 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.3812662908523407e-07, | |
| "loss": 0.1866, | |
| "step": 785 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.3603710265041176e-07, | |
| "loss": 0.1754, | |
| "step": 786 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.339555568810221e-07, | |
| "loss": 0.1988, | |
| "step": 787 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.3188201352173408e-07, | |
| "loss": 0.1872, | |
| "step": 788 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.298164942336205e-07, | |
| "loss": 0.1845, | |
| "step": 789 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.277590205939308e-07, | |
| "loss": 0.1793, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.2570961409586754e-07, | |
| "loss": 0.1878, | |
| "step": 791 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.2366829614835991e-07, | |
| "loss": 0.1876, | |
| "step": 792 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.2163508807583998e-07, | |
| "loss": 0.1844, | |
| "step": 793 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.1961001111802245e-07, | |
| "loss": 0.1798, | |
| "step": 794 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.1759308642968023e-07, | |
| "loss": 0.1791, | |
| "step": 795 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.1558433508042427e-07, | |
| "loss": 0.1833, | |
| "step": 796 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.135837780544838e-07, | |
| "loss": 0.1814, | |
| "step": 797 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.115914362504865e-07, | |
| "loss": 0.1923, | |
| "step": 798 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.096073304812408e-07, | |
| "loss": 0.1866, | |
| "step": 799 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.0763148147351782e-07, | |
| "loss": 0.1842, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.0566390986783644e-07, | |
| "loss": 0.1742, | |
| "step": 801 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.037046362182444e-07, | |
| "loss": 0.1907, | |
| "step": 802 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.0175368099210698e-07, | |
| "loss": 0.1848, | |
| "step": 803 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.9981106456989206e-07, | |
| "loss": 0.1717, | |
| "step": 804 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.9787680724495614e-07, | |
| "loss": 0.1834, | |
| "step": 805 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.9595092922333412e-07, | |
| "loss": 0.195, | |
| "step": 806 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.9403345062352572e-07, | |
| "loss": 0.1853, | |
| "step": 807 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.921243914762889e-07, | |
| "loss": 0.1736, | |
| "step": 808 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.9022377172442748e-07, | |
| "loss": 0.1795, | |
| "step": 809 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8833161122258402e-07, | |
| "loss": 0.1701, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.864479297370325e-07, | |
| "loss": 0.1818, | |
| "step": 811 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8457274694547175e-07, | |
| "loss": 0.1846, | |
| "step": 812 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.827060824368195e-07, | |
| "loss": 0.1904, | |
| "step": 813 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8084795571100809e-07, | |
| "loss": 0.1827, | |
| "step": 814 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.789983861787816e-07, | |
| "loss": 0.1823, | |
| "step": 815 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.771573931614908e-07, | |
| "loss": 0.1843, | |
| "step": 816 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.753249958908932e-07, | |
| "loss": 0.1793, | |
| "step": 817 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.7350121350895252e-07, | |
| "loss": 0.1807, | |
| "step": 818 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.7168606506763695e-07, | |
| "loss": 0.1861, | |
| "step": 819 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.6987956952872118e-07, | |
| "loss": 0.176, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.6808174576358846e-07, | |
| "loss": 0.1942, | |
| "step": 821 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.6629261255303283e-07, | |
| "loss": 0.1861, | |
| "step": 822 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.6451218858706372e-07, | |
| "loss": 0.1726, | |
| "step": 823 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.6274049246470956e-07, | |
| "loss": 0.1862, | |
| "step": 824 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.6097754269382534e-07, | |
| "loss": 0.1757, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.59223357690897e-07, | |
| "loss": 0.1923, | |
| "step": 826 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.5747795578085042e-07, | |
| "loss": 0.1853, | |
| "step": 827 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.557413551968604e-07, | |
| "loss": 0.1788, | |
| "step": 828 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5401357408015892e-07, | |
| "loss": 0.1785, | |
