{ "best_metric": null, "best_model_checkpoint": null, "epoch": 1.0, "eval_steps": 500, "global_step": 17629, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 3.780718336483932e-07, "loss": 0.9942, "step": 1 }, { "epoch": 0.0, "learning_rate": 7.561436672967864e-07, "loss": 1.186, "step": 2 }, { "epoch": 0.0, "learning_rate": 1.1342155009451797e-06, "loss": 0.7816, "step": 3 }, { "epoch": 0.0, "learning_rate": 1.5122873345935729e-06, "loss": 1.0956, "step": 4 }, { "epoch": 0.0, "learning_rate": 1.890359168241966e-06, "loss": 1.1652, "step": 5 }, { "epoch": 0.0, "learning_rate": 2.2684310018903594e-06, "loss": 1.1566, "step": 6 }, { "epoch": 0.0, "learning_rate": 2.6465028355387526e-06, "loss": 1.062, "step": 7 }, { "epoch": 0.0, "learning_rate": 3.0245746691871457e-06, "loss": 1.1208, "step": 8 }, { "epoch": 0.0, "learning_rate": 3.402646502835539e-06, "loss": 1.0688, "step": 9 }, { "epoch": 0.0, "learning_rate": 3.780718336483932e-06, "loss": 1.0519, "step": 10 }, { "epoch": 0.0, "learning_rate": 4.158790170132326e-06, "loss": 1.0398, "step": 11 }, { "epoch": 0.0, "learning_rate": 4.536862003780719e-06, "loss": 1.0662, "step": 12 }, { "epoch": 0.0, "learning_rate": 4.914933837429111e-06, "loss": 1.0447, "step": 13 }, { "epoch": 0.0, "learning_rate": 5.293005671077505e-06, "loss": 1.0491, "step": 14 }, { "epoch": 0.0, "learning_rate": 5.671077504725898e-06, "loss": 0.8047, "step": 15 }, { "epoch": 0.0, "learning_rate": 6.0491493383742914e-06, "loss": 1.0296, "step": 16 }, { "epoch": 0.0, "learning_rate": 6.427221172022685e-06, "loss": 0.9716, "step": 17 }, { "epoch": 0.0, "learning_rate": 6.805293005671078e-06, "loss": 1.1077, "step": 18 }, { "epoch": 0.0, "learning_rate": 7.183364839319472e-06, "loss": 0.7845, "step": 19 }, { "epoch": 0.0, "learning_rate": 7.561436672967864e-06, "loss": 1.0398, "step": 20 }, { "epoch": 0.0, "learning_rate": 7.939508506616256e-06, "loss": 0.9949, "step": 21 }, { "epoch": 0.0, "learning_rate": 8.317580340264651e-06, "loss": 1.0673, "step": 22 }, { "epoch": 0.0, "learning_rate": 8.695652173913044e-06, "loss": 1.1504, "step": 23 }, { "epoch": 0.0, "learning_rate": 9.073724007561438e-06, "loss": 1.0377, "step": 24 }, { "epoch": 0.0, "learning_rate": 9.45179584120983e-06, "loss": 1.0069, "step": 25 }, { "epoch": 0.0, "learning_rate": 9.829867674858222e-06, "loss": 0.9581, "step": 26 }, { "epoch": 0.0, "learning_rate": 1.0207939508506617e-05, "loss": 0.9665, "step": 27 }, { "epoch": 0.0, "learning_rate": 1.058601134215501e-05, "loss": 1.0166, "step": 28 }, { "epoch": 0.0, "learning_rate": 1.0964083175803403e-05, "loss": 1.104, "step": 29 }, { "epoch": 0.0, "learning_rate": 1.1342155009451797e-05, "loss": 0.8117, "step": 30 }, { "epoch": 0.0, "learning_rate": 1.172022684310019e-05, "loss": 1.0256, "step": 31 }, { "epoch": 0.0, "learning_rate": 1.2098298676748583e-05, "loss": 1.1432, "step": 32 }, { "epoch": 0.0, "learning_rate": 1.2476370510396976e-05, "loss": 1.1489, "step": 33 }, { "epoch": 0.0, "learning_rate": 1.285444234404537e-05, "loss": 1.0742, "step": 34 }, { "epoch": 0.0, "learning_rate": 1.323251417769376e-05, "loss": 1.0565, "step": 35 }, { "epoch": 0.0, "learning_rate": 1.3610586011342156e-05, "loss": 1.0476, "step": 36 }, { "epoch": 0.0, "learning_rate": 1.3988657844990549e-05, "loss": 1.0618, "step": 37 }, { "epoch": 0.0, "learning_rate": 1.4366729678638944e-05, "loss": 1.1947, "step": 38 }, { "epoch": 0.0, "learning_rate": 1.4744801512287335e-05, "loss": 1.0535, "step": 39 }, { "epoch": 0.0, "learning_rate": 1.5122873345935728e-05, "loss": 1.03, "step": 40 }, { "epoch": 0.0, "learning_rate": 1.5500945179584123e-05, "loss": 0.9602, "step": 41 }, { "epoch": 0.0, "learning_rate": 1.5879017013232513e-05, "loss": 1.0257, "step": 42 }, { "epoch": 0.0, "learning_rate": 1.625708884688091e-05, "loss": 1.0307, "step": 43 }, { "epoch": 0.0, "learning_rate": 1.6635160680529302e-05, "loss": 1.0305, "step": 44 }, { "epoch": 0.0, "learning_rate": 1.7013232514177692e-05, "loss": 0.7618, "step": 45 }, { "epoch": 0.0, "learning_rate": 1.739130434782609e-05, "loss": 1.0671, "step": 46 }, { "epoch": 0.0, "learning_rate": 1.7769376181474482e-05, "loss": 0.9879, "step": 47 }, { "epoch": 0.0, "learning_rate": 1.8147448015122875e-05, "loss": 0.9583, "step": 48 }, { "epoch": 0.0, "learning_rate": 1.8525519848771268e-05, "loss": 0.9303, "step": 49 }, { "epoch": 0.0, "learning_rate": 1.890359168241966e-05, "loss": 0.8626, "step": 50 }, { "epoch": 0.0, "learning_rate": 1.9281663516068055e-05, "loss": 1.0766, "step": 51 }, { "epoch": 0.0, "learning_rate": 1.9659735349716444e-05, "loss": 0.9663, "step": 52 }, { "epoch": 0.0, "learning_rate": 2.003780718336484e-05, "loss": 0.9017, "step": 53 }, { "epoch": 0.0, "learning_rate": 2.0415879017013234e-05, "loss": 0.8976, "step": 54 }, { "epoch": 0.0, "learning_rate": 2.0793950850661624e-05, "loss": 1.0174, "step": 55 }, { "epoch": 0.0, "learning_rate": 2.117202268431002e-05, "loss": 0.9185, "step": 56 }, { "epoch": 0.0, "learning_rate": 2.1550094517958414e-05, "loss": 0.9969, "step": 57 }, { "epoch": 0.0, "learning_rate": 2.1928166351606807e-05, "loss": 0.9531, "step": 58 }, { "epoch": 0.0, "learning_rate": 2.23062381852552e-05, "loss": 1.0973, "step": 59 }, { "epoch": 0.0, "learning_rate": 2.2684310018903593e-05, "loss": 0.9399, "step": 60 }, { "epoch": 0.0, "learning_rate": 2.3062381852551986e-05, "loss": 1.0349, "step": 61 }, { "epoch": 0.0, "learning_rate": 2.344045368620038e-05, "loss": 0.9696, "step": 62 }, { "epoch": 0.0, "learning_rate": 2.3818525519848773e-05, "loss": 0.8665, "step": 63 }, { "epoch": 0.0, "learning_rate": 2.4196597353497166e-05, "loss": 0.7526, "step": 64 }, { "epoch": 0.0, "learning_rate": 2.457466918714556e-05, "loss": 0.8596, "step": 65 }, { "epoch": 0.0, "learning_rate": 2.4952741020793952e-05, "loss": 0.7888, "step": 66 }, { "epoch": 0.0, "learning_rate": 2.5330812854442342e-05, "loss": 0.9507, "step": 67 }, { "epoch": 0.0, "learning_rate": 2.570888468809074e-05, "loss": 0.9369, "step": 68 }, { "epoch": 0.0, "learning_rate": 2.608695652173913e-05, "loss": 0.8152, "step": 69 }, { "epoch": 0.0, "learning_rate": 2.646502835538752e-05, "loss": 0.9377, "step": 70 }, { "epoch": 0.0, "learning_rate": 2.6843100189035918e-05, "loss": 0.7819, "step": 71 }, { "epoch": 0.0, "learning_rate": 2.722117202268431e-05, "loss": 0.9361, "step": 72 }, { "epoch": 0.0, "learning_rate": 2.7599243856332708e-05, "loss": 0.925, "step": 73 }, { "epoch": 0.0, "learning_rate": 2.7977315689981097e-05, "loss": 0.9496, "step": 74 }, { "epoch": 0.0, "learning_rate": 2.835538752362949e-05, "loss": 0.8773, "step": 75 }, { "epoch": 0.0, "learning_rate": 2.8733459357277887e-05, "loss": 0.9184, "step": 76 }, { "epoch": 0.0, "learning_rate": 2.9111531190926277e-05, "loss": 0.9475, "step": 77 }, { "epoch": 0.0, "learning_rate": 2.948960302457467e-05, "loss": 0.9029, "step": 78 }, { "epoch": 0.0, "learning_rate": 2.9867674858223067e-05, "loss": 0.9605, "step": 79 }, { "epoch": 0.0, "learning_rate": 3.0245746691871456e-05, "loss": 0.9128, "step": 80 }, { "epoch": 0.0, "learning_rate": 3.0623818525519846e-05, "loss": 0.8634, "step": 81 }, { "epoch": 0.0, "learning_rate": 3.1001890359168246e-05, "loss": 0.6919, "step": 82 }, { "epoch": 0.0, "learning_rate": 3.137996219281664e-05, "loss": 0.9366, "step": 83 }, { "epoch": 0.0, "learning_rate": 3.1758034026465026e-05, "loss": 0.8869, "step": 84 }, { "epoch": 0.0, "learning_rate": 3.2136105860113426e-05, "loss": 0.8967, "step": 85 }, { "epoch": 0.0, "learning_rate": 3.251417769376182e-05, "loss": 0.8176, "step": 86 }, { "epoch": 0.0, "learning_rate": 3.2892249527410205e-05, "loss": 0.9247, "step": 87 }, { "epoch": 0.0, "learning_rate": 3.3270321361058605e-05, "loss": 0.853, "step": 88 }, { "epoch": 0.01, "learning_rate": 3.3648393194707e-05, "loss": 0.7597, "step": 89 }, { "epoch": 0.01, "learning_rate": 3.4026465028355385e-05, "loss": 0.7731, "step": 90 }, { "epoch": 0.01, "learning_rate": 3.4404536862003784e-05, "loss": 0.7826, "step": 91 }, { "epoch": 0.01, "learning_rate": 3.478260869565218e-05, "loss": 0.785, "step": 92 }, { "epoch": 0.01, "learning_rate": 3.516068052930057e-05, "loss": 0.788, "step": 93 }, { "epoch": 0.01, "learning_rate": 3.5538752362948964e-05, "loss": 0.8599, "step": 94 }, { "epoch": 0.01, "learning_rate": 3.591682419659736e-05, "loss": 0.8639, "step": 95 }, { "epoch": 0.01, "learning_rate": 3.629489603024575e-05, "loss": 0.7195, "step": 96 }, { "epoch": 0.01, "learning_rate": 3.6672967863894143e-05, "loss": 0.7491, "step": 97 }, { "epoch": 0.01, "learning_rate": 3.7051039697542537e-05, "loss": 0.9036, "step": 98 }, { "epoch": 0.01, "learning_rate": 3.742911153119093e-05, "loss": 0.7987, "step": 99 }, { "epoch": 0.01, "learning_rate": 3.780718336483932e-05, "loss": 0.7803, "step": 100 }, { "epoch": 0.01, "learning_rate": 3.8185255198487716e-05, "loss": 0.8248, "step": 101 }, { "epoch": 0.01, "learning_rate": 3.856332703213611e-05, "loss": 0.9346, "step": 102 }, { "epoch": 0.01, "learning_rate": 3.89413988657845e-05, "loss": 0.8276, "step": 103 }, { "epoch": 0.01, "learning_rate": 3.931947069943289e-05, "loss": 0.7939, "step": 104 }, { "epoch": 0.01, "learning_rate": 3.969754253308129e-05, "loss": 0.7352, "step": 105 }, { "epoch": 0.01, "learning_rate": 4.007561436672968e-05, "loss": 0.8962, "step": 106 }, { "epoch": 0.01, "learning_rate": 4.045368620037807e-05, "loss": 0.7757, "step": 107 }, { "epoch": 0.01, "learning_rate": 4.083175803402647e-05, "loss": 0.8364, "step": 108 }, { "epoch": 0.01, "learning_rate": 4.120982986767486e-05, "loss": 0.9954, "step": 109 }, { "epoch": 0.01, "learning_rate": 4.158790170132325e-05, "loss": 0.8837, "step": 110 }, { "epoch": 0.01, "learning_rate": 4.196597353497165e-05, "loss": 0.8631, "step": 111 }, { "epoch": 0.01, "learning_rate": 4.234404536862004e-05, "loss": 0.7751, "step": 112 }, { "epoch": 0.01, "learning_rate": 4.2722117202268434e-05, "loss": 0.866, "step": 113 }, { "epoch": 0.01, "learning_rate": 4.310018903591683e-05, "loss": 0.7762, "step": 114 }, { "epoch": 0.01, "learning_rate": 4.347826086956522e-05, "loss": 0.8579, "step": 115 }, { "epoch": 0.01, "learning_rate": 4.3856332703213614e-05, "loss": 0.9105, "step": 116 }, { "epoch": 0.01, "learning_rate": 4.423440453686201e-05, "loss": 0.8564, "step": 117 }, { "epoch": 0.01, "learning_rate": 4.46124763705104e-05, "loss": 0.8737, "step": 118 }, { "epoch": 0.01, "learning_rate": 4.499054820415879e-05, "loss": 0.7436, "step": 119 }, { "epoch": 0.01, "learning_rate": 4.5368620037807186e-05, "loss": 0.8344, "step": 120 }, { "epoch": 0.01, "learning_rate": 4.574669187145558e-05, "loss": 0.8473, "step": 121 }, { "epoch": 0.01, "learning_rate": 4.612476370510397e-05, "loss": 0.8236, "step": 122 }, { "epoch": 0.01, "learning_rate": 4.6502835538752366e-05, "loss": 0.8306, "step": 123 }, { "epoch": 0.01, "learning_rate": 4.688090737240076e-05, "loss": 0.8587, "step": 124 }, { "epoch": 0.01, "learning_rate": 4.725897920604915e-05, "loss": 0.7958, "step": 125 }, { "epoch": 0.01, "learning_rate": 4.7637051039697545e-05, "loss": 0.8524, "step": 126 }, { "epoch": 0.01, "learning_rate": 4.801512287334594e-05, "loss": 0.9097, "step": 127 }, { "epoch": 0.01, "learning_rate": 4.839319470699433e-05, "loss": 0.9649, "step": 128 }, { "epoch": 0.01, "learning_rate": 4.8771266540642725e-05, "loss": 0.7831, "step": 129 }, { "epoch": 0.01, "learning_rate": 4.914933837429112e-05, "loss": 0.9123, "step": 130 }, { "epoch": 0.01, "learning_rate": 4.952741020793951e-05, "loss": 0.7646, "step": 131 }, { "epoch": 0.01, "learning_rate": 4.9905482041587904e-05, "loss": 0.9485, "step": 132 }, { "epoch": 0.01, "learning_rate": 5.02835538752363e-05, "loss": 0.8175, "step": 133 }, { "epoch": 0.01, "learning_rate": 5.0661625708884684e-05, "loss": 0.8015, "step": 134 }, { "epoch": 0.01, "learning_rate": 5.1039697542533084e-05, "loss": 0.7727, "step": 135 }, { "epoch": 0.01, "learning_rate": 5.141776937618148e-05, "loss": 0.8151, "step": 136 }, { "epoch": 0.01, "learning_rate": 5.179584120982986e-05, "loss": 0.7858, "step": 137 }, { "epoch": 0.01, "learning_rate": 5.217391304347826e-05, "loss": 0.7213, "step": 138 }, { "epoch": 0.01, "learning_rate": 5.2551984877126656e-05, "loss": 0.8068, "step": 139 }, { "epoch": 0.01, "learning_rate": 5.293005671077504e-05, "loss": 0.7059, "step": 140 }, { "epoch": 0.01, "learning_rate": 5.330812854442344e-05, "loss": 0.7212, "step": 141 }, { "epoch": 0.01, "learning_rate": 5.3686200378071836e-05, "loss": 0.8979, "step": 142 }, { "epoch": 0.01, "learning_rate": 5.4064272211720236e-05, "loss": 0.6804, "step": 143 }, { "epoch": 0.01, "learning_rate": 5.444234404536862e-05, "loss": 0.9708, "step": 144 }, { "epoch": 0.01, "learning_rate": 5.4820415879017015e-05, "loss": 0.7814, "step": 145 }, { "epoch": 0.01, "learning_rate": 5.5198487712665415e-05, "loss": 0.7526, "step": 146 }, { "epoch": 0.01, "learning_rate": 5.55765595463138e-05, "loss": 0.785, "step": 147 }, { "epoch": 0.01, "learning_rate": 5.5954631379962195e-05, "loss": 0.8069, "step": 148 }, { "epoch": 0.01, "learning_rate": 5.6332703213610595e-05, "loss": 0.7573, "step": 149 }, { "epoch": 0.01, "learning_rate": 5.671077504725898e-05, "loss": 0.8468, "step": 150 }, { "epoch": 0.01, "learning_rate": 5.7088846880907374e-05, "loss": 0.793, "step": 151 }, { "epoch": 0.01, "learning_rate": 5.7466918714555774e-05, "loss": 0.8021, "step": 152 }, { "epoch": 0.01, "learning_rate": 5.784499054820416e-05, "loss": 0.8249, "step": 153 }, { "epoch": 0.01, "learning_rate": 5.8223062381852554e-05, "loss": 0.764, "step": 154 }, { "epoch": 0.01, "learning_rate": 5.8601134215500954e-05, "loss": 0.7887, "step": 155 }, { "epoch": 0.01, "learning_rate": 5.897920604914934e-05, "loss": 0.7563, "step": 156 }, { "epoch": 0.01, "learning_rate": 5.935727788279773e-05, "loss": 0.8354, "step": 157 }, { "epoch": 0.01, "learning_rate": 5.973534971644613e-05, "loss": 0.7723, "step": 158 }, { "epoch": 0.01, "learning_rate": 6.011342155009452e-05, "loss": 0.7455, "step": 159 }, { "epoch": 0.01, "learning_rate": 6.049149338374291e-05, "loss": 0.7655, "step": 160 }, { "epoch": 0.01, "learning_rate": 6.086956521739131e-05, "loss": 0.8144, "step": 161 }, { "epoch": 0.01, "learning_rate": 6.124763705103969e-05, "loss": 0.7718, "step": 162 }, { "epoch": 0.01, "learning_rate": 6.16257088846881e-05, "loss": 0.8373, "step": 163 }, { "epoch": 0.01, "learning_rate": 6.200378071833649e-05, "loss": 0.8353, "step": 164 }, { "epoch": 0.01, "learning_rate": 6.238185255198487e-05, "loss": 0.7683, "step": 165 }, { "epoch": 0.01, "learning_rate": 6.275992438563328e-05, "loss": 0.7856, "step": 166 }, { "epoch": 0.01, "learning_rate": 6.313799621928167e-05, "loss": 0.8346, "step": 167 }, { "epoch": 0.01, "learning_rate": 6.351606805293005e-05, "loss": 0.8528, "step": 168 }, { "epoch": 0.01, "learning_rate": 6.389413988657846e-05, "loss": 0.7765, "step": 169 }, { "epoch": 0.01, "learning_rate": 6.427221172022685e-05, "loss": 0.7727, "step": 170 }, { "epoch": 0.01, "learning_rate": 6.465028355387523e-05, "loss": 0.8243, "step": 171 }, { "epoch": 0.01, "learning_rate": 6.502835538752364e-05, "loss": 0.8843, "step": 172 }, { "epoch": 0.01, "learning_rate": 6.540642722117203e-05, "loss": 0.8214, "step": 173 }, { "epoch": 0.01, "learning_rate": 6.578449905482041e-05, "loss": 0.7581, "step": 174 }, { "epoch": 0.01, "learning_rate": 6.616257088846882e-05, "loss": 0.7969, "step": 175 }, { "epoch": 0.01, "learning_rate": 6.654064272211721e-05, "loss": 0.8198, "step": 176 }, { "epoch": 0.01, "learning_rate": 6.691871455576559e-05, "loss": 0.8869, "step": 177 }, { "epoch": 0.01, "learning_rate": 6.7296786389414e-05, "loss": 0.7037, "step": 178 }, { "epoch": 0.01, "learning_rate": 6.767485822306239e-05, "loss": 0.8832, "step": 179 }, { "epoch": 0.01, "learning_rate": 6.805293005671077e-05, "loss": 0.7906, "step": 180 }, { "epoch": 0.01, "learning_rate": 6.843100189035918e-05, "loss": 0.7779, "step": 181 }, { "epoch": 0.01, "learning_rate": 6.880907372400757e-05, "loss": 0.8687, "step": 182 }, { "epoch": 0.01, "learning_rate": 6.918714555765595e-05, "loss": 0.9126, "step": 183 }, { "epoch": 0.01, "learning_rate": 6.956521739130436e-05, "loss": 0.7644, "step": 184 }, { "epoch": 0.01, "learning_rate": 6.994328922495275e-05, "loss": 0.8284, "step": 185 }, { "epoch": 0.01, "learning_rate": 7.032136105860114e-05, "loss": 0.781, "step": 186 }, { "epoch": 0.01, "learning_rate": 7.069943289224953e-05, "loss": 0.8645, "step": 187 }, { "epoch": 0.01, "learning_rate": 7.107750472589793e-05, "loss": 0.824, "step": 188 }, { "epoch": 0.01, "learning_rate": 7.145557655954632e-05, "loss": 0.8563, "step": 189 }, { "epoch": 0.01, "learning_rate": 7.183364839319471e-05, "loss": 0.8171, "step": 190 }, { "epoch": 0.01, "learning_rate": 7.221172022684311e-05, "loss": 0.8377, "step": 191 }, { "epoch": 0.01, "learning_rate": 7.25897920604915e-05, "loss": 0.818, "step": 192 }, { "epoch": 0.01, "learning_rate": 7.29678638941399e-05, "loss": 0.7107, "step": 193 }, { "epoch": 0.01, "learning_rate": 7.334593572778829e-05, "loss": 0.8238, "step": 194 }, { "epoch": 0.01, "learning_rate": 7.372400756143668e-05, "loss": 0.8171, "step": 195 }, { "epoch": 0.01, "learning_rate": 7.410207939508507e-05, "loss": 0.8068, "step": 196 }, { "epoch": 0.01, "learning_rate": 7.448015122873347e-05, "loss": 0.8557, "step": 197 }, { "epoch": 0.01, "learning_rate": 7.485822306238186e-05, "loss": 0.7559, "step": 198 }, { "epoch": 0.01, "learning_rate": 7.523629489603025e-05, "loss": 0.7922, "step": 199 }, { "epoch": 0.01, "learning_rate": 7.561436672967865e-05, "loss": 0.7273, "step": 200 }, { "epoch": 0.01, "learning_rate": 7.599243856332704e-05, "loss": 0.807, "step": 201 }, { "epoch": 0.01, "learning_rate": 7.637051039697543e-05, "loss": 0.7367, "step": 202 }, { "epoch": 0.01, "learning_rate": 7.674858223062383e-05, "loss": 0.7642, "step": 203 }, { "epoch": 0.01, "learning_rate": 7.712665406427222e-05, "loss": 0.721, "step": 204 }, { "epoch": 0.01, "learning_rate": 7.75047258979206e-05, "loss": 0.7212, "step": 205 }, { "epoch": 0.01, "learning_rate": 7.7882797731569e-05, "loss": 0.7766, "step": 206 }, { "epoch": 0.01, "learning_rate": 7.82608695652174e-05, "loss": 0.7286, "step": 207 }, { "epoch": 0.01, "learning_rate": 7.863894139886578e-05, "loss": 0.6945, "step": 208 }, { "epoch": 0.01, "learning_rate": 7.901701323251418e-05, "loss": 0.8086, "step": 209 }, { "epoch": 0.01, "learning_rate": 7.939508506616258e-05, "loss": 0.7999, "step": 210 }, { "epoch": 0.01, "learning_rate": 7.977315689981096e-05, "loss": 0.8036, "step": 211 }, { "epoch": 0.01, "learning_rate": 8.015122873345936e-05, "loss": 0.7186, "step": 212 }, { "epoch": 0.01, "learning_rate": 8.052930056710776e-05, "loss": 0.8682, "step": 213 }, { "epoch": 0.01, "learning_rate": 8.090737240075614e-05, "loss": 0.7464, "step": 214 }, { "epoch": 0.01, "learning_rate": 8.128544423440454e-05, "loss": 0.8147, "step": 215 }, { "epoch": 0.01, "learning_rate": 8.166351606805294e-05, "loss": 0.7729, "step": 216 }, { "epoch": 0.01, "learning_rate": 8.204158790170132e-05, "loss": 0.7332, "step": 217 }, { "epoch": 0.01, "learning_rate": 8.241965973534972e-05, "loss": 0.7158, "step": 218 }, { "epoch": 0.01, "learning_rate": 8.279773156899812e-05, "loss": 0.8053, "step": 219 }, { "epoch": 0.01, "learning_rate": 8.31758034026465e-05, "loss": 0.7396, "step": 220 }, { "epoch": 0.01, "learning_rate": 8.35538752362949e-05, "loss": 0.9114, "step": 221 }, { "epoch": 0.01, "learning_rate": 8.39319470699433e-05, "loss": 0.7537, "step": 222 }, { "epoch": 0.01, "learning_rate": 8.431001890359168e-05, "loss": 0.7455, "step": 223 }, { "epoch": 0.01, "learning_rate": 8.468809073724008e-05, "loss": 0.7709, "step": 224 }, { "epoch": 0.01, "learning_rate": 8.506616257088847e-05, "loss": 0.654, "step": 225 }, { "epoch": 0.01, "learning_rate": 8.544423440453687e-05, "loss": 0.865, "step": 226 }, { "epoch": 0.01, "learning_rate": 8.582230623818526e-05, "loss": 0.7905, "step": 227 }, { "epoch": 0.01, "learning_rate": 8.620037807183365e-05, "loss": 0.7276, "step": 228 }, { "epoch": 0.01, "learning_rate": 8.657844990548205e-05, "loss": 0.8015, "step": 229 }, { "epoch": 0.01, "learning_rate": 8.695652173913044e-05, "loss": 0.7306, "step": 230 }, { "epoch": 0.01, "learning_rate": 8.733459357277883e-05, "loss": 0.7668, "step": 231 }, { "epoch": 0.01, "learning_rate": 8.771266540642723e-05, "loss": 0.7788, "step": 232 }, { "epoch": 0.01, "learning_rate": 8.809073724007562e-05, "loss": 0.7493, "step": 233 }, { "epoch": 0.01, "learning_rate": 8.846880907372401e-05, "loss": 0.7832, "step": 234 }, { "epoch": 0.01, "learning_rate": 8.88468809073724e-05, "loss": 0.79, "step": 235 }, { "epoch": 0.01, "learning_rate": 8.92249527410208e-05, "loss": 0.7606, "step": 236 }, { "epoch": 0.01, "learning_rate": 8.960302457466919e-05, "loss": 0.8204, "step": 237 }, { "epoch": 0.01, "learning_rate": 8.998109640831759e-05, "loss": 0.8119, "step": 238 }, { "epoch": 0.01, "learning_rate": 9.035916824196598e-05, "loss": 0.7517, "step": 239 }, { "epoch": 0.01, "learning_rate": 9.073724007561437e-05, "loss": 0.837, "step": 240 }, { "epoch": 0.01, "learning_rate": 9.111531190926277e-05, "loss": 0.7086, "step": 241 }, { "epoch": 0.01, "learning_rate": 9.149338374291116e-05, "loss": 0.782, "step": 242 }, { "epoch": 0.01, "learning_rate": 9.187145557655955e-05, "loss": 0.7932, "step": 243 }, { "epoch": 0.01, "learning_rate": 9.224952741020794e-05, "loss": 0.7455, "step": 244 }, { "epoch": 0.01, "learning_rate": 9.262759924385634e-05, "loss": 0.7015, "step": 245 }, { "epoch": 0.01, "learning_rate": 9.300567107750473e-05, "loss": 0.7233, "step": 246 }, { "epoch": 0.01, "learning_rate": 9.338374291115312e-05, "loss": 0.786, "step": 247 }, { "epoch": 0.01, "learning_rate": 9.376181474480152e-05, "loss": 0.7863, "step": 248 }, { "epoch": 0.01, "learning_rate": 9.413988657844991e-05, "loss": 0.7864, "step": 249 }, { "epoch": 0.01, "learning_rate": 9.45179584120983e-05, "loss": 0.7399, "step": 250 }, { "epoch": 0.01, "learning_rate": 9.48960302457467e-05, "loss": 0.7849, "step": 251 }, { "epoch": 0.01, "learning_rate": 9.527410207939509e-05, "loss": 0.7351, "step": 252 }, { "epoch": 0.01, "learning_rate": 9.565217391304348e-05, "loss": 0.6828, "step": 253 }, { "epoch": 0.01, "learning_rate": 9.603024574669188e-05, "loss": 0.8177, "step": 254 }, { "epoch": 0.01, "learning_rate": 9.640831758034027e-05, "loss": 0.7653, "step": 255 }, { "epoch": 0.01, "learning_rate": 9.678638941398866e-05, "loss": 0.7928, "step": 256 }, { "epoch": 0.01, "learning_rate": 9.716446124763706e-05, "loss": 0.7601, "step": 257 }, { "epoch": 0.01, "learning_rate": 9.754253308128545e-05, "loss": 0.7372, "step": 258 }, { "epoch": 0.01, "learning_rate": 9.792060491493384e-05, "loss": 0.7513, "step": 259 }, { "epoch": 0.01, "learning_rate": 9.829867674858224e-05, "loss": 0.7956, "step": 260 }, { "epoch": 0.01, "learning_rate": 9.867674858223063e-05, "loss": 0.8067, "step": 261 }, { "epoch": 0.01, "learning_rate": 9.905482041587902e-05, "loss": 0.7436, "step": 262 }, { "epoch": 0.01, "learning_rate": 9.943289224952742e-05, "loss": 0.7129, "step": 263 }, { "epoch": 0.01, "learning_rate": 9.981096408317581e-05, "loss": 0.8051, "step": 264 }, { "epoch": 0.02, "learning_rate": 0.0001001890359168242, "loss": 0.7728, "step": 265 }, { "epoch": 0.02, "learning_rate": 0.0001005671077504726, "loss": 0.7609, "step": 266 }, { "epoch": 0.02, "learning_rate": 0.000100945179584121, "loss": 0.7859, "step": 267 }, { "epoch": 0.02, "learning_rate": 0.00010132325141776937, "loss": 0.8832, "step": 268 }, { "epoch": 0.02, "learning_rate": 0.00010170132325141777, "loss": 0.7846, "step": 269 }, { "epoch": 0.02, "learning_rate": 0.00010207939508506617, "loss": 0.7463, "step": 270 }, { "epoch": 0.02, "learning_rate": 0.00010245746691871456, "loss": 0.7232, "step": 271 }, { "epoch": 0.02, "learning_rate": 0.00010283553875236295, "loss": 0.7346, "step": 272 }, { "epoch": 0.02, "learning_rate": 0.00010321361058601136, "loss": 0.768, "step": 273 }, { "epoch": 0.02, "learning_rate": 0.00010359168241965973, "loss": 0.7324, "step": 274 }, { "epoch": 0.02, "learning_rate": 0.00010396975425330813, "loss": 0.7984, "step": 275 }, { "epoch": 0.02, "learning_rate": 0.00010434782608695653, "loss": 0.7399, "step": 276 }, { "epoch": 0.02, "learning_rate": 0.00010472589792060492, "loss": 0.7372, "step": 277 }, { "epoch": 0.02, "learning_rate": 0.00010510396975425331, "loss": 0.6848, "step": 278 }, { "epoch": 0.02, "learning_rate": 0.00010548204158790172, "loss": 0.9239, "step": 279 }, { "epoch": 0.02, "learning_rate": 0.00010586011342155009, "loss": 0.7942, "step": 280 }, { "epoch": 0.02, "learning_rate": 0.00010623818525519849, "loss": 0.7674, "step": 281 }, { "epoch": 0.02, "learning_rate": 0.00010661625708884689, "loss": 0.7701, "step": 282 }, { "epoch": 0.02, "learning_rate": 0.00010699432892249528, "loss": 0.8783, "step": 283 }, { "epoch": 0.02, "learning_rate": 0.00010737240075614367, "loss": 0.743, "step": 284 }, { "epoch": 0.02, "learning_rate": 0.00010775047258979208, "loss": 0.7057, "step": 285 }, { "epoch": 0.02, "learning_rate": 0.00010812854442344047, "loss": 0.6562, "step": 286 }, { "epoch": 0.02, "learning_rate": 0.00010850661625708885, "loss": 0.8288, "step": 287 }, { "epoch": 0.02, "learning_rate": 0.00010888468809073724, "loss": 0.8103, "step": 288 }, { "epoch": 0.02, "learning_rate": 0.00010926275992438564, "loss": 0.7568, "step": 289 }, { "epoch": 0.02, "learning_rate": 0.00010964083175803403, "loss": 0.7258, "step": 290 }, { "epoch": 0.02, "learning_rate": 0.00011001890359168244, "loss": 0.7762, "step": 291 }, { "epoch": 0.02, "learning_rate": 0.00011039697542533083, "loss": 0.9137, "step": 292 }, { "epoch": 0.02, "learning_rate": 0.00011077504725897921, "loss": 0.7523, "step": 293 }, { "epoch": 0.02, "learning_rate": 0.0001111531190926276, "loss": 0.7189, "step": 294 }, { "epoch": 0.02, "learning_rate": 0.000111531190926276, "loss": 0.883, "step": 295 }, { "epoch": 0.02, "learning_rate": 0.00011190926275992439, "loss": 0.7056, "step": 296 }, { "epoch": 0.02, "learning_rate": 0.0001122873345935728, "loss": 0.6642, "step": 297 }, { "epoch": 0.02, "learning_rate": 0.00011266540642722119, "loss": 0.7314, "step": 298 }, { "epoch": 0.02, "learning_rate": 0.00011304347826086956, "loss": 0.7569, "step": 299 }, { "epoch": 0.02, "learning_rate": 0.00011342155009451796, "loss": 0.7016, "step": 300 }, { "epoch": 0.02, "learning_rate": 0.00011379962192816636, "loss": 0.7376, "step": 301 }, { "epoch": 0.02, "learning_rate": 0.00011417769376181475, "loss": 0.6992, "step": 302 }, { "epoch": 0.02, "learning_rate": 0.00011455576559546316, "loss": 0.6923, "step": 303 }, { "epoch": 0.02, "learning_rate": 0.00011493383742911155, "loss": 0.706, "step": 304 }, { "epoch": 0.02, "learning_rate": 0.00011531190926275991, "loss": 0.7788, "step": 305 }, { "epoch": 0.02, "learning_rate": 0.00011568998109640832, "loss": 0.7782, "step": 306 }, { "epoch": 0.02, "learning_rate": 0.00011606805293005671, "loss": 0.819, "step": 307 }, { "epoch": 0.02, "learning_rate": 0.00011644612476370511, "loss": 0.7665, "step": 308 }, { "epoch": 0.02, "learning_rate": 0.00011682419659735351, "loss": 0.7721, "step": 309 }, { "epoch": 0.02, "learning_rate": 0.00011720226843100191, "loss": 0.7923, "step": 310 }, { "epoch": 0.02, "learning_rate": 0.00011758034026465027, "loss": 0.7888, "step": 311 }, { "epoch": 0.02, "learning_rate": 0.00011795841209829868, "loss": 0.7604, "step": 312 }, { "epoch": 0.02, "learning_rate": 0.00011833648393194707, "loss": 0.7475, "step": 313 }, { "epoch": 0.02, "learning_rate": 0.00011871455576559547, "loss": 0.7452, "step": 314 }, { "epoch": 0.02, "learning_rate": 0.00011909262759924387, "loss": 0.8524, "step": 315 }, { "epoch": 0.02, "learning_rate": 0.00011947069943289227, "loss": 0.6892, "step": 316 }, { "epoch": 0.02, "learning_rate": 0.00011984877126654063, "loss": 0.8106, "step": 317 }, { "epoch": 0.02, "learning_rate": 0.00012022684310018904, "loss": 0.6472, "step": 318 }, { "epoch": 0.02, "learning_rate": 0.00012060491493383743, "loss": 0.806, "step": 319 }, { "epoch": 0.02, "learning_rate": 0.00012098298676748583, "loss": 0.8756, "step": 320 }, { "epoch": 0.02, "learning_rate": 0.00012136105860113423, "loss": 0.7527, "step": 321 }, { "epoch": 0.02, "learning_rate": 0.00012173913043478263, "loss": 0.7876, "step": 322 }, { "epoch": 0.02, "learning_rate": 0.000122117202268431, "loss": 0.7784, "step": 323 }, { "epoch": 0.02, "learning_rate": 0.00012249527410207938, "loss": 0.803, "step": 324 }, { "epoch": 0.02, "learning_rate": 0.00012287334593572778, "loss": 0.8312, "step": 325 }, { "epoch": 0.02, "learning_rate": 0.0001232514177693762, "loss": 0.7448, "step": 326 }, { "epoch": 0.02, "learning_rate": 0.0001236294896030246, "loss": 0.8357, "step": 327 }, { "epoch": 0.02, "learning_rate": 0.00012400756143667298, "loss": 0.8747, "step": 328 }, { "epoch": 0.02, "learning_rate": 0.00012438563327032138, "loss": 0.7374, "step": 329 }, { "epoch": 0.02, "learning_rate": 0.00012476370510396974, "loss": 0.8152, "step": 330 }, { "epoch": 0.02, "learning_rate": 0.00012514177693761814, "loss": 0.7999, "step": 331 }, { "epoch": 0.02, "learning_rate": 0.00012551984877126656, "loss": 0.7081, "step": 332 }, { "epoch": 0.02, "learning_rate": 0.00012589792060491495, "loss": 0.7587, "step": 333 }, { "epoch": 0.02, "learning_rate": 0.00012627599243856334, "loss": 0.785, "step": 334 }, { "epoch": 0.02, "learning_rate": 0.00012665406427221174, "loss": 0.7822, "step": 335 }, { "epoch": 0.02, "learning_rate": 0.0001270321361058601, "loss": 0.8207, "step": 336 }, { "epoch": 0.02, "learning_rate": 0.0001274102079395085, "loss": 0.7499, "step": 337 }, { "epoch": 0.02, "learning_rate": 0.00012778827977315692, "loss": 0.8434, "step": 338 }, { "epoch": 0.02, "learning_rate": 0.0001281663516068053, "loss": 0.6938, "step": 339 }, { "epoch": 0.02, "learning_rate": 0.0001285444234404537, "loss": 0.7837, "step": 340 }, { "epoch": 0.02, "learning_rate": 0.0001289224952741021, "loss": 0.8135, "step": 341 }, { "epoch": 0.02, "learning_rate": 0.00012930056710775046, "loss": 0.6807, "step": 342 }, { "epoch": 0.02, "learning_rate": 0.00012967863894139885, "loss": 0.7096, "step": 343 }, { "epoch": 0.02, "learning_rate": 0.00013005671077504727, "loss": 0.7162, "step": 344 }, { "epoch": 0.02, "learning_rate": 0.00013043478260869567, "loss": 0.8306, "step": 345 }, { "epoch": 0.02, "learning_rate": 0.00013081285444234406, "loss": 0.7884, "step": 346 }, { "epoch": 0.02, "learning_rate": 0.00013119092627599245, "loss": 0.7579, "step": 347 }, { "epoch": 0.02, "learning_rate": 0.00013156899810964082, "loss": 0.6791, "step": 348 }, { "epoch": 0.02, "learning_rate": 0.0001319470699432892, "loss": 0.7746, "step": 349 }, { "epoch": 0.02, "learning_rate": 0.00013232514177693763, "loss": 0.7018, "step": 350 }, { "epoch": 0.02, "learning_rate": 0.00013270321361058603, "loss": 0.7726, "step": 351 }, { "epoch": 0.02, "learning_rate": 0.00013308128544423442, "loss": 0.7557, "step": 352 }, { "epoch": 0.02, "learning_rate": 0.0001334593572778828, "loss": 0.742, "step": 353 }, { "epoch": 0.02, "learning_rate": 0.00013383742911153118, "loss": 0.732, "step": 354 }, { "epoch": 0.02, "learning_rate": 0.00013421550094517957, "loss": 0.8174, "step": 355 }, { "epoch": 0.02, "learning_rate": 0.000134593572778828, "loss": 0.8188, "step": 356 }, { "epoch": 0.02, "learning_rate": 0.00013497164461247639, "loss": 0.8223, "step": 357 }, { "epoch": 0.02, "learning_rate": 0.00013534971644612478, "loss": 0.7609, "step": 358 }, { "epoch": 0.02, "learning_rate": 0.00013572778827977317, "loss": 0.8144, "step": 359 }, { "epoch": 0.02, "learning_rate": 0.00013610586011342154, "loss": 0.6955, "step": 360 }, { "epoch": 0.02, "learning_rate": 0.00013648393194706993, "loss": 0.692, "step": 361 }, { "epoch": 0.02, "learning_rate": 0.00013686200378071835, "loss": 0.7275, "step": 362 }, { "epoch": 0.02, "learning_rate": 0.00013724007561436674, "loss": 0.7141, "step": 363 }, { "epoch": 0.02, "learning_rate": 0.00013761814744801514, "loss": 0.9163, "step": 364 }, { "epoch": 0.02, "learning_rate": 0.00013799621928166353, "loss": 0.7509, "step": 365 }, { "epoch": 0.02, "learning_rate": 0.0001383742911153119, "loss": 0.6881, "step": 366 }, { "epoch": 0.02, "learning_rate": 0.0001387523629489603, "loss": 0.7414, "step": 367 }, { "epoch": 0.02, "learning_rate": 0.0001391304347826087, "loss": 0.7223, "step": 368 }, { "epoch": 0.02, "learning_rate": 0.0001395085066162571, "loss": 0.716, "step": 369 }, { "epoch": 0.02, "learning_rate": 0.0001398865784499055, "loss": 0.8151, "step": 370 }, { "epoch": 0.02, "learning_rate": 0.0001402646502835539, "loss": 0.8119, "step": 371 }, { "epoch": 0.02, "learning_rate": 0.00014064272211720228, "loss": 0.7539, "step": 372 }, { "epoch": 0.02, "learning_rate": 0.00014102079395085065, "loss": 0.7497, "step": 373 }, { "epoch": 0.02, "learning_rate": 0.00014139886578449907, "loss": 0.6819, "step": 374 }, { "epoch": 0.02, "learning_rate": 0.00014177693761814746, "loss": 0.8433, "step": 375 }, { "epoch": 0.02, "learning_rate": 0.00014215500945179586, "loss": 0.6986, "step": 376 }, { "epoch": 0.02, "learning_rate": 0.00014253308128544425, "loss": 0.827, "step": 377 }, { "epoch": 0.02, "learning_rate": 0.00014291115311909264, "loss": 0.6814, "step": 378 }, { "epoch": 0.02, "learning_rate": 0.000143289224952741, "loss": 0.772, "step": 379 }, { "epoch": 0.02, "learning_rate": 0.00014366729678638943, "loss": 0.7516, "step": 380 }, { "epoch": 0.02, "learning_rate": 0.00014404536862003782, "loss": 0.7909, "step": 381 }, { "epoch": 0.02, "learning_rate": 0.00014442344045368621, "loss": 0.7744, "step": 382 }, { "epoch": 0.02, "learning_rate": 0.0001448015122873346, "loss": 0.7809, "step": 383 }, { "epoch": 0.02, "learning_rate": 0.000145179584120983, "loss": 0.6953, "step": 384 }, { "epoch": 0.02, "learning_rate": 0.00014555765595463137, "loss": 0.7916, "step": 385 }, { "epoch": 0.02, "learning_rate": 0.0001459357277882798, "loss": 0.7973, "step": 386 }, { "epoch": 0.02, "learning_rate": 0.00014631379962192818, "loss": 0.6805, "step": 387 }, { "epoch": 0.02, "learning_rate": 0.00014669187145557657, "loss": 0.7311, "step": 388 }, { "epoch": 0.02, "learning_rate": 0.00014706994328922497, "loss": 0.7445, "step": 389 }, { "epoch": 0.02, "learning_rate": 0.00014744801512287336, "loss": 0.865, "step": 390 }, { "epoch": 0.02, "learning_rate": 0.00014782608695652173, "loss": 0.7427, "step": 391 }, { "epoch": 0.02, "learning_rate": 0.00014820415879017015, "loss": 0.7512, "step": 392 }, { "epoch": 0.02, "learning_rate": 0.00014858223062381854, "loss": 0.7215, "step": 393 }, { "epoch": 0.02, "learning_rate": 0.00014896030245746693, "loss": 0.7322, "step": 394 }, { "epoch": 0.02, "learning_rate": 0.00014933837429111533, "loss": 0.7145, "step": 395 }, { "epoch": 0.02, "learning_rate": 0.00014971644612476372, "loss": 0.6939, "step": 396 }, { "epoch": 0.02, "learning_rate": 0.00015009451795841209, "loss": 0.7653, "step": 397 }, { "epoch": 0.02, "learning_rate": 0.0001504725897920605, "loss": 0.8597, "step": 398 }, { "epoch": 0.02, "learning_rate": 0.0001508506616257089, "loss": 0.8206, "step": 399 }, { "epoch": 0.02, "learning_rate": 0.0001512287334593573, "loss": 0.8256, "step": 400 }, { "epoch": 0.02, "learning_rate": 0.00015160680529300568, "loss": 0.7832, "step": 401 }, { "epoch": 0.02, "learning_rate": 0.00015198487712665408, "loss": 0.7338, "step": 402 }, { "epoch": 0.02, "learning_rate": 0.00015236294896030244, "loss": 0.7217, "step": 403 }, { "epoch": 0.02, "learning_rate": 0.00015274102079395086, "loss": 0.7976, "step": 404 }, { "epoch": 0.02, "learning_rate": 0.00015311909262759926, "loss": 0.8505, "step": 405 }, { "epoch": 0.02, "learning_rate": 0.00015349716446124765, "loss": 0.8333, "step": 406 }, { "epoch": 0.02, "learning_rate": 0.00015387523629489604, "loss": 0.7478, "step": 407 }, { "epoch": 0.02, "learning_rate": 0.00015425330812854444, "loss": 0.742, "step": 408 }, { "epoch": 0.02, "learning_rate": 0.00015463137996219283, "loss": 0.6976, "step": 409 }, { "epoch": 0.02, "learning_rate": 0.0001550094517958412, "loss": 0.7441, "step": 410 }, { "epoch": 0.02, "learning_rate": 0.00015538752362948962, "loss": 0.6726, "step": 411 }, { "epoch": 0.02, "learning_rate": 0.000155765595463138, "loss": 0.8219, "step": 412 }, { "epoch": 0.02, "learning_rate": 0.0001561436672967864, "loss": 0.7447, "step": 413 }, { "epoch": 0.02, "learning_rate": 0.0001565217391304348, "loss": 0.6856, "step": 414 }, { "epoch": 0.02, "learning_rate": 0.0001568998109640832, "loss": 0.7757, "step": 415 }, { "epoch": 0.02, "learning_rate": 0.00015727788279773156, "loss": 0.652, "step": 416 }, { "epoch": 0.02, "learning_rate": 0.00015765595463137998, "loss": 0.8359, "step": 417 }, { "epoch": 0.02, "learning_rate": 0.00015803402646502837, "loss": 0.742, "step": 418 }, { "epoch": 0.02, "learning_rate": 0.00015841209829867676, "loss": 0.7283, "step": 419 }, { "epoch": 0.02, "learning_rate": 0.00015879017013232515, "loss": 0.7728, "step": 420 }, { "epoch": 0.02, "learning_rate": 0.00015916824196597355, "loss": 0.6323, "step": 421 }, { "epoch": 0.02, "learning_rate": 0.00015954631379962191, "loss": 0.7615, "step": 422 }, { "epoch": 0.02, "learning_rate": 0.00015992438563327033, "loss": 0.7267, "step": 423 }, { "epoch": 0.02, "learning_rate": 0.00016030245746691873, "loss": 0.854, "step": 424 }, { "epoch": 0.02, "learning_rate": 0.00016068052930056712, "loss": 0.6493, "step": 425 }, { "epoch": 0.02, "learning_rate": 0.00016105860113421551, "loss": 0.7221, "step": 426 }, { "epoch": 0.02, "learning_rate": 0.0001614366729678639, "loss": 0.7105, "step": 427 }, { "epoch": 0.02, "learning_rate": 0.00016181474480151227, "loss": 0.7159, "step": 428 }, { "epoch": 0.02, "learning_rate": 0.0001621928166351607, "loss": 0.7045, "step": 429 }, { "epoch": 0.02, "learning_rate": 0.0001625708884688091, "loss": 0.7615, "step": 430 }, { "epoch": 0.02, "learning_rate": 0.00016294896030245748, "loss": 0.6443, "step": 431 }, { "epoch": 0.02, "learning_rate": 0.00016332703213610587, "loss": 0.7156, "step": 432 }, { "epoch": 0.02, "learning_rate": 0.00016370510396975427, "loss": 0.7369, "step": 433 }, { "epoch": 0.02, "learning_rate": 0.00016408317580340263, "loss": 0.7105, "step": 434 }, { "epoch": 0.02, "learning_rate": 0.00016446124763705105, "loss": 0.7744, "step": 435 }, { "epoch": 0.02, "learning_rate": 0.00016483931947069945, "loss": 0.8192, "step": 436 }, { "epoch": 0.02, "learning_rate": 0.00016521739130434784, "loss": 0.9627, "step": 437 }, { "epoch": 0.02, "learning_rate": 0.00016559546313799623, "loss": 0.7714, "step": 438 }, { "epoch": 0.02, "learning_rate": 0.00016597353497164463, "loss": 0.8793, "step": 439 }, { "epoch": 0.02, "learning_rate": 0.000166351606805293, "loss": 0.6631, "step": 440 }, { "epoch": 0.03, "learning_rate": 0.0001667296786389414, "loss": 0.8016, "step": 441 }, { "epoch": 0.03, "learning_rate": 0.0001671077504725898, "loss": 0.6858, "step": 442 }, { "epoch": 0.03, "learning_rate": 0.0001674858223062382, "loss": 0.8587, "step": 443 }, { "epoch": 0.03, "learning_rate": 0.0001678638941398866, "loss": 0.7206, "step": 444 }, { "epoch": 0.03, "learning_rate": 0.00016824196597353498, "loss": 0.7491, "step": 445 }, { "epoch": 0.03, "learning_rate": 0.00016862003780718335, "loss": 0.6381, "step": 446 }, { "epoch": 0.03, "learning_rate": 0.00016899810964083177, "loss": 0.8495, "step": 447 }, { "epoch": 0.03, "learning_rate": 0.00016937618147448016, "loss": 0.7851, "step": 448 }, { "epoch": 0.03, "learning_rate": 0.00016975425330812856, "loss": 0.7151, "step": 449 }, { "epoch": 0.03, "learning_rate": 0.00017013232514177695, "loss": 0.6806, "step": 450 }, { "epoch": 0.03, "learning_rate": 0.00017051039697542534, "loss": 0.7357, "step": 451 }, { "epoch": 0.03, "learning_rate": 0.00017088846880907374, "loss": 0.798, "step": 452 }, { "epoch": 0.03, "learning_rate": 0.00017126654064272213, "loss": 0.6492, "step": 453 }, { "epoch": 0.03, "learning_rate": 0.00017164461247637052, "loss": 0.689, "step": 454 }, { "epoch": 0.03, "learning_rate": 0.00017202268431001892, "loss": 0.7367, "step": 455 }, { "epoch": 0.03, "learning_rate": 0.0001724007561436673, "loss": 0.7661, "step": 456 }, { "epoch": 0.03, "learning_rate": 0.0001727788279773157, "loss": 0.7847, "step": 457 }, { "epoch": 0.03, "learning_rate": 0.0001731568998109641, "loss": 0.742, "step": 458 }, { "epoch": 0.03, "learning_rate": 0.0001735349716446125, "loss": 0.786, "step": 459 }, { "epoch": 0.03, "learning_rate": 0.00017391304347826088, "loss": 0.7552, "step": 460 }, { "epoch": 0.03, "learning_rate": 0.00017429111531190927, "loss": 0.668, "step": 461 }, { "epoch": 0.03, "learning_rate": 0.00017466918714555767, "loss": 0.7361, "step": 462 }, { "epoch": 0.03, "learning_rate": 0.00017504725897920606, "loss": 0.6988, "step": 463 }, { "epoch": 0.03, "learning_rate": 0.00017542533081285445, "loss": 0.7347, "step": 464 }, { "epoch": 0.03, "learning_rate": 0.00017580340264650285, "loss": 0.7535, "step": 465 }, { "epoch": 0.03, "learning_rate": 0.00017618147448015124, "loss": 0.76, "step": 466 }, { "epoch": 0.03, "learning_rate": 0.00017655954631379963, "loss": 0.7541, "step": 467 }, { "epoch": 0.03, "learning_rate": 0.00017693761814744803, "loss": 0.8276, "step": 468 }, { "epoch": 0.03, "learning_rate": 0.00017731568998109642, "loss": 0.768, "step": 469 }, { "epoch": 0.03, "learning_rate": 0.0001776937618147448, "loss": 0.7265, "step": 470 }, { "epoch": 0.03, "learning_rate": 0.0001780718336483932, "loss": 0.7449, "step": 471 }, { "epoch": 0.03, "learning_rate": 0.0001784499054820416, "loss": 0.8693, "step": 472 }, { "epoch": 0.03, "learning_rate": 0.00017882797731569, "loss": 0.7793, "step": 473 }, { "epoch": 0.03, "learning_rate": 0.00017920604914933839, "loss": 0.8006, "step": 474 }, { "epoch": 0.03, "learning_rate": 0.00017958412098298678, "loss": 0.7447, "step": 475 }, { "epoch": 0.03, "learning_rate": 0.00017996219281663517, "loss": 0.7013, "step": 476 }, { "epoch": 0.03, "learning_rate": 0.00018034026465028357, "loss": 0.7461, "step": 477 }, { "epoch": 0.03, "learning_rate": 0.00018071833648393196, "loss": 0.6779, "step": 478 }, { "epoch": 0.03, "learning_rate": 0.00018109640831758035, "loss": 0.7201, "step": 479 }, { "epoch": 0.03, "learning_rate": 0.00018147448015122874, "loss": 0.7978, "step": 480 }, { "epoch": 0.03, "learning_rate": 0.00018185255198487714, "loss": 0.852, "step": 481 }, { "epoch": 0.03, "learning_rate": 0.00018223062381852553, "loss": 0.644, "step": 482 }, { "epoch": 0.03, "learning_rate": 0.00018260869565217392, "loss": 0.7835, "step": 483 }, { "epoch": 0.03, "learning_rate": 0.00018298676748582232, "loss": 0.7961, "step": 484 }, { "epoch": 0.03, "learning_rate": 0.0001833648393194707, "loss": 0.7547, "step": 485 }, { "epoch": 0.03, "learning_rate": 0.0001837429111531191, "loss": 0.7265, "step": 486 }, { "epoch": 0.03, "learning_rate": 0.0001841209829867675, "loss": 0.7022, "step": 487 }, { "epoch": 0.03, "learning_rate": 0.0001844990548204159, "loss": 0.7545, "step": 488 }, { "epoch": 0.03, "learning_rate": 0.00018487712665406428, "loss": 0.8499, "step": 489 }, { "epoch": 0.03, "learning_rate": 0.00018525519848771268, "loss": 0.7723, "step": 490 }, { "epoch": 0.03, "learning_rate": 0.00018563327032136107, "loss": 0.6731, "step": 491 }, { "epoch": 0.03, "learning_rate": 0.00018601134215500946, "loss": 0.7111, "step": 492 }, { "epoch": 0.03, "learning_rate": 0.00018638941398865786, "loss": 0.7398, "step": 493 }, { "epoch": 0.03, "learning_rate": 0.00018676748582230625, "loss": 0.6574, "step": 494 }, { "epoch": 0.03, "learning_rate": 0.00018714555765595464, "loss": 0.6958, "step": 495 }, { "epoch": 0.03, "learning_rate": 0.00018752362948960304, "loss": 0.7863, "step": 496 }, { "epoch": 0.03, "learning_rate": 0.00018790170132325143, "loss": 0.7219, "step": 497 }, { "epoch": 0.03, "learning_rate": 0.00018827977315689982, "loss": 0.7535, "step": 498 }, { "epoch": 0.03, "learning_rate": 0.00018865784499054821, "loss": 0.8022, "step": 499 }, { "epoch": 0.03, "learning_rate": 0.0001890359168241966, "loss": 0.7348, "step": 500 }, { "epoch": 0.03, "learning_rate": 0.000189413988657845, "loss": 0.6373, "step": 501 }, { "epoch": 0.03, "learning_rate": 0.0001897920604914934, "loss": 0.849, "step": 502 }, { "epoch": 0.03, "learning_rate": 0.0001901701323251418, "loss": 0.7987, "step": 503 }, { "epoch": 0.03, "learning_rate": 0.00019054820415879018, "loss": 0.7888, "step": 504 }, { "epoch": 0.03, "learning_rate": 0.00019092627599243857, "loss": 0.856, "step": 505 }, { "epoch": 0.03, "learning_rate": 0.00019130434782608697, "loss": 0.7104, "step": 506 }, { "epoch": 0.03, "learning_rate": 0.00019168241965973536, "loss": 0.7832, "step": 507 }, { "epoch": 0.03, "learning_rate": 0.00019206049149338375, "loss": 0.7891, "step": 508 }, { "epoch": 0.03, "learning_rate": 0.00019243856332703215, "loss": 0.743, "step": 509 }, { "epoch": 0.03, "learning_rate": 0.00019281663516068054, "loss": 0.7353, "step": 510 }, { "epoch": 0.03, "learning_rate": 0.00019319470699432893, "loss": 0.7519, "step": 511 }, { "epoch": 0.03, "learning_rate": 0.00019357277882797733, "loss": 0.7927, "step": 512 }, { "epoch": 0.03, "learning_rate": 0.00019395085066162572, "loss": 0.7391, "step": 513 }, { "epoch": 0.03, "learning_rate": 0.0001943289224952741, "loss": 0.7394, "step": 514 }, { "epoch": 0.03, "learning_rate": 0.0001947069943289225, "loss": 0.7619, "step": 515 }, { "epoch": 0.03, "learning_rate": 0.0001950850661625709, "loss": 0.7437, "step": 516 }, { "epoch": 0.03, "learning_rate": 0.0001954631379962193, "loss": 0.6868, "step": 517 }, { "epoch": 0.03, "learning_rate": 0.00019584120982986768, "loss": 0.7711, "step": 518 }, { "epoch": 0.03, "learning_rate": 0.00019621928166351608, "loss": 0.8135, "step": 519 }, { "epoch": 0.03, "learning_rate": 0.00019659735349716447, "loss": 0.7324, "step": 520 }, { "epoch": 0.03, "learning_rate": 0.00019697542533081286, "loss": 0.7296, "step": 521 }, { "epoch": 0.03, "learning_rate": 0.00019735349716446126, "loss": 0.7823, "step": 522 }, { "epoch": 0.03, "learning_rate": 0.00019773156899810965, "loss": 0.8658, "step": 523 }, { "epoch": 0.03, "learning_rate": 0.00019810964083175804, "loss": 0.7415, "step": 524 }, { "epoch": 0.03, "learning_rate": 0.00019848771266540644, "loss": 0.7415, "step": 525 }, { "epoch": 0.03, "learning_rate": 0.00019886578449905483, "loss": 0.6833, "step": 526 }, { "epoch": 0.03, "learning_rate": 0.00019924385633270322, "loss": 0.7189, "step": 527 }, { "epoch": 0.03, "learning_rate": 0.00019962192816635162, "loss": 0.8257, "step": 528 }, { "epoch": 0.03, "learning_rate": 0.0002, "loss": 0.6833, "step": 529 }, { "epoch": 0.03, "learning_rate": 0.00019999999831236885, "loss": 0.7394, "step": 530 }, { "epoch": 0.03, "learning_rate": 0.00019999999324947555, "loss": 0.7536, "step": 531 }, { "epoch": 0.03, "learning_rate": 0.0001999999848113202, "loss": 0.8425, "step": 532 }, { "epoch": 0.03, "learning_rate": 0.00019999997299790308, "loss": 0.8247, "step": 533 }, { "epoch": 0.03, "learning_rate": 0.00019999995780922461, "loss": 0.7779, "step": 534 }, { "epoch": 0.03, "learning_rate": 0.00019999993924528533, "loss": 0.8063, "step": 535 }, { "epoch": 0.03, "learning_rate": 0.0001999999173060858, "loss": 0.7847, "step": 536 }, { "epoch": 0.03, "learning_rate": 0.00019999989199162682, "loss": 0.7274, "step": 537 }, { "epoch": 0.03, "learning_rate": 0.00019999986330190926, "loss": 0.7614, "step": 538 }, { "epoch": 0.03, "learning_rate": 0.00019999983123693401, "loss": 0.7469, "step": 539 }, { "epoch": 0.03, "learning_rate": 0.00019999979579670224, "loss": 0.7836, "step": 540 }, { "epoch": 0.03, "learning_rate": 0.00019999975698121506, "loss": 0.832, "step": 541 }, { "epoch": 0.03, "learning_rate": 0.00019999971479047385, "loss": 0.6952, "step": 542 }, { "epoch": 0.03, "learning_rate": 0.00019999966922448, "loss": 0.7213, "step": 543 }, { "epoch": 0.03, "learning_rate": 0.00019999962028323503, "loss": 0.6913, "step": 544 }, { "epoch": 0.03, "learning_rate": 0.00019999956796674066, "loss": 0.72, "step": 545 }, { "epoch": 0.03, "learning_rate": 0.0001999995122749986, "loss": 0.6912, "step": 546 }, { "epoch": 0.03, "learning_rate": 0.00019999945320801072, "loss": 0.749, "step": 547 }, { "epoch": 0.03, "learning_rate": 0.00019999939076577905, "loss": 0.7436, "step": 548 }, { "epoch": 0.03, "learning_rate": 0.0001999993249483057, "loss": 0.6657, "step": 549 }, { "epoch": 0.03, "learning_rate": 0.00019999925575559283, "loss": 0.7532, "step": 550 }, { "epoch": 0.03, "learning_rate": 0.00019999918318764286, "loss": 0.7636, "step": 551 }, { "epoch": 0.03, "learning_rate": 0.00019999910724445818, "loss": 0.8463, "step": 552 }, { "epoch": 0.03, "learning_rate": 0.00019999902792604138, "loss": 0.673, "step": 553 }, { "epoch": 0.03, "learning_rate": 0.00019999894523239515, "loss": 0.7636, "step": 554 }, { "epoch": 0.03, "learning_rate": 0.00019999885916352223, "loss": 0.678, "step": 555 }, { "epoch": 0.03, "learning_rate": 0.00019999876971942557, "loss": 0.8294, "step": 556 }, { "epoch": 0.03, "learning_rate": 0.00019999867690010818, "loss": 0.7825, "step": 557 }, { "epoch": 0.03, "learning_rate": 0.00019999858070557318, "loss": 0.7335, "step": 558 }, { "epoch": 0.03, "learning_rate": 0.00019999848113582384, "loss": 0.7403, "step": 559 }, { "epoch": 0.03, "learning_rate": 0.0001999983781908635, "loss": 0.7712, "step": 560 }, { "epoch": 0.03, "learning_rate": 0.00019999827187069564, "loss": 0.8162, "step": 561 }, { "epoch": 0.03, "learning_rate": 0.00019999816217532382, "loss": 0.7227, "step": 562 }, { "epoch": 0.03, "learning_rate": 0.00019999804910475182, "loss": 0.7409, "step": 563 }, { "epoch": 0.03, "learning_rate": 0.0001999979326589834, "loss": 0.7171, "step": 564 }, { "epoch": 0.03, "learning_rate": 0.0001999978128380225, "loss": 0.817, "step": 565 }, { "epoch": 0.03, "learning_rate": 0.0001999976896418731, "loss": 0.8022, "step": 566 }, { "epoch": 0.03, "learning_rate": 0.00019999756307053948, "loss": 0.8586, "step": 567 }, { "epoch": 0.03, "learning_rate": 0.00019999743312402584, "loss": 0.7495, "step": 568 }, { "epoch": 0.03, "learning_rate": 0.0001999972998023366, "loss": 0.7503, "step": 569 }, { "epoch": 0.03, "learning_rate": 0.00019999716310547622, "loss": 0.8176, "step": 570 }, { "epoch": 0.03, "learning_rate": 0.00019999702303344932, "loss": 0.6689, "step": 571 }, { "epoch": 0.03, "learning_rate": 0.00019999687958626065, "loss": 0.6191, "step": 572 }, { "epoch": 0.03, "learning_rate": 0.00019999673276391506, "loss": 0.8295, "step": 573 }, { "epoch": 0.03, "learning_rate": 0.00019999658256641747, "loss": 0.7805, "step": 574 }, { "epoch": 0.03, "learning_rate": 0.000199996428993773, "loss": 0.7556, "step": 575 }, { "epoch": 0.03, "learning_rate": 0.00019999627204598673, "loss": 0.7585, "step": 576 }, { "epoch": 0.03, "learning_rate": 0.0001999961117230641, "loss": 0.7412, "step": 577 }, { "epoch": 0.03, "learning_rate": 0.00019999594802501042, "loss": 0.7405, "step": 578 }, { "epoch": 0.03, "learning_rate": 0.00019999578095183124, "loss": 0.736, "step": 579 }, { "epoch": 0.03, "learning_rate": 0.0001999956105035322, "loss": 0.8063, "step": 580 }, { "epoch": 0.03, "learning_rate": 0.00019999543668011908, "loss": 0.8848, "step": 581 }, { "epoch": 0.03, "learning_rate": 0.0001999952594815977, "loss": 0.7567, "step": 582 }, { "epoch": 0.03, "learning_rate": 0.00019999507890797408, "loss": 0.7591, "step": 583 }, { "epoch": 0.03, "learning_rate": 0.00019999489495925433, "loss": 0.7066, "step": 584 }, { "epoch": 0.03, "learning_rate": 0.00019999470763544457, "loss": 0.8139, "step": 585 }, { "epoch": 0.03, "learning_rate": 0.00019999451693655123, "loss": 0.8238, "step": 586 }, { "epoch": 0.03, "learning_rate": 0.0001999943228625807, "loss": 0.789, "step": 587 }, { "epoch": 0.03, "learning_rate": 0.00019999412541353947, "loss": 0.7843, "step": 588 }, { "epoch": 0.03, "learning_rate": 0.00019999392458943432, "loss": 0.7157, "step": 589 }, { "epoch": 0.03, "learning_rate": 0.00019999372039027194, "loss": 0.6616, "step": 590 }, { "epoch": 0.03, "learning_rate": 0.00019999351281605926, "loss": 0.7388, "step": 591 }, { "epoch": 0.03, "learning_rate": 0.0001999933018668033, "loss": 0.7759, "step": 592 }, { "epoch": 0.03, "learning_rate": 0.00019999308754251114, "loss": 0.8275, "step": 593 }, { "epoch": 0.03, "learning_rate": 0.00019999286984319004, "loss": 0.7319, "step": 594 }, { "epoch": 0.03, "learning_rate": 0.00019999264876884734, "loss": 0.7792, "step": 595 }, { "epoch": 0.03, "learning_rate": 0.00019999242431949048, "loss": 0.6831, "step": 596 }, { "epoch": 0.03, "learning_rate": 0.0001999921964951271, "loss": 0.794, "step": 597 }, { "epoch": 0.03, "learning_rate": 0.00019999196529576482, "loss": 0.75, "step": 598 }, { "epoch": 0.03, "learning_rate": 0.0001999917307214115, "loss": 0.7499, "step": 599 }, { "epoch": 0.03, "learning_rate": 0.00019999149277207502, "loss": 0.7107, "step": 600 }, { "epoch": 0.03, "learning_rate": 0.0001999912514477634, "loss": 0.7176, "step": 601 }, { "epoch": 0.03, "learning_rate": 0.00019999100674848483, "loss": 0.7424, "step": 602 }, { "epoch": 0.03, "learning_rate": 0.00019999075867424754, "loss": 0.7256, "step": 603 }, { "epoch": 0.03, "learning_rate": 0.00019999050722505993, "loss": 0.724, "step": 604 }, { "epoch": 0.03, "learning_rate": 0.00019999025240093044, "loss": 0.7149, "step": 605 }, { "epoch": 0.03, "learning_rate": 0.0001999899942018677, "loss": 0.7571, "step": 606 }, { "epoch": 0.03, "learning_rate": 0.00019998973262788043, "loss": 0.7918, "step": 607 }, { "epoch": 0.03, "learning_rate": 0.00019998946767897744, "loss": 0.6789, "step": 608 }, { "epoch": 0.03, "learning_rate": 0.00019998919935516768, "loss": 0.8104, "step": 609 }, { "epoch": 0.03, "learning_rate": 0.00019998892765646026, "loss": 0.7803, "step": 610 }, { "epoch": 0.03, "learning_rate": 0.00019998865258286422, "loss": 0.7634, "step": 611 }, { "epoch": 0.03, "learning_rate": 0.00019998837413438898, "loss": 0.8564, "step": 612 }, { "epoch": 0.03, "learning_rate": 0.00019998809231104388, "loss": 0.6888, "step": 613 }, { "epoch": 0.03, "learning_rate": 0.00019998780711283842, "loss": 0.797, "step": 614 }, { "epoch": 0.03, "learning_rate": 0.00019998751853978223, "loss": 0.8302, "step": 615 }, { "epoch": 0.03, "learning_rate": 0.0001999872265918851, "loss": 0.7648, "step": 616 }, { "epoch": 0.03, "learning_rate": 0.00019998693126915682, "loss": 0.7352, "step": 617 }, { "epoch": 0.04, "learning_rate": 0.0001999866325716074, "loss": 0.6563, "step": 618 }, { "epoch": 0.04, "learning_rate": 0.0001999863304992469, "loss": 0.764, "step": 619 }, { "epoch": 0.04, "learning_rate": 0.00019998602505208552, "loss": 0.7302, "step": 620 }, { "epoch": 0.04, "learning_rate": 0.00019998571623013357, "loss": 0.7475, "step": 621 }, { "epoch": 0.04, "learning_rate": 0.00019998540403340148, "loss": 0.786, "step": 622 }, { "epoch": 0.04, "learning_rate": 0.0001999850884618998, "loss": 0.6831, "step": 623 }, { "epoch": 0.04, "learning_rate": 0.00019998476951563915, "loss": 0.7615, "step": 624 }, { "epoch": 0.04, "learning_rate": 0.00019998444719463029, "loss": 0.7126, "step": 625 }, { "epoch": 0.04, "learning_rate": 0.00019998412149888412, "loss": 0.8215, "step": 626 }, { "epoch": 0.04, "learning_rate": 0.00019998379242841167, "loss": 0.8811, "step": 627 }, { "epoch": 0.04, "learning_rate": 0.00019998345998322397, "loss": 0.707, "step": 628 }, { "epoch": 0.04, "learning_rate": 0.00019998312416333227, "loss": 0.7653, "step": 629 }, { "epoch": 0.04, "learning_rate": 0.00019998278496874793, "loss": 0.773, "step": 630 }, { "epoch": 0.04, "learning_rate": 0.00019998244239948241, "loss": 0.7968, "step": 631 }, { "epoch": 0.04, "learning_rate": 0.00019998209645554721, "loss": 0.7454, "step": 632 }, { "epoch": 0.04, "learning_rate": 0.00019998174713695407, "loss": 0.8389, "step": 633 }, { "epoch": 0.04, "learning_rate": 0.00019998139444371473, "loss": 0.812, "step": 634 }, { "epoch": 0.04, "learning_rate": 0.00019998103837584114, "loss": 0.7154, "step": 635 }, { "epoch": 0.04, "learning_rate": 0.00019998067893334527, "loss": 0.7698, "step": 636 }, { "epoch": 0.04, "learning_rate": 0.0001999803161162393, "loss": 0.7977, "step": 637 }, { "epoch": 0.04, "learning_rate": 0.00019997994992453544, "loss": 0.6974, "step": 638 }, { "epoch": 0.04, "learning_rate": 0.00019997958035824606, "loss": 0.8081, "step": 639 }, { "epoch": 0.04, "learning_rate": 0.00019997920741738362, "loss": 0.7217, "step": 640 }, { "epoch": 0.04, "learning_rate": 0.00019997883110196076, "loss": 0.7946, "step": 641 }, { "epoch": 0.04, "learning_rate": 0.00019997845141199013, "loss": 0.8673, "step": 642 }, { "epoch": 0.04, "learning_rate": 0.00019997806834748456, "loss": 0.7699, "step": 643 }, { "epoch": 0.04, "learning_rate": 0.00019997768190845697, "loss": 0.697, "step": 644 }, { "epoch": 0.04, "learning_rate": 0.00019997729209492044, "loss": 0.7111, "step": 645 }, { "epoch": 0.04, "learning_rate": 0.0001999768989068881, "loss": 0.7474, "step": 646 }, { "epoch": 0.04, "learning_rate": 0.0001999765023443732, "loss": 0.6913, "step": 647 }, { "epoch": 0.04, "learning_rate": 0.00019997610240738917, "loss": 0.7157, "step": 648 }, { "epoch": 0.04, "learning_rate": 0.00019997569909594947, "loss": 0.75, "step": 649 }, { "epoch": 0.04, "learning_rate": 0.00019997529241006773, "loss": 0.7109, "step": 650 }, { "epoch": 0.04, "learning_rate": 0.00019997488234975766, "loss": 0.8038, "step": 651 }, { "epoch": 0.04, "learning_rate": 0.00019997446891503317, "loss": 0.767, "step": 652 }, { "epoch": 0.04, "learning_rate": 0.00019997405210590814, "loss": 0.7561, "step": 653 }, { "epoch": 0.04, "learning_rate": 0.00019997363192239664, "loss": 0.7091, "step": 654 }, { "epoch": 0.04, "learning_rate": 0.0001999732083645129, "loss": 0.6963, "step": 655 }, { "epoch": 0.04, "learning_rate": 0.00019997278143227116, "loss": 0.7364, "step": 656 }, { "epoch": 0.04, "learning_rate": 0.00019997235112568588, "loss": 0.594, "step": 657 }, { "epoch": 0.04, "learning_rate": 0.00019997191744477153, "loss": 0.7173, "step": 658 }, { "epoch": 0.04, "learning_rate": 0.0001999714803895428, "loss": 0.7801, "step": 659 }, { "epoch": 0.04, "learning_rate": 0.00019997103996001445, "loss": 0.7262, "step": 660 }, { "epoch": 0.04, "learning_rate": 0.00019997059615620128, "loss": 0.7762, "step": 661 }, { "epoch": 0.04, "learning_rate": 0.00019997014897811833, "loss": 0.7289, "step": 662 }, { "epoch": 0.04, "learning_rate": 0.00019996969842578065, "loss": 0.7143, "step": 663 }, { "epoch": 0.04, "learning_rate": 0.0001999692444992035, "loss": 0.768, "step": 664 }, { "epoch": 0.04, "learning_rate": 0.00019996878719840213, "loss": 0.7468, "step": 665 }, { "epoch": 0.04, "learning_rate": 0.00019996832652339203, "loss": 0.681, "step": 666 }, { "epoch": 0.04, "learning_rate": 0.00019996786247418875, "loss": 0.6802, "step": 667 }, { "epoch": 0.04, "learning_rate": 0.0001999673950508079, "loss": 0.8638, "step": 668 }, { "epoch": 0.04, "learning_rate": 0.00019996692425326533, "loss": 0.7162, "step": 669 }, { "epoch": 0.04, "learning_rate": 0.00019996645008157687, "loss": 0.6875, "step": 670 }, { "epoch": 0.04, "learning_rate": 0.00019996597253575855, "loss": 0.7656, "step": 671 }, { "epoch": 0.04, "learning_rate": 0.0001999654916158265, "loss": 0.8669, "step": 672 }, { "epoch": 0.04, "learning_rate": 0.00019996500732179695, "loss": 0.8672, "step": 673 }, { "epoch": 0.04, "learning_rate": 0.00019996451965368622, "loss": 0.7437, "step": 674 }, { "epoch": 0.04, "learning_rate": 0.00019996402861151077, "loss": 0.7346, "step": 675 }, { "epoch": 0.04, "learning_rate": 0.0001999635341952872, "loss": 0.7312, "step": 676 }, { "epoch": 0.04, "learning_rate": 0.00019996303640503217, "loss": 0.8144, "step": 677 }, { "epoch": 0.04, "learning_rate": 0.00019996253524076253, "loss": 0.6515, "step": 678 }, { "epoch": 0.04, "learning_rate": 0.00019996203070249516, "loss": 0.7021, "step": 679 }, { "epoch": 0.04, "learning_rate": 0.00019996152279024707, "loss": 0.8277, "step": 680 }, { "epoch": 0.04, "learning_rate": 0.00019996101150403543, "loss": 0.717, "step": 681 }, { "epoch": 0.04, "learning_rate": 0.0001999604968438775, "loss": 0.7006, "step": 682 }, { "epoch": 0.04, "learning_rate": 0.00019995997880979067, "loss": 0.7947, "step": 683 }, { "epoch": 0.04, "learning_rate": 0.00019995945740179237, "loss": 0.7324, "step": 684 }, { "epoch": 0.04, "learning_rate": 0.00019995893261990025, "loss": 0.7389, "step": 685 }, { "epoch": 0.04, "learning_rate": 0.000199958404464132, "loss": 0.7328, "step": 686 }, { "epoch": 0.04, "learning_rate": 0.00019995787293450543, "loss": 0.6743, "step": 687 }, { "epoch": 0.04, "learning_rate": 0.00019995733803103853, "loss": 0.7632, "step": 688 }, { "epoch": 0.04, "learning_rate": 0.0001999567997537493, "loss": 0.755, "step": 689 }, { "epoch": 0.04, "learning_rate": 0.00019995625810265594, "loss": 0.7098, "step": 690 }, { "epoch": 0.04, "learning_rate": 0.0001999557130777767, "loss": 0.7428, "step": 691 }, { "epoch": 0.04, "learning_rate": 0.00019995516467913004, "loss": 0.838, "step": 692 }, { "epoch": 0.04, "learning_rate": 0.00019995461290673442, "loss": 0.7216, "step": 693 }, { "epoch": 0.04, "learning_rate": 0.00019995405776060846, "loss": 0.746, "step": 694 }, { "epoch": 0.04, "learning_rate": 0.0001999534992407709, "loss": 0.7921, "step": 695 }, { "epoch": 0.04, "learning_rate": 0.00019995293734724064, "loss": 0.688, "step": 696 }, { "epoch": 0.04, "learning_rate": 0.00019995237208003658, "loss": 0.8074, "step": 697 }, { "epoch": 0.04, "learning_rate": 0.00019995180343917784, "loss": 0.8191, "step": 698 }, { "epoch": 0.04, "learning_rate": 0.00019995123142468359, "loss": 0.6685, "step": 699 }, { "epoch": 0.04, "learning_rate": 0.00019995065603657316, "loss": 0.7194, "step": 700 }, { "epoch": 0.04, "learning_rate": 0.00019995007727486596, "loss": 0.7529, "step": 701 }, { "epoch": 0.04, "learning_rate": 0.0001999494951395815, "loss": 0.8191, "step": 702 }, { "epoch": 0.04, "learning_rate": 0.00019994890963073947, "loss": 0.8197, "step": 703 }, { "epoch": 0.04, "learning_rate": 0.00019994832074835963, "loss": 1.0235, "step": 704 }, { "epoch": 0.04, "learning_rate": 0.00019994772849246182, "loss": 0.6767, "step": 705 }, { "epoch": 0.04, "learning_rate": 0.00019994713286306605, "loss": 0.6445, "step": 706 }, { "epoch": 0.04, "learning_rate": 0.0001999465338601924, "loss": 0.8069, "step": 707 }, { "epoch": 0.04, "learning_rate": 0.00019994593148386117, "loss": 0.8647, "step": 708 }, { "epoch": 0.04, "learning_rate": 0.00019994532573409262, "loss": 0.7295, "step": 709 }, { "epoch": 0.04, "learning_rate": 0.0001999447166109072, "loss": 0.8574, "step": 710 }, { "epoch": 0.04, "learning_rate": 0.00019994410411432547, "loss": 0.7932, "step": 711 }, { "epoch": 0.04, "learning_rate": 0.00019994348824436812, "loss": 0.8193, "step": 712 }, { "epoch": 0.04, "learning_rate": 0.00019994286900105592, "loss": 0.6837, "step": 713 }, { "epoch": 0.04, "learning_rate": 0.0001999422463844098, "loss": 0.7518, "step": 714 }, { "epoch": 0.04, "learning_rate": 0.00019994162039445074, "loss": 0.6279, "step": 715 }, { "epoch": 0.04, "learning_rate": 0.0001999409910311999, "loss": 0.8235, "step": 716 }, { "epoch": 0.04, "learning_rate": 0.0001999403582946785, "loss": 0.7708, "step": 717 }, { "epoch": 0.04, "learning_rate": 0.0001999397221849079, "loss": 0.715, "step": 718 }, { "epoch": 0.04, "learning_rate": 0.0001999390827019096, "loss": 0.8184, "step": 719 }, { "epoch": 0.04, "learning_rate": 0.00019993843984570511, "loss": 0.697, "step": 720 }, { "epoch": 0.04, "learning_rate": 0.0001999377936163162, "loss": 0.7083, "step": 721 }, { "epoch": 0.04, "learning_rate": 0.0001999371440137647, "loss": 0.7587, "step": 722 }, { "epoch": 0.04, "learning_rate": 0.00019993649103807243, "loss": 0.6323, "step": 723 }, { "epoch": 0.04, "learning_rate": 0.00019993583468926155, "loss": 0.7841, "step": 724 }, { "epoch": 0.04, "learning_rate": 0.00019993517496735412, "loss": 0.7084, "step": 725 }, { "epoch": 0.04, "learning_rate": 0.00019993451187237248, "loss": 0.7197, "step": 726 }, { "epoch": 0.04, "learning_rate": 0.00019993384540433894, "loss": 0.7642, "step": 727 }, { "epoch": 0.04, "learning_rate": 0.00019993317556327603, "loss": 0.7807, "step": 728 }, { "epoch": 0.04, "learning_rate": 0.00019993250234920636, "loss": 0.7437, "step": 729 }, { "epoch": 0.04, "learning_rate": 0.0001999318257621527, "loss": 0.7565, "step": 730 }, { "epoch": 0.04, "learning_rate": 0.00019993114580213779, "loss": 0.8735, "step": 731 }, { "epoch": 0.04, "learning_rate": 0.00019993046246918466, "loss": 0.7612, "step": 732 }, { "epoch": 0.04, "learning_rate": 0.00019992977576331633, "loss": 0.639, "step": 733 }, { "epoch": 0.04, "learning_rate": 0.000199929085684556, "loss": 0.7033, "step": 734 }, { "epoch": 0.04, "learning_rate": 0.00019992839223292695, "loss": 0.6156, "step": 735 }, { "epoch": 0.04, "learning_rate": 0.00019992769540845258, "loss": 0.7596, "step": 736 }, { "epoch": 0.04, "learning_rate": 0.00019992699521115644, "loss": 0.7316, "step": 737 }, { "epoch": 0.04, "learning_rate": 0.0001999262916410621, "loss": 0.7271, "step": 738 }, { "epoch": 0.04, "learning_rate": 0.00019992558469819342, "loss": 0.621, "step": 739 }, { "epoch": 0.04, "learning_rate": 0.00019992487438257414, "loss": 0.7823, "step": 740 }, { "epoch": 0.04, "learning_rate": 0.00019992416069422833, "loss": 0.7904, "step": 741 }, { "epoch": 0.04, "learning_rate": 0.00019992344363318, "loss": 0.6666, "step": 742 }, { "epoch": 0.04, "learning_rate": 0.00019992272319945337, "loss": 0.7463, "step": 743 }, { "epoch": 0.04, "learning_rate": 0.00019992199939307281, "loss": 0.7638, "step": 744 }, { "epoch": 0.04, "learning_rate": 0.00019992127221406275, "loss": 0.8649, "step": 745 }, { "epoch": 0.04, "learning_rate": 0.0001999205416624476, "loss": 0.7269, "step": 746 }, { "epoch": 0.04, "learning_rate": 0.00019991980773825221, "loss": 0.8053, "step": 747 }, { "epoch": 0.04, "learning_rate": 0.00019991907044150122, "loss": 0.6935, "step": 748 }, { "epoch": 0.04, "learning_rate": 0.00019991832977221956, "loss": 0.7946, "step": 749 }, { "epoch": 0.04, "learning_rate": 0.00019991758573043223, "loss": 0.7702, "step": 750 }, { "epoch": 0.04, "learning_rate": 0.00019991683831616436, "loss": 0.8265, "step": 751 }, { "epoch": 0.04, "learning_rate": 0.00019991608752944112, "loss": 0.7677, "step": 752 }, { "epoch": 0.04, "learning_rate": 0.00019991533337028786, "loss": 0.7176, "step": 753 }, { "epoch": 0.04, "learning_rate": 0.0001999145758387301, "loss": 0.7919, "step": 754 }, { "epoch": 0.04, "learning_rate": 0.00019991381493479336, "loss": 0.6741, "step": 755 }, { "epoch": 0.04, "learning_rate": 0.00019991305065850337, "loss": 0.7282, "step": 756 }, { "epoch": 0.04, "learning_rate": 0.00019991228300988585, "loss": 0.7304, "step": 757 }, { "epoch": 0.04, "learning_rate": 0.00019991151198896673, "loss": 0.7949, "step": 758 }, { "epoch": 0.04, "learning_rate": 0.00019991073759577211, "loss": 0.7375, "step": 759 }, { "epoch": 0.04, "learning_rate": 0.00019990995983032802, "loss": 0.6679, "step": 760 }, { "epoch": 0.04, "learning_rate": 0.00019990917869266078, "loss": 0.7771, "step": 761 }, { "epoch": 0.04, "learning_rate": 0.00019990839418279675, "loss": 0.6819, "step": 762 }, { "epoch": 0.04, "learning_rate": 0.00019990760630076237, "loss": 0.773, "step": 763 }, { "epoch": 0.04, "learning_rate": 0.00019990681504658428, "loss": 0.6519, "step": 764 }, { "epoch": 0.04, "learning_rate": 0.00019990602042028915, "loss": 0.8086, "step": 765 }, { "epoch": 0.04, "learning_rate": 0.00019990522242190387, "loss": 0.7437, "step": 766 }, { "epoch": 0.04, "learning_rate": 0.00019990442105145525, "loss": 0.7577, "step": 767 }, { "epoch": 0.04, "learning_rate": 0.00019990361630897047, "loss": 0.7422, "step": 768 }, { "epoch": 0.04, "learning_rate": 0.0001999028081944766, "loss": 0.8053, "step": 769 }, { "epoch": 0.04, "learning_rate": 0.00019990199670800096, "loss": 0.7849, "step": 770 }, { "epoch": 0.04, "learning_rate": 0.0001999011818495709, "loss": 0.7641, "step": 771 }, { "epoch": 0.04, "learning_rate": 0.000199900363619214, "loss": 0.6814, "step": 772 }, { "epoch": 0.04, "learning_rate": 0.0001998995420169578, "loss": 0.7401, "step": 773 }, { "epoch": 0.04, "learning_rate": 0.00019989871704283006, "loss": 0.7085, "step": 774 }, { "epoch": 0.04, "learning_rate": 0.00019989788869685863, "loss": 0.7021, "step": 775 }, { "epoch": 0.04, "learning_rate": 0.00019989705697907149, "loss": 0.6708, "step": 776 }, { "epoch": 0.04, "learning_rate": 0.00019989622188949663, "loss": 0.8314, "step": 777 }, { "epoch": 0.04, "learning_rate": 0.00019989538342816236, "loss": 0.7257, "step": 778 }, { "epoch": 0.04, "learning_rate": 0.0001998945415950969, "loss": 0.8087, "step": 779 }, { "epoch": 0.04, "learning_rate": 0.00019989369639032863, "loss": 0.7582, "step": 780 }, { "epoch": 0.04, "learning_rate": 0.00019989284781388617, "loss": 0.833, "step": 781 }, { "epoch": 0.04, "learning_rate": 0.0001998919958657981, "loss": 0.7344, "step": 782 }, { "epoch": 0.04, "learning_rate": 0.00019989114054609318, "loss": 0.8061, "step": 783 }, { "epoch": 0.04, "learning_rate": 0.0001998902818548003, "loss": 0.7998, "step": 784 }, { "epoch": 0.04, "learning_rate": 0.00019988941979194846, "loss": 0.7682, "step": 785 }, { "epoch": 0.04, "learning_rate": 0.00019988855435756672, "loss": 0.7424, "step": 786 }, { "epoch": 0.04, "learning_rate": 0.0001998876855516843, "loss": 0.7187, "step": 787 }, { "epoch": 0.04, "learning_rate": 0.00019988681337433054, "loss": 0.7182, "step": 788 }, { "epoch": 0.04, "learning_rate": 0.00019988593782553483, "loss": 0.7577, "step": 789 }, { "epoch": 0.04, "learning_rate": 0.0001998850589053268, "loss": 0.8574, "step": 790 }, { "epoch": 0.04, "learning_rate": 0.00019988417661373603, "loss": 0.6783, "step": 791 }, { "epoch": 0.04, "learning_rate": 0.00019988329095079239, "loss": 0.7468, "step": 792 }, { "epoch": 0.04, "learning_rate": 0.0001998824019165257, "loss": 0.9002, "step": 793 }, { "epoch": 0.05, "learning_rate": 0.000199881509510966, "loss": 0.6674, "step": 794 }, { "epoch": 0.05, "learning_rate": 0.0001998806137341434, "loss": 0.8131, "step": 795 }, { "epoch": 0.05, "learning_rate": 0.00019987971458608817, "loss": 0.8132, "step": 796 }, { "epoch": 0.05, "learning_rate": 0.0001998788120668306, "loss": 0.8159, "step": 797 }, { "epoch": 0.05, "learning_rate": 0.00019987790617640117, "loss": 0.7789, "step": 798 }, { "epoch": 0.05, "learning_rate": 0.00019987699691483048, "loss": 0.7073, "step": 799 }, { "epoch": 0.05, "learning_rate": 0.00019987608428214922, "loss": 0.7544, "step": 800 }, { "epoch": 0.05, "learning_rate": 0.00019987516827838818, "loss": 0.7113, "step": 801 }, { "epoch": 0.05, "learning_rate": 0.00019987424890357825, "loss": 0.7757, "step": 802 }, { "epoch": 0.05, "learning_rate": 0.0001998733261577505, "loss": 0.8215, "step": 803 }, { "epoch": 0.05, "learning_rate": 0.0001998724000409361, "loss": 0.7067, "step": 804 }, { "epoch": 0.05, "learning_rate": 0.0001998714705531662, "loss": 0.7913, "step": 805 }, { "epoch": 0.05, "learning_rate": 0.00019987053769447233, "loss": 0.6642, "step": 806 }, { "epoch": 0.05, "learning_rate": 0.00019986960146488585, "loss": 0.7222, "step": 807 }, { "epoch": 0.05, "learning_rate": 0.0001998686618644384, "loss": 0.8425, "step": 808 }, { "epoch": 0.05, "learning_rate": 0.00019986771889316172, "loss": 0.7744, "step": 809 }, { "epoch": 0.05, "learning_rate": 0.0001998667725510876, "loss": 0.7533, "step": 810 }, { "epoch": 0.05, "learning_rate": 0.000199865822838248, "loss": 0.7354, "step": 811 }, { "epoch": 0.05, "learning_rate": 0.00019986486975467495, "loss": 0.8434, "step": 812 }, { "epoch": 0.05, "learning_rate": 0.00019986391330040066, "loss": 0.7426, "step": 813 }, { "epoch": 0.05, "learning_rate": 0.0001998629534754574, "loss": 0.7362, "step": 814 }, { "epoch": 0.05, "learning_rate": 0.00019986199027987757, "loss": 0.8317, "step": 815 }, { "epoch": 0.05, "learning_rate": 0.00019986102371369362, "loss": 0.7374, "step": 816 }, { "epoch": 0.05, "learning_rate": 0.00019986005377693825, "loss": 0.6741, "step": 817 }, { "epoch": 0.05, "learning_rate": 0.0001998590804696442, "loss": 0.6995, "step": 818 }, { "epoch": 0.05, "learning_rate": 0.00019985810379184426, "loss": 0.6832, "step": 819 }, { "epoch": 0.05, "learning_rate": 0.00019985712374357144, "loss": 0.7027, "step": 820 }, { "epoch": 0.05, "learning_rate": 0.00019985614032485882, "loss": 0.7401, "step": 821 }, { "epoch": 0.05, "learning_rate": 0.00019985515353573957, "loss": 0.7695, "step": 822 }, { "epoch": 0.05, "learning_rate": 0.000199854163376247, "loss": 0.6936, "step": 823 }, { "epoch": 0.05, "learning_rate": 0.00019985316984641455, "loss": 0.8704, "step": 824 }, { "epoch": 0.05, "learning_rate": 0.00019985217294627577, "loss": 0.7844, "step": 825 }, { "epoch": 0.05, "learning_rate": 0.00019985117267586424, "loss": 0.798, "step": 826 }, { "epoch": 0.05, "learning_rate": 0.00019985016903521378, "loss": 0.7155, "step": 827 }, { "epoch": 0.05, "learning_rate": 0.00019984916202435825, "loss": 0.7077, "step": 828 }, { "epoch": 0.05, "learning_rate": 0.00019984815164333163, "loss": 0.7694, "step": 829 }, { "epoch": 0.05, "learning_rate": 0.00019984713789216802, "loss": 0.7846, "step": 830 }, { "epoch": 0.05, "learning_rate": 0.00019984612077090169, "loss": 0.8021, "step": 831 }, { "epoch": 0.05, "learning_rate": 0.00019984510027956689, "loss": 0.74, "step": 832 }, { "epoch": 0.05, "learning_rate": 0.00019984407641819812, "loss": 0.6782, "step": 833 }, { "epoch": 0.05, "learning_rate": 0.0001998430491868299, "loss": 0.8044, "step": 834 }, { "epoch": 0.05, "learning_rate": 0.00019984201858549693, "loss": 0.7415, "step": 835 }, { "epoch": 0.05, "learning_rate": 0.000199840984614234, "loss": 0.8336, "step": 836 }, { "epoch": 0.05, "learning_rate": 0.000199839947273076, "loss": 0.8165, "step": 837 }, { "epoch": 0.05, "learning_rate": 0.00019983890656205793, "loss": 0.7178, "step": 838 }, { "epoch": 0.05, "learning_rate": 0.0001998378624812149, "loss": 0.7961, "step": 839 }, { "epoch": 0.05, "learning_rate": 0.0001998368150305822, "loss": 0.7743, "step": 840 }, { "epoch": 0.05, "learning_rate": 0.00019983576421019517, "loss": 0.6848, "step": 841 }, { "epoch": 0.05, "learning_rate": 0.00019983471002008924, "loss": 0.7588, "step": 842 }, { "epoch": 0.05, "learning_rate": 0.00019983365246030007, "loss": 0.7391, "step": 843 }, { "epoch": 0.05, "learning_rate": 0.00019983259153086327, "loss": 0.7452, "step": 844 }, { "epoch": 0.05, "learning_rate": 0.00019983152723181472, "loss": 0.773, "step": 845 }, { "epoch": 0.05, "learning_rate": 0.00019983045956319025, "loss": 0.7931, "step": 846 }, { "epoch": 0.05, "learning_rate": 0.000199829388525026, "loss": 0.714, "step": 847 }, { "epoch": 0.05, "learning_rate": 0.00019982831411735806, "loss": 0.8368, "step": 848 }, { "epoch": 0.05, "learning_rate": 0.00019982723634022272, "loss": 0.7884, "step": 849 }, { "epoch": 0.05, "learning_rate": 0.0001998261551936563, "loss": 0.723, "step": 850 }, { "epoch": 0.05, "learning_rate": 0.0001998250706776954, "loss": 0.8065, "step": 851 }, { "epoch": 0.05, "learning_rate": 0.00019982398279237655, "loss": 0.7562, "step": 852 }, { "epoch": 0.05, "learning_rate": 0.00019982289153773646, "loss": 0.6875, "step": 853 }, { "epoch": 0.05, "learning_rate": 0.000199821796913812, "loss": 0.8078, "step": 854 }, { "epoch": 0.05, "learning_rate": 0.0001998206989206401, "loss": 0.7367, "step": 855 }, { "epoch": 0.05, "learning_rate": 0.00019981959755825782, "loss": 0.7462, "step": 856 }, { "epoch": 0.05, "learning_rate": 0.00019981849282670234, "loss": 0.697, "step": 857 }, { "epoch": 0.05, "learning_rate": 0.00019981738472601092, "loss": 0.7946, "step": 858 }, { "epoch": 0.05, "learning_rate": 0.00019981627325622102, "loss": 0.7051, "step": 859 }, { "epoch": 0.05, "learning_rate": 0.00019981515841737012, "loss": 0.8624, "step": 860 }, { "epoch": 0.05, "learning_rate": 0.00019981404020949582, "loss": 0.7656, "step": 861 }, { "epoch": 0.05, "learning_rate": 0.00019981291863263592, "loss": 0.7805, "step": 862 }, { "epoch": 0.05, "learning_rate": 0.00019981179368682824, "loss": 0.7363, "step": 863 }, { "epoch": 0.05, "learning_rate": 0.00019981066537211073, "loss": 0.7309, "step": 864 }, { "epoch": 0.05, "learning_rate": 0.0001998095336885215, "loss": 0.6779, "step": 865 }, { "epoch": 0.05, "learning_rate": 0.0001998083986360988, "loss": 0.8641, "step": 866 }, { "epoch": 0.05, "learning_rate": 0.00019980726021488082, "loss": 0.6775, "step": 867 }, { "epoch": 0.05, "learning_rate": 0.00019980611842490613, "loss": 0.8176, "step": 868 }, { "epoch": 0.05, "learning_rate": 0.00019980497326621316, "loss": 0.803, "step": 869 }, { "epoch": 0.05, "learning_rate": 0.00019980382473884056, "loss": 0.6757, "step": 870 }, { "epoch": 0.05, "learning_rate": 0.00019980267284282717, "loss": 0.7276, "step": 871 }, { "epoch": 0.05, "learning_rate": 0.0001998015175782118, "loss": 0.854, "step": 872 }, { "epoch": 0.05, "learning_rate": 0.0001998003589450335, "loss": 0.76, "step": 873 }, { "epoch": 0.05, "learning_rate": 0.00019979919694333134, "loss": 0.7699, "step": 874 }, { "epoch": 0.05, "learning_rate": 0.00019979803157314457, "loss": 0.7002, "step": 875 }, { "epoch": 0.05, "learning_rate": 0.00019979686283451247, "loss": 0.6984, "step": 876 }, { "epoch": 0.05, "learning_rate": 0.00019979569072747452, "loss": 0.7504, "step": 877 }, { "epoch": 0.05, "learning_rate": 0.0001997945152520703, "loss": 0.7513, "step": 878 }, { "epoch": 0.05, "learning_rate": 0.00019979333640833947, "loss": 0.7761, "step": 879 }, { "epoch": 0.05, "learning_rate": 0.00019979215419632182, "loss": 0.7658, "step": 880 }, { "epoch": 0.05, "learning_rate": 0.00019979096861605724, "loss": 0.831, "step": 881 }, { "epoch": 0.05, "learning_rate": 0.00019978977966758578, "loss": 0.7822, "step": 882 }, { "epoch": 0.05, "learning_rate": 0.00019978858735094753, "loss": 0.689, "step": 883 }, { "epoch": 0.05, "learning_rate": 0.00019978739166618277, "loss": 0.6182, "step": 884 }, { "epoch": 0.05, "learning_rate": 0.00019978619261333183, "loss": 0.7169, "step": 885 }, { "epoch": 0.05, "learning_rate": 0.00019978499019243518, "loss": 0.6652, "step": 886 }, { "epoch": 0.05, "learning_rate": 0.00019978378440353343, "loss": 0.6882, "step": 887 }, { "epoch": 0.05, "learning_rate": 0.00019978257524666724, "loss": 0.8629, "step": 888 }, { "epoch": 0.05, "learning_rate": 0.00019978136272187747, "loss": 0.7027, "step": 889 }, { "epoch": 0.05, "learning_rate": 0.000199780146829205, "loss": 0.7876, "step": 890 }, { "epoch": 0.05, "learning_rate": 0.0001997789275686909, "loss": 0.7217, "step": 891 }, { "epoch": 0.05, "learning_rate": 0.00019977770494037633, "loss": 0.78, "step": 892 }, { "epoch": 0.05, "learning_rate": 0.00019977647894430255, "loss": 0.7401, "step": 893 }, { "epoch": 0.05, "learning_rate": 0.0001997752495805109, "loss": 0.7401, "step": 894 }, { "epoch": 0.05, "learning_rate": 0.00019977401684904291, "loss": 0.7867, "step": 895 }, { "epoch": 0.05, "learning_rate": 0.0001997727807499402, "loss": 0.7164, "step": 896 }, { "epoch": 0.05, "learning_rate": 0.00019977154128324447, "loss": 0.7104, "step": 897 }, { "epoch": 0.05, "learning_rate": 0.00019977029844899758, "loss": 0.7321, "step": 898 }, { "epoch": 0.05, "learning_rate": 0.00019976905224724142, "loss": 0.7247, "step": 899 }, { "epoch": 0.05, "learning_rate": 0.00019976780267801813, "loss": 0.734, "step": 900 }, { "epoch": 0.05, "learning_rate": 0.00019976654974136984, "loss": 0.6461, "step": 901 }, { "epoch": 0.05, "learning_rate": 0.00019976529343733882, "loss": 0.7709, "step": 902 }, { "epoch": 0.05, "learning_rate": 0.00019976403376596754, "loss": 0.7653, "step": 903 }, { "epoch": 0.05, "learning_rate": 0.00019976277072729845, "loss": 0.7961, "step": 904 }, { "epoch": 0.05, "learning_rate": 0.00019976150432137423, "loss": 0.7298, "step": 905 }, { "epoch": 0.05, "learning_rate": 0.0001997602345482376, "loss": 0.7151, "step": 906 }, { "epoch": 0.05, "learning_rate": 0.00019975896140793142, "loss": 0.833, "step": 907 }, { "epoch": 0.05, "learning_rate": 0.00019975768490049868, "loss": 0.7735, "step": 908 }, { "epoch": 0.05, "learning_rate": 0.00019975640502598244, "loss": 0.7516, "step": 909 }, { "epoch": 0.05, "learning_rate": 0.00019975512178442591, "loss": 0.6627, "step": 910 }, { "epoch": 0.05, "learning_rate": 0.0001997538351758724, "loss": 0.728, "step": 911 }, { "epoch": 0.05, "learning_rate": 0.00019975254520036533, "loss": 0.741, "step": 912 }, { "epoch": 0.05, "learning_rate": 0.00019975125185794825, "loss": 0.848, "step": 913 }, { "epoch": 0.05, "learning_rate": 0.00019974995514866485, "loss": 0.7586, "step": 914 }, { "epoch": 0.05, "learning_rate": 0.00019974865507255883, "loss": 0.7438, "step": 915 }, { "epoch": 0.05, "learning_rate": 0.0001997473516296741, "loss": 0.7305, "step": 916 }, { "epoch": 0.05, "learning_rate": 0.00019974604482005464, "loss": 0.7298, "step": 917 }, { "epoch": 0.05, "learning_rate": 0.0001997447346437446, "loss": 0.6967, "step": 918 }, { "epoch": 0.05, "learning_rate": 0.00019974342110078817, "loss": 0.6973, "step": 919 }, { "epoch": 0.05, "learning_rate": 0.00019974210419122969, "loss": 0.7226, "step": 920 }, { "epoch": 0.05, "learning_rate": 0.0001997407839151136, "loss": 0.7895, "step": 921 }, { "epoch": 0.05, "learning_rate": 0.00019973946027248448, "loss": 0.6389, "step": 922 }, { "epoch": 0.05, "learning_rate": 0.00019973813326338698, "loss": 0.7339, "step": 923 }, { "epoch": 0.05, "learning_rate": 0.00019973680288786595, "loss": 0.7491, "step": 924 }, { "epoch": 0.05, "learning_rate": 0.00019973546914596623, "loss": 0.8163, "step": 925 }, { "epoch": 0.05, "learning_rate": 0.00019973413203773287, "loss": 0.755, "step": 926 }, { "epoch": 0.05, "learning_rate": 0.00019973279156321097, "loss": 0.7497, "step": 927 }, { "epoch": 0.05, "learning_rate": 0.00019973144772244582, "loss": 0.7587, "step": 928 }, { "epoch": 0.05, "learning_rate": 0.00019973010051548275, "loss": 0.8273, "step": 929 }, { "epoch": 0.05, "learning_rate": 0.00019972874994236722, "loss": 0.8351, "step": 930 }, { "epoch": 0.05, "learning_rate": 0.00019972739600314485, "loss": 0.7409, "step": 931 }, { "epoch": 0.05, "learning_rate": 0.0001997260386978613, "loss": 0.8279, "step": 932 }, { "epoch": 0.05, "learning_rate": 0.00019972467802656244, "loss": 0.7422, "step": 933 }, { "epoch": 0.05, "learning_rate": 0.0001997233139892941, "loss": 0.784, "step": 934 }, { "epoch": 0.05, "learning_rate": 0.00019972194658610245, "loss": 0.7041, "step": 935 }, { "epoch": 0.05, "learning_rate": 0.0001997205758170335, "loss": 0.6972, "step": 936 }, { "epoch": 0.05, "learning_rate": 0.00019971920168213364, "loss": 0.7508, "step": 937 }, { "epoch": 0.05, "learning_rate": 0.0001997178241814492, "loss": 0.7924, "step": 938 }, { "epoch": 0.05, "learning_rate": 0.00019971644331502665, "loss": 0.7531, "step": 939 }, { "epoch": 0.05, "learning_rate": 0.0001997150590829126, "loss": 0.749, "step": 940 }, { "epoch": 0.05, "learning_rate": 0.00019971367148515385, "loss": 0.778, "step": 941 }, { "epoch": 0.05, "learning_rate": 0.00019971228052179714, "loss": 0.7219, "step": 942 }, { "epoch": 0.05, "learning_rate": 0.0001997108861928895, "loss": 0.692, "step": 943 }, { "epoch": 0.05, "learning_rate": 0.0001997094884984779, "loss": 0.785, "step": 944 }, { "epoch": 0.05, "learning_rate": 0.0001997080874386096, "loss": 0.7516, "step": 945 }, { "epoch": 0.05, "learning_rate": 0.00019970668301333183, "loss": 0.72, "step": 946 }, { "epoch": 0.05, "learning_rate": 0.00019970527522269205, "loss": 0.7129, "step": 947 }, { "epoch": 0.05, "learning_rate": 0.00019970386406673772, "loss": 0.8121, "step": 948 }, { "epoch": 0.05, "learning_rate": 0.0001997024495455165, "loss": 0.7369, "step": 949 }, { "epoch": 0.05, "learning_rate": 0.00019970103165907612, "loss": 0.8516, "step": 950 }, { "epoch": 0.05, "learning_rate": 0.00019969961040746444, "loss": 0.7664, "step": 951 }, { "epoch": 0.05, "learning_rate": 0.00019969818579072945, "loss": 0.7157, "step": 952 }, { "epoch": 0.05, "learning_rate": 0.00019969675780891923, "loss": 0.6885, "step": 953 }, { "epoch": 0.05, "learning_rate": 0.00019969532646208195, "loss": 0.7725, "step": 954 }, { "epoch": 0.05, "learning_rate": 0.00019969389175026597, "loss": 0.8012, "step": 955 }, { "epoch": 0.05, "learning_rate": 0.00019969245367351966, "loss": 0.7812, "step": 956 }, { "epoch": 0.05, "learning_rate": 0.0001996910122318916, "loss": 0.7769, "step": 957 }, { "epoch": 0.05, "learning_rate": 0.00019968956742543042, "loss": 0.6854, "step": 958 }, { "epoch": 0.05, "learning_rate": 0.0001996881192541849, "loss": 0.5981, "step": 959 }, { "epoch": 0.05, "learning_rate": 0.00019968666771820391, "loss": 0.7177, "step": 960 }, { "epoch": 0.05, "learning_rate": 0.00019968521281753642, "loss": 0.79, "step": 961 }, { "epoch": 0.05, "learning_rate": 0.0001996837545522316, "loss": 0.7635, "step": 962 }, { "epoch": 0.05, "learning_rate": 0.0001996822929223386, "loss": 0.6759, "step": 963 }, { "epoch": 0.05, "learning_rate": 0.00019968082792790685, "loss": 0.7776, "step": 964 }, { "epoch": 0.05, "learning_rate": 0.0001996793595689857, "loss": 0.6666, "step": 965 }, { "epoch": 0.05, "learning_rate": 0.00019967788784562473, "loss": 0.7803, "step": 966 }, { "epoch": 0.05, "learning_rate": 0.00019967641275787362, "loss": 0.7288, "step": 967 }, { "epoch": 0.05, "learning_rate": 0.0001996749343057822, "loss": 0.6387, "step": 968 }, { "epoch": 0.05, "learning_rate": 0.00019967345248940034, "loss": 0.6912, "step": 969 }, { "epoch": 0.06, "learning_rate": 0.00019967196730877803, "loss": 0.7124, "step": 970 }, { "epoch": 0.06, "learning_rate": 0.00019967047876396545, "loss": 0.8044, "step": 971 }, { "epoch": 0.06, "learning_rate": 0.00019966898685501281, "loss": 0.7117, "step": 972 }, { "epoch": 0.06, "learning_rate": 0.00019966749158197047, "loss": 0.7879, "step": 973 }, { "epoch": 0.06, "learning_rate": 0.0001996659929448889, "loss": 0.7679, "step": 974 }, { "epoch": 0.06, "learning_rate": 0.0001996644909438187, "loss": 0.7816, "step": 975 }, { "epoch": 0.06, "learning_rate": 0.0001996629855788105, "loss": 0.7575, "step": 976 }, { "epoch": 0.06, "learning_rate": 0.0001996614768499152, "loss": 0.7015, "step": 977 }, { "epoch": 0.06, "learning_rate": 0.00019965996475718368, "loss": 0.7934, "step": 978 }, { "epoch": 0.06, "learning_rate": 0.000199658449300667, "loss": 0.7257, "step": 979 }, { "epoch": 0.06, "learning_rate": 0.00019965693048041628, "loss": 0.7516, "step": 980 }, { "epoch": 0.06, "learning_rate": 0.00019965540829648278, "loss": 0.8475, "step": 981 }, { "epoch": 0.06, "learning_rate": 0.00019965388274891793, "loss": 0.6342, "step": 982 }, { "epoch": 0.06, "learning_rate": 0.00019965235383777313, "loss": 0.6828, "step": 983 }, { "epoch": 0.06, "learning_rate": 0.00019965082156310008, "loss": 0.6942, "step": 984 }, { "epoch": 0.06, "learning_rate": 0.00019964928592495045, "loss": 0.7328, "step": 985 }, { "epoch": 0.06, "learning_rate": 0.00019964774692337608, "loss": 0.8081, "step": 986 }, { "epoch": 0.06, "learning_rate": 0.00019964620455842892, "loss": 0.7633, "step": 987 }, { "epoch": 0.06, "learning_rate": 0.000199644658830161, "loss": 0.688, "step": 988 }, { "epoch": 0.06, "learning_rate": 0.00019964310973862455, "loss": 0.7476, "step": 989 }, { "epoch": 0.06, "learning_rate": 0.0001996415572838718, "loss": 0.6451, "step": 990 }, { "epoch": 0.06, "learning_rate": 0.00019964000146595518, "loss": 0.6669, "step": 991 }, { "epoch": 0.06, "learning_rate": 0.00019963844228492722, "loss": 0.6851, "step": 992 }, { "epoch": 0.06, "learning_rate": 0.00019963687974084048, "loss": 0.7943, "step": 993 }, { "epoch": 0.06, "learning_rate": 0.00019963531383374776, "loss": 0.6309, "step": 994 }, { "epoch": 0.06, "learning_rate": 0.0001996337445637019, "loss": 0.792, "step": 995 }, { "epoch": 0.06, "learning_rate": 0.00019963217193075586, "loss": 0.7871, "step": 996 }, { "epoch": 0.06, "learning_rate": 0.00019963059593496268, "loss": 0.6972, "step": 997 }, { "epoch": 0.06, "learning_rate": 0.00019962901657637566, "loss": 0.7982, "step": 998 }, { "epoch": 0.06, "learning_rate": 0.00019962743385504802, "loss": 0.7516, "step": 999 }, { "epoch": 0.06, "learning_rate": 0.00019962584777103318, "loss": 0.7982, "step": 1000 }, { "epoch": 0.06, "learning_rate": 0.00019962425832438472, "loss": 0.7375, "step": 1001 }, { "epoch": 0.06, "learning_rate": 0.0001996226655151563, "loss": 0.7594, "step": 1002 }, { "epoch": 0.06, "learning_rate": 0.0001996210693434016, "loss": 0.6724, "step": 1003 }, { "epoch": 0.06, "learning_rate": 0.00019961946980917456, "loss": 0.7702, "step": 1004 }, { "epoch": 0.06, "learning_rate": 0.00019961786691252917, "loss": 0.9138, "step": 1005 }, { "epoch": 0.06, "learning_rate": 0.00019961626065351947, "loss": 0.8298, "step": 1006 }, { "epoch": 0.06, "learning_rate": 0.00019961465103219975, "loss": 0.746, "step": 1007 }, { "epoch": 0.06, "learning_rate": 0.00019961303804862432, "loss": 0.779, "step": 1008 }, { "epoch": 0.06, "learning_rate": 0.00019961142170284762, "loss": 0.7243, "step": 1009 }, { "epoch": 0.06, "learning_rate": 0.00019960980199492414, "loss": 0.7271, "step": 1010 }, { "epoch": 0.06, "learning_rate": 0.00019960817892490865, "loss": 0.7646, "step": 1011 }, { "epoch": 0.06, "learning_rate": 0.0001996065524928559, "loss": 0.7972, "step": 1012 }, { "epoch": 0.06, "learning_rate": 0.00019960492269882073, "loss": 0.7527, "step": 1013 }, { "epoch": 0.06, "learning_rate": 0.00019960328954285824, "loss": 0.7941, "step": 1014 }, { "epoch": 0.06, "learning_rate": 0.0001996016530250235, "loss": 0.8104, "step": 1015 }, { "epoch": 0.06, "learning_rate": 0.00019960001314537173, "loss": 0.6829, "step": 1016 }, { "epoch": 0.06, "learning_rate": 0.00019959836990395832, "loss": 0.6831, "step": 1017 }, { "epoch": 0.06, "learning_rate": 0.00019959672330083873, "loss": 0.7192, "step": 1018 }, { "epoch": 0.06, "learning_rate": 0.00019959507333606853, "loss": 0.8008, "step": 1019 }, { "epoch": 0.06, "learning_rate": 0.0001995934200097034, "loss": 0.7055, "step": 1020 }, { "epoch": 0.06, "learning_rate": 0.00019959176332179915, "loss": 0.7998, "step": 1021 }, { "epoch": 0.06, "learning_rate": 0.00019959010327241168, "loss": 0.7504, "step": 1022 }, { "epoch": 0.06, "learning_rate": 0.00019958843986159704, "loss": 0.65, "step": 1023 }, { "epoch": 0.06, "learning_rate": 0.00019958677308941139, "loss": 0.7417, "step": 1024 }, { "epoch": 0.06, "learning_rate": 0.00019958510295591098, "loss": 0.6536, "step": 1025 }, { "epoch": 0.06, "learning_rate": 0.00019958342946115216, "loss": 0.7523, "step": 1026 }, { "epoch": 0.06, "learning_rate": 0.00019958175260519144, "loss": 0.7764, "step": 1027 }, { "epoch": 0.06, "learning_rate": 0.0001995800723880854, "loss": 0.6648, "step": 1028 }, { "epoch": 0.06, "learning_rate": 0.00019957838880989078, "loss": 0.7403, "step": 1029 }, { "epoch": 0.06, "learning_rate": 0.00019957670187066432, "loss": 0.6898, "step": 1030 }, { "epoch": 0.06, "learning_rate": 0.0001995750115704631, "loss": 0.8559, "step": 1031 }, { "epoch": 0.06, "learning_rate": 0.00019957331790934407, "loss": 0.7822, "step": 1032 }, { "epoch": 0.06, "learning_rate": 0.0001995716208873644, "loss": 0.7558, "step": 1033 }, { "epoch": 0.06, "learning_rate": 0.00019956992050458139, "loss": 0.7638, "step": 1034 }, { "epoch": 0.06, "learning_rate": 0.00019956821676105245, "loss": 0.8225, "step": 1035 }, { "epoch": 0.06, "learning_rate": 0.00019956650965683505, "loss": 0.593, "step": 1036 }, { "epoch": 0.06, "learning_rate": 0.00019956479919198682, "loss": 0.795, "step": 1037 }, { "epoch": 0.06, "learning_rate": 0.00019956308536656553, "loss": 0.8797, "step": 1038 }, { "epoch": 0.06, "learning_rate": 0.00019956136818062898, "loss": 0.7308, "step": 1039 }, { "epoch": 0.06, "learning_rate": 0.00019955964763423514, "loss": 0.8061, "step": 1040 }, { "epoch": 0.06, "learning_rate": 0.00019955792372744209, "loss": 0.7292, "step": 1041 }, { "epoch": 0.06, "learning_rate": 0.00019955619646030802, "loss": 0.7852, "step": 1042 }, { "epoch": 0.06, "learning_rate": 0.0001995544658328912, "loss": 0.8367, "step": 1043 }, { "epoch": 0.06, "learning_rate": 0.0001995527318452501, "loss": 0.6144, "step": 1044 }, { "epoch": 0.06, "learning_rate": 0.0001995509944974432, "loss": 0.795, "step": 1045 }, { "epoch": 0.06, "learning_rate": 0.00019954925378952914, "loss": 0.6599, "step": 1046 }, { "epoch": 0.06, "learning_rate": 0.00019954750972156672, "loss": 0.7679, "step": 1047 }, { "epoch": 0.06, "learning_rate": 0.00019954576229361473, "loss": 0.6981, "step": 1048 }, { "epoch": 0.06, "learning_rate": 0.00019954401150573222, "loss": 0.8283, "step": 1049 }, { "epoch": 0.06, "learning_rate": 0.00019954225735797825, "loss": 0.7697, "step": 1050 }, { "epoch": 0.06, "learning_rate": 0.00019954049985041204, "loss": 0.739, "step": 1051 }, { "epoch": 0.06, "learning_rate": 0.00019953873898309293, "loss": 0.7649, "step": 1052 }, { "epoch": 0.06, "learning_rate": 0.0001995369747560803, "loss": 0.754, "step": 1053 }, { "epoch": 0.06, "learning_rate": 0.00019953520716943371, "loss": 0.7377, "step": 1054 }, { "epoch": 0.06, "learning_rate": 0.00019953343622321288, "loss": 0.7937, "step": 1055 }, { "epoch": 0.06, "learning_rate": 0.00019953166191747753, "loss": 0.7777, "step": 1056 }, { "epoch": 0.06, "learning_rate": 0.00019952988425228754, "loss": 0.8087, "step": 1057 }, { "epoch": 0.06, "learning_rate": 0.00019952810322770295, "loss": 0.6077, "step": 1058 }, { "epoch": 0.06, "learning_rate": 0.00019952631884378384, "loss": 0.7387, "step": 1059 }, { "epoch": 0.06, "learning_rate": 0.00019952453110059045, "loss": 0.7384, "step": 1060 }, { "epoch": 0.06, "learning_rate": 0.0001995227399981831, "loss": 0.7835, "step": 1061 }, { "epoch": 0.06, "learning_rate": 0.00019952094553662232, "loss": 0.7519, "step": 1062 }, { "epoch": 0.06, "learning_rate": 0.0001995191477159686, "loss": 0.7106, "step": 1063 }, { "epoch": 0.06, "learning_rate": 0.00019951734653628265, "loss": 0.6758, "step": 1064 }, { "epoch": 0.06, "learning_rate": 0.00019951554199762526, "loss": 0.6786, "step": 1065 }, { "epoch": 0.06, "learning_rate": 0.00019951373410005734, "loss": 0.7134, "step": 1066 }, { "epoch": 0.06, "learning_rate": 0.00019951192284363992, "loss": 0.6925, "step": 1067 }, { "epoch": 0.06, "learning_rate": 0.00019951010822843407, "loss": 0.7501, "step": 1068 }, { "epoch": 0.06, "learning_rate": 0.00019950829025450114, "loss": 0.7135, "step": 1069 }, { "epoch": 0.06, "learning_rate": 0.00019950646892190243, "loss": 0.7764, "step": 1070 }, { "epoch": 0.06, "learning_rate": 0.00019950464423069945, "loss": 0.6561, "step": 1071 }, { "epoch": 0.06, "learning_rate": 0.00019950281618095374, "loss": 0.742, "step": 1072 }, { "epoch": 0.06, "learning_rate": 0.00019950098477272704, "loss": 0.679, "step": 1073 }, { "epoch": 0.06, "learning_rate": 0.00019949915000608116, "loss": 0.7436, "step": 1074 }, { "epoch": 0.06, "learning_rate": 0.00019949731188107804, "loss": 0.6496, "step": 1075 }, { "epoch": 0.06, "learning_rate": 0.00019949547039777965, "loss": 0.7623, "step": 1076 }, { "epoch": 0.06, "learning_rate": 0.00019949362555624824, "loss": 0.7526, "step": 1077 }, { "epoch": 0.06, "learning_rate": 0.00019949177735654601, "loss": 0.7861, "step": 1078 }, { "epoch": 0.06, "learning_rate": 0.00019948992579873538, "loss": 0.9274, "step": 1079 }, { "epoch": 0.06, "learning_rate": 0.00019948807088287883, "loss": 0.6947, "step": 1080 }, { "epoch": 0.06, "learning_rate": 0.00019948621260903896, "loss": 0.7745, "step": 1081 }, { "epoch": 0.06, "learning_rate": 0.00019948435097727853, "loss": 0.6991, "step": 1082 }, { "epoch": 0.06, "learning_rate": 0.00019948248598766034, "loss": 0.6711, "step": 1083 }, { "epoch": 0.06, "learning_rate": 0.00019948061764024733, "loss": 0.7739, "step": 1084 }, { "epoch": 0.06, "learning_rate": 0.0001994787459351026, "loss": 0.7595, "step": 1085 }, { "epoch": 0.06, "learning_rate": 0.00019947687087228927, "loss": 0.8289, "step": 1086 }, { "epoch": 0.06, "learning_rate": 0.00019947499245187068, "loss": 0.8313, "step": 1087 }, { "epoch": 0.06, "learning_rate": 0.0001994731106739102, "loss": 0.7476, "step": 1088 }, { "epoch": 0.06, "learning_rate": 0.00019947122553847136, "loss": 0.7595, "step": 1089 }, { "epoch": 0.06, "learning_rate": 0.00019946933704561779, "loss": 0.7661, "step": 1090 }, { "epoch": 0.06, "learning_rate": 0.00019946744519541322, "loss": 0.7299, "step": 1091 }, { "epoch": 0.06, "learning_rate": 0.00019946554998792154, "loss": 0.7973, "step": 1092 }, { "epoch": 0.06, "learning_rate": 0.00019946365142320665, "loss": 0.8041, "step": 1093 }, { "epoch": 0.06, "learning_rate": 0.00019946174950133267, "loss": 0.686, "step": 1094 }, { "epoch": 0.06, "learning_rate": 0.00019945984422236384, "loss": 0.7766, "step": 1095 }, { "epoch": 0.06, "learning_rate": 0.00019945793558636437, "loss": 0.8514, "step": 1096 }, { "epoch": 0.06, "learning_rate": 0.00019945602359339878, "loss": 0.6501, "step": 1097 }, { "epoch": 0.06, "learning_rate": 0.00019945410824353155, "loss": 0.7313, "step": 1098 }, { "epoch": 0.06, "learning_rate": 0.00019945218953682734, "loss": 0.7131, "step": 1099 }, { "epoch": 0.06, "learning_rate": 0.0001994502674733509, "loss": 0.8151, "step": 1100 }, { "epoch": 0.06, "learning_rate": 0.00019944834205316711, "loss": 0.6821, "step": 1101 }, { "epoch": 0.06, "learning_rate": 0.000199446413276341, "loss": 0.8436, "step": 1102 }, { "epoch": 0.06, "learning_rate": 0.0001994444811429376, "loss": 0.749, "step": 1103 }, { "epoch": 0.06, "learning_rate": 0.00019944254565302217, "loss": 0.7998, "step": 1104 }, { "epoch": 0.06, "learning_rate": 0.00019944060680666002, "loss": 0.7114, "step": 1105 }, { "epoch": 0.06, "learning_rate": 0.0001994386646039166, "loss": 0.8128, "step": 1106 }, { "epoch": 0.06, "learning_rate": 0.00019943671904485748, "loss": 0.7026, "step": 1107 }, { "epoch": 0.06, "learning_rate": 0.00019943477012954828, "loss": 0.6796, "step": 1108 }, { "epoch": 0.06, "learning_rate": 0.0001994328178580548, "loss": 0.652, "step": 1109 }, { "epoch": 0.06, "learning_rate": 0.000199430862230443, "loss": 0.8686, "step": 1110 }, { "epoch": 0.06, "learning_rate": 0.00019942890324677878, "loss": 0.7524, "step": 1111 }, { "epoch": 0.06, "learning_rate": 0.00019942694090712836, "loss": 0.7862, "step": 1112 }, { "epoch": 0.06, "learning_rate": 0.0001994249752115579, "loss": 0.7986, "step": 1113 }, { "epoch": 0.06, "learning_rate": 0.0001994230061601338, "loss": 0.7396, "step": 1114 }, { "epoch": 0.06, "learning_rate": 0.00019942103375292244, "loss": 0.7388, "step": 1115 }, { "epoch": 0.06, "learning_rate": 0.00019941905798999046, "loss": 0.7531, "step": 1116 }, { "epoch": 0.06, "learning_rate": 0.0001994170788714046, "loss": 0.7328, "step": 1117 }, { "epoch": 0.06, "learning_rate": 0.00019941509639723155, "loss": 0.8177, "step": 1118 }, { "epoch": 0.06, "learning_rate": 0.00019941311056753826, "loss": 0.7457, "step": 1119 }, { "epoch": 0.06, "learning_rate": 0.0001994111213823918, "loss": 0.783, "step": 1120 }, { "epoch": 0.06, "learning_rate": 0.00019940912884185927, "loss": 0.7378, "step": 1121 }, { "epoch": 0.06, "learning_rate": 0.00019940713294600793, "loss": 0.7219, "step": 1122 }, { "epoch": 0.06, "learning_rate": 0.00019940513369490516, "loss": 0.742, "step": 1123 }, { "epoch": 0.06, "learning_rate": 0.00019940313108861838, "loss": 0.8731, "step": 1124 }, { "epoch": 0.06, "learning_rate": 0.0001994011251272153, "loss": 0.826, "step": 1125 }, { "epoch": 0.06, "learning_rate": 0.0001993991158107635, "loss": 0.7634, "step": 1126 }, { "epoch": 0.06, "learning_rate": 0.00019939710313933087, "loss": 0.6861, "step": 1127 }, { "epoch": 0.06, "learning_rate": 0.00019939508711298536, "loss": 0.6697, "step": 1128 }, { "epoch": 0.06, "learning_rate": 0.00019939306773179497, "loss": 0.7186, "step": 1129 }, { "epoch": 0.06, "learning_rate": 0.00019939104499582785, "loss": 0.6285, "step": 1130 }, { "epoch": 0.06, "learning_rate": 0.00019938901890515233, "loss": 0.7111, "step": 1131 }, { "epoch": 0.06, "learning_rate": 0.00019938698945983676, "loss": 0.7208, "step": 1132 }, { "epoch": 0.06, "learning_rate": 0.00019938495665994962, "loss": 0.856, "step": 1133 }, { "epoch": 0.06, "learning_rate": 0.0001993829205055596, "loss": 0.7331, "step": 1134 }, { "epoch": 0.06, "learning_rate": 0.00019938088099673534, "loss": 0.761, "step": 1135 }, { "epoch": 0.06, "learning_rate": 0.0001993788381335457, "loss": 0.8211, "step": 1136 }, { "epoch": 0.06, "learning_rate": 0.00019937679191605963, "loss": 0.738, "step": 1137 }, { "epoch": 0.06, "learning_rate": 0.00019937474234434625, "loss": 0.6923, "step": 1138 }, { "epoch": 0.06, "learning_rate": 0.00019937268941847466, "loss": 0.716, "step": 1139 }, { "epoch": 0.06, "learning_rate": 0.0001993706331385142, "loss": 0.782, "step": 1140 }, { "epoch": 0.06, "learning_rate": 0.0001993685735045343, "loss": 0.8011, "step": 1141 }, { "epoch": 0.06, "learning_rate": 0.0001993665105166044, "loss": 0.8387, "step": 1142 }, { "epoch": 0.06, "learning_rate": 0.00019936444417479416, "loss": 0.8215, "step": 1143 }, { "epoch": 0.06, "learning_rate": 0.00019936237447917336, "loss": 0.7941, "step": 1144 }, { "epoch": 0.06, "learning_rate": 0.00019936030142981182, "loss": 0.6616, "step": 1145 }, { "epoch": 0.07, "learning_rate": 0.00019935822502677956, "loss": 0.7456, "step": 1146 }, { "epoch": 0.07, "learning_rate": 0.00019935614527014659, "loss": 0.825, "step": 1147 }, { "epoch": 0.07, "learning_rate": 0.00019935406215998316, "loss": 0.7809, "step": 1148 }, { "epoch": 0.07, "learning_rate": 0.00019935197569635954, "loss": 0.7935, "step": 1149 }, { "epoch": 0.07, "learning_rate": 0.00019934988587934623, "loss": 0.7634, "step": 1150 }, { "epoch": 0.07, "learning_rate": 0.00019934779270901367, "loss": 0.6511, "step": 1151 }, { "epoch": 0.07, "learning_rate": 0.00019934569618543257, "loss": 0.7497, "step": 1152 }, { "epoch": 0.07, "learning_rate": 0.0001993435963086737, "loss": 0.8237, "step": 1153 }, { "epoch": 0.07, "learning_rate": 0.00019934149307880791, "loss": 0.7854, "step": 1154 }, { "epoch": 0.07, "learning_rate": 0.0001993393864959062, "loss": 0.7551, "step": 1155 }, { "epoch": 0.07, "learning_rate": 0.00019933727656003963, "loss": 0.7226, "step": 1156 }, { "epoch": 0.07, "learning_rate": 0.0001993351632712795, "loss": 0.744, "step": 1157 }, { "epoch": 0.07, "learning_rate": 0.00019933304662969706, "loss": 0.8729, "step": 1158 }, { "epoch": 0.07, "learning_rate": 0.00019933092663536382, "loss": 0.8826, "step": 1159 }, { "epoch": 0.07, "learning_rate": 0.0001993288032883513, "loss": 0.7508, "step": 1160 }, { "epoch": 0.07, "learning_rate": 0.00019932667658873114, "loss": 0.7399, "step": 1161 }, { "epoch": 0.07, "learning_rate": 0.00019932454653657518, "loss": 0.725, "step": 1162 }, { "epoch": 0.07, "learning_rate": 0.00019932241313195528, "loss": 0.7326, "step": 1163 }, { "epoch": 0.07, "learning_rate": 0.00019932027637494342, "loss": 0.8143, "step": 1164 }, { "epoch": 0.07, "learning_rate": 0.0001993181362656118, "loss": 0.7038, "step": 1165 }, { "epoch": 0.07, "learning_rate": 0.0001993159928040326, "loss": 0.6404, "step": 1166 }, { "epoch": 0.07, "learning_rate": 0.00019931384599027814, "loss": 0.6874, "step": 1167 }, { "epoch": 0.07, "learning_rate": 0.00019931169582442095, "loss": 0.7124, "step": 1168 }, { "epoch": 0.07, "learning_rate": 0.00019930954230653355, "loss": 0.744, "step": 1169 }, { "epoch": 0.07, "learning_rate": 0.0001993073854366887, "loss": 0.6503, "step": 1170 }, { "epoch": 0.07, "learning_rate": 0.0001993052252149591, "loss": 0.7735, "step": 1171 }, { "epoch": 0.07, "learning_rate": 0.00019930306164141773, "loss": 0.7307, "step": 1172 }, { "epoch": 0.07, "learning_rate": 0.0001993008947161376, "loss": 0.7459, "step": 1173 }, { "epoch": 0.07, "learning_rate": 0.0001992987244391918, "loss": 0.664, "step": 1174 }, { "epoch": 0.07, "learning_rate": 0.0001992965508106537, "loss": 0.699, "step": 1175 }, { "epoch": 0.07, "learning_rate": 0.00019929437383059656, "loss": 0.7165, "step": 1176 }, { "epoch": 0.07, "learning_rate": 0.00019929219349909392, "loss": 0.7303, "step": 1177 }, { "epoch": 0.07, "learning_rate": 0.00019929000981621931, "loss": 0.827, "step": 1178 }, { "epoch": 0.07, "learning_rate": 0.0001992878227820465, "loss": 0.774, "step": 1179 }, { "epoch": 0.07, "learning_rate": 0.00019928563239664926, "loss": 0.665, "step": 1180 }, { "epoch": 0.07, "learning_rate": 0.00019928343866010154, "loss": 0.7547, "step": 1181 }, { "epoch": 0.07, "learning_rate": 0.0001992812415724774, "loss": 0.7191, "step": 1182 }, { "epoch": 0.07, "learning_rate": 0.00019927904113385098, "loss": 0.7331, "step": 1183 }, { "epoch": 0.07, "learning_rate": 0.00019927683734429654, "loss": 0.7241, "step": 1184 }, { "epoch": 0.07, "learning_rate": 0.0001992746302038885, "loss": 0.7665, "step": 1185 }, { "epoch": 0.07, "learning_rate": 0.0001992724197127013, "loss": 0.6606, "step": 1186 }, { "epoch": 0.07, "learning_rate": 0.0001992702058708096, "loss": 0.7851, "step": 1187 }, { "epoch": 0.07, "learning_rate": 0.0001992679886782881, "loss": 0.8091, "step": 1188 }, { "epoch": 0.07, "learning_rate": 0.00019926576813521164, "loss": 0.8406, "step": 1189 }, { "epoch": 0.07, "learning_rate": 0.0001992635442416552, "loss": 0.7369, "step": 1190 }, { "epoch": 0.07, "learning_rate": 0.0001992613169976938, "loss": 0.7025, "step": 1191 }, { "epoch": 0.07, "learning_rate": 0.0001992590864034026, "loss": 0.7158, "step": 1192 }, { "epoch": 0.07, "learning_rate": 0.00019925685245885695, "loss": 0.733, "step": 1193 }, { "epoch": 0.07, "learning_rate": 0.00019925461516413223, "loss": 0.7153, "step": 1194 }, { "epoch": 0.07, "learning_rate": 0.0001992523745193039, "loss": 0.7229, "step": 1195 }, { "epoch": 0.07, "learning_rate": 0.00019925013052444766, "loss": 0.7473, "step": 1196 }, { "epoch": 0.07, "learning_rate": 0.0001992478831796392, "loss": 0.7823, "step": 1197 }, { "epoch": 0.07, "learning_rate": 0.00019924563248495443, "loss": 0.7396, "step": 1198 }, { "epoch": 0.07, "learning_rate": 0.00019924337844046925, "loss": 0.7287, "step": 1199 }, { "epoch": 0.07, "learning_rate": 0.00019924112104625978, "loss": 0.8, "step": 1200 }, { "epoch": 0.07, "learning_rate": 0.00019923886030240225, "loss": 0.6337, "step": 1201 }, { "epoch": 0.07, "learning_rate": 0.00019923659620897286, "loss": 0.7454, "step": 1202 }, { "epoch": 0.07, "learning_rate": 0.00019923432876604811, "loss": 0.8511, "step": 1203 }, { "epoch": 0.07, "learning_rate": 0.0001992320579737045, "loss": 0.5782, "step": 1204 }, { "epoch": 0.07, "learning_rate": 0.00019922978383201873, "loss": 0.8037, "step": 1205 }, { "epoch": 0.07, "learning_rate": 0.00019922750634106747, "loss": 0.7027, "step": 1206 }, { "epoch": 0.07, "learning_rate": 0.00019922522550092767, "loss": 0.8066, "step": 1207 }, { "epoch": 0.07, "learning_rate": 0.00019922294131167627, "loss": 0.7728, "step": 1208 }, { "epoch": 0.07, "learning_rate": 0.00019922065377339036, "loss": 0.8205, "step": 1209 }, { "epoch": 0.07, "learning_rate": 0.0001992183628861472, "loss": 0.8074, "step": 1210 }, { "epoch": 0.07, "learning_rate": 0.00019921606865002406, "loss": 0.7516, "step": 1211 }, { "epoch": 0.07, "learning_rate": 0.0001992137710650984, "loss": 0.6698, "step": 1212 }, { "epoch": 0.07, "learning_rate": 0.0001992114701314478, "loss": 0.8298, "step": 1213 }, { "epoch": 0.07, "learning_rate": 0.00019920916584914987, "loss": 0.8247, "step": 1214 }, { "epoch": 0.07, "learning_rate": 0.00019920685821828239, "loss": 0.7043, "step": 1215 }, { "epoch": 0.07, "learning_rate": 0.00019920454723892328, "loss": 0.7063, "step": 1216 }, { "epoch": 0.07, "learning_rate": 0.00019920223291115054, "loss": 0.7042, "step": 1217 }, { "epoch": 0.07, "learning_rate": 0.00019919991523504224, "loss": 0.7429, "step": 1218 }, { "epoch": 0.07, "learning_rate": 0.0001991975942106767, "loss": 0.7453, "step": 1219 }, { "epoch": 0.07, "learning_rate": 0.00019919526983813214, "loss": 0.9567, "step": 1220 }, { "epoch": 0.07, "learning_rate": 0.00019919294211748708, "loss": 0.8021, "step": 1221 }, { "epoch": 0.07, "learning_rate": 0.0001991906110488201, "loss": 0.7331, "step": 1222 }, { "epoch": 0.07, "learning_rate": 0.00019918827663220986, "loss": 0.779, "step": 1223 }, { "epoch": 0.07, "learning_rate": 0.00019918593886773515, "loss": 0.7315, "step": 1224 }, { "epoch": 0.07, "learning_rate": 0.00019918359775547489, "loss": 0.7131, "step": 1225 }, { "epoch": 0.07, "learning_rate": 0.00019918125329550807, "loss": 0.6654, "step": 1226 }, { "epoch": 0.07, "learning_rate": 0.00019917890548791384, "loss": 0.7779, "step": 1227 }, { "epoch": 0.07, "learning_rate": 0.00019917655433277144, "loss": 0.7979, "step": 1228 }, { "epoch": 0.07, "learning_rate": 0.00019917419983016025, "loss": 0.8135, "step": 1229 }, { "epoch": 0.07, "learning_rate": 0.00019917184198015971, "loss": 0.6917, "step": 1230 }, { "epoch": 0.07, "learning_rate": 0.0001991694807828494, "loss": 0.7915, "step": 1231 }, { "epoch": 0.07, "learning_rate": 0.00019916711623830903, "loss": 0.636, "step": 1232 }, { "epoch": 0.07, "learning_rate": 0.00019916474834661844, "loss": 0.68, "step": 1233 }, { "epoch": 0.07, "learning_rate": 0.00019916237710785752, "loss": 0.754, "step": 1234 }, { "epoch": 0.07, "learning_rate": 0.00019916000252210631, "loss": 0.695, "step": 1235 }, { "epoch": 0.07, "learning_rate": 0.00019915762458944496, "loss": 0.7361, "step": 1236 }, { "epoch": 0.07, "learning_rate": 0.00019915524330995371, "loss": 0.756, "step": 1237 }, { "epoch": 0.07, "learning_rate": 0.00019915285868371297, "loss": 0.8324, "step": 1238 }, { "epoch": 0.07, "learning_rate": 0.0001991504707108032, "loss": 0.6363, "step": 1239 }, { "epoch": 0.07, "learning_rate": 0.000199148079391305, "loss": 0.6866, "step": 1240 }, { "epoch": 0.07, "learning_rate": 0.00019914568472529913, "loss": 0.7619, "step": 1241 }, { "epoch": 0.07, "learning_rate": 0.00019914328671286638, "loss": 0.7786, "step": 1242 }, { "epoch": 0.07, "learning_rate": 0.00019914088535408767, "loss": 0.6947, "step": 1243 }, { "epoch": 0.07, "learning_rate": 0.00019913848064904407, "loss": 0.7375, "step": 1244 }, { "epoch": 0.07, "learning_rate": 0.00019913607259781677, "loss": 0.7578, "step": 1245 }, { "epoch": 0.07, "learning_rate": 0.000199133661200487, "loss": 0.7055, "step": 1246 }, { "epoch": 0.07, "learning_rate": 0.00019913124645713618, "loss": 0.6671, "step": 1247 }, { "epoch": 0.07, "learning_rate": 0.0001991288283678458, "loss": 0.7345, "step": 1248 }, { "epoch": 0.07, "learning_rate": 0.00019912640693269752, "loss": 0.8143, "step": 1249 }, { "epoch": 0.07, "learning_rate": 0.00019912398215177305, "loss": 0.7861, "step": 1250 }, { "epoch": 0.07, "learning_rate": 0.00019912155402515417, "loss": 0.8244, "step": 1251 }, { "epoch": 0.07, "learning_rate": 0.00019911912255292293, "loss": 0.699, "step": 1252 }, { "epoch": 0.07, "learning_rate": 0.00019911668773516135, "loss": 0.7943, "step": 1253 }, { "epoch": 0.07, "learning_rate": 0.00019911424957195158, "loss": 0.7656, "step": 1254 }, { "epoch": 0.07, "learning_rate": 0.00019911180806337598, "loss": 0.7848, "step": 1255 }, { "epoch": 0.07, "learning_rate": 0.0001991093632095169, "loss": 0.644, "step": 1256 }, { "epoch": 0.07, "learning_rate": 0.00019910691501045692, "loss": 0.7554, "step": 1257 }, { "epoch": 0.07, "learning_rate": 0.00019910446346627862, "loss": 0.7639, "step": 1258 }, { "epoch": 0.07, "learning_rate": 0.00019910200857706478, "loss": 0.7376, "step": 1259 }, { "epoch": 0.07, "learning_rate": 0.00019909955034289823, "loss": 0.685, "step": 1260 }, { "epoch": 0.07, "learning_rate": 0.000199097088763862, "loss": 0.7283, "step": 1261 }, { "epoch": 0.07, "learning_rate": 0.00019909462384003906, "loss": 0.6963, "step": 1262 }, { "epoch": 0.07, "learning_rate": 0.00019909215557151275, "loss": 0.7656, "step": 1263 }, { "epoch": 0.07, "learning_rate": 0.0001990896839583663, "loss": 0.6279, "step": 1264 }, { "epoch": 0.07, "learning_rate": 0.0001990872090006831, "loss": 0.721, "step": 1265 }, { "epoch": 0.07, "learning_rate": 0.00019908473069854676, "loss": 0.7381, "step": 1266 }, { "epoch": 0.07, "learning_rate": 0.0001990822490520409, "loss": 0.8094, "step": 1267 }, { "epoch": 0.07, "learning_rate": 0.00019907976406124926, "loss": 0.7743, "step": 1268 }, { "epoch": 0.07, "learning_rate": 0.0001990772757262558, "loss": 0.794, "step": 1269 }, { "epoch": 0.07, "learning_rate": 0.00019907478404714436, "loss": 0.8177, "step": 1270 }, { "epoch": 0.07, "learning_rate": 0.00019907228902399919, "loss": 0.69, "step": 1271 }, { "epoch": 0.07, "learning_rate": 0.00019906979065690438, "loss": 0.8181, "step": 1272 }, { "epoch": 0.07, "learning_rate": 0.00019906728894594438, "loss": 0.7173, "step": 1273 }, { "epoch": 0.07, "learning_rate": 0.0001990647838912035, "loss": 0.7818, "step": 1274 }, { "epoch": 0.07, "learning_rate": 0.00019906227549276642, "loss": 0.6544, "step": 1275 }, { "epoch": 0.07, "learning_rate": 0.00019905976375071772, "loss": 0.7625, "step": 1276 }, { "epoch": 0.07, "learning_rate": 0.00019905724866514218, "loss": 0.6201, "step": 1277 }, { "epoch": 0.07, "learning_rate": 0.0001990547302361247, "loss": 0.8192, "step": 1278 }, { "epoch": 0.07, "learning_rate": 0.00019905220846375032, "loss": 0.7195, "step": 1279 }, { "epoch": 0.07, "learning_rate": 0.00019904968334810415, "loss": 0.5994, "step": 1280 }, { "epoch": 0.07, "learning_rate": 0.00019904715488927138, "loss": 0.7469, "step": 1281 }, { "epoch": 0.07, "learning_rate": 0.00019904462308733737, "loss": 0.6309, "step": 1282 }, { "epoch": 0.07, "learning_rate": 0.00019904208794238758, "loss": 0.7911, "step": 1283 }, { "epoch": 0.07, "learning_rate": 0.00019903954945450756, "loss": 0.7161, "step": 1284 }, { "epoch": 0.07, "learning_rate": 0.000199037007623783, "loss": 0.7217, "step": 1285 }, { "epoch": 0.07, "learning_rate": 0.00019903446245029974, "loss": 0.7892, "step": 1286 }, { "epoch": 0.07, "learning_rate": 0.0001990319139341436, "loss": 0.5909, "step": 1287 }, { "epoch": 0.07, "learning_rate": 0.00019902936207540065, "loss": 0.774, "step": 1288 }, { "epoch": 0.07, "learning_rate": 0.00019902680687415705, "loss": 0.7699, "step": 1289 }, { "epoch": 0.07, "learning_rate": 0.00019902424833049897, "loss": 0.731, "step": 1290 }, { "epoch": 0.07, "learning_rate": 0.00019902168644451282, "loss": 0.8804, "step": 1291 }, { "epoch": 0.07, "learning_rate": 0.00019901912121628507, "loss": 0.7759, "step": 1292 }, { "epoch": 0.07, "learning_rate": 0.0001990165526459023, "loss": 0.8047, "step": 1293 }, { "epoch": 0.07, "learning_rate": 0.00019901398073345118, "loss": 0.6443, "step": 1294 }, { "epoch": 0.07, "learning_rate": 0.00019901140547901855, "loss": 0.7708, "step": 1295 }, { "epoch": 0.07, "learning_rate": 0.0001990088268826913, "loss": 0.7526, "step": 1296 }, { "epoch": 0.07, "learning_rate": 0.0001990062449445565, "loss": 0.7133, "step": 1297 }, { "epoch": 0.07, "learning_rate": 0.00019900365966470126, "loss": 0.6747, "step": 1298 }, { "epoch": 0.07, "learning_rate": 0.00019900107104321287, "loss": 0.8247, "step": 1299 }, { "epoch": 0.07, "learning_rate": 0.0001989984790801787, "loss": 0.7662, "step": 1300 }, { "epoch": 0.07, "learning_rate": 0.00019899588377568622, "loss": 0.7146, "step": 1301 }, { "epoch": 0.07, "learning_rate": 0.00019899328512982306, "loss": 0.7187, "step": 1302 }, { "epoch": 0.07, "learning_rate": 0.00019899068314267688, "loss": 0.6538, "step": 1303 }, { "epoch": 0.07, "learning_rate": 0.00019898807781433555, "loss": 0.7965, "step": 1304 }, { "epoch": 0.07, "learning_rate": 0.00019898546914488697, "loss": 0.6832, "step": 1305 }, { "epoch": 0.07, "learning_rate": 0.00019898285713441921, "loss": 0.7379, "step": 1306 }, { "epoch": 0.07, "learning_rate": 0.00019898024178302044, "loss": 0.757, "step": 1307 }, { "epoch": 0.07, "learning_rate": 0.0001989776230907789, "loss": 0.7038, "step": 1308 }, { "epoch": 0.07, "learning_rate": 0.000198975001057783, "loss": 0.7146, "step": 1309 }, { "epoch": 0.07, "learning_rate": 0.0001989723756841213, "loss": 0.8081, "step": 1310 }, { "epoch": 0.07, "learning_rate": 0.0001989697469698823, "loss": 0.8034, "step": 1311 }, { "epoch": 0.07, "learning_rate": 0.00019896711491515482, "loss": 0.7258, "step": 1312 }, { "epoch": 0.07, "learning_rate": 0.00019896447952002762, "loss": 0.7927, "step": 1313 }, { "epoch": 0.07, "learning_rate": 0.00019896184078458973, "loss": 0.7148, "step": 1314 }, { "epoch": 0.07, "learning_rate": 0.00019895919870893013, "loss": 0.8017, "step": 1315 }, { "epoch": 0.07, "learning_rate": 0.0001989565532931381, "loss": 0.802, "step": 1316 }, { "epoch": 0.07, "learning_rate": 0.00019895390453730284, "loss": 0.747, "step": 1317 }, { "epoch": 0.07, "learning_rate": 0.0001989512524415138, "loss": 0.827, "step": 1318 }, { "epoch": 0.07, "learning_rate": 0.00019894859700586047, "loss": 0.6567, "step": 1319 }, { "epoch": 0.07, "learning_rate": 0.0001989459382304325, "loss": 0.7736, "step": 1320 }, { "epoch": 0.07, "learning_rate": 0.0001989432761153196, "loss": 0.8153, "step": 1321 }, { "epoch": 0.07, "learning_rate": 0.00019894061066061167, "loss": 0.7456, "step": 1322 }, { "epoch": 0.08, "learning_rate": 0.00019893794186639862, "loss": 0.6937, "step": 1323 }, { "epoch": 0.08, "learning_rate": 0.00019893526973277056, "loss": 0.7878, "step": 1324 }, { "epoch": 0.08, "learning_rate": 0.0001989325942598177, "loss": 0.6457, "step": 1325 }, { "epoch": 0.08, "learning_rate": 0.00019892991544763033, "loss": 0.7185, "step": 1326 }, { "epoch": 0.08, "learning_rate": 0.00019892723329629887, "loss": 0.7578, "step": 1327 }, { "epoch": 0.08, "learning_rate": 0.00019892454780591378, "loss": 0.8476, "step": 1328 }, { "epoch": 0.08, "learning_rate": 0.00019892185897656578, "loss": 0.7939, "step": 1329 }, { "epoch": 0.08, "learning_rate": 0.00019891916680834566, "loss": 0.6923, "step": 1330 }, { "epoch": 0.08, "learning_rate": 0.0001989164713013442, "loss": 0.7521, "step": 1331 }, { "epoch": 0.08, "learning_rate": 0.00019891377245565242, "loss": 0.7065, "step": 1332 }, { "epoch": 0.08, "learning_rate": 0.00019891107027136137, "loss": 0.7104, "step": 1333 }, { "epoch": 0.08, "learning_rate": 0.00019890836474856233, "loss": 0.7582, "step": 1334 }, { "epoch": 0.08, "learning_rate": 0.00019890565588734658, "loss": 0.6676, "step": 1335 }, { "epoch": 0.08, "learning_rate": 0.00019890294368780557, "loss": 0.7426, "step": 1336 }, { "epoch": 0.08, "learning_rate": 0.00019890022815003083, "loss": 0.7424, "step": 1337 }, { "epoch": 0.08, "learning_rate": 0.000198897509274114, "loss": 0.828, "step": 1338 }, { "epoch": 0.08, "learning_rate": 0.00019889478706014687, "loss": 0.749, "step": 1339 }, { "epoch": 0.08, "learning_rate": 0.0001988920615082213, "loss": 0.708, "step": 1340 }, { "epoch": 0.08, "learning_rate": 0.0001988893326184293, "loss": 0.7733, "step": 1341 }, { "epoch": 0.08, "learning_rate": 0.00019888660039086298, "loss": 0.7258, "step": 1342 }, { "epoch": 0.08, "learning_rate": 0.00019888386482561456, "loss": 0.7637, "step": 1343 }, { "epoch": 0.08, "learning_rate": 0.0001988811259227764, "loss": 0.6321, "step": 1344 }, { "epoch": 0.08, "learning_rate": 0.00019887838368244087, "loss": 0.7427, "step": 1345 }, { "epoch": 0.08, "learning_rate": 0.0001988756381047006, "loss": 0.7827, "step": 1346 }, { "epoch": 0.08, "learning_rate": 0.0001988728891896482, "loss": 0.8127, "step": 1347 }, { "epoch": 0.08, "learning_rate": 0.00019887013693737653, "loss": 0.6663, "step": 1348 }, { "epoch": 0.08, "learning_rate": 0.00019886738134797843, "loss": 0.6561, "step": 1349 }, { "epoch": 0.08, "learning_rate": 0.0001988646224215469, "loss": 0.7683, "step": 1350 }, { "epoch": 0.08, "learning_rate": 0.0001988618601581751, "loss": 0.7447, "step": 1351 }, { "epoch": 0.08, "learning_rate": 0.00019885909455795623, "loss": 0.7518, "step": 1352 }, { "epoch": 0.08, "learning_rate": 0.00019885632562098368, "loss": 0.6797, "step": 1353 }, { "epoch": 0.08, "learning_rate": 0.00019885355334735082, "loss": 0.7319, "step": 1354 }, { "epoch": 0.08, "learning_rate": 0.00019885077773715133, "loss": 0.7111, "step": 1355 }, { "epoch": 0.08, "learning_rate": 0.00019884799879047886, "loss": 0.6545, "step": 1356 }, { "epoch": 0.08, "learning_rate": 0.00019884521650742715, "loss": 0.684, "step": 1357 }, { "epoch": 0.08, "learning_rate": 0.00019884243088809017, "loss": 0.7589, "step": 1358 }, { "epoch": 0.08, "learning_rate": 0.00019883964193256196, "loss": 0.7444, "step": 1359 }, { "epoch": 0.08, "learning_rate": 0.00019883684964093655, "loss": 0.6735, "step": 1360 }, { "epoch": 0.08, "learning_rate": 0.0001988340540133083, "loss": 0.7163, "step": 1361 }, { "epoch": 0.08, "learning_rate": 0.0001988312550497715, "loss": 0.7429, "step": 1362 }, { "epoch": 0.08, "learning_rate": 0.0001988284527504207, "loss": 0.8951, "step": 1363 }, { "epoch": 0.08, "learning_rate": 0.00019882564711535037, "loss": 0.8531, "step": 1364 }, { "epoch": 0.08, "learning_rate": 0.0001988228381446553, "loss": 0.7732, "step": 1365 }, { "epoch": 0.08, "learning_rate": 0.00019882002583843024, "loss": 0.7382, "step": 1366 }, { "epoch": 0.08, "learning_rate": 0.0001988172101967702, "loss": 0.7766, "step": 1367 }, { "epoch": 0.08, "learning_rate": 0.0001988143912197701, "loss": 0.7856, "step": 1368 }, { "epoch": 0.08, "learning_rate": 0.00019881156890752517, "loss": 0.7924, "step": 1369 }, { "epoch": 0.08, "learning_rate": 0.00019880874326013066, "loss": 0.783, "step": 1370 }, { "epoch": 0.08, "learning_rate": 0.00019880591427768192, "loss": 0.7, "step": 1371 }, { "epoch": 0.08, "learning_rate": 0.00019880308196027445, "loss": 0.7228, "step": 1372 }, { "epoch": 0.08, "learning_rate": 0.0001988002463080038, "loss": 0.7317, "step": 1373 }, { "epoch": 0.08, "learning_rate": 0.0001987974073209658, "loss": 0.714, "step": 1374 }, { "epoch": 0.08, "learning_rate": 0.00019879456499925614, "loss": 0.6976, "step": 1375 }, { "epoch": 0.08, "learning_rate": 0.00019879171934297082, "loss": 0.7505, "step": 1376 }, { "epoch": 0.08, "learning_rate": 0.0001987888703522059, "loss": 0.7698, "step": 1377 }, { "epoch": 0.08, "learning_rate": 0.00019878601802705753, "loss": 0.7995, "step": 1378 }, { "epoch": 0.08, "learning_rate": 0.00019878316236762196, "loss": 0.7591, "step": 1379 }, { "epoch": 0.08, "learning_rate": 0.0001987803033739956, "loss": 0.7731, "step": 1380 }, { "epoch": 0.08, "learning_rate": 0.0001987774410462749, "loss": 0.7452, "step": 1381 }, { "epoch": 0.08, "learning_rate": 0.00019877457538455655, "loss": 0.8198, "step": 1382 }, { "epoch": 0.08, "learning_rate": 0.00019877170638893724, "loss": 0.6633, "step": 1383 }, { "epoch": 0.08, "learning_rate": 0.00019876883405951377, "loss": 0.7995, "step": 1384 }, { "epoch": 0.08, "learning_rate": 0.00019876595839638314, "loss": 0.6193, "step": 1385 }, { "epoch": 0.08, "learning_rate": 0.0001987630793996424, "loss": 0.7973, "step": 1386 }, { "epoch": 0.08, "learning_rate": 0.00019876019706938872, "loss": 0.9458, "step": 1387 }, { "epoch": 0.08, "learning_rate": 0.00019875731140571933, "loss": 0.7106, "step": 1388 }, { "epoch": 0.08, "learning_rate": 0.00019875442240873173, "loss": 0.7223, "step": 1389 }, { "epoch": 0.08, "learning_rate": 0.0001987515300785234, "loss": 0.7555, "step": 1390 }, { "epoch": 0.08, "learning_rate": 0.00019874863441519188, "loss": 0.6703, "step": 1391 }, { "epoch": 0.08, "learning_rate": 0.000198745735418835, "loss": 0.761, "step": 1392 }, { "epoch": 0.08, "learning_rate": 0.00019874283308955057, "loss": 0.779, "step": 1393 }, { "epoch": 0.08, "learning_rate": 0.0001987399274274366, "loss": 0.726, "step": 1394 }, { "epoch": 0.08, "learning_rate": 0.00019873701843259108, "loss": 0.8051, "step": 1395 }, { "epoch": 0.08, "learning_rate": 0.00019873410610511224, "loss": 0.7293, "step": 1396 }, { "epoch": 0.08, "learning_rate": 0.0001987311904450984, "loss": 0.8437, "step": 1397 }, { "epoch": 0.08, "learning_rate": 0.00019872827145264794, "loss": 0.7721, "step": 1398 }, { "epoch": 0.08, "learning_rate": 0.00019872534912785943, "loss": 0.8051, "step": 1399 }, { "epoch": 0.08, "learning_rate": 0.0001987224234708314, "loss": 0.7657, "step": 1400 }, { "epoch": 0.08, "learning_rate": 0.00019871949448166273, "loss": 0.8171, "step": 1401 }, { "epoch": 0.08, "learning_rate": 0.0001987165621604522, "loss": 0.8208, "step": 1402 }, { "epoch": 0.08, "learning_rate": 0.0001987136265072988, "loss": 0.849, "step": 1403 }, { "epoch": 0.08, "learning_rate": 0.00019871068752230162, "loss": 0.7305, "step": 1404 }, { "epoch": 0.08, "learning_rate": 0.00019870774520555986, "loss": 0.8359, "step": 1405 }, { "epoch": 0.08, "learning_rate": 0.00019870479955717282, "loss": 0.6565, "step": 1406 }, { "epoch": 0.08, "learning_rate": 0.00019870185057723996, "loss": 0.8074, "step": 1407 }, { "epoch": 0.08, "learning_rate": 0.00019869889826586076, "loss": 0.768, "step": 1408 }, { "epoch": 0.08, "learning_rate": 0.0001986959426231349, "loss": 0.8292, "step": 1409 }, { "epoch": 0.08, "learning_rate": 0.00019869298364916214, "loss": 0.7863, "step": 1410 }, { "epoch": 0.08, "learning_rate": 0.00019869002134404235, "loss": 0.6221, "step": 1411 }, { "epoch": 0.08, "learning_rate": 0.00019868705570787553, "loss": 0.7958, "step": 1412 }, { "epoch": 0.08, "learning_rate": 0.00019868408674076173, "loss": 0.7656, "step": 1413 }, { "epoch": 0.08, "learning_rate": 0.00019868111444280121, "loss": 0.7937, "step": 1414 }, { "epoch": 0.08, "learning_rate": 0.0001986781388140943, "loss": 0.7439, "step": 1415 }, { "epoch": 0.08, "learning_rate": 0.0001986751598547414, "loss": 0.7448, "step": 1416 }, { "epoch": 0.08, "learning_rate": 0.00019867217756484306, "loss": 0.7071, "step": 1417 }, { "epoch": 0.08, "learning_rate": 0.00019866919194449993, "loss": 0.7869, "step": 1418 }, { "epoch": 0.08, "learning_rate": 0.00019866620299381285, "loss": 0.7614, "step": 1419 }, { "epoch": 0.08, "learning_rate": 0.0001986632107128826, "loss": 0.7972, "step": 1420 }, { "epoch": 0.08, "learning_rate": 0.0001986602151018103, "loss": 0.712, "step": 1421 }, { "epoch": 0.08, "learning_rate": 0.00019865721616069696, "loss": 0.7545, "step": 1422 }, { "epoch": 0.08, "learning_rate": 0.00019865421388964383, "loss": 0.7788, "step": 1423 }, { "epoch": 0.08, "learning_rate": 0.00019865120828875226, "loss": 0.8366, "step": 1424 }, { "epoch": 0.08, "learning_rate": 0.00019864819935812368, "loss": 0.7232, "step": 1425 }, { "epoch": 0.08, "learning_rate": 0.00019864518709785967, "loss": 0.7871, "step": 1426 }, { "epoch": 0.08, "learning_rate": 0.0001986421715080619, "loss": 0.8289, "step": 1427 }, { "epoch": 0.08, "learning_rate": 0.00019863915258883213, "loss": 0.6836, "step": 1428 }, { "epoch": 0.08, "learning_rate": 0.00019863613034027224, "loss": 0.6383, "step": 1429 }, { "epoch": 0.08, "learning_rate": 0.0001986331047624843, "loss": 0.7151, "step": 1430 }, { "epoch": 0.08, "learning_rate": 0.0001986300758555704, "loss": 0.8442, "step": 1431 }, { "epoch": 0.08, "learning_rate": 0.00019862704361963275, "loss": 0.7219, "step": 1432 }, { "epoch": 0.08, "learning_rate": 0.00019862400805477375, "loss": 0.6897, "step": 1433 }, { "epoch": 0.08, "learning_rate": 0.0001986209691610958, "loss": 0.7795, "step": 1434 }, { "epoch": 0.08, "learning_rate": 0.00019861792693870155, "loss": 0.7545, "step": 1435 }, { "epoch": 0.08, "learning_rate": 0.00019861488138769358, "loss": 0.6969, "step": 1436 }, { "epoch": 0.08, "learning_rate": 0.00019861183250817472, "loss": 0.773, "step": 1437 }, { "epoch": 0.08, "learning_rate": 0.0001986087803002479, "loss": 0.7113, "step": 1438 }, { "epoch": 0.08, "learning_rate": 0.0001986057247640162, "loss": 0.7818, "step": 1439 }, { "epoch": 0.08, "learning_rate": 0.00019860266589958263, "loss": 0.8362, "step": 1440 }, { "epoch": 0.08, "learning_rate": 0.0001985996037070505, "loss": 0.7023, "step": 1441 }, { "epoch": 0.08, "learning_rate": 0.00019859653818652317, "loss": 0.6854, "step": 1442 }, { "epoch": 0.08, "learning_rate": 0.00019859346933810408, "loss": 0.6562, "step": 1443 }, { "epoch": 0.08, "learning_rate": 0.00019859039716189686, "loss": 0.5922, "step": 1444 }, { "epoch": 0.08, "learning_rate": 0.00019858732165800512, "loss": 0.8722, "step": 1445 }, { "epoch": 0.08, "learning_rate": 0.00019858424282653277, "loss": 0.6886, "step": 1446 }, { "epoch": 0.08, "learning_rate": 0.00019858116066758362, "loss": 0.8635, "step": 1447 }, { "epoch": 0.08, "learning_rate": 0.0001985780751812618, "loss": 0.6191, "step": 1448 }, { "epoch": 0.08, "learning_rate": 0.00019857498636767143, "loss": 0.6716, "step": 1449 }, { "epoch": 0.08, "learning_rate": 0.00019857189422691672, "loss": 0.6414, "step": 1450 }, { "epoch": 0.08, "learning_rate": 0.00019856879875910206, "loss": 0.7398, "step": 1451 }, { "epoch": 0.08, "learning_rate": 0.00019856569996433196, "loss": 0.7935, "step": 1452 }, { "epoch": 0.08, "learning_rate": 0.000198562597842711, "loss": 0.7714, "step": 1453 }, { "epoch": 0.08, "learning_rate": 0.0001985594923943438, "loss": 0.813, "step": 1454 }, { "epoch": 0.08, "learning_rate": 0.00019855638361933534, "loss": 0.7666, "step": 1455 }, { "epoch": 0.08, "learning_rate": 0.00019855327151779042, "loss": 0.7497, "step": 1456 }, { "epoch": 0.08, "learning_rate": 0.00019855015608981412, "loss": 0.7814, "step": 1457 }, { "epoch": 0.08, "learning_rate": 0.00019854703733551156, "loss": 0.7043, "step": 1458 }, { "epoch": 0.08, "learning_rate": 0.0001985439152549881, "loss": 0.7865, "step": 1459 }, { "epoch": 0.08, "learning_rate": 0.00019854078984834903, "loss": 0.6759, "step": 1460 }, { "epoch": 0.08, "learning_rate": 0.00019853766111569988, "loss": 0.7526, "step": 1461 }, { "epoch": 0.08, "learning_rate": 0.00019853452905714623, "loss": 0.7431, "step": 1462 }, { "epoch": 0.08, "learning_rate": 0.0001985313936727938, "loss": 0.8478, "step": 1463 }, { "epoch": 0.08, "learning_rate": 0.00019852825496274843, "loss": 0.7462, "step": 1464 }, { "epoch": 0.08, "learning_rate": 0.00019852511292711608, "loss": 0.8625, "step": 1465 }, { "epoch": 0.08, "learning_rate": 0.00019852196756600277, "loss": 0.7544, "step": 1466 }, { "epoch": 0.08, "learning_rate": 0.00019851881887951467, "loss": 0.7784, "step": 1467 }, { "epoch": 0.08, "learning_rate": 0.00019851566686775806, "loss": 0.7641, "step": 1468 }, { "epoch": 0.08, "learning_rate": 0.0001985125115308393, "loss": 0.7841, "step": 1469 }, { "epoch": 0.08, "learning_rate": 0.00019850935286886494, "loss": 0.6611, "step": 1470 }, { "epoch": 0.08, "learning_rate": 0.00019850619088194156, "loss": 0.7501, "step": 1471 }, { "epoch": 0.08, "learning_rate": 0.0001985030255701759, "loss": 0.8028, "step": 1472 }, { "epoch": 0.08, "learning_rate": 0.00019849985693367477, "loss": 0.7891, "step": 1473 }, { "epoch": 0.08, "learning_rate": 0.0001984966849725452, "loss": 0.8329, "step": 1474 }, { "epoch": 0.08, "learning_rate": 0.00019849350968689414, "loss": 0.7662, "step": 1475 }, { "epoch": 0.08, "learning_rate": 0.00019849033107682885, "loss": 0.8613, "step": 1476 }, { "epoch": 0.08, "learning_rate": 0.00019848714914245656, "loss": 0.718, "step": 1477 }, { "epoch": 0.08, "learning_rate": 0.00019848396388388474, "loss": 0.623, "step": 1478 }, { "epoch": 0.08, "learning_rate": 0.00019848077530122083, "loss": 0.8318, "step": 1479 }, { "epoch": 0.08, "learning_rate": 0.00019847758339457244, "loss": 0.9068, "step": 1480 }, { "epoch": 0.08, "learning_rate": 0.0001984743881640474, "loss": 0.7767, "step": 1481 }, { "epoch": 0.08, "learning_rate": 0.0001984711896097535, "loss": 0.7315, "step": 1482 }, { "epoch": 0.08, "learning_rate": 0.00019846798773179866, "loss": 0.8836, "step": 1483 }, { "epoch": 0.08, "learning_rate": 0.000198464782530291, "loss": 0.6713, "step": 1484 }, { "epoch": 0.08, "learning_rate": 0.00019846157400533874, "loss": 0.6291, "step": 1485 }, { "epoch": 0.08, "learning_rate": 0.00019845836215705008, "loss": 0.7482, "step": 1486 }, { "epoch": 0.08, "learning_rate": 0.0001984551469855335, "loss": 0.723, "step": 1487 }, { "epoch": 0.08, "learning_rate": 0.00019845192849089752, "loss": 0.6929, "step": 1488 }, { "epoch": 0.08, "learning_rate": 0.00019844870667325073, "loss": 0.7097, "step": 1489 }, { "epoch": 0.08, "learning_rate": 0.0001984454815327019, "loss": 0.7346, "step": 1490 }, { "epoch": 0.08, "learning_rate": 0.00019844225306935988, "loss": 0.7834, "step": 1491 }, { "epoch": 0.08, "learning_rate": 0.00019843902128333367, "loss": 0.8113, "step": 1492 }, { "epoch": 0.08, "learning_rate": 0.0001984357861747323, "loss": 0.7149, "step": 1493 }, { "epoch": 0.08, "learning_rate": 0.000198432547743665, "loss": 0.7881, "step": 1494 }, { "epoch": 0.08, "learning_rate": 0.0001984293059902411, "loss": 0.7572, "step": 1495 }, { "epoch": 0.08, "learning_rate": 0.0001984260609145699, "loss": 0.7914, "step": 1496 }, { "epoch": 0.08, "learning_rate": 0.00019842281251676109, "loss": 0.7055, "step": 1497 }, { "epoch": 0.08, "learning_rate": 0.0001984195607969242, "loss": 0.7154, "step": 1498 }, { "epoch": 0.09, "learning_rate": 0.00019841630575516902, "loss": 0.6405, "step": 1499 }, { "epoch": 0.09, "learning_rate": 0.0001984130473916054, "loss": 0.6734, "step": 1500 }, { "epoch": 0.09, "learning_rate": 0.0001984097857063434, "loss": 0.7612, "step": 1501 }, { "epoch": 0.09, "learning_rate": 0.00019840652069949297, "loss": 0.7159, "step": 1502 }, { "epoch": 0.09, "learning_rate": 0.0001984032523711644, "loss": 0.7991, "step": 1503 }, { "epoch": 0.09, "learning_rate": 0.000198399980721468, "loss": 0.759, "step": 1504 }, { "epoch": 0.09, "learning_rate": 0.0001983967057505142, "loss": 0.8391, "step": 1505 }, { "epoch": 0.09, "learning_rate": 0.00019839342745841352, "loss": 0.795, "step": 1506 }, { "epoch": 0.09, "learning_rate": 0.0001983901458452766, "loss": 0.7052, "step": 1507 }, { "epoch": 0.09, "learning_rate": 0.00019838686091121425, "loss": 0.7168, "step": 1508 }, { "epoch": 0.09, "learning_rate": 0.00019838357265633728, "loss": 0.6789, "step": 1509 }, { "epoch": 0.09, "learning_rate": 0.00019838028108075671, "loss": 0.7493, "step": 1510 }, { "epoch": 0.09, "learning_rate": 0.00019837698618458367, "loss": 0.7364, "step": 1511 }, { "epoch": 0.09, "learning_rate": 0.00019837368796792935, "loss": 0.7725, "step": 1512 }, { "epoch": 0.09, "learning_rate": 0.00019837038643090503, "loss": 0.699, "step": 1513 }, { "epoch": 0.09, "learning_rate": 0.00019836708157362223, "loss": 0.7675, "step": 1514 }, { "epoch": 0.09, "learning_rate": 0.0001983637733961924, "loss": 0.7979, "step": 1515 }, { "epoch": 0.09, "learning_rate": 0.00019836046189872727, "loss": 0.7667, "step": 1516 }, { "epoch": 0.09, "learning_rate": 0.00019835714708133862, "loss": 0.746, "step": 1517 }, { "epoch": 0.09, "learning_rate": 0.00019835382894413826, "loss": 0.7428, "step": 1518 }, { "epoch": 0.09, "learning_rate": 0.00019835050748723824, "loss": 0.7996, "step": 1519 }, { "epoch": 0.09, "learning_rate": 0.00019834718271075067, "loss": 0.7071, "step": 1520 }, { "epoch": 0.09, "learning_rate": 0.00019834385461478776, "loss": 0.749, "step": 1521 }, { "epoch": 0.09, "learning_rate": 0.00019834052319946184, "loss": 0.7819, "step": 1522 }, { "epoch": 0.09, "learning_rate": 0.00019833718846488534, "loss": 0.6704, "step": 1523 }, { "epoch": 0.09, "learning_rate": 0.00019833385041117084, "loss": 0.706, "step": 1524 }, { "epoch": 0.09, "learning_rate": 0.000198330509038431, "loss": 0.7207, "step": 1525 }, { "epoch": 0.09, "learning_rate": 0.0001983271643467786, "loss": 0.6005, "step": 1526 }, { "epoch": 0.09, "learning_rate": 0.00019832381633632653, "loss": 0.6445, "step": 1527 }, { "epoch": 0.09, "learning_rate": 0.0001983204650071878, "loss": 0.6549, "step": 1528 }, { "epoch": 0.09, "learning_rate": 0.0001983171103594755, "loss": 0.7372, "step": 1529 }, { "epoch": 0.09, "learning_rate": 0.00019831375239330292, "loss": 0.7385, "step": 1530 }, { "epoch": 0.09, "learning_rate": 0.00019831039110878332, "loss": 0.7977, "step": 1531 }, { "epoch": 0.09, "learning_rate": 0.00019830702650603022, "loss": 0.7464, "step": 1532 }, { "epoch": 0.09, "learning_rate": 0.00019830365858515713, "loss": 0.7257, "step": 1533 }, { "epoch": 0.09, "learning_rate": 0.00019830028734627776, "loss": 0.8189, "step": 1534 }, { "epoch": 0.09, "learning_rate": 0.00019829691278950589, "loss": 0.7207, "step": 1535 }, { "epoch": 0.09, "learning_rate": 0.00019829353491495545, "loss": 0.8001, "step": 1536 }, { "epoch": 0.09, "learning_rate": 0.00019829015372274038, "loss": 0.6812, "step": 1537 }, { "epoch": 0.09, "learning_rate": 0.00019828676921297486, "loss": 0.7784, "step": 1538 }, { "epoch": 0.09, "learning_rate": 0.0001982833813857731, "loss": 0.8326, "step": 1539 }, { "epoch": 0.09, "learning_rate": 0.0001982799902412495, "loss": 0.7582, "step": 1540 }, { "epoch": 0.09, "learning_rate": 0.00019827659577951844, "loss": 0.8103, "step": 1541 }, { "epoch": 0.09, "learning_rate": 0.00019827319800069457, "loss": 0.8036, "step": 1542 }, { "epoch": 0.09, "learning_rate": 0.00019826979690489252, "loss": 0.8115, "step": 1543 }, { "epoch": 0.09, "learning_rate": 0.0001982663924922271, "loss": 0.7261, "step": 1544 }, { "epoch": 0.09, "learning_rate": 0.0001982629847628132, "loss": 0.7475, "step": 1545 }, { "epoch": 0.09, "learning_rate": 0.00019825957371676588, "loss": 0.7108, "step": 1546 }, { "epoch": 0.09, "learning_rate": 0.00019825615935420025, "loss": 0.6135, "step": 1547 }, { "epoch": 0.09, "learning_rate": 0.00019825274167523156, "loss": 0.7239, "step": 1548 }, { "epoch": 0.09, "learning_rate": 0.00019824932067997515, "loss": 0.6355, "step": 1549 }, { "epoch": 0.09, "learning_rate": 0.00019824589636854652, "loss": 0.7111, "step": 1550 }, { "epoch": 0.09, "learning_rate": 0.00019824246874106121, "loss": 0.7738, "step": 1551 }, { "epoch": 0.09, "learning_rate": 0.00019823903779763493, "loss": 0.7816, "step": 1552 }, { "epoch": 0.09, "learning_rate": 0.0001982356035383835, "loss": 0.7382, "step": 1553 }, { "epoch": 0.09, "learning_rate": 0.0001982321659634228, "loss": 0.7448, "step": 1554 }, { "epoch": 0.09, "learning_rate": 0.0001982287250728689, "loss": 0.7989, "step": 1555 }, { "epoch": 0.09, "learning_rate": 0.00019822528086683788, "loss": 0.7272, "step": 1556 }, { "epoch": 0.09, "learning_rate": 0.00019822183334544603, "loss": 0.7856, "step": 1557 }, { "epoch": 0.09, "learning_rate": 0.0001982183825088097, "loss": 0.7336, "step": 1558 }, { "epoch": 0.09, "learning_rate": 0.0001982149283570454, "loss": 0.835, "step": 1559 }, { "epoch": 0.09, "learning_rate": 0.0001982114708902697, "loss": 0.7302, "step": 1560 }, { "epoch": 0.09, "learning_rate": 0.0001982080101085993, "loss": 0.6727, "step": 1561 }, { "epoch": 0.09, "learning_rate": 0.00019820454601215094, "loss": 0.7769, "step": 1562 }, { "epoch": 0.09, "learning_rate": 0.00019820107860104162, "loss": 0.778, "step": 1563 }, { "epoch": 0.09, "learning_rate": 0.0001981976078753884, "loss": 0.723, "step": 1564 }, { "epoch": 0.09, "learning_rate": 0.00019819413383530835, "loss": 0.657, "step": 1565 }, { "epoch": 0.09, "learning_rate": 0.00019819065648091875, "loss": 0.8065, "step": 1566 }, { "epoch": 0.09, "learning_rate": 0.00019818717581233697, "loss": 0.7287, "step": 1567 }, { "epoch": 0.09, "learning_rate": 0.00019818369182968054, "loss": 0.6302, "step": 1568 }, { "epoch": 0.09, "learning_rate": 0.00019818020453306697, "loss": 0.6972, "step": 1569 }, { "epoch": 0.09, "learning_rate": 0.00019817671392261403, "loss": 0.7482, "step": 1570 }, { "epoch": 0.09, "learning_rate": 0.00019817321999843952, "loss": 0.7584, "step": 1571 }, { "epoch": 0.09, "learning_rate": 0.00019816972276066137, "loss": 0.8039, "step": 1572 }, { "epoch": 0.09, "learning_rate": 0.0001981662222093976, "loss": 0.819, "step": 1573 }, { "epoch": 0.09, "learning_rate": 0.00019816271834476642, "loss": 0.8415, "step": 1574 }, { "epoch": 0.09, "learning_rate": 0.000198159211166886, "loss": 0.7319, "step": 1575 }, { "epoch": 0.09, "learning_rate": 0.0001981557006758748, "loss": 0.7776, "step": 1576 }, { "epoch": 0.09, "learning_rate": 0.0001981521868718513, "loss": 0.7844, "step": 1577 }, { "epoch": 0.09, "learning_rate": 0.00019814866975493404, "loss": 0.8178, "step": 1578 }, { "epoch": 0.09, "learning_rate": 0.0001981451493252418, "loss": 0.7987, "step": 1579 }, { "epoch": 0.09, "learning_rate": 0.00019814162558289336, "loss": 0.8104, "step": 1580 }, { "epoch": 0.09, "learning_rate": 0.00019813809852800769, "loss": 0.7469, "step": 1581 }, { "epoch": 0.09, "learning_rate": 0.0001981345681607038, "loss": 0.6907, "step": 1582 }, { "epoch": 0.09, "learning_rate": 0.00019813103448110085, "loss": 0.6895, "step": 1583 }, { "epoch": 0.09, "learning_rate": 0.00019812749748931817, "loss": 0.7141, "step": 1584 }, { "epoch": 0.09, "learning_rate": 0.00019812395718547506, "loss": 0.6711, "step": 1585 }, { "epoch": 0.09, "learning_rate": 0.0001981204135696911, "loss": 0.7824, "step": 1586 }, { "epoch": 0.09, "learning_rate": 0.0001981168666420858, "loss": 0.7002, "step": 1587 }, { "epoch": 0.09, "learning_rate": 0.00019811331640277898, "loss": 0.6787, "step": 1588 }, { "epoch": 0.09, "learning_rate": 0.00019810976285189038, "loss": 0.7458, "step": 1589 }, { "epoch": 0.09, "learning_rate": 0.00019810620598954, "loss": 0.7103, "step": 1590 }, { "epoch": 0.09, "learning_rate": 0.00019810264581584787, "loss": 0.6099, "step": 1591 }, { "epoch": 0.09, "learning_rate": 0.00019809908233093417, "loss": 0.8913, "step": 1592 }, { "epoch": 0.09, "learning_rate": 0.00019809551553491916, "loss": 0.7709, "step": 1593 }, { "epoch": 0.09, "learning_rate": 0.00019809194542792324, "loss": 0.7075, "step": 1594 }, { "epoch": 0.09, "learning_rate": 0.0001980883720100669, "loss": 0.8302, "step": 1595 }, { "epoch": 0.09, "learning_rate": 0.00019808479528147076, "loss": 0.8193, "step": 1596 }, { "epoch": 0.09, "learning_rate": 0.00019808121524225554, "loss": 0.7602, "step": 1597 }, { "epoch": 0.09, "learning_rate": 0.0001980776318925421, "loss": 0.7048, "step": 1598 }, { "epoch": 0.09, "learning_rate": 0.00019807404523245134, "loss": 0.723, "step": 1599 }, { "epoch": 0.09, "learning_rate": 0.00019807045526210436, "loss": 0.7305, "step": 1600 }, { "epoch": 0.09, "learning_rate": 0.0001980668619816223, "loss": 0.7289, "step": 1601 }, { "epoch": 0.09, "learning_rate": 0.00019806326539112648, "loss": 0.8033, "step": 1602 }, { "epoch": 0.09, "learning_rate": 0.00019805966549073825, "loss": 0.7408, "step": 1603 }, { "epoch": 0.09, "learning_rate": 0.00019805606228057916, "loss": 0.7192, "step": 1604 }, { "epoch": 0.09, "learning_rate": 0.00019805245576077075, "loss": 0.7629, "step": 1605 }, { "epoch": 0.09, "learning_rate": 0.00019804884593143486, "loss": 0.7138, "step": 1606 }, { "epoch": 0.09, "learning_rate": 0.0001980452327926933, "loss": 0.8541, "step": 1607 }, { "epoch": 0.09, "learning_rate": 0.00019804161634466797, "loss": 0.844, "step": 1608 }, { "epoch": 0.09, "learning_rate": 0.00019803799658748094, "loss": 0.7462, "step": 1609 }, { "epoch": 0.09, "learning_rate": 0.00019803437352125444, "loss": 0.7212, "step": 1610 }, { "epoch": 0.09, "learning_rate": 0.0001980307471461107, "loss": 0.7458, "step": 1611 }, { "epoch": 0.09, "learning_rate": 0.00019802711746217218, "loss": 0.7388, "step": 1612 }, { "epoch": 0.09, "learning_rate": 0.00019802348446956136, "loss": 0.8463, "step": 1613 }, { "epoch": 0.09, "learning_rate": 0.00019801984816840085, "loss": 0.7251, "step": 1614 }, { "epoch": 0.09, "learning_rate": 0.0001980162085588134, "loss": 0.7054, "step": 1615 }, { "epoch": 0.09, "learning_rate": 0.00019801256564092187, "loss": 0.7759, "step": 1616 }, { "epoch": 0.09, "learning_rate": 0.0001980089194148492, "loss": 0.775, "step": 1617 }, { "epoch": 0.09, "learning_rate": 0.00019800526988071843, "loss": 0.72, "step": 1618 }, { "epoch": 0.09, "learning_rate": 0.00019800161703865282, "loss": 0.788, "step": 1619 }, { "epoch": 0.09, "learning_rate": 0.00019799796088877558, "loss": 0.6888, "step": 1620 }, { "epoch": 0.09, "learning_rate": 0.00019799430143121015, "loss": 0.7546, "step": 1621 }, { "epoch": 0.09, "learning_rate": 0.00019799063866608006, "loss": 0.6722, "step": 1622 }, { "epoch": 0.09, "learning_rate": 0.00019798697259350894, "loss": 0.622, "step": 1623 }, { "epoch": 0.09, "learning_rate": 0.0001979833032136205, "loss": 0.7496, "step": 1624 }, { "epoch": 0.09, "learning_rate": 0.0001979796305265386, "loss": 0.7316, "step": 1625 }, { "epoch": 0.09, "learning_rate": 0.0001979759545323872, "loss": 0.6537, "step": 1626 }, { "epoch": 0.09, "learning_rate": 0.0001979722752312904, "loss": 0.8999, "step": 1627 }, { "epoch": 0.09, "learning_rate": 0.00019796859262337239, "loss": 0.7275, "step": 1628 }, { "epoch": 0.09, "learning_rate": 0.0001979649067087574, "loss": 0.6826, "step": 1629 }, { "epoch": 0.09, "learning_rate": 0.00019796121748756992, "loss": 0.7842, "step": 1630 }, { "epoch": 0.09, "learning_rate": 0.0001979575249599344, "loss": 0.7615, "step": 1631 }, { "epoch": 0.09, "learning_rate": 0.00019795382912597557, "loss": 0.7033, "step": 1632 }, { "epoch": 0.09, "learning_rate": 0.00019795012998581805, "loss": 0.7426, "step": 1633 }, { "epoch": 0.09, "learning_rate": 0.0001979464275395868, "loss": 0.8084, "step": 1634 }, { "epoch": 0.09, "learning_rate": 0.00019794272178740675, "loss": 0.7325, "step": 1635 }, { "epoch": 0.09, "learning_rate": 0.00019793901272940293, "loss": 0.733, "step": 1636 }, { "epoch": 0.09, "learning_rate": 0.00019793530036570062, "loss": 0.7087, "step": 1637 }, { "epoch": 0.09, "learning_rate": 0.00019793158469642505, "loss": 0.751, "step": 1638 }, { "epoch": 0.09, "learning_rate": 0.0001979278657217017, "loss": 0.8125, "step": 1639 }, { "epoch": 0.09, "learning_rate": 0.00019792414344165603, "loss": 0.5868, "step": 1640 }, { "epoch": 0.09, "learning_rate": 0.0001979204178564137, "loss": 0.7794, "step": 1641 }, { "epoch": 0.09, "learning_rate": 0.00019791668896610047, "loss": 0.7371, "step": 1642 }, { "epoch": 0.09, "learning_rate": 0.00019791295677084217, "loss": 0.8266, "step": 1643 }, { "epoch": 0.09, "learning_rate": 0.00019790922127076483, "loss": 0.6327, "step": 1644 }, { "epoch": 0.09, "learning_rate": 0.00019790548246599447, "loss": 0.7699, "step": 1645 }, { "epoch": 0.09, "learning_rate": 0.00019790174035665733, "loss": 0.7838, "step": 1646 }, { "epoch": 0.09, "learning_rate": 0.00019789799494287968, "loss": 0.8129, "step": 1647 }, { "epoch": 0.09, "learning_rate": 0.00019789424622478795, "loss": 0.7845, "step": 1648 }, { "epoch": 0.09, "learning_rate": 0.0001978904942025087, "loss": 0.7464, "step": 1649 }, { "epoch": 0.09, "learning_rate": 0.0001978867388761685, "loss": 0.7585, "step": 1650 }, { "epoch": 0.09, "learning_rate": 0.00019788298024589417, "loss": 0.653, "step": 1651 }, { "epoch": 0.09, "learning_rate": 0.00019787921831181256, "loss": 0.8124, "step": 1652 }, { "epoch": 0.09, "learning_rate": 0.00019787545307405063, "loss": 0.6822, "step": 1653 }, { "epoch": 0.09, "learning_rate": 0.00019787168453273544, "loss": 0.7411, "step": 1654 }, { "epoch": 0.09, "learning_rate": 0.00019786791268799428, "loss": 0.7326, "step": 1655 }, { "epoch": 0.09, "learning_rate": 0.00019786413753995435, "loss": 0.747, "step": 1656 }, { "epoch": 0.09, "learning_rate": 0.00019786035908874312, "loss": 0.7793, "step": 1657 }, { "epoch": 0.09, "learning_rate": 0.00019785657733448813, "loss": 0.7515, "step": 1658 }, { "epoch": 0.09, "learning_rate": 0.00019785279227731704, "loss": 0.7883, "step": 1659 }, { "epoch": 0.09, "learning_rate": 0.00019784900391735754, "loss": 0.6953, "step": 1660 }, { "epoch": 0.09, "learning_rate": 0.0001978452122547376, "loss": 0.7138, "step": 1661 }, { "epoch": 0.09, "learning_rate": 0.00019784141728958508, "loss": 0.7333, "step": 1662 }, { "epoch": 0.09, "learning_rate": 0.00019783761902202813, "loss": 0.7695, "step": 1663 }, { "epoch": 0.09, "learning_rate": 0.00019783381745219498, "loss": 0.7312, "step": 1664 }, { "epoch": 0.09, "learning_rate": 0.00019783001258021388, "loss": 0.6569, "step": 1665 }, { "epoch": 0.09, "learning_rate": 0.00019782620440621332, "loss": 0.7867, "step": 1666 }, { "epoch": 0.09, "learning_rate": 0.0001978223929303218, "loss": 0.7524, "step": 1667 }, { "epoch": 0.09, "learning_rate": 0.000197818578152668, "loss": 0.7709, "step": 1668 }, { "epoch": 0.09, "learning_rate": 0.00019781476007338058, "loss": 0.7884, "step": 1669 }, { "epoch": 0.09, "learning_rate": 0.0001978109386925885, "loss": 0.7337, "step": 1670 }, { "epoch": 0.09, "learning_rate": 0.00019780711401042075, "loss": 0.7326, "step": 1671 }, { "epoch": 0.09, "learning_rate": 0.00019780328602700636, "loss": 0.8075, "step": 1672 }, { "epoch": 0.09, "learning_rate": 0.00019779945474247457, "loss": 0.743, "step": 1673 }, { "epoch": 0.09, "learning_rate": 0.00019779562015695472, "loss": 0.7729, "step": 1674 }, { "epoch": 0.1, "learning_rate": 0.00019779178227057617, "loss": 0.7195, "step": 1675 }, { "epoch": 0.1, "learning_rate": 0.00019778794108346853, "loss": 0.7174, "step": 1676 }, { "epoch": 0.1, "learning_rate": 0.00019778409659576142, "loss": 0.6502, "step": 1677 }, { "epoch": 0.1, "learning_rate": 0.0001977802488075846, "loss": 0.6526, "step": 1678 }, { "epoch": 0.1, "learning_rate": 0.00019777639771906795, "loss": 0.7392, "step": 1679 }, { "epoch": 0.1, "learning_rate": 0.00019777254333034144, "loss": 0.7213, "step": 1680 }, { "epoch": 0.1, "learning_rate": 0.00019776868564153516, "loss": 0.797, "step": 1681 }, { "epoch": 0.1, "learning_rate": 0.00019776482465277937, "loss": 0.768, "step": 1682 }, { "epoch": 0.1, "learning_rate": 0.00019776096036420432, "loss": 0.7777, "step": 1683 }, { "epoch": 0.1, "learning_rate": 0.00019775709277594045, "loss": 0.6167, "step": 1684 }, { "epoch": 0.1, "learning_rate": 0.00019775322188811837, "loss": 0.7037, "step": 1685 }, { "epoch": 0.1, "learning_rate": 0.00019774934770086866, "loss": 0.7973, "step": 1686 }, { "epoch": 0.1, "learning_rate": 0.00019774547021432213, "loss": 0.7138, "step": 1687 }, { "epoch": 0.1, "learning_rate": 0.0001977415894286096, "loss": 0.7658, "step": 1688 }, { "epoch": 0.1, "learning_rate": 0.0001977377053438621, "loss": 0.7353, "step": 1689 }, { "epoch": 0.1, "learning_rate": 0.0001977338179602107, "loss": 0.7424, "step": 1690 }, { "epoch": 0.1, "learning_rate": 0.00019772992727778668, "loss": 0.742, "step": 1691 }, { "epoch": 0.1, "learning_rate": 0.0001977260332967213, "loss": 0.759, "step": 1692 }, { "epoch": 0.1, "learning_rate": 0.00019772213601714595, "loss": 0.897, "step": 1693 }, { "epoch": 0.1, "learning_rate": 0.00019771823543919225, "loss": 0.7676, "step": 1694 }, { "epoch": 0.1, "learning_rate": 0.00019771433156299184, "loss": 0.6548, "step": 1695 }, { "epoch": 0.1, "learning_rate": 0.00019771042438867647, "loss": 0.7475, "step": 1696 }, { "epoch": 0.1, "learning_rate": 0.00019770651391637803, "loss": 0.6914, "step": 1697 }, { "epoch": 0.1, "learning_rate": 0.00019770260014622846, "loss": 0.7065, "step": 1698 }, { "epoch": 0.1, "learning_rate": 0.00019769868307835994, "loss": 0.8097, "step": 1699 }, { "epoch": 0.1, "learning_rate": 0.0001976947627129046, "loss": 0.7427, "step": 1700 }, { "epoch": 0.1, "learning_rate": 0.00019769083904999486, "loss": 0.7369, "step": 1701 }, { "epoch": 0.1, "learning_rate": 0.0001976869120897631, "loss": 0.6597, "step": 1702 }, { "epoch": 0.1, "learning_rate": 0.00019768298183234182, "loss": 0.7314, "step": 1703 }, { "epoch": 0.1, "learning_rate": 0.00019767904827786375, "loss": 0.776, "step": 1704 }, { "epoch": 0.1, "learning_rate": 0.0001976751114264616, "loss": 0.6836, "step": 1705 }, { "epoch": 0.1, "learning_rate": 0.00019767117127826827, "loss": 0.6975, "step": 1706 }, { "epoch": 0.1, "learning_rate": 0.0001976672278334168, "loss": 0.6739, "step": 1707 }, { "epoch": 0.1, "learning_rate": 0.00019766328109204024, "loss": 0.7895, "step": 1708 }, { "epoch": 0.1, "learning_rate": 0.0001976593310542718, "loss": 0.7947, "step": 1709 }, { "epoch": 0.1, "learning_rate": 0.00019765537772024482, "loss": 0.7036, "step": 1710 }, { "epoch": 0.1, "learning_rate": 0.00019765142109009272, "loss": 0.7078, "step": 1711 }, { "epoch": 0.1, "learning_rate": 0.00019764746116394906, "loss": 0.729, "step": 1712 }, { "epoch": 0.1, "learning_rate": 0.0001976434979419475, "loss": 0.8399, "step": 1713 }, { "epoch": 0.1, "learning_rate": 0.0001976395314242218, "loss": 0.7322, "step": 1714 }, { "epoch": 0.1, "learning_rate": 0.00019763556161090585, "loss": 0.7729, "step": 1715 }, { "epoch": 0.1, "learning_rate": 0.00019763158850213364, "loss": 0.7733, "step": 1716 }, { "epoch": 0.1, "learning_rate": 0.00019762761209803927, "loss": 0.7123, "step": 1717 }, { "epoch": 0.1, "learning_rate": 0.00019762363239875697, "loss": 0.8365, "step": 1718 }, { "epoch": 0.1, "learning_rate": 0.00019761964940442103, "loss": 0.7047, "step": 1719 }, { "epoch": 0.1, "learning_rate": 0.0001976156631151659, "loss": 0.7514, "step": 1720 }, { "epoch": 0.1, "learning_rate": 0.00019761167353112615, "loss": 0.7748, "step": 1721 }, { "epoch": 0.1, "learning_rate": 0.0001976076806524364, "loss": 0.6861, "step": 1722 }, { "epoch": 0.1, "learning_rate": 0.00019760368447923146, "loss": 0.7744, "step": 1723 }, { "epoch": 0.1, "learning_rate": 0.0001975996850116462, "loss": 0.7514, "step": 1724 }, { "epoch": 0.1, "learning_rate": 0.0001975956822498156, "loss": 0.7293, "step": 1725 }, { "epoch": 0.1, "learning_rate": 0.00019759167619387476, "loss": 0.7246, "step": 1726 }, { "epoch": 0.1, "learning_rate": 0.0001975876668439589, "loss": 0.6408, "step": 1727 }, { "epoch": 0.1, "learning_rate": 0.00019758365420020338, "loss": 0.6882, "step": 1728 }, { "epoch": 0.1, "learning_rate": 0.00019757963826274357, "loss": 0.7193, "step": 1729 }, { "epoch": 0.1, "learning_rate": 0.00019757561903171507, "loss": 0.7923, "step": 1730 }, { "epoch": 0.1, "learning_rate": 0.00019757159650725354, "loss": 0.7537, "step": 1731 }, { "epoch": 0.1, "learning_rate": 0.00019756757068949474, "loss": 0.7623, "step": 1732 }, { "epoch": 0.1, "learning_rate": 0.00019756354157857454, "loss": 0.7686, "step": 1733 }, { "epoch": 0.1, "learning_rate": 0.00019755950917462893, "loss": 0.6461, "step": 1734 }, { "epoch": 0.1, "learning_rate": 0.00019755547347779403, "loss": 0.7649, "step": 1735 }, { "epoch": 0.1, "learning_rate": 0.00019755143448820607, "loss": 0.7661, "step": 1736 }, { "epoch": 0.1, "learning_rate": 0.00019754739220600133, "loss": 0.7447, "step": 1737 }, { "epoch": 0.1, "learning_rate": 0.00019754334663131632, "loss": 0.6582, "step": 1738 }, { "epoch": 0.1, "learning_rate": 0.0001975392977642875, "loss": 0.7922, "step": 1739 }, { "epoch": 0.1, "learning_rate": 0.00019753524560505158, "loss": 0.7449, "step": 1740 }, { "epoch": 0.1, "learning_rate": 0.00019753119015374535, "loss": 0.6949, "step": 1741 }, { "epoch": 0.1, "learning_rate": 0.00019752713141050565, "loss": 0.8093, "step": 1742 }, { "epoch": 0.1, "learning_rate": 0.0001975230693754695, "loss": 0.8365, "step": 1743 }, { "epoch": 0.1, "learning_rate": 0.000197519004048774, "loss": 0.7706, "step": 1744 }, { "epoch": 0.1, "learning_rate": 0.00019751493543055632, "loss": 0.7351, "step": 1745 }, { "epoch": 0.1, "learning_rate": 0.00019751086352095388, "loss": 0.6502, "step": 1746 }, { "epoch": 0.1, "learning_rate": 0.00019750678832010404, "loss": 0.6957, "step": 1747 }, { "epoch": 0.1, "learning_rate": 0.00019750270982814435, "loss": 0.7476, "step": 1748 }, { "epoch": 0.1, "learning_rate": 0.00019749862804521254, "loss": 0.7452, "step": 1749 }, { "epoch": 0.1, "learning_rate": 0.0001974945429714463, "loss": 0.7068, "step": 1750 }, { "epoch": 0.1, "learning_rate": 0.00019749045460698358, "loss": 0.7175, "step": 1751 }, { "epoch": 0.1, "learning_rate": 0.0001974863629519623, "loss": 0.6969, "step": 1752 }, { "epoch": 0.1, "learning_rate": 0.0001974822680065206, "loss": 0.7104, "step": 1753 }, { "epoch": 0.1, "learning_rate": 0.00019747816977079671, "loss": 0.7971, "step": 1754 }, { "epoch": 0.1, "learning_rate": 0.00019747406824492897, "loss": 0.8164, "step": 1755 }, { "epoch": 0.1, "learning_rate": 0.00019746996342905577, "loss": 0.7392, "step": 1756 }, { "epoch": 0.1, "learning_rate": 0.0001974658553233157, "loss": 0.7431, "step": 1757 }, { "epoch": 0.1, "learning_rate": 0.00019746174392784737, "loss": 0.7609, "step": 1758 }, { "epoch": 0.1, "learning_rate": 0.00019745762924278958, "loss": 0.6589, "step": 1759 }, { "epoch": 0.1, "learning_rate": 0.00019745351126828123, "loss": 0.7268, "step": 1760 }, { "epoch": 0.1, "learning_rate": 0.0001974493900044613, "loss": 0.7029, "step": 1761 }, { "epoch": 0.1, "learning_rate": 0.00019744526545146886, "loss": 0.7486, "step": 1762 }, { "epoch": 0.1, "learning_rate": 0.00019744113760944315, "loss": 0.7347, "step": 1763 }, { "epoch": 0.1, "learning_rate": 0.00019743700647852354, "loss": 0.8218, "step": 1764 }, { "epoch": 0.1, "learning_rate": 0.00019743287205884938, "loss": 0.7042, "step": 1765 }, { "epoch": 0.1, "learning_rate": 0.0001974287343505603, "loss": 0.7166, "step": 1766 }, { "epoch": 0.1, "learning_rate": 0.0001974245933537959, "loss": 0.6374, "step": 1767 }, { "epoch": 0.1, "learning_rate": 0.00019742044906869596, "loss": 0.6642, "step": 1768 }, { "epoch": 0.1, "learning_rate": 0.00019741630149540035, "loss": 0.8164, "step": 1769 }, { "epoch": 0.1, "learning_rate": 0.00019741215063404913, "loss": 0.8149, "step": 1770 }, { "epoch": 0.1, "learning_rate": 0.00019740799648478233, "loss": 0.7253, "step": 1771 }, { "epoch": 0.1, "learning_rate": 0.0001974038390477402, "loss": 0.7804, "step": 1772 }, { "epoch": 0.1, "learning_rate": 0.000197399678323063, "loss": 0.7862, "step": 1773 }, { "epoch": 0.1, "learning_rate": 0.00019739551431089124, "loss": 0.7251, "step": 1774 }, { "epoch": 0.1, "learning_rate": 0.0001973913470113655, "loss": 0.8283, "step": 1775 }, { "epoch": 0.1, "learning_rate": 0.00019738717642462633, "loss": 0.7084, "step": 1776 }, { "epoch": 0.1, "learning_rate": 0.00019738300255081454, "loss": 0.8083, "step": 1777 }, { "epoch": 0.1, "learning_rate": 0.00019737882539007102, "loss": 0.7268, "step": 1778 }, { "epoch": 0.1, "learning_rate": 0.0001973746449425368, "loss": 0.7314, "step": 1779 }, { "epoch": 0.1, "learning_rate": 0.0001973704612083529, "loss": 0.8047, "step": 1780 }, { "epoch": 0.1, "learning_rate": 0.00019736627418766057, "loss": 0.7668, "step": 1781 }, { "epoch": 0.1, "learning_rate": 0.00019736208388060114, "loss": 0.726, "step": 1782 }, { "epoch": 0.1, "learning_rate": 0.00019735789028731604, "loss": 0.5881, "step": 1783 }, { "epoch": 0.1, "learning_rate": 0.0001973536934079468, "loss": 0.8263, "step": 1784 }, { "epoch": 0.1, "learning_rate": 0.0001973494932426351, "loss": 0.7552, "step": 1785 }, { "epoch": 0.1, "learning_rate": 0.0001973452897915227, "loss": 0.7495, "step": 1786 }, { "epoch": 0.1, "learning_rate": 0.00019734108305475146, "loss": 0.7869, "step": 1787 }, { "epoch": 0.1, "learning_rate": 0.0001973368730324634, "loss": 0.774, "step": 1788 }, { "epoch": 0.1, "learning_rate": 0.0001973326597248006, "loss": 0.7499, "step": 1789 }, { "epoch": 0.1, "learning_rate": 0.00019732844313190525, "loss": 0.7514, "step": 1790 }, { "epoch": 0.1, "learning_rate": 0.00019732422325391972, "loss": 0.775, "step": 1791 }, { "epoch": 0.1, "learning_rate": 0.00019732000009098637, "loss": 0.708, "step": 1792 }, { "epoch": 0.1, "learning_rate": 0.00019731577364324782, "loss": 0.702, "step": 1793 }, { "epoch": 0.1, "learning_rate": 0.00019731154391084668, "loss": 0.7396, "step": 1794 }, { "epoch": 0.1, "learning_rate": 0.0001973073108939257, "loss": 0.7378, "step": 1795 }, { "epoch": 0.1, "learning_rate": 0.00019730307459262783, "loss": 0.8169, "step": 1796 }, { "epoch": 0.1, "learning_rate": 0.00019729883500709596, "loss": 0.8462, "step": 1797 }, { "epoch": 0.1, "learning_rate": 0.00019729459213747326, "loss": 0.7144, "step": 1798 }, { "epoch": 0.1, "learning_rate": 0.00019729034598390288, "loss": 0.6732, "step": 1799 }, { "epoch": 0.1, "learning_rate": 0.0001972860965465282, "loss": 0.6941, "step": 1800 }, { "epoch": 0.1, "learning_rate": 0.00019728184382549263, "loss": 0.6962, "step": 1801 }, { "epoch": 0.1, "learning_rate": 0.00019727758782093967, "loss": 0.7447, "step": 1802 }, { "epoch": 0.1, "learning_rate": 0.000197273328533013, "loss": 0.719, "step": 1803 }, { "epoch": 0.1, "learning_rate": 0.0001972690659618564, "loss": 0.7056, "step": 1804 }, { "epoch": 0.1, "learning_rate": 0.00019726480010761374, "loss": 0.6603, "step": 1805 }, { "epoch": 0.1, "learning_rate": 0.00019726053097042897, "loss": 0.8238, "step": 1806 }, { "epoch": 0.1, "learning_rate": 0.00019725625855044617, "loss": 0.796, "step": 1807 }, { "epoch": 0.1, "learning_rate": 0.00019725198284780963, "loss": 0.675, "step": 1808 }, { "epoch": 0.1, "learning_rate": 0.00019724770386266363, "loss": 0.6866, "step": 1809 }, { "epoch": 0.1, "learning_rate": 0.00019724342159515256, "loss": 0.7377, "step": 1810 }, { "epoch": 0.1, "learning_rate": 0.00019723913604542098, "loss": 0.7913, "step": 1811 }, { "epoch": 0.1, "learning_rate": 0.0001972348472136135, "loss": 0.747, "step": 1812 }, { "epoch": 0.1, "learning_rate": 0.00019723055509987498, "loss": 0.7596, "step": 1813 }, { "epoch": 0.1, "learning_rate": 0.0001972262597043502, "loss": 0.7388, "step": 1814 }, { "epoch": 0.1, "learning_rate": 0.00019722196102718416, "loss": 0.8085, "step": 1815 }, { "epoch": 0.1, "learning_rate": 0.00019721765906852197, "loss": 0.7606, "step": 1816 }, { "epoch": 0.1, "learning_rate": 0.00019721335382850885, "loss": 0.8058, "step": 1817 }, { "epoch": 0.1, "learning_rate": 0.00019720904530729004, "loss": 0.7106, "step": 1818 }, { "epoch": 0.1, "learning_rate": 0.00019720473350501105, "loss": 0.6829, "step": 1819 }, { "epoch": 0.1, "learning_rate": 0.00019720041842181735, "loss": 0.7415, "step": 1820 }, { "epoch": 0.1, "learning_rate": 0.00019719610005785465, "loss": 0.7013, "step": 1821 }, { "epoch": 0.1, "learning_rate": 0.00019719177841326865, "loss": 0.7297, "step": 1822 }, { "epoch": 0.1, "learning_rate": 0.0001971874534882052, "loss": 0.7469, "step": 1823 }, { "epoch": 0.1, "learning_rate": 0.00019718312528281033, "loss": 0.7253, "step": 1824 }, { "epoch": 0.1, "learning_rate": 0.00019717879379723012, "loss": 0.723, "step": 1825 }, { "epoch": 0.1, "learning_rate": 0.00019717445903161073, "loss": 0.7822, "step": 1826 }, { "epoch": 0.1, "learning_rate": 0.00019717012098609854, "loss": 0.697, "step": 1827 }, { "epoch": 0.1, "learning_rate": 0.00019716577966083992, "loss": 0.8229, "step": 1828 }, { "epoch": 0.1, "learning_rate": 0.0001971614350559814, "loss": 0.7237, "step": 1829 }, { "epoch": 0.1, "learning_rate": 0.00019715708717166962, "loss": 0.7613, "step": 1830 }, { "epoch": 0.1, "learning_rate": 0.00019715273600805135, "loss": 0.705, "step": 1831 }, { "epoch": 0.1, "learning_rate": 0.00019714838156527345, "loss": 0.8174, "step": 1832 }, { "epoch": 0.1, "learning_rate": 0.00019714402384348291, "loss": 0.6638, "step": 1833 }, { "epoch": 0.1, "learning_rate": 0.00019713966284282678, "loss": 0.8315, "step": 1834 }, { "epoch": 0.1, "learning_rate": 0.00019713529856345226, "loss": 0.6654, "step": 1835 }, { "epoch": 0.1, "learning_rate": 0.0001971309310055067, "loss": 0.7411, "step": 1836 }, { "epoch": 0.1, "learning_rate": 0.00019712656016913744, "loss": 0.6834, "step": 1837 }, { "epoch": 0.1, "learning_rate": 0.00019712218605449206, "loss": 0.6983, "step": 1838 }, { "epoch": 0.1, "learning_rate": 0.0001971178086617182, "loss": 0.6959, "step": 1839 }, { "epoch": 0.1, "learning_rate": 0.00019711342799096361, "loss": 0.7648, "step": 1840 }, { "epoch": 0.1, "learning_rate": 0.00019710904404237614, "loss": 0.7304, "step": 1841 }, { "epoch": 0.1, "learning_rate": 0.00019710465681610373, "loss": 0.7365, "step": 1842 }, { "epoch": 0.1, "learning_rate": 0.0001971002663122945, "loss": 0.7932, "step": 1843 }, { "epoch": 0.1, "learning_rate": 0.0001970958725310966, "loss": 0.6981, "step": 1844 }, { "epoch": 0.1, "learning_rate": 0.0001970914754726584, "loss": 0.6341, "step": 1845 }, { "epoch": 0.1, "learning_rate": 0.00019708707513712824, "loss": 0.7163, "step": 1846 }, { "epoch": 0.1, "learning_rate": 0.00019708267152465468, "loss": 0.8397, "step": 1847 }, { "epoch": 0.1, "learning_rate": 0.00019707826463538636, "loss": 0.7614, "step": 1848 }, { "epoch": 0.1, "learning_rate": 0.000197073854469472, "loss": 0.7792, "step": 1849 }, { "epoch": 0.1, "learning_rate": 0.00019706944102706049, "loss": 0.7396, "step": 1850 }, { "epoch": 0.1, "learning_rate": 0.00019706502430830077, "loss": 0.7756, "step": 1851 }, { "epoch": 0.11, "learning_rate": 0.00019706060431334187, "loss": 0.7369, "step": 1852 }, { "epoch": 0.11, "learning_rate": 0.00019705618104233307, "loss": 0.7712, "step": 1853 }, { "epoch": 0.11, "learning_rate": 0.00019705175449542358, "loss": 0.7398, "step": 1854 }, { "epoch": 0.11, "learning_rate": 0.00019704732467276288, "loss": 0.6215, "step": 1855 }, { "epoch": 0.11, "learning_rate": 0.00019704289157450045, "loss": 0.7337, "step": 1856 }, { "epoch": 0.11, "learning_rate": 0.0001970384552007859, "loss": 0.7887, "step": 1857 }, { "epoch": 0.11, "learning_rate": 0.00019703401555176906, "loss": 0.7823, "step": 1858 }, { "epoch": 0.11, "learning_rate": 0.00019702957262759965, "loss": 0.7872, "step": 1859 }, { "epoch": 0.11, "learning_rate": 0.0001970251264284277, "loss": 0.6703, "step": 1860 }, { "epoch": 0.11, "learning_rate": 0.00019702067695440332, "loss": 0.7335, "step": 1861 }, { "epoch": 0.11, "learning_rate": 0.0001970162242056766, "loss": 0.7956, "step": 1862 }, { "epoch": 0.11, "learning_rate": 0.00019701176818239793, "loss": 0.6784, "step": 1863 }, { "epoch": 0.11, "learning_rate": 0.0001970073088847176, "loss": 0.7582, "step": 1864 }, { "epoch": 0.11, "learning_rate": 0.00019700284631278623, "loss": 0.6609, "step": 1865 }, { "epoch": 0.11, "learning_rate": 0.0001969983804667544, "loss": 0.8012, "step": 1866 }, { "epoch": 0.11, "learning_rate": 0.00019699391134677278, "loss": 0.7304, "step": 1867 }, { "epoch": 0.11, "learning_rate": 0.00019698943895299234, "loss": 0.7063, "step": 1868 }, { "epoch": 0.11, "learning_rate": 0.00019698496328556396, "loss": 0.6165, "step": 1869 }, { "epoch": 0.11, "learning_rate": 0.0001969804843446387, "loss": 0.7795, "step": 1870 }, { "epoch": 0.11, "learning_rate": 0.00019697600213036776, "loss": 0.6956, "step": 1871 }, { "epoch": 0.11, "learning_rate": 0.0001969715166429024, "loss": 0.7733, "step": 1872 }, { "epoch": 0.11, "learning_rate": 0.00019696702788239408, "loss": 0.7151, "step": 1873 }, { "epoch": 0.11, "learning_rate": 0.0001969625358489942, "loss": 0.6776, "step": 1874 }, { "epoch": 0.11, "learning_rate": 0.00019695804054285448, "loss": 0.6855, "step": 1875 }, { "epoch": 0.11, "learning_rate": 0.00019695354196412664, "loss": 0.7385, "step": 1876 }, { "epoch": 0.11, "learning_rate": 0.00019694904011296246, "loss": 0.7496, "step": 1877 }, { "epoch": 0.11, "learning_rate": 0.0001969445349895139, "loss": 0.7937, "step": 1878 }, { "epoch": 0.11, "learning_rate": 0.00019694002659393305, "loss": 0.7087, "step": 1879 }, { "epoch": 0.11, "learning_rate": 0.00019693551492637207, "loss": 0.734, "step": 1880 }, { "epoch": 0.11, "learning_rate": 0.00019693099998698322, "loss": 0.7472, "step": 1881 }, { "epoch": 0.11, "learning_rate": 0.00019692648177591895, "loss": 0.7133, "step": 1882 }, { "epoch": 0.11, "learning_rate": 0.0001969219602933317, "loss": 0.7268, "step": 1883 }, { "epoch": 0.11, "learning_rate": 0.00019691743553937408, "loss": 0.7336, "step": 1884 }, { "epoch": 0.11, "learning_rate": 0.0001969129075141989, "loss": 0.8367, "step": 1885 }, { "epoch": 0.11, "learning_rate": 0.0001969083762179589, "loss": 0.7116, "step": 1886 }, { "epoch": 0.11, "learning_rate": 0.00019690384165080703, "loss": 0.8297, "step": 1887 }, { "epoch": 0.11, "learning_rate": 0.00019689930381289634, "loss": 0.733, "step": 1888 }, { "epoch": 0.11, "learning_rate": 0.00019689476270438004, "loss": 0.7821, "step": 1889 }, { "epoch": 0.11, "learning_rate": 0.0001968902183254114, "loss": 0.7368, "step": 1890 }, { "epoch": 0.11, "learning_rate": 0.0001968856706761438, "loss": 0.7634, "step": 1891 }, { "epoch": 0.11, "learning_rate": 0.00019688111975673072, "loss": 0.8457, "step": 1892 }, { "epoch": 0.11, "learning_rate": 0.00019687656556732574, "loss": 0.8574, "step": 1893 }, { "epoch": 0.11, "learning_rate": 0.00019687200810808263, "loss": 0.8612, "step": 1894 }, { "epoch": 0.11, "learning_rate": 0.00019686744737915514, "loss": 0.7623, "step": 1895 }, { "epoch": 0.11, "learning_rate": 0.00019686288338069733, "loss": 0.7061, "step": 1896 }, { "epoch": 0.11, "learning_rate": 0.0001968583161128631, "loss": 0.8258, "step": 1897 }, { "epoch": 0.11, "learning_rate": 0.00019685374557580675, "loss": 0.7821, "step": 1898 }, { "epoch": 0.11, "learning_rate": 0.00019684917176968246, "loss": 0.7854, "step": 1899 }, { "epoch": 0.11, "learning_rate": 0.00019684459469464458, "loss": 0.7157, "step": 1900 }, { "epoch": 0.11, "learning_rate": 0.00019684001435084768, "loss": 0.6857, "step": 1901 }, { "epoch": 0.11, "learning_rate": 0.00019683543073844632, "loss": 0.706, "step": 1902 }, { "epoch": 0.11, "learning_rate": 0.00019683084385759523, "loss": 0.784, "step": 1903 }, { "epoch": 0.11, "learning_rate": 0.00019682625370844918, "loss": 0.7962, "step": 1904 }, { "epoch": 0.11, "learning_rate": 0.00019682166029116315, "loss": 0.8108, "step": 1905 }, { "epoch": 0.11, "learning_rate": 0.00019681706360589216, "loss": 0.7955, "step": 1906 }, { "epoch": 0.11, "learning_rate": 0.00019681246365279136, "loss": 0.8288, "step": 1907 }, { "epoch": 0.11, "learning_rate": 0.000196807860432016, "loss": 0.6904, "step": 1908 }, { "epoch": 0.11, "learning_rate": 0.0001968032539437215, "loss": 0.635, "step": 1909 }, { "epoch": 0.11, "learning_rate": 0.00019679864418806325, "loss": 0.7726, "step": 1910 }, { "epoch": 0.11, "learning_rate": 0.00019679403116519692, "loss": 0.7695, "step": 1911 }, { "epoch": 0.11, "learning_rate": 0.0001967894148752782, "loss": 0.7464, "step": 1912 }, { "epoch": 0.11, "learning_rate": 0.00019678479531846287, "loss": 0.6438, "step": 1913 }, { "epoch": 0.11, "learning_rate": 0.0001967801724949069, "loss": 0.7436, "step": 1914 }, { "epoch": 0.11, "learning_rate": 0.00019677554640476624, "loss": 0.7106, "step": 1915 }, { "epoch": 0.11, "learning_rate": 0.00019677091704819715, "loss": 0.6997, "step": 1916 }, { "epoch": 0.11, "learning_rate": 0.00019676628442535576, "loss": 0.7953, "step": 1917 }, { "epoch": 0.11, "learning_rate": 0.00019676164853639852, "loss": 0.7024, "step": 1918 }, { "epoch": 0.11, "learning_rate": 0.00019675700938148188, "loss": 0.7209, "step": 1919 }, { "epoch": 0.11, "learning_rate": 0.0001967523669607624, "loss": 0.7741, "step": 1920 }, { "epoch": 0.11, "learning_rate": 0.0001967477212743968, "loss": 0.8105, "step": 1921 }, { "epoch": 0.11, "learning_rate": 0.0001967430723225419, "loss": 0.7502, "step": 1922 }, { "epoch": 0.11, "learning_rate": 0.00019673842010535454, "loss": 0.8316, "step": 1923 }, { "epoch": 0.11, "learning_rate": 0.00019673376462299184, "loss": 0.7265, "step": 1924 }, { "epoch": 0.11, "learning_rate": 0.00019672910587561088, "loss": 0.7701, "step": 1925 }, { "epoch": 0.11, "learning_rate": 0.0001967244438633689, "loss": 0.8404, "step": 1926 }, { "epoch": 0.11, "learning_rate": 0.0001967197785864233, "loss": 0.7408, "step": 1927 }, { "epoch": 0.11, "learning_rate": 0.0001967151100449315, "loss": 0.7953, "step": 1928 }, { "epoch": 0.11, "learning_rate": 0.0001967104382390511, "loss": 0.6514, "step": 1929 }, { "epoch": 0.11, "learning_rate": 0.00019670576316893973, "loss": 0.6442, "step": 1930 }, { "epoch": 0.11, "learning_rate": 0.0001967010848347553, "loss": 0.6762, "step": 1931 }, { "epoch": 0.11, "learning_rate": 0.0001966964032366556, "loss": 0.7797, "step": 1932 }, { "epoch": 0.11, "learning_rate": 0.00019669171837479873, "loss": 0.7986, "step": 1933 }, { "epoch": 0.11, "learning_rate": 0.00019668703024934274, "loss": 0.7277, "step": 1934 }, { "epoch": 0.11, "learning_rate": 0.00019668233886044597, "loss": 0.733, "step": 1935 }, { "epoch": 0.11, "learning_rate": 0.00019667764420826665, "loss": 0.8131, "step": 1936 }, { "epoch": 0.11, "learning_rate": 0.00019667294629296332, "loss": 0.6934, "step": 1937 }, { "epoch": 0.11, "learning_rate": 0.0001966682451146945, "loss": 0.7114, "step": 1938 }, { "epoch": 0.11, "learning_rate": 0.00019666354067361888, "loss": 0.7513, "step": 1939 }, { "epoch": 0.11, "learning_rate": 0.0001966588329698953, "loss": 0.8072, "step": 1940 }, { "epoch": 0.11, "learning_rate": 0.00019665412200368257, "loss": 0.7355, "step": 1941 }, { "epoch": 0.11, "learning_rate": 0.00019664940777513974, "loss": 0.7177, "step": 1942 }, { "epoch": 0.11, "learning_rate": 0.00019664469028442592, "loss": 0.7424, "step": 1943 }, { "epoch": 0.11, "learning_rate": 0.00019663996953170036, "loss": 0.622, "step": 1944 }, { "epoch": 0.11, "learning_rate": 0.00019663524551712236, "loss": 0.8723, "step": 1945 }, { "epoch": 0.11, "learning_rate": 0.0001966305182408514, "loss": 0.8068, "step": 1946 }, { "epoch": 0.11, "learning_rate": 0.00019662578770304702, "loss": 0.7946, "step": 1947 }, { "epoch": 0.11, "learning_rate": 0.0001966210539038689, "loss": 0.6983, "step": 1948 }, { "epoch": 0.11, "learning_rate": 0.00019661631684347685, "loss": 0.7057, "step": 1949 }, { "epoch": 0.11, "learning_rate": 0.00019661157652203065, "loss": 0.8144, "step": 1950 }, { "epoch": 0.11, "learning_rate": 0.00019660683293969041, "loss": 0.7563, "step": 1951 }, { "epoch": 0.11, "learning_rate": 0.00019660208609661618, "loss": 0.7379, "step": 1952 }, { "epoch": 0.11, "learning_rate": 0.00019659733599296824, "loss": 0.7726, "step": 1953 }, { "epoch": 0.11, "learning_rate": 0.00019659258262890683, "loss": 0.76, "step": 1954 }, { "epoch": 0.11, "learning_rate": 0.00019658782600459244, "loss": 0.7503, "step": 1955 }, { "epoch": 0.11, "learning_rate": 0.00019658306612018565, "loss": 0.8006, "step": 1956 }, { "epoch": 0.11, "learning_rate": 0.00019657830297584703, "loss": 0.7487, "step": 1957 }, { "epoch": 0.11, "learning_rate": 0.00019657353657173742, "loss": 0.9353, "step": 1958 }, { "epoch": 0.11, "learning_rate": 0.00019656876690801766, "loss": 0.78, "step": 1959 }, { "epoch": 0.11, "learning_rate": 0.0001965639939848488, "loss": 0.7147, "step": 1960 }, { "epoch": 0.11, "learning_rate": 0.00019655921780239183, "loss": 0.7028, "step": 1961 }, { "epoch": 0.11, "learning_rate": 0.0001965544383608081, "loss": 0.712, "step": 1962 }, { "epoch": 0.11, "learning_rate": 0.0001965496556602588, "loss": 0.7228, "step": 1963 }, { "epoch": 0.11, "learning_rate": 0.0001965448697009054, "loss": 0.7055, "step": 1964 }, { "epoch": 0.11, "learning_rate": 0.0001965400804829095, "loss": 0.725, "step": 1965 }, { "epoch": 0.11, "learning_rate": 0.00019653528800643266, "loss": 0.7776, "step": 1966 }, { "epoch": 0.11, "learning_rate": 0.00019653049227163667, "loss": 0.7099, "step": 1967 }, { "epoch": 0.11, "learning_rate": 0.00019652569327868342, "loss": 0.6692, "step": 1968 }, { "epoch": 0.11, "learning_rate": 0.00019652089102773488, "loss": 0.7251, "step": 1969 }, { "epoch": 0.11, "learning_rate": 0.00019651608551895313, "loss": 0.7696, "step": 1970 }, { "epoch": 0.11, "learning_rate": 0.00019651127675250035, "loss": 0.7265, "step": 1971 }, { "epoch": 0.11, "learning_rate": 0.00019650646472853887, "loss": 0.8387, "step": 1972 }, { "epoch": 0.11, "learning_rate": 0.00019650164944723115, "loss": 0.7293, "step": 1973 }, { "epoch": 0.11, "learning_rate": 0.00019649683090873965, "loss": 0.8569, "step": 1974 }, { "epoch": 0.11, "learning_rate": 0.00019649200911322706, "loss": 0.6882, "step": 1975 }, { "epoch": 0.11, "learning_rate": 0.00019648718406085607, "loss": 0.7609, "step": 1976 }, { "epoch": 0.11, "learning_rate": 0.00019648235575178957, "loss": 0.7539, "step": 1977 }, { "epoch": 0.11, "learning_rate": 0.00019647752418619054, "loss": 0.7487, "step": 1978 }, { "epoch": 0.11, "learning_rate": 0.00019647268936422206, "loss": 0.7315, "step": 1979 }, { "epoch": 0.11, "learning_rate": 0.00019646785128604727, "loss": 0.803, "step": 1980 }, { "epoch": 0.11, "learning_rate": 0.00019646300995182953, "loss": 0.7332, "step": 1981 }, { "epoch": 0.11, "learning_rate": 0.0001964581653617322, "loss": 0.6686, "step": 1982 }, { "epoch": 0.11, "learning_rate": 0.00019645331751591886, "loss": 0.7327, "step": 1983 }, { "epoch": 0.11, "learning_rate": 0.00019644846641455308, "loss": 0.7206, "step": 1984 }, { "epoch": 0.11, "learning_rate": 0.0001964436120577986, "loss": 0.7903, "step": 1985 }, { "epoch": 0.11, "learning_rate": 0.00019643875444581927, "loss": 0.7356, "step": 1986 }, { "epoch": 0.11, "learning_rate": 0.00019643389357877907, "loss": 0.7805, "step": 1987 }, { "epoch": 0.11, "learning_rate": 0.00019642902945684205, "loss": 0.6787, "step": 1988 }, { "epoch": 0.11, "learning_rate": 0.0001964241620801724, "loss": 0.7373, "step": 1989 }, { "epoch": 0.11, "learning_rate": 0.0001964192914489344, "loss": 0.6903, "step": 1990 }, { "epoch": 0.11, "learning_rate": 0.00019641441756329244, "loss": 0.7343, "step": 1991 }, { "epoch": 0.11, "learning_rate": 0.00019640954042341103, "loss": 0.734, "step": 1992 }, { "epoch": 0.11, "learning_rate": 0.00019640466002945477, "loss": 0.7136, "step": 1993 }, { "epoch": 0.11, "learning_rate": 0.00019639977638158843, "loss": 0.6956, "step": 1994 }, { "epoch": 0.11, "learning_rate": 0.0001963948894799768, "loss": 0.7609, "step": 1995 }, { "epoch": 0.11, "learning_rate": 0.00019638999932478486, "loss": 0.8388, "step": 1996 }, { "epoch": 0.11, "learning_rate": 0.00019638510591617764, "loss": 0.795, "step": 1997 }, { "epoch": 0.11, "learning_rate": 0.00019638020925432032, "loss": 0.6909, "step": 1998 }, { "epoch": 0.11, "learning_rate": 0.00019637530933937819, "loss": 0.6731, "step": 1999 }, { "epoch": 0.11, "learning_rate": 0.00019637040617151658, "loss": 0.824, "step": 2000 }, { "epoch": 0.11, "learning_rate": 0.00019636549975090106, "loss": 0.7306, "step": 2001 }, { "epoch": 0.11, "learning_rate": 0.00019636059007769718, "loss": 0.8318, "step": 2002 }, { "epoch": 0.11, "learning_rate": 0.00019635567715207066, "loss": 0.7025, "step": 2003 }, { "epoch": 0.11, "learning_rate": 0.00019635076097418734, "loss": 0.7232, "step": 2004 }, { "epoch": 0.11, "learning_rate": 0.00019634584154421317, "loss": 0.7649, "step": 2005 }, { "epoch": 0.11, "learning_rate": 0.00019634091886231412, "loss": 0.7641, "step": 2006 }, { "epoch": 0.11, "learning_rate": 0.00019633599292865643, "loss": 0.7302, "step": 2007 }, { "epoch": 0.11, "learning_rate": 0.00019633106374340633, "loss": 0.731, "step": 2008 }, { "epoch": 0.11, "learning_rate": 0.0001963261313067302, "loss": 0.8302, "step": 2009 }, { "epoch": 0.11, "learning_rate": 0.0001963211956187945, "loss": 0.7307, "step": 2010 }, { "epoch": 0.11, "learning_rate": 0.00019631625667976583, "loss": 0.7078, "step": 2011 }, { "epoch": 0.11, "learning_rate": 0.00019631131448981091, "loss": 0.8005, "step": 2012 }, { "epoch": 0.11, "learning_rate": 0.00019630636904909653, "loss": 0.7969, "step": 2013 }, { "epoch": 0.11, "learning_rate": 0.00019630142035778964, "loss": 0.7175, "step": 2014 }, { "epoch": 0.11, "learning_rate": 0.0001962964684160572, "loss": 0.7075, "step": 2015 }, { "epoch": 0.11, "learning_rate": 0.00019629151322406646, "loss": 0.7405, "step": 2016 }, { "epoch": 0.11, "learning_rate": 0.0001962865547819846, "loss": 0.7607, "step": 2017 }, { "epoch": 0.11, "learning_rate": 0.00019628159308997898, "loss": 0.8112, "step": 2018 }, { "epoch": 0.11, "learning_rate": 0.00019627662814821712, "loss": 0.8207, "step": 2019 }, { "epoch": 0.11, "learning_rate": 0.00019627165995686653, "loss": 0.8025, "step": 2020 }, { "epoch": 0.11, "learning_rate": 0.00019626668851609495, "loss": 0.7621, "step": 2021 }, { "epoch": 0.11, "learning_rate": 0.00019626171382607018, "loss": 0.7935, "step": 2022 }, { "epoch": 0.11, "learning_rate": 0.00019625673588696008, "loss": 0.6335, "step": 2023 }, { "epoch": 0.11, "learning_rate": 0.00019625175469893273, "loss": 0.7266, "step": 2024 }, { "epoch": 0.11, "learning_rate": 0.0001962467702621562, "loss": 0.7606, "step": 2025 }, { "epoch": 0.11, "learning_rate": 0.00019624178257679877, "loss": 0.7594, "step": 2026 }, { "epoch": 0.11, "learning_rate": 0.0001962367916430288, "loss": 0.7556, "step": 2027 }, { "epoch": 0.12, "learning_rate": 0.0001962317974610147, "loss": 0.7321, "step": 2028 }, { "epoch": 0.12, "learning_rate": 0.00019622680003092503, "loss": 0.7613, "step": 2029 }, { "epoch": 0.12, "learning_rate": 0.00019622179935292855, "loss": 0.7461, "step": 2030 }, { "epoch": 0.12, "learning_rate": 0.00019621679542719395, "loss": 0.8584, "step": 2031 }, { "epoch": 0.12, "learning_rate": 0.0001962117882538902, "loss": 0.7313, "step": 2032 }, { "epoch": 0.12, "learning_rate": 0.00019620677783318625, "loss": 0.765, "step": 2033 }, { "epoch": 0.12, "learning_rate": 0.00019620176416525125, "loss": 0.6557, "step": 2034 }, { "epoch": 0.12, "learning_rate": 0.0001961967472502544, "loss": 0.7684, "step": 2035 }, { "epoch": 0.12, "learning_rate": 0.00019619172708836504, "loss": 0.6966, "step": 2036 }, { "epoch": 0.12, "learning_rate": 0.00019618670367975262, "loss": 0.7492, "step": 2037 }, { "epoch": 0.12, "learning_rate": 0.0001961816770245867, "loss": 0.8106, "step": 2038 }, { "epoch": 0.12, "learning_rate": 0.0001961766471230369, "loss": 0.648, "step": 2039 }, { "epoch": 0.12, "learning_rate": 0.00019617161397527308, "loss": 0.8148, "step": 2040 }, { "epoch": 0.12, "learning_rate": 0.00019616657758146503, "loss": 0.7987, "step": 2041 }, { "epoch": 0.12, "learning_rate": 0.00019616153794178282, "loss": 0.7045, "step": 2042 }, { "epoch": 0.12, "learning_rate": 0.0001961564950563965, "loss": 0.7608, "step": 2043 }, { "epoch": 0.12, "learning_rate": 0.0001961514489254763, "loss": 0.7648, "step": 2044 }, { "epoch": 0.12, "learning_rate": 0.00019614639954919247, "loss": 0.7505, "step": 2045 }, { "epoch": 0.12, "learning_rate": 0.00019614134692771555, "loss": 0.7134, "step": 2046 }, { "epoch": 0.12, "learning_rate": 0.00019613629106121603, "loss": 0.6989, "step": 2047 }, { "epoch": 0.12, "learning_rate": 0.00019613123194986456, "loss": 0.7698, "step": 2048 }, { "epoch": 0.12, "learning_rate": 0.0001961261695938319, "loss": 0.742, "step": 2049 }, { "epoch": 0.12, "learning_rate": 0.00019612110399328892, "loss": 0.6737, "step": 2050 }, { "epoch": 0.12, "learning_rate": 0.00019611603514840657, "loss": 0.8479, "step": 2051 }, { "epoch": 0.12, "learning_rate": 0.00019611096305935597, "loss": 0.7559, "step": 2052 }, { "epoch": 0.12, "learning_rate": 0.0001961058877263083, "loss": 0.8253, "step": 2053 }, { "epoch": 0.12, "learning_rate": 0.00019610080914943492, "loss": 0.724, "step": 2054 }, { "epoch": 0.12, "learning_rate": 0.00019609572732890714, "loss": 0.6052, "step": 2055 }, { "epoch": 0.12, "learning_rate": 0.00019609064226489657, "loss": 0.7363, "step": 2056 }, { "epoch": 0.12, "learning_rate": 0.00019608555395757483, "loss": 0.668, "step": 2057 }, { "epoch": 0.12, "learning_rate": 0.0001960804624071136, "loss": 0.7958, "step": 2058 }, { "epoch": 0.12, "learning_rate": 0.00019607536761368484, "loss": 0.7118, "step": 2059 }, { "epoch": 0.12, "learning_rate": 0.0001960702695774604, "loss": 0.7098, "step": 2060 }, { "epoch": 0.12, "learning_rate": 0.00019606516829861245, "loss": 0.6896, "step": 2061 }, { "epoch": 0.12, "learning_rate": 0.00019606006377731312, "loss": 0.7188, "step": 2062 }, { "epoch": 0.12, "learning_rate": 0.0001960549560137347, "loss": 0.6685, "step": 2063 }, { "epoch": 0.12, "learning_rate": 0.00019604984500804965, "loss": 0.7822, "step": 2064 }, { "epoch": 0.12, "learning_rate": 0.00019604473076043037, "loss": 0.6913, "step": 2065 }, { "epoch": 0.12, "learning_rate": 0.00019603961327104957, "loss": 0.722, "step": 2066 }, { "epoch": 0.12, "learning_rate": 0.00019603449254007992, "loss": 0.7411, "step": 2067 }, { "epoch": 0.12, "learning_rate": 0.0001960293685676943, "loss": 0.7058, "step": 2068 }, { "epoch": 0.12, "learning_rate": 0.00019602424135406569, "loss": 0.7497, "step": 2069 }, { "epoch": 0.12, "learning_rate": 0.00019601911089936705, "loss": 0.7369, "step": 2070 }, { "epoch": 0.12, "learning_rate": 0.0001960139772037716, "loss": 0.8156, "step": 2071 }, { "epoch": 0.12, "learning_rate": 0.00019600884026745263, "loss": 0.72, "step": 2072 }, { "epoch": 0.12, "learning_rate": 0.0001960037000905835, "loss": 0.7823, "step": 2073 }, { "epoch": 0.12, "learning_rate": 0.00019599855667333773, "loss": 0.7095, "step": 2074 }, { "epoch": 0.12, "learning_rate": 0.00019599341001588888, "loss": 0.7908, "step": 2075 }, { "epoch": 0.12, "learning_rate": 0.0001959882601184107, "loss": 0.7147, "step": 2076 }, { "epoch": 0.12, "learning_rate": 0.00019598310698107702, "loss": 0.6942, "step": 2077 }, { "epoch": 0.12, "learning_rate": 0.00019597795060406173, "loss": 0.7297, "step": 2078 }, { "epoch": 0.12, "learning_rate": 0.0001959727909875389, "loss": 0.8407, "step": 2079 }, { "epoch": 0.12, "learning_rate": 0.0001959676281316827, "loss": 0.8409, "step": 2080 }, { "epoch": 0.12, "learning_rate": 0.00019596246203666735, "loss": 0.7444, "step": 2081 }, { "epoch": 0.12, "learning_rate": 0.00019595729270266722, "loss": 0.7642, "step": 2082 }, { "epoch": 0.12, "learning_rate": 0.00019595212012985682, "loss": 0.708, "step": 2083 }, { "epoch": 0.12, "learning_rate": 0.00019594694431841072, "loss": 0.7047, "step": 2084 }, { "epoch": 0.12, "learning_rate": 0.00019594176526850364, "loss": 0.7392, "step": 2085 }, { "epoch": 0.12, "learning_rate": 0.00019593658298031034, "loss": 0.7656, "step": 2086 }, { "epoch": 0.12, "learning_rate": 0.00019593139745400576, "loss": 0.786, "step": 2087 }, { "epoch": 0.12, "learning_rate": 0.00019592620868976495, "loss": 0.7216, "step": 2088 }, { "epoch": 0.12, "learning_rate": 0.00019592101668776298, "loss": 0.7253, "step": 2089 }, { "epoch": 0.12, "learning_rate": 0.00019591582144817517, "loss": 0.7466, "step": 2090 }, { "epoch": 0.12, "learning_rate": 0.00019591062297117683, "loss": 0.8395, "step": 2091 }, { "epoch": 0.12, "learning_rate": 0.00019590542125694343, "loss": 0.7381, "step": 2092 }, { "epoch": 0.12, "learning_rate": 0.00019590021630565052, "loss": 0.7879, "step": 2093 }, { "epoch": 0.12, "learning_rate": 0.00019589500811747384, "loss": 0.5617, "step": 2094 }, { "epoch": 0.12, "learning_rate": 0.0001958897966925891, "loss": 0.6973, "step": 2095 }, { "epoch": 0.12, "learning_rate": 0.0001958845820311723, "loss": 0.8272, "step": 2096 }, { "epoch": 0.12, "learning_rate": 0.00019587936413339934, "loss": 0.7293, "step": 2097 }, { "epoch": 0.12, "learning_rate": 0.0001958741429994464, "loss": 0.7914, "step": 2098 }, { "epoch": 0.12, "learning_rate": 0.0001958689186294897, "loss": 0.8364, "step": 2099 }, { "epoch": 0.12, "learning_rate": 0.00019586369102370555, "loss": 0.7069, "step": 2100 }, { "epoch": 0.12, "learning_rate": 0.00019585846018227044, "loss": 0.7886, "step": 2101 }, { "epoch": 0.12, "learning_rate": 0.0001958532261053609, "loss": 0.7743, "step": 2102 }, { "epoch": 0.12, "learning_rate": 0.00019584798879315358, "loss": 0.7335, "step": 2103 }, { "epoch": 0.12, "learning_rate": 0.0001958427482458253, "loss": 0.8201, "step": 2104 }, { "epoch": 0.12, "learning_rate": 0.00019583750446355286, "loss": 0.8848, "step": 2105 }, { "epoch": 0.12, "learning_rate": 0.00019583225744651333, "loss": 0.7818, "step": 2106 }, { "epoch": 0.12, "learning_rate": 0.00019582700719488377, "loss": 0.7996, "step": 2107 }, { "epoch": 0.12, "learning_rate": 0.00019582175370884143, "loss": 0.7104, "step": 2108 }, { "epoch": 0.12, "learning_rate": 0.00019581649698856358, "loss": 0.7181, "step": 2109 }, { "epoch": 0.12, "learning_rate": 0.00019581123703422767, "loss": 0.802, "step": 2110 }, { "epoch": 0.12, "learning_rate": 0.00019580597384601122, "loss": 0.7618, "step": 2111 }, { "epoch": 0.12, "learning_rate": 0.00019580070742409192, "loss": 0.7638, "step": 2112 }, { "epoch": 0.12, "learning_rate": 0.0001957954377686475, "loss": 0.8148, "step": 2113 }, { "epoch": 0.12, "learning_rate": 0.0001957901648798558, "loss": 0.7417, "step": 2114 }, { "epoch": 0.12, "learning_rate": 0.00019578488875789482, "loss": 0.8273, "step": 2115 }, { "epoch": 0.12, "learning_rate": 0.00019577960940294264, "loss": 0.7634, "step": 2116 }, { "epoch": 0.12, "learning_rate": 0.00019577432681517747, "loss": 0.7071, "step": 2117 }, { "epoch": 0.12, "learning_rate": 0.00019576904099477758, "loss": 0.7723, "step": 2118 }, { "epoch": 0.12, "learning_rate": 0.00019576375194192138, "loss": 0.7117, "step": 2119 }, { "epoch": 0.12, "learning_rate": 0.0001957584596567874, "loss": 0.7143, "step": 2120 }, { "epoch": 0.12, "learning_rate": 0.00019575316413955428, "loss": 0.7785, "step": 2121 }, { "epoch": 0.12, "learning_rate": 0.00019574786539040077, "loss": 0.7325, "step": 2122 }, { "epoch": 0.12, "learning_rate": 0.00019574256340950568, "loss": 0.7997, "step": 2123 }, { "epoch": 0.12, "learning_rate": 0.00019573725819704797, "loss": 0.677, "step": 2124 }, { "epoch": 0.12, "learning_rate": 0.00019573194975320673, "loss": 0.7646, "step": 2125 }, { "epoch": 0.12, "learning_rate": 0.00019572663807816113, "loss": 0.7917, "step": 2126 }, { "epoch": 0.12, "learning_rate": 0.00019572132317209043, "loss": 0.7737, "step": 2127 }, { "epoch": 0.12, "learning_rate": 0.00019571600503517401, "loss": 0.6935, "step": 2128 }, { "epoch": 0.12, "learning_rate": 0.00019571068366759143, "loss": 0.6762, "step": 2129 }, { "epoch": 0.12, "learning_rate": 0.00019570535906952225, "loss": 0.7558, "step": 2130 }, { "epoch": 0.12, "learning_rate": 0.00019570003124114619, "loss": 0.7567, "step": 2131 }, { "epoch": 0.12, "learning_rate": 0.00019569470018264312, "loss": 0.7155, "step": 2132 }, { "epoch": 0.12, "learning_rate": 0.00019568936589419296, "loss": 0.7948, "step": 2133 }, { "epoch": 0.12, "learning_rate": 0.00019568402837597573, "loss": 0.622, "step": 2134 }, { "epoch": 0.12, "learning_rate": 0.0001956786876281716, "loss": 0.7186, "step": 2135 }, { "epoch": 0.12, "learning_rate": 0.00019567334365096082, "loss": 0.7987, "step": 2136 }, { "epoch": 0.12, "learning_rate": 0.00019566799644452377, "loss": 0.6779, "step": 2137 }, { "epoch": 0.12, "learning_rate": 0.00019566264600904098, "loss": 0.6743, "step": 2138 }, { "epoch": 0.12, "learning_rate": 0.000195657292344693, "loss": 0.709, "step": 2139 }, { "epoch": 0.12, "learning_rate": 0.00019565193545166052, "loss": 0.768, "step": 2140 }, { "epoch": 0.12, "learning_rate": 0.0001956465753301244, "loss": 0.7238, "step": 2141 }, { "epoch": 0.12, "learning_rate": 0.00019564121198026543, "loss": 0.7851, "step": 2142 }, { "epoch": 0.12, "learning_rate": 0.00019563584540226481, "loss": 0.6608, "step": 2143 }, { "epoch": 0.12, "learning_rate": 0.00019563047559630357, "loss": 0.633, "step": 2144 }, { "epoch": 0.12, "learning_rate": 0.00019562510256256296, "loss": 0.8493, "step": 2145 }, { "epoch": 0.12, "learning_rate": 0.00019561972630122436, "loss": 0.7472, "step": 2146 }, { "epoch": 0.12, "learning_rate": 0.00019561434681246925, "loss": 0.6758, "step": 2147 }, { "epoch": 0.12, "learning_rate": 0.00019560896409647917, "loss": 0.6643, "step": 2148 }, { "epoch": 0.12, "learning_rate": 0.00019560357815343577, "loss": 0.7486, "step": 2149 }, { "epoch": 0.12, "learning_rate": 0.00019559818898352092, "loss": 0.7097, "step": 2150 }, { "epoch": 0.12, "learning_rate": 0.00019559279658691646, "loss": 0.709, "step": 2151 }, { "epoch": 0.12, "learning_rate": 0.0001955874009638044, "loss": 0.7126, "step": 2152 }, { "epoch": 0.12, "learning_rate": 0.00019558200211436687, "loss": 0.7538, "step": 2153 }, { "epoch": 0.12, "learning_rate": 0.00019557660003878614, "loss": 0.7038, "step": 2154 }, { "epoch": 0.12, "learning_rate": 0.00019557119473724447, "loss": 0.7142, "step": 2155 }, { "epoch": 0.12, "learning_rate": 0.00019556578620992431, "loss": 0.7195, "step": 2156 }, { "epoch": 0.12, "learning_rate": 0.00019556037445700827, "loss": 0.7069, "step": 2157 }, { "epoch": 0.12, "learning_rate": 0.00019555495947867895, "loss": 0.6716, "step": 2158 }, { "epoch": 0.12, "learning_rate": 0.00019554954127511918, "loss": 0.7347, "step": 2159 }, { "epoch": 0.12, "learning_rate": 0.00019554411984651178, "loss": 0.8377, "step": 2160 }, { "epoch": 0.12, "learning_rate": 0.00019553869519303976, "loss": 0.7971, "step": 2161 }, { "epoch": 0.12, "learning_rate": 0.00019553326731488625, "loss": 0.7386, "step": 2162 }, { "epoch": 0.12, "learning_rate": 0.00019552783621223436, "loss": 0.8712, "step": 2163 }, { "epoch": 0.12, "learning_rate": 0.00019552240188526753, "loss": 0.7041, "step": 2164 }, { "epoch": 0.12, "learning_rate": 0.0001955169643341691, "loss": 0.7795, "step": 2165 }, { "epoch": 0.12, "learning_rate": 0.0001955115235591226, "loss": 0.8248, "step": 2166 }, { "epoch": 0.12, "learning_rate": 0.0001955060795603117, "loss": 0.7509, "step": 2167 }, { "epoch": 0.12, "learning_rate": 0.00019550063233792015, "loss": 0.7439, "step": 2168 }, { "epoch": 0.12, "learning_rate": 0.0001954951818921318, "loss": 0.7893, "step": 2169 }, { "epoch": 0.12, "learning_rate": 0.0001954897282231306, "loss": 0.7899, "step": 2170 }, { "epoch": 0.12, "learning_rate": 0.00019548427133110065, "loss": 0.7529, "step": 2171 }, { "epoch": 0.12, "learning_rate": 0.00019547881121622615, "loss": 0.6508, "step": 2172 }, { "epoch": 0.12, "learning_rate": 0.00019547334787869134, "loss": 0.8067, "step": 2173 }, { "epoch": 0.12, "learning_rate": 0.00019546788131868066, "loss": 0.6748, "step": 2174 }, { "epoch": 0.12, "learning_rate": 0.0001954624115363786, "loss": 0.7974, "step": 2175 }, { "epoch": 0.12, "learning_rate": 0.00019545693853196983, "loss": 0.7141, "step": 2176 }, { "epoch": 0.12, "learning_rate": 0.00019545146230563897, "loss": 0.7028, "step": 2177 }, { "epoch": 0.12, "learning_rate": 0.00019544598285757098, "loss": 0.751, "step": 2178 }, { "epoch": 0.12, "learning_rate": 0.00019544050018795075, "loss": 0.7261, "step": 2179 }, { "epoch": 0.12, "learning_rate": 0.00019543501429696334, "loss": 0.7536, "step": 2180 }, { "epoch": 0.12, "learning_rate": 0.0001954295251847939, "loss": 0.7453, "step": 2181 }, { "epoch": 0.12, "learning_rate": 0.0001954240328516277, "loss": 0.8355, "step": 2182 }, { "epoch": 0.12, "learning_rate": 0.00019541853729765013, "loss": 0.8008, "step": 2183 }, { "epoch": 0.12, "learning_rate": 0.0001954130385230467, "loss": 0.7692, "step": 2184 }, { "epoch": 0.12, "learning_rate": 0.000195407536528003, "loss": 0.8107, "step": 2185 }, { "epoch": 0.12, "learning_rate": 0.0001954020313127047, "loss": 0.6868, "step": 2186 }, { "epoch": 0.12, "learning_rate": 0.00019539652287733764, "loss": 0.6642, "step": 2187 }, { "epoch": 0.12, "learning_rate": 0.00019539101122208777, "loss": 0.7616, "step": 2188 }, { "epoch": 0.12, "learning_rate": 0.00019538549634714108, "loss": 0.712, "step": 2189 }, { "epoch": 0.12, "learning_rate": 0.00019537997825268375, "loss": 0.741, "step": 2190 }, { "epoch": 0.12, "learning_rate": 0.000195374456938902, "loss": 0.7723, "step": 2191 }, { "epoch": 0.12, "learning_rate": 0.0001953689324059822, "loss": 0.7665, "step": 2192 }, { "epoch": 0.12, "learning_rate": 0.0001953634046541108, "loss": 0.7464, "step": 2193 }, { "epoch": 0.12, "learning_rate": 0.00019535787368347442, "loss": 0.5989, "step": 2194 }, { "epoch": 0.12, "learning_rate": 0.0001953523394942597, "loss": 0.6981, "step": 2195 }, { "epoch": 0.12, "learning_rate": 0.0001953468020866535, "loss": 0.7327, "step": 2196 }, { "epoch": 0.12, "learning_rate": 0.0001953412614608426, "loss": 0.783, "step": 2197 }, { "epoch": 0.12, "learning_rate": 0.00019533571761701414, "loss": 0.7341, "step": 2198 }, { "epoch": 0.12, "learning_rate": 0.00019533017055535515, "loss": 0.6936, "step": 2199 }, { "epoch": 0.12, "learning_rate": 0.0001953246202760529, "loss": 0.6952, "step": 2200 }, { "epoch": 0.12, "learning_rate": 0.0001953190667792947, "loss": 0.7198, "step": 2201 }, { "epoch": 0.12, "learning_rate": 0.00019531351006526806, "loss": 0.7173, "step": 2202 }, { "epoch": 0.12, "learning_rate": 0.00019530795013416046, "loss": 0.7936, "step": 2203 }, { "epoch": 0.13, "learning_rate": 0.00019530238698615957, "loss": 0.8158, "step": 2204 }, { "epoch": 0.13, "learning_rate": 0.0001952968206214532, "loss": 0.6799, "step": 2205 }, { "epoch": 0.13, "learning_rate": 0.0001952912510402292, "loss": 0.6157, "step": 2206 }, { "epoch": 0.13, "learning_rate": 0.00019528567824267558, "loss": 0.7685, "step": 2207 }, { "epoch": 0.13, "learning_rate": 0.00019528010222898042, "loss": 0.7866, "step": 2208 }, { "epoch": 0.13, "learning_rate": 0.0001952745229993319, "loss": 0.7064, "step": 2209 }, { "epoch": 0.13, "learning_rate": 0.0001952689405539184, "loss": 0.6861, "step": 2210 }, { "epoch": 0.13, "learning_rate": 0.0001952633548929283, "loss": 0.7273, "step": 2211 }, { "epoch": 0.13, "learning_rate": 0.00019525776601655014, "loss": 0.6834, "step": 2212 }, { "epoch": 0.13, "learning_rate": 0.0001952521739249725, "loss": 0.7677, "step": 2213 }, { "epoch": 0.13, "learning_rate": 0.0001952465786183842, "loss": 0.6793, "step": 2214 }, { "epoch": 0.13, "learning_rate": 0.00019524098009697413, "loss": 0.7649, "step": 2215 }, { "epoch": 0.13, "learning_rate": 0.00019523537836093116, "loss": 0.7016, "step": 2216 }, { "epoch": 0.13, "learning_rate": 0.00019522977341044444, "loss": 0.7529, "step": 2217 }, { "epoch": 0.13, "learning_rate": 0.0001952241652457031, "loss": 0.6497, "step": 2218 }, { "epoch": 0.13, "learning_rate": 0.00019521855386689645, "loss": 0.782, "step": 2219 }, { "epoch": 0.13, "learning_rate": 0.00019521293927421388, "loss": 0.7093, "step": 2220 }, { "epoch": 0.13, "learning_rate": 0.00019520732146784491, "loss": 0.8162, "step": 2221 }, { "epoch": 0.13, "learning_rate": 0.00019520170044797915, "loss": 