| "step": 829 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5229463047984648e-07, | |
| "loss": 0.1985, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.505845423527027e-07, | |
| "loss": 0.1821, | |
| "step": 831 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.4888332756300027e-07, | |
| "loss": 0.1845, | |
| "step": 832 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.4719100388231697e-07, | |
| "loss": 0.1711, | |
| "step": 833 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.4550758898935044e-07, | |
| "loss": 0.1682, | |
| "step": 834 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.4383310046973362e-07, | |
| "loss": 0.1864, | |
| "step": 835 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.421675558158507e-07, | |
| "loss": 0.1813, | |
| "step": 836 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.40510972426655e-07, | |
| "loss": 0.1894, | |
| "step": 837 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.3886336760748618e-07, | |
| "loss": 0.1742, | |
| "step": 838 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.3722475856989158e-07, | |
| "loss": 0.1726, | |
| "step": 839 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.355951624314432e-07, | |
| "loss": 0.1771, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.3397459621556128e-07, | |
| "loss": 0.1761, | |
| "step": 841 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.3236307685133674e-07, | |
| "loss": 0.1815, | |
| "step": 842 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.307606211733522e-07, | |
| "loss": 0.1868, | |
| "step": 843 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.2916724592150797e-07, | |
| "loss": 0.1953, | |
| "step": 844 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.2758296774084633e-07, | |
| "loss": 0.1817, | |
| "step": 845 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.2600780318137805e-07, | |
| "loss": 0.1758, | |
| "step": 846 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.2444176869790923e-07, | |
| "loss": 0.1783, | |
| "step": 847 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.2288488064986945e-07, | |
| "loss": 0.1914, | |
| "step": 848 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.2133715530114107e-07, | |
| "loss": 0.1792, | |
| "step": 849 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.1979860881988902e-07, | |
| "loss": 0.1895, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.18269257278392e-07, | |
| "loss": 0.1859, | |
| "step": 851 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.1674911665287512e-07, | |
| "loss": 0.1969, | |
| "step": 852 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.1523820282334218e-07, | |
| "loss": 0.1753, | |
| "step": 853 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.1373653157341035e-07, | |
| "loss": 0.1852, | |
| "step": 854 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.1224411859014416e-07, | |
| "loss": 0.187, | |
| "step": 855 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.1076097946389395e-07, | |
| "loss": 0.1841, | |
| "step": 856 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.092871296881307e-07, | |
| "loss": 0.1763, | |
| "step": 857 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.0782258465928539e-07, | |
| "loss": 0.1741, | |
| "step": 858 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.0636735967658784e-07, | |
| "loss": 0.1878, | |
| "step": 859 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.0492146994190664e-07, | |
| "loss": 0.1808, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.0348493055959062e-07, | |
| "loss": 0.1838, | |
| "step": 861 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.0205775653631176e-07, | |
| "loss": 0.1964, | |
| "step": 862 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.0063996278090703e-07, | |
| "loss": 0.1737, | |
| "step": 863 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 9.923156410422317e-08, | |
| "loss": 0.1721, | |
| "step": 864 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 9.783257521896227e-08, | |
| "loss": 0.1722, | |
| "step": 865 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 9.644301073952865e-08, | |
| "loss": 0.1821, | |
| "step": 866 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 9.506288518187466e-08, | |
| "loss": 0.1734, | |
| "step": 867 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 9.369221296335006e-08, | |
| "loss": 0.1892, | |
| "step": 868 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 9.233100840255125e-08, | |
| "loss": 0.1888, | |
| "step": 869 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 9.09792857191718e-08, | |
| "loss": 0.1803, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.963705903385343e-08, | |
| "loss": 0.1798, | |
| "step": 871 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.830434236803908e-08, | |
| "loss": 0.181, | |
| "step": 872 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.698114964382597e-08, | |
| "loss": 0.175, | |
| "step": 873 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.566749468382072e-08, | |
| "loss": 0.1858, | |
| "step": 874 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.436339121099412e-08, | |
| "loss": 0.1756, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.30688528485387e-08, | |
| "loss": 0.1829, | |
| "step": 876 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.178389311972612e-08, | |
| "loss": 0.1823, | |
| "step": 877 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 8.050852544776488e-08, | |
| "loss": 0.1753, | |
| "step": 878 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.92427631556617e-08, | |
| "loss": 0.1734, | |
| "step": 879 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.798661946608164e-08, | |
| "loss": 0.1929, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.674010750120963e-08, | |
| "loss": 0.1833, | |
| "step": 881 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.55032402826139e-08, | |
| "loss": 0.1907, | |
| "step": 882 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.427603073110966e-08, | |
| "loss": 0.1816, | |
| "step": 883 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.305849166662426e-08, | |
| "loss": 0.1858, | |
| "step": 884 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.185063580806317e-08, | |
| "loss": 0.1866, | |
| "step": 885 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.065247577317745e-08, | |
| "loss": 0.1867, | |
| "step": 886 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 6.946402407843155e-08, | |
| "loss": 0.1799, | |
| "step": 887 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.828529313887211e-08, | |
| "loss": 0.1746, | |
| "step": 888 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.711629526799944e-08, | |
| "loss": 0.1706, | |
| "step": 889 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.595704267763846e-08, | |
| "loss": 0.1814, | |
| "step": 890 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.480754747781036e-08, | |
| "loss": 0.1745, | |
| "step": 891 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.36678216766069e-08, | |
| "loss": 0.1837, | |
| "step": 892 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.253787718006498e-08, | |
| "loss": 0.1938, | |
| "step": 893 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.141772579204174e-08, | |
| "loss": 0.1835, | |
| "step": 894 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.030737921409168e-08, | |
| "loss": 0.1913, | |
| "step": 895 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 5.920684904534423e-08, | |
| "loss": 0.1928, | |
| "step": 896 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 5.811614678238275e-08, | |
| "loss": 0.1749, | |
| "step": 897 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.7035283819124146e-08, | |
| "loss": 0.1939, | |
| "step": 898 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.596427144670002e-08, | |
| "loss": 0.1806, | |
| "step": 899 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.490312085333904e-08, | |
| "loss": 0.1842, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.385184312424973e-08, | |
| "loss": 0.1925, | |
| "step": 901 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.2810449241503905e-08, | |
| "loss": 0.183, | |
| "step": 902 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.177895008392352e-08, | |
| "loss": 0.1866, | |
| "step": 903 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.0757356426966105e-08, | |
| "loss": 0.1812, | |
| "step": 904 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 4.974567894261217e-08, | |
| "loss": 0.1863, | |
| "step": 905 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 4.874392819925388e-08, | |
| "loss": 0.1779, | |
| "step": 906 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 4.7752114661584685e-08, | |
| "loss": 0.182, | |
| "step": 907 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.677024869049007e-08, | |
| "loss": 0.1998, | |
| "step": 908 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.579834054293885e-08, | |
| "loss": 0.187, | |
| "step": 909 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.483640037187697e-08, | |
| "loss": 0.1706, | |
| "step": 910 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.3884438226120427e-08, | |
| "loss": 0.194, | |
| "step": 911 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.2942464050250703e-08, | |
| "loss": 0.1796, | |
| "step": 912 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.20104876845111e-08, | |
| "loss": 0.185, | |
| "step": 913 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.108851886470388e-08, | |
| "loss": 0.1882, | |
| "step": 914 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.017656722208806e-08, | |
| "loss": 0.1889, | |
| "step": 915 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 3.927464228327948e-08, | |
| "loss": 0.1729, | |
| "step": 916 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 3.838275347015085e-08, | |
| "loss": 0.1754, | |
| "step": 917 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.750091009973344e-08, | |
| "loss": 0.1958, | |
| "step": 918 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.6629121384119666e-08, | |