0.6988, "step": 2222 }, { "epoch": 0.13, "learning_rate": 0.00019519607621480632, "loss": 0.7219, "step": 2223 }, { "epoch": 0.13, "learning_rate": 0.00019519044876851625, "loss": 0.7624, "step": 2224 }, { "epoch": 0.13, "learning_rate": 0.0001951848181092989, "loss": 0.6991, "step": 2225 }, { "epoch": 0.13, "learning_rate": 0.00019517918423734432, "loss": 0.7194, "step": 2226 }, { "epoch": 0.13, "learning_rate": 0.00019517354715284263, "loss": 0.6387, "step": 2227 }, { "epoch": 0.13, "learning_rate": 0.00019516790685598413, "loss": 0.7466, "step": 2228 }, { "epoch": 0.13, "learning_rate": 0.0001951622633469592, "loss": 0.7505, "step": 2229 }, { "epoch": 0.13, "learning_rate": 0.0001951566166259583, "loss": 0.6141, "step": 2230 }, { "epoch": 0.13, "learning_rate": 0.00019515096669317204, "loss": 0.7155, "step": 2231 }, { "epoch": 0.13, "learning_rate": 0.00019514531354879107, "loss": 0.786, "step": 2232 }, { "epoch": 0.13, "learning_rate": 0.00019513965719300628, "loss": 0.8232, "step": 2233 }, { "epoch": 0.13, "learning_rate": 0.00019513399762600855, "loss": 0.7708, "step": 2234 }, { "epoch": 0.13, "learning_rate": 0.00019512833484798886, "loss": 0.7785, "step": 2235 }, { "epoch": 0.13, "learning_rate": 0.00019512266885913845, "loss": 0.691, "step": 2236 }, { "epoch": 0.13, "learning_rate": 0.00019511699965964847, "loss": 0.7848, "step": 2237 }, { "epoch": 0.13, "learning_rate": 0.00019511132724971025, "loss": 0.6796, "step": 2238 }, { "epoch": 0.13, "learning_rate": 0.00019510565162951537, "loss": 0.7354, "step": 2239 }, { "epoch": 0.13, "learning_rate": 0.0001950999727992553, "loss": 0.7605, "step": 2240 }, { "epoch": 0.13, "learning_rate": 0.00019509429075912173, "loss": 0.6409, "step": 2241 }, { "epoch": 0.13, "learning_rate": 0.00019508860550930646, "loss": 0.713, "step": 2242 }, { "epoch": 0.13, "learning_rate": 0.00019508291705000135, "loss": 0.7427, "step": 2243 }, { "epoch": 0.13, "learning_rate": 0.00019507722538139846, "loss": 0.7845, "step": 2244 }, { "epoch": 0.13, "learning_rate": 0.00019507153050368984, "loss": 0.6326, "step": 2245 }, { "epoch": 0.13, "learning_rate": 0.00019506583241706772, "loss": 0.7134, "step": 2246 }, { "epoch": 0.13, "learning_rate": 0.00019506013112172447, "loss": 0.7149, "step": 2247 }, { "epoch": 0.13, "learning_rate": 0.0001950544266178525, "loss": 0.7241, "step": 2248 }, { "epoch": 0.13, "learning_rate": 0.0001950487189056443, "loss": 0.7732, "step": 2249 }, { "epoch": 0.13, "learning_rate": 0.0001950430079852926, "loss": 0.7015, "step": 2250 }, { "epoch": 0.13, "learning_rate": 0.00019503729385699007, "loss": 0.8034, "step": 2251 }, { "epoch": 0.13, "learning_rate": 0.00019503157652092968, "loss": 0.7214, "step": 2252 }, { "epoch": 0.13, "learning_rate": 0.00019502585597730435, "loss": 0.7131, "step": 2253 }, { "epoch": 0.13, "learning_rate": 0.00019502013222630712, "loss": 0.7256, "step": 2254 }, { "epoch": 0.13, "learning_rate": 0.00019501440526813126, "loss": 0.6581, "step": 2255 }, { "epoch": 0.13, "learning_rate": 0.00019500867510297004, "loss": 0.8177, "step": 2256 }, { "epoch": 0.13, "learning_rate": 0.00019500294173101687, "loss": 0.6783, "step": 2257 }, { "epoch": 0.13, "learning_rate": 0.00019499720515246525, "loss": 0.8624, "step": 2258 }, { "epoch": 0.13, "learning_rate": 0.00019499146536750884, "loss": 0.728, "step": 2259 }, { "epoch": 0.13, "learning_rate": 0.00019498572237634134, "loss": 0.7063, "step": 2260 }, { "epoch": 0.13, "learning_rate": 0.00019497997617915657, "loss": 0.6955, "step": 2261 }, { "epoch": 0.13, "learning_rate": 0.00019497422677614853, "loss": 0.7366, "step": 2262 }, { "epoch": 0.13, "learning_rate": 0.00019496847416751125, "loss": 0.7627, "step": 2263 }, { "epoch": 0.13, "learning_rate": 0.0001949627183534389, "loss": 0.683, "step": 2264 }, { "epoch": 0.13, "learning_rate": 0.0001949569593341258, "loss": 0.7599, "step": 2265 }, { "epoch": 0.13, "learning_rate": 0.00019495119710976626, "loss": 0.7191, "step": 2266 }, { "epoch": 0.13, "learning_rate": 0.0001949454316805548, "loss": 0.7231, "step": 2267 }, { "epoch": 0.13, "learning_rate": 0.000194939663046686, "loss": 0.6613, "step": 2268 }, { "epoch": 0.13, "learning_rate": 0.00019493389120835462, "loss": 0.7291, "step": 2269 }, { "epoch": 0.13, "learning_rate": 0.0001949281161657554, "loss": 0.6385, "step": 2270 }, { "epoch": 0.13, "learning_rate": 0.0001949223379190833, "loss": 0.6794, "step": 2271 }, { "epoch": 0.13, "learning_rate": 0.0001949165564685334, "loss": 0.6579, "step": 2272 }, { "epoch": 0.13, "learning_rate": 0.00019491077181430075, "loss": 0.6838, "step": 2273 }, { "epoch": 0.13, "learning_rate": 0.00019490498395658063, "loss": 0.7295, "step": 2274 }, { "epoch": 0.13, "learning_rate": 0.00019489919289556845, "loss": 0.8589, "step": 2275 }, { "epoch": 0.13, "learning_rate": 0.0001948933986314596, "loss": 0.7114, "step": 2276 }, { "epoch": 0.13, "learning_rate": 0.00019488760116444966, "loss": 0.6917, "step": 2277 }, { "epoch": 0.13, "learning_rate": 0.00019488180049473438, "loss": 0.865, "step": 2278 }, { "epoch": 0.13, "learning_rate": 0.00019487599662250943, "loss": 0.7465, "step": 2279 }, { "epoch": 0.13, "learning_rate": 0.0001948701895479708, "loss": 0.6393, "step": 2280 }, { "epoch": 0.13, "learning_rate": 0.0001948643792713145, "loss": 0.7264, "step": 2281 }, { "epoch": 0.13, "learning_rate": 0.00019485856579273657, "loss": 0.7139, "step": 2282 }, { "epoch": 0.13, "learning_rate": 0.00019485274911243327, "loss": 0.8448, "step": 2283 }, { "epoch": 0.13, "learning_rate": 0.00019484692923060095, "loss": 0.7994, "step": 2284 }, { "epoch": 0.13, "learning_rate": 0.00019484110614743602, "loss": 0.7514, "step": 2285 }, { "epoch": 0.13, "learning_rate": 0.000194835279863135, "loss": 0.7678, "step": 2286 }, { "epoch": 0.13, "learning_rate": 0.0001948294503778946, "loss": 0.7966, "step": 2287 }, { "epoch": 0.13, "learning_rate": 0.00019482361769191156, "loss": 0.7247, "step": 2288 }, { "epoch": 0.13, "learning_rate": 0.0001948177818053827, "loss": 0.6763, "step": 2289 }, { "epoch": 0.13, "learning_rate": 0.00019481194271850507, "loss": 0.7136, "step": 2290 }, { "epoch": 0.13, "learning_rate": 0.0001948061004314757, "loss": 0.7048, "step": 2291 }, { "epoch": 0.13, "learning_rate": 0.00019480025494449178, "loss": 0.7672, "step": 2292 }, { "epoch": 0.13, "learning_rate": 0.0001947944062577507, "loss": 0.751, "step": 2293 }, { "epoch": 0.13, "learning_rate": 0.00019478855437144973, "loss": 0.7806, "step": 2294 }, { "epoch": 0.13, "learning_rate": 0.0001947826992857865, "loss": 0.7201, "step": 2295 }, { "epoch": 0.13, "learning_rate": 0.0001947768410009586, "loss": 0.8723, "step": 2296 }, { "epoch": 0.13, "learning_rate": 0.00019477097951716373, "loss": 0.6956, "step": 2297 }, { "epoch": 0.13, "learning_rate": 0.00019476511483459974, "loss": 0.7099, "step": 2298 }, { "epoch": 0.13, "learning_rate": 0.00019475924695346462, "loss": 0.7179, "step": 2299 }, { "epoch": 0.13, "learning_rate": 0.0001947533758739564, "loss": 0.782, "step": 2300 }, { "epoch": 0.13, "learning_rate": 0.00019474750159627325, "loss": 0.7088, "step": 2301 }, { "epoch": 0.13, "learning_rate": 0.0001947416241206134, "loss": 0.8032, "step": 2302 }, { "epoch": 0.13, "learning_rate": 0.00019473574344717532, "loss": 0.6964, "step": 2303 }, { "epoch": 0.13, "learning_rate": 0.0001947298595761574, "loss": 0.6655, "step": 2304 }, { "epoch": 0.13, "learning_rate": 0.00019472397250775832, "loss": 0.7964, "step": 2305 }, { "epoch": 0.13, "learning_rate": 0.00019471808224217672, "loss": 0.7489, "step": 2306 }, { "epoch": 0.13, "learning_rate": 0.00019471218877961142, "loss": 0.7813, "step": 2307 }, { "epoch": 0.13, "learning_rate": 0.00019470629212026136, "loss": 0.8432, "step": 2308 }, { "epoch": 0.13, "learning_rate": 0.0001947003922643256, "loss": 0.7457, "step": 2309 }, { "epoch": 0.13, "learning_rate": 0.00019469448921200324, "loss": 0.7762, "step": 2310 }, { "epoch": 0.13, "learning_rate": 0.0001946885829634935, "loss": 0.7879, "step": 2311 }, { "epoch": 0.13, "learning_rate": 0.00019468267351899576, "loss": 0.6835, "step": 2312 }, { "epoch": 0.13, "learning_rate": 0.00019467676087870948, "loss": 0.682, "step": 2313 }, { "epoch": 0.13, "learning_rate": 0.00019467084504283422, "loss": 0.7206, "step": 2314 }, { "epoch": 0.13, "learning_rate": 0.00019466492601156966, "loss": 0.7392, "step": 2315 }, { "epoch": 0.13, "learning_rate": 0.00019465900378511558, "loss": 0.8134, "step": 2316 }, { "epoch": 0.13, "learning_rate": 0.00019465307836367184, "loss": 0.7129, "step": 2317 }, { "epoch": 0.13, "learning_rate": 0.00019464714974743848, "loss": 0.7224, "step": 2318 }, { "epoch": 0.13, "learning_rate": 0.00019464121793661562, "loss": 0.7483, "step": 2319 }, { "epoch": 0.13, "learning_rate": 0.00019463528293140345, "loss": 0.6771, "step": 2320 }, { "epoch": 0.13, "learning_rate": 0.00019462934473200229, "loss": 0.771, "step": 2321 }, { "epoch": 0.13, "learning_rate": 0.00019462340333861253, "loss": 0.7191, "step": 2322 }, { "epoch": 0.13, "learning_rate": 0.00019461745875143477, "loss": 0.7132, "step": 2323 }, { "epoch": 0.13, "learning_rate": 0.00019461151097066966, "loss": 0.7122, "step": 2324 }, { "epoch": 0.13, "learning_rate": 0.0001946055599965179, "loss": 0.7533, "step": 2325 }, { "epoch": 0.13, "learning_rate": 0.00019459960582918039, "loss": 0.7277, "step": 2326 }, { "epoch": 0.13, "learning_rate": 0.00019459364846885805, "loss": 0.6834, "step": 2327 }, { "epoch": 0.13, "learning_rate": 0.00019458768791575203, "loss": 0.622, "step": 2328 }, { "epoch": 0.13, "learning_rate": 0.00019458172417006347, "loss": 0.7855, "step": 2329 }, { "epoch": 0.13, "learning_rate": 0.0001945757572319937, "loss": 0.7317, "step": 2330 }, { "epoch": 0.13, "learning_rate": 0.00019456978710174404, "loss": 0.7145, "step": 2331 }, { "epoch": 0.13, "learning_rate": 0.00019456381377951608, "loss": 0.6867, "step": 2332 }, { "epoch": 0.13, "learning_rate": 0.00019455783726551139, "loss": 0.7873, "step": 2333 }, { "epoch": 0.13, "learning_rate": 0.0001945518575599317, "loss": 0.8785, "step": 2334 }, { "epoch": 0.13, "learning_rate": 0.00019454587466297883, "loss": 0.7935, "step": 2335 }, { "epoch": 0.13, "learning_rate": 0.00019453988857485476, "loss": 0.7616, "step": 2336 }, { "epoch": 0.13, "learning_rate": 0.00019453389929576152, "loss": 0.8338, "step": 2337 }, { "epoch": 0.13, "learning_rate": 0.00019452790682590124, "loss": 0.7605, "step": 2338 }, { "epoch": 0.13, "learning_rate": 0.00019452191116547618, "loss": 0.6508, "step": 2339 }, { "epoch": 0.13, "learning_rate": 0.00019451591231468875, "loss": 0.6974, "step": 2340 }, { "epoch": 0.13, "learning_rate": 0.0001945099102737414, "loss": 0.7024, "step": 2341 }, { "epoch": 0.13, "learning_rate": 0.0001945039050428367, "loss": 0.7088, "step": 2342 }, { "epoch": 0.13, "learning_rate": 0.00019449789662217737, "loss": 0.7405, "step": 2343 }, { "epoch": 0.13, "learning_rate": 0.00019449188501196622, "loss": 0.7081, "step": 2344 }, { "epoch": 0.13, "learning_rate": 0.00019448587021240611, "loss": 0.8315, "step": 2345 }, { "epoch": 0.13, "learning_rate": 0.0001944798522237001, "loss": 0.7166, "step": 2346 }, { "epoch": 0.13, "learning_rate": 0.00019447383104605125, "loss": 0.7242, "step": 2347 }, { "epoch": 0.13, "learning_rate": 0.00019446780667966288, "loss": 0.7503, "step": 2348 }, { "epoch": 0.13, "learning_rate": 0.0001944617791247383, "loss": 0.8092, "step": 2349 }, { "epoch": 0.13, "learning_rate": 0.00019445574838148087, "loss": 0.7153, "step": 2350 }, { "epoch": 0.13, "learning_rate": 0.00019444971445009427, "loss": 0.822, "step": 2351 }, { "epoch": 0.13, "learning_rate": 0.0001944436773307821, "loss": 0.6793, "step": 2352 }, { "epoch": 0.13, "learning_rate": 0.00019443763702374812, "loss": 0.7001, "step": 2353 }, { "epoch": 0.13, "learning_rate": 0.00019443159352919623, "loss": 0.7535, "step": 2354 }, { "epoch": 0.13, "learning_rate": 0.00019442554684733038, "loss": 0.7505, "step": 2355 }, { "epoch": 0.13, "learning_rate": 0.0001944194969783547, "loss": 0.8014, "step": 2356 }, { "epoch": 0.13, "learning_rate": 0.00019441344392247335, "loss": 0.646, "step": 2357 }, { "epoch": 0.13, "learning_rate": 0.0001944073876798907, "loss": 0.7668, "step": 2358 }, { "epoch": 0.13, "learning_rate": 0.00019440132825081108, "loss": 0.671, "step": 2359 }, { "epoch": 0.13, "learning_rate": 0.00019439526563543909, "loss": 0.8627, "step": 2360 }, { "epoch": 0.13, "learning_rate": 0.0001943891998339793, "loss": 0.7552, "step": 2361 }, { "epoch": 0.13, "learning_rate": 0.0001943831308466365, "loss": 0.8169, "step": 2362 }, { "epoch": 0.13, "learning_rate": 0.00019437705867361547, "loss": 0.7417, "step": 2363 }, { "epoch": 0.13, "learning_rate": 0.00019437098331512122, "loss": 0.7017, "step": 2364 }, { "epoch": 0.13, "learning_rate": 0.00019436490477135878, "loss": 0.7278, "step": 2365 }, { "epoch": 0.13, "learning_rate": 0.00019435882304253332, "loss": 0.7265, "step": 2366 }, { "epoch": 0.13, "learning_rate": 0.00019435273812885013, "loss": 0.7462, "step": 2367 }, { "epoch": 0.13, "learning_rate": 0.00019434665003051458, "loss": 0.7602, "step": 2368 }, { "epoch": 0.13, "learning_rate": 0.00019434055874773215, "loss": 0.7716, "step": 2369 }, { "epoch": 0.13, "learning_rate": 0.00019433446428070842, "loss": 0.7944, "step": 2370 }, { "epoch": 0.13, "learning_rate": 0.00019432836662964915, "loss": 0.7289, "step": 2371 }, { "epoch": 0.13, "learning_rate": 0.0001943222657947601, "loss": 0.8101, "step": 2372 }, { "epoch": 0.13, "learning_rate": 0.00019431616177624723, "loss": 0.7199, "step": 2373 }, { "epoch": 0.13, "learning_rate": 0.00019431005457431653, "loss": 0.7655, "step": 2374 }, { "epoch": 0.13, "learning_rate": 0.00019430394418917417, "loss": 0.8577, "step": 2375 }, { "epoch": 0.13, "learning_rate": 0.00019429783062102635, "loss": 0.7098, "step": 2376 }, { "epoch": 0.13, "learning_rate": 0.00019429171387007948, "loss": 0.7172, "step": 2377 }, { "epoch": 0.13, "learning_rate": 0.00019428559393653997, "loss": 0.7637, "step": 2378 }, { "epoch": 0.13, "learning_rate": 0.00019427947082061432, "loss": 0.7266, "step": 2379 }, { "epoch": 0.14, "learning_rate": 0.00019427334452250933, "loss": 0.7444, "step": 2380 }, { "epoch": 0.14, "learning_rate": 0.00019426721504243175, "loss": 0.8021, "step": 2381 }, { "epoch": 0.14, "learning_rate": 0.0001942610823805884, "loss": 0.8099, "step": 2382 }, { "epoch": 0.14, "learning_rate": 0.0001942549465371863, "loss": 0.7407, "step": 2383 }, { "epoch": 0.14, "learning_rate": 0.0001942488075124326, "loss": 0.6813, "step": 2384 }, { "epoch": 0.14, "learning_rate": 0.00019424266530653443, "loss": 0.5766, "step": 2385 }, { "epoch": 0.14, "learning_rate": 0.00019423651991969915, "loss": 0.7293, "step": 2386 }, { "epoch": 0.14, "learning_rate": 0.0001942303713521342, "loss": 0.5815, "step": 2387 }, { "epoch": 0.14, "learning_rate": 0.00019422421960404708, "loss": 0.7704, "step": 2388 }, { "epoch": 0.14, "learning_rate": 0.00019421806467564544, "loss": 0.7497, "step": 2389 }, { "epoch": 0.14, "learning_rate": 0.00019421190656713702, "loss": 0.7995, "step": 2390 }, { "epoch": 0.14, "learning_rate": 0.00019420574527872968, "loss": 0.7333, "step": 2391 }, { "epoch": 0.14, "learning_rate": 0.00019419958081063138, "loss": 0.7918, "step": 2392 }, { "epoch": 0.14, "learning_rate": 0.00019419341316305016, "loss": 0.7967, "step": 2393 }, { "epoch": 0.14, "learning_rate": 0.0001941872423361942, "loss": 0.7433, "step": 2394 }, { "epoch": 0.14, "learning_rate": 0.00019418106833027184, "loss": 0.6255, "step": 2395 }, { "epoch": 0.14, "learning_rate": 0.00019417489114549139, "loss": 0.7492, "step": 2396 }, { "epoch": 0.14, "learning_rate": 0.00019416871078206137, "loss": 0.72, "step": 2397 }, { "epoch": 0.14, "learning_rate": 0.00019416252724019045, "loss": 0.7216, "step": 2398 }, { "epoch": 0.14, "learning_rate": 0.00019415634052008724, "loss": 0.7095, "step": 2399 }, { "epoch": 0.14, "learning_rate": 0.0001941501506219606, "loss": 0.681, "step": 2400 }, { "epoch": 0.14, "learning_rate": 0.00019414395754601947, "loss": 0.7222, "step": 2401 }, { "epoch": 0.14, "learning_rate": 0.0001941377612924729, "loss": 0.6869, "step": 2402 }, { "epoch": 0.14, "learning_rate": 0.00019413156186152996, "loss": 0.7477, "step": 2403 }, { "epoch": 0.14, "learning_rate": 0.00019412535925339997, "loss": 0.7139, "step": 2404 }, { "epoch": 0.14, "learning_rate": 0.00019411915346829225, "loss": 0.7308, "step": 2405 }, { "epoch": 0.14, "learning_rate": 0.00019411294450641622, "loss": 0.666, "step": 2406 }, { "epoch": 0.14, "learning_rate": 0.00019410673236798154, "loss": 0.8389, "step": 2407 }, { "epoch": 0.14, "learning_rate": 0.0001941005170531978, "loss": 0.7617, "step": 2408 }, { "epoch": 0.14, "learning_rate": 0.00019409429856227485, "loss": 0.6792, "step": 2409 }, { "epoch": 0.14, "learning_rate": 0.00019408807689542257, "loss": 0.779, "step": 2410 }, { "epoch": 0.14, "learning_rate": 0.0001940818520528509, "loss": 0.7687, "step": 2411 }, { "epoch": 0.14, "learning_rate": 0.00019407562403477, "loss": 0.7724, "step": 2412 }, { "epoch": 0.14, "learning_rate": 0.00019406939284139006, "loss": 0.6593, "step": 2413 }, { "epoch": 0.14, "learning_rate": 0.00019406315847292143, "loss": 0.7864, "step": 2414 }, { "epoch": 0.14, "learning_rate": 0.00019405692092957448, "loss": 0.8328, "step": 2415 }, { "epoch": 0.14, "learning_rate": 0.00019405068021155982, "loss": 0.8277, "step": 2416 }, { "epoch": 0.14, "learning_rate": 0.000194044436319088, "loss": 0.6138, "step": 2417 }, { "epoch": 0.14, "learning_rate": 0.0001940381892523698, "loss": 0.6734, "step": 2418 }, { "epoch": 0.14, "learning_rate": 0.00019403193901161613, "loss": 0.8014, "step": 2419 }, { "epoch": 0.14, "learning_rate": 0.00019402568559703792, "loss": 0.6545, "step": 2420 }, { "epoch": 0.14, "learning_rate": 0.0001940194290088462, "loss": 0.8314, "step": 2421 }, { "epoch": 0.14, "learning_rate": 0.00019401316924725216, "loss": 0.7228, "step": 2422 }, { "epoch": 0.14, "learning_rate": 0.00019400690631246714, "loss": 0.7664, "step": 2423 }, { "epoch": 0.14, "learning_rate": 0.00019400064020470248, "loss": 0.6205, "step": 2424 }, { "epoch": 0.14, "learning_rate": 0.00019399437092416967, "loss": 0.6669, "step": 2425 }, { "epoch": 0.14, "learning_rate": 0.00019398809847108036, "loss": 0.7316, "step": 2426 }, { "epoch": 0.14, "learning_rate": 0.00019398182284564622, "loss": 0.7925, "step": 2427 }, { "epoch": 0.14, "learning_rate": 0.00019397554404807906, "loss": 0.6572, "step": 2428 }, { "epoch": 0.14, "learning_rate": 0.00019396926207859084, "loss": 0.7737, "step": 2429 }, { "epoch": 0.14, "learning_rate": 0.00019396297693739358, "loss": 0.7049, "step": 2430 }, { "epoch": 0.14, "learning_rate": 0.00019395668862469946, "loss": 0.7688, "step": 2431 }, { "epoch": 0.14, "learning_rate": 0.00019395039714072064, "loss": 0.7132, "step": 2432 }, { "epoch": 0.14, "learning_rate": 0.00019394410248566954, "loss": 0.6475, "step": 2433 }, { "epoch": 0.14, "learning_rate": 0.00019393780465975862, "loss": 0.7007, "step": 2434 }, { "epoch": 0.14, "learning_rate": 0.00019393150366320042, "loss": 0.7737, "step": 2435 }, { "epoch": 0.14, "learning_rate": 0.0001939251994962076, "loss": 0.8007, "step": 2436 }, { "epoch": 0.14, "learning_rate": 0.00019391889215899299, "loss": 0.7471, "step": 2437 }, { "epoch": 0.14, "learning_rate": 0.00019391258165176947, "loss": 0.7753, "step": 2438 }, { "epoch": 0.14, "learning_rate": 0.00019390626797475, "loss": 0.7162, "step": 2439 }, { "epoch": 0.14, "learning_rate": 0.00019389995112814774, "loss": 0.7521, "step": 2440 }, { "epoch": 0.14, "learning_rate": 0.00019389363111217586, "loss": 0.6954, "step": 2441 }, { "epoch": 0.14, "learning_rate": 0.00019388730792704767, "loss": 0.7449, "step": 2442 }, { "epoch": 0.14, "learning_rate": 0.0001938809815729766, "loss": 0.736, "step": 2443 }, { "epoch": 0.14, "learning_rate": 0.00019387465205017624, "loss": 0.7511, "step": 2444 }, { "epoch": 0.14, "learning_rate": 0.0001938683193588601, "loss": 0.6907, "step": 2445 }, { "epoch": 0.14, "learning_rate": 0.00019386198349924207, "loss": 0.7773, "step": 2446 }, { "epoch": 0.14, "learning_rate": 0.0001938556444715359, "loss": 0.8102, "step": 2447 }, { "epoch": 0.14, "learning_rate": 0.0001938493022759556, "loss": 0.7101, "step": 2448 }, { "epoch": 0.14, "learning_rate": 0.00019384295691271522, "loss": 0.7921, "step": 2449 }, { "epoch": 0.14, "learning_rate": 0.00019383660838202894, "loss": 0.7157, "step": 2450 }, { "epoch": 0.14, "learning_rate": 0.00019383025668411098, "loss": 0.7766, "step": 2451 }, { "epoch": 0.14, "learning_rate": 0.00019382390181917582, "loss": 0.6539, "step": 2452 }, { "epoch": 0.14, "learning_rate": 0.00019381754378743792, "loss": 0.7154, "step": 2453 }, { "epoch": 0.14, "learning_rate": 0.00019381118258911186, "loss": 0.7399, "step": 2454 }, { "epoch": 0.14, "learning_rate": 0.00019380481822441235, "loss": 0.636, "step": 2455 }, { "epoch": 0.14, "learning_rate": 0.00019379845069355422, "loss": 0.7676, "step": 2456 }, { "epoch": 0.14, "learning_rate": 0.0001937920799967524, "loss": 0.7339, "step": 2457 }, { "epoch": 0.14, "learning_rate": 0.00019378570613422187, "loss": 0.7255, "step": 2458 }, { "epoch": 0.14, "learning_rate": 0.00019377932910617782, "loss": 0.7656, "step": 2459 }, { "epoch": 0.14, "learning_rate": 0.00019377294891283545, "loss": 0.7381, "step": 2460 }, { "epoch": 0.14, "learning_rate": 0.00019376656555441014, "loss": 0.7094, "step": 2461 }, { "epoch": 0.14, "learning_rate": 0.0001937601790311173, "loss": 0.7534, "step": 2462 }, { "epoch": 0.14, "learning_rate": 0.00019375378934317257, "loss": 0.7937, "step": 2463 }, { "epoch": 0.14, "learning_rate": 0.00019374739649079153, "loss": 0.7406, "step": 2464 }, { "epoch": 0.14, "learning_rate": 0.00019374100047419, "loss": 0.6979, "step": 2465 }, { "epoch": 0.14, "learning_rate": 0.0001937346012935839, "loss": 0.6868, "step": 2466 }, { "epoch": 0.14, "learning_rate": 0.00019372819894918915, "loss": 0.6512, "step": 2467 }, { "epoch": 0.14, "learning_rate": 0.00019372179344122187, "loss": 0.8813, "step": 2468 }, { "epoch": 0.14, "learning_rate": 0.0001937153847698983, "loss": 0.6608, "step": 2469 }, { "epoch": 0.14, "learning_rate": 0.00019370897293543467, "loss": 0.8189, "step": 2470 }, { "epoch": 0.14, "learning_rate": 0.00019370255793804746, "loss": 0.7743, "step": 2471 }, { "epoch": 0.14, "learning_rate": 0.00019369613977795317, "loss": 0.7316, "step": 2472 }, { "epoch": 0.14, "learning_rate": 0.00019368971845536845, "loss": 0.7888, "step": 2473 }, { "epoch": 0.14, "learning_rate": 0.00019368329397051004, "loss": 0.797, "step": 2474 }, { "epoch": 0.14, "learning_rate": 0.00019367686632359475, "loss": 0.7903, "step": 2475 }, { "epoch": 0.14, "learning_rate": 0.00019367043551483953, "loss": 0.8271, "step": 2476 }, { "epoch": 0.14, "learning_rate": 0.00019366400154446144, "loss": 0.7213, "step": 2477 }, { "epoch": 0.14, "learning_rate": 0.0001936575644126777, "loss": 0.7686, "step": 2478 }, { "epoch": 0.14, "learning_rate": 0.0001936511241197055, "loss": 0.8095, "step": 2479 }, { "epoch": 0.14, "learning_rate": 0.00019364468066576226, "loss": 0.7468, "step": 2480 }, { "epoch": 0.14, "learning_rate": 0.00019363823405106544, "loss": 0.7591, "step": 2481 }, { "epoch": 0.14, "learning_rate": 0.00019363178427583266, "loss": 0.7823, "step": 2482 }, { "epoch": 0.14, "learning_rate": 0.0001936253313402816, "loss": 0.7329, "step": 2483 }, { "epoch": 0.14, "learning_rate": 0.00019361887524463008, "loss": 0.7043, "step": 2484 }, { "epoch": 0.14, "learning_rate": 0.000193612415989096, "loss": 0.6124, "step": 2485 }, { "epoch": 0.14, "learning_rate": 0.00019360595357389735, "loss": 0.6646, "step": 2486 }, { "epoch": 0.14, "learning_rate": 0.00019359948799925228, "loss": 0.6955, "step": 2487 }, { "epoch": 0.14, "learning_rate": 0.00019359301926537902, "loss": 0.7948, "step": 2488 }, { "epoch": 0.14, "learning_rate": 0.00019358654737249592, "loss": 0.8325, "step": 2489 }, { "epoch": 0.14, "learning_rate": 0.00019358007232082137, "loss": 0.7498, "step": 2490 }, { "epoch": 0.14, "learning_rate": 0.000193573594110574, "loss": 0.7124, "step": 2491 }, { "epoch": 0.14, "learning_rate": 0.00019356711274197237, "loss": 0.7373, "step": 2492 }, { "epoch": 0.14, "learning_rate": 0.00019356062821523532, "loss": 0.7907, "step": 2493 }, { "epoch": 0.14, "learning_rate": 0.00019355414053058172, "loss": 0.7388, "step": 2494 }, { "epoch": 0.14, "learning_rate": 0.00019354764968823048, "loss": 0.7037, "step": 2495 }, { "epoch": 0.14, "learning_rate": 0.00019354115568840077, "loss": 0.8059, "step": 2496 }, { "epoch": 0.14, "learning_rate": 0.0001935346585313117, "loss": 0.7189, "step": 2497 }, { "epoch": 0.14, "learning_rate": 0.00019352815821718262, "loss": 0.698, "step": 2498 }, { "epoch": 0.14, "learning_rate": 0.0001935216547462329, "loss": 0.8029, "step": 2499 }, { "epoch": 0.14, "learning_rate": 0.00019351514811868207, "loss": 0.8277, "step": 2500 }, { "epoch": 0.14, "learning_rate": 0.00019350863833474975, "loss": 0.749, "step": 2501 }, { "epoch": 0.14, "learning_rate": 0.00019350212539465564, "loss": 0.7758, "step": 2502 }, { "epoch": 0.14, "learning_rate": 0.00019349560929861958, "loss": 0.674, "step": 2503 }, { "epoch": 0.14, "learning_rate": 0.00019348909004686152, "loss": 0.6952, "step": 2504 }, { "epoch": 0.14, "learning_rate": 0.00019348256763960145, "loss": 0.7, "step": 2505 }, { "epoch": 0.14, "learning_rate": 0.00019347604207705957, "loss": 0.7587, "step": 2506 }, { "epoch": 0.14, "learning_rate": 0.00019346951335945614, "loss": 0.7566, "step": 2507 }, { "epoch": 0.14, "learning_rate": 0.00019346298148701147, "loss": 0.7494, "step": 2508 }, { "epoch": 0.14, "learning_rate": 0.0001934564464599461, "loss": 0.7601, "step": 2509 }, { "epoch": 0.14, "learning_rate": 0.00019344990827848055, "loss": 0.6429, "step": 2510 }, { "epoch": 0.14, "learning_rate": 0.0001934433669428355, "loss": 0.7822, "step": 2511 }, { "epoch": 0.14, "learning_rate": 0.00019343682245323175, "loss": 0.6129, "step": 2512 }, { "epoch": 0.14, "learning_rate": 0.0001934302748098902, "loss": 0.7748, "step": 2513 }, { "epoch": 0.14, "learning_rate": 0.00019342372401303184, "loss": 0.7383, "step": 2514 }, { "epoch": 0.14, "learning_rate": 0.0001934171700628778, "loss": 0.7152, "step": 2515 }, { "epoch": 0.14, "learning_rate": 0.00019341061295964927, "loss": 0.7144, "step": 2516 }, { "epoch": 0.14, "learning_rate": 0.00019340405270356755, "loss": 0.6, "step": 2517 }, { "epoch": 0.14, "learning_rate": 0.0001933974892948541, "loss": 0.7578, "step": 2518 }, { "epoch": 0.14, "learning_rate": 0.00019339092273373048, "loss": 0.7097, "step": 2519 }, { "epoch": 0.14, "learning_rate": 0.00019338435302041824, "loss": 0.6787, "step": 2520 }, { "epoch": 0.14, "learning_rate": 0.00019337778015513922, "loss": 0.8308, "step": 2521 }, { "epoch": 0.14, "learning_rate": 0.00019337120413811517, "loss": 0.7522, "step": 2522 }, { "epoch": 0.14, "learning_rate": 0.00019336462496956816, "loss": 0.8322, "step": 2523 }, { "epoch": 0.14, "learning_rate": 0.00019335804264972018, "loss": 0.7077, "step": 2524 }, { "epoch": 0.14, "learning_rate": 0.0001933514571787934, "loss": 0.7189, "step": 2525 }, { "epoch": 0.14, "learning_rate": 0.00019334486855701014, "loss": 0.6941, "step": 2526 }, { "epoch": 0.14, "learning_rate": 0.0001933382767845928, "loss": 0.7451, "step": 2527 }, { "epoch": 0.14, "learning_rate": 0.00019333168186176378, "loss": 0.7577, "step": 2528 }, { "epoch": 0.14, "learning_rate": 0.0001933250837887457, "loss": 0.698, "step": 2529 }, { "epoch": 0.14, "learning_rate": 0.00019331848256576135, "loss": 0.7215, "step": 2530 }, { "epoch": 0.14, "learning_rate": 0.00019331187819303345, "loss": 0.6939, "step": 2531 }, { "epoch": 0.14, "learning_rate": 0.00019330527067078492, "loss": 0.7512, "step": 2532 }, { "epoch": 0.14, "learning_rate": 0.00019329865999923883, "loss": 0.7089, "step": 2533 }, { "epoch": 0.14, "learning_rate": 0.0001932920461786183, "loss": 0.7725, "step": 2534 }, { "epoch": 0.14, "learning_rate": 0.00019328542920914649, "loss": 0.7662, "step": 2535 }, { "epoch": 0.14, "learning_rate": 0.00019327880909104683, "loss": 0.7902, "step": 2536 }, { "epoch": 0.14, "learning_rate": 0.0001932721858245427, "loss": 0.7606, "step": 2537 }, { "epoch": 0.14, "learning_rate": 0.0001932655594098577, "loss": 0.7828, "step": 2538 }, { "epoch": 0.14, "learning_rate": 0.00019325892984721553, "loss": 0.7213, "step": 2539 }, { "epoch": 0.14, "learning_rate": 0.00019325229713683983, "loss": 0.6846, "step": 2540 }, { "epoch": 0.14, "learning_rate": 0.00019324566127895454, "loss": 0.775, "step": 2541 }, { "epoch": 0.14, "learning_rate": 0.00019323902227378365, "loss": 0.7156, "step": 2542 }, { "epoch": 0.14, "learning_rate": 0.00019323238012155123, "loss": 0.6907, "step": 2543 }, { "epoch": 0.14, "learning_rate": 0.00019322573482248148, "loss": 0.724, "step": 2544 }, { "epoch": 0.14, "learning_rate": 0.00019321908637679865, "loss": 0.8372, "step": 2545 }, { "epoch": 0.14, "learning_rate": 0.00019321243478472722, "loss": 0.7859, "step": 2546 }, { "epoch": 0.14, "learning_rate": 0.00019320578004649162, "loss": 0.7703, "step": 2547 }, { "epoch": 0.14, "learning_rate": 0.00019319912216231655, "loss": 0.8245, "step": 2548 }, { "epoch": 0.14, "learning_rate": 0.00019319246113242664, "loss": 0.7788, "step": 2549 }, { "epoch": 0.14, "learning_rate": 0.00019318579695704678, "loss": 0.7247, "step": 2550 }, { "epoch": 0.14, "learning_rate": 0.00019317912963640188, "loss": 0.7032, "step": 2551 }, { "epoch": 0.14, "learning_rate": 0.000193172459170717, "loss": 0.7284, "step": 2552 }, { "epoch": 0.14, "learning_rate": 0.00019316578556021724, "loss": 0.701, "step": 2553 }, { "epoch": 0.14, "learning_rate": 0.0001931591088051279, "loss": 0.7008, "step": 2554 }, { "epoch": 0.14, "learning_rate": 0.0001931524289056743, "loss": 0.7575, "step": 2555 }, { "epoch": 0.14, "learning_rate": 0.00019314574586208196, "loss": 0.7028, "step": 2556 }, { "epoch": 0.15, "learning_rate": 0.0001931390596745764, "loss": 0.7021, "step": 2557 }, { "epoch": 0.15, "learning_rate": 0.0001931323703433833, "loss": 0.6881, "step": 2558 }, { "epoch": 0.15, "learning_rate": 0.00019312567786872843, "loss": 0.6939, "step": 2559 }, { "epoch": 0.15, "learning_rate": 0.00019311898225083773, "loss": 0.7786, "step": 2560 }, { "epoch": 0.15, "learning_rate": 0.00019311228348993718, "loss": 0.807, "step": 2561 }, { "epoch": 0.15, "learning_rate": 0.00019310558158625285, "loss": 0.771, "step": 2562 }, { "epoch": 0.15, "learning_rate": 0.00019309887654001096, "loss": 0.7325, "step": 2563 }, { "epoch": 0.15, "learning_rate": 0.0001930921683514378, "loss": 0.7438, "step": 2564 }, { "epoch": 0.15, "learning_rate": 0.00019308545702075987, "loss": 0.7347, "step": 2565 }, { "epoch": 0.15, "learning_rate": 0.0001930787425482036, "loss": 0.7355, "step": 2566 }, { "epoch": 0.15, "learning_rate": 0.00019307202493399567, "loss": 0.7455, "step": 2567 }, { "epoch": 0.15, "learning_rate": 0.0001930653041783628, "loss": 0.6641, "step": 2568 }, { "epoch": 0.15, "learning_rate": 0.00019305858028153186, "loss": 0.7694, "step": 2569 }, { "epoch": 0.15, "learning_rate": 0.00019305185324372976, "loss": 0.771, "step": 2570 }, { "epoch": 0.15, "learning_rate": 0.00019304512306518362, "loss": 0.762, "step": 2571 }, { "epoch": 0.15, "learning_rate": 0.0001930383897461205, "loss": 0.6972, "step": 2572 }, { "epoch": 0.15, "learning_rate": 0.00019303165328676776, "loss": 0.7614, "step": 2573 }, { "epoch": 0.15, "learning_rate": 0.00019302491368735272, "loss": 0.755, "step": 2574 }, { "epoch": 0.15, "learning_rate": 0.00019301817094810288, "loss": 0.7487, "step": 2575 }, { "epoch": 0.15, "learning_rate": 0.00019301142506924582, "loss": 0.7309, "step": 2576 }, { "epoch": 0.15, "learning_rate": 0.00019300467605100924, "loss": 0.6843, "step": 2577 }, { "epoch": 0.15, "learning_rate": 0.0001929979238936209, "loss": 0.7193, "step": 2578 }, { "epoch": 0.15, "learning_rate": 0.0001929911685973088, "loss": 0.7459, "step": 2579 }, { "epoch": 0.15, "learning_rate": 0.00019298441016230082, "loss": 0.7196, "step": 2580 }, { "epoch": 0.15, "learning_rate": 0.00019297764858882514, "loss": 0.6704, "step": 2581 }, { "epoch": 0.15, "learning_rate": 0.00019297088387711, "loss": 0.6634, "step": 2582 }, { "epoch": 0.15, "learning_rate": 0.00019296411602738372, "loss": 0.7445, "step": 2583 }, { "epoch": 0.15, "learning_rate": 0.00019295734503987469, "loss": 0.7534, "step": 2584 }, { "epoch": 0.15, "learning_rate": 0.00019295057091481147, "loss": 0.7443, "step": 2585 }, { "epoch": 0.15, "learning_rate": 0.00019294379365242272, "loss": 0.7361, "step": 2586 }, { "epoch": 0.15, "learning_rate": 0.00019293701325293717, "loss": 0.7184, "step": 2587 }, { "epoch": 0.15, "learning_rate": 0.0001929302297165837, "loss": 0.8537, "step": 2588 }, { "epoch": 0.15, "learning_rate": 0.00019292344304359124, "loss": 0.7699, "step": 2589 }, { "epoch": 0.15, "learning_rate": 0.0001929166532341889, "loss": 0.8026, "step": 2590 }, { "epoch": 0.15, "learning_rate": 0.00019290986028860582, "loss": 0.7417, "step": 2591 }, { "epoch": 0.15, "learning_rate": 0.0001929030642070713, "loss": 0.6212, "step": 2592 }, { "epoch": 0.15, "learning_rate": 0.0001928962649898147, "loss": 0.7428, "step": 2593 }, { "epoch": 0.15, "learning_rate": 0.00019288946263706553, "loss": 0.7421, "step": 2594 }, { "epoch": 0.15, "learning_rate": 0.0001928826571490534, "loss": 0.8023, "step": 2595 }, { "epoch": 0.15, "learning_rate": 0.000192875848526008, "loss": 0.7699, "step": 2596 }, { "epoch": 0.15, "learning_rate": 0.0001928690367681591, "loss": 0.6838, "step": 2597 }, { "epoch": 0.15, "learning_rate": 0.0001928622218757367, "loss": 0.6847, "step": 2598 }, { "epoch": 0.15, "learning_rate": 0.00019285540384897073, "loss": 0.6513, "step": 2599 }, { "epoch": 0.15, "learning_rate": 0.00019284858268809137, "loss": 0.7602, "step": 2600 }, { "epoch": 0.15, "learning_rate": 0.00019284175839332884, "loss": 0.7776, "step": 2601 }, { "epoch": 0.15, "learning_rate": 0.00019283493096491347, "loss": 0.7165, "step": 2602 }, { "epoch": 0.15, "learning_rate": 0.00019282810040307572, "loss": 0.6794, "step": 2603 }, { "epoch": 0.15, "learning_rate": 0.00019282126670804614, "loss": 0.7439, "step": 2604 }, { "epoch": 0.15, "learning_rate": 0.00019281442988005538, "loss": 0.8003, "step": 2605 }, { "epoch": 0.15, "learning_rate": 0.0001928075899193342, "loss": 0.71, "step": 2606 }, { "epoch": 0.15, "learning_rate": 0.00019280074682611345, "loss": 0.768, "step": 2607 }, { "epoch": 0.15, "learning_rate": 0.00019279390060062407, "loss": 0.8449, "step": 2608 }, { "epoch": 0.15, "learning_rate": 0.00019278705124309723, "loss": 0.7921, "step": 2609 }, { "epoch": 0.15, "learning_rate": 0.00019278019875376408, "loss": 0.5549, "step": 2610 }, { "epoch": 0.15, "learning_rate": 0.00019277334313285587, "loss": 0.7473, "step": 2611 }, { "epoch": 0.15, "learning_rate": 0.00019276648438060404, "loss": 0.734, "step": 2612 }, { "epoch": 0.15, "learning_rate": 0.0001927596224972401, "loss": 0.666, "step": 2613 }, { "epoch": 0.15, "learning_rate": 0.00019275275748299557, "loss": 0.6481, "step": 2614 }, { "epoch": 0.15, "learning_rate": 0.00019274588933810226, "loss": 0.7263, "step": 2615 }, { "epoch": 0.15, "learning_rate": 0.00019273901806279194, "loss": 0.7699, "step": 2616 }, { "epoch": 0.15, "learning_rate": 0.00019273214365729655, "loss": 0.6973, "step": 2617 }, { "epoch": 0.15, "learning_rate": 0.00019272526612184812, "loss": 0.6988, "step": 2618 }, { "epoch": 0.15, "learning_rate": 0.00019271838545667876, "loss": 0.7552, "step": 2619 }, { "epoch": 0.15, "learning_rate": 0.0001927115016620207, "loss": 0.7512, "step": 2620 }, { "epoch": 0.15, "learning_rate": 0.00019270461473810637, "loss": 0.9268, "step": 2621 }, { "epoch": 0.15, "learning_rate": 0.00019269772468516813, "loss": 0.6143, "step": 2622 }, { "epoch": 0.15, "learning_rate": 0.00019269083150343859, "loss": 0.8094, "step": 2623 }, { "epoch": 0.15, "learning_rate": 0.00019268393519315038, "loss": 0.7379, "step": 2624 }, { "epoch": 0.15, "learning_rate": 0.0001926770357545363, "loss": 0.7757, "step": 2625 }, { "epoch": 0.15, "learning_rate": 0.0001926701331878292, "loss": 0.7665, "step": 2626 }, { "epoch": 0.15, "learning_rate": 0.00019266322749326208, "loss": 0.5851, "step": 2627 }, { "epoch": 0.15, "learning_rate": 0.000192656318671068, "loss": 0.7532, "step": 2628 }, { "epoch": 0.15, "learning_rate": 0.00019264940672148018, "loss": 0.8037, "step": 2629 }, { "epoch": 0.15, "learning_rate": 0.0001926424916447319, "loss": 0.6179, "step": 2630 }, { "epoch": 0.15, "learning_rate": 0.00019263557344105655, "loss": 0.7614, "step": 2631 }, { "epoch": 0.15, "learning_rate": 0.00019262865211068766, "loss": 0.7835, "step": 2632 }, { "epoch": 0.15, "learning_rate": 0.00019262172765385885, "loss": 0.7329, "step": 2633 }, { "epoch": 0.15, "learning_rate": 0.0001926148000708038, "loss": 0.7889, "step": 2634 }, { "epoch": 0.15, "learning_rate": 0.00019260786936175635, "loss": 0.7271, "step": 2635 }, { "epoch": 0.15, "learning_rate": 0.00019260093552695044, "loss": 0.7568, "step": 2636 }, { "epoch": 0.15, "learning_rate": 0.00019259399856662012, "loss": 0.7515, "step": 2637 }, { "epoch": 0.15, "learning_rate": 0.0001925870584809995, "loss": 0.7777, "step": 2638 }, { "epoch": 0.15, "learning_rate": 0.00019258011527032283, "loss": 0.7332, "step": 2639 }, { "epoch": 0.15, "learning_rate": 0.00019257316893482446, "loss": 0.7449, "step": 2640 }, { "epoch": 0.15, "learning_rate": 0.00019256621947473886, "loss": 0.7454, "step": 2641 }, { "epoch": 0.15, "learning_rate": 0.00019255926689030063, "loss": 0.7416, "step": 2642 }, { "epoch": 0.15, "learning_rate": 0.00019255231118174435, "loss": 0.766, "step": 2643 }, { "epoch": 0.15, "learning_rate": 0.00019254535234930486, "loss": 0.6858, "step": 2644 }, { "epoch": 0.15, "learning_rate": 0.000192538390393217, "loss": 0.8001, "step": 2645 }, { "epoch": 0.15, "learning_rate": 0.00019253142531371582, "loss": 0.7726, "step": 2646 }, { "epoch": 0.15, "learning_rate": 0.00019252445711103632, "loss": 0.6885, "step": 2647 }, { "epoch": 0.15, "learning_rate": 0.00019251748578541375, "loss": 0.765, "step": 2648 }, { "epoch": 0.15, "learning_rate": 0.0001925105113370834, "loss": 0.7848, "step": 2649 }, { "epoch": 0.15, "learning_rate": 0.0001925035337662807, "loss": 0.6602, "step": 2650 }, { "epoch": 0.15, "learning_rate": 0.00019249655307324108, "loss": 0.7129, "step": 2651 }, { "epoch": 0.15, "learning_rate": 0.00019248956925820026, "loss": 0.6867, "step": 2652 }, { "epoch": 0.15, "learning_rate": 0.00019248258232139388, "loss": 0.6859, "step": 2653 }, { "epoch": 0.15, "learning_rate": 0.00019247559226305785, "loss": 0.6795, "step": 2654 }, { "epoch": 0.15, "learning_rate": 0.00019246859908342802, "loss": 0.685, "step": 2655 }, { "epoch": 0.15, "learning_rate": 0.00019246160278274047, "loss": 0.6804, "step": 2656 }, { "epoch": 0.15, "learning_rate": 0.00019245460336123134, "loss": 0.7607, "step": 2657 }, { "epoch": 0.15, "learning_rate": 0.00019244760081913684, "loss": 0.7345, "step": 2658 }, { "epoch": 0.15, "learning_rate": 0.0001924405951566934, "loss": 0.6192, "step": 2659 }, { "epoch": 0.15, "learning_rate": 0.00019243358637413742, "loss": 0.8653, "step": 2660 }, { "epoch": 0.15, "learning_rate": 0.00019242657447170552, "loss": 0.7195, "step": 2661 }, { "epoch": 0.15, "learning_rate": 0.0001924195594496343, "loss": 0.7519, "step": 2662 }, { "epoch": 0.15, "learning_rate": 0.00019241254130816058, "loss": 0.7591, "step": 2663 }, { "epoch": 0.15, "learning_rate": 0.00019240552004752123, "loss": 0.6452, "step": 2664 }, { "epoch": 0.15, "learning_rate": 0.00019239849566795323, "loss": 0.7479, "step": 2665 }, { "epoch": 0.15, "learning_rate": 0.00019239146816969367, "loss": 0.7846, "step": 2666 }, { "epoch": 0.15, "learning_rate": 0.00019238443755297978, "loss": 0.7675, "step": 2667 }, { "epoch": 0.15, "learning_rate": 0.0001923774038180488, "loss": 0.6917, "step": 2668 }, { "epoch": 0.15, "learning_rate": 0.00019237036696513818, "loss": 0.7472, "step": 2669 }, { "epoch": 0.15, "learning_rate": 0.00019236332699448547, "loss": 0.7383, "step": 2670 }, { "epoch": 0.15, "learning_rate": 0.00019235628390632822, "loss": 0.8192, "step": 2671 }, { "epoch": 0.15, "learning_rate": 0.00019234923770090415, "loss": 0.6376, "step": 2672 }, { "epoch": 0.15, "learning_rate": 0.00019234218837845113, "loss": 0.723, "step": 2673 }, { "epoch": 0.15, "learning_rate": 0.00019233513593920707, "loss": 0.6684, "step": 2674 }, { "epoch": 0.15, "learning_rate": 0.00019232808038341, "loss": 0.8057, "step": 2675 }, { "epoch": 0.15, "learning_rate": 0.00019232102171129811, "loss": 0.7412, "step": 2676 }, { "epoch": 0.15, "learning_rate": 0.0001923139599231096, "loss": 0.7628, "step": 2677 }, { "epoch": 0.15, "learning_rate": 0.00019230689501908283, "loss": 0.7329, "step": 2678 }, { "epoch": 0.15, "learning_rate": 0.0001922998269994563, "loss": 0.7073, "step": 2679 }, { "epoch": 0.15, "learning_rate": 0.0001922927558644685, "loss": 0.7508, "step": 2680 }, { "epoch": 0.15, "learning_rate": 0.00019228568161435818, "loss": 0.6076, "step": 2681 }, { "epoch": 0.15, "learning_rate": 0.00019227860424936405, "loss": 0.7152, "step": 2682 }, { "epoch": 0.15, "learning_rate": 0.00019227152376972506, "loss": 0.7283, "step": 2683 }, { "epoch": 0.15, "learning_rate": 0.0001922644401756801, "loss": 0.7675, "step": 2684 }, { "epoch": 0.15, "learning_rate": 0.00019225735346746836, "loss": 0.7601, "step": 2685 }, { "epoch": 0.15, "learning_rate": 0.00019225026364532894, "loss": 0.7354, "step": 2686 }, { "epoch": 0.15, "learning_rate": 0.00019224317070950123, "loss": 0.714, "step": 2687 }, { "epoch": 0.15, "learning_rate": 0.00019223607466022457, "loss": 0.7894, "step": 2688 }, { "epoch": 0.15, "learning_rate": 0.00019222897549773848, "loss": 0.8567, "step": 2689 }, { "epoch": 0.15, "learning_rate": 0.0001922218732222826, "loss": 0.6953, "step": 2690 }, { "epoch": 0.15, "learning_rate": 0.00019221476783409664, "loss": 0.7505, "step": 2691 }, { "epoch": 0.15, "learning_rate": 0.00019220765933342042, "loss": 0.6124, "step": 2692 }, { "epoch": 0.15, "learning_rate": 0.00019220054772049386, "loss": 0.7544, "step": 2693 }, { "epoch": 0.15, "learning_rate": 0.00019219343299555703, "loss": 0.7191, "step": 2694 }, { "epoch": 0.15, "learning_rate": 0.00019218631515885006, "loss": 0.6738, "step": 2695 }, { "epoch": 0.15, "learning_rate": 0.00019217919421061317, "loss": 0.7237, "step": 2696 }, { "epoch": 0.15, "learning_rate": 0.00019217207015108674, "loss": 0.8019, "step": 2697 }, { "epoch": 0.15, "learning_rate": 0.0001921649429805112, "loss": 0.7873, "step": 2698 }, { "epoch": 0.15, "learning_rate": 0.0001921578126991271, "loss": 0.5799, "step": 2699 }, { "epoch": 0.15, "learning_rate": 0.0001921506793071752, "loss": 0.679, "step": 2700 }, { "epoch": 0.15, "learning_rate": 0.00019214354280489612, "loss": 0.7504, "step": 2701 }, { "epoch": 0.15, "learning_rate": 0.00019213640319253085, "loss": 0.7283, "step": 2702 }, { "epoch": 0.15, "learning_rate": 0.00019212926047032034, "loss": 0.697, "step": 2703 }, { "epoch": 0.15, "learning_rate": 0.00019212211463850567, "loss": 0.8085, "step": 2704 }, { "epoch": 0.15, "learning_rate": 0.000192114965697328, "loss": 0.6777, "step": 2705 }, { "epoch": 0.15, "learning_rate": 0.0001921078136470287, "loss": 0.8156, "step": 2706 }, { "epoch": 0.15, "learning_rate": 0.00019210065848784913, "loss": 0.745, "step": 2707 }, { "epoch": 0.15, "learning_rate": 0.00019209350022003077, "loss": 0.788, "step": 2708 }, { "epoch": 0.15, "learning_rate": 0.00019208633884381526, "loss": 0.8178, "step": 2709 }, { "epoch": 0.15, "learning_rate": 0.00019207917435944428, "loss": 0.6626, "step": 2710 }, { "epoch": 0.15, "learning_rate": 0.0001920720067671597, "loss": 0.7947, "step": 2711 }, { "epoch": 0.15, "learning_rate": 0.00019206483606720343, "loss": 0.8499, "step": 2712 }, { "epoch": 0.15, "learning_rate": 0.0001920576622598175, "loss": 0.7214, "step": 2713 }, { "epoch": 0.15, "learning_rate": 0.00019205048534524406, "loss": 0.6745, "step": 2714 }, { "epoch": 0.15, "learning_rate": 0.0001920433053237253, "loss": 0.7654, "step": 2715 }, { "epoch": 0.15, "learning_rate": 0.00019203612219550358, "loss": 0.7309, "step": 2716 }, { "epoch": 0.15, "learning_rate": 0.0001920289359608214, "loss": 0.8346, "step": 2717 }, { "epoch": 0.15, "learning_rate": 0.00019202174661992123, "loss": 0.7021, "step": 2718 }, { "epoch": 0.15, "learning_rate": 0.00019201455417304579, "loss": 0.6852, "step": 2719 }, { "epoch": 0.15, "learning_rate": 0.00019200735862043785, "loss": 0.7924, "step": 2720 }, { "epoch": 0.15, "learning_rate": 0.00019200015996234023, "loss": 0.6881, "step": 2721 }, { "epoch": 0.15, "learning_rate": 0.00019199295819899597, "loss": 0.775, "step": 2722 }, { "epoch": 0.15, "learning_rate": 0.00019198575333064806, "loss": 0.6693, "step": 2723 }, { "epoch": 0.15, "learning_rate": 0.0001919785453575398, "loss": 0.7038, "step": 2724 }, { "epoch": 0.15, "learning_rate": 0.00019197133427991436, "loss": 0.7553, "step": 2725 }, { "epoch": 0.15, "learning_rate": 0.0001919641200980152, "loss": 0.8294, "step": 2726 }, { "epoch": 0.15, "learning_rate": 0.0001919569028120858, "loss": 0.7301, "step": 2727 }, { "epoch": 0.15, "learning_rate": 0.00019194968242236974, "loss": 0.8078, "step": 2728 }, { "epoch": 0.15, "learning_rate": 0.0001919424589291108, "loss": 0.6959, "step": 2729 }, { "epoch": 0.15, "learning_rate": 0.0001919352323325527, "loss": 0.8286, "step": 2730 }, { "epoch": 0.15, "learning_rate": 0.00019192800263293945, "loss": 0.7456, "step": 2731 }, { "epoch": 0.15, "learning_rate": 0.00019192076983051497, "loss": 0.7169, "step": 2732 }, { "epoch": 0.16, "learning_rate": 0.00019191353392552344, "loss": 0.6943, "step": 2733 }, { "epoch": 0.16, "learning_rate": 0.00019190629491820912, "loss": 0.7107, "step": 2734 }, { "epoch": 0.16, "learning_rate": 0.0001918990528088163, "loss": 0.7889, "step": 2735 }, { "epoch": 0.16, "learning_rate": 0.0001918918075975894, "loss": 0.6095, "step": 2736 }, { "epoch": 0.16, "learning_rate": 0.00019188455928477306, "loss": 0.7566, "step": 2737 }, { "epoch": 0.16, "learning_rate": 0.00019187730787061184, "loss": 0.7929, "step": 2738 }, { "epoch": 0.16, "learning_rate": 0.00019187005335535054, "loss": 0.6706, "step": 2739 }, { "epoch": 0.16, "learning_rate": 0.00019186279573923395, "loss": 0.7665, "step": 2740 }, { "epoch": 0.16, "learning_rate": 0.00019185553502250712, "loss": 0.8115, "step": 2741 }, { "epoch": 0.16, "learning_rate": 0.00019184827120541508, "loss": 0.6841, "step": 2742 }, { "epoch": 0.16, "learning_rate": 0.000191841004288203, "loss": 0.7702, "step": 2743 }, { "epoch": 0.16, "learning_rate": 0.0001918337342711162, "loss": 0.7146, "step": 2744 }, { "epoch": 0.16, "learning_rate": 0.00019182646115439996, "loss": 0.7227, "step": 2745 }, { "epoch": 0.16, "learning_rate": 0.0001918191849382999, "loss": 0.5965, "step": 2746 }, { "epoch": 0.16, "learning_rate": 0.00019181190562306147, "loss": 0.7382, "step": 2747 }, { "epoch": 0.16, "learning_rate": 0.00019180462320893052, "loss": 0.6915, "step": 2748 }, { "epoch": 0.16, "learning_rate": 0.0001917973376961527, "loss": 0.6978, "step": 2749 }, { "epoch": 0.16, "learning_rate": 0.00019179004908497405, "loss": 0.7195, "step": 2750 }, { "epoch": 0.16, "learning_rate": 0.00019178275737564046, "loss": 0.7931, "step": 2751 }, { "epoch": 0.16, "learning_rate": 0.00019177546256839812, "loss": 0.8166, "step": 2752 }, { "epoch": 0.16, "learning_rate": 0.00019176816466349322, "loss": 0.8375, "step": 2753 }, { "epoch": 0.16, "learning_rate": 0.00019176086366117211, "loss": 0.8479, "step": 2754 }, { "epoch": 0.16, "learning_rate": 0.0001917535595616812, "loss": 0.8256, "step": 2755 }, { "epoch": 0.16, "learning_rate": 0.00019174625236526702, "loss": 0.6868, "step": 2756 }, { "epoch": 0.16, "learning_rate": 0.0001917389420721762, "loss": 0.7918, "step": 2757 }, { "epoch": 0.16, "learning_rate": 0.0001917316286826555, "loss": 0.7922, "step": 2758 }, { "epoch": 0.16, "learning_rate": 0.00019172431219695172, "loss": 0.7825, "step": 2759 }, { "epoch": 0.16, "learning_rate": 0.0001917169926153119, "loss": 0.7321, "step": 2760 }, { "epoch": 0.16, "learning_rate": 0.000191709669937983, "loss": 0.7079, "step": 2761 }, { "epoch": 0.16, "learning_rate": 0.00019170234416521225, "loss": 0.6954, "step": 2762 }, { "epoch": 0.16, "learning_rate": 0.0001916950152972469, "loss": 0.6536, "step": 2763 }, { "epoch": 0.16, "learning_rate": 0.00019168768333433428, "loss": 0.7077, "step": 2764 }, { "epoch": 0.16, "learning_rate": 0.00019168034827672187, "loss": 0.7807, "step": 2765 }, { "epoch": 0.16, "learning_rate": 0.00019167301012465728, "loss": 0.6915, "step": 2766 }, { "epoch": 0.16, "learning_rate": 0.00019166566887838819, "loss": 0.6952, "step": 2767 }, { "epoch": 0.16, "learning_rate": 0.00019165832453816236, "loss": 0.6775, "step": 2768 }, { "epoch": 0.16, "learning_rate": 0.0001916509771042277, "loss": 0.8163, "step": 2769 }, { "epoch": 0.16, "learning_rate": 0.00019164362657683222, "loss": 0.771, "step": 2770 }, { "epoch": 0.16, "learning_rate": 0.00019163627295622397, "loss": 0.781, "step": 2771 }, { "epoch": 0.16, "learning_rate": 0.00019162891624265117, "loss": 0.7292, "step": 2772 }, { "epoch": 0.16, "learning_rate": 0.00019162155643636218, "loss": 0.7252, "step": 2773 }, { "epoch": 0.16, "learning_rate": 0.00019161419353760537, "loss": 0.7251, "step": 2774 }, { "epoch": 0.16, "learning_rate": 0.00019160682754662924, "loss": 0.8071, "step": 2775 }, { "epoch": 0.16, "learning_rate": 0.00019159945846368244, "loss": 0.7584, "step": 2776 }, { "epoch": 0.16, "learning_rate": 0.00019159208628901367, "loss": 0.7736, "step": 2777 }, { "epoch": 0.16, "learning_rate": 0.0001915847110228718, "loss": 0.7055, "step": 2778 }, { "epoch": 0.16, "learning_rate": 0.00019157733266550575, "loss": 0.7297, "step": 2779 }, { "epoch": 0.16, "learning_rate": 0.00019156995121716455, "loss": 0.6312, "step": 2780 }, { "epoch": 0.16, "learning_rate": 0.00019156256667809734, "loss": 0.6662, "step": 2781 }, { "epoch": 0.16, "learning_rate": 0.00019155517904855337, "loss": 0.6076, "step": 2782 }, { "epoch": 0.16, "learning_rate": 0.000191547788328782, "loss": 0.6434, "step": 2783 }, { "epoch": 0.16, "learning_rate": 0.00019154039451903265, "loss": 0.7449, "step": 2784 }, { "epoch": 0.16, "learning_rate": 0.00019153299761955495, "loss": 0.6594, "step": 2785 }, { "epoch": 0.16, "learning_rate": 0.00019152559763059853, "loss": 0.7722, "step": 2786 }, { "epoch": 0.16, "learning_rate": 0.00019151819455241314, "loss": 0.7517, "step": 2787 }, { "epoch": 0.16, "learning_rate": 0.00019151078838524867, "loss": 0.6992, "step": 2788 }, { "epoch": 0.16, "learning_rate": 0.0001915033791293551, "loss": 0.7389, "step": 2789 }, { "epoch": 0.16, "learning_rate": 0.0001914959667849825, "loss": 0.7266, "step": 2790 }, { "epoch": 0.16, "learning_rate": 0.00019148855135238106, "loss": 0.778, "step": 2791 }, { "epoch": 0.16, "learning_rate": 0.00019148113283180108, "loss": 0.7719, "step": 2792 }, { "epoch": 0.16, "learning_rate": 0.00019147371122349297, "loss": 0.7454, "step": 2793 }, { "epoch": 0.16, "learning_rate": 0.00019146628652770719, "loss": 0.7448, "step": 2794 }, { "epoch": 0.16, "learning_rate": 0.00019145885874469434, "loss": 0.6934, "step": 2795 }, { "epoch": 0.16, "learning_rate": 0.00019145142787470516, "loss": 0.6932, "step": 2796 }, { "epoch": 0.16, "learning_rate": 0.00019144399391799043, "loss": 0.6702, "step": 2797 }, { "epoch": 0.16, "learning_rate": 0.00019143655687480114, "loss": 0.7557, "step": 2798 }, { "epoch": 0.16, "learning_rate": 0.00019142911674538822, "loss": 0.7659, "step": 2799 }, { "epoch": 0.16, "learning_rate": 0.00019142167353000282, "loss": 0.7347, "step": 2800 }, { "epoch": 0.16, "learning_rate": 0.0001914142272288962, "loss": 0.7237, "step": 2801 }, { "epoch": 0.16, "learning_rate": 0.00019140677784231964, "loss": 0.6938, "step": 2802 }, { "epoch": 0.16, "learning_rate": 0.00019139932537052463, "loss": 0.8078, "step": 2803 }, { "epoch": 0.16, "learning_rate": 0.00019139186981376267, "loss": 0.7834, "step": 2804 }, { "epoch": 0.16, "learning_rate": 0.00019138441117228544, "loss": 0.7425, "step": 2805 }, { "epoch": 0.16, "learning_rate": 0.00019137694944634464, "loss": 0.8588, "step": 2806 }, { "epoch": 0.16, "learning_rate": 0.00019136948463619217, "loss": 0.7247, "step": 2807 }, { "epoch": 0.16, "learning_rate": 0.00019136201674208, "loss": 0.7152, "step": 2808 }, { "epoch": 0.16, "learning_rate": 0.0001913545457642601, "loss": 0.7984, "step": 2809 }, { "epoch": 0.16, "learning_rate": 0.00019134707170298472, "loss": 0.6556, "step": 2810 }, { "epoch": 0.16, "learning_rate": 0.00019133959455850609, "loss": 0.7658, "step": 2811 }, { "epoch": 0.16, "learning_rate": 0.00019133211433107662, "loss": 0.7702, "step": 2812 }, { "epoch": 0.16, "learning_rate": 0.00019132463102094875, "loss": 0.711, "step": 2813 }, { "epoch": 0.16, "learning_rate": 0.00019131714462837507, "loss": 0.6857, "step": 2814 }, { "epoch": 0.16, "learning_rate": 0.0001913096551536083, "loss": 0.731, "step": 2815 }, { "epoch": 0.16, "learning_rate": 0.00019130216259690116, "loss": 0.7705, "step": 2816 }, { "epoch": 0.16, "learning_rate": 0.0001912946669585066, "loss": 0.736, "step": 2817 }, { "epoch": 0.16, "learning_rate": 0.0001912871682386776, "loss": 0.6513, "step": 2818 }, { "epoch": 0.16, "learning_rate": 0.0001912796664376673, "loss": 0.6356, "step": 2819 }, { "epoch": 0.16, "learning_rate": 0.00019127216155572882, "loss": 0.7817, "step": 2820 }, { "epoch": 0.16, "learning_rate": 0.00019126465359311552, "loss": 0.7829, "step": 2821 }, { "epoch": 0.16, "learning_rate": 0.00019125714255008083, "loss": 0.7733, "step": 2822 }, { "epoch": 0.16, "learning_rate": 0.00019124962842687824, "loss": 0.7142, "step": 2823 }, { "epoch": 0.16, "learning_rate": 0.00019124211122376137, "loss": 0.6919, "step": 2824 }, { "epoch": 0.16, "learning_rate": 0.00019123459094098398, "loss": 0.7982, "step": 2825 }, { "epoch": 0.16, "learning_rate": 0.00019122706757879985, "loss": 0.757, "step": 2826 }, { "epoch": 0.16, "learning_rate": 0.00019121954113746298, "loss": 0.7401, "step": 2827 }, { "epoch": 0.16, "learning_rate": 0.0001912120116172273, "loss": 0.8202, "step": 2828 }, { "epoch": 0.16, "learning_rate": 0.00019120447901834706, "loss": 0.6971, "step": 2829 }, { "epoch": 0.16, "learning_rate": 0.00019119694334107645, "loss": 0.6147, "step": 2830 }, { "epoch": 0.16, "learning_rate": 0.00019118940458566982, "loss": 0.7579, "step": 2831 }, { "epoch": 0.16, "learning_rate": 0.00019118186275238165, "loss": 0.7038, "step": 2832 }, { "epoch": 0.16, "learning_rate": 0.00019117431784146645, "loss": 0.7703, "step": 2833 }, { "epoch": 0.16, "learning_rate": 0.00019116676985317895, "loss": 0.6943, "step": 2834 }, { "epoch": 0.16, "learning_rate": 0.00019115921878777386, "loss": 0.7608, "step": 2835 }, { "epoch": 0.16, "learning_rate": 0.00019115166464550606, "loss": 0.8512, "step": 2836 }, { "epoch": 0.16, "learning_rate": 0.00019114410742663052, "loss": 0.6787, "step": 2837 }, { "epoch": 0.16, "learning_rate": 0.0001911365471314023, "loss": 0.7452, "step": 2838 }, { "epoch": 0.16, "learning_rate": 0.00019112898376007662, "loss": 0.7814, "step": 2839 }, { "epoch": 0.16, "learning_rate": 0.00019112141731290875, "loss": 0.6699, "step": 2840 }, { "epoch": 0.16, "learning_rate": 0.00019111384779015405, "loss": 0.7571, "step": 2841 }, { "epoch": 0.16, "learning_rate": 0.00019110627519206805, "loss": 0.6761, "step": 2842 }, { "epoch": 0.16, "learning_rate": 0.00019109869951890632, "loss": 0.7816, "step": 2843 }, { "epoch": 0.16, "learning_rate": 0.00019109112077092457, "loss": 0.7043, "step": 2844 }, { "epoch": 0.16, "learning_rate": 0.00019108353894837861, "loss": 0.6788, "step": 2845 }, { "epoch": 0.16, "learning_rate": 0.0001910759540515243, "loss": 0.7096, "step": 2846 }, { "epoch": 0.16, "learning_rate": 0.00019106836608061772, "loss": 0.702, "step": 2847 }, { "epoch": 0.16, "learning_rate": 0.00019106077503591493, "loss": 0.7716, "step": 2848 }, { "epoch": 0.16, "learning_rate": 0.0001910531809176722, "loss": 0.6781, "step": 2849 }, { "epoch": 0.16, "learning_rate": 0.00019104558372614577, "loss": 0.7223, "step": 2850 }, { "epoch": 0.16, "learning_rate": 0.00019103798346159213, "loss": 0.786, "step": 2851 }, { "epoch": 0.16, "learning_rate": 0.00019103038012426783, "loss": 0.743, "step": 2852 }, { "epoch": 0.16, "learning_rate": 0.00019102277371442941, "loss": 0.7682, "step": 2853 }, { "epoch": 0.16, "learning_rate": 0.00019101516423233368, "loss": 0.715, "step": 2854 }, { "epoch": 0.16, "learning_rate": 0.00019100755167823747, "loss": 0.6953, "step": 2855 }, { "epoch": 0.16, "learning_rate": 0.0001909999360523977, "loss": 0.6888, "step": 2856 }, { "epoch": 0.16, "learning_rate": 0.00019099231735507147, "loss": 0.7345, "step": 2857 }, { "epoch": 0.16, "learning_rate": 0.00019098469558651585, "loss": 0.5804, "step": 2858 }, { "epoch": 0.16, "learning_rate": 0.00019097707074698815, "loss": 0.7119, "step": 2859 }, { "epoch": 0.16, "learning_rate": 0.00019096944283674571, "loss": 0.7181, "step": 2860 }, { "epoch": 0.16, "learning_rate": 0.00019096181185604603, "loss": 0.7272, "step": 2861 }, { "epoch": 0.16, "learning_rate": 0.00019095417780514663, "loss": 0.8551, "step": 2862 }, { "epoch": 0.16, "learning_rate": 0.0001909465406843052, "loss": 0.7419, "step": 2863 }, { "epoch": 0.16, "learning_rate": 0.00019093890049377946, "loss": 0.6779, "step": 2864 }, { "epoch": 0.16, "learning_rate": 0.00019093125723382737, "loss": 0.6874, "step": 2865 }, { "epoch": 0.16, "learning_rate": 0.00019092361090470688, "loss": 0.7439, "step": 2866 }, { "epoch": 0.16, "learning_rate": 0.00019091596150667603, "loss": 0.684, "step": 2867 }, { "epoch": 0.16, "learning_rate": 0.00019090830903999305, "loss": 0.8336, "step": 2868 }, { "epoch": 0.16, "learning_rate": 0.00019090065350491626, "loss": 0.7538, "step": 2869 }, { "epoch": 0.16, "learning_rate": 0.00019089299490170398, "loss": 0.7048, "step": 2870 }, { "epoch": 0.16, "learning_rate": 0.00019088533323061477, "loss": 0.7285, "step": 2871 }, { "epoch": 0.16, "learning_rate": 0.00019087766849190722, "loss": 0.7164, "step": 2872 }, { "epoch": 0.16, "learning_rate": 0.00019087000068584, "loss": 0.7274, "step": 2873 }, { "epoch": 0.16, "learning_rate": 0.00019086232981267194, "loss": 0.7872, "step": 2874 }, { "epoch": 0.16, "learning_rate": 0.00019085465587266197, "loss": 0.7481, "step": 2875 }, { "epoch": 0.16, "learning_rate": 0.0001908469788660691, "loss": 0.7533, "step": 2876 }, { "epoch": 0.16, "learning_rate": 0.00019083929879315238, "loss": 0.7145, "step": 2877 }, { "epoch": 0.16, "learning_rate": 0.00019083161565417115, "loss": 0.7412, "step": 2878 }, { "epoch": 0.16, "learning_rate": 0.00019082392944938466, "loss": 0.7953, "step": 2879 }, { "epoch": 0.16, "learning_rate": 0.00019081624017905236, "loss": 0.766, "step": 2880 }, { "epoch": 0.16, "learning_rate": 0.00019080854784343375, "loss": 0.7772, "step": 2881 }, { "epoch": 0.16, "learning_rate": 0.0001908008524427885, "loss": 0.7515, "step": 2882 }, { "epoch": 0.16, "learning_rate": 0.00019079315397737637, "loss": 0.7997, "step": 2883 }, { "epoch": 0.16, "learning_rate": 0.00019078545244745717, "loss": 0.7404, "step": 2884 }, { "epoch": 0.16, "learning_rate": 0.00019077774785329087, "loss": 0.7421, "step": 2885 }, { "epoch": 0.16, "learning_rate": 0.00019077004019513752, "loss": 0.7441, "step": 2886 }, { "epoch": 0.16, "learning_rate": 0.00019076232947325722, "loss": 0.807, "step": 2887 }, { "epoch": 0.16, "learning_rate": 0.00019075461568791028, "loss": 0.6974, "step": 2888 }, { "epoch": 0.16, "learning_rate": 0.00019074689883935705, "loss": 0.7391, "step": 2889 }, { "epoch": 0.16, "learning_rate": 0.000190739178927858, "loss": 0.683, "step": 2890 }, { "epoch": 0.16, "learning_rate": 0.0001907314559536737, "loss": 0.662, "step": 2891 }, { "epoch": 0.16, "learning_rate": 0.0001907237299170648, "loss": 0.7876, "step": 2892 }, { "epoch": 0.16, "learning_rate": 0.00019071600081829203, "loss": 0.6876, "step": 2893 }, { "epoch": 0.16, "learning_rate": 0.0001907082686576164, "loss": 0.7752, "step": 2894 }, { "epoch": 0.16, "learning_rate": 0.00019070053343529875, "loss": 0.7632, "step": 2895 }, { "epoch": 0.16, "learning_rate": 0.00019069279515160025, "loss": 0.7097, "step": 2896 }, { "epoch": 0.16, "learning_rate": 0.0001906850538067821, "loss": 0.6291, "step": 2897 }, { "epoch": 0.16, "learning_rate": 0.0001906773094011055, "loss": 0.7478, "step": 2898 }, { "epoch": 0.16, "learning_rate": 0.00019066956193483195, "loss": 0.7423, "step": 2899 }, { "epoch": 0.16, "learning_rate": 0.00019066181140822285, "loss": 0.6907, "step": 2900 }, { "epoch": 0.16, "learning_rate": 0.00019065405782153985, "loss": 0.6826, "step": 2901 }, { "epoch": 0.16, "learning_rate": 0.00019064630117504464, "loss": 0.7354, "step": 2902 }, { "epoch": 0.16, "learning_rate": 0.00019063854146899906, "loss": 0.7074, "step": 2903 }, { "epoch": 0.16, "learning_rate": 0.000190630778703665, "loss": 0.723, "step": 2904 }, { "epoch": 0.16, "learning_rate": 0.00019062301287930446, "loss": 0.6482, "step": 2905 }, { "epoch": 0.16, "learning_rate": 0.0001906152439961796, "loss": 0.7896, "step": 2906 }, { "epoch": 0.16, "learning_rate": 0.00019060747205455256, "loss": 0.7716, "step": 2907 }, { "epoch": 0.16, "learning_rate": 0.0001905996970546857, "loss": 0.7855, "step": 2908 }, { "epoch": 0.17, "learning_rate": 0.00019059191899684154, "loss": 0.8203, "step": 2909 }, { "epoch": 0.17, "learning_rate": 0.00019058413788128244, "loss": 0.6878, "step": 2910 }, { "epoch": 0.17, "learning_rate": 0.00019057635370827117, "loss": 0.7986, "step": 2911 }, { "epoch": 0.17, "learning_rate": 0.0001905685664780704, "loss": 0.7197, "step": 2912 }, { "epoch": 0.17, "learning_rate": 0.00019056077619094303, "loss": 0.7137, "step": 2913 }, { "epoch": 0.17, "learning_rate": 0.00019055298284715192, "loss": 0.7513, "step": 2914 }, { "epoch": 0.17, "learning_rate": 0.0001905451864469601, "loss": 0.7426, "step": 2915 }, { "epoch": 0.17, "learning_rate": 0.00019053738699063085, "loss": 0.7636, "step": 2916 }, { "epoch": 0.17, "learning_rate": 0.00019052958447842732, "loss": 0.6698, "step": 2917 }, { "epoch": 0.17, "learning_rate": 0.0001905217789106129, "loss": 0.6569, "step": 2918 }, { "epoch": 0.17, "learning_rate": 0.00019051397028745104, "loss": 0.709, "step": 2919 }, { "epoch": 0.17, "learning_rate": 0.00019050615860920531, "loss": 0.7499, "step": 2920 }, { "epoch": 0.17, "learning_rate": 0.00019049834387613933, "loss": 0.7503, "step": 2921 }, { "epoch": 0.17, "learning_rate": 0.00019049052608851693, "loss": 0.6476, "step": 2922 }, { "epoch": 0.17, "learning_rate": 0.00019048270524660196, "loss": 0.7305, "step": 2923 }, { "epoch": 0.17, "learning_rate": 0.00019047488135065838, "loss": 0.7196, "step": 2924 }, { "epoch": 0.17, "learning_rate": 0.00019046705440095026, "loss": 0.7664, "step": 2925 }, { "epoch": 0.17, "learning_rate": 0.00019045922439774184, "loss": 0.7001, "step": 2926 }, { "epoch": 0.17, "learning_rate": 0.0001904513913412973, "loss": 0.662, "step": 2927 }, { "epoch": 0.17, "learning_rate": 0.00019044355523188113, "loss": 0.704, "step": 2928 }, { "epoch": 0.17, "learning_rate": 0.00019043571606975777, "loss": 0.7109, "step": 2929 }, { "epoch": 0.17, "learning_rate": 0.0001904278738551918, "loss": 0.7338, "step": 2930 }, { "epoch": 0.17, "learning_rate": 0.00019042002858844792, "loss": 0.7636, "step": 2931 }, { "epoch": 0.17, "learning_rate": 0.00019041218026979095, "loss": 0.738, "step": 2932 }, { "epoch": 0.17, "learning_rate": 0.0001904043288994858, "loss": 0.6815, "step": 2933 }, { "epoch": 0.17, "learning_rate": 0.00019039647447779742, "loss": 0.768, "step": 2934 }, { "epoch": 0.17, "learning_rate": 0.000190388617004991, "loss": 0.6507, "step": 2935 }, { "epoch": 0.17, "learning_rate": 0.00019038075648133168, "loss": 0.6975, "step": 2936 }, { "epoch": 0.17, "learning_rate": 0.0001903728929070848, "loss": 0.6964, "step": 2937 }, { "epoch": 0.17, "learning_rate": 0.00019036502628251575, "loss": 0.7093, "step": 2938 }, { "epoch": 0.17, "learning_rate": 0.00019035715660789012, "loss": 0.7697, "step": 2939 }, { "epoch": 0.17, "learning_rate": 0.00019034928388347343, "loss": 0.7742, "step": 2940 }, { "epoch": 0.17, "learning_rate": 0.0001903414081095315, "loss": 0.7073, "step": 2941 }, { "epoch": 0.17, "learning_rate": 0.0001903335292863301, "loss": 0.7713, "step": 2942 }, { "epoch": 0.17, "learning_rate": 0.00019032564741413516, "loss": 0.7628, "step": 2943 }, { "epoch": 0.17, "learning_rate": 0.00019031776249321276, "loss": 0.8023, "step": 2944 }, { "epoch": 0.17, "learning_rate": 0.000190309874523829, "loss": 0.7521, "step": 2945 }, { "epoch": 0.17, "learning_rate": 0.00019030198350625013, "loss": 0.7821, "step": 2946 }, { "epoch": 0.17, "learning_rate": 0.00019029408944074247, "loss": 0.7062, "step": 2947 }, { "epoch": 0.17, "learning_rate": 0.00019028619232757253, "loss": 0.6922, "step": 2948 }, { "epoch": 0.17, "learning_rate": 0.00019027829216700678, "loss": 0.8084, "step": 2949 }, { "epoch": 0.17, "learning_rate": 0.0001902703889593119, "loss": 0.7722, "step": 2950 }, { "epoch": 0.17, "learning_rate": 0.00019026248270475468, "loss": 0.6683, "step": 2951 }, { "epoch": 0.17, "learning_rate": 0.00019025457340360194, "loss": 0.6593, "step": 2952 }, { "epoch": 0.17, "learning_rate": 0.00019024666105612066, "loss": 0.7513, "step": 2953 }, { "epoch": 0.17, "learning_rate": 0.00019023874566257784, "loss": 0.7857, "step": 2954 }, { "epoch": 0.17, "learning_rate": 0.00019023082722324073, "loss": 0.7643, "step": 2955 }, { "epoch": 0.17, "learning_rate": 0.00019022290573837656, "loss": 0.7293, "step": 2956 }, { "epoch": 0.17, "learning_rate": 0.0001902149812082527, "loss": 0.7714, "step": 2957 }, { "epoch": 0.17, "learning_rate": 0.00019020705363313663, "loss": 0.8449, "step": 2958 }, { "epoch": 0.17, "learning_rate": 0.00019019912301329592, "loss": 0.8284, "step": 2959 }, { "epoch": 0.17, "learning_rate": 0.00019019118934899821, "loss": 0.7599, "step": 2960 }, { "epoch": 0.17, "learning_rate": 0.0001901832526405114, "loss": 0.7699, "step": 2961 }, { "epoch": 0.17, "learning_rate": 0.00019017531288810323, "loss": 0.7796, "step": 2962 }, { "epoch": 0.17, "learning_rate": 0.0001901673700920418, "loss": 0.709, "step": 2963 }, { "epoch": 0.17, "learning_rate": 0.00019015942425259515, "loss": 0.7674, "step": 2964 }, { "epoch": 0.17, "learning_rate": 0.00019015147537003148, "loss": 0.62, "step": 2965 }, { "epoch": 0.17, "learning_rate": 0.00019014352344461907, "loss": 0.7757, "step": 2966 }, { "epoch": 0.17, "learning_rate": 0.00019013556847662634, "loss": 0.7457, "step": 2967 }, { "epoch": 0.17, "learning_rate": 0.0001901276104663218, "loss": 0.7039, "step": 2968 }, { "epoch": 0.17, "learning_rate": 0.00019011964941397405, "loss": 0.7722, "step": 2969 }, { "epoch": 0.17, "learning_rate": 0.00019011168531985175, "loss": 0.7464, "step": 2970 }, { "epoch": 0.17, "learning_rate": 0.00019010371818422377, "loss": 0.6775, "step": 2971 }, { "epoch": 0.17, "learning_rate": 0.00019009574800735898, "loss": 0.7718, "step": 2972 }, { "epoch": 0.17, "learning_rate": 0.0001900877747895264, "loss": 0.7379, "step": 2973 }, { "epoch": 0.17, "learning_rate": 0.0001900797985309952, "loss": 0.7446, "step": 2974 }, { "epoch": 0.17, "learning_rate": 0.0001900718192320345, "loss": 0.8097, "step": 2975 }, { "epoch": 0.17, "learning_rate": 0.00019006383689291372, "loss": 0.732, "step": 2976 }, { "epoch": 0.17, "learning_rate": 0.00019005585151390223, "loss": 0.7282, "step": 2977 }, { "epoch": 0.17, "learning_rate": 0.00019004786309526956, "loss": 0.8317, "step": 2978 }, { "epoch": 0.17, "learning_rate": 0.00019003987163728535, "loss": 0.6768, "step": 2979 }, { "epoch": 0.17, "learning_rate": 0.00019003187714021938, "loss": 0.7384, "step": 2980 }, { "epoch": 0.17, "learning_rate": 0.00019002387960434137, "loss": 0.6808, "step": 2981 }, { "epoch": 0.17, "learning_rate": 0.0001900158790299214, "loss": 0.7768, "step": 2982 }, { "epoch": 0.17, "learning_rate": 0.0001900078754172294, "loss": 0.7264, "step": 2983 }, { "epoch": 0.17, "learning_rate": 0.00018999986876653554, "loss": 0.7553, "step": 2984 }, { "epoch": 0.17, "learning_rate": 0.00018999185907811009, "loss": 0.7328, "step": 2985 }, { "epoch": 0.17, "learning_rate": 0.00018998384635222334, "loss": 0.7652, "step": 2986 }, { "epoch": 0.17, "learning_rate": 0.00018997583058914585, "loss": 0.5986, "step": 2987 }, { "epoch": 0.17, "learning_rate": 0.0001899678117891481, "loss": 0.7288, "step": 2988 }, { "epoch": 0.17, "learning_rate": 0.0001899597899525007, "loss": 0.6787, "step": 2989 }, { "epoch": 0.17, "learning_rate": 0.00018995176507947452, "loss": 0.728, "step": 2990 }, { "epoch": 0.17, "learning_rate": 0.00018994373717034032, "loss": 0.827, "step": 2991 }, { "epoch": 0.17, "learning_rate": 0.00018993570622536914, "loss": 0.6999, "step": 2992 }, { "epoch": 0.17, "learning_rate": 0.00018992767224483198, "loss": 0.7313, "step": 2993 }, { "epoch": 0.17, "learning_rate": 0.00018991963522900005, "loss": 0.6136, "step": 2994 }, { "epoch": 0.17, "learning_rate": 0.0001899115951781446, "loss": 0.7125, "step": 2995 }, { "epoch": 0.17, "learning_rate": 0.00018990355209253703, "loss": 0.7367, "step": 2996 }, { "epoch": 0.17, "learning_rate": 0.00018989550597244876, "loss": 0.7637, "step": 2997 }, { "epoch": 0.17, "learning_rate": 0.00018988745681815144, "loss": 0.641, "step": 2998 }, { "epoch": 0.17, "learning_rate": 0.0001898794046299167, "loss": 0.7306, "step": 2999 }, { "epoch": 0.17, "learning_rate": 0.00018987134940801634, "loss": 0.6481, "step": 3000 }, { "epoch": 0.17, "learning_rate": 0.00018986329115272223, "loss": 0.7121, "step": 3001 }, { "epoch": 0.17, "learning_rate": 0.00018985522986430639, "loss": 0.7363, "step": 3002 }, { "epoch": 0.17, "learning_rate": 0.0001898471655430409, "loss": 0.7309, "step": 3003 }, { "epoch": 0.17, "learning_rate": 0.0001898390981891979, "loss": 0.7294, "step": 3004 }, { "epoch": 0.17, "learning_rate": 0.00018983102780304974, "loss": 0.7514, "step": 3005 }, { "epoch": 0.17, "learning_rate": 0.00018982295438486882, "loss": 0.7042, "step": 3006 }, { "epoch": 0.17, "learning_rate": 0.0001898148779349276, "loss": 0.7678, "step": 3007 }, { "epoch": 0.17, "learning_rate": 0.00018980679845349868, "loss": 0.7307, "step": 3008 }, { "epoch": 0.17, "learning_rate": 0.0001897987159408548, "loss": 0.7709, "step": 3009 }, { "epoch": 0.17, "learning_rate": 0.0001897906303972688, "loss": 0.7577, "step": 3010 }, { "epoch": 0.17, "learning_rate": 0.00018978254182301348, "loss": 0.7037, "step": 3011 }, { "epoch": 0.17, "learning_rate": 0.00018977445021836192, "loss": 0.7584, "step": 3012 }, { "epoch": 0.17, "learning_rate": 0.00018976635558358722, "loss": 0.8185, "step": 3013 }, { "epoch": 0.17, "learning_rate": 0.00018975825791896263, "loss": 0.7583, "step": 3014 }, { "epoch": 0.17, "learning_rate": 0.00018975015722476143, "loss": 0.741, "step": 3015 }, { "epoch": 0.17, "learning_rate": 0.00018974205350125701, "loss": 0.7073, "step": 3016 }, { "epoch": 0.17, "learning_rate": 0.00018973394674872296, "loss": 0.7313, "step": 3017 }, { "epoch": 0.17, "learning_rate": 0.00018972583696743285, "loss": 0.7393, "step": 3018 }, { "epoch": 0.17, "learning_rate": 0.00018971772415766043, "loss": 0.6962, "step": 3019 }, { "epoch": 0.17, "learning_rate": 0.00018970960831967955, "loss": 0.7585, "step": 3020 }, { "epoch": 0.17, "learning_rate": 0.0001897014894537641, "loss": 0.7573, "step": 3021 }, { "epoch": 0.17, "learning_rate": 0.00018969336756018815, "loss": 0.7384, "step": 3022 }, { "epoch": 0.17, "learning_rate": 0.00018968524263922575, "loss": 0.6772, "step": 3023 }, { "epoch": 0.17, "learning_rate": 0.00018967711469115126, "loss": 0.7366, "step": 3024 }, { "epoch": 0.17, "learning_rate": 0.00018966898371623896, "loss": 0.6564, "step": 3025 }, { "epoch": 0.17, "learning_rate": 0.0001896608497147633, "loss": 0.7892, "step": 3026 }, { "epoch": 0.17, "learning_rate": 0.0001896527126869988, "loss": 0.6295, "step": 3027 }, { "epoch": 0.17, "learning_rate": 0.0001896445726332201, "loss": 0.7508, "step": 3028 }, { "epoch": 0.17, "learning_rate": 0.00018963642955370201, "loss": 0.8041, "step": 3029 }, { "epoch": 0.17, "learning_rate": 0.00018962828344871934, "loss": 0.8618, "step": 3030 }, { "epoch": 0.17, "learning_rate": 0.00018962013431854702, "loss": 0.6927, "step": 3031 }, { "epoch": 0.17, "learning_rate": 0.00018961198216346016, "loss": 0.7109, "step": 3032 }, { "epoch": 0.17, "learning_rate": 0.00018960382698373387, "loss": 0.7073, "step": 3033 }, { "epoch": 0.17, "learning_rate": 0.00018959566877964343, "loss": 0.6236, "step": 3034 }, { "epoch": 0.17, "learning_rate": 0.0001895875075514642, "loss": 0.6994, "step": 3035 }, { "epoch": 0.17, "learning_rate": 0.00018957934329947164, "loss": 0.7397, "step": 3036 }, { "epoch": 0.17, "learning_rate": 0.0001895711760239413, "loss": 0.786, "step": 3037 }, { "epoch": 0.17, "learning_rate": 0.00018956300572514886, "loss": 0.6813, "step": 3038 }, { "epoch": 0.17, "learning_rate": 0.00018955483240337012, "loss": 0.7245, "step": 3039 }, { "epoch": 0.17, "learning_rate": 0.00018954665605888088, "loss": 0.8674, "step": 3040 }, { "epoch": 0.17, "learning_rate": 0.00018953847669195715, "loss": 0.7153, "step": 3041 }, { "epoch": 0.17, "learning_rate": 0.00018953029430287502, "loss": 0.757, "step": 3042 }, { "epoch": 0.17, "learning_rate": 0.00018952210889191067, "loss": 0.7244, "step": 3043 }, { "epoch": 0.17, "learning_rate": 0.00018951392045934035, "loss": 0.6781, "step": 3044 }, { "epoch": 0.17, "learning_rate": 0.00018950572900544046, "loss": 0.7233, "step": 3045 }, { "epoch": 0.17, "learning_rate": 0.00018949753453048748, "loss": 0.6708, "step": 3046 }, { "epoch": 0.17, "learning_rate": 0.000189489337034758, "loss": 0.7345, "step": 3047 }, { "epoch": 0.17, "learning_rate": 0.00018948113651852868, "loss": 0.6974, "step": 3048 }, { "epoch": 0.17, "learning_rate": 0.00018947293298207635, "loss": 0.7335, "step": 3049 }, { "epoch": 0.17, "learning_rate": 0.00018946472642567788, "loss": 0.6386, "step": 3050 }, { "epoch": 0.17, "learning_rate": 0.00018945651684961027, "loss": 0.7318, "step": 3051 }, { "epoch": 0.17, "learning_rate": 0.0001894483042541506, "loss": 0.7057, "step": 3052 }, { "epoch": 0.17, "learning_rate": 0.00018944008863957604, "loss": 0.7284, "step": 3053 }, { "epoch": 0.17, "learning_rate": 0.00018943187000616395, "loss": 0.7454, "step": 3054 }, { "epoch": 0.17, "learning_rate": 0.00018942364835419174, "loss": 0.8444, "step": 3055 }, { "epoch": 0.17, "learning_rate": 0.0001894154236839368, "loss": 0.7911, "step": 3056 }, { "epoch": 0.17, "learning_rate": 0.00018940719599567688, "loss": 0.6745, "step": 3057 }, { "epoch": 0.17, "learning_rate": 0.0001893989652896896, "loss": 0.6406, "step": 3058 }, { "epoch": 0.17, "learning_rate": 0.00018939073156625278, "loss": 0.7464, "step": 3059 }, { "epoch": 0.17, "learning_rate": 0.0001893824948256443, "loss": 0.6927, "step": 3060 }, { "epoch": 0.17, "learning_rate": 0.00018937425506814225, "loss": 0.7564, "step": 3061 }, { "epoch": 0.17, "learning_rate": 0.00018936601229402466, "loss": 0.7702, "step": 3062 }, { "epoch": 0.17, "learning_rate": 0.0001893577665035698, "loss": 0.7652, "step": 3063 }, { "epoch": 0.17, "learning_rate": 0.000189349517697056, "loss": 0.7757, "step": 3064 }, { "epoch": 0.17, "learning_rate": 0.00018934126587476162, "loss": 0.7074, "step": 3065 }, { "epoch": 0.17, "learning_rate": 0.00018933301103696523, "loss": 0.8045, "step": 3066 }, { "epoch": 0.17, "learning_rate": 0.0001893247531839454, "loss": 0.7868, "step": 3067 }, { "epoch": 0.17, "learning_rate": 0.00018931649231598093, "loss": 0.8091, "step": 3068 }, { "epoch": 0.17, "learning_rate": 0.00018930822843335056, "loss": 0.7172, "step": 3069 }, { "epoch": 0.17, "learning_rate": 0.00018929996153633328, "loss": 0.7689, "step": 3070 }, { "epoch": 0.17, "learning_rate": 0.0001892916916252081, "loss": 0.7272, "step": 3071 }, { "epoch": 0.17, "learning_rate": 0.00018928341870025418, "loss": 0.6704, "step": 3072 }, { "epoch": 0.17, "learning_rate": 0.0001892751427617507, "loss": 0.7283, "step": 3073 }, { "epoch": 0.17, "learning_rate": 0.000189266863809977, "loss": 0.6613, "step": 3074 }, { "epoch": 0.17, "learning_rate": 0.00018925858184521256, "loss": 0.727, "step": 3075 }, { "epoch": 0.17, "learning_rate": 0.0001892502968677369, "loss": 0.7413, "step": 3076 }, { "epoch": 0.17, "learning_rate": 0.00018924200887782963, "loss": 0.7059, "step": 3077 }, { "epoch": 0.17, "learning_rate": 0.00018923371787577052, "loss": 0.7261, "step": 3078 }, { "epoch": 0.17, "learning_rate": 0.0001892254238618394, "loss": 0.8012, "step": 3079 }, { "epoch": 0.17, "learning_rate": 0.00018921712683631626, "loss": 0.7637, "step": 3080 }, { "epoch": 0.17, "learning_rate": 0.00018920882679948107, "loss": 0.6844, "step": 3081 }, { "epoch": 0.17, "learning_rate": 0.00018920052375161404, "loss": 0.8009, "step": 3082 }, { "epoch": 0.17, "learning_rate": 0.00018919221769299539, "loss": 0.7078, "step": 3083 }, { "epoch": 0.17, "learning_rate": 0.00018918390862390547, "loss": 0.7538, "step": 3084 }, { "epoch": 0.17, "learning_rate": 0.00018917559654462474, "loss": 0.7651, "step": 3085 }, { "epoch": 0.18, "learning_rate": 0.00018916728145543376, "loss": 0.7938, "step": 3086 }, { "epoch": 0.18, "learning_rate": 0.00018915896335661318, "loss": 0.685, "step": 3087 }, { "epoch": 0.18, "learning_rate": 0.00018915064224844373, "loss": 0.7628, "step": 3088 }, { "epoch": 0.18, "learning_rate": 0.00018914231813120635, "loss": 0.7408, "step": 3089 }, { "epoch": 0.18, "learning_rate": 0.0001891339910051819, "loss": 0.6841, "step": 3090 }, { "epoch": 0.18, "learning_rate": 0.00018912566087065153, "loss": 0.7946, "step": 3091 }, { "epoch": 0.18, "learning_rate": 0.00018911732772789637, "loss": 0.6357, "step": 3092 }, { "epoch": 0.18, "learning_rate": 0.00018910899157719766, "loss": 0.6376, "step": 3093 }, { "epoch": 0.18, "learning_rate": 0.0001891006524188368, "loss": 0.7763, "step": 3094 }, { "epoch": 0.18, "learning_rate": 0.0001890923102530952, "loss": 0.7523, "step": 3095 }, { "epoch": 0.18, "learning_rate": 0.00018908396508025455, "loss": 0.7475, "step": 3096 }, { "epoch": 0.18, "learning_rate": 0.00018907561690059642, "loss": 0.7323, "step": 3097 }, { "epoch": 0.18, "learning_rate": 0.00018906726571440259, "loss": 0.6828, "step": 3098 }, { "epoch": 0.18, "learning_rate": 0.00018905891152195495, "loss": 0.8168, "step": 3099 }, { "epoch": 0.18, "learning_rate": 0.00018905055432353552, "loss": 0.6743, "step": 3100 }, { "epoch": 0.18, "learning_rate": 0.0001890421941194263, "loss": 0.7283, "step": 3101 }, { "epoch": 0.18, "learning_rate": 0.00018903383090990953, "loss": 0.7906, "step": 3102 }, { "epoch": 0.18, "learning_rate": 0.00018902546469526743, "loss": 0.8012, "step": 3103 }, { "epoch": 0.18, "learning_rate": 0.00018901709547578245, "loss": 0.7003, "step": 3104 }, { "epoch": 0.18, "learning_rate": 0.00018900872325173707, "loss": 0.7088, "step": 3105 }, { "epoch": 0.18, "learning_rate": 0.00018900034802341378, "loss": 0.7638, "step": 3106 }, { "epoch": 0.18, "learning_rate": 0.00018899196979109536, "loss": 0.7567, "step": 3107 }, { "epoch": 0.18, "learning_rate": 0.0001889835885550646, "loss": 0.7773, "step": 3108 }, { "epoch": 0.18, "learning_rate": 0.00018897520431560434, "loss": 0.8073, "step": 3109 }, { "epoch": 0.18, "learning_rate": 0.00018896681707299755, "loss": 0.7416, "step": 3110 }, { "epoch": 0.18, "learning_rate": 0.0001889584268275274, "loss": 0.7565, "step": 3111 }, { "epoch": 0.18, "learning_rate": 0.00018895003357947705, "loss": 0.7008, "step": 3112 }, { "epoch": 0.18, "learning_rate": 0.00018894163732912977, "loss": 0.7202, "step": 3113 }, { "epoch": 0.18, "learning_rate": 0.00018893323807676898, "loss": 0.7915, "step": 3114 }, { "epoch": 0.18, "learning_rate": 0.00018892483582267814, "loss": 0.6843, "step": 3115 }, { "epoch": 0.18, "learning_rate": 0.00018891643056714093, "loss": 0.7576, "step": 3116 }, { "epoch": 0.18, "learning_rate": 0.00018890802231044096, "loss": 0.7421, "step": 3117 }, { "epoch": 0.18, "learning_rate": 0.00018889961105286206, "loss": 0.77, "step": 3118 }, { "epoch": 0.18, "learning_rate": 0.00018889119679468816, "loss": 0.7962, "step": 3119 }, { "epoch": 0.18, "learning_rate": 0.00018888277953620325, "loss": 0.7362, "step": 3120 }, { "epoch": 0.18, "learning_rate": 0.00018887435927769137, "loss": 0.7658, "step": 3121 }, { "epoch": 0.18, "learning_rate": 0.00018886593601943682, "loss": 0.7643, "step": 3122 }, { "epoch": 0.18, "learning_rate": 0.0001888575097617239, "loss": 0.7554, "step": 3123 }, { "epoch": 0.18, "learning_rate": 0.00018884908050483695, "loss": 0.6762, "step": 3124 }, { "epoch": 0.18, "learning_rate": 0.00018884064824906054, "loss": 0.6147, "step": 3125 }, { "epoch": 0.18, "learning_rate": 0.00018883221299467924, "loss": 0.6445, "step": 3126 }, { "epoch": 0.18, "learning_rate": 0.00018882377474197778, "loss": 0.797, "step": 3127 }, { "epoch": 0.18, "learning_rate": 0.000188815333491241, "loss": 0.676, "step": 3128 }, { "epoch": 0.18, "learning_rate": 0.00018880688924275378, "loss": 0.7821, "step": 3129 }, { "epoch": 0.18, "learning_rate": 0.0001887984419968011, "loss": 0.7757, "step": 3130 }, { "epoch": 0.18, "learning_rate": 0.00018878999175366817, "loss": 0.6863, "step": 3131 }, { "epoch": 0.18, "learning_rate": 0.00018878153851364013, "loss": 0.6476, "step": 3132 }, { "epoch": 0.18, "learning_rate": 0.00018877308227700237, "loss": 0.7565, "step": 3133 }, { "epoch": 0.18, "learning_rate": 0.0001887646230440402, "loss": 0.7479, "step": 3134 }, { "epoch": 0.18, "learning_rate": 0.00018875616081503926, "loss": 0.7717, "step": 3135 }, { "epoch": 0.18, "learning_rate": 0.00018874769559028512, "loss": 0.6371, "step": 3136 }, { "epoch": 0.18, "learning_rate": 0.00018873922737006347, "loss": 0.719, "step": 3137 }, { "epoch": 0.18, "learning_rate": 0.0001887307561546602, "loss": 0.8071, "step": 3138 }, { "epoch": 0.18, "learning_rate": 0.0001887222819443612, "loss": 0.7549, "step": 3139 }, { "epoch": 0.18, "learning_rate": 0.0001887138047394525, "loss": 0.6514, "step": 3140 }, { "epoch": 0.18, "learning_rate": 0.0001887053245402202, "loss": 0.7287, "step": 3141 }, { "epoch": 0.18, "learning_rate": 0.0001886968413469506, "loss": 0.7331, "step": 3142 }, { "epoch": 0.18, "learning_rate": 0.00018868835515992998, "loss": 0.7658, "step": 3143 }, { "epoch": 0.18, "learning_rate": 0.00018867986597944476, "loss": 0.81, "step": 3144 }, { "epoch": 0.18, "learning_rate": 0.0001886713738057815, "loss": 0.7817, "step": 3145 }, { "epoch": 0.18, "learning_rate": 0.00018866287863922684, "loss": 0.7293, "step": 3146 }, { "epoch": 0.18, "learning_rate": 0.00018865438048006747, "loss": 0.7956, "step": 3147 }, { "epoch": 0.18, "learning_rate": 0.00018864587932859028, "loss": 0.7158, "step": 3148 }, { "epoch": 0.18, "learning_rate": 0.00018863737518508218, "loss": 0.7635, "step": 3149 }, { "epoch": 0.18, "learning_rate": 0.0001886288680498302, "loss": 0.7164, "step": 3150 }, { "epoch": 0.18, "learning_rate": 0.00018862035792312147, "loss": 0.7693, "step": 3151 }, { "epoch": 0.18, "learning_rate": 0.00018861184480524327, "loss": 0.7822, "step": 3152 }, { "epoch": 0.18, "learning_rate": 0.00018860332869648292, "loss": 0.7808, "step": 3153 }, { "epoch": 0.18, "learning_rate": 0.0001885948095971278, "loss": 0.7544, "step": 3154 }, { "epoch": 0.18, "learning_rate": 0.0001885862875074656, "loss": 0.6861, "step": 3155 }, { "epoch": 0.18, "learning_rate": 0.0001885777624277838, "loss": 0.7104, "step": 3156 }, { "epoch": 0.18, "learning_rate": 0.00018856923435837022, "loss": 0.7739, "step": 3157 }, { "epoch": 0.18, "learning_rate": 0.00018856070329951273, "loss": 0.7008, "step": 3158 }, { "epoch": 0.18, "learning_rate": 0.00018855216925149922, "loss": 0.749, "step": 3159 }, { "epoch": 0.18, "learning_rate": 0.00018854363221461776, "loss": 0.7778, "step": 3160 }, { "epoch": 0.18, "learning_rate": 0.00018853509218915652, "loss": 0.7967, "step": 3161 }, { "epoch": 0.18, "learning_rate": 0.00018852654917540372, "loss": 0.7422, "step": 3162 }, { "epoch": 0.18, "learning_rate": 0.0001885180031736477, "loss": 0.7347, "step": 3163 }, { "epoch": 0.18, "learning_rate": 0.00018850945418417693, "loss": 0.6763, "step": 3164 }, { "epoch": 0.18, "learning_rate": 0.00018850090220727997, "loss": 0.7853, "step": 3165 }, { "epoch": 0.18, "learning_rate": 0.00018849234724324543, "loss": 0.8651, "step": 3166 }, { "epoch": 0.18, "learning_rate": 0.00018848378929236213, "loss": 0.7606, "step": 3167 }, { "epoch": 0.18, "learning_rate": 0.00018847522835491884, "loss": 0.7652, "step": 3168 }, { "epoch": 0.18, "learning_rate": 0.0001884666644312046, "loss": 0.745, "step": 3169 }, { "epoch": 0.18, "learning_rate": 0.0001884580975215084, "loss": 0.8248, "step": 3170 }, { "epoch": 0.18, "learning_rate": 0.00018844952762611943, "loss": 0.7083, "step": 3171 }, { "epoch": 0.18, "learning_rate": 0.00018844095474532694, "loss": 0.705, "step": 3172 }, { "epoch": 0.18, "learning_rate": 0.00018843237887942026, "loss": 0.7138, "step": 3173 }, { "epoch": 0.18, "learning_rate": 0.0001884238000286889, "loss": 0.7527, "step": 3174 }, { "epoch": 0.18, "learning_rate": 0.00018841521819342236, "loss": 0.6867, "step": 3175 }, { "epoch": 0.18, "learning_rate": 0.00018840663337391032, "loss": 0.8124, "step": 3176 }, { "epoch": 0.18, "learning_rate": 0.0001883980455704426, "loss": 0.8009, "step": 3177 }, { "epoch": 0.18, "learning_rate": 0.00018838945478330898, "loss": 0.7287, "step": 3178 }, { "epoch": 0.18, "learning_rate": 0.00018838086101279945, "loss": 0.8139, "step": 3179 }, { "epoch": 0.18, "learning_rate": 0.00018837226425920406, "loss": 0.7522, "step": 3180 }, { "epoch": 0.18, "learning_rate": 0.000188363664522813, "loss": 0.6827, "step": 3181 }, { "epoch": 0.18, "learning_rate": 0.00018835506180391653, "loss": 0.7641, "step": 3182 }, { "epoch": 0.18, "learning_rate": 0.00018834645610280498, "loss": 0.7011, "step": 3183 }, { "epoch": 0.18, "learning_rate": 0.0001883378474197689, "loss": 0.8073, "step": 3184 }, { "epoch": 0.18, "learning_rate": 0.0001883292357550987, "loss": 0.7442, "step": 3185 }, { "epoch": 0.18, "learning_rate": 0.00018832062110908518, "loss": 0.7237, "step": 3186 }, { "epoch": 0.18, "learning_rate": 0.0001883120034820191, "loss": 0.7618, "step": 3187 }, { "epoch": 0.18, "learning_rate": 0.0001883033828741913, "loss": 0.7618, "step": 3188 }, { "epoch": 0.18, "learning_rate": 0.00018829475928589271, "loss": 0.6914, "step": 3189 }, { "epoch": 0.18, "learning_rate": 0.00018828613271741444, "loss": 0.6915, "step": 3190 }, { "epoch": 0.18, "learning_rate": 0.00018827750316904767, "loss": 0.742, "step": 3191 }, { "epoch": 0.18, "learning_rate": 0.0001882688706410836, "loss": 0.701, "step": 3192 }, { "epoch": 0.18, "learning_rate": 0.0001882602351338137, "loss": 0.7202, "step": 3193 }, { "epoch": 0.18, "learning_rate": 0.00018825159664752938, "loss": 0.6784, "step": 3194 }, { "epoch": 0.18, "learning_rate": 0.00018824295518252222, "loss": 0.8005, "step": 3195 }, { "epoch": 0.18, "learning_rate": 0.00018823431073908392, "loss": 0.7746, "step": 3196 }, { "epoch": 0.18, "learning_rate": 0.0001882256633175062, "loss": 0.723, "step": 3197 }, { "epoch": 0.18, "learning_rate": 0.00018821701291808098, "loss": 0.681, "step": 3198 }, { "epoch": 0.18, "learning_rate": 0.00018820835954110022, "loss": 0.6792, "step": 3199 }, { "epoch": 0.18, "learning_rate": 0.00018819970318685597, "loss": 0.7491, "step": 3200 }, { "epoch": 0.18, "learning_rate": 0.00018819104385564046, "loss": 0.7089, "step": 3201 }, { "epoch": 0.18, "learning_rate": 0.00018818238154774588, "loss": 0.7246, "step": 3202 }, { "epoch": 0.18, "learning_rate": 0.0001881737162634647, "loss": 0.7753, "step": 3203 }, { "epoch": 0.18, "learning_rate": 0.00018816504800308934, "loss": 0.6395, "step": 3204 }, { "epoch": 0.18, "learning_rate": 0.00018815637676691238, "loss": 0.8019, "step": 3205 }, { "epoch": 0.18, "learning_rate": 0.0001881477025552265, "loss": 0.7646, "step": 3206 }, { "epoch": 0.18, "learning_rate": 0.00018813902536832447, "loss": 0.7208, "step": 3207 }, { "epoch": 0.18, "learning_rate": 0.0001881303452064992, "loss": 0.748, "step": 3208 }, { "epoch": 0.18, "learning_rate": 0.00018812166207004367, "loss": 0.7388, "step": 3209 }, { "epoch": 0.18, "learning_rate": 0.00018811297595925096, "loss": 0.75, "step": 3210 }, { "epoch": 0.18, "learning_rate": 0.00018810428687441414, "loss": 0.7437, "step": 3211 }, { "epoch": 0.18, "learning_rate": 0.00018809559481582663, "loss": 0.7584, "step": 3212 }, { "epoch": 0.18, "learning_rate": 0.00018808689978378179, "loss": 0.7456, "step": 3213 }, { "epoch": 0.18, "learning_rate": 0.00018807820177857301, "loss": 0.638, "step": 3214 }, { "epoch": 0.18, "learning_rate": 0.00018806950080049392, "loss": 0.7812, "step": 3215 }, { "epoch": 0.18, "learning_rate": 0.00018806079684983824, "loss": 0.8299, "step": 3216 }, { "epoch": 0.18, "learning_rate": 0.0001880520899268997, "loss": 0.8232, "step": 3217 }, { "epoch": 0.18, "learning_rate": 0.00018804338003197222, "loss": 0.7203, "step": 3218 }, { "epoch": 0.18, "learning_rate": 0.00018803466716534975, "loss": 0.7771, "step": 3219 }, { "epoch": 0.18, "learning_rate": 0.00018802595132732635, "loss": 0.7612, "step": 3220 }, { "epoch": 0.18, "learning_rate": 0.00018801723251819628, "loss": 0.7191, "step": 3221 }, { "epoch": 0.18, "learning_rate": 0.00018800851073825378, "loss": 0.6678, "step": 3222 }, { "epoch": 0.18, "learning_rate": 0.00018799978598779322, "loss": 0.7109, "step": 3223 }, { "epoch": 0.18, "learning_rate": 0.00018799105826710908, "loss": 0.6653, "step": 3224 }, { "epoch": 0.18, "learning_rate": 0.000187982327576496, "loss": 0.7389, "step": 3225 }, { "epoch": 0.18, "learning_rate": 0.0001879735939162486, "loss": 0.7292, "step": 3226 }, { "epoch": 0.18, "learning_rate": 0.00018796485728666165, "loss": 0.7493, "step": 3227 }, { "epoch": 0.18, "learning_rate": 0.00018795611768803008, "loss": 0.7096, "step": 3228 }, { "epoch": 0.18, "learning_rate": 0.0001879473751206489, "loss": 0.7226, "step": 3229 }, { "epoch": 0.18, "learning_rate": 0.00018793862958481317, "loss": 0.7121, "step": 3230 }, { "epoch": 0.18, "learning_rate": 0.00018792988108081801, "loss": 0.7428, "step": 3231 }, { "epoch": 0.18, "learning_rate": 0.0001879211296089588, "loss": 0.7303, "step": 3232 }, { "epoch": 0.18, "learning_rate": 0.0001879123751695309, "loss": 0.7429, "step": 3233 }, { "epoch": 0.18, "learning_rate": 0.00018790361776282972, "loss": 0.7167, "step": 3234 }, { "epoch": 0.18, "learning_rate": 0.00018789485738915094, "loss": 0.7302, "step": 3235 }, { "epoch": 0.18, "learning_rate": 0.00018788609404879022, "loss": 0.7576, "step": 3236 }, { "epoch": 0.18, "learning_rate": 0.0001878773277420433, "loss": 0.8294, "step": 3237 }, { "epoch": 0.18, "learning_rate": 0.00018786855846920615, "loss": 0.7305, "step": 3238 }, { "epoch": 0.18, "learning_rate": 0.00018785978623057467, "loss": 0.7768, "step": 3239 }, { "epoch": 0.18, "learning_rate": 0.000187851011026445, "loss": 0.7359, "step": 3240 }, { "epoch": 0.18, "learning_rate": 0.0001878422328571133, "loss": 0.6791, "step": 3241 }, { "epoch": 0.18, "learning_rate": 0.00018783345172287586, "loss": 0.8077, "step": 3242 }, { "epoch": 0.18, "learning_rate": 0.0001878246676240291, "loss": 0.833, "step": 3243 }, { "epoch": 0.18, "learning_rate": 0.00018781588056086948, "loss": 0.7807, "step": 3244 }, { "epoch": 0.18, "learning_rate": 0.00018780709053369356, "loss": 0.7007, "step": 3245 }, { "epoch": 0.18, "learning_rate": 0.00018779829754279805, "loss": 0.6601, "step": 3246 }, { "epoch": 0.18, "learning_rate": 0.00018778950158847976, "loss": 0.7393, "step": 3247 }, { "epoch": 0.18, "learning_rate": 0.00018778070267103556, "loss": 0.6626, "step": 3248 }, { "epoch": 0.18, "learning_rate": 0.00018777190079076242, "loss": 0.7948, "step": 3249 }, { "epoch": 0.18, "learning_rate": 0.00018776309594795742, "loss": 0.7563, "step": 3250 }, { "epoch": 0.18, "learning_rate": 0.00018775428814291777, "loss": 0.7386, "step": 3251 }, { "epoch": 0.18, "learning_rate": 0.0001877454773759408, "loss": 0.67, "step": 3252 }, { "epoch": 0.18, "learning_rate": 0.0001877366636473238, "loss": 0.7599, "step": 3253 }, { "epoch": 0.18, "learning_rate": 0.0001877278469573643, "loss": 0.7372, "step": 3254 }, { "epoch": 0.18, "learning_rate": 0.0001877190273063599, "loss": 0.7585, "step": 3255 }, { "epoch": 0.18, "learning_rate": 0.0001877102046946083, "loss": 0.7314, "step": 3256 }, { "epoch": 0.18, "learning_rate": 0.00018770137912240724, "loss": 0.6351, "step": 3257 }, { "epoch": 0.18, "learning_rate": 0.0001876925505900546, "loss": 0.7375, "step": 3258 }, { "epoch": 0.18, "learning_rate": 0.00018768371909784845, "loss": 0.7973, "step": 3259 }, { "epoch": 0.18, "learning_rate": 0.00018767488464608678, "loss": 0.7453, "step": 3260 }, { "epoch": 0.18, "learning_rate": 0.00018766604723506784, "loss": 0.6342, "step": 3261 }, { "epoch": 0.19, "learning_rate": 0.00018765720686508993, "loss": 0.7663, "step": 3262 }, { "epoch": 0.19, "learning_rate": 0.00018764836353645135, "loss": 0.7227, "step": 3263 }, { "epoch": 0.19, "learning_rate": 0.00018763951724945062, "loss": 0.6906, "step": 3264 }, { "epoch": 0.19, "learning_rate": 0.00018763066800438636, "loss": 0.7268, "step": 3265 }, { "epoch": 0.19, "learning_rate": 0.00018762181580155723, "loss": 0.7968, "step": 3266 }, { "epoch": 0.19, "learning_rate": 0.00018761296064126205, "loss": 0.7011, "step": 3267 }, { "epoch": 0.19, "learning_rate": 0.00018760410252379966, "loss": 0.6969, "step": 3268 }, { "epoch": 0.19, "learning_rate": 0.00018759524144946904, "loss": 0.7198, "step": 3269 }, { "epoch": 0.19, "learning_rate": 0.0001875863774185693, "loss": 0.6958, "step": 3270 }, { "epoch": 0.19, "learning_rate": 0.00018757751043139962, "loss": 0.7363, "step": 3271 }, { "epoch": 0.19, "learning_rate": 0.00018756864048825932, "loss": 0.5996, "step": 3272 }, { "epoch": 0.19, "learning_rate": 0.0001875597675894477, "loss": 0.6359, "step": 3273 }, { "epoch": 0.19, "learning_rate": 0.0001875508917352643, "loss": 0.5834, "step": 3274 }, { "epoch": 0.19, "learning_rate": 0.0001875420129260087, "loss": 0.6885, "step": 3275 }, { "epoch": 0.19, "learning_rate": 0.00018753313116198056, "loss": 0.683, "step": 3276 }, { "epoch": 0.19, "learning_rate": 0.0001875242464434797, "loss": 0.8059, "step": 3277 }, { "epoch": 0.19, "learning_rate": 0.00018751535877080597, "loss": 0.7315, "step": 3278 }, { "epoch": 0.19, "learning_rate": 0.00018750646814425938, "loss": 0.7866, "step": 3279 }, { "epoch": 0.19, "learning_rate": 0.00018749757456413998, "loss": 0.7417, "step": 3280 }, { "epoch": 0.19, "learning_rate": 0.00018748867803074798, "loss": 0.69, "step": 3281 }, { "epoch": 0.19, "learning_rate": 0.00018747977854438364, "loss": 0.7051, "step": 3282 }, { "epoch": 0.19, "learning_rate": 0.00018747087610534736, "loss": 0.8138, "step": 3283 }, { "epoch": 0.19, "learning_rate": 0.00018746197071393958, "loss": 0.7692, "step": 3284 }, { "epoch": 0.19, "learning_rate": 0.00018745306237046095, "loss": 0.7298, "step": 3285 }, { "epoch": 0.19, "learning_rate": 0.00018744415107521206, "loss": 0.8345, "step": 3286 }, { "epoch": 0.19, "learning_rate": 0.0001874352368284938, "loss": 0.6942, "step": 3287 }, { "epoch": 0.19, "learning_rate": 0.00018742631963060698, "loss": 0.7687, "step": 3288 }, { "epoch": 0.19, "learning_rate": 0.00018741739948185257, "loss": 0.6935, "step": 3289 }, { "epoch": 0.19, "learning_rate": 0.00018740847638253167, "loss": 0.6307, "step": 3290 }, { "epoch": 0.19, "learning_rate": 0.00018739955033294546, "loss": 0.6929, "step": 3291 }, { "epoch": 0.19, "learning_rate": 0.00018739062133339517, "loss": 0.7442, "step": 3292 }, { "epoch": 0.19, "learning_rate": 0.00018738168938418228, "loss": 0.7116, "step": 3293 }, { "epoch": 0.19, "learning_rate": 0.00018737275448560818, "loss": 0.6428, "step": 3294 }, { "epoch": 0.19, "learning_rate": 0.0001873638166379745, "loss": 0.7345, "step": 3295 }, { "epoch": 0.19, "learning_rate": 0.00018735487584158283, "loss": 0.7799, "step": 3296 }, { "epoch": 0.19, "learning_rate": 0.00018734593209673506, "loss": 0.7714, "step": 3297 }, { "epoch": 0.19, "learning_rate": 0.00018733698540373297, "loss": 0.69, "step": 3298 }, { "epoch": 0.19, "learning_rate": 0.0001873280357628786, "loss": 0.6955, "step": 3299 }, { "epoch": 0.19, "learning_rate": 0.000187319083174474, "loss": 0.6931, "step": 3300 }, { "epoch": 0.19, "learning_rate": 0.00018731012763882133, "loss": 0.7231, "step": 3301 }, { "epoch": 0.19, "learning_rate": 0.00018730116915622284, "loss": 0.7558, "step": 3302 }, { "epoch": 0.19, "learning_rate": 0.00018729220772698097, "loss": 0.8173, "step": 3303 }, { "epoch": 0.19, "learning_rate": 0.00018728324335139814, "loss": 0.8047, "step": 3304 }, { "epoch": 0.19, "learning_rate": 0.000187274276029777, "loss": 0.7766, "step": 3305 }, { "epoch": 0.19, "learning_rate": 0.00018726530576242007, "loss": 0.7133, "step": 3306 }, { "epoch": 0.19, "learning_rate": 0.00018725633254963023, "loss": 0.7335, "step": 3307 }, { "epoch": 0.19, "learning_rate": 0.00018724735639171036, "loss": 0.726, "step": 3308 }, { "epoch": 0.19, "learning_rate": 0.0001872383772889634, "loss": 0.6306, "step": 3309 }, { "epoch": 0.19, "learning_rate": 0.0001872293952416924, "loss": 0.6538, "step": 3310 }, { "epoch": 0.19, "learning_rate": 0.00018722041025020055, "loss": 0.7432, "step": 3311 }, { "epoch": 0.19, "learning_rate": 0.0001872114223147911, "loss": 0.9077, "step": 3312 }, { "epoch": 0.19, "learning_rate": 0.00018720243143576743, "loss": 0.6078, "step": 3313 }, { "epoch": 0.19, "learning_rate": 0.000187193437613433, "loss": 0.7591, "step": 3314 }, { "epoch": 0.19, "learning_rate": 0.00018718444084809136, "loss": 0.7212, "step": 3315 }, { "epoch": 0.19, "learning_rate": 0.00018717544114004624, "loss": 0.7457, "step": 3316 }, { "epoch": 0.19, "learning_rate": 0.0001871664384896013, "loss": 0.7413, "step": 3317 }, { "epoch": 0.19, "learning_rate": 0.00018715743289706048, "loss": 0.776, "step": 3318 }, { "epoch": 0.19, "learning_rate": 0.00018714842436272773, "loss": 0.677, "step": 3319 }, { "epoch": 0.19, "learning_rate": 0.0001871394128869071, "loss": 0.7279, "step": 3320 }, { "epoch": 0.19, "learning_rate": 0.00018713039846990275, "loss": 0.8154, "step": 3321 }, { "epoch": 0.19, "learning_rate": 0.00018712138111201895, "loss": 0.7046, "step": 3322 }, { "epoch": 0.19, "learning_rate": 0.00018711236081356004, "loss": 0.6985, "step": 3323 }, { "epoch": 0.19, "learning_rate": 0.00018710333757483052, "loss": 0.7823, "step": 3324 }, { "epoch": 0.19, "learning_rate": 0.00018709431139613488, "loss": 0.7233, "step": 3325 }, { "epoch": 0.19, "learning_rate": 0.00018708528227777787, "loss": 0.8058, "step": 3326 }, { "epoch": 0.19, "learning_rate": 0.00018707625022006415, "loss": 0.7389, "step": 3327 }, { "epoch": 0.19, "learning_rate": 0.00018706721522329862, "loss": 0.7933, "step": 3328 }, { "epoch": 0.19, "learning_rate": 0.00018705817728778624, "loss": 0.6401, "step": 3329 }, { "epoch": 0.19, "learning_rate": 0.00018704913641383208, "loss": 0.7994, "step": 3330 }, { "epoch": 0.19, "learning_rate": 0.00018704009260174124, "loss": 0.7784, "step": 3331 }, { "epoch": 0.19, "learning_rate": 0.000187031045851819, "loss": 0.7686, "step": 3332 }, { "epoch": 0.19, "learning_rate": 0.00018702199616437074, "loss": 0.8853, "step": 3333 }, { "epoch": 0.19, "learning_rate": 0.0001870129435397019, "loss": 0.705, "step": 3334 }, { "epoch": 0.19, "learning_rate": 0.000187003887978118, "loss": 0.7532, "step": 3335 }, { "epoch": 0.19, "learning_rate": 0.0001869948294799247, "loss": 0.8056, "step": 3336 }, { "epoch": 0.19, "learning_rate": 0.00018698576804542777, "loss": 0.7469, "step": 3337 }, { "epoch": 0.19, "learning_rate": 0.00018697670367493299, "loss": 0.6622, "step": 3338 }, { "epoch": 0.19, "learning_rate": 0.0001869676363687464, "loss": 0.7545, "step": 3339 }, { "epoch": 0.19, "learning_rate": 0.00018695856612717398, "loss": 0.6326, "step": 3340 }, { "epoch": 0.19, "learning_rate": 0.0001869494929505219, "loss": 0.7715, "step": 3341 }, { "epoch": 0.19, "learning_rate": 0.0001869404168390964, "loss": 0.7607, "step": 3342 }, { "epoch": 0.19, "learning_rate": 0.00018693133779320385, "loss": 0.7803, "step": 3343 }, { "epoch": 0.19, "learning_rate": 0.00018692225581315065, "loss": 0.6072, "step": 3344 }, { "epoch": 0.19, "learning_rate": 0.00018691317089924333, "loss": 0.7335, "step": 3345 }, { "epoch": 0.19, "learning_rate": 0.00018690408305178856, "loss": 0.7238, "step": 3346 }, { "epoch": 0.19, "learning_rate": 0.0001868949922710931, "loss": 0.763, "step": 3347 }, { "epoch": 0.19, "learning_rate": 0.00018688589855746376, "loss": 0.6519, "step": 3348 }, { "epoch": 0.19, "learning_rate": 0.00018687680191120743, "loss": 0.5769, "step": 3349 }, { "epoch": 0.19, "learning_rate": 0.00018686770233263122, "loss": 0.7881, "step": 3350 }, { "epoch": 0.19, "learning_rate": 0.00018685859982204225, "loss": 0.792, "step": 3351 }, { "epoch": 0.19, "learning_rate": 0.00018684949437974772, "loss": 0.7348, "step": 3352 }, { "epoch": 0.19, "learning_rate": 0.000186840386006055, "loss": 0.8653, "step": 3353 }, { "epoch": 0.19, "learning_rate": 0.0001868312747012715, "loss": 0.6925, "step": 3354 }, { "epoch": 0.19, "learning_rate": 0.00018682216046570475, "loss": 0.7685, "step": 3355 }, { "epoch": 0.19, "learning_rate": 0.0001868130432996624, "loss": 0.652, "step": 3356 }, { "epoch": 0.19, "learning_rate": 0.00018680392320345214, "loss": 0.7266, "step": 3357 }, { "epoch": 0.19, "learning_rate": 0.00018679480017738185, "loss": 0.7905, "step": 3358 }, { "epoch": 0.19, "learning_rate": 0.00018678567422175942, "loss": 0.8116, "step": 3359 }, { "epoch": 0.19, "learning_rate": 0.00018677654533689287, "loss": 0.7405, "step": 3360 }, { "epoch": 0.19, "learning_rate": 0.00018676741352309032, "loss": 0.757, "step": 3361 }, { "epoch": 0.19, "learning_rate": 0.00018675827878066002, "loss": 0.7897, "step": 3362 }, { "epoch": 0.19, "learning_rate": 0.0001867491411099103, "loss": 0.7904, "step": 3363 }, { "epoch": 0.19, "learning_rate": 0.00018674000051114952, "loss": 0.7938, "step": 3364 }, { "epoch": 0.19, "learning_rate": 0.00018673085698468625, "loss": 0.7007, "step": 3365 }, { "epoch": 0.19, "learning_rate": 0.00018672171053082912, "loss": 0.8305, "step": 3366 }, { "epoch": 0.19, "learning_rate": 0.00018671256114988682, "loss": 0.73, "step": 3367 }, { "epoch": 0.19, "learning_rate": 0.00018670340884216816, "loss": 0.6832, "step": 3368 }, { "epoch": 0.19, "learning_rate": 0.00018669425360798205, "loss": 0.639, "step": 3369 }, { "epoch": 0.19, "learning_rate": 0.00018668509544763752, "loss": 0.7375, "step": 3370 }, { "epoch": 0.19, "learning_rate": 0.0001866759343614437, "loss": 0.7296, "step": 3371 }, { "epoch": 0.19, "learning_rate": 0.0001866667703497098, "loss": 0.7837, "step": 3372 }, { "epoch": 0.19, "learning_rate": 0.00018665760341274505, "loss": 0.7384, "step": 3373 }, { "epoch": 0.19, "learning_rate": 0.00018664843355085896, "loss": 0.7761, "step": 3374 }, { "epoch": 0.19, "learning_rate": 0.00018663926076436095, "loss": 0.7596, "step": 3375 }, { "epoch": 0.19, "learning_rate": 0.00018663008505356073, "loss": 0.7562, "step": 3376 }, { "epoch": 0.19, "learning_rate": 0.0001866209064187679, "loss": 0.5796, "step": 3377 }, { "epoch": 0.19, "learning_rate": 0.00018661172486029233, "loss": 0.7451, "step": 3378 }, { "epoch": 0.19, "learning_rate": 0.00018660254037844388, "loss": 0.584, "step": 3379 }, { "epoch": 0.19, "learning_rate": 0.00018659335297353257, "loss": 0.7997, "step": 3380 }, { "epoch": 0.19, "learning_rate": 0.0001865841626458685, "loss": 0.7466, "step": 3381 }, { "epoch": 0.19, "learning_rate": 0.00018657496939576186, "loss": 0.7229, "step": 3382 }, { "epoch": 0.19, "learning_rate": 0.00018656577322352294, "loss": 0.6874, "step": 3383 }, { "epoch": 0.19, "learning_rate": 0.0001865565741294622, "loss": 0.6209, "step": 3384 }, { "epoch": 0.19, "learning_rate": 0.00018654737211389004, "loss": 0.753, "step": 3385 }, { "epoch": 0.19, "learning_rate": 0.0001865381671771171, "loss": 0.7535, "step": 3386 }, { "epoch": 0.19, "learning_rate": 0.00018652895931945408, "loss": 0.7919, "step": 3387 }, { "epoch": 0.19, "learning_rate": 0.00018651974854121172, "loss": 0.6639, "step": 3388 }, { "epoch": 0.19, "learning_rate": 0.00018651053484270095, "loss": 0.7585, "step": 3389 }, { "epoch": 0.19, "learning_rate": 0.00018650131822423277, "loss": 0.7878, "step": 3390 }, { "epoch": 0.19, "learning_rate": 0.0001864920986861182, "loss": 0.7468, "step": 3391 }, { "epoch": 0.19, "learning_rate": 0.0001864828762286685, "loss": 0.6994, "step": 3392 }, { "epoch": 0.19, "learning_rate": 0.00018647365085219494, "loss": 0.7345, "step": 3393 }, { "epoch": 0.19, "learning_rate": 0.00018646442255700882, "loss": 0.7414, "step": 3394 }, { "epoch": 0.19, "learning_rate": 0.0001864551913434217, "loss": 0.8047, "step": 3395 }, { "epoch": 0.19, "learning_rate": 0.00018644595721174512, "loss": 0.7181, "step": 3396 }, { "epoch": 0.19, "learning_rate": 0.00018643672016229078, "loss": 0.7718, "step": 3397 }, { "epoch": 0.19, "learning_rate": 0.0001864274801953705, "loss": 0.7044, "step": 3398 }, { "epoch": 0.19, "learning_rate": 0.00018641823731129604, "loss": 0.6855, "step": 3399 }, { "epoch": 0.19, "learning_rate": 0.00018640899151037944, "loss": 0.6797, "step": 3400 }, { "epoch": 0.19, "learning_rate": 0.00018639974279293278, "loss": 0.7669, "step": 3401 }, { "epoch": 0.19, "learning_rate": 0.00018639049115926816, "loss": 0.7759, "step": 3402 }, { "epoch": 0.19, "learning_rate": 0.00018638123660969796, "loss": 0.7841, "step": 3403 }, { "epoch": 0.19, "learning_rate": 0.00018637197914453445, "loss": 0.8111, "step": 3404 }, { "epoch": 0.19, "learning_rate": 0.00018636271876409013, "loss": 0.7474, "step": 3405 }, { "epoch": 0.19, "learning_rate": 0.00018635345546867753, "loss": 0.7572, "step": 3406 }, { "epoch": 0.19, "learning_rate": 0.00018634418925860937, "loss": 0.7438, "step": 3407 }, { "epoch": 0.19, "learning_rate": 0.00018633492013419838, "loss": 0.7401, "step": 3408 }, { "epoch": 0.19, "learning_rate": 0.00018632564809575742, "loss": 0.7993, "step": 3409 }, { "epoch": 0.19, "learning_rate": 0.0001863163731435994, "loss": 0.6793, "step": 3410 }, { "epoch": 0.19, "learning_rate": 0.00018630709527803742, "loss": 0.7866, "step": 3411 }, { "epoch": 0.19, "learning_rate": 0.00018629781449938464, "loss": 0.6558, "step": 3412 }, { "epoch": 0.19, "learning_rate": 0.00018628853080795432, "loss": 0.7035, "step": 3413 }, { "epoch": 0.19, "learning_rate": 0.00018627924420405975, "loss": 0.7155, "step": 3414 }, { "epoch": 0.19, "learning_rate": 0.00018626995468801444, "loss": 0.7107, "step": 3415 }, { "epoch": 0.19, "learning_rate": 0.00018626066226013187, "loss": 0.6766, "step": 3416 }, { "epoch": 0.19, "learning_rate": 0.00018625136692072575, "loss": 0.6694, "step": 3417 }, { "epoch": 0.19, "learning_rate": 0.0001862420686701098, "loss": 0.6597, "step": 3418 }, { "epoch": 0.19, "learning_rate": 0.00018623276750859784, "loss": 0.7091, "step": 3419 }, { "epoch": 0.19, "learning_rate": 0.00018622346343650384, "loss": 0.7212, "step": 3420 }, { "epoch": 0.19, "learning_rate": 0.0001862141564541418, "loss": 0.7378, "step": 3421 }, { "epoch": 0.19, "learning_rate": 0.00018620484656182588, "loss": 0.7127, "step": 3422 }, { "epoch": 0.19, "learning_rate": 0.0001861955337598703, "loss": 0.6755, "step": 3423 }, { "epoch": 0.19, "learning_rate": 0.00018618621804858943, "loss": 0.8171, "step": 3424 }, { "epoch": 0.19, "learning_rate": 0.0001861768994282976, "loss": 0.737, "step": 3425 }, { "epoch": 0.19, "learning_rate": 0.00018616757789930948, "loss": 0.6945, "step": 3426 }, { "epoch": 0.19, "learning_rate": 0.0001861582534619396, "loss": 0.7207, "step": 3427 }, { "epoch": 0.19, "learning_rate": 0.0001861489261165027, "loss": 0.6832, "step": 3428 }, { "epoch": 0.19, "learning_rate": 0.00018613959586331362, "loss": 0.7451, "step": 3429 }, { "epoch": 0.19, "learning_rate": 0.00018613026270268725, "loss": 0.6353, "step": 3430 }, { "epoch": 0.19, "learning_rate": 0.00018612092663493866, "loss": 0.6931, "step": 3431 }, { "epoch": 0.19, "learning_rate": 0.0001861115876603829, "loss": 0.7154, "step": 3432 }, { "epoch": 0.19, "learning_rate": 0.00018610224577933523, "loss": 0.609, "step": 3433 }, { "epoch": 0.19, "learning_rate": 0.00018609290099211098, "loss": 0.7881, "step": 3434 }, { "epoch": 0.19, "learning_rate": 0.00018608355329902555, "loss": 0.6903, "step": 3435 }, { "epoch": 0.19, "learning_rate": 0.0001860742027003944, "loss": 0.7516, "step": 3436 }, { "epoch": 0.19, "learning_rate": 0.00018606484919653316, "loss": 0.8096, "step": 3437 }, { "epoch": 0.2, "learning_rate": 0.00018605549278775753, "loss": 0.7087, "step": 3438 }, { "epoch": 0.2, "learning_rate": 0.0001860461334743834, "loss": 0.6856, "step": 3439 }, { "epoch": 0.2, "learning_rate": 0.00018603677125672653, "loss": 0.8874, "step": 3440 }, { "epoch": 0.2, "learning_rate": 0.00018602740613510302, "loss": 0.7111, "step": 3441 }, { "epoch": 0.2, "learning_rate": 0.00018601803810982894, "loss": 0.8737, "step": 3442 }, { "epoch": 0.2, "learning_rate": 0.0001860086671812205, "loss": 0.7366, "step": 3443 }, { "epoch": 0.2, "learning_rate": 0.00018599929334959394, "loss": 0.8125, "step": 3444 }, { "epoch": 0.2, "learning_rate": 0.00018598991661526572, "loss": 0.727, "step": 3445 }, { "epoch": 0.2, "learning_rate": 0.00018598053697855226, "loss": 0.7194, "step": 3446 }, { "epoch": 0.2, "learning_rate": 0.0001859711544397702, "loss": 0.7465, "step": 3447 }, { "epoch": 0.2, "learning_rate": 0.00018596176899923625, "loss": 0.6706, "step": 3448 }, { "epoch": 0.2, "learning_rate": 0.00018595238065726713, "loss": 0.6944, "step": 3449 }, { "epoch": 0.2, "learning_rate": 0.0001859429894141797, "loss": 0.6288, "step": 3450 }, { "epoch": 0.2, "learning_rate": 0.00018593359527029104, "loss": 0.7608, "step": 3451 }, { "epoch": 0.2, "learning_rate": 0.00018592419822591815, "loss": 0.7465, "step": 3452 }, { "epoch": 0.2, "learning_rate": 0.0001859147982813782, "loss": 0.6793, "step": 3453 }, { "epoch": 0.2, "learning_rate": 0.00018590539543698854, "loss": 0.769, "step": 3454 }, { "epoch": 0.2, "learning_rate": 0.00018589598969306645, "loss": 0.6955, "step": 3455 }, { "epoch": 0.2, "learning_rate": 0.00018588658104992946, "loss": 0.8022, "step": 3456 }, { "epoch": 0.2, "learning_rate": 0.0001858771695078951, "loss": 0.7739, "step": 3457 }, { "epoch": 0.2, "learning_rate": 0.00018586775506728104, "loss": 0.7827, "step": 3458 }, { "epoch": 0.2, "learning_rate": 0.00018585833772840506, "loss": 0.774, "step": 3459 }, { "epoch": 0.2, "learning_rate": 0.00018584891749158503, "loss": 0.6964, "step": 3460 }, { "epoch": 0.2, "learning_rate": 0.00018583949435713886, "loss": 0.6953, "step": 3461 }, { "epoch": 0.2, "learning_rate": 0.00018583006832538463, "loss": 0.664, "step": 3462 }, { "epoch": 0.2, "learning_rate": 0.0001858206393966405, "loss": 0.6541, "step": 3463 }, { "epoch": 0.2, "learning_rate": 0.00018581120757122475, "loss": 0.6807, "step": 3464 }, { "epoch": 0.2, "learning_rate": 0.00018580177284945566, "loss": 0.6267, "step": 3465 }, { "epoch": 0.2, "learning_rate": 0.00018579233523165174, "loss": 0.7393, "step": 3466 }, { "epoch": 0.2, "learning_rate": 0.00018578289471813148, "loss": 0.7127, "step": 3467 }, { "epoch": 0.2, "learning_rate": 0.00018577345130921358, "loss": 0.7664, "step": 3468 }, { "epoch": 0.2, "learning_rate": 0.00018576400500521672, "loss": 0.7797, "step": 3469 }, { "epoch": 0.2, "learning_rate": 0.00018575455580645977, "loss": 0.7451, "step": 3470 }, { "epoch": 0.2, "learning_rate": 0.00018574510371326168, "loss": 0.7441, "step": 3471 }, { "epoch": 0.2, "learning_rate": 0.00018573564872594145, "loss": 0.8053, "step": 3472 }, { "epoch": 0.2, "learning_rate": 0.0001857261908448182, "loss": 0.6919, "step": 3473 }, { "epoch": 0.2, "learning_rate": 0.00018571673007021123, "loss": 0.7191, "step": 3474 }, { "epoch": 0.2, "learning_rate": 0.00018570726640243984, "loss": 0.7386, "step": 3475 }, { "epoch": 0.2, "learning_rate": 0.00018569779984182336, "loss": 0.7126, "step": 3476 }, { "epoch": 0.2, "learning_rate": 0.00018568833038868143, "loss": 0.6186, "step": 3477 }, { "epoch": 0.2, "learning_rate": 0.00018567885804333362, "loss": 0.7222, "step": 3478 }, { "epoch": 0.2, "learning_rate": 0.00018566938280609966, "loss": 0.8007, "step": 3479 }, { "epoch": 0.2, "learning_rate": 0.00018565990467729932, "loss": 0.6701, "step": 3480 }, { "epoch": 0.2, "learning_rate": 0.00018565042365725258, "loss": 0.7165, "step": 3481 }, { "epoch": 0.2, "learning_rate": 0.00018564093974627937, "loss": 0.745, "step": 3482 }, { "epoch": 0.2, "learning_rate": 0.0001856314529446999, "loss": 0.7207, "step": 3483 }, { "epoch": 0.2, "learning_rate": 0.00018562196325283428, "loss": 0.7065, "step": 3484 }, { "epoch": 0.2, "learning_rate": 0.00018561247067100286, "loss": 0.7007, "step": 3485 }, { "epoch": 0.2, "learning_rate": 0.00018560297519952602, "loss": 0.7216, "step": 3486 }, { "epoch": 0.2, "learning_rate": 0.00018559347683872428, "loss": 0.7349, "step": 3487 }, { "epoch": 0.2, "learning_rate": 0.0001855839755889182, "loss": 0.8454, "step": 3488 }, { "epoch": 0.2, "learning_rate": 0.0001855744714504285, "loss": 0.6351, "step": 3489 }, { "epoch": 0.2, "learning_rate": 0.00018556496442357595, "loss": 0.6262, "step": 3490 }, { "epoch": 0.2, "learning_rate": 0.00018555545450868146, "loss": 0.7255, "step": 3491 }, { "epoch": 0.2, "learning_rate": 0.000185545941706066, "loss": 0.5939, "step": 3492 }, { "epoch": 0.2, "learning_rate": 0.00018553642601605068, "loss": 0.6307, "step": 3493 }, { "epoch": 0.2, "learning_rate": 0.00018552690743895662, "loss": 0.7285, "step": 3494 }, { "epoch": 0.2, "learning_rate": 0.00018551738597510515, "loss": 0.7679, "step": 3495 }, { "epoch": 0.2, "learning_rate": 0.00018550786162481765, "loss": 0.6126, "step": 3496 }, { "epoch": 0.2, "learning_rate": 0.0001854983343884155, "loss": 0.7041, "step": 3497 }, { "epoch": 0.2, "learning_rate": 0.0001854888042662204, "loss": 0.8603, "step": 3498 }, { "epoch": 0.2, "learning_rate": 0.0001854792712585539, "loss": 0.7459, "step": 3499 }, { "epoch": 0.2, "learning_rate": 0.00018546973536573785, "loss": 0.7879, "step": 3500 }, { "epoch": 0.2, "learning_rate": 0.0001854601965880941, "loss": 0.6816, "step": 3501 }, { "epoch": 0.2, "learning_rate": 0.00018545065492594454, "loss": 0.76, "step": 3502 }, { "epoch": 0.2, "learning_rate": 0.0001854411103796113, "loss": 0.7169, "step": 3503 }, { "epoch": 0.2, "learning_rate": 0.0001854315629494165, "loss": 0.7098, "step": 3504 }, { "epoch": 0.2, "learning_rate": 0.0001854220126356824, "loss": 0.7318, "step": 3505 }, { "epoch": 0.2, "learning_rate": 0.00018541245943873134, "loss": 0.75, "step": 3506 }, { "epoch": 0.2, "learning_rate": 0.00018540290335888578, "loss": 0.757, "step": 3507 }, { "epoch": 0.2, "learning_rate": 0.00018539334439646824, "loss": 0.7189, "step": 3508 }, { "epoch": 0.2, "learning_rate": 0.0001853837825518014, "loss": 0.7944, "step": 3509 }, { "epoch": 0.2, "learning_rate": 0.00018537421782520796, "loss": 0.7043, "step": 3510 }, { "epoch": 0.2, "learning_rate": 0.00018536465021701078, "loss": 0.727, "step": 3511 }, { "epoch": 0.2, "learning_rate": 0.00018535507972753274, "loss": 0.7345, "step": 3512 }, { "epoch": 0.2, "learning_rate": 0.00018534550635709694, "loss": 0.7929, "step": 3513 }, { "epoch": 0.2, "learning_rate": 0.00018533593010602645, "loss": 0.845, "step": 3514 }, { "epoch": 0.2, "learning_rate": 0.00018532635097464454, "loss": 0.7356, "step": 3515 }, { "epoch": 0.2, "learning_rate": 0.00018531676896327452, "loss": 0.724, "step": 3516 }, { "epoch": 0.2, "learning_rate": 0.00018530718407223974, "loss": 0.6545, "step": 3517 }, { "epoch": 0.2, "learning_rate": 0.00018529759630186383, "loss": 0.8269, "step": 3518 }, { "epoch": 0.2, "learning_rate": 0.0001852880056524703, "loss": 0.8623, "step": 3519 }, { "epoch": 0.2, "learning_rate": 0.00018527841212438294, "loss": 0.7958, "step": 3520 }, { "epoch": 0.2, "learning_rate": 0.00018526881571792551, "loss": 0.6766, "step": 3521 }, { "epoch": 0.2, "learning_rate": 0.00018525921643342192, "loss": 0.6547, "step": 3522 }, { "epoch": 0.2, "learning_rate": 0.00018524961427119615, "loss": 0.7577, "step": 3523 }, { "epoch": 0.2, "learning_rate": 0.00018524000923157236, "loss": 0.8352, "step": 3524 }, { "epoch": 0.2, "learning_rate": 0.00018523040131487468, "loss": 0.7499, "step": 3525 }, { "epoch": 0.2, "learning_rate": 0.00018522079052142747, "loss": 0.6669, "step": 3526 }, { "epoch": 0.2, "learning_rate": 0.00018521117685155503, "loss": 0.6328, "step": 3527 }, { "epoch": 0.2, "learning_rate": 0.00018520156030558192, "loss": 0.693, "step": 3528 }, { "epoch": 0.2, "learning_rate": 0.00018519194088383273, "loss": 0.6192, "step": 3529 }, { "epoch": 0.2, "learning_rate": 0.00018518231858663205, "loss": 0.7407, "step": 3530 }, { "epoch": 0.2, "learning_rate": 0.00018517269341430476, "loss": 0.7192, "step": 3531 }, { "epoch": 0.2, "learning_rate": 0.0001851630653671757, "loss": 0.7681, "step": 3532 }, { "epoch": 0.2, "learning_rate": 0.00018515343444556982, "loss": 0.7534, "step": 3533 }, { "epoch": 0.2, "learning_rate": 0.0001851438006498122, "loss": 0.7461, "step": 3534 }, { "epoch": 0.2, "learning_rate": 0.00018513416398022802, "loss": 0.8462, "step": 3535 }, { "epoch": 0.2, "learning_rate": 0.00018512452443714253, "loss": 0.7279, "step": 3536 }, { "epoch": 0.2, "learning_rate": 0.00018511488202088109, "loss": 0.7613, "step": 3537 }, { "epoch": 0.2, "learning_rate": 0.00018510523673176917, "loss": 0.7619, "step": 3538 }, { "epoch": 0.2, "learning_rate": 0.0001850955885701323, "loss": 0.8178, "step": 3539 }, { "epoch": 0.2, "learning_rate": 0.00018508593753629616, "loss": 0.6944, "step": 3540 }, { "epoch": 0.2, "learning_rate": 0.00018507628363058647, "loss": 0.6231, "step": 3541 }, { "epoch": 0.2, "learning_rate": 0.00018506662685332907, "loss": 0.6915, "step": 3542 }, { "epoch": 0.2, "learning_rate": 0.00018505696720484994, "loss": 0.7015, "step": 3543 }, { "epoch": 0.2, "learning_rate": 0.0001850473046854751, "loss": 0.7267, "step": 3544 }, { "epoch": 0.2, "learning_rate": 0.0001850376392955307, "loss": 0.7686, "step": 3545 }, { "epoch": 0.2, "learning_rate": 0.00018502797103534292, "loss": 0.6759, "step": 3546 }, { "epoch": 0.2, "learning_rate": 0.0001850182999052381, "loss": 0.725, "step": 3547 }, { "epoch": 0.2, "learning_rate": 0.00018500862590554274, "loss": 0.8211, "step": 3548 }, { "epoch": 0.2, "learning_rate": 0.00018499894903658328, "loss": 0.7038, "step": 3549 }, { "epoch": 0.2, "learning_rate": 0.00018498926929868642, "loss": 0.7341, "step": 3550 }, { "epoch": 0.2, "learning_rate": 0.0001849795866921788, "loss": 0.7434, "step": 3551 }, { "epoch": 0.2, "learning_rate": 0.00018496990121738725, "loss": 0.6483, "step": 3552 }, { "epoch": 0.2, "learning_rate": 0.0001849602128746387, "loss": 0.7376, "step": 3553 }, { "epoch": 0.2, "learning_rate": 0.00018495052166426015, "loss": 0.7177, "step": 3554 }, { "epoch": 0.2, "learning_rate": 0.0001849408275865787, "loss": 0.7675, "step": 3555 }, { "epoch": 0.2, "learning_rate": 0.00018493113064192157, "loss": 0.6625, "step": 3556 }, { "epoch": 0.2, "learning_rate": 0.00018492143083061603, "loss": 0.6585, "step": 3557 }, { "epoch": 0.2, "learning_rate": 0.00018491172815298946, "loss": 0.7683, "step": 3558 }, { "epoch": 0.2, "learning_rate": 0.0001849020226093694, "loss": 0.7429, "step": 3559 }, { "epoch": 0.2, "learning_rate": 0.00018489231420008338, "loss": 0.6472, "step": 3560 }, { "epoch": 0.2, "learning_rate": 0.00018488260292545916, "loss": 0.6794, "step": 3561 }, { "epoch": 0.2, "learning_rate": 0.00018487288878582447, "loss": 0.69, "step": 3562 }, { "epoch": 0.2, "learning_rate": 0.0001848631717815072, "loss": 0.6659, "step": 3563 }, { "epoch": 0.2, "learning_rate": 0.00018485345191283532, "loss": 0.7336, "step": 3564 }, { "epoch": 0.2, "learning_rate": 0.0001848437291801369, "loss": 0.7626, "step": 3565 }, { "epoch": 0.2, "learning_rate": 0.0001848340035837401, "loss": 0.7309, "step": 3566 }, { "epoch": 0.2, "learning_rate": 0.00018482427512397318, "loss": 0.7426, "step": 3567 }, { "epoch": 0.2, "learning_rate": 0.00018481454380116455, "loss": 0.8138, "step": 3568 }, { "epoch": 0.2, "learning_rate": 0.0001848048096156426, "loss": 0.7269, "step": 3569 }, { "epoch": 0.2, "learning_rate": 0.00018479507256773594, "loss": 0.737, "step": 3570 }, { "epoch": 0.2, "learning_rate": 0.00018478533265777318, "loss": 0.7555, "step": 3571 }, { "epoch": 0.2, "learning_rate": 0.0001847755898860831, "loss": 0.8503, "step": 3572 }, { "epoch": 0.2, "learning_rate": 0.00018476584425299454, "loss": 0.7771, "step": 3573 }, { "epoch": 0.2, "learning_rate": 0.00018475609575883638, "loss": 0.6634, "step": 3574 }, { "epoch": 0.2, "learning_rate": 0.00018474634440393778, "loss": 0.7548, "step": 3575 }, { "epoch": 0.2, "learning_rate": 0.00018473659018862775, "loss": 0.7423, "step": 3576 }, { "epoch": 0.2, "learning_rate": 0.00018472683311323556, "loss": 0.6919, "step": 3577 }, { "epoch": 0.2, "learning_rate": 0.00018471707317809058, "loss": 0.6837, "step": 3578 }, { "epoch": 0.2, "learning_rate": 0.0001847073103835222, "loss": 0.7437, "step": 3579 }, { "epoch": 0.2, "learning_rate": 0.00018469754472985993, "loss": 0.7419, "step": 3580 }, { "epoch": 0.2, "learning_rate": 0.0001846877762174334, "loss": 0.7486, "step": 3581 }, { "epoch": 0.2, "learning_rate": 0.00018467800484657233, "loss": 0.6514, "step": 3582 }, { "epoch": 0.2, "learning_rate": 0.00018466823061760653, "loss": 0.6694, "step": 3583 }, { "epoch": 0.2, "learning_rate": 0.00018465845353086586, "loss": 0.707, "step": 3584 }, { "epoch": 0.2, "learning_rate": 0.00018464867358668037, "loss": 0.7197, "step": 3585 }, { "epoch": 0.2, "learning_rate": 0.00018463889078538018, "loss": 0.7473, "step": 3586 }, { "epoch": 0.2, "learning_rate": 0.00018462910512729542, "loss": 0.6672, "step": 3587 }, { "epoch": 0.2, "learning_rate": 0.00018461931661275643, "loss": 0.7361, "step": 3588 }, { "epoch": 0.2, "learning_rate": 0.00018460952524209355, "loss": 0.7514, "step": 3589 }, { "epoch": 0.2, "learning_rate": 0.00018459973101563732, "loss": 0.8591, "step": 3590 }, { "epoch": 0.2, "learning_rate": 0.00018458993393371832, "loss": 0.8708, "step": 3591 }, { "epoch": 0.2, "learning_rate": 0.00018458013399666717, "loss": 0.7605, "step": 3592 }, { "epoch": 0.2, "learning_rate": 0.0001845703312048147, "loss": 0.7307, "step": 3593 }, { "epoch": 0.2, "learning_rate": 0.00018456052555849173, "loss": 0.7699, "step": 3594 }, { "epoch": 0.2, "learning_rate": 0.00018455071705802926, "loss": 0.6842, "step": 3595 }, { "epoch": 0.2, "learning_rate": 0.00018454090570375838, "loss": 0.7599, "step": 3596 }, { "epoch": 0.2, "learning_rate": 0.0001845310914960102, "loss": 0.6418, "step": 3597 }, { "epoch": 0.2, "learning_rate": 0.00018452127443511598, "loss": 0.7004, "step": 3598 }, { "epoch": 0.2, "learning_rate": 0.00018451145452140708, "loss": 0.755, "step": 3599 }, { "epoch": 0.2, "learning_rate": 0.00018450163175521496, "loss": 0.7525, "step": 3600 }, { "epoch": 0.2, "learning_rate": 0.00018449180613687116, "loss": 0.6972, "step": 3601 }, { "epoch": 0.2, "learning_rate": 0.0001844819776667073, "loss": 0.5983, "step": 3602 }, { "epoch": 0.2, "learning_rate": 0.00018447214634505514, "loss": 0.6391, "step": 3603 }, { "epoch": 0.2, "learning_rate": 0.0001844623121722465, "loss": 0.7146, "step": 3604 }, { "epoch": 0.2, "learning_rate": 0.00018445247514861331, "loss": 0.6314, "step": 3605 }, { "epoch": 0.2, "learning_rate": 0.00018444263527448763, "loss": 0.732, "step": 3606 }, { "epoch": 0.2, "learning_rate": 0.00018443279255020152, "loss": 0.7148, "step": 3607 }, { "epoch": 0.2, "learning_rate": 0.00018442294697608723, "loss": 0.6492, "step": 3608 }, { "epoch": 0.2, "learning_rate": 0.00018441309855247708, "loss": 0.7904, "step": 3609 }, { "epoch": 0.2, "learning_rate": 0.00018440324727970347, "loss": 0.7657, "step": 3610 }, { "epoch": 0.2, "learning_rate": 0.0001843933931580989, "loss": 0.7585, "step": 3611 }, { "epoch": 0.2, "learning_rate": 0.000184383536187996, "loss": 0.7149, "step": 3612 }, { "epoch": 0.2, "learning_rate": 0.00018437367636972744, "loss": 0.7313, "step": 3613 }, { "epoch": 0.21, "learning_rate": 0.00018436381370362604, "loss": 0.7366, "step": 3614 }, { "epoch": 0.21, "learning_rate": 0.00018435394819002465, "loss": 0.7481, "step": 3615 }, { "epoch": 0.21, "learning_rate": 0.0001843440798292563, "loss": 0.7392, "step": 3616 }, { "epoch": 0.21, "learning_rate": 0.00018433420862165405, "loss": 0.7636, "step": 3617 }, { "epoch": 0.21, "learning_rate": 0.00018432433456755108, "loss": 0.699, "step": 3618 }, { "epoch": 0.21, "learning_rate": 0.00018431445766728066, "loss": 0.742, "step": 3619 }, { "epoch": 0.21, "learning_rate": 0.00018430457792117618, "loss": 0.8445, "step": 3620 }, { "epoch": 0.21, "learning_rate": 0.0001842946953295711, "loss": 0.8392, "step": 3621 }, { "epoch": 0.21, "learning_rate": 0.00018428480989279898, "loss": 0.7687, "step": 3622 }, { "epoch": 0.21, "learning_rate": 0.00018427492161119348, "loss": 0.7141, "step": 3623 }, { "epoch": 0.21, "learning_rate": 0.00018426503048508836, "loss": 0.6808, "step": 3624 }, { "epoch": 0.21, "learning_rate": 0.00018425513651481747, "loss": 0.7425, "step": 3625 }, { "epoch": 0.21, "learning_rate": 0.00018424523970071477, "loss": 0.7374, "step": 3626 }, { "epoch": 0.21, "learning_rate": 0.00018423534004311426, "loss": 0.7415, "step": 3627 }, { "epoch": 0.21, "learning_rate": 0.00018422543754235014, "loss": 0.7857, "step": 3628 }, { "epoch": 0.21, "learning_rate": 0.00018421553219875658, "loss": 0.6938, "step": 3629 }, { "epoch": 0.21, "learning_rate": 0.00018420562401266796, "loss": 0.8091, "step": 3630 }, { "epoch": 0.21, "learning_rate": 0.0001841957129844187, "loss": 0.6981, "step": 3631 }, { "epoch": 0.21, "learning_rate": 0.00018418579911434331, "loss": 0.6116, "step": 3632 }, { "epoch": 0.21, "learning_rate": 0.0001841758824027764, "loss": 0.7381, "step": 3633 }, { "epoch": 0.21, "learning_rate": 0.00018416596285005272, "loss": 0.7524, "step": 3634 }, { "epoch": 0.21, "learning_rate": 0.00018415604045650706, "loss": 0.6544, "step": 3635 }, { "epoch": 0.21, "learning_rate": 0.0001841461152224743, "loss": 0.7288, "step": 3636 }, { "epoch": 0.21, "learning_rate": 0.0001841361871482895, "loss": 0.7047, "step": 3637 }, { "epoch": 0.21, "learning_rate": 0.00018412625623428773, "loss": 0.6801, "step": 3638 }, { "epoch": 0.21, "learning_rate": 0.00018411632248080414, "loss": 0.7546, "step": 3639 }, { "epoch": 0.21, "learning_rate": 0.0001841063858881741, "loss": 0.7406, "step": 3640 }, { "epoch": 0.21, "learning_rate": 0.00018409644645673296, "loss": 0.6478, "step": 3641 }, { "epoch": 0.21, "learning_rate": 0.00018408650418681618, "loss": 0.7185, "step": 3642 }, { "epoch": 0.21, "learning_rate": 0.0001840765590787594, "loss": 0.6908, "step": 3643 }, { "epoch": 0.21, "learning_rate": 0.00018406661113289823, "loss": 0.7082, "step": 3644 }, { "epoch": 0.21, "learning_rate": 0.00018405666034956844, "loss": 0.7844, "step": 3645 }, { "epoch": 0.21, "learning_rate": 0.00018404670672910592, "loss": 0.709, "step": 3646 }, { "epoch": 0.21, "learning_rate": 0.00018403675027184666, "loss": 0.7509, "step": 3647 }, { "epoch": 0.21, "learning_rate": 0.00018402679097812664, "loss": 0.7267, "step": 3648 }, { "epoch": 0.21, "learning_rate": 0.00018401682884828212, "loss": 0.7071, "step": 3649 }, { "epoch": 0.21, "learning_rate": 0.00018400686388264924, "loss": 0.7312, "step": 3650 }, { "epoch": 0.21, "learning_rate": 0.00018399689608156438, "loss": 0.6859, "step": 3651 }, { "epoch": 0.21, "learning_rate": 0.00018398692544536397, "loss": 0.7368, "step": 3652 }, { "epoch": 0.21, "learning_rate": 0.00018397695197438463, "loss": 0.5768, "step": 3653 }, { "epoch": 0.21, "learning_rate": 0.00018396697566896286, "loss": 0.6904, "step": 3654 }, { "epoch": 0.21, "learning_rate": 0.00018395699652943548, "loss": 0.7624, "step": 3655 }, { "epoch": 0.21, "learning_rate": 0.00018394701455613929, "loss": 0.8039, "step": 3656 }, { "epoch": 0.21, "learning_rate": 0.00018393702974941117, "loss": 0.7928, "step": 3657 }, { "epoch": 0.21, "learning_rate": 0.00018392704210958819, "loss": 0.7966, "step": 3658 }, { "epoch": 0.21, "learning_rate": 0.0001839170516370074, "loss": 0.7168, "step": 3659 }, { "epoch": 0.21, "learning_rate": 0.00018390705833200608, "loss": 0.7216, "step": 3660 }, { "epoch": 0.21, "learning_rate": 0.00018389706219492147, "loss": 0.7457, "step": 3661 }, { "epoch": 0.21, "learning_rate": 0.00018388706322609099, "loss": 0.7896, "step": 3662 }, { "epoch": 0.21, "learning_rate": 0.00018387706142585208, "loss": 0.7613, "step": 3663 }, { "epoch": 0.21, "learning_rate": 0.00018386705679454242, "loss": 0.7782, "step": 3664 }, { "epoch": 0.21, "learning_rate": 0.00018385704933249962, "loss": 0.6945, "step": 3665 }, { "epoch": 0.21, "learning_rate": 0.0001838470390400615, "loss": 0.8187, "step": 3666 }, { "epoch": 0.21, "learning_rate": 0.0001838370259175659, "loss": 0.7171, "step": 3667 }, { "epoch": 0.21, "learning_rate": 0.0001838270099653508, "loss": 0.759, "step": 3668 }, { "epoch": 0.21, "learning_rate": 0.00018381699118375427, "loss": 0.6996, "step": 3669 }, { "epoch": 0.21, "learning_rate": 0.00018380696957311449, "loss": 0.7377, "step": 3670 }, { "epoch": 0.21, "learning_rate": 0.00018379694513376968, "loss": 0.7571, "step": 3671 }, { "epoch": 0.21, "learning_rate": 0.00018378691786605822, "loss": 0.8894, "step": 3672 }, { "epoch": 0.21, "learning_rate": 0.00018377688777031852, "loss": 0.7265, "step": 3673 }, { "epoch": 0.21, "learning_rate": 0.00018376685484688916, "loss": 0.7654, "step": 3674 }, { "epoch": 0.21, "learning_rate": 0.00018375681909610876, "loss": 0.778, "step": 3675 }, { "epoch": 0.21, "learning_rate": 0.00018374678051831603, "loss": 0.7601, "step": 3676 }, { "epoch": 0.21, "learning_rate": 0.00018373673911384987, "loss": 0.751, "step": 3677 }, { "epoch": 0.21, "learning_rate": 0.00018372669488304915, "loss": 0.7514, "step": 3678 }, { "epoch": 0.21, "learning_rate": 0.00018371664782625287, "loss": 0.7026, "step": 3679 }, { "epoch": 0.21, "learning_rate": 0.0001837065979438002, "loss": 0.669, "step": 3680 }, { "epoch": 0.21, "learning_rate": 0.00018369654523603027, "loss": 0.7837, "step": 3681 }, { "epoch": 0.21, "learning_rate": 0.00018368648970328247, "loss": 0.7212, "step": 3682 }, { "epoch": 0.21, "learning_rate": 0.00018367643134589617, "loss": 0.7192, "step": 3683 }, { "epoch": 0.21, "learning_rate": 0.00018366637016421086, "loss": 0.6904, "step": 3684 }, { "epoch": 0.21, "learning_rate": 0.00018365630615856614, "loss": 0.7671, "step": 3685 }, { "epoch": 0.21, "learning_rate": 0.0001836462393293017, "loss": 0.8087, "step": 3686 }, { "epoch": 0.21, "learning_rate": 0.00018363616967675728, "loss": 0.8018, "step": 3687 }, { "epoch": 0.21, "learning_rate": 0.0001836260972012728, "loss": 0.738, "step": 3688 }, { "epoch": 0.21, "learning_rate": 0.00018361602190318822, "loss": 0.7272, "step": 3689 }, { "epoch": 0.21, "learning_rate": 0.00018360594378284364, "loss": 0.7128, "step": 3690 }, { "epoch": 0.21, "learning_rate": 0.00018359586284057917, "loss": 0.754, "step": 3691 }, { "epoch": 0.21, "learning_rate": 0.00018358577907673506, "loss": 0.6786, "step": 3692 }, { "epoch": 0.21, "learning_rate": 0.00018357569249165175, "loss": 0.6991, "step": 3693 }, { "epoch": 0.21, "learning_rate": 0.0001835656030856696, "loss": 0.8365, "step": 3694 }, { "epoch": 0.21, "learning_rate": 0.00018355551085912918, "loss": 0.7164, "step": 3695 }, { "epoch": 0.21, "learning_rate": 0.00018354541581237115, "loss": 0.7331, "step": 3696 }, { "epoch": 0.21, "learning_rate": 0.00018353531794573625, "loss": 0.8245, "step": 3697 }, { "epoch": 0.21, "learning_rate": 0.00018352521725956527, "loss": 0.8176, "step": 3698 }, { "epoch": 0.21, "learning_rate": 0.00018351511375419913, "loss": 0.6914, "step": 3699 }, { "epoch": 0.21, "learning_rate": 0.00018350500742997893, "loss": 0.748, "step": 3700 }, { "epoch": 0.21, "learning_rate": 0.0001834948982872457, "loss": 0.6824, "step": 3701 }, { "epoch": 0.21, "learning_rate": 0.00018348478632634066, "loss": 0.6591, "step": 3702 }, { "epoch": 0.21, "learning_rate": 0.00018347467154760516, "loss": 0.7923, "step": 3703 }, { "epoch": 0.21, "learning_rate": 0.00018346455395138058, "loss": 0.6798, "step": 3704 }, { "epoch": 0.21, "learning_rate": 0.00018345443353800839, "loss": 0.7029, "step": 3705 }, { "epoch": 0.21, "learning_rate": 0.0001834443103078302, "loss": 0.7244, "step": 3706 }, { "epoch": 0.21, "learning_rate": 0.0001834341842611877, "loss": 0.8185, "step": 3707 }, { "epoch": 0.21, "learning_rate": 0.00018342405539842268, "loss": 0.6982, "step": 3708 }, { "epoch": 0.21, "learning_rate": 0.000183413923719877, "loss": 0.7592, "step": 3709 }, { "epoch": 0.21, "learning_rate": 0.00018340378922589263, "loss": 0.7494, "step": 3710 }, { "epoch": 0.21, "learning_rate": 0.00018339365191681164, "loss": 0.7839, "step": 3711 }, { "epoch": 0.21, "learning_rate": 0.00018338351179297618, "loss": 0.6624, "step": 3712 }, { "epoch": 0.21, "learning_rate": 0.00018337336885472852, "loss": 0.7199, "step": 3713 }, { "epoch": 0.21, "learning_rate": 0.00018336322310241103, "loss": 0.6326, "step": 3714 }, { "epoch": 0.21, "learning_rate": 0.0001833530745363661, "loss": 0.7085, "step": 3715 }, { "epoch": 0.21, "learning_rate": 0.0001833429231569363, "loss": 0.6127, "step": 3716 }, { "epoch": 0.21, "learning_rate": 0.0001833327689644643, "loss": 0.6195, "step": 3717 }, { "epoch": 0.21, "learning_rate": 0.0001833226119592928, "loss": 0.6921, "step": 3718 }, { "epoch": 0.21, "learning_rate": 0.00018331245214176458, "loss": 0.738, "step": 3719 }, { "epoch": 0.21, "learning_rate": 0.00018330228951222264, "loss": 0.6975, "step": 3720 }, { "epoch": 0.21, "learning_rate": 0.00018329212407100994, "loss": 0.6457, "step": 3721 }, { "epoch": 0.21, "learning_rate": 0.00018328195581846966, "loss": 0.641, "step": 3722 }, { "epoch": 0.21, "learning_rate": 0.0001832717847549449, "loss": 0.7122, "step": 3723 }, { "epoch": 0.21, "learning_rate": 0.00018326161088077903, "loss": 0.8463, "step": 3724 }, { "epoch": 0.21, "learning_rate": 0.00018325143419631546, "loss": 0.8019, "step": 3725 }, { "epoch": 0.21, "learning_rate": 0.00018324125470189762, "loss": 0.7404, "step": 3726 }, { "epoch": 0.21, "learning_rate": 0.00018323107239786915, "loss": 0.8004, "step": 3727 }, { "epoch": 0.21, "learning_rate": 0.00018322088728457367, "loss": 0.7871, "step": 3728 }, { "epoch": 0.21, "learning_rate": 0.00018321069936235503, "loss": 0.6986, "step": 3729 }, { "epoch": 0.21, "learning_rate": 0.00018320050863155702, "loss": 0.7784, "step": 3730 }, { "epoch": 0.21, "learning_rate": 0.00018319031509252367, "loss": 0.9659, "step": 3731 }, { "epoch": 0.21, "learning_rate": 0.00018318011874559898, "loss": 0.7166, "step": 3732 }, { "epoch": 0.21, "learning_rate": 0.00018316991959112716, "loss": 0.8939, "step": 3733 }, { "epoch": 0.21, "learning_rate": 0.00018315971762945243, "loss": 0.7261, "step": 3734 }, { "epoch": 0.21, "learning_rate": 0.00018314951286091914, "loss": 0.7687, "step": 3735 }, { "epoch": 0.21, "learning_rate": 0.00018313930528587172, "loss": 0.6583, "step": 3736 }, { "epoch": 0.21, "learning_rate": 0.00018312909490465472, "loss": 0.8011, "step": 3737 }, { "epoch": 0.21, "learning_rate": 0.00018311888171761273, "loss": 0.824, "step": 3738 }, { "epoch": 0.21, "learning_rate": 0.0001831086657250905, "loss": 0.7315, "step": 3739 }, { "epoch": 0.21, "learning_rate": 0.00018309844692743283, "loss": 0.7384, "step": 3740 }, { "epoch": 0.21, "learning_rate": 0.00018308822532498467, "loss": 0.72, "step": 3741 }, { "epoch": 0.21, "learning_rate": 0.00018307800091809097, "loss": 0.8121, "step": 3742 }, { "epoch": 0.21, "learning_rate": 0.0001830677737070969, "loss": 0.6256, "step": 3743 }, { "epoch": 0.21, "learning_rate": 0.00018305754369234758, "loss": 0.8243, "step": 3744 }, { "epoch": 0.21, "learning_rate": 0.00018304731087418833, "loss": 0.7202, "step": 3745 }, { "epoch": 0.21, "learning_rate": 0.00018303707525296457, "loss": 0.8234, "step": 3746 }, { "epoch": 0.21, "learning_rate": 0.0001830268368290217, "loss": 0.6695, "step": 3747 }, { "epoch": 0.21, "learning_rate": 0.00018301659560270538, "loss": 0.8048, "step": 3748 }, { "epoch": 0.21, "learning_rate": 0.00018300635157436125, "loss": 0.6776, "step": 3749 }, { "epoch": 0.21, "learning_rate": 0.00018299610474433503, "loss": 0.7291, "step": 3750 }, { "epoch": 0.21, "learning_rate": 0.0001829858551129726, "loss": 0.7355, "step": 3751 }, { "epoch": 0.21, "learning_rate": 0.00018297560268061992, "loss": 0.7323, "step": 3752 }, { "epoch": 0.21, "learning_rate": 0.00018296534744762308, "loss": 0.7292, "step": 3753 }, { "epoch": 0.21, "learning_rate": 0.00018295508941432815, "loss": 0.841, "step": 3754 }, { "epoch": 0.21, "learning_rate": 0.0001829448285810814, "loss": 0.7599, "step": 3755 }, { "epoch": 0.21, "learning_rate": 0.00018293456494822917, "loss": 0.7913, "step": 3756 }, { "epoch": 0.21, "learning_rate": 0.00018292429851611783, "loss": 0.7178, "step": 3757 }, { "epoch": 0.21, "learning_rate": 0.00018291402928509397, "loss": 0.6865, "step": 3758 }, { "epoch": 0.21, "learning_rate": 0.00018290375725550417, "loss": 0.75, "step": 3759 }, { "epoch": 0.21, "learning_rate": 0.00018289348242769515, "loss": 0.7191, "step": 3760 }, { "epoch": 0.21, "learning_rate": 0.00018288320480201368, "loss": 0.6882, "step": 3761 }, { "epoch": 0.21, "learning_rate": 0.00018287292437880667, "loss": 0.7691, "step": 3762 }, { "epoch": 0.21, "learning_rate": 0.00018286264115842117, "loss": 0.777, "step": 3763 }, { "epoch": 0.21, "learning_rate": 0.00018285235514120415, "loss": 0.7904, "step": 3764 }, { "epoch": 0.21, "learning_rate": 0.0001828420663275029, "loss": 0.7551, "step": 3765 }, { "epoch": 0.21, "learning_rate": 0.00018283177471766464, "loss": 0.7849, "step": 3766 }, { "epoch": 0.21, "learning_rate": 0.00018282148031203676, "loss": 0.7675, "step": 3767 }, { "epoch": 0.21, "learning_rate": 0.00018281118311096667, "loss": 0.7277, "step": 3768 }, { "epoch": 0.21, "learning_rate": 0.00018280088311480201, "loss": 0.6946, "step": 3769 }, { "epoch": 0.21, "learning_rate": 0.00018279058032389037, "loss": 0.7844, "step": 3770 }, { "epoch": 0.21, "learning_rate": 0.00018278027473857952, "loss": 0.7506, "step": 3771 }, { "epoch": 0.21, "learning_rate": 0.00018276996635921734, "loss": 0.7656, "step": 3772 }, { "epoch": 0.21, "learning_rate": 0.00018275965518615167, "loss": 0.6852, "step": 3773 }, { "epoch": 0.21, "learning_rate": 0.00018274934121973062, "loss": 0.7794, "step": 3774 }, { "epoch": 0.21, "learning_rate": 0.00018273902446030226, "loss": 0.7092, "step": 3775 }, { "epoch": 0.21, "learning_rate": 0.00018272870490821486, "loss": 0.6517, "step": 3776 }, { "epoch": 0.21, "learning_rate": 0.0001827183825638167, "loss": 0.7838, "step": 3777 }, { "epoch": 0.21, "learning_rate": 0.00018270805742745617, "loss": 0.7399, "step": 3778 }, { "epoch": 0.21, "learning_rate": 0.00018269772949948182, "loss": 0.7366, "step": 3779 }, { "epoch": 0.21, "learning_rate": 0.0001826873987802422, "loss": 0.7139, "step": 3780 }, { "epoch": 0.21, "learning_rate": 0.00018267706527008603, "loss": 0.7915, "step": 3781 }, { "epoch": 0.21, "learning_rate": 0.0001826667289693621, "loss": 0.7059, "step": 3782 }, { "epoch": 0.21, "learning_rate": 0.00018265638987841922, "loss": 0.7938, "step": 3783 }, { "epoch": 0.21, "learning_rate": 0.00018264604799760642, "loss": 0.7609, "step": 3784 }, { "epoch": 0.21, "learning_rate": 0.00018263570332727275, "loss": 0.7902, "step": 3785 }, { "epoch": 0.21, "learning_rate": 0.0001826253558677674, "loss": 0.7851, "step": 3786 }, { "epoch": 0.21, "learning_rate": 0.00018261500561943955, "loss": 0.8005, "step": 3787 }, { "epoch": 0.21, "learning_rate": 0.00018260465258263865, "loss": 0.7348, "step": 3788 }, { "epoch": 0.21, "learning_rate": 0.00018259429675771403, "loss": 0.6224, "step": 3789 }, { "epoch": 0.21, "learning_rate": 0.0001825839381450153, "loss": 0.7034, "step": 3790 }, { "epoch": 0.22, "learning_rate": 0.0001825735767448921, "loss": 0.7276, "step": 3791 }, { "epoch": 0.22, "learning_rate": 0.0001825632125576941, "loss": 0.8232, "step": 3792 }, { "epoch": 0.22, "learning_rate": 0.00018255284558377116, "loss": 0.757, "step": 3793 }, { "epoch": 0.22, "learning_rate": 0.00018254247582347318, "loss": 0.718, "step": 3794 }, { "epoch": 0.22, "learning_rate": 0.00018253210327715018, "loss": 0.7403, "step": 3795 }, { "epoch": 0.22, "learning_rate": 0.00018252172794515223, "loss": 0.6728, "step": 3796 }, { "epoch": 0.22, "learning_rate": 0.00018251134982782952, "loss": 0.6841, "step": 3797 }, { "epoch": 0.22, "learning_rate": 0.00018250096892553235, "loss": 0.7341, "step": 3798 }, { "epoch": 0.22, "learning_rate": 0.00018249058523861118, "loss": 0.8388, "step": 3799 }, { "epoch": 0.22, "learning_rate": 0.00018248019876741637, "loss": 0.737, "step": 3800 }, { "epoch": 0.22, "learning_rate": 0.00018246980951229852, "loss": 0.7993, "step": 3801 }, { "epoch": 0.22, "learning_rate": 0.00018245941747360839, "loss": 0.8639, "step": 3802 }, { "epoch": 0.22, "learning_rate": 0.0001824490226516966, "loss": 0.7782, "step": 3803 }, { "epoch": 0.22, "learning_rate": 0.00018243862504691407, "loss": 0.7285, "step": 3804 }, { "epoch": 0.22, "learning_rate": 0.00018242822465961176, "loss": 0.7394, "step": 3805 }, { "epoch": 0.22, "learning_rate": 0.00018241782149014068, "loss": 0.6122, "step": 3806 }, { "epoch": 0.22, "learning_rate": 0.00018240741553885196, "loss": 0.7901, "step": 3807 }, { "epoch": 0.22, "learning_rate": 0.00018239700680609686, "loss": 0.7922, "step": 3808 }, { "epoch": 0.22, "learning_rate": 0.0001823865952922267, "loss": 0.7898, "step": 3809 }, { "epoch": 0.22, "learning_rate": 0.00018237618099759288, "loss": 0.7596, "step": 3810 }, { "epoch": 0.22, "learning_rate": 0.0001823657639225469, "loss": 0.7616, "step": 3811 }, { "epoch": 0.22, "learning_rate": 0.00018235534406744038, "loss": 0.666, "step": 3812 }, { "epoch": 0.22, "learning_rate": 0.000182344921432625, "loss": 0.7079, "step": 3813 }, { "epoch": 0.22, "learning_rate": 0.00018233449601845258, "loss": 0.5438, "step": 3814 }, { "epoch": 0.22, "learning_rate": 0.00018232406782527498, "loss": 0.726, "step": 3815 }, { "epoch": 0.22, "learning_rate": 0.0001823136368534442, "loss": 0.87, "step": 3816 }, { "epoch": 0.22, "learning_rate": 0.00018230320310331228, "loss": 0.6784, "step": 3817 }, { "epoch": 0.22, "learning_rate": 0.00018229276657523143, "loss": 0.776, "step": 3818 }, { "epoch": 0.22, "learning_rate": 0.00018228232726955388, "loss": 0.7796, "step": 3819 }, { "epoch": 0.22, "learning_rate": 0.000182271885186632, "loss": 0.7088, "step": 3820 }, { "epoch": 0.22, "learning_rate": 0.0001822614403268182, "loss": 0.7392, "step": 3821 }, { "epoch": 0.22, "learning_rate": 0.00018225099269046508, "loss": 0.8127, "step": 3822 }, { "epoch": 0.22, "learning_rate": 0.00018224054227792524, "loss": 0.748, "step": 3823 }, { "epoch": 0.22, "learning_rate": 0.0001822300890895514, "loss": 0.7883, "step": 3824 }, { "epoch": 0.22, "learning_rate": 0.0001822196331256964, "loss": 0.7218, "step": 3825 }, { "epoch": 0.22, "learning_rate": 0.00018220917438671318, "loss": 0.7231, "step": 3826 }, { "epoch": 0.22, "learning_rate": 0.00018219871287295468, "loss": 0.8202, "step": 3827 }, { "epoch": 0.22, "learning_rate": 0.0001821882485847741, "loss": 0.7659, "step": 3828 }, { "epoch": 0.22, "learning_rate": 0.0001821777815225245, "loss": 0.6718, "step": 3829 }, { "epoch": 0.22, "learning_rate": 0.00018216731168655933, "loss": 0.7662, "step": 3830 }, { "epoch": 0.22, "learning_rate": 0.00018215683907723187, "loss": 0.8151, "step": 3831 }, { "epoch": 0.22, "learning_rate": 0.00018214636369489563, "loss": 0.724, "step": 3832 }, { "epoch": 0.22, "learning_rate": 0.0001821358855399042, "loss": 0.7587, "step": 3833 }, { "epoch": 0.22, "learning_rate": 0.0001821254046126112, "loss": 0.7524, "step": 3834 }, { "epoch": 0.22, "learning_rate": 0.00018211492091337042, "loss": 0.8218, "step": 3835 }, { "epoch": 0.22, "learning_rate": 0.00018210443444253567, "loss": 0.749, "step": 3836 }, { "epoch": 0.22, "learning_rate": 0.00018209394520046098, "loss": 0.8019, "step": 3837 }, { "epoch": 0.22, "learning_rate": 0.00018208345318750034, "loss": 0.7152, "step": 3838 }, { "epoch": 0.22, "learning_rate": 0.00018207295840400788, "loss": 0.7536, "step": 3839 }, { "epoch": 0.22, "learning_rate": 0.0001820624608503378, "loss": 0.7623, "step": 3840 }, { "epoch": 0.22, "learning_rate": 0.00018205196052684445, "loss": 0.7356, "step": 3841 }, { "epoch": 0.22, "learning_rate": 0.00018204145743388228, "loss": 0.6645, "step": 3842 }, { "epoch": 0.22, "learning_rate": 0.00018203095157180572, "loss": 0.7104, "step": 3843 }, { "epoch": 0.22, "learning_rate": 0.00018202044294096946, "loss": 0.7265, "step": 3844 }, { "epoch": 0.22, "learning_rate": 0.00018200993154172807, "loss": 0.6665, "step": 3845 }, { "epoch": 0.22, "learning_rate": 0.00018199941737443646, "loss": 0.7188, "step": 3846 }, { "epoch": 0.22, "learning_rate": 0.00018198890043944946, "loss": 0.7799, "step": 3847 }, { "epoch": 0.22, "learning_rate": 0.00018197838073712204, "loss": 0.6797, "step": 3848 }, { "epoch": 0.22, "learning_rate": 0.00018196785826780928, "loss": 0.8081, "step": 3849 }, { "epoch": 0.22, "learning_rate": 0.00018195733303186633, "loss": 0.7134, "step": 3850 }, { "epoch": 0.22, "learning_rate": 0.00018194680502964846, "loss": 0.7802, "step": 3851 }, { "epoch": 0.22, "learning_rate": 0.000181936274261511, "loss": 0.7844, "step": 3852 }, { "epoch": 0.22, "learning_rate": 0.0001819257407278094, "loss": 0.7025, "step": 3853 }, { "epoch": 0.22, "learning_rate": 0.0001819152044288992, "loss": 0.7134, "step": 3854 }, { "epoch": 0.22, "learning_rate": 0.000181904665365136, "loss": 0.6413, "step": 3855 }, { "epoch": 0.22, "learning_rate": 0.00018189412353687554, "loss": 0.5528, "step": 3856 }, { "epoch": 0.22, "learning_rate": 0.00018188357894447367, "loss": 0.8072, "step": 3857 }, { "epoch": 0.22, "learning_rate": 0.00018187303158828622, "loss": 0.7441, "step": 3858 }, { "epoch": 0.22, "learning_rate": 0.00018186248146866927, "loss": 0.6894, "step": 3859 }, { "epoch": 0.22, "learning_rate": 0.00018185192858597883, "loss": 0.8143, "step": 3860 }, { "epoch": 0.22, "learning_rate": 0.00018184137294057116, "loss": 0.7178, "step": 3861 }, { "epoch": 0.22, "learning_rate": 0.00018183081453280252, "loss": 0.6897, "step": 3862 }, { "epoch": 0.22, "learning_rate": 0.0001818202533630293, "loss": 0.8791, "step": 3863 }, { "epoch": 0.22, "learning_rate": 0.00018180968943160795, "loss": 0.7419, "step": 3864 }, { "epoch": 0.22, "learning_rate": 0.00018179912273889501, "loss": 0.7158, "step": 3865 }, { "epoch": 0.22, "learning_rate": 0.00018178855328524713, "loss": 0.8426, "step": 3866 }, { "epoch": 0.22, "learning_rate": 0.0001817779810710211, "loss": 0.7822, "step": 3867 }, { "epoch": 0.22, "learning_rate": 0.0001817674060965737, "loss": 0.7422, "step": 3868 }, { "epoch": 0.22, "learning_rate": 0.000181756828362262, "loss": 0.6523, "step": 3869 }, { "epoch": 0.22, "learning_rate": 0.00018174624786844286, "loss": 0.7703, "step": 3870 }, { "epoch": 0.22, "learning_rate": 0.00018173566461547345, "loss": 0.7723, "step": 3871 }, { "epoch": 0.22, "learning_rate": 0.000181725078603711, "loss": 0.7271, "step": 3872 }, { "epoch": 0.22, "learning_rate": 0.00018171448983351284, "loss": 0.7521, "step": 3873 }, { "epoch": 0.22, "learning_rate": 0.00018170389830523637, "loss": 0.7003, "step": 3874 }, { "epoch": 0.22, "learning_rate": 0.000181693304019239, "loss": 0.7525, "step": 3875 }, { "epoch": 0.22, "learning_rate": 0.0001816827069758784, "loss": 0.6772, "step": 3876 }, { "epoch": 0.22, "learning_rate": 0.00018167210717551224, "loss": 0.6822, "step": 3877 }, { "epoch": 0.22, "learning_rate": 0.00018166150461849825, "loss": 0.6941, "step": 3878 }, { "epoch": 0.22, "learning_rate": 0.0001816508993051943, "loss": 0.7188, "step": 3879 }, { "epoch": 0.22, "learning_rate": 0.00018164029123595837, "loss": 0.7675, "step": 3880 }, { "epoch": 0.22, "learning_rate": 0.0001816296804111485, "loss": 0.7404, "step": 3881 }, { "epoch": 0.22, "learning_rate": 0.00018161906683112285, "loss": 0.7522, "step": 3882 }, { "epoch": 0.22, "learning_rate": 0.00018160845049623964, "loss": 0.7682, "step": 3883 }, { "epoch": 0.22, "learning_rate": 0.00018159783140685715, "loss": 0.6548, "step": 3884 }, { "epoch": 0.22, "learning_rate": 0.0001815872095633339, "loss": 0.6171, "step": 3885 }, { "epoch": 0.22, "learning_rate": 0.00018157658496602833, "loss": 0.7737, "step": 3886 }, { "epoch": 0.22, "learning_rate": 0.00018156595761529914, "loss": 0.6828, "step": 3887 }, { "epoch": 0.22, "learning_rate": 0.00018155532751150491, "loss": 0.7482, "step": 3888 }, { "epoch": 0.22, "learning_rate": 0.00018154469465500448, "loss": 0.771, "step": 3889 }, { "epoch": 0.22, "learning_rate": 0.00018153405904615676, "loss": 0.7861, "step": 3890 }, { "epoch": 0.22, "learning_rate": 0.00018152342068532073, "loss": 0.7265, "step": 3891 }, { "epoch": 0.22, "learning_rate": 0.00018151277957285543, "loss": 0.776, "step": 3892 }, { "epoch": 0.22, "learning_rate": 0.00018150213570912005, "loss": 0.72, "step": 3893 }, { "epoch": 0.22, "learning_rate": 0.00018149148909447384, "loss": 0.7761, "step": 3894 }, { "epoch": 0.22, "learning_rate": 0.00018148083972927616, "loss": 0.7851, "step": 3895 }, { "epoch": 0.22, "learning_rate": 0.00018147018761388646, "loss": 0.7819, "step": 3896 }, { "epoch": 0.22, "learning_rate": 0.00018145953274866425, "loss": 0.8126, "step": 3897 }, { "epoch": 0.22, "learning_rate": 0.00018144887513396919, "loss": 0.6336, "step": 3898 }, { "epoch": 0.22, "learning_rate": 0.00018143821477016096, "loss": 0.7576, "step": 3899 }, { "epoch": 0.22, "learning_rate": 0.0001814275516575994, "loss": 0.7166, "step": 3900 }, { "epoch": 0.22, "learning_rate": 0.00018141688579664448, "loss": 0.8552, "step": 3901 }, { "epoch": 0.22, "learning_rate": 0.00018140621718765605, "loss": 0.836, "step": 3902 }, { "epoch": 0.22, "learning_rate": 0.00018139554583099436, "loss": 0.7172, "step": 3903 }, { "epoch": 0.22, "learning_rate": 0.0001813848717270195, "loss": 0.7536, "step": 3904 }, { "epoch": 0.22, "learning_rate": 0.0001813741948760918, "loss": 0.7088, "step": 3905 }, { "epoch": 0.22, "learning_rate": 0.00018136351527857158, "loss": 0.6726, "step": 3906 }, { "epoch": 0.22, "learning_rate": 0.00018135283293481936, "loss": 0.7514, "step": 3907 }, { "epoch": 0.22, "learning_rate": 0.00018134214784519567, "loss": 0.8048, "step": 3908 }, { "epoch": 0.22, "learning_rate": 0.00018133146001006117, "loss": 0.7291, "step": 3909 }, { "epoch": 0.22, "learning_rate": 0.00018132076942977657, "loss": 0.8054, "step": 3910 }, { "epoch": 0.22, "learning_rate": 0.00018131007610470276, "loss": 0.6687, "step": 3911 }, { "epoch": 0.22, "learning_rate": 0.00018129938003520063, "loss": 0.705, "step": 3912 }, { "epoch": 0.22, "learning_rate": 0.00018128868122163123, "loss": 0.7047, "step": 3913 }, { "epoch": 0.22, "learning_rate": 0.0001812779796643556, "loss": 0.6744, "step": 3914 }, { "epoch": 0.22, "learning_rate": 0.00018126727536373505, "loss": 0.707, "step": 3915 }, { "epoch": 0.22, "learning_rate": 0.00018125656832013078, "loss": 0.7516, "step": 3916 }, { "epoch": 0.22, "learning_rate": 0.00018124585853390425, "loss": 0.7129, "step": 3917 }, { "epoch": 0.22, "learning_rate": 0.0001812351460054169, "loss": 0.7527, "step": 3918 }, { "epoch": 0.22, "learning_rate": 0.00018122443073503034, "loss": 0.7515, "step": 3919 }, { "epoch": 0.22, "learning_rate": 0.00018121371272310623, "loss": 0.7308, "step": 3920 }, { "epoch": 0.22, "learning_rate": 0.00018120299197000632, "loss": 0.7598, "step": 3921 }, { "epoch": 0.22, "learning_rate": 0.00018119226847609245, "loss": 0.7866, "step": 3922 }, { "epoch": 0.22, "learning_rate": 0.0001811815422417266, "loss": 0.7385, "step": 3923 }, { "epoch": 0.22, "learning_rate": 0.00018117081326727078, "loss": 0.8732, "step": 3924 }, { "epoch": 0.22, "learning_rate": 0.00018116008155308718, "loss": 0.8075, "step": 3925 }, { "epoch": 0.22, "learning_rate": 0.00018114934709953794, "loss": 0.6801, "step": 3926 }, { "epoch": 0.22, "learning_rate": 0.00018113860990698544, "loss": 0.809, "step": 3927 }, { "epoch": 0.22, "learning_rate": 0.00018112786997579204, "loss": 0.636, "step": 3928 }, { "epoch": 0.22, "learning_rate": 0.00018111712730632022, "loss": 0.7033, "step": 3929 }, { "epoch": 0.22, "learning_rate": 0.00018110638189893267, "loss": 0.7122, "step": 3930 }, { "epoch": 0.22, "learning_rate": 0.000181095633753992, "loss": 0.6818, "step": 3931 }, { "epoch": 0.22, "learning_rate": 0.00018108488287186102, "loss": 0.863, "step": 3932 }, { "epoch": 0.22, "learning_rate": 0.0001810741292529026, "loss": 0.6769, "step": 3933 }, { "epoch": 0.22, "learning_rate": 0.00018106337289747968, "loss": 0.6948, "step": 3934 }, { "epoch": 0.22, "learning_rate": 0.00018105261380595535, "loss": 0.7537, "step": 3935 }, { "epoch": 0.22, "learning_rate": 0.00018104185197869273, "loss": 0.762, "step": 3936 }, { "epoch": 0.22, "learning_rate": 0.000181031087416055, "loss": 0.677, "step": 3937 }, { "epoch": 0.22, "learning_rate": 0.00018102032011840563, "loss": 0.7321, "step": 3938 }, { "epoch": 0.22, "learning_rate": 0.00018100955008610797, "loss": 0.7604, "step": 3939 }, { "epoch": 0.22, "learning_rate": 0.0001809987773195255, "loss": 0.7215, "step": 3940 }, { "epoch": 0.22, "learning_rate": 0.00018098800181902188, "loss": 0.8676, "step": 3941 }, { "epoch": 0.22, "learning_rate": 0.00018097722358496082, "loss": 0.7501, "step": 3942 }, { "epoch": 0.22, "learning_rate": 0.0001809664426177061, "loss": 0.7752, "step": 3943 }, { "epoch": 0.22, "learning_rate": 0.0001809556589176216, "loss": 0.7477, "step": 3944 }, { "epoch": 0.22, "learning_rate": 0.00018094487248507127, "loss": 0.7421, "step": 3945 }, { "epoch": 0.22, "learning_rate": 0.00018093408332041922, "loss": 0.8875, "step": 3946 }, { "epoch": 0.22, "learning_rate": 0.00018092329142402956, "loss": 0.6367, "step": 3947 }, { "epoch": 0.22, "learning_rate": 0.00018091249679626667, "loss": 0.7658, "step": 3948 }, { "epoch": 0.22, "learning_rate": 0.00018090169943749476, "loss": 0.7385, "step": 3949 }, { "epoch": 0.22, "learning_rate": 0.00018089089934807833, "loss": 0.7841, "step": 3950 }, { "epoch": 0.22, "learning_rate": 0.0001808800965283819, "loss": 0.7742, "step": 3951 }, { "epoch": 0.22, "learning_rate": 0.0001808692909787701, "loss": 0.7262, "step": 3952 }, { "epoch": 0.22, "learning_rate": 0.00018085848269960763, "loss": 0.8744, "step": 3953 }, { "epoch": 0.22, "learning_rate": 0.00018084767169125932, "loss": 0.6428, "step": 3954 }, { "epoch": 0.22, "learning_rate": 0.00018083685795409004, "loss": 0.8406, "step": 3955 }, { "epoch": 0.22, "learning_rate": 0.00018082604148846482, "loss": 0.7085, "step": 3956 }, { "epoch": 0.22, "learning_rate": 0.00018081522229474874, "loss": 0.8237, "step": 3957 }, { "epoch": 0.22, "learning_rate": 0.00018080440037330695, "loss": 0.8113, "step": 3958 }, { "epoch": 0.22, "learning_rate": 0.00018079357572450472, "loss": 0.7365, "step": 3959 }, { "epoch": 0.22, "learning_rate": 0.00018078274834870743, "loss": 0.73, "step": 3960 }, { "epoch": 0.22, "learning_rate": 0.00018077191824628053, "loss": 0.6717, "step": 3961 }, { "epoch": 0.22, "learning_rate": 0.00018076108541758955, "loss": 0.8346, "step": 3962 }, { "epoch": 0.22, "learning_rate": 0.00018075024986300014, "loss": 0.7304, "step": 3963 }, { "epoch": 0.22, "learning_rate": 0.000180739411582878, "loss": 0.6742, "step": 3964 }, { "epoch": 0.22, "learning_rate": 0.000180728570577589, "loss": 0.7354, "step": 3965 }, { "epoch": 0.22, "learning_rate": 0.00018071772684749902, "loss": 0.7026, "step": 3966 }, { "epoch": 0.23, "learning_rate": 0.00018070688039297403, "loss": 0.6907, "step": 3967 }, { "epoch": 0.23, "learning_rate": 0.00018069603121438022, "loss": 0.8065, "step": 3968 }, { "epoch": 0.23, "learning_rate": 0.0001806851793120837, "loss": 0.8081, "step": 3969 }, { "epoch": 0.23, "learning_rate": 0.00018067432468645077, "loss": 0.6725, "step": 3970 }, { "epoch": 0.23, "learning_rate": 0.00018066346733784777, "loss": 0.721, "step": 3971 }, { "epoch": 0.23, "learning_rate": 0.00018065260726664127, "loss": 0.6727, "step": 3972 }, { "epoch": 0.23, "learning_rate": 0.0001806417444731977, "loss": 0.7353, "step": 3973 }, { "epoch": 0.23, "learning_rate": 0.0001806308789578838, "loss": 0.7682, "step": 3974 }, { "epoch": 0.23, "learning_rate": 0.00018062001072106628, "loss": 0.7565, "step": 3975 }, { "epoch": 0.23, "learning_rate": 0.00018060913976311192, "loss": 0.7102, "step": 3976 }, { "epoch": 0.23, "learning_rate": 0.00018059826608438773, "loss": 0.6757, "step": 3977 }, { "epoch": 0.23, "learning_rate": 0.00018058738968526066, "loss": 0.7408, "step": 3978 }, { "epoch": 0.23, "learning_rate": 0.00018057651056609784, "loss": 0.7322, "step": 3979 }, { "epoch": 0.23, "learning_rate": 0.00018056562872726645, "loss": 0.7203, "step": 3980 }, { "epoch": 0.23, "learning_rate": 0.00018055474416913382, "loss": 0.79, "step": 3981 }, { "epoch": 0.23, "learning_rate": 0.0001805438568920673, "loss": 0.6294, "step": 3982 }, { "epoch": 0.23, "learning_rate": 0.00018053296689643443, "loss": 0.8445, "step": 3983 }, { "epoch": 0.23, "learning_rate": 0.00018052207418260268, "loss": 0.8209, "step": 3984 }, { "epoch": 0.23, "learning_rate": 0.00018051117875093976, "loss": 0.7927, "step": 3985 }, { "epoch": 0.23, "learning_rate": 0.00018050028060181337, "loss": 0.7202, "step": 3986 }, { "epoch": 0.23, "learning_rate": 0.0001804893797355914, "loss": 0.7838, "step": 3987 }, { "epoch": 0.23, "learning_rate": 0.0001804784761526418, "loss": 0.7218, "step": 3988 }, { "epoch": 0.23, "learning_rate": 0.00018046756985333256, "loss": 0.6725, "step": 3989 }, { "epoch": 0.23, "learning_rate": 0.00018045666083803182, "loss": 0.7025, "step": 3990 }, { "epoch": 0.23, "learning_rate": 0.00018044574910710775, "loss": 0.7012, "step": 3991 }, { "epoch": 0.23, "learning_rate": 0.00018043483466092868, "loss": 0.7905, "step": 3992 }, { "epoch": 0.23, "learning_rate": 0.00018042391749986298, "loss": 0.842, "step": 3993 }, { "epoch": 0.23, "learning_rate": 0.00018041299762427916, "loss": 0.5995, "step": 3994 }, { "epoch": 0.23, "learning_rate": 0.00018040207503454577, "loss": 0.8086, "step": 3995 }, { "epoch": 0.23, "learning_rate": 0.00018039114973103152, "loss": 0.7935, "step": 3996 }, { "epoch": 0.23, "learning_rate": 0.00018038022171410506, "loss": 0.8124, "step": 3997 }, { "epoch": 0.23, "learning_rate": 0.00018036929098413538, "loss": 0.7632, "step": 3998 }, { "epoch": 0.23, "learning_rate": 0.0001803583575414913, "loss": 0.7217, "step": 3999 }, { "epoch": 0.23, "learning_rate": 0.00018034742138654192, "loss": 0.7205, "step": 4000 }, { "epoch": 0.23, "learning_rate": 0.00018033648251965638, "loss": 0.731, "step": 4001 }, { "epoch": 0.23, "learning_rate": 0.00018032554094120384, "loss": 0.7842, "step": 4002 }, { "epoch": 0.23, "learning_rate": 0.00018031459665155363, "loss": 0.7231, "step": 4003 }, { "epoch": 0.23, "learning_rate": 0.0001803036496510752, "loss": 0.7129, "step": 4004 }, { "epoch": 0.23, "learning_rate": 0.00018029269994013791, "loss": 0.7436, "step": 4005 }, { "epoch": 0.23, "learning_rate": 0.00018028174751911146, "loss": 0.7289, "step": 4006 }, { "epoch": 0.23, "learning_rate": 0.0001802707923883655, "loss": 0.8122, "step": 4007 }, { "epoch": 0.23, "learning_rate": 0.00018025983454826973, "loss": 0.6883, "step": 4008 }, { "epoch": 0.23, "learning_rate": 0.0001802488739991941, "loss": 0.6608, "step": 4009 }, { "epoch": 0.23, "learning_rate": 0.0001802379107415085, "loss": 0.6884, "step": 4010 }, { "epoch": 0.23, "learning_rate": 0.00018022694477558298, "loss": 0.7959, "step": 4011 }, { "epoch": 0.23, "learning_rate": 0.00018021597610178768, "loss": 0.6922, "step": 4012 }, { "epoch": 0.23, "learning_rate": 0.0001802050047204928, "loss": 0.8659, "step": 4013 }, { "epoch": 0.23, "learning_rate": 0.0001801940306320687, "loss": 0.8297, "step": 4014 }, { "epoch": 0.23, "learning_rate": 0.0001801830538368857, "loss": 0.5944, "step": 4015 }, { "epoch": 0.23, "learning_rate": 0.00018017207433531436, "loss": 0.741, "step": 4016 }, { "epoch": 0.23, "learning_rate": 0.00018016109212772526, "loss": 0.7699, "step": 4017 }, { "epoch": 0.23, "learning_rate": 0.00018015010721448907, "loss": 0.8546, "step": 4018 }, { "epoch": 0.23, "learning_rate": 0.00018013911959597655, "loss": 0.7964, "step": 4019 }, { "epoch": 0.23, "learning_rate": 0.00018012812927255858, "loss": 0.639, "step": 4020 }, { "epoch": 0.23, "learning_rate": 0.00018011713624460608, "loss": 0.7713, "step": 4021 }, { "epoch": 0.23, "learning_rate": 0.00018010614051249015, "loss": 0.7458, "step": 4022 }, { "epoch": 0.23, "learning_rate": 0.00018009514207658188, "loss": 0.7529, "step": 4023 }, { "epoch": 0.23, "learning_rate": 0.0001800841409372525, "loss": 0.7239, "step": 4024 }, { "epoch": 0.23, "learning_rate": 0.00018007313709487334, "loss": 0.7883, "step": 4025 }, { "epoch": 0.23, "learning_rate": 0.00018006213054981582, "loss": 0.7385, "step": 4026 }, { "epoch": 0.23, "learning_rate": 0.00018005112130245143, "loss": 0.758, "step": 4027 }, { "epoch": 0.23, "learning_rate": 0.00018004010935315171, "loss": 0.6911, "step": 4028 }, { "epoch": 0.23, "learning_rate": 0.00018002909470228842, "loss": 0.7618, "step": 4029 }, { "epoch": 0.23, "learning_rate": 0.0001800180773502333, "loss": 0.706, "step": 4030 }, { "epoch": 0.23, "learning_rate": 0.0001800070572973582, "loss": 0.7242, "step": 4031 }, { "epoch": 0.23, "learning_rate": 0.0001799960345440351, "loss": 0.7635, "step": 4032 }, { "epoch": 0.23, "learning_rate": 0.00017998500909063603, "loss": 0.7421, "step": 4033 }, { "epoch": 0.23, "learning_rate": 0.00017997398093753313, "loss": 0.7707, "step": 4034 }, { "epoch": 0.23, "learning_rate": 0.00017996295008509866, "loss": 0.7722, "step": 4035 }, { "epoch": 0.23, "learning_rate": 0.00017995191653370486, "loss": 0.6805, "step": 4036 }, { "epoch": 0.23, "learning_rate": 0.00017994088028372424, "loss": 0.7718, "step": 4037 }, { "epoch": 0.23, "learning_rate": 0.00017992984133552927, "loss": 0.6612, "step": 4038 }, { "epoch": 0.23, "learning_rate": 0.0001799187996894925, "loss": 0.7255, "step": 4039 }, { "epoch": 0.23, "learning_rate": 0.00017990775534598664, "loss": 0.7556, "step": 4040 }, { "epoch": 0.23, "learning_rate": 0.0001798967083053845, "loss": 0.5869, "step": 4041 }, { "epoch": 0.23, "learning_rate": 0.00017988565856805888, "loss": 0.7297, "step": 4042 }, { "epoch": 0.23, "learning_rate": 0.00017987460613438282, "loss": 0.6358, "step": 4043 }, { "epoch": 0.23, "learning_rate": 0.00017986355100472928, "loss": 0.7896, "step": 4044 }, { "epoch": 0.23, "learning_rate": 0.00017985249317947148, "loss": 0.7455, "step": 4045 }, { "epoch": 0.23, "learning_rate": 0.00017984143265898262, "loss": 0.7014, "step": 4046 }, { "epoch": 0.23, "learning_rate": 0.000179830369443636, "loss": 0.7137, "step": 4047 }, { "epoch": 0.23, "learning_rate": 0.00017981930353380503, "loss": 0.7484, "step": 4048 }, { "epoch": 0.23, "learning_rate": 0.00017980823492986324, "loss": 0.7663, "step": 4049 }, { "epoch": 0.23, "learning_rate": 0.00017979716363218423, "loss": 0.7557, "step": 4050 }, { "epoch": 0.23, "learning_rate": 0.00017978608964114168, "loss": 0.7835, "step": 4051 }, { "epoch": 0.23, "learning_rate": 0.00017977501295710934, "loss": 0.7058, "step": 4052 }, { "epoch": 0.23, "learning_rate": 0.00017976393358046109, "loss": 0.7726, "step": 4053 }, { "epoch": 0.23, "learning_rate": 0.0001797528515115709, "loss": 0.8072, "step": 4054 }, { "epoch": 0.23, "learning_rate": 0.00017974176675081282, "loss": 0.7093, "step": 4055 }, { "epoch": 0.23, "learning_rate": 0.00017973067929856097, "loss": 0.6814, "step": 4056 }, { "epoch": 0.23, "learning_rate": 0.0001797195891551896, "loss": 0.6512, "step": 4057 }, { "epoch": 0.23, "learning_rate": 0.00017970849632107302, "loss": 0.7593, "step": 4058 }, { "epoch": 0.23, "learning_rate": 0.00017969740079658566, "loss": 0.7254, "step": 4059 }, { "epoch": 0.23, "learning_rate": 0.00017968630258210198, "loss": 0.7867, "step": 4060 }, { "epoch": 0.23, "learning_rate": 0.0001796752016779966, "loss": 0.7796, "step": 4061 }, { "epoch": 0.23, "learning_rate": 0.00017966409808464423, "loss": 0.7239, "step": 4062 }, { "epoch": 0.23, "learning_rate": 0.00017965299180241963, "loss": 0.7006, "step": 4063 }, { "epoch": 0.23, "learning_rate": 0.00017964188283169766, "loss": 0.6713, "step": 4064 }, { "epoch": 0.23, "learning_rate": 0.00017963077117285326, "loss": 0.7606, "step": 4065 }, { "epoch": 0.23, "learning_rate": 0.0001796196568262615, "loss": 0.7332, "step": 4066 }, { "epoch": 0.23, "learning_rate": 0.00017960853979229752, "loss": 0.7955, "step": 4067 }, { "epoch": 0.23, "learning_rate": 0.00017959742007133653, "loss": 0.6056, "step": 4068 }, { "epoch": 0.23, "learning_rate": 0.00017958629766375386, "loss": 0.7164, "step": 4069 }, { "epoch": 0.23, "learning_rate": 0.00017957517256992494, "loss": 0.6726, "step": 4070 }, { "epoch": 0.23, "learning_rate": 0.0001795640447902252, "loss": 0.7663, "step": 4071 }, { "epoch": 0.23, "learning_rate": 0.00017955291432503035, "loss": 0.7127, "step": 4072 }, { "epoch": 0.23, "learning_rate": 0.00017954178117471597, "loss": 0.6697, "step": 4073 }, { "epoch": 0.23, "learning_rate": 0.00017953064533965786, "loss": 0.8052, "step": 4074 }, { "epoch": 0.23, "learning_rate": 0.00017951950682023191, "loss": 0.7306, "step": 4075 }, { "epoch": 0.23, "learning_rate": 0.00017950836561681405, "loss": 0.6687, "step": 4076 }, { "epoch": 0.23, "learning_rate": 0.00017949722172978035, "loss": 0.8068, "step": 4077 }, { "epoch": 0.23, "learning_rate": 0.0001794860751595069, "loss": 0.829, "step": 4078 }, { "epoch": 0.23, "learning_rate": 0.00017947492590637, "loss": 0.681, "step": 4079 }, { "epoch": 0.23, "learning_rate": 0.00017946377397074588, "loss": 0.7835, "step": 4080 }, { "epoch": 0.23, "learning_rate": 0.00017945261935301096, "loss": 0.6408, "step": 4081 }, { "epoch": 0.23, "learning_rate": 0.00017944146205354182, "loss": 0.5794, "step": 4082 }, { "epoch": 0.23, "learning_rate": 0.00017943030207271497, "loss": 0.7173, "step": 4083 }, { "epoch": 0.23, "learning_rate": 0.0001794191394109071, "loss": 0.616, "step": 4084 }, { "epoch": 0.23, "learning_rate": 0.00017940797406849502, "loss": 0.6978, "step": 4085 }, { "epoch": 0.23, "learning_rate": 0.00017939680604585556, "loss": 0.7315, "step": 4086 }, { "epoch": 0.23, "learning_rate": 0.00017938563534336565, "loss": 0.7597, "step": 4087 }, { "epoch": 0.23, "learning_rate": 0.00017937446196140234, "loss": 0.7426, "step": 4088 }, { "epoch": 0.23, "learning_rate": 0.0001793632859003428, "loss": 0.8064, "step": 4089 }, { "epoch": 0.23, "learning_rate": 0.00017935210716056422, "loss": 0.7116, "step": 4090 }, { "epoch": 0.23, "learning_rate": 0.0001793409257424439, "loss": 0.767, "step": 4091 }, { "epoch": 0.23, "learning_rate": 0.00017932974164635925, "loss": 0.7239, "step": 4092 }, { "epoch": 0.23, "learning_rate": 0.00017931855487268782, "loss": 0.7126, "step": 4093 }, { "epoch": 0.23, "learning_rate": 0.00017930736542180707, "loss": 0.6522, "step": 4094 }, { "epoch": 0.23, "learning_rate": 0.00017929617329409477, "loss": 0.8135, "step": 4095 }, { "epoch": 0.23, "learning_rate": 0.0001792849784899287, "loss": 0.6601, "step": 4096 }, { "epoch": 0.23, "learning_rate": 0.00017927378100968664, "loss": 0.709, "step": 4097 }, { "epoch": 0.23, "learning_rate": 0.00017926258085374658, "loss": 0.7052, "step": 4098 }, { "epoch": 0.23, "learning_rate": 0.00017925137802248653, "loss": 0.8138, "step": 4099 }, { "epoch": 0.23, "learning_rate": 0.00017924017251628462, "loss": 0.7862, "step": 4100 }, { "epoch": 0.23, "learning_rate": 0.00017922896433551907, "loss": 0.6196, "step": 4101 }, { "epoch": 0.23, "learning_rate": 0.0001792177534805682, "loss": 0.7428, "step": 4102 }, { "epoch": 0.23, "learning_rate": 0.0001792065399518104, "loss": 0.6516, "step": 4103 }, { "epoch": 0.23, "learning_rate": 0.00017919532374962416, "loss": 0.801, "step": 4104 }, { "epoch": 0.23, "learning_rate": 0.00017918410487438805, "loss": 0.7136, "step": 4105 }, { "epoch": 0.23, "learning_rate": 0.0001791728833264807, "loss": 0.6365, "step": 4106 }, { "epoch": 0.23, "learning_rate": 0.00017916165910628092, "loss": 0.792, "step": 4107 }, { "epoch": 0.23, "learning_rate": 0.0001791504322141675, "loss": 0.7299, "step": 4108 }, { "epoch": 0.23, "learning_rate": 0.00017913920265051947, "loss": 0.6426, "step": 4109 }, { "epoch": 0.23, "learning_rate": 0.00017912797041571576, "loss": 0.7384, "step": 4110 }, { "epoch": 0.23, "learning_rate": 0.00017911673551013551, "loss": 0.8013, "step": 4111 }, { "epoch": 0.23, "learning_rate": 0.000179105497934158, "loss": 0.7934, "step": 4112 }, { "epoch": 0.23, "learning_rate": 0.00017909425768816245, "loss": 0.6183, "step": 4113 }, { "epoch": 0.23, "learning_rate": 0.00017908301477252827, "loss": 0.6609, "step": 4114 }, { "epoch": 0.23, "learning_rate": 0.00017907176918763494, "loss": 0.7652, "step": 4115 }, { "epoch": 0.23, "learning_rate": 0.00017906052093386203, "loss": 0.6004, "step": 4116 }, { "epoch": 0.23, "learning_rate": 0.0001790492700115892, "loss": 0.6769, "step": 4117 }, { "epoch": 0.23, "learning_rate": 0.00017903801642119618, "loss": 0.8274, "step": 4118 }, { "epoch": 0.23, "learning_rate": 0.0001790267601630628, "loss": 0.7323, "step": 4119 }, { "epoch": 0.23, "learning_rate": 0.00017901550123756906, "loss": 0.6996, "step": 4120 }, { "epoch": 0.23, "learning_rate": 0.00017900423964509486, "loss": 0.8448, "step": 4121 }, { "epoch": 0.23, "learning_rate": 0.00017899297538602046, "loss": 0.6379, "step": 4122 }, { "epoch": 0.23, "learning_rate": 0.00017898170846072592, "loss": 0.7398, "step": 4123 }, { "epoch": 0.23, "learning_rate": 0.0001789704388695916, "loss": 0.7429, "step": 4124 }, { "epoch": 0.23, "learning_rate": 0.00017895916661299786, "loss": 0.7012, "step": 4125 }, { "epoch": 0.23, "learning_rate": 0.00017894789169132515, "loss": 0.8599, "step": 4126 }, { "epoch": 0.23, "learning_rate": 0.00017893661410495405, "loss": 0.7492, "step": 4127 }, { "epoch": 0.23, "learning_rate": 0.00017892533385426522, "loss": 0.7286, "step": 4128 }, { "epoch": 0.23, "learning_rate": 0.00017891405093963938, "loss": 0.6926, "step": 4129 }, { "epoch": 0.23, "learning_rate": 0.00017890276536145736, "loss": 0.7422, "step": 4130 }, { "epoch": 0.23, "learning_rate": 0.00017889147712010005, "loss": 0.7136, "step": 4131 }, { "epoch": 0.23, "learning_rate": 0.00017888018621594853, "loss": 0.7871, "step": 4132 }, { "epoch": 0.23, "learning_rate": 0.00017886889264938385, "loss": 0.7158, "step": 4133 }, { "epoch": 0.23, "learning_rate": 0.00017885759642078717, "loss": 0.7344, "step": 4134 }, { "epoch": 0.23, "learning_rate": 0.00017884629753053984, "loss": 0.6908, "step": 4135 }, { "epoch": 0.23, "learning_rate": 0.00017883499597902316, "loss": 0.7529, "step": 4136 }, { "epoch": 0.23, "learning_rate": 0.00017882369176661862, "loss": 0.9182, "step": 4137 }, { "epoch": 0.23, "learning_rate": 0.00017881238489370776, "loss": 0.7564, "step": 4138 }, { "epoch": 0.23, "learning_rate": 0.00017880107536067218, "loss": 0.8013, "step": 4139 }, { "epoch": 0.23, "learning_rate": 0.0001787897631678937, "loss": 0.7178, "step": 4140 }, { "epoch": 0.23, "learning_rate": 0.00017877844831575406, "loss": 0.7223, "step": 4141 }, { "epoch": 0.23, "learning_rate": 0.0001787671308046352, "loss": 0.7527, "step": 4142 }, { "epoch": 0.24, "learning_rate": 0.00017875581063491906, "loss": 0.8219, "step": 4143 }, { "epoch": 0.24, "learning_rate": 0.0001787444878069878, "loss": 0.8221, "step": 4144 }, { "epoch": 0.24, "learning_rate": 0.00017873316232122356, "loss": 0.6675, "step": 4145 }, { "epoch": 0.24, "learning_rate": 0.00017872183417800858, "loss": 0.6901, "step": 4146 }, { "epoch": 0.24, "learning_rate": 0.00017871050337772525, "loss": 0.7793, "step": 4147 }, { "epoch": 0.24, "learning_rate": 0.00017869916992075603, "loss": 0.7541, "step": 4148 }, { "epoch": 0.24, "learning_rate": 0.00017868783380748342, "loss": 0.6979, "step": 4149 }, { "epoch": 0.24, "learning_rate": 0.00017867649503829005, "loss": 0.7152, "step": 4150 }, { "epoch": 0.24, "learning_rate": 0.0001786651536135586, "loss": 0.6561, "step": 4151 }, { "epoch": 0.24, "learning_rate": 0.00017865380953367198, "loss": 0.7577, "step": 4152 }, { "epoch": 0.24, "learning_rate": 0.00017864246279901297, "loss": 0.7683, "step": 4153 }, { "epoch": 0.24, "learning_rate": 0.00017863111340996458, "loss": 0.733, "step": 4154 }, { "epoch": 0.24, "learning_rate": 0.00017861976136690993, "loss": 0.7004, "step": 4155 }, { "epoch": 0.24, "learning_rate": 0.00017860840667023212, "loss": 0.6761, "step": 4156 }, { "epoch": 0.24, "learning_rate": 0.00017859704932031445, "loss": 0.644, "step": 4157 }, { "epoch": 0.24, "learning_rate": 0.0001785856893175402, "loss": 0.6281, "step": 4158 }, { "epoch": 0.24, "learning_rate": 0.00017857432666229287, "loss": 0.8114, "step": 4159 }, { "epoch": 0.24, "learning_rate": 0.0001785629613549559, "loss": 0.7825, "step": 4160 }, { "epoch": 0.24, "learning_rate": 0.000178551593395913, "loss": 0.6827, "step": 4161 }, { "epoch": 0.24, "learning_rate": 0.00017854022278554778, "loss": 0.739, "step": 4162 }, { "epoch": 0.24, "learning_rate": 0.0001785288495242441, "loss": 0.7348, "step": 4163 }, { "epoch": 0.24, "learning_rate": 0.00017851747361238574, "loss": 0.6948, "step": 4164 }, { "epoch": 0.24, "learning_rate": 0.0001785060950503568, "loss": 0.7746, "step": 4165 }, { "epoch": 0.24, "learning_rate": 0.0001784947138385412, "loss": 0.6632, "step": 4166 }, { "epoch": 0.24, "learning_rate": 0.00017848332997732317, "loss": 0.7141, "step": 4167 }, { "epoch": 0.24, "learning_rate": 0.0001784719434670869, "loss": 0.6652, "step": 4168 }, { "epoch": 0.24, "learning_rate": 0.00017846055430821678, "loss": 0.7582, "step": 4169 }, { "epoch": 0.24, "learning_rate": 0.00017844916250109714, "loss": 0.6625, "step": 4170 }, { "epoch": 0.24, "learning_rate": 0.00017843776804611256, "loss": 0.7112, "step": 4171 }, { "epoch": 0.24, "learning_rate": 0.0001784263709436476, "loss": 0.8554, "step": 4172 }, { "epoch": 0.24, "learning_rate": 0.00017841497119408693, "loss": 0.7698, "step": 4173 }, { "epoch": 0.24, "learning_rate": 0.0001784035687978153, "loss": 0.6944, "step": 4174 }, { "epoch": 0.24, "learning_rate": 0.00017839216375521765, "loss": 0.7731, "step": 4175 }, { "epoch": 0.24, "learning_rate": 0.00017838075606667887, "loss": 0.807, "step": 4176 }, { "epoch": 0.24, "learning_rate": 0.000178369345732584, "loss": 0.6148, "step": 4177 }, { "epoch": 0.24, "learning_rate": 0.00017835793275331816, "loss": 0.7694, "step": 4178 }, { "epoch": 0.24, "learning_rate": 0.00017834651712926662, "loss": 0.7526, "step": 4179 }, { "epoch": 0.24, "learning_rate": 0.00017833509886081464, "loss": 0.7258, "step": 4180 }, { "epoch": 0.24, "learning_rate": 0.00017832367794834767, "loss": 0.7884, "step": 4181 }, { "epoch": 0.24, "learning_rate": 0.0001783122543922511, "loss": 0.6267, "step": 4182 }, { "epoch": 0.24, "learning_rate": 0.0001783008281929106, "loss": 0.6946, "step": 4183 }, { "epoch": 0.24, "learning_rate": 0.00017828939935071177, "loss": 0.6682, "step": 4184 }, { "epoch": 0.24, "learning_rate": 0.00017827796786604042, "loss": 0.7332, "step": 4185 }, { "epoch": 0.24, "learning_rate": 0.00017826653373928235, "loss": 0.7738, "step": 4186 }, { "epoch": 0.24, "learning_rate": 0.0001782550969708235, "loss": 0.7115, "step": 4187 }, { "epoch": 0.24, "learning_rate": 0.0001782436575610499, "loss": 0.7568, "step": 4188 }, { "epoch": 0.24, "learning_rate": 0.00017823221551034764, "loss": 0.6195, "step": 4189 }, { "epoch": 0.24, "learning_rate": 0.00017822077081910298, "loss": 0.7526, "step": 4190 }, { "epoch": 0.24, "learning_rate": 0.00017820932348770212, "loss": 0.7901, "step": 4191 }, { "epoch": 0.24, "learning_rate": 0.0001781978735165315, "loss": 0.8235, "step": 4192 }, { "epoch": 0.24, "learning_rate": 0.00017818642090597757, "loss": 0.7201, "step": 4193 }, { "epoch": 0.24, "learning_rate": 0.00017817496565642688, "loss": 0.6607, "step": 4194 }, { "epoch": 0.24, "learning_rate": 0.00017816350776826608, "loss": 0.7113, "step": 4195 }, { "epoch": 0.24, "learning_rate": 0.00017815204724188187, "loss": 0.799, "step": 4196 }, { "epoch": 0.24, "learning_rate": 0.00017814058407766115, "loss": 0.7379, "step": 4197 }, { "epoch": 0.24, "learning_rate": 0.00017812911827599077, "loss": 0.6713, "step": 4198 }, { "epoch": 0.24, "learning_rate": 0.00017811764983725775, "loss": 0.6871, "step": 4199 }, { "epoch": 0.24, "learning_rate": 0.00017810617876184916, "loss": 0.7387, "step": 4200 }, { "epoch": 0.24, "learning_rate": 0.0001780947050501522, "loss": 0.7413, "step": 4201 }, { "epoch": 0.24, "learning_rate": 0.00017808322870255415, "loss": 0.6617, "step": 4202 }, { "epoch": 0.24, "learning_rate": 0.00017807174971944234, "loss": 0.7986, "step": 4203 }, { "epoch": 0.24, "learning_rate": 0.00017806026810120423, "loss": 0.6344, "step": 4204 }, { "epoch": 0.24, "learning_rate": 0.00017804878384822733, "loss": 0.7174, "step": 4205 }, { "epoch": 0.24, "learning_rate": 0.00017803729696089931, "loss": 0.5741, "step": 4206 }, { "epoch": 0.24, "learning_rate": 0.00017802580743960785, "loss": 0.7458, "step": 4207 }, { "epoch": 0.24, "learning_rate": 0.00017801431528474075, "loss": 0.7636, "step": 4208 }, { "epoch": 0.24, "learning_rate": 0.00017800282049668594, "loss": 0.7568, "step": 4209 }, { "epoch": 0.24, "learning_rate": 0.00017799132307583132, "loss": 0.7139, "step": 4210 }, { "epoch": 0.24, "learning_rate": 0.00017797982302256504, "loss": 0.8002, "step": 4211 }, { "epoch": 0.24, "learning_rate": 0.0001779683203372752, "loss": 0.784, "step": 4212 }, { "epoch": 0.24, "learning_rate": 0.00017795681502035012, "loss": 0.8558, "step": 4213 }, { "epoch": 0.24, "learning_rate": 0.00017794530707217803, "loss": 0.7024, "step": 4214 }, { "epoch": 0.24, "learning_rate": 0.00017793379649314744, "loss": 0.7987, "step": 4215 }, { "epoch": 0.24, "learning_rate": 0.0001779222832836468, "loss": 0.7145, "step": 4216 }, { "epoch": 0.24, "learning_rate": 0.00017791076744406477, "loss": 0.6699, "step": 4217 }, { "epoch": 0.24, "learning_rate": 0.00017789924897478998, "loss": 0.7246, "step": 4218 }, { "epoch": 0.24, "learning_rate": 0.00017788772787621126, "loss": 0.7026, "step": 4219 }, { "epoch": 0.24, "learning_rate": 0.00017787620414871745, "loss": 0.7499, "step": 4220 }, { "epoch": 0.24, "learning_rate": 0.0001778646777926975, "loss": 0.7948, "step": 4221 }, { "epoch": 0.24, "learning_rate": 0.0001778531488085405, "loss": 0.7569, "step": 4222 }, { "epoch": 0.24, "learning_rate": 0.00017784161719663552, "loss": 0.7133, "step": 4223 }, { "epoch": 0.24, "learning_rate": 0.00017783008295737182, "loss": 0.7055, "step": 4224 }, { "epoch": 0.24, "learning_rate": 0.0001778185460911387, "loss": 0.7946, "step": 4225 }, { "epoch": 0.24, "learning_rate": 0.00017780700659832558, "loss": 0.6646, "step": 4226 }, { "epoch": 0.24, "learning_rate": 0.0001777954644793219, "loss": 0.768, "step": 4227 }, { "epoch": 0.24, "learning_rate": 0.0001777839197345173, "loss": 0.766, "step": 4228 }, { "epoch": 0.24, "learning_rate": 0.0001777723723643014, "loss": 0.6536, "step": 4229 }, { "epoch": 0.24, "learning_rate": 0.00017776082236906397, "loss": 0.7731, "step": 4230 }, { "epoch": 0.24, "learning_rate": 0.00017774926974919485, "loss": 0.6764, "step": 4231 }, { "epoch": 0.24, "learning_rate": 0.00017773771450508397, "loss": 0.7702, "step": 4232 }, { "epoch": 0.24, "learning_rate": 0.00017772615663712136, "loss": 0.6693, "step": 4233 }, { "epoch": 0.24, "learning_rate": 0.0001777145961456971, "loss": 0.7784, "step": 4234 }, { "epoch": 0.24, "learning_rate": 0.00017770303303120142, "loss": 0.6964, "step": 4235 }, { "epoch": 0.24, "learning_rate": 0.0001776914672940246, "loss": 0.8461, "step": 4236 }, { "epoch": 0.24, "learning_rate": 0.00017767989893455698, "loss": 0.6333, "step": 4237 }, { "epoch": 0.24, "learning_rate": 0.00017766832795318904, "loss": 0.7365, "step": 4238 }, { "epoch": 0.24, "learning_rate": 0.00017765675435031138, "loss": 0.6178, "step": 4239 }, { "epoch": 0.24, "learning_rate": 0.00017764517812631457, "loss": 0.7266, "step": 4240 }, { "epoch": 0.24, "learning_rate": 0.00017763359928158937, "loss": 0.7407, "step": 4241 }, { "epoch": 0.24, "learning_rate": 0.0001776220178165266, "loss": 0.7595, "step": 4242 }, { "epoch": 0.24, "learning_rate": 0.00017761043373151715, "loss": 0.7643, "step": 4243 }, { "epoch": 0.24, "learning_rate": 0.00017759884702695203, "loss": 0.7552, "step": 4244 }, { "epoch": 0.24, "learning_rate": 0.0001775872577032223, "loss": 0.7984, "step": 4245 }, { "epoch": 0.24, "learning_rate": 0.00017757566576071914, "loss": 0.7892, "step": 4246 }, { "epoch": 0.24, "learning_rate": 0.00017756407119983383, "loss": 0.7218, "step": 4247 }, { "epoch": 0.24, "learning_rate": 0.0001775524740209577, "loss": 0.6856, "step": 4248 }, { "epoch": 0.24, "learning_rate": 0.00017754087422448215, "loss": 0.7201, "step": 4249 }, { "epoch": 0.24, "learning_rate": 0.00017752927181079878, "loss": 0.7682, "step": 4250 }, { "epoch": 0.24, "learning_rate": 0.00017751766678029913, "loss": 0.6129, "step": 4251 }, { "epoch": 0.24, "learning_rate": 0.00017750605913337495, "loss": 0.7594, "step": 4252 }, { "epoch": 0.24, "learning_rate": 0.00017749444887041799, "loss": 0.7216, "step": 4253 }, { "epoch": 0.24, "learning_rate": 0.00017748283599182014, "loss": 0.6942, "step": 4254 }, { "epoch": 0.24, "learning_rate": 0.00017747122049797335, "loss": 0.7516, "step": 4255 }, { "epoch": 0.24, "learning_rate": 0.00017745960238926974, "loss": 0.553, "step": 4256 }, { "epoch": 0.24, "learning_rate": 0.00017744798166610136, "loss": 0.8057, "step": 4257 }, { "epoch": 0.24, "learning_rate": 0.0001774363583288605, "loss": 0.7667, "step": 4258 }, { "epoch": 0.24, "learning_rate": 0.0001774247323779395, "loss": 0.7662, "step": 4259 }, { "epoch": 0.24, "learning_rate": 0.0001774131038137307, "loss": 0.7741, "step": 4260 }, { "epoch": 0.24, "learning_rate": 0.0001774014726366266, "loss": 0.7122, "step": 4261 }, { "epoch": 0.24, "learning_rate": 0.00017738983884701983, "loss": 0.7898, "step": 4262 }, { "epoch": 0.24, "learning_rate": 0.00017737820244530302, "loss": 0.7099, "step": 4263 }, { "epoch": 0.24, "learning_rate": 0.00017736656343186896, "loss": 0.7446, "step": 4264 }, { "epoch": 0.24, "learning_rate": 0.0001773549218071105, "loss": 0.6932, "step": 4265 }, { "epoch": 0.24, "learning_rate": 0.00017734327757142052, "loss": 0.7275, "step": 4266 }, { "epoch": 0.24, "learning_rate": 0.0001773316307251921, "loss": 0.8131, "step": 4267 }, { "epoch": 0.24, "learning_rate": 0.00017731998126881832, "loss": 0.7675, "step": 4268 }, { "epoch": 0.24, "learning_rate": 0.0001773083292026924, "loss": 0.7056, "step": 4269 }, { "epoch": 0.24, "learning_rate": 0.00017729667452720765, "loss": 0.7602, "step": 4270 }, { "epoch": 0.24, "learning_rate": 0.00017728501724275736, "loss": 0.676, "step": 4271 }, { "epoch": 0.24, "learning_rate": 0.00017727335734973512, "loss": 0.7752, "step": 4272 }, { "epoch": 0.24, "learning_rate": 0.00017726169484853438, "loss": 0.8836, "step": 4273 }, { "epoch": 0.24, "learning_rate": 0.0001772500297395488, "loss": 0.6906, "step": 4274 }, { "epoch": 0.24, "learning_rate": 0.00017723836202317213, "loss": 0.7777, "step": 4275 }, { "epoch": 0.24, "learning_rate": 0.0001772266916997982, "loss": 0.702, "step": 4276 }, { "epoch": 0.24, "learning_rate": 0.00017721501876982084, "loss": 0.8026, "step": 4277 }, { "epoch": 0.24, "learning_rate": 0.00017720334323363413, "loss": 0.6363, "step": 4278 }, { "epoch": 0.24, "learning_rate": 0.0001771916650916321, "loss": 0.7049, "step": 4279 }, { "epoch": 0.24, "learning_rate": 0.00017717998434420892, "loss": 0.7808, "step": 4280 }, { "epoch": 0.24, "learning_rate": 0.0001771683009917589, "loss": 0.7308, "step": 4281 }, { "epoch": 0.24, "learning_rate": 0.0001771566150346763, "loss": 0.7994, "step": 4282 }, { "epoch": 0.24, "learning_rate": 0.00017714492647335562, "loss": 0.7175, "step": 4283 }, { "epoch": 0.24, "learning_rate": 0.0001771332353081913, "loss": 0.7318, "step": 4284 }, { "epoch": 0.24, "learning_rate": 0.00017712154153957805, "loss": 0.7495, "step": 4285 }, { "epoch": 0.24, "learning_rate": 0.00017710984516791048, "loss": 0.7204, "step": 4286 }, { "epoch": 0.24, "learning_rate": 0.00017709814619358343, "loss": 0.7128, "step": 4287 }, { "epoch": 0.24, "learning_rate": 0.00017708644461699176, "loss": 0.6542, "step": 4288 }, { "epoch": 0.24, "learning_rate": 0.00017707474043853041, "loss": 0.7062, "step": 4289 }, { "epoch": 0.24, "learning_rate": 0.00017706303365859442, "loss": 0.7596, "step": 4290 }, { "epoch": 0.24, "learning_rate": 0.00017705132427757895, "loss": 0.6992, "step": 4291 }, { "epoch": 0.24, "learning_rate": 0.0001770396122958792, "loss": 0.7851, "step": 4292 }, { "epoch": 0.24, "learning_rate": 0.00017702789771389049, "loss": 0.7714, "step": 4293 }, { "epoch": 0.24, "learning_rate": 0.0001770161805320082, "loss": 0.7628, "step": 4294 }, { "epoch": 0.24, "learning_rate": 0.00017700446075062786, "loss": 0.7141, "step": 4295 }, { "epoch": 0.24, "learning_rate": 0.00017699273837014503, "loss": 0.6676, "step": 4296 }, { "epoch": 0.24, "learning_rate": 0.00017698101339095536, "loss": 0.8898, "step": 4297 }, { "epoch": 0.24, "learning_rate": 0.00017696928581345457, "loss": 0.7988, "step": 4298 }, { "epoch": 0.24, "learning_rate": 0.00017695755563803856, "loss": 0.6404, "step": 4299 }, { "epoch": 0.24, "learning_rate": 0.0001769458228651032, "loss": 0.6748, "step": 4300 }, { "epoch": 0.24, "learning_rate": 0.00017693408749504453, "loss": 0.7588, "step": 4301 }, { "epoch": 0.24, "learning_rate": 0.00017692234952825862, "loss": 0.7844, "step": 4302 }, { "epoch": 0.24, "learning_rate": 0.0001769106089651417, "loss": 0.7842, "step": 4303 }, { "epoch": 0.24, "learning_rate": 0.00017689886580608998, "loss": 0.84, "step": 4304 }, { "epoch": 0.24, "learning_rate": 0.0001768871200514999, "loss": 0.7397, "step": 4305 }, { "epoch": 0.24, "learning_rate": 0.00017687537170176787, "loss": 0.8642, "step": 4306 }, { "epoch": 0.24, "learning_rate": 0.00017686362075729044, "loss": 0.7685, "step": 4307 }, { "epoch": 0.24, "learning_rate": 0.0001768518672184642, "loss": 0.7096, "step": 4308 }, { "epoch": 0.24, "learning_rate": 0.00017684011108568592, "loss": 0.704, "step": 4309 }, { "epoch": 0.24, "learning_rate": 0.00017682835235935236, "loss": 0.863, "step": 4310 }, { "epoch": 0.24, "learning_rate": 0.0001768165910398604, "loss": 0.7871, "step": 4311 }, { "epoch": 0.24, "learning_rate": 0.00017680482712760704, "loss": 0.6094, "step": 4312 }, { "epoch": 0.24, "learning_rate": 0.00017679306062298933, "loss": 0.7437, "step": 4313 }, { "epoch": 0.24, "learning_rate": 0.00017678129152640448, "loss": 0.6102, "step": 4314 }, { "epoch": 0.24, "learning_rate": 0.00017676951983824964, "loss": 0.7682, "step": 4315 }, { "epoch": 0.24, "learning_rate": 0.00017675774555892215, "loss": 0.6765, "step": 4316 }, { "epoch": 0.24, "learning_rate": 0.00017674596868881944, "loss": 0.7852, "step": 4317 }, { "epoch": 0.24, "learning_rate": 0.00017673418922833903, "loss": 0.7102, "step": 4318 }, { "epoch": 0.24, "learning_rate": 0.00017672240717787852, "loss": 0.8515, "step": 4319 }, { "epoch": 0.25, "learning_rate": 0.00017671062253783553, "loss": 0.7032, "step": 4320 }, { "epoch": 0.25, "learning_rate": 0.00017669883530860783, "loss": 0.7575, "step": 4321 }, { "epoch": 0.25, "learning_rate": 0.0001766870454905933, "loss": 0.7826, "step": 4322 }, { "epoch": 0.25, "learning_rate": 0.0001766752530841899, "loss": 0.6959, "step": 4323 }, { "epoch": 0.25, "learning_rate": 0.0001766634580897956, "loss": 0.7366, "step": 4324 }, { "epoch": 0.25, "learning_rate": 0.0001766516605078085, "loss": 0.7572, "step": 4325 }, { "epoch": 0.25, "learning_rate": 0.00017663986033862684, "loss": 0.7903, "step": 4326 }, { "epoch": 0.25, "learning_rate": 0.00017662805758264893, "loss": 0.7652, "step": 4327 }, { "epoch": 0.25, "learning_rate": 0.0001766162522402731, "loss": 0.7646, "step": 4328 }, { "epoch": 0.25, "learning_rate": 0.0001766044443118978, "loss": 0.6881, "step": 4329 }, { "epoch": 0.25, "learning_rate": 0.00017659263379792164, "loss": 0.7295, "step": 4330 }, { "epoch": 0.25, "learning_rate": 0.0001765808206987432, "loss": 0.7205, "step": 4331 }, { "epoch": 0.25, "learning_rate": 0.0001765690050147612, "loss": 0.6301, "step": 4332 }, { "epoch": 0.25, "learning_rate": 0.0001765571867463745, "loss": 0.686, "step": 4333 }, { "epoch": 0.25, "learning_rate": 0.00017654536589398195, "loss": 0.8561, "step": 4334 }, { "epoch": 0.25, "learning_rate": 0.00017653354245798257, "loss": 0.6773, "step": 4335 }, { "epoch": 0.25, "learning_rate": 0.0001765217164387754, "loss": 0.7829, "step": 4336 }, { "epoch": 0.25, "learning_rate": 0.00017650988783675964, "loss": 0.7301, "step": 4337 }, { "epoch": 0.25, "learning_rate": 0.00017649805665233447, "loss": 0.6888, "step": 4338 }, { "epoch": 0.25, "learning_rate": 0.0001764862228858993, "loss": 0.7615, "step": 4339 }, { "epoch": 0.25, "learning_rate": 0.00017647438653785351, "loss": 0.7627, "step": 4340 }, { "epoch": 0.25, "learning_rate": 0.0001764625476085966, "loss": 0.7697, "step": 4341 }, { "epoch": 0.25, "learning_rate": 0.00017645070609852818, "loss": 0.7062, "step": 4342 }, { "epoch": 0.25, "learning_rate": 0.00017643886200804793, "loss": 0.7634, "step": 4343 }, { "epoch": 0.25, "learning_rate": 0.00017642701533755566, "loss": 0.8001, "step": 4344 }, { "epoch": 0.25, "learning_rate": 0.00017641516608745114, "loss": 0.8038, "step": 4345 }, { "epoch": 0.25, "learning_rate": 0.00017640331425813438, "loss": 0.7686, "step": 4346 }, { "epoch": 0.25, "learning_rate": 0.00017639145985000532, "loss": 0.6961, "step": 4347 }, { "epoch": 0.25, "learning_rate": 0.00017637960286346425, "loss": 0.6768, "step": 4348 }, { "epoch": 0.25, "learning_rate": 0.00017636774329891122, "loss": 0.8066, "step": 4349 }, { "epoch": 0.25, "learning_rate": 0.00017635588115674656, "loss": 0.7834, "step": 4350 }, { "epoch": 0.25, "learning_rate": 0.00017634401643737066, "loss": 0.7982, "step": 4351 }, { "epoch": 0.25, "learning_rate": 0.000176332149141184, "loss": 0.7553, "step": 4352 }, { "epoch": 0.25, "learning_rate": 0.00017632027926858714, "loss": 0.7075, "step": 4353 }, { "epoch": 0.25, "learning_rate": 0.00017630840681998066, "loss": 0.721, "step": 4354 }, { "epoch": 0.25, "learning_rate": 0.00017629653179576535, "loss": 0.7315, "step": 4355 }, { "epoch": 0.25, "learning_rate": 0.00017628465419634198, "loss": 0.7668, "step": 4356 }, { "epoch": 0.25, "learning_rate": 0.00017627277402211146, "loss": 0.7911, "step": 4357 }, { "epoch": 0.25, "learning_rate": 0.0001762608912734748, "loss": 0.7742, "step": 4358 }, { "epoch": 0.25, "learning_rate": 0.00017624900595083307, "loss": 0.7219, "step": 4359 }, { "epoch": 0.25, "learning_rate": 0.0001762371180545874, "loss": 0.7416, "step": 4360 }, { "epoch": 0.25, "learning_rate": 0.00017622522758513904, "loss": 0.7467, "step": 4361 }, { "epoch": 0.25, "learning_rate": 0.00017621333454288936, "loss": 0.7119, "step": 4362 }, { "epoch": 0.25, "learning_rate": 0.00017620143892823977, "loss": 0.7546, "step": 4363 }, { "epoch": 0.25, "learning_rate": 0.00017618954074159177, "loss": 0.809, "step": 4364 }, { "epoch": 0.25, "learning_rate": 0.00017617763998334696, "loss": 0.8454, "step": 4365 }, { "epoch": 0.25, "learning_rate": 0.000176165736653907, "loss": 0.6329, "step": 4366 }, { "epoch": 0.25, "learning_rate": 0.0001761538307536737, "loss": 0.7289, "step": 4367 }, { "epoch": 0.25, "learning_rate": 0.00017614192228304883, "loss": 0.778, "step": 4368 }, { "epoch": 0.25, "learning_rate": 0.00017613001124243446, "loss": 0.6862, "step": 4369 }, { "epoch": 0.25, "learning_rate": 0.00017611809763223252, "loss": 0.8041, "step": 4370 }, { "epoch": 0.25, "learning_rate": 0.00017610618145284516, "loss": 0.719, "step": 4371 }, { "epoch": 0.25, "learning_rate": 0.00017609426270467462, "loss": 0.6484, "step": 4372 }, { "epoch": 0.25, "learning_rate": 0.0001760823413881231, "loss": 0.7801, "step": 4373 }, { "epoch": 0.25, "learning_rate": 0.00017607041750359304, "loss": 0.8262, "step": 4374 }, { "epoch": 0.25, "learning_rate": 0.0001760584910514869, "loss": 0.7128, "step": 4375 }, { "epoch": 0.25, "learning_rate": 0.0001760465620322072, "loss": 0.6917, "step": 4376 }, { "epoch": 0.25, "learning_rate": 0.0001760346304461566, "loss": 0.8358, "step": 4377 }, { "epoch": 0.25, "learning_rate": 0.00017602269629373783, "loss": 0.7851, "step": 4378 }, { "epoch": 0.25, "learning_rate": 0.00017601075957535364, "loss": 0.7657, "step": 4379 }, { "epoch": 0.25, "learning_rate": 0.000175998820291407, "loss": 0.7039, "step": 4380 }, { "epoch": 0.25, "learning_rate": 0.00017598687844230088, "loss": 0.7084, "step": 4381 }, { "epoch": 0.25, "learning_rate": 0.00017597493402843827, "loss": 0.8064, "step": 4382 }, { "epoch": 0.25, "learning_rate": 0.00017596298705022245, "loss": 0.7244, "step": 4383 }, { "epoch": 0.25, "learning_rate": 0.00017595103750805654, "loss": 0.7294, "step": 4384 }, { "epoch": 0.25, "learning_rate": 0.00017593908540234398, "loss": 0.6953, "step": 4385 }, { "epoch": 0.25, "learning_rate": 0.00017592713073348807, "loss": 0.7373, "step": 4386 }, { "epoch": 0.25, "learning_rate": 0.00017591517350189242, "loss": 0.7382, "step": 4387 }, { "epoch": 0.25, "learning_rate": 0.00017590321370796055, "loss": 0.6879, "step": 4388 }, { "epoch": 0.25, "learning_rate": 0.00017589125135209616, "loss": 0.7454, "step": 4389 }, { "epoch": 0.25, "learning_rate": 0.000175879286434703, "loss": 0.7312, "step": 4390 }, { "epoch": 0.25, "learning_rate": 0.00017586731895618494, "loss": 0.7902, "step": 4391 }, { "epoch": 0.25, "learning_rate": 0.0001758553489169459, "loss": 0.673, "step": 4392 }, { "epoch": 0.25, "learning_rate": 0.00017584337631738986, "loss": 0.7255, "step": 4393 }, { "epoch": 0.25, "learning_rate": 0.00017583140115792097, "loss": 0.7172, "step": 4394 }, { "epoch": 0.25, "learning_rate": 0.00017581942343894343, "loss": 0.7626, "step": 4395 }, { "epoch": 0.25, "learning_rate": 0.0001758074431608615, "loss": 0.7828, "step": 4396 }, { "epoch": 0.25, "learning_rate": 0.00017579546032407955, "loss": 0.759, "step": 4397 }, { "epoch": 0.25, "learning_rate": 0.00017578347492900202, "loss": 0.6389, "step": 4398 }, { "epoch": 0.25, "learning_rate": 0.0001757714869760335, "loss": 0.6097, "step": 4399 }, { "epoch": 0.25, "learning_rate": 0.00017575949646557852, "loss": 0.6533, "step": 4400 }, { "epoch": 0.25, "learning_rate": 0.0001757475033980419, "loss": 0.7002, "step": 4401 }, { "epoch": 0.25, "learning_rate": 0.00017573550777382835, "loss": 0.7907, "step": 4402 }, { "epoch": 0.25, "learning_rate": 0.0001757235095933428, "loss": 0.7933, "step": 4403 }, { "epoch": 0.25, "learning_rate": 0.00017571150885699023, "loss": 0.6607, "step": 4404 }, { "epoch": 0.25, "learning_rate": 0.00017569950556517566, "loss": 0.7414, "step": 4405 }, { "epoch": 0.25, "learning_rate": 0.00017568749971830423, "loss": 0.6793, "step": 4406 }, { "epoch": 0.25, "learning_rate": 0.0001756754913167812, "loss": 0.7651, "step": 4407 }, { "epoch": 0.25, "learning_rate": 0.00017566348036101187, "loss": 0.7, "step": 4408 }, { "epoch": 0.25, "learning_rate": 0.00017565146685140167, "loss": 0.8573, "step": 4409 }, { "epoch": 0.25, "learning_rate": 0.00017563945078835603, "loss": 0.7172, "step": 4410 }, { "epoch": 0.25, "learning_rate": 0.00017562743217228058, "loss": 0.7281, "step": 4411 }, { "epoch": 0.25, "learning_rate": 0.00017561541100358092, "loss": 0.6472, "step": 4412 }, { "epoch": 0.25, "learning_rate": 0.00017560338728266284, "loss": 0.8111, "step": 4413 }, { "epoch": 0.25, "learning_rate": 0.0001755913610099322, "loss": 0.7223, "step": 4414 }, { "epoch": 0.25, "learning_rate": 0.00017557933218579483, "loss": 0.8076, "step": 4415 }, { "epoch": 0.25, "learning_rate": 0.0001755673008106568, "loss": 0.6978, "step": 4416 }, { "epoch": 0.25, "learning_rate": 0.0001755552668849242, "loss": 0.702, "step": 4417 }, { "epoch": 0.25, "learning_rate": 0.00017554323040900317, "loss": 0.7906, "step": 4418 }, { "epoch": 0.25, "learning_rate": 0.0001755311913833, "loss": 0.7323, "step": 4419 }, { "epoch": 0.25, "learning_rate": 0.00017551914980822107, "loss": 0.719, "step": 4420 }, { "epoch": 0.25, "learning_rate": 0.0001755071056841727, "loss": 0.675, "step": 4421 }, { "epoch": 0.25, "learning_rate": 0.00017549505901156155, "loss": 0.7269, "step": 4422 }, { "epoch": 0.25, "learning_rate": 0.00017548300979079414, "loss": 0.7319, "step": 4423 }, { "epoch": 0.25, "learning_rate": 0.00017547095802227723, "loss": 0.6436, "step": 4424 }, { "epoch": 0.25, "learning_rate": 0.0001754589037064175, "loss": 0.6486, "step": 4425 }, { "epoch": 0.25, "learning_rate": 0.0001754468468436219, "loss": 0.7057, "step": 4426 }, { "epoch": 0.25, "learning_rate": 0.00017543478743429732, "loss": 0.7354, "step": 4427 }, { "epoch": 0.25, "learning_rate": 0.00017542272547885088, "loss": 0.7535, "step": 4428 }, { "epoch": 0.25, "learning_rate": 0.00017541066097768963, "loss": 0.7317, "step": 4429 }, { "epoch": 0.25, "learning_rate": 0.0001753985939312208, "loss": 0.6909, "step": 4430 }, { "epoch": 0.25, "learning_rate": 0.0001753865243398517, "loss": 0.7823, "step": 4431 }, { "epoch": 0.25, "learning_rate": 0.00017537445220398967, "loss": 0.7634, "step": 4432 }, { "epoch": 0.25, "learning_rate": 0.0001753623775240422, "loss": 0.6874, "step": 4433 }, { "epoch": 0.25, "learning_rate": 0.0001753503003004169, "loss": 0.6867, "step": 4434 }, { "epoch": 0.25, "learning_rate": 0.00017533822053352128, "loss": 0.6705, "step": 4435 }, { "epoch": 0.25, "learning_rate": 0.00017532613822376315, "loss": 0.776, "step": 4436 }, { "epoch": 0.25, "learning_rate": 0.00017531405337155033, "loss": 0.7545, "step": 4437 }, { "epoch": 0.25, "learning_rate": 0.00017530196597729066, "loss": 0.7604, "step": 4438 }, { "epoch": 0.25, "learning_rate": 0.00017528987604139217, "loss": 0.6848, "step": 4439 }, { "epoch": 0.25, "learning_rate": 0.0001752777835642629, "loss": 0.6332, "step": 4440 }, { "epoch": 0.25, "learning_rate": 0.00017526568854631103, "loss": 0.6582, "step": 4441 }, { "epoch": 0.25, "learning_rate": 0.00017525359098794478, "loss": 0.7483, "step": 4442 }, { "epoch": 0.25, "learning_rate": 0.00017524149088957245, "loss": 0.8515, "step": 4443 }, { "epoch": 0.25, "learning_rate": 0.0001752293882516025, "loss": 0.7461, "step": 4444 }, { "epoch": 0.25, "learning_rate": 0.00017521728307444336, "loss": 0.7112, "step": 4445 }, { "epoch": 0.25, "learning_rate": 0.0001752051753585037, "loss": 0.7134, "step": 4446 }, { "epoch": 0.25, "learning_rate": 0.0001751930651041921, "loss": 0.7531, "step": 4447 }, { "epoch": 0.25, "learning_rate": 0.00017518095231191738, "loss": 0.77, "step": 4448 }, { "epoch": 0.25, "learning_rate": 0.00017516883698208836, "loss": 0.6934, "step": 4449 }, { "epoch": 0.25, "learning_rate": 0.0001751567191151139, "loss": 0.7007, "step": 4450 }, { "epoch": 0.25, "learning_rate": 0.0001751445987114031, "loss": 0.7016, "step": 4451 }, { "epoch": 0.25, "learning_rate": 0.00017513247577136506, "loss": 0.7669, "step": 4452 }, { "epoch": 0.25, "learning_rate": 0.00017512035029540885, "loss": 0.7546, "step": 4453 }, { "epoch": 0.25, "learning_rate": 0.00017510822228394385, "loss": 0.5772, "step": 4454 }, { "epoch": 0.25, "learning_rate": 0.00017509609173737935, "loss": 0.7558, "step": 4455 }, { "epoch": 0.25, "learning_rate": 0.0001750839586561248, "loss": 0.5768, "step": 4456 }, { "epoch": 0.25, "learning_rate": 0.00017507182304058976, "loss": 0.7416, "step": 4457 }, { "epoch": 0.25, "learning_rate": 0.00017505968489118376, "loss": 0.7401, "step": 4458 }, { "epoch": 0.25, "learning_rate": 0.0001750475442083166, "loss": 0.6595, "step": 4459 }, { "epoch": 0.25, "learning_rate": 0.00017503540099239796, "loss": 0.7072, "step": 4460 }, { "epoch": 0.25, "learning_rate": 0.0001750232552438378, "loss": 0.7615, "step": 4461 }, { "epoch": 0.25, "learning_rate": 0.00017501110696304596, "loss": 0.7464, "step": 4462 }, { "epoch": 0.25, "learning_rate": 0.00017499895615043256, "loss": 0.6379, "step": 4463 }, { "epoch": 0.25, "learning_rate": 0.00017498680280640772, "loss": 0.5846, "step": 4464 }, { "epoch": 0.25, "learning_rate": 0.0001749746469313816, "loss": 0.6951, "step": 4465 }, { "epoch": 0.25, "learning_rate": 0.00017496248852576454, "loss": 0.7764, "step": 4466 }, { "epoch": 0.25, "learning_rate": 0.00017495032758996687, "loss": 0.6936, "step": 4467 }, { "epoch": 0.25, "learning_rate": 0.0001749381641243991, "loss": 0.7809, "step": 4468 }, { "epoch": 0.25, "learning_rate": 0.00017492599812947174, "loss": 0.5984, "step": 4469 }, { "epoch": 0.25, "learning_rate": 0.00017491382960559544, "loss": 0.7964, "step": 4470 }, { "epoch": 0.25, "learning_rate": 0.00017490165855318094, "loss": 0.7277, "step": 4471 }, { "epoch": 0.25, "learning_rate": 0.00017488948497263902, "loss": 0.7935, "step": 4472 }, { "epoch": 0.25, "learning_rate": 0.00017487730886438056, "loss": 0.7089, "step": 4473 }, { "epoch": 0.25, "learning_rate": 0.0001748651302288166, "loss": 0.7205, "step": 4474 }, { "epoch": 0.25, "learning_rate": 0.0001748529490663581, "loss": 0.6983, "step": 4475 }, { "epoch": 0.25, "learning_rate": 0.00017484076537741628, "loss": 0.6252, "step": 4476 }, { "epoch": 0.25, "learning_rate": 0.00017482857916240236, "loss": 0.7574, "step": 4477 }, { "epoch": 0.25, "learning_rate": 0.00017481639042172766, "loss": 0.7036, "step": 4478 }, { "epoch": 0.25, "learning_rate": 0.00017480419915580356, "loss": 0.6684, "step": 4479 }, { "epoch": 0.25, "learning_rate": 0.00017479200536504156, "loss": 0.7437, "step": 4480 }, { "epoch": 0.25, "learning_rate": 0.0001747798090498532, "loss": 0.8354, "step": 4481 }, { "epoch": 0.25, "learning_rate": 0.00017476761021065017, "loss": 0.6974, "step": 4482 }, { "epoch": 0.25, "learning_rate": 0.00017475540884784424, "loss": 0.734, "step": 4483 }, { "epoch": 0.25, "learning_rate": 0.0001747432049618472, "loss": 0.7386, "step": 4484 }, { "epoch": 0.25, "learning_rate": 0.00017473099855307096, "loss": 0.7564, "step": 4485 }, { "epoch": 0.25, "learning_rate": 0.00017471878962192755, "loss": 0.5895, "step": 4486 }, { "epoch": 0.25, "learning_rate": 0.00017470657816882903, "loss": 0.7577, "step": 4487 }, { "epoch": 0.25, "learning_rate": 0.00017469436419418753, "loss": 0.6656, "step": 4488 }, { "epoch": 0.25, "learning_rate": 0.0001746821476984154, "loss": 0.7466, "step": 4489 }, { "epoch": 0.25, "learning_rate": 0.0001746699286819249, "loss": 0.7103, "step": 4490 }, { "epoch": 0.25, "learning_rate": 0.00017465770714512847, "loss": 0.7017, "step": 4491 }, { "epoch": 0.25, "learning_rate": 0.00017464548308843866, "loss": 0.8233, "step": 4492 }, { "epoch": 0.25, "learning_rate": 0.000174633256512268, "loss": 0.6621, "step": 4493 }, { "epoch": 0.25, "learning_rate": 0.0001746210274170292, "loss": 0.7536, "step": 4494 }, { "epoch": 0.25, "learning_rate": 0.00017460879580313504, "loss": 0.8011, "step": 4495 }, { "epoch": 0.26, "learning_rate": 0.00017459656167099835, "loss": 0.789, "step": 4496 }, { "epoch": 0.26, "learning_rate": 0.00017458432502103204, "loss": 0.7076, "step": 4497 }, { "epoch": 0.26, "learning_rate": 0.00017457208585364918, "loss": 0.7366, "step": 4498 }, { "epoch": 0.26, "learning_rate": 0.00017455984416926286, "loss": 0.7051, "step": 4499 }, { "epoch": 0.26, "learning_rate": 0.00017454759996828623, "loss": 0.6983, "step": 4500 }, { "epoch": 0.26, "learning_rate": 0.0001745353532511326, "loss": 0.8168, "step": 4501 }, { "epoch": 0.26, "learning_rate": 0.00017452310401821534, "loss": 0.6607, "step": 4502 }, { "epoch": 0.26, "learning_rate": 0.00017451085226994785, "loss": 0.8059, "step": 4503 }, { "epoch": 0.26, "learning_rate": 0.00017449859800674371, "loss": 0.678, "step": 4504 }, { "epoch": 0.26, "learning_rate": 0.0001744863412290165, "loss": 0.7567, "step": 4505 }, { "epoch": 0.26, "learning_rate": 0.00017447408193717992, "loss": 0.7837, "step": 4506 }, { "epoch": 0.26, "learning_rate": 0.00017446182013164778, "loss": 0.7054, "step": 4507 }, { "epoch": 0.26, "learning_rate": 0.00017444955581283388, "loss": 0.7737, "step": 4508 }, { "epoch": 0.26, "learning_rate": 0.00017443728898115226, "loss": 0.7617, "step": 4509 }, { "epoch": 0.26, "learning_rate": 0.00017442501963701692, "loss": 0.6978, "step": 4510 }, { "epoch": 0.26, "learning_rate": 0.00017441274778084195, "loss": 0.6235, "step": 4511 }, { "epoch": 0.26, "learning_rate": 0.0001744004734130416, "loss": 0.7606, "step": 4512 }, { "epoch": 0.26, "learning_rate": 0.00017438819653403016, "loss": 0.6978, "step": 4513 }, { "epoch": 0.26, "learning_rate": 0.000174375917144222, "loss": 0.6203, "step": 4514 }, { "epoch": 0.26, "learning_rate": 0.00017436363524403156, "loss": 0.7765, "step": 4515 }, { "epoch": 0.26, "learning_rate": 0.00017435135083387345, "loss": 0.7744, "step": 4516 }, { "epoch": 0.26, "learning_rate": 0.00017433906391416218, "loss": 0.7739, "step": 4517 }, { "epoch": 0.26, "learning_rate": 0.0001743267744853126, "loss": 0.7685, "step": 4518 }, { "epoch": 0.26, "learning_rate": 0.00017431448254773944, "loss": 0.8057, "step": 4519 }, { "epoch": 0.26, "learning_rate": 0.0001743021881018576, "loss": 0.788, "step": 4520 }, { "epoch": 0.26, "learning_rate": 0.000174289891148082, "loss": 0.7157, "step": 4521 }, { "epoch": 0.26, "learning_rate": 0.00017427759168682781, "loss": 0.778, "step": 4522 }, { "epoch": 0.26, "learning_rate": 0.00017426528971851006, "loss": 0.6144, "step": 4523 }, { "epoch": 0.26, "learning_rate": 0.00017425298524354404, "loss": 0.7671, "step": 4524 }, { "epoch": 0.26, "learning_rate": 0.000174240678262345, "loss": 0.7997, "step": 4525 }, { "epoch": 0.26, "learning_rate": 0.0001742283687753284, "loss": 0.7648, "step": 4526 }, { "epoch": 0.26, "learning_rate": 0.00017421605678290963, "loss": 0.7157, "step": 4527 }, { "epoch": 0.26, "learning_rate": 0.00017420374228550438, "loss": 0.7116, "step": 4528 }, { "epoch": 0.26, "learning_rate": 0.00017419142528352817, "loss": 0.7564, "step": 4529 }, { "epoch": 0.26, "learning_rate": 0.0001741791057773968, "loss": 0.7407, "step": 4530 }, { "epoch": 0.26, "learning_rate": 0.00017416678376752604, "loss": 0.7725, "step": 4531 }, { "epoch": 0.26, "learning_rate": 0.00017415445925433184, "loss": 0.7425, "step": 4532 }, { "epoch": 0.26, "learning_rate": 0.00017414213223823017, "loss": 0.7004, "step": 4533 }, { "epoch": 0.26, "learning_rate": 0.0001741298027196371, "loss": 0.8057, "step": 4534 }, { "epoch": 0.26, "learning_rate": 0.00017411747069896876, "loss": 0.7276, "step": 4535 }, { "epoch": 0.26, "learning_rate": 0.0001741051361766414, "loss": 0.713, "step": 4536 }, { "epoch": 0.26, "learning_rate": 0.00017409279915307134, "loss": 0.6705, "step": 4537 }, { "epoch": 0.26, "learning_rate": 0.000174080459628675, "loss": 0.637, "step": 4538 }, { "epoch": 0.26, "learning_rate": 0.0001740681176038689, "loss": 0.742, "step": 4539 }, { "epoch": 0.26, "learning_rate": 0.00017405577307906954, "loss": 0.7433, "step": 4540 }, { "epoch": 0.26, "learning_rate": 0.00017404342605469361, "loss": 0.7395, "step": 4541 }, { "epoch": 0.26, "learning_rate": 0.0001740310765311579, "loss": 0.7062, "step": 4542 }, { "epoch": 0.26, "learning_rate": 0.00017401872450887917, "loss": 0.6824, "step": 4543 }, { "epoch": 0.26, "learning_rate": 0.0001740063699882744, "loss": 0.7854, "step": 4544 }, { "epoch": 0.26, "learning_rate": 0.00017399401296976053, "loss": 0.8126, "step": 4545 }, { "epoch": 0.26, "learning_rate": 0.00017398165345375468, "loss": 0.7598, "step": 4546 }, { "epoch": 0.26, "learning_rate": 0.000173969291440674, "loss": 0.7049, "step": 4547 }, { "epoch": 0.26, "learning_rate": 0.00017395692693093575, "loss": 0.7865, "step": 4548 }, { "epoch": 0.26, "learning_rate": 0.00017394455992495722, "loss": 0.8098, "step": 4549 }, { "epoch": 0.26, "learning_rate": 0.00017393219042315588, "loss": 0.6862, "step": 4550 }, { "epoch": 0.26, "learning_rate": 0.00017391981842594924, "loss": 0.7906, "step": 4551 }, { "epoch": 0.26, "learning_rate": 0.00017390744393375486, "loss": 0.7133, "step": 4552 }, { "epoch": 0.26, "learning_rate": 0.0001738950669469904, "loss": 0.6308, "step": 4553 }, { "epoch": 0.26, "learning_rate": 0.0001738826874660737, "loss": 0.6622, "step": 4554 }, { "epoch": 0.26, "learning_rate": 0.00017387030549142246, "loss": 0.7334, "step": 4555 }, { "epoch": 0.26, "learning_rate": 0.00017385792102345469, "loss": 0.7381, "step": 4556 }, { "epoch": 0.26, "learning_rate": 0.00017384553406258842, "loss": 0.7366, "step": 4557 }, { "epoch": 0.26, "learning_rate": 0.00017383314460924166, "loss": 0.7584, "step": 4558 }, { "epoch": 0.26, "learning_rate": 0.00017382075266383264, "loss": 0.6244, "step": 4559 }, { "epoch": 0.26, "learning_rate": 0.00017380835822677963, "loss": 0.8593, "step": 4560 }, { "epoch": 0.26, "learning_rate": 0.00017379596129850098, "loss": 0.7286, "step": 4561 }, { "epoch": 0.26, "learning_rate": 0.00017378356187941507, "loss": 0.8417, "step": 4562 }, { "epoch": 0.26, "learning_rate": 0.00017377115996994048, "loss": 0.7168, "step": 4563 }, { "epoch": 0.26, "learning_rate": 0.00017375875557049572, "loss": 0.7473, "step": 4564 }, { "epoch": 0.26, "learning_rate": 0.00017374634868149954, "loss": 0.7351, "step": 4565 }, { "epoch": 0.26, "learning_rate": 0.0001737339393033707, "loss": 0.6775, "step": 4566 }, { "epoch": 0.26, "learning_rate": 0.00017372152743652802, "loss": 0.734, "step": 4567 }, { "epoch": 0.26, "learning_rate": 0.00017370911308139047, "loss": 0.6205, "step": 4568 }, { "epoch": 0.26, "learning_rate": 0.00017369669623837702, "loss": 0.6147, "step": 4569 }, { "epoch": 0.26, "learning_rate": 0.00017368427690790677, "loss": 0.8652, "step": 4570 }, { "epoch": 0.26, "learning_rate": 0.00017367185509039898, "loss": 0.7029, "step": 4571 }, { "epoch": 0.26, "learning_rate": 0.00017365943078627284, "loss": 0.6309, "step": 4572 }, { "epoch": 0.26, "learning_rate": 0.00017364700399594775, "loss": 0.6316, "step": 4573 }, { "epoch": 0.26, "learning_rate": 0.00017363457471984313, "loss": 0.6938, "step": 4574 }, { "epoch": 0.26, "learning_rate": 0.0001736221429583785, "loss": 0.6402, "step": 4575 }, { "epoch": 0.26, "learning_rate": 0.00017360970871197346, "loss": 0.7416, "step": 4576 }, { "epoch": 0.26, "learning_rate": 0.00017359727198104768, "loss": 0.7794, "step": 4577 }, { "epoch": 0.26, "learning_rate": 0.00017358483276602097, "loss": 0.675, "step": 4578 }, { "epoch": 0.26, "learning_rate": 0.00017357239106731317, "loss": 0.772, "step": 4579 }, { "epoch": 0.26, "learning_rate": 0.00017355994688534422, "loss": 0.8252, "step": 4580 }, { "epoch": 0.26, "learning_rate": 0.00017354750022053415, "loss": 0.7203, "step": 4581 }, { "epoch": 0.26, "learning_rate": 0.00017353505107330306, "loss": 0.8423, "step": 4582 }, { "epoch": 0.26, "learning_rate": 0.00017352259944407113, "loss": 0.7373, "step": 4583 }, { "epoch": 0.26, "learning_rate": 0.00017351014533325868, "loss": 0.8211, "step": 4584 }, { "epoch": 0.26, "learning_rate": 0.00017349768874128603, "loss": 0.7023, "step": 4585 }, { "epoch": 0.26, "learning_rate": 0.00017348522966857363, "loss": 0.692, "step": 4586 }, { "epoch": 0.26, "learning_rate": 0.00017347276811554196, "loss": 0.8024, "step": 4587 }, { "epoch": 0.26, "learning_rate": 0.00017346030408261172, "loss": 0.824, "step": 4588 }, { "epoch": 0.26, "learning_rate": 0.00017344783757020356, "loss": 0.6953, "step": 4589 }, { "epoch": 0.26, "learning_rate": 0.00017343536857873824, "loss": 0.8142, "step": 4590 }, { "epoch": 0.26, "learning_rate": 0.00017342289710863665, "loss": 0.6629, "step": 4591 }, { "epoch": 0.26, "learning_rate": 0.00017341042316031975, "loss": 0.7221, "step": 4592 }, { "epoch": 0.26, "learning_rate": 0.0001733979467342085, "loss": 0.8482, "step": 4593 }, { "epoch": 0.26, "learning_rate": 0.00017338546783072408, "loss": 0.7468, "step": 4594 }, { "epoch": 0.26, "learning_rate": 0.00017337298645028764, "loss": 0.6904, "step": 4595 }, { "epoch": 0.26, "learning_rate": 0.0001733605025933205, "loss": 0.7029, "step": 4596 }, { "epoch": 0.26, "learning_rate": 0.000173348016260244, "loss": 0.8158, "step": 4597 }, { "epoch": 0.26, "learning_rate": 0.0001733355274514796, "loss": 0.7292, "step": 4598 }, { "epoch": 0.26, "learning_rate": 0.00017332303616744878, "loss": 0.7416, "step": 4599 }, { "epoch": 0.26, "learning_rate": 0.0001733105424085732, "loss": 0.701, "step": 4600 }, { "epoch": 0.26, "learning_rate": 0.00017329804617527458, "loss": 0.6992, "step": 4601 }, { "epoch": 0.26, "learning_rate": 0.00017328554746797465, "loss": 0.729, "step": 4602 }, { "epoch": 0.26, "learning_rate": 0.0001732730462870953, "loss": 0.7522, "step": 4603 }, { "epoch": 0.26, "learning_rate": 0.00017326054263305847, "loss": 0.7794, "step": 4604 }, { "epoch": 0.26, "learning_rate": 0.0001732480365062862, "loss": 0.7, "step": 4605 }, { "epoch": 0.26, "learning_rate": 0.00017323552790720058, "loss": 0.7895, "step": 4606 }, { "epoch": 0.26, "learning_rate": 0.00017322301683622384, "loss": 0.6063, "step": 4607 }, { "epoch": 0.26, "learning_rate": 0.00017321050329377827, "loss": 0.7172, "step": 4608 }, { "epoch": 0.26, "learning_rate": 0.00017319798728028619, "loss": 0.6622, "step": 4609 }, { "epoch": 0.26, "learning_rate": 0.00017318546879617006, "loss": 0.797, "step": 4610 }, { "epoch": 0.26, "learning_rate": 0.00017317294784185242, "loss": 0.7136, "step": 4611 }, { "epoch": 0.26, "learning_rate": 0.0001731604244177559, "loss": 0.666, "step": 4612 }, { "epoch": 0.26, "learning_rate": 0.00017314789852430318, "loss": 0.788, "step": 4613 }, { "epoch": 0.26, "learning_rate": 0.00017313537016191706, "loss": 0.7384, "step": 4614 }, { "epoch": 0.26, "learning_rate": 0.00017312283933102038, "loss": 0.7064, "step": 4615 }, { "epoch": 0.26, "learning_rate": 0.0001731103060320361, "loss": 0.6328, "step": 4616 }, { "epoch": 0.26, "learning_rate": 0.00017309777026538727, "loss": 0.7339, "step": 4617 }, { "epoch": 0.26, "learning_rate": 0.00017308523203149698, "loss": 0.8139, "step": 4618 }, { "epoch": 0.26, "learning_rate": 0.00017307269133078843, "loss": 0.7283, "step": 4619 }, { "epoch": 0.26, "learning_rate": 0.0001730601481636849, "loss": 0.8208, "step": 4620 }, { "epoch": 0.26, "learning_rate": 0.00017304760253060975, "loss": 0.738, "step": 4621 }, { "epoch": 0.26, "learning_rate": 0.0001730350544319865, "loss": 0.8895, "step": 4622 }, { "epoch": 0.26, "learning_rate": 0.00017302250386823858, "loss": 0.7496, "step": 4623 }, { "epoch": 0.26, "learning_rate": 0.00017300995083978965, "loss": 0.7613, "step": 4624 }, { "epoch": 0.26, "learning_rate": 0.0001729973953470634, "loss": 0.6799, "step": 4625 }, { "epoch": 0.26, "learning_rate": 0.0001729848373904836, "loss": 0.6345, "step": 4626 }, { "epoch": 0.26, "learning_rate": 0.00017297227697047414, "loss": 0.6656, "step": 4627 }, { "epoch": 0.26, "learning_rate": 0.00017295971408745896, "loss": 0.6799, "step": 4628 }, { "epoch": 0.26, "learning_rate": 0.0001729471487418621, "loss": 0.7887, "step": 4629 }, { "epoch": 0.26, "learning_rate": 0.00017293458093410764, "loss": 0.721, "step": 4630 }, { "epoch": 0.26, "learning_rate": 0.0001729220106646198, "loss": 0.739, "step": 4631 }, { "epoch": 0.26, "learning_rate": 0.00017290943793382284, "loss": 0.7, "step": 4632 }, { "epoch": 0.26, "learning_rate": 0.00017289686274214118, "loss": 0.7138, "step": 4633 }, { "epoch": 0.26, "learning_rate": 0.00017288428508999915, "loss": 0.7598, "step": 4634 }, { "epoch": 0.26, "learning_rate": 0.0001728717049778214, "loss": 0.7022, "step": 4635 }, { "epoch": 0.26, "learning_rate": 0.00017285912240603253, "loss": 0.7406, "step": 4636 }, { "epoch": 0.26, "learning_rate": 0.00017284653737505715, "loss": 0.6183, "step": 4637 }, { "epoch": 0.26, "learning_rate": 0.00017283394988532012, "loss": 0.7723, "step": 4638 }, { "epoch": 0.26, "learning_rate": 0.00017282135993724622, "loss": 0.7166, "step": 4639 }, { "epoch": 0.26, "learning_rate": 0.00017280876753126046, "loss": 0.7347, "step": 4640 }, { "epoch": 0.26, "learning_rate": 0.00017279617266778788, "loss": 0.7191, "step": 4641 }, { "epoch": 0.26, "learning_rate": 0.0001727835753472535, "loss": 0.6863, "step": 4642 }, { "epoch": 0.26, "learning_rate": 0.0001727709755700826, "loss": 0.6871, "step": 4643 }, { "epoch": 0.26, "learning_rate": 0.00017275837333670046, "loss": 0.6883, "step": 4644 }, { "epoch": 0.26, "learning_rate": 0.00017274576864753237, "loss": 0.8007, "step": 4645 }, { "epoch": 0.26, "learning_rate": 0.00017273316150300382, "loss": 0.7057, "step": 4646 }, { "epoch": 0.26, "learning_rate": 0.0001727205519035403, "loss": 0.6551, "step": 4647 }, { "epoch": 0.26, "learning_rate": 0.00017270793984956742, "loss": 0.672, "step": 4648 }, { "epoch": 0.26, "learning_rate": 0.00017269532534151092, "loss": 0.7352, "step": 4649 }, { "epoch": 0.26, "learning_rate": 0.0001726827083797965, "loss": 0.6231, "step": 4650 }, { "epoch": 0.26, "learning_rate": 0.0001726700889648501, "loss": 0.7242, "step": 4651 }, { "epoch": 0.26, "learning_rate": 0.0001726574670970976, "loss": 0.6936, "step": 4652 }, { "epoch": 0.26, "learning_rate": 0.00017264484277696503, "loss": 0.719, "step": 4653 }, { "epoch": 0.26, "learning_rate": 0.00017263221600487852, "loss": 0.7119, "step": 4654 }, { "epoch": 0.26, "learning_rate": 0.0001726195867812642, "loss": 0.725, "step": 4655 }, { "epoch": 0.26, "learning_rate": 0.00017260695510654838, "loss": 0.7888, "step": 4656 }, { "epoch": 0.26, "learning_rate": 0.00017259432098115743, "loss": 0.7716, "step": 4657 }, { "epoch": 0.26, "learning_rate": 0.00017258168440551774, "loss": 1.0328, "step": 4658 }, { "epoch": 0.26, "learning_rate": 0.00017256904538005585, "loss": 0.7547, "step": 4659 }, { "epoch": 0.26, "learning_rate": 0.00017255640390519836, "loss": 0.7451, "step": 4660 }, { "epoch": 0.26, "learning_rate": 0.00017254375998137196, "loss": 0.7162, "step": 4661 }, { "epoch": 0.26, "learning_rate": 0.0001725311136090034, "loss": 0.697, "step": 4662 }, { "epoch": 0.26, "learning_rate": 0.00017251846478851955, "loss": 0.729, "step": 4663 }, { "epoch": 0.26, "learning_rate": 0.00017250581352034732, "loss": 0.7472, "step": 4664 }, { "epoch": 0.26, "learning_rate": 0.00017249315980491373, "loss": 0.7381, "step": 4665 }, { "epoch": 0.26, "learning_rate": 0.00017248050364264585, "loss": 0.7661, "step": 4666 }, { "epoch": 0.26, "learning_rate": 0.0001724678450339709, "loss": 0.7161, "step": 4667 }, { "epoch": 0.26, "learning_rate": 0.00017245518397931617, "loss": 0.6548, "step": 4668 }, { "epoch": 0.26, "learning_rate": 0.00017244252047910892, "loss": 0.7106, "step": 4669 }, { "epoch": 0.26, "learning_rate": 0.00017242985453377662, "loss": 0.7179, "step": 4670 }, { "epoch": 0.26, "learning_rate": 0.00017241718614374678, "loss": 0.7739, "step": 4671 }, { "epoch": 0.27, "learning_rate": 0.00017240451530944697, "loss": 0.7341, "step": 4672 }, { "epoch": 0.27, "learning_rate": 0.00017239184203130488, "loss": 0.7943, "step": 4673 }, { "epoch": 0.27, "learning_rate": 0.0001723791663097483, "loss": 0.7489, "step": 4674 }, { "epoch": 0.27, "learning_rate": 0.000172366488145205, "loss": 0.7233, "step": 4675 }, { "epoch": 0.27, "learning_rate": 0.00017235380753810296, "loss": 0.8054, "step": 4676 }, { "epoch": 0.27, "learning_rate": 0.00017234112448887015, "loss": 0.7409, "step": 4677 }, { "epoch": 0.27, "learning_rate": 0.00017232843899793468, "loss": 0.8562, "step": 4678 }, { "epoch": 0.27, "learning_rate": 0.00017231575106572467, "loss": 0.764, "step": 4679 }, { "epoch": 0.27, "learning_rate": 0.00017230306069266843, "loss": 0.7791, "step": 4680 }, { "epoch": 0.27, "learning_rate": 0.00017229036787919428, "loss": 0.6866, "step": 4681 }, { "epoch": 0.27, "learning_rate": 0.00017227767262573062, "loss": 0.7372, "step": 4682 }, { "epoch": 0.27, "learning_rate": 0.00017226497493270598, "loss": 0.7287, "step": 4683 }, { "epoch": 0.27, "learning_rate": 0.00017225227480054887, "loss": 0.6996, "step": 4684 }, { "epoch": 0.27, "learning_rate": 0.000172239572229688, "loss": 0.7239, "step": 4685 }, { "epoch": 0.27, "learning_rate": 0.0001722268672205521, "loss": 0.7556, "step": 4686 }, { "epoch": 0.27, "learning_rate": 0.00017221415977357007, "loss": 0.6631, "step": 4687 }, { "epoch": 0.27, "learning_rate": 0.0001722014498891707, "loss": 0.6801, "step": 4688 }, { "epoch": 0.27, "learning_rate": 0.0001721887375677831, "loss": 0.6351, "step": 4689 }, { "epoch": 0.27, "learning_rate": 0.00017217602280983623, "loss": 0.7634, "step": 4690 }, { "epoch": 0.27, "learning_rate": 0.00017216330561575931, "loss": 0.8655, "step": 4691 }, { "epoch": 0.27, "learning_rate": 0.0001721505859859816, "loss": 0.6782, "step": 4692 }, { "epoch": 0.27, "learning_rate": 0.00017213786392093237, "loss": 0.7525, "step": 4693 }, { "epoch": 0.27, "learning_rate": 0.00017212513942104106, "loss": 0.6233, "step": 4694 }, { "epoch": 0.27, "learning_rate": 0.0001721124124867371, "loss": 0.7718, "step": 4695 }, { "epoch": 0.27, "learning_rate": 0.00017209968311845012, "loss": 0.6473, "step": 4696 }, { "epoch": 0.27, "learning_rate": 0.00017208695131660978, "loss": 0.6955, "step": 4697 }, { "epoch": 0.27, "learning_rate": 0.00017207421708164572, "loss": 0.674, "step": 4698 }, { "epoch": 0.27, "learning_rate": 0.00017206148041398785, "loss": 0.6776, "step": 4699 }, { "epoch": 0.27, "learning_rate": 0.00017204874131406602, "loss": 0.6067, "step": 4700 }, { "epoch": 0.27, "learning_rate": 0.0001720359997823102, "loss": 0.6629, "step": 4701 }, { "epoch": 0.27, "learning_rate": 0.0001720232558191505, "loss": 0.7855, "step": 4702 }, { "epoch": 0.27, "learning_rate": 0.00017201050942501698, "loss": 0.7662, "step": 4703 }, { "epoch": 0.27, "learning_rate": 0.00017199776060033997, "loss": 0.7337, "step": 4704 }, { "epoch": 0.27, "learning_rate": 0.00017198500934554966, "loss": 0.7268, "step": 4705 }, { "epoch": 0.27, "learning_rate": 0.00017197225566107654, "loss": 0.7982, "step": 4706 }, { "epoch": 0.27, "learning_rate": 0.00017195949954735103, "loss": 0.7375, "step": 4707 }, { "epoch": 0.27, "learning_rate": 0.00017194674100480368, "loss": 0.7585, "step": 4708 }, { "epoch": 0.27, "learning_rate": 0.0001719339800338651, "loss": 0.7043, "step": 4709 }, { "epoch": 0.27, "learning_rate": 0.0001719212166349661, "loss": 0.6773, "step": 4710 }, { "epoch": 0.27, "learning_rate": 0.00017190845080853739, "loss": 0.8061, "step": 4711 }, { "epoch": 0.27, "learning_rate": 0.0001718956825550099, "loss": 0.7625, "step": 4712 }, { "epoch": 0.27, "learning_rate": 0.00017188291187481455, "loss": 0.7103, "step": 4713 }, { "epoch": 0.27, "learning_rate": 0.0001718701387683824, "loss": 0.6856, "step": 4714 }, { "epoch": 0.27, "learning_rate": 0.0001718573632361446, "loss": 0.7818, "step": 4715 }, { "epoch": 0.27, "learning_rate": 0.0001718445852785323, "loss": 0.6967, "step": 4716 }, { "epoch": 0.27, "learning_rate": 0.00017183180489597688, "loss": 0.6915, "step": 4717 }, { "epoch": 0.27, "learning_rate": 0.00017181902208890964, "loss": 0.6622, "step": 4718 }, { "epoch": 0.27, "learning_rate": 0.00017180623685776202, "loss": 0.7808, "step": 4719 }, { "epoch": 0.27, "learning_rate": 0.00017179344920296562, "loss": 0.6774, "step": 4720 }, { "epoch": 0.27, "learning_rate": 0.000171780659124952, "loss": 0.7356, "step": 4721 }, { "epoch": 0.27, "learning_rate": 0.0001717678666241529, "loss": 0.6842, "step": 4722 }, { "epoch": 0.27, "learning_rate": 0.0001717550717010001, "loss": 0.7298, "step": 4723 }, { "epoch": 0.27, "learning_rate": 0.0001717422743559254, "loss": 0.8034, "step": 4724 }, { "epoch": 0.27, "learning_rate": 0.0001717294745893608, "loss": 0.7317, "step": 4725 }, { "epoch": 0.27, "learning_rate": 0.00017171667240173835, "loss": 0.6761, "step": 4726 }, { "epoch": 0.27, "learning_rate": 0.00017170386779349014, "loss": 0.7615, "step": 4727 }, { "epoch": 0.27, "learning_rate": 0.0001716910607650483, "loss": 0.7427, "step": 4728 }, { "epoch": 0.27, "learning_rate": 0.00017167825131684513, "loss": 0.6169, "step": 4729 }, { "epoch": 0.27, "learning_rate": 0.000171665439449313, "loss": 0.7661, "step": 4730 }, { "epoch": 0.27, "learning_rate": 0.00017165262516288435, "loss": 0.6782, "step": 4731 }, { "epoch": 0.27, "learning_rate": 0.0001716398084579917, "loss": 0.7448, "step": 4732 }, { "epoch": 0.27, "learning_rate": 0.00017162698933506762, "loss": 0.6671, "step": 4733 }, { "epoch": 0.27, "learning_rate": 0.00017161416779454483, "loss": 0.7275, "step": 4734 }, { "epoch": 0.27, "learning_rate": 0.000171601343836856, "loss": 0.7475, "step": 4735 }, { "epoch": 0.27, "learning_rate": 0.0001715885174624341, "loss": 0.6792, "step": 4736 }, { "epoch": 0.27, "learning_rate": 0.00017157568867171192, "loss": 0.7249, "step": 4737 }, { "epoch": 0.27, "learning_rate": 0.00017156285746512258, "loss": 0.8242, "step": 4738 }, { "epoch": 0.27, "learning_rate": 0.0001715500238430991, "loss": 0.7987, "step": 4739 }, { "epoch": 0.27, "learning_rate": 0.00017153718780607468, "loss": 0.6525, "step": 4740 }, { "epoch": 0.27, "learning_rate": 0.00017152434935448256, "loss": 0.7142, "step": 4741 }, { "epoch": 0.27, "learning_rate": 0.00017151150848875604, "loss": 0.753, "step": 4742 }, { "epoch": 0.27, "learning_rate": 0.00017149866520932864, "loss": 0.7255, "step": 4743 }, { "epoch": 0.27, "learning_rate": 0.0001714858195166337, "loss": 0.7731, "step": 4744 }, { "epoch": 0.27, "learning_rate": 0.0001714729714111049, "loss": 0.6619, "step": 4745 }, { "epoch": 0.27, "learning_rate": 0.00017146012089317586, "loss": 0.7474, "step": 4746 }, { "epoch": 0.27, "learning_rate": 0.00017144726796328034, "loss": 0.7808, "step": 4747 }, { "epoch": 0.27, "learning_rate": 0.00017143441262185217, "loss": 0.6787, "step": 4748 }, { "epoch": 0.27, "learning_rate": 0.00017142155486932518, "loss": 0.7549, "step": 4749 }, { "epoch": 0.27, "learning_rate": 0.00017140869470613342, "loss": 0.7536, "step": 4750 }, { "epoch": 0.27, "learning_rate": 0.00017139583213271098, "loss": 0.726, "step": 4751 }, { "epoch": 0.27, "learning_rate": 0.0001713829671494919, "loss": 0.7581, "step": 4752 }, { "epoch": 0.27, "learning_rate": 0.0001713700997569105, "loss": 0.741, "step": 4753 }, { "epoch": 0.27, "learning_rate": 0.00017135722995540107, "loss": 0.8198, "step": 4754 }, { "epoch": 0.27, "learning_rate": 0.00017134435774539798, "loss": 0.7862, "step": 4755 }, { "epoch": 0.27, "learning_rate": 0.00017133148312733568, "loss": 0.7535, "step": 4756 }, { "epoch": 0.27, "learning_rate": 0.0001713186061016488, "loss": 0.7618, "step": 4757 }, { "epoch": 0.27, "learning_rate": 0.00017130572666877188, "loss": 0.7203, "step": 4758 }, { "epoch": 0.27, "learning_rate": 0.00017129284482913972, "loss": 0.7261, "step": 4759 }, { "epoch": 0.27, "learning_rate": 0.00017127996058318706, "loss": 0.6987, "step": 4760 }, { "epoch": 0.27, "learning_rate": 0.00017126707393134877, "loss": 0.7145, "step": 4761 }, { "epoch": 0.27, "learning_rate": 0.00017125418487405985, "loss": 0.6331, "step": 4762 }, { "epoch": 0.27, "learning_rate": 0.00017124129341175532, "loss": 0.7902, "step": 4763 }, { "epoch": 0.27, "learning_rate": 0.0001712283995448703, "loss": 0.6061, "step": 4764 }, { "epoch": 0.27, "learning_rate": 0.00017121550327383998, "loss": 0.7159, "step": 4765 }, { "epoch": 0.27, "learning_rate": 0.00017120260459909967, "loss": 0.6774, "step": 4766 }, { "epoch": 0.27, "learning_rate": 0.0001711897035210847, "loss": 0.8269, "step": 4767 }, { "epoch": 0.27, "learning_rate": 0.00017117680004023056, "loss": 0.7131, "step": 4768 }, { "epoch": 0.27, "learning_rate": 0.00017116389415697272, "loss": 0.6759, "step": 4769 }, { "epoch": 0.27, "learning_rate": 0.00017115098587174683, "loss": 0.7598, "step": 4770 }, { "epoch": 0.27, "learning_rate": 0.00017113807518498857, "loss": 0.7053, "step": 4771 }, { "epoch": 0.27, "learning_rate": 0.0001711251620971337, "loss": 0.7511, "step": 4772 }, { "epoch": 0.27, "learning_rate": 0.00017111224660861809, "loss": 0.7165, "step": 4773 }, { "epoch": 0.27, "learning_rate": 0.00017109932871987765, "loss": 0.6785, "step": 4774 }, { "epoch": 0.27, "learning_rate": 0.00017108640843134843, "loss": 0.6222, "step": 4775 }, { "epoch": 0.27, "learning_rate": 0.00017107348574346644, "loss": 0.7362, "step": 4776 }, { "epoch": 0.27, "learning_rate": 0.00017106056065666793, "loss": 0.6862, "step": 4777 }, { "epoch": 0.27, "learning_rate": 0.00017104763317138914, "loss": 0.6615, "step": 4778 }, { "epoch": 0.27, "learning_rate": 0.0001710347032880664, "loss": 0.7551, "step": 4779 }, { "epoch": 0.27, "learning_rate": 0.00017102177100713613, "loss": 0.7639, "step": 4780 }, { "epoch": 0.27, "learning_rate": 0.00017100883632903483, "loss": 0.7127, "step": 4781 }, { "epoch": 0.27, "learning_rate": 0.00017099589925419908, "loss": 0.74, "step": 4782 }, { "epoch": 0.27, "learning_rate": 0.00017098295978306552, "loss": 0.7894, "step": 4783 }, { "epoch": 0.27, "learning_rate": 0.00017097001791607091, "loss": 0.7244, "step": 4784 }, { "epoch": 0.27, "learning_rate": 0.0001709570736536521, "loss": 0.7922, "step": 4785 }, { "epoch": 0.27, "learning_rate": 0.00017094412699624595, "loss": 0.6586, "step": 4786 }, { "epoch": 0.27, "learning_rate": 0.00017093117794428944, "loss": 0.7086, "step": 4787 }, { "epoch": 0.27, "learning_rate": 0.00017091822649821964, "loss": 0.6826, "step": 4788 }, { "epoch": 0.27, "learning_rate": 0.00017090527265847374, "loss": 0.695, "step": 4789 }, { "epoch": 0.27, "learning_rate": 0.0001708923164254889, "loss": 0.7092, "step": 4790 }, { "epoch": 0.27, "learning_rate": 0.0001708793577997025, "loss": 0.6453, "step": 4791 }, { "epoch": 0.27, "learning_rate": 0.00017086639678155186, "loss": 0.8255, "step": 4792 }, { "epoch": 0.27, "learning_rate": 0.00017085343337147445, "loss": 0.7234, "step": 4793 }, { "epoch": 0.27, "learning_rate": 0.00017084046756990788, "loss": 0.7422, "step": 4794 }, { "epoch": 0.27, "learning_rate": 0.00017082749937728973, "loss": 0.7135, "step": 4795 }, { "epoch": 0.27, "learning_rate": 0.00017081452879405773, "loss": 0.7406, "step": 4796 }, { "epoch": 0.27, "learning_rate": 0.00017080155582064967, "loss": 0.7029, "step": 4797 }, { "epoch": 0.27, "learning_rate": 0.00017078858045750338, "loss": 0.7357, "step": 4798 }, { "epoch": 0.27, "learning_rate": 0.0001707756027050569, "loss": 0.8465, "step": 4799 }, { "epoch": 0.27, "learning_rate": 0.0001707626225637482, "loss": 0.7942, "step": 4800 }, { "epoch": 0.27, "learning_rate": 0.00017074964003401538, "loss": 0.7295, "step": 4801 }, { "epoch": 0.27, "learning_rate": 0.0001707366551162967, "loss": 0.7152, "step": 4802 }, { "epoch": 0.27, "learning_rate": 0.00017072366781103036, "loss": 0.7412, "step": 4803 }, { "epoch": 0.27, "learning_rate": 0.00017071067811865476, "loss": 0.669, "step": 4804 }, { "epoch": 0.27, "learning_rate": 0.00017069768603960835, "loss": 0.7359, "step": 4805 }, { "epoch": 0.27, "learning_rate": 0.0001706846915743296, "loss": 0.6374, "step": 4806 }, { "epoch": 0.27, "learning_rate": 0.00017067169472325715, "loss": 0.8022, "step": 4807 }, { "epoch": 0.27, "learning_rate": 0.00017065869548682966, "loss": 0.7456, "step": 4808 }, { "epoch": 0.27, "learning_rate": 0.00017064569386548585, "loss": 0.8425, "step": 4809 }, { "epoch": 0.27, "learning_rate": 0.00017063268985966463, "loss": 0.6756, "step": 4810 }, { "epoch": 0.27, "learning_rate": 0.00017061968346980487, "loss": 0.8271, "step": 4811 }, { "epoch": 0.27, "learning_rate": 0.00017060667469634563, "loss": 0.663, "step": 4812 }, { "epoch": 0.27, "learning_rate": 0.0001705936635397259, "loss": 0.7155, "step": 4813 }, { "epoch": 0.27, "learning_rate": 0.0001705806500003849, "loss": 0.6187, "step": 4814 }, { "epoch": 0.27, "learning_rate": 0.00017056763407876185, "loss": 0.8053, "step": 4815 }, { "epoch": 0.27, "learning_rate": 0.0001705546157752961, "loss": 0.7906, "step": 4816 }, { "epoch": 0.27, "learning_rate": 0.00017054159509042704, "loss": 0.7975, "step": 4817 }, { "epoch": 0.27, "learning_rate": 0.00017052857202459412, "loss": 0.6856, "step": 4818 }, { "epoch": 0.27, "learning_rate": 0.00017051554657823692, "loss": 0.7291, "step": 4819 }, { "epoch": 0.27, "learning_rate": 0.00017050251875179508, "loss": 0.7481, "step": 4820 }, { "epoch": 0.27, "learning_rate": 0.00017048948854570838, "loss": 0.7815, "step": 4821 }, { "epoch": 0.27, "learning_rate": 0.00017047645596041653, "loss": 0.6856, "step": 4822 }, { "epoch": 0.27, "learning_rate": 0.00017046342099635948, "loss": 0.8193, "step": 4823 }, { "epoch": 0.27, "learning_rate": 0.00017045038365397715, "loss": 0.655, "step": 4824 }, { "epoch": 0.27, "learning_rate": 0.00017043734393370965, "loss": 0.769, "step": 4825 }, { "epoch": 0.27, "learning_rate": 0.00017042430183599705, "loss": 0.7602, "step": 4826 }, { "epoch": 0.27, "learning_rate": 0.00017041125736127954, "loss": 0.758, "step": 4827 }, { "epoch": 0.27, "learning_rate": 0.00017039821050999747, "loss": 0.6518, "step": 4828 }, { "epoch": 0.27, "learning_rate": 0.00017038516128259115, "loss": 0.6668, "step": 4829 }, { "epoch": 0.27, "learning_rate": 0.00017037210967950106, "loss": 0.6987, "step": 4830 }, { "epoch": 0.27, "learning_rate": 0.0001703590557011677, "loss": 0.747, "step": 4831 }, { "epoch": 0.27, "learning_rate": 0.00017034599934803168, "loss": 0.7934, "step": 4832 }, { "epoch": 0.27, "learning_rate": 0.0001703329406205337, "loss": 0.7503, "step": 4833 }, { "epoch": 0.27, "learning_rate": 0.00017031987951911454, "loss": 0.7067, "step": 4834 }, { "epoch": 0.27, "learning_rate": 0.00017030681604421498, "loss": 0.7445, "step": 4835 }, { "epoch": 0.27, "learning_rate": 0.00017029375019627602, "loss": 0.7423, "step": 4836 }, { "epoch": 0.27, "learning_rate": 0.00017028068197573866, "loss": 0.7353, "step": 4837 }, { "epoch": 0.27, "learning_rate": 0.00017026761138304392, "loss": 0.5904, "step": 4838 }, { "epoch": 0.27, "learning_rate": 0.00017025453841863306, "loss": 0.6998, "step": 4839 }, { "epoch": 0.27, "learning_rate": 0.00017024146308294724, "loss": 0.8449, "step": 4840 }, { "epoch": 0.27, "learning_rate": 0.00017022838537642784, "loss": 0.6942, "step": 4841 }, { "epoch": 0.27, "learning_rate": 0.00017021530529951625, "loss": 0.7029, "step": 4842 }, { "epoch": 0.27, "learning_rate": 0.00017020222285265397, "loss": 0.8037, "step": 4843 }, { "epoch": 0.27, "learning_rate": 0.00017018913803628255, "loss": 0.7186, "step": 4844 }, { "epoch": 0.27, "learning_rate": 0.00017017605085084363, "loss": 0.7116, "step": 4845 }, { "epoch": 0.27, "learning_rate": 0.00017016296129677896, "loss": 0.8157, "step": 4846 }, { "epoch": 0.27, "learning_rate": 0.00017014986937453035, "loss": 0.6692, "step": 4847 }, { "epoch": 0.28, "learning_rate": 0.00017013677508453964, "loss": 0.7288, "step": 4848 }, { "epoch": 0.28, "learning_rate": 0.00017012367842724887, "loss": 0.7509, "step": 4849 }, { "epoch": 0.28, "learning_rate": 0.00017011057940310002, "loss": 0.7132, "step": 4850 }, { "epoch": 0.28, "learning_rate": 0.00017009747801253528, "loss": 0.7339, "step": 4851 }, { "epoch": 0.28, "learning_rate": 0.00017008437425599678, "loss": 0.8034, "step": 4852 }, { "epoch": 0.28, "learning_rate": 0.00017007126813392684, "loss": 0.6949, "step": 4853 }, { "epoch": 0.28, "learning_rate": 0.00017005815964676787, "loss": 0.7891, "step": 4854 }, { "epoch": 0.28, "learning_rate": 0.00017004504879496225, "loss": 0.675, "step": 4855 }, { "epoch": 0.28, "learning_rate": 0.00017003193557895254, "loss": 0.6985, "step": 4856 }, { "epoch": 0.28, "learning_rate": 0.00017001881999918137, "loss": 0.841, "step": 4857 }, { "epoch": 0.28, "learning_rate": 0.00017000570205609136, "loss": 0.6692, "step": 4858 }, { "epoch": 0.28, "learning_rate": 0.0001699925817501253, "loss": 0.7081, "step": 4859 }, { "epoch": 0.28, "learning_rate": 0.0001699794590817261, "loss": 0.7215, "step": 4860 }, { "epoch": 0.28, "learning_rate": 0.00016996633405133655, "loss": 0.7442, "step": 4861 }, { "epoch": 0.28, "learning_rate": 0.00016995320665939977, "loss": 0.8374, "step": 4862 }, { "epoch": 0.28, "learning_rate": 0.00016994007690635882, "loss": 0.5372, "step": 4863 }, { "epoch": 0.28, "learning_rate": 0.0001699269447926568, "loss": 0.7138, "step": 4864 }, { "epoch": 0.28, "learning_rate": 0.00016991381031873705, "loss": 0.764, "step": 4865 }, { "epoch": 0.28, "learning_rate": 0.00016990067348504284, "loss": 0.6639, "step": 4866 }, { "epoch": 0.28, "learning_rate": 0.00016988753429201755, "loss": 0.7057, "step": 4867 }, { "epoch": 0.28, "learning_rate": 0.00016987439274010474, "loss": 0.8076, "step": 4868 }, { "epoch": 0.28, "learning_rate": 0.0001698612488297479, "loss": 0.7924, "step": 4869 }, { "epoch": 0.28, "learning_rate": 0.00016984810256139067, "loss": 0.6837, "step": 4870 }, { "epoch": 0.28, "learning_rate": 0.00016983495393547684, "loss": 0.7601, "step": 4871 }, { "epoch": 0.28, "learning_rate": 0.00016982180295245012, "loss": 0.7732, "step": 4872 }, { "epoch": 0.28, "learning_rate": 0.00016980864961275446, "loss": 0.7625, "step": 4873 }, { "epoch": 0.28, "learning_rate": 0.0001697954939168338, "loss": 0.8143, "step": 4874 }, { "epoch": 0.28, "learning_rate": 0.00016978233586513217, "loss": 0.6294, "step": 4875 }, { "epoch": 0.28, "learning_rate": 0.00016976917545809367, "loss": 0.799, "step": 4876 }, { "epoch": 0.28, "learning_rate": 0.00016975601269616255, "loss": 0.756, "step": 4877 }, { "epoch": 0.28, "learning_rate": 0.00016974284757978304, "loss": 0.7558, "step": 4878 }, { "epoch": 0.28, "learning_rate": 0.00016972968010939954, "loss": 0.7846, "step": 4879 }, { "epoch": 0.28, "learning_rate": 0.00016971651028545648, "loss": 0.6473, "step": 4880 }, { "epoch": 0.28, "learning_rate": 0.00016970333810839835, "loss": 0.6117, "step": 4881 }, { "epoch": 0.28, "learning_rate": 0.00016969016357866972, "loss": 0.7388, "step": 4882 }, { "epoch": 0.28, "learning_rate": 0.0001696769866967153, "loss": 0.793, "step": 4883 }, { "epoch": 0.28, "learning_rate": 0.00016966380746297985, "loss": 0.6887, "step": 4884 }, { "epoch": 0.28, "learning_rate": 0.00016965062587790823, "loss": 0.7528, "step": 4885 }, { "epoch": 0.28, "learning_rate": 0.0001696374419419453, "loss": 0.7728, "step": 4886 }, { "epoch": 0.28, "learning_rate": 0.00016962425565553608, "loss": 0.691, "step": 4887 }, { "epoch": 0.28, "learning_rate": 0.00016961106701912564, "loss": 0.6791, "step": 4888 }, { "epoch": 0.28, "learning_rate": 0.0001695978760331591, "loss": 0.7195, "step": 4889 }, { "epoch": 0.28, "learning_rate": 0.0001695846826980817, "loss": 0.763, "step": 4890 }, { "epoch": 0.28, "learning_rate": 0.00016957148701433883, "loss": 0.6809, "step": 4891 }, { "epoch": 0.28, "learning_rate": 0.00016955828898237578, "loss": 0.7014, "step": 4892 }, { "epoch": 0.28, "learning_rate": 0.00016954508860263806, "loss": 0.6667, "step": 4893 }, { "epoch": 0.28, "learning_rate": 0.00016953188587557122, "loss": 0.7223, "step": 4894 }, { "epoch": 0.28, "learning_rate": 0.00016951868080162087, "loss": 0.6859, "step": 4895 }, { "epoch": 0.28, "learning_rate": 0.00016950547338123273, "loss": 0.6979, "step": 4896 }, { "epoch": 0.28, "learning_rate": 0.00016949226361485256, "loss": 0.7296, "step": 4897 }, { "epoch": 0.28, "learning_rate": 0.00016947905150292624, "loss": 0.7503, "step": 4898 }, { "epoch": 0.28, "learning_rate": 0.00016946583704589973, "loss": 0.8014, "step": 4899 }, { "epoch": 0.28, "learning_rate": 0.00016945262024421905, "loss": 0.6824, "step": 4900 }, { "epoch": 0.28, "learning_rate": 0.00016943940109833026, "loss": 0.7082, "step": 4901 }, { "epoch": 0.28, "learning_rate": 0.0001694261796086796, "loss": 0.6004, "step": 4902 }, { "epoch": 0.28, "learning_rate": 0.0001694129557757133, "loss": 0.7227, "step": 4903 }, { "epoch": 0.28, "learning_rate": 0.0001693997295998777, "loss": 0.8207, "step": 4904 }, { "epoch": 0.28, "learning_rate": 0.0001693865010816192, "loss": 0.6603, "step": 4905 }, { "epoch": 0.28, "learning_rate": 0.00016937327022138434, "loss": 0.7412, "step": 4906 }, { "epoch": 0.28, "learning_rate": 0.00016936003701961965, "loss": 0.626, "step": 4907 }, { "epoch": 0.28, "learning_rate": 0.0001693468014767718, "loss": 0.7613, "step": 4908 }, { "epoch": 0.28, "learning_rate": 0.00016933356359328757, "loss": 0.7515, "step": 4909 }, { "epoch": 0.28, "learning_rate": 0.0001693203233696137, "loss": 0.7739, "step": 4910 }, { "epoch": 0.28, "learning_rate": 0.00016930708080619713, "loss": 0.6912, "step": 4911 }, { "epoch": 0.28, "learning_rate": 0.0001692938359034848, "loss": 0.7551, "step": 4912 }, { "epoch": 0.28, "learning_rate": 0.0001692805886619238, "loss": 0.6706, "step": 4913 }, { "epoch": 0.28, "learning_rate": 0.00016926733908196123, "loss": 0.7418, "step": 4914 }, { "epoch": 0.28, "learning_rate": 0.00016925408716404432, "loss": 0.7005, "step": 4915 }, { "epoch": 0.28, "learning_rate": 0.00016924083290862033, "loss": 0.6977, "step": 4916 }, { "epoch": 0.28, "learning_rate": 0.00016922757631613667, "loss": 0.6778, "step": 4917 }, { "epoch": 0.28, "learning_rate": 0.0001692143173870407, "loss": 0.7387, "step": 4918 }, { "epoch": 0.28, "learning_rate": 0.00016920105612178, "loss": 0.7351, "step": 4919 }, { "epoch": 0.28, "learning_rate": 0.0001691877925208022, "loss": 0.6821, "step": 4920 }, { "epoch": 0.28, "learning_rate": 0.00016917452658455495, "loss": 0.6954, "step": 4921 }, { "epoch": 0.28, "learning_rate": 0.000169161258313486, "loss": 0.6649, "step": 4922 }, { "epoch": 0.28, "learning_rate": 0.0001691479877080432, "loss": 0.7424, "step": 4923 }, { "epoch": 0.28, "learning_rate": 0.00016913471476867446, "loss": 0.6808, "step": 4924 }, { "epoch": 0.28, "learning_rate": 0.00016912143949582783, "loss": 0.8326, "step": 4925 }, { "epoch": 0.28, "learning_rate": 0.00016910816188995127, "loss": 0.672, "step": 4926 }, { "epoch": 0.28, "learning_rate": 0.00016909488195149304, "loss": 0.9328, "step": 4927 }, { "epoch": 0.28, "learning_rate": 0.0001690815996809013, "loss": 0.7523, "step": 4928 }, { "epoch": 0.28, "learning_rate": 0.00016906831507862443, "loss": 0.7887, "step": 4929 }, { "epoch": 0.28, "learning_rate": 0.00016905502814511082, "loss": 0.7077, "step": 4930 }, { "epoch": 0.28, "learning_rate": 0.00016904173888080884, "loss": 0.6966, "step": 4931 }, { "epoch": 0.28, "learning_rate": 0.0001690284472861672, "loss": 0.6754, "step": 4932 }, { "epoch": 0.28, "learning_rate": 0.00016901515336163433, "loss": 0.6576, "step": 4933 }, { "epoch": 0.28, "learning_rate": 0.00016900185710765907, "loss": 0.7648, "step": 4934 }, { "epoch": 0.28, "learning_rate": 0.0001689885585246902, "loss": 0.7794, "step": 4935 }, { "epoch": 0.28, "learning_rate": 0.0001689752576131765, "loss": 0.7555, "step": 4936 }, { "epoch": 0.28, "learning_rate": 0.000168961954373567, "loss": 0.7279, "step": 4937 }, { "epoch": 0.28, "learning_rate": 0.00016894864880631066, "loss": 0.6613, "step": 4938 }, { "epoch": 0.28, "learning_rate": 0.0001689353409118566, "loss": 0.7266, "step": 4939 }, { "epoch": 0.28, "learning_rate": 0.000168922030690654, "loss": 0.7276, "step": 4940 }, { "epoch": 0.28, "learning_rate": 0.0001689087181431521, "loss": 0.7042, "step": 4941 }, { "epoch": 0.28, "learning_rate": 0.00016889540326980025, "loss": 0.6855, "step": 4942 }, { "epoch": 0.28, "learning_rate": 0.00016888208607104789, "loss": 0.796, "step": 4943 }, { "epoch": 0.28, "learning_rate": 0.00016886876654734442, "loss": 0.7435, "step": 4944 }, { "epoch": 0.28, "learning_rate": 0.00016885544469913945, "loss": 0.729, "step": 4945 }, { "epoch": 0.28, "learning_rate": 0.00016884212052688268, "loss": 0.8036, "step": 4946 }, { "epoch": 0.28, "learning_rate": 0.0001688287940310238, "loss": 0.75, "step": 4947 }, { "epoch": 0.28, "learning_rate": 0.0001688154652120126, "loss": 0.6831, "step": 4948 }, { "epoch": 0.28, "learning_rate": 0.00016880213407029899, "loss": 0.7744, "step": 4949 }, { "epoch": 0.28, "learning_rate": 0.00016878880060633292, "loss": 0.6871, "step": 4950 }, { "epoch": 0.28, "learning_rate": 0.00016877546482056439, "loss": 0.7044, "step": 4951 }, { "epoch": 0.28, "learning_rate": 0.0001687621267134436, "loss": 0.7757, "step": 4952 }, { "epoch": 0.28, "learning_rate": 0.00016874878628542065, "loss": 0.6881, "step": 4953 }, { "epoch": 0.28, "learning_rate": 0.00016873544353694588, "loss": 0.7059, "step": 4954 }, { "epoch": 0.28, "learning_rate": 0.00016872209846846963, "loss": 0.6872, "step": 4955 }, { "epoch": 0.28, "learning_rate": 0.0001687087510804423, "loss": 0.692, "step": 4956 }, { "epoch": 0.28, "learning_rate": 0.00016869540137331445, "loss": 0.72, "step": 4957 }, { "epoch": 0.28, "learning_rate": 0.00016868204934753663, "loss": 0.6755, "step": 4958 }, { "epoch": 0.28, "learning_rate": 0.00016866869500355955, "loss": 0.8144, "step": 4959 }, { "epoch": 0.28, "learning_rate": 0.00016865533834183388, "loss": 0.7342, "step": 4960 }, { "epoch": 0.28, "learning_rate": 0.00016864197936281055, "loss": 0.8684, "step": 4961 }, { "epoch": 0.28, "learning_rate": 0.00016862861806694036, "loss": 0.7257, "step": 4962 }, { "epoch": 0.28, "learning_rate": 0.0001686152544546743, "loss": 0.7097, "step": 4963 }, { "epoch": 0.28, "learning_rate": 0.00016860188852646347, "loss": 0.7273, "step": 4964 }, { "epoch": 0.28, "learning_rate": 0.00016858852028275897, "loss": 0.6921, "step": 4965 }, { "epoch": 0.28, "learning_rate": 0.00016857514972401207, "loss": 0.7066, "step": 4966 }, { "epoch": 0.28, "learning_rate": 0.000168561776850674, "loss": 0.7454, "step": 4967 }, { "epoch": 0.28, "learning_rate": 0.00016854840166319614, "loss": 0.7067, "step": 4968 }, { "epoch": 0.28, "learning_rate": 0.00016853502416203, "loss": 0.7275, "step": 4969 }, { "epoch": 0.28, "learning_rate": 0.000168521644347627, "loss": 0.7358, "step": 4970 }, { "epoch": 0.28, "learning_rate": 0.0001685082622204388, "loss": 0.6986, "step": 4971 }, { "epoch": 0.28, "learning_rate": 0.00016849487778091712, "loss": 0.7582, "step": 4972 }, { "epoch": 0.28, "learning_rate": 0.00016848149102951366, "loss": 0.8193, "step": 4973 }, { "epoch": 0.28, "learning_rate": 0.00016846810196668028, "loss": 0.6814, "step": 4974 }, { "epoch": 0.28, "learning_rate": 0.00016845471059286887, "loss": 0.7396, "step": 4975 }, { "epoch": 0.28, "learning_rate": 0.00016844131690853148, "loss": 0.8485, "step": 4976 }, { "epoch": 0.28, "learning_rate": 0.00016842792091412015, "loss": 0.728, "step": 4977 }, { "epoch": 0.28, "learning_rate": 0.00016841452261008702, "loss": 0.7318, "step": 4978 }, { "epoch": 0.28, "learning_rate": 0.00016840112199688432, "loss": 0.8037, "step": 4979 }, { "epoch": 0.28, "learning_rate": 0.00016838771907496435, "loss": 0.6854, "step": 4980 }, { "epoch": 0.28, "learning_rate": 0.00016837431384477956, "loss": 0.8529, "step": 4981 }, { "epoch": 0.28, "learning_rate": 0.00016836090630678233, "loss": 0.6711, "step": 4982 }, { "epoch": 0.28, "learning_rate": 0.0001683474964614252, "loss": 0.6839, "step": 4983 }, { "epoch": 0.28, "learning_rate": 0.00016833408430916085, "loss": 0.6785, "step": 4984 }, { "epoch": 0.28, "learning_rate": 0.00016832066985044195, "loss": 0.734, "step": 4985 }, { "epoch": 0.28, "learning_rate": 0.00016830725308572127, "loss": 0.7946, "step": 4986 }, { "epoch": 0.28, "learning_rate": 0.0001682938340154516, "loss": 0.7114, "step": 4987 }, { "epoch": 0.28, "learning_rate": 0.00016828041264008598, "loss": 0.9628, "step": 4988 }, { "epoch": 0.28, "learning_rate": 0.00016826698896007733, "loss": 0.7562, "step": 4989 }, { "epoch": 0.28, "learning_rate": 0.0001682535629758788, "loss": 0.6995, "step": 4990 }, { "epoch": 0.28, "learning_rate": 0.00016824013468794347, "loss": 0.702, "step": 4991 }, { "epoch": 0.28, "learning_rate": 0.0001682267040967247, "loss": 0.6549, "step": 4992 }, { "epoch": 0.28, "learning_rate": 0.00016821327120267567, "loss": 0.858, "step": 4993 }, { "epoch": 0.28, "learning_rate": 0.00016819983600624986, "loss": 0.7946, "step": 4994 }, { "epoch": 0.28, "learning_rate": 0.00016818639850790072, "loss": 0.7114, "step": 4995 }, { "epoch": 0.28, "learning_rate": 0.0001681729587080818, "loss": 0.612, "step": 4996 }, { "epoch": 0.28, "learning_rate": 0.00016815951660724675, "loss": 0.7436, "step": 4997 }, { "epoch": 0.28, "learning_rate": 0.00016814607220584925, "loss": 0.7978, "step": 4998 }, { "epoch": 0.28, "learning_rate": 0.00016813262550434313, "loss": 0.7487, "step": 4999 }, { "epoch": 0.28, "learning_rate": 0.00016811917650318215, "loss": 0.8018, "step": 5000 }, { "epoch": 0.28, "learning_rate": 0.00016810572520282037, "loss": 0.7605, "step": 5001 }, { "epoch": 0.28, "learning_rate": 0.0001680922716037117, "loss": 0.8229, "step": 5002 }, { "epoch": 0.28, "learning_rate": 0.00016807881570631033, "loss": 0.6941, "step": 5003 }, { "epoch": 0.28, "learning_rate": 0.00016806535751107037, "loss": 0.6719, "step": 5004 }, { "epoch": 0.28, "learning_rate": 0.0001680518970184461, "loss": 0.7445, "step": 5005 }, { "epoch": 0.28, "learning_rate": 0.00016803843422889182, "loss": 0.7344, "step": 5006 }, { "epoch": 0.28, "learning_rate": 0.00016802496914286195, "loss": 0.7448, "step": 5007 }, { "epoch": 0.28, "learning_rate": 0.00016801150176081096, "loss": 0.707, "step": 5008 }, { "epoch": 0.28, "learning_rate": 0.0001679980320831934, "loss": 0.6986, "step": 5009 }, { "epoch": 0.28, "learning_rate": 0.00016798456011046396, "loss": 0.7464, "step": 5010 }, { "epoch": 0.28, "learning_rate": 0.00016797108584307732, "loss": 0.7209, "step": 5011 }, { "epoch": 0.28, "learning_rate": 0.00016795760928148824, "loss": 0.7164, "step": 5012 }, { "epoch": 0.28, "learning_rate": 0.00016794413042615168, "loss": 0.6849, "step": 5013 }, { "epoch": 0.28, "learning_rate": 0.00016793064927752248, "loss": 0.7715, "step": 5014 }, { "epoch": 0.28, "learning_rate": 0.00016791716583605573, "loss": 0.6553, "step": 5015 }, { "epoch": 0.28, "learning_rate": 0.0001679036801022065, "loss": 0.6692, "step": 5016 }, { "epoch": 0.28, "learning_rate": 0.00016789019207642998, "loss": 0.7093, "step": 5017 }, { "epoch": 0.28, "learning_rate": 0.00016787670175918144, "loss": 0.7179, "step": 5018 }, { "epoch": 0.28, "learning_rate": 0.0001678632091509162, "loss": 0.7236, "step": 5019 }, { "epoch": 0.28, "learning_rate": 0.00016784971425208965, "loss": 0.7069, "step": 5020 }, { "epoch": 0.28, "learning_rate": 0.00016783621706315736, "loss": 0.6806, "step": 5021 }, { "epoch": 0.28, "learning_rate": 0.0001678227175845748, "loss": 0.6987, "step": 5022 }, { "epoch": 0.28, "learning_rate": 0.00016780921581679764, "loss": 0.7755, "step": 5023 }, { "epoch": 0.28, "learning_rate": 0.00016779571176028164, "loss": 0.5603, "step": 5024 }, { "epoch": 0.29, "learning_rate": 0.00016778220541548256, "loss": 0.7227, "step": 5025 }, { "epoch": 0.29, "learning_rate": 0.00016776869678285626, "loss": 0.7673, "step": 5026 }, { "epoch": 0.29, "learning_rate": 0.00016775518586285872, "loss": 0.7312, "step": 5027 }, { "epoch": 0.29, "learning_rate": 0.00016774167265594598, "loss": 0.7078, "step": 5028 }, { "epoch": 0.29, "learning_rate": 0.00016772815716257412, "loss": 0.6631, "step": 5029 }, { "epoch": 0.29, "learning_rate": 0.00016771463938319934, "loss": 0.7402, "step": 5030 }, { "epoch": 0.29, "learning_rate": 0.00016770111931827786, "loss": 0.7177, "step": 5031 }, { "epoch": 0.29, "learning_rate": 0.00016768759696826608, "loss": 0.7682, "step": 5032 }, { "epoch": 0.29, "learning_rate": 0.00016767407233362038, "loss": 0.7369, "step": 5033 }, { "epoch": 0.29, "learning_rate": 0.00016766054541479726, "loss": 0.6393, "step": 5034 }, { "epoch": 0.29, "learning_rate": 0.00016764701621225326, "loss": 0.7506, "step": 5035 }, { "epoch": 0.29, "learning_rate": 0.00016763348472644508, "loss": 0.6953, "step": 5036 }, { "epoch": 0.29, "learning_rate": 0.00016761995095782942, "loss": 0.841, "step": 5037 }, { "epoch": 0.29, "learning_rate": 0.00016760641490686307, "loss": 0.7668, "step": 5038 }, { "epoch": 0.29, "learning_rate": 0.0001675928765740029, "loss": 0.6338, "step": 5039 }, { "epoch": 0.29, "learning_rate": 0.00016757933595970589, "loss": 0.7268, "step": 5040 }, { "epoch": 0.29, "learning_rate": 0.00016756579306442904, "loss": 0.7552, "step": 5041 }, { "epoch": 0.29, "learning_rate": 0.00016755224788862947, "loss": 0.6827, "step": 5042 }, { "epoch": 0.29, "learning_rate": 0.00016753870043276442, "loss": 0.7115, "step": 5043 }, { "epoch": 0.29, "learning_rate": 0.00016752515069729106, "loss": 0.6559, "step": 5044 }, { "epoch": 0.29, "learning_rate": 0.0001675115986826668, "loss": 0.8273, "step": 5045 }, { "epoch": 0.29, "learning_rate": 0.000167498044389349, "loss": 0.7451, "step": 5046 }, { "epoch": 0.29, "learning_rate": 0.0001674844878177952, "loss": 0.748, "step": 5047 }, { "epoch": 0.29, "learning_rate": 0.00016747092896846294, "loss": 0.7389, "step": 5048 }, { "epoch": 0.29, "learning_rate": 0.0001674573678418099, "loss": 0.6558, "step": 5049 }, { "epoch": 0.29, "learning_rate": 0.00016744380443829379, "loss": 0.6781, "step": 5050 }, { "epoch": 0.29, "learning_rate": 0.00016743023875837233, "loss": 0.711, "step": 5051 }, { "epoch": 0.29, "learning_rate": 0.00016741667080250353, "loss": 0.7573, "step": 5052 }, { "epoch": 0.29, "learning_rate": 0.0001674031005711453, "loss": 0.6097, "step": 5053 }, { "epoch": 0.29, "learning_rate": 0.0001673895280647556, "loss": 0.8447, "step": 5054 }, { "epoch": 0.29, "learning_rate": 0.00016737595328379263, "loss": 0.7705, "step": 5055 }, { "epoch": 0.29, "learning_rate": 0.00016736237622871452, "loss": 0.7591, "step": 5056 }, { "epoch": 0.29, "learning_rate": 0.00016734879689997954, "loss": 0.6713, "step": 5057 }, { "epoch": 0.29, "learning_rate": 0.00016733521529804608, "loss": 0.6275, "step": 5058 }, { "epoch": 0.29, "learning_rate": 0.0001673216314233725, "loss": 0.7907, "step": 5059 }, { "epoch": 0.29, "learning_rate": 0.00016730804527641726, "loss": 0.7533, "step": 5060 }, { "epoch": 0.29, "learning_rate": 0.00016729445685763897, "loss": 0.7948, "step": 5061 }, { "epoch": 0.29, "learning_rate": 0.00016728086616749633, "loss": 0.6295, "step": 5062 }, { "epoch": 0.29, "learning_rate": 0.00016726727320644797, "loss": 0.8138, "step": 5063 }, { "epoch": 0.29, "learning_rate": 0.00016725367797495273, "loss": 0.7088, "step": 5064 }, { "epoch": 0.29, "learning_rate": 0.00016724008047346947, "loss": 0.8004, "step": 5065 }, { "epoch": 0.29, "learning_rate": 0.00016722648070245717, "loss": 0.759, "step": 5066 }, { "epoch": 0.29, "learning_rate": 0.00016721287866237482, "loss": 0.7522, "step": 5067 }, { "epoch": 0.29, "learning_rate": 0.00016719927435368155, "loss": 0.6541, "step": 5068 }, { "epoch": 0.29, "learning_rate": 0.00016718566777683655, "loss": 0.8095, "step": 5069 }, { "epoch": 0.29, "learning_rate": 0.00016717205893229903, "loss": 0.6956, "step": 5070 }, { "epoch": 0.29, "learning_rate": 0.00016715844782052837, "loss": 0.8505, "step": 5071 }, { "epoch": 0.29, "learning_rate": 0.00016714483444198397, "loss": 0.8265, "step": 5072 }, { "epoch": 0.29, "learning_rate": 0.00016713121879712532, "loss": 0.8411, "step": 5073 }, { "epoch": 0.29, "learning_rate": 0.00016711760088641196, "loss": 0.7558, "step": 5074 }, { "epoch": 0.29, "learning_rate": 0.00016710398071030356, "loss": 0.6907, "step": 5075 }, { "epoch": 0.29, "learning_rate": 0.00016709035826925983, "loss": 0.6724, "step": 5076 }, { "epoch": 0.29, "learning_rate": 0.00016707673356374054, "loss": 0.7978, "step": 5077 }, { "epoch": 0.29, "learning_rate": 0.00016706310659420559, "loss": 0.6686, "step": 5078 }, { "epoch": 0.29, "learning_rate": 0.00016704947736111492, "loss": 0.7336, "step": 5079 }, { "epoch": 0.29, "learning_rate": 0.00016703584586492853, "loss": 0.6969, "step": 5080 }, { "epoch": 0.29, "learning_rate": 0.00016702221210610654, "loss": 0.7425, "step": 5081 }, { "epoch": 0.29, "learning_rate": 0.0001670085760851091, "loss": 0.7249, "step": 5082 }, { "epoch": 0.29, "learning_rate": 0.0001669949378023965, "loss": 0.7786, "step": 5083 }, { "epoch": 0.29, "learning_rate": 0.00016698129725842905, "loss": 0.6097, "step": 5084 }, { "epoch": 0.29, "learning_rate": 0.00016696765445366717, "loss": 0.708, "step": 5085 }, { "epoch": 0.29, "learning_rate": 0.00016695400938857127, "loss": 0.6926, "step": 5086 }, { "epoch": 0.29, "learning_rate": 0.000166940362063602, "loss": 0.7816, "step": 5087 }, { "epoch": 0.29, "learning_rate": 0.00016692671247921996, "loss": 0.7423, "step": 5088 }, { "epoch": 0.29, "learning_rate": 0.00016691306063588583, "loss": 0.8279, "step": 5089 }, { "epoch": 0.29, "learning_rate": 0.00016689940653406043, "loss": 0.7705, "step": 5090 }, { "epoch": 0.29, "learning_rate": 0.00016688575017420457, "loss": 0.6568, "step": 5091 }, { "epoch": 0.29, "learning_rate": 0.00016687209155677929, "loss": 0.7627, "step": 5092 }, { "epoch": 0.29, "learning_rate": 0.00016685843068224552, "loss": 0.8151, "step": 5093 }, { "epoch": 0.29, "learning_rate": 0.00016684476755106435, "loss": 0.6693, "step": 5094 }, { "epoch": 0.29, "learning_rate": 0.000166831102163697, "loss": 0.7864, "step": 5095 }, { "epoch": 0.29, "learning_rate": 0.00016681743452060465, "loss": 0.7042, "step": 5096 }, { "epoch": 0.29, "learning_rate": 0.00016680376462224866, "loss": 0.6951, "step": 5097 }, { "epoch": 0.29, "learning_rate": 0.00016679009246909043, "loss": 0.7254, "step": 5098 }, { "epoch": 0.29, "learning_rate": 0.0001667764180615914, "loss": 0.6698, "step": 5099 }, { "epoch": 0.29, "learning_rate": 0.00016676274140021315, "loss": 0.7134, "step": 5100 }, { "epoch": 0.29, "learning_rate": 0.00016674906248541726, "loss": 0.7635, "step": 5101 }, { "epoch": 0.29, "learning_rate": 0.00016673538131766548, "loss": 0.7656, "step": 5102 }, { "epoch": 0.29, "learning_rate": 0.00016672169789741953, "loss": 0.6717, "step": 5103 }, { "epoch": 0.29, "learning_rate": 0.00016670801222514134, "loss": 0.7225, "step": 5104 }, { "epoch": 0.29, "learning_rate": 0.00016669432430129274, "loss": 0.761, "step": 5105 }, { "epoch": 0.29, "learning_rate": 0.00016668063412633583, "loss": 0.7263, "step": 5106 }, { "epoch": 0.29, "learning_rate": 0.0001666669417007326, "loss": 0.749, "step": 5107 }, { "epoch": 0.29, "learning_rate": 0.00016665324702494524, "loss": 0.704, "step": 5108 }, { "epoch": 0.29, "learning_rate": 0.00016663955009943603, "loss": 0.719, "step": 5109 }, { "epoch": 0.29, "learning_rate": 0.00016662585092466723, "loss": 0.7779, "step": 5110 }, { "epoch": 0.29, "learning_rate": 0.0001666121495011012, "loss": 0.8089, "step": 5111 }, { "epoch": 0.29, "learning_rate": 0.00016659844582920044, "loss": 0.7352, "step": 5112 }, { "epoch": 0.29, "learning_rate": 0.00016658473990942746, "loss": 0.7334, "step": 5113 }, { "epoch": 0.29, "learning_rate": 0.00016657103174224492, "loss": 0.6618, "step": 5114 }, { "epoch": 0.29, "learning_rate": 0.00016655732132811544, "loss": 0.7383, "step": 5115 }, { "epoch": 0.29, "learning_rate": 0.00016654360866750182, "loss": 0.7353, "step": 5116 }, { "epoch": 0.29, "learning_rate": 0.0001665298937608669, "loss": 0.7091, "step": 5117 }, { "epoch": 0.29, "learning_rate": 0.00016651617660867358, "loss": 0.7706, "step": 5118 }, { "epoch": 0.29, "learning_rate": 0.0001665024572113848, "loss": 0.801, "step": 5119 }, { "epoch": 0.29, "learning_rate": 0.00016648873556946377, "loss": 0.824, "step": 5120 }, { "epoch": 0.29, "learning_rate": 0.0001664750116833735, "loss": 0.7357, "step": 5121 }, { "epoch": 0.29, "learning_rate": 0.00016646128555357725, "loss": 0.7991, "step": 5122 }, { "epoch": 0.29, "learning_rate": 0.0001664475571805383, "loss": 0.656, "step": 5123 }, { "epoch": 0.29, "learning_rate": 0.00016643382656472005, "loss": 0.7681, "step": 5124 }, { "epoch": 0.29, "learning_rate": 0.00016642009370658592, "loss": 0.709, "step": 5125 }, { "epoch": 0.29, "learning_rate": 0.0001664063586065994, "loss": 0.7385, "step": 5126 }, { "epoch": 0.29, "learning_rate": 0.00016639262126522418, "loss": 0.7148, "step": 5127 }, { "epoch": 0.29, "learning_rate": 0.00016637888168292384, "loss": 0.7332, "step": 5128 }, { "epoch": 0.29, "learning_rate": 0.00016636513986016213, "loss": 0.6917, "step": 5129 }, { "epoch": 0.29, "learning_rate": 0.00016635139579740296, "loss": 0.7045, "step": 5130 }, { "epoch": 0.29, "learning_rate": 0.00016633764949511014, "loss": 0.7555, "step": 5131 }, { "epoch": 0.29, "learning_rate": 0.00016632390095374766, "loss": 0.8083, "step": 5132 }, { "epoch": 0.29, "learning_rate": 0.0001663101501737796, "loss": 0.7645, "step": 5133 }, { "epoch": 0.29, "learning_rate": 0.00016629639715567005, "loss": 0.7645, "step": 5134 }, { "epoch": 0.29, "learning_rate": 0.00016628264189988323, "loss": 0.7588, "step": 5135 }, { "epoch": 0.29, "learning_rate": 0.00016626888440688344, "loss": 0.8264, "step": 5136 }, { "epoch": 0.29, "learning_rate": 0.000166255124677135, "loss": 0.7566, "step": 5137 }, { "epoch": 0.29, "learning_rate": 0.0001662413627111023, "loss": 0.7288, "step": 5138 }, { "epoch": 0.29, "learning_rate": 0.00016622759850924988, "loss": 0.69, "step": 5139 }, { "epoch": 0.29, "learning_rate": 0.0001662138320720424, "loss": 0.8002, "step": 5140 }, { "epoch": 0.29, "learning_rate": 0.00016620006339994437, "loss": 0.7483, "step": 5141 }, { "epoch": 0.29, "learning_rate": 0.0001661862924934206, "loss": 0.7119, "step": 5142 }, { "epoch": 0.29, "learning_rate": 0.0001661725193529359, "loss": 0.7535, "step": 5143 }, { "epoch": 0.29, "learning_rate": 0.0001661587439789551, "loss": 0.8005, "step": 5144 }, { "epoch": 0.29, "learning_rate": 0.0001661449663719432, "loss": 0.6078, "step": 5145 }, { "epoch": 0.29, "learning_rate": 0.00016613118653236518, "loss": 0.6766, "step": 5146 }, { "epoch": 0.29, "learning_rate": 0.0001661174044606862, "loss": 0.6387, "step": 5147 }, { "epoch": 0.29, "learning_rate": 0.00016610362015737144, "loss": 0.7505, "step": 5148 }, { "epoch": 0.29, "learning_rate": 0.00016608983362288612, "loss": 0.7167, "step": 5149 }, { "epoch": 0.29, "learning_rate": 0.0001660760448576956, "loss": 0.6757, "step": 5150 }, { "epoch": 0.29, "learning_rate": 0.00016606225386226527, "loss": 0.7275, "step": 5151 }, { "epoch": 0.29, "learning_rate": 0.0001660484606370606, "loss": 0.7047, "step": 5152 }, { "epoch": 0.29, "learning_rate": 0.0001660346651825472, "loss": 0.6725, "step": 5153 }, { "epoch": 0.29, "learning_rate": 0.00016602086749919063, "loss": 0.7556, "step": 5154 }, { "epoch": 0.29, "learning_rate": 0.00016600706758745668, "loss": 0.7917, "step": 5155 }, { "epoch": 0.29, "learning_rate": 0.00016599326544781105, "loss": 0.796, "step": 5156 }, { "epoch": 0.29, "learning_rate": 0.00016597946108071963, "loss": 0.6759, "step": 5157 }, { "epoch": 0.29, "learning_rate": 0.00016596565448664838, "loss": 0.7397, "step": 5158 }, { "epoch": 0.29, "learning_rate": 0.00016595184566606328, "loss": 0.6669, "step": 5159 }, { "epoch": 0.29, "learning_rate": 0.00016593803461943045, "loss": 0.6539, "step": 5160 }, { "epoch": 0.29, "learning_rate": 0.000165924221347216, "loss": 0.657, "step": 5161 }, { "epoch": 0.29, "learning_rate": 0.00016591040584988618, "loss": 0.6574, "step": 5162 }, { "epoch": 0.29, "learning_rate": 0.00016589658812790734, "loss": 0.705, "step": 5163 }, { "epoch": 0.29, "learning_rate": 0.0001658827681817458, "loss": 0.8144, "step": 5164 }, { "epoch": 0.29, "learning_rate": 0.00016586894601186805, "loss": 0.7092, "step": 5165 }, { "epoch": 0.29, "learning_rate": 0.00016585512161874065, "loss": 0.689, "step": 5166 }, { "epoch": 0.29, "learning_rate": 0.00016584129500283017, "loss": 0.7105, "step": 5167 }, { "epoch": 0.29, "learning_rate": 0.0001658274661646033, "loss": 0.8377, "step": 5168 }, { "epoch": 0.29, "learning_rate": 0.00016581363510452684, "loss": 0.783, "step": 5169 }, { "epoch": 0.29, "learning_rate": 0.00016579980182306753, "loss": 0.6441, "step": 5170 }, { "epoch": 0.29, "learning_rate": 0.00016578596632069238, "loss": 0.6376, "step": 5171 }, { "epoch": 0.29, "learning_rate": 0.00016577212859786834, "loss": 0.6997, "step": 5172 }, { "epoch": 0.29, "learning_rate": 0.00016575828865506245, "loss": 0.7225, "step": 5173 }, { "epoch": 0.29, "learning_rate": 0.00016574444649274187, "loss": 0.7035, "step": 5174 }, { "epoch": 0.29, "learning_rate": 0.0001657306021113738, "loss": 0.7655, "step": 5175 }, { "epoch": 0.29, "learning_rate": 0.00016571675551142553, "loss": 0.6709, "step": 5176 }, { "epoch": 0.29, "learning_rate": 0.00016570290669336442, "loss": 0.7645, "step": 5177 }, { "epoch": 0.29, "learning_rate": 0.00016568905565765788, "loss": 0.77, "step": 5178 }, { "epoch": 0.29, "learning_rate": 0.00016567520240477344, "loss": 0.758, "step": 5179 }, { "epoch": 0.29, "learning_rate": 0.0001656613469351787, "loss": 0.7546, "step": 5180 }, { "epoch": 0.29, "learning_rate": 0.00016564748924934128, "loss": 0.7951, "step": 5181 }, { "epoch": 0.29, "learning_rate": 0.00016563362934772892, "loss": 0.6818, "step": 5182 }, { "epoch": 0.29, "learning_rate": 0.00016561976723080944, "loss": 0.7335, "step": 5183 }, { "epoch": 0.29, "learning_rate": 0.00016560590289905073, "loss": 0.7876, "step": 5184 }, { "epoch": 0.29, "learning_rate": 0.00016559203635292072, "loss": 0.7785, "step": 5185 }, { "epoch": 0.29, "learning_rate": 0.0001655781675928875, "loss": 0.6778, "step": 5186 }, { "epoch": 0.29, "learning_rate": 0.00016556429661941912, "loss": 0.6734, "step": 5187 }, { "epoch": 0.29, "learning_rate": 0.00016555042343298378, "loss": 0.71, "step": 5188 }, { "epoch": 0.29, "learning_rate": 0.00016553654803404974, "loss": 0.7227, "step": 5189 }, { "epoch": 0.29, "learning_rate": 0.00016552267042308533, "loss": 0.8106, "step": 5190 }, { "epoch": 0.29, "learning_rate": 0.00016550879060055895, "loss": 0.7953, "step": 5191 }, { "epoch": 0.29, "learning_rate": 0.0001654949085669391, "loss": 0.7604, "step": 5192 }, { "epoch": 0.29, "learning_rate": 0.00016548102432269431, "loss": 0.7444, "step": 5193 }, { "epoch": 0.29, "learning_rate": 0.00016546713786829318, "loss": 0.6905, "step": 5194 }, { "epoch": 0.29, "learning_rate": 0.0001654532492042045, "loss": 0.7356, "step": 5195 }, { "epoch": 0.29, "learning_rate": 0.000165439358330897, "loss": 0.7329, "step": 5196 }, { "epoch": 0.29, "learning_rate": 0.0001654254652488395, "loss": 0.7343, "step": 5197 }, { "epoch": 0.29, "learning_rate": 0.000165411569958501, "loss": 0.7388, "step": 5198 }, { "epoch": 0.29, "learning_rate": 0.00016539767246035044, "loss": 0.7409, "step": 5199 }, { "epoch": 0.29, "learning_rate": 0.00016538377275485691, "loss": 0.7676, "step": 5200 }, { "epoch": 0.3, "learning_rate": 0.0001653698708424896, "loss": 0.7729, "step": 5201 }, { "epoch": 0.3, "learning_rate": 0.0001653559667237177, "loss": 0.6682, "step": 5202 }, { "epoch": 0.3, "learning_rate": 0.00016534206039901057, "loss": 0.7204, "step": 5203 }, { "epoch": 0.3, "learning_rate": 0.00016532815186883748, "loss": 0.7691, "step": 5204 }, { "epoch": 0.3, "learning_rate": 0.00016531424113366795, "loss": 0.8226, "step": 5205 }, { "epoch": 0.3, "learning_rate": 0.00016530032819397146, "loss": 0.6217, "step": 5206 }, { "epoch": 0.3, "learning_rate": 0.00016528641305021768, "loss": 0.7819, "step": 5207 }, { "epoch": 0.3, "learning_rate": 0.00016527249570287623, "loss": 0.6869, "step": 5208 }, { "epoch": 0.3, "learning_rate": 0.00016525857615241687, "loss": 0.6542, "step": 5209 }, { "epoch": 0.3, "learning_rate": 0.00016524465439930942, "loss": 0.8, "step": 5210 }, { "epoch": 0.3, "learning_rate": 0.00016523073044402375, "loss": 0.6964, "step": 5211 }, { "epoch": 0.3, "learning_rate": 0.00016521680428702986, "loss": 0.76, "step": 5212 }, { "epoch": 0.3, "learning_rate": 0.0001652028759287978, "loss": 0.7103, "step": 5213 }, { "epoch": 0.3, "learning_rate": 0.00016518894536979767, "loss": 0.7379, "step": 5214 }, { "epoch": 0.3, "learning_rate": 0.00016517501261049968, "loss": 0.8847, "step": 5215 }, { "epoch": 0.3, "learning_rate": 0.0001651610776513741, "loss": 0.6475, "step": 5216 }, { "epoch": 0.3, "learning_rate": 0.0001651471404928912, "loss": 0.7318, "step": 5217 }, { "epoch": 0.3, "learning_rate": 0.00016513320113552152, "loss": 0.6967, "step": 5218 }, { "epoch": 0.3, "learning_rate": 0.00016511925957973542, "loss": 0.7643, "step": 5219 }, { "epoch": 0.3, "learning_rate": 0.00016510531582600356, "loss": 0.7327, "step": 5220 }, { "epoch": 0.3, "learning_rate": 0.00016509136987479656, "loss": 0.8005, "step": 5221 }, { "epoch": 0.3, "learning_rate": 0.0001650774217265851, "loss": 0.659, "step": 5222 }, { "epoch": 0.3, "learning_rate": 0.00016506347138184, "loss": 0.7684, "step": 5223 }, { "epoch": 0.3, "learning_rate": 0.0001650495188410321, "loss": 0.8234, "step": 5224 }, { "epoch": 0.3, "learning_rate": 0.00016503556410463234, "loss": 0.7275, "step": 5225 }, { "epoch": 0.3, "learning_rate": 0.0001650216071731117, "loss": 0.7666, "step": 5226 }, { "epoch": 0.3, "learning_rate": 0.0001650076480469413, "loss": 0.8123, "step": 5227 }, { "epoch": 0.3, "learning_rate": 0.00016499368672659234, "loss": 0.7384, "step": 5228 }, { "epoch": 0.3, "learning_rate": 0.000164979723212536, "loss": 0.7399, "step": 5229 }, { "epoch": 0.3, "learning_rate": 0.00016496575750524353, "loss": 0.6799, "step": 5230 }, { "epoch": 0.3, "learning_rate": 0.00016495178960518637, "loss": 0.7781, "step": 5231 }, { "epoch": 0.3, "learning_rate": 0.000164937819512836, "loss": 0.8155, "step": 5232 }, { "epoch": 0.3, "learning_rate": 0.0001649238472286639, "loss": 0.7404, "step": 5233 }, { "epoch": 0.3, "learning_rate": 0.0001649098727531417, "loss": 0.7319, "step": 5234 }, { "epoch": 0.3, "learning_rate": 0.00016489589608674105, "loss": 0.795, "step": 5235 }, { "epoch": 0.3, "learning_rate": 0.0001648819172299337, "loss": 0.6608, "step": 5236 }, { "epoch": 0.3, "learning_rate": 0.0001648679361831915, "loss": 0.655, "step": 5237 }, { "epoch": 0.3, "learning_rate": 0.00016485395294698634, "loss": 0.8035, "step": 5238 }, { "epoch": 0.3, "learning_rate": 0.00016483996752179015, "loss": 0.7255, "step": 5239 }, { "epoch": 0.3, "learning_rate": 0.00016482597990807505, "loss": 0.6769, "step": 5240 }, { "epoch": 0.3, "learning_rate": 0.0001648119901063131, "loss": 0.7719, "step": 5241 }, { "epoch": 0.3, "learning_rate": 0.0001647979981169765, "loss": 0.7319, "step": 5242 }, { "epoch": 0.3, "learning_rate": 0.00016478400394053755, "loss": 0.6943, "step": 5243 }, { "epoch": 0.3, "learning_rate": 0.00016477000757746853, "loss": 0.899, "step": 5244 }, { "epoch": 0.3, "learning_rate": 0.0001647560090282419, "loss": 0.6924, "step": 5245 }, { "epoch": 0.3, "learning_rate": 0.00016474200829333014, "loss": 0.6727, "step": 5246 }, { "epoch": 0.3, "learning_rate": 0.0001647280053732058, "loss": 0.6074, "step": 5247 }, { "epoch": 0.3, "learning_rate": 0.00016471400026834152, "loss": 0.6937, "step": 5248 }, { "epoch": 0.3, "learning_rate": 0.00016469999297921004, "loss": 0.7191, "step": 5249 }, { "epoch": 0.3, "learning_rate": 0.0001646859835062841, "loss": 0.7481, "step": 5250 }, { "epoch": 0.3, "learning_rate": 0.00016467197185003656, "loss": 0.7465, "step": 5251 }, { "epoch": 0.3, "learning_rate": 0.00016465795801094037, "loss": 0.6751, "step": 5252 }, { "epoch": 0.3, "learning_rate": 0.00016464394198946854, "loss": 0.7224, "step": 5253 }, { "epoch": 0.3, "learning_rate": 0.00016462992378609407, "loss": 0.6201, "step": 5254 }, { "epoch": 0.3, "learning_rate": 0.00016461590340129028, "loss": 0.8479, "step": 5255 }, { "epoch": 0.3, "learning_rate": 0.0001646018808355302, "loss": 0.7238, "step": 5256 }, { "epoch": 0.3, "learning_rate": 0.00016458785608928724, "loss": 0.7595, "step": 5257 }, { "epoch": 0.3, "learning_rate": 0.00016457382916303475, "loss": 0.7593, "step": 5258 }, { "epoch": 0.3, "learning_rate": 0.0001645598000572462, "loss": 0.7619, "step": 5259 }, { "epoch": 0.3, "learning_rate": 0.00016454576877239507, "loss": 0.6358, "step": 5260 }, { "epoch": 0.3, "learning_rate": 0.00016453173530895496, "loss": 0.73, "step": 5261 }, { "epoch": 0.3, "learning_rate": 0.00016451769966739955, "loss": 0.7604, "step": 5262 }, { "epoch": 0.3, "learning_rate": 0.00016450366184820255, "loss": 0.7351, "step": 5263 }, { "epoch": 0.3, "learning_rate": 0.00016448962185183782, "loss": 0.8761, "step": 5264 }, { "epoch": 0.3, "learning_rate": 0.00016447557967877923, "loss": 0.7236, "step": 5265 }, { "epoch": 0.3, "learning_rate": 0.0001644615353295007, "loss": 0.6429, "step": 5266 }, { "epoch": 0.3, "learning_rate": 0.00016444748880447633, "loss": 0.8768, "step": 5267 }, { "epoch": 0.3, "learning_rate": 0.00016443344010418017, "loss": 0.6722, "step": 5268 }, { "epoch": 0.3, "learning_rate": 0.00016441938922908645, "loss": 0.773, "step": 5269 }, { "epoch": 0.3, "learning_rate": 0.00016440533617966934, "loss": 0.7627, "step": 5270 }, { "epoch": 0.3, "learning_rate": 0.00016439128095640325, "loss": 0.7487, "step": 5271 }, { "epoch": 0.3, "learning_rate": 0.00016437722355976258, "loss": 0.7222, "step": 5272 }, { "epoch": 0.3, "learning_rate": 0.0001643631639902218, "loss": 0.7984, "step": 5273 }, { "epoch": 0.3, "learning_rate": 0.0001643491022482554, "loss": 0.7678, "step": 5274 }, { "epoch": 0.3, "learning_rate": 0.000164335038334338, "loss": 0.7104, "step": 5275 }, { "epoch": 0.3, "learning_rate": 0.0001643209722489444, "loss": 0.7262, "step": 5276 }, { "epoch": 0.3, "learning_rate": 0.00016430690399254927, "loss": 0.5837, "step": 5277 }, { "epoch": 0.3, "learning_rate": 0.00016429283356562748, "loss": 0.7103, "step": 5278 }, { "epoch": 0.3, "learning_rate": 0.00016427876096865394, "loss": 0.6759, "step": 5279 }, { "epoch": 0.3, "learning_rate": 0.00016426468620210364, "loss": 0.76, "step": 5280 }, { "epoch": 0.3, "learning_rate": 0.00016425060926645167, "loss": 0.6957, "step": 5281 }, { "epoch": 0.3, "learning_rate": 0.00016423653016217312, "loss": 0.7771, "step": 5282 }, { "epoch": 0.3, "learning_rate": 0.00016422244888974317, "loss": 0.6995, "step": 5283 }, { "epoch": 0.3, "learning_rate": 0.00016420836544963718, "loss": 0.7847, "step": 5284 }, { "epoch": 0.3, "learning_rate": 0.00016419427984233043, "loss": 0.6816, "step": 5285 }, { "epoch": 0.3, "learning_rate": 0.00016418019206829843, "loss": 0.764, "step": 5286 }, { "epoch": 0.3, "learning_rate": 0.0001641661021280166, "loss": 0.6883, "step": 5287 }, { "epoch": 0.3, "learning_rate": 0.0001641520100219605, "loss": 0.7104, "step": 5288 }, { "epoch": 0.3, "learning_rate": 0.0001641379157506059, "loss": 0.707, "step": 5289 }, { "epoch": 0.3, "learning_rate": 0.00016412381931442838, "loss": 0.6681, "step": 5290 }, { "epoch": 0.3, "learning_rate": 0.00016410972071390377, "loss": 0.6197, "step": 5291 }, { "epoch": 0.3, "learning_rate": 0.00016409561994950802, "loss": 0.7643, "step": 5292 }, { "epoch": 0.3, "learning_rate": 0.00016408151702171697, "loss": 0.8238, "step": 5293 }, { "epoch": 0.3, "learning_rate": 0.00016406741193100667, "loss": 0.7207, "step": 5294 }, { "epoch": 0.3, "learning_rate": 0.00016405330467785315, "loss": 0.6843, "step": 5295 }, { "epoch": 0.3, "learning_rate": 0.00016403919526273265, "loss": 0.6439, "step": 5296 }, { "epoch": 0.3, "learning_rate": 0.00016402508368612134, "loss": 0.775, "step": 5297 }, { "epoch": 0.3, "learning_rate": 0.00016401096994849557, "loss": 0.6483, "step": 5298 }, { "epoch": 0.3, "learning_rate": 0.00016399685405033167, "loss": 0.7164, "step": 5299 }, { "epoch": 0.3, "learning_rate": 0.0001639827359921061, "loss": 0.6719, "step": 5300 }, { "epoch": 0.3, "learning_rate": 0.00016396861577429542, "loss": 0.7119, "step": 5301 }, { "epoch": 0.3, "learning_rate": 0.00016395449339737613, "loss": 0.7637, "step": 5302 }, { "epoch": 0.3, "learning_rate": 0.00016394036886182501, "loss": 0.7784, "step": 5303 }, { "epoch": 0.3, "learning_rate": 0.00016392624216811879, "loss": 0.5984, "step": 5304 }, { "epoch": 0.3, "learning_rate": 0.0001639121133167342, "loss": 0.8122, "step": 5305 }, { "epoch": 0.3, "learning_rate": 0.00016389798230814816, "loss": 0.8946, "step": 5306 }, { "epoch": 0.3, "learning_rate": 0.00016388384914283764, "loss": 0.8223, "step": 5307 }, { "epoch": 0.3, "learning_rate": 0.0001638697138212797, "loss": 0.6196, "step": 5308 }, { "epoch": 0.3, "learning_rate": 0.00016385557634395137, "loss": 0.655, "step": 5309 }, { "epoch": 0.3, "learning_rate": 0.0001638414367113299, "loss": 0.7058, "step": 5310 }, { "epoch": 0.3, "learning_rate": 0.0001638272949238925, "loss": 0.6114, "step": 5311 }, { "epoch": 0.3, "learning_rate": 0.00016381315098211652, "loss": 0.8105, "step": 5312 }, { "epoch": 0.3, "learning_rate": 0.00016379900488647933, "loss": 0.6374, "step": 5313 }, { "epoch": 0.3, "learning_rate": 0.0001637848566374584, "loss": 0.7099, "step": 5314 }, { "epoch": 0.3, "learning_rate": 0.00016377070623553128, "loss": 0.6921, "step": 5315 }, { "epoch": 0.3, "learning_rate": 0.00016375655368117557, "loss": 0.7761, "step": 5316 }, { "epoch": 0.3, "learning_rate": 0.000163742398974869, "loss": 0.6986, "step": 5317 }, { "epoch": 0.3, "learning_rate": 0.00016372824211708926, "loss": 0.7656, "step": 5318 }, { "epoch": 0.3, "learning_rate": 0.0001637140831083142, "loss": 0.7473, "step": 5319 }, { "epoch": 0.3, "learning_rate": 0.00016369992194902177, "loss": 0.7213, "step": 5320 }, { "epoch": 0.3, "learning_rate": 0.00016368575863968992, "loss": 0.67, "step": 5321 }, { "epoch": 0.3, "learning_rate": 0.0001636715931807967, "loss": 0.856, "step": 5322 }, { "epoch": 0.3, "learning_rate": 0.00016365742557282017, "loss": 0.7375, "step": 5323 }, { "epoch": 0.3, "learning_rate": 0.00016364325581623863, "loss": 0.7352, "step": 5324 }, { "epoch": 0.3, "learning_rate": 0.0001636290839115303, "loss": 0.7543, "step": 5325 }, { "epoch": 0.3, "learning_rate": 0.0001636149098591735, "loss": 0.6918, "step": 5326 }, { "epoch": 0.3, "learning_rate": 0.00016360073365964664, "loss": 0.7796, "step": 5327 }, { "epoch": 0.3, "learning_rate": 0.0001635865553134282, "loss": 0.6869, "step": 5328 }, { "epoch": 0.3, "learning_rate": 0.00016357237482099684, "loss": 0.6442, "step": 5329 }, { "epoch": 0.3, "learning_rate": 0.00016355819218283105, "loss": 0.6981, "step": 5330 }, { "epoch": 0.3, "learning_rate": 0.00016354400739940957, "loss": 0.6845, "step": 5331 }, { "epoch": 0.3, "learning_rate": 0.0001635298204712112, "loss": 0.7996, "step": 5332 }, { "epoch": 0.3, "learning_rate": 0.0001635156313987148, "loss": 0.7187, "step": 5333 }, { "epoch": 0.3, "learning_rate": 0.00016350144018239924, "loss": 0.7705, "step": 5334 }, { "epoch": 0.3, "learning_rate": 0.00016348724682274353, "loss": 0.7683, "step": 5335 }, { "epoch": 0.3, "learning_rate": 0.00016347305132022677, "loss": 0.8033, "step": 5336 }, { "epoch": 0.3, "learning_rate": 0.00016345885367532805, "loss": 0.802, "step": 5337 }, { "epoch": 0.3, "learning_rate": 0.00016344465388852658, "loss": 0.7161, "step": 5338 }, { "epoch": 0.3, "learning_rate": 0.00016343045196030168, "loss": 0.7448, "step": 5339 }, { "epoch": 0.3, "learning_rate": 0.00016341624789113266, "loss": 0.7886, "step": 5340 }, { "epoch": 0.3, "learning_rate": 0.00016340204168149893, "loss": 0.6079, "step": 5341 }, { "epoch": 0.3, "learning_rate": 0.00016338783333188004, "loss": 0.7493, "step": 5342 }, { "epoch": 0.3, "learning_rate": 0.00016337362284275556, "loss": 0.789, "step": 5343 }, { "epoch": 0.3, "learning_rate": 0.00016335941021460506, "loss": 0.7109, "step": 5344 }, { "epoch": 0.3, "learning_rate": 0.0001633451954479083, "loss": 0.794, "step": 5345 }, { "epoch": 0.3, "learning_rate": 0.00016333097854314508, "loss": 0.7894, "step": 5346 }, { "epoch": 0.3, "learning_rate": 0.00016331675950079526, "loss": 0.7352, "step": 5347 }, { "epoch": 0.3, "learning_rate": 0.00016330253832133872, "loss": 0.6633, "step": 5348 }, { "epoch": 0.3, "learning_rate": 0.00016328831500525554, "loss": 0.7344, "step": 5349 }, { "epoch": 0.3, "learning_rate": 0.00016327408955302573, "loss": 0.7901, "step": 5350 }, { "epoch": 0.3, "learning_rate": 0.00016325986196512944, "loss": 0.7062, "step": 5351 }, { "epoch": 0.3, "learning_rate": 0.0001632456322420469, "loss": 0.6758, "step": 5352 }, { "epoch": 0.3, "learning_rate": 0.00016323140038425842, "loss": 0.7418, "step": 5353 }, { "epoch": 0.3, "learning_rate": 0.00016321716639224434, "loss": 0.6913, "step": 5354 }, { "epoch": 0.3, "learning_rate": 0.0001632029302664851, "loss": 0.7482, "step": 5355 }, { "epoch": 0.3, "learning_rate": 0.00016318869200746122, "loss": 0.7612, "step": 5356 }, { "epoch": 0.3, "learning_rate": 0.00016317445161565325, "loss": 0.6728, "step": 5357 }, { "epoch": 0.3, "learning_rate": 0.00016316020909154185, "loss": 0.7347, "step": 5358 }, { "epoch": 0.3, "learning_rate": 0.00016314596443560778, "loss": 0.8193, "step": 5359 }, { "epoch": 0.3, "learning_rate": 0.00016313171764833177, "loss": 0.7626, "step": 5360 }, { "epoch": 0.3, "learning_rate": 0.00016311746873019474, "loss": 0.8088, "step": 5361 }, { "epoch": 0.3, "learning_rate": 0.00016310321768167762, "loss": 0.6785, "step": 5362 }, { "epoch": 0.3, "learning_rate": 0.0001630889645032614, "loss": 0.7483, "step": 5363 }, { "epoch": 0.3, "learning_rate": 0.00016307470919542713, "loss": 0.7908, "step": 5364 }, { "epoch": 0.3, "learning_rate": 0.00016306045175865606, "loss": 0.8122, "step": 5365 }, { "epoch": 0.3, "learning_rate": 0.00016304619219342933, "loss": 0.6127, "step": 5366 }, { "epoch": 0.3, "learning_rate": 0.00016303193050022828, "loss": 0.7474, "step": 5367 }, { "epoch": 0.3, "learning_rate": 0.00016301766667953425, "loss": 0.7945, "step": 5368 }, { "epoch": 0.3, "learning_rate": 0.00016300340073182877, "loss": 0.6905, "step": 5369 }, { "epoch": 0.3, "learning_rate": 0.0001629891326575932, "loss": 0.7888, "step": 5370 }, { "epoch": 0.3, "learning_rate": 0.00016297486245730927, "loss": 0.7161, "step": 5371 }, { "epoch": 0.3, "learning_rate": 0.0001629605901314585, "loss": 0.7257, "step": 5372 }, { "epoch": 0.3, "learning_rate": 0.00016294631568052276, "loss": 0.755, "step": 5373 }, { "epoch": 0.3, "learning_rate": 0.00016293203910498376, "loss": 0.7684, "step": 5374 }, { "epoch": 0.3, "learning_rate": 0.0001629177604053234, "loss": 0.685, "step": 5375 }, { "epoch": 0.3, "learning_rate": 0.00016290347958202358, "loss": 0.6763, "step": 5376 }, { "epoch": 0.31, "learning_rate": 0.0001628891966355664, "loss": 0.749, "step": 5377 }, { "epoch": 0.31, "learning_rate": 0.0001628749115664339, "loss": 0.7045, "step": 5378 }, { "epoch": 0.31, "learning_rate": 0.0001628606243751082, "loss": 0.7334, "step": 5379 }, { "epoch": 0.31, "learning_rate": 0.0001628463350620716, "loss": 0.7511, "step": 5380 }, { "epoch": 0.31, "learning_rate": 0.0001628320436278064, "loss": 0.7322, "step": 5381 }, { "epoch": 0.31, "learning_rate": 0.0001628177500727949, "loss": 0.6934, "step": 5382 }, { "epoch": 0.31, "learning_rate": 0.00016280345439751958, "loss": 0.6493, "step": 5383 }, { "epoch": 0.31, "learning_rate": 0.000162789156602463, "loss": 0.7983, "step": 5384 }, { "epoch": 0.31, "learning_rate": 0.0001627748566881077, "loss": 0.6901, "step": 5385 }, { "epoch": 0.31, "learning_rate": 0.00016276055465493637, "loss": 0.7776, "step": 5386 }, { "epoch": 0.31, "learning_rate": 0.0001627462505034317, "loss": 0.7559, "step": 5387 }, { "epoch": 0.31, "learning_rate": 0.00016273194423407653, "loss": 0.7876, "step": 5388 }, { "epoch": 0.31, "learning_rate": 0.0001627176358473537, "loss": 0.7289, "step": 5389 }, { "epoch": 0.31, "learning_rate": 0.0001627033253437462, "loss": 0.6567, "step": 5390 }, { "epoch": 0.31, "learning_rate": 0.000162689012723737, "loss": 0.6195, "step": 5391 }, { "epoch": 0.31, "learning_rate": 0.00016267469798780928, "loss": 0.6311, "step": 5392 }, { "epoch": 0.31, "learning_rate": 0.00016266038113644607, "loss": 0.7506, "step": 5393 }, { "epoch": 0.31, "learning_rate": 0.00016264606217013068, "loss": 0.7154, "step": 5394 }, { "epoch": 0.31, "learning_rate": 0.0001626317410893464, "loss": 0.6834, "step": 5395 }, { "epoch": 0.31, "learning_rate": 0.00016261741789457657, "loss": 0.6837, "step": 5396 }, { "epoch": 0.31, "learning_rate": 0.00016260309258630469, "loss": 0.8326, "step": 5397 }, { "epoch": 0.31, "learning_rate": 0.00016258876516501424, "loss": 0.7416, "step": 5398 }, { "epoch": 0.31, "learning_rate": 0.00016257443563118886, "loss": 0.666, "step": 5399 }, { "epoch": 0.31, "learning_rate": 0.00016256010398531213, "loss": 0.6999, "step": 5400 }, { "epoch": 0.31, "learning_rate": 0.00016254577022786784, "loss": 0.7635, "step": 5401 }, { "epoch": 0.31, "learning_rate": 0.00016253143435933978, "loss": 0.7029, "step": 5402 }, { "epoch": 0.31, "learning_rate": 0.0001625170963802118, "loss": 0.7037, "step": 5403 }, { "epoch": 0.31, "learning_rate": 0.00016250275629096786, "loss": 0.7392, "step": 5404 }, { "epoch": 0.31, "learning_rate": 0.00016248841409209196, "loss": 0.7854, "step": 5405 }, { "epoch": 0.31, "learning_rate": 0.00016247406978406822, "loss": 0.7892, "step": 5406 }, { "epoch": 0.31, "learning_rate": 0.0001624597233673808, "loss": 0.7921, "step": 5407 }, { "epoch": 0.31, "learning_rate": 0.0001624453748425139, "loss": 0.7855, "step": 5408 }, { "epoch": 0.31, "learning_rate": 0.00016243102420995182, "loss": 0.7578, "step": 5409 }, { "epoch": 0.31, "learning_rate": 0.00016241667147017896, "loss": 0.6909, "step": 5410 }, { "epoch": 0.31, "learning_rate": 0.00016240231662367972, "loss": 0.7252, "step": 5411 }, { "epoch": 0.31, "learning_rate": 0.00016238795967093864, "loss": 0.6882, "step": 5412 }, { "epoch": 0.31, "learning_rate": 0.0001623736006124403, "loss": 0.8601, "step": 5413 }, { "epoch": 0.31, "learning_rate": 0.00016235923944866936, "loss": 0.6165, "step": 5414 }, { "epoch": 0.31, "learning_rate": 0.00016234487618011056, "loss": 0.7244, "step": 5415 }, { "epoch": 0.31, "learning_rate": 0.00016233051080724868, "loss": 0.735, "step": 5416 }, { "epoch": 0.31, "learning_rate": 0.00016231614333056857, "loss": 0.7122, "step": 5417 }, { "epoch": 0.31, "learning_rate": 0.00016230177375055522, "loss": 0.7135, "step": 5418 }, { "epoch": 0.31, "learning_rate": 0.00016228740206769362, "loss": 0.593, "step": 5419 }, { "epoch": 0.31, "learning_rate": 0.00016227302828246878, "loss": 0.813, "step": 5420 }, { "epoch": 0.31, "learning_rate": 0.000162258652395366, "loss": 0.7426, "step": 5421 }, { "epoch": 0.31, "learning_rate": 0.0001622442744068704, "loss": 0.7323, "step": 5422 }, { "epoch": 0.31, "learning_rate": 0.00016222989431746728, "loss": 0.7798, "step": 5423 }, { "epoch": 0.31, "learning_rate": 0.00016221551212764203, "loss": 0.7353, "step": 5424 }, { "epoch": 0.31, "learning_rate": 0.0001622011278378801, "loss": 0.7967, "step": 5425 }, { "epoch": 0.31, "learning_rate": 0.00016218674144866696, "loss": 0.7012, "step": 5426 }, { "epoch": 0.31, "learning_rate": 0.00016217235296048818, "loss": 0.7186, "step": 5427 }, { "epoch": 0.31, "learning_rate": 0.00016215796237382946, "loss": 0.7414, "step": 5428 }, { "epoch": 0.31, "learning_rate": 0.00016214356968917648, "loss": 0.7372, "step": 5429 }, { "epoch": 0.31, "learning_rate": 0.00016212917490701507, "loss": 0.7255, "step": 5430 }, { "epoch": 0.31, "learning_rate": 0.00016211477802783103, "loss": 0.6683, "step": 5431 }, { "epoch": 0.31, "learning_rate": 0.00016210037905211037, "loss": 0.8469, "step": 5432 }, { "epoch": 0.31, "learning_rate": 0.00016208597798033904, "loss": 0.7481, "step": 5433 }, { "epoch": 0.31, "learning_rate": 0.00016207157481300312, "loss": 0.6907, "step": 5434 }, { "epoch": 0.31, "learning_rate": 0.00016205716955058875, "loss": 0.7105, "step": 5435 }, { "epoch": 0.31, "learning_rate": 0.0001620427621935822, "loss": 0.621, "step": 5436 }, { "epoch": 0.31, "learning_rate": 0.0001620283527424697, "loss": 0.7479, "step": 5437 }, { "epoch": 0.31, "learning_rate": 0.00016201394119773758, "loss": 0.717, "step": 5438 }, { "epoch": 0.31, "learning_rate": 0.00016199952755987233, "loss": 0.6982, "step": 5439 }, { "epoch": 0.31, "learning_rate": 0.00016198511182936043, "loss": 0.7645, "step": 5440 }, { "epoch": 0.31, "learning_rate": 0.00016197069400668843, "loss": 0.8387, "step": 5441 }, { "epoch": 0.31, "learning_rate": 0.000161956274092343, "loss": 0.7254, "step": 5442 }, { "epoch": 0.31, "learning_rate": 0.00016194185208681083, "loss": 0.7074, "step": 5443 }, { "epoch": 0.31, "learning_rate": 0.00016192742799057868, "loss": 0.742, "step": 5444 }, { "epoch": 0.31, "learning_rate": 0.00016191300180413344, "loss": 0.6753, "step": 5445 }, { "epoch": 0.31, "learning_rate": 0.00016189857352796202, "loss": 0.6772, "step": 5446 }, { "epoch": 0.31, "learning_rate": 0.00016188414316255142, "loss": 0.73, "step": 5447 }, { "epoch": 0.31, "learning_rate": 0.00016186971070838868, "loss": 0.7365, "step": 5448 }, { "epoch": 0.31, "learning_rate": 0.00016185527616596095, "loss": 0.6905, "step": 5449 }, { "epoch": 0.31, "learning_rate": 0.0001618408395357554, "loss": 0.7017, "step": 5450 }, { "epoch": 0.31, "learning_rate": 0.00016182640081825937, "loss": 0.6871, "step": 5451 }, { "epoch": 0.31, "learning_rate": 0.00016181196001396019, "loss": 0.6796, "step": 5452 }, { "epoch": 0.31, "learning_rate": 0.00016179751712334519, "loss": 0.7568, "step": 5453 }, { "epoch": 0.31, "learning_rate": 0.00016178307214690193, "loss": 0.7771, "step": 5454 }, { "epoch": 0.31, "learning_rate": 0.00016176862508511799, "loss": 0.7458, "step": 5455 }, { "epoch": 0.31, "learning_rate": 0.0001617541759384809, "loss": 0.7525, "step": 5456 }, { "epoch": 0.31, "learning_rate": 0.00016173972470747848, "loss": 0.7505, "step": 5457 }, { "epoch": 0.31, "learning_rate": 0.00016172527139259838, "loss": 0.7374, "step": 5458 }, { "epoch": 0.31, "learning_rate": 0.00016171081599432853, "loss": 0.6667, "step": 5459 }, { "epoch": 0.31, "learning_rate": 0.00016169635851315676, "loss": 0.7143, "step": 5460 }, { "epoch": 0.31, "learning_rate": 0.0001616818989495711, "loss": 0.7567, "step": 5461 }, { "epoch": 0.31, "learning_rate": 0.0001616674373040596, "loss": 0.7374, "step": 5462 }, { "epoch": 0.31, "learning_rate": 0.00016165297357711035, "loss": 0.6757, "step": 5463 }, { "epoch": 0.31, "learning_rate": 0.00016163850776921155, "loss": 0.7463, "step": 5464 }, { "epoch": 0.31, "learning_rate": 0.00016162403988085147, "loss": 0.6644, "step": 5465 }, { "epoch": 0.31, "learning_rate": 0.0001616095699125184, "loss": 0.6838, "step": 5466 }, { "epoch": 0.31, "learning_rate": 0.0001615950978647008, "loss": 0.7357, "step": 5467 }, { "epoch": 0.31, "learning_rate": 0.00016158062373788709, "loss": 0.7176, "step": 5468 }, { "epoch": 0.31, "learning_rate": 0.0001615661475325658, "loss": 0.8377, "step": 5469 }, { "epoch": 0.31, "learning_rate": 0.00016155166924922566, "loss": 0.7341, "step": 5470 }, { "epoch": 0.31, "learning_rate": 0.0001615371888883552, "loss": 0.6963, "step": 5471 }, { "epoch": 0.31, "learning_rate": 0.00016152270645044322, "loss": 0.772, "step": 5472 }, { "epoch": 0.31, "learning_rate": 0.00016150822193597856, "loss": 0.7675, "step": 5473 }, { "epoch": 0.31, "learning_rate": 0.00016149373534545008, "loss": 0.6616, "step": 5474 }, { "epoch": 0.31, "learning_rate": 0.00016147924667934675, "loss": 0.6281, "step": 5475 }, { "epoch": 0.31, "learning_rate": 0.00016146475593815764, "loss": 0.7829, "step": 5476 }, { "epoch": 0.31, "learning_rate": 0.0001614502631223718, "loss": 0.7659, "step": 5477 }, { "epoch": 0.31, "learning_rate": 0.00016143576823247845, "loss": 0.5869, "step": 5478 }, { "epoch": 0.31, "learning_rate": 0.0001614212712689668, "loss": 0.7661, "step": 5479 }, { "epoch": 0.31, "learning_rate": 0.00016140677223232613, "loss": 0.6856, "step": 5480 }, { "epoch": 0.31, "learning_rate": 0.00016139227112304586, "loss": 0.8651, "step": 5481 }, { "epoch": 0.31, "learning_rate": 0.00016137776794161545, "loss": 0.7482, "step": 5482 }, { "epoch": 0.31, "learning_rate": 0.00016136326268852437, "loss": 0.6458, "step": 5483 }, { "epoch": 0.31, "learning_rate": 0.0001613487553642623, "loss": 0.6747, "step": 5484 }, { "epoch": 0.31, "learning_rate": 0.0001613342459693188, "loss": 0.7446, "step": 5485 }, { "epoch": 0.31, "learning_rate": 0.00016131973450418365, "loss": 0.6445, "step": 5486 }, { "epoch": 0.31, "learning_rate": 0.00016130522096934665, "loss": 0.7158, "step": 5487 }, { "epoch": 0.31, "learning_rate": 0.00016129070536529766, "loss": 0.7725, "step": 5488 }, { "epoch": 0.31, "learning_rate": 0.0001612761876925266, "loss": 0.6128, "step": 5489 }, { "epoch": 0.31, "learning_rate": 0.00016126166795152355, "loss": 0.6876, "step": 5490 }, { "epoch": 0.31, "learning_rate": 0.0001612471461427785, "loss": 0.7085, "step": 5491 }, { "epoch": 0.31, "learning_rate": 0.00016123262226678167, "loss": 0.7414, "step": 5492 }, { "epoch": 0.31, "learning_rate": 0.00016121809632402325, "loss": 0.7909, "step": 5493 }, { "epoch": 0.31, "learning_rate": 0.0001612035683149935, "loss": 0.7575, "step": 5494 }, { "epoch": 0.31, "learning_rate": 0.00016118903824018283, "loss": 0.7328, "step": 5495 }, { "epoch": 0.31, "learning_rate": 0.00016117450610008164, "loss": 0.7622, "step": 5496 }, { "epoch": 0.31, "learning_rate": 0.00016115997189518043, "loss": 0.7488, "step": 5497 }, { "epoch": 0.31, "learning_rate": 0.00016114543562596978, "loss": 0.6641, "step": 5498 }, { "epoch": 0.31, "learning_rate": 0.0001611308972929403, "loss": 0.7702, "step": 5499 }, { "epoch": 0.31, "learning_rate": 0.00016111635689658273, "loss": 0.6844, "step": 5500 }, { "epoch": 0.31, "learning_rate": 0.00016110181443738784, "loss": 0.831, "step": 5501 }, { "epoch": 0.31, "learning_rate": 0.00016108726991584643, "loss": 0.573, "step": 5502 }, { "epoch": 0.31, "learning_rate": 0.00016107272333244953, "loss": 0.7811, "step": 5503 }, { "epoch": 0.31, "learning_rate": 0.00016105817468768798, "loss": 0.781, "step": 5504 }, { "epoch": 0.31, "learning_rate": 0.0001610436239820529, "loss": 0.679, "step": 5505 }, { "epoch": 0.31, "learning_rate": 0.00016102907121603543, "loss": 0.6977, "step": 5506 }, { "epoch": 0.31, "learning_rate": 0.0001610145163901268, "loss": 0.73, "step": 5507 }, { "epoch": 0.31, "learning_rate": 0.00016099995950481818, "loss": 0.7895, "step": 5508 }, { "epoch": 0.31, "learning_rate": 0.00016098540056060093, "loss": 0.7555, "step": 5509 }, { "epoch": 0.31, "learning_rate": 0.0001609708395579665, "loss": 0.7674, "step": 5510 }, { "epoch": 0.31, "learning_rate": 0.00016095627649740632, "loss": 0.747, "step": 5511 }, { "epoch": 0.31, "learning_rate": 0.00016094171137941195, "loss": 0.7617, "step": 5512 }, { "epoch": 0.31, "learning_rate": 0.000160927144204475, "loss": 0.71, "step": 5513 }, { "epoch": 0.31, "learning_rate": 0.00016091257497308713, "loss": 0.7775, "step": 5514 }, { "epoch": 0.31, "learning_rate": 0.00016089800368574014, "loss": 0.7564, "step": 5515 }, { "epoch": 0.31, "learning_rate": 0.0001608834303429258, "loss": 0.7614, "step": 5516 }, { "epoch": 0.31, "learning_rate": 0.000160868854945136, "loss": 0.7215, "step": 5517 }, { "epoch": 0.31, "learning_rate": 0.00016085427749286271, "loss": 0.751, "step": 5518 }, { "epoch": 0.31, "learning_rate": 0.00016083969798659796, "loss": 0.7842, "step": 5519 }, { "epoch": 0.31, "learning_rate": 0.00016082511642683388, "loss": 0.824, "step": 5520 }, { "epoch": 0.31, "learning_rate": 0.00016081053281406256, "loss": 0.7237, "step": 5521 }, { "epoch": 0.31, "learning_rate": 0.00016079594714877627, "loss": 0.5915, "step": 5522 }, { "epoch": 0.31, "learning_rate": 0.00016078135943146734, "loss": 0.7748, "step": 5523 }, { "epoch": 0.31, "learning_rate": 0.00016076676966262813, "loss": 0.7619, "step": 5524 }, { "epoch": 0.31, "learning_rate": 0.00016075217784275106, "loss": 0.6921, "step": 5525 }, { "epoch": 0.31, "learning_rate": 0.00016073758397232868, "loss": 0.769, "step": 5526 }, { "epoch": 0.31, "learning_rate": 0.00016072298805185352, "loss": 0.754, "step": 5527 }, { "epoch": 0.31, "learning_rate": 0.0001607083900818183, "loss": 0.7091, "step": 5528 }, { "epoch": 0.31, "learning_rate": 0.00016069379006271566, "loss": 0.6905, "step": 5529 }, { "epoch": 0.31, "learning_rate": 0.00016067918799503848, "loss": 0.7005, "step": 5530 }, { "epoch": 0.31, "learning_rate": 0.0001606645838792795, "loss": 0.7161, "step": 5531 }, { "epoch": 0.31, "learning_rate": 0.0001606499777159318, "loss": 0.6786, "step": 5532 }, { "epoch": 0.31, "learning_rate": 0.00016063536950548826, "loss": 0.6512, "step": 5533 }, { "epoch": 0.31, "learning_rate": 0.00016062075924844194, "loss": 0.7702, "step": 5534 }, { "epoch": 0.31, "learning_rate": 0.00016060614694528603, "loss": 0.7039, "step": 5535 }, { "epoch": 0.31, "learning_rate": 0.00016059153259651373, "loss": 0.7331, "step": 5536 }, { "epoch": 0.31, "learning_rate": 0.0001605769162026183, "loss": 0.7256, "step": 5537 }, { "epoch": 0.31, "learning_rate": 0.00016056229776409306, "loss": 0.7006, "step": 5538 }, { "epoch": 0.31, "learning_rate": 0.00016054767728143147, "loss": 0.741, "step": 5539 }, { "epoch": 0.31, "learning_rate": 0.00016053305475512695, "loss": 0.6737, "step": 5540 }, { "epoch": 0.31, "learning_rate": 0.00016051843018567306, "loss": 0.7619, "step": 5541 }, { "epoch": 0.31, "learning_rate": 0.0001605038035735635, "loss": 0.8002, "step": 5542 }, { "epoch": 0.31, "learning_rate": 0.00016048917491929182, "loss": 0.7719, "step": 5543 }, { "epoch": 0.31, "learning_rate": 0.0001604745442233519, "loss": 0.7216, "step": 5544 }, { "epoch": 0.31, "learning_rate": 0.0001604599114862375, "loss": 0.7344, "step": 5545 }, { "epoch": 0.31, "learning_rate": 0.0001604452767084425, "loss": 0.8487, "step": 5546 }, { "epoch": 0.31, "learning_rate": 0.0001604306398904609, "loss": 0.7364, "step": 5547 }, { "epoch": 0.31, "learning_rate": 0.00016041600103278673, "loss": 0.7404, "step": 5548 }, { "epoch": 0.31, "learning_rate": 0.0001604013601359141, "loss": 0.7791, "step": 5549 }, { "epoch": 0.31, "learning_rate": 0.00016038671720033714, "loss": 0.7867, "step": 5550 }, { "epoch": 0.31, "learning_rate": 0.0001603720722265501, "loss": 0.7273, "step": 5551 }, { "epoch": 0.31, "learning_rate": 0.00016035742521504725, "loss": 0.7223, "step": 5552 }, { "epoch": 0.31, "learning_rate": 0.00016034277616632308, "loss": 0.7478, "step": 5553 }, { "epoch": 0.32, "learning_rate": 0.0001603281250808719, "loss": 0.7191, "step": 5554 }, { "epoch": 0.32, "learning_rate": 0.00016031347195918835, "loss": 0.7889, "step": 5555 }, { "epoch": 0.32, "learning_rate": 0.00016029881680176686, "loss": 0.8369, "step": 5556 }, { "epoch": 0.32, "learning_rate": 0.0001602841596091022, "loss": 0.7621, "step": 5557 }, { "epoch": 0.32, "learning_rate": 0.00016026950038168907, "loss": 0.6527, "step": 5558 }, { "epoch": 0.32, "learning_rate": 0.00016025483912002222, "loss": 0.7445, "step": 5559 }, { "epoch": 0.32, "learning_rate": 0.00016024017582459652, "loss": 0.7211, "step": 5560 }, { "epoch": 0.32, "learning_rate": 0.00016022551049590692, "loss": 0.7824, "step": 5561 }, { "epoch": 0.32, "learning_rate": 0.0001602108431344484, "loss": 0.6653, "step": 5562 }, { "epoch": 0.32, "learning_rate": 0.00016019617374071597, "loss": 0.6881, "step": 5563 }, { "epoch": 0.32, "learning_rate": 0.00016018150231520486, "loss": 0.8487, "step": 5564 }, { "epoch": 0.32, "learning_rate": 0.00016016682885841015, "loss": 0.7316, "step": 5565 }, { "epoch": 0.32, "learning_rate": 0.00016015215337082724, "loss": 0.5569, "step": 5566 }, { "epoch": 0.32, "learning_rate": 0.00016013747585295136, "loss": 0.7273, "step": 5567 }, { "epoch": 0.32, "learning_rate": 0.00016012279630527797, "loss": 0.6986, "step": 5568 }, { "epoch": 0.32, "learning_rate": 0.00016010811472830252, "loss": 0.8385, "step": 5569 }, { "epoch": 0.32, "learning_rate": 0.00016009343112252056, "loss": 0.7226, "step": 5570 }, { "epoch": 0.32, "learning_rate": 0.00016007874548842768, "loss": 0.768, "step": 5571 }, { "epoch": 0.32, "learning_rate": 0.0001600640578265196, "loss": 0.7697, "step": 5572 }, { "epoch": 0.32, "learning_rate": 0.00016004936813729205, "loss": 0.691, "step": 5573 }, { "epoch": 0.32, "learning_rate": 0.00016003467642124086, "loss": 0.6835, "step": 5574 }, { "epoch": 0.32, "learning_rate": 0.00016001998267886185, "loss": 0.6906, "step": 5575 }, { "epoch": 0.32, "learning_rate": 0.000160005286910651, "loss": 0.7528, "step": 5576 }, { "epoch": 0.32, "learning_rate": 0.0001599905891171044, "loss": 0.8257, "step": 5577 }, { "epoch": 0.32, "learning_rate": 0.00015997588929871808, "loss": 0.7054, "step": 5578 }, { "epoch": 0.32, "learning_rate": 0.00015996118745598817, "loss": 0.7243, "step": 5579 }, { "epoch": 0.32, "learning_rate": 0.00015994648358941095, "loss": 0.723, "step": 5580 }, { "epoch": 0.32, "learning_rate": 0.0001599317776994827, "loss": 0.6941, "step": 5581 }, { "epoch": 0.32, "learning_rate": 0.00015991706978669976, "loss": 0.6625, "step": 5582 }, { "epoch": 0.32, "learning_rate": 0.0001599023598515586, "loss": 0.7359, "step": 5583 }, { "epoch": 0.32, "learning_rate": 0.00015988764789455568, "loss": 0.6681, "step": 5584 }, { "epoch": 0.32, "learning_rate": 0.00015987293391618755, "loss": 0.7906, "step": 5585 }, { "epoch": 0.32, "learning_rate": 0.00015985821791695095, "loss": 0.7809, "step": 5586 }, { "epoch": 0.32, "learning_rate": 0.00015984349989734247, "loss": 0.7838, "step": 5587 }, { "epoch": 0.32, "learning_rate": 0.00015982877985785892, "loss": 0.7192, "step": 5588 }, { "epoch": 0.32, "learning_rate": 0.00015981405779899715, "loss": 0.7049, "step": 5589 }, { "epoch": 0.32, "learning_rate": 0.00015979933372125405, "loss": 0.7092, "step": 5590 }, { "epoch": 0.32, "learning_rate": 0.0001597846076251266, "loss": 0.6568, "step": 5591 }, { "epoch": 0.32, "learning_rate": 0.0001597698795111119, "loss": 0.7444, "step": 5592 }, { "epoch": 0.32, "learning_rate": 0.00015975514937970696, "loss": 0.7224, "step": 5593 }, { "epoch": 0.32, "learning_rate": 0.00015974041723140902, "loss": 0.6801, "step": 5594 }, { "epoch": 0.32, "learning_rate": 0.00015972568306671535, "loss": 0.7355, "step": 5595 }, { "epoch": 0.32, "learning_rate": 0.00015971094688612322, "loss": 0.7842, "step": 5596 }, { "epoch": 0.32, "learning_rate": 0.00015969620869013006, "loss": 0.694, "step": 5597 }, { "epoch": 0.32, "learning_rate": 0.00015968146847923328, "loss": 0.6985, "step": 5598 }, { "epoch": 0.32, "learning_rate": 0.00015966672625393043, "loss": 0.7087, "step": 5599 }, { "epoch": 0.32, "learning_rate": 0.0001596519820147191, "loss": 0.703, "step": 5600 }, { "epoch": 0.32, "learning_rate": 0.00015963723576209687, "loss": 0.7701, "step": 5601 }, { "epoch": 0.32, "learning_rate": 0.0001596224874965616, "loss": 0.7688, "step": 5602 }, { "epoch": 0.32, "learning_rate": 0.00015960773721861098, "loss": 0.8061, "step": 5603 }, { "epoch": 0.32, "learning_rate": 0.00015959298492874288, "loss": 0.757, "step": 5604 }, { "epoch": 0.32, "learning_rate": 0.0001595782306274553, "loss": 0.8144, "step": 5605 }, { "epoch": 0.32, "learning_rate": 0.00015956347431524614, "loss": 0.7921, "step": 5606 }, { "epoch": 0.32, "learning_rate": 0.00015954871599261353, "loss": 0.6811, "step": 5607 }, { "epoch": 0.32, "learning_rate": 0.0001595339556600556, "loss": 0.7396, "step": 5608 }, { "epoch": 0.32, "learning_rate": 0.0001595191933180705, "loss": 0.715, "step": 5609 }, { "epoch": 0.32, "learning_rate": 0.00015950442896715656, "loss": 0.8093, "step": 5610 }, { "epoch": 0.32, "learning_rate": 0.0001594896626078121, "loss": 0.749, "step": 5611 }, { "epoch": 0.32, "learning_rate": 0.0001594748942405355, "loss": 0.6476, "step": 5612 }, { "epoch": 0.32, "learning_rate": 0.00015946012386582517, "loss": 0.7226, "step": 5613 }, { "epoch": 0.32, "learning_rate": 0.00015944535148417982, "loss": 0.7092, "step": 5614 }, { "epoch": 0.32, "learning_rate": 0.0001594305770960979, "loss": 0.6814, "step": 5615 }, { "epoch": 0.32, "learning_rate": 0.00015941580070207815, "loss": 0.7152, "step": 5616 }, { "epoch": 0.32, "learning_rate": 0.00015940102230261932, "loss": 0.6682, "step": 5617 }, { "epoch": 0.32, "learning_rate": 0.00015938624189822015, "loss": 0.7506, "step": 5618 }, { "epoch": 0.32, "learning_rate": 0.00015937145948937963, "loss": 0.7032, "step": 5619 }, { "epoch": 0.32, "learning_rate": 0.0001593566750765966, "loss": 0.8729, "step": 5620 }, { "epoch": 0.32, "learning_rate": 0.00015934188866037016, "loss": 0.6766, "step": 5621 }, { "epoch": 0.32, "learning_rate": 0.00015932710024119933, "loss": 0.7341, "step": 5622 }, { "epoch": 0.32, "learning_rate": 0.00015931230981958326, "loss": 0.6882, "step": 5623 }, { "epoch": 0.32, "learning_rate": 0.00015929751739602119, "loss": 0.8331, "step": 5624 }, { "epoch": 0.32, "learning_rate": 0.0001592827229710124, "loss": 0.5527, "step": 5625 }, { "epoch": 0.32, "learning_rate": 0.00015926792654505625, "loss": 0.7234, "step": 5626 }, { "epoch": 0.32, "learning_rate": 0.00015925312811865212, "loss": 0.6806, "step": 5627 }, { "epoch": 0.32, "learning_rate": 0.0001592383276922995, "loss": 0.7426, "step": 5628 }, { "epoch": 0.32, "learning_rate": 0.00015922352526649803, "loss": 0.7497, "step": 5629 }, { "epoch": 0.32, "learning_rate": 0.0001592087208417472, "loss": 0.8047, "step": 5630 }, { "epoch": 0.32, "learning_rate": 0.0001591939144185468, "loss": 0.7678, "step": 5631 }, { "epoch": 0.32, "learning_rate": 0.0001591791059973965, "loss": 0.7019, "step": 5632 }, { "epoch": 0.32, "learning_rate": 0.00015916429557879622, "loss": 0.6814, "step": 5633 }, { "epoch": 0.32, "learning_rate": 0.00015914948316324577, "loss": 0.6972, "step": 5634 }, { "epoch": 0.32, "learning_rate": 0.00015913466875124517, "loss": 0.7983, "step": 5635 }, { "epoch": 0.32, "learning_rate": 0.00015911985234329435, "loss": 0.7767, "step": 5636 }, { "epoch": 0.32, "learning_rate": 0.00015910503393989353, "loss": 0.7497, "step": 5637 }, { "epoch": 0.32, "learning_rate": 0.00015909021354154276, "loss": 0.7582, "step": 5638 }, { "epoch": 0.32, "learning_rate": 0.00015907539114874236, "loss": 0.5811, "step": 5639 }, { "epoch": 0.32, "learning_rate": 0.00015906056676199255, "loss": 0.7325, "step": 5640 }, { "epoch": 0.32, "learning_rate": 0.0001590457403817937, "loss": 0.9059, "step": 5641 }, { "epoch": 0.32, "learning_rate": 0.0001590309120086463, "loss": 0.7089, "step": 5642 }, { "epoch": 0.32, "learning_rate": 0.00015901608164305077, "loss": 0.7194, "step": 5643 }, { "epoch": 0.32, "learning_rate": 0.00015900124928550774, "loss": 0.7531, "step": 5644 }, { "epoch": 0.32, "learning_rate": 0.0001589864149365178, "loss": 0.6871, "step": 5645 }, { "epoch": 0.32, "learning_rate": 0.00015897157859658164, "loss": 0.7624, "step": 5646 }, { "epoch": 0.32, "learning_rate": 0.0001589567402662001, "loss": 0.7199, "step": 5647 }, { "epoch": 0.32, "learning_rate": 0.0001589418999458739, "loss": 0.6985, "step": 5648 }, { "epoch": 0.32, "learning_rate": 0.00015892705763610398, "loss": 0.661, "step": 5649 }, { "epoch": 0.32, "learning_rate": 0.00015891221333739136, "loss": 0.6723, "step": 5650 }, { "epoch": 0.32, "learning_rate": 0.00015889736705023704, "loss": 0.7706, "step": 5651 }, { "epoch": 0.32, "learning_rate": 0.00015888251877514212, "loss": 0.8038, "step": 5652 }, { "epoch": 0.32, "learning_rate": 0.00015886766851260775, "loss": 0.7146, "step": 5653 }, { "epoch": 0.32, "learning_rate": 0.00015885281626313517, "loss": 0.796, "step": 5654 }, { "epoch": 0.32, "learning_rate": 0.00015883796202722573, "loss": 0.7348, "step": 5655 }, { "epoch": 0.32, "learning_rate": 0.00015882310580538075, "loss": 0.74, "step": 5656 }, { "epoch": 0.32, "learning_rate": 0.00015880824759810164, "loss": 0.7729, "step": 5657 }, { "epoch": 0.32, "learning_rate": 0.00015879338740589, "loss": 0.6708, "step": 5658 }, { "epoch": 0.32, "learning_rate": 0.00015877852522924732, "loss": 0.7519, "step": 5659 }, { "epoch": 0.32, "learning_rate": 0.00015876366106867528, "loss": 0.6049, "step": 5660 }, { "epoch": 0.32, "learning_rate": 0.00015874879492467554, "loss": 0.7961, "step": 5661 }, { "epoch": 0.32, "learning_rate": 0.00015873392679774989, "loss": 0.7294, "step": 5662 }, { "epoch": 0.32, "learning_rate": 0.0001587190566884002, "loss": 0.7134, "step": 5663 }, { "epoch": 0.32, "learning_rate": 0.00015870418459712834, "loss": 0.6915, "step": 5664 }, { "epoch": 0.32, "learning_rate": 0.0001586893105244363, "loss": 0.7917, "step": 5665 }, { "epoch": 0.32, "learning_rate": 0.0001586744344708261, "loss": 0.7676, "step": 5666 }, { "epoch": 0.32, "learning_rate": 0.00015865955643679986, "loss": 0.6763, "step": 5667 }, { "epoch": 0.32, "learning_rate": 0.00015864467642285975, "loss": 0.6892, "step": 5668 }, { "epoch": 0.32, "learning_rate": 0.000158629794429508, "loss": 0.7042, "step": 5669 }, { "epoch": 0.32, "learning_rate": 0.00015861491045724696, "loss": 0.6815, "step": 5670 }, { "epoch": 0.32, "learning_rate": 0.00015860002450657893, "loss": 0.7336, "step": 5671 }, { "epoch": 0.32, "learning_rate": 0.0001585851365780064, "loss": 0.843, "step": 5672 }, { "epoch": 0.32, "learning_rate": 0.0001585702466720319, "loss": 0.6212, "step": 5673 }, { "epoch": 0.32, "learning_rate": 0.00015855535478915793, "loss": 0.7355, "step": 5674 }, { "epoch": 0.32, "learning_rate": 0.00015854046092988715, "loss": 0.6897, "step": 5675 }, { "epoch": 0.32, "learning_rate": 0.0001585255650947223, "loss": 0.7428, "step": 5676 }, { "epoch": 0.32, "learning_rate": 0.00015851066728416618, "loss": 0.6937, "step": 5677 }, { "epoch": 0.32, "learning_rate": 0.00015849576749872157, "loss": 0.7121, "step": 5678 }, { "epoch": 0.32, "learning_rate": 0.00015848086573889137, "loss": 0.6468, "step": 5679 }, { "epoch": 0.32, "learning_rate": 0.00015846596200517858, "loss": 0.7983, "step": 5680 }, { "epoch": 0.32, "learning_rate": 0.00015845105629808627, "loss": 0.7075, "step": 5681 }, { "epoch": 0.32, "learning_rate": 0.0001584361486181175, "loss": 0.6966, "step": 5682 }, { "epoch": 0.32, "learning_rate": 0.00015842123896577545, "loss": 0.7381, "step": 5683 }, { "epoch": 0.32, "learning_rate": 0.00015840632734156343, "loss": 0.656, "step": 5684 }, { "epoch": 0.32, "learning_rate": 0.0001583914137459846, "loss": 0.6881, "step": 5685 }, { "epoch": 0.32, "learning_rate": 0.00015837649817954246, "loss": 0.6927, "step": 5686 }, { "epoch": 0.32, "learning_rate": 0.00015836158064274042, "loss": 0.6731, "step": 5687 }, { "epoch": 0.32, "learning_rate": 0.00015834666113608193, "loss": 0.7264, "step": 5688 }, { "epoch": 0.32, "learning_rate": 0.00015833173966007066, "loss": 0.7769, "step": 5689 }, { "epoch": 0.32, "learning_rate": 0.00015831681621521017, "loss": 0.7485, "step": 5690 }, { "epoch": 0.32, "learning_rate": 0.0001583018908020042, "loss": 0.5349, "step": 5691 }, { "epoch": 0.32, "learning_rate": 0.0001582869634209565, "loss": 0.8184, "step": 5692 }, { "epoch": 0.32, "learning_rate": 0.00015827203407257093, "loss": 0.7707, "step": 5693 }, { "epoch": 0.32, "learning_rate": 0.0001582571027573514, "loss": 0.7475, "step": 5694 }, { "epoch": 0.32, "learning_rate": 0.00015824216947580183, "loss": 0.6824, "step": 5695 }, { "epoch": 0.32, "learning_rate": 0.00015822723422842633, "loss": 0.7091, "step": 5696 }, { "epoch": 0.32, "learning_rate": 0.00015821229701572896, "loss": 0.7142, "step": 5697 }, { "epoch": 0.32, "learning_rate": 0.00015819735783821387, "loss": 0.7069, "step": 5698 }, { "epoch": 0.32, "learning_rate": 0.00015818241669638536, "loss": 0.7269, "step": 5699 }, { "epoch": 0.32, "learning_rate": 0.0001581674735907477, "loss": 0.7368, "step": 5700 }, { "epoch": 0.32, "learning_rate": 0.00015815252852180523, "loss": 0.7881, "step": 5701 }, { "epoch": 0.32, "learning_rate": 0.0001581375814900624, "loss": 0.6477, "step": 5702 }, { "epoch": 0.32, "learning_rate": 0.0001581226324960238, "loss": 0.7922, "step": 5703 }, { "epoch": 0.32, "learning_rate": 0.00015810768154019385, "loss": 0.6992, "step": 5704 }, { "epoch": 0.32, "learning_rate": 0.00015809272862307724, "loss": 0.7875, "step": 5705 }, { "epoch": 0.32, "learning_rate": 0.00015807777374517873, "loss": 0.7343, "step": 5706 }, { "epoch": 0.32, "learning_rate": 0.00015806281690700305, "loss": 0.7781, "step": 5707 }, { "epoch": 0.32, "learning_rate": 0.00015804785810905498, "loss": 0.7568, "step": 5708 }, { "epoch": 0.32, "learning_rate": 0.00015803289735183952, "loss": 0.6857, "step": 5709 }, { "epoch": 0.32, "learning_rate": 0.00015801793463586153, "loss": 0.7566, "step": 5710 }, { "epoch": 0.32, "learning_rate": 0.00015800296996162612, "loss": 0.7898, "step": 5711 }, { "epoch": 0.32, "learning_rate": 0.00015798800332963832, "loss": 0.7018, "step": 5712 }, { "epoch": 0.32, "learning_rate": 0.00015797303474040332, "loss": 0.8378, "step": 5713 }, { "epoch": 0.32, "learning_rate": 0.0001579580641944264, "loss": 0.6523, "step": 5714 }, { "epoch": 0.32, "learning_rate": 0.00015794309169221278, "loss": 0.8038, "step": 5715 }, { "epoch": 0.32, "learning_rate": 0.0001579281172342679, "loss": 0.7585, "step": 5716 }, { "epoch": 0.32, "learning_rate": 0.0001579131408210971, "loss": 0.7873, "step": 5717 }, { "epoch": 0.32, "learning_rate": 0.00015789816245320593, "loss": 0.7779, "step": 5718 }, { "epoch": 0.32, "learning_rate": 0.00015788318213109993, "loss": 0.7406, "step": 5719 }, { "epoch": 0.32, "learning_rate": 0.00015786819985528472, "loss": 0.6338, "step": 5720 }, { "epoch": 0.32, "learning_rate": 0.00015785321562626604, "loss": 0.6869, "step": 5721 }, { "epoch": 0.32, "learning_rate": 0.00015783822944454956, "loss": 0.752, "step": 5722 }, { "epoch": 0.32, "learning_rate": 0.00015782324131064117, "loss": 0.7602, "step": 5723 }, { "epoch": 0.32, "learning_rate": 0.0001578082512250467, "loss": 0.7643, "step": 5724 }, { "epoch": 0.32, "learning_rate": 0.0001577932591882722, "loss": 0.6645, "step": 5725 }, { "epoch": 0.32, "learning_rate": 0.00015777826520082364, "loss": 0.8582, "step": 5726 }, { "epoch": 0.32, "learning_rate": 0.00015776326926320705, "loss": 0.5951, "step": 5727 }, { "epoch": 0.32, "learning_rate": 0.00015774827137592865, "loss": 0.6437, "step": 5728 }, { "epoch": 0.32, "learning_rate": 0.00015773327153949465, "loss": 0.8014, "step": 5729 }, { "epoch": 0.33, "learning_rate": 0.00015771826975441132, "loss": 0.7188, "step": 5730 }, { "epoch": 0.33, "learning_rate": 0.000157703266021185, "loss": 0.6926, "step": 5731 }, { "epoch": 0.33, "learning_rate": 0.00015768826034032215, "loss": 0.6736, "step": 5732 }, { "epoch": 0.33, "learning_rate": 0.0001576732527123292, "loss": 0.7061, "step": 5733 }, { "epoch": 0.33, "learning_rate": 0.00015765824313771274, "loss": 0.6874, "step": 5734 }, { "epoch": 0.33, "learning_rate": 0.00015764323161697935, "loss": 0.6981, "step": 5735 }, { "epoch": 0.33, "learning_rate": 0.0001576282181506357, "loss": 0.7489, "step": 5736 }, { "epoch": 0.33, "learning_rate": 0.00015761320273918855, "loss": 0.694, "step": 5737 }, { "epoch": 0.33, "learning_rate": 0.0001575981853831447, "loss": 0.7316, "step": 5738 }, { "epoch": 0.33, "learning_rate": 0.00015758316608301107, "loss": 0.7513, "step": 5739 }, { "epoch": 0.33, "learning_rate": 0.00015756814483929458, "loss": 0.7065, "step": 5740 }, { "epoch": 0.33, "learning_rate": 0.0001575531216525022, "loss": 0.7018, "step": 5741 }, { "epoch": 0.33, "learning_rate": 0.00015753809652314102, "loss": 0.7127, "step": 5742 }, { "epoch": 0.33, "learning_rate": 0.0001575230694517182, "loss": 0.7582, "step": 5743 }, { "epoch": 0.33, "learning_rate": 0.00015750804043874092, "loss": 0.6572, "step": 5744 }, { "epoch": 0.33, "learning_rate": 0.00015749300948471646, "loss": 0.8089, "step": 5745 }, { "epoch": 0.33, "learning_rate": 0.0001574779765901521, "loss": 0.7307, "step": 5746 }, { "epoch": 0.33, "learning_rate": 0.00015746294175555535, "loss": 0.7373, "step": 5747 }, { "epoch": 0.33, "learning_rate": 0.0001574479049814336, "loss": 0.6787, "step": 5748 }, { "epoch": 0.33, "learning_rate": 0.00015743286626829437, "loss": 0.7423, "step": 5749 }, { "epoch": 0.33, "learning_rate": 0.00015741782561664528, "loss": 0.7287, "step": 5750 }, { "epoch": 0.33, "learning_rate": 0.00015740278302699406, "loss": 0.7298, "step": 5751 }, { "epoch": 0.33, "learning_rate": 0.00015738773849984829, "loss": 0.7358, "step": 5752 }, { "epoch": 0.33, "learning_rate": 0.00015737269203571587, "loss": 0.7656, "step": 5753 }, { "epoch": 0.33, "learning_rate": 0.0001573576436351046, "loss": 0.7399, "step": 5754 }, { "epoch": 0.33, "learning_rate": 0.00015734259329852247, "loss": 0.7573, "step": 5755 }, { "epoch": 0.33, "learning_rate": 0.00015732754102647744, "loss": 0.7603, "step": 5756 }, { "epoch": 0.33, "learning_rate": 0.00015731248681947753, "loss": 0.6873, "step": 5757 }, { "epoch": 0.33, "learning_rate": 0.00015729743067803087, "loss": 0.7144, "step": 5758 }, { "epoch": 0.33, "learning_rate": 0.00015728237260264567, "loss": 0.6607, "step": 5759 }, { "epoch": 0.33, "learning_rate": 0.00015726731259383016, "loss": 0.7097, "step": 5760 }, { "epoch": 0.33, "learning_rate": 0.00015725225065209265, "loss": 0.6319, "step": 5761 }, { "epoch": 0.33, "learning_rate": 0.00015723718677794157, "loss": 0.745, "step": 5762 }, { "epoch": 0.33, "learning_rate": 0.00015722212097188528, "loss": 0.8085, "step": 5763 }, { "epoch": 0.33, "learning_rate": 0.00015720705323443238, "loss": 0.7733, "step": 5764 }, { "epoch": 0.33, "learning_rate": 0.00015719198356609136, "loss": 0.7031, "step": 5765 }, { "epoch": 0.33, "learning_rate": 0.00015717691196737095, "loss": 0.7227, "step": 5766 }, { "epoch": 0.33, "learning_rate": 0.00015716183843877976, "loss": 0.6793, "step": 5767 }, { "epoch": 0.33, "learning_rate": 0.00015714676298082665, "loss": 0.7052, "step": 5768 }, { "epoch": 0.33, "learning_rate": 0.0001571316855940204, "loss": 0.6978, "step": 5769 }, { "epoch": 0.33, "learning_rate": 0.00015711660627886994, "loss": 0.7107, "step": 5770 }, { "epoch": 0.33, "learning_rate": 0.00015710152503588422, "loss": 0.8034, "step": 5771 }, { "epoch": 0.33, "learning_rate": 0.00015708644186557226, "loss": 0.7384, "step": 5772 }, { "epoch": 0.33, "learning_rate": 0.0001570713567684432, "loss": 0.6336, "step": 5773 }, { "epoch": 0.33, "learning_rate": 0.00015705626974500613, "loss": 0.6711, "step": 5774 }, { "epoch": 0.33, "learning_rate": 0.00015704118079577035, "loss": 0.713, "step": 5775 }, { "epoch": 0.33, "learning_rate": 0.0001570260899212451, "loss": 0.7979, "step": 5776 }, { "epoch": 0.33, "learning_rate": 0.00015701099712193977, "loss": 0.6775, "step": 5777 }, { "epoch": 0.33, "learning_rate": 0.00015699590239836382, "loss": 0.7211, "step": 5778 }, { "epoch": 0.33, "learning_rate": 0.00015698080575102661, "loss": 0.6633, "step": 5779 }, { "epoch": 0.33, "learning_rate": 0.0001569657071804378, "loss": 0.7748, "step": 5780 }, { "epoch": 0.33, "learning_rate": 0.000156950606687107, "loss": 0.6834, "step": 5781 }, { "epoch": 0.33, "learning_rate": 0.00015693550427154385, "loss": 0.6952, "step": 5782 }, { "epoch": 0.33, "learning_rate": 0.00015692039993425808, "loss": 0.6852, "step": 5783 }, { "epoch": 0.33, "learning_rate": 0.00015690529367575956, "loss": 0.8151, "step": 5784 }, { "epoch": 0.33, "learning_rate": 0.00015689018549655813, "loss": 0.6703, "step": 5785 }, { "epoch": 0.33, "learning_rate": 0.00015687507539716378, "loss": 0.7472, "step": 5786 }, { "epoch": 0.33, "learning_rate": 0.00015685996337808642, "loss": 0.7522, "step": 5787 }, { "epoch": 0.33, "learning_rate": 0.00015684484943983618, "loss": 0.6414, "step": 5788 }, { "epoch": 0.33, "learning_rate": 0.00015682973358292323, "loss": 0.7322, "step": 5789 }, { "epoch": 0.33, "learning_rate": 0.0001568146158078577, "loss": 0.6907, "step": 5790 }, { "epoch": 0.33, "learning_rate": 0.0001567994961151499, "loss": 0.7165, "step": 5791 }, { "epoch": 0.33, "learning_rate": 0.00015678437450531013, "loss": 0.6667, "step": 5792 }, { "epoch": 0.33, "learning_rate": 0.0001567692509788488, "loss": 0.7783, "step": 5793 }, { "epoch": 0.33, "learning_rate": 0.00015675412553627639, "loss": 0.6256, "step": 5794 }, { "epoch": 0.33, "learning_rate": 0.00015673899817810337, "loss": 0.6632, "step": 5795 }, { "epoch": 0.33, "learning_rate": 0.00015672386890484038, "loss": 0.7176, "step": 5796 }, { "epoch": 0.33, "learning_rate": 0.00015670873771699804, "loss": 0.6593, "step": 5797 }, { "epoch": 0.33, "learning_rate": 0.00015669360461508707, "loss": 0.8326, "step": 5798 }, { "epoch": 0.33, "learning_rate": 0.00015667846959961827, "loss": 0.799, "step": 5799 }, { "epoch": 0.33, "learning_rate": 0.00015666333267110247, "loss": 0.704, "step": 5800 }, { "epoch": 0.33, "learning_rate": 0.00015664819383005057, "loss": 0.7884, "step": 5801 }, { "epoch": 0.33, "learning_rate": 0.0001566330530769736, "loss": 0.7336, "step": 5802 }, { "epoch": 0.33, "learning_rate": 0.00015661791041238256, "loss": 0.7276, "step": 5803 }, { "epoch": 0.33, "learning_rate": 0.00015660276583678853, "loss": 0.8346, "step": 5804 }, { "epoch": 0.33, "learning_rate": 0.00015658761935070272, "loss": 0.638, "step": 5805 }, { "epoch": 0.33, "learning_rate": 0.00015657247095463635, "loss": 0.7786, "step": 5806 }, { "epoch": 0.33, "learning_rate": 0.0001565573206491007, "loss": 0.7026, "step": 5807 }, { "epoch": 0.33, "learning_rate": 0.0001565421684346072, "loss": 0.6093, "step": 5808 }, { "epoch": 0.33, "learning_rate": 0.0001565270143116672, "loss": 0.809, "step": 5809 }, { "epoch": 0.33, "learning_rate": 0.00015651185828079221, "loss": 0.7501, "step": 5810 }, { "epoch": 0.33, "learning_rate": 0.0001564967003424938, "loss": 0.7111, "step": 5811 }, { "epoch": 0.33, "learning_rate": 0.00015648154049728358, "loss": 0.6341, "step": 5812 }, { "epoch": 0.33, "learning_rate": 0.00015646637874567327, "loss": 0.7646, "step": 5813 }, { "epoch": 0.33, "learning_rate": 0.00015645121508817456, "loss": 0.7381, "step": 5814 }, { "epoch": 0.33, "learning_rate": 0.0001564360495252993, "loss": 0.6831, "step": 5815 }, { "epoch": 0.33, "learning_rate": 0.00015642088205755937, "loss": 0.7487, "step": 5816 }, { "epoch": 0.33, "learning_rate": 0.0001564057126854667, "loss": 0.6502, "step": 5817 }, { "epoch": 0.33, "learning_rate": 0.0001563905414095333, "loss": 0.761, "step": 5818 }, { "epoch": 0.33, "learning_rate": 0.0001563753682302712, "loss": 0.7092, "step": 5819 }, { "epoch": 0.33, "learning_rate": 0.0001563601931481926, "loss": 0.7576, "step": 5820 }, { "epoch": 0.33, "learning_rate": 0.00015634501616380967, "loss": 0.748, "step": 5821 }, { "epoch": 0.33, "learning_rate": 0.00015632983727763468, "loss": 0.7475, "step": 5822 }, { "epoch": 0.33, "learning_rate": 0.00015631465649017993, "loss": 0.6396, "step": 5823 }, { "epoch": 0.33, "learning_rate": 0.00015629947380195786, "loss": 0.7313, "step": 5824 }, { "epoch": 0.33, "learning_rate": 0.00015628428921348088, "loss": 0.5715, "step": 5825 }, { "epoch": 0.33, "learning_rate": 0.00015626910272526154, "loss": 0.6849, "step": 5826 }, { "epoch": 0.33, "learning_rate": 0.0001562539143378124, "loss": 0.6961, "step": 5827 }, { "epoch": 0.33, "learning_rate": 0.00015623872405164612, "loss": 0.7781, "step": 5828 }, { "epoch": 0.33, "learning_rate": 0.00015622353186727544, "loss": 0.7416, "step": 5829 }, { "epoch": 0.33, "learning_rate": 0.00015620833778521307, "loss": 0.6566, "step": 5830 }, { "epoch": 0.33, "learning_rate": 0.00015619314180597189, "loss": 0.8555, "step": 5831 }, { "epoch": 0.33, "learning_rate": 0.00015617794393006483, "loss": 0.7274, "step": 5832 }, { "epoch": 0.33, "learning_rate": 0.00015616274415800483, "loss": 0.701, "step": 5833 }, { "epoch": 0.33, "learning_rate": 0.0001561475424903049, "loss": 0.6602, "step": 5834 }, { "epoch": 0.33, "learning_rate": 0.00015613233892747817, "loss": 0.71, "step": 5835 }, { "epoch": 0.33, "learning_rate": 0.0001561171334700378, "loss": 0.5692, "step": 5836 }, { "epoch": 0.33, "learning_rate": 0.00015610192611849696, "loss": 0.6998, "step": 5837 }, { "epoch": 0.33, "learning_rate": 0.000156086716873369, "loss": 0.7924, "step": 5838 }, { "epoch": 0.33, "learning_rate": 0.0001560715057351673, "loss": 0.8459, "step": 5839 }, { "epoch": 0.33, "learning_rate": 0.00015605629270440519, "loss": 0.7285, "step": 5840 }, { "epoch": 0.33, "learning_rate": 0.00015604107778159615, "loss": 0.7783, "step": 5841 }, { "epoch": 0.33, "learning_rate": 0.00015602586096725383, "loss": 0.7287, "step": 5842 }, { "epoch": 0.33, "learning_rate": 0.00015601064226189172, "loss": 0.7891, "step": 5843 }, { "epoch": 0.33, "learning_rate": 0.00015599542166602358, "loss": 0.7189, "step": 5844 }, { "epoch": 0.33, "learning_rate": 0.0001559801991801631, "loss": 0.6329, "step": 5845 }, { "epoch": 0.33, "learning_rate": 0.00015596497480482407, "loss": 0.7214, "step": 5846 }, { "epoch": 0.33, "learning_rate": 0.00015594974854052035, "loss": 0.7575, "step": 5847 }, { "epoch": 0.33, "learning_rate": 0.0001559345203877659, "loss": 0.7425, "step": 5848 }, { "epoch": 0.33, "learning_rate": 0.0001559192903470747, "loss": 0.7153, "step": 5849 }, { "epoch": 0.33, "learning_rate": 0.00015590405841896078, "loss": 0.7588, "step": 5850 }, { "epoch": 0.33, "learning_rate": 0.00015588882460393827, "loss": 0.7367, "step": 5851 }, { "epoch": 0.33, "learning_rate": 0.0001558735889025214, "loss": 0.6844, "step": 5852 }, { "epoch": 0.33, "learning_rate": 0.00015585835131522432, "loss": 0.7292, "step": 5853 }, { "epoch": 0.33, "learning_rate": 0.0001558431118425614, "loss": 0.7312, "step": 5854 }, { "epoch": 0.33, "learning_rate": 0.00015582787048504707, "loss": 0.6884, "step": 5855 }, { "epoch": 0.33, "learning_rate": 0.0001558126272431956, "loss": 0.7397, "step": 5856 }, { "epoch": 0.33, "learning_rate": 0.00015579738211752165, "loss": 0.6158, "step": 5857 }, { "epoch": 0.33, "learning_rate": 0.0001557821351085397, "loss": 0.6814, "step": 5858 }, { "epoch": 0.33, "learning_rate": 0.00015576688621676436, "loss": 0.7707, "step": 5859 }, { "epoch": 0.33, "learning_rate": 0.0001557516354427104, "loss": 0.8511, "step": 5860 }, { "epoch": 0.33, "learning_rate": 0.0001557363827868925, "loss": 0.779, "step": 5861 }, { "epoch": 0.33, "learning_rate": 0.00015572112824982555, "loss": 0.7305, "step": 5862 }, { "epoch": 0.33, "learning_rate": 0.00015570587183202434, "loss": 0.7192, "step": 5863 }, { "epoch": 0.33, "learning_rate": 0.0001556906135340039, "loss": 0.7611, "step": 5864 }, { "epoch": 0.33, "learning_rate": 0.00015567535335627916, "loss": 0.7228, "step": 5865 }, { "epoch": 0.33, "learning_rate": 0.00015566009129936527, "loss": 0.7318, "step": 5866 }, { "epoch": 0.33, "learning_rate": 0.0001556448273637773, "loss": 0.7268, "step": 5867 }, { "epoch": 0.33, "learning_rate": 0.0001556295615500305, "loss": 0.8186, "step": 5868 }, { "epoch": 0.33, "learning_rate": 0.00015561429385864005, "loss": 0.7084, "step": 5869 }, { "epoch": 0.33, "learning_rate": 0.0001555990242901214, "loss": 0.7453, "step": 5870 }, { "epoch": 0.33, "learning_rate": 0.00015558375284498982, "loss": 0.8065, "step": 5871 }, { "epoch": 0.33, "learning_rate": 0.0001555684795237608, "loss": 0.7082, "step": 5872 }, { "epoch": 0.33, "learning_rate": 0.0001555532043269499, "loss": 0.6968, "step": 5873 }, { "epoch": 0.33, "learning_rate": 0.00015553792725507264, "loss": 0.7387, "step": 5874 }, { "epoch": 0.33, "learning_rate": 0.00015552264830864468, "loss": 0.8174, "step": 5875 }, { "epoch": 0.33, "learning_rate": 0.00015550736748818172, "loss": 0.7345, "step": 5876 }, { "epoch": 0.33, "learning_rate": 0.00015549208479419954, "loss": 0.6394, "step": 5877 }, { "epoch": 0.33, "learning_rate": 0.00015547680022721396, "loss": 0.8161, "step": 5878 }, { "epoch": 0.33, "learning_rate": 0.00015546151378774086, "loss": 0.7706, "step": 5879 }, { "epoch": 0.33, "learning_rate": 0.00015544622547629627, "loss": 0.7449, "step": 5880 }, { "epoch": 0.33, "learning_rate": 0.00015543093529339614, "loss": 0.769, "step": 5881 }, { "epoch": 0.33, "learning_rate": 0.00015541564323955656, "loss": 0.6837, "step": 5882 }, { "epoch": 0.33, "learning_rate": 0.00015540034931529366, "loss": 0.7509, "step": 5883 }, { "epoch": 0.33, "learning_rate": 0.00015538505352112375, "loss": 0.649, "step": 5884 }, { "epoch": 0.33, "learning_rate": 0.00015536975585756296, "loss": 0.5653, "step": 5885 }, { "epoch": 0.33, "learning_rate": 0.00015535445632512774, "loss": 0.7452, "step": 5886 }, { "epoch": 0.33, "learning_rate": 0.00015533915492433443, "loss": 0.7407, "step": 5887 }, { "epoch": 0.33, "learning_rate": 0.00015532385165569948, "loss": 0.6464, "step": 5888 }, { "epoch": 0.33, "learning_rate": 0.00015530854651973948, "loss": 0.6332, "step": 5889 }, { "epoch": 0.33, "learning_rate": 0.000155293239516971, "loss": 0.9134, "step": 5890 }, { "epoch": 0.33, "learning_rate": 0.00015527793064791062, "loss": 0.703, "step": 5891 }, { "epoch": 0.33, "learning_rate": 0.00015526261991307517, "loss": 0.7123, "step": 5892 }, { "epoch": 0.33, "learning_rate": 0.00015524730731298134, "loss": 0.7376, "step": 5893 }, { "epoch": 0.33, "learning_rate": 0.000155231992848146, "loss": 0.7357, "step": 5894 }, { "epoch": 0.33, "learning_rate": 0.00015521667651908602, "loss": 0.6851, "step": 5895 }, { "epoch": 0.33, "learning_rate": 0.00015520135832631843, "loss": 0.675, "step": 5896 }, { "epoch": 0.33, "learning_rate": 0.00015518603827036022, "loss": 0.6992, "step": 5897 }, { "epoch": 0.33, "learning_rate": 0.00015517071635172847, "loss": 0.7057, "step": 5898 }, { "epoch": 0.33, "learning_rate": 0.00015515539257094036, "loss": 0.6973, "step": 5899 }, { "epoch": 0.33, "learning_rate": 0.00015514006692851311, "loss": 0.5757, "step": 5900 }, { "epoch": 0.33, "learning_rate": 0.000155124739424964, "loss": 0.6378, "step": 5901 }, { "epoch": 0.33, "learning_rate": 0.00015510941006081032, "loss": 0.7285, "step": 5902 }, { "epoch": 0.33, "learning_rate": 0.00015509407883656959, "loss": 0.7305, "step": 5903 }, { "epoch": 0.33, "learning_rate": 0.00015507874575275917, "loss": 0.7338, "step": 5904 }, { "epoch": 0.33, "learning_rate": 0.00015506341080989664, "loss": 0.7685, "step": 5905 }, { "epoch": 0.34, "learning_rate": 0.00015504807400849958, "loss": 0.7381, "step": 5906 }, { "epoch": 0.34, "learning_rate": 0.00015503273534908565, "loss": 0.7152, "step": 5907 }, { "epoch": 0.34, "learning_rate": 0.00015501739483217257, "loss": 0.7579, "step": 5908 }, { "epoch": 0.34, "learning_rate": 0.00015500205245827812, "loss": 0.6913, "step": 5909 }, { "epoch": 0.34, "learning_rate": 0.00015498670822792018, "loss": 0.7479, "step": 5910 }, { "epoch": 0.34, "learning_rate": 0.00015497136214161664, "loss": 0.6324, "step": 5911 }, { "epoch": 0.34, "learning_rate": 0.0001549560141998854, "loss": 0.8649, "step": 5912 }, { "epoch": 0.34, "learning_rate": 0.00015494066440324458, "loss": 0.726, "step": 5913 }, { "epoch": 0.34, "learning_rate": 0.00015492531275221227, "loss": 0.6739, "step": 5914 }, { "epoch": 0.34, "learning_rate": 0.0001549099592473066, "loss": 0.8069, "step": 5915 }, { "epoch": 0.34, "learning_rate": 0.00015489460388904577, "loss": 0.732, "step": 5916 }, { "epoch": 0.34, "learning_rate": 0.0001548792466779481, "loss": 0.7453, "step": 5917 }, { "epoch": 0.34, "learning_rate": 0.000154863887614532, "loss": 0.7578, "step": 5918 }, { "epoch": 0.34, "learning_rate": 0.00015484852669931575, "loss": 0.7449, "step": 5919 }, { "epoch": 0.34, "learning_rate": 0.00015483316393281784, "loss": 0.7403, "step": 5920 }, { "epoch": 0.34, "learning_rate": 0.00015481779931555688, "loss": 0.718, "step": 5921 }, { "epoch": 0.34, "learning_rate": 0.00015480243284805143, "loss": 0.7652, "step": 5922 }, { "epoch": 0.34, "learning_rate": 0.00015478706453082017, "loss": 0.7584, "step": 5923 }, { "epoch": 0.34, "learning_rate": 0.00015477169436438177, "loss": 0.7605, "step": 5924 }, { "epoch": 0.34, "learning_rate": 0.00015475632234925504, "loss": 0.6043, "step": 5925 }, { "epoch": 0.34, "learning_rate": 0.00015474094848595883, "loss": 0.7873, "step": 5926 }, { "epoch": 0.34, "learning_rate": 0.00015472557277501205, "loss": 0.6897, "step": 5927 }, { "epoch": 0.34, "learning_rate": 0.00015471019521693364, "loss": 0.7726, "step": 5928 }, { "epoch": 0.34, "learning_rate": 0.00015469481581224272, "loss": 0.7669, "step": 5929 }, { "epoch": 0.34, "learning_rate": 0.00015467943456145826, "loss": 0.7122, "step": 5930 }, { "epoch": 0.34, "learning_rate": 0.0001546640514650995, "loss": 0.7023, "step": 5931 }, { "epoch": 0.34, "learning_rate": 0.00015464866652368565, "loss": 0.7322, "step": 5932 }, { "epoch": 0.34, "learning_rate": 0.00015463327973773598, "loss": 0.7492, "step": 5933 }, { "epoch": 0.34, "learning_rate": 0.00015461789110776986, "loss": 0.6609, "step": 5934 }, { "epoch": 0.34, "learning_rate": 0.00015460250063430663, "loss": 0.7353, "step": 5935 }, { "epoch": 0.34, "learning_rate": 0.00015458710831786586, "loss": 0.7605, "step": 5936 }, { "epoch": 0.34, "learning_rate": 0.00015457171415896696, "loss": 0.7326, "step": 5937 }, { "epoch": 0.34, "learning_rate": 0.00015455631815812962, "loss": 0.7559, "step": 5938 }, { "epoch": 0.34, "learning_rate": 0.00015454092031587347, "loss": 0.744, "step": 5939 }, { "epoch": 0.34, "learning_rate": 0.0001545255206327182, "loss": 0.859, "step": 5940 }, { "epoch": 0.34, "learning_rate": 0.00015451011910918365, "loss": 0.6535, "step": 5941 }, { "epoch": 0.34, "learning_rate": 0.00015449471574578957, "loss": 0.6437, "step": 5942 }, { "epoch": 0.34, "learning_rate": 0.00015447931054305596, "loss": 0.7795, "step": 5943 }, { "epoch": 0.34, "learning_rate": 0.00015446390350150273, "loss": 0.7946, "step": 5944 }, { "epoch": 0.34, "learning_rate": 0.0001544484946216499, "loss": 0.7614, "step": 5945 }, { "epoch": 0.34, "learning_rate": 0.0001544330839040176, "loss": 0.7362, "step": 5946 }, { "epoch": 0.34, "learning_rate": 0.00015441767134912596, "loss": 0.7668, "step": 5947 }, { "epoch": 0.34, "learning_rate": 0.00015440225695749517, "loss": 0.7074, "step": 5948 }, { "epoch": 0.34, "learning_rate": 0.00015438684072964555, "loss": 0.7624, "step": 5949 }, { "epoch": 0.34, "learning_rate": 0.00015437142266609742, "loss": 0.7569, "step": 5950 }, { "epoch": 0.34, "learning_rate": 0.0001543560027673712, "loss": 0.7129, "step": 5951 }, { "epoch": 0.34, "learning_rate": 0.00015434058103398735, "loss": 0.7486, "step": 5952 }, { "epoch": 0.34, "learning_rate": 0.00015432515746646632, "loss": 0.6694, "step": 5953 }, { "epoch": 0.34, "learning_rate": 0.00015430973206532878, "loss": 0.7133, "step": 5954 }, { "epoch": 0.34, "learning_rate": 0.00015429430483109534, "loss": 0.8029, "step": 5955 }, { "epoch": 0.34, "learning_rate": 0.00015427887576428674, "loss": 0.7479, "step": 5956 }, { "epoch": 0.34, "learning_rate": 0.0001542634448654237, "loss": 0.6782, "step": 5957 }, { "epoch": 0.34, "learning_rate": 0.00015424801213502712, "loss": 0.7368, "step": 5958 }, { "epoch": 0.34, "learning_rate": 0.00015423257757361784, "loss": 0.7804, "step": 5959 }, { "epoch": 0.34, "learning_rate": 0.00015421714118171685, "loss": 0.7636, "step": 5960 }, { "epoch": 0.34, "learning_rate": 0.00015420170295984514, "loss": 0.7503, "step": 5961 }, { "epoch": 0.34, "learning_rate": 0.0001541862629085238, "loss": 0.7468, "step": 5962 }, { "epoch": 0.34, "learning_rate": 0.000154170821028274, "loss": 0.7549, "step": 5963 }, { "epoch": 0.34, "learning_rate": 0.0001541553773196169, "loss": 0.7542, "step": 5964 }, { "epoch": 0.34, "learning_rate": 0.0001541399317830738, "loss": 0.7303, "step": 5965 }, { "epoch": 0.34, "learning_rate": 0.00015412448441916605, "loss": 0.6744, "step": 5966 }, { "epoch": 0.34, "learning_rate": 0.00015410903522841495, "loss": 0.7597, "step": 5967 }, { "epoch": 0.34, "learning_rate": 0.000154093584211342, "loss": 0.7799, "step": 5968 }, { "epoch": 0.34, "learning_rate": 0.00015407813136846877, "loss": 0.811, "step": 5969 }, { "epoch": 0.34, "learning_rate": 0.00015406267670031674, "loss": 0.7159, "step": 5970 }, { "epoch": 0.34, "learning_rate": 0.0001540472202074076, "loss": 0.7509, "step": 5971 }, { "epoch": 0.34, "learning_rate": 0.00015403176189026304, "loss": 0.8018, "step": 5972 }, { "epoch": 0.34, "learning_rate": 0.00015401630174940482, "loss": 0.7815, "step": 5973 }, { "epoch": 0.34, "learning_rate": 0.00015400083978535473, "loss": 0.7101, "step": 5974 }, { "epoch": 0.34, "learning_rate": 0.00015398537599863466, "loss": 0.7059, "step": 5975 }, { "epoch": 0.34, "learning_rate": 0.00015396991038976661, "loss": 0.6807, "step": 5976 }, { "epoch": 0.34, "learning_rate": 0.00015395444295927252, "loss": 0.74, "step": 5977 }, { "epoch": 0.34, "learning_rate": 0.0001539389737076745, "loss": 0.7078, "step": 5978 }, { "epoch": 0.34, "learning_rate": 0.0001539235026354946, "loss": 0.8038, "step": 5979 }, { "epoch": 0.34, "learning_rate": 0.00015390802974325513, "loss": 0.7541, "step": 5980 }, { "epoch": 0.34, "learning_rate": 0.00015389255503147825, "loss": 0.7402, "step": 5981 }, { "epoch": 0.34, "learning_rate": 0.0001538770785006863, "loss": 0.6851, "step": 5982 }, { "epoch": 0.34, "learning_rate": 0.00015386160015140168, "loss": 0.693, "step": 5983 }, { "epoch": 0.34, "learning_rate": 0.00015384611998414677, "loss": 0.6431, "step": 5984 }, { "epoch": 0.34, "learning_rate": 0.00015383063799944408, "loss": 0.6022, "step": 5985 }, { "epoch": 0.34, "learning_rate": 0.00015381515419781624, "loss": 0.7938, "step": 5986 }, { "epoch": 0.34, "learning_rate": 0.00015379966857978573, "loss": 0.7102, "step": 5987 }, { "epoch": 0.34, "learning_rate": 0.0001537841811458754, "loss": 0.6963, "step": 5988 }, { "epoch": 0.34, "learning_rate": 0.00015376869189660783, "loss": 0.7967, "step": 5989 }, { "epoch": 0.34, "learning_rate": 0.00015375320083250592, "loss": 0.6884, "step": 5990 }, { "epoch": 0.34, "learning_rate": 0.0001537377079540925, "loss": 0.7704, "step": 5991 }, { "epoch": 0.34, "learning_rate": 0.0001537222132618905, "loss": 0.6488, "step": 5992 }, { "epoch": 0.34, "learning_rate": 0.00015370671675642294, "loss": 0.7222, "step": 5993 }, { "epoch": 0.34, "learning_rate": 0.00015369121843821282, "loss": 0.7309, "step": 5994 }, { "epoch": 0.34, "learning_rate": 0.00015367571830778327, "loss": 0.7307, "step": 5995 }, { "epoch": 0.34, "learning_rate": 0.00015366021636565745, "loss": 0.8238, "step": 5996 }, { "epoch": 0.34, "learning_rate": 0.0001536447126123586, "loss": 0.7918, "step": 5997 }, { "epoch": 0.34, "learning_rate": 0.00015362920704841, "loss": 0.7617, "step": 5998 }, { "epoch": 0.34, "learning_rate": 0.00015361369967433505, "loss": 0.7162, "step": 5999 }, { "epoch": 0.34, "learning_rate": 0.00015359819049065712, "loss": 0.6691, "step": 6000 }, { "epoch": 0.34, "learning_rate": 0.00015358267949789966, "loss": 0.689, "step": 6001 }, { "epoch": 0.34, "learning_rate": 0.0001535671666965863, "loss": 0.7339, "step": 6002 }, { "epoch": 0.34, "learning_rate": 0.00015355165208724052, "loss": 0.632, "step": 6003 }, { "epoch": 0.34, "learning_rate": 0.00015353613567038607, "loss": 0.8238, "step": 6004 }, { "epoch": 0.34, "learning_rate": 0.00015352061744654662, "loss": 0.7942, "step": 6005 }, { "epoch": 0.34, "learning_rate": 0.00015350509741624596, "loss": 0.7048, "step": 6006 }, { "epoch": 0.34, "learning_rate": 0.00015348957558000795, "loss": 0.7959, "step": 6007 }, { "epoch": 0.34, "learning_rate": 0.0001534740519383565, "loss": 0.7382, "step": 6008 }, { "epoch": 0.34, "learning_rate": 0.00015345852649181556, "loss": 0.7816, "step": 6009 }, { "epoch": 0.34, "learning_rate": 0.00015344299924090915, "loss": 0.6964, "step": 6010 }, { "epoch": 0.34, "learning_rate": 0.00015342747018616133, "loss": 0.7579, "step": 6011 }, { "epoch": 0.34, "learning_rate": 0.0001534119393280963, "loss": 0.6741, "step": 6012 }, { "epoch": 0.34, "learning_rate": 0.0001533964066672382, "loss": 0.6988, "step": 6013 }, { "epoch": 0.34, "learning_rate": 0.00015338087220411137, "loss": 0.878, "step": 6014 }, { "epoch": 0.34, "learning_rate": 0.00015336533593924008, "loss": 0.6989, "step": 6015 }, { "epoch": 0.34, "learning_rate": 0.0001533497978731488, "loss": 0.7532, "step": 6016 }, { "epoch": 0.34, "learning_rate": 0.0001533342580063619, "loss": 0.7851, "step": 6017 }, { "epoch": 0.34, "learning_rate": 0.00015331871633940386, "loss": 0.6918, "step": 6018 }, { "epoch": 0.34, "learning_rate": 0.0001533031728727994, "loss": 0.606, "step": 6019 }, { "epoch": 0.34, "learning_rate": 0.000153287627607073, "loss": 0.7411, "step": 6020 }, { "epoch": 0.34, "learning_rate": 0.00015327208054274945, "loss": 0.7852, "step": 6021 }, { "epoch": 0.34, "learning_rate": 0.00015325653168035344, "loss": 0.6467, "step": 6022 }, { "epoch": 0.34, "learning_rate": 0.00015324098102040983, "loss": 0.7187, "step": 6023 }, { "epoch": 0.34, "learning_rate": 0.0001532254285634435, "loss": 0.7452, "step": 6024 }, { "epoch": 0.34, "learning_rate": 0.00015320987430997939, "loss": 0.8356, "step": 6025 }, { "epoch": 0.34, "learning_rate": 0.00015319431826054243, "loss": 0.637, "step": 6026 }, { "epoch": 0.34, "learning_rate": 0.0001531787604156577, "loss": 0.7592, "step": 6027 }, { "epoch": 0.34, "learning_rate": 0.00015316320077585037, "loss": 0.6305, "step": 6028 }, { "epoch": 0.34, "learning_rate": 0.0001531476393416456, "loss": 0.7231, "step": 6029 }, { "epoch": 0.34, "learning_rate": 0.00015313207611356862, "loss": 0.789, "step": 6030 }, { "epoch": 0.34, "learning_rate": 0.0001531165110921447, "loss": 0.7689, "step": 6031 }, { "epoch": 0.34, "learning_rate": 0.00015310094427789923, "loss": 0.7037, "step": 6032 }, { "epoch": 0.34, "learning_rate": 0.00015308537567135764, "loss": 0.7447, "step": 6033 }, { "epoch": 0.34, "learning_rate": 0.00015306980527304538, "loss": 0.745, "step": 6034 }, { "epoch": 0.34, "learning_rate": 0.00015305423308348805, "loss": 0.79, "step": 6035 }, { "epoch": 0.34, "learning_rate": 0.00015303865910321117, "loss": 0.7511, "step": 6036 }, { "epoch": 0.34, "learning_rate": 0.0001530230833327405, "loss": 0.6622, "step": 6037 }, { "epoch": 0.34, "learning_rate": 0.00015300750577260165, "loss": 0.6731, "step": 6038 }, { "epoch": 0.34, "learning_rate": 0.0001529919264233205, "loss": 0.8238, "step": 6039 }, { "epoch": 0.34, "learning_rate": 0.00015297634528542285, "loss": 0.6344, "step": 6040 }, { "epoch": 0.34, "learning_rate": 0.0001529607623594346, "loss": 0.7166, "step": 6041 }, { "epoch": 0.34, "learning_rate": 0.00015294517764588174, "loss": 0.6563, "step": 6042 }, { "epoch": 0.34, "learning_rate": 0.00015292959114529027, "loss": 0.7619, "step": 6043 }, { "epoch": 0.34, "learning_rate": 0.00015291400285818632, "loss": 0.6926, "step": 6044 }, { "epoch": 0.34, "learning_rate": 0.00015289841278509595, "loss": 0.6842, "step": 6045 }, { "epoch": 0.34, "learning_rate": 0.00015288282092654543, "loss": 0.6538, "step": 6046 }, { "epoch": 0.34, "learning_rate": 0.00015286722728306105, "loss": 0.6871, "step": 6047 }, { "epoch": 0.34, "learning_rate": 0.00015285163185516906, "loss": 0.6636, "step": 6048 }, { "epoch": 0.34, "learning_rate": 0.0001528360346433959, "loss": 0.7408, "step": 6049 }, { "epoch": 0.34, "learning_rate": 0.000152820435648268, "loss": 0.7104, "step": 6050 }, { "epoch": 0.34, "learning_rate": 0.0001528048348703119, "loss": 0.6895, "step": 6051 }, { "epoch": 0.34, "learning_rate": 0.00015278923231005408, "loss": 0.7605, "step": 6052 }, { "epoch": 0.34, "learning_rate": 0.00015277362796802123, "loss": 0.8369, "step": 6053 }, { "epoch": 0.34, "learning_rate": 0.0001527580218447401, "loss": 0.7704, "step": 6054 }, { "epoch": 0.34, "learning_rate": 0.00015274241394073733, "loss": 0.7761, "step": 6055 }, { "epoch": 0.34, "learning_rate": 0.00015272680425653973, "loss": 0.7047, "step": 6056 }, { "epoch": 0.34, "learning_rate": 0.00015271119279267425, "loss": 0.7253, "step": 6057 }, { "epoch": 0.34, "learning_rate": 0.00015269557954966778, "loss": 0.7381, "step": 6058 }, { "epoch": 0.34, "learning_rate": 0.00015267996452804727, "loss": 0.7994, "step": 6059 }, { "epoch": 0.34, "learning_rate": 0.00015266434772833983, "loss": 0.7647, "step": 6060 }, { "epoch": 0.34, "learning_rate": 0.0001526487291510725, "loss": 0.7466, "step": 6061 }, { "epoch": 0.34, "learning_rate": 0.0001526331087967725, "loss": 0.6627, "step": 6062 }, { "epoch": 0.34, "learning_rate": 0.00015261748666596704, "loss": 0.7266, "step": 6063 }, { "epoch": 0.34, "learning_rate": 0.00015260186275918342, "loss": 0.7194, "step": 6064 }, { "epoch": 0.34, "learning_rate": 0.000152586237076949, "loss": 0.7775, "step": 6065 }, { "epoch": 0.34, "learning_rate": 0.00015257060961979113, "loss": 0.6708, "step": 6066 }, { "epoch": 0.34, "learning_rate": 0.0001525549803882373, "loss": 0.7051, "step": 6067 }, { "epoch": 0.34, "learning_rate": 0.0001525393493828151, "loss": 0.7773, "step": 6068 }, { "epoch": 0.34, "learning_rate": 0.00015252371660405203, "loss": 0.7942, "step": 6069 }, { "epoch": 0.34, "learning_rate": 0.00015250808205247583, "loss": 0.6992, "step": 6070 }, { "epoch": 0.34, "learning_rate": 0.0001524924457286141, "loss": 0.7047, "step": 6071 }, { "epoch": 0.34, "learning_rate": 0.00015247680763299466, "loss": 0.7626, "step": 6072 }, { "epoch": 0.34, "learning_rate": 0.00015246116776614538, "loss": 0.697, "step": 6073 }, { "epoch": 0.34, "learning_rate": 0.00015244552612859406, "loss": 0.7528, "step": 6074 }, { "epoch": 0.34, "learning_rate": 0.0001524298827208687, "loss": 0.7564, "step": 6075 }, { "epoch": 0.34, "learning_rate": 0.00015241423754349732, "loss": 0.7128, "step": 6076 }, { "epoch": 0.34, "learning_rate": 0.00015239859059700794, "loss": 0.7606, "step": 6077 }, { "epoch": 0.34, "learning_rate": 0.00015238294188192869, "loss": 0.7576, "step": 6078 }, { "epoch": 0.34, "learning_rate": 0.00015236729139878782, "loss": 0.7478, "step": 6079 }, { "epoch": 0.34, "learning_rate": 0.00015235163914811346, "loss": 0.6314, "step": 6080 }, { "epoch": 0.34, "learning_rate": 0.00015233598513043407, "loss": 0.7485, "step": 6081 }, { "epoch": 0.34, "learning_rate": 0.00015232032934627782, "loss": 0.7764, "step": 6082 }, { "epoch": 0.35, "learning_rate": 0.00015230467179617332, "loss": 0.7592, "step": 6083 }, { "epoch": 0.35, "learning_rate": 0.00015228901248064895, "loss": 0.7771, "step": 6084 }, { "epoch": 0.35, "learning_rate": 0.00015227335140023326, "loss": 0.8415, "step": 6085 }, { "epoch": 0.35, "learning_rate": 0.0001522576885554549, "loss": 0.7756, "step": 6086 }, { "epoch": 0.35, "learning_rate": 0.00015224202394684247, "loss": 0.751, "step": 6087 }, { "epoch": 0.35, "learning_rate": 0.00015222635757492476, "loss": 0.7136, "step": 6088 }, { "epoch": 0.35, "learning_rate": 0.00015221068944023047, "loss": 0.6577, "step": 6089 }, { "epoch": 0.35, "learning_rate": 0.0001521950195432885, "loss": 0.6828, "step": 6090 }, { "epoch": 0.35, "learning_rate": 0.00015217934788462774, "loss": 0.6365, "step": 6091 }, { "epoch": 0.35, "learning_rate": 0.00015216367446477716, "loss": 0.734, "step": 6092 }, { "epoch": 0.35, "learning_rate": 0.00015214799928426573, "loss": 0.7243, "step": 6093 }, { "epoch": 0.35, "learning_rate": 0.00015213232234362255, "loss": 0.6313, "step": 6094 }, { "epoch": 0.35, "learning_rate": 0.00015211664364337682, "loss": 0.68, "step": 6095 }, { "epoch": 0.35, "learning_rate": 0.00015210096318405767, "loss": 0.7947, "step": 6096 }, { "epoch": 0.35, "learning_rate": 0.00015208528096619434, "loss": 0.7058, "step": 6097 }, { "epoch": 0.35, "learning_rate": 0.0001520695969903162, "loss": 0.7929, "step": 6098 }, { "epoch": 0.35, "learning_rate": 0.0001520539112569526, "loss": 0.611, "step": 6099 }, { "epoch": 0.35, "learning_rate": 0.000152038223766633, "loss": 0.6076, "step": 6100 }, { "epoch": 0.35, "learning_rate": 0.00015202253451988682, "loss": 0.6753, "step": 6101 }, { "epoch": 0.35, "learning_rate": 0.00015200684351724373, "loss": 0.733, "step": 6102 }, { "epoch": 0.35, "learning_rate": 0.00015199115075923325, "loss": 0.6642, "step": 6103 }, { "epoch": 0.35, "learning_rate": 0.00015197545624638504, "loss": 0.6948, "step": 6104 }, { "epoch": 0.35, "learning_rate": 0.00015195975997922892, "loss": 0.7545, "step": 6105 }, { "epoch": 0.35, "learning_rate": 0.00015194406195829464, "loss": 0.7395, "step": 6106 }, { "epoch": 0.35, "learning_rate": 0.000151928362184112, "loss": 0.7539, "step": 6107 }, { "epoch": 0.35, "learning_rate": 0.000151912660657211, "loss": 0.7478, "step": 6108 }, { "epoch": 0.35, "learning_rate": 0.00015189695737812152, "loss": 0.5923, "step": 6109 }, { "epoch": 0.35, "learning_rate": 0.00015188125234737366, "loss": 0.6963, "step": 6110 }, { "epoch": 0.35, "learning_rate": 0.00015186554556549742, "loss": 0.6857, "step": 6111 }, { "epoch": 0.35, "learning_rate": 0.00015184983703302301, "loss": 0.6095, "step": 6112 }, { "epoch": 0.35, "learning_rate": 0.00015183412675048063, "loss": 0.764, "step": 6113 }, { "epoch": 0.35, "learning_rate": 0.00015181841471840053, "loss": 0.7517, "step": 6114 }, { "epoch": 0.35, "learning_rate": 0.00015180270093731303, "loss": 0.7563, "step": 6115 }, { "epoch": 0.35, "learning_rate": 0.0001517869854077485, "loss": 0.7236, "step": 6116 }, { "epoch": 0.35, "learning_rate": 0.00015177126813023742, "loss": 0.7835, "step": 6117 }, { "epoch": 0.35, "learning_rate": 0.0001517555491053103, "loss": 0.7461, "step": 6118 }, { "epoch": 0.35, "learning_rate": 0.0001517398283334976, "loss": 0.7681, "step": 6119 }, { "epoch": 0.35, "learning_rate": 0.00015172410581533002, "loss": 0.7558, "step": 6120 }, { "epoch": 0.35, "learning_rate": 0.00015170838155133823, "loss": 0.7569, "step": 6121 }, { "epoch": 0.35, "learning_rate": 0.00015169265554205294, "loss": 0.7321, "step": 6122 }, { "epoch": 0.35, "learning_rate": 0.00015167692778800495, "loss": 0.7206, "step": 6123 }, { "epoch": 0.35, "learning_rate": 0.00015166119828972514, "loss": 0.772, "step": 6124 }, { "epoch": 0.35, "learning_rate": 0.0001516454670477444, "loss": 0.7955, "step": 6125 }, { "epoch": 0.35, "learning_rate": 0.0001516297340625937, "loss": 0.711, "step": 6126 }, { "epoch": 0.35, "learning_rate": 0.00015161399933480402, "loss": 0.6542, "step": 6127 }, { "epoch": 0.35, "learning_rate": 0.00015159826286490657, "loss": 0.7194, "step": 6128 }, { "epoch": 0.35, "learning_rate": 0.00015158252465343242, "loss": 0.8295, "step": 6129 }, { "epoch": 0.35, "learning_rate": 0.00015156678470091273, "loss": 0.7146, "step": 6130 }, { "epoch": 0.35, "learning_rate": 0.00015155104300787887, "loss": 0.7802, "step": 6131 }, { "epoch": 0.35, "learning_rate": 0.0001515352995748621, "loss": 0.8739, "step": 6132 }, { "epoch": 0.35, "learning_rate": 0.0001515195544023938, "loss": 0.7651, "step": 6133 }, { "epoch": 0.35, "learning_rate": 0.00015150380749100545, "loss": 0.6746, "step": 6134 }, { "epoch": 0.35, "learning_rate": 0.0001514880588412285, "loss": 0.7877, "step": 6135 }, { "epoch": 0.35, "learning_rate": 0.00015147230845359452, "loss": 0.6355, "step": 6136 }, { "epoch": 0.35, "learning_rate": 0.00015145655632863516, "loss": 0.742, "step": 6137 }, { "epoch": 0.35, "learning_rate": 0.00015144080246688206, "loss": 0.8059, "step": 6138 }, { "epoch": 0.35, "learning_rate": 0.00015142504686886698, "loss": 0.7662, "step": 6139 }, { "epoch": 0.35, "learning_rate": 0.0001514092895351217, "loss": 0.861, "step": 6140 }, { "epoch": 0.35, "learning_rate": 0.00015139353046617808, "loss": 0.6388, "step": 6141 }, { "epoch": 0.35, "learning_rate": 0.000151377769662568, "loss": 0.5886, "step": 6142 }, { "epoch": 0.35, "learning_rate": 0.00015136200712482346, "loss": 0.7155, "step": 6143 }, { "epoch": 0.35, "learning_rate": 0.0001513462428534765, "loss": 0.7649, "step": 6144 }, { "epoch": 0.35, "learning_rate": 0.00015133047684905916, "loss": 0.7497, "step": 6145 }, { "epoch": 0.35, "learning_rate": 0.00015131470911210356, "loss": 0.722, "step": 6146 }, { "epoch": 0.35, "learning_rate": 0.00015129893964314204, "loss": 0.7545, "step": 6147 }, { "epoch": 0.35, "learning_rate": 0.00015128316844270674, "loss": 0.7147, "step": 6148 }, { "epoch": 0.35, "learning_rate": 0.00015126739551132997, "loss": 0.7482, "step": 6149 }, { "epoch": 0.35, "learning_rate": 0.0001512516208495442, "loss": 0.7502, "step": 6150 }, { "epoch": 0.35, "learning_rate": 0.00015123584445788182, "loss": 0.7137, "step": 6151 }, { "epoch": 0.35, "learning_rate": 0.00015122006633687526, "loss": 0.7515, "step": 6152 }, { "epoch": 0.35, "learning_rate": 0.00015120428648705717, "loss": 0.7485, "step": 6153 }, { "epoch": 0.35, "learning_rate": 0.00015118850490896012, "loss": 0.8391, "step": 6154 }, { "epoch": 0.35, "learning_rate": 0.0001511727216031168, "loss": 0.7645, "step": 6155 }, { "epoch": 0.35, "learning_rate": 0.00015115693657005984, "loss": 0.6669, "step": 6156 }, { "epoch": 0.35, "learning_rate": 0.00015114114981032222, "loss": 0.7263, "step": 6157 }, { "epoch": 0.35, "learning_rate": 0.0001511253613244366, "loss": 0.7595, "step": 6158 }, { "epoch": 0.35, "learning_rate": 0.00015110957111293596, "loss": 0.7859, "step": 6159 }, { "epoch": 0.35, "learning_rate": 0.00015109377917635328, "loss": 0.7146, "step": 6160 }, { "epoch": 0.35, "learning_rate": 0.00015107798551522152, "loss": 0.7341, "step": 6161 }, { "epoch": 0.35, "learning_rate": 0.00015106219013007383, "loss": 0.7154, "step": 6162 }, { "epoch": 0.35, "learning_rate": 0.00015104639302144327, "loss": 0.6666, "step": 6163 }, { "epoch": 0.35, "learning_rate": 0.0001510305941898631, "loss": 0.7126, "step": 6164 }, { "epoch": 0.35, "learning_rate": 0.0001510147936358665, "loss": 0.7066, "step": 6165 }, { "epoch": 0.35, "learning_rate": 0.0001509989913599869, "loss": 0.7063, "step": 6166 }, { "epoch": 0.35, "learning_rate": 0.00015098318736275757, "loss": 0.7429, "step": 6167 }, { "epoch": 0.35, "learning_rate": 0.0001509673816447119, "loss": 0.6386, "step": 6168 }, { "epoch": 0.35, "learning_rate": 0.00015095157420638348, "loss": 0.7613, "step": 6169 }, { "epoch": 0.35, "learning_rate": 0.0001509357650483058, "loss": 0.7517, "step": 6170 }, { "epoch": 0.35, "learning_rate": 0.00015091995417101248, "loss": 0.7779, "step": 6171 }, { "epoch": 0.35, "learning_rate": 0.00015090414157503714, "loss": 0.7988, "step": 6172 }, { "epoch": 0.35, "learning_rate": 0.00015088832726091353, "loss": 0.7586, "step": 6173 }, { "epoch": 0.35, "learning_rate": 0.00015087251122917542, "loss": 0.7948, "step": 6174 }, { "epoch": 0.35, "learning_rate": 0.00015085669348035663, "loss": 0.8747, "step": 6175 }, { "epoch": 0.35, "learning_rate": 0.00015084087401499106, "loss": 0.6786, "step": 6176 }, { "epoch": 0.35, "learning_rate": 0.00015082505283361267, "loss": 0.7194, "step": 6177 }, { "epoch": 0.35, "learning_rate": 0.0001508092299367554, "loss": 0.7241, "step": 6178 }, { "epoch": 0.35, "learning_rate": 0.00015079340532495343, "loss": 0.7744, "step": 6179 }, { "epoch": 0.35, "learning_rate": 0.0001507775789987408, "loss": 0.759, "step": 6180 }, { "epoch": 0.35, "learning_rate": 0.0001507617509586517, "loss": 0.7487, "step": 6181 }, { "epoch": 0.35, "learning_rate": 0.00015074592120522038, "loss": 0.6273, "step": 6182 }, { "epoch": 0.35, "learning_rate": 0.00015073008973898113, "loss": 0.6693, "step": 6183 }, { "epoch": 0.35, "learning_rate": 0.00015071425656046833, "loss": 0.7044, "step": 6184 }, { "epoch": 0.35, "learning_rate": 0.00015069842167021635, "loss": 0.6543, "step": 6185 }, { "epoch": 0.35, "learning_rate": 0.0001506825850687597, "loss": 0.7591, "step": 6186 }, { "epoch": 0.35, "learning_rate": 0.00015066674675663283, "loss": 0.7602, "step": 6187 }, { "epoch": 0.35, "learning_rate": 0.00015065090673437042, "loss": 0.7014, "step": 6188 }, { "epoch": 0.35, "learning_rate": 0.00015063506500250708, "loss": 0.6218, "step": 6189 }, { "epoch": 0.35, "learning_rate": 0.00015061922156157746, "loss": 0.6606, "step": 6190 }, { "epoch": 0.35, "learning_rate": 0.00015060337641211637, "loss": 0.7436, "step": 6191 }, { "epoch": 0.35, "learning_rate": 0.00015058752955465866, "loss": 0.7524, "step": 6192 }, { "epoch": 0.35, "learning_rate": 0.0001505716809897391, "loss": 0.7124, "step": 6193 }, { "epoch": 0.35, "learning_rate": 0.0001505558307178927, "loss": 0.6899, "step": 6194 }, { "epoch": 0.35, "learning_rate": 0.00015053997873965444, "loss": 0.7129, "step": 6195 }, { "epoch": 0.35, "learning_rate": 0.00015052412505555936, "loss": 0.8497, "step": 6196 }, { "epoch": 0.35, "learning_rate": 0.00015050826966614251, "loss": 0.6801, "step": 6197 }, { "epoch": 0.35, "learning_rate": 0.00015049241257193908, "loss": 0.5967, "step": 6198 }, { "epoch": 0.35, "learning_rate": 0.0001504765537734844, "loss": 0.6761, "step": 6199 }, { "epoch": 0.35, "learning_rate": 0.00015046069327131359, "loss": 0.7116, "step": 6200 }, { "epoch": 0.35, "learning_rate": 0.00015044483106596205, "loss": 0.7014, "step": 6201 }, { "epoch": 0.35, "learning_rate": 0.00015042896715796516, "loss": 0.7223, "step": 6202 }, { "epoch": 0.35, "learning_rate": 0.00015041310154785838, "loss": 0.7342, "step": 6203 }, { "epoch": 0.35, "learning_rate": 0.0001503972342361772, "loss": 0.7517, "step": 6204 }, { "epoch": 0.35, "learning_rate": 0.00015038136522345719, "loss": 0.8012, "step": 6205 }, { "epoch": 0.35, "learning_rate": 0.00015036549451023396, "loss": 0.7577, "step": 6206 }, { "epoch": 0.35, "learning_rate": 0.00015034962209704324, "loss": 0.7492, "step": 6207 }, { "epoch": 0.35, "learning_rate": 0.00015033374798442074, "loss": 0.7479, "step": 6208 }, { "epoch": 0.35, "learning_rate": 0.0001503178721729022, "loss": 0.678, "step": 6209 }, { "epoch": 0.35, "learning_rate": 0.00015030199466302353, "loss": 0.7942, "step": 6210 }, { "epoch": 0.35, "learning_rate": 0.00015028611545532061, "loss": 0.7215, "step": 6211 }, { "epoch": 0.35, "learning_rate": 0.0001502702345503294, "loss": 0.6845, "step": 6212 }, { "epoch": 0.35, "learning_rate": 0.00015025435194858595, "loss": 0.6773, "step": 6213 }, { "epoch": 0.35, "learning_rate": 0.00015023846765062633, "loss": 0.7018, "step": 6214 }, { "epoch": 0.35, "learning_rate": 0.00015022258165698667, "loss": 0.7301, "step": 6215 }, { "epoch": 0.35, "learning_rate": 0.00015020669396820314, "loss": 0.6658, "step": 6216 }, { "epoch": 0.35, "learning_rate": 0.00015019080458481202, "loss": 0.6783, "step": 6217 }, { "epoch": 0.35, "learning_rate": 0.00015017491350734967, "loss": 0.6234, "step": 6218 }, { "epoch": 0.35, "learning_rate": 0.00015015902073635236, "loss": 0.6378, "step": 6219 }, { "epoch": 0.35, "learning_rate": 0.00015014312627235655, "loss": 0.7596, "step": 6220 }, { "epoch": 0.35, "learning_rate": 0.00015012723011589873, "loss": 0.7175, "step": 6221 }, { "epoch": 0.35, "learning_rate": 0.00015011133226751545, "loss": 0.7171, "step": 6222 }, { "epoch": 0.35, "learning_rate": 0.00015009543272774325, "loss": 0.7744, "step": 6223 }, { "epoch": 0.35, "learning_rate": 0.00015007953149711884, "loss": 0.7034, "step": 6224 }, { "epoch": 0.35, "learning_rate": 0.0001500636285761789, "loss": 0.6796, "step": 6225 }, { "epoch": 0.35, "learning_rate": 0.0001500477239654602, "loss": 0.671, "step": 6226 }, { "epoch": 0.35, "learning_rate": 0.00015003181766549953, "loss": 0.7124, "step": 6227 }, { "epoch": 0.35, "learning_rate": 0.00015001590967683384, "loss": 0.7132, "step": 6228 }, { "epoch": 0.35, "learning_rate": 0.00015000000000000001, "loss": 0.8065, "step": 6229 }, { "epoch": 0.35, "learning_rate": 0.00014998408863553503, "loss": 0.7243, "step": 6230 }, { "epoch": 0.35, "learning_rate": 0.000149968175583976, "loss": 0.8071, "step": 6231 }, { "epoch": 0.35, "learning_rate": 0.00014995226084586, "loss": 0.7676, "step": 6232 }, { "epoch": 0.35, "learning_rate": 0.0001499363444217242, "loss": 0.6616, "step": 6233 }, { "epoch": 0.35, "learning_rate": 0.0001499204263121058, "loss": 0.7499, "step": 6234 }, { "epoch": 0.35, "learning_rate": 0.00014990450651754207, "loss": 0.7693, "step": 6235 }, { "epoch": 0.35, "learning_rate": 0.00014988858503857036, "loss": 0.8487, "step": 6236 }, { "epoch": 0.35, "learning_rate": 0.0001498726618757281, "loss": 0.7993, "step": 6237 }, { "epoch": 0.35, "learning_rate": 0.00014985673702955273, "loss": 0.686, "step": 6238 }, { "epoch": 0.35, "learning_rate": 0.00014984081050058165, "loss": 0.7576, "step": 6239 }, { "epoch": 0.35, "learning_rate": 0.00014982488228935256, "loss": 0.7338, "step": 6240 }, { "epoch": 0.35, "learning_rate": 0.00014980895239640303, "loss": 0.6297, "step": 6241 }, { "epoch": 0.35, "learning_rate": 0.0001497930208222707, "loss": 0.7222, "step": 6242 }, { "epoch": 0.35, "learning_rate": 0.00014977708756749332, "loss": 0.751, "step": 6243 }, { "epoch": 0.35, "learning_rate": 0.00014976115263260875, "loss": 0.8042, "step": 6244 }, { "epoch": 0.35, "learning_rate": 0.0001497452160181547, "loss": 0.7161, "step": 6245 }, { "epoch": 0.35, "learning_rate": 0.0001497292777246692, "loss": 0.7964, "step": 6246 }, { "epoch": 0.35, "learning_rate": 0.00014971333775269012, "loss": 0.7659, "step": 6247 }, { "epoch": 0.35, "learning_rate": 0.00014969739610275556, "loss": 0.6556, "step": 6248 }, { "epoch": 0.35, "learning_rate": 0.0001496814527754035, "loss": 0.7012, "step": 6249 }, { "epoch": 0.35, "learning_rate": 0.00014966550777117214, "loss": 0.7217, "step": 6250 }, { "epoch": 0.35, "learning_rate": 0.00014964956109059964, "loss": 0.697, "step": 6251 }, { "epoch": 0.35, "learning_rate": 0.00014963361273422423, "loss": 0.6414, "step": 6252 }, { "epoch": 0.35, "learning_rate": 0.00014961766270258422, "loss": 0.7614, "step": 6253 }, { "epoch": 0.35, "learning_rate": 0.00014960171099621795, "loss": 0.7674, "step": 6254 }, { "epoch": 0.35, "learning_rate": 0.0001495857576156639, "loss": 0.7236, "step": 6255 }, { "epoch": 0.35, "learning_rate": 0.00014956980256146044, "loss": 0.8052, "step": 6256 }, { "epoch": 0.35, "learning_rate": 0.00014955384583414616, "loss": 0.7253, "step": 6257 }, { "epoch": 0.35, "learning_rate": 0.00014953788743425964, "loss": 0.7851, "step": 6258 }, { "epoch": 0.36, "learning_rate": 0.0001495219273623395, "loss": 0.7221, "step": 6259 }, { "epoch": 0.36, "learning_rate": 0.0001495059656189244, "loss": 0.7058, "step": 6260 }, { "epoch": 0.36, "learning_rate": 0.00014949000220455314, "loss": 0.6935, "step": 6261 }, { "epoch": 0.36, "learning_rate": 0.00014947403711976454, "loss": 0.7447, "step": 6262 }, { "epoch": 0.36, "learning_rate": 0.0001494580703650974, "loss": 0.6704, "step": 6263 }, { "epoch": 0.36, "learning_rate": 0.00014944210194109067, "loss": 0.6195, "step": 6264 }, { "epoch": 0.36, "learning_rate": 0.00014942613184828335, "loss": 0.7383, "step": 6265 }, { "epoch": 0.36, "learning_rate": 0.00014941016008721447, "loss": 0.6692, "step": 6266 }, { "epoch": 0.36, "learning_rate": 0.0001493941866584231, "loss": 0.7062, "step": 6267 }, { "epoch": 0.36, "learning_rate": 0.00014937821156244837, "loss": 0.7169, "step": 6268 }, { "epoch": 0.36, "learning_rate": 0.00014936223479982953, "loss": 0.7883, "step": 6269 }, { "epoch": 0.36, "learning_rate": 0.0001493462563711058, "loss": 0.7451, "step": 6270 }, { "epoch": 0.36, "learning_rate": 0.0001493302762768165, "loss": 0.7765, "step": 6271 }, { "epoch": 0.36, "learning_rate": 0.000149314294517501, "loss": 0.7326, "step": 6272 }, { "epoch": 0.36, "learning_rate": 0.00014929831109369873, "loss": 0.5429, "step": 6273 }, { "epoch": 0.36, "learning_rate": 0.00014928232600594916, "loss": 0.6988, "step": 6274 }, { "epoch": 0.36, "learning_rate": 0.00014926633925479182, "loss": 0.7314, "step": 6275 }, { "epoch": 0.36, "learning_rate": 0.00014925035084076637, "loss": 0.8123, "step": 6276 }, { "epoch": 0.36, "learning_rate": 0.00014923436076441242, "loss": 0.6448, "step": 6277 }, { "epoch": 0.36, "learning_rate": 0.00014921836902626962, "loss": 0.676, "step": 6278 }, { "epoch": 0.36, "learning_rate": 0.00014920237562687785, "loss": 0.6634, "step": 6279 }, { "epoch": 0.36, "learning_rate": 0.0001491863805667768, "loss": 0.6592, "step": 6280 }, { "epoch": 0.36, "learning_rate": 0.00014917038384650646, "loss": 0.78, "step": 6281 }, { "epoch": 0.36, "learning_rate": 0.00014915438546660666, "loss": 0.713, "step": 6282 }, { "epoch": 0.36, "learning_rate": 0.0001491383854276175, "loss": 0.6811, "step": 6283 }, { "epoch": 0.36, "learning_rate": 0.00014912238373007892, "loss": 0.7371, "step": 6284 }, { "epoch": 0.36, "learning_rate": 0.00014910638037453106, "loss": 0.7019, "step": 6285 }, { "epoch": 0.36, "learning_rate": 0.00014909037536151409, "loss": 0.7058, "step": 6286 }, { "epoch": 0.36, "learning_rate": 0.00014907436869156818, "loss": 0.7506, "step": 6287 }, { "epoch": 0.36, "learning_rate": 0.00014905836036523367, "loss": 0.7079, "step": 6288 }, { "epoch": 0.36, "learning_rate": 0.00014904235038305083, "loss": 0.6564, "step": 6289 }, { "epoch": 0.36, "learning_rate": 0.00014902633874556002, "loss": 0.6764, "step": 6290 }, { "epoch": 0.36, "learning_rate": 0.00014901032545330172, "loss": 0.6589, "step": 6291 }, { "epoch": 0.36, "learning_rate": 0.0001489943105068164, "loss": 0.7319, "step": 6292 }, { "epoch": 0.36, "learning_rate": 0.0001489782939066446, "loss": 0.7664, "step": 6293 }, { "epoch": 0.36, "learning_rate": 0.00014896227565332692, "loss": 0.7479, "step": 6294 }, { "epoch": 0.36, "learning_rate": 0.00014894625574740406, "loss": 0.8076, "step": 6295 }, { "epoch": 0.36, "learning_rate": 0.0001489302341894167, "loss": 0.6559, "step": 6296 }, { "epoch": 0.36, "learning_rate": 0.0001489142109799056, "loss": 0.703, "step": 6297 }, { "epoch": 0.36, "learning_rate": 0.00014889818611941156, "loss": 0.6428, "step": 6298 }, { "epoch": 0.36, "learning_rate": 0.00014888215960847555, "loss": 0.6813, "step": 6299 }, { "epoch": 0.36, "learning_rate": 0.00014886613144763844, "loss": 0.8103, "step": 6300 }, { "epoch": 0.36, "learning_rate": 0.0001488501016374412, "loss": 0.6411, "step": 6301 }, { "epoch": 0.36, "learning_rate": 0.00014883407017842498, "loss": 0.7005, "step": 6302 }, { "epoch": 0.36, "learning_rate": 0.0001488180370711308, "loss": 0.719, "step": 6303 }, { "epoch": 0.36, "learning_rate": 0.00014880200231609983, "loss": 0.7098, "step": 6304 }, { "epoch": 0.36, "learning_rate": 0.0001487859659138733, "loss": 0.7847, "step": 6305 }, { "epoch": 0.36, "learning_rate": 0.00014876992786499245, "loss": 0.7852, "step": 6306 }, { "epoch": 0.36, "learning_rate": 0.00014875388816999865, "loss": 0.6861, "step": 6307 }, { "epoch": 0.36, "learning_rate": 0.00014873784682943326, "loss": 0.7372, "step": 6308 }, { "epoch": 0.36, "learning_rate": 0.00014872180384383773, "loss": 0.7315, "step": 6309 }, { "epoch": 0.36, "learning_rate": 0.0001487057592137535, "loss": 0.7259, "step": 6310 }, { "epoch": 0.36, "learning_rate": 0.0001486897129397222, "loss": 0.7746, "step": 6311 }, { "epoch": 0.36, "learning_rate": 0.00014867366502228539, "loss": 0.7619, "step": 6312 }, { "epoch": 0.36, "learning_rate": 0.00014865761546198474, "loss": 0.7502, "step": 6313 }, { "epoch": 0.36, "learning_rate": 0.00014864156425936195, "loss": 0.8404, "step": 6314 }, { "epoch": 0.36, "learning_rate": 0.00014862551141495882, "loss": 0.6295, "step": 6315 }, { "epoch": 0.36, "learning_rate": 0.00014860945692931712, "loss": 0.7742, "step": 6316 }, { "epoch": 0.36, "learning_rate": 0.00014859340080297876, "loss": 0.7459, "step": 6317 }, { "epoch": 0.36, "learning_rate": 0.00014857734303648574, "loss": 0.7913, "step": 6318 }, { "epoch": 0.36, "learning_rate": 0.00014856128363037994, "loss": 0.726, "step": 6319 }, { "epoch": 0.36, "learning_rate": 0.00014854522258520348, "loss": 0.7739, "step": 6320 }, { "epoch": 0.36, "learning_rate": 0.00014852915990149846, "loss": 0.741, "step": 6321 }, { "epoch": 0.36, "learning_rate": 0.000148513095579807, "loss": 0.6759, "step": 6322 }, { "epoch": 0.36, "learning_rate": 0.00014849702962067132, "loss": 0.8033, "step": 6323 }, { "epoch": 0.36, "learning_rate": 0.00014848096202463372, "loss": 0.7272, "step": 6324 }, { "epoch": 0.36, "learning_rate": 0.00014846489279223652, "loss": 0.7314, "step": 6325 }, { "epoch": 0.36, "learning_rate": 0.00014844882192402206, "loss": 0.5987, "step": 6326 }, { "epoch": 0.36, "learning_rate": 0.00014843274942053278, "loss": 0.7346, "step": 6327 }, { "epoch": 0.36, "learning_rate": 0.00014841667528231123, "loss": 0.691, "step": 6328 }, { "epoch": 0.36, "learning_rate": 0.0001484005995098999, "loss": 0.7764, "step": 6329 }, { "epoch": 0.36, "learning_rate": 0.00014838452210384138, "loss": 0.7048, "step": 6330 }, { "epoch": 0.36, "learning_rate": 0.00014836844306467838, "loss": 0.6804, "step": 6331 }, { "epoch": 0.36, "learning_rate": 0.00014835236239295354, "loss": 0.7052, "step": 6332 }, { "epoch": 0.36, "learning_rate": 0.0001483362800892097, "loss": 0.7507, "step": 6333 }, { "epoch": 0.36, "learning_rate": 0.00014832019615398963, "loss": 0.7707, "step": 6334 }, { "epoch": 0.36, "learning_rate": 0.00014830411058783618, "loss": 0.5952, "step": 6335 }, { "epoch": 0.36, "learning_rate": 0.00014828802339129235, "loss": 0.7462, "step": 6336 }, { "epoch": 0.36, "learning_rate": 0.00014827193456490108, "loss": 0.7008, "step": 6337 }, { "epoch": 0.36, "learning_rate": 0.0001482558441092054, "loss": 0.7412, "step": 6338 }, { "epoch": 0.36, "learning_rate": 0.00014823975202474845, "loss": 0.7352, "step": 6339 }, { "epoch": 0.36, "learning_rate": 0.00014822365831207337, "loss": 0.6262, "step": 6340 }, { "epoch": 0.36, "learning_rate": 0.00014820756297172333, "loss": 0.6766, "step": 6341 }, { "epoch": 0.36, "learning_rate": 0.00014819146600424162, "loss": 0.6994, "step": 6342 }, { "epoch": 0.36, "learning_rate": 0.00014817536741017152, "loss": 0.7404, "step": 6343 }, { "epoch": 0.36, "learning_rate": 0.00014815926719005649, "loss": 0.7372, "step": 6344 }, { "epoch": 0.36, "learning_rate": 0.00014814316534443982, "loss": 0.7004, "step": 6345 }, { "epoch": 0.36, "learning_rate": 0.00014812706187386507, "loss": 0.7157, "step": 6346 }, { "epoch": 0.36, "learning_rate": 0.0001481109567788758, "loss": 0.7415, "step": 6347 }, { "epoch": 0.36, "learning_rate": 0.00014809485006001552, "loss": 0.7285, "step": 6348 }, { "epoch": 0.36, "learning_rate": 0.00014807874171782795, "loss": 0.7209, "step": 6349 }, { "epoch": 0.36, "learning_rate": 0.00014806263175285675, "loss": 0.73, "step": 6350 }, { "epoch": 0.36, "learning_rate": 0.00014804652016564568, "loss": 0.7762, "step": 6351 }, { "epoch": 0.36, "learning_rate": 0.00014803040695673852, "loss": 0.6849, "step": 6352 }, { "epoch": 0.36, "learning_rate": 0.00014801429212667918, "loss": 0.5558, "step": 6353 }, { "epoch": 0.36, "learning_rate": 0.00014799817567601157, "loss": 0.6913, "step": 6354 }, { "epoch": 0.36, "learning_rate": 0.00014798205760527966, "loss": 0.7806, "step": 6355 }, { "epoch": 0.36, "learning_rate": 0.00014796593791502745, "loss": 0.8594, "step": 6356 }, { "epoch": 0.36, "learning_rate": 0.00014794981660579904, "loss": 0.6734, "step": 6357 }, { "epoch": 0.36, "learning_rate": 0.00014793369367813855, "loss": 0.7169, "step": 6358 }, { "epoch": 0.36, "learning_rate": 0.00014791756913259022, "loss": 0.6663, "step": 6359 }, { "epoch": 0.36, "learning_rate": 0.00014790144296969825, "loss": 0.7956, "step": 6360 }, { "epoch": 0.36, "learning_rate": 0.00014788531519000696, "loss": 0.7357, "step": 6361 }, { "epoch": 0.36, "learning_rate": 0.0001478691857940607, "loss": 0.7593, "step": 6362 }, { "epoch": 0.36, "learning_rate": 0.00014785305478240388, "loss": 0.7307, "step": 6363 }, { "epoch": 0.36, "learning_rate": 0.00014783692215558094, "loss": 0.7528, "step": 6364 }, { "epoch": 0.36, "learning_rate": 0.0001478207879141364, "loss": 0.6747, "step": 6365 }, { "epoch": 0.36, "learning_rate": 0.0001478046520586149, "loss": 0.7925, "step": 6366 }, { "epoch": 0.36, "learning_rate": 0.00014778851458956103, "loss": 0.6016, "step": 6367 }, { "epoch": 0.36, "learning_rate": 0.0001477723755075194, "loss": 0.8019, "step": 6368 }, { "epoch": 0.36, "learning_rate": 0.00014775623481303487, "loss": 0.7672, "step": 6369 }, { "epoch": 0.36, "learning_rate": 0.00014774009250665215, "loss": 0.7462, "step": 6370 }, { "epoch": 0.36, "learning_rate": 0.0001477239485889161, "loss": 0.6291, "step": 6371 }, { "epoch": 0.36, "learning_rate": 0.0001477078030603716, "loss": 0.7672, "step": 6372 }, { "epoch": 0.36, "learning_rate": 0.00014769165592156368, "loss": 0.6774, "step": 6373 }, { "epoch": 0.36, "learning_rate": 0.00014767550717303726, "loss": 0.6558, "step": 6374 }, { "epoch": 0.36, "learning_rate": 0.00014765935681533744, "loss": 0.7691, "step": 6375 }, { "epoch": 0.36, "learning_rate": 0.00014764320484900937, "loss": 0.7264, "step": 6376 }, { "epoch": 0.36, "learning_rate": 0.00014762705127459818, "loss": 0.7374, "step": 6377 }, { "epoch": 0.36, "learning_rate": 0.00014761089609264906, "loss": 0.7905, "step": 6378 }, { "epoch": 0.36, "learning_rate": 0.00014759473930370736, "loss": 0.7377, "step": 6379 }, { "epoch": 0.36, "learning_rate": 0.00014757858090831837, "loss": 0.6957, "step": 6380 }, { "epoch": 0.36, "learning_rate": 0.00014756242090702756, "loss": 0.8003, "step": 6381 }, { "epoch": 0.36, "learning_rate": 0.0001475462593003802, "loss": 0.725, "step": 6382 }, { "epoch": 0.36, "learning_rate": 0.000147530096088922, "loss": 0.7737, "step": 6383 }, { "epoch": 0.36, "learning_rate": 0.00014751393127319834, "loss": 0.8136, "step": 6384 }, { "epoch": 0.36, "learning_rate": 0.00014749776485375492, "loss": 0.6699, "step": 6385 }, { "epoch": 0.36, "learning_rate": 0.00014748159683113736, "loss": 0.7433, "step": 6386 }, { "epoch": 0.36, "learning_rate": 0.00014746542720589135, "loss": 0.7312, "step": 6387 }, { "epoch": 0.36, "learning_rate": 0.00014744925597856275, "loss": 0.8095, "step": 6388 }, { "epoch": 0.36, "learning_rate": 0.0001474330831496973, "loss": 0.6991, "step": 6389 }, { "epoch": 0.36, "learning_rate": 0.00014741690871984084, "loss": 0.7553, "step": 6390 }, { "epoch": 0.36, "learning_rate": 0.0001474007326895394, "loss": 0.8165, "step": 6391 }, { "epoch": 0.36, "learning_rate": 0.00014738455505933893, "loss": 0.7413, "step": 6392 }, { "epoch": 0.36, "learning_rate": 0.00014736837582978544, "loss": 0.6895, "step": 6393 }, { "epoch": 0.36, "learning_rate": 0.00014735219500142504, "loss": 0.7852, "step": 6394 }, { "epoch": 0.36, "learning_rate": 0.00014733601257480388, "loss": 0.6598, "step": 6395 }, { "epoch": 0.36, "learning_rate": 0.00014731982855046815, "loss": 0.6926, "step": 6396 }, { "epoch": 0.36, "learning_rate": 0.0001473036429289641, "loss": 0.7157, "step": 6397 }, { "epoch": 0.36, "learning_rate": 0.00014728745571083803, "loss": 0.7726, "step": 6398 }, { "epoch": 0.36, "learning_rate": 0.00014727126689663636, "loss": 0.7251, "step": 6399 }, { "epoch": 0.36, "learning_rate": 0.00014725507648690543, "loss": 0.7833, "step": 6400 }, { "epoch": 0.36, "learning_rate": 0.0001472388844821917, "loss": 0.7307, "step": 6401 }, { "epoch": 0.36, "learning_rate": 0.00014722269088304177, "loss": 0.6247, "step": 6402 }, { "epoch": 0.36, "learning_rate": 0.0001472064956900022, "loss": 0.6934, "step": 6403 }, { "epoch": 0.36, "learning_rate": 0.00014719029890361955, "loss": 0.7697, "step": 6404 }, { "epoch": 0.36, "learning_rate": 0.00014717410052444058, "loss": 0.7247, "step": 6405 }, { "epoch": 0.36, "learning_rate": 0.00014715790055301196, "loss": 0.6944, "step": 6406 }, { "epoch": 0.36, "learning_rate": 0.00014714169898988056, "loss": 0.7142, "step": 6407 }, { "epoch": 0.36, "learning_rate": 0.00014712549583559322, "loss": 0.7223, "step": 6408 }, { "epoch": 0.36, "learning_rate": 0.00014710929109069674, "loss": 0.806, "step": 6409 }, { "epoch": 0.36, "learning_rate": 0.00014709308475573815, "loss": 0.7751, "step": 6410 }, { "epoch": 0.36, "learning_rate": 0.00014707687683126444, "loss": 0.7499, "step": 6411 }, { "epoch": 0.36, "learning_rate": 0.00014706066731782268, "loss": 0.8189, "step": 6412 }, { "epoch": 0.36, "learning_rate": 0.00014704445621595996, "loss": 0.812, "step": 6413 }, { "epoch": 0.36, "learning_rate": 0.0001470282435262235, "loss": 0.7146, "step": 6414 }, { "epoch": 0.36, "learning_rate": 0.0001470120292491605, "loss": 0.6522, "step": 6415 }, { "epoch": 0.36, "learning_rate": 0.00014699581338531816, "loss": 0.7528, "step": 6416 }, { "epoch": 0.36, "learning_rate": 0.00014697959593524388, "loss": 0.6069, "step": 6417 }, { "epoch": 0.36, "learning_rate": 0.00014696337689948507, "loss": 0.6623, "step": 6418 }, { "epoch": 0.36, "learning_rate": 0.00014694715627858908, "loss": 0.7481, "step": 6419 }, { "epoch": 0.36, "learning_rate": 0.00014693093407310346, "loss": 0.7257, "step": 6420 }, { "epoch": 0.36, "learning_rate": 0.00014691471028357574, "loss": 0.6756, "step": 6421 }, { "epoch": 0.36, "learning_rate": 0.0001468984849105535, "loss": 0.6439, "step": 6422 }, { "epoch": 0.36, "learning_rate": 0.0001468822579545844, "loss": 0.8178, "step": 6423 }, { "epoch": 0.36, "learning_rate": 0.00014686602941621615, "loss": 0.6615, "step": 6424 }, { "epoch": 0.36, "learning_rate": 0.0001468497992959965, "loss": 0.6492, "step": 6425 }, { "epoch": 0.36, "learning_rate": 0.00014683356759447323, "loss": 0.8258, "step": 6426 }, { "epoch": 0.36, "learning_rate": 0.00014681733431219423, "loss": 0.7031, "step": 6427 }, { "epoch": 0.36, "learning_rate": 0.00014680109944970743, "loss": 0.7916, "step": 6428 }, { "epoch": 0.36, "learning_rate": 0.0001467848630075608, "loss": 0.7567, "step": 6429 }, { "epoch": 0.36, "learning_rate": 0.0001467686249863023, "loss": 0.636, "step": 6430 }, { "epoch": 0.36, "learning_rate": 0.0001467523853864801, "loss": 0.713, "step": 6431 }, { "epoch": 0.36, "learning_rate": 0.00014673614420864225, "loss": 0.7818, "step": 6432 }, { "epoch": 0.36, "learning_rate": 0.00014671990145333696, "loss": 0.685, "step": 6433 }, { "epoch": 0.36, "learning_rate": 0.0001467036571211125, "loss": 0.6334, "step": 6434 }, { "epoch": 0.37, "learning_rate": 0.0001466874112125171, "loss": 0.7714, "step": 6435 }, { "epoch": 0.37, "learning_rate": 0.00014667116372809916, "loss": 0.8404, "step": 6436 }, { "epoch": 0.37, "learning_rate": 0.00014665491466840698, "loss": 0.7292, "step": 6437 }, { "epoch": 0.37, "learning_rate": 0.00014663866403398913, "loss": 0.6419, "step": 6438 }, { "epoch": 0.37, "learning_rate": 0.00014662241182539402, "loss": 0.7593, "step": 6439 }, { "epoch": 0.37, "learning_rate": 0.00014660615804317026, "loss": 0.6336, "step": 6440 }, { "epoch": 0.37, "learning_rate": 0.00014658990268786646, "loss": 0.6234, "step": 6441 }, { "epoch": 0.37, "learning_rate": 0.00014657364576003122, "loss": 0.7284, "step": 6442 }, { "epoch": 0.37, "learning_rate": 0.00014655738726021328, "loss": 0.6596, "step": 6443 }, { "epoch": 0.37, "learning_rate": 0.0001465411271889615, "loss": 0.711, "step": 6444 }, { "epoch": 0.37, "learning_rate": 0.00014652486554682454, "loss": 0.793, "step": 6445 }, { "epoch": 0.37, "learning_rate": 0.00014650860233435134, "loss": 0.648, "step": 6446 }, { "epoch": 0.37, "learning_rate": 0.0001464923375520909, "loss": 0.7119, "step": 6447 }, { "epoch": 0.37, "learning_rate": 0.0001464760712005921, "loss": 0.7439, "step": 6448 }, { "epoch": 0.37, "learning_rate": 0.00014645980328040401, "loss": 0.738, "step": 6449 }, { "epoch": 0.37, "learning_rate": 0.00014644353379207573, "loss": 0.8143, "step": 6450 }, { "epoch": 0.37, "learning_rate": 0.0001464272627361564, "loss": 0.7633, "step": 6451 }, { "epoch": 0.37, "learning_rate": 0.00014641099011319516, "loss": 0.6812, "step": 6452 }, { "epoch": 0.37, "learning_rate": 0.00014639471592374126, "loss": 0.6424, "step": 6453 }, { "epoch": 0.37, "learning_rate": 0.00014637844016834406, "loss": 0.7564, "step": 6454 }, { "epoch": 0.37, "learning_rate": 0.00014636216284755289, "loss": 0.7359, "step": 6455 }, { "epoch": 0.37, "learning_rate": 0.00014634588396191707, "loss": 0.7943, "step": 6456 }, { "epoch": 0.37, "learning_rate": 0.00014632960351198618, "loss": 0.7691, "step": 6457 }, { "epoch": 0.37, "learning_rate": 0.00014631332149830963, "loss": 0.7468, "step": 6458 }, { "epoch": 0.37, "learning_rate": 0.00014629703792143705, "loss": 0.7052, "step": 6459 }, { "epoch": 0.37, "learning_rate": 0.000146280752781918, "loss": 0.6958, "step": 6460 }, { "epoch": 0.37, "learning_rate": 0.0001462644660803022, "loss": 0.7896, "step": 6461 }, { "epoch": 0.37, "learning_rate": 0.00014624817781713932, "loss": 0.69, "step": 6462 }, { "epoch": 0.37, "learning_rate": 0.00014623188799297915, "loss": 0.7711, "step": 6463 }, { "epoch": 0.37, "learning_rate": 0.0001462155966083715, "loss": 0.7558, "step": 6464 }, { "epoch": 0.37, "learning_rate": 0.00014619930366386626, "loss": 0.743, "step": 6465 }, { "epoch": 0.37, "learning_rate": 0.00014618300916001336, "loss": 0.7193, "step": 6466 }, { "epoch": 0.37, "learning_rate": 0.00014616671309736277, "loss": 0.7317, "step": 6467 }, { "epoch": 0.37, "learning_rate": 0.00014615041547646454, "loss": 0.7054, "step": 6468 }, { "epoch": 0.37, "learning_rate": 0.0001461341162978688, "loss": 0.7458, "step": 6469 }, { "epoch": 0.37, "learning_rate": 0.00014611781556212562, "loss": 0.7269, "step": 6470 }, { "epoch": 0.37, "learning_rate": 0.0001461015132697852, "loss": 0.704, "step": 6471 }, { "epoch": 0.37, "learning_rate": 0.0001460852094213978, "loss": 0.6382, "step": 6472 }, { "epoch": 0.37, "learning_rate": 0.00014606890401751375, "loss": 0.6827, "step": 6473 }, { "epoch": 0.37, "learning_rate": 0.00014605259705868334, "loss": 0.6269, "step": 6474 }, { "epoch": 0.37, "learning_rate": 0.00014603628854545697, "loss": 0.6532, "step": 6475 }, { "epoch": 0.37, "learning_rate": 0.00014601997847838518, "loss": 0.5683, "step": 6476 }, { "epoch": 0.37, "learning_rate": 0.0001460036668580184, "loss": 0.7489, "step": 6477 }, { "epoch": 0.37, "learning_rate": 0.00014598735368490725, "loss": 0.8147, "step": 6478 }, { "epoch": 0.37, "learning_rate": 0.00014597103895960226, "loss": 0.7811, "step": 6479 }, { "epoch": 0.37, "learning_rate": 0.00014595472268265415, "loss": 0.7261, "step": 6480 }, { "epoch": 0.37, "learning_rate": 0.00014593840485461364, "loss": 0.7734, "step": 6481 }, { "epoch": 0.37, "learning_rate": 0.0001459220854760315, "loss": 0.6881, "step": 6482 }, { "epoch": 0.37, "learning_rate": 0.0001459057645474585, "loss": 0.7792, "step": 6483 }, { "epoch": 0.37, "learning_rate": 0.00014588944206944562, "loss": 0.7495, "step": 6484 }, { "epoch": 0.37, "learning_rate": 0.00014587311804254365, "loss": 0.7777, "step": 6485 }, { "epoch": 0.37, "learning_rate": 0.0001458567924673037, "loss": 0.756, "step": 6486 }, { "epoch": 0.37, "learning_rate": 0.0001458404653442767, "loss": 0.6616, "step": 6487 }, { "epoch": 0.37, "learning_rate": 0.0001458241366740138, "loss": 0.7168, "step": 6488 }, { "epoch": 0.37, "learning_rate": 0.0001458078064570661, "loss": 0.706, "step": 6489 }, { "epoch": 0.37, "learning_rate": 0.0001457914746939848, "loss": 0.6167, "step": 6490 }, { "epoch": 0.37, "learning_rate": 0.0001457751413853211, "loss": 0.7299, "step": 6491 }, { "epoch": 0.37, "learning_rate": 0.00014575880653162638, "loss": 0.8055, "step": 6492 }, { "epoch": 0.37, "learning_rate": 0.0001457424701334519, "loss": 0.6847, "step": 6493 }, { "epoch": 0.37, "learning_rate": 0.0001457261321913491, "loss": 0.6557, "step": 6494 }, { "epoch": 0.37, "learning_rate": 0.00014570979270586945, "loss": 0.7398, "step": 6495 }, { "epoch": 0.37, "learning_rate": 0.00014569345167756437, "loss": 0.7762, "step": 6496 }, { "epoch": 0.37, "learning_rate": 0.00014567710910698547, "loss": 0.7083, "step": 6497 }, { "epoch": 0.37, "learning_rate": 0.00014566076499468434, "loss": 0.7988, "step": 6498 }, { "epoch": 0.37, "learning_rate": 0.00014564441934121266, "loss": 0.6819, "step": 6499 }, { "epoch": 0.37, "learning_rate": 0.00014562807214712212, "loss": 0.6553, "step": 6500 }, { "epoch": 0.37, "learning_rate": 0.00014561172341296446, "loss": 0.7537, "step": 6501 }, { "epoch": 0.37, "learning_rate": 0.00014559537313929154, "loss": 0.7837, "step": 6502 }, { "epoch": 0.37, "learning_rate": 0.0001455790213266552, "loss": 0.8, "step": 6503 }, { "epoch": 0.37, "learning_rate": 0.00014556266797560732, "loss": 0.7804, "step": 6504 }, { "epoch": 0.37, "learning_rate": 0.00014554631308669994, "loss": 0.6887, "step": 6505 }, { "epoch": 0.37, "learning_rate": 0.00014552995666048502, "loss": 0.7043, "step": 6506 }, { "epoch": 0.37, "learning_rate": 0.00014551359869751466, "loss": 0.6034, "step": 6507 }, { "epoch": 0.37, "learning_rate": 0.000145497239198341, "loss": 0.7224, "step": 6508 }, { "epoch": 0.37, "learning_rate": 0.00014548087816351616, "loss": 0.6673, "step": 6509 }, { "epoch": 0.37, "learning_rate": 0.00014546451559359244, "loss": 0.7286, "step": 6510 }, { "epoch": 0.37, "learning_rate": 0.00014544815148912206, "loss": 0.7322, "step": 6511 }, { "epoch": 0.37, "learning_rate": 0.00014543178585065737, "loss": 0.6907, "step": 6512 }, { "epoch": 0.37, "learning_rate": 0.00014541541867875077, "loss": 0.6751, "step": 6513 }, { "epoch": 0.37, "learning_rate": 0.00014539904997395468, "loss": 0.745, "step": 6514 }, { "epoch": 0.37, "learning_rate": 0.00014538267973682162, "loss": 0.646, "step": 6515 }, { "epoch": 0.37, "learning_rate": 0.00014536630796790404, "loss": 0.6314, "step": 6516 }, { "epoch": 0.37, "learning_rate": 0.00014534993466775462, "loss": 0.7197, "step": 6517 }, { "epoch": 0.37, "learning_rate": 0.00014533355983692602, "loss": 0.6422, "step": 6518 }, { "epoch": 0.37, "learning_rate": 0.0001453171834759708, "loss": 0.7539, "step": 6519 }, { "epoch": 0.37, "learning_rate": 0.00014530080558544181, "loss": 0.7862, "step": 6520 }, { "epoch": 0.37, "learning_rate": 0.00014528442616589186, "loss": 0.696, "step": 6521 }, { "epoch": 0.37, "learning_rate": 0.00014526804521787376, "loss": 0.7242, "step": 6522 }, { "epoch": 0.37, "learning_rate": 0.00014525166274194037, "loss": 0.7701, "step": 6523 }, { "epoch": 0.37, "learning_rate": 0.00014523527873864473, "loss": 0.7793, "step": 6524 }, { "epoch": 0.37, "learning_rate": 0.00014521889320853978, "loss": 0.7971, "step": 6525 }, { "epoch": 0.37, "learning_rate": 0.00014520250615217858, "loss": 0.669, "step": 6526 }, { "epoch": 0.37, "learning_rate": 0.00014518611757011426, "loss": 0.7051, "step": 6527 }, { "epoch": 0.37, "learning_rate": 0.00014516972746289998, "loss": 0.7366, "step": 6528 }, { "epoch": 0.37, "learning_rate": 0.00014515333583108896, "loss": 0.7539, "step": 6529 }, { "epoch": 0.37, "learning_rate": 0.0001451369426752344, "loss": 0.7799, "step": 6530 }, { "epoch": 0.37, "learning_rate": 0.00014512054799588965, "loss": 0.7606, "step": 6531 }, { "epoch": 0.37, "learning_rate": 0.00014510415179360805, "loss": 0.6783, "step": 6532 }, { "epoch": 0.37, "learning_rate": 0.00014508775406894307, "loss": 0.8564, "step": 6533 }, { "epoch": 0.37, "learning_rate": 0.00014507135482244815, "loss": 0.7533, "step": 6534 }, { "epoch": 0.37, "learning_rate": 0.0001450549540546768, "loss": 0.7407, "step": 6535 }, { "epoch": 0.37, "learning_rate": 0.00014503855176618258, "loss": 0.7283, "step": 6536 }, { "epoch": 0.37, "learning_rate": 0.0001450221479575191, "loss": 0.697, "step": 6537 }, { "epoch": 0.37, "learning_rate": 0.00014500574262924008, "loss": 0.6445, "step": 6538 }, { "epoch": 0.37, "learning_rate": 0.00014498933578189918, "loss": 0.7516, "step": 6539 }, { "epoch": 0.37, "learning_rate": 0.00014497292741605025, "loss": 0.7638, "step": 6540 }, { "epoch": 0.37, "learning_rate": 0.00014495651753224705, "loss": 0.6972, "step": 6541 }, { "epoch": 0.37, "learning_rate": 0.0001449401061310435, "loss": 0.813, "step": 6542 }, { "epoch": 0.37, "learning_rate": 0.00014492369321299347, "loss": 0.7894, "step": 6543 }, { "epoch": 0.37, "learning_rate": 0.00014490727877865102, "loss": 0.7077, "step": 6544 }, { "epoch": 0.37, "learning_rate": 0.00014489086282857012, "loss": 0.7167, "step": 6545 }, { "epoch": 0.37, "learning_rate": 0.00014487444536330485, "loss": 0.5752, "step": 6546 }, { "epoch": 0.37, "learning_rate": 0.00014485802638340942, "loss": 0.7373, "step": 6547 }, { "epoch": 0.37, "learning_rate": 0.0001448416058894379, "loss": 0.6234, "step": 6548 }, { "epoch": 0.37, "learning_rate": 0.00014482518388194457, "loss": 0.7259, "step": 6549 }, { "epoch": 0.37, "learning_rate": 0.00014480876036148378, "loss": 0.7763, "step": 6550 }, { "epoch": 0.37, "learning_rate": 0.00014479233532860977, "loss": 0.7574, "step": 6551 }, { "epoch": 0.37, "learning_rate": 0.00014477590878387696, "loss": 0.8109, "step": 6552 }, { "epoch": 0.37, "learning_rate": 0.00014475948072783983, "loss": 0.6482, "step": 6553 }, { "epoch": 0.37, "learning_rate": 0.00014474305116105284, "loss": 0.7733, "step": 6554 }, { "epoch": 0.37, "learning_rate": 0.00014472662008407054, "loss": 0.6648, "step": 6555 }, { "epoch": 0.37, "learning_rate": 0.00014471018749744746, "loss": 0.8424, "step": 6556 }, { "epoch": 0.37, "learning_rate": 0.00014469375340173834, "loss": 0.7629, "step": 6557 }, { "epoch": 0.37, "learning_rate": 0.00014467731779749783, "loss": 0.6668, "step": 6558 }, { "epoch": 0.37, "learning_rate": 0.00014466088068528068, "loss": 0.7161, "step": 6559 }, { "epoch": 0.37, "learning_rate": 0.00014464444206564164, "loss": 0.6366, "step": 6560 }, { "epoch": 0.37, "learning_rate": 0.00014462800193913562, "loss": 0.777, "step": 6561 }, { "epoch": 0.37, "learning_rate": 0.0001446115603063175, "loss": 0.7639, "step": 6562 }, { "epoch": 0.37, "learning_rate": 0.00014459511716774222, "loss": 0.692, "step": 6563 }, { "epoch": 0.37, "learning_rate": 0.00014457867252396476, "loss": 0.8105, "step": 6564 }, { "epoch": 0.37, "learning_rate": 0.0001445622263755402, "loss": 0.6355, "step": 6565 }, { "epoch": 0.37, "learning_rate": 0.00014454577872302364, "loss": 0.738, "step": 6566 }, { "epoch": 0.37, "learning_rate": 0.00014452932956697022, "loss": 0.7378, "step": 6567 }, { "epoch": 0.37, "learning_rate": 0.00014451287890793513, "loss": 0.795, "step": 6568 }, { "epoch": 0.37, "learning_rate": 0.00014449642674647367, "loss": 0.7343, "step": 6569 }, { "epoch": 0.37, "learning_rate": 0.0001444799730831411, "loss": 0.8132, "step": 6570 }, { "epoch": 0.37, "learning_rate": 0.00014446351791849276, "loss": 0.8107, "step": 6571 }, { "epoch": 0.37, "learning_rate": 0.00014444706125308406, "loss": 0.6795, "step": 6572 }, { "epoch": 0.37, "learning_rate": 0.00014443060308747056, "loss": 0.7062, "step": 6573 }, { "epoch": 0.37, "learning_rate": 0.00014441414342220764, "loss": 0.6661, "step": 6574 }, { "epoch": 0.37, "learning_rate": 0.00014439768225785088, "loss": 0.7146, "step": 6575 }, { "epoch": 0.37, "learning_rate": 0.00014438121959495592, "loss": 0.7457, "step": 6576 }, { "epoch": 0.37, "learning_rate": 0.00014436475543407843, "loss": 0.6876, "step": 6577 }, { "epoch": 0.37, "learning_rate": 0.00014434828977577405, "loss": 0.6903, "step": 6578 }, { "epoch": 0.37, "learning_rate": 0.0001443318226205986, "loss": 0.7602, "step": 6579 }, { "epoch": 0.37, "learning_rate": 0.00014431535396910788, "loss": 0.7598, "step": 6580 }, { "epoch": 0.37, "learning_rate": 0.00014429888382185772, "loss": 0.7977, "step": 6581 }, { "epoch": 0.37, "learning_rate": 0.00014428241217940405, "loss": 0.6921, "step": 6582 }, { "epoch": 0.37, "learning_rate": 0.00014426593904230288, "loss": 0.8892, "step": 6583 }, { "epoch": 0.37, "learning_rate": 0.00014424946441111015, "loss": 0.7593, "step": 6584 }, { "epoch": 0.37, "learning_rate": 0.00014423298828638195, "loss": 0.5677, "step": 6585 }, { "epoch": 0.37, "learning_rate": 0.00014421651066867438, "loss": 0.8126, "step": 6586 }, { "epoch": 0.37, "learning_rate": 0.0001442000315585436, "loss": 0.6855, "step": 6587 }, { "epoch": 0.37, "learning_rate": 0.0001441835509565459, "loss": 0.7438, "step": 6588 }, { "epoch": 0.37, "learning_rate": 0.0001441670688632374, "loss": 0.6316, "step": 6589 }, { "epoch": 0.37, "learning_rate": 0.00014415058527917452, "loss": 0.7414, "step": 6590 }, { "epoch": 0.37, "learning_rate": 0.0001441341002049136, "loss": 0.6603, "step": 6591 }, { "epoch": 0.37, "learning_rate": 0.00014411761364101108, "loss": 0.7774, "step": 6592 }, { "epoch": 0.37, "learning_rate": 0.00014410112558802337, "loss": 0.7029, "step": 6593 }, { "epoch": 0.37, "learning_rate": 0.00014408463604650702, "loss": 0.6971, "step": 6594 }, { "epoch": 0.37, "learning_rate": 0.00014406814501701857, "loss": 0.7144, "step": 6595 }, { "epoch": 0.37, "learning_rate": 0.00014405165250011467, "loss": 0.5965, "step": 6596 }, { "epoch": 0.37, "learning_rate": 0.00014403515849635196, "loss": 0.6086, "step": 6597 }, { "epoch": 0.37, "learning_rate": 0.0001440186630062872, "loss": 0.7888, "step": 6598 }, { "epoch": 0.37, "learning_rate": 0.00014400216603047708, "loss": 0.7982, "step": 6599 }, { "epoch": 0.37, "learning_rate": 0.00014398566756947847, "loss": 0.6907, "step": 6600 }, { "epoch": 0.37, "learning_rate": 0.0001439691676238482, "loss": 0.6975, "step": 6601 }, { "epoch": 0.37, "learning_rate": 0.00014395266619414323, "loss": 0.6794, "step": 6602 }, { "epoch": 0.37, "learning_rate": 0.00014393616328092052, "loss": 0.7026, "step": 6603 }, { "epoch": 0.37, "learning_rate": 0.00014391965888473703, "loss": 0.7256, "step": 6604 }, { "epoch": 0.37, "learning_rate": 0.00014390315300614993, "loss": 0.6358, "step": 6605 }, { "epoch": 0.37, "learning_rate": 0.00014388664564571622, "loss": 0.7414, "step": 6606 }, { "epoch": 0.37, "learning_rate": 0.00014387013680399315, "loss": 0.8319, "step": 6607 }, { "epoch": 0.37, "learning_rate": 0.00014385362648153792, "loss": 0.7818, "step": 6608 }, { "epoch": 0.37, "learning_rate": 0.00014383711467890774, "loss": 0.7156, "step": 6609 }, { "epoch": 0.37, "learning_rate": 0.00014382060139666003, "loss": 0.707, "step": 6610 }, { "epoch": 0.38, "learning_rate": 0.00014380408663535206, "loss": 0.7466, "step": 6611 }, { "epoch": 0.38, "learning_rate": 0.0001437875703955413, "loss": 0.758, "step": 6612 }, { "epoch": 0.38, "learning_rate": 0.00014377105267778518, "loss": 0.672, "step": 6613 }, { "epoch": 0.38, "learning_rate": 0.00014375453348264126, "loss": 0.7545, "step": 6614 }, { "epoch": 0.38, "learning_rate": 0.0001437380128106671, "loss": 0.7041, "step": 6615 }, { "epoch": 0.38, "learning_rate": 0.00014372149066242026, "loss": 0.7233, "step": 6616 }, { "epoch": 0.38, "learning_rate": 0.00014370496703845846, "loss": 0.8398, "step": 6617 }, { "epoch": 0.38, "learning_rate": 0.00014368844193933942, "loss": 0.7624, "step": 6618 }, { "epoch": 0.38, "learning_rate": 0.00014367191536562087, "loss": 0.7602, "step": 6619 }, { "epoch": 0.38, "learning_rate": 0.00014365538731786066, "loss": 0.7735, "step": 6620 }, { "epoch": 0.38, "learning_rate": 0.00014363885779661664, "loss": 0.7584, "step": 6621 }, { "epoch": 0.38, "learning_rate": 0.00014362232680244672, "loss": 0.6665, "step": 6622 }, { "epoch": 0.38, "learning_rate": 0.00014360579433590884, "loss": 0.7802, "step": 6623 }, { "epoch": 0.38, "learning_rate": 0.00014358926039756105, "loss": 0.7349, "step": 6624 }, { "epoch": 0.38, "learning_rate": 0.00014357272498796144, "loss": 0.6846, "step": 6625 }, { "epoch": 0.38, "learning_rate": 0.00014355618810766804, "loss": 0.755, "step": 6626 }, { "epoch": 0.38, "learning_rate": 0.00014353964975723907, "loss": 0.7937, "step": 6627 }, { "epoch": 0.38, "learning_rate": 0.00014352310993723277, "loss": 0.6473, "step": 6628 }, { "epoch": 0.38, "learning_rate": 0.00014350656864820733, "loss": 0.754, "step": 6629 }, { "epoch": 0.38, "learning_rate": 0.00014349002589072112, "loss": 0.7486, "step": 6630 }, { "epoch": 0.38, "learning_rate": 0.00014347348166533248, "loss": 0.7226, "step": 6631 }, { "epoch": 0.38, "learning_rate": 0.00014345693597259978, "loss": 0.8441, "step": 6632 }, { "epoch": 0.38, "learning_rate": 0.00014344038881308158, "loss": 0.7151, "step": 6633 }, { "epoch": 0.38, "learning_rate": 0.0001434238401873363, "loss": 0.7167, "step": 6634 }, { "epoch": 0.38, "learning_rate": 0.0001434072900959225, "loss": 0.6973, "step": 6635 }, { "epoch": 0.38, "learning_rate": 0.00014339073853939885, "loss": 0.7064, "step": 6636 }, { "epoch": 0.38, "learning_rate": 0.000143374185518324, "loss": 0.6751, "step": 6637 }, { "epoch": 0.38, "learning_rate": 0.00014335763103325662, "loss": 0.7559, "step": 6638 }, { "epoch": 0.38, "learning_rate": 0.00014334107508475545, "loss": 0.7401, "step": 6639 }, { "epoch": 0.38, "learning_rate": 0.00014332451767337938, "loss": 0.8744, "step": 6640 }, { "epoch": 0.38, "learning_rate": 0.0001433079587996872, "loss": 0.7182, "step": 6641 }, { "epoch": 0.38, "learning_rate": 0.0001432913984642378, "loss": 0.8921, "step": 6642 }, { "epoch": 0.38, "learning_rate": 0.0001432748366675902, "loss": 0.7726, "step": 6643 }, { "epoch": 0.38, "learning_rate": 0.00014325827341030338, "loss": 0.6909, "step": 6644 }, { "epoch": 0.38, "learning_rate": 0.00014324170869293635, "loss": 0.6698, "step": 6645 }, { "epoch": 0.38, "learning_rate": 0.00014322514251604825, "loss": 0.7843, "step": 6646 }, { "epoch": 0.38, "learning_rate": 0.00014320857488019824, "loss": 0.6496, "step": 6647 }, { "epoch": 0.38, "learning_rate": 0.00014319200578594553, "loss": 0.714, "step": 6648 }, { "epoch": 0.38, "learning_rate": 0.00014317543523384928, "loss": 0.7022, "step": 6649 }, { "epoch": 0.38, "learning_rate": 0.00014315886322446893, "loss": 0.7588, "step": 6650 }, { "epoch": 0.38, "learning_rate": 0.00014314228975836375, "loss": 0.7111, "step": 6651 }, { "epoch": 0.38, "learning_rate": 0.00014312571483609312, "loss": 0.6502, "step": 6652 }, { "epoch": 0.38, "learning_rate": 0.0001431091384582165, "loss": 0.6593, "step": 6653 }, { "epoch": 0.38, "learning_rate": 0.00014309256062529344, "loss": 0.814, "step": 6654 }, { "epoch": 0.38, "learning_rate": 0.00014307598133788343, "loss": 0.7303, "step": 6655 }, { "epoch": 0.38, "learning_rate": 0.00014305940059654605, "loss": 0.6605, "step": 6656 }, { "epoch": 0.38, "learning_rate": 0.00014304281840184102, "loss": 0.711, "step": 6657 }, { "epoch": 0.38, "learning_rate": 0.00014302623475432794, "loss": 0.7175, "step": 6658 }, { "epoch": 0.38, "learning_rate": 0.00014300964965456662, "loss": 0.7949, "step": 6659 }, { "epoch": 0.38, "learning_rate": 0.00014299306310311684, "loss": 0.6709, "step": 6660 }, { "epoch": 0.38, "learning_rate": 0.00014297647510053843, "loss": 0.7062, "step": 6661 }, { "epoch": 0.38, "learning_rate": 0.00014295988564739124, "loss": 0.6552, "step": 6662 }, { "epoch": 0.38, "learning_rate": 0.0001429432947442353, "loss": 0.6924, "step": 6663 }, { "epoch": 0.38, "learning_rate": 0.00014292670239163046, "loss": 0.7242, "step": 6664 }, { "epoch": 0.38, "learning_rate": 0.00014291010859013688, "loss": 0.7235, "step": 6665 }, { "epoch": 0.38, "learning_rate": 0.0001428935133403146, "loss": 0.7081, "step": 6666 }, { "epoch": 0.38, "learning_rate": 0.00014287691664272375, "loss": 0.6033, "step": 6667 }, { "epoch": 0.38, "learning_rate": 0.0001428603184979245, "loss": 0.7427, "step": 6668 }, { "epoch": 0.38, "learning_rate": 0.00014284371890647713, "loss": 0.7336, "step": 6669 }, { "epoch": 0.38, "learning_rate": 0.00014282711786894184, "loss": 0.6513, "step": 6670 }, { "epoch": 0.38, "learning_rate": 0.00014281051538587904, "loss": 0.704, "step": 6671 }, { "epoch": 0.38, "learning_rate": 0.00014279391145784903, "loss": 0.7433, "step": 6672 }, { "epoch": 0.38, "learning_rate": 0.00014277730608541232, "loss": 0.7101, "step": 6673 }, { "epoch": 0.38, "learning_rate": 0.00014276069926912933, "loss": 0.7657, "step": 6674 }, { "epoch": 0.38, "learning_rate": 0.00014274409100956058, "loss": 0.8074, "step": 6675 }, { "epoch": 0.38, "learning_rate": 0.00014272748130726668, "loss": 0.7268, "step": 6676 }, { "epoch": 0.38, "learning_rate": 0.00014271087016280823, "loss": 0.7184, "step": 6677 }, { "epoch": 0.38, "learning_rate": 0.0001426942575767459, "loss": 0.7173, "step": 6678 }, { "epoch": 0.38, "learning_rate": 0.00014267764354964038, "loss": 0.7276, "step": 6679 }, { "epoch": 0.38, "learning_rate": 0.00014266102808205248, "loss": 0.6619, "step": 6680 }, { "epoch": 0.38, "learning_rate": 0.000142644411174543, "loss": 0.6498, "step": 6681 }, { "epoch": 0.38, "learning_rate": 0.00014262779282767278, "loss": 0.6514, "step": 6682 }, { "epoch": 0.38, "learning_rate": 0.0001426111730420028, "loss": 0.7085, "step": 6683 }, { "epoch": 0.38, "learning_rate": 0.00014259455181809394, "loss": 0.6526, "step": 6684 }, { "epoch": 0.38, "learning_rate": 0.00014257792915650728, "loss": 0.651, "step": 6685 }, { "epoch": 0.38, "learning_rate": 0.00014256130505780384, "loss": 0.5875, "step": 6686 }, { "epoch": 0.38, "learning_rate": 0.00014254467952254472, "loss": 0.8239, "step": 6687 }, { "epoch": 0.38, "learning_rate": 0.0001425280525512911, "loss": 0.6901, "step": 6688 }, { "epoch": 0.38, "learning_rate": 0.00014251142414460415, "loss": 0.7683, "step": 6689 }, { "epoch": 0.38, "learning_rate": 0.00014249479430304517, "loss": 0.7217, "step": 6690 }, { "epoch": 0.38, "learning_rate": 0.0001424781630271754, "loss": 0.7493, "step": 6691 }, { "epoch": 0.38, "learning_rate": 0.00014246153031755626, "loss": 0.7456, "step": 6692 }, { "epoch": 0.38, "learning_rate": 0.0001424448961747491, "loss": 0.6882, "step": 6693 }, { "epoch": 0.38, "learning_rate": 0.00014242826059931537, "loss": 0.7875, "step": 6694 }, { "epoch": 0.38, "learning_rate": 0.00014241162359181661, "loss": 0.7524, "step": 6695 }, { "epoch": 0.38, "learning_rate": 0.0001423949851528143, "loss": 0.6896, "step": 6696 }, { "epoch": 0.38, "learning_rate": 0.00014237834528287003, "loss": 0.7113, "step": 6697 }, { "epoch": 0.38, "learning_rate": 0.00014236170398254552, "loss": 0.609, "step": 6698 }, { "epoch": 0.38, "learning_rate": 0.00014234506125240237, "loss": 0.6114, "step": 6699 }, { "epoch": 0.38, "learning_rate": 0.00014232841709300233, "loss": 0.7725, "step": 6700 }, { "epoch": 0.38, "learning_rate": 0.00014231177150490721, "loss": 0.7394, "step": 6701 }, { "epoch": 0.38, "learning_rate": 0.00014229512448867888, "loss": 0.7273, "step": 6702 }, { "epoch": 0.38, "learning_rate": 0.00014227847604487913, "loss": 0.6276, "step": 6703 }, { "epoch": 0.38, "learning_rate": 0.00014226182617406996, "loss": 0.7953, "step": 6704 }, { "epoch": 0.38, "learning_rate": 0.0001422451748768133, "loss": 0.7298, "step": 6705 }, { "epoch": 0.38, "learning_rate": 0.0001422285221536712, "loss": 0.8031, "step": 6706 }, { "epoch": 0.38, "learning_rate": 0.00014221186800520574, "loss": 0.7185, "step": 6707 }, { "epoch": 0.38, "learning_rate": 0.00014219521243197903, "loss": 0.8129, "step": 6708 }, { "epoch": 0.38, "learning_rate": 0.00014217855543455322, "loss": 0.6883, "step": 6709 }, { "epoch": 0.38, "learning_rate": 0.00014216189701349058, "loss": 0.7831, "step": 6710 }, { "epoch": 0.38, "learning_rate": 0.00014214523716935332, "loss": 0.7426, "step": 6711 }, { "epoch": 0.38, "learning_rate": 0.0001421285759027038, "loss": 0.6826, "step": 6712 }, { "epoch": 0.38, "learning_rate": 0.00014211191321410432, "loss": 0.5839, "step": 6713 }, { "epoch": 0.38, "learning_rate": 0.00014209524910411734, "loss": 0.821, "step": 6714 }, { "epoch": 0.38, "learning_rate": 0.0001420785835733053, "loss": 0.7561, "step": 6715 }, { "epoch": 0.38, "learning_rate": 0.00014206191662223072, "loss": 0.7409, "step": 6716 }, { "epoch": 0.38, "learning_rate": 0.00014204524825145613, "loss": 0.6645, "step": 6717 }, { "epoch": 0.38, "learning_rate": 0.00014202857846154417, "loss": 0.646, "step": 6718 }, { "epoch": 0.38, "learning_rate": 0.00014201190725305745, "loss": 0.7195, "step": 6719 }, { "epoch": 0.38, "learning_rate": 0.00014199523462655864, "loss": 0.7538, "step": 6720 }, { "epoch": 0.38, "learning_rate": 0.0001419785605826106, "loss": 0.739, "step": 6721 }, { "epoch": 0.38, "learning_rate": 0.000141961885121776, "loss": 0.6742, "step": 6722 }, { "epoch": 0.38, "learning_rate": 0.00014194520824461771, "loss": 0.6766, "step": 6723 }, { "epoch": 0.38, "learning_rate": 0.00014192852995169867, "loss": 0.7848, "step": 6724 }, { "epoch": 0.38, "learning_rate": 0.0001419118502435818, "loss": 0.7144, "step": 6725 }, { "epoch": 0.38, "learning_rate": 0.00014189516912083003, "loss": 0.6903, "step": 6726 }, { "epoch": 0.38, "learning_rate": 0.00014187848658400642, "loss": 0.7224, "step": 6727 }, { "epoch": 0.38, "learning_rate": 0.00014186180263367409, "loss": 0.7244, "step": 6728 }, { "epoch": 0.38, "learning_rate": 0.00014184511727039612, "loss": 0.6881, "step": 6729 }, { "epoch": 0.38, "learning_rate": 0.00014182843049473568, "loss": 0.787, "step": 6730 }, { "epoch": 0.38, "learning_rate": 0.00014181174230725604, "loss": 0.8085, "step": 6731 }, { "epoch": 0.38, "learning_rate": 0.00014179505270852042, "loss": 0.685, "step": 6732 }, { "epoch": 0.38, "learning_rate": 0.0001417783616990922, "loss": 0.7726, "step": 6733 }, { "epoch": 0.38, "learning_rate": 0.00014176166927953467, "loss": 0.7714, "step": 6734 }, { "epoch": 0.38, "learning_rate": 0.00014174497545041127, "loss": 0.7331, "step": 6735 }, { "epoch": 0.38, "learning_rate": 0.0001417282802122855, "loss": 0.6663, "step": 6736 }, { "epoch": 0.38, "learning_rate": 0.00014171158356572082, "loss": 0.6782, "step": 6737 }, { "epoch": 0.38, "learning_rate": 0.0001416948855112808, "loss": 0.695, "step": 6738 }, { "epoch": 0.38, "learning_rate": 0.00014167818604952906, "loss": 0.5972, "step": 6739 }, { "epoch": 0.38, "learning_rate": 0.0001416614851810292, "loss": 0.5968, "step": 6740 }, { "epoch": 0.38, "learning_rate": 0.000141644782906345, "loss": 0.7093, "step": 6741 }, { "epoch": 0.38, "learning_rate": 0.00014162807922604012, "loss": 0.7688, "step": 6742 }, { "epoch": 0.38, "learning_rate": 0.0001416113741406784, "loss": 0.7583, "step": 6743 }, { "epoch": 0.38, "learning_rate": 0.0001415946676508237, "loss": 0.7543, "step": 6744 }, { "epoch": 0.38, "learning_rate": 0.00014157795975703986, "loss": 0.8211, "step": 6745 }, { "epoch": 0.38, "learning_rate": 0.00014156125045989085, "loss": 0.7191, "step": 6746 }, { "epoch": 0.38, "learning_rate": 0.0001415445397599406, "loss": 0.7472, "step": 6747 }, { "epoch": 0.38, "learning_rate": 0.00014152782765775322, "loss": 0.6847, "step": 6748 }, { "epoch": 0.38, "learning_rate": 0.00014151111415389273, "loss": 0.6518, "step": 6749 }, { "epoch": 0.38, "learning_rate": 0.00014149439924892327, "loss": 0.7613, "step": 6750 }, { "epoch": 0.38, "learning_rate": 0.00014147768294340903, "loss": 0.6916, "step": 6751 }, { "epoch": 0.38, "learning_rate": 0.0001414609652379142, "loss": 0.7783, "step": 6752 }, { "epoch": 0.38, "learning_rate": 0.00014144424613300304, "loss": 0.6244, "step": 6753 }, { "epoch": 0.38, "learning_rate": 0.00014142752562923988, "loss": 0.6593, "step": 6754 }, { "epoch": 0.38, "learning_rate": 0.00014141080372718912, "loss": 0.692, "step": 6755 }, { "epoch": 0.38, "learning_rate": 0.00014139408042741512, "loss": 0.6736, "step": 6756 }, { "epoch": 0.38, "learning_rate": 0.00014137735573048233, "loss": 0.699, "step": 6757 }, { "epoch": 0.38, "learning_rate": 0.00014136062963695525, "loss": 0.756, "step": 6758 }, { "epoch": 0.38, "learning_rate": 0.00014134390214739847, "loss": 0.7592, "step": 6759 }, { "epoch": 0.38, "learning_rate": 0.0001413271732623766, "loss": 0.7234, "step": 6760 }, { "epoch": 0.38, "learning_rate": 0.0001413104429824542, "loss": 0.7035, "step": 6761 }, { "epoch": 0.38, "learning_rate": 0.000141293711308196, "loss": 0.6349, "step": 6762 }, { "epoch": 0.38, "learning_rate": 0.0001412769782401668, "loss": 0.663, "step": 6763 }, { "epoch": 0.38, "learning_rate": 0.00014126024377893127, "loss": 0.6785, "step": 6764 }, { "epoch": 0.38, "learning_rate": 0.00014124350792505433, "loss": 0.6671, "step": 6765 }, { "epoch": 0.38, "learning_rate": 0.00014122677067910085, "loss": 0.801, "step": 6766 }, { "epoch": 0.38, "learning_rate": 0.00014121003204163571, "loss": 0.7116, "step": 6767 }, { "epoch": 0.38, "learning_rate": 0.00014119329201322392, "loss": 0.7718, "step": 6768 }, { "epoch": 0.38, "learning_rate": 0.0001411765505944305, "loss": 0.744, "step": 6769 }, { "epoch": 0.38, "learning_rate": 0.00014115980778582051, "loss": 0.7294, "step": 6770 }, { "epoch": 0.38, "learning_rate": 0.00014114306358795905, "loss": 0.6636, "step": 6771 }, { "epoch": 0.38, "learning_rate": 0.0001411263180014113, "loss": 0.7196, "step": 6772 }, { "epoch": 0.38, "learning_rate": 0.00014110957102674246, "loss": 0.6194, "step": 6773 }, { "epoch": 0.38, "learning_rate": 0.00014109282266451778, "loss": 0.6269, "step": 6774 }, { "epoch": 0.38, "learning_rate": 0.00014107607291530256, "loss": 0.7099, "step": 6775 }, { "epoch": 0.38, "learning_rate": 0.00014105932177966218, "loss": 0.7268, "step": 6776 }, { "epoch": 0.38, "learning_rate": 0.00014104256925816198, "loss": 0.6707, "step": 6777 }, { "epoch": 0.38, "learning_rate": 0.00014102581535136742, "loss": 0.711, "step": 6778 }, { "epoch": 0.38, "learning_rate": 0.00014100906005984403, "loss": 0.8119, "step": 6779 }, { "epoch": 0.38, "learning_rate": 0.00014099230338415728, "loss": 0.7488, "step": 6780 }, { "epoch": 0.38, "learning_rate": 0.0001409755453248728, "loss": 0.69, "step": 6781 }, { "epoch": 0.38, "learning_rate": 0.00014095878588255622, "loss": 0.6857, "step": 6782 }, { "epoch": 0.38, "learning_rate": 0.00014094202505777318, "loss": 0.7421, "step": 6783 }, { "epoch": 0.38, "learning_rate": 0.0001409252628510894, "loss": 0.6827, "step": 6784 }, { "epoch": 0.38, "learning_rate": 0.0001409084992630707, "loss": 0.755, "step": 6785 }, { "epoch": 0.38, "learning_rate": 0.00014089173429428285, "loss": 0.755, "step": 6786 }, { "epoch": 0.38, "learning_rate": 0.00014087496794529174, "loss": 0.7217, "step": 6787 }, { "epoch": 0.39, "learning_rate": 0.00014085820021666326, "loss": 0.7772, "step": 6788 }, { "epoch": 0.39, "learning_rate": 0.00014084143110896336, "loss": 0.7517, "step": 6789 }, { "epoch": 0.39, "learning_rate": 0.00014082466062275805, "loss": 0.7071, "step": 6790 }, { "epoch": 0.39, "learning_rate": 0.00014080788875861337, "loss": 0.7363, "step": 6791 }, { "epoch": 0.39, "learning_rate": 0.00014079111551709546, "loss": 0.645, "step": 6792 }, { "epoch": 0.39, "learning_rate": 0.00014077434089877037, "loss": 0.7182, "step": 6793 }, { "epoch": 0.39, "learning_rate": 0.00014075756490420434, "loss": 0.7322, "step": 6794 }, { "epoch": 0.39, "learning_rate": 0.00014074078753396364, "loss": 0.6883, "step": 6795 }, { "epoch": 0.39, "learning_rate": 0.0001407240087886145, "loss": 0.7057, "step": 6796 }, { "epoch": 0.39, "learning_rate": 0.00014070722866872326, "loss": 0.7232, "step": 6797 }, { "epoch": 0.39, "learning_rate": 0.00014069044717485632, "loss": 0.6755, "step": 6798 }, { "epoch": 0.39, "learning_rate": 0.00014067366430758004, "loss": 0.7211, "step": 6799 }, { "epoch": 0.39, "learning_rate": 0.00014065688006746093, "loss": 0.7154, "step": 6800 }, { "epoch": 0.39, "learning_rate": 0.00014064009445506547, "loss": 0.7222, "step": 6801 }, { "epoch": 0.39, "learning_rate": 0.00014062330747096027, "loss": 0.6249, "step": 6802 }, { "epoch": 0.39, "learning_rate": 0.0001406065191157119, "loss": 0.751, "step": 6803 }, { "epoch": 0.39, "learning_rate": 0.000140589729389887, "loss": 0.7176, "step": 6804 }, { "epoch": 0.39, "learning_rate": 0.00014057293829405229, "loss": 0.7884, "step": 6805 }, { "epoch": 0.39, "learning_rate": 0.0001405561458287745, "loss": 0.6467, "step": 6806 }, { "epoch": 0.39, "learning_rate": 0.00014053935199462043, "loss": 0.7476, "step": 6807 }, { "epoch": 0.39, "learning_rate": 0.00014052255679215694, "loss": 0.7207, "step": 6808 }, { "epoch": 0.39, "learning_rate": 0.00014050576022195084, "loss": 0.7338, "step": 6809 }, { "epoch": 0.39, "learning_rate": 0.0001404889622845691, "loss": 0.7281, "step": 6810 }, { "epoch": 0.39, "learning_rate": 0.00014047216298057873, "loss": 0.7158, "step": 6811 }, { "epoch": 0.39, "learning_rate": 0.0001404553623105467, "loss": 0.6638, "step": 6812 }, { "epoch": 0.39, "learning_rate": 0.0001404385602750401, "loss": 0.6911, "step": 6813 }, { "epoch": 0.39, "learning_rate": 0.00014042175687462605, "loss": 0.7679, "step": 6814 }, { "epoch": 0.39, "learning_rate": 0.0001404049521098717, "loss": 0.7133, "step": 6815 }, { "epoch": 0.39, "learning_rate": 0.0001403881459813442, "loss": 0.7258, "step": 6816 }, { "epoch": 0.39, "learning_rate": 0.00014037133848961085, "loss": 0.7898, "step": 6817 }, { "epoch": 0.39, "learning_rate": 0.00014035452963523902, "loss": 0.6886, "step": 6818 }, { "epoch": 0.39, "learning_rate": 0.00014033771941879595, "loss": 0.8172, "step": 6819 }, { "epoch": 0.39, "learning_rate": 0.00014032090784084903, "loss": 0.8581, "step": 6820 }, { "epoch": 0.39, "learning_rate": 0.00014030409490196574, "loss": 0.7445, "step": 6821 }, { "epoch": 0.39, "learning_rate": 0.00014028728060271357, "loss": 0.716, "step": 6822 }, { "epoch": 0.39, "learning_rate": 0.00014027046494366, "loss": 0.7275, "step": 6823 }, { "epoch": 0.39, "learning_rate": 0.00014025364792537263, "loss": 0.6567, "step": 6824 }, { "epoch": 0.39, "learning_rate": 0.00014023682954841907, "loss": 0.7128, "step": 6825 }, { "epoch": 0.39, "learning_rate": 0.000140220009813367, "loss": 0.7146, "step": 6826 }, { "epoch": 0.39, "learning_rate": 0.0001402031887207841, "loss": 0.8006, "step": 6827 }, { "epoch": 0.39, "learning_rate": 0.00014018636627123815, "loss": 0.613, "step": 6828 }, { "epoch": 0.39, "learning_rate": 0.00014016954246529696, "loss": 0.6655, "step": 6829 }, { "epoch": 0.39, "learning_rate": 0.00014015271730352834, "loss": 0.7163, "step": 6830 }, { "epoch": 0.39, "learning_rate": 0.00014013589078650023, "loss": 0.7884, "step": 6831 }, { "epoch": 0.39, "learning_rate": 0.00014011906291478053, "loss": 0.739, "step": 6832 }, { "epoch": 0.39, "learning_rate": 0.00014010223368893725, "loss": 0.7746, "step": 6833 }, { "epoch": 0.39, "learning_rate": 0.00014008540310953842, "loss": 0.6562, "step": 6834 }, { "epoch": 0.39, "learning_rate": 0.00014006857117715207, "loss": 0.628, "step": 6835 }, { "epoch": 0.39, "learning_rate": 0.0001400517378923464, "loss": 0.7924, "step": 6836 }, { "epoch": 0.39, "learning_rate": 0.00014003490325568954, "loss": 0.6594, "step": 6837 }, { "epoch": 0.39, "learning_rate": 0.00014001806726774964, "loss": 0.7082, "step": 6838 }, { "epoch": 0.39, "learning_rate": 0.00014000122992909505, "loss": 0.7349, "step": 6839 }, { "epoch": 0.39, "learning_rate": 0.00013998439124029405, "loss": 0.7413, "step": 6840 }, { "epoch": 0.39, "learning_rate": 0.000139967551201915, "loss": 0.7506, "step": 6841 }, { "epoch": 0.39, "learning_rate": 0.00013995070981452626, "loss": 0.6551, "step": 6842 }, { "epoch": 0.39, "learning_rate": 0.0001399338670786963, "loss": 0.7814, "step": 6843 }, { "epoch": 0.39, "learning_rate": 0.0001399170229949936, "loss": 0.7603, "step": 6844 }, { "epoch": 0.39, "learning_rate": 0.00013990017756398664, "loss": 0.7527, "step": 6845 }, { "epoch": 0.39, "learning_rate": 0.00013988333078624406, "loss": 0.8197, "step": 6846 }, { "epoch": 0.39, "learning_rate": 0.00013986648266233452, "loss": 0.718, "step": 6847 }, { "epoch": 0.39, "learning_rate": 0.0001398496331928266, "loss": 0.7399, "step": 6848 }, { "epoch": 0.39, "learning_rate": 0.00013983278237828905, "loss": 0.7207, "step": 6849 }, { "epoch": 0.39, "learning_rate": 0.00013981593021929062, "loss": 0.7011, "step": 6850 }, { "epoch": 0.39, "learning_rate": 0.00013979907671640015, "loss": 0.716, "step": 6851 }, { "epoch": 0.39, "learning_rate": 0.00013978222187018645, "loss": 0.6692, "step": 6852 }, { "epoch": 0.39, "learning_rate": 0.00013976536568121843, "loss": 0.7248, "step": 6853 }, { "epoch": 0.39, "learning_rate": 0.00013974850815006503, "loss": 0.756, "step": 6854 }, { "epoch": 0.39, "learning_rate": 0.00013973164927729524, "loss": 0.7434, "step": 6855 }, { "epoch": 0.39, "learning_rate": 0.00013971478906347806, "loss": 0.6905, "step": 6856 }, { "epoch": 0.39, "learning_rate": 0.00013969792750918262, "loss": 0.6522, "step": 6857 }, { "epoch": 0.39, "learning_rate": 0.000139681064614978, "loss": 0.6944, "step": 6858 }, { "epoch": 0.39, "learning_rate": 0.0001396642003814334, "loss": 0.6975, "step": 6859 }, { "epoch": 0.39, "learning_rate": 0.00013964733480911802, "loss": 0.8878, "step": 6860 }, { "epoch": 0.39, "learning_rate": 0.0001396304678986011, "loss": 0.7437, "step": 6861 }, { "epoch": 0.39, "learning_rate": 0.00013961359965045195, "loss": 0.7858, "step": 6862 }, { "epoch": 0.39, "learning_rate": 0.00013959673006523992, "loss": 0.6356, "step": 6863 }, { "epoch": 0.39, "learning_rate": 0.0001395798591435344, "loss": 0.7123, "step": 6864 }, { "epoch": 0.39, "learning_rate": 0.00013956298688590484, "loss": 0.7673, "step": 6865 }, { "epoch": 0.39, "learning_rate": 0.00013954611329292072, "loss": 0.6811, "step": 6866 }, { "epoch": 0.39, "learning_rate": 0.00013952923836515154, "loss": 0.7268, "step": 6867 }, { "epoch": 0.39, "learning_rate": 0.00013951236210316692, "loss": 0.6852, "step": 6868 }, { "epoch": 0.39, "learning_rate": 0.00013949548450753645, "loss": 0.7322, "step": 6869 }, { "epoch": 0.39, "learning_rate": 0.0001394786055788298, "loss": 0.7405, "step": 6870 }, { "epoch": 0.39, "learning_rate": 0.00013946172531761666, "loss": 0.7501, "step": 6871 }, { "epoch": 0.39, "learning_rate": 0.0001394448437244668, "loss": 0.8345, "step": 6872 }, { "epoch": 0.39, "learning_rate": 0.00013942796079995004, "loss": 0.7249, "step": 6873 }, { "epoch": 0.39, "learning_rate": 0.0001394110765446362, "loss": 0.7342, "step": 6874 }, { "epoch": 0.39, "learning_rate": 0.00013939419095909512, "loss": 0.6521, "step": 6875 }, { "epoch": 0.39, "learning_rate": 0.00013937730404389682, "loss": 0.6918, "step": 6876 }, { "epoch": 0.39, "learning_rate": 0.00013936041579961123, "loss": 0.7606, "step": 6877 }, { "epoch": 0.39, "learning_rate": 0.00013934352622680835, "loss": 0.7827, "step": 6878 }, { "epoch": 0.39, "learning_rate": 0.0001393266353260583, "loss": 0.6864, "step": 6879 }, { "epoch": 0.39, "learning_rate": 0.0001393097430979312, "loss": 0.6642, "step": 6880 }, { "epoch": 0.39, "learning_rate": 0.00013929284954299716, "loss": 0.6942, "step": 6881 }, { "epoch": 0.39, "learning_rate": 0.00013927595466182638, "loss": 0.6351, "step": 6882 }, { "epoch": 0.39, "learning_rate": 0.00013925905845498914, "loss": 0.7414, "step": 6883 }, { "epoch": 0.39, "learning_rate": 0.00013924216092305568, "loss": 0.711, "step": 6884 }, { "epoch": 0.39, "learning_rate": 0.00013922526206659643, "loss": 0.6405, "step": 6885 }, { "epoch": 0.39, "learning_rate": 0.0001392083618861817, "loss": 0.7369, "step": 6886 }, { "epoch": 0.39, "learning_rate": 0.00013919146038238192, "loss": 0.7346, "step": 6887 }, { "epoch": 0.39, "learning_rate": 0.00013917455755576755, "loss": 0.6881, "step": 6888 }, { "epoch": 0.39, "learning_rate": 0.00013915765340690917, "loss": 0.7115, "step": 6889 }, { "epoch": 0.39, "learning_rate": 0.00013914074793637725, "loss": 0.6522, "step": 6890 }, { "epoch": 0.39, "learning_rate": 0.00013912384114474243, "loss": 0.548, "step": 6891 }, { "epoch": 0.39, "learning_rate": 0.00013910693303257541, "loss": 0.738, "step": 6892 }, { "epoch": 0.39, "learning_rate": 0.00013909002360044682, "loss": 0.8062, "step": 6893 }, { "epoch": 0.39, "learning_rate": 0.00013907311284892736, "loss": 0.7741, "step": 6894 }, { "epoch": 0.39, "learning_rate": 0.00013905620077858793, "loss": 0.7808, "step": 6895 }, { "epoch": 0.39, "learning_rate": 0.0001390392873899993, "loss": 0.7251, "step": 6896 }, { "epoch": 0.39, "learning_rate": 0.0001390223726837323, "loss": 0.7642, "step": 6897 }, { "epoch": 0.39, "learning_rate": 0.00013900545666035793, "loss": 0.7312, "step": 6898 }, { "epoch": 0.39, "learning_rate": 0.00013898853932044707, "loss": 0.715, "step": 6899 }, { "epoch": 0.39, "learning_rate": 0.00013897162066457076, "loss": 0.702, "step": 6900 }, { "epoch": 0.39, "learning_rate": 0.00013895470069330004, "loss": 0.7232, "step": 6901 }, { "epoch": 0.39, "learning_rate": 0.000138937779407206, "loss": 0.6593, "step": 6902 }, { "epoch": 0.39, "learning_rate": 0.00013892085680685981, "loss": 0.7684, "step": 6903 }, { "epoch": 0.39, "learning_rate": 0.0001389039328928326, "loss": 0.809, "step": 6904 }, { "epoch": 0.39, "learning_rate": 0.00013888700766569566, "loss": 0.7141, "step": 6905 }, { "epoch": 0.39, "learning_rate": 0.0001388700811260202, "loss": 0.6404, "step": 6906 }, { "epoch": 0.39, "learning_rate": 0.00013885315327437757, "loss": 0.5851, "step": 6907 }, { "epoch": 0.39, "learning_rate": 0.00013883622411133914, "loss": 0.7228, "step": 6908 }, { "epoch": 0.39, "learning_rate": 0.00013881929363747627, "loss": 0.7675, "step": 6909 }, { "epoch": 0.39, "learning_rate": 0.00013880236185336044, "loss": 0.6736, "step": 6910 }, { "epoch": 0.39, "learning_rate": 0.00013878542875956314, "loss": 0.7752, "step": 6911 }, { "epoch": 0.39, "learning_rate": 0.00013876849435665592, "loss": 0.7089, "step": 6912 }, { "epoch": 0.39, "learning_rate": 0.0001387515586452103, "loss": 0.6974, "step": 6913 }, { "epoch": 0.39, "learning_rate": 0.00013873462162579799, "loss": 0.6749, "step": 6914 }, { "epoch": 0.39, "learning_rate": 0.00013871768329899058, "loss": 0.7434, "step": 6915 }, { "epoch": 0.39, "learning_rate": 0.0001387007436653598, "loss": 0.8078, "step": 6916 }, { "epoch": 0.39, "learning_rate": 0.00013868380272547744, "loss": 0.7669, "step": 6917 }, { "epoch": 0.39, "learning_rate": 0.00013866686047991534, "loss": 0.7711, "step": 6918 }, { "epoch": 0.39, "learning_rate": 0.00013864991692924523, "loss": 0.6489, "step": 6919 }, { "epoch": 0.39, "learning_rate": 0.00013863297207403907, "loss": 0.596, "step": 6920 }, { "epoch": 0.39, "learning_rate": 0.00013861602591486882, "loss": 0.8026, "step": 6921 }, { "epoch": 0.39, "learning_rate": 0.00013859907845230638, "loss": 0.6288, "step": 6922 }, { "epoch": 0.39, "learning_rate": 0.00013858212968692385, "loss": 0.8241, "step": 6923 }, { "epoch": 0.39, "learning_rate": 0.00013856517961929323, "loss": 0.8414, "step": 6924 }, { "epoch": 0.39, "learning_rate": 0.00013854822824998665, "loss": 0.7308, "step": 6925 }, { "epoch": 0.39, "learning_rate": 0.00013853127557957628, "loss": 0.7781, "step": 6926 }, { "epoch": 0.39, "learning_rate": 0.00013851432160863432, "loss": 0.7135, "step": 6927 }, { "epoch": 0.39, "learning_rate": 0.00013849736633773297, "loss": 0.7906, "step": 6928 }, { "epoch": 0.39, "learning_rate": 0.00013848040976744457, "loss": 0.704, "step": 6929 }, { "epoch": 0.39, "learning_rate": 0.0001384634518983414, "loss": 0.7297, "step": 6930 }, { "epoch": 0.39, "learning_rate": 0.0001384464927309959, "loss": 0.7317, "step": 6931 }, { "epoch": 0.39, "learning_rate": 0.00013842953226598037, "loss": 0.8532, "step": 6932 }, { "epoch": 0.39, "learning_rate": 0.00013841257050386738, "loss": 0.7183, "step": 6933 }, { "epoch": 0.39, "learning_rate": 0.0001383956074452294, "loss": 0.7161, "step": 6934 }, { "epoch": 0.39, "learning_rate": 0.00013837864309063896, "loss": 0.734, "step": 6935 }, { "epoch": 0.39, "learning_rate": 0.00013836167744066868, "loss": 0.7155, "step": 6936 }, { "epoch": 0.39, "learning_rate": 0.00013834471049589117, "loss": 0.7306, "step": 6937 }, { "epoch": 0.39, "learning_rate": 0.00013832774225687913, "loss": 0.7665, "step": 6938 }, { "epoch": 0.39, "learning_rate": 0.00013831077272420523, "loss": 0.6932, "step": 6939 }, { "epoch": 0.39, "learning_rate": 0.00013829380189844231, "loss": 0.8137, "step": 6940 }, { "epoch": 0.39, "learning_rate": 0.00013827682978016315, "loss": 0.7339, "step": 6941 }, { "epoch": 0.39, "learning_rate": 0.0001382598563699406, "loss": 0.7776, "step": 6942 }, { "epoch": 0.39, "learning_rate": 0.00013824288166834752, "loss": 0.7007, "step": 6943 }, { "epoch": 0.39, "learning_rate": 0.00013822590567595696, "loss": 0.7012, "step": 6944 }, { "epoch": 0.39, "learning_rate": 0.00013820892839334177, "loss": 0.7273, "step": 6945 }, { "epoch": 0.39, "learning_rate": 0.00013819194982107503, "loss": 0.7752, "step": 6946 }, { "epoch": 0.39, "learning_rate": 0.00013817496995972988, "loss": 0.7077, "step": 6947 }, { "epoch": 0.39, "learning_rate": 0.00013815798880987935, "loss": 0.745, "step": 6948 }, { "epoch": 0.39, "learning_rate": 0.0001381410063720966, "loss": 0.8637, "step": 6949 }, { "epoch": 0.39, "learning_rate": 0.0001381240226469549, "loss": 0.6108, "step": 6950 }, { "epoch": 0.39, "learning_rate": 0.00013810703763502744, "loss": 0.6412, "step": 6951 }, { "epoch": 0.39, "learning_rate": 0.0001380900513368875, "loss": 0.804, "step": 6952 }, { "epoch": 0.39, "learning_rate": 0.00013807306375310846, "loss": 0.7564, "step": 6953 }, { "epoch": 0.39, "learning_rate": 0.00013805607488426362, "loss": 0.7396, "step": 6954 }, { "epoch": 0.39, "learning_rate": 0.00013803908473092647, "loss": 0.696, "step": 6955 }, { "epoch": 0.39, "learning_rate": 0.00013802209329367047, "loss": 0.6731, "step": 6956 }, { "epoch": 0.39, "learning_rate": 0.0001380051005730691, "loss": 0.6085, "step": 6957 }, { "epoch": 0.39, "learning_rate": 0.00013798810656969588, "loss": 0.7016, "step": 6958 }, { "epoch": 0.39, "learning_rate": 0.00013797111128412446, "loss": 0.7108, "step": 6959 }, { "epoch": 0.39, "learning_rate": 0.00013795411471692848, "loss": 0.6625, "step": 6960 }, { "epoch": 0.39, "learning_rate": 0.00013793711686868157, "loss": 0.7293, "step": 6961 }, { "epoch": 0.39, "learning_rate": 0.00013792011773995747, "loss": 0.7464, "step": 6962 }, { "epoch": 0.39, "learning_rate": 0.0001379031173313299, "loss": 0.739, "step": 6963 }, { "epoch": 0.4, "learning_rate": 0.00013788611564337277, "loss": 0.7496, "step": 6964 }, { "epoch": 0.4, "learning_rate": 0.00013786911267665985, "loss": 0.8206, "step": 6965 }, { "epoch": 0.4, "learning_rate": 0.00013785210843176513, "loss": 0.8061, "step": 6966 }, { "epoch": 0.4, "learning_rate": 0.0001378351029092624, "loss": 0.705, "step": 6967 }, { "epoch": 0.4, "learning_rate": 0.00013781809610972573, "loss": 0.6465, "step": 6968 }, { "epoch": 0.4, "learning_rate": 0.00013780108803372916, "loss": 0.7672, "step": 6969 }, { "epoch": 0.4, "learning_rate": 0.00013778407868184672, "loss": 0.748, "step": 6970 }, { "epoch": 0.4, "learning_rate": 0.00013776706805465253, "loss": 0.6582, "step": 6971 }, { "epoch": 0.4, "learning_rate": 0.00013775005615272075, "loss": 0.8275, "step": 6972 }, { "epoch": 0.4, "learning_rate": 0.00013773304297662559, "loss": 0.7582, "step": 6973 }, { "epoch": 0.4, "learning_rate": 0.00013771602852694125, "loss": 0.7037, "step": 6974 }, { "epoch": 0.4, "learning_rate": 0.000137699012804242, "loss": 0.706, "step": 6975 }, { "epoch": 0.4, "learning_rate": 0.00013768199580910226, "loss": 0.6904, "step": 6976 }, { "epoch": 0.4, "learning_rate": 0.00013766497754209633, "loss": 0.6002, "step": 6977 }, { "epoch": 0.4, "learning_rate": 0.0001376479580037986, "loss": 0.7221, "step": 6978 }, { "epoch": 0.4, "learning_rate": 0.00013763093719478358, "loss": 0.7133, "step": 6979 }, { "epoch": 0.4, "learning_rate": 0.0001376139151156257, "loss": 0.7576, "step": 6980 }, { "epoch": 0.4, "learning_rate": 0.00013759689176689956, "loss": 0.6707, "step": 6981 }, { "epoch": 0.4, "learning_rate": 0.00013757986714917972, "loss": 0.7386, "step": 6982 }, { "epoch": 0.4, "learning_rate": 0.0001375628412630408, "loss": 0.7555, "step": 6983 }, { "epoch": 0.4, "learning_rate": 0.00013754581410905747, "loss": 0.8101, "step": 6984 }, { "epoch": 0.4, "learning_rate": 0.00013752878568780446, "loss": 0.8239, "step": 6985 }, { "epoch": 0.4, "learning_rate": 0.00013751175599985654, "loss": 0.7629, "step": 6986 }, { "epoch": 0.4, "learning_rate": 0.00013749472504578843, "loss": 0.8108, "step": 6987 }, { "epoch": 0.4, "learning_rate": 0.00013747769282617504, "loss": 0.7453, "step": 6988 }, { "epoch": 0.4, "learning_rate": 0.00013746065934159123, "loss": 0.6688, "step": 6989 }, { "epoch": 0.4, "learning_rate": 0.00013744362459261188, "loss": 0.7206, "step": 6990 }, { "epoch": 0.4, "learning_rate": 0.00013742658857981204, "loss": 0.6857, "step": 6991 }, { "epoch": 0.4, "learning_rate": 0.0001374095513037667, "loss": 0.6201, "step": 6992 }, { "epoch": 0.4, "learning_rate": 0.00013739251276505084, "loss": 0.7198, "step": 6993 }, { "epoch": 0.4, "learning_rate": 0.00013737547296423965, "loss": 0.7293, "step": 6994 }, { "epoch": 0.4, "learning_rate": 0.0001373584319019082, "loss": 0.8436, "step": 6995 }, { "epoch": 0.4, "learning_rate": 0.00013734138957863175, "loss": 0.7123, "step": 6996 }, { "epoch": 0.4, "learning_rate": 0.0001373243459949854, "loss": 0.772, "step": 6997 }, { "epoch": 0.4, "learning_rate": 0.00013730730115154454, "loss": 0.7453, "step": 6998 }, { "epoch": 0.4, "learning_rate": 0.0001372902550488844, "loss": 0.6889, "step": 6999 }, { "epoch": 0.4, "learning_rate": 0.00013727320768758033, "loss": 0.7954, "step": 7000 }, { "epoch": 0.4, "learning_rate": 0.00013725615906820776, "loss": 0.7047, "step": 7001 }, { "epoch": 0.4, "learning_rate": 0.00013723910919134216, "loss": 0.7172, "step": 7002 }, { "epoch": 0.4, "learning_rate": 0.00013722205805755892, "loss": 0.798, "step": 7003 }, { "epoch": 0.4, "learning_rate": 0.00013720500566743362, "loss": 0.6876, "step": 7004 }, { "epoch": 0.4, "learning_rate": 0.0001371879520215418, "loss": 0.7434, "step": 7005 }, { "epoch": 0.4, "learning_rate": 0.00013717089712045905, "loss": 0.7589, "step": 7006 }, { "epoch": 0.4, "learning_rate": 0.00013715384096476106, "loss": 0.7788, "step": 7007 }, { "epoch": 0.4, "learning_rate": 0.00013713678355502351, "loss": 0.7453, "step": 7008 }, { "epoch": 0.4, "learning_rate": 0.00013711972489182208, "loss": 0.7321, "step": 7009 }, { "epoch": 0.4, "learning_rate": 0.00013710266497573258, "loss": 0.7286, "step": 7010 }, { "epoch": 0.4, "learning_rate": 0.0001370856038073309, "loss": 0.7139, "step": 7011 }, { "epoch": 0.4, "learning_rate": 0.0001370685413871928, "loss": 0.7414, "step": 7012 }, { "epoch": 0.4, "learning_rate": 0.00013705147771589417, "loss": 0.7106, "step": 7013 }, { "epoch": 0.4, "learning_rate": 0.00013703441279401104, "loss": 0.7637, "step": 7014 }, { "epoch": 0.4, "learning_rate": 0.00013701734662211935, "loss": 0.6738, "step": 7015 }, { "epoch": 0.4, "learning_rate": 0.00013700027920079513, "loss": 0.8143, "step": 7016 }, { "epoch": 0.4, "learning_rate": 0.0001369832105306144, "loss": 0.7508, "step": 7017 }, { "epoch": 0.4, "learning_rate": 0.0001369661406121534, "loss": 0.8059, "step": 7018 }, { "epoch": 0.4, "learning_rate": 0.00013694906944598816, "loss": 0.761, "step": 7019 }, { "epoch": 0.4, "learning_rate": 0.00013693199703269492, "loss": 0.7462, "step": 7020 }, { "epoch": 0.4, "learning_rate": 0.00013691492337284994, "loss": 0.7472, "step": 7021 }, { "epoch": 0.4, "learning_rate": 0.0001368978484670295, "loss": 0.7467, "step": 7022 }, { "epoch": 0.4, "learning_rate": 0.00013688077231580986, "loss": 0.7374, "step": 7023 }, { "epoch": 0.4, "learning_rate": 0.00013686369491976746, "loss": 0.7427, "step": 7024 }, { "epoch": 0.4, "learning_rate": 0.0001368466162794787, "loss": 0.7074, "step": 7025 }, { "epoch": 0.4, "learning_rate": 0.00013682953639551997, "loss": 0.7618, "step": 7026 }, { "epoch": 0.4, "learning_rate": 0.00013681245526846783, "loss": 0.7297, "step": 7027 }, { "epoch": 0.4, "learning_rate": 0.00013679537289889874, "loss": 0.7988, "step": 7028 }, { "epoch": 0.4, "learning_rate": 0.00013677828928738934, "loss": 0.6165, "step": 7029 }, { "epoch": 0.4, "learning_rate": 0.00013676120443451622, "loss": 0.6764, "step": 7030 }, { "epoch": 0.4, "learning_rate": 0.00013674411834085603, "loss": 0.6952, "step": 7031 }, { "epoch": 0.4, "learning_rate": 0.00013672703100698548, "loss": 0.784, "step": 7032 }, { "epoch": 0.4, "learning_rate": 0.00013670994243348132, "loss": 0.7611, "step": 7033 }, { "epoch": 0.4, "learning_rate": 0.00013669285262092034, "loss": 0.7539, "step": 7034 }, { "epoch": 0.4, "learning_rate": 0.00013667576156987935, "loss": 0.8272, "step": 7035 }, { "epoch": 0.4, "learning_rate": 0.0001366586692809352, "loss": 0.6673, "step": 7036 }, { "epoch": 0.4, "learning_rate": 0.00013664157575466485, "loss": 0.7084, "step": 7037 }, { "epoch": 0.4, "learning_rate": 0.0001366244809916452, "loss": 0.6936, "step": 7038 }, { "epoch": 0.4, "learning_rate": 0.00013660738499245324, "loss": 0.691, "step": 7039 }, { "epoch": 0.4, "learning_rate": 0.0001365902877576661, "loss": 0.761, "step": 7040 }, { "epoch": 0.4, "learning_rate": 0.00013657318928786074, "loss": 0.6951, "step": 7041 }, { "epoch": 0.4, "learning_rate": 0.00013655608958361432, "loss": 0.709, "step": 7042 }, { "epoch": 0.4, "learning_rate": 0.00013653898864550402, "loss": 0.902, "step": 7043 }, { "epoch": 0.4, "learning_rate": 0.00013652188647410705, "loss": 0.758, "step": 7044 }, { "epoch": 0.4, "learning_rate": 0.00013650478307000057, "loss": 0.6925, "step": 7045 }, { "epoch": 0.4, "learning_rate": 0.00013648767843376196, "loss": 0.6741, "step": 7046 }, { "epoch": 0.4, "learning_rate": 0.00013647057256596851, "loss": 0.6687, "step": 7047 }, { "epoch": 0.4, "learning_rate": 0.0001364534654671976, "loss": 0.8068, "step": 7048 }, { "epoch": 0.4, "learning_rate": 0.0001364363571380266, "loss": 0.6856, "step": 7049 }, { "epoch": 0.4, "learning_rate": 0.00013641924757903302, "loss": 0.7564, "step": 7050 }, { "epoch": 0.4, "learning_rate": 0.0001364021367907943, "loss": 0.7169, "step": 7051 }, { "epoch": 0.4, "learning_rate": 0.00013638502477388798, "loss": 0.7199, "step": 7052 }, { "epoch": 0.4, "learning_rate": 0.0001363679115288917, "loss": 0.785, "step": 7053 }, { "epoch": 0.4, "learning_rate": 0.00013635079705638298, "loss": 0.7465, "step": 7054 }, { "epoch": 0.4, "learning_rate": 0.00013633368135693955, "loss": 0.7578, "step": 7055 }, { "epoch": 0.4, "learning_rate": 0.0001363165644311391, "loss": 0.7322, "step": 7056 }, { "epoch": 0.4, "learning_rate": 0.00013629944627955933, "loss": 0.7119, "step": 7057 }, { "epoch": 0.4, "learning_rate": 0.00013628232690277803, "loss": 0.6656, "step": 7058 }, { "epoch": 0.4, "learning_rate": 0.0001362652063013731, "loss": 0.6936, "step": 7059 }, { "epoch": 0.4, "learning_rate": 0.00013624808447592232, "loss": 0.6871, "step": 7060 }, { "epoch": 0.4, "learning_rate": 0.0001362309614270036, "loss": 0.7657, "step": 7061 }, { "epoch": 0.4, "learning_rate": 0.00013621383715519496, "loss": 0.7856, "step": 7062 }, { "epoch": 0.4, "learning_rate": 0.0001361967116610743, "loss": 0.7485, "step": 7063 }, { "epoch": 0.4, "learning_rate": 0.0001361795849452197, "loss": 0.7137, "step": 7064 }, { "epoch": 0.4, "learning_rate": 0.00013616245700820922, "loss": 0.6079, "step": 7065 }, { "epoch": 0.4, "learning_rate": 0.000136145327850621, "loss": 0.7565, "step": 7066 }, { "epoch": 0.4, "learning_rate": 0.0001361281974730332, "loss": 0.7234, "step": 7067 }, { "epoch": 0.4, "learning_rate": 0.0001361110658760239, "loss": 0.714, "step": 7068 }, { "epoch": 0.4, "learning_rate": 0.0001360939330601715, "loss": 0.8065, "step": 7069 }, { "epoch": 0.4, "learning_rate": 0.00013607679902605417, "loss": 0.6789, "step": 7070 }, { "epoch": 0.4, "learning_rate": 0.00013605966377425022, "loss": 0.7984, "step": 7071 }, { "epoch": 0.4, "learning_rate": 0.0001360425273053381, "loss": 0.7176, "step": 7072 }, { "epoch": 0.4, "learning_rate": 0.00013602538961989612, "loss": 0.8533, "step": 7073 }, { "epoch": 0.4, "learning_rate": 0.0001360082507185028, "loss": 0.6279, "step": 7074 }, { "epoch": 0.4, "learning_rate": 0.00013599111060173656, "loss": 0.6877, "step": 7075 }, { "epoch": 0.4, "learning_rate": 0.00013597396927017593, "loss": 0.7426, "step": 7076 }, { "epoch": 0.4, "learning_rate": 0.00013595682672439953, "loss": 0.7485, "step": 7077 }, { "epoch": 0.4, "learning_rate": 0.0001359396829649859, "loss": 0.7089, "step": 7078 }, { "epoch": 0.4, "learning_rate": 0.00013592253799251376, "loss": 0.6377, "step": 7079 }, { "epoch": 0.4, "learning_rate": 0.00013590539180756168, "loss": 0.6671, "step": 7080 }, { "epoch": 0.4, "learning_rate": 0.00013588824441070852, "loss": 0.7346, "step": 7081 }, { "epoch": 0.4, "learning_rate": 0.00013587109580253295, "loss": 0.7694, "step": 7082 }, { "epoch": 0.4, "learning_rate": 0.00013585394598361386, "loss": 0.6818, "step": 7083 }, { "epoch": 0.4, "learning_rate": 0.00013583679495453, "loss": 0.7128, "step": 7084 }, { "epoch": 0.4, "learning_rate": 0.0001358196427158604, "loss": 0.7662, "step": 7085 }, { "epoch": 0.4, "learning_rate": 0.0001358024892681839, "loss": 0.6202, "step": 7086 }, { "epoch": 0.4, "learning_rate": 0.00013578533461207947, "loss": 0.7486, "step": 7087 }, { "epoch": 0.4, "learning_rate": 0.00013576817874812614, "loss": 0.7178, "step": 7088 }, { "epoch": 0.4, "learning_rate": 0.000135751021676903, "loss": 0.7488, "step": 7089 }, { "epoch": 0.4, "learning_rate": 0.0001357338633989891, "loss": 0.7316, "step": 7090 }, { "epoch": 0.4, "learning_rate": 0.0001357167039149636, "loss": 0.6891, "step": 7091 }, { "epoch": 0.4, "learning_rate": 0.0001356995432254057, "loss": 0.7961, "step": 7092 }, { "epoch": 0.4, "learning_rate": 0.00013568238133089455, "loss": 0.7409, "step": 7093 }, { "epoch": 0.4, "learning_rate": 0.00013566521823200943, "loss": 0.7637, "step": 7094 }, { "epoch": 0.4, "learning_rate": 0.0001356480539293297, "loss": 0.7492, "step": 7095 }, { "epoch": 0.4, "learning_rate": 0.00013563088842343465, "loss": 0.6313, "step": 7096 }, { "epoch": 0.4, "learning_rate": 0.00013561372171490366, "loss": 0.819, "step": 7097 }, { "epoch": 0.4, "learning_rate": 0.0001355965538043162, "loss": 0.692, "step": 7098 }, { "epoch": 0.4, "learning_rate": 0.00013557938469225167, "loss": 0.7592, "step": 7099 }, { "epoch": 0.4, "learning_rate": 0.00013556221437928956, "loss": 0.6349, "step": 7100 }, { "epoch": 0.4, "learning_rate": 0.00013554504286600948, "loss": 0.7558, "step": 7101 }, { "epoch": 0.4, "learning_rate": 0.000135527870152991, "loss": 0.7327, "step": 7102 }, { "epoch": 0.4, "learning_rate": 0.0001355106962408137, "loss": 0.799, "step": 7103 }, { "epoch": 0.4, "learning_rate": 0.00013549352113005728, "loss": 0.7971, "step": 7104 }, { "epoch": 0.4, "learning_rate": 0.00013547634482130144, "loss": 0.7867, "step": 7105 }, { "epoch": 0.4, "learning_rate": 0.00013545916731512594, "loss": 0.6985, "step": 7106 }, { "epoch": 0.4, "learning_rate": 0.00013544198861211053, "loss": 0.8077, "step": 7107 }, { "epoch": 0.4, "learning_rate": 0.00013542480871283507, "loss": 0.6889, "step": 7108 }, { "epoch": 0.4, "learning_rate": 0.00013540762761787937, "loss": 0.6899, "step": 7109 }, { "epoch": 0.4, "learning_rate": 0.0001353904453278234, "loss": 0.7438, "step": 7110 }, { "epoch": 0.4, "learning_rate": 0.0001353732618432471, "loss": 0.6574, "step": 7111 }, { "epoch": 0.4, "learning_rate": 0.00013535607716473044, "loss": 0.6323, "step": 7112 }, { "epoch": 0.4, "learning_rate": 0.00013533889129285342, "loss": 0.8773, "step": 7113 }, { "epoch": 0.4, "learning_rate": 0.0001353217042281962, "loss": 0.6784, "step": 7114 }, { "epoch": 0.4, "learning_rate": 0.00013530451597133878, "loss": 0.7294, "step": 7115 }, { "epoch": 0.4, "learning_rate": 0.00013528732652286137, "loss": 0.7288, "step": 7116 }, { "epoch": 0.4, "learning_rate": 0.00013527013588334415, "loss": 0.7483, "step": 7117 }, { "epoch": 0.4, "learning_rate": 0.0001352529440533673, "loss": 0.7616, "step": 7118 }, { "epoch": 0.4, "learning_rate": 0.0001352357510335112, "loss": 0.6904, "step": 7119 }, { "epoch": 0.4, "learning_rate": 0.000135218556824356, "loss": 0.714, "step": 7120 }, { "epoch": 0.4, "learning_rate": 0.00013520136142648224, "loss": 0.6877, "step": 7121 }, { "epoch": 0.4, "learning_rate": 0.00013518416484047018, "loss": 0.7087, "step": 7122 }, { "epoch": 0.4, "learning_rate": 0.0001351669670669003, "loss": 0.7305, "step": 7123 }, { "epoch": 0.4, "learning_rate": 0.00013514976810635303, "loss": 0.7191, "step": 7124 }, { "epoch": 0.4, "learning_rate": 0.0001351325679594089, "loss": 0.7065, "step": 7125 }, { "epoch": 0.4, "learning_rate": 0.00013511536662664847, "loss": 0.6102, "step": 7126 }, { "epoch": 0.4, "learning_rate": 0.00013509816410865234, "loss": 0.7856, "step": 7127 }, { "epoch": 0.4, "learning_rate": 0.00013508096040600108, "loss": 0.6763, "step": 7128 }, { "epoch": 0.4, "learning_rate": 0.00013506375551927547, "loss": 0.7772, "step": 7129 }, { "epoch": 0.4, "learning_rate": 0.00013504654944905607, "loss": 0.7648, "step": 7130 }, { "epoch": 0.4, "learning_rate": 0.00013502934219592378, "loss": 0.7535, "step": 7131 }, { "epoch": 0.4, "learning_rate": 0.00013501213376045928, "loss": 0.7058, "step": 7132 }, { "epoch": 0.4, "learning_rate": 0.00013499492414324347, "loss": 0.6539, "step": 7133 }, { "epoch": 0.4, "learning_rate": 0.00013497771334485722, "loss": 0.6808, "step": 7134 }, { "epoch": 0.4, "learning_rate": 0.00013496050136588134, "loss": 0.6681, "step": 7135 }, { "epoch": 0.4, "learning_rate": 0.0001349432882068969, "loss": 0.7059, "step": 7136 }, { "epoch": 0.4, "learning_rate": 0.00013492607386848484, "loss": 0.6846, "step": 7137 }, { "epoch": 0.4, "learning_rate": 0.00013490885835122619, "loss": 0.7257, "step": 7138 }, { "epoch": 0.4, "learning_rate": 0.00013489164165570195, "loss": 0.8025, "step": 7139 }, { "epoch": 0.41, "learning_rate": 0.0001348744237824934, "loss": 0.7144, "step": 7140 }, { "epoch": 0.41, "learning_rate": 0.00013485720473218154, "loss": 0.8255, "step": 7141 }, { "epoch": 0.41, "learning_rate": 0.0001348399845053476, "loss": 0.7259, "step": 7142 }, { "epoch": 0.41, "learning_rate": 0.00013482276310257282, "loss": 0.6136, "step": 7143 }, { "epoch": 0.41, "learning_rate": 0.00013480554052443846, "loss": 0.665, "step": 7144 }, { "epoch": 0.41, "learning_rate": 0.0001347883167715258, "loss": 0.8342, "step": 7145 }, { "epoch": 0.41, "learning_rate": 0.00013477109184441624, "loss": 0.6994, "step": 7146 }, { "epoch": 0.41, "learning_rate": 0.00013475386574369113, "loss": 0.704, "step": 7147 }, { "epoch": 0.41, "learning_rate": 0.00013473663846993192, "loss": 0.7578, "step": 7148 }, { "epoch": 0.41, "learning_rate": 0.00013471941002372005, "loss": 0.7693, "step": 7149 }, { "epoch": 0.41, "learning_rate": 0.00013470218040563704, "loss": 0.7728, "step": 7150 }, { "epoch": 0.41, "learning_rate": 0.00013468494961626446, "loss": 0.7388, "step": 7151 }, { "epoch": 0.41, "learning_rate": 0.00013466771765618382, "loss": 0.6928, "step": 7152 }, { "epoch": 0.41, "learning_rate": 0.00013465048452597682, "loss": 0.7465, "step": 7153 }, { "epoch": 0.41, "learning_rate": 0.00013463325022622507, "loss": 0.7355, "step": 7154 }, { "epoch": 0.41, "learning_rate": 0.00013461601475751032, "loss": 0.7435, "step": 7155 }, { "epoch": 0.41, "learning_rate": 0.00013459877812041425, "loss": 0.7655, "step": 7156 }, { "epoch": 0.41, "learning_rate": 0.00013458154031551872, "loss": 0.7009, "step": 7157 }, { "epoch": 0.41, "learning_rate": 0.00013456430134340546, "loss": 0.6358, "step": 7158 }, { "epoch": 0.41, "learning_rate": 0.00013454706120465644, "loss": 0.7064, "step": 7159 }, { "epoch": 0.41, "learning_rate": 0.00013452981989985348, "loss": 0.6873, "step": 7160 }, { "epoch": 0.41, "learning_rate": 0.00013451257742957853, "loss": 0.5966, "step": 7161 }, { "epoch": 0.41, "learning_rate": 0.0001344953337944136, "loss": 0.7059, "step": 7162 }, { "epoch": 0.41, "learning_rate": 0.00013447808899494067, "loss": 0.6525, "step": 7163 }, { "epoch": 0.41, "learning_rate": 0.0001344608430317418, "loss": 0.7724, "step": 7164 }, { "epoch": 0.41, "learning_rate": 0.00013444359590539912, "loss": 0.7188, "step": 7165 }, { "epoch": 0.41, "learning_rate": 0.00013442634761649477, "loss": 0.8738, "step": 7166 }, { "epoch": 0.41, "learning_rate": 0.00013440909816561086, "loss": 0.7769, "step": 7167 }, { "epoch": 0.41, "learning_rate": 0.00013439184755332966, "loss": 0.7264, "step": 7168 }, { "epoch": 0.41, "learning_rate": 0.00013437459578023343, "loss": 0.7252, "step": 7169 }, { "epoch": 0.41, "learning_rate": 0.00013435734284690442, "loss": 0.7925, "step": 7170 }, { "epoch": 0.41, "learning_rate": 0.000134340088753925, "loss": 0.6007, "step": 7171 }, { "epoch": 0.41, "learning_rate": 0.0001343228335018775, "loss": 0.6805, "step": 7172 }, { "epoch": 0.41, "learning_rate": 0.00013430557709134434, "loss": 0.6029, "step": 7173 }, { "epoch": 0.41, "learning_rate": 0.00013428831952290804, "loss": 0.6794, "step": 7174 }, { "epoch": 0.41, "learning_rate": 0.000134271060797151, "loss": 0.684, "step": 7175 }, { "epoch": 0.41, "learning_rate": 0.00013425380091465575, "loss": 0.7388, "step": 7176 }, { "epoch": 0.41, "learning_rate": 0.00013423653987600497, "loss": 0.616, "step": 7177 }, { "epoch": 0.41, "learning_rate": 0.0001342192776817811, "loss": 0.8063, "step": 7178 }, { "epoch": 0.41, "learning_rate": 0.00013420201433256689, "loss": 0.7214, "step": 7179 }, { "epoch": 0.41, "learning_rate": 0.00013418474982894498, "loss": 0.6453, "step": 7180 }, { "epoch": 0.41, "learning_rate": 0.00013416748417149815, "loss": 0.7799, "step": 7181 }, { "epoch": 0.41, "learning_rate": 0.0001341502173608091, "loss": 0.642, "step": 7182 }, { "epoch": 0.41, "learning_rate": 0.00013413294939746063, "loss": 0.7644, "step": 7183 }, { "epoch": 0.41, "learning_rate": 0.00013411568028203562, "loss": 0.7984, "step": 7184 }, { "epoch": 0.41, "learning_rate": 0.00013409841001511694, "loss": 0.681, "step": 7185 }, { "epoch": 0.41, "learning_rate": 0.00013408113859728752, "loss": 0.7631, "step": 7186 }, { "epoch": 0.41, "learning_rate": 0.00013406386602913024, "loss": 0.6873, "step": 7187 }, { "epoch": 0.41, "learning_rate": 0.0001340465923112282, "loss": 0.7084, "step": 7188 }, { "epoch": 0.41, "learning_rate": 0.00013402931744416433, "loss": 0.7159, "step": 7189 }, { "epoch": 0.41, "learning_rate": 0.00013401204142852175, "loss": 0.6655, "step": 7190 }, { "epoch": 0.41, "learning_rate": 0.00013399476426488358, "loss": 0.6957, "step": 7191 }, { "epoch": 0.41, "learning_rate": 0.000133977485953833, "loss": 0.7457, "step": 7192 }, { "epoch": 0.41, "learning_rate": 0.00013396020649595312, "loss": 0.6141, "step": 7193 }, { "epoch": 0.41, "learning_rate": 0.0001339429258918272, "loss": 0.73, "step": 7194 }, { "epoch": 0.41, "learning_rate": 0.00013392564414203855, "loss": 0.7044, "step": 7195 }, { "epoch": 0.41, "learning_rate": 0.0001339083612471704, "loss": 0.6735, "step": 7196 }, { "epoch": 0.41, "learning_rate": 0.00013389107720780614, "loss": 0.7939, "step": 7197 }, { "epoch": 0.41, "learning_rate": 0.00013387379202452917, "loss": 0.7197, "step": 7198 }, { "epoch": 0.41, "learning_rate": 0.00013385650569792284, "loss": 0.7474, "step": 7199 }, { "epoch": 0.41, "learning_rate": 0.00013383921822857068, "loss": 0.7994, "step": 7200 }, { "epoch": 0.41, "learning_rate": 0.00013382192961705616, "loss": 0.709, "step": 7201 }, { "epoch": 0.41, "learning_rate": 0.0001338046398639628, "loss": 0.7204, "step": 7202 }, { "epoch": 0.41, "learning_rate": 0.00013378734896987417, "loss": 0.7409, "step": 7203 }, { "epoch": 0.41, "learning_rate": 0.0001337700569353739, "loss": 0.6679, "step": 7204 }, { "epoch": 0.41, "learning_rate": 0.00013375276376104568, "loss": 0.7562, "step": 7205 }, { "epoch": 0.41, "learning_rate": 0.00013373546944747314, "loss": 0.7038, "step": 7206 }, { "epoch": 0.41, "learning_rate": 0.00013371817399524005, "loss": 0.7511, "step": 7207 }, { "epoch": 0.41, "learning_rate": 0.00013370087740493015, "loss": 0.7631, "step": 7208 }, { "epoch": 0.41, "learning_rate": 0.00013368357967712726, "loss": 0.718, "step": 7209 }, { "epoch": 0.41, "learning_rate": 0.00013366628081241518, "loss": 0.6397, "step": 7210 }, { "epoch": 0.41, "learning_rate": 0.00013364898081137788, "loss": 0.702, "step": 7211 }, { "epoch": 0.41, "learning_rate": 0.00013363167967459923, "loss": 0.6794, "step": 7212 }, { "epoch": 0.41, "learning_rate": 0.00013361437740266313, "loss": 0.6997, "step": 7213 }, { "epoch": 0.41, "learning_rate": 0.00013359707399615372, "loss": 0.6858, "step": 7214 }, { "epoch": 0.41, "learning_rate": 0.0001335797694556549, "loss": 0.7628, "step": 7215 }, { "epoch": 0.41, "learning_rate": 0.0001335624637817508, "loss": 0.7604, "step": 7216 }, { "epoch": 0.41, "learning_rate": 0.00013354515697502553, "loss": 0.7534, "step": 7217 }, { "epoch": 0.41, "learning_rate": 0.00013352784903606323, "loss": 0.6702, "step": 7218 }, { "epoch": 0.41, "learning_rate": 0.0001335105399654481, "loss": 0.694, "step": 7219 }, { "epoch": 0.41, "learning_rate": 0.00013349322976376432, "loss": 0.6317, "step": 7220 }, { "epoch": 0.41, "learning_rate": 0.00013347591843159627, "loss": 0.6883, "step": 7221 }, { "epoch": 0.41, "learning_rate": 0.00013345860596952815, "loss": 0.7313, "step": 7222 }, { "epoch": 0.41, "learning_rate": 0.0001334412923781443, "loss": 0.6868, "step": 7223 }, { "epoch": 0.41, "learning_rate": 0.00013342397765802916, "loss": 0.8354, "step": 7224 }, { "epoch": 0.41, "learning_rate": 0.00013340666180976712, "loss": 0.7387, "step": 7225 }, { "epoch": 0.41, "learning_rate": 0.0001333893448339426, "loss": 0.8173, "step": 7226 }, { "epoch": 0.41, "learning_rate": 0.00013337202673114018, "loss": 0.7794, "step": 7227 }, { "epoch": 0.41, "learning_rate": 0.0001333547075019443, "loss": 0.7803, "step": 7228 }, { "epoch": 0.41, "learning_rate": 0.00013333738714693956, "loss": 0.6314, "step": 7229 }, { "epoch": 0.41, "learning_rate": 0.00013332006566671057, "loss": 0.6739, "step": 7230 }, { "epoch": 0.41, "learning_rate": 0.000133302743061842, "loss": 0.882, "step": 7231 }, { "epoch": 0.41, "learning_rate": 0.00013328541933291848, "loss": 0.7095, "step": 7232 }, { "epoch": 0.41, "learning_rate": 0.00013326809448052483, "loss": 0.6986, "step": 7233 }, { "epoch": 0.41, "learning_rate": 0.0001332507685052457, "loss": 0.6706, "step": 7234 }, { "epoch": 0.41, "learning_rate": 0.00013323344140766592, "loss": 0.7387, "step": 7235 }, { "epoch": 0.41, "learning_rate": 0.00013321611318837032, "loss": 0.7417, "step": 7236 }, { "epoch": 0.41, "learning_rate": 0.00013319878384794384, "loss": 0.7594, "step": 7237 }, { "epoch": 0.41, "learning_rate": 0.0001331814533869713, "loss": 0.7316, "step": 7238 }, { "epoch": 0.41, "learning_rate": 0.00013316412180603768, "loss": 0.6896, "step": 7239 }, { "epoch": 0.41, "learning_rate": 0.000133146789105728, "loss": 0.7815, "step": 7240 }, { "epoch": 0.41, "learning_rate": 0.00013312945528662725, "loss": 0.7552, "step": 7241 }, { "epoch": 0.41, "learning_rate": 0.0001331121203493205, "loss": 0.6934, "step": 7242 }, { "epoch": 0.41, "learning_rate": 0.00013309478429439283, "loss": 0.7695, "step": 7243 }, { "epoch": 0.41, "learning_rate": 0.0001330774471224294, "loss": 0.7888, "step": 7244 }, { "epoch": 0.41, "learning_rate": 0.0001330601088340154, "loss": 0.6981, "step": 7245 }, { "epoch": 0.41, "learning_rate": 0.000133042769429736, "loss": 0.7303, "step": 7246 }, { "epoch": 0.41, "learning_rate": 0.00013302542891017652, "loss": 0.7805, "step": 7247 }, { "epoch": 0.41, "learning_rate": 0.00013300808727592216, "loss": 0.7521, "step": 7248 }, { "epoch": 0.41, "learning_rate": 0.0001329907445275583, "loss": 0.8182, "step": 7249 }, { "epoch": 0.41, "learning_rate": 0.0001329734006656703, "loss": 0.7328, "step": 7250 }, { "epoch": 0.41, "learning_rate": 0.00013295605569084353, "loss": 0.8658, "step": 7251 }, { "epoch": 0.41, "learning_rate": 0.00013293870960366346, "loss": 0.6282, "step": 7252 }, { "epoch": 0.41, "learning_rate": 0.00013292136240471557, "loss": 0.7235, "step": 7253 }, { "epoch": 0.41, "learning_rate": 0.00013290401409458532, "loss": 0.6151, "step": 7254 }, { "epoch": 0.41, "learning_rate": 0.00013288666467385833, "loss": 0.7777, "step": 7255 }, { "epoch": 0.41, "learning_rate": 0.00013286931414312019, "loss": 0.7239, "step": 7256 }, { "epoch": 0.41, "learning_rate": 0.00013285196250295642, "loss": 0.7909, "step": 7257 }, { "epoch": 0.41, "learning_rate": 0.00013283460975395278, "loss": 0.7557, "step": 7258 }, { "epoch": 0.41, "learning_rate": 0.00013281725589669497, "loss": 0.7752, "step": 7259 }, { "epoch": 0.41, "learning_rate": 0.00013279990093176874, "loss": 0.7514, "step": 7260 }, { "epoch": 0.41, "learning_rate": 0.00013278254485975976, "loss": 0.5943, "step": 7261 }, { "epoch": 0.41, "learning_rate": 0.00013276518768125397, "loss": 0.6644, "step": 7262 }, { "epoch": 0.41, "learning_rate": 0.00013274782939683716, "loss": 0.7311, "step": 7263 }, { "epoch": 0.41, "learning_rate": 0.00013273047000709519, "loss": 0.7427, "step": 7264 }, { "epoch": 0.41, "learning_rate": 0.00013271310951261405, "loss": 0.7218, "step": 7265 }, { "epoch": 0.41, "learning_rate": 0.00013269574791397967, "loss": 0.7346, "step": 7266 }, { "epoch": 0.41, "learning_rate": 0.000132678385211778, "loss": 0.7563, "step": 7267 }, { "epoch": 0.41, "learning_rate": 0.00013266102140659517, "loss": 0.773, "step": 7268 }, { "epoch": 0.41, "learning_rate": 0.00013264365649901723, "loss": 0.7472, "step": 7269 }, { "epoch": 0.41, "learning_rate": 0.00013262629048963026, "loss": 0.6964, "step": 7270 }, { "epoch": 0.41, "learning_rate": 0.0001326089233790204, "loss": 0.7675, "step": 7271 }, { "epoch": 0.41, "learning_rate": 0.00013259155516777389, "loss": 0.7743, "step": 7272 }, { "epoch": 0.41, "learning_rate": 0.00013257418585647688, "loss": 0.6138, "step": 7273 }, { "epoch": 0.41, "learning_rate": 0.00013255681544571568, "loss": 0.7575, "step": 7274 }, { "epoch": 0.41, "learning_rate": 0.00013253944393607658, "loss": 0.5818, "step": 7275 }, { "epoch": 0.41, "learning_rate": 0.0001325220713281459, "loss": 0.7622, "step": 7276 }, { "epoch": 0.41, "learning_rate": 0.00013250469762251003, "loss": 0.748, "step": 7277 }, { "epoch": 0.41, "learning_rate": 0.00013248732281975536, "loss": 0.7186, "step": 7278 }, { "epoch": 0.41, "learning_rate": 0.00013246994692046836, "loss": 0.7286, "step": 7279 }, { "epoch": 0.41, "learning_rate": 0.00013245256992523548, "loss": 0.7142, "step": 7280 }, { "epoch": 0.41, "learning_rate": 0.00013243519183464328, "loss": 0.7013, "step": 7281 }, { "epoch": 0.41, "learning_rate": 0.0001324178126492783, "loss": 0.8184, "step": 7282 }, { "epoch": 0.41, "learning_rate": 0.00013240043236972706, "loss": 0.7531, "step": 7283 }, { "epoch": 0.41, "learning_rate": 0.00013238305099657626, "loss": 0.6386, "step": 7284 }, { "epoch": 0.41, "learning_rate": 0.0001323656685304126, "loss": 0.7787, "step": 7285 }, { "epoch": 0.41, "learning_rate": 0.00013234828497182276, "loss": 0.5985, "step": 7286 }, { "epoch": 0.41, "learning_rate": 0.00013233090032139345, "loss": 0.6516, "step": 7287 }, { "epoch": 0.41, "learning_rate": 0.00013231351457971145, "loss": 0.6885, "step": 7288 }, { "epoch": 0.41, "learning_rate": 0.00013229612774736359, "loss": 0.6209, "step": 7289 }, { "epoch": 0.41, "learning_rate": 0.00013227873982493672, "loss": 0.7106, "step": 7290 }, { "epoch": 0.41, "learning_rate": 0.00013226135081301771, "loss": 0.7938, "step": 7291 }, { "epoch": 0.41, "learning_rate": 0.00013224396071219357, "loss": 0.6363, "step": 7292 }, { "epoch": 0.41, "learning_rate": 0.00013222656952305113, "loss": 0.6648, "step": 7293 }, { "epoch": 0.41, "learning_rate": 0.00013220917724617745, "loss": 0.7668, "step": 7294 }, { "epoch": 0.41, "learning_rate": 0.00013219178388215958, "loss": 0.7492, "step": 7295 }, { "epoch": 0.41, "learning_rate": 0.0001321743894315846, "loss": 0.7016, "step": 7296 }, { "epoch": 0.41, "learning_rate": 0.00013215699389503954, "loss": 0.7298, "step": 7297 }, { "epoch": 0.41, "learning_rate": 0.00013213959727311167, "loss": 0.7721, "step": 7298 }, { "epoch": 0.41, "learning_rate": 0.00013212219956638808, "loss": 0.754, "step": 7299 }, { "epoch": 0.41, "learning_rate": 0.000132104800775456, "loss": 0.7554, "step": 7300 }, { "epoch": 0.41, "learning_rate": 0.00013208740090090267, "loss": 0.6694, "step": 7301 }, { "epoch": 0.41, "learning_rate": 0.00013206999994331544, "loss": 0.756, "step": 7302 }, { "epoch": 0.41, "learning_rate": 0.00013205259790328162, "loss": 0.7674, "step": 7303 }, { "epoch": 0.41, "learning_rate": 0.00013203519478138852, "loss": 0.6579, "step": 7304 }, { "epoch": 0.41, "learning_rate": 0.0001320177905782236, "loss": 0.642, "step": 7305 }, { "epoch": 0.41, "learning_rate": 0.00013200038529437427, "loss": 0.8196, "step": 7306 }, { "epoch": 0.41, "learning_rate": 0.00013198297893042804, "loss": 0.6297, "step": 7307 }, { "epoch": 0.41, "learning_rate": 0.00013196557148697238, "loss": 0.6882, "step": 7308 }, { "epoch": 0.41, "learning_rate": 0.0001319481629645948, "loss": 0.5804, "step": 7309 }, { "epoch": 0.41, "learning_rate": 0.000131930753363883, "loss": 0.5764, "step": 7310 }, { "epoch": 0.41, "learning_rate": 0.00013191334268542452, "loss": 0.7478, "step": 7311 }, { "epoch": 0.41, "learning_rate": 0.00013189593092980702, "loss": 0.7307, "step": 7312 }, { "epoch": 0.41, "learning_rate": 0.00013187851809761818, "loss": 0.6644, "step": 7313 }, { "epoch": 0.41, "learning_rate": 0.00013186110418944577, "loss": 0.6723, "step": 7314 }, { "epoch": 0.41, "learning_rate": 0.00013184368920587754, "loss": 0.7862, "step": 7315 }, { "epoch": 0.41, "learning_rate": 0.00013182627314750126, "loss": 0.7719, "step": 7316 }, { "epoch": 0.42, "learning_rate": 0.00013180885601490482, "loss": 0.695, "step": 7317 }, { "epoch": 0.42, "learning_rate": 0.00013179143780867606, "loss": 0.786, "step": 7318 }, { "epoch": 0.42, "learning_rate": 0.0001317740185294029, "loss": 0.7573, "step": 7319 }, { "epoch": 0.42, "learning_rate": 0.00013175659817767326, "loss": 0.7573, "step": 7320 }, { "epoch": 0.42, "learning_rate": 0.00013173917675407518, "loss": 0.7906, "step": 7321 }, { "epoch": 0.42, "learning_rate": 0.00013172175425919664, "loss": 0.6826, "step": 7322 }, { "epoch": 0.42, "learning_rate": 0.00013170433069362568, "loss": 0.6479, "step": 7323 }, { "epoch": 0.42, "learning_rate": 0.00013168690605795045, "loss": 0.6464, "step": 7324 }, { "epoch": 0.42, "learning_rate": 0.00013166948035275904, "loss": 0.7751, "step": 7325 }, { "epoch": 0.42, "learning_rate": 0.00013165205357863955, "loss": 0.6958, "step": 7326 }, { "epoch": 0.42, "learning_rate": 0.00013163462573618029, "loss": 0.6843, "step": 7327 }, { "epoch": 0.42, "learning_rate": 0.00013161719682596942, "loss": 0.7486, "step": 7328 }, { "epoch": 0.42, "learning_rate": 0.00013159976684859527, "loss": 0.7939, "step": 7329 }, { "epoch": 0.42, "learning_rate": 0.0001315823358046461, "loss": 0.7229, "step": 7330 }, { "epoch": 0.42, "learning_rate": 0.00013156490369471027, "loss": 0.7261, "step": 7331 }, { "epoch": 0.42, "learning_rate": 0.00013154747051937613, "loss": 0.7753, "step": 7332 }, { "epoch": 0.42, "learning_rate": 0.00013153003627923218, "loss": 0.8082, "step": 7333 }, { "epoch": 0.42, "learning_rate": 0.00013151260097486677, "loss": 0.6926, "step": 7334 }, { "epoch": 0.42, "learning_rate": 0.00013149516460686845, "loss": 0.7181, "step": 7335 }, { "epoch": 0.42, "learning_rate": 0.0001314777271758257, "loss": 0.6789, "step": 7336 }, { "epoch": 0.42, "learning_rate": 0.00013146028868232717, "loss": 0.7523, "step": 7337 }, { "epoch": 0.42, "learning_rate": 0.0001314428491269613, "loss": 0.7206, "step": 7338 }, { "epoch": 0.42, "learning_rate": 0.00013142540851031687, "loss": 0.6763, "step": 7339 }, { "epoch": 0.42, "learning_rate": 0.0001314079668329825, "loss": 0.7037, "step": 7340 }, { "epoch": 0.42, "learning_rate": 0.00013139052409554683, "loss": 0.7588, "step": 7341 }, { "epoch": 0.42, "learning_rate": 0.00013137308029859864, "loss": 0.7167, "step": 7342 }, { "epoch": 0.42, "learning_rate": 0.00013135563544272675, "loss": 0.6974, "step": 7343 }, { "epoch": 0.42, "learning_rate": 0.00013133818952851996, "loss": 0.6696, "step": 7344 }, { "epoch": 0.42, "learning_rate": 0.000131320742556567, "loss": 0.7759, "step": 7345 }, { "epoch": 0.42, "learning_rate": 0.0001313032945274569, "loss": 0.7151, "step": 7346 }, { "epoch": 0.42, "learning_rate": 0.0001312858454417785, "loss": 0.7612, "step": 7347 }, { "epoch": 0.42, "learning_rate": 0.00013126839530012077, "loss": 0.7226, "step": 7348 }, { "epoch": 0.42, "learning_rate": 0.00013125094410307265, "loss": 0.728, "step": 7349 }, { "epoch": 0.42, "learning_rate": 0.00013123349185122327, "loss": 0.7647, "step": 7350 }, { "epoch": 0.42, "learning_rate": 0.0001312160385451616, "loss": 0.6371, "step": 7351 }, { "epoch": 0.42, "learning_rate": 0.00013119858418547678, "loss": 0.7131, "step": 7352 }, { "epoch": 0.42, "learning_rate": 0.0001311811287727579, "loss": 0.7333, "step": 7353 }, { "epoch": 0.42, "learning_rate": 0.00013116367230759415, "loss": 0.7097, "step": 7354 }, { "epoch": 0.42, "learning_rate": 0.00013114621479057474, "loss": 0.7357, "step": 7355 }, { "epoch": 0.42, "learning_rate": 0.00013112875622228889, "loss": 0.757, "step": 7356 }, { "epoch": 0.42, "learning_rate": 0.00013111129660332586, "loss": 0.7146, "step": 7357 }, { "epoch": 0.42, "learning_rate": 0.000131093835934275, "loss": 0.6111, "step": 7358 }, { "epoch": 0.42, "learning_rate": 0.00013107637421572562, "loss": 0.7459, "step": 7359 }, { "epoch": 0.42, "learning_rate": 0.00013105891144826715, "loss": 0.6791, "step": 7360 }, { "epoch": 0.42, "learning_rate": 0.00013104144763248892, "loss": 0.6652, "step": 7361 }, { "epoch": 0.42, "learning_rate": 0.00013102398276898042, "loss": 0.6315, "step": 7362 }, { "epoch": 0.42, "learning_rate": 0.00013100651685833117, "loss": 0.6996, "step": 7363 }, { "epoch": 0.42, "learning_rate": 0.00013098904990113062, "loss": 0.6985, "step": 7364 }, { "epoch": 0.42, "learning_rate": 0.00013097158189796839, "loss": 0.6706, "step": 7365 }, { "epoch": 0.42, "learning_rate": 0.00013095411284943406, "loss": 0.8105, "step": 7366 }, { "epoch": 0.42, "learning_rate": 0.00013093664275611722, "loss": 0.6944, "step": 7367 }, { "epoch": 0.42, "learning_rate": 0.00013091917161860755, "loss": 0.6527, "step": 7368 }, { "epoch": 0.42, "learning_rate": 0.00013090169943749476, "loss": 0.6968, "step": 7369 }, { "epoch": 0.42, "learning_rate": 0.00013088422621336857, "loss": 0.7457, "step": 7370 }, { "epoch": 0.42, "learning_rate": 0.0001308667519468187, "loss": 0.6794, "step": 7371 }, { "epoch": 0.42, "learning_rate": 0.00013084927663843508, "loss": 0.7244, "step": 7372 }, { "epoch": 0.42, "learning_rate": 0.00013083180028880743, "loss": 0.6682, "step": 7373 }, { "epoch": 0.42, "learning_rate": 0.00013081432289852566, "loss": 0.7036, "step": 7374 }, { "epoch": 0.42, "learning_rate": 0.00013079684446817968, "loss": 0.7259, "step": 7375 }, { "epoch": 0.42, "learning_rate": 0.00013077936499835942, "loss": 0.755, "step": 7376 }, { "epoch": 0.42, "learning_rate": 0.00013076188448965493, "loss": 0.8636, "step": 7377 }, { "epoch": 0.42, "learning_rate": 0.0001307444029426561, "loss": 0.8125, "step": 7378 }, { "epoch": 0.42, "learning_rate": 0.00013072692035795305, "loss": 0.7128, "step": 7379 }, { "epoch": 0.42, "learning_rate": 0.00013070943673613585, "loss": 0.8146, "step": 7380 }, { "epoch": 0.42, "learning_rate": 0.00013069195207779464, "loss": 0.7557, "step": 7381 }, { "epoch": 0.42, "learning_rate": 0.00013067446638351953, "loss": 0.7104, "step": 7382 }, { "epoch": 0.42, "learning_rate": 0.00013065697965390077, "loss": 0.8221, "step": 7383 }, { "epoch": 0.42, "learning_rate": 0.0001306394918895285, "loss": 0.7114, "step": 7384 }, { "epoch": 0.42, "learning_rate": 0.0001306220030909931, "loss": 0.7346, "step": 7385 }, { "epoch": 0.42, "learning_rate": 0.0001306045132588847, "loss": 0.7471, "step": 7386 }, { "epoch": 0.42, "learning_rate": 0.00013058702239379376, "loss": 0.6846, "step": 7387 }, { "epoch": 0.42, "learning_rate": 0.00013056953049631057, "loss": 0.5901, "step": 7388 }, { "epoch": 0.42, "learning_rate": 0.0001305520375670256, "loss": 0.6358, "step": 7389 }, { "epoch": 0.42, "learning_rate": 0.0001305345436065292, "loss": 0.7227, "step": 7390 }, { "epoch": 0.42, "learning_rate": 0.00013051704861541185, "loss": 0.6763, "step": 7391 }, { "epoch": 0.42, "learning_rate": 0.00013049955259426414, "loss": 0.6258, "step": 7392 }, { "epoch": 0.42, "learning_rate": 0.0001304820555436765, "loss": 0.6767, "step": 7393 }, { "epoch": 0.42, "learning_rate": 0.00013046455746423954, "loss": 0.6722, "step": 7394 }, { "epoch": 0.42, "learning_rate": 0.00013044705835654386, "loss": 0.677, "step": 7395 }, { "epoch": 0.42, "learning_rate": 0.00013042955822118016, "loss": 0.7063, "step": 7396 }, { "epoch": 0.42, "learning_rate": 0.00013041205705873902, "loss": 0.7084, "step": 7397 }, { "epoch": 0.42, "learning_rate": 0.00013039455486981122, "loss": 0.6599, "step": 7398 }, { "epoch": 0.42, "learning_rate": 0.00013037705165498746, "loss": 0.7715, "step": 7399 }, { "epoch": 0.42, "learning_rate": 0.00013035954741485856, "loss": 0.7629, "step": 7400 }, { "epoch": 0.42, "learning_rate": 0.0001303420421500153, "loss": 0.7067, "step": 7401 }, { "epoch": 0.42, "learning_rate": 0.00013032453586104856, "loss": 0.6779, "step": 7402 }, { "epoch": 0.42, "learning_rate": 0.00013030702854854918, "loss": 0.7421, "step": 7403 }, { "epoch": 0.42, "learning_rate": 0.00013028952021310812, "loss": 0.72, "step": 7404 }, { "epoch": 0.42, "learning_rate": 0.00013027201085531634, "loss": 0.7602, "step": 7405 }, { "epoch": 0.42, "learning_rate": 0.00013025450047576476, "loss": 0.6925, "step": 7406 }, { "epoch": 0.42, "learning_rate": 0.00013023698907504446, "loss": 0.7462, "step": 7407 }, { "epoch": 0.42, "learning_rate": 0.00013021947665374647, "loss": 0.7406, "step": 7408 }, { "epoch": 0.42, "learning_rate": 0.0001302019632124619, "loss": 0.7387, "step": 7409 }, { "epoch": 0.42, "learning_rate": 0.00013018444875178187, "loss": 0.7793, "step": 7410 }, { "epoch": 0.42, "learning_rate": 0.00013016693327229755, "loss": 0.6866, "step": 7411 }, { "epoch": 0.42, "learning_rate": 0.00013014941677460005, "loss": 0.6258, "step": 7412 }, { "epoch": 0.42, "learning_rate": 0.00013013189925928072, "loss": 0.7725, "step": 7413 }, { "epoch": 0.42, "learning_rate": 0.00013011438072693077, "loss": 0.7559, "step": 7414 }, { "epoch": 0.42, "learning_rate": 0.00013009686117814145, "loss": 0.7011, "step": 7415 }, { "epoch": 0.42, "learning_rate": 0.00013007934061350418, "loss": 0.654, "step": 7416 }, { "epoch": 0.42, "learning_rate": 0.00013006181903361025, "loss": 0.701, "step": 7417 }, { "epoch": 0.42, "learning_rate": 0.0001300442964390511, "loss": 0.6541, "step": 7418 }, { "epoch": 0.42, "learning_rate": 0.00013002677283041814, "loss": 0.7246, "step": 7419 }, { "epoch": 0.42, "learning_rate": 0.00013000924820830284, "loss": 0.6997, "step": 7420 }, { "epoch": 0.42, "learning_rate": 0.00012999172257329672, "loss": 0.7219, "step": 7421 }, { "epoch": 0.42, "learning_rate": 0.00012997419592599135, "loss": 0.6277, "step": 7422 }, { "epoch": 0.42, "learning_rate": 0.00012995666826697819, "loss": 0.7666, "step": 7423 }, { "epoch": 0.42, "learning_rate": 0.00012993913959684895, "loss": 0.6807, "step": 7424 }, { "epoch": 0.42, "learning_rate": 0.00012992160991619523, "loss": 0.7081, "step": 7425 }, { "epoch": 0.42, "learning_rate": 0.00012990407922560868, "loss": 0.7474, "step": 7426 }, { "epoch": 0.42, "learning_rate": 0.00012988654752568105, "loss": 0.8158, "step": 7427 }, { "epoch": 0.42, "learning_rate": 0.00012986901481700405, "loss": 0.798, "step": 7428 }, { "epoch": 0.42, "learning_rate": 0.00012985148110016947, "loss": 0.7605, "step": 7429 }, { "epoch": 0.42, "learning_rate": 0.00012983394637576908, "loss": 0.7842, "step": 7430 }, { "epoch": 0.42, "learning_rate": 0.00012981641064439484, "loss": 0.7504, "step": 7431 }, { "epoch": 0.42, "learning_rate": 0.00012979887390663847, "loss": 0.7404, "step": 7432 }, { "epoch": 0.42, "learning_rate": 0.00012978133616309197, "loss": 0.8073, "step": 7433 }, { "epoch": 0.42, "learning_rate": 0.0001297637974143473, "loss": 0.8394, "step": 7434 }, { "epoch": 0.42, "learning_rate": 0.00012974625766099638, "loss": 0.7482, "step": 7435 }, { "epoch": 0.42, "learning_rate": 0.00012972871690363123, "loss": 0.7028, "step": 7436 }, { "epoch": 0.42, "learning_rate": 0.000129711175142844, "loss": 0.7084, "step": 7437 }, { "epoch": 0.42, "learning_rate": 0.00012969363237922663, "loss": 0.7826, "step": 7438 }, { "epoch": 0.42, "learning_rate": 0.00012967608861337132, "loss": 0.8131, "step": 7439 }, { "epoch": 0.42, "learning_rate": 0.00012965854384587016, "loss": 0.769, "step": 7440 }, { "epoch": 0.42, "learning_rate": 0.0001296409980773154, "loss": 0.6854, "step": 7441 }, { "epoch": 0.42, "learning_rate": 0.00012962345130829915, "loss": 0.6327, "step": 7442 }, { "epoch": 0.42, "learning_rate": 0.0001296059035394138, "loss": 0.6659, "step": 7443 }, { "epoch": 0.42, "learning_rate": 0.00012958835477125153, "loss": 0.669, "step": 7444 }, { "epoch": 0.42, "learning_rate": 0.00012957080500440468, "loss": 0.7764, "step": 7445 }, { "epoch": 0.42, "learning_rate": 0.00012955325423946563, "loss": 0.7495, "step": 7446 }, { "epoch": 0.42, "learning_rate": 0.00012953570247702673, "loss": 0.7193, "step": 7447 }, { "epoch": 0.42, "learning_rate": 0.0001295181497176804, "loss": 0.6515, "step": 7448 }, { "epoch": 0.42, "learning_rate": 0.0001295005959620191, "loss": 0.7881, "step": 7449 }, { "epoch": 0.42, "learning_rate": 0.00012948304121063536, "loss": 0.7481, "step": 7450 }, { "epoch": 0.42, "learning_rate": 0.00012946548546412164, "loss": 0.7184, "step": 7451 }, { "epoch": 0.42, "learning_rate": 0.0001294479287230705, "loss": 0.7496, "step": 7452 }, { "epoch": 0.42, "learning_rate": 0.00012943037098807453, "loss": 0.7287, "step": 7453 }, { "epoch": 0.42, "learning_rate": 0.00012941281225972636, "loss": 0.7833, "step": 7454 }, { "epoch": 0.42, "learning_rate": 0.00012939525253861866, "loss": 0.7887, "step": 7455 }, { "epoch": 0.42, "learning_rate": 0.00012937769182534407, "loss": 0.63, "step": 7456 }, { "epoch": 0.42, "learning_rate": 0.00012936013012049532, "loss": 0.6471, "step": 7457 }, { "epoch": 0.42, "learning_rate": 0.0001293425674246652, "loss": 0.7464, "step": 7458 }, { "epoch": 0.42, "learning_rate": 0.0001293250037384465, "loss": 0.7488, "step": 7459 }, { "epoch": 0.42, "learning_rate": 0.000129307439062432, "loss": 0.774, "step": 7460 }, { "epoch": 0.42, "learning_rate": 0.00012928987339721454, "loss": 0.6994, "step": 7461 }, { "epoch": 0.42, "learning_rate": 0.00012927230674338706, "loss": 0.6966, "step": 7462 }, { "epoch": 0.42, "learning_rate": 0.00012925473910154245, "loss": 0.7627, "step": 7463 }, { "epoch": 0.42, "learning_rate": 0.00012923717047227368, "loss": 0.7823, "step": 7464 }, { "epoch": 0.42, "learning_rate": 0.00012921960085617373, "loss": 0.7054, "step": 7465 }, { "epoch": 0.42, "learning_rate": 0.00012920203025383562, "loss": 0.7775, "step": 7466 }, { "epoch": 0.42, "learning_rate": 0.0001291844586658524, "loss": 0.7341, "step": 7467 }, { "epoch": 0.42, "learning_rate": 0.00012916688609281715, "loss": 0.7779, "step": 7468 }, { "epoch": 0.42, "learning_rate": 0.00012914931253532304, "loss": 0.8576, "step": 7469 }, { "epoch": 0.42, "learning_rate": 0.00012913173799396316, "loss": 0.8027, "step": 7470 }, { "epoch": 0.42, "learning_rate": 0.0001291141624693307, "loss": 0.7806, "step": 7471 }, { "epoch": 0.42, "learning_rate": 0.00012909658596201894, "loss": 0.7277, "step": 7472 }, { "epoch": 0.42, "learning_rate": 0.00012907900847262106, "loss": 0.8133, "step": 7473 }, { "epoch": 0.42, "learning_rate": 0.0001290614300017304, "loss": 0.7515, "step": 7474 }, { "epoch": 0.42, "learning_rate": 0.00012904385054994026, "loss": 0.7009, "step": 7475 }, { "epoch": 0.42, "learning_rate": 0.000129026270117844, "loss": 0.7244, "step": 7476 }, { "epoch": 0.42, "learning_rate": 0.00012900868870603503, "loss": 0.7229, "step": 7477 }, { "epoch": 0.42, "learning_rate": 0.00012899110631510666, "loss": 0.6891, "step": 7478 }, { "epoch": 0.42, "learning_rate": 0.0001289735229456525, "loss": 0.7533, "step": 7479 }, { "epoch": 0.42, "learning_rate": 0.00012895593859826588, "loss": 0.7115, "step": 7480 }, { "epoch": 0.42, "learning_rate": 0.00012893835327354046, "loss": 0.6892, "step": 7481 }, { "epoch": 0.42, "learning_rate": 0.00012892076697206968, "loss": 0.7232, "step": 7482 }, { "epoch": 0.42, "learning_rate": 0.00012890317969444716, "loss": 0.7992, "step": 7483 }, { "epoch": 0.42, "learning_rate": 0.00012888559144126656, "loss": 0.7783, "step": 7484 }, { "epoch": 0.42, "learning_rate": 0.0001288680022131215, "loss": 0.7454, "step": 7485 }, { "epoch": 0.42, "learning_rate": 0.0001288504120106056, "loss": 0.7211, "step": 7486 }, { "epoch": 0.42, "learning_rate": 0.0001288328208343127, "loss": 0.713, "step": 7487 }, { "epoch": 0.42, "learning_rate": 0.00012881522868483643, "loss": 0.6542, "step": 7488 }, { "epoch": 0.42, "learning_rate": 0.00012879763556277062, "loss": 0.7968, "step": 7489 }, { "epoch": 0.42, "learning_rate": 0.0001287800414687091, "loss": 0.6647, "step": 7490 }, { "epoch": 0.42, "learning_rate": 0.0001287624464032457, "loss": 0.7121, "step": 7491 }, { "epoch": 0.42, "learning_rate": 0.00012874485036697432, "loss": 0.6247, "step": 7492 }, { "epoch": 0.43, "learning_rate": 0.00012872725336048884, "loss": 0.7091, "step": 7493 }, { "epoch": 0.43, "learning_rate": 0.0001287096553843832, "loss": 0.7623, "step": 7494 }, { "epoch": 0.43, "learning_rate": 0.0001286920564392514, "loss": 0.7483, "step": 7495 }, { "epoch": 0.43, "learning_rate": 0.0001286744565256875, "loss": 0.7085, "step": 7496 }, { "epoch": 0.43, "learning_rate": 0.00012865685564428546, "loss": 0.7443, "step": 7497 }, { "epoch": 0.43, "learning_rate": 0.00012863925379563937, "loss": 0.7371, "step": 7498 }, { "epoch": 0.43, "learning_rate": 0.00012862165098034332, "loss": 0.7005, "step": 7499 }, { "epoch": 0.43, "learning_rate": 0.00012860404719899159, "loss": 0.6899, "step": 7500 }, { "epoch": 0.43, "learning_rate": 0.0001285864424521782, "loss": 0.7139, "step": 7501 }, { "epoch": 0.43, "learning_rate": 0.00012856883674049736, "loss": 0.7596, "step": 7502 }, { "epoch": 0.43, "learning_rate": 0.00012855123006454342, "loss": 0.7387, "step": 7503 }, { "epoch": 0.43, "learning_rate": 0.00012853362242491053, "loss": 0.6267, "step": 7504 }, { "epoch": 0.43, "learning_rate": 0.0001285160138221931, "loss": 0.6808, "step": 7505 }, { "epoch": 0.43, "learning_rate": 0.0001284984042569854, "loss": 0.7552, "step": 7506 }, { "epoch": 0.43, "learning_rate": 0.0001284807937298818, "loss": 0.6753, "step": 7507 }, { "epoch": 0.43, "learning_rate": 0.00012846318224147676, "loss": 0.7359, "step": 7508 }, { "epoch": 0.43, "learning_rate": 0.0001284455697923646, "loss": 0.7491, "step": 7509 }, { "epoch": 0.43, "learning_rate": 0.00012842795638313991, "loss": 0.7645, "step": 7510 }, { "epoch": 0.43, "learning_rate": 0.00012841034201439716, "loss": 0.6207, "step": 7511 }, { "epoch": 0.43, "learning_rate": 0.00012839272668673082, "loss": 0.7192, "step": 7512 }, { "epoch": 0.43, "learning_rate": 0.0001283751104007355, "loss": 0.785, "step": 7513 }, { "epoch": 0.43, "learning_rate": 0.00012835749315700577, "loss": 0.6476, "step": 7514 }, { "epoch": 0.43, "learning_rate": 0.00012833987495613632, "loss": 0.7066, "step": 7515 }, { "epoch": 0.43, "learning_rate": 0.00012832225579872172, "loss": 0.6925, "step": 7516 }, { "epoch": 0.43, "learning_rate": 0.00012830463568535677, "loss": 0.6582, "step": 7517 }, { "epoch": 0.43, "learning_rate": 0.0001282870146166361, "loss": 0.6902, "step": 7518 }, { "epoch": 0.43, "learning_rate": 0.00012826939259315447, "loss": 0.8193, "step": 7519 }, { "epoch": 0.43, "learning_rate": 0.0001282517696155067, "loss": 0.6752, "step": 7520 }, { "epoch": 0.43, "learning_rate": 0.00012823414568428768, "loss": 0.7734, "step": 7521 }, { "epoch": 0.43, "learning_rate": 0.00012821652080009213, "loss": 0.7025, "step": 7522 }, { "epoch": 0.43, "learning_rate": 0.00012819889496351502, "loss": 0.737, "step": 7523 }, { "epoch": 0.43, "learning_rate": 0.00012818126817515125, "loss": 0.7569, "step": 7524 }, { "epoch": 0.43, "learning_rate": 0.00012816364043559582, "loss": 0.6262, "step": 7525 }, { "epoch": 0.43, "learning_rate": 0.0001281460117454436, "loss": 0.7156, "step": 7526 }, { "epoch": 0.43, "learning_rate": 0.00012812838210528971, "loss": 0.6585, "step": 7527 }, { "epoch": 0.43, "learning_rate": 0.00012811075151572911, "loss": 0.7695, "step": 7528 }, { "epoch": 0.43, "learning_rate": 0.00012809311997735696, "loss": 0.7062, "step": 7529 }, { "epoch": 0.43, "learning_rate": 0.00012807548749076833, "loss": 0.6928, "step": 7530 }, { "epoch": 0.43, "learning_rate": 0.00012805785405655833, "loss": 0.6956, "step": 7531 }, { "epoch": 0.43, "learning_rate": 0.00012804021967532218, "loss": 0.6637, "step": 7532 }, { "epoch": 0.43, "learning_rate": 0.0001280225843476551, "loss": 0.7713, "step": 7533 }, { "epoch": 0.43, "learning_rate": 0.00012800494807415234, "loss": 0.7712, "step": 7534 }, { "epoch": 0.43, "learning_rate": 0.00012798731085540907, "loss": 0.7556, "step": 7535 }, { "epoch": 0.43, "learning_rate": 0.0001279696726920207, "loss": 0.7449, "step": 7536 }, { "epoch": 0.43, "learning_rate": 0.00012795203358458251, "loss": 0.682, "step": 7537 }, { "epoch": 0.43, "learning_rate": 0.0001279343935336899, "loss": 0.7135, "step": 7538 }, { "epoch": 0.43, "learning_rate": 0.0001279167525399382, "loss": 0.7074, "step": 7539 }, { "epoch": 0.43, "learning_rate": 0.00012789911060392294, "loss": 0.7364, "step": 7540 }, { "epoch": 0.43, "learning_rate": 0.0001278814677262395, "loss": 0.7451, "step": 7541 }, { "epoch": 0.43, "learning_rate": 0.00012786382390748342, "loss": 0.8284, "step": 7542 }, { "epoch": 0.43, "learning_rate": 0.0001278461791482502, "loss": 0.7577, "step": 7543 }, { "epoch": 0.43, "learning_rate": 0.00012782853344913542, "loss": 0.7476, "step": 7544 }, { "epoch": 0.43, "learning_rate": 0.0001278108868107346, "loss": 0.7469, "step": 7545 }, { "epoch": 0.43, "learning_rate": 0.00012779323923364346, "loss": 0.8, "step": 7546 }, { "epoch": 0.43, "learning_rate": 0.00012777559071845758, "loss": 0.9127, "step": 7547 }, { "epoch": 0.43, "learning_rate": 0.0001277579412657727, "loss": 0.7876, "step": 7548 }, { "epoch": 0.43, "learning_rate": 0.00012774029087618446, "loss": 0.752, "step": 7549 }, { "epoch": 0.43, "learning_rate": 0.00012772263955028869, "loss": 0.796, "step": 7550 }, { "epoch": 0.43, "learning_rate": 0.00012770498728868112, "loss": 0.7019, "step": 7551 }, { "epoch": 0.43, "learning_rate": 0.00012768733409195758, "loss": 0.7048, "step": 7552 }, { "epoch": 0.43, "learning_rate": 0.0001276696799607139, "loss": 0.5679, "step": 7553 }, { "epoch": 0.43, "learning_rate": 0.0001276520248955459, "loss": 0.803, "step": 7554 }, { "epoch": 0.43, "learning_rate": 0.0001276343688970496, "loss": 0.7798, "step": 7555 }, { "epoch": 0.43, "learning_rate": 0.00012761671196582087, "loss": 0.7742, "step": 7556 }, { "epoch": 0.43, "learning_rate": 0.00012759905410245564, "loss": 0.8645, "step": 7557 }, { "epoch": 0.43, "learning_rate": 0.00012758139530755, "loss": 0.6887, "step": 7558 }, { "epoch": 0.43, "learning_rate": 0.0001275637355816999, "loss": 0.7498, "step": 7559 }, { "epoch": 0.43, "learning_rate": 0.0001275460749255015, "loss": 0.8609, "step": 7560 }, { "epoch": 0.43, "learning_rate": 0.00012752841333955075, "loss": 0.7669, "step": 7561 }, { "epoch": 0.43, "learning_rate": 0.0001275107508244439, "loss": 0.7908, "step": 7562 }, { "epoch": 0.43, "learning_rate": 0.00012749308738077705, "loss": 0.7794, "step": 7563 }, { "epoch": 0.43, "learning_rate": 0.0001274754230091464, "loss": 0.6877, "step": 7564 }, { "epoch": 0.43, "learning_rate": 0.00012745775771014816, "loss": 0.7309, "step": 7565 }, { "epoch": 0.43, "learning_rate": 0.0001274400914843786, "loss": 0.7015, "step": 7566 }, { "epoch": 0.43, "learning_rate": 0.00012742242433243396, "loss": 0.6983, "step": 7567 }, { "epoch": 0.43, "learning_rate": 0.0001274047562549106, "loss": 0.661, "step": 7568 }, { "epoch": 0.43, "learning_rate": 0.00012738708725240484, "loss": 0.6889, "step": 7569 }, { "epoch": 0.43, "learning_rate": 0.00012736941732551307, "loss": 0.802, "step": 7570 }, { "epoch": 0.43, "learning_rate": 0.00012735174647483168, "loss": 0.7828, "step": 7571 }, { "epoch": 0.43, "learning_rate": 0.00012733407470095714, "loss": 0.7721, "step": 7572 }, { "epoch": 0.43, "learning_rate": 0.00012731640200448584, "loss": 0.6559, "step": 7573 }, { "epoch": 0.43, "learning_rate": 0.00012729872838601435, "loss": 0.6346, "step": 7574 }, { "epoch": 0.43, "learning_rate": 0.0001272810538461392, "loss": 0.807, "step": 7575 }, { "epoch": 0.43, "learning_rate": 0.0001272633783854569, "loss": 0.7096, "step": 7576 }, { "epoch": 0.43, "learning_rate": 0.00012724570200456413, "loss": 0.7129, "step": 7577 }, { "epoch": 0.43, "learning_rate": 0.00012722802470405744, "loss": 0.7414, "step": 7578 }, { "epoch": 0.43, "learning_rate": 0.00012721034648453353, "loss": 0.7914, "step": 7579 }, { "epoch": 0.43, "learning_rate": 0.00012719266734658903, "loss": 0.7365, "step": 7580 }, { "epoch": 0.43, "learning_rate": 0.00012717498729082068, "loss": 0.6631, "step": 7581 }, { "epoch": 0.43, "learning_rate": 0.0001271573063178253, "loss": 0.7484, "step": 7582 }, { "epoch": 0.43, "learning_rate": 0.00012713962442819957, "loss": 0.7277, "step": 7583 }, { "epoch": 0.43, "learning_rate": 0.0001271219416225403, "loss": 0.6937, "step": 7584 }, { "epoch": 0.43, "learning_rate": 0.00012710425790144446, "loss": 0.6459, "step": 7585 }, { "epoch": 0.43, "learning_rate": 0.00012708657326550874, "loss": 0.7114, "step": 7586 }, { "epoch": 0.43, "learning_rate": 0.00012706888771533018, "loss": 0.8075, "step": 7587 }, { "epoch": 0.43, "learning_rate": 0.00012705120125150568, "loss": 0.7355, "step": 7588 }, { "epoch": 0.43, "learning_rate": 0.0001270335138746322, "loss": 0.7447, "step": 7589 }, { "epoch": 0.43, "learning_rate": 0.00012701582558530668, "loss": 0.639, "step": 7590 }, { "epoch": 0.43, "learning_rate": 0.00012699813638412626, "loss": 0.7722, "step": 7591 }, { "epoch": 0.43, "learning_rate": 0.00012698044627168788, "loss": 0.7283, "step": 7592 }, { "epoch": 0.43, "learning_rate": 0.0001269627552485887, "loss": 0.7899, "step": 7593 }, { "epoch": 0.43, "learning_rate": 0.0001269450633154258, "loss": 0.805, "step": 7594 }, { "epoch": 0.43, "learning_rate": 0.00012692737047279636, "loss": 0.5942, "step": 7595 }, { "epoch": 0.43, "learning_rate": 0.00012690967672129757, "loss": 0.7001, "step": 7596 }, { "epoch": 0.43, "learning_rate": 0.00012689198206152657, "loss": 0.6804, "step": 7597 }, { "epoch": 0.43, "learning_rate": 0.00012687428649408068, "loss": 0.7292, "step": 7598 }, { "epoch": 0.43, "learning_rate": 0.00012685659001955712, "loss": 0.7394, "step": 7599 }, { "epoch": 0.43, "learning_rate": 0.00012683889263855324, "loss": 0.7572, "step": 7600 }, { "epoch": 0.43, "learning_rate": 0.00012682119435166635, "loss": 0.7882, "step": 7601 }, { "epoch": 0.43, "learning_rate": 0.00012680349515949377, "loss": 0.8257, "step": 7602 }, { "epoch": 0.43, "learning_rate": 0.00012678579506263297, "loss": 0.6467, "step": 7603 }, { "epoch": 0.43, "learning_rate": 0.00012676809406168133, "loss": 0.6724, "step": 7604 }, { "epoch": 0.43, "learning_rate": 0.0001267503921572363, "loss": 0.7282, "step": 7605 }, { "epoch": 0.43, "learning_rate": 0.00012673268934989537, "loss": 0.7993, "step": 7606 }, { "epoch": 0.43, "learning_rate": 0.0001267149856402561, "loss": 0.694, "step": 7607 }, { "epoch": 0.43, "learning_rate": 0.000126697281028916, "loss": 0.7621, "step": 7608 }, { "epoch": 0.43, "learning_rate": 0.00012667957551647262, "loss": 0.7218, "step": 7609 }, { "epoch": 0.43, "learning_rate": 0.0001266618691035236, "loss": 0.7376, "step": 7610 }, { "epoch": 0.43, "learning_rate": 0.00012664416179066661, "loss": 0.6835, "step": 7611 }, { "epoch": 0.43, "learning_rate": 0.00012662645357849924, "loss": 0.6612, "step": 7612 }, { "epoch": 0.43, "learning_rate": 0.00012660874446761926, "loss": 0.635, "step": 7613 }, { "epoch": 0.43, "learning_rate": 0.00012659103445862436, "loss": 0.688, "step": 7614 }, { "epoch": 0.43, "learning_rate": 0.00012657332355211232, "loss": 0.7617, "step": 7615 }, { "epoch": 0.43, "learning_rate": 0.00012655561174868088, "loss": 0.6843, "step": 7616 }, { "epoch": 0.43, "learning_rate": 0.00012653789904892794, "loss": 0.6691, "step": 7617 }, { "epoch": 0.43, "learning_rate": 0.0001265201854534513, "loss": 0.7152, "step": 7618 }, { "epoch": 0.43, "learning_rate": 0.00012650247096284884, "loss": 0.7307, "step": 7619 }, { "epoch": 0.43, "learning_rate": 0.00012648475557771845, "loss": 0.7588, "step": 7620 }, { "epoch": 0.43, "learning_rate": 0.00012646703929865817, "loss": 0.7478, "step": 7621 }, { "epoch": 0.43, "learning_rate": 0.00012644932212626587, "loss": 0.678, "step": 7622 }, { "epoch": 0.43, "learning_rate": 0.00012643160406113955, "loss": 0.6555, "step": 7623 }, { "epoch": 0.43, "learning_rate": 0.00012641388510387734, "loss": 0.7329, "step": 7624 }, { "epoch": 0.43, "learning_rate": 0.00012639616525507717, "loss": 0.7322, "step": 7625 }, { "epoch": 0.43, "learning_rate": 0.00012637844451533723, "loss": 0.6723, "step": 7626 }, { "epoch": 0.43, "learning_rate": 0.0001263607228852556, "loss": 0.7436, "step": 7627 }, { "epoch": 0.43, "learning_rate": 0.00012634300036543043, "loss": 0.6791, "step": 7628 }, { "epoch": 0.43, "learning_rate": 0.00012632527695645993, "loss": 0.786, "step": 7629 }, { "epoch": 0.43, "learning_rate": 0.0001263075526589423, "loss": 0.7644, "step": 7630 }, { "epoch": 0.43, "learning_rate": 0.00012628982747347575, "loss": 0.7381, "step": 7631 }, { "epoch": 0.43, "learning_rate": 0.00012627210140065857, "loss": 0.6582, "step": 7632 }, { "epoch": 0.43, "learning_rate": 0.00012625437444108907, "loss": 0.7107, "step": 7633 }, { "epoch": 0.43, "learning_rate": 0.0001262366465953656, "loss": 0.7281, "step": 7634 }, { "epoch": 0.43, "learning_rate": 0.00012621891786408648, "loss": 0.6631, "step": 7635 }, { "epoch": 0.43, "learning_rate": 0.0001262011882478501, "loss": 0.7086, "step": 7636 }, { "epoch": 0.43, "learning_rate": 0.00012618345774725498, "loss": 0.6746, "step": 7637 }, { "epoch": 0.43, "learning_rate": 0.00012616572636289943, "loss": 0.7446, "step": 7638 }, { "epoch": 0.43, "learning_rate": 0.00012614799409538198, "loss": 0.7196, "step": 7639 }, { "epoch": 0.43, "learning_rate": 0.00012613026094530122, "loss": 0.7121, "step": 7640 }, { "epoch": 0.43, "learning_rate": 0.00012611252691325554, "loss": 0.7346, "step": 7641 }, { "epoch": 0.43, "learning_rate": 0.00012609479199984365, "loss": 0.7715, "step": 7642 }, { "epoch": 0.43, "learning_rate": 0.00012607705620566408, "loss": 0.6584, "step": 7643 }, { "epoch": 0.43, "learning_rate": 0.00012605931953131548, "loss": 0.7302, "step": 7644 }, { "epoch": 0.43, "learning_rate": 0.00012604158197739647, "loss": 0.7584, "step": 7645 }, { "epoch": 0.43, "learning_rate": 0.0001260238435445058, "loss": 0.7552, "step": 7646 }, { "epoch": 0.43, "learning_rate": 0.00012600610423324214, "loss": 0.6101, "step": 7647 }, { "epoch": 0.43, "learning_rate": 0.0001259883640442043, "loss": 0.7551, "step": 7648 }, { "epoch": 0.43, "learning_rate": 0.000125970622977991, "loss": 0.7823, "step": 7649 }, { "epoch": 0.43, "learning_rate": 0.00012595288103520104, "loss": 0.6621, "step": 7650 }, { "epoch": 0.43, "learning_rate": 0.00012593513821643331, "loss": 0.7548, "step": 7651 }, { "epoch": 0.43, "learning_rate": 0.0001259173945222866, "loss": 0.8494, "step": 7652 }, { "epoch": 0.43, "learning_rate": 0.0001258996499533599, "loss": 0.6537, "step": 7653 }, { "epoch": 0.43, "learning_rate": 0.00012588190451025207, "loss": 0.7359, "step": 7654 }, { "epoch": 0.43, "learning_rate": 0.00012586415819356212, "loss": 0.7142, "step": 7655 }, { "epoch": 0.43, "learning_rate": 0.00012584641100388898, "loss": 0.5836, "step": 7656 }, { "epoch": 0.43, "learning_rate": 0.00012582866294183167, "loss": 0.7031, "step": 7657 }, { "epoch": 0.43, "learning_rate": 0.00012581091400798924, "loss": 0.7328, "step": 7658 }, { "epoch": 0.43, "learning_rate": 0.0001257931642029608, "loss": 0.8041, "step": 7659 }, { "epoch": 0.43, "learning_rate": 0.0001257754135273454, "loss": 0.6963, "step": 7660 }, { "epoch": 0.43, "learning_rate": 0.0001257576619817422, "loss": 0.737, "step": 7661 }, { "epoch": 0.43, "learning_rate": 0.00012573990956675038, "loss": 0.7041, "step": 7662 }, { "epoch": 0.43, "learning_rate": 0.00012572215628296908, "loss": 0.7635, "step": 7663 }, { "epoch": 0.43, "learning_rate": 0.00012570440213099756, "loss": 0.7382, "step": 7664 }, { "epoch": 0.43, "learning_rate": 0.000125686647111435, "loss": 0.707, "step": 7665 }, { "epoch": 0.43, "learning_rate": 0.00012566889122488084, "loss": 0.7784, "step": 7666 }, { "epoch": 0.43, "learning_rate": 0.0001256511344719342, "loss": 0.7195, "step": 7667 }, { "epoch": 0.43, "learning_rate": 0.0001256333768531945, "loss": 0.7288, "step": 7668 }, { "epoch": 0.44, "learning_rate": 0.00012561561836926113, "loss": 0.6944, "step": 7669 }, { "epoch": 0.44, "learning_rate": 0.00012559785902073348, "loss": 0.777, "step": 7670 }, { "epoch": 0.44, "learning_rate": 0.00012558009880821094, "loss": 0.8285, "step": 7671 }, { "epoch": 0.44, "learning_rate": 0.00012556233773229297, "loss": 0.7293, "step": 7672 }, { "epoch": 0.44, "learning_rate": 0.00012554457579357905, "loss": 0.7273, "step": 7673 }, { "epoch": 0.44, "learning_rate": 0.00012552681299266873, "loss": 0.6994, "step": 7674 }, { "epoch": 0.44, "learning_rate": 0.00012550904933016153, "loss": 0.7471, "step": 7675 }, { "epoch": 0.44, "learning_rate": 0.000125491284806657, "loss": 0.6402, "step": 7676 }, { "epoch": 0.44, "learning_rate": 0.00012547351942275475, "loss": 0.6871, "step": 7677 }, { "epoch": 0.44, "learning_rate": 0.0001254557531790544, "loss": 0.8327, "step": 7678 }, { "epoch": 0.44, "learning_rate": 0.00012543798607615565, "loss": 0.7084, "step": 7679 }, { "epoch": 0.44, "learning_rate": 0.00012542021811465815, "loss": 0.7514, "step": 7680 }, { "epoch": 0.44, "learning_rate": 0.0001254024492951616, "loss": 0.7229, "step": 7681 }, { "epoch": 0.44, "learning_rate": 0.00012538467961826578, "loss": 0.7848, "step": 7682 }, { "epoch": 0.44, "learning_rate": 0.00012536690908457044, "loss": 0.6483, "step": 7683 }, { "epoch": 0.44, "learning_rate": 0.0001253491376946754, "loss": 0.6732, "step": 7684 }, { "epoch": 0.44, "learning_rate": 0.0001253313654491805, "loss": 0.7732, "step": 7685 }, { "epoch": 0.44, "learning_rate": 0.0001253135923486856, "loss": 0.7741, "step": 7686 }, { "epoch": 0.44, "learning_rate": 0.0001252958183937905, "loss": 0.7905, "step": 7687 }, { "epoch": 0.44, "learning_rate": 0.00012527804358509524, "loss": 0.6769, "step": 7688 }, { "epoch": 0.44, "learning_rate": 0.0001252602679231997, "loss": 0.6596, "step": 7689 }, { "epoch": 0.44, "learning_rate": 0.00012524249140870387, "loss": 0.668, "step": 7690 }, { "epoch": 0.44, "learning_rate": 0.00012522471404220775, "loss": 0.7129, "step": 7691 }, { "epoch": 0.44, "learning_rate": 0.0001252069358243114, "loss": 0.7822, "step": 7692 }, { "epoch": 0.44, "learning_rate": 0.00012518915675561483, "loss": 0.7338, "step": 7693 }, { "epoch": 0.44, "learning_rate": 0.00012517137683671815, "loss": 0.625, "step": 7694 }, { "epoch": 0.44, "learning_rate": 0.0001251535960682215, "loss": 0.7998, "step": 7695 }, { "epoch": 0.44, "learning_rate": 0.00012513581445072503, "loss": 0.7571, "step": 7696 }, { "epoch": 0.44, "learning_rate": 0.00012511803198482887, "loss": 0.719, "step": 7697 }, { "epoch": 0.44, "learning_rate": 0.0001251002486711333, "loss": 0.71, "step": 7698 }, { "epoch": 0.44, "learning_rate": 0.00012508246451023845, "loss": 0.6796, "step": 7699 }, { "epoch": 0.44, "learning_rate": 0.00012506467950274471, "loss": 0.6609, "step": 7700 }, { "epoch": 0.44, "learning_rate": 0.00012504689364925228, "loss": 0.6769, "step": 7701 }, { "epoch": 0.44, "learning_rate": 0.00012502910695036146, "loss": 0.7402, "step": 7702 }, { "epoch": 0.44, "learning_rate": 0.0001250113194066727, "loss": 0.6356, "step": 7703 }, { "epoch": 0.44, "learning_rate": 0.0001249935310187863, "loss": 0.7061, "step": 7704 }, { "epoch": 0.44, "learning_rate": 0.00012497574178730266, "loss": 0.8317, "step": 7705 }, { "epoch": 0.44, "learning_rate": 0.00012495795171282224, "loss": 0.6663, "step": 7706 }, { "epoch": 0.44, "learning_rate": 0.0001249401607959455, "loss": 0.7189, "step": 7707 }, { "epoch": 0.44, "learning_rate": 0.00012492236903727293, "loss": 0.7127, "step": 7708 }, { "epoch": 0.44, "learning_rate": 0.000124904576437405, "loss": 0.6765, "step": 7709 }, { "epoch": 0.44, "learning_rate": 0.00012488678299694232, "loss": 0.7702, "step": 7710 }, { "epoch": 0.44, "learning_rate": 0.0001248689887164855, "loss": 0.8043, "step": 7711 }, { "epoch": 0.44, "learning_rate": 0.00012485119359663505, "loss": 0.7554, "step": 7712 }, { "epoch": 0.44, "learning_rate": 0.00012483339763799163, "loss": 0.7169, "step": 7713 }, { "epoch": 0.44, "learning_rate": 0.00012481560084115593, "loss": 0.7211, "step": 7714 }, { "epoch": 0.44, "learning_rate": 0.00012479780320672862, "loss": 0.7214, "step": 7715 }, { "epoch": 0.44, "learning_rate": 0.0001247800047353104, "loss": 0.7628, "step": 7716 }, { "epoch": 0.44, "learning_rate": 0.00012476220542750204, "loss": 0.6151, "step": 7717 }, { "epoch": 0.44, "learning_rate": 0.00012474440528390432, "loss": 0.7413, "step": 7718 }, { "epoch": 0.44, "learning_rate": 0.000124726604305118, "loss": 0.6544, "step": 7719 }, { "epoch": 0.44, "learning_rate": 0.00012470880249174397, "loss": 0.6782, "step": 7720 }, { "epoch": 0.44, "learning_rate": 0.00012469099984438301, "loss": 0.8028, "step": 7721 }, { "epoch": 0.44, "learning_rate": 0.0001246731963636361, "loss": 0.7517, "step": 7722 }, { "epoch": 0.44, "learning_rate": 0.00012465539205010407, "loss": 0.7625, "step": 7723 }, { "epoch": 0.44, "learning_rate": 0.00012463758690438792, "loss": 0.8104, "step": 7724 }, { "epoch": 0.44, "learning_rate": 0.00012461978092708856, "loss": 0.7263, "step": 7725 }, { "epoch": 0.44, "learning_rate": 0.00012460197411880708, "loss": 0.7172, "step": 7726 }, { "epoch": 0.44, "learning_rate": 0.00012458416648014443, "loss": 0.6341, "step": 7727 }, { "epoch": 0.44, "learning_rate": 0.00012456635801170166, "loss": 0.7466, "step": 7728 }, { "epoch": 0.44, "learning_rate": 0.00012454854871407994, "loss": 0.8375, "step": 7729 }, { "epoch": 0.44, "learning_rate": 0.00012453073858788026, "loss": 0.7338, "step": 7730 }, { "epoch": 0.44, "learning_rate": 0.00012451292763370386, "loss": 0.7017, "step": 7731 }, { "epoch": 0.44, "learning_rate": 0.0001244951158521518, "loss": 0.6816, "step": 7732 }, { "epoch": 0.44, "learning_rate": 0.00012447730324382543, "loss": 0.6431, "step": 7733 }, { "epoch": 0.44, "learning_rate": 0.00012445948980932583, "loss": 0.6107, "step": 7734 }, { "epoch": 0.44, "learning_rate": 0.0001244416755492543, "loss": 0.7703, "step": 7735 }, { "epoch": 0.44, "learning_rate": 0.00012442386046421211, "loss": 0.7086, "step": 7736 }, { "epoch": 0.44, "learning_rate": 0.00012440604455480063, "loss": 0.6644, "step": 7737 }, { "epoch": 0.44, "learning_rate": 0.0001243882278216211, "loss": 0.651, "step": 7738 }, { "epoch": 0.44, "learning_rate": 0.00012437041026527488, "loss": 0.6966, "step": 7739 }, { "epoch": 0.44, "learning_rate": 0.0001243525918863635, "loss": 0.8072, "step": 7740 }, { "epoch": 0.44, "learning_rate": 0.00012433477268548822, "loss": 0.7224, "step": 7741 }, { "epoch": 0.44, "learning_rate": 0.00012431695266325053, "loss": 0.8713, "step": 7742 }, { "epoch": 0.44, "learning_rate": 0.0001242991318202519, "loss": 0.8224, "step": 7743 }, { "epoch": 0.44, "learning_rate": 0.00012428131015709387, "loss": 0.7039, "step": 7744 }, { "epoch": 0.44, "learning_rate": 0.00012426348767437793, "loss": 0.7841, "step": 7745 }, { "epoch": 0.44, "learning_rate": 0.00012424566437270568, "loss": 0.7371, "step": 7746 }, { "epoch": 0.44, "learning_rate": 0.00012422784025267864, "loss": 0.7007, "step": 7747 }, { "epoch": 0.44, "learning_rate": 0.0001242100153148985, "loss": 0.6852, "step": 7748 }, { "epoch": 0.44, "learning_rate": 0.00012419218955996676, "loss": 0.7204, "step": 7749 }, { "epoch": 0.44, "learning_rate": 0.00012417436298848525, "loss": 0.7551, "step": 7750 }, { "epoch": 0.44, "learning_rate": 0.00012415653560105558, "loss": 0.7052, "step": 7751 }, { "epoch": 0.44, "learning_rate": 0.00012413870739827948, "loss": 0.7543, "step": 7752 }, { "epoch": 0.44, "learning_rate": 0.00012412087838075867, "loss": 0.7213, "step": 7753 }, { "epoch": 0.44, "learning_rate": 0.00012410304854909495, "loss": 0.7646, "step": 7754 }, { "epoch": 0.44, "learning_rate": 0.00012408521790389015, "loss": 0.7975, "step": 7755 }, { "epoch": 0.44, "learning_rate": 0.00012406738644574612, "loss": 0.7927, "step": 7756 }, { "epoch": 0.44, "learning_rate": 0.00012404955417526461, "loss": 0.7148, "step": 7757 }, { "epoch": 0.44, "learning_rate": 0.00012403172109304758, "loss": 0.6574, "step": 7758 }, { "epoch": 0.44, "learning_rate": 0.00012401388719969698, "loss": 0.7477, "step": 7759 }, { "epoch": 0.44, "learning_rate": 0.0001239960524958147, "loss": 0.5876, "step": 7760 }, { "epoch": 0.44, "learning_rate": 0.0001239782169820027, "loss": 0.7993, "step": 7761 }, { "epoch": 0.44, "learning_rate": 0.000123960380658863, "loss": 0.7839, "step": 7762 }, { "epoch": 0.44, "learning_rate": 0.00012394254352699765, "loss": 0.6553, "step": 7763 }, { "epoch": 0.44, "learning_rate": 0.00012392470558700858, "loss": 0.7287, "step": 7764 }, { "epoch": 0.44, "learning_rate": 0.00012390686683949798, "loss": 0.6976, "step": 7765 }, { "epoch": 0.44, "learning_rate": 0.000123889027285068, "loss": 0.7298, "step": 7766 }, { "epoch": 0.44, "learning_rate": 0.00012387118692432062, "loss": 0.761, "step": 7767 }, { "epoch": 0.44, "learning_rate": 0.0001238533457578581, "loss": 0.7412, "step": 7768 }, { "epoch": 0.44, "learning_rate": 0.00012383550378628258, "loss": 0.558, "step": 7769 }, { "epoch": 0.44, "learning_rate": 0.00012381766101019634, "loss": 0.6892, "step": 7770 }, { "epoch": 0.44, "learning_rate": 0.00012379981743020156, "loss": 0.8436, "step": 7771 }, { "epoch": 0.44, "learning_rate": 0.00012378197304690051, "loss": 0.7211, "step": 7772 }, { "epoch": 0.44, "learning_rate": 0.0001237641278608955, "loss": 0.7245, "step": 7773 }, { "epoch": 0.44, "learning_rate": 0.00012374628187278888, "loss": 0.7462, "step": 7774 }, { "epoch": 0.44, "learning_rate": 0.00012372843508318295, "loss": 0.7221, "step": 7775 }, { "epoch": 0.44, "learning_rate": 0.0001237105874926801, "loss": 0.6446, "step": 7776 }, { "epoch": 0.44, "learning_rate": 0.00012369273910188276, "loss": 0.736, "step": 7777 }, { "epoch": 0.44, "learning_rate": 0.00012367488991139332, "loss": 0.7536, "step": 7778 }, { "epoch": 0.44, "learning_rate": 0.00012365703992181425, "loss": 0.7042, "step": 7779 }, { "epoch": 0.44, "learning_rate": 0.00012363918913374804, "loss": 0.7779, "step": 7780 }, { "epoch": 0.44, "learning_rate": 0.00012362133754779722, "loss": 0.798, "step": 7781 }, { "epoch": 0.44, "learning_rate": 0.0001236034851645643, "loss": 0.6552, "step": 7782 }, { "epoch": 0.44, "learning_rate": 0.00012358563198465182, "loss": 0.651, "step": 7783 }, { "epoch": 0.44, "learning_rate": 0.00012356777800866245, "loss": 0.6536, "step": 7784 }, { "epoch": 0.44, "learning_rate": 0.00012354992323719877, "loss": 0.7079, "step": 7785 }, { "epoch": 0.44, "learning_rate": 0.0001235320676708634, "loss": 0.745, "step": 7786 }, { "epoch": 0.44, "learning_rate": 0.000123514211310259, "loss": 0.6877, "step": 7787 }, { "epoch": 0.44, "learning_rate": 0.00012349635415598835, "loss": 0.6766, "step": 7788 }, { "epoch": 0.44, "learning_rate": 0.0001234784962086541, "loss": 0.7184, "step": 7789 }, { "epoch": 0.44, "learning_rate": 0.00012346063746885902, "loss": 0.6356, "step": 7790 }, { "epoch": 0.44, "learning_rate": 0.00012344277793720592, "loss": 0.7587, "step": 7791 }, { "epoch": 0.44, "learning_rate": 0.0001234249176142976, "loss": 0.7823, "step": 7792 }, { "epoch": 0.44, "learning_rate": 0.00012340705650073683, "loss": 0.656, "step": 7793 }, { "epoch": 0.44, "learning_rate": 0.00012338919459712653, "loss": 0.7128, "step": 7794 }, { "epoch": 0.44, "learning_rate": 0.00012337133190406957, "loss": 0.7129, "step": 7795 }, { "epoch": 0.44, "learning_rate": 0.00012335346842216887, "loss": 0.7871, "step": 7796 }, { "epoch": 0.44, "learning_rate": 0.0001233356041520274, "loss": 0.8414, "step": 7797 }, { "epoch": 0.44, "learning_rate": 0.00012331773909424803, "loss": 0.6504, "step": 7798 }, { "epoch": 0.44, "learning_rate": 0.00012329987324943384, "loss": 0.795, "step": 7799 }, { "epoch": 0.44, "learning_rate": 0.00012328200661818783, "loss": 0.7006, "step": 7800 }, { "epoch": 0.44, "learning_rate": 0.00012326413920111303, "loss": 0.6891, "step": 7801 }, { "epoch": 0.44, "learning_rate": 0.0001232462709988125, "loss": 0.7423, "step": 7802 }, { "epoch": 0.44, "learning_rate": 0.0001232284020118894, "loss": 0.6939, "step": 7803 }, { "epoch": 0.44, "learning_rate": 0.0001232105322409468, "loss": 0.7107, "step": 7804 }, { "epoch": 0.44, "learning_rate": 0.00012319266168658784, "loss": 0.7482, "step": 7805 }, { "epoch": 0.44, "learning_rate": 0.00012317479034941573, "loss": 0.6821, "step": 7806 }, { "epoch": 0.44, "learning_rate": 0.00012315691823003367, "loss": 0.7341, "step": 7807 }, { "epoch": 0.44, "learning_rate": 0.00012313904532904494, "loss": 0.8264, "step": 7808 }, { "epoch": 0.44, "learning_rate": 0.00012312117164705265, "loss": 0.7372, "step": 7809 }, { "epoch": 0.44, "learning_rate": 0.00012310329718466024, "loss": 0.6519, "step": 7810 }, { "epoch": 0.44, "learning_rate": 0.00012308542194247097, "loss": 0.8122, "step": 7811 }, { "epoch": 0.44, "learning_rate": 0.00012306754592108813, "loss": 0.7101, "step": 7812 }, { "epoch": 0.44, "learning_rate": 0.00012304966912111512, "loss": 0.7499, "step": 7813 }, { "epoch": 0.44, "learning_rate": 0.00012303179154315538, "loss": 0.761, "step": 7814 }, { "epoch": 0.44, "learning_rate": 0.0001230139131878122, "loss": 0.6665, "step": 7815 }, { "epoch": 0.44, "learning_rate": 0.00012299603405568912, "loss": 0.7239, "step": 7816 }, { "epoch": 0.44, "learning_rate": 0.0001229781541473896, "loss": 0.6337, "step": 7817 }, { "epoch": 0.44, "learning_rate": 0.00012296027346351713, "loss": 0.7057, "step": 7818 }, { "epoch": 0.44, "learning_rate": 0.00012294239200467516, "loss": 0.675, "step": 7819 }, { "epoch": 0.44, "learning_rate": 0.00012292450977146734, "loss": 0.7073, "step": 7820 }, { "epoch": 0.44, "learning_rate": 0.00012290662676449715, "loss": 0.6996, "step": 7821 }, { "epoch": 0.44, "learning_rate": 0.0001228887429843683, "loss": 0.7153, "step": 7822 }, { "epoch": 0.44, "learning_rate": 0.00012287085843168426, "loss": 0.715, "step": 7823 }, { "epoch": 0.44, "learning_rate": 0.0001228529731070488, "loss": 0.676, "step": 7824 }, { "epoch": 0.44, "learning_rate": 0.00012283508701106557, "loss": 0.6371, "step": 7825 }, { "epoch": 0.44, "learning_rate": 0.00012281720014433828, "loss": 0.7476, "step": 7826 }, { "epoch": 0.44, "learning_rate": 0.00012279931250747063, "loss": 0.7505, "step": 7827 }, { "epoch": 0.44, "learning_rate": 0.0001227814241010664, "loss": 0.7222, "step": 7828 }, { "epoch": 0.44, "learning_rate": 0.00012276353492572935, "loss": 0.7404, "step": 7829 }, { "epoch": 0.44, "learning_rate": 0.00012274564498206331, "loss": 0.7629, "step": 7830 }, { "epoch": 0.44, "learning_rate": 0.00012272775427067206, "loss": 0.6493, "step": 7831 }, { "epoch": 0.44, "learning_rate": 0.00012270986279215953, "loss": 0.7061, "step": 7832 }, { "epoch": 0.44, "learning_rate": 0.00012269197054712958, "loss": 0.6394, "step": 7833 }, { "epoch": 0.44, "learning_rate": 0.0001226740775361861, "loss": 0.7963, "step": 7834 }, { "epoch": 0.44, "learning_rate": 0.000122656183759933, "loss": 0.6705, "step": 7835 }, { "epoch": 0.44, "learning_rate": 0.00012263828921897435, "loss": 0.7782, "step": 7836 }, { "epoch": 0.44, "learning_rate": 0.00012262039391391404, "loss": 0.8002, "step": 7837 }, { "epoch": 0.44, "learning_rate": 0.00012260249784535612, "loss": 0.6149, "step": 7838 }, { "epoch": 0.44, "learning_rate": 0.0001225846010139046, "loss": 0.711, "step": 7839 }, { "epoch": 0.44, "learning_rate": 0.0001225667034201636, "loss": 0.8441, "step": 7840 }, { "epoch": 0.44, "learning_rate": 0.00012254880506473712, "loss": 0.6692, "step": 7841 }, { "epoch": 0.44, "learning_rate": 0.00012253090594822937, "loss": 0.7962, "step": 7842 }, { "epoch": 0.44, "learning_rate": 0.00012251300607124443, "loss": 0.7047, "step": 7843 }, { "epoch": 0.44, "learning_rate": 0.0001224951054343865, "loss": 0.7467, "step": 7844 }, { "epoch": 0.45, "learning_rate": 0.00012247720403825977, "loss": 0.6597, "step": 7845 }, { "epoch": 0.45, "learning_rate": 0.00012245930188346847, "loss": 0.7034, "step": 7846 }, { "epoch": 0.45, "learning_rate": 0.00012244139897061681, "loss": 0.6803, "step": 7847 }, { "epoch": 0.45, "learning_rate": 0.0001224234953003091, "loss": 0.6693, "step": 7848 }, { "epoch": 0.45, "learning_rate": 0.0001224055908731496, "loss": 0.5911, "step": 7849 }, { "epoch": 0.45, "learning_rate": 0.00012238768568974265, "loss": 0.758, "step": 7850 }, { "epoch": 0.45, "learning_rate": 0.0001223697797506926, "loss": 0.7626, "step": 7851 }, { "epoch": 0.45, "learning_rate": 0.00012235187305660383, "loss": 0.7109, "step": 7852 }, { "epoch": 0.45, "learning_rate": 0.0001223339656080807, "loss": 0.6962, "step": 7853 }, { "epoch": 0.45, "learning_rate": 0.00012231605740572766, "loss": 0.6737, "step": 7854 }, { "epoch": 0.45, "learning_rate": 0.0001222981484501492, "loss": 0.7177, "step": 7855 }, { "epoch": 0.45, "learning_rate": 0.0001222802387419497, "loss": 0.8543, "step": 7856 }, { "epoch": 0.45, "learning_rate": 0.0001222623282817337, "loss": 0.7206, "step": 7857 }, { "epoch": 0.45, "learning_rate": 0.00012224441707010575, "loss": 0.7252, "step": 7858 }, { "epoch": 0.45, "learning_rate": 0.00012222650510767042, "loss": 0.7001, "step": 7859 }, { "epoch": 0.45, "learning_rate": 0.0001222085923950322, "loss": 0.7164, "step": 7860 }, { "epoch": 0.45, "learning_rate": 0.0001221906789327958, "loss": 0.7381, "step": 7861 }, { "epoch": 0.45, "learning_rate": 0.00012217276472156573, "loss": 0.7125, "step": 7862 }, { "epoch": 0.45, "learning_rate": 0.00012215484976194676, "loss": 0.6306, "step": 7863 }, { "epoch": 0.45, "learning_rate": 0.00012213693405454344, "loss": 0.6302, "step": 7864 }, { "epoch": 0.45, "learning_rate": 0.0001221190175999606, "loss": 0.7325, "step": 7865 }, { "epoch": 0.45, "learning_rate": 0.00012210110039880288, "loss": 0.7592, "step": 7866 }, { "epoch": 0.45, "learning_rate": 0.00012208318245167504, "loss": 0.6942, "step": 7867 }, { "epoch": 0.45, "learning_rate": 0.0001220652637591819, "loss": 0.8253, "step": 7868 }, { "epoch": 0.45, "learning_rate": 0.00012204734432192823, "loss": 0.806, "step": 7869 }, { "epoch": 0.45, "learning_rate": 0.0001220294241405189, "loss": 0.7467, "step": 7870 }, { "epoch": 0.45, "learning_rate": 0.0001220115032155587, "loss": 0.6657, "step": 7871 }, { "epoch": 0.45, "learning_rate": 0.00012199358154765255, "loss": 0.7779, "step": 7872 }, { "epoch": 0.45, "learning_rate": 0.00012197565913740531, "loss": 0.7724, "step": 7873 }, { "epoch": 0.45, "learning_rate": 0.00012195773598542199, "loss": 0.6828, "step": 7874 }, { "epoch": 0.45, "learning_rate": 0.00012193981209230749, "loss": 0.744, "step": 7875 }, { "epoch": 0.45, "learning_rate": 0.00012192188745866675, "loss": 0.7152, "step": 7876 }, { "epoch": 0.45, "learning_rate": 0.00012190396208510486, "loss": 0.7149, "step": 7877 }, { "epoch": 0.45, "learning_rate": 0.00012188603597222676, "loss": 0.8041, "step": 7878 }, { "epoch": 0.45, "learning_rate": 0.0001218681091206376, "loss": 0.7356, "step": 7879 }, { "epoch": 0.45, "learning_rate": 0.00012185018153094236, "loss": 0.653, "step": 7880 }, { "epoch": 0.45, "learning_rate": 0.00012183225320374623, "loss": 0.6852, "step": 7881 }, { "epoch": 0.45, "learning_rate": 0.00012181432413965428, "loss": 0.744, "step": 7882 }, { "epoch": 0.45, "learning_rate": 0.00012179639433927166, "loss": 0.7026, "step": 7883 }, { "epoch": 0.45, "learning_rate": 0.00012177846380320356, "loss": 0.7748, "step": 7884 }, { "epoch": 0.45, "learning_rate": 0.00012176053253205526, "loss": 0.7813, "step": 7885 }, { "epoch": 0.45, "learning_rate": 0.00012174260052643185, "loss": 0.7989, "step": 7886 }, { "epoch": 0.45, "learning_rate": 0.00012172466778693868, "loss": 0.8185, "step": 7887 }, { "epoch": 0.45, "learning_rate": 0.00012170673431418098, "loss": 0.7273, "step": 7888 }, { "epoch": 0.45, "learning_rate": 0.00012168880010876409, "loss": 0.7337, "step": 7889 }, { "epoch": 0.45, "learning_rate": 0.0001216708651712933, "loss": 0.5884, "step": 7890 }, { "epoch": 0.45, "learning_rate": 0.00012165292950237399, "loss": 0.8209, "step": 7891 }, { "epoch": 0.45, "learning_rate": 0.00012163499310261151, "loss": 0.7334, "step": 7892 }, { "epoch": 0.45, "learning_rate": 0.00012161705597261126, "loss": 0.7078, "step": 7893 }, { "epoch": 0.45, "learning_rate": 0.00012159911811297868, "loss": 0.7663, "step": 7894 }, { "epoch": 0.45, "learning_rate": 0.00012158117952431924, "loss": 0.756, "step": 7895 }, { "epoch": 0.45, "learning_rate": 0.00012156324020723842, "loss": 0.7606, "step": 7896 }, { "epoch": 0.45, "learning_rate": 0.00012154530016234163, "loss": 0.7401, "step": 7897 }, { "epoch": 0.45, "learning_rate": 0.0001215273593902345, "loss": 0.7112, "step": 7898 }, { "epoch": 0.45, "learning_rate": 0.00012150941789152248, "loss": 0.636, "step": 7899 }, { "epoch": 0.45, "learning_rate": 0.00012149147566681124, "loss": 0.8022, "step": 7900 }, { "epoch": 0.45, "learning_rate": 0.00012147353271670634, "loss": 0.6904, "step": 7901 }, { "epoch": 0.45, "learning_rate": 0.00012145558904181335, "loss": 0.6863, "step": 7902 }, { "epoch": 0.45, "learning_rate": 0.00012143764464273802, "loss": 0.6094, "step": 7903 }, { "epoch": 0.45, "learning_rate": 0.00012141969952008591, "loss": 0.7778, "step": 7904 }, { "epoch": 0.45, "learning_rate": 0.00012140175367446278, "loss": 0.7683, "step": 7905 }, { "epoch": 0.45, "learning_rate": 0.00012138380710647435, "loss": 0.6488, "step": 7906 }, { "epoch": 0.45, "learning_rate": 0.00012136585981672632, "loss": 0.724, "step": 7907 }, { "epoch": 0.45, "learning_rate": 0.00012134791180582452, "loss": 0.7709, "step": 7908 }, { "epoch": 0.45, "learning_rate": 0.0001213299630743747, "loss": 0.6987, "step": 7909 }, { "epoch": 0.45, "learning_rate": 0.00012131201362298266, "loss": 0.6442, "step": 7910 }, { "epoch": 0.45, "learning_rate": 0.00012129406345225431, "loss": 0.703, "step": 7911 }, { "epoch": 0.45, "learning_rate": 0.00012127611256279545, "loss": 0.7233, "step": 7912 }, { "epoch": 0.45, "learning_rate": 0.00012125816095521199, "loss": 0.7505, "step": 7913 }, { "epoch": 0.45, "learning_rate": 0.00012124020863010988, "loss": 0.7273, "step": 7914 }, { "epoch": 0.45, "learning_rate": 0.00012122225558809497, "loss": 0.6446, "step": 7915 }, { "epoch": 0.45, "learning_rate": 0.00012120430182977329, "loss": 0.7586, "step": 7916 }, { "epoch": 0.45, "learning_rate": 0.00012118634735575083, "loss": 0.8514, "step": 7917 }, { "epoch": 0.45, "learning_rate": 0.00012116839216663359, "loss": 0.6846, "step": 7918 }, { "epoch": 0.45, "learning_rate": 0.00012115043626302758, "loss": 0.6461, "step": 7919 }, { "epoch": 0.45, "learning_rate": 0.00012113247964553888, "loss": 0.7439, "step": 7920 }, { "epoch": 0.45, "learning_rate": 0.00012111452231477356, "loss": 0.7004, "step": 7921 }, { "epoch": 0.45, "learning_rate": 0.00012109656427133775, "loss": 0.7372, "step": 7922 }, { "epoch": 0.45, "learning_rate": 0.00012107860551583757, "loss": 0.6476, "step": 7923 }, { "epoch": 0.45, "learning_rate": 0.00012106064604887918, "loss": 0.7718, "step": 7924 }, { "epoch": 0.45, "learning_rate": 0.00012104268587106873, "loss": 0.7003, "step": 7925 }, { "epoch": 0.45, "learning_rate": 0.00012102472498301248, "loss": 0.7817, "step": 7926 }, { "epoch": 0.45, "learning_rate": 0.0001210067633853166, "loss": 0.7643, "step": 7927 }, { "epoch": 0.45, "learning_rate": 0.00012098880107858735, "loss": 0.6648, "step": 7928 }, { "epoch": 0.45, "learning_rate": 0.00012097083806343103, "loss": 0.6883, "step": 7929 }, { "epoch": 0.45, "learning_rate": 0.00012095287434045395, "loss": 0.6169, "step": 7930 }, { "epoch": 0.45, "learning_rate": 0.00012093490991026238, "loss": 0.7633, "step": 7931 }, { "epoch": 0.45, "learning_rate": 0.00012091694477346269, "loss": 0.7264, "step": 7932 }, { "epoch": 0.45, "learning_rate": 0.00012089897893066128, "loss": 0.7915, "step": 7933 }, { "epoch": 0.45, "learning_rate": 0.00012088101238246454, "loss": 0.7216, "step": 7934 }, { "epoch": 0.45, "learning_rate": 0.00012086304512947882, "loss": 0.7371, "step": 7935 }, { "epoch": 0.45, "learning_rate": 0.00012084507717231064, "loss": 0.7489, "step": 7936 }, { "epoch": 0.45, "learning_rate": 0.00012082710851156645, "loss": 0.7506, "step": 7937 }, { "epoch": 0.45, "learning_rate": 0.0001208091391478527, "loss": 0.6312, "step": 7938 }, { "epoch": 0.45, "learning_rate": 0.00012079116908177593, "loss": 0.8282, "step": 7939 }, { "epoch": 0.45, "learning_rate": 0.00012077319831394272, "loss": 0.7343, "step": 7940 }, { "epoch": 0.45, "learning_rate": 0.00012075522684495954, "loss": 0.6763, "step": 7941 }, { "epoch": 0.45, "learning_rate": 0.00012073725467543302, "loss": 0.7359, "step": 7942 }, { "epoch": 0.45, "learning_rate": 0.00012071928180596979, "loss": 0.697, "step": 7943 }, { "epoch": 0.45, "learning_rate": 0.00012070130823717647, "loss": 0.7866, "step": 7944 }, { "epoch": 0.45, "learning_rate": 0.00012068333396965968, "loss": 0.7121, "step": 7945 }, { "epoch": 0.45, "learning_rate": 0.00012066535900402616, "loss": 0.7973, "step": 7946 }, { "epoch": 0.45, "learning_rate": 0.00012064738334088253, "loss": 0.7113, "step": 7947 }, { "epoch": 0.45, "learning_rate": 0.0001206294069808356, "loss": 0.7231, "step": 7948 }, { "epoch": 0.45, "learning_rate": 0.00012061142992449203, "loss": 0.7467, "step": 7949 }, { "epoch": 0.45, "learning_rate": 0.00012059345217245873, "loss": 0.7537, "step": 7950 }, { "epoch": 0.45, "learning_rate": 0.00012057547372534235, "loss": 0.8262, "step": 7951 }, { "epoch": 0.45, "learning_rate": 0.00012055749458374976, "loss": 0.6913, "step": 7952 }, { "epoch": 0.45, "learning_rate": 0.00012053951474828786, "loss": 0.7769, "step": 7953 }, { "epoch": 0.45, "learning_rate": 0.00012052153421956342, "loss": 0.7798, "step": 7954 }, { "epoch": 0.45, "learning_rate": 0.00012050355299818343, "loss": 0.704, "step": 7955 }, { "epoch": 0.45, "learning_rate": 0.00012048557108475473, "loss": 0.6662, "step": 7956 }, { "epoch": 0.45, "learning_rate": 0.00012046758847988429, "loss": 0.6509, "step": 7957 }, { "epoch": 0.45, "learning_rate": 0.00012044960518417903, "loss": 0.7351, "step": 7958 }, { "epoch": 0.45, "learning_rate": 0.00012043162119824603, "loss": 0.6452, "step": 7959 }, { "epoch": 0.45, "learning_rate": 0.00012041363652269218, "loss": 0.6154, "step": 7960 }, { "epoch": 0.45, "learning_rate": 0.00012039565115812455, "loss": 0.7422, "step": 7961 }, { "epoch": 0.45, "learning_rate": 0.00012037766510515022, "loss": 0.784, "step": 7962 }, { "epoch": 0.45, "learning_rate": 0.00012035967836437625, "loss": 0.7294, "step": 7963 }, { "epoch": 0.45, "learning_rate": 0.00012034169093640975, "loss": 0.6599, "step": 7964 }, { "epoch": 0.45, "learning_rate": 0.00012032370282185783, "loss": 0.8195, "step": 7965 }, { "epoch": 0.45, "learning_rate": 0.00012030571402132764, "loss": 0.6981, "step": 7966 }, { "epoch": 0.45, "learning_rate": 0.00012028772453542632, "loss": 0.7257, "step": 7967 }, { "epoch": 0.45, "learning_rate": 0.0001202697343647611, "loss": 0.7021, "step": 7968 }, { "epoch": 0.45, "learning_rate": 0.00012025174350993922, "loss": 0.662, "step": 7969 }, { "epoch": 0.45, "learning_rate": 0.00012023375197156786, "loss": 0.6803, "step": 7970 }, { "epoch": 0.45, "learning_rate": 0.0001202157597502543, "loss": 0.7075, "step": 7971 }, { "epoch": 0.45, "learning_rate": 0.00012019776684660584, "loss": 0.7287, "step": 7972 }, { "epoch": 0.45, "learning_rate": 0.00012017977326122978, "loss": 0.7743, "step": 7973 }, { "epoch": 0.45, "learning_rate": 0.00012016177899473344, "loss": 0.704, "step": 7974 }, { "epoch": 0.45, "learning_rate": 0.00012014378404772421, "loss": 0.6606, "step": 7975 }, { "epoch": 0.45, "learning_rate": 0.00012012578842080941, "loss": 0.7515, "step": 7976 }, { "epoch": 0.45, "learning_rate": 0.00012010779211459648, "loss": 0.7449, "step": 7977 }, { "epoch": 0.45, "learning_rate": 0.00012008979512969282, "loss": 0.7515, "step": 7978 }, { "epoch": 0.45, "learning_rate": 0.00012007179746670592, "loss": 0.636, "step": 7979 }, { "epoch": 0.45, "learning_rate": 0.00012005379912624318, "loss": 0.6245, "step": 7980 }, { "epoch": 0.45, "learning_rate": 0.00012003580010891213, "loss": 0.5947, "step": 7981 }, { "epoch": 0.45, "learning_rate": 0.00012001780041532029, "loss": 0.7835, "step": 7982 }, { "epoch": 0.45, "learning_rate": 0.00011999980004607518, "loss": 0.7247, "step": 7983 }, { "epoch": 0.45, "learning_rate": 0.00011998179900178436, "loss": 0.7543, "step": 7984 }, { "epoch": 0.45, "learning_rate": 0.00011996379728305547, "loss": 0.5407, "step": 7985 }, { "epoch": 0.45, "learning_rate": 0.00011994579489049603, "loss": 0.7053, "step": 7986 }, { "epoch": 0.45, "learning_rate": 0.00011992779182471367, "loss": 0.7496, "step": 7987 }, { "epoch": 0.45, "learning_rate": 0.00011990978808631612, "loss": 0.6528, "step": 7988 }, { "epoch": 0.45, "learning_rate": 0.00011989178367591098, "loss": 0.7153, "step": 7989 }, { "epoch": 0.45, "learning_rate": 0.00011987377859410597, "loss": 0.5783, "step": 7990 }, { "epoch": 0.45, "learning_rate": 0.00011985577284150884, "loss": 0.6619, "step": 7991 }, { "epoch": 0.45, "learning_rate": 0.00011983776641872731, "loss": 0.7623, "step": 7992 }, { "epoch": 0.45, "learning_rate": 0.00011981975932636909, "loss": 0.7663, "step": 7993 }, { "epoch": 0.45, "learning_rate": 0.00011980175156504203, "loss": 0.728, "step": 7994 }, { "epoch": 0.45, "learning_rate": 0.00011978374313535394, "loss": 0.6976, "step": 7995 }, { "epoch": 0.45, "learning_rate": 0.00011976573403791262, "loss": 0.6677, "step": 7996 }, { "epoch": 0.45, "learning_rate": 0.00011974772427332594, "loss": 0.7747, "step": 7997 }, { "epoch": 0.45, "learning_rate": 0.0001197297138422018, "loss": 0.7794, "step": 7998 }, { "epoch": 0.45, "learning_rate": 0.00011971170274514802, "loss": 0.6288, "step": 7999 }, { "epoch": 0.45, "learning_rate": 0.00011969369098277263, "loss": 0.7409, "step": 8000 }, { "epoch": 0.45, "learning_rate": 0.0001196756785556835, "loss": 0.7028, "step": 8001 }, { "epoch": 0.45, "learning_rate": 0.00011965766546448862, "loss": 0.6388, "step": 8002 }, { "epoch": 0.45, "learning_rate": 0.00011963965170979599, "loss": 0.7435, "step": 8003 }, { "epoch": 0.45, "learning_rate": 0.0001196216372922136, "loss": 0.7194, "step": 8004 }, { "epoch": 0.45, "learning_rate": 0.00011960362221234946, "loss": 0.7624, "step": 8005 }, { "epoch": 0.45, "learning_rate": 0.00011958560647081167, "loss": 0.7031, "step": 8006 }, { "epoch": 0.45, "learning_rate": 0.00011956759006820833, "loss": 0.7588, "step": 8007 }, { "epoch": 0.45, "learning_rate": 0.00011954957300514748, "loss": 0.7636, "step": 8008 }, { "epoch": 0.45, "learning_rate": 0.00011953155528223727, "loss": 0.726, "step": 8009 }, { "epoch": 0.45, "learning_rate": 0.00011951353690008587, "loss": 0.6559, "step": 8010 }, { "epoch": 0.45, "learning_rate": 0.00011949551785930141, "loss": 0.7967, "step": 8011 }, { "epoch": 0.45, "learning_rate": 0.00011947749816049207, "loss": 0.746, "step": 8012 }, { "epoch": 0.45, "learning_rate": 0.00011945947780426609, "loss": 0.6813, "step": 8013 }, { "epoch": 0.45, "learning_rate": 0.00011944145679123175, "loss": 0.7017, "step": 8014 }, { "epoch": 0.45, "learning_rate": 0.0001194234351219972, "loss": 0.7092, "step": 8015 }, { "epoch": 0.45, "learning_rate": 0.0001194054127971708, "loss": 0.6815, "step": 8016 }, { "epoch": 0.45, "learning_rate": 0.00011938738981736085, "loss": 0.7686, "step": 8017 }, { "epoch": 0.45, "learning_rate": 0.00011936936618317561, "loss": 0.7493, "step": 8018 }, { "epoch": 0.45, "learning_rate": 0.00011935134189522345, "loss": 0.7164, "step": 8019 }, { "epoch": 0.45, "learning_rate": 0.00011933331695411278, "loss": 0.7268, "step": 8020 }, { "epoch": 0.45, "learning_rate": 0.00011931529136045194, "loss": 0.7257, "step": 8021 }, { "epoch": 0.46, "learning_rate": 0.0001192972651148494, "loss": 0.7279, "step": 8022 }, { "epoch": 0.46, "learning_rate": 0.00011927923821791352, "loss": 0.7707, "step": 8023 }, { "epoch": 0.46, "learning_rate": 0.00011926121067025278, "loss": 0.8078, "step": 8024 }, { "epoch": 0.46, "learning_rate": 0.00011924318247247568, "loss": 0.7088, "step": 8025 }, { "epoch": 0.46, "learning_rate": 0.0001192251536251907, "loss": 0.6428, "step": 8026 }, { "epoch": 0.46, "learning_rate": 0.00011920712412900639, "loss": 0.8798, "step": 8027 }, { "epoch": 0.46, "learning_rate": 0.00011918909398453124, "loss": 0.7339, "step": 8028 }, { "epoch": 0.46, "learning_rate": 0.00011917106319237386, "loss": 0.7393, "step": 8029 }, { "epoch": 0.46, "learning_rate": 0.0001191530317531428, "loss": 0.7145, "step": 8030 }, { "epoch": 0.46, "learning_rate": 0.0001191349996674467, "loss": 0.7093, "step": 8031 }, { "epoch": 0.46, "learning_rate": 0.00011911696693589416, "loss": 0.672, "step": 8032 }, { "epoch": 0.46, "learning_rate": 0.00011909893355909387, "loss": 0.5869, "step": 8033 }, { "epoch": 0.46, "learning_rate": 0.00011908089953765449, "loss": 0.7206, "step": 8034 }, { "epoch": 0.46, "learning_rate": 0.0001190628648721847, "loss": 0.6406, "step": 8035 }, { "epoch": 0.46, "learning_rate": 0.00011904482956329324, "loss": 0.7104, "step": 8036 }, { "epoch": 0.46, "learning_rate": 0.00011902679361158882, "loss": 0.6871, "step": 8037 }, { "epoch": 0.46, "learning_rate": 0.0001190087570176802, "loss": 0.653, "step": 8038 }, { "epoch": 0.46, "learning_rate": 0.00011899071978217617, "loss": 0.7664, "step": 8039 }, { "epoch": 0.46, "learning_rate": 0.0001189726819056856, "loss": 0.6875, "step": 8040 }, { "epoch": 0.46, "learning_rate": 0.00011895464338881721, "loss": 0.6712, "step": 8041 }, { "epoch": 0.46, "learning_rate": 0.00011893660423217992, "loss": 0.6327, "step": 8042 }, { "epoch": 0.46, "learning_rate": 0.00011891856443638256, "loss": 0.6772, "step": 8043 }, { "epoch": 0.46, "learning_rate": 0.00011890052400203404, "loss": 0.7223, "step": 8044 }, { "epoch": 0.46, "learning_rate": 0.00011888248292974326, "loss": 0.733, "step": 8045 }, { "epoch": 0.46, "learning_rate": 0.00011886444122011918, "loss": 0.7272, "step": 8046 }, { "epoch": 0.46, "learning_rate": 0.00011884639887377069, "loss": 0.7218, "step": 8047 }, { "epoch": 0.46, "learning_rate": 0.00011882835589130686, "loss": 0.7324, "step": 8048 }, { "epoch": 0.46, "learning_rate": 0.00011881031227333663, "loss": 0.7231, "step": 8049 }, { "epoch": 0.46, "learning_rate": 0.00011879226802046902, "loss": 0.7732, "step": 8050 }, { "epoch": 0.46, "learning_rate": 0.00011877422313331308, "loss": 0.7391, "step": 8051 }, { "epoch": 0.46, "learning_rate": 0.00011875617761247787, "loss": 0.5839, "step": 8052 }, { "epoch": 0.46, "learning_rate": 0.00011873813145857249, "loss": 0.8068, "step": 8053 }, { "epoch": 0.46, "learning_rate": 0.00011872008467220599, "loss": 0.8269, "step": 8054 }, { "epoch": 0.46, "learning_rate": 0.00011870203725398759, "loss": 0.7928, "step": 8055 }, { "epoch": 0.46, "learning_rate": 0.00011868398920452634, "loss": 0.6874, "step": 8056 }, { "epoch": 0.46, "learning_rate": 0.00011866594052443148, "loss": 0.6504, "step": 8057 }, { "epoch": 0.46, "learning_rate": 0.00011864789121431215, "loss": 0.7728, "step": 8058 }, { "epoch": 0.46, "learning_rate": 0.00011862984127477763, "loss": 0.6797, "step": 8059 }, { "epoch": 0.46, "learning_rate": 0.00011861179070643709, "loss": 0.6579, "step": 8060 }, { "epoch": 0.46, "learning_rate": 0.00011859373950989976, "loss": 0.8068, "step": 8061 }, { "epoch": 0.46, "learning_rate": 0.000118575687685775, "loss": 0.7012, "step": 8062 }, { "epoch": 0.46, "learning_rate": 0.00011855763523467207, "loss": 0.7331, "step": 8063 }, { "epoch": 0.46, "learning_rate": 0.00011853958215720027, "loss": 0.6911, "step": 8064 }, { "epoch": 0.46, "learning_rate": 0.00011852152845396895, "loss": 0.7747, "step": 8065 }, { "epoch": 0.46, "learning_rate": 0.00011850347412558749, "loss": 0.6757, "step": 8066 }, { "epoch": 0.46, "learning_rate": 0.00011848541917266521, "loss": 0.7159, "step": 8067 }, { "epoch": 0.46, "learning_rate": 0.00011846736359581157, "loss": 0.709, "step": 8068 }, { "epoch": 0.46, "learning_rate": 0.00011844930739563598, "loss": 0.6664, "step": 8069 }, { "epoch": 0.46, "learning_rate": 0.00011843125057274789, "loss": 0.762, "step": 8070 }, { "epoch": 0.46, "learning_rate": 0.00011841319312775671, "loss": 0.6812, "step": 8071 }, { "epoch": 0.46, "learning_rate": 0.00011839513506127203, "loss": 0.7726, "step": 8072 }, { "epoch": 0.46, "learning_rate": 0.00011837707637390328, "loss": 0.7559, "step": 8073 }, { "epoch": 0.46, "learning_rate": 0.00011835901706626001, "loss": 0.7868, "step": 8074 }, { "epoch": 0.46, "learning_rate": 0.00011834095713895175, "loss": 0.7796, "step": 8075 }, { "epoch": 0.46, "learning_rate": 0.00011832289659258808, "loss": 0.6943, "step": 8076 }, { "epoch": 0.46, "learning_rate": 0.00011830483542777862, "loss": 0.7053, "step": 8077 }, { "epoch": 0.46, "learning_rate": 0.00011828677364513294, "loss": 0.7489, "step": 8078 }, { "epoch": 0.46, "learning_rate": 0.00011826871124526071, "loss": 0.7, "step": 8079 }, { "epoch": 0.46, "learning_rate": 0.00011825064822877152, "loss": 0.6738, "step": 8080 }, { "epoch": 0.46, "learning_rate": 0.00011823258459627516, "loss": 0.7703, "step": 8081 }, { "epoch": 0.46, "learning_rate": 0.00011821452034838121, "loss": 0.7114, "step": 8082 }, { "epoch": 0.46, "learning_rate": 0.00011819645548569941, "loss": 0.7123, "step": 8083 }, { "epoch": 0.46, "learning_rate": 0.00011817839000883952, "loss": 0.7225, "step": 8084 }, { "epoch": 0.46, "learning_rate": 0.00011816032391841135, "loss": 0.8318, "step": 8085 }, { "epoch": 0.46, "learning_rate": 0.00011814225721502457, "loss": 0.7246, "step": 8086 }, { "epoch": 0.46, "learning_rate": 0.00011812418989928904, "loss": 0.7595, "step": 8087 }, { "epoch": 0.46, "learning_rate": 0.00011810612197181456, "loss": 0.6804, "step": 8088 }, { "epoch": 0.46, "learning_rate": 0.000118088053433211, "loss": 0.6116, "step": 8089 }, { "epoch": 0.46, "learning_rate": 0.00011806998428408818, "loss": 0.6953, "step": 8090 }, { "epoch": 0.46, "learning_rate": 0.00011805191452505602, "loss": 0.7212, "step": 8091 }, { "epoch": 0.46, "learning_rate": 0.00011803384415672438, "loss": 0.7588, "step": 8092 }, { "epoch": 0.46, "learning_rate": 0.0001180157731797032, "loss": 0.8102, "step": 8093 }, { "epoch": 0.46, "learning_rate": 0.00011799770159460246, "loss": 0.7391, "step": 8094 }, { "epoch": 0.46, "learning_rate": 0.00011797962940203207, "loss": 0.6683, "step": 8095 }, { "epoch": 0.46, "learning_rate": 0.00011796155660260204, "loss": 0.7142, "step": 8096 }, { "epoch": 0.46, "learning_rate": 0.00011794348319692234, "loss": 0.7579, "step": 8097 }, { "epoch": 0.46, "learning_rate": 0.00011792540918560306, "loss": 0.7038, "step": 8098 }, { "epoch": 0.46, "learning_rate": 0.00011790733456925419, "loss": 0.7195, "step": 8099 }, { "epoch": 0.46, "learning_rate": 0.00011788925934848583, "loss": 0.7304, "step": 8100 }, { "epoch": 0.46, "learning_rate": 0.00011787118352390805, "loss": 0.6923, "step": 8101 }, { "epoch": 0.46, "learning_rate": 0.00011785310709613094, "loss": 0.7851, "step": 8102 }, { "epoch": 0.46, "learning_rate": 0.00011783503006576464, "loss": 0.6559, "step": 8103 }, { "epoch": 0.46, "learning_rate": 0.00011781695243341932, "loss": 0.7465, "step": 8104 }, { "epoch": 0.46, "learning_rate": 0.00011779887419970512, "loss": 0.7189, "step": 8105 }, { "epoch": 0.46, "learning_rate": 0.00011778079536523223, "loss": 0.7387, "step": 8106 }, { "epoch": 0.46, "learning_rate": 0.00011776271593061089, "loss": 0.637, "step": 8107 }, { "epoch": 0.46, "learning_rate": 0.00011774463589645129, "loss": 0.7393, "step": 8108 }, { "epoch": 0.46, "learning_rate": 0.00011772655526336367, "loss": 0.8653, "step": 8109 }, { "epoch": 0.46, "learning_rate": 0.00011770847403195834, "loss": 0.7959, "step": 8110 }, { "epoch": 0.46, "learning_rate": 0.00011769039220284557, "loss": 0.6146, "step": 8111 }, { "epoch": 0.46, "learning_rate": 0.00011767230977663565, "loss": 0.6205, "step": 8112 }, { "epoch": 0.46, "learning_rate": 0.00011765422675393893, "loss": 0.7201, "step": 8113 }, { "epoch": 0.46, "learning_rate": 0.00011763614313536582, "loss": 0.6538, "step": 8114 }, { "epoch": 0.46, "learning_rate": 0.00011761805892152655, "loss": 0.7584, "step": 8115 }, { "epoch": 0.46, "learning_rate": 0.00011759997411303161, "loss": 0.7559, "step": 8116 }, { "epoch": 0.46, "learning_rate": 0.0001175818887104914, "loss": 0.6823, "step": 8117 }, { "epoch": 0.46, "learning_rate": 0.00011756380271451633, "loss": 0.7076, "step": 8118 }, { "epoch": 0.46, "learning_rate": 0.00011754571612571683, "loss": 0.6581, "step": 8119 }, { "epoch": 0.46, "learning_rate": 0.00011752762894470343, "loss": 0.7535, "step": 8120 }, { "epoch": 0.46, "learning_rate": 0.00011750954117208657, "loss": 0.7146, "step": 8121 }, { "epoch": 0.46, "learning_rate": 0.00011749145280847678, "loss": 0.7438, "step": 8122 }, { "epoch": 0.46, "learning_rate": 0.00011747336385448458, "loss": 0.6918, "step": 8123 }, { "epoch": 0.46, "learning_rate": 0.00011745527431072052, "loss": 0.6676, "step": 8124 }, { "epoch": 0.46, "learning_rate": 0.00011743718417779517, "loss": 0.7528, "step": 8125 }, { "epoch": 0.46, "learning_rate": 0.00011741909345631914, "loss": 0.7988, "step": 8126 }, { "epoch": 0.46, "learning_rate": 0.00011740100214690301, "loss": 0.7773, "step": 8127 }, { "epoch": 0.46, "learning_rate": 0.00011738291025015744, "loss": 0.7109, "step": 8128 }, { "epoch": 0.46, "learning_rate": 0.00011736481776669306, "loss": 0.6856, "step": 8129 }, { "epoch": 0.46, "learning_rate": 0.00011734672469712053, "loss": 0.672, "step": 8130 }, { "epoch": 0.46, "learning_rate": 0.00011732863104205052, "loss": 0.7975, "step": 8131 }, { "epoch": 0.46, "learning_rate": 0.00011731053680209378, "loss": 0.6535, "step": 8132 }, { "epoch": 0.46, "learning_rate": 0.00011729244197786105, "loss": 0.6648, "step": 8133 }, { "epoch": 0.46, "learning_rate": 0.00011727434656996305, "loss": 0.7682, "step": 8134 }, { "epoch": 0.46, "learning_rate": 0.00011725625057901053, "loss": 0.8223, "step": 8135 }, { "epoch": 0.46, "learning_rate": 0.00011723815400561431, "loss": 0.7953, "step": 8136 }, { "epoch": 0.46, "learning_rate": 0.00011722005685038518, "loss": 0.713, "step": 8137 }, { "epoch": 0.46, "learning_rate": 0.00011720195911393396, "loss": 0.7086, "step": 8138 }, { "epoch": 0.46, "learning_rate": 0.00011718386079687152, "loss": 0.672, "step": 8139 }, { "epoch": 0.46, "learning_rate": 0.00011716576189980873, "loss": 0.706, "step": 8140 }, { "epoch": 0.46, "learning_rate": 0.00011714766242335642, "loss": 0.7168, "step": 8141 }, { "epoch": 0.46, "learning_rate": 0.00011712956236812553, "loss": 0.7139, "step": 8142 }, { "epoch": 0.46, "learning_rate": 0.000117111461734727, "loss": 0.7213, "step": 8143 }, { "epoch": 0.46, "learning_rate": 0.00011709336052377178, "loss": 0.7013, "step": 8144 }, { "epoch": 0.46, "learning_rate": 0.00011707525873587077, "loss": 0.8097, "step": 8145 }, { "epoch": 0.46, "learning_rate": 0.00011705715637163503, "loss": 0.7523, "step": 8146 }, { "epoch": 0.46, "learning_rate": 0.00011703905343167552, "loss": 0.7268, "step": 8147 }, { "epoch": 0.46, "learning_rate": 0.00011702094991660326, "loss": 0.7799, "step": 8148 }, { "epoch": 0.46, "learning_rate": 0.00011700284582702932, "loss": 0.6705, "step": 8149 }, { "epoch": 0.46, "learning_rate": 0.00011698474116356473, "loss": 0.8159, "step": 8150 }, { "epoch": 0.46, "learning_rate": 0.00011696663592682058, "loss": 0.6961, "step": 8151 }, { "epoch": 0.46, "learning_rate": 0.00011694853011740798, "loss": 0.769, "step": 8152 }, { "epoch": 0.46, "learning_rate": 0.00011693042373593803, "loss": 0.6279, "step": 8153 }, { "epoch": 0.46, "learning_rate": 0.00011691231678302187, "loss": 0.7933, "step": 8154 }, { "epoch": 0.46, "learning_rate": 0.00011689420925927069, "loss": 0.6692, "step": 8155 }, { "epoch": 0.46, "learning_rate": 0.00011687610116529563, "loss": 0.6352, "step": 8156 }, { "epoch": 0.46, "learning_rate": 0.00011685799250170788, "loss": 0.8437, "step": 8157 }, { "epoch": 0.46, "learning_rate": 0.00011683988326911868, "loss": 0.7217, "step": 8158 }, { "epoch": 0.46, "learning_rate": 0.0001168217734681393, "loss": 0.7543, "step": 8159 }, { "epoch": 0.46, "learning_rate": 0.00011680366309938089, "loss": 0.7539, "step": 8160 }, { "epoch": 0.46, "learning_rate": 0.00011678555216345477, "loss": 0.6541, "step": 8161 }, { "epoch": 0.46, "learning_rate": 0.0001167674406609723, "loss": 0.7257, "step": 8162 }, { "epoch": 0.46, "learning_rate": 0.00011674932859254471, "loss": 0.743, "step": 8163 }, { "epoch": 0.46, "learning_rate": 0.00011673121595878333, "loss": 0.7787, "step": 8164 }, { "epoch": 0.46, "learning_rate": 0.00011671310276029957, "loss": 0.7413, "step": 8165 }, { "epoch": 0.46, "learning_rate": 0.00011669498899770474, "loss": 0.692, "step": 8166 }, { "epoch": 0.46, "learning_rate": 0.00011667687467161024, "loss": 0.6483, "step": 8167 }, { "epoch": 0.46, "learning_rate": 0.00011665875978262749, "loss": 0.7133, "step": 8168 }, { "epoch": 0.46, "learning_rate": 0.0001166406443313679, "loss": 0.7643, "step": 8169 }, { "epoch": 0.46, "learning_rate": 0.00011662252831844296, "loss": 0.6749, "step": 8170 }, { "epoch": 0.46, "learning_rate": 0.00011660441174446404, "loss": 0.7688, "step": 8171 }, { "epoch": 0.46, "learning_rate": 0.00011658629461004271, "loss": 0.726, "step": 8172 }, { "epoch": 0.46, "learning_rate": 0.00011656817691579043, "loss": 0.5935, "step": 8173 }, { "epoch": 0.46, "learning_rate": 0.00011655005866231874, "loss": 0.6252, "step": 8174 }, { "epoch": 0.46, "learning_rate": 0.00011653193985023915, "loss": 0.755, "step": 8175 }, { "epoch": 0.46, "learning_rate": 0.00011651382048016324, "loss": 0.5982, "step": 8176 }, { "epoch": 0.46, "learning_rate": 0.00011649570055270257, "loss": 0.6622, "step": 8177 }, { "epoch": 0.46, "learning_rate": 0.00011647758006846877, "loss": 0.6635, "step": 8178 }, { "epoch": 0.46, "learning_rate": 0.00011645945902807341, "loss": 0.6537, "step": 8179 }, { "epoch": 0.46, "learning_rate": 0.00011644133743212813, "loss": 0.7445, "step": 8180 }, { "epoch": 0.46, "learning_rate": 0.0001164232152812446, "loss": 0.771, "step": 8181 }, { "epoch": 0.46, "learning_rate": 0.00011640509257603452, "loss": 0.7061, "step": 8182 }, { "epoch": 0.46, "learning_rate": 0.00011638696931710948, "loss": 0.6968, "step": 8183 }, { "epoch": 0.46, "learning_rate": 0.00011636884550508127, "loss": 0.7198, "step": 8184 }, { "epoch": 0.46, "learning_rate": 0.00011635072114056162, "loss": 0.8147, "step": 8185 }, { "epoch": 0.46, "learning_rate": 0.00011633259622416224, "loss": 0.6986, "step": 8186 }, { "epoch": 0.46, "learning_rate": 0.00011631447075649489, "loss": 0.8432, "step": 8187 }, { "epoch": 0.46, "learning_rate": 0.00011629634473817137, "loss": 0.699, "step": 8188 }, { "epoch": 0.46, "learning_rate": 0.00011627821816980349, "loss": 0.6753, "step": 8189 }, { "epoch": 0.46, "learning_rate": 0.00011626009105200304, "loss": 0.7038, "step": 8190 }, { "epoch": 0.46, "learning_rate": 0.00011624196338538189, "loss": 0.8884, "step": 8191 }, { "epoch": 0.46, "learning_rate": 0.00011622383517055189, "loss": 0.7614, "step": 8192 }, { "epoch": 0.46, "learning_rate": 0.00011620570640812484, "loss": 0.7236, "step": 8193 }, { "epoch": 0.46, "learning_rate": 0.00011618757709871277, "loss": 0.772, "step": 8194 }, { "epoch": 0.46, "learning_rate": 0.00011616944724292746, "loss": 0.7323, "step": 8195 }, { "epoch": 0.46, "learning_rate": 0.00011615131684138096, "loss": 0.6568, "step": 8196 }, { "epoch": 0.46, "learning_rate": 0.00011613318589468511, "loss": 0.748, "step": 8197 }, { "epoch": 0.47, "learning_rate": 0.00011611505440345197, "loss": 0.6917, "step": 8198 }, { "epoch": 0.47, "learning_rate": 0.00011609692236829342, "loss": 0.7559, "step": 8199 }, { "epoch": 0.47, "learning_rate": 0.00011607878978982159, "loss": 0.7952, "step": 8200 }, { "epoch": 0.47, "learning_rate": 0.00011606065666864841, "loss": 0.7872, "step": 8201 }, { "epoch": 0.47, "learning_rate": 0.00011604252300538592, "loss": 0.7368, "step": 8202 }, { "epoch": 0.47, "learning_rate": 0.00011602438880064623, "loss": 0.6752, "step": 8203 }, { "epoch": 0.47, "learning_rate": 0.0001160062540550414, "loss": 0.7563, "step": 8204 }, { "epoch": 0.47, "learning_rate": 0.0001159881187691835, "loss": 0.7573, "step": 8205 }, { "epoch": 0.47, "learning_rate": 0.00011596998294368466, "loss": 0.8019, "step": 8206 }, { "epoch": 0.47, "learning_rate": 0.00011595184657915703, "loss": 0.6196, "step": 8207 }, { "epoch": 0.47, "learning_rate": 0.00011593370967621274, "loss": 0.7996, "step": 8208 }, { "epoch": 0.47, "learning_rate": 0.00011591557223546395, "loss": 0.7802, "step": 8209 }, { "epoch": 0.47, "learning_rate": 0.00011589743425752286, "loss": 0.783, "step": 8210 }, { "epoch": 0.47, "learning_rate": 0.0001158792957430017, "loss": 0.6347, "step": 8211 }, { "epoch": 0.47, "learning_rate": 0.00011586115669251261, "loss": 0.7397, "step": 8212 }, { "epoch": 0.47, "learning_rate": 0.0001158430171066679, "loss": 0.6981, "step": 8213 }, { "epoch": 0.47, "learning_rate": 0.00011582487698607985, "loss": 0.7398, "step": 8214 }, { "epoch": 0.47, "learning_rate": 0.00011580673633136065, "loss": 0.7673, "step": 8215 }, { "epoch": 0.47, "learning_rate": 0.00011578859514312266, "loss": 0.6641, "step": 8216 }, { "epoch": 0.47, "learning_rate": 0.00011577045342197821, "loss": 0.8038, "step": 8217 }, { "epoch": 0.47, "learning_rate": 0.00011575231116853957, "loss": 0.6553, "step": 8218 }, { "epoch": 0.47, "learning_rate": 0.0001157341683834191, "loss": 0.6976, "step": 8219 }, { "epoch": 0.47, "learning_rate": 0.0001157160250672292, "loss": 0.707, "step": 8220 }, { "epoch": 0.47, "learning_rate": 0.00011569788122058225, "loss": 0.7348, "step": 8221 }, { "epoch": 0.47, "learning_rate": 0.00011567973684409064, "loss": 0.8038, "step": 8222 }, { "epoch": 0.47, "learning_rate": 0.00011566159193836675, "loss": 0.7016, "step": 8223 }, { "epoch": 0.47, "learning_rate": 0.0001156434465040231, "loss": 0.6345, "step": 8224 }, { "epoch": 0.47, "learning_rate": 0.00011562530054167207, "loss": 0.6943, "step": 8225 }, { "epoch": 0.47, "learning_rate": 0.0001156071540519262, "loss": 0.6829, "step": 8226 }, { "epoch": 0.47, "learning_rate": 0.00011558900703539795, "loss": 0.6725, "step": 8227 }, { "epoch": 0.47, "learning_rate": 0.00011557085949269981, "loss": 0.6891, "step": 8228 }, { "epoch": 0.47, "learning_rate": 0.00011555271142444433, "loss": 0.7336, "step": 8229 }, { "epoch": 0.47, "learning_rate": 0.00011553456283124406, "loss": 0.6881, "step": 8230 }, { "epoch": 0.47, "learning_rate": 0.00011551641371371153, "loss": 0.7116, "step": 8231 }, { "epoch": 0.47, "learning_rate": 0.00011549826407245936, "loss": 0.7363, "step": 8232 }, { "epoch": 0.47, "learning_rate": 0.00011548011390810017, "loss": 0.713, "step": 8233 }, { "epoch": 0.47, "learning_rate": 0.00011546196322124647, "loss": 0.7297, "step": 8234 }, { "epoch": 0.47, "learning_rate": 0.00011544381201251098, "loss": 0.7398, "step": 8235 }, { "epoch": 0.47, "learning_rate": 0.00011542566028250634, "loss": 0.7377, "step": 8236 }, { "epoch": 0.47, "learning_rate": 0.00011540750803184523, "loss": 0.615, "step": 8237 }, { "epoch": 0.47, "learning_rate": 0.00011538935526114029, "loss": 0.727, "step": 8238 }, { "epoch": 0.47, "learning_rate": 0.00011537120197100426, "loss": 0.6492, "step": 8239 }, { "epoch": 0.47, "learning_rate": 0.00011535304816204986, "loss": 0.7285, "step": 8240 }, { "epoch": 0.47, "learning_rate": 0.0001153348938348898, "loss": 0.7404, "step": 8241 }, { "epoch": 0.47, "learning_rate": 0.00011531673899013686, "loss": 0.7162, "step": 8242 }, { "epoch": 0.47, "learning_rate": 0.00011529858362840382, "loss": 0.7109, "step": 8243 }, { "epoch": 0.47, "learning_rate": 0.00011528042775030346, "loss": 0.7802, "step": 8244 }, { "epoch": 0.47, "learning_rate": 0.00011526227135644858, "loss": 0.7989, "step": 8245 }, { "epoch": 0.47, "learning_rate": 0.00011524411444745203, "loss": 0.7438, "step": 8246 }, { "epoch": 0.47, "learning_rate": 0.0001152259570239266, "loss": 0.7334, "step": 8247 }, { "epoch": 0.47, "learning_rate": 0.00011520779908648522, "loss": 0.7024, "step": 8248 }, { "epoch": 0.47, "learning_rate": 0.00011518964063574076, "loss": 0.7898, "step": 8249 }, { "epoch": 0.47, "learning_rate": 0.00011517148167230604, "loss": 0.7536, "step": 8250 }, { "epoch": 0.47, "learning_rate": 0.00011515332219679404, "loss": 0.7112, "step": 8251 }, { "epoch": 0.47, "learning_rate": 0.00011513516220981771, "loss": 0.7161, "step": 8252 }, { "epoch": 0.47, "learning_rate": 0.00011511700171198994, "loss": 0.7829, "step": 8253 }, { "epoch": 0.47, "learning_rate": 0.00011509884070392369, "loss": 0.6956, "step": 8254 }, { "epoch": 0.47, "learning_rate": 0.000115080679186232, "loss": 0.7927, "step": 8255 }, { "epoch": 0.47, "learning_rate": 0.00011506251715952785, "loss": 0.6301, "step": 8256 }, { "epoch": 0.47, "learning_rate": 0.0001150443546244242, "loss": 0.7975, "step": 8257 }, { "epoch": 0.47, "learning_rate": 0.00011502619158153413, "loss": 0.8548, "step": 8258 }, { "epoch": 0.47, "learning_rate": 0.00011500802803147075, "loss": 0.6278, "step": 8259 }, { "epoch": 0.47, "learning_rate": 0.000114989863974847, "loss": 0.7539, "step": 8260 }, { "epoch": 0.47, "learning_rate": 0.00011497169941227604, "loss": 0.6692, "step": 8261 }, { "epoch": 0.47, "learning_rate": 0.00011495353434437098, "loss": 0.7832, "step": 8262 }, { "epoch": 0.47, "learning_rate": 0.0001149353687717449, "loss": 0.7045, "step": 8263 }, { "epoch": 0.47, "learning_rate": 0.00011491720269501093, "loss": 0.7074, "step": 8264 }, { "epoch": 0.47, "learning_rate": 0.00011489903611478229, "loss": 0.7567, "step": 8265 }, { "epoch": 0.47, "learning_rate": 0.00011488086903167213, "loss": 0.705, "step": 8266 }, { "epoch": 0.47, "learning_rate": 0.00011486270144629355, "loss": 0.7367, "step": 8267 }, { "epoch": 0.47, "learning_rate": 0.00011484453335925982, "loss": 0.7135, "step": 8268 }, { "epoch": 0.47, "learning_rate": 0.0001148263647711842, "loss": 0.7118, "step": 8269 }, { "epoch": 0.47, "learning_rate": 0.00011480819568267989, "loss": 0.7549, "step": 8270 }, { "epoch": 0.47, "learning_rate": 0.00011479002609436009, "loss": 0.7607, "step": 8271 }, { "epoch": 0.47, "learning_rate": 0.00011477185600683817, "loss": 0.7331, "step": 8272 }, { "epoch": 0.47, "learning_rate": 0.00011475368542072733, "loss": 0.7269, "step": 8273 }, { "epoch": 0.47, "learning_rate": 0.00011473551433664094, "loss": 0.6829, "step": 8274 }, { "epoch": 0.47, "learning_rate": 0.0001147173427551923, "loss": 0.7005, "step": 8275 }, { "epoch": 0.47, "learning_rate": 0.00011469917067699472, "loss": 0.651, "step": 8276 }, { "epoch": 0.47, "learning_rate": 0.0001146809981026616, "loss": 0.6752, "step": 8277 }, { "epoch": 0.47, "learning_rate": 0.0001146628250328063, "loss": 0.7005, "step": 8278 }, { "epoch": 0.47, "learning_rate": 0.00011464465146804217, "loss": 0.6704, "step": 8279 }, { "epoch": 0.47, "learning_rate": 0.00011462647740898265, "loss": 0.6917, "step": 8280 }, { "epoch": 0.47, "learning_rate": 0.00011460830285624118, "loss": 0.6912, "step": 8281 }, { "epoch": 0.47, "learning_rate": 0.00011459012781043117, "loss": 0.6315, "step": 8282 }, { "epoch": 0.47, "learning_rate": 0.00011457195227216607, "loss": 0.7642, "step": 8283 }, { "epoch": 0.47, "learning_rate": 0.00011455377624205934, "loss": 0.728, "step": 8284 }, { "epoch": 0.47, "learning_rate": 0.00011453559972072457, "loss": 0.7167, "step": 8285 }, { "epoch": 0.47, "learning_rate": 0.00011451742270877513, "loss": 0.5914, "step": 8286 }, { "epoch": 0.47, "learning_rate": 0.0001144992452068246, "loss": 0.6975, "step": 8287 }, { "epoch": 0.47, "learning_rate": 0.00011448106721548654, "loss": 0.7103, "step": 8288 }, { "epoch": 0.47, "learning_rate": 0.0001144628887353745, "loss": 0.7522, "step": 8289 }, { "epoch": 0.47, "learning_rate": 0.00011444470976710198, "loss": 0.6604, "step": 8290 }, { "epoch": 0.47, "learning_rate": 0.00011442653031128267, "loss": 0.7044, "step": 8291 }, { "epoch": 0.47, "learning_rate": 0.00011440835036853012, "loss": 0.661, "step": 8292 }, { "epoch": 0.47, "learning_rate": 0.00011439016993945793, "loss": 0.7445, "step": 8293 }, { "epoch": 0.47, "learning_rate": 0.0001143719890246798, "loss": 0.7327, "step": 8294 }, { "epoch": 0.47, "learning_rate": 0.00011435380762480932, "loss": 0.574, "step": 8295 }, { "epoch": 0.47, "learning_rate": 0.00011433562574046019, "loss": 0.6611, "step": 8296 }, { "epoch": 0.47, "learning_rate": 0.0001143174433722461, "loss": 0.7247, "step": 8297 }, { "epoch": 0.47, "learning_rate": 0.00011429926052078077, "loss": 0.7846, "step": 8298 }, { "epoch": 0.47, "learning_rate": 0.00011428107718667787, "loss": 0.6486, "step": 8299 }, { "epoch": 0.47, "learning_rate": 0.00011426289337055119, "loss": 0.7657, "step": 8300 }, { "epoch": 0.47, "learning_rate": 0.00011424470907301443, "loss": 0.7742, "step": 8301 }, { "epoch": 0.47, "learning_rate": 0.00011422652429468137, "loss": 0.823, "step": 8302 }, { "epoch": 0.47, "learning_rate": 0.00011420833903616583, "loss": 0.6837, "step": 8303 }, { "epoch": 0.47, "learning_rate": 0.00011419015329808157, "loss": 0.7645, "step": 8304 }, { "epoch": 0.47, "learning_rate": 0.00011417196708104243, "loss": 0.5892, "step": 8305 }, { "epoch": 0.47, "learning_rate": 0.0001141537803856622, "loss": 0.7474, "step": 8306 }, { "epoch": 0.47, "learning_rate": 0.0001141355932125548, "loss": 0.7412, "step": 8307 }, { "epoch": 0.47, "learning_rate": 0.00011411740556233406, "loss": 0.7069, "step": 8308 }, { "epoch": 0.47, "learning_rate": 0.00011409921743561383, "loss": 0.6726, "step": 8309 }, { "epoch": 0.47, "learning_rate": 0.00011408102883300806, "loss": 0.7256, "step": 8310 }, { "epoch": 0.47, "learning_rate": 0.00011406283975513063, "loss": 0.6984, "step": 8311 }, { "epoch": 0.47, "learning_rate": 0.00011404465020259547, "loss": 0.733, "step": 8312 }, { "epoch": 0.47, "learning_rate": 0.00011402646017601652, "loss": 0.5906, "step": 8313 }, { "epoch": 0.47, "learning_rate": 0.0001140082696760078, "loss": 0.7472, "step": 8314 }, { "epoch": 0.47, "learning_rate": 0.00011399007870318321, "loss": 0.7429, "step": 8315 }, { "epoch": 0.47, "learning_rate": 0.00011397188725815677, "loss": 0.6829, "step": 8316 }, { "epoch": 0.47, "learning_rate": 0.0001139536953415425, "loss": 0.6953, "step": 8317 }, { "epoch": 0.47, "learning_rate": 0.00011393550295395443, "loss": 0.6441, "step": 8318 }, { "epoch": 0.47, "learning_rate": 0.00011391731009600654, "loss": 0.8023, "step": 8319 }, { "epoch": 0.47, "learning_rate": 0.000113899116768313, "loss": 0.8709, "step": 8320 }, { "epoch": 0.47, "learning_rate": 0.00011388092297148776, "loss": 0.7566, "step": 8321 }, { "epoch": 0.47, "learning_rate": 0.000113862728706145, "loss": 0.7815, "step": 8322 }, { "epoch": 0.47, "learning_rate": 0.00011384453397289876, "loss": 0.6703, "step": 8323 }, { "epoch": 0.47, "learning_rate": 0.00011382633877236322, "loss": 0.7391, "step": 8324 }, { "epoch": 0.47, "learning_rate": 0.00011380814310515246, "loss": 0.7393, "step": 8325 }, { "epoch": 0.47, "learning_rate": 0.0001137899469718807, "loss": 0.7053, "step": 8326 }, { "epoch": 0.47, "learning_rate": 0.00011377175037316204, "loss": 0.7081, "step": 8327 }, { "epoch": 0.47, "learning_rate": 0.00011375355330961066, "loss": 0.7874, "step": 8328 }, { "epoch": 0.47, "learning_rate": 0.00011373535578184082, "loss": 0.7413, "step": 8329 }, { "epoch": 0.47, "learning_rate": 0.00011371715779046672, "loss": 0.6561, "step": 8330 }, { "epoch": 0.47, "learning_rate": 0.00011369895933610251, "loss": 0.7157, "step": 8331 }, { "epoch": 0.47, "learning_rate": 0.00011368076041936251, "loss": 0.7451, "step": 8332 }, { "epoch": 0.47, "learning_rate": 0.00011366256104086101, "loss": 0.7192, "step": 8333 }, { "epoch": 0.47, "learning_rate": 0.00011364436120121222, "loss": 0.6239, "step": 8334 }, { "epoch": 0.47, "learning_rate": 0.00011362616090103044, "loss": 0.7625, "step": 8335 }, { "epoch": 0.47, "learning_rate": 0.00011360796014093, "loss": 0.7083, "step": 8336 }, { "epoch": 0.47, "learning_rate": 0.00011358975892152523, "loss": 0.6809, "step": 8337 }, { "epoch": 0.47, "learning_rate": 0.00011357155724343045, "loss": 0.7526, "step": 8338 }, { "epoch": 0.47, "learning_rate": 0.00011355335510726002, "loss": 0.8132, "step": 8339 }, { "epoch": 0.47, "learning_rate": 0.00011353515251362832, "loss": 0.7463, "step": 8340 }, { "epoch": 0.47, "learning_rate": 0.0001135169494631497, "loss": 0.6233, "step": 8341 }, { "epoch": 0.47, "learning_rate": 0.00011349874595643859, "loss": 0.8095, "step": 8342 }, { "epoch": 0.47, "learning_rate": 0.00011348054199410943, "loss": 0.7164, "step": 8343 }, { "epoch": 0.47, "learning_rate": 0.0001134623375767766, "loss": 0.7217, "step": 8344 }, { "epoch": 0.47, "learning_rate": 0.00011344413270505457, "loss": 0.7573, "step": 8345 }, { "epoch": 0.47, "learning_rate": 0.00011342592737955782, "loss": 0.73, "step": 8346 }, { "epoch": 0.47, "learning_rate": 0.00011340772160090078, "loss": 0.7164, "step": 8347 }, { "epoch": 0.47, "learning_rate": 0.00011338951536969802, "loss": 0.7465, "step": 8348 }, { "epoch": 0.47, "learning_rate": 0.00011337130868656398, "loss": 0.7357, "step": 8349 }, { "epoch": 0.47, "learning_rate": 0.00011335310155211318, "loss": 0.7114, "step": 8350 }, { "epoch": 0.47, "learning_rate": 0.00011333489396696017, "loss": 0.6751, "step": 8351 }, { "epoch": 0.47, "learning_rate": 0.00011331668593171958, "loss": 0.6957, "step": 8352 }, { "epoch": 0.47, "learning_rate": 0.00011329847744700588, "loss": 0.7311, "step": 8353 }, { "epoch": 0.47, "learning_rate": 0.00011328026851343367, "loss": 0.7252, "step": 8354 }, { "epoch": 0.47, "learning_rate": 0.00011326205913161763, "loss": 0.7564, "step": 8355 }, { "epoch": 0.47, "learning_rate": 0.00011324384930217226, "loss": 0.6476, "step": 8356 }, { "epoch": 0.47, "learning_rate": 0.00011322563902571226, "loss": 0.8422, "step": 8357 }, { "epoch": 0.47, "learning_rate": 0.00011320742830285223, "loss": 0.7117, "step": 8358 }, { "epoch": 0.47, "learning_rate": 0.00011318921713420691, "loss": 0.7261, "step": 8359 }, { "epoch": 0.47, "learning_rate": 0.00011317100552039088, "loss": 0.7445, "step": 8360 }, { "epoch": 0.47, "learning_rate": 0.00011315279346201888, "loss": 0.734, "step": 8361 }, { "epoch": 0.47, "learning_rate": 0.00011313458095970561, "loss": 0.6731, "step": 8362 }, { "epoch": 0.47, "learning_rate": 0.0001131163680140658, "loss": 0.7248, "step": 8363 }, { "epoch": 0.47, "learning_rate": 0.00011309815462571416, "loss": 0.6738, "step": 8364 }, { "epoch": 0.47, "learning_rate": 0.00011307994079526546, "loss": 0.6862, "step": 8365 }, { "epoch": 0.47, "learning_rate": 0.00011306172652333445, "loss": 0.6401, "step": 8366 }, { "epoch": 0.47, "learning_rate": 0.00011304351181053587, "loss": 0.6506, "step": 8367 }, { "epoch": 0.47, "learning_rate": 0.0001130252966574846, "loss": 0.7982, "step": 8368 }, { "epoch": 0.47, "learning_rate": 0.0001130070810647954, "loss": 0.6855, "step": 8369 }, { "epoch": 0.47, "learning_rate": 0.00011298886503308312, "loss": 0.7059, "step": 8370 }, { "epoch": 0.47, "learning_rate": 0.00011297064856296256, "loss": 0.6974, "step": 8371 }, { "epoch": 0.47, "learning_rate": 0.0001129524316550486, "loss": 0.7166, "step": 8372 }, { "epoch": 0.47, "learning_rate": 0.00011293421430995609, "loss": 0.6897, "step": 8373 }, { "epoch": 0.48, "learning_rate": 0.00011291599652829994, "loss": 0.6178, "step": 8374 }, { "epoch": 0.48, "learning_rate": 0.00011289777831069505, "loss": 0.7758, "step": 8375 }, { "epoch": 0.48, "learning_rate": 0.0001128795596577563, "loss": 0.6485, "step": 8376 }, { "epoch": 0.48, "learning_rate": 0.00011286134057009863, "loss": 0.7382, "step": 8377 }, { "epoch": 0.48, "learning_rate": 0.00011284312104833698, "loss": 0.7459, "step": 8378 }, { "epoch": 0.48, "learning_rate": 0.00011282490109308633, "loss": 0.7216, "step": 8379 }, { "epoch": 0.48, "learning_rate": 0.00011280668070496161, "loss": 0.7657, "step": 8380 }, { "epoch": 0.48, "learning_rate": 0.00011278845988457785, "loss": 0.6797, "step": 8381 }, { "epoch": 0.48, "learning_rate": 0.00011277023863255003, "loss": 0.6227, "step": 8382 }, { "epoch": 0.48, "learning_rate": 0.00011275201694949313, "loss": 0.7342, "step": 8383 }, { "epoch": 0.48, "learning_rate": 0.00011273379483602227, "loss": 0.6885, "step": 8384 }, { "epoch": 0.48, "learning_rate": 0.00011271557229275241, "loss": 0.8102, "step": 8385 }, { "epoch": 0.48, "learning_rate": 0.00011269734932029863, "loss": 0.7041, "step": 8386 }, { "epoch": 0.48, "learning_rate": 0.000112679125919276, "loss": 0.7482, "step": 8387 }, { "epoch": 0.48, "learning_rate": 0.00011266090209029964, "loss": 0.7102, "step": 8388 }, { "epoch": 0.48, "learning_rate": 0.00011264267783398463, "loss": 0.6988, "step": 8389 }, { "epoch": 0.48, "learning_rate": 0.00011262445315094607, "loss": 0.6976, "step": 8390 }, { "epoch": 0.48, "learning_rate": 0.00011260622804179915, "loss": 0.6665, "step": 8391 }, { "epoch": 0.48, "learning_rate": 0.00011258800250715895, "loss": 0.7234, "step": 8392 }, { "epoch": 0.48, "learning_rate": 0.00011256977654764063, "loss": 0.7547, "step": 8393 }, { "epoch": 0.48, "learning_rate": 0.0001125515501638594, "loss": 0.6681, "step": 8394 }, { "epoch": 0.48, "learning_rate": 0.00011253332335643043, "loss": 0.7324, "step": 8395 }, { "epoch": 0.48, "learning_rate": 0.00011251509612596894, "loss": 0.6916, "step": 8396 }, { "epoch": 0.48, "learning_rate": 0.00011249686847309012, "loss": 0.6949, "step": 8397 }, { "epoch": 0.48, "learning_rate": 0.00011247864039840923, "loss": 0.7774, "step": 8398 }, { "epoch": 0.48, "learning_rate": 0.00011246041190254146, "loss": 0.7543, "step": 8399 }, { "epoch": 0.48, "learning_rate": 0.00011244218298610216, "loss": 0.7064, "step": 8400 }, { "epoch": 0.48, "learning_rate": 0.00011242395364970654, "loss": 0.7091, "step": 8401 }, { "epoch": 0.48, "learning_rate": 0.00011240572389396988, "loss": 0.7301, "step": 8402 }, { "epoch": 0.48, "learning_rate": 0.00011238749371950754, "loss": 0.7617, "step": 8403 }, { "epoch": 0.48, "learning_rate": 0.00011236926312693479, "loss": 0.7692, "step": 8404 }, { "epoch": 0.48, "learning_rate": 0.00011235103211686695, "loss": 0.7397, "step": 8405 }, { "epoch": 0.48, "learning_rate": 0.00011233280068991936, "loss": 0.7201, "step": 8406 }, { "epoch": 0.48, "learning_rate": 0.00011231456884670746, "loss": 0.7691, "step": 8407 }, { "epoch": 0.48, "learning_rate": 0.00011229633658784655, "loss": 0.7131, "step": 8408 }, { "epoch": 0.48, "learning_rate": 0.000112278103913952, "loss": 0.7702, "step": 8409 }, { "epoch": 0.48, "learning_rate": 0.00011225987082563927, "loss": 0.7413, "step": 8410 }, { "epoch": 0.48, "learning_rate": 0.00011224163732352375, "loss": 0.7874, "step": 8411 }, { "epoch": 0.48, "learning_rate": 0.00011222340340822086, "loss": 0.6731, "step": 8412 }, { "epoch": 0.48, "learning_rate": 0.00011220516908034601, "loss": 0.761, "step": 8413 }, { "epoch": 0.48, "learning_rate": 0.00011218693434051475, "loss": 0.6535, "step": 8414 }, { "epoch": 0.48, "learning_rate": 0.00011216869918934249, "loss": 0.7168, "step": 8415 }, { "epoch": 0.48, "learning_rate": 0.00011215046362744468, "loss": 0.6977, "step": 8416 }, { "epoch": 0.48, "learning_rate": 0.00011213222765543689, "loss": 0.6983, "step": 8417 }, { "epoch": 0.48, "learning_rate": 0.0001121139912739346, "loss": 0.7203, "step": 8418 }, { "epoch": 0.48, "learning_rate": 0.00011209575448355331, "loss": 0.7732, "step": 8419 }, { "epoch": 0.48, "learning_rate": 0.00011207751728490861, "loss": 0.6787, "step": 8420 }, { "epoch": 0.48, "learning_rate": 0.000112059279678616, "loss": 0.6171, "step": 8421 }, { "epoch": 0.48, "learning_rate": 0.0001120410416652911, "loss": 0.757, "step": 8422 }, { "epoch": 0.48, "learning_rate": 0.00011202280324554945, "loss": 0.7939, "step": 8423 }, { "epoch": 0.48, "learning_rate": 0.00011200456442000666, "loss": 0.7822, "step": 8424 }, { "epoch": 0.48, "learning_rate": 0.00011198632518927832, "loss": 0.7624, "step": 8425 }, { "epoch": 0.48, "learning_rate": 0.0001119680855539801, "loss": 0.7261, "step": 8426 }, { "epoch": 0.48, "learning_rate": 0.0001119498455147276, "loss": 0.6765, "step": 8427 }, { "epoch": 0.48, "learning_rate": 0.00011193160507213647, "loss": 0.5785, "step": 8428 }, { "epoch": 0.48, "learning_rate": 0.00011191336422682237, "loss": 0.7231, "step": 8429 }, { "epoch": 0.48, "learning_rate": 0.00011189512297940101, "loss": 0.8011, "step": 8430 }, { "epoch": 0.48, "learning_rate": 0.00011187688133048801, "loss": 0.6897, "step": 8431 }, { "epoch": 0.48, "learning_rate": 0.00011185863928069912, "loss": 0.7158, "step": 8432 }, { "epoch": 0.48, "learning_rate": 0.00011184039683065013, "loss": 0.7595, "step": 8433 }, { "epoch": 0.48, "learning_rate": 0.00011182215398095661, "loss": 0.6112, "step": 8434 }, { "epoch": 0.48, "learning_rate": 0.00011180391073223442, "loss": 0.652, "step": 8435 }, { "epoch": 0.48, "learning_rate": 0.00011178566708509928, "loss": 0.7945, "step": 8436 }, { "epoch": 0.48, "learning_rate": 0.00011176742304016699, "loss": 0.7501, "step": 8437 }, { "epoch": 0.48, "learning_rate": 0.00011174917859805327, "loss": 0.7354, "step": 8438 }, { "epoch": 0.48, "learning_rate": 0.00011173093375937398, "loss": 0.6991, "step": 8439 }, { "epoch": 0.48, "learning_rate": 0.00011171268852474493, "loss": 0.7729, "step": 8440 }, { "epoch": 0.48, "learning_rate": 0.00011169444289478188, "loss": 0.7412, "step": 8441 }, { "epoch": 0.48, "learning_rate": 0.00011167619687010074, "loss": 0.7738, "step": 8442 }, { "epoch": 0.48, "learning_rate": 0.00011165795045131733, "loss": 0.7453, "step": 8443 }, { "epoch": 0.48, "learning_rate": 0.00011163970363904752, "loss": 0.6449, "step": 8444 }, { "epoch": 0.48, "learning_rate": 0.00011162145643390718, "loss": 0.7205, "step": 8445 }, { "epoch": 0.48, "learning_rate": 0.00011160320883651223, "loss": 0.7402, "step": 8446 }, { "epoch": 0.48, "learning_rate": 0.00011158496084747852, "loss": 0.6732, "step": 8447 }, { "epoch": 0.48, "learning_rate": 0.00011156671246742201, "loss": 0.718, "step": 8448 }, { "epoch": 0.48, "learning_rate": 0.00011154846369695863, "loss": 0.7155, "step": 8449 }, { "epoch": 0.48, "learning_rate": 0.00011153021453670429, "loss": 0.7476, "step": 8450 }, { "epoch": 0.48, "learning_rate": 0.00011151196498727497, "loss": 0.7957, "step": 8451 }, { "epoch": 0.48, "learning_rate": 0.00011149371504928668, "loss": 0.81, "step": 8452 }, { "epoch": 0.48, "learning_rate": 0.00011147546472335532, "loss": 0.6956, "step": 8453 }, { "epoch": 0.48, "learning_rate": 0.00011145721401009694, "loss": 0.7244, "step": 8454 }, { "epoch": 0.48, "learning_rate": 0.00011143896291012757, "loss": 0.7162, "step": 8455 }, { "epoch": 0.48, "learning_rate": 0.00011142071142406319, "loss": 0.8062, "step": 8456 }, { "epoch": 0.48, "learning_rate": 0.00011140245955251982, "loss": 0.6798, "step": 8457 }, { "epoch": 0.48, "learning_rate": 0.00011138420729611355, "loss": 0.7239, "step": 8458 }, { "epoch": 0.48, "learning_rate": 0.00011136595465546045, "loss": 0.7871, "step": 8459 }, { "epoch": 0.48, "learning_rate": 0.00011134770163117654, "loss": 0.6467, "step": 8460 }, { "epoch": 0.48, "learning_rate": 0.00011132944822387794, "loss": 0.6623, "step": 8461 }, { "epoch": 0.48, "learning_rate": 0.00011131119443418077, "loss": 0.7578, "step": 8462 }, { "epoch": 0.48, "learning_rate": 0.00011129294026270115, "loss": 0.7786, "step": 8463 }, { "epoch": 0.48, "learning_rate": 0.00011127468571005511, "loss": 0.709, "step": 8464 }, { "epoch": 0.48, "learning_rate": 0.0001112564307768589, "loss": 0.8122, "step": 8465 }, { "epoch": 0.48, "learning_rate": 0.00011123817546372864, "loss": 0.7778, "step": 8466 }, { "epoch": 0.48, "learning_rate": 0.00011121991977128045, "loss": 0.7594, "step": 8467 }, { "epoch": 0.48, "learning_rate": 0.00011120166370013055, "loss": 0.75, "step": 8468 }, { "epoch": 0.48, "learning_rate": 0.00011118340725089515, "loss": 0.7033, "step": 8469 }, { "epoch": 0.48, "learning_rate": 0.00011116515042419042, "loss": 0.7358, "step": 8470 }, { "epoch": 0.48, "learning_rate": 0.00011114689322063255, "loss": 0.7081, "step": 8471 }, { "epoch": 0.48, "learning_rate": 0.00011112863564083784, "loss": 0.7453, "step": 8472 }, { "epoch": 0.48, "learning_rate": 0.00011111037768542245, "loss": 0.6663, "step": 8473 }, { "epoch": 0.48, "learning_rate": 0.0001110921193550027, "loss": 0.7282, "step": 8474 }, { "epoch": 0.48, "learning_rate": 0.00011107386065019485, "loss": 0.7644, "step": 8475 }, { "epoch": 0.48, "learning_rate": 0.00011105560157161513, "loss": 0.637, "step": 8476 }, { "epoch": 0.48, "learning_rate": 0.00011103734211987986, "loss": 0.7638, "step": 8477 }, { "epoch": 0.48, "learning_rate": 0.00011101908229560538, "loss": 0.7873, "step": 8478 }, { "epoch": 0.48, "learning_rate": 0.00011100082209940795, "loss": 0.6753, "step": 8479 }, { "epoch": 0.48, "learning_rate": 0.00011098256153190392, "loss": 0.6256, "step": 8480 }, { "epoch": 0.48, "learning_rate": 0.00011096430059370964, "loss": 0.668, "step": 8481 }, { "epoch": 0.48, "learning_rate": 0.00011094603928544147, "loss": 0.7095, "step": 8482 }, { "epoch": 0.48, "learning_rate": 0.00011092777760771573, "loss": 0.7816, "step": 8483 }, { "epoch": 0.48, "learning_rate": 0.00011090951556114889, "loss": 0.6592, "step": 8484 }, { "epoch": 0.48, "learning_rate": 0.00011089125314635726, "loss": 0.7194, "step": 8485 }, { "epoch": 0.48, "learning_rate": 0.00011087299036395728, "loss": 0.7416, "step": 8486 }, { "epoch": 0.48, "learning_rate": 0.00011085472721456533, "loss": 0.6601, "step": 8487 }, { "epoch": 0.48, "learning_rate": 0.00011083646369879793, "loss": 0.6695, "step": 8488 }, { "epoch": 0.48, "learning_rate": 0.00011081819981727142, "loss": 0.6453, "step": 8489 }, { "epoch": 0.48, "learning_rate": 0.0001107999355706023, "loss": 0.6768, "step": 8490 }, { "epoch": 0.48, "learning_rate": 0.00011078167095940703, "loss": 0.7484, "step": 8491 }, { "epoch": 0.48, "learning_rate": 0.00011076340598430211, "loss": 0.7928, "step": 8492 }, { "epoch": 0.48, "learning_rate": 0.00011074514064590398, "loss": 0.7509, "step": 8493 }, { "epoch": 0.48, "learning_rate": 0.00011072687494482919, "loss": 0.7167, "step": 8494 }, { "epoch": 0.48, "learning_rate": 0.00011070860888169421, "loss": 0.785, "step": 8495 }, { "epoch": 0.48, "learning_rate": 0.00011069034245711563, "loss": 0.6991, "step": 8496 }, { "epoch": 0.48, "learning_rate": 0.00011067207567170992, "loss": 0.6379, "step": 8497 }, { "epoch": 0.48, "learning_rate": 0.0001106538085260937, "loss": 0.7494, "step": 8498 }, { "epoch": 0.48, "learning_rate": 0.00011063554102088349, "loss": 0.6364, "step": 8499 }, { "epoch": 0.48, "learning_rate": 0.00011061727315669587, "loss": 0.7386, "step": 8500 }, { "epoch": 0.48, "learning_rate": 0.00011059900493414748, "loss": 0.7162, "step": 8501 }, { "epoch": 0.48, "learning_rate": 0.00011058073635385482, "loss": 0.7093, "step": 8502 }, { "epoch": 0.48, "learning_rate": 0.0001105624674164346, "loss": 0.7371, "step": 8503 }, { "epoch": 0.48, "learning_rate": 0.00011054419812250338, "loss": 0.758, "step": 8504 }, { "epoch": 0.48, "learning_rate": 0.00011052592847267781, "loss": 0.6889, "step": 8505 }, { "epoch": 0.48, "learning_rate": 0.00011050765846757456, "loss": 0.7571, "step": 8506 }, { "epoch": 0.48, "learning_rate": 0.00011048938810781029, "loss": 0.6834, "step": 8507 }, { "epoch": 0.48, "learning_rate": 0.00011047111739400167, "loss": 0.5833, "step": 8508 }, { "epoch": 0.48, "learning_rate": 0.00011045284632676536, "loss": 0.7502, "step": 8509 }, { "epoch": 0.48, "learning_rate": 0.00011043457490671806, "loss": 0.7332, "step": 8510 }, { "epoch": 0.48, "learning_rate": 0.00011041630313447653, "loss": 0.727, "step": 8511 }, { "epoch": 0.48, "learning_rate": 0.00011039803101065742, "loss": 0.7226, "step": 8512 }, { "epoch": 0.48, "learning_rate": 0.00011037975853587747, "loss": 0.6921, "step": 8513 }, { "epoch": 0.48, "learning_rate": 0.00011036148571075351, "loss": 0.7532, "step": 8514 }, { "epoch": 0.48, "learning_rate": 0.0001103432125359022, "loss": 0.8256, "step": 8515 }, { "epoch": 0.48, "learning_rate": 0.00011032493901194034, "loss": 0.8014, "step": 8516 }, { "epoch": 0.48, "learning_rate": 0.0001103066651394847, "loss": 0.6633, "step": 8517 }, { "epoch": 0.48, "learning_rate": 0.00011028839091915212, "loss": 0.7924, "step": 8518 }, { "epoch": 0.48, "learning_rate": 0.00011027011635155933, "loss": 0.7358, "step": 8519 }, { "epoch": 0.48, "learning_rate": 0.00011025184143732319, "loss": 0.6827, "step": 8520 }, { "epoch": 0.48, "learning_rate": 0.00011023356617706052, "loss": 0.7264, "step": 8521 }, { "epoch": 0.48, "learning_rate": 0.00011021529057138813, "loss": 0.716, "step": 8522 }, { "epoch": 0.48, "learning_rate": 0.00011019701462092294, "loss": 0.7111, "step": 8523 }, { "epoch": 0.48, "learning_rate": 0.00011017873832628172, "loss": 0.8005, "step": 8524 }, { "epoch": 0.48, "learning_rate": 0.00011016046168808136, "loss": 0.7335, "step": 8525 }, { "epoch": 0.48, "learning_rate": 0.00011014218470693882, "loss": 0.8176, "step": 8526 }, { "epoch": 0.48, "learning_rate": 0.00011012390738347096, "loss": 0.6535, "step": 8527 }, { "epoch": 0.48, "learning_rate": 0.00011010562971829463, "loss": 0.6341, "step": 8528 }, { "epoch": 0.48, "learning_rate": 0.00011008735171202684, "loss": 0.7491, "step": 8529 }, { "epoch": 0.48, "learning_rate": 0.00011006907336528445, "loss": 0.6548, "step": 8530 }, { "epoch": 0.48, "learning_rate": 0.00011005079467868442, "loss": 0.6502, "step": 8531 }, { "epoch": 0.48, "learning_rate": 0.00011003251565284371, "loss": 0.6933, "step": 8532 }, { "epoch": 0.48, "learning_rate": 0.00011001423628837932, "loss": 0.7258, "step": 8533 }, { "epoch": 0.48, "learning_rate": 0.00010999595658590816, "loss": 0.7558, "step": 8534 }, { "epoch": 0.48, "learning_rate": 0.00010997767654604722, "loss": 0.7469, "step": 8535 }, { "epoch": 0.48, "learning_rate": 0.0001099593961694136, "loss": 0.6935, "step": 8536 }, { "epoch": 0.48, "learning_rate": 0.0001099411154566242, "loss": 0.707, "step": 8537 }, { "epoch": 0.48, "learning_rate": 0.00010992283440829607, "loss": 0.7108, "step": 8538 }, { "epoch": 0.48, "learning_rate": 0.0001099045530250463, "loss": 0.6711, "step": 8539 }, { "epoch": 0.48, "learning_rate": 0.00010988627130749185, "loss": 0.7779, "step": 8540 }, { "epoch": 0.48, "learning_rate": 0.00010986798925624982, "loss": 0.7166, "step": 8541 }, { "epoch": 0.48, "learning_rate": 0.00010984970687193727, "loss": 0.6833, "step": 8542 }, { "epoch": 0.48, "learning_rate": 0.0001098314241551713, "loss": 0.7064, "step": 8543 }, { "epoch": 0.48, "learning_rate": 0.00010981314110656898, "loss": 0.7908, "step": 8544 }, { "epoch": 0.48, "learning_rate": 0.0001097948577267474, "loss": 0.773, "step": 8545 }, { "epoch": 0.48, "learning_rate": 0.00010977657401632369, "loss": 0.6537, "step": 8546 }, { "epoch": 0.48, "learning_rate": 0.00010975828997591495, "loss": 0.6973, "step": 8547 }, { "epoch": 0.48, "learning_rate": 0.00010974000560613835, "loss": 0.7847, "step": 8548 }, { "epoch": 0.48, "learning_rate": 0.00010972172090761103, "loss": 0.7283, "step": 8549 }, { "epoch": 0.48, "learning_rate": 0.00010970343588095008, "loss": 0.7097, "step": 8550 }, { "epoch": 0.49, "learning_rate": 0.00010968515052677277, "loss": 0.8011, "step": 8551 }, { "epoch": 0.49, "learning_rate": 0.00010966686484569621, "loss": 0.7106, "step": 8552 }, { "epoch": 0.49, "learning_rate": 0.00010964857883833763, "loss": 0.6431, "step": 8553 }, { "epoch": 0.49, "learning_rate": 0.00010963029250531418, "loss": 0.6479, "step": 8554 }, { "epoch": 0.49, "learning_rate": 0.00010961200584724313, "loss": 0.7257, "step": 8555 }, { "epoch": 0.49, "learning_rate": 0.00010959371886474168, "loss": 0.8637, "step": 8556 }, { "epoch": 0.49, "learning_rate": 0.00010957543155842702, "loss": 0.7018, "step": 8557 }, { "epoch": 0.49, "learning_rate": 0.00010955714392891643, "loss": 0.7811, "step": 8558 }, { "epoch": 0.49, "learning_rate": 0.00010953885597682723, "loss": 0.6988, "step": 8559 }, { "epoch": 0.49, "learning_rate": 0.00010952056770277656, "loss": 0.7163, "step": 8560 }, { "epoch": 0.49, "learning_rate": 0.00010950227910738177, "loss": 0.6775, "step": 8561 }, { "epoch": 0.49, "learning_rate": 0.00010948399019126017, "loss": 0.7476, "step": 8562 }, { "epoch": 0.49, "learning_rate": 0.00010946570095502902, "loss": 0.7498, "step": 8563 }, { "epoch": 0.49, "learning_rate": 0.00010944741139930561, "loss": 0.6961, "step": 8564 }, { "epoch": 0.49, "learning_rate": 0.00010942912152470733, "loss": 0.7835, "step": 8565 }, { "epoch": 0.49, "learning_rate": 0.00010941083133185146, "loss": 0.7901, "step": 8566 }, { "epoch": 0.49, "learning_rate": 0.00010939254082135532, "loss": 0.7187, "step": 8567 }, { "epoch": 0.49, "learning_rate": 0.00010937424999383631, "loss": 0.7623, "step": 8568 }, { "epoch": 0.49, "learning_rate": 0.00010935595884991178, "loss": 0.6534, "step": 8569 }, { "epoch": 0.49, "learning_rate": 0.00010933766739019912, "loss": 0.6806, "step": 8570 }, { "epoch": 0.49, "learning_rate": 0.00010931937561531567, "loss": 0.6593, "step": 8571 }, { "epoch": 0.49, "learning_rate": 0.00010930108352587887, "loss": 0.7173, "step": 8572 }, { "epoch": 0.49, "learning_rate": 0.0001092827911225061, "loss": 0.768, "step": 8573 }, { "epoch": 0.49, "learning_rate": 0.00010926449840581479, "loss": 0.6898, "step": 8574 }, { "epoch": 0.49, "learning_rate": 0.00010924620537642236, "loss": 0.7051, "step": 8575 }, { "epoch": 0.49, "learning_rate": 0.00010922791203494624, "loss": 0.6895, "step": 8576 }, { "epoch": 0.49, "learning_rate": 0.00010920961838200388, "loss": 0.7532, "step": 8577 }, { "epoch": 0.49, "learning_rate": 0.0001091913244182128, "loss": 0.6781, "step": 8578 }, { "epoch": 0.49, "learning_rate": 0.00010917303014419036, "loss": 0.7729, "step": 8579 }, { "epoch": 0.49, "learning_rate": 0.0001091547355605541, "loss": 0.7814, "step": 8580 }, { "epoch": 0.49, "learning_rate": 0.00010913644066792153, "loss": 0.673, "step": 8581 }, { "epoch": 0.49, "learning_rate": 0.00010911814546691013, "loss": 0.7133, "step": 8582 }, { "epoch": 0.49, "learning_rate": 0.00010909984995813738, "loss": 0.8065, "step": 8583 }, { "epoch": 0.49, "learning_rate": 0.00010908155414222083, "loss": 0.5879, "step": 8584 }, { "epoch": 0.49, "learning_rate": 0.00010906325801977804, "loss": 0.6677, "step": 8585 }, { "epoch": 0.49, "learning_rate": 0.0001090449615914265, "loss": 0.7966, "step": 8586 }, { "epoch": 0.49, "learning_rate": 0.00010902666485778377, "loss": 0.7585, "step": 8587 }, { "epoch": 0.49, "learning_rate": 0.00010900836781946747, "loss": 0.8471, "step": 8588 }, { "epoch": 0.49, "learning_rate": 0.0001089900704770951, "loss": 0.7011, "step": 8589 }, { "epoch": 0.49, "learning_rate": 0.00010897177283128426, "loss": 0.7564, "step": 8590 }, { "epoch": 0.49, "learning_rate": 0.0001089534748826526, "loss": 0.7662, "step": 8591 }, { "epoch": 0.49, "learning_rate": 0.00010893517663181767, "loss": 0.6737, "step": 8592 }, { "epoch": 0.49, "learning_rate": 0.00010891687807939707, "loss": 0.7105, "step": 8593 }, { "epoch": 0.49, "learning_rate": 0.00010889857922600848, "loss": 0.8106, "step": 8594 }, { "epoch": 0.49, "learning_rate": 0.00010888028007226947, "loss": 0.7581, "step": 8595 }, { "epoch": 0.49, "learning_rate": 0.00010886198061879774, "loss": 0.6703, "step": 8596 }, { "epoch": 0.49, "learning_rate": 0.0001088436808662109, "loss": 0.6944, "step": 8597 }, { "epoch": 0.49, "learning_rate": 0.00010882538081512668, "loss": 0.6896, "step": 8598 }, { "epoch": 0.49, "learning_rate": 0.00010880708046616268, "loss": 0.7457, "step": 8599 }, { "epoch": 0.49, "learning_rate": 0.00010878877981993663, "loss": 0.6327, "step": 8600 }, { "epoch": 0.49, "learning_rate": 0.00010877047887706622, "loss": 0.7383, "step": 8601 }, { "epoch": 0.49, "learning_rate": 0.00010875217763816911, "loss": 0.8295, "step": 8602 }, { "epoch": 0.49, "learning_rate": 0.0001087338761038631, "loss": 0.7951, "step": 8603 }, { "epoch": 0.49, "learning_rate": 0.00010871557427476583, "loss": 0.7777, "step": 8604 }, { "epoch": 0.49, "learning_rate": 0.00010869727215149507, "loss": 0.7772, "step": 8605 }, { "epoch": 0.49, "learning_rate": 0.00010867896973466857, "loss": 0.6634, "step": 8606 }, { "epoch": 0.49, "learning_rate": 0.00010866066702490412, "loss": 0.7278, "step": 8607 }, { "epoch": 0.49, "learning_rate": 0.0001086423640228194, "loss": 0.674, "step": 8608 }, { "epoch": 0.49, "learning_rate": 0.00010862406072903224, "loss": 0.7506, "step": 8609 }, { "epoch": 0.49, "learning_rate": 0.00010860575714416042, "loss": 0.6047, "step": 8610 }, { "epoch": 0.49, "learning_rate": 0.00010858745326882171, "loss": 0.6847, "step": 8611 }, { "epoch": 0.49, "learning_rate": 0.00010856914910363394, "loss": 0.6745, "step": 8612 }, { "epoch": 0.49, "learning_rate": 0.00010855084464921492, "loss": 0.7291, "step": 8613 }, { "epoch": 0.49, "learning_rate": 0.00010853253990618246, "loss": 0.6752, "step": 8614 }, { "epoch": 0.49, "learning_rate": 0.00010851423487515443, "loss": 0.7695, "step": 8615 }, { "epoch": 0.49, "learning_rate": 0.00010849592955674859, "loss": 0.7799, "step": 8616 }, { "epoch": 0.49, "learning_rate": 0.00010847762395158288, "loss": 0.6479, "step": 8617 }, { "epoch": 0.49, "learning_rate": 0.00010845931806027514, "loss": 0.7769, "step": 8618 }, { "epoch": 0.49, "learning_rate": 0.0001084410118834432, "loss": 0.7232, "step": 8619 }, { "epoch": 0.49, "learning_rate": 0.00010842270542170499, "loss": 0.6974, "step": 8620 }, { "epoch": 0.49, "learning_rate": 0.00010840439867567835, "loss": 0.6771, "step": 8621 }, { "epoch": 0.49, "learning_rate": 0.00010838609164598125, "loss": 0.7999, "step": 8622 }, { "epoch": 0.49, "learning_rate": 0.00010836778433323158, "loss": 0.7688, "step": 8623 }, { "epoch": 0.49, "learning_rate": 0.00010834947673804721, "loss": 0.7058, "step": 8624 }, { "epoch": 0.49, "learning_rate": 0.00010833116886104608, "loss": 0.7327, "step": 8625 }, { "epoch": 0.49, "learning_rate": 0.00010831286070284618, "loss": 0.6884, "step": 8626 }, { "epoch": 0.49, "learning_rate": 0.00010829455226406547, "loss": 0.671, "step": 8627 }, { "epoch": 0.49, "learning_rate": 0.0001082762435453218, "loss": 0.7096, "step": 8628 }, { "epoch": 0.49, "learning_rate": 0.00010825793454723325, "loss": 0.6904, "step": 8629 }, { "epoch": 0.49, "learning_rate": 0.00010823962527041774, "loss": 0.6698, "step": 8630 }, { "epoch": 0.49, "learning_rate": 0.00010822131571549325, "loss": 0.703, "step": 8631 }, { "epoch": 0.49, "learning_rate": 0.00010820300588307779, "loss": 0.8382, "step": 8632 }, { "epoch": 0.49, "learning_rate": 0.00010818469577378942, "loss": 0.7432, "step": 8633 }, { "epoch": 0.49, "learning_rate": 0.00010816638538824606, "loss": 0.7411, "step": 8634 }, { "epoch": 0.49, "learning_rate": 0.00010814807472706576, "loss": 0.807, "step": 8635 }, { "epoch": 0.49, "learning_rate": 0.00010812976379086661, "loss": 0.7577, "step": 8636 }, { "epoch": 0.49, "learning_rate": 0.00010811145258026661, "loss": 0.7348, "step": 8637 }, { "epoch": 0.49, "learning_rate": 0.00010809314109588378, "loss": 0.7105, "step": 8638 }, { "epoch": 0.49, "learning_rate": 0.00010807482933833624, "loss": 0.7086, "step": 8639 }, { "epoch": 0.49, "learning_rate": 0.000108056517308242, "loss": 0.7582, "step": 8640 }, { "epoch": 0.49, "learning_rate": 0.00010803820500621922, "loss": 0.7265, "step": 8641 }, { "epoch": 0.49, "learning_rate": 0.00010801989243288589, "loss": 0.7374, "step": 8642 }, { "epoch": 0.49, "learning_rate": 0.0001080015795888602, "loss": 0.7727, "step": 8643 }, { "epoch": 0.49, "learning_rate": 0.0001079832664747602, "loss": 0.7792, "step": 8644 }, { "epoch": 0.49, "learning_rate": 0.00010796495309120401, "loss": 0.6592, "step": 8645 }, { "epoch": 0.49, "learning_rate": 0.00010794663943880976, "loss": 0.7631, "step": 8646 }, { "epoch": 0.49, "learning_rate": 0.00010792832551819558, "loss": 0.711, "step": 8647 }, { "epoch": 0.49, "learning_rate": 0.00010791001132997967, "loss": 0.6781, "step": 8648 }, { "epoch": 0.49, "learning_rate": 0.00010789169687478012, "loss": 0.7549, "step": 8649 }, { "epoch": 0.49, "learning_rate": 0.00010787338215321505, "loss": 0.7363, "step": 8650 }, { "epoch": 0.49, "learning_rate": 0.0001078550671659027, "loss": 0.6004, "step": 8651 }, { "epoch": 0.49, "learning_rate": 0.00010783675191346128, "loss": 0.767, "step": 8652 }, { "epoch": 0.49, "learning_rate": 0.0001078184363965089, "loss": 0.7013, "step": 8653 }, { "epoch": 0.49, "learning_rate": 0.00010780012061566378, "loss": 0.7477, "step": 8654 }, { "epoch": 0.49, "learning_rate": 0.00010778180457154414, "loss": 0.7246, "step": 8655 }, { "epoch": 0.49, "learning_rate": 0.00010776348826476821, "loss": 0.7792, "step": 8656 }, { "epoch": 0.49, "learning_rate": 0.00010774517169595415, "loss": 0.753, "step": 8657 }, { "epoch": 0.49, "learning_rate": 0.00010772685486572025, "loss": 0.7641, "step": 8658 }, { "epoch": 0.49, "learning_rate": 0.00010770853777468475, "loss": 0.7293, "step": 8659 }, { "epoch": 0.49, "learning_rate": 0.00010769022042346586, "loss": 0.7734, "step": 8660 }, { "epoch": 0.49, "learning_rate": 0.00010767190281268187, "loss": 0.6603, "step": 8661 }, { "epoch": 0.49, "learning_rate": 0.00010765358494295103, "loss": 0.698, "step": 8662 }, { "epoch": 0.49, "learning_rate": 0.00010763526681489167, "loss": 0.6632, "step": 8663 }, { "epoch": 0.49, "learning_rate": 0.00010761694842912198, "loss": 0.7073, "step": 8664 }, { "epoch": 0.49, "learning_rate": 0.00010759862978626031, "loss": 0.6912, "step": 8665 }, { "epoch": 0.49, "learning_rate": 0.00010758031088692499, "loss": 0.6283, "step": 8666 }, { "epoch": 0.49, "learning_rate": 0.00010756199173173427, "loss": 0.6617, "step": 8667 }, { "epoch": 0.49, "learning_rate": 0.00010754367232130651, "loss": 0.75, "step": 8668 }, { "epoch": 0.49, "learning_rate": 0.00010752535265626, "loss": 0.7705, "step": 8669 }, { "epoch": 0.49, "learning_rate": 0.00010750703273721314, "loss": 0.7437, "step": 8670 }, { "epoch": 0.49, "learning_rate": 0.00010748871256478421, "loss": 0.7528, "step": 8671 }, { "epoch": 0.49, "learning_rate": 0.0001074703921395916, "loss": 0.7095, "step": 8672 }, { "epoch": 0.49, "learning_rate": 0.00010745207146225365, "loss": 0.6721, "step": 8673 }, { "epoch": 0.49, "learning_rate": 0.00010743375053338877, "loss": 0.7621, "step": 8674 }, { "epoch": 0.49, "learning_rate": 0.00010741542935361531, "loss": 0.695, "step": 8675 }, { "epoch": 0.49, "learning_rate": 0.00010739710792355165, "loss": 0.8194, "step": 8676 }, { "epoch": 0.49, "learning_rate": 0.00010737878624381616, "loss": 0.74, "step": 8677 }, { "epoch": 0.49, "learning_rate": 0.00010736046431502737, "loss": 0.6788, "step": 8678 }, { "epoch": 0.49, "learning_rate": 0.00010734214213780354, "loss": 0.7342, "step": 8679 }, { "epoch": 0.49, "learning_rate": 0.00010732381971276318, "loss": 0.7389, "step": 8680 }, { "epoch": 0.49, "learning_rate": 0.00010730549704052468, "loss": 0.7012, "step": 8681 }, { "epoch": 0.49, "learning_rate": 0.00010728717412170652, "loss": 0.6897, "step": 8682 }, { "epoch": 0.49, "learning_rate": 0.00010726885095692712, "loss": 0.6374, "step": 8683 }, { "epoch": 0.49, "learning_rate": 0.00010725052754680494, "loss": 0.7767, "step": 8684 }, { "epoch": 0.49, "learning_rate": 0.00010723220389195843, "loss": 0.6099, "step": 8685 }, { "epoch": 0.49, "learning_rate": 0.00010721387999300605, "loss": 0.7402, "step": 8686 }, { "epoch": 0.49, "learning_rate": 0.0001071955558505663, "loss": 0.6859, "step": 8687 }, { "epoch": 0.49, "learning_rate": 0.00010717723146525771, "loss": 0.705, "step": 8688 }, { "epoch": 0.49, "learning_rate": 0.00010715890683769872, "loss": 0.7295, "step": 8689 }, { "epoch": 0.49, "learning_rate": 0.00010714058196850783, "loss": 0.706, "step": 8690 }, { "epoch": 0.49, "learning_rate": 0.0001071222568583036, "loss": 0.7091, "step": 8691 }, { "epoch": 0.49, "learning_rate": 0.0001071039315077045, "loss": 0.6928, "step": 8692 }, { "epoch": 0.49, "learning_rate": 0.00010708560591732908, "loss": 0.7514, "step": 8693 }, { "epoch": 0.49, "learning_rate": 0.00010706728008779589, "loss": 0.8768, "step": 8694 }, { "epoch": 0.49, "learning_rate": 0.00010704895401972346, "loss": 0.7265, "step": 8695 }, { "epoch": 0.49, "learning_rate": 0.00010703062771373036, "loss": 0.7595, "step": 8696 }, { "epoch": 0.49, "learning_rate": 0.0001070123011704351, "loss": 0.775, "step": 8697 }, { "epoch": 0.49, "learning_rate": 0.00010699397439045631, "loss": 0.7655, "step": 8698 }, { "epoch": 0.49, "learning_rate": 0.00010697564737441252, "loss": 0.6608, "step": 8699 }, { "epoch": 0.49, "learning_rate": 0.00010695732012292236, "loss": 0.7923, "step": 8700 }, { "epoch": 0.49, "learning_rate": 0.00010693899263660441, "loss": 0.7335, "step": 8701 }, { "epoch": 0.49, "learning_rate": 0.00010692066491607723, "loss": 0.7294, "step": 8702 }, { "epoch": 0.49, "learning_rate": 0.00010690233696195948, "loss": 0.7998, "step": 8703 }, { "epoch": 0.49, "learning_rate": 0.00010688400877486978, "loss": 0.7602, "step": 8704 }, { "epoch": 0.49, "learning_rate": 0.00010686568035542667, "loss": 0.6577, "step": 8705 }, { "epoch": 0.49, "learning_rate": 0.00010684735170424887, "loss": 0.7423, "step": 8706 }, { "epoch": 0.49, "learning_rate": 0.00010682902282195502, "loss": 0.6871, "step": 8707 }, { "epoch": 0.49, "learning_rate": 0.0001068106937091637, "loss": 0.6386, "step": 8708 }, { "epoch": 0.49, "learning_rate": 0.0001067923643664936, "loss": 0.7489, "step": 8709 }, { "epoch": 0.49, "learning_rate": 0.00010677403479456341, "loss": 0.6658, "step": 8710 }, { "epoch": 0.49, "learning_rate": 0.00010675570499399179, "loss": 0.7854, "step": 8711 }, { "epoch": 0.49, "learning_rate": 0.0001067373749653974, "loss": 0.7826, "step": 8712 }, { "epoch": 0.49, "learning_rate": 0.00010671904470939894, "loss": 0.7081, "step": 8713 }, { "epoch": 0.49, "learning_rate": 0.00010670071422661507, "loss": 0.7456, "step": 8714 }, { "epoch": 0.49, "learning_rate": 0.00010668238351766456, "loss": 0.7508, "step": 8715 }, { "epoch": 0.49, "learning_rate": 0.00010666405258316603, "loss": 0.7116, "step": 8716 }, { "epoch": 0.49, "learning_rate": 0.0001066457214237383, "loss": 0.6753, "step": 8717 }, { "epoch": 0.49, "learning_rate": 0.00010662739004000005, "loss": 0.6571, "step": 8718 }, { "epoch": 0.49, "learning_rate": 0.00010660905843256994, "loss": 0.7387, "step": 8719 }, { "epoch": 0.49, "learning_rate": 0.00010659072660206684, "loss": 0.679, "step": 8720 }, { "epoch": 0.49, "learning_rate": 0.00010657239454910939, "loss": 0.7312, "step": 8721 }, { "epoch": 0.49, "learning_rate": 0.00010655406227431641, "loss": 0.706, "step": 8722 }, { "epoch": 0.49, "learning_rate": 0.00010653572977830666, "loss": 0.736, "step": 8723 }, { "epoch": 0.49, "learning_rate": 0.00010651739706169884, "loss": 0.8376, "step": 8724 }, { "epoch": 0.49, "learning_rate": 0.0001064990641251118, "loss": 0.648, "step": 8725 }, { "epoch": 0.49, "learning_rate": 0.0001064807309691643, "loss": 0.7793, "step": 8726 }, { "epoch": 0.5, "learning_rate": 0.00010646239759447514, "loss": 0.7997, "step": 8727 }, { "epoch": 0.5, "learning_rate": 0.00010644406400166309, "loss": 0.7335, "step": 8728 }, { "epoch": 0.5, "learning_rate": 0.00010642573019134703, "loss": 0.698, "step": 8729 }, { "epoch": 0.5, "learning_rate": 0.00010640739616414568, "loss": 0.6725, "step": 8730 }, { "epoch": 0.5, "learning_rate": 0.00010638906192067792, "loss": 0.7456, "step": 8731 }, { "epoch": 0.5, "learning_rate": 0.00010637072746156253, "loss": 0.6728, "step": 8732 }, { "epoch": 0.5, "learning_rate": 0.00010635239278741846, "loss": 0.7138, "step": 8733 }, { "epoch": 0.5, "learning_rate": 0.00010633405789886442, "loss": 0.7738, "step": 8734 }, { "epoch": 0.5, "learning_rate": 0.00010631572279651929, "loss": 0.6784, "step": 8735 }, { "epoch": 0.5, "learning_rate": 0.000106297387481002, "loss": 0.752, "step": 8736 }, { "epoch": 0.5, "learning_rate": 0.00010627905195293135, "loss": 0.7682, "step": 8737 }, { "epoch": 0.5, "learning_rate": 0.00010626071621292621, "loss": 0.7217, "step": 8738 }, { "epoch": 0.5, "learning_rate": 0.0001062423802616055, "loss": 0.7175, "step": 8739 }, { "epoch": 0.5, "learning_rate": 0.00010622404409958808, "loss": 0.7107, "step": 8740 }, { "epoch": 0.5, "learning_rate": 0.00010620570772749288, "loss": 0.7228, "step": 8741 }, { "epoch": 0.5, "learning_rate": 0.00010618737114593871, "loss": 0.792, "step": 8742 }, { "epoch": 0.5, "learning_rate": 0.00010616903435554457, "loss": 0.6337, "step": 8743 }, { "epoch": 0.5, "learning_rate": 0.00010615069735692936, "loss": 0.674, "step": 8744 }, { "epoch": 0.5, "learning_rate": 0.00010613236015071195, "loss": 0.7524, "step": 8745 }, { "epoch": 0.5, "learning_rate": 0.00010611402273751132, "loss": 0.6925, "step": 8746 }, { "epoch": 0.5, "learning_rate": 0.00010609568511794636, "loss": 0.6782, "step": 8747 }, { "epoch": 0.5, "learning_rate": 0.00010607734729263608, "loss": 0.6639, "step": 8748 }, { "epoch": 0.5, "learning_rate": 0.00010605900926219939, "loss": 0.6902, "step": 8749 }, { "epoch": 0.5, "learning_rate": 0.00010604067102725521, "loss": 0.7187, "step": 8750 }, { "epoch": 0.5, "learning_rate": 0.00010602233258842254, "loss": 0.7169, "step": 8751 }, { "epoch": 0.5, "learning_rate": 0.00010600399394632041, "loss": 0.7494, "step": 8752 }, { "epoch": 0.5, "learning_rate": 0.00010598565510156769, "loss": 0.619, "step": 8753 }, { "epoch": 0.5, "learning_rate": 0.0001059673160547834, "loss": 0.7238, "step": 8754 }, { "epoch": 0.5, "learning_rate": 0.00010594897680658658, "loss": 0.7059, "step": 8755 }, { "epoch": 0.5, "learning_rate": 0.00010593063735759618, "loss": 0.7353, "step": 8756 }, { "epoch": 0.5, "learning_rate": 0.0001059122977084312, "loss": 0.7258, "step": 8757 }, { "epoch": 0.5, "learning_rate": 0.00010589395785971068, "loss": 0.7229, "step": 8758 }, { "epoch": 0.5, "learning_rate": 0.00010587561781205364, "loss": 0.6303, "step": 8759 }, { "epoch": 0.5, "learning_rate": 0.00010585727756607906, "loss": 0.6524, "step": 8760 }, { "epoch": 0.5, "learning_rate": 0.000105838937122406, "loss": 0.8697, "step": 8761 }, { "epoch": 0.5, "learning_rate": 0.00010582059648165353, "loss": 0.7384, "step": 8762 }, { "epoch": 0.5, "learning_rate": 0.00010580225564444064, "loss": 0.6962, "step": 8763 }, { "epoch": 0.5, "learning_rate": 0.00010578391461138641, "loss": 0.6204, "step": 8764 }, { "epoch": 0.5, "learning_rate": 0.0001057655733831099, "loss": 0.7341, "step": 8765 }, { "epoch": 0.5, "learning_rate": 0.00010574723196023017, "loss": 0.7021, "step": 8766 }, { "epoch": 0.5, "learning_rate": 0.00010572889034336625, "loss": 0.7002, "step": 8767 }, { "epoch": 0.5, "learning_rate": 0.0001057105485331373, "loss": 0.7277, "step": 8768 }, { "epoch": 0.5, "learning_rate": 0.00010569220653016235, "loss": 0.8322, "step": 8769 }, { "epoch": 0.5, "learning_rate": 0.0001056738643350605, "loss": 0.7326, "step": 8770 }, { "epoch": 0.5, "learning_rate": 0.00010565552194845083, "loss": 0.6857, "step": 8771 }, { "epoch": 0.5, "learning_rate": 0.0001056371793709525, "loss": 0.6427, "step": 8772 }, { "epoch": 0.5, "learning_rate": 0.00010561883660318455, "loss": 0.8089, "step": 8773 }, { "epoch": 0.5, "learning_rate": 0.00010560049364576614, "loss": 0.6564, "step": 8774 }, { "epoch": 0.5, "learning_rate": 0.00010558215049931638, "loss": 0.6792, "step": 8775 }, { "epoch": 0.5, "learning_rate": 0.00010556380716445442, "loss": 0.7767, "step": 8776 }, { "epoch": 0.5, "learning_rate": 0.00010554546364179933, "loss": 0.735, "step": 8777 }, { "epoch": 0.5, "learning_rate": 0.00010552711993197037, "loss": 0.6505, "step": 8778 }, { "epoch": 0.5, "learning_rate": 0.00010550877603558655, "loss": 0.6785, "step": 8779 }, { "epoch": 0.5, "learning_rate": 0.00010549043195326711, "loss": 0.7172, "step": 8780 }, { "epoch": 0.5, "learning_rate": 0.00010547208768563121, "loss": 0.7185, "step": 8781 }, { "epoch": 0.5, "learning_rate": 0.000105453743233298, "loss": 0.6405, "step": 8782 }, { "epoch": 0.5, "learning_rate": 0.00010543539859688663, "loss": 0.6972, "step": 8783 }, { "epoch": 0.5, "learning_rate": 0.00010541705377701631, "loss": 0.8215, "step": 8784 }, { "epoch": 0.5, "learning_rate": 0.00010539870877430624, "loss": 0.7449, "step": 8785 }, { "epoch": 0.5, "learning_rate": 0.00010538036358937555, "loss": 0.6987, "step": 8786 }, { "epoch": 0.5, "learning_rate": 0.00010536201822284348, "loss": 0.6735, "step": 8787 }, { "epoch": 0.5, "learning_rate": 0.00010534367267532927, "loss": 0.7655, "step": 8788 }, { "epoch": 0.5, "learning_rate": 0.0001053253269474521, "loss": 0.7681, "step": 8789 }, { "epoch": 0.5, "learning_rate": 0.00010530698103983114, "loss": 0.6233, "step": 8790 }, { "epoch": 0.5, "learning_rate": 0.00010528863495308567, "loss": 0.6981, "step": 8791 }, { "epoch": 0.5, "learning_rate": 0.00010527028868783491, "loss": 0.8402, "step": 8792 }, { "epoch": 0.5, "learning_rate": 0.00010525194224469805, "loss": 0.8566, "step": 8793 }, { "epoch": 0.5, "learning_rate": 0.0001052335956242944, "loss": 0.6972, "step": 8794 }, { "epoch": 0.5, "learning_rate": 0.00010521524882724314, "loss": 0.6322, "step": 8795 }, { "epoch": 0.5, "learning_rate": 0.0001051969018541636, "loss": 0.7418, "step": 8796 }, { "epoch": 0.5, "learning_rate": 0.00010517855470567496, "loss": 0.7218, "step": 8797 }, { "epoch": 0.5, "learning_rate": 0.00010516020738239651, "loss": 0.7701, "step": 8798 }, { "epoch": 0.5, "learning_rate": 0.00010514185988494754, "loss": 0.7314, "step": 8799 }, { "epoch": 0.5, "learning_rate": 0.00010512351221394731, "loss": 0.6755, "step": 8800 }, { "epoch": 0.5, "learning_rate": 0.00010510516437001513, "loss": 0.8068, "step": 8801 }, { "epoch": 0.5, "learning_rate": 0.00010508681635377024, "loss": 0.6955, "step": 8802 }, { "epoch": 0.5, "learning_rate": 0.00010506846816583195, "loss": 0.8643, "step": 8803 }, { "epoch": 0.5, "learning_rate": 0.00010505011980681962, "loss": 0.8014, "step": 8804 }, { "epoch": 0.5, "learning_rate": 0.00010503177127735244, "loss": 0.7633, "step": 8805 }, { "epoch": 0.5, "learning_rate": 0.0001050134225780498, "loss": 0.7972, "step": 8806 }, { "epoch": 0.5, "learning_rate": 0.00010499507370953104, "loss": 0.6349, "step": 8807 }, { "epoch": 0.5, "learning_rate": 0.0001049767246724154, "loss": 0.6709, "step": 8808 }, { "epoch": 0.5, "learning_rate": 0.00010495837546732224, "loss": 0.7386, "step": 8809 }, { "epoch": 0.5, "learning_rate": 0.00010494002609487093, "loss": 0.7208, "step": 8810 }, { "epoch": 0.5, "learning_rate": 0.00010492167655568078, "loss": 0.7148, "step": 8811 }, { "epoch": 0.5, "learning_rate": 0.00010490332685037112, "loss": 0.6511, "step": 8812 }, { "epoch": 0.5, "learning_rate": 0.00010488497697956135, "loss": 0.699, "step": 8813 }, { "epoch": 0.5, "learning_rate": 0.00010486662694387077, "loss": 0.757, "step": 8814 }, { "epoch": 0.5, "learning_rate": 0.00010484827674391877, "loss": 0.6378, "step": 8815 }, { "epoch": 0.5, "learning_rate": 0.00010482992638032471, "loss": 0.7374, "step": 8816 }, { "epoch": 0.5, "learning_rate": 0.000104811575853708, "loss": 0.6507, "step": 8817 }, { "epoch": 0.5, "learning_rate": 0.00010479322516468797, "loss": 0.7122, "step": 8818 }, { "epoch": 0.5, "learning_rate": 0.000104774874313884, "loss": 0.6909, "step": 8819 }, { "epoch": 0.5, "learning_rate": 0.00010475652330191552, "loss": 0.8002, "step": 8820 }, { "epoch": 0.5, "learning_rate": 0.00010473817212940189, "loss": 0.7333, "step": 8821 }, { "epoch": 0.5, "learning_rate": 0.00010471982079696254, "loss": 0.7546, "step": 8822 }, { "epoch": 0.5, "learning_rate": 0.00010470146930521687, "loss": 0.7698, "step": 8823 }, { "epoch": 0.5, "learning_rate": 0.00010468311765478424, "loss": 0.6192, "step": 8824 }, { "epoch": 0.5, "learning_rate": 0.00010466476584628413, "loss": 0.7463, "step": 8825 }, { "epoch": 0.5, "learning_rate": 0.00010464641388033594, "loss": 0.66, "step": 8826 }, { "epoch": 0.5, "learning_rate": 0.0001046280617575591, "loss": 0.7184, "step": 8827 }, { "epoch": 0.5, "learning_rate": 0.00010460970947857302, "loss": 0.7306, "step": 8828 }, { "epoch": 0.5, "learning_rate": 0.00010459135704399718, "loss": 0.738, "step": 8829 }, { "epoch": 0.5, "learning_rate": 0.00010457300445445101, "loss": 0.6332, "step": 8830 }, { "epoch": 0.5, "learning_rate": 0.00010455465171055392, "loss": 0.7123, "step": 8831 }, { "epoch": 0.5, "learning_rate": 0.00010453629881292538, "loss": 0.7818, "step": 8832 }, { "epoch": 0.5, "learning_rate": 0.0001045179457621849, "loss": 0.7174, "step": 8833 }, { "epoch": 0.5, "learning_rate": 0.00010449959255895185, "loss": 0.6542, "step": 8834 }, { "epoch": 0.5, "learning_rate": 0.00010448123920384578, "loss": 0.6741, "step": 8835 }, { "epoch": 0.5, "learning_rate": 0.00010446288569748613, "loss": 0.7938, "step": 8836 }, { "epoch": 0.5, "learning_rate": 0.00010444453204049238, "loss": 0.6121, "step": 8837 }, { "epoch": 0.5, "learning_rate": 0.00010442617823348402, "loss": 0.6497, "step": 8838 }, { "epoch": 0.5, "learning_rate": 0.00010440782427708056, "loss": 0.7594, "step": 8839 }, { "epoch": 0.5, "learning_rate": 0.00010438947017190142, "loss": 0.6677, "step": 8840 }, { "epoch": 0.5, "learning_rate": 0.00010437111591856619, "loss": 0.633, "step": 8841 }, { "epoch": 0.5, "learning_rate": 0.00010435276151769431, "loss": 0.6309, "step": 8842 }, { "epoch": 0.5, "learning_rate": 0.00010433440696990533, "loss": 0.7313, "step": 8843 }, { "epoch": 0.5, "learning_rate": 0.00010431605227581876, "loss": 0.7599, "step": 8844 }, { "epoch": 0.5, "learning_rate": 0.00010429769743605407, "loss": 0.7301, "step": 8845 }, { "epoch": 0.5, "learning_rate": 0.00010427934245123086, "loss": 0.7375, "step": 8846 }, { "epoch": 0.5, "learning_rate": 0.00010426098732196856, "loss": 0.7566, "step": 8847 }, { "epoch": 0.5, "learning_rate": 0.00010424263204888683, "loss": 0.6293, "step": 8848 }, { "epoch": 0.5, "learning_rate": 0.00010422427663260513, "loss": 0.6899, "step": 8849 }, { "epoch": 0.5, "learning_rate": 0.00010420592107374296, "loss": 0.7605, "step": 8850 }, { "epoch": 0.5, "learning_rate": 0.00010418756537291996, "loss": 0.6513, "step": 8851 }, { "epoch": 0.5, "learning_rate": 0.00010416920953075569, "loss": 0.8046, "step": 8852 }, { "epoch": 0.5, "learning_rate": 0.0001041508535478696, "loss": 0.7109, "step": 8853 }, { "epoch": 0.5, "learning_rate": 0.00010413249742488131, "loss": 0.7166, "step": 8854 }, { "epoch": 0.5, "learning_rate": 0.00010411414116241046, "loss": 0.806, "step": 8855 }, { "epoch": 0.5, "learning_rate": 0.00010409578476107651, "loss": 0.6449, "step": 8856 }, { "epoch": 0.5, "learning_rate": 0.00010407742822149909, "loss": 0.7667, "step": 8857 }, { "epoch": 0.5, "learning_rate": 0.00010405907154429778, "loss": 0.6933, "step": 8858 }, { "epoch": 0.5, "learning_rate": 0.00010404071473009215, "loss": 0.7109, "step": 8859 }, { "epoch": 0.5, "learning_rate": 0.0001040223577795018, "loss": 0.7331, "step": 8860 }, { "epoch": 0.5, "learning_rate": 0.00010400400069314633, "loss": 0.7284, "step": 8861 }, { "epoch": 0.5, "learning_rate": 0.00010398564347164534, "loss": 0.7176, "step": 8862 }, { "epoch": 0.5, "learning_rate": 0.00010396728611561844, "loss": 0.7143, "step": 8863 }, { "epoch": 0.5, "learning_rate": 0.00010394892862568519, "loss": 0.7911, "step": 8864 }, { "epoch": 0.5, "learning_rate": 0.00010393057100246528, "loss": 0.728, "step": 8865 }, { "epoch": 0.5, "learning_rate": 0.00010391221324657829, "loss": 0.7027, "step": 8866 }, { "epoch": 0.5, "learning_rate": 0.00010389385535864382, "loss": 0.7565, "step": 8867 }, { "epoch": 0.5, "learning_rate": 0.00010387549733928155, "loss": 0.7482, "step": 8868 }, { "epoch": 0.5, "learning_rate": 0.00010385713918911105, "loss": 0.728, "step": 8869 }, { "epoch": 0.5, "learning_rate": 0.00010383878090875201, "loss": 0.7259, "step": 8870 }, { "epoch": 0.5, "learning_rate": 0.00010382042249882402, "loss": 0.6687, "step": 8871 }, { "epoch": 0.5, "learning_rate": 0.00010380206395994679, "loss": 0.5919, "step": 8872 }, { "epoch": 0.5, "learning_rate": 0.00010378370529273988, "loss": 0.7417, "step": 8873 }, { "epoch": 0.5, "learning_rate": 0.00010376534649782304, "loss": 0.6871, "step": 8874 }, { "epoch": 0.5, "learning_rate": 0.00010374698757581586, "loss": 0.73, "step": 8875 }, { "epoch": 0.5, "learning_rate": 0.00010372862852733803, "loss": 0.6037, "step": 8876 }, { "epoch": 0.5, "learning_rate": 0.00010371026935300918, "loss": 0.7781, "step": 8877 }, { "epoch": 0.5, "learning_rate": 0.00010369191005344907, "loss": 0.6805, "step": 8878 }, { "epoch": 0.5, "learning_rate": 0.00010367355062927726, "loss": 0.693, "step": 8879 }, { "epoch": 0.5, "learning_rate": 0.00010365519108111348, "loss": 0.6965, "step": 8880 }, { "epoch": 0.5, "learning_rate": 0.00010363683140957745, "loss": 0.6821, "step": 8881 }, { "epoch": 0.5, "learning_rate": 0.00010361847161528881, "loss": 0.7119, "step": 8882 }, { "epoch": 0.5, "learning_rate": 0.00010360011169886723, "loss": 0.787, "step": 8883 }, { "epoch": 0.5, "learning_rate": 0.0001035817516609325, "loss": 0.7367, "step": 8884 }, { "epoch": 0.5, "learning_rate": 0.00010356339150210421, "loss": 0.6913, "step": 8885 }, { "epoch": 0.5, "learning_rate": 0.00010354503122300211, "loss": 0.9406, "step": 8886 }, { "epoch": 0.5, "learning_rate": 0.00010352667082424592, "loss": 0.6831, "step": 8887 }, { "epoch": 0.5, "learning_rate": 0.00010350831030645536, "loss": 0.6913, "step": 8888 }, { "epoch": 0.5, "learning_rate": 0.00010348994967025012, "loss": 0.696, "step": 8889 }, { "epoch": 0.5, "learning_rate": 0.00010347158891624991, "loss": 0.7121, "step": 8890 }, { "epoch": 0.5, "learning_rate": 0.00010345322804507447, "loss": 0.7241, "step": 8891 }, { "epoch": 0.5, "learning_rate": 0.00010343486705734355, "loss": 0.7202, "step": 8892 }, { "epoch": 0.5, "learning_rate": 0.00010341650595367682, "loss": 0.741, "step": 8893 }, { "epoch": 0.5, "learning_rate": 0.0001033981447346941, "loss": 0.6871, "step": 8894 }, { "epoch": 0.5, "learning_rate": 0.00010337978340101504, "loss": 0.6647, "step": 8895 }, { "epoch": 0.5, "learning_rate": 0.00010336142195325944, "loss": 0.7597, "step": 8896 }, { "epoch": 0.5, "learning_rate": 0.00010334306039204707, "loss": 0.7023, "step": 8897 }, { "epoch": 0.5, "learning_rate": 0.00010332469871799759, "loss": 0.7894, "step": 8898 }, { "epoch": 0.5, "learning_rate": 0.00010330633693173082, "loss": 0.7117, "step": 8899 }, { "epoch": 0.5, "learning_rate": 0.00010328797503386654, "loss": 0.7481, "step": 8900 }, { "epoch": 0.5, "learning_rate": 0.00010326961302502445, "loss": 0.6875, "step": 8901 }, { "epoch": 0.5, "learning_rate": 0.00010325125090582435, "loss": 0.5725, "step": 8902 }, { "epoch": 0.51, "learning_rate": 0.000103232888676886, "loss": 0.6289, "step": 8903 }, { "epoch": 0.51, "learning_rate": 0.00010321452633882922, "loss": 0.7581, "step": 8904 }, { "epoch": 0.51, "learning_rate": 0.00010319616389227369, "loss": 0.6764, "step": 8905 }, { "epoch": 0.51, "learning_rate": 0.00010317780133783927, "loss": 0.7235, "step": 8906 }, { "epoch": 0.51, "learning_rate": 0.00010315943867614575, "loss": 0.7164, "step": 8907 }, { "epoch": 0.51, "learning_rate": 0.00010314107590781284, "loss": 0.8128, "step": 8908 }, { "epoch": 0.51, "learning_rate": 0.0001031227130334604, "loss": 0.6584, "step": 8909 }, { "epoch": 0.51, "learning_rate": 0.0001031043500537082, "loss": 0.7542, "step": 8910 }, { "epoch": 0.51, "learning_rate": 0.00010308598696917606, "loss": 0.7053, "step": 8911 }, { "epoch": 0.51, "learning_rate": 0.00010306762378048376, "loss": 0.6754, "step": 8912 }, { "epoch": 0.51, "learning_rate": 0.0001030492604882511, "loss": 0.7161, "step": 8913 }, { "epoch": 0.51, "learning_rate": 0.0001030308970930979, "loss": 0.7516, "step": 8914 }, { "epoch": 0.51, "learning_rate": 0.000103012533595644, "loss": 0.6662, "step": 8915 }, { "epoch": 0.51, "learning_rate": 0.00010299416999650918, "loss": 0.7183, "step": 8916 }, { "epoch": 0.51, "learning_rate": 0.00010297580629631325, "loss": 0.727, "step": 8917 }, { "epoch": 0.51, "learning_rate": 0.0001029574424956761, "loss": 0.6826, "step": 8918 }, { "epoch": 0.51, "learning_rate": 0.00010293907859521745, "loss": 0.6046, "step": 8919 }, { "epoch": 0.51, "learning_rate": 0.00010292071459555723, "loss": 0.6419, "step": 8920 }, { "epoch": 0.51, "learning_rate": 0.0001029023504973152, "loss": 0.7194, "step": 8921 }, { "epoch": 0.51, "learning_rate": 0.00010288398630111127, "loss": 0.7422, "step": 8922 }, { "epoch": 0.51, "learning_rate": 0.00010286562200756522, "loss": 0.7189, "step": 8923 }, { "epoch": 0.51, "learning_rate": 0.00010284725761729689, "loss": 0.7227, "step": 8924 }, { "epoch": 0.51, "learning_rate": 0.00010282889313092616, "loss": 0.7187, "step": 8925 }, { "epoch": 0.51, "learning_rate": 0.00010281052854907289, "loss": 0.6499, "step": 8926 }, { "epoch": 0.51, "learning_rate": 0.0001027921638723569, "loss": 0.7165, "step": 8927 }, { "epoch": 0.51, "learning_rate": 0.00010277379910139804, "loss": 0.6641, "step": 8928 }, { "epoch": 0.51, "learning_rate": 0.00010275543423681621, "loss": 0.7129, "step": 8929 }, { "epoch": 0.51, "learning_rate": 0.00010273706927923125, "loss": 0.7794, "step": 8930 }, { "epoch": 0.51, "learning_rate": 0.000102718704229263, "loss": 0.7071, "step": 8931 }, { "epoch": 0.51, "learning_rate": 0.00010270033908753134, "loss": 0.7248, "step": 8932 }, { "epoch": 0.51, "learning_rate": 0.0001026819738546562, "loss": 0.6701, "step": 8933 }, { "epoch": 0.51, "learning_rate": 0.00010266360853125738, "loss": 0.795, "step": 8934 }, { "epoch": 0.51, "learning_rate": 0.00010264524311795478, "loss": 0.6828, "step": 8935 }, { "epoch": 0.51, "learning_rate": 0.0001026268776153683, "loss": 0.6041, "step": 8936 }, { "epoch": 0.51, "learning_rate": 0.00010260851202411782, "loss": 0.6993, "step": 8937 }, { "epoch": 0.51, "learning_rate": 0.0001025901463448232, "loss": 0.7605, "step": 8938 }, { "epoch": 0.51, "learning_rate": 0.00010257178057810437, "loss": 0.7158, "step": 8939 }, { "epoch": 0.51, "learning_rate": 0.00010255341472458118, "loss": 0.714, "step": 8940 }, { "epoch": 0.51, "learning_rate": 0.00010253504878487357, "loss": 0.7926, "step": 8941 }, { "epoch": 0.51, "learning_rate": 0.0001025166827596014, "loss": 0.7421, "step": 8942 }, { "epoch": 0.51, "learning_rate": 0.00010249831664938456, "loss": 0.7116, "step": 8943 }, { "epoch": 0.51, "learning_rate": 0.00010247995045484302, "loss": 0.7442, "step": 8944 }, { "epoch": 0.51, "learning_rate": 0.00010246158417659664, "loss": 0.7552, "step": 8945 }, { "epoch": 0.51, "learning_rate": 0.00010244321781526533, "loss": 0.6467, "step": 8946 }, { "epoch": 0.51, "learning_rate": 0.000102424851371469, "loss": 0.7472, "step": 8947 }, { "epoch": 0.51, "learning_rate": 0.0001024064848458276, "loss": 0.661, "step": 8948 }, { "epoch": 0.51, "learning_rate": 0.00010238811823896102, "loss": 0.6879, "step": 8949 }, { "epoch": 0.51, "learning_rate": 0.00010236975155148918, "loss": 0.7486, "step": 8950 }, { "epoch": 0.51, "learning_rate": 0.00010235138478403199, "loss": 0.6968, "step": 8951 }, { "epoch": 0.51, "learning_rate": 0.00010233301793720947, "loss": 0.723, "step": 8952 }, { "epoch": 0.51, "learning_rate": 0.00010231465101164139, "loss": 0.6096, "step": 8953 }, { "epoch": 0.51, "learning_rate": 0.0001022962840079478, "loss": 0.7009, "step": 8954 }, { "epoch": 0.51, "learning_rate": 0.00010227791692674862, "loss": 0.7068, "step": 8955 }, { "epoch": 0.51, "learning_rate": 0.00010225954976866376, "loss": 0.6934, "step": 8956 }, { "epoch": 0.51, "learning_rate": 0.00010224118253431316, "loss": 0.7257, "step": 8957 }, { "epoch": 0.51, "learning_rate": 0.00010222281522431678, "loss": 0.7167, "step": 8958 }, { "epoch": 0.51, "learning_rate": 0.00010220444783929456, "loss": 0.7626, "step": 8959 }, { "epoch": 0.51, "learning_rate": 0.00010218608037986643, "loss": 0.6414, "step": 8960 }, { "epoch": 0.51, "learning_rate": 0.00010216771284665234, "loss": 0.8199, "step": 8961 }, { "epoch": 0.51, "learning_rate": 0.00010214934524027228, "loss": 0.7286, "step": 8962 }, { "epoch": 0.51, "learning_rate": 0.0001021309775613462, "loss": 0.6545, "step": 8963 }, { "epoch": 0.51, "learning_rate": 0.000102112609810494, "loss": 0.8036, "step": 8964 }, { "epoch": 0.51, "learning_rate": 0.0001020942419883357, "loss": 0.7656, "step": 8965 }, { "epoch": 0.51, "learning_rate": 0.00010207587409549125, "loss": 0.6779, "step": 8966 }, { "epoch": 0.51, "learning_rate": 0.00010205750613258058, "loss": 0.6898, "step": 8967 }, { "epoch": 0.51, "learning_rate": 0.0001020391381002237, "loss": 0.7193, "step": 8968 }, { "epoch": 0.51, "learning_rate": 0.00010202076999904055, "loss": 0.7053, "step": 8969 }, { "epoch": 0.51, "learning_rate": 0.00010200240182965113, "loss": 0.6597, "step": 8970 }, { "epoch": 0.51, "learning_rate": 0.00010198403359267537, "loss": 0.6641, "step": 8971 }, { "epoch": 0.51, "learning_rate": 0.0001019656652887333, "loss": 0.6342, "step": 8972 }, { "epoch": 0.51, "learning_rate": 0.00010194729691844485, "loss": 0.6244, "step": 8973 }, { "epoch": 0.51, "learning_rate": 0.00010192892848243004, "loss": 0.6186, "step": 8974 }, { "epoch": 0.51, "learning_rate": 0.00010191055998130883, "loss": 0.7403, "step": 8975 }, { "epoch": 0.51, "learning_rate": 0.0001018921914157012, "loss": 0.6947, "step": 8976 }, { "epoch": 0.51, "learning_rate": 0.00010187382278622712, "loss": 0.7193, "step": 8977 }, { "epoch": 0.51, "learning_rate": 0.00010185545409350667, "loss": 0.6679, "step": 8978 }, { "epoch": 0.51, "learning_rate": 0.00010183708533815974, "loss": 0.7384, "step": 8979 }, { "epoch": 0.51, "learning_rate": 0.00010181871652080637, "loss": 0.7057, "step": 8980 }, { "epoch": 0.51, "learning_rate": 0.00010180034764206659, "loss": 0.7476, "step": 8981 }, { "epoch": 0.51, "learning_rate": 0.00010178197870256027, "loss": 0.7318, "step": 8982 }, { "epoch": 0.51, "learning_rate": 0.00010176360970290755, "loss": 0.7823, "step": 8983 }, { "epoch": 0.51, "learning_rate": 0.00010174524064372837, "loss": 0.7405, "step": 8984 }, { "epoch": 0.51, "learning_rate": 0.00010172687152564273, "loss": 0.6445, "step": 8985 }, { "epoch": 0.51, "learning_rate": 0.00010170850234927066, "loss": 0.7558, "step": 8986 }, { "epoch": 0.51, "learning_rate": 0.00010169013311523212, "loss": 0.6864, "step": 8987 }, { "epoch": 0.51, "learning_rate": 0.0001016717638241472, "loss": 0.7921, "step": 8988 }, { "epoch": 0.51, "learning_rate": 0.00010165339447663587, "loss": 0.7696, "step": 8989 }, { "epoch": 0.51, "learning_rate": 0.00010163502507331811, "loss": 0.5843, "step": 8990 }, { "epoch": 0.51, "learning_rate": 0.00010161665561481397, "loss": 0.6604, "step": 8991 }, { "epoch": 0.51, "learning_rate": 0.00010159828610174349, "loss": 0.6742, "step": 8992 }, { "epoch": 0.51, "learning_rate": 0.00010157991653472662, "loss": 0.7412, "step": 8993 }, { "epoch": 0.51, "learning_rate": 0.00010156154691438346, "loss": 0.7948, "step": 8994 }, { "epoch": 0.51, "learning_rate": 0.00010154317724133398, "loss": 0.76, "step": 8995 }, { "epoch": 0.51, "learning_rate": 0.00010152480751619822, "loss": 0.6534, "step": 8996 }, { "epoch": 0.51, "learning_rate": 0.00010150643773959623, "loss": 0.6935, "step": 8997 }, { "epoch": 0.51, "learning_rate": 0.00010148806791214798, "loss": 0.6636, "step": 8998 }, { "epoch": 0.51, "learning_rate": 0.00010146969803447353, "loss": 0.7012, "step": 8999 }, { "epoch": 0.51, "learning_rate": 0.00010145132810719296, "loss": 0.7511, "step": 9000 }, { "epoch": 0.51, "learning_rate": 0.00010143295813092624, "loss": 0.5979, "step": 9001 }, { "epoch": 0.51, "learning_rate": 0.0001014145881062934, "loss": 0.7103, "step": 9002 }, { "epoch": 0.51, "learning_rate": 0.00010139621803391455, "loss": 0.7664, "step": 9003 }, { "epoch": 0.51, "learning_rate": 0.00010137784791440965, "loss": 0.6673, "step": 9004 }, { "epoch": 0.51, "learning_rate": 0.00010135947774839876, "loss": 0.7288, "step": 9005 }, { "epoch": 0.51, "learning_rate": 0.00010134110753650195, "loss": 0.7794, "step": 9006 }, { "epoch": 0.51, "learning_rate": 0.00010132273727933925, "loss": 0.6858, "step": 9007 }, { "epoch": 0.51, "learning_rate": 0.00010130436697753067, "loss": 0.7066, "step": 9008 }, { "epoch": 0.51, "learning_rate": 0.0001012859966316963, "loss": 0.6455, "step": 9009 }, { "epoch": 0.51, "learning_rate": 0.00010126762624245619, "loss": 0.669, "step": 9010 }, { "epoch": 0.51, "learning_rate": 0.00010124925581043034, "loss": 0.7257, "step": 9011 }, { "epoch": 0.51, "learning_rate": 0.00010123088533623883, "loss": 0.7576, "step": 9012 }, { "epoch": 0.51, "learning_rate": 0.00010121251482050173, "loss": 0.75, "step": 9013 }, { "epoch": 0.51, "learning_rate": 0.00010119414426383907, "loss": 0.731, "step": 9014 }, { "epoch": 0.51, "learning_rate": 0.0001011757736668709, "loss": 0.6614, "step": 9015 }, { "epoch": 0.51, "learning_rate": 0.00010115740303021728, "loss": 0.7188, "step": 9016 }, { "epoch": 0.51, "learning_rate": 0.00010113903235449828, "loss": 0.6443, "step": 9017 }, { "epoch": 0.51, "learning_rate": 0.00010112066164033399, "loss": 0.6629, "step": 9018 }, { "epoch": 0.51, "learning_rate": 0.00010110229088834437, "loss": 0.6966, "step": 9019 }, { "epoch": 0.51, "learning_rate": 0.00010108392009914957, "loss": 0.8156, "step": 9020 }, { "epoch": 0.51, "learning_rate": 0.00010106554927336962, "loss": 0.7556, "step": 9021 }, { "epoch": 0.51, "learning_rate": 0.00010104717841162458, "loss": 0.7088, "step": 9022 }, { "epoch": 0.51, "learning_rate": 0.00010102880751453456, "loss": 0.6754, "step": 9023 }, { "epoch": 0.51, "learning_rate": 0.00010101043658271953, "loss": 0.7409, "step": 9024 }, { "epoch": 0.51, "learning_rate": 0.00010099206561679963, "loss": 0.6658, "step": 9025 }, { "epoch": 0.51, "learning_rate": 0.00010097369461739494, "loss": 0.6736, "step": 9026 }, { "epoch": 0.51, "learning_rate": 0.00010095532358512552, "loss": 0.7412, "step": 9027 }, { "epoch": 0.51, "learning_rate": 0.00010093695252061137, "loss": 0.7269, "step": 9028 }, { "epoch": 0.51, "learning_rate": 0.00010091858142447265, "loss": 0.6909, "step": 9029 }, { "epoch": 0.51, "learning_rate": 0.0001009002102973294, "loss": 0.7895, "step": 9030 }, { "epoch": 0.51, "learning_rate": 0.00010088183913980167, "loss": 0.7059, "step": 9031 }, { "epoch": 0.51, "learning_rate": 0.00010086346795250953, "loss": 0.7575, "step": 9032 }, { "epoch": 0.51, "learning_rate": 0.00010084509673607316, "loss": 0.7355, "step": 9033 }, { "epoch": 0.51, "learning_rate": 0.0001008267254911125, "loss": 0.708, "step": 9034 }, { "epoch": 0.51, "learning_rate": 0.0001008083542182477, "loss": 0.6383, "step": 9035 }, { "epoch": 0.51, "learning_rate": 0.00010078998291809883, "loss": 0.7651, "step": 9036 }, { "epoch": 0.51, "learning_rate": 0.00010077161159128596, "loss": 0.8183, "step": 9037 }, { "epoch": 0.51, "learning_rate": 0.00010075324023842918, "loss": 0.7557, "step": 9038 }, { "epoch": 0.51, "learning_rate": 0.00010073486886014858, "loss": 0.8061, "step": 9039 }, { "epoch": 0.51, "learning_rate": 0.00010071649745706419, "loss": 0.6565, "step": 9040 }, { "epoch": 0.51, "learning_rate": 0.00010069812602979615, "loss": 0.7567, "step": 9041 }, { "epoch": 0.51, "learning_rate": 0.00010067975457896454, "loss": 0.7208, "step": 9042 }, { "epoch": 0.51, "learning_rate": 0.00010066138310518942, "loss": 0.7816, "step": 9043 }, { "epoch": 0.51, "learning_rate": 0.0001006430116090909, "loss": 0.6524, "step": 9044 }, { "epoch": 0.51, "learning_rate": 0.00010062464009128902, "loss": 0.6829, "step": 9045 }, { "epoch": 0.51, "learning_rate": 0.00010060626855240394, "loss": 0.7027, "step": 9046 }, { "epoch": 0.51, "learning_rate": 0.00010058789699305567, "loss": 0.7251, "step": 9047 }, { "epoch": 0.51, "learning_rate": 0.00010056952541386435, "loss": 0.7448, "step": 9048 }, { "epoch": 0.51, "learning_rate": 0.00010055115381545006, "loss": 0.7244, "step": 9049 }, { "epoch": 0.51, "learning_rate": 0.00010053278219843285, "loss": 0.6571, "step": 9050 }, { "epoch": 0.51, "learning_rate": 0.00010051441056343287, "loss": 0.6828, "step": 9051 }, { "epoch": 0.51, "learning_rate": 0.0001004960389110702, "loss": 0.7983, "step": 9052 }, { "epoch": 0.51, "learning_rate": 0.00010047766724196491, "loss": 0.7583, "step": 9053 }, { "epoch": 0.51, "learning_rate": 0.00010045929555673705, "loss": 0.7632, "step": 9054 }, { "epoch": 0.51, "learning_rate": 0.00010044092385600681, "loss": 0.7539, "step": 9055 }, { "epoch": 0.51, "learning_rate": 0.00010042255214039425, "loss": 0.6876, "step": 9056 }, { "epoch": 0.51, "learning_rate": 0.0001004041804105194, "loss": 0.8329, "step": 9057 }, { "epoch": 0.51, "learning_rate": 0.00010038580866700243, "loss": 0.7062, "step": 9058 }, { "epoch": 0.51, "learning_rate": 0.0001003674369104634, "loss": 0.5914, "step": 9059 }, { "epoch": 0.51, "learning_rate": 0.00010034906514152238, "loss": 0.6818, "step": 9060 }, { "epoch": 0.51, "learning_rate": 0.00010033069336079952, "loss": 0.5872, "step": 9061 }, { "epoch": 0.51, "learning_rate": 0.0001003123215689149, "loss": 0.6799, "step": 9062 }, { "epoch": 0.51, "learning_rate": 0.00010029394976648861, "loss": 0.7026, "step": 9063 }, { "epoch": 0.51, "learning_rate": 0.00010027557795414071, "loss": 0.6739, "step": 9064 }, { "epoch": 0.51, "learning_rate": 0.00010025720613249136, "loss": 0.6389, "step": 9065 }, { "epoch": 0.51, "learning_rate": 0.00010023883430216064, "loss": 0.7221, "step": 9066 }, { "epoch": 0.51, "learning_rate": 0.00010022046246376861, "loss": 0.7888, "step": 9067 }, { "epoch": 0.51, "learning_rate": 0.00010020209061793542, "loss": 0.7226, "step": 9068 }, { "epoch": 0.51, "learning_rate": 0.00010018371876528113, "loss": 0.8232, "step": 9069 }, { "epoch": 0.51, "learning_rate": 0.00010016534690642583, "loss": 0.7749, "step": 9070 }, { "epoch": 0.51, "learning_rate": 0.00010014697504198964, "loss": 0.6058, "step": 9071 }, { "epoch": 0.51, "learning_rate": 0.0001001286031725927, "loss": 0.8035, "step": 9072 }, { "epoch": 0.51, "learning_rate": 0.00010011023129885502, "loss": 0.6529, "step": 9073 }, { "epoch": 0.51, "learning_rate": 0.00010009185942139676, "loss": 0.8526, "step": 9074 }, { "epoch": 0.51, "learning_rate": 0.00010007348754083801, "loss": 0.7143, "step": 9075 }, { "epoch": 0.51, "learning_rate": 0.00010005511565779885, "loss": 0.7113, "step": 9076 }, { "epoch": 0.51, "learning_rate": 0.0001000367437728994, "loss": 0.6383, "step": 9077 }, { "epoch": 0.51, "learning_rate": 0.00010001837188675978, "loss": 0.7769, "step": 9078 }, { "epoch": 0.52, "learning_rate": 0.0001, "loss": 0.6912, "step": 9079 }, { "epoch": 0.52, "learning_rate": 9.998162811324024e-05, "loss": 0.6807, "step": 9080 }, { "epoch": 0.52, "learning_rate": 9.99632562271006e-05, "loss": 0.6605, "step": 9081 }, { "epoch": 0.52, "learning_rate": 9.994488434220116e-05, "loss": 0.7073, "step": 9082 }, { "epoch": 0.52, "learning_rate": 9.992651245916202e-05, "loss": 0.6516, "step": 9083 }, { "epoch": 0.52, "learning_rate": 9.990814057860326e-05, "loss": 0.7378, "step": 9084 }, { "epoch": 0.52, "learning_rate": 9.988976870114502e-05, "loss": 0.7417, "step": 9085 }, { "epoch": 0.52, "learning_rate": 9.987139682740734e-05, "loss": 0.7366, "step": 9086 }, { "epoch": 0.52, "learning_rate": 9.985302495801037e-05, "loss": 0.804, "step": 9087 }, { "epoch": 0.52, "learning_rate": 9.983465309357418e-05, "loss": 0.6649, "step": 9088 }, { "epoch": 0.52, "learning_rate": 9.981628123471888e-05, "loss": 0.7375, "step": 9089 }, { "epoch": 0.52, "learning_rate": 9.97979093820646e-05, "loss": 0.643, "step": 9090 }, { "epoch": 0.52, "learning_rate": 9.977953753623142e-05, "loss": 0.6873, "step": 9091 }, { "epoch": 0.52, "learning_rate": 9.97611656978394e-05, "loss": 0.6082, "step": 9092 }, { "epoch": 0.52, "learning_rate": 9.974279386750864e-05, "loss": 0.6968, "step": 9093 }, { "epoch": 0.52, "learning_rate": 9.97244220458593e-05, "loss": 0.717, "step": 9094 }, { "epoch": 0.52, "learning_rate": 9.970605023351141e-05, "loss": 0.6178, "step": 9095 }, { "epoch": 0.52, "learning_rate": 9.968767843108511e-05, "loss": 0.7337, "step": 9096 }, { "epoch": 0.52, "learning_rate": 9.966930663920049e-05, "loss": 0.7385, "step": 9097 }, { "epoch": 0.52, "learning_rate": 9.965093485847767e-05, "loss": 0.7174, "step": 9098 }, { "epoch": 0.52, "learning_rate": 9.963256308953665e-05, "loss": 0.6924, "step": 9099 }, { "epoch": 0.52, "learning_rate": 9.96141913329976e-05, "loss": 0.7199, "step": 9100 }, { "epoch": 0.52, "learning_rate": 9.959581958948063e-05, "loss": 0.6134, "step": 9101 }, { "epoch": 0.52, "learning_rate": 9.957744785960579e-05, "loss": 0.7928, "step": 9102 }, { "epoch": 0.52, "learning_rate": 9.95590761439932e-05, "loss": 0.6454, "step": 9103 }, { "epoch": 0.52, "learning_rate": 9.954070444326293e-05, "loss": 0.6171, "step": 9104 }, { "epoch": 0.52, "learning_rate": 9.952233275803514e-05, "loss": 0.7699, "step": 9105 }, { "epoch": 0.52, "learning_rate": 9.950396108892983e-05, "loss": 0.6941, "step": 9106 }, { "epoch": 0.52, "learning_rate": 9.948558943656716e-05, "loss": 0.7062, "step": 9107 }, { "epoch": 0.52, "learning_rate": 9.946721780156716e-05, "loss": 0.729, "step": 9108 }, { "epoch": 0.52, "learning_rate": 9.944884618454996e-05, "loss": 0.7921, "step": 9109 }, { "epoch": 0.52, "learning_rate": 9.943047458613566e-05, "loss": 0.6611, "step": 9110 }, { "epoch": 0.52, "learning_rate": 9.941210300694434e-05, "loss": 0.7264, "step": 9111 }, { "epoch": 0.52, "learning_rate": 9.939373144759607e-05, "loss": 0.6227, "step": 9112 }, { "epoch": 0.52, "learning_rate": 9.937535990871099e-05, "loss": 0.6688, "step": 9113 }, { "epoch": 0.52, "learning_rate": 9.935698839090915e-05, "loss": 0.655, "step": 9114 }, { "epoch": 0.52, "learning_rate": 9.93386168948106e-05, "loss": 0.7338, "step": 9115 }, { "epoch": 0.52, "learning_rate": 9.932024542103547e-05, "loss": 0.7023, "step": 9116 }, { "epoch": 0.52, "learning_rate": 9.930187397020386e-05, "loss": 0.6979, "step": 9117 }, { "epoch": 0.52, "learning_rate": 9.92835025429358e-05, "loss": 0.765, "step": 9118 }, { "epoch": 0.52, "learning_rate": 9.926513113985145e-05, "loss": 0.7285, "step": 9119 }, { "epoch": 0.52, "learning_rate": 9.924675976157087e-05, "loss": 0.7405, "step": 9120 }, { "epoch": 0.52, "learning_rate": 9.922838840871407e-05, "loss": 0.7385, "step": 9121 }, { "epoch": 0.52, "learning_rate": 9.921001708190118e-05, "loss": 0.7007, "step": 9122 }, { "epoch": 0.52, "learning_rate": 9.919164578175233e-05, "loss": 0.7813, "step": 9123 }, { "epoch": 0.52, "learning_rate": 9.917327450888751e-05, "loss": 0.7677, "step": 9124 }, { "epoch": 0.52, "learning_rate": 9.915490326392686e-05, "loss": 0.7193, "step": 9125 }, { "epoch": 0.52, "learning_rate": 9.913653204749046e-05, "loss": 0.7083, "step": 9126 }, { "epoch": 0.52, "learning_rate": 9.911816086019834e-05, "loss": 0.7316, "step": 9127 }, { "epoch": 0.52, "learning_rate": 9.909978970267065e-05, "loss": 0.6575, "step": 9128 }, { "epoch": 0.52, "learning_rate": 9.908141857552737e-05, "loss": 0.757, "step": 9129 }, { "epoch": 0.52, "learning_rate": 9.906304747938865e-05, "loss": 0.7115, "step": 9130 }, { "epoch": 0.52, "learning_rate": 9.904467641487452e-05, "loss": 0.7612, "step": 9131 }, { "epoch": 0.52, "learning_rate": 9.902630538260508e-05, "loss": 0.8094, "step": 9132 }, { "epoch": 0.52, "learning_rate": 9.900793438320037e-05, "loss": 0.7078, "step": 9133 }, { "epoch": 0.52, "learning_rate": 9.898956341728047e-05, "loss": 0.7849, "step": 9134 }, { "epoch": 0.52, "learning_rate": 9.897119248546549e-05, "loss": 0.728, "step": 9135 }, { "epoch": 0.52, "learning_rate": 9.895282158837545e-05, "loss": 0.7997, "step": 9136 }, { "epoch": 0.52, "learning_rate": 9.893445072663041e-05, "loss": 0.6787, "step": 9137 }, { "epoch": 0.52, "learning_rate": 9.891607990085045e-05, "loss": 0.6851, "step": 9138 }, { "epoch": 0.52, "learning_rate": 9.889770911165565e-05, "loss": 0.6682, "step": 9139 }, { "epoch": 0.52, "learning_rate": 9.887933835966604e-05, "loss": 0.7722, "step": 9140 }, { "epoch": 0.52, "learning_rate": 9.88609676455017e-05, "loss": 0.6828, "step": 9141 }, { "epoch": 0.52, "learning_rate": 9.884259696978274e-05, "loss": 0.6906, "step": 9142 }, { "epoch": 0.52, "learning_rate": 9.882422633312915e-05, "loss": 0.6549, "step": 9143 }, { "epoch": 0.52, "learning_rate": 9.880585573616097e-05, "loss": 0.6989, "step": 9144 }, { "epoch": 0.52, "learning_rate": 9.878748517949829e-05, "loss": 0.7304, "step": 9145 }, { "epoch": 0.52, "learning_rate": 9.876911466376118e-05, "loss": 0.7661, "step": 9146 }, { "epoch": 0.52, "learning_rate": 9.875074418956968e-05, "loss": 0.6875, "step": 9147 }, { "epoch": 0.52, "learning_rate": 9.873237375754385e-05, "loss": 0.6517, "step": 9148 }, { "epoch": 0.52, "learning_rate": 9.871400336830371e-05, "loss": 0.6999, "step": 9149 }, { "epoch": 0.52, "learning_rate": 9.869563302246937e-05, "loss": 0.7738, "step": 9150 }, { "epoch": 0.52, "learning_rate": 9.86772627206608e-05, "loss": 0.7006, "step": 9151 }, { "epoch": 0.52, "learning_rate": 9.865889246349809e-05, "loss": 0.7036, "step": 9152 }, { "epoch": 0.52, "learning_rate": 9.864052225160125e-05, "loss": 0.737, "step": 9153 }, { "epoch": 0.52, "learning_rate": 9.862215208559037e-05, "loss": 0.793, "step": 9154 }, { "epoch": 0.52, "learning_rate": 9.860378196608549e-05, "loss": 0.6752, "step": 9155 }, { "epoch": 0.52, "learning_rate": 9.85854118937066e-05, "loss": 0.7061, "step": 9156 }, { "epoch": 0.52, "learning_rate": 9.856704186907381e-05, "loss": 0.7415, "step": 9157 }, { "epoch": 0.52, "learning_rate": 9.854867189280708e-05, "loss": 0.7122, "step": 9158 }, { "epoch": 0.52, "learning_rate": 9.85303019655265e-05, "loss": 0.7113, "step": 9159 }, { "epoch": 0.52, "learning_rate": 9.851193208785206e-05, "loss": 0.6348, "step": 9160 }, { "epoch": 0.52, "learning_rate": 9.849356226040381e-05, "loss": 0.694, "step": 9161 }, { "epoch": 0.52, "learning_rate": 9.84751924838018e-05, "loss": 0.6861, "step": 9162 }, { "epoch": 0.52, "learning_rate": 9.845682275866603e-05, "loss": 0.7411, "step": 9163 }, { "epoch": 0.52, "learning_rate": 9.843845308561656e-05, "loss": 0.673, "step": 9164 }, { "epoch": 0.52, "learning_rate": 9.842008346527341e-05, "loss": 0.7857, "step": 9165 }, { "epoch": 0.52, "learning_rate": 9.840171389825656e-05, "loss": 0.7322, "step": 9166 }, { "epoch": 0.52, "learning_rate": 9.838334438518605e-05, "loss": 0.6427, "step": 9167 }, { "epoch": 0.52, "learning_rate": 9.836497492668191e-05, "loss": 0.7947, "step": 9168 }, { "epoch": 0.52, "learning_rate": 9.834660552336415e-05, "loss": 0.6769, "step": 9169 }, { "epoch": 0.52, "learning_rate": 9.83282361758528e-05, "loss": 0.6833, "step": 9170 }, { "epoch": 0.52, "learning_rate": 9.830986688476788e-05, "loss": 0.8043, "step": 9171 }, { "epoch": 0.52, "learning_rate": 9.82914976507294e-05, "loss": 0.8152, "step": 9172 }, { "epoch": 0.52, "learning_rate": 9.82731284743573e-05, "loss": 0.738, "step": 9173 }, { "epoch": 0.52, "learning_rate": 9.825475935627165e-05, "loss": 0.7116, "step": 9174 }, { "epoch": 0.52, "learning_rate": 9.823639029709248e-05, "loss": 0.7046, "step": 9175 }, { "epoch": 0.52, "learning_rate": 9.821802129743974e-05, "loss": 0.6777, "step": 9176 }, { "epoch": 0.52, "learning_rate": 9.819965235793345e-05, "loss": 0.6354, "step": 9177 }, { "epoch": 0.52, "learning_rate": 9.818128347919362e-05, "loss": 0.7684, "step": 9178 }, { "epoch": 0.52, "learning_rate": 9.816291466184026e-05, "loss": 0.7531, "step": 9179 }, { "epoch": 0.52, "learning_rate": 9.814454590649335e-05, "loss": 0.7135, "step": 9180 }, { "epoch": 0.52, "learning_rate": 9.81261772137729e-05, "loss": 0.7304, "step": 9181 }, { "epoch": 0.52, "learning_rate": 9.810780858429883e-05, "loss": 0.717, "step": 9182 }, { "epoch": 0.52, "learning_rate": 9.808944001869119e-05, "loss": 0.7481, "step": 9183 }, { "epoch": 0.52, "learning_rate": 9.807107151756999e-05, "loss": 0.6537, "step": 9184 }, { "epoch": 0.52, "learning_rate": 9.805270308155515e-05, "loss": 0.7404, "step": 9185 }, { "epoch": 0.52, "learning_rate": 9.803433471126671e-05, "loss": 0.7636, "step": 9186 }, { "epoch": 0.52, "learning_rate": 9.801596640732465e-05, "loss": 0.6537, "step": 9187 }, { "epoch": 0.52, "learning_rate": 9.799759817034891e-05, "loss": 0.6685, "step": 9188 }, { "epoch": 0.52, "learning_rate": 9.797923000095947e-05, "loss": 0.755, "step": 9189 }, { "epoch": 0.52, "learning_rate": 9.796086189977631e-05, "loss": 0.756, "step": 9190 }, { "epoch": 0.52, "learning_rate": 9.794249386741943e-05, "loss": 0.8006, "step": 9191 }, { "epoch": 0.52, "learning_rate": 9.792412590450876e-05, "loss": 0.7037, "step": 9192 }, { "epoch": 0.52, "learning_rate": 9.790575801166432e-05, "loss": 0.6442, "step": 9193 }, { "epoch": 0.52, "learning_rate": 9.788739018950601e-05, "loss": 0.7425, "step": 9194 }, { "epoch": 0.52, "learning_rate": 9.786902243865385e-05, "loss": 0.7129, "step": 9195 }, { "epoch": 0.52, "learning_rate": 9.785065475972774e-05, "loss": 0.7613, "step": 9196 }, { "epoch": 0.52, "learning_rate": 9.783228715334768e-05, "loss": 0.8035, "step": 9197 }, { "epoch": 0.52, "learning_rate": 9.78139196201336e-05, "loss": 0.7382, "step": 9198 }, { "epoch": 0.52, "learning_rate": 9.779555216070546e-05, "loss": 0.7029, "step": 9199 }, { "epoch": 0.52, "learning_rate": 9.777718477568323e-05, "loss": 0.6901, "step": 9200 }, { "epoch": 0.52, "learning_rate": 9.775881746568685e-05, "loss": 0.6922, "step": 9201 }, { "epoch": 0.52, "learning_rate": 9.774045023133629e-05, "loss": 0.8093, "step": 9202 }, { "epoch": 0.52, "learning_rate": 9.77220830732514e-05, "loss": 0.599, "step": 9203 }, { "epoch": 0.52, "learning_rate": 9.770371599205222e-05, "loss": 0.6565, "step": 9204 }, { "epoch": 0.52, "learning_rate": 9.768534898835862e-05, "loss": 0.7417, "step": 9205 }, { "epoch": 0.52, "learning_rate": 9.766698206279057e-05, "loss": 0.7356, "step": 9206 }, { "epoch": 0.52, "learning_rate": 9.764861521596801e-05, "loss": 0.7269, "step": 9207 }, { "epoch": 0.52, "learning_rate": 9.763024844851081e-05, "loss": 0.8251, "step": 9208 }, { "epoch": 0.52, "learning_rate": 9.761188176103902e-05, "loss": 0.6904, "step": 9209 }, { "epoch": 0.52, "learning_rate": 9.759351515417243e-05, "loss": 0.7555, "step": 9210 }, { "epoch": 0.52, "learning_rate": 9.757514862853102e-05, "loss": 0.7392, "step": 9211 }, { "epoch": 0.52, "learning_rate": 9.755678218473469e-05, "loss": 0.8028, "step": 9212 }, { "epoch": 0.52, "learning_rate": 9.75384158234034e-05, "loss": 0.7932, "step": 9213 }, { "epoch": 0.52, "learning_rate": 9.7520049545157e-05, "loss": 0.7686, "step": 9214 }, { "epoch": 0.52, "learning_rate": 9.750168335061542e-05, "loss": 0.6925, "step": 9215 }, { "epoch": 0.52, "learning_rate": 9.748331724039863e-05, "loss": 0.7013, "step": 9216 }, { "epoch": 0.52, "learning_rate": 9.746495121512648e-05, "loss": 0.8107, "step": 9217 }, { "epoch": 0.52, "learning_rate": 9.744658527541884e-05, "loss": 0.6958, "step": 9218 }, { "epoch": 0.52, "learning_rate": 9.742821942189565e-05, "loss": 0.712, "step": 9219 }, { "epoch": 0.52, "learning_rate": 9.740985365517682e-05, "loss": 0.6693, "step": 9220 }, { "epoch": 0.52, "learning_rate": 9.739148797588218e-05, "loss": 0.5913, "step": 9221 }, { "epoch": 0.52, "learning_rate": 9.737312238463169e-05, "loss": 0.7312, "step": 9222 }, { "epoch": 0.52, "learning_rate": 9.735475688204521e-05, "loss": 0.7262, "step": 9223 }, { "epoch": 0.52, "learning_rate": 9.733639146874267e-05, "loss": 0.7234, "step": 9224 }, { "epoch": 0.52, "learning_rate": 9.731802614534383e-05, "loss": 0.683, "step": 9225 }, { "epoch": 0.52, "learning_rate": 9.729966091246868e-05, "loss": 0.6628, "step": 9226 }, { "epoch": 0.52, "learning_rate": 9.728129577073702e-05, "loss": 0.7617, "step": 9227 }, { "epoch": 0.52, "learning_rate": 9.726293072076877e-05, "loss": 0.7245, "step": 9228 }, { "epoch": 0.52, "learning_rate": 9.724456576318381e-05, "loss": 0.6117, "step": 9229 }, { "epoch": 0.52, "learning_rate": 9.722620089860196e-05, "loss": 0.6827, "step": 9230 }, { "epoch": 0.52, "learning_rate": 9.720783612764314e-05, "loss": 0.6676, "step": 9231 }, { "epoch": 0.52, "learning_rate": 9.718947145092713e-05, "loss": 0.5676, "step": 9232 }, { "epoch": 0.52, "learning_rate": 9.717110686907386e-05, "loss": 0.7152, "step": 9233 }, { "epoch": 0.52, "learning_rate": 9.715274238270312e-05, "loss": 0.7444, "step": 9234 }, { "epoch": 0.52, "learning_rate": 9.71343779924348e-05, "loss": 0.7497, "step": 9235 }, { "epoch": 0.52, "learning_rate": 9.711601369888877e-05, "loss": 0.7281, "step": 9236 }, { "epoch": 0.52, "learning_rate": 9.70976495026848e-05, "loss": 0.6715, "step": 9237 }, { "epoch": 0.52, "learning_rate": 9.70792854044428e-05, "loss": 0.6034, "step": 9238 }, { "epoch": 0.52, "learning_rate": 9.706092140478257e-05, "loss": 0.6613, "step": 9239 }, { "epoch": 0.52, "learning_rate": 9.704255750432397e-05, "loss": 0.8384, "step": 9240 }, { "epoch": 0.52, "learning_rate": 9.702419370368676e-05, "loss": 0.6777, "step": 9241 }, { "epoch": 0.52, "learning_rate": 9.700583000349086e-05, "loss": 0.6717, "step": 9242 }, { "epoch": 0.52, "learning_rate": 9.698746640435603e-05, "loss": 0.7948, "step": 9243 }, { "epoch": 0.52, "learning_rate": 9.69691029069021e-05, "loss": 0.7428, "step": 9244 }, { "epoch": 0.52, "learning_rate": 9.695073951174891e-05, "loss": 0.6822, "step": 9245 }, { "epoch": 0.52, "learning_rate": 9.69323762195163e-05, "loss": 0.6075, "step": 9246 }, { "epoch": 0.52, "learning_rate": 9.691401303082398e-05, "loss": 0.6761, "step": 9247 }, { "epoch": 0.52, "learning_rate": 9.689564994629181e-05, "loss": 0.7554, "step": 9248 }, { "epoch": 0.52, "learning_rate": 9.687728696653963e-05, "loss": 0.6541, "step": 9249 }, { "epoch": 0.52, "learning_rate": 9.685892409218717e-05, "loss": 0.7857, "step": 9250 }, { "epoch": 0.52, "learning_rate": 9.684056132385427e-05, "loss": 0.7759, "step": 9251 }, { "epoch": 0.52, "learning_rate": 9.682219866216074e-05, "loss": 0.6967, "step": 9252 }, { "epoch": 0.52, "learning_rate": 9.68038361077263e-05, "loss": 0.5776, "step": 9253 }, { "epoch": 0.52, "learning_rate": 9.678547366117083e-05, "loss": 0.6413, "step": 9254 }, { "epoch": 0.52, "learning_rate": 9.6767111323114e-05, "loss": 0.7345, "step": 9255 }, { "epoch": 0.53, "learning_rate": 9.674874909417568e-05, "loss": 0.7065, "step": 9256 }, { "epoch": 0.53, "learning_rate": 9.673038697497556e-05, "loss": 0.7212, "step": 9257 }, { "epoch": 0.53, "learning_rate": 9.67120249661335e-05, "loss": 0.7724, "step": 9258 }, { "epoch": 0.53, "learning_rate": 9.669366306826919e-05, "loss": 0.7332, "step": 9259 }, { "epoch": 0.53, "learning_rate": 9.66753012820024e-05, "loss": 0.7385, "step": 9260 }, { "epoch": 0.53, "learning_rate": 9.665693960795298e-05, "loss": 0.7304, "step": 9261 }, { "epoch": 0.53, "learning_rate": 9.663857804674059e-05, "loss": 0.8463, "step": 9262 }, { "epoch": 0.53, "learning_rate": 9.662021659898498e-05, "loss": 0.6738, "step": 9263 }, { "epoch": 0.53, "learning_rate": 9.660185526530593e-05, "loss": 0.6285, "step": 9264 }, { "epoch": 0.53, "learning_rate": 9.65834940463232e-05, "loss": 0.7517, "step": 9265 }, { "epoch": 0.53, "learning_rate": 9.656513294265647e-05, "loss": 0.8187, "step": 9266 }, { "epoch": 0.53, "learning_rate": 9.654677195492553e-05, "loss": 0.6124, "step": 9267 }, { "epoch": 0.53, "learning_rate": 9.652841108375012e-05, "loss": 0.6324, "step": 9268 }, { "epoch": 0.53, "learning_rate": 9.651005032974994e-05, "loss": 0.6799, "step": 9269 }, { "epoch": 0.53, "learning_rate": 9.649168969354467e-05, "loss": 0.7019, "step": 9270 }, { "epoch": 0.53, "learning_rate": 9.64733291757541e-05, "loss": 0.7406, "step": 9271 }, { "epoch": 0.53, "learning_rate": 9.64549687769979e-05, "loss": 0.695, "step": 9272 }, { "epoch": 0.53, "learning_rate": 9.643660849789581e-05, "loss": 0.6614, "step": 9273 }, { "epoch": 0.53, "learning_rate": 9.641824833906754e-05, "loss": 0.6455, "step": 9274 }, { "epoch": 0.53, "learning_rate": 9.639988830113276e-05, "loss": 0.7376, "step": 9275 }, { "epoch": 0.53, "learning_rate": 9.638152838471124e-05, "loss": 0.6822, "step": 9276 }, { "epoch": 0.53, "learning_rate": 9.636316859042259e-05, "loss": 0.6779, "step": 9277 }, { "epoch": 0.53, "learning_rate": 9.634480891888654e-05, "loss": 0.7096, "step": 9278 }, { "epoch": 0.53, "learning_rate": 9.632644937072277e-05, "loss": 0.6421, "step": 9279 }, { "epoch": 0.53, "learning_rate": 9.630808994655097e-05, "loss": 0.6452, "step": 9280 }, { "epoch": 0.53, "learning_rate": 9.628973064699081e-05, "loss": 0.6919, "step": 9281 }, { "epoch": 0.53, "learning_rate": 9.627137147266197e-05, "loss": 0.6976, "step": 9282 }, { "epoch": 0.53, "learning_rate": 9.625301242418417e-05, "loss": 0.7386, "step": 9283 }, { "epoch": 0.53, "learning_rate": 9.623465350217699e-05, "loss": 0.7208, "step": 9284 }, { "epoch": 0.53, "learning_rate": 9.621629470726013e-05, "loss": 0.8171, "step": 9285 }, { "epoch": 0.53, "learning_rate": 9.619793604005323e-05, "loss": 0.817, "step": 9286 }, { "epoch": 0.53, "learning_rate": 9.6179577501176e-05, "loss": 0.6616, "step": 9287 }, { "epoch": 0.53, "learning_rate": 9.616121909124801e-05, "loss": 0.6493, "step": 9288 }, { "epoch": 0.53, "learning_rate": 9.614286081088895e-05, "loss": 0.8082, "step": 9289 }, { "epoch": 0.53, "learning_rate": 9.612450266071847e-05, "loss": 0.6843, "step": 9290 }, { "epoch": 0.53, "learning_rate": 9.610614464135622e-05, "loss": 0.7163, "step": 9291 }, { "epoch": 0.53, "learning_rate": 9.608778675342175e-05, "loss": 0.7231, "step": 9292 }, { "epoch": 0.53, "learning_rate": 9.606942899753473e-05, "loss": 0.6798, "step": 9293 }, { "epoch": 0.53, "learning_rate": 9.605107137431482e-05, "loss": 0.7625, "step": 9294 }, { "epoch": 0.53, "learning_rate": 9.603271388438159e-05, "loss": 0.7332, "step": 9295 }, { "epoch": 0.53, "learning_rate": 9.601435652835467e-05, "loss": 0.6778, "step": 9296 }, { "epoch": 0.53, "learning_rate": 9.599599930685367e-05, "loss": 0.7257, "step": 9297 }, { "epoch": 0.53, "learning_rate": 9.597764222049824e-05, "loss": 0.6938, "step": 9298 }, { "epoch": 0.53, "learning_rate": 9.595928526990788e-05, "loss": 0.7576, "step": 9299 }, { "epoch": 0.53, "learning_rate": 9.594092845570225e-05, "loss": 0.702, "step": 9300 }, { "epoch": 0.53, "learning_rate": 9.592257177850095e-05, "loss": 0.7468, "step": 9301 }, { "epoch": 0.53, "learning_rate": 9.590421523892351e-05, "loss": 0.6229, "step": 9302 }, { "epoch": 0.53, "learning_rate": 9.588585883758958e-05, "loss": 0.7927, "step": 9303 }, { "epoch": 0.53, "learning_rate": 9.586750257511867e-05, "loss": 0.7781, "step": 9304 }, { "epoch": 0.53, "learning_rate": 9.584914645213045e-05, "loss": 0.662, "step": 9305 }, { "epoch": 0.53, "learning_rate": 9.583079046924436e-05, "loss": 0.7769, "step": 9306 }, { "epoch": 0.53, "learning_rate": 9.581243462708006e-05, "loss": 0.7137, "step": 9307 }, { "epoch": 0.53, "learning_rate": 9.579407892625705e-05, "loss": 0.7103, "step": 9308 }, { "epoch": 0.53, "learning_rate": 9.577572336739491e-05, "loss": 0.7309, "step": 9309 }, { "epoch": 0.53, "learning_rate": 9.57573679511132e-05, "loss": 0.6718, "step": 9310 }, { "epoch": 0.53, "learning_rate": 9.573901267803142e-05, "loss": 0.7391, "step": 9311 }, { "epoch": 0.53, "learning_rate": 9.572065754876915e-05, "loss": 0.7769, "step": 9312 }, { "epoch": 0.53, "learning_rate": 9.570230256394596e-05, "loss": 0.6508, "step": 9313 }, { "epoch": 0.53, "learning_rate": 9.568394772418129e-05, "loss": 0.6933, "step": 9314 }, { "epoch": 0.53, "learning_rate": 9.566559303009469e-05, "loss": 0.7731, "step": 9315 }, { "epoch": 0.53, "learning_rate": 9.564723848230571e-05, "loss": 0.6918, "step": 9316 }, { "epoch": 0.53, "learning_rate": 9.562888408143383e-05, "loss": 0.6997, "step": 9317 }, { "epoch": 0.53, "learning_rate": 9.561052982809858e-05, "loss": 0.7059, "step": 9318 }, { "epoch": 0.53, "learning_rate": 9.559217572291948e-05, "loss": 0.8058, "step": 9319 }, { "epoch": 0.53, "learning_rate": 9.557382176651603e-05, "loss": 0.7785, "step": 9320 }, { "epoch": 0.53, "learning_rate": 9.555546795950765e-05, "loss": 0.7491, "step": 9321 }, { "epoch": 0.53, "learning_rate": 9.553711430251388e-05, "loss": 0.6361, "step": 9322 }, { "epoch": 0.53, "learning_rate": 9.551876079615425e-05, "loss": 0.732, "step": 9323 }, { "epoch": 0.53, "learning_rate": 9.550040744104816e-05, "loss": 0.7022, "step": 9324 }, { "epoch": 0.53, "learning_rate": 9.548205423781512e-05, "loss": 0.7422, "step": 9325 }, { "epoch": 0.53, "learning_rate": 9.546370118707463e-05, "loss": 0.5847, "step": 9326 }, { "epoch": 0.53, "learning_rate": 9.544534828944609e-05, "loss": 0.7174, "step": 9327 }, { "epoch": 0.53, "learning_rate": 9.542699554554904e-05, "loss": 0.7484, "step": 9328 }, { "epoch": 0.53, "learning_rate": 9.540864295600283e-05, "loss": 0.7903, "step": 9329 }, { "epoch": 0.53, "learning_rate": 9.539029052142699e-05, "loss": 0.6585, "step": 9330 }, { "epoch": 0.53, "learning_rate": 9.537193824244091e-05, "loss": 0.6724, "step": 9331 }, { "epoch": 0.53, "learning_rate": 9.535358611966407e-05, "loss": 0.8153, "step": 9332 }, { "epoch": 0.53, "learning_rate": 9.533523415371587e-05, "loss": 0.7324, "step": 9333 }, { "epoch": 0.53, "learning_rate": 9.531688234521575e-05, "loss": 0.75, "step": 9334 }, { "epoch": 0.53, "learning_rate": 9.529853069478317e-05, "loss": 0.7014, "step": 9335 }, { "epoch": 0.53, "learning_rate": 9.52801792030375e-05, "loss": 0.7405, "step": 9336 }, { "epoch": 0.53, "learning_rate": 9.526182787059813e-05, "loss": 0.6183, "step": 9337 }, { "epoch": 0.53, "learning_rate": 9.52434766980845e-05, "loss": 0.6765, "step": 9338 }, { "epoch": 0.53, "learning_rate": 9.522512568611603e-05, "loss": 0.7243, "step": 9339 }, { "epoch": 0.53, "learning_rate": 9.520677483531205e-05, "loss": 0.7123, "step": 9340 }, { "epoch": 0.53, "learning_rate": 9.5188424146292e-05, "loss": 0.644, "step": 9341 }, { "epoch": 0.53, "learning_rate": 9.517007361967528e-05, "loss": 0.7605, "step": 9342 }, { "epoch": 0.53, "learning_rate": 9.515172325608127e-05, "loss": 0.7537, "step": 9343 }, { "epoch": 0.53, "learning_rate": 9.513337305612925e-05, "loss": 0.6899, "step": 9344 }, { "epoch": 0.53, "learning_rate": 9.511502302043868e-05, "loss": 0.766, "step": 9345 }, { "epoch": 0.53, "learning_rate": 9.50966731496289e-05, "loss": 0.7561, "step": 9346 }, { "epoch": 0.53, "learning_rate": 9.507832344431923e-05, "loss": 0.711, "step": 9347 }, { "epoch": 0.53, "learning_rate": 9.50599739051291e-05, "loss": 0.8621, "step": 9348 }, { "epoch": 0.53, "learning_rate": 9.504162453267777e-05, "loss": 0.8181, "step": 9349 }, { "epoch": 0.53, "learning_rate": 9.502327532758467e-05, "loss": 0.6842, "step": 9350 }, { "epoch": 0.53, "learning_rate": 9.500492629046901e-05, "loss": 0.6907, "step": 9351 }, { "epoch": 0.53, "learning_rate": 9.498657742195024e-05, "loss": 0.7198, "step": 9352 }, { "epoch": 0.53, "learning_rate": 9.496822872264758e-05, "loss": 0.7315, "step": 9353 }, { "epoch": 0.53, "learning_rate": 9.49498801931804e-05, "loss": 0.6685, "step": 9354 }, { "epoch": 0.53, "learning_rate": 9.493153183416806e-05, "loss": 0.7188, "step": 9355 }, { "epoch": 0.53, "learning_rate": 9.491318364622975e-05, "loss": 0.7465, "step": 9356 }, { "epoch": 0.53, "learning_rate": 9.48948356299849e-05, "loss": 0.6734, "step": 9357 }, { "epoch": 0.53, "learning_rate": 9.48764877860527e-05, "loss": 0.6145, "step": 9358 }, { "epoch": 0.53, "learning_rate": 9.485814011505249e-05, "loss": 0.6435, "step": 9359 }, { "epoch": 0.53, "learning_rate": 9.48397926176035e-05, "loss": 0.7214, "step": 9360 }, { "epoch": 0.53, "learning_rate": 9.482144529432505e-05, "loss": 0.7726, "step": 9361 }, { "epoch": 0.53, "learning_rate": 9.480309814583644e-05, "loss": 0.7326, "step": 9362 }, { "epoch": 0.53, "learning_rate": 9.478475117275684e-05, "loss": 0.6489, "step": 9363 }, { "epoch": 0.53, "learning_rate": 9.476640437570562e-05, "loss": 0.7691, "step": 9364 }, { "epoch": 0.53, "learning_rate": 9.474805775530199e-05, "loss": 0.7163, "step": 9365 }, { "epoch": 0.53, "learning_rate": 9.472971131216514e-05, "loss": 0.7459, "step": 9366 }, { "epoch": 0.53, "learning_rate": 9.471136504691436e-05, "loss": 0.6181, "step": 9367 }, { "epoch": 0.53, "learning_rate": 9.46930189601689e-05, "loss": 0.7374, "step": 9368 }, { "epoch": 0.53, "learning_rate": 9.467467305254793e-05, "loss": 0.7415, "step": 9369 }, { "epoch": 0.53, "learning_rate": 9.465632732467073e-05, "loss": 0.7011, "step": 9370 }, { "epoch": 0.53, "learning_rate": 9.46379817771565e-05, "loss": 0.7216, "step": 9371 }, { "epoch": 0.53, "learning_rate": 9.461963641062449e-05, "loss": 0.7102, "step": 9372 }, { "epoch": 0.53, "learning_rate": 9.460129122569381e-05, "loss": 0.6831, "step": 9373 }, { "epoch": 0.53, "learning_rate": 9.45829462229837e-05, "loss": 0.7626, "step": 9374 }, { "epoch": 0.53, "learning_rate": 9.45646014031134e-05, "loss": 0.6548, "step": 9375 }, { "epoch": 0.53, "learning_rate": 9.454625676670203e-05, "loss": 0.6831, "step": 9376 }, { "epoch": 0.53, "learning_rate": 9.452791231436881e-05, "loss": 0.7461, "step": 9377 }, { "epoch": 0.53, "learning_rate": 9.450956804673289e-05, "loss": 0.7297, "step": 9378 }, { "epoch": 0.53, "learning_rate": 9.449122396441345e-05, "loss": 0.6852, "step": 9379 }, { "epoch": 0.53, "learning_rate": 9.447288006802968e-05, "loss": 0.697, "step": 9380 }, { "epoch": 0.53, "learning_rate": 9.445453635820068e-05, "loss": 0.6305, "step": 9381 }, { "epoch": 0.53, "learning_rate": 9.443619283554562e-05, "loss": 0.6841, "step": 9382 }, { "epoch": 0.53, "learning_rate": 9.441784950068362e-05, "loss": 0.7539, "step": 9383 }, { "epoch": 0.53, "learning_rate": 9.439950635423387e-05, "loss": 0.6288, "step": 9384 }, { "epoch": 0.53, "learning_rate": 9.438116339681545e-05, "loss": 0.5727, "step": 9385 }, { "epoch": 0.53, "learning_rate": 9.43628206290475e-05, "loss": 0.774, "step": 9386 }, { "epoch": 0.53, "learning_rate": 9.434447805154918e-05, "loss": 0.7017, "step": 9387 }, { "epoch": 0.53, "learning_rate": 9.432613566493955e-05, "loss": 0.7638, "step": 9388 }, { "epoch": 0.53, "learning_rate": 9.430779346983766e-05, "loss": 0.7316, "step": 9389 }, { "epoch": 0.53, "learning_rate": 9.42894514668627e-05, "loss": 0.7677, "step": 9390 }, { "epoch": 0.53, "learning_rate": 9.427110965663375e-05, "loss": 0.6917, "step": 9391 }, { "epoch": 0.53, "learning_rate": 9.425276803976985e-05, "loss": 0.7616, "step": 9392 }, { "epoch": 0.53, "learning_rate": 9.423442661689012e-05, "loss": 0.7372, "step": 9393 }, { "epoch": 0.53, "learning_rate": 9.421608538861361e-05, "loss": 0.608, "step": 9394 }, { "epoch": 0.53, "learning_rate": 9.419774435555939e-05, "loss": 0.7133, "step": 9395 }, { "epoch": 0.53, "learning_rate": 9.41794035183465e-05, "loss": 0.7778, "step": 9396 }, { "epoch": 0.53, "learning_rate": 9.416106287759402e-05, "loss": 0.7372, "step": 9397 }, { "epoch": 0.53, "learning_rate": 9.414272243392096e-05, "loss": 0.7042, "step": 9398 }, { "epoch": 0.53, "learning_rate": 9.412438218794638e-05, "loss": 0.7454, "step": 9399 }, { "epoch": 0.53, "learning_rate": 9.410604214028933e-05, "loss": 0.6364, "step": 9400 }, { "epoch": 0.53, "learning_rate": 9.408770229156881e-05, "loss": 0.7128, "step": 9401 }, { "epoch": 0.53, "learning_rate": 9.406936264240386e-05, "loss": 0.6153, "step": 9402 }, { "epoch": 0.53, "learning_rate": 9.405102319341344e-05, "loss": 0.6766, "step": 9403 }, { "epoch": 0.53, "learning_rate": 9.403268394521662e-05, "loss": 0.6105, "step": 9404 }, { "epoch": 0.53, "learning_rate": 9.401434489843234e-05, "loss": 0.6123, "step": 9405 }, { "epoch": 0.53, "learning_rate": 9.399600605367962e-05, "loss": 0.6619, "step": 9406 }, { "epoch": 0.53, "learning_rate": 9.397766741157746e-05, "loss": 0.7642, "step": 9407 }, { "epoch": 0.53, "learning_rate": 9.39593289727448e-05, "loss": 0.6743, "step": 9408 }, { "epoch": 0.53, "learning_rate": 9.394099073780066e-05, "loss": 0.6392, "step": 9409 }, { "epoch": 0.53, "learning_rate": 9.392265270736397e-05, "loss": 0.7599, "step": 9410 }, { "epoch": 0.53, "learning_rate": 9.390431488205366e-05, "loss": 0.7234, "step": 9411 }, { "epoch": 0.53, "learning_rate": 9.38859772624887e-05, "loss": 0.6627, "step": 9412 }, { "epoch": 0.53, "learning_rate": 9.386763984928808e-05, "loss": 0.7272, "step": 9413 }, { "epoch": 0.53, "learning_rate": 9.384930264307067e-05, "loss": 0.7556, "step": 9414 }, { "epoch": 0.53, "learning_rate": 9.383096564445542e-05, "loss": 0.7057, "step": 9415 }, { "epoch": 0.53, "learning_rate": 9.381262885406129e-05, "loss": 0.6111, "step": 9416 }, { "epoch": 0.53, "learning_rate": 9.379429227250718e-05, "loss": 0.6673, "step": 9417 }, { "epoch": 0.53, "learning_rate": 9.377595590041194e-05, "loss": 0.6784, "step": 9418 }, { "epoch": 0.53, "learning_rate": 9.37576197383945e-05, "loss": 0.7201, "step": 9419 }, { "epoch": 0.53, "learning_rate": 9.37392837870738e-05, "loss": 0.7008, "step": 9420 }, { "epoch": 0.53, "learning_rate": 9.372094804706867e-05, "loss": 0.6846, "step": 9421 }, { "epoch": 0.53, "learning_rate": 9.370261251899802e-05, "loss": 0.7408, "step": 9422 }, { "epoch": 0.53, "learning_rate": 9.36842772034807e-05, "loss": 0.695, "step": 9423 }, { "epoch": 0.53, "learning_rate": 9.366594210113563e-05, "loss": 0.7003, "step": 9424 }, { "epoch": 0.53, "learning_rate": 9.364760721258159e-05, "loss": 0.6895, "step": 9425 }, { "epoch": 0.53, "learning_rate": 9.362927253843748e-05, "loss": 0.7398, "step": 9426 }, { "epoch": 0.53, "learning_rate": 9.361093807932211e-05, "loss": 0.7387, "step": 9427 }, { "epoch": 0.53, "learning_rate": 9.359260383585433e-05, "loss": 0.5783, "step": 9428 }, { "epoch": 0.53, "learning_rate": 9.357426980865301e-05, "loss": 0.7782, "step": 9429 }, { "epoch": 0.53, "learning_rate": 9.35559359983369e-05, "loss": 0.583, "step": 9430 }, { "epoch": 0.53, "learning_rate": 9.35376024055249e-05, "loss": 0.7732, "step": 9431 }, { "epoch": 0.54, "learning_rate": 9.351926903083572e-05, "loss": 0.7985, "step": 9432 }, { "epoch": 0.54, "learning_rate": 9.350093587488823e-05, "loss": 0.7909, "step": 9433 }, { "epoch": 0.54, "learning_rate": 9.348260293830117e-05, "loss": 0.7176, "step": 9434 }, { "epoch": 0.54, "learning_rate": 9.346427022169337e-05, "loss": 0.6942, "step": 9435 }, { "epoch": 0.54, "learning_rate": 9.344593772568361e-05, "loss": 0.7006, "step": 9436 }, { "epoch": 0.54, "learning_rate": 9.34276054508906e-05, "loss": 0.6566, "step": 9437 }, { "epoch": 0.54, "learning_rate": 9.340927339793318e-05, "loss": 0.6259, "step": 9438 }, { "epoch": 0.54, "learning_rate": 9.339094156743007e-05, "loss": 0.6122, "step": 9439 }, { "epoch": 0.54, "learning_rate": 9.337260996000002e-05, "loss": 0.8328, "step": 9440 }, { "epoch": 0.54, "learning_rate": 9.335427857626172e-05, "loss": 0.5883, "step": 9441 }, { "epoch": 0.54, "learning_rate": 9.333594741683398e-05, "loss": 0.6748, "step": 9442 }, { "epoch": 0.54, "learning_rate": 9.331761648233548e-05, "loss": 0.6332, "step": 9443 }, { "epoch": 0.54, "learning_rate": 9.329928577338494e-05, "loss": 0.7438, "step": 9444 }, { "epoch": 0.54, "learning_rate": 9.32809552906011e-05, "loss": 0.6779, "step": 9445 }, { "epoch": 0.54, "learning_rate": 9.326262503460266e-05, "loss": 0.6912, "step": 9446 }, { "epoch": 0.54, "learning_rate": 9.324429500600826e-05, "loss": 0.6585, "step": 9447 }, { "epoch": 0.54, "learning_rate": 9.32259652054366e-05, "loss": 0.7373, "step": 9448 }, { "epoch": 0.54, "learning_rate": 9.320763563350643e-05, "loss": 0.6656, "step": 9449 }, { "epoch": 0.54, "learning_rate": 9.318930629083633e-05, "loss": 0.7258, "step": 9450 }, { "epoch": 0.54, "learning_rate": 9.3170977178045e-05, "loss": 0.7158, "step": 9451 }, { "epoch": 0.54, "learning_rate": 9.315264829575114e-05, "loss": 0.7295, "step": 9452 }, { "epoch": 0.54, "learning_rate": 9.313431964457334e-05, "loss": 0.744, "step": 9453 }, { "epoch": 0.54, "learning_rate": 9.311599122513029e-05, "loss": 0.7799, "step": 9454 }, { "epoch": 0.54, "learning_rate": 9.309766303804055e-05, "loss": 0.7377, "step": 9455 }, { "epoch": 0.54, "learning_rate": 9.307933508392279e-05, "loss": 0.6983, "step": 9456 }, { "epoch": 0.54, "learning_rate": 9.30610073633956e-05, "loss": 0.6859, "step": 9457 }, { "epoch": 0.54, "learning_rate": 9.304267987707766e-05, "loss": 0.6791, "step": 9458 }, { "epoch": 0.54, "learning_rate": 9.302435262558747e-05, "loss": 0.7005, "step": 9459 }, { "epoch": 0.54, "learning_rate": 9.300602560954368e-05, "loss": 0.6286, "step": 9460 }, { "epoch": 0.54, "learning_rate": 9.298769882956492e-05, "loss": 0.684, "step": 9461 }, { "epoch": 0.54, "learning_rate": 9.29693722862697e-05, "loss": 0.6792, "step": 9462 }, { "epoch": 0.54, "learning_rate": 9.295104598027656e-05, "loss": 0.8016, "step": 9463 }, { "epoch": 0.54, "learning_rate": 9.293271991220412e-05, "loss": 0.7859, "step": 9464 }, { "epoch": 0.54, "learning_rate": 9.291439408267093e-05, "loss": 0.7273, "step": 9465 }, { "epoch": 0.54, "learning_rate": 9.289606849229551e-05, "loss": 0.7194, "step": 9466 }, { "epoch": 0.54, "learning_rate": 9.287774314169643e-05, "loss": 0.7601, "step": 9467 }, { "epoch": 0.54, "learning_rate": 9.28594180314922e-05, "loss": 0.8116, "step": 9468 }, { "epoch": 0.54, "learning_rate": 9.284109316230133e-05, "loss": 0.6786, "step": 9469 }, { "epoch": 0.54, "learning_rate": 9.282276853474232e-05, "loss": 0.6943, "step": 9470 }, { "epoch": 0.54, "learning_rate": 9.280444414943372e-05, "loss": 0.6926, "step": 9471 }, { "epoch": 0.54, "learning_rate": 9.278612000699397e-05, "loss": 0.8197, "step": 9472 }, { "epoch": 0.54, "learning_rate": 9.27677961080416e-05, "loss": 0.7072, "step": 9473 }, { "epoch": 0.54, "learning_rate": 9.27494724531951e-05, "loss": 0.7377, "step": 9474 }, { "epoch": 0.54, "learning_rate": 9.273114904307289e-05, "loss": 0.7738, "step": 9475 }, { "epoch": 0.54, "learning_rate": 9.271282587829352e-05, "loss": 0.7033, "step": 9476 }, { "epoch": 0.54, "learning_rate": 9.269450295947533e-05, "loss": 0.7123, "step": 9477 }, { "epoch": 0.54, "learning_rate": 9.267618028723686e-05, "loss": 0.6597, "step": 9478 }, { "epoch": 0.54, "learning_rate": 9.265785786219647e-05, "loss": 0.6984, "step": 9479 }, { "epoch": 0.54, "learning_rate": 9.263953568497265e-05, "loss": 0.7258, "step": 9480 }, { "epoch": 0.54, "learning_rate": 9.262121375618382e-05, "loss": 0.7574, "step": 9481 }, { "epoch": 0.54, "learning_rate": 9.260289207644836e-05, "loss": 0.6193, "step": 9482 }, { "epoch": 0.54, "learning_rate": 9.258457064638474e-05, "loss": 0.6499, "step": 9483 }, { "epoch": 0.54, "learning_rate": 9.256624946661125e-05, "loss": 0.6466, "step": 9484 }, { "epoch": 0.54, "learning_rate": 9.254792853774636e-05, "loss": 0.7519, "step": 9485 }, { "epoch": 0.54, "learning_rate": 9.252960786040842e-05, "loss": 0.6038, "step": 9486 }, { "epoch": 0.54, "learning_rate": 9.251128743521582e-05, "loss": 0.7959, "step": 9487 }, { "epoch": 0.54, "learning_rate": 9.249296726278688e-05, "loss": 0.7784, "step": 9488 }, { "epoch": 0.54, "learning_rate": 9.247464734373998e-05, "loss": 0.7024, "step": 9489 }, { "epoch": 0.54, "learning_rate": 9.245632767869352e-05, "loss": 0.7451, "step": 9490 }, { "epoch": 0.54, "learning_rate": 9.243800826826578e-05, "loss": 0.7415, "step": 9491 }, { "epoch": 0.54, "learning_rate": 9.241968911307504e-05, "loss": 0.7381, "step": 9492 }, { "epoch": 0.54, "learning_rate": 9.24013702137397e-05, "loss": 0.7325, "step": 9493 }, { "epoch": 0.54, "learning_rate": 9.238305157087805e-05, "loss": 0.6991, "step": 9494 }, { "epoch": 0.54, "learning_rate": 9.236473318510837e-05, "loss": 0.7183, "step": 9495 }, { "epoch": 0.54, "learning_rate": 9.234641505704895e-05, "loss": 0.6591, "step": 9496 }, { "epoch": 0.54, "learning_rate": 9.232809718731814e-05, "loss": 0.7284, "step": 9497 }, { "epoch": 0.54, "learning_rate": 9.230977957653419e-05, "loss": 0.6838, "step": 9498 }, { "epoch": 0.54, "learning_rate": 9.229146222531529e-05, "loss": 0.7622, "step": 9499 }, { "epoch": 0.54, "learning_rate": 9.227314513427979e-05, "loss": 0.6981, "step": 9500 }, { "epoch": 0.54, "learning_rate": 9.225482830404586e-05, "loss": 0.7177, "step": 9501 }, { "epoch": 0.54, "learning_rate": 9.223651173523182e-05, "loss": 0.6266, "step": 9502 }, { "epoch": 0.54, "learning_rate": 9.221819542845587e-05, "loss": 0.7262, "step": 9503 }, { "epoch": 0.54, "learning_rate": 9.219987938433621e-05, "loss": 0.7206, "step": 9504 }, { "epoch": 0.54, "learning_rate": 9.21815636034911e-05, "loss": 0.7356, "step": 9505 }, { "epoch": 0.54, "learning_rate": 9.216324808653876e-05, "loss": 0.7806, "step": 9506 }, { "epoch": 0.54, "learning_rate": 9.214493283409731e-05, "loss": 0.6715, "step": 9507 }, { "epoch": 0.54, "learning_rate": 9.212661784678497e-05, "loss": 0.6994, "step": 9508 }, { "epoch": 0.54, "learning_rate": 9.210830312521992e-05, "loss": 0.7092, "step": 9509 }, { "epoch": 0.54, "learning_rate": 9.208998867002036e-05, "loss": 0.7399, "step": 9510 }, { "epoch": 0.54, "learning_rate": 9.20716744818044e-05, "loss": 0.6917, "step": 9511 }, { "epoch": 0.54, "learning_rate": 9.205336056119023e-05, "loss": 0.7917, "step": 9512 }, { "epoch": 0.54, "learning_rate": 9.203504690879602e-05, "loss": 0.6727, "step": 9513 }, { "epoch": 0.54, "learning_rate": 9.201673352523985e-05, "loss": 0.7052, "step": 9514 }, { "epoch": 0.54, "learning_rate": 9.199842041113983e-05, "loss": 0.5913, "step": 9515 }, { "epoch": 0.54, "learning_rate": 9.198010756711412e-05, "loss": 0.7194, "step": 9516 }, { "epoch": 0.54, "learning_rate": 9.196179499378082e-05, "loss": 0.7229, "step": 9517 }, { "epoch": 0.54, "learning_rate": 9.1943482691758e-05, "loss": 0.6148, "step": 9518 }, { "epoch": 0.54, "learning_rate": 9.192517066166378e-05, "loss": 0.6935, "step": 9519 }, { "epoch": 0.54, "learning_rate": 9.190685890411627e-05, "loss": 0.8273, "step": 9520 }, { "epoch": 0.54, "learning_rate": 9.188854741973344e-05, "loss": 0.7263, "step": 9521 }, { "epoch": 0.54, "learning_rate": 9.187023620913342e-05, "loss": 0.7224, "step": 9522 }, { "epoch": 0.54, "learning_rate": 9.185192527293425e-05, "loss": 0.6513, "step": 9523 }, { "epoch": 0.54, "learning_rate": 9.183361461175396e-05, "loss": 0.6942, "step": 9524 }, { "epoch": 0.54, "learning_rate": 9.181530422621059e-05, "loss": 0.7044, "step": 9525 }, { "epoch": 0.54, "learning_rate": 9.17969941169222e-05, "loss": 0.724, "step": 9526 }, { "epoch": 0.54, "learning_rate": 9.177868428450674e-05, "loss": 0.7782, "step": 9527 }, { "epoch": 0.54, "learning_rate": 9.17603747295823e-05, "loss": 0.7619, "step": 9528 }, { "epoch": 0.54, "learning_rate": 9.174206545276677e-05, "loss": 0.7108, "step": 9529 }, { "epoch": 0.54, "learning_rate": 9.172375645467822e-05, "loss": 0.7101, "step": 9530 }, { "epoch": 0.54, "learning_rate": 9.170544773593457e-05, "loss": 0.7973, "step": 9531 }, { "epoch": 0.54, "learning_rate": 9.168713929715383e-05, "loss": 0.7284, "step": 9532 }, { "epoch": 0.54, "learning_rate": 9.166883113895392e-05, "loss": 0.6809, "step": 9533 }, { "epoch": 0.54, "learning_rate": 9.16505232619528e-05, "loss": 0.7163, "step": 9534 }, { "epoch": 0.54, "learning_rate": 9.163221566676847e-05, "loss": 0.6541, "step": 9535 }, { "epoch": 0.54, "learning_rate": 9.161390835401878e-05, "loss": 0.7051, "step": 9536 }, { "epoch": 0.54, "learning_rate": 9.159560132432166e-05, "loss": 0.7478, "step": 9537 }, { "epoch": 0.54, "learning_rate": 9.157729457829503e-05, "loss": 0.6329, "step": 9538 }, { "epoch": 0.54, "learning_rate": 9.155898811655683e-05, "loss": 0.729, "step": 9539 }, { "epoch": 0.54, "learning_rate": 9.15406819397249e-05, "loss": 0.7367, "step": 9540 }, { "epoch": 0.54, "learning_rate": 9.152237604841712e-05, "loss": 0.6573, "step": 9541 }, { "epoch": 0.54, "learning_rate": 9.150407044325141e-05, "loss": 0.7347, "step": 9542 }, { "epoch": 0.54, "learning_rate": 9.148576512484564e-05, "loss": 0.6944, "step": 9543 }, { "epoch": 0.54, "learning_rate": 9.146746009381756e-05, "loss": 0.7071, "step": 9544 }, { "epoch": 0.54, "learning_rate": 9.144915535078509e-05, "loss": 0.6842, "step": 9545 }, { "epoch": 0.54, "learning_rate": 9.143085089636609e-05, "loss": 0.7127, "step": 9546 }, { "epoch": 0.54, "learning_rate": 9.14125467311783e-05, "loss": 0.7351, "step": 9547 }, { "epoch": 0.54, "learning_rate": 9.13942428558396e-05, "loss": 0.6436, "step": 9548 }, { "epoch": 0.54, "learning_rate": 9.137593927096776e-05, "loss": 0.8032, "step": 9549 }, { "epoch": 0.54, "learning_rate": 9.135763597718064e-05, "loss": 0.806, "step": 9550 }, { "epoch": 0.54, "learning_rate": 9.133933297509592e-05, "loss": 0.6908, "step": 9551 }, { "epoch": 0.54, "learning_rate": 9.132103026533145e-05, "loss": 0.7257, "step": 9552 }, { "epoch": 0.54, "learning_rate": 9.130272784850494e-05, "loss": 0.6229, "step": 9553 }, { "epoch": 0.54, "learning_rate": 9.128442572523417e-05, "loss": 0.7872, "step": 9554 }, { "epoch": 0.54, "learning_rate": 9.126612389613694e-05, "loss": 0.8244, "step": 9555 }, { "epoch": 0.54, "learning_rate": 9.124782236183088e-05, "loss": 0.6131, "step": 9556 }, { "epoch": 0.54, "learning_rate": 9.122952112293383e-05, "loss": 0.7195, "step": 9557 }, { "epoch": 0.54, "learning_rate": 9.12112201800634e-05, "loss": 0.7048, "step": 9558 }, { "epoch": 0.54, "learning_rate": 9.119291953383736e-05, "loss": 0.7123, "step": 9559 }, { "epoch": 0.54, "learning_rate": 9.117461918487334e-05, "loss": 0.7535, "step": 9560 }, { "epoch": 0.54, "learning_rate": 9.115631913378912e-05, "loss": 0.6342, "step": 9561 }, { "epoch": 0.54, "learning_rate": 9.113801938120227e-05, "loss": 0.7874, "step": 9562 }, { "epoch": 0.54, "learning_rate": 9.111971992773053e-05, "loss": 0.8116, "step": 9563 }, { "epoch": 0.54, "learning_rate": 9.110142077399156e-05, "loss": 0.669, "step": 9564 }, { "epoch": 0.54, "learning_rate": 9.108312192060298e-05, "loss": 0.68, "step": 9565 }, { "epoch": 0.54, "learning_rate": 9.106482336818239e-05, "loss": 0.7151, "step": 9566 }, { "epoch": 0.54, "learning_rate": 9.104652511734743e-05, "loss": 0.6781, "step": 9567 }, { "epoch": 0.54, "learning_rate": 9.102822716871575e-05, "loss": 0.6878, "step": 9568 }, { "epoch": 0.54, "learning_rate": 9.100992952290492e-05, "loss": 0.7864, "step": 9569 }, { "epoch": 0.54, "learning_rate": 9.099163218053254e-05, "loss": 0.5746, "step": 9570 }, { "epoch": 0.54, "learning_rate": 9.097333514221624e-05, "loss": 0.6246, "step": 9571 }, { "epoch": 0.54, "learning_rate": 9.095503840857356e-05, "loss": 0.6127, "step": 9572 }, { "epoch": 0.54, "learning_rate": 9.093674198022201e-05, "loss": 0.6948, "step": 9573 }, { "epoch": 0.54, "learning_rate": 9.091844585777918e-05, "loss": 0.7341, "step": 9574 }, { "epoch": 0.54, "learning_rate": 9.090015004186266e-05, "loss": 0.7465, "step": 9575 }, { "epoch": 0.54, "learning_rate": 9.08818545330899e-05, "loss": 0.6958, "step": 9576 }, { "epoch": 0.54, "learning_rate": 9.08635593320785e-05, "loss": 0.7595, "step": 9577 }, { "epoch": 0.54, "learning_rate": 9.08452644394459e-05, "loss": 0.6975, "step": 9578 }, { "epoch": 0.54, "learning_rate": 9.082696985580964e-05, "loss": 0.7004, "step": 9579 }, { "epoch": 0.54, "learning_rate": 9.080867558178725e-05, "loss": 0.6273, "step": 9580 }, { "epoch": 0.54, "learning_rate": 9.079038161799615e-05, "loss": 0.6835, "step": 9581 }, { "epoch": 0.54, "learning_rate": 9.077208796505378e-05, "loss": 0.799, "step": 9582 }, { "epoch": 0.54, "learning_rate": 9.075379462357766e-05, "loss": 0.7229, "step": 9583 }, { "epoch": 0.54, "learning_rate": 9.073550159418523e-05, "loss": 0.683, "step": 9584 }, { "epoch": 0.54, "learning_rate": 9.07172088774939e-05, "loss": 0.7535, "step": 9585 }, { "epoch": 0.54, "learning_rate": 9.069891647412113e-05, "loss": 0.7368, "step": 9586 }, { "epoch": 0.54, "learning_rate": 9.068062438468436e-05, "loss": 0.594, "step": 9587 }, { "epoch": 0.54, "learning_rate": 9.066233260980093e-05, "loss": 0.8172, "step": 9588 }, { "epoch": 0.54, "learning_rate": 9.064404115008823e-05, "loss": 0.667, "step": 9589 }, { "epoch": 0.54, "learning_rate": 9.06257500061637e-05, "loss": 0.6674, "step": 9590 }, { "epoch": 0.54, "learning_rate": 9.060745917864469e-05, "loss": 0.6999, "step": 9591 }, { "epoch": 0.54, "learning_rate": 9.058916866814858e-05, "loss": 0.7618, "step": 9592 }, { "epoch": 0.54, "learning_rate": 9.05708784752927e-05, "loss": 0.6859, "step": 9593 }, { "epoch": 0.54, "learning_rate": 9.055258860069443e-05, "loss": 0.689, "step": 9594 }, { "epoch": 0.54, "learning_rate": 9.053429904497102e-05, "loss": 0.656, "step": 9595 }, { "epoch": 0.54, "learning_rate": 9.051600980873985e-05, "loss": 0.7013, "step": 9596 }, { "epoch": 0.54, "learning_rate": 9.049772089261825e-05, "loss": 0.7273, "step": 9597 }, { "epoch": 0.54, "learning_rate": 9.047943229722345e-05, "loss": 0.6987, "step": 9598 }, { "epoch": 0.54, "learning_rate": 9.04611440231728e-05, "loss": 0.7512, "step": 9599 }, { "epoch": 0.54, "learning_rate": 9.044285607108355e-05, "loss": 0.6885, "step": 9600 }, { "epoch": 0.54, "learning_rate": 9.042456844157299e-05, "loss": 0.7301, "step": 9601 }, { "epoch": 0.54, "learning_rate": 9.040628113525837e-05, "loss": 0.7292, "step": 9602 }, { "epoch": 0.54, "learning_rate": 9.038799415275689e-05, "loss": 0.7885, "step": 9603 }, { "epoch": 0.54, "learning_rate": 9.036970749468584e-05, "loss": 0.7248, "step": 9604 }, { "epoch": 0.54, "learning_rate": 9.035142116166239e-05, "loss": 0.7173, "step": 9605 }, { "epoch": 0.54, "learning_rate": 9.033313515430381e-05, "loss": 0.773, "step": 9606 }, { "epoch": 0.54, "learning_rate": 9.031484947322724e-05, "loss": 0.7843, "step": 9607 }, { "epoch": 0.55, "learning_rate": 9.02965641190499e-05, "loss": 0.7087, "step": 9608 }, { "epoch": 0.55, "learning_rate": 9.027827909238902e-05, "loss": 0.7065, "step": 9609 }, { "epoch": 0.55, "learning_rate": 9.025999439386169e-05, "loss": 0.7332, "step": 9610 }, { "epoch": 0.55, "learning_rate": 9.024171002408506e-05, "loss": 0.7751, "step": 9611 }, { "epoch": 0.55, "learning_rate": 9.022342598367633e-05, "loss": 0.6208, "step": 9612 }, { "epoch": 0.55, "learning_rate": 9.020514227325264e-05, "loss": 0.7583, "step": 9613 }, { "epoch": 0.55, "learning_rate": 9.018685889343103e-05, "loss": 0.6911, "step": 9614 }, { "epoch": 0.55, "learning_rate": 9.01685758448287e-05, "loss": 0.6681, "step": 9615 }, { "epoch": 0.55, "learning_rate": 9.015029312806274e-05, "loss": 0.7474, "step": 9616 }, { "epoch": 0.55, "learning_rate": 9.013201074375022e-05, "loss": 0.6711, "step": 9617 }, { "epoch": 0.55, "learning_rate": 9.011372869250819e-05, "loss": 0.781, "step": 9618 }, { "epoch": 0.55, "learning_rate": 9.009544697495374e-05, "loss": 0.7508, "step": 9619 }, { "epoch": 0.55, "learning_rate": 9.007716559170395e-05, "loss": 0.6986, "step": 9620 }, { "epoch": 0.55, "learning_rate": 9.005888454337583e-05, "loss": 0.7998, "step": 9621 }, { "epoch": 0.55, "learning_rate": 9.004060383058643e-05, "loss": 0.6932, "step": 9622 }, { "epoch": 0.55, "learning_rate": 9.002232345395276e-05, "loss": 0.6571, "step": 9623 }, { "epoch": 0.55, "learning_rate": 9.000404341409189e-05, "loss": 0.6441, "step": 9624 }, { "epoch": 0.55, "learning_rate": 8.998576371162073e-05, "loss": 0.7452, "step": 9625 }, { "epoch": 0.55, "learning_rate": 8.996748434715632e-05, "loss": 0.7643, "step": 9626 }, { "epoch": 0.55, "learning_rate": 8.99492053213156e-05, "loss": 0.6909, "step": 9627 }, { "epoch": 0.55, "learning_rate": 8.993092663471556e-05, "loss": 0.5988, "step": 9628 }, { "epoch": 0.55, "learning_rate": 8.991264828797319e-05, "loss": 0.5565, "step": 9629 }, { "epoch": 0.55, "learning_rate": 8.989437028170537e-05, "loss": 0.7525, "step": 9630 }, { "epoch": 0.55, "learning_rate": 8.987609261652909e-05, "loss": 0.5934, "step": 9631 }, { "epoch": 0.55, "learning_rate": 8.985781529306119e-05, "loss": 0.6496, "step": 9632 }, { "epoch": 0.55, "learning_rate": 8.983953831191865e-05, "loss": 0.6729, "step": 9633 }, { "epoch": 0.55, "learning_rate": 8.982126167371832e-05, "loss": 0.6909, "step": 9634 }, { "epoch": 0.55, "learning_rate": 8.98029853790771e-05, "loss": 0.7565, "step": 9635 }, { "epoch": 0.55, "learning_rate": 8.978470942861188e-05, "loss": 0.7022, "step": 9636 }, { "epoch": 0.55, "learning_rate": 8.97664338229395e-05, "loss": 0.5776, "step": 9637 }, { "epoch": 0.55, "learning_rate": 8.974815856267683e-05, "loss": 0.6574, "step": 9638 }, { "epoch": 0.55, "learning_rate": 8.972988364844071e-05, "loss": 0.7459, "step": 9639 }, { "epoch": 0.55, "learning_rate": 8.971160908084793e-05, "loss": 0.7513, "step": 9640 }, { "epoch": 0.55, "learning_rate": 8.96933348605153e-05, "loss": 0.8279, "step": 9641 }, { "epoch": 0.55, "learning_rate": 8.967506098805969e-05, "loss": 0.6259, "step": 9642 }, { "epoch": 0.55, "learning_rate": 8.965678746409782e-05, "loss": 0.755, "step": 9643 }, { "epoch": 0.55, "learning_rate": 8.96385142892465e-05, "loss": 0.7998, "step": 9644 }, { "epoch": 0.55, "learning_rate": 8.962024146412251e-05, "loss": 0.7892, "step": 9645 }, { "epoch": 0.55, "learning_rate": 8.960196898934263e-05, "loss": 0.6979, "step": 9646 }, { "epoch": 0.55, "learning_rate": 8.958369686552351e-05, "loss": 0.8043, "step": 9647 }, { "epoch": 0.55, "learning_rate": 8.956542509328195e-05, "loss": 0.7653, "step": 9648 }, { "epoch": 0.55, "learning_rate": 8.954715367323468e-05, "loss": 0.6902, "step": 9649 }, { "epoch": 0.55, "learning_rate": 8.952888260599835e-05, "loss": 0.7359, "step": 9650 }, { "epoch": 0.55, "learning_rate": 8.951061189218972e-05, "loss": 0.7782, "step": 9651 }, { "epoch": 0.55, "learning_rate": 8.949234153242544e-05, "loss": 0.7466, "step": 9652 }, { "epoch": 0.55, "learning_rate": 8.947407152732218e-05, "loss": 0.7727, "step": 9653 }, { "epoch": 0.55, "learning_rate": 8.945580187749666e-05, "loss": 0.8051, "step": 9654 }, { "epoch": 0.55, "learning_rate": 8.943753258356546e-05, "loss": 0.7251, "step": 9655 }, { "epoch": 0.55, "learning_rate": 8.94192636461452e-05, "loss": 0.7185, "step": 9656 }, { "epoch": 0.55, "learning_rate": 8.940099506585256e-05, "loss": 0.8113, "step": 9657 }, { "epoch": 0.55, "learning_rate": 8.938272684330414e-05, "loss": 0.7465, "step": 9658 }, { "epoch": 0.55, "learning_rate": 8.936445897911652e-05, "loss": 0.7812, "step": 9659 }, { "epoch": 0.55, "learning_rate": 8.934619147390629e-05, "loss": 0.787, "step": 9660 }, { "epoch": 0.55, "learning_rate": 8.932792432829009e-05, "loss": 0.717, "step": 9661 }, { "epoch": 0.55, "learning_rate": 8.930965754288442e-05, "loss": 0.7023, "step": 9662 }, { "epoch": 0.55, "learning_rate": 8.929139111830582e-05, "loss": 0.6971, "step": 9663 }, { "epoch": 0.55, "learning_rate": 8.927312505517085e-05, "loss": 0.7526, "step": 9664 }, { "epoch": 0.55, "learning_rate": 8.925485935409606e-05, "loss": 0.7134, "step": 9665 }, { "epoch": 0.55, "learning_rate": 8.923659401569793e-05, "loss": 0.7619, "step": 9666 }, { "epoch": 0.55, "learning_rate": 8.9218329040593e-05, "loss": 0.6571, "step": 9667 }, { "epoch": 0.55, "learning_rate": 8.920006442939772e-05, "loss": 0.6694, "step": 9668 }, { "epoch": 0.55, "learning_rate": 8.918180018272863e-05, "loss": 0.7154, "step": 9669 }, { "epoch": 0.55, "learning_rate": 8.916353630120211e-05, "loss": 0.6818, "step": 9670 }, { "epoch": 0.55, "learning_rate": 8.914527278543469e-05, "loss": 0.6542, "step": 9671 }, { "epoch": 0.55, "learning_rate": 8.912700963604275e-05, "loss": 0.7079, "step": 9672 }, { "epoch": 0.55, "learning_rate": 8.910874685364275e-05, "loss": 0.6236, "step": 9673 }, { "epoch": 0.55, "learning_rate": 8.909048443885112e-05, "loss": 0.7482, "step": 9674 }, { "epoch": 0.55, "learning_rate": 8.907222239228425e-05, "loss": 0.716, "step": 9675 }, { "epoch": 0.55, "learning_rate": 8.905396071455858e-05, "loss": 0.6801, "step": 9676 }, { "epoch": 0.55, "learning_rate": 8.903569940629039e-05, "loss": 0.7355, "step": 9677 }, { "epoch": 0.55, "learning_rate": 8.901743846809612e-05, "loss": 0.714, "step": 9678 }, { "epoch": 0.55, "learning_rate": 8.899917790059208e-05, "loss": 0.6207, "step": 9679 }, { "epoch": 0.55, "learning_rate": 8.898091770439465e-05, "loss": 0.7532, "step": 9680 }, { "epoch": 0.55, "learning_rate": 8.896265788012015e-05, "loss": 0.8157, "step": 9681 }, { "epoch": 0.55, "learning_rate": 8.894439842838488e-05, "loss": 0.7818, "step": 9682 }, { "epoch": 0.55, "learning_rate": 8.892613934980521e-05, "loss": 0.6605, "step": 9683 }, { "epoch": 0.55, "learning_rate": 8.890788064499734e-05, "loss": 0.6094, "step": 9684 }, { "epoch": 0.55, "learning_rate": 8.888962231457756e-05, "loss": 0.7628, "step": 9685 }, { "epoch": 0.55, "learning_rate": 8.88713643591622e-05, "loss": 0.6428, "step": 9686 }, { "epoch": 0.55, "learning_rate": 8.885310677936746e-05, "loss": 0.6551, "step": 9687 }, { "epoch": 0.55, "learning_rate": 8.883484957580962e-05, "loss": 0.7209, "step": 9688 }, { "epoch": 0.55, "learning_rate": 8.881659274910486e-05, "loss": 0.6595, "step": 9689 }, { "epoch": 0.55, "learning_rate": 8.879833629986945e-05, "loss": 0.6925, "step": 9690 }, { "epoch": 0.55, "learning_rate": 8.878008022871958e-05, "loss": 0.6746, "step": 9691 }, { "epoch": 0.55, "learning_rate": 8.876182453627141e-05, "loss": 0.704, "step": 9692 }, { "epoch": 0.55, "learning_rate": 8.87435692231411e-05, "loss": 0.6959, "step": 9693 }, { "epoch": 0.55, "learning_rate": 8.87253142899449e-05, "loss": 0.7032, "step": 9694 }, { "epoch": 0.55, "learning_rate": 8.870705973729889e-05, "loss": 0.627, "step": 9695 }, { "epoch": 0.55, "learning_rate": 8.868880556581923e-05, "loss": 0.6764, "step": 9696 }, { "epoch": 0.55, "learning_rate": 8.867055177612206e-05, "loss": 0.673, "step": 9697 }, { "epoch": 0.55, "learning_rate": 8.86522983688235e-05, "loss": 0.6821, "step": 9698 }, { "epoch": 0.55, "learning_rate": 8.863404534453958e-05, "loss": 0.7538, "step": 9699 }, { "epoch": 0.55, "learning_rate": 8.861579270388648e-05, "loss": 0.7231, "step": 9700 }, { "epoch": 0.55, "learning_rate": 8.85975404474802e-05, "loss": 0.7773, "step": 9701 }, { "epoch": 0.55, "learning_rate": 8.857928857593684e-05, "loss": 0.638, "step": 9702 }, { "epoch": 0.55, "learning_rate": 8.856103708987245e-05, "loss": 0.7194, "step": 9703 }, { "epoch": 0.55, "learning_rate": 8.854278598990305e-05, "loss": 0.723, "step": 9704 }, { "epoch": 0.55, "learning_rate": 8.852453527664466e-05, "loss": 0.8769, "step": 9705 }, { "epoch": 0.55, "learning_rate": 8.850628495071336e-05, "loss": 0.7011, "step": 9706 }, { "epoch": 0.55, "learning_rate": 8.848803501272504e-05, "loss": 0.643, "step": 9707 }, { "epoch": 0.55, "learning_rate": 8.846978546329573e-05, "loss": 0.8373, "step": 9708 }, { "epoch": 0.55, "learning_rate": 8.845153630304139e-05, "loss": 0.8036, "step": 9709 }, { "epoch": 0.55, "learning_rate": 8.843328753257801e-05, "loss": 0.6534, "step": 9710 }, { "epoch": 0.55, "learning_rate": 8.841503915252149e-05, "loss": 0.7218, "step": 9711 }, { "epoch": 0.55, "learning_rate": 8.83967911634878e-05, "loss": 0.7105, "step": 9712 }, { "epoch": 0.55, "learning_rate": 8.837854356609284e-05, "loss": 0.7016, "step": 9713 }, { "epoch": 0.55, "learning_rate": 8.836029636095252e-05, "loss": 0.6437, "step": 9714 }, { "epoch": 0.55, "learning_rate": 8.834204954868269e-05, "loss": 0.6406, "step": 9715 }, { "epoch": 0.55, "learning_rate": 8.83238031298993e-05, "loss": 0.7696, "step": 9716 }, { "epoch": 0.55, "learning_rate": 8.830555710521813e-05, "loss": 0.7217, "step": 9717 }, { "epoch": 0.55, "learning_rate": 8.82873114752551e-05, "loss": 0.6845, "step": 9718 }, { "epoch": 0.55, "learning_rate": 8.826906624062603e-05, "loss": 0.6619, "step": 9719 }, { "epoch": 0.55, "learning_rate": 8.825082140194678e-05, "loss": 0.6645, "step": 9720 }, { "epoch": 0.55, "learning_rate": 8.823257695983306e-05, "loss": 0.7878, "step": 9721 }, { "epoch": 0.55, "learning_rate": 8.821433291490074e-05, "loss": 0.6774, "step": 9722 }, { "epoch": 0.55, "learning_rate": 8.819608926776562e-05, "loss": 0.7668, "step": 9723 }, { "epoch": 0.55, "learning_rate": 8.817784601904342e-05, "loss": 0.6785, "step": 9724 }, { "epoch": 0.55, "learning_rate": 8.81596031693499e-05, "loss": 0.6593, "step": 9725 }, { "epoch": 0.55, "learning_rate": 8.814136071930086e-05, "loss": 0.7959, "step": 9726 }, { "epoch": 0.55, "learning_rate": 8.812311866951198e-05, "loss": 0.7429, "step": 9727 }, { "epoch": 0.55, "learning_rate": 8.810487702059903e-05, "loss": 0.7697, "step": 9728 }, { "epoch": 0.55, "learning_rate": 8.808663577317764e-05, "loss": 0.6468, "step": 9729 }, { "epoch": 0.55, "learning_rate": 8.806839492786355e-05, "loss": 0.6365, "step": 9730 }, { "epoch": 0.55, "learning_rate": 8.80501544852724e-05, "loss": 0.7452, "step": 9731 }, { "epoch": 0.55, "learning_rate": 8.803191444601991e-05, "loss": 0.6558, "step": 9732 }, { "epoch": 0.55, "learning_rate": 8.801367481072166e-05, "loss": 0.6668, "step": 9733 }, { "epoch": 0.55, "learning_rate": 8.799543557999335e-05, "loss": 0.7697, "step": 9734 }, { "epoch": 0.55, "learning_rate": 8.797719675445058e-05, "loss": 0.7598, "step": 9735 }, { "epoch": 0.55, "learning_rate": 8.795895833470895e-05, "loss": 0.7827, "step": 9736 }, { "epoch": 0.55, "learning_rate": 8.794072032138402e-05, "loss": 0.7669, "step": 9737 }, { "epoch": 0.55, "learning_rate": 8.792248271509141e-05, "loss": 0.7893, "step": 9738 }, { "epoch": 0.55, "learning_rate": 8.790424551644671e-05, "loss": 0.7321, "step": 9739 }, { "epoch": 0.55, "learning_rate": 8.788600872606542e-05, "loss": 0.6536, "step": 9740 }, { "epoch": 0.55, "learning_rate": 8.78677723445631e-05, "loss": 0.6622, "step": 9741 }, { "epoch": 0.55, "learning_rate": 8.78495363725553e-05, "loss": 0.7371, "step": 9742 }, { "epoch": 0.55, "learning_rate": 8.783130081065756e-05, "loss": 0.7202, "step": 9743 }, { "epoch": 0.55, "learning_rate": 8.781306565948528e-05, "loss": 0.7527, "step": 9744 }, { "epoch": 0.55, "learning_rate": 8.7794830919654e-05, "loss": 0.6817, "step": 9745 }, { "epoch": 0.55, "learning_rate": 8.777659659177917e-05, "loss": 0.6778, "step": 9746 }, { "epoch": 0.55, "learning_rate": 8.775836267647626e-05, "loss": 0.7307, "step": 9747 }, { "epoch": 0.55, "learning_rate": 8.774012917436074e-05, "loss": 0.7037, "step": 9748 }, { "epoch": 0.55, "learning_rate": 8.7721896086048e-05, "loss": 0.7271, "step": 9749 }, { "epoch": 0.55, "learning_rate": 8.77036634121535e-05, "loss": 0.7403, "step": 9750 }, { "epoch": 0.55, "learning_rate": 8.768543115329256e-05, "loss": 0.7523, "step": 9751 }, { "epoch": 0.55, "learning_rate": 8.766719931008065e-05, "loss": 0.6645, "step": 9752 }, { "epoch": 0.55, "learning_rate": 8.764896788313308e-05, "loss": 0.6521, "step": 9753 }, { "epoch": 0.55, "learning_rate": 8.763073687306524e-05, "loss": 0.7545, "step": 9754 }, { "epoch": 0.55, "learning_rate": 8.761250628049249e-05, "loss": 0.7307, "step": 9755 }, { "epoch": 0.55, "learning_rate": 8.75942761060301e-05, "loss": 0.7302, "step": 9756 }, { "epoch": 0.55, "learning_rate": 8.757604635029349e-05, "loss": 0.6759, "step": 9757 }, { "epoch": 0.55, "learning_rate": 8.755781701389786e-05, "loss": 0.7134, "step": 9758 }, { "epoch": 0.55, "learning_rate": 8.753958809745855e-05, "loss": 0.7608, "step": 9759 }, { "epoch": 0.55, "learning_rate": 8.752135960159079e-05, "loss": 0.7077, "step": 9760 }, { "epoch": 0.55, "learning_rate": 8.75031315269099e-05, "loss": 0.7207, "step": 9761 }, { "epoch": 0.55, "learning_rate": 8.748490387403107e-05, "loss": 0.6203, "step": 9762 }, { "epoch": 0.55, "learning_rate": 8.746667664356956e-05, "loss": 0.6934, "step": 9763 }, { "epoch": 0.55, "learning_rate": 8.744844983614061e-05, "loss": 0.6478, "step": 9764 }, { "epoch": 0.55, "learning_rate": 8.743022345235942e-05, "loss": 0.7129, "step": 9765 }, { "epoch": 0.55, "learning_rate": 8.74119974928411e-05, "loss": 0.7683, "step": 9766 }, { "epoch": 0.55, "learning_rate": 8.739377195820088e-05, "loss": 0.7808, "step": 9767 }, { "epoch": 0.55, "learning_rate": 8.737554684905394e-05, "loss": 0.6796, "step": 9768 }, { "epoch": 0.55, "learning_rate": 8.735732216601538e-05, "loss": 0.8094, "step": 9769 }, { "epoch": 0.55, "learning_rate": 8.733909790970035e-05, "loss": 0.7522, "step": 9770 }, { "epoch": 0.55, "learning_rate": 8.7320874080724e-05, "loss": 0.8056, "step": 9771 }, { "epoch": 0.55, "learning_rate": 8.730265067970142e-05, "loss": 0.7732, "step": 9772 }, { "epoch": 0.55, "learning_rate": 8.728442770724763e-05, "loss": 0.6744, "step": 9773 }, { "epoch": 0.55, "learning_rate": 8.726620516397777e-05, "loss": 0.7398, "step": 9774 }, { "epoch": 0.55, "learning_rate": 8.724798305050688e-05, "loss": 0.6881, "step": 9775 }, { "epoch": 0.55, "learning_rate": 8.722976136745e-05, "loss": 0.6591, "step": 9776 }, { "epoch": 0.55, "learning_rate": 8.721154011542218e-05, "loss": 0.8988, "step": 9777 }, { "epoch": 0.55, "learning_rate": 8.71933192950384e-05, "loss": 0.6796, "step": 9778 }, { "epoch": 0.55, "learning_rate": 8.717509890691368e-05, "loss": 0.6698, "step": 9779 }, { "epoch": 0.55, "learning_rate": 8.715687895166304e-05, "loss": 0.7289, "step": 9780 }, { "epoch": 0.55, "learning_rate": 8.713865942990141e-05, "loss": 0.7382, "step": 9781 }, { "epoch": 0.55, "learning_rate": 8.712044034224374e-05, "loss": 0.7167, "step": 9782 }, { "epoch": 0.55, "learning_rate": 8.710222168930497e-05, "loss": 0.7129, "step": 9783 }, { "epoch": 0.55, "learning_rate": 8.708400347170007e-05, "loss": 0.7044, "step": 9784 }, { "epoch": 0.56, "learning_rate": 8.706578569004392e-05, "loss": 0.7759, "step": 9785 }, { "epoch": 0.56, "learning_rate": 8.70475683449514e-05, "loss": 0.6758, "step": 9786 }, { "epoch": 0.56, "learning_rate": 8.702935143703747e-05, "loss": 0.6999, "step": 9787 }, { "epoch": 0.56, "learning_rate": 8.701113496691691e-05, "loss": 0.6923, "step": 9788 }, { "epoch": 0.56, "learning_rate": 8.699291893520462e-05, "loss": 0.6778, "step": 9789 }, { "epoch": 0.56, "learning_rate": 8.697470334251543e-05, "loss": 0.6998, "step": 9790 }, { "epoch": 0.56, "learning_rate": 8.695648818946414e-05, "loss": 0.706, "step": 9791 }, { "epoch": 0.56, "learning_rate": 8.693827347666559e-05, "loss": 0.6733, "step": 9792 }, { "epoch": 0.56, "learning_rate": 8.692005920473457e-05, "loss": 0.745, "step": 9793 }, { "epoch": 0.56, "learning_rate": 8.69018453742859e-05, "loss": 0.6789, "step": 9794 }, { "epoch": 0.56, "learning_rate": 8.688363198593424e-05, "loss": 0.7021, "step": 9795 }, { "epoch": 0.56, "learning_rate": 8.686541904029441e-05, "loss": 0.6794, "step": 9796 }, { "epoch": 0.56, "learning_rate": 8.684720653798114e-05, "loss": 0.7618, "step": 9797 }, { "epoch": 0.56, "learning_rate": 8.682899447960914e-05, "loss": 0.7588, "step": 9798 }, { "epoch": 0.56, "learning_rate": 8.681078286579311e-05, "loss": 0.681, "step": 9799 }, { "epoch": 0.56, "learning_rate": 8.679257169714776e-05, "loss": 0.7177, "step": 9800 }, { "epoch": 0.56, "learning_rate": 8.677436097428775e-05, "loss": 0.821, "step": 9801 }, { "epoch": 0.56, "learning_rate": 8.675615069782777e-05, "loss": 0.7168, "step": 9802 }, { "epoch": 0.56, "learning_rate": 8.673794086838241e-05, "loss": 0.6811, "step": 9803 }, { "epoch": 0.56, "learning_rate": 8.671973148656634e-05, "loss": 0.7272, "step": 9804 }, { "epoch": 0.56, "learning_rate": 8.670152255299414e-05, "loss": 0.7555, "step": 9805 }, { "epoch": 0.56, "learning_rate": 8.668331406828045e-05, "loss": 0.7558, "step": 9806 }, { "epoch": 0.56, "learning_rate": 8.666510603303981e-05, "loss": 0.7249, "step": 9807 }, { "epoch": 0.56, "learning_rate": 8.664689844788683e-05, "loss": 0.7286, "step": 9808 }, { "epoch": 0.56, "learning_rate": 8.662869131343607e-05, "loss": 0.6759, "step": 9809 }, { "epoch": 0.56, "learning_rate": 8.661048463030203e-05, "loss": 0.7101, "step": 9810 }, { "epoch": 0.56, "learning_rate": 8.659227839909924e-05, "loss": 0.7478, "step": 9811 }, { "epoch": 0.56, "learning_rate": 8.65740726204422e-05, "loss": 0.7705, "step": 9812 }, { "epoch": 0.56, "learning_rate": 8.655586729494544e-05, "loss": 0.7402, "step": 9813 }, { "epoch": 0.56, "learning_rate": 8.653766242322341e-05, "loss": 0.6925, "step": 9814 }, { "epoch": 0.56, "learning_rate": 8.651945800589058e-05, "loss": 0.6743, "step": 9815 }, { "epoch": 0.56, "learning_rate": 8.65012540435614e-05, "loss": 0.7699, "step": 9816 }, { "epoch": 0.56, "learning_rate": 8.648305053685034e-05, "loss": 0.7058, "step": 9817 }, { "epoch": 0.56, "learning_rate": 8.646484748637172e-05, "loss": 0.7422, "step": 9818 }, { "epoch": 0.56, "learning_rate": 8.644664489273999e-05, "loss": 0.82, "step": 9819 }, { "epoch": 0.56, "learning_rate": 8.642844275656957e-05, "loss": 0.6288, "step": 9820 }, { "epoch": 0.56, "learning_rate": 8.641024107847478e-05, "loss": 0.7892, "step": 9821 }, { "epoch": 0.56, "learning_rate": 8.639203985907e-05, "loss": 0.7602, "step": 9822 }, { "epoch": 0.56, "learning_rate": 8.637383909896957e-05, "loss": 0.7597, "step": 9823 }, { "epoch": 0.56, "learning_rate": 8.635563879878783e-05, "loss": 0.7163, "step": 9824 }, { "epoch": 0.56, "learning_rate": 8.633743895913901e-05, "loss": 0.7475, "step": 9825 }, { "epoch": 0.56, "learning_rate": 8.63192395806375e-05, "loss": 0.6933, "step": 9826 }, { "epoch": 0.56, "learning_rate": 8.63010406638975e-05, "loss": 0.6711, "step": 9827 }, { "epoch": 0.56, "learning_rate": 8.628284220953331e-05, "loss": 0.7307, "step": 9828 }, { "epoch": 0.56, "learning_rate": 8.626464421815919e-05, "loss": 0.6672, "step": 9829 }, { "epoch": 0.56, "learning_rate": 8.624644669038932e-05, "loss": 0.7537, "step": 9830 }, { "epoch": 0.56, "learning_rate": 8.622824962683796e-05, "loss": 0.6512, "step": 9831 }, { "epoch": 0.56, "learning_rate": 8.621005302811933e-05, "loss": 0.6865, "step": 9832 }, { "epoch": 0.56, "learning_rate": 8.619185689484755e-05, "loss": 0.7173, "step": 9833 }, { "epoch": 0.56, "learning_rate": 8.617366122763679e-05, "loss": 0.761, "step": 9834 }, { "epoch": 0.56, "learning_rate": 8.615546602710125e-05, "loss": 0.803, "step": 9835 }, { "epoch": 0.56, "learning_rate": 8.613727129385502e-05, "loss": 0.8181, "step": 9836 }, { "epoch": 0.56, "learning_rate": 8.611907702851224e-05, "loss": 0.7261, "step": 9837 }, { "epoch": 0.56, "learning_rate": 8.610088323168704e-05, "loss": 0.6934, "step": 9838 }, { "epoch": 0.56, "learning_rate": 8.608268990399349e-05, "loss": 0.8578, "step": 9839 }, { "epoch": 0.56, "learning_rate": 8.606449704604562e-05, "loss": 0.6347, "step": 9840 }, { "epoch": 0.56, "learning_rate": 8.604630465845752e-05, "loss": 0.7169, "step": 9841 }, { "epoch": 0.56, "learning_rate": 8.602811274184326e-05, "loss": 0.7449, "step": 9842 }, { "epoch": 0.56, "learning_rate": 8.600992129681681e-05, "loss": 0.7621, "step": 9843 }, { "epoch": 0.56, "learning_rate": 8.599173032399221e-05, "loss": 0.7195, "step": 9844 }, { "epoch": 0.56, "learning_rate": 8.597353982398347e-05, "loss": 0.7139, "step": 9845 }, { "epoch": 0.56, "learning_rate": 8.595534979740457e-05, "loss": 0.7811, "step": 9846 }, { "epoch": 0.56, "learning_rate": 8.59371602448694e-05, "loss": 0.8165, "step": 9847 }, { "epoch": 0.56, "learning_rate": 8.591897116699196e-05, "loss": 0.6883, "step": 9848 }, { "epoch": 0.56, "learning_rate": 8.590078256438618e-05, "loss": 0.6932, "step": 9849 }, { "epoch": 0.56, "learning_rate": 8.588259443766596e-05, "loss": 0.8121, "step": 9850 }, { "epoch": 0.56, "learning_rate": 8.586440678744522e-05, "loss": 0.6999, "step": 9851 }, { "epoch": 0.56, "learning_rate": 8.58462196143378e-05, "loss": 0.7072, "step": 9852 }, { "epoch": 0.56, "learning_rate": 8.582803291895758e-05, "loss": 0.7097, "step": 9853 }, { "epoch": 0.56, "learning_rate": 8.580984670191848e-05, "loss": 0.6765, "step": 9854 }, { "epoch": 0.56, "learning_rate": 8.579166096383422e-05, "loss": 0.5759, "step": 9855 }, { "epoch": 0.56, "learning_rate": 8.577347570531865e-05, "loss": 0.7576, "step": 9856 }, { "epoch": 0.56, "learning_rate": 8.57552909269856e-05, "loss": 0.7541, "step": 9857 }, { "epoch": 0.56, "learning_rate": 8.573710662944885e-05, "loss": 0.6705, "step": 9858 }, { "epoch": 0.56, "learning_rate": 8.571892281332213e-05, "loss": 0.8148, "step": 9859 }, { "epoch": 0.56, "learning_rate": 8.570073947921923e-05, "loss": 0.6481, "step": 9860 }, { "epoch": 0.56, "learning_rate": 8.568255662775391e-05, "loss": 0.6317, "step": 9861 }, { "epoch": 0.56, "learning_rate": 8.566437425953983e-05, "loss": 0.5535, "step": 9862 }, { "epoch": 0.56, "learning_rate": 8.56461923751907e-05, "loss": 0.6568, "step": 9863 }, { "epoch": 0.56, "learning_rate": 8.562801097532022e-05, "loss": 0.7162, "step": 9864 }, { "epoch": 0.56, "learning_rate": 8.560983006054208e-05, "loss": 0.715, "step": 9865 }, { "epoch": 0.56, "learning_rate": 8.55916496314699e-05, "loss": 0.7195, "step": 9866 }, { "epoch": 0.56, "learning_rate": 8.557346968871735e-05, "loss": 0.7779, "step": 9867 }, { "epoch": 0.56, "learning_rate": 8.555529023289801e-05, "loss": 0.6714, "step": 9868 }, { "epoch": 0.56, "learning_rate": 8.553711126462556e-05, "loss": 0.709, "step": 9869 }, { "epoch": 0.56, "learning_rate": 8.551893278451348e-05, "loss": 0.8502, "step": 9870 }, { "epoch": 0.56, "learning_rate": 8.550075479317542e-05, "loss": 0.7256, "step": 9871 }, { "epoch": 0.56, "learning_rate": 8.54825772912249e-05, "loss": 0.7801, "step": 9872 }, { "epoch": 0.56, "learning_rate": 8.546440027927546e-05, "loss": 0.8107, "step": 9873 }, { "epoch": 0.56, "learning_rate": 8.544622375794065e-05, "loss": 0.5994, "step": 9874 }, { "epoch": 0.56, "learning_rate": 8.542804772783394e-05, "loss": 0.7365, "step": 9875 }, { "epoch": 0.56, "learning_rate": 8.540987218956888e-05, "loss": 0.7552, "step": 9876 }, { "epoch": 0.56, "learning_rate": 8.539169714375885e-05, "loss": 0.7243, "step": 9877 }, { "epoch": 0.56, "learning_rate": 8.537352259101739e-05, "loss": 0.6096, "step": 9878 }, { "epoch": 0.56, "learning_rate": 8.535534853195786e-05, "loss": 0.5845, "step": 9879 }, { "epoch": 0.56, "learning_rate": 8.533717496719373e-05, "loss": 0.7533, "step": 9880 }, { "epoch": 0.56, "learning_rate": 8.531900189733842e-05, "loss": 0.819, "step": 9881 }, { "epoch": 0.56, "learning_rate": 8.530082932300529e-05, "loss": 0.7414, "step": 9882 }, { "epoch": 0.56, "learning_rate": 8.528265724480775e-05, "loss": 0.7404, "step": 9883 }, { "epoch": 0.56, "learning_rate": 8.52644856633591e-05, "loss": 0.716, "step": 9884 }, { "epoch": 0.56, "learning_rate": 8.52463145792727e-05, "loss": 0.654, "step": 9885 }, { "epoch": 0.56, "learning_rate": 8.522814399316185e-05, "loss": 0.6743, "step": 9886 }, { "epoch": 0.56, "learning_rate": 8.520997390563993e-05, "loss": 0.7634, "step": 9887 }, { "epoch": 0.56, "learning_rate": 8.519180431732015e-05, "loss": 0.6061, "step": 9888 }, { "epoch": 0.56, "learning_rate": 8.517363522881579e-05, "loss": 0.6966, "step": 9889 }, { "epoch": 0.56, "learning_rate": 8.515546664074016e-05, "loss": 0.7048, "step": 9890 }, { "epoch": 0.56, "learning_rate": 8.513729855370649e-05, "loss": 0.7134, "step": 9891 }, { "epoch": 0.56, "learning_rate": 8.511913096832793e-05, "loss": 0.6829, "step": 9892 }, { "epoch": 0.56, "learning_rate": 8.510096388521772e-05, "loss": 0.7001, "step": 9893 }, { "epoch": 0.56, "learning_rate": 8.508279730498908e-05, "loss": 0.7356, "step": 9894 }, { "epoch": 0.56, "learning_rate": 8.506463122825512e-05, "loss": 0.7311, "step": 9895 }, { "epoch": 0.56, "learning_rate": 8.504646565562906e-05, "loss": 0.6104, "step": 9896 }, { "epoch": 0.56, "learning_rate": 8.502830058772397e-05, "loss": 0.7331, "step": 9897 }, { "epoch": 0.56, "learning_rate": 8.501013602515306e-05, "loss": 0.7325, "step": 9898 }, { "epoch": 0.56, "learning_rate": 8.49919719685293e-05, "loss": 0.733, "step": 9899 }, { "epoch": 0.56, "learning_rate": 8.497380841846589e-05, "loss": 0.7714, "step": 9900 }, { "epoch": 0.56, "learning_rate": 8.495564537557582e-05, "loss": 0.6474, "step": 9901 }, { "epoch": 0.56, "learning_rate": 8.493748284047218e-05, "loss": 0.7161, "step": 9902 }, { "epoch": 0.56, "learning_rate": 8.491932081376803e-05, "loss": 0.6987, "step": 9903 }, { "epoch": 0.56, "learning_rate": 8.490115929607631e-05, "loss": 0.658, "step": 9904 }, { "epoch": 0.56, "learning_rate": 8.488299828801007e-05, "loss": 0.7321, "step": 9905 }, { "epoch": 0.56, "learning_rate": 8.486483779018233e-05, "loss": 0.7721, "step": 9906 }, { "epoch": 0.56, "learning_rate": 8.484667780320597e-05, "loss": 0.6948, "step": 9907 }, { "epoch": 0.56, "learning_rate": 8.482851832769398e-05, "loss": 0.6787, "step": 9908 }, { "epoch": 0.56, "learning_rate": 8.481035936425927e-05, "loss": 0.7479, "step": 9909 }, { "epoch": 0.56, "learning_rate": 8.479220091351479e-05, "loss": 0.6996, "step": 9910 }, { "epoch": 0.56, "learning_rate": 8.47740429760734e-05, "loss": 0.6207, "step": 9911 }, { "epoch": 0.56, "learning_rate": 8.4755885552548e-05, "loss": 0.6836, "step": 9912 }, { "epoch": 0.56, "learning_rate": 8.473772864355145e-05, "loss": 0.6827, "step": 9913 }, { "epoch": 0.56, "learning_rate": 8.471957224969658e-05, "loss": 0.6922, "step": 9914 }, { "epoch": 0.56, "learning_rate": 8.47014163715962e-05, "loss": 0.787, "step": 9915 }, { "epoch": 0.56, "learning_rate": 8.468326100986317e-05, "loss": 0.696, "step": 9916 }, { "epoch": 0.56, "learning_rate": 8.466510616511022e-05, "loss": 0.701, "step": 9917 }, { "epoch": 0.56, "learning_rate": 8.464695183795015e-05, "loss": 0.6959, "step": 9918 }, { "epoch": 0.56, "learning_rate": 8.462879802899575e-05, "loss": 0.7143, "step": 9919 }, { "epoch": 0.56, "learning_rate": 8.461064473885976e-05, "loss": 0.6982, "step": 9920 }, { "epoch": 0.56, "learning_rate": 8.45924919681548e-05, "loss": 0.7028, "step": 9921 }, { "epoch": 0.56, "learning_rate": 8.457433971749367e-05, "loss": 0.6766, "step": 9922 }, { "epoch": 0.56, "learning_rate": 8.455618798748904e-05, "loss": 0.6701, "step": 9923 }, { "epoch": 0.56, "learning_rate": 8.453803677875355e-05, "loss": 0.6536, "step": 9924 }, { "epoch": 0.56, "learning_rate": 8.451988609189987e-05, "loss": 0.743, "step": 9925 }, { "epoch": 0.56, "learning_rate": 8.450173592754065e-05, "loss": 0.6857, "step": 9926 }, { "epoch": 0.56, "learning_rate": 8.448358628628846e-05, "loss": 0.6707, "step": 9927 }, { "epoch": 0.56, "learning_rate": 8.446543716875597e-05, "loss": 0.641, "step": 9928 }, { "epoch": 0.56, "learning_rate": 8.444728857555572e-05, "loss": 0.7309, "step": 9929 }, { "epoch": 0.56, "learning_rate": 8.442914050730021e-05, "loss": 0.7576, "step": 9930 }, { "epoch": 0.56, "learning_rate": 8.441099296460208e-05, "loss": 0.6714, "step": 9931 }, { "epoch": 0.56, "learning_rate": 8.439284594807381e-05, "loss": 0.6712, "step": 9932 }, { "epoch": 0.56, "learning_rate": 8.437469945832792e-05, "loss": 0.7114, "step": 9933 }, { "epoch": 0.56, "learning_rate": 8.435655349597689e-05, "loss": 0.6907, "step": 9934 }, { "epoch": 0.56, "learning_rate": 8.433840806163326e-05, "loss": 0.7513, "step": 9935 }, { "epoch": 0.56, "learning_rate": 8.432026315590942e-05, "loss": 0.7022, "step": 9936 }, { "epoch": 0.56, "learning_rate": 8.430211877941779e-05, "loss": 0.775, "step": 9937 }, { "epoch": 0.56, "learning_rate": 8.42839749327708e-05, "loss": 0.7289, "step": 9938 }, { "epoch": 0.56, "learning_rate": 8.426583161658092e-05, "loss": 0.7461, "step": 9939 }, { "epoch": 0.56, "learning_rate": 8.424768883146045e-05, "loss": 0.6628, "step": 9940 }, { "epoch": 0.56, "learning_rate": 8.422954657802183e-05, "loss": 0.7079, "step": 9941 }, { "epoch": 0.56, "learning_rate": 8.421140485687734e-05, "loss": 0.7692, "step": 9942 }, { "epoch": 0.56, "learning_rate": 8.419326366863938e-05, "loss": 0.6787, "step": 9943 }, { "epoch": 0.56, "learning_rate": 8.417512301392019e-05, "loss": 0.79, "step": 9944 }, { "epoch": 0.56, "learning_rate": 8.415698289333213e-05, "loss": 0.6448, "step": 9945 }, { "epoch": 0.56, "learning_rate": 8.413884330748741e-05, "loss": 0.7095, "step": 9946 }, { "epoch": 0.56, "learning_rate": 8.412070425699834e-05, "loss": 0.6128, "step": 9947 }, { "epoch": 0.56, "learning_rate": 8.410256574247716e-05, "loss": 0.7117, "step": 9948 }, { "epoch": 0.56, "learning_rate": 8.408442776453605e-05, "loss": 0.6581, "step": 9949 }, { "epoch": 0.56, "learning_rate": 8.40662903237873e-05, "loss": 0.695, "step": 9950 }, { "epoch": 0.56, "learning_rate": 8.4048153420843e-05, "loss": 0.7886, "step": 9951 }, { "epoch": 0.56, "learning_rate": 8.403001705631536e-05, "loss": 0.7127, "step": 9952 }, { "epoch": 0.56, "learning_rate": 8.401188123081653e-05, "loss": 0.7123, "step": 9953 }, { "epoch": 0.56, "learning_rate": 8.399374594495861e-05, "loss": 0.6563, "step": 9954 }, { "epoch": 0.56, "learning_rate": 8.397561119935379e-05, "loss": 0.6705, "step": 9955 }, { "epoch": 0.56, "learning_rate": 8.395747699461408e-05, "loss": 0.7106, "step": 9956 }, { "epoch": 0.56, "learning_rate": 8.393934333135165e-05, "loss": 0.681, "step": 9957 }, { "epoch": 0.56, "learning_rate": 8.392121021017845e-05, "loss": 0.8124, "step": 9958 }, { "epoch": 0.56, "learning_rate": 8.390307763170659e-05, "loss": 0.635, "step": 9959 }, { "epoch": 0.56, "learning_rate": 8.388494559654807e-05, "loss": 0.7026, "step": 9960 }, { "epoch": 0.57, "learning_rate": 8.386681410531491e-05, "loss": 0.7159, "step": 9961 }, { "epoch": 0.57, "learning_rate": 8.384868315861907e-05, "loss": 0.6851, "step": 9962 }, { "epoch": 0.57, "learning_rate": 8.383055275707252e-05, "loss": 0.7146, "step": 9963 }, { "epoch": 0.57, "learning_rate": 8.381242290128726e-05, "loss": 0.7098, "step": 9964 }, { "epoch": 0.57, "learning_rate": 8.379429359187518e-05, "loss": 0.7469, "step": 9965 }, { "epoch": 0.57, "learning_rate": 8.377616482944816e-05, "loss": 0.7217, "step": 9966 }, { "epoch": 0.57, "learning_rate": 8.375803661461812e-05, "loss": 0.6878, "step": 9967 }, { "epoch": 0.57, "learning_rate": 8.373990894799698e-05, "loss": 0.7529, "step": 9968 }, { "epoch": 0.57, "learning_rate": 8.372178183019652e-05, "loss": 0.7484, "step": 9969 }, { "epoch": 0.57, "learning_rate": 8.370365526182861e-05, "loss": 0.7458, "step": 9970 }, { "epoch": 0.57, "learning_rate": 8.368552924350512e-05, "loss": 0.7275, "step": 9971 }, { "epoch": 0.57, "learning_rate": 8.366740377583781e-05, "loss": 0.7428, "step": 9972 }, { "epoch": 0.57, "learning_rate": 8.36492788594384e-05, "loss": 0.7186, "step": 9973 }, { "epoch": 0.57, "learning_rate": 8.363115449491876e-05, "loss": 0.7567, "step": 9974 }, { "epoch": 0.57, "learning_rate": 8.361303068289054e-05, "loss": 0.7838, "step": 9975 }, { "epoch": 0.57, "learning_rate": 8.359490742396552e-05, "loss": 0.7369, "step": 9976 }, { "epoch": 0.57, "learning_rate": 8.35767847187554e-05, "loss": 0.6698, "step": 9977 }, { "epoch": 0.57, "learning_rate": 8.355866256787188e-05, "loss": 0.7737, "step": 9978 }, { "epoch": 0.57, "learning_rate": 8.35405409719266e-05, "loss": 0.7219, "step": 9979 }, { "epoch": 0.57, "learning_rate": 8.352241993153127e-05, "loss": 0.7316, "step": 9980 }, { "epoch": 0.57, "learning_rate": 8.350429944729746e-05, "loss": 0.7756, "step": 9981 }, { "epoch": 0.57, "learning_rate": 8.348617951983679e-05, "loss": 0.705, "step": 9982 }, { "epoch": 0.57, "learning_rate": 8.346806014976086e-05, "loss": 0.7145, "step": 9983 }, { "epoch": 0.57, "learning_rate": 8.344994133768129e-05, "loss": 0.7312, "step": 9984 }, { "epoch": 0.57, "learning_rate": 8.343182308420958e-05, "loss": 0.7836, "step": 9985 }, { "epoch": 0.57, "learning_rate": 8.341370538995731e-05, "loss": 0.6712, "step": 9986 }, { "epoch": 0.57, "learning_rate": 8.339558825553597e-05, "loss": 0.7493, "step": 9987 }, { "epoch": 0.57, "learning_rate": 8.337747168155709e-05, "loss": 0.6436, "step": 9988 }, { "epoch": 0.57, "learning_rate": 8.33593556686321e-05, "loss": 0.8215, "step": 9989 }, { "epoch": 0.57, "learning_rate": 8.334124021737254e-05, "loss": 0.7892, "step": 9990 }, { "epoch": 0.57, "learning_rate": 8.332312532838978e-05, "loss": 0.7298, "step": 9991 }, { "epoch": 0.57, "learning_rate": 8.330501100229527e-05, "loss": 0.7024, "step": 9992 }, { "epoch": 0.57, "learning_rate": 8.328689723970045e-05, "loss": 0.6553, "step": 9993 }, { "epoch": 0.57, "learning_rate": 8.326878404121666e-05, "loss": 0.7005, "step": 9994 }, { "epoch": 0.57, "learning_rate": 8.325067140745534e-05, "loss": 0.681, "step": 9995 }, { "epoch": 0.57, "learning_rate": 8.323255933902772e-05, "loss": 0.6895, "step": 9996 }, { "epoch": 0.57, "learning_rate": 8.321444783654524e-05, "loss": 0.7763, "step": 9997 }, { "epoch": 0.57, "learning_rate": 8.319633690061915e-05, "loss": 0.7454, "step": 9998 }, { "epoch": 0.57, "learning_rate": 8.317822653186074e-05, "loss": 0.7824, "step": 9999 }, { "epoch": 0.57, "learning_rate": 8.316011673088132e-05, "loss": 0.6493, "step": 10000 }, { "epoch": 0.57, "learning_rate": 8.314200749829212e-05, "loss": 0.5991, "step": 10001 }, { "epoch": 0.57, "learning_rate": 8.31238988347044e-05, "loss": 0.6204, "step": 10002 }, { "epoch": 0.57, "learning_rate": 8.310579074072932e-05, "loss": 0.7127, "step": 10003 }, { "epoch": 0.57, "learning_rate": 8.308768321697815e-05, "loss": 0.6951, "step": 10004 }, { "epoch": 0.57, "learning_rate": 8.306957626406198e-05, "loss": 0.7404, "step": 10005 }, { "epoch": 0.57, "learning_rate": 8.305146988259205e-05, "loss": 0.7767, "step": 10006 }, { "epoch": 0.57, "learning_rate": 8.303336407317943e-05, "loss": 0.7008, "step": 10007 }, { "epoch": 0.57, "learning_rate": 8.301525883643527e-05, "loss": 0.7811, "step": 10008 }, { "epoch": 0.57, "learning_rate": 8.299715417297071e-05, "loss": 0.6469, "step": 10009 }, { "epoch": 0.57, "learning_rate": 8.297905008339677e-05, "loss": 0.6728, "step": 10010 }, { "epoch": 0.57, "learning_rate": 8.296094656832451e-05, "loss": 0.8254, "step": 10011 }, { "epoch": 0.57, "learning_rate": 8.294284362836498e-05, "loss": 0.7306, "step": 10012 }, { "epoch": 0.57, "learning_rate": 8.292474126412924e-05, "loss": 0.6877, "step": 10013 }, { "epoch": 0.57, "learning_rate": 8.290663947622826e-05, "loss": 0.6474, "step": 10014 }, { "epoch": 0.57, "learning_rate": 8.2888538265273e-05, "loss": 0.6696, "step": 10015 }, { "epoch": 0.57, "learning_rate": 8.287043763187448e-05, "loss": 0.687, "step": 10016 }, { "epoch": 0.57, "learning_rate": 8.285233757664364e-05, "loss": 0.6843, "step": 10017 }, { "epoch": 0.57, "learning_rate": 8.283423810019132e-05, "loss": 0.6925, "step": 10018 }, { "epoch": 0.57, "learning_rate": 8.281613920312852e-05, "loss": 0.7319, "step": 10019 }, { "epoch": 0.57, "learning_rate": 8.279804088606606e-05, "loss": 0.6676, "step": 10020 }, { "epoch": 0.57, "learning_rate": 8.277994314961483e-05, "loss": 0.7064, "step": 10021 }, { "epoch": 0.57, "learning_rate": 8.276184599438571e-05, "loss": 0.842, "step": 10022 }, { "epoch": 0.57, "learning_rate": 8.274374942098948e-05, "loss": 0.6515, "step": 10023 }, { "epoch": 0.57, "learning_rate": 8.2725653430037e-05, "loss": 0.7178, "step": 10024 }, { "epoch": 0.57, "learning_rate": 8.270755802213896e-05, "loss": 0.6747, "step": 10025 }, { "epoch": 0.57, "learning_rate": 8.268946319790624e-05, "loss": 0.693, "step": 10026 }, { "epoch": 0.57, "learning_rate": 8.267136895794949e-05, "loss": 0.8126, "step": 10027 }, { "epoch": 0.57, "learning_rate": 8.26532753028795e-05, "loss": 0.7125, "step": 10028 }, { "epoch": 0.57, "learning_rate": 8.263518223330697e-05, "loss": 0.6189, "step": 10029 }, { "epoch": 0.57, "learning_rate": 8.261708974984257e-05, "loss": 0.6418, "step": 10030 }, { "epoch": 0.57, "learning_rate": 8.259899785309697e-05, "loss": 0.7366, "step": 10031 }, { "epoch": 0.57, "learning_rate": 8.258090654368087e-05, "loss": 0.6807, "step": 10032 }, { "epoch": 0.57, "learning_rate": 8.256281582220485e-05, "loss": 0.634, "step": 10033 }, { "epoch": 0.57, "learning_rate": 8.25447256892795e-05, "loss": 0.7023, "step": 10034 }, { "epoch": 0.57, "learning_rate": 8.252663614551546e-05, "loss": 0.6383, "step": 10035 }, { "epoch": 0.57, "learning_rate": 8.250854719152324e-05, "loss": 0.6952, "step": 10036 }, { "epoch": 0.57, "learning_rate": 8.249045882791344e-05, "loss": 0.8067, "step": 10037 }, { "epoch": 0.57, "learning_rate": 8.247237105529659e-05, "loss": 0.6542, "step": 10038 }, { "epoch": 0.57, "learning_rate": 8.24542838742832e-05, "loss": 0.7522, "step": 10039 }, { "epoch": 0.57, "learning_rate": 8.243619728548371e-05, "loss": 0.7101, "step": 10040 }, { "epoch": 0.57, "learning_rate": 8.241811128950862e-05, "loss": 0.7272, "step": 10041 }, { "epoch": 0.57, "learning_rate": 8.240002588696842e-05, "loss": 0.7263, "step": 10042 }, { "epoch": 0.57, "learning_rate": 8.238194107847346e-05, "loss": 0.834, "step": 10043 }, { "epoch": 0.57, "learning_rate": 8.23638568646342e-05, "loss": 0.6514, "step": 10044 }, { "epoch": 0.57, "learning_rate": 8.234577324606105e-05, "loss": 0.5963, "step": 10045 }, { "epoch": 0.57, "learning_rate": 8.232769022336438e-05, "loss": 0.7881, "step": 10046 }, { "epoch": 0.57, "learning_rate": 8.230960779715446e-05, "loss": 0.6309, "step": 10047 }, { "epoch": 0.57, "learning_rate": 8.229152596804168e-05, "loss": 0.6852, "step": 10048 }, { "epoch": 0.57, "learning_rate": 8.227344473663636e-05, "loss": 0.7493, "step": 10049 }, { "epoch": 0.57, "learning_rate": 8.225536410354873e-05, "loss": 0.7491, "step": 10050 }, { "epoch": 0.57, "learning_rate": 8.223728406938914e-05, "loss": 0.7031, "step": 10051 }, { "epoch": 0.57, "learning_rate": 8.221920463476778e-05, "loss": 0.7337, "step": 10052 }, { "epoch": 0.57, "learning_rate": 8.220112580029489e-05, "loss": 0.7223, "step": 10053 }, { "epoch": 0.57, "learning_rate": 8.218304756658072e-05, "loss": 0.7087, "step": 10054 }, { "epoch": 0.57, "learning_rate": 8.21649699342354e-05, "loss": 0.6942, "step": 10055 }, { "epoch": 0.57, "learning_rate": 8.21468929038691e-05, "loss": 0.7066, "step": 10056 }, { "epoch": 0.57, "learning_rate": 8.212881647609199e-05, "loss": 0.6894, "step": 10057 }, { "epoch": 0.57, "learning_rate": 8.211074065151421e-05, "loss": 0.7704, "step": 10058 }, { "epoch": 0.57, "learning_rate": 8.209266543074582e-05, "loss": 0.5559, "step": 10059 }, { "epoch": 0.57, "learning_rate": 8.207459081439695e-05, "loss": 0.7633, "step": 10060 }, { "epoch": 0.57, "learning_rate": 8.205651680307768e-05, "loss": 0.7218, "step": 10061 }, { "epoch": 0.57, "learning_rate": 8.203844339739801e-05, "loss": 0.6886, "step": 10062 }, { "epoch": 0.57, "learning_rate": 8.202037059796797e-05, "loss": 0.7537, "step": 10063 }, { "epoch": 0.57, "learning_rate": 8.200229840539757e-05, "loss": 0.7293, "step": 10064 }, { "epoch": 0.57, "learning_rate": 8.19842268202968e-05, "loss": 0.7267, "step": 10065 }, { "epoch": 0.57, "learning_rate": 8.196615584327563e-05, "loss": 0.7289, "step": 10066 }, { "epoch": 0.57, "learning_rate": 8.194808547494401e-05, "loss": 0.6947, "step": 10067 }, { "epoch": 0.57, "learning_rate": 8.193001571591182e-05, "loss": 0.672, "step": 10068 }, { "epoch": 0.57, "learning_rate": 8.191194656678904e-05, "loss": 0.7568, "step": 10069 }, { "epoch": 0.57, "learning_rate": 8.189387802818546e-05, "loss": 0.6919, "step": 10070 }, { "epoch": 0.57, "learning_rate": 8.1875810100711e-05, "loss": 0.6382, "step": 10071 }, { "epoch": 0.57, "learning_rate": 8.185774278497544e-05, "loss": 0.7478, "step": 10072 }, { "epoch": 0.57, "learning_rate": 8.183967608158866e-05, "loss": 0.7963, "step": 10073 }, { "epoch": 0.57, "learning_rate": 8.182160999116046e-05, "loss": 0.8092, "step": 10074 }, { "epoch": 0.57, "learning_rate": 8.18035445143006e-05, "loss": 0.7632, "step": 10075 }, { "epoch": 0.57, "learning_rate": 8.178547965161883e-05, "loss": 0.7654, "step": 10076 }, { "epoch": 0.57, "learning_rate": 8.176741540372488e-05, "loss": 0.6721, "step": 10077 }, { "epoch": 0.57, "learning_rate": 8.174935177122849e-05, "loss": 0.7501, "step": 10078 }, { "epoch": 0.57, "learning_rate": 8.173128875473932e-05, "loss": 0.8047, "step": 10079 }, { "epoch": 0.57, "learning_rate": 8.17132263548671e-05, "loss": 0.7138, "step": 10080 }, { "epoch": 0.57, "learning_rate": 8.16951645722214e-05, "loss": 0.6972, "step": 10081 }, { "epoch": 0.57, "learning_rate": 8.167710340741192e-05, "loss": 0.7072, "step": 10082 }, { "epoch": 0.57, "learning_rate": 8.16590428610483e-05, "loss": 0.7303, "step": 10083 }, { "epoch": 0.57, "learning_rate": 8.164098293374005e-05, "loss": 0.7312, "step": 10084 }, { "epoch": 0.57, "learning_rate": 8.162292362609676e-05, "loss": 0.7673, "step": 10085 }, { "epoch": 0.57, "learning_rate": 8.160486493872798e-05, "loss": 0.6193, "step": 10086 }, { "epoch": 0.57, "learning_rate": 8.15868068722433e-05, "loss": 0.7028, "step": 10087 }, { "epoch": 0.57, "learning_rate": 8.156874942725215e-05, "loss": 0.7787, "step": 10088 }, { "epoch": 0.57, "learning_rate": 8.155069260436403e-05, "loss": 0.7385, "step": 10089 }, { "epoch": 0.57, "learning_rate": 8.153263640418844e-05, "loss": 0.7109, "step": 10090 }, { "epoch": 0.57, "learning_rate": 8.151458082733484e-05, "loss": 0.7123, "step": 10091 }, { "epoch": 0.57, "learning_rate": 8.149652587441257e-05, "loss": 0.7227, "step": 10092 }, { "epoch": 0.57, "learning_rate": 8.147847154603107e-05, "loss": 0.7213, "step": 10093 }, { "epoch": 0.57, "learning_rate": 8.146041784279976e-05, "loss": 0.765, "step": 10094 }, { "epoch": 0.57, "learning_rate": 8.144236476532794e-05, "loss": 0.7336, "step": 10095 }, { "epoch": 0.57, "learning_rate": 8.1424312314225e-05, "loss": 0.7992, "step": 10096 }, { "epoch": 0.57, "learning_rate": 8.140626049010023e-05, "loss": 0.584, "step": 10097 }, { "epoch": 0.57, "learning_rate": 8.138820929356296e-05, "loss": 0.822, "step": 10098 }, { "epoch": 0.57, "learning_rate": 8.137015872522241e-05, "loss": 0.7923, "step": 10099 }, { "epoch": 0.57, "learning_rate": 8.135210878568786e-05, "loss": 0.7812, "step": 10100 }, { "epoch": 0.57, "learning_rate": 8.133405947556853e-05, "loss": 0.6603, "step": 10101 }, { "epoch": 0.57, "learning_rate": 8.131601079547366e-05, "loss": 0.7187, "step": 10102 }, { "epoch": 0.57, "learning_rate": 8.129796274601245e-05, "loss": 0.7818, "step": 10103 }, { "epoch": 0.57, "learning_rate": 8.127991532779401e-05, "loss": 0.6168, "step": 10104 }, { "epoch": 0.57, "learning_rate": 8.126186854142752e-05, "loss": 0.6267, "step": 10105 }, { "epoch": 0.57, "learning_rate": 8.124382238752216e-05, "loss": 0.7012, "step": 10106 }, { "epoch": 0.57, "learning_rate": 8.122577686668697e-05, "loss": 0.7314, "step": 10107 }, { "epoch": 0.57, "learning_rate": 8.120773197953101e-05, "loss": 0.7364, "step": 10108 }, { "epoch": 0.57, "learning_rate": 8.118968772666339e-05, "loss": 0.7838, "step": 10109 }, { "epoch": 0.57, "learning_rate": 8.117164410869315e-05, "loss": 0.7089, "step": 10110 }, { "epoch": 0.57, "learning_rate": 8.115360112622929e-05, "loss": 0.7882, "step": 10111 }, { "epoch": 0.57, "learning_rate": 8.113555877988083e-05, "loss": 0.6569, "step": 10112 }, { "epoch": 0.57, "learning_rate": 8.111751707025679e-05, "loss": 0.8204, "step": 10113 }, { "epoch": 0.57, "learning_rate": 8.1099475997966e-05, "loss": 0.7491, "step": 10114 }, { "epoch": 0.57, "learning_rate": 8.108143556361746e-05, "loss": 0.7379, "step": 10115 }, { "epoch": 0.57, "learning_rate": 8.106339576782012e-05, "loss": 0.7066, "step": 10116 }, { "epoch": 0.57, "learning_rate": 8.10453566111828e-05, "loss": 0.7273, "step": 10117 }, { "epoch": 0.57, "learning_rate": 8.10273180943144e-05, "loss": 0.6346, "step": 10118 }, { "epoch": 0.57, "learning_rate": 8.100928021782381e-05, "loss": 0.8155, "step": 10119 }, { "epoch": 0.57, "learning_rate": 8.099124298231984e-05, "loss": 0.7316, "step": 10120 }, { "epoch": 0.57, "learning_rate": 8.097320638841123e-05, "loss": 0.7449, "step": 10121 }, { "epoch": 0.57, "learning_rate": 8.09551704367068e-05, "loss": 0.6832, "step": 10122 }, { "epoch": 0.57, "learning_rate": 8.093713512781534e-05, "loss": 0.7243, "step": 10123 }, { "epoch": 0.57, "learning_rate": 8.091910046234552e-05, "loss": 0.7327, "step": 10124 }, { "epoch": 0.57, "learning_rate": 8.090106644090614e-05, "loss": 0.7127, "step": 10125 }, { "epoch": 0.57, "learning_rate": 8.088303306410583e-05, "loss": 0.6968, "step": 10126 }, { "epoch": 0.57, "learning_rate": 8.08650003325533e-05, "loss": 0.6523, "step": 10127 }, { "epoch": 0.57, "learning_rate": 8.084696824685722e-05, "loss": 0.6487, "step": 10128 }, { "epoch": 0.57, "learning_rate": 8.082893680762619e-05, "loss": 0.785, "step": 10129 }, { "epoch": 0.57, "learning_rate": 8.081090601546879e-05, "loss": 0.7219, "step": 10130 }, { "epoch": 0.57, "learning_rate": 8.079287587099363e-05, "loss": 0.7155, "step": 10131 }, { "epoch": 0.57, "learning_rate": 8.07748463748093e-05, "loss": 0.718, "step": 10132 }, { "epoch": 0.57, "learning_rate": 8.075681752752433e-05, "loss": 0.7056, "step": 10133 }, { "epoch": 0.57, "learning_rate": 8.07387893297472e-05, "loss": 0.8282, "step": 10134 }, { "epoch": 0.57, "learning_rate": 8.072076178208652e-05, "loss": 0.6948, "step": 10135 }, { "epoch": 0.57, "learning_rate": 8.070273488515066e-05, "loss": 0.8683, "step": 10136 }, { "epoch": 0.58, "learning_rate": 8.068470863954807e-05, "loss": 0.8747, "step": 10137 }, { "epoch": 0.58, "learning_rate": 8.066668304588723e-05, "loss": 0.7265, "step": 10138 }, { "epoch": 0.58, "learning_rate": 8.064865810477657e-05, "loss": 0.5796, "step": 10139 }, { "epoch": 0.58, "learning_rate": 8.063063381682442e-05, "loss": 0.659, "step": 10140 }, { "epoch": 0.58, "learning_rate": 8.061261018263919e-05, "loss": 0.735, "step": 10141 }, { "epoch": 0.58, "learning_rate": 8.05945872028292e-05, "loss": 0.6762, "step": 10142 }, { "epoch": 0.58, "learning_rate": 8.057656487800282e-05, "loss": 0.6843, "step": 10143 }, { "epoch": 0.58, "learning_rate": 8.055854320876829e-05, "loss": 0.7487, "step": 10144 }, { "epoch": 0.58, "learning_rate": 8.054052219573393e-05, "loss": 0.7149, "step": 10145 }, { "epoch": 0.58, "learning_rate": 8.052250183950795e-05, "loss": 0.6119, "step": 10146 }, { "epoch": 0.58, "learning_rate": 8.050448214069861e-05, "loss": 0.7141, "step": 10147 }, { "epoch": 0.58, "learning_rate": 8.048646309991416e-05, "loss": 0.7405, "step": 10148 }, { "epoch": 0.58, "learning_rate": 8.046844471776273e-05, "loss": 0.7445, "step": 10149 }, { "epoch": 0.58, "learning_rate": 8.045042699485255e-05, "loss": 0.643, "step": 10150 }, { "epoch": 0.58, "learning_rate": 8.04324099317917e-05, "loss": 0.7114, "step": 10151 }, { "epoch": 0.58, "learning_rate": 8.041439352918834e-05, "loss": 0.7203, "step": 10152 }, { "epoch": 0.58, "learning_rate": 8.039637778765056e-05, "loss": 0.6929, "step": 10153 }, { "epoch": 0.58, "learning_rate": 8.037836270778642e-05, "loss": 0.6669, "step": 10154 }, { "epoch": 0.58, "learning_rate": 8.036034829020405e-05, "loss": 0.7192, "step": 10155 }, { "epoch": 0.58, "learning_rate": 8.034233453551139e-05, "loss": 0.7352, "step": 10156 }, { "epoch": 0.58, "learning_rate": 8.03243214443165e-05, "loss": 0.7296, "step": 10157 }, { "epoch": 0.58, "learning_rate": 8.030630901722742e-05, "loss": 0.7513, "step": 10158 }, { "epoch": 0.58, "learning_rate": 8.028829725485199e-05, "loss": 0.6766, "step": 10159 }, { "epoch": 0.58, "learning_rate": 8.027028615779824e-05, "loss": 0.7646, "step": 10160 }, { "epoch": 0.58, "learning_rate": 8.025227572667408e-05, "loss": 0.7378, "step": 10161 }, { "epoch": 0.58, "learning_rate": 8.023426596208739e-05, "loss": 0.7115, "step": 10162 }, { "epoch": 0.58, "learning_rate": 8.021625686464607e-05, "loss": 0.6871, "step": 10163 }, { "epoch": 0.58, "learning_rate": 8.019824843495797e-05, "loss": 0.7516, "step": 10164 }, { "epoch": 0.58, "learning_rate": 8.018024067363095e-05, "loss": 0.6636, "step": 10165 }, { "epoch": 0.58, "learning_rate": 8.016223358127274e-05, "loss": 0.747, "step": 10166 }, { "epoch": 0.58, "learning_rate": 8.014422715849117e-05, "loss": 0.7402, "step": 10167 }, { "epoch": 0.58, "learning_rate": 8.012622140589404e-05, "loss": 0.6714, "step": 10168 }, { "epoch": 0.58, "learning_rate": 8.010821632408903e-05, "loss": 0.7033, "step": 10169 }, { "epoch": 0.58, "learning_rate": 8.009021191368391e-05, "loss": 0.7385, "step": 10170 }, { "epoch": 0.58, "learning_rate": 8.007220817528634e-05, "loss": 0.6968, "step": 10171 }, { "epoch": 0.58, "learning_rate": 8.005420510950404e-05, "loss": 0.724, "step": 10172 }, { "epoch": 0.58, "learning_rate": 8.003620271694458e-05, "loss": 0.72, "step": 10173 }, { "epoch": 0.58, "learning_rate": 8.001820099821565e-05, "loss": 0.6768, "step": 10174 }, { "epoch": 0.58, "learning_rate": 8.000019995392484e-05, "loss": 0.6857, "step": 10175 }, { "epoch": 0.58, "learning_rate": 7.998219958467972e-05, "loss": 0.6725, "step": 10176 }, { "epoch": 0.58, "learning_rate": 7.996419989108789e-05, "loss": 0.774, "step": 10177 }, { "epoch": 0.58, "learning_rate": 7.994620087375683e-05, "loss": 0.6341, "step": 10178 }, { "epoch": 0.58, "learning_rate": 7.992820253329409e-05, "loss": 0.7169, "step": 10179 }, { "epoch": 0.58, "learning_rate": 7.99102048703072e-05, "loss": 0.6818, "step": 10180 }, { "epoch": 0.58, "learning_rate": 7.989220788540355e-05, "loss": 0.6402, "step": 10181 }, { "epoch": 0.58, "learning_rate": 7.987421157919061e-05, "loss": 0.7702, "step": 10182 }, { "epoch": 0.58, "learning_rate": 7.985621595227581e-05, "loss": 0.6892, "step": 10183 }, { "epoch": 0.58, "learning_rate": 7.983822100526657e-05, "loss": 0.8049, "step": 10184 }, { "epoch": 0.58, "learning_rate": 7.982022673877022e-05, "loss": 0.7583, "step": 10185 }, { "epoch": 0.58, "learning_rate": 7.980223315339417e-05, "loss": 0.6718, "step": 10186 }, { "epoch": 0.58, "learning_rate": 7.978424024974572e-05, "loss": 0.792, "step": 10187 }, { "epoch": 0.58, "learning_rate": 7.976624802843218e-05, "loss": 0.6311, "step": 10188 }, { "epoch": 0.58, "learning_rate": 7.974825649006081e-05, "loss": 0.7535, "step": 10189 }, { "epoch": 0.58, "learning_rate": 7.973026563523891e-05, "loss": 0.6978, "step": 10190 }, { "epoch": 0.58, "learning_rate": 7.971227546457369e-05, "loss": 0.7904, "step": 10191 }, { "epoch": 0.58, "learning_rate": 7.969428597867238e-05, "loss": 0.6745, "step": 10192 }, { "epoch": 0.58, "learning_rate": 7.96762971781422e-05, "loss": 0.6422, "step": 10193 }, { "epoch": 0.58, "learning_rate": 7.965830906359026e-05, "loss": 0.6425, "step": 10194 }, { "epoch": 0.58, "learning_rate": 7.964032163562378e-05, "loss": 0.6412, "step": 10195 }, { "epoch": 0.58, "learning_rate": 7.96223348948498e-05, "loss": 0.7047, "step": 10196 }, { "epoch": 0.58, "learning_rate": 7.960434884187549e-05, "loss": 0.7536, "step": 10197 }, { "epoch": 0.58, "learning_rate": 7.958636347730786e-05, "loss": 0.784, "step": 10198 }, { "epoch": 0.58, "learning_rate": 7.956837880175401e-05, "loss": 0.7519, "step": 10199 }, { "epoch": 0.58, "learning_rate": 7.955039481582097e-05, "loss": 0.6912, "step": 10200 }, { "epoch": 0.58, "learning_rate": 7.953241152011572e-05, "loss": 0.7449, "step": 10201 }, { "epoch": 0.58, "learning_rate": 7.951442891524529e-05, "loss": 0.7429, "step": 10202 }, { "epoch": 0.58, "learning_rate": 7.949644700181662e-05, "loss": 0.6638, "step": 10203 }, { "epoch": 0.58, "learning_rate": 7.947846578043659e-05, "loss": 0.7827, "step": 10204 }, { "epoch": 0.58, "learning_rate": 7.946048525171217e-05, "loss": 0.7492, "step": 10205 }, { "epoch": 0.58, "learning_rate": 7.944250541625025e-05, "loss": 0.7353, "step": 10206 }, { "epoch": 0.58, "learning_rate": 7.942452627465767e-05, "loss": 0.6112, "step": 10207 }, { "epoch": 0.58, "learning_rate": 7.940654782754128e-05, "loss": 0.6711, "step": 10208 }, { "epoch": 0.58, "learning_rate": 7.938857007550798e-05, "loss": 0.7036, "step": 10209 }, { "epoch": 0.58, "learning_rate": 7.937059301916445e-05, "loss": 0.7115, "step": 10210 }, { "epoch": 0.58, "learning_rate": 7.93526166591175e-05, "loss": 0.69, "step": 10211 }, { "epoch": 0.58, "learning_rate": 7.933464099597388e-05, "loss": 0.6454, "step": 10212 }, { "epoch": 0.58, "learning_rate": 7.931666603034033e-05, "loss": 0.7559, "step": 10213 }, { "epoch": 0.58, "learning_rate": 7.929869176282355e-05, "loss": 0.7589, "step": 10214 }, { "epoch": 0.58, "learning_rate": 7.928071819403021e-05, "loss": 0.7129, "step": 10215 }, { "epoch": 0.58, "learning_rate": 7.926274532456698e-05, "loss": 0.6175, "step": 10216 }, { "epoch": 0.58, "learning_rate": 7.924477315504051e-05, "loss": 0.7381, "step": 10217 }, { "epoch": 0.58, "learning_rate": 7.922680168605733e-05, "loss": 0.6036, "step": 10218 }, { "epoch": 0.58, "learning_rate": 7.920883091822408e-05, "loss": 0.7552, "step": 10219 }, { "epoch": 0.58, "learning_rate": 7.919086085214732e-05, "loss": 0.6532, "step": 10220 }, { "epoch": 0.58, "learning_rate": 7.917289148843356e-05, "loss": 0.7215, "step": 10221 }, { "epoch": 0.58, "learning_rate": 7.915492282768937e-05, "loss": 0.8463, "step": 10222 }, { "epoch": 0.58, "learning_rate": 7.913695487052118e-05, "loss": 0.7084, "step": 10223 }, { "epoch": 0.58, "learning_rate": 7.91189876175355e-05, "loss": 0.6316, "step": 10224 }, { "epoch": 0.58, "learning_rate": 7.910102106933874e-05, "loss": 0.7289, "step": 10225 }, { "epoch": 0.58, "learning_rate": 7.908305522653733e-05, "loss": 0.5931, "step": 10226 }, { "epoch": 0.58, "learning_rate": 7.906509008973763e-05, "loss": 0.7117, "step": 10227 }, { "epoch": 0.58, "learning_rate": 7.904712565954607e-05, "loss": 0.7417, "step": 10228 }, { "epoch": 0.58, "learning_rate": 7.902916193656898e-05, "loss": 0.6431, "step": 10229 }, { "epoch": 0.58, "learning_rate": 7.901119892141266e-05, "loss": 0.6507, "step": 10230 }, { "epoch": 0.58, "learning_rate": 7.899323661468343e-05, "loss": 0.7143, "step": 10231 }, { "epoch": 0.58, "learning_rate": 7.897527501698756e-05, "loss": 0.6989, "step": 10232 }, { "epoch": 0.58, "learning_rate": 7.895731412893129e-05, "loss": 0.7223, "step": 10233 }, { "epoch": 0.58, "learning_rate": 7.893935395112084e-05, "loss": 0.7103, "step": 10234 }, { "epoch": 0.58, "learning_rate": 7.892139448416244e-05, "loss": 0.5808, "step": 10235 }, { "epoch": 0.58, "learning_rate": 7.890343572866226e-05, "loss": 0.7243, "step": 10236 }, { "epoch": 0.58, "learning_rate": 7.888547768522644e-05, "loss": 0.7289, "step": 10237 }, { "epoch": 0.58, "learning_rate": 7.886752035446114e-05, "loss": 0.7063, "step": 10238 }, { "epoch": 0.58, "learning_rate": 7.884956373697247e-05, "loss": 0.7095, "step": 10239 }, { "epoch": 0.58, "learning_rate": 7.883160783336645e-05, "loss": 0.6901, "step": 10240 }, { "epoch": 0.58, "learning_rate": 7.88136526442492e-05, "loss": 0.6913, "step": 10241 }, { "epoch": 0.58, "learning_rate": 7.879569817022673e-05, "loss": 0.6378, "step": 10242 }, { "epoch": 0.58, "learning_rate": 7.877774441190505e-05, "loss": 0.6666, "step": 10243 }, { "epoch": 0.58, "learning_rate": 7.875979136989016e-05, "loss": 0.7268, "step": 10244 }, { "epoch": 0.58, "learning_rate": 7.874183904478802e-05, "loss": 0.7112, "step": 10245 }, { "epoch": 0.58, "learning_rate": 7.87238874372046e-05, "loss": 0.6387, "step": 10246 }, { "epoch": 0.58, "learning_rate": 7.870593654774573e-05, "loss": 0.7203, "step": 10247 }, { "epoch": 0.58, "learning_rate": 7.868798637701736e-05, "loss": 0.6141, "step": 10248 }, { "epoch": 0.58, "learning_rate": 7.867003692562534e-05, "loss": 0.6977, "step": 10249 }, { "epoch": 0.58, "learning_rate": 7.865208819417549e-05, "loss": 0.6609, "step": 10250 }, { "epoch": 0.58, "learning_rate": 7.863414018327369e-05, "loss": 0.6862, "step": 10251 }, { "epoch": 0.58, "learning_rate": 7.861619289352566e-05, "loss": 0.7622, "step": 10252 }, { "epoch": 0.58, "learning_rate": 7.85982463255372e-05, "loss": 0.7332, "step": 10253 }, { "epoch": 0.58, "learning_rate": 7.858030047991411e-05, "loss": 0.7493, "step": 10254 }, { "epoch": 0.58, "learning_rate": 7.856235535726205e-05, "loss": 0.6804, "step": 10255 }, { "epoch": 0.58, "learning_rate": 7.854441095818666e-05, "loss": 0.7878, "step": 10256 }, { "epoch": 0.58, "learning_rate": 7.852646728329368e-05, "loss": 0.6483, "step": 10257 }, { "epoch": 0.58, "learning_rate": 7.850852433318877e-05, "loss": 0.8023, "step": 10258 }, { "epoch": 0.58, "learning_rate": 7.849058210847751e-05, "loss": 0.7274, "step": 10259 }, { "epoch": 0.58, "learning_rate": 7.847264060976551e-05, "loss": 0.5845, "step": 10260 }, { "epoch": 0.58, "learning_rate": 7.84546998376584e-05, "loss": 0.6934, "step": 10261 }, { "epoch": 0.58, "learning_rate": 7.843675979276164e-05, "loss": 0.7395, "step": 10262 }, { "epoch": 0.58, "learning_rate": 7.841882047568077e-05, "loss": 0.7774, "step": 10263 }, { "epoch": 0.58, "learning_rate": 7.840088188702133e-05, "loss": 0.8, "step": 10264 }, { "epoch": 0.58, "learning_rate": 7.838294402738875e-05, "loss": 0.6684, "step": 10265 }, { "epoch": 0.58, "learning_rate": 7.83650068973885e-05, "loss": 0.7912, "step": 10266 }, { "epoch": 0.58, "learning_rate": 7.834707049762603e-05, "loss": 0.615, "step": 10267 }, { "epoch": 0.58, "learning_rate": 7.832913482870671e-05, "loss": 0.6202, "step": 10268 }, { "epoch": 0.58, "learning_rate": 7.831119989123596e-05, "loss": 0.6715, "step": 10269 }, { "epoch": 0.58, "learning_rate": 7.829326568581904e-05, "loss": 0.6388, "step": 10270 }, { "epoch": 0.58, "learning_rate": 7.827533221306136e-05, "loss": 0.6461, "step": 10271 }, { "epoch": 0.58, "learning_rate": 7.825739947356816e-05, "loss": 0.6762, "step": 10272 }, { "epoch": 0.58, "learning_rate": 7.823946746794477e-05, "loss": 0.7834, "step": 10273 }, { "epoch": 0.58, "learning_rate": 7.822153619679643e-05, "loss": 0.7434, "step": 10274 }, { "epoch": 0.58, "learning_rate": 7.820360566072834e-05, "loss": 0.6196, "step": 10275 }, { "epoch": 0.58, "learning_rate": 7.818567586034577e-05, "loss": 0.6972, "step": 10276 }, { "epoch": 0.58, "learning_rate": 7.816774679625381e-05, "loss": 0.7101, "step": 10277 }, { "epoch": 0.58, "learning_rate": 7.814981846905765e-05, "loss": 0.792, "step": 10278 }, { "epoch": 0.58, "learning_rate": 7.813189087936243e-05, "loss": 0.7732, "step": 10279 }, { "epoch": 0.58, "learning_rate": 7.811396402777325e-05, "loss": 0.6509, "step": 10280 }, { "epoch": 0.58, "learning_rate": 7.809603791489516e-05, "loss": 0.6374, "step": 10281 }, { "epoch": 0.58, "learning_rate": 7.807811254133323e-05, "loss": 0.6353, "step": 10282 }, { "epoch": 0.58, "learning_rate": 7.806018790769256e-05, "loss": 0.7092, "step": 10283 }, { "epoch": 0.58, "learning_rate": 7.804226401457804e-05, "loss": 0.7585, "step": 10284 }, { "epoch": 0.58, "learning_rate": 7.80243408625947e-05, "loss": 0.7856, "step": 10285 }, { "epoch": 0.58, "learning_rate": 7.800641845234748e-05, "loss": 0.6873, "step": 10286 }, { "epoch": 0.58, "learning_rate": 7.798849678444133e-05, "loss": 0.741, "step": 10287 }, { "epoch": 0.58, "learning_rate": 7.797057585948113e-05, "loss": 0.7429, "step": 10288 }, { "epoch": 0.58, "learning_rate": 7.795265567807175e-05, "loss": 0.6769, "step": 10289 }, { "epoch": 0.58, "learning_rate": 7.79347362408181e-05, "loss": 0.6389, "step": 10290 }, { "epoch": 0.58, "learning_rate": 7.791681754832498e-05, "loss": 0.7316, "step": 10291 }, { "epoch": 0.58, "learning_rate": 7.789889960119716e-05, "loss": 0.7001, "step": 10292 }, { "epoch": 0.58, "learning_rate": 7.788098240003943e-05, "loss": 0.756, "step": 10293 }, { "epoch": 0.58, "learning_rate": 7.786306594545657e-05, "loss": 0.7658, "step": 10294 }, { "epoch": 0.58, "learning_rate": 7.784515023805328e-05, "loss": 0.7303, "step": 10295 }, { "epoch": 0.58, "learning_rate": 7.782723527843428e-05, "loss": 0.7361, "step": 10296 }, { "epoch": 0.58, "learning_rate": 7.780932106720422e-05, "loss": 0.6887, "step": 10297 }, { "epoch": 0.58, "learning_rate": 7.779140760496783e-05, "loss": 0.6807, "step": 10298 }, { "epoch": 0.58, "learning_rate": 7.77734948923296e-05, "loss": 0.699, "step": 10299 }, { "epoch": 0.58, "learning_rate": 7.775558292989427e-05, "loss": 0.7746, "step": 10300 }, { "epoch": 0.58, "learning_rate": 7.773767171826631e-05, "loss": 0.7426, "step": 10301 }, { "epoch": 0.58, "learning_rate": 7.771976125805032e-05, "loss": 0.7353, "step": 10302 }, { "epoch": 0.58, "learning_rate": 7.770185154985085e-05, "loss": 0.7454, "step": 10303 }, { "epoch": 0.58, "learning_rate": 7.768394259427234e-05, "loss": 0.7497, "step": 10304 }, { "epoch": 0.58, "learning_rate": 7.76660343919193e-05, "loss": 0.78, "step": 10305 }, { "epoch": 0.58, "learning_rate": 7.764812694339621e-05, "loss": 0.7897, "step": 10306 }, { "epoch": 0.58, "learning_rate": 7.763022024930741e-05, "loss": 0.7092, "step": 10307 }, { "epoch": 0.58, "learning_rate": 7.761231431025736e-05, "loss": 0.7017, "step": 10308 }, { "epoch": 0.58, "learning_rate": 7.759440912685042e-05, "loss": 0.769, "step": 10309 }, { "epoch": 0.58, "learning_rate": 7.757650469969091e-05, "loss": 0.663, "step": 10310 }, { "epoch": 0.58, "learning_rate": 7.755860102938318e-05, "loss": 0.6737, "step": 10311 }, { "epoch": 0.58, "learning_rate": 7.754069811653154e-05, "loss": 0.6554, "step": 10312 }, { "epoch": 0.59, "learning_rate": 7.752279596174025e-05, "loss": 0.7075, "step": 10313 }, { "epoch": 0.59, "learning_rate": 7.750489456561352e-05, "loss": 0.5931, "step": 10314 }, { "epoch": 0.59, "learning_rate": 7.74869939287556e-05, "loss": 0.67, "step": 10315 }, { "epoch": 0.59, "learning_rate": 7.746909405177067e-05, "loss": 0.6855, "step": 10316 }, { "epoch": 0.59, "learning_rate": 7.74511949352629e-05, "loss": 0.6343, "step": 10317 }, { "epoch": 0.59, "learning_rate": 7.743329657983644e-05, "loss": 0.7006, "step": 10318 }, { "epoch": 0.59, "learning_rate": 7.741539898609543e-05, "loss": 0.7648, "step": 10319 }, { "epoch": 0.59, "learning_rate": 7.73975021546439e-05, "loss": 0.7788, "step": 10320 }, { "epoch": 0.59, "learning_rate": 7.7379606086086e-05, "loss": 0.8137, "step": 10321 }, { "epoch": 0.59, "learning_rate": 7.736171078102566e-05, "loss": 0.7691, "step": 10322 }, { "epoch": 0.59, "learning_rate": 7.734381624006699e-05, "loss": 0.7665, "step": 10323 }, { "epoch": 0.59, "learning_rate": 7.732592246381392e-05, "loss": 0.8106, "step": 10324 }, { "epoch": 0.59, "learning_rate": 7.730802945287044e-05, "loss": 0.7071, "step": 10325 }, { "epoch": 0.59, "learning_rate": 7.729013720784047e-05, "loss": 0.7048, "step": 10326 }, { "epoch": 0.59, "learning_rate": 7.727224572932792e-05, "loss": 0.7589, "step": 10327 }, { "epoch": 0.59, "learning_rate": 7.725435501793674e-05, "loss": 0.7806, "step": 10328 }, { "epoch": 0.59, "learning_rate": 7.72364650742707e-05, "loss": 0.7516, "step": 10329 }, { "epoch": 0.59, "learning_rate": 7.721857589893364e-05, "loss": 0.7371, "step": 10330 }, { "epoch": 0.59, "learning_rate": 7.720068749252938e-05, "loss": 0.7031, "step": 10331 }, { "epoch": 0.59, "learning_rate": 7.718279985566175e-05, "loss": 0.7191, "step": 10332 }, { "epoch": 0.59, "learning_rate": 7.716491298893442e-05, "loss": 0.7808, "step": 10333 }, { "epoch": 0.59, "learning_rate": 7.714702689295119e-05, "loss": 0.6635, "step": 10334 }, { "epoch": 0.59, "learning_rate": 7.712914156831577e-05, "loss": 0.7487, "step": 10335 }, { "epoch": 0.59, "learning_rate": 7.711125701563177e-05, "loss": 0.8088, "step": 10336 }, { "epoch": 0.59, "learning_rate": 7.709337323550286e-05, "loss": 0.7859, "step": 10337 }, { "epoch": 0.59, "learning_rate": 7.707549022853268e-05, "loss": 0.6899, "step": 10338 }, { "epoch": 0.59, "learning_rate": 7.705760799532485e-05, "loss": 0.6292, "step": 10339 }, { "epoch": 0.59, "learning_rate": 7.70397265364829e-05, "loss": 0.7959, "step": 10340 }, { "epoch": 0.59, "learning_rate": 7.702184585261041e-05, "loss": 0.7178, "step": 10341 }, { "epoch": 0.59, "learning_rate": 7.700396594431086e-05, "loss": 0.7342, "step": 10342 }, { "epoch": 0.59, "learning_rate": 7.698608681218782e-05, "loss": 0.6735, "step": 10343 }, { "epoch": 0.59, "learning_rate": 7.696820845684465e-05, "loss": 0.6562, "step": 10344 }, { "epoch": 0.59, "learning_rate": 7.695033087888489e-05, "loss": 0.7199, "step": 10345 }, { "epoch": 0.59, "learning_rate": 7.693245407891188e-05, "loss": 0.6622, "step": 10346 }, { "epoch": 0.59, "learning_rate": 7.691457805752904e-05, "loss": 0.796, "step": 10347 }, { "epoch": 0.59, "learning_rate": 7.689670281533976e-05, "loss": 0.743, "step": 10348 }, { "epoch": 0.59, "learning_rate": 7.687882835294733e-05, "loss": 0.662, "step": 10349 }, { "epoch": 0.59, "learning_rate": 7.686095467095512e-05, "loss": 0.7635, "step": 10350 }, { "epoch": 0.59, "learning_rate": 7.684308176996632e-05, "loss": 0.7688, "step": 10351 }, { "epoch": 0.59, "learning_rate": 7.682520965058428e-05, "loss": 0.7259, "step": 10352 }, { "epoch": 0.59, "learning_rate": 7.680733831341218e-05, "loss": 0.6691, "step": 10353 }, { "epoch": 0.59, "learning_rate": 7.678946775905324e-05, "loss": 0.6548, "step": 10354 }, { "epoch": 0.59, "learning_rate": 7.677159798811062e-05, "loss": 0.671, "step": 10355 }, { "epoch": 0.59, "learning_rate": 7.675372900118748e-05, "loss": 0.6973, "step": 10356 }, { "epoch": 0.59, "learning_rate": 7.673586079888698e-05, "loss": 0.6515, "step": 10357 }, { "epoch": 0.59, "learning_rate": 7.67179933818122e-05, "loss": 0.7157, "step": 10358 }, { "epoch": 0.59, "learning_rate": 7.670012675056619e-05, "loss": 0.7098, "step": 10359 }, { "epoch": 0.59, "learning_rate": 7.668226090575198e-05, "loss": 0.579, "step": 10360 }, { "epoch": 0.59, "learning_rate": 7.666439584797264e-05, "loss": 0.6875, "step": 10361 }, { "epoch": 0.59, "learning_rate": 7.664653157783112e-05, "loss": 0.6611, "step": 10362 }, { "epoch": 0.59, "learning_rate": 7.662866809593042e-05, "loss": 0.6937, "step": 10363 }, { "epoch": 0.59, "learning_rate": 7.661080540287348e-05, "loss": 0.669, "step": 10364 }, { "epoch": 0.59, "learning_rate": 7.659294349926322e-05, "loss": 0.771, "step": 10365 }, { "epoch": 0.59, "learning_rate": 7.657508238570246e-05, "loss": 0.6669, "step": 10366 }, { "epoch": 0.59, "learning_rate": 7.65572220627941e-05, "loss": 0.7905, "step": 10367 }, { "epoch": 0.59, "learning_rate": 7.6539362531141e-05, "loss": 0.6376, "step": 10368 }, { "epoch": 0.59, "learning_rate": 7.652150379134592e-05, "loss": 0.5949, "step": 10369 }, { "epoch": 0.59, "learning_rate": 7.650364584401169e-05, "loss": 0.6948, "step": 10370 }, { "epoch": 0.59, "learning_rate": 7.6485788689741e-05, "loss": 0.7131, "step": 10371 }, { "epoch": 0.59, "learning_rate": 7.646793232913666e-05, "loss": 0.6313, "step": 10372 }, { "epoch": 0.59, "learning_rate": 7.645007676280128e-05, "loss": 0.697, "step": 10373 }, { "epoch": 0.59, "learning_rate": 7.643222199133759e-05, "loss": 0.8518, "step": 10374 }, { "epoch": 0.59, "learning_rate": 7.641436801534818e-05, "loss": 0.6491, "step": 10375 }, { "epoch": 0.59, "learning_rate": 7.639651483543572e-05, "loss": 0.7235, "step": 10376 }, { "epoch": 0.59, "learning_rate": 7.637866245220281e-05, "loss": 0.671, "step": 10377 }, { "epoch": 0.59, "learning_rate": 7.636081086625196e-05, "loss": 0.7638, "step": 10378 }, { "epoch": 0.59, "learning_rate": 7.634296007818576e-05, "loss": 0.7901, "step": 10379 }, { "epoch": 0.59, "learning_rate": 7.632511008860673e-05, "loss": 0.6711, "step": 10380 }, { "epoch": 0.59, "learning_rate": 7.630726089811729e-05, "loss": 0.7431, "step": 10381 }, { "epoch": 0.59, "learning_rate": 7.628941250731993e-05, "loss": 0.6852, "step": 10382 }, { "epoch": 0.59, "learning_rate": 7.627156491681708e-05, "loss": 0.6954, "step": 10383 }, { "epoch": 0.59, "learning_rate": 7.625371812721114e-05, "loss": 0.6334, "step": 10384 }, { "epoch": 0.59, "learning_rate": 7.62358721391045e-05, "loss": 0.705, "step": 10385 }, { "epoch": 0.59, "learning_rate": 7.62180269530995e-05, "loss": 0.6795, "step": 10386 }, { "epoch": 0.59, "learning_rate": 7.62001825697985e-05, "loss": 0.7468, "step": 10387 }, { "epoch": 0.59, "learning_rate": 7.618233898980369e-05, "loss": 0.726, "step": 10388 }, { "epoch": 0.59, "learning_rate": 7.616449621371741e-05, "loss": 0.7316, "step": 10389 }, { "epoch": 0.59, "learning_rate": 7.614665424214193e-05, "loss": 0.7671, "step": 10390 }, { "epoch": 0.59, "learning_rate": 7.612881307567939e-05, "loss": 0.7356, "step": 10391 }, { "epoch": 0.59, "learning_rate": 7.611097271493203e-05, "loss": 0.6139, "step": 10392 }, { "epoch": 0.59, "learning_rate": 7.6093133160502e-05, "loss": 0.6569, "step": 10393 }, { "epoch": 0.59, "learning_rate": 7.60752944129914e-05, "loss": 0.7605, "step": 10394 }, { "epoch": 0.59, "learning_rate": 7.605745647300242e-05, "loss": 0.7858, "step": 10395 }, { "epoch": 0.59, "learning_rate": 7.603961934113702e-05, "loss": 0.6822, "step": 10396 }, { "epoch": 0.59, "learning_rate": 7.602178301799732e-05, "loss": 0.7825, "step": 10397 }, { "epoch": 0.59, "learning_rate": 7.600394750418531e-05, "loss": 0.7101, "step": 10398 }, { "epoch": 0.59, "learning_rate": 7.598611280030303e-05, "loss": 0.7443, "step": 10399 }, { "epoch": 0.59, "learning_rate": 7.596827890695241e-05, "loss": 0.6701, "step": 10400 }, { "epoch": 0.59, "learning_rate": 7.595044582473538e-05, "loss": 0.6722, "step": 10401 }, { "epoch": 0.59, "learning_rate": 7.593261355425394e-05, "loss": 0.7458, "step": 10402 }, { "epoch": 0.59, "learning_rate": 7.591478209610987e-05, "loss": 0.6799, "step": 10403 }, { "epoch": 0.59, "learning_rate": 7.589695145090506e-05, "loss": 0.7331, "step": 10404 }, { "epoch": 0.59, "learning_rate": 7.587912161924135e-05, "loss": 0.5721, "step": 10405 }, { "epoch": 0.59, "learning_rate": 7.586129260172057e-05, "loss": 0.6579, "step": 10406 }, { "epoch": 0.59, "learning_rate": 7.584346439894444e-05, "loss": 0.701, "step": 10407 }, { "epoch": 0.59, "learning_rate": 7.582563701151474e-05, "loss": 0.6775, "step": 10408 }, { "epoch": 0.59, "learning_rate": 7.580781044003324e-05, "loss": 0.6539, "step": 10409 }, { "epoch": 0.59, "learning_rate": 7.578998468510157e-05, "loss": 0.8111, "step": 10410 }, { "epoch": 0.59, "learning_rate": 7.577215974732137e-05, "loss": 0.6962, "step": 10411 }, { "epoch": 0.59, "learning_rate": 7.575433562729434e-05, "loss": 0.669, "step": 10412 }, { "epoch": 0.59, "learning_rate": 7.573651232562207e-05, "loss": 0.7348, "step": 10413 }, { "epoch": 0.59, "learning_rate": 7.571868984290613e-05, "loss": 0.701, "step": 10414 }, { "epoch": 0.59, "learning_rate": 7.570086817974811e-05, "loss": 0.7359, "step": 10415 }, { "epoch": 0.59, "learning_rate": 7.56830473367495e-05, "loss": 0.6321, "step": 10416 }, { "epoch": 0.59, "learning_rate": 7.566522731451185e-05, "loss": 0.8057, "step": 10417 }, { "epoch": 0.59, "learning_rate": 7.564740811363656e-05, "loss": 0.7207, "step": 10418 }, { "epoch": 0.59, "learning_rate": 7.562958973472513e-05, "loss": 0.7277, "step": 10419 }, { "epoch": 0.59, "learning_rate": 7.561177217837894e-05, "loss": 0.7712, "step": 10420 }, { "epoch": 0.59, "learning_rate": 7.559395544519939e-05, "loss": 0.6436, "step": 10421 }, { "epoch": 0.59, "learning_rate": 7.557613953578788e-05, "loss": 0.6918, "step": 10422 }, { "epoch": 0.59, "learning_rate": 7.55583244507457e-05, "loss": 0.7171, "step": 10423 }, { "epoch": 0.59, "learning_rate": 7.554051019067421e-05, "loss": 0.8001, "step": 10424 }, { "epoch": 0.59, "learning_rate": 7.55226967561746e-05, "loss": 0.6907, "step": 10425 }, { "epoch": 0.59, "learning_rate": 7.55048841478482e-05, "loss": 0.6002, "step": 10426 }, { "epoch": 0.59, "learning_rate": 7.548707236629617e-05, "loss": 0.7268, "step": 10427 }, { "epoch": 0.59, "learning_rate": 7.546926141211974e-05, "loss": 0.8062, "step": 10428 }, { "epoch": 0.59, "learning_rate": 7.54514512859201e-05, "loss": 0.7209, "step": 10429 }, { "epoch": 0.59, "learning_rate": 7.543364198829832e-05, "loss": 0.7451, "step": 10430 }, { "epoch": 0.59, "learning_rate": 7.541583351985559e-05, "loss": 0.6854, "step": 10431 }, { "epoch": 0.59, "learning_rate": 7.539802588119295e-05, "loss": 0.6832, "step": 10432 }, { "epoch": 0.59, "learning_rate": 7.538021907291144e-05, "loss": 0.7431, "step": 10433 }, { "epoch": 0.59, "learning_rate": 7.53624130956121e-05, "loss": 0.6928, "step": 10434 }, { "epoch": 0.59, "learning_rate": 7.534460794989595e-05, "loss": 0.7467, "step": 10435 }, { "epoch": 0.59, "learning_rate": 7.532680363636392e-05, "loss": 0.7188, "step": 10436 }, { "epoch": 0.59, "learning_rate": 7.530900015561697e-05, "loss": 0.697, "step": 10437 }, { "epoch": 0.59, "learning_rate": 7.529119750825606e-05, "loss": 0.62, "step": 10438 }, { "epoch": 0.59, "learning_rate": 7.527339569488203e-05, "loss": 0.6819, "step": 10439 }, { "epoch": 0.59, "learning_rate": 7.525559471609572e-05, "loss": 0.6378, "step": 10440 }, { "epoch": 0.59, "learning_rate": 7.523779457249798e-05, "loss": 0.6358, "step": 10441 }, { "epoch": 0.59, "learning_rate": 7.521999526468962e-05, "loss": 0.5961, "step": 10442 }, { "epoch": 0.59, "learning_rate": 7.52021967932714e-05, "loss": 0.6894, "step": 10443 }, { "epoch": 0.59, "learning_rate": 7.518439915884407e-05, "loss": 0.7183, "step": 10444 }, { "epoch": 0.59, "learning_rate": 7.516660236200838e-05, "loss": 0.6268, "step": 10445 }, { "epoch": 0.59, "learning_rate": 7.5148806403365e-05, "loss": 0.7963, "step": 10446 }, { "epoch": 0.59, "learning_rate": 7.513101128351454e-05, "loss": 0.7886, "step": 10447 }, { "epoch": 0.59, "learning_rate": 7.511321700305769e-05, "loss": 0.7014, "step": 10448 }, { "epoch": 0.59, "learning_rate": 7.509542356259502e-05, "loss": 0.7083, "step": 10449 }, { "epoch": 0.59, "learning_rate": 7.507763096272711e-05, "loss": 0.7034, "step": 10450 }, { "epoch": 0.59, "learning_rate": 7.505983920405454e-05, "loss": 0.6874, "step": 10451 }, { "epoch": 0.59, "learning_rate": 7.504204828717778e-05, "loss": 0.7267, "step": 10452 }, { "epoch": 0.59, "learning_rate": 7.502425821269734e-05, "loss": 0.7566, "step": 10453 }, { "epoch": 0.59, "learning_rate": 7.500646898121373e-05, "loss": 0.7421, "step": 10454 }, { "epoch": 0.59, "learning_rate": 7.498868059332734e-05, "loss": 0.6798, "step": 10455 }, { "epoch": 0.59, "learning_rate": 7.497089304963853e-05, "loss": 0.7679, "step": 10456 }, { "epoch": 0.59, "learning_rate": 7.495310635074773e-05, "loss": 0.7972, "step": 10457 }, { "epoch": 0.59, "learning_rate": 7.493532049725531e-05, "loss": 0.5847, "step": 10458 }, { "epoch": 0.59, "learning_rate": 7.491753548976153e-05, "loss": 0.7177, "step": 10459 }, { "epoch": 0.59, "learning_rate": 7.489975132886672e-05, "loss": 0.7333, "step": 10460 }, { "epoch": 0.59, "learning_rate": 7.488196801517113e-05, "loss": 0.7045, "step": 10461 }, { "epoch": 0.59, "learning_rate": 7.486418554927501e-05, "loss": 0.7106, "step": 10462 }, { "epoch": 0.59, "learning_rate": 7.484640393177851e-05, "loss": 0.6913, "step": 10463 }, { "epoch": 0.59, "learning_rate": 7.482862316328188e-05, "loss": 0.5863, "step": 10464 }, { "epoch": 0.59, "learning_rate": 7.48108432443852e-05, "loss": 0.6577, "step": 10465 }, { "epoch": 0.59, "learning_rate": 7.479306417568864e-05, "loss": 0.718, "step": 10466 }, { "epoch": 0.59, "learning_rate": 7.477528595779228e-05, "loss": 0.7155, "step": 10467 }, { "epoch": 0.59, "learning_rate": 7.475750859129614e-05, "loss": 0.653, "step": 10468 }, { "epoch": 0.59, "learning_rate": 7.473973207680033e-05, "loss": 0.6984, "step": 10469 }, { "epoch": 0.59, "learning_rate": 7.47219564149048e-05, "loss": 0.6697, "step": 10470 }, { "epoch": 0.59, "learning_rate": 7.470418160620953e-05, "loss": 0.7558, "step": 10471 }, { "epoch": 0.59, "learning_rate": 7.468640765131445e-05, "loss": 0.7292, "step": 10472 }, { "epoch": 0.59, "learning_rate": 7.46686345508195e-05, "loss": 0.5858, "step": 10473 }, { "epoch": 0.59, "learning_rate": 7.46508623053246e-05, "loss": 0.5944, "step": 10474 }, { "epoch": 0.59, "learning_rate": 7.463309091542955e-05, "loss": 0.7299, "step": 10475 }, { "epoch": 0.59, "learning_rate": 7.461532038173425e-05, "loss": 0.7071, "step": 10476 }, { "epoch": 0.59, "learning_rate": 7.459755070483841e-05, "loss": 0.7666, "step": 10477 }, { "epoch": 0.59, "learning_rate": 7.457978188534189e-05, "loss": 0.8311, "step": 10478 }, { "epoch": 0.59, "learning_rate": 7.456201392384436e-05, "loss": 0.6349, "step": 10479 }, { "epoch": 0.59, "learning_rate": 7.454424682094562e-05, "loss": 0.7083, "step": 10480 }, { "epoch": 0.59, "learning_rate": 7.452648057724527e-05, "loss": 0.7115, "step": 10481 }, { "epoch": 0.59, "learning_rate": 7.450871519334301e-05, "loss": 0.7497, "step": 10482 }, { "epoch": 0.59, "learning_rate": 7.449095066983849e-05, "loss": 0.5707, "step": 10483 }, { "epoch": 0.59, "learning_rate": 7.44731870073313e-05, "loss": 0.6315, "step": 10484 }, { "epoch": 0.59, "learning_rate": 7.445542420642097e-05, "loss": 0.6808, "step": 10485 }, { "epoch": 0.59, "learning_rate": 7.443766226770706e-05, "loss": 0.6968, "step": 10486 }, { "epoch": 0.59, "learning_rate": 7.441990119178909e-05, "loss": 0.7121, "step": 10487 }, { "epoch": 0.59, "learning_rate": 7.440214097926653e-05, "loss": 0.7766, "step": 10488 }, { "epoch": 0.59, "learning_rate": 7.438438163073884e-05, "loss": 0.6483, "step": 10489 }, { "epoch": 0.6, "learning_rate": 7.436662314680549e-05, "loss": 0.7359, "step": 10490 }, { "epoch": 0.6, "learning_rate": 7.434886552806583e-05, "loss": 0.827, "step": 10491 }, { "epoch": 0.6, "learning_rate": 7.433110877511922e-05, "loss": 0.788, "step": 10492 }, { "epoch": 0.6, "learning_rate": 7.4313352888565e-05, "loss": 0.6842, "step": 10493 }, { "epoch": 0.6, "learning_rate": 7.429559786900248e-05, "loss": 0.6189, "step": 10494 }, { "epoch": 0.6, "learning_rate": 7.427784371703095e-05, "loss": 0.6985, "step": 10495 }, { "epoch": 0.6, "learning_rate": 7.426009043324966e-05, "loss": 0.6525, "step": 10496 }, { "epoch": 0.6, "learning_rate": 7.42423380182578e-05, "loss": 0.8068, "step": 10497 }, { "epoch": 0.6, "learning_rate": 7.422458647265465e-05, "loss": 0.7354, "step": 10498 }, { "epoch": 0.6, "learning_rate": 7.420683579703924e-05, "loss": 0.6912, "step": 10499 }, { "epoch": 0.6, "learning_rate": 7.418908599201078e-05, "loss": 0.7843, "step": 10500 }, { "epoch": 0.6, "learning_rate": 7.417133705816837e-05, "loss": 0.6942, "step": 10501 }, { "epoch": 0.6, "learning_rate": 7.415358899611105e-05, "loss": 0.7292, "step": 10502 }, { "epoch": 0.6, "learning_rate": 7.41358418064379e-05, "loss": 0.5939, "step": 10503 }, { "epoch": 0.6, "learning_rate": 7.411809548974792e-05, "loss": 0.676, "step": 10504 }, { "epoch": 0.6, "learning_rate": 7.410035004664011e-05, "loss": 0.7479, "step": 10505 }, { "epoch": 0.6, "learning_rate": 7.40826054777134e-05, "loss": 0.7029, "step": 10506 }, { "epoch": 0.6, "learning_rate": 7.406486178356674e-05, "loss": 0.707, "step": 10507 }, { "epoch": 0.6, "learning_rate": 7.4047118964799e-05, "loss": 0.7574, "step": 10508 }, { "epoch": 0.6, "learning_rate": 7.402937702200905e-05, "loss": 0.6937, "step": 10509 }, { "epoch": 0.6, "learning_rate": 7.401163595579572e-05, "loss": 0.7044, "step": 10510 }, { "epoch": 0.6, "learning_rate": 7.399389576675784e-05, "loss": 0.7082, "step": 10511 }, { "epoch": 0.6, "learning_rate": 7.397615645549421e-05, "loss": 0.6914, "step": 10512 }, { "epoch": 0.6, "learning_rate": 7.395841802260356e-05, "loss": 0.6923, "step": 10513 }, { "epoch": 0.6, "learning_rate": 7.394068046868457e-05, "loss": 0.6338, "step": 10514 }, { "epoch": 0.6, "learning_rate": 7.392294379433595e-05, "loss": 0.6296, "step": 10515 }, { "epoch": 0.6, "learning_rate": 7.390520800015638e-05, "loss": 0.6192, "step": 10516 }, { "epoch": 0.6, "learning_rate": 7.388747308674447e-05, "loss": 0.8628, "step": 10517 }, { "epoch": 0.6, "learning_rate": 7.386973905469882e-05, "loss": 0.8018, "step": 10518 }, { "epoch": 0.6, "learning_rate": 7.385200590461803e-05, "loss": 0.6787, "step": 10519 }, { "epoch": 0.6, "learning_rate": 7.383427363710058e-05, "loss": 0.7242, "step": 10520 }, { "epoch": 0.6, "learning_rate": 7.381654225274508e-05, "loss": 0.8086, "step": 10521 }, { "epoch": 0.6, "learning_rate": 7.379881175214989e-05, "loss": 0.7573, "step": 10522 }, { "epoch": 0.6, "learning_rate": 7.378108213591355e-05, "loss": 0.7106, "step": 10523 }, { "epoch": 0.6, "learning_rate": 7.376335340463442e-05, "loss": 0.7521, "step": 10524 }, { "epoch": 0.6, "learning_rate": 7.374562555891094e-05, "loss": 0.7682, "step": 10525 }, { "epoch": 0.6, "learning_rate": 7.372789859934144e-05, "loss": 0.6854, "step": 10526 }, { "epoch": 0.6, "learning_rate": 7.371017252652426e-05, "loss": 0.7166, "step": 10527 }, { "epoch": 0.6, "learning_rate": 7.369244734105774e-05, "loss": 0.7081, "step": 10528 }, { "epoch": 0.6, "learning_rate": 7.36747230435401e-05, "loss": 0.7136, "step": 10529 }, { "epoch": 0.6, "learning_rate": 7.365699963456958e-05, "loss": 0.761, "step": 10530 }, { "epoch": 0.6, "learning_rate": 7.36392771147444e-05, "loss": 0.7377, "step": 10531 }, { "epoch": 0.6, "learning_rate": 7.362155548466279e-05, "loss": 0.7299, "step": 10532 }, { "epoch": 0.6, "learning_rate": 7.360383474492282e-05, "loss": 0.7487, "step": 10533 }, { "epoch": 0.6, "learning_rate": 7.358611489612268e-05, "loss": 0.5921, "step": 10534 }, { "epoch": 0.6, "learning_rate": 7.356839593886046e-05, "loss": 0.6941, "step": 10535 }, { "epoch": 0.6, "learning_rate": 7.355067787373418e-05, "loss": 0.6584, "step": 10536 }, { "epoch": 0.6, "learning_rate": 7.353296070134186e-05, "loss": 0.6654, "step": 10537 }, { "epoch": 0.6, "learning_rate": 7.351524442228154e-05, "loss": 0.6976, "step": 10538 }, { "epoch": 0.6, "learning_rate": 7.34975290371512e-05, "loss": 0.6593, "step": 10539 }, { "epoch": 0.6, "learning_rate": 7.34798145465487e-05, "loss": 0.6423, "step": 10540 }, { "epoch": 0.6, "learning_rate": 7.346210095107208e-05, "loss": 0.7099, "step": 10541 }, { "epoch": 0.6, "learning_rate": 7.344438825131911e-05, "loss": 0.7274, "step": 10542 }, { "epoch": 0.6, "learning_rate": 7.342667644788773e-05, "loss": 0.7302, "step": 10543 }, { "epoch": 0.6, "learning_rate": 7.340896554137567e-05, "loss": 0.6509, "step": 10544 }, { "epoch": 0.6, "learning_rate": 7.339125553238078e-05, "loss": 0.6482, "step": 10545 }, { "epoch": 0.6, "learning_rate": 7.337354642150077e-05, "loss": 0.6894, "step": 10546 }, { "epoch": 0.6, "learning_rate": 7.335583820933341e-05, "loss": 0.736, "step": 10547 }, { "epoch": 0.6, "learning_rate": 7.33381308964764e-05, "loss": 0.8133, "step": 10548 }, { "epoch": 0.6, "learning_rate": 7.332042448352738e-05, "loss": 0.6574, "step": 10549 }, { "epoch": 0.6, "learning_rate": 7.330271897108405e-05, "loss": 0.6037, "step": 10550 }, { "epoch": 0.6, "learning_rate": 7.328501435974392e-05, "loss": 0.6014, "step": 10551 }, { "epoch": 0.6, "learning_rate": 7.326731065010464e-05, "loss": 0.794, "step": 10552 }, { "epoch": 0.6, "learning_rate": 7.324960784276371e-05, "loss": 0.7752, "step": 10553 }, { "epoch": 0.6, "learning_rate": 7.32319059383187e-05, "loss": 0.6837, "step": 10554 }, { "epoch": 0.6, "learning_rate": 7.321420493736705e-05, "loss": 0.7154, "step": 10555 }, { "epoch": 0.6, "learning_rate": 7.319650484050621e-05, "loss": 0.685, "step": 10556 }, { "epoch": 0.6, "learning_rate": 7.317880564833367e-05, "loss": 0.7734, "step": 10557 }, { "epoch": 0.6, "learning_rate": 7.316110736144679e-05, "loss": 0.6578, "step": 10558 }, { "epoch": 0.6, "learning_rate": 7.314340998044289e-05, "loss": 0.7016, "step": 10559 }, { "epoch": 0.6, "learning_rate": 7.312571350591932e-05, "loss": 0.7333, "step": 10560 }, { "epoch": 0.6, "learning_rate": 7.310801793847344e-05, "loss": 0.6937, "step": 10561 }, { "epoch": 0.6, "learning_rate": 7.309032327870247e-05, "loss": 0.6749, "step": 10562 }, { "epoch": 0.6, "learning_rate": 7.307262952720364e-05, "loss": 0.7698, "step": 10563 }, { "epoch": 0.6, "learning_rate": 7.30549366845742e-05, "loss": 0.6865, "step": 10564 }, { "epoch": 0.6, "learning_rate": 7.303724475141134e-05, "loss": 0.6434, "step": 10565 }, { "epoch": 0.6, "learning_rate": 7.301955372831214e-05, "loss": 0.6819, "step": 10566 }, { "epoch": 0.6, "learning_rate": 7.300186361587378e-05, "loss": 0.6943, "step": 10567 }, { "epoch": 0.6, "learning_rate": 7.298417441469333e-05, "loss": 0.7109, "step": 10568 }, { "epoch": 0.6, "learning_rate": 7.296648612536783e-05, "loss": 0.6799, "step": 10569 }, { "epoch": 0.6, "learning_rate": 7.294879874849434e-05, "loss": 0.678, "step": 10570 }, { "epoch": 0.6, "learning_rate": 7.293111228466983e-05, "loss": 0.7076, "step": 10571 }, { "epoch": 0.6, "learning_rate": 7.291342673449129e-05, "loss": 0.755, "step": 10572 }, { "epoch": 0.6, "learning_rate": 7.289574209855559e-05, "loss": 0.5819, "step": 10573 }, { "epoch": 0.6, "learning_rate": 7.287805837745972e-05, "loss": 0.8147, "step": 10574 }, { "epoch": 0.6, "learning_rate": 7.286037557180047e-05, "loss": 0.6645, "step": 10575 }, { "epoch": 0.6, "learning_rate": 7.284269368217474e-05, "loss": 0.8261, "step": 10576 }, { "epoch": 0.6, "learning_rate": 7.282501270917933e-05, "loss": 0.8074, "step": 10577 }, { "epoch": 0.6, "learning_rate": 7.280733265341098e-05, "loss": 0.6326, "step": 10578 }, { "epoch": 0.6, "learning_rate": 7.278965351546648e-05, "loss": 0.7468, "step": 10579 }, { "epoch": 0.6, "learning_rate": 7.277197529594257e-05, "loss": 0.6712, "step": 10580 }, { "epoch": 0.6, "learning_rate": 7.275429799543589e-05, "loss": 0.7799, "step": 10581 }, { "epoch": 0.6, "learning_rate": 7.273662161454309e-05, "loss": 0.6364, "step": 10582 }, { "epoch": 0.6, "learning_rate": 7.271894615386083e-05, "loss": 0.6111, "step": 10583 }, { "epoch": 0.6, "learning_rate": 7.270127161398565e-05, "loss": 0.6873, "step": 10584 }, { "epoch": 0.6, "learning_rate": 7.268359799551416e-05, "loss": 0.7371, "step": 10585 }, { "epoch": 0.6, "learning_rate": 7.266592529904291e-05, "loss": 0.655, "step": 10586 }, { "epoch": 0.6, "learning_rate": 7.264825352516836e-05, "loss": 0.6366, "step": 10587 }, { "epoch": 0.6, "learning_rate": 7.263058267448696e-05, "loss": 0.7184, "step": 10588 }, { "epoch": 0.6, "learning_rate": 7.261291274759516e-05, "loss": 0.6855, "step": 10589 }, { "epoch": 0.6, "learning_rate": 7.259524374508941e-05, "loss": 0.7703, "step": 10590 }, { "epoch": 0.6, "learning_rate": 7.257757566756605e-05, "loss": 0.7445, "step": 10591 }, { "epoch": 0.6, "learning_rate": 7.255990851562141e-05, "loss": 0.7938, "step": 10592 }, { "epoch": 0.6, "learning_rate": 7.254224228985185e-05, "loss": 0.6079, "step": 10593 }, { "epoch": 0.6, "learning_rate": 7.25245769908536e-05, "loss": 0.7069, "step": 10594 }, { "epoch": 0.6, "learning_rate": 7.250691261922298e-05, "loss": 0.7714, "step": 10595 }, { "epoch": 0.6, "learning_rate": 7.24892491755561e-05, "loss": 0.7347, "step": 10596 }, { "epoch": 0.6, "learning_rate": 7.247158666044927e-05, "loss": 0.6271, "step": 10597 }, { "epoch": 0.6, "learning_rate": 7.245392507449854e-05, "loss": 0.7417, "step": 10598 }, { "epoch": 0.6, "learning_rate": 7.243626441830009e-05, "loss": 0.6914, "step": 10599 }, { "epoch": 0.6, "learning_rate": 7.241860469245001e-05, "loss": 0.678, "step": 10600 }, { "epoch": 0.6, "learning_rate": 7.240094589754433e-05, "loss": 0.6713, "step": 10601 }, { "epoch": 0.6, "learning_rate": 7.238328803417918e-05, "loss": 0.745, "step": 10602 }, { "epoch": 0.6, "learning_rate": 7.236563110295045e-05, "loss": 0.6785, "step": 10603 }, { "epoch": 0.6, "learning_rate": 7.234797510445411e-05, "loss": 0.7676, "step": 10604 }, { "epoch": 0.6, "learning_rate": 7.233032003928615e-05, "loss": 0.7256, "step": 10605 }, { "epoch": 0.6, "learning_rate": 7.231266590804246e-05, "loss": 0.7076, "step": 10606 }, { "epoch": 0.6, "learning_rate": 7.22950127113189e-05, "loss": 0.759, "step": 10607 }, { "epoch": 0.6, "learning_rate": 7.227736044971132e-05, "loss": 0.698, "step": 10608 }, { "epoch": 0.6, "learning_rate": 7.225970912381556e-05, "loss": 0.7041, "step": 10609 }, { "epoch": 0.6, "learning_rate": 7.224205873422735e-05, "loss": 0.7367, "step": 10610 }, { "epoch": 0.6, "learning_rate": 7.222440928154243e-05, "loss": 0.6944, "step": 10611 }, { "epoch": 0.6, "learning_rate": 7.220676076635656e-05, "loss": 0.5471, "step": 10612 }, { "epoch": 0.6, "learning_rate": 7.218911318926542e-05, "loss": 0.6616, "step": 10613 }, { "epoch": 0.6, "learning_rate": 7.217146655086462e-05, "loss": 0.7377, "step": 10614 }, { "epoch": 0.6, "learning_rate": 7.215382085174984e-05, "loss": 0.7625, "step": 10615 }, { "epoch": 0.6, "learning_rate": 7.213617609251659e-05, "loss": 0.7074, "step": 10616 }, { "epoch": 0.6, "learning_rate": 7.211853227376053e-05, "loss": 0.7413, "step": 10617 }, { "epoch": 0.6, "learning_rate": 7.210088939607708e-05, "loss": 0.6764, "step": 10618 }, { "epoch": 0.6, "learning_rate": 7.208324746006181e-05, "loss": 0.6857, "step": 10619 }, { "epoch": 0.6, "learning_rate": 7.206560646631014e-05, "loss": 0.6987, "step": 10620 }, { "epoch": 0.6, "learning_rate": 7.20479664154175e-05, "loss": 0.8114, "step": 10621 }, { "epoch": 0.6, "learning_rate": 7.203032730797932e-05, "loss": 0.6981, "step": 10622 }, { "epoch": 0.6, "learning_rate": 7.201268914459092e-05, "loss": 0.6753, "step": 10623 }, { "epoch": 0.6, "learning_rate": 7.199505192584771e-05, "loss": 0.7295, "step": 10624 }, { "epoch": 0.6, "learning_rate": 7.19774156523449e-05, "loss": 0.7176, "step": 10625 }, { "epoch": 0.6, "learning_rate": 7.195978032467783e-05, "loss": 0.6579, "step": 10626 }, { "epoch": 0.6, "learning_rate": 7.194214594344168e-05, "loss": 0.7664, "step": 10627 }, { "epoch": 0.6, "learning_rate": 7.192451250923171e-05, "loss": 0.7568, "step": 10628 }, { "epoch": 0.6, "learning_rate": 7.190688002264308e-05, "loss": 0.7159, "step": 10629 }, { "epoch": 0.6, "learning_rate": 7.18892484842709e-05, "loss": 0.7635, "step": 10630 }, { "epoch": 0.6, "learning_rate": 7.187161789471034e-05, "loss": 0.6932, "step": 10631 }, { "epoch": 0.6, "learning_rate": 7.185398825455645e-05, "loss": 0.7597, "step": 10632 }, { "epoch": 0.6, "learning_rate": 7.183635956440425e-05, "loss": 0.7568, "step": 10633 }, { "epoch": 0.6, "learning_rate": 7.181873182484875e-05, "loss": 0.6795, "step": 10634 }, { "epoch": 0.6, "learning_rate": 7.1801105036485e-05, "loss": 0.703, "step": 10635 }, { "epoch": 0.6, "learning_rate": 7.178347919990789e-05, "loss": 0.6861, "step": 10636 }, { "epoch": 0.6, "learning_rate": 7.176585431571235e-05, "loss": 0.7358, "step": 10637 }, { "epoch": 0.6, "learning_rate": 7.17482303844933e-05, "loss": 0.7899, "step": 10638 }, { "epoch": 0.6, "learning_rate": 7.173060740684558e-05, "loss": 0.6162, "step": 10639 }, { "epoch": 0.6, "learning_rate": 7.171298538336396e-05, "loss": 0.7053, "step": 10640 }, { "epoch": 0.6, "learning_rate": 7.169536431464328e-05, "loss": 0.739, "step": 10641 }, { "epoch": 0.6, "learning_rate": 7.167774420127829e-05, "loss": 0.7398, "step": 10642 }, { "epoch": 0.6, "learning_rate": 7.16601250438637e-05, "loss": 0.76, "step": 10643 }, { "epoch": 0.6, "learning_rate": 7.164250684299423e-05, "loss": 0.6598, "step": 10644 }, { "epoch": 0.6, "learning_rate": 7.16248895992645e-05, "loss": 0.6794, "step": 10645 }, { "epoch": 0.6, "learning_rate": 7.160727331326917e-05, "loss": 0.6682, "step": 10646 }, { "epoch": 0.6, "learning_rate": 7.158965798560287e-05, "loss": 0.6868, "step": 10647 }, { "epoch": 0.6, "learning_rate": 7.157204361686011e-05, "loss": 0.6255, "step": 10648 }, { "epoch": 0.6, "learning_rate": 7.15544302076354e-05, "loss": 0.6029, "step": 10649 }, { "epoch": 0.6, "learning_rate": 7.153681775852328e-05, "loss": 0.8189, "step": 10650 }, { "epoch": 0.6, "learning_rate": 7.15192062701182e-05, "loss": 0.713, "step": 10651 }, { "epoch": 0.6, "learning_rate": 7.150159574301462e-05, "loss": 0.653, "step": 10652 }, { "epoch": 0.6, "learning_rate": 7.14839861778069e-05, "loss": 0.7046, "step": 10653 }, { "epoch": 0.6, "learning_rate": 7.146637757508949e-05, "loss": 0.7321, "step": 10654 }, { "epoch": 0.6, "learning_rate": 7.144876993545664e-05, "loss": 0.7387, "step": 10655 }, { "epoch": 0.6, "learning_rate": 7.143116325950265e-05, "loss": 0.5981, "step": 10656 }, { "epoch": 0.6, "learning_rate": 7.141355754782184e-05, "loss": 0.7269, "step": 10657 }, { "epoch": 0.6, "learning_rate": 7.139595280100846e-05, "loss": 0.6148, "step": 10658 }, { "epoch": 0.6, "learning_rate": 7.137834901965665e-05, "loss": 0.7258, "step": 10659 }, { "epoch": 0.6, "learning_rate": 7.136074620436067e-05, "loss": 0.6605, "step": 10660 }, { "epoch": 0.6, "learning_rate": 7.134314435571458e-05, "loss": 0.7194, "step": 10661 }, { "epoch": 0.6, "learning_rate": 7.132554347431255e-05, "loss": 0.7132, "step": 10662 }, { "epoch": 0.6, "learning_rate": 7.130794356074859e-05, "loss": 0.6856, "step": 10663 }, { "epoch": 0.6, "learning_rate": 7.129034461561682e-05, "loss": 0.7197, "step": 10664 }, { "epoch": 0.6, "learning_rate": 7.12727466395112e-05, "loss": 0.898, "step": 10665 }, { "epoch": 0.61, "learning_rate": 7.125514963302569e-05, "loss": 0.7043, "step": 10666 }, { "epoch": 0.61, "learning_rate": 7.12375535967543e-05, "loss": 0.7074, "step": 10667 }, { "epoch": 0.61, "learning_rate": 7.12199585312909e-05, "loss": 0.6689, "step": 10668 }, { "epoch": 0.61, "learning_rate": 7.12023644372294e-05, "loss": 0.7042, "step": 10669 }, { "epoch": 0.61, "learning_rate": 7.118477131516361e-05, "loss": 0.7763, "step": 10670 }, { "epoch": 0.61, "learning_rate": 7.116717916568736e-05, "loss": 0.7338, "step": 10671 }, { "epoch": 0.61, "learning_rate": 7.114958798939442e-05, "loss": 0.7635, "step": 10672 }, { "epoch": 0.61, "learning_rate": 7.113199778687853e-05, "loss": 0.6959, "step": 10673 }, { "epoch": 0.61, "learning_rate": 7.111440855873346e-05, "loss": 0.7034, "step": 10674 }, { "epoch": 0.61, "learning_rate": 7.109682030555283e-05, "loss": 0.7152, "step": 10675 }, { "epoch": 0.61, "learning_rate": 7.107923302793035e-05, "loss": 0.6745, "step": 10676 }, { "epoch": 0.61, "learning_rate": 7.10616467264596e-05, "loss": 0.7579, "step": 10677 }, { "epoch": 0.61, "learning_rate": 7.104406140173414e-05, "loss": 0.74, "step": 10678 }, { "epoch": 0.61, "learning_rate": 7.102647705434756e-05, "loss": 0.7254, "step": 10679 }, { "epoch": 0.61, "learning_rate": 7.100889368489336e-05, "loss": 0.6807, "step": 10680 }, { "epoch": 0.61, "learning_rate": 7.099131129396501e-05, "loss": 0.7074, "step": 10681 }, { "epoch": 0.61, "learning_rate": 7.0973729882156e-05, "loss": 0.6925, "step": 10682 }, { "epoch": 0.61, "learning_rate": 7.095614945005974e-05, "loss": 0.7834, "step": 10683 }, { "epoch": 0.61, "learning_rate": 7.093856999826963e-05, "loss": 0.7346, "step": 10684 }, { "epoch": 0.61, "learning_rate": 7.092099152737897e-05, "loss": 0.7874, "step": 10685 }, { "epoch": 0.61, "learning_rate": 7.090341403798109e-05, "loss": 0.7649, "step": 10686 }, { "epoch": 0.61, "learning_rate": 7.088583753066933e-05, "loss": 0.7424, "step": 10687 }, { "epoch": 0.61, "learning_rate": 7.086826200603686e-05, "loss": 0.7294, "step": 10688 }, { "epoch": 0.61, "learning_rate": 7.0850687464677e-05, "loss": 0.704, "step": 10689 }, { "epoch": 0.61, "learning_rate": 7.083311390718284e-05, "loss": 0.7554, "step": 10690 }, { "epoch": 0.61, "learning_rate": 7.081554133414764e-05, "loss": 0.6676, "step": 10691 }, { "epoch": 0.61, "learning_rate": 7.07979697461644e-05, "loss": 0.7571, "step": 10692 }, { "epoch": 0.61, "learning_rate": 7.07803991438263e-05, "loss": 0.7557, "step": 10693 }, { "epoch": 0.61, "learning_rate": 7.076282952772633e-05, "loss": 0.6261, "step": 10694 }, { "epoch": 0.61, "learning_rate": 7.074526089845756e-05, "loss": 0.7483, "step": 10695 }, { "epoch": 0.61, "learning_rate": 7.072769325661297e-05, "loss": 0.7229, "step": 10696 }, { "epoch": 0.61, "learning_rate": 7.071012660278547e-05, "loss": 0.6738, "step": 10697 }, { "epoch": 0.61, "learning_rate": 7.069256093756806e-05, "loss": 0.7244, "step": 10698 }, { "epoch": 0.61, "learning_rate": 7.067499626155354e-05, "loss": 0.6966, "step": 10699 }, { "epoch": 0.61, "learning_rate": 7.065743257533483e-05, "loss": 0.6926, "step": 10700 }, { "epoch": 0.61, "learning_rate": 7.063986987950468e-05, "loss": 0.7197, "step": 10701 }, { "epoch": 0.61, "learning_rate": 7.062230817465595e-05, "loss": 0.7364, "step": 10702 }, { "epoch": 0.61, "learning_rate": 7.060474746138137e-05, "loss": 0.7971, "step": 10703 }, { "epoch": 0.61, "learning_rate": 7.058718774027364e-05, "loss": 0.6952, "step": 10704 }, { "epoch": 0.61, "learning_rate": 7.056962901192548e-05, "loss": 0.7311, "step": 10705 }, { "epoch": 0.61, "learning_rate": 7.055207127692953e-05, "loss": 0.6283, "step": 10706 }, { "epoch": 0.61, "learning_rate": 7.053451453587841e-05, "loss": 0.7592, "step": 10707 }, { "epoch": 0.61, "learning_rate": 7.051695878936466e-05, "loss": 0.7652, "step": 10708 }, { "epoch": 0.61, "learning_rate": 7.049940403798089e-05, "loss": 0.6332, "step": 10709 }, { "epoch": 0.61, "learning_rate": 7.048185028231962e-05, "loss": 0.6857, "step": 10710 }, { "epoch": 0.61, "learning_rate": 7.046429752297328e-05, "loss": 0.6933, "step": 10711 }, { "epoch": 0.61, "learning_rate": 7.044674576053439e-05, "loss": 0.674, "step": 10712 }, { "epoch": 0.61, "learning_rate": 7.042919499559537e-05, "loss": 0.6824, "step": 10713 }, { "epoch": 0.61, "learning_rate": 7.041164522874851e-05, "loss": 0.74, "step": 10714 }, { "epoch": 0.61, "learning_rate": 7.039409646058624e-05, "loss": 0.6503, "step": 10715 }, { "epoch": 0.61, "learning_rate": 7.037654869170086e-05, "loss": 0.7808, "step": 10716 }, { "epoch": 0.61, "learning_rate": 7.035900192268464e-05, "loss": 0.6093, "step": 10717 }, { "epoch": 0.61, "learning_rate": 7.034145615412985e-05, "loss": 0.616, "step": 10718 }, { "epoch": 0.61, "learning_rate": 7.03239113866287e-05, "loss": 0.7884, "step": 10719 }, { "epoch": 0.61, "learning_rate": 7.030636762077336e-05, "loss": 0.6732, "step": 10720 }, { "epoch": 0.61, "learning_rate": 7.028882485715604e-05, "loss": 0.6682, "step": 10721 }, { "epoch": 0.61, "learning_rate": 7.027128309636877e-05, "loss": 0.6975, "step": 10722 }, { "epoch": 0.61, "learning_rate": 7.025374233900364e-05, "loss": 0.755, "step": 10723 }, { "epoch": 0.61, "learning_rate": 7.023620258565272e-05, "loss": 0.7288, "step": 10724 }, { "epoch": 0.61, "learning_rate": 7.021866383690802e-05, "loss": 0.7571, "step": 10725 }, { "epoch": 0.61, "learning_rate": 7.020112609336153e-05, "loss": 0.8673, "step": 10726 }, { "epoch": 0.61, "learning_rate": 7.018358935560519e-05, "loss": 0.7952, "step": 10727 }, { "epoch": 0.61, "learning_rate": 7.016605362423092e-05, "loss": 0.7119, "step": 10728 }, { "epoch": 0.61, "learning_rate": 7.014851889983057e-05, "loss": 0.7304, "step": 10729 }, { "epoch": 0.61, "learning_rate": 7.013098518299598e-05, "loss": 0.6544, "step": 10730 }, { "epoch": 0.61, "learning_rate": 7.011345247431897e-05, "loss": 0.6862, "step": 10731 }, { "epoch": 0.61, "learning_rate": 7.009592077439134e-05, "loss": 0.6915, "step": 10732 }, { "epoch": 0.61, "learning_rate": 7.007839008380479e-05, "loss": 0.8156, "step": 10733 }, { "epoch": 0.61, "learning_rate": 7.006086040315107e-05, "loss": 0.6797, "step": 10734 }, { "epoch": 0.61, "learning_rate": 7.004333173302185e-05, "loss": 0.7518, "step": 10735 }, { "epoch": 0.61, "learning_rate": 7.002580407400872e-05, "loss": 0.6585, "step": 10736 }, { "epoch": 0.61, "learning_rate": 7.00082774267033e-05, "loss": 0.6653, "step": 10737 }, { "epoch": 0.61, "learning_rate": 6.999075179169719e-05, "loss": 0.8018, "step": 10738 }, { "epoch": 0.61, "learning_rate": 6.99732271695819e-05, "loss": 0.708, "step": 10739 }, { "epoch": 0.61, "learning_rate": 6.995570356094891e-05, "loss": 0.7415, "step": 10740 }, { "epoch": 0.61, "learning_rate": 6.993818096638978e-05, "loss": 0.77, "step": 10741 }, { "epoch": 0.61, "learning_rate": 6.992065938649584e-05, "loss": 0.7696, "step": 10742 }, { "epoch": 0.61, "learning_rate": 6.990313882185857e-05, "loss": 0.7453, "step": 10743 }, { "epoch": 0.61, "learning_rate": 6.988561927306927e-05, "loss": 0.7337, "step": 10744 }, { "epoch": 0.61, "learning_rate": 6.986810074071932e-05, "loss": 0.7029, "step": 10745 }, { "epoch": 0.61, "learning_rate": 6.985058322539994e-05, "loss": 0.7338, "step": 10746 }, { "epoch": 0.61, "learning_rate": 6.983306672770249e-05, "loss": 0.7135, "step": 10747 }, { "epoch": 0.61, "learning_rate": 6.981555124821814e-05, "loss": 0.77, "step": 10748 }, { "epoch": 0.61, "learning_rate": 6.979803678753809e-05, "loss": 0.742, "step": 10749 }, { "epoch": 0.61, "learning_rate": 6.978052334625355e-05, "loss": 0.6529, "step": 10750 }, { "epoch": 0.61, "learning_rate": 6.976301092495556e-05, "loss": 0.7351, "step": 10751 }, { "epoch": 0.61, "learning_rate": 6.974549952423527e-05, "loss": 0.6223, "step": 10752 }, { "epoch": 0.61, "learning_rate": 6.972798914468369e-05, "loss": 0.812, "step": 10753 }, { "epoch": 0.61, "learning_rate": 6.971047978689189e-05, "loss": 0.751, "step": 10754 }, { "epoch": 0.61, "learning_rate": 6.969297145145081e-05, "loss": 0.746, "step": 10755 }, { "epoch": 0.61, "learning_rate": 6.967546413895143e-05, "loss": 0.7002, "step": 10756 }, { "epoch": 0.61, "learning_rate": 6.96579578499847e-05, "loss": 0.73, "step": 10757 }, { "epoch": 0.61, "learning_rate": 6.964045258514147e-05, "loss": 0.7696, "step": 10758 }, { "epoch": 0.61, "learning_rate": 6.962294834501255e-05, "loss": 0.664, "step": 10759 }, { "epoch": 0.61, "learning_rate": 6.96054451301888e-05, "loss": 0.6813, "step": 10760 }, { "epoch": 0.61, "learning_rate": 6.9587942941261e-05, "loss": 0.6926, "step": 10761 }, { "epoch": 0.61, "learning_rate": 6.957044177881987e-05, "loss": 0.6996, "step": 10762 }, { "epoch": 0.61, "learning_rate": 6.955294164345612e-05, "loss": 0.689, "step": 10763 }, { "epoch": 0.61, "learning_rate": 6.953544253576048e-05, "loss": 0.7512, "step": 10764 }, { "epoch": 0.61, "learning_rate": 6.951794445632357e-05, "loss": 0.6916, "step": 10765 }, { "epoch": 0.61, "learning_rate": 6.950044740573591e-05, "loss": 0.7091, "step": 10766 }, { "epoch": 0.61, "learning_rate": 6.948295138458817e-05, "loss": 0.7141, "step": 10767 }, { "epoch": 0.61, "learning_rate": 6.946545639347084e-05, "loss": 0.6405, "step": 10768 }, { "epoch": 0.61, "learning_rate": 6.944796243297444e-05, "loss": 0.7559, "step": 10769 }, { "epoch": 0.61, "learning_rate": 6.943046950368944e-05, "loss": 0.6419, "step": 10770 }, { "epoch": 0.61, "learning_rate": 6.941297760620627e-05, "loss": 0.7098, "step": 10771 }, { "epoch": 0.61, "learning_rate": 6.939548674111533e-05, "loss": 0.7192, "step": 10772 }, { "epoch": 0.61, "learning_rate": 6.937799690900696e-05, "loss": 0.6465, "step": 10773 }, { "epoch": 0.61, "learning_rate": 6.936050811047152e-05, "loss": 0.7566, "step": 10774 }, { "epoch": 0.61, "learning_rate": 6.934302034609927e-05, "loss": 0.699, "step": 10775 }, { "epoch": 0.61, "learning_rate": 6.932553361648046e-05, "loss": 0.6887, "step": 10776 }, { "epoch": 0.61, "learning_rate": 6.930804792220538e-05, "loss": 0.6565, "step": 10777 }, { "epoch": 0.61, "learning_rate": 6.929056326386415e-05, "loss": 0.6984, "step": 10778 }, { "epoch": 0.61, "learning_rate": 6.927307964204694e-05, "loss": 0.8056, "step": 10779 }, { "epoch": 0.61, "learning_rate": 6.925559705734394e-05, "loss": 0.7066, "step": 10780 }, { "epoch": 0.61, "learning_rate": 6.923811551034513e-05, "loss": 0.7562, "step": 10781 }, { "epoch": 0.61, "learning_rate": 6.922063500164058e-05, "loss": 0.6707, "step": 10782 }, { "epoch": 0.61, "learning_rate": 6.920315553182035e-05, "loss": 0.7109, "step": 10783 }, { "epoch": 0.61, "learning_rate": 6.918567710147436e-05, "loss": 0.7119, "step": 10784 }, { "epoch": 0.61, "learning_rate": 6.916819971119257e-05, "loss": 0.745, "step": 10785 }, { "epoch": 0.61, "learning_rate": 6.915072336156494e-05, "loss": 0.7916, "step": 10786 }, { "epoch": 0.61, "learning_rate": 6.913324805318131e-05, "loss": 0.6359, "step": 10787 }, { "epoch": 0.61, "learning_rate": 6.911577378663148e-05, "loss": 0.7522, "step": 10788 }, { "epoch": 0.61, "learning_rate": 6.909830056250527e-05, "loss": 0.6478, "step": 10789 }, { "epoch": 0.61, "learning_rate": 6.908082838139248e-05, "loss": 0.7042, "step": 10790 }, { "epoch": 0.61, "learning_rate": 6.90633572438828e-05, "loss": 0.7207, "step": 10791 }, { "epoch": 0.61, "learning_rate": 6.904588715056594e-05, "loss": 0.6288, "step": 10792 }, { "epoch": 0.61, "learning_rate": 6.902841810203161e-05, "loss": 0.7351, "step": 10793 }, { "epoch": 0.61, "learning_rate": 6.901095009886936e-05, "loss": 0.6445, "step": 10794 }, { "epoch": 0.61, "learning_rate": 6.899348314166887e-05, "loss": 0.6788, "step": 10795 }, { "epoch": 0.61, "learning_rate": 6.897601723101959e-05, "loss": 0.6953, "step": 10796 }, { "epoch": 0.61, "learning_rate": 6.895855236751112e-05, "loss": 0.7292, "step": 10797 }, { "epoch": 0.61, "learning_rate": 6.894108855173288e-05, "loss": 0.6634, "step": 10798 }, { "epoch": 0.61, "learning_rate": 6.892362578427439e-05, "loss": 0.643, "step": 10799 }, { "epoch": 0.61, "learning_rate": 6.890616406572502e-05, "loss": 0.7332, "step": 10800 }, { "epoch": 0.61, "learning_rate": 6.888870339667413e-05, "loss": 0.6997, "step": 10801 }, { "epoch": 0.61, "learning_rate": 6.887124377771115e-05, "loss": 0.5942, "step": 10802 }, { "epoch": 0.61, "learning_rate": 6.885378520942531e-05, "loss": 0.8239, "step": 10803 }, { "epoch": 0.61, "learning_rate": 6.883632769240589e-05, "loss": 0.6565, "step": 10804 }, { "epoch": 0.61, "learning_rate": 6.881887122724213e-05, "loss": 0.7068, "step": 10805 }, { "epoch": 0.61, "learning_rate": 6.880141581452326e-05, "loss": 0.6391, "step": 10806 }, { "epoch": 0.61, "learning_rate": 6.878396145483841e-05, "loss": 0.7659, "step": 10807 }, { "epoch": 0.61, "learning_rate": 6.876650814877674e-05, "loss": 0.6154, "step": 10808 }, { "epoch": 0.61, "learning_rate": 6.874905589692733e-05, "loss": 0.5532, "step": 10809 }, { "epoch": 0.61, "learning_rate": 6.87316046998793e-05, "loss": 0.7173, "step": 10810 }, { "epoch": 0.61, "learning_rate": 6.871415455822155e-05, "loss": 0.7123, "step": 10811 }, { "epoch": 0.61, "learning_rate": 6.869670547254312e-05, "loss": 0.7591, "step": 10812 }, { "epoch": 0.61, "learning_rate": 6.867925744343301e-05, "loss": 0.7025, "step": 10813 }, { "epoch": 0.61, "learning_rate": 6.866181047148009e-05, "loss": 0.8078, "step": 10814 }, { "epoch": 0.61, "learning_rate": 6.864436455727326e-05, "loss": 0.7832, "step": 10815 }, { "epoch": 0.61, "learning_rate": 6.862691970140134e-05, "loss": 0.6297, "step": 10816 }, { "epoch": 0.61, "learning_rate": 6.860947590445321e-05, "loss": 0.7804, "step": 10817 }, { "epoch": 0.61, "learning_rate": 6.859203316701755e-05, "loss": 0.6613, "step": 10818 }, { "epoch": 0.61, "learning_rate": 6.857459148968317e-05, "loss": 0.7103, "step": 10819 }, { "epoch": 0.61, "learning_rate": 6.855715087303871e-05, "loss": 0.7138, "step": 10820 }, { "epoch": 0.61, "learning_rate": 6.853971131767287e-05, "loss": 0.7358, "step": 10821 }, { "epoch": 0.61, "learning_rate": 6.852227282417431e-05, "loss": 0.6965, "step": 10822 }, { "epoch": 0.61, "learning_rate": 6.850483539313158e-05, "loss": 0.7201, "step": 10823 }, { "epoch": 0.61, "learning_rate": 6.848739902513327e-05, "loss": 0.6849, "step": 10824 }, { "epoch": 0.61, "learning_rate": 6.846996372076786e-05, "loss": 0.6868, "step": 10825 }, { "epoch": 0.61, "learning_rate": 6.845252948062388e-05, "loss": 0.7321, "step": 10826 }, { "epoch": 0.61, "learning_rate": 6.843509630528977e-05, "loss": 0.6699, "step": 10827 }, { "epoch": 0.61, "learning_rate": 6.841766419535394e-05, "loss": 0.7338, "step": 10828 }, { "epoch": 0.61, "learning_rate": 6.840023315140475e-05, "loss": 0.7665, "step": 10829 }, { "epoch": 0.61, "learning_rate": 6.838280317403058e-05, "loss": 0.6972, "step": 10830 }, { "epoch": 0.61, "learning_rate": 6.836537426381972e-05, "loss": 0.7059, "step": 10831 }, { "epoch": 0.61, "learning_rate": 6.834794642136049e-05, "loss": 0.6559, "step": 10832 }, { "epoch": 0.61, "learning_rate": 6.833051964724101e-05, "loss": 0.6219, "step": 10833 }, { "epoch": 0.61, "learning_rate": 6.831309394204957e-05, "loss": 0.7335, "step": 10834 }, { "epoch": 0.61, "learning_rate": 6.829566930637432e-05, "loss": 0.6544, "step": 10835 }, { "epoch": 0.61, "learning_rate": 6.827824574080337e-05, "loss": 0.6956, "step": 10836 }, { "epoch": 0.61, "learning_rate": 6.826082324592481e-05, "loss": 0.6692, "step": 10837 }, { "epoch": 0.61, "learning_rate": 6.824340182232673e-05, "loss": 0.7058, "step": 10838 }, { "epoch": 0.61, "learning_rate": 6.822598147059713e-05, "loss": 0.7127, "step": 10839 }, { "epoch": 0.61, "learning_rate": 6.820856219132397e-05, "loss": 0.6826, "step": 10840 }, { "epoch": 0.61, "learning_rate": 6.819114398509518e-05, "loss": 0.79, "step": 10841 }, { "epoch": 0.62, "learning_rate": 6.817372685249875e-05, "loss": 0.6833, "step": 10842 }, { "epoch": 0.62, "learning_rate": 6.815631079412248e-05, "loss": 0.7865, "step": 10843 }, { "epoch": 0.62, "learning_rate": 6.813889581055425e-05, "loss": 0.6721, "step": 10844 }, { "epoch": 0.62, "learning_rate": 6.812148190238183e-05, "loss": 0.6753, "step": 10845 }, { "epoch": 0.62, "learning_rate": 6.8104069070193e-05, "loss": 0.6866, "step": 10846 }, { "epoch": 0.62, "learning_rate": 6.808665731457552e-05, "loss": 0.7984, "step": 10847 }, { "epoch": 0.62, "learning_rate": 6.806924663611704e-05, "loss": 0.6796, "step": 10848 }, { "epoch": 0.62, "learning_rate": 6.80518370354052e-05, "loss": 0.7557, "step": 10849 }, { "epoch": 0.62, "learning_rate": 6.803442851302765e-05, "loss": 0.636, "step": 10850 }, { "epoch": 0.62, "learning_rate": 6.8017021069572e-05, "loss": 0.7494, "step": 10851 }, { "epoch": 0.62, "learning_rate": 6.799961470562573e-05, "loss": 0.6929, "step": 10852 }, { "epoch": 0.62, "learning_rate": 6.79822094217764e-05, "loss": 0.6083, "step": 10853 }, { "epoch": 0.62, "learning_rate": 6.79648052186115e-05, "loss": 0.6965, "step": 10854 }, { "epoch": 0.62, "learning_rate": 6.794740209671843e-05, "loss": 0.6665, "step": 10855 }, { "epoch": 0.62, "learning_rate": 6.793000005668458e-05, "loss": 0.6957, "step": 10856 }, { "epoch": 0.62, "learning_rate": 6.791259909909733e-05, "loss": 0.5827, "step": 10857 }, { "epoch": 0.62, "learning_rate": 6.789519922454403e-05, "loss": 0.7041, "step": 10858 }, { "epoch": 0.62, "learning_rate": 6.787780043361194e-05, "loss": 0.702, "step": 10859 }, { "epoch": 0.62, "learning_rate": 6.786040272688837e-05, "loss": 0.8232, "step": 10860 }, { "epoch": 0.62, "learning_rate": 6.784300610496048e-05, "loss": 0.7078, "step": 10861 }, { "epoch": 0.62, "learning_rate": 6.782561056841546e-05, "loss": 0.7054, "step": 10862 }, { "epoch": 0.62, "learning_rate": 6.780821611784043e-05, "loss": 0.7116, "step": 10863 }, { "epoch": 0.62, "learning_rate": 6.779082275382258e-05, "loss": 0.8681, "step": 10864 }, { "epoch": 0.62, "learning_rate": 6.77734304769489e-05, "loss": 0.6947, "step": 10865 }, { "epoch": 0.62, "learning_rate": 6.775603928780647e-05, "loss": 0.5904, "step": 10866 }, { "epoch": 0.62, "learning_rate": 6.773864918698228e-05, "loss": 0.6565, "step": 10867 }, { "epoch": 0.62, "learning_rate": 6.772126017506328e-05, "loss": 0.7289, "step": 10868 }, { "epoch": 0.62, "learning_rate": 6.770387225263644e-05, "loss": 0.6747, "step": 10869 }, { "epoch": 0.62, "learning_rate": 6.768648542028858e-05, "loss": 0.6774, "step": 10870 }, { "epoch": 0.62, "learning_rate": 6.76690996786066e-05, "loss": 0.7254, "step": 10871 }, { "epoch": 0.62, "learning_rate": 6.765171502817726e-05, "loss": 0.7322, "step": 10872 }, { "epoch": 0.62, "learning_rate": 6.763433146958741e-05, "loss": 0.7108, "step": 10873 }, { "epoch": 0.62, "learning_rate": 6.761694900342372e-05, "loss": 0.6938, "step": 10874 }, { "epoch": 0.62, "learning_rate": 6.759956763027296e-05, "loss": 0.6817, "step": 10875 }, { "epoch": 0.62, "learning_rate": 6.758218735072177e-05, "loss": 0.7259, "step": 10876 }, { "epoch": 0.62, "learning_rate": 6.756480816535677e-05, "loss": 0.7256, "step": 10877 }, { "epoch": 0.62, "learning_rate": 6.754743007476453e-05, "loss": 0.7468, "step": 10878 }, { "epoch": 0.62, "learning_rate": 6.753005307953167e-05, "loss": 0.6995, "step": 10879 }, { "epoch": 0.62, "learning_rate": 6.751267718024466e-05, "loss": 0.5989, "step": 10880 }, { "epoch": 0.62, "learning_rate": 6.749530237748998e-05, "loss": 0.7122, "step": 10881 }, { "epoch": 0.62, "learning_rate": 6.74779286718541e-05, "loss": 0.6661, "step": 10882 }, { "epoch": 0.62, "learning_rate": 6.746055606392343e-05, "loss": 0.7163, "step": 10883 }, { "epoch": 0.62, "learning_rate": 6.744318455428436e-05, "loss": 0.7461, "step": 10884 }, { "epoch": 0.62, "learning_rate": 6.742581414352316e-05, "loss": 0.7147, "step": 10885 }, { "epoch": 0.62, "learning_rate": 6.740844483222615e-05, "loss": 0.7721, "step": 10886 }, { "epoch": 0.62, "learning_rate": 6.739107662097963e-05, "loss": 0.6536, "step": 10887 }, { "epoch": 0.62, "learning_rate": 6.737370951036976e-05, "loss": 0.7581, "step": 10888 }, { "epoch": 0.62, "learning_rate": 6.735634350098278e-05, "loss": 0.7461, "step": 10889 }, { "epoch": 0.62, "learning_rate": 6.733897859340481e-05, "loss": 0.6997, "step": 10890 }, { "epoch": 0.62, "learning_rate": 6.732161478822201e-05, "loss": 0.7753, "step": 10891 }, { "epoch": 0.62, "learning_rate": 6.730425208602037e-05, "loss": 0.6318, "step": 10892 }, { "epoch": 0.62, "learning_rate": 6.728689048738598e-05, "loss": 0.7381, "step": 10893 }, { "epoch": 0.62, "learning_rate": 6.726952999290482e-05, "loss": 0.6543, "step": 10894 }, { "epoch": 0.62, "learning_rate": 6.725217060316286e-05, "loss": 0.7109, "step": 10895 }, { "epoch": 0.62, "learning_rate": 6.723481231874604e-05, "loss": 0.6806, "step": 10896 }, { "epoch": 0.62, "learning_rate": 6.721745514024022e-05, "loss": 0.7188, "step": 10897 }, { "epoch": 0.62, "learning_rate": 6.720009906823131e-05, "loss": 0.7081, "step": 10898 }, { "epoch": 0.62, "learning_rate": 6.718274410330502e-05, "loss": 0.8059, "step": 10899 }, { "epoch": 0.62, "learning_rate": 6.716539024604723e-05, "loss": 0.6694, "step": 10900 }, { "epoch": 0.62, "learning_rate": 6.71480374970436e-05, "loss": 0.7307, "step": 10901 }, { "epoch": 0.62, "learning_rate": 6.713068585687985e-05, "loss": 0.6799, "step": 10902 }, { "epoch": 0.62, "learning_rate": 6.711333532614168e-05, "loss": 0.6635, "step": 10903 }, { "epoch": 0.62, "learning_rate": 6.709598590541469e-05, "loss": 0.7378, "step": 10904 }, { "epoch": 0.62, "learning_rate": 6.707863759528446e-05, "loss": 0.8289, "step": 10905 }, { "epoch": 0.62, "learning_rate": 6.706129039633659e-05, "loss": 0.6894, "step": 10906 }, { "epoch": 0.62, "learning_rate": 6.70439443091565e-05, "loss": 0.7225, "step": 10907 }, { "epoch": 0.62, "learning_rate": 6.702659933432973e-05, "loss": 0.7584, "step": 10908 }, { "epoch": 0.62, "learning_rate": 6.700925547244173e-05, "loss": 0.6642, "step": 10909 }, { "epoch": 0.62, "learning_rate": 6.699191272407786e-05, "loss": 0.6575, "step": 10910 }, { "epoch": 0.62, "learning_rate": 6.697457108982349e-05, "loss": 0.7185, "step": 10911 }, { "epoch": 0.62, "learning_rate": 6.695723057026399e-05, "loss": 0.779, "step": 10912 }, { "epoch": 0.62, "learning_rate": 6.693989116598463e-05, "loss": 0.6515, "step": 10913 }, { "epoch": 0.62, "learning_rate": 6.692255287757061e-05, "loss": 0.5733, "step": 10914 }, { "epoch": 0.62, "learning_rate": 6.690521570560717e-05, "loss": 0.7493, "step": 10915 }, { "epoch": 0.62, "learning_rate": 6.688787965067953e-05, "loss": 0.6843, "step": 10916 }, { "epoch": 0.62, "learning_rate": 6.687054471337276e-05, "loss": 0.6682, "step": 10917 }, { "epoch": 0.62, "learning_rate": 6.685321089427201e-05, "loss": 0.7096, "step": 10918 }, { "epoch": 0.62, "learning_rate": 6.683587819396231e-05, "loss": 0.7277, "step": 10919 }, { "epoch": 0.62, "learning_rate": 6.68185466130287e-05, "loss": 0.7729, "step": 10920 }, { "epoch": 0.62, "learning_rate": 6.680121615205619e-05, "loss": 0.7769, "step": 10921 }, { "epoch": 0.62, "learning_rate": 6.67838868116297e-05, "loss": 0.8108, "step": 10922 }, { "epoch": 0.62, "learning_rate": 6.676655859233412e-05, "loss": 0.7205, "step": 10923 }, { "epoch": 0.62, "learning_rate": 6.674923149475432e-05, "loss": 0.7048, "step": 10924 }, { "epoch": 0.62, "learning_rate": 6.673190551947521e-05, "loss": 0.6299, "step": 10925 }, { "epoch": 0.62, "learning_rate": 6.67145806670815e-05, "loss": 0.6429, "step": 10926 }, { "epoch": 0.62, "learning_rate": 6.6697256938158e-05, "loss": 0.7444, "step": 10927 }, { "epoch": 0.62, "learning_rate": 6.667993433328945e-05, "loss": 0.684, "step": 10928 }, { "epoch": 0.62, "learning_rate": 6.666261285306047e-05, "loss": 0.7783, "step": 10929 }, { "epoch": 0.62, "learning_rate": 6.664529249805573e-05, "loss": 0.6179, "step": 10930 }, { "epoch": 0.62, "learning_rate": 6.662797326885986e-05, "loss": 0.7744, "step": 10931 }, { "epoch": 0.62, "learning_rate": 6.661065516605741e-05, "loss": 0.6063, "step": 10932 }, { "epoch": 0.62, "learning_rate": 6.65933381902329e-05, "loss": 0.744, "step": 10933 }, { "epoch": 0.62, "learning_rate": 6.657602234197085e-05, "loss": 0.774, "step": 10934 }, { "epoch": 0.62, "learning_rate": 6.655870762185572e-05, "loss": 0.7252, "step": 10935 }, { "epoch": 0.62, "learning_rate": 6.65413940304719e-05, "loss": 0.6941, "step": 10936 }, { "epoch": 0.62, "learning_rate": 6.652408156840377e-05, "loss": 0.6927, "step": 10937 }, { "epoch": 0.62, "learning_rate": 6.650677023623568e-05, "loss": 0.6586, "step": 10938 }, { "epoch": 0.62, "learning_rate": 6.648946003455194e-05, "loss": 0.761, "step": 10939 }, { "epoch": 0.62, "learning_rate": 6.64721509639368e-05, "loss": 0.6968, "step": 10940 }, { "epoch": 0.62, "learning_rate": 6.64548430249745e-05, "loss": 0.7057, "step": 10941 }, { "epoch": 0.62, "learning_rate": 6.643753621824922e-05, "loss": 0.603, "step": 10942 }, { "epoch": 0.62, "learning_rate": 6.642023054434515e-05, "loss": 0.7004, "step": 10943 }, { "epoch": 0.62, "learning_rate": 6.640292600384632e-05, "loss": 0.7778, "step": 10944 }, { "epoch": 0.62, "learning_rate": 6.638562259733687e-05, "loss": 0.6004, "step": 10945 }, { "epoch": 0.62, "learning_rate": 6.63683203254008e-05, "loss": 0.7468, "step": 10946 }, { "epoch": 0.62, "learning_rate": 6.635101918862212e-05, "loss": 0.7032, "step": 10947 }, { "epoch": 0.62, "learning_rate": 6.633371918758481e-05, "loss": 0.7221, "step": 10948 }, { "epoch": 0.62, "learning_rate": 6.631642032287275e-05, "loss": 0.6656, "step": 10949 }, { "epoch": 0.62, "learning_rate": 6.629912259506988e-05, "loss": 0.7449, "step": 10950 }, { "epoch": 0.62, "learning_rate": 6.628182600475999e-05, "loss": 0.5897, "step": 10951 }, { "epoch": 0.62, "learning_rate": 6.626453055252688e-05, "loss": 0.7316, "step": 10952 }, { "epoch": 0.62, "learning_rate": 6.624723623895432e-05, "loss": 0.6528, "step": 10953 }, { "epoch": 0.62, "learning_rate": 6.622994306462611e-05, "loss": 0.5909, "step": 10954 }, { "epoch": 0.62, "learning_rate": 6.621265103012584e-05, "loss": 0.6287, "step": 10955 }, { "epoch": 0.62, "learning_rate": 6.619536013603722e-05, "loss": 0.7585, "step": 10956 }, { "epoch": 0.62, "learning_rate": 6.617807038294387e-05, "loss": 0.7092, "step": 10957 }, { "epoch": 0.62, "learning_rate": 6.616078177142937e-05, "loss": 0.6967, "step": 10958 }, { "epoch": 0.62, "learning_rate": 6.61434943020772e-05, "loss": 0.7794, "step": 10959 }, { "epoch": 0.62, "learning_rate": 6.612620797547087e-05, "loss": 0.6443, "step": 10960 }, { "epoch": 0.62, "learning_rate": 6.610892279219388e-05, "loss": 0.6696, "step": 10961 }, { "epoch": 0.62, "learning_rate": 6.609163875282962e-05, "loss": 0.7093, "step": 10962 }, { "epoch": 0.62, "learning_rate": 6.607435585796147e-05, "loss": 0.6323, "step": 10963 }, { "epoch": 0.62, "learning_rate": 6.605707410817281e-05, "loss": 0.8296, "step": 10964 }, { "epoch": 0.62, "learning_rate": 6.603979350404693e-05, "loss": 0.6881, "step": 10965 }, { "epoch": 0.62, "learning_rate": 6.602251404616705e-05, "loss": 0.6858, "step": 10966 }, { "epoch": 0.62, "learning_rate": 6.600523573511646e-05, "loss": 0.6497, "step": 10967 }, { "epoch": 0.62, "learning_rate": 6.598795857147829e-05, "loss": 0.7559, "step": 10968 }, { "epoch": 0.62, "learning_rate": 6.59706825558357e-05, "loss": 0.6618, "step": 10969 }, { "epoch": 0.62, "learning_rate": 6.595340768877184e-05, "loss": 0.7707, "step": 10970 }, { "epoch": 0.62, "learning_rate": 6.593613397086977e-05, "loss": 0.6968, "step": 10971 }, { "epoch": 0.62, "learning_rate": 6.59188614027125e-05, "loss": 0.7292, "step": 10972 }, { "epoch": 0.62, "learning_rate": 6.590158998488307e-05, "loss": 0.7792, "step": 10973 }, { "epoch": 0.62, "learning_rate": 6.588431971796439e-05, "loss": 0.6885, "step": 10974 }, { "epoch": 0.62, "learning_rate": 6.586705060253937e-05, "loss": 0.5991, "step": 10975 }, { "epoch": 0.62, "learning_rate": 6.584978263919092e-05, "loss": 0.7429, "step": 10976 }, { "epoch": 0.62, "learning_rate": 6.583251582850188e-05, "loss": 0.7906, "step": 10977 }, { "epoch": 0.62, "learning_rate": 6.581525017105501e-05, "loss": 0.7327, "step": 10978 }, { "epoch": 0.62, "learning_rate": 6.579798566743314e-05, "loss": 0.703, "step": 10979 }, { "epoch": 0.62, "learning_rate": 6.578072231821893e-05, "loss": 0.7541, "step": 10980 }, { "epoch": 0.62, "learning_rate": 6.57634601239951e-05, "loss": 0.634, "step": 10981 }, { "epoch": 0.62, "learning_rate": 6.574619908534426e-05, "loss": 0.7474, "step": 10982 }, { "epoch": 0.62, "learning_rate": 6.572893920284904e-05, "loss": 0.6279, "step": 10983 }, { "epoch": 0.62, "learning_rate": 6.5711680477092e-05, "loss": 0.7159, "step": 10984 }, { "epoch": 0.62, "learning_rate": 6.569442290865564e-05, "loss": 0.7966, "step": 10985 }, { "epoch": 0.62, "learning_rate": 6.567716649812253e-05, "loss": 0.7667, "step": 10986 }, { "epoch": 0.62, "learning_rate": 6.565991124607507e-05, "loss": 0.642, "step": 10987 }, { "epoch": 0.62, "learning_rate": 6.56426571530956e-05, "loss": 0.7926, "step": 10988 }, { "epoch": 0.62, "learning_rate": 6.56254042197666e-05, "loss": 0.689, "step": 10989 }, { "epoch": 0.62, "learning_rate": 6.560815244667036e-05, "loss": 0.6464, "step": 10990 }, { "epoch": 0.62, "learning_rate": 6.559090183438915e-05, "loss": 0.7286, "step": 10991 }, { "epoch": 0.62, "learning_rate": 6.557365238350524e-05, "loss": 0.8323, "step": 10992 }, { "epoch": 0.62, "learning_rate": 6.555640409460089e-05, "loss": 0.7402, "step": 10993 }, { "epoch": 0.62, "learning_rate": 6.553915696825819e-05, "loss": 0.7585, "step": 10994 }, { "epoch": 0.62, "learning_rate": 6.552191100505937e-05, "loss": 0.6712, "step": 10995 }, { "epoch": 0.62, "learning_rate": 6.550466620558643e-05, "loss": 0.6785, "step": 10996 }, { "epoch": 0.62, "learning_rate": 6.54874225704215e-05, "loss": 0.7198, "step": 10997 }, { "epoch": 0.62, "learning_rate": 6.547018010014654e-05, "loss": 0.7498, "step": 10998 }, { "epoch": 0.62, "learning_rate": 6.545293879534358e-05, "loss": 0.7536, "step": 10999 }, { "epoch": 0.62, "learning_rate": 6.543569865659453e-05, "loss": 0.6265, "step": 11000 }, { "epoch": 0.62, "learning_rate": 6.541845968448129e-05, "loss": 0.6788, "step": 11001 }, { "epoch": 0.62, "learning_rate": 6.540122187958577e-05, "loss": 0.8078, "step": 11002 }, { "epoch": 0.62, "learning_rate": 6.538398524248975e-05, "loss": 0.6696, "step": 11003 }, { "epoch": 0.62, "learning_rate": 6.536674977377496e-05, "loss": 0.7203, "step": 11004 }, { "epoch": 0.62, "learning_rate": 6.534951547402322e-05, "loss": 0.7381, "step": 11005 }, { "epoch": 0.62, "learning_rate": 6.53322823438162e-05, "loss": 0.7044, "step": 11006 }, { "epoch": 0.62, "learning_rate": 6.531505038373558e-05, "loss": 0.7894, "step": 11007 }, { "epoch": 0.62, "learning_rate": 6.529781959436294e-05, "loss": 0.6762, "step": 11008 }, { "epoch": 0.62, "learning_rate": 6.528058997627995e-05, "loss": 0.8221, "step": 11009 }, { "epoch": 0.62, "learning_rate": 6.526336153006812e-05, "loss": 0.7026, "step": 11010 }, { "epoch": 0.62, "learning_rate": 6.524613425630887e-05, "loss": 0.8166, "step": 11011 }, { "epoch": 0.62, "learning_rate": 6.522890815558378e-05, "loss": 0.6944, "step": 11012 }, { "epoch": 0.62, "learning_rate": 6.521168322847421e-05, "loss": 0.5081, "step": 11013 }, { "epoch": 0.62, "learning_rate": 6.519445947556155e-05, "loss": 0.7111, "step": 11014 }, { "epoch": 0.62, "learning_rate": 6.51772368974272e-05, "loss": 0.7626, "step": 11015 }, { "epoch": 0.62, "learning_rate": 6.51600154946524e-05, "loss": 0.7856, "step": 11016 }, { "epoch": 0.62, "learning_rate": 6.51427952678185e-05, "loss": 0.7335, "step": 11017 }, { "epoch": 0.62, "learning_rate": 6.512557621750664e-05, "loss": 0.8424, "step": 11018 }, { "epoch": 0.63, "learning_rate": 6.510835834429804e-05, "loss": 0.7182, "step": 11019 }, { "epoch": 0.63, "learning_rate": 6.509114164877385e-05, "loss": 0.7588, "step": 11020 }, { "epoch": 0.63, "learning_rate": 6.507392613151518e-05, "loss": 0.723, "step": 11021 }, { "epoch": 0.63, "learning_rate": 6.505671179310312e-05, "loss": 0.7563, "step": 11022 }, { "epoch": 0.63, "learning_rate": 6.503949863411865e-05, "loss": 0.7708, "step": 11023 }, { "epoch": 0.63, "learning_rate": 6.502228665514283e-05, "loss": 0.6558, "step": 11024 }, { "epoch": 0.63, "learning_rate": 6.500507585675655e-05, "loss": 0.6564, "step": 11025 }, { "epoch": 0.63, "learning_rate": 6.498786623954074e-05, "loss": 0.815, "step": 11026 }, { "epoch": 0.63, "learning_rate": 6.497065780407623e-05, "loss": 0.6212, "step": 11027 }, { "epoch": 0.63, "learning_rate": 6.495345055094394e-05, "loss": 0.6898, "step": 11028 }, { "epoch": 0.63, "learning_rate": 6.493624448072457e-05, "loss": 0.6723, "step": 11029 }, { "epoch": 0.63, "learning_rate": 6.49190395939989e-05, "loss": 0.6815, "step": 11030 }, { "epoch": 0.63, "learning_rate": 6.490183589134768e-05, "loss": 0.7179, "step": 11031 }, { "epoch": 0.63, "learning_rate": 6.488463337335157e-05, "loss": 0.6799, "step": 11032 }, { "epoch": 0.63, "learning_rate": 6.486743204059113e-05, "loss": 0.6623, "step": 11033 }, { "epoch": 0.63, "learning_rate": 6.485023189364699e-05, "loss": 0.8305, "step": 11034 }, { "epoch": 0.63, "learning_rate": 6.483303293309972e-05, "loss": 0.7799, "step": 11035 }, { "epoch": 0.63, "learning_rate": 6.481583515952983e-05, "loss": 0.7438, "step": 11036 }, { "epoch": 0.63, "learning_rate": 6.479863857351775e-05, "loss": 0.7387, "step": 11037 }, { "epoch": 0.63, "learning_rate": 6.478144317564396e-05, "loss": 0.5554, "step": 11038 }, { "epoch": 0.63, "learning_rate": 6.476424896648886e-05, "loss": 0.6532, "step": 11039 }, { "epoch": 0.63, "learning_rate": 6.474705594663271e-05, "loss": 0.7355, "step": 11040 }, { "epoch": 0.63, "learning_rate": 6.47298641166559e-05, "loss": 0.7891, "step": 11041 }, { "epoch": 0.63, "learning_rate": 6.471267347713868e-05, "loss": 0.7633, "step": 11042 }, { "epoch": 0.63, "learning_rate": 6.469548402866124e-05, "loss": 0.6276, "step": 11043 }, { "epoch": 0.63, "learning_rate": 6.467829577180384e-05, "loss": 0.6723, "step": 11044 }, { "epoch": 0.63, "learning_rate": 6.466110870714657e-05, "loss": 0.665, "step": 11045 }, { "epoch": 0.63, "learning_rate": 6.464392283526957e-05, "loss": 0.7238, "step": 11046 }, { "epoch": 0.63, "learning_rate": 6.462673815675293e-05, "loss": 0.7338, "step": 11047 }, { "epoch": 0.63, "learning_rate": 6.460955467217662e-05, "loss": 0.7075, "step": 11048 }, { "epoch": 0.63, "learning_rate": 6.459237238212063e-05, "loss": 0.708, "step": 11049 }, { "epoch": 0.63, "learning_rate": 6.457519128716497e-05, "loss": 0.6781, "step": 11050 }, { "epoch": 0.63, "learning_rate": 6.455801138788948e-05, "loss": 0.7276, "step": 11051 }, { "epoch": 0.63, "learning_rate": 6.454083268487407e-05, "loss": 0.7055, "step": 11052 }, { "epoch": 0.63, "learning_rate": 6.452365517869854e-05, "loss": 0.8119, "step": 11053 }, { "epoch": 0.63, "learning_rate": 6.450647886994272e-05, "loss": 0.5697, "step": 11054 }, { "epoch": 0.63, "learning_rate": 6.448930375918631e-05, "loss": 0.6137, "step": 11055 }, { "epoch": 0.63, "learning_rate": 6.447212984700902e-05, "loss": 0.6474, "step": 11056 }, { "epoch": 0.63, "learning_rate": 6.445495713399053e-05, "loss": 0.6637, "step": 11057 }, { "epoch": 0.63, "learning_rate": 6.443778562071043e-05, "loss": 0.6879, "step": 11058 }, { "epoch": 0.63, "learning_rate": 6.442061530774834e-05, "loss": 0.8028, "step": 11059 }, { "epoch": 0.63, "learning_rate": 6.440344619568382e-05, "loss": 0.7357, "step": 11060 }, { "epoch": 0.63, "learning_rate": 6.438627828509637e-05, "loss": 0.8252, "step": 11061 }, { "epoch": 0.63, "learning_rate": 6.436911157656538e-05, "loss": 0.6618, "step": 11062 }, { "epoch": 0.63, "learning_rate": 6.435194607067033e-05, "loss": 0.783, "step": 11063 }, { "epoch": 0.63, "learning_rate": 6.433478176799058e-05, "loss": 0.721, "step": 11064 }, { "epoch": 0.63, "learning_rate": 6.431761866910549e-05, "loss": 0.7622, "step": 11065 }, { "epoch": 0.63, "learning_rate": 6.430045677459434e-05, "loss": 0.7498, "step": 11066 }, { "epoch": 0.63, "learning_rate": 6.42832960850364e-05, "loss": 0.6795, "step": 11067 }, { "epoch": 0.63, "learning_rate": 6.426613660101091e-05, "loss": 0.7503, "step": 11068 }, { "epoch": 0.63, "learning_rate": 6.424897832309703e-05, "loss": 0.7647, "step": 11069 }, { "epoch": 0.63, "learning_rate": 6.423182125187385e-05, "loss": 0.7352, "step": 11070 }, { "epoch": 0.63, "learning_rate": 6.421466538792056e-05, "loss": 0.7553, "step": 11071 }, { "epoch": 0.63, "learning_rate": 6.419751073181613e-05, "loss": 0.7085, "step": 11072 }, { "epoch": 0.63, "learning_rate": 6.418035728413962e-05, "loss": 0.7143, "step": 11073 }, { "epoch": 0.63, "learning_rate": 6.416320504546997e-05, "loss": 0.5976, "step": 11074 }, { "epoch": 0.63, "learning_rate": 6.414605401638615e-05, "loss": 0.7235, "step": 11075 }, { "epoch": 0.63, "learning_rate": 6.412890419746707e-05, "loss": 0.6197, "step": 11076 }, { "epoch": 0.63, "learning_rate": 6.411175558929152e-05, "loss": 0.6935, "step": 11077 }, { "epoch": 0.63, "learning_rate": 6.409460819243833e-05, "loss": 0.6308, "step": 11078 }, { "epoch": 0.63, "learning_rate": 6.407746200748628e-05, "loss": 0.6847, "step": 11079 }, { "epoch": 0.63, "learning_rate": 6.406031703501412e-05, "loss": 0.743, "step": 11080 }, { "epoch": 0.63, "learning_rate": 6.404317327560048e-05, "loss": 0.7099, "step": 11081 }, { "epoch": 0.63, "learning_rate": 6.402603072982405e-05, "loss": 0.6447, "step": 11082 }, { "epoch": 0.63, "learning_rate": 6.400888939826345e-05, "loss": 0.6715, "step": 11083 }, { "epoch": 0.63, "learning_rate": 6.399174928149724e-05, "loss": 0.7081, "step": 11084 }, { "epoch": 0.63, "learning_rate": 6.397461038010389e-05, "loss": 0.7313, "step": 11085 }, { "epoch": 0.63, "learning_rate": 6.395747269466192e-05, "loss": 0.703, "step": 11086 }, { "epoch": 0.63, "learning_rate": 6.39403362257498e-05, "loss": 0.6134, "step": 11087 }, { "epoch": 0.63, "learning_rate": 6.392320097394586e-05, "loss": 0.7573, "step": 11088 }, { "epoch": 0.63, "learning_rate": 6.390606693982855e-05, "loss": 0.6782, "step": 11089 }, { "epoch": 0.63, "learning_rate": 6.38889341239761e-05, "loss": 0.7011, "step": 11090 }, { "epoch": 0.63, "learning_rate": 6.387180252696688e-05, "loss": 0.6988, "step": 11091 }, { "epoch": 0.63, "learning_rate": 6.385467214937902e-05, "loss": 0.6789, "step": 11092 }, { "epoch": 0.63, "learning_rate": 6.383754299179079e-05, "loss": 0.7638, "step": 11093 }, { "epoch": 0.63, "learning_rate": 6.382041505478031e-05, "loss": 0.6622, "step": 11094 }, { "epoch": 0.63, "learning_rate": 6.38032883389257e-05, "loss": 0.6927, "step": 11095 }, { "epoch": 0.63, "learning_rate": 6.378616284480508e-05, "loss": 0.6871, "step": 11096 }, { "epoch": 0.63, "learning_rate": 6.376903857299639e-05, "loss": 0.666, "step": 11097 }, { "epoch": 0.63, "learning_rate": 6.375191552407773e-05, "loss": 0.7363, "step": 11098 }, { "epoch": 0.63, "learning_rate": 6.373479369862693e-05, "loss": 0.7881, "step": 11099 }, { "epoch": 0.63, "learning_rate": 6.371767309722199e-05, "loss": 0.663, "step": 11100 }, { "epoch": 0.63, "learning_rate": 6.370055372044069e-05, "loss": 0.688, "step": 11101 }, { "epoch": 0.63, "learning_rate": 6.368343556886093e-05, "loss": 0.6865, "step": 11102 }, { "epoch": 0.63, "learning_rate": 6.366631864306045e-05, "loss": 0.6528, "step": 11103 }, { "epoch": 0.63, "learning_rate": 6.3649202943617e-05, "loss": 0.8105, "step": 11104 }, { "epoch": 0.63, "learning_rate": 6.36320884711083e-05, "loss": 0.6134, "step": 11105 }, { "epoch": 0.63, "learning_rate": 6.361497522611204e-05, "loss": 0.7093, "step": 11106 }, { "epoch": 0.63, "learning_rate": 6.359786320920573e-05, "loss": 0.6471, "step": 11107 }, { "epoch": 0.63, "learning_rate": 6.358075242096699e-05, "loss": 0.7162, "step": 11108 }, { "epoch": 0.63, "learning_rate": 6.35636428619734e-05, "loss": 0.7383, "step": 11109 }, { "epoch": 0.63, "learning_rate": 6.354653453280241e-05, "loss": 0.7807, "step": 11110 }, { "epoch": 0.63, "learning_rate": 6.352942743403147e-05, "loss": 0.7016, "step": 11111 }, { "epoch": 0.63, "learning_rate": 6.351232156623803e-05, "loss": 0.7043, "step": 11112 }, { "epoch": 0.63, "learning_rate": 6.349521692999945e-05, "loss": 0.8259, "step": 11113 }, { "epoch": 0.63, "learning_rate": 6.3478113525893e-05, "loss": 0.7276, "step": 11114 }, { "epoch": 0.63, "learning_rate": 6.346101135449598e-05, "loss": 0.6933, "step": 11115 }, { "epoch": 0.63, "learning_rate": 6.34439104163857e-05, "loss": 0.6475, "step": 11116 }, { "epoch": 0.63, "learning_rate": 6.342681071213928e-05, "loss": 0.7224, "step": 11117 }, { "epoch": 0.63, "learning_rate": 6.340971224233393e-05, "loss": 0.7768, "step": 11118 }, { "epoch": 0.63, "learning_rate": 6.339261500754674e-05, "loss": 0.6892, "step": 11119 }, { "epoch": 0.63, "learning_rate": 6.337551900835482e-05, "loss": 0.6453, "step": 11120 }, { "epoch": 0.63, "learning_rate": 6.335842424533519e-05, "loss": 0.715, "step": 11121 }, { "epoch": 0.63, "learning_rate": 6.334133071906483e-05, "loss": 0.7349, "step": 11122 }, { "epoch": 0.63, "learning_rate": 6.332423843012069e-05, "loss": 0.7139, "step": 11123 }, { "epoch": 0.63, "learning_rate": 6.330714737907968e-05, "loss": 0.6946, "step": 11124 }, { "epoch": 0.63, "learning_rate": 6.329005756651869e-05, "loss": 0.6416, "step": 11125 }, { "epoch": 0.63, "learning_rate": 6.327296899301453e-05, "loss": 0.6291, "step": 11126 }, { "epoch": 0.63, "learning_rate": 6.325588165914398e-05, "loss": 0.7181, "step": 11127 }, { "epoch": 0.63, "learning_rate": 6.323879556548382e-05, "loss": 0.661, "step": 11128 }, { "epoch": 0.63, "learning_rate": 6.322171071261071e-05, "loss": 0.8118, "step": 11129 }, { "epoch": 0.63, "learning_rate": 6.320462710110129e-05, "loss": 0.8104, "step": 11130 }, { "epoch": 0.63, "learning_rate": 6.318754473153221e-05, "loss": 0.7721, "step": 11131 }, { "epoch": 0.63, "learning_rate": 6.317046360448005e-05, "loss": 0.7161, "step": 11132 }, { "epoch": 0.63, "learning_rate": 6.315338372052134e-05, "loss": 0.7263, "step": 11133 }, { "epoch": 0.63, "learning_rate": 6.313630508023253e-05, "loss": 0.744, "step": 11134 }, { "epoch": 0.63, "learning_rate": 6.311922768419014e-05, "loss": 0.6601, "step": 11135 }, { "epoch": 0.63, "learning_rate": 6.310215153297055e-05, "loss": 0.6698, "step": 11136 }, { "epoch": 0.63, "learning_rate": 6.308507662715007e-05, "loss": 0.7024, "step": 11137 }, { "epoch": 0.63, "learning_rate": 6.306800296730511e-05, "loss": 0.6009, "step": 11138 }, { "epoch": 0.63, "learning_rate": 6.305093055401186e-05, "loss": 0.6718, "step": 11139 }, { "epoch": 0.63, "learning_rate": 6.303385938784663e-05, "loss": 0.675, "step": 11140 }, { "epoch": 0.63, "learning_rate": 6.301678946938558e-05, "loss": 0.7108, "step": 11141 }, { "epoch": 0.63, "learning_rate": 6.299972079920488e-05, "loss": 0.7525, "step": 11142 }, { "epoch": 0.63, "learning_rate": 6.298265337788069e-05, "loss": 0.7089, "step": 11143 }, { "epoch": 0.63, "learning_rate": 6.296558720598897e-05, "loss": 0.7438, "step": 11144 }, { "epoch": 0.63, "learning_rate": 6.294852228410585e-05, "loss": 0.7343, "step": 11145 }, { "epoch": 0.63, "learning_rate": 6.293145861280723e-05, "loss": 0.6866, "step": 11146 }, { "epoch": 0.63, "learning_rate": 6.291439619266913e-05, "loss": 0.7383, "step": 11147 }, { "epoch": 0.63, "learning_rate": 6.28973350242674e-05, "loss": 0.6651, "step": 11148 }, { "epoch": 0.63, "learning_rate": 6.28802751081779e-05, "loss": 0.6846, "step": 11149 }, { "epoch": 0.63, "learning_rate": 6.286321644497655e-05, "loss": 0.6847, "step": 11150 }, { "epoch": 0.63, "learning_rate": 6.284615903523899e-05, "loss": 0.6432, "step": 11151 }, { "epoch": 0.63, "learning_rate": 6.282910287954098e-05, "loss": 0.7948, "step": 11152 }, { "epoch": 0.63, "learning_rate": 6.281204797845824e-05, "loss": 0.771, "step": 11153 }, { "epoch": 0.63, "learning_rate": 6.279499433256642e-05, "loss": 0.7057, "step": 11154 }, { "epoch": 0.63, "learning_rate": 6.277794194244109e-05, "loss": 0.7041, "step": 11155 }, { "epoch": 0.63, "learning_rate": 6.276089080865785e-05, "loss": 0.7349, "step": 11156 }, { "epoch": 0.63, "learning_rate": 6.274384093179222e-05, "loss": 0.7844, "step": 11157 }, { "epoch": 0.63, "learning_rate": 6.27267923124197e-05, "loss": 0.6995, "step": 11158 }, { "epoch": 0.63, "learning_rate": 6.270974495111564e-05, "loss": 0.7344, "step": 11159 }, { "epoch": 0.63, "learning_rate": 6.26926988484555e-05, "loss": 0.6894, "step": 11160 }, { "epoch": 0.63, "learning_rate": 6.267565400501462e-05, "loss": 0.6557, "step": 11161 }, { "epoch": 0.63, "learning_rate": 6.265861042136829e-05, "loss": 0.6995, "step": 11162 }, { "epoch": 0.63, "learning_rate": 6.26415680980918e-05, "loss": 0.836, "step": 11163 }, { "epoch": 0.63, "learning_rate": 6.262452703576036e-05, "loss": 0.8175, "step": 11164 }, { "epoch": 0.63, "learning_rate": 6.260748723494918e-05, "loss": 0.7, "step": 11165 }, { "epoch": 0.63, "learning_rate": 6.259044869623333e-05, "loss": 0.779, "step": 11166 }, { "epoch": 0.63, "learning_rate": 6.257341142018798e-05, "loss": 0.7484, "step": 11167 }, { "epoch": 0.63, "learning_rate": 6.255637540738812e-05, "loss": 0.6959, "step": 11168 }, { "epoch": 0.63, "learning_rate": 6.25393406584088e-05, "loss": 0.7254, "step": 11169 }, { "epoch": 0.63, "learning_rate": 6.252230717382499e-05, "loss": 0.669, "step": 11170 }, { "epoch": 0.63, "learning_rate": 6.250527495421158e-05, "loss": 0.7563, "step": 11171 }, { "epoch": 0.63, "learning_rate": 6.248824400014347e-05, "loss": 0.7993, "step": 11172 }, { "epoch": 0.63, "learning_rate": 6.247121431219555e-05, "loss": 0.7322, "step": 11173 }, { "epoch": 0.63, "learning_rate": 6.245418589094254e-05, "loss": 0.6472, "step": 11174 }, { "epoch": 0.63, "learning_rate": 6.243715873695922e-05, "loss": 0.7163, "step": 11175 }, { "epoch": 0.63, "learning_rate": 6.242013285082029e-05, "loss": 0.6843, "step": 11176 }, { "epoch": 0.63, "learning_rate": 6.240310823310047e-05, "loss": 0.8238, "step": 11177 }, { "epoch": 0.63, "learning_rate": 6.23860848843743e-05, "loss": 0.6818, "step": 11178 }, { "epoch": 0.63, "learning_rate": 6.236906280521646e-05, "loss": 0.734, "step": 11179 }, { "epoch": 0.63, "learning_rate": 6.235204199620143e-05, "loss": 0.7083, "step": 11180 }, { "epoch": 0.63, "learning_rate": 6.233502245790372e-05, "loss": 0.7785, "step": 11181 }, { "epoch": 0.63, "learning_rate": 6.231800419089775e-05, "loss": 0.6888, "step": 11182 }, { "epoch": 0.63, "learning_rate": 6.2300987195758e-05, "loss": 0.727, "step": 11183 }, { "epoch": 0.63, "learning_rate": 6.228397147305878e-05, "loss": 0.8012, "step": 11184 }, { "epoch": 0.63, "learning_rate": 6.226695702337442e-05, "loss": 0.6765, "step": 11185 }, { "epoch": 0.63, "learning_rate": 6.224994384727926e-05, "loss": 0.7519, "step": 11186 }, { "epoch": 0.63, "learning_rate": 6.223293194534751e-05, "loss": 0.6397, "step": 11187 }, { "epoch": 0.63, "learning_rate": 6.22159213181533e-05, "loss": 0.6871, "step": 11188 }, { "epoch": 0.63, "learning_rate": 6.219891196627085e-05, "loss": 0.6873, "step": 11189 }, { "epoch": 0.63, "learning_rate": 6.21819038902743e-05, "loss": 0.6853, "step": 11190 }, { "epoch": 0.63, "learning_rate": 6.216489709073763e-05, "loss": 0.7848, "step": 11191 }, { "epoch": 0.63, "learning_rate": 6.21478915682349e-05, "loss": 0.6029, "step": 11192 }, { "epoch": 0.63, "learning_rate": 6.213088732334013e-05, "loss": 0.6666, "step": 11193 }, { "epoch": 0.63, "learning_rate": 6.211388435662721e-05, "loss": 0.7103, "step": 11194 }, { "epoch": 0.64, "learning_rate": 6.209688266867011e-05, "loss": 0.6763, "step": 11195 }, { "epoch": 0.64, "learning_rate": 6.20798822600426e-05, "loss": 0.6723, "step": 11196 }, { "epoch": 0.64, "learning_rate": 6.206288313131847e-05, "loss": 0.723, "step": 11197 }, { "epoch": 0.64, "learning_rate": 6.204588528307153e-05, "loss": 0.762, "step": 11198 }, { "epoch": 0.64, "learning_rate": 6.202888871587555e-05, "loss": 0.7521, "step": 11199 }, { "epoch": 0.64, "learning_rate": 6.201189343030412e-05, "loss": 0.7022, "step": 11200 }, { "epoch": 0.64, "learning_rate": 6.199489942693091e-05, "loss": 0.7932, "step": 11201 }, { "epoch": 0.64, "learning_rate": 6.197790670632956e-05, "loss": 0.7493, "step": 11202 }, { "epoch": 0.64, "learning_rate": 6.196091526907355e-05, "loss": 0.6053, "step": 11203 }, { "epoch": 0.64, "learning_rate": 6.19439251157364e-05, "loss": 0.7206, "step": 11204 }, { "epoch": 0.64, "learning_rate": 6.192693624689158e-05, "loss": 0.7401, "step": 11205 }, { "epoch": 0.64, "learning_rate": 6.190994866311252e-05, "loss": 0.6678, "step": 11206 }, { "epoch": 0.64, "learning_rate": 6.18929623649726e-05, "loss": 0.7379, "step": 11207 }, { "epoch": 0.64, "learning_rate": 6.187597735304513e-05, "loss": 0.6804, "step": 11208 }, { "epoch": 0.64, "learning_rate": 6.185899362790339e-05, "loss": 0.6846, "step": 11209 }, { "epoch": 0.64, "learning_rate": 6.18420111901207e-05, "loss": 0.7715, "step": 11210 }, { "epoch": 0.64, "learning_rate": 6.182503004027014e-05, "loss": 0.7117, "step": 11211 }, { "epoch": 0.64, "learning_rate": 6.180805017892498e-05, "loss": 0.706, "step": 11212 }, { "epoch": 0.64, "learning_rate": 6.179107160665826e-05, "loss": 0.6881, "step": 11213 }, { "epoch": 0.64, "learning_rate": 6.177409432404308e-05, "loss": 0.7497, "step": 11214 }, { "epoch": 0.64, "learning_rate": 6.175711833165247e-05, "loss": 0.6863, "step": 11215 }, { "epoch": 0.64, "learning_rate": 6.174014363005943e-05, "loss": 0.7011, "step": 11216 }, { "epoch": 0.64, "learning_rate": 6.172317021983689e-05, "loss": 0.7283, "step": 11217 }, { "epoch": 0.64, "learning_rate": 6.170619810155771e-05, "loss": 0.662, "step": 11218 }, { "epoch": 0.64, "learning_rate": 6.168922727579478e-05, "loss": 0.7408, "step": 11219 }, { "epoch": 0.64, "learning_rate": 6.167225774312092e-05, "loss": 0.7548, "step": 11220 }, { "epoch": 0.64, "learning_rate": 6.165528950410884e-05, "loss": 0.6929, "step": 11221 }, { "epoch": 0.64, "learning_rate": 6.163832255933133e-05, "loss": 0.755, "step": 11222 }, { "epoch": 0.64, "learning_rate": 6.162135690936103e-05, "loss": 0.7231, "step": 11223 }, { "epoch": 0.64, "learning_rate": 6.160439255477063e-05, "loss": 0.759, "step": 11224 }, { "epoch": 0.64, "learning_rate": 6.158742949613263e-05, "loss": 0.7186, "step": 11225 }, { "epoch": 0.64, "learning_rate": 6.157046773401964e-05, "loss": 0.6542, "step": 11226 }, { "epoch": 0.64, "learning_rate": 6.155350726900414e-05, "loss": 0.7438, "step": 11227 }, { "epoch": 0.64, "learning_rate": 6.153654810165861e-05, "loss": 0.7265, "step": 11228 }, { "epoch": 0.64, "learning_rate": 6.151959023255545e-05, "loss": 0.6977, "step": 11229 }, { "epoch": 0.64, "learning_rate": 6.150263366226703e-05, "loss": 0.681, "step": 11230 }, { "epoch": 0.64, "learning_rate": 6.148567839136569e-05, "loss": 0.7679, "step": 11231 }, { "epoch": 0.64, "learning_rate": 6.146872442042377e-05, "loss": 0.6941, "step": 11232 }, { "epoch": 0.64, "learning_rate": 6.145177175001337e-05, "loss": 0.6595, "step": 11233 }, { "epoch": 0.64, "learning_rate": 6.14348203807068e-05, "loss": 0.6946, "step": 11234 }, { "epoch": 0.64, "learning_rate": 6.14178703130762e-05, "loss": 0.7667, "step": 11235 }, { "epoch": 0.64, "learning_rate": 6.140092154769364e-05, "loss": 0.7358, "step": 11236 }, { "epoch": 0.64, "learning_rate": 6.138397408513119e-05, "loss": 0.7791, "step": 11237 }, { "epoch": 0.64, "learning_rate": 6.136702792596091e-05, "loss": 0.6985, "step": 11238 }, { "epoch": 0.64, "learning_rate": 6.135008307075481e-05, "loss": 0.6668, "step": 11239 }, { "epoch": 0.64, "learning_rate": 6.13331395200847e-05, "loss": 0.7605, "step": 11240 }, { "epoch": 0.64, "learning_rate": 6.131619727452257e-05, "loss": 0.7328, "step": 11241 }, { "epoch": 0.64, "learning_rate": 6.12992563346402e-05, "loss": 0.7428, "step": 11242 }, { "epoch": 0.64, "learning_rate": 6.128231670100944e-05, "loss": 0.6786, "step": 11243 }, { "epoch": 0.64, "learning_rate": 6.126537837420205e-05, "loss": 0.7861, "step": 11244 }, { "epoch": 0.64, "learning_rate": 6.12484413547897e-05, "loss": 0.7054, "step": 11245 }, { "epoch": 0.64, "learning_rate": 6.123150564334409e-05, "loss": 0.6742, "step": 11246 }, { "epoch": 0.64, "learning_rate": 6.121457124043688e-05, "loss": 0.6257, "step": 11247 }, { "epoch": 0.64, "learning_rate": 6.119763814663958e-05, "loss": 0.7425, "step": 11248 }, { "epoch": 0.64, "learning_rate": 6.118070636252374e-05, "loss": 0.6747, "step": 11249 }, { "epoch": 0.64, "learning_rate": 6.116377588866087e-05, "loss": 0.706, "step": 11250 }, { "epoch": 0.64, "learning_rate": 6.114684672562244e-05, "loss": 0.7587, "step": 11251 }, { "epoch": 0.64, "learning_rate": 6.112991887397978e-05, "loss": 0.7883, "step": 11252 }, { "epoch": 0.64, "learning_rate": 6.111299233430435e-05, "loss": 0.6467, "step": 11253 }, { "epoch": 0.64, "learning_rate": 6.109606710716741e-05, "loss": 0.6632, "step": 11254 }, { "epoch": 0.64, "learning_rate": 6.107914319314024e-05, "loss": 0.7429, "step": 11255 }, { "epoch": 0.64, "learning_rate": 6.106222059279402e-05, "loss": 0.7374, "step": 11256 }, { "epoch": 0.64, "learning_rate": 6.10452993067e-05, "loss": 0.6791, "step": 11257 }, { "epoch": 0.64, "learning_rate": 6.1028379335429265e-05, "loss": 0.5925, "step": 11258 }, { "epoch": 0.64, "learning_rate": 6.1011460679552945e-05, "loss": 0.835, "step": 11259 }, { "epoch": 0.64, "learning_rate": 6.09945433396421e-05, "loss": 0.689, "step": 11260 }, { "epoch": 0.64, "learning_rate": 6.097762731626773e-05, "loss": 0.6641, "step": 11261 }, { "epoch": 0.64, "learning_rate": 6.096071261000074e-05, "loss": 0.776, "step": 11262 }, { "epoch": 0.64, "learning_rate": 6.0943799221412076e-05, "loss": 0.6515, "step": 11263 }, { "epoch": 0.64, "learning_rate": 6.092688715107264e-05, "loss": 0.7531, "step": 11264 }, { "epoch": 0.64, "learning_rate": 6.0909976399553226e-05, "loss": 0.8707, "step": 11265 }, { "epoch": 0.64, "learning_rate": 6.089306696742461e-05, "loss": 0.6734, "step": 11266 }, { "epoch": 0.64, "learning_rate": 6.087615885525757e-05, "loss": 0.7081, "step": 11267 }, { "epoch": 0.64, "learning_rate": 6.085925206362276e-05, "loss": 0.6955, "step": 11268 }, { "epoch": 0.64, "learning_rate": 6.084234659309088e-05, "loss": 0.6461, "step": 11269 }, { "epoch": 0.64, "learning_rate": 6.0825442444232464e-05, "loss": 0.7332, "step": 11270 }, { "epoch": 0.64, "learning_rate": 6.080853961761812e-05, "loss": 0.7494, "step": 11271 }, { "epoch": 0.64, "learning_rate": 6.079163811381832e-05, "loss": 0.6682, "step": 11272 }, { "epoch": 0.64, "learning_rate": 6.077473793340358e-05, "loss": 0.6697, "step": 11273 }, { "epoch": 0.64, "learning_rate": 6.07578390769443e-05, "loss": 0.6843, "step": 11274 }, { "epoch": 0.64, "learning_rate": 6.074094154501087e-05, "loss": 0.7848, "step": 11275 }, { "epoch": 0.64, "learning_rate": 6.072404533817364e-05, "loss": 0.6659, "step": 11276 }, { "epoch": 0.64, "learning_rate": 6.070715045700288e-05, "loss": 0.7558, "step": 11277 }, { "epoch": 0.64, "learning_rate": 6.069025690206882e-05, "loss": 0.7445, "step": 11278 }, { "epoch": 0.64, "learning_rate": 6.067336467394169e-05, "loss": 0.7495, "step": 11279 }, { "epoch": 0.64, "learning_rate": 6.0656473773191655e-05, "loss": 0.6203, "step": 11280 }, { "epoch": 0.64, "learning_rate": 6.06395842003888e-05, "loss": 0.6239, "step": 11281 }, { "epoch": 0.64, "learning_rate": 6.0622695956103196e-05, "loss": 0.7226, "step": 11282 }, { "epoch": 0.64, "learning_rate": 6.0605809040904894e-05, "loss": 0.6276, "step": 11283 }, { "epoch": 0.64, "learning_rate": 6.0588923455363864e-05, "loss": 0.6362, "step": 11284 }, { "epoch": 0.64, "learning_rate": 6.057203920004999e-05, "loss": 0.7708, "step": 11285 }, { "epoch": 0.64, "learning_rate": 6.055515627553323e-05, "loss": 0.71, "step": 11286 }, { "epoch": 0.64, "learning_rate": 6.0538274682383354e-05, "loss": 0.6694, "step": 11287 }, { "epoch": 0.64, "learning_rate": 6.0521394421170216e-05, "loss": 0.7345, "step": 11288 }, { "epoch": 0.64, "learning_rate": 6.0504515492463564e-05, "loss": 0.7009, "step": 11289 }, { "epoch": 0.64, "learning_rate": 6.0487637896833085e-05, "loss": 0.7498, "step": 11290 }, { "epoch": 0.64, "learning_rate": 6.047076163484848e-05, "loss": 0.7386, "step": 11291 }, { "epoch": 0.64, "learning_rate": 6.045388670707931e-05, "loss": 0.6211, "step": 11292 }, { "epoch": 0.64, "learning_rate": 6.0437013114095195e-05, "loss": 0.6872, "step": 11293 }, { "epoch": 0.64, "learning_rate": 6.042014085646561e-05, "loss": 0.6634, "step": 11294 }, { "epoch": 0.64, "learning_rate": 6.040326993476009e-05, "loss": 0.7297, "step": 11295 }, { "epoch": 0.64, "learning_rate": 6.0386400349548076e-05, "loss": 0.7525, "step": 11296 }, { "epoch": 0.64, "learning_rate": 6.036953210139891e-05, "loss": 0.6976, "step": 11297 }, { "epoch": 0.64, "learning_rate": 6.035266519088202e-05, "loss": 0.7494, "step": 11298 }, { "epoch": 0.64, "learning_rate": 6.0335799618566616e-05, "loss": 0.7454, "step": 11299 }, { "epoch": 0.64, "learning_rate": 6.0318935385022016e-05, "loss": 0.6028, "step": 11300 }, { "epoch": 0.64, "learning_rate": 6.0302072490817385e-05, "loss": 0.7863, "step": 11301 }, { "epoch": 0.64, "learning_rate": 6.0285210936521955e-05, "loss": 0.6553, "step": 11302 }, { "epoch": 0.64, "learning_rate": 6.026835072270478e-05, "loss": 0.7683, "step": 11303 }, { "epoch": 0.64, "learning_rate": 6.025149184993498e-05, "loss": 0.6711, "step": 11304 }, { "epoch": 0.64, "learning_rate": 6.023463431878159e-05, "loss": 0.6905, "step": 11305 }, { "epoch": 0.64, "learning_rate": 6.02177781298136e-05, "loss": 0.6637, "step": 11306 }, { "epoch": 0.64, "learning_rate": 6.020092328359988e-05, "loss": 0.7607, "step": 11307 }, { "epoch": 0.64, "learning_rate": 6.018406978070939e-05, "loss": 0.7359, "step": 11308 }, { "epoch": 0.64, "learning_rate": 6.016721762171098e-05, "loss": 0.7642, "step": 11309 }, { "epoch": 0.64, "learning_rate": 6.015036680717343e-05, "loss": 0.6897, "step": 11310 }, { "epoch": 0.64, "learning_rate": 6.01335173376655e-05, "loss": 0.6681, "step": 11311 }, { "epoch": 0.64, "learning_rate": 6.0116669213755935e-05, "loss": 0.7137, "step": 11312 }, { "epoch": 0.64, "learning_rate": 6.00998224360134e-05, "loss": 0.6919, "step": 11313 }, { "epoch": 0.64, "learning_rate": 6.008297700500646e-05, "loss": 0.6766, "step": 11314 }, { "epoch": 0.64, "learning_rate": 6.006613292130373e-05, "loss": 0.726, "step": 11315 }, { "epoch": 0.64, "learning_rate": 6.0049290185473775e-05, "loss": 0.7493, "step": 11316 }, { "epoch": 0.64, "learning_rate": 6.003244879808503e-05, "loss": 0.6631, "step": 11317 }, { "epoch": 0.64, "learning_rate": 6.0015608759705955e-05, "loss": 0.7821, "step": 11318 }, { "epoch": 0.64, "learning_rate": 5.9998770070904955e-05, "loss": 0.756, "step": 11319 }, { "epoch": 0.64, "learning_rate": 5.998193273225036e-05, "loss": 0.7782, "step": 11320 }, { "epoch": 0.64, "learning_rate": 5.9965096744310526e-05, "loss": 0.6914, "step": 11321 }, { "epoch": 0.64, "learning_rate": 5.994826210765365e-05, "loss": 0.7512, "step": 11322 }, { "epoch": 0.64, "learning_rate": 5.9931428822847944e-05, "loss": 0.6229, "step": 11323 }, { "epoch": 0.64, "learning_rate": 5.9914596890461614e-05, "loss": 0.6939, "step": 11324 }, { "epoch": 0.64, "learning_rate": 5.9897766311062775e-05, "loss": 0.6239, "step": 11325 }, { "epoch": 0.64, "learning_rate": 5.988093708521948e-05, "loss": 0.6865, "step": 11326 }, { "epoch": 0.64, "learning_rate": 5.986410921349976e-05, "loss": 0.6873, "step": 11327 }, { "epoch": 0.64, "learning_rate": 5.9847282696471685e-05, "loss": 0.7218, "step": 11328 }, { "epoch": 0.64, "learning_rate": 5.983045753470308e-05, "loss": 0.7361, "step": 11329 }, { "epoch": 0.64, "learning_rate": 5.981363372876186e-05, "loss": 0.6414, "step": 11330 }, { "epoch": 0.64, "learning_rate": 5.97968112792159e-05, "loss": 0.6583, "step": 11331 }, { "epoch": 0.64, "learning_rate": 5.977999018663303e-05, "loss": 0.699, "step": 11332 }, { "epoch": 0.64, "learning_rate": 5.976317045158093e-05, "loss": 0.6609, "step": 11333 }, { "epoch": 0.64, "learning_rate": 5.974635207462739e-05, "loss": 0.7486, "step": 11334 }, { "epoch": 0.64, "learning_rate": 5.9729535056340004e-05, "loss": 0.7516, "step": 11335 }, { "epoch": 0.64, "learning_rate": 5.9712719397286466e-05, "loss": 0.805, "step": 11336 }, { "epoch": 0.64, "learning_rate": 5.9695905098034264e-05, "loss": 0.7009, "step": 11337 }, { "epoch": 0.64, "learning_rate": 5.967909215915099e-05, "loss": 0.7499, "step": 11338 }, { "epoch": 0.64, "learning_rate": 5.966228058120409e-05, "loss": 0.7891, "step": 11339 }, { "epoch": 0.64, "learning_rate": 5.964547036476099e-05, "loss": 0.6916, "step": 11340 }, { "epoch": 0.64, "learning_rate": 5.962866151038913e-05, "loss": 0.7047, "step": 11341 }, { "epoch": 0.64, "learning_rate": 5.96118540186558e-05, "loss": 0.7191, "step": 11342 }, { "epoch": 0.64, "learning_rate": 5.959504789012836e-05, "loss": 0.7055, "step": 11343 }, { "epoch": 0.64, "learning_rate": 5.957824312537398e-05, "loss": 0.6879, "step": 11344 }, { "epoch": 0.64, "learning_rate": 5.956143972495992e-05, "loss": 0.6707, "step": 11345 }, { "epoch": 0.64, "learning_rate": 5.954463768945331e-05, "loss": 0.634, "step": 11346 }, { "epoch": 0.64, "learning_rate": 5.95278370194213e-05, "loss": 0.6874, "step": 11347 }, { "epoch": 0.64, "learning_rate": 5.9511037715430896e-05, "loss": 0.679, "step": 11348 }, { "epoch": 0.64, "learning_rate": 5.9494239778049165e-05, "loss": 0.6359, "step": 11349 }, { "epoch": 0.64, "learning_rate": 5.947744320784312e-05, "loss": 0.7364, "step": 11350 }, { "epoch": 0.64, "learning_rate": 5.9460648005379604e-05, "loss": 0.7019, "step": 11351 }, { "epoch": 0.64, "learning_rate": 5.944385417122553e-05, "loss": 0.6297, "step": 11352 }, { "epoch": 0.64, "learning_rate": 5.942706170594774e-05, "loss": 0.6912, "step": 11353 }, { "epoch": 0.64, "learning_rate": 5.941027061011303e-05, "loss": 0.7248, "step": 11354 }, { "epoch": 0.64, "learning_rate": 5.939348088428813e-05, "loss": 0.6108, "step": 11355 }, { "epoch": 0.64, "learning_rate": 5.937669252903973e-05, "loss": 0.7076, "step": 11356 }, { "epoch": 0.64, "learning_rate": 5.9359905544934534e-05, "loss": 0.8036, "step": 11357 }, { "epoch": 0.64, "learning_rate": 5.934311993253912e-05, "loss": 0.7724, "step": 11358 }, { "epoch": 0.64, "learning_rate": 5.9326335692419995e-05, "loss": 0.6795, "step": 11359 }, { "epoch": 0.64, "learning_rate": 5.930955282514371e-05, "loss": 0.6745, "step": 11360 }, { "epoch": 0.64, "learning_rate": 5.929277133127675e-05, "loss": 0.678, "step": 11361 }, { "epoch": 0.64, "learning_rate": 5.92759912113855e-05, "loss": 0.8047, "step": 11362 }, { "epoch": 0.64, "learning_rate": 5.925921246603637e-05, "loss": 0.7139, "step": 11363 }, { "epoch": 0.64, "learning_rate": 5.9242435095795655e-05, "loss": 0.7179, "step": 11364 }, { "epoch": 0.64, "learning_rate": 5.922565910122967e-05, "loss": 0.7266, "step": 11365 }, { "epoch": 0.64, "learning_rate": 5.9208884482904594e-05, "loss": 0.7384, "step": 11366 }, { "epoch": 0.64, "learning_rate": 5.919211124138666e-05, "loss": 0.6893, "step": 11367 }, { "epoch": 0.64, "learning_rate": 5.9175339377241975e-05, "loss": 0.7068, "step": 11368 }, { "epoch": 0.64, "learning_rate": 5.915856889103666e-05, "loss": 0.6685, "step": 11369 }, { "epoch": 0.64, "learning_rate": 5.914179978333676e-05, "loss": 0.7023, "step": 11370 }, { "epoch": 0.65, "learning_rate": 5.912503205470826e-05, "loss": 0.734, "step": 11371 }, { "epoch": 0.65, "learning_rate": 5.9108265705717123e-05, "loss": 0.7067, "step": 11372 }, { "epoch": 0.65, "learning_rate": 5.909150073692932e-05, "loss": 0.7982, "step": 11373 }, { "epoch": 0.65, "learning_rate": 5.907473714891061e-05, "loss": 0.746, "step": 11374 }, { "epoch": 0.65, "learning_rate": 5.9057974942226845e-05, "loss": 0.7121, "step": 11375 }, { "epoch": 0.65, "learning_rate": 5.90412141174438e-05, "loss": 0.7173, "step": 11376 }, { "epoch": 0.65, "learning_rate": 5.902445467512721e-05, "loss": 0.6581, "step": 11377 }, { "epoch": 0.65, "learning_rate": 5.900769661584272e-05, "loss": 0.6886, "step": 11378 }, { "epoch": 0.65, "learning_rate": 5.8990939940156e-05, "loss": 0.7225, "step": 11379 }, { "epoch": 0.65, "learning_rate": 5.897418464863262e-05, "loss": 0.7536, "step": 11380 }, { "epoch": 0.65, "learning_rate": 5.895743074183806e-05, "loss": 0.6755, "step": 11381 }, { "epoch": 0.65, "learning_rate": 5.8940678220337863e-05, "loss": 0.7442, "step": 11382 }, { "epoch": 0.65, "learning_rate": 5.8923927084697475e-05, "loss": 0.7073, "step": 11383 }, { "epoch": 0.65, "learning_rate": 5.890717733548225e-05, "loss": 0.7488, "step": 11384 }, { "epoch": 0.65, "learning_rate": 5.889042897325755e-05, "loss": 0.6845, "step": 11385 }, { "epoch": 0.65, "learning_rate": 5.887368199858872e-05, "loss": 0.7563, "step": 11386 }, { "epoch": 0.65, "learning_rate": 5.8856936412040994e-05, "loss": 0.7409, "step": 11387 }, { "epoch": 0.65, "learning_rate": 5.884019221417952e-05, "loss": 0.7292, "step": 11388 }, { "epoch": 0.65, "learning_rate": 5.8823449405569516e-05, "loss": 0.7752, "step": 11389 }, { "epoch": 0.65, "learning_rate": 5.88067079867761e-05, "loss": 0.6978, "step": 11390 }, { "epoch": 0.65, "learning_rate": 5.878996795836431e-05, "loss": 0.7493, "step": 11391 }, { "epoch": 0.65, "learning_rate": 5.877322932089919e-05, "loss": 0.8027, "step": 11392 }, { "epoch": 0.65, "learning_rate": 5.875649207494568e-05, "loss": 0.7312, "step": 11393 }, { "epoch": 0.65, "learning_rate": 5.873975622106873e-05, "loss": 0.7421, "step": 11394 }, { "epoch": 0.65, "learning_rate": 5.872302175983326e-05, "loss": 0.7268, "step": 11395 }, { "epoch": 0.65, "learning_rate": 5.8706288691804034e-05, "loss": 0.7683, "step": 11396 }, { "epoch": 0.65, "learning_rate": 5.868955701754584e-05, "loss": 0.7619, "step": 11397 }, { "epoch": 0.65, "learning_rate": 5.8672826737623446e-05, "loss": 0.7616, "step": 11398 }, { "epoch": 0.65, "learning_rate": 5.865609785260153e-05, "loss": 0.6964, "step": 11399 }, { "epoch": 0.65, "learning_rate": 5.863937036304474e-05, "loss": 0.6636, "step": 11400 }, { "epoch": 0.65, "learning_rate": 5.862264426951768e-05, "loss": 0.7083, "step": 11401 }, { "epoch": 0.65, "learning_rate": 5.860591957258492e-05, "loss": 0.6733, "step": 11402 }, { "epoch": 0.65, "learning_rate": 5.858919627281092e-05, "loss": 0.6533, "step": 11403 }, { "epoch": 0.65, "learning_rate": 5.857247437076012e-05, "loss": 0.7175, "step": 11404 }, { "epoch": 0.65, "learning_rate": 5.855575386699696e-05, "loss": 0.6915, "step": 11405 }, { "epoch": 0.65, "learning_rate": 5.853903476208583e-05, "loss": 0.7707, "step": 11406 }, { "epoch": 0.65, "learning_rate": 5.852231705659099e-05, "loss": 0.7031, "step": 11407 }, { "epoch": 0.65, "learning_rate": 5.8505600751076736e-05, "loss": 0.6845, "step": 11408 }, { "epoch": 0.65, "learning_rate": 5.848888584610727e-05, "loss": 0.6492, "step": 11409 }, { "epoch": 0.65, "learning_rate": 5.847217234224681e-05, "loss": 0.606, "step": 11410 }, { "epoch": 0.65, "learning_rate": 5.845546024005941e-05, "loss": 0.7349, "step": 11411 }, { "epoch": 0.65, "learning_rate": 5.84387495401092e-05, "loss": 0.6372, "step": 11412 }, { "epoch": 0.65, "learning_rate": 5.8422040242960166e-05, "loss": 0.7371, "step": 11413 }, { "epoch": 0.65, "learning_rate": 5.840533234917631e-05, "loss": 0.6552, "step": 11414 }, { "epoch": 0.65, "learning_rate": 5.83886258593216e-05, "loss": 0.6922, "step": 11415 }, { "epoch": 0.65, "learning_rate": 5.83719207739599e-05, "loss": 0.7305, "step": 11416 }, { "epoch": 0.65, "learning_rate": 5.835521709365506e-05, "loss": 0.7945, "step": 11417 }, { "epoch": 0.65, "learning_rate": 5.8338514818970836e-05, "loss": 0.6203, "step": 11418 }, { "epoch": 0.65, "learning_rate": 5.832181395047098e-05, "loss": 0.7261, "step": 11419 }, { "epoch": 0.65, "learning_rate": 5.830511448871922e-05, "loss": 0.7614, "step": 11420 }, { "epoch": 0.65, "learning_rate": 5.828841643427918e-05, "loss": 0.6834, "step": 11421 }, { "epoch": 0.65, "learning_rate": 5.8271719787714497e-05, "loss": 0.7682, "step": 11422 }, { "epoch": 0.65, "learning_rate": 5.825502454958873e-05, "loss": 0.6289, "step": 11423 }, { "epoch": 0.65, "learning_rate": 5.8238330720465384e-05, "loss": 0.569, "step": 11424 }, { "epoch": 0.65, "learning_rate": 5.8221638300907854e-05, "loss": 0.6448, "step": 11425 }, { "epoch": 0.65, "learning_rate": 5.8204947291479586e-05, "loss": 0.7033, "step": 11426 }, { "epoch": 0.65, "learning_rate": 5.8188257692743974e-05, "loss": 0.687, "step": 11427 }, { "epoch": 0.65, "learning_rate": 5.8171569505264314e-05, "loss": 0.5965, "step": 11428 }, { "epoch": 0.65, "learning_rate": 5.8154882729603876e-05, "loss": 0.6687, "step": 11429 }, { "epoch": 0.65, "learning_rate": 5.8138197366325944e-05, "loss": 0.7872, "step": 11430 }, { "epoch": 0.65, "learning_rate": 5.812151341599359e-05, "loss": 0.6774, "step": 11431 }, { "epoch": 0.65, "learning_rate": 5.810483087917002e-05, "loss": 0.6495, "step": 11432 }, { "epoch": 0.65, "learning_rate": 5.8088149756418255e-05, "loss": 0.6819, "step": 11433 }, { "epoch": 0.65, "learning_rate": 5.807147004830135e-05, "loss": 0.7081, "step": 11434 }, { "epoch": 0.65, "learning_rate": 5.805479175538229e-05, "loss": 0.7998, "step": 11435 }, { "epoch": 0.65, "learning_rate": 5.803811487822405e-05, "loss": 0.7586, "step": 11436 }, { "epoch": 0.65, "learning_rate": 5.8021439417389444e-05, "loss": 0.7382, "step": 11437 }, { "epoch": 0.65, "learning_rate": 5.800476537344135e-05, "loss": 0.736, "step": 11438 }, { "epoch": 0.65, "learning_rate": 5.7988092746942614e-05, "loss": 0.7816, "step": 11439 }, { "epoch": 0.65, "learning_rate": 5.797142153845587e-05, "loss": 0.8072, "step": 11440 }, { "epoch": 0.65, "learning_rate": 5.795475174854388e-05, "loss": 0.7172, "step": 11441 }, { "epoch": 0.65, "learning_rate": 5.7938083377769295e-05, "loss": 0.7578, "step": 11442 }, { "epoch": 0.65, "learning_rate": 5.792141642669473e-05, "loss": 0.7482, "step": 11443 }, { "epoch": 0.65, "learning_rate": 5.7904750895882684e-05, "loss": 0.6907, "step": 11444 }, { "epoch": 0.65, "learning_rate": 5.78880867858957e-05, "loss": 0.6191, "step": 11445 }, { "epoch": 0.65, "learning_rate": 5.7871424097296224e-05, "loss": 0.6972, "step": 11446 }, { "epoch": 0.65, "learning_rate": 5.7854762830646705e-05, "loss": 0.7393, "step": 11447 }, { "epoch": 0.65, "learning_rate": 5.783810298650945e-05, "loss": 0.7555, "step": 11448 }, { "epoch": 0.65, "learning_rate": 5.78214445654468e-05, "loss": 0.6883, "step": 11449 }, { "epoch": 0.65, "learning_rate": 5.7804787568021e-05, "loss": 0.6217, "step": 11450 }, { "epoch": 0.65, "learning_rate": 5.778813199479427e-05, "loss": 0.7141, "step": 11451 }, { "epoch": 0.65, "learning_rate": 5.777147784632879e-05, "loss": 0.6302, "step": 11452 }, { "epoch": 0.65, "learning_rate": 5.7754825123186685e-05, "loss": 0.6747, "step": 11453 }, { "epoch": 0.65, "learning_rate": 5.773817382593008e-05, "loss": 0.6473, "step": 11454 }, { "epoch": 0.65, "learning_rate": 5.772152395512087e-05, "loss": 0.6709, "step": 11455 }, { "epoch": 0.65, "learning_rate": 5.770487551132118e-05, "loss": 0.8252, "step": 11456 }, { "epoch": 0.65, "learning_rate": 5.76882284950928e-05, "loss": 0.6427, "step": 11457 }, { "epoch": 0.65, "learning_rate": 5.767158290699768e-05, "loss": 0.7765, "step": 11458 }, { "epoch": 0.65, "learning_rate": 5.7654938747597656e-05, "loss": 0.6826, "step": 11459 }, { "epoch": 0.65, "learning_rate": 5.76382960174545e-05, "loss": 0.7313, "step": 11460 }, { "epoch": 0.65, "learning_rate": 5.762165471712998e-05, "loss": 0.6786, "step": 11461 }, { "epoch": 0.65, "learning_rate": 5.7605014847185725e-05, "loss": 0.6412, "step": 11462 }, { "epoch": 0.65, "learning_rate": 5.758837640818344e-05, "loss": 0.6949, "step": 11463 }, { "epoch": 0.65, "learning_rate": 5.757173940068464e-05, "loss": 0.7258, "step": 11464 }, { "epoch": 0.65, "learning_rate": 5.7555103825250914e-05, "loss": 0.7052, "step": 11465 }, { "epoch": 0.65, "learning_rate": 5.753846968244374e-05, "loss": 0.7466, "step": 11466 }, { "epoch": 0.65, "learning_rate": 5.7521836972824584e-05, "loss": 0.7044, "step": 11467 }, { "epoch": 0.65, "learning_rate": 5.7505205696954864e-05, "loss": 0.7237, "step": 11468 }, { "epoch": 0.65, "learning_rate": 5.748857585539591e-05, "loss": 0.6961, "step": 11469 }, { "epoch": 0.65, "learning_rate": 5.7471947448708964e-05, "loss": 0.6738, "step": 11470 }, { "epoch": 0.65, "learning_rate": 5.745532047745532e-05, "loss": 0.7032, "step": 11471 }, { "epoch": 0.65, "learning_rate": 5.743869494219619e-05, "loss": 0.7013, "step": 11472 }, { "epoch": 0.65, "learning_rate": 5.7422070843492734e-05, "loss": 0.6023, "step": 11473 }, { "epoch": 0.65, "learning_rate": 5.740544818190604e-05, "loss": 0.6988, "step": 11474 }, { "epoch": 0.65, "learning_rate": 5.738882695799723e-05, "loss": 0.737, "step": 11475 }, { "epoch": 0.65, "learning_rate": 5.737220717232725e-05, "loss": 0.7139, "step": 11476 }, { "epoch": 0.65, "learning_rate": 5.7355588825457044e-05, "loss": 0.6713, "step": 11477 }, { "epoch": 0.65, "learning_rate": 5.733897191794755e-05, "loss": 0.749, "step": 11478 }, { "epoch": 0.65, "learning_rate": 5.732235645035964e-05, "loss": 0.6509, "step": 11479 }, { "epoch": 0.65, "learning_rate": 5.730574242325413e-05, "loss": 0.7615, "step": 11480 }, { "epoch": 0.65, "learning_rate": 5.728912983719181e-05, "loss": 0.6931, "step": 11481 }, { "epoch": 0.65, "learning_rate": 5.727251869273333e-05, "loss": 0.6553, "step": 11482 }, { "epoch": 0.65, "learning_rate": 5.725590899043941e-05, "loss": 0.5462, "step": 11483 }, { "epoch": 0.65, "learning_rate": 5.7239300730870706e-05, "loss": 0.6674, "step": 11484 }, { "epoch": 0.65, "learning_rate": 5.72226939145877e-05, "loss": 0.7786, "step": 11485 }, { "epoch": 0.65, "learning_rate": 5.720608854215096e-05, "loss": 0.7998, "step": 11486 }, { "epoch": 0.65, "learning_rate": 5.718948461412097e-05, "loss": 0.6717, "step": 11487 }, { "epoch": 0.65, "learning_rate": 5.717288213105818e-05, "loss": 0.742, "step": 11488 }, { "epoch": 0.65, "learning_rate": 5.71562810935229e-05, "loss": 0.7519, "step": 11489 }, { "epoch": 0.65, "learning_rate": 5.713968150207549e-05, "loss": 0.7379, "step": 11490 }, { "epoch": 0.65, "learning_rate": 5.7123083357276283e-05, "loss": 0.6733, "step": 11491 }, { "epoch": 0.65, "learning_rate": 5.710648665968543e-05, "loss": 0.6869, "step": 11492 }, { "epoch": 0.65, "learning_rate": 5.708989140986313e-05, "loss": 0.6349, "step": 11493 }, { "epoch": 0.65, "learning_rate": 5.707329760836957e-05, "loss": 0.7455, "step": 11494 }, { "epoch": 0.65, "learning_rate": 5.705670525576476e-05, "loss": 0.7149, "step": 11495 }, { "epoch": 0.65, "learning_rate": 5.7040114352608765e-05, "loss": 0.7161, "step": 11496 }, { "epoch": 0.65, "learning_rate": 5.702352489946159e-05, "loss": 0.7315, "step": 11497 }, { "epoch": 0.65, "learning_rate": 5.700693689688314e-05, "loss": 0.7303, "step": 11498 }, { "epoch": 0.65, "learning_rate": 5.6990350345433384e-05, "loss": 0.6148, "step": 11499 }, { "epoch": 0.65, "learning_rate": 5.697376524567205e-05, "loss": 0.7683, "step": 11500 }, { "epoch": 0.65, "learning_rate": 5.695718159815901e-05, "loss": 0.6365, "step": 11501 }, { "epoch": 0.65, "learning_rate": 5.6940599403453955e-05, "loss": 0.7447, "step": 11502 }, { "epoch": 0.65, "learning_rate": 5.692401866211658e-05, "loss": 0.7162, "step": 11503 }, { "epoch": 0.65, "learning_rate": 5.690743937470657e-05, "loss": 0.6152, "step": 11504 }, { "epoch": 0.65, "learning_rate": 5.689086154178348e-05, "loss": 0.7244, "step": 11505 }, { "epoch": 0.65, "learning_rate": 5.687428516390692e-05, "loss": 0.7483, "step": 11506 }, { "epoch": 0.65, "learning_rate": 5.685771024163628e-05, "loss": 0.7149, "step": 11507 }, { "epoch": 0.65, "learning_rate": 5.6841136775531114e-05, "loss": 0.8367, "step": 11508 }, { "epoch": 0.65, "learning_rate": 5.6824564766150726e-05, "loss": 0.6765, "step": 11509 }, { "epoch": 0.65, "learning_rate": 5.680799421405451e-05, "loss": 0.7061, "step": 11510 }, { "epoch": 0.65, "learning_rate": 5.679142511980175e-05, "loss": 0.6983, "step": 11511 }, { "epoch": 0.65, "learning_rate": 5.677485748395174e-05, "loss": 0.7145, "step": 11512 }, { "epoch": 0.65, "learning_rate": 5.675829130706367e-05, "loss": 0.7037, "step": 11513 }, { "epoch": 0.65, "learning_rate": 5.67417265896967e-05, "loss": 0.6572, "step": 11514 }, { "epoch": 0.65, "learning_rate": 5.6725163332409836e-05, "loss": 0.7509, "step": 11515 }, { "epoch": 0.65, "learning_rate": 5.670860153576222e-05, "loss": 0.6748, "step": 11516 }, { "epoch": 0.65, "learning_rate": 5.669204120031283e-05, "loss": 0.6257, "step": 11517 }, { "epoch": 0.65, "learning_rate": 5.6675482326620635e-05, "loss": 0.6453, "step": 11518 }, { "epoch": 0.65, "learning_rate": 5.6658924915244524e-05, "loss": 0.6825, "step": 11519 }, { "epoch": 0.65, "learning_rate": 5.664236896674342e-05, "loss": 0.7514, "step": 11520 }, { "epoch": 0.65, "learning_rate": 5.6625814481676054e-05, "loss": 0.8205, "step": 11521 }, { "epoch": 0.65, "learning_rate": 5.660926146060117e-05, "loss": 0.7277, "step": 11522 }, { "epoch": 0.65, "learning_rate": 5.6592709904077515e-05, "loss": 0.7408, "step": 11523 }, { "epoch": 0.65, "learning_rate": 5.657615981266373e-05, "loss": 0.7325, "step": 11524 }, { "epoch": 0.65, "learning_rate": 5.6559611186918436e-05, "loss": 0.6944, "step": 11525 }, { "epoch": 0.65, "learning_rate": 5.654306402740024e-05, "loss": 0.7587, "step": 11526 }, { "epoch": 0.65, "learning_rate": 5.652651833466755e-05, "loss": 0.7255, "step": 11527 }, { "epoch": 0.65, "learning_rate": 5.650997410927894e-05, "loss": 0.7519, "step": 11528 }, { "epoch": 0.65, "learning_rate": 5.64934313517927e-05, "loss": 0.6902, "step": 11529 }, { "epoch": 0.65, "learning_rate": 5.647689006276726e-05, "loss": 0.6348, "step": 11530 }, { "epoch": 0.65, "learning_rate": 5.6460350242760926e-05, "loss": 0.7049, "step": 11531 }, { "epoch": 0.65, "learning_rate": 5.644381189233196e-05, "loss": 0.7079, "step": 11532 }, { "epoch": 0.65, "learning_rate": 5.642727501203861e-05, "loss": 0.6725, "step": 11533 }, { "epoch": 0.65, "learning_rate": 5.6410739602438955e-05, "loss": 0.6736, "step": 11534 }, { "epoch": 0.65, "learning_rate": 5.639420566409116e-05, "loss": 0.7795, "step": 11535 }, { "epoch": 0.65, "learning_rate": 5.6377673197553325e-05, "loss": 0.82, "step": 11536 }, { "epoch": 0.65, "learning_rate": 5.636114220338338e-05, "loss": 0.6195, "step": 11537 }, { "epoch": 0.65, "learning_rate": 5.6344612682139345e-05, "loss": 0.6502, "step": 11538 }, { "epoch": 0.65, "learning_rate": 5.632808463437915e-05, "loss": 0.7065, "step": 11539 }, { "epoch": 0.65, "learning_rate": 5.631155806066059e-05, "loss": 0.7756, "step": 11540 }, { "epoch": 0.65, "learning_rate": 5.6295032961541525e-05, "loss": 0.7316, "step": 11541 }, { "epoch": 0.65, "learning_rate": 5.6278509337579734e-05, "loss": 0.8496, "step": 11542 }, { "epoch": 0.65, "learning_rate": 5.626198718933294e-05, "loss": 0.697, "step": 11543 }, { "epoch": 0.65, "learning_rate": 5.6245466517358755e-05, "loss": 0.6692, "step": 11544 }, { "epoch": 0.65, "learning_rate": 5.622894732221482e-05, "loss": 0.7395, "step": 11545 }, { "epoch": 0.65, "learning_rate": 5.621242960445874e-05, "loss": 0.6917, "step": 11546 }, { "epoch": 0.66, "learning_rate": 5.619591336464798e-05, "loss": 0.6868, "step": 11547 }, { "epoch": 0.66, "learning_rate": 5.6179398603340004e-05, "loss": 0.7069, "step": 11548 }, { "epoch": 0.66, "learning_rate": 5.616288532109225e-05, "loss": 0.6817, "step": 11549 }, { "epoch": 0.66, "learning_rate": 5.6146373518462126e-05, "loss": 0.7165, "step": 11550 }, { "epoch": 0.66, "learning_rate": 5.612986319600687e-05, "loss": 0.5679, "step": 11551 }, { "epoch": 0.66, "learning_rate": 5.611335435428377e-05, "loss": 0.7004, "step": 11552 }, { "epoch": 0.66, "learning_rate": 5.609684699385012e-05, "loss": 0.613, "step": 11553 }, { "epoch": 0.66, "learning_rate": 5.608034111526298e-05, "loss": 0.6783, "step": 11554 }, { "epoch": 0.66, "learning_rate": 5.606383671907951e-05, "loss": 0.7415, "step": 11555 }, { "epoch": 0.66, "learning_rate": 5.604733380585676e-05, "loss": 0.6835, "step": 11556 }, { "epoch": 0.66, "learning_rate": 5.603083237615179e-05, "loss": 0.644, "step": 11557 }, { "epoch": 0.66, "learning_rate": 5.601433243052156e-05, "loss": 0.6559, "step": 11558 }, { "epoch": 0.66, "learning_rate": 5.599783396952297e-05, "loss": 0.7004, "step": 11559 }, { "epoch": 0.66, "learning_rate": 5.5981336993712854e-05, "loss": 0.5799, "step": 11560 }, { "epoch": 0.66, "learning_rate": 5.596484150364805e-05, "loss": 0.8069, "step": 11561 }, { "epoch": 0.66, "learning_rate": 5.594834749988533e-05, "loss": 0.6452, "step": 11562 }, { "epoch": 0.66, "learning_rate": 5.593185498298141e-05, "loss": 0.7074, "step": 11563 }, { "epoch": 0.66, "learning_rate": 5.591536395349297e-05, "loss": 0.8701, "step": 11564 }, { "epoch": 0.66, "learning_rate": 5.5898874411976656e-05, "loss": 0.775, "step": 11565 }, { "epoch": 0.66, "learning_rate": 5.588238635898897e-05, "loss": 0.79, "step": 11566 }, { "epoch": 0.66, "learning_rate": 5.5865899795086416e-05, "loss": 0.6937, "step": 11567 }, { "epoch": 0.66, "learning_rate": 5.584941472082549e-05, "loss": 0.7344, "step": 11568 }, { "epoch": 0.66, "learning_rate": 5.58329311367626e-05, "loss": 0.7161, "step": 11569 }, { "epoch": 0.66, "learning_rate": 5.5816449043454136e-05, "loss": 0.6851, "step": 11570 }, { "epoch": 0.66, "learning_rate": 5.579996844145641e-05, "loss": 0.5895, "step": 11571 }, { "epoch": 0.66, "learning_rate": 5.578348933132565e-05, "loss": 0.729, "step": 11572 }, { "epoch": 0.66, "learning_rate": 5.576701171361811e-05, "loss": 0.6686, "step": 11573 }, { "epoch": 0.66, "learning_rate": 5.5750535588889895e-05, "loss": 0.6911, "step": 11574 }, { "epoch": 0.66, "learning_rate": 5.5734060957697154e-05, "loss": 0.8157, "step": 11575 }, { "epoch": 0.66, "learning_rate": 5.571758782059593e-05, "loss": 0.7209, "step": 11576 }, { "epoch": 0.66, "learning_rate": 5.5701116178142276e-05, "loss": 0.8427, "step": 11577 }, { "epoch": 0.66, "learning_rate": 5.5684646030892164e-05, "loss": 0.6327, "step": 11578 }, { "epoch": 0.66, "learning_rate": 5.566817737940142e-05, "loss": 0.655, "step": 11579 }, { "epoch": 0.66, "learning_rate": 5.5651710224225994e-05, "loss": 0.7609, "step": 11580 }, { "epoch": 0.66, "learning_rate": 5.563524456592163e-05, "loss": 0.7664, "step": 11581 }, { "epoch": 0.66, "learning_rate": 5.5618780405044103e-05, "loss": 0.6993, "step": 11582 }, { "epoch": 0.66, "learning_rate": 5.560231774214912e-05, "loss": 0.7686, "step": 11583 }, { "epoch": 0.66, "learning_rate": 5.5585856577792404e-05, "loss": 0.6639, "step": 11584 }, { "epoch": 0.66, "learning_rate": 5.556939691252947e-05, "loss": 0.7774, "step": 11585 }, { "epoch": 0.66, "learning_rate": 5.5552938746915915e-05, "loss": 0.6249, "step": 11586 }, { "epoch": 0.66, "learning_rate": 5.553648208150728e-05, "loss": 0.7598, "step": 11587 }, { "epoch": 0.66, "learning_rate": 5.552002691685895e-05, "loss": 0.7746, "step": 11588 }, { "epoch": 0.66, "learning_rate": 5.5503573253526354e-05, "loss": 0.6878, "step": 11589 }, { "epoch": 0.66, "learning_rate": 5.548712109206486e-05, "loss": 0.7523, "step": 11590 }, { "epoch": 0.66, "learning_rate": 5.5470670433029814e-05, "loss": 0.7238, "step": 11591 }, { "epoch": 0.66, "learning_rate": 5.545422127697638e-05, "loss": 0.6924, "step": 11592 }, { "epoch": 0.66, "learning_rate": 5.543777362445981e-05, "loss": 0.8191, "step": 11593 }, { "epoch": 0.66, "learning_rate": 5.542132747603523e-05, "loss": 0.7259, "step": 11594 }, { "epoch": 0.66, "learning_rate": 5.540488283225782e-05, "loss": 0.7086, "step": 11595 }, { "epoch": 0.66, "learning_rate": 5.5388439693682524e-05, "loss": 0.7587, "step": 11596 }, { "epoch": 0.66, "learning_rate": 5.5371998060864385e-05, "loss": 0.6755, "step": 11597 }, { "epoch": 0.66, "learning_rate": 5.535555793435838e-05, "loss": 0.6805, "step": 11598 }, { "epoch": 0.66, "learning_rate": 5.533911931471936e-05, "loss": 0.7461, "step": 11599 }, { "epoch": 0.66, "learning_rate": 5.532268220250219e-05, "loss": 0.7247, "step": 11600 }, { "epoch": 0.66, "learning_rate": 5.530624659826165e-05, "loss": 0.6706, "step": 11601 }, { "epoch": 0.66, "learning_rate": 5.528981250255254e-05, "loss": 0.7063, "step": 11602 }, { "epoch": 0.66, "learning_rate": 5.527337991592949e-05, "loss": 0.6926, "step": 11603 }, { "epoch": 0.66, "learning_rate": 5.52569488389472e-05, "loss": 0.7197, "step": 11604 }, { "epoch": 0.66, "learning_rate": 5.524051927216018e-05, "loss": 0.6696, "step": 11605 }, { "epoch": 0.66, "learning_rate": 5.522409121612304e-05, "loss": 0.8587, "step": 11606 }, { "epoch": 0.66, "learning_rate": 5.520766467139025e-05, "loss": 0.6274, "step": 11607 }, { "epoch": 0.66, "learning_rate": 5.519123963851623e-05, "loss": 0.6469, "step": 11608 }, { "epoch": 0.66, "learning_rate": 5.5174816118055396e-05, "loss": 0.6402, "step": 11609 }, { "epoch": 0.66, "learning_rate": 5.515839411056212e-05, "loss": 0.6855, "step": 11610 }, { "epoch": 0.66, "learning_rate": 5.514197361659065e-05, "loss": 0.6365, "step": 11611 }, { "epoch": 0.66, "learning_rate": 5.5125554636695165e-05, "loss": 0.7456, "step": 11612 }, { "epoch": 0.66, "learning_rate": 5.5109137171429906e-05, "loss": 0.6356, "step": 11613 }, { "epoch": 0.66, "learning_rate": 5.509272122134899e-05, "loss": 0.6951, "step": 11614 }, { "epoch": 0.66, "learning_rate": 5.5076306787006504e-05, "loss": 0.7393, "step": 11615 }, { "epoch": 0.66, "learning_rate": 5.505989386895654e-05, "loss": 0.684, "step": 11616 }, { "epoch": 0.66, "learning_rate": 5.504348246775299e-05, "loss": 0.6807, "step": 11617 }, { "epoch": 0.66, "learning_rate": 5.502707258394979e-05, "loss": 0.6756, "step": 11618 }, { "epoch": 0.66, "learning_rate": 5.501066421810084e-05, "loss": 0.7183, "step": 11619 }, { "epoch": 0.66, "learning_rate": 5.499425737075994e-05, "loss": 0.5901, "step": 11620 }, { "epoch": 0.66, "learning_rate": 5.49778520424809e-05, "loss": 0.8408, "step": 11621 }, { "epoch": 0.66, "learning_rate": 5.4961448233817424e-05, "loss": 0.6845, "step": 11622 }, { "epoch": 0.66, "learning_rate": 5.494504594532324e-05, "loss": 0.8022, "step": 11623 }, { "epoch": 0.66, "learning_rate": 5.492864517755189e-05, "loss": 0.7788, "step": 11624 }, { "epoch": 0.66, "learning_rate": 5.491224593105695e-05, "loss": 0.6696, "step": 11625 }, { "epoch": 0.66, "learning_rate": 5.489584820639195e-05, "loss": 0.601, "step": 11626 }, { "epoch": 0.66, "learning_rate": 5.487945200411038e-05, "loss": 0.7289, "step": 11627 }, { "epoch": 0.66, "learning_rate": 5.486305732476562e-05, "loss": 0.5545, "step": 11628 }, { "epoch": 0.66, "learning_rate": 5.484666416891109e-05, "loss": 0.6939, "step": 11629 }, { "epoch": 0.66, "learning_rate": 5.483027253710003e-05, "loss": 0.6197, "step": 11630 }, { "epoch": 0.66, "learning_rate": 5.481388242988572e-05, "loss": 0.7696, "step": 11631 }, { "epoch": 0.66, "learning_rate": 5.4797493847821446e-05, "loss": 0.7198, "step": 11632 }, { "epoch": 0.66, "learning_rate": 5.478110679146025e-05, "loss": 0.692, "step": 11633 }, { "epoch": 0.66, "learning_rate": 5.476472126135529e-05, "loss": 0.6538, "step": 11634 }, { "epoch": 0.66, "learning_rate": 5.4748337258059626e-05, "loss": 0.6752, "step": 11635 }, { "epoch": 0.66, "learning_rate": 5.4731954782126294e-05, "loss": 0.6853, "step": 11636 }, { "epoch": 0.66, "learning_rate": 5.471557383410816e-05, "loss": 0.7115, "step": 11637 }, { "epoch": 0.66, "learning_rate": 5.4699194414558175e-05, "loss": 0.7564, "step": 11638 }, { "epoch": 0.66, "learning_rate": 5.468281652402923e-05, "loss": 0.6433, "step": 11639 }, { "epoch": 0.66, "learning_rate": 5.466644016307403e-05, "loss": 0.7677, "step": 11640 }, { "epoch": 0.66, "learning_rate": 5.465006533224537e-05, "loss": 0.7824, "step": 11641 }, { "epoch": 0.66, "learning_rate": 5.463369203209594e-05, "loss": 0.7696, "step": 11642 }, { "epoch": 0.66, "learning_rate": 5.461732026317843e-05, "loss": 0.7611, "step": 11643 }, { "epoch": 0.66, "learning_rate": 5.4600950026045326e-05, "loss": 0.7018, "step": 11644 }, { "epoch": 0.66, "learning_rate": 5.458458132124924e-05, "loss": 0.618, "step": 11645 }, { "epoch": 0.66, "learning_rate": 5.4568214149342624e-05, "loss": 0.7401, "step": 11646 }, { "epoch": 0.66, "learning_rate": 5.455184851087798e-05, "loss": 0.7129, "step": 11647 }, { "epoch": 0.66, "learning_rate": 5.453548440640759e-05, "loss": 0.7876, "step": 11648 }, { "epoch": 0.66, "learning_rate": 5.4519121836483886e-05, "loss": 0.8288, "step": 11649 }, { "epoch": 0.66, "learning_rate": 5.450276080165906e-05, "loss": 0.6632, "step": 11650 }, { "epoch": 0.66, "learning_rate": 5.448640130248537e-05, "loss": 0.7739, "step": 11651 }, { "epoch": 0.66, "learning_rate": 5.4470043339515e-05, "loss": 0.7147, "step": 11652 }, { "epoch": 0.66, "learning_rate": 5.4453686913300074e-05, "loss": 0.6033, "step": 11653 }, { "epoch": 0.66, "learning_rate": 5.4437332024392694e-05, "loss": 0.7365, "step": 11654 }, { "epoch": 0.66, "learning_rate": 5.4420978673344836e-05, "loss": 0.7521, "step": 11655 }, { "epoch": 0.66, "learning_rate": 5.4404626860708506e-05, "loss": 0.7275, "step": 11656 }, { "epoch": 0.66, "learning_rate": 5.438827658703556e-05, "loss": 0.6999, "step": 11657 }, { "epoch": 0.66, "learning_rate": 5.43719278528779e-05, "loss": 0.7106, "step": 11658 }, { "epoch": 0.66, "learning_rate": 5.435558065878734e-05, "loss": 0.7122, "step": 11659 }, { "epoch": 0.66, "learning_rate": 5.4339235005315646e-05, "loss": 0.6856, "step": 11660 }, { "epoch": 0.66, "learning_rate": 5.432289089301456e-05, "loss": 0.6545, "step": 11661 }, { "epoch": 0.66, "learning_rate": 5.430654832243569e-05, "loss": 0.8124, "step": 11662 }, { "epoch": 0.66, "learning_rate": 5.4290207294130615e-05, "loss": 0.7121, "step": 11663 }, { "epoch": 0.66, "learning_rate": 5.4273867808650915e-05, "loss": 0.7539, "step": 11664 }, { "epoch": 0.66, "learning_rate": 5.4257529866548105e-05, "loss": 0.5937, "step": 11665 }, { "epoch": 0.66, "learning_rate": 5.4241193468373617e-05, "loss": 0.7206, "step": 11666 }, { "epoch": 0.66, "learning_rate": 5.4224858614678876e-05, "loss": 0.7129, "step": 11667 }, { "epoch": 0.66, "learning_rate": 5.420852530601523e-05, "loss": 0.7022, "step": 11668 }, { "epoch": 0.66, "learning_rate": 5.4192193542933943e-05, "loss": 0.6977, "step": 11669 }, { "epoch": 0.66, "learning_rate": 5.417586332598623e-05, "loss": 0.7106, "step": 11670 }, { "epoch": 0.66, "learning_rate": 5.415953465572332e-05, "loss": 0.7649, "step": 11671 }, { "epoch": 0.66, "learning_rate": 5.4143207532696316e-05, "loss": 0.6914, "step": 11672 }, { "epoch": 0.66, "learning_rate": 5.412688195745632e-05, "loss": 0.8042, "step": 11673 }, { "epoch": 0.66, "learning_rate": 5.411055793055442e-05, "loss": 0.6738, "step": 11674 }, { "epoch": 0.66, "learning_rate": 5.4094235452541484e-05, "loss": 0.6953, "step": 11675 }, { "epoch": 0.66, "learning_rate": 5.4077914523968555e-05, "loss": 0.7367, "step": 11676 }, { "epoch": 0.66, "learning_rate": 5.406159514538638e-05, "loss": 0.7481, "step": 11677 }, { "epoch": 0.66, "learning_rate": 5.404527731734587e-05, "loss": 0.6856, "step": 11678 }, { "epoch": 0.66, "learning_rate": 5.402896104039776e-05, "loss": 0.7539, "step": 11679 }, { "epoch": 0.66, "learning_rate": 5.401264631509277e-05, "loss": 0.7935, "step": 11680 }, { "epoch": 0.66, "learning_rate": 5.399633314198163e-05, "loss": 0.6753, "step": 11681 }, { "epoch": 0.66, "learning_rate": 5.398002152161484e-05, "loss": 0.7805, "step": 11682 }, { "epoch": 0.66, "learning_rate": 5.396371145454302e-05, "loss": 0.7534, "step": 11683 }, { "epoch": 0.66, "learning_rate": 5.394740294131673e-05, "loss": 0.6483, "step": 11684 }, { "epoch": 0.66, "learning_rate": 5.393109598248631e-05, "loss": 0.7506, "step": 11685 }, { "epoch": 0.66, "learning_rate": 5.391479057860222e-05, "loss": 0.5576, "step": 11686 }, { "epoch": 0.66, "learning_rate": 5.3898486730214826e-05, "loss": 0.6888, "step": 11687 }, { "epoch": 0.66, "learning_rate": 5.388218443787443e-05, "loss": 0.7381, "step": 11688 }, { "epoch": 0.66, "learning_rate": 5.386588370213124e-05, "loss": 0.7158, "step": 11689 }, { "epoch": 0.66, "learning_rate": 5.3849584523535435e-05, "loss": 0.7308, "step": 11690 }, { "epoch": 0.66, "learning_rate": 5.3833286902637245e-05, "loss": 0.6502, "step": 11691 }, { "epoch": 0.66, "learning_rate": 5.381699083998666e-05, "loss": 0.669, "step": 11692 }, { "epoch": 0.66, "learning_rate": 5.380069633613376e-05, "loss": 0.7538, "step": 11693 }, { "epoch": 0.66, "learning_rate": 5.378440339162856e-05, "loss": 0.8147, "step": 11694 }, { "epoch": 0.66, "learning_rate": 5.37681120070209e-05, "loss": 0.7537, "step": 11695 }, { "epoch": 0.66, "learning_rate": 5.37518221828607e-05, "loss": 0.6693, "step": 11696 }, { "epoch": 0.66, "learning_rate": 5.3735533919697814e-05, "loss": 0.6977, "step": 11697 }, { "epoch": 0.66, "learning_rate": 5.371924721808197e-05, "loss": 0.8723, "step": 11698 }, { "epoch": 0.66, "learning_rate": 5.3702962078562956e-05, "loss": 0.7108, "step": 11699 }, { "epoch": 0.66, "learning_rate": 5.368667850169036e-05, "loss": 0.7295, "step": 11700 }, { "epoch": 0.66, "learning_rate": 5.3670396488013854e-05, "loss": 0.6787, "step": 11701 }, { "epoch": 0.66, "learning_rate": 5.3654116038082936e-05, "loss": 0.7303, "step": 11702 }, { "epoch": 0.66, "learning_rate": 5.3637837152447145e-05, "loss": 0.7211, "step": 11703 }, { "epoch": 0.66, "learning_rate": 5.362155983165594e-05, "loss": 0.7129, "step": 11704 }, { "epoch": 0.66, "learning_rate": 5.360528407625873e-05, "loss": 0.7227, "step": 11705 }, { "epoch": 0.66, "learning_rate": 5.358900988680489e-05, "loss": 0.7497, "step": 11706 }, { "epoch": 0.66, "learning_rate": 5.357273726384367e-05, "loss": 0.698, "step": 11707 }, { "epoch": 0.66, "learning_rate": 5.3556466207924317e-05, "loss": 0.7495, "step": 11708 }, { "epoch": 0.66, "learning_rate": 5.3540196719596e-05, "loss": 0.6431, "step": 11709 }, { "epoch": 0.66, "learning_rate": 5.352392879940791e-05, "loss": 0.7365, "step": 11710 }, { "epoch": 0.66, "learning_rate": 5.350766244790911e-05, "loss": 0.7119, "step": 11711 }, { "epoch": 0.66, "learning_rate": 5.3491397665648634e-05, "loss": 0.7235, "step": 11712 }, { "epoch": 0.66, "learning_rate": 5.3475134453175494e-05, "loss": 0.6605, "step": 11713 }, { "epoch": 0.66, "learning_rate": 5.345887281103859e-05, "loss": 0.7591, "step": 11714 }, { "epoch": 0.66, "learning_rate": 5.3442612739786726e-05, "loss": 0.7529, "step": 11715 }, { "epoch": 0.66, "learning_rate": 5.34263542399688e-05, "loss": 0.6813, "step": 11716 }, { "epoch": 0.66, "learning_rate": 5.3410097312133556e-05, "loss": 0.6835, "step": 11717 }, { "epoch": 0.66, "learning_rate": 5.3393841956829724e-05, "loss": 0.7733, "step": 11718 }, { "epoch": 0.66, "learning_rate": 5.3377588174605985e-05, "loss": 0.6289, "step": 11719 }, { "epoch": 0.66, "learning_rate": 5.33613359660109e-05, "loss": 0.7461, "step": 11720 }, { "epoch": 0.66, "learning_rate": 5.334508533159305e-05, "loss": 0.7009, "step": 11721 }, { "epoch": 0.66, "learning_rate": 5.3328836271900905e-05, "loss": 0.6982, "step": 11722 }, { "epoch": 0.66, "learning_rate": 5.331258878748293e-05, "loss": 0.771, "step": 11723 }, { "epoch": 0.67, "learning_rate": 5.329634287888752e-05, "loss": 0.6458, "step": 11724 }, { "epoch": 0.67, "learning_rate": 5.328009854666303e-05, "loss": 0.6714, "step": 11725 }, { "epoch": 0.67, "learning_rate": 5.326385579135778e-05, "loss": 0.6846, "step": 11726 }, { "epoch": 0.67, "learning_rate": 5.324761461351992e-05, "loss": 0.6874, "step": 11727 }, { "epoch": 0.67, "learning_rate": 5.323137501369773e-05, "loss": 0.7194, "step": 11728 }, { "epoch": 0.67, "learning_rate": 5.321513699243924e-05, "loss": 0.761, "step": 11729 }, { "epoch": 0.67, "learning_rate": 5.3198900550292576e-05, "loss": 0.7085, "step": 11730 }, { "epoch": 0.67, "learning_rate": 5.318266568780577e-05, "loss": 0.6954, "step": 11731 }, { "epoch": 0.67, "learning_rate": 5.316643240552677e-05, "loss": 0.6817, "step": 11732 }, { "epoch": 0.67, "learning_rate": 5.315020070400354e-05, "loss": 0.6684, "step": 11733 }, { "epoch": 0.67, "learning_rate": 5.313397058378386e-05, "loss": 0.7408, "step": 11734 }, { "epoch": 0.67, "learning_rate": 5.31177420454156e-05, "loss": 0.6662, "step": 11735 }, { "epoch": 0.67, "learning_rate": 5.310151508944653e-05, "loss": 0.6097, "step": 11736 }, { "epoch": 0.67, "learning_rate": 5.308528971642428e-05, "loss": 0.6688, "step": 11737 }, { "epoch": 0.67, "learning_rate": 5.306906592689655e-05, "loss": 0.7846, "step": 11738 }, { "epoch": 0.67, "learning_rate": 5.305284372141095e-05, "loss": 0.7639, "step": 11739 }, { "epoch": 0.67, "learning_rate": 5.3036623100514956e-05, "loss": 0.7876, "step": 11740 }, { "epoch": 0.67, "learning_rate": 5.3020404064756125e-05, "loss": 0.6936, "step": 11741 }, { "epoch": 0.67, "learning_rate": 5.300418661468184e-05, "loss": 0.7158, "step": 11742 }, { "epoch": 0.67, "learning_rate": 5.2987970750839555e-05, "loss": 0.5972, "step": 11743 }, { "epoch": 0.67, "learning_rate": 5.2971756473776515e-05, "loss": 0.7316, "step": 11744 }, { "epoch": 0.67, "learning_rate": 5.295554378404002e-05, "loss": 0.7672, "step": 11745 }, { "epoch": 0.67, "learning_rate": 5.2939332682177355e-05, "loss": 0.7201, "step": 11746 }, { "epoch": 0.67, "learning_rate": 5.2923123168735574e-05, "loss": 0.6774, "step": 11747 }, { "epoch": 0.67, "learning_rate": 5.290691524426188e-05, "loss": 0.6239, "step": 11748 }, { "epoch": 0.67, "learning_rate": 5.289070890930328e-05, "loss": 0.6668, "step": 11749 }, { "epoch": 0.67, "learning_rate": 5.2874504164406846e-05, "loss": 0.7259, "step": 11750 }, { "epoch": 0.67, "learning_rate": 5.2858301010119446e-05, "loss": 0.6345, "step": 11751 }, { "epoch": 0.67, "learning_rate": 5.284209944698806e-05, "loss": 0.7493, "step": 11752 }, { "epoch": 0.67, "learning_rate": 5.282589947555946e-05, "loss": 0.6924, "step": 11753 }, { "epoch": 0.67, "learning_rate": 5.280970109638047e-05, "loss": 0.8322, "step": 11754 }, { "epoch": 0.67, "learning_rate": 5.279350430999783e-05, "loss": 0.743, "step": 11755 }, { "epoch": 0.67, "learning_rate": 5.277730911695822e-05, "loss": 0.6796, "step": 11756 }, { "epoch": 0.67, "learning_rate": 5.276111551780827e-05, "loss": 0.6537, "step": 11757 }, { "epoch": 0.67, "learning_rate": 5.274492351309461e-05, "loss": 0.7505, "step": 11758 }, { "epoch": 0.67, "learning_rate": 5.27287331033637e-05, "loss": 0.7725, "step": 11759 }, { "epoch": 0.67, "learning_rate": 5.271254428916199e-05, "loss": 0.5164, "step": 11760 }, { "epoch": 0.67, "learning_rate": 5.269635707103593e-05, "loss": 0.7581, "step": 11761 }, { "epoch": 0.67, "learning_rate": 5.268017144953187e-05, "loss": 0.7168, "step": 11762 }, { "epoch": 0.67, "learning_rate": 5.266398742519611e-05, "loss": 0.7312, "step": 11763 }, { "epoch": 0.67, "learning_rate": 5.2647804998574946e-05, "loss": 0.6788, "step": 11764 }, { "epoch": 0.67, "learning_rate": 5.263162417021457e-05, "loss": 0.7678, "step": 11765 }, { "epoch": 0.67, "learning_rate": 5.261544494066111e-05, "loss": 0.6998, "step": 11766 }, { "epoch": 0.67, "learning_rate": 5.259926731046062e-05, "loss": 0.7646, "step": 11767 }, { "epoch": 0.67, "learning_rate": 5.2583091280159166e-05, "loss": 0.6384, "step": 11768 }, { "epoch": 0.67, "learning_rate": 5.256691685030274e-05, "loss": 0.7927, "step": 11769 }, { "epoch": 0.67, "learning_rate": 5.255074402143727e-05, "loss": 0.7494, "step": 11770 }, { "epoch": 0.67, "learning_rate": 5.253457279410866e-05, "loss": 0.687, "step": 11771 }, { "epoch": 0.67, "learning_rate": 5.2518403168862676e-05, "loss": 0.7172, "step": 11772 }, { "epoch": 0.67, "learning_rate": 5.250223514624513e-05, "loss": 0.7638, "step": 11773 }, { "epoch": 0.67, "learning_rate": 5.248606872680169e-05, "loss": 0.7295, "step": 11774 }, { "epoch": 0.67, "learning_rate": 5.2469903911078035e-05, "loss": 0.7445, "step": 11775 }, { "epoch": 0.67, "learning_rate": 5.245374069961977e-05, "loss": 0.7466, "step": 11776 }, { "epoch": 0.67, "learning_rate": 5.243757909297247e-05, "loss": 0.7043, "step": 11777 }, { "epoch": 0.67, "learning_rate": 5.242141909168163e-05, "loss": 0.6732, "step": 11778 }, { "epoch": 0.67, "learning_rate": 5.240526069629265e-05, "loss": 0.7167, "step": 11779 }, { "epoch": 0.67, "learning_rate": 5.238910390735098e-05, "loss": 0.779, "step": 11780 }, { "epoch": 0.67, "learning_rate": 5.237294872540187e-05, "loss": 0.7161, "step": 11781 }, { "epoch": 0.67, "learning_rate": 5.2356795150990656e-05, "loss": 0.681, "step": 11782 }, { "epoch": 0.67, "learning_rate": 5.234064318466255e-05, "loss": 0.7394, "step": 11783 }, { "epoch": 0.67, "learning_rate": 5.232449282696277e-05, "loss": 0.7625, "step": 11784 }, { "epoch": 0.67, "learning_rate": 5.2308344078436344e-05, "loss": 0.7071, "step": 11785 }, { "epoch": 0.67, "learning_rate": 5.229219693962839e-05, "loss": 0.6967, "step": 11786 }, { "epoch": 0.67, "learning_rate": 5.2276051411083946e-05, "loss": 0.6762, "step": 11787 }, { "epoch": 0.67, "learning_rate": 5.225990749334788e-05, "loss": 0.6489, "step": 11788 }, { "epoch": 0.67, "learning_rate": 5.2243765186965156e-05, "loss": 0.6588, "step": 11789 }, { "epoch": 0.67, "learning_rate": 5.222762449248059e-05, "loss": 0.7691, "step": 11790 }, { "epoch": 0.67, "learning_rate": 5.221148541043903e-05, "loss": 0.7224, "step": 11791 }, { "epoch": 0.67, "learning_rate": 5.219534794138512e-05, "loss": 0.8046, "step": 11792 }, { "epoch": 0.67, "learning_rate": 5.217921208586359e-05, "loss": 0.6472, "step": 11793 }, { "epoch": 0.67, "learning_rate": 5.216307784441907e-05, "loss": 0.7228, "step": 11794 }, { "epoch": 0.67, "learning_rate": 5.2146945217596174e-05, "loss": 0.674, "step": 11795 }, { "epoch": 0.67, "learning_rate": 5.213081420593933e-05, "loss": 0.673, "step": 11796 }, { "epoch": 0.67, "learning_rate": 5.2114684809993044e-05, "loss": 0.7472, "step": 11797 }, { "epoch": 0.67, "learning_rate": 5.2098557030301776e-05, "loss": 0.7804, "step": 11798 }, { "epoch": 0.67, "learning_rate": 5.2082430867409805e-05, "loss": 0.6977, "step": 11799 }, { "epoch": 0.67, "learning_rate": 5.2066306321861444e-05, "loss": 0.7517, "step": 11800 }, { "epoch": 0.67, "learning_rate": 5.205018339420096e-05, "loss": 0.6398, "step": 11801 }, { "epoch": 0.67, "learning_rate": 5.2034062084972576e-05, "loss": 0.6981, "step": 11802 }, { "epoch": 0.67, "learning_rate": 5.201794239472035e-05, "loss": 0.7981, "step": 11803 }, { "epoch": 0.67, "learning_rate": 5.2001824323988455e-05, "loss": 0.7503, "step": 11804 }, { "epoch": 0.67, "learning_rate": 5.198570787332082e-05, "loss": 0.6583, "step": 11805 }, { "epoch": 0.67, "learning_rate": 5.1969593043261475e-05, "loss": 0.7099, "step": 11806 }, { "epoch": 0.67, "learning_rate": 5.1953479834354334e-05, "loss": 0.6714, "step": 11807 }, { "epoch": 0.67, "learning_rate": 5.193736824714325e-05, "loss": 0.7188, "step": 11808 }, { "epoch": 0.67, "learning_rate": 5.1921258282172024e-05, "loss": 0.6907, "step": 11809 }, { "epoch": 0.67, "learning_rate": 5.190514993998448e-05, "loss": 0.6446, "step": 11810 }, { "epoch": 0.67, "learning_rate": 5.188904322112425e-05, "loss": 0.7076, "step": 11811 }, { "epoch": 0.67, "learning_rate": 5.187293812613494e-05, "loss": 0.7008, "step": 11812 }, { "epoch": 0.67, "learning_rate": 5.1856834655560195e-05, "loss": 0.6625, "step": 11813 }, { "epoch": 0.67, "learning_rate": 5.184073280994355e-05, "loss": 0.654, "step": 11814 }, { "epoch": 0.67, "learning_rate": 5.182463258982846e-05, "loss": 0.7672, "step": 11815 }, { "epoch": 0.67, "learning_rate": 5.18085339957584e-05, "loss": 0.7881, "step": 11816 }, { "epoch": 0.67, "learning_rate": 5.179243702827672e-05, "loss": 0.7469, "step": 11817 }, { "epoch": 0.67, "learning_rate": 5.1776341687926664e-05, "loss": 0.6474, "step": 11818 }, { "epoch": 0.67, "learning_rate": 5.1760247975251565e-05, "loss": 0.8497, "step": 11819 }, { "epoch": 0.67, "learning_rate": 5.17441558907946e-05, "loss": 0.8144, "step": 11820 }, { "epoch": 0.67, "learning_rate": 5.172806543509894e-05, "loss": 0.7262, "step": 11821 }, { "epoch": 0.67, "learning_rate": 5.1711976608707654e-05, "loss": 0.7006, "step": 11822 }, { "epoch": 0.67, "learning_rate": 5.169588941216383e-05, "loss": 0.741, "step": 11823 }, { "epoch": 0.67, "learning_rate": 5.167980384601041e-05, "loss": 0.7493, "step": 11824 }, { "epoch": 0.67, "learning_rate": 5.1663719910790354e-05, "loss": 0.7446, "step": 11825 }, { "epoch": 0.67, "learning_rate": 5.1647637607046475e-05, "loss": 0.7871, "step": 11826 }, { "epoch": 0.67, "learning_rate": 5.1631556935321646e-05, "loss": 0.6931, "step": 11827 }, { "epoch": 0.67, "learning_rate": 5.161547789615861e-05, "loss": 0.6499, "step": 11828 }, { "epoch": 0.67, "learning_rate": 5.159940049010015e-05, "loss": 0.7157, "step": 11829 }, { "epoch": 0.67, "learning_rate": 5.15833247176888e-05, "loss": 0.7779, "step": 11830 }, { "epoch": 0.67, "learning_rate": 5.156725057946722e-05, "loss": 0.7519, "step": 11831 }, { "epoch": 0.67, "learning_rate": 5.155117807597799e-05, "loss": 0.6587, "step": 11832 }, { "epoch": 0.67, "learning_rate": 5.1535107207763534e-05, "loss": 0.7185, "step": 11833 }, { "epoch": 0.67, "learning_rate": 5.15190379753663e-05, "loss": 0.6947, "step": 11834 }, { "epoch": 0.67, "learning_rate": 5.15029703793287e-05, "loss": 0.7468, "step": 11835 }, { "epoch": 0.67, "learning_rate": 5.148690442019306e-05, "loss": 0.6621, "step": 11836 }, { "epoch": 0.67, "learning_rate": 5.147084009850158e-05, "loss": 0.76, "step": 11837 }, { "epoch": 0.67, "learning_rate": 5.145477741479654e-05, "loss": 0.688, "step": 11838 }, { "epoch": 0.67, "learning_rate": 5.14387163696201e-05, "loss": 0.7875, "step": 11839 }, { "epoch": 0.67, "learning_rate": 5.142265696351431e-05, "loss": 0.7155, "step": 11840 }, { "epoch": 0.67, "learning_rate": 5.140659919702123e-05, "loss": 0.7877, "step": 11841 }, { "epoch": 0.67, "learning_rate": 5.1390543070682887e-05, "loss": 0.7771, "step": 11842 }, { "epoch": 0.67, "learning_rate": 5.137448858504123e-05, "loss": 0.8027, "step": 11843 }, { "epoch": 0.67, "learning_rate": 5.135843574063807e-05, "loss": 0.6381, "step": 11844 }, { "epoch": 0.67, "learning_rate": 5.134238453801528e-05, "loss": 0.7203, "step": 11845 }, { "epoch": 0.67, "learning_rate": 5.1326334977714596e-05, "loss": 0.7564, "step": 11846 }, { "epoch": 0.67, "learning_rate": 5.131028706027782e-05, "loss": 0.7165, "step": 11847 }, { "epoch": 0.67, "learning_rate": 5.129424078624649e-05, "loss": 0.5995, "step": 11848 }, { "epoch": 0.67, "learning_rate": 5.127819615616233e-05, "loss": 0.7333, "step": 11849 }, { "epoch": 0.67, "learning_rate": 5.126215317056677e-05, "loss": 0.7931, "step": 11850 }, { "epoch": 0.67, "learning_rate": 5.124611183000138e-05, "loss": 0.7753, "step": 11851 }, { "epoch": 0.67, "learning_rate": 5.1230072135007564e-05, "loss": 0.798, "step": 11852 }, { "epoch": 0.67, "learning_rate": 5.121403408612672e-05, "loss": 0.6065, "step": 11853 }, { "epoch": 0.67, "learning_rate": 5.1197997683900214e-05, "loss": 0.7401, "step": 11854 }, { "epoch": 0.67, "learning_rate": 5.1181962928869234e-05, "loss": 0.7521, "step": 11855 }, { "epoch": 0.67, "learning_rate": 5.1165929821575064e-05, "loss": 0.6364, "step": 11856 }, { "epoch": 0.67, "learning_rate": 5.11498983625588e-05, "loss": 0.6884, "step": 11857 }, { "epoch": 0.67, "learning_rate": 5.1133868552361584e-05, "loss": 0.692, "step": 11858 }, { "epoch": 0.67, "learning_rate": 5.111784039152446e-05, "loss": 0.6863, "step": 11859 }, { "epoch": 0.67, "learning_rate": 5.110181388058842e-05, "loss": 0.639, "step": 11860 }, { "epoch": 0.67, "learning_rate": 5.108578902009444e-05, "loss": 0.7034, "step": 11861 }, { "epoch": 0.67, "learning_rate": 5.106976581058337e-05, "loss": 0.7153, "step": 11862 }, { "epoch": 0.67, "learning_rate": 5.105374425259598e-05, "loss": 0.8231, "step": 11863 }, { "epoch": 0.67, "learning_rate": 5.103772434667309e-05, "loss": 0.6339, "step": 11864 }, { "epoch": 0.67, "learning_rate": 5.1021706093355414e-05, "loss": 0.757, "step": 11865 }, { "epoch": 0.67, "learning_rate": 5.100568949318361e-05, "loss": 0.721, "step": 11866 }, { "epoch": 0.67, "learning_rate": 5.098967454669827e-05, "loss": 0.7092, "step": 11867 }, { "epoch": 0.67, "learning_rate": 5.0973661254439986e-05, "loss": 0.7563, "step": 11868 }, { "epoch": 0.67, "learning_rate": 5.095764961694922e-05, "loss": 0.6614, "step": 11869 }, { "epoch": 0.67, "learning_rate": 5.0941639634766355e-05, "loss": 0.7549, "step": 11870 }, { "epoch": 0.67, "learning_rate": 5.092563130843182e-05, "loss": 0.7533, "step": 11871 }, { "epoch": 0.67, "learning_rate": 5.090962463848592e-05, "loss": 0.7659, "step": 11872 }, { "epoch": 0.67, "learning_rate": 5.089361962546894e-05, "loss": 0.6302, "step": 11873 }, { "epoch": 0.67, "learning_rate": 5.0877616269921115e-05, "loss": 0.7481, "step": 11874 }, { "epoch": 0.67, "learning_rate": 5.086161457238254e-05, "loss": 0.6598, "step": 11875 }, { "epoch": 0.67, "learning_rate": 5.084561453339337e-05, "loss": 0.6275, "step": 11876 }, { "epoch": 0.67, "learning_rate": 5.08296161534936e-05, "loss": 0.6756, "step": 11877 }, { "epoch": 0.67, "learning_rate": 5.081361943322322e-05, "loss": 0.7032, "step": 11878 }, { "epoch": 0.67, "learning_rate": 5.079762437312219e-05, "loss": 0.7228, "step": 11879 }, { "epoch": 0.67, "learning_rate": 5.0781630973730364e-05, "loss": 0.7638, "step": 11880 }, { "epoch": 0.67, "learning_rate": 5.0765639235587626e-05, "loss": 0.7569, "step": 11881 }, { "epoch": 0.67, "learning_rate": 5.0749649159233636e-05, "loss": 0.7476, "step": 11882 }, { "epoch": 0.67, "learning_rate": 5.073366074520816e-05, "loss": 0.6465, "step": 11883 }, { "epoch": 0.67, "learning_rate": 5.0717673994050874e-05, "loss": 0.7477, "step": 11884 }, { "epoch": 0.67, "learning_rate": 5.07016889063013e-05, "loss": 0.6952, "step": 11885 }, { "epoch": 0.67, "learning_rate": 5.0685705482499025e-05, "loss": 0.6594, "step": 11886 }, { "epoch": 0.67, "learning_rate": 5.0669723723183506e-05, "loss": 0.6745, "step": 11887 }, { "epoch": 0.67, "learning_rate": 5.065374362889423e-05, "loss": 0.6533, "step": 11888 }, { "epoch": 0.67, "learning_rate": 5.063776520017047e-05, "loss": 0.7424, "step": 11889 }, { "epoch": 0.67, "learning_rate": 5.062178843755161e-05, "loss": 0.7305, "step": 11890 }, { "epoch": 0.67, "learning_rate": 5.0605813341576924e-05, "loss": 0.7243, "step": 11891 }, { "epoch": 0.67, "learning_rate": 5.058983991278554e-05, "loss": 0.6885, "step": 11892 }, { "epoch": 0.67, "learning_rate": 5.057386815171664e-05, "loss": 0.6953, "step": 11893 }, { "epoch": 0.67, "learning_rate": 5.055789805890936e-05, "loss": 0.7363, "step": 11894 }, { "epoch": 0.67, "learning_rate": 5.054192963490263e-05, "loss": 0.7119, "step": 11895 }, { "epoch": 0.67, "learning_rate": 5.05259628802355e-05, "loss": 0.6568, "step": 11896 }, { "epoch": 0.67, "learning_rate": 5.050999779544687e-05, "loss": 0.7445, "step": 11897 }, { "epoch": 0.67, "learning_rate": 5.0494034381075604e-05, "loss": 0.6763, "step": 11898 }, { "epoch": 0.67, "learning_rate": 5.0478072637660555e-05, "loss": 0.6879, "step": 11899 }, { "epoch": 0.68, "learning_rate": 5.0462112565740385e-05, "loss": 0.6724, "step": 11900 }, { "epoch": 0.68, "learning_rate": 5.0446154165853875e-05, "loss": 0.6917, "step": 11901 }, { "epoch": 0.68, "learning_rate": 5.043019743853958e-05, "loss": 0.7243, "step": 11902 }, { "epoch": 0.68, "learning_rate": 5.041424238433613e-05, "loss": 0.7396, "step": 11903 }, { "epoch": 0.68, "learning_rate": 5.039828900378204e-05, "loss": 0.7247, "step": 11904 }, { "epoch": 0.68, "learning_rate": 5.0382337297415773e-05, "loss": 0.7097, "step": 11905 }, { "epoch": 0.68, "learning_rate": 5.03663872657758e-05, "loss": 0.7331, "step": 11906 }, { "epoch": 0.68, "learning_rate": 5.035043890940042e-05, "loss": 0.7003, "step": 11907 }, { "epoch": 0.68, "learning_rate": 5.033449222882791e-05, "loss": 0.6995, "step": 11908 }, { "epoch": 0.68, "learning_rate": 5.0318547224596523e-05, "loss": 0.7167, "step": 11909 }, { "epoch": 0.68, "learning_rate": 5.0302603897244474e-05, "loss": 0.6943, "step": 11910 }, { "epoch": 0.68, "learning_rate": 5.028666224730987e-05, "loss": 0.7195, "step": 11911 }, { "epoch": 0.68, "learning_rate": 5.0270722275330804e-05, "loss": 0.6498, "step": 11912 }, { "epoch": 0.68, "learning_rate": 5.025478398184531e-05, "loss": 0.7312, "step": 11913 }, { "epoch": 0.68, "learning_rate": 5.023884736739132e-05, "loss": 0.691, "step": 11914 }, { "epoch": 0.68, "learning_rate": 5.0222912432506694e-05, "loss": 0.6488, "step": 11915 }, { "epoch": 0.68, "learning_rate": 5.0206979177729316e-05, "loss": 0.7184, "step": 11916 }, { "epoch": 0.68, "learning_rate": 5.019104760359699e-05, "loss": 0.7756, "step": 11917 }, { "epoch": 0.68, "learning_rate": 5.0175117710647436e-05, "loss": 0.6343, "step": 11918 }, { "epoch": 0.68, "learning_rate": 5.015918949941836e-05, "loss": 0.7574, "step": 11919 }, { "epoch": 0.68, "learning_rate": 5.0143262970447316e-05, "loss": 0.7135, "step": 11920 }, { "epoch": 0.68, "learning_rate": 5.012733812427194e-05, "loss": 0.7026, "step": 11921 }, { "epoch": 0.68, "learning_rate": 5.011141496142966e-05, "loss": 0.7477, "step": 11922 }, { "epoch": 0.68, "learning_rate": 5.0095493482457955e-05, "loss": 0.7014, "step": 11923 }, { "epoch": 0.68, "learning_rate": 5.007957368789422e-05, "loss": 0.6503, "step": 11924 }, { "epoch": 0.68, "learning_rate": 5.0063655578275814e-05, "loss": 0.7329, "step": 11925 }, { "epoch": 0.68, "learning_rate": 5.0047739154140025e-05, "loss": 0.7557, "step": 11926 }, { "epoch": 0.68, "learning_rate": 5.003182441602401e-05, "loss": 0.6976, "step": 11927 }, { "epoch": 0.68, "learning_rate": 5.0015911364464996e-05, "loss": 0.6919, "step": 11928 }, { "epoch": 0.68, "learning_rate": 5.000000000000002e-05, "loss": 0.7593, "step": 11929 }, { "epoch": 0.68, "learning_rate": 4.998409032316619e-05, "loss": 0.6945, "step": 11930 }, { "epoch": 0.68, "learning_rate": 4.996818233450047e-05, "loss": 0.7075, "step": 11931 }, { "epoch": 0.68, "learning_rate": 4.9952276034539815e-05, "loss": 0.6646, "step": 11932 }, { "epoch": 0.68, "learning_rate": 4.993637142382113e-05, "loss": 0.7183, "step": 11933 }, { "epoch": 0.68, "learning_rate": 4.992046850288118e-05, "loss": 0.7942, "step": 11934 }, { "epoch": 0.68, "learning_rate": 4.9904567272256745e-05, "loss": 0.677, "step": 11935 }, { "epoch": 0.68, "learning_rate": 4.988866773248458e-05, "loss": 0.7252, "step": 11936 }, { "epoch": 0.68, "learning_rate": 4.987276988410128e-05, "loss": 0.7644, "step": 11937 }, { "epoch": 0.68, "learning_rate": 4.985687372764345e-05, "loss": 0.7262, "step": 11938 }, { "epoch": 0.68, "learning_rate": 4.984097926364767e-05, "loss": 0.5627, "step": 11939 }, { "epoch": 0.68, "learning_rate": 4.9825086492650354e-05, "loss": 0.7309, "step": 11940 }, { "epoch": 0.68, "learning_rate": 4.980919541518796e-05, "loss": 0.7058, "step": 11941 }, { "epoch": 0.68, "learning_rate": 4.979330603179685e-05, "loss": 0.7085, "step": 11942 }, { "epoch": 0.68, "learning_rate": 4.977741834301337e-05, "loss": 0.7753, "step": 11943 }, { "epoch": 0.68, "learning_rate": 4.9761532349373696e-05, "loss": 0.7644, "step": 11944 }, { "epoch": 0.68, "learning_rate": 4.974564805141405e-05, "loss": 0.6925, "step": 11945 }, { "epoch": 0.68, "learning_rate": 4.972976544967063e-05, "loss": 0.6529, "step": 11946 }, { "epoch": 0.68, "learning_rate": 4.971388454467942e-05, "loss": 0.6744, "step": 11947 }, { "epoch": 0.68, "learning_rate": 4.969800533697649e-05, "loss": 0.7521, "step": 11948 }, { "epoch": 0.68, "learning_rate": 4.9682127827097814e-05, "loss": 0.7787, "step": 11949 }, { "epoch": 0.68, "learning_rate": 4.9666252015579305e-05, "loss": 0.7283, "step": 11950 }, { "epoch": 0.68, "learning_rate": 4.965037790295677e-05, "loss": 0.5381, "step": 11951 }, { "epoch": 0.68, "learning_rate": 4.963450548976606e-05, "loss": 0.5589, "step": 11952 }, { "epoch": 0.68, "learning_rate": 4.961863477654285e-05, "loss": 0.6692, "step": 11953 }, { "epoch": 0.68, "learning_rate": 4.960276576382283e-05, "loss": 0.7008, "step": 11954 }, { "epoch": 0.68, "learning_rate": 4.958689845214165e-05, "loss": 0.7382, "step": 11955 }, { "epoch": 0.68, "learning_rate": 4.957103284203486e-05, "loss": 0.7764, "step": 11956 }, { "epoch": 0.68, "learning_rate": 4.955516893403796e-05, "loss": 0.7126, "step": 11957 }, { "epoch": 0.68, "learning_rate": 4.9539306728686444e-05, "loss": 0.6219, "step": 11958 }, { "epoch": 0.68, "learning_rate": 4.952344622651566e-05, "loss": 0.6983, "step": 11959 }, { "epoch": 0.68, "learning_rate": 4.950758742806092e-05, "loss": 0.5974, "step": 11960 }, { "epoch": 0.68, "learning_rate": 4.949173033385751e-05, "loss": 0.762, "step": 11961 }, { "epoch": 0.68, "learning_rate": 4.947587494444067e-05, "loss": 0.6098, "step": 11962 }, { "epoch": 0.68, "learning_rate": 4.946002126034556e-05, "loss": 0.7258, "step": 11963 }, { "epoch": 0.68, "learning_rate": 4.944416928210731e-05, "loss": 0.718, "step": 11964 }, { "epoch": 0.68, "learning_rate": 4.9428319010260904e-05, "loss": 0.6148, "step": 11965 }, { "epoch": 0.68, "learning_rate": 4.9412470445341395e-05, "loss": 0.6882, "step": 11966 }, { "epoch": 0.68, "learning_rate": 4.939662358788364e-05, "loss": 0.5978, "step": 11967 }, { "epoch": 0.68, "learning_rate": 4.9380778438422546e-05, "loss": 0.6659, "step": 11968 }, { "epoch": 0.68, "learning_rate": 4.936493499749294e-05, "loss": 0.6818, "step": 11969 }, { "epoch": 0.68, "learning_rate": 4.934909326562958e-05, "loss": 0.628, "step": 11970 }, { "epoch": 0.68, "learning_rate": 4.9333253243367184e-05, "loss": 0.6781, "step": 11971 }, { "epoch": 0.68, "learning_rate": 4.9317414931240336e-05, "loss": 0.5681, "step": 11972 }, { "epoch": 0.68, "learning_rate": 4.9301578329783695e-05, "loss": 0.7649, "step": 11973 }, { "epoch": 0.68, "learning_rate": 4.9285743439531704e-05, "loss": 0.7083, "step": 11974 }, { "epoch": 0.68, "learning_rate": 4.926991026101888e-05, "loss": 0.6975, "step": 11975 }, { "epoch": 0.68, "learning_rate": 4.925407879477962e-05, "loss": 0.7851, "step": 11976 }, { "epoch": 0.68, "learning_rate": 4.923824904134829e-05, "loss": 0.7339, "step": 11977 }, { "epoch": 0.68, "learning_rate": 4.9222421001259235e-05, "loss": 0.6802, "step": 11978 }, { "epoch": 0.68, "learning_rate": 4.920659467504659e-05, "loss": 0.7246, "step": 11979 }, { "epoch": 0.68, "learning_rate": 4.919077006324462e-05, "loss": 0.7584, "step": 11980 }, { "epoch": 0.68, "learning_rate": 4.917494716638739e-05, "loss": 0.6633, "step": 11981 }, { "epoch": 0.68, "learning_rate": 4.9159125985008966e-05, "loss": 0.6304, "step": 11982 }, { "epoch": 0.68, "learning_rate": 4.914330651964339e-05, "loss": 0.7212, "step": 11983 }, { "epoch": 0.68, "learning_rate": 4.912748877082463e-05, "loss": 0.7044, "step": 11984 }, { "epoch": 0.68, "learning_rate": 4.91116727390865e-05, "loss": 0.7383, "step": 11985 }, { "epoch": 0.68, "learning_rate": 4.909585842496287e-05, "loss": 0.7114, "step": 11986 }, { "epoch": 0.68, "learning_rate": 4.908004582898753e-05, "loss": 0.6889, "step": 11987 }, { "epoch": 0.68, "learning_rate": 4.906423495169422e-05, "loss": 0.7259, "step": 11988 }, { "epoch": 0.68, "learning_rate": 4.904842579361653e-05, "loss": 0.7446, "step": 11989 }, { "epoch": 0.68, "learning_rate": 4.903261835528808e-05, "loss": 0.6899, "step": 11990 }, { "epoch": 0.68, "learning_rate": 4.9016812637242484e-05, "loss": 0.7669, "step": 11991 }, { "epoch": 0.68, "learning_rate": 4.9001008640013124e-05, "loss": 0.7171, "step": 11992 }, { "epoch": 0.68, "learning_rate": 4.898520636413348e-05, "loss": 0.7557, "step": 11993 }, { "epoch": 0.68, "learning_rate": 4.896940581013689e-05, "loss": 0.6428, "step": 11994 }, { "epoch": 0.68, "learning_rate": 4.895360697855674e-05, "loss": 0.7362, "step": 11995 }, { "epoch": 0.68, "learning_rate": 4.893780986992619e-05, "loss": 0.7808, "step": 11996 }, { "epoch": 0.68, "learning_rate": 4.892201448477851e-05, "loss": 0.7162, "step": 11997 }, { "epoch": 0.68, "learning_rate": 4.8906220823646765e-05, "loss": 0.6666, "step": 11998 }, { "epoch": 0.68, "learning_rate": 4.8890428887064055e-05, "loss": 0.7248, "step": 11999 }, { "epoch": 0.68, "learning_rate": 4.8874638675563425e-05, "loss": 0.6385, "step": 12000 }, { "epoch": 0.68, "learning_rate": 4.8858850189677806e-05, "loss": 0.5885, "step": 12001 }, { "epoch": 0.68, "learning_rate": 4.884306342994016e-05, "loss": 0.6971, "step": 12002 }, { "epoch": 0.68, "learning_rate": 4.882727839688325e-05, "loss": 0.6846, "step": 12003 }, { "epoch": 0.68, "learning_rate": 4.8811495091039926e-05, "loss": 0.7282, "step": 12004 }, { "epoch": 0.68, "learning_rate": 4.8795713512942865e-05, "loss": 0.7023, "step": 12005 }, { "epoch": 0.68, "learning_rate": 4.8779933663124755e-05, "loss": 0.7938, "step": 12006 }, { "epoch": 0.68, "learning_rate": 4.876415554211822e-05, "loss": 0.6643, "step": 12007 }, { "epoch": 0.68, "learning_rate": 4.8748379150455804e-05, "loss": 0.7709, "step": 12008 }, { "epoch": 0.68, "learning_rate": 4.8732604488670044e-05, "loss": 0.7438, "step": 12009 }, { "epoch": 0.68, "learning_rate": 4.8716831557293285e-05, "loss": 0.6329, "step": 12010 }, { "epoch": 0.68, "learning_rate": 4.870106035685801e-05, "loss": 0.6828, "step": 12011 }, { "epoch": 0.68, "learning_rate": 4.868529088789644e-05, "loss": 0.6632, "step": 12012 }, { "epoch": 0.68, "learning_rate": 4.866952315094088e-05, "loss": 0.7533, "step": 12013 }, { "epoch": 0.68, "learning_rate": 4.865375714652353e-05, "loss": 0.7722, "step": 12014 }, { "epoch": 0.68, "learning_rate": 4.863799287517654e-05, "loss": 0.8052, "step": 12015 }, { "epoch": 0.68, "learning_rate": 4.8622230337432026e-05, "loss": 0.6533, "step": 12016 }, { "epoch": 0.68, "learning_rate": 4.860646953382199e-05, "loss": 0.7251, "step": 12017 }, { "epoch": 0.68, "learning_rate": 4.859071046487834e-05, "loss": 0.651, "step": 12018 }, { "epoch": 0.68, "learning_rate": 4.857495313113305e-05, "loss": 0.6721, "step": 12019 }, { "epoch": 0.68, "learning_rate": 4.855919753311795e-05, "loss": 0.6889, "step": 12020 }, { "epoch": 0.68, "learning_rate": 4.854344367136485e-05, "loss": 0.6889, "step": 12021 }, { "epoch": 0.68, "learning_rate": 4.852769154640547e-05, "loss": 0.7514, "step": 12022 }, { "epoch": 0.68, "learning_rate": 4.851194115877153e-05, "loss": 0.7031, "step": 12023 }, { "epoch": 0.68, "learning_rate": 4.8496192508994576e-05, "loss": 0.6482, "step": 12024 }, { "epoch": 0.68, "learning_rate": 4.848044559760624e-05, "loss": 0.6428, "step": 12025 }, { "epoch": 0.68, "learning_rate": 4.846470042513793e-05, "loss": 0.776, "step": 12026 }, { "epoch": 0.68, "learning_rate": 4.844895699212114e-05, "loss": 0.6558, "step": 12027 }, { "epoch": 0.68, "learning_rate": 4.8433215299087256e-05, "loss": 0.6468, "step": 12028 }, { "epoch": 0.68, "learning_rate": 4.841747534656763e-05, "loss": 0.6865, "step": 12029 }, { "epoch": 0.68, "learning_rate": 4.840173713509345e-05, "loss": 0.7752, "step": 12030 }, { "epoch": 0.68, "learning_rate": 4.838600066519596e-05, "loss": 0.7626, "step": 12031 }, { "epoch": 0.68, "learning_rate": 4.837026593740636e-05, "loss": 0.636, "step": 12032 }, { "epoch": 0.68, "learning_rate": 4.8354532952255636e-05, "loss": 0.7458, "step": 12033 }, { "epoch": 0.68, "learning_rate": 4.833880171027487e-05, "loss": 0.6679, "step": 12034 }, { "epoch": 0.68, "learning_rate": 4.832307221199505e-05, "loss": 0.7549, "step": 12035 }, { "epoch": 0.68, "learning_rate": 4.8307344457947094e-05, "loss": 0.7677, "step": 12036 }, { "epoch": 0.68, "learning_rate": 4.829161844866179e-05, "loss": 0.7192, "step": 12037 }, { "epoch": 0.68, "learning_rate": 4.827589418466999e-05, "loss": 0.6379, "step": 12038 }, { "epoch": 0.68, "learning_rate": 4.826017166650244e-05, "loss": 0.664, "step": 12039 }, { "epoch": 0.68, "learning_rate": 4.824445089468975e-05, "loss": 0.7523, "step": 12040 }, { "epoch": 0.68, "learning_rate": 4.822873186976258e-05, "loss": 0.6768, "step": 12041 }, { "epoch": 0.68, "learning_rate": 4.821301459225153e-05, "loss": 0.644, "step": 12042 }, { "epoch": 0.68, "learning_rate": 4.8197299062686995e-05, "loss": 0.7464, "step": 12043 }, { "epoch": 0.68, "learning_rate": 4.81815852815995e-05, "loss": 0.6967, "step": 12044 }, { "epoch": 0.68, "learning_rate": 4.816587324951938e-05, "loss": 0.7479, "step": 12045 }, { "epoch": 0.68, "learning_rate": 4.815016296697699e-05, "loss": 0.7232, "step": 12046 }, { "epoch": 0.68, "learning_rate": 4.8134454434502605e-05, "loss": 0.6616, "step": 12047 }, { "epoch": 0.68, "learning_rate": 4.811874765262637e-05, "loss": 0.7772, "step": 12048 }, { "epoch": 0.68, "learning_rate": 4.810304262187852e-05, "loss": 0.6447, "step": 12049 }, { "epoch": 0.68, "learning_rate": 4.8087339342789026e-05, "loss": 0.7052, "step": 12050 }, { "epoch": 0.68, "learning_rate": 4.807163781588799e-05, "loss": 0.6778, "step": 12051 }, { "epoch": 0.68, "learning_rate": 4.805593804170536e-05, "loss": 0.7016, "step": 12052 }, { "epoch": 0.68, "learning_rate": 4.8040240020771054e-05, "loss": 0.7692, "step": 12053 }, { "epoch": 0.68, "learning_rate": 4.802454375361495e-05, "loss": 0.6898, "step": 12054 }, { "epoch": 0.68, "learning_rate": 4.800884924076677e-05, "loss": 0.6623, "step": 12055 }, { "epoch": 0.68, "learning_rate": 4.799315648275632e-05, "loss": 0.811, "step": 12056 }, { "epoch": 0.68, "learning_rate": 4.797746548011319e-05, "loss": 0.6231, "step": 12057 }, { "epoch": 0.68, "learning_rate": 4.796177623336703e-05, "loss": 0.7997, "step": 12058 }, { "epoch": 0.68, "learning_rate": 4.79460887430474e-05, "loss": 0.6959, "step": 12059 }, { "epoch": 0.68, "learning_rate": 4.79304030096838e-05, "loss": 0.6315, "step": 12060 }, { "epoch": 0.68, "learning_rate": 4.791471903380569e-05, "loss": 0.6526, "step": 12061 }, { "epoch": 0.68, "learning_rate": 4.78990368159424e-05, "loss": 0.7235, "step": 12062 }, { "epoch": 0.68, "learning_rate": 4.7883356356623235e-05, "loss": 0.6543, "step": 12063 }, { "epoch": 0.68, "learning_rate": 4.7867677656377465e-05, "loss": 0.6886, "step": 12064 }, { "epoch": 0.68, "learning_rate": 4.785200071573429e-05, "loss": 0.7012, "step": 12065 }, { "epoch": 0.68, "learning_rate": 4.783632553522287e-05, "loss": 0.6471, "step": 12066 }, { "epoch": 0.68, "learning_rate": 4.7820652115372253e-05, "loss": 0.7786, "step": 12067 }, { "epoch": 0.68, "learning_rate": 4.780498045671151e-05, "loss": 0.748, "step": 12068 }, { "epoch": 0.68, "learning_rate": 4.778931055976957e-05, "loss": 0.7145, "step": 12069 }, { "epoch": 0.68, "learning_rate": 4.777364242507529e-05, "loss": 0.7301, "step": 12070 }, { "epoch": 0.68, "learning_rate": 4.775797605315755e-05, "loss": 0.7298, "step": 12071 }, { "epoch": 0.68, "learning_rate": 4.774231144454512e-05, "loss": 0.6949, "step": 12072 }, { "epoch": 0.68, "learning_rate": 4.772664859976673e-05, "loss": 0.6373, "step": 12073 }, { "epoch": 0.68, "learning_rate": 4.771098751935108e-05, "loss": 0.7105, "step": 12074 }, { "epoch": 0.68, "learning_rate": 4.769532820382671e-05, "loss": 0.6823, "step": 12075 }, { "epoch": 0.69, "learning_rate": 4.767967065372221e-05, "loss": 0.7175, "step": 12076 }, { "epoch": 0.69, "learning_rate": 4.7664014869566e-05, "loss": 0.6769, "step": 12077 }, { "epoch": 0.69, "learning_rate": 4.7648360851886554e-05, "loss": 0.651, "step": 12078 }, { "epoch": 0.69, "learning_rate": 4.763270860121222e-05, "loss": 0.7543, "step": 12079 }, { "epoch": 0.69, "learning_rate": 4.761705811807131e-05, "loss": 0.7547, "step": 12080 }, { "epoch": 0.69, "learning_rate": 4.7601409402992106e-05, "loss": 0.7394, "step": 12081 }, { "epoch": 0.69, "learning_rate": 4.758576245650271e-05, "loss": 0.6404, "step": 12082 }, { "epoch": 0.69, "learning_rate": 4.757011727913131e-05, "loss": 0.7164, "step": 12083 }, { "epoch": 0.69, "learning_rate": 4.755447387140598e-05, "loss": 0.6789, "step": 12084 }, { "epoch": 0.69, "learning_rate": 4.753883223385467e-05, "loss": 0.6662, "step": 12085 }, { "epoch": 0.69, "learning_rate": 4.752319236700534e-05, "loss": 0.7078, "step": 12086 }, { "epoch": 0.69, "learning_rate": 4.750755427138595e-05, "loss": 0.6969, "step": 12087 }, { "epoch": 0.69, "learning_rate": 4.749191794752422e-05, "loss": 0.8292, "step": 12088 }, { "epoch": 0.69, "learning_rate": 4.7476283395947975e-05, "loss": 0.7419, "step": 12089 }, { "epoch": 0.69, "learning_rate": 4.746065061718491e-05, "loss": 0.7233, "step": 12090 }, { "epoch": 0.69, "learning_rate": 4.744501961176271e-05, "loss": 0.6905, "step": 12091 }, { "epoch": 0.69, "learning_rate": 4.7429390380208895e-05, "loss": 0.7537, "step": 12092 }, { "epoch": 0.69, "learning_rate": 4.741376292305103e-05, "loss": 0.6925, "step": 12093 }, { "epoch": 0.69, "learning_rate": 4.739813724081661e-05, "loss": 0.6733, "step": 12094 }, { "epoch": 0.69, "learning_rate": 4.738251333403298e-05, "loss": 0.7751, "step": 12095 }, { "epoch": 0.69, "learning_rate": 4.7366891203227515e-05, "loss": 0.6927, "step": 12096 }, { "epoch": 0.69, "learning_rate": 4.735127084892751e-05, "loss": 0.711, "step": 12097 }, { "epoch": 0.69, "learning_rate": 4.733565227166018e-05, "loss": 0.7882, "step": 12098 }, { "epoch": 0.69, "learning_rate": 4.732003547195275e-05, "loss": 0.745, "step": 12099 }, { "epoch": 0.69, "learning_rate": 4.7304420450332244e-05, "loss": 0.6889, "step": 12100 }, { "epoch": 0.69, "learning_rate": 4.728880720732578e-05, "loss": 0.868, "step": 12101 }, { "epoch": 0.69, "learning_rate": 4.7273195743460286e-05, "loss": 0.7346, "step": 12102 }, { "epoch": 0.69, "learning_rate": 4.72575860592627e-05, "loss": 0.914, "step": 12103 }, { "epoch": 0.69, "learning_rate": 4.7241978155259925e-05, "loss": 0.7446, "step": 12104 }, { "epoch": 0.69, "learning_rate": 4.7226372031978735e-05, "loss": 0.594, "step": 12105 }, { "epoch": 0.69, "learning_rate": 4.721076768994593e-05, "loss": 0.6519, "step": 12106 }, { "epoch": 0.69, "learning_rate": 4.719516512968818e-05, "loss": 0.7336, "step": 12107 }, { "epoch": 0.69, "learning_rate": 4.717956435173203e-05, "loss": 0.7246, "step": 12108 }, { "epoch": 0.69, "learning_rate": 4.7163965356604125e-05, "loss": 0.7293, "step": 12109 }, { "epoch": 0.69, "learning_rate": 4.7148368144830955e-05, "loss": 0.7318, "step": 12110 }, { "epoch": 0.69, "learning_rate": 4.713277271693897e-05, "loss": 0.6319, "step": 12111 }, { "epoch": 0.69, "learning_rate": 4.7117179073454554e-05, "loss": 0.7401, "step": 12112 }, { "epoch": 0.69, "learning_rate": 4.7101587214904064e-05, "loss": 0.7457, "step": 12113 }, { "epoch": 0.69, "learning_rate": 4.708599714181374e-05, "loss": 0.6659, "step": 12114 }, { "epoch": 0.69, "learning_rate": 4.707040885470976e-05, "loss": 0.6856, "step": 12115 }, { "epoch": 0.69, "learning_rate": 4.705482235411828e-05, "loss": 0.7239, "step": 12116 }, { "epoch": 0.69, "learning_rate": 4.70392376405654e-05, "loss": 0.7736, "step": 12117 }, { "epoch": 0.69, "learning_rate": 4.702365471457715e-05, "loss": 0.6284, "step": 12118 }, { "epoch": 0.69, "learning_rate": 4.700807357667952e-05, "loss": 0.6661, "step": 12119 }, { "epoch": 0.69, "learning_rate": 4.699249422739835e-05, "loss": 0.7689, "step": 12120 }, { "epoch": 0.69, "learning_rate": 4.697691666725955e-05, "loss": 0.733, "step": 12121 }, { "epoch": 0.69, "learning_rate": 4.696134089678884e-05, "loss": 0.597, "step": 12122 }, { "epoch": 0.69, "learning_rate": 4.694576691651198e-05, "loss": 0.797, "step": 12123 }, { "epoch": 0.69, "learning_rate": 4.693019472695461e-05, "loss": 0.6191, "step": 12124 }, { "epoch": 0.69, "learning_rate": 4.691462432864235e-05, "loss": 0.7777, "step": 12125 }, { "epoch": 0.69, "learning_rate": 4.6899055722100795e-05, "loss": 0.7174, "step": 12126 }, { "epoch": 0.69, "learning_rate": 4.688348890785532e-05, "loss": 0.6434, "step": 12127 }, { "epoch": 0.69, "learning_rate": 4.686792388643144e-05, "loss": 0.7121, "step": 12128 }, { "epoch": 0.69, "learning_rate": 4.685236065835443e-05, "loss": 0.7262, "step": 12129 }, { "epoch": 0.69, "learning_rate": 4.683679922414964e-05, "loss": 0.7628, "step": 12130 }, { "epoch": 0.69, "learning_rate": 4.68212395843423e-05, "loss": 0.6835, "step": 12131 }, { "epoch": 0.69, "learning_rate": 4.680568173945759e-05, "loss": 0.7406, "step": 12132 }, { "epoch": 0.69, "learning_rate": 4.679012569002066e-05, "loss": 0.7281, "step": 12133 }, { "epoch": 0.69, "learning_rate": 4.6774571436556504e-05, "loss": 0.717, "step": 12134 }, { "epoch": 0.69, "learning_rate": 4.6759018979590154e-05, "loss": 0.6595, "step": 12135 }, { "epoch": 0.69, "learning_rate": 4.674346831964657e-05, "loss": 0.7342, "step": 12136 }, { "epoch": 0.69, "learning_rate": 4.6727919457250566e-05, "loss": 0.7752, "step": 12137 }, { "epoch": 0.69, "learning_rate": 4.6712372392927e-05, "loss": 0.6948, "step": 12138 }, { "epoch": 0.69, "learning_rate": 4.669682712720065e-05, "loss": 0.7155, "step": 12139 }, { "epoch": 0.69, "learning_rate": 4.668128366059614e-05, "loss": 0.6794, "step": 12140 }, { "epoch": 0.69, "learning_rate": 4.6665741993638136e-05, "loss": 0.7426, "step": 12141 }, { "epoch": 0.69, "learning_rate": 4.6650202126851215e-05, "loss": 0.6958, "step": 12142 }, { "epoch": 0.69, "learning_rate": 4.6634664060759934e-05, "loss": 0.7299, "step": 12143 }, { "epoch": 0.69, "learning_rate": 4.6619127795888653e-05, "loss": 0.7004, "step": 12144 }, { "epoch": 0.69, "learning_rate": 4.6603593332761796e-05, "loss": 0.742, "step": 12145 }, { "epoch": 0.69, "learning_rate": 4.6588060671903755e-05, "loss": 0.8108, "step": 12146 }, { "epoch": 0.69, "learning_rate": 4.6572529813838696e-05, "loss": 0.6597, "step": 12147 }, { "epoch": 0.69, "learning_rate": 4.655700075909087e-05, "loss": 0.7934, "step": 12148 }, { "epoch": 0.69, "learning_rate": 4.654147350818444e-05, "loss": 0.6489, "step": 12149 }, { "epoch": 0.69, "learning_rate": 4.652594806164348e-05, "loss": 0.6179, "step": 12150 }, { "epoch": 0.69, "learning_rate": 4.651042441999205e-05, "loss": 0.7255, "step": 12151 }, { "epoch": 0.69, "learning_rate": 4.649490258375406e-05, "loss": 0.6717, "step": 12152 }, { "epoch": 0.69, "learning_rate": 4.647938255345341e-05, "loss": 0.8025, "step": 12153 }, { "epoch": 0.69, "learning_rate": 4.646386432961396e-05, "loss": 0.7305, "step": 12154 }, { "epoch": 0.69, "learning_rate": 4.6448347912759494e-05, "loss": 0.7209, "step": 12155 }, { "epoch": 0.69, "learning_rate": 4.6432833303413725e-05, "loss": 0.6767, "step": 12156 }, { "epoch": 0.69, "learning_rate": 4.6417320502100316e-05, "loss": 0.831, "step": 12157 }, { "epoch": 0.69, "learning_rate": 4.6401809509342905e-05, "loss": 0.7376, "step": 12158 }, { "epoch": 0.69, "learning_rate": 4.638630032566499e-05, "loss": 0.7427, "step": 12159 }, { "epoch": 0.69, "learning_rate": 4.6370792951590015e-05, "loss": 0.7373, "step": 12160 }, { "epoch": 0.69, "learning_rate": 4.6355287387641414e-05, "loss": 0.7213, "step": 12161 }, { "epoch": 0.69, "learning_rate": 4.633978363434256e-05, "loss": 0.742, "step": 12162 }, { "epoch": 0.69, "learning_rate": 4.632428169221673e-05, "loss": 0.7148, "step": 12163 }, { "epoch": 0.69, "learning_rate": 4.6308781561787203e-05, "loss": 0.6629, "step": 12164 }, { "epoch": 0.69, "learning_rate": 4.6293283243577076e-05, "loss": 0.6889, "step": 12165 }, { "epoch": 0.69, "learning_rate": 4.627778673810953e-05, "loss": 0.6391, "step": 12166 }, { "epoch": 0.69, "learning_rate": 4.626229204590753e-05, "loss": 0.7113, "step": 12167 }, { "epoch": 0.69, "learning_rate": 4.62467991674941e-05, "loss": 0.7977, "step": 12168 }, { "epoch": 0.69, "learning_rate": 4.623130810339219e-05, "loss": 0.6668, "step": 12169 }, { "epoch": 0.69, "learning_rate": 4.621581885412463e-05, "loss": 0.7063, "step": 12170 }, { "epoch": 0.69, "learning_rate": 4.620033142021427e-05, "loss": 0.7519, "step": 12171 }, { "epoch": 0.69, "learning_rate": 4.6184845802183795e-05, "loss": 0.6517, "step": 12172 }, { "epoch": 0.69, "learning_rate": 4.616936200055594e-05, "loss": 0.7555, "step": 12173 }, { "epoch": 0.69, "learning_rate": 4.6153880015853256e-05, "loss": 0.6979, "step": 12174 }, { "epoch": 0.69, "learning_rate": 4.613839984859835e-05, "loss": 0.7334, "step": 12175 }, { "epoch": 0.69, "learning_rate": 4.612292149931369e-05, "loss": 0.6948, "step": 12176 }, { "epoch": 0.69, "learning_rate": 4.610744496852174e-05, "loss": 0.6205, "step": 12177 }, { "epoch": 0.69, "learning_rate": 4.6091970256744885e-05, "loss": 0.7013, "step": 12178 }, { "epoch": 0.69, "learning_rate": 4.6076497364505386e-05, "loss": 0.7017, "step": 12179 }, { "epoch": 0.69, "learning_rate": 4.606102629232556e-05, "loss": 0.703, "step": 12180 }, { "epoch": 0.69, "learning_rate": 4.6045557040727507e-05, "loss": 0.6796, "step": 12181 }, { "epoch": 0.69, "learning_rate": 4.6030089610233416e-05, "loss": 0.8212, "step": 12182 }, { "epoch": 0.69, "learning_rate": 4.601462400136533e-05, "loss": 0.7192, "step": 12183 }, { "epoch": 0.69, "learning_rate": 4.599916021464531e-05, "loss": 0.7502, "step": 12184 }, { "epoch": 0.69, "learning_rate": 4.598369825059522e-05, "loss": 0.7208, "step": 12185 }, { "epoch": 0.69, "learning_rate": 4.596823810973698e-05, "loss": 0.7754, "step": 12186 }, { "epoch": 0.69, "learning_rate": 4.5952779792592406e-05, "loss": 0.6903, "step": 12187 }, { "epoch": 0.69, "learning_rate": 4.593732329968329e-05, "loss": 0.7449, "step": 12188 }, { "epoch": 0.69, "learning_rate": 4.592186863153126e-05, "loss": 0.617, "step": 12189 }, { "epoch": 0.69, "learning_rate": 4.5906415788657984e-05, "loss": 0.7284, "step": 12190 }, { "epoch": 0.69, "learning_rate": 4.5890964771585086e-05, "loss": 0.7031, "step": 12191 }, { "epoch": 0.69, "learning_rate": 4.587551558083399e-05, "loss": 0.6273, "step": 12192 }, { "epoch": 0.69, "learning_rate": 4.58600682169262e-05, "loss": 0.7272, "step": 12193 }, { "epoch": 0.69, "learning_rate": 4.584462268038309e-05, "loss": 0.6937, "step": 12194 }, { "epoch": 0.69, "learning_rate": 4.582917897172603e-05, "loss": 0.6916, "step": 12195 }, { "epoch": 0.69, "learning_rate": 4.5813737091476203e-05, "loss": 0.7136, "step": 12196 }, { "epoch": 0.69, "learning_rate": 4.57982970401549e-05, "loss": 0.7079, "step": 12197 }, { "epoch": 0.69, "learning_rate": 4.578285881828319e-05, "loss": 0.7019, "step": 12198 }, { "epoch": 0.69, "learning_rate": 4.576742242638218e-05, "loss": 0.6764, "step": 12199 }, { "epoch": 0.69, "learning_rate": 4.575198786497289e-05, "loss": 0.6446, "step": 12200 }, { "epoch": 0.69, "learning_rate": 4.5736555134576285e-05, "loss": 0.7575, "step": 12201 }, { "epoch": 0.69, "learning_rate": 4.5721124235713295e-05, "loss": 0.7236, "step": 12202 }, { "epoch": 0.69, "learning_rate": 4.570569516890467e-05, "loss": 0.8382, "step": 12203 }, { "epoch": 0.69, "learning_rate": 4.569026793467126e-05, "loss": 0.6111, "step": 12204 }, { "epoch": 0.69, "learning_rate": 4.5674842533533714e-05, "loss": 0.7072, "step": 12205 }, { "epoch": 0.69, "learning_rate": 4.565941896601269e-05, "loss": 0.6615, "step": 12206 }, { "epoch": 0.69, "learning_rate": 4.56439972326288e-05, "loss": 0.7215, "step": 12207 }, { "epoch": 0.69, "learning_rate": 4.562857733390256e-05, "loss": 0.7419, "step": 12208 }, { "epoch": 0.69, "learning_rate": 4.561315927035446e-05, "loss": 0.735, "step": 12209 }, { "epoch": 0.69, "learning_rate": 4.5597743042504835e-05, "loss": 0.579, "step": 12210 }, { "epoch": 0.69, "learning_rate": 4.55823286508741e-05, "loss": 0.6614, "step": 12211 }, { "epoch": 0.69, "learning_rate": 4.5566916095982436e-05, "loss": 0.6655, "step": 12212 }, { "epoch": 0.69, "learning_rate": 4.5551505378350115e-05, "loss": 0.6737, "step": 12213 }, { "epoch": 0.69, "learning_rate": 4.5536096498497295e-05, "loss": 0.743, "step": 12214 }, { "epoch": 0.69, "learning_rate": 4.552068945694404e-05, "loss": 0.6868, "step": 12215 }, { "epoch": 0.69, "learning_rate": 4.550528425421043e-05, "loss": 0.7246, "step": 12216 }, { "epoch": 0.69, "learning_rate": 4.5489880890816404e-05, "loss": 0.5469, "step": 12217 }, { "epoch": 0.69, "learning_rate": 4.5474479367281816e-05, "loss": 0.658, "step": 12218 }, { "epoch": 0.69, "learning_rate": 4.545907968412655e-05, "loss": 0.7269, "step": 12219 }, { "epoch": 0.69, "learning_rate": 4.544368184187039e-05, "loss": 0.791, "step": 12220 }, { "epoch": 0.69, "learning_rate": 4.542828584103304e-05, "loss": 0.6898, "step": 12221 }, { "epoch": 0.69, "learning_rate": 4.5412891682134154e-05, "loss": 0.5937, "step": 12222 }, { "epoch": 0.69, "learning_rate": 4.5397499365693374e-05, "loss": 0.6778, "step": 12223 }, { "epoch": 0.69, "learning_rate": 4.538210889223016e-05, "loss": 0.6972, "step": 12224 }, { "epoch": 0.69, "learning_rate": 4.536672026226405e-05, "loss": 0.7516, "step": 12225 }, { "epoch": 0.69, "learning_rate": 4.535133347631437e-05, "loss": 0.7816, "step": 12226 }, { "epoch": 0.69, "learning_rate": 4.5335948534900505e-05, "loss": 0.6942, "step": 12227 }, { "epoch": 0.69, "learning_rate": 4.532056543854174e-05, "loss": 0.68, "step": 12228 }, { "epoch": 0.69, "learning_rate": 4.530518418775733e-05, "loss": 0.7456, "step": 12229 }, { "epoch": 0.69, "learning_rate": 4.5289804783066366e-05, "loss": 0.6531, "step": 12230 }, { "epoch": 0.69, "learning_rate": 4.527442722498797e-05, "loss": 0.6255, "step": 12231 }, { "epoch": 0.69, "learning_rate": 4.52590515140412e-05, "loss": 0.7024, "step": 12232 }, { "epoch": 0.69, "learning_rate": 4.524367765074499e-05, "loss": 0.6309, "step": 12233 }, { "epoch": 0.69, "learning_rate": 4.522830563561825e-05, "loss": 0.7441, "step": 12234 }, { "epoch": 0.69, "learning_rate": 4.521293546917985e-05, "loss": 0.7425, "step": 12235 }, { "epoch": 0.69, "learning_rate": 4.5197567151948595e-05, "loss": 0.5867, "step": 12236 }, { "epoch": 0.69, "learning_rate": 4.518220068444313e-05, "loss": 0.7041, "step": 12237 }, { "epoch": 0.69, "learning_rate": 4.516683606718216e-05, "loss": 0.7056, "step": 12238 }, { "epoch": 0.69, "learning_rate": 4.515147330068431e-05, "loss": 0.7444, "step": 12239 }, { "epoch": 0.69, "learning_rate": 4.5136112385468054e-05, "loss": 0.7269, "step": 12240 }, { "epoch": 0.69, "learning_rate": 4.512075332205189e-05, "loss": 0.6936, "step": 12241 }, { "epoch": 0.69, "learning_rate": 4.5105396110954254e-05, "loss": 0.6589, "step": 12242 }, { "epoch": 0.69, "learning_rate": 4.509004075269344e-05, "loss": 0.7375, "step": 12243 }, { "epoch": 0.69, "learning_rate": 4.507468724778775e-05, "loss": 0.7277, "step": 12244 }, { "epoch": 0.69, "learning_rate": 4.505933559675541e-05, "loss": 0.7873, "step": 12245 }, { "epoch": 0.69, "learning_rate": 4.5043985800114596e-05, "loss": 0.6744, "step": 12246 }, { "epoch": 0.69, "learning_rate": 4.502863785838342e-05, "loss": 0.7702, "step": 12247 }, { "epoch": 0.69, "learning_rate": 4.5013291772079835e-05, "loss": 0.7284, "step": 12248 }, { "epoch": 0.69, "learning_rate": 4.499794754172191e-05, "loss": 0.6783, "step": 12249 }, { "epoch": 0.69, "learning_rate": 4.498260516782746e-05, "loss": 0.6751, "step": 12250 }, { "epoch": 0.69, "learning_rate": 4.4967264650914375e-05, "loss": 0.6525, "step": 12251 }, { "epoch": 0.69, "learning_rate": 4.495192599150044e-05, "loss": 0.6362, "step": 12252 }, { "epoch": 0.7, "learning_rate": 4.4936589190103375e-05, "loss": 0.584, "step": 12253 }, { "epoch": 0.7, "learning_rate": 4.492125424724086e-05, "loss": 0.7506, "step": 12254 }, { "epoch": 0.7, "learning_rate": 4.490592116343043e-05, "loss": 0.6902, "step": 12255 }, { "epoch": 0.7, "learning_rate": 4.489058993918969e-05, "loss": 0.7567, "step": 12256 }, { "epoch": 0.7, "learning_rate": 4.487526057503603e-05, "loss": 0.6649, "step": 12257 }, { "epoch": 0.7, "learning_rate": 4.48599330714869e-05, "loss": 0.7054, "step": 12258 }, { "epoch": 0.7, "learning_rate": 4.484460742905964e-05, "loss": 0.6655, "step": 12259 }, { "epoch": 0.7, "learning_rate": 4.4829283648271524e-05, "loss": 0.7339, "step": 12260 }, { "epoch": 0.7, "learning_rate": 4.4813961729639807e-05, "loss": 0.7071, "step": 12261 }, { "epoch": 0.7, "learning_rate": 4.479864167368162e-05, "loss": 0.7239, "step": 12262 }, { "epoch": 0.7, "learning_rate": 4.4783323480914006e-05, "loss": 0.6899, "step": 12263 }, { "epoch": 0.7, "learning_rate": 4.476800715185404e-05, "loss": 0.6091, "step": 12264 }, { "epoch": 0.7, "learning_rate": 4.475269268701868e-05, "loss": 0.6519, "step": 12265 }, { "epoch": 0.7, "learning_rate": 4.473738008692484e-05, "loss": 0.6144, "step": 12266 }, { "epoch": 0.7, "learning_rate": 4.4722069352089355e-05, "loss": 0.7454, "step": 12267 }, { "epoch": 0.7, "learning_rate": 4.470676048302903e-05, "loss": 0.6767, "step": 12268 }, { "epoch": 0.7, "learning_rate": 4.469145348026056e-05, "loss": 0.6983, "step": 12269 }, { "epoch": 0.7, "learning_rate": 4.467614834430054e-05, "loss": 0.7261, "step": 12270 }, { "epoch": 0.7, "learning_rate": 4.46608450756656e-05, "loss": 0.6708, "step": 12271 }, { "epoch": 0.7, "learning_rate": 4.464554367487228e-05, "loss": 0.7367, "step": 12272 }, { "epoch": 0.7, "learning_rate": 4.463024414243704e-05, "loss": 0.64, "step": 12273 }, { "epoch": 0.7, "learning_rate": 4.461494647887631e-05, "loss": 0.7056, "step": 12274 }, { "epoch": 0.7, "learning_rate": 4.459965068470634e-05, "loss": 0.7193, "step": 12275 }, { "epoch": 0.7, "learning_rate": 4.458435676044349e-05, "loss": 0.7324, "step": 12276 }, { "epoch": 0.7, "learning_rate": 4.4569064706603905e-05, "loss": 0.6334, "step": 12277 }, { "epoch": 0.7, "learning_rate": 4.455377452370375e-05, "loss": 0.7654, "step": 12278 }, { "epoch": 0.7, "learning_rate": 4.453848621225912e-05, "loss": 0.7985, "step": 12279 }, { "epoch": 0.7, "learning_rate": 4.452319977278603e-05, "loss": 0.729, "step": 12280 }, { "epoch": 0.7, "learning_rate": 4.450791520580049e-05, "loss": 0.5925, "step": 12281 }, { "epoch": 0.7, "learning_rate": 4.44926325118183e-05, "loss": 0.6817, "step": 12282 }, { "epoch": 0.7, "learning_rate": 4.447735169135533e-05, "loss": 0.7391, "step": 12283 }, { "epoch": 0.7, "learning_rate": 4.4462072744927394e-05, "loss": 0.6903, "step": 12284 }, { "epoch": 0.7, "learning_rate": 4.444679567305013e-05, "loss": 0.6582, "step": 12285 }, { "epoch": 0.7, "learning_rate": 4.44315204762392e-05, "loss": 0.6915, "step": 12286 }, { "epoch": 0.7, "learning_rate": 4.4416247155010226e-05, "loss": 0.7333, "step": 12287 }, { "epoch": 0.7, "learning_rate": 4.4400975709878644e-05, "loss": 0.6805, "step": 12288 }, { "epoch": 0.7, "learning_rate": 4.438570614135994e-05, "loss": 0.7002, "step": 12289 }, { "epoch": 0.7, "learning_rate": 4.437043844996952e-05, "loss": 0.7732, "step": 12290 }, { "epoch": 0.7, "learning_rate": 4.435517263622272e-05, "loss": 0.6775, "step": 12291 }, { "epoch": 0.7, "learning_rate": 4.433990870063475e-05, "loss": 0.634, "step": 12292 }, { "epoch": 0.7, "learning_rate": 4.432464664372083e-05, "loss": 0.6483, "step": 12293 }, { "epoch": 0.7, "learning_rate": 4.430938646599614e-05, "loss": 0.7542, "step": 12294 }, { "epoch": 0.7, "learning_rate": 4.4294128167975676e-05, "loss": 0.6708, "step": 12295 }, { "epoch": 0.7, "learning_rate": 4.427887175017448e-05, "loss": 0.6896, "step": 12296 }, { "epoch": 0.7, "learning_rate": 4.4263617213107486e-05, "loss": 0.7996, "step": 12297 }, { "epoch": 0.7, "learning_rate": 4.42483645572896e-05, "loss": 0.6379, "step": 12298 }, { "epoch": 0.7, "learning_rate": 4.423311378323565e-05, "loss": 0.6976, "step": 12299 }, { "epoch": 0.7, "learning_rate": 4.421786489146034e-05, "loss": 0.6885, "step": 12300 }, { "epoch": 0.7, "learning_rate": 4.4202617882478405e-05, "loss": 0.7112, "step": 12301 }, { "epoch": 0.7, "learning_rate": 4.418737275680442e-05, "loss": 0.7251, "step": 12302 }, { "epoch": 0.7, "learning_rate": 4.417212951495298e-05, "loss": 0.7037, "step": 12303 }, { "epoch": 0.7, "learning_rate": 4.415688815743858e-05, "loss": 0.7562, "step": 12304 }, { "epoch": 0.7, "learning_rate": 4.4141648684775664e-05, "loss": 0.7168, "step": 12305 }, { "epoch": 0.7, "learning_rate": 4.412641109747863e-05, "loss": 0.6512, "step": 12306 }, { "epoch": 0.7, "learning_rate": 4.4111175396061754e-05, "loss": 0.7163, "step": 12307 }, { "epoch": 0.7, "learning_rate": 4.409594158103924e-05, "loss": 0.6615, "step": 12308 }, { "epoch": 0.7, "learning_rate": 4.4080709652925336e-05, "loss": 0.7349, "step": 12309 }, { "epoch": 0.7, "learning_rate": 4.4065479612234116e-05, "loss": 0.6182, "step": 12310 }, { "epoch": 0.7, "learning_rate": 4.405025145947965e-05, "loss": 0.6391, "step": 12311 }, { "epoch": 0.7, "learning_rate": 4.4035025195175937e-05, "loss": 0.7097, "step": 12312 }, { "epoch": 0.7, "learning_rate": 4.401980081983693e-05, "loss": 0.6281, "step": 12313 }, { "epoch": 0.7, "learning_rate": 4.400457833397647e-05, "loss": 0.7348, "step": 12314 }, { "epoch": 0.7, "learning_rate": 4.3989357738108297e-05, "loss": 0.6579, "step": 12315 }, { "epoch": 0.7, "learning_rate": 4.3974139032746195e-05, "loss": 0.6456, "step": 12316 }, { "epoch": 0.7, "learning_rate": 4.3958922218403844e-05, "loss": 0.7625, "step": 12317 }, { "epoch": 0.7, "learning_rate": 4.394370729559483e-05, "loss": 0.6903, "step": 12318 }, { "epoch": 0.7, "learning_rate": 4.392849426483274e-05, "loss": 0.73, "step": 12319 }, { "epoch": 0.7, "learning_rate": 4.3913283126631e-05, "loss": 0.6703, "step": 12320 }, { "epoch": 0.7, "learning_rate": 4.3898073881503074e-05, "loss": 0.6361, "step": 12321 }, { "epoch": 0.7, "learning_rate": 4.388286652996225e-05, "loss": 0.6341, "step": 12322 }, { "epoch": 0.7, "learning_rate": 4.3867661072521857e-05, "loss": 0.7654, "step": 12323 }, { "epoch": 0.7, "learning_rate": 4.385245750969511e-05, "loss": 0.7567, "step": 12324 }, { "epoch": 0.7, "learning_rate": 4.3837255841995186e-05, "loss": 0.7023, "step": 12325 }, { "epoch": 0.7, "learning_rate": 4.382205606993519e-05, "loss": 0.7798, "step": 12326 }, { "epoch": 0.7, "learning_rate": 4.380685819402811e-05, "loss": 0.7105, "step": 12327 }, { "epoch": 0.7, "learning_rate": 4.379166221478697e-05, "loss": 0.726, "step": 12328 }, { "epoch": 0.7, "learning_rate": 4.3776468132724604e-05, "loss": 0.7611, "step": 12329 }, { "epoch": 0.7, "learning_rate": 4.37612759483539e-05, "loss": 0.7036, "step": 12330 }, { "epoch": 0.7, "learning_rate": 4.374608566218761e-05, "loss": 0.7856, "step": 12331 }, { "epoch": 0.7, "learning_rate": 4.37308972747385e-05, "loss": 0.6384, "step": 12332 }, { "epoch": 0.7, "learning_rate": 4.371571078651915e-05, "loss": 0.662, "step": 12333 }, { "epoch": 0.7, "learning_rate": 4.3700526198042155e-05, "loss": 0.763, "step": 12334 }, { "epoch": 0.7, "learning_rate": 4.3685343509820075e-05, "loss": 0.7125, "step": 12335 }, { "epoch": 0.7, "learning_rate": 4.367016272236536e-05, "loss": 0.6813, "step": 12336 }, { "epoch": 0.7, "learning_rate": 4.365498383619036e-05, "loss": 0.8053, "step": 12337 }, { "epoch": 0.7, "learning_rate": 4.363980685180742e-05, "loss": 0.7198, "step": 12338 }, { "epoch": 0.7, "learning_rate": 4.362463176972884e-05, "loss": 0.7248, "step": 12339 }, { "epoch": 0.7, "learning_rate": 4.360945859046676e-05, "loss": 0.6878, "step": 12340 }, { "epoch": 0.7, "learning_rate": 4.359428731453333e-05, "loss": 0.7188, "step": 12341 }, { "epoch": 0.7, "learning_rate": 4.357911794244064e-05, "loss": 0.755, "step": 12342 }, { "epoch": 0.7, "learning_rate": 4.356395047470073e-05, "loss": 0.6698, "step": 12343 }, { "epoch": 0.7, "learning_rate": 4.354878491182546e-05, "loss": 0.5368, "step": 12344 }, { "epoch": 0.7, "learning_rate": 4.353362125432674e-05, "loss": 0.6893, "step": 12345 }, { "epoch": 0.7, "learning_rate": 4.3518459502716434e-05, "loss": 0.7803, "step": 12346 }, { "epoch": 0.7, "learning_rate": 4.350329965750621e-05, "loss": 0.7567, "step": 12347 }, { "epoch": 0.7, "learning_rate": 4.3488141719207795e-05, "loss": 0.6412, "step": 12348 }, { "epoch": 0.7, "learning_rate": 4.3472985688332815e-05, "loss": 0.6973, "step": 12349 }, { "epoch": 0.7, "learning_rate": 4.345783156539279e-05, "loss": 0.7274, "step": 12350 }, { "epoch": 0.7, "learning_rate": 4.3442679350899295e-05, "loss": 0.7464, "step": 12351 }, { "epoch": 0.7, "learning_rate": 4.342752904536369e-05, "loss": 0.6399, "step": 12352 }, { "epoch": 0.7, "learning_rate": 4.341238064929731e-05, "loss": 0.8007, "step": 12353 }, { "epoch": 0.7, "learning_rate": 4.3397234163211483e-05, "loss": 0.656, "step": 12354 }, { "epoch": 0.7, "learning_rate": 4.338208958761747e-05, "loss": 0.727, "step": 12355 }, { "epoch": 0.7, "learning_rate": 4.33669469230264e-05, "loss": 0.7362, "step": 12356 }, { "epoch": 0.7, "learning_rate": 4.3351806169949406e-05, "loss": 0.7065, "step": 12357 }, { "epoch": 0.7, "learning_rate": 4.333666732889755e-05, "loss": 0.7706, "step": 12358 }, { "epoch": 0.7, "learning_rate": 4.3321530400381785e-05, "loss": 0.7136, "step": 12359 }, { "epoch": 0.7, "learning_rate": 4.330639538491297e-05, "loss": 0.6931, "step": 12360 }, { "epoch": 0.7, "learning_rate": 4.329126228300199e-05, "loss": 0.6726, "step": 12361 }, { "epoch": 0.7, "learning_rate": 4.327613109515965e-05, "loss": 0.7509, "step": 12362 }, { "epoch": 0.7, "learning_rate": 4.326100182189663e-05, "loss": 0.7667, "step": 12363 }, { "epoch": 0.7, "learning_rate": 4.3245874463723645e-05, "loss": 0.7078, "step": 12364 }, { "epoch": 0.7, "learning_rate": 4.323074902115124e-05, "loss": 0.7726, "step": 12365 }, { "epoch": 0.7, "learning_rate": 4.32156254946899e-05, "loss": 0.6397, "step": 12366 }, { "epoch": 0.7, "learning_rate": 4.320050388485013e-05, "loss": 0.7, "step": 12367 }, { "epoch": 0.7, "learning_rate": 4.318538419214231e-05, "loss": 0.6626, "step": 12368 }, { "epoch": 0.7, "learning_rate": 4.3170266417076776e-05, "loss": 0.7046, "step": 12369 }, { "epoch": 0.7, "learning_rate": 4.315515056016379e-05, "loss": 0.7386, "step": 12370 }, { "epoch": 0.7, "learning_rate": 4.3140036621913595e-05, "loss": 0.8251, "step": 12371 }, { "epoch": 0.7, "learning_rate": 4.3124924602836245e-05, "loss": 0.6255, "step": 12372 }, { "epoch": 0.7, "learning_rate": 4.310981450344189e-05, "loss": 0.688, "step": 12373 }, { "epoch": 0.7, "learning_rate": 4.3094706324240454e-05, "loss": 0.7595, "step": 12374 }, { "epoch": 0.7, "learning_rate": 4.307960006574192e-05, "loss": 0.7337, "step": 12375 }, { "epoch": 0.7, "learning_rate": 4.306449572845617e-05, "loss": 0.6994, "step": 12376 }, { "epoch": 0.7, "learning_rate": 4.304939331289304e-05, "loss": 0.6932, "step": 12377 }, { "epoch": 0.7, "learning_rate": 4.3034292819562206e-05, "loss": 0.6802, "step": 12378 }, { "epoch": 0.7, "learning_rate": 4.301919424897338e-05, "loss": 0.7222, "step": 12379 }, { "epoch": 0.7, "learning_rate": 4.300409760163624e-05, "loss": 0.8005, "step": 12380 }, { "epoch": 0.7, "learning_rate": 4.2989002878060235e-05, "loss": 0.6351, "step": 12381 }, { "epoch": 0.7, "learning_rate": 4.29739100787549e-05, "loss": 0.7075, "step": 12382 }, { "epoch": 0.7, "learning_rate": 4.295881920422965e-05, "loss": 0.7419, "step": 12383 }, { "epoch": 0.7, "learning_rate": 4.29437302549939e-05, "loss": 0.7249, "step": 12384 }, { "epoch": 0.7, "learning_rate": 4.2928643231556844e-05, "loss": 0.7151, "step": 12385 }, { "epoch": 0.7, "learning_rate": 4.2913558134427754e-05, "loss": 0.6821, "step": 12386 }, { "epoch": 0.7, "learning_rate": 4.2898474964115796e-05, "loss": 0.6511, "step": 12387 }, { "epoch": 0.7, "learning_rate": 4.2883393721130096e-05, "loss": 0.6985, "step": 12388 }, { "epoch": 0.7, "learning_rate": 4.2868314405979614e-05, "loss": 0.6082, "step": 12389 }, { "epoch": 0.7, "learning_rate": 4.2853237019173356e-05, "loss": 0.633, "step": 12390 }, { "epoch": 0.7, "learning_rate": 4.2838161561220245e-05, "loss": 0.7879, "step": 12391 }, { "epoch": 0.7, "learning_rate": 4.282308803262908e-05, "loss": 0.6787, "step": 12392 }, { "epoch": 0.7, "learning_rate": 4.2808016433908624e-05, "loss": 0.7082, "step": 12393 }, { "epoch": 0.7, "learning_rate": 4.279294676556762e-05, "loss": 0.6805, "step": 12394 }, { "epoch": 0.7, "learning_rate": 4.277787902811473e-05, "loss": 0.7968, "step": 12395 }, { "epoch": 0.7, "learning_rate": 4.276281322205845e-05, "loss": 0.7132, "step": 12396 }, { "epoch": 0.7, "learning_rate": 4.274774934790737e-05, "loss": 0.638, "step": 12397 }, { "epoch": 0.7, "learning_rate": 4.273268740616987e-05, "loss": 0.7464, "step": 12398 }, { "epoch": 0.7, "learning_rate": 4.2717627397354356e-05, "loss": 0.6747, "step": 12399 }, { "epoch": 0.7, "learning_rate": 4.2702569321969145e-05, "loss": 0.7519, "step": 12400 }, { "epoch": 0.7, "learning_rate": 4.2687513180522484e-05, "loss": 0.7547, "step": 12401 }, { "epoch": 0.7, "learning_rate": 4.26724589735226e-05, "loss": 0.6987, "step": 12402 }, { "epoch": 0.7, "learning_rate": 4.265740670147753e-05, "loss": 0.7072, "step": 12403 }, { "epoch": 0.7, "learning_rate": 4.264235636489542e-05, "loss": 0.6366, "step": 12404 }, { "epoch": 0.7, "learning_rate": 4.262730796428417e-05, "loss": 0.7059, "step": 12405 }, { "epoch": 0.7, "learning_rate": 4.261226150015173e-05, "loss": 0.8076, "step": 12406 }, { "epoch": 0.7, "learning_rate": 4.259721697300598e-05, "loss": 0.7146, "step": 12407 }, { "epoch": 0.7, "learning_rate": 4.258217438335469e-05, "loss": 0.7824, "step": 12408 }, { "epoch": 0.7, "learning_rate": 4.256713373170564e-05, "loss": 0.8215, "step": 12409 }, { "epoch": 0.7, "learning_rate": 4.255209501856646e-05, "loss": 0.7326, "step": 12410 }, { "epoch": 0.7, "learning_rate": 4.253705824444469e-05, "loss": 0.5662, "step": 12411 }, { "epoch": 0.7, "learning_rate": 4.252202340984791e-05, "loss": 0.682, "step": 12412 }, { "epoch": 0.7, "learning_rate": 4.250699051528357e-05, "loss": 0.7793, "step": 12413 }, { "epoch": 0.7, "learning_rate": 4.24919595612591e-05, "loss": 0.6801, "step": 12414 }, { "epoch": 0.7, "learning_rate": 4.24769305482818e-05, "loss": 0.7557, "step": 12415 }, { "epoch": 0.7, "learning_rate": 4.2461903476859e-05, "loss": 0.713, "step": 12416 }, { "epoch": 0.7, "learning_rate": 4.244687834749786e-05, "loss": 0.7173, "step": 12417 }, { "epoch": 0.7, "learning_rate": 4.243185516070546e-05, "loss": 0.7229, "step": 12418 }, { "epoch": 0.7, "learning_rate": 4.241683391698894e-05, "loss": 0.7056, "step": 12419 }, { "epoch": 0.7, "learning_rate": 4.2401814616855284e-05, "loss": 0.699, "step": 12420 }, { "epoch": 0.7, "learning_rate": 4.238679726081145e-05, "loss": 0.7393, "step": 12421 }, { "epoch": 0.7, "learning_rate": 4.2371781849364335e-05, "loss": 0.709, "step": 12422 }, { "epoch": 0.7, "learning_rate": 4.235676838302068e-05, "loss": 0.7438, "step": 12423 }, { "epoch": 0.7, "learning_rate": 4.2341756862287273e-05, "loss": 0.7125, "step": 12424 }, { "epoch": 0.7, "learning_rate": 4.232674728767082e-05, "loss": 0.7091, "step": 12425 }, { "epoch": 0.7, "learning_rate": 4.2311739659677875e-05, "loss": 0.7044, "step": 12426 }, { "epoch": 0.7, "learning_rate": 4.229673397881499e-05, "loss": 0.7067, "step": 12427 }, { "epoch": 0.7, "learning_rate": 4.228173024558868e-05, "loss": 0.7021, "step": 12428 }, { "epoch": 0.71, "learning_rate": 4.2266728460505375e-05, "loss": 0.654, "step": 12429 }, { "epoch": 0.71, "learning_rate": 4.225172862407136e-05, "loss": 0.7571, "step": 12430 }, { "epoch": 0.71, "learning_rate": 4.223673073679295e-05, "loss": 0.7879, "step": 12431 }, { "epoch": 0.71, "learning_rate": 4.2221734799176415e-05, "loss": 0.6832, "step": 12432 }, { "epoch": 0.71, "learning_rate": 4.220674081172782e-05, "loss": 0.658, "step": 12433 }, { "epoch": 0.71, "learning_rate": 4.2191748774953285e-05, "loss": 0.7379, "step": 12434 }, { "epoch": 0.71, "learning_rate": 4.217675868935884e-05, "loss": 0.5739, "step": 12435 }, { "epoch": 0.71, "learning_rate": 4.216177055545047e-05, "loss": 0.7648, "step": 12436 }, { "epoch": 0.71, "learning_rate": 4.214678437373399e-05, "loss": 0.6979, "step": 12437 }, { "epoch": 0.71, "learning_rate": 4.213180014471527e-05, "loss": 0.7475, "step": 12438 }, { "epoch": 0.71, "learning_rate": 4.21168178689001e-05, "loss": 0.6909, "step": 12439 }, { "epoch": 0.71, "learning_rate": 4.210183754679409e-05, "loss": 0.6805, "step": 12440 }, { "epoch": 0.71, "learning_rate": 4.2086859178902906e-05, "loss": 0.7247, "step": 12441 }, { "epoch": 0.71, "learning_rate": 4.207188276573214e-05, "loss": 0.707, "step": 12442 }, { "epoch": 0.71, "learning_rate": 4.2056908307787225e-05, "loss": 0.7391, "step": 12443 }, { "epoch": 0.71, "learning_rate": 4.204193580557361e-05, "loss": 0.6378, "step": 12444 }, { "epoch": 0.71, "learning_rate": 4.2026965259596666e-05, "loss": 0.6652, "step": 12445 }, { "epoch": 0.71, "learning_rate": 4.201199667036169e-05, "loss": 0.7038, "step": 12446 }, { "epoch": 0.71, "learning_rate": 4.199703003837393e-05, "loss": 0.6172, "step": 12447 }, { "epoch": 0.71, "learning_rate": 4.198206536413849e-05, "loss": 0.7451, "step": 12448 }, { "epoch": 0.71, "learning_rate": 4.196710264816054e-05, "loss": 0.7859, "step": 12449 }, { "epoch": 0.71, "learning_rate": 4.195214189094503e-05, "loss": 0.7743, "step": 12450 }, { "epoch": 0.71, "learning_rate": 4.1937183092996976e-05, "loss": 0.6514, "step": 12451 }, { "epoch": 0.71, "learning_rate": 4.1922226254821264e-05, "loss": 0.7714, "step": 12452 }, { "epoch": 0.71, "learning_rate": 4.190727137692273e-05, "loss": 0.6752, "step": 12453 }, { "epoch": 0.71, "learning_rate": 4.189231845980618e-05, "loss": 0.6828, "step": 12454 }, { "epoch": 0.71, "learning_rate": 4.187736750397627e-05, "loss": 0.6186, "step": 12455 }, { "epoch": 0.71, "learning_rate": 4.1862418509937616e-05, "loss": 0.7311, "step": 12456 }, { "epoch": 0.71, "learning_rate": 4.18474714781948e-05, "loss": 0.6866, "step": 12457 }, { "epoch": 0.71, "learning_rate": 4.1832526409252324e-05, "loss": 0.7463, "step": 12458 }, { "epoch": 0.71, "learning_rate": 4.181758330361464e-05, "loss": 0.6792, "step": 12459 }, { "epoch": 0.71, "learning_rate": 4.180264216178611e-05, "loss": 0.6689, "step": 12460 }, { "epoch": 0.71, "learning_rate": 4.1787702984271074e-05, "loss": 0.6726, "step": 12461 }, { "epoch": 0.71, "learning_rate": 4.177276577157372e-05, "loss": 0.6227, "step": 12462 }, { "epoch": 0.71, "learning_rate": 4.1757830524198195e-05, "loss": 0.7236, "step": 12463 }, { "epoch": 0.71, "learning_rate": 4.1742897242648636e-05, "loss": 0.6506, "step": 12464 }, { "epoch": 0.71, "learning_rate": 4.172796592742908e-05, "loss": 0.6972, "step": 12465 }, { "epoch": 0.71, "learning_rate": 4.17130365790435e-05, "loss": 0.7642, "step": 12466 }, { "epoch": 0.71, "learning_rate": 4.169810919799579e-05, "loss": 0.6473, "step": 12467 }, { "epoch": 0.71, "learning_rate": 4.168318378478985e-05, "loss": 0.7037, "step": 12468 }, { "epoch": 0.71, "learning_rate": 4.1668260339929385e-05, "loss": 0.7678, "step": 12469 }, { "epoch": 0.71, "learning_rate": 4.165333886391808e-05, "loss": 0.7642, "step": 12470 }, { "epoch": 0.71, "learning_rate": 4.163841935725961e-05, "loss": 0.7329, "step": 12471 }, { "epoch": 0.71, "learning_rate": 4.162350182045755e-05, "loss": 0.713, "step": 12472 }, { "epoch": 0.71, "learning_rate": 4.160858625401539e-05, "loss": 0.7162, "step": 12473 }, { "epoch": 0.71, "learning_rate": 4.159367265843662e-05, "loss": 0.7312, "step": 12474 }, { "epoch": 0.71, "learning_rate": 4.1578761034224545e-05, "loss": 0.6618, "step": 12475 }, { "epoch": 0.71, "learning_rate": 4.15638513818825e-05, "loss": 0.7829, "step": 12476 }, { "epoch": 0.71, "learning_rate": 4.1548943701913756e-05, "loss": 0.7448, "step": 12477 }, { "epoch": 0.71, "learning_rate": 4.153403799482143e-05, "loss": 0.6945, "step": 12478 }, { "epoch": 0.71, "learning_rate": 4.151913426110864e-05, "loss": 0.7121, "step": 12479 }, { "epoch": 0.71, "learning_rate": 4.150423250127845e-05, "loss": 0.7086, "step": 12480 }, { "epoch": 0.71, "learning_rate": 4.148933271583385e-05, "loss": 0.6295, "step": 12481 }, { "epoch": 0.71, "learning_rate": 4.1474434905277694e-05, "loss": 0.7038, "step": 12482 }, { "epoch": 0.71, "learning_rate": 4.145953907011285e-05, "loss": 0.6928, "step": 12483 }, { "epoch": 0.71, "learning_rate": 4.144464521084212e-05, "loss": 0.7416, "step": 12484 }, { "epoch": 0.71, "learning_rate": 4.142975332796815e-05, "loss": 0.6768, "step": 12485 }, { "epoch": 0.71, "learning_rate": 4.141486342199361e-05, "loss": 0.6493, "step": 12486 }, { "epoch": 0.71, "learning_rate": 4.139997549342111e-05, "loss": 0.6521, "step": 12487 }, { "epoch": 0.71, "learning_rate": 4.1385089542753086e-05, "loss": 0.7724, "step": 12488 }, { "epoch": 0.71, "learning_rate": 4.1370205570492004e-05, "loss": 0.6614, "step": 12489 }, { "epoch": 0.71, "learning_rate": 4.135532357714026e-05, "loss": 0.7831, "step": 12490 }, { "epoch": 0.71, "learning_rate": 4.134044356320017e-05, "loss": 0.6608, "step": 12491 }, { "epoch": 0.71, "learning_rate": 4.132556552917393e-05, "loss": 0.7094, "step": 12492 }, { "epoch": 0.71, "learning_rate": 4.131068947556372e-05, "loss": 0.523, "step": 12493 }, { "epoch": 0.71, "learning_rate": 4.12958154028717e-05, "loss": 0.689, "step": 12494 }, { "epoch": 0.71, "learning_rate": 4.128094331159983e-05, "loss": 0.5791, "step": 12495 }, { "epoch": 0.71, "learning_rate": 4.126607320225012e-05, "loss": 0.6752, "step": 12496 }, { "epoch": 0.71, "learning_rate": 4.1251205075324474e-05, "loss": 0.6044, "step": 12497 }, { "epoch": 0.71, "learning_rate": 4.123633893132473e-05, "loss": 0.6325, "step": 12498 }, { "epoch": 0.71, "learning_rate": 4.12214747707527e-05, "loss": 0.6493, "step": 12499 }, { "epoch": 0.71, "learning_rate": 4.1206612594110006e-05, "loss": 0.7512, "step": 12500 }, { "epoch": 0.71, "learning_rate": 4.1191752401898366e-05, "loss": 0.6644, "step": 12501 }, { "epoch": 0.71, "learning_rate": 4.117689419461929e-05, "loss": 0.7195, "step": 12502 }, { "epoch": 0.71, "learning_rate": 4.116203797277428e-05, "loss": 0.7448, "step": 12503 }, { "epoch": 0.71, "learning_rate": 4.114718373686481e-05, "loss": 0.7598, "step": 12504 }, { "epoch": 0.71, "learning_rate": 4.113233148739224e-05, "loss": 0.6555, "step": 12505 }, { "epoch": 0.71, "learning_rate": 4.11174812248579e-05, "loss": 0.6492, "step": 12506 }, { "epoch": 0.71, "learning_rate": 4.1102632949762996e-05, "loss": 0.7107, "step": 12507 }, { "epoch": 0.71, "learning_rate": 4.1087786662608665e-05, "loss": 0.7506, "step": 12508 }, { "epoch": 0.71, "learning_rate": 4.1072942363896026e-05, "loss": 0.6227, "step": 12509 }, { "epoch": 0.71, "learning_rate": 4.105810005412612e-05, "loss": 0.6118, "step": 12510 }, { "epoch": 0.71, "learning_rate": 4.104325973379993e-05, "loss": 0.662, "step": 12511 }, { "epoch": 0.71, "learning_rate": 4.102842140341834e-05, "loss": 0.5929, "step": 12512 }, { "epoch": 0.71, "learning_rate": 4.101358506348222e-05, "loss": 0.6383, "step": 12513 }, { "epoch": 0.71, "learning_rate": 4.0998750714492296e-05, "loss": 0.6966, "step": 12514 }, { "epoch": 0.71, "learning_rate": 4.098391835694925e-05, "loss": 0.7453, "step": 12515 }, { "epoch": 0.71, "learning_rate": 4.096908799135372e-05, "loss": 0.6745, "step": 12516 }, { "epoch": 0.71, "learning_rate": 4.0954259618206295e-05, "loss": 0.7394, "step": 12517 }, { "epoch": 0.71, "learning_rate": 4.093943323800745e-05, "loss": 0.721, "step": 12518 }, { "epoch": 0.71, "learning_rate": 4.0924608851257676e-05, "loss": 0.5803, "step": 12519 }, { "epoch": 0.71, "learning_rate": 4.0909786458457234e-05, "loss": 0.7246, "step": 12520 }, { "epoch": 0.71, "learning_rate": 4.089496606010652e-05, "loss": 0.6738, "step": 12521 }, { "epoch": 0.71, "learning_rate": 4.088014765670566e-05, "loss": 0.7595, "step": 12522 }, { "epoch": 0.71, "learning_rate": 4.086533124875487e-05, "loss": 0.6914, "step": 12523 }, { "epoch": 0.71, "learning_rate": 4.085051683675424e-05, "loss": 0.6418, "step": 12524 }, { "epoch": 0.71, "learning_rate": 4.083570442120379e-05, "loss": 0.6865, "step": 12525 }, { "epoch": 0.71, "learning_rate": 4.0820894002603506e-05, "loss": 0.6203, "step": 12526 }, { "epoch": 0.71, "learning_rate": 4.080608558145322e-05, "loss": 0.8057, "step": 12527 }, { "epoch": 0.71, "learning_rate": 4.079127915825284e-05, "loss": 0.6728, "step": 12528 }, { "epoch": 0.71, "learning_rate": 4.077647473350201e-05, "loss": 0.6345, "step": 12529 }, { "epoch": 0.71, "learning_rate": 4.0761672307700496e-05, "loss": 0.6236, "step": 12530 }, { "epoch": 0.71, "learning_rate": 4.0746871881347894e-05, "loss": 0.7327, "step": 12531 }, { "epoch": 0.71, "learning_rate": 4.07320734549438e-05, "loss": 0.7719, "step": 12532 }, { "epoch": 0.71, "learning_rate": 4.071727702898762e-05, "loss": 0.6973, "step": 12533 }, { "epoch": 0.71, "learning_rate": 4.070248260397882e-05, "loss": 0.7821, "step": 12534 }, { "epoch": 0.71, "learning_rate": 4.0687690180416736e-05, "loss": 0.7604, "step": 12535 }, { "epoch": 0.71, "learning_rate": 4.067289975880071e-05, "loss": 0.758, "step": 12536 }, { "epoch": 0.71, "learning_rate": 4.065811133962987e-05, "loss": 0.662, "step": 12537 }, { "epoch": 0.71, "learning_rate": 4.064332492340339e-05, "loss": 0.7169, "step": 12538 }, { "epoch": 0.71, "learning_rate": 4.062854051062042e-05, "loss": 0.6832, "step": 12539 }, { "epoch": 0.71, "learning_rate": 4.0613758101779864e-05, "loss": 0.7501, "step": 12540 }, { "epoch": 0.71, "learning_rate": 4.059897769738071e-05, "loss": 0.6421, "step": 12541 }, { "epoch": 0.71, "learning_rate": 4.058419929792186e-05, "loss": 0.7045, "step": 12542 }, { "epoch": 0.71, "learning_rate": 4.0569422903902135e-05, "loss": 0.7346, "step": 12543 }, { "epoch": 0.71, "learning_rate": 4.055464851582021e-05, "loss": 0.6101, "step": 12544 }, { "epoch": 0.71, "learning_rate": 4.053987613417481e-05, "loss": 0.6347, "step": 12545 }, { "epoch": 0.71, "learning_rate": 4.052510575946456e-05, "loss": 0.7809, "step": 12546 }, { "epoch": 0.71, "learning_rate": 4.0510337392187935e-05, "loss": 0.6835, "step": 12547 }, { "epoch": 0.71, "learning_rate": 4.049557103284344e-05, "loss": 0.8064, "step": 12548 }, { "epoch": 0.71, "learning_rate": 4.0480806681929485e-05, "loss": 0.6783, "step": 12549 }, { "epoch": 0.71, "learning_rate": 4.0466044339944383e-05, "loss": 0.7186, "step": 12550 }, { "epoch": 0.71, "learning_rate": 4.045128400738647e-05, "loss": 0.6923, "step": 12551 }, { "epoch": 0.71, "learning_rate": 4.043652568475389e-05, "loss": 0.696, "step": 12552 }, { "epoch": 0.71, "learning_rate": 4.0421769372544736e-05, "loss": 0.6958, "step": 12553 }, { "epoch": 0.71, "learning_rate": 4.040701507125712e-05, "loss": 0.7205, "step": 12554 }, { "epoch": 0.71, "learning_rate": 4.039226278138903e-05, "loss": 0.7288, "step": 12555 }, { "epoch": 0.71, "learning_rate": 4.037751250343841e-05, "loss": 0.7112, "step": 12556 }, { "epoch": 0.71, "learning_rate": 4.0362764237903095e-05, "loss": 0.6248, "step": 12557 }, { "epoch": 0.71, "learning_rate": 4.034801798528094e-05, "loss": 0.768, "step": 12558 }, { "epoch": 0.71, "learning_rate": 4.033327374606961e-05, "loss": 0.6831, "step": 12559 }, { "epoch": 0.71, "learning_rate": 4.031853152076675e-05, "loss": 0.6966, "step": 12560 }, { "epoch": 0.71, "learning_rate": 4.0303791309869955e-05, "loss": 0.7216, "step": 12561 }, { "epoch": 0.71, "learning_rate": 4.0289053113876775e-05, "loss": 0.6983, "step": 12562 }, { "epoch": 0.71, "learning_rate": 4.027431693328464e-05, "loss": 0.7018, "step": 12563 }, { "epoch": 0.71, "learning_rate": 4.025958276859099e-05, "loss": 0.6643, "step": 12564 }, { "epoch": 0.71, "learning_rate": 4.0244850620293084e-05, "loss": 0.7017, "step": 12565 }, { "epoch": 0.71, "learning_rate": 4.023012048888816e-05, "loss": 0.7581, "step": 12566 }, { "epoch": 0.71, "learning_rate": 4.021539237487342e-05, "loss": 0.7646, "step": 12567 }, { "epoch": 0.71, "learning_rate": 4.020066627874597e-05, "loss": 0.7022, "step": 12568 }, { "epoch": 0.71, "learning_rate": 4.018594220100287e-05, "loss": 0.7063, "step": 12569 }, { "epoch": 0.71, "learning_rate": 4.017122014214109e-05, "loss": 0.7245, "step": 12570 }, { "epoch": 0.71, "learning_rate": 4.015650010265757e-05, "loss": 0.7527, "step": 12571 }, { "epoch": 0.71, "learning_rate": 4.014178208304907e-05, "loss": 0.7664, "step": 12572 }, { "epoch": 0.71, "learning_rate": 4.012706608381246e-05, "loss": 0.7095, "step": 12573 }, { "epoch": 0.71, "learning_rate": 4.011235210544435e-05, "loss": 0.6184, "step": 12574 }, { "epoch": 0.71, "learning_rate": 4.009764014844143e-05, "loss": 0.7412, "step": 12575 }, { "epoch": 0.71, "learning_rate": 4.008293021330024e-05, "loss": 0.6445, "step": 12576 }, { "epoch": 0.71, "learning_rate": 4.006822230051733e-05, "loss": 0.664, "step": 12577 }, { "epoch": 0.71, "learning_rate": 4.0053516410589066e-05, "loss": 0.6414, "step": 12578 }, { "epoch": 0.71, "learning_rate": 4.003881254401183e-05, "loss": 0.7268, "step": 12579 }, { "epoch": 0.71, "learning_rate": 4.002411070128197e-05, "loss": 0.7001, "step": 12580 }, { "epoch": 0.71, "learning_rate": 4.0009410882895615e-05, "loss": 0.7018, "step": 12581 }, { "epoch": 0.71, "learning_rate": 3.9994713089348987e-05, "loss": 0.6631, "step": 12582 }, { "epoch": 0.71, "learning_rate": 3.998001732113816e-05, "loss": 0.6696, "step": 12583 }, { "epoch": 0.71, "learning_rate": 3.996532357875919e-05, "loss": 0.651, "step": 12584 }, { "epoch": 0.71, "learning_rate": 3.9950631862707964e-05, "loss": 0.8029, "step": 12585 }, { "epoch": 0.71, "learning_rate": 3.993594217348039e-05, "loss": 0.642, "step": 12586 }, { "epoch": 0.71, "learning_rate": 3.9921254511572305e-05, "loss": 0.7728, "step": 12587 }, { "epoch": 0.71, "learning_rate": 3.9906568877479466e-05, "loss": 0.7906, "step": 12588 }, { "epoch": 0.71, "learning_rate": 3.9891885271697496e-05, "loss": 0.7704, "step": 12589 }, { "epoch": 0.71, "learning_rate": 3.9877203694722044e-05, "loss": 0.7066, "step": 12590 }, { "epoch": 0.71, "learning_rate": 3.9862524147048666e-05, "loss": 0.7076, "step": 12591 }, { "epoch": 0.71, "learning_rate": 3.9847846629172794e-05, "loss": 0.6375, "step": 12592 }, { "epoch": 0.71, "learning_rate": 3.983317114158984e-05, "loss": 0.7212, "step": 12593 }, { "epoch": 0.71, "learning_rate": 3.981849768479517e-05, "loss": 0.6648, "step": 12594 }, { "epoch": 0.71, "learning_rate": 3.980382625928405e-05, "loss": 0.7566, "step": 12595 }, { "epoch": 0.71, "learning_rate": 3.978915686555164e-05, "loss": 0.7454, "step": 12596 }, { "epoch": 0.71, "learning_rate": 3.9774489504093125e-05, "loss": 0.8177, "step": 12597 }, { "epoch": 0.71, "learning_rate": 3.9759824175403506e-05, "loss": 0.6437, "step": 12598 }, { "epoch": 0.71, "learning_rate": 3.97451608799778e-05, "loss": 0.7047, "step": 12599 }, { "epoch": 0.71, "learning_rate": 3.9730499618310944e-05, "loss": 0.6273, "step": 12600 }, { "epoch": 0.71, "learning_rate": 3.97158403908978e-05, "loss": 0.7232, "step": 12601 }, { "epoch": 0.71, "learning_rate": 3.970118319823316e-05, "loss": 0.6623, "step": 12602 }, { "epoch": 0.71, "learning_rate": 3.96865280408117e-05, "loss": 0.7244, "step": 12603 }, { "epoch": 0.71, "learning_rate": 3.967187491912813e-05, "loss": 0.6942, "step": 12604 }, { "epoch": 0.72, "learning_rate": 3.965722383367696e-05, "loss": 0.7084, "step": 12605 }, { "epoch": 0.72, "learning_rate": 3.9642574784952745e-05, "loss": 0.7641, "step": 12606 }, { "epoch": 0.72, "learning_rate": 3.962792777344992e-05, "loss": 0.7099, "step": 12607 }, { "epoch": 0.72, "learning_rate": 3.961328279966287e-05, "loss": 0.6384, "step": 12608 }, { "epoch": 0.72, "learning_rate": 3.959863986408593e-05, "loss": 0.7486, "step": 12609 }, { "epoch": 0.72, "learning_rate": 3.9583998967213296e-05, "loss": 0.7214, "step": 12610 }, { "epoch": 0.72, "learning_rate": 3.956936010953911e-05, "loss": 0.7579, "step": 12611 }, { "epoch": 0.72, "learning_rate": 3.955472329155751e-05, "loss": 0.7464, "step": 12612 }, { "epoch": 0.72, "learning_rate": 3.954008851376252e-05, "loss": 0.7508, "step": 12613 }, { "epoch": 0.72, "learning_rate": 3.9525455776648104e-05, "loss": 0.7373, "step": 12614 }, { "epoch": 0.72, "learning_rate": 3.9510825080708155e-05, "loss": 0.6855, "step": 12615 }, { "epoch": 0.72, "learning_rate": 3.949619642643654e-05, "loss": 0.7122, "step": 12616 }, { "epoch": 0.72, "learning_rate": 3.9481569814326966e-05, "loss": 0.7113, "step": 12617 }, { "epoch": 0.72, "learning_rate": 3.946694524487309e-05, "loss": 0.6847, "step": 12618 }, { "epoch": 0.72, "learning_rate": 3.945232271856857e-05, "loss": 0.596, "step": 12619 }, { "epoch": 0.72, "learning_rate": 3.9437702235906945e-05, "loss": 0.7175, "step": 12620 }, { "epoch": 0.72, "learning_rate": 3.942308379738171e-05, "loss": 0.7616, "step": 12621 }, { "epoch": 0.72, "learning_rate": 3.9408467403486296e-05, "loss": 0.7173, "step": 12622 }, { "epoch": 0.72, "learning_rate": 3.939385305471398e-05, "loss": 0.6621, "step": 12623 }, { "epoch": 0.72, "learning_rate": 3.937924075155806e-05, "loss": 0.7415, "step": 12624 }, { "epoch": 0.72, "learning_rate": 3.936463049451179e-05, "loss": 0.7332, "step": 12625 }, { "epoch": 0.72, "learning_rate": 3.9350022284068235e-05, "loss": 0.7193, "step": 12626 }, { "epoch": 0.72, "learning_rate": 3.9335416120720484e-05, "loss": 0.7682, "step": 12627 }, { "epoch": 0.72, "learning_rate": 3.932081200496153e-05, "loss": 0.6555, "step": 12628 }, { "epoch": 0.72, "learning_rate": 3.9306209937284346e-05, "loss": 0.714, "step": 12629 }, { "epoch": 0.72, "learning_rate": 3.9291609918181715e-05, "loss": 0.6041, "step": 12630 }, { "epoch": 0.72, "learning_rate": 3.927701194814647e-05, "loss": 0.7629, "step": 12631 }, { "epoch": 0.72, "learning_rate": 3.9262416027671356e-05, "loss": 0.6845, "step": 12632 }, { "epoch": 0.72, "learning_rate": 3.924782215724896e-05, "loss": 0.7118, "step": 12633 }, { "epoch": 0.72, "learning_rate": 3.9233230337371886e-05, "loss": 0.681, "step": 12634 }, { "epoch": 0.72, "learning_rate": 3.921864056853266e-05, "loss": 0.5806, "step": 12635 }, { "epoch": 0.72, "learning_rate": 3.920405285122375e-05, "loss": 0.6788, "step": 12636 }, { "epoch": 0.72, "learning_rate": 3.9189467185937465e-05, "loss": 0.7169, "step": 12637 }, { "epoch": 0.72, "learning_rate": 3.917488357316615e-05, "loss": 0.5911, "step": 12638 }, { "epoch": 0.72, "learning_rate": 3.916030201340203e-05, "loss": 0.6548, "step": 12639 }, { "epoch": 0.72, "learning_rate": 3.9145722507137315e-05, "loss": 0.7313, "step": 12640 }, { "epoch": 0.72, "learning_rate": 3.9131145054864017e-05, "loss": 0.6744, "step": 12641 }, { "epoch": 0.72, "learning_rate": 3.9116569657074255e-05, "loss": 0.727, "step": 12642 }, { "epoch": 0.72, "learning_rate": 3.910199631425989e-05, "loss": 0.5632, "step": 12643 }, { "epoch": 0.72, "learning_rate": 3.908742502691287e-05, "loss": 0.6186, "step": 12644 }, { "epoch": 0.72, "learning_rate": 3.907285579552501e-05, "loss": 0.6828, "step": 12645 }, { "epoch": 0.72, "learning_rate": 3.9058288620588046e-05, "loss": 0.6921, "step": 12646 }, { "epoch": 0.72, "learning_rate": 3.90437235025937e-05, "loss": 0.7238, "step": 12647 }, { "epoch": 0.72, "learning_rate": 3.902916044203352e-05, "loss": 0.7119, "step": 12648 }, { "epoch": 0.72, "learning_rate": 3.90145994393991e-05, "loss": 0.7119, "step": 12649 }, { "epoch": 0.72, "learning_rate": 3.900004049518187e-05, "loss": 0.7232, "step": 12650 }, { "epoch": 0.72, "learning_rate": 3.8985483609873244e-05, "loss": 0.7583, "step": 12651 }, { "epoch": 0.72, "learning_rate": 3.897092878396457e-05, "loss": 0.5956, "step": 12652 }, { "epoch": 0.72, "learning_rate": 3.8956376017947096e-05, "loss": 0.6123, "step": 12653 }, { "epoch": 0.72, "learning_rate": 3.8941825312312054e-05, "loss": 0.5522, "step": 12654 }, { "epoch": 0.72, "learning_rate": 3.892727666755055e-05, "loss": 0.6849, "step": 12655 }, { "epoch": 0.72, "learning_rate": 3.8912730084153584e-05, "loss": 0.6742, "step": 12656 }, { "epoch": 0.72, "learning_rate": 3.889818556261219e-05, "loss": 0.7039, "step": 12657 }, { "epoch": 0.72, "learning_rate": 3.888364310341728e-05, "loss": 0.7295, "step": 12658 }, { "epoch": 0.72, "learning_rate": 3.886910270705969e-05, "loss": 0.7131, "step": 12659 }, { "epoch": 0.72, "learning_rate": 3.8854564374030214e-05, "loss": 0.7392, "step": 12660 }, { "epoch": 0.72, "learning_rate": 3.884002810481958e-05, "loss": 0.6859, "step": 12661 }, { "epoch": 0.72, "learning_rate": 3.8825493899918395e-05, "loss": 0.8143, "step": 12662 }, { "epoch": 0.72, "learning_rate": 3.88109617598172e-05, "loss": 0.6641, "step": 12663 }, { "epoch": 0.72, "learning_rate": 3.879643168500652e-05, "loss": 0.6864, "step": 12664 }, { "epoch": 0.72, "learning_rate": 3.8781903675976775e-05, "loss": 0.7756, "step": 12665 }, { "epoch": 0.72, "learning_rate": 3.8767377733218324e-05, "loss": 0.661, "step": 12666 }, { "epoch": 0.72, "learning_rate": 3.875285385722152e-05, "loss": 0.7427, "step": 12667 }, { "epoch": 0.72, "learning_rate": 3.8738332048476464e-05, "loss": 0.6023, "step": 12668 }, { "epoch": 0.72, "learning_rate": 3.872381230747342e-05, "loss": 0.7044, "step": 12669 }, { "epoch": 0.72, "learning_rate": 3.8709294634702376e-05, "loss": 0.7753, "step": 12670 }, { "epoch": 0.72, "learning_rate": 3.8694779030653374e-05, "loss": 0.6202, "step": 12671 }, { "epoch": 0.72, "learning_rate": 3.868026549581636e-05, "loss": 0.7429, "step": 12672 }, { "epoch": 0.72, "learning_rate": 3.86657540306812e-05, "loss": 0.7291, "step": 12673 }, { "epoch": 0.72, "learning_rate": 3.865124463573774e-05, "loss": 0.7446, "step": 12674 }, { "epoch": 0.72, "learning_rate": 3.863673731147562e-05, "loss": 0.722, "step": 12675 }, { "epoch": 0.72, "learning_rate": 3.8622232058384557e-05, "loss": 0.6183, "step": 12676 }, { "epoch": 0.72, "learning_rate": 3.860772887695415e-05, "loss": 0.6125, "step": 12677 }, { "epoch": 0.72, "learning_rate": 3.859322776767389e-05, "loss": 0.8017, "step": 12678 }, { "epoch": 0.72, "learning_rate": 3.857872873103322e-05, "loss": 0.7255, "step": 12679 }, { "epoch": 0.72, "learning_rate": 3.856423176752154e-05, "loss": 0.749, "step": 12680 }, { "epoch": 0.72, "learning_rate": 3.85497368776282e-05, "loss": 0.7001, "step": 12681 }, { "epoch": 0.72, "learning_rate": 3.853524406184237e-05, "loss": 0.7322, "step": 12682 }, { "epoch": 0.72, "learning_rate": 3.852075332065324e-05, "loss": 0.7481, "step": 12683 }, { "epoch": 0.72, "learning_rate": 3.850626465454995e-05, "loss": 0.7398, "step": 12684 }, { "epoch": 0.72, "learning_rate": 3.849177806402148e-05, "loss": 0.6676, "step": 12685 }, { "epoch": 0.72, "learning_rate": 3.8477293549556805e-05, "loss": 0.8245, "step": 12686 }, { "epoch": 0.72, "learning_rate": 3.8462811111644855e-05, "loss": 0.637, "step": 12687 }, { "epoch": 0.72, "learning_rate": 3.8448330750774386e-05, "loss": 0.7659, "step": 12688 }, { "epoch": 0.72, "learning_rate": 3.843385246743417e-05, "loss": 0.6642, "step": 12689 }, { "epoch": 0.72, "learning_rate": 3.841937626211292e-05, "loss": 0.8178, "step": 12690 }, { "epoch": 0.72, "learning_rate": 3.840490213529924e-05, "loss": 0.6566, "step": 12691 }, { "epoch": 0.72, "learning_rate": 3.839043008748161e-05, "loss": 0.7738, "step": 12692 }, { "epoch": 0.72, "learning_rate": 3.8375960119148544e-05, "loss": 0.6629, "step": 12693 }, { "epoch": 0.72, "learning_rate": 3.8361492230788485e-05, "loss": 0.6056, "step": 12694 }, { "epoch": 0.72, "learning_rate": 3.834702642288968e-05, "loss": 0.6903, "step": 12695 }, { "epoch": 0.72, "learning_rate": 3.833256269594042e-05, "loss": 0.7775, "step": 12696 }, { "epoch": 0.72, "learning_rate": 3.8318101050428904e-05, "loss": 0.6492, "step": 12697 }, { "epoch": 0.72, "learning_rate": 3.8303641486843234e-05, "loss": 0.7386, "step": 12698 }, { "epoch": 0.72, "learning_rate": 3.828918400567151e-05, "loss": 0.5794, "step": 12699 }, { "epoch": 0.72, "learning_rate": 3.8274728607401666e-05, "loss": 0.718, "step": 12700 }, { "epoch": 0.72, "learning_rate": 3.826027529252158e-05, "loss": 0.7247, "step": 12701 }, { "epoch": 0.72, "learning_rate": 3.8245824061519106e-05, "loss": 0.7271, "step": 12702 }, { "epoch": 0.72, "learning_rate": 3.8231374914882045e-05, "loss": 0.7008, "step": 12703 }, { "epoch": 0.72, "learning_rate": 3.821692785309807e-05, "loss": 0.7836, "step": 12704 }, { "epoch": 0.72, "learning_rate": 3.820248287665481e-05, "loss": 0.6473, "step": 12705 }, { "epoch": 0.72, "learning_rate": 3.818803998603986e-05, "loss": 0.8175, "step": 12706 }, { "epoch": 0.72, "learning_rate": 3.817359918174066e-05, "loss": 0.7494, "step": 12707 }, { "epoch": 0.72, "learning_rate": 3.8159160464244606e-05, "loss": 0.7403, "step": 12708 }, { "epoch": 0.72, "learning_rate": 3.814472383403907e-05, "loss": 0.6892, "step": 12709 }, { "epoch": 0.72, "learning_rate": 3.813028929161132e-05, "loss": 0.7231, "step": 12710 }, { "epoch": 0.72, "learning_rate": 3.811585683744857e-05, "loss": 0.6489, "step": 12711 }, { "epoch": 0.72, "learning_rate": 3.8101426472037996e-05, "loss": 0.7053, "step": 12712 }, { "epoch": 0.72, "learning_rate": 3.808699819586656e-05, "loss": 0.6717, "step": 12713 }, { "epoch": 0.72, "learning_rate": 3.8072572009421346e-05, "loss": 0.6816, "step": 12714 }, { "epoch": 0.72, "learning_rate": 3.805814791318921e-05, "loss": 0.6551, "step": 12715 }, { "epoch": 0.72, "learning_rate": 3.8043725907657025e-05, "loss": 0.6269, "step": 12716 }, { "epoch": 0.72, "learning_rate": 3.802930599331158e-05, "loss": 0.7382, "step": 12717 }, { "epoch": 0.72, "learning_rate": 3.801488817063957e-05, "loss": 0.6446, "step": 12718 }, { "epoch": 0.72, "learning_rate": 3.80004724401277e-05, "loss": 0.7086, "step": 12719 }, { "epoch": 0.72, "learning_rate": 3.7986058802262435e-05, "loss": 0.732, "step": 12720 }, { "epoch": 0.72, "learning_rate": 3.7971647257530375e-05, "loss": 0.7823, "step": 12721 }, { "epoch": 0.72, "learning_rate": 3.7957237806417844e-05, "loss": 0.6267, "step": 12722 }, { "epoch": 0.72, "learning_rate": 3.794283044941126e-05, "loss": 0.6291, "step": 12723 }, { "epoch": 0.72, "learning_rate": 3.7928425186996885e-05, "loss": 0.7673, "step": 12724 }, { "epoch": 0.72, "learning_rate": 3.7914022019660966e-05, "loss": 0.7097, "step": 12725 }, { "epoch": 0.72, "learning_rate": 3.7899620947889655e-05, "loss": 0.7643, "step": 12726 }, { "epoch": 0.72, "learning_rate": 3.788522197216897e-05, "loss": 0.7261, "step": 12727 }, { "epoch": 0.72, "learning_rate": 3.787082509298497e-05, "loss": 0.5947, "step": 12728 }, { "epoch": 0.72, "learning_rate": 3.7856430310823545e-05, "loss": 0.7092, "step": 12729 }, { "epoch": 0.72, "learning_rate": 3.7842037626170566e-05, "loss": 0.7384, "step": 12730 }, { "epoch": 0.72, "learning_rate": 3.782764703951183e-05, "loss": 0.6778, "step": 12731 }, { "epoch": 0.72, "learning_rate": 3.78132585513331e-05, "loss": 0.7337, "step": 12732 }, { "epoch": 0.72, "learning_rate": 3.779887216211995e-05, "loss": 0.6714, "step": 12733 }, { "epoch": 0.72, "learning_rate": 3.778448787235799e-05, "loss": 0.6357, "step": 12734 }, { "epoch": 0.72, "learning_rate": 3.777010568253273e-05, "loss": 0.6039, "step": 12735 }, { "epoch": 0.72, "learning_rate": 3.775572559312964e-05, "loss": 0.5842, "step": 12736 }, { "epoch": 0.72, "learning_rate": 3.774134760463401e-05, "loss": 0.778, "step": 12737 }, { "epoch": 0.72, "learning_rate": 3.77269717175312e-05, "loss": 0.6435, "step": 12738 }, { "epoch": 0.72, "learning_rate": 3.7712597932306424e-05, "loss": 0.6088, "step": 12739 }, { "epoch": 0.72, "learning_rate": 3.76982262494448e-05, "loss": 0.7706, "step": 12740 }, { "epoch": 0.72, "learning_rate": 3.7683856669431436e-05, "loss": 0.7012, "step": 12741 }, { "epoch": 0.72, "learning_rate": 3.766948919275133e-05, "loss": 0.7056, "step": 12742 }, { "epoch": 0.72, "learning_rate": 3.765512381988947e-05, "loss": 0.6155, "step": 12743 }, { "epoch": 0.72, "learning_rate": 3.764076055133065e-05, "loss": 0.6213, "step": 12744 }, { "epoch": 0.72, "learning_rate": 3.762639938755974e-05, "loss": 0.7106, "step": 12745 }, { "epoch": 0.72, "learning_rate": 3.7612040329061405e-05, "loss": 0.6017, "step": 12746 }, { "epoch": 0.72, "learning_rate": 3.7597683376320314e-05, "loss": 0.7078, "step": 12747 }, { "epoch": 0.72, "learning_rate": 3.758332852982107e-05, "loss": 0.6602, "step": 12748 }, { "epoch": 0.72, "learning_rate": 3.7568975790048186e-05, "loss": 0.7135, "step": 12749 }, { "epoch": 0.72, "learning_rate": 3.75546251574861e-05, "loss": 0.8676, "step": 12750 }, { "epoch": 0.72, "learning_rate": 3.754027663261922e-05, "loss": 0.6796, "step": 12751 }, { "epoch": 0.72, "learning_rate": 3.752593021593181e-05, "loss": 0.7703, "step": 12752 }, { "epoch": 0.72, "learning_rate": 3.7511585907908054e-05, "loss": 0.6886, "step": 12753 }, { "epoch": 0.72, "learning_rate": 3.749724370903216e-05, "loss": 0.7308, "step": 12754 }, { "epoch": 0.72, "learning_rate": 3.748290361978821e-05, "loss": 0.7351, "step": 12755 }, { "epoch": 0.72, "learning_rate": 3.746856564066023e-05, "loss": 0.6656, "step": 12756 }, { "epoch": 0.72, "learning_rate": 3.745422977213218e-05, "loss": 0.8188, "step": 12757 }, { "epoch": 0.72, "learning_rate": 3.743989601468787e-05, "loss": 0.6806, "step": 12758 }, { "epoch": 0.72, "learning_rate": 3.742556436881118e-05, "loss": 0.8438, "step": 12759 }, { "epoch": 0.72, "learning_rate": 3.7411234834985765e-05, "loss": 0.771, "step": 12760 }, { "epoch": 0.72, "learning_rate": 3.7396907413695317e-05, "loss": 0.7407, "step": 12761 }, { "epoch": 0.72, "learning_rate": 3.7382582105423436e-05, "loss": 0.7185, "step": 12762 }, { "epoch": 0.72, "learning_rate": 3.736825891065362e-05, "loss": 0.686, "step": 12763 }, { "epoch": 0.72, "learning_rate": 3.735393782986936e-05, "loss": 0.7332, "step": 12764 }, { "epoch": 0.72, "learning_rate": 3.733961886355398e-05, "loss": 0.6443, "step": 12765 }, { "epoch": 0.72, "learning_rate": 3.7325302012190775e-05, "loss": 0.7448, "step": 12766 }, { "epoch": 0.72, "learning_rate": 3.7310987276263e-05, "loss": 0.782, "step": 12767 }, { "epoch": 0.72, "learning_rate": 3.729667465625381e-05, "loss": 0.6948, "step": 12768 }, { "epoch": 0.72, "learning_rate": 3.7282364152646297e-05, "loss": 0.7474, "step": 12769 }, { "epoch": 0.72, "learning_rate": 3.726805576592347e-05, "loss": 0.6769, "step": 12770 }, { "epoch": 0.72, "learning_rate": 3.725374949656832e-05, "loss": 0.7948, "step": 12771 }, { "epoch": 0.72, "learning_rate": 3.723944534506365e-05, "loss": 0.8291, "step": 12772 }, { "epoch": 0.72, "learning_rate": 3.722514331189233e-05, "loss": 0.7028, "step": 12773 }, { "epoch": 0.72, "learning_rate": 3.721084339753702e-05, "loss": 0.7377, "step": 12774 }, { "epoch": 0.72, "learning_rate": 3.7196545602480426e-05, "loss": 0.6299, "step": 12775 }, { "epoch": 0.72, "learning_rate": 3.7182249927205116e-05, "loss": 0.7233, "step": 12776 }, { "epoch": 0.72, "learning_rate": 3.716795637219364e-05, "loss": 0.6951, "step": 12777 }, { "epoch": 0.72, "learning_rate": 3.71536649379284e-05, "loss": 0.6397, "step": 12778 }, { "epoch": 0.72, "learning_rate": 3.713937562489179e-05, "loss": 0.7201, "step": 12779 }, { "epoch": 0.72, "learning_rate": 3.7125088433566135e-05, "loss": 0.7144, "step": 12780 }, { "epoch": 0.72, "learning_rate": 3.711080336443362e-05, "loss": 0.6328, "step": 12781 }, { "epoch": 0.73, "learning_rate": 3.709652041797641e-05, "loss": 0.8081, "step": 12782 }, { "epoch": 0.73, "learning_rate": 3.708223959467661e-05, "loss": 0.701, "step": 12783 }, { "epoch": 0.73, "learning_rate": 3.7067960895016275e-05, "loss": 0.6588, "step": 12784 }, { "epoch": 0.73, "learning_rate": 3.7053684319477263e-05, "loss": 0.7475, "step": 12785 }, { "epoch": 0.73, "learning_rate": 3.7039409868541487e-05, "loss": 0.7146, "step": 12786 }, { "epoch": 0.73, "learning_rate": 3.702513754269076e-05, "loss": 0.7122, "step": 12787 }, { "epoch": 0.73, "learning_rate": 3.701086734240682e-05, "loss": 0.608, "step": 12788 }, { "epoch": 0.73, "learning_rate": 3.699659926817127e-05, "loss": 0.6744, "step": 12789 }, { "epoch": 0.73, "learning_rate": 3.698233332046576e-05, "loss": 0.6661, "step": 12790 }, { "epoch": 0.73, "learning_rate": 3.696806949977174e-05, "loss": 0.8081, "step": 12791 }, { "epoch": 0.73, "learning_rate": 3.695380780657068e-05, "loss": 0.6569, "step": 12792 }, { "epoch": 0.73, "learning_rate": 3.693954824134396e-05, "loss": 0.6939, "step": 12793 }, { "epoch": 0.73, "learning_rate": 3.6925290804572856e-05, "loss": 0.7267, "step": 12794 }, { "epoch": 0.73, "learning_rate": 3.6911035496738645e-05, "loss": 0.7288, "step": 12795 }, { "epoch": 0.73, "learning_rate": 3.689678231832241e-05, "loss": 0.6982, "step": 12796 }, { "epoch": 0.73, "learning_rate": 3.68825312698053e-05, "loss": 0.7638, "step": 12797 }, { "epoch": 0.73, "learning_rate": 3.686828235166825e-05, "loss": 0.6467, "step": 12798 }, { "epoch": 0.73, "learning_rate": 3.6854035564392255e-05, "loss": 0.7545, "step": 12799 }, { "epoch": 0.73, "learning_rate": 3.683979090845815e-05, "loss": 0.6984, "step": 12800 }, { "epoch": 0.73, "learning_rate": 3.6825548384346756e-05, "loss": 0.7466, "step": 12801 }, { "epoch": 0.73, "learning_rate": 3.681130799253878e-05, "loss": 0.7089, "step": 12802 }, { "epoch": 0.73, "learning_rate": 3.679706973351491e-05, "loss": 0.6963, "step": 12803 }, { "epoch": 0.73, "learning_rate": 3.678283360775571e-05, "loss": 0.739, "step": 12804 }, { "epoch": 0.73, "learning_rate": 3.676859961574162e-05, "loss": 0.7997, "step": 12805 }, { "epoch": 0.73, "learning_rate": 3.675436775795311e-05, "loss": 0.6116, "step": 12806 }, { "epoch": 0.73, "learning_rate": 3.674013803487057e-05, "loss": 0.63, "step": 12807 }, { "epoch": 0.73, "learning_rate": 3.6725910446974286e-05, "loss": 0.764, "step": 12808 }, { "epoch": 0.73, "learning_rate": 3.671168499474449e-05, "loss": 0.7521, "step": 12809 }, { "epoch": 0.73, "learning_rate": 3.66974616786613e-05, "loss": 0.6826, "step": 12810 }, { "epoch": 0.73, "learning_rate": 3.6683240499204774e-05, "loss": 0.7247, "step": 12811 }, { "epoch": 0.73, "learning_rate": 3.6669021456854926e-05, "loss": 0.6998, "step": 12812 }, { "epoch": 0.73, "learning_rate": 3.66548045520917e-05, "loss": 0.7288, "step": 12813 }, { "epoch": 0.73, "learning_rate": 3.6640589785394955e-05, "loss": 0.649, "step": 12814 }, { "epoch": 0.73, "learning_rate": 3.662637715724446e-05, "loss": 0.7221, "step": 12815 }, { "epoch": 0.73, "learning_rate": 3.6612166668119973e-05, "loss": 0.7188, "step": 12816 }, { "epoch": 0.73, "learning_rate": 3.6597958318501115e-05, "loss": 0.8041, "step": 12817 }, { "epoch": 0.73, "learning_rate": 3.6583752108867385e-05, "loss": 0.6879, "step": 12818 }, { "epoch": 0.73, "learning_rate": 3.6569548039698355e-05, "loss": 0.7127, "step": 12819 }, { "epoch": 0.73, "learning_rate": 3.655534611147343e-05, "loss": 0.7359, "step": 12820 }, { "epoch": 0.73, "learning_rate": 3.654114632467196e-05, "loss": 0.6477, "step": 12821 }, { "epoch": 0.73, "learning_rate": 3.6526948679773257e-05, "loss": 0.7186, "step": 12822 }, { "epoch": 0.73, "learning_rate": 3.6512753177256476e-05, "loss": 0.7935, "step": 12823 }, { "epoch": 0.73, "learning_rate": 3.6498559817600766e-05, "loss": 0.7745, "step": 12824 }, { "epoch": 0.73, "learning_rate": 3.648436860128525e-05, "loss": 0.7464, "step": 12825 }, { "epoch": 0.73, "learning_rate": 3.6470179528788816e-05, "loss": 0.7517, "step": 12826 }, { "epoch": 0.73, "learning_rate": 3.645599260059045e-05, "loss": 0.6616, "step": 12827 }, { "epoch": 0.73, "learning_rate": 3.6441807817168984e-05, "loss": 0.7741, "step": 12828 }, { "epoch": 0.73, "learning_rate": 3.642762517900322e-05, "loss": 0.8113, "step": 12829 }, { "epoch": 0.73, "learning_rate": 3.64134446865718e-05, "loss": 0.7108, "step": 12830 }, { "epoch": 0.73, "learning_rate": 3.639926634035338e-05, "loss": 0.7228, "step": 12831 }, { "epoch": 0.73, "learning_rate": 3.6385090140826564e-05, "loss": 0.6428, "step": 12832 }, { "epoch": 0.73, "learning_rate": 3.6370916088469743e-05, "loss": 0.7098, "step": 12833 }, { "epoch": 0.73, "learning_rate": 3.635674418376138e-05, "loss": 0.708, "step": 12834 }, { "epoch": 0.73, "learning_rate": 3.634257442717982e-05, "loss": 0.6641, "step": 12835 }, { "epoch": 0.73, "learning_rate": 3.6328406819203354e-05, "loss": 0.7398, "step": 12836 }, { "epoch": 0.73, "learning_rate": 3.631424136031011e-05, "loss": 0.694, "step": 12837 }, { "epoch": 0.73, "learning_rate": 3.6300078050978225e-05, "loss": 0.6494, "step": 12838 }, { "epoch": 0.73, "learning_rate": 3.628591689168578e-05, "loss": 0.778, "step": 12839 }, { "epoch": 0.73, "learning_rate": 3.627175788291077e-05, "loss": 0.7789, "step": 12840 }, { "epoch": 0.73, "learning_rate": 3.6257601025131026e-05, "loss": 0.6904, "step": 12841 }, { "epoch": 0.73, "learning_rate": 3.624344631882446e-05, "loss": 0.7533, "step": 12842 }, { "epoch": 0.73, "learning_rate": 3.622929376446875e-05, "loss": 0.6678, "step": 12843 }, { "epoch": 0.73, "learning_rate": 3.621514336254162e-05, "loss": 0.7682, "step": 12844 }, { "epoch": 0.73, "learning_rate": 3.6200995113520685e-05, "loss": 0.7271, "step": 12845 }, { "epoch": 0.73, "learning_rate": 3.618684901788347e-05, "loss": 0.7507, "step": 12846 }, { "epoch": 0.73, "learning_rate": 3.617270507610752e-05, "loss": 0.642, "step": 12847 }, { "epoch": 0.73, "learning_rate": 3.615856328867011e-05, "loss": 0.7317, "step": 12848 }, { "epoch": 0.73, "learning_rate": 3.6144423656048656e-05, "loss": 0.5664, "step": 12849 }, { "epoch": 0.73, "learning_rate": 3.6130286178720344e-05, "loss": 0.7487, "step": 12850 }, { "epoch": 0.73, "learning_rate": 3.6116150857162365e-05, "loss": 0.6824, "step": 12851 }, { "epoch": 0.73, "learning_rate": 3.610201769185185e-05, "loss": 0.6881, "step": 12852 }, { "epoch": 0.73, "learning_rate": 3.6087886683265814e-05, "loss": 0.7122, "step": 12853 }, { "epoch": 0.73, "learning_rate": 3.607375783188125e-05, "loss": 0.5558, "step": 12854 }, { "epoch": 0.73, "learning_rate": 3.605963113817501e-05, "loss": 0.7442, "step": 12855 }, { "epoch": 0.73, "learning_rate": 3.604550660262387e-05, "loss": 0.7134, "step": 12856 }, { "epoch": 0.73, "learning_rate": 3.6031384225704626e-05, "loss": 0.6831, "step": 12857 }, { "epoch": 0.73, "learning_rate": 3.601726400789391e-05, "loss": 0.7123, "step": 12858 }, { "epoch": 0.73, "learning_rate": 3.600314594966834e-05, "loss": 0.6558, "step": 12859 }, { "epoch": 0.73, "learning_rate": 3.5989030051504434e-05, "loss": 0.6898, "step": 12860 }, { "epoch": 0.73, "learning_rate": 3.597491631387867e-05, "loss": 0.6396, "step": 12861 }, { "epoch": 0.73, "learning_rate": 3.5960804737267405e-05, "loss": 0.696, "step": 12862 }, { "epoch": 0.73, "learning_rate": 3.594669532214688e-05, "loss": 0.6017, "step": 12863 }, { "epoch": 0.73, "learning_rate": 3.593258806899338e-05, "loss": 0.6799, "step": 12864 }, { "epoch": 0.73, "learning_rate": 3.591848297828305e-05, "loss": 0.736, "step": 12865 }, { "epoch": 0.73, "learning_rate": 3.590438005049199e-05, "loss": 0.782, "step": 12866 }, { "epoch": 0.73, "learning_rate": 3.589027928609623e-05, "loss": 0.7659, "step": 12867 }, { "epoch": 0.73, "learning_rate": 3.5876180685571634e-05, "loss": 0.597, "step": 12868 }, { "epoch": 0.73, "learning_rate": 3.586208424939416e-05, "loss": 0.7078, "step": 12869 }, { "epoch": 0.73, "learning_rate": 3.584798997803951e-05, "loss": 0.7041, "step": 12870 }, { "epoch": 0.73, "learning_rate": 3.5833897871983435e-05, "loss": 0.6186, "step": 12871 }, { "epoch": 0.73, "learning_rate": 3.58198079317016e-05, "loss": 0.6654, "step": 12872 }, { "epoch": 0.73, "learning_rate": 3.580572015766956e-05, "loss": 0.6849, "step": 12873 }, { "epoch": 0.73, "learning_rate": 3.579163455036286e-05, "loss": 0.7568, "step": 12874 }, { "epoch": 0.73, "learning_rate": 3.577755111025685e-05, "loss": 0.724, "step": 12875 }, { "epoch": 0.73, "learning_rate": 3.576346983782691e-05, "loss": 0.677, "step": 12876 }, { "epoch": 0.73, "learning_rate": 3.574939073354838e-05, "loss": 0.6863, "step": 12877 }, { "epoch": 0.73, "learning_rate": 3.5735313797896375e-05, "loss": 0.8078, "step": 12878 }, { "epoch": 0.73, "learning_rate": 3.5721239031346066e-05, "loss": 0.7441, "step": 12879 }, { "epoch": 0.73, "learning_rate": 3.570716643437252e-05, "loss": 0.6548, "step": 12880 }, { "epoch": 0.73, "learning_rate": 3.5693096007450754e-05, "loss": 0.6254, "step": 12881 }, { "epoch": 0.73, "learning_rate": 3.567902775105562e-05, "loss": 0.8044, "step": 12882 }, { "epoch": 0.73, "learning_rate": 3.5664961665661975e-05, "loss": 0.5367, "step": 12883 }, { "epoch": 0.73, "learning_rate": 3.565089775174465e-05, "loss": 0.7976, "step": 12884 }, { "epoch": 0.73, "learning_rate": 3.563683600977824e-05, "loss": 0.7455, "step": 12885 }, { "epoch": 0.73, "learning_rate": 3.5622776440237414e-05, "loss": 0.6534, "step": 12886 }, { "epoch": 0.73, "learning_rate": 3.560871904359676e-05, "loss": 0.7306, "step": 12887 }, { "epoch": 0.73, "learning_rate": 3.559466382033067e-05, "loss": 0.7199, "step": 12888 }, { "epoch": 0.73, "learning_rate": 3.558061077091359e-05, "loss": 0.8642, "step": 12889 }, { "epoch": 0.73, "learning_rate": 3.556655989581984e-05, "loss": 0.7584, "step": 12890 }, { "epoch": 0.73, "learning_rate": 3.555251119552371e-05, "loss": 0.6942, "step": 12891 }, { "epoch": 0.73, "learning_rate": 3.5538464670499314e-05, "loss": 0.6611, "step": 12892 }, { "epoch": 0.73, "learning_rate": 3.5524420321220795e-05, "loss": 0.6979, "step": 12893 }, { "epoch": 0.73, "learning_rate": 3.5510378148162215e-05, "loss": 0.7799, "step": 12894 }, { "epoch": 0.73, "learning_rate": 3.549633815179746e-05, "loss": 0.6722, "step": 12895 }, { "epoch": 0.73, "learning_rate": 3.5482300332600474e-05, "loss": 0.7802, "step": 12896 }, { "epoch": 0.73, "learning_rate": 3.546826469104505e-05, "loss": 0.4983, "step": 12897 }, { "epoch": 0.73, "learning_rate": 3.545423122760493e-05, "loss": 0.6967, "step": 12898 }, { "epoch": 0.73, "learning_rate": 3.5440199942753814e-05, "loss": 0.6805, "step": 12899 }, { "epoch": 0.73, "learning_rate": 3.542617083696528e-05, "loss": 0.663, "step": 12900 }, { "epoch": 0.73, "learning_rate": 3.541214391071278e-05, "loss": 0.6776, "step": 12901 }, { "epoch": 0.73, "learning_rate": 3.5398119164469825e-05, "loss": 0.7088, "step": 12902 }, { "epoch": 0.73, "learning_rate": 3.5384096598709756e-05, "loss": 0.7807, "step": 12903 }, { "epoch": 0.73, "learning_rate": 3.53700762139059e-05, "loss": 0.6354, "step": 12904 }, { "epoch": 0.73, "learning_rate": 3.535605801053147e-05, "loss": 0.7954, "step": 12905 }, { "epoch": 0.73, "learning_rate": 3.534204198905965e-05, "loss": 0.6834, "step": 12906 }, { "epoch": 0.73, "learning_rate": 3.5328028149963485e-05, "loss": 0.7855, "step": 12907 }, { "epoch": 0.73, "learning_rate": 3.531401649371594e-05, "loss": 0.6554, "step": 12908 }, { "epoch": 0.73, "learning_rate": 3.530000702078999e-05, "loss": 0.7059, "step": 12909 }, { "epoch": 0.73, "learning_rate": 3.528599973165848e-05, "loss": 0.7355, "step": 12910 }, { "epoch": 0.73, "learning_rate": 3.5271994626794204e-05, "loss": 0.6525, "step": 12911 }, { "epoch": 0.73, "learning_rate": 3.525799170666989e-05, "loss": 0.7049, "step": 12912 }, { "epoch": 0.73, "learning_rate": 3.5243990971758125e-05, "loss": 0.7089, "step": 12913 }, { "epoch": 0.73, "learning_rate": 3.522999242253151e-05, "loss": 0.6616, "step": 12914 }, { "epoch": 0.73, "learning_rate": 3.52159960594625e-05, "loss": 0.7521, "step": 12915 }, { "epoch": 0.73, "learning_rate": 3.5202001883023515e-05, "loss": 0.6368, "step": 12916 }, { "epoch": 0.73, "learning_rate": 3.518800989368691e-05, "loss": 0.7822, "step": 12917 }, { "epoch": 0.73, "learning_rate": 3.517402009192494e-05, "loss": 0.7311, "step": 12918 }, { "epoch": 0.73, "learning_rate": 3.516003247820985e-05, "loss": 0.7501, "step": 12919 }, { "epoch": 0.73, "learning_rate": 3.514604705301368e-05, "loss": 0.7323, "step": 12920 }, { "epoch": 0.73, "learning_rate": 3.513206381680852e-05, "loss": 0.7166, "step": 12921 }, { "epoch": 0.73, "learning_rate": 3.511808277006632e-05, "loss": 0.6879, "step": 12922 }, { "epoch": 0.73, "learning_rate": 3.510410391325897e-05, "loss": 0.6083, "step": 12923 }, { "epoch": 0.73, "learning_rate": 3.509012724685832e-05, "loss": 0.6773, "step": 12924 }, { "epoch": 0.73, "learning_rate": 3.50761527713361e-05, "loss": 0.701, "step": 12925 }, { "epoch": 0.73, "learning_rate": 3.506218048716402e-05, "loss": 0.7045, "step": 12926 }, { "epoch": 0.73, "learning_rate": 3.504821039481363e-05, "loss": 0.7338, "step": 12927 }, { "epoch": 0.73, "learning_rate": 3.503424249475652e-05, "loss": 0.6773, "step": 12928 }, { "epoch": 0.73, "learning_rate": 3.5020276787464056e-05, "loss": 0.7251, "step": 12929 }, { "epoch": 0.73, "learning_rate": 3.500631327340767e-05, "loss": 0.7659, "step": 12930 }, { "epoch": 0.73, "learning_rate": 3.499235195305868e-05, "loss": 0.7098, "step": 12931 }, { "epoch": 0.73, "learning_rate": 3.497839282688832e-05, "loss": 0.6644, "step": 12932 }, { "epoch": 0.73, "learning_rate": 3.49644358953677e-05, "loss": 0.634, "step": 12933 }, { "epoch": 0.73, "learning_rate": 3.4950481158967916e-05, "loss": 0.7262, "step": 12934 }, { "epoch": 0.73, "learning_rate": 3.493652861816001e-05, "loss": 0.7622, "step": 12935 }, { "epoch": 0.73, "learning_rate": 3.492257827341492e-05, "loss": 0.7323, "step": 12936 }, { "epoch": 0.73, "learning_rate": 3.490863012520346e-05, "loss": 0.6771, "step": 12937 }, { "epoch": 0.73, "learning_rate": 3.489468417399644e-05, "loss": 0.7352, "step": 12938 }, { "epoch": 0.73, "learning_rate": 3.48807404202646e-05, "loss": 0.7455, "step": 12939 }, { "epoch": 0.73, "learning_rate": 3.486679886447852e-05, "loss": 0.754, "step": 12940 }, { "epoch": 0.73, "learning_rate": 3.48528595071088e-05, "loss": 0.6734, "step": 12941 }, { "epoch": 0.73, "learning_rate": 3.483892234862592e-05, "loss": 0.6283, "step": 12942 }, { "epoch": 0.73, "learning_rate": 3.482498738950035e-05, "loss": 0.6614, "step": 12943 }, { "epoch": 0.73, "learning_rate": 3.481105463020234e-05, "loss": 0.6289, "step": 12944 }, { "epoch": 0.73, "learning_rate": 3.479712407120224e-05, "loss": 0.6737, "step": 12945 }, { "epoch": 0.73, "learning_rate": 3.4783195712970165e-05, "loss": 0.6133, "step": 12946 }, { "epoch": 0.73, "learning_rate": 3.4769269555976266e-05, "loss": 0.762, "step": 12947 }, { "epoch": 0.73, "learning_rate": 3.4755345600690605e-05, "loss": 0.76, "step": 12948 }, { "epoch": 0.73, "learning_rate": 3.4741423847583134e-05, "loss": 0.685, "step": 12949 }, { "epoch": 0.73, "learning_rate": 3.4727504297123756e-05, "loss": 0.7017, "step": 12950 }, { "epoch": 0.73, "learning_rate": 3.471358694978233e-05, "loss": 0.7496, "step": 12951 }, { "epoch": 0.73, "learning_rate": 3.469967180602855e-05, "loss": 0.7558, "step": 12952 }, { "epoch": 0.73, "learning_rate": 3.468575886633208e-05, "loss": 0.7649, "step": 12953 }, { "epoch": 0.73, "learning_rate": 3.4671848131162544e-05, "loss": 0.6737, "step": 12954 }, { "epoch": 0.73, "learning_rate": 3.465793960098945e-05, "loss": 0.791, "step": 12955 }, { "epoch": 0.73, "learning_rate": 3.464403327628227e-05, "loss": 0.6981, "step": 12956 }, { "epoch": 0.73, "learning_rate": 3.46301291575104e-05, "loss": 0.716, "step": 12957 }, { "epoch": 0.74, "learning_rate": 3.461622724514309e-05, "loss": 0.6736, "step": 12958 }, { "epoch": 0.74, "learning_rate": 3.46023275396496e-05, "loss": 0.6919, "step": 12959 }, { "epoch": 0.74, "learning_rate": 3.458843004149904e-05, "loss": 0.6494, "step": 12960 }, { "epoch": 0.74, "learning_rate": 3.457453475116051e-05, "loss": 0.6448, "step": 12961 }, { "epoch": 0.74, "learning_rate": 3.456064166910303e-05, "loss": 0.6164, "step": 12962 }, { "epoch": 0.74, "learning_rate": 3.45467507957955e-05, "loss": 0.6654, "step": 12963 }, { "epoch": 0.74, "learning_rate": 3.453286213170683e-05, "loss": 0.6574, "step": 12964 }, { "epoch": 0.74, "learning_rate": 3.451897567730572e-05, "loss": 0.7468, "step": 12965 }, { "epoch": 0.74, "learning_rate": 3.4505091433060954e-05, "loss": 0.7083, "step": 12966 }, { "epoch": 0.74, "learning_rate": 3.449120939944107e-05, "loss": 0.6926, "step": 12967 }, { "epoch": 0.74, "learning_rate": 3.4477329576914685e-05, "loss": 0.689, "step": 12968 }, { "epoch": 0.74, "learning_rate": 3.446345196595025e-05, "loss": 0.7033, "step": 12969 }, { "epoch": 0.74, "learning_rate": 3.4449576567016205e-05, "loss": 0.6659, "step": 12970 }, { "epoch": 0.74, "learning_rate": 3.4435703380580897e-05, "loss": 0.6918, "step": 12971 }, { "epoch": 0.74, "learning_rate": 3.44218324071125e-05, "loss": 0.7086, "step": 12972 }, { "epoch": 0.74, "learning_rate": 3.4407963647079286e-05, "loss": 0.6663, "step": 12973 }, { "epoch": 0.74, "learning_rate": 3.439409710094929e-05, "loss": 0.6757, "step": 12974 }, { "epoch": 0.74, "learning_rate": 3.4380232769190566e-05, "loss": 0.6195, "step": 12975 }, { "epoch": 0.74, "learning_rate": 3.4366370652271085e-05, "loss": 0.6843, "step": 12976 }, { "epoch": 0.74, "learning_rate": 3.435251075065876e-05, "loss": 0.7165, "step": 12977 }, { "epoch": 0.74, "learning_rate": 3.4338653064821326e-05, "loss": 0.6717, "step": 12978 }, { "epoch": 0.74, "learning_rate": 3.4324797595226565e-05, "loss": 0.6788, "step": 12979 }, { "epoch": 0.74, "learning_rate": 3.4310944342342146e-05, "loss": 0.6216, "step": 12980 }, { "epoch": 0.74, "learning_rate": 3.42970933066356e-05, "loss": 0.7591, "step": 12981 }, { "epoch": 0.74, "learning_rate": 3.428324448857447e-05, "loss": 0.7208, "step": 12982 }, { "epoch": 0.74, "learning_rate": 3.426939788862619e-05, "loss": 0.7632, "step": 12983 }, { "epoch": 0.74, "learning_rate": 3.4255553507258145e-05, "loss": 0.6699, "step": 12984 }, { "epoch": 0.74, "learning_rate": 3.424171134493756e-05, "loss": 0.7174, "step": 12985 }, { "epoch": 0.74, "learning_rate": 3.4227871402131675e-05, "loss": 0.7151, "step": 12986 }, { "epoch": 0.74, "learning_rate": 3.421403367930761e-05, "loss": 0.6356, "step": 12987 }, { "epoch": 0.74, "learning_rate": 3.4200198176932483e-05, "loss": 0.7403, "step": 12988 }, { "epoch": 0.74, "learning_rate": 3.41863648954732e-05, "loss": 0.6836, "step": 12989 }, { "epoch": 0.74, "learning_rate": 3.4172533835396736e-05, "loss": 0.5852, "step": 12990 }, { "epoch": 0.74, "learning_rate": 3.415870499716987e-05, "loss": 0.6756, "step": 12991 }, { "epoch": 0.74, "learning_rate": 3.4144878381259374e-05, "loss": 0.7094, "step": 12992 }, { "epoch": 0.74, "learning_rate": 3.413105398813195e-05, "loss": 0.6979, "step": 12993 }, { "epoch": 0.74, "learning_rate": 3.41172318182542e-05, "loss": 0.8257, "step": 12994 }, { "epoch": 0.74, "learning_rate": 3.4103411872092694e-05, "loss": 0.6986, "step": 12995 }, { "epoch": 0.74, "learning_rate": 3.4089594150113825e-05, "loss": 0.6111, "step": 12996 }, { "epoch": 0.74, "learning_rate": 3.407577865278404e-05, "loss": 0.7386, "step": 12997 }, { "epoch": 0.74, "learning_rate": 3.4061965380569583e-05, "loss": 0.7224, "step": 12998 }, { "epoch": 0.74, "learning_rate": 3.404815433393672e-05, "loss": 0.6372, "step": 12999 }, { "epoch": 0.74, "learning_rate": 3.4034345513351626e-05, "loss": 0.5703, "step": 13000 }, { "epoch": 0.74, "learning_rate": 3.4020538919280366e-05, "loss": 0.7301, "step": 13001 }, { "epoch": 0.74, "learning_rate": 3.400673455218899e-05, "loss": 0.6695, "step": 13002 }, { "epoch": 0.74, "learning_rate": 3.399293241254336e-05, "loss": 0.7279, "step": 13003 }, { "epoch": 0.74, "learning_rate": 3.3979132500809405e-05, "loss": 0.7316, "step": 13004 }, { "epoch": 0.74, "learning_rate": 3.396533481745284e-05, "loss": 0.6716, "step": 13005 }, { "epoch": 0.74, "learning_rate": 3.395153936293941e-05, "loss": 0.6785, "step": 13006 }, { "epoch": 0.74, "learning_rate": 3.393774613773474e-05, "loss": 0.63, "step": 13007 }, { "epoch": 0.74, "learning_rate": 3.39239551423044e-05, "loss": 0.7483, "step": 13008 }, { "epoch": 0.74, "learning_rate": 3.3910166377113894e-05, "loss": 0.6901, "step": 13009 }, { "epoch": 0.74, "learning_rate": 3.38963798426286e-05, "loss": 0.7331, "step": 13010 }, { "epoch": 0.74, "learning_rate": 3.388259553931381e-05, "loss": 0.7538, "step": 13011 }, { "epoch": 0.74, "learning_rate": 3.386881346763483e-05, "loss": 0.6946, "step": 13012 }, { "epoch": 0.74, "learning_rate": 3.385503362805682e-05, "loss": 0.709, "step": 13013 }, { "epoch": 0.74, "learning_rate": 3.384125602104491e-05, "loss": 0.7139, "step": 13014 }, { "epoch": 0.74, "learning_rate": 3.38274806470641e-05, "loss": 0.8348, "step": 13015 }, { "epoch": 0.74, "learning_rate": 3.381370750657941e-05, "loss": 0.7682, "step": 13016 }, { "epoch": 0.74, "learning_rate": 3.379993660005567e-05, "loss": 0.6908, "step": 13017 }, { "epoch": 0.74, "learning_rate": 3.378616792795764e-05, "loss": 0.8404, "step": 13018 }, { "epoch": 0.74, "learning_rate": 3.377240149075012e-05, "loss": 0.7094, "step": 13019 }, { "epoch": 0.74, "learning_rate": 3.375863728889771e-05, "loss": 0.6513, "step": 13020 }, { "epoch": 0.74, "learning_rate": 3.3744875322865034e-05, "loss": 0.705, "step": 13021 }, { "epoch": 0.74, "learning_rate": 3.37311155931166e-05, "loss": 0.7695, "step": 13022 }, { "epoch": 0.74, "learning_rate": 3.371735810011678e-05, "loss": 0.7487, "step": 13023 }, { "epoch": 0.74, "learning_rate": 3.370360284432996e-05, "loss": 0.698, "step": 13024 }, { "epoch": 0.74, "learning_rate": 3.3689849826220444e-05, "loss": 0.6624, "step": 13025 }, { "epoch": 0.74, "learning_rate": 3.367609904625236e-05, "loss": 0.6361, "step": 13026 }, { "epoch": 0.74, "learning_rate": 3.366235050488989e-05, "loss": 0.7152, "step": 13027 }, { "epoch": 0.74, "learning_rate": 3.364860420259705e-05, "loss": 0.6715, "step": 13028 }, { "epoch": 0.74, "learning_rate": 3.363486013983788e-05, "loss": 0.7326, "step": 13029 }, { "epoch": 0.74, "learning_rate": 3.3621118317076184e-05, "loss": 0.6502, "step": 13030 }, { "epoch": 0.74, "learning_rate": 3.360737873477584e-05, "loss": 0.6852, "step": 13031 }, { "epoch": 0.74, "learning_rate": 3.359364139340061e-05, "loss": 0.6741, "step": 13032 }, { "epoch": 0.74, "learning_rate": 3.357990629341411e-05, "loss": 0.6697, "step": 13033 }, { "epoch": 0.74, "learning_rate": 3.356617343527997e-05, "loss": 0.6728, "step": 13034 }, { "epoch": 0.74, "learning_rate": 3.355244281946173e-05, "loss": 0.6138, "step": 13035 }, { "epoch": 0.74, "learning_rate": 3.353871444642278e-05, "loss": 0.7791, "step": 13036 }, { "epoch": 0.74, "learning_rate": 3.352498831662653e-05, "loss": 0.7192, "step": 13037 }, { "epoch": 0.74, "learning_rate": 3.3511264430536236e-05, "loss": 0.7045, "step": 13038 }, { "epoch": 0.74, "learning_rate": 3.349754278861517e-05, "loss": 0.6649, "step": 13039 }, { "epoch": 0.74, "learning_rate": 3.348382339132645e-05, "loss": 0.7112, "step": 13040 }, { "epoch": 0.74, "learning_rate": 3.3470106239133115e-05, "loss": 0.6697, "step": 13041 }, { "epoch": 0.74, "learning_rate": 3.345639133249822e-05, "loss": 0.7333, "step": 13042 }, { "epoch": 0.74, "learning_rate": 3.344267867188458e-05, "loss": 0.636, "step": 13043 }, { "epoch": 0.74, "learning_rate": 3.34289682577551e-05, "loss": 0.6672, "step": 13044 }, { "epoch": 0.74, "learning_rate": 3.341526009057253e-05, "loss": 0.7464, "step": 13045 }, { "epoch": 0.74, "learning_rate": 3.3401554170799555e-05, "loss": 0.7085, "step": 13046 }, { "epoch": 0.74, "learning_rate": 3.3387850498898824e-05, "loss": 0.6044, "step": 13047 }, { "epoch": 0.74, "learning_rate": 3.3374149075332796e-05, "loss": 0.6164, "step": 13048 }, { "epoch": 0.74, "learning_rate": 3.336044990056401e-05, "loss": 0.7483, "step": 13049 }, { "epoch": 0.74, "learning_rate": 3.334675297505476e-05, "loss": 0.6939, "step": 13050 }, { "epoch": 0.74, "learning_rate": 3.333305829926742e-05, "loss": 0.6677, "step": 13051 }, { "epoch": 0.74, "learning_rate": 3.33193658736642e-05, "loss": 0.8013, "step": 13052 }, { "epoch": 0.74, "learning_rate": 3.330567569870725e-05, "loss": 0.7435, "step": 13053 }, { "epoch": 0.74, "learning_rate": 3.329198777485869e-05, "loss": 0.7066, "step": 13054 }, { "epoch": 0.74, "learning_rate": 3.3278302102580496e-05, "loss": 0.6979, "step": 13055 }, { "epoch": 0.74, "learning_rate": 3.3264618682334556e-05, "loss": 0.7009, "step": 13056 }, { "epoch": 0.74, "learning_rate": 3.325093751458276e-05, "loss": 0.7552, "step": 13057 }, { "epoch": 0.74, "learning_rate": 3.323725859978687e-05, "loss": 0.6969, "step": 13058 }, { "epoch": 0.74, "learning_rate": 3.32235819384086e-05, "loss": 0.7089, "step": 13059 }, { "epoch": 0.74, "learning_rate": 3.320990753090956e-05, "loss": 0.7176, "step": 13060 }, { "epoch": 0.74, "learning_rate": 3.319623537775134e-05, "loss": 0.7668, "step": 13061 }, { "epoch": 0.74, "learning_rate": 3.318256547939538e-05, "loss": 0.6428, "step": 13062 }, { "epoch": 0.74, "learning_rate": 3.316889783630304e-05, "loss": 0.7041, "step": 13063 }, { "epoch": 0.74, "learning_rate": 3.315523244893566e-05, "loss": 0.6255, "step": 13064 }, { "epoch": 0.74, "learning_rate": 3.314156931775449e-05, "loss": 0.7236, "step": 13065 }, { "epoch": 0.74, "learning_rate": 3.312790844322071e-05, "loss": 0.6569, "step": 13066 }, { "epoch": 0.74, "learning_rate": 3.311424982579543e-05, "loss": 0.8228, "step": 13067 }, { "epoch": 0.74, "learning_rate": 3.310059346593959e-05, "loss": 0.7065, "step": 13068 }, { "epoch": 0.74, "learning_rate": 3.308693936411421e-05, "loss": 0.7166, "step": 13069 }, { "epoch": 0.74, "learning_rate": 3.307328752078007e-05, "loss": 0.6598, "step": 13070 }, { "epoch": 0.74, "learning_rate": 3.305963793639801e-05, "loss": 0.5222, "step": 13071 }, { "epoch": 0.74, "learning_rate": 3.304599061142872e-05, "loss": 0.6329, "step": 13072 }, { "epoch": 0.74, "learning_rate": 3.303234554633285e-05, "loss": 0.7602, "step": 13073 }, { "epoch": 0.74, "learning_rate": 3.301870274157097e-05, "loss": 0.7323, "step": 13074 }, { "epoch": 0.74, "learning_rate": 3.3005062197603506e-05, "loss": 0.5893, "step": 13075 }, { "epoch": 0.74, "learning_rate": 3.29914239148909e-05, "loss": 0.7272, "step": 13076 }, { "epoch": 0.74, "learning_rate": 3.29777878938935e-05, "loss": 0.7823, "step": 13077 }, { "epoch": 0.74, "learning_rate": 3.29641541350715e-05, "loss": 0.6871, "step": 13078 }, { "epoch": 0.74, "learning_rate": 3.2950522638885106e-05, "loss": 0.7572, "step": 13079 }, { "epoch": 0.74, "learning_rate": 3.2936893405794444e-05, "loss": 0.7158, "step": 13080 }, { "epoch": 0.74, "learning_rate": 3.292326643625948e-05, "loss": 0.652, "step": 13081 }, { "epoch": 0.74, "learning_rate": 3.2909641730740195e-05, "loss": 0.8025, "step": 13082 }, { "epoch": 0.74, "learning_rate": 3.2896019289696446e-05, "loss": 0.6211, "step": 13083 }, { "epoch": 0.74, "learning_rate": 3.2882399113588066e-05, "loss": 0.654, "step": 13084 }, { "epoch": 0.74, "learning_rate": 3.286878120287471e-05, "loss": 0.7744, "step": 13085 }, { "epoch": 0.74, "learning_rate": 3.2855165558016044e-05, "loss": 0.7729, "step": 13086 }, { "epoch": 0.74, "learning_rate": 3.2841552179471654e-05, "loss": 0.7684, "step": 13087 }, { "epoch": 0.74, "learning_rate": 3.2827941067700996e-05, "loss": 0.747, "step": 13088 }, { "epoch": 0.74, "learning_rate": 3.281433222316348e-05, "loss": 0.7265, "step": 13089 }, { "epoch": 0.74, "learning_rate": 3.2800725646318456e-05, "loss": 0.7169, "step": 13090 }, { "epoch": 0.74, "learning_rate": 3.2787121337625206e-05, "loss": 0.6602, "step": 13091 }, { "epoch": 0.74, "learning_rate": 3.277351929754285e-05, "loss": 0.6207, "step": 13092 }, { "epoch": 0.74, "learning_rate": 3.275991952653054e-05, "loss": 0.5745, "step": 13093 }, { "epoch": 0.74, "learning_rate": 3.27463220250473e-05, "loss": 0.6886, "step": 13094 }, { "epoch": 0.74, "learning_rate": 3.273272679355206e-05, "loss": 0.6268, "step": 13095 }, { "epoch": 0.74, "learning_rate": 3.271913383250369e-05, "loss": 0.7193, "step": 13096 }, { "epoch": 0.74, "learning_rate": 3.270554314236101e-05, "loss": 0.7665, "step": 13097 }, { "epoch": 0.74, "learning_rate": 3.269195472358273e-05, "loss": 0.7488, "step": 13098 }, { "epoch": 0.74, "learning_rate": 3.267836857662754e-05, "loss": 0.7335, "step": 13099 }, { "epoch": 0.74, "learning_rate": 3.2664784701953966e-05, "loss": 0.7201, "step": 13100 }, { "epoch": 0.74, "learning_rate": 3.265120310002047e-05, "loss": 0.7296, "step": 13101 }, { "epoch": 0.74, "learning_rate": 3.26376237712855e-05, "loss": 0.696, "step": 13102 }, { "epoch": 0.74, "learning_rate": 3.262404671620738e-05, "loss": 0.7489, "step": 13103 }, { "epoch": 0.74, "learning_rate": 3.261047193524439e-05, "loss": 0.7418, "step": 13104 }, { "epoch": 0.74, "learning_rate": 3.259689942885471e-05, "loss": 0.733, "step": 13105 }, { "epoch": 0.74, "learning_rate": 3.258332919749648e-05, "loss": 0.7746, "step": 13106 }, { "epoch": 0.74, "learning_rate": 3.2569761241627696e-05, "loss": 0.7053, "step": 13107 }, { "epoch": 0.74, "learning_rate": 3.255619556170627e-05, "loss": 0.7241, "step": 13108 }, { "epoch": 0.74, "learning_rate": 3.2542632158190135e-05, "loss": 0.7283, "step": 13109 }, { "epoch": 0.74, "learning_rate": 3.252907103153706e-05, "loss": 0.7975, "step": 13110 }, { "epoch": 0.74, "learning_rate": 3.25155121822048e-05, "loss": 0.7462, "step": 13111 }, { "epoch": 0.74, "learning_rate": 3.250195561065103e-05, "loss": 0.7109, "step": 13112 }, { "epoch": 0.74, "learning_rate": 3.248840131733323e-05, "loss": 0.7562, "step": 13113 }, { "epoch": 0.74, "learning_rate": 3.247484930270898e-05, "loss": 0.635, "step": 13114 }, { "epoch": 0.74, "learning_rate": 3.2461299567235614e-05, "loss": 0.6947, "step": 13115 }, { "epoch": 0.74, "learning_rate": 3.2447752111370535e-05, "loss": 0.7654, "step": 13116 }, { "epoch": 0.74, "learning_rate": 3.243420693557097e-05, "loss": 0.6984, "step": 13117 }, { "epoch": 0.74, "learning_rate": 3.242066404029411e-05, "loss": 0.7016, "step": 13118 }, { "epoch": 0.74, "learning_rate": 3.240712342599712e-05, "loss": 0.7497, "step": 13119 }, { "epoch": 0.74, "learning_rate": 3.2393585093136944e-05, "loss": 0.624, "step": 13120 }, { "epoch": 0.74, "learning_rate": 3.238004904217061e-05, "loss": 0.6695, "step": 13121 }, { "epoch": 0.74, "learning_rate": 3.236651527355493e-05, "loss": 0.6927, "step": 13122 }, { "epoch": 0.74, "learning_rate": 3.235298378774674e-05, "loss": 0.7462, "step": 13123 }, { "epoch": 0.74, "learning_rate": 3.2339454585202756e-05, "loss": 0.6703, "step": 13124 }, { "epoch": 0.74, "learning_rate": 3.232592766637965e-05, "loss": 0.7199, "step": 13125 }, { "epoch": 0.74, "learning_rate": 3.231240303173394e-05, "loss": 0.6826, "step": 13126 }, { "epoch": 0.74, "learning_rate": 3.2298880681722144e-05, "loss": 0.685, "step": 13127 }, { "epoch": 0.74, "learning_rate": 3.2285360616800676e-05, "loss": 0.7172, "step": 13128 }, { "epoch": 0.74, "learning_rate": 3.227184283742591e-05, "loss": 0.6994, "step": 13129 }, { "epoch": 0.74, "learning_rate": 3.225832734405404e-05, "loss": 0.7426, "step": 13130 }, { "epoch": 0.74, "learning_rate": 3.224481413714128e-05, "loss": 0.688, "step": 13131 }, { "epoch": 0.74, "learning_rate": 3.223130321714377e-05, "loss": 0.6903, "step": 13132 }, { "epoch": 0.74, "learning_rate": 3.2217794584517466e-05, "loss": 0.6669, "step": 13133 }, { "epoch": 0.75, "learning_rate": 3.2204288239718374e-05, "loss": 0.7836, "step": 13134 }, { "epoch": 0.75, "learning_rate": 3.219078418320235e-05, "loss": 0.7528, "step": 13135 }, { "epoch": 0.75, "learning_rate": 3.217728241542523e-05, "loss": 0.7002, "step": 13136 }, { "epoch": 0.75, "learning_rate": 3.2163782936842654e-05, "loss": 0.8147, "step": 13137 }, { "epoch": 0.75, "learning_rate": 3.215028574791033e-05, "loss": 0.7281, "step": 13138 }, { "epoch": 0.75, "learning_rate": 3.213679084908382e-05, "loss": 0.687, "step": 13139 }, { "epoch": 0.75, "learning_rate": 3.212329824081858e-05, "loss": 0.5882, "step": 13140 }, { "epoch": 0.75, "learning_rate": 3.2109807923570026e-05, "loss": 0.7181, "step": 13141 }, { "epoch": 0.75, "learning_rate": 3.20963198977935e-05, "loss": 0.7274, "step": 13142 }, { "epoch": 0.75, "learning_rate": 3.208283416394431e-05, "loss": 0.6523, "step": 13143 }, { "epoch": 0.75, "learning_rate": 3.206935072247754e-05, "loss": 0.7791, "step": 13144 }, { "epoch": 0.75, "learning_rate": 3.205586957384838e-05, "loss": 0.7058, "step": 13145 }, { "epoch": 0.75, "learning_rate": 3.2042390718511774e-05, "loss": 0.6421, "step": 13146 }, { "epoch": 0.75, "learning_rate": 3.2028914156922705e-05, "loss": 0.7363, "step": 13147 }, { "epoch": 0.75, "learning_rate": 3.201543988953605e-05, "loss": 0.7399, "step": 13148 }, { "epoch": 0.75, "learning_rate": 3.200196791680659e-05, "loss": 0.6602, "step": 13149 }, { "epoch": 0.75, "learning_rate": 3.1988498239189035e-05, "loss": 0.6767, "step": 13150 }, { "epoch": 0.75, "learning_rate": 3.197503085713808e-05, "loss": 0.6389, "step": 13151 }, { "epoch": 0.75, "learning_rate": 3.196156577110823e-05, "loss": 0.7321, "step": 13152 }, { "epoch": 0.75, "learning_rate": 3.194810298155394e-05, "loss": 0.6604, "step": 13153 }, { "epoch": 0.75, "learning_rate": 3.193464248892964e-05, "loss": 0.7016, "step": 13154 }, { "epoch": 0.75, "learning_rate": 3.1921184293689665e-05, "loss": 0.7348, "step": 13155 }, { "epoch": 0.75, "learning_rate": 3.190772839628827e-05, "loss": 0.6948, "step": 13156 }, { "epoch": 0.75, "learning_rate": 3.189427479717966e-05, "loss": 0.7206, "step": 13157 }, { "epoch": 0.75, "learning_rate": 3.188082349681788e-05, "loss": 0.5782, "step": 13158 }, { "epoch": 0.75, "learning_rate": 3.186737449565692e-05, "loss": 0.7579, "step": 13159 }, { "epoch": 0.75, "learning_rate": 3.185392779415076e-05, "loss": 0.7813, "step": 13160 }, { "epoch": 0.75, "learning_rate": 3.184048339275325e-05, "loss": 0.6828, "step": 13161 }, { "epoch": 0.75, "learning_rate": 3.182704129191819e-05, "loss": 0.6559, "step": 13162 }, { "epoch": 0.75, "learning_rate": 3.1813601492099274e-05, "loss": 0.6644, "step": 13163 }, { "epoch": 0.75, "learning_rate": 3.1800163993750166e-05, "loss": 0.7092, "step": 13164 }, { "epoch": 0.75, "learning_rate": 3.178672879732435e-05, "loss": 0.7436, "step": 13165 }, { "epoch": 0.75, "learning_rate": 3.1773295903275355e-05, "loss": 0.7369, "step": 13166 }, { "epoch": 0.75, "learning_rate": 3.175986531205654e-05, "loss": 0.7115, "step": 13167 }, { "epoch": 0.75, "learning_rate": 3.174643702412122e-05, "loss": 0.7396, "step": 13168 }, { "epoch": 0.75, "learning_rate": 3.173301103992267e-05, "loss": 0.7815, "step": 13169 }, { "epoch": 0.75, "learning_rate": 3.1719587359914015e-05, "loss": 0.7494, "step": 13170 }, { "epoch": 0.75, "learning_rate": 3.170616598454841e-05, "loss": 0.7535, "step": 13171 }, { "epoch": 0.75, "learning_rate": 3.169274691427877e-05, "loss": 0.7583, "step": 13172 }, { "epoch": 0.75, "learning_rate": 3.1679330149558096e-05, "loss": 0.7753, "step": 13173 }, { "epoch": 0.75, "learning_rate": 3.166591569083916e-05, "loss": 0.5933, "step": 13174 }, { "epoch": 0.75, "learning_rate": 3.1652503538574804e-05, "loss": 0.5836, "step": 13175 }, { "epoch": 0.75, "learning_rate": 3.1639093693217694e-05, "loss": 0.6902, "step": 13176 }, { "epoch": 0.75, "learning_rate": 3.162568615522048e-05, "loss": 0.7174, "step": 13177 }, { "epoch": 0.75, "learning_rate": 3.1612280925035657e-05, "loss": 0.7258, "step": 13178 }, { "epoch": 0.75, "learning_rate": 3.159887800311569e-05, "loss": 0.6987, "step": 13179 }, { "epoch": 0.75, "learning_rate": 3.158547738991303e-05, "loss": 0.7494, "step": 13180 }, { "epoch": 0.75, "learning_rate": 3.157207908587989e-05, "loss": 0.6237, "step": 13181 }, { "epoch": 0.75, "learning_rate": 3.155868309146855e-05, "loss": 0.8023, "step": 13182 }, { "epoch": 0.75, "learning_rate": 3.154528940713113e-05, "loss": 0.5956, "step": 13183 }, { "epoch": 0.75, "learning_rate": 3.1531898033319766e-05, "loss": 0.8589, "step": 13184 }, { "epoch": 0.75, "learning_rate": 3.151850897048637e-05, "loss": 0.5984, "step": 13185 }, { "epoch": 0.75, "learning_rate": 3.1505122219082894e-05, "loss": 0.6806, "step": 13186 }, { "epoch": 0.75, "learning_rate": 3.149173777956118e-05, "loss": 0.6922, "step": 13187 }, { "epoch": 0.75, "learning_rate": 3.147835565237302e-05, "loss": 0.6116, "step": 13188 }, { "epoch": 0.75, "learning_rate": 3.1464975837970036e-05, "loss": 0.716, "step": 13189 }, { "epoch": 0.75, "learning_rate": 3.145159833680388e-05, "loss": 0.7447, "step": 13190 }, { "epoch": 0.75, "learning_rate": 3.1438223149326016e-05, "loss": 0.7179, "step": 13191 }, { "epoch": 0.75, "learning_rate": 3.142485027598795e-05, "loss": 0.5843, "step": 13192 }, { "epoch": 0.75, "learning_rate": 3.141147971724102e-05, "loss": 0.7509, "step": 13193 }, { "epoch": 0.75, "learning_rate": 3.139811147353653e-05, "loss": 0.7088, "step": 13194 }, { "epoch": 0.75, "learning_rate": 3.138474554532572e-05, "loss": 0.7329, "step": 13195 }, { "epoch": 0.75, "learning_rate": 3.137138193305967e-05, "loss": 0.7238, "step": 13196 }, { "epoch": 0.75, "learning_rate": 3.13580206371895e-05, "loss": 0.8784, "step": 13197 }, { "epoch": 0.75, "learning_rate": 3.134466165816612e-05, "loss": 0.7049, "step": 13198 }, { "epoch": 0.75, "learning_rate": 3.133130499644047e-05, "loss": 0.6789, "step": 13199 }, { "epoch": 0.75, "learning_rate": 3.131795065246336e-05, "loss": 0.6951, "step": 13200 }, { "epoch": 0.75, "learning_rate": 3.1304598626685545e-05, "loss": 0.8229, "step": 13201 }, { "epoch": 0.75, "learning_rate": 3.129124891955771e-05, "loss": 0.7102, "step": 13202 }, { "epoch": 0.75, "learning_rate": 3.127790153153043e-05, "loss": 0.6778, "step": 13203 }, { "epoch": 0.75, "learning_rate": 3.126455646305416e-05, "loss": 0.6556, "step": 13204 }, { "epoch": 0.75, "learning_rate": 3.125121371457939e-05, "loss": 0.7325, "step": 13205 }, { "epoch": 0.75, "learning_rate": 3.123787328655644e-05, "loss": 0.6797, "step": 13206 }, { "epoch": 0.75, "learning_rate": 3.122453517943561e-05, "loss": 0.618, "step": 13207 }, { "epoch": 0.75, "learning_rate": 3.121119939366709e-05, "loss": 0.73, "step": 13208 }, { "epoch": 0.75, "learning_rate": 3.119786592970102e-05, "loss": 0.7202, "step": 13209 }, { "epoch": 0.75, "learning_rate": 3.118453478798743e-05, "loss": 0.6771, "step": 13210 }, { "epoch": 0.75, "learning_rate": 3.1171205968976226e-05, "loss": 0.8139, "step": 13211 }, { "epoch": 0.75, "learning_rate": 3.1157879473117325e-05, "loss": 0.709, "step": 13212 }, { "epoch": 0.75, "learning_rate": 3.114455530086054e-05, "loss": 0.8007, "step": 13213 }, { "epoch": 0.75, "learning_rate": 3.1131233452655594e-05, "loss": 0.7119, "step": 13214 }, { "epoch": 0.75, "learning_rate": 3.111791392895214e-05, "loss": 0.657, "step": 13215 }, { "epoch": 0.75, "learning_rate": 3.1104596730199764e-05, "loss": 0.7076, "step": 13216 }, { "epoch": 0.75, "learning_rate": 3.1091281856847945e-05, "loss": 0.68, "step": 13217 }, { "epoch": 0.75, "learning_rate": 3.107796930934603e-05, "loss": 0.7248, "step": 13218 }, { "epoch": 0.75, "learning_rate": 3.106465908814342e-05, "loss": 0.7214, "step": 13219 }, { "epoch": 0.75, "learning_rate": 3.105135119368936e-05, "loss": 0.6673, "step": 13220 }, { "epoch": 0.75, "learning_rate": 3.103804562643302e-05, "loss": 0.6985, "step": 13221 }, { "epoch": 0.75, "learning_rate": 3.1024742386823524e-05, "loss": 0.668, "step": 13222 }, { "epoch": 0.75, "learning_rate": 3.101144147530983e-05, "loss": 0.7403, "step": 13223 }, { "epoch": 0.75, "learning_rate": 3.099814289234093e-05, "loss": 0.6857, "step": 13224 }, { "epoch": 0.75, "learning_rate": 3.09848466383657e-05, "loss": 0.7403, "step": 13225 }, { "epoch": 0.75, "learning_rate": 3.097155271383285e-05, "loss": 0.6092, "step": 13226 }, { "epoch": 0.75, "learning_rate": 3.095826111919115e-05, "loss": 0.8025, "step": 13227 }, { "epoch": 0.75, "learning_rate": 3.0944971854889195e-05, "loss": 0.6953, "step": 13228 }, { "epoch": 0.75, "learning_rate": 3.093168492137557e-05, "loss": 0.7726, "step": 13229 }, { "epoch": 0.75, "learning_rate": 3.0918400319098696e-05, "loss": 0.6341, "step": 13230 }, { "epoch": 0.75, "learning_rate": 3.0905118048506974e-05, "loss": 0.6585, "step": 13231 }, { "epoch": 0.75, "learning_rate": 3.089183811004876e-05, "loss": 0.6567, "step": 13232 }, { "epoch": 0.75, "learning_rate": 3.087856050417223e-05, "loss": 0.691, "step": 13233 }, { "epoch": 0.75, "learning_rate": 3.086528523132555e-05, "loss": 0.6078, "step": 13234 }, { "epoch": 0.75, "learning_rate": 3.085201229195684e-05, "loss": 0.7161, "step": 13235 }, { "epoch": 0.75, "learning_rate": 3.083874168651403e-05, "loss": 0.6659, "step": 13236 }, { "epoch": 0.75, "learning_rate": 3.0825473415445074e-05, "loss": 0.7544, "step": 13237 }, { "epoch": 0.75, "learning_rate": 3.0812207479197806e-05, "loss": 0.8418, "step": 13238 }, { "epoch": 0.75, "learning_rate": 3.079894387821999e-05, "loss": 0.6342, "step": 13239 }, { "epoch": 0.75, "learning_rate": 3.078568261295933e-05, "loss": 0.8028, "step": 13240 }, { "epoch": 0.75, "learning_rate": 3.077242368386337e-05, "loss": 0.7191, "step": 13241 }, { "epoch": 0.75, "learning_rate": 3.07591670913797e-05, "loss": 0.7163, "step": 13242 }, { "epoch": 0.75, "learning_rate": 3.07459128359557e-05, "loss": 0.6584, "step": 13243 }, { "epoch": 0.75, "learning_rate": 3.073266091803878e-05, "loss": 0.5437, "step": 13244 }, { "epoch": 0.75, "learning_rate": 3.07194113380762e-05, "loss": 0.6292, "step": 13245 }, { "epoch": 0.75, "learning_rate": 3.070616409651519e-05, "loss": 0.7013, "step": 13246 }, { "epoch": 0.75, "learning_rate": 3.069291919380289e-05, "loss": 0.6931, "step": 13247 }, { "epoch": 0.75, "learning_rate": 3.067967663038631e-05, "loss": 0.7052, "step": 13248 }, { "epoch": 0.75, "learning_rate": 3.0666436406712485e-05, "loss": 0.7106, "step": 13249 }, { "epoch": 0.75, "learning_rate": 3.0653198523228224e-05, "loss": 0.7185, "step": 13250 }, { "epoch": 0.75, "learning_rate": 3.0639962980380376e-05, "loss": 0.6944, "step": 13251 }, { "epoch": 0.75, "learning_rate": 3.062672977861568e-05, "loss": 0.6734, "step": 13252 }, { "epoch": 0.75, "learning_rate": 3.06134989183808e-05, "loss": 0.6677, "step": 13253 }, { "epoch": 0.75, "learning_rate": 3.0600270400122335e-05, "loss": 0.6713, "step": 13254 }, { "epoch": 0.75, "learning_rate": 3.0587044224286746e-05, "loss": 0.631, "step": 13255 }, { "epoch": 0.75, "learning_rate": 3.057382039132043e-05, "loss": 0.6107, "step": 13256 }, { "epoch": 0.75, "learning_rate": 3.0560598901669745e-05, "loss": 0.5877, "step": 13257 }, { "epoch": 0.75, "learning_rate": 3.054737975578097e-05, "loss": 0.6887, "step": 13258 }, { "epoch": 0.75, "learning_rate": 3.053416295410026e-05, "loss": 0.6548, "step": 13259 }, { "epoch": 0.75, "learning_rate": 3.052094849707374e-05, "loss": 0.7142, "step": 13260 }, { "epoch": 0.75, "learning_rate": 3.0507736385147458e-05, "loss": 0.756, "step": 13261 }, { "epoch": 0.75, "learning_rate": 3.0494526618767328e-05, "loss": 0.6338, "step": 13262 }, { "epoch": 0.75, "learning_rate": 3.0481319198379166e-05, "loss": 0.6843, "step": 13263 }, { "epoch": 0.75, "learning_rate": 3.0468114124428803e-05, "loss": 0.6968, "step": 13264 }, { "epoch": 0.75, "learning_rate": 3.045491139736194e-05, "loss": 0.7505, "step": 13265 }, { "epoch": 0.75, "learning_rate": 3.0441711017624208e-05, "loss": 0.7022, "step": 13266 }, { "epoch": 0.75, "learning_rate": 3.0428512985661197e-05, "loss": 0.6532, "step": 13267 }, { "epoch": 0.75, "learning_rate": 3.0415317301918277e-05, "loss": 0.6287, "step": 13268 }, { "epoch": 0.75, "learning_rate": 3.0402123966840934e-05, "loss": 0.6431, "step": 13269 }, { "epoch": 0.75, "learning_rate": 3.0388932980874406e-05, "loss": 0.6774, "step": 13270 }, { "epoch": 0.75, "learning_rate": 3.037574434446395e-05, "loss": 0.6912, "step": 13271 }, { "epoch": 0.75, "learning_rate": 3.0362558058054715e-05, "loss": 0.6558, "step": 13272 }, { "epoch": 0.75, "learning_rate": 3.034937412209178e-05, "loss": 0.711, "step": 13273 }, { "epoch": 0.75, "learning_rate": 3.033619253702016e-05, "loss": 0.738, "step": 13274 }, { "epoch": 0.75, "learning_rate": 3.0323013303284707e-05, "loss": 0.6672, "step": 13275 }, { "epoch": 0.75, "learning_rate": 3.0309836421330296e-05, "loss": 0.8257, "step": 13276 }, { "epoch": 0.75, "learning_rate": 3.0296661891601708e-05, "loss": 0.7759, "step": 13277 }, { "epoch": 0.75, "learning_rate": 3.0283489714543556e-05, "loss": 0.7328, "step": 13278 }, { "epoch": 0.75, "learning_rate": 3.0270319890600462e-05, "loss": 0.6323, "step": 13279 }, { "epoch": 0.75, "learning_rate": 3.0257152420216973e-05, "loss": 0.7621, "step": 13280 }, { "epoch": 0.75, "learning_rate": 3.0243987303837462e-05, "loss": 0.7202, "step": 13281 }, { "epoch": 0.75, "learning_rate": 3.0230824541906334e-05, "loss": 0.7088, "step": 13282 }, { "epoch": 0.75, "learning_rate": 3.021766413486784e-05, "loss": 0.7657, "step": 13283 }, { "epoch": 0.75, "learning_rate": 3.020450608316623e-05, "loss": 0.6274, "step": 13284 }, { "epoch": 0.75, "learning_rate": 3.0191350387245554e-05, "loss": 0.7157, "step": 13285 }, { "epoch": 0.75, "learning_rate": 3.0178197047549882e-05, "loss": 0.689, "step": 13286 }, { "epoch": 0.75, "learning_rate": 3.0165046064523205e-05, "loss": 0.7245, "step": 13287 }, { "epoch": 0.75, "learning_rate": 3.0151897438609344e-05, "loss": 0.7379, "step": 13288 }, { "epoch": 0.75, "learning_rate": 3.013875117025212e-05, "loss": 0.7585, "step": 13289 }, { "epoch": 0.75, "learning_rate": 3.0125607259895273e-05, "loss": 0.7174, "step": 13290 }, { "epoch": 0.75, "learning_rate": 3.0112465707982417e-05, "loss": 0.75, "step": 13291 }, { "epoch": 0.75, "learning_rate": 3.0099326514957183e-05, "loss": 0.6766, "step": 13292 }, { "epoch": 0.75, "learning_rate": 3.0086189681262955e-05, "loss": 0.639, "step": 13293 }, { "epoch": 0.75, "learning_rate": 3.0073055207343215e-05, "loss": 0.7359, "step": 13294 }, { "epoch": 0.75, "learning_rate": 3.0059923093641216e-05, "loss": 0.6511, "step": 13295 }, { "epoch": 0.75, "learning_rate": 3.0046793340600245e-05, "loss": 0.7063, "step": 13296 }, { "epoch": 0.75, "learning_rate": 3.0033665948663448e-05, "loss": 0.7822, "step": 13297 }, { "epoch": 0.75, "learning_rate": 3.002054091827393e-05, "loss": 0.6803, "step": 13298 }, { "epoch": 0.75, "learning_rate": 3.000741824987471e-05, "loss": 0.7419, "step": 13299 }, { "epoch": 0.75, "learning_rate": 2.999429794390869e-05, "loss": 0.66, "step": 13300 }, { "epoch": 0.75, "learning_rate": 2.9981180000818676e-05, "loss": 0.669, "step": 13301 }, { "epoch": 0.75, "learning_rate": 2.996806442104746e-05, "loss": 0.6116, "step": 13302 }, { "epoch": 0.75, "learning_rate": 2.995495120503775e-05, "loss": 0.6976, "step": 13303 }, { "epoch": 0.75, "learning_rate": 2.994184035323213e-05, "loss": 0.7043, "step": 13304 }, { "epoch": 0.75, "learning_rate": 2.9928731866073135e-05, "loss": 0.6381, "step": 13305 }, { "epoch": 0.75, "learning_rate": 2.9915625744003228e-05, "loss": 0.7155, "step": 13306 }, { "epoch": 0.75, "learning_rate": 2.990252198746477e-05, "loss": 0.7164, "step": 13307 }, { "epoch": 0.75, "learning_rate": 2.9889420596899998e-05, "loss": 0.6904, "step": 13308 }, { "epoch": 0.75, "learning_rate": 2.9876321572751144e-05, "loss": 0.6638, "step": 13309 }, { "epoch": 0.76, "learning_rate": 2.9863224915460353e-05, "loss": 0.713, "step": 13310 }, { "epoch": 0.76, "learning_rate": 2.9850130625469653e-05, "loss": 0.6822, "step": 13311 }, { "epoch": 0.76, "learning_rate": 2.9837038703221054e-05, "loss": 0.7524, "step": 13312 }, { "epoch": 0.76, "learning_rate": 2.982394914915638e-05, "loss": 0.6348, "step": 13313 }, { "epoch": 0.76, "learning_rate": 2.9810861963717495e-05, "loss": 0.628, "step": 13314 }, { "epoch": 0.76, "learning_rate": 2.9797777147346062e-05, "loss": 0.7142, "step": 13315 }, { "epoch": 0.76, "learning_rate": 2.9784694700483762e-05, "loss": 0.6757, "step": 13316 }, { "epoch": 0.76, "learning_rate": 2.9771614623572174e-05, "loss": 0.6141, "step": 13317 }, { "epoch": 0.76, "learning_rate": 2.9758536917052758e-05, "loss": 0.711, "step": 13318 }, { "epoch": 0.76, "learning_rate": 2.9745461581366974e-05, "loss": 0.72, "step": 13319 }, { "epoch": 0.76, "learning_rate": 2.9732388616956085e-05, "loss": 0.6196, "step": 13320 }, { "epoch": 0.76, "learning_rate": 2.971931802426139e-05, "loss": 0.5587, "step": 13321 }, { "epoch": 0.76, "learning_rate": 2.9706249803723994e-05, "loss": 0.6814, "step": 13322 }, { "epoch": 0.76, "learning_rate": 2.969318395578502e-05, "loss": 0.6326, "step": 13323 }, { "epoch": 0.76, "learning_rate": 2.9680120480885475e-05, "loss": 0.7185, "step": 13324 }, { "epoch": 0.76, "learning_rate": 2.9667059379466322e-05, "loss": 0.6873, "step": 13325 }, { "epoch": 0.76, "learning_rate": 2.9654000651968328e-05, "loss": 0.6879, "step": 13326 }, { "epoch": 0.76, "learning_rate": 2.964094429883231e-05, "loss": 0.6051, "step": 13327 }, { "epoch": 0.76, "learning_rate": 2.962789032049894e-05, "loss": 0.8024, "step": 13328 }, { "epoch": 0.76, "learning_rate": 2.9614838717408867e-05, "loss": 0.734, "step": 13329 }, { "epoch": 0.76, "learning_rate": 2.960178949000254e-05, "loss": 0.6753, "step": 13330 }, { "epoch": 0.76, "learning_rate": 2.9588742638720446e-05, "loss": 0.6648, "step": 13331 }, { "epoch": 0.76, "learning_rate": 2.9575698164002987e-05, "loss": 0.6875, "step": 13332 }, { "epoch": 0.76, "learning_rate": 2.9562656066290374e-05, "loss": 0.7609, "step": 13333 }, { "epoch": 0.76, "learning_rate": 2.9549616346022845e-05, "loss": 0.7573, "step": 13334 }, { "epoch": 0.76, "learning_rate": 2.953657900364053e-05, "loss": 0.7413, "step": 13335 }, { "epoch": 0.76, "learning_rate": 2.9523544039583496e-05, "loss": 0.7198, "step": 13336 }, { "epoch": 0.76, "learning_rate": 2.9510511454291655e-05, "loss": 0.6892, "step": 13337 }, { "epoch": 0.76, "learning_rate": 2.9497481248204917e-05, "loss": 0.6704, "step": 13338 }, { "epoch": 0.76, "learning_rate": 2.9484453421763116e-05, "loss": 0.7011, "step": 13339 }, { "epoch": 0.76, "learning_rate": 2.9471427975405918e-05, "loss": 0.7637, "step": 13340 }, { "epoch": 0.76, "learning_rate": 2.945840490957299e-05, "loss": 0.7266, "step": 13341 }, { "epoch": 0.76, "learning_rate": 2.94453842247039e-05, "loss": 0.6965, "step": 13342 }, { "epoch": 0.76, "learning_rate": 2.943236592123817e-05, "loss": 0.6624, "step": 13343 }, { "epoch": 0.76, "learning_rate": 2.941934999961512e-05, "loss": 0.7485, "step": 13344 }, { "epoch": 0.76, "learning_rate": 2.940633646027414e-05, "loss": 0.773, "step": 13345 }, { "epoch": 0.76, "learning_rate": 2.9393325303654417e-05, "loss": 0.7828, "step": 13346 }, { "epoch": 0.76, "learning_rate": 2.9380316530195128e-05, "loss": 0.7259, "step": 13347 }, { "epoch": 0.76, "learning_rate": 2.936731014033538e-05, "loss": 0.6794, "step": 13348 }, { "epoch": 0.76, "learning_rate": 2.9354306134514142e-05, "loss": 0.7331, "step": 13349 }, { "epoch": 0.76, "learning_rate": 2.9341304513170355e-05, "loss": 0.6942, "step": 13350 }, { "epoch": 0.76, "learning_rate": 2.9328305276742874e-05, "loss": 0.7119, "step": 13351 }, { "epoch": 0.76, "learning_rate": 2.931530842567044e-05, "loss": 0.6475, "step": 13352 }, { "epoch": 0.76, "learning_rate": 2.9302313960391693e-05, "loss": 0.6947, "step": 13353 }, { "epoch": 0.76, "learning_rate": 2.9289321881345254e-05, "loss": 0.5902, "step": 13354 }, { "epoch": 0.76, "learning_rate": 2.9276332188969656e-05, "loss": 0.6698, "step": 13355 }, { "epoch": 0.76, "learning_rate": 2.9263344883703313e-05, "loss": 0.7572, "step": 13356 }, { "epoch": 0.76, "learning_rate": 2.9250359965984642e-05, "loss": 0.6566, "step": 13357 }, { "epoch": 0.76, "learning_rate": 2.9237377436251857e-05, "loss": 0.7129, "step": 13358 }, { "epoch": 0.76, "learning_rate": 2.922439729494314e-05, "loss": 0.6271, "step": 13359 }, { "epoch": 0.76, "learning_rate": 2.9211419542496632e-05, "loss": 0.6884, "step": 13360 }, { "epoch": 0.76, "learning_rate": 2.919844417935036e-05, "loss": 0.7465, "step": 13361 }, { "epoch": 0.76, "learning_rate": 2.9185471205942283e-05, "loss": 0.7061, "step": 13362 }, { "epoch": 0.76, "learning_rate": 2.9172500622710263e-05, "loss": 0.7602, "step": 13363 }, { "epoch": 0.76, "learning_rate": 2.915953243009214e-05, "loss": 0.7569, "step": 13364 }, { "epoch": 0.76, "learning_rate": 2.9146566628525552e-05, "loss": 0.7844, "step": 13365 }, { "epoch": 0.76, "learning_rate": 2.913360321844819e-05, "loss": 0.728, "step": 13366 }, { "epoch": 0.76, "learning_rate": 2.9120642200297544e-05, "loss": 0.8262, "step": 13367 }, { "epoch": 0.76, "learning_rate": 2.91076835745111e-05, "loss": 0.6167, "step": 13368 }, { "epoch": 0.76, "learning_rate": 2.9094727341526275e-05, "loss": 0.7606, "step": 13369 }, { "epoch": 0.76, "learning_rate": 2.9081773501780373e-05, "loss": 0.706, "step": 13370 }, { "epoch": 0.76, "learning_rate": 2.9068822055710586e-05, "loss": 0.7961, "step": 13371 }, { "epoch": 0.76, "learning_rate": 2.9055873003754064e-05, "loss": 0.704, "step": 13372 }, { "epoch": 0.76, "learning_rate": 2.904292634634793e-05, "loss": 0.7042, "step": 13373 }, { "epoch": 0.76, "learning_rate": 2.9029982083929085e-05, "loss": 0.6399, "step": 13374 }, { "epoch": 0.76, "learning_rate": 2.901704021693448e-05, "loss": 0.6464, "step": 13375 }, { "epoch": 0.76, "learning_rate": 2.900410074580092e-05, "loss": 0.7246, "step": 13376 }, { "epoch": 0.76, "learning_rate": 2.8991163670965182e-05, "loss": 0.6183, "step": 13377 }, { "epoch": 0.76, "learning_rate": 2.8978228992863875e-05, "loss": 0.779, "step": 13378 }, { "epoch": 0.76, "learning_rate": 2.89652967119336e-05, "loss": 0.825, "step": 13379 }, { "epoch": 0.76, "learning_rate": 2.8952366828610876e-05, "loss": 0.7646, "step": 13380 }, { "epoch": 0.76, "learning_rate": 2.8939439343332086e-05, "loss": 0.663, "step": 13381 }, { "epoch": 0.76, "learning_rate": 2.8926514256533567e-05, "loss": 0.6089, "step": 13382 }, { "epoch": 0.76, "learning_rate": 2.8913591568651597e-05, "loss": 0.6557, "step": 13383 }, { "epoch": 0.76, "learning_rate": 2.890067128012237e-05, "loss": 0.6732, "step": 13384 }, { "epoch": 0.76, "learning_rate": 2.8887753391381924e-05, "loss": 0.6838, "step": 13385 }, { "epoch": 0.76, "learning_rate": 2.8874837902866304e-05, "loss": 0.7015, "step": 13386 }, { "epoch": 0.76, "learning_rate": 2.886192481501142e-05, "loss": 0.7656, "step": 13387 }, { "epoch": 0.76, "learning_rate": 2.8849014128253195e-05, "loss": 0.6777, "step": 13388 }, { "epoch": 0.76, "learning_rate": 2.8836105843027295e-05, "loss": 0.6878, "step": 13389 }, { "epoch": 0.76, "learning_rate": 2.8823199959769488e-05, "loss": 0.755, "step": 13390 }, { "epoch": 0.76, "learning_rate": 2.8810296478915322e-05, "loss": 0.6419, "step": 13391 }, { "epoch": 0.76, "learning_rate": 2.879739540090036e-05, "loss": 0.7831, "step": 13392 }, { "epoch": 0.76, "learning_rate": 2.878449672616004e-05, "loss": 0.7237, "step": 13393 }, { "epoch": 0.76, "learning_rate": 2.877160045512971e-05, "loss": 0.7066, "step": 13394 }, { "epoch": 0.76, "learning_rate": 2.875870658824471e-05, "loss": 0.7327, "step": 13395 }, { "epoch": 0.76, "learning_rate": 2.8745815125940158e-05, "loss": 0.6564, "step": 13396 }, { "epoch": 0.76, "learning_rate": 2.8732926068651256e-05, "loss": 0.7005, "step": 13397 }, { "epoch": 0.76, "learning_rate": 2.8720039416812973e-05, "loss": 0.6762, "step": 13398 }, { "epoch": 0.76, "learning_rate": 2.87071551708603e-05, "loss": 0.6941, "step": 13399 }, { "epoch": 0.76, "learning_rate": 2.8694273331228104e-05, "loss": 0.7682, "step": 13400 }, { "epoch": 0.76, "learning_rate": 2.8681393898351194e-05, "loss": 0.6231, "step": 13401 }, { "epoch": 0.76, "learning_rate": 2.866851687266432e-05, "loss": 0.632, "step": 13402 }, { "epoch": 0.76, "learning_rate": 2.8655642254602067e-05, "loss": 0.6471, "step": 13403 }, { "epoch": 0.76, "learning_rate": 2.8642770044598966e-05, "loss": 0.8053, "step": 13404 }, { "epoch": 0.76, "learning_rate": 2.8629900243089513e-05, "loss": 0.6718, "step": 13405 }, { "epoch": 0.76, "learning_rate": 2.8617032850508098e-05, "loss": 0.7417, "step": 13406 }, { "epoch": 0.76, "learning_rate": 2.8604167867289045e-05, "loss": 0.6749, "step": 13407 }, { "epoch": 0.76, "learning_rate": 2.8591305293866553e-05, "loss": 0.7072, "step": 13408 }, { "epoch": 0.76, "learning_rate": 2.8578445130674836e-05, "loss": 0.7859, "step": 13409 }, { "epoch": 0.76, "learning_rate": 2.8565587378147885e-05, "loss": 0.6906, "step": 13410 }, { "epoch": 0.76, "learning_rate": 2.8552732036719687e-05, "loss": 0.6838, "step": 13411 }, { "epoch": 0.76, "learning_rate": 2.8539879106824153e-05, "loss": 0.6984, "step": 13412 }, { "epoch": 0.76, "learning_rate": 2.852702858889511e-05, "loss": 0.749, "step": 13413 }, { "epoch": 0.76, "learning_rate": 2.8514180483366305e-05, "loss": 0.7293, "step": 13414 }, { "epoch": 0.76, "learning_rate": 2.850133479067141e-05, "loss": 0.7612, "step": 13415 }, { "epoch": 0.76, "learning_rate": 2.8488491511243942e-05, "loss": 0.6824, "step": 13416 }, { "epoch": 0.76, "learning_rate": 2.8475650645517472e-05, "loss": 0.6842, "step": 13417 }, { "epoch": 0.76, "learning_rate": 2.8462812193925338e-05, "loss": 0.7644, "step": 13418 }, { "epoch": 0.76, "learning_rate": 2.8449976156900905e-05, "loss": 0.8191, "step": 13419 }, { "epoch": 0.76, "learning_rate": 2.8437142534877424e-05, "loss": 0.6775, "step": 13420 }, { "epoch": 0.76, "learning_rate": 2.8424311328288068e-05, "loss": 0.727, "step": 13421 }, { "epoch": 0.76, "learning_rate": 2.8411482537565948e-05, "loss": 0.6714, "step": 13422 }, { "epoch": 0.76, "learning_rate": 2.8398656163144e-05, "loss": 0.7042, "step": 13423 }, { "epoch": 0.76, "learning_rate": 2.83858322054552e-05, "loss": 0.6101, "step": 13424 }, { "epoch": 0.76, "learning_rate": 2.83730106649324e-05, "loss": 0.7636, "step": 13425 }, { "epoch": 0.76, "learning_rate": 2.8360191542008308e-05, "loss": 0.748, "step": 13426 }, { "epoch": 0.76, "learning_rate": 2.8347374837115636e-05, "loss": 0.7109, "step": 13427 }, { "epoch": 0.76, "learning_rate": 2.8334560550686984e-05, "loss": 0.7911, "step": 13428 }, { "epoch": 0.76, "learning_rate": 2.8321748683154893e-05, "loss": 0.801, "step": 13429 }, { "epoch": 0.76, "learning_rate": 2.8308939234951726e-05, "loss": 0.6799, "step": 13430 }, { "epoch": 0.76, "learning_rate": 2.8296132206509884e-05, "loss": 0.7637, "step": 13431 }, { "epoch": 0.76, "learning_rate": 2.8283327598261666e-05, "loss": 0.594, "step": 13432 }, { "epoch": 0.76, "learning_rate": 2.8270525410639193e-05, "loss": 0.7537, "step": 13433 }, { "epoch": 0.76, "learning_rate": 2.8257725644074594e-05, "loss": 0.5606, "step": 13434 }, { "epoch": 0.76, "learning_rate": 2.824492829899994e-05, "loss": 0.7988, "step": 13435 }, { "epoch": 0.76, "learning_rate": 2.8232133375847113e-05, "loss": 0.7291, "step": 13436 }, { "epoch": 0.76, "learning_rate": 2.8219340875048005e-05, "loss": 0.7178, "step": 13437 }, { "epoch": 0.76, "learning_rate": 2.8206550797034393e-05, "loss": 0.6844, "step": 13438 }, { "epoch": 0.76, "learning_rate": 2.8193763142237974e-05, "loss": 0.7363, "step": 13439 }, { "epoch": 0.76, "learning_rate": 2.81809779110904e-05, "loss": 0.6479, "step": 13440 }, { "epoch": 0.76, "learning_rate": 2.8168195104023143e-05, "loss": 0.6864, "step": 13441 }, { "epoch": 0.76, "learning_rate": 2.8155414721467722e-05, "loss": 0.7211, "step": 13442 }, { "epoch": 0.76, "learning_rate": 2.8142636763855436e-05, "loss": 0.7267, "step": 13443 }, { "epoch": 0.76, "learning_rate": 2.8129861231617615e-05, "loss": 0.7039, "step": 13444 }, { "epoch": 0.76, "learning_rate": 2.8117088125185465e-05, "loss": 0.7201, "step": 13445 }, { "epoch": 0.76, "learning_rate": 2.8104317444990115e-05, "loss": 0.6007, "step": 13446 }, { "epoch": 0.76, "learning_rate": 2.809154919146263e-05, "loss": 0.7429, "step": 13447 }, { "epoch": 0.76, "learning_rate": 2.8078783365033944e-05, "loss": 0.7221, "step": 13448 }, { "epoch": 0.76, "learning_rate": 2.8066019966134904e-05, "loss": 0.6649, "step": 13449 }, { "epoch": 0.76, "learning_rate": 2.8053258995196353e-05, "loss": 0.6954, "step": 13450 }, { "epoch": 0.76, "learning_rate": 2.8040500452649e-05, "loss": 0.7155, "step": 13451 }, { "epoch": 0.76, "learning_rate": 2.8027744338923466e-05, "loss": 0.7566, "step": 13452 }, { "epoch": 0.76, "learning_rate": 2.8014990654450325e-05, "loss": 0.6085, "step": 13453 }, { "epoch": 0.76, "learning_rate": 2.800223939966007e-05, "loss": 0.6212, "step": 13454 }, { "epoch": 0.76, "learning_rate": 2.7989490574983047e-05, "loss": 0.6685, "step": 13455 }, { "epoch": 0.76, "learning_rate": 2.7976744180849547e-05, "loss": 0.73, "step": 13456 }, { "epoch": 0.76, "learning_rate": 2.7964000217689824e-05, "loss": 0.6854, "step": 13457 }, { "epoch": 0.76, "learning_rate": 2.7951258685934002e-05, "loss": 0.6338, "step": 13458 }, { "epoch": 0.76, "learning_rate": 2.7938519586012157e-05, "loss": 0.6883, "step": 13459 }, { "epoch": 0.76, "learning_rate": 2.7925782918354304e-05, "loss": 0.6048, "step": 13460 }, { "epoch": 0.76, "learning_rate": 2.7913048683390264e-05, "loss": 0.6658, "step": 13461 }, { "epoch": 0.76, "learning_rate": 2.7900316881549914e-05, "loss": 0.7581, "step": 13462 }, { "epoch": 0.76, "learning_rate": 2.7887587513262926e-05, "loss": 0.7189, "step": 13463 }, { "epoch": 0.76, "learning_rate": 2.7874860578958983e-05, "loss": 0.7618, "step": 13464 }, { "epoch": 0.76, "learning_rate": 2.7862136079067646e-05, "loss": 0.6468, "step": 13465 }, { "epoch": 0.76, "learning_rate": 2.7849414014018415e-05, "loss": 0.7376, "step": 13466 }, { "epoch": 0.76, "learning_rate": 2.7836694384240702e-05, "loss": 0.6694, "step": 13467 }, { "epoch": 0.76, "learning_rate": 2.7823977190163786e-05, "loss": 0.7139, "step": 13468 }, { "epoch": 0.76, "learning_rate": 2.7811262432216954e-05, "loss": 0.6494, "step": 13469 }, { "epoch": 0.76, "learning_rate": 2.7798550110829313e-05, "loss": 0.6634, "step": 13470 }, { "epoch": 0.76, "learning_rate": 2.778584022642996e-05, "loss": 0.6919, "step": 13471 }, { "epoch": 0.76, "learning_rate": 2.7773132779447884e-05, "loss": 0.6656, "step": 13472 }, { "epoch": 0.76, "learning_rate": 2.7760427770312003e-05, "loss": 0.7166, "step": 13473 }, { "epoch": 0.76, "learning_rate": 2.7747725199451157e-05, "loss": 0.6913, "step": 13474 }, { "epoch": 0.76, "learning_rate": 2.7735025067294064e-05, "loss": 0.6729, "step": 13475 }, { "epoch": 0.76, "learning_rate": 2.7722327374269386e-05, "loss": 0.6822, "step": 13476 }, { "epoch": 0.76, "learning_rate": 2.770963212080574e-05, "loss": 0.7013, "step": 13477 }, { "epoch": 0.76, "learning_rate": 2.7696939307331572e-05, "loss": 0.6582, "step": 13478 }, { "epoch": 0.76, "learning_rate": 2.7684248934275325e-05, "loss": 0.5702, "step": 13479 }, { "epoch": 0.76, "learning_rate": 2.767156100206536e-05, "loss": 0.7778, "step": 13480 }, { "epoch": 0.76, "learning_rate": 2.7658875511129877e-05, "loss": 0.6452, "step": 13481 }, { "epoch": 0.76, "learning_rate": 2.764619246189706e-05, "loss": 0.6603, "step": 13482 }, { "epoch": 0.76, "learning_rate": 2.7633511854795004e-05, "loss": 0.7233, "step": 13483 }, { "epoch": 0.76, "learning_rate": 2.7620833690251745e-05, "loss": 0.6075, "step": 13484 }, { "epoch": 0.76, "learning_rate": 2.7608157968695136e-05, "loss": 0.6818, "step": 13485 }, { "epoch": 0.76, "learning_rate": 2.7595484690553043e-05, "loss": 0.7253, "step": 13486 }, { "epoch": 0.77, "learning_rate": 2.7582813856253275e-05, "loss": 0.6624, "step": 13487 }, { "epoch": 0.77, "learning_rate": 2.757014546622342e-05, "loss": 0.6793, "step": 13488 }, { "epoch": 0.77, "learning_rate": 2.7557479520891104e-05, "loss": 0.66, "step": 13489 }, { "epoch": 0.77, "learning_rate": 2.754481602068385e-05, "loss": 0.7442, "step": 13490 }, { "epoch": 0.77, "learning_rate": 2.7532154966029077e-05, "loss": 0.6658, "step": 13491 }, { "epoch": 0.77, "learning_rate": 2.7519496357354156e-05, "loss": 0.743, "step": 13492 }, { "epoch": 0.77, "learning_rate": 2.7506840195086326e-05, "loss": 0.731, "step": 13493 }, { "epoch": 0.77, "learning_rate": 2.749418647965272e-05, "loss": 0.7357, "step": 13494 }, { "epoch": 0.77, "learning_rate": 2.748153521148048e-05, "loss": 0.6712, "step": 13495 }, { "epoch": 0.77, "learning_rate": 2.7468886390996608e-05, "loss": 0.7787, "step": 13496 }, { "epoch": 0.77, "learning_rate": 2.7456240018628043e-05, "loss": 0.6855, "step": 13497 }, { "epoch": 0.77, "learning_rate": 2.7443596094801627e-05, "loss": 0.7369, "step": 13498 }, { "epoch": 0.77, "learning_rate": 2.7430954619944173e-05, "loss": 0.7068, "step": 13499 }, { "epoch": 0.77, "learning_rate": 2.74183155944823e-05, "loss": 0.6988, "step": 13500 }, { "epoch": 0.77, "learning_rate": 2.740567901884261e-05, "loss": 0.6945, "step": 13501 }, { "epoch": 0.77, "learning_rate": 2.7393044893451637e-05, "loss": 0.7065, "step": 13502 }, { "epoch": 0.77, "learning_rate": 2.738041321873581e-05, "loss": 0.7048, "step": 13503 }, { "epoch": 0.77, "learning_rate": 2.73677839951215e-05, "loss": 0.7147, "step": 13504 }, { "epoch": 0.77, "learning_rate": 2.7355157223034967e-05, "loss": 0.6896, "step": 13505 }, { "epoch": 0.77, "learning_rate": 2.734253290290242e-05, "loss": 0.7334, "step": 13506 }, { "epoch": 0.77, "learning_rate": 2.7329911035149937e-05, "loss": 0.6995, "step": 13507 }, { "epoch": 0.77, "learning_rate": 2.731729162020351e-05, "loss": 0.6416, "step": 13508 }, { "epoch": 0.77, "learning_rate": 2.7304674658489104e-05, "loss": 0.7345, "step": 13509 }, { "epoch": 0.77, "learning_rate": 2.7292060150432587e-05, "loss": 0.6845, "step": 13510 }, { "epoch": 0.77, "learning_rate": 2.7279448096459725e-05, "loss": 0.7666, "step": 13511 }, { "epoch": 0.77, "learning_rate": 2.7266838496996228e-05, "loss": 0.6496, "step": 13512 }, { "epoch": 0.77, "learning_rate": 2.7254231352467664e-05, "loss": 0.7657, "step": 13513 }, { "epoch": 0.77, "learning_rate": 2.724162666329959e-05, "loss": 0.6077, "step": 13514 }, { "epoch": 0.77, "learning_rate": 2.7229024429917405e-05, "loss": 0.7365, "step": 13515 }, { "epoch": 0.77, "learning_rate": 2.7216424652746498e-05, "loss": 0.6285, "step": 13516 }, { "epoch": 0.77, "learning_rate": 2.720382733221214e-05, "loss": 0.7265, "step": 13517 }, { "epoch": 0.77, "learning_rate": 2.7191232468739524e-05, "loss": 0.7097, "step": 13518 }, { "epoch": 0.77, "learning_rate": 2.7178640062753802e-05, "loss": 0.6879, "step": 13519 }, { "epoch": 0.77, "learning_rate": 2.716605011467992e-05, "loss": 0.6785, "step": 13520 }, { "epoch": 0.77, "learning_rate": 2.7153462624942883e-05, "loss": 0.5946, "step": 13521 }, { "epoch": 0.77, "learning_rate": 2.7140877593967506e-05, "loss": 0.7033, "step": 13522 }, { "epoch": 0.77, "learning_rate": 2.7128295022178584e-05, "loss": 0.6641, "step": 13523 }, { "epoch": 0.77, "learning_rate": 2.7115714910000835e-05, "loss": 0.683, "step": 13524 }, { "epoch": 0.77, "learning_rate": 2.7103137257858868e-05, "loss": 0.6103, "step": 13525 }, { "epoch": 0.77, "learning_rate": 2.7090562066177174e-05, "loss": 0.8003, "step": 13526 }, { "epoch": 0.77, "learning_rate": 2.7077989335380216e-05, "loss": 0.6984, "step": 13527 }, { "epoch": 0.77, "learning_rate": 2.7065419065892372e-05, "loss": 0.6443, "step": 13528 }, { "epoch": 0.77, "learning_rate": 2.7052851258137935e-05, "loss": 0.7206, "step": 13529 }, { "epoch": 0.77, "learning_rate": 2.7040285912541054e-05, "loss": 0.6778, "step": 13530 }, { "epoch": 0.77, "learning_rate": 2.7027723029525866e-05, "loss": 0.7118, "step": 13531 }, { "epoch": 0.77, "learning_rate": 2.701516260951643e-05, "loss": 0.673, "step": 13532 }, { "epoch": 0.77, "learning_rate": 2.7002604652936637e-05, "loss": 0.7088, "step": 13533 }, { "epoch": 0.77, "learning_rate": 2.6990049160210385e-05, "loss": 0.7043, "step": 13534 }, { "epoch": 0.77, "learning_rate": 2.6977496131761436e-05, "loss": 0.7525, "step": 13535 }, { "epoch": 0.77, "learning_rate": 2.6964945568013545e-05, "loss": 0.7107, "step": 13536 }, { "epoch": 0.77, "learning_rate": 2.6952397469390245e-05, "loss": 0.717, "step": 13537 }, { "epoch": 0.77, "learning_rate": 2.6939851836315133e-05, "loss": 0.6716, "step": 13538 }, { "epoch": 0.77, "learning_rate": 2.6927308669211605e-05, "loss": 0.7339, "step": 13539 }, { "epoch": 0.77, "learning_rate": 2.691476796850305e-05, "loss": 0.7564, "step": 13540 }, { "epoch": 0.77, "learning_rate": 2.690222973461275e-05, "loss": 0.689, "step": 13541 }, { "epoch": 0.77, "learning_rate": 2.688969396796389e-05, "loss": 0.605, "step": 13542 }, { "epoch": 0.77, "learning_rate": 2.687716066897964e-05, "loss": 0.6615, "step": 13543 }, { "epoch": 0.77, "learning_rate": 2.6864629838082956e-05, "loss": 0.7468, "step": 13544 }, { "epoch": 0.77, "learning_rate": 2.6852101475696843e-05, "loss": 0.716, "step": 13545 }, { "epoch": 0.77, "learning_rate": 2.6839575582244124e-05, "loss": 0.6798, "step": 13546 }, { "epoch": 0.77, "learning_rate": 2.6827052158147592e-05, "loss": 0.6648, "step": 13547 }, { "epoch": 0.77, "learning_rate": 2.681453120382995e-05, "loss": 0.7007, "step": 13548 }, { "epoch": 0.77, "learning_rate": 2.680201271971383e-05, "loss": 0.6583, "step": 13549 }, { "epoch": 0.77, "learning_rate": 2.678949670622174e-05, "loss": 0.6929, "step": 13550 }, { "epoch": 0.77, "learning_rate": 2.677698316377616e-05, "loss": 0.6688, "step": 13551 }, { "epoch": 0.77, "learning_rate": 2.6764472092799443e-05, "loss": 0.699, "step": 13552 }, { "epoch": 0.77, "learning_rate": 2.6751963493713827e-05, "loss": 0.799, "step": 13553 }, { "epoch": 0.77, "learning_rate": 2.6739457366941543e-05, "loss": 0.6224, "step": 13554 }, { "epoch": 0.77, "learning_rate": 2.6726953712904712e-05, "loss": 0.6585, "step": 13555 }, { "epoch": 0.77, "learning_rate": 2.6714452532025347e-05, "loss": 0.7145, "step": 13556 }, { "epoch": 0.77, "learning_rate": 2.6701953824725444e-05, "loss": 0.6807, "step": 13557 }, { "epoch": 0.77, "learning_rate": 2.6689457591426825e-05, "loss": 0.6994, "step": 13558 }, { "epoch": 0.77, "learning_rate": 2.667696383255126e-05, "loss": 0.7596, "step": 13559 }, { "epoch": 0.77, "learning_rate": 2.6664472548520446e-05, "loss": 0.7793, "step": 13560 }, { "epoch": 0.77, "learning_rate": 2.6651983739756026e-05, "loss": 0.7625, "step": 13561 }, { "epoch": 0.77, "learning_rate": 2.6639497406679505e-05, "loss": 0.7935, "step": 13562 }, { "epoch": 0.77, "learning_rate": 2.6627013549712355e-05, "loss": 0.7295, "step": 13563 }, { "epoch": 0.77, "learning_rate": 2.661453216927594e-05, "loss": 0.7472, "step": 13564 }, { "epoch": 0.77, "learning_rate": 2.6602053265791504e-05, "loss": 0.8022, "step": 13565 }, { "epoch": 0.77, "learning_rate": 2.6589576839680296e-05, "loss": 0.7095, "step": 13566 }, { "epoch": 0.77, "learning_rate": 2.657710289136336e-05, "loss": 0.6304, "step": 13567 }, { "epoch": 0.77, "learning_rate": 2.6564631421261765e-05, "loss": 0.7381, "step": 13568 }, { "epoch": 0.77, "learning_rate": 2.6552162429796446e-05, "loss": 0.7159, "step": 13569 }, { "epoch": 0.77, "learning_rate": 2.6539695917388296e-05, "loss": 0.7198, "step": 13570 }, { "epoch": 0.77, "learning_rate": 2.6527231884458048e-05, "loss": 0.8116, "step": 13571 }, { "epoch": 0.77, "learning_rate": 2.6514770331426408e-05, "loss": 0.7194, "step": 13572 }, { "epoch": 0.77, "learning_rate": 2.650231125871402e-05, "loss": 0.6534, "step": 13573 }, { "epoch": 0.77, "learning_rate": 2.6489854666741343e-05, "loss": 0.6696, "step": 13574 }, { "epoch": 0.77, "learning_rate": 2.6477400555928867e-05, "loss": 0.6634, "step": 13575 }, { "epoch": 0.77, "learning_rate": 2.646494892669694e-05, "loss": 0.7073, "step": 13576 }, { "epoch": 0.77, "learning_rate": 2.6452499779465876e-05, "loss": 0.7447, "step": 13577 }, { "epoch": 0.77, "learning_rate": 2.6440053114655795e-05, "loss": 0.7306, "step": 13578 }, { "epoch": 0.77, "learning_rate": 2.6427608932686843e-05, "loss": 0.7869, "step": 13579 }, { "epoch": 0.77, "learning_rate": 2.6415167233979066e-05, "loss": 0.6691, "step": 13580 }, { "epoch": 0.77, "learning_rate": 2.640272801895234e-05, "loss": 0.727, "step": 13581 }, { "epoch": 0.77, "learning_rate": 2.639029128802657e-05, "loss": 0.7721, "step": 13582 }, { "epoch": 0.77, "learning_rate": 2.637785704162151e-05, "loss": 0.6893, "step": 13583 }, { "epoch": 0.77, "learning_rate": 2.6365425280156898e-05, "loss": 0.7528, "step": 13584 }, { "epoch": 0.77, "learning_rate": 2.6352996004052255e-05, "loss": 0.7361, "step": 13585 }, { "epoch": 0.77, "learning_rate": 2.634056921372715e-05, "loss": 0.702, "step": 13586 }, { "epoch": 0.77, "learning_rate": 2.6328144909601014e-05, "loss": 0.7659, "step": 13587 }, { "epoch": 0.77, "learning_rate": 2.6315723092093226e-05, "loss": 0.5798, "step": 13588 }, { "epoch": 0.77, "learning_rate": 2.6303303761623e-05, "loss": 0.6524, "step": 13589 }, { "epoch": 0.77, "learning_rate": 2.629088691860957e-05, "loss": 0.7262, "step": 13590 }, { "epoch": 0.77, "learning_rate": 2.6278472563472e-05, "loss": 0.7474, "step": 13591 }, { "epoch": 0.77, "learning_rate": 2.6266060696629315e-05, "loss": 0.6119, "step": 13592 }, { "epoch": 0.77, "learning_rate": 2.625365131850045e-05, "loss": 0.741, "step": 13593 }, { "epoch": 0.77, "learning_rate": 2.6241244429504264e-05, "loss": 0.7127, "step": 13594 }, { "epoch": 0.77, "learning_rate": 2.6228840030059554e-05, "loss": 0.6754, "step": 13595 }, { "epoch": 0.77, "learning_rate": 2.621643812058493e-05, "loss": 0.7203, "step": 13596 }, { "epoch": 0.77, "learning_rate": 2.6204038701499056e-05, "loss": 0.6951, "step": 13597 }, { "epoch": 0.77, "learning_rate": 2.6191641773220376e-05, "loss": 0.7677, "step": 13598 }, { "epoch": 0.77, "learning_rate": 2.6179247336167367e-05, "loss": 0.7389, "step": 13599 }, { "epoch": 0.77, "learning_rate": 2.616685539075835e-05, "loss": 0.6546, "step": 13600 }, { "epoch": 0.77, "learning_rate": 2.615446593741161e-05, "loss": 0.6262, "step": 13601 }, { "epoch": 0.77, "learning_rate": 2.6142078976545337e-05, "loss": 0.6955, "step": 13602 }, { "epoch": 0.77, "learning_rate": 2.6129694508577594e-05, "loss": 0.6757, "step": 13603 }, { "epoch": 0.77, "learning_rate": 2.6117312533926362e-05, "loss": 0.7674, "step": 13604 }, { "epoch": 0.77, "learning_rate": 2.6104933053009607e-05, "loss": 0.6224, "step": 13605 }, { "epoch": 0.77, "learning_rate": 2.6092556066245154e-05, "loss": 0.7771, "step": 13606 }, { "epoch": 0.77, "learning_rate": 2.6080181574050755e-05, "loss": 0.6845, "step": 13607 }, { "epoch": 0.77, "learning_rate": 2.6067809576844093e-05, "loss": 0.6322, "step": 13608 }, { "epoch": 0.77, "learning_rate": 2.6055440075042793e-05, "loss": 0.733, "step": 13609 }, { "epoch": 0.77, "learning_rate": 2.6043073069064306e-05, "loss": 0.7428, "step": 13610 }, { "epoch": 0.77, "learning_rate": 2.6030708559326033e-05, "loss": 0.7926, "step": 13611 }, { "epoch": 0.77, "learning_rate": 2.601834654624534e-05, "loss": 0.7218, "step": 13612 }, { "epoch": 0.77, "learning_rate": 2.6005987030239475e-05, "loss": 0.6705, "step": 13613 }, { "epoch": 0.77, "learning_rate": 2.5993630011725613e-05, "loss": 0.7607, "step": 13614 }, { "epoch": 0.77, "learning_rate": 2.598127549112084e-05, "loss": 0.632, "step": 13615 }, { "epoch": 0.77, "learning_rate": 2.5968923468842122e-05, "loss": 0.6779, "step": 13616 }, { "epoch": 0.77, "learning_rate": 2.5956573945306385e-05, "loss": 0.7205, "step": 13617 }, { "epoch": 0.77, "learning_rate": 2.5944226920930493e-05, "loss": 0.6652, "step": 13618 }, { "epoch": 0.77, "learning_rate": 2.5931882396131135e-05, "loss": 0.6648, "step": 13619 }, { "epoch": 0.77, "learning_rate": 2.5919540371325e-05, "loss": 0.7422, "step": 13620 }, { "epoch": 0.77, "learning_rate": 2.590720084692865e-05, "loss": 0.6709, "step": 13621 }, { "epoch": 0.77, "learning_rate": 2.5894863823358628e-05, "loss": 0.652, "step": 13622 }, { "epoch": 0.77, "learning_rate": 2.5882529301031265e-05, "loss": 0.6997, "step": 13623 }, { "epoch": 0.77, "learning_rate": 2.587019728036292e-05, "loss": 0.7138, "step": 13624 }, { "epoch": 0.77, "learning_rate": 2.585786776176985e-05, "loss": 0.6465, "step": 13625 }, { "epoch": 0.77, "learning_rate": 2.5845540745668163e-05, "loss": 0.6608, "step": 13626 }, { "epoch": 0.77, "learning_rate": 2.5833216232473955e-05, "loss": 0.6925, "step": 13627 }, { "epoch": 0.77, "learning_rate": 2.582089422260321e-05, "loss": 0.7685, "step": 13628 }, { "epoch": 0.77, "learning_rate": 2.5808574716471856e-05, "loss": 0.7221, "step": 13629 }, { "epoch": 0.77, "learning_rate": 2.5796257714495654e-05, "loss": 0.7188, "step": 13630 }, { "epoch": 0.77, "learning_rate": 2.5783943217090357e-05, "loss": 0.7466, "step": 13631 }, { "epoch": 0.77, "learning_rate": 2.5771631224671644e-05, "loss": 0.6713, "step": 13632 }, { "epoch": 0.77, "learning_rate": 2.5759321737655017e-05, "loss": 0.6058, "step": 13633 }, { "epoch": 0.77, "learning_rate": 2.5747014756455978e-05, "loss": 0.6719, "step": 13634 }, { "epoch": 0.77, "learning_rate": 2.573471028148997e-05, "loss": 0.7558, "step": 13635 }, { "epoch": 0.77, "learning_rate": 2.572240831317222e-05, "loss": 0.7534, "step": 13636 }, { "epoch": 0.77, "learning_rate": 2.571010885191799e-05, "loss": 0.7229, "step": 13637 }, { "epoch": 0.77, "learning_rate": 2.569781189814242e-05, "loss": 0.7821, "step": 13638 }, { "epoch": 0.77, "learning_rate": 2.5685517452260567e-05, "loss": 0.6639, "step": 13639 }, { "epoch": 0.77, "learning_rate": 2.5673225514687416e-05, "loss": 0.7423, "step": 13640 }, { "epoch": 0.77, "learning_rate": 2.5660936085837816e-05, "loss": 0.5861, "step": 13641 }, { "epoch": 0.77, "learning_rate": 2.5648649166126605e-05, "loss": 0.6385, "step": 13642 }, { "epoch": 0.77, "learning_rate": 2.5636364755968458e-05, "loss": 0.7062, "step": 13643 }, { "epoch": 0.77, "learning_rate": 2.562408285577802e-05, "loss": 0.5576, "step": 13644 }, { "epoch": 0.77, "learning_rate": 2.5611803465969842e-05, "loss": 0.7238, "step": 13645 }, { "epoch": 0.77, "learning_rate": 2.5599526586958388e-05, "loss": 0.6935, "step": 13646 }, { "epoch": 0.77, "learning_rate": 2.5587252219158066e-05, "loss": 0.7099, "step": 13647 }, { "epoch": 0.77, "learning_rate": 2.5574980362983136e-05, "loss": 0.7352, "step": 13648 }, { "epoch": 0.77, "learning_rate": 2.5562711018847772e-05, "loss": 0.7144, "step": 13649 }, { "epoch": 0.77, "learning_rate": 2.5550444187166135e-05, "loss": 0.6461, "step": 13650 }, { "epoch": 0.77, "learning_rate": 2.5538179868352253e-05, "loss": 0.712, "step": 13651 }, { "epoch": 0.77, "learning_rate": 2.552591806282009e-05, "loss": 0.7034, "step": 13652 }, { "epoch": 0.77, "learning_rate": 2.5513658770983495e-05, "loss": 0.6412, "step": 13653 }, { "epoch": 0.77, "learning_rate": 2.55014019932563e-05, "loss": 0.6837, "step": 13654 }, { "epoch": 0.77, "learning_rate": 2.5489147730052176e-05, "loss": 0.7287, "step": 13655 }, { "epoch": 0.77, "learning_rate": 2.5476895981784698e-05, "loss": 0.7289, "step": 13656 }, { "epoch": 0.77, "learning_rate": 2.5464646748867415e-05, "loss": 0.7439, "step": 13657 }, { "epoch": 0.77, "learning_rate": 2.5452400031713785e-05, "loss": 0.6754, "step": 13658 }, { "epoch": 0.77, "learning_rate": 2.5440155830737156e-05, "loss": 0.6467, "step": 13659 }, { "epoch": 0.77, "learning_rate": 2.542791414635085e-05, "loss": 0.6174, "step": 13660 }, { "epoch": 0.77, "learning_rate": 2.5415674978967975e-05, "loss": 0.6456, "step": 13661 }, { "epoch": 0.77, "learning_rate": 2.5403438329001707e-05, "loss": 0.7292, "step": 13662 }, { "epoch": 0.78, "learning_rate": 2.5391204196865005e-05, "loss": 0.7529, "step": 13663 }, { "epoch": 0.78, "learning_rate": 2.5378972582970828e-05, "loss": 0.6806, "step": 13664 }, { "epoch": 0.78, "learning_rate": 2.5366743487732027e-05, "loss": 0.6462, "step": 13665 }, { "epoch": 0.78, "learning_rate": 2.5354516911561367e-05, "loss": 0.6808, "step": 13666 }, { "epoch": 0.78, "learning_rate": 2.534229285487155e-05, "loss": 0.7013, "step": 13667 }, { "epoch": 0.78, "learning_rate": 2.5330071318075123e-05, "loss": 0.6785, "step": 13668 }, { "epoch": 0.78, "learning_rate": 2.5317852301584643e-05, "loss": 0.706, "step": 13669 }, { "epoch": 0.78, "learning_rate": 2.5305635805812478e-05, "loss": 0.5648, "step": 13670 }, { "epoch": 0.78, "learning_rate": 2.5293421831171003e-05, "loss": 0.6948, "step": 13671 }, { "epoch": 0.78, "learning_rate": 2.5281210378072463e-05, "loss": 0.6916, "step": 13672 }, { "epoch": 0.78, "learning_rate": 2.526900144692903e-05, "loss": 0.7057, "step": 13673 }, { "epoch": 0.78, "learning_rate": 2.5256795038152825e-05, "loss": 0.5651, "step": 13674 }, { "epoch": 0.78, "learning_rate": 2.524459115215577e-05, "loss": 0.7437, "step": 13675 }, { "epoch": 0.78, "learning_rate": 2.523238978934982e-05, "loss": 0.7423, "step": 13676 }, { "epoch": 0.78, "learning_rate": 2.5220190950146827e-05, "loss": 0.7144, "step": 13677 }, { "epoch": 0.78, "learning_rate": 2.5207994634958476e-05, "loss": 0.5956, "step": 13678 }, { "epoch": 0.78, "learning_rate": 2.519580084419646e-05, "loss": 0.7088, "step": 13679 }, { "epoch": 0.78, "learning_rate": 2.5183609578272383e-05, "loss": 0.714, "step": 13680 }, { "epoch": 0.78, "learning_rate": 2.5171420837597646e-05, "loss": 0.7079, "step": 13681 }, { "epoch": 0.78, "learning_rate": 2.5159234622583715e-05, "loss": 0.7488, "step": 13682 }, { "epoch": 0.78, "learning_rate": 2.514705093364189e-05, "loss": 0.6422, "step": 13683 }, { "epoch": 0.78, "learning_rate": 2.5134869771183445e-05, "loss": 0.7742, "step": 13684 }, { "epoch": 0.78, "learning_rate": 2.5122691135619446e-05, "loss": 0.79, "step": 13685 }, { "epoch": 0.78, "learning_rate": 2.511051502736099e-05, "loss": 0.6567, "step": 13686 }, { "epoch": 0.78, "learning_rate": 2.5098341446819097e-05, "loss": 0.7136, "step": 13687 }, { "epoch": 0.78, "learning_rate": 2.5086170394404573e-05, "loss": 0.6728, "step": 13688 }, { "epoch": 0.78, "learning_rate": 2.5074001870528286e-05, "loss": 0.752, "step": 13689 }, { "epoch": 0.78, "learning_rate": 2.5061835875600926e-05, "loss": 0.7433, "step": 13690 }, { "epoch": 0.78, "learning_rate": 2.504967241003313e-05, "loss": 0.7838, "step": 13691 }, { "epoch": 0.78, "learning_rate": 2.5037511474235498e-05, "loss": 0.7299, "step": 13692 }, { "epoch": 0.78, "learning_rate": 2.502535306861844e-05, "loss": 0.6753, "step": 13693 }, { "epoch": 0.78, "learning_rate": 2.501319719359232e-05, "loss": 0.7082, "step": 13694 }, { "epoch": 0.78, "learning_rate": 2.500104384956745e-05, "loss": 0.7177, "step": 13695 }, { "epoch": 0.78, "learning_rate": 2.4988893036954043e-05, "loss": 0.7101, "step": 13696 }, { "epoch": 0.78, "learning_rate": 2.497674475616223e-05, "loss": 0.6354, "step": 13697 }, { "epoch": 0.78, "learning_rate": 2.496459900760203e-05, "loss": 0.6883, "step": 13698 }, { "epoch": 0.78, "learning_rate": 2.4952455791683426e-05, "loss": 0.7939, "step": 13699 }, { "epoch": 0.78, "learning_rate": 2.494031510881626e-05, "loss": 0.6325, "step": 13700 }, { "epoch": 0.78, "learning_rate": 2.4928176959410286e-05, "loss": 0.6946, "step": 13701 }, { "epoch": 0.78, "learning_rate": 2.491604134387522e-05, "loss": 0.7242, "step": 13702 }, { "epoch": 0.78, "learning_rate": 2.4903908262620677e-05, "loss": 0.747, "step": 13703 }, { "epoch": 0.78, "learning_rate": 2.4891777716056176e-05, "loss": 0.8311, "step": 13704 }, { "epoch": 0.78, "learning_rate": 2.487964970459118e-05, "loss": 0.7386, "step": 13705 }, { "epoch": 0.78, "learning_rate": 2.4867524228634986e-05, "loss": 0.5716, "step": 13706 }, { "epoch": 0.78, "learning_rate": 2.4855401288596924e-05, "loss": 0.6452, "step": 13707 }, { "epoch": 0.78, "learning_rate": 2.484328088488611e-05, "loss": 0.7117, "step": 13708 }, { "epoch": 0.78, "learning_rate": 2.4831163017911686e-05, "loss": 0.6461, "step": 13709 }, { "epoch": 0.78, "learning_rate": 2.481904768808263e-05, "loss": 0.7493, "step": 13710 }, { "epoch": 0.78, "learning_rate": 2.4806934895807887e-05, "loss": 0.6602, "step": 13711 }, { "epoch": 0.78, "learning_rate": 2.4794824641496328e-05, "loss": 0.8394, "step": 13712 }, { "epoch": 0.78, "learning_rate": 2.4782716925556638e-05, "loss": 0.7279, "step": 13713 }, { "epoch": 0.78, "learning_rate": 2.4770611748397553e-05, "loss": 0.7426, "step": 13714 }, { "epoch": 0.78, "learning_rate": 2.4758509110427575e-05, "loss": 0.632, "step": 13715 }, { "epoch": 0.78, "learning_rate": 2.474640901205525e-05, "loss": 0.658, "step": 13716 }, { "epoch": 0.78, "learning_rate": 2.4734311453688985e-05, "loss": 0.6786, "step": 13717 }, { "epoch": 0.78, "learning_rate": 2.4722216435737088e-05, "loss": 0.6311, "step": 13718 }, { "epoch": 0.78, "learning_rate": 2.4710123958607846e-05, "loss": 0.717, "step": 13719 }, { "epoch": 0.78, "learning_rate": 2.469803402270935e-05, "loss": 0.6306, "step": 13720 }, { "epoch": 0.78, "learning_rate": 2.4685946628449718e-05, "loss": 0.7276, "step": 13721 }, { "epoch": 0.78, "learning_rate": 2.4673861776236874e-05, "loss": 0.6296, "step": 13722 }, { "epoch": 0.78, "learning_rate": 2.466177946647874e-05, "loss": 0.6911, "step": 13723 }, { "epoch": 0.78, "learning_rate": 2.464969969958314e-05, "loss": 0.5925, "step": 13724 }, { "epoch": 0.78, "learning_rate": 2.463762247595782e-05, "loss": 0.7272, "step": 13725 }, { "epoch": 0.78, "learning_rate": 2.462554779601035e-05, "loss": 0.7199, "step": 13726 }, { "epoch": 0.78, "learning_rate": 2.4613475660148323e-05, "loss": 0.7416, "step": 13727 }, { "epoch": 0.78, "learning_rate": 2.46014060687792e-05, "loss": 0.6709, "step": 13728 }, { "epoch": 0.78, "learning_rate": 2.4589339022310386e-05, "loss": 0.6947, "step": 13729 }, { "epoch": 0.78, "learning_rate": 2.4577274521149128e-05, "loss": 0.659, "step": 13730 }, { "epoch": 0.78, "learning_rate": 2.4565212565702656e-05, "loss": 0.676, "step": 13731 }, { "epoch": 0.78, "learning_rate": 2.4553153156378127e-05, "loss": 0.6393, "step": 13732 }, { "epoch": 0.78, "learning_rate": 2.454109629358251e-05, "loss": 0.7945, "step": 13733 }, { "epoch": 0.78, "learning_rate": 2.45290419777228e-05, "loss": 0.642, "step": 13734 }, { "epoch": 0.78, "learning_rate": 2.4516990209205847e-05, "loss": 0.7321, "step": 13735 }, { "epoch": 0.78, "learning_rate": 2.4504940988438473e-05, "loss": 0.6174, "step": 13736 }, { "epoch": 0.78, "learning_rate": 2.4492894315827297e-05, "loss": 0.762, "step": 13737 }, { "epoch": 0.78, "learning_rate": 2.448085019177899e-05, "loss": 0.7464, "step": 13738 }, { "epoch": 0.78, "learning_rate": 2.4468808616700024e-05, "loss": 0.6464, "step": 13739 }, { "epoch": 0.78, "learning_rate": 2.4456769590996854e-05, "loss": 0.7115, "step": 13740 }, { "epoch": 0.78, "learning_rate": 2.4444733115075823e-05, "loss": 0.7468, "step": 13741 }, { "epoch": 0.78, "learning_rate": 2.44326991893432e-05, "loss": 0.7848, "step": 13742 }, { "epoch": 0.78, "learning_rate": 2.442066781420519e-05, "loss": 0.6803, "step": 13743 }, { "epoch": 0.78, "learning_rate": 2.4408638990067833e-05, "loss": 0.6665, "step": 13744 }, { "epoch": 0.78, "learning_rate": 2.4396612717337186e-05, "loss": 0.7119, "step": 13745 }, { "epoch": 0.78, "learning_rate": 2.43845889964191e-05, "loss": 0.6974, "step": 13746 }, { "epoch": 0.78, "learning_rate": 2.437256782771945e-05, "loss": 0.6414, "step": 13747 }, { "epoch": 0.78, "learning_rate": 2.4360549211643978e-05, "loss": 0.6938, "step": 13748 }, { "epoch": 0.78, "learning_rate": 2.434853314859834e-05, "loss": 0.8114, "step": 13749 }, { "epoch": 0.78, "learning_rate": 2.4336519638988132e-05, "loss": 0.7281, "step": 13750 }, { "epoch": 0.78, "learning_rate": 2.4324508683218804e-05, "loss": 0.6594, "step": 13751 }, { "epoch": 0.78, "learning_rate": 2.4312500281695794e-05, "loss": 0.7115, "step": 13752 }, { "epoch": 0.78, "learning_rate": 2.4300494434824373e-05, "loss": 0.7501, "step": 13753 }, { "epoch": 0.78, "learning_rate": 2.4288491143009795e-05, "loss": 0.8145, "step": 13754 }, { "epoch": 0.78, "learning_rate": 2.4276490406657194e-05, "loss": 0.6329, "step": 13755 }, { "epoch": 0.78, "learning_rate": 2.4264492226171644e-05, "loss": 0.6639, "step": 13756 }, { "epoch": 0.78, "learning_rate": 2.425249660195813e-05, "loss": 0.6911, "step": 13757 }, { "epoch": 0.78, "learning_rate": 2.424050353442151e-05, "loss": 0.7618, "step": 13758 }, { "epoch": 0.78, "learning_rate": 2.422851302396655e-05, "loss": 0.712, "step": 13759 }, { "epoch": 0.78, "learning_rate": 2.4216525070997996e-05, "loss": 0.7677, "step": 13760 }, { "epoch": 0.78, "learning_rate": 2.420453967592048e-05, "loss": 0.5935, "step": 13761 }, { "epoch": 0.78, "learning_rate": 2.4192556839138515e-05, "loss": 0.6976, "step": 13762 }, { "epoch": 0.78, "learning_rate": 2.4180576561056577e-05, "loss": 0.7759, "step": 13763 }, { "epoch": 0.78, "learning_rate": 2.416859884207905e-05, "loss": 0.6347, "step": 13764 }, { "epoch": 0.78, "learning_rate": 2.415662368261017e-05, "loss": 0.7857, "step": 13765 }, { "epoch": 0.78, "learning_rate": 2.414465108305416e-05, "loss": 0.8092, "step": 13766 }, { "epoch": 0.78, "learning_rate": 2.4132681043815098e-05, "loss": 0.7751, "step": 13767 }, { "epoch": 0.78, "learning_rate": 2.4120713565297005e-05, "loss": 0.7533, "step": 13768 }, { "epoch": 0.78, "learning_rate": 2.4108748647903846e-05, "loss": 0.8418, "step": 13769 }, { "epoch": 0.78, "learning_rate": 2.4096786292039465e-05, "loss": 0.8028, "step": 13770 }, { "epoch": 0.78, "learning_rate": 2.4084826498107592e-05, "loss": 0.6179, "step": 13771 }, { "epoch": 0.78, "learning_rate": 2.407286926651192e-05, "loss": 0.738, "step": 13772 }, { "epoch": 0.78, "learning_rate": 2.406091459765606e-05, "loss": 0.749, "step": 13773 }, { "epoch": 0.78, "learning_rate": 2.404896249194346e-05, "loss": 0.666, "step": 13774 }, { "epoch": 0.78, "learning_rate": 2.403701294977757e-05, "loss": 0.6526, "step": 13775 }, { "epoch": 0.78, "learning_rate": 2.402506597156171e-05, "loss": 0.7366, "step": 13776 }, { "epoch": 0.78, "learning_rate": 2.4013121557699157e-05, "loss": 0.7769, "step": 13777 }, { "epoch": 0.78, "learning_rate": 2.4001179708593014e-05, "loss": 0.7237, "step": 13778 }, { "epoch": 0.78, "learning_rate": 2.3989240424646355e-05, "loss": 0.6357, "step": 13779 }, { "epoch": 0.78, "learning_rate": 2.3977303706262177e-05, "loss": 0.6241, "step": 13780 }, { "epoch": 0.78, "learning_rate": 2.396536955384341e-05, "loss": 0.7597, "step": 13781 }, { "epoch": 0.78, "learning_rate": 2.3953437967792803e-05, "loss": 0.6813, "step": 13782 }, { "epoch": 0.78, "learning_rate": 2.3941508948513125e-05, "loss": 0.7278, "step": 13783 }, { "epoch": 0.78, "learning_rate": 2.392958249640698e-05, "loss": 0.6565, "step": 13784 }, { "epoch": 0.78, "learning_rate": 2.3917658611876904e-05, "loss": 0.6624, "step": 13785 }, { "epoch": 0.78, "learning_rate": 2.390573729532539e-05, "loss": 0.5652, "step": 13786 }, { "epoch": 0.78, "learning_rate": 2.389381854715481e-05, "loss": 0.6799, "step": 13787 }, { "epoch": 0.78, "learning_rate": 2.3881902367767484e-05, "loss": 0.5958, "step": 13788 }, { "epoch": 0.78, "learning_rate": 2.3869988757565543e-05, "loss": 0.7615, "step": 13789 }, { "epoch": 0.78, "learning_rate": 2.385807771695118e-05, "loss": 0.6651, "step": 13790 }, { "epoch": 0.78, "learning_rate": 2.3846169246326343e-05, "loss": 0.6339, "step": 13791 }, { "epoch": 0.78, "learning_rate": 2.3834263346093012e-05, "loss": 0.6224, "step": 13792 }, { "epoch": 0.78, "learning_rate": 2.3822360016653055e-05, "loss": 0.7821, "step": 13793 }, { "epoch": 0.78, "learning_rate": 2.3810459258408224e-05, "loss": 0.659, "step": 13794 }, { "epoch": 0.78, "learning_rate": 2.3798561071760238e-05, "loss": 0.7588, "step": 13795 }, { "epoch": 0.78, "learning_rate": 2.3786665457110635e-05, "loss": 0.6912, "step": 13796 }, { "epoch": 0.78, "learning_rate": 2.377477241486098e-05, "loss": 0.7021, "step": 13797 }, { "epoch": 0.78, "learning_rate": 2.376288194541263e-05, "loss": 0.7387, "step": 13798 }, { "epoch": 0.78, "learning_rate": 2.3750994049166953e-05, "loss": 0.7081, "step": 13799 }, { "epoch": 0.78, "learning_rate": 2.3739108726525207e-05, "loss": 0.7573, "step": 13800 }, { "epoch": 0.78, "learning_rate": 2.3727225977888533e-05, "loss": 0.6799, "step": 13801 }, { "epoch": 0.78, "learning_rate": 2.371534580365804e-05, "loss": 0.6613, "step": 13802 }, { "epoch": 0.78, "learning_rate": 2.3703468204234704e-05, "loss": 0.8231, "step": 13803 }, { "epoch": 0.78, "learning_rate": 2.3691593180019366e-05, "loss": 0.7232, "step": 13804 }, { "epoch": 0.78, "learning_rate": 2.3679720731412904e-05, "loss": 0.7594, "step": 13805 }, { "epoch": 0.78, "learning_rate": 2.366785085881601e-05, "loss": 0.6902, "step": 13806 }, { "epoch": 0.78, "learning_rate": 2.3655983562629335e-05, "loss": 0.6537, "step": 13807 }, { "epoch": 0.78, "learning_rate": 2.3644118843253448e-05, "loss": 0.7743, "step": 13808 }, { "epoch": 0.78, "learning_rate": 2.363225670108882e-05, "loss": 0.668, "step": 13809 }, { "epoch": 0.78, "learning_rate": 2.362039713653581e-05, "loss": 0.6777, "step": 13810 }, { "epoch": 0.78, "learning_rate": 2.3608540149994684e-05, "loss": 0.7013, "step": 13811 }, { "epoch": 0.78, "learning_rate": 2.3596685741865665e-05, "loss": 0.6975, "step": 13812 }, { "epoch": 0.78, "learning_rate": 2.3584833912548888e-05, "loss": 0.6442, "step": 13813 }, { "epoch": 0.78, "learning_rate": 2.3572984662444365e-05, "loss": 0.6403, "step": 13814 }, { "epoch": 0.78, "learning_rate": 2.3561137991952077e-05, "loss": 0.7815, "step": 13815 }, { "epoch": 0.78, "learning_rate": 2.3549293901471825e-05, "loss": 0.7178, "step": 13816 }, { "epoch": 0.78, "learning_rate": 2.3537452391403403e-05, "loss": 0.6867, "step": 13817 }, { "epoch": 0.78, "learning_rate": 2.3525613462146522e-05, "loss": 0.7157, "step": 13818 }, { "epoch": 0.78, "learning_rate": 2.351377711410071e-05, "loss": 0.6645, "step": 13819 }, { "epoch": 0.78, "learning_rate": 2.3501943347665524e-05, "loss": 0.5971, "step": 13820 }, { "epoch": 0.78, "learning_rate": 2.3490112163240373e-05, "loss": 0.6736, "step": 13821 }, { "epoch": 0.78, "learning_rate": 2.3478283561224612e-05, "loss": 0.7102, "step": 13822 }, { "epoch": 0.78, "learning_rate": 2.346645754201744e-05, "loss": 0.6898, "step": 13823 }, { "epoch": 0.78, "learning_rate": 2.345463410601805e-05, "loss": 0.8151, "step": 13824 }, { "epoch": 0.78, "learning_rate": 2.3442813253625528e-05, "loss": 0.6637, "step": 13825 }, { "epoch": 0.78, "learning_rate": 2.3430994985238807e-05, "loss": 0.7553, "step": 13826 }, { "epoch": 0.78, "learning_rate": 2.3419179301256822e-05, "loss": 0.7128, "step": 13827 }, { "epoch": 0.78, "learning_rate": 2.3407366202078394e-05, "loss": 0.8094, "step": 13828 }, { "epoch": 0.78, "learning_rate": 2.339555568810221e-05, "loss": 0.6157, "step": 13829 }, { "epoch": 0.78, "learning_rate": 2.3383747759726916e-05, "loss": 0.6688, "step": 13830 }, { "epoch": 0.78, "learning_rate": 2.3371942417351077e-05, "loss": 0.7191, "step": 13831 }, { "epoch": 0.78, "learning_rate": 2.336013966137317e-05, "loss": 0.7867, "step": 13832 }, { "epoch": 0.78, "learning_rate": 2.3348339492191518e-05, "loss": 0.6584, "step": 13833 }, { "epoch": 0.78, "learning_rate": 2.3336541910204436e-05, "loss": 0.6847, "step": 13834 }, { "epoch": 0.78, "learning_rate": 2.332474691581015e-05, "loss": 0.6493, "step": 13835 }, { "epoch": 0.78, "learning_rate": 2.3312954509406705e-05, "loss": 0.7103, "step": 13836 }, { "epoch": 0.78, "learning_rate": 2.3301164691392176e-05, "loss": 0.7032, "step": 13837 }, { "epoch": 0.78, "learning_rate": 2.3289377462164484e-05, "loss": 0.6356, "step": 13838 }, { "epoch": 0.79, "learning_rate": 2.3277592822121485e-05, "loss": 0.7137, "step": 13839 }, { "epoch": 0.79, "learning_rate": 2.3265810771660968e-05, "loss": 0.6783, "step": 13840 }, { "epoch": 0.79, "learning_rate": 2.3254031311180548e-05, "loss": 0.6709, "step": 13841 }, { "epoch": 0.79, "learning_rate": 2.3242254441077883e-05, "loss": 0.6369, "step": 13842 }, { "epoch": 0.79, "learning_rate": 2.32304801617504e-05, "loss": 0.6906, "step": 13843 }, { "epoch": 0.79, "learning_rate": 2.3218708473595542e-05, "loss": 0.6703, "step": 13844 }, { "epoch": 0.79, "learning_rate": 2.320693937701065e-05, "loss": 0.6649, "step": 13845 }, { "epoch": 0.79, "learning_rate": 2.3195172872392945e-05, "loss": 0.6806, "step": 13846 }, { "epoch": 0.79, "learning_rate": 2.318340896013962e-05, "loss": 0.6037, "step": 13847 }, { "epoch": 0.79, "learning_rate": 2.3171647640647687e-05, "loss": 0.7079, "step": 13848 }, { "epoch": 0.79, "learning_rate": 2.315988891431412e-05, "loss": 0.7062, "step": 13849 }, { "epoch": 0.79, "learning_rate": 2.3148132781535813e-05, "loss": 0.7002, "step": 13850 }, { "epoch": 0.79, "learning_rate": 2.3136379242709583e-05, "loss": 0.7343, "step": 13851 }, { "epoch": 0.79, "learning_rate": 2.312462829823213e-05, "loss": 0.7416, "step": 13852 }, { "epoch": 0.79, "learning_rate": 2.3112879948500097e-05, "loss": 0.6973, "step": 13853 }, { "epoch": 0.79, "learning_rate": 2.3101134193910024e-05, "loss": 0.7604, "step": 13854 }, { "epoch": 0.79, "learning_rate": 2.3089391034858355e-05, "loss": 0.7785, "step": 13855 }, { "epoch": 0.79, "learning_rate": 2.3077650471741418e-05, "loss": 0.6189, "step": 13856 }, { "epoch": 0.79, "learning_rate": 2.3065912504955512e-05, "loss": 0.832, "step": 13857 }, { "epoch": 0.79, "learning_rate": 2.3054177134896814e-05, "loss": 0.6868, "step": 13858 }, { "epoch": 0.79, "learning_rate": 2.3042444361961446e-05, "loss": 0.7713, "step": 13859 }, { "epoch": 0.79, "learning_rate": 2.3030714186545442e-05, "loss": 0.7505, "step": 13860 }, { "epoch": 0.79, "learning_rate": 2.3018986609044657e-05, "loss": 0.6604, "step": 13861 }, { "epoch": 0.79, "learning_rate": 2.3007261629854993e-05, "loss": 0.6081, "step": 13862 }, { "epoch": 0.79, "learning_rate": 2.2995539249372145e-05, "loss": 0.7173, "step": 13863 }, { "epoch": 0.79, "learning_rate": 2.2983819467991796e-05, "loss": 0.6959, "step": 13864 }, { "epoch": 0.79, "learning_rate": 2.297210228610952e-05, "loss": 0.7716, "step": 13865 }, { "epoch": 0.79, "learning_rate": 2.2960387704120812e-05, "loss": 0.7807, "step": 13866 }, { "epoch": 0.79, "learning_rate": 2.2948675722421086e-05, "loss": 0.6855, "step": 13867 }, { "epoch": 0.79, "learning_rate": 2.293696634140561e-05, "loss": 0.7431, "step": 13868 }, { "epoch": 0.79, "learning_rate": 2.2925259561469636e-05, "loss": 0.6733, "step": 13869 }, { "epoch": 0.79, "learning_rate": 2.291355538300828e-05, "loss": 0.8114, "step": 13870 }, { "epoch": 0.79, "learning_rate": 2.2901853806416586e-05, "loss": 0.6622, "step": 13871 }, { "epoch": 0.79, "learning_rate": 2.2890154832089518e-05, "loss": 0.677, "step": 13872 }, { "epoch": 0.79, "learning_rate": 2.287845846042198e-05, "loss": 0.7228, "step": 13873 }, { "epoch": 0.79, "learning_rate": 2.2866764691808706e-05, "loss": 0.8007, "step": 13874 }, { "epoch": 0.79, "learning_rate": 2.2855073526644422e-05, "loss": 0.6538, "step": 13875 }, { "epoch": 0.79, "learning_rate": 2.2843384965323712e-05, "loss": 0.6877, "step": 13876 }, { "epoch": 0.79, "learning_rate": 2.2831699008241137e-05, "loss": 0.6965, "step": 13877 }, { "epoch": 0.79, "learning_rate": 2.282001565579108e-05, "loss": 0.7438, "step": 13878 }, { "epoch": 0.79, "learning_rate": 2.2808334908367914e-05, "loss": 0.6265, "step": 13879 }, { "epoch": 0.79, "learning_rate": 2.2796656766365898e-05, "loss": 0.7291, "step": 13880 }, { "epoch": 0.79, "learning_rate": 2.278498123017918e-05, "loss": 0.7102, "step": 13881 }, { "epoch": 0.79, "learning_rate": 2.2773308300201835e-05, "loss": 0.8043, "step": 13882 }, { "epoch": 0.79, "learning_rate": 2.2761637976827875e-05, "loss": 0.6704, "step": 13883 }, { "epoch": 0.79, "learning_rate": 2.2749970260451225e-05, "loss": 0.6791, "step": 13884 }, { "epoch": 0.79, "learning_rate": 2.2738305151465645e-05, "loss": 0.7224, "step": 13885 }, { "epoch": 0.79, "learning_rate": 2.2726642650264895e-05, "loss": 0.648, "step": 13886 }, { "epoch": 0.79, "learning_rate": 2.2714982757242643e-05, "loss": 0.6287, "step": 13887 }, { "epoch": 0.79, "learning_rate": 2.270332547279238e-05, "loss": 0.7422, "step": 13888 }, { "epoch": 0.79, "learning_rate": 2.2691670797307596e-05, "loss": 0.677, "step": 13889 }, { "epoch": 0.79, "learning_rate": 2.2680018731181675e-05, "loss": 0.7773, "step": 13890 }, { "epoch": 0.79, "learning_rate": 2.26683692748079e-05, "loss": 0.7077, "step": 13891 }, { "epoch": 0.79, "learning_rate": 2.26567224285795e-05, "loss": 0.8036, "step": 13892 }, { "epoch": 0.79, "learning_rate": 2.2645078192889556e-05, "loss": 0.7618, "step": 13893 }, { "epoch": 0.79, "learning_rate": 2.2633436568131074e-05, "loss": 0.6764, "step": 13894 }, { "epoch": 0.79, "learning_rate": 2.2621797554697e-05, "loss": 0.6372, "step": 13895 }, { "epoch": 0.79, "learning_rate": 2.261016115298019e-05, "loss": 0.6998, "step": 13896 }, { "epoch": 0.79, "learning_rate": 2.2598527363373412e-05, "loss": 0.7971, "step": 13897 }, { "epoch": 0.79, "learning_rate": 2.2586896186269312e-05, "loss": 0.806, "step": 13898 }, { "epoch": 0.79, "learning_rate": 2.2575267622060536e-05, "loss": 0.7269, "step": 13899 }, { "epoch": 0.79, "learning_rate": 2.2563641671139523e-05, "loss": 0.6577, "step": 13900 }, { "epoch": 0.79, "learning_rate": 2.2552018333898663e-05, "loss": 0.7343, "step": 13901 }, { "epoch": 0.79, "learning_rate": 2.254039761073029e-05, "loss": 0.6994, "step": 13902 }, { "epoch": 0.79, "learning_rate": 2.2528779502026652e-05, "loss": 0.6789, "step": 13903 }, { "epoch": 0.79, "learning_rate": 2.2517164008179882e-05, "loss": 0.7007, "step": 13904 }, { "epoch": 0.79, "learning_rate": 2.2505551129582047e-05, "loss": 0.6748, "step": 13905 }, { "epoch": 0.79, "learning_rate": 2.249394086662512e-05, "loss": 0.6574, "step": 13906 }, { "epoch": 0.79, "learning_rate": 2.2482333219700912e-05, "loss": 0.7435, "step": 13907 }, { "epoch": 0.79, "learning_rate": 2.2470728189201263e-05, "loss": 0.7301, "step": 13908 }, { "epoch": 0.79, "learning_rate": 2.2459125775517852e-05, "loss": 0.6717, "step": 13909 }, { "epoch": 0.79, "learning_rate": 2.2447525979042317e-05, "loss": 0.7422, "step": 13910 }, { "epoch": 0.79, "learning_rate": 2.2435928800166174e-05, "loss": 0.6925, "step": 13911 }, { "epoch": 0.79, "learning_rate": 2.2424334239280876e-05, "loss": 0.7436, "step": 13912 }, { "epoch": 0.79, "learning_rate": 2.241274229677771e-05, "loss": 0.6834, "step": 13913 }, { "epoch": 0.79, "learning_rate": 2.240115297304801e-05, "loss": 0.6844, "step": 13914 }, { "epoch": 0.79, "learning_rate": 2.2389566268482875e-05, "loss": 0.7033, "step": 13915 }, { "epoch": 0.79, "learning_rate": 2.2377982183473424e-05, "loss": 0.7574, "step": 13916 }, { "epoch": 0.79, "learning_rate": 2.2366400718410642e-05, "loss": 0.7015, "step": 13917 }, { "epoch": 0.79, "learning_rate": 2.2354821873685438e-05, "loss": 0.8437, "step": 13918 }, { "epoch": 0.79, "learning_rate": 2.234324564968865e-05, "loss": 0.632, "step": 13919 }, { "epoch": 0.79, "learning_rate": 2.233167204681096e-05, "loss": 0.6115, "step": 13920 }, { "epoch": 0.79, "learning_rate": 2.2320101065443056e-05, "loss": 0.7322, "step": 13921 }, { "epoch": 0.79, "learning_rate": 2.230853270597545e-05, "loss": 0.6591, "step": 13922 }, { "epoch": 0.79, "learning_rate": 2.2296966968798606e-05, "loss": 0.7407, "step": 13923 }, { "epoch": 0.79, "learning_rate": 2.2285403854302912e-05, "loss": 0.746, "step": 13924 }, { "epoch": 0.79, "learning_rate": 2.2273843362878677e-05, "loss": 0.6832, "step": 13925 }, { "epoch": 0.79, "learning_rate": 2.2262285494916046e-05, "loss": 0.6747, "step": 13926 }, { "epoch": 0.79, "learning_rate": 2.2250730250805164e-05, "loss": 0.7358, "step": 13927 }, { "epoch": 0.79, "learning_rate": 2.2239177630936024e-05, "loss": 0.69, "step": 13928 }, { "epoch": 0.79, "learning_rate": 2.222762763569862e-05, "loss": 0.7237, "step": 13929 }, { "epoch": 0.79, "learning_rate": 2.221608026548271e-05, "loss": 0.7391, "step": 13930 }, { "epoch": 0.79, "learning_rate": 2.2204535520678095e-05, "loss": 0.6833, "step": 13931 }, { "epoch": 0.79, "learning_rate": 2.2192993401674445e-05, "loss": 0.7537, "step": 13932 }, { "epoch": 0.79, "learning_rate": 2.2181453908861317e-05, "loss": 0.7278, "step": 13933 }, { "epoch": 0.79, "learning_rate": 2.2169917042628186e-05, "loss": 0.6939, "step": 13934 }, { "epoch": 0.79, "learning_rate": 2.2158382803364487e-05, "loss": 0.6827, "step": 13935 }, { "epoch": 0.79, "learning_rate": 2.214685119145954e-05, "loss": 0.799, "step": 13936 }, { "epoch": 0.79, "learning_rate": 2.21353222073025e-05, "loss": 0.7222, "step": 13937 }, { "epoch": 0.79, "learning_rate": 2.2123795851282592e-05, "loss": 0.6718, "step": 13938 }, { "epoch": 0.79, "learning_rate": 2.2112272123788768e-05, "loss": 0.7427, "step": 13939 }, { "epoch": 0.79, "learning_rate": 2.210075102521004e-05, "loss": 0.7111, "step": 13940 }, { "epoch": 0.79, "learning_rate": 2.2089232555935258e-05, "loss": 0.6681, "step": 13941 }, { "epoch": 0.79, "learning_rate": 2.2077716716353203e-05, "loss": 0.6668, "step": 13942 }, { "epoch": 0.79, "learning_rate": 2.2066203506852566e-05, "loss": 0.6605, "step": 13943 }, { "epoch": 0.79, "learning_rate": 2.2054692927821986e-05, "loss": 0.658, "step": 13944 }, { "epoch": 0.79, "learning_rate": 2.2043184979649933e-05, "loss": 0.6105, "step": 13945 }, { "epoch": 0.79, "learning_rate": 2.2031679662724812e-05, "loss": 0.8206, "step": 13946 }, { "epoch": 0.79, "learning_rate": 2.202017697743498e-05, "loss": 0.6607, "step": 13947 }, { "epoch": 0.79, "learning_rate": 2.200867692416868e-05, "loss": 0.7213, "step": 13948 }, { "epoch": 0.79, "learning_rate": 2.199717950331408e-05, "loss": 0.7127, "step": 13949 }, { "epoch": 0.79, "learning_rate": 2.198568471525927e-05, "loss": 0.665, "step": 13950 }, { "epoch": 0.79, "learning_rate": 2.1974192560392172e-05, "loss": 0.7781, "step": 13951 }, { "epoch": 0.79, "learning_rate": 2.196270303910073e-05, "loss": 0.6726, "step": 13952 }, { "epoch": 0.79, "learning_rate": 2.195121615177269e-05, "loss": 0.7289, "step": 13953 }, { "epoch": 0.79, "learning_rate": 2.1939731898795802e-05, "loss": 0.6536, "step": 13954 }, { "epoch": 0.79, "learning_rate": 2.1928250280557683e-05, "loss": 0.73, "step": 13955 }, { "epoch": 0.79, "learning_rate": 2.191677129744586e-05, "loss": 0.7251, "step": 13956 }, { "epoch": 0.79, "learning_rate": 2.190529494984782e-05, "loss": 0.7021, "step": 13957 }, { "epoch": 0.79, "learning_rate": 2.189382123815088e-05, "loss": 0.8641, "step": 13958 }, { "epoch": 0.79, "learning_rate": 2.1882350162742292e-05, "loss": 0.7736, "step": 13959 }, { "epoch": 0.79, "learning_rate": 2.187088172400926e-05, "loss": 0.7383, "step": 13960 }, { "epoch": 0.79, "learning_rate": 2.1859415922338866e-05, "loss": 0.7176, "step": 13961 }, { "epoch": 0.79, "learning_rate": 2.1847952758118117e-05, "loss": 0.6633, "step": 13962 }, { "epoch": 0.79, "learning_rate": 2.1836492231733928e-05, "loss": 0.6841, "step": 13963 }, { "epoch": 0.79, "learning_rate": 2.1825034343573147e-05, "loss": 0.7323, "step": 13964 }, { "epoch": 0.79, "learning_rate": 2.181357909402244e-05, "loss": 0.7436, "step": 13965 }, { "epoch": 0.79, "learning_rate": 2.180212648346852e-05, "loss": 0.7654, "step": 13966 }, { "epoch": 0.79, "learning_rate": 2.1790676512297892e-05, "loss": 0.801, "step": 13967 }, { "epoch": 0.79, "learning_rate": 2.177922918089704e-05, "loss": 0.674, "step": 13968 }, { "epoch": 0.79, "learning_rate": 2.1767784489652343e-05, "loss": 0.7279, "step": 13969 }, { "epoch": 0.79, "learning_rate": 2.1756342438950117e-05, "loss": 0.7124, "step": 13970 }, { "epoch": 0.79, "learning_rate": 2.174490302917651e-05, "loss": 0.6947, "step": 13971 }, { "epoch": 0.79, "learning_rate": 2.173346626071766e-05, "loss": 0.676, "step": 13972 }, { "epoch": 0.79, "learning_rate": 2.1722032133959613e-05, "loss": 0.6233, "step": 13973 }, { "epoch": 0.79, "learning_rate": 2.171060064928824e-05, "loss": 0.7051, "step": 13974 }, { "epoch": 0.79, "learning_rate": 2.1699171807089412e-05, "loss": 0.7736, "step": 13975 }, { "epoch": 0.79, "learning_rate": 2.16877456077489e-05, "loss": 0.7519, "step": 13976 }, { "epoch": 0.79, "learning_rate": 2.1676322051652375e-05, "loss": 0.7657, "step": 13977 }, { "epoch": 0.79, "learning_rate": 2.1664901139185368e-05, "loss": 0.7343, "step": 13978 }, { "epoch": 0.79, "learning_rate": 2.165348287073339e-05, "loss": 0.8152, "step": 13979 }, { "epoch": 0.79, "learning_rate": 2.164206724668183e-05, "loss": 0.7014, "step": 13980 }, { "epoch": 0.79, "learning_rate": 2.163065426741603e-05, "loss": 0.7103, "step": 13981 }, { "epoch": 0.79, "learning_rate": 2.1619243933321155e-05, "loss": 0.8097, "step": 13982 }, { "epoch": 0.79, "learning_rate": 2.1607836244782386e-05, "loss": 0.7037, "step": 13983 }, { "epoch": 0.79, "learning_rate": 2.1596431202184708e-05, "loss": 0.6908, "step": 13984 }, { "epoch": 0.79, "learning_rate": 2.15850288059131e-05, "loss": 0.6685, "step": 13985 }, { "epoch": 0.79, "learning_rate": 2.1573629056352416e-05, "loss": 0.6515, "step": 13986 }, { "epoch": 0.79, "learning_rate": 2.1562231953887434e-05, "loss": 0.7575, "step": 13987 }, { "epoch": 0.79, "learning_rate": 2.1550837498902855e-05, "loss": 0.7576, "step": 13988 }, { "epoch": 0.79, "learning_rate": 2.153944569178323e-05, "loss": 0.6492, "step": 13989 }, { "epoch": 0.79, "learning_rate": 2.152805653291311e-05, "loss": 0.6035, "step": 13990 }, { "epoch": 0.79, "learning_rate": 2.1516670022676856e-05, "loss": 0.7507, "step": 13991 }, { "epoch": 0.79, "learning_rate": 2.1505286161458816e-05, "loss": 0.7504, "step": 13992 }, { "epoch": 0.79, "learning_rate": 2.149390494964323e-05, "loss": 0.5911, "step": 13993 }, { "epoch": 0.79, "learning_rate": 2.1482526387614244e-05, "loss": 0.6591, "step": 13994 }, { "epoch": 0.79, "learning_rate": 2.147115047575593e-05, "loss": 0.793, "step": 13995 }, { "epoch": 0.79, "learning_rate": 2.1459777214452226e-05, "loss": 0.7499, "step": 13996 }, { "epoch": 0.79, "learning_rate": 2.1448406604087034e-05, "loss": 0.7425, "step": 13997 }, { "epoch": 0.79, "learning_rate": 2.1437038645044106e-05, "loss": 0.7918, "step": 13998 }, { "epoch": 0.79, "learning_rate": 2.1425673337707164e-05, "loss": 0.6547, "step": 13999 }, { "epoch": 0.79, "learning_rate": 2.1414310682459802e-05, "loss": 0.7513, "step": 14000 }, { "epoch": 0.79, "learning_rate": 2.1402950679685574e-05, "loss": 0.711, "step": 14001 }, { "epoch": 0.79, "learning_rate": 2.13915933297679e-05, "loss": 0.6307, "step": 14002 }, { "epoch": 0.79, "learning_rate": 2.138023863309012e-05, "loss": 0.7006, "step": 14003 }, { "epoch": 0.79, "learning_rate": 2.1368886590035443e-05, "loss": 0.7077, "step": 14004 }, { "epoch": 0.79, "learning_rate": 2.1357537200987055e-05, "loss": 0.7178, "step": 14005 }, { "epoch": 0.79, "learning_rate": 2.134619046632804e-05, "loss": 0.7785, "step": 14006 }, { "epoch": 0.79, "learning_rate": 2.1334846386441378e-05, "loss": 0.6339, "step": 14007 }, { "epoch": 0.79, "learning_rate": 2.132350496170995e-05, "loss": 0.5755, "step": 14008 }, { "epoch": 0.79, "learning_rate": 2.1312166192516592e-05, "loss": 0.6995, "step": 14009 }, { "epoch": 0.79, "learning_rate": 2.1300830079243994e-05, "loss": 0.7625, "step": 14010 }, { "epoch": 0.79, "learning_rate": 2.1289496622274753e-05, "loss": 0.7194, "step": 14011 }, { "epoch": 0.79, "learning_rate": 2.1278165821991424e-05, "loss": 0.7293, "step": 14012 }, { "epoch": 0.79, "learning_rate": 2.1266837678776465e-05, "loss": 0.5589, "step": 14013 }, { "epoch": 0.79, "learning_rate": 2.1255512193012205e-05, "loss": 0.7271, "step": 14014 }, { "epoch": 0.79, "learning_rate": 2.1244189365080957e-05, "loss": 0.6455, "step": 14015 }, { "epoch": 0.8, "learning_rate": 2.1232869195364836e-05, "loss": 0.6652, "step": 14016 }, { "epoch": 0.8, "learning_rate": 2.1221551684245944e-05, "loss": 0.8293, "step": 14017 }, { "epoch": 0.8, "learning_rate": 2.1210236832106323e-05, "loss": 0.7218, "step": 14018 }, { "epoch": 0.8, "learning_rate": 2.119892463932781e-05, "loss": 0.8009, "step": 14019 }, { "epoch": 0.8, "learning_rate": 2.118761510629226e-05, "loss": 0.6909, "step": 14020 }, { "epoch": 0.8, "learning_rate": 2.1176308233381393e-05, "loss": 0.7282, "step": 14021 }, { "epoch": 0.8, "learning_rate": 2.1165004020976875e-05, "loss": 0.6704, "step": 14022 }, { "epoch": 0.8, "learning_rate": 2.115370246946019e-05, "loss": 0.7178, "step": 14023 }, { "epoch": 0.8, "learning_rate": 2.1142403579212833e-05, "loss": 0.7787, "step": 14024 }, { "epoch": 0.8, "learning_rate": 2.1131107350616187e-05, "loss": 0.716, "step": 14025 }, { "epoch": 0.8, "learning_rate": 2.111981378405149e-05, "loss": 0.693, "step": 14026 }, { "epoch": 0.8, "learning_rate": 2.110852287989995e-05, "loss": 0.7522, "step": 14027 }, { "epoch": 0.8, "learning_rate": 2.1097234638542684e-05, "loss": 0.7053, "step": 14028 }, { "epoch": 0.8, "learning_rate": 2.1085949060360654e-05, "loss": 0.7503, "step": 14029 }, { "epoch": 0.8, "learning_rate": 2.10746661457348e-05, "loss": 0.6349, "step": 14030 }, { "epoch": 0.8, "learning_rate": 2.106338589504596e-05, "loss": 0.6641, "step": 14031 }, { "epoch": 0.8, "learning_rate": 2.105210830867489e-05, "loss": 0.6721, "step": 14032 }, { "epoch": 0.8, "learning_rate": 2.1040833387002178e-05, "loss": 0.6605, "step": 14033 }, { "epoch": 0.8, "learning_rate": 2.1029561130408426e-05, "loss": 0.6698, "step": 14034 }, { "epoch": 0.8, "learning_rate": 2.1018291539274114e-05, "loss": 0.7012, "step": 14035 }, { "epoch": 0.8, "learning_rate": 2.1007024613979577e-05, "loss": 0.712, "step": 14036 }, { "epoch": 0.8, "learning_rate": 2.0995760354905125e-05, "loss": 0.7306, "step": 14037 }, { "epoch": 0.8, "learning_rate": 2.098449876243096e-05, "loss": 0.7734, "step": 14038 }, { "epoch": 0.8, "learning_rate": 2.0973239836937186e-05, "loss": 0.7151, "step": 14039 }, { "epoch": 0.8, "learning_rate": 2.0961983578803844e-05, "loss": 0.6744, "step": 14040 }, { "epoch": 0.8, "learning_rate": 2.0950729988410822e-05, "loss": 0.7348, "step": 14041 }, { "epoch": 0.8, "learning_rate": 2.0939479066138e-05, "loss": 0.7089, "step": 14042 }, { "epoch": 0.8, "learning_rate": 2.092823081236508e-05, "loss": 0.6192, "step": 14043 }, { "epoch": 0.8, "learning_rate": 2.091698522747174e-05, "loss": 0.8132, "step": 14044 }, { "epoch": 0.8, "learning_rate": 2.090574231183755e-05, "loss": 0.6519, "step": 14045 }, { "epoch": 0.8, "learning_rate": 2.0894502065841992e-05, "loss": 0.7289, "step": 14046 }, { "epoch": 0.8, "learning_rate": 2.0883264489864475e-05, "loss": 0.6967, "step": 14047 }, { "epoch": 0.8, "learning_rate": 2.087202958428428e-05, "loss": 0.7582, "step": 14048 }, { "epoch": 0.8, "learning_rate": 2.0860797349480577e-05, "loss": 0.6331, "step": 14049 }, { "epoch": 0.8, "learning_rate": 2.0849567785832514e-05, "loss": 0.663, "step": 14050 }, { "epoch": 0.8, "learning_rate": 2.083834089371911e-05, "loss": 0.6556, "step": 14051 }, { "epoch": 0.8, "learning_rate": 2.082711667351932e-05, "loss": 0.7504, "step": 14052 }, { "epoch": 0.8, "learning_rate": 2.0815895125611972e-05, "loss": 0.7745, "step": 14053 }, { "epoch": 0.8, "learning_rate": 2.0804676250375867e-05, "loss": 0.6435, "step": 14054 }, { "epoch": 0.8, "learning_rate": 2.079346004818963e-05, "loss": 0.663, "step": 14055 }, { "epoch": 0.8, "learning_rate": 2.0782246519431813e-05, "loss": 0.7479, "step": 14056 }, { "epoch": 0.8, "learning_rate": 2.0771035664480942e-05, "loss": 0.6969, "step": 14057 }, { "epoch": 0.8, "learning_rate": 2.0759827483715398e-05, "loss": 0.8119, "step": 14058 }, { "epoch": 0.8, "learning_rate": 2.0748621977513495e-05, "loss": 0.766, "step": 14059 }, { "epoch": 0.8, "learning_rate": 2.0737419146253467e-05, "loss": 0.6386, "step": 14060 }, { "epoch": 0.8, "learning_rate": 2.0726218990313386e-05, "loss": 0.7322, "step": 14061 }, { "epoch": 0.8, "learning_rate": 2.0715021510071353e-05, "loss": 0.702, "step": 14062 }, { "epoch": 0.8, "learning_rate": 2.0703826705905238e-05, "loss": 0.7738, "step": 14063 }, { "epoch": 0.8, "learning_rate": 2.069263457819294e-05, "loss": 0.6507, "step": 14064 }, { "epoch": 0.8, "learning_rate": 2.0681445127312214e-05, "loss": 0.6979, "step": 14065 }, { "epoch": 0.8, "learning_rate": 2.0670258353640748e-05, "loss": 0.7504, "step": 14066 }, { "epoch": 0.8, "learning_rate": 2.065907425755612e-05, "loss": 0.6623, "step": 14067 }, { "epoch": 0.8, "learning_rate": 2.06478928394358e-05, "loss": 0.6254, "step": 14068 }, { "epoch": 0.8, "learning_rate": 2.0636714099657238e-05, "loss": 0.672, "step": 14069 }, { "epoch": 0.8, "learning_rate": 2.062553803859767e-05, "loss": 0.702, "step": 14070 }, { "epoch": 0.8, "learning_rate": 2.0614364656634376e-05, "loss": 0.662, "step": 14071 }, { "epoch": 0.8, "learning_rate": 2.0603193954144462e-05, "loss": 0.6902, "step": 14072 }, { "epoch": 0.8, "learning_rate": 2.0592025931505e-05, "loss": 0.6664, "step": 14073 }, { "epoch": 0.8, "learning_rate": 2.0580860589092897e-05, "loss": 0.7267, "step": 14074 }, { "epoch": 0.8, "learning_rate": 2.0569697927285035e-05, "loss": 0.6886, "step": 14075 }, { "epoch": 0.8, "learning_rate": 2.0558537946458177e-05, "loss": 0.6935, "step": 14076 }, { "epoch": 0.8, "learning_rate": 2.0547380646989045e-05, "loss": 0.7033, "step": 14077 }, { "epoch": 0.8, "learning_rate": 2.0536226029254146e-05, "loss": 0.6217, "step": 14078 }, { "epoch": 0.8, "learning_rate": 2.0525074093630036e-05, "loss": 0.7714, "step": 14079 }, { "epoch": 0.8, "learning_rate": 2.0513924840493113e-05, "loss": 0.6298, "step": 14080 }, { "epoch": 0.8, "learning_rate": 2.050277827021968e-05, "loss": 0.6078, "step": 14081 }, { "epoch": 0.8, "learning_rate": 2.0491634383185954e-05, "loss": 0.7083, "step": 14082 }, { "epoch": 0.8, "learning_rate": 2.048049317976809e-05, "loss": 0.7521, "step": 14083 }, { "epoch": 0.8, "learning_rate": 2.0469354660342164e-05, "loss": 0.6965, "step": 14084 }, { "epoch": 0.8, "learning_rate": 2.0458218825284058e-05, "loss": 0.7284, "step": 14085 }, { "epoch": 0.8, "learning_rate": 2.0447085674969678e-05, "loss": 0.6036, "step": 14086 }, { "epoch": 0.8, "learning_rate": 2.0435955209774817e-05, "loss": 0.6921, "step": 14087 }, { "epoch": 0.8, "learning_rate": 2.04248274300751e-05, "loss": 0.726, "step": 14088 }, { "epoch": 0.8, "learning_rate": 2.0413702336246154e-05, "loss": 0.6666, "step": 14089 }, { "epoch": 0.8, "learning_rate": 2.040257992866348e-05, "loss": 0.5741, "step": 14090 }, { "epoch": 0.8, "learning_rate": 2.0391460207702484e-05, "loss": 0.7055, "step": 14091 }, { "epoch": 0.8, "learning_rate": 2.0380343173738515e-05, "loss": 0.6921, "step": 14092 }, { "epoch": 0.8, "learning_rate": 2.036922882714677e-05, "loss": 0.6359, "step": 14093 }, { "epoch": 0.8, "learning_rate": 2.035811716830237e-05, "loss": 0.7616, "step": 14094 }, { "epoch": 0.8, "learning_rate": 2.0347008197580374e-05, "loss": 0.6769, "step": 14095 }, { "epoch": 0.8, "learning_rate": 2.033590191535576e-05, "loss": 0.7128, "step": 14096 }, { "epoch": 0.8, "learning_rate": 2.032479832200338e-05, "loss": 0.7987, "step": 14097 }, { "epoch": 0.8, "learning_rate": 2.0313697417898013e-05, "loss": 0.5633, "step": 14098 }, { "epoch": 0.8, "learning_rate": 2.0302599203414374e-05, "loss": 0.7422, "step": 14099 }, { "epoch": 0.8, "learning_rate": 2.0291503678927016e-05, "loss": 0.7123, "step": 14100 }, { "epoch": 0.8, "learning_rate": 2.0280410844810428e-05, "loss": 0.6754, "step": 14101 }, { "epoch": 0.8, "learning_rate": 2.0269320701439053e-05, "loss": 0.6135, "step": 14102 }, { "epoch": 0.8, "learning_rate": 2.02582332491872e-05, "loss": 0.6514, "step": 14103 }, { "epoch": 0.8, "learning_rate": 2.02471484884291e-05, "loss": 0.7204, "step": 14104 }, { "epoch": 0.8, "learning_rate": 2.0236066419538934e-05, "loss": 0.6923, "step": 14105 }, { "epoch": 0.8, "learning_rate": 2.0224987042890687e-05, "loss": 0.6246, "step": 14106 }, { "epoch": 0.8, "learning_rate": 2.0213910358858366e-05, "loss": 0.7707, "step": 14107 }, { "epoch": 0.8, "learning_rate": 2.020283636781579e-05, "loss": 0.7239, "step": 14108 }, { "epoch": 0.8, "learning_rate": 2.019176507013677e-05, "loss": 0.6849, "step": 14109 }, { "epoch": 0.8, "learning_rate": 2.018069646619498e-05, "loss": 0.7235, "step": 14110 }, { "epoch": 0.8, "learning_rate": 2.0169630556364016e-05, "loss": 0.663, "step": 14111 }, { "epoch": 0.8, "learning_rate": 2.015856734101742e-05, "loss": 0.6626, "step": 14112 }, { "epoch": 0.8, "learning_rate": 2.0147506820528527e-05, "loss": 0.6203, "step": 14113 }, { "epoch": 0.8, "learning_rate": 2.013644899527074e-05, "loss": 0.7093, "step": 14114 }, { "epoch": 0.8, "learning_rate": 2.0125393865617216e-05, "loss": 0.7342, "step": 14115 }, { "epoch": 0.8, "learning_rate": 2.0114341431941118e-05, "loss": 0.7233, "step": 14116 }, { "epoch": 0.8, "learning_rate": 2.010329169461552e-05, "loss": 0.6776, "step": 14117 }, { "epoch": 0.8, "learning_rate": 2.009224465401338e-05, "loss": 0.7959, "step": 14118 }, { "epoch": 0.8, "learning_rate": 2.008120031050753e-05, "loss": 0.7349, "step": 14119 }, { "epoch": 0.8, "learning_rate": 2.0070158664470762e-05, "loss": 0.7018, "step": 14120 }, { "epoch": 0.8, "learning_rate": 2.0059119716275788e-05, "loss": 0.6281, "step": 14121 }, { "epoch": 0.8, "learning_rate": 2.0048083466295153e-05, "loss": 0.6866, "step": 14122 }, { "epoch": 0.8, "learning_rate": 2.0037049914901373e-05, "loss": 0.6496, "step": 14123 }, { "epoch": 0.8, "learning_rate": 2.0026019062466884e-05, "loss": 0.6669, "step": 14124 }, { "epoch": 0.8, "learning_rate": 2.0014990909364008e-05, "loss": 0.7523, "step": 14125 }, { "epoch": 0.8, "learning_rate": 2.0003965455964935e-05, "loss": 0.6228, "step": 14126 }, { "epoch": 0.8, "learning_rate": 1.9992942702641814e-05, "loss": 0.7492, "step": 14127 }, { "epoch": 0.8, "learning_rate": 1.9981922649766715e-05, "loss": 0.8568, "step": 14128 }, { "epoch": 0.8, "learning_rate": 1.9970905297711606e-05, "loss": 0.7877, "step": 14129 }, { "epoch": 0.8, "learning_rate": 1.9959890646848302e-05, "loss": 0.7196, "step": 14130 }, { "epoch": 0.8, "learning_rate": 1.994887869754859e-05, "loss": 0.7589, "step": 14131 }, { "epoch": 0.8, "learning_rate": 1.9937869450184198e-05, "loss": 0.5621, "step": 14132 }, { "epoch": 0.8, "learning_rate": 1.9926862905126665e-05, "loss": 0.7121, "step": 14133 }, { "epoch": 0.8, "learning_rate": 1.99158590627475e-05, "loss": 0.7066, "step": 14134 }, { "epoch": 0.8, "learning_rate": 1.990485792341812e-05, "loss": 0.6755, "step": 14135 }, { "epoch": 0.8, "learning_rate": 1.989385948750987e-05, "loss": 0.7456, "step": 14136 }, { "epoch": 0.8, "learning_rate": 1.988286375539391e-05, "loss": 0.7867, "step": 14137 }, { "epoch": 0.8, "learning_rate": 1.9871870727441456e-05, "loss": 0.6663, "step": 14138 }, { "epoch": 0.8, "learning_rate": 1.9860880404023474e-05, "loss": 0.7054, "step": 14139 }, { "epoch": 0.8, "learning_rate": 1.9849892785510958e-05, "loss": 0.674, "step": 14140 }, { "epoch": 0.8, "learning_rate": 1.983890787227476e-05, "loss": 0.6584, "step": 14141 }, { "epoch": 0.8, "learning_rate": 1.982792566468564e-05, "loss": 0.7263, "step": 14142 }, { "epoch": 0.8, "learning_rate": 1.9816946163114303e-05, "loss": 0.7545, "step": 14143 }, { "epoch": 0.8, "learning_rate": 1.9805969367931344e-05, "loss": 0.7099, "step": 14144 }, { "epoch": 0.8, "learning_rate": 1.979499527950722e-05, "loss": 0.7322, "step": 14145 }, { "epoch": 0.8, "learning_rate": 1.9784023898212343e-05, "loss": 0.7504, "step": 14146 }, { "epoch": 0.8, "learning_rate": 1.9773055224417025e-05, "loss": 0.616, "step": 14147 }, { "epoch": 0.8, "learning_rate": 1.97620892584915e-05, "loss": 0.743, "step": 14148 }, { "epoch": 0.8, "learning_rate": 1.9751126000805897e-05, "loss": 0.7314, "step": 14149 }, { "epoch": 0.8, "learning_rate": 1.9740165451730276e-05, "loss": 0.638, "step": 14150 }, { "epoch": 0.8, "learning_rate": 1.972920761163455e-05, "loss": 0.7395, "step": 14151 }, { "epoch": 0.8, "learning_rate": 1.9718252480888566e-05, "loss": 0.7071, "step": 14152 }, { "epoch": 0.8, "learning_rate": 1.970730005986211e-05, "loss": 0.6603, "step": 14153 }, { "epoch": 0.8, "learning_rate": 1.969635034892485e-05, "loss": 0.7658, "step": 14154 }, { "epoch": 0.8, "learning_rate": 1.9685403348446374e-05, "loss": 0.6704, "step": 14155 }, { "epoch": 0.8, "learning_rate": 1.967445905879616e-05, "loss": 0.7036, "step": 14156 }, { "epoch": 0.8, "learning_rate": 1.9663517480343642e-05, "loss": 0.6692, "step": 14157 }, { "epoch": 0.8, "learning_rate": 1.9652578613458105e-05, "loss": 0.7805, "step": 14158 }, { "epoch": 0.8, "learning_rate": 1.9641642458508726e-05, "loss": 0.7209, "step": 14159 }, { "epoch": 0.8, "learning_rate": 1.963070901586467e-05, "loss": 0.656, "step": 14160 }, { "epoch": 0.8, "learning_rate": 1.9619778285894953e-05, "loss": 0.6933, "step": 14161 }, { "epoch": 0.8, "learning_rate": 1.9608850268968527e-05, "loss": 0.6896, "step": 14162 }, { "epoch": 0.8, "learning_rate": 1.9597924965454262e-05, "loss": 0.7489, "step": 14163 }, { "epoch": 0.8, "learning_rate": 1.9587002375720864e-05, "loss": 0.6676, "step": 14164 }, { "epoch": 0.8, "learning_rate": 1.957608250013703e-05, "loss": 0.6729, "step": 14165 }, { "epoch": 0.8, "learning_rate": 1.9565165339071356e-05, "loss": 0.6816, "step": 14166 }, { "epoch": 0.8, "learning_rate": 1.9554250892892266e-05, "loss": 0.7259, "step": 14167 }, { "epoch": 0.8, "learning_rate": 1.95433391619682e-05, "loss": 0.665, "step": 14168 }, { "epoch": 0.8, "learning_rate": 1.9532430146667436e-05, "loss": 0.7043, "step": 14169 }, { "epoch": 0.8, "learning_rate": 1.9521523847358214e-05, "loss": 0.6969, "step": 14170 }, { "epoch": 0.8, "learning_rate": 1.9510620264408596e-05, "loss": 0.6494, "step": 14171 }, { "epoch": 0.8, "learning_rate": 1.9499719398186632e-05, "loss": 0.7022, "step": 14172 }, { "epoch": 0.8, "learning_rate": 1.9488821249060297e-05, "loss": 0.7535, "step": 14173 }, { "epoch": 0.8, "learning_rate": 1.947792581739736e-05, "loss": 0.765, "step": 14174 }, { "epoch": 0.8, "learning_rate": 1.9467033103565602e-05, "loss": 0.6752, "step": 14175 }, { "epoch": 0.8, "learning_rate": 1.9456143107932678e-05, "loss": 0.6977, "step": 14176 }, { "epoch": 0.8, "learning_rate": 1.9445255830866194e-05, "loss": 0.7503, "step": 14177 }, { "epoch": 0.8, "learning_rate": 1.9434371272733566e-05, "loss": 0.8063, "step": 14178 }, { "epoch": 0.8, "learning_rate": 1.9423489433902186e-05, "loss": 0.693, "step": 14179 }, { "epoch": 0.8, "learning_rate": 1.9412610314739366e-05, "loss": 0.6846, "step": 14180 }, { "epoch": 0.8, "learning_rate": 1.9401733915612307e-05, "loss": 0.6559, "step": 14181 }, { "epoch": 0.8, "learning_rate": 1.9390860236888097e-05, "loss": 0.7529, "step": 14182 }, { "epoch": 0.8, "learning_rate": 1.9379989278933774e-05, "loss": 0.6801, "step": 14183 }, { "epoch": 0.8, "learning_rate": 1.9369121042116235e-05, "loss": 0.6366, "step": 14184 }, { "epoch": 0.8, "learning_rate": 1.9358255526802303e-05, "loss": 0.6943, "step": 14185 }, { "epoch": 0.8, "learning_rate": 1.934739273335875e-05, "loss": 0.6776, "step": 14186 }, { "epoch": 0.8, "learning_rate": 1.9336532662152205e-05, "loss": 0.6783, "step": 14187 }, { "epoch": 0.8, "learning_rate": 1.9325675313549264e-05, "loss": 0.7313, "step": 14188 }, { "epoch": 0.8, "learning_rate": 1.9314820687916336e-05, "loss": 0.7139, "step": 14189 }, { "epoch": 0.8, "learning_rate": 1.930396878561983e-05, "loss": 0.6771, "step": 14190 }, { "epoch": 0.8, "learning_rate": 1.9293119607025987e-05, "loss": 0.6716, "step": 14191 }, { "epoch": 0.81, "learning_rate": 1.9282273152501017e-05, "loss": 0.7276, "step": 14192 }, { "epoch": 0.81, "learning_rate": 1.9271429422411013e-05, "loss": 0.6934, "step": 14193 }, { "epoch": 0.81, "learning_rate": 1.9260588417121995e-05, "loss": 0.7362, "step": 14194 }, { "epoch": 0.81, "learning_rate": 1.9249750136999888e-05, "loss": 0.7183, "step": 14195 }, { "epoch": 0.81, "learning_rate": 1.9238914582410484e-05, "loss": 0.6928, "step": 14196 }, { "epoch": 0.81, "learning_rate": 1.9228081753719496e-05, "loss": 0.7214, "step": 14197 }, { "epoch": 0.81, "learning_rate": 1.9217251651292577e-05, "loss": 0.7058, "step": 14198 }, { "epoch": 0.81, "learning_rate": 1.9206424275495283e-05, "loss": 0.6918, "step": 14199 }, { "epoch": 0.81, "learning_rate": 1.919559962669305e-05, "loss": 0.6022, "step": 14200 }, { "epoch": 0.81, "learning_rate": 1.9184777705251257e-05, "loss": 0.7127, "step": 14201 }, { "epoch": 0.81, "learning_rate": 1.9173958511535182e-05, "loss": 0.7206, "step": 14202 }, { "epoch": 0.81, "learning_rate": 1.9163142045909976e-05, "loss": 0.7041, "step": 14203 }, { "epoch": 0.81, "learning_rate": 1.9152328308740707e-05, "loss": 0.705, "step": 14204 }, { "epoch": 0.81, "learning_rate": 1.9141517300392387e-05, "loss": 0.6409, "step": 14205 }, { "epoch": 0.81, "learning_rate": 1.913070902122992e-05, "loss": 0.6417, "step": 14206 }, { "epoch": 0.81, "learning_rate": 1.9119903471618117e-05, "loss": 0.7079, "step": 14207 }, { "epoch": 0.81, "learning_rate": 1.9109100651921706e-05, "loss": 0.8246, "step": 14208 }, { "epoch": 0.81, "learning_rate": 1.9098300562505266e-05, "loss": 0.8037, "step": 14209 }, { "epoch": 0.81, "learning_rate": 1.908750320373338e-05, "loss": 0.7709, "step": 14210 }, { "epoch": 0.81, "learning_rate": 1.9076708575970436e-05, "loss": 0.6773, "step": 14211 }, { "epoch": 0.81, "learning_rate": 1.906591667958081e-05, "loss": 0.7063, "step": 14212 }, { "epoch": 0.81, "learning_rate": 1.905512751492875e-05, "loss": 0.5845, "step": 14213 }, { "epoch": 0.81, "learning_rate": 1.9044341082378425e-05, "loss": 0.7086, "step": 14214 }, { "epoch": 0.81, "learning_rate": 1.903355738229392e-05, "loss": 0.705, "step": 14215 }, { "epoch": 0.81, "learning_rate": 1.902277641503918e-05, "loss": 0.6385, "step": 14216 }, { "epoch": 0.81, "learning_rate": 1.9011998180978108e-05, "loss": 0.643, "step": 14217 }, { "epoch": 0.81, "learning_rate": 1.9001222680474517e-05, "loss": 0.64, "step": 14218 }, { "epoch": 0.81, "learning_rate": 1.899044991389206e-05, "loss": 0.6447, "step": 14219 }, { "epoch": 0.81, "learning_rate": 1.897967988159437e-05, "loss": 0.6497, "step": 14220 }, { "epoch": 0.81, "learning_rate": 1.8968912583944975e-05, "loss": 0.6594, "step": 14221 }, { "epoch": 0.81, "learning_rate": 1.895814802130732e-05, "loss": 0.652, "step": 14222 }, { "epoch": 0.81, "learning_rate": 1.894738619404468e-05, "loss": 0.7341, "step": 14223 }, { "epoch": 0.81, "learning_rate": 1.8936627102520323e-05, "loss": 0.5713, "step": 14224 }, { "epoch": 0.81, "learning_rate": 1.8925870747097417e-05, "loss": 0.7304, "step": 14225 }, { "epoch": 0.81, "learning_rate": 1.8915117128138983e-05, "loss": 0.6771, "step": 14226 }, { "epoch": 0.81, "learning_rate": 1.8904366246008e-05, "loss": 0.7262, "step": 14227 }, { "epoch": 0.81, "learning_rate": 1.8893618101067355e-05, "loss": 0.7282, "step": 14228 }, { "epoch": 0.81, "learning_rate": 1.888287269367979e-05, "loss": 0.6929, "step": 14229 }, { "epoch": 0.81, "learning_rate": 1.8872130024208002e-05, "loss": 0.633, "step": 14230 }, { "epoch": 0.81, "learning_rate": 1.8861390093014597e-05, "loss": 0.6779, "step": 14231 }, { "epoch": 0.81, "learning_rate": 1.8850652900462086e-05, "loss": 0.7942, "step": 14232 }, { "epoch": 0.81, "learning_rate": 1.883991844691284e-05, "loss": 0.7653, "step": 14233 }, { "epoch": 0.81, "learning_rate": 1.882918673272921e-05, "loss": 0.6959, "step": 14234 }, { "epoch": 0.81, "learning_rate": 1.881845775827341e-05, "loss": 0.6999, "step": 14235 }, { "epoch": 0.81, "learning_rate": 1.8807731523907558e-05, "loss": 0.7324, "step": 14236 }, { "epoch": 0.81, "learning_rate": 1.8797008029993703e-05, "loss": 0.6966, "step": 14237 }, { "epoch": 0.81, "learning_rate": 1.8786287276893777e-05, "loss": 0.6354, "step": 14238 }, { "epoch": 0.81, "learning_rate": 1.8775569264969652e-05, "loss": 0.7128, "step": 14239 }, { "epoch": 0.81, "learning_rate": 1.876485399458311e-05, "loss": 0.6847, "step": 14240 }, { "epoch": 0.81, "learning_rate": 1.8754141466095786e-05, "loss": 0.7367, "step": 14241 }, { "epoch": 0.81, "learning_rate": 1.874343167986925e-05, "loss": 0.7496, "step": 14242 }, { "epoch": 0.81, "learning_rate": 1.8732724636264987e-05, "loss": 0.7486, "step": 14243 }, { "epoch": 0.81, "learning_rate": 1.8722020335644408e-05, "loss": 0.581, "step": 14244 }, { "epoch": 0.81, "learning_rate": 1.871131877836879e-05, "loss": 0.6305, "step": 14245 }, { "epoch": 0.81, "learning_rate": 1.870061996479936e-05, "loss": 0.5806, "step": 14246 }, { "epoch": 0.81, "learning_rate": 1.8689923895297245e-05, "loss": 0.701, "step": 14247 }, { "epoch": 0.81, "learning_rate": 1.867923057022344e-05, "loss": 0.6743, "step": 14248 }, { "epoch": 0.81, "learning_rate": 1.866853998993885e-05, "loss": 0.6404, "step": 14249 }, { "epoch": 0.81, "learning_rate": 1.8657852154804333e-05, "loss": 0.6887, "step": 14250 }, { "epoch": 0.81, "learning_rate": 1.864716706518064e-05, "loss": 0.7231, "step": 14251 }, { "epoch": 0.81, "learning_rate": 1.8636484721428403e-05, "loss": 0.8523, "step": 14252 }, { "epoch": 0.81, "learning_rate": 1.8625805123908202e-05, "loss": 0.7217, "step": 14253 }, { "epoch": 0.81, "learning_rate": 1.861512827298051e-05, "loss": 0.6711, "step": 14254 }, { "epoch": 0.81, "learning_rate": 1.860445416900568e-05, "loss": 0.594, "step": 14255 }, { "epoch": 0.81, "learning_rate": 1.8593782812343962e-05, "loss": 0.7433, "step": 14256 }, { "epoch": 0.81, "learning_rate": 1.8583114203355568e-05, "loss": 0.6822, "step": 14257 }, { "epoch": 0.81, "learning_rate": 1.8572448342400594e-05, "loss": 0.567, "step": 14258 }, { "epoch": 0.81, "learning_rate": 1.8561785229839048e-05, "loss": 0.7087, "step": 14259 }, { "epoch": 0.81, "learning_rate": 1.8551124866030844e-05, "loss": 0.7163, "step": 14260 }, { "epoch": 0.81, "learning_rate": 1.854046725133577e-05, "loss": 0.784, "step": 14261 }, { "epoch": 0.81, "learning_rate": 1.8529812386113575e-05, "loss": 0.6981, "step": 14262 }, { "epoch": 0.81, "learning_rate": 1.8519160270723857e-05, "loss": 0.6915, "step": 14263 }, { "epoch": 0.81, "learning_rate": 1.8508510905526168e-05, "loss": 0.6589, "step": 14264 }, { "epoch": 0.81, "learning_rate": 1.8497864290879953e-05, "loss": 0.7426, "step": 14265 }, { "epoch": 0.81, "learning_rate": 1.848722042714457e-05, "loss": 0.7073, "step": 14266 }, { "epoch": 0.81, "learning_rate": 1.8476579314679297e-05, "loss": 0.6877, "step": 14267 }, { "epoch": 0.81, "learning_rate": 1.846594095384325e-05, "loss": 0.7138, "step": 14268 }, { "epoch": 0.81, "learning_rate": 1.8455305344995523e-05, "loss": 0.6521, "step": 14269 }, { "epoch": 0.81, "learning_rate": 1.844467248849513e-05, "loss": 0.6484, "step": 14270 }, { "epoch": 0.81, "learning_rate": 1.84340423847009e-05, "loss": 0.8119, "step": 14271 }, { "epoch": 0.81, "learning_rate": 1.8423415033971657e-05, "loss": 0.8225, "step": 14272 }, { "epoch": 0.81, "learning_rate": 1.8412790436666115e-05, "loss": 0.6575, "step": 14273 }, { "epoch": 0.81, "learning_rate": 1.8402168593142855e-05, "loss": 0.7249, "step": 14274 }, { "epoch": 0.81, "learning_rate": 1.8391549503760396e-05, "loss": 0.7538, "step": 14275 }, { "epoch": 0.81, "learning_rate": 1.838093316887717e-05, "loss": 0.6727, "step": 14276 }, { "epoch": 0.81, "learning_rate": 1.8370319588851516e-05, "loss": 0.725, "step": 14277 }, { "epoch": 0.81, "learning_rate": 1.835970876404164e-05, "loss": 0.7204, "step": 14278 }, { "epoch": 0.81, "learning_rate": 1.834910069480571e-05, "loss": 0.6929, "step": 14279 }, { "epoch": 0.81, "learning_rate": 1.8338495381501786e-05, "loss": 0.6654, "step": 14280 }, { "epoch": 0.81, "learning_rate": 1.8327892824487792e-05, "loss": 0.7508, "step": 14281 }, { "epoch": 0.81, "learning_rate": 1.8317293024121597e-05, "loss": 0.7273, "step": 14282 }, { "epoch": 0.81, "learning_rate": 1.830669598076099e-05, "loss": 0.7514, "step": 14283 }, { "epoch": 0.81, "learning_rate": 1.8296101694763656e-05, "loss": 0.6921, "step": 14284 }, { "epoch": 0.81, "learning_rate": 1.8285510166487152e-05, "loss": 0.7623, "step": 14285 }, { "epoch": 0.81, "learning_rate": 1.827492139628898e-05, "loss": 0.6396, "step": 14286 }, { "epoch": 0.81, "learning_rate": 1.8264335384526578e-05, "loss": 0.6833, "step": 14287 }, { "epoch": 0.81, "learning_rate": 1.825375213155718e-05, "loss": 0.7232, "step": 14288 }, { "epoch": 0.81, "learning_rate": 1.8243171637738034e-05, "loss": 0.7123, "step": 14289 }, { "epoch": 0.81, "learning_rate": 1.8232593903426266e-05, "loss": 0.6958, "step": 14290 }, { "epoch": 0.81, "learning_rate": 1.82220189289789e-05, "loss": 0.6925, "step": 14291 }, { "epoch": 0.81, "learning_rate": 1.8211446714752878e-05, "loss": 0.5972, "step": 14292 }, { "epoch": 0.81, "learning_rate": 1.8200877261105042e-05, "loss": 0.6594, "step": 14293 }, { "epoch": 0.81, "learning_rate": 1.819031056839209e-05, "loss": 0.7545, "step": 14294 }, { "epoch": 0.81, "learning_rate": 1.817974663697072e-05, "loss": 0.7657, "step": 14295 }, { "epoch": 0.81, "learning_rate": 1.816918546719747e-05, "loss": 0.6594, "step": 14296 }, { "epoch": 0.81, "learning_rate": 1.8158627059428824e-05, "loss": 0.678, "step": 14297 }, { "epoch": 0.81, "learning_rate": 1.814807141402115e-05, "loss": 0.6907, "step": 14298 }, { "epoch": 0.81, "learning_rate": 1.8137518531330767e-05, "loss": 0.6951, "step": 14299 }, { "epoch": 0.81, "learning_rate": 1.812696841171381e-05, "loss": 0.693, "step": 14300 }, { "epoch": 0.81, "learning_rate": 1.8116421055526366e-05, "loss": 0.6575, "step": 14301 }, { "epoch": 0.81, "learning_rate": 1.8105876463124473e-05, "loss": 0.6956, "step": 14302 }, { "epoch": 0.81, "learning_rate": 1.8095334634864014e-05, "loss": 0.6765, "step": 14303 }, { "epoch": 0.81, "learning_rate": 1.808479557110081e-05, "loss": 0.7683, "step": 14304 }, { "epoch": 0.81, "learning_rate": 1.8074259272190618e-05, "loss": 0.6902, "step": 14305 }, { "epoch": 0.81, "learning_rate": 1.8063725738489012e-05, "loss": 0.6691, "step": 14306 }, { "epoch": 0.81, "learning_rate": 1.805319497035156e-05, "loss": 0.6732, "step": 14307 }, { "epoch": 0.81, "learning_rate": 1.804266696813368e-05, "loss": 0.7703, "step": 14308 }, { "epoch": 0.81, "learning_rate": 1.803214173219072e-05, "loss": 0.6205, "step": 14309 }, { "epoch": 0.81, "learning_rate": 1.8021619262877955e-05, "loss": 0.7541, "step": 14310 }, { "epoch": 0.81, "learning_rate": 1.8011099560550526e-05, "loss": 0.5947, "step": 14311 }, { "epoch": 0.81, "learning_rate": 1.8000582625563545e-05, "loss": 0.6742, "step": 14312 }, { "epoch": 0.81, "learning_rate": 1.7990068458271924e-05, "loss": 0.8295, "step": 14313 }, { "epoch": 0.81, "learning_rate": 1.7979557059030596e-05, "loss": 0.7417, "step": 14314 }, { "epoch": 0.81, "learning_rate": 1.796904842819429e-05, "loss": 0.714, "step": 14315 }, { "epoch": 0.81, "learning_rate": 1.7958542566117742e-05, "loss": 0.6942, "step": 14316 }, { "epoch": 0.81, "learning_rate": 1.7948039473155554e-05, "loss": 0.7598, "step": 14317 }, { "epoch": 0.81, "learning_rate": 1.793753914966223e-05, "loss": 0.73, "step": 14318 }, { "epoch": 0.81, "learning_rate": 1.7927041595992157e-05, "loss": 0.7799, "step": 14319 }, { "epoch": 0.81, "learning_rate": 1.7916546812499678e-05, "loss": 0.6937, "step": 14320 }, { "epoch": 0.81, "learning_rate": 1.790605479953904e-05, "loss": 0.6299, "step": 14321 }, { "epoch": 0.81, "learning_rate": 1.7895565557464322e-05, "loss": 0.6756, "step": 14322 }, { "epoch": 0.81, "learning_rate": 1.78850790866296e-05, "loss": 0.7162, "step": 14323 }, { "epoch": 0.81, "learning_rate": 1.7874595387388814e-05, "loss": 0.6633, "step": 14324 }, { "epoch": 0.81, "learning_rate": 1.7864114460095838e-05, "loss": 0.6855, "step": 14325 }, { "epoch": 0.81, "learning_rate": 1.785363630510438e-05, "loss": 0.6249, "step": 14326 }, { "epoch": 0.81, "learning_rate": 1.784316092276813e-05, "loss": 0.6409, "step": 14327 }, { "epoch": 0.81, "learning_rate": 1.783268831344067e-05, "loss": 0.6834, "step": 14328 }, { "epoch": 0.81, "learning_rate": 1.7822218477475494e-05, "loss": 0.6868, "step": 14329 }, { "epoch": 0.81, "learning_rate": 1.7811751415225943e-05, "loss": 0.6256, "step": 14330 }, { "epoch": 0.81, "learning_rate": 1.780128712704532e-05, "loss": 0.7136, "step": 14331 }, { "epoch": 0.81, "learning_rate": 1.7790825613286854e-05, "loss": 0.7058, "step": 14332 }, { "epoch": 0.81, "learning_rate": 1.778036687430361e-05, "loss": 0.792, "step": 14333 }, { "epoch": 0.81, "learning_rate": 1.7769910910448605e-05, "loss": 0.7114, "step": 14334 }, { "epoch": 0.81, "learning_rate": 1.7759457722074768e-05, "loss": 0.7624, "step": 14335 }, { "epoch": 0.81, "learning_rate": 1.7749007309534948e-05, "loss": 0.7663, "step": 14336 }, { "epoch": 0.81, "learning_rate": 1.773855967318181e-05, "loss": 0.6368, "step": 14337 }, { "epoch": 0.81, "learning_rate": 1.7728114813368045e-05, "loss": 0.7465, "step": 14338 }, { "epoch": 0.81, "learning_rate": 1.7717672730446145e-05, "loss": 0.7995, "step": 14339 }, { "epoch": 0.81, "learning_rate": 1.7707233424768587e-05, "loss": 0.7211, "step": 14340 }, { "epoch": 0.81, "learning_rate": 1.7696796896687718e-05, "loss": 0.6647, "step": 14341 }, { "epoch": 0.81, "learning_rate": 1.7686363146555805e-05, "loss": 0.6358, "step": 14342 }, { "epoch": 0.81, "learning_rate": 1.7675932174725006e-05, "loss": 0.7722, "step": 14343 }, { "epoch": 0.81, "learning_rate": 1.7665503981547428e-05, "loss": 0.7764, "step": 14344 }, { "epoch": 0.81, "learning_rate": 1.7655078567375028e-05, "loss": 0.6558, "step": 14345 }, { "epoch": 0.81, "learning_rate": 1.764465593255965e-05, "loss": 0.7358, "step": 14346 }, { "epoch": 0.81, "learning_rate": 1.7634236077453116e-05, "loss": 0.6738, "step": 14347 }, { "epoch": 0.81, "learning_rate": 1.762381900240714e-05, "loss": 0.7134, "step": 14348 }, { "epoch": 0.81, "learning_rate": 1.7613404707773295e-05, "loss": 0.7525, "step": 14349 }, { "epoch": 0.81, "learning_rate": 1.760299319390315e-05, "loss": 0.7324, "step": 14350 }, { "epoch": 0.81, "learning_rate": 1.7592584461148065e-05, "loss": 0.7551, "step": 14351 }, { "epoch": 0.81, "learning_rate": 1.758217850985936e-05, "loss": 0.6803, "step": 14352 }, { "epoch": 0.81, "learning_rate": 1.7571775340388276e-05, "loss": 0.7195, "step": 14353 }, { "epoch": 0.81, "learning_rate": 1.756137495308594e-05, "loss": 0.7164, "step": 14354 }, { "epoch": 0.81, "learning_rate": 1.755097734830342e-05, "loss": 0.6701, "step": 14355 }, { "epoch": 0.81, "learning_rate": 1.7540582526391637e-05, "loss": 0.7168, "step": 14356 }, { "epoch": 0.81, "learning_rate": 1.753019048770147e-05, "loss": 0.8491, "step": 14357 }, { "epoch": 0.81, "learning_rate": 1.7519801232583667e-05, "loss": 0.8009, "step": 14358 }, { "epoch": 0.81, "learning_rate": 1.7509414761388853e-05, "loss": 0.7088, "step": 14359 }, { "epoch": 0.81, "learning_rate": 1.749903107446764e-05, "loss": 0.7057, "step": 14360 }, { "epoch": 0.81, "learning_rate": 1.7488650172170496e-05, "loss": 0.6818, "step": 14361 }, { "epoch": 0.81, "learning_rate": 1.7478272054847787e-05, "loss": 0.7529, "step": 14362 }, { "epoch": 0.81, "learning_rate": 1.7467896722849852e-05, "loss": 0.7758, "step": 14363 }, { "epoch": 0.81, "learning_rate": 1.745752417652683e-05, "loss": 0.7014, "step": 14364 }, { "epoch": 0.81, "learning_rate": 1.7447154416228838e-05, "loss": 0.7344, "step": 14365 }, { "epoch": 0.81, "learning_rate": 1.7436787442305914e-05, "loss": 0.6775, "step": 14366 }, { "epoch": 0.81, "learning_rate": 1.7426423255107915e-05, "loss": 0.7249, "step": 14367 }, { "epoch": 0.82, "learning_rate": 1.741606185498469e-05, "loss": 0.7334, "step": 14368 }, { "epoch": 0.82, "learning_rate": 1.7405703242285965e-05, "loss": 0.6243, "step": 14369 }, { "epoch": 0.82, "learning_rate": 1.7395347417361395e-05, "loss": 0.5945, "step": 14370 }, { "epoch": 0.82, "learning_rate": 1.738499438056045e-05, "loss": 0.6565, "step": 14371 }, { "epoch": 0.82, "learning_rate": 1.737464413223262e-05, "loss": 0.7945, "step": 14372 }, { "epoch": 0.82, "learning_rate": 1.7364296672727266e-05, "loss": 0.7537, "step": 14373 }, { "epoch": 0.82, "learning_rate": 1.7353952002393602e-05, "loss": 0.8569, "step": 14374 }, { "epoch": 0.82, "learning_rate": 1.73436101215808e-05, "loss": 0.5832, "step": 14375 }, { "epoch": 0.82, "learning_rate": 1.733327103063792e-05, "loss": 0.7527, "step": 14376 }, { "epoch": 0.82, "learning_rate": 1.7322934729913986e-05, "loss": 0.6628, "step": 14377 }, { "epoch": 0.82, "learning_rate": 1.7312601219757796e-05, "loss": 0.677, "step": 14378 }, { "epoch": 0.82, "learning_rate": 1.7302270500518182e-05, "loss": 0.7321, "step": 14379 }, { "epoch": 0.82, "learning_rate": 1.7291942572543807e-05, "loss": 0.6932, "step": 14380 }, { "epoch": 0.82, "learning_rate": 1.7281617436183317e-05, "loss": 0.6831, "step": 14381 }, { "epoch": 0.82, "learning_rate": 1.7271295091785146e-05, "loss": 0.7493, "step": 14382 }, { "epoch": 0.82, "learning_rate": 1.7260975539697753e-05, "loss": 0.6789, "step": 14383 }, { "epoch": 0.82, "learning_rate": 1.7250658780269403e-05, "loss": 0.6617, "step": 14384 }, { "epoch": 0.82, "learning_rate": 1.7240344813848343e-05, "loss": 0.656, "step": 14385 }, { "epoch": 0.82, "learning_rate": 1.7230033640782684e-05, "loss": 0.6862, "step": 14386 }, { "epoch": 0.82, "learning_rate": 1.7219725261420462e-05, "loss": 0.7142, "step": 14387 }, { "epoch": 0.82, "learning_rate": 1.7209419676109627e-05, "loss": 0.6941, "step": 14388 }, { "epoch": 0.82, "learning_rate": 1.7199116885197995e-05, "loss": 0.6812, "step": 14389 }, { "epoch": 0.82, "learning_rate": 1.7188816889033345e-05, "loss": 0.7003, "step": 14390 }, { "epoch": 0.82, "learning_rate": 1.717851968796328e-05, "loss": 0.5907, "step": 14391 }, { "epoch": 0.82, "learning_rate": 1.716822528233537e-05, "loss": 0.5941, "step": 14392 }, { "epoch": 0.82, "learning_rate": 1.7157933672497107e-05, "loss": 0.6079, "step": 14393 }, { "epoch": 0.82, "learning_rate": 1.714764485879583e-05, "loss": 0.7299, "step": 14394 }, { "epoch": 0.82, "learning_rate": 1.7137358841578866e-05, "loss": 0.6634, "step": 14395 }, { "epoch": 0.82, "learning_rate": 1.7127075621193346e-05, "loss": 0.6512, "step": 14396 }, { "epoch": 0.82, "learning_rate": 1.7116795197986347e-05, "loss": 0.6831, "step": 14397 }, { "epoch": 0.82, "learning_rate": 1.710651757230488e-05, "loss": 0.6542, "step": 14398 }, { "epoch": 0.82, "learning_rate": 1.7096242744495837e-05, "loss": 0.6265, "step": 14399 }, { "epoch": 0.82, "learning_rate": 1.708597071490602e-05, "loss": 0.7042, "step": 14400 }, { "epoch": 0.82, "learning_rate": 1.7075701483882146e-05, "loss": 0.7064, "step": 14401 }, { "epoch": 0.82, "learning_rate": 1.706543505177085e-05, "loss": 0.7444, "step": 14402 }, { "epoch": 0.82, "learning_rate": 1.7055171418918626e-05, "loss": 0.6712, "step": 14403 }, { "epoch": 0.82, "learning_rate": 1.704491058567187e-05, "loss": 0.6448, "step": 14404 }, { "epoch": 0.82, "learning_rate": 1.703465255237694e-05, "loss": 0.6795, "step": 14405 }, { "epoch": 0.82, "learning_rate": 1.702439731938007e-05, "loss": 0.6727, "step": 14406 }, { "epoch": 0.82, "learning_rate": 1.7014144887027404e-05, "loss": 0.666, "step": 14407 }, { "epoch": 0.82, "learning_rate": 1.7003895255665014e-05, "loss": 0.6869, "step": 14408 }, { "epoch": 0.82, "learning_rate": 1.6993648425638797e-05, "loss": 0.6806, "step": 14409 }, { "epoch": 0.82, "learning_rate": 1.6983404397294666e-05, "loss": 0.7355, "step": 14410 }, { "epoch": 0.82, "learning_rate": 1.6973163170978324e-05, "loss": 0.754, "step": 14411 }, { "epoch": 0.82, "learning_rate": 1.696292474703547e-05, "loss": 0.6973, "step": 14412 }, { "epoch": 0.82, "learning_rate": 1.6952689125811682e-05, "loss": 0.7204, "step": 14413 }, { "epoch": 0.82, "learning_rate": 1.6942456307652445e-05, "loss": 0.6249, "step": 14414 }, { "epoch": 0.82, "learning_rate": 1.693222629290314e-05, "loss": 0.6944, "step": 14415 }, { "epoch": 0.82, "learning_rate": 1.6921999081909034e-05, "loss": 0.6447, "step": 14416 }, { "epoch": 0.82, "learning_rate": 1.6911774675015347e-05, "loss": 0.7297, "step": 14417 }, { "epoch": 0.82, "learning_rate": 1.690155307256719e-05, "loss": 0.5967, "step": 14418 }, { "epoch": 0.82, "learning_rate": 1.6891334274909522e-05, "loss": 0.7511, "step": 14419 }, { "epoch": 0.82, "learning_rate": 1.6881118282387286e-05, "loss": 0.6597, "step": 14420 }, { "epoch": 0.82, "learning_rate": 1.687090509534529e-05, "loss": 0.6837, "step": 14421 }, { "epoch": 0.82, "learning_rate": 1.68606947141283e-05, "loss": 0.6615, "step": 14422 }, { "epoch": 0.82, "learning_rate": 1.6850487139080874e-05, "loss": 0.7119, "step": 14423 }, { "epoch": 0.82, "learning_rate": 1.6840282370547566e-05, "loss": 0.7488, "step": 14424 }, { "epoch": 0.82, "learning_rate": 1.683008040887285e-05, "loss": 0.7547, "step": 14425 }, { "epoch": 0.82, "learning_rate": 1.6819881254401016e-05, "loss": 0.8337, "step": 14426 }, { "epoch": 0.82, "learning_rate": 1.6809684907476343e-05, "loss": 0.7402, "step": 14427 }, { "epoch": 0.82, "learning_rate": 1.6799491368442997e-05, "loss": 0.777, "step": 14428 }, { "epoch": 0.82, "learning_rate": 1.6789300637645e-05, "loss": 0.8037, "step": 14429 }, { "epoch": 0.82, "learning_rate": 1.6779112715426338e-05, "loss": 0.7428, "step": 14430 }, { "epoch": 0.82, "learning_rate": 1.676892760213087e-05, "loss": 0.6161, "step": 14431 }, { "epoch": 0.82, "learning_rate": 1.6758745298102372e-05, "loss": 0.6631, "step": 14432 }, { "epoch": 0.82, "learning_rate": 1.6748565803684568e-05, "loss": 0.7586, "step": 14433 }, { "epoch": 0.82, "learning_rate": 1.6738389119220964e-05, "loss": 0.6996, "step": 14434 }, { "epoch": 0.82, "learning_rate": 1.6728215245055124e-05, "loss": 0.7015, "step": 14435 }, { "epoch": 0.82, "learning_rate": 1.671804418153038e-05, "loss": 0.6618, "step": 14436 }, { "epoch": 0.82, "learning_rate": 1.6707875928990058e-05, "loss": 0.7114, "step": 14437 }, { "epoch": 0.82, "learning_rate": 1.6697710487777362e-05, "loss": 0.7296, "step": 14438 }, { "epoch": 0.82, "learning_rate": 1.668754785823541e-05, "loss": 0.6702, "step": 14439 }, { "epoch": 0.82, "learning_rate": 1.6677388040707232e-05, "loss": 0.6713, "step": 14440 }, { "epoch": 0.82, "learning_rate": 1.6667231035535736e-05, "loss": 0.6447, "step": 14441 }, { "epoch": 0.82, "learning_rate": 1.665707684306371e-05, "loss": 0.6925, "step": 14442 }, { "epoch": 0.82, "learning_rate": 1.6646925463633922e-05, "loss": 0.6638, "step": 14443 }, { "epoch": 0.82, "learning_rate": 1.663677689758899e-05, "loss": 0.6902, "step": 14444 }, { "epoch": 0.82, "learning_rate": 1.6626631145271476e-05, "loss": 0.6385, "step": 14445 }, { "epoch": 0.82, "learning_rate": 1.6616488207023817e-05, "loss": 0.7764, "step": 14446 }, { "epoch": 0.82, "learning_rate": 1.6606348083188385e-05, "loss": 0.6401, "step": 14447 }, { "epoch": 0.82, "learning_rate": 1.65962107741074e-05, "loss": 0.7406, "step": 14448 }, { "epoch": 0.82, "learning_rate": 1.658607628012303e-05, "loss": 0.7406, "step": 14449 }, { "epoch": 0.82, "learning_rate": 1.6575944601577333e-05, "loss": 0.747, "step": 14450 }, { "epoch": 0.82, "learning_rate": 1.6565815738812295e-05, "loss": 0.7725, "step": 14451 }, { "epoch": 0.82, "learning_rate": 1.6555689692169796e-05, "loss": 0.6358, "step": 14452 }, { "epoch": 0.82, "learning_rate": 1.6545566461991634e-05, "loss": 0.7651, "step": 14453 }, { "epoch": 0.82, "learning_rate": 1.653544604861945e-05, "loss": 0.7618, "step": 14454 }, { "epoch": 0.82, "learning_rate": 1.6525328452394872e-05, "loss": 0.7384, "step": 14455 }, { "epoch": 0.82, "learning_rate": 1.6515213673659357e-05, "loss": 0.7525, "step": 14456 }, { "epoch": 0.82, "learning_rate": 1.6505101712754334e-05, "loss": 0.7006, "step": 14457 }, { "epoch": 0.82, "learning_rate": 1.6494992570021097e-05, "loss": 0.6728, "step": 14458 }, { "epoch": 0.82, "learning_rate": 1.6484886245800857e-05, "loss": 0.6636, "step": 14459 }, { "epoch": 0.82, "learning_rate": 1.6474782740434756e-05, "loss": 0.632, "step": 14460 }, { "epoch": 0.82, "learning_rate": 1.646468205426377e-05, "loss": 0.6507, "step": 14461 }, { "epoch": 0.82, "learning_rate": 1.6454584187628873e-05, "loss": 0.6899, "step": 14462 }, { "epoch": 0.82, "learning_rate": 1.6444489140870834e-05, "loss": 0.6713, "step": 14463 }, { "epoch": 0.82, "learning_rate": 1.643439691433042e-05, "loss": 0.656, "step": 14464 }, { "epoch": 0.82, "learning_rate": 1.642430750834827e-05, "loss": 0.6492, "step": 14465 }, { "epoch": 0.82, "learning_rate": 1.6414220923264934e-05, "loss": 0.6766, "step": 14466 }, { "epoch": 0.82, "learning_rate": 1.6404137159420863e-05, "loss": 0.5844, "step": 14467 }, { "epoch": 0.82, "learning_rate": 1.6394056217156394e-05, "loss": 0.7471, "step": 14468 }, { "epoch": 0.82, "learning_rate": 1.638397809681178e-05, "loss": 0.6752, "step": 14469 }, { "epoch": 0.82, "learning_rate": 1.6373902798727225e-05, "loss": 0.7799, "step": 14470 }, { "epoch": 0.82, "learning_rate": 1.6363830323242734e-05, "loss": 0.7969, "step": 14471 }, { "epoch": 0.82, "learning_rate": 1.6353760670698326e-05, "loss": 0.7859, "step": 14472 }, { "epoch": 0.82, "learning_rate": 1.634369384143388e-05, "loss": 0.6524, "step": 14473 }, { "epoch": 0.82, "learning_rate": 1.6333629835789156e-05, "loss": 0.655, "step": 14474 }, { "epoch": 0.82, "learning_rate": 1.632356865410384e-05, "loss": 0.7083, "step": 14475 }, { "epoch": 0.82, "learning_rate": 1.6313510296717528e-05, "loss": 0.62, "step": 14476 }, { "epoch": 0.82, "learning_rate": 1.630345476396974e-05, "loss": 0.6947, "step": 14477 }, { "epoch": 0.82, "learning_rate": 1.6293402056199834e-05, "loss": 0.7209, "step": 14478 }, { "epoch": 0.82, "learning_rate": 1.6283352173747145e-05, "loss": 0.6982, "step": 14479 }, { "epoch": 0.82, "learning_rate": 1.627330511695089e-05, "loss": 0.7621, "step": 14480 }, { "epoch": 0.82, "learning_rate": 1.6263260886150143e-05, "loss": 0.7746, "step": 14481 }, { "epoch": 0.82, "learning_rate": 1.6253219481683958e-05, "loss": 0.7271, "step": 14482 }, { "epoch": 0.82, "learning_rate": 1.6243180903891243e-05, "loss": 0.7426, "step": 14483 }, { "epoch": 0.82, "learning_rate": 1.6233145153110863e-05, "loss": 0.6651, "step": 14484 }, { "epoch": 0.82, "learning_rate": 1.622311222968149e-05, "loss": 0.7323, "step": 14485 }, { "epoch": 0.82, "learning_rate": 1.6213082133941825e-05, "loss": 0.761, "step": 14486 }, { "epoch": 0.82, "learning_rate": 1.6203054866230337e-05, "loss": 0.7406, "step": 14487 }, { "epoch": 0.82, "learning_rate": 1.6193030426885525e-05, "loss": 0.7741, "step": 14488 }, { "epoch": 0.82, "learning_rate": 1.6183008816245725e-05, "loss": 0.5864, "step": 14489 }, { "epoch": 0.82, "learning_rate": 1.6172990034649206e-05, "loss": 0.6884, "step": 14490 }, { "epoch": 0.82, "learning_rate": 1.61629740824341e-05, "loss": 0.7263, "step": 14491 }, { "epoch": 0.82, "learning_rate": 1.6152960959938522e-05, "loss": 0.6768, "step": 14492 }, { "epoch": 0.82, "learning_rate": 1.6142950667500412e-05, "loss": 0.6374, "step": 14493 }, { "epoch": 0.82, "learning_rate": 1.6132943205457606e-05, "loss": 0.6525, "step": 14494 }, { "epoch": 0.82, "learning_rate": 1.612293857414793e-05, "loss": 0.7777, "step": 14495 }, { "epoch": 0.82, "learning_rate": 1.6112936773909038e-05, "loss": 0.6785, "step": 14496 }, { "epoch": 0.82, "learning_rate": 1.6102937805078544e-05, "loss": 0.7112, "step": 14497 }, { "epoch": 0.82, "learning_rate": 1.6092941667993945e-05, "loss": 0.7147, "step": 14498 }, { "epoch": 0.82, "learning_rate": 1.60829483629926e-05, "loss": 0.6139, "step": 14499 }, { "epoch": 0.82, "learning_rate": 1.6072957890411845e-05, "loss": 0.6223, "step": 14500 }, { "epoch": 0.82, "learning_rate": 1.606297025058885e-05, "loss": 0.7802, "step": 14501 }, { "epoch": 0.82, "learning_rate": 1.6052985443860734e-05, "loss": 0.6956, "step": 14502 }, { "epoch": 0.82, "learning_rate": 1.6043003470564533e-05, "loss": 0.649, "step": 14503 }, { "epoch": 0.82, "learning_rate": 1.6033024331037138e-05, "loss": 0.6912, "step": 14504 }, { "epoch": 0.82, "learning_rate": 1.6023048025615405e-05, "loss": 0.756, "step": 14505 }, { "epoch": 0.82, "learning_rate": 1.601307455463602e-05, "loss": 0.6924, "step": 14506 }, { "epoch": 0.82, "learning_rate": 1.6003103918435657e-05, "loss": 0.6024, "step": 14507 }, { "epoch": 0.82, "learning_rate": 1.5993136117350804e-05, "loss": 0.6715, "step": 14508 }, { "epoch": 0.82, "learning_rate": 1.5983171151717923e-05, "loss": 0.7439, "step": 14509 }, { "epoch": 0.82, "learning_rate": 1.5973209021873358e-05, "loss": 0.6764, "step": 14510 }, { "epoch": 0.82, "learning_rate": 1.596324972815335e-05, "loss": 0.6427, "step": 14511 }, { "epoch": 0.82, "learning_rate": 1.595329327089409e-05, "loss": 0.7278, "step": 14512 }, { "epoch": 0.82, "learning_rate": 1.5943339650431576e-05, "loss": 0.6756, "step": 14513 }, { "epoch": 0.82, "learning_rate": 1.5933388867101817e-05, "loss": 0.7415, "step": 14514 }, { "epoch": 0.82, "learning_rate": 1.592344092124064e-05, "loss": 0.5873, "step": 14515 }, { "epoch": 0.82, "learning_rate": 1.5913495813183822e-05, "loss": 0.7026, "step": 14516 }, { "epoch": 0.82, "learning_rate": 1.5903553543267047e-05, "loss": 0.5925, "step": 14517 }, { "epoch": 0.82, "learning_rate": 1.5893614111825918e-05, "loss": 0.795, "step": 14518 }, { "epoch": 0.82, "learning_rate": 1.588367751919586e-05, "loss": 0.722, "step": 14519 }, { "epoch": 0.82, "learning_rate": 1.58737437657123e-05, "loss": 0.7295, "step": 14520 }, { "epoch": 0.82, "learning_rate": 1.5863812851710536e-05, "loss": 0.7007, "step": 14521 }, { "epoch": 0.82, "learning_rate": 1.585388477752572e-05, "loss": 0.6686, "step": 14522 }, { "epoch": 0.82, "learning_rate": 1.584395954349297e-05, "loss": 0.6502, "step": 14523 }, { "epoch": 0.82, "learning_rate": 1.583403714994729e-05, "loss": 0.7331, "step": 14524 }, { "epoch": 0.82, "learning_rate": 1.582411759722362e-05, "loss": 0.736, "step": 14525 }, { "epoch": 0.82, "learning_rate": 1.5814200885656716e-05, "loss": 0.6787, "step": 14526 }, { "epoch": 0.82, "learning_rate": 1.5804287015581322e-05, "loss": 0.6403, "step": 14527 }, { "epoch": 0.82, "learning_rate": 1.5794375987332043e-05, "loss": 0.714, "step": 14528 }, { "epoch": 0.82, "learning_rate": 1.578446780124344e-05, "loss": 0.5971, "step": 14529 }, { "epoch": 0.82, "learning_rate": 1.5774562457649888e-05, "loss": 0.5379, "step": 14530 }, { "epoch": 0.82, "learning_rate": 1.5764659956885764e-05, "loss": 0.7642, "step": 14531 }, { "epoch": 0.82, "learning_rate": 1.5754760299285252e-05, "loss": 0.6913, "step": 14532 }, { "epoch": 0.82, "learning_rate": 1.5744863485182537e-05, "loss": 0.6977, "step": 14533 }, { "epoch": 0.82, "learning_rate": 1.5734969514911636e-05, "loss": 0.6976, "step": 14534 }, { "epoch": 0.82, "learning_rate": 1.5725078388806513e-05, "loss": 0.7209, "step": 14535 }, { "epoch": 0.82, "learning_rate": 1.5715190107201026e-05, "loss": 0.7092, "step": 14536 }, { "epoch": 0.82, "learning_rate": 1.570530467042891e-05, "loss": 0.7272, "step": 14537 }, { "epoch": 0.82, "learning_rate": 1.569542207882384e-05, "loss": 0.7715, "step": 14538 }, { "epoch": 0.82, "learning_rate": 1.5685542332719362e-05, "loss": 0.6667, "step": 14539 }, { "epoch": 0.82, "learning_rate": 1.5675665432448948e-05, "loss": 0.5975, "step": 14540 }, { "epoch": 0.82, "learning_rate": 1.5665791378345972e-05, "loss": 0.793, "step": 14541 }, { "epoch": 0.82, "learning_rate": 1.565592017074371e-05, "loss": 0.7528, "step": 14542 }, { "epoch": 0.82, "learning_rate": 1.564605180997537e-05, "loss": 0.6317, "step": 14543 }, { "epoch": 0.83, "learning_rate": 1.563618629637398e-05, "loss": 0.7255, "step": 14544 }, { "epoch": 0.83, "learning_rate": 1.5626323630272587e-05, "loss": 0.695, "step": 14545 }, { "epoch": 0.83, "learning_rate": 1.5616463812004022e-05, "loss": 0.797, "step": 14546 }, { "epoch": 0.83, "learning_rate": 1.5606606841901107e-05, "loss": 0.7095, "step": 14547 }, { "epoch": 0.83, "learning_rate": 1.559675272029654e-05, "loss": 0.7033, "step": 14548 }, { "epoch": 0.83, "learning_rate": 1.5586901447522916e-05, "loss": 0.7538, "step": 14549 }, { "epoch": 0.83, "learning_rate": 1.557705302391278e-05, "loss": 0.7949, "step": 14550 }, { "epoch": 0.83, "learning_rate": 1.5567207449798515e-05, "loss": 0.685, "step": 14551 }, { "epoch": 0.83, "learning_rate": 1.5557364725512413e-05, "loss": 0.7302, "step": 14552 }, { "epoch": 0.83, "learning_rate": 1.5547524851386695e-05, "loss": 0.6492, "step": 14553 }, { "epoch": 0.83, "learning_rate": 1.553768782775351e-05, "loss": 0.6398, "step": 14554 }, { "epoch": 0.83, "learning_rate": 1.5527853654944868e-05, "loss": 0.6901, "step": 14555 }, { "epoch": 0.83, "learning_rate": 1.5518022333292702e-05, "loss": 0.6848, "step": 14556 }, { "epoch": 0.83, "learning_rate": 1.5508193863128863e-05, "loss": 0.6525, "step": 14557 }, { "epoch": 0.83, "learning_rate": 1.5498368244785078e-05, "loss": 0.7307, "step": 14558 }, { "epoch": 0.83, "learning_rate": 1.5488545478592954e-05, "loss": 0.7067, "step": 14559 }, { "epoch": 0.83, "learning_rate": 1.5478725564884043e-05, "loss": 0.7375, "step": 14560 }, { "epoch": 0.83, "learning_rate": 1.546890850398983e-05, "loss": 0.6231, "step": 14561 }, { "epoch": 0.83, "learning_rate": 1.5459094296241638e-05, "loss": 0.7486, "step": 14562 }, { "epoch": 0.83, "learning_rate": 1.5449282941970757e-05, "loss": 0.6151, "step": 14563 }, { "epoch": 0.83, "learning_rate": 1.5439474441508283e-05, "loss": 0.7301, "step": 14564 }, { "epoch": 0.83, "learning_rate": 1.5429668795185327e-05, "loss": 0.6435, "step": 14565 }, { "epoch": 0.83, "learning_rate": 1.5419866003332863e-05, "loss": 0.6338, "step": 14566 }, { "epoch": 0.83, "learning_rate": 1.5410066066281714e-05, "loss": 0.6252, "step": 14567 }, { "epoch": 0.83, "learning_rate": 1.540026898436269e-05, "loss": 0.7385, "step": 14568 }, { "epoch": 0.83, "learning_rate": 1.5390474757906446e-05, "loss": 0.732, "step": 14569 }, { "epoch": 0.83, "learning_rate": 1.538068338724361e-05, "loss": 0.7811, "step": 14570 }, { "epoch": 0.83, "learning_rate": 1.537089487270461e-05, "loss": 0.7221, "step": 14571 }, { "epoch": 0.83, "learning_rate": 1.5361109214619852e-05, "loss": 0.7896, "step": 14572 }, { "epoch": 0.83, "learning_rate": 1.5351326413319645e-05, "loss": 0.6554, "step": 14573 }, { "epoch": 0.83, "learning_rate": 1.534154646913416e-05, "loss": 0.6317, "step": 14574 }, { "epoch": 0.83, "learning_rate": 1.53317693823935e-05, "loss": 0.6919, "step": 14575 }, { "epoch": 0.83, "learning_rate": 1.5321995153427706e-05, "loss": 0.6616, "step": 14576 }, { "epoch": 0.83, "learning_rate": 1.5312223782566613e-05, "loss": 0.7533, "step": 14577 }, { "epoch": 0.83, "learning_rate": 1.5302455270140082e-05, "loss": 0.7189, "step": 14578 }, { "epoch": 0.83, "learning_rate": 1.5292689616477806e-05, "loss": 0.7279, "step": 14579 }, { "epoch": 0.83, "learning_rate": 1.5282926821909415e-05, "loss": 0.682, "step": 14580 }, { "epoch": 0.83, "learning_rate": 1.5273166886764446e-05, "loss": 0.5564, "step": 14581 }, { "epoch": 0.83, "learning_rate": 1.5263409811372265e-05, "loss": 0.623, "step": 14582 }, { "epoch": 0.83, "learning_rate": 1.5253655596062267e-05, "loss": 0.6778, "step": 14583 }, { "epoch": 0.83, "learning_rate": 1.524390424116362e-05, "loss": 0.7707, "step": 14584 }, { "epoch": 0.83, "learning_rate": 1.5234155747005486e-05, "loss": 0.7765, "step": 14585 }, { "epoch": 0.83, "learning_rate": 1.5224410113916909e-05, "loss": 0.7656, "step": 14586 }, { "epoch": 0.83, "learning_rate": 1.5214667342226818e-05, "loss": 0.6892, "step": 14587 }, { "epoch": 0.83, "learning_rate": 1.5204927432264082e-05, "loss": 0.8421, "step": 14588 }, { "epoch": 0.83, "learning_rate": 1.5195190384357404e-05, "loss": 0.6994, "step": 14589 }, { "epoch": 0.83, "learning_rate": 1.518545619883549e-05, "loss": 0.7336, "step": 14590 }, { "epoch": 0.83, "learning_rate": 1.5175724876026831e-05, "loss": 0.7344, "step": 14591 }, { "epoch": 0.83, "learning_rate": 1.516599641625992e-05, "loss": 0.7501, "step": 14592 }, { "epoch": 0.83, "learning_rate": 1.5156270819863116e-05, "loss": 0.5998, "step": 14593 }, { "epoch": 0.83, "learning_rate": 1.5146548087164692e-05, "loss": 0.6006, "step": 14594 }, { "epoch": 0.83, "learning_rate": 1.5136828218492815e-05, "loss": 0.6541, "step": 14595 }, { "epoch": 0.83, "learning_rate": 1.512711121417556e-05, "loss": 0.6507, "step": 14596 }, { "epoch": 0.83, "learning_rate": 1.5117397074540862e-05, "loss": 0.7378, "step": 14597 }, { "epoch": 0.83, "learning_rate": 1.5107685799916616e-05, "loss": 0.6677, "step": 14598 }, { "epoch": 0.83, "learning_rate": 1.5097977390630625e-05, "loss": 0.5787, "step": 14599 }, { "epoch": 0.83, "learning_rate": 1.5088271847010548e-05, "loss": 0.6492, "step": 14600 }, { "epoch": 0.83, "learning_rate": 1.5078569169383994e-05, "loss": 0.6664, "step": 14601 }, { "epoch": 0.83, "learning_rate": 1.5068869358078475e-05, "loss": 0.6934, "step": 14602 }, { "epoch": 0.83, "learning_rate": 1.505917241342134e-05, "loss": 0.7336, "step": 14603 }, { "epoch": 0.83, "learning_rate": 1.5049478335739886e-05, "loss": 0.6613, "step": 14604 }, { "epoch": 0.83, "learning_rate": 1.5039787125361326e-05, "loss": 0.7238, "step": 14605 }, { "epoch": 0.83, "learning_rate": 1.5030098782612778e-05, "loss": 0.7308, "step": 14606 }, { "epoch": 0.83, "learning_rate": 1.5020413307821224e-05, "loss": 0.7542, "step": 14607 }, { "epoch": 0.83, "learning_rate": 1.5010730701313625e-05, "loss": 0.7077, "step": 14608 }, { "epoch": 0.83, "learning_rate": 1.5001050963416719e-05, "loss": 0.6827, "step": 14609 }, { "epoch": 0.83, "learning_rate": 1.4991374094457288e-05, "loss": 0.6632, "step": 14610 }, { "epoch": 0.83, "learning_rate": 1.4981700094761908e-05, "loss": 0.672, "step": 14611 }, { "epoch": 0.83, "learning_rate": 1.4972028964657114e-05, "loss": 0.6933, "step": 14612 }, { "epoch": 0.83, "learning_rate": 1.4962360704469337e-05, "loss": 0.7765, "step": 14613 }, { "epoch": 0.83, "learning_rate": 1.4952695314524912e-05, "loss": 0.7346, "step": 14614 }, { "epoch": 0.83, "learning_rate": 1.494303279515008e-05, "loss": 0.6815, "step": 14615 }, { "epoch": 0.83, "learning_rate": 1.4933373146670937e-05, "loss": 0.6118, "step": 14616 }, { "epoch": 0.83, "learning_rate": 1.4923716369413553e-05, "loss": 0.6737, "step": 14617 }, { "epoch": 0.83, "learning_rate": 1.4914062463703882e-05, "loss": 0.7003, "step": 14618 }, { "epoch": 0.83, "learning_rate": 1.4904411429867726e-05, "loss": 0.6707, "step": 14619 }, { "epoch": 0.83, "learning_rate": 1.4894763268230849e-05, "loss": 0.74, "step": 14620 }, { "epoch": 0.83, "learning_rate": 1.4885117979118912e-05, "loss": 0.6649, "step": 14621 }, { "epoch": 0.83, "learning_rate": 1.4875475562857488e-05, "loss": 0.6817, "step": 14622 }, { "epoch": 0.83, "learning_rate": 1.4865836019771995e-05, "loss": 0.7235, "step": 14623 }, { "epoch": 0.83, "learning_rate": 1.4856199350187806e-05, "loss": 0.729, "step": 14624 }, { "epoch": 0.83, "learning_rate": 1.4846565554430203e-05, "loss": 0.7278, "step": 14625 }, { "epoch": 0.83, "learning_rate": 1.4836934632824317e-05, "loss": 0.7617, "step": 14626 }, { "epoch": 0.83, "learning_rate": 1.4827306585695234e-05, "loss": 0.7339, "step": 14627 }, { "epoch": 0.83, "learning_rate": 1.4817681413367957e-05, "loss": 0.7244, "step": 14628 }, { "epoch": 0.83, "learning_rate": 1.4808059116167305e-05, "loss": 0.6942, "step": 14629 }, { "epoch": 0.83, "learning_rate": 1.4798439694418077e-05, "loss": 0.7119, "step": 14630 }, { "epoch": 0.83, "learning_rate": 1.4788823148444963e-05, "loss": 0.7109, "step": 14631 }, { "epoch": 0.83, "learning_rate": 1.4779209478572542e-05, "loss": 0.7796, "step": 14632 }, { "epoch": 0.83, "learning_rate": 1.476959868512532e-05, "loss": 0.7131, "step": 14633 }, { "epoch": 0.83, "learning_rate": 1.4759990768427645e-05, "loss": 0.7327, "step": 14634 }, { "epoch": 0.83, "learning_rate": 1.4750385728803862e-05, "loss": 0.6578, "step": 14635 }, { "epoch": 0.83, "learning_rate": 1.4740783566578109e-05, "loss": 0.5994, "step": 14636 }, { "epoch": 0.83, "learning_rate": 1.4731184282074506e-05, "loss": 0.695, "step": 14637 }, { "epoch": 0.83, "learning_rate": 1.472158787561706e-05, "loss": 0.7708, "step": 14638 }, { "epoch": 0.83, "learning_rate": 1.471199434752968e-05, "loss": 0.732, "step": 14639 }, { "epoch": 0.83, "learning_rate": 1.4702403698136192e-05, "loss": 0.7771, "step": 14640 }, { "epoch": 0.83, "learning_rate": 1.4692815927760273e-05, "loss": 0.6603, "step": 14641 }, { "epoch": 0.83, "learning_rate": 1.4683231036725519e-05, "loss": 0.6653, "step": 14642 }, { "epoch": 0.83, "learning_rate": 1.4673649025355474e-05, "loss": 0.7705, "step": 14643 }, { "epoch": 0.83, "learning_rate": 1.4664069893973542e-05, "loss": 0.6434, "step": 14644 }, { "epoch": 0.83, "learning_rate": 1.4654493642903066e-05, "loss": 0.7682, "step": 14645 }, { "epoch": 0.83, "learning_rate": 1.4644920272467244e-05, "loss": 0.7123, "step": 14646 }, { "epoch": 0.83, "learning_rate": 1.463534978298925e-05, "loss": 0.7239, "step": 14647 }, { "epoch": 0.83, "learning_rate": 1.4625782174792068e-05, "loss": 0.7181, "step": 14648 }, { "epoch": 0.83, "learning_rate": 1.4616217448198621e-05, "loss": 0.6989, "step": 14649 }, { "epoch": 0.83, "learning_rate": 1.460665560353176e-05, "loss": 0.6829, "step": 14650 }, { "epoch": 0.83, "learning_rate": 1.4597096641114227e-05, "loss": 0.7459, "step": 14651 }, { "epoch": 0.83, "learning_rate": 1.4587540561268654e-05, "loss": 0.7028, "step": 14652 }, { "epoch": 0.83, "learning_rate": 1.4577987364317625e-05, "loss": 0.7111, "step": 14653 }, { "epoch": 0.83, "learning_rate": 1.4568437050583517e-05, "loss": 0.6843, "step": 14654 }, { "epoch": 0.83, "learning_rate": 1.4558889620388739e-05, "loss": 0.6691, "step": 14655 }, { "epoch": 0.83, "learning_rate": 1.4549345074055487e-05, "loss": 0.7229, "step": 14656 }, { "epoch": 0.83, "learning_rate": 1.453980341190594e-05, "loss": 0.7112, "step": 14657 }, { "epoch": 0.83, "learning_rate": 1.4530264634262159e-05, "loss": 0.7321, "step": 14658 }, { "epoch": 0.83, "learning_rate": 1.4520728741446089e-05, "loss": 0.6523, "step": 14659 }, { "epoch": 0.83, "learning_rate": 1.4511195733779637e-05, "loss": 0.697, "step": 14660 }, { "epoch": 0.83, "learning_rate": 1.4501665611584503e-05, "loss": 0.754, "step": 14661 }, { "epoch": 0.83, "learning_rate": 1.4492138375182407e-05, "loss": 0.7377, "step": 14662 }, { "epoch": 0.83, "learning_rate": 1.448261402489487e-05, "loss": 0.7414, "step": 14663 }, { "epoch": 0.83, "learning_rate": 1.447309256104339e-05, "loss": 0.768, "step": 14664 }, { "epoch": 0.83, "learning_rate": 1.4463573983949341e-05, "loss": 0.6991, "step": 14665 }, { "epoch": 0.83, "learning_rate": 1.4454058293933992e-05, "loss": 0.639, "step": 14666 }, { "epoch": 0.83, "learning_rate": 1.444454549131855e-05, "loss": 0.7287, "step": 14667 }, { "epoch": 0.83, "learning_rate": 1.443503557642405e-05, "loss": 0.709, "step": 14668 }, { "epoch": 0.83, "learning_rate": 1.442552854957151e-05, "loss": 0.6824, "step": 14669 }, { "epoch": 0.83, "learning_rate": 1.441602441108183e-05, "loss": 0.6776, "step": 14670 }, { "epoch": 0.83, "learning_rate": 1.4406523161275754e-05, "loss": 0.7047, "step": 14671 }, { "epoch": 0.83, "learning_rate": 1.4397024800473991e-05, "loss": 0.7144, "step": 14672 }, { "epoch": 0.83, "learning_rate": 1.4387529328997173e-05, "loss": 0.7557, "step": 14673 }, { "epoch": 0.83, "learning_rate": 1.4378036747165746e-05, "loss": 0.628, "step": 14674 }, { "epoch": 0.83, "learning_rate": 1.436854705530013e-05, "loss": 0.7441, "step": 14675 }, { "epoch": 0.83, "learning_rate": 1.4359060253720623e-05, "loss": 0.6615, "step": 14676 }, { "epoch": 0.83, "learning_rate": 1.4349576342747462e-05, "loss": 0.6928, "step": 14677 }, { "epoch": 0.83, "learning_rate": 1.4340095322700698e-05, "loss": 0.6666, "step": 14678 }, { "epoch": 0.83, "learning_rate": 1.4330617193900364e-05, "loss": 0.6458, "step": 14679 }, { "epoch": 0.83, "learning_rate": 1.4321141956666407e-05, "loss": 0.7253, "step": 14680 }, { "epoch": 0.83, "learning_rate": 1.431166961131859e-05, "loss": 0.64, "step": 14681 }, { "epoch": 0.83, "learning_rate": 1.4302200158176637e-05, "loss": 0.6684, "step": 14682 }, { "epoch": 0.83, "learning_rate": 1.4292733597560192e-05, "loss": 0.7275, "step": 14683 }, { "epoch": 0.83, "learning_rate": 1.4283269929788779e-05, "loss": 0.758, "step": 14684 }, { "epoch": 0.83, "learning_rate": 1.4273809155181794e-05, "loss": 0.6781, "step": 14685 }, { "epoch": 0.83, "learning_rate": 1.4264351274058585e-05, "loss": 0.6973, "step": 14686 }, { "epoch": 0.83, "learning_rate": 1.4254896286738351e-05, "loss": 0.6361, "step": 14687 }, { "epoch": 0.83, "learning_rate": 1.4245444193540247e-05, "loss": 0.6937, "step": 14688 }, { "epoch": 0.83, "learning_rate": 1.4235994994783297e-05, "loss": 0.7096, "step": 14689 }, { "epoch": 0.83, "learning_rate": 1.4226548690786446e-05, "loss": 0.6703, "step": 14690 }, { "epoch": 0.83, "learning_rate": 1.4217105281868514e-05, "loss": 0.6772, "step": 14691 }, { "epoch": 0.83, "learning_rate": 1.4207664768348283e-05, "loss": 0.8147, "step": 14692 }, { "epoch": 0.83, "learning_rate": 1.4198227150544363e-05, "loss": 0.8172, "step": 14693 }, { "epoch": 0.83, "learning_rate": 1.4188792428775278e-05, "loss": 0.7156, "step": 14694 }, { "epoch": 0.83, "learning_rate": 1.4179360603359504e-05, "loss": 0.6498, "step": 14695 }, { "epoch": 0.83, "learning_rate": 1.4169931674615377e-05, "loss": 0.7439, "step": 14696 }, { "epoch": 0.83, "learning_rate": 1.4160505642861144e-05, "loss": 0.7223, "step": 14697 }, { "epoch": 0.83, "learning_rate": 1.4151082508415004e-05, "loss": 0.6923, "step": 14698 }, { "epoch": 0.83, "learning_rate": 1.4141662271594946e-05, "loss": 0.7508, "step": 14699 }, { "epoch": 0.83, "learning_rate": 1.4132244932718985e-05, "loss": 0.7289, "step": 14700 }, { "epoch": 0.83, "learning_rate": 1.4122830492104933e-05, "loss": 0.6749, "step": 14701 }, { "epoch": 0.83, "learning_rate": 1.4113418950070567e-05, "loss": 0.7419, "step": 14702 }, { "epoch": 0.83, "learning_rate": 1.4104010306933557e-05, "loss": 0.68, "step": 14703 }, { "epoch": 0.83, "learning_rate": 1.4094604563011472e-05, "loss": 0.6277, "step": 14704 }, { "epoch": 0.83, "learning_rate": 1.4085201718621798e-05, "loss": 0.5346, "step": 14705 }, { "epoch": 0.83, "learning_rate": 1.4075801774081877e-05, "loss": 0.7014, "step": 14706 }, { "epoch": 0.83, "learning_rate": 1.4066404729708993e-05, "loss": 0.6841, "step": 14707 }, { "epoch": 0.83, "learning_rate": 1.4057010585820307e-05, "loss": 0.7832, "step": 14708 }, { "epoch": 0.83, "learning_rate": 1.404761934273291e-05, "loss": 0.733, "step": 14709 }, { "epoch": 0.83, "learning_rate": 1.403823100076378e-05, "loss": 0.7827, "step": 14710 }, { "epoch": 0.83, "learning_rate": 1.4028845560229787e-05, "loss": 0.6841, "step": 14711 }, { "epoch": 0.83, "learning_rate": 1.4019463021447754e-05, "loss": 0.6733, "step": 14712 }, { "epoch": 0.83, "learning_rate": 1.4010083384734308e-05, "loss": 0.7917, "step": 14713 }, { "epoch": 0.83, "learning_rate": 1.4000706650406093e-05, "loss": 0.6358, "step": 14714 }, { "epoch": 0.83, "learning_rate": 1.3991332818779535e-05, "loss": 0.711, "step": 14715 }, { "epoch": 0.83, "learning_rate": 1.3981961890171069e-05, "loss": 0.8011, "step": 14716 }, { "epoch": 0.83, "learning_rate": 1.3972593864896987e-05, "loss": 0.6633, "step": 14717 }, { "epoch": 0.83, "learning_rate": 1.3963228743273494e-05, "loss": 0.7213, "step": 14718 }, { "epoch": 0.83, "learning_rate": 1.3953866525616643e-05, "loss": 0.6508, "step": 14719 }, { "epoch": 0.83, "learning_rate": 1.3944507212242464e-05, "loss": 0.7262, "step": 14720 }, { "epoch": 0.84, "learning_rate": 1.3935150803466878e-05, "loss": 0.6524, "step": 14721 }, { "epoch": 0.84, "learning_rate": 1.3925797299605647e-05, "loss": 0.7473, "step": 14722 }, { "epoch": 0.84, "learning_rate": 1.3916446700974494e-05, "loss": 0.671, "step": 14723 }, { "epoch": 0.84, "learning_rate": 1.3907099007889024e-05, "loss": 0.7319, "step": 14724 }, { "epoch": 0.84, "learning_rate": 1.3897754220664771e-05, "loss": 0.5846, "step": 14725 }, { "epoch": 0.84, "learning_rate": 1.3888412339617107e-05, "loss": 0.6872, "step": 14726 }, { "epoch": 0.84, "learning_rate": 1.3879073365061356e-05, "loss": 0.6433, "step": 14727 }, { "epoch": 0.84, "learning_rate": 1.3869737297312745e-05, "loss": 0.6993, "step": 14728 }, { "epoch": 0.84, "learning_rate": 1.3860404136686411e-05, "loss": 0.6656, "step": 14729 }, { "epoch": 0.84, "learning_rate": 1.3851073883497312e-05, "loss": 0.741, "step": 14730 }, { "epoch": 0.84, "learning_rate": 1.384174653806044e-05, "loss": 0.7237, "step": 14731 }, { "epoch": 0.84, "learning_rate": 1.3832422100690545e-05, "loss": 0.7109, "step": 14732 }, { "epoch": 0.84, "learning_rate": 1.3823100571702397e-05, "loss": 0.6652, "step": 14733 }, { "epoch": 0.84, "learning_rate": 1.381378195141061e-05, "loss": 0.5784, "step": 14734 }, { "epoch": 0.84, "learning_rate": 1.380446624012971e-05, "loss": 0.7324, "step": 14735 }, { "epoch": 0.84, "learning_rate": 1.3795153438174146e-05, "loss": 0.6199, "step": 14736 }, { "epoch": 0.84, "learning_rate": 1.3785843545858223e-05, "loss": 0.5793, "step": 14737 }, { "epoch": 0.84, "learning_rate": 1.3776536563496201e-05, "loss": 0.6924, "step": 14738 }, { "epoch": 0.84, "learning_rate": 1.3767232491402182e-05, "loss": 0.634, "step": 14739 }, { "epoch": 0.84, "learning_rate": 1.3757931329890216e-05, "loss": 0.6487, "step": 14740 }, { "epoch": 0.84, "learning_rate": 1.3748633079274253e-05, "loss": 0.6928, "step": 14741 }, { "epoch": 0.84, "learning_rate": 1.373933773986812e-05, "loss": 0.6837, "step": 14742 }, { "epoch": 0.84, "learning_rate": 1.3730045311985596e-05, "loss": 0.7394, "step": 14743 }, { "epoch": 0.84, "learning_rate": 1.3720755795940255e-05, "loss": 0.6915, "step": 14744 }, { "epoch": 0.84, "learning_rate": 1.3711469192045723e-05, "loss": 0.7242, "step": 14745 }, { "epoch": 0.84, "learning_rate": 1.3702185500615372e-05, "loss": 0.7409, "step": 14746 }, { "epoch": 0.84, "learning_rate": 1.3692904721962585e-05, "loss": 0.6418, "step": 14747 }, { "epoch": 0.84, "learning_rate": 1.3683626856400612e-05, "loss": 0.7163, "step": 14748 }, { "epoch": 0.84, "learning_rate": 1.3674351904242611e-05, "loss": 0.722, "step": 14749 }, { "epoch": 0.84, "learning_rate": 1.3665079865801644e-05, "loss": 0.6802, "step": 14750 }, { "epoch": 0.84, "learning_rate": 1.3655810741390662e-05, "loss": 0.6589, "step": 14751 }, { "epoch": 0.84, "learning_rate": 1.3646544531322481e-05, "loss": 0.6891, "step": 14752 }, { "epoch": 0.84, "learning_rate": 1.3637281235909904e-05, "loss": 0.7886, "step": 14753 }, { "epoch": 0.84, "learning_rate": 1.3628020855465572e-05, "loss": 0.7832, "step": 14754 }, { "epoch": 0.84, "learning_rate": 1.3618763390302058e-05, "loss": 0.5723, "step": 14755 }, { "epoch": 0.84, "learning_rate": 1.3609508840731821e-05, "loss": 0.7612, "step": 14756 }, { "epoch": 0.84, "learning_rate": 1.3600257207067258e-05, "loss": 0.7628, "step": 14757 }, { "epoch": 0.84, "learning_rate": 1.359100848962057e-05, "loss": 0.6953, "step": 14758 }, { "epoch": 0.84, "learning_rate": 1.3581762688703992e-05, "loss": 0.8021, "step": 14759 }, { "epoch": 0.84, "learning_rate": 1.3572519804629536e-05, "loss": 0.7148, "step": 14760 }, { "epoch": 0.84, "learning_rate": 1.3563279837709209e-05, "loss": 0.7803, "step": 14761 }, { "epoch": 0.84, "learning_rate": 1.3554042788254872e-05, "loss": 0.7459, "step": 14762 }, { "epoch": 0.84, "learning_rate": 1.3544808656578322e-05, "loss": 0.5824, "step": 14763 }, { "epoch": 0.84, "learning_rate": 1.35355774429912e-05, "loss": 0.66, "step": 14764 }, { "epoch": 0.84, "learning_rate": 1.3526349147805095e-05, "loss": 0.7385, "step": 14765 }, { "epoch": 0.84, "learning_rate": 1.351712377133152e-05, "loss": 0.7047, "step": 14766 }, { "epoch": 0.84, "learning_rate": 1.350790131388181e-05, "loss": 0.7883, "step": 14767 }, { "epoch": 0.84, "learning_rate": 1.3498681775767252e-05, "loss": 0.7132, "step": 14768 }, { "epoch": 0.84, "learning_rate": 1.3489465157299053e-05, "loss": 0.7947, "step": 14769 }, { "epoch": 0.84, "learning_rate": 1.3480251458788296e-05, "loss": 0.5865, "step": 14770 }, { "epoch": 0.84, "learning_rate": 1.3471040680545943e-05, "loss": 0.7525, "step": 14771 }, { "epoch": 0.84, "learning_rate": 1.3461832822882903e-05, "loss": 0.6038, "step": 14772 }, { "epoch": 0.84, "learning_rate": 1.3452627886109982e-05, "loss": 0.7167, "step": 14773 }, { "epoch": 0.84, "learning_rate": 1.3443425870537818e-05, "loss": 0.6603, "step": 14774 }, { "epoch": 0.84, "learning_rate": 1.3434226776477043e-05, "loss": 0.717, "step": 14775 }, { "epoch": 0.84, "learning_rate": 1.3425030604238154e-05, "loss": 0.692, "step": 14776 }, { "epoch": 0.84, "learning_rate": 1.3415837354131521e-05, "loss": 0.6193, "step": 14777 }, { "epoch": 0.84, "learning_rate": 1.3406647026467445e-05, "loss": 0.624, "step": 14778 }, { "epoch": 0.84, "learning_rate": 1.339745962155613e-05, "loss": 0.6941, "step": 14779 }, { "epoch": 0.84, "learning_rate": 1.3388275139707685e-05, "loss": 0.7336, "step": 14780 }, { "epoch": 0.84, "learning_rate": 1.3379093581232116e-05, "loss": 0.735, "step": 14781 }, { "epoch": 0.84, "learning_rate": 1.3369914946439288e-05, "loss": 0.6472, "step": 14782 }, { "epoch": 0.84, "learning_rate": 1.3360739235639052e-05, "loss": 0.7991, "step": 14783 }, { "epoch": 0.84, "learning_rate": 1.3351566449141073e-05, "loss": 0.7833, "step": 14784 }, { "epoch": 0.84, "learning_rate": 1.3342396587254958e-05, "loss": 0.6317, "step": 14785 }, { "epoch": 0.84, "learning_rate": 1.3333229650290225e-05, "loss": 0.7656, "step": 14786 }, { "epoch": 0.84, "learning_rate": 1.3324065638556293e-05, "loss": 0.6177, "step": 14787 }, { "epoch": 0.84, "learning_rate": 1.3314904552362473e-05, "loss": 0.6735, "step": 14788 }, { "epoch": 0.84, "learning_rate": 1.3305746392017949e-05, "loss": 0.6761, "step": 14789 }, { "epoch": 0.84, "learning_rate": 1.3296591157831872e-05, "loss": 0.7228, "step": 14790 }, { "epoch": 0.84, "learning_rate": 1.32874388501132e-05, "loss": 0.6842, "step": 14791 }, { "epoch": 0.84, "learning_rate": 1.327828946917089e-05, "loss": 0.6451, "step": 14792 }, { "epoch": 0.84, "learning_rate": 1.3269143015313735e-05, "loss": 0.7487, "step": 14793 }, { "epoch": 0.84, "learning_rate": 1.3259999488850472e-05, "loss": 0.8065, "step": 14794 }, { "epoch": 0.84, "learning_rate": 1.3250858890089735e-05, "loss": 0.6858, "step": 14795 }, { "epoch": 0.84, "learning_rate": 1.3241721219340008e-05, "loss": 0.7665, "step": 14796 }, { "epoch": 0.84, "learning_rate": 1.3232586476909703e-05, "loss": 0.6419, "step": 14797 }, { "epoch": 0.84, "learning_rate": 1.3223454663107172e-05, "loss": 0.6529, "step": 14798 }, { "epoch": 0.84, "learning_rate": 1.321432577824061e-05, "loss": 0.7107, "step": 14799 }, { "epoch": 0.84, "learning_rate": 1.3205199822618164e-05, "loss": 0.7101, "step": 14800 }, { "epoch": 0.84, "learning_rate": 1.3196076796547852e-05, "loss": 0.6777, "step": 14801 }, { "epoch": 0.84, "learning_rate": 1.3186956700337616e-05, "loss": 0.6839, "step": 14802 }, { "epoch": 0.84, "learning_rate": 1.3177839534295277e-05, "loss": 0.7707, "step": 14803 }, { "epoch": 0.84, "learning_rate": 1.3168725298728524e-05, "loss": 0.7018, "step": 14804 }, { "epoch": 0.84, "learning_rate": 1.3159613993945019e-05, "loss": 0.7647, "step": 14805 }, { "epoch": 0.84, "learning_rate": 1.3150505620252286e-05, "loss": 0.8086, "step": 14806 }, { "epoch": 0.84, "learning_rate": 1.3141400177957752e-05, "loss": 0.7315, "step": 14807 }, { "epoch": 0.84, "learning_rate": 1.3132297667368799e-05, "loss": 0.7088, "step": 14808 }, { "epoch": 0.84, "learning_rate": 1.3123198088792576e-05, "loss": 0.6602, "step": 14809 }, { "epoch": 0.84, "learning_rate": 1.3114101442536297e-05, "loss": 0.6461, "step": 14810 }, { "epoch": 0.84, "learning_rate": 1.3105007728906927e-05, "loss": 0.605, "step": 14811 }, { "epoch": 0.84, "learning_rate": 1.3095916948211451e-05, "loss": 0.7193, "step": 14812 }, { "epoch": 0.84, "learning_rate": 1.308682910075668e-05, "loss": 0.7538, "step": 14813 }, { "epoch": 0.84, "learning_rate": 1.3077744186849372e-05, "loss": 0.6423, "step": 14814 }, { "epoch": 0.84, "learning_rate": 1.3068662206796179e-05, "loss": 0.6457, "step": 14815 }, { "epoch": 0.84, "learning_rate": 1.30595831609036e-05, "loss": 0.6935, "step": 14816 }, { "epoch": 0.84, "learning_rate": 1.30505070494781e-05, "loss": 0.7377, "step": 14817 }, { "epoch": 0.84, "learning_rate": 1.3041433872826036e-05, "loss": 0.6298, "step": 14818 }, { "epoch": 0.84, "learning_rate": 1.303236363125362e-05, "loss": 0.7619, "step": 14819 }, { "epoch": 0.84, "learning_rate": 1.3023296325067013e-05, "loss": 0.6783, "step": 14820 }, { "epoch": 0.84, "learning_rate": 1.3014231954572287e-05, "loss": 0.6948, "step": 14821 }, { "epoch": 0.84, "learning_rate": 1.3005170520075328e-05, "loss": 0.6989, "step": 14822 }, { "epoch": 0.84, "learning_rate": 1.2996112021882022e-05, "loss": 0.7285, "step": 14823 }, { "epoch": 0.84, "learning_rate": 1.2987056460298108e-05, "loss": 0.7344, "step": 14824 }, { "epoch": 0.84, "learning_rate": 1.297800383562926e-05, "loss": 0.7152, "step": 14825 }, { "epoch": 0.84, "learning_rate": 1.2968954148180989e-05, "loss": 0.7819, "step": 14826 }, { "epoch": 0.84, "learning_rate": 1.2959907398258764e-05, "loss": 0.6781, "step": 14827 }, { "epoch": 0.84, "learning_rate": 1.2950863586167949e-05, "loss": 0.7981, "step": 14828 }, { "epoch": 0.84, "learning_rate": 1.294182271221377e-05, "loss": 0.7767, "step": 14829 }, { "epoch": 0.84, "learning_rate": 1.293278477670139e-05, "loss": 0.6176, "step": 14830 }, { "epoch": 0.84, "learning_rate": 1.2923749779935868e-05, "loss": 0.6957, "step": 14831 }, { "epoch": 0.84, "learning_rate": 1.2914717722222148e-05, "loss": 0.7057, "step": 14832 }, { "epoch": 0.84, "learning_rate": 1.2905688603865117e-05, "loss": 0.7027, "step": 14833 }, { "epoch": 0.84, "learning_rate": 1.2896662425169493e-05, "loss": 0.707, "step": 14834 }, { "epoch": 0.84, "learning_rate": 1.288763918643997e-05, "loss": 0.7095, "step": 14835 }, { "epoch": 0.84, "learning_rate": 1.2878618887981064e-05, "loss": 0.7464, "step": 14836 }, { "epoch": 0.84, "learning_rate": 1.2869601530097253e-05, "loss": 0.7775, "step": 14837 }, { "epoch": 0.84, "learning_rate": 1.2860587113092903e-05, "loss": 0.7414, "step": 14838 }, { "epoch": 0.84, "learning_rate": 1.2851575637272262e-05, "loss": 0.6617, "step": 14839 }, { "epoch": 0.84, "learning_rate": 1.2842567102939517e-05, "loss": 0.7188, "step": 14840 }, { "epoch": 0.84, "learning_rate": 1.2833561510398729e-05, "loss": 0.7771, "step": 14841 }, { "epoch": 0.84, "learning_rate": 1.2824558859953806e-05, "loss": 0.7006, "step": 14842 }, { "epoch": 0.84, "learning_rate": 1.2815559151908652e-05, "loss": 0.755, "step": 14843 }, { "epoch": 0.84, "learning_rate": 1.2806562386567023e-05, "loss": 0.7636, "step": 14844 }, { "epoch": 0.84, "learning_rate": 1.2797568564232587e-05, "loss": 0.7481, "step": 14845 }, { "epoch": 0.84, "learning_rate": 1.2788577685208902e-05, "loss": 0.6978, "step": 14846 }, { "epoch": 0.84, "learning_rate": 1.2779589749799469e-05, "loss": 0.7455, "step": 14847 }, { "epoch": 0.84, "learning_rate": 1.2770604758307635e-05, "loss": 0.7006, "step": 14848 }, { "epoch": 0.84, "learning_rate": 1.2761622711036626e-05, "loss": 0.771, "step": 14849 }, { "epoch": 0.84, "learning_rate": 1.2752643608289638e-05, "loss": 0.6956, "step": 14850 }, { "epoch": 0.84, "learning_rate": 1.2743667450369757e-05, "loss": 0.7248, "step": 14851 }, { "epoch": 0.84, "learning_rate": 1.2734694237579925e-05, "loss": 0.7191, "step": 14852 }, { "epoch": 0.84, "learning_rate": 1.2725723970223046e-05, "loss": 0.6895, "step": 14853 }, { "epoch": 0.84, "learning_rate": 1.2716756648601857e-05, "loss": 0.6687, "step": 14854 }, { "epoch": 0.84, "learning_rate": 1.2707792273019048e-05, "loss": 0.6494, "step": 14855 }, { "epoch": 0.84, "learning_rate": 1.2698830843777166e-05, "loss": 0.6445, "step": 14856 }, { "epoch": 0.84, "learning_rate": 1.2689872361178701e-05, "loss": 0.7241, "step": 14857 }, { "epoch": 0.84, "learning_rate": 1.2680916825526023e-05, "loss": 0.7525, "step": 14858 }, { "epoch": 0.84, "learning_rate": 1.2671964237121403e-05, "loss": 0.6507, "step": 14859 }, { "epoch": 0.84, "learning_rate": 1.2663014596267043e-05, "loss": 0.6496, "step": 14860 }, { "epoch": 0.84, "learning_rate": 1.2654067903264965e-05, "loss": 0.8033, "step": 14861 }, { "epoch": 0.84, "learning_rate": 1.2645124158417188e-05, "loss": 0.7981, "step": 14862 }, { "epoch": 0.84, "learning_rate": 1.2636183362025544e-05, "loss": 0.9071, "step": 14863 }, { "epoch": 0.84, "learning_rate": 1.262724551439184e-05, "loss": 0.7312, "step": 14864 }, { "epoch": 0.84, "learning_rate": 1.261831061581773e-05, "loss": 0.7444, "step": 14865 }, { "epoch": 0.84, "learning_rate": 1.2609378666604833e-05, "loss": 0.6387, "step": 14866 }, { "epoch": 0.84, "learning_rate": 1.260044966705457e-05, "loss": 0.7063, "step": 14867 }, { "epoch": 0.84, "learning_rate": 1.2591523617468347e-05, "loss": 0.6991, "step": 14868 }, { "epoch": 0.84, "learning_rate": 1.2582600518147447e-05, "loss": 0.6965, "step": 14869 }, { "epoch": 0.84, "learning_rate": 1.2573680369393059e-05, "loss": 0.7456, "step": 14870 }, { "epoch": 0.84, "learning_rate": 1.256476317150621e-05, "loss": 0.6549, "step": 14871 }, { "epoch": 0.84, "learning_rate": 1.2555848924787916e-05, "loss": 0.7303, "step": 14872 }, { "epoch": 0.84, "learning_rate": 1.2546937629539079e-05, "loss": 0.7063, "step": 14873 }, { "epoch": 0.84, "learning_rate": 1.2538029286060426e-05, "loss": 0.7519, "step": 14874 }, { "epoch": 0.84, "learning_rate": 1.2529123894652661e-05, "loss": 0.6623, "step": 14875 }, { "epoch": 0.84, "learning_rate": 1.2520221455616366e-05, "loss": 0.7216, "step": 14876 }, { "epoch": 0.84, "learning_rate": 1.2511321969252043e-05, "loss": 0.7314, "step": 14877 }, { "epoch": 0.84, "learning_rate": 1.2502425435860033e-05, "loss": 0.7291, "step": 14878 }, { "epoch": 0.84, "learning_rate": 1.2493531855740625e-05, "loss": 0.7004, "step": 14879 }, { "epoch": 0.84, "learning_rate": 1.2484641229194049e-05, "loss": 0.8587, "step": 14880 }, { "epoch": 0.84, "learning_rate": 1.247575355652031e-05, "loss": 0.6974, "step": 14881 }, { "epoch": 0.84, "learning_rate": 1.2466868838019441e-05, "loss": 0.7084, "step": 14882 }, { "epoch": 0.84, "learning_rate": 1.2457987073991306e-05, "loss": 0.7307, "step": 14883 }, { "epoch": 0.84, "learning_rate": 1.244910826473572e-05, "loss": 0.7298, "step": 14884 }, { "epoch": 0.84, "learning_rate": 1.2440232410552322e-05, "loss": 0.7554, "step": 14885 }, { "epoch": 0.84, "learning_rate": 1.2431359511740726e-05, "loss": 0.749, "step": 14886 }, { "epoch": 0.84, "learning_rate": 1.2422489568600393e-05, "loss": 0.6183, "step": 14887 }, { "epoch": 0.84, "learning_rate": 1.2413622581430706e-05, "loss": 0.6606, "step": 14888 }, { "epoch": 0.84, "learning_rate": 1.240475855053097e-05, "loss": 0.8148, "step": 14889 }, { "epoch": 0.84, "learning_rate": 1.2395897476200357e-05, "loss": 0.6166, "step": 14890 }, { "epoch": 0.84, "learning_rate": 1.238703935873795e-05, "loss": 0.7619, "step": 14891 }, { "epoch": 0.84, "learning_rate": 1.2378184198442766e-05, "loss": 0.6809, "step": 14892 }, { "epoch": 0.84, "learning_rate": 1.2369331995613665e-05, "loss": 0.7375, "step": 14893 }, { "epoch": 0.84, "learning_rate": 1.2360482750549397e-05, "loss": 0.7569, "step": 14894 }, { "epoch": 0.84, "learning_rate": 1.2351636463548688e-05, "loss": 0.6667, "step": 14895 }, { "epoch": 0.84, "learning_rate": 1.2342793134910103e-05, "loss": 0.7051, "step": 14896 }, { "epoch": 0.85, "learning_rate": 1.2333952764932156e-05, "loss": 0.8415, "step": 14897 }, { "epoch": 0.85, "learning_rate": 1.232511535391322e-05, "loss": 0.7211, "step": 14898 }, { "epoch": 0.85, "learning_rate": 1.231628090215159e-05, "loss": 0.6519, "step": 14899 }, { "epoch": 0.85, "learning_rate": 1.2307449409945404e-05, "loss": 0.7045, "step": 14900 }, { "epoch": 0.85, "learning_rate": 1.2298620877592792e-05, "loss": 0.7738, "step": 14901 }, { "epoch": 0.85, "learning_rate": 1.2289795305391727e-05, "loss": 0.7241, "step": 14902 }, { "epoch": 0.85, "learning_rate": 1.2280972693640103e-05, "loss": 0.6646, "step": 14903 }, { "epoch": 0.85, "learning_rate": 1.2272153042635704e-05, "loss": 0.6768, "step": 14904 }, { "epoch": 0.85, "learning_rate": 1.2263336352676235e-05, "loss": 0.7449, "step": 14905 }, { "epoch": 0.85, "learning_rate": 1.225452262405924e-05, "loss": 0.7036, "step": 14906 }, { "epoch": 0.85, "learning_rate": 1.2245711857082243e-05, "loss": 0.5932, "step": 14907 }, { "epoch": 0.85, "learning_rate": 1.2236904052042598e-05, "loss": 0.7467, "step": 14908 }, { "epoch": 0.85, "learning_rate": 1.2228099209237609e-05, "loss": 0.7651, "step": 14909 }, { "epoch": 0.85, "learning_rate": 1.2219297328964453e-05, "loss": 0.6749, "step": 14910 }, { "epoch": 0.85, "learning_rate": 1.2210498411520255e-05, "loss": 0.7899, "step": 14911 }, { "epoch": 0.85, "learning_rate": 1.2201702457201947e-05, "loss": 0.7128, "step": 14912 }, { "epoch": 0.85, "learning_rate": 1.2192909466306446e-05, "loss": 0.7487, "step": 14913 }, { "epoch": 0.85, "learning_rate": 1.2184119439130548e-05, "loss": 0.5793, "step": 14914 }, { "epoch": 0.85, "learning_rate": 1.2175332375970916e-05, "loss": 0.7673, "step": 14915 }, { "epoch": 0.85, "learning_rate": 1.2166548277124134e-05, "loss": 0.7141, "step": 14916 }, { "epoch": 0.85, "learning_rate": 1.21577671428867e-05, "loss": 0.7072, "step": 14917 }, { "epoch": 0.85, "learning_rate": 1.214898897355502e-05, "loss": 0.7299, "step": 14918 }, { "epoch": 0.85, "learning_rate": 1.2140213769425336e-05, "loss": 0.6196, "step": 14919 }, { "epoch": 0.85, "learning_rate": 1.2131441530793863e-05, "loss": 0.6435, "step": 14920 }, { "epoch": 0.85, "learning_rate": 1.2122672257956691e-05, "loss": 0.7604, "step": 14921 }, { "epoch": 0.85, "learning_rate": 1.2113905951209804e-05, "loss": 0.631, "step": 14922 }, { "epoch": 0.85, "learning_rate": 1.2105142610849062e-05, "loss": 0.6869, "step": 14923 }, { "epoch": 0.85, "learning_rate": 1.2096382237170278e-05, "loss": 0.7832, "step": 14924 }, { "epoch": 0.85, "learning_rate": 1.2087624830469135e-05, "loss": 0.6384, "step": 14925 }, { "epoch": 0.85, "learning_rate": 1.2078870391041209e-05, "loss": 0.7648, "step": 14926 }, { "epoch": 0.85, "learning_rate": 1.2070118919181983e-05, "loss": 0.7131, "step": 14927 }, { "epoch": 0.85, "learning_rate": 1.2061370415186845e-05, "loss": 0.6246, "step": 14928 }, { "epoch": 0.85, "learning_rate": 1.2052624879351104e-05, "loss": 0.7411, "step": 14929 }, { "epoch": 0.85, "learning_rate": 1.2043882311969901e-05, "loss": 0.6933, "step": 14930 }, { "epoch": 0.85, "learning_rate": 1.2035142713338366e-05, "loss": 0.6263, "step": 14931 }, { "epoch": 0.85, "learning_rate": 1.202640608375144e-05, "loss": 0.6979, "step": 14932 }, { "epoch": 0.85, "learning_rate": 1.2017672423504034e-05, "loss": 0.6102, "step": 14933 }, { "epoch": 0.85, "learning_rate": 1.200894173289092e-05, "loss": 0.7222, "step": 14934 }, { "epoch": 0.85, "learning_rate": 1.2000214012206789e-05, "loss": 0.729, "step": 14935 }, { "epoch": 0.85, "learning_rate": 1.1991489261746235e-05, "loss": 0.7252, "step": 14936 }, { "epoch": 0.85, "learning_rate": 1.1982767481803726e-05, "loss": 0.7644, "step": 14937 }, { "epoch": 0.85, "learning_rate": 1.1974048672673655e-05, "loss": 0.7956, "step": 14938 }, { "epoch": 0.85, "learning_rate": 1.1965332834650277e-05, "loss": 0.66, "step": 14939 }, { "epoch": 0.85, "learning_rate": 1.1956619968027804e-05, "loss": 0.665, "step": 14940 }, { "epoch": 0.85, "learning_rate": 1.1947910073100299e-05, "loss": 0.8116, "step": 14941 }, { "epoch": 0.85, "learning_rate": 1.193920315016176e-05, "loss": 0.7438, "step": 14942 }, { "epoch": 0.85, "learning_rate": 1.1930499199506084e-05, "loss": 0.7454, "step": 14943 }, { "epoch": 0.85, "learning_rate": 1.1921798221427038e-05, "loss": 0.8403, "step": 14944 }, { "epoch": 0.85, "learning_rate": 1.1913100216218265e-05, "loss": 0.7389, "step": 14945 }, { "epoch": 0.85, "learning_rate": 1.1904405184173373e-05, "loss": 0.6217, "step": 14946 }, { "epoch": 0.85, "learning_rate": 1.189571312558585e-05, "loss": 0.5491, "step": 14947 }, { "epoch": 0.85, "learning_rate": 1.1887024040749073e-05, "loss": 0.7014, "step": 14948 }, { "epoch": 0.85, "learning_rate": 1.1878337929956318e-05, "loss": 0.7253, "step": 14949 }, { "epoch": 0.85, "learning_rate": 1.1869654793500784e-05, "loss": 0.7415, "step": 14950 }, { "epoch": 0.85, "learning_rate": 1.1860974631675536e-05, "loss": 0.679, "step": 14951 }, { "epoch": 0.85, "learning_rate": 1.185229744477352e-05, "loss": 0.7137, "step": 14952 }, { "epoch": 0.85, "learning_rate": 1.1843623233087642e-05, "loss": 0.673, "step": 14953 }, { "epoch": 0.85, "learning_rate": 1.183495199691068e-05, "loss": 0.7755, "step": 14954 }, { "epoch": 0.85, "learning_rate": 1.1826283736535315e-05, "loss": 0.7143, "step": 14955 }, { "epoch": 0.85, "learning_rate": 1.1817618452254109e-05, "loss": 0.597, "step": 14956 }, { "epoch": 0.85, "learning_rate": 1.1808956144359572e-05, "loss": 0.6931, "step": 14957 }, { "epoch": 0.85, "learning_rate": 1.180029681314404e-05, "loss": 0.66, "step": 14958 }, { "epoch": 0.85, "learning_rate": 1.1791640458899811e-05, "loss": 0.698, "step": 14959 }, { "epoch": 0.85, "learning_rate": 1.178298708191904e-05, "loss": 0.6828, "step": 14960 }, { "epoch": 0.85, "learning_rate": 1.1774336682493802e-05, "loss": 0.6577, "step": 14961 }, { "epoch": 0.85, "learning_rate": 1.176568926091609e-05, "loss": 0.7021, "step": 14962 }, { "epoch": 0.85, "learning_rate": 1.175704481747779e-05, "loss": 0.619, "step": 14963 }, { "epoch": 0.85, "learning_rate": 1.1748403352470627e-05, "loss": 0.7703, "step": 14964 }, { "epoch": 0.85, "learning_rate": 1.173976486618631e-05, "loss": 0.6552, "step": 14965 }, { "epoch": 0.85, "learning_rate": 1.1731129358916404e-05, "loss": 0.7503, "step": 14966 }, { "epoch": 0.85, "learning_rate": 1.1722496830952368e-05, "loss": 0.7736, "step": 14967 }, { "epoch": 0.85, "learning_rate": 1.1713867282585578e-05, "loss": 0.6548, "step": 14968 }, { "epoch": 0.85, "learning_rate": 1.1705240714107302e-05, "loss": 0.7306, "step": 14969 }, { "epoch": 0.85, "learning_rate": 1.169661712580874e-05, "loss": 0.6635, "step": 14970 }, { "epoch": 0.85, "learning_rate": 1.1687996517980905e-05, "loss": 0.7774, "step": 14971 }, { "epoch": 0.85, "learning_rate": 1.1679378890914794e-05, "loss": 0.67, "step": 14972 }, { "epoch": 0.85, "learning_rate": 1.16707642449013e-05, "loss": 0.7084, "step": 14973 }, { "epoch": 0.85, "learning_rate": 1.1662152580231145e-05, "loss": 0.6547, "step": 14974 }, { "epoch": 0.85, "learning_rate": 1.1653543897195018e-05, "loss": 0.7453, "step": 14975 }, { "epoch": 0.85, "learning_rate": 1.1644938196083499e-05, "loss": 0.7442, "step": 14976 }, { "epoch": 0.85, "learning_rate": 1.163633547718701e-05, "loss": 0.7574, "step": 14977 }, { "epoch": 0.85, "learning_rate": 1.1627735740795954e-05, "loss": 0.7468, "step": 14978 }, { "epoch": 0.85, "learning_rate": 1.1619138987200562e-05, "loss": 0.7152, "step": 14979 }, { "epoch": 0.85, "learning_rate": 1.1610545216691026e-05, "loss": 0.7333, "step": 14980 }, { "epoch": 0.85, "learning_rate": 1.1601954429557427e-05, "loss": 0.6526, "step": 14981 }, { "epoch": 0.85, "learning_rate": 1.1593366626089675e-05, "loss": 0.6997, "step": 14982 }, { "epoch": 0.85, "learning_rate": 1.158478180657766e-05, "loss": 0.7065, "step": 14983 }, { "epoch": 0.85, "learning_rate": 1.157619997131113e-05, "loss": 0.7286, "step": 14984 }, { "epoch": 0.85, "learning_rate": 1.1567621120579753e-05, "loss": 0.7205, "step": 14985 }, { "epoch": 0.85, "learning_rate": 1.1559045254673073e-05, "loss": 0.7668, "step": 14986 }, { "epoch": 0.85, "learning_rate": 1.1550472373880572e-05, "loss": 0.6842, "step": 14987 }, { "epoch": 0.85, "learning_rate": 1.1541902478491606e-05, "loss": 0.7503, "step": 14988 }, { "epoch": 0.85, "learning_rate": 1.1533335568795412e-05, "loss": 0.6866, "step": 14989 }, { "epoch": 0.85, "learning_rate": 1.1524771645081167e-05, "loss": 0.743, "step": 14990 }, { "epoch": 0.85, "learning_rate": 1.1516210707637898e-05, "loss": 0.6351, "step": 14991 }, { "epoch": 0.85, "learning_rate": 1.1507652756754572e-05, "loss": 0.6949, "step": 14992 }, { "epoch": 0.85, "learning_rate": 1.1499097792720048e-05, "loss": 0.6689, "step": 14993 }, { "epoch": 0.85, "learning_rate": 1.1490545815823072e-05, "loss": 0.6878, "step": 14994 }, { "epoch": 0.85, "learning_rate": 1.1481996826352326e-05, "loss": 0.6639, "step": 14995 }, { "epoch": 0.85, "learning_rate": 1.1473450824596333e-05, "loss": 0.7268, "step": 14996 }, { "epoch": 0.85, "learning_rate": 1.1464907810843517e-05, "loss": 0.725, "step": 14997 }, { "epoch": 0.85, "learning_rate": 1.145636778538226e-05, "loss": 0.7177, "step": 14998 }, { "epoch": 0.85, "learning_rate": 1.1447830748500799e-05, "loss": 0.7508, "step": 14999 }, { "epoch": 0.85, "learning_rate": 1.1439296700487289e-05, "loss": 0.6821, "step": 15000 }, { "epoch": 0.85, "learning_rate": 1.143076564162977e-05, "loss": 0.7269, "step": 15001 }, { "epoch": 0.85, "learning_rate": 1.1422237572216221e-05, "loss": 0.6997, "step": 15002 }, { "epoch": 0.85, "learning_rate": 1.1413712492534456e-05, "loss": 0.6088, "step": 15003 }, { "epoch": 0.85, "learning_rate": 1.1405190402872202e-05, "loss": 0.756, "step": 15004 }, { "epoch": 0.85, "learning_rate": 1.1396671303517114e-05, "loss": 0.7324, "step": 15005 }, { "epoch": 0.85, "learning_rate": 1.1388155194756745e-05, "loss": 0.6963, "step": 15006 }, { "epoch": 0.85, "learning_rate": 1.1379642076878527e-05, "loss": 0.742, "step": 15007 }, { "epoch": 0.85, "learning_rate": 1.1371131950169822e-05, "loss": 0.7225, "step": 15008 }, { "epoch": 0.85, "learning_rate": 1.1362624814917843e-05, "loss": 0.7401, "step": 15009 }, { "epoch": 0.85, "learning_rate": 1.135412067140974e-05, "loss": 0.7415, "step": 15010 }, { "epoch": 0.85, "learning_rate": 1.1345619519932537e-05, "loss": 0.6999, "step": 15011 }, { "epoch": 0.85, "learning_rate": 1.1337121360773183e-05, "loss": 0.7227, "step": 15012 }, { "epoch": 0.85, "learning_rate": 1.1328626194218505e-05, "loss": 0.699, "step": 15013 }, { "epoch": 0.85, "learning_rate": 1.132013402055524e-05, "loss": 0.6685, "step": 15014 }, { "epoch": 0.85, "learning_rate": 1.1311644840070046e-05, "loss": 0.6611, "step": 15015 }, { "epoch": 0.85, "learning_rate": 1.1303158653049418e-05, "loss": 0.7234, "step": 15016 }, { "epoch": 0.85, "learning_rate": 1.1294675459779791e-05, "loss": 0.6483, "step": 15017 }, { "epoch": 0.85, "learning_rate": 1.1286195260547538e-05, "loss": 0.5989, "step": 15018 }, { "epoch": 0.85, "learning_rate": 1.1277718055638819e-05, "loss": 0.7557, "step": 15019 }, { "epoch": 0.85, "learning_rate": 1.1269243845339816e-05, "loss": 0.7289, "step": 15020 }, { "epoch": 0.85, "learning_rate": 1.1260772629936545e-05, "loss": 0.76, "step": 15021 }, { "epoch": 0.85, "learning_rate": 1.1252304409714909e-05, "loss": 0.6575, "step": 15022 }, { "epoch": 0.85, "learning_rate": 1.1243839184960748e-05, "loss": 0.7589, "step": 15023 }, { "epoch": 0.85, "learning_rate": 1.1235376955959787e-05, "loss": 0.685, "step": 15024 }, { "epoch": 0.85, "learning_rate": 1.1226917722997665e-05, "loss": 0.6218, "step": 15025 }, { "epoch": 0.85, "learning_rate": 1.1218461486359877e-05, "loss": 0.6258, "step": 15026 }, { "epoch": 0.85, "learning_rate": 1.1210008246331838e-05, "loss": 0.7503, "step": 15027 }, { "epoch": 0.85, "learning_rate": 1.120155800319891e-05, "loss": 0.8169, "step": 15028 }, { "epoch": 0.85, "learning_rate": 1.119311075724625e-05, "loss": 0.7238, "step": 15029 }, { "epoch": 0.85, "learning_rate": 1.1184666508759024e-05, "loss": 0.742, "step": 15030 }, { "epoch": 0.85, "learning_rate": 1.117622525802222e-05, "loss": 0.7428, "step": 15031 }, { "epoch": 0.85, "learning_rate": 1.1167787005320762e-05, "loss": 0.68, "step": 15032 }, { "epoch": 0.85, "learning_rate": 1.1159351750939485e-05, "loss": 0.6968, "step": 15033 }, { "epoch": 0.85, "learning_rate": 1.115091949516306e-05, "loss": 0.6674, "step": 15034 }, { "epoch": 0.85, "learning_rate": 1.1142490238276138e-05, "loss": 0.688, "step": 15035 }, { "epoch": 0.85, "learning_rate": 1.1134063980563181e-05, "loss": 0.7228, "step": 15036 }, { "epoch": 0.85, "learning_rate": 1.1125640722308628e-05, "loss": 0.7124, "step": 15037 }, { "epoch": 0.85, "learning_rate": 1.1117220463796786e-05, "loss": 0.7112, "step": 15038 }, { "epoch": 0.85, "learning_rate": 1.1108803205311858e-05, "loss": 0.7986, "step": 15039 }, { "epoch": 0.85, "learning_rate": 1.1100388947137962e-05, "loss": 0.728, "step": 15040 }, { "epoch": 0.85, "learning_rate": 1.1091977689559085e-05, "loss": 0.74, "step": 15041 }, { "epoch": 0.85, "learning_rate": 1.108356943285912e-05, "loss": 0.8237, "step": 15042 }, { "epoch": 0.85, "learning_rate": 1.1075164177321874e-05, "loss": 0.7052, "step": 15043 }, { "epoch": 0.85, "learning_rate": 1.106676192323105e-05, "loss": 0.7667, "step": 15044 }, { "epoch": 0.85, "learning_rate": 1.1058362670870249e-05, "loss": 0.7453, "step": 15045 }, { "epoch": 0.85, "learning_rate": 1.1049966420522962e-05, "loss": 0.6702, "step": 15046 }, { "epoch": 0.85, "learning_rate": 1.1041573172472607e-05, "loss": 0.6177, "step": 15047 }, { "epoch": 0.85, "learning_rate": 1.1033182927002462e-05, "loss": 0.6708, "step": 15048 }, { "epoch": 0.85, "learning_rate": 1.1024795684395694e-05, "loss": 0.6593, "step": 15049 }, { "epoch": 0.85, "learning_rate": 1.1016411444935426e-05, "loss": 0.7139, "step": 15050 }, { "epoch": 0.85, "learning_rate": 1.1008030208904641e-05, "loss": 0.7032, "step": 15051 }, { "epoch": 0.85, "learning_rate": 1.0999651976586211e-05, "loss": 0.7197, "step": 15052 }, { "epoch": 0.85, "learning_rate": 1.0991276748262968e-05, "loss": 0.6371, "step": 15053 }, { "epoch": 0.85, "learning_rate": 1.0982904524217551e-05, "loss": 0.6961, "step": 15054 }, { "epoch": 0.85, "learning_rate": 1.097453530473258e-05, "loss": 0.7128, "step": 15055 }, { "epoch": 0.85, "learning_rate": 1.0966169090090506e-05, "loss": 0.5895, "step": 15056 }, { "epoch": 0.85, "learning_rate": 1.0957805880573723e-05, "loss": 0.7044, "step": 15057 }, { "epoch": 0.85, "learning_rate": 1.0949445676464508e-05, "loss": 0.6682, "step": 15058 }, { "epoch": 0.85, "learning_rate": 1.0941088478045048e-05, "loss": 0.6621, "step": 15059 }, { "epoch": 0.85, "learning_rate": 1.0932734285597435e-05, "loss": 0.7441, "step": 15060 }, { "epoch": 0.85, "learning_rate": 1.0924383099403613e-05, "loss": 0.712, "step": 15061 }, { "epoch": 0.85, "learning_rate": 1.0916034919745477e-05, "loss": 0.7132, "step": 15062 }, { "epoch": 0.85, "learning_rate": 1.0907689746904792e-05, "loss": 0.7309, "step": 15063 }, { "epoch": 0.85, "learning_rate": 1.0899347581163221e-05, "loss": 0.6834, "step": 15064 }, { "epoch": 0.85, "learning_rate": 1.089100842280234e-05, "loss": 0.6425, "step": 15065 }, { "epoch": 0.85, "learning_rate": 1.0882672272103656e-05, "loss": 0.7867, "step": 15066 }, { "epoch": 0.85, "learning_rate": 1.0874339129348465e-05, "loss": 0.6096, "step": 15067 }, { "epoch": 0.85, "learning_rate": 1.0866008994818078e-05, "loss": 0.6831, "step": 15068 }, { "epoch": 0.85, "learning_rate": 1.0857681868793646e-05, "loss": 0.5941, "step": 15069 }, { "epoch": 0.85, "learning_rate": 1.0849357751556266e-05, "loss": 0.6837, "step": 15070 }, { "epoch": 0.85, "learning_rate": 1.0841036643386837e-05, "loss": 0.7166, "step": 15071 }, { "epoch": 0.85, "learning_rate": 1.0832718544566257e-05, "loss": 0.7081, "step": 15072 }, { "epoch": 0.86, "learning_rate": 1.0824403455375288e-05, "loss": 0.7011, "step": 15073 }, { "epoch": 0.86, "learning_rate": 1.081609137609455e-05, "loss": 0.7226, "step": 15074 }, { "epoch": 0.86, "learning_rate": 1.0807782307004633e-05, "loss": 0.6434, "step": 15075 }, { "epoch": 0.86, "learning_rate": 1.0799476248385976e-05, "loss": 0.7912, "step": 15076 }, { "epoch": 0.86, "learning_rate": 1.0791173200518945e-05, "loss": 0.6304, "step": 15077 }, { "epoch": 0.86, "learning_rate": 1.0782873163683771e-05, "loss": 0.7643, "step": 15078 }, { "epoch": 0.86, "learning_rate": 1.0774576138160597e-05, "loss": 0.6646, "step": 15079 }, { "epoch": 0.86, "learning_rate": 1.0766282124229499e-05, "loss": 0.744, "step": 15080 }, { "epoch": 0.86, "learning_rate": 1.0757991122170396e-05, "loss": 0.7617, "step": 15081 }, { "epoch": 0.86, "learning_rate": 1.0749703132263122e-05, "loss": 0.7119, "step": 15082 }, { "epoch": 0.86, "learning_rate": 1.0741418154787442e-05, "loss": 0.6577, "step": 15083 }, { "epoch": 0.86, "learning_rate": 1.0733136190022996e-05, "loss": 0.8125, "step": 15084 }, { "epoch": 0.86, "learning_rate": 1.0724857238249331e-05, "loss": 0.6598, "step": 15085 }, { "epoch": 0.86, "learning_rate": 1.0716581299745854e-05, "loss": 0.737, "step": 15086 }, { "epoch": 0.86, "learning_rate": 1.0708308374791909e-05, "loss": 0.7053, "step": 15087 }, { "epoch": 0.86, "learning_rate": 1.0700038463666728e-05, "loss": 0.6785, "step": 15088 }, { "epoch": 0.86, "learning_rate": 1.0691771566649445e-05, "loss": 0.7787, "step": 15089 }, { "epoch": 0.86, "learning_rate": 1.068350768401909e-05, "loss": 0.6147, "step": 15090 }, { "epoch": 0.86, "learning_rate": 1.0675246816054586e-05, "loss": 0.6859, "step": 15091 }, { "epoch": 0.86, "learning_rate": 1.0666988963034796e-05, "loss": 0.6178, "step": 15092 }, { "epoch": 0.86, "learning_rate": 1.0658734125238402e-05, "loss": 0.7068, "step": 15093 }, { "epoch": 0.86, "learning_rate": 1.065048230294402e-05, "loss": 0.6859, "step": 15094 }, { "epoch": 0.86, "learning_rate": 1.0642233496430198e-05, "loss": 0.6554, "step": 15095 }, { "epoch": 0.86, "learning_rate": 1.0633987705975334e-05, "loss": 0.6543, "step": 15096 }, { "epoch": 0.86, "learning_rate": 1.0625744931857762e-05, "loss": 0.7459, "step": 15097 }, { "epoch": 0.86, "learning_rate": 1.0617505174355713e-05, "loss": 0.82, "step": 15098 }, { "epoch": 0.86, "learning_rate": 1.0609268433747266e-05, "loss": 0.6739, "step": 15099 }, { "epoch": 0.86, "learning_rate": 1.060103471031043e-05, "loss": 0.6712, "step": 15100 }, { "epoch": 0.86, "learning_rate": 1.0592804004323143e-05, "loss": 0.7079, "step": 15101 }, { "epoch": 0.86, "learning_rate": 1.0584576316063188e-05, "loss": 0.6601, "step": 15102 }, { "epoch": 0.86, "learning_rate": 1.057635164580828e-05, "loss": 0.6754, "step": 15103 }, { "epoch": 0.86, "learning_rate": 1.056812999383604e-05, "loss": 0.811, "step": 15104 }, { "epoch": 0.86, "learning_rate": 1.0559911360423958e-05, "loss": 0.6846, "step": 15105 }, { "epoch": 0.86, "learning_rate": 1.0551695745849433e-05, "loss": 0.6204, "step": 15106 }, { "epoch": 0.86, "learning_rate": 1.0543483150389766e-05, "loss": 0.7534, "step": 15107 }, { "epoch": 0.86, "learning_rate": 1.0535273574322135e-05, "loss": 0.6322, "step": 15108 }, { "epoch": 0.86, "learning_rate": 1.0527067017923654e-05, "loss": 0.7358, "step": 15109 }, { "epoch": 0.86, "learning_rate": 1.0518863481471319e-05, "loss": 0.617, "step": 15110 }, { "epoch": 0.86, "learning_rate": 1.0510662965242025e-05, "loss": 0.6949, "step": 15111 }, { "epoch": 0.86, "learning_rate": 1.0502465469512535e-05, "loss": 0.6859, "step": 15112 }, { "epoch": 0.86, "learning_rate": 1.049427099455954e-05, "loss": 0.7306, "step": 15113 }, { "epoch": 0.86, "learning_rate": 1.0486079540659665e-05, "loss": 0.739, "step": 15114 }, { "epoch": 0.86, "learning_rate": 1.0477891108089345e-05, "loss": 0.7064, "step": 15115 }, { "epoch": 0.86, "learning_rate": 1.0469705697124988e-05, "loss": 0.6968, "step": 15116 }, { "epoch": 0.86, "learning_rate": 1.0461523308042853e-05, "loss": 0.6867, "step": 15117 }, { "epoch": 0.86, "learning_rate": 1.0453343941119154e-05, "loss": 0.7323, "step": 15118 }, { "epoch": 0.86, "learning_rate": 1.0445167596629924e-05, "loss": 0.7768, "step": 15119 }, { "epoch": 0.86, "learning_rate": 1.0436994274851153e-05, "loss": 0.7397, "step": 15120 }, { "epoch": 0.86, "learning_rate": 1.042882397605871e-05, "loss": 0.6762, "step": 15121 }, { "epoch": 0.86, "learning_rate": 1.0420656700528387e-05, "loss": 0.6814, "step": 15122 }, { "epoch": 0.86, "learning_rate": 1.0412492448535815e-05, "loss": 0.6701, "step": 15123 }, { "epoch": 0.86, "learning_rate": 1.0404331220356568e-05, "loss": 0.726, "step": 15124 }, { "epoch": 0.86, "learning_rate": 1.0396173016266141e-05, "loss": 0.7035, "step": 15125 }, { "epoch": 0.86, "learning_rate": 1.0388017836539843e-05, "loss": 0.6839, "step": 15126 }, { "epoch": 0.86, "learning_rate": 1.0379865681452971e-05, "loss": 0.7459, "step": 15127 }, { "epoch": 0.86, "learning_rate": 1.0371716551280663e-05, "loss": 0.7287, "step": 15128 }, { "epoch": 0.86, "learning_rate": 1.0363570446297999e-05, "loss": 0.6975, "step": 15129 }, { "epoch": 0.86, "learning_rate": 1.035542736677989e-05, "loss": 0.7681, "step": 15130 }, { "epoch": 0.86, "learning_rate": 1.034728731300123e-05, "loss": 0.5679, "step": 15131 }, { "epoch": 0.86, "learning_rate": 1.0339150285236731e-05, "loss": 0.7526, "step": 15132 }, { "epoch": 0.86, "learning_rate": 1.0331016283761052e-05, "loss": 0.6605, "step": 15133 }, { "epoch": 0.86, "learning_rate": 1.032288530884874e-05, "loss": 0.6628, "step": 15134 }, { "epoch": 0.86, "learning_rate": 1.031475736077423e-05, "loss": 0.7798, "step": 15135 }, { "epoch": 0.86, "learning_rate": 1.0306632439811891e-05, "loss": 0.5614, "step": 15136 }, { "epoch": 0.86, "learning_rate": 1.0298510546235917e-05, "loss": 0.6587, "step": 15137 }, { "epoch": 0.86, "learning_rate": 1.0290391680320478e-05, "loss": 0.718, "step": 15138 }, { "epoch": 0.86, "learning_rate": 1.0282275842339583e-05, "loss": 0.7429, "step": 15139 }, { "epoch": 0.86, "learning_rate": 1.0274163032567163e-05, "loss": 0.7061, "step": 15140 }, { "epoch": 0.86, "learning_rate": 1.0266053251277053e-05, "loss": 0.7002, "step": 15141 }, { "epoch": 0.86, "learning_rate": 1.0257946498742987e-05, "loss": 0.7304, "step": 15142 }, { "epoch": 0.86, "learning_rate": 1.0249842775238605e-05, "loss": 0.5879, "step": 15143 }, { "epoch": 0.86, "learning_rate": 1.024174208103741e-05, "loss": 0.6652, "step": 15144 }, { "epoch": 0.86, "learning_rate": 1.0233644416412791e-05, "loss": 0.6239, "step": 15145 }, { "epoch": 0.86, "learning_rate": 1.02255497816381e-05, "loss": 0.7014, "step": 15146 }, { "epoch": 0.86, "learning_rate": 1.0217458176986538e-05, "loss": 0.6702, "step": 15147 }, { "epoch": 0.86, "learning_rate": 1.0209369602731233e-05, "loss": 0.6579, "step": 15148 }, { "epoch": 0.86, "learning_rate": 1.0201284059145188e-05, "loss": 0.6726, "step": 15149 }, { "epoch": 0.86, "learning_rate": 1.019320154650133e-05, "loss": 0.7063, "step": 15150 }, { "epoch": 0.86, "learning_rate": 1.018512206507245e-05, "loss": 0.6892, "step": 15151 }, { "epoch": 0.86, "learning_rate": 1.0177045615131219e-05, "loss": 0.6839, "step": 15152 }, { "epoch": 0.86, "learning_rate": 1.0168972196950277e-05, "loss": 0.7283, "step": 15153 }, { "epoch": 0.86, "learning_rate": 1.0160901810802115e-05, "loss": 0.738, "step": 15154 }, { "epoch": 0.86, "learning_rate": 1.0152834456959127e-05, "loss": 0.7024, "step": 15155 }, { "epoch": 0.86, "learning_rate": 1.0144770135693627e-05, "loss": 0.5998, "step": 15156 }, { "epoch": 0.86, "learning_rate": 1.0136708847277764e-05, "loss": 0.6924, "step": 15157 }, { "epoch": 0.86, "learning_rate": 1.0128650591983658e-05, "loss": 0.6704, "step": 15158 }, { "epoch": 0.86, "learning_rate": 1.0120595370083318e-05, "loss": 0.6852, "step": 15159 }, { "epoch": 0.86, "learning_rate": 1.0112543181848566e-05, "loss": 0.6906, "step": 15160 }, { "epoch": 0.86, "learning_rate": 1.0104494027551237e-05, "loss": 0.7723, "step": 15161 }, { "epoch": 0.86, "learning_rate": 1.009644790746298e-05, "loss": 0.6587, "step": 15162 }, { "epoch": 0.86, "learning_rate": 1.0088404821855412e-05, "loss": 0.7218, "step": 15163 }, { "epoch": 0.86, "learning_rate": 1.0080364770999962e-05, "loss": 0.7147, "step": 15164 }, { "epoch": 0.86, "learning_rate": 1.0072327755168032e-05, "loss": 0.7247, "step": 15165 }, { "epoch": 0.86, "learning_rate": 1.0064293774630896e-05, "loss": 0.6842, "step": 15166 }, { "epoch": 0.86, "learning_rate": 1.005626282965969e-05, "loss": 0.6168, "step": 15167 }, { "epoch": 0.86, "learning_rate": 1.0048234920525501e-05, "loss": 0.7057, "step": 15168 }, { "epoch": 0.86, "learning_rate": 1.0040210047499288e-05, "loss": 0.6836, "step": 15169 }, { "epoch": 0.86, "learning_rate": 1.0032188210851934e-05, "loss": 0.6532, "step": 15170 }, { "epoch": 0.86, "learning_rate": 1.0024169410854168e-05, "loss": 0.6845, "step": 15171 }, { "epoch": 0.86, "learning_rate": 1.0016153647776638e-05, "loss": 0.6654, "step": 15172 }, { "epoch": 0.86, "learning_rate": 1.000814092188994e-05, "loss": 0.6506, "step": 15173 }, { "epoch": 0.86, "learning_rate": 1.0000131233464482e-05, "loss": 0.6904, "step": 15174 }, { "epoch": 0.86, "learning_rate": 9.992124582770634e-06, "loss": 0.7379, "step": 15175 }, { "epoch": 0.86, "learning_rate": 9.984120970078648e-06, "loss": 0.6571, "step": 15176 }, { "epoch": 0.86, "learning_rate": 9.976120395658628e-06, "loss": 0.6339, "step": 15177 }, { "epoch": 0.86, "learning_rate": 9.968122859780648e-06, "loss": 0.741, "step": 15178 }, { "epoch": 0.86, "learning_rate": 9.960128362714637e-06, "loss": 0.7146, "step": 15179 }, { "epoch": 0.86, "learning_rate": 9.952136904730435e-06, "loss": 0.6872, "step": 15180 }, { "epoch": 0.86, "learning_rate": 9.944148486097793e-06, "loss": 0.6331, "step": 15181 }, { "epoch": 0.86, "learning_rate": 9.936163107086293e-06, "loss": 0.7002, "step": 15182 }, { "epoch": 0.86, "learning_rate": 9.928180767965523e-06, "loss": 0.7252, "step": 15183 }, { "epoch": 0.86, "learning_rate": 9.920201469004841e-06, "loss": 0.5754, "step": 15184 }, { "epoch": 0.86, "learning_rate": 9.912225210473603e-06, "loss": 0.6704, "step": 15185 }, { "epoch": 0.86, "learning_rate": 9.904251992641033e-06, "loss": 0.849, "step": 15186 }, { "epoch": 0.86, "learning_rate": 9.896281815776242e-06, "loss": 0.6939, "step": 15187 }, { "epoch": 0.86, "learning_rate": 9.88831468014826e-06, "loss": 0.7699, "step": 15188 }, { "epoch": 0.86, "learning_rate": 9.880350586025988e-06, "loss": 0.5948, "step": 15189 }, { "epoch": 0.86, "learning_rate": 9.872389533678206e-06, "loss": 0.7455, "step": 15190 }, { "epoch": 0.86, "learning_rate": 9.864431523373652e-06, "loss": 0.7519, "step": 15191 }, { "epoch": 0.86, "learning_rate": 9.856476555380922e-06, "loss": 0.724, "step": 15192 }, { "epoch": 0.86, "learning_rate": 9.848524629968525e-06, "loss": 0.7474, "step": 15193 }, { "epoch": 0.86, "learning_rate": 9.840575747404846e-06, "loss": 0.6821, "step": 15194 }, { "epoch": 0.86, "learning_rate": 9.832629907958202e-06, "loss": 0.6984, "step": 15195 }, { "epoch": 0.86, "learning_rate": 9.82468711189678e-06, "loss": 0.8115, "step": 15196 }, { "epoch": 0.86, "learning_rate": 9.816747359488632e-06, "loss": 0.7019, "step": 15197 }, { "epoch": 0.86, "learning_rate": 9.808810651001787e-06, "loss": 0.7253, "step": 15198 }, { "epoch": 0.86, "learning_rate": 9.80087698670411e-06, "loss": 0.7528, "step": 15199 }, { "epoch": 0.86, "learning_rate": 9.792946366863386e-06, "loss": 0.7217, "step": 15200 }, { "epoch": 0.86, "learning_rate": 9.785018791747325e-06, "loss": 0.7506, "step": 15201 }, { "epoch": 0.86, "learning_rate": 9.777094261623465e-06, "loss": 0.6877, "step": 15202 }, { "epoch": 0.86, "learning_rate": 9.769172776759294e-06, "loss": 0.7355, "step": 15203 }, { "epoch": 0.86, "learning_rate": 9.761254337422176e-06, "loss": 0.5827, "step": 15204 }, { "epoch": 0.86, "learning_rate": 9.753338943879375e-06, "loss": 0.731, "step": 15205 }, { "epoch": 0.86, "learning_rate": 9.745426596398066e-06, "loss": 0.6505, "step": 15206 }, { "epoch": 0.86, "learning_rate": 9.737517295245314e-06, "loss": 0.6938, "step": 15207 }, { "epoch": 0.86, "learning_rate": 9.729611040688103e-06, "loss": 0.7988, "step": 15208 }, { "epoch": 0.86, "learning_rate": 9.721707832993232e-06, "loss": 0.6852, "step": 15209 }, { "epoch": 0.86, "learning_rate": 9.7138076724275e-06, "loss": 0.7006, "step": 15210 }, { "epoch": 0.86, "learning_rate": 9.705910559257536e-06, "loss": 0.526, "step": 15211 }, { "epoch": 0.86, "learning_rate": 9.698016493749884e-06, "loss": 0.7063, "step": 15212 }, { "epoch": 0.86, "learning_rate": 9.690125476171009e-06, "loss": 0.7614, "step": 15213 }, { "epoch": 0.86, "learning_rate": 9.682237506787239e-06, "loss": 0.6559, "step": 15214 }, { "epoch": 0.86, "learning_rate": 9.674352585864855e-06, "loss": 0.7824, "step": 15215 }, { "epoch": 0.86, "learning_rate": 9.666470713669918e-06, "loss": 0.6746, "step": 15216 }, { "epoch": 0.86, "learning_rate": 9.658591890468515e-06, "loss": 0.6998, "step": 15217 }, { "epoch": 0.86, "learning_rate": 9.650716116526581e-06, "loss": 0.6768, "step": 15218 }, { "epoch": 0.86, "learning_rate": 9.642843392109913e-06, "loss": 0.6197, "step": 15219 }, { "epoch": 0.86, "learning_rate": 9.634973717484253e-06, "loss": 0.8084, "step": 15220 }, { "epoch": 0.86, "learning_rate": 9.627107092915232e-06, "loss": 0.76, "step": 15221 }, { "epoch": 0.86, "learning_rate": 9.619243518668342e-06, "loss": 0.717, "step": 15222 }, { "epoch": 0.86, "learning_rate": 9.61138299500901e-06, "loss": 0.6994, "step": 15223 }, { "epoch": 0.86, "learning_rate": 9.603525522202562e-06, "loss": 0.6031, "step": 15224 }, { "epoch": 0.86, "learning_rate": 9.595671100514214e-06, "loss": 0.6471, "step": 15225 }, { "epoch": 0.86, "learning_rate": 9.587819730209047e-06, "loss": 0.7016, "step": 15226 }, { "epoch": 0.86, "learning_rate": 9.57997141155208e-06, "loss": 0.7415, "step": 15227 }, { "epoch": 0.86, "learning_rate": 9.572126144808225e-06, "loss": 0.7993, "step": 15228 }, { "epoch": 0.86, "learning_rate": 9.564283930242257e-06, "loss": 0.7212, "step": 15229 }, { "epoch": 0.86, "learning_rate": 9.556444768118878e-06, "loss": 0.7407, "step": 15230 }, { "epoch": 0.86, "learning_rate": 9.548608658702685e-06, "loss": 0.6314, "step": 15231 }, { "epoch": 0.86, "learning_rate": 9.540775602258168e-06, "loss": 0.7042, "step": 15232 }, { "epoch": 0.86, "learning_rate": 9.532945599049737e-06, "loss": 0.7071, "step": 15233 }, { "epoch": 0.86, "learning_rate": 9.525118649341646e-06, "loss": 0.8662, "step": 15234 }, { "epoch": 0.86, "learning_rate": 9.517294753398064e-06, "loss": 0.6992, "step": 15235 }, { "epoch": 0.86, "learning_rate": 9.50947391148308e-06, "loss": 0.7239, "step": 15236 }, { "epoch": 0.86, "learning_rate": 9.50165612386067e-06, "loss": 0.6067, "step": 15237 }, { "epoch": 0.86, "learning_rate": 9.493841390794699e-06, "loss": 0.6334, "step": 15238 }, { "epoch": 0.86, "learning_rate": 9.48602971254896e-06, "loss": 0.737, "step": 15239 }, { "epoch": 0.86, "learning_rate": 9.478221089387107e-06, "loss": 0.7189, "step": 15240 }, { "epoch": 0.86, "learning_rate": 9.470415521572695e-06, "loss": 0.6357, "step": 15241 }, { "epoch": 0.86, "learning_rate": 9.46261300936917e-06, "loss": 0.674, "step": 15242 }, { "epoch": 0.86, "learning_rate": 9.454813553039887e-06, "loss": 0.7262, "step": 15243 }, { "epoch": 0.86, "learning_rate": 9.447017152848125e-06, "loss": 0.6704, "step": 15244 }, { "epoch": 0.86, "learning_rate": 9.439223809057007e-06, "loss": 0.7198, "step": 15245 }, { "epoch": 0.86, "learning_rate": 9.431433521929611e-06, "loss": 0.6381, "step": 15246 }, { "epoch": 0.86, "learning_rate": 9.423646291728838e-06, "loss": 0.6913, "step": 15247 }, { "epoch": 0.86, "learning_rate": 9.415862118717566e-06, "loss": 0.6607, "step": 15248 }, { "epoch": 0.86, "learning_rate": 9.408081003158498e-06, "loss": 0.5987, "step": 15249 }, { "epoch": 0.87, "learning_rate": 9.400302945314288e-06, "loss": 0.6822, "step": 15250 }, { "epoch": 0.87, "learning_rate": 9.39252794544745e-06, "loss": 0.7443, "step": 15251 }, { "epoch": 0.87, "learning_rate": 9.384756003820427e-06, "loss": 0.6506, "step": 15252 }, { "epoch": 0.87, "learning_rate": 9.376987120695545e-06, "loss": 0.7204, "step": 15253 }, { "epoch": 0.87, "learning_rate": 9.369221296335006e-06, "loss": 0.7118, "step": 15254 }, { "epoch": 0.87, "learning_rate": 9.361458531000955e-06, "loss": 0.6351, "step": 15255 }, { "epoch": 0.87, "learning_rate": 9.35369882495536e-06, "loss": 0.7962, "step": 15256 }, { "epoch": 0.87, "learning_rate": 9.345942178460165e-06, "loss": 0.6491, "step": 15257 }, { "epoch": 0.87, "learning_rate": 9.338188591777175e-06, "loss": 0.7001, "step": 15258 }, { "epoch": 0.87, "learning_rate": 9.33043806516808e-06, "loss": 0.799, "step": 15259 }, { "epoch": 0.87, "learning_rate": 9.322690598894512e-06, "loss": 0.7062, "step": 15260 }, { "epoch": 0.87, "learning_rate": 9.314946193217921e-06, "loss": 0.6871, "step": 15261 }, { "epoch": 0.87, "learning_rate": 9.307204848399752e-06, "loss": 0.8417, "step": 15262 }, { "epoch": 0.87, "learning_rate": 9.299466564701253e-06, "loss": 0.7621, "step": 15263 }, { "epoch": 0.87, "learning_rate": 9.291731342383625e-06, "loss": 0.6786, "step": 15264 }, { "epoch": 0.87, "learning_rate": 9.28399918170796e-06, "loss": 0.7559, "step": 15265 }, { "epoch": 0.87, "learning_rate": 9.276270082935245e-06, "loss": 0.66, "step": 15266 }, { "epoch": 0.87, "learning_rate": 9.268544046326332e-06, "loss": 0.768, "step": 15267 }, { "epoch": 0.87, "learning_rate": 9.260821072142012e-06, "loss": 0.7357, "step": 15268 }, { "epoch": 0.87, "learning_rate": 9.253101160642952e-06, "loss": 0.6929, "step": 15269 }, { "epoch": 0.87, "learning_rate": 9.245384312089732e-06, "loss": 0.648, "step": 15270 }, { "epoch": 0.87, "learning_rate": 9.237670526742793e-06, "loss": 0.5635, "step": 15271 }, { "epoch": 0.87, "learning_rate": 9.229959804862508e-06, "loss": 0.6871, "step": 15272 }, { "epoch": 0.87, "learning_rate": 9.222252146709142e-06, "loss": 0.7107, "step": 15273 }, { "epoch": 0.87, "learning_rate": 9.214547552542829e-06, "loss": 0.7622, "step": 15274 }, { "epoch": 0.87, "learning_rate": 9.206846022623627e-06, "loss": 0.6224, "step": 15275 }, { "epoch": 0.87, "learning_rate": 9.199147557211485e-06, "loss": 0.7651, "step": 15276 }, { "epoch": 0.87, "learning_rate": 9.191452156566272e-06, "loss": 0.7492, "step": 15277 }, { "epoch": 0.87, "learning_rate": 9.18375982094768e-06, "loss": 0.6815, "step": 15278 }, { "epoch": 0.87, "learning_rate": 9.176070550615378e-06, "loss": 0.7066, "step": 15279 }, { "epoch": 0.87, "learning_rate": 9.168384345828885e-06, "loss": 0.7518, "step": 15280 }, { "epoch": 0.87, "learning_rate": 9.160701206847622e-06, "loss": 0.6929, "step": 15281 }, { "epoch": 0.87, "learning_rate": 9.15302113393094e-06, "loss": 0.7173, "step": 15282 }, { "epoch": 0.87, "learning_rate": 9.14534412733804e-06, "loss": 0.805, "step": 15283 }, { "epoch": 0.87, "learning_rate": 9.137670187328052e-06, "loss": 0.6703, "step": 15284 }, { "epoch": 0.87, "learning_rate": 9.12999931416001e-06, "loss": 0.6888, "step": 15285 }, { "epoch": 0.87, "learning_rate": 9.122331508092807e-06, "loss": 0.6582, "step": 15286 }, { "epoch": 0.87, "learning_rate": 9.114666769385238e-06, "loss": 0.6432, "step": 15287 }, { "epoch": 0.87, "learning_rate": 9.107005098296018e-06, "loss": 0.7593, "step": 15288 }, { "epoch": 0.87, "learning_rate": 9.09934649508375e-06, "loss": 0.6974, "step": 15289 }, { "epoch": 0.87, "learning_rate": 9.091690960006938e-06, "loss": 0.7526, "step": 15290 }, { "epoch": 0.87, "learning_rate": 9.084038493323966e-06, "loss": 0.6717, "step": 15291 }, { "epoch": 0.87, "learning_rate": 9.076389095293148e-06, "loss": 0.7321, "step": 15292 }, { "epoch": 0.87, "learning_rate": 9.068742766172655e-06, "loss": 0.6789, "step": 15293 }, { "epoch": 0.87, "learning_rate": 9.06109950622055e-06, "loss": 0.7441, "step": 15294 }, { "epoch": 0.87, "learning_rate": 9.053459315694846e-06, "loss": 0.7032, "step": 15295 }, { "epoch": 0.87, "learning_rate": 9.045822194853393e-06, "loss": 0.7919, "step": 15296 }, { "epoch": 0.87, "learning_rate": 9.038188143953974e-06, "loss": 0.7326, "step": 15297 }, { "epoch": 0.87, "learning_rate": 9.030557163254283e-06, "loss": 0.6732, "step": 15298 }, { "epoch": 0.87, "learning_rate": 9.02292925301187e-06, "loss": 0.692, "step": 15299 }, { "epoch": 0.87, "learning_rate": 9.01530441348417e-06, "loss": 0.6745, "step": 15300 }, { "epoch": 0.87, "learning_rate": 9.007682644928561e-06, "loss": 0.6276, "step": 15301 }, { "epoch": 0.87, "learning_rate": 9.0000639476023e-06, "loss": 0.712, "step": 15302 }, { "epoch": 0.87, "learning_rate": 8.992448321762536e-06, "loss": 0.7214, "step": 15303 }, { "epoch": 0.87, "learning_rate": 8.98483576766631e-06, "loss": 0.6549, "step": 15304 }, { "epoch": 0.87, "learning_rate": 8.977226285570606e-06, "loss": 0.7433, "step": 15305 }, { "epoch": 0.87, "learning_rate": 8.969619875732204e-06, "loss": 0.7794, "step": 15306 }, { "epoch": 0.87, "learning_rate": 8.96201653840788e-06, "loss": 0.8228, "step": 15307 }, { "epoch": 0.87, "learning_rate": 8.954416273854238e-06, "loss": 0.6994, "step": 15308 }, { "epoch": 0.87, "learning_rate": 8.946819082327829e-06, "loss": 0.682, "step": 15309 }, { "epoch": 0.87, "learning_rate": 8.939224964085069e-06, "loss": 0.6206, "step": 15310 }, { "epoch": 0.87, "learning_rate": 8.931633919382298e-06, "loss": 0.7049, "step": 15311 }, { "epoch": 0.87, "learning_rate": 8.924045948475701e-06, "loss": 0.7479, "step": 15312 }, { "epoch": 0.87, "learning_rate": 8.916461051621417e-06, "loss": 0.7465, "step": 15313 }, { "epoch": 0.87, "learning_rate": 8.90887922907545e-06, "loss": 0.6734, "step": 15314 }, { "epoch": 0.87, "learning_rate": 8.9013004810937e-06, "loss": 0.6889, "step": 15315 }, { "epoch": 0.87, "learning_rate": 8.893724807931968e-06, "loss": 0.6898, "step": 15316 }, { "epoch": 0.87, "learning_rate": 8.886152209845955e-06, "loss": 0.6787, "step": 15317 }, { "epoch": 0.87, "learning_rate": 8.878582687091286e-06, "loss": 0.6655, "step": 15318 }, { "epoch": 0.87, "learning_rate": 8.871016239923403e-06, "loss": 0.7039, "step": 15319 }, { "epoch": 0.87, "learning_rate": 8.863452868597711e-06, "loss": 0.6938, "step": 15320 }, { "epoch": 0.87, "learning_rate": 8.855892573369507e-06, "loss": 0.6942, "step": 15321 }, { "epoch": 0.87, "learning_rate": 8.848335354493975e-06, "loss": 0.7418, "step": 15322 }, { "epoch": 0.87, "learning_rate": 8.840781212226168e-06, "loss": 0.7941, "step": 15323 }, { "epoch": 0.87, "learning_rate": 8.833230146821058e-06, "loss": 0.6683, "step": 15324 }, { "epoch": 0.87, "learning_rate": 8.825682158533554e-06, "loss": 0.6169, "step": 15325 }, { "epoch": 0.87, "learning_rate": 8.818137247618375e-06, "loss": 0.7145, "step": 15326 }, { "epoch": 0.87, "learning_rate": 8.81059541433018e-06, "loss": 0.6139, "step": 15327 }, { "epoch": 0.87, "learning_rate": 8.803056658923559e-06, "loss": 0.649, "step": 15328 }, { "epoch": 0.87, "learning_rate": 8.795520981652961e-06, "loss": 0.7985, "step": 15329 }, { "epoch": 0.87, "learning_rate": 8.787988382772705e-06, "loss": 0.5851, "step": 15330 }, { "epoch": 0.87, "learning_rate": 8.780458862537067e-06, "loss": 0.8387, "step": 15331 }, { "epoch": 0.87, "learning_rate": 8.772932421200163e-06, "loss": 0.6146, "step": 15332 }, { "epoch": 0.87, "learning_rate": 8.765409059016038e-06, "loss": 0.7431, "step": 15333 }, { "epoch": 0.87, "learning_rate": 8.757888776238621e-06, "loss": 0.8006, "step": 15334 }, { "epoch": 0.87, "learning_rate": 8.750371573121763e-06, "loss": 0.6443, "step": 15335 }, { "epoch": 0.87, "learning_rate": 8.742857449919185e-06, "loss": 0.7122, "step": 15336 }, { "epoch": 0.87, "learning_rate": 8.735346406884482e-06, "loss": 0.6645, "step": 15337 }, { "epoch": 0.87, "learning_rate": 8.727838444271209e-06, "loss": 0.6561, "step": 15338 }, { "epoch": 0.87, "learning_rate": 8.720333562332738e-06, "loss": 0.6751, "step": 15339 }, { "epoch": 0.87, "learning_rate": 8.712831761322394e-06, "loss": 0.6842, "step": 15340 }, { "epoch": 0.87, "learning_rate": 8.705333041493403e-06, "loss": 0.7269, "step": 15341 }, { "epoch": 0.87, "learning_rate": 8.69783740309883e-06, "loss": 0.6961, "step": 15342 }, { "epoch": 0.87, "learning_rate": 8.690344846391729e-06, "loss": 0.6779, "step": 15343 }, { "epoch": 0.87, "learning_rate": 8.682855371624953e-06, "loss": 0.6223, "step": 15344 }, { "epoch": 0.87, "learning_rate": 8.675368979051268e-06, "loss": 0.7276, "step": 15345 }, { "epoch": 0.87, "learning_rate": 8.667885668923404e-06, "loss": 0.7607, "step": 15346 }, { "epoch": 0.87, "learning_rate": 8.660405441493912e-06, "loss": 0.6668, "step": 15347 }, { "epoch": 0.87, "learning_rate": 8.652928297015294e-06, "loss": 0.7708, "step": 15348 }, { "epoch": 0.87, "learning_rate": 8.645454235739903e-06, "loss": 0.6604, "step": 15349 }, { "epoch": 0.87, "learning_rate": 8.637983257920034e-06, "loss": 0.8114, "step": 15350 }, { "epoch": 0.87, "learning_rate": 8.630515363807845e-06, "loss": 0.6312, "step": 15351 }, { "epoch": 0.87, "learning_rate": 8.623050553655365e-06, "loss": 0.69, "step": 15352 }, { "epoch": 0.87, "learning_rate": 8.615588827714582e-06, "loss": 0.7, "step": 15353 }, { "epoch": 0.87, "learning_rate": 8.608130186237329e-06, "loss": 0.7171, "step": 15354 }, { "epoch": 0.87, "learning_rate": 8.60067462947537e-06, "loss": 0.7268, "step": 15355 }, { "epoch": 0.87, "learning_rate": 8.59322215768037e-06, "loss": 0.5969, "step": 15356 }, { "epoch": 0.87, "learning_rate": 8.585772771103818e-06, "loss": 0.5607, "step": 15357 }, { "epoch": 0.87, "learning_rate": 8.578326469997178e-06, "loss": 0.7405, "step": 15358 }, { "epoch": 0.87, "learning_rate": 8.570883254611805e-06, "loss": 0.6196, "step": 15359 }, { "epoch": 0.87, "learning_rate": 8.563443125198889e-06, "loss": 0.6839, "step": 15360 }, { "epoch": 0.87, "learning_rate": 8.55600608200956e-06, "loss": 0.683, "step": 15361 }, { "epoch": 0.87, "learning_rate": 8.548572125294852e-06, "loss": 0.71, "step": 15362 }, { "epoch": 0.87, "learning_rate": 8.541141255305685e-06, "loss": 0.6566, "step": 15363 }, { "epoch": 0.87, "learning_rate": 8.533713472292849e-06, "loss": 0.6202, "step": 15364 }, { "epoch": 0.87, "learning_rate": 8.526288776507064e-06, "loss": 0.7421, "step": 15365 }, { "epoch": 0.87, "learning_rate": 8.518867168198941e-06, "loss": 0.7313, "step": 15366 }, { "epoch": 0.87, "learning_rate": 8.511448647618958e-06, "loss": 0.7128, "step": 15367 }, { "epoch": 0.87, "learning_rate": 8.504033215017527e-06, "loss": 0.6788, "step": 15368 }, { "epoch": 0.87, "learning_rate": 8.496620870644923e-06, "loss": 0.5961, "step": 15369 }, { "epoch": 0.87, "learning_rate": 8.489211614751358e-06, "loss": 0.7466, "step": 15370 }, { "epoch": 0.87, "learning_rate": 8.481805447586878e-06, "loss": 0.6214, "step": 15371 }, { "epoch": 0.87, "learning_rate": 8.47440236940149e-06, "loss": 0.7538, "step": 15372 }, { "epoch": 0.87, "learning_rate": 8.467002380445066e-06, "loss": 0.707, "step": 15373 }, { "epoch": 0.87, "learning_rate": 8.459605480967347e-06, "loss": 0.6238, "step": 15374 }, { "epoch": 0.87, "learning_rate": 8.452211671218024e-06, "loss": 0.7983, "step": 15375 }, { "epoch": 0.87, "learning_rate": 8.444820951446652e-06, "loss": 0.7068, "step": 15376 }, { "epoch": 0.87, "learning_rate": 8.437433321902688e-06, "loss": 0.5722, "step": 15377 }, { "epoch": 0.87, "learning_rate": 8.430048782835465e-06, "loss": 0.6813, "step": 15378 }, { "epoch": 0.87, "learning_rate": 8.422667334494249e-06, "loss": 0.7615, "step": 15379 }, { "epoch": 0.87, "learning_rate": 8.415288977128188e-06, "loss": 0.6335, "step": 15380 }, { "epoch": 0.87, "learning_rate": 8.407913710986325e-06, "loss": 0.5897, "step": 15381 }, { "epoch": 0.87, "learning_rate": 8.400541536317575e-06, "loss": 0.7185, "step": 15382 }, { "epoch": 0.87, "learning_rate": 8.39317245337079e-06, "loss": 0.7153, "step": 15383 }, { "epoch": 0.87, "learning_rate": 8.385806462394662e-06, "loss": 0.7107, "step": 15384 }, { "epoch": 0.87, "learning_rate": 8.378443563637828e-06, "loss": 0.7013, "step": 15385 }, { "epoch": 0.87, "learning_rate": 8.371083757348818e-06, "loss": 0.7201, "step": 15386 }, { "epoch": 0.87, "learning_rate": 8.363727043776038e-06, "loss": 0.7144, "step": 15387 }, { "epoch": 0.87, "learning_rate": 8.356373423167807e-06, "loss": 0.651, "step": 15388 }, { "epoch": 0.87, "learning_rate": 8.349022895772318e-06, "loss": 0.6287, "step": 15389 }, { "epoch": 0.87, "learning_rate": 8.34167546183765e-06, "loss": 0.7263, "step": 15390 }, { "epoch": 0.87, "learning_rate": 8.334331121611828e-06, "loss": 0.6846, "step": 15391 }, { "epoch": 0.87, "learning_rate": 8.326989875342716e-06, "loss": 0.6499, "step": 15392 }, { "epoch": 0.87, "learning_rate": 8.319651723278132e-06, "loss": 0.6414, "step": 15393 }, { "epoch": 0.87, "learning_rate": 8.312316665665743e-06, "loss": 0.74, "step": 15394 }, { "epoch": 0.87, "learning_rate": 8.304984702753138e-06, "loss": 0.703, "step": 15395 }, { "epoch": 0.87, "learning_rate": 8.297655834787777e-06, "loss": 0.7217, "step": 15396 }, { "epoch": 0.87, "learning_rate": 8.290330062017016e-06, "loss": 0.6584, "step": 15397 }, { "epoch": 0.87, "learning_rate": 8.283007384688123e-06, "loss": 0.6715, "step": 15398 }, { "epoch": 0.87, "learning_rate": 8.275687803048282e-06, "loss": 0.6422, "step": 15399 }, { "epoch": 0.87, "learning_rate": 8.268371317344525e-06, "loss": 0.6283, "step": 15400 }, { "epoch": 0.87, "learning_rate": 8.261057927823834e-06, "loss": 0.7838, "step": 15401 }, { "epoch": 0.87, "learning_rate": 8.253747634733011e-06, "loss": 0.6508, "step": 15402 }, { "epoch": 0.87, "learning_rate": 8.246440438318836e-06, "loss": 0.6845, "step": 15403 }, { "epoch": 0.87, "learning_rate": 8.239136338827903e-06, "loss": 0.6885, "step": 15404 }, { "epoch": 0.87, "learning_rate": 8.231835336506788e-06, "loss": 0.7571, "step": 15405 }, { "epoch": 0.87, "learning_rate": 8.224537431601886e-06, "loss": 0.7209, "step": 15406 }, { "epoch": 0.87, "learning_rate": 8.217242624359544e-06, "loss": 0.7603, "step": 15407 }, { "epoch": 0.87, "learning_rate": 8.209950915025987e-06, "loss": 0.7157, "step": 15408 }, { "epoch": 0.87, "learning_rate": 8.202662303847298e-06, "loss": 0.7022, "step": 15409 }, { "epoch": 0.87, "learning_rate": 8.195376791069498e-06, "loss": 0.6978, "step": 15410 }, { "epoch": 0.87, "learning_rate": 8.188094376938527e-06, "loss": 0.6821, "step": 15411 }, { "epoch": 0.87, "learning_rate": 8.18081506170013e-06, "loss": 0.7981, "step": 15412 }, { "epoch": 0.87, "learning_rate": 8.173538845600036e-06, "loss": 0.7149, "step": 15413 }, { "epoch": 0.87, "learning_rate": 8.166265728883826e-06, "loss": 0.6971, "step": 15414 }, { "epoch": 0.87, "learning_rate": 8.158995711797002e-06, "loss": 0.7131, "step": 15415 }, { "epoch": 0.87, "learning_rate": 8.151728794584923e-06, "loss": 0.7551, "step": 15416 }, { "epoch": 0.87, "learning_rate": 8.144464977492883e-06, "loss": 0.6472, "step": 15417 }, { "epoch": 0.87, "learning_rate": 8.137204260766062e-06, "loss": 0.6731, "step": 15418 }, { "epoch": 0.87, "learning_rate": 8.129946644649499e-06, "loss": 0.6335, "step": 15419 }, { "epoch": 0.87, "learning_rate": 8.122692129388176e-06, "loss": 0.7496, "step": 15420 }, { "epoch": 0.87, "learning_rate": 8.115440715226963e-06, "loss": 0.6101, "step": 15421 }, { "epoch": 0.87, "learning_rate": 8.108192402410586e-06, "loss": 0.7626, "step": 15422 }, { "epoch": 0.87, "learning_rate": 8.100947191183716e-06, "loss": 0.66, "step": 15423 }, { "epoch": 0.87, "learning_rate": 8.093705081790893e-06, "loss": 0.6483, "step": 15424 }, { "epoch": 0.87, "learning_rate": 8.086466074476563e-06, "loss": 0.6787, "step": 15425 }, { "epoch": 0.88, "learning_rate": 8.079230169485053e-06, "loss": 0.8093, "step": 15426 }, { "epoch": 0.88, "learning_rate": 8.07199736706058e-06, "loss": 0.8258, "step": 15427 }, { "epoch": 0.88, "learning_rate": 8.064767667447315e-06, "loss": 0.8145, "step": 15428 }, { "epoch": 0.88, "learning_rate": 8.05754107088923e-06, "loss": 0.7287, "step": 15429 }, { "epoch": 0.88, "learning_rate": 8.050317577630262e-06, "loss": 0.7733, "step": 15430 }, { "epoch": 0.88, "learning_rate": 8.043097187914216e-06, "loss": 0.7586, "step": 15431 }, { "epoch": 0.88, "learning_rate": 8.03587990198481e-06, "loss": 0.6571, "step": 15432 }, { "epoch": 0.88, "learning_rate": 8.02866572008566e-06, "loss": 0.7762, "step": 15433 }, { "epoch": 0.88, "learning_rate": 8.021454642460247e-06, "loss": 0.6692, "step": 15434 }, { "epoch": 0.88, "learning_rate": 8.014246669351943e-06, "loss": 0.7938, "step": 15435 }, { "epoch": 0.88, "learning_rate": 8.007041801004056e-06, "loss": 0.6561, "step": 15436 }, { "epoch": 0.88, "learning_rate": 7.999840037659778e-06, "loss": 0.7522, "step": 15437 }, { "epoch": 0.88, "learning_rate": 7.99264137956216e-06, "loss": 0.6477, "step": 15438 }, { "epoch": 0.88, "learning_rate": 7.985445826954207e-06, "loss": 0.6108, "step": 15439 }, { "epoch": 0.88, "learning_rate": 7.978253380078792e-06, "loss": 0.6937, "step": 15440 }, { "epoch": 0.88, "learning_rate": 7.971064039178656e-06, "loss": 0.6536, "step": 15441 }, { "epoch": 0.88, "learning_rate": 7.963877804496444e-06, "loss": 0.693, "step": 15442 }, { "epoch": 0.88, "learning_rate": 7.956694676274734e-06, "loss": 0.6907, "step": 15443 }, { "epoch": 0.88, "learning_rate": 7.949514654755962e-06, "loss": 0.6633, "step": 15444 }, { "epoch": 0.88, "learning_rate": 7.942337740182494e-06, "loss": 0.653, "step": 15445 }, { "epoch": 0.88, "learning_rate": 7.935163932796575e-06, "loss": 0.7594, "step": 15446 }, { "epoch": 0.88, "learning_rate": 7.927993232840292e-06, "loss": 0.6893, "step": 15447 }, { "epoch": 0.88, "learning_rate": 7.92082564055573e-06, "loss": 0.6578, "step": 15448 }, { "epoch": 0.88, "learning_rate": 7.913661156184771e-06, "loss": 0.7315, "step": 15449 }, { "epoch": 0.88, "learning_rate": 7.906499779969256e-06, "loss": 0.7544, "step": 15450 }, { "epoch": 0.88, "learning_rate": 7.899341512150894e-06, "loss": 0.6328, "step": 15451 }, { "epoch": 0.88, "learning_rate": 7.89218635297131e-06, "loss": 0.7088, "step": 15452 }, { "epoch": 0.88, "learning_rate": 7.885034302672e-06, "loss": 0.6154, "step": 15453 }, { "epoch": 0.88, "learning_rate": 7.877885361494353e-06, "loss": 0.7702, "step": 15454 }, { "epoch": 0.88, "learning_rate": 7.870739529679694e-06, "loss": 0.5887, "step": 15455 }, { "epoch": 0.88, "learning_rate": 7.863596807469164e-06, "loss": 0.6858, "step": 15456 }, { "epoch": 0.88, "learning_rate": 7.856457195103894e-06, "loss": 0.7221, "step": 15457 }, { "epoch": 0.88, "learning_rate": 7.849320692824835e-06, "loss": 0.7147, "step": 15458 }, { "epoch": 0.88, "learning_rate": 7.842187300872883e-06, "loss": 0.6894, "step": 15459 }, { "epoch": 0.88, "learning_rate": 7.835057019488823e-06, "loss": 0.6707, "step": 15460 }, { "epoch": 0.88, "learning_rate": 7.827929848913273e-06, "loss": 0.7656, "step": 15461 }, { "epoch": 0.88, "learning_rate": 7.820805789386853e-06, "loss": 0.6308, "step": 15462 }, { "epoch": 0.88, "learning_rate": 7.81368484114996e-06, "loss": 0.7084, "step": 15463 }, { "epoch": 0.88, "learning_rate": 7.806567004442966e-06, "loss": 0.7701, "step": 15464 }, { "epoch": 0.88, "learning_rate": 7.799452279506125e-06, "loss": 0.6281, "step": 15465 }, { "epoch": 0.88, "learning_rate": 7.7923406665796e-06, "loss": 0.7435, "step": 15466 }, { "epoch": 0.88, "learning_rate": 7.785232165903379e-06, "loss": 0.7247, "step": 15467 }, { "epoch": 0.88, "learning_rate": 7.778126777717409e-06, "loss": 0.6979, "step": 15468 }, { "epoch": 0.88, "learning_rate": 7.771024502261526e-06, "loss": 0.6734, "step": 15469 }, { "epoch": 0.88, "learning_rate": 7.763925339775458e-06, "loss": 0.5962, "step": 15470 }, { "epoch": 0.88, "learning_rate": 7.756829290498802e-06, "loss": 0.6914, "step": 15471 }, { "epoch": 0.88, "learning_rate": 7.749736354671067e-06, "loss": 0.6019, "step": 15472 }, { "epoch": 0.88, "learning_rate": 7.742646532531672e-06, "loss": 0.7235, "step": 15473 }, { "epoch": 0.88, "learning_rate": 7.735559824319905e-06, "loss": 0.7582, "step": 15474 }, { "epoch": 0.88, "learning_rate": 7.728476230274961e-06, "loss": 0.7438, "step": 15475 }, { "epoch": 0.88, "learning_rate": 7.72139575063594e-06, "loss": 0.5845, "step": 15476 }, { "epoch": 0.88, "learning_rate": 7.714318385641838e-06, "loss": 0.7642, "step": 15477 }, { "epoch": 0.88, "learning_rate": 7.707244135531499e-06, "loss": 0.5723, "step": 15478 }, { "epoch": 0.88, "learning_rate": 7.700173000543742e-06, "loss": 0.6818, "step": 15479 }, { "epoch": 0.88, "learning_rate": 7.693104980917188e-06, "loss": 0.6428, "step": 15480 }, { "epoch": 0.88, "learning_rate": 7.686040076890422e-06, "loss": 0.6713, "step": 15481 }, { "epoch": 0.88, "learning_rate": 7.67897828870191e-06, "loss": 0.6438, "step": 15482 }, { "epoch": 0.88, "learning_rate": 7.671919616589996e-06, "loss": 0.6902, "step": 15483 }, { "epoch": 0.88, "learning_rate": 7.664864060792942e-06, "loss": 0.6868, "step": 15484 }, { "epoch": 0.88, "learning_rate": 7.657811621548894e-06, "loss": 0.8503, "step": 15485 }, { "epoch": 0.88, "learning_rate": 7.650762299095881e-06, "loss": 0.6956, "step": 15486 }, { "epoch": 0.88, "learning_rate": 7.643716093671827e-06, "loss": 0.6303, "step": 15487 }, { "epoch": 0.88, "learning_rate": 7.636673005514561e-06, "loss": 0.8202, "step": 15488 }, { "epoch": 0.88, "learning_rate": 7.629633034861815e-06, "loss": 0.7311, "step": 15489 }, { "epoch": 0.88, "learning_rate": 7.622596181951203e-06, "loss": 0.6383, "step": 15490 }, { "epoch": 0.88, "learning_rate": 7.615562447020253e-06, "loss": 0.6672, "step": 15491 }, { "epoch": 0.88, "learning_rate": 7.608531830306343e-06, "loss": 0.7423, "step": 15492 }, { "epoch": 0.88, "learning_rate": 7.6015043320468045e-06, "loss": 0.6011, "step": 15493 }, { "epoch": 0.88, "learning_rate": 7.594479952478795e-06, "loss": 0.7548, "step": 15494 }, { "epoch": 0.88, "learning_rate": 7.587458691839444e-06, "loss": 0.6542, "step": 15495 }, { "epoch": 0.88, "learning_rate": 7.580440550365708e-06, "loss": 0.6961, "step": 15496 }, { "epoch": 0.88, "learning_rate": 7.573425528294498e-06, "loss": 0.6011, "step": 15497 }, { "epoch": 0.88, "learning_rate": 7.566413625862578e-06, "loss": 0.6915, "step": 15498 }, { "epoch": 0.88, "learning_rate": 7.5594048433066275e-06, "loss": 0.7514, "step": 15499 }, { "epoch": 0.88, "learning_rate": 7.552399180863168e-06, "loss": 0.676, "step": 15500 }, { "epoch": 0.88, "learning_rate": 7.545396638768698e-06, "loss": 0.6143, "step": 15501 }, { "epoch": 0.88, "learning_rate": 7.538397217259552e-06, "loss": 0.6181, "step": 15502 }, { "epoch": 0.88, "learning_rate": 7.531400916571995e-06, "loss": 0.698, "step": 15503 }, { "epoch": 0.88, "learning_rate": 7.524407736942174e-06, "loss": 0.7181, "step": 15504 }, { "epoch": 0.88, "learning_rate": 7.51741767860612e-06, "loss": 0.7279, "step": 15505 }, { "epoch": 0.88, "learning_rate": 7.510430741799756e-06, "loss": 0.7532, "step": 15506 }, { "epoch": 0.88, "learning_rate": 7.503446926758928e-06, "loss": 0.5986, "step": 15507 }, { "epoch": 0.88, "learning_rate": 7.496466233719335e-06, "loss": 0.7, "step": 15508 }, { "epoch": 0.88, "learning_rate": 7.4894886629166106e-06, "loss": 0.6034, "step": 15509 }, { "epoch": 0.88, "learning_rate": 7.482514214586256e-06, "loss": 0.8122, "step": 15510 }, { "epoch": 0.88, "learning_rate": 7.4755428889636935e-06, "loss": 0.6718, "step": 15511 }, { "epoch": 0.88, "learning_rate": 7.468574686284202e-06, "loss": 0.7353, "step": 15512 }, { "epoch": 0.88, "learning_rate": 7.461609606782993e-06, "loss": 0.6757, "step": 15513 }, { "epoch": 0.88, "learning_rate": 7.454647650695157e-06, "loss": 0.739, "step": 15514 }, { "epoch": 0.88, "learning_rate": 7.4476888182556605e-06, "loss": 0.7168, "step": 15515 }, { "epoch": 0.88, "learning_rate": 7.440733109699394e-06, "loss": 0.7076, "step": 15516 }, { "epoch": 0.88, "learning_rate": 7.4337805252611244e-06, "loss": 0.7794, "step": 15517 }, { "epoch": 0.88, "learning_rate": 7.4268310651755545e-06, "loss": 0.6849, "step": 15518 }, { "epoch": 0.88, "learning_rate": 7.419884729677196e-06, "loss": 0.7011, "step": 15519 }, { "epoch": 0.88, "learning_rate": 7.412941519000527e-06, "loss": 0.7397, "step": 15520 }, { "epoch": 0.88, "learning_rate": 7.406001433379895e-06, "loss": 0.699, "step": 15521 }, { "epoch": 0.88, "learning_rate": 7.399064473049577e-06, "loss": 0.7471, "step": 15522 }, { "epoch": 0.88, "learning_rate": 7.392130638243666e-06, "loss": 0.6982, "step": 15523 }, { "epoch": 0.88, "learning_rate": 7.385199929196241e-06, "loss": 0.7285, "step": 15524 }, { "epoch": 0.88, "learning_rate": 7.378272346141191e-06, "loss": 0.5641, "step": 15525 }, { "epoch": 0.88, "learning_rate": 7.371347889312352e-06, "loss": 0.6352, "step": 15526 }, { "epoch": 0.88, "learning_rate": 7.364426558943449e-06, "loss": 0.8017, "step": 15527 }, { "epoch": 0.88, "learning_rate": 7.357508355268106e-06, "loss": 0.6468, "step": 15528 }, { "epoch": 0.88, "learning_rate": 7.350593278519824e-06, "loss": 0.7003, "step": 15529 }, { "epoch": 0.88, "learning_rate": 7.343681328931995e-06, "loss": 0.6837, "step": 15530 }, { "epoch": 0.88, "learning_rate": 7.336772506737932e-06, "loss": 0.7636, "step": 15531 }, { "epoch": 0.88, "learning_rate": 7.329866812170804e-06, "loss": 0.6211, "step": 15532 }, { "epoch": 0.88, "learning_rate": 7.322964245463704e-06, "loss": 0.6207, "step": 15533 }, { "epoch": 0.88, "learning_rate": 7.316064806849609e-06, "loss": 0.7023, "step": 15534 }, { "epoch": 0.88, "learning_rate": 7.309168496561414e-06, "loss": 0.6335, "step": 15535 }, { "epoch": 0.88, "learning_rate": 7.302275314831875e-06, "loss": 0.6858, "step": 15536 }, { "epoch": 0.88, "learning_rate": 7.2953852618936395e-06, "loss": 0.7341, "step": 15537 }, { "epoch": 0.88, "learning_rate": 7.2884983379793e-06, "loss": 0.6087, "step": 15538 }, { "epoch": 0.88, "learning_rate": 7.281614543321269e-06, "loss": 0.7145, "step": 15539 }, { "epoch": 0.88, "learning_rate": 7.274733878151907e-06, "loss": 0.5737, "step": 15540 }, { "epoch": 0.88, "learning_rate": 7.267856342703461e-06, "loss": 0.656, "step": 15541 }, { "epoch": 0.88, "learning_rate": 7.2609819372080555e-06, "loss": 0.7991, "step": 15542 }, { "epoch": 0.88, "learning_rate": 7.254110661897751e-06, "loss": 0.7218, "step": 15543 }, { "epoch": 0.88, "learning_rate": 7.247242517004437e-06, "loss": 0.6591, "step": 15544 }, { "epoch": 0.88, "learning_rate": 7.240377502759932e-06, "loss": 0.7331, "step": 15545 }, { "epoch": 0.88, "learning_rate": 7.2335156193959586e-06, "loss": 0.7561, "step": 15546 }, { "epoch": 0.88, "learning_rate": 7.226656867144121e-06, "loss": 0.6504, "step": 15547 }, { "epoch": 0.88, "learning_rate": 7.2198012462359245e-06, "loss": 0.6401, "step": 15548 }, { "epoch": 0.88, "learning_rate": 7.2129487569027485e-06, "loss": 0.681, "step": 15549 }, { "epoch": 0.88, "learning_rate": 7.20609939937592e-06, "loss": 0.6405, "step": 15550 }, { "epoch": 0.88, "learning_rate": 7.1992531738866e-06, "loss": 0.7163, "step": 15551 }, { "epoch": 0.88, "learning_rate": 7.192410080665846e-06, "loss": 0.7739, "step": 15552 }, { "epoch": 0.88, "learning_rate": 7.18557011994464e-06, "loss": 0.758, "step": 15553 }, { "epoch": 0.88, "learning_rate": 7.178733291953865e-06, "loss": 0.7493, "step": 15554 }, { "epoch": 0.88, "learning_rate": 7.171899596924281e-06, "loss": 0.6432, "step": 15555 }, { "epoch": 0.88, "learning_rate": 7.165069035086536e-06, "loss": 0.6499, "step": 15556 }, { "epoch": 0.88, "learning_rate": 7.158241606671179e-06, "loss": 0.7254, "step": 15557 }, { "epoch": 0.88, "learning_rate": 7.151417311908648e-06, "loss": 0.7449, "step": 15558 }, { "epoch": 0.88, "learning_rate": 7.144596151029303e-06, "loss": 0.7572, "step": 15559 }, { "epoch": 0.88, "learning_rate": 7.137778124263339e-06, "loss": 0.7671, "step": 15560 }, { "epoch": 0.88, "learning_rate": 7.130963231840915e-06, "loss": 0.6804, "step": 15561 }, { "epoch": 0.88, "learning_rate": 7.124151473992025e-06, "loss": 0.6267, "step": 15562 }, { "epoch": 0.88, "learning_rate": 7.117342850946629e-06, "loss": 0.6968, "step": 15563 }, { "epoch": 0.88, "learning_rate": 7.110537362934477e-06, "loss": 0.6813, "step": 15564 }, { "epoch": 0.88, "learning_rate": 7.103735010185309e-06, "loss": 0.6828, "step": 15565 }, { "epoch": 0.88, "learning_rate": 7.096935792928728e-06, "loss": 0.6734, "step": 15566 }, { "epoch": 0.88, "learning_rate": 7.0901397113941946e-06, "loss": 0.6641, "step": 15567 }, { "epoch": 0.88, "learning_rate": 7.083346765811116e-06, "loss": 0.7633, "step": 15568 }, { "epoch": 0.88, "learning_rate": 7.076556956408775e-06, "loss": 0.7763, "step": 15569 }, { "epoch": 0.88, "learning_rate": 7.069770283416321e-06, "loss": 0.6285, "step": 15570 }, { "epoch": 0.88, "learning_rate": 7.062986747062838e-06, "loss": 0.6864, "step": 15571 }, { "epoch": 0.88, "learning_rate": 7.056206347577298e-06, "loss": 0.742, "step": 15572 }, { "epoch": 0.88, "learning_rate": 7.049429085188552e-06, "loss": 0.6725, "step": 15573 }, { "epoch": 0.88, "learning_rate": 7.0426549601253385e-06, "loss": 0.6972, "step": 15574 }, { "epoch": 0.88, "learning_rate": 7.0358839726163085e-06, "loss": 0.718, "step": 15575 }, { "epoch": 0.88, "learning_rate": 7.029116122890022e-06, "loss": 0.6812, "step": 15576 }, { "epoch": 0.88, "learning_rate": 7.022351411174866e-06, "loss": 0.6957, "step": 15577 }, { "epoch": 0.88, "learning_rate": 7.015589837699199e-06, "loss": 0.6973, "step": 15578 }, { "epoch": 0.88, "learning_rate": 7.00883140269123e-06, "loss": 0.5666, "step": 15579 }, { "epoch": 0.88, "learning_rate": 7.002076106379085e-06, "loss": 0.7785, "step": 15580 }, { "epoch": 0.88, "learning_rate": 6.995323948990784e-06, "loss": 0.6783, "step": 15581 }, { "epoch": 0.88, "learning_rate": 6.988574930754188e-06, "loss": 0.6915, "step": 15582 }, { "epoch": 0.88, "learning_rate": 6.981829051897149e-06, "loss": 0.6476, "step": 15583 }, { "epoch": 0.88, "learning_rate": 6.975086312647306e-06, "loss": 0.5885, "step": 15584 }, { "epoch": 0.88, "learning_rate": 6.968346713232265e-06, "loss": 0.6579, "step": 15585 }, { "epoch": 0.88, "learning_rate": 6.9616102538795025e-06, "loss": 0.6522, "step": 15586 }, { "epoch": 0.88, "learning_rate": 6.954876934816401e-06, "loss": 0.6593, "step": 15587 }, { "epoch": 0.88, "learning_rate": 6.948146756270246e-06, "loss": 0.5847, "step": 15588 }, { "epoch": 0.88, "learning_rate": 6.941419718468168e-06, "loss": 0.6509, "step": 15589 }, { "epoch": 0.88, "learning_rate": 6.934695821637216e-06, "loss": 0.7253, "step": 15590 }, { "epoch": 0.88, "learning_rate": 6.927975066004344e-06, "loss": 0.6537, "step": 15591 }, { "epoch": 0.88, "learning_rate": 6.921257451796414e-06, "loss": 0.6102, "step": 15592 }, { "epoch": 0.88, "learning_rate": 6.914542979240157e-06, "loss": 0.7598, "step": 15593 }, { "epoch": 0.88, "learning_rate": 6.907831648562191e-06, "loss": 0.725, "step": 15594 }, { "epoch": 0.88, "learning_rate": 6.901123459989067e-06, "loss": 0.725, "step": 15595 }, { "epoch": 0.88, "learning_rate": 6.894418413747183e-06, "loss": 0.7269, "step": 15596 }, { "epoch": 0.88, "learning_rate": 6.887716510062847e-06, "loss": 0.6732, "step": 15597 }, { "epoch": 0.88, "learning_rate": 6.881017749162278e-06, "loss": 0.7489, "step": 15598 }, { "epoch": 0.88, "learning_rate": 6.87432213127156e-06, "loss": 0.6983, "step": 15599 }, { "epoch": 0.88, "learning_rate": 6.867629656616714e-06, "loss": 0.7266, "step": 15600 }, { "epoch": 0.88, "learning_rate": 6.860940325423626e-06, "loss": 0.7405, "step": 15601 }, { "epoch": 0.89, "learning_rate": 6.854254137918048e-06, "loss": 0.7127, "step": 15602 }, { "epoch": 0.89, "learning_rate": 6.8475710943257e-06, "loss": 0.6779, "step": 15603 }, { "epoch": 0.89, "learning_rate": 6.840891194872112e-06, "loss": 0.6157, "step": 15604 }, { "epoch": 0.89, "learning_rate": 6.83421443978276e-06, "loss": 0.7004, "step": 15605 }, { "epoch": 0.89, "learning_rate": 6.827540829283019e-06, "loss": 0.6224, "step": 15606 }, { "epoch": 0.89, "learning_rate": 6.82087036359812e-06, "loss": 0.7733, "step": 15607 }, { "epoch": 0.89, "learning_rate": 6.814203042953227e-06, "loss": 0.6879, "step": 15608 }, { "epoch": 0.89, "learning_rate": 6.80753886757336e-06, "loss": 0.6901, "step": 15609 }, { "epoch": 0.89, "learning_rate": 6.800877837683461e-06, "loss": 0.7186, "step": 15610 }, { "epoch": 0.89, "learning_rate": 6.794219953508385e-06, "loss": 0.6983, "step": 15611 }, { "epoch": 0.89, "learning_rate": 6.787565215272795e-06, "loss": 0.6804, "step": 15612 }, { "epoch": 0.89, "learning_rate": 6.780913623201346e-06, "loss": 0.739, "step": 15613 }, { "epoch": 0.89, "learning_rate": 6.774265177518558e-06, "loss": 0.6882, "step": 15614 }, { "epoch": 0.89, "learning_rate": 6.767619878448783e-06, "loss": 0.7021, "step": 15615 }, { "epoch": 0.89, "learning_rate": 6.760977726216366e-06, "loss": 0.6324, "step": 15616 }, { "epoch": 0.89, "learning_rate": 6.754338721045461e-06, "loss": 0.7301, "step": 15617 }, { "epoch": 0.89, "learning_rate": 6.747702863160199e-06, "loss": 0.6529, "step": 15618 }, { "epoch": 0.89, "learning_rate": 6.7410701527845125e-06, "loss": 0.7908, "step": 15619 }, { "epoch": 0.89, "learning_rate": 6.734440590142288e-06, "loss": 0.7062, "step": 15620 }, { "epoch": 0.89, "learning_rate": 6.727814175457303e-06, "loss": 0.6564, "step": 15621 }, { "epoch": 0.89, "learning_rate": 6.72119090895319e-06, "loss": 0.715, "step": 15622 }, { "epoch": 0.89, "learning_rate": 6.714570790853514e-06, "loss": 0.7122, "step": 15623 }, { "epoch": 0.89, "learning_rate": 6.70795382138173e-06, "loss": 0.6505, "step": 15624 }, { "epoch": 0.89, "learning_rate": 6.70134000076118e-06, "loss": 0.6175, "step": 15625 }, { "epoch": 0.89, "learning_rate": 6.694729329215077e-06, "loss": 0.6523, "step": 15626 }, { "epoch": 0.89, "learning_rate": 6.688121806966563e-06, "loss": 0.6639, "step": 15627 }, { "epoch": 0.89, "learning_rate": 6.681517434238671e-06, "loss": 0.7987, "step": 15628 }, { "epoch": 0.89, "learning_rate": 6.674916211254289e-06, "loss": 0.6241, "step": 15629 }, { "epoch": 0.89, "learning_rate": 6.6683181382362406e-06, "loss": 0.7038, "step": 15630 }, { "epoch": 0.89, "learning_rate": 6.661723215407223e-06, "loss": 0.7075, "step": 15631 }, { "epoch": 0.89, "learning_rate": 6.655131442989837e-06, "loss": 0.6485, "step": 15632 }, { "epoch": 0.89, "learning_rate": 6.648542821206582e-06, "loss": 0.7224, "step": 15633 }, { "epoch": 0.89, "learning_rate": 6.6419573502798374e-06, "loss": 0.7403, "step": 15634 }, { "epoch": 0.89, "learning_rate": 6.635375030431857e-06, "loss": 0.6376, "step": 15635 }, { "epoch": 0.89, "learning_rate": 6.6287958618848175e-06, "loss": 0.6774, "step": 15636 }, { "epoch": 0.89, "learning_rate": 6.6222198448608e-06, "loss": 0.663, "step": 15637 }, { "epoch": 0.89, "learning_rate": 6.6156469795817576e-06, "loss": 0.7647, "step": 15638 }, { "epoch": 0.89, "learning_rate": 6.609077266269536e-06, "loss": 0.671, "step": 15639 }, { "epoch": 0.89, "learning_rate": 6.602510705145892e-06, "loss": 0.7218, "step": 15640 }, { "epoch": 0.89, "learning_rate": 6.59594729643247e-06, "loss": 0.6859, "step": 15641 }, { "epoch": 0.89, "learning_rate": 6.589387040350758e-06, "loss": 0.7723, "step": 15642 }, { "epoch": 0.89, "learning_rate": 6.5828299371222145e-06, "loss": 0.6848, "step": 15643 }, { "epoch": 0.89, "learning_rate": 6.576275986968161e-06, "loss": 0.6844, "step": 15644 }, { "epoch": 0.89, "learning_rate": 6.56972519010981e-06, "loss": 0.7116, "step": 15645 }, { "epoch": 0.89, "learning_rate": 6.563177546768273e-06, "loss": 0.6697, "step": 15646 }, { "epoch": 0.89, "learning_rate": 6.556633057164518e-06, "loss": 0.6452, "step": 15647 }, { "epoch": 0.89, "learning_rate": 6.550091721519491e-06, "loss": 0.6411, "step": 15648 }, { "epoch": 0.89, "learning_rate": 6.543553540053926e-06, "loss": 0.6289, "step": 15649 }, { "epoch": 0.89, "learning_rate": 6.537018512988524e-06, "loss": 0.6379, "step": 15650 }, { "epoch": 0.89, "learning_rate": 6.530486640543865e-06, "loss": 0.6817, "step": 15651 }, { "epoch": 0.89, "learning_rate": 6.523957922940427e-06, "loss": 0.6496, "step": 15652 }, { "epoch": 0.89, "learning_rate": 6.517432360398556e-06, "loss": 0.7204, "step": 15653 }, { "epoch": 0.89, "learning_rate": 6.510909953138511e-06, "loss": 0.7876, "step": 15654 }, { "epoch": 0.89, "learning_rate": 6.504390701380447e-06, "loss": 0.7246, "step": 15655 }, { "epoch": 0.89, "learning_rate": 6.497874605344378e-06, "loss": 0.7043, "step": 15656 }, { "epoch": 0.89, "learning_rate": 6.491361665250262e-06, "loss": 0.7334, "step": 15657 }, { "epoch": 0.89, "learning_rate": 6.484851881317933e-06, "loss": 0.6519, "step": 15658 }, { "epoch": 0.89, "learning_rate": 6.478345253767115e-06, "loss": 0.6389, "step": 15659 }, { "epoch": 0.89, "learning_rate": 6.4718417828174e-06, "loss": 0.6372, "step": 15660 }, { "epoch": 0.89, "learning_rate": 6.465341468688313e-06, "loss": 0.695, "step": 15661 }, { "epoch": 0.89, "learning_rate": 6.458844311599266e-06, "loss": 0.7204, "step": 15662 }, { "epoch": 0.89, "learning_rate": 6.452350311769529e-06, "loss": 0.7275, "step": 15663 }, { "epoch": 0.89, "learning_rate": 6.445859469418303e-06, "loss": 0.7332, "step": 15664 }, { "epoch": 0.89, "learning_rate": 6.439371784764681e-06, "loss": 0.6971, "step": 15665 }, { "epoch": 0.89, "learning_rate": 6.432887258027653e-06, "loss": 0.7605, "step": 15666 }, { "epoch": 0.89, "learning_rate": 6.426405889426046e-06, "loss": 0.6574, "step": 15667 }, { "epoch": 0.89, "learning_rate": 6.41992767917865e-06, "loss": 0.7916, "step": 15668 }, { "epoch": 0.89, "learning_rate": 6.413452627504113e-06, "loss": 0.6656, "step": 15669 }, { "epoch": 0.89, "learning_rate": 6.406980734621004e-06, "loss": 0.6736, "step": 15670 }, { "epoch": 0.89, "learning_rate": 6.400512000747738e-06, "loss": 0.7469, "step": 15671 }, { "epoch": 0.89, "learning_rate": 6.394046426102674e-06, "loss": 0.6915, "step": 15672 }, { "epoch": 0.89, "learning_rate": 6.387584010904036e-06, "loss": 0.6984, "step": 15673 }, { "epoch": 0.89, "learning_rate": 6.381124755369938e-06, "loss": 0.6989, "step": 15674 }, { "epoch": 0.89, "learning_rate": 6.374668659718408e-06, "loss": 0.5984, "step": 15675 }, { "epoch": 0.89, "learning_rate": 6.368215724167337e-06, "loss": 0.6666, "step": 15676 }, { "epoch": 0.89, "learning_rate": 6.3617659489345725e-06, "loss": 0.7106, "step": 15677 }, { "epoch": 0.89, "learning_rate": 6.355319334237764e-06, "loss": 0.6629, "step": 15678 }, { "epoch": 0.89, "learning_rate": 6.3488758802945354e-06, "loss": 0.7242, "step": 15679 }, { "epoch": 0.89, "learning_rate": 6.342435587322338e-06, "loss": 0.7233, "step": 15680 }, { "epoch": 0.89, "learning_rate": 6.335998455538561e-06, "loss": 0.5902, "step": 15681 }, { "epoch": 0.89, "learning_rate": 6.329564485160488e-06, "loss": 0.639, "step": 15682 }, { "epoch": 0.89, "learning_rate": 6.323133676405268e-06, "loss": 0.824, "step": 15683 }, { "epoch": 0.89, "learning_rate": 6.316706029489972e-06, "loss": 0.8195, "step": 15684 }, { "epoch": 0.89, "learning_rate": 6.310281544631546e-06, "loss": 0.7284, "step": 15685 }, { "epoch": 0.89, "learning_rate": 6.303860222046831e-06, "loss": 0.7626, "step": 15686 }, { "epoch": 0.89, "learning_rate": 6.297442061952541e-06, "loss": 0.7039, "step": 15687 }, { "epoch": 0.89, "learning_rate": 6.2910270645653355e-06, "loss": 0.6698, "step": 15688 }, { "epoch": 0.89, "learning_rate": 6.2846152301017205e-06, "loss": 0.7835, "step": 15689 }, { "epoch": 0.89, "learning_rate": 6.2782065587781215e-06, "loss": 0.7054, "step": 15690 }, { "epoch": 0.89, "learning_rate": 6.2718010508108545e-06, "loss": 0.6585, "step": 15691 }, { "epoch": 0.89, "learning_rate": 6.265398706416115e-06, "loss": 0.6886, "step": 15692 }, { "epoch": 0.89, "learning_rate": 6.258999525809994e-06, "loss": 0.6137, "step": 15693 }, { "epoch": 0.89, "learning_rate": 6.252603509208466e-06, "loss": 0.7557, "step": 15694 }, { "epoch": 0.89, "learning_rate": 6.246210656827445e-06, "loss": 0.6735, "step": 15695 }, { "epoch": 0.89, "learning_rate": 6.239820968882693e-06, "loss": 0.757, "step": 15696 }, { "epoch": 0.89, "learning_rate": 6.233434445589869e-06, "loss": 0.7317, "step": 15697 }, { "epoch": 0.89, "learning_rate": 6.227051087164559e-06, "loss": 0.6518, "step": 15698 }, { "epoch": 0.89, "learning_rate": 6.220670893822211e-06, "loss": 0.6934, "step": 15699 }, { "epoch": 0.89, "learning_rate": 6.214293865778142e-06, "loss": 0.6345, "step": 15700 }, { "epoch": 0.89, "learning_rate": 6.207920003247625e-06, "loss": 0.6328, "step": 15701 }, { "epoch": 0.89, "learning_rate": 6.201549306445786e-06, "loss": 0.7169, "step": 15702 }, { "epoch": 0.89, "learning_rate": 6.195181775587655e-06, "loss": 0.673, "step": 15703 }, { "epoch": 0.89, "learning_rate": 6.188817410888148e-06, "loss": 0.6697, "step": 15704 }, { "epoch": 0.89, "learning_rate": 6.182456212562093e-06, "loss": 0.7251, "step": 15705 }, { "epoch": 0.89, "learning_rate": 6.176098180824175e-06, "loss": 0.7711, "step": 15706 }, { "epoch": 0.89, "learning_rate": 6.1697433158890205e-06, "loss": 0.6044, "step": 15707 }, { "epoch": 0.89, "learning_rate": 6.1633916179710925e-06, "loss": 0.6807, "step": 15708 }, { "epoch": 0.89, "learning_rate": 6.157043087284798e-06, "loss": 0.7093, "step": 15709 }, { "epoch": 0.89, "learning_rate": 6.1506977240444074e-06, "loss": 0.7681, "step": 15710 }, { "epoch": 0.89, "learning_rate": 6.144355528464108e-06, "loss": 0.7193, "step": 15711 }, { "epoch": 0.89, "learning_rate": 6.1380165007579484e-06, "loss": 0.7372, "step": 15712 }, { "epoch": 0.89, "learning_rate": 6.131680641139892e-06, "loss": 0.6316, "step": 15713 }, { "epoch": 0.89, "learning_rate": 6.125347949823801e-06, "loss": 0.7357, "step": 15714 }, { "epoch": 0.89, "learning_rate": 6.119018427023404e-06, "loss": 0.6529, "step": 15715 }, { "epoch": 0.89, "learning_rate": 6.112692072952342e-06, "loss": 0.6591, "step": 15716 }, { "epoch": 0.89, "learning_rate": 6.106368887824155e-06, "loss": 0.7638, "step": 15717 }, { "epoch": 0.89, "learning_rate": 6.100048871852271e-06, "loss": 0.643, "step": 15718 }, { "epoch": 0.89, "learning_rate": 6.093732025249999e-06, "loss": 0.7114, "step": 15719 }, { "epoch": 0.89, "learning_rate": 6.087418348230534e-06, "loss": 0.7354, "step": 15720 }, { "epoch": 0.89, "learning_rate": 6.081107841007006e-06, "loss": 0.6666, "step": 15721 }, { "epoch": 0.89, "learning_rate": 6.074800503792399e-06, "loss": 0.6969, "step": 15722 }, { "epoch": 0.89, "learning_rate": 6.068496336799601e-06, "loss": 0.653, "step": 15723 }, { "epoch": 0.89, "learning_rate": 6.062195340241406e-06, "loss": 0.6843, "step": 15724 }, { "epoch": 0.89, "learning_rate": 6.0558975143304665e-06, "loss": 0.6701, "step": 15725 }, { "epoch": 0.89, "learning_rate": 6.049602859279368e-06, "loss": 0.7333, "step": 15726 }, { "epoch": 0.89, "learning_rate": 6.043311375300564e-06, "loss": 0.6028, "step": 15727 }, { "epoch": 0.89, "learning_rate": 6.037023062606406e-06, "loss": 0.6864, "step": 15728 }, { "epoch": 0.89, "learning_rate": 6.030737921409169e-06, "loss": 0.6515, "step": 15729 }, { "epoch": 0.89, "learning_rate": 6.024455951920949e-06, "loss": 0.6606, "step": 15730 }, { "epoch": 0.89, "learning_rate": 6.018177154353821e-06, "loss": 0.7506, "step": 15731 }, { "epoch": 0.89, "learning_rate": 6.011901528919672e-06, "loss": 0.639, "step": 15732 }, { "epoch": 0.89, "learning_rate": 6.005629075830343e-06, "loss": 0.7441, "step": 15733 }, { "epoch": 0.89, "learning_rate": 5.999359795297532e-06, "loss": 0.7159, "step": 15734 }, { "epoch": 0.89, "learning_rate": 5.993093687532869e-06, "loss": 0.7014, "step": 15735 }, { "epoch": 0.89, "learning_rate": 5.986830752747841e-06, "loss": 0.7438, "step": 15736 }, { "epoch": 0.89, "learning_rate": 5.980570991153822e-06, "loss": 0.6919, "step": 15737 }, { "epoch": 0.89, "learning_rate": 5.974314402962111e-06, "loss": 0.7709, "step": 15738 }, { "epoch": 0.89, "learning_rate": 5.968060988383883e-06, "loss": 0.7431, "step": 15739 }, { "epoch": 0.89, "learning_rate": 5.961810747630192e-06, "loss": 0.7221, "step": 15740 }, { "epoch": 0.89, "learning_rate": 5.955563680912013e-06, "loss": 0.5618, "step": 15741 }, { "epoch": 0.89, "learning_rate": 5.949319788440211e-06, "loss": 0.7486, "step": 15742 }, { "epoch": 0.89, "learning_rate": 5.943079070425528e-06, "loss": 0.6847, "step": 15743 }, { "epoch": 0.89, "learning_rate": 5.936841527078607e-06, "loss": 0.7062, "step": 15744 }, { "epoch": 0.89, "learning_rate": 5.930607158609946e-06, "loss": 0.7215, "step": 15745 }, { "epoch": 0.89, "learning_rate": 5.924375965230011e-06, "loss": 0.676, "step": 15746 }, { "epoch": 0.89, "learning_rate": 5.9181479471491105e-06, "loss": 0.7126, "step": 15747 }, { "epoch": 0.89, "learning_rate": 5.911923104577455e-06, "loss": 0.6155, "step": 15748 }, { "epoch": 0.89, "learning_rate": 5.905701437725142e-06, "loss": 0.7121, "step": 15749 }, { "epoch": 0.89, "learning_rate": 5.8994829468021925e-06, "loss": 0.6404, "step": 15750 }, { "epoch": 0.89, "learning_rate": 5.893267632018484e-06, "loss": 0.7234, "step": 15751 }, { "epoch": 0.89, "learning_rate": 5.887055493583793e-06, "loss": 0.7266, "step": 15752 }, { "epoch": 0.89, "learning_rate": 5.880846531707784e-06, "loss": 0.7071, "step": 15753 }, { "epoch": 0.89, "learning_rate": 5.8746407466000464e-06, "loss": 0.6918, "step": 15754 }, { "epoch": 0.89, "learning_rate": 5.8684381384700335e-06, "loss": 0.6943, "step": 15755 }, { "epoch": 0.89, "learning_rate": 5.862238707527123e-06, "loss": 0.7106, "step": 15756 }, { "epoch": 0.89, "learning_rate": 5.856042453980526e-06, "loss": 0.6816, "step": 15757 }, { "epoch": 0.89, "learning_rate": 5.849849378039396e-06, "loss": 0.6591, "step": 15758 }, { "epoch": 0.89, "learning_rate": 5.843659479912778e-06, "loss": 0.6854, "step": 15759 }, { "epoch": 0.89, "learning_rate": 5.837472759809581e-06, "loss": 0.7273, "step": 15760 }, { "epoch": 0.89, "learning_rate": 5.831289217938618e-06, "loss": 0.7003, "step": 15761 }, { "epoch": 0.89, "learning_rate": 5.825108854508621e-06, "loss": 0.7019, "step": 15762 }, { "epoch": 0.89, "learning_rate": 5.818931669728189e-06, "loss": 0.6757, "step": 15763 }, { "epoch": 0.89, "learning_rate": 5.812757663805801e-06, "loss": 0.6676, "step": 15764 }, { "epoch": 0.89, "learning_rate": 5.8065868369498565e-06, "loss": 0.7269, "step": 15765 }, { "epoch": 0.89, "learning_rate": 5.8004191893686554e-06, "loss": 0.7241, "step": 15766 }, { "epoch": 0.89, "learning_rate": 5.7942547212703315e-06, "loss": 0.734, "step": 15767 }, { "epoch": 0.89, "learning_rate": 5.788093432862985e-06, "loss": 0.663, "step": 15768 }, { "epoch": 0.89, "learning_rate": 5.78193532435457e-06, "loss": 0.635, "step": 15769 }, { "epoch": 0.89, "learning_rate": 5.775780395952923e-06, "loss": 0.6764, "step": 15770 }, { "epoch": 0.89, "learning_rate": 5.769628647865799e-06, "loss": 0.7221, "step": 15771 }, { "epoch": 0.89, "learning_rate": 5.763480080300832e-06, "loss": 0.7031, "step": 15772 }, { "epoch": 0.89, "learning_rate": 5.757334693465566e-06, "loss": 0.6153, "step": 15773 }, { "epoch": 0.89, "learning_rate": 5.751192487567425e-06, "loss": 0.6605, "step": 15774 }, { "epoch": 0.89, "learning_rate": 5.745053462813699e-06, "loss": 0.6457, "step": 15775 }, { "epoch": 0.89, "learning_rate": 5.7389176194116215e-06, "loss": 0.6861, "step": 15776 }, { "epoch": 0.89, "learning_rate": 5.732784957568282e-06, "loss": 0.7085, "step": 15777 }, { "epoch": 0.9, "learning_rate": 5.7266554774906615e-06, "loss": 0.6722, "step": 15778 }, { "epoch": 0.9, "learning_rate": 5.720529179385659e-06, "loss": 0.6183, "step": 15779 }, { "epoch": 0.9, "learning_rate": 5.714406063460065e-06, "loss": 0.6582, "step": 15780 }, { "epoch": 0.9, "learning_rate": 5.708286129920548e-06, "loss": 0.6525, "step": 15781 }, { "epoch": 0.9, "learning_rate": 5.702169378973643e-06, "loss": 0.6518, "step": 15782 }, { "epoch": 0.9, "learning_rate": 5.696055810825851e-06, "loss": 0.7751, "step": 15783 }, { "epoch": 0.9, "learning_rate": 5.689945425683474e-06, "loss": 0.8136, "step": 15784 }, { "epoch": 0.9, "learning_rate": 5.68383822375278e-06, "loss": 0.6742, "step": 15785 }, { "epoch": 0.9, "learning_rate": 5.6777342052399045e-06, "loss": 0.6771, "step": 15786 }, { "epoch": 0.9, "learning_rate": 5.6716333703508594e-06, "loss": 0.702, "step": 15787 }, { "epoch": 0.9, "learning_rate": 5.665535719291593e-06, "loss": 0.6882, "step": 15788 }, { "epoch": 0.9, "learning_rate": 5.659441252267894e-06, "loss": 0.6588, "step": 15789 }, { "epoch": 0.9, "learning_rate": 5.653349969485455e-06, "loss": 0.6656, "step": 15790 }, { "epoch": 0.9, "learning_rate": 5.64726187114989e-06, "loss": 0.6249, "step": 15791 }, { "epoch": 0.9, "learning_rate": 5.641176957466688e-06, "loss": 0.6944, "step": 15792 }, { "epoch": 0.9, "learning_rate": 5.63509522864123e-06, "loss": 0.6952, "step": 15793 }, { "epoch": 0.9, "learning_rate": 5.6290166848787854e-06, "loss": 0.6316, "step": 15794 }, { "epoch": 0.9, "learning_rate": 5.6229413263845345e-06, "loss": 0.7243, "step": 15795 }, { "epoch": 0.9, "learning_rate": 5.616869153363535e-06, "loss": 0.6936, "step": 15796 }, { "epoch": 0.9, "learning_rate": 5.610800166020713e-06, "loss": 0.7118, "step": 15797 }, { "epoch": 0.9, "learning_rate": 5.604734364560926e-06, "loss": 0.7909, "step": 15798 }, { "epoch": 0.9, "learning_rate": 5.5986717491889216e-06, "loss": 0.7013, "step": 15799 }, { "epoch": 0.9, "learning_rate": 5.592612320109314e-06, "loss": 0.6568, "step": 15800 }, { "epoch": 0.9, "learning_rate": 5.58655607752665e-06, "loss": 0.7011, "step": 15801 }, { "epoch": 0.9, "learning_rate": 5.580503021645322e-06, "loss": 0.7047, "step": 15802 }, { "epoch": 0.9, "learning_rate": 5.574453152669646e-06, "loss": 0.727, "step": 15803 }, { "epoch": 0.9, "learning_rate": 5.568406470803799e-06, "loss": 0.6635, "step": 15804 }, { "epoch": 0.9, "learning_rate": 5.562362976251901e-06, "loss": 0.7013, "step": 15805 }, { "epoch": 0.9, "learning_rate": 5.5563226692179175e-06, "loss": 0.7, "step": 15806 }, { "epoch": 0.9, "learning_rate": 5.5502855499057315e-06, "loss": 0.7615, "step": 15807 }, { "epoch": 0.9, "learning_rate": 5.544251618519125e-06, "loss": 0.7774, "step": 15808 }, { "epoch": 0.9, "learning_rate": 5.538220875261735e-06, "loss": 0.7927, "step": 15809 }, { "epoch": 0.9, "learning_rate": 5.5321933203371205e-06, "loss": 0.6739, "step": 15810 }, { "epoch": 0.9, "learning_rate": 5.5261689539487515e-06, "loss": 0.7575, "step": 15811 }, { "epoch": 0.9, "learning_rate": 5.520147776299933e-06, "loss": 0.6795, "step": 15812 }, { "epoch": 0.9, "learning_rate": 5.514129787593902e-06, "loss": 0.7083, "step": 15813 }, { "epoch": 0.9, "learning_rate": 5.508114988033808e-06, "loss": 0.6825, "step": 15814 }, { "epoch": 0.9, "learning_rate": 5.502103377822643e-06, "loss": 0.7647, "step": 15815 }, { "epoch": 0.9, "learning_rate": 5.4960949571633e-06, "loss": 0.7874, "step": 15816 }, { "epoch": 0.9, "learning_rate": 5.490089726258607e-06, "loss": 0.6748, "step": 15817 }, { "epoch": 0.9, "learning_rate": 5.4840876853112675e-06, "loss": 0.7625, "step": 15818 }, { "epoch": 0.9, "learning_rate": 5.4780888345238205e-06, "loss": 0.7041, "step": 15819 }, { "epoch": 0.9, "learning_rate": 5.47209317409878e-06, "loss": 0.7068, "step": 15820 }, { "epoch": 0.9, "learning_rate": 5.4661007042385085e-06, "loss": 0.6625, "step": 15821 }, { "epoch": 0.9, "learning_rate": 5.460111425145253e-06, "loss": 0.7274, "step": 15822 }, { "epoch": 0.9, "learning_rate": 5.4541253370211656e-06, "loss": 0.6011, "step": 15823 }, { "epoch": 0.9, "learning_rate": 5.448142440068316e-06, "loss": 0.7592, "step": 15824 }, { "epoch": 0.9, "learning_rate": 5.442162734488643e-06, "loss": 0.6769, "step": 15825 }, { "epoch": 0.9, "learning_rate": 5.436186220483952e-06, "loss": 0.7859, "step": 15826 }, { "epoch": 0.9, "learning_rate": 5.430212898255971e-06, "loss": 0.7058, "step": 15827 }, { "epoch": 0.9, "learning_rate": 5.424242768006338e-06, "loss": 0.6715, "step": 15828 }, { "epoch": 0.9, "learning_rate": 5.418275829936537e-06, "loss": 0.6601, "step": 15829 }, { "epoch": 0.9, "learning_rate": 5.412312084247972e-06, "loss": 0.6626, "step": 15830 }, { "epoch": 0.9, "learning_rate": 5.40635153114194e-06, "loss": 0.7292, "step": 15831 }, { "epoch": 0.9, "learning_rate": 5.400394170819634e-06, "loss": 0.6499, "step": 15832 }, { "epoch": 0.9, "learning_rate": 5.394440003482126e-06, "loss": 0.6966, "step": 15833 }, { "epoch": 0.9, "learning_rate": 5.388489029330379e-06, "loss": 0.6379, "step": 15834 }, { "epoch": 0.9, "learning_rate": 5.382541248565242e-06, "loss": 0.6557, "step": 15835 }, { "epoch": 0.9, "learning_rate": 5.37659666138749e-06, "loss": 0.7192, "step": 15836 }, { "epoch": 0.9, "learning_rate": 5.370655267997749e-06, "loss": 0.6345, "step": 15837 }, { "epoch": 0.9, "learning_rate": 5.36471706859657e-06, "loss": 0.7252, "step": 15838 }, { "epoch": 0.9, "learning_rate": 5.358782063384382e-06, "loss": 0.6508, "step": 15839 }, { "epoch": 0.9, "learning_rate": 5.352850252561515e-06, "loss": 0.6047, "step": 15840 }, { "epoch": 0.9, "learning_rate": 5.3469216363281835e-06, "loss": 0.7085, "step": 15841 }, { "epoch": 0.9, "learning_rate": 5.3409962148844636e-06, "loss": 0.6398, "step": 15842 }, { "epoch": 0.9, "learning_rate": 5.335073988430372e-06, "loss": 0.7592, "step": 15843 }, { "epoch": 0.9, "learning_rate": 5.329154957165794e-06, "loss": 0.7497, "step": 15844 }, { "epoch": 0.9, "learning_rate": 5.323239121290535e-06, "loss": 0.7174, "step": 15845 }, { "epoch": 0.9, "learning_rate": 5.317326481004259e-06, "loss": 0.7768, "step": 15846 }, { "epoch": 0.9, "learning_rate": 5.3114170365065164e-06, "loss": 0.687, "step": 15847 }, { "epoch": 0.9, "learning_rate": 5.305510787996792e-06, "loss": 0.6758, "step": 15848 }, { "epoch": 0.9, "learning_rate": 5.299607735674406e-06, "loss": 0.6826, "step": 15849 }, { "epoch": 0.9, "learning_rate": 5.29370787973863e-06, "loss": 0.7579, "step": 15850 }, { "epoch": 0.9, "learning_rate": 5.287811220388583e-06, "loss": 0.6939, "step": 15851 }, { "epoch": 0.9, "learning_rate": 5.281917757823296e-06, "loss": 0.6979, "step": 15852 }, { "epoch": 0.9, "learning_rate": 5.27602749224172e-06, "loss": 0.6604, "step": 15853 }, { "epoch": 0.9, "learning_rate": 5.270140423842607e-06, "loss": 0.6833, "step": 15854 }, { "epoch": 0.9, "learning_rate": 5.26425655282472e-06, "loss": 0.6382, "step": 15855 }, { "epoch": 0.9, "learning_rate": 5.258375879386601e-06, "loss": 0.7103, "step": 15856 }, { "epoch": 0.9, "learning_rate": 5.25249840372678e-06, "loss": 0.7495, "step": 15857 }, { "epoch": 0.9, "learning_rate": 5.246624126043609e-06, "loss": 0.7065, "step": 15858 }, { "epoch": 0.9, "learning_rate": 5.240753046535396e-06, "loss": 0.7352, "step": 15859 }, { "epoch": 0.9, "learning_rate": 5.234885165400272e-06, "loss": 0.7403, "step": 15860 }, { "epoch": 0.9, "learning_rate": 5.2290204828363e-06, "loss": 0.7429, "step": 15861 }, { "epoch": 0.9, "learning_rate": 5.223158999041444e-06, "loss": 0.6929, "step": 15862 }, { "epoch": 0.9, "learning_rate": 5.217300714213513e-06, "loss": 0.6221, "step": 15863 }, { "epoch": 0.9, "learning_rate": 5.211445628550271e-06, "loss": 0.6432, "step": 15864 }, { "epoch": 0.9, "learning_rate": 5.205593742249326e-06, "loss": 0.6728, "step": 15865 }, { "epoch": 0.9, "learning_rate": 5.199745055508221e-06, "loss": 0.6387, "step": 15866 }, { "epoch": 0.9, "learning_rate": 5.193899568524319e-06, "loss": 0.8408, "step": 15867 }, { "epoch": 0.9, "learning_rate": 5.188057281494952e-06, "loss": 0.7033, "step": 15868 }, { "epoch": 0.9, "learning_rate": 5.182218194617306e-06, "loss": 0.6933, "step": 15869 }, { "epoch": 0.9, "learning_rate": 5.176382308088467e-06, "loss": 0.6469, "step": 15870 }, { "epoch": 0.9, "learning_rate": 5.170549622105403e-06, "loss": 0.6129, "step": 15871 }, { "epoch": 0.9, "learning_rate": 5.164720136864987e-06, "loss": 0.6698, "step": 15872 }, { "epoch": 0.9, "learning_rate": 5.158893852564006e-06, "loss": 0.7783, "step": 15873 }, { "epoch": 0.9, "learning_rate": 5.15307076939906e-06, "loss": 0.6557, "step": 15874 }, { "epoch": 0.9, "learning_rate": 5.1472508875667236e-06, "loss": 0.718, "step": 15875 }, { "epoch": 0.9, "learning_rate": 5.1414342072634295e-06, "loss": 0.7153, "step": 15876 }, { "epoch": 0.9, "learning_rate": 5.13562072868553e-06, "loss": 0.7145, "step": 15877 }, { "epoch": 0.9, "learning_rate": 5.129810452029194e-06, "loss": 0.6984, "step": 15878 }, { "epoch": 0.9, "learning_rate": 5.124003377490582e-06, "loss": 0.7068, "step": 15879 }, { "epoch": 0.9, "learning_rate": 5.118199505265664e-06, "loss": 0.6999, "step": 15880 }, { "epoch": 0.9, "learning_rate": 5.1123988355503475e-06, "loss": 0.6975, "step": 15881 }, { "epoch": 0.9, "learning_rate": 5.10660136854042e-06, "loss": 0.6973, "step": 15882 }, { "epoch": 0.9, "learning_rate": 5.100807104431571e-06, "loss": 0.6436, "step": 15883 }, { "epoch": 0.9, "learning_rate": 5.095016043419365e-06, "loss": 0.5487, "step": 15884 }, { "epoch": 0.9, "learning_rate": 5.0892281856992685e-06, "loss": 0.6459, "step": 15885 }, { "epoch": 0.9, "learning_rate": 5.083443531466636e-06, "loss": 0.7126, "step": 15886 }, { "epoch": 0.9, "learning_rate": 5.077662080916701e-06, "loss": 0.7151, "step": 15887 }, { "epoch": 0.9, "learning_rate": 5.071883834244617e-06, "loss": 0.679, "step": 15888 }, { "epoch": 0.9, "learning_rate": 5.066108791645408e-06, "loss": 0.5956, "step": 15889 }, { "epoch": 0.9, "learning_rate": 5.060336953314005e-06, "loss": 0.7086, "step": 15890 }, { "epoch": 0.9, "learning_rate": 5.05456831944523e-06, "loss": 0.7058, "step": 15891 }, { "epoch": 0.9, "learning_rate": 5.0488028902337726e-06, "loss": 0.7375, "step": 15892 }, { "epoch": 0.9, "learning_rate": 5.043040665874221e-06, "loss": 0.773, "step": 15893 }, { "epoch": 0.9, "learning_rate": 5.037281646561087e-06, "loss": 0.6799, "step": 15894 }, { "epoch": 0.9, "learning_rate": 5.031525832488748e-06, "loss": 0.8072, "step": 15895 }, { "epoch": 0.9, "learning_rate": 5.025773223851471e-06, "loss": 0.6971, "step": 15896 }, { "epoch": 0.9, "learning_rate": 5.020023820843434e-06, "loss": 0.7353, "step": 15897 }, { "epoch": 0.9, "learning_rate": 5.014277623658692e-06, "loss": 0.6423, "step": 15898 }, { "epoch": 0.9, "learning_rate": 5.008534632491191e-06, "loss": 0.7763, "step": 15899 }, { "epoch": 0.9, "learning_rate": 5.002794847534764e-06, "loss": 0.693, "step": 15900 }, { "epoch": 0.9, "learning_rate": 4.997058268983135e-06, "loss": 0.6573, "step": 15901 }, { "epoch": 0.9, "learning_rate": 4.991324897029959e-06, "loss": 0.7127, "step": 15902 }, { "epoch": 0.9, "learning_rate": 4.985594731868726e-06, "loss": 0.6292, "step": 15903 }, { "epoch": 0.9, "learning_rate": 4.979867773692881e-06, "loss": 0.763, "step": 15904 }, { "epoch": 0.9, "learning_rate": 4.97414402269567e-06, "loss": 0.653, "step": 15905 }, { "epoch": 0.9, "learning_rate": 4.968423479070327e-06, "loss": 0.7389, "step": 15906 }, { "epoch": 0.9, "learning_rate": 4.962706143009932e-06, "loss": 0.6051, "step": 15907 }, { "epoch": 0.9, "learning_rate": 4.9569920147074285e-06, "loss": 0.681, "step": 15908 }, { "epoch": 0.9, "learning_rate": 4.951281094355709e-06, "loss": 0.7202, "step": 15909 }, { "epoch": 0.9, "learning_rate": 4.945573382147517e-06, "loss": 0.6298, "step": 15910 }, { "epoch": 0.9, "learning_rate": 4.939868878275544e-06, "loss": 0.7272, "step": 15911 }, { "epoch": 0.9, "learning_rate": 4.9341675829322695e-06, "loss": 0.7295, "step": 15912 }, { "epoch": 0.9, "learning_rate": 4.928469496310173e-06, "loss": 0.6694, "step": 15913 }, { "epoch": 0.9, "learning_rate": 4.922774618601567e-06, "loss": 0.7146, "step": 15914 }, { "epoch": 0.9, "learning_rate": 4.9170829499986635e-06, "loss": 0.649, "step": 15915 }, { "epoch": 0.9, "learning_rate": 4.911394490693566e-06, "loss": 0.5595, "step": 15916 }, { "epoch": 0.9, "learning_rate": 4.905709240878287e-06, "loss": 0.7067, "step": 15917 }, { "epoch": 0.9, "learning_rate": 4.9000272007447275e-06, "loss": 0.641, "step": 15918 }, { "epoch": 0.9, "learning_rate": 4.8943483704846475e-06, "loss": 0.7026, "step": 15919 }, { "epoch": 0.9, "learning_rate": 4.888672750289736e-06, "loss": 0.7397, "step": 15920 }, { "epoch": 0.9, "learning_rate": 4.883000340351551e-06, "loss": 0.6379, "step": 15921 }, { "epoch": 0.9, "learning_rate": 4.877331140861574e-06, "loss": 0.6816, "step": 15922 }, { "epoch": 0.9, "learning_rate": 4.871665152011129e-06, "loss": 0.6447, "step": 15923 }, { "epoch": 0.9, "learning_rate": 4.866002373991474e-06, "loss": 0.7773, "step": 15924 }, { "epoch": 0.9, "learning_rate": 4.860342806993734e-06, "loss": 0.7139, "step": 15925 }, { "epoch": 0.9, "learning_rate": 4.854686451208923e-06, "loss": 0.6878, "step": 15926 }, { "epoch": 0.9, "learning_rate": 4.849033306827988e-06, "loss": 0.7465, "step": 15927 }, { "epoch": 0.9, "learning_rate": 4.84338337404171e-06, "loss": 0.7356, "step": 15928 }, { "epoch": 0.9, "learning_rate": 4.8377366530408254e-06, "loss": 0.6679, "step": 15929 }, { "epoch": 0.9, "learning_rate": 4.8320931440158814e-06, "loss": 0.7568, "step": 15930 }, { "epoch": 0.9, "learning_rate": 4.826452847157392e-06, "loss": 0.7481, "step": 15931 }, { "epoch": 0.9, "learning_rate": 4.820815762655706e-06, "loss": 0.7065, "step": 15932 }, { "epoch": 0.9, "learning_rate": 4.815181890701104e-06, "loss": 0.7332, "step": 15933 }, { "epoch": 0.9, "learning_rate": 4.809551231483744e-06, "loss": 0.7015, "step": 15934 }, { "epoch": 0.9, "learning_rate": 4.803923785193687e-06, "loss": 0.7293, "step": 15935 }, { "epoch": 0.9, "learning_rate": 4.798299552020869e-06, "loss": 0.7216, "step": 15936 }, { "epoch": 0.9, "learning_rate": 4.792678532155115e-06, "loss": 0.7199, "step": 15937 }, { "epoch": 0.9, "learning_rate": 4.7870607257861415e-06, "loss": 0.6994, "step": 15938 }, { "epoch": 0.9, "learning_rate": 4.7814461331035735e-06, "loss": 0.7229, "step": 15939 }, { "epoch": 0.9, "learning_rate": 4.775834754296915e-06, "loss": 0.6609, "step": 15940 }, { "epoch": 0.9, "learning_rate": 4.770226589555571e-06, "loss": 0.7138, "step": 15941 }, { "epoch": 0.9, "learning_rate": 4.764621639068823e-06, "loss": 0.7575, "step": 15942 }, { "epoch": 0.9, "learning_rate": 4.759019903025874e-06, "loss": 0.7057, "step": 15943 }, { "epoch": 0.9, "learning_rate": 4.753421381615786e-06, "loss": 0.6358, "step": 15944 }, { "epoch": 0.9, "learning_rate": 4.747826075027506e-06, "loss": 0.7499, "step": 15945 }, { "epoch": 0.9, "learning_rate": 4.742233983449906e-06, "loss": 0.7179, "step": 15946 }, { "epoch": 0.9, "learning_rate": 4.736645107071724e-06, "loss": 0.636, "step": 15947 }, { "epoch": 0.9, "learning_rate": 4.731059446081609e-06, "loss": 0.7308, "step": 15948 }, { "epoch": 0.9, "learning_rate": 4.72547700066811e-06, "loss": 0.7852, "step": 15949 }, { "epoch": 0.9, "learning_rate": 4.7198977710196105e-06, "loss": 0.6503, "step": 15950 }, { "epoch": 0.9, "learning_rate": 4.714321757324447e-06, "loss": 0.6562, "step": 15951 }, { "epoch": 0.9, "learning_rate": 4.708748959770826e-06, "loss": 0.6899, "step": 15952 }, { "epoch": 0.9, "learning_rate": 4.703179378546818e-06, "loss": 0.7288, "step": 15953 }, { "epoch": 0.9, "learning_rate": 4.697613013840441e-06, "loss": 0.7129, "step": 15954 }, { "epoch": 0.91, "learning_rate": 4.692049865839565e-06, "loss": 0.7072, "step": 15955 }, { "epoch": 0.91, "learning_rate": 4.686489934731975e-06, "loss": 0.7123, "step": 15956 }, { "epoch": 0.91, "learning_rate": 4.680933220705308e-06, "loss": 0.6943, "step": 15957 }, { "epoch": 0.91, "learning_rate": 4.675379723947115e-06, "loss": 0.7048, "step": 15958 }, { "epoch": 0.91, "learning_rate": 4.6698294446448795e-06, "loss": 0.6888, "step": 15959 }, { "epoch": 0.91, "learning_rate": 4.664282382985896e-06, "loss": 0.6477, "step": 15960 }, { "epoch": 0.91, "learning_rate": 4.658738539157403e-06, "loss": 0.6923, "step": 15961 }, { "epoch": 0.91, "learning_rate": 4.653197913346541e-06, "loss": 0.6989, "step": 15962 }, { "epoch": 0.91, "learning_rate": 4.647660505740303e-06, "loss": 0.7686, "step": 15963 }, { "epoch": 0.91, "learning_rate": 4.642126316525586e-06, "loss": 0.7434, "step": 15964 }, { "epoch": 0.91, "learning_rate": 4.636595345889194e-06, "loss": 0.5967, "step": 15965 }, { "epoch": 0.91, "learning_rate": 4.631067594017824e-06, "loss": 0.5919, "step": 15966 }, { "epoch": 0.91, "learning_rate": 4.625543061098015e-06, "loss": 0.5718, "step": 15967 }, { "epoch": 0.91, "learning_rate": 4.620021747316261e-06, "loss": 0.7109, "step": 15968 }, { "epoch": 0.91, "learning_rate": 4.614503652858926e-06, "loss": 0.7232, "step": 15969 }, { "epoch": 0.91, "learning_rate": 4.608988777912249e-06, "loss": 0.7468, "step": 15970 }, { "epoch": 0.91, "learning_rate": 4.6034771226623585e-06, "loss": 0.6409, "step": 15971 }, { "epoch": 0.91, "learning_rate": 4.5979686872953065e-06, "loss": 0.6522, "step": 15972 }, { "epoch": 0.91, "learning_rate": 4.592463471997022e-06, "loss": 0.6855, "step": 15973 }, { "epoch": 0.91, "learning_rate": 4.586961476953311e-06, "loss": 0.6394, "step": 15974 }, { "epoch": 0.91, "learning_rate": 4.58146270234987e-06, "loss": 0.679, "step": 15975 }, { "epoch": 0.91, "learning_rate": 4.575967148372317e-06, "loss": 0.682, "step": 15976 }, { "epoch": 0.91, "learning_rate": 4.5704748152061275e-06, "loss": 0.7327, "step": 15977 }, { "epoch": 0.91, "learning_rate": 4.564985703036684e-06, "loss": 0.7521, "step": 15978 }, { "epoch": 0.91, "learning_rate": 4.559499812049251e-06, "loss": 0.6375, "step": 15979 }, { "epoch": 0.91, "learning_rate": 4.5540171424290125e-06, "loss": 0.6839, "step": 15980 }, { "epoch": 0.91, "learning_rate": 4.548537694361021e-06, "loss": 0.6772, "step": 15981 }, { "epoch": 0.91, "learning_rate": 4.543061468030208e-06, "loss": 0.6868, "step": 15982 }, { "epoch": 0.91, "learning_rate": 4.537588463621412e-06, "loss": 0.6826, "step": 15983 }, { "epoch": 0.91, "learning_rate": 4.532118681319353e-06, "loss": 0.7459, "step": 15984 }, { "epoch": 0.91, "learning_rate": 4.526652121308672e-06, "loss": 0.7383, "step": 15985 }, { "epoch": 0.91, "learning_rate": 4.521188783773867e-06, "loss": 0.7219, "step": 15986 }, { "epoch": 0.91, "learning_rate": 4.515728668899344e-06, "loss": 0.5883, "step": 15987 }, { "epoch": 0.91, "learning_rate": 4.510271776869401e-06, "loss": 0.7123, "step": 15988 }, { "epoch": 0.91, "learning_rate": 4.504818107868225e-06, "loss": 0.6332, "step": 15989 }, { "epoch": 0.91, "learning_rate": 4.499367662079868e-06, "loss": 0.7404, "step": 15990 }, { "epoch": 0.91, "learning_rate": 4.493920439688315e-06, "loss": 0.6593, "step": 15991 }, { "epoch": 0.91, "learning_rate": 4.48847644087741e-06, "loss": 0.7, "step": 15992 }, { "epoch": 0.91, "learning_rate": 4.483035665830926e-06, "loss": 0.7185, "step": 15993 }, { "epoch": 0.91, "learning_rate": 4.477598114732495e-06, "loss": 0.6183, "step": 15994 }, { "epoch": 0.91, "learning_rate": 4.4721637877656375e-06, "loss": 0.6859, "step": 15995 }, { "epoch": 0.91, "learning_rate": 4.4667326851137945e-06, "loss": 0.7811, "step": 15996 }, { "epoch": 0.91, "learning_rate": 4.461304806960254e-06, "loss": 0.7281, "step": 15997 }, { "epoch": 0.91, "learning_rate": 4.455880153488234e-06, "loss": 0.6991, "step": 15998 }, { "epoch": 0.91, "learning_rate": 4.450458724880835e-06, "loss": 0.7006, "step": 15999 }, { "epoch": 0.91, "learning_rate": 4.445040521321042e-06, "loss": 0.7035, "step": 16000 }, { "epoch": 0.91, "learning_rate": 4.439625542991743e-06, "loss": 0.7523, "step": 16001 }, { "epoch": 0.91, "learning_rate": 4.434213790075692e-06, "loss": 0.6514, "step": 16002 }, { "epoch": 0.91, "learning_rate": 4.428805262755564e-06, "loss": 0.7549, "step": 16003 }, { "epoch": 0.91, "learning_rate": 4.423399961213892e-06, "loss": 0.8326, "step": 16004 }, { "epoch": 0.91, "learning_rate": 4.41799788563313e-06, "loss": 0.6641, "step": 16005 }, { "epoch": 0.91, "learning_rate": 4.412599036195608e-06, "loss": 0.6192, "step": 16006 }, { "epoch": 0.91, "learning_rate": 4.40720341308356e-06, "loss": 0.7874, "step": 16007 }, { "epoch": 0.91, "learning_rate": 4.401811016479107e-06, "loss": 0.6002, "step": 16008 }, { "epoch": 0.91, "learning_rate": 4.3964218465642355e-06, "loss": 0.697, "step": 16009 }, { "epoch": 0.91, "learning_rate": 4.391035903520857e-06, "loss": 0.8127, "step": 16010 }, { "epoch": 0.91, "learning_rate": 4.38565318753078e-06, "loss": 0.6714, "step": 16011 }, { "epoch": 0.91, "learning_rate": 4.3802736987756496e-06, "loss": 0.7336, "step": 16012 }, { "epoch": 0.91, "learning_rate": 4.374897437437053e-06, "loss": 0.6915, "step": 16013 }, { "epoch": 0.91, "learning_rate": 4.369524403696457e-06, "loss": 0.7085, "step": 16014 }, { "epoch": 0.91, "learning_rate": 4.364154597735215e-06, "loss": 0.6372, "step": 16015 }, { "epoch": 0.91, "learning_rate": 4.358788019734561e-06, "loss": 0.6248, "step": 16016 }, { "epoch": 0.91, "learning_rate": 4.353424669875639e-06, "loss": 0.7315, "step": 16017 }, { "epoch": 0.91, "learning_rate": 4.348064548339492e-06, "loss": 0.7836, "step": 16018 }, { "epoch": 0.91, "learning_rate": 4.34270765530701e-06, "loss": 0.7297, "step": 16019 }, { "epoch": 0.91, "learning_rate": 4.3373539909590144e-06, "loss": 0.6688, "step": 16020 }, { "epoch": 0.91, "learning_rate": 4.332003555476227e-06, "loss": 0.7359, "step": 16021 }, { "epoch": 0.91, "learning_rate": 4.3266563490391934e-06, "loss": 0.8199, "step": 16022 }, { "epoch": 0.91, "learning_rate": 4.321312371828423e-06, "loss": 0.6894, "step": 16023 }, { "epoch": 0.91, "learning_rate": 4.315971624024296e-06, "loss": 0.6729, "step": 16024 }, { "epoch": 0.91, "learning_rate": 4.310634105807065e-06, "loss": 0.6613, "step": 16025 }, { "epoch": 0.91, "learning_rate": 4.3052998173568895e-06, "loss": 0.7077, "step": 16026 }, { "epoch": 0.91, "learning_rate": 4.299968758853812e-06, "loss": 0.6034, "step": 16027 }, { "epoch": 0.91, "learning_rate": 4.294640930477767e-06, "loss": 0.6798, "step": 16028 }, { "epoch": 0.91, "learning_rate": 4.2893163324085885e-06, "loss": 0.7056, "step": 16029 }, { "epoch": 0.91, "learning_rate": 4.283994964825988e-06, "loss": 0.7796, "step": 16030 }, { "epoch": 0.91, "learning_rate": 4.278676827909589e-06, "loss": 0.6438, "step": 16031 }, { "epoch": 0.91, "learning_rate": 4.273361921838881e-06, "loss": 0.7414, "step": 16032 }, { "epoch": 0.91, "learning_rate": 4.268050246793276e-06, "loss": 0.7079, "step": 16033 }, { "epoch": 0.91, "learning_rate": 4.262741802952042e-06, "loss": 0.6975, "step": 16034 }, { "epoch": 0.91, "learning_rate": 4.257436590494334e-06, "loss": 0.6729, "step": 16035 }, { "epoch": 0.91, "learning_rate": 4.252134609599245e-06, "loss": 0.7023, "step": 16036 }, { "epoch": 0.91, "learning_rate": 4.2468358604457085e-06, "loss": 0.7093, "step": 16037 }, { "epoch": 0.91, "learning_rate": 4.241540343212591e-06, "loss": 0.6573, "step": 16038 }, { "epoch": 0.91, "learning_rate": 4.236248058078618e-06, "loss": 0.7245, "step": 16039 }, { "epoch": 0.91, "learning_rate": 4.230959005222445e-06, "loss": 0.6635, "step": 16040 }, { "epoch": 0.91, "learning_rate": 4.225673184822554e-06, "loss": 0.6729, "step": 16041 }, { "epoch": 0.91, "learning_rate": 4.220390597057366e-06, "loss": 0.7023, "step": 16042 }, { "epoch": 0.91, "learning_rate": 4.215111242105196e-06, "loss": 0.6667, "step": 16043 }, { "epoch": 0.91, "learning_rate": 4.209835120144212e-06, "loss": 0.7162, "step": 16044 }, { "epoch": 0.91, "learning_rate": 4.204562231352516e-06, "loss": 0.6304, "step": 16045 }, { "epoch": 0.91, "learning_rate": 4.199292575908098e-06, "loss": 0.684, "step": 16046 }, { "epoch": 0.91, "learning_rate": 4.194026153988783e-06, "loss": 0.7277, "step": 16047 }, { "epoch": 0.91, "learning_rate": 4.1887629657723524e-06, "loss": 0.7094, "step": 16048 }, { "epoch": 0.91, "learning_rate": 4.18350301143644e-06, "loss": 0.5399, "step": 16049 }, { "epoch": 0.91, "learning_rate": 4.178246291158594e-06, "loss": 0.7493, "step": 16050 }, { "epoch": 0.91, "learning_rate": 4.172992805116227e-06, "loss": 0.6785, "step": 16051 }, { "epoch": 0.91, "learning_rate": 4.167742553486675e-06, "loss": 0.6733, "step": 16052 }, { "epoch": 0.91, "learning_rate": 4.162495536447153e-06, "loss": 0.722, "step": 16053 }, { "epoch": 0.91, "learning_rate": 4.1572517541747294e-06, "loss": 0.7683, "step": 16054 }, { "epoch": 0.91, "learning_rate": 4.1520112068464405e-06, "loss": 0.6284, "step": 16055 }, { "epoch": 0.91, "learning_rate": 4.146773894639122e-06, "loss": 0.6343, "step": 16056 }, { "epoch": 0.91, "learning_rate": 4.141539817729567e-06, "loss": 0.7412, "step": 16057 }, { "epoch": 0.91, "learning_rate": 4.136308976294457e-06, "loss": 0.7078, "step": 16058 }, { "epoch": 0.91, "learning_rate": 4.131081370510326e-06, "loss": 0.6546, "step": 16059 }, { "epoch": 0.91, "learning_rate": 4.125857000553624e-06, "loss": 0.6645, "step": 16060 }, { "epoch": 0.91, "learning_rate": 4.120635866600675e-06, "loss": 0.7554, "step": 16061 }, { "epoch": 0.91, "learning_rate": 4.11541796882774e-06, "loss": 0.6712, "step": 16062 }, { "epoch": 0.91, "learning_rate": 4.1102033074108985e-06, "loss": 0.6427, "step": 16063 }, { "epoch": 0.91, "learning_rate": 4.1049918825261765e-06, "loss": 0.735, "step": 16064 }, { "epoch": 0.91, "learning_rate": 4.0997836943494795e-06, "loss": 0.6974, "step": 16065 }, { "epoch": 0.91, "learning_rate": 4.094578743056598e-06, "loss": 0.6463, "step": 16066 }, { "epoch": 0.91, "learning_rate": 4.0893770288231825e-06, "loss": 0.7861, "step": 16067 }, { "epoch": 0.91, "learning_rate": 4.084178551824835e-06, "loss": 0.7313, "step": 16068 }, { "epoch": 0.91, "learning_rate": 4.078983312237017e-06, "loss": 0.702, "step": 16069 }, { "epoch": 0.91, "learning_rate": 4.073791310235075e-06, "loss": 0.598, "step": 16070 }, { "epoch": 0.91, "learning_rate": 4.068602545994249e-06, "loss": 0.707, "step": 16071 }, { "epoch": 0.91, "learning_rate": 4.0634170196896745e-06, "loss": 0.6016, "step": 16072 }, { "epoch": 0.91, "learning_rate": 4.058234731496391e-06, "loss": 0.611, "step": 16073 }, { "epoch": 0.91, "learning_rate": 4.05305568158928e-06, "loss": 0.657, "step": 16074 }, { "epoch": 0.91, "learning_rate": 4.047879870143179e-06, "loss": 0.8294, "step": 16075 }, { "epoch": 0.91, "learning_rate": 4.04270729733277e-06, "loss": 0.8237, "step": 16076 }, { "epoch": 0.91, "learning_rate": 4.037537963332672e-06, "loss": 0.7002, "step": 16077 }, { "epoch": 0.91, "learning_rate": 4.0323718683173194e-06, "loss": 0.6827, "step": 16078 }, { "epoch": 0.91, "learning_rate": 4.027209012461108e-06, "loss": 0.7025, "step": 16079 }, { "epoch": 0.91, "learning_rate": 4.0220493959382765e-06, "loss": 0.7235, "step": 16080 }, { "epoch": 0.91, "learning_rate": 4.016893018922996e-06, "loss": 0.6524, "step": 16081 }, { "epoch": 0.91, "learning_rate": 4.011739881589305e-06, "loss": 0.7001, "step": 16082 }, { "epoch": 0.91, "learning_rate": 4.00658998411112e-06, "loss": 0.7105, "step": 16083 }, { "epoch": 0.91, "learning_rate": 4.00144332666228e-06, "loss": 0.7273, "step": 16084 }, { "epoch": 0.91, "learning_rate": 3.996299909416501e-06, "loss": 0.596, "step": 16085 }, { "epoch": 0.91, "learning_rate": 3.991159732547389e-06, "loss": 0.7365, "step": 16086 }, { "epoch": 0.91, "learning_rate": 3.9860227962284055e-06, "loss": 0.755, "step": 16087 }, { "epoch": 0.91, "learning_rate": 3.9808891006329675e-06, "loss": 0.8129, "step": 16088 }, { "epoch": 0.91, "learning_rate": 3.975758645934336e-06, "loss": 0.7253, "step": 16089 }, { "epoch": 0.91, "learning_rate": 3.970631432305694e-06, "loss": 0.773, "step": 16090 }, { "epoch": 0.91, "learning_rate": 3.9655074599200815e-06, "loss": 0.6956, "step": 16091 }, { "epoch": 0.91, "learning_rate": 3.96038672895046e-06, "loss": 0.7411, "step": 16092 }, { "epoch": 0.91, "learning_rate": 3.955269239569659e-06, "loss": 0.7134, "step": 16093 }, { "epoch": 0.91, "learning_rate": 3.9501549919503945e-06, "loss": 0.7197, "step": 16094 }, { "epoch": 0.91, "learning_rate": 3.945043986265295e-06, "loss": 0.6467, "step": 16095 }, { "epoch": 0.91, "learning_rate": 3.93993622268688e-06, "loss": 0.6973, "step": 16096 }, { "epoch": 0.91, "learning_rate": 3.934831701387554e-06, "loss": 0.7466, "step": 16097 }, { "epoch": 0.91, "learning_rate": 3.929730422539602e-06, "loss": 0.7247, "step": 16098 }, { "epoch": 0.91, "learning_rate": 3.924632386315186e-06, "loss": 0.7056, "step": 16099 }, { "epoch": 0.91, "learning_rate": 3.919537592886413e-06, "loss": 0.7084, "step": 16100 }, { "epoch": 0.91, "learning_rate": 3.914446042425201e-06, "loss": 0.8707, "step": 16101 }, { "epoch": 0.91, "learning_rate": 3.909357735103447e-06, "loss": 0.6312, "step": 16102 }, { "epoch": 0.91, "learning_rate": 3.9042726710928675e-06, "loss": 0.712, "step": 16103 }, { "epoch": 0.91, "learning_rate": 3.899190850565115e-06, "loss": 0.5578, "step": 16104 }, { "epoch": 0.91, "learning_rate": 3.894112273691697e-06, "loss": 0.7077, "step": 16105 }, { "epoch": 0.91, "learning_rate": 3.889036940644031e-06, "loss": 0.7761, "step": 16106 }, { "epoch": 0.91, "learning_rate": 3.883964851593447e-06, "loss": 0.75, "step": 16107 }, { "epoch": 0.91, "learning_rate": 3.878896006711108e-06, "loss": 0.6623, "step": 16108 }, { "epoch": 0.91, "learning_rate": 3.873830406168111e-06, "loss": 0.7217, "step": 16109 }, { "epoch": 0.91, "learning_rate": 3.868768050135452e-06, "loss": 0.6696, "step": 16110 }, { "epoch": 0.91, "learning_rate": 3.863708938783983e-06, "loss": 0.674, "step": 16111 }, { "epoch": 0.91, "learning_rate": 3.858653072284457e-06, "loss": 0.6912, "step": 16112 }, { "epoch": 0.91, "learning_rate": 3.853600450807526e-06, "loss": 0.6897, "step": 16113 }, { "epoch": 0.91, "learning_rate": 3.848551074523754e-06, "loss": 0.7437, "step": 16114 }, { "epoch": 0.91, "learning_rate": 3.843504943603527e-06, "loss": 0.7626, "step": 16115 }, { "epoch": 0.91, "learning_rate": 3.838462058217196e-06, "loss": 0.663, "step": 16116 }, { "epoch": 0.91, "learning_rate": 3.83342241853496e-06, "loss": 0.6218, "step": 16117 }, { "epoch": 0.91, "learning_rate": 3.828386024726938e-06, "loss": 0.7415, "step": 16118 }, { "epoch": 0.91, "learning_rate": 3.823352876963094e-06, "loss": 0.7671, "step": 16119 }, { "epoch": 0.91, "learning_rate": 3.8183229754133245e-06, "loss": 0.6497, "step": 16120 }, { "epoch": 0.91, "learning_rate": 3.813296320247395e-06, "loss": 0.7153, "step": 16121 }, { "epoch": 0.91, "learning_rate": 3.80827291163498e-06, "loss": 0.7175, "step": 16122 }, { "epoch": 0.91, "learning_rate": 3.803252749745623e-06, "loss": 0.7856, "step": 16123 }, { "epoch": 0.91, "learning_rate": 3.798235834748787e-06, "loss": 0.7578, "step": 16124 }, { "epoch": 0.91, "learning_rate": 3.79322216681377e-06, "loss": 0.6677, "step": 16125 }, { "epoch": 0.91, "learning_rate": 3.7882117461098263e-06, "loss": 0.6415, "step": 16126 }, { "epoch": 0.91, "learning_rate": 3.783204572806054e-06, "loss": 0.7484, "step": 16127 }, { "epoch": 0.91, "learning_rate": 3.7782006470714616e-06, "loss": 0.7564, "step": 16128 }, { "epoch": 0.91, "learning_rate": 3.7731999690749585e-06, "loss": 0.5904, "step": 16129 }, { "epoch": 0.91, "learning_rate": 3.7682025389853215e-06, "loss": 0.6673, "step": 16130 }, { "epoch": 0.92, "learning_rate": 3.7632083569712262e-06, "loss": 0.6522, "step": 16131 }, { "epoch": 0.92, "learning_rate": 3.7582174232012377e-06, "loss": 0.7835, "step": 16132 }, { "epoch": 0.92, "learning_rate": 3.7532297378438e-06, "loss": 0.6585, "step": 16133 }, { "epoch": 0.92, "learning_rate": 3.748245301067288e-06, "loss": 0.7132, "step": 16134 }, { "epoch": 0.92, "learning_rate": 3.7432641130399236e-06, "loss": 0.6977, "step": 16135 }, { "epoch": 0.92, "learning_rate": 3.7382861739298393e-06, "loss": 0.7465, "step": 16136 }, { "epoch": 0.92, "learning_rate": 3.7333114839050663e-06, "loss": 0.8097, "step": 16137 }, { "epoch": 0.92, "learning_rate": 3.7283400431334827e-06, "loss": 0.7148, "step": 16138 }, { "epoch": 0.92, "learning_rate": 3.7233718517828976e-06, "loss": 0.6891, "step": 16139 }, { "epoch": 0.92, "learning_rate": 3.718406910021022e-06, "loss": 0.7233, "step": 16140 }, { "epoch": 0.92, "learning_rate": 3.71344521801541e-06, "loss": 0.7671, "step": 16141 }, { "epoch": 0.92, "learning_rate": 3.7084867759335397e-06, "loss": 0.6951, "step": 16142 }, { "epoch": 0.92, "learning_rate": 3.7035315839427986e-06, "loss": 0.8043, "step": 16143 }, { "epoch": 0.92, "learning_rate": 3.698579642210398e-06, "loss": 0.685, "step": 16144 }, { "epoch": 0.92, "learning_rate": 3.6936309509034927e-06, "loss": 0.7174, "step": 16145 }, { "epoch": 0.92, "learning_rate": 3.688685510189116e-06, "loss": 0.688, "step": 16146 }, { "epoch": 0.92, "learning_rate": 3.68374332023419e-06, "loss": 0.8083, "step": 16147 }, { "epoch": 0.92, "learning_rate": 3.678804381205514e-06, "loss": 0.7809, "step": 16148 }, { "epoch": 0.92, "learning_rate": 3.673868693269822e-06, "loss": 0.7292, "step": 16149 }, { "epoch": 0.92, "learning_rate": 3.6689362565936803e-06, "loss": 0.7319, "step": 16150 }, { "epoch": 0.92, "learning_rate": 3.664007071343578e-06, "loss": 0.6338, "step": 16151 }, { "epoch": 0.92, "learning_rate": 3.6590811376858823e-06, "loss": 0.6939, "step": 16152 }, { "epoch": 0.92, "learning_rate": 3.6541584557868604e-06, "loss": 0.6676, "step": 16153 }, { "epoch": 0.92, "learning_rate": 3.6492390258126673e-06, "loss": 0.6752, "step": 16154 }, { "epoch": 0.92, "learning_rate": 3.644322847929338e-06, "loss": 0.7556, "step": 16155 }, { "epoch": 0.92, "learning_rate": 3.639409922302839e-06, "loss": 0.6968, "step": 16156 }, { "epoch": 0.92, "learning_rate": 3.6345002490989487e-06, "loss": 0.6021, "step": 16157 }, { "epoch": 0.92, "learning_rate": 3.629593828483413e-06, "loss": 0.7455, "step": 16158 }, { "epoch": 0.92, "learning_rate": 3.624690660621832e-06, "loss": 0.7228, "step": 16159 }, { "epoch": 0.92, "learning_rate": 3.6197907456796743e-06, "loss": 0.6639, "step": 16160 }, { "epoch": 0.92, "learning_rate": 3.614894083822362e-06, "loss": 0.732, "step": 16161 }, { "epoch": 0.92, "learning_rate": 3.6100006752151415e-06, "loss": 0.7174, "step": 16162 }, { "epoch": 0.92, "learning_rate": 3.605110520023203e-06, "loss": 0.6745, "step": 16163 }, { "epoch": 0.92, "learning_rate": 3.6002236184115803e-06, "loss": 0.6584, "step": 16164 }, { "epoch": 0.92, "learning_rate": 3.5953399705452305e-06, "loss": 0.7239, "step": 16165 }, { "epoch": 0.92, "learning_rate": 3.5904595765890005e-06, "loss": 0.6853, "step": 16166 }, { "epoch": 0.92, "learning_rate": 3.58558243670758e-06, "loss": 0.702, "step": 16167 }, { "epoch": 0.92, "learning_rate": 3.5807085510656146e-06, "loss": 0.694, "step": 16168 }, { "epoch": 0.92, "learning_rate": 3.5758379198276183e-06, "loss": 0.632, "step": 16169 }, { "epoch": 0.92, "learning_rate": 3.5709705431579587e-06, "loss": 0.656, "step": 16170 }, { "epoch": 0.92, "learning_rate": 3.5661064212209493e-06, "loss": 0.6728, "step": 16171 }, { "epoch": 0.92, "learning_rate": 3.5612455541807367e-06, "loss": 0.6526, "step": 16172 }, { "epoch": 0.92, "learning_rate": 3.5563879422014224e-06, "loss": 0.7166, "step": 16173 }, { "epoch": 0.92, "learning_rate": 3.5515335854469534e-06, "loss": 0.7528, "step": 16174 }, { "epoch": 0.92, "learning_rate": 3.546682484081154e-06, "loss": 0.6975, "step": 16175 }, { "epoch": 0.92, "learning_rate": 3.5418346382677935e-06, "loss": 0.6709, "step": 16176 }, { "epoch": 0.92, "learning_rate": 3.5369900481704744e-06, "loss": 0.7805, "step": 16177 }, { "epoch": 0.92, "learning_rate": 3.532148713952732e-06, "loss": 0.7186, "step": 16178 }, { "epoch": 0.92, "learning_rate": 3.5273106357779585e-06, "loss": 0.6864, "step": 16179 }, { "epoch": 0.92, "learning_rate": 3.522475813809467e-06, "loss": 0.7904, "step": 16180 }, { "epoch": 0.92, "learning_rate": 3.517644248210439e-06, "loss": 0.691, "step": 16181 }, { "epoch": 0.92, "learning_rate": 3.512815939143954e-06, "loss": 0.6924, "step": 16182 }, { "epoch": 0.92, "learning_rate": 3.5079908867729717e-06, "loss": 0.6649, "step": 16183 }, { "epoch": 0.92, "learning_rate": 3.503169091260361e-06, "loss": 0.7253, "step": 16184 }, { "epoch": 0.92, "learning_rate": 3.4983505527688586e-06, "loss": 0.7474, "step": 16185 }, { "epoch": 0.92, "learning_rate": 3.4935352714611123e-06, "loss": 0.6399, "step": 16186 }, { "epoch": 0.92, "learning_rate": 3.4887232474996367e-06, "loss": 0.7046, "step": 16187 }, { "epoch": 0.92, "learning_rate": 3.4839144810468906e-06, "loss": 0.7438, "step": 16188 }, { "epoch": 0.92, "learning_rate": 3.4791089722651436e-06, "loss": 0.6512, "step": 16189 }, { "epoch": 0.92, "learning_rate": 3.474306721316589e-06, "loss": 0.6136, "step": 16190 }, { "epoch": 0.92, "learning_rate": 3.4695077283633416e-06, "loss": 0.7182, "step": 16191 }, { "epoch": 0.92, "learning_rate": 3.46471199356736e-06, "loss": 0.7226, "step": 16192 }, { "epoch": 0.92, "learning_rate": 3.4599195170905263e-06, "loss": 0.6902, "step": 16193 }, { "epoch": 0.92, "learning_rate": 3.455130299094611e-06, "loss": 0.7408, "step": 16194 }, { "epoch": 0.92, "learning_rate": 3.4503443397412295e-06, "loss": 0.6258, "step": 16195 }, { "epoch": 0.92, "learning_rate": 3.445561639191941e-06, "loss": 0.6855, "step": 16196 }, { "epoch": 0.92, "learning_rate": 3.440782197608172e-06, "loss": 0.6629, "step": 16197 }, { "epoch": 0.92, "learning_rate": 3.4360060151512273e-06, "loss": 0.7228, "step": 16198 }, { "epoch": 0.92, "learning_rate": 3.431233091982333e-06, "loss": 0.6914, "step": 16199 }, { "epoch": 0.92, "learning_rate": 3.426463428262583e-06, "loss": 0.6615, "step": 16200 }, { "epoch": 0.92, "learning_rate": 3.421697024152981e-06, "loss": 0.7798, "step": 16201 }, { "epoch": 0.92, "learning_rate": 3.416933879814377e-06, "loss": 0.7271, "step": 16202 }, { "epoch": 0.92, "learning_rate": 3.412173995407564e-06, "loss": 0.7123, "step": 16203 }, { "epoch": 0.92, "learning_rate": 3.40741737109318e-06, "loss": 0.6294, "step": 16204 }, { "epoch": 0.92, "learning_rate": 3.402664007031775e-06, "loss": 0.7139, "step": 16205 }, { "epoch": 0.92, "learning_rate": 3.397913903383809e-06, "loss": 0.7311, "step": 16206 }, { "epoch": 0.92, "learning_rate": 3.393167060309588e-06, "loss": 0.7644, "step": 16207 }, { "epoch": 0.92, "learning_rate": 3.388423477969349e-06, "loss": 0.6861, "step": 16208 }, { "epoch": 0.92, "learning_rate": 3.3836831565231875e-06, "loss": 0.7079, "step": 16209 }, { "epoch": 0.92, "learning_rate": 3.378946096131097e-06, "loss": 0.7017, "step": 16210 }, { "epoch": 0.92, "learning_rate": 3.374212296952983e-06, "loss": 0.6971, "step": 16211 }, { "epoch": 0.92, "learning_rate": 3.3694817591486074e-06, "loss": 0.6998, "step": 16212 }, { "epoch": 0.92, "learning_rate": 3.364754482877652e-06, "loss": 0.6972, "step": 16213 }, { "epoch": 0.92, "learning_rate": 3.3600304682996684e-06, "loss": 0.72, "step": 16214 }, { "epoch": 0.92, "learning_rate": 3.3553097155740953e-06, "loss": 0.7211, "step": 16215 }, { "epoch": 0.92, "learning_rate": 3.350592224860283e-06, "loss": 0.7405, "step": 16216 }, { "epoch": 0.92, "learning_rate": 3.3458779963174593e-06, "loss": 0.6561, "step": 16217 }, { "epoch": 0.92, "learning_rate": 3.3411670301047415e-06, "loss": 0.7671, "step": 16218 }, { "epoch": 0.92, "learning_rate": 3.3364593263811252e-06, "loss": 0.6923, "step": 16219 }, { "epoch": 0.92, "learning_rate": 3.3317548853055156e-06, "loss": 0.6405, "step": 16220 }, { "epoch": 0.92, "learning_rate": 3.3270537070367093e-06, "loss": 0.7437, "step": 16221 }, { "epoch": 0.92, "learning_rate": 3.322355791733367e-06, "loss": 0.7522, "step": 16222 }, { "epoch": 0.92, "learning_rate": 3.3176611395540626e-06, "loss": 0.6649, "step": 16223 }, { "epoch": 0.92, "learning_rate": 3.3129697506572464e-06, "loss": 0.6882, "step": 16224 }, { "epoch": 0.92, "learning_rate": 3.3082816252012926e-06, "loss": 0.5813, "step": 16225 }, { "epoch": 0.92, "learning_rate": 3.3035967633444076e-06, "loss": 0.6926, "step": 16226 }, { "epoch": 0.92, "learning_rate": 3.2989151652447313e-06, "loss": 0.6792, "step": 16227 }, { "epoch": 0.92, "learning_rate": 3.294236831060271e-06, "loss": 0.7641, "step": 16228 }, { "epoch": 0.92, "learning_rate": 3.2895617609489336e-06, "loss": 0.6464, "step": 16229 }, { "epoch": 0.92, "learning_rate": 3.2848899550685154e-06, "loss": 0.6968, "step": 16230 }, { "epoch": 0.92, "learning_rate": 3.2802214135767117e-06, "loss": 0.6798, "step": 16231 }, { "epoch": 0.92, "learning_rate": 3.2755561366310973e-06, "loss": 0.682, "step": 16232 }, { "epoch": 0.92, "learning_rate": 3.2708941243891343e-06, "loss": 0.6744, "step": 16233 }, { "epoch": 0.92, "learning_rate": 3.2662353770081756e-06, "loss": 0.6793, "step": 16234 }, { "epoch": 0.92, "learning_rate": 3.261579894645461e-06, "loss": 0.7551, "step": 16235 }, { "epoch": 0.92, "learning_rate": 3.2569276774581326e-06, "loss": 0.69, "step": 16236 }, { "epoch": 0.92, "learning_rate": 3.252278725603208e-06, "loss": 0.6975, "step": 16237 }, { "epoch": 0.92, "learning_rate": 3.2476330392376077e-06, "loss": 0.6968, "step": 16238 }, { "epoch": 0.92, "learning_rate": 3.242990618518138e-06, "loss": 0.7334, "step": 16239 }, { "epoch": 0.92, "learning_rate": 3.2383514636014857e-06, "loss": 0.6658, "step": 16240 }, { "epoch": 0.92, "learning_rate": 3.2337155746442584e-06, "loss": 0.6251, "step": 16241 }, { "epoch": 0.92, "learning_rate": 3.2290829518028862e-06, "loss": 0.7314, "step": 16242 }, { "epoch": 0.92, "learning_rate": 3.2244535952337562e-06, "loss": 0.7803, "step": 16243 }, { "epoch": 0.92, "learning_rate": 3.2198275050931205e-06, "loss": 0.7055, "step": 16244 }, { "epoch": 0.92, "learning_rate": 3.2152046815371316e-06, "loss": 0.8516, "step": 16245 }, { "epoch": 0.92, "learning_rate": 3.2105851247218098e-06, "loss": 0.7332, "step": 16246 }, { "epoch": 0.92, "learning_rate": 3.2059688348030746e-06, "loss": 0.7283, "step": 16247 }, { "epoch": 0.92, "learning_rate": 3.201355811936757e-06, "loss": 0.6525, "step": 16248 }, { "epoch": 0.92, "learning_rate": 3.1967460562785324e-06, "loss": 0.6796, "step": 16249 }, { "epoch": 0.92, "learning_rate": 3.192139567983998e-06, "loss": 0.6225, "step": 16250 }, { "epoch": 0.92, "learning_rate": 3.187536347208653e-06, "loss": 0.6721, "step": 16251 }, { "epoch": 0.92, "learning_rate": 3.1829363941078385e-06, "loss": 0.6651, "step": 16252 }, { "epoch": 0.92, "learning_rate": 3.1783397088368636e-06, "loss": 0.7057, "step": 16253 }, { "epoch": 0.92, "learning_rate": 3.1737462915508277e-06, "loss": 0.7837, "step": 16254 }, { "epoch": 0.92, "learning_rate": 3.1691561424047945e-06, "loss": 0.7027, "step": 16255 }, { "epoch": 0.92, "learning_rate": 3.1645692615536846e-06, "loss": 0.65, "step": 16256 }, { "epoch": 0.92, "learning_rate": 3.1599856491523196e-06, "loss": 0.6336, "step": 16257 }, { "epoch": 0.92, "learning_rate": 3.155405305355419e-06, "loss": 0.5672, "step": 16258 }, { "epoch": 0.92, "learning_rate": 3.150828230317582e-06, "loss": 0.775, "step": 16259 }, { "epoch": 0.92, "learning_rate": 3.1462544241932735e-06, "loss": 0.5469, "step": 16260 }, { "epoch": 0.92, "learning_rate": 3.1416838871368924e-06, "loss": 0.7047, "step": 16261 }, { "epoch": 0.92, "learning_rate": 3.1371166193026934e-06, "loss": 0.6819, "step": 16262 }, { "epoch": 0.92, "learning_rate": 3.132552620844853e-06, "loss": 0.7106, "step": 16263 }, { "epoch": 0.92, "learning_rate": 3.1279918919173925e-06, "loss": 0.7604, "step": 16264 }, { "epoch": 0.92, "learning_rate": 3.1234344326742657e-06, "loss": 0.7059, "step": 16265 }, { "epoch": 0.92, "learning_rate": 3.1188802432693064e-06, "loss": 0.6695, "step": 16266 }, { "epoch": 0.92, "learning_rate": 3.1143293238562132e-06, "loss": 0.7154, "step": 16267 }, { "epoch": 0.92, "learning_rate": 3.1097816745885965e-06, "loss": 0.6299, "step": 16268 }, { "epoch": 0.92, "learning_rate": 3.105237295619945e-06, "loss": 0.672, "step": 16269 }, { "epoch": 0.92, "learning_rate": 3.100696187103658e-06, "loss": 0.7103, "step": 16270 }, { "epoch": 0.92, "learning_rate": 3.0961583491930012e-06, "loss": 0.7774, "step": 16271 }, { "epoch": 0.92, "learning_rate": 3.091623782041142e-06, "loss": 0.688, "step": 16272 }, { "epoch": 0.92, "learning_rate": 3.087092485801135e-06, "loss": 0.702, "step": 16273 }, { "epoch": 0.92, "learning_rate": 3.082564460625903e-06, "loss": 0.6505, "step": 16274 }, { "epoch": 0.92, "learning_rate": 3.0780397066683008e-06, "loss": 0.7039, "step": 16275 }, { "epoch": 0.92, "learning_rate": 3.0735182240810512e-06, "loss": 0.7622, "step": 16276 }, { "epoch": 0.92, "learning_rate": 3.069000013016776e-06, "loss": 0.7311, "step": 16277 }, { "epoch": 0.92, "learning_rate": 3.064485073627943e-06, "loss": 0.6415, "step": 16278 }, { "epoch": 0.92, "learning_rate": 3.059973406066963e-06, "loss": 0.6162, "step": 16279 }, { "epoch": 0.92, "learning_rate": 3.0554650104861136e-06, "loss": 0.7478, "step": 16280 }, { "epoch": 0.92, "learning_rate": 3.0509598870375743e-06, "loss": 0.6397, "step": 16281 }, { "epoch": 0.92, "learning_rate": 3.0464580358733895e-06, "loss": 0.6287, "step": 16282 }, { "epoch": 0.92, "learning_rate": 3.0419594571455157e-06, "loss": 0.6827, "step": 16283 }, { "epoch": 0.92, "learning_rate": 3.0374641510057977e-06, "loss": 0.7211, "step": 16284 }, { "epoch": 0.92, "learning_rate": 3.032972117605948e-06, "loss": 0.6459, "step": 16285 }, { "epoch": 0.92, "learning_rate": 3.028483357097611e-06, "loss": 0.7425, "step": 16286 }, { "epoch": 0.92, "learning_rate": 3.023997869632267e-06, "loss": 0.6509, "step": 16287 }, { "epoch": 0.92, "learning_rate": 3.0195156553613267e-06, "loss": 0.7048, "step": 16288 }, { "epoch": 0.92, "learning_rate": 3.0150367144360703e-06, "loss": 0.6339, "step": 16289 }, { "epoch": 0.92, "learning_rate": 3.010561047007676e-06, "loss": 0.7789, "step": 16290 }, { "epoch": 0.92, "learning_rate": 3.0060886532272235e-06, "loss": 0.7273, "step": 16291 }, { "epoch": 0.92, "learning_rate": 3.0016195332456474e-06, "loss": 0.7612, "step": 16292 }, { "epoch": 0.92, "learning_rate": 2.9971536872137938e-06, "loss": 0.6725, "step": 16293 }, { "epoch": 0.92, "learning_rate": 2.992691115282409e-06, "loss": 0.7621, "step": 16294 }, { "epoch": 0.92, "learning_rate": 2.988231817602105e-06, "loss": 0.6683, "step": 16295 }, { "epoch": 0.92, "learning_rate": 2.9837757943233957e-06, "loss": 0.6676, "step": 16296 }, { "epoch": 0.92, "learning_rate": 2.9793230455966937e-06, "loss": 0.6711, "step": 16297 }, { "epoch": 0.92, "learning_rate": 2.97487357157229e-06, "loss": 0.6562, "step": 16298 }, { "epoch": 0.92, "learning_rate": 2.970427372400353e-06, "loss": 0.6909, "step": 16299 }, { "epoch": 0.92, "learning_rate": 2.965984448230974e-06, "loss": 0.7204, "step": 16300 }, { "epoch": 0.92, "learning_rate": 2.9615447992140888e-06, "loss": 0.6441, "step": 16301 }, { "epoch": 0.92, "learning_rate": 2.957108425499555e-06, "loss": 0.6188, "step": 16302 }, { "epoch": 0.92, "learning_rate": 2.9526753272371198e-06, "loss": 0.7503, "step": 16303 }, { "epoch": 0.92, "learning_rate": 2.948245504576419e-06, "loss": 0.758, "step": 16304 }, { "epoch": 0.92, "learning_rate": 2.943818957666955e-06, "loss": 0.634, "step": 16305 }, { "epoch": 0.92, "learning_rate": 2.9393956866581306e-06, "loss": 0.796, "step": 16306 }, { "epoch": 0.93, "learning_rate": 2.934975691699271e-06, "loss": 0.6712, "step": 16307 }, { "epoch": 0.93, "learning_rate": 2.9305589729395342e-06, "loss": 0.6728, "step": 16308 }, { "epoch": 0.93, "learning_rate": 2.926145530528002e-06, "loss": 0.726, "step": 16309 }, { "epoch": 0.93, "learning_rate": 2.9217353646136426e-06, "loss": 0.7875, "step": 16310 }, { "epoch": 0.93, "learning_rate": 2.917328475345327e-06, "loss": 0.6743, "step": 16311 }, { "epoch": 0.93, "learning_rate": 2.9129248628717687e-06, "loss": 0.7248, "step": 16312 }, { "epoch": 0.93, "learning_rate": 2.9085245273416163e-06, "loss": 0.7307, "step": 16313 }, { "epoch": 0.93, "learning_rate": 2.904127468903406e-06, "loss": 0.6672, "step": 16314 }, { "epoch": 0.93, "learning_rate": 2.8997336877055194e-06, "loss": 0.7058, "step": 16315 }, { "epoch": 0.93, "learning_rate": 2.895343183896282e-06, "loss": 0.6859, "step": 16316 }, { "epoch": 0.93, "learning_rate": 2.8909559576238865e-06, "loss": 0.6903, "step": 16317 }, { "epoch": 0.93, "learning_rate": 2.8865720090364034e-06, "loss": 0.6614, "step": 16318 }, { "epoch": 0.93, "learning_rate": 2.882191338281792e-06, "loss": 0.9006, "step": 16319 }, { "epoch": 0.93, "learning_rate": 2.8778139455079346e-06, "loss": 0.7679, "step": 16320 }, { "epoch": 0.93, "learning_rate": 2.8734398308625567e-06, "loss": 0.7252, "step": 16321 }, { "epoch": 0.93, "learning_rate": 2.869068994493329e-06, "loss": 0.671, "step": 16322 }, { "epoch": 0.93, "learning_rate": 2.8647014365477453e-06, "loss": 0.7808, "step": 16323 }, { "epoch": 0.93, "learning_rate": 2.8603371571732428e-06, "loss": 0.7407, "step": 16324 }, { "epoch": 0.93, "learning_rate": 2.855976156517104e-06, "loss": 0.6637, "step": 16325 }, { "epoch": 0.93, "learning_rate": 2.851618434726555e-06, "loss": 0.694, "step": 16326 }, { "epoch": 0.93, "learning_rate": 2.8472639919486453e-06, "loss": 0.6101, "step": 16327 }, { "epoch": 0.93, "learning_rate": 2.8429128283303796e-06, "loss": 0.6555, "step": 16328 }, { "epoch": 0.93, "learning_rate": 2.838564944018618e-06, "loss": 0.6948, "step": 16329 }, { "epoch": 0.93, "learning_rate": 2.8342203391600986e-06, "loss": 0.6766, "step": 16330 }, { "epoch": 0.93, "learning_rate": 2.8298790139014817e-06, "loss": 0.6515, "step": 16331 }, { "epoch": 0.93, "learning_rate": 2.825540968389262e-06, "loss": 0.7058, "step": 16332 }, { "epoch": 0.93, "learning_rate": 2.821206202769899e-06, "loss": 0.7571, "step": 16333 }, { "epoch": 0.93, "learning_rate": 2.816874717189677e-06, "loss": 0.6754, "step": 16334 }, { "epoch": 0.93, "learning_rate": 2.8125465117948113e-06, "loss": 0.7209, "step": 16335 }, { "epoch": 0.93, "learning_rate": 2.8082215867313857e-06, "loss": 0.7425, "step": 16336 }, { "epoch": 0.93, "learning_rate": 2.8038999421453826e-06, "loss": 0.6967, "step": 16337 }, { "epoch": 0.93, "learning_rate": 2.799581578182653e-06, "loss": 0.7663, "step": 16338 }, { "epoch": 0.93, "learning_rate": 2.7952664949889685e-06, "loss": 0.7019, "step": 16339 }, { "epoch": 0.93, "learning_rate": 2.790954692709957e-06, "loss": 0.6655, "step": 16340 }, { "epoch": 0.93, "learning_rate": 2.7866461714911694e-06, "loss": 0.7002, "step": 16341 }, { "epoch": 0.93, "learning_rate": 2.7823409314780226e-06, "loss": 0.7214, "step": 16342 }, { "epoch": 0.93, "learning_rate": 2.778038972815855e-06, "loss": 0.6593, "step": 16343 }, { "epoch": 0.93, "learning_rate": 2.7737402956498292e-06, "loss": 0.7259, "step": 16344 }, { "epoch": 0.93, "learning_rate": 2.7694449001250512e-06, "loss": 0.6039, "step": 16345 }, { "epoch": 0.93, "learning_rate": 2.7651527863865044e-06, "loss": 0.6867, "step": 16346 }, { "epoch": 0.93, "learning_rate": 2.7608639545790516e-06, "loss": 0.7411, "step": 16347 }, { "epoch": 0.93, "learning_rate": 2.7565784048474654e-06, "loss": 0.7273, "step": 16348 }, { "epoch": 0.93, "learning_rate": 2.7522961373364078e-06, "loss": 0.7302, "step": 16349 }, { "epoch": 0.93, "learning_rate": 2.748017152190374e-06, "loss": 0.6995, "step": 16350 }, { "epoch": 0.93, "learning_rate": 2.743741449553827e-06, "loss": 0.6944, "step": 16351 }, { "epoch": 0.93, "learning_rate": 2.7394690295710624e-06, "loss": 0.6971, "step": 16352 }, { "epoch": 0.93, "learning_rate": 2.7351998923862865e-06, "loss": 0.7664, "step": 16353 }, { "epoch": 0.93, "learning_rate": 2.730934038143607e-06, "loss": 0.7447, "step": 16354 }, { "epoch": 0.93, "learning_rate": 2.7266714669869964e-06, "loss": 0.7567, "step": 16355 }, { "epoch": 0.93, "learning_rate": 2.7224121790603517e-06, "loss": 0.7636, "step": 16356 }, { "epoch": 0.93, "learning_rate": 2.7181561745073915e-06, "loss": 0.672, "step": 16357 }, { "epoch": 0.93, "learning_rate": 2.7139034534718e-06, "loss": 0.7242, "step": 16358 }, { "epoch": 0.93, "learning_rate": 2.7096540160971184e-06, "loss": 0.6901, "step": 16359 }, { "epoch": 0.93, "learning_rate": 2.7054078625267655e-06, "loss": 0.772, "step": 16360 }, { "epoch": 0.93, "learning_rate": 2.701164992904048e-06, "loss": 0.6907, "step": 16361 }, { "epoch": 0.93, "learning_rate": 2.696925407372208e-06, "loss": 0.7846, "step": 16362 }, { "epoch": 0.93, "learning_rate": 2.6926891060742976e-06, "loss": 0.6536, "step": 16363 }, { "epoch": 0.93, "learning_rate": 2.688456089153346e-06, "loss": 0.6586, "step": 16364 }, { "epoch": 0.93, "learning_rate": 2.6842263567521953e-06, "loss": 0.5767, "step": 16365 }, { "epoch": 0.93, "learning_rate": 2.679999909013642e-06, "loss": 0.6816, "step": 16366 }, { "epoch": 0.93, "learning_rate": 2.675776746080316e-06, "loss": 0.7442, "step": 16367 }, { "epoch": 0.93, "learning_rate": 2.67155686809476e-06, "loss": 0.6532, "step": 16368 }, { "epoch": 0.93, "learning_rate": 2.667340275199426e-06, "loss": 0.7115, "step": 16369 }, { "epoch": 0.93, "learning_rate": 2.663126967536611e-06, "loss": 0.7246, "step": 16370 }, { "epoch": 0.93, "learning_rate": 2.658916945248535e-06, "loss": 0.7306, "step": 16371 }, { "epoch": 0.93, "learning_rate": 2.654710208477307e-06, "loss": 0.7324, "step": 16372 }, { "epoch": 0.93, "learning_rate": 2.65050675736489e-06, "loss": 0.7123, "step": 16373 }, { "epoch": 0.93, "learning_rate": 2.646306592053205e-06, "loss": 0.588, "step": 16374 }, { "epoch": 0.93, "learning_rate": 2.6421097126839712e-06, "loss": 0.7657, "step": 16375 }, { "epoch": 0.93, "learning_rate": 2.6379161193988868e-06, "loss": 0.7361, "step": 16376 }, { "epoch": 0.93, "learning_rate": 2.633725812339449e-06, "loss": 0.6846, "step": 16377 }, { "epoch": 0.93, "learning_rate": 2.6295387916471236e-06, "loss": 0.7461, "step": 16378 }, { "epoch": 0.93, "learning_rate": 2.6253550574632303e-06, "loss": 0.6962, "step": 16379 }, { "epoch": 0.93, "learning_rate": 2.621174609928978e-06, "loss": 0.7562, "step": 16380 }, { "epoch": 0.93, "learning_rate": 2.6169974491854766e-06, "loss": 0.6897, "step": 16381 }, { "epoch": 0.93, "learning_rate": 2.612823575373713e-06, "loss": 0.6868, "step": 16382 }, { "epoch": 0.93, "learning_rate": 2.608652988634541e-06, "loss": 0.6287, "step": 16383 }, { "epoch": 0.93, "learning_rate": 2.6044856891087597e-06, "loss": 0.7576, "step": 16384 }, { "epoch": 0.93, "learning_rate": 2.600321676937012e-06, "loss": 0.6929, "step": 16385 }, { "epoch": 0.93, "learning_rate": 2.5961609522598405e-06, "loss": 0.7083, "step": 16386 }, { "epoch": 0.93, "learning_rate": 2.5920035152176892e-06, "loss": 0.7072, "step": 16387 }, { "epoch": 0.93, "learning_rate": 2.5878493659509006e-06, "loss": 0.7213, "step": 16388 }, { "epoch": 0.93, "learning_rate": 2.583698504599663e-06, "loss": 0.7686, "step": 16389 }, { "epoch": 0.93, "learning_rate": 2.5795509313040758e-06, "loss": 0.7717, "step": 16390 }, { "epoch": 0.93, "learning_rate": 2.575406646204137e-06, "loss": 0.6866, "step": 16391 }, { "epoch": 0.93, "learning_rate": 2.5712656494397245e-06, "loss": 0.6262, "step": 16392 }, { "epoch": 0.93, "learning_rate": 2.5671279411506156e-06, "loss": 0.6724, "step": 16393 }, { "epoch": 0.93, "learning_rate": 2.5629935214764865e-06, "loss": 0.6776, "step": 16394 }, { "epoch": 0.93, "learning_rate": 2.5588623905568377e-06, "loss": 0.7692, "step": 16395 }, { "epoch": 0.93, "learning_rate": 2.554734548531157e-06, "loss": 0.6758, "step": 16396 }, { "epoch": 0.93, "learning_rate": 2.550609995538722e-06, "loss": 0.6889, "step": 16397 }, { "epoch": 0.93, "learning_rate": 2.546488731718777e-06, "loss": 0.7149, "step": 16398 }, { "epoch": 0.93, "learning_rate": 2.5423707572104217e-06, "loss": 0.7416, "step": 16399 }, { "epoch": 0.93, "learning_rate": 2.538256072152634e-06, "loss": 0.6313, "step": 16400 }, { "epoch": 0.93, "learning_rate": 2.5341446766843246e-06, "loss": 0.7313, "step": 16401 }, { "epoch": 0.93, "learning_rate": 2.5300365709442387e-06, "loss": 0.57, "step": 16402 }, { "epoch": 0.93, "learning_rate": 2.5259317550710427e-06, "loss": 0.694, "step": 16403 }, { "epoch": 0.93, "learning_rate": 2.5218302292032816e-06, "loss": 0.6577, "step": 16404 }, { "epoch": 0.93, "learning_rate": 2.5177319934794e-06, "loss": 0.7498, "step": 16405 }, { "epoch": 0.93, "learning_rate": 2.513637048037709e-06, "loss": 0.7033, "step": 16406 }, { "epoch": 0.93, "learning_rate": 2.5095453930164435e-06, "loss": 0.6981, "step": 16407 }, { "epoch": 0.93, "learning_rate": 2.5054570285537147e-06, "loss": 0.7013, "step": 16408 }, { "epoch": 0.93, "learning_rate": 2.5013719547874793e-06, "loss": 0.7772, "step": 16409 }, { "epoch": 0.93, "learning_rate": 2.4972901718556484e-06, "loss": 0.7722, "step": 16410 }, { "epoch": 0.93, "learning_rate": 2.4932116798959904e-06, "loss": 0.6994, "step": 16411 }, { "epoch": 0.93, "learning_rate": 2.489136479046139e-06, "loss": 0.6045, "step": 16412 }, { "epoch": 0.93, "learning_rate": 2.4850645694436736e-06, "loss": 0.7259, "step": 16413 }, { "epoch": 0.93, "learning_rate": 2.4809959512260285e-06, "loss": 0.7272, "step": 16414 }, { "epoch": 0.93, "learning_rate": 2.476930624530516e-06, "loss": 0.7414, "step": 16415 }, { "epoch": 0.93, "learning_rate": 2.4728685894943593e-06, "loss": 0.6286, "step": 16416 }, { "epoch": 0.93, "learning_rate": 2.4688098462546604e-06, "loss": 0.6676, "step": 16417 }, { "epoch": 0.93, "learning_rate": 2.4647543949484207e-06, "loss": 0.6582, "step": 16418 }, { "epoch": 0.93, "learning_rate": 2.4607022357125086e-06, "loss": 0.6775, "step": 16419 }, { "epoch": 0.93, "learning_rate": 2.4566533686837033e-06, "loss": 0.671, "step": 16420 }, { "epoch": 0.93, "learning_rate": 2.4526077939986737e-06, "loss": 0.7859, "step": 16421 }, { "epoch": 0.93, "learning_rate": 2.448565511793943e-06, "loss": 0.6371, "step": 16422 }, { "epoch": 0.93, "learning_rate": 2.44452652220597e-06, "loss": 0.6441, "step": 16423 }, { "epoch": 0.93, "learning_rate": 2.4404908253710778e-06, "loss": 0.6528, "step": 16424 }, { "epoch": 0.93, "learning_rate": 2.4364584214254695e-06, "loss": 0.6526, "step": 16425 }, { "epoch": 0.93, "learning_rate": 2.4324293105052797e-06, "loss": 0.7268, "step": 16426 }, { "epoch": 0.93, "learning_rate": 2.428403492746478e-06, "loss": 0.697, "step": 16427 }, { "epoch": 0.93, "learning_rate": 2.424380968284934e-06, "loss": 0.7881, "step": 16428 }, { "epoch": 0.93, "learning_rate": 2.420361737256438e-06, "loss": 0.76, "step": 16429 }, { "epoch": 0.93, "learning_rate": 2.4163457997966487e-06, "loss": 0.5681, "step": 16430 }, { "epoch": 0.93, "learning_rate": 2.412333156041102e-06, "loss": 0.779, "step": 16431 }, { "epoch": 0.93, "learning_rate": 2.4083238061252567e-06, "loss": 0.7395, "step": 16432 }, { "epoch": 0.93, "learning_rate": 2.404317750184426e-06, "loss": 0.7119, "step": 16433 }, { "epoch": 0.93, "learning_rate": 2.4003149883538356e-06, "loss": 0.6777, "step": 16434 }, { "epoch": 0.93, "learning_rate": 2.396315520768555e-06, "loss": 0.728, "step": 16435 }, { "epoch": 0.93, "learning_rate": 2.3923193475636097e-06, "loss": 0.6885, "step": 16436 }, { "epoch": 0.93, "learning_rate": 2.3883264688738694e-06, "loss": 0.7038, "step": 16437 }, { "epoch": 0.93, "learning_rate": 2.384336884834104e-06, "loss": 0.67, "step": 16438 }, { "epoch": 0.93, "learning_rate": 2.380350595578995e-06, "loss": 0.7742, "step": 16439 }, { "epoch": 0.93, "learning_rate": 2.376367601243057e-06, "loss": 0.7048, "step": 16440 }, { "epoch": 0.93, "learning_rate": 2.3723879019607374e-06, "loss": 0.6303, "step": 16441 }, { "epoch": 0.93, "learning_rate": 2.368411497866363e-06, "loss": 0.7408, "step": 16442 }, { "epoch": 0.93, "learning_rate": 2.364438389094159e-06, "loss": 0.7002, "step": 16443 }, { "epoch": 0.93, "learning_rate": 2.3604685757782074e-06, "loss": 0.5818, "step": 16444 }, { "epoch": 0.93, "learning_rate": 2.356502058052512e-06, "loss": 0.7296, "step": 16445 }, { "epoch": 0.93, "learning_rate": 2.3525388360509547e-06, "loss": 0.7983, "step": 16446 }, { "epoch": 0.93, "learning_rate": 2.3485789099072954e-06, "loss": 0.6903, "step": 16447 }, { "epoch": 0.93, "learning_rate": 2.3446222797552154e-06, "loss": 0.6784, "step": 16448 }, { "epoch": 0.93, "learning_rate": 2.3406689457282304e-06, "loss": 0.7163, "step": 16449 }, { "epoch": 0.93, "learning_rate": 2.3367189079597786e-06, "loss": 0.611, "step": 16450 }, { "epoch": 0.93, "learning_rate": 2.332772166583208e-06, "loss": 0.7648, "step": 16451 }, { "epoch": 0.93, "learning_rate": 2.3288287217317127e-06, "loss": 0.7391, "step": 16452 }, { "epoch": 0.93, "learning_rate": 2.32488857353842e-06, "loss": 0.8036, "step": 16453 }, { "epoch": 0.93, "learning_rate": 2.3209517221362777e-06, "loss": 0.5963, "step": 16454 }, { "epoch": 0.93, "learning_rate": 2.3170181676582026e-06, "loss": 0.7306, "step": 16455 }, { "epoch": 0.93, "learning_rate": 2.313087910236933e-06, "loss": 0.6802, "step": 16456 }, { "epoch": 0.93, "learning_rate": 2.3091609500051515e-06, "loss": 0.7018, "step": 16457 }, { "epoch": 0.93, "learning_rate": 2.3052372870953855e-06, "loss": 0.6466, "step": 16458 }, { "epoch": 0.93, "learning_rate": 2.3013169216400733e-06, "loss": 0.6951, "step": 16459 }, { "epoch": 0.93, "learning_rate": 2.2973998537715424e-06, "loss": 0.7697, "step": 16460 }, { "epoch": 0.93, "learning_rate": 2.2934860836219985e-06, "loss": 0.693, "step": 16461 }, { "epoch": 0.93, "learning_rate": 2.289575611323547e-06, "loss": 0.6373, "step": 16462 }, { "epoch": 0.93, "learning_rate": 2.2856684370081816e-06, "loss": 0.6832, "step": 16463 }, { "epoch": 0.93, "learning_rate": 2.281764560807753e-06, "loss": 0.6639, "step": 16464 }, { "epoch": 0.93, "learning_rate": 2.277863982854056e-06, "loss": 0.7738, "step": 16465 }, { "epoch": 0.93, "learning_rate": 2.2739667032787405e-06, "loss": 0.6642, "step": 16466 }, { "epoch": 0.93, "learning_rate": 2.2700727222133346e-06, "loss": 0.7044, "step": 16467 }, { "epoch": 0.93, "learning_rate": 2.2661820397892883e-06, "loss": 0.6909, "step": 16468 }, { "epoch": 0.93, "learning_rate": 2.262294656137909e-06, "loss": 0.708, "step": 16469 }, { "epoch": 0.93, "learning_rate": 2.2584105713904125e-06, "loss": 0.7275, "step": 16470 }, { "epoch": 0.93, "learning_rate": 2.2545297856778946e-06, "loss": 0.7373, "step": 16471 }, { "epoch": 0.93, "learning_rate": 2.250652299131362e-06, "loss": 0.7079, "step": 16472 }, { "epoch": 0.93, "learning_rate": 2.2467781118816424e-06, "loss": 0.7712, "step": 16473 }, { "epoch": 0.93, "learning_rate": 2.2429072240595427e-06, "loss": 0.7129, "step": 16474 }, { "epoch": 0.93, "learning_rate": 2.239039635795703e-06, "loss": 0.6289, "step": 16475 }, { "epoch": 0.93, "learning_rate": 2.2351753472206528e-06, "loss": 0.6356, "step": 16476 }, { "epoch": 0.93, "learning_rate": 2.2313143584648423e-06, "loss": 0.7276, "step": 16477 }, { "epoch": 0.93, "learning_rate": 2.2274566696585784e-06, "loss": 0.7146, "step": 16478 }, { "epoch": 0.93, "learning_rate": 2.22360228093208e-06, "loss": 0.6654, "step": 16479 }, { "epoch": 0.93, "learning_rate": 2.21975119241542e-06, "loss": 0.6902, "step": 16480 }, { "epoch": 0.93, "learning_rate": 2.215903404238595e-06, "loss": 0.7356, "step": 16481 }, { "epoch": 0.93, "learning_rate": 2.212058916531479e-06, "loss": 0.6809, "step": 16482 }, { "epoch": 0.93, "learning_rate": 2.208217729423823e-06, "loss": 0.6514, "step": 16483 }, { "epoch": 0.94, "learning_rate": 2.2043798430453123e-06, "loss": 0.6736, "step": 16484 }, { "epoch": 0.94, "learning_rate": 2.200545257525444e-06, "loss": 0.7688, "step": 16485 }, { "epoch": 0.94, "learning_rate": 2.196713972993669e-06, "loss": 0.54, "step": 16486 }, { "epoch": 0.94, "learning_rate": 2.192885989579285e-06, "loss": 0.7222, "step": 16487 }, { "epoch": 0.94, "learning_rate": 2.189061307411511e-06, "loss": 0.6866, "step": 16488 }, { "epoch": 0.94, "learning_rate": 2.1852399266194314e-06, "loss": 0.6134, "step": 16489 }, { "epoch": 0.94, "learning_rate": 2.1814218473320434e-06, "loss": 0.6323, "step": 16490 }, { "epoch": 0.94, "learning_rate": 2.1776070696782115e-06, "loss": 0.7067, "step": 16491 }, { "epoch": 0.94, "learning_rate": 2.173795593786687e-06, "loss": 0.6977, "step": 16492 }, { "epoch": 0.94, "learning_rate": 2.1699874197861124e-06, "loss": 0.7053, "step": 16493 }, { "epoch": 0.94, "learning_rate": 2.166182547805029e-06, "loss": 0.5787, "step": 16494 }, { "epoch": 0.94, "learning_rate": 2.162380977971867e-06, "loss": 0.6651, "step": 16495 }, { "epoch": 0.94, "learning_rate": 2.1585827104149247e-06, "loss": 0.6871, "step": 16496 }, { "epoch": 0.94, "learning_rate": 2.154787745262432e-06, "loss": 0.7209, "step": 16497 }, { "epoch": 0.94, "learning_rate": 2.1509960826424535e-06, "loss": 0.6772, "step": 16498 }, { "epoch": 0.94, "learning_rate": 2.1472077226829756e-06, "loss": 0.6568, "step": 16499 }, { "epoch": 0.94, "learning_rate": 2.143422665511874e-06, "loss": 0.6491, "step": 16500 }, { "epoch": 0.94, "learning_rate": 2.1396409112568904e-06, "loss": 0.7081, "step": 16501 }, { "epoch": 0.94, "learning_rate": 2.1358624600456677e-06, "loss": 0.7207, "step": 16502 }, { "epoch": 0.94, "learning_rate": 2.1320873120057484e-06, "loss": 0.7452, "step": 16503 }, { "epoch": 0.94, "learning_rate": 2.128315467264552e-06, "loss": 0.7061, "step": 16504 }, { "epoch": 0.94, "learning_rate": 2.124546925949389e-06, "loss": 0.5957, "step": 16505 }, { "epoch": 0.94, "learning_rate": 2.1207816881874456e-06, "loss": 0.5247, "step": 16506 }, { "epoch": 0.94, "learning_rate": 2.1170197541058314e-06, "loss": 0.8312, "step": 16507 }, { "epoch": 0.94, "learning_rate": 2.1132611238315003e-06, "loss": 0.6444, "step": 16508 }, { "epoch": 0.94, "learning_rate": 2.109505797491318e-06, "loss": 0.7337, "step": 16509 }, { "epoch": 0.94, "learning_rate": 2.105753775212049e-06, "loss": 0.6738, "step": 16510 }, { "epoch": 0.94, "learning_rate": 2.1020050571203375e-06, "loss": 0.7509, "step": 16511 }, { "epoch": 0.94, "learning_rate": 2.0982596433426927e-06, "loss": 0.7713, "step": 16512 }, { "epoch": 0.94, "learning_rate": 2.0945175340055357e-06, "loss": 0.7065, "step": 16513 }, { "epoch": 0.94, "learning_rate": 2.0907787292351878e-06, "loss": 0.7347, "step": 16514 }, { "epoch": 0.94, "learning_rate": 2.0870432291578256e-06, "loss": 0.6175, "step": 16515 }, { "epoch": 0.94, "learning_rate": 2.0833110338995377e-06, "loss": 0.6916, "step": 16516 }, { "epoch": 0.94, "learning_rate": 2.0795821435863116e-06, "loss": 0.7441, "step": 16517 }, { "epoch": 0.94, "learning_rate": 2.0758565583439916e-06, "loss": 0.684, "step": 16518 }, { "epoch": 0.94, "learning_rate": 2.0721342782983213e-06, "loss": 0.7136, "step": 16519 }, { "epoch": 0.94, "learning_rate": 2.0684153035749445e-06, "loss": 0.6965, "step": 16520 }, { "epoch": 0.94, "learning_rate": 2.064699634299383e-06, "loss": 0.6348, "step": 16521 }, { "epoch": 0.94, "learning_rate": 2.06098727059707e-06, "loss": 0.749, "step": 16522 }, { "epoch": 0.94, "learning_rate": 2.057278212593272e-06, "loss": 0.671, "step": 16523 }, { "epoch": 0.94, "learning_rate": 2.0535724604132222e-06, "loss": 0.7216, "step": 16524 }, { "epoch": 0.94, "learning_rate": 2.0498700141819536e-06, "loss": 0.7471, "step": 16525 }, { "epoch": 0.94, "learning_rate": 2.0461708740244555e-06, "loss": 0.6972, "step": 16526 }, { "epoch": 0.94, "learning_rate": 2.0424750400655947e-06, "loss": 0.7427, "step": 16527 }, { "epoch": 0.94, "learning_rate": 2.0387825124300928e-06, "loss": 0.7204, "step": 16528 }, { "epoch": 0.94, "learning_rate": 2.035093291242607e-06, "loss": 0.7453, "step": 16529 }, { "epoch": 0.94, "learning_rate": 2.0314073766276364e-06, "loss": 0.5879, "step": 16530 }, { "epoch": 0.94, "learning_rate": 2.0277247687096155e-06, "loss": 0.6943, "step": 16531 }, { "epoch": 0.94, "learning_rate": 2.0240454676128117e-06, "loss": 0.7626, "step": 16532 }, { "epoch": 0.94, "learning_rate": 2.0203694734614143e-06, "loss": 0.6307, "step": 16533 }, { "epoch": 0.94, "learning_rate": 2.016696786379524e-06, "loss": 0.6929, "step": 16534 }, { "epoch": 0.94, "learning_rate": 2.0130274064910747e-06, "loss": 0.6937, "step": 16535 }, { "epoch": 0.94, "learning_rate": 2.0093613339199455e-06, "loss": 0.722, "step": 16536 }, { "epoch": 0.94, "learning_rate": 2.00569856878986e-06, "loss": 0.7024, "step": 16537 }, { "epoch": 0.94, "learning_rate": 2.0020391112244406e-06, "loss": 0.7179, "step": 16538 }, { "epoch": 0.94, "learning_rate": 1.9983829613472116e-06, "loss": 0.8089, "step": 16539 }, { "epoch": 0.94, "learning_rate": 1.994730119281574e-06, "loss": 0.6695, "step": 16540 }, { "epoch": 0.94, "learning_rate": 1.9910805851508173e-06, "loss": 0.6589, "step": 16541 }, { "epoch": 0.94, "learning_rate": 1.987434359078133e-06, "loss": 0.6159, "step": 16542 }, { "epoch": 0.94, "learning_rate": 1.9837914411866e-06, "loss": 0.6641, "step": 16543 }, { "epoch": 0.94, "learning_rate": 1.980151831599153e-06, "loss": 0.7892, "step": 16544 }, { "epoch": 0.94, "learning_rate": 1.9765155304386494e-06, "loss": 0.5653, "step": 16545 }, { "epoch": 0.94, "learning_rate": 1.9728825378278246e-06, "loss": 0.6607, "step": 16546 }, { "epoch": 0.94, "learning_rate": 1.9692528538892917e-06, "loss": 0.7235, "step": 16547 }, { "epoch": 0.94, "learning_rate": 1.965626478745575e-06, "loss": 0.6743, "step": 16548 }, { "epoch": 0.94, "learning_rate": 1.9620034125190644e-06, "loss": 0.7022, "step": 16549 }, { "epoch": 0.94, "learning_rate": 1.9583836553320635e-06, "loss": 0.7001, "step": 16550 }, { "epoch": 0.94, "learning_rate": 1.9547672073067293e-06, "loss": 0.6739, "step": 16551 }, { "epoch": 0.94, "learning_rate": 1.951154068565142e-06, "loss": 0.6484, "step": 16552 }, { "epoch": 0.94, "learning_rate": 1.9475442392292377e-06, "loss": 0.7174, "step": 16553 }, { "epoch": 0.94, "learning_rate": 1.943937719420863e-06, "loss": 0.7103, "step": 16554 }, { "epoch": 0.94, "learning_rate": 1.9403345092617654e-06, "loss": 0.6834, "step": 16555 }, { "epoch": 0.94, "learning_rate": 1.936734608873547e-06, "loss": 0.6605, "step": 16556 }, { "epoch": 0.94, "learning_rate": 1.9331380183777113e-06, "loss": 0.6854, "step": 16557 }, { "epoch": 0.94, "learning_rate": 1.929544737895661e-06, "loss": 0.6215, "step": 16558 }, { "epoch": 0.94, "learning_rate": 1.925954767548677e-06, "loss": 0.7323, "step": 16559 }, { "epoch": 0.94, "learning_rate": 1.922368107457917e-06, "loss": 0.7278, "step": 16560 }, { "epoch": 0.94, "learning_rate": 1.918784757744463e-06, "loss": 0.5823, "step": 16561 }, { "epoch": 0.94, "learning_rate": 1.915204718529251e-06, "loss": 0.7027, "step": 16562 }, { "epoch": 0.94, "learning_rate": 1.911627989933107e-06, "loss": 0.5851, "step": 16563 }, { "epoch": 0.94, "learning_rate": 1.9080545720767674e-06, "loss": 0.7034, "step": 16564 }, { "epoch": 0.94, "learning_rate": 1.904484465080847e-06, "loss": 0.6082, "step": 16565 }, { "epoch": 0.94, "learning_rate": 1.900917669065838e-06, "loss": 0.7203, "step": 16566 }, { "epoch": 0.94, "learning_rate": 1.8973541841521335e-06, "loss": 0.6559, "step": 16567 }, { "epoch": 0.94, "learning_rate": 1.8937940104600037e-06, "loss": 0.7056, "step": 16568 }, { "epoch": 0.94, "learning_rate": 1.8902371481096304e-06, "loss": 0.5986, "step": 16569 }, { "epoch": 0.94, "learning_rate": 1.8866835972210396e-06, "loss": 0.6906, "step": 16570 }, { "epoch": 0.94, "learning_rate": 1.8831333579141908e-06, "loss": 0.7232, "step": 16571 }, { "epoch": 0.94, "learning_rate": 1.8795864303089217e-06, "loss": 0.8107, "step": 16572 }, { "epoch": 0.94, "learning_rate": 1.8760428145249364e-06, "loss": 0.7546, "step": 16573 }, { "epoch": 0.94, "learning_rate": 1.8725025106818506e-06, "loss": 0.6858, "step": 16574 }, { "epoch": 0.94, "learning_rate": 1.868965518899146e-06, "loss": 0.7541, "step": 16575 }, { "epoch": 0.94, "learning_rate": 1.8654318392962277e-06, "loss": 0.5768, "step": 16576 }, { "epoch": 0.94, "learning_rate": 1.8619014719923332e-06, "loss": 0.613, "step": 16577 }, { "epoch": 0.94, "learning_rate": 1.8583744171066454e-06, "loss": 0.6791, "step": 16578 }, { "epoch": 0.94, "learning_rate": 1.8548506747582129e-06, "loss": 0.7541, "step": 16579 }, { "epoch": 0.94, "learning_rate": 1.851330245065952e-06, "loss": 0.6919, "step": 16580 }, { "epoch": 0.94, "learning_rate": 1.8478131281487232e-06, "loss": 0.6674, "step": 16581 }, { "epoch": 0.94, "learning_rate": 1.8442993241252094e-06, "loss": 0.7215, "step": 16582 }, { "epoch": 0.94, "learning_rate": 1.8407888331140043e-06, "loss": 0.674, "step": 16583 }, { "epoch": 0.94, "learning_rate": 1.8372816552336026e-06, "loss": 0.7276, "step": 16584 }, { "epoch": 0.94, "learning_rate": 1.8337777906023978e-06, "loss": 0.6683, "step": 16585 }, { "epoch": 0.94, "learning_rate": 1.8302772393386402e-06, "loss": 0.7125, "step": 16586 }, { "epoch": 0.94, "learning_rate": 1.8267800015604796e-06, "loss": 0.6144, "step": 16587 }, { "epoch": 0.94, "learning_rate": 1.8232860773859772e-06, "loss": 0.6801, "step": 16588 }, { "epoch": 0.94, "learning_rate": 1.8197954669330498e-06, "loss": 0.6464, "step": 16589 }, { "epoch": 0.94, "learning_rate": 1.8163081703195028e-06, "loss": 0.6714, "step": 16590 }, { "epoch": 0.94, "learning_rate": 1.8128241876630426e-06, "loss": 0.5681, "step": 16591 }, { "epoch": 0.94, "learning_rate": 1.8093435190812747e-06, "loss": 0.6842, "step": 16592 }, { "epoch": 0.94, "learning_rate": 1.805866164691683e-06, "loss": 0.5599, "step": 16593 }, { "epoch": 0.94, "learning_rate": 1.8023921246116405e-06, "loss": 0.6952, "step": 16594 }, { "epoch": 0.94, "learning_rate": 1.7989213989583754e-06, "loss": 0.6916, "step": 16595 }, { "epoch": 0.94, "learning_rate": 1.7954539878490828e-06, "loss": 0.6613, "step": 16596 }, { "epoch": 0.94, "learning_rate": 1.7919898914007472e-06, "loss": 0.6806, "step": 16597 }, { "epoch": 0.94, "learning_rate": 1.7885291097303081e-06, "loss": 0.7376, "step": 16598 }, { "epoch": 0.94, "learning_rate": 1.7850716429545944e-06, "loss": 0.7153, "step": 16599 }, { "epoch": 0.94, "learning_rate": 1.7816174911902794e-06, "loss": 0.6378, "step": 16600 }, { "epoch": 0.94, "learning_rate": 1.778166654553981e-06, "loss": 0.7029, "step": 16601 }, { "epoch": 0.94, "learning_rate": 1.7747191331621392e-06, "loss": 0.6808, "step": 16602 }, { "epoch": 0.94, "learning_rate": 1.771274927131139e-06, "loss": 0.6673, "step": 16603 }, { "epoch": 0.94, "learning_rate": 1.7678340365772206e-06, "loss": 0.7031, "step": 16604 }, { "epoch": 0.94, "learning_rate": 1.7643964616165243e-06, "loss": 0.7587, "step": 16605 }, { "epoch": 0.94, "learning_rate": 1.7609622023650795e-06, "loss": 0.7533, "step": 16606 }, { "epoch": 0.94, "learning_rate": 1.757531258938805e-06, "loss": 0.8458, "step": 16607 }, { "epoch": 0.94, "learning_rate": 1.7541036314534964e-06, "loss": 0.695, "step": 16608 }, { "epoch": 0.94, "learning_rate": 1.7506793200248506e-06, "loss": 0.6709, "step": 16609 }, { "epoch": 0.94, "learning_rate": 1.7472583247684526e-06, "loss": 0.7638, "step": 16610 }, { "epoch": 0.94, "learning_rate": 1.7438406457997546e-06, "loss": 0.6991, "step": 16611 }, { "epoch": 0.94, "learning_rate": 1.7404262832341312e-06, "loss": 0.6039, "step": 16612 }, { "epoch": 0.94, "learning_rate": 1.737015237186801e-06, "loss": 0.7247, "step": 16613 }, { "epoch": 0.94, "learning_rate": 1.7336075077729275e-06, "loss": 0.6899, "step": 16614 }, { "epoch": 0.94, "learning_rate": 1.7302030951075076e-06, "loss": 0.9016, "step": 16615 }, { "epoch": 0.94, "learning_rate": 1.7268019993054497e-06, "loss": 0.9047, "step": 16616 }, { "epoch": 0.94, "learning_rate": 1.7234042204815615e-06, "loss": 0.7281, "step": 16617 }, { "epoch": 0.94, "learning_rate": 1.720009758750518e-06, "loss": 0.7135, "step": 16618 }, { "epoch": 0.94, "learning_rate": 1.7166186142268947e-06, "loss": 0.7135, "step": 16619 }, { "epoch": 0.94, "learning_rate": 1.7132307870251441e-06, "loss": 0.7521, "step": 16620 }, { "epoch": 0.94, "learning_rate": 1.7098462772596302e-06, "loss": 0.7529, "step": 16621 }, { "epoch": 0.94, "learning_rate": 1.706465085044584e-06, "loss": 0.6641, "step": 16622 }, { "epoch": 0.94, "learning_rate": 1.7030872104941142e-06, "loss": 0.6132, "step": 16623 }, { "epoch": 0.94, "learning_rate": 1.6997126537222407e-06, "loss": 0.6731, "step": 16624 }, { "epoch": 0.94, "learning_rate": 1.6963414148428724e-06, "loss": 0.6827, "step": 16625 }, { "epoch": 0.94, "learning_rate": 1.6929734939697962e-06, "loss": 0.6894, "step": 16626 }, { "epoch": 0.94, "learning_rate": 1.6896088912166986e-06, "loss": 0.7314, "step": 16627 }, { "epoch": 0.94, "learning_rate": 1.6862476066971e-06, "loss": 0.6857, "step": 16628 }, { "epoch": 0.94, "learning_rate": 1.6828896405244988e-06, "loss": 0.7055, "step": 16629 }, { "epoch": 0.94, "learning_rate": 1.679534992812215e-06, "loss": 0.588, "step": 16630 }, { "epoch": 0.94, "learning_rate": 1.6761836636734696e-06, "loss": 0.7686, "step": 16631 }, { "epoch": 0.94, "learning_rate": 1.672835653221405e-06, "loss": 0.7063, "step": 16632 }, { "epoch": 0.94, "learning_rate": 1.669490961569009e-06, "loss": 0.7043, "step": 16633 }, { "epoch": 0.94, "learning_rate": 1.666149588829169e-06, "loss": 0.7468, "step": 16634 }, { "epoch": 0.94, "learning_rate": 1.6628115351146722e-06, "loss": 0.6806, "step": 16635 }, { "epoch": 0.94, "learning_rate": 1.6594768005381733e-06, "loss": 0.7118, "step": 16636 }, { "epoch": 0.94, "learning_rate": 1.6561453852122488e-06, "loss": 0.6567, "step": 16637 }, { "epoch": 0.94, "learning_rate": 1.652817289249331e-06, "loss": 0.705, "step": 16638 }, { "epoch": 0.94, "learning_rate": 1.6494925127617634e-06, "loss": 0.7082, "step": 16639 }, { "epoch": 0.94, "learning_rate": 1.646171055861767e-06, "loss": 0.7067, "step": 16640 }, { "epoch": 0.94, "learning_rate": 1.6428529186614195e-06, "loss": 0.5797, "step": 16641 }, { "epoch": 0.94, "learning_rate": 1.6395381012727418e-06, "loss": 0.8025, "step": 16642 }, { "epoch": 0.94, "learning_rate": 1.6362266038076113e-06, "loss": 0.6302, "step": 16643 }, { "epoch": 0.94, "learning_rate": 1.632918426377794e-06, "loss": 0.6601, "step": 16644 }, { "epoch": 0.94, "learning_rate": 1.6296135690949677e-06, "loss": 0.7705, "step": 16645 }, { "epoch": 0.94, "learning_rate": 1.626312032070676e-06, "loss": 0.7739, "step": 16646 }, { "epoch": 0.94, "learning_rate": 1.6230138154163298e-06, "loss": 0.7882, "step": 16647 }, { "epoch": 0.94, "learning_rate": 1.6197189192432849e-06, "loss": 0.5675, "step": 16648 }, { "epoch": 0.94, "learning_rate": 1.6164273436627409e-06, "loss": 0.7301, "step": 16649 }, { "epoch": 0.94, "learning_rate": 1.6131390887857755e-06, "loss": 0.7001, "step": 16650 }, { "epoch": 0.94, "learning_rate": 1.6098541547234114e-06, "loss": 0.6925, "step": 16651 }, { "epoch": 0.94, "learning_rate": 1.606572541586504e-06, "loss": 0.7498, "step": 16652 }, { "epoch": 0.94, "learning_rate": 1.6032942494858206e-06, "loss": 0.7141, "step": 16653 }, { "epoch": 0.94, "learning_rate": 1.6000192785320057e-06, "loss": 0.7579, "step": 16654 }, { "epoch": 0.94, "learning_rate": 1.5967476288356042e-06, "loss": 0.6832, "step": 16655 }, { "epoch": 0.94, "learning_rate": 1.5934793005070504e-06, "loss": 0.6544, "step": 16656 }, { "epoch": 0.94, "learning_rate": 1.5902142936566334e-06, "loss": 0.7562, "step": 16657 }, { "epoch": 0.94, "learning_rate": 1.5869526083945874e-06, "loss": 0.6732, "step": 16658 }, { "epoch": 0.94, "learning_rate": 1.583694244830991e-06, "loss": 0.6253, "step": 16659 }, { "epoch": 0.95, "learning_rate": 1.580439203075812e-06, "loss": 0.7151, "step": 16660 }, { "epoch": 0.95, "learning_rate": 1.5771874832389176e-06, "loss": 0.6933, "step": 16661 }, { "epoch": 0.95, "learning_rate": 1.5739390854300762e-06, "loss": 0.6722, "step": 16662 }, { "epoch": 0.95, "learning_rate": 1.5706940097589329e-06, "loss": 0.7364, "step": 16663 }, { "epoch": 0.95, "learning_rate": 1.5674522563350003e-06, "loss": 0.6518, "step": 16664 }, { "epoch": 0.95, "learning_rate": 1.5642138252677019e-06, "loss": 0.5983, "step": 16665 }, { "epoch": 0.95, "learning_rate": 1.5609787166663393e-06, "loss": 0.6811, "step": 16666 }, { "epoch": 0.95, "learning_rate": 1.5577469306401137e-06, "loss": 0.6051, "step": 16667 }, { "epoch": 0.95, "learning_rate": 1.554518467298105e-06, "loss": 0.6557, "step": 16668 }, { "epoch": 0.95, "learning_rate": 1.5512933267492813e-06, "loss": 0.6623, "step": 16669 }, { "epoch": 0.95, "learning_rate": 1.5480715091025e-06, "loss": 0.7146, "step": 16670 }, { "epoch": 0.95, "learning_rate": 1.5448530144665075e-06, "loss": 0.6885, "step": 16671 }, { "epoch": 0.95, "learning_rate": 1.5416378429499278e-06, "loss": 0.6755, "step": 16672 }, { "epoch": 0.95, "learning_rate": 1.5384259946612856e-06, "loss": 0.7324, "step": 16673 }, { "epoch": 0.95, "learning_rate": 1.5352174697089938e-06, "loss": 0.7076, "step": 16674 }, { "epoch": 0.95, "learning_rate": 1.532012268201344e-06, "loss": 0.5387, "step": 16675 }, { "epoch": 0.95, "learning_rate": 1.5288103902465267e-06, "loss": 0.6608, "step": 16676 }, { "epoch": 0.95, "learning_rate": 1.5256118359526117e-06, "loss": 0.6759, "step": 16677 }, { "epoch": 0.95, "learning_rate": 1.5224166054275456e-06, "loss": 0.6702, "step": 16678 }, { "epoch": 0.95, "learning_rate": 1.5192246987791981e-06, "loss": 0.633, "step": 16679 }, { "epoch": 0.95, "learning_rate": 1.5160361161152826e-06, "loss": 0.6437, "step": 16680 }, { "epoch": 0.95, "learning_rate": 1.5128508575434353e-06, "loss": 0.7648, "step": 16681 }, { "epoch": 0.95, "learning_rate": 1.509668923171159e-06, "loss": 0.7251, "step": 16682 }, { "epoch": 0.95, "learning_rate": 1.506490313105857e-06, "loss": 0.7857, "step": 16683 }, { "epoch": 0.95, "learning_rate": 1.5033150274548324e-06, "loss": 0.6855, "step": 16684 }, { "epoch": 0.95, "learning_rate": 1.5001430663252325e-06, "loss": 0.6695, "step": 16685 }, { "epoch": 0.95, "learning_rate": 1.4969744298241272e-06, "loss": 0.7421, "step": 16686 }, { "epoch": 0.95, "learning_rate": 1.4938091180584646e-06, "loss": 0.6543, "step": 16687 }, { "epoch": 0.95, "learning_rate": 1.4906471311350812e-06, "loss": 0.726, "step": 16688 }, { "epoch": 0.95, "learning_rate": 1.4874884691607139e-06, "loss": 0.63, "step": 16689 }, { "epoch": 0.95, "learning_rate": 1.4843331322419774e-06, "loss": 0.6562, "step": 16690 }, { "epoch": 0.95, "learning_rate": 1.481181120485353e-06, "loss": 0.7123, "step": 16691 }, { "epoch": 0.95, "learning_rate": 1.4780324339972562e-06, "loss": 0.6857, "step": 16692 }, { "epoch": 0.95, "learning_rate": 1.4748870728839347e-06, "loss": 0.7044, "step": 16693 }, { "epoch": 0.95, "learning_rate": 1.4717450372515707e-06, "loss": 0.6271, "step": 16694 }, { "epoch": 0.95, "learning_rate": 1.4686063272062013e-06, "loss": 0.7295, "step": 16695 }, { "epoch": 0.95, "learning_rate": 1.465470942853786e-06, "loss": 0.7776, "step": 16696 }, { "epoch": 0.95, "learning_rate": 1.4623388843001518e-06, "loss": 0.6717, "step": 16697 }, { "epoch": 0.95, "learning_rate": 1.4592101516509914e-06, "loss": 0.6703, "step": 16698 }, { "epoch": 0.95, "learning_rate": 1.4560847450119209e-06, "loss": 0.7771, "step": 16699 }, { "epoch": 0.95, "learning_rate": 1.4529626644884443e-06, "loss": 0.7466, "step": 16700 }, { "epoch": 0.95, "learning_rate": 1.4498439101859107e-06, "loss": 0.7297, "step": 16701 }, { "epoch": 0.95, "learning_rate": 1.4467284822096028e-06, "loss": 0.6518, "step": 16702 }, { "epoch": 0.95, "learning_rate": 1.4436163806646807e-06, "loss": 0.7053, "step": 16703 }, { "epoch": 0.95, "learning_rate": 1.4405076056561828e-06, "loss": 0.6787, "step": 16704 }, { "epoch": 0.95, "learning_rate": 1.4374021572890361e-06, "loss": 0.7152, "step": 16705 }, { "epoch": 0.95, "learning_rate": 1.4343000356680458e-06, "loss": 0.6886, "step": 16706 }, { "epoch": 0.95, "learning_rate": 1.4312012408979392e-06, "loss": 0.7018, "step": 16707 }, { "epoch": 0.95, "learning_rate": 1.4281057730832881e-06, "loss": 0.6938, "step": 16708 }, { "epoch": 0.95, "learning_rate": 1.4250136323285868e-06, "loss": 0.6837, "step": 16709 }, { "epoch": 0.95, "learning_rate": 1.4219248187381851e-06, "loss": 0.6996, "step": 16710 }, { "epoch": 0.95, "learning_rate": 1.4188393324163663e-06, "loss": 0.7088, "step": 16711 }, { "epoch": 0.95, "learning_rate": 1.4157571734672581e-06, "loss": 0.6644, "step": 16712 }, { "epoch": 0.95, "learning_rate": 1.4126783419948774e-06, "loss": 0.8176, "step": 16713 }, { "epoch": 0.95, "learning_rate": 1.4096028381031745e-06, "loss": 0.6585, "step": 16714 }, { "epoch": 0.95, "learning_rate": 1.406530661895933e-06, "loss": 0.7886, "step": 16715 }, { "epoch": 0.95, "learning_rate": 1.4034618134768474e-06, "loss": 0.8393, "step": 16716 }, { "epoch": 0.95, "learning_rate": 1.400396292949513e-06, "loss": 0.7378, "step": 16717 }, { "epoch": 0.95, "learning_rate": 1.3973341004173801e-06, "loss": 0.5738, "step": 16718 }, { "epoch": 0.95, "learning_rate": 1.3942752359838218e-06, "loss": 0.6985, "step": 16719 }, { "epoch": 0.95, "learning_rate": 1.3912196997520777e-06, "loss": 0.7402, "step": 16720 }, { "epoch": 0.95, "learning_rate": 1.3881674918252763e-06, "loss": 0.682, "step": 16721 }, { "epoch": 0.95, "learning_rate": 1.3851186123064464e-06, "loss": 0.6589, "step": 16722 }, { "epoch": 0.95, "learning_rate": 1.3820730612984833e-06, "loss": 0.6409, "step": 16723 }, { "epoch": 0.95, "learning_rate": 1.379030838904205e-06, "loss": 0.7063, "step": 16724 }, { "epoch": 0.95, "learning_rate": 1.3759919452262626e-06, "loss": 0.6966, "step": 16725 }, { "epoch": 0.95, "learning_rate": 1.3729563803672519e-06, "loss": 0.6135, "step": 16726 }, { "epoch": 0.95, "learning_rate": 1.369924144429613e-06, "loss": 0.5981, "step": 16727 }, { "epoch": 0.95, "learning_rate": 1.3668952375157086e-06, "loss": 0.6144, "step": 16728 }, { "epoch": 0.95, "learning_rate": 1.3638696597277679e-06, "loss": 0.668, "step": 16729 }, { "epoch": 0.95, "learning_rate": 1.3608474111679093e-06, "loss": 0.7215, "step": 16730 }, { "epoch": 0.95, "learning_rate": 1.357828491938129e-06, "loss": 0.6675, "step": 16731 }, { "epoch": 0.95, "learning_rate": 1.3548129021403455e-06, "loss": 0.7731, "step": 16732 }, { "epoch": 0.95, "learning_rate": 1.3518006418763219e-06, "loss": 0.6938, "step": 16733 }, { "epoch": 0.95, "learning_rate": 1.3487917112477544e-06, "loss": 0.6207, "step": 16734 }, { "epoch": 0.95, "learning_rate": 1.345786110356173e-06, "loss": 0.7596, "step": 16735 }, { "epoch": 0.95, "learning_rate": 1.3427838393030633e-06, "loss": 0.7816, "step": 16736 }, { "epoch": 0.95, "learning_rate": 1.3397848981897222e-06, "loss": 0.6998, "step": 16737 }, { "epoch": 0.95, "learning_rate": 1.3367892871173904e-06, "loss": 0.6731, "step": 16738 }, { "epoch": 0.95, "learning_rate": 1.3337970061871763e-06, "loss": 0.7257, "step": 16739 }, { "epoch": 0.95, "learning_rate": 1.3308080555000657e-06, "loss": 0.7238, "step": 16740 }, { "epoch": 0.95, "learning_rate": 1.3278224351569556e-06, "loss": 0.6556, "step": 16741 }, { "epoch": 0.95, "learning_rate": 1.3248401452586212e-06, "loss": 0.6976, "step": 16742 }, { "epoch": 0.95, "learning_rate": 1.321861185905715e-06, "loss": 0.756, "step": 16743 }, { "epoch": 0.95, "learning_rate": 1.3188855571987902e-06, "loss": 0.707, "step": 16744 }, { "epoch": 0.95, "learning_rate": 1.3159132592382772e-06, "loss": 0.6031, "step": 16745 }, { "epoch": 0.95, "learning_rate": 1.3129442921244962e-06, "loss": 0.719, "step": 16746 }, { "epoch": 0.95, "learning_rate": 1.3099786559576554e-06, "loss": 0.7025, "step": 16747 }, { "epoch": 0.95, "learning_rate": 1.3070163508378642e-06, "loss": 0.7058, "step": 16748 }, { "epoch": 0.95, "learning_rate": 1.3040573768651087e-06, "loss": 0.7625, "step": 16749 }, { "epoch": 0.95, "learning_rate": 1.301101734139254e-06, "loss": 0.766, "step": 16750 }, { "epoch": 0.95, "learning_rate": 1.2981494227600643e-06, "loss": 0.6906, "step": 16751 }, { "epoch": 0.95, "learning_rate": 1.2952004428271825e-06, "loss": 0.7424, "step": 16752 }, { "epoch": 0.95, "learning_rate": 1.292254794440151e-06, "loss": 0.6513, "step": 16753 }, { "epoch": 0.95, "learning_rate": 1.28931247769839e-06, "loss": 0.6657, "step": 16754 }, { "epoch": 0.95, "learning_rate": 1.2863734927012095e-06, "loss": 0.6302, "step": 16755 }, { "epoch": 0.95, "learning_rate": 1.2834378395478186e-06, "loss": 0.7742, "step": 16756 }, { "epoch": 0.95, "learning_rate": 1.2805055183372828e-06, "loss": 0.6487, "step": 16757 }, { "epoch": 0.95, "learning_rate": 1.2775765291685894e-06, "loss": 0.6966, "step": 16758 }, { "epoch": 0.95, "learning_rate": 1.2746508721406036e-06, "loss": 0.6973, "step": 16759 }, { "epoch": 0.95, "learning_rate": 1.271728547352058e-06, "loss": 0.66, "step": 16760 }, { "epoch": 0.95, "learning_rate": 1.2688095549016066e-06, "loss": 0.6138, "step": 16761 }, { "epoch": 0.95, "learning_rate": 1.2658938948877597e-06, "loss": 0.7103, "step": 16762 }, { "epoch": 0.95, "learning_rate": 1.2629815674089384e-06, "loss": 0.6635, "step": 16763 }, { "epoch": 0.95, "learning_rate": 1.2600725725634311e-06, "loss": 0.6686, "step": 16764 }, { "epoch": 0.95, "learning_rate": 1.2571669104494256e-06, "loss": 0.7194, "step": 16765 }, { "epoch": 0.95, "learning_rate": 1.25426458116501e-06, "loss": 0.6687, "step": 16766 }, { "epoch": 0.95, "learning_rate": 1.2513655848081284e-06, "loss": 0.6846, "step": 16767 }, { "epoch": 0.95, "learning_rate": 1.248469921476636e-06, "loss": 0.6681, "step": 16768 }, { "epoch": 0.95, "learning_rate": 1.2455775912682766e-06, "loss": 0.7891, "step": 16769 }, { "epoch": 0.95, "learning_rate": 1.2426885942806609e-06, "loss": 0.6864, "step": 16770 }, { "epoch": 0.95, "learning_rate": 1.2398029306113e-06, "loss": 0.7686, "step": 16771 }, { "epoch": 0.95, "learning_rate": 1.236920600357605e-06, "loss": 0.6672, "step": 16772 }, { "epoch": 0.95, "learning_rate": 1.2340416036168534e-06, "loss": 0.7256, "step": 16773 }, { "epoch": 0.95, "learning_rate": 1.231165940486234e-06, "loss": 0.6483, "step": 16774 }, { "epoch": 0.95, "learning_rate": 1.2282936110627808e-06, "loss": 0.6716, "step": 16775 }, { "epoch": 0.95, "learning_rate": 1.22542461544346e-06, "loss": 0.7991, "step": 16776 }, { "epoch": 0.95, "learning_rate": 1.222558953725106e-06, "loss": 0.7141, "step": 16777 }, { "epoch": 0.95, "learning_rate": 1.2196966260044295e-06, "loss": 0.7386, "step": 16778 }, { "epoch": 0.95, "learning_rate": 1.216837632378065e-06, "loss": 0.7474, "step": 16779 }, { "epoch": 0.95, "learning_rate": 1.2139819729424906e-06, "loss": 0.7537, "step": 16780 }, { "epoch": 0.95, "learning_rate": 1.211129647794107e-06, "loss": 0.642, "step": 16781 }, { "epoch": 0.95, "learning_rate": 1.2082806570291815e-06, "loss": 0.7392, "step": 16782 }, { "epoch": 0.95, "learning_rate": 1.2054350007438707e-06, "loss": 0.6775, "step": 16783 }, { "epoch": 0.95, "learning_rate": 1.2025926790342313e-06, "loss": 0.7118, "step": 16784 }, { "epoch": 0.95, "learning_rate": 1.1997536919961861e-06, "loss": 0.7328, "step": 16785 }, { "epoch": 0.95, "learning_rate": 1.1969180397255697e-06, "loss": 0.6973, "step": 16786 }, { "epoch": 0.95, "learning_rate": 1.1940857223180946e-06, "loss": 0.6479, "step": 16787 }, { "epoch": 0.95, "learning_rate": 1.1912567398693508e-06, "loss": 0.7231, "step": 16788 }, { "epoch": 0.95, "learning_rate": 1.18843109247484e-06, "loss": 0.7306, "step": 16789 }, { "epoch": 0.95, "learning_rate": 1.1856087802299076e-06, "loss": 0.7699, "step": 16790 }, { "epoch": 0.95, "learning_rate": 1.1827898032298334e-06, "loss": 0.7015, "step": 16791 }, { "epoch": 0.95, "learning_rate": 1.1799741615697524e-06, "loss": 0.6897, "step": 16792 }, { "epoch": 0.95, "learning_rate": 1.1771618553447216e-06, "loss": 0.7674, "step": 16793 }, { "epoch": 0.95, "learning_rate": 1.1743528846496433e-06, "loss": 0.7084, "step": 16794 }, { "epoch": 0.95, "learning_rate": 1.1715472495793412e-06, "loss": 0.6623, "step": 16795 }, { "epoch": 0.95, "learning_rate": 1.1687449502285063e-06, "loss": 0.6281, "step": 16796 }, { "epoch": 0.95, "learning_rate": 1.1659459866917187e-06, "loss": 0.7212, "step": 16797 }, { "epoch": 0.95, "learning_rate": 1.1631503590634474e-06, "loss": 0.6363, "step": 16798 }, { "epoch": 0.95, "learning_rate": 1.160358067438072e-06, "loss": 0.6987, "step": 16799 }, { "epoch": 0.95, "learning_rate": 1.1575691119098286e-06, "loss": 0.7422, "step": 16800 }, { "epoch": 0.95, "learning_rate": 1.1547834925728528e-06, "loss": 0.6679, "step": 16801 }, { "epoch": 0.95, "learning_rate": 1.1520012095211586e-06, "loss": 0.7183, "step": 16802 }, { "epoch": 0.95, "learning_rate": 1.1492222628486705e-06, "loss": 0.7646, "step": 16803 }, { "epoch": 0.95, "learning_rate": 1.146446652649169e-06, "loss": 0.7149, "step": 16804 }, { "epoch": 0.95, "learning_rate": 1.1436743790163461e-06, "loss": 0.6442, "step": 16805 }, { "epoch": 0.95, "learning_rate": 1.1409054420437715e-06, "loss": 0.7626, "step": 16806 }, { "epoch": 0.95, "learning_rate": 1.1381398418249145e-06, "loss": 0.6988, "step": 16807 }, { "epoch": 0.95, "learning_rate": 1.1353775784531117e-06, "loss": 0.7118, "step": 16808 }, { "epoch": 0.95, "learning_rate": 1.1326186520215886e-06, "loss": 0.697, "step": 16809 }, { "epoch": 0.95, "learning_rate": 1.129863062623482e-06, "loss": 0.6691, "step": 16810 }, { "epoch": 0.95, "learning_rate": 1.1271108103517946e-06, "loss": 0.7355, "step": 16811 }, { "epoch": 0.95, "learning_rate": 1.1243618952994195e-06, "loss": 0.7312, "step": 16812 }, { "epoch": 0.95, "learning_rate": 1.1216163175591377e-06, "loss": 0.7062, "step": 16813 }, { "epoch": 0.95, "learning_rate": 1.1188740772236305e-06, "loss": 0.7336, "step": 16814 }, { "epoch": 0.95, "learning_rate": 1.1161351743854353e-06, "loss": 0.6328, "step": 16815 }, { "epoch": 0.95, "learning_rate": 1.1133996091370224e-06, "loss": 0.6576, "step": 16816 }, { "epoch": 0.95, "learning_rate": 1.110667381570707e-06, "loss": 0.6397, "step": 16817 }, { "epoch": 0.95, "learning_rate": 1.107938491778715e-06, "loss": 0.7157, "step": 16818 }, { "epoch": 0.95, "learning_rate": 1.1052129398531507e-06, "loss": 0.6973, "step": 16819 }, { "epoch": 0.95, "learning_rate": 1.1024907258860184e-06, "loss": 0.659, "step": 16820 }, { "epoch": 0.95, "learning_rate": 1.099771849969189e-06, "loss": 0.6104, "step": 16821 }, { "epoch": 0.95, "learning_rate": 1.0970563121944332e-06, "loss": 0.6248, "step": 16822 }, { "epoch": 0.95, "learning_rate": 1.0943441126534116e-06, "loss": 0.6506, "step": 16823 }, { "epoch": 0.95, "learning_rate": 1.0916352514376616e-06, "loss": 0.6999, "step": 16824 }, { "epoch": 0.95, "learning_rate": 1.0889297286386102e-06, "loss": 0.6841, "step": 16825 }, { "epoch": 0.95, "learning_rate": 1.0862275443476067e-06, "loss": 0.7447, "step": 16826 }, { "epoch": 0.95, "learning_rate": 1.0835286986558225e-06, "loss": 0.7168, "step": 16827 }, { "epoch": 0.95, "learning_rate": 1.0808331916543623e-06, "loss": 0.6859, "step": 16828 }, { "epoch": 0.95, "learning_rate": 1.0781410234342094e-06, "loss": 0.6806, "step": 16829 }, { "epoch": 0.95, "learning_rate": 1.0754521940862239e-06, "loss": 0.669, "step": 16830 }, { "epoch": 0.95, "learning_rate": 1.0727667037011668e-06, "loss": 0.7021, "step": 16831 }, { "epoch": 0.95, "learning_rate": 1.0700845523696767e-06, "loss": 0.7197, "step": 16832 }, { "epoch": 0.95, "learning_rate": 1.0674057401823035e-06, "loss": 0.6842, "step": 16833 }, { "epoch": 0.95, "learning_rate": 1.0647302672294303e-06, "loss": 0.6759, "step": 16834 }, { "epoch": 0.95, "learning_rate": 1.0620581336013845e-06, "loss": 0.679, "step": 16835 }, { "epoch": 0.96, "learning_rate": 1.0593893393883504e-06, "loss": 0.6815, "step": 16836 }, { "epoch": 0.96, "learning_rate": 1.0567238846803996e-06, "loss": 0.7193, "step": 16837 }, { "epoch": 0.96, "learning_rate": 1.0540617695675158e-06, "loss": 0.73, "step": 16838 }, { "epoch": 0.96, "learning_rate": 1.0514029941395388e-06, "loss": 0.6486, "step": 16839 }, { "epoch": 0.96, "learning_rate": 1.0487475584862182e-06, "loss": 0.7902, "step": 16840 }, { "epoch": 0.96, "learning_rate": 1.046095462697172e-06, "loss": 0.7866, "step": 16841 }, { "epoch": 0.96, "learning_rate": 1.0434467068619169e-06, "loss": 0.7596, "step": 16842 }, { "epoch": 0.96, "learning_rate": 1.0408012910698594e-06, "loss": 0.6929, "step": 16843 }, { "epoch": 0.96, "learning_rate": 1.0381592154102837e-06, "loss": 0.6439, "step": 16844 }, { "epoch": 0.96, "learning_rate": 1.0355204799723739e-06, "loss": 0.7639, "step": 16845 }, { "epoch": 0.96, "learning_rate": 1.0328850848452031e-06, "loss": 0.7203, "step": 16846 }, { "epoch": 0.96, "learning_rate": 1.0302530301177005e-06, "loss": 0.6498, "step": 16847 }, { "epoch": 0.96, "learning_rate": 1.027624315878728e-06, "loss": 0.6216, "step": 16848 }, { "epoch": 0.96, "learning_rate": 1.0249989422169926e-06, "loss": 0.6176, "step": 16849 }, { "epoch": 0.96, "learning_rate": 1.0223769092211012e-06, "loss": 0.7145, "step": 16850 }, { "epoch": 0.96, "learning_rate": 1.0197582169795828e-06, "loss": 0.6708, "step": 16851 }, { "epoch": 0.96, "learning_rate": 1.0171428655808113e-06, "loss": 0.7054, "step": 16852 }, { "epoch": 0.96, "learning_rate": 1.0145308551130495e-06, "loss": 0.694, "step": 16853 }, { "epoch": 0.96, "learning_rate": 1.0119221856644712e-06, "loss": 0.6407, "step": 16854 }, { "epoch": 0.96, "learning_rate": 1.0093168573231393e-06, "loss": 0.6877, "step": 16855 }, { "epoch": 0.96, "learning_rate": 1.0067148701769613e-06, "loss": 0.7064, "step": 16856 }, { "epoch": 0.96, "learning_rate": 1.004116224313778e-06, "loss": 0.7326, "step": 16857 }, { "epoch": 0.96, "learning_rate": 1.0015209198212972e-06, "loss": 0.7031, "step": 16858 }, { "epoch": 0.96, "learning_rate": 9.989289567871262e-07, "loss": 0.6768, "step": 16859 }, { "epoch": 0.96, "learning_rate": 9.9634033529874e-07, "loss": 0.7128, "step": 16860 }, { "epoch": 0.96, "learning_rate": 9.937550554435126e-07, "loss": 0.687, "step": 16861 }, { "epoch": 0.96, "learning_rate": 9.911731173086968e-07, "loss": 0.796, "step": 16862 }, { "epoch": 0.96, "learning_rate": 9.885945209814674e-07, "loss": 0.7419, "step": 16863 }, { "epoch": 0.96, "learning_rate": 9.86019266548821e-07, "loss": 0.7158, "step": 16864 }, { "epoch": 0.96, "learning_rate": 9.834473540977108e-07, "loss": 0.6197, "step": 16865 }, { "epoch": 0.96, "learning_rate": 9.808787837149335e-07, "loss": 0.7132, "step": 16866 }, { "epoch": 0.96, "learning_rate": 9.783135554871758e-07, "loss": 0.71, "step": 16867 }, { "epoch": 0.96, "learning_rate": 9.75751669501035e-07, "loss": 0.7126, "step": 16868 }, { "epoch": 0.96, "learning_rate": 9.731931258429638e-07, "loss": 0.618, "step": 16869 }, { "epoch": 0.96, "learning_rate": 9.70637924599349e-07, "loss": 0.7056, "step": 16870 }, { "epoch": 0.96, "learning_rate": 9.680860658564105e-07, "loss": 0.6686, "step": 16871 }, { "epoch": 0.96, "learning_rate": 9.655375497002905e-07, "loss": 0.6235, "step": 16872 }, { "epoch": 0.96, "learning_rate": 9.62992376217009e-07, "loss": 0.7167, "step": 16873 }, { "epoch": 0.96, "learning_rate": 9.60450545492453e-07, "loss": 0.7211, "step": 16874 }, { "epoch": 0.96, "learning_rate": 9.579120576124423e-07, "loss": 0.701, "step": 16875 }, { "epoch": 0.96, "learning_rate": 9.553769126626532e-07, "loss": 0.698, "step": 16876 }, { "epoch": 0.96, "learning_rate": 9.528451107286396e-07, "loss": 0.6449, "step": 16877 }, { "epoch": 0.96, "learning_rate": 9.503166518958662e-07, "loss": 0.739, "step": 16878 }, { "epoch": 0.96, "learning_rate": 9.477915362496758e-07, "loss": 0.6674, "step": 16879 }, { "epoch": 0.96, "learning_rate": 9.452697638752894e-07, "loss": 0.7211, "step": 16880 }, { "epoch": 0.96, "learning_rate": 9.427513348578276e-07, "loss": 0.6897, "step": 16881 }, { "epoch": 0.96, "learning_rate": 9.402362492823002e-07, "loss": 0.7297, "step": 16882 }, { "epoch": 0.96, "learning_rate": 9.37724507233595e-07, "loss": 0.6957, "step": 16883 }, { "epoch": 0.96, "learning_rate": 9.352161087964883e-07, "loss": 0.7123, "step": 16884 }, { "epoch": 0.96, "learning_rate": 9.327110540556461e-07, "loss": 0.6497, "step": 16885 }, { "epoch": 0.96, "learning_rate": 9.302093430956116e-07, "loss": 0.7031, "step": 16886 }, { "epoch": 0.96, "learning_rate": 9.277109760008395e-07, "loss": 0.737, "step": 16887 }, { "epoch": 0.96, "learning_rate": 9.252159528556403e-07, "loss": 0.732, "step": 16888 }, { "epoch": 0.96, "learning_rate": 9.227242737442355e-07, "loss": 0.7158, "step": 16889 }, { "epoch": 0.96, "learning_rate": 9.202359387507353e-07, "loss": 0.5877, "step": 16890 }, { "epoch": 0.96, "learning_rate": 9.177509479591173e-07, "loss": 0.7134, "step": 16891 }, { "epoch": 0.96, "learning_rate": 9.152693014532587e-07, "loss": 0.6424, "step": 16892 }, { "epoch": 0.96, "learning_rate": 9.127909993169148e-07, "loss": 0.6969, "step": 16893 }, { "epoch": 0.96, "learning_rate": 9.103160416337409e-07, "loss": 0.7794, "step": 16894 }, { "epoch": 0.96, "learning_rate": 9.0784442848727e-07, "loss": 0.6358, "step": 16895 }, { "epoch": 0.96, "learning_rate": 9.053761599609245e-07, "loss": 0.7085, "step": 16896 }, { "epoch": 0.96, "learning_rate": 9.029112361380376e-07, "loss": 0.6354, "step": 16897 }, { "epoch": 0.96, "learning_rate": 9.004496571017651e-07, "loss": 0.6681, "step": 16898 }, { "epoch": 0.96, "learning_rate": 8.979914229352293e-07, "loss": 0.7158, "step": 16899 }, { "epoch": 0.96, "learning_rate": 8.955365337213861e-07, "loss": 0.7012, "step": 16900 }, { "epoch": 0.96, "learning_rate": 8.930849895430915e-07, "loss": 0.7771, "step": 16901 }, { "epoch": 0.96, "learning_rate": 8.906367904831015e-07, "loss": 0.7174, "step": 16902 }, { "epoch": 0.96, "learning_rate": 8.881919366240388e-07, "loss": 0.6224, "step": 16903 }, { "epoch": 0.96, "learning_rate": 8.857504280484375e-07, "loss": 0.7195, "step": 16904 }, { "epoch": 0.96, "learning_rate": 8.833122648386871e-07, "loss": 0.7482, "step": 16905 }, { "epoch": 0.96, "learning_rate": 8.808774470770886e-07, "loss": 0.7332, "step": 16906 }, { "epoch": 0.96, "learning_rate": 8.784459748458318e-07, "loss": 0.7671, "step": 16907 }, { "epoch": 0.96, "learning_rate": 8.760178482269732e-07, "loss": 0.6474, "step": 16908 }, { "epoch": 0.96, "learning_rate": 8.735930673024806e-07, "loss": 0.6499, "step": 16909 }, { "epoch": 0.96, "learning_rate": 8.711716321541885e-07, "loss": 0.7071, "step": 16910 }, { "epoch": 0.96, "learning_rate": 8.687535428638316e-07, "loss": 0.6956, "step": 16911 }, { "epoch": 0.96, "learning_rate": 8.663387995130112e-07, "loss": 0.6536, "step": 16912 }, { "epoch": 0.96, "learning_rate": 8.639274021832511e-07, "loss": 0.6897, "step": 16913 }, { "epoch": 0.96, "learning_rate": 8.615193509559416e-07, "loss": 0.6868, "step": 16914 }, { "epoch": 0.96, "learning_rate": 8.591146459123511e-07, "loss": 0.6599, "step": 16915 }, { "epoch": 0.96, "learning_rate": 8.56713287133648e-07, "loss": 0.5924, "step": 16916 }, { "epoch": 0.96, "learning_rate": 8.543152747008897e-07, "loss": 0.7667, "step": 16917 }, { "epoch": 0.96, "learning_rate": 8.519206086950004e-07, "loss": 0.6761, "step": 16918 }, { "epoch": 0.96, "learning_rate": 8.495292891968154e-07, "loss": 0.7801, "step": 16919 }, { "epoch": 0.96, "learning_rate": 8.471413162870479e-07, "loss": 0.737, "step": 16920 }, { "epoch": 0.96, "learning_rate": 8.447566900462999e-07, "loss": 0.7025, "step": 16921 }, { "epoch": 0.96, "learning_rate": 8.423754105550629e-07, "loss": 0.7003, "step": 16922 }, { "epoch": 0.96, "learning_rate": 8.399974778937059e-07, "loss": 0.8047, "step": 16923 }, { "epoch": 0.96, "learning_rate": 8.376228921424867e-07, "loss": 0.6867, "step": 16924 }, { "epoch": 0.96, "learning_rate": 8.352516533815635e-07, "loss": 0.7699, "step": 16925 }, { "epoch": 0.96, "learning_rate": 8.328837616909613e-07, "loss": 0.688, "step": 16926 }, { "epoch": 0.96, "learning_rate": 8.305192171506049e-07, "loss": 0.7779, "step": 16927 }, { "epoch": 0.96, "learning_rate": 8.281580198403083e-07, "loss": 0.7353, "step": 16928 }, { "epoch": 0.96, "learning_rate": 8.258001698397744e-07, "loss": 0.7504, "step": 16929 }, { "epoch": 0.96, "learning_rate": 8.234456672285729e-07, "loss": 0.7091, "step": 16930 }, { "epoch": 0.96, "learning_rate": 8.210945120861735e-07, "loss": 0.7304, "step": 16931 }, { "epoch": 0.96, "learning_rate": 8.187467044919462e-07, "loss": 0.8258, "step": 16932 }, { "epoch": 0.96, "learning_rate": 8.164022445251274e-07, "loss": 0.7117, "step": 16933 }, { "epoch": 0.96, "learning_rate": 8.14061132264854e-07, "loss": 0.728, "step": 16934 }, { "epoch": 0.96, "learning_rate": 8.117233677901403e-07, "loss": 0.7275, "step": 16935 }, { "epoch": 0.96, "learning_rate": 8.093889511799013e-07, "loss": 0.6371, "step": 16936 }, { "epoch": 0.96, "learning_rate": 8.070578825129182e-07, "loss": 0.7029, "step": 16937 }, { "epoch": 0.96, "learning_rate": 8.047301618678726e-07, "loss": 0.6525, "step": 16938 }, { "epoch": 0.96, "learning_rate": 8.02405789323335e-07, "loss": 0.715, "step": 16939 }, { "epoch": 0.96, "learning_rate": 8.000847649577537e-07, "loss": 0.668, "step": 16940 }, { "epoch": 0.96, "learning_rate": 7.977670888494659e-07, "loss": 0.6368, "step": 16941 }, { "epoch": 0.96, "learning_rate": 7.954527610767204e-07, "loss": 0.7344, "step": 16942 }, { "epoch": 0.96, "learning_rate": 7.931417817176101e-07, "loss": 0.6673, "step": 16943 }, { "epoch": 0.96, "learning_rate": 7.908341508501505e-07, "loss": 0.6694, "step": 16944 }, { "epoch": 0.96, "learning_rate": 7.885298685522235e-07, "loss": 0.7493, "step": 16945 }, { "epoch": 0.96, "learning_rate": 7.862289349016005e-07, "loss": 0.7466, "step": 16946 }, { "epoch": 0.96, "learning_rate": 7.839313499759415e-07, "loss": 0.6334, "step": 16947 }, { "epoch": 0.96, "learning_rate": 7.816371138528067e-07, "loss": 0.6846, "step": 16948 }, { "epoch": 0.96, "learning_rate": 7.793462266096452e-07, "loss": 0.703, "step": 16949 }, { "epoch": 0.96, "learning_rate": 7.770586883237507e-07, "loss": 0.832, "step": 16950 }, { "epoch": 0.96, "learning_rate": 7.747744990723394e-07, "loss": 0.7255, "step": 16951 }, { "epoch": 0.96, "learning_rate": 7.724936589325383e-07, "loss": 0.77, "step": 16952 }, { "epoch": 0.96, "learning_rate": 7.702161679812969e-07, "loss": 0.673, "step": 16953 }, { "epoch": 0.96, "learning_rate": 7.679420262954984e-07, "loss": 0.7742, "step": 16954 }, { "epoch": 0.96, "learning_rate": 7.656712339518923e-07, "loss": 0.6443, "step": 16955 }, { "epoch": 0.96, "learning_rate": 7.634037910271507e-07, "loss": 0.6782, "step": 16956 }, { "epoch": 0.96, "learning_rate": 7.611396975977902e-07, "loss": 0.6758, "step": 16957 }, { "epoch": 0.96, "learning_rate": 7.588789537402163e-07, "loss": 0.7017, "step": 16958 }, { "epoch": 0.96, "learning_rate": 7.566215595307568e-07, "loss": 0.7185, "step": 16959 }, { "epoch": 0.96, "learning_rate": 7.543675150455843e-07, "loss": 0.6625, "step": 16960 }, { "epoch": 0.96, "learning_rate": 7.521168203607931e-07, "loss": 0.7427, "step": 16961 }, { "epoch": 0.96, "learning_rate": 7.498694755523561e-07, "loss": 0.615, "step": 16962 }, { "epoch": 0.96, "learning_rate": 7.476254806961014e-07, "loss": 0.7907, "step": 16963 }, { "epoch": 0.96, "learning_rate": 7.453848358678017e-07, "loss": 0.6507, "step": 16964 }, { "epoch": 0.96, "learning_rate": 7.431475411430522e-07, "loss": 0.7265, "step": 16965 }, { "epoch": 0.96, "learning_rate": 7.409135965974035e-07, "loss": 0.6616, "step": 16966 }, { "epoch": 0.96, "learning_rate": 7.386830023062174e-07, "loss": 0.6783, "step": 16967 }, { "epoch": 0.96, "learning_rate": 7.364557583448117e-07, "loss": 0.6447, "step": 16968 }, { "epoch": 0.96, "learning_rate": 7.342318647883595e-07, "loss": 0.8385, "step": 16969 }, { "epoch": 0.96, "learning_rate": 7.320113217119007e-07, "loss": 0.7756, "step": 16970 }, { "epoch": 0.96, "learning_rate": 7.297941291904086e-07, "loss": 0.7336, "step": 16971 }, { "epoch": 0.96, "learning_rate": 7.275802872987014e-07, "loss": 0.6845, "step": 16972 }, { "epoch": 0.96, "learning_rate": 7.253697961115191e-07, "loss": 0.6902, "step": 16973 }, { "epoch": 0.96, "learning_rate": 7.231626557034688e-07, "loss": 0.6584, "step": 16974 }, { "epoch": 0.96, "learning_rate": 7.209588661490352e-07, "loss": 0.6978, "step": 16975 }, { "epoch": 0.96, "learning_rate": 7.187584275226144e-07, "loss": 0.6701, "step": 16976 }, { "epoch": 0.96, "learning_rate": 7.165613398984583e-07, "loss": 0.5957, "step": 16977 }, { "epoch": 0.96, "learning_rate": 7.143676033507518e-07, "loss": 0.6995, "step": 16978 }, { "epoch": 0.96, "learning_rate": 7.121772179535135e-07, "loss": 0.6631, "step": 16979 }, { "epoch": 0.96, "learning_rate": 7.099901837806844e-07, "loss": 0.5846, "step": 16980 }, { "epoch": 0.96, "learning_rate": 7.078065009060941e-07, "loss": 0.717, "step": 16981 }, { "epoch": 0.96, "learning_rate": 7.056261694034394e-07, "loss": 0.6555, "step": 16982 }, { "epoch": 0.96, "learning_rate": 7.034491893463058e-07, "loss": 0.7403, "step": 16983 }, { "epoch": 0.96, "learning_rate": 7.012755608081789e-07, "loss": 0.7017, "step": 16984 }, { "epoch": 0.96, "learning_rate": 6.991052838624113e-07, "loss": 0.7796, "step": 16985 }, { "epoch": 0.96, "learning_rate": 6.969383585822775e-07, "loss": 0.6858, "step": 16986 }, { "epoch": 0.96, "learning_rate": 6.947747850409081e-07, "loss": 0.7225, "step": 16987 }, { "epoch": 0.96, "learning_rate": 6.926145633113223e-07, "loss": 0.6762, "step": 16988 }, { "epoch": 0.96, "learning_rate": 6.904576934664397e-07, "loss": 0.6569, "step": 16989 }, { "epoch": 0.96, "learning_rate": 6.883041755790576e-07, "loss": 0.6524, "step": 16990 }, { "epoch": 0.96, "learning_rate": 6.861540097218621e-07, "loss": 0.6777, "step": 16991 }, { "epoch": 0.96, "learning_rate": 6.840071959674288e-07, "loss": 0.7078, "step": 16992 }, { "epoch": 0.96, "learning_rate": 6.818637343882217e-07, "loss": 0.7656, "step": 16993 }, { "epoch": 0.96, "learning_rate": 6.79723625056583e-07, "loss": 0.7623, "step": 16994 }, { "epoch": 0.96, "learning_rate": 6.775868680447439e-07, "loss": 0.6672, "step": 16995 }, { "epoch": 0.96, "learning_rate": 6.754534634248466e-07, "loss": 0.6134, "step": 16996 }, { "epoch": 0.96, "learning_rate": 6.733234112688669e-07, "loss": 0.7094, "step": 16997 }, { "epoch": 0.96, "learning_rate": 6.711967116487139e-07, "loss": 0.7204, "step": 16998 }, { "epoch": 0.96, "learning_rate": 6.690733646361857e-07, "loss": 0.6869, "step": 16999 }, { "epoch": 0.96, "learning_rate": 6.669533703029251e-07, "loss": 0.7339, "step": 17000 }, { "epoch": 0.96, "learning_rate": 6.648367287205081e-07, "loss": 0.7847, "step": 17001 }, { "epoch": 0.96, "learning_rate": 6.627234399603555e-07, "loss": 0.6641, "step": 17002 }, { "epoch": 0.96, "learning_rate": 6.60613504093821e-07, "loss": 0.8041, "step": 17003 }, { "epoch": 0.96, "learning_rate": 6.585069211921035e-07, "loss": 0.6112, "step": 17004 }, { "epoch": 0.96, "learning_rate": 6.564036913263017e-07, "loss": 0.6142, "step": 17005 }, { "epoch": 0.96, "learning_rate": 6.543038145674251e-07, "loss": 0.6756, "step": 17006 }, { "epoch": 0.96, "learning_rate": 6.522072909863286e-07, "loss": 0.6958, "step": 17007 }, { "epoch": 0.96, "learning_rate": 6.501141206537886e-07, "loss": 0.7346, "step": 17008 }, { "epoch": 0.96, "learning_rate": 6.480243036404599e-07, "loss": 0.6823, "step": 17009 }, { "epoch": 0.96, "learning_rate": 6.459378400168637e-07, "loss": 0.6926, "step": 17010 }, { "epoch": 0.96, "learning_rate": 6.438547298534326e-07, "loss": 0.7222, "step": 17011 }, { "epoch": 0.97, "learning_rate": 6.41774973220477e-07, "loss": 0.687, "step": 17012 }, { "epoch": 0.97, "learning_rate": 6.396985701881852e-07, "loss": 0.6919, "step": 17013 }, { "epoch": 0.97, "learning_rate": 6.376255208266568e-07, "loss": 0.7158, "step": 17014 }, { "epoch": 0.97, "learning_rate": 6.355558252058469e-07, "loss": 0.6174, "step": 17015 }, { "epoch": 0.97, "learning_rate": 6.334894833956218e-07, "loss": 0.6848, "step": 17016 }, { "epoch": 0.97, "learning_rate": 6.314264954657256e-07, "loss": 0.7046, "step": 17017 }, { "epoch": 0.97, "learning_rate": 6.293668614858029e-07, "loss": 0.7523, "step": 17018 }, { "epoch": 0.97, "learning_rate": 6.273105815253422e-07, "loss": 0.7064, "step": 17019 }, { "epoch": 0.97, "learning_rate": 6.25257655653766e-07, "loss": 0.7795, "step": 17020 }, { "epoch": 0.97, "learning_rate": 6.232080839403631e-07, "loss": 0.7156, "step": 17021 }, { "epoch": 0.97, "learning_rate": 6.211618664543118e-07, "loss": 0.7493, "step": 17022 }, { "epoch": 0.97, "learning_rate": 6.191190032646788e-07, "loss": 0.7386, "step": 17023 }, { "epoch": 0.97, "learning_rate": 6.170794944404201e-07, "loss": 0.7061, "step": 17024 }, { "epoch": 0.97, "learning_rate": 6.150433400503697e-07, "loss": 0.75, "step": 17025 }, { "epoch": 0.97, "learning_rate": 6.130105401632503e-07, "loss": 0.6856, "step": 17026 }, { "epoch": 0.97, "learning_rate": 6.109810948476846e-07, "loss": 0.6249, "step": 17027 }, { "epoch": 0.97, "learning_rate": 6.089550041721515e-07, "loss": 0.7129, "step": 17028 }, { "epoch": 0.97, "learning_rate": 6.069322682050516e-07, "loss": 0.7259, "step": 17029 }, { "epoch": 0.97, "learning_rate": 6.049128870146636e-07, "loss": 0.7233, "step": 17030 }, { "epoch": 0.97, "learning_rate": 6.028968606691332e-07, "loss": 0.6782, "step": 17031 }, { "epoch": 0.97, "learning_rate": 6.008841892365169e-07, "loss": 0.648, "step": 17032 }, { "epoch": 0.97, "learning_rate": 5.988748727847382e-07, "loss": 0.7086, "step": 17033 }, { "epoch": 0.97, "learning_rate": 5.968689113816206e-07, "loss": 0.7033, "step": 17034 }, { "epoch": 0.97, "learning_rate": 5.948663050948767e-07, "loss": 0.7618, "step": 17035 }, { "epoch": 0.97, "learning_rate": 5.928670539920855e-07, "loss": 0.7094, "step": 17036 }, { "epoch": 0.97, "learning_rate": 5.908711581407378e-07, "loss": 0.7477, "step": 17037 }, { "epoch": 0.97, "learning_rate": 5.888786176082018e-07, "loss": 0.6664, "step": 17038 }, { "epoch": 0.97, "learning_rate": 5.868894324617346e-07, "loss": 0.6545, "step": 17039 }, { "epoch": 0.97, "learning_rate": 5.849036027684606e-07, "loss": 0.7371, "step": 17040 }, { "epoch": 0.97, "learning_rate": 5.829211285954261e-07, "loss": 0.6349, "step": 17041 }, { "epoch": 0.97, "learning_rate": 5.80942010009522e-07, "loss": 0.7955, "step": 17042 }, { "epoch": 0.97, "learning_rate": 5.789662470775614e-07, "loss": 0.6801, "step": 17043 }, { "epoch": 0.97, "learning_rate": 5.769938398662355e-07, "loss": 0.6402, "step": 17044 }, { "epoch": 0.97, "learning_rate": 5.750247884421134e-07, "loss": 0.7528, "step": 17045 }, { "epoch": 0.97, "learning_rate": 5.730590928716639e-07, "loss": 0.7402, "step": 17046 }, { "epoch": 0.97, "learning_rate": 5.71096753221212e-07, "loss": 0.6912, "step": 17047 }, { "epoch": 0.97, "learning_rate": 5.691377695570155e-07, "loss": 0.6467, "step": 17048 }, { "epoch": 0.97, "learning_rate": 5.671821419451884e-07, "loss": 0.635, "step": 17049 }, { "epoch": 0.97, "learning_rate": 5.652298704517334e-07, "loss": 0.7371, "step": 17050 }, { "epoch": 0.97, "learning_rate": 5.632809551425422e-07, "loss": 0.6722, "step": 17051 }, { "epoch": 0.97, "learning_rate": 5.613353960834067e-07, "loss": 0.719, "step": 17052 }, { "epoch": 0.97, "learning_rate": 5.593931933399854e-07, "loss": 0.7699, "step": 17053 }, { "epoch": 0.97, "learning_rate": 5.57454346977837e-07, "loss": 0.7028, "step": 17054 }, { "epoch": 0.97, "learning_rate": 5.555188570624092e-07, "loss": 0.7268, "step": 17055 }, { "epoch": 0.97, "learning_rate": 5.535867236590164e-07, "loss": 0.7688, "step": 17056 }, { "epoch": 0.97, "learning_rate": 5.516579468328842e-07, "loss": 0.6725, "step": 17057 }, { "epoch": 0.97, "learning_rate": 5.497325266491049e-07, "loss": 0.6598, "step": 17058 }, { "epoch": 0.97, "learning_rate": 5.478104631726711e-07, "loss": 0.7151, "step": 17059 }, { "epoch": 0.97, "learning_rate": 5.458917564684529e-07, "loss": 0.7742, "step": 17060 }, { "epoch": 0.97, "learning_rate": 5.43976406601221e-07, "loss": 0.6729, "step": 17061 }, { "epoch": 0.97, "learning_rate": 5.420644136356123e-07, "loss": 0.7074, "step": 17062 }, { "epoch": 0.97, "learning_rate": 5.401557776361754e-07, "loss": 0.7229, "step": 17063 }, { "epoch": 0.97, "learning_rate": 5.382504986673254e-07, "loss": 0.7032, "step": 17064 }, { "epoch": 0.97, "learning_rate": 5.363485767933663e-07, "loss": 0.6844, "step": 17065 }, { "epoch": 0.97, "learning_rate": 5.344500120784912e-07, "loss": 0.6635, "step": 17066 }, { "epoch": 0.97, "learning_rate": 5.325548045867935e-07, "loss": 0.7184, "step": 17067 }, { "epoch": 0.97, "learning_rate": 5.306629543822216e-07, "loss": 0.8403, "step": 17068 }, { "epoch": 0.97, "learning_rate": 5.287744615286583e-07, "loss": 0.6891, "step": 17069 }, { "epoch": 0.97, "learning_rate": 5.268893260898189e-07, "loss": 0.5927, "step": 17070 }, { "epoch": 0.97, "learning_rate": 5.250075481293526e-07, "loss": 0.6097, "step": 17071 }, { "epoch": 0.97, "learning_rate": 5.231291277107531e-07, "loss": 0.7, "step": 17072 }, { "epoch": 0.97, "learning_rate": 5.212540648974362e-07, "loss": 0.7057, "step": 17073 }, { "epoch": 0.97, "learning_rate": 5.193823597526959e-07, "loss": 0.7428, "step": 17074 }, { "epoch": 0.97, "learning_rate": 5.175140123396815e-07, "loss": 0.6382, "step": 17075 }, { "epoch": 0.97, "learning_rate": 5.156490227214871e-07, "loss": 0.7189, "step": 17076 }, { "epoch": 0.97, "learning_rate": 5.137873909610403e-07, "loss": 0.6708, "step": 17077 }, { "epoch": 0.97, "learning_rate": 5.119291171211793e-07, "loss": 0.7231, "step": 17078 }, { "epoch": 0.97, "learning_rate": 5.10074201264632e-07, "loss": 0.6541, "step": 17079 }, { "epoch": 0.97, "learning_rate": 5.082226434540039e-07, "loss": 0.7952, "step": 17080 }, { "epoch": 0.97, "learning_rate": 5.063744437517781e-07, "loss": 0.7177, "step": 17081 }, { "epoch": 0.97, "learning_rate": 5.045296022203494e-07, "loss": 0.6361, "step": 17082 }, { "epoch": 0.97, "learning_rate": 5.0268811892199e-07, "loss": 0.7735, "step": 17083 }, { "epoch": 0.97, "learning_rate": 5.008499939188505e-07, "loss": 0.5975, "step": 17084 }, { "epoch": 0.97, "learning_rate": 4.990152272729587e-07, "loss": 0.6826, "step": 17085 }, { "epoch": 0.97, "learning_rate": 4.971838190462652e-07, "loss": 0.7593, "step": 17086 }, { "epoch": 0.97, "learning_rate": 4.95355769300565e-07, "loss": 0.6915, "step": 17087 }, { "epoch": 0.97, "learning_rate": 4.935310780975644e-07, "loss": 0.6942, "step": 17088 }, { "epoch": 0.97, "learning_rate": 4.917097454988584e-07, "loss": 0.6962, "step": 17089 }, { "epoch": 0.97, "learning_rate": 4.898917715659201e-07, "loss": 0.6281, "step": 17090 }, { "epoch": 0.97, "learning_rate": 4.880771563601117e-07, "loss": 0.7456, "step": 17091 }, { "epoch": 0.97, "learning_rate": 4.862658999426839e-07, "loss": 0.8428, "step": 17092 }, { "epoch": 0.97, "learning_rate": 4.844580023747547e-07, "loss": 0.6321, "step": 17093 }, { "epoch": 0.97, "learning_rate": 4.826534637173641e-07, "loss": 0.7013, "step": 17094 }, { "epoch": 0.97, "learning_rate": 4.808522840314078e-07, "loss": 0.6952, "step": 17095 }, { "epoch": 0.97, "learning_rate": 4.790544633776817e-07, "loss": 0.6871, "step": 17096 }, { "epoch": 0.97, "learning_rate": 4.772600018168816e-07, "loss": 0.6679, "step": 17097 }, { "epoch": 0.97, "learning_rate": 4.754688994095591e-07, "loss": 0.6354, "step": 17098 }, { "epoch": 0.97, "learning_rate": 4.7368115621617694e-07, "loss": 0.6718, "step": 17099 }, { "epoch": 0.97, "learning_rate": 4.7189677229707576e-07, "loss": 0.7554, "step": 17100 }, { "epoch": 0.97, "learning_rate": 4.7011574771247404e-07, "loss": 0.6906, "step": 17101 }, { "epoch": 0.97, "learning_rate": 4.683380825224903e-07, "loss": 0.7246, "step": 17102 }, { "epoch": 0.97, "learning_rate": 4.6656377678713226e-07, "loss": 0.6048, "step": 17103 }, { "epoch": 0.97, "learning_rate": 4.647928305662852e-07, "loss": 0.66, "step": 17104 }, { "epoch": 0.97, "learning_rate": 4.630252439197125e-07, "loss": 0.732, "step": 17105 }, { "epoch": 0.97, "learning_rate": 4.6126101690708856e-07, "loss": 0.798, "step": 17106 }, { "epoch": 0.97, "learning_rate": 4.5950014958795475e-07, "loss": 0.6737, "step": 17107 }, { "epoch": 0.97, "learning_rate": 4.577426420217523e-07, "loss": 0.7502, "step": 17108 }, { "epoch": 0.97, "learning_rate": 4.5598849426777833e-07, "loss": 0.7229, "step": 17109 }, { "epoch": 0.97, "learning_rate": 4.5423770638527427e-07, "loss": 0.7167, "step": 17110 }, { "epoch": 0.97, "learning_rate": 4.5249027843330405e-07, "loss": 0.7147, "step": 17111 }, { "epoch": 0.97, "learning_rate": 4.5074621047086487e-07, "loss": 0.6875, "step": 17112 }, { "epoch": 0.97, "learning_rate": 4.4900550255682073e-07, "loss": 0.766, "step": 17113 }, { "epoch": 0.97, "learning_rate": 4.4726815474991356e-07, "loss": 0.5588, "step": 17114 }, { "epoch": 0.97, "learning_rate": 4.455341671087965e-07, "loss": 0.7632, "step": 17115 }, { "epoch": 0.97, "learning_rate": 4.438035396920004e-07, "loss": 0.6449, "step": 17116 }, { "epoch": 0.97, "learning_rate": 4.4207627255792305e-07, "loss": 0.6485, "step": 17117 }, { "epoch": 0.97, "learning_rate": 4.403523657648734e-07, "loss": 0.6854, "step": 17118 }, { "epoch": 0.97, "learning_rate": 4.3863181937103816e-07, "loss": 0.7117, "step": 17119 }, { "epoch": 0.97, "learning_rate": 4.3691463343448204e-07, "loss": 0.6307, "step": 17120 }, { "epoch": 0.97, "learning_rate": 4.3520080801318084e-07, "loss": 0.6514, "step": 17121 }, { "epoch": 0.97, "learning_rate": 4.3349034316496615e-07, "loss": 0.766, "step": 17122 }, { "epoch": 0.97, "learning_rate": 4.317832389475696e-07, "loss": 0.7113, "step": 17123 }, { "epoch": 0.97, "learning_rate": 4.300794954186227e-07, "loss": 0.7024, "step": 17124 }, { "epoch": 0.97, "learning_rate": 4.2837911263562404e-07, "loss": 0.7386, "step": 17125 }, { "epoch": 0.97, "learning_rate": 4.2668209065596097e-07, "loss": 0.6979, "step": 17126 }, { "epoch": 0.97, "learning_rate": 4.2498842953692107e-07, "loss": 0.771, "step": 17127 }, { "epoch": 0.97, "learning_rate": 4.232981293356697e-07, "loss": 0.6951, "step": 17128 }, { "epoch": 0.97, "learning_rate": 4.216111901092501e-07, "loss": 0.6105, "step": 17129 }, { "epoch": 0.97, "learning_rate": 4.199276119146167e-07, "loss": 0.7011, "step": 17130 }, { "epoch": 0.97, "learning_rate": 4.1824739480857966e-07, "loss": 0.6848, "step": 17131 }, { "epoch": 0.97, "learning_rate": 4.1657053884784914e-07, "loss": 0.7079, "step": 17132 }, { "epoch": 0.97, "learning_rate": 4.148970440890354e-07, "loss": 0.6779, "step": 17133 }, { "epoch": 0.97, "learning_rate": 4.132269105886155e-07, "loss": 0.697, "step": 17134 }, { "epoch": 0.97, "learning_rate": 4.115601384029666e-07, "loss": 0.691, "step": 17135 }, { "epoch": 0.97, "learning_rate": 4.0989672758834366e-07, "loss": 0.7574, "step": 17136 }, { "epoch": 0.97, "learning_rate": 4.082366782008906e-07, "loss": 0.7021, "step": 17137 }, { "epoch": 0.97, "learning_rate": 4.065799902966294e-07, "loss": 0.7017, "step": 17138 }, { "epoch": 0.97, "learning_rate": 4.04926663931493e-07, "loss": 0.6461, "step": 17139 }, { "epoch": 0.97, "learning_rate": 4.032766991612813e-07, "loss": 0.7764, "step": 17140 }, { "epoch": 0.97, "learning_rate": 4.01630096041683e-07, "loss": 0.6692, "step": 17141 }, { "epoch": 0.97, "learning_rate": 3.999868546282759e-07, "loss": 0.7656, "step": 17142 }, { "epoch": 0.97, "learning_rate": 3.983469749765267e-07, "loss": 0.7022, "step": 17143 }, { "epoch": 0.97, "learning_rate": 3.967104571417801e-07, "loss": 0.7221, "step": 17144 }, { "epoch": 0.97, "learning_rate": 3.9507730117926967e-07, "loss": 0.6346, "step": 17145 }, { "epoch": 0.97, "learning_rate": 3.9344750714412904e-07, "loss": 0.7325, "step": 17146 }, { "epoch": 0.97, "learning_rate": 3.918210750913587e-07, "loss": 0.7815, "step": 17147 }, { "epoch": 0.97, "learning_rate": 3.901980050758591e-07, "loss": 0.7385, "step": 17148 }, { "epoch": 0.97, "learning_rate": 3.885782971524088e-07, "loss": 0.6389, "step": 17149 }, { "epoch": 0.97, "learning_rate": 3.869619513756861e-07, "loss": 0.717, "step": 17150 }, { "epoch": 0.97, "learning_rate": 3.8534896780024756e-07, "loss": 0.739, "step": 17151 }, { "epoch": 0.97, "learning_rate": 3.837393464805272e-07, "loss": 0.807, "step": 17152 }, { "epoch": 0.97, "learning_rate": 3.8213308747084843e-07, "loss": 0.6338, "step": 17153 }, { "epoch": 0.97, "learning_rate": 3.805301908254455e-07, "loss": 0.7003, "step": 17154 }, { "epoch": 0.97, "learning_rate": 3.789306565984085e-07, "loss": 0.652, "step": 17155 }, { "epoch": 0.97, "learning_rate": 3.773344848437277e-07, "loss": 0.6171, "step": 17156 }, { "epoch": 0.97, "learning_rate": 3.757416756152821e-07, "loss": 0.7605, "step": 17157 }, { "epoch": 0.97, "learning_rate": 3.741522289668176e-07, "loss": 0.7393, "step": 17158 }, { "epoch": 0.97, "learning_rate": 3.7256614495200235e-07, "loss": 0.6724, "step": 17159 }, { "epoch": 0.97, "learning_rate": 3.7098342362436035e-07, "loss": 0.7502, "step": 17160 }, { "epoch": 0.97, "learning_rate": 3.694040650373154e-07, "loss": 0.6685, "step": 17161 }, { "epoch": 0.97, "learning_rate": 3.6782806924416933e-07, "loss": 0.7255, "step": 17162 }, { "epoch": 0.97, "learning_rate": 3.662554362981241e-07, "loss": 0.6745, "step": 17163 }, { "epoch": 0.97, "learning_rate": 3.6468616625225936e-07, "loss": 0.6577, "step": 17164 }, { "epoch": 0.97, "learning_rate": 3.6312025915953286e-07, "loss": 0.7955, "step": 17165 }, { "epoch": 0.97, "learning_rate": 3.615577150728133e-07, "loss": 0.6639, "step": 17166 }, { "epoch": 0.97, "learning_rate": 3.599985340448364e-07, "loss": 0.5862, "step": 17167 }, { "epoch": 0.97, "learning_rate": 3.5844271612821557e-07, "loss": 0.6863, "step": 17168 }, { "epoch": 0.97, "learning_rate": 3.568902613754754e-07, "loss": 0.7085, "step": 17169 }, { "epoch": 0.97, "learning_rate": 3.553411698390074e-07, "loss": 0.6896, "step": 17170 }, { "epoch": 0.97, "learning_rate": 3.537954415711031e-07, "loss": 0.663, "step": 17171 }, { "epoch": 0.97, "learning_rate": 3.522530766239318e-07, "loss": 0.7704, "step": 17172 }, { "epoch": 0.97, "learning_rate": 3.50714075049563e-07, "loss": 0.6377, "step": 17173 }, { "epoch": 0.97, "learning_rate": 3.491784368999329e-07, "loss": 0.7438, "step": 17174 }, { "epoch": 0.97, "learning_rate": 3.4764616222687785e-07, "loss": 0.7052, "step": 17175 }, { "epoch": 0.97, "learning_rate": 3.46117251082112e-07, "loss": 0.6902, "step": 17176 }, { "epoch": 0.97, "learning_rate": 3.4459170351722746e-07, "loss": 0.6738, "step": 17177 }, { "epoch": 0.97, "learning_rate": 3.4306951958373854e-07, "loss": 0.7371, "step": 17178 }, { "epoch": 0.97, "learning_rate": 3.415506993330153e-07, "loss": 0.6883, "step": 17179 }, { "epoch": 0.97, "learning_rate": 3.400352428163167e-07, "loss": 0.6338, "step": 17180 }, { "epoch": 0.97, "learning_rate": 3.3852315008480185e-07, "loss": 0.8606, "step": 17181 }, { "epoch": 0.97, "learning_rate": 3.3701442118949654e-07, "loss": 0.7092, "step": 17182 }, { "epoch": 0.97, "learning_rate": 3.3550905618132676e-07, "loss": 0.7089, "step": 17183 }, { "epoch": 0.97, "learning_rate": 3.340070551111185e-07, "loss": 0.7034, "step": 17184 }, { "epoch": 0.97, "learning_rate": 3.325084180295424e-07, "loss": 0.7799, "step": 17185 }, { "epoch": 0.97, "learning_rate": 3.3101314498720226e-07, "loss": 0.6968, "step": 17186 }, { "epoch": 0.97, "learning_rate": 3.2952123603455787e-07, "loss": 0.716, "step": 17187 }, { "epoch": 0.97, "learning_rate": 3.280326912219689e-07, "loss": 0.6586, "step": 17188 }, { "epoch": 0.98, "learning_rate": 3.265475105996729e-07, "loss": 0.6578, "step": 17189 }, { "epoch": 0.98, "learning_rate": 3.250656942178076e-07, "loss": 0.7657, "step": 17190 }, { "epoch": 0.98, "learning_rate": 3.235872421263775e-07, "loss": 0.5781, "step": 17191 }, { "epoch": 0.98, "learning_rate": 3.2211215437528694e-07, "loss": 0.7509, "step": 17192 }, { "epoch": 0.98, "learning_rate": 3.2064043101431853e-07, "loss": 0.8212, "step": 17193 }, { "epoch": 0.98, "learning_rate": 3.1917207209317676e-07, "loss": 0.7118, "step": 17194 }, { "epoch": 0.98, "learning_rate": 3.1770707766137777e-07, "loss": 0.7644, "step": 17195 }, { "epoch": 0.98, "learning_rate": 3.1624544776840403e-07, "loss": 0.7612, "step": 17196 }, { "epoch": 0.98, "learning_rate": 3.1478718246357173e-07, "loss": 0.7151, "step": 17197 }, { "epoch": 0.98, "learning_rate": 3.133322817961082e-07, "loss": 0.7448, "step": 17198 }, { "epoch": 0.98, "learning_rate": 3.118807458151185e-07, "loss": 0.7092, "step": 17199 }, { "epoch": 0.98, "learning_rate": 3.1043257456959686e-07, "loss": 0.7001, "step": 17200 }, { "epoch": 0.98, "learning_rate": 3.089877681084152e-07, "loss": 0.5967, "step": 17201 }, { "epoch": 0.98, "learning_rate": 3.0754632648035687e-07, "loss": 0.7246, "step": 17202 }, { "epoch": 0.98, "learning_rate": 3.061082497340495e-07, "loss": 0.7935, "step": 17203 }, { "epoch": 0.98, "learning_rate": 3.046735379180543e-07, "loss": 0.7084, "step": 17204 }, { "epoch": 0.98, "learning_rate": 3.032421910807881e-07, "loss": 0.7292, "step": 17205 }, { "epoch": 0.98, "learning_rate": 3.018142092705567e-07, "loss": 0.789, "step": 17206 }, { "epoch": 0.98, "learning_rate": 3.0038959253556597e-07, "loss": 0.7247, "step": 17207 }, { "epoch": 0.98, "learning_rate": 2.989683409238997e-07, "loss": 0.7246, "step": 17208 }, { "epoch": 0.98, "learning_rate": 2.9755045448351946e-07, "loss": 0.7917, "step": 17209 }, { "epoch": 0.98, "learning_rate": 2.9613593326229817e-07, "loss": 0.7096, "step": 17210 }, { "epoch": 0.98, "learning_rate": 2.947247773079753e-07, "loss": 0.6689, "step": 17211 }, { "epoch": 0.98, "learning_rate": 2.933169866681795e-07, "loss": 0.752, "step": 17212 }, { "epoch": 0.98, "learning_rate": 2.9191256139041723e-07, "loss": 0.6187, "step": 17213 }, { "epoch": 0.98, "learning_rate": 2.9051150152210604e-07, "loss": 0.6218, "step": 17214 }, { "epoch": 0.98, "learning_rate": 2.891138071105193e-07, "loss": 0.7022, "step": 17215 }, { "epoch": 0.98, "learning_rate": 2.877194782028525e-07, "loss": 0.7141, "step": 17216 }, { "epoch": 0.98, "learning_rate": 2.8632851484615696e-07, "loss": 0.6921, "step": 17217 }, { "epoch": 0.98, "learning_rate": 2.84940917087384e-07, "loss": 0.6818, "step": 17218 }, { "epoch": 0.98, "learning_rate": 2.8355668497336285e-07, "loss": 0.7185, "step": 17219 }, { "epoch": 0.98, "learning_rate": 2.8217581855082275e-07, "loss": 0.6769, "step": 17220 }, { "epoch": 0.98, "learning_rate": 2.807983178663709e-07, "loss": 0.838, "step": 17221 }, { "epoch": 0.98, "learning_rate": 2.794241829664923e-07, "loss": 0.679, "step": 17222 }, { "epoch": 0.98, "learning_rate": 2.780534138975721e-07, "loss": 0.6863, "step": 17223 }, { "epoch": 0.98, "learning_rate": 2.766860107058844e-07, "loss": 0.6639, "step": 17224 }, { "epoch": 0.98, "learning_rate": 2.7532197343758115e-07, "loss": 0.7154, "step": 17225 }, { "epoch": 0.98, "learning_rate": 2.739613021387033e-07, "loss": 0.6222, "step": 17226 }, { "epoch": 0.98, "learning_rate": 2.726039968551697e-07, "loss": 0.7659, "step": 17227 }, { "epoch": 0.98, "learning_rate": 2.712500576327881e-07, "loss": 0.7023, "step": 17228 }, { "epoch": 0.98, "learning_rate": 2.6989948451726643e-07, "loss": 0.7, "step": 17229 }, { "epoch": 0.98, "learning_rate": 2.685522775541904e-07, "loss": 0.712, "step": 17230 }, { "epoch": 0.98, "learning_rate": 2.6720843678903487e-07, "loss": 0.632, "step": 17231 }, { "epoch": 0.98, "learning_rate": 2.658679622671523e-07, "loss": 0.6557, "step": 17232 }, { "epoch": 0.98, "learning_rate": 2.645308540337843e-07, "loss": 0.7774, "step": 17233 }, { "epoch": 0.98, "learning_rate": 2.631971121340615e-07, "loss": 0.7639, "step": 17234 }, { "epoch": 0.98, "learning_rate": 2.6186673661301455e-07, "loss": 0.6659, "step": 17235 }, { "epoch": 0.98, "learning_rate": 2.605397275155297e-07, "loss": 0.722, "step": 17236 }, { "epoch": 0.98, "learning_rate": 2.592160848864156e-07, "loss": 0.7536, "step": 17237 }, { "epoch": 0.98, "learning_rate": 2.578958087703365e-07, "loss": 0.6975, "step": 17238 }, { "epoch": 0.98, "learning_rate": 2.565788992118567e-07, "loss": 0.69, "step": 17239 }, { "epoch": 0.98, "learning_rate": 2.552653562554186e-07, "loss": 0.6541, "step": 17240 }, { "epoch": 0.98, "learning_rate": 2.5395517994537545e-07, "loss": 0.5681, "step": 17241 }, { "epoch": 0.98, "learning_rate": 2.5264837032592525e-07, "loss": 0.7214, "step": 17242 }, { "epoch": 0.98, "learning_rate": 2.513449274411994e-07, "loss": 0.7233, "step": 17243 }, { "epoch": 0.98, "learning_rate": 2.5004485133517385e-07, "loss": 0.6885, "step": 17244 }, { "epoch": 0.98, "learning_rate": 2.487481420517468e-07, "loss": 0.7247, "step": 17245 }, { "epoch": 0.98, "learning_rate": 2.474547996346832e-07, "loss": 0.6172, "step": 17246 }, { "epoch": 0.98, "learning_rate": 2.461648241276149e-07, "loss": 0.7031, "step": 17247 }, { "epoch": 0.98, "learning_rate": 2.4487821557410696e-07, "loss": 0.8286, "step": 17248 }, { "epoch": 0.98, "learning_rate": 2.4359497401758024e-07, "loss": 0.8198, "step": 17249 }, { "epoch": 0.98, "learning_rate": 2.4231509950133347e-07, "loss": 0.7742, "step": 17250 }, { "epoch": 0.98, "learning_rate": 2.410385920685765e-07, "loss": 0.6531, "step": 17251 }, { "epoch": 0.98, "learning_rate": 2.3976545176240816e-07, "loss": 0.5319, "step": 17252 }, { "epoch": 0.98, "learning_rate": 2.384956786257719e-07, "loss": 0.6476, "step": 17253 }, { "epoch": 0.98, "learning_rate": 2.372292727015557e-07, "loss": 0.7267, "step": 17254 }, { "epoch": 0.98, "learning_rate": 2.3596623403248087e-07, "loss": 0.6941, "step": 17255 }, { "epoch": 0.98, "learning_rate": 2.3470656266117995e-07, "loss": 0.7617, "step": 17256 }, { "epoch": 0.98, "learning_rate": 2.3345025863018566e-07, "loss": 0.7093, "step": 17257 }, { "epoch": 0.98, "learning_rate": 2.3219732198188625e-07, "loss": 0.6092, "step": 17258 }, { "epoch": 0.98, "learning_rate": 2.3094775275858128e-07, "loss": 0.6935, "step": 17259 }, { "epoch": 0.98, "learning_rate": 2.2970155100244804e-07, "loss": 0.6215, "step": 17260 }, { "epoch": 0.98, "learning_rate": 2.2845871675553075e-07, "loss": 0.7236, "step": 17261 }, { "epoch": 0.98, "learning_rate": 2.2721925005980693e-07, "loss": 0.6876, "step": 17262 }, { "epoch": 0.98, "learning_rate": 2.2598315095708755e-07, "loss": 0.7177, "step": 17263 }, { "epoch": 0.98, "learning_rate": 2.2475041948910592e-07, "loss": 0.7018, "step": 17264 }, { "epoch": 0.98, "learning_rate": 2.2352105569747318e-07, "loss": 0.6361, "step": 17265 }, { "epoch": 0.98, "learning_rate": 2.2229505962367835e-07, "loss": 0.6442, "step": 17266 }, { "epoch": 0.98, "learning_rate": 2.2107243130909949e-07, "loss": 0.7383, "step": 17267 }, { "epoch": 0.98, "learning_rate": 2.1985317079500356e-07, "loss": 0.6891, "step": 17268 }, { "epoch": 0.98, "learning_rate": 2.1863727812254653e-07, "loss": 0.6883, "step": 17269 }, { "epoch": 0.98, "learning_rate": 2.1742475333276225e-07, "loss": 0.7121, "step": 17270 }, { "epoch": 0.98, "learning_rate": 2.1621559646659572e-07, "loss": 0.8042, "step": 17271 }, { "epoch": 0.98, "learning_rate": 2.1500980756483657e-07, "loss": 0.7648, "step": 17272 }, { "epoch": 0.98, "learning_rate": 2.1380738666818557e-07, "loss": 0.6886, "step": 17273 }, { "epoch": 0.98, "learning_rate": 2.1260833381724355e-07, "loss": 0.6504, "step": 17274 }, { "epoch": 0.98, "learning_rate": 2.1141264905246706e-07, "loss": 0.7749, "step": 17275 }, { "epoch": 0.98, "learning_rate": 2.1022033241422378e-07, "loss": 0.6917, "step": 17276 }, { "epoch": 0.98, "learning_rate": 2.0903138394275934e-07, "loss": 0.7072, "step": 17277 }, { "epoch": 0.98, "learning_rate": 2.0784580367818608e-07, "loss": 0.6748, "step": 17278 }, { "epoch": 0.98, "learning_rate": 2.066635916605386e-07, "loss": 0.6874, "step": 17279 }, { "epoch": 0.98, "learning_rate": 2.0548474792970728e-07, "loss": 0.6255, "step": 17280 }, { "epoch": 0.98, "learning_rate": 2.043092725254825e-07, "loss": 0.624, "step": 17281 }, { "epoch": 0.98, "learning_rate": 2.0313716548754358e-07, "loss": 0.6471, "step": 17282 }, { "epoch": 0.98, "learning_rate": 2.0196842685545892e-07, "loss": 0.7033, "step": 17283 }, { "epoch": 0.98, "learning_rate": 2.0080305666867472e-07, "loss": 0.7205, "step": 17284 }, { "epoch": 0.98, "learning_rate": 1.996410549665151e-07, "loss": 0.7501, "step": 17285 }, { "epoch": 0.98, "learning_rate": 1.984824217882042e-07, "loss": 0.6497, "step": 17286 }, { "epoch": 0.98, "learning_rate": 1.973271571728441e-07, "loss": 0.6745, "step": 17287 }, { "epoch": 0.98, "learning_rate": 1.9617526115943697e-07, "loss": 0.7362, "step": 17288 }, { "epoch": 0.98, "learning_rate": 1.9502673378686277e-07, "loss": 0.6792, "step": 17289 }, { "epoch": 0.98, "learning_rate": 1.938815750938905e-07, "loss": 0.5849, "step": 17290 }, { "epoch": 0.98, "learning_rate": 1.92739785119167e-07, "loss": 0.6785, "step": 17291 }, { "epoch": 0.98, "learning_rate": 1.9160136390121708e-07, "loss": 0.6435, "step": 17292 }, { "epoch": 0.98, "learning_rate": 1.904663114784877e-07, "loss": 0.7097, "step": 17293 }, { "epoch": 0.98, "learning_rate": 1.893346278892816e-07, "loss": 0.801, "step": 17294 }, { "epoch": 0.98, "learning_rate": 1.882063131717904e-07, "loss": 0.6552, "step": 17295 }, { "epoch": 0.98, "learning_rate": 1.8708136736409477e-07, "loss": 0.764, "step": 17296 }, { "epoch": 0.98, "learning_rate": 1.8595979050418655e-07, "loss": 0.7236, "step": 17297 }, { "epoch": 0.98, "learning_rate": 1.8484158262990214e-07, "loss": 0.6828, "step": 17298 }, { "epoch": 0.98, "learning_rate": 1.8372674377898913e-07, "loss": 0.6952, "step": 17299 }, { "epoch": 0.98, "learning_rate": 1.8261527398907297e-07, "loss": 0.7758, "step": 17300 }, { "epoch": 0.98, "learning_rate": 1.815071732976792e-07, "loss": 0.6964, "step": 17301 }, { "epoch": 0.98, "learning_rate": 1.80402441742189e-07, "loss": 0.7871, "step": 17302 }, { "epoch": 0.98, "learning_rate": 1.7930107935991703e-07, "loss": 0.6392, "step": 17303 }, { "epoch": 0.98, "learning_rate": 1.782030861880113e-07, "loss": 0.778, "step": 17304 }, { "epoch": 0.98, "learning_rate": 1.7710846226355328e-07, "loss": 0.634, "step": 17305 }, { "epoch": 0.98, "learning_rate": 1.7601720762346897e-07, "loss": 0.7393, "step": 17306 }, { "epoch": 0.98, "learning_rate": 1.749293223046178e-07, "loss": 0.746, "step": 17307 }, { "epoch": 0.98, "learning_rate": 1.738448063436926e-07, "loss": 0.6674, "step": 17308 }, { "epoch": 0.98, "learning_rate": 1.7276365977730856e-07, "loss": 0.5973, "step": 17309 }, { "epoch": 0.98, "learning_rate": 1.7168588264195872e-07, "loss": 0.6604, "step": 17310 }, { "epoch": 0.98, "learning_rate": 1.7061147497402507e-07, "loss": 0.7626, "step": 17311 }, { "epoch": 0.98, "learning_rate": 1.695404368097564e-07, "loss": 0.7209, "step": 17312 }, { "epoch": 0.98, "learning_rate": 1.6847276818531266e-07, "loss": 0.6989, "step": 17313 }, { "epoch": 0.98, "learning_rate": 1.674084691367428e-07, "loss": 0.7328, "step": 17314 }, { "epoch": 0.98, "learning_rate": 1.6634753969995142e-07, "loss": 0.7162, "step": 17315 }, { "epoch": 0.98, "learning_rate": 1.6528997991075433e-07, "loss": 0.7127, "step": 17316 }, { "epoch": 0.98, "learning_rate": 1.6423578980484522e-07, "loss": 0.7216, "step": 17317 }, { "epoch": 0.98, "learning_rate": 1.6318496941780669e-07, "loss": 0.6534, "step": 17318 }, { "epoch": 0.98, "learning_rate": 1.6213751878509932e-07, "loss": 0.7414, "step": 17319 }, { "epoch": 0.98, "learning_rate": 1.6109343794209476e-07, "loss": 0.6994, "step": 17320 }, { "epoch": 0.98, "learning_rate": 1.600527269240204e-07, "loss": 0.6865, "step": 17321 }, { "epoch": 0.98, "learning_rate": 1.5901538576601482e-07, "loss": 0.7288, "step": 17322 }, { "epoch": 0.98, "learning_rate": 1.5798141450307225e-07, "loss": 0.753, "step": 17323 }, { "epoch": 0.98, "learning_rate": 1.5695081317010917e-07, "loss": 0.7398, "step": 17324 }, { "epoch": 0.98, "learning_rate": 1.5592358180189782e-07, "loss": 0.6713, "step": 17325 }, { "epoch": 0.98, "learning_rate": 1.5489972043312152e-07, "loss": 0.5467, "step": 17326 }, { "epoch": 0.98, "learning_rate": 1.5387922909833042e-07, "loss": 0.7648, "step": 17327 }, { "epoch": 0.98, "learning_rate": 1.5286210783197474e-07, "loss": 0.833, "step": 17328 }, { "epoch": 0.98, "learning_rate": 1.518483566683826e-07, "loss": 0.7678, "step": 17329 }, { "epoch": 0.98, "learning_rate": 1.5083797564175994e-07, "loss": 0.774, "step": 17330 }, { "epoch": 0.98, "learning_rate": 1.4983096478622394e-07, "loss": 0.7036, "step": 17331 }, { "epoch": 0.98, "learning_rate": 1.488273241357696e-07, "loss": 0.6041, "step": 17332 }, { "epoch": 0.98, "learning_rate": 1.4782705372424766e-07, "loss": 0.6671, "step": 17333 }, { "epoch": 0.98, "learning_rate": 1.4683015358544217e-07, "loss": 0.741, "step": 17334 }, { "epoch": 0.98, "learning_rate": 1.458366237529929e-07, "loss": 0.6649, "step": 17335 }, { "epoch": 0.98, "learning_rate": 1.448464642604397e-07, "loss": 0.686, "step": 17336 }, { "epoch": 0.98, "learning_rate": 1.4385967514118915e-07, "loss": 0.7389, "step": 17337 }, { "epoch": 0.98, "learning_rate": 1.4287625642855906e-07, "loss": 0.7128, "step": 17338 }, { "epoch": 0.98, "learning_rate": 1.4189620815574511e-07, "loss": 0.6682, "step": 17339 }, { "epoch": 0.98, "learning_rate": 1.4091953035580973e-07, "loss": 0.7296, "step": 17340 }, { "epoch": 0.98, "learning_rate": 1.3994622306173765e-07, "loss": 0.7192, "step": 17341 }, { "epoch": 0.98, "learning_rate": 1.3897628630638038e-07, "loss": 0.8179, "step": 17342 }, { "epoch": 0.98, "learning_rate": 1.3800972012245616e-07, "loss": 0.6529, "step": 17343 }, { "epoch": 0.98, "learning_rate": 1.3704652454261668e-07, "loss": 0.6294, "step": 17344 }, { "epoch": 0.98, "learning_rate": 1.3608669959934705e-07, "loss": 0.7569, "step": 17345 }, { "epoch": 0.98, "learning_rate": 1.351302453250547e-07, "loss": 0.7206, "step": 17346 }, { "epoch": 0.98, "learning_rate": 1.341771617520249e-07, "loss": 0.6797, "step": 17347 }, { "epoch": 0.98, "learning_rate": 1.3322744891242078e-07, "loss": 0.6395, "step": 17348 }, { "epoch": 0.98, "learning_rate": 1.3228110683829454e-07, "loss": 0.7238, "step": 17349 }, { "epoch": 0.98, "learning_rate": 1.3133813556159835e-07, "loss": 0.6023, "step": 17350 }, { "epoch": 0.98, "learning_rate": 1.3039853511416234e-07, "loss": 0.6972, "step": 17351 }, { "epoch": 0.98, "learning_rate": 1.2946230552768333e-07, "loss": 0.6693, "step": 17352 }, { "epoch": 0.98, "learning_rate": 1.2852944683378054e-07, "loss": 0.6988, "step": 17353 }, { "epoch": 0.98, "learning_rate": 1.2759995906392874e-07, "loss": 0.7385, "step": 17354 }, { "epoch": 0.98, "learning_rate": 1.2667384224950284e-07, "loss": 0.717, "step": 17355 }, { "epoch": 0.98, "learning_rate": 1.2575109642175565e-07, "loss": 0.772, "step": 17356 }, { "epoch": 0.98, "learning_rate": 1.2483172161184e-07, "loss": 0.7015, "step": 17357 }, { "epoch": 0.98, "learning_rate": 1.239157178507866e-07, "loss": 0.6397, "step": 17358 }, { "epoch": 0.98, "learning_rate": 1.230030851695263e-07, "loss": 0.6074, "step": 17359 }, { "epoch": 0.98, "learning_rate": 1.2209382359883448e-07, "loss": 0.7018, "step": 17360 }, { "epoch": 0.98, "learning_rate": 1.2118793316941989e-07, "loss": 0.7216, "step": 17361 }, { "epoch": 0.98, "learning_rate": 1.202854139118581e-07, "loss": 0.7315, "step": 17362 }, { "epoch": 0.98, "learning_rate": 1.193862658566025e-07, "loss": 0.7782, "step": 17363 }, { "epoch": 0.98, "learning_rate": 1.1849048903400661e-07, "loss": 0.6369, "step": 17364 }, { "epoch": 0.99, "learning_rate": 1.1759808347430179e-07, "loss": 0.6368, "step": 17365 }, { "epoch": 0.99, "learning_rate": 1.167090492076195e-07, "loss": 0.6124, "step": 17366 }, { "epoch": 0.99, "learning_rate": 1.1582338626395794e-07, "loss": 0.7473, "step": 17367 }, { "epoch": 0.99, "learning_rate": 1.1494109467321545e-07, "loss": 0.6446, "step": 17368 }, { "epoch": 0.99, "learning_rate": 1.1406217446516821e-07, "loss": 0.7189, "step": 17369 }, { "epoch": 0.99, "learning_rate": 1.1318662566948135e-07, "loss": 0.7358, "step": 17370 }, { "epoch": 0.99, "learning_rate": 1.1231444831570902e-07, "loss": 0.7443, "step": 17371 }, { "epoch": 0.99, "learning_rate": 1.1144564243328326e-07, "loss": 0.7029, "step": 17372 }, { "epoch": 0.99, "learning_rate": 1.1058020805154722e-07, "loss": 0.7541, "step": 17373 }, { "epoch": 0.99, "learning_rate": 1.097181451996887e-07, "loss": 0.7413, "step": 17374 }, { "epoch": 0.99, "learning_rate": 1.0885945390681773e-07, "loss": 0.7415, "step": 17375 }, { "epoch": 0.99, "learning_rate": 1.0800413420192224e-07, "loss": 0.7361, "step": 17376 }, { "epoch": 0.99, "learning_rate": 1.0715218611384581e-07, "loss": 0.6755, "step": 17377 }, { "epoch": 0.99, "learning_rate": 1.0630360967137653e-07, "loss": 0.696, "step": 17378 }, { "epoch": 0.99, "learning_rate": 1.0545840490313596e-07, "loss": 0.6447, "step": 17379 }, { "epoch": 0.99, "learning_rate": 1.0461657183765683e-07, "loss": 0.6824, "step": 17380 }, { "epoch": 0.99, "learning_rate": 1.0377811050336084e-07, "loss": 0.7756, "step": 17381 }, { "epoch": 0.99, "learning_rate": 1.0294302092853647e-07, "loss": 0.7705, "step": 17382 }, { "epoch": 0.99, "learning_rate": 1.021113031413723e-07, "loss": 0.7003, "step": 17383 }, { "epoch": 0.99, "learning_rate": 1.0128295716993474e-07, "loss": 0.7274, "step": 17384 }, { "epoch": 0.99, "learning_rate": 1.0045798304220145e-07, "loss": 0.7084, "step": 17385 }, { "epoch": 0.99, "learning_rate": 9.963638078600567e-08, "loss": 0.5909, "step": 17386 }, { "epoch": 0.99, "learning_rate": 9.881815042909193e-08, "loss": 0.6618, "step": 17387 }, { "epoch": 0.99, "learning_rate": 9.800329199904923e-08, "loss": 0.7284, "step": 17388 }, { "epoch": 0.99, "learning_rate": 9.719180552341111e-08, "loss": 0.7679, "step": 17389 }, { "epoch": 0.99, "learning_rate": 9.638369102954459e-08, "loss": 0.6619, "step": 17390 }, { "epoch": 0.99, "learning_rate": 9.557894854473893e-08, "loss": 0.691, "step": 17391 }, { "epoch": 0.99, "learning_rate": 9.477757809615017e-08, "loss": 0.7244, "step": 17392 }, { "epoch": 0.99, "learning_rate": 9.397957971083449e-08, "loss": 0.6788, "step": 17393 }, { "epoch": 0.99, "learning_rate": 9.318495341571476e-08, "loss": 0.7245, "step": 17394 }, { "epoch": 0.99, "learning_rate": 9.239369923762508e-08, "loss": 0.679, "step": 17395 }, { "epoch": 0.99, "learning_rate": 9.160581720325523e-08, "loss": 0.7913, "step": 17396 }, { "epoch": 0.99, "learning_rate": 9.082130733921723e-08, "loss": 0.6882, "step": 17397 }, { "epoch": 0.99, "learning_rate": 9.004016967197881e-08, "loss": 0.7536, "step": 17398 }, { "epoch": 0.99, "learning_rate": 8.926240422790777e-08, "loss": 0.7158, "step": 17399 }, { "epoch": 0.99, "learning_rate": 8.848801103326087e-08, "loss": 0.6958, "step": 17400 }, { "epoch": 0.99, "learning_rate": 8.771699011416168e-08, "loss": 0.7423, "step": 17401 }, { "epoch": 0.99, "learning_rate": 8.694934149665601e-08, "loss": 0.657, "step": 17402 }, { "epoch": 0.99, "learning_rate": 8.618506520663428e-08, "loss": 0.6497, "step": 17403 }, { "epoch": 0.99, "learning_rate": 8.542416126989805e-08, "loss": 0.7431, "step": 17404 }, { "epoch": 0.99, "learning_rate": 8.46666297121379e-08, "loss": 0.682, "step": 17405 }, { "epoch": 0.99, "learning_rate": 8.391247055891116e-08, "loss": 0.6757, "step": 17406 }, { "epoch": 0.99, "learning_rate": 8.316168383567524e-08, "loss": 0.6977, "step": 17407 }, { "epoch": 0.99, "learning_rate": 8.241426956777653e-08, "loss": 0.7123, "step": 17408 }, { "epoch": 0.99, "learning_rate": 8.167022778045042e-08, "loss": 0.7329, "step": 17409 }, { "epoch": 0.99, "learning_rate": 8.092955849878792e-08, "loss": 0.699, "step": 17410 }, { "epoch": 0.99, "learning_rate": 8.019226174780236e-08, "loss": 0.739, "step": 17411 }, { "epoch": 0.99, "learning_rate": 7.945833755237386e-08, "loss": 0.6353, "step": 17412 }, { "epoch": 0.99, "learning_rate": 7.872778593728258e-08, "loss": 0.7157, "step": 17413 }, { "epoch": 0.99, "learning_rate": 7.800060692718657e-08, "loss": 0.7034, "step": 17414 }, { "epoch": 0.99, "learning_rate": 7.727680054662178e-08, "loss": 0.695, "step": 17415 }, { "epoch": 0.99, "learning_rate": 7.65563668200131e-08, "loss": 0.619, "step": 17416 }, { "epoch": 0.99, "learning_rate": 7.583930577169662e-08, "loss": 0.7697, "step": 17417 }, { "epoch": 0.99, "learning_rate": 7.51256174258641e-08, "loss": 0.6811, "step": 17418 }, { "epoch": 0.99, "learning_rate": 7.441530180659628e-08, "loss": 0.714, "step": 17419 }, { "epoch": 0.99, "learning_rate": 7.370835893788508e-08, "loss": 0.7193, "step": 17420 }, { "epoch": 0.99, "learning_rate": 7.300478884357808e-08, "loss": 0.6995, "step": 17421 }, { "epoch": 0.99, "learning_rate": 7.230459154742298e-08, "loss": 0.7133, "step": 17422 }, { "epoch": 0.99, "learning_rate": 7.160776707306749e-08, "loss": 0.7187, "step": 17423 }, { "epoch": 0.99, "learning_rate": 7.091431544401505e-08, "loss": 0.7203, "step": 17424 }, { "epoch": 0.99, "learning_rate": 7.022423668368028e-08, "loss": 0.8115, "step": 17425 }, { "epoch": 0.99, "learning_rate": 6.953753081535564e-08, "loss": 0.6078, "step": 17426 }, { "epoch": 0.99, "learning_rate": 6.885419786221148e-08, "loss": 0.7325, "step": 17427 }, { "epoch": 0.99, "learning_rate": 6.817423784731825e-08, "loss": 0.7696, "step": 17428 }, { "epoch": 0.99, "learning_rate": 6.749765079363534e-08, "loss": 0.64, "step": 17429 }, { "epoch": 0.99, "learning_rate": 6.682443672397786e-08, "loss": 0.7623, "step": 17430 }, { "epoch": 0.99, "learning_rate": 6.615459566108317e-08, "loss": 0.703, "step": 17431 }, { "epoch": 0.99, "learning_rate": 6.54881276275554e-08, "loss": 0.6961, "step": 17432 }, { "epoch": 0.99, "learning_rate": 6.482503264588769e-08, "loss": 0.7582, "step": 17433 }, { "epoch": 0.99, "learning_rate": 6.41653107384732e-08, "loss": 0.7906, "step": 17434 }, { "epoch": 0.99, "learning_rate": 6.350896192757194e-08, "loss": 0.6898, "step": 17435 }, { "epoch": 0.99, "learning_rate": 6.285598623533284e-08, "loss": 0.7713, "step": 17436 }, { "epoch": 0.99, "learning_rate": 6.220638368379384e-08, "loss": 0.7024, "step": 17437 }, { "epoch": 0.99, "learning_rate": 6.156015429489293e-08, "loss": 0.6099, "step": 17438 }, { "epoch": 0.99, "learning_rate": 6.09172980904238e-08, "loss": 0.6592, "step": 17439 }, { "epoch": 0.99, "learning_rate": 6.02778150921024e-08, "loss": 0.7073, "step": 17440 }, { "epoch": 0.99, "learning_rate": 5.964170532151147e-08, "loss": 0.6556, "step": 17441 }, { "epoch": 0.99, "learning_rate": 5.900896880011164e-08, "loss": 0.694, "step": 17442 }, { "epoch": 0.99, "learning_rate": 5.8379605549263584e-08, "loss": 0.6594, "step": 17443 }, { "epoch": 0.99, "learning_rate": 5.775361559021697e-08, "loss": 0.676, "step": 17444 }, { "epoch": 0.99, "learning_rate": 5.713099894408824e-08, "loss": 0.6385, "step": 17445 }, { "epoch": 0.99, "learning_rate": 5.6511755631893925e-08, "loss": 0.7042, "step": 17446 }, { "epoch": 0.99, "learning_rate": 5.589588567455062e-08, "loss": 0.706, "step": 17447 }, { "epoch": 0.99, "learning_rate": 5.528338909281949e-08, "loss": 0.6017, "step": 17448 }, { "epoch": 0.99, "learning_rate": 5.467426590739511e-08, "loss": 0.6968, "step": 17449 }, { "epoch": 0.99, "learning_rate": 5.406851613883879e-08, "loss": 0.685, "step": 17450 }, { "epoch": 0.99, "learning_rate": 5.346613980757864e-08, "loss": 0.7207, "step": 17451 }, { "epoch": 0.99, "learning_rate": 5.2867136933953956e-08, "loss": 0.6461, "step": 17452 }, { "epoch": 0.99, "learning_rate": 5.227150753819299e-08, "loss": 0.6999, "step": 17453 }, { "epoch": 0.99, "learning_rate": 5.167925164037968e-08, "loss": 0.7197, "step": 17454 }, { "epoch": 0.99, "learning_rate": 5.1090369260531345e-08, "loss": 0.7118, "step": 17455 }, { "epoch": 0.99, "learning_rate": 5.050486041849878e-08, "loss": 0.7589, "step": 17456 }, { "epoch": 0.99, "learning_rate": 4.9922725134055045e-08, "loss": 0.733, "step": 17457 }, { "epoch": 0.99, "learning_rate": 4.934396342684e-08, "loss": 0.6843, "step": 17458 }, { "epoch": 0.99, "learning_rate": 4.876857531640466e-08, "loss": 0.7863, "step": 17459 }, { "epoch": 0.99, "learning_rate": 4.8196560822166836e-08, "loss": 0.7487, "step": 17460 }, { "epoch": 0.99, "learning_rate": 4.7627919963422195e-08, "loss": 0.7117, "step": 17461 }, { "epoch": 0.99, "learning_rate": 4.70626527593665e-08, "loss": 0.6898, "step": 17462 }, { "epoch": 0.99, "learning_rate": 4.650075922909558e-08, "loss": 0.7304, "step": 17463 }, { "epoch": 0.99, "learning_rate": 4.594223939154985e-08, "loss": 0.698, "step": 17464 }, { "epoch": 0.99, "learning_rate": 4.5387093265591986e-08, "loss": 0.7395, "step": 17465 }, { "epoch": 0.99, "learning_rate": 4.483532086996256e-08, "loss": 0.7516, "step": 17466 }, { "epoch": 0.99, "learning_rate": 4.428692222329112e-08, "loss": 0.5928, "step": 17467 }, { "epoch": 0.99, "learning_rate": 4.374189734407397e-08, "loss": 0.7948, "step": 17468 }, { "epoch": 0.99, "learning_rate": 4.320024625070751e-08, "loss": 0.6562, "step": 17469 }, { "epoch": 0.99, "learning_rate": 4.266196896148822e-08, "loss": 0.7832, "step": 17470 }, { "epoch": 0.99, "learning_rate": 4.212706549456824e-08, "loss": 0.7067, "step": 17471 }, { "epoch": 0.99, "learning_rate": 4.159553586801091e-08, "loss": 0.7276, "step": 17472 }, { "epoch": 0.99, "learning_rate": 4.1067380099757415e-08, "loss": 0.6535, "step": 17473 }, { "epoch": 0.99, "learning_rate": 4.0542598207626846e-08, "loss": 0.7282, "step": 17474 }, { "epoch": 0.99, "learning_rate": 4.002119020933837e-08, "loss": 0.6953, "step": 17475 }, { "epoch": 0.99, "learning_rate": 3.950315612248901e-08, "loss": 0.7435, "step": 17476 }, { "epoch": 0.99, "learning_rate": 3.898849596456478e-08, "loss": 0.7081, "step": 17477 }, { "epoch": 0.99, "learning_rate": 3.847720975294067e-08, "loss": 0.68, "step": 17478 }, { "epoch": 0.99, "learning_rate": 3.796929750485845e-08, "loss": 0.6743, "step": 17479 }, { "epoch": 0.99, "learning_rate": 3.746475923748216e-08, "loss": 0.669, "step": 17480 }, { "epoch": 0.99, "learning_rate": 3.696359496783153e-08, "loss": 0.7892, "step": 17481 }, { "epoch": 0.99, "learning_rate": 3.6465804712815244e-08, "loss": 0.7241, "step": 17482 }, { "epoch": 0.99, "learning_rate": 3.5971388489242084e-08, "loss": 0.6972, "step": 17483 }, { "epoch": 0.99, "learning_rate": 3.548034631380981e-08, "loss": 0.7277, "step": 17484 }, { "epoch": 0.99, "learning_rate": 3.499267820307184e-08, "loss": 0.6034, "step": 17485 }, { "epoch": 0.99, "learning_rate": 3.4508384173503883e-08, "loss": 0.6152, "step": 17486 }, { "epoch": 0.99, "learning_rate": 3.402746424144843e-08, "loss": 0.6325, "step": 17487 }, { "epoch": 0.99, "learning_rate": 3.354991842313693e-08, "loss": 0.6502, "step": 17488 }, { "epoch": 0.99, "learning_rate": 3.307574673467873e-08, "loss": 0.7221, "step": 17489 }, { "epoch": 0.99, "learning_rate": 3.2604949192094334e-08, "loss": 0.7098, "step": 17490 }, { "epoch": 0.99, "learning_rate": 3.2137525811271055e-08, "loss": 0.729, "step": 17491 }, { "epoch": 0.99, "learning_rate": 3.167347660797404e-08, "loss": 0.6312, "step": 17492 }, { "epoch": 0.99, "learning_rate": 3.121280159787965e-08, "loss": 0.7815, "step": 17493 }, { "epoch": 0.99, "learning_rate": 3.0755500796531004e-08, "loss": 0.7226, "step": 17494 }, { "epoch": 0.99, "learning_rate": 3.03015742193602e-08, "loss": 0.6919, "step": 17495 }, { "epoch": 0.99, "learning_rate": 2.985102188168831e-08, "loss": 0.7015, "step": 17496 }, { "epoch": 0.99, "learning_rate": 2.9403843798725407e-08, "loss": 0.6607, "step": 17497 }, { "epoch": 0.99, "learning_rate": 2.8960039985570508e-08, "loss": 0.7649, "step": 17498 }, { "epoch": 0.99, "learning_rate": 2.8519610457189426e-08, "loss": 0.6905, "step": 17499 }, { "epoch": 0.99, "learning_rate": 2.8082555228470254e-08, "loss": 0.6998, "step": 17500 }, { "epoch": 0.99, "learning_rate": 2.7648874314145645e-08, "loss": 0.7373, "step": 17501 }, { "epoch": 0.99, "learning_rate": 2.7218567728848344e-08, "loss": 0.6799, "step": 17502 }, { "epoch": 0.99, "learning_rate": 2.6791635487122268e-08, "loss": 0.6546, "step": 17503 }, { "epoch": 0.99, "learning_rate": 2.6368077603367015e-08, "loss": 0.6437, "step": 17504 }, { "epoch": 0.99, "learning_rate": 2.594789409188225e-08, "loss": 0.7555, "step": 17505 }, { "epoch": 0.99, "learning_rate": 2.5531084966834428e-08, "loss": 0.6576, "step": 17506 }, { "epoch": 0.99, "learning_rate": 2.5117650242323376e-08, "loss": 0.6785, "step": 17507 }, { "epoch": 0.99, "learning_rate": 2.4707589932271292e-08, "loss": 0.665, "step": 17508 }, { "epoch": 0.99, "learning_rate": 2.430090405054486e-08, "loss": 0.6409, "step": 17509 }, { "epoch": 0.99, "learning_rate": 2.389759261084423e-08, "loss": 0.6928, "step": 17510 }, { "epoch": 0.99, "learning_rate": 2.349765562680295e-08, "loss": 0.6786, "step": 17511 }, { "epoch": 0.99, "learning_rate": 2.3101093111921325e-08, "loss": 0.6063, "step": 17512 }, { "epoch": 0.99, "learning_rate": 2.2707905079566437e-08, "loss": 0.7707, "step": 17513 }, { "epoch": 0.99, "learning_rate": 2.2318091543027665e-08, "loss": 0.6469, "step": 17514 }, { "epoch": 0.99, "learning_rate": 2.193165251545004e-08, "loss": 0.6531, "step": 17515 }, { "epoch": 0.99, "learning_rate": 2.1548588009878688e-08, "loss": 0.6602, "step": 17516 }, { "epoch": 0.99, "learning_rate": 2.1168898039258812e-08, "loss": 0.7551, "step": 17517 }, { "epoch": 0.99, "learning_rate": 2.0792582616380174e-08, "loss": 0.7181, "step": 17518 }, { "epoch": 0.99, "learning_rate": 2.041964175395483e-08, "loss": 0.6037, "step": 17519 }, { "epoch": 0.99, "learning_rate": 2.005007546458382e-08, "loss": 0.5671, "step": 17520 }, { "epoch": 0.99, "learning_rate": 1.9683883760723832e-08, "loss": 0.7315, "step": 17521 }, { "epoch": 0.99, "learning_rate": 1.932106665474276e-08, "loss": 0.6464, "step": 17522 }, { "epoch": 0.99, "learning_rate": 1.8961624158875257e-08, "loss": 0.727, "step": 17523 }, { "epoch": 0.99, "learning_rate": 1.860555628527827e-08, "loss": 0.6405, "step": 17524 }, { "epoch": 0.99, "learning_rate": 1.8252863045942204e-08, "loss": 0.6508, "step": 17525 }, { "epoch": 0.99, "learning_rate": 1.7903544452790855e-08, "loss": 0.7066, "step": 17526 }, { "epoch": 0.99, "learning_rate": 1.7557600517603688e-08, "loss": 0.7233, "step": 17527 }, { "epoch": 0.99, "learning_rate": 1.7215031252060255e-08, "loss": 0.7038, "step": 17528 }, { "epoch": 0.99, "learning_rate": 1.687583666772907e-08, "loss": 0.6531, "step": 17529 }, { "epoch": 0.99, "learning_rate": 1.6540016776045442e-08, "loss": 0.6353, "step": 17530 }, { "epoch": 0.99, "learning_rate": 1.6207571588355842e-08, "loss": 0.627, "step": 17531 }, { "epoch": 0.99, "learning_rate": 1.5878501115884625e-08, "loss": 0.7338, "step": 17532 }, { "epoch": 0.99, "learning_rate": 1.555280536972292e-08, "loss": 0.674, "step": 17533 }, { "epoch": 0.99, "learning_rate": 1.5230484360873044e-08, "loss": 0.673, "step": 17534 }, { "epoch": 0.99, "learning_rate": 1.491153810021517e-08, "loss": 0.7758, "step": 17535 }, { "epoch": 0.99, "learning_rate": 1.4595966598518473e-08, "loss": 0.6656, "step": 17536 }, { "epoch": 0.99, "learning_rate": 1.4283769866429985e-08, "loss": 0.7122, "step": 17537 }, { "epoch": 0.99, "learning_rate": 1.3974947914485725e-08, "loss": 0.7913, "step": 17538 }, { "epoch": 0.99, "learning_rate": 1.3669500753099585e-08, "loss": 0.6482, "step": 17539 }, { "epoch": 0.99, "learning_rate": 1.3367428392607739e-08, "loss": 0.5564, "step": 17540 }, { "epoch": 1.0, "learning_rate": 1.3068730843179833e-08, "loss": 0.6731, "step": 17541 }, { "epoch": 1.0, "learning_rate": 1.2773408114907792e-08, "loss": 0.634, "step": 17542 }, { "epoch": 1.0, "learning_rate": 1.2481460217761421e-08, "loss": 0.6971, "step": 17543 }, { "epoch": 1.0, "learning_rate": 1.2192887161588395e-08, "loss": 0.6791, "step": 17544 }, { "epoch": 1.0, "learning_rate": 1.1907688956136477e-08, "loss": 0.7686, "step": 17545 }, { "epoch": 1.0, "learning_rate": 1.1625865611031295e-08, "loss": 0.7204, "step": 17546 }, { "epoch": 1.0, "learning_rate": 1.1347417135776362e-08, "loss": 0.6924, "step": 17547 }, { "epoch": 1.0, "learning_rate": 1.1072343539775264e-08, "loss": 0.715, "step": 17548 }, { "epoch": 1.0, "learning_rate": 1.0800644832320572e-08, "loss": 0.7137, "step": 17549 }, { "epoch": 1.0, "learning_rate": 1.0532321022571623e-08, "loss": 0.7497, "step": 17550 }, { "epoch": 1.0, "learning_rate": 1.0267372119587837e-08, "loss": 0.7403, "step": 17551 }, { "epoch": 1.0, "learning_rate": 1.0005798132317611e-08, "loss": 0.6132, "step": 17552 }, { "epoch": 1.0, "learning_rate": 9.747599069576119e-09, "loss": 0.7178, "step": 17553 }, { "epoch": 1.0, "learning_rate": 9.49277494008971e-09, "loss": 0.7382, "step": 17554 }, { "epoch": 1.0, "learning_rate": 9.241325752462615e-09, "loss": 0.653, "step": 17555 }, { "epoch": 1.0, "learning_rate": 8.993251515176937e-09, "loss": 0.7667, "step": 17556 }, { "epoch": 1.0, "learning_rate": 8.748552236603757e-09, "loss": 0.6497, "step": 17557 }, { "epoch": 1.0, "learning_rate": 8.507227924992034e-09, "loss": 0.7236, "step": 17558 }, { "epoch": 1.0, "learning_rate": 8.269278588513008e-09, "loss": 0.6968, "step": 17559 }, { "epoch": 1.0, "learning_rate": 8.034704235182488e-09, "loss": 0.7042, "step": 17560 }, { "epoch": 1.0, "learning_rate": 7.803504872916367e-09, "loss": 0.6757, "step": 17561 }, { "epoch": 1.0, "learning_rate": 7.575680509519512e-09, "loss": 0.6531, "step": 17562 }, { "epoch": 1.0, "learning_rate": 7.3512311526746645e-09, "loss": 0.6385, "step": 17563 }, { "epoch": 1.0, "learning_rate": 7.130156809975752e-09, "loss": 0.6271, "step": 17564 }, { "epoch": 1.0, "learning_rate": 6.9124574888723706e-09, "loss": 0.7298, "step": 17565 }, { "epoch": 1.0, "learning_rate": 6.698133196725298e-09, "loss": 0.7729, "step": 17566 }, { "epoch": 1.0, "learning_rate": 6.4871839407509846e-09, "loss": 0.7175, "step": 17567 }, { "epoch": 1.0, "learning_rate": 6.279609728077063e-09, "loss": 0.6761, "step": 17568 }, { "epoch": 1.0, "learning_rate": 6.075410565697937e-09, "loss": 0.7006, "step": 17569 }, { "epoch": 1.0, "learning_rate": 5.874586460530296e-09, "loss": 0.669, "step": 17570 }, { "epoch": 1.0, "learning_rate": 5.677137419335399e-09, "loss": 0.7837, "step": 17571 }, { "epoch": 1.0, "learning_rate": 5.483063448785686e-09, "loss": 0.5719, "step": 17572 }, { "epoch": 1.0, "learning_rate": 5.292364555431473e-09, "loss": 0.719, "step": 17573 }, { "epoch": 1.0, "learning_rate": 5.105040745689849e-09, "loss": 0.5918, "step": 17574 }, { "epoch": 1.0, "learning_rate": 4.921092025911289e-09, "loss": 0.6249, "step": 17575 }, { "epoch": 1.0, "learning_rate": 4.740518402290839e-09, "loss": 0.7372, "step": 17576 }, { "epoch": 1.0, "learning_rate": 4.563319880923622e-09, "loss": 0.6424, "step": 17577 }, { "epoch": 1.0, "learning_rate": 4.389496467793741e-09, "loss": 0.6953, "step": 17578 }, { "epoch": 1.0, "learning_rate": 4.219048168763174e-09, "loss": 0.6162, "step": 17579 }, { "epoch": 1.0, "learning_rate": 4.051974989593976e-09, "loss": 0.7265, "step": 17580 }, { "epoch": 1.0, "learning_rate": 3.888276935914981e-09, "loss": 0.7129, "step": 17581 }, { "epoch": 1.0, "learning_rate": 3.727954013266199e-09, "loss": 0.6248, "step": 17582 }, { "epoch": 1.0, "learning_rate": 3.571006227032214e-09, "loss": 0.6917, "step": 17583 }, { "epoch": 1.0, "learning_rate": 3.4174335825420957e-09, "loss": 0.713, "step": 17584 }, { "epoch": 1.0, "learning_rate": 3.2672360849472782e-09, "loss": 0.7213, "step": 17585 }, { "epoch": 1.0, "learning_rate": 3.120413739343686e-09, "loss": 0.6898, "step": 17586 }, { "epoch": 1.0, "learning_rate": 2.9769665506829137e-09, "loss": 0.6773, "step": 17587 }, { "epoch": 1.0, "learning_rate": 2.836894523794431e-09, "loss": 0.6198, "step": 17588 }, { "epoch": 1.0, "learning_rate": 2.7001976634188907e-09, "loss": 0.6836, "step": 17589 }, { "epoch": 1.0, "learning_rate": 2.5668759741637184e-09, "loss": 0.7636, "step": 17590 }, { "epoch": 1.0, "learning_rate": 2.4369294605253166e-09, "loss": 0.8194, "step": 17591 }, { "epoch": 1.0, "learning_rate": 2.310358126889067e-09, "loss": 0.6807, "step": 17592 }, { "epoch": 1.0, "learning_rate": 2.187161977540431e-09, "loss": 0.6232, "step": 17593 }, { "epoch": 1.0, "learning_rate": 2.0673410166205387e-09, "loss": 0.7249, "step": 17594 }, { "epoch": 1.0, "learning_rate": 1.9508952481928077e-09, "loss": 0.7069, "step": 17595 }, { "epoch": 1.0, "learning_rate": 1.8378246761652229e-09, "loss": 0.6999, "step": 17596 }, { "epoch": 1.0, "learning_rate": 1.7281293043791557e-09, "loss": 0.7256, "step": 17597 }, { "epoch": 1.0, "learning_rate": 1.6218091365094445e-09, "loss": 0.6517, "step": 17598 }, { "epoch": 1.0, "learning_rate": 1.5188641761643142e-09, "loss": 0.7171, "step": 17599 }, { "epoch": 1.0, "learning_rate": 1.4192944268187625e-09, "loss": 0.7895, "step": 17600 }, { "epoch": 1.0, "learning_rate": 1.3230998918256631e-09, "loss": 0.7976, "step": 17601 }, { "epoch": 1.0, "learning_rate": 1.2302805744268674e-09, "loss": 0.7357, "step": 17602 }, { "epoch": 1.0, "learning_rate": 1.1408364777754088e-09, "loss": 0.6304, "step": 17603 }, { "epoch": 1.0, "learning_rate": 1.0547676048688892e-09, "loss": 0.7214, "step": 17604 }, { "epoch": 1.0, "learning_rate": 9.720739586160931e-10, "loss": 0.6061, "step": 17605 }, { "epoch": 1.0, "learning_rate": 8.927555418258848e-10, "loss": 0.6737, "step": 17606 }, { "epoch": 1.0, "learning_rate": 8.168123571516972e-10, "loss": 0.6766, "step": 17607 }, { "epoch": 1.0, "learning_rate": 7.442444071692478e-10, "loss": 0.6473, "step": 17608 }, { "epoch": 1.0, "learning_rate": 6.750516943321295e-10, "loss": 0.6675, "step": 17609 }, { "epoch": 1.0, "learning_rate": 6.092342209607083e-10, "loss": 0.6621, "step": 17610 }, { "epoch": 1.0, "learning_rate": 5.467919892865325e-10, "loss": 0.7433, "step": 17611 }, { "epoch": 1.0, "learning_rate": 4.877250014301283e-10, "loss": 0.7158, "step": 17612 }, { "epoch": 1.0, "learning_rate": 4.320332593565901e-10, "loss": 0.6709, "step": 17613 }, { "epoch": 1.0, "learning_rate": 3.7971676496439956e-10, "loss": 0.7198, "step": 17614 }, { "epoch": 1.0, "learning_rate": 3.307755200188112e-10, "loss": 0.6047, "step": 17615 }, { "epoch": 1.0, "learning_rate": 2.8520952616295503e-10, "loss": 0.7419, "step": 17616 }, { "epoch": 1.0, "learning_rate": 2.4301878494004116e-10, "loss": 0.5218, "step": 17617 }, { "epoch": 1.0, "learning_rate": 2.0420329778225722e-10, "loss": 0.5928, "step": 17618 }, { "epoch": 1.0, "learning_rate": 1.6876306598856417e-10, "loss": 0.6283, "step": 17619 }, { "epoch": 1.0, "learning_rate": 1.3669809075800288e-10, "loss": 0.7463, "step": 17620 }, { "epoch": 1.0, "learning_rate": 1.0800837316748968e-10, "loss": 0.6502, "step": 17621 }, { "epoch": 1.0, "learning_rate": 8.269391419402083e-11, "loss": 0.6956, "step": 17622 }, { "epoch": 1.0, "learning_rate": 6.075471469246808e-11, "loss": 0.6771, "step": 17623 }, { "epoch": 1.0, "learning_rate": 4.2190775395578584e-11, "loss": 0.73, "step": 17624 }, { "epoch": 1.0, "learning_rate": 2.70020969361795e-11, "loss": 0.6667, "step": 17625 }, { "epoch": 1.0, "learning_rate": 1.5188679824973405e-11, "loss": 0.6819, "step": 17626 }, { "epoch": 1.0, "learning_rate": 6.750524461640595e-12, "loss": 0.6962, "step": 17627 }, { "epoch": 1.0, "learning_rate": 1.6876311348390517e-12, "loss": 0.6265, "step": 17628 }, { "epoch": 1.0, "learning_rate": 0.0, "loss": 0.6693, "step": 17629 }, { "epoch": 1.0, "step": 17629, "total_flos": 2311539419627520.0, "train_loss": 0.7206904479721453, "train_runtime": 58029.5997, "train_samples_per_second": 19.442, "train_steps_per_second": 0.304 } ], "logging_steps": 1.0, "max_steps": 17629, "num_input_tokens_seen": 0, "num_train_epochs": 1, "save_steps": 1000, "total_flos": 2311539419627520.0, "train_batch_size": 8, "trial_name": null, "trial_params": null }