| "loss": 0.1856, | |
| "step": 919 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.57673964303673e-08, | |
| "loss": 0.1791, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.491574424040355e-08, | |
| "loss": 0.1845, | |
| "step": 921 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.4074173710931796e-08, | |
| "loss": 0.1858, | |
| "step": 922 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.324269363333798e-08, | |
| "loss": 0.1882, | |
| "step": 923 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.24213126935996e-08, | |
| "loss": 0.1865, | |
| "step": 924 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.1610039472194205e-08, | |
| "loss": 0.1873, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.0808882444009586e-08, | |
| "loss": 0.1772, | |
| "step": 926 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.0017849978256515e-08, | |
| "loss": 0.1749, | |
| "step": 927 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.9236950338380027e-08, | |
| "loss": 0.1793, | |
| "step": 928 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.8466191681973506e-08, | |
| "loss": 0.1869, | |
| "step": 929 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.7705582060693622e-08, | |
| "loss": 0.1745, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.6955129420176193e-08, | |
| "loss": 0.1884, | |
| "step": 931 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.621484159995302e-08, | |
| "loss": 0.1848, | |
| "step": 932 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.5484726333370067e-08, | |
| "loss": 0.1827, | |
| "step": 933 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.4764791247506968e-08, | |
| "loss": 0.1782, | |
| "step": 934 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.4055043863096424e-08, | |
| "loss": 0.1808, | |
| "step": 935 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.335549159444705e-08, | |
| "loss": 0.1832, | |
| "step": 936 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.266614174936443e-08, | |
| "loss": 0.1856, | |
| "step": 937 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.1987001529076288e-08, | |
| "loss": 0.1926, | |
| "step": 938 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.1318078028155882e-08, | |
| "loss": 0.1769, | |
| "step": 939 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.065937823444852e-08, | |
| "loss": 0.1825, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.0010909028998825e-08, | |
| "loss": 0.1759, | |
| "step": 941 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.937267718597846e-08, | |
| "loss": 0.1877, | |
| "step": 942 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.8744689372615306e-08, | |
| "loss": 0.1757, | |
| "step": 943 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.8126952149124186e-08, | |
| "loss": 0.1896, | |
| "step": 944 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.751947196863779e-08, | |
| "loss": 0.1929, | |
| "step": 945 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.6922255177140076e-08, | |
| "loss": 0.1874, | |
| "step": 946 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.6335308013398886e-08, | |
| "loss": 0.1797, | |
| "step": 947 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.5758636608901978e-08, | |
| "loss": 0.1743, | |
| "step": 948 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.519224698779198e-08, | |
| "loss": 0.1922, | |
| "step": 949 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.4636145066803885e-08, | |
| "loss": 0.1956, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.4090336655203538e-08, | |
| "loss": 0.1821, | |
| "step": 951 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.3554827454726136e-08, | |
| "loss": 0.1816, | |
| "step": 952 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.3029623059517492e-08, | |
| "loss": 0.1799, | |
| "step": 953 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.2514728956075083e-08, | |
| "loss": 0.1696, | |
| "step": 954 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.2010150523190988e-08, | |
| "loss": 0.1913, | |
| "step": 955 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.1515893031895817e-08, | |
| "loss": 0.1808, | |
| "step": 956 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.1031961645402877e-08, | |
| "loss": 0.1823, | |
| "step": 957 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.0558361419055529e-08, | |
| "loss": 0.1775, | |
| "step": 958 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.0095097300273026e-08, | |
| "loss": 0.1817, | |
| "step": 959 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 9.64217412849988e-09, | |
| "loss": 0.1885, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 9.199596635154683e-09, | |
| "loss": 0.1857, | |
| "step": 961 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 8.76736944358103e-09, | |
| "loss": 0.1858, | |
| "step": 962 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 8.345497068998896e-09, | |
| "loss": 0.1915, | |
| "step": 963 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 7.933983918457677e-09, | |
| "loss": 0.173, | |
| "step": 964 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 7.532834290790435e-09, | |
| "loss": 0.1787, | |
| "step": 965 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 7.142052376568175e-09, | |
| "loss": 0.1761, | |
| "step": 966 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 6.761642258056976e-09, | |
| "loss": 0.1797, | |
| "step": 967 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 6.391607909174812e-09, | |
| "loss": 0.1908, | |
| "step": 968 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 6.031953195450134e-09, | |
| "loss": 0.1814, | |
| "step": 969 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 5.682681873981576e-09, | |
| "loss": 0.1936, | |
| "step": 970 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 5.343797593398536e-09, | |
| "loss": 0.1837, | |
| "step": 971 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 5.015303893823431e-09, | |
| "loss": 0.1754, | |
| "step": 972 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 4.697204206834171e-09, | |
| "loss": 0.1782, | |
| "step": 973 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 4.389501855428857e-09, | |
| "loss": 0.1836, | |
| "step": 974 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 4.092200053990691e-09, | |
| "loss": 0.1737, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 3.805301908254455e-09, | |
| "loss": 0.1838, | |
| "step": 976 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 3.5288104152743082e-09, | |
| "loss": 0.1779, | |
| "step": 977 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 3.262728463392261e-09, | |
| "loss": 0.1849, | |
| "step": 978 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 3.007058832207976e-09, | |
| "loss": 0.1935, | |
| "step": 979 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.761804192549677e-09, | |
| "loss": 0.195, | |
| "step": 980 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.5269671064467313e-09, | |
| "loss": 0.1853, | |
| "step": 981 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.3025500271023346e-09, | |
| "loss": 0.1816, | |
| "step": 982 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.088555298867978e-09, | |
| "loss": 0.1882, | |
| "step": 983 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.884985157219132e-09, | |
| "loss": 0.184, | |
| "step": 984 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.6918417287318243e-09, | |
| "loss": 0.1775, | |
| "step": 985 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.509127031060431e-09, | |
| "loss": 0.186, | |
| "step": 986 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.3368429729168073e-09, | |
| "loss": 0.1799, | |
| "step": 987 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.174991354049637e-09, | |
| "loss": 0.1823, | |
| "step": 988 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.0235738652264458e-09, | |
| "loss": 0.1761, | |
| "step": 989 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 8.825920882157278e-10, | |
| "loss": 0.1938, | |
| "step": 990 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 7.520474957699585e-10, | |
| "loss": 0.1741, | |
| "step": 991 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 6.31941451610829e-10, | |
| "loss": 0.1855, | |
| "step": 992 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 5.222752104147021e-10, | |
| "loss": 0.1783, | |
| "step": 993 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 4.230499177994007e-10, | |
| "loss": 0.1884, | |
| "step": 994 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 3.342666103125502e-10, | |
| "loss": 0.1789, | |
| "step": 995 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 2.5592621542047664e-10, | |
| "loss": 0.1763, | |
| "step": 996 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.8802955149865852e-10, | |
| "loss": 0.1925, | |
| "step": 997 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.3057732782295606e-10, | |
| "loss": 0.1879, | |
| "step": 998 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 8.357014456272793e-11, | |
| "loss": 0.1822, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 4.7008492773836785e-11, | |
| "loss": 0.1809, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 2.0892754394208345e-11, | |
| "loss": 0.191, | |
| "step": 1001 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 5.223202239501568e-12, | |
| "loss": 0.1853, | |
| "step": 1002 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 0.0, | |
| "loss": 0.1821, | |
| "step": 1003 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "step": 1003, | |
| "total_flos": 1949480219836416.0, | |
| "train_loss": 0.24595857647099975, | |
| "train_runtime": 38421.9977, | |
| "train_samples_per_second": 1.044, | |
| "train_steps_per_second": 0.026 | |
| } | |
| ], | |
| "logging_steps": 1.0, | |
| "max_steps": 1003, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 1, | |
| "save_steps": 50000, | |
| "total_flos": 1949480219836416.0, | |
| "train_batch_size": 2, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |