{ "best_metric": null, "best_model_checkpoint": null, "epoch": 1.0, "eval_steps": 10.0, "global_step": 15596, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 4.273504273504274e-07, "loss": 2.2024, "step": 1 }, { "epoch": 0.0, "learning_rate": 8.547008547008548e-07, "loss": 2.0973, "step": 2 }, { "epoch": 0.0, "learning_rate": 1.282051282051282e-06, "loss": 2.1711, "step": 3 }, { "epoch": 0.0, "learning_rate": 1.7094017094017097e-06, "loss": 2.1481, "step": 4 }, { "epoch": 0.0, "learning_rate": 2.136752136752137e-06, "loss": 1.9849, "step": 5 }, { "epoch": 0.0, "learning_rate": 2.564102564102564e-06, "loss": 2.2166, "step": 6 }, { "epoch": 0.0, "learning_rate": 2.991452991452992e-06, "loss": 2.2209, "step": 7 }, { "epoch": 0.0, "learning_rate": 3.4188034188034193e-06, "loss": 2.1768, "step": 8 }, { "epoch": 0.0, "learning_rate": 3.846153846153847e-06, "loss": 2.175, "step": 9 }, { "epoch": 0.0, "learning_rate": 4.273504273504274e-06, "loss": 1.9895, "step": 10 }, { "epoch": 0.0, "learning_rate": 4.700854700854701e-06, "loss": 2.2848, "step": 11 }, { "epoch": 0.0, "learning_rate": 5.128205128205128e-06, "loss": 2.0633, "step": 12 }, { "epoch": 0.0, "learning_rate": 5.555555555555556e-06, "loss": 2.161, "step": 13 }, { "epoch": 0.0, "learning_rate": 5.982905982905984e-06, "loss": 2.0673, "step": 14 }, { "epoch": 0.0, "learning_rate": 6.41025641025641e-06, "loss": 2.0508, "step": 15 }, { "epoch": 0.0, "learning_rate": 6.837606837606839e-06, "loss": 1.9866, "step": 16 }, { "epoch": 0.0, "learning_rate": 7.264957264957266e-06, "loss": 2.08, "step": 17 }, { "epoch": 0.0, "learning_rate": 7.692307692307694e-06, "loss": 1.8909, "step": 18 }, { "epoch": 0.0, "learning_rate": 8.11965811965812e-06, "loss": 1.857, "step": 19 }, { "epoch": 0.0, "learning_rate": 8.547008547008548e-06, "loss": 1.9067, "step": 20 }, { "epoch": 0.0, "learning_rate": 8.974358974358976e-06, "loss": 1.8523, "step": 21 }, { "epoch": 0.0, "learning_rate": 9.401709401709402e-06, "loss": 1.83, "step": 22 }, { "epoch": 0.0, "learning_rate": 9.82905982905983e-06, "loss": 1.8249, "step": 23 }, { "epoch": 0.0, "learning_rate": 1.0256410256410256e-05, "loss": 1.798, "step": 24 }, { "epoch": 0.0, "learning_rate": 1.0683760683760684e-05, "loss": 1.6432, "step": 25 }, { "epoch": 0.0, "learning_rate": 1.1111111111111112e-05, "loss": 1.6308, "step": 26 }, { "epoch": 0.0, "learning_rate": 1.153846153846154e-05, "loss": 1.5276, "step": 27 }, { "epoch": 0.0, "learning_rate": 1.1965811965811967e-05, "loss": 1.4847, "step": 28 }, { "epoch": 0.0, "learning_rate": 1.2393162393162394e-05, "loss": 1.3408, "step": 29 }, { "epoch": 0.0, "learning_rate": 1.282051282051282e-05, "loss": 1.3173, "step": 30 }, { "epoch": 0.0, "learning_rate": 1.324786324786325e-05, "loss": 1.3263, "step": 31 }, { "epoch": 0.0, "learning_rate": 1.3675213675213677e-05, "loss": 1.3267, "step": 32 }, { "epoch": 0.0, "learning_rate": 1.4102564102564104e-05, "loss": 1.3007, "step": 33 }, { "epoch": 0.0, "learning_rate": 1.4529914529914531e-05, "loss": 1.1593, "step": 34 }, { "epoch": 0.0, "learning_rate": 1.4957264957264958e-05, "loss": 1.1569, "step": 35 }, { "epoch": 0.0, "learning_rate": 1.5384615384615387e-05, "loss": 1.0768, "step": 36 }, { "epoch": 0.0, "learning_rate": 1.581196581196581e-05, "loss": 0.9651, "step": 37 }, { "epoch": 0.0, "learning_rate": 1.623931623931624e-05, "loss": 0.9634, "step": 38 }, { "epoch": 0.0, "learning_rate": 1.6666666666666667e-05, "loss": 1.0149, "step": 39 }, { "epoch": 0.0, "learning_rate": 1.7094017094017095e-05, "loss": 1.0115, "step": 40 }, { "epoch": 0.0, "learning_rate": 1.752136752136752e-05, "loss": 0.8136, "step": 41 }, { "epoch": 0.0, "learning_rate": 1.794871794871795e-05, "loss": 0.7424, "step": 42 }, { "epoch": 0.0, "learning_rate": 1.837606837606838e-05, "loss": 0.7743, "step": 43 }, { "epoch": 0.0, "learning_rate": 1.8803418803418804e-05, "loss": 0.7617, "step": 44 }, { "epoch": 0.0, "learning_rate": 1.923076923076923e-05, "loss": 0.8008, "step": 45 }, { "epoch": 0.0, "learning_rate": 1.965811965811966e-05, "loss": 0.6991, "step": 46 }, { "epoch": 0.0, "learning_rate": 2.0085470085470087e-05, "loss": 0.616, "step": 47 }, { "epoch": 0.0, "learning_rate": 2.0512820512820512e-05, "loss": 0.6265, "step": 48 }, { "epoch": 0.0, "learning_rate": 2.0940170940170943e-05, "loss": 0.7199, "step": 49 }, { "epoch": 0.0, "learning_rate": 2.1367521367521368e-05, "loss": 0.6932, "step": 50 }, { "epoch": 0.0, "learning_rate": 2.1794871794871795e-05, "loss": 0.6399, "step": 51 }, { "epoch": 0.0, "learning_rate": 2.2222222222222223e-05, "loss": 0.625, "step": 52 }, { "epoch": 0.0, "learning_rate": 2.264957264957265e-05, "loss": 0.5366, "step": 53 }, { "epoch": 0.0, "learning_rate": 2.307692307692308e-05, "loss": 0.546, "step": 54 }, { "epoch": 0.0, "learning_rate": 2.3504273504273504e-05, "loss": 0.6041, "step": 55 }, { "epoch": 0.0, "learning_rate": 2.3931623931623935e-05, "loss": 0.5764, "step": 56 }, { "epoch": 0.0, "learning_rate": 2.435897435897436e-05, "loss": 0.5875, "step": 57 }, { "epoch": 0.0, "learning_rate": 2.4786324786324787e-05, "loss": 0.5612, "step": 58 }, { "epoch": 0.0, "learning_rate": 2.5213675213675215e-05, "loss": 0.4755, "step": 59 }, { "epoch": 0.0, "learning_rate": 2.564102564102564e-05, "loss": 0.5238, "step": 60 }, { "epoch": 0.0, "learning_rate": 2.606837606837607e-05, "loss": 0.5042, "step": 61 }, { "epoch": 0.0, "learning_rate": 2.64957264957265e-05, "loss": 0.4918, "step": 62 }, { "epoch": 0.0, "learning_rate": 2.6923076923076923e-05, "loss": 0.5399, "step": 63 }, { "epoch": 0.0, "learning_rate": 2.7350427350427355e-05, "loss": 0.5545, "step": 64 }, { "epoch": 0.0, "learning_rate": 2.777777777777778e-05, "loss": 0.49, "step": 65 }, { "epoch": 0.0, "learning_rate": 2.8205128205128207e-05, "loss": 0.4359, "step": 66 }, { "epoch": 0.0, "learning_rate": 2.863247863247863e-05, "loss": 0.4028, "step": 67 }, { "epoch": 0.0, "learning_rate": 2.9059829059829063e-05, "loss": 0.4014, "step": 68 }, { "epoch": 0.0, "learning_rate": 2.948717948717949e-05, "loss": 0.4163, "step": 69 }, { "epoch": 0.0, "learning_rate": 2.9914529914529915e-05, "loss": 0.3771, "step": 70 }, { "epoch": 0.0, "learning_rate": 3.034188034188034e-05, "loss": 0.3697, "step": 71 }, { "epoch": 0.0, "learning_rate": 3.0769230769230774e-05, "loss": 0.3598, "step": 72 }, { "epoch": 0.0, "learning_rate": 3.1196581196581195e-05, "loss": 0.4058, "step": 73 }, { "epoch": 0.0, "learning_rate": 3.162393162393162e-05, "loss": 0.334, "step": 74 }, { "epoch": 0.0, "learning_rate": 3.205128205128206e-05, "loss": 0.3833, "step": 75 }, { "epoch": 0.0, "learning_rate": 3.247863247863248e-05, "loss": 0.3409, "step": 76 }, { "epoch": 0.0, "learning_rate": 3.290598290598291e-05, "loss": 0.3046, "step": 77 }, { "epoch": 0.01, "learning_rate": 3.3333333333333335e-05, "loss": 0.4491, "step": 78 }, { "epoch": 0.01, "learning_rate": 3.376068376068376e-05, "loss": 0.3423, "step": 79 }, { "epoch": 0.01, "learning_rate": 3.418803418803419e-05, "loss": 0.325, "step": 80 }, { "epoch": 0.01, "learning_rate": 3.461538461538462e-05, "loss": 0.3713, "step": 81 }, { "epoch": 0.01, "learning_rate": 3.504273504273504e-05, "loss": 0.3057, "step": 82 }, { "epoch": 0.01, "learning_rate": 3.5470085470085474e-05, "loss": 0.3446, "step": 83 }, { "epoch": 0.01, "learning_rate": 3.58974358974359e-05, "loss": 0.3262, "step": 84 }, { "epoch": 0.01, "learning_rate": 3.6324786324786323e-05, "loss": 0.3346, "step": 85 }, { "epoch": 0.01, "learning_rate": 3.675213675213676e-05, "loss": 0.3198, "step": 86 }, { "epoch": 0.01, "learning_rate": 3.717948717948718e-05, "loss": 0.226, "step": 87 }, { "epoch": 0.01, "learning_rate": 3.760683760683761e-05, "loss": 0.3139, "step": 88 }, { "epoch": 0.01, "learning_rate": 3.8034188034188035e-05, "loss": 0.3114, "step": 89 }, { "epoch": 0.01, "learning_rate": 3.846153846153846e-05, "loss": 0.2975, "step": 90 }, { "epoch": 0.01, "learning_rate": 3.888888888888889e-05, "loss": 0.2665, "step": 91 }, { "epoch": 0.01, "learning_rate": 3.931623931623932e-05, "loss": 0.2797, "step": 92 }, { "epoch": 0.01, "learning_rate": 3.974358974358974e-05, "loss": 0.2971, "step": 93 }, { "epoch": 0.01, "learning_rate": 4.0170940170940174e-05, "loss": 0.3003, "step": 94 }, { "epoch": 0.01, "learning_rate": 4.05982905982906e-05, "loss": 0.2782, "step": 95 }, { "epoch": 0.01, "learning_rate": 4.1025641025641023e-05, "loss": 0.2787, "step": 96 }, { "epoch": 0.01, "learning_rate": 4.145299145299146e-05, "loss": 0.3264, "step": 97 }, { "epoch": 0.01, "learning_rate": 4.1880341880341886e-05, "loss": 0.2774, "step": 98 }, { "epoch": 0.01, "learning_rate": 4.230769230769231e-05, "loss": 0.2611, "step": 99 }, { "epoch": 0.01, "learning_rate": 4.2735042735042735e-05, "loss": 0.2821, "step": 100 }, { "epoch": 0.01, "learning_rate": 4.316239316239317e-05, "loss": 0.2594, "step": 101 }, { "epoch": 0.01, "learning_rate": 4.358974358974359e-05, "loss": 0.262, "step": 102 }, { "epoch": 0.01, "learning_rate": 4.401709401709402e-05, "loss": 0.2825, "step": 103 }, { "epoch": 0.01, "learning_rate": 4.4444444444444447e-05, "loss": 0.2761, "step": 104 }, { "epoch": 0.01, "learning_rate": 4.4871794871794874e-05, "loss": 0.2676, "step": 105 }, { "epoch": 0.01, "learning_rate": 4.52991452991453e-05, "loss": 0.316, "step": 106 }, { "epoch": 0.01, "learning_rate": 4.572649572649573e-05, "loss": 0.2214, "step": 107 }, { "epoch": 0.01, "learning_rate": 4.615384615384616e-05, "loss": 0.2368, "step": 108 }, { "epoch": 0.01, "learning_rate": 4.6581196581196586e-05, "loss": 0.2876, "step": 109 }, { "epoch": 0.01, "learning_rate": 4.700854700854701e-05, "loss": 0.2387, "step": 110 }, { "epoch": 0.01, "learning_rate": 4.7435897435897435e-05, "loss": 0.2767, "step": 111 }, { "epoch": 0.01, "learning_rate": 4.786324786324787e-05, "loss": 0.2635, "step": 112 }, { "epoch": 0.01, "learning_rate": 4.829059829059829e-05, "loss": 0.2935, "step": 113 }, { "epoch": 0.01, "learning_rate": 4.871794871794872e-05, "loss": 0.2636, "step": 114 }, { "epoch": 0.01, "learning_rate": 4.9145299145299147e-05, "loss": 0.2678, "step": 115 }, { "epoch": 0.01, "learning_rate": 4.9572649572649575e-05, "loss": 0.2339, "step": 116 }, { "epoch": 0.01, "learning_rate": 5e-05, "loss": 0.2466, "step": 117 }, { "epoch": 0.01, "learning_rate": 5.042735042735043e-05, "loss": 0.3245, "step": 118 }, { "epoch": 0.01, "learning_rate": 5.085470085470085e-05, "loss": 0.2429, "step": 119 }, { "epoch": 0.01, "learning_rate": 5.128205128205128e-05, "loss": 0.2507, "step": 120 }, { "epoch": 0.01, "learning_rate": 5.1709401709401714e-05, "loss": 0.2643, "step": 121 }, { "epoch": 0.01, "learning_rate": 5.213675213675214e-05, "loss": 0.2461, "step": 122 }, { "epoch": 0.01, "learning_rate": 5.256410256410257e-05, "loss": 0.2296, "step": 123 }, { "epoch": 0.01, "learning_rate": 5.2991452991453e-05, "loss": 0.2261, "step": 124 }, { "epoch": 0.01, "learning_rate": 5.341880341880342e-05, "loss": 0.2582, "step": 125 }, { "epoch": 0.01, "learning_rate": 5.384615384615385e-05, "loss": 0.2367, "step": 126 }, { "epoch": 0.01, "learning_rate": 5.4273504273504275e-05, "loss": 0.2339, "step": 127 }, { "epoch": 0.01, "learning_rate": 5.470085470085471e-05, "loss": 0.2391, "step": 128 }, { "epoch": 0.01, "learning_rate": 5.512820512820514e-05, "loss": 0.228, "step": 129 }, { "epoch": 0.01, "learning_rate": 5.555555555555556e-05, "loss": 0.1913, "step": 130 }, { "epoch": 0.01, "learning_rate": 5.5982905982905986e-05, "loss": 0.2987, "step": 131 }, { "epoch": 0.01, "learning_rate": 5.6410256410256414e-05, "loss": 0.2525, "step": 132 }, { "epoch": 0.01, "learning_rate": 5.6837606837606835e-05, "loss": 0.238, "step": 133 }, { "epoch": 0.01, "learning_rate": 5.726495726495726e-05, "loss": 0.2679, "step": 134 }, { "epoch": 0.01, "learning_rate": 5.769230769230769e-05, "loss": 0.2252, "step": 135 }, { "epoch": 0.01, "learning_rate": 5.8119658119658126e-05, "loss": 0.2325, "step": 136 }, { "epoch": 0.01, "learning_rate": 5.8547008547008553e-05, "loss": 0.2247, "step": 137 }, { "epoch": 0.01, "learning_rate": 5.897435897435898e-05, "loss": 0.218, "step": 138 }, { "epoch": 0.01, "learning_rate": 5.94017094017094e-05, "loss": 0.2221, "step": 139 }, { "epoch": 0.01, "learning_rate": 5.982905982905983e-05, "loss": 0.2218, "step": 140 }, { "epoch": 0.01, "learning_rate": 6.025641025641026e-05, "loss": 0.2451, "step": 141 }, { "epoch": 0.01, "learning_rate": 6.068376068376068e-05, "loss": 0.215, "step": 142 }, { "epoch": 0.01, "learning_rate": 6.111111111111112e-05, "loss": 0.2509, "step": 143 }, { "epoch": 0.01, "learning_rate": 6.153846153846155e-05, "loss": 0.2303, "step": 144 }, { "epoch": 0.01, "learning_rate": 6.196581196581198e-05, "loss": 0.2204, "step": 145 }, { "epoch": 0.01, "learning_rate": 6.239316239316239e-05, "loss": 0.261, "step": 146 }, { "epoch": 0.01, "learning_rate": 6.282051282051282e-05, "loss": 0.2085, "step": 147 }, { "epoch": 0.01, "learning_rate": 6.324786324786325e-05, "loss": 0.1935, "step": 148 }, { "epoch": 0.01, "learning_rate": 6.367521367521367e-05, "loss": 0.2316, "step": 149 }, { "epoch": 0.01, "learning_rate": 6.410256410256412e-05, "loss": 0.2248, "step": 150 }, { "epoch": 0.01, "learning_rate": 6.452991452991453e-05, "loss": 0.2507, "step": 151 }, { "epoch": 0.01, "learning_rate": 6.495726495726496e-05, "loss": 0.2106, "step": 152 }, { "epoch": 0.01, "learning_rate": 6.538461538461539e-05, "loss": 0.1884, "step": 153 }, { "epoch": 0.01, "learning_rate": 6.581196581196581e-05, "loss": 0.2191, "step": 154 }, { "epoch": 0.01, "learning_rate": 6.623931623931624e-05, "loss": 0.2129, "step": 155 }, { "epoch": 0.01, "learning_rate": 6.666666666666667e-05, "loss": 0.2018, "step": 156 }, { "epoch": 0.01, "learning_rate": 6.70940170940171e-05, "loss": 0.2201, "step": 157 }, { "epoch": 0.01, "learning_rate": 6.752136752136753e-05, "loss": 0.253, "step": 158 }, { "epoch": 0.01, "learning_rate": 6.794871794871795e-05, "loss": 0.2196, "step": 159 }, { "epoch": 0.01, "learning_rate": 6.837606837606838e-05, "loss": 0.2238, "step": 160 }, { "epoch": 0.01, "learning_rate": 6.880341880341881e-05, "loss": 0.2362, "step": 161 }, { "epoch": 0.01, "learning_rate": 6.923076923076924e-05, "loss": 0.1969, "step": 162 }, { "epoch": 0.01, "learning_rate": 6.965811965811965e-05, "loss": 0.2309, "step": 163 }, { "epoch": 0.01, "learning_rate": 7.008547008547008e-05, "loss": 0.206, "step": 164 }, { "epoch": 0.01, "learning_rate": 7.051282051282052e-05, "loss": 0.2209, "step": 165 }, { "epoch": 0.01, "learning_rate": 7.094017094017095e-05, "loss": 0.2135, "step": 166 }, { "epoch": 0.01, "learning_rate": 7.136752136752138e-05, "loss": 0.2274, "step": 167 }, { "epoch": 0.01, "learning_rate": 7.17948717948718e-05, "loss": 0.2085, "step": 168 }, { "epoch": 0.01, "learning_rate": 7.222222222222222e-05, "loss": 0.1974, "step": 169 }, { "epoch": 0.01, "learning_rate": 7.264957264957265e-05, "loss": 0.2223, "step": 170 }, { "epoch": 0.01, "learning_rate": 7.307692307692307e-05, "loss": 0.2262, "step": 171 }, { "epoch": 0.01, "learning_rate": 7.350427350427352e-05, "loss": 0.2154, "step": 172 }, { "epoch": 0.01, "learning_rate": 7.393162393162394e-05, "loss": 0.199, "step": 173 }, { "epoch": 0.01, "learning_rate": 7.435897435897436e-05, "loss": 0.1885, "step": 174 }, { "epoch": 0.01, "learning_rate": 7.478632478632479e-05, "loss": 0.193, "step": 175 }, { "epoch": 0.01, "learning_rate": 7.521367521367521e-05, "loss": 0.2212, "step": 176 }, { "epoch": 0.01, "learning_rate": 7.564102564102564e-05, "loss": 0.2076, "step": 177 }, { "epoch": 0.01, "learning_rate": 7.606837606837607e-05, "loss": 0.2235, "step": 178 }, { "epoch": 0.01, "learning_rate": 7.64957264957265e-05, "loss": 0.2175, "step": 179 }, { "epoch": 0.01, "learning_rate": 7.692307692307693e-05, "loss": 0.1879, "step": 180 }, { "epoch": 0.01, "learning_rate": 7.735042735042735e-05, "loss": 0.2014, "step": 181 }, { "epoch": 0.01, "learning_rate": 7.777777777777778e-05, "loss": 0.2194, "step": 182 }, { "epoch": 0.01, "learning_rate": 7.820512820512821e-05, "loss": 0.2017, "step": 183 }, { "epoch": 0.01, "learning_rate": 7.863247863247864e-05, "loss": 0.2098, "step": 184 }, { "epoch": 0.01, "learning_rate": 7.905982905982907e-05, "loss": 0.2002, "step": 185 }, { "epoch": 0.01, "learning_rate": 7.948717948717948e-05, "loss": 0.2134, "step": 186 }, { "epoch": 0.01, "learning_rate": 7.991452991452992e-05, "loss": 0.2148, "step": 187 }, { "epoch": 0.01, "learning_rate": 8.034188034188035e-05, "loss": 0.2394, "step": 188 }, { "epoch": 0.01, "learning_rate": 8.076923076923078e-05, "loss": 0.2157, "step": 189 }, { "epoch": 0.01, "learning_rate": 8.11965811965812e-05, "loss": 0.176, "step": 190 }, { "epoch": 0.01, "learning_rate": 8.162393162393163e-05, "loss": 0.2229, "step": 191 }, { "epoch": 0.01, "learning_rate": 8.205128205128205e-05, "loss": 0.2052, "step": 192 }, { "epoch": 0.01, "learning_rate": 8.247863247863247e-05, "loss": 0.2058, "step": 193 }, { "epoch": 0.01, "learning_rate": 8.290598290598292e-05, "loss": 0.1862, "step": 194 }, { "epoch": 0.01, "learning_rate": 8.333333333333334e-05, "loss": 0.1957, "step": 195 }, { "epoch": 0.01, "learning_rate": 8.376068376068377e-05, "loss": 0.1909, "step": 196 }, { "epoch": 0.01, "learning_rate": 8.418803418803419e-05, "loss": 0.1986, "step": 197 }, { "epoch": 0.01, "learning_rate": 8.461538461538461e-05, "loss": 0.2081, "step": 198 }, { "epoch": 0.01, "learning_rate": 8.504273504273504e-05, "loss": 0.1768, "step": 199 }, { "epoch": 0.01, "learning_rate": 8.547008547008547e-05, "loss": 0.183, "step": 200 }, { "epoch": 0.01, "learning_rate": 8.58974358974359e-05, "loss": 0.2077, "step": 201 }, { "epoch": 0.01, "learning_rate": 8.632478632478634e-05, "loss": 0.2396, "step": 202 }, { "epoch": 0.01, "learning_rate": 8.675213675213675e-05, "loss": 0.204, "step": 203 }, { "epoch": 0.01, "learning_rate": 8.717948717948718e-05, "loss": 0.1948, "step": 204 }, { "epoch": 0.01, "learning_rate": 8.760683760683761e-05, "loss": 0.1908, "step": 205 }, { "epoch": 0.01, "learning_rate": 8.803418803418804e-05, "loss": 0.1671, "step": 206 }, { "epoch": 0.01, "learning_rate": 8.846153846153847e-05, "loss": 0.1928, "step": 207 }, { "epoch": 0.01, "learning_rate": 8.888888888888889e-05, "loss": 0.2057, "step": 208 }, { "epoch": 0.01, "learning_rate": 8.931623931623932e-05, "loss": 0.2159, "step": 209 }, { "epoch": 0.01, "learning_rate": 8.974358974358975e-05, "loss": 0.2019, "step": 210 }, { "epoch": 0.01, "learning_rate": 9.017094017094018e-05, "loss": 0.231, "step": 211 }, { "epoch": 0.01, "learning_rate": 9.05982905982906e-05, "loss": 0.2127, "step": 212 }, { "epoch": 0.01, "learning_rate": 9.102564102564103e-05, "loss": 0.19, "step": 213 }, { "epoch": 0.01, "learning_rate": 9.145299145299146e-05, "loss": 0.2093, "step": 214 }, { "epoch": 0.01, "learning_rate": 9.188034188034187e-05, "loss": 0.1787, "step": 215 }, { "epoch": 0.01, "learning_rate": 9.230769230769232e-05, "loss": 0.1981, "step": 216 }, { "epoch": 0.01, "learning_rate": 9.273504273504274e-05, "loss": 0.1918, "step": 217 }, { "epoch": 0.01, "learning_rate": 9.316239316239317e-05, "loss": 0.2051, "step": 218 }, { "epoch": 0.01, "learning_rate": 9.35897435897436e-05, "loss": 0.1981, "step": 219 }, { "epoch": 0.01, "learning_rate": 9.401709401709401e-05, "loss": 0.2133, "step": 220 }, { "epoch": 0.01, "learning_rate": 9.444444444444444e-05, "loss": 0.1746, "step": 221 }, { "epoch": 0.01, "learning_rate": 9.487179487179487e-05, "loss": 0.1887, "step": 222 }, { "epoch": 0.01, "learning_rate": 9.52991452991453e-05, "loss": 0.1985, "step": 223 }, { "epoch": 0.01, "learning_rate": 9.572649572649574e-05, "loss": 0.2242, "step": 224 }, { "epoch": 0.01, "learning_rate": 9.615384615384617e-05, "loss": 0.1936, "step": 225 }, { "epoch": 0.01, "learning_rate": 9.658119658119658e-05, "loss": 0.1822, "step": 226 }, { "epoch": 0.01, "learning_rate": 9.700854700854701e-05, "loss": 0.1774, "step": 227 }, { "epoch": 0.01, "learning_rate": 9.743589743589744e-05, "loss": 0.2074, "step": 228 }, { "epoch": 0.01, "learning_rate": 9.786324786324787e-05, "loss": 0.2111, "step": 229 }, { "epoch": 0.01, "learning_rate": 9.829059829059829e-05, "loss": 0.1961, "step": 230 }, { "epoch": 0.01, "learning_rate": 9.871794871794872e-05, "loss": 0.2001, "step": 231 }, { "epoch": 0.01, "learning_rate": 9.914529914529915e-05, "loss": 0.2172, "step": 232 }, { "epoch": 0.01, "learning_rate": 9.957264957264958e-05, "loss": 0.1839, "step": 233 }, { "epoch": 0.02, "learning_rate": 0.0001, "loss": 0.1904, "step": 234 }, { "epoch": 0.02, "learning_rate": 0.00010042735042735043, "loss": 0.2019, "step": 235 }, { "epoch": 0.02, "learning_rate": 0.00010085470085470086, "loss": 0.1899, "step": 236 }, { "epoch": 0.02, "learning_rate": 0.00010128205128205129, "loss": 0.208, "step": 237 }, { "epoch": 0.02, "learning_rate": 0.0001017094017094017, "loss": 0.2056, "step": 238 }, { "epoch": 0.02, "learning_rate": 0.00010213675213675213, "loss": 0.1792, "step": 239 }, { "epoch": 0.02, "learning_rate": 0.00010256410256410256, "loss": 0.1773, "step": 240 }, { "epoch": 0.02, "learning_rate": 0.00010299145299145299, "loss": 0.2033, "step": 241 }, { "epoch": 0.02, "learning_rate": 0.00010341880341880343, "loss": 0.176, "step": 242 }, { "epoch": 0.02, "learning_rate": 0.00010384615384615386, "loss": 0.205, "step": 243 }, { "epoch": 0.02, "learning_rate": 0.00010427350427350428, "loss": 0.1971, "step": 244 }, { "epoch": 0.02, "learning_rate": 0.00010470085470085471, "loss": 0.2094, "step": 245 }, { "epoch": 0.02, "learning_rate": 0.00010512820512820514, "loss": 0.1937, "step": 246 }, { "epoch": 0.02, "learning_rate": 0.00010555555555555557, "loss": 0.1999, "step": 247 }, { "epoch": 0.02, "learning_rate": 0.000105982905982906, "loss": 0.2, "step": 248 }, { "epoch": 0.02, "learning_rate": 0.00010641025641025641, "loss": 0.1855, "step": 249 }, { "epoch": 0.02, "learning_rate": 0.00010683760683760684, "loss": 0.1765, "step": 250 }, { "epoch": 0.02, "learning_rate": 0.00010726495726495727, "loss": 0.1923, "step": 251 }, { "epoch": 0.02, "learning_rate": 0.0001076923076923077, "loss": 0.2059, "step": 252 }, { "epoch": 0.02, "learning_rate": 0.00010811965811965812, "loss": 0.2078, "step": 253 }, { "epoch": 0.02, "learning_rate": 0.00010854700854700855, "loss": 0.1919, "step": 254 }, { "epoch": 0.02, "learning_rate": 0.00010897435897435896, "loss": 0.1943, "step": 255 }, { "epoch": 0.02, "learning_rate": 0.00010940170940170942, "loss": 0.1889, "step": 256 }, { "epoch": 0.02, "learning_rate": 0.00010982905982905985, "loss": 0.1777, "step": 257 }, { "epoch": 0.02, "learning_rate": 0.00011025641025641027, "loss": 0.2197, "step": 258 }, { "epoch": 0.02, "learning_rate": 0.0001106837606837607, "loss": 0.1823, "step": 259 }, { "epoch": 0.02, "learning_rate": 0.00011111111111111112, "loss": 0.1861, "step": 260 }, { "epoch": 0.02, "learning_rate": 0.00011153846153846154, "loss": 0.2104, "step": 261 }, { "epoch": 0.02, "learning_rate": 0.00011196581196581197, "loss": 0.1747, "step": 262 }, { "epoch": 0.02, "learning_rate": 0.0001123931623931624, "loss": 0.178, "step": 263 }, { "epoch": 0.02, "learning_rate": 0.00011282051282051283, "loss": 0.2207, "step": 264 }, { "epoch": 0.02, "learning_rate": 0.00011324786324786326, "loss": 0.1835, "step": 265 }, { "epoch": 0.02, "learning_rate": 0.00011367521367521367, "loss": 0.1954, "step": 266 }, { "epoch": 0.02, "learning_rate": 0.0001141025641025641, "loss": 0.2233, "step": 267 }, { "epoch": 0.02, "learning_rate": 0.00011452991452991453, "loss": 0.1809, "step": 268 }, { "epoch": 0.02, "learning_rate": 0.00011495726495726495, "loss": 0.1948, "step": 269 }, { "epoch": 0.02, "learning_rate": 0.00011538461538461538, "loss": 0.1998, "step": 270 }, { "epoch": 0.02, "learning_rate": 0.00011581196581196582, "loss": 0.2008, "step": 271 }, { "epoch": 0.02, "learning_rate": 0.00011623931623931625, "loss": 0.1808, "step": 272 }, { "epoch": 0.02, "learning_rate": 0.00011666666666666668, "loss": 0.1801, "step": 273 }, { "epoch": 0.02, "learning_rate": 0.00011709401709401711, "loss": 0.1656, "step": 274 }, { "epoch": 0.02, "learning_rate": 0.00011752136752136753, "loss": 0.1766, "step": 275 }, { "epoch": 0.02, "learning_rate": 0.00011794871794871796, "loss": 0.1678, "step": 276 }, { "epoch": 0.02, "learning_rate": 0.00011837606837606838, "loss": 0.2036, "step": 277 }, { "epoch": 0.02, "learning_rate": 0.0001188034188034188, "loss": 0.1744, "step": 278 }, { "epoch": 0.02, "learning_rate": 0.00011923076923076923, "loss": 0.1746, "step": 279 }, { "epoch": 0.02, "learning_rate": 0.00011965811965811966, "loss": 0.1757, "step": 280 }, { "epoch": 0.02, "learning_rate": 0.00012008547008547009, "loss": 0.1939, "step": 281 }, { "epoch": 0.02, "learning_rate": 0.00012051282051282052, "loss": 0.1776, "step": 282 }, { "epoch": 0.02, "learning_rate": 0.00012094017094017094, "loss": 0.2072, "step": 283 }, { "epoch": 0.02, "learning_rate": 0.00012136752136752136, "loss": 0.1718, "step": 284 }, { "epoch": 0.02, "learning_rate": 0.00012179487179487179, "loss": 0.188, "step": 285 }, { "epoch": 0.02, "learning_rate": 0.00012222222222222224, "loss": 0.1782, "step": 286 }, { "epoch": 0.02, "learning_rate": 0.00012264957264957266, "loss": 0.1683, "step": 287 }, { "epoch": 0.02, "learning_rate": 0.0001230769230769231, "loss": 0.202, "step": 288 }, { "epoch": 0.02, "learning_rate": 0.0001235042735042735, "loss": 0.1837, "step": 289 }, { "epoch": 0.02, "learning_rate": 0.00012393162393162395, "loss": 0.2008, "step": 290 }, { "epoch": 0.02, "learning_rate": 0.00012435897435897437, "loss": 0.2176, "step": 291 }, { "epoch": 0.02, "learning_rate": 0.00012478632478632478, "loss": 0.218, "step": 292 }, { "epoch": 0.02, "learning_rate": 0.00012521367521367522, "loss": 0.1882, "step": 293 }, { "epoch": 0.02, "learning_rate": 0.00012564102564102564, "loss": 0.2091, "step": 294 }, { "epoch": 0.02, "learning_rate": 0.00012606837606837608, "loss": 0.2073, "step": 295 }, { "epoch": 0.02, "learning_rate": 0.0001264957264957265, "loss": 0.1826, "step": 296 }, { "epoch": 0.02, "learning_rate": 0.00012692307692307693, "loss": 0.2067, "step": 297 }, { "epoch": 0.02, "learning_rate": 0.00012735042735042735, "loss": 0.2234, "step": 298 }, { "epoch": 0.02, "learning_rate": 0.00012777777777777776, "loss": 0.1895, "step": 299 }, { "epoch": 0.02, "learning_rate": 0.00012820512820512823, "loss": 0.1822, "step": 300 }, { "epoch": 0.02, "learning_rate": 0.00012863247863247865, "loss": 0.199, "step": 301 }, { "epoch": 0.02, "learning_rate": 0.00012905982905982906, "loss": 0.1964, "step": 302 }, { "epoch": 0.02, "learning_rate": 0.0001294871794871795, "loss": 0.1721, "step": 303 }, { "epoch": 0.02, "learning_rate": 0.00012991452991452992, "loss": 0.181, "step": 304 }, { "epoch": 0.02, "learning_rate": 0.00013034188034188036, "loss": 0.1924, "step": 305 }, { "epoch": 0.02, "learning_rate": 0.00013076923076923077, "loss": 0.2048, "step": 306 }, { "epoch": 0.02, "learning_rate": 0.00013119658119658121, "loss": 0.1698, "step": 307 }, { "epoch": 0.02, "learning_rate": 0.00013162393162393163, "loss": 0.1942, "step": 308 }, { "epoch": 0.02, "learning_rate": 0.00013205128205128204, "loss": 0.1964, "step": 309 }, { "epoch": 0.02, "learning_rate": 0.00013247863247863248, "loss": 0.1872, "step": 310 }, { "epoch": 0.02, "learning_rate": 0.0001329059829059829, "loss": 0.1955, "step": 311 }, { "epoch": 0.02, "learning_rate": 0.00013333333333333334, "loss": 0.1979, "step": 312 }, { "epoch": 0.02, "learning_rate": 0.00013376068376068375, "loss": 0.1875, "step": 313 }, { "epoch": 0.02, "learning_rate": 0.0001341880341880342, "loss": 0.1966, "step": 314 }, { "epoch": 0.02, "learning_rate": 0.00013461538461538464, "loss": 0.1911, "step": 315 }, { "epoch": 0.02, "learning_rate": 0.00013504273504273505, "loss": 0.1889, "step": 316 }, { "epoch": 0.02, "learning_rate": 0.0001354700854700855, "loss": 0.1748, "step": 317 }, { "epoch": 0.02, "learning_rate": 0.0001358974358974359, "loss": 0.1911, "step": 318 }, { "epoch": 0.02, "learning_rate": 0.00013632478632478635, "loss": 0.2121, "step": 319 }, { "epoch": 0.02, "learning_rate": 0.00013675213675213676, "loss": 0.173, "step": 320 }, { "epoch": 0.02, "learning_rate": 0.00013717948717948718, "loss": 0.182, "step": 321 }, { "epoch": 0.02, "learning_rate": 0.00013760683760683762, "loss": 0.1766, "step": 322 }, { "epoch": 0.02, "learning_rate": 0.00013803418803418803, "loss": 0.1689, "step": 323 }, { "epoch": 0.02, "learning_rate": 0.00013846153846153847, "loss": 0.171, "step": 324 }, { "epoch": 0.02, "learning_rate": 0.0001388888888888889, "loss": 0.1802, "step": 325 }, { "epoch": 0.02, "learning_rate": 0.0001393162393162393, "loss": 0.1877, "step": 326 }, { "epoch": 0.02, "learning_rate": 0.00013974358974358974, "loss": 0.1737, "step": 327 }, { "epoch": 0.02, "learning_rate": 0.00014017094017094016, "loss": 0.1937, "step": 328 }, { "epoch": 0.02, "learning_rate": 0.0001405982905982906, "loss": 0.175, "step": 329 }, { "epoch": 0.02, "learning_rate": 0.00014102564102564104, "loss": 0.1968, "step": 330 }, { "epoch": 0.02, "learning_rate": 0.00014145299145299146, "loss": 0.1847, "step": 331 }, { "epoch": 0.02, "learning_rate": 0.0001418803418803419, "loss": 0.2078, "step": 332 }, { "epoch": 0.02, "learning_rate": 0.0001423076923076923, "loss": 0.1874, "step": 333 }, { "epoch": 0.02, "learning_rate": 0.00014273504273504275, "loss": 0.1992, "step": 334 }, { "epoch": 0.02, "learning_rate": 0.00014316239316239317, "loss": 0.178, "step": 335 }, { "epoch": 0.02, "learning_rate": 0.0001435897435897436, "loss": 0.2117, "step": 336 }, { "epoch": 0.02, "learning_rate": 0.00014401709401709402, "loss": 0.1994, "step": 337 }, { "epoch": 0.02, "learning_rate": 0.00014444444444444444, "loss": 0.189, "step": 338 }, { "epoch": 0.02, "learning_rate": 0.00014487179487179488, "loss": 0.2009, "step": 339 }, { "epoch": 0.02, "learning_rate": 0.0001452991452991453, "loss": 0.2058, "step": 340 }, { "epoch": 0.02, "learning_rate": 0.00014572649572649574, "loss": 0.197, "step": 341 }, { "epoch": 0.02, "learning_rate": 0.00014615384615384615, "loss": 0.1835, "step": 342 }, { "epoch": 0.02, "learning_rate": 0.0001465811965811966, "loss": 0.2083, "step": 343 }, { "epoch": 0.02, "learning_rate": 0.00014700854700854703, "loss": 0.1942, "step": 344 }, { "epoch": 0.02, "learning_rate": 0.00014743589743589745, "loss": 0.1623, "step": 345 }, { "epoch": 0.02, "learning_rate": 0.0001478632478632479, "loss": 0.2086, "step": 346 }, { "epoch": 0.02, "learning_rate": 0.0001482905982905983, "loss": 0.1744, "step": 347 }, { "epoch": 0.02, "learning_rate": 0.00014871794871794872, "loss": 0.1945, "step": 348 }, { "epoch": 0.02, "learning_rate": 0.00014914529914529916, "loss": 0.2024, "step": 349 }, { "epoch": 0.02, "learning_rate": 0.00014957264957264957, "loss": 0.1886, "step": 350 }, { "epoch": 0.02, "learning_rate": 0.00015000000000000001, "loss": 0.178, "step": 351 }, { "epoch": 0.02, "learning_rate": 0.00015042735042735043, "loss": 0.175, "step": 352 }, { "epoch": 0.02, "learning_rate": 0.00015085470085470087, "loss": 0.1867, "step": 353 }, { "epoch": 0.02, "learning_rate": 0.00015128205128205128, "loss": 0.1844, "step": 354 }, { "epoch": 0.02, "learning_rate": 0.0001517094017094017, "loss": 0.1931, "step": 355 }, { "epoch": 0.02, "learning_rate": 0.00015213675213675214, "loss": 0.1949, "step": 356 }, { "epoch": 0.02, "learning_rate": 0.00015256410256410255, "loss": 0.2036, "step": 357 }, { "epoch": 0.02, "learning_rate": 0.000152991452991453, "loss": 0.2106, "step": 358 }, { "epoch": 0.02, "learning_rate": 0.00015341880341880344, "loss": 0.2007, "step": 359 }, { "epoch": 0.02, "learning_rate": 0.00015384615384615385, "loss": 0.1715, "step": 360 }, { "epoch": 0.02, "learning_rate": 0.0001542735042735043, "loss": 0.1793, "step": 361 }, { "epoch": 0.02, "learning_rate": 0.0001547008547008547, "loss": 0.1745, "step": 362 }, { "epoch": 0.02, "learning_rate": 0.00015512820512820515, "loss": 0.2084, "step": 363 }, { "epoch": 0.02, "learning_rate": 0.00015555555555555556, "loss": 0.1743, "step": 364 }, { "epoch": 0.02, "learning_rate": 0.000155982905982906, "loss": 0.1848, "step": 365 }, { "epoch": 0.02, "learning_rate": 0.00015641025641025642, "loss": 0.1789, "step": 366 }, { "epoch": 0.02, "learning_rate": 0.00015683760683760683, "loss": 0.1652, "step": 367 }, { "epoch": 0.02, "learning_rate": 0.00015726495726495727, "loss": 0.1676, "step": 368 }, { "epoch": 0.02, "learning_rate": 0.0001576923076923077, "loss": 0.2041, "step": 369 }, { "epoch": 0.02, "learning_rate": 0.00015811965811965813, "loss": 0.1627, "step": 370 }, { "epoch": 0.02, "learning_rate": 0.00015854700854700854, "loss": 0.1739, "step": 371 }, { "epoch": 0.02, "learning_rate": 0.00015897435897435896, "loss": 0.2074, "step": 372 }, { "epoch": 0.02, "learning_rate": 0.00015940170940170943, "loss": 0.1762, "step": 373 }, { "epoch": 0.02, "learning_rate": 0.00015982905982905984, "loss": 0.1667, "step": 374 }, { "epoch": 0.02, "learning_rate": 0.00016025641025641028, "loss": 0.1723, "step": 375 }, { "epoch": 0.02, "learning_rate": 0.0001606837606837607, "loss": 0.2029, "step": 376 }, { "epoch": 0.02, "learning_rate": 0.0001611111111111111, "loss": 0.1747, "step": 377 }, { "epoch": 0.02, "learning_rate": 0.00016153846153846155, "loss": 0.1743, "step": 378 }, { "epoch": 0.02, "learning_rate": 0.00016196581196581197, "loss": 0.1861, "step": 379 }, { "epoch": 0.02, "learning_rate": 0.0001623931623931624, "loss": 0.1643, "step": 380 }, { "epoch": 0.02, "learning_rate": 0.00016282051282051282, "loss": 0.1729, "step": 381 }, { "epoch": 0.02, "learning_rate": 0.00016324786324786327, "loss": 0.1767, "step": 382 }, { "epoch": 0.02, "learning_rate": 0.00016367521367521368, "loss": 0.1763, "step": 383 }, { "epoch": 0.02, "learning_rate": 0.0001641025641025641, "loss": 0.1972, "step": 384 }, { "epoch": 0.02, "learning_rate": 0.00016452991452991454, "loss": 0.1834, "step": 385 }, { "epoch": 0.02, "learning_rate": 0.00016495726495726495, "loss": 0.1953, "step": 386 }, { "epoch": 0.02, "learning_rate": 0.0001653846153846154, "loss": 0.1845, "step": 387 }, { "epoch": 0.02, "learning_rate": 0.00016581196581196583, "loss": 0.1701, "step": 388 }, { "epoch": 0.02, "learning_rate": 0.00016623931623931625, "loss": 0.1907, "step": 389 }, { "epoch": 0.03, "learning_rate": 0.0001666666666666667, "loss": 0.1815, "step": 390 }, { "epoch": 0.03, "learning_rate": 0.0001670940170940171, "loss": 0.1918, "step": 391 }, { "epoch": 0.03, "learning_rate": 0.00016752136752136754, "loss": 0.1893, "step": 392 }, { "epoch": 0.03, "learning_rate": 0.00016794871794871796, "loss": 0.1606, "step": 393 }, { "epoch": 0.03, "learning_rate": 0.00016837606837606837, "loss": 0.1926, "step": 394 }, { "epoch": 0.03, "learning_rate": 0.00016880341880341881, "loss": 0.2165, "step": 395 }, { "epoch": 0.03, "learning_rate": 0.00016923076923076923, "loss": 0.1726, "step": 396 }, { "epoch": 0.03, "learning_rate": 0.00016965811965811967, "loss": 0.167, "step": 397 }, { "epoch": 0.03, "learning_rate": 0.00017008547008547008, "loss": 0.1898, "step": 398 }, { "epoch": 0.03, "learning_rate": 0.00017051282051282053, "loss": 0.2034, "step": 399 }, { "epoch": 0.03, "learning_rate": 0.00017094017094017094, "loss": 0.186, "step": 400 }, { "epoch": 0.03, "learning_rate": 0.00017136752136752135, "loss": 0.1953, "step": 401 }, { "epoch": 0.03, "learning_rate": 0.0001717948717948718, "loss": 0.1989, "step": 402 }, { "epoch": 0.03, "learning_rate": 0.00017222222222222224, "loss": 0.1922, "step": 403 }, { "epoch": 0.03, "learning_rate": 0.00017264957264957268, "loss": 0.167, "step": 404 }, { "epoch": 0.03, "learning_rate": 0.0001730769230769231, "loss": 0.1625, "step": 405 }, { "epoch": 0.03, "learning_rate": 0.0001735042735042735, "loss": 0.1913, "step": 406 }, { "epoch": 0.03, "learning_rate": 0.00017393162393162395, "loss": 0.167, "step": 407 }, { "epoch": 0.03, "learning_rate": 0.00017435897435897436, "loss": 0.163, "step": 408 }, { "epoch": 0.03, "learning_rate": 0.0001747863247863248, "loss": 0.1654, "step": 409 }, { "epoch": 0.03, "learning_rate": 0.00017521367521367522, "loss": 0.1774, "step": 410 }, { "epoch": 0.03, "learning_rate": 0.00017564102564102566, "loss": 0.1668, "step": 411 }, { "epoch": 0.03, "learning_rate": 0.00017606837606837607, "loss": 0.1982, "step": 412 }, { "epoch": 0.03, "learning_rate": 0.0001764957264957265, "loss": 0.1816, "step": 413 }, { "epoch": 0.03, "learning_rate": 0.00017692307692307693, "loss": 0.1655, "step": 414 }, { "epoch": 0.03, "learning_rate": 0.00017735042735042734, "loss": 0.1961, "step": 415 }, { "epoch": 0.03, "learning_rate": 0.00017777777777777779, "loss": 0.1976, "step": 416 }, { "epoch": 0.03, "learning_rate": 0.00017820512820512823, "loss": 0.1759, "step": 417 }, { "epoch": 0.03, "learning_rate": 0.00017863247863247864, "loss": 0.1902, "step": 418 }, { "epoch": 0.03, "learning_rate": 0.00017905982905982908, "loss": 0.18, "step": 419 }, { "epoch": 0.03, "learning_rate": 0.0001794871794871795, "loss": 0.1767, "step": 420 }, { "epoch": 0.03, "learning_rate": 0.00017991452991452994, "loss": 0.204, "step": 421 }, { "epoch": 0.03, "learning_rate": 0.00018034188034188035, "loss": 0.1767, "step": 422 }, { "epoch": 0.03, "learning_rate": 0.00018076923076923077, "loss": 0.2004, "step": 423 }, { "epoch": 0.03, "learning_rate": 0.0001811965811965812, "loss": 0.1829, "step": 424 }, { "epoch": 0.03, "learning_rate": 0.00018162393162393162, "loss": 0.17, "step": 425 }, { "epoch": 0.03, "learning_rate": 0.00018205128205128207, "loss": 0.1888, "step": 426 }, { "epoch": 0.03, "learning_rate": 0.00018247863247863248, "loss": 0.202, "step": 427 }, { "epoch": 0.03, "learning_rate": 0.00018290598290598292, "loss": 0.1908, "step": 428 }, { "epoch": 0.03, "learning_rate": 0.00018333333333333334, "loss": 0.1642, "step": 429 }, { "epoch": 0.03, "learning_rate": 0.00018376068376068375, "loss": 0.176, "step": 430 }, { "epoch": 0.03, "learning_rate": 0.0001841880341880342, "loss": 0.1843, "step": 431 }, { "epoch": 0.03, "learning_rate": 0.00018461538461538463, "loss": 0.184, "step": 432 }, { "epoch": 0.03, "learning_rate": 0.00018504273504273505, "loss": 0.176, "step": 433 }, { "epoch": 0.03, "learning_rate": 0.0001854700854700855, "loss": 0.2076, "step": 434 }, { "epoch": 0.03, "learning_rate": 0.0001858974358974359, "loss": 0.1779, "step": 435 }, { "epoch": 0.03, "learning_rate": 0.00018632478632478634, "loss": 0.1662, "step": 436 }, { "epoch": 0.03, "learning_rate": 0.00018675213675213676, "loss": 0.1734, "step": 437 }, { "epoch": 0.03, "learning_rate": 0.0001871794871794872, "loss": 0.1959, "step": 438 }, { "epoch": 0.03, "learning_rate": 0.00018760683760683761, "loss": 0.1883, "step": 439 }, { "epoch": 0.03, "learning_rate": 0.00018803418803418803, "loss": 0.1792, "step": 440 }, { "epoch": 0.03, "learning_rate": 0.00018846153846153847, "loss": 0.1955, "step": 441 }, { "epoch": 0.03, "learning_rate": 0.00018888888888888888, "loss": 0.1778, "step": 442 }, { "epoch": 0.03, "learning_rate": 0.00018931623931623933, "loss": 0.1751, "step": 443 }, { "epoch": 0.03, "learning_rate": 0.00018974358974358974, "loss": 0.2048, "step": 444 }, { "epoch": 0.03, "learning_rate": 0.00019017094017094018, "loss": 0.1856, "step": 445 }, { "epoch": 0.03, "learning_rate": 0.0001905982905982906, "loss": 0.1943, "step": 446 }, { "epoch": 0.03, "learning_rate": 0.00019102564102564104, "loss": 0.1714, "step": 447 }, { "epoch": 0.03, "learning_rate": 0.00019145299145299148, "loss": 0.1701, "step": 448 }, { "epoch": 0.03, "learning_rate": 0.0001918803418803419, "loss": 0.1739, "step": 449 }, { "epoch": 0.03, "learning_rate": 0.00019230769230769233, "loss": 0.1822, "step": 450 }, { "epoch": 0.03, "learning_rate": 0.00019273504273504275, "loss": 0.1649, "step": 451 }, { "epoch": 0.03, "learning_rate": 0.00019316239316239316, "loss": 0.1709, "step": 452 }, { "epoch": 0.03, "learning_rate": 0.0001935897435897436, "loss": 0.1922, "step": 453 }, { "epoch": 0.03, "learning_rate": 0.00019401709401709402, "loss": 0.1858, "step": 454 }, { "epoch": 0.03, "learning_rate": 0.00019444444444444446, "loss": 0.1686, "step": 455 }, { "epoch": 0.03, "learning_rate": 0.00019487179487179487, "loss": 0.185, "step": 456 }, { "epoch": 0.03, "learning_rate": 0.00019529914529914532, "loss": 0.1853, "step": 457 }, { "epoch": 0.03, "learning_rate": 0.00019572649572649573, "loss": 0.1937, "step": 458 }, { "epoch": 0.03, "learning_rate": 0.00019615384615384615, "loss": 0.205, "step": 459 }, { "epoch": 0.03, "learning_rate": 0.00019658119658119659, "loss": 0.1695, "step": 460 }, { "epoch": 0.03, "learning_rate": 0.00019700854700854703, "loss": 0.1642, "step": 461 }, { "epoch": 0.03, "learning_rate": 0.00019743589743589744, "loss": 0.1655, "step": 462 }, { "epoch": 0.03, "learning_rate": 0.00019786324786324788, "loss": 0.1694, "step": 463 }, { "epoch": 0.03, "learning_rate": 0.0001982905982905983, "loss": 0.2171, "step": 464 }, { "epoch": 0.03, "learning_rate": 0.00019871794871794874, "loss": 0.1942, "step": 465 }, { "epoch": 0.03, "learning_rate": 0.00019914529914529915, "loss": 0.1969, "step": 466 }, { "epoch": 0.03, "learning_rate": 0.0001995726495726496, "loss": 0.1774, "step": 467 }, { "epoch": 0.03, "learning_rate": 0.0002, "loss": 0.1593, "step": 468 }, { "epoch": 0.03, "learning_rate": 0.00019999999784371225, "loss": 0.1653, "step": 469 }, { "epoch": 0.03, "learning_rate": 0.0001999999913748491, "loss": 0.1728, "step": 470 }, { "epoch": 0.03, "learning_rate": 0.00019999998059341077, "loss": 0.1804, "step": 471 }, { "epoch": 0.03, "learning_rate": 0.00019999996549939782, "loss": 0.1493, "step": 472 }, { "epoch": 0.03, "learning_rate": 0.0001999999460928108, "loss": 0.2014, "step": 473 }, { "epoch": 0.03, "learning_rate": 0.00019999992237365062, "loss": 0.1789, "step": 474 }, { "epoch": 0.03, "learning_rate": 0.00019999989434191828, "loss": 0.2105, "step": 475 }, { "epoch": 0.03, "learning_rate": 0.000199999861997615, "loss": 0.1795, "step": 476 }, { "epoch": 0.03, "learning_rate": 0.00019999982534074214, "loss": 0.1898, "step": 477 }, { "epoch": 0.03, "learning_rate": 0.0001999997843713013, "loss": 0.1808, "step": 478 }, { "epoch": 0.03, "learning_rate": 0.0001999997390892943, "loss": 0.1806, "step": 479 }, { "epoch": 0.03, "learning_rate": 0.00019999968949472297, "loss": 0.1741, "step": 480 }, { "epoch": 0.03, "learning_rate": 0.00019999963558758958, "loss": 0.1789, "step": 481 }, { "epoch": 0.03, "learning_rate": 0.00019999957736789638, "loss": 0.1979, "step": 482 }, { "epoch": 0.03, "learning_rate": 0.0001999995148356459, "loss": 0.172, "step": 483 }, { "epoch": 0.03, "learning_rate": 0.00019999944799084082, "loss": 0.1845, "step": 484 }, { "epoch": 0.03, "learning_rate": 0.00019999937683348403, "loss": 0.1755, "step": 485 }, { "epoch": 0.03, "learning_rate": 0.00019999930136357862, "loss": 0.1869, "step": 486 }, { "epoch": 0.03, "learning_rate": 0.00019999922158112784, "loss": 0.1782, "step": 487 }, { "epoch": 0.03, "learning_rate": 0.00019999913748613515, "loss": 0.1558, "step": 488 }, { "epoch": 0.03, "learning_rate": 0.0001999990490786041, "loss": 0.1645, "step": 489 }, { "epoch": 0.03, "learning_rate": 0.00019999895635853856, "loss": 0.1813, "step": 490 }, { "epoch": 0.03, "learning_rate": 0.00019999885932594252, "loss": 0.1813, "step": 491 }, { "epoch": 0.03, "learning_rate": 0.00019999875798082018, "loss": 0.1691, "step": 492 }, { "epoch": 0.03, "learning_rate": 0.00019999865232317587, "loss": 0.1709, "step": 493 }, { "epoch": 0.03, "learning_rate": 0.00019999854235301418, "loss": 0.1744, "step": 494 }, { "epoch": 0.03, "learning_rate": 0.00019999842807033987, "loss": 0.1726, "step": 495 }, { "epoch": 0.03, "learning_rate": 0.0001999983094751578, "loss": 0.1694, "step": 496 }, { "epoch": 0.03, "learning_rate": 0.00019999818656747315, "loss": 0.1679, "step": 497 }, { "epoch": 0.03, "learning_rate": 0.00019999805934729122, "loss": 0.1942, "step": 498 }, { "epoch": 0.03, "learning_rate": 0.00019999792781461744, "loss": 0.194, "step": 499 }, { "epoch": 0.03, "learning_rate": 0.0001999977919694575, "loss": 0.1922, "step": 500 }, { "epoch": 0.03, "learning_rate": 0.00019999765181181732, "loss": 0.2241, "step": 501 }, { "epoch": 0.03, "learning_rate": 0.00019999750734170287, "loss": 0.171, "step": 502 }, { "epoch": 0.03, "learning_rate": 0.00019999735855912043, "loss": 0.206, "step": 503 }, { "epoch": 0.03, "learning_rate": 0.00019999720546407638, "loss": 0.2047, "step": 504 }, { "epoch": 0.03, "learning_rate": 0.00019999704805657734, "loss": 0.1702, "step": 505 }, { "epoch": 0.03, "learning_rate": 0.00019999688633663012, "loss": 0.1869, "step": 506 }, { "epoch": 0.03, "learning_rate": 0.00019999672030424163, "loss": 0.1722, "step": 507 }, { "epoch": 0.03, "learning_rate": 0.0001999965499594191, "loss": 0.1595, "step": 508 }, { "epoch": 0.03, "learning_rate": 0.00019999637530216985, "loss": 0.1759, "step": 509 }, { "epoch": 0.03, "learning_rate": 0.0001999961963325014, "loss": 0.1853, "step": 510 }, { "epoch": 0.03, "learning_rate": 0.00019999601305042145, "loss": 0.2159, "step": 511 }, { "epoch": 0.03, "learning_rate": 0.000199995825455938, "loss": 0.1727, "step": 512 }, { "epoch": 0.03, "learning_rate": 0.000199995633549059, "loss": 0.1834, "step": 513 }, { "epoch": 0.03, "learning_rate": 0.00019999543732979283, "loss": 0.1717, "step": 514 }, { "epoch": 0.03, "learning_rate": 0.00019999523679814792, "loss": 0.1769, "step": 515 }, { "epoch": 0.03, "learning_rate": 0.00019999503195413293, "loss": 0.1643, "step": 516 }, { "epoch": 0.03, "learning_rate": 0.00019999482279775665, "loss": 0.1663, "step": 517 }, { "epoch": 0.03, "learning_rate": 0.00019999460932902815, "loss": 0.1963, "step": 518 }, { "epoch": 0.03, "learning_rate": 0.0001999943915479566, "loss": 0.1708, "step": 519 }, { "epoch": 0.03, "learning_rate": 0.00019999416945455146, "loss": 0.1659, "step": 520 }, { "epoch": 0.03, "learning_rate": 0.00019999394304882223, "loss": 0.1806, "step": 521 }, { "epoch": 0.03, "learning_rate": 0.00019999371233077868, "loss": 0.1683, "step": 522 }, { "epoch": 0.03, "learning_rate": 0.00019999347730043082, "loss": 0.175, "step": 523 }, { "epoch": 0.03, "learning_rate": 0.00019999323795778874, "loss": 0.173, "step": 524 }, { "epoch": 0.03, "learning_rate": 0.00019999299430286275, "loss": 0.2002, "step": 525 }, { "epoch": 0.03, "learning_rate": 0.00019999274633566337, "loss": 0.1854, "step": 526 }, { "epoch": 0.03, "learning_rate": 0.00019999249405620132, "loss": 0.1912, "step": 527 }, { "epoch": 0.03, "learning_rate": 0.00019999223746448746, "loss": 0.1828, "step": 528 }, { "epoch": 0.03, "learning_rate": 0.00019999197656053288, "loss": 0.1782, "step": 529 }, { "epoch": 0.03, "learning_rate": 0.00019999171134434877, "loss": 0.1705, "step": 530 }, { "epoch": 0.03, "learning_rate": 0.00019999144181594662, "loss": 0.1772, "step": 531 }, { "epoch": 0.03, "learning_rate": 0.00019999116797533804, "loss": 0.1626, "step": 532 }, { "epoch": 0.03, "learning_rate": 0.00019999088982253484, "loss": 0.1666, "step": 533 }, { "epoch": 0.03, "learning_rate": 0.00019999060735754903, "loss": 0.1802, "step": 534 }, { "epoch": 0.03, "learning_rate": 0.00019999032058039276, "loss": 0.2064, "step": 535 }, { "epoch": 0.03, "learning_rate": 0.0001999900294910784, "loss": 0.2069, "step": 536 }, { "epoch": 0.03, "learning_rate": 0.00019998973408961855, "loss": 0.1894, "step": 537 }, { "epoch": 0.03, "learning_rate": 0.0001999894343760259, "loss": 0.1613, "step": 538 }, { "epoch": 0.03, "learning_rate": 0.0001999891303503134, "loss": 0.1585, "step": 539 }, { "epoch": 0.03, "learning_rate": 0.00019998882201249413, "loss": 0.1658, "step": 540 }, { "epoch": 0.03, "learning_rate": 0.00019998850936258145, "loss": 0.1649, "step": 541 }, { "epoch": 0.03, "learning_rate": 0.00019998819240058877, "loss": 0.1906, "step": 542 }, { "epoch": 0.03, "learning_rate": 0.00019998787112652982, "loss": 0.1739, "step": 543 }, { "epoch": 0.03, "learning_rate": 0.00019998754554041842, "loss": 0.1855, "step": 544 }, { "epoch": 0.03, "learning_rate": 0.00019998721564226858, "loss": 0.1677, "step": 545 }, { "epoch": 0.04, "learning_rate": 0.0001999868814320946, "loss": 0.1778, "step": 546 }, { "epoch": 0.04, "learning_rate": 0.00019998654290991088, "loss": 0.1782, "step": 547 }, { "epoch": 0.04, "learning_rate": 0.00019998620007573198, "loss": 0.1631, "step": 548 }, { "epoch": 0.04, "learning_rate": 0.0001999858529295727, "loss": 0.1687, "step": 549 }, { "epoch": 0.04, "learning_rate": 0.00019998550147144802, "loss": 0.1679, "step": 550 }, { "epoch": 0.04, "learning_rate": 0.0001999851457013731, "loss": 0.1741, "step": 551 }, { "epoch": 0.04, "learning_rate": 0.00019998478561936324, "loss": 0.198, "step": 552 }, { "epoch": 0.04, "learning_rate": 0.00019998442122543404, "loss": 0.178, "step": 553 }, { "epoch": 0.04, "learning_rate": 0.00019998405251960114, "loss": 0.1642, "step": 554 }, { "epoch": 0.04, "learning_rate": 0.00019998367950188053, "loss": 0.2076, "step": 555 }, { "epoch": 0.04, "learning_rate": 0.0001999833021722882, "loss": 0.1963, "step": 556 }, { "epoch": 0.04, "learning_rate": 0.00019998292053084052, "loss": 0.1867, "step": 557 }, { "epoch": 0.04, "learning_rate": 0.00019998253457755385, "loss": 0.1862, "step": 558 }, { "epoch": 0.04, "learning_rate": 0.00019998214431244489, "loss": 0.154, "step": 559 }, { "epoch": 0.04, "learning_rate": 0.00019998174973553048, "loss": 0.1646, "step": 560 }, { "epoch": 0.04, "learning_rate": 0.0001999813508468276, "loss": 0.1675, "step": 561 }, { "epoch": 0.04, "learning_rate": 0.00019998094764635347, "loss": 0.1712, "step": 562 }, { "epoch": 0.04, "learning_rate": 0.0001999805401341255, "loss": 0.1663, "step": 563 }, { "epoch": 0.04, "learning_rate": 0.0001999801283101612, "loss": 0.2004, "step": 564 }, { "epoch": 0.04, "learning_rate": 0.0001999797121744784, "loss": 0.1676, "step": 565 }, { "epoch": 0.04, "learning_rate": 0.000199979291727095, "loss": 0.1868, "step": 566 }, { "epoch": 0.04, "learning_rate": 0.00019997886696802918, "loss": 0.2074, "step": 567 }, { "epoch": 0.04, "learning_rate": 0.00019997843789729925, "loss": 0.1923, "step": 568 }, { "epoch": 0.04, "learning_rate": 0.00019997800451492366, "loss": 0.1883, "step": 569 }, { "epoch": 0.04, "learning_rate": 0.00019997756682092112, "loss": 0.161, "step": 570 }, { "epoch": 0.04, "learning_rate": 0.00019997712481531057, "loss": 0.195, "step": 571 }, { "epoch": 0.04, "learning_rate": 0.00019997667849811098, "loss": 0.1772, "step": 572 }, { "epoch": 0.04, "learning_rate": 0.00019997622786934165, "loss": 0.1852, "step": 573 }, { "epoch": 0.04, "learning_rate": 0.000199975772929022, "loss": 0.1834, "step": 574 }, { "epoch": 0.04, "learning_rate": 0.00019997531367717163, "loss": 0.1806, "step": 575 }, { "epoch": 0.04, "learning_rate": 0.0001999748501138104, "loss": 0.2151, "step": 576 }, { "epoch": 0.04, "learning_rate": 0.0001999743822389583, "loss": 0.1767, "step": 577 }, { "epoch": 0.04, "learning_rate": 0.0001999739100526354, "loss": 0.1702, "step": 578 }, { "epoch": 0.04, "learning_rate": 0.0001999734335548622, "loss": 0.1931, "step": 579 }, { "epoch": 0.04, "learning_rate": 0.00019997295274565917, "loss": 0.1905, "step": 580 }, { "epoch": 0.04, "learning_rate": 0.00019997246762504703, "loss": 0.1777, "step": 581 }, { "epoch": 0.04, "learning_rate": 0.0001999719781930468, "loss": 0.1752, "step": 582 }, { "epoch": 0.04, "learning_rate": 0.00019997148444967947, "loss": 0.1739, "step": 583 }, { "epoch": 0.04, "learning_rate": 0.0001999709863949664, "loss": 0.1794, "step": 584 }, { "epoch": 0.04, "learning_rate": 0.00019997048402892908, "loss": 0.1781, "step": 585 }, { "epoch": 0.04, "learning_rate": 0.00019996997735158913, "loss": 0.1763, "step": 586 }, { "epoch": 0.04, "learning_rate": 0.00019996946636296843, "loss": 0.1592, "step": 587 }, { "epoch": 0.04, "learning_rate": 0.000199968951063089, "loss": 0.1675, "step": 588 }, { "epoch": 0.04, "learning_rate": 0.00019996843145197307, "loss": 0.1658, "step": 589 }, { "epoch": 0.04, "learning_rate": 0.00019996790752964305, "loss": 0.195, "step": 590 }, { "epoch": 0.04, "learning_rate": 0.00019996737929612156, "loss": 0.1863, "step": 591 }, { "epoch": 0.04, "learning_rate": 0.0001999668467514313, "loss": 0.2108, "step": 592 }, { "epoch": 0.04, "learning_rate": 0.00019996630989559534, "loss": 0.1689, "step": 593 }, { "epoch": 0.04, "learning_rate": 0.00019996576872863674, "loss": 0.1754, "step": 594 }, { "epoch": 0.04, "learning_rate": 0.0001999652232505789, "loss": 0.1869, "step": 595 }, { "epoch": 0.04, "learning_rate": 0.00019996467346144537, "loss": 0.1807, "step": 596 }, { "epoch": 0.04, "learning_rate": 0.00019996411936125977, "loss": 0.1706, "step": 597 }, { "epoch": 0.04, "learning_rate": 0.00019996356095004604, "loss": 0.18, "step": 598 }, { "epoch": 0.04, "learning_rate": 0.00019996299822782826, "loss": 0.2002, "step": 599 }, { "epoch": 0.04, "learning_rate": 0.00019996243119463074, "loss": 0.15, "step": 600 }, { "epoch": 0.04, "learning_rate": 0.00019996185985047787, "loss": 0.1793, "step": 601 }, { "epoch": 0.04, "learning_rate": 0.00019996128419539431, "loss": 0.159, "step": 602 }, { "epoch": 0.04, "learning_rate": 0.00019996070422940488, "loss": 0.1615, "step": 603 }, { "epoch": 0.04, "learning_rate": 0.00019996011995253462, "loss": 0.1854, "step": 604 }, { "epoch": 0.04, "learning_rate": 0.0001999595313648087, "loss": 0.1711, "step": 605 }, { "epoch": 0.04, "learning_rate": 0.0001999589384662525, "loss": 0.1663, "step": 606 }, { "epoch": 0.04, "learning_rate": 0.0001999583412568916, "loss": 0.1737, "step": 607 }, { "epoch": 0.04, "learning_rate": 0.0001999577397367518, "loss": 0.1882, "step": 608 }, { "epoch": 0.04, "learning_rate": 0.00019995713390585896, "loss": 0.1715, "step": 609 }, { "epoch": 0.04, "learning_rate": 0.00019995652376423925, "loss": 0.154, "step": 610 }, { "epoch": 0.04, "learning_rate": 0.00019995590931191898, "loss": 0.1469, "step": 611 }, { "epoch": 0.04, "learning_rate": 0.00019995529054892463, "loss": 0.1748, "step": 612 }, { "epoch": 0.04, "learning_rate": 0.00019995466747528293, "loss": 0.1312, "step": 613 }, { "epoch": 0.04, "learning_rate": 0.0001999540400910207, "loss": 0.1858, "step": 614 }, { "epoch": 0.04, "learning_rate": 0.00019995340839616503, "loss": 0.1674, "step": 615 }, { "epoch": 0.04, "learning_rate": 0.00019995277239074315, "loss": 0.1778, "step": 616 }, { "epoch": 0.04, "learning_rate": 0.00019995213207478247, "loss": 0.166, "step": 617 }, { "epoch": 0.04, "learning_rate": 0.00019995148744831063, "loss": 0.1585, "step": 618 }, { "epoch": 0.04, "learning_rate": 0.00019995083851135544, "loss": 0.1845, "step": 619 }, { "epoch": 0.04, "learning_rate": 0.00019995018526394485, "loss": 0.1901, "step": 620 }, { "epoch": 0.04, "learning_rate": 0.00019994952770610708, "loss": 0.1946, "step": 621 }, { "epoch": 0.04, "learning_rate": 0.0001999488658378704, "loss": 0.163, "step": 622 }, { "epoch": 0.04, "learning_rate": 0.00019994819965926346, "loss": 0.1466, "step": 623 }, { "epoch": 0.04, "learning_rate": 0.00019994752917031492, "loss": 0.1325, "step": 624 }, { "epoch": 0.04, "learning_rate": 0.00019994685437105372, "loss": 0.1891, "step": 625 }, { "epoch": 0.04, "learning_rate": 0.00019994617526150897, "loss": 0.2057, "step": 626 }, { "epoch": 0.04, "learning_rate": 0.0001999454918417099, "loss": 0.158, "step": 627 }, { "epoch": 0.04, "learning_rate": 0.00019994480411168609, "loss": 0.2165, "step": 628 }, { "epoch": 0.04, "learning_rate": 0.0001999441120714671, "loss": 0.2055, "step": 629 }, { "epoch": 0.04, "learning_rate": 0.00019994341572108278, "loss": 0.1743, "step": 630 }, { "epoch": 0.04, "learning_rate": 0.00019994271506056324, "loss": 0.2047, "step": 631 }, { "epoch": 0.04, "learning_rate": 0.00019994201008993864, "loss": 0.176, "step": 632 }, { "epoch": 0.04, "learning_rate": 0.00019994130080923936, "loss": 0.2094, "step": 633 }, { "epoch": 0.04, "learning_rate": 0.00019994058721849606, "loss": 0.2009, "step": 634 }, { "epoch": 0.04, "learning_rate": 0.00019993986931773945, "loss": 0.1674, "step": 635 }, { "epoch": 0.04, "learning_rate": 0.00019993914710700052, "loss": 0.1911, "step": 636 }, { "epoch": 0.04, "learning_rate": 0.00019993842058631042, "loss": 0.1658, "step": 637 }, { "epoch": 0.04, "learning_rate": 0.00019993768975570044, "loss": 0.1618, "step": 638 }, { "epoch": 0.04, "learning_rate": 0.00019993695461520216, "loss": 0.1767, "step": 639 }, { "epoch": 0.04, "learning_rate": 0.0001999362151648472, "loss": 0.1939, "step": 640 }, { "epoch": 0.04, "learning_rate": 0.00019993547140466754, "loss": 0.1693, "step": 641 }, { "epoch": 0.04, "learning_rate": 0.00019993472333469523, "loss": 0.1884, "step": 642 }, { "epoch": 0.04, "learning_rate": 0.00019993397095496252, "loss": 0.2159, "step": 643 }, { "epoch": 0.04, "learning_rate": 0.0001999332142655018, "loss": 0.1986, "step": 644 }, { "epoch": 0.04, "learning_rate": 0.0001999324532663458, "loss": 0.1663, "step": 645 }, { "epoch": 0.04, "learning_rate": 0.0001999316879575273, "loss": 0.1648, "step": 646 }, { "epoch": 0.04, "learning_rate": 0.00019993091833907928, "loss": 0.1852, "step": 647 }, { "epoch": 0.04, "learning_rate": 0.00019993014441103496, "loss": 0.1794, "step": 648 }, { "epoch": 0.04, "learning_rate": 0.0001999293661734277, "loss": 0.1701, "step": 649 }, { "epoch": 0.04, "learning_rate": 0.0001999285836262911, "loss": 0.1667, "step": 650 }, { "epoch": 0.04, "learning_rate": 0.00019992779676965885, "loss": 0.1712, "step": 651 }, { "epoch": 0.04, "learning_rate": 0.0001999270056035649, "loss": 0.1899, "step": 652 }, { "epoch": 0.04, "learning_rate": 0.00019992621012804337, "loss": 0.1634, "step": 653 }, { "epoch": 0.04, "learning_rate": 0.0001999254103431286, "loss": 0.1502, "step": 654 }, { "epoch": 0.04, "learning_rate": 0.00019992460624885508, "loss": 0.1865, "step": 655 }, { "epoch": 0.04, "learning_rate": 0.0001999237978452574, "loss": 0.1639, "step": 656 }, { "epoch": 0.04, "learning_rate": 0.00019992298513237056, "loss": 0.1624, "step": 657 }, { "epoch": 0.04, "learning_rate": 0.00019992216811022951, "loss": 0.16, "step": 658 }, { "epoch": 0.04, "learning_rate": 0.00019992134677886947, "loss": 0.1583, "step": 659 }, { "epoch": 0.04, "learning_rate": 0.00019992052113832593, "loss": 0.1677, "step": 660 }, { "epoch": 0.04, "learning_rate": 0.00019991969118863447, "loss": 0.1529, "step": 661 }, { "epoch": 0.04, "learning_rate": 0.0001999188569298309, "loss": 0.1702, "step": 662 }, { "epoch": 0.04, "learning_rate": 0.00019991801836195117, "loss": 0.1906, "step": 663 }, { "epoch": 0.04, "learning_rate": 0.00019991717548503143, "loss": 0.1636, "step": 664 }, { "epoch": 0.04, "learning_rate": 0.00019991632829910807, "loss": 0.1707, "step": 665 }, { "epoch": 0.04, "learning_rate": 0.00019991547680421758, "loss": 0.1584, "step": 666 }, { "epoch": 0.04, "learning_rate": 0.00019991462100039673, "loss": 0.178, "step": 667 }, { "epoch": 0.04, "learning_rate": 0.0001999137608876824, "loss": 0.1574, "step": 668 }, { "epoch": 0.04, "learning_rate": 0.0001999128964661117, "loss": 0.1897, "step": 669 }, { "epoch": 0.04, "learning_rate": 0.00019991202773572187, "loss": 0.1738, "step": 670 }, { "epoch": 0.04, "learning_rate": 0.00019991115469655043, "loss": 0.1706, "step": 671 }, { "epoch": 0.04, "learning_rate": 0.000199910277348635, "loss": 0.1647, "step": 672 }, { "epoch": 0.04, "learning_rate": 0.0001999093956920134, "loss": 0.1737, "step": 673 }, { "epoch": 0.04, "learning_rate": 0.0001999085097267237, "loss": 0.1811, "step": 674 }, { "epoch": 0.04, "learning_rate": 0.00019990761945280406, "loss": 0.1592, "step": 675 }, { "epoch": 0.04, "learning_rate": 0.00019990672487029288, "loss": 0.1592, "step": 676 }, { "epoch": 0.04, "learning_rate": 0.00019990582597922876, "loss": 0.1874, "step": 677 }, { "epoch": 0.04, "learning_rate": 0.00019990492277965048, "loss": 0.1597, "step": 678 }, { "epoch": 0.04, "learning_rate": 0.0001999040152715969, "loss": 0.1609, "step": 679 }, { "epoch": 0.04, "learning_rate": 0.00019990310345510733, "loss": 0.1622, "step": 680 }, { "epoch": 0.04, "learning_rate": 0.00019990218733022093, "loss": 0.1888, "step": 681 }, { "epoch": 0.04, "learning_rate": 0.00019990126689697725, "loss": 0.1702, "step": 682 }, { "epoch": 0.04, "learning_rate": 0.000199900342155416, "loss": 0.1713, "step": 683 }, { "epoch": 0.04, "learning_rate": 0.0001998994131055771, "loss": 0.1682, "step": 684 }, { "epoch": 0.04, "learning_rate": 0.00019989847974750057, "loss": 0.1845, "step": 685 }, { "epoch": 0.04, "learning_rate": 0.00019989754208122665, "loss": 0.1854, "step": 686 }, { "epoch": 0.04, "learning_rate": 0.00019989660010679582, "loss": 0.2032, "step": 687 }, { "epoch": 0.04, "learning_rate": 0.00019989565382424869, "loss": 0.1641, "step": 688 }, { "epoch": 0.04, "learning_rate": 0.00019989470323362603, "loss": 0.1909, "step": 689 }, { "epoch": 0.04, "learning_rate": 0.00019989374833496887, "loss": 0.1956, "step": 690 }, { "epoch": 0.04, "learning_rate": 0.0001998927891283184, "loss": 0.1789, "step": 691 }, { "epoch": 0.04, "learning_rate": 0.00019989182561371593, "loss": 0.1485, "step": 692 }, { "epoch": 0.04, "learning_rate": 0.0001998908577912031, "loss": 0.1666, "step": 693 }, { "epoch": 0.04, "learning_rate": 0.00019988988566082156, "loss": 0.1643, "step": 694 }, { "epoch": 0.04, "learning_rate": 0.00019988890922261333, "loss": 0.1576, "step": 695 }, { "epoch": 0.04, "learning_rate": 0.00019988792847662042, "loss": 0.1322, "step": 696 }, { "epoch": 0.04, "learning_rate": 0.0001998869434228852, "loss": 0.1779, "step": 697 }, { "epoch": 0.04, "learning_rate": 0.00019988595406145008, "loss": 0.1745, "step": 698 }, { "epoch": 0.04, "learning_rate": 0.00019988496039235782, "loss": 0.1787, "step": 699 }, { "epoch": 0.04, "learning_rate": 0.00019988396241565117, "loss": 0.1775, "step": 700 }, { "epoch": 0.04, "learning_rate": 0.00019988296013137327, "loss": 0.1772, "step": 701 }, { "epoch": 0.05, "learning_rate": 0.00019988195353956727, "loss": 0.1713, "step": 702 }, { "epoch": 0.05, "learning_rate": 0.00019988094264027662, "loss": 0.1895, "step": 703 }, { "epoch": 0.05, "learning_rate": 0.00019987992743354487, "loss": 0.1542, "step": 704 }, { "epoch": 0.05, "learning_rate": 0.00019987890791941582, "loss": 0.1568, "step": 705 }, { "epoch": 0.05, "learning_rate": 0.0001998778840979335, "loss": 0.1626, "step": 706 }, { "epoch": 0.05, "learning_rate": 0.000199876855969142, "loss": 0.1712, "step": 707 }, { "epoch": 0.05, "learning_rate": 0.00019987582353308566, "loss": 0.1845, "step": 708 }, { "epoch": 0.05, "learning_rate": 0.000199874786789809, "loss": 0.1627, "step": 709 }, { "epoch": 0.05, "learning_rate": 0.00019987374573935678, "loss": 0.1739, "step": 710 }, { "epoch": 0.05, "learning_rate": 0.00019987270038177384, "loss": 0.1633, "step": 711 }, { "epoch": 0.05, "learning_rate": 0.00019987165071710527, "loss": 0.1515, "step": 712 }, { "epoch": 0.05, "learning_rate": 0.0001998705967453964, "loss": 0.192, "step": 713 }, { "epoch": 0.05, "learning_rate": 0.0001998695384666926, "loss": 0.2018, "step": 714 }, { "epoch": 0.05, "learning_rate": 0.00019986847588103954, "loss": 0.1886, "step": 715 }, { "epoch": 0.05, "learning_rate": 0.00019986740898848306, "loss": 0.1625, "step": 716 }, { "epoch": 0.05, "learning_rate": 0.00019986633778906916, "loss": 0.1597, "step": 717 }, { "epoch": 0.05, "learning_rate": 0.00019986526228284404, "loss": 0.2159, "step": 718 }, { "epoch": 0.05, "learning_rate": 0.00019986418246985407, "loss": 0.1657, "step": 719 }, { "epoch": 0.05, "learning_rate": 0.0001998630983501458, "loss": 0.1596, "step": 720 }, { "epoch": 0.05, "learning_rate": 0.00019986200992376605, "loss": 0.1906, "step": 721 }, { "epoch": 0.05, "learning_rate": 0.0001998609171907617, "loss": 0.1691, "step": 722 }, { "epoch": 0.05, "learning_rate": 0.00019985982015117987, "loss": 0.169, "step": 723 }, { "epoch": 0.05, "learning_rate": 0.00019985871880506793, "loss": 0.1786, "step": 724 }, { "epoch": 0.05, "learning_rate": 0.00019985761315247333, "loss": 0.1767, "step": 725 }, { "epoch": 0.05, "learning_rate": 0.00019985650319344375, "loss": 0.2059, "step": 726 }, { "epoch": 0.05, "learning_rate": 0.00019985538892802708, "loss": 0.2064, "step": 727 }, { "epoch": 0.05, "learning_rate": 0.00019985427035627137, "loss": 0.1978, "step": 728 }, { "epoch": 0.05, "learning_rate": 0.00019985314747822486, "loss": 0.1565, "step": 729 }, { "epoch": 0.05, "learning_rate": 0.00019985202029393592, "loss": 0.1542, "step": 730 }, { "epoch": 0.05, "learning_rate": 0.00019985088880345326, "loss": 0.159, "step": 731 }, { "epoch": 0.05, "learning_rate": 0.0001998497530068256, "loss": 0.1889, "step": 732 }, { "epoch": 0.05, "learning_rate": 0.00019984861290410196, "loss": 0.1811, "step": 733 }, { "epoch": 0.05, "learning_rate": 0.0001998474684953315, "loss": 0.1562, "step": 734 }, { "epoch": 0.05, "learning_rate": 0.00019984631978056354, "loss": 0.1702, "step": 735 }, { "epoch": 0.05, "learning_rate": 0.0001998451667598477, "loss": 0.1686, "step": 736 }, { "epoch": 0.05, "learning_rate": 0.0001998440094332336, "loss": 0.1689, "step": 737 }, { "epoch": 0.05, "learning_rate": 0.00019984284780077117, "loss": 0.1593, "step": 738 }, { "epoch": 0.05, "learning_rate": 0.00019984168186251063, "loss": 0.2214, "step": 739 }, { "epoch": 0.05, "learning_rate": 0.0001998405116185021, "loss": 0.1748, "step": 740 }, { "epoch": 0.05, "learning_rate": 0.00019983933706879613, "loss": 0.1885, "step": 741 }, { "epoch": 0.05, "learning_rate": 0.00019983815821344338, "loss": 0.1652, "step": 742 }, { "epoch": 0.05, "learning_rate": 0.00019983697505249467, "loss": 0.1714, "step": 743 }, { "epoch": 0.05, "learning_rate": 0.00019983578758600104, "loss": 0.1832, "step": 744 }, { "epoch": 0.05, "learning_rate": 0.00019983459581401366, "loss": 0.1821, "step": 745 }, { "epoch": 0.05, "learning_rate": 0.00019983339973658396, "loss": 0.161, "step": 746 }, { "epoch": 0.05, "learning_rate": 0.00019983219935376353, "loss": 0.204, "step": 747 }, { "epoch": 0.05, "learning_rate": 0.0001998309946656041, "loss": 0.1915, "step": 748 }, { "epoch": 0.05, "learning_rate": 0.00019982978567215765, "loss": 0.1742, "step": 749 }, { "epoch": 0.05, "learning_rate": 0.00019982857237347632, "loss": 0.1699, "step": 750 }, { "epoch": 0.05, "learning_rate": 0.00019982735476961242, "loss": 0.1869, "step": 751 }, { "epoch": 0.05, "learning_rate": 0.00019982613286061846, "loss": 0.1937, "step": 752 }, { "epoch": 0.05, "learning_rate": 0.00019982490664654715, "loss": 0.1665, "step": 753 }, { "epoch": 0.05, "learning_rate": 0.00019982367612745138, "loss": 0.193, "step": 754 }, { "epoch": 0.05, "learning_rate": 0.0001998224413033842, "loss": 0.1842, "step": 755 }, { "epoch": 0.05, "learning_rate": 0.00019982120217439884, "loss": 0.1566, "step": 756 }, { "epoch": 0.05, "learning_rate": 0.00019981995874054876, "loss": 0.1746, "step": 757 }, { "epoch": 0.05, "learning_rate": 0.0001998187110018876, "loss": 0.1588, "step": 758 }, { "epoch": 0.05, "learning_rate": 0.00019981745895846919, "loss": 0.1757, "step": 759 }, { "epoch": 0.05, "learning_rate": 0.00019981620261034743, "loss": 0.183, "step": 760 }, { "epoch": 0.05, "learning_rate": 0.00019981494195757662, "loss": 0.1708, "step": 761 }, { "epoch": 0.05, "learning_rate": 0.00019981367700021104, "loss": 0.1523, "step": 762 }, { "epoch": 0.05, "learning_rate": 0.00019981240773830528, "loss": 0.1808, "step": 763 }, { "epoch": 0.05, "learning_rate": 0.00019981113417191407, "loss": 0.1986, "step": 764 }, { "epoch": 0.05, "learning_rate": 0.0001998098563010923, "loss": 0.1394, "step": 765 }, { "epoch": 0.05, "learning_rate": 0.00019980857412589515, "loss": 0.1693, "step": 766 }, { "epoch": 0.05, "learning_rate": 0.00019980728764637786, "loss": 0.1586, "step": 767 }, { "epoch": 0.05, "learning_rate": 0.00019980599686259594, "loss": 0.1792, "step": 768 }, { "epoch": 0.05, "learning_rate": 0.00019980470177460508, "loss": 0.1651, "step": 769 }, { "epoch": 0.05, "learning_rate": 0.00019980340238246103, "loss": 0.1876, "step": 770 }, { "epoch": 0.05, "learning_rate": 0.0001998020986862199, "loss": 0.1485, "step": 771 }, { "epoch": 0.05, "learning_rate": 0.00019980079068593793, "loss": 0.188, "step": 772 }, { "epoch": 0.05, "learning_rate": 0.0001997994783816715, "loss": 0.1802, "step": 773 }, { "epoch": 0.05, "learning_rate": 0.0001997981617734772, "loss": 0.1899, "step": 774 }, { "epoch": 0.05, "learning_rate": 0.00019979684086141183, "loss": 0.1594, "step": 775 }, { "epoch": 0.05, "learning_rate": 0.00019979551564553233, "loss": 0.1625, "step": 776 }, { "epoch": 0.05, "learning_rate": 0.00019979418612589586, "loss": 0.1586, "step": 777 }, { "epoch": 0.05, "learning_rate": 0.0001997928523025598, "loss": 0.1694, "step": 778 }, { "epoch": 0.05, "learning_rate": 0.00019979151417558159, "loss": 0.1512, "step": 779 }, { "epoch": 0.05, "learning_rate": 0.00019979017174501896, "loss": 0.1762, "step": 780 }, { "epoch": 0.05, "learning_rate": 0.00019978882501092989, "loss": 0.1835, "step": 781 }, { "epoch": 0.05, "learning_rate": 0.00019978747397337234, "loss": 0.1587, "step": 782 }, { "epoch": 0.05, "learning_rate": 0.00019978611863240463, "loss": 0.1793, "step": 783 }, { "epoch": 0.05, "learning_rate": 0.00019978475898808524, "loss": 0.1589, "step": 784 }, { "epoch": 0.05, "learning_rate": 0.00019978339504047275, "loss": 0.1994, "step": 785 }, { "epoch": 0.05, "learning_rate": 0.000199782026789626, "loss": 0.1636, "step": 786 }, { "epoch": 0.05, "learning_rate": 0.00019978065423560402, "loss": 0.1546, "step": 787 }, { "epoch": 0.05, "learning_rate": 0.00019977927737846595, "loss": 0.1705, "step": 788 }, { "epoch": 0.05, "learning_rate": 0.00019977789621827125, "loss": 0.1823, "step": 789 }, { "epoch": 0.05, "learning_rate": 0.0001997765107550794, "loss": 0.1636, "step": 790 }, { "epoch": 0.05, "learning_rate": 0.00019977512098895019, "loss": 0.1529, "step": 791 }, { "epoch": 0.05, "learning_rate": 0.00019977372691994355, "loss": 0.1498, "step": 792 }, { "epoch": 0.05, "learning_rate": 0.00019977232854811962, "loss": 0.1847, "step": 793 }, { "epoch": 0.05, "learning_rate": 0.00019977092587353864, "loss": 0.1525, "step": 794 }, { "epoch": 0.05, "learning_rate": 0.0001997695188962612, "loss": 0.1666, "step": 795 }, { "epoch": 0.05, "learning_rate": 0.00019976810761634787, "loss": 0.1899, "step": 796 }, { "epoch": 0.05, "learning_rate": 0.0001997666920338596, "loss": 0.1506, "step": 797 }, { "epoch": 0.05, "learning_rate": 0.00019976527214885737, "loss": 0.1562, "step": 798 }, { "epoch": 0.05, "learning_rate": 0.00019976384796140248, "loss": 0.166, "step": 799 }, { "epoch": 0.05, "learning_rate": 0.00019976241947155634, "loss": 0.1689, "step": 800 }, { "epoch": 0.05, "learning_rate": 0.0001997609866793805, "loss": 0.1699, "step": 801 }, { "epoch": 0.05, "learning_rate": 0.00019975954958493676, "loss": 0.1636, "step": 802 }, { "epoch": 0.05, "learning_rate": 0.00019975810818828715, "loss": 0.1626, "step": 803 }, { "epoch": 0.05, "learning_rate": 0.00019975666248949378, "loss": 0.1601, "step": 804 }, { "epoch": 0.05, "learning_rate": 0.00019975521248861904, "loss": 0.1916, "step": 805 }, { "epoch": 0.05, "learning_rate": 0.0001997537581857254, "loss": 0.1533, "step": 806 }, { "epoch": 0.05, "learning_rate": 0.00019975229958087566, "loss": 0.1948, "step": 807 }, { "epoch": 0.05, "learning_rate": 0.00019975083667413265, "loss": 0.1729, "step": 808 }, { "epoch": 0.05, "learning_rate": 0.00019974936946555948, "loss": 0.1537, "step": 809 }, { "epoch": 0.05, "learning_rate": 0.00019974789795521947, "loss": 0.1485, "step": 810 }, { "epoch": 0.05, "learning_rate": 0.00019974642214317601, "loss": 0.1599, "step": 811 }, { "epoch": 0.05, "learning_rate": 0.0001997449420294928, "loss": 0.1848, "step": 812 }, { "epoch": 0.05, "learning_rate": 0.00019974345761423362, "loss": 0.1739, "step": 813 }, { "epoch": 0.05, "learning_rate": 0.00019974196889746256, "loss": 0.1664, "step": 814 }, { "epoch": 0.05, "learning_rate": 0.00019974047587924373, "loss": 0.1583, "step": 815 }, { "epoch": 0.05, "learning_rate": 0.0001997389785596416, "loss": 0.1612, "step": 816 }, { "epoch": 0.05, "learning_rate": 0.00019973747693872068, "loss": 0.1886, "step": 817 }, { "epoch": 0.05, "learning_rate": 0.0001997359710165458, "loss": 0.1694, "step": 818 }, { "epoch": 0.05, "learning_rate": 0.00019973446079318185, "loss": 0.1519, "step": 819 }, { "epoch": 0.05, "learning_rate": 0.00019973294626869393, "loss": 0.1636, "step": 820 }, { "epoch": 0.05, "learning_rate": 0.00019973142744314742, "loss": 0.1729, "step": 821 }, { "epoch": 0.05, "learning_rate": 0.0001997299043166078, "loss": 0.1661, "step": 822 }, { "epoch": 0.05, "learning_rate": 0.00019972837688914076, "loss": 0.1653, "step": 823 }, { "epoch": 0.05, "learning_rate": 0.00019972684516081214, "loss": 0.1738, "step": 824 }, { "epoch": 0.05, "learning_rate": 0.000199725309131688, "loss": 0.1562, "step": 825 }, { "epoch": 0.05, "learning_rate": 0.00019972376880183465, "loss": 0.1668, "step": 826 }, { "epoch": 0.05, "learning_rate": 0.00019972222417131844, "loss": 0.1643, "step": 827 }, { "epoch": 0.05, "learning_rate": 0.00019972067524020605, "loss": 0.173, "step": 828 }, { "epoch": 0.05, "learning_rate": 0.00019971912200856423, "loss": 0.1894, "step": 829 }, { "epoch": 0.05, "learning_rate": 0.00019971756447645996, "loss": 0.1519, "step": 830 }, { "epoch": 0.05, "learning_rate": 0.00019971600264396045, "loss": 0.174, "step": 831 }, { "epoch": 0.05, "learning_rate": 0.000199714436511133, "loss": 0.1774, "step": 832 }, { "epoch": 0.05, "learning_rate": 0.0001997128660780452, "loss": 0.1752, "step": 833 }, { "epoch": 0.05, "learning_rate": 0.00019971129134476473, "loss": 0.1739, "step": 834 }, { "epoch": 0.05, "learning_rate": 0.0001997097123113596, "loss": 0.1602, "step": 835 }, { "epoch": 0.05, "learning_rate": 0.0001997081289778978, "loss": 0.1737, "step": 836 }, { "epoch": 0.05, "learning_rate": 0.0001997065413444476, "loss": 0.1703, "step": 837 }, { "epoch": 0.05, "learning_rate": 0.0001997049494110776, "loss": 0.152, "step": 838 }, { "epoch": 0.05, "learning_rate": 0.00019970335317785634, "loss": 0.1797, "step": 839 }, { "epoch": 0.05, "learning_rate": 0.00019970175264485266, "loss": 0.1571, "step": 840 }, { "epoch": 0.05, "learning_rate": 0.00019970014781213563, "loss": 0.1555, "step": 841 }, { "epoch": 0.05, "learning_rate": 0.00019969853867977448, "loss": 0.1832, "step": 842 }, { "epoch": 0.05, "learning_rate": 0.00019969692524783854, "loss": 0.182, "step": 843 }, { "epoch": 0.05, "learning_rate": 0.00019969530751639742, "loss": 0.1623, "step": 844 }, { "epoch": 0.05, "learning_rate": 0.00019969368548552089, "loss": 0.1979, "step": 845 }, { "epoch": 0.05, "learning_rate": 0.00019969205915527889, "loss": 0.1915, "step": 846 }, { "epoch": 0.05, "learning_rate": 0.0001996904285257416, "loss": 0.1557, "step": 847 }, { "epoch": 0.05, "learning_rate": 0.00019968879359697926, "loss": 0.161, "step": 848 }, { "epoch": 0.05, "learning_rate": 0.00019968715436906241, "loss": 0.1648, "step": 849 }, { "epoch": 0.05, "learning_rate": 0.0001996855108420618, "loss": 0.1481, "step": 850 }, { "epoch": 0.05, "learning_rate": 0.00019968386301604825, "loss": 0.1628, "step": 851 }, { "epoch": 0.05, "learning_rate": 0.00019968221089109283, "loss": 0.18, "step": 852 }, { "epoch": 0.05, "learning_rate": 0.00019968055446726676, "loss": 0.177, "step": 853 }, { "epoch": 0.05, "learning_rate": 0.00019967889374464156, "loss": 0.1736, "step": 854 }, { "epoch": 0.05, "learning_rate": 0.0001996772287232888, "loss": 0.1637, "step": 855 }, { "epoch": 0.05, "learning_rate": 0.00019967555940328028, "loss": 0.1809, "step": 856 }, { "epoch": 0.05, "learning_rate": 0.000199673885784688, "loss": 0.1804, "step": 857 }, { "epoch": 0.06, "learning_rate": 0.00019967220786758412, "loss": 0.1847, "step": 858 }, { "epoch": 0.06, "learning_rate": 0.00019967052565204104, "loss": 0.1657, "step": 859 }, { "epoch": 0.06, "learning_rate": 0.00019966883913813125, "loss": 0.1667, "step": 860 }, { "epoch": 0.06, "learning_rate": 0.00019966714832592755, "loss": 0.1723, "step": 861 }, { "epoch": 0.06, "learning_rate": 0.00019966545321550277, "loss": 0.1698, "step": 862 }, { "epoch": 0.06, "learning_rate": 0.0001996637538069301, "loss": 0.1756, "step": 863 }, { "epoch": 0.06, "learning_rate": 0.0001996620501002828, "loss": 0.1787, "step": 864 }, { "epoch": 0.06, "learning_rate": 0.00019966034209563431, "loss": 0.1698, "step": 865 }, { "epoch": 0.06, "learning_rate": 0.00019965862979305835, "loss": 0.1692, "step": 866 }, { "epoch": 0.06, "learning_rate": 0.00019965691319262868, "loss": 0.1514, "step": 867 }, { "epoch": 0.06, "learning_rate": 0.00019965519229441943, "loss": 0.1799, "step": 868 }, { "epoch": 0.06, "learning_rate": 0.00019965346709850475, "loss": 0.1793, "step": 869 }, { "epoch": 0.06, "learning_rate": 0.00019965173760495906, "loss": 0.1551, "step": 870 }, { "epoch": 0.06, "learning_rate": 0.0001996500038138569, "loss": 0.1732, "step": 871 }, { "epoch": 0.06, "learning_rate": 0.00019964826572527314, "loss": 0.1784, "step": 872 }, { "epoch": 0.06, "learning_rate": 0.00019964652333928263, "loss": 0.1756, "step": 873 }, { "epoch": 0.06, "learning_rate": 0.0001996447766559606, "loss": 0.1628, "step": 874 }, { "epoch": 0.06, "learning_rate": 0.00019964302567538228, "loss": 0.1732, "step": 875 }, { "epoch": 0.06, "learning_rate": 0.0001996412703976233, "loss": 0.1848, "step": 876 }, { "epoch": 0.06, "learning_rate": 0.00019963951082275925, "loss": 0.1589, "step": 877 }, { "epoch": 0.06, "learning_rate": 0.00019963774695086608, "loss": 0.1619, "step": 878 }, { "epoch": 0.06, "learning_rate": 0.00019963597878201985, "loss": 0.1645, "step": 879 }, { "epoch": 0.06, "learning_rate": 0.00019963420631629678, "loss": 0.1606, "step": 880 }, { "epoch": 0.06, "learning_rate": 0.00019963242955377335, "loss": 0.1552, "step": 881 }, { "epoch": 0.06, "learning_rate": 0.00019963064849452616, "loss": 0.1895, "step": 882 }, { "epoch": 0.06, "learning_rate": 0.000199628863138632, "loss": 0.1721, "step": 883 }, { "epoch": 0.06, "learning_rate": 0.0001996270734861679, "loss": 0.1575, "step": 884 }, { "epoch": 0.06, "learning_rate": 0.00019962527953721105, "loss": 0.1935, "step": 885 }, { "epoch": 0.06, "learning_rate": 0.00019962348129183878, "loss": 0.139, "step": 886 }, { "epoch": 0.06, "learning_rate": 0.00019962167875012864, "loss": 0.1972, "step": 887 }, { "epoch": 0.06, "learning_rate": 0.00019961987191215841, "loss": 0.1655, "step": 888 }, { "epoch": 0.06, "learning_rate": 0.00019961806077800596, "loss": 0.2231, "step": 889 }, { "epoch": 0.06, "learning_rate": 0.00019961624534774942, "loss": 0.1689, "step": 890 }, { "epoch": 0.06, "learning_rate": 0.0001996144256214671, "loss": 0.1751, "step": 891 }, { "epoch": 0.06, "learning_rate": 0.00019961260159923744, "loss": 0.1682, "step": 892 }, { "epoch": 0.06, "learning_rate": 0.00019961077328113915, "loss": 0.1571, "step": 893 }, { "epoch": 0.06, "learning_rate": 0.000199608940667251, "loss": 0.1523, "step": 894 }, { "epoch": 0.06, "learning_rate": 0.0001996071037576521, "loss": 0.1801, "step": 895 }, { "epoch": 0.06, "learning_rate": 0.00019960526255242164, "loss": 0.1528, "step": 896 }, { "epoch": 0.06, "learning_rate": 0.00019960341705163904, "loss": 0.1736, "step": 897 }, { "epoch": 0.06, "learning_rate": 0.00019960156725538385, "loss": 0.1531, "step": 898 }, { "epoch": 0.06, "learning_rate": 0.00019959971316373584, "loss": 0.1901, "step": 899 }, { "epoch": 0.06, "learning_rate": 0.00019959785477677503, "loss": 0.158, "step": 900 }, { "epoch": 0.06, "learning_rate": 0.00019959599209458148, "loss": 0.1694, "step": 901 }, { "epoch": 0.06, "learning_rate": 0.0001995941251172356, "loss": 0.2038, "step": 902 }, { "epoch": 0.06, "learning_rate": 0.0001995922538448179, "loss": 0.1479, "step": 903 }, { "epoch": 0.06, "learning_rate": 0.00019959037827740902, "loss": 0.1645, "step": 904 }, { "epoch": 0.06, "learning_rate": 0.00019958849841508988, "loss": 0.1831, "step": 905 }, { "epoch": 0.06, "learning_rate": 0.00019958661425794157, "loss": 0.1743, "step": 906 }, { "epoch": 0.06, "learning_rate": 0.0001995847258060453, "loss": 0.139, "step": 907 }, { "epoch": 0.06, "learning_rate": 0.00019958283305948252, "loss": 0.214, "step": 908 }, { "epoch": 0.06, "learning_rate": 0.00019958093601833493, "loss": 0.172, "step": 909 }, { "epoch": 0.06, "learning_rate": 0.00019957903468268424, "loss": 0.1731, "step": 910 }, { "epoch": 0.06, "learning_rate": 0.0001995771290526125, "loss": 0.1709, "step": 911 }, { "epoch": 0.06, "learning_rate": 0.00019957521912820187, "loss": 0.2, "step": 912 }, { "epoch": 0.06, "learning_rate": 0.00019957330490953472, "loss": 0.1776, "step": 913 }, { "epoch": 0.06, "learning_rate": 0.00019957138639669362, "loss": 0.1728, "step": 914 }, { "epoch": 0.06, "learning_rate": 0.0001995694635897613, "loss": 0.1607, "step": 915 }, { "epoch": 0.06, "learning_rate": 0.00019956753648882068, "loss": 0.1712, "step": 916 }, { "epoch": 0.06, "learning_rate": 0.00019956560509395486, "loss": 0.1751, "step": 917 }, { "epoch": 0.06, "learning_rate": 0.00019956366940524712, "loss": 0.1618, "step": 918 }, { "epoch": 0.06, "learning_rate": 0.000199561729422781, "loss": 0.1704, "step": 919 }, { "epoch": 0.06, "learning_rate": 0.00019955978514664012, "loss": 0.1621, "step": 920 }, { "epoch": 0.06, "learning_rate": 0.0001995578365769083, "loss": 0.1516, "step": 921 }, { "epoch": 0.06, "learning_rate": 0.00019955588371366965, "loss": 0.1526, "step": 922 }, { "epoch": 0.06, "learning_rate": 0.00019955392655700827, "loss": 0.1898, "step": 923 }, { "epoch": 0.06, "learning_rate": 0.0001995519651070087, "loss": 0.1637, "step": 924 }, { "epoch": 0.06, "learning_rate": 0.00019954999936375542, "loss": 0.1569, "step": 925 }, { "epoch": 0.06, "learning_rate": 0.0001995480293273333, "loss": 0.1722, "step": 926 }, { "epoch": 0.06, "learning_rate": 0.0001995460549978272, "loss": 0.1599, "step": 927 }, { "epoch": 0.06, "learning_rate": 0.00019954407637532234, "loss": 0.1592, "step": 928 }, { "epoch": 0.06, "learning_rate": 0.000199542093459904, "loss": 0.1836, "step": 929 }, { "epoch": 0.06, "learning_rate": 0.00019954010625165773, "loss": 0.1749, "step": 930 }, { "epoch": 0.06, "learning_rate": 0.00019953811475066923, "loss": 0.1551, "step": 931 }, { "epoch": 0.06, "learning_rate": 0.00019953611895702435, "loss": 0.1774, "step": 932 }, { "epoch": 0.06, "learning_rate": 0.00019953411887080918, "loss": 0.177, "step": 933 }, { "epoch": 0.06, "learning_rate": 0.00019953211449210998, "loss": 0.186, "step": 934 }, { "epoch": 0.06, "learning_rate": 0.0001995301058210132, "loss": 0.152, "step": 935 }, { "epoch": 0.06, "learning_rate": 0.00019952809285760545, "loss": 0.1633, "step": 936 }, { "epoch": 0.06, "learning_rate": 0.00019952607560197355, "loss": 0.1713, "step": 937 }, { "epoch": 0.06, "learning_rate": 0.00019952405405420447, "loss": 0.168, "step": 938 }, { "epoch": 0.06, "learning_rate": 0.0001995220282143854, "loss": 0.1563, "step": 939 }, { "epoch": 0.06, "learning_rate": 0.00019951999808260376, "loss": 0.1662, "step": 940 }, { "epoch": 0.06, "learning_rate": 0.00019951796365894704, "loss": 0.161, "step": 941 }, { "epoch": 0.06, "learning_rate": 0.00019951592494350296, "loss": 0.165, "step": 942 }, { "epoch": 0.06, "learning_rate": 0.0001995138819363595, "loss": 0.1634, "step": 943 }, { "epoch": 0.06, "learning_rate": 0.00019951183463760475, "loss": 0.1644, "step": 944 }, { "epoch": 0.06, "learning_rate": 0.00019950978304732696, "loss": 0.1632, "step": 945 }, { "epoch": 0.06, "learning_rate": 0.00019950772716561467, "loss": 0.1744, "step": 946 }, { "epoch": 0.06, "learning_rate": 0.0001995056669925565, "loss": 0.1734, "step": 947 }, { "epoch": 0.06, "learning_rate": 0.0001995036025282413, "loss": 0.1729, "step": 948 }, { "epoch": 0.06, "learning_rate": 0.0001995015337727581, "loss": 0.1699, "step": 949 }, { "epoch": 0.06, "learning_rate": 0.00019949946072619614, "loss": 0.1938, "step": 950 }, { "epoch": 0.06, "learning_rate": 0.00019949738338864484, "loss": 0.1606, "step": 951 }, { "epoch": 0.06, "learning_rate": 0.00019949530176019368, "loss": 0.1712, "step": 952 }, { "epoch": 0.06, "learning_rate": 0.00019949321584093255, "loss": 0.1664, "step": 953 }, { "epoch": 0.06, "learning_rate": 0.00019949112563095133, "loss": 0.154, "step": 954 }, { "epoch": 0.06, "learning_rate": 0.00019948903113034024, "loss": 0.1684, "step": 955 }, { "epoch": 0.06, "learning_rate": 0.00019948693233918952, "loss": 0.1861, "step": 956 }, { "epoch": 0.06, "learning_rate": 0.00019948482925758972, "loss": 0.1773, "step": 957 }, { "epoch": 0.06, "learning_rate": 0.00019948272188563158, "loss": 0.1627, "step": 958 }, { "epoch": 0.06, "learning_rate": 0.0001994806102234059, "loss": 0.1652, "step": 959 }, { "epoch": 0.06, "learning_rate": 0.00019947849427100382, "loss": 0.1749, "step": 960 }, { "epoch": 0.06, "learning_rate": 0.00019947637402851655, "loss": 0.1655, "step": 961 }, { "epoch": 0.06, "learning_rate": 0.0001994742494960355, "loss": 0.1495, "step": 962 }, { "epoch": 0.06, "learning_rate": 0.00019947212067365238, "loss": 0.1483, "step": 963 }, { "epoch": 0.06, "learning_rate": 0.0001994699875614589, "loss": 0.1715, "step": 964 }, { "epoch": 0.06, "learning_rate": 0.00019946785015954714, "loss": 0.1697, "step": 965 }, { "epoch": 0.06, "learning_rate": 0.0001994657084680092, "loss": 0.1602, "step": 966 }, { "epoch": 0.06, "learning_rate": 0.00019946356248693752, "loss": 0.1453, "step": 967 }, { "epoch": 0.06, "learning_rate": 0.00019946141221642458, "loss": 0.1754, "step": 968 }, { "epoch": 0.06, "learning_rate": 0.00019945925765656313, "loss": 0.1653, "step": 969 }, { "epoch": 0.06, "learning_rate": 0.0001994570988074461, "loss": 0.1643, "step": 970 }, { "epoch": 0.06, "learning_rate": 0.00019945493566916658, "loss": 0.1733, "step": 971 }, { "epoch": 0.06, "learning_rate": 0.00019945276824181787, "loss": 0.1375, "step": 972 }, { "epoch": 0.06, "learning_rate": 0.00019945059652549344, "loss": 0.1621, "step": 973 }, { "epoch": 0.06, "learning_rate": 0.00019944842052028693, "loss": 0.1614, "step": 974 }, { "epoch": 0.06, "learning_rate": 0.00019944624022629222, "loss": 0.1508, "step": 975 }, { "epoch": 0.06, "learning_rate": 0.00019944405564360326, "loss": 0.1467, "step": 976 }, { "epoch": 0.06, "learning_rate": 0.00019944186677231437, "loss": 0.1472, "step": 977 }, { "epoch": 0.06, "learning_rate": 0.00019943967361251983, "loss": 0.167, "step": 978 }, { "epoch": 0.06, "learning_rate": 0.00019943747616431429, "loss": 0.1705, "step": 979 }, { "epoch": 0.06, "learning_rate": 0.00019943527442779256, "loss": 0.1704, "step": 980 }, { "epoch": 0.06, "learning_rate": 0.0001994330684030495, "loss": 0.1785, "step": 981 }, { "epoch": 0.06, "learning_rate": 0.0001994308580901803, "loss": 0.1703, "step": 982 }, { "epoch": 0.06, "learning_rate": 0.00019942864348928023, "loss": 0.1756, "step": 983 }, { "epoch": 0.06, "learning_rate": 0.00019942642460044486, "loss": 0.1731, "step": 984 }, { "epoch": 0.06, "learning_rate": 0.00019942420142376986, "loss": 0.1782, "step": 985 }, { "epoch": 0.06, "learning_rate": 0.00019942197395935108, "loss": 0.1735, "step": 986 }, { "epoch": 0.06, "learning_rate": 0.00019941974220728464, "loss": 0.1655, "step": 987 }, { "epoch": 0.06, "learning_rate": 0.0001994175061676667, "loss": 0.1762, "step": 988 }, { "epoch": 0.06, "learning_rate": 0.00019941526584059375, "loss": 0.1578, "step": 989 }, { "epoch": 0.06, "learning_rate": 0.0001994130212261624, "loss": 0.1554, "step": 990 }, { "epoch": 0.06, "learning_rate": 0.00019941077232446945, "loss": 0.1695, "step": 991 }, { "epoch": 0.06, "learning_rate": 0.00019940851913561187, "loss": 0.1712, "step": 992 }, { "epoch": 0.06, "learning_rate": 0.00019940626165968685, "loss": 0.1819, "step": 993 }, { "epoch": 0.06, "learning_rate": 0.00019940399989679174, "loss": 0.1919, "step": 994 }, { "epoch": 0.06, "learning_rate": 0.00019940173384702407, "loss": 0.1598, "step": 995 }, { "epoch": 0.06, "learning_rate": 0.00019939946351048158, "loss": 0.1648, "step": 996 }, { "epoch": 0.06, "learning_rate": 0.00019939718888726214, "loss": 0.1585, "step": 997 }, { "epoch": 0.06, "learning_rate": 0.00019939490997746391, "loss": 0.1754, "step": 998 }, { "epoch": 0.06, "learning_rate": 0.00019939262678118512, "loss": 0.1684, "step": 999 }, { "epoch": 0.06, "learning_rate": 0.00019939033929852425, "loss": 0.183, "step": 1000 }, { "epoch": 0.06, "learning_rate": 0.00019938804752957996, "loss": 0.1461, "step": 1001 }, { "epoch": 0.06, "learning_rate": 0.00019938575147445106, "loss": 0.1954, "step": 1002 }, { "epoch": 0.06, "learning_rate": 0.0001993834511332366, "loss": 0.1555, "step": 1003 }, { "epoch": 0.06, "learning_rate": 0.00019938114650603573, "loss": 0.1788, "step": 1004 }, { "epoch": 0.06, "learning_rate": 0.0001993788375929479, "loss": 0.1563, "step": 1005 }, { "epoch": 0.06, "learning_rate": 0.00019937652439407268, "loss": 0.1655, "step": 1006 }, { "epoch": 0.06, "learning_rate": 0.00019937420690950978, "loss": 0.1902, "step": 1007 }, { "epoch": 0.06, "learning_rate": 0.00019937188513935921, "loss": 0.1461, "step": 1008 }, { "epoch": 0.06, "learning_rate": 0.00019936955908372102, "loss": 0.1977, "step": 1009 }, { "epoch": 0.06, "learning_rate": 0.00019936722874269557, "loss": 0.1764, "step": 1010 }, { "epoch": 0.06, "learning_rate": 0.00019936489411638336, "loss": 0.1557, "step": 1011 }, { "epoch": 0.06, "learning_rate": 0.00019936255520488504, "loss": 0.1671, "step": 1012 }, { "epoch": 0.06, "learning_rate": 0.00019936021200830153, "loss": 0.1808, "step": 1013 }, { "epoch": 0.07, "learning_rate": 0.0001993578645267338, "loss": 0.1443, "step": 1014 }, { "epoch": 0.07, "learning_rate": 0.00019935551276028323, "loss": 0.1618, "step": 1015 }, { "epoch": 0.07, "learning_rate": 0.00019935315670905105, "loss": 0.148, "step": 1016 }, { "epoch": 0.07, "learning_rate": 0.00019935079637313906, "loss": 0.2013, "step": 1017 }, { "epoch": 0.07, "learning_rate": 0.00019934843175264887, "loss": 0.1529, "step": 1018 }, { "epoch": 0.07, "learning_rate": 0.0001993460628476826, "loss": 0.1717, "step": 1019 }, { "epoch": 0.07, "learning_rate": 0.00019934368965834236, "loss": 0.1431, "step": 1020 }, { "epoch": 0.07, "learning_rate": 0.00019934131218473047, "loss": 0.1548, "step": 1021 }, { "epoch": 0.07, "learning_rate": 0.00019933893042694946, "loss": 0.1811, "step": 1022 }, { "epoch": 0.07, "learning_rate": 0.00019933654438510212, "loss": 0.1614, "step": 1023 }, { "epoch": 0.07, "learning_rate": 0.00019933415405929124, "loss": 0.1703, "step": 1024 }, { "epoch": 0.07, "learning_rate": 0.00019933175944962, "loss": 0.1676, "step": 1025 }, { "epoch": 0.07, "learning_rate": 0.00019932936055619164, "loss": 0.174, "step": 1026 }, { "epoch": 0.07, "learning_rate": 0.00019932695737910955, "loss": 0.1749, "step": 1027 }, { "epoch": 0.07, "learning_rate": 0.00019932454991847745, "loss": 0.1745, "step": 1028 }, { "epoch": 0.07, "learning_rate": 0.00019932213817439913, "loss": 0.1862, "step": 1029 }, { "epoch": 0.07, "learning_rate": 0.00019931972214697863, "loss": 0.1611, "step": 1030 }, { "epoch": 0.07, "learning_rate": 0.0001993173018363201, "loss": 0.1769, "step": 1031 }, { "epoch": 0.07, "learning_rate": 0.00019931487724252796, "loss": 0.1823, "step": 1032 }, { "epoch": 0.07, "learning_rate": 0.0001993124483657067, "loss": 0.1612, "step": 1033 }, { "epoch": 0.07, "learning_rate": 0.00019931001520596115, "loss": 0.1764, "step": 1034 }, { "epoch": 0.07, "learning_rate": 0.0001993075777633962, "loss": 0.1481, "step": 1035 }, { "epoch": 0.07, "learning_rate": 0.00019930513603811699, "loss": 0.1553, "step": 1036 }, { "epoch": 0.07, "learning_rate": 0.00019930269003022875, "loss": 0.1414, "step": 1037 }, { "epoch": 0.07, "learning_rate": 0.00019930023973983708, "loss": 0.1816, "step": 1038 }, { "epoch": 0.07, "learning_rate": 0.00019929778516704756, "loss": 0.1896, "step": 1039 }, { "epoch": 0.07, "learning_rate": 0.00019929532631196608, "loss": 0.191, "step": 1040 }, { "epoch": 0.07, "learning_rate": 0.00019929286317469868, "loss": 0.159, "step": 1041 }, { "epoch": 0.07, "learning_rate": 0.00019929039575535158, "loss": 0.169, "step": 1042 }, { "epoch": 0.07, "learning_rate": 0.00019928792405403121, "loss": 0.1576, "step": 1043 }, { "epoch": 0.07, "learning_rate": 0.0001992854480708441, "loss": 0.1816, "step": 1044 }, { "epoch": 0.07, "learning_rate": 0.00019928296780589712, "loss": 0.1562, "step": 1045 }, { "epoch": 0.07, "learning_rate": 0.00019928048325929717, "loss": 0.1692, "step": 1046 }, { "epoch": 0.07, "learning_rate": 0.0001992779944311514, "loss": 0.1653, "step": 1047 }, { "epoch": 0.07, "learning_rate": 0.00019927550132156716, "loss": 0.1833, "step": 1048 }, { "epoch": 0.07, "learning_rate": 0.00019927300393065197, "loss": 0.1616, "step": 1049 }, { "epoch": 0.07, "learning_rate": 0.00019927050225851357, "loss": 0.1511, "step": 1050 }, { "epoch": 0.07, "learning_rate": 0.00019926799630525972, "loss": 0.1651, "step": 1051 }, { "epoch": 0.07, "learning_rate": 0.00019926548607099865, "loss": 0.144, "step": 1052 }, { "epoch": 0.07, "learning_rate": 0.00019926297155583848, "loss": 0.1745, "step": 1053 }, { "epoch": 0.07, "learning_rate": 0.00019926045275988773, "loss": 0.1666, "step": 1054 }, { "epoch": 0.07, "learning_rate": 0.00019925792968325504, "loss": 0.1629, "step": 1055 }, { "epoch": 0.07, "learning_rate": 0.00019925540232604915, "loss": 0.1699, "step": 1056 }, { "epoch": 0.07, "learning_rate": 0.00019925287068837908, "loss": 0.17, "step": 1057 }, { "epoch": 0.07, "learning_rate": 0.000199250334770354, "loss": 0.1702, "step": 1058 }, { "epoch": 0.07, "learning_rate": 0.00019924779457208334, "loss": 0.1529, "step": 1059 }, { "epoch": 0.07, "learning_rate": 0.00019924525009367658, "loss": 0.1735, "step": 1060 }, { "epoch": 0.07, "learning_rate": 0.0001992427013352435, "loss": 0.1785, "step": 1061 }, { "epoch": 0.07, "learning_rate": 0.0001992401482968939, "loss": 0.1797, "step": 1062 }, { "epoch": 0.07, "learning_rate": 0.00019923759097873806, "loss": 0.1669, "step": 1063 }, { "epoch": 0.07, "learning_rate": 0.00019923502938088615, "loss": 0.1701, "step": 1064 }, { "epoch": 0.07, "learning_rate": 0.00019923246350344863, "loss": 0.1494, "step": 1065 }, { "epoch": 0.07, "learning_rate": 0.00019922989334653624, "loss": 0.1675, "step": 1066 }, { "epoch": 0.07, "learning_rate": 0.00019922731891025975, "loss": 0.1633, "step": 1067 }, { "epoch": 0.07, "learning_rate": 0.00019922474019473023, "loss": 0.1803, "step": 1068 }, { "epoch": 0.07, "learning_rate": 0.00019922215720005886, "loss": 0.127, "step": 1069 }, { "epoch": 0.07, "learning_rate": 0.00019921956992635705, "loss": 0.1775, "step": 1070 }, { "epoch": 0.07, "learning_rate": 0.00019921697837373633, "loss": 0.1794, "step": 1071 }, { "epoch": 0.07, "learning_rate": 0.0001992143825423085, "loss": 0.1742, "step": 1072 }, { "epoch": 0.07, "learning_rate": 0.00019921178243218556, "loss": 0.1931, "step": 1073 }, { "epoch": 0.07, "learning_rate": 0.00019920917804347953, "loss": 0.1797, "step": 1074 }, { "epoch": 0.07, "learning_rate": 0.0001992065693763028, "loss": 0.1711, "step": 1075 }, { "epoch": 0.07, "learning_rate": 0.0001992039564307679, "loss": 0.1924, "step": 1076 }, { "epoch": 0.07, "learning_rate": 0.00019920133920698743, "loss": 0.1715, "step": 1077 }, { "epoch": 0.07, "learning_rate": 0.0001991987177050743, "loss": 0.1595, "step": 1078 }, { "epoch": 0.07, "learning_rate": 0.00019919609192514158, "loss": 0.1673, "step": 1079 }, { "epoch": 0.07, "learning_rate": 0.0001991934618673025, "loss": 0.1771, "step": 1080 }, { "epoch": 0.07, "learning_rate": 0.00019919082753167046, "loss": 0.1681, "step": 1081 }, { "epoch": 0.07, "learning_rate": 0.0001991881889183591, "loss": 0.1692, "step": 1082 }, { "epoch": 0.07, "learning_rate": 0.0001991855460274822, "loss": 0.1778, "step": 1083 }, { "epoch": 0.07, "learning_rate": 0.0001991828988591537, "loss": 0.1666, "step": 1084 }, { "epoch": 0.07, "learning_rate": 0.00019918024741348784, "loss": 0.1526, "step": 1085 }, { "epoch": 0.07, "learning_rate": 0.0001991775916905989, "loss": 0.1789, "step": 1086 }, { "epoch": 0.07, "learning_rate": 0.00019917493169060145, "loss": 0.1728, "step": 1087 }, { "epoch": 0.07, "learning_rate": 0.00019917226741361015, "loss": 0.1667, "step": 1088 }, { "epoch": 0.07, "learning_rate": 0.00019916959885973997, "loss": 0.1541, "step": 1089 }, { "epoch": 0.07, "learning_rate": 0.00019916692602910593, "loss": 0.1836, "step": 1090 }, { "epoch": 0.07, "learning_rate": 0.00019916424892182336, "loss": 0.1845, "step": 1091 }, { "epoch": 0.07, "learning_rate": 0.00019916156753800764, "loss": 0.1665, "step": 1092 }, { "epoch": 0.07, "learning_rate": 0.00019915888187777445, "loss": 0.1505, "step": 1093 }, { "epoch": 0.07, "learning_rate": 0.00019915619194123962, "loss": 0.1683, "step": 1094 }, { "epoch": 0.07, "learning_rate": 0.00019915349772851913, "loss": 0.1741, "step": 1095 }, { "epoch": 0.07, "learning_rate": 0.00019915079923972918, "loss": 0.1654, "step": 1096 }, { "epoch": 0.07, "learning_rate": 0.0001991480964749862, "loss": 0.1568, "step": 1097 }, { "epoch": 0.07, "learning_rate": 0.00019914538943440664, "loss": 0.1661, "step": 1098 }, { "epoch": 0.07, "learning_rate": 0.0001991426781181073, "loss": 0.1825, "step": 1099 }, { "epoch": 0.07, "learning_rate": 0.0001991399625262051, "loss": 0.1437, "step": 1100 }, { "epoch": 0.07, "learning_rate": 0.00019913724265881717, "loss": 0.1483, "step": 1101 }, { "epoch": 0.07, "learning_rate": 0.00019913451851606077, "loss": 0.1657, "step": 1102 }, { "epoch": 0.07, "learning_rate": 0.00019913179009805346, "loss": 0.1592, "step": 1103 }, { "epoch": 0.07, "learning_rate": 0.0001991290574049128, "loss": 0.1963, "step": 1104 }, { "epoch": 0.07, "learning_rate": 0.0001991263204367567, "loss": 0.1581, "step": 1105 }, { "epoch": 0.07, "learning_rate": 0.0001991235791937032, "loss": 0.1684, "step": 1106 }, { "epoch": 0.07, "learning_rate": 0.00019912083367587048, "loss": 0.1597, "step": 1107 }, { "epoch": 0.07, "learning_rate": 0.00019911808388337697, "loss": 0.1756, "step": 1108 }, { "epoch": 0.07, "learning_rate": 0.00019911532981634124, "loss": 0.1737, "step": 1109 }, { "epoch": 0.07, "learning_rate": 0.0001991125714748821, "loss": 0.1681, "step": 1110 }, { "epoch": 0.07, "learning_rate": 0.00019910980885911846, "loss": 0.179, "step": 1111 }, { "epoch": 0.07, "learning_rate": 0.00019910704196916948, "loss": 0.1681, "step": 1112 }, { "epoch": 0.07, "learning_rate": 0.00019910427080515447, "loss": 0.1414, "step": 1113 }, { "epoch": 0.07, "learning_rate": 0.00019910149536719296, "loss": 0.1748, "step": 1114 }, { "epoch": 0.07, "learning_rate": 0.00019909871565540463, "loss": 0.1338, "step": 1115 }, { "epoch": 0.07, "learning_rate": 0.00019909593166990934, "loss": 0.162, "step": 1116 }, { "epoch": 0.07, "learning_rate": 0.0001990931434108272, "loss": 0.1478, "step": 1117 }, { "epoch": 0.07, "learning_rate": 0.00019909035087827844, "loss": 0.1828, "step": 1118 }, { "epoch": 0.07, "learning_rate": 0.00019908755407238343, "loss": 0.1625, "step": 1119 }, { "epoch": 0.07, "learning_rate": 0.00019908475299326286, "loss": 0.1758, "step": 1120 }, { "epoch": 0.07, "learning_rate": 0.00019908194764103749, "loss": 0.172, "step": 1121 }, { "epoch": 0.07, "learning_rate": 0.00019907913801582832, "loss": 0.1591, "step": 1122 }, { "epoch": 0.07, "learning_rate": 0.0001990763241177565, "loss": 0.2033, "step": 1123 }, { "epoch": 0.07, "learning_rate": 0.0001990735059469434, "loss": 0.1715, "step": 1124 }, { "epoch": 0.07, "learning_rate": 0.00019907068350351055, "loss": 0.1467, "step": 1125 }, { "epoch": 0.07, "learning_rate": 0.00019906785678757967, "loss": 0.192, "step": 1126 }, { "epoch": 0.07, "learning_rate": 0.00019906502579927264, "loss": 0.1335, "step": 1127 }, { "epoch": 0.07, "learning_rate": 0.0001990621905387116, "loss": 0.1807, "step": 1128 }, { "epoch": 0.07, "learning_rate": 0.00019905935100601876, "loss": 0.1797, "step": 1129 }, { "epoch": 0.07, "learning_rate": 0.00019905650720131664, "loss": 0.1517, "step": 1130 }, { "epoch": 0.07, "learning_rate": 0.00019905365912472784, "loss": 0.1653, "step": 1131 }, { "epoch": 0.07, "learning_rate": 0.00019905080677637518, "loss": 0.171, "step": 1132 }, { "epoch": 0.07, "learning_rate": 0.0001990479501563817, "loss": 0.1758, "step": 1133 }, { "epoch": 0.07, "learning_rate": 0.0001990450892648706, "loss": 0.1563, "step": 1134 }, { "epoch": 0.07, "learning_rate": 0.0001990422241019652, "loss": 0.1903, "step": 1135 }, { "epoch": 0.07, "learning_rate": 0.00019903935466778915, "loss": 0.1778, "step": 1136 }, { "epoch": 0.07, "learning_rate": 0.00019903648096246613, "loss": 0.1719, "step": 1137 }, { "epoch": 0.07, "learning_rate": 0.00019903360298612004, "loss": 0.1633, "step": 1138 }, { "epoch": 0.07, "learning_rate": 0.00019903072073887507, "loss": 0.1937, "step": 1139 }, { "epoch": 0.07, "learning_rate": 0.0001990278342208555, "loss": 0.1478, "step": 1140 }, { "epoch": 0.07, "learning_rate": 0.00019902494343218582, "loss": 0.1708, "step": 1141 }, { "epoch": 0.07, "learning_rate": 0.00019902204837299068, "loss": 0.1471, "step": 1142 }, { "epoch": 0.07, "learning_rate": 0.00019901914904339488, "loss": 0.1726, "step": 1143 }, { "epoch": 0.07, "learning_rate": 0.00019901624544352357, "loss": 0.1686, "step": 1144 }, { "epoch": 0.07, "learning_rate": 0.0001990133375735019, "loss": 0.1727, "step": 1145 }, { "epoch": 0.07, "learning_rate": 0.00019901042543345526, "loss": 0.192, "step": 1146 }, { "epoch": 0.07, "learning_rate": 0.00019900750902350928, "loss": 0.1601, "step": 1147 }, { "epoch": 0.07, "learning_rate": 0.00019900458834378972, "loss": 0.1659, "step": 1148 }, { "epoch": 0.07, "learning_rate": 0.0001990016633944225, "loss": 0.182, "step": 1149 }, { "epoch": 0.07, "learning_rate": 0.00019899873417553382, "loss": 0.1629, "step": 1150 }, { "epoch": 0.07, "learning_rate": 0.00019899580068724997, "loss": 0.1768, "step": 1151 }, { "epoch": 0.07, "learning_rate": 0.00019899286292969747, "loss": 0.1509, "step": 1152 }, { "epoch": 0.07, "learning_rate": 0.00019898992090300303, "loss": 0.1629, "step": 1153 }, { "epoch": 0.07, "learning_rate": 0.00019898697460729348, "loss": 0.1399, "step": 1154 }, { "epoch": 0.07, "learning_rate": 0.0001989840240426959, "loss": 0.149, "step": 1155 }, { "epoch": 0.07, "learning_rate": 0.00019898106920933755, "loss": 0.1819, "step": 1156 }, { "epoch": 0.07, "learning_rate": 0.00019897811010734584, "loss": 0.1447, "step": 1157 }, { "epoch": 0.07, "learning_rate": 0.00019897514673684842, "loss": 0.1958, "step": 1158 }, { "epoch": 0.07, "learning_rate": 0.000198972179097973, "loss": 0.1342, "step": 1159 }, { "epoch": 0.07, "learning_rate": 0.0001989692071908477, "loss": 0.1472, "step": 1160 }, { "epoch": 0.07, "learning_rate": 0.00019896623101560058, "loss": 0.1823, "step": 1161 }, { "epoch": 0.07, "learning_rate": 0.00019896325057236, "loss": 0.151, "step": 1162 }, { "epoch": 0.07, "learning_rate": 0.00019896026586125457, "loss": 0.204, "step": 1163 }, { "epoch": 0.07, "learning_rate": 0.00019895727688241287, "loss": 0.1862, "step": 1164 }, { "epoch": 0.07, "learning_rate": 0.00019895428363596397, "loss": 0.1877, "step": 1165 }, { "epoch": 0.07, "learning_rate": 0.0001989512861220368, "loss": 0.1886, "step": 1166 }, { "epoch": 0.07, "learning_rate": 0.00019894828434076071, "loss": 0.1813, "step": 1167 }, { "epoch": 0.07, "learning_rate": 0.00019894527829226517, "loss": 0.1739, "step": 1168 }, { "epoch": 0.07, "learning_rate": 0.0001989422679766798, "loss": 0.1778, "step": 1169 }, { "epoch": 0.08, "learning_rate": 0.0001989392533941344, "loss": 0.1633, "step": 1170 }, { "epoch": 0.08, "learning_rate": 0.00019893623454475897, "loss": 0.1791, "step": 1171 }, { "epoch": 0.08, "learning_rate": 0.00019893321142868377, "loss": 0.1452, "step": 1172 }, { "epoch": 0.08, "learning_rate": 0.00019893018404603906, "loss": 0.1698, "step": 1173 }, { "epoch": 0.08, "learning_rate": 0.0001989271523969555, "loss": 0.1841, "step": 1174 }, { "epoch": 0.08, "learning_rate": 0.0001989241164815638, "loss": 0.1766, "step": 1175 }, { "epoch": 0.08, "learning_rate": 0.00019892107629999485, "loss": 0.1676, "step": 1176 }, { "epoch": 0.08, "learning_rate": 0.00019891803185237983, "loss": 0.1693, "step": 1177 }, { "epoch": 0.08, "learning_rate": 0.00019891498313884997, "loss": 0.1801, "step": 1178 }, { "epoch": 0.08, "learning_rate": 0.00019891193015953676, "loss": 0.1967, "step": 1179 }, { "epoch": 0.08, "learning_rate": 0.0001989088729145719, "loss": 0.1561, "step": 1180 }, { "epoch": 0.08, "learning_rate": 0.00019890581140408718, "loss": 0.1556, "step": 1181 }, { "epoch": 0.08, "learning_rate": 0.0001989027456282147, "loss": 0.1677, "step": 1182 }, { "epoch": 0.08, "learning_rate": 0.00019889967558708662, "loss": 0.1428, "step": 1183 }, { "epoch": 0.08, "learning_rate": 0.00019889660128083537, "loss": 0.1674, "step": 1184 }, { "epoch": 0.08, "learning_rate": 0.0001988935227095935, "loss": 0.1773, "step": 1185 }, { "epoch": 0.08, "learning_rate": 0.0001988904398734938, "loss": 0.159, "step": 1186 }, { "epoch": 0.08, "learning_rate": 0.00019888735277266917, "loss": 0.1826, "step": 1187 }, { "epoch": 0.08, "learning_rate": 0.00019888426140725283, "loss": 0.1774, "step": 1188 }, { "epoch": 0.08, "learning_rate": 0.000198881165777378, "loss": 0.1743, "step": 1189 }, { "epoch": 0.08, "learning_rate": 0.00019887806588317827, "loss": 0.1679, "step": 1190 }, { "epoch": 0.08, "learning_rate": 0.0001988749617247873, "loss": 0.1738, "step": 1191 }, { "epoch": 0.08, "learning_rate": 0.00019887185330233893, "loss": 0.1664, "step": 1192 }, { "epoch": 0.08, "learning_rate": 0.00019886874061596723, "loss": 0.1599, "step": 1193 }, { "epoch": 0.08, "learning_rate": 0.00019886562366580645, "loss": 0.1915, "step": 1194 }, { "epoch": 0.08, "learning_rate": 0.00019886250245199096, "loss": 0.1545, "step": 1195 }, { "epoch": 0.08, "learning_rate": 0.00019885937697465545, "loss": 0.1462, "step": 1196 }, { "epoch": 0.08, "learning_rate": 0.00019885624723393463, "loss": 0.1822, "step": 1197 }, { "epoch": 0.08, "learning_rate": 0.0001988531132299635, "loss": 0.1815, "step": 1198 }, { "epoch": 0.08, "learning_rate": 0.00019884997496287722, "loss": 0.1525, "step": 1199 }, { "epoch": 0.08, "learning_rate": 0.00019884683243281116, "loss": 0.1596, "step": 1200 }, { "epoch": 0.08, "learning_rate": 0.0001988436856399008, "loss": 0.1533, "step": 1201 }, { "epoch": 0.08, "learning_rate": 0.00019884053458428183, "loss": 0.1638, "step": 1202 }, { "epoch": 0.08, "learning_rate": 0.0001988373792660902, "loss": 0.1843, "step": 1203 }, { "epoch": 0.08, "learning_rate": 0.00019883421968546196, "loss": 0.1785, "step": 1204 }, { "epoch": 0.08, "learning_rate": 0.00019883105584253338, "loss": 0.1365, "step": 1205 }, { "epoch": 0.08, "learning_rate": 0.00019882788773744085, "loss": 0.1955, "step": 1206 }, { "epoch": 0.08, "learning_rate": 0.0001988247153703211, "loss": 0.1527, "step": 1207 }, { "epoch": 0.08, "learning_rate": 0.00019882153874131083, "loss": 0.1675, "step": 1208 }, { "epoch": 0.08, "learning_rate": 0.0001988183578505471, "loss": 0.1566, "step": 1209 }, { "epoch": 0.08, "learning_rate": 0.00019881517269816705, "loss": 0.1439, "step": 1210 }, { "epoch": 0.08, "learning_rate": 0.00019881198328430808, "loss": 0.1789, "step": 1211 }, { "epoch": 0.08, "learning_rate": 0.00019880878960910772, "loss": 0.1811, "step": 1212 }, { "epoch": 0.08, "learning_rate": 0.00019880559167270373, "loss": 0.1618, "step": 1213 }, { "epoch": 0.08, "learning_rate": 0.00019880238947523395, "loss": 0.1638, "step": 1214 }, { "epoch": 0.08, "learning_rate": 0.00019879918301683654, "loss": 0.159, "step": 1215 }, { "epoch": 0.08, "learning_rate": 0.00019879597229764974, "loss": 0.1398, "step": 1216 }, { "epoch": 0.08, "learning_rate": 0.00019879275731781205, "loss": 0.1517, "step": 1217 }, { "epoch": 0.08, "learning_rate": 0.0001987895380774621, "loss": 0.1683, "step": 1218 }, { "epoch": 0.08, "learning_rate": 0.0001987863145767387, "loss": 0.1904, "step": 1219 }, { "epoch": 0.08, "learning_rate": 0.00019878308681578095, "loss": 0.1551, "step": 1220 }, { "epoch": 0.08, "learning_rate": 0.00019877985479472795, "loss": 0.153, "step": 1221 }, { "epoch": 0.08, "learning_rate": 0.00019877661851371915, "loss": 0.1936, "step": 1222 }, { "epoch": 0.08, "learning_rate": 0.00019877337797289405, "loss": 0.1465, "step": 1223 }, { "epoch": 0.08, "learning_rate": 0.0001987701331723925, "loss": 0.1767, "step": 1224 }, { "epoch": 0.08, "learning_rate": 0.00019876688411235436, "loss": 0.1681, "step": 1225 }, { "epoch": 0.08, "learning_rate": 0.00019876363079291976, "loss": 0.1636, "step": 1226 }, { "epoch": 0.08, "learning_rate": 0.00019876037321422897, "loss": 0.1556, "step": 1227 }, { "epoch": 0.08, "learning_rate": 0.00019875711137642258, "loss": 0.145, "step": 1228 }, { "epoch": 0.08, "learning_rate": 0.00019875384527964115, "loss": 0.1551, "step": 1229 }, { "epoch": 0.08, "learning_rate": 0.00019875057492402558, "loss": 0.1669, "step": 1230 }, { "epoch": 0.08, "learning_rate": 0.00019874730030971689, "loss": 0.1874, "step": 1231 }, { "epoch": 0.08, "learning_rate": 0.00019874402143685633, "loss": 0.1603, "step": 1232 }, { "epoch": 0.08, "learning_rate": 0.00019874073830558527, "loss": 0.1529, "step": 1233 }, { "epoch": 0.08, "learning_rate": 0.00019873745091604535, "loss": 0.1594, "step": 1234 }, { "epoch": 0.08, "learning_rate": 0.00019873415926837827, "loss": 0.1644, "step": 1235 }, { "epoch": 0.08, "learning_rate": 0.00019873086336272603, "loss": 0.1696, "step": 1236 }, { "epoch": 0.08, "learning_rate": 0.00019872756319923074, "loss": 0.1603, "step": 1237 }, { "epoch": 0.08, "learning_rate": 0.00019872425877803476, "loss": 0.1543, "step": 1238 }, { "epoch": 0.08, "learning_rate": 0.00019872095009928055, "loss": 0.1921, "step": 1239 }, { "epoch": 0.08, "learning_rate": 0.00019871763716311085, "loss": 0.1595, "step": 1240 }, { "epoch": 0.08, "learning_rate": 0.00019871431996966853, "loss": 0.1649, "step": 1241 }, { "epoch": 0.08, "learning_rate": 0.00019871099851909656, "loss": 0.1848, "step": 1242 }, { "epoch": 0.08, "learning_rate": 0.0001987076728115383, "loss": 0.1641, "step": 1243 }, { "epoch": 0.08, "learning_rate": 0.0001987043428471371, "loss": 0.1673, "step": 1244 }, { "epoch": 0.08, "learning_rate": 0.00019870100862603656, "loss": 0.1701, "step": 1245 }, { "epoch": 0.08, "learning_rate": 0.0001986976701483805, "loss": 0.1597, "step": 1246 }, { "epoch": 0.08, "learning_rate": 0.00019869432741431294, "loss": 0.1597, "step": 1247 }, { "epoch": 0.08, "learning_rate": 0.00019869098042397796, "loss": 0.145, "step": 1248 }, { "epoch": 0.08, "learning_rate": 0.00019868762917751992, "loss": 0.1689, "step": 1249 }, { "epoch": 0.08, "learning_rate": 0.00019868427367508335, "loss": 0.183, "step": 1250 }, { "epoch": 0.08, "learning_rate": 0.00019868091391681296, "loss": 0.199, "step": 1251 }, { "epoch": 0.08, "learning_rate": 0.00019867754990285366, "loss": 0.1569, "step": 1252 }, { "epoch": 0.08, "learning_rate": 0.00019867418163335053, "loss": 0.1742, "step": 1253 }, { "epoch": 0.08, "learning_rate": 0.00019867080910844877, "loss": 0.1568, "step": 1254 }, { "epoch": 0.08, "learning_rate": 0.00019866743232829388, "loss": 0.1828, "step": 1255 }, { "epoch": 0.08, "learning_rate": 0.00019866405129303146, "loss": 0.1441, "step": 1256 }, { "epoch": 0.08, "learning_rate": 0.00019866066600280732, "loss": 0.1763, "step": 1257 }, { "epoch": 0.08, "learning_rate": 0.0001986572764577675, "loss": 0.1576, "step": 1258 }, { "epoch": 0.08, "learning_rate": 0.0001986538826580581, "loss": 0.18, "step": 1259 }, { "epoch": 0.08, "learning_rate": 0.00019865048460382555, "loss": 0.1595, "step": 1260 }, { "epoch": 0.08, "learning_rate": 0.00019864708229521636, "loss": 0.1636, "step": 1261 }, { "epoch": 0.08, "learning_rate": 0.00019864367573237723, "loss": 0.1901, "step": 1262 }, { "epoch": 0.08, "learning_rate": 0.00019864026491545511, "loss": 0.166, "step": 1263 }, { "epoch": 0.08, "learning_rate": 0.0001986368498445971, "loss": 0.1619, "step": 1264 }, { "epoch": 0.08, "learning_rate": 0.00019863343051995045, "loss": 0.1656, "step": 1265 }, { "epoch": 0.08, "learning_rate": 0.00019863000694166262, "loss": 0.1558, "step": 1266 }, { "epoch": 0.08, "learning_rate": 0.00019862657910988128, "loss": 0.1656, "step": 1267 }, { "epoch": 0.08, "learning_rate": 0.00019862314702475425, "loss": 0.1735, "step": 1268 }, { "epoch": 0.08, "learning_rate": 0.00019861971068642955, "loss": 0.1465, "step": 1269 }, { "epoch": 0.08, "learning_rate": 0.00019861627009505533, "loss": 0.1721, "step": 1270 }, { "epoch": 0.08, "learning_rate": 0.00019861282525077999, "loss": 0.1674, "step": 1271 }, { "epoch": 0.08, "learning_rate": 0.00019860937615375214, "loss": 0.1593, "step": 1272 }, { "epoch": 0.08, "learning_rate": 0.00019860592280412047, "loss": 0.1599, "step": 1273 }, { "epoch": 0.08, "learning_rate": 0.0001986024652020339, "loss": 0.1606, "step": 1274 }, { "epoch": 0.08, "learning_rate": 0.00019859900334764162, "loss": 0.144, "step": 1275 }, { "epoch": 0.08, "learning_rate": 0.00019859553724109282, "loss": 0.1487, "step": 1276 }, { "epoch": 0.08, "learning_rate": 0.00019859206688253704, "loss": 0.1694, "step": 1277 }, { "epoch": 0.08, "learning_rate": 0.00019858859227212393, "loss": 0.1558, "step": 1278 }, { "epoch": 0.08, "learning_rate": 0.00019858511341000332, "loss": 0.1606, "step": 1279 }, { "epoch": 0.08, "learning_rate": 0.00019858163029632528, "loss": 0.1759, "step": 1280 }, { "epoch": 0.08, "learning_rate": 0.00019857814293123999, "loss": 0.1391, "step": 1281 }, { "epoch": 0.08, "learning_rate": 0.00019857465131489785, "loss": 0.1655, "step": 1282 }, { "epoch": 0.08, "learning_rate": 0.00019857115544744943, "loss": 0.19, "step": 1283 }, { "epoch": 0.08, "learning_rate": 0.0001985676553290455, "loss": 0.1314, "step": 1284 }, { "epoch": 0.08, "learning_rate": 0.00019856415095983706, "loss": 0.1692, "step": 1285 }, { "epoch": 0.08, "learning_rate": 0.00019856064233997512, "loss": 0.1633, "step": 1286 }, { "epoch": 0.08, "learning_rate": 0.0001985571294696111, "loss": 0.1837, "step": 1287 }, { "epoch": 0.08, "learning_rate": 0.0001985536123488964, "loss": 0.1978, "step": 1288 }, { "epoch": 0.08, "learning_rate": 0.0001985500909779828, "loss": 0.1638, "step": 1289 }, { "epoch": 0.08, "learning_rate": 0.00019854656535702212, "loss": 0.1502, "step": 1290 }, { "epoch": 0.08, "learning_rate": 0.0001985430354861664, "loss": 0.1615, "step": 1291 }, { "epoch": 0.08, "learning_rate": 0.00019853950136556781, "loss": 0.1561, "step": 1292 }, { "epoch": 0.08, "learning_rate": 0.00019853596299537887, "loss": 0.1963, "step": 1293 }, { "epoch": 0.08, "learning_rate": 0.0001985324203757521, "loss": 0.1716, "step": 1294 }, { "epoch": 0.08, "learning_rate": 0.00019852887350684027, "loss": 0.1451, "step": 1295 }, { "epoch": 0.08, "learning_rate": 0.00019852532238879645, "loss": 0.1737, "step": 1296 }, { "epoch": 0.08, "learning_rate": 0.00019852176702177366, "loss": 0.1595, "step": 1297 }, { "epoch": 0.08, "learning_rate": 0.0001985182074059253, "loss": 0.1599, "step": 1298 }, { "epoch": 0.08, "learning_rate": 0.00019851464354140482, "loss": 0.1617, "step": 1299 }, { "epoch": 0.08, "learning_rate": 0.00019851107542836598, "loss": 0.1523, "step": 1300 }, { "epoch": 0.08, "learning_rate": 0.0001985075030669626, "loss": 0.1734, "step": 1301 }, { "epoch": 0.08, "learning_rate": 0.0001985039264573488, "loss": 0.163, "step": 1302 }, { "epoch": 0.08, "learning_rate": 0.00019850034559967876, "loss": 0.1408, "step": 1303 }, { "epoch": 0.08, "learning_rate": 0.00019849676049410694, "loss": 0.1821, "step": 1304 }, { "epoch": 0.08, "learning_rate": 0.00019849317114078798, "loss": 0.1613, "step": 1305 }, { "epoch": 0.08, "learning_rate": 0.00019848957753987664, "loss": 0.1808, "step": 1306 }, { "epoch": 0.08, "learning_rate": 0.0001984859796915279, "loss": 0.1546, "step": 1307 }, { "epoch": 0.08, "learning_rate": 0.00019848237759589692, "loss": 0.1707, "step": 1308 }, { "epoch": 0.08, "learning_rate": 0.00019847877125313902, "loss": 0.1572, "step": 1309 }, { "epoch": 0.08, "learning_rate": 0.00019847516066340978, "loss": 0.16, "step": 1310 }, { "epoch": 0.08, "learning_rate": 0.00019847154582686486, "loss": 0.1388, "step": 1311 }, { "epoch": 0.08, "learning_rate": 0.00019846792674366016, "loss": 0.1784, "step": 1312 }, { "epoch": 0.08, "learning_rate": 0.00019846430341395182, "loss": 0.1626, "step": 1313 }, { "epoch": 0.08, "learning_rate": 0.000198460675837896, "loss": 0.1535, "step": 1314 }, { "epoch": 0.08, "learning_rate": 0.0001984570440156492, "loss": 0.1852, "step": 1315 }, { "epoch": 0.08, "learning_rate": 0.00019845340794736802, "loss": 0.1549, "step": 1316 }, { "epoch": 0.08, "learning_rate": 0.00019844976763320931, "loss": 0.1681, "step": 1317 }, { "epoch": 0.08, "learning_rate": 0.00019844612307333003, "loss": 0.1604, "step": 1318 }, { "epoch": 0.08, "learning_rate": 0.00019844247426788734, "loss": 0.1679, "step": 1319 }, { "epoch": 0.08, "learning_rate": 0.00019843882121703863, "loss": 0.1615, "step": 1320 }, { "epoch": 0.08, "learning_rate": 0.00019843516392094142, "loss": 0.1499, "step": 1321 }, { "epoch": 0.08, "learning_rate": 0.00019843150237975344, "loss": 0.1748, "step": 1322 }, { "epoch": 0.08, "learning_rate": 0.00019842783659363257, "loss": 0.1641, "step": 1323 }, { "epoch": 0.08, "learning_rate": 0.00019842416656273696, "loss": 0.1755, "step": 1324 }, { "epoch": 0.08, "learning_rate": 0.00019842049228722483, "loss": 0.1605, "step": 1325 }, { "epoch": 0.09, "learning_rate": 0.00019841681376725468, "loss": 0.1508, "step": 1326 }, { "epoch": 0.09, "learning_rate": 0.00019841313100298508, "loss": 0.1773, "step": 1327 }, { "epoch": 0.09, "learning_rate": 0.00019840944399457492, "loss": 0.1411, "step": 1328 }, { "epoch": 0.09, "learning_rate": 0.0001984057527421832, "loss": 0.166, "step": 1329 }, { "epoch": 0.09, "learning_rate": 0.00019840205724596906, "loss": 0.1599, "step": 1330 }, { "epoch": 0.09, "learning_rate": 0.00019839835750609192, "loss": 0.1673, "step": 1331 }, { "epoch": 0.09, "learning_rate": 0.0001983946535227113, "loss": 0.1913, "step": 1332 }, { "epoch": 0.09, "learning_rate": 0.00019839094529598694, "loss": 0.1593, "step": 1333 }, { "epoch": 0.09, "learning_rate": 0.0001983872328260788, "loss": 0.1749, "step": 1334 }, { "epoch": 0.09, "learning_rate": 0.0001983835161131469, "loss": 0.1534, "step": 1335 }, { "epoch": 0.09, "learning_rate": 0.00019837979515735166, "loss": 0.1582, "step": 1336 }, { "epoch": 0.09, "learning_rate": 0.0001983760699588534, "loss": 0.1673, "step": 1337 }, { "epoch": 0.09, "learning_rate": 0.00019837234051781288, "loss": 0.1738, "step": 1338 }, { "epoch": 0.09, "learning_rate": 0.00019836860683439088, "loss": 0.1629, "step": 1339 }, { "epoch": 0.09, "learning_rate": 0.00019836486890874845, "loss": 0.1739, "step": 1340 }, { "epoch": 0.09, "learning_rate": 0.00019836112674104676, "loss": 0.1675, "step": 1341 }, { "epoch": 0.09, "learning_rate": 0.0001983573803314472, "loss": 0.1448, "step": 1342 }, { "epoch": 0.09, "learning_rate": 0.00019835362968011135, "loss": 0.1713, "step": 1343 }, { "epoch": 0.09, "learning_rate": 0.00019834987478720096, "loss": 0.1861, "step": 1344 }, { "epoch": 0.09, "learning_rate": 0.00019834611565287794, "loss": 0.1522, "step": 1345 }, { "epoch": 0.09, "learning_rate": 0.00019834235227730447, "loss": 0.1657, "step": 1346 }, { "epoch": 0.09, "learning_rate": 0.00019833858466064276, "loss": 0.1731, "step": 1347 }, { "epoch": 0.09, "learning_rate": 0.00019833481280305535, "loss": 0.161, "step": 1348 }, { "epoch": 0.09, "learning_rate": 0.00019833103670470486, "loss": 0.1753, "step": 1349 }, { "epoch": 0.09, "learning_rate": 0.0001983272563657542, "loss": 0.174, "step": 1350 }, { "epoch": 0.09, "learning_rate": 0.00019832347178636635, "loss": 0.1632, "step": 1351 }, { "epoch": 0.09, "learning_rate": 0.00019831968296670454, "loss": 0.1509, "step": 1352 }, { "epoch": 0.09, "learning_rate": 0.00019831588990693218, "loss": 0.1648, "step": 1353 }, { "epoch": 0.09, "learning_rate": 0.0001983120926072128, "loss": 0.1408, "step": 1354 }, { "epoch": 0.09, "learning_rate": 0.00019830829106771018, "loss": 0.1575, "step": 1355 }, { "epoch": 0.09, "learning_rate": 0.00019830448528858832, "loss": 0.1457, "step": 1356 }, { "epoch": 0.09, "learning_rate": 0.0001983006752700113, "loss": 0.1643, "step": 1357 }, { "epoch": 0.09, "learning_rate": 0.0001982968610121434, "loss": 0.1482, "step": 1358 }, { "epoch": 0.09, "learning_rate": 0.00019829304251514916, "loss": 0.1422, "step": 1359 }, { "epoch": 0.09, "learning_rate": 0.00019828921977919323, "loss": 0.1629, "step": 1360 }, { "epoch": 0.09, "learning_rate": 0.0001982853928044405, "loss": 0.1588, "step": 1361 }, { "epoch": 0.09, "learning_rate": 0.000198281561591056, "loss": 0.1464, "step": 1362 }, { "epoch": 0.09, "learning_rate": 0.00019827772613920496, "loss": 0.1694, "step": 1363 }, { "epoch": 0.09, "learning_rate": 0.0001982738864490527, "loss": 0.1748, "step": 1364 }, { "epoch": 0.09, "learning_rate": 0.00019827004252076494, "loss": 0.1735, "step": 1365 }, { "epoch": 0.09, "learning_rate": 0.0001982661943545074, "loss": 0.157, "step": 1366 }, { "epoch": 0.09, "learning_rate": 0.000198262341950446, "loss": 0.1653, "step": 1367 }, { "epoch": 0.09, "learning_rate": 0.00019825848530874692, "loss": 0.1558, "step": 1368 }, { "epoch": 0.09, "learning_rate": 0.00019825462442957646, "loss": 0.1721, "step": 1369 }, { "epoch": 0.09, "learning_rate": 0.00019825075931310116, "loss": 0.1632, "step": 1370 }, { "epoch": 0.09, "learning_rate": 0.00019824688995948763, "loss": 0.1733, "step": 1371 }, { "epoch": 0.09, "learning_rate": 0.0001982430163689028, "loss": 0.194, "step": 1372 }, { "epoch": 0.09, "learning_rate": 0.00019823913854151375, "loss": 0.1813, "step": 1373 }, { "epoch": 0.09, "learning_rate": 0.0001982352564774876, "loss": 0.165, "step": 1374 }, { "epoch": 0.09, "learning_rate": 0.0001982313701769919, "loss": 0.1683, "step": 1375 }, { "epoch": 0.09, "learning_rate": 0.00019822747964019416, "loss": 0.1728, "step": 1376 }, { "epoch": 0.09, "learning_rate": 0.0001982235848672622, "loss": 0.1508, "step": 1377 }, { "epoch": 0.09, "learning_rate": 0.00019821968585836397, "loss": 0.173, "step": 1378 }, { "epoch": 0.09, "learning_rate": 0.00019821578261366762, "loss": 0.1544, "step": 1379 }, { "epoch": 0.09, "learning_rate": 0.0001982118751333415, "loss": 0.1845, "step": 1380 }, { "epoch": 0.09, "learning_rate": 0.0001982079634175541, "loss": 0.1945, "step": 1381 }, { "epoch": 0.09, "learning_rate": 0.00019820404746647409, "loss": 0.1631, "step": 1382 }, { "epoch": 0.09, "learning_rate": 0.00019820012728027044, "loss": 0.2064, "step": 1383 }, { "epoch": 0.09, "learning_rate": 0.00019819620285911212, "loss": 0.1712, "step": 1384 }, { "epoch": 0.09, "learning_rate": 0.0001981922742031684, "loss": 0.1851, "step": 1385 }, { "epoch": 0.09, "learning_rate": 0.0001981883413126087, "loss": 0.1879, "step": 1386 }, { "epoch": 0.09, "learning_rate": 0.0001981844041876027, "loss": 0.1608, "step": 1387 }, { "epoch": 0.09, "learning_rate": 0.00019818046282832005, "loss": 0.1585, "step": 1388 }, { "epoch": 0.09, "learning_rate": 0.00019817651723493085, "loss": 0.1875, "step": 1389 }, { "epoch": 0.09, "learning_rate": 0.00019817256740760523, "loss": 0.153, "step": 1390 }, { "epoch": 0.09, "learning_rate": 0.00019816861334651347, "loss": 0.169, "step": 1391 }, { "epoch": 0.09, "learning_rate": 0.00019816465505182617, "loss": 0.155, "step": 1392 }, { "epoch": 0.09, "learning_rate": 0.00019816069252371398, "loss": 0.191, "step": 1393 }, { "epoch": 0.09, "learning_rate": 0.00019815672576234782, "loss": 0.1618, "step": 1394 }, { "epoch": 0.09, "learning_rate": 0.00019815275476789873, "loss": 0.1783, "step": 1395 }, { "epoch": 0.09, "learning_rate": 0.00019814877954053798, "loss": 0.1625, "step": 1396 }, { "epoch": 0.09, "learning_rate": 0.000198144800080437, "loss": 0.1688, "step": 1397 }, { "epoch": 0.09, "learning_rate": 0.00019814081638776742, "loss": 0.162, "step": 1398 }, { "epoch": 0.09, "learning_rate": 0.00019813682846270102, "loss": 0.1803, "step": 1399 }, { "epoch": 0.09, "learning_rate": 0.0001981328363054098, "loss": 0.1561, "step": 1400 }, { "epoch": 0.09, "learning_rate": 0.0001981288399160659, "loss": 0.1531, "step": 1401 }, { "epoch": 0.09, "learning_rate": 0.0001981248392948417, "loss": 0.1821, "step": 1402 }, { "epoch": 0.09, "learning_rate": 0.0001981208344419097, "loss": 0.1763, "step": 1403 }, { "epoch": 0.09, "learning_rate": 0.00019811682535744263, "loss": 0.1834, "step": 1404 }, { "epoch": 0.09, "learning_rate": 0.00019811281204161336, "loss": 0.1653, "step": 1405 }, { "epoch": 0.09, "learning_rate": 0.000198108794494595, "loss": 0.1656, "step": 1406 }, { "epoch": 0.09, "learning_rate": 0.0001981047727165608, "loss": 0.1621, "step": 1407 }, { "epoch": 0.09, "learning_rate": 0.00019810074670768423, "loss": 0.1596, "step": 1408 }, { "epoch": 0.09, "learning_rate": 0.00019809671646813884, "loss": 0.1933, "step": 1409 }, { "epoch": 0.09, "learning_rate": 0.0001980926819980985, "loss": 0.1717, "step": 1410 }, { "epoch": 0.09, "learning_rate": 0.00019808864329773717, "loss": 0.1637, "step": 1411 }, { "epoch": 0.09, "learning_rate": 0.00019808460036722906, "loss": 0.1671, "step": 1412 }, { "epoch": 0.09, "learning_rate": 0.00019808055320674848, "loss": 0.1661, "step": 1413 }, { "epoch": 0.09, "learning_rate": 0.00019807650181647, "loss": 0.1763, "step": 1414 }, { "epoch": 0.09, "learning_rate": 0.00019807244619656828, "loss": 0.1658, "step": 1415 }, { "epoch": 0.09, "learning_rate": 0.00019806838634721832, "loss": 0.1637, "step": 1416 }, { "epoch": 0.09, "learning_rate": 0.0001980643222685951, "loss": 0.1434, "step": 1417 }, { "epoch": 0.09, "learning_rate": 0.00019806025396087396, "loss": 0.165, "step": 1418 }, { "epoch": 0.09, "learning_rate": 0.00019805618142423032, "loss": 0.1482, "step": 1419 }, { "epoch": 0.09, "learning_rate": 0.0001980521046588398, "loss": 0.1877, "step": 1420 }, { "epoch": 0.09, "learning_rate": 0.00019804802366487826, "loss": 0.1671, "step": 1421 }, { "epoch": 0.09, "learning_rate": 0.00019804393844252165, "loss": 0.1989, "step": 1422 }, { "epoch": 0.09, "learning_rate": 0.00019803984899194617, "loss": 0.1806, "step": 1423 }, { "epoch": 0.09, "learning_rate": 0.00019803575531332816, "loss": 0.1982, "step": 1424 }, { "epoch": 0.09, "learning_rate": 0.00019803165740684417, "loss": 0.183, "step": 1425 }, { "epoch": 0.09, "learning_rate": 0.00019802755527267095, "loss": 0.1607, "step": 1426 }, { "epoch": 0.09, "learning_rate": 0.0001980234489109854, "loss": 0.1613, "step": 1427 }, { "epoch": 0.09, "learning_rate": 0.00019801933832196456, "loss": 0.1754, "step": 1428 }, { "epoch": 0.09, "learning_rate": 0.00019801522350578577, "loss": 0.1606, "step": 1429 }, { "epoch": 0.09, "learning_rate": 0.00019801110446262648, "loss": 0.1732, "step": 1430 }, { "epoch": 0.09, "learning_rate": 0.00019800698119266428, "loss": 0.1637, "step": 1431 }, { "epoch": 0.09, "learning_rate": 0.000198002853696077, "loss": 0.1675, "step": 1432 }, { "epoch": 0.09, "learning_rate": 0.00019799872197304267, "loss": 0.1548, "step": 1433 }, { "epoch": 0.09, "learning_rate": 0.00019799458602373947, "loss": 0.1745, "step": 1434 }, { "epoch": 0.09, "learning_rate": 0.0001979904458483457, "loss": 0.1392, "step": 1435 }, { "epoch": 0.09, "learning_rate": 0.00019798630144704, "loss": 0.1625, "step": 1436 }, { "epoch": 0.09, "learning_rate": 0.00019798215282000107, "loss": 0.1937, "step": 1437 }, { "epoch": 0.09, "learning_rate": 0.0001979779999674078, "loss": 0.1873, "step": 1438 }, { "epoch": 0.09, "learning_rate": 0.0001979738428894393, "loss": 0.1499, "step": 1439 }, { "epoch": 0.09, "learning_rate": 0.00019796968158627488, "loss": 0.1721, "step": 1440 }, { "epoch": 0.09, "learning_rate": 0.00019796551605809391, "loss": 0.193, "step": 1441 }, { "epoch": 0.09, "learning_rate": 0.0001979613463050761, "loss": 0.1533, "step": 1442 }, { "epoch": 0.09, "learning_rate": 0.0001979571723274013, "loss": 0.1463, "step": 1443 }, { "epoch": 0.09, "learning_rate": 0.00019795299412524945, "loss": 0.1496, "step": 1444 }, { "epoch": 0.09, "learning_rate": 0.00019794881169880077, "loss": 0.1749, "step": 1445 }, { "epoch": 0.09, "learning_rate": 0.00019794462504823564, "loss": 0.1876, "step": 1446 }, { "epoch": 0.09, "learning_rate": 0.00019794043417373458, "loss": 0.1768, "step": 1447 }, { "epoch": 0.09, "learning_rate": 0.00019793623907547834, "loss": 0.1666, "step": 1448 }, { "epoch": 0.09, "learning_rate": 0.00019793203975364786, "loss": 0.1912, "step": 1449 }, { "epoch": 0.09, "learning_rate": 0.0001979278362084242, "loss": 0.2897, "step": 1450 }, { "epoch": 0.09, "learning_rate": 0.00019792362843998866, "loss": 0.1698, "step": 1451 }, { "epoch": 0.09, "learning_rate": 0.00019791941644852273, "loss": 0.1865, "step": 1452 }, { "epoch": 0.09, "learning_rate": 0.000197915200234208, "loss": 0.1884, "step": 1453 }, { "epoch": 0.09, "learning_rate": 0.00019791097979722632, "loss": 0.1798, "step": 1454 }, { "epoch": 0.09, "learning_rate": 0.00019790675513775975, "loss": 0.1789, "step": 1455 }, { "epoch": 0.09, "learning_rate": 0.0001979025262559904, "loss": 0.1601, "step": 1456 }, { "epoch": 0.09, "learning_rate": 0.0001978982931521007, "loss": 0.1767, "step": 1457 }, { "epoch": 0.09, "learning_rate": 0.00019789405582627313, "loss": 0.162, "step": 1458 }, { "epoch": 0.09, "learning_rate": 0.00019788981427869057, "loss": 0.1575, "step": 1459 }, { "epoch": 0.09, "learning_rate": 0.0001978855685095358, "loss": 0.1894, "step": 1460 }, { "epoch": 0.09, "learning_rate": 0.00019788131851899198, "loss": 0.1721, "step": 1461 }, { "epoch": 0.09, "learning_rate": 0.0001978770643072424, "loss": 0.1735, "step": 1462 }, { "epoch": 0.09, "learning_rate": 0.00019787280587447051, "loss": 0.1646, "step": 1463 }, { "epoch": 0.09, "learning_rate": 0.00019786854322085997, "loss": 0.1639, "step": 1464 }, { "epoch": 0.09, "learning_rate": 0.0001978642763465946, "loss": 0.1743, "step": 1465 }, { "epoch": 0.09, "learning_rate": 0.00019786000525185844, "loss": 0.1371, "step": 1466 }, { "epoch": 0.09, "learning_rate": 0.00019785572993683562, "loss": 0.1514, "step": 1467 }, { "epoch": 0.09, "learning_rate": 0.0001978514504017106, "loss": 0.1766, "step": 1468 }, { "epoch": 0.09, "learning_rate": 0.00019784716664666786, "loss": 0.1623, "step": 1469 }, { "epoch": 0.09, "learning_rate": 0.00019784287867189218, "loss": 0.1685, "step": 1470 }, { "epoch": 0.09, "learning_rate": 0.00019783858647756853, "loss": 0.1557, "step": 1471 }, { "epoch": 0.09, "learning_rate": 0.0001978342900638819, "loss": 0.1822, "step": 1472 }, { "epoch": 0.09, "learning_rate": 0.00019782998943101765, "loss": 0.1742, "step": 1473 }, { "epoch": 0.09, "learning_rate": 0.00019782568457916123, "loss": 0.1861, "step": 1474 }, { "epoch": 0.09, "learning_rate": 0.00019782137550849834, "loss": 0.2046, "step": 1475 }, { "epoch": 0.09, "learning_rate": 0.00019781706221921473, "loss": 0.1814, "step": 1476 }, { "epoch": 0.09, "learning_rate": 0.00019781274471149645, "loss": 0.1854, "step": 1477 }, { "epoch": 0.09, "learning_rate": 0.00019780842298552971, "loss": 0.1799, "step": 1478 }, { "epoch": 0.09, "learning_rate": 0.00019780409704150088, "loss": 0.1747, "step": 1479 }, { "epoch": 0.09, "learning_rate": 0.00019779976687959652, "loss": 0.167, "step": 1480 }, { "epoch": 0.09, "learning_rate": 0.00019779543250000335, "loss": 0.1728, "step": 1481 }, { "epoch": 0.1, "learning_rate": 0.00019779109390290832, "loss": 0.1893, "step": 1482 }, { "epoch": 0.1, "learning_rate": 0.00019778675108849848, "loss": 0.165, "step": 1483 }, { "epoch": 0.1, "learning_rate": 0.00019778240405696123, "loss": 0.1737, "step": 1484 }, { "epoch": 0.1, "learning_rate": 0.00019777805280848395, "loss": 0.1841, "step": 1485 }, { "epoch": 0.1, "learning_rate": 0.00019777369734325429, "loss": 0.1728, "step": 1486 }, { "epoch": 0.1, "learning_rate": 0.00019776933766146014, "loss": 0.1829, "step": 1487 }, { "epoch": 0.1, "learning_rate": 0.00019776497376328944, "loss": 0.1572, "step": 1488 }, { "epoch": 0.1, "learning_rate": 0.00019776060564893045, "loss": 0.1817, "step": 1489 }, { "epoch": 0.1, "learning_rate": 0.00019775623331857152, "loss": 0.1746, "step": 1490 }, { "epoch": 0.1, "learning_rate": 0.0001977518567724012, "loss": 0.1707, "step": 1491 }, { "epoch": 0.1, "learning_rate": 0.00019774747601060825, "loss": 0.1565, "step": 1492 }, { "epoch": 0.1, "learning_rate": 0.0001977430910333816, "loss": 0.1818, "step": 1493 }, { "epoch": 0.1, "learning_rate": 0.00019773870184091035, "loss": 0.19, "step": 1494 }, { "epoch": 0.1, "learning_rate": 0.00019773430843338376, "loss": 0.1734, "step": 1495 }, { "epoch": 0.1, "learning_rate": 0.00019772991081099136, "loss": 0.1652, "step": 1496 }, { "epoch": 0.1, "learning_rate": 0.00019772550897392274, "loss": 0.13, "step": 1497 }, { "epoch": 0.1, "learning_rate": 0.00019772110292236774, "loss": 0.1548, "step": 1498 }, { "epoch": 0.1, "learning_rate": 0.00019771669265651641, "loss": 0.1897, "step": 1499 }, { "epoch": 0.1, "learning_rate": 0.00019771227817655892, "loss": 0.1454, "step": 1500 }, { "epoch": 0.1, "learning_rate": 0.00019770785948268563, "loss": 0.1758, "step": 1501 }, { "epoch": 0.1, "learning_rate": 0.00019770343657508717, "loss": 0.1901, "step": 1502 }, { "epoch": 0.1, "learning_rate": 0.00019769900945395419, "loss": 0.1702, "step": 1503 }, { "epoch": 0.1, "learning_rate": 0.00019769457811947766, "loss": 0.1626, "step": 1504 }, { "epoch": 0.1, "learning_rate": 0.0001976901425718487, "loss": 0.1862, "step": 1505 }, { "epoch": 0.1, "learning_rate": 0.0001976857028112586, "loss": 0.1635, "step": 1506 }, { "epoch": 0.1, "learning_rate": 0.00019768125883789877, "loss": 0.1889, "step": 1507 }, { "epoch": 0.1, "learning_rate": 0.00019767681065196085, "loss": 0.1579, "step": 1508 }, { "epoch": 0.1, "learning_rate": 0.0001976723582536368, "loss": 0.1827, "step": 1509 }, { "epoch": 0.1, "learning_rate": 0.00019766790164311851, "loss": 0.1613, "step": 1510 }, { "epoch": 0.1, "learning_rate": 0.00019766344082059823, "loss": 0.1789, "step": 1511 }, { "epoch": 0.1, "learning_rate": 0.0001976589757862683, "loss": 0.1761, "step": 1512 }, { "epoch": 0.1, "learning_rate": 0.0001976545065403213, "loss": 0.1646, "step": 1513 }, { "epoch": 0.1, "learning_rate": 0.00019765003308295001, "loss": 0.1704, "step": 1514 }, { "epoch": 0.1, "learning_rate": 0.00019764555541434725, "loss": 0.1841, "step": 1515 }, { "epoch": 0.1, "learning_rate": 0.0001976410735347062, "loss": 0.1488, "step": 1516 }, { "epoch": 0.1, "learning_rate": 0.00019763658744422014, "loss": 0.1658, "step": 1517 }, { "epoch": 0.1, "learning_rate": 0.00019763209714308255, "loss": 0.1575, "step": 1518 }, { "epoch": 0.1, "learning_rate": 0.000197627602631487, "loss": 0.158, "step": 1519 }, { "epoch": 0.1, "learning_rate": 0.0001976231039096274, "loss": 0.1705, "step": 1520 }, { "epoch": 0.1, "learning_rate": 0.00019761860097769772, "loss": 0.172, "step": 1521 }, { "epoch": 0.1, "learning_rate": 0.00019761409383589216, "loss": 0.1831, "step": 1522 }, { "epoch": 0.1, "learning_rate": 0.0001976095824844051, "loss": 0.1597, "step": 1523 }, { "epoch": 0.1, "learning_rate": 0.00019760506692343107, "loss": 0.1825, "step": 1524 }, { "epoch": 0.1, "learning_rate": 0.00019760054715316488, "loss": 0.1554, "step": 1525 }, { "epoch": 0.1, "learning_rate": 0.00019759602317380133, "loss": 0.1733, "step": 1526 }, { "epoch": 0.1, "learning_rate": 0.00019759149498553566, "loss": 0.1792, "step": 1527 }, { "epoch": 0.1, "learning_rate": 0.00019758696258856303, "loss": 0.145, "step": 1528 }, { "epoch": 0.1, "learning_rate": 0.00019758242598307896, "loss": 0.1682, "step": 1529 }, { "epoch": 0.1, "learning_rate": 0.0001975778851692791, "loss": 0.1731, "step": 1530 }, { "epoch": 0.1, "learning_rate": 0.00019757334014735927, "loss": 0.1611, "step": 1531 }, { "epoch": 0.1, "learning_rate": 0.00019756879091751544, "loss": 0.1777, "step": 1532 }, { "epoch": 0.1, "learning_rate": 0.00019756423747994384, "loss": 0.165, "step": 1533 }, { "epoch": 0.1, "learning_rate": 0.00019755967983484083, "loss": 0.1653, "step": 1534 }, { "epoch": 0.1, "learning_rate": 0.00019755511798240298, "loss": 0.158, "step": 1535 }, { "epoch": 0.1, "learning_rate": 0.000197550551922827, "loss": 0.1606, "step": 1536 }, { "epoch": 0.1, "learning_rate": 0.0001975459816563098, "loss": 0.1534, "step": 1537 }, { "epoch": 0.1, "learning_rate": 0.00019754140718304845, "loss": 0.1697, "step": 1538 }, { "epoch": 0.1, "learning_rate": 0.00019753682850324032, "loss": 0.1456, "step": 1539 }, { "epoch": 0.1, "learning_rate": 0.0001975322456170828, "loss": 0.167, "step": 1540 }, { "epoch": 0.1, "learning_rate": 0.00019752765852477352, "loss": 0.1545, "step": 1541 }, { "epoch": 0.1, "learning_rate": 0.00019752306722651034, "loss": 0.1689, "step": 1542 }, { "epoch": 0.1, "learning_rate": 0.00019751847172249124, "loss": 0.1548, "step": 1543 }, { "epoch": 0.1, "learning_rate": 0.00019751387201291442, "loss": 0.1595, "step": 1544 }, { "epoch": 0.1, "learning_rate": 0.00019750926809797826, "loss": 0.1831, "step": 1545 }, { "epoch": 0.1, "learning_rate": 0.00019750465997788122, "loss": 0.1658, "step": 1546 }, { "epoch": 0.1, "learning_rate": 0.00019750004765282215, "loss": 0.1353, "step": 1547 }, { "epoch": 0.1, "learning_rate": 0.0001974954311229999, "loss": 0.1554, "step": 1548 }, { "epoch": 0.1, "learning_rate": 0.00019749081038861354, "loss": 0.1398, "step": 1549 }, { "epoch": 0.1, "learning_rate": 0.00019748618544986242, "loss": 0.1712, "step": 1550 }, { "epoch": 0.1, "learning_rate": 0.0001974815563069459, "loss": 0.18, "step": 1551 }, { "epoch": 0.1, "learning_rate": 0.00019747692296006366, "loss": 0.1527, "step": 1552 }, { "epoch": 0.1, "learning_rate": 0.00019747228540941553, "loss": 0.1602, "step": 1553 }, { "epoch": 0.1, "learning_rate": 0.00019746764365520148, "loss": 0.2086, "step": 1554 }, { "epoch": 0.1, "learning_rate": 0.00019746299769762168, "loss": 0.1656, "step": 1555 }, { "epoch": 0.1, "learning_rate": 0.00019745834753687652, "loss": 0.1665, "step": 1556 }, { "epoch": 0.1, "learning_rate": 0.00019745369317316657, "loss": 0.1532, "step": 1557 }, { "epoch": 0.1, "learning_rate": 0.0001974490346066925, "loss": 0.1625, "step": 1558 }, { "epoch": 0.1, "learning_rate": 0.00019744437183765522, "loss": 0.1728, "step": 1559 }, { "epoch": 0.1, "learning_rate": 0.00019743970486625582, "loss": 0.1635, "step": 1560 }, { "epoch": 0.1, "learning_rate": 0.00019743503369269562, "loss": 0.1619, "step": 1561 }, { "epoch": 0.1, "learning_rate": 0.00019743035831717596, "loss": 0.1526, "step": 1562 }, { "epoch": 0.1, "learning_rate": 0.00019742567873989856, "loss": 0.1493, "step": 1563 }, { "epoch": 0.1, "learning_rate": 0.0001974209949610652, "loss": 0.1679, "step": 1564 }, { "epoch": 0.1, "learning_rate": 0.00019741630698087785, "loss": 0.1654, "step": 1565 }, { "epoch": 0.1, "learning_rate": 0.0001974116147995387, "loss": 0.1791, "step": 1566 }, { "epoch": 0.1, "learning_rate": 0.00019740691841725014, "loss": 0.1675, "step": 1567 }, { "epoch": 0.1, "learning_rate": 0.00019740221783421467, "loss": 0.1628, "step": 1568 }, { "epoch": 0.1, "learning_rate": 0.000197397513050635, "loss": 0.1567, "step": 1569 }, { "epoch": 0.1, "learning_rate": 0.00019739280406671405, "loss": 0.1679, "step": 1570 }, { "epoch": 0.1, "learning_rate": 0.00019738809088265484, "loss": 0.1631, "step": 1571 }, { "epoch": 0.1, "learning_rate": 0.00019738337349866071, "loss": 0.1514, "step": 1572 }, { "epoch": 0.1, "learning_rate": 0.00019737865191493505, "loss": 0.1601, "step": 1573 }, { "epoch": 0.1, "learning_rate": 0.0001973739261316815, "loss": 0.1645, "step": 1574 }, { "epoch": 0.1, "learning_rate": 0.00019736919614910388, "loss": 0.1843, "step": 1575 }, { "epoch": 0.1, "learning_rate": 0.00019736446196740612, "loss": 0.1743, "step": 1576 }, { "epoch": 0.1, "learning_rate": 0.00019735972358679246, "loss": 0.1541, "step": 1577 }, { "epoch": 0.1, "learning_rate": 0.0001973549810074672, "loss": 0.1519, "step": 1578 }, { "epoch": 0.1, "learning_rate": 0.00019735023422963485, "loss": 0.176, "step": 1579 }, { "epoch": 0.1, "learning_rate": 0.00019734548325350013, "loss": 0.1602, "step": 1580 }, { "epoch": 0.1, "learning_rate": 0.00019734072807926797, "loss": 0.1971, "step": 1581 }, { "epoch": 0.1, "learning_rate": 0.0001973359687071434, "loss": 0.1763, "step": 1582 }, { "epoch": 0.1, "learning_rate": 0.00019733120513733168, "loss": 0.1741, "step": 1583 }, { "epoch": 0.1, "learning_rate": 0.00019732643737003827, "loss": 0.1548, "step": 1584 }, { "epoch": 0.1, "learning_rate": 0.00019732166540546874, "loss": 0.1926, "step": 1585 }, { "epoch": 0.1, "learning_rate": 0.0001973168892438289, "loss": 0.1748, "step": 1586 }, { "epoch": 0.1, "learning_rate": 0.00019731210888532472, "loss": 0.1761, "step": 1587 }, { "epoch": 0.1, "learning_rate": 0.00019730732433016236, "loss": 0.1675, "step": 1588 }, { "epoch": 0.1, "learning_rate": 0.00019730253557854817, "loss": 0.1618, "step": 1589 }, { "epoch": 0.1, "learning_rate": 0.0001972977426306887, "loss": 0.1615, "step": 1590 }, { "epoch": 0.1, "learning_rate": 0.0001972929454867906, "loss": 0.1707, "step": 1591 }, { "epoch": 0.1, "learning_rate": 0.00019728814414706074, "loss": 0.1667, "step": 1592 }, { "epoch": 0.1, "learning_rate": 0.0001972833386117062, "loss": 0.1615, "step": 1593 }, { "epoch": 0.1, "learning_rate": 0.00019727852888093422, "loss": 0.1881, "step": 1594 }, { "epoch": 0.1, "learning_rate": 0.00019727371495495226, "loss": 0.1739, "step": 1595 }, { "epoch": 0.1, "learning_rate": 0.00019726889683396786, "loss": 0.1501, "step": 1596 }, { "epoch": 0.1, "learning_rate": 0.00019726407451818885, "loss": 0.1677, "step": 1597 }, { "epoch": 0.1, "learning_rate": 0.0001972592480078232, "loss": 0.1477, "step": 1598 }, { "epoch": 0.1, "learning_rate": 0.00019725441730307903, "loss": 0.1421, "step": 1599 }, { "epoch": 0.1, "learning_rate": 0.00019724958240416467, "loss": 0.1772, "step": 1600 }, { "epoch": 0.1, "learning_rate": 0.00019724474331128864, "loss": 0.1453, "step": 1601 }, { "epoch": 0.1, "learning_rate": 0.00019723990002465963, "loss": 0.17, "step": 1602 }, { "epoch": 0.1, "learning_rate": 0.0001972350525444865, "loss": 0.1643, "step": 1603 }, { "epoch": 0.1, "learning_rate": 0.00019723020087097833, "loss": 0.1536, "step": 1604 }, { "epoch": 0.1, "learning_rate": 0.0001972253450043443, "loss": 0.1477, "step": 1605 }, { "epoch": 0.1, "learning_rate": 0.0001972204849447939, "loss": 0.1769, "step": 1606 }, { "epoch": 0.1, "learning_rate": 0.00019721562069253664, "loss": 0.1571, "step": 1607 }, { "epoch": 0.1, "learning_rate": 0.0001972107522477823, "loss": 0.1704, "step": 1608 }, { "epoch": 0.1, "learning_rate": 0.0001972058796107409, "loss": 0.1501, "step": 1609 }, { "epoch": 0.1, "learning_rate": 0.00019720100278162257, "loss": 0.1627, "step": 1610 }, { "epoch": 0.1, "learning_rate": 0.00019719612176063758, "loss": 0.1779, "step": 1611 }, { "epoch": 0.1, "learning_rate": 0.00019719123654799646, "loss": 0.1513, "step": 1612 }, { "epoch": 0.1, "learning_rate": 0.00019718634714390986, "loss": 0.1451, "step": 1613 }, { "epoch": 0.1, "learning_rate": 0.00019718145354858865, "loss": 0.1809, "step": 1614 }, { "epoch": 0.1, "learning_rate": 0.0001971765557622439, "loss": 0.1773, "step": 1615 }, { "epoch": 0.1, "learning_rate": 0.00019717165378508678, "loss": 0.1644, "step": 1616 }, { "epoch": 0.1, "learning_rate": 0.00019716674761732872, "loss": 0.1487, "step": 1617 }, { "epoch": 0.1, "learning_rate": 0.00019716183725918132, "loss": 0.1723, "step": 1618 }, { "epoch": 0.1, "learning_rate": 0.0001971569227108563, "loss": 0.1843, "step": 1619 }, { "epoch": 0.1, "learning_rate": 0.00019715200397256563, "loss": 0.1763, "step": 1620 }, { "epoch": 0.1, "learning_rate": 0.00019714708104452146, "loss": 0.1571, "step": 1621 }, { "epoch": 0.1, "learning_rate": 0.00019714215392693603, "loss": 0.1636, "step": 1622 }, { "epoch": 0.1, "learning_rate": 0.00019713722262002187, "loss": 0.1646, "step": 1623 }, { "epoch": 0.1, "learning_rate": 0.00019713228712399167, "loss": 0.1399, "step": 1624 }, { "epoch": 0.1, "learning_rate": 0.0001971273474390582, "loss": 0.1533, "step": 1625 }, { "epoch": 0.1, "learning_rate": 0.00019712240356543456, "loss": 0.1919, "step": 1626 }, { "epoch": 0.1, "learning_rate": 0.0001971174555033339, "loss": 0.1639, "step": 1627 }, { "epoch": 0.1, "learning_rate": 0.00019711250325296967, "loss": 0.1747, "step": 1628 }, { "epoch": 0.1, "learning_rate": 0.0001971075468145554, "loss": 0.1742, "step": 1629 }, { "epoch": 0.1, "learning_rate": 0.00019710258618830487, "loss": 0.1854, "step": 1630 }, { "epoch": 0.1, "learning_rate": 0.00019709762137443194, "loss": 0.1973, "step": 1631 }, { "epoch": 0.1, "learning_rate": 0.0001970926523731508, "loss": 0.1471, "step": 1632 }, { "epoch": 0.1, "learning_rate": 0.0001970876791846757, "loss": 0.196, "step": 1633 }, { "epoch": 0.1, "learning_rate": 0.00019708270180922118, "loss": 0.1854, "step": 1634 }, { "epoch": 0.1, "learning_rate": 0.0001970777202470018, "loss": 0.1668, "step": 1635 }, { "epoch": 0.1, "learning_rate": 0.00019707273449823243, "loss": 0.1727, "step": 1636 }, { "epoch": 0.1, "learning_rate": 0.0001970677445631281, "loss": 0.1467, "step": 1637 }, { "epoch": 0.11, "learning_rate": 0.00019706275044190398, "loss": 0.1852, "step": 1638 }, { "epoch": 0.11, "learning_rate": 0.00019705775213477544, "loss": 0.1617, "step": 1639 }, { "epoch": 0.11, "learning_rate": 0.00019705274964195807, "loss": 0.1558, "step": 1640 }, { "epoch": 0.11, "learning_rate": 0.00019704774296366758, "loss": 0.1239, "step": 1641 }, { "epoch": 0.11, "learning_rate": 0.00019704273210011992, "loss": 0.1686, "step": 1642 }, { "epoch": 0.11, "learning_rate": 0.00019703771705153114, "loss": 0.1806, "step": 1643 }, { "epoch": 0.11, "learning_rate": 0.00019703269781811756, "loss": 0.1637, "step": 1644 }, { "epoch": 0.11, "learning_rate": 0.0001970276744000956, "loss": 0.1783, "step": 1645 }, { "epoch": 0.11, "learning_rate": 0.0001970226467976819, "loss": 0.1571, "step": 1646 }, { "epoch": 0.11, "learning_rate": 0.00019701761501109334, "loss": 0.1583, "step": 1647 }, { "epoch": 0.11, "learning_rate": 0.00019701257904054686, "loss": 0.1818, "step": 1648 }, { "epoch": 0.11, "learning_rate": 0.00019700753888625964, "loss": 0.1753, "step": 1649 }, { "epoch": 0.11, "learning_rate": 0.0001970024945484491, "loss": 0.1426, "step": 1650 }, { "epoch": 0.11, "learning_rate": 0.0001969974460273327, "loss": 0.1837, "step": 1651 }, { "epoch": 0.11, "learning_rate": 0.0001969923933231282, "loss": 0.1713, "step": 1652 }, { "epoch": 0.11, "learning_rate": 0.0001969873364360535, "loss": 0.1432, "step": 1653 }, { "epoch": 0.11, "learning_rate": 0.0001969822753663267, "loss": 0.1613, "step": 1654 }, { "epoch": 0.11, "learning_rate": 0.00019697721011416604, "loss": 0.1577, "step": 1655 }, { "epoch": 0.11, "learning_rate": 0.00019697214067978999, "loss": 0.1506, "step": 1656 }, { "epoch": 0.11, "learning_rate": 0.0001969670670634171, "loss": 0.1687, "step": 1657 }, { "epoch": 0.11, "learning_rate": 0.00019696198926526622, "loss": 0.1934, "step": 1658 }, { "epoch": 0.11, "learning_rate": 0.0001969569072855564, "loss": 0.1689, "step": 1659 }, { "epoch": 0.11, "learning_rate": 0.0001969518211245067, "loss": 0.141, "step": 1660 }, { "epoch": 0.11, "learning_rate": 0.0001969467307823365, "loss": 0.1484, "step": 1661 }, { "epoch": 0.11, "learning_rate": 0.00019694163625926536, "loss": 0.1342, "step": 1662 }, { "epoch": 0.11, "learning_rate": 0.00019693653755551295, "loss": 0.1708, "step": 1663 }, { "epoch": 0.11, "learning_rate": 0.00019693143467129916, "loss": 0.1787, "step": 1664 }, { "epoch": 0.11, "learning_rate": 0.00019692632760684407, "loss": 0.1741, "step": 1665 }, { "epoch": 0.11, "learning_rate": 0.00019692121636236788, "loss": 0.1585, "step": 1666 }, { "epoch": 0.11, "learning_rate": 0.0001969161009380911, "loss": 0.1665, "step": 1667 }, { "epoch": 0.11, "learning_rate": 0.00019691098133423423, "loss": 0.1587, "step": 1668 }, { "epoch": 0.11, "learning_rate": 0.0001969058575510182, "loss": 0.1691, "step": 1669 }, { "epoch": 0.11, "learning_rate": 0.00019690072958866378, "loss": 0.1519, "step": 1670 }, { "epoch": 0.11, "learning_rate": 0.00019689559744739233, "loss": 0.1604, "step": 1671 }, { "epoch": 0.11, "learning_rate": 0.00019689046112742503, "loss": 0.1562, "step": 1672 }, { "epoch": 0.11, "learning_rate": 0.0001968853206289834, "loss": 0.1616, "step": 1673 }, { "epoch": 0.11, "learning_rate": 0.0001968801759522892, "loss": 0.1686, "step": 1674 }, { "epoch": 0.11, "learning_rate": 0.00019687502709756424, "loss": 0.1615, "step": 1675 }, { "epoch": 0.11, "learning_rate": 0.00019686987406503063, "loss": 0.1637, "step": 1676 }, { "epoch": 0.11, "learning_rate": 0.00019686471685491052, "loss": 0.1624, "step": 1677 }, { "epoch": 0.11, "learning_rate": 0.00019685955546742636, "loss": 0.1549, "step": 1678 }, { "epoch": 0.11, "learning_rate": 0.00019685438990280074, "loss": 0.1518, "step": 1679 }, { "epoch": 0.11, "learning_rate": 0.0001968492201612564, "loss": 0.1662, "step": 1680 }, { "epoch": 0.11, "learning_rate": 0.00019684404624301636, "loss": 0.1605, "step": 1681 }, { "epoch": 0.11, "learning_rate": 0.00019683886814830367, "loss": 0.1643, "step": 1682 }, { "epoch": 0.11, "learning_rate": 0.00019683368587734165, "loss": 0.198, "step": 1683 }, { "epoch": 0.11, "learning_rate": 0.00019682849943035384, "loss": 0.1618, "step": 1684 }, { "epoch": 0.11, "learning_rate": 0.00019682330880756385, "loss": 0.181, "step": 1685 }, { "epoch": 0.11, "learning_rate": 0.00019681811400919558, "loss": 0.1783, "step": 1686 }, { "epoch": 0.11, "learning_rate": 0.00019681291503547302, "loss": 0.1493, "step": 1687 }, { "epoch": 0.11, "learning_rate": 0.00019680771188662044, "loss": 0.171, "step": 1688 }, { "epoch": 0.11, "learning_rate": 0.00019680250456286216, "loss": 0.1718, "step": 1689 }, { "epoch": 0.11, "learning_rate": 0.00019679729306442277, "loss": 0.1814, "step": 1690 }, { "epoch": 0.11, "learning_rate": 0.00019679207739152702, "loss": 0.1747, "step": 1691 }, { "epoch": 0.11, "learning_rate": 0.00019678685754439987, "loss": 0.1599, "step": 1692 }, { "epoch": 0.11, "learning_rate": 0.00019678163352326638, "loss": 0.1757, "step": 1693 }, { "epoch": 0.11, "learning_rate": 0.0001967764053283519, "loss": 0.1885, "step": 1694 }, { "epoch": 0.11, "learning_rate": 0.00019677117295988184, "loss": 0.1912, "step": 1695 }, { "epoch": 0.11, "learning_rate": 0.0001967659364180819, "loss": 0.1502, "step": 1696 }, { "epoch": 0.11, "learning_rate": 0.00019676069570317785, "loss": 0.146, "step": 1697 }, { "epoch": 0.11, "learning_rate": 0.00019675545081539574, "loss": 0.1817, "step": 1698 }, { "epoch": 0.11, "learning_rate": 0.0001967502017549618, "loss": 0.1577, "step": 1699 }, { "epoch": 0.11, "learning_rate": 0.0001967449485221023, "loss": 0.1475, "step": 1700 }, { "epoch": 0.11, "learning_rate": 0.00019673969111704386, "loss": 0.1959, "step": 1701 }, { "epoch": 0.11, "learning_rate": 0.0001967344295400132, "loss": 0.1621, "step": 1702 }, { "epoch": 0.11, "learning_rate": 0.00019672916379123723, "loss": 0.1573, "step": 1703 }, { "epoch": 0.11, "learning_rate": 0.00019672389387094302, "loss": 0.145, "step": 1704 }, { "epoch": 0.11, "learning_rate": 0.00019671861977935788, "loss": 0.1684, "step": 1705 }, { "epoch": 0.11, "learning_rate": 0.00019671334151670923, "loss": 0.1807, "step": 1706 }, { "epoch": 0.11, "learning_rate": 0.00019670805908322464, "loss": 0.1647, "step": 1707 }, { "epoch": 0.11, "learning_rate": 0.00019670277247913205, "loss": 0.1675, "step": 1708 }, { "epoch": 0.11, "learning_rate": 0.00019669748170465936, "loss": 0.1773, "step": 1709 }, { "epoch": 0.11, "learning_rate": 0.00019669218676003473, "loss": 0.1686, "step": 1710 }, { "epoch": 0.11, "learning_rate": 0.00019668688764548655, "loss": 0.1485, "step": 1711 }, { "epoch": 0.11, "learning_rate": 0.00019668158436124332, "loss": 0.1528, "step": 1712 }, { "epoch": 0.11, "learning_rate": 0.00019667627690753378, "loss": 0.1672, "step": 1713 }, { "epoch": 0.11, "learning_rate": 0.00019667096528458683, "loss": 0.1418, "step": 1714 }, { "epoch": 0.11, "learning_rate": 0.00019666564949263147, "loss": 0.1341, "step": 1715 }, { "epoch": 0.11, "learning_rate": 0.000196660329531897, "loss": 0.1803, "step": 1716 }, { "epoch": 0.11, "learning_rate": 0.0001966550054026128, "loss": 0.158, "step": 1717 }, { "epoch": 0.11, "learning_rate": 0.00019664967710500856, "loss": 0.1864, "step": 1718 }, { "epoch": 0.11, "learning_rate": 0.00019664434463931396, "loss": 0.1756, "step": 1719 }, { "epoch": 0.11, "learning_rate": 0.00019663900800575906, "loss": 0.172, "step": 1720 }, { "epoch": 0.11, "learning_rate": 0.00019663366720457394, "loss": 0.1822, "step": 1721 }, { "epoch": 0.11, "learning_rate": 0.000196628322235989, "loss": 0.1496, "step": 1722 }, { "epoch": 0.11, "learning_rate": 0.00019662297310023464, "loss": 0.1648, "step": 1723 }, { "epoch": 0.11, "learning_rate": 0.00019661761979754166, "loss": 0.172, "step": 1724 }, { "epoch": 0.11, "learning_rate": 0.00019661226232814088, "loss": 0.1647, "step": 1725 }, { "epoch": 0.11, "learning_rate": 0.00019660690069226328, "loss": 0.1523, "step": 1726 }, { "epoch": 0.11, "learning_rate": 0.00019660153489014018, "loss": 0.1722, "step": 1727 }, { "epoch": 0.11, "learning_rate": 0.00019659616492200295, "loss": 0.1739, "step": 1728 }, { "epoch": 0.11, "learning_rate": 0.00019659079078808317, "loss": 0.1577, "step": 1729 }, { "epoch": 0.11, "learning_rate": 0.00019658541248861256, "loss": 0.179, "step": 1730 }, { "epoch": 0.11, "learning_rate": 0.00019658003002382315, "loss": 0.153, "step": 1731 }, { "epoch": 0.11, "learning_rate": 0.000196574643393947, "loss": 0.1496, "step": 1732 }, { "epoch": 0.11, "learning_rate": 0.00019656925259921643, "loss": 0.1558, "step": 1733 }, { "epoch": 0.11, "learning_rate": 0.00019656385763986393, "loss": 0.1605, "step": 1734 }, { "epoch": 0.11, "learning_rate": 0.00019655845851612218, "loss": 0.1643, "step": 1735 }, { "epoch": 0.11, "learning_rate": 0.00019655305522822399, "loss": 0.1698, "step": 1736 }, { "epoch": 0.11, "learning_rate": 0.00019654764777640236, "loss": 0.1481, "step": 1737 }, { "epoch": 0.11, "learning_rate": 0.00019654223616089053, "loss": 0.1515, "step": 1738 }, { "epoch": 0.11, "learning_rate": 0.00019653682038192188, "loss": 0.1679, "step": 1739 }, { "epoch": 0.11, "learning_rate": 0.00019653140043972994, "loss": 0.1527, "step": 1740 }, { "epoch": 0.11, "learning_rate": 0.00019652597633454847, "loss": 0.171, "step": 1741 }, { "epoch": 0.11, "learning_rate": 0.00019652054806661138, "loss": 0.1595, "step": 1742 }, { "epoch": 0.11, "learning_rate": 0.00019651511563615278, "loss": 0.1649, "step": 1743 }, { "epoch": 0.11, "learning_rate": 0.00019650967904340694, "loss": 0.1776, "step": 1744 }, { "epoch": 0.11, "learning_rate": 0.00019650423828860834, "loss": 0.1528, "step": 1745 }, { "epoch": 0.11, "learning_rate": 0.00019649879337199156, "loss": 0.1343, "step": 1746 }, { "epoch": 0.11, "learning_rate": 0.00019649334429379148, "loss": 0.1506, "step": 1747 }, { "epoch": 0.11, "learning_rate": 0.00019648789105424305, "loss": 0.1547, "step": 1748 }, { "epoch": 0.11, "learning_rate": 0.00019648243365358146, "loss": 0.1503, "step": 1749 }, { "epoch": 0.11, "learning_rate": 0.00019647697209204207, "loss": 0.1474, "step": 1750 }, { "epoch": 0.11, "learning_rate": 0.0001964715063698604, "loss": 0.1583, "step": 1751 }, { "epoch": 0.11, "learning_rate": 0.00019646603648727223, "loss": 0.1485, "step": 1752 }, { "epoch": 0.11, "learning_rate": 0.00019646056244451335, "loss": 0.1366, "step": 1753 }, { "epoch": 0.11, "learning_rate": 0.00019645508424181987, "loss": 0.1781, "step": 1754 }, { "epoch": 0.11, "learning_rate": 0.00019644960187942807, "loss": 0.158, "step": 1755 }, { "epoch": 0.11, "learning_rate": 0.00019644411535757437, "loss": 0.1738, "step": 1756 }, { "epoch": 0.11, "learning_rate": 0.00019643862467649535, "loss": 0.1804, "step": 1757 }, { "epoch": 0.11, "learning_rate": 0.00019643312983642783, "loss": 0.1414, "step": 1758 }, { "epoch": 0.11, "learning_rate": 0.00019642763083760876, "loss": 0.173, "step": 1759 }, { "epoch": 0.11, "learning_rate": 0.0001964221276802753, "loss": 0.1459, "step": 1760 }, { "epoch": 0.11, "learning_rate": 0.0001964166203646648, "loss": 0.1665, "step": 1761 }, { "epoch": 0.11, "learning_rate": 0.00019641110889101474, "loss": 0.1708, "step": 1762 }, { "epoch": 0.11, "learning_rate": 0.0001964055932595628, "loss": 0.1636, "step": 1763 }, { "epoch": 0.11, "learning_rate": 0.00019640007347054684, "loss": 0.1659, "step": 1764 }, { "epoch": 0.11, "learning_rate": 0.00019639454952420494, "loss": 0.178, "step": 1765 }, { "epoch": 0.11, "learning_rate": 0.0001963890214207753, "loss": 0.1392, "step": 1766 }, { "epoch": 0.11, "learning_rate": 0.00019638348916049635, "loss": 0.1733, "step": 1767 }, { "epoch": 0.11, "learning_rate": 0.00019637795274360663, "loss": 0.1752, "step": 1768 }, { "epoch": 0.11, "learning_rate": 0.00019637241217034492, "loss": 0.1647, "step": 1769 }, { "epoch": 0.11, "learning_rate": 0.0001963668674409502, "loss": 0.1532, "step": 1770 }, { "epoch": 0.11, "learning_rate": 0.00019636131855566152, "loss": 0.1556, "step": 1771 }, { "epoch": 0.11, "learning_rate": 0.0001963557655147182, "loss": 0.1642, "step": 1772 }, { "epoch": 0.11, "learning_rate": 0.0001963502083183598, "loss": 0.1638, "step": 1773 }, { "epoch": 0.11, "learning_rate": 0.00019634464696682586, "loss": 0.1417, "step": 1774 }, { "epoch": 0.11, "learning_rate": 0.0001963390814603563, "loss": 0.1619, "step": 1775 }, { "epoch": 0.11, "learning_rate": 0.00019633351179919109, "loss": 0.152, "step": 1776 }, { "epoch": 0.11, "learning_rate": 0.00019632793798357045, "loss": 0.1641, "step": 1777 }, { "epoch": 0.11, "learning_rate": 0.00019632236001373477, "loss": 0.1398, "step": 1778 }, { "epoch": 0.11, "learning_rate": 0.00019631677788992455, "loss": 0.158, "step": 1779 }, { "epoch": 0.11, "learning_rate": 0.0001963111916123806, "loss": 0.1715, "step": 1780 }, { "epoch": 0.11, "learning_rate": 0.00019630560118134375, "loss": 0.1572, "step": 1781 }, { "epoch": 0.11, "learning_rate": 0.00019630000659705512, "loss": 0.1927, "step": 1782 }, { "epoch": 0.11, "learning_rate": 0.00019629440785975602, "loss": 0.1659, "step": 1783 }, { "epoch": 0.11, "learning_rate": 0.00019628880496968786, "loss": 0.156, "step": 1784 }, { "epoch": 0.11, "learning_rate": 0.00019628319792709225, "loss": 0.1683, "step": 1785 }, { "epoch": 0.11, "learning_rate": 0.0001962775867322111, "loss": 0.1653, "step": 1786 }, { "epoch": 0.11, "learning_rate": 0.00019627197138528626, "loss": 0.1526, "step": 1787 }, { "epoch": 0.11, "learning_rate": 0.00019626635188656, "loss": 0.167, "step": 1788 }, { "epoch": 0.11, "learning_rate": 0.00019626072823627458, "loss": 0.1448, "step": 1789 }, { "epoch": 0.11, "learning_rate": 0.00019625510043467262, "loss": 0.1541, "step": 1790 }, { "epoch": 0.11, "learning_rate": 0.00019624946848199675, "loss": 0.168, "step": 1791 }, { "epoch": 0.11, "learning_rate": 0.00019624383237848986, "loss": 0.1461, "step": 1792 }, { "epoch": 0.11, "learning_rate": 0.00019623819212439506, "loss": 0.1386, "step": 1793 }, { "epoch": 0.12, "learning_rate": 0.00019623254771995553, "loss": 0.146, "step": 1794 }, { "epoch": 0.12, "learning_rate": 0.00019622689916541473, "loss": 0.152, "step": 1795 }, { "epoch": 0.12, "learning_rate": 0.00019622124646101626, "loss": 0.1735, "step": 1796 }, { "epoch": 0.12, "learning_rate": 0.00019621558960700386, "loss": 0.164, "step": 1797 }, { "epoch": 0.12, "learning_rate": 0.00019620992860362152, "loss": 0.1745, "step": 1798 }, { "epoch": 0.12, "learning_rate": 0.00019620426345111334, "loss": 0.1683, "step": 1799 }, { "epoch": 0.12, "learning_rate": 0.00019619859414972366, "loss": 0.1916, "step": 1800 }, { "epoch": 0.12, "learning_rate": 0.000196192920699697, "loss": 0.1849, "step": 1801 }, { "epoch": 0.12, "learning_rate": 0.000196187243101278, "loss": 0.1636, "step": 1802 }, { "epoch": 0.12, "learning_rate": 0.00019618156135471146, "loss": 0.165, "step": 1803 }, { "epoch": 0.12, "learning_rate": 0.0001961758754602425, "loss": 0.1545, "step": 1804 }, { "epoch": 0.12, "learning_rate": 0.00019617018541811626, "loss": 0.1656, "step": 1805 }, { "epoch": 0.12, "learning_rate": 0.0001961644912285782, "loss": 0.1717, "step": 1806 }, { "epoch": 0.12, "learning_rate": 0.0001961587928918738, "loss": 0.1624, "step": 1807 }, { "epoch": 0.12, "learning_rate": 0.00019615309040824888, "loss": 0.1498, "step": 1808 }, { "epoch": 0.12, "learning_rate": 0.0001961473837779493, "loss": 0.1668, "step": 1809 }, { "epoch": 0.12, "learning_rate": 0.00019614167300122126, "loss": 0.1525, "step": 1810 }, { "epoch": 0.12, "learning_rate": 0.00019613595807831095, "loss": 0.1721, "step": 1811 }, { "epoch": 0.12, "learning_rate": 0.00019613023900946484, "loss": 0.1609, "step": 1812 }, { "epoch": 0.12, "learning_rate": 0.00019612451579492958, "loss": 0.1491, "step": 1813 }, { "epoch": 0.12, "learning_rate": 0.00019611878843495202, "loss": 0.1737, "step": 1814 }, { "epoch": 0.12, "learning_rate": 0.0001961130569297791, "loss": 0.1676, "step": 1815 }, { "epoch": 0.12, "learning_rate": 0.00019610732127965804, "loss": 0.1691, "step": 1816 }, { "epoch": 0.12, "learning_rate": 0.0001961015814848362, "loss": 0.1706, "step": 1817 }, { "epoch": 0.12, "learning_rate": 0.00019609583754556106, "loss": 0.1576, "step": 1818 }, { "epoch": 0.12, "learning_rate": 0.00019609008946208038, "loss": 0.1572, "step": 1819 }, { "epoch": 0.12, "learning_rate": 0.00019608433723464206, "loss": 0.1693, "step": 1820 }, { "epoch": 0.12, "learning_rate": 0.0001960785808634941, "loss": 0.1585, "step": 1821 }, { "epoch": 0.12, "learning_rate": 0.00019607282034888484, "loss": 0.1803, "step": 1822 }, { "epoch": 0.12, "learning_rate": 0.00019606705569106262, "loss": 0.1715, "step": 1823 }, { "epoch": 0.12, "learning_rate": 0.0001960612868902761, "loss": 0.1957, "step": 1824 }, { "epoch": 0.12, "learning_rate": 0.00019605551394677407, "loss": 0.1963, "step": 1825 }, { "epoch": 0.12, "learning_rate": 0.00019604973686080542, "loss": 0.1596, "step": 1826 }, { "epoch": 0.12, "learning_rate": 0.0001960439556326194, "loss": 0.1523, "step": 1827 }, { "epoch": 0.12, "learning_rate": 0.00019603817026246523, "loss": 0.1957, "step": 1828 }, { "epoch": 0.12, "learning_rate": 0.00019603238075059247, "loss": 0.1726, "step": 1829 }, { "epoch": 0.12, "learning_rate": 0.00019602658709725078, "loss": 0.1572, "step": 1830 }, { "epoch": 0.12, "learning_rate": 0.00019602078930269, "loss": 0.1603, "step": 1831 }, { "epoch": 0.12, "learning_rate": 0.00019601498736716017, "loss": 0.1698, "step": 1832 }, { "epoch": 0.12, "learning_rate": 0.00019600918129091154, "loss": 0.1675, "step": 1833 }, { "epoch": 0.12, "learning_rate": 0.00019600337107419447, "loss": 0.164, "step": 1834 }, { "epoch": 0.12, "learning_rate": 0.00019599755671725953, "loss": 0.1262, "step": 1835 }, { "epoch": 0.12, "learning_rate": 0.00019599173822035744, "loss": 0.1513, "step": 1836 }, { "epoch": 0.12, "learning_rate": 0.00019598591558373917, "loss": 0.15, "step": 1837 }, { "epoch": 0.12, "learning_rate": 0.0001959800888076558, "loss": 0.154, "step": 1838 }, { "epoch": 0.12, "learning_rate": 0.0001959742578923587, "loss": 0.1529, "step": 1839 }, { "epoch": 0.12, "learning_rate": 0.00019596842283809917, "loss": 0.1395, "step": 1840 }, { "epoch": 0.12, "learning_rate": 0.000195962583645129, "loss": 0.1509, "step": 1841 }, { "epoch": 0.12, "learning_rate": 0.00019595674031369992, "loss": 0.174, "step": 1842 }, { "epoch": 0.12, "learning_rate": 0.00019595089284406395, "loss": 0.1764, "step": 1843 }, { "epoch": 0.12, "learning_rate": 0.0001959450412364733, "loss": 0.1596, "step": 1844 }, { "epoch": 0.12, "learning_rate": 0.00019593918549118029, "loss": 0.1626, "step": 1845 }, { "epoch": 0.12, "learning_rate": 0.00019593332560843748, "loss": 0.1838, "step": 1846 }, { "epoch": 0.12, "learning_rate": 0.00019592746158849752, "loss": 0.1579, "step": 1847 }, { "epoch": 0.12, "learning_rate": 0.00019592159343161342, "loss": 0.1504, "step": 1848 }, { "epoch": 0.12, "learning_rate": 0.0001959157211380381, "loss": 0.1699, "step": 1849 }, { "epoch": 0.12, "learning_rate": 0.00019590984470802493, "loss": 0.1802, "step": 1850 }, { "epoch": 0.12, "learning_rate": 0.00019590396414182726, "loss": 0.146, "step": 1851 }, { "epoch": 0.12, "learning_rate": 0.00019589807943969873, "loss": 0.1612, "step": 1852 }, { "epoch": 0.12, "learning_rate": 0.00019589219060189313, "loss": 0.1504, "step": 1853 }, { "epoch": 0.12, "learning_rate": 0.0001958862976286644, "loss": 0.162, "step": 1854 }, { "epoch": 0.12, "learning_rate": 0.00019588040052026667, "loss": 0.1601, "step": 1855 }, { "epoch": 0.12, "learning_rate": 0.00019587449927695424, "loss": 0.1658, "step": 1856 }, { "epoch": 0.12, "learning_rate": 0.00019586859389898167, "loss": 0.1708, "step": 1857 }, { "epoch": 0.12, "learning_rate": 0.0001958626843866036, "loss": 0.1917, "step": 1858 }, { "epoch": 0.12, "learning_rate": 0.00019585677074007487, "loss": 0.1448, "step": 1859 }, { "epoch": 0.12, "learning_rate": 0.00019585085295965055, "loss": 0.1875, "step": 1860 }, { "epoch": 0.12, "learning_rate": 0.0001958449310455858, "loss": 0.1802, "step": 1861 }, { "epoch": 0.12, "learning_rate": 0.000195839004998136, "loss": 0.1568, "step": 1862 }, { "epoch": 0.12, "learning_rate": 0.0001958330748175568, "loss": 0.159, "step": 1863 }, { "epoch": 0.12, "learning_rate": 0.00019582714050410386, "loss": 0.1758, "step": 1864 }, { "epoch": 0.12, "learning_rate": 0.0001958212020580331, "loss": 0.1966, "step": 1865 }, { "epoch": 0.12, "learning_rate": 0.00019581525947960068, "loss": 0.1932, "step": 1866 }, { "epoch": 0.12, "learning_rate": 0.00019580931276906284, "loss": 0.1461, "step": 1867 }, { "epoch": 0.12, "learning_rate": 0.00019580336192667606, "loss": 0.1777, "step": 1868 }, { "epoch": 0.12, "learning_rate": 0.00019579740695269693, "loss": 0.1469, "step": 1869 }, { "epoch": 0.12, "learning_rate": 0.0001957914478473823, "loss": 0.155, "step": 1870 }, { "epoch": 0.12, "learning_rate": 0.00019578548461098914, "loss": 0.1732, "step": 1871 }, { "epoch": 0.12, "learning_rate": 0.00019577951724377462, "loss": 0.1745, "step": 1872 }, { "epoch": 0.12, "learning_rate": 0.00019577354574599616, "loss": 0.17, "step": 1873 }, { "epoch": 0.12, "learning_rate": 0.00019576757011791117, "loss": 0.1717, "step": 1874 }, { "epoch": 0.12, "learning_rate": 0.00019576159035977742, "loss": 0.1553, "step": 1875 }, { "epoch": 0.12, "learning_rate": 0.00019575560647185278, "loss": 0.1598, "step": 1876 }, { "epoch": 0.12, "learning_rate": 0.0001957496184543953, "loss": 0.146, "step": 1877 }, { "epoch": 0.12, "learning_rate": 0.00019574362630766322, "loss": 0.1912, "step": 1878 }, { "epoch": 0.12, "learning_rate": 0.00019573763003191496, "loss": 0.1599, "step": 1879 }, { "epoch": 0.12, "learning_rate": 0.00019573162962740914, "loss": 0.1651, "step": 1880 }, { "epoch": 0.12, "learning_rate": 0.0001957256250944045, "loss": 0.1442, "step": 1881 }, { "epoch": 0.12, "learning_rate": 0.00019571961643315996, "loss": 0.162, "step": 1882 }, { "epoch": 0.12, "learning_rate": 0.00019571360364393475, "loss": 0.195, "step": 1883 }, { "epoch": 0.12, "learning_rate": 0.00019570758672698805, "loss": 0.1366, "step": 1884 }, { "epoch": 0.12, "learning_rate": 0.00019570156568257945, "loss": 0.1532, "step": 1885 }, { "epoch": 0.12, "learning_rate": 0.00019569554051096852, "loss": 0.1491, "step": 1886 }, { "epoch": 0.12, "learning_rate": 0.00019568951121241518, "loss": 0.1359, "step": 1887 }, { "epoch": 0.12, "learning_rate": 0.0001956834777871794, "loss": 0.1753, "step": 1888 }, { "epoch": 0.12, "learning_rate": 0.0001956774402355214, "loss": 0.1413, "step": 1889 }, { "epoch": 0.12, "learning_rate": 0.0001956713985577016, "loss": 0.1884, "step": 1890 }, { "epoch": 0.12, "learning_rate": 0.0001956653527539804, "loss": 0.1623, "step": 1891 }, { "epoch": 0.12, "learning_rate": 0.00019565930282461867, "loss": 0.1658, "step": 1892 }, { "epoch": 0.12, "learning_rate": 0.00019565324876987726, "loss": 0.1631, "step": 1893 }, { "epoch": 0.12, "learning_rate": 0.00019564719059001734, "loss": 0.1548, "step": 1894 }, { "epoch": 0.12, "learning_rate": 0.00019564112828530003, "loss": 0.1914, "step": 1895 }, { "epoch": 0.12, "learning_rate": 0.00019563506185598685, "loss": 0.1884, "step": 1896 }, { "epoch": 0.12, "learning_rate": 0.00019562899130233941, "loss": 0.1604, "step": 1897 }, { "epoch": 0.12, "learning_rate": 0.00019562291662461956, "loss": 0.2034, "step": 1898 }, { "epoch": 0.12, "learning_rate": 0.0001956168378230892, "loss": 0.1831, "step": 1899 }, { "epoch": 0.12, "learning_rate": 0.00019561075489801054, "loss": 0.1335, "step": 1900 }, { "epoch": 0.12, "learning_rate": 0.00019560466784964583, "loss": 0.175, "step": 1901 }, { "epoch": 0.12, "learning_rate": 0.00019559857667825764, "loss": 0.162, "step": 1902 }, { "epoch": 0.12, "learning_rate": 0.00019559248138410865, "loss": 0.1554, "step": 1903 }, { "epoch": 0.12, "learning_rate": 0.00019558638196746176, "loss": 0.1711, "step": 1904 }, { "epoch": 0.12, "learning_rate": 0.00019558027842857994, "loss": 0.1638, "step": 1905 }, { "epoch": 0.12, "learning_rate": 0.0001955741707677264, "loss": 0.1624, "step": 1906 }, { "epoch": 0.12, "learning_rate": 0.00019556805898516465, "loss": 0.1925, "step": 1907 }, { "epoch": 0.12, "learning_rate": 0.00019556194308115816, "loss": 0.1693, "step": 1908 }, { "epoch": 0.12, "learning_rate": 0.00019555582305597074, "loss": 0.1518, "step": 1909 }, { "epoch": 0.12, "learning_rate": 0.00019554969890986628, "loss": 0.1568, "step": 1910 }, { "epoch": 0.12, "learning_rate": 0.0001955435706431089, "loss": 0.1913, "step": 1911 }, { "epoch": 0.12, "learning_rate": 0.0001955374382559629, "loss": 0.1781, "step": 1912 }, { "epoch": 0.12, "learning_rate": 0.00019553130174869275, "loss": 0.1432, "step": 1913 }, { "epoch": 0.12, "learning_rate": 0.00019552516112156305, "loss": 0.1439, "step": 1914 }, { "epoch": 0.12, "learning_rate": 0.00019551901637483864, "loss": 0.1781, "step": 1915 }, { "epoch": 0.12, "learning_rate": 0.00019551286750878454, "loss": 0.1833, "step": 1916 }, { "epoch": 0.12, "learning_rate": 0.00019550671452366592, "loss": 0.1408, "step": 1917 }, { "epoch": 0.12, "learning_rate": 0.00019550055741974814, "loss": 0.1562, "step": 1918 }, { "epoch": 0.12, "learning_rate": 0.00019549439619729667, "loss": 0.1578, "step": 1919 }, { "epoch": 0.12, "learning_rate": 0.00019548823085657727, "loss": 0.1685, "step": 1920 }, { "epoch": 0.12, "learning_rate": 0.00019548206139785582, "loss": 0.1516, "step": 1921 }, { "epoch": 0.12, "learning_rate": 0.0001954758878213984, "loss": 0.1603, "step": 1922 }, { "epoch": 0.12, "learning_rate": 0.00019546971012747118, "loss": 0.137, "step": 1923 }, { "epoch": 0.12, "learning_rate": 0.00019546352831634062, "loss": 0.1625, "step": 1924 }, { "epoch": 0.12, "learning_rate": 0.00019545734238827334, "loss": 0.1636, "step": 1925 }, { "epoch": 0.12, "learning_rate": 0.0001954511523435361, "loss": 0.1794, "step": 1926 }, { "epoch": 0.12, "learning_rate": 0.00019544495818239583, "loss": 0.1537, "step": 1927 }, { "epoch": 0.12, "learning_rate": 0.00019543875990511969, "loss": 0.1532, "step": 1928 }, { "epoch": 0.12, "learning_rate": 0.00019543255751197492, "loss": 0.167, "step": 1929 }, { "epoch": 0.12, "learning_rate": 0.00019542635100322907, "loss": 0.1562, "step": 1930 }, { "epoch": 0.12, "learning_rate": 0.0001954201403791498, "loss": 0.1624, "step": 1931 }, { "epoch": 0.12, "learning_rate": 0.00019541392564000488, "loss": 0.1692, "step": 1932 }, { "epoch": 0.12, "learning_rate": 0.0001954077067860624, "loss": 0.161, "step": 1933 }, { "epoch": 0.12, "learning_rate": 0.00019540148381759053, "loss": 0.1836, "step": 1934 }, { "epoch": 0.12, "learning_rate": 0.00019539525673485764, "loss": 0.1394, "step": 1935 }, { "epoch": 0.12, "learning_rate": 0.00019538902553813226, "loss": 0.1452, "step": 1936 }, { "epoch": 0.12, "learning_rate": 0.00019538279022768312, "loss": 0.1553, "step": 1937 }, { "epoch": 0.12, "learning_rate": 0.00019537655080377911, "loss": 0.1436, "step": 1938 }, { "epoch": 0.12, "learning_rate": 0.00019537030726668938, "loss": 0.1555, "step": 1939 }, { "epoch": 0.12, "learning_rate": 0.00019536405961668313, "loss": 0.1585, "step": 1940 }, { "epoch": 0.12, "learning_rate": 0.00019535780785402977, "loss": 0.1631, "step": 1941 }, { "epoch": 0.12, "learning_rate": 0.00019535155197899896, "loss": 0.1476, "step": 1942 }, { "epoch": 0.12, "learning_rate": 0.00019534529199186044, "loss": 0.1627, "step": 1943 }, { "epoch": 0.12, "learning_rate": 0.00019533902789288426, "loss": 0.1588, "step": 1944 }, { "epoch": 0.12, "learning_rate": 0.0001953327596823405, "loss": 0.1545, "step": 1945 }, { "epoch": 0.12, "learning_rate": 0.00019532648736049946, "loss": 0.1582, "step": 1946 }, { "epoch": 0.12, "learning_rate": 0.00019532021092763172, "loss": 0.1462, "step": 1947 }, { "epoch": 0.12, "learning_rate": 0.00019531393038400786, "loss": 0.1544, "step": 1948 }, { "epoch": 0.12, "learning_rate": 0.00019530764572989884, "loss": 0.1464, "step": 1949 }, { "epoch": 0.13, "learning_rate": 0.00019530135696557558, "loss": 0.1529, "step": 1950 }, { "epoch": 0.13, "learning_rate": 0.0001952950640913094, "loss": 0.1706, "step": 1951 }, { "epoch": 0.13, "learning_rate": 0.00019528876710737155, "loss": 0.17, "step": 1952 }, { "epoch": 0.13, "learning_rate": 0.00019528246601403372, "loss": 0.1605, "step": 1953 }, { "epoch": 0.13, "learning_rate": 0.0001952761608115676, "loss": 0.1568, "step": 1954 }, { "epoch": 0.13, "learning_rate": 0.00019526985150024507, "loss": 0.1719, "step": 1955 }, { "epoch": 0.13, "learning_rate": 0.00019526353808033825, "loss": 0.1716, "step": 1956 }, { "epoch": 0.13, "learning_rate": 0.00019525722055211946, "loss": 0.1626, "step": 1957 }, { "epoch": 0.13, "learning_rate": 0.00019525089891586107, "loss": 0.1522, "step": 1958 }, { "epoch": 0.13, "learning_rate": 0.00019524457317183578, "loss": 0.1765, "step": 1959 }, { "epoch": 0.13, "learning_rate": 0.00019523824332031632, "loss": 0.1591, "step": 1960 }, { "epoch": 0.13, "learning_rate": 0.0001952319093615757, "loss": 0.1488, "step": 1961 }, { "epoch": 0.13, "learning_rate": 0.00019522557129588713, "loss": 0.1733, "step": 1962 }, { "epoch": 0.13, "learning_rate": 0.00019521922912352387, "loss": 0.1674, "step": 1963 }, { "epoch": 0.13, "learning_rate": 0.00019521288284475943, "loss": 0.1366, "step": 1964 }, { "epoch": 0.13, "learning_rate": 0.00019520653245986755, "loss": 0.1451, "step": 1965 }, { "epoch": 0.13, "learning_rate": 0.00019520017796912208, "loss": 0.1753, "step": 1966 }, { "epoch": 0.13, "learning_rate": 0.00019519381937279704, "loss": 0.1525, "step": 1967 }, { "epoch": 0.13, "learning_rate": 0.00019518745667116663, "loss": 0.1726, "step": 1968 }, { "epoch": 0.13, "learning_rate": 0.00019518108986450532, "loss": 0.1661, "step": 1969 }, { "epoch": 0.13, "learning_rate": 0.0001951747189530876, "loss": 0.1655, "step": 1970 }, { "epoch": 0.13, "learning_rate": 0.0001951683439371883, "loss": 0.1604, "step": 1971 }, { "epoch": 0.13, "learning_rate": 0.00019516196481708228, "loss": 0.1664, "step": 1972 }, { "epoch": 0.13, "learning_rate": 0.0001951555815930447, "loss": 0.1646, "step": 1973 }, { "epoch": 0.13, "learning_rate": 0.00019514919426535075, "loss": 0.1608, "step": 1974 }, { "epoch": 0.13, "learning_rate": 0.000195142802834276, "loss": 0.1737, "step": 1975 }, { "epoch": 0.13, "learning_rate": 0.00019513640730009604, "loss": 0.198, "step": 1976 }, { "epoch": 0.13, "learning_rate": 0.0001951300076630867, "loss": 0.1554, "step": 1977 }, { "epoch": 0.13, "learning_rate": 0.00019512360392352388, "loss": 0.1688, "step": 1978 }, { "epoch": 0.13, "learning_rate": 0.00019511719608168387, "loss": 0.1479, "step": 1979 }, { "epoch": 0.13, "learning_rate": 0.00019511078413784295, "loss": 0.1533, "step": 1980 }, { "epoch": 0.13, "learning_rate": 0.00019510436809227764, "loss": 0.1659, "step": 1981 }, { "epoch": 0.13, "learning_rate": 0.00019509794794526464, "loss": 0.1697, "step": 1982 }, { "epoch": 0.13, "learning_rate": 0.00019509152369708084, "loss": 0.175, "step": 1983 }, { "epoch": 0.13, "learning_rate": 0.00019508509534800326, "loss": 0.1923, "step": 1984 }, { "epoch": 0.13, "learning_rate": 0.00019507866289830917, "loss": 0.1634, "step": 1985 }, { "epoch": 0.13, "learning_rate": 0.00019507222634827592, "loss": 0.1686, "step": 1986 }, { "epoch": 0.13, "learning_rate": 0.00019506578569818115, "loss": 0.1735, "step": 1987 }, { "epoch": 0.13, "learning_rate": 0.00019505934094830258, "loss": 0.1683, "step": 1988 }, { "epoch": 0.13, "learning_rate": 0.00019505289209891812, "loss": 0.1603, "step": 1989 }, { "epoch": 0.13, "learning_rate": 0.00019504643915030595, "loss": 0.1651, "step": 1990 }, { "epoch": 0.13, "learning_rate": 0.00019503998210274436, "loss": 0.1728, "step": 1991 }, { "epoch": 0.13, "learning_rate": 0.00019503352095651174, "loss": 0.1531, "step": 1992 }, { "epoch": 0.13, "learning_rate": 0.00019502705571188672, "loss": 0.1666, "step": 1993 }, { "epoch": 0.13, "learning_rate": 0.00019502058636914824, "loss": 0.1587, "step": 1994 }, { "epoch": 0.13, "learning_rate": 0.0001950141129285752, "loss": 0.1608, "step": 1995 }, { "epoch": 0.13, "learning_rate": 0.00019500763539044683, "loss": 0.152, "step": 1996 }, { "epoch": 0.13, "learning_rate": 0.00019500115375504237, "loss": 0.1774, "step": 1997 }, { "epoch": 0.13, "learning_rate": 0.00019499466802264147, "loss": 0.1439, "step": 1998 }, { "epoch": 0.13, "learning_rate": 0.0001949881781935238, "loss": 0.1876, "step": 1999 }, { "epoch": 0.13, "learning_rate": 0.00019498168426796919, "loss": 0.1478, "step": 2000 }, { "epoch": 0.13, "learning_rate": 0.00019497518624625774, "loss": 0.1786, "step": 2001 }, { "epoch": 0.13, "learning_rate": 0.00019496868412866966, "loss": 0.1469, "step": 2002 }, { "epoch": 0.13, "learning_rate": 0.00019496217791548537, "loss": 0.1745, "step": 2003 }, { "epoch": 0.13, "learning_rate": 0.00019495566760698547, "loss": 0.176, "step": 2004 }, { "epoch": 0.13, "learning_rate": 0.0001949491532034507, "loss": 0.1615, "step": 2005 }, { "epoch": 0.13, "learning_rate": 0.000194942634705162, "loss": 0.1616, "step": 2006 }, { "epoch": 0.13, "learning_rate": 0.00019493611211240046, "loss": 0.1571, "step": 2007 }, { "epoch": 0.13, "learning_rate": 0.00019492958542544742, "loss": 0.1498, "step": 2008 }, { "epoch": 0.13, "learning_rate": 0.00019492305464458433, "loss": 0.1375, "step": 2009 }, { "epoch": 0.13, "learning_rate": 0.00019491651977009285, "loss": 0.1656, "step": 2010 }, { "epoch": 0.13, "learning_rate": 0.00019490998080225476, "loss": 0.1519, "step": 2011 }, { "epoch": 0.13, "learning_rate": 0.00019490343774135208, "loss": 0.148, "step": 2012 }, { "epoch": 0.13, "learning_rate": 0.000194896890587667, "loss": 0.1397, "step": 2013 }, { "epoch": 0.13, "learning_rate": 0.00019489033934148186, "loss": 0.1841, "step": 2014 }, { "epoch": 0.13, "learning_rate": 0.0001948837840030792, "loss": 0.1824, "step": 2015 }, { "epoch": 0.13, "learning_rate": 0.00019487722457274168, "loss": 0.1701, "step": 2016 }, { "epoch": 0.13, "learning_rate": 0.00019487066105075223, "loss": 0.1333, "step": 2017 }, { "epoch": 0.13, "learning_rate": 0.00019486409343739389, "loss": 0.1465, "step": 2018 }, { "epoch": 0.13, "learning_rate": 0.0001948575217329499, "loss": 0.146, "step": 2019 }, { "epoch": 0.13, "learning_rate": 0.00019485094593770362, "loss": 0.1519, "step": 2020 }, { "epoch": 0.13, "learning_rate": 0.00019484436605193869, "loss": 0.1475, "step": 2021 }, { "epoch": 0.13, "learning_rate": 0.00019483778207593887, "loss": 0.1913, "step": 2022 }, { "epoch": 0.13, "learning_rate": 0.0001948311940099881, "loss": 0.1741, "step": 2023 }, { "epoch": 0.13, "learning_rate": 0.00019482460185437045, "loss": 0.1585, "step": 2024 }, { "epoch": 0.13, "learning_rate": 0.00019481800560937029, "loss": 0.1619, "step": 2025 }, { "epoch": 0.13, "learning_rate": 0.000194811405275272, "loss": 0.152, "step": 2026 }, { "epoch": 0.13, "learning_rate": 0.00019480480085236029, "loss": 0.1617, "step": 2027 }, { "epoch": 0.13, "learning_rate": 0.00019479819234091993, "loss": 0.1512, "step": 2028 }, { "epoch": 0.13, "learning_rate": 0.000194791579741236, "loss": 0.1691, "step": 2029 }, { "epoch": 0.13, "learning_rate": 0.00019478496305359356, "loss": 0.1597, "step": 2030 }, { "epoch": 0.13, "learning_rate": 0.00019477834227827804, "loss": 0.1595, "step": 2031 }, { "epoch": 0.13, "learning_rate": 0.00019477171741557494, "loss": 0.1535, "step": 2032 }, { "epoch": 0.13, "learning_rate": 0.00019476508846576998, "loss": 0.1586, "step": 2033 }, { "epoch": 0.13, "learning_rate": 0.00019475845542914902, "loss": 0.1381, "step": 2034 }, { "epoch": 0.13, "learning_rate": 0.0001947518183059981, "loss": 0.1486, "step": 2035 }, { "epoch": 0.13, "learning_rate": 0.00019474517709660346, "loss": 0.1288, "step": 2036 }, { "epoch": 0.13, "learning_rate": 0.00019473853180125157, "loss": 0.1655, "step": 2037 }, { "epoch": 0.13, "learning_rate": 0.00019473188242022892, "loss": 0.1765, "step": 2038 }, { "epoch": 0.13, "learning_rate": 0.00019472522895382232, "loss": 0.1822, "step": 2039 }, { "epoch": 0.13, "learning_rate": 0.00019471857140231872, "loss": 0.1714, "step": 2040 }, { "epoch": 0.13, "learning_rate": 0.00019471190976600515, "loss": 0.1745, "step": 2041 }, { "epoch": 0.13, "learning_rate": 0.00019470524404516902, "loss": 0.1523, "step": 2042 }, { "epoch": 0.13, "learning_rate": 0.0001946985742400977, "loss": 0.1586, "step": 2043 }, { "epoch": 0.13, "learning_rate": 0.00019469190035107886, "loss": 0.147, "step": 2044 }, { "epoch": 0.13, "learning_rate": 0.00019468522237840033, "loss": 0.1596, "step": 2045 }, { "epoch": 0.13, "learning_rate": 0.00019467854032235006, "loss": 0.1575, "step": 2046 }, { "epoch": 0.13, "learning_rate": 0.00019467185418321626, "loss": 0.1932, "step": 2047 }, { "epoch": 0.13, "learning_rate": 0.00019466516396128726, "loss": 0.1697, "step": 2048 }, { "epoch": 0.13, "learning_rate": 0.00019465846965685158, "loss": 0.172, "step": 2049 }, { "epoch": 0.13, "learning_rate": 0.00019465177127019797, "loss": 0.1451, "step": 2050 }, { "epoch": 0.13, "learning_rate": 0.00019464506880161522, "loss": 0.163, "step": 2051 }, { "epoch": 0.13, "learning_rate": 0.0001946383622513924, "loss": 0.1664, "step": 2052 }, { "epoch": 0.13, "learning_rate": 0.00019463165161981877, "loss": 0.1784, "step": 2053 }, { "epoch": 0.13, "learning_rate": 0.0001946249369071837, "loss": 0.1625, "step": 2054 }, { "epoch": 0.13, "learning_rate": 0.0001946182181137768, "loss": 0.1589, "step": 2055 }, { "epoch": 0.13, "learning_rate": 0.00019461149523988772, "loss": 0.1755, "step": 2056 }, { "epoch": 0.13, "learning_rate": 0.00019460476828580655, "loss": 0.1601, "step": 2057 }, { "epoch": 0.13, "learning_rate": 0.00019459803725182327, "loss": 0.1636, "step": 2058 }, { "epoch": 0.13, "learning_rate": 0.00019459130213822824, "loss": 0.166, "step": 2059 }, { "epoch": 0.13, "learning_rate": 0.00019458456294531185, "loss": 0.1812, "step": 2060 }, { "epoch": 0.13, "learning_rate": 0.0001945778196733648, "loss": 0.1327, "step": 2061 }, { "epoch": 0.13, "learning_rate": 0.0001945710723226778, "loss": 0.1724, "step": 2062 }, { "epoch": 0.13, "learning_rate": 0.00019456432089354195, "loss": 0.1819, "step": 2063 }, { "epoch": 0.13, "learning_rate": 0.00019455756538624833, "loss": 0.1543, "step": 2064 }, { "epoch": 0.13, "learning_rate": 0.00019455080580108833, "loss": 0.2122, "step": 2065 }, { "epoch": 0.13, "learning_rate": 0.00019454404213835343, "loss": 0.1343, "step": 2066 }, { "epoch": 0.13, "learning_rate": 0.00019453727439833532, "loss": 0.1703, "step": 2067 }, { "epoch": 0.13, "learning_rate": 0.00019453050258132586, "loss": 0.1704, "step": 2068 }, { "epoch": 0.13, "learning_rate": 0.00019452372668761708, "loss": 0.1616, "step": 2069 }, { "epoch": 0.13, "learning_rate": 0.00019451694671750126, "loss": 0.1774, "step": 2070 }, { "epoch": 0.13, "learning_rate": 0.00019451016267127072, "loss": 0.1641, "step": 2071 }, { "epoch": 0.13, "learning_rate": 0.00019450337454921806, "loss": 0.1771, "step": 2072 }, { "epoch": 0.13, "learning_rate": 0.000194496582351636, "loss": 0.1357, "step": 2073 }, { "epoch": 0.13, "learning_rate": 0.00019448978607881746, "loss": 0.1575, "step": 2074 }, { "epoch": 0.13, "learning_rate": 0.0001944829857310556, "loss": 0.1637, "step": 2075 }, { "epoch": 0.13, "learning_rate": 0.00019447618130864357, "loss": 0.1618, "step": 2076 }, { "epoch": 0.13, "learning_rate": 0.00019446937281187494, "loss": 0.1792, "step": 2077 }, { "epoch": 0.13, "learning_rate": 0.00019446256024104326, "loss": 0.1993, "step": 2078 }, { "epoch": 0.13, "learning_rate": 0.00019445574359644232, "loss": 0.1613, "step": 2079 }, { "epoch": 0.13, "learning_rate": 0.00019444892287836613, "loss": 0.186, "step": 2080 }, { "epoch": 0.13, "learning_rate": 0.00019444209808710885, "loss": 0.1671, "step": 2081 }, { "epoch": 0.13, "learning_rate": 0.00019443526922296474, "loss": 0.1521, "step": 2082 }, { "epoch": 0.13, "learning_rate": 0.00019442843628622835, "loss": 0.146, "step": 2083 }, { "epoch": 0.13, "learning_rate": 0.00019442159927719434, "loss": 0.1494, "step": 2084 }, { "epoch": 0.13, "learning_rate": 0.00019441475819615756, "loss": 0.1405, "step": 2085 }, { "epoch": 0.13, "learning_rate": 0.00019440791304341305, "loss": 0.151, "step": 2086 }, { "epoch": 0.13, "learning_rate": 0.00019440106381925603, "loss": 0.1761, "step": 2087 }, { "epoch": 0.13, "learning_rate": 0.00019439421052398182, "loss": 0.16, "step": 2088 }, { "epoch": 0.13, "learning_rate": 0.000194387353157886, "loss": 0.1658, "step": 2089 }, { "epoch": 0.13, "learning_rate": 0.00019438049172126434, "loss": 0.1438, "step": 2090 }, { "epoch": 0.13, "learning_rate": 0.00019437362621441267, "loss": 0.1575, "step": 2091 }, { "epoch": 0.13, "learning_rate": 0.00019436675663762716, "loss": 0.1724, "step": 2092 }, { "epoch": 0.13, "learning_rate": 0.00019435988299120399, "loss": 0.1744, "step": 2093 }, { "epoch": 0.13, "learning_rate": 0.0001943530052754396, "loss": 0.1504, "step": 2094 }, { "epoch": 0.13, "learning_rate": 0.00019434612349063062, "loss": 0.1712, "step": 2095 }, { "epoch": 0.13, "learning_rate": 0.00019433923763707383, "loss": 0.146, "step": 2096 }, { "epoch": 0.13, "learning_rate": 0.00019433234771506618, "loss": 0.1451, "step": 2097 }, { "epoch": 0.13, "learning_rate": 0.0001943254537249048, "loss": 0.1561, "step": 2098 }, { "epoch": 0.13, "learning_rate": 0.00019431855566688703, "loss": 0.176, "step": 2099 }, { "epoch": 0.13, "learning_rate": 0.0001943116535413103, "loss": 0.1735, "step": 2100 }, { "epoch": 0.13, "learning_rate": 0.0001943047473484723, "loss": 0.1427, "step": 2101 }, { "epoch": 0.13, "learning_rate": 0.00019429783708867088, "loss": 0.1706, "step": 2102 }, { "epoch": 0.13, "learning_rate": 0.000194290922762204, "loss": 0.1536, "step": 2103 }, { "epoch": 0.13, "learning_rate": 0.00019428400436936994, "loss": 0.1486, "step": 2104 }, { "epoch": 0.13, "learning_rate": 0.00019427708191046696, "loss": 0.1752, "step": 2105 }, { "epoch": 0.14, "learning_rate": 0.00019427015538579366, "loss": 0.1688, "step": 2106 }, { "epoch": 0.14, "learning_rate": 0.00019426322479564867, "loss": 0.1785, "step": 2107 }, { "epoch": 0.14, "learning_rate": 0.00019425629014033097, "loss": 0.1612, "step": 2108 }, { "epoch": 0.14, "learning_rate": 0.00019424935142013961, "loss": 0.1601, "step": 2109 }, { "epoch": 0.14, "learning_rate": 0.00019424240863537376, "loss": 0.1664, "step": 2110 }, { "epoch": 0.14, "learning_rate": 0.0001942354617863329, "loss": 0.1772, "step": 2111 }, { "epoch": 0.14, "learning_rate": 0.0001942285108733166, "loss": 0.1593, "step": 2112 }, { "epoch": 0.14, "learning_rate": 0.0001942215558966246, "loss": 0.1576, "step": 2113 }, { "epoch": 0.14, "learning_rate": 0.00019421459685655686, "loss": 0.1479, "step": 2114 }, { "epoch": 0.14, "learning_rate": 0.0001942076337534135, "loss": 0.1556, "step": 2115 }, { "epoch": 0.14, "learning_rate": 0.00019420066658749481, "loss": 0.1702, "step": 2116 }, { "epoch": 0.14, "learning_rate": 0.00019419369535910122, "loss": 0.1852, "step": 2117 }, { "epoch": 0.14, "learning_rate": 0.0001941867200685334, "loss": 0.1449, "step": 2118 }, { "epoch": 0.14, "learning_rate": 0.00019417974071609217, "loss": 0.1564, "step": 2119 }, { "epoch": 0.14, "learning_rate": 0.0001941727573020785, "loss": 0.1542, "step": 2120 }, { "epoch": 0.14, "learning_rate": 0.0001941657698267936, "loss": 0.1363, "step": 2121 }, { "epoch": 0.14, "learning_rate": 0.0001941587782905387, "loss": 0.1497, "step": 2122 }, { "epoch": 0.14, "learning_rate": 0.00019415178269361542, "loss": 0.1906, "step": 2123 }, { "epoch": 0.14, "learning_rate": 0.00019414478303632544, "loss": 0.1807, "step": 2124 }, { "epoch": 0.14, "learning_rate": 0.00019413777931897056, "loss": 0.1643, "step": 2125 }, { "epoch": 0.14, "learning_rate": 0.0001941307715418529, "loss": 0.1417, "step": 2126 }, { "epoch": 0.14, "learning_rate": 0.00019412375970527467, "loss": 0.1566, "step": 2127 }, { "epoch": 0.14, "learning_rate": 0.00019411674380953817, "loss": 0.1738, "step": 2128 }, { "epoch": 0.14, "learning_rate": 0.00019410972385494607, "loss": 0.1774, "step": 2129 }, { "epoch": 0.14, "learning_rate": 0.00019410269984180105, "loss": 0.1718, "step": 2130 }, { "epoch": 0.14, "learning_rate": 0.00019409567177040602, "loss": 0.1679, "step": 2131 }, { "epoch": 0.14, "learning_rate": 0.00019408863964106413, "loss": 0.1659, "step": 2132 }, { "epoch": 0.14, "learning_rate": 0.00019408160345407862, "loss": 0.1732, "step": 2133 }, { "epoch": 0.14, "learning_rate": 0.00019407456320975287, "loss": 0.1395, "step": 2134 }, { "epoch": 0.14, "learning_rate": 0.00019406751890839062, "loss": 0.1615, "step": 2135 }, { "epoch": 0.14, "learning_rate": 0.00019406047055029553, "loss": 0.1739, "step": 2136 }, { "epoch": 0.14, "learning_rate": 0.00019405341813577166, "loss": 0.1483, "step": 2137 }, { "epoch": 0.14, "learning_rate": 0.00019404636166512308, "loss": 0.1378, "step": 2138 }, { "epoch": 0.14, "learning_rate": 0.00019403930113865415, "loss": 0.1521, "step": 2139 }, { "epoch": 0.14, "learning_rate": 0.00019403223655666937, "loss": 0.1794, "step": 2140 }, { "epoch": 0.14, "learning_rate": 0.00019402516791947337, "loss": 0.1497, "step": 2141 }, { "epoch": 0.14, "learning_rate": 0.000194018095227371, "loss": 0.1414, "step": 2142 }, { "epoch": 0.14, "learning_rate": 0.00019401101848066728, "loss": 0.1791, "step": 2143 }, { "epoch": 0.14, "learning_rate": 0.00019400393767966744, "loss": 0.1533, "step": 2144 }, { "epoch": 0.14, "learning_rate": 0.00019399685282467675, "loss": 0.1593, "step": 2145 }, { "epoch": 0.14, "learning_rate": 0.00019398976391600087, "loss": 0.1598, "step": 2146 }, { "epoch": 0.14, "learning_rate": 0.0001939826709539454, "loss": 0.1536, "step": 2147 }, { "epoch": 0.14, "learning_rate": 0.00019397557393881633, "loss": 0.1602, "step": 2148 }, { "epoch": 0.14, "learning_rate": 0.00019396847287091965, "loss": 0.1625, "step": 2149 }, { "epoch": 0.14, "learning_rate": 0.00019396136775056157, "loss": 0.1525, "step": 2150 }, { "epoch": 0.14, "learning_rate": 0.00019395425857804863, "loss": 0.1747, "step": 2151 }, { "epoch": 0.14, "learning_rate": 0.00019394714535368732, "loss": 0.1672, "step": 2152 }, { "epoch": 0.14, "learning_rate": 0.00019394002807778443, "loss": 0.1446, "step": 2153 }, { "epoch": 0.14, "learning_rate": 0.00019393290675064688, "loss": 0.1887, "step": 2154 }, { "epoch": 0.14, "learning_rate": 0.00019392578137258183, "loss": 0.1548, "step": 2155 }, { "epoch": 0.14, "learning_rate": 0.0001939186519438965, "loss": 0.161, "step": 2156 }, { "epoch": 0.14, "learning_rate": 0.00019391151846489844, "loss": 0.1709, "step": 2157 }, { "epoch": 0.14, "learning_rate": 0.0001939043809358952, "loss": 0.1706, "step": 2158 }, { "epoch": 0.14, "learning_rate": 0.0001938972393571946, "loss": 0.1668, "step": 2159 }, { "epoch": 0.14, "learning_rate": 0.0001938900937291047, "loss": 0.1532, "step": 2160 }, { "epoch": 0.14, "learning_rate": 0.0001938829440519336, "loss": 0.1451, "step": 2161 }, { "epoch": 0.14, "learning_rate": 0.00019387579032598967, "loss": 0.1627, "step": 2162 }, { "epoch": 0.14, "learning_rate": 0.00019386863255158135, "loss": 0.1622, "step": 2163 }, { "epoch": 0.14, "learning_rate": 0.0001938614707290174, "loss": 0.1516, "step": 2164 }, { "epoch": 0.14, "learning_rate": 0.00019385430485860665, "loss": 0.1758, "step": 2165 }, { "epoch": 0.14, "learning_rate": 0.00019384713494065812, "loss": 0.1496, "step": 2166 }, { "epoch": 0.14, "learning_rate": 0.00019383996097548106, "loss": 0.1946, "step": 2167 }, { "epoch": 0.14, "learning_rate": 0.0001938327829633848, "loss": 0.1613, "step": 2168 }, { "epoch": 0.14, "learning_rate": 0.00019382560090467894, "loss": 0.1492, "step": 2169 }, { "epoch": 0.14, "learning_rate": 0.00019381841479967319, "loss": 0.1482, "step": 2170 }, { "epoch": 0.14, "learning_rate": 0.00019381122464867746, "loss": 0.1693, "step": 2171 }, { "epoch": 0.14, "learning_rate": 0.00019380403045200182, "loss": 0.1461, "step": 2172 }, { "epoch": 0.14, "learning_rate": 0.00019379683220995656, "loss": 0.1631, "step": 2173 }, { "epoch": 0.14, "learning_rate": 0.0001937896299228521, "loss": 0.1583, "step": 2174 }, { "epoch": 0.14, "learning_rate": 0.00019378242359099898, "loss": 0.1335, "step": 2175 }, { "epoch": 0.14, "learning_rate": 0.00019377521321470805, "loss": 0.152, "step": 2176 }, { "epoch": 0.14, "learning_rate": 0.00019376799879429024, "loss": 0.156, "step": 2177 }, { "epoch": 0.14, "learning_rate": 0.0001937607803300567, "loss": 0.1679, "step": 2178 }, { "epoch": 0.14, "learning_rate": 0.0001937535578223187, "loss": 0.1678, "step": 2179 }, { "epoch": 0.14, "learning_rate": 0.00019374633127138772, "loss": 0.1504, "step": 2180 }, { "epoch": 0.14, "learning_rate": 0.00019373910067757542, "loss": 0.1454, "step": 2181 }, { "epoch": 0.14, "learning_rate": 0.0001937318660411936, "loss": 0.1547, "step": 2182 }, { "epoch": 0.14, "learning_rate": 0.00019372462736255434, "loss": 0.1536, "step": 2183 }, { "epoch": 0.14, "learning_rate": 0.0001937173846419697, "loss": 0.1591, "step": 2184 }, { "epoch": 0.14, "learning_rate": 0.00019371013787975212, "loss": 0.1746, "step": 2185 }, { "epoch": 0.14, "learning_rate": 0.0001937028870762141, "loss": 0.1521, "step": 2186 }, { "epoch": 0.14, "learning_rate": 0.00019369563223166827, "loss": 0.1512, "step": 2187 }, { "epoch": 0.14, "learning_rate": 0.00019368837334642758, "loss": 0.1634, "step": 2188 }, { "epoch": 0.14, "learning_rate": 0.00019368111042080503, "loss": 0.1588, "step": 2189 }, { "epoch": 0.14, "learning_rate": 0.00019367384345511389, "loss": 0.1496, "step": 2190 }, { "epoch": 0.14, "learning_rate": 0.00019366657244966747, "loss": 0.1535, "step": 2191 }, { "epoch": 0.14, "learning_rate": 0.0001936592974047794, "loss": 0.159, "step": 2192 }, { "epoch": 0.14, "learning_rate": 0.00019365201832076344, "loss": 0.1836, "step": 2193 }, { "epoch": 0.14, "learning_rate": 0.00019364473519793343, "loss": 0.1638, "step": 2194 }, { "epoch": 0.14, "learning_rate": 0.00019363744803660356, "loss": 0.1651, "step": 2195 }, { "epoch": 0.14, "learning_rate": 0.000193630156837088, "loss": 0.1741, "step": 2196 }, { "epoch": 0.14, "learning_rate": 0.00019362286159970124, "loss": 0.1764, "step": 2197 }, { "epoch": 0.14, "learning_rate": 0.0001936155623247579, "loss": 0.1406, "step": 2198 }, { "epoch": 0.14, "learning_rate": 0.0001936082590125727, "loss": 0.1496, "step": 2199 }, { "epoch": 0.14, "learning_rate": 0.00019360095166346067, "loss": 0.147, "step": 2200 }, { "epoch": 0.14, "learning_rate": 0.00019359364027773693, "loss": 0.1699, "step": 2201 }, { "epoch": 0.14, "learning_rate": 0.0001935863248557168, "loss": 0.1772, "step": 2202 }, { "epoch": 0.14, "learning_rate": 0.0001935790053977157, "loss": 0.1579, "step": 2203 }, { "epoch": 0.14, "learning_rate": 0.00019357168190404936, "loss": 0.1414, "step": 2204 }, { "epoch": 0.14, "learning_rate": 0.0001935643543750336, "loss": 0.1684, "step": 2205 }, { "epoch": 0.14, "learning_rate": 0.00019355702281098437, "loss": 0.1691, "step": 2206 }, { "epoch": 0.14, "learning_rate": 0.0001935496872122179, "loss": 0.1591, "step": 2207 }, { "epoch": 0.14, "learning_rate": 0.00019354234757905055, "loss": 0.1645, "step": 2208 }, { "epoch": 0.14, "learning_rate": 0.0001935350039117988, "loss": 0.1754, "step": 2209 }, { "epoch": 0.14, "learning_rate": 0.0001935276562107794, "loss": 0.1614, "step": 2210 }, { "epoch": 0.14, "learning_rate": 0.00019352030447630921, "loss": 0.1594, "step": 2211 }, { "epoch": 0.14, "learning_rate": 0.00019351294870870528, "loss": 0.1521, "step": 2212 }, { "epoch": 0.14, "learning_rate": 0.0001935055889082848, "loss": 0.2028, "step": 2213 }, { "epoch": 0.14, "learning_rate": 0.0001934982250753652, "loss": 0.1698, "step": 2214 }, { "epoch": 0.14, "learning_rate": 0.00019349085721026405, "loss": 0.1724, "step": 2215 }, { "epoch": 0.14, "learning_rate": 0.0001934834853132991, "loss": 0.155, "step": 2216 }, { "epoch": 0.14, "learning_rate": 0.00019347610938478824, "loss": 0.1487, "step": 2217 }, { "epoch": 0.14, "learning_rate": 0.00019346872942504958, "loss": 0.1613, "step": 2218 }, { "epoch": 0.14, "learning_rate": 0.00019346134543440143, "loss": 0.1602, "step": 2219 }, { "epoch": 0.14, "learning_rate": 0.00019345395741316216, "loss": 0.1588, "step": 2220 }, { "epoch": 0.14, "learning_rate": 0.00019344656536165038, "loss": 0.1771, "step": 2221 }, { "epoch": 0.14, "learning_rate": 0.00019343916928018492, "loss": 0.1713, "step": 2222 }, { "epoch": 0.14, "learning_rate": 0.00019343176916908475, "loss": 0.1509, "step": 2223 }, { "epoch": 0.14, "learning_rate": 0.00019342436502866898, "loss": 0.1698, "step": 2224 }, { "epoch": 0.14, "learning_rate": 0.00019341695685925694, "loss": 0.1628, "step": 2225 }, { "epoch": 0.14, "learning_rate": 0.00019340954466116807, "loss": 0.1541, "step": 2226 }, { "epoch": 0.14, "learning_rate": 0.00019340212843472205, "loss": 0.1593, "step": 2227 }, { "epoch": 0.14, "learning_rate": 0.00019339470818023876, "loss": 0.1954, "step": 2228 }, { "epoch": 0.14, "learning_rate": 0.00019338728389803813, "loss": 0.1529, "step": 2229 }, { "epoch": 0.14, "learning_rate": 0.00019337985558844035, "loss": 0.1796, "step": 2230 }, { "epoch": 0.14, "learning_rate": 0.0001933724232517658, "loss": 0.1758, "step": 2231 }, { "epoch": 0.14, "learning_rate": 0.000193364986888335, "loss": 0.1546, "step": 2232 }, { "epoch": 0.14, "learning_rate": 0.00019335754649846864, "loss": 0.1655, "step": 2233 }, { "epoch": 0.14, "learning_rate": 0.00019335010208248757, "loss": 0.1642, "step": 2234 }, { "epoch": 0.14, "learning_rate": 0.00019334265364071292, "loss": 0.1556, "step": 2235 }, { "epoch": 0.14, "learning_rate": 0.0001933352011734658, "loss": 0.1499, "step": 2236 }, { "epoch": 0.14, "learning_rate": 0.00019332774468106768, "loss": 0.1621, "step": 2237 }, { "epoch": 0.14, "learning_rate": 0.00019332028416384008, "loss": 0.1758, "step": 2238 }, { "epoch": 0.14, "learning_rate": 0.0001933128196221048, "loss": 0.1787, "step": 2239 }, { "epoch": 0.14, "learning_rate": 0.00019330535105618367, "loss": 0.1771, "step": 2240 }, { "epoch": 0.14, "learning_rate": 0.00019329787846639884, "loss": 0.1705, "step": 2241 }, { "epoch": 0.14, "learning_rate": 0.00019329040185307253, "loss": 0.1678, "step": 2242 }, { "epoch": 0.14, "learning_rate": 0.0001932829212165272, "loss": 0.1721, "step": 2243 }, { "epoch": 0.14, "learning_rate": 0.0001932754365570855, "loss": 0.1495, "step": 2244 }, { "epoch": 0.14, "learning_rate": 0.00019326794787507013, "loss": 0.161, "step": 2245 }, { "epoch": 0.14, "learning_rate": 0.0001932604551708041, "loss": 0.1549, "step": 2246 }, { "epoch": 0.14, "learning_rate": 0.0001932529584446105, "loss": 0.1455, "step": 2247 }, { "epoch": 0.14, "learning_rate": 0.0001932454576968127, "loss": 0.1573, "step": 2248 }, { "epoch": 0.14, "learning_rate": 0.0001932379529277341, "loss": 0.1626, "step": 2249 }, { "epoch": 0.14, "learning_rate": 0.00019323044413769837, "loss": 0.1786, "step": 2250 }, { "epoch": 0.14, "learning_rate": 0.00019322293132702933, "loss": 0.1533, "step": 2251 }, { "epoch": 0.14, "learning_rate": 0.000193215414496051, "loss": 0.182, "step": 2252 }, { "epoch": 0.14, "learning_rate": 0.00019320789364508754, "loss": 0.1894, "step": 2253 }, { "epoch": 0.14, "learning_rate": 0.0001932003687744633, "loss": 0.1524, "step": 2254 }, { "epoch": 0.14, "learning_rate": 0.00019319283988450276, "loss": 0.155, "step": 2255 }, { "epoch": 0.14, "learning_rate": 0.00019318530697553062, "loss": 0.1614, "step": 2256 }, { "epoch": 0.14, "learning_rate": 0.00019317777004787182, "loss": 0.178, "step": 2257 }, { "epoch": 0.14, "learning_rate": 0.00019317022910185126, "loss": 0.1561, "step": 2258 }, { "epoch": 0.14, "learning_rate": 0.00019316268413779425, "loss": 0.1757, "step": 2259 }, { "epoch": 0.14, "learning_rate": 0.00019315513515602613, "loss": 0.1851, "step": 2260 }, { "epoch": 0.14, "learning_rate": 0.00019314758215687244, "loss": 0.1598, "step": 2261 }, { "epoch": 0.15, "learning_rate": 0.000193140025140659, "loss": 0.1597, "step": 2262 }, { "epoch": 0.15, "learning_rate": 0.0001931324641077116, "loss": 0.1775, "step": 2263 }, { "epoch": 0.15, "learning_rate": 0.00019312489905835636, "loss": 0.1364, "step": 2264 }, { "epoch": 0.15, "learning_rate": 0.00019311732999291955, "loss": 0.1733, "step": 2265 }, { "epoch": 0.15, "learning_rate": 0.00019310975691172753, "loss": 0.1699, "step": 2266 }, { "epoch": 0.15, "learning_rate": 0.000193102179815107, "loss": 0.1715, "step": 2267 }, { "epoch": 0.15, "learning_rate": 0.00019309459870338463, "loss": 0.1678, "step": 2268 }, { "epoch": 0.15, "learning_rate": 0.0001930870135768874, "loss": 0.1752, "step": 2269 }, { "epoch": 0.15, "learning_rate": 0.00019307942443594245, "loss": 0.1535, "step": 2270 }, { "epoch": 0.15, "learning_rate": 0.00019307183128087701, "loss": 0.1402, "step": 2271 }, { "epoch": 0.15, "learning_rate": 0.00019306423411201857, "loss": 0.1857, "step": 2272 }, { "epoch": 0.15, "learning_rate": 0.00019305663292969478, "loss": 0.1688, "step": 2273 }, { "epoch": 0.15, "learning_rate": 0.00019304902773423338, "loss": 0.156, "step": 2274 }, { "epoch": 0.15, "learning_rate": 0.00019304141852596245, "loss": 0.1665, "step": 2275 }, { "epoch": 0.15, "learning_rate": 0.00019303380530521008, "loss": 0.1542, "step": 2276 }, { "epoch": 0.15, "learning_rate": 0.00019302618807230462, "loss": 0.1595, "step": 2277 }, { "epoch": 0.15, "learning_rate": 0.00019301856682757453, "loss": 0.1646, "step": 2278 }, { "epoch": 0.15, "learning_rate": 0.00019301094157134853, "loss": 0.1747, "step": 2279 }, { "epoch": 0.15, "learning_rate": 0.00019300331230395545, "loss": 0.1722, "step": 2280 }, { "epoch": 0.15, "learning_rate": 0.00019299567902572427, "loss": 0.1744, "step": 2281 }, { "epoch": 0.15, "learning_rate": 0.00019298804173698423, "loss": 0.1525, "step": 2282 }, { "epoch": 0.15, "learning_rate": 0.0001929804004380647, "loss": 0.1696, "step": 2283 }, { "epoch": 0.15, "learning_rate": 0.00019297275512929516, "loss": 0.1733, "step": 2284 }, { "epoch": 0.15, "learning_rate": 0.00019296510581100535, "loss": 0.1544, "step": 2285 }, { "epoch": 0.15, "learning_rate": 0.00019295745248352518, "loss": 0.1586, "step": 2286 }, { "epoch": 0.15, "learning_rate": 0.00019294979514718465, "loss": 0.1479, "step": 2287 }, { "epoch": 0.15, "learning_rate": 0.00019294213380231402, "loss": 0.1721, "step": 2288 }, { "epoch": 0.15, "learning_rate": 0.00019293446844924374, "loss": 0.1538, "step": 2289 }, { "epoch": 0.15, "learning_rate": 0.00019292679908830432, "loss": 0.1735, "step": 2290 }, { "epoch": 0.15, "learning_rate": 0.00019291912571982648, "loss": 0.1502, "step": 2291 }, { "epoch": 0.15, "learning_rate": 0.0001929114483441412, "loss": 0.1925, "step": 2292 }, { "epoch": 0.15, "learning_rate": 0.00019290376696157958, "loss": 0.1734, "step": 2293 }, { "epoch": 0.15, "learning_rate": 0.00019289608157247283, "loss": 0.1564, "step": 2294 }, { "epoch": 0.15, "learning_rate": 0.00019288839217715243, "loss": 0.1491, "step": 2295 }, { "epoch": 0.15, "learning_rate": 0.00019288069877594998, "loss": 0.1343, "step": 2296 }, { "epoch": 0.15, "learning_rate": 0.00019287300136919727, "loss": 0.1363, "step": 2297 }, { "epoch": 0.15, "learning_rate": 0.00019286529995722623, "loss": 0.1634, "step": 2298 }, { "epoch": 0.15, "learning_rate": 0.000192857594540369, "loss": 0.1528, "step": 2299 }, { "epoch": 0.15, "learning_rate": 0.00019284988511895793, "loss": 0.1687, "step": 2300 }, { "epoch": 0.15, "learning_rate": 0.00019284217169332544, "loss": 0.16, "step": 2301 }, { "epoch": 0.15, "learning_rate": 0.0001928344542638042, "loss": 0.191, "step": 2302 }, { "epoch": 0.15, "learning_rate": 0.00019282673283072698, "loss": 0.1392, "step": 2303 }, { "epoch": 0.15, "learning_rate": 0.00019281900739442684, "loss": 0.1599, "step": 2304 }, { "epoch": 0.15, "learning_rate": 0.00019281127795523693, "loss": 0.1831, "step": 2305 }, { "epoch": 0.15, "learning_rate": 0.00019280354451349058, "loss": 0.157, "step": 2306 }, { "epoch": 0.15, "learning_rate": 0.00019279580706952127, "loss": 0.149, "step": 2307 }, { "epoch": 0.15, "learning_rate": 0.00019278806562366275, "loss": 0.1782, "step": 2308 }, { "epoch": 0.15, "learning_rate": 0.0001927803201762488, "loss": 0.1651, "step": 2309 }, { "epoch": 0.15, "learning_rate": 0.0001927725707276135, "loss": 0.1619, "step": 2310 }, { "epoch": 0.15, "learning_rate": 0.00019276481727809104, "loss": 0.1817, "step": 2311 }, { "epoch": 0.15, "learning_rate": 0.00019275705982801578, "loss": 0.1697, "step": 2312 }, { "epoch": 0.15, "learning_rate": 0.00019274929837772228, "loss": 0.1736, "step": 2313 }, { "epoch": 0.15, "learning_rate": 0.00019274153292754526, "loss": 0.1492, "step": 2314 }, { "epoch": 0.15, "learning_rate": 0.00019273376347781962, "loss": 0.177, "step": 2315 }, { "epoch": 0.15, "learning_rate": 0.0001927259900288804, "loss": 0.1825, "step": 2316 }, { "epoch": 0.15, "learning_rate": 0.0001927182125810628, "loss": 0.1632, "step": 2317 }, { "epoch": 0.15, "learning_rate": 0.00019271043113470234, "loss": 0.1696, "step": 2318 }, { "epoch": 0.15, "learning_rate": 0.0001927026456901345, "loss": 0.1417, "step": 2319 }, { "epoch": 0.15, "learning_rate": 0.00019269485624769507, "loss": 0.1401, "step": 2320 }, { "epoch": 0.15, "learning_rate": 0.00019268706280771997, "loss": 0.1693, "step": 2321 }, { "epoch": 0.15, "learning_rate": 0.00019267926537054527, "loss": 0.1726, "step": 2322 }, { "epoch": 0.15, "learning_rate": 0.00019267146393650732, "loss": 0.1888, "step": 2323 }, { "epoch": 0.15, "learning_rate": 0.0001926636585059425, "loss": 0.1746, "step": 2324 }, { "epoch": 0.15, "learning_rate": 0.00019265584907918743, "loss": 0.1695, "step": 2325 }, { "epoch": 0.15, "learning_rate": 0.00019264803565657892, "loss": 0.1687, "step": 2326 }, { "epoch": 0.15, "learning_rate": 0.0001926402182384539, "loss": 0.1867, "step": 2327 }, { "epoch": 0.15, "learning_rate": 0.00019263239682514952, "loss": 0.1634, "step": 2328 }, { "epoch": 0.15, "learning_rate": 0.00019262457141700306, "loss": 0.154, "step": 2329 }, { "epoch": 0.15, "learning_rate": 0.00019261674201435203, "loss": 0.1754, "step": 2330 }, { "epoch": 0.15, "learning_rate": 0.0001926089086175341, "loss": 0.1537, "step": 2331 }, { "epoch": 0.15, "learning_rate": 0.00019260107122688705, "loss": 0.1464, "step": 2332 }, { "epoch": 0.15, "learning_rate": 0.00019259322984274885, "loss": 0.1493, "step": 2333 }, { "epoch": 0.15, "learning_rate": 0.0001925853844654577, "loss": 0.1459, "step": 2334 }, { "epoch": 0.15, "learning_rate": 0.00019257753509535196, "loss": 0.1617, "step": 2335 }, { "epoch": 0.15, "learning_rate": 0.0001925696817327701, "loss": 0.1809, "step": 2336 }, { "epoch": 0.15, "learning_rate": 0.00019256182437805082, "loss": 0.1651, "step": 2337 }, { "epoch": 0.15, "learning_rate": 0.000192553963031533, "loss": 0.171, "step": 2338 }, { "epoch": 0.15, "learning_rate": 0.0001925460976935556, "loss": 0.167, "step": 2339 }, { "epoch": 0.15, "learning_rate": 0.00019253822836445787, "loss": 0.1577, "step": 2340 }, { "epoch": 0.15, "learning_rate": 0.00019253035504457916, "loss": 0.1766, "step": 2341 }, { "epoch": 0.15, "learning_rate": 0.00019252247773425904, "loss": 0.1869, "step": 2342 }, { "epoch": 0.15, "learning_rate": 0.0001925145964338372, "loss": 0.156, "step": 2343 }, { "epoch": 0.15, "learning_rate": 0.00019250671114365352, "loss": 0.1531, "step": 2344 }, { "epoch": 0.15, "learning_rate": 0.00019249882186404808, "loss": 0.1581, "step": 2345 }, { "epoch": 0.15, "learning_rate": 0.0001924909285953611, "loss": 0.1429, "step": 2346 }, { "epoch": 0.15, "learning_rate": 0.00019248303133793297, "loss": 0.1471, "step": 2347 }, { "epoch": 0.15, "learning_rate": 0.00019247513009210433, "loss": 0.1627, "step": 2348 }, { "epoch": 0.15, "learning_rate": 0.00019246722485821583, "loss": 0.1627, "step": 2349 }, { "epoch": 0.15, "learning_rate": 0.00019245931563660845, "loss": 0.1671, "step": 2350 }, { "epoch": 0.15, "learning_rate": 0.00019245140242762327, "loss": 0.1878, "step": 2351 }, { "epoch": 0.15, "learning_rate": 0.00019244348523160155, "loss": 0.1612, "step": 2352 }, { "epoch": 0.15, "learning_rate": 0.00019243556404888474, "loss": 0.1386, "step": 2353 }, { "epoch": 0.15, "learning_rate": 0.00019242763887981443, "loss": 0.1574, "step": 2354 }, { "epoch": 0.15, "learning_rate": 0.0001924197097247324, "loss": 0.1421, "step": 2355 }, { "epoch": 0.15, "learning_rate": 0.0001924117765839806, "loss": 0.1797, "step": 2356 }, { "epoch": 0.15, "learning_rate": 0.00019240383945790115, "loss": 0.165, "step": 2357 }, { "epoch": 0.15, "learning_rate": 0.00019239589834683636, "loss": 0.1805, "step": 2358 }, { "epoch": 0.15, "learning_rate": 0.0001923879532511287, "loss": 0.1741, "step": 2359 }, { "epoch": 0.15, "learning_rate": 0.00019238000417112078, "loss": 0.1428, "step": 2360 }, { "epoch": 0.15, "learning_rate": 0.0001923720511071554, "loss": 0.1909, "step": 2361 }, { "epoch": 0.15, "learning_rate": 0.0001923640940595756, "loss": 0.157, "step": 2362 }, { "epoch": 0.15, "learning_rate": 0.0001923561330287245, "loss": 0.156, "step": 2363 }, { "epoch": 0.15, "learning_rate": 0.00019234816801494542, "loss": 0.1654, "step": 2364 }, { "epoch": 0.15, "learning_rate": 0.00019234019901858188, "loss": 0.1702, "step": 2365 }, { "epoch": 0.15, "learning_rate": 0.00019233222603997748, "loss": 0.1712, "step": 2366 }, { "epoch": 0.15, "learning_rate": 0.00019232424907947618, "loss": 0.1597, "step": 2367 }, { "epoch": 0.15, "learning_rate": 0.0001923162681374219, "loss": 0.1904, "step": 2368 }, { "epoch": 0.15, "learning_rate": 0.00019230828321415886, "loss": 0.1549, "step": 2369 }, { "epoch": 0.15, "learning_rate": 0.00019230029431003138, "loss": 0.164, "step": 2370 }, { "epoch": 0.15, "learning_rate": 0.00019229230142538402, "loss": 0.1395, "step": 2371 }, { "epoch": 0.15, "learning_rate": 0.00019228430456056148, "loss": 0.1669, "step": 2372 }, { "epoch": 0.15, "learning_rate": 0.00019227630371590864, "loss": 0.1455, "step": 2373 }, { "epoch": 0.15, "learning_rate": 0.0001922682988917705, "loss": 0.1509, "step": 2374 }, { "epoch": 0.15, "learning_rate": 0.00019226029008849232, "loss": 0.1458, "step": 2375 }, { "epoch": 0.15, "learning_rate": 0.00019225227730641943, "loss": 0.1533, "step": 2376 }, { "epoch": 0.15, "learning_rate": 0.00019224426054589746, "loss": 0.1734, "step": 2377 }, { "epoch": 0.15, "learning_rate": 0.00019223623980727208, "loss": 0.154, "step": 2378 }, { "epoch": 0.15, "learning_rate": 0.0001922282150908892, "loss": 0.1678, "step": 2379 }, { "epoch": 0.15, "learning_rate": 0.00019222018639709497, "loss": 0.1623, "step": 2380 }, { "epoch": 0.15, "learning_rate": 0.0001922121537262355, "loss": 0.15, "step": 2381 }, { "epoch": 0.15, "learning_rate": 0.0001922041170786573, "loss": 0.1668, "step": 2382 }, { "epoch": 0.15, "learning_rate": 0.00019219607645470693, "loss": 0.1594, "step": 2383 }, { "epoch": 0.15, "learning_rate": 0.00019218803185473116, "loss": 0.1544, "step": 2384 }, { "epoch": 0.15, "learning_rate": 0.0001921799832790769, "loss": 0.1658, "step": 2385 }, { "epoch": 0.15, "learning_rate": 0.00019217193072809124, "loss": 0.1651, "step": 2386 }, { "epoch": 0.15, "learning_rate": 0.00019216387420212144, "loss": 0.1326, "step": 2387 }, { "epoch": 0.15, "learning_rate": 0.00019215581370151505, "loss": 0.1614, "step": 2388 }, { "epoch": 0.15, "learning_rate": 0.00019214774922661956, "loss": 0.1568, "step": 2389 }, { "epoch": 0.15, "learning_rate": 0.0001921396807777828, "loss": 0.1557, "step": 2390 }, { "epoch": 0.15, "learning_rate": 0.00019213160835535275, "loss": 0.1511, "step": 2391 }, { "epoch": 0.15, "learning_rate": 0.00019212353195967752, "loss": 0.1842, "step": 2392 }, { "epoch": 0.15, "learning_rate": 0.0001921154515911054, "loss": 0.1587, "step": 2393 }, { "epoch": 0.15, "learning_rate": 0.0001921073672499849, "loss": 0.1694, "step": 2394 }, { "epoch": 0.15, "learning_rate": 0.00019209927893666464, "loss": 0.1381, "step": 2395 }, { "epoch": 0.15, "learning_rate": 0.00019209118665149344, "loss": 0.159, "step": 2396 }, { "epoch": 0.15, "learning_rate": 0.00019208309039482027, "loss": 0.1643, "step": 2397 }, { "epoch": 0.15, "learning_rate": 0.00019207499016699427, "loss": 0.1718, "step": 2398 }, { "epoch": 0.15, "learning_rate": 0.0001920668859683648, "loss": 0.1488, "step": 2399 }, { "epoch": 0.15, "learning_rate": 0.00019205877779928136, "loss": 0.151, "step": 2400 }, { "epoch": 0.15, "learning_rate": 0.00019205066566009365, "loss": 0.1831, "step": 2401 }, { "epoch": 0.15, "learning_rate": 0.00019204254955115143, "loss": 0.1486, "step": 2402 }, { "epoch": 0.15, "learning_rate": 0.00019203442947280478, "loss": 0.1531, "step": 2403 }, { "epoch": 0.15, "learning_rate": 0.00019202630542540386, "loss": 0.1714, "step": 2404 }, { "epoch": 0.15, "learning_rate": 0.00019201817740929905, "loss": 0.1644, "step": 2405 }, { "epoch": 0.15, "learning_rate": 0.00019201004542484084, "loss": 0.1687, "step": 2406 }, { "epoch": 0.15, "learning_rate": 0.00019200190947237994, "loss": 0.1428, "step": 2407 }, { "epoch": 0.15, "learning_rate": 0.00019199376955226724, "loss": 0.1614, "step": 2408 }, { "epoch": 0.15, "learning_rate": 0.00019198562566485376, "loss": 0.1482, "step": 2409 }, { "epoch": 0.15, "learning_rate": 0.0001919774778104907, "loss": 0.169, "step": 2410 }, { "epoch": 0.15, "learning_rate": 0.00019196932598952948, "loss": 0.1516, "step": 2411 }, { "epoch": 0.15, "learning_rate": 0.00019196117020232165, "loss": 0.1752, "step": 2412 }, { "epoch": 0.15, "learning_rate": 0.00019195301044921888, "loss": 0.1681, "step": 2413 }, { "epoch": 0.15, "learning_rate": 0.00019194484673057313, "loss": 0.1511, "step": 2414 }, { "epoch": 0.15, "learning_rate": 0.00019193667904673643, "loss": 0.142, "step": 2415 }, { "epoch": 0.15, "learning_rate": 0.000191928507398061, "loss": 0.1814, "step": 2416 }, { "epoch": 0.15, "learning_rate": 0.0001919203317848993, "loss": 0.1647, "step": 2417 }, { "epoch": 0.16, "learning_rate": 0.0001919121522076039, "loss": 0.1623, "step": 2418 }, { "epoch": 0.16, "learning_rate": 0.00019190396866652752, "loss": 0.1669, "step": 2419 }, { "epoch": 0.16, "learning_rate": 0.00019189578116202307, "loss": 0.16, "step": 2420 }, { "epoch": 0.16, "learning_rate": 0.00019188758969444368, "loss": 0.1607, "step": 2421 }, { "epoch": 0.16, "learning_rate": 0.00019187939426414263, "loss": 0.1722, "step": 2422 }, { "epoch": 0.16, "learning_rate": 0.00019187119487147332, "loss": 0.1502, "step": 2423 }, { "epoch": 0.16, "learning_rate": 0.00019186299151678933, "loss": 0.1673, "step": 2424 }, { "epoch": 0.16, "learning_rate": 0.00019185478420044447, "loss": 0.1757, "step": 2425 }, { "epoch": 0.16, "learning_rate": 0.00019184657292279272, "loss": 0.1393, "step": 2426 }, { "epoch": 0.16, "learning_rate": 0.00019183835768418815, "loss": 0.1699, "step": 2427 }, { "epoch": 0.16, "learning_rate": 0.00019183013848498505, "loss": 0.1528, "step": 2428 }, { "epoch": 0.16, "learning_rate": 0.00019182191532553786, "loss": 0.1614, "step": 2429 }, { "epoch": 0.16, "learning_rate": 0.00019181368820620127, "loss": 0.1725, "step": 2430 }, { "epoch": 0.16, "learning_rate": 0.00019180545712733005, "loss": 0.1312, "step": 2431 }, { "epoch": 0.16, "learning_rate": 0.00019179722208927916, "loss": 0.1695, "step": 2432 }, { "epoch": 0.16, "learning_rate": 0.00019178898309240374, "loss": 0.1689, "step": 2433 }, { "epoch": 0.16, "learning_rate": 0.00019178074013705913, "loss": 0.1512, "step": 2434 }, { "epoch": 0.16, "learning_rate": 0.00019177249322360082, "loss": 0.1438, "step": 2435 }, { "epoch": 0.16, "learning_rate": 0.00019176424235238445, "loss": 0.1807, "step": 2436 }, { "epoch": 0.16, "learning_rate": 0.00019175598752376577, "loss": 0.1764, "step": 2437 }, { "epoch": 0.16, "learning_rate": 0.00019174772873810093, "loss": 0.1666, "step": 2438 }, { "epoch": 0.16, "learning_rate": 0.00019173946599574594, "loss": 0.1615, "step": 2439 }, { "epoch": 0.16, "learning_rate": 0.00019173119929705726, "loss": 0.1634, "step": 2440 }, { "epoch": 0.16, "learning_rate": 0.00019172292864239133, "loss": 0.1619, "step": 2441 }, { "epoch": 0.16, "learning_rate": 0.00019171465403210482, "loss": 0.1585, "step": 2442 }, { "epoch": 0.16, "learning_rate": 0.00019170637546655464, "loss": 0.138, "step": 2443 }, { "epoch": 0.16, "learning_rate": 0.00019169809294609772, "loss": 0.1467, "step": 2444 }, { "epoch": 0.16, "learning_rate": 0.00019168980647109134, "loss": 0.1584, "step": 2445 }, { "epoch": 0.16, "learning_rate": 0.0001916815160418928, "loss": 0.1743, "step": 2446 }, { "epoch": 0.16, "learning_rate": 0.0001916732216588597, "loss": 0.1709, "step": 2447 }, { "epoch": 0.16, "learning_rate": 0.00019166492332234968, "loss": 0.1569, "step": 2448 }, { "epoch": 0.16, "learning_rate": 0.0001916566210327206, "loss": 0.1587, "step": 2449 }, { "epoch": 0.16, "learning_rate": 0.00019164831479033057, "loss": 0.1604, "step": 2450 }, { "epoch": 0.16, "learning_rate": 0.00019164000459553774, "loss": 0.1708, "step": 2451 }, { "epoch": 0.16, "learning_rate": 0.0001916316904487005, "loss": 0.1461, "step": 2452 }, { "epoch": 0.16, "learning_rate": 0.00019162337235017742, "loss": 0.1623, "step": 2453 }, { "epoch": 0.16, "learning_rate": 0.00019161505030032726, "loss": 0.1844, "step": 2454 }, { "epoch": 0.16, "learning_rate": 0.0001916067242995088, "loss": 0.1671, "step": 2455 }, { "epoch": 0.16, "learning_rate": 0.00019159839434808125, "loss": 0.1585, "step": 2456 }, { "epoch": 0.16, "learning_rate": 0.00019159006044640377, "loss": 0.1525, "step": 2457 }, { "epoch": 0.16, "learning_rate": 0.0001915817225948358, "loss": 0.1699, "step": 2458 }, { "epoch": 0.16, "learning_rate": 0.00019157338079373685, "loss": 0.1775, "step": 2459 }, { "epoch": 0.16, "learning_rate": 0.00019156503504346673, "loss": 0.1462, "step": 2460 }, { "epoch": 0.16, "learning_rate": 0.0001915566853443853, "loss": 0.1576, "step": 2461 }, { "epoch": 0.16, "learning_rate": 0.0001915483316968527, "loss": 0.1562, "step": 2462 }, { "epoch": 0.16, "learning_rate": 0.0001915399741012292, "loss": 0.1608, "step": 2463 }, { "epoch": 0.16, "learning_rate": 0.00019153161255787518, "loss": 0.1802, "step": 2464 }, { "epoch": 0.16, "learning_rate": 0.00019152324706715125, "loss": 0.1714, "step": 2465 }, { "epoch": 0.16, "learning_rate": 0.00019151487762941816, "loss": 0.1758, "step": 2466 }, { "epoch": 0.16, "learning_rate": 0.0001915065042450369, "loss": 0.1557, "step": 2467 }, { "epoch": 0.16, "learning_rate": 0.00019149812691436854, "loss": 0.1561, "step": 2468 }, { "epoch": 0.16, "learning_rate": 0.00019148974563777438, "loss": 0.1887, "step": 2469 }, { "epoch": 0.16, "learning_rate": 0.00019148136041561583, "loss": 0.1611, "step": 2470 }, { "epoch": 0.16, "learning_rate": 0.00019147297124825455, "loss": 0.1259, "step": 2471 }, { "epoch": 0.16, "learning_rate": 0.00019146457813605232, "loss": 0.1682, "step": 2472 }, { "epoch": 0.16, "learning_rate": 0.0001914561810793711, "loss": 0.1857, "step": 2473 }, { "epoch": 0.16, "learning_rate": 0.000191447780078573, "loss": 0.1558, "step": 2474 }, { "epoch": 0.16, "learning_rate": 0.00019143937513402035, "loss": 0.1812, "step": 2475 }, { "epoch": 0.16, "learning_rate": 0.00019143096624607558, "loss": 0.1804, "step": 2476 }, { "epoch": 0.16, "learning_rate": 0.00019142255341510138, "loss": 0.1392, "step": 2477 }, { "epoch": 0.16, "learning_rate": 0.0001914141366414605, "loss": 0.1536, "step": 2478 }, { "epoch": 0.16, "learning_rate": 0.00019140571592551596, "loss": 0.1582, "step": 2479 }, { "epoch": 0.16, "learning_rate": 0.00019139729126763092, "loss": 0.1393, "step": 2480 }, { "epoch": 0.16, "learning_rate": 0.00019138886266816866, "loss": 0.1567, "step": 2481 }, { "epoch": 0.16, "learning_rate": 0.0001913804301274927, "loss": 0.1786, "step": 2482 }, { "epoch": 0.16, "learning_rate": 0.00019137199364596672, "loss": 0.162, "step": 2483 }, { "epoch": 0.16, "learning_rate": 0.0001913635532239545, "loss": 0.1693, "step": 2484 }, { "epoch": 0.16, "learning_rate": 0.00019135510886182004, "loss": 0.1447, "step": 2485 }, { "epoch": 0.16, "learning_rate": 0.00019134666055992754, "loss": 0.1854, "step": 2486 }, { "epoch": 0.16, "learning_rate": 0.00019133820831864133, "loss": 0.1489, "step": 2487 }, { "epoch": 0.16, "learning_rate": 0.00019132975213832594, "loss": 0.1515, "step": 2488 }, { "epoch": 0.16, "learning_rate": 0.00019132129201934598, "loss": 0.1533, "step": 2489 }, { "epoch": 0.16, "learning_rate": 0.0001913128279620664, "loss": 0.1878, "step": 2490 }, { "epoch": 0.16, "learning_rate": 0.00019130435996685214, "loss": 0.1585, "step": 2491 }, { "epoch": 0.16, "learning_rate": 0.00019129588803406842, "loss": 0.1707, "step": 2492 }, { "epoch": 0.16, "learning_rate": 0.0001912874121640806, "loss": 0.165, "step": 2493 }, { "epoch": 0.16, "learning_rate": 0.0001912789323572542, "loss": 0.1635, "step": 2494 }, { "epoch": 0.16, "learning_rate": 0.00019127044861395488, "loss": 0.1482, "step": 2495 }, { "epoch": 0.16, "learning_rate": 0.0001912619609345486, "loss": 0.1784, "step": 2496 }, { "epoch": 0.16, "learning_rate": 0.0001912534693194013, "loss": 0.1678, "step": 2497 }, { "epoch": 0.16, "learning_rate": 0.00019124497376887925, "loss": 0.1596, "step": 2498 }, { "epoch": 0.16, "learning_rate": 0.0001912364742833488, "loss": 0.1535, "step": 2499 }, { "epoch": 0.16, "learning_rate": 0.00019122797086317653, "loss": 0.1696, "step": 2500 }, { "epoch": 0.16, "learning_rate": 0.0001912194635087291, "loss": 0.1523, "step": 2501 }, { "epoch": 0.16, "learning_rate": 0.00019121095222037347, "loss": 0.1683, "step": 2502 }, { "epoch": 0.16, "learning_rate": 0.00019120243699847662, "loss": 0.1682, "step": 2503 }, { "epoch": 0.16, "learning_rate": 0.0001911939178434058, "loss": 0.1662, "step": 2504 }, { "epoch": 0.16, "learning_rate": 0.00019118539475552844, "loss": 0.1592, "step": 2505 }, { "epoch": 0.16, "learning_rate": 0.00019117686773521207, "loss": 0.1488, "step": 2506 }, { "epoch": 0.16, "learning_rate": 0.00019116833678282442, "loss": 0.1809, "step": 2507 }, { "epoch": 0.16, "learning_rate": 0.00019115980189873341, "loss": 0.1794, "step": 2508 }, { "epoch": 0.16, "learning_rate": 0.00019115126308330712, "loss": 0.1696, "step": 2509 }, { "epoch": 0.16, "learning_rate": 0.00019114272033691378, "loss": 0.1603, "step": 2510 }, { "epoch": 0.16, "learning_rate": 0.00019113417365992184, "loss": 0.1659, "step": 2511 }, { "epoch": 0.16, "learning_rate": 0.0001911256230526998, "loss": 0.1627, "step": 2512 }, { "epoch": 0.16, "learning_rate": 0.00019111706851561647, "loss": 0.1665, "step": 2513 }, { "epoch": 0.16, "learning_rate": 0.00019110851004904076, "loss": 0.1534, "step": 2514 }, { "epoch": 0.16, "learning_rate": 0.00019109994765334175, "loss": 0.1511, "step": 2515 }, { "epoch": 0.16, "learning_rate": 0.00019109138132888872, "loss": 0.1582, "step": 2516 }, { "epoch": 0.16, "learning_rate": 0.0001910828110760511, "loss": 0.1819, "step": 2517 }, { "epoch": 0.16, "learning_rate": 0.00019107423689519845, "loss": 0.1463, "step": 2518 }, { "epoch": 0.16, "learning_rate": 0.0001910656587867006, "loss": 0.155, "step": 2519 }, { "epoch": 0.16, "learning_rate": 0.00019105707675092743, "loss": 0.156, "step": 2520 }, { "epoch": 0.16, "learning_rate": 0.00019104849078824908, "loss": 0.1678, "step": 2521 }, { "epoch": 0.16, "learning_rate": 0.0001910399008990358, "loss": 0.149, "step": 2522 }, { "epoch": 0.16, "learning_rate": 0.00019103130708365807, "loss": 0.1518, "step": 2523 }, { "epoch": 0.16, "learning_rate": 0.00019102270934248647, "loss": 0.158, "step": 2524 }, { "epoch": 0.16, "learning_rate": 0.00019101410767589181, "loss": 0.1628, "step": 2525 }, { "epoch": 0.16, "learning_rate": 0.00019100550208424503, "loss": 0.1499, "step": 2526 }, { "epoch": 0.16, "learning_rate": 0.00019099689256791725, "loss": 0.1438, "step": 2527 }, { "epoch": 0.16, "learning_rate": 0.00019098827912727976, "loss": 0.1435, "step": 2528 }, { "epoch": 0.16, "learning_rate": 0.00019097966176270404, "loss": 0.1574, "step": 2529 }, { "epoch": 0.16, "learning_rate": 0.0001909710404745617, "loss": 0.1583, "step": 2530 }, { "epoch": 0.16, "learning_rate": 0.00019096241526322458, "loss": 0.1592, "step": 2531 }, { "epoch": 0.16, "learning_rate": 0.0001909537861290646, "loss": 0.1738, "step": 2532 }, { "epoch": 0.16, "learning_rate": 0.00019094515307245392, "loss": 0.1703, "step": 2533 }, { "epoch": 0.16, "learning_rate": 0.00019093651609376484, "loss": 0.1764, "step": 2534 }, { "epoch": 0.16, "learning_rate": 0.0001909278751933698, "loss": 0.1772, "step": 2535 }, { "epoch": 0.16, "learning_rate": 0.00019091923037164156, "loss": 0.1683, "step": 2536 }, { "epoch": 0.16, "learning_rate": 0.0001909105816289528, "loss": 0.1542, "step": 2537 }, { "epoch": 0.16, "learning_rate": 0.00019090192896567657, "loss": 0.1681, "step": 2538 }, { "epoch": 0.16, "learning_rate": 0.00019089327238218604, "loss": 0.1527, "step": 2539 }, { "epoch": 0.16, "learning_rate": 0.0001908846118788545, "loss": 0.1833, "step": 2540 }, { "epoch": 0.16, "learning_rate": 0.00019087594745605542, "loss": 0.1741, "step": 2541 }, { "epoch": 0.16, "learning_rate": 0.0001908672791141625, "loss": 0.1764, "step": 2542 }, { "epoch": 0.16, "learning_rate": 0.00019085860685354958, "loss": 0.1629, "step": 2543 }, { "epoch": 0.16, "learning_rate": 0.0001908499306745906, "loss": 0.1631, "step": 2544 }, { "epoch": 0.16, "learning_rate": 0.00019084125057765978, "loss": 0.1643, "step": 2545 }, { "epoch": 0.16, "learning_rate": 0.00019083256656313142, "loss": 0.1584, "step": 2546 }, { "epoch": 0.16, "learning_rate": 0.00019082387863138007, "loss": 0.1792, "step": 2547 }, { "epoch": 0.16, "learning_rate": 0.00019081518678278035, "loss": 0.1534, "step": 2548 }, { "epoch": 0.16, "learning_rate": 0.00019080649101770711, "loss": 0.1336, "step": 2549 }, { "epoch": 0.16, "learning_rate": 0.00019079779133653543, "loss": 0.1477, "step": 2550 }, { "epoch": 0.16, "learning_rate": 0.0001907890877396404, "loss": 0.1541, "step": 2551 }, { "epoch": 0.16, "learning_rate": 0.0001907803802273974, "loss": 0.133, "step": 2552 }, { "epoch": 0.16, "learning_rate": 0.000190771668800182, "loss": 0.1495, "step": 2553 }, { "epoch": 0.16, "learning_rate": 0.0001907629534583698, "loss": 0.1789, "step": 2554 }, { "epoch": 0.16, "learning_rate": 0.0001907542342023367, "loss": 0.1472, "step": 2555 }, { "epoch": 0.16, "learning_rate": 0.00019074551103245876, "loss": 0.16, "step": 2556 }, { "epoch": 0.16, "learning_rate": 0.00019073678394911209, "loss": 0.164, "step": 2557 }, { "epoch": 0.16, "learning_rate": 0.00019072805295267315, "loss": 0.1585, "step": 2558 }, { "epoch": 0.16, "learning_rate": 0.0001907193180435184, "loss": 0.1574, "step": 2559 }, { "epoch": 0.16, "learning_rate": 0.00019071057922202456, "loss": 0.1448, "step": 2560 }, { "epoch": 0.16, "learning_rate": 0.00019070183648856848, "loss": 0.1498, "step": 2561 }, { "epoch": 0.16, "learning_rate": 0.0001906930898435272, "loss": 0.1431, "step": 2562 }, { "epoch": 0.16, "learning_rate": 0.00019068433928727795, "loss": 0.1652, "step": 2563 }, { "epoch": 0.16, "learning_rate": 0.00019067558482019811, "loss": 0.1646, "step": 2564 }, { "epoch": 0.16, "learning_rate": 0.0001906668264426652, "loss": 0.1504, "step": 2565 }, { "epoch": 0.16, "learning_rate": 0.00019065806415505696, "loss": 0.1462, "step": 2566 }, { "epoch": 0.16, "learning_rate": 0.00019064929795775123, "loss": 0.1651, "step": 2567 }, { "epoch": 0.16, "learning_rate": 0.00019064052785112607, "loss": 0.1615, "step": 2568 }, { "epoch": 0.16, "learning_rate": 0.0001906317538355597, "loss": 0.162, "step": 2569 }, { "epoch": 0.16, "learning_rate": 0.00019062297591143054, "loss": 0.1786, "step": 2570 }, { "epoch": 0.16, "learning_rate": 0.00019061419407911712, "loss": 0.1307, "step": 2571 }, { "epoch": 0.16, "learning_rate": 0.00019060540833899814, "loss": 0.1554, "step": 2572 }, { "epoch": 0.16, "learning_rate": 0.00019059661869145252, "loss": 0.1641, "step": 2573 }, { "epoch": 0.17, "learning_rate": 0.00019058782513685927, "loss": 0.1589, "step": 2574 }, { "epoch": 0.17, "learning_rate": 0.0001905790276755977, "loss": 0.1528, "step": 2575 }, { "epoch": 0.17, "learning_rate": 0.00019057022630804716, "loss": 0.1507, "step": 2576 }, { "epoch": 0.17, "learning_rate": 0.00019056142103458722, "loss": 0.1482, "step": 2577 }, { "epoch": 0.17, "learning_rate": 0.00019055261185559763, "loss": 0.1611, "step": 2578 }, { "epoch": 0.17, "learning_rate": 0.00019054379877145826, "loss": 0.1757, "step": 2579 }, { "epoch": 0.17, "learning_rate": 0.0001905349817825492, "loss": 0.1681, "step": 2580 }, { "epoch": 0.17, "learning_rate": 0.0001905261608892507, "loss": 0.1375, "step": 2581 }, { "epoch": 0.17, "learning_rate": 0.00019051733609194314, "loss": 0.1671, "step": 2582 }, { "epoch": 0.17, "learning_rate": 0.00019050850739100713, "loss": 0.1784, "step": 2583 }, { "epoch": 0.17, "learning_rate": 0.00019049967478682338, "loss": 0.1795, "step": 2584 }, { "epoch": 0.17, "learning_rate": 0.00019049083827977285, "loss": 0.1504, "step": 2585 }, { "epoch": 0.17, "learning_rate": 0.00019048199787023658, "loss": 0.1796, "step": 2586 }, { "epoch": 0.17, "learning_rate": 0.0001904731535585958, "loss": 0.1425, "step": 2587 }, { "epoch": 0.17, "learning_rate": 0.00019046430534523198, "loss": 0.1672, "step": 2588 }, { "epoch": 0.17, "learning_rate": 0.0001904554532305267, "loss": 0.1713, "step": 2589 }, { "epoch": 0.17, "learning_rate": 0.0001904465972148617, "loss": 0.1492, "step": 2590 }, { "epoch": 0.17, "learning_rate": 0.00019043773729861884, "loss": 0.1532, "step": 2591 }, { "epoch": 0.17, "learning_rate": 0.00019042887348218033, "loss": 0.1759, "step": 2592 }, { "epoch": 0.17, "learning_rate": 0.00019042000576592835, "loss": 0.1569, "step": 2593 }, { "epoch": 0.17, "learning_rate": 0.00019041113415024536, "loss": 0.1864, "step": 2594 }, { "epoch": 0.17, "learning_rate": 0.00019040225863551394, "loss": 0.1586, "step": 2595 }, { "epoch": 0.17, "learning_rate": 0.00019039337922211683, "loss": 0.1444, "step": 2596 }, { "epoch": 0.17, "learning_rate": 0.00019038449591043702, "loss": 0.1694, "step": 2597 }, { "epoch": 0.17, "learning_rate": 0.00019037560870085755, "loss": 0.1615, "step": 2598 }, { "epoch": 0.17, "learning_rate": 0.00019036671759376176, "loss": 0.1891, "step": 2599 }, { "epoch": 0.17, "learning_rate": 0.00019035782258953297, "loss": 0.1608, "step": 2600 }, { "epoch": 0.17, "learning_rate": 0.0001903489236885549, "loss": 0.1639, "step": 2601 }, { "epoch": 0.17, "learning_rate": 0.00019034002089121126, "loss": 0.1687, "step": 2602 }, { "epoch": 0.17, "learning_rate": 0.00019033111419788597, "loss": 0.138, "step": 2603 }, { "epoch": 0.17, "learning_rate": 0.00019032220360896325, "loss": 0.1382, "step": 2604 }, { "epoch": 0.17, "learning_rate": 0.00019031328912482725, "loss": 0.1415, "step": 2605 }, { "epoch": 0.17, "learning_rate": 0.00019030437074586248, "loss": 0.1627, "step": 2606 }, { "epoch": 0.17, "learning_rate": 0.0001902954484724535, "loss": 0.1609, "step": 2607 }, { "epoch": 0.17, "learning_rate": 0.00019028652230498516, "loss": 0.161, "step": 2608 }, { "epoch": 0.17, "learning_rate": 0.00019027759224384235, "loss": 0.1691, "step": 2609 }, { "epoch": 0.17, "learning_rate": 0.00019026865828941023, "loss": 0.1585, "step": 2610 }, { "epoch": 0.17, "learning_rate": 0.00019025972044207405, "loss": 0.156, "step": 2611 }, { "epoch": 0.17, "learning_rate": 0.00019025077870221929, "loss": 0.1581, "step": 2612 }, { "epoch": 0.17, "learning_rate": 0.00019024183307023152, "loss": 0.1519, "step": 2613 }, { "epoch": 0.17, "learning_rate": 0.00019023288354649655, "loss": 0.1375, "step": 2614 }, { "epoch": 0.17, "learning_rate": 0.0001902239301314004, "loss": 0.1563, "step": 2615 }, { "epoch": 0.17, "learning_rate": 0.00019021497282532907, "loss": 0.1564, "step": 2616 }, { "epoch": 0.17, "learning_rate": 0.00019020601162866895, "loss": 0.1487, "step": 2617 }, { "epoch": 0.17, "learning_rate": 0.00019019704654180648, "loss": 0.1546, "step": 2618 }, { "epoch": 0.17, "learning_rate": 0.00019018807756512827, "loss": 0.1715, "step": 2619 }, { "epoch": 0.17, "learning_rate": 0.00019017910469902108, "loss": 0.1425, "step": 2620 }, { "epoch": 0.17, "learning_rate": 0.00019017012794387195, "loss": 0.1594, "step": 2621 }, { "epoch": 0.17, "learning_rate": 0.00019016114730006792, "loss": 0.1644, "step": 2622 }, { "epoch": 0.17, "learning_rate": 0.00019015216276799634, "loss": 0.1528, "step": 2623 }, { "epoch": 0.17, "learning_rate": 0.00019014317434804472, "loss": 0.1467, "step": 2624 }, { "epoch": 0.17, "learning_rate": 0.0001901341820406006, "loss": 0.1641, "step": 2625 }, { "epoch": 0.17, "learning_rate": 0.00019012518584605182, "loss": 0.1838, "step": 2626 }, { "epoch": 0.17, "learning_rate": 0.00019011618576478636, "loss": 0.1674, "step": 2627 }, { "epoch": 0.17, "learning_rate": 0.00019010718179719236, "loss": 0.1847, "step": 2628 }, { "epoch": 0.17, "learning_rate": 0.00019009817394365807, "loss": 0.1766, "step": 2629 }, { "epoch": 0.17, "learning_rate": 0.00019008916220457201, "loss": 0.1777, "step": 2630 }, { "epoch": 0.17, "learning_rate": 0.0001900801465803228, "loss": 0.1567, "step": 2631 }, { "epoch": 0.17, "learning_rate": 0.00019007112707129927, "loss": 0.1395, "step": 2632 }, { "epoch": 0.17, "learning_rate": 0.00019006210367789037, "loss": 0.1593, "step": 2633 }, { "epoch": 0.17, "learning_rate": 0.00019005307640048523, "loss": 0.1312, "step": 2634 }, { "epoch": 0.17, "learning_rate": 0.00019004404523947316, "loss": 0.1793, "step": 2635 }, { "epoch": 0.17, "learning_rate": 0.00019003501019524368, "loss": 0.1666, "step": 2636 }, { "epoch": 0.17, "learning_rate": 0.0001900259712681864, "loss": 0.1562, "step": 2637 }, { "epoch": 0.17, "learning_rate": 0.00019001692845869113, "loss": 0.174, "step": 2638 }, { "epoch": 0.17, "learning_rate": 0.00019000788176714784, "loss": 0.1645, "step": 2639 }, { "epoch": 0.17, "learning_rate": 0.00018999883119394668, "loss": 0.1495, "step": 2640 }, { "epoch": 0.17, "learning_rate": 0.00018998977673947797, "loss": 0.1546, "step": 2641 }, { "epoch": 0.17, "learning_rate": 0.00018998071840413218, "loss": 0.1632, "step": 2642 }, { "epoch": 0.17, "learning_rate": 0.0001899716561883, "loss": 0.1598, "step": 2643 }, { "epoch": 0.17, "learning_rate": 0.0001899625900923722, "loss": 0.1619, "step": 2644 }, { "epoch": 0.17, "learning_rate": 0.00018995352011673977, "loss": 0.1928, "step": 2645 }, { "epoch": 0.17, "learning_rate": 0.00018994444626179384, "loss": 0.1723, "step": 2646 }, { "epoch": 0.17, "learning_rate": 0.00018993536852792578, "loss": 0.1461, "step": 2647 }, { "epoch": 0.17, "learning_rate": 0.00018992628691552702, "loss": 0.1521, "step": 2648 }, { "epoch": 0.17, "learning_rate": 0.00018991720142498927, "loss": 0.1558, "step": 2649 }, { "epoch": 0.17, "learning_rate": 0.00018990811205670425, "loss": 0.1836, "step": 2650 }, { "epoch": 0.17, "learning_rate": 0.00018989901881106405, "loss": 0.1541, "step": 2651 }, { "epoch": 0.17, "learning_rate": 0.0001898899216884608, "loss": 0.1678, "step": 2652 }, { "epoch": 0.17, "learning_rate": 0.00018988082068928676, "loss": 0.1526, "step": 2653 }, { "epoch": 0.17, "learning_rate": 0.00018987171581393448, "loss": 0.1732, "step": 2654 }, { "epoch": 0.17, "learning_rate": 0.0001898626070627966, "loss": 0.1619, "step": 2655 }, { "epoch": 0.17, "learning_rate": 0.00018985349443626594, "loss": 0.1824, "step": 2656 }, { "epoch": 0.17, "learning_rate": 0.00018984437793473548, "loss": 0.1517, "step": 2657 }, { "epoch": 0.17, "learning_rate": 0.00018983525755859838, "loss": 0.1595, "step": 2658 }, { "epoch": 0.17, "learning_rate": 0.00018982613330824797, "loss": 0.1563, "step": 2659 }, { "epoch": 0.17, "learning_rate": 0.00018981700518407775, "loss": 0.1503, "step": 2660 }, { "epoch": 0.17, "learning_rate": 0.00018980787318648133, "loss": 0.1408, "step": 2661 }, { "epoch": 0.17, "learning_rate": 0.00018979873731585255, "loss": 0.1678, "step": 2662 }, { "epoch": 0.17, "learning_rate": 0.00018978959757258549, "loss": 0.1633, "step": 2663 }, { "epoch": 0.17, "learning_rate": 0.00018978045395707418, "loss": 0.1312, "step": 2664 }, { "epoch": 0.17, "learning_rate": 0.000189771306469713, "loss": 0.1665, "step": 2665 }, { "epoch": 0.17, "learning_rate": 0.0001897621551108965, "loss": 0.1772, "step": 2666 }, { "epoch": 0.17, "learning_rate": 0.0001897529998810192, "loss": 0.1393, "step": 2667 }, { "epoch": 0.17, "learning_rate": 0.00018974384078047606, "loss": 0.1519, "step": 2668 }, { "epoch": 0.17, "learning_rate": 0.00018973467780966202, "loss": 0.1628, "step": 2669 }, { "epoch": 0.17, "learning_rate": 0.00018972551096897223, "loss": 0.1506, "step": 2670 }, { "epoch": 0.17, "learning_rate": 0.00018971634025880202, "loss": 0.1649, "step": 2671 }, { "epoch": 0.17, "learning_rate": 0.0001897071656795469, "loss": 0.1576, "step": 2672 }, { "epoch": 0.17, "learning_rate": 0.00018969798723160255, "loss": 0.1589, "step": 2673 }, { "epoch": 0.17, "learning_rate": 0.00018968880491536475, "loss": 0.1455, "step": 2674 }, { "epoch": 0.17, "learning_rate": 0.00018967961873122952, "loss": 0.1548, "step": 2675 }, { "epoch": 0.17, "learning_rate": 0.000189670428679593, "loss": 0.1617, "step": 2676 }, { "epoch": 0.17, "learning_rate": 0.00018966123476085156, "loss": 0.159, "step": 2677 }, { "epoch": 0.17, "learning_rate": 0.00018965203697540167, "loss": 0.1588, "step": 2678 }, { "epoch": 0.17, "learning_rate": 0.00018964283532363997, "loss": 0.1189, "step": 2679 }, { "epoch": 0.17, "learning_rate": 0.00018963362980596333, "loss": 0.1737, "step": 2680 }, { "epoch": 0.17, "learning_rate": 0.0001896244204227687, "loss": 0.147, "step": 2681 }, { "epoch": 0.17, "learning_rate": 0.00018961520717445329, "loss": 0.1756, "step": 2682 }, { "epoch": 0.17, "learning_rate": 0.00018960599006141436, "loss": 0.1677, "step": 2683 }, { "epoch": 0.17, "learning_rate": 0.00018959676908404948, "loss": 0.1446, "step": 2684 }, { "epoch": 0.17, "learning_rate": 0.0001895875442427563, "loss": 0.1756, "step": 2685 }, { "epoch": 0.17, "learning_rate": 0.00018957831553793262, "loss": 0.1714, "step": 2686 }, { "epoch": 0.17, "learning_rate": 0.00018956908296997643, "loss": 0.1676, "step": 2687 }, { "epoch": 0.17, "learning_rate": 0.0001895598465392859, "loss": 0.1535, "step": 2688 }, { "epoch": 0.17, "learning_rate": 0.00018955060624625936, "loss": 0.1406, "step": 2689 }, { "epoch": 0.17, "learning_rate": 0.0001895413620912953, "loss": 0.1529, "step": 2690 }, { "epoch": 0.17, "learning_rate": 0.0001895321140747924, "loss": 0.1406, "step": 2691 }, { "epoch": 0.17, "learning_rate": 0.00018952286219714952, "loss": 0.1565, "step": 2692 }, { "epoch": 0.17, "learning_rate": 0.00018951360645876557, "loss": 0.1819, "step": 2693 }, { "epoch": 0.17, "learning_rate": 0.00018950434686003975, "loss": 0.1404, "step": 2694 }, { "epoch": 0.17, "learning_rate": 0.0001894950834013714, "loss": 0.1745, "step": 2695 }, { "epoch": 0.17, "learning_rate": 0.00018948581608315999, "loss": 0.1684, "step": 2696 }, { "epoch": 0.17, "learning_rate": 0.00018947654490580518, "loss": 0.1371, "step": 2697 }, { "epoch": 0.17, "learning_rate": 0.00018946726986970682, "loss": 0.1364, "step": 2698 }, { "epoch": 0.17, "learning_rate": 0.0001894579909752649, "loss": 0.1786, "step": 2699 }, { "epoch": 0.17, "learning_rate": 0.00018944870822287956, "loss": 0.1571, "step": 2700 }, { "epoch": 0.17, "learning_rate": 0.00018943942161295114, "loss": 0.1593, "step": 2701 }, { "epoch": 0.17, "learning_rate": 0.00018943013114588016, "loss": 0.1575, "step": 2702 }, { "epoch": 0.17, "learning_rate": 0.00018942083682206722, "loss": 0.1686, "step": 2703 }, { "epoch": 0.17, "learning_rate": 0.00018941153864191317, "loss": 0.1568, "step": 2704 }, { "epoch": 0.17, "learning_rate": 0.00018940223660581898, "loss": 0.166, "step": 2705 }, { "epoch": 0.17, "learning_rate": 0.00018939293071418587, "loss": 0.1561, "step": 2706 }, { "epoch": 0.17, "learning_rate": 0.00018938362096741507, "loss": 0.1605, "step": 2707 }, { "epoch": 0.17, "learning_rate": 0.0001893743073659082, "loss": 0.1665, "step": 2708 }, { "epoch": 0.17, "learning_rate": 0.0001893649899100668, "loss": 0.1601, "step": 2709 }, { "epoch": 0.17, "learning_rate": 0.00018935566860029272, "loss": 0.1686, "step": 2710 }, { "epoch": 0.17, "learning_rate": 0.00018934634343698798, "loss": 0.1604, "step": 2711 }, { "epoch": 0.17, "learning_rate": 0.0001893370144205547, "loss": 0.1442, "step": 2712 }, { "epoch": 0.17, "learning_rate": 0.00018932768155139525, "loss": 0.1433, "step": 2713 }, { "epoch": 0.17, "learning_rate": 0.00018931834482991205, "loss": 0.1586, "step": 2714 }, { "epoch": 0.17, "learning_rate": 0.0001893090042565078, "loss": 0.1651, "step": 2715 }, { "epoch": 0.17, "learning_rate": 0.00018929965983158532, "loss": 0.1556, "step": 2716 }, { "epoch": 0.17, "learning_rate": 0.00018929031155554755, "loss": 0.1537, "step": 2717 }, { "epoch": 0.17, "learning_rate": 0.0001892809594287977, "loss": 0.1457, "step": 2718 }, { "epoch": 0.17, "learning_rate": 0.00018927160345173902, "loss": 0.1571, "step": 2719 }, { "epoch": 0.17, "learning_rate": 0.00018926224362477508, "loss": 0.1576, "step": 2720 }, { "epoch": 0.17, "learning_rate": 0.00018925287994830945, "loss": 0.1609, "step": 2721 }, { "epoch": 0.17, "learning_rate": 0.000189243512422746, "loss": 0.1617, "step": 2722 }, { "epoch": 0.17, "learning_rate": 0.00018923414104848866, "loss": 0.1673, "step": 2723 }, { "epoch": 0.17, "learning_rate": 0.00018922476582594163, "loss": 0.1655, "step": 2724 }, { "epoch": 0.17, "learning_rate": 0.0001892153867555092, "loss": 0.1687, "step": 2725 }, { "epoch": 0.17, "learning_rate": 0.00018920600383759588, "loss": 0.1521, "step": 2726 }, { "epoch": 0.17, "learning_rate": 0.00018919661707260624, "loss": 0.1669, "step": 2727 }, { "epoch": 0.17, "learning_rate": 0.00018918722646094516, "loss": 0.1461, "step": 2728 }, { "epoch": 0.17, "learning_rate": 0.00018917783200301757, "loss": 0.145, "step": 2729 }, { "epoch": 0.18, "learning_rate": 0.00018916843369922869, "loss": 0.1793, "step": 2730 }, { "epoch": 0.18, "learning_rate": 0.00018915903154998373, "loss": 0.1572, "step": 2731 }, { "epoch": 0.18, "learning_rate": 0.00018914962555568825, "loss": 0.1495, "step": 2732 }, { "epoch": 0.18, "learning_rate": 0.00018914021571674782, "loss": 0.1578, "step": 2733 }, { "epoch": 0.18, "learning_rate": 0.0001891308020335683, "loss": 0.1593, "step": 2734 }, { "epoch": 0.18, "learning_rate": 0.00018912138450655563, "loss": 0.159, "step": 2735 }, { "epoch": 0.18, "learning_rate": 0.00018911196313611597, "loss": 0.1705, "step": 2736 }, { "epoch": 0.18, "learning_rate": 0.0001891025379226556, "loss": 0.1633, "step": 2737 }, { "epoch": 0.18, "learning_rate": 0.000189093108866581, "loss": 0.1739, "step": 2738 }, { "epoch": 0.18, "learning_rate": 0.0001890836759682988, "loss": 0.1791, "step": 2739 }, { "epoch": 0.18, "learning_rate": 0.00018907423922821584, "loss": 0.1447, "step": 2740 }, { "epoch": 0.18, "learning_rate": 0.00018906479864673904, "loss": 0.157, "step": 2741 }, { "epoch": 0.18, "learning_rate": 0.00018905535422427555, "loss": 0.1694, "step": 2742 }, { "epoch": 0.18, "learning_rate": 0.00018904590596123265, "loss": 0.156, "step": 2743 }, { "epoch": 0.18, "learning_rate": 0.00018903645385801783, "loss": 0.1482, "step": 2744 }, { "epoch": 0.18, "learning_rate": 0.00018902699791503868, "loss": 0.162, "step": 2745 }, { "epoch": 0.18, "learning_rate": 0.00018901753813270307, "loss": 0.1667, "step": 2746 }, { "epoch": 0.18, "learning_rate": 0.00018900807451141888, "loss": 0.1666, "step": 2747 }, { "epoch": 0.18, "learning_rate": 0.00018899860705159428, "loss": 0.1624, "step": 2748 }, { "epoch": 0.18, "learning_rate": 0.00018898913575363753, "loss": 0.1417, "step": 2749 }, { "epoch": 0.18, "learning_rate": 0.00018897966061795713, "loss": 0.148, "step": 2750 }, { "epoch": 0.18, "learning_rate": 0.0001889701816449617, "loss": 0.1724, "step": 2751 }, { "epoch": 0.18, "learning_rate": 0.00018896069883505992, "loss": 0.1606, "step": 2752 }, { "epoch": 0.18, "learning_rate": 0.00018895121218866093, "loss": 0.1731, "step": 2753 }, { "epoch": 0.18, "learning_rate": 0.0001889417217061737, "loss": 0.1497, "step": 2754 }, { "epoch": 0.18, "learning_rate": 0.00018893222738800757, "loss": 0.1573, "step": 2755 }, { "epoch": 0.18, "learning_rate": 0.00018892272923457198, "loss": 0.1629, "step": 2756 }, { "epoch": 0.18, "learning_rate": 0.00018891322724627654, "loss": 0.1678, "step": 2757 }, { "epoch": 0.18, "learning_rate": 0.00018890372142353105, "loss": 0.1587, "step": 2758 }, { "epoch": 0.18, "learning_rate": 0.00018889421176674544, "loss": 0.1653, "step": 2759 }, { "epoch": 0.18, "learning_rate": 0.00018888469827632985, "loss": 0.1746, "step": 2760 }, { "epoch": 0.18, "learning_rate": 0.0001888751809526945, "loss": 0.1529, "step": 2761 }, { "epoch": 0.18, "learning_rate": 0.0001888656597962499, "loss": 0.1541, "step": 2762 }, { "epoch": 0.18, "learning_rate": 0.00018885613480740655, "loss": 0.1805, "step": 2763 }, { "epoch": 0.18, "learning_rate": 0.00018884660598657535, "loss": 0.1594, "step": 2764 }, { "epoch": 0.18, "learning_rate": 0.00018883707333416716, "loss": 0.1517, "step": 2765 }, { "epoch": 0.18, "learning_rate": 0.00018882753685059311, "loss": 0.1955, "step": 2766 }, { "epoch": 0.18, "learning_rate": 0.00018881799653626446, "loss": 0.1783, "step": 2767 }, { "epoch": 0.18, "learning_rate": 0.00018880845239159264, "loss": 0.1543, "step": 2768 }, { "epoch": 0.18, "learning_rate": 0.00018879890441698928, "loss": 0.1526, "step": 2769 }, { "epoch": 0.18, "learning_rate": 0.00018878935261286607, "loss": 0.1677, "step": 2770 }, { "epoch": 0.18, "learning_rate": 0.00018877979697963501, "loss": 0.1523, "step": 2771 }, { "epoch": 0.18, "learning_rate": 0.00018877023751770816, "loss": 0.1591, "step": 2772 }, { "epoch": 0.18, "learning_rate": 0.00018876067422749778, "loss": 0.1812, "step": 2773 }, { "epoch": 0.18, "learning_rate": 0.00018875110710941636, "loss": 0.1652, "step": 2774 }, { "epoch": 0.18, "learning_rate": 0.00018874153616387637, "loss": 0.1777, "step": 2775 }, { "epoch": 0.18, "learning_rate": 0.00018873196139129067, "loss": 0.1473, "step": 2776 }, { "epoch": 0.18, "learning_rate": 0.0001887223827920721, "loss": 0.1545, "step": 2777 }, { "epoch": 0.18, "learning_rate": 0.00018871280036663376, "loss": 0.1579, "step": 2778 }, { "epoch": 0.18, "learning_rate": 0.00018870321411538897, "loss": 0.1527, "step": 2779 }, { "epoch": 0.18, "learning_rate": 0.00018869362403875107, "loss": 0.1961, "step": 2780 }, { "epoch": 0.18, "learning_rate": 0.00018868403013713367, "loss": 0.1802, "step": 2781 }, { "epoch": 0.18, "learning_rate": 0.00018867443241095047, "loss": 0.1927, "step": 2782 }, { "epoch": 0.18, "learning_rate": 0.00018866483086061545, "loss": 0.1723, "step": 2783 }, { "epoch": 0.18, "learning_rate": 0.00018865522548654266, "loss": 0.1601, "step": 2784 }, { "epoch": 0.18, "learning_rate": 0.00018864561628914632, "loss": 0.1677, "step": 2785 }, { "epoch": 0.18, "learning_rate": 0.00018863600326884082, "loss": 0.1826, "step": 2786 }, { "epoch": 0.18, "learning_rate": 0.00018862638642604078, "loss": 0.172, "step": 2787 }, { "epoch": 0.18, "learning_rate": 0.0001886167657611609, "loss": 0.1545, "step": 2788 }, { "epoch": 0.18, "learning_rate": 0.00018860714127461606, "loss": 0.1539, "step": 2789 }, { "epoch": 0.18, "learning_rate": 0.00018859751296682137, "loss": 0.164, "step": 2790 }, { "epoch": 0.18, "learning_rate": 0.00018858788083819203, "loss": 0.1475, "step": 2791 }, { "epoch": 0.18, "learning_rate": 0.00018857824488914345, "loss": 0.1573, "step": 2792 }, { "epoch": 0.18, "learning_rate": 0.00018856860512009115, "loss": 0.1419, "step": 2793 }, { "epoch": 0.18, "learning_rate": 0.0001885589615314509, "loss": 0.1676, "step": 2794 }, { "epoch": 0.18, "learning_rate": 0.00018854931412363855, "loss": 0.1689, "step": 2795 }, { "epoch": 0.18, "learning_rate": 0.00018853966289707018, "loss": 0.1776, "step": 2796 }, { "epoch": 0.18, "learning_rate": 0.00018853000785216197, "loss": 0.1662, "step": 2797 }, { "epoch": 0.18, "learning_rate": 0.00018852034898933034, "loss": 0.1704, "step": 2798 }, { "epoch": 0.18, "learning_rate": 0.0001885106863089918, "loss": 0.1658, "step": 2799 }, { "epoch": 0.18, "learning_rate": 0.0001885010198115631, "loss": 0.21, "step": 2800 }, { "epoch": 0.18, "learning_rate": 0.0001884913494974611, "loss": 0.1422, "step": 2801 }, { "epoch": 0.18, "learning_rate": 0.00018848167536710282, "loss": 0.1649, "step": 2802 }, { "epoch": 0.18, "learning_rate": 0.00018847199742090548, "loss": 0.1318, "step": 2803 }, { "epoch": 0.18, "learning_rate": 0.00018846231565928644, "loss": 0.1618, "step": 2804 }, { "epoch": 0.18, "learning_rate": 0.00018845263008266327, "loss": 0.1666, "step": 2805 }, { "epoch": 0.18, "learning_rate": 0.00018844294069145362, "loss": 0.1683, "step": 2806 }, { "epoch": 0.18, "learning_rate": 0.0001884332474860754, "loss": 0.1567, "step": 2807 }, { "epoch": 0.18, "learning_rate": 0.00018842355046694656, "loss": 0.1736, "step": 2808 }, { "epoch": 0.18, "learning_rate": 0.00018841384963448534, "loss": 0.1639, "step": 2809 }, { "epoch": 0.18, "learning_rate": 0.0001884041449891101, "loss": 0.1813, "step": 2810 }, { "epoch": 0.18, "learning_rate": 0.0001883944365312394, "loss": 0.1658, "step": 2811 }, { "epoch": 0.18, "learning_rate": 0.00018838472426129185, "loss": 0.1667, "step": 2812 }, { "epoch": 0.18, "learning_rate": 0.00018837500817968632, "loss": 0.157, "step": 2813 }, { "epoch": 0.18, "learning_rate": 0.00018836528828684184, "loss": 0.166, "step": 2814 }, { "epoch": 0.18, "learning_rate": 0.00018835556458317757, "loss": 0.1657, "step": 2815 }, { "epoch": 0.18, "learning_rate": 0.00018834583706911286, "loss": 0.18, "step": 2816 }, { "epoch": 0.18, "learning_rate": 0.00018833610574506724, "loss": 0.1513, "step": 2817 }, { "epoch": 0.18, "learning_rate": 0.00018832637061146036, "loss": 0.1552, "step": 2818 }, { "epoch": 0.18, "learning_rate": 0.00018831663166871202, "loss": 0.1519, "step": 2819 }, { "epoch": 0.18, "learning_rate": 0.00018830688891724226, "loss": 0.1822, "step": 2820 }, { "epoch": 0.18, "learning_rate": 0.00018829714235747127, "loss": 0.1319, "step": 2821 }, { "epoch": 0.18, "learning_rate": 0.00018828739198981933, "loss": 0.1562, "step": 2822 }, { "epoch": 0.18, "learning_rate": 0.00018827763781470693, "loss": 0.1711, "step": 2823 }, { "epoch": 0.18, "learning_rate": 0.00018826787983255473, "loss": 0.1726, "step": 2824 }, { "epoch": 0.18, "learning_rate": 0.00018825811804378358, "loss": 0.1655, "step": 2825 }, { "epoch": 0.18, "learning_rate": 0.00018824835244881448, "loss": 0.1702, "step": 2826 }, { "epoch": 0.18, "learning_rate": 0.0001882385830480685, "loss": 0.1602, "step": 2827 }, { "epoch": 0.18, "learning_rate": 0.00018822880984196703, "loss": 0.1563, "step": 2828 }, { "epoch": 0.18, "learning_rate": 0.0001882190328309315, "loss": 0.1629, "step": 2829 }, { "epoch": 0.18, "learning_rate": 0.00018820925201538358, "loss": 0.1757, "step": 2830 }, { "epoch": 0.18, "learning_rate": 0.00018819946739574502, "loss": 0.167, "step": 2831 }, { "epoch": 0.18, "learning_rate": 0.00018818967897243785, "loss": 0.1513, "step": 2832 }, { "epoch": 0.18, "learning_rate": 0.0001881798867458842, "loss": 0.1465, "step": 2833 }, { "epoch": 0.18, "learning_rate": 0.00018817009071650636, "loss": 0.1514, "step": 2834 }, { "epoch": 0.18, "learning_rate": 0.00018816029088472677, "loss": 0.1494, "step": 2835 }, { "epoch": 0.18, "learning_rate": 0.00018815048725096805, "loss": 0.1577, "step": 2836 }, { "epoch": 0.18, "learning_rate": 0.00018814067981565304, "loss": 0.1633, "step": 2837 }, { "epoch": 0.18, "learning_rate": 0.0001881308685792046, "loss": 0.1427, "step": 2838 }, { "epoch": 0.18, "learning_rate": 0.00018812105354204598, "loss": 0.1662, "step": 2839 }, { "epoch": 0.18, "learning_rate": 0.00018811123470460035, "loss": 0.1565, "step": 2840 }, { "epoch": 0.18, "learning_rate": 0.00018810141206729118, "loss": 0.1632, "step": 2841 }, { "epoch": 0.18, "learning_rate": 0.00018809158563054212, "loss": 0.1637, "step": 2842 }, { "epoch": 0.18, "learning_rate": 0.0001880817553947769, "loss": 0.1655, "step": 2843 }, { "epoch": 0.18, "learning_rate": 0.00018807192136041943, "loss": 0.183, "step": 2844 }, { "epoch": 0.18, "learning_rate": 0.00018806208352789393, "loss": 0.1954, "step": 2845 }, { "epoch": 0.18, "learning_rate": 0.00018805224189762454, "loss": 0.1593, "step": 2846 }, { "epoch": 0.18, "learning_rate": 0.00018804239647003573, "loss": 0.1653, "step": 2847 }, { "epoch": 0.18, "learning_rate": 0.0001880325472455521, "loss": 0.1311, "step": 2848 }, { "epoch": 0.18, "learning_rate": 0.0001880226942245984, "loss": 0.1637, "step": 2849 }, { "epoch": 0.18, "learning_rate": 0.00018801283740759952, "loss": 0.1568, "step": 2850 }, { "epoch": 0.18, "learning_rate": 0.00018800297679498061, "loss": 0.1628, "step": 2851 }, { "epoch": 0.18, "learning_rate": 0.00018799311238716683, "loss": 0.1769, "step": 2852 }, { "epoch": 0.18, "learning_rate": 0.00018798324418458368, "loss": 0.1712, "step": 2853 }, { "epoch": 0.18, "learning_rate": 0.00018797337218765664, "loss": 0.161, "step": 2854 }, { "epoch": 0.18, "learning_rate": 0.00018796349639681154, "loss": 0.1424, "step": 2855 }, { "epoch": 0.18, "learning_rate": 0.00018795361681247426, "loss": 0.1347, "step": 2856 }, { "epoch": 0.18, "learning_rate": 0.0001879437334350708, "loss": 0.1304, "step": 2857 }, { "epoch": 0.18, "learning_rate": 0.00018793384626502745, "loss": 0.1701, "step": 2858 }, { "epoch": 0.18, "learning_rate": 0.00018792395530277056, "loss": 0.1722, "step": 2859 }, { "epoch": 0.18, "learning_rate": 0.00018791406054872672, "loss": 0.1644, "step": 2860 }, { "epoch": 0.18, "learning_rate": 0.00018790416200332264, "loss": 0.1522, "step": 2861 }, { "epoch": 0.18, "learning_rate": 0.0001878942596669852, "loss": 0.1402, "step": 2862 }, { "epoch": 0.18, "learning_rate": 0.00018788435354014145, "loss": 0.1559, "step": 2863 }, { "epoch": 0.18, "learning_rate": 0.00018787444362321857, "loss": 0.1615, "step": 2864 }, { "epoch": 0.18, "learning_rate": 0.00018786452991664397, "loss": 0.152, "step": 2865 }, { "epoch": 0.18, "learning_rate": 0.00018785461242084518, "loss": 0.1438, "step": 2866 }, { "epoch": 0.18, "learning_rate": 0.00018784469113624987, "loss": 0.1361, "step": 2867 }, { "epoch": 0.18, "learning_rate": 0.00018783476606328594, "loss": 0.182, "step": 2868 }, { "epoch": 0.18, "learning_rate": 0.00018782483720238143, "loss": 0.1444, "step": 2869 }, { "epoch": 0.18, "learning_rate": 0.00018781490455396446, "loss": 0.1697, "step": 2870 }, { "epoch": 0.18, "learning_rate": 0.00018780496811846345, "loss": 0.1547, "step": 2871 }, { "epoch": 0.18, "learning_rate": 0.00018779502789630686, "loss": 0.1644, "step": 2872 }, { "epoch": 0.18, "learning_rate": 0.00018778508388792343, "loss": 0.1455, "step": 2873 }, { "epoch": 0.18, "learning_rate": 0.00018777513609374193, "loss": 0.1779, "step": 2874 }, { "epoch": 0.18, "learning_rate": 0.00018776518451419144, "loss": 0.1496, "step": 2875 }, { "epoch": 0.18, "learning_rate": 0.0001877552291497011, "loss": 0.1894, "step": 2876 }, { "epoch": 0.18, "learning_rate": 0.00018774527000070022, "loss": 0.1684, "step": 2877 }, { "epoch": 0.18, "learning_rate": 0.00018773530706761834, "loss": 0.1604, "step": 2878 }, { "epoch": 0.18, "learning_rate": 0.00018772534035088508, "loss": 0.1693, "step": 2879 }, { "epoch": 0.18, "learning_rate": 0.00018771536985093026, "loss": 0.1606, "step": 2880 }, { "epoch": 0.18, "learning_rate": 0.00018770539556818393, "loss": 0.1653, "step": 2881 }, { "epoch": 0.18, "learning_rate": 0.00018769541750307615, "loss": 0.1897, "step": 2882 }, { "epoch": 0.18, "learning_rate": 0.0001876854356560373, "loss": 0.1659, "step": 2883 }, { "epoch": 0.18, "learning_rate": 0.00018767545002749782, "loss": 0.1688, "step": 2884 }, { "epoch": 0.18, "learning_rate": 0.00018766546061788834, "loss": 0.1486, "step": 2885 }, { "epoch": 0.19, "learning_rate": 0.0001876554674276397, "loss": 0.1565, "step": 2886 }, { "epoch": 0.19, "learning_rate": 0.00018764547045718282, "loss": 0.1496, "step": 2887 }, { "epoch": 0.19, "learning_rate": 0.00018763546970694886, "loss": 0.1552, "step": 2888 }, { "epoch": 0.19, "learning_rate": 0.00018762546517736908, "loss": 0.1626, "step": 2889 }, { "epoch": 0.19, "learning_rate": 0.00018761545686887495, "loss": 0.1531, "step": 2890 }, { "epoch": 0.19, "learning_rate": 0.0001876054447818981, "loss": 0.1588, "step": 2891 }, { "epoch": 0.19, "learning_rate": 0.0001875954289168703, "loss": 0.1627, "step": 2892 }, { "epoch": 0.19, "learning_rate": 0.00018758540927422347, "loss": 0.1474, "step": 2893 }, { "epoch": 0.19, "learning_rate": 0.00018757538585438977, "loss": 0.1611, "step": 2894 }, { "epoch": 0.19, "learning_rate": 0.00018756535865780137, "loss": 0.1589, "step": 2895 }, { "epoch": 0.19, "learning_rate": 0.00018755532768489078, "loss": 0.172, "step": 2896 }, { "epoch": 0.19, "learning_rate": 0.0001875452929360906, "loss": 0.163, "step": 2897 }, { "epoch": 0.19, "learning_rate": 0.00018753525441183352, "loss": 0.1646, "step": 2898 }, { "epoch": 0.19, "learning_rate": 0.00018752521211255253, "loss": 0.1787, "step": 2899 }, { "epoch": 0.19, "learning_rate": 0.00018751516603868066, "loss": 0.142, "step": 2900 }, { "epoch": 0.19, "learning_rate": 0.00018750511619065118, "loss": 0.1684, "step": 2901 }, { "epoch": 0.19, "learning_rate": 0.00018749506256889747, "loss": 0.1577, "step": 2902 }, { "epoch": 0.19, "learning_rate": 0.00018748500517385314, "loss": 0.146, "step": 2903 }, { "epoch": 0.19, "learning_rate": 0.00018747494400595191, "loss": 0.1634, "step": 2904 }, { "epoch": 0.19, "learning_rate": 0.00018746487906562767, "loss": 0.1656, "step": 2905 }, { "epoch": 0.19, "learning_rate": 0.00018745481035331444, "loss": 0.1385, "step": 2906 }, { "epoch": 0.19, "learning_rate": 0.00018744473786944652, "loss": 0.15, "step": 2907 }, { "epoch": 0.19, "learning_rate": 0.00018743466161445823, "loss": 0.1524, "step": 2908 }, { "epoch": 0.19, "learning_rate": 0.00018742458158878413, "loss": 0.1357, "step": 2909 }, { "epoch": 0.19, "learning_rate": 0.00018741449779285897, "loss": 0.1658, "step": 2910 }, { "epoch": 0.19, "learning_rate": 0.00018740441022711756, "loss": 0.1654, "step": 2911 }, { "epoch": 0.19, "learning_rate": 0.00018739431889199497, "loss": 0.1581, "step": 2912 }, { "epoch": 0.19, "learning_rate": 0.00018738422378792636, "loss": 0.1482, "step": 2913 }, { "epoch": 0.19, "learning_rate": 0.00018737412491534716, "loss": 0.152, "step": 2914 }, { "epoch": 0.19, "learning_rate": 0.0001873640222746928, "loss": 0.1727, "step": 2915 }, { "epoch": 0.19, "learning_rate": 0.00018735391586639904, "loss": 0.151, "step": 2916 }, { "epoch": 0.19, "learning_rate": 0.00018734380569090174, "loss": 0.1407, "step": 2917 }, { "epoch": 0.19, "learning_rate": 0.0001873336917486368, "loss": 0.1611, "step": 2918 }, { "epoch": 0.19, "learning_rate": 0.00018732357404004047, "loss": 0.1956, "step": 2919 }, { "epoch": 0.19, "learning_rate": 0.0001873134525655491, "loss": 0.1406, "step": 2920 }, { "epoch": 0.19, "learning_rate": 0.00018730332732559915, "loss": 0.1468, "step": 2921 }, { "epoch": 0.19, "learning_rate": 0.0001872931983206273, "loss": 0.1436, "step": 2922 }, { "epoch": 0.19, "learning_rate": 0.00018728306555107035, "loss": 0.1555, "step": 2923 }, { "epoch": 0.19, "learning_rate": 0.00018727292901736528, "loss": 0.1604, "step": 2924 }, { "epoch": 0.19, "learning_rate": 0.00018726278871994927, "loss": 0.1746, "step": 2925 }, { "epoch": 0.19, "learning_rate": 0.0001872526446592596, "loss": 0.1673, "step": 2926 }, { "epoch": 0.19, "learning_rate": 0.00018724249683573372, "loss": 0.1759, "step": 2927 }, { "epoch": 0.19, "learning_rate": 0.0001872323452498093, "loss": 0.1614, "step": 2928 }, { "epoch": 0.19, "learning_rate": 0.00018722218990192415, "loss": 0.1645, "step": 2929 }, { "epoch": 0.19, "learning_rate": 0.00018721203079251616, "loss": 0.1353, "step": 2930 }, { "epoch": 0.19, "learning_rate": 0.00018720186792202354, "loss": 0.172, "step": 2931 }, { "epoch": 0.19, "learning_rate": 0.00018719170129088449, "loss": 0.1547, "step": 2932 }, { "epoch": 0.19, "learning_rate": 0.00018718153089953746, "loss": 0.1514, "step": 2933 }, { "epoch": 0.19, "learning_rate": 0.00018717135674842112, "loss": 0.1486, "step": 2934 }, { "epoch": 0.19, "learning_rate": 0.00018716117883797417, "loss": 0.146, "step": 2935 }, { "epoch": 0.19, "learning_rate": 0.0001871509971686356, "loss": 0.1462, "step": 2936 }, { "epoch": 0.19, "learning_rate": 0.00018714081174084446, "loss": 0.1722, "step": 2937 }, { "epoch": 0.19, "learning_rate": 0.00018713062255504002, "loss": 0.1444, "step": 2938 }, { "epoch": 0.19, "learning_rate": 0.00018712042961166169, "loss": 0.1418, "step": 2939 }, { "epoch": 0.19, "learning_rate": 0.00018711023291114902, "loss": 0.1505, "step": 2940 }, { "epoch": 0.19, "learning_rate": 0.00018710003245394182, "loss": 0.1646, "step": 2941 }, { "epoch": 0.19, "learning_rate": 0.00018708982824047993, "loss": 0.2087, "step": 2942 }, { "epoch": 0.19, "learning_rate": 0.00018707962027120342, "loss": 0.1582, "step": 2943 }, { "epoch": 0.19, "learning_rate": 0.00018706940854655257, "loss": 0.1728, "step": 2944 }, { "epoch": 0.19, "learning_rate": 0.0001870591930669677, "loss": 0.1433, "step": 2945 }, { "epoch": 0.19, "learning_rate": 0.00018704897383288943, "loss": 0.1465, "step": 2946 }, { "epoch": 0.19, "learning_rate": 0.0001870387508447584, "loss": 0.1609, "step": 2947 }, { "epoch": 0.19, "learning_rate": 0.00018702852410301554, "loss": 0.171, "step": 2948 }, { "epoch": 0.19, "learning_rate": 0.00018701829360810184, "loss": 0.164, "step": 2949 }, { "epoch": 0.19, "learning_rate": 0.00018700805936045853, "loss": 0.1587, "step": 2950 }, { "epoch": 0.19, "learning_rate": 0.00018699782136052698, "loss": 0.1366, "step": 2951 }, { "epoch": 0.19, "learning_rate": 0.0001869875796087487, "loss": 0.185, "step": 2952 }, { "epoch": 0.19, "learning_rate": 0.00018697733410556535, "loss": 0.1495, "step": 2953 }, { "epoch": 0.19, "learning_rate": 0.00018696708485141877, "loss": 0.1558, "step": 2954 }, { "epoch": 0.19, "learning_rate": 0.00018695683184675102, "loss": 0.1407, "step": 2955 }, { "epoch": 0.19, "learning_rate": 0.00018694657509200422, "loss": 0.1628, "step": 2956 }, { "epoch": 0.19, "learning_rate": 0.0001869363145876207, "loss": 0.1605, "step": 2957 }, { "epoch": 0.19, "learning_rate": 0.000186926050334043, "loss": 0.1551, "step": 2958 }, { "epoch": 0.19, "learning_rate": 0.00018691578233171377, "loss": 0.1616, "step": 2959 }, { "epoch": 0.19, "learning_rate": 0.00018690551058107575, "loss": 0.1621, "step": 2960 }, { "epoch": 0.19, "learning_rate": 0.00018689523508257197, "loss": 0.1607, "step": 2961 }, { "epoch": 0.19, "learning_rate": 0.00018688495583664557, "loss": 0.1863, "step": 2962 }, { "epoch": 0.19, "learning_rate": 0.00018687467284373988, "loss": 0.1559, "step": 2963 }, { "epoch": 0.19, "learning_rate": 0.0001868643861042983, "loss": 0.1317, "step": 2964 }, { "epoch": 0.19, "learning_rate": 0.00018685409561876445, "loss": 0.1523, "step": 2965 }, { "epoch": 0.19, "learning_rate": 0.00018684380138758218, "loss": 0.141, "step": 2966 }, { "epoch": 0.19, "learning_rate": 0.00018683350341119537, "loss": 0.1428, "step": 2967 }, { "epoch": 0.19, "learning_rate": 0.00018682320169004818, "loss": 0.1387, "step": 2968 }, { "epoch": 0.19, "learning_rate": 0.00018681289622458485, "loss": 0.1574, "step": 2969 }, { "epoch": 0.19, "learning_rate": 0.00018680258701524985, "loss": 0.1412, "step": 2970 }, { "epoch": 0.19, "learning_rate": 0.00018679227406248768, "loss": 0.1553, "step": 2971 }, { "epoch": 0.19, "learning_rate": 0.00018678195736674319, "loss": 0.1726, "step": 2972 }, { "epoch": 0.19, "learning_rate": 0.0001867716369284613, "loss": 0.1703, "step": 2973 }, { "epoch": 0.19, "learning_rate": 0.000186761312748087, "loss": 0.1556, "step": 2974 }, { "epoch": 0.19, "learning_rate": 0.00018675098482606557, "loss": 0.1337, "step": 2975 }, { "epoch": 0.19, "learning_rate": 0.00018674065316284243, "loss": 0.1589, "step": 2976 }, { "epoch": 0.19, "learning_rate": 0.0001867303177588631, "loss": 0.1444, "step": 2977 }, { "epoch": 0.19, "learning_rate": 0.00018671997861457337, "loss": 0.1523, "step": 2978 }, { "epoch": 0.19, "learning_rate": 0.0001867096357304191, "loss": 0.1428, "step": 2979 }, { "epoch": 0.19, "learning_rate": 0.00018669928910684626, "loss": 0.1536, "step": 2980 }, { "epoch": 0.19, "learning_rate": 0.00018668893874430116, "loss": 0.1728, "step": 2981 }, { "epoch": 0.19, "learning_rate": 0.00018667858464323009, "loss": 0.1436, "step": 2982 }, { "epoch": 0.19, "learning_rate": 0.0001866682268040796, "loss": 0.1526, "step": 2983 }, { "epoch": 0.19, "learning_rate": 0.00018665786522729637, "loss": 0.1563, "step": 2984 }, { "epoch": 0.19, "learning_rate": 0.00018664749991332733, "loss": 0.1798, "step": 2985 }, { "epoch": 0.19, "learning_rate": 0.0001866371308626194, "loss": 0.1759, "step": 2986 }, { "epoch": 0.19, "learning_rate": 0.00018662675807561976, "loss": 0.1712, "step": 2987 }, { "epoch": 0.19, "learning_rate": 0.0001866163815527758, "loss": 0.1457, "step": 2988 }, { "epoch": 0.19, "learning_rate": 0.00018660600129453498, "loss": 0.1521, "step": 2989 }, { "epoch": 0.19, "learning_rate": 0.00018659561730134495, "loss": 0.1833, "step": 2990 }, { "epoch": 0.19, "learning_rate": 0.0001865852295736535, "loss": 0.1602, "step": 2991 }, { "epoch": 0.19, "learning_rate": 0.0001865748381119087, "loss": 0.1797, "step": 2992 }, { "epoch": 0.19, "learning_rate": 0.00018656444291655864, "loss": 0.1502, "step": 2993 }, { "epoch": 0.19, "learning_rate": 0.00018655404398805158, "loss": 0.1337, "step": 2994 }, { "epoch": 0.19, "learning_rate": 0.00018654364132683604, "loss": 0.1579, "step": 2995 }, { "epoch": 0.19, "learning_rate": 0.00018653323493336062, "loss": 0.1587, "step": 2996 }, { "epoch": 0.19, "learning_rate": 0.0001865228248080741, "loss": 0.1544, "step": 2997 }, { "epoch": 0.19, "learning_rate": 0.0001865124109514254, "loss": 0.185, "step": 2998 }, { "epoch": 0.19, "learning_rate": 0.00018650199336386368, "loss": 0.1657, "step": 2999 }, { "epoch": 0.19, "learning_rate": 0.0001864915720458382, "loss": 0.1542, "step": 3000 }, { "epoch": 0.19, "learning_rate": 0.00018648114699779833, "loss": 0.1523, "step": 3001 }, { "epoch": 0.19, "learning_rate": 0.00018647071822019374, "loss": 0.1604, "step": 3002 }, { "epoch": 0.19, "learning_rate": 0.00018646028571347407, "loss": 0.161, "step": 3003 }, { "epoch": 0.19, "learning_rate": 0.00018644984947808934, "loss": 0.1665, "step": 3004 }, { "epoch": 0.19, "learning_rate": 0.00018643940951448956, "loss": 0.1608, "step": 3005 }, { "epoch": 0.19, "learning_rate": 0.000186428965823125, "loss": 0.1801, "step": 3006 }, { "epoch": 0.19, "learning_rate": 0.000186418518404446, "loss": 0.161, "step": 3007 }, { "epoch": 0.19, "learning_rate": 0.0001864080672589031, "loss": 0.1546, "step": 3008 }, { "epoch": 0.19, "learning_rate": 0.00018639761238694713, "loss": 0.1516, "step": 3009 }, { "epoch": 0.19, "learning_rate": 0.00018638715378902886, "loss": 0.1382, "step": 3010 }, { "epoch": 0.19, "learning_rate": 0.00018637669146559936, "loss": 0.1594, "step": 3011 }, { "epoch": 0.19, "learning_rate": 0.00018636622541710982, "loss": 0.1563, "step": 3012 }, { "epoch": 0.19, "learning_rate": 0.00018635575564401158, "loss": 0.1376, "step": 3013 }, { "epoch": 0.19, "learning_rate": 0.00018634528214675616, "loss": 0.1535, "step": 3014 }, { "epoch": 0.19, "learning_rate": 0.00018633480492579528, "loss": 0.1404, "step": 3015 }, { "epoch": 0.19, "learning_rate": 0.00018632432398158074, "loss": 0.1589, "step": 3016 }, { "epoch": 0.19, "learning_rate": 0.00018631383931456453, "loss": 0.157, "step": 3017 }, { "epoch": 0.19, "learning_rate": 0.00018630335092519883, "loss": 0.1633, "step": 3018 }, { "epoch": 0.19, "learning_rate": 0.00018629285881393597, "loss": 0.1386, "step": 3019 }, { "epoch": 0.19, "learning_rate": 0.00018628236298122842, "loss": 0.1778, "step": 3020 }, { "epoch": 0.19, "learning_rate": 0.0001862718634275288, "loss": 0.165, "step": 3021 }, { "epoch": 0.19, "learning_rate": 0.00018626136015328993, "loss": 0.144, "step": 3022 }, { "epoch": 0.19, "learning_rate": 0.00018625085315896478, "loss": 0.1694, "step": 3023 }, { "epoch": 0.19, "learning_rate": 0.00018624034244500647, "loss": 0.1643, "step": 3024 }, { "epoch": 0.19, "learning_rate": 0.00018622982801186824, "loss": 0.1395, "step": 3025 }, { "epoch": 0.19, "learning_rate": 0.00018621930986000357, "loss": 0.1556, "step": 3026 }, { "epoch": 0.19, "learning_rate": 0.00018620878798986608, "loss": 0.1566, "step": 3027 }, { "epoch": 0.19, "learning_rate": 0.0001861982624019095, "loss": 0.1654, "step": 3028 }, { "epoch": 0.19, "learning_rate": 0.00018618773309658776, "loss": 0.1585, "step": 3029 }, { "epoch": 0.19, "learning_rate": 0.00018617720007435497, "loss": 0.1509, "step": 3030 }, { "epoch": 0.19, "learning_rate": 0.00018616666333566535, "loss": 0.1461, "step": 3031 }, { "epoch": 0.19, "learning_rate": 0.0001861561228809733, "loss": 0.1495, "step": 3032 }, { "epoch": 0.19, "learning_rate": 0.00018614557871073341, "loss": 0.1712, "step": 3033 }, { "epoch": 0.19, "learning_rate": 0.0001861350308254004, "loss": 0.166, "step": 3034 }, { "epoch": 0.19, "learning_rate": 0.00018612447922542914, "loss": 0.1549, "step": 3035 }, { "epoch": 0.19, "learning_rate": 0.00018611392391127467, "loss": 0.1505, "step": 3036 }, { "epoch": 0.19, "learning_rate": 0.0001861033648833922, "loss": 0.1593, "step": 3037 }, { "epoch": 0.19, "learning_rate": 0.00018609280214223713, "loss": 0.17, "step": 3038 }, { "epoch": 0.19, "learning_rate": 0.00018608223568826492, "loss": 0.1542, "step": 3039 }, { "epoch": 0.19, "learning_rate": 0.00018607166552193135, "loss": 0.1691, "step": 3040 }, { "epoch": 0.19, "learning_rate": 0.00018606109164369218, "loss": 0.1547, "step": 3041 }, { "epoch": 0.2, "learning_rate": 0.00018605051405400346, "loss": 0.1816, "step": 3042 }, { "epoch": 0.2, "learning_rate": 0.00018603993275332132, "loss": 0.1572, "step": 3043 }, { "epoch": 0.2, "learning_rate": 0.00018602934774210214, "loss": 0.1706, "step": 3044 }, { "epoch": 0.2, "learning_rate": 0.00018601875902080238, "loss": 0.1464, "step": 3045 }, { "epoch": 0.2, "learning_rate": 0.0001860081665898787, "loss": 0.166, "step": 3046 }, { "epoch": 0.2, "learning_rate": 0.00018599757044978789, "loss": 0.1514, "step": 3047 }, { "epoch": 0.2, "learning_rate": 0.00018598697060098686, "loss": 0.1757, "step": 3048 }, { "epoch": 0.2, "learning_rate": 0.00018597636704393284, "loss": 0.1786, "step": 3049 }, { "epoch": 0.2, "learning_rate": 0.00018596575977908306, "loss": 0.1551, "step": 3050 }, { "epoch": 0.2, "learning_rate": 0.000185955148806895, "loss": 0.1437, "step": 3051 }, { "epoch": 0.2, "learning_rate": 0.00018594453412782623, "loss": 0.1572, "step": 3052 }, { "epoch": 0.2, "learning_rate": 0.00018593391574233455, "loss": 0.1697, "step": 3053 }, { "epoch": 0.2, "learning_rate": 0.00018592329365087784, "loss": 0.1523, "step": 3054 }, { "epoch": 0.2, "learning_rate": 0.00018591266785391423, "loss": 0.1658, "step": 3055 }, { "epoch": 0.2, "learning_rate": 0.00018590203835190192, "loss": 0.1577, "step": 3056 }, { "epoch": 0.2, "learning_rate": 0.00018589140514529938, "loss": 0.1496, "step": 3057 }, { "epoch": 0.2, "learning_rate": 0.0001858807682345651, "loss": 0.176, "step": 3058 }, { "epoch": 0.2, "learning_rate": 0.00018587012762015786, "loss": 0.151, "step": 3059 }, { "epoch": 0.2, "learning_rate": 0.00018585948330253652, "loss": 0.1699, "step": 3060 }, { "epoch": 0.2, "learning_rate": 0.00018584883528216018, "loss": 0.1699, "step": 3061 }, { "epoch": 0.2, "learning_rate": 0.00018583818355948791, "loss": 0.1759, "step": 3062 }, { "epoch": 0.2, "learning_rate": 0.00018582752813497923, "loss": 0.1771, "step": 3063 }, { "epoch": 0.2, "learning_rate": 0.00018581686900909355, "loss": 0.1356, "step": 3064 }, { "epoch": 0.2, "learning_rate": 0.00018580620618229064, "loss": 0.1731, "step": 3065 }, { "epoch": 0.2, "learning_rate": 0.00018579553965503024, "loss": 0.1601, "step": 3066 }, { "epoch": 0.2, "learning_rate": 0.00018578486942777243, "loss": 0.1588, "step": 3067 }, { "epoch": 0.2, "learning_rate": 0.0001857741955009774, "loss": 0.1553, "step": 3068 }, { "epoch": 0.2, "learning_rate": 0.00018576351787510538, "loss": 0.2097, "step": 3069 }, { "epoch": 0.2, "learning_rate": 0.00018575283655061688, "loss": 0.1956, "step": 3070 }, { "epoch": 0.2, "learning_rate": 0.0001857421515279726, "loss": 0.1585, "step": 3071 }, { "epoch": 0.2, "learning_rate": 0.00018573146280763324, "loss": 0.1879, "step": 3072 }, { "epoch": 0.2, "learning_rate": 0.00018572077039005983, "loss": 0.1458, "step": 3073 }, { "epoch": 0.2, "learning_rate": 0.0001857100742757135, "loss": 0.1668, "step": 3074 }, { "epoch": 0.2, "learning_rate": 0.00018569937446505545, "loss": 0.1477, "step": 3075 }, { "epoch": 0.2, "learning_rate": 0.0001856886709585472, "loss": 0.1636, "step": 3076 }, { "epoch": 0.2, "learning_rate": 0.0001856779637566503, "loss": 0.1549, "step": 3077 }, { "epoch": 0.2, "learning_rate": 0.00018566725285982654, "loss": 0.155, "step": 3078 }, { "epoch": 0.2, "learning_rate": 0.00018565653826853776, "loss": 0.1859, "step": 3079 }, { "epoch": 0.2, "learning_rate": 0.00018564581998324614, "loss": 0.1635, "step": 3080 }, { "epoch": 0.2, "learning_rate": 0.00018563509800441385, "loss": 0.1655, "step": 3081 }, { "epoch": 0.2, "learning_rate": 0.0001856243723325033, "loss": 0.1618, "step": 3082 }, { "epoch": 0.2, "learning_rate": 0.00018561364296797704, "loss": 0.1784, "step": 3083 }, { "epoch": 0.2, "learning_rate": 0.00018560290991129777, "loss": 0.1679, "step": 3084 }, { "epoch": 0.2, "learning_rate": 0.00018559217316292838, "loss": 0.1592, "step": 3085 }, { "epoch": 0.2, "learning_rate": 0.0001855814327233319, "loss": 0.1812, "step": 3086 }, { "epoch": 0.2, "learning_rate": 0.00018557068859297152, "loss": 0.1413, "step": 3087 }, { "epoch": 0.2, "learning_rate": 0.00018555994077231055, "loss": 0.1583, "step": 3088 }, { "epoch": 0.2, "learning_rate": 0.00018554918926181254, "loss": 0.1727, "step": 3089 }, { "epoch": 0.2, "learning_rate": 0.00018553843406194116, "loss": 0.1552, "step": 3090 }, { "epoch": 0.2, "learning_rate": 0.00018552767517316022, "loss": 0.1621, "step": 3091 }, { "epoch": 0.2, "learning_rate": 0.00018551691259593368, "loss": 0.1538, "step": 3092 }, { "epoch": 0.2, "learning_rate": 0.00018550614633072577, "loss": 0.1956, "step": 3093 }, { "epoch": 0.2, "learning_rate": 0.0001854953763780007, "loss": 0.1458, "step": 3094 }, { "epoch": 0.2, "learning_rate": 0.00018548460273822294, "loss": 0.1619, "step": 3095 }, { "epoch": 0.2, "learning_rate": 0.00018547382541185718, "loss": 0.1572, "step": 3096 }, { "epoch": 0.2, "learning_rate": 0.00018546304439936813, "loss": 0.1612, "step": 3097 }, { "epoch": 0.2, "learning_rate": 0.0001854522597012208, "loss": 0.1797, "step": 3098 }, { "epoch": 0.2, "learning_rate": 0.0001854414713178802, "loss": 0.1896, "step": 3099 }, { "epoch": 0.2, "learning_rate": 0.00018543067924981166, "loss": 0.1342, "step": 3100 }, { "epoch": 0.2, "learning_rate": 0.00018541988349748056, "loss": 0.1581, "step": 3101 }, { "epoch": 0.2, "learning_rate": 0.00018540908406135248, "loss": 0.1667, "step": 3102 }, { "epoch": 0.2, "learning_rate": 0.00018539828094189318, "loss": 0.17, "step": 3103 }, { "epoch": 0.2, "learning_rate": 0.00018538747413956853, "loss": 0.1549, "step": 3104 }, { "epoch": 0.2, "learning_rate": 0.00018537666365484458, "loss": 0.1669, "step": 3105 }, { "epoch": 0.2, "learning_rate": 0.00018536584948818753, "loss": 0.1715, "step": 3106 }, { "epoch": 0.2, "learning_rate": 0.00018535503164006375, "loss": 0.1568, "step": 3107 }, { "epoch": 0.2, "learning_rate": 0.00018534421011093982, "loss": 0.1729, "step": 3108 }, { "epoch": 0.2, "learning_rate": 0.00018533338490128238, "loss": 0.1546, "step": 3109 }, { "epoch": 0.2, "learning_rate": 0.00018532255601155825, "loss": 0.1715, "step": 3110 }, { "epoch": 0.2, "learning_rate": 0.0001853117234422345, "loss": 0.1566, "step": 3111 }, { "epoch": 0.2, "learning_rate": 0.00018530088719377825, "loss": 0.1444, "step": 3112 }, { "epoch": 0.2, "learning_rate": 0.00018529004726665683, "loss": 0.1501, "step": 3113 }, { "epoch": 0.2, "learning_rate": 0.00018527920366133769, "loss": 0.1711, "step": 3114 }, { "epoch": 0.2, "learning_rate": 0.00018526835637828852, "loss": 0.149, "step": 3115 }, { "epoch": 0.2, "learning_rate": 0.00018525750541797712, "loss": 0.1499, "step": 3116 }, { "epoch": 0.2, "learning_rate": 0.0001852466507808714, "loss": 0.1759, "step": 3117 }, { "epoch": 0.2, "learning_rate": 0.0001852357924674395, "loss": 0.1599, "step": 3118 }, { "epoch": 0.2, "learning_rate": 0.00018522493047814972, "loss": 0.1449, "step": 3119 }, { "epoch": 0.2, "learning_rate": 0.0001852140648134704, "loss": 0.1717, "step": 3120 }, { "epoch": 0.2, "learning_rate": 0.0001852031954738702, "loss": 0.1499, "step": 3121 }, { "epoch": 0.2, "learning_rate": 0.0001851923224598179, "loss": 0.1588, "step": 3122 }, { "epoch": 0.2, "learning_rate": 0.00018518144577178236, "loss": 0.1392, "step": 3123 }, { "epoch": 0.2, "learning_rate": 0.00018517056541023262, "loss": 0.1532, "step": 3124 }, { "epoch": 0.2, "learning_rate": 0.00018515968137563797, "loss": 0.151, "step": 3125 }, { "epoch": 0.2, "learning_rate": 0.00018514879366846775, "loss": 0.1626, "step": 3126 }, { "epoch": 0.2, "learning_rate": 0.0001851379022891915, "loss": 0.1539, "step": 3127 }, { "epoch": 0.2, "learning_rate": 0.00018512700723827892, "loss": 0.1582, "step": 3128 }, { "epoch": 0.2, "learning_rate": 0.0001851161085161999, "loss": 0.153, "step": 3129 }, { "epoch": 0.2, "learning_rate": 0.00018510520612342442, "loss": 0.1359, "step": 3130 }, { "epoch": 0.2, "learning_rate": 0.00018509430006042265, "loss": 0.1565, "step": 3131 }, { "epoch": 0.2, "learning_rate": 0.00018508339032766494, "loss": 0.1299, "step": 3132 }, { "epoch": 0.2, "learning_rate": 0.0001850724769256218, "loss": 0.1486, "step": 3133 }, { "epoch": 0.2, "learning_rate": 0.00018506155985476385, "loss": 0.1723, "step": 3134 }, { "epoch": 0.2, "learning_rate": 0.00018505063911556188, "loss": 0.1464, "step": 3135 }, { "epoch": 0.2, "learning_rate": 0.00018503971470848688, "loss": 0.173, "step": 3136 }, { "epoch": 0.2, "learning_rate": 0.00018502878663401, "loss": 0.1723, "step": 3137 }, { "epoch": 0.2, "learning_rate": 0.00018501785489260248, "loss": 0.1557, "step": 3138 }, { "epoch": 0.2, "learning_rate": 0.00018500691948473576, "loss": 0.1597, "step": 3139 }, { "epoch": 0.2, "learning_rate": 0.00018499598041088144, "loss": 0.1504, "step": 3140 }, { "epoch": 0.2, "learning_rate": 0.00018498503767151131, "loss": 0.1748, "step": 3141 }, { "epoch": 0.2, "learning_rate": 0.00018497409126709725, "loss": 0.1858, "step": 3142 }, { "epoch": 0.2, "learning_rate": 0.00018496314119811135, "loss": 0.1712, "step": 3143 }, { "epoch": 0.2, "learning_rate": 0.00018495218746502582, "loss": 0.1617, "step": 3144 }, { "epoch": 0.2, "learning_rate": 0.00018494123006831308, "loss": 0.1456, "step": 3145 }, { "epoch": 0.2, "learning_rate": 0.00018493026900844563, "loss": 0.1494, "step": 3146 }, { "epoch": 0.2, "learning_rate": 0.0001849193042858962, "loss": 0.1519, "step": 3147 }, { "epoch": 0.2, "learning_rate": 0.00018490833590113767, "loss": 0.1677, "step": 3148 }, { "epoch": 0.2, "learning_rate": 0.00018489736385464306, "loss": 0.1421, "step": 3149 }, { "epoch": 0.2, "learning_rate": 0.00018488638814688547, "loss": 0.1605, "step": 3150 }, { "epoch": 0.2, "learning_rate": 0.00018487540877833834, "loss": 0.1634, "step": 3151 }, { "epoch": 0.2, "learning_rate": 0.00018486442574947511, "loss": 0.1414, "step": 3152 }, { "epoch": 0.2, "learning_rate": 0.00018485343906076942, "loss": 0.1524, "step": 3153 }, { "epoch": 0.2, "learning_rate": 0.00018484244871269515, "loss": 0.1632, "step": 3154 }, { "epoch": 0.2, "learning_rate": 0.00018483145470572616, "loss": 0.1588, "step": 3155 }, { "epoch": 0.2, "learning_rate": 0.00018482045704033663, "loss": 0.1688, "step": 3156 }, { "epoch": 0.2, "learning_rate": 0.0001848094557170009, "loss": 0.1422, "step": 3157 }, { "epoch": 0.2, "learning_rate": 0.00018479845073619328, "loss": 0.1479, "step": 3158 }, { "epoch": 0.2, "learning_rate": 0.00018478744209838853, "loss": 0.1574, "step": 3159 }, { "epoch": 0.2, "learning_rate": 0.00018477642980406126, "loss": 0.1838, "step": 3160 }, { "epoch": 0.2, "learning_rate": 0.00018476541385368644, "loss": 0.1595, "step": 3161 }, { "epoch": 0.2, "learning_rate": 0.00018475439424773914, "loss": 0.158, "step": 3162 }, { "epoch": 0.2, "learning_rate": 0.00018474337098669461, "loss": 0.1548, "step": 3163 }, { "epoch": 0.2, "learning_rate": 0.0001847323440710282, "loss": 0.173, "step": 3164 }, { "epoch": 0.2, "learning_rate": 0.0001847213135012155, "loss": 0.1725, "step": 3165 }, { "epoch": 0.2, "learning_rate": 0.00018471027927773216, "loss": 0.1551, "step": 3166 }, { "epoch": 0.2, "learning_rate": 0.00018469924140105407, "loss": 0.1551, "step": 3167 }, { "epoch": 0.2, "learning_rate": 0.00018468819987165725, "loss": 0.1583, "step": 3168 }, { "epoch": 0.2, "learning_rate": 0.00018467715469001786, "loss": 0.1527, "step": 3169 }, { "epoch": 0.2, "learning_rate": 0.00018466610585661224, "loss": 0.1661, "step": 3170 }, { "epoch": 0.2, "learning_rate": 0.00018465505337191687, "loss": 0.1546, "step": 3171 }, { "epoch": 0.2, "learning_rate": 0.00018464399723640837, "loss": 0.1682, "step": 3172 }, { "epoch": 0.2, "learning_rate": 0.00018463293745056363, "loss": 0.1582, "step": 3173 }, { "epoch": 0.2, "learning_rate": 0.00018462187401485953, "loss": 0.1454, "step": 3174 }, { "epoch": 0.2, "learning_rate": 0.0001846108069297732, "loss": 0.1386, "step": 3175 }, { "epoch": 0.2, "learning_rate": 0.00018459973619578193, "loss": 0.157, "step": 3176 }, { "epoch": 0.2, "learning_rate": 0.00018458866181336321, "loss": 0.1583, "step": 3177 }, { "epoch": 0.2, "learning_rate": 0.00018457758378299453, "loss": 0.1567, "step": 3178 }, { "epoch": 0.2, "learning_rate": 0.0001845665021051537, "loss": 0.1593, "step": 3179 }, { "epoch": 0.2, "learning_rate": 0.0001845554167803186, "loss": 0.1709, "step": 3180 }, { "epoch": 0.2, "learning_rate": 0.00018454432780896734, "loss": 0.1583, "step": 3181 }, { "epoch": 0.2, "learning_rate": 0.0001845332351915781, "loss": 0.1762, "step": 3182 }, { "epoch": 0.2, "learning_rate": 0.00018452213892862921, "loss": 0.1607, "step": 3183 }, { "epoch": 0.2, "learning_rate": 0.0001845110390205993, "loss": 0.1399, "step": 3184 }, { "epoch": 0.2, "learning_rate": 0.00018449993546796703, "loss": 0.1811, "step": 3185 }, { "epoch": 0.2, "learning_rate": 0.0001844888282712112, "loss": 0.1585, "step": 3186 }, { "epoch": 0.2, "learning_rate": 0.00018447771743081092, "loss": 0.1444, "step": 3187 }, { "epoch": 0.2, "learning_rate": 0.0001844666029472452, "loss": 0.1566, "step": 3188 }, { "epoch": 0.2, "learning_rate": 0.00018445548482099347, "loss": 0.1513, "step": 3189 }, { "epoch": 0.2, "learning_rate": 0.00018444436305253524, "loss": 0.1636, "step": 3190 }, { "epoch": 0.2, "learning_rate": 0.00018443323764235007, "loss": 0.142, "step": 3191 }, { "epoch": 0.2, "learning_rate": 0.00018442210859091776, "loss": 0.1651, "step": 3192 }, { "epoch": 0.2, "learning_rate": 0.00018441097589871826, "loss": 0.1621, "step": 3193 }, { "epoch": 0.2, "learning_rate": 0.00018439983956623172, "loss": 0.1501, "step": 3194 }, { "epoch": 0.2, "learning_rate": 0.00018438869959393835, "loss": 0.1489, "step": 3195 }, { "epoch": 0.2, "learning_rate": 0.00018437755598231856, "loss": 0.1575, "step": 3196 }, { "epoch": 0.2, "learning_rate": 0.000184366408731853, "loss": 0.1593, "step": 3197 }, { "epoch": 0.21, "learning_rate": 0.00018435525784302236, "loss": 0.145, "step": 3198 }, { "epoch": 0.21, "learning_rate": 0.0001843441033163075, "loss": 0.1836, "step": 3199 }, { "epoch": 0.21, "learning_rate": 0.0001843329451521895, "loss": 0.151, "step": 3200 }, { "epoch": 0.21, "learning_rate": 0.0001843217833511496, "loss": 0.1536, "step": 3201 }, { "epoch": 0.21, "learning_rate": 0.0001843106179136691, "loss": 0.1628, "step": 3202 }, { "epoch": 0.21, "learning_rate": 0.00018429944884022951, "loss": 0.1602, "step": 3203 }, { "epoch": 0.21, "learning_rate": 0.0001842882761313126, "loss": 0.151, "step": 3204 }, { "epoch": 0.21, "learning_rate": 0.00018427709978740007, "loss": 0.1361, "step": 3205 }, { "epoch": 0.21, "learning_rate": 0.00018426591980897402, "loss": 0.1558, "step": 3206 }, { "epoch": 0.21, "learning_rate": 0.00018425473619651655, "loss": 0.1593, "step": 3207 }, { "epoch": 0.21, "learning_rate": 0.00018424354895050994, "loss": 0.1723, "step": 3208 }, { "epoch": 0.21, "learning_rate": 0.00018423235807143667, "loss": 0.1516, "step": 3209 }, { "epoch": 0.21, "learning_rate": 0.0001842211635597794, "loss": 0.1557, "step": 3210 }, { "epoch": 0.21, "learning_rate": 0.0001842099654160208, "loss": 0.1547, "step": 3211 }, { "epoch": 0.21, "learning_rate": 0.00018419876364064388, "loss": 0.1726, "step": 3212 }, { "epoch": 0.21, "learning_rate": 0.0001841875582341317, "loss": 0.1548, "step": 3213 }, { "epoch": 0.21, "learning_rate": 0.00018417634919696752, "loss": 0.1804, "step": 3214 }, { "epoch": 0.21, "learning_rate": 0.0001841651365296347, "loss": 0.1557, "step": 3215 }, { "epoch": 0.21, "learning_rate": 0.00018415392023261683, "loss": 0.1397, "step": 3216 }, { "epoch": 0.21, "learning_rate": 0.00018414270030639757, "loss": 0.137, "step": 3217 }, { "epoch": 0.21, "learning_rate": 0.00018413147675146085, "loss": 0.1527, "step": 3218 }, { "epoch": 0.21, "learning_rate": 0.0001841202495682907, "loss": 0.1427, "step": 3219 }, { "epoch": 0.21, "learning_rate": 0.0001841090187573712, "loss": 0.1622, "step": 3220 }, { "epoch": 0.21, "learning_rate": 0.0001840977843191868, "loss": 0.1563, "step": 3221 }, { "epoch": 0.21, "learning_rate": 0.00018408654625422198, "loss": 0.1665, "step": 3222 }, { "epoch": 0.21, "learning_rate": 0.00018407530456296132, "loss": 0.1371, "step": 3223 }, { "epoch": 0.21, "learning_rate": 0.00018406405924588967, "loss": 0.1509, "step": 3224 }, { "epoch": 0.21, "learning_rate": 0.00018405281030349198, "loss": 0.1498, "step": 3225 }, { "epoch": 0.21, "learning_rate": 0.0001840415577362534, "loss": 0.1434, "step": 3226 }, { "epoch": 0.21, "learning_rate": 0.00018403030154465915, "loss": 0.1833, "step": 3227 }, { "epoch": 0.21, "learning_rate": 0.00018401904172919475, "loss": 0.145, "step": 3228 }, { "epoch": 0.21, "learning_rate": 0.00018400777829034572, "loss": 0.1452, "step": 3229 }, { "epoch": 0.21, "learning_rate": 0.00018399651122859778, "loss": 0.1532, "step": 3230 }, { "epoch": 0.21, "learning_rate": 0.00018398524054443692, "loss": 0.165, "step": 3231 }, { "epoch": 0.21, "learning_rate": 0.0001839739662383491, "loss": 0.1676, "step": 3232 }, { "epoch": 0.21, "learning_rate": 0.0001839626883108206, "loss": 0.1367, "step": 3233 }, { "epoch": 0.21, "learning_rate": 0.00018395140676233775, "loss": 0.1716, "step": 3234 }, { "epoch": 0.21, "learning_rate": 0.00018394012159338714, "loss": 0.1696, "step": 3235 }, { "epoch": 0.21, "learning_rate": 0.00018392883280445537, "loss": 0.1666, "step": 3236 }, { "epoch": 0.21, "learning_rate": 0.0001839175403960293, "loss": 0.1682, "step": 3237 }, { "epoch": 0.21, "learning_rate": 0.00018390624436859598, "loss": 0.1545, "step": 3238 }, { "epoch": 0.21, "learning_rate": 0.0001838949447226425, "loss": 0.1391, "step": 3239 }, { "epoch": 0.21, "learning_rate": 0.00018388364145865613, "loss": 0.1568, "step": 3240 }, { "epoch": 0.21, "learning_rate": 0.00018387233457712443, "loss": 0.1716, "step": 3241 }, { "epoch": 0.21, "learning_rate": 0.00018386102407853496, "loss": 0.1438, "step": 3242 }, { "epoch": 0.21, "learning_rate": 0.0001838497099633755, "loss": 0.1673, "step": 3243 }, { "epoch": 0.21, "learning_rate": 0.000183838392232134, "loss": 0.1721, "step": 3244 }, { "epoch": 0.21, "learning_rate": 0.0001838270708852985, "loss": 0.1619, "step": 3245 }, { "epoch": 0.21, "learning_rate": 0.0001838157459233573, "loss": 0.1527, "step": 3246 }, { "epoch": 0.21, "learning_rate": 0.00018380441734679875, "loss": 0.1452, "step": 3247 }, { "epoch": 0.21, "learning_rate": 0.00018379308515611146, "loss": 0.1493, "step": 3248 }, { "epoch": 0.21, "learning_rate": 0.00018378174935178407, "loss": 0.1571, "step": 3249 }, { "epoch": 0.21, "learning_rate": 0.00018377040993430548, "loss": 0.1605, "step": 3250 }, { "epoch": 0.21, "learning_rate": 0.00018375906690416472, "loss": 0.1346, "step": 3251 }, { "epoch": 0.21, "learning_rate": 0.00018374772026185095, "loss": 0.177, "step": 3252 }, { "epoch": 0.21, "learning_rate": 0.0001837363700078535, "loss": 0.174, "step": 3253 }, { "epoch": 0.21, "learning_rate": 0.00018372501614266184, "loss": 0.1754, "step": 3254 }, { "epoch": 0.21, "learning_rate": 0.0001837136586667657, "loss": 0.1617, "step": 3255 }, { "epoch": 0.21, "learning_rate": 0.00018370229758065478, "loss": 0.1641, "step": 3256 }, { "epoch": 0.21, "learning_rate": 0.0001836909328848191, "loss": 0.1651, "step": 3257 }, { "epoch": 0.21, "learning_rate": 0.00018367956457974872, "loss": 0.1891, "step": 3258 }, { "epoch": 0.21, "learning_rate": 0.0001836681926659339, "loss": 0.1507, "step": 3259 }, { "epoch": 0.21, "learning_rate": 0.00018365681714386516, "loss": 0.1648, "step": 3260 }, { "epoch": 0.21, "learning_rate": 0.00018364543801403297, "loss": 0.1724, "step": 3261 }, { "epoch": 0.21, "learning_rate": 0.00018363405527692815, "loss": 0.1683, "step": 3262 }, { "epoch": 0.21, "learning_rate": 0.00018362266893304148, "loss": 0.1605, "step": 3263 }, { "epoch": 0.21, "learning_rate": 0.00018361127898286416, "loss": 0.1726, "step": 3264 }, { "epoch": 0.21, "learning_rate": 0.00018359988542688722, "loss": 0.1703, "step": 3265 }, { "epoch": 0.21, "learning_rate": 0.00018358848826560212, "loss": 0.1435, "step": 3266 }, { "epoch": 0.21, "learning_rate": 0.00018357708749950038, "loss": 0.1792, "step": 3267 }, { "epoch": 0.21, "learning_rate": 0.0001835656831290736, "loss": 0.1629, "step": 3268 }, { "epoch": 0.21, "learning_rate": 0.00018355427515481365, "loss": 0.1425, "step": 3269 }, { "epoch": 0.21, "learning_rate": 0.00018354286357721248, "loss": 0.1474, "step": 3270 }, { "epoch": 0.21, "learning_rate": 0.00018353144839676223, "loss": 0.1552, "step": 3271 }, { "epoch": 0.21, "learning_rate": 0.0001835200296139552, "loss": 0.1582, "step": 3272 }, { "epoch": 0.21, "learning_rate": 0.00018350860722928384, "loss": 0.1458, "step": 3273 }, { "epoch": 0.21, "learning_rate": 0.00018349718124324076, "loss": 0.1402, "step": 3274 }, { "epoch": 0.21, "learning_rate": 0.00018348575165631864, "loss": 0.1496, "step": 3275 }, { "epoch": 0.21, "learning_rate": 0.00018347431846901048, "loss": 0.175, "step": 3276 }, { "epoch": 0.21, "learning_rate": 0.00018346288168180928, "loss": 0.16, "step": 3277 }, { "epoch": 0.21, "learning_rate": 0.0001834514412952083, "loss": 0.1545, "step": 3278 }, { "epoch": 0.21, "learning_rate": 0.00018343999730970092, "loss": 0.1498, "step": 3279 }, { "epoch": 0.21, "learning_rate": 0.00018342854972578062, "loss": 0.1612, "step": 3280 }, { "epoch": 0.21, "learning_rate": 0.00018341709854394114, "loss": 0.1583, "step": 3281 }, { "epoch": 0.21, "learning_rate": 0.0001834056437646763, "loss": 0.1593, "step": 3282 }, { "epoch": 0.21, "learning_rate": 0.00018339418538848007, "loss": 0.1617, "step": 3283 }, { "epoch": 0.21, "learning_rate": 0.00018338272341584662, "loss": 0.1629, "step": 3284 }, { "epoch": 0.21, "learning_rate": 0.0001833712578472703, "loss": 0.1407, "step": 3285 }, { "epoch": 0.21, "learning_rate": 0.0001833597886832455, "loss": 0.1756, "step": 3286 }, { "epoch": 0.21, "learning_rate": 0.00018334831592426687, "loss": 0.1685, "step": 3287 }, { "epoch": 0.21, "learning_rate": 0.0001833368395708292, "loss": 0.1436, "step": 3288 }, { "epoch": 0.21, "learning_rate": 0.0001833253596234274, "loss": 0.1482, "step": 3289 }, { "epoch": 0.21, "learning_rate": 0.00018331387608255651, "loss": 0.1885, "step": 3290 }, { "epoch": 0.21, "learning_rate": 0.00018330238894871185, "loss": 0.171, "step": 3291 }, { "epoch": 0.21, "learning_rate": 0.00018329089822238874, "loss": 0.1569, "step": 3292 }, { "epoch": 0.21, "learning_rate": 0.0001832794039040828, "loss": 0.1354, "step": 3293 }, { "epoch": 0.21, "learning_rate": 0.00018326790599428963, "loss": 0.1624, "step": 3294 }, { "epoch": 0.21, "learning_rate": 0.00018325640449350515, "loss": 0.1484, "step": 3295 }, { "epoch": 0.21, "learning_rate": 0.00018324489940222538, "loss": 0.1876, "step": 3296 }, { "epoch": 0.21, "learning_rate": 0.00018323339072094644, "loss": 0.1469, "step": 3297 }, { "epoch": 0.21, "learning_rate": 0.0001832218784501647, "loss": 0.1944, "step": 3298 }, { "epoch": 0.21, "learning_rate": 0.0001832103625903766, "loss": 0.1448, "step": 3299 }, { "epoch": 0.21, "learning_rate": 0.00018319884314207878, "loss": 0.1726, "step": 3300 }, { "epoch": 0.21, "learning_rate": 0.00018318732010576803, "loss": 0.1676, "step": 3301 }, { "epoch": 0.21, "learning_rate": 0.0001831757934819413, "loss": 0.1566, "step": 3302 }, { "epoch": 0.21, "learning_rate": 0.00018316426327109566, "loss": 0.1669, "step": 3303 }, { "epoch": 0.21, "learning_rate": 0.00018315272947372837, "loss": 0.1928, "step": 3304 }, { "epoch": 0.21, "learning_rate": 0.00018314119209033686, "loss": 0.1322, "step": 3305 }, { "epoch": 0.21, "learning_rate": 0.0001831296511214186, "loss": 0.1621, "step": 3306 }, { "epoch": 0.21, "learning_rate": 0.00018311810656747142, "loss": 0.1632, "step": 3307 }, { "epoch": 0.21, "learning_rate": 0.00018310655842899312, "loss": 0.151, "step": 3308 }, { "epoch": 0.21, "learning_rate": 0.00018309500670648173, "loss": 0.1767, "step": 3309 }, { "epoch": 0.21, "learning_rate": 0.00018308345140043543, "loss": 0.1792, "step": 3310 }, { "epoch": 0.21, "learning_rate": 0.00018307189251135254, "loss": 0.1648, "step": 3311 }, { "epoch": 0.21, "learning_rate": 0.0001830603300397316, "loss": 0.1611, "step": 3312 }, { "epoch": 0.21, "learning_rate": 0.00018304876398607118, "loss": 0.146, "step": 3313 }, { "epoch": 0.21, "learning_rate": 0.0001830371943508701, "loss": 0.1768, "step": 3314 }, { "epoch": 0.21, "learning_rate": 0.00018302562113462734, "loss": 0.1679, "step": 3315 }, { "epoch": 0.21, "learning_rate": 0.00018301404433784194, "loss": 0.1408, "step": 3316 }, { "epoch": 0.21, "learning_rate": 0.00018300246396101322, "loss": 0.139, "step": 3317 }, { "epoch": 0.21, "learning_rate": 0.00018299088000464056, "loss": 0.152, "step": 3318 }, { "epoch": 0.21, "learning_rate": 0.00018297929246922356, "loss": 0.1755, "step": 3319 }, { "epoch": 0.21, "learning_rate": 0.0001829677013552619, "loss": 0.1743, "step": 3320 }, { "epoch": 0.21, "learning_rate": 0.00018295610666325547, "loss": 0.1603, "step": 3321 }, { "epoch": 0.21, "learning_rate": 0.0001829445083937043, "loss": 0.1673, "step": 3322 }, { "epoch": 0.21, "learning_rate": 0.00018293290654710858, "loss": 0.182, "step": 3323 }, { "epoch": 0.21, "learning_rate": 0.00018292130112396865, "loss": 0.159, "step": 3324 }, { "epoch": 0.21, "learning_rate": 0.00018290969212478498, "loss": 0.1568, "step": 3325 }, { "epoch": 0.21, "learning_rate": 0.00018289807955005821, "loss": 0.1647, "step": 3326 }, { "epoch": 0.21, "learning_rate": 0.00018288646340028922, "loss": 0.1437, "step": 3327 }, { "epoch": 0.21, "learning_rate": 0.00018287484367597888, "loss": 0.1494, "step": 3328 }, { "epoch": 0.21, "learning_rate": 0.00018286322037762832, "loss": 0.1473, "step": 3329 }, { "epoch": 0.21, "learning_rate": 0.00018285159350573885, "loss": 0.1437, "step": 3330 }, { "epoch": 0.21, "learning_rate": 0.0001828399630608118, "loss": 0.1568, "step": 3331 }, { "epoch": 0.21, "learning_rate": 0.00018282832904334883, "loss": 0.164, "step": 3332 }, { "epoch": 0.21, "learning_rate": 0.00018281669145385162, "loss": 0.1502, "step": 3333 }, { "epoch": 0.21, "learning_rate": 0.000182805050292822, "loss": 0.1712, "step": 3334 }, { "epoch": 0.21, "learning_rate": 0.00018279340556076216, "loss": 0.1529, "step": 3335 }, { "epoch": 0.21, "learning_rate": 0.00018278175725817413, "loss": 0.1652, "step": 3336 }, { "epoch": 0.21, "learning_rate": 0.00018277010538556035, "loss": 0.1505, "step": 3337 }, { "epoch": 0.21, "learning_rate": 0.00018275844994342322, "loss": 0.1521, "step": 3338 }, { "epoch": 0.21, "learning_rate": 0.00018274679093226547, "loss": 0.1564, "step": 3339 }, { "epoch": 0.21, "learning_rate": 0.0001827351283525899, "loss": 0.1685, "step": 3340 }, { "epoch": 0.21, "learning_rate": 0.00018272346220489944, "loss": 0.1478, "step": 3341 }, { "epoch": 0.21, "learning_rate": 0.0001827117924896972, "loss": 0.1702, "step": 3342 }, { "epoch": 0.21, "learning_rate": 0.00018270011920748646, "loss": 0.157, "step": 3343 }, { "epoch": 0.21, "learning_rate": 0.00018268844235877062, "loss": 0.144, "step": 3344 }, { "epoch": 0.21, "learning_rate": 0.00018267676194405329, "loss": 0.17, "step": 3345 }, { "epoch": 0.21, "learning_rate": 0.00018266507796383817, "loss": 0.1683, "step": 3346 }, { "epoch": 0.21, "learning_rate": 0.00018265339041862914, "loss": 0.1624, "step": 3347 }, { "epoch": 0.21, "learning_rate": 0.00018264169930893022, "loss": 0.1476, "step": 3348 }, { "epoch": 0.21, "learning_rate": 0.00018263000463524564, "loss": 0.1523, "step": 3349 }, { "epoch": 0.21, "learning_rate": 0.0001826183063980797, "loss": 0.1673, "step": 3350 }, { "epoch": 0.21, "learning_rate": 0.00018260660459793693, "loss": 0.1418, "step": 3351 }, { "epoch": 0.21, "learning_rate": 0.00018259489923532195, "loss": 0.151, "step": 3352 }, { "epoch": 0.21, "learning_rate": 0.00018258319031073958, "loss": 0.1671, "step": 3353 }, { "epoch": 0.22, "learning_rate": 0.00018257147782469477, "loss": 0.1667, "step": 3354 }, { "epoch": 0.22, "learning_rate": 0.00018255976177769263, "loss": 0.1712, "step": 3355 }, { "epoch": 0.22, "learning_rate": 0.00018254804217023843, "loss": 0.1671, "step": 3356 }, { "epoch": 0.22, "learning_rate": 0.00018253631900283756, "loss": 0.1391, "step": 3357 }, { "epoch": 0.22, "learning_rate": 0.00018252459227599563, "loss": 0.158, "step": 3358 }, { "epoch": 0.22, "learning_rate": 0.00018251286199021834, "loss": 0.1538, "step": 3359 }, { "epoch": 0.22, "learning_rate": 0.00018250112814601158, "loss": 0.1714, "step": 3360 }, { "epoch": 0.22, "learning_rate": 0.00018248939074388135, "loss": 0.1541, "step": 3361 }, { "epoch": 0.22, "learning_rate": 0.00018247764978433388, "loss": 0.1652, "step": 3362 }, { "epoch": 0.22, "learning_rate": 0.00018246590526787548, "loss": 0.1557, "step": 3363 }, { "epoch": 0.22, "learning_rate": 0.00018245415719501265, "loss": 0.1492, "step": 3364 }, { "epoch": 0.22, "learning_rate": 0.00018244240556625202, "loss": 0.1833, "step": 3365 }, { "epoch": 0.22, "learning_rate": 0.00018243065038210042, "loss": 0.1541, "step": 3366 }, { "epoch": 0.22, "learning_rate": 0.00018241889164306475, "loss": 0.1537, "step": 3367 }, { "epoch": 0.22, "learning_rate": 0.00018240712934965218, "loss": 0.1591, "step": 3368 }, { "epoch": 0.22, "learning_rate": 0.0001823953635023699, "loss": 0.1629, "step": 3369 }, { "epoch": 0.22, "learning_rate": 0.00018238359410172535, "loss": 0.1581, "step": 3370 }, { "epoch": 0.22, "learning_rate": 0.00018237182114822612, "loss": 0.1591, "step": 3371 }, { "epoch": 0.22, "learning_rate": 0.00018236004464237992, "loss": 0.1578, "step": 3372 }, { "epoch": 0.22, "learning_rate": 0.00018234826458469458, "loss": 0.1527, "step": 3373 }, { "epoch": 0.22, "learning_rate": 0.00018233648097567816, "loss": 0.1755, "step": 3374 }, { "epoch": 0.22, "learning_rate": 0.0001823246938158388, "loss": 0.1551, "step": 3375 }, { "epoch": 0.22, "learning_rate": 0.0001823129031056849, "loss": 0.1428, "step": 3376 }, { "epoch": 0.22, "learning_rate": 0.0001823011088457249, "loss": 0.1661, "step": 3377 }, { "epoch": 0.22, "learning_rate": 0.0001822893110364674, "loss": 0.1604, "step": 3378 }, { "epoch": 0.22, "learning_rate": 0.00018227750967842124, "loss": 0.1587, "step": 3379 }, { "epoch": 0.22, "learning_rate": 0.00018226570477209534, "loss": 0.1528, "step": 3380 }, { "epoch": 0.22, "learning_rate": 0.00018225389631799883, "loss": 0.1696, "step": 3381 }, { "epoch": 0.22, "learning_rate": 0.00018224208431664093, "loss": 0.1707, "step": 3382 }, { "epoch": 0.22, "learning_rate": 0.00018223026876853103, "loss": 0.1521, "step": 3383 }, { "epoch": 0.22, "learning_rate": 0.0001822184496741787, "loss": 0.1527, "step": 3384 }, { "epoch": 0.22, "learning_rate": 0.00018220662703409367, "loss": 0.1682, "step": 3385 }, { "epoch": 0.22, "learning_rate": 0.00018219480084878575, "loss": 0.1783, "step": 3386 }, { "epoch": 0.22, "learning_rate": 0.000182182971118765, "loss": 0.173, "step": 3387 }, { "epoch": 0.22, "learning_rate": 0.00018217113784454151, "loss": 0.143, "step": 3388 }, { "epoch": 0.22, "learning_rate": 0.00018215930102662573, "loss": 0.1572, "step": 3389 }, { "epoch": 0.22, "learning_rate": 0.000182147460665528, "loss": 0.1514, "step": 3390 }, { "epoch": 0.22, "learning_rate": 0.00018213561676175903, "loss": 0.1613, "step": 3391 }, { "epoch": 0.22, "learning_rate": 0.00018212376931582956, "loss": 0.1686, "step": 3392 }, { "epoch": 0.22, "learning_rate": 0.0001821119183282505, "loss": 0.1649, "step": 3393 }, { "epoch": 0.22, "learning_rate": 0.00018210006379953302, "loss": 0.1714, "step": 3394 }, { "epoch": 0.22, "learning_rate": 0.00018208820573018827, "loss": 0.1645, "step": 3395 }, { "epoch": 0.22, "learning_rate": 0.00018207634412072764, "loss": 0.155, "step": 3396 }, { "epoch": 0.22, "learning_rate": 0.00018206447897166272, "loss": 0.1486, "step": 3397 }, { "epoch": 0.22, "learning_rate": 0.00018205261028350517, "loss": 0.173, "step": 3398 }, { "epoch": 0.22, "learning_rate": 0.00018204073805676684, "loss": 0.1806, "step": 3399 }, { "epoch": 0.22, "learning_rate": 0.00018202886229195975, "loss": 0.1664, "step": 3400 }, { "epoch": 0.22, "learning_rate": 0.000182016982989596, "loss": 0.1659, "step": 3401 }, { "epoch": 0.22, "learning_rate": 0.00018200510015018798, "loss": 0.1824, "step": 3402 }, { "epoch": 0.22, "learning_rate": 0.00018199321377424806, "loss": 0.1604, "step": 3403 }, { "epoch": 0.22, "learning_rate": 0.0001819813238622889, "loss": 0.156, "step": 3404 }, { "epoch": 0.22, "learning_rate": 0.00018196943041482325, "loss": 0.1731, "step": 3405 }, { "epoch": 0.22, "learning_rate": 0.00018195753343236401, "loss": 0.1543, "step": 3406 }, { "epoch": 0.22, "learning_rate": 0.00018194563291542426, "loss": 0.1527, "step": 3407 }, { "epoch": 0.22, "learning_rate": 0.0001819337288645172, "loss": 0.1523, "step": 3408 }, { "epoch": 0.22, "learning_rate": 0.00018192182128015625, "loss": 0.1628, "step": 3409 }, { "epoch": 0.22, "learning_rate": 0.00018190991016285487, "loss": 0.1592, "step": 3410 }, { "epoch": 0.22, "learning_rate": 0.00018189799551312675, "loss": 0.1492, "step": 3411 }, { "epoch": 0.22, "learning_rate": 0.00018188607733148578, "loss": 0.1503, "step": 3412 }, { "epoch": 0.22, "learning_rate": 0.00018187415561844586, "loss": 0.1802, "step": 3413 }, { "epoch": 0.22, "learning_rate": 0.00018186223037452116, "loss": 0.1474, "step": 3414 }, { "epoch": 0.22, "learning_rate": 0.00018185030160022597, "loss": 0.165, "step": 3415 }, { "epoch": 0.22, "learning_rate": 0.00018183836929607473, "loss": 0.1505, "step": 3416 }, { "epoch": 0.22, "learning_rate": 0.00018182643346258202, "loss": 0.1662, "step": 3417 }, { "epoch": 0.22, "learning_rate": 0.00018181449410026255, "loss": 0.1535, "step": 3418 }, { "epoch": 0.22, "learning_rate": 0.00018180255120963128, "loss": 0.1729, "step": 3419 }, { "epoch": 0.22, "learning_rate": 0.0001817906047912032, "loss": 0.1711, "step": 3420 }, { "epoch": 0.22, "learning_rate": 0.00018177865484549354, "loss": 0.1779, "step": 3421 }, { "epoch": 0.22, "learning_rate": 0.00018176670137301764, "loss": 0.1417, "step": 3422 }, { "epoch": 0.22, "learning_rate": 0.00018175474437429097, "loss": 0.1456, "step": 3423 }, { "epoch": 0.22, "learning_rate": 0.00018174278384982926, "loss": 0.1372, "step": 3424 }, { "epoch": 0.22, "learning_rate": 0.00018173081980014827, "loss": 0.1319, "step": 3425 }, { "epoch": 0.22, "learning_rate": 0.00018171885222576395, "loss": 0.1501, "step": 3426 }, { "epoch": 0.22, "learning_rate": 0.00018170688112719243, "loss": 0.1926, "step": 3427 }, { "epoch": 0.22, "learning_rate": 0.00018169490650494995, "loss": 0.1393, "step": 3428 }, { "epoch": 0.22, "learning_rate": 0.00018168292835955295, "loss": 0.1576, "step": 3429 }, { "epoch": 0.22, "learning_rate": 0.000181670946691518, "loss": 0.1828, "step": 3430 }, { "epoch": 0.22, "learning_rate": 0.00018165896150136178, "loss": 0.1386, "step": 3431 }, { "epoch": 0.22, "learning_rate": 0.00018164697278960122, "loss": 0.1639, "step": 3432 }, { "epoch": 0.22, "learning_rate": 0.0001816349805567533, "loss": 0.1535, "step": 3433 }, { "epoch": 0.22, "learning_rate": 0.0001816229848033352, "loss": 0.1593, "step": 3434 }, { "epoch": 0.22, "learning_rate": 0.00018161098552986422, "loss": 0.1532, "step": 3435 }, { "epoch": 0.22, "learning_rate": 0.00018159898273685787, "loss": 0.1626, "step": 3436 }, { "epoch": 0.22, "learning_rate": 0.00018158697642483383, "loss": 0.1601, "step": 3437 }, { "epoch": 0.22, "learning_rate": 0.00018157496659430974, "loss": 0.1773, "step": 3438 }, { "epoch": 0.22, "learning_rate": 0.00018156295324580368, "loss": 0.1641, "step": 3439 }, { "epoch": 0.22, "learning_rate": 0.00018155093637983367, "loss": 0.1487, "step": 3440 }, { "epoch": 0.22, "learning_rate": 0.00018153891599691793, "loss": 0.1381, "step": 3441 }, { "epoch": 0.22, "learning_rate": 0.0001815268920975749, "loss": 0.1558, "step": 3442 }, { "epoch": 0.22, "learning_rate": 0.00018151486468232305, "loss": 0.1513, "step": 3443 }, { "epoch": 0.22, "learning_rate": 0.00018150283375168114, "loss": 0.1603, "step": 3444 }, { "epoch": 0.22, "learning_rate": 0.000181490799306168, "loss": 0.1621, "step": 3445 }, { "epoch": 0.22, "learning_rate": 0.00018147876134630257, "loss": 0.1789, "step": 3446 }, { "epoch": 0.22, "learning_rate": 0.00018146671987260405, "loss": 0.1568, "step": 3447 }, { "epoch": 0.22, "learning_rate": 0.0001814546748855917, "loss": 0.1549, "step": 3448 }, { "epoch": 0.22, "learning_rate": 0.00018144262638578499, "loss": 0.1462, "step": 3449 }, { "epoch": 0.22, "learning_rate": 0.00018143057437370358, "loss": 0.1323, "step": 3450 }, { "epoch": 0.22, "learning_rate": 0.00018141851884986713, "loss": 0.1694, "step": 3451 }, { "epoch": 0.22, "learning_rate": 0.00018140645981479554, "loss": 0.1534, "step": 3452 }, { "epoch": 0.22, "learning_rate": 0.0001813943972690089, "loss": 0.1537, "step": 3453 }, { "epoch": 0.22, "learning_rate": 0.00018138233121302747, "loss": 0.1619, "step": 3454 }, { "epoch": 0.22, "learning_rate": 0.00018137026164737155, "loss": 0.1302, "step": 3455 }, { "epoch": 0.22, "learning_rate": 0.00018135818857256163, "loss": 0.1656, "step": 3456 }, { "epoch": 0.22, "learning_rate": 0.0001813461119891184, "loss": 0.1886, "step": 3457 }, { "epoch": 0.22, "learning_rate": 0.00018133403189756267, "loss": 0.1542, "step": 3458 }, { "epoch": 0.22, "learning_rate": 0.0001813219482984154, "loss": 0.1913, "step": 3459 }, { "epoch": 0.22, "learning_rate": 0.00018130986119219774, "loss": 0.1719, "step": 3460 }, { "epoch": 0.22, "learning_rate": 0.00018129777057943085, "loss": 0.1709, "step": 3461 }, { "epoch": 0.22, "learning_rate": 0.00018128567646063628, "loss": 0.1619, "step": 3462 }, { "epoch": 0.22, "learning_rate": 0.0001812735788363355, "loss": 0.1598, "step": 3463 }, { "epoch": 0.22, "learning_rate": 0.00018126147770705027, "loss": 0.1408, "step": 3464 }, { "epoch": 0.22, "learning_rate": 0.00018124937307330245, "loss": 0.1744, "step": 3465 }, { "epoch": 0.22, "learning_rate": 0.00018123726493561408, "loss": 0.1547, "step": 3466 }, { "epoch": 0.22, "learning_rate": 0.00018122515329450732, "loss": 0.159, "step": 3467 }, { "epoch": 0.22, "learning_rate": 0.00018121303815050449, "loss": 0.1698, "step": 3468 }, { "epoch": 0.22, "learning_rate": 0.00018120091950412804, "loss": 0.1475, "step": 3469 }, { "epoch": 0.22, "learning_rate": 0.00018118879735590064, "loss": 0.1483, "step": 3470 }, { "epoch": 0.22, "learning_rate": 0.0001811766717063451, "loss": 0.1581, "step": 3471 }, { "epoch": 0.22, "learning_rate": 0.00018116454255598423, "loss": 0.1409, "step": 3472 }, { "epoch": 0.22, "learning_rate": 0.0001811524099053412, "loss": 0.157, "step": 3473 }, { "epoch": 0.22, "learning_rate": 0.0001811402737549392, "loss": 0.1762, "step": 3474 }, { "epoch": 0.22, "learning_rate": 0.00018112813410530167, "loss": 0.1551, "step": 3475 }, { "epoch": 0.22, "learning_rate": 0.00018111599095695205, "loss": 0.1363, "step": 3476 }, { "epoch": 0.22, "learning_rate": 0.0001811038443104141, "loss": 0.1756, "step": 3477 }, { "epoch": 0.22, "learning_rate": 0.0001810916941662116, "loss": 0.1497, "step": 3478 }, { "epoch": 0.22, "learning_rate": 0.0001810795405248686, "loss": 0.151, "step": 3479 }, { "epoch": 0.22, "learning_rate": 0.00018106738338690917, "loss": 0.1611, "step": 3480 }, { "epoch": 0.22, "learning_rate": 0.0001810552227528576, "loss": 0.1626, "step": 3481 }, { "epoch": 0.22, "learning_rate": 0.00018104305862323835, "loss": 0.1491, "step": 3482 }, { "epoch": 0.22, "learning_rate": 0.00018103089099857602, "loss": 0.1642, "step": 3483 }, { "epoch": 0.22, "learning_rate": 0.00018101871987939535, "loss": 0.1526, "step": 3484 }, { "epoch": 0.22, "learning_rate": 0.00018100654526622117, "loss": 0.1564, "step": 3485 }, { "epoch": 0.22, "learning_rate": 0.0001809943671595786, "loss": 0.1614, "step": 3486 }, { "epoch": 0.22, "learning_rate": 0.00018098218555999274, "loss": 0.1491, "step": 3487 }, { "epoch": 0.22, "learning_rate": 0.00018097000046798898, "loss": 0.1681, "step": 3488 }, { "epoch": 0.22, "learning_rate": 0.00018095781188409284, "loss": 0.1525, "step": 3489 }, { "epoch": 0.22, "learning_rate": 0.00018094561980882992, "loss": 0.1538, "step": 3490 }, { "epoch": 0.22, "learning_rate": 0.00018093342424272603, "loss": 0.1664, "step": 3491 }, { "epoch": 0.22, "learning_rate": 0.00018092122518630709, "loss": 0.1759, "step": 3492 }, { "epoch": 0.22, "learning_rate": 0.00018090902264009925, "loss": 0.151, "step": 3493 }, { "epoch": 0.22, "learning_rate": 0.0001808968166046287, "loss": 0.1482, "step": 3494 }, { "epoch": 0.22, "learning_rate": 0.00018088460708042182, "loss": 0.1498, "step": 3495 }, { "epoch": 0.22, "learning_rate": 0.00018087239406800522, "loss": 0.1562, "step": 3496 }, { "epoch": 0.22, "learning_rate": 0.00018086017756790556, "loss": 0.1519, "step": 3497 }, { "epoch": 0.22, "learning_rate": 0.0001808479575806497, "loss": 0.1651, "step": 3498 }, { "epoch": 0.22, "learning_rate": 0.00018083573410676457, "loss": 0.1378, "step": 3499 }, { "epoch": 0.22, "learning_rate": 0.00018082350714677739, "loss": 0.1552, "step": 3500 }, { "epoch": 0.22, "learning_rate": 0.00018081127670121545, "loss": 0.1535, "step": 3501 }, { "epoch": 0.22, "learning_rate": 0.00018079904277060616, "loss": 0.1594, "step": 3502 }, { "epoch": 0.22, "learning_rate": 0.00018078680535547719, "loss": 0.1661, "step": 3503 }, { "epoch": 0.22, "learning_rate": 0.0001807745644563562, "loss": 0.1745, "step": 3504 }, { "epoch": 0.22, "learning_rate": 0.00018076232007377112, "loss": 0.1653, "step": 3505 }, { "epoch": 0.22, "learning_rate": 0.00018075007220825002, "loss": 0.1725, "step": 3506 }, { "epoch": 0.22, "learning_rate": 0.00018073782086032108, "loss": 0.1387, "step": 3507 }, { "epoch": 0.22, "learning_rate": 0.00018072556603051265, "loss": 0.1445, "step": 3508 }, { "epoch": 0.22, "learning_rate": 0.00018071330771935322, "loss": 0.1684, "step": 3509 }, { "epoch": 0.23, "learning_rate": 0.00018070104592737146, "loss": 0.1707, "step": 3510 }, { "epoch": 0.23, "learning_rate": 0.00018068878065509615, "loss": 0.1396, "step": 3511 }, { "epoch": 0.23, "learning_rate": 0.00018067651190305627, "loss": 0.1544, "step": 3512 }, { "epoch": 0.23, "learning_rate": 0.00018066423967178088, "loss": 0.1641, "step": 3513 }, { "epoch": 0.23, "learning_rate": 0.00018065196396179923, "loss": 0.15, "step": 3514 }, { "epoch": 0.23, "learning_rate": 0.00018063968477364078, "loss": 0.1642, "step": 3515 }, { "epoch": 0.23, "learning_rate": 0.000180627402107835, "loss": 0.1704, "step": 3516 }, { "epoch": 0.23, "learning_rate": 0.00018061511596491163, "loss": 0.1614, "step": 3517 }, { "epoch": 0.23, "learning_rate": 0.00018060282634540053, "loss": 0.1426, "step": 3518 }, { "epoch": 0.23, "learning_rate": 0.00018059053324983167, "loss": 0.1615, "step": 3519 }, { "epoch": 0.23, "learning_rate": 0.00018057823667873518, "loss": 0.1625, "step": 3520 }, { "epoch": 0.23, "learning_rate": 0.00018056593663264142, "loss": 0.1559, "step": 3521 }, { "epoch": 0.23, "learning_rate": 0.00018055363311208082, "loss": 0.1643, "step": 3522 }, { "epoch": 0.23, "learning_rate": 0.00018054132611758394, "loss": 0.1522, "step": 3523 }, { "epoch": 0.23, "learning_rate": 0.00018052901564968158, "loss": 0.1643, "step": 3524 }, { "epoch": 0.23, "learning_rate": 0.0001805167017089046, "loss": 0.1689, "step": 3525 }, { "epoch": 0.23, "learning_rate": 0.00018050438429578406, "loss": 0.173, "step": 3526 }, { "epoch": 0.23, "learning_rate": 0.00018049206341085117, "loss": 0.159, "step": 3527 }, { "epoch": 0.23, "learning_rate": 0.00018047973905463728, "loss": 0.1445, "step": 3528 }, { "epoch": 0.23, "learning_rate": 0.00018046741122767387, "loss": 0.15, "step": 3529 }, { "epoch": 0.23, "learning_rate": 0.00018045507993049256, "loss": 0.1603, "step": 3530 }, { "epoch": 0.23, "learning_rate": 0.00018044274516362517, "loss": 0.1364, "step": 3531 }, { "epoch": 0.23, "learning_rate": 0.00018043040692760368, "loss": 0.1543, "step": 3532 }, { "epoch": 0.23, "learning_rate": 0.00018041806522296018, "loss": 0.1582, "step": 3533 }, { "epoch": 0.23, "learning_rate": 0.00018040572005022687, "loss": 0.1537, "step": 3534 }, { "epoch": 0.23, "learning_rate": 0.00018039337140993616, "loss": 0.177, "step": 3535 }, { "epoch": 0.23, "learning_rate": 0.00018038101930262064, "loss": 0.1647, "step": 3536 }, { "epoch": 0.23, "learning_rate": 0.00018036866372881294, "loss": 0.1666, "step": 3537 }, { "epoch": 0.23, "learning_rate": 0.00018035630468904592, "loss": 0.168, "step": 3538 }, { "epoch": 0.23, "learning_rate": 0.00018034394218385263, "loss": 0.1592, "step": 3539 }, { "epoch": 0.23, "learning_rate": 0.00018033157621376612, "loss": 0.1629, "step": 3540 }, { "epoch": 0.23, "learning_rate": 0.00018031920677931976, "loss": 0.1914, "step": 3541 }, { "epoch": 0.23, "learning_rate": 0.00018030683388104695, "loss": 0.1618, "step": 3542 }, { "epoch": 0.23, "learning_rate": 0.0001802944575194813, "loss": 0.1584, "step": 3543 }, { "epoch": 0.23, "learning_rate": 0.0001802820776951565, "loss": 0.1372, "step": 3544 }, { "epoch": 0.23, "learning_rate": 0.0001802696944086065, "loss": 0.1603, "step": 3545 }, { "epoch": 0.23, "learning_rate": 0.00018025730766036536, "loss": 0.1597, "step": 3546 }, { "epoch": 0.23, "learning_rate": 0.0001802449174509672, "loss": 0.1801, "step": 3547 }, { "epoch": 0.23, "learning_rate": 0.00018023252378094635, "loss": 0.1831, "step": 3548 }, { "epoch": 0.23, "learning_rate": 0.00018022012665083737, "loss": 0.1432, "step": 3549 }, { "epoch": 0.23, "learning_rate": 0.00018020772606117485, "loss": 0.165, "step": 3550 }, { "epoch": 0.23, "learning_rate": 0.00018019532201249357, "loss": 0.1688, "step": 3551 }, { "epoch": 0.23, "learning_rate": 0.00018018291450532848, "loss": 0.1669, "step": 3552 }, { "epoch": 0.23, "learning_rate": 0.00018017050354021463, "loss": 0.1566, "step": 3553 }, { "epoch": 0.23, "learning_rate": 0.00018015808911768733, "loss": 0.1703, "step": 3554 }, { "epoch": 0.23, "learning_rate": 0.00018014567123828185, "loss": 0.14, "step": 3555 }, { "epoch": 0.23, "learning_rate": 0.00018013324990253385, "loss": 0.1581, "step": 3556 }, { "epoch": 0.23, "learning_rate": 0.0001801208251109789, "loss": 0.1346, "step": 3557 }, { "epoch": 0.23, "learning_rate": 0.00018010839686415287, "loss": 0.1386, "step": 3558 }, { "epoch": 0.23, "learning_rate": 0.00018009596516259174, "loss": 0.1749, "step": 3559 }, { "epoch": 0.23, "learning_rate": 0.00018008353000683163, "loss": 0.1896, "step": 3560 }, { "epoch": 0.23, "learning_rate": 0.0001800710913974088, "loss": 0.1543, "step": 3561 }, { "epoch": 0.23, "learning_rate": 0.00018005864933485973, "loss": 0.1627, "step": 3562 }, { "epoch": 0.23, "learning_rate": 0.00018004620381972095, "loss": 0.1622, "step": 3563 }, { "epoch": 0.23, "learning_rate": 0.00018003375485252919, "loss": 0.1486, "step": 3564 }, { "epoch": 0.23, "learning_rate": 0.0001800213024338213, "loss": 0.1554, "step": 3565 }, { "epoch": 0.23, "learning_rate": 0.00018000884656413436, "loss": 0.1351, "step": 3566 }, { "epoch": 0.23, "learning_rate": 0.00017999638724400545, "loss": 0.1473, "step": 3567 }, { "epoch": 0.23, "learning_rate": 0.00017998392447397197, "loss": 0.1487, "step": 3568 }, { "epoch": 0.23, "learning_rate": 0.00017997145825457132, "loss": 0.1686, "step": 3569 }, { "epoch": 0.23, "learning_rate": 0.00017995898858634118, "loss": 0.1549, "step": 3570 }, { "epoch": 0.23, "learning_rate": 0.00017994651546981926, "loss": 0.1415, "step": 3571 }, { "epoch": 0.23, "learning_rate": 0.0001799340389055435, "loss": 0.1608, "step": 3572 }, { "epoch": 0.23, "learning_rate": 0.00017992155889405196, "loss": 0.1643, "step": 3573 }, { "epoch": 0.23, "learning_rate": 0.00017990907543588284, "loss": 0.1616, "step": 3574 }, { "epoch": 0.23, "learning_rate": 0.0001798965885315745, "loss": 0.1577, "step": 3575 }, { "epoch": 0.23, "learning_rate": 0.00017988409818166546, "loss": 0.1489, "step": 3576 }, { "epoch": 0.23, "learning_rate": 0.00017987160438669438, "loss": 0.1193, "step": 3577 }, { "epoch": 0.23, "learning_rate": 0.0001798591071472, "loss": 0.1561, "step": 3578 }, { "epoch": 0.23, "learning_rate": 0.0001798466064637214, "loss": 0.1358, "step": 3579 }, { "epoch": 0.23, "learning_rate": 0.00017983410233679754, "loss": 0.1471, "step": 3580 }, { "epoch": 0.23, "learning_rate": 0.00017982159476696776, "loss": 0.1767, "step": 3581 }, { "epoch": 0.23, "learning_rate": 0.00017980908375477144, "loss": 0.1565, "step": 3582 }, { "epoch": 0.23, "learning_rate": 0.00017979656930074812, "loss": 0.1444, "step": 3583 }, { "epoch": 0.23, "learning_rate": 0.00017978405140543747, "loss": 0.1705, "step": 3584 }, { "epoch": 0.23, "learning_rate": 0.00017977153006937938, "loss": 0.1381, "step": 3585 }, { "epoch": 0.23, "learning_rate": 0.00017975900529311383, "loss": 0.1702, "step": 3586 }, { "epoch": 0.23, "learning_rate": 0.00017974647707718093, "loss": 0.1322, "step": 3587 }, { "epoch": 0.23, "learning_rate": 0.000179733945422121, "loss": 0.1652, "step": 3588 }, { "epoch": 0.23, "learning_rate": 0.00017972141032847447, "loss": 0.1634, "step": 3589 }, { "epoch": 0.23, "learning_rate": 0.00017970887179678195, "loss": 0.1562, "step": 3590 }, { "epoch": 0.23, "learning_rate": 0.00017969632982758413, "loss": 0.157, "step": 3591 }, { "epoch": 0.23, "learning_rate": 0.0001796837844214219, "loss": 0.1597, "step": 3592 }, { "epoch": 0.23, "learning_rate": 0.00017967123557883629, "loss": 0.1656, "step": 3593 }, { "epoch": 0.23, "learning_rate": 0.00017965868330036848, "loss": 0.1505, "step": 3594 }, { "epoch": 0.23, "learning_rate": 0.00017964612758655983, "loss": 0.1656, "step": 3595 }, { "epoch": 0.23, "learning_rate": 0.00017963356843795177, "loss": 0.1664, "step": 3596 }, { "epoch": 0.23, "learning_rate": 0.00017962100585508595, "loss": 0.1526, "step": 3597 }, { "epoch": 0.23, "learning_rate": 0.00017960843983850415, "loss": 0.1362, "step": 3598 }, { "epoch": 0.23, "learning_rate": 0.00017959587038874822, "loss": 0.1726, "step": 3599 }, { "epoch": 0.23, "learning_rate": 0.00017958329750636037, "loss": 0.1691, "step": 3600 }, { "epoch": 0.23, "learning_rate": 0.00017957072119188262, "loss": 0.1639, "step": 3601 }, { "epoch": 0.23, "learning_rate": 0.00017955814144585749, "loss": 0.1498, "step": 3602 }, { "epoch": 0.23, "learning_rate": 0.0001795455582688274, "loss": 0.1468, "step": 3603 }, { "epoch": 0.23, "learning_rate": 0.00017953297166133512, "loss": 0.1319, "step": 3604 }, { "epoch": 0.23, "learning_rate": 0.00017952038162392331, "loss": 0.1556, "step": 3605 }, { "epoch": 0.23, "learning_rate": 0.00017950778815713502, "loss": 0.1589, "step": 3606 }, { "epoch": 0.23, "learning_rate": 0.00017949519126151336, "loss": 0.1686, "step": 3607 }, { "epoch": 0.23, "learning_rate": 0.00017948259093760155, "loss": 0.1507, "step": 3608 }, { "epoch": 0.23, "learning_rate": 0.00017946998718594293, "loss": 0.1576, "step": 3609 }, { "epoch": 0.23, "learning_rate": 0.00017945738000708117, "loss": 0.1499, "step": 3610 }, { "epoch": 0.23, "learning_rate": 0.00017944476940155986, "loss": 0.1733, "step": 3611 }, { "epoch": 0.23, "learning_rate": 0.00017943215536992289, "loss": 0.1738, "step": 3612 }, { "epoch": 0.23, "learning_rate": 0.00017941953791271423, "loss": 0.1474, "step": 3613 }, { "epoch": 0.23, "learning_rate": 0.00017940691703047804, "loss": 0.1648, "step": 3614 }, { "epoch": 0.23, "learning_rate": 0.00017939429272375863, "loss": 0.1657, "step": 3615 }, { "epoch": 0.23, "learning_rate": 0.00017938166499310035, "loss": 0.1636, "step": 3616 }, { "epoch": 0.23, "learning_rate": 0.00017936903383904785, "loss": 0.1447, "step": 3617 }, { "epoch": 0.23, "learning_rate": 0.00017935639926214581, "loss": 0.152, "step": 3618 }, { "epoch": 0.23, "learning_rate": 0.00017934376126293918, "loss": 0.1426, "step": 3619 }, { "epoch": 0.23, "learning_rate": 0.0001793311198419729, "loss": 0.1913, "step": 3620 }, { "epoch": 0.23, "learning_rate": 0.00017931847499979216, "loss": 0.1858, "step": 3621 }, { "epoch": 0.23, "learning_rate": 0.00017930582673694235, "loss": 0.1363, "step": 3622 }, { "epoch": 0.23, "learning_rate": 0.00017929317505396884, "loss": 0.1693, "step": 3623 }, { "epoch": 0.23, "learning_rate": 0.0001792805199514173, "loss": 0.171, "step": 3624 }, { "epoch": 0.23, "learning_rate": 0.00017926786142983346, "loss": 0.1374, "step": 3625 }, { "epoch": 0.23, "learning_rate": 0.00017925519948976324, "loss": 0.1396, "step": 3626 }, { "epoch": 0.23, "learning_rate": 0.0001792425341317527, "loss": 0.1722, "step": 3627 }, { "epoch": 0.23, "learning_rate": 0.00017922986535634804, "loss": 0.1622, "step": 3628 }, { "epoch": 0.23, "learning_rate": 0.0001792171931640956, "loss": 0.1588, "step": 3629 }, { "epoch": 0.23, "learning_rate": 0.0001792045175555419, "loss": 0.1777, "step": 3630 }, { "epoch": 0.23, "learning_rate": 0.00017919183853123358, "loss": 0.1693, "step": 3631 }, { "epoch": 0.23, "learning_rate": 0.00017917915609171742, "loss": 0.1568, "step": 3632 }, { "epoch": 0.23, "learning_rate": 0.00017916647023754037, "loss": 0.1447, "step": 3633 }, { "epoch": 0.23, "learning_rate": 0.00017915378096924954, "loss": 0.1423, "step": 3634 }, { "epoch": 0.23, "learning_rate": 0.00017914108828739208, "loss": 0.1431, "step": 3635 }, { "epoch": 0.23, "learning_rate": 0.00017912839219251547, "loss": 0.1515, "step": 3636 }, { "epoch": 0.23, "learning_rate": 0.00017911569268516718, "loss": 0.148, "step": 3637 }, { "epoch": 0.23, "learning_rate": 0.00017910298976589492, "loss": 0.1653, "step": 3638 }, { "epoch": 0.23, "learning_rate": 0.00017909028343524649, "loss": 0.1744, "step": 3639 }, { "epoch": 0.23, "learning_rate": 0.00017907757369376985, "loss": 0.1837, "step": 3640 }, { "epoch": 0.23, "learning_rate": 0.00017906486054201316, "loss": 0.1855, "step": 3641 }, { "epoch": 0.23, "learning_rate": 0.00017905214398052465, "loss": 0.1413, "step": 3642 }, { "epoch": 0.23, "learning_rate": 0.00017903942400985273, "loss": 0.1855, "step": 3643 }, { "epoch": 0.23, "learning_rate": 0.00017902670063054597, "loss": 0.1619, "step": 3644 }, { "epoch": 0.23, "learning_rate": 0.0001790139738431531, "loss": 0.1637, "step": 3645 }, { "epoch": 0.23, "learning_rate": 0.00017900124364822294, "loss": 0.1515, "step": 3646 }, { "epoch": 0.23, "learning_rate": 0.00017898851004630447, "loss": 0.1513, "step": 3647 }, { "epoch": 0.23, "learning_rate": 0.0001789757730379469, "loss": 0.1684, "step": 3648 }, { "epoch": 0.23, "learning_rate": 0.00017896303262369945, "loss": 0.157, "step": 3649 }, { "epoch": 0.23, "learning_rate": 0.0001789502888041116, "loss": 0.17, "step": 3650 }, { "epoch": 0.23, "learning_rate": 0.00017893754157973294, "loss": 0.1482, "step": 3651 }, { "epoch": 0.23, "learning_rate": 0.00017892479095111318, "loss": 0.1754, "step": 3652 }, { "epoch": 0.23, "learning_rate": 0.00017891203691880223, "loss": 0.1496, "step": 3653 }, { "epoch": 0.23, "learning_rate": 0.00017889927948335011, "loss": 0.1765, "step": 3654 }, { "epoch": 0.23, "learning_rate": 0.00017888651864530697, "loss": 0.1744, "step": 3655 }, { "epoch": 0.23, "learning_rate": 0.00017887375440522313, "loss": 0.1777, "step": 3656 }, { "epoch": 0.23, "learning_rate": 0.0001788609867636491, "loss": 0.1538, "step": 3657 }, { "epoch": 0.23, "learning_rate": 0.0001788482157211355, "loss": 0.15, "step": 3658 }, { "epoch": 0.23, "learning_rate": 0.00017883544127823302, "loss": 0.1578, "step": 3659 }, { "epoch": 0.23, "learning_rate": 0.00017882266343549258, "loss": 0.151, "step": 3660 }, { "epoch": 0.23, "learning_rate": 0.0001788098821934653, "loss": 0.1481, "step": 3661 }, { "epoch": 0.23, "learning_rate": 0.00017879709755270234, "loss": 0.1556, "step": 3662 }, { "epoch": 0.23, "learning_rate": 0.00017878430951375504, "loss": 0.1542, "step": 3663 }, { "epoch": 0.23, "learning_rate": 0.00017877151807717492, "loss": 0.1385, "step": 3664 }, { "epoch": 0.23, "learning_rate": 0.00017875872324351358, "loss": 0.1586, "step": 3665 }, { "epoch": 0.24, "learning_rate": 0.00017874592501332287, "loss": 0.179, "step": 3666 }, { "epoch": 0.24, "learning_rate": 0.00017873312338715467, "loss": 0.1872, "step": 3667 }, { "epoch": 0.24, "learning_rate": 0.00017872031836556104, "loss": 0.1623, "step": 3668 }, { "epoch": 0.24, "learning_rate": 0.0001787075099490943, "loss": 0.1496, "step": 3669 }, { "epoch": 0.24, "learning_rate": 0.0001786946981383067, "loss": 0.1357, "step": 3670 }, { "epoch": 0.24, "learning_rate": 0.00017868188293375083, "loss": 0.1724, "step": 3671 }, { "epoch": 0.24, "learning_rate": 0.0001786690643359794, "loss": 0.1603, "step": 3672 }, { "epoch": 0.24, "learning_rate": 0.00017865624234554513, "loss": 0.1668, "step": 3673 }, { "epoch": 0.24, "learning_rate": 0.000178643416963001, "loss": 0.1583, "step": 3674 }, { "epoch": 0.24, "learning_rate": 0.00017863058818890016, "loss": 0.1414, "step": 3675 }, { "epoch": 0.24, "learning_rate": 0.00017861775602379581, "loss": 0.1579, "step": 3676 }, { "epoch": 0.24, "learning_rate": 0.0001786049204682414, "loss": 0.149, "step": 3677 }, { "epoch": 0.24, "learning_rate": 0.00017859208152279043, "loss": 0.1541, "step": 3678 }, { "epoch": 0.24, "learning_rate": 0.00017857923918799658, "loss": 0.1546, "step": 3679 }, { "epoch": 0.24, "learning_rate": 0.00017856639346441373, "loss": 0.1822, "step": 3680 }, { "epoch": 0.24, "learning_rate": 0.00017855354435259582, "loss": 0.1884, "step": 3681 }, { "epoch": 0.24, "learning_rate": 0.000178540691853097, "loss": 0.1743, "step": 3682 }, { "epoch": 0.24, "learning_rate": 0.00017852783596647155, "loss": 0.1357, "step": 3683 }, { "epoch": 0.24, "learning_rate": 0.00017851497669327387, "loss": 0.1553, "step": 3684 }, { "epoch": 0.24, "learning_rate": 0.00017850211403405854, "loss": 0.1409, "step": 3685 }, { "epoch": 0.24, "learning_rate": 0.00017848924798938027, "loss": 0.1614, "step": 3686 }, { "epoch": 0.24, "learning_rate": 0.00017847637855979392, "loss": 0.1538, "step": 3687 }, { "epoch": 0.24, "learning_rate": 0.00017846350574585445, "loss": 0.1509, "step": 3688 }, { "epoch": 0.24, "learning_rate": 0.0001784506295481171, "loss": 0.1525, "step": 3689 }, { "epoch": 0.24, "learning_rate": 0.00017843774996713707, "loss": 0.1662, "step": 3690 }, { "epoch": 0.24, "learning_rate": 0.00017842486700346988, "loss": 0.1618, "step": 3691 }, { "epoch": 0.24, "learning_rate": 0.00017841198065767107, "loss": 0.1752, "step": 3692 }, { "epoch": 0.24, "learning_rate": 0.00017839909093029637, "loss": 0.1645, "step": 3693 }, { "epoch": 0.24, "learning_rate": 0.00017838619782190168, "loss": 0.1642, "step": 3694 }, { "epoch": 0.24, "learning_rate": 0.00017837330133304305, "loss": 0.1449, "step": 3695 }, { "epoch": 0.24, "learning_rate": 0.00017836040146427658, "loss": 0.1625, "step": 3696 }, { "epoch": 0.24, "learning_rate": 0.00017834749821615866, "loss": 0.1643, "step": 3697 }, { "epoch": 0.24, "learning_rate": 0.0001783345915892457, "loss": 0.1484, "step": 3698 }, { "epoch": 0.24, "learning_rate": 0.00017832168158409437, "loss": 0.1603, "step": 3699 }, { "epoch": 0.24, "learning_rate": 0.00017830876820126133, "loss": 0.1529, "step": 3700 }, { "epoch": 0.24, "learning_rate": 0.00017829585144130356, "loss": 0.1325, "step": 3701 }, { "epoch": 0.24, "learning_rate": 0.00017828293130477807, "loss": 0.1651, "step": 3702 }, { "epoch": 0.24, "learning_rate": 0.00017827000779224204, "loss": 0.1754, "step": 3703 }, { "epoch": 0.24, "learning_rate": 0.00017825708090425283, "loss": 0.1827, "step": 3704 }, { "epoch": 0.24, "learning_rate": 0.00017824415064136793, "loss": 0.1648, "step": 3705 }, { "epoch": 0.24, "learning_rate": 0.00017823121700414495, "loss": 0.1652, "step": 3706 }, { "epoch": 0.24, "learning_rate": 0.0001782182799931417, "loss": 0.1576, "step": 3707 }, { "epoch": 0.24, "learning_rate": 0.00017820533960891601, "loss": 0.1576, "step": 3708 }, { "epoch": 0.24, "learning_rate": 0.00017819239585202602, "loss": 0.1759, "step": 3709 }, { "epoch": 0.24, "learning_rate": 0.0001781794487230299, "loss": 0.1674, "step": 3710 }, { "epoch": 0.24, "learning_rate": 0.00017816649822248603, "loss": 0.1535, "step": 3711 }, { "epoch": 0.24, "learning_rate": 0.00017815354435095292, "loss": 0.1591, "step": 3712 }, { "epoch": 0.24, "learning_rate": 0.00017814058710898916, "loss": 0.1631, "step": 3713 }, { "epoch": 0.24, "learning_rate": 0.00017812762649715365, "loss": 0.1482, "step": 3714 }, { "epoch": 0.24, "learning_rate": 0.0001781146625160052, "loss": 0.1809, "step": 3715 }, { "epoch": 0.24, "learning_rate": 0.00017810169516610294, "loss": 0.1743, "step": 3716 }, { "epoch": 0.24, "learning_rate": 0.0001780887244480061, "loss": 0.1785, "step": 3717 }, { "epoch": 0.24, "learning_rate": 0.00017807575036227408, "loss": 0.1753, "step": 3718 }, { "epoch": 0.24, "learning_rate": 0.00017806277290946637, "loss": 0.148, "step": 3719 }, { "epoch": 0.24, "learning_rate": 0.00017804979209014264, "loss": 0.1485, "step": 3720 }, { "epoch": 0.24, "learning_rate": 0.00017803680790486265, "loss": 0.1793, "step": 3721 }, { "epoch": 0.24, "learning_rate": 0.00017802382035418646, "loss": 0.1524, "step": 3722 }, { "epoch": 0.24, "learning_rate": 0.00017801082943867406, "loss": 0.1509, "step": 3723 }, { "epoch": 0.24, "learning_rate": 0.00017799783515888576, "loss": 0.1739, "step": 3724 }, { "epoch": 0.24, "learning_rate": 0.00017798483751538187, "loss": 0.1531, "step": 3725 }, { "epoch": 0.24, "learning_rate": 0.00017797183650872302, "loss": 0.1623, "step": 3726 }, { "epoch": 0.24, "learning_rate": 0.00017795883213946983, "loss": 0.1424, "step": 3727 }, { "epoch": 0.24, "learning_rate": 0.00017794582440818318, "loss": 0.1404, "step": 3728 }, { "epoch": 0.24, "learning_rate": 0.00017793281331542394, "loss": 0.152, "step": 3729 }, { "epoch": 0.24, "learning_rate": 0.00017791979886175332, "loss": 0.1678, "step": 3730 }, { "epoch": 0.24, "learning_rate": 0.00017790678104773254, "loss": 0.1599, "step": 3731 }, { "epoch": 0.24, "learning_rate": 0.00017789375987392294, "loss": 0.1527, "step": 3732 }, { "epoch": 0.24, "learning_rate": 0.0001778807353408862, "loss": 0.1528, "step": 3733 }, { "epoch": 0.24, "learning_rate": 0.0001778677074491839, "loss": 0.1623, "step": 3734 }, { "epoch": 0.24, "learning_rate": 0.00017785467619937793, "loss": 0.1627, "step": 3735 }, { "epoch": 0.24, "learning_rate": 0.00017784164159203027, "loss": 0.1575, "step": 3736 }, { "epoch": 0.24, "learning_rate": 0.00017782860362770304, "loss": 0.1525, "step": 3737 }, { "epoch": 0.24, "learning_rate": 0.00017781556230695851, "loss": 0.1393, "step": 3738 }, { "epoch": 0.24, "learning_rate": 0.00017780251763035908, "loss": 0.1691, "step": 3739 }, { "epoch": 0.24, "learning_rate": 0.00017778946959846737, "loss": 0.1604, "step": 3740 }, { "epoch": 0.24, "learning_rate": 0.00017777641821184603, "loss": 0.1472, "step": 3741 }, { "epoch": 0.24, "learning_rate": 0.0001777633634710579, "loss": 0.1423, "step": 3742 }, { "epoch": 0.24, "learning_rate": 0.00017775030537666605, "loss": 0.144, "step": 3743 }, { "epoch": 0.24, "learning_rate": 0.00017773724392923352, "loss": 0.1375, "step": 3744 }, { "epoch": 0.24, "learning_rate": 0.00017772417912932366, "loss": 0.1748, "step": 3745 }, { "epoch": 0.24, "learning_rate": 0.00017771111097749992, "loss": 0.1397, "step": 3746 }, { "epoch": 0.24, "learning_rate": 0.0001776980394743258, "loss": 0.1358, "step": 3747 }, { "epoch": 0.24, "learning_rate": 0.00017768496462036509, "loss": 0.1406, "step": 3748 }, { "epoch": 0.24, "learning_rate": 0.0001776718864161816, "loss": 0.1361, "step": 3749 }, { "epoch": 0.24, "learning_rate": 0.00017765880486233937, "loss": 0.1423, "step": 3750 }, { "epoch": 0.24, "learning_rate": 0.00017764571995940254, "loss": 0.1459, "step": 3751 }, { "epoch": 0.24, "learning_rate": 0.00017763263170793541, "loss": 0.1616, "step": 3752 }, { "epoch": 0.24, "learning_rate": 0.00017761954010850242, "loss": 0.1314, "step": 3753 }, { "epoch": 0.24, "learning_rate": 0.00017760644516166815, "loss": 0.1699, "step": 3754 }, { "epoch": 0.24, "learning_rate": 0.00017759334686799733, "loss": 0.1508, "step": 3755 }, { "epoch": 0.24, "learning_rate": 0.00017758024522805484, "loss": 0.1587, "step": 3756 }, { "epoch": 0.24, "learning_rate": 0.00017756714024240572, "loss": 0.1456, "step": 3757 }, { "epoch": 0.24, "learning_rate": 0.00017755403191161505, "loss": 0.1489, "step": 3758 }, { "epoch": 0.24, "learning_rate": 0.00017754092023624824, "loss": 0.1709, "step": 3759 }, { "epoch": 0.24, "learning_rate": 0.0001775278052168707, "loss": 0.1828, "step": 3760 }, { "epoch": 0.24, "learning_rate": 0.00017751468685404798, "loss": 0.1743, "step": 3761 }, { "epoch": 0.24, "learning_rate": 0.0001775015651483459, "loss": 0.1766, "step": 3762 }, { "epoch": 0.24, "learning_rate": 0.0001774884401003303, "loss": 0.1772, "step": 3763 }, { "epoch": 0.24, "learning_rate": 0.00017747531171056718, "loss": 0.146, "step": 3764 }, { "epoch": 0.24, "learning_rate": 0.00017746217997962273, "loss": 0.1396, "step": 3765 }, { "epoch": 0.24, "learning_rate": 0.00017744904490806334, "loss": 0.1657, "step": 3766 }, { "epoch": 0.24, "learning_rate": 0.00017743590649645533, "loss": 0.182, "step": 3767 }, { "epoch": 0.24, "learning_rate": 0.00017742276474536545, "loss": 0.1679, "step": 3768 }, { "epoch": 0.24, "learning_rate": 0.00017740961965536034, "loss": 0.1567, "step": 3769 }, { "epoch": 0.24, "learning_rate": 0.00017739647122700695, "loss": 0.1692, "step": 3770 }, { "epoch": 0.24, "learning_rate": 0.0001773833194608723, "loss": 0.1727, "step": 3771 }, { "epoch": 0.24, "learning_rate": 0.00017737016435752357, "loss": 0.1495, "step": 3772 }, { "epoch": 0.24, "learning_rate": 0.00017735700591752805, "loss": 0.1743, "step": 3773 }, { "epoch": 0.24, "learning_rate": 0.0001773438441414533, "loss": 0.1374, "step": 3774 }, { "epoch": 0.24, "learning_rate": 0.00017733067902986682, "loss": 0.1566, "step": 3775 }, { "epoch": 0.24, "learning_rate": 0.0001773175105833364, "loss": 0.1766, "step": 3776 }, { "epoch": 0.24, "learning_rate": 0.00017730433880243, "loss": 0.141, "step": 3777 }, { "epoch": 0.24, "learning_rate": 0.0001772911636877156, "loss": 0.1735, "step": 3778 }, { "epoch": 0.24, "learning_rate": 0.0001772779852397614, "loss": 0.1645, "step": 3779 }, { "epoch": 0.24, "learning_rate": 0.00017726480345913573, "loss": 0.1846, "step": 3780 }, { "epoch": 0.24, "learning_rate": 0.0001772516183464071, "loss": 0.1532, "step": 3781 }, { "epoch": 0.24, "learning_rate": 0.00017723842990214408, "loss": 0.1507, "step": 3782 }, { "epoch": 0.24, "learning_rate": 0.00017722523812691544, "loss": 0.1772, "step": 3783 }, { "epoch": 0.24, "learning_rate": 0.0001772120430212901, "loss": 0.1841, "step": 3784 }, { "epoch": 0.24, "learning_rate": 0.0001771988445858371, "loss": 0.2019, "step": 3785 }, { "epoch": 0.24, "learning_rate": 0.00017718564282112565, "loss": 0.1521, "step": 3786 }, { "epoch": 0.24, "learning_rate": 0.00017717243772772504, "loss": 0.162, "step": 3787 }, { "epoch": 0.24, "learning_rate": 0.0001771592293062048, "loss": 0.1928, "step": 3788 }, { "epoch": 0.24, "learning_rate": 0.00017714601755713455, "loss": 0.1695, "step": 3789 }, { "epoch": 0.24, "learning_rate": 0.000177132802481084, "loss": 0.142, "step": 3790 }, { "epoch": 0.24, "learning_rate": 0.00017711958407862316, "loss": 0.1723, "step": 3791 }, { "epoch": 0.24, "learning_rate": 0.000177106362350322, "loss": 0.201, "step": 3792 }, { "epoch": 0.24, "learning_rate": 0.00017709313729675073, "loss": 0.1771, "step": 3793 }, { "epoch": 0.24, "learning_rate": 0.00017707990891847972, "loss": 0.1596, "step": 3794 }, { "epoch": 0.24, "learning_rate": 0.0001770666772160794, "loss": 0.1641, "step": 3795 }, { "epoch": 0.24, "learning_rate": 0.00017705344219012047, "loss": 0.1523, "step": 3796 }, { "epoch": 0.24, "learning_rate": 0.00017704020384117365, "loss": 0.1644, "step": 3797 }, { "epoch": 0.24, "learning_rate": 0.00017702696216980989, "loss": 0.1452, "step": 3798 }, { "epoch": 0.24, "learning_rate": 0.0001770137171766002, "loss": 0.1507, "step": 3799 }, { "epoch": 0.24, "learning_rate": 0.0001770004688621158, "loss": 0.1474, "step": 3800 }, { "epoch": 0.24, "learning_rate": 0.00017698721722692805, "loss": 0.1548, "step": 3801 }, { "epoch": 0.24, "learning_rate": 0.0001769739622716084, "loss": 0.1552, "step": 3802 }, { "epoch": 0.24, "learning_rate": 0.00017696070399672855, "loss": 0.1555, "step": 3803 }, { "epoch": 0.24, "learning_rate": 0.0001769474424028602, "loss": 0.1601, "step": 3804 }, { "epoch": 0.24, "learning_rate": 0.0001769341774905753, "loss": 0.1488, "step": 3805 }, { "epoch": 0.24, "learning_rate": 0.00017692090926044591, "loss": 0.1509, "step": 3806 }, { "epoch": 0.24, "learning_rate": 0.00017690763771304423, "loss": 0.1636, "step": 3807 }, { "epoch": 0.24, "learning_rate": 0.00017689436284894257, "loss": 0.166, "step": 3808 }, { "epoch": 0.24, "learning_rate": 0.0001768810846687135, "loss": 0.1425, "step": 3809 }, { "epoch": 0.24, "learning_rate": 0.00017686780317292957, "loss": 0.1626, "step": 3810 }, { "epoch": 0.24, "learning_rate": 0.00017685451836216357, "loss": 0.168, "step": 3811 }, { "epoch": 0.24, "learning_rate": 0.00017684123023698847, "loss": 0.1484, "step": 3812 }, { "epoch": 0.24, "learning_rate": 0.0001768279387979773, "loss": 0.1735, "step": 3813 }, { "epoch": 0.24, "learning_rate": 0.00017681464404570323, "loss": 0.1546, "step": 3814 }, { "epoch": 0.24, "learning_rate": 0.00017680134598073964, "loss": 0.1507, "step": 3815 }, { "epoch": 0.24, "learning_rate": 0.00017678804460366, "loss": 0.1555, "step": 3816 }, { "epoch": 0.24, "learning_rate": 0.000176774739915038, "loss": 0.1817, "step": 3817 }, { "epoch": 0.24, "learning_rate": 0.0001767614319154473, "loss": 0.1733, "step": 3818 }, { "epoch": 0.24, "learning_rate": 0.00017674812060546196, "loss": 0.168, "step": 3819 }, { "epoch": 0.24, "learning_rate": 0.00017673480598565596, "loss": 0.1819, "step": 3820 }, { "epoch": 0.24, "learning_rate": 0.00017672148805660345, "loss": 0.1912, "step": 3821 }, { "epoch": 0.25, "learning_rate": 0.0001767081668188789, "loss": 0.167, "step": 3822 }, { "epoch": 0.25, "learning_rate": 0.0001766948422730567, "loss": 0.1596, "step": 3823 }, { "epoch": 0.25, "learning_rate": 0.0001766815144197115, "loss": 0.173, "step": 3824 }, { "epoch": 0.25, "learning_rate": 0.00017666818325941814, "loss": 0.1629, "step": 3825 }, { "epoch": 0.25, "learning_rate": 0.00017665484879275144, "loss": 0.1423, "step": 3826 }, { "epoch": 0.25, "learning_rate": 0.00017664151102028653, "loss": 0.1324, "step": 3827 }, { "epoch": 0.25, "learning_rate": 0.00017662816994259858, "loss": 0.1648, "step": 3828 }, { "epoch": 0.25, "learning_rate": 0.00017661482556026296, "loss": 0.1533, "step": 3829 }, { "epoch": 0.25, "learning_rate": 0.00017660147787385512, "loss": 0.1638, "step": 3830 }, { "epoch": 0.25, "learning_rate": 0.0001765881268839507, "loss": 0.1682, "step": 3831 }, { "epoch": 0.25, "learning_rate": 0.0001765747725911255, "loss": 0.17, "step": 3832 }, { "epoch": 0.25, "learning_rate": 0.00017656141499595537, "loss": 0.1645, "step": 3833 }, { "epoch": 0.25, "learning_rate": 0.00017654805409901645, "loss": 0.1662, "step": 3834 }, { "epoch": 0.25, "learning_rate": 0.00017653468990088488, "loss": 0.1587, "step": 3835 }, { "epoch": 0.25, "learning_rate": 0.000176521322402137, "loss": 0.1635, "step": 3836 }, { "epoch": 0.25, "learning_rate": 0.00017650795160334935, "loss": 0.1541, "step": 3837 }, { "epoch": 0.25, "learning_rate": 0.00017649457750509848, "loss": 0.1748, "step": 3838 }, { "epoch": 0.25, "learning_rate": 0.00017648120010796121, "loss": 0.1507, "step": 3839 }, { "epoch": 0.25, "learning_rate": 0.00017646781941251443, "loss": 0.1365, "step": 3840 }, { "epoch": 0.25, "learning_rate": 0.0001764544354193352, "loss": 0.1368, "step": 3841 }, { "epoch": 0.25, "learning_rate": 0.00017644104812900073, "loss": 0.1678, "step": 3842 }, { "epoch": 0.25, "learning_rate": 0.00017642765754208835, "loss": 0.1546, "step": 3843 }, { "epoch": 0.25, "learning_rate": 0.0001764142636591755, "loss": 0.1302, "step": 3844 }, { "epoch": 0.25, "learning_rate": 0.00017640086648083983, "loss": 0.145, "step": 3845 }, { "epoch": 0.25, "learning_rate": 0.00017638746600765913, "loss": 0.1565, "step": 3846 }, { "epoch": 0.25, "learning_rate": 0.00017637406224021125, "loss": 0.1425, "step": 3847 }, { "epoch": 0.25, "learning_rate": 0.00017636065517907427, "loss": 0.1703, "step": 3848 }, { "epoch": 0.25, "learning_rate": 0.0001763472448248264, "loss": 0.1708, "step": 3849 }, { "epoch": 0.25, "learning_rate": 0.00017633383117804594, "loss": 0.1438, "step": 3850 }, { "epoch": 0.25, "learning_rate": 0.00017632041423931136, "loss": 0.1485, "step": 3851 }, { "epoch": 0.25, "learning_rate": 0.0001763069940092013, "loss": 0.1681, "step": 3852 }, { "epoch": 0.25, "learning_rate": 0.0001762935704882945, "loss": 0.164, "step": 3853 }, { "epoch": 0.25, "learning_rate": 0.00017628014367716984, "loss": 0.1611, "step": 3854 }, { "epoch": 0.25, "learning_rate": 0.00017626671357640641, "loss": 0.1612, "step": 3855 }, { "epoch": 0.25, "learning_rate": 0.0001762532801865834, "loss": 0.1478, "step": 3856 }, { "epoch": 0.25, "learning_rate": 0.00017623984350828006, "loss": 0.1502, "step": 3857 }, { "epoch": 0.25, "learning_rate": 0.00017622640354207592, "loss": 0.1529, "step": 3858 }, { "epoch": 0.25, "learning_rate": 0.00017621296028855055, "loss": 0.1723, "step": 3859 }, { "epoch": 0.25, "learning_rate": 0.00017619951374828373, "loss": 0.1678, "step": 3860 }, { "epoch": 0.25, "learning_rate": 0.00017618606392185538, "loss": 0.1371, "step": 3861 }, { "epoch": 0.25, "learning_rate": 0.00017617261080984546, "loss": 0.1576, "step": 3862 }, { "epoch": 0.25, "learning_rate": 0.00017615915441283417, "loss": 0.1693, "step": 3863 }, { "epoch": 0.25, "learning_rate": 0.00017614569473140188, "loss": 0.141, "step": 3864 }, { "epoch": 0.25, "learning_rate": 0.00017613223176612895, "loss": 0.1701, "step": 3865 }, { "epoch": 0.25, "learning_rate": 0.0001761187655175961, "loss": 0.1702, "step": 3866 }, { "epoch": 0.25, "learning_rate": 0.000176105295986384, "loss": 0.16, "step": 3867 }, { "epoch": 0.25, "learning_rate": 0.0001760918231730735, "loss": 0.1482, "step": 3868 }, { "epoch": 0.25, "learning_rate": 0.0001760783470782457, "loss": 0.1683, "step": 3869 }, { "epoch": 0.25, "learning_rate": 0.00017606486770248175, "loss": 0.1551, "step": 3870 }, { "epoch": 0.25, "learning_rate": 0.00017605138504636294, "loss": 0.1676, "step": 3871 }, { "epoch": 0.25, "learning_rate": 0.0001760378991104707, "loss": 0.1539, "step": 3872 }, { "epoch": 0.25, "learning_rate": 0.00017602440989538667, "loss": 0.1755, "step": 3873 }, { "epoch": 0.25, "learning_rate": 0.00017601091740169253, "loss": 0.1631, "step": 3874 }, { "epoch": 0.25, "learning_rate": 0.00017599742162997022, "loss": 0.1536, "step": 3875 }, { "epoch": 0.25, "learning_rate": 0.00017598392258080168, "loss": 0.1365, "step": 3876 }, { "epoch": 0.25, "learning_rate": 0.00017597042025476912, "loss": 0.1635, "step": 3877 }, { "epoch": 0.25, "learning_rate": 0.00017595691465245486, "loss": 0.1546, "step": 3878 }, { "epoch": 0.25, "learning_rate": 0.00017594340577444125, "loss": 0.1561, "step": 3879 }, { "epoch": 0.25, "learning_rate": 0.00017592989362131097, "loss": 0.1613, "step": 3880 }, { "epoch": 0.25, "learning_rate": 0.00017591637819364665, "loss": 0.1492, "step": 3881 }, { "epoch": 0.25, "learning_rate": 0.0001759028594920312, "loss": 0.1548, "step": 3882 }, { "epoch": 0.25, "learning_rate": 0.00017588933751704766, "loss": 0.1514, "step": 3883 }, { "epoch": 0.25, "learning_rate": 0.0001758758122692791, "loss": 0.1705, "step": 3884 }, { "epoch": 0.25, "learning_rate": 0.00017586228374930887, "loss": 0.1633, "step": 3885 }, { "epoch": 0.25, "learning_rate": 0.00017584875195772033, "loss": 0.1497, "step": 3886 }, { "epoch": 0.25, "learning_rate": 0.00017583521689509712, "loss": 0.1528, "step": 3887 }, { "epoch": 0.25, "learning_rate": 0.00017582167856202295, "loss": 0.1608, "step": 3888 }, { "epoch": 0.25, "learning_rate": 0.00017580813695908156, "loss": 0.1645, "step": 3889 }, { "epoch": 0.25, "learning_rate": 0.0001757945920868571, "loss": 0.1344, "step": 3890 }, { "epoch": 0.25, "learning_rate": 0.00017578104394593358, "loss": 0.171, "step": 3891 }, { "epoch": 0.25, "learning_rate": 0.00017576749253689532, "loss": 0.1413, "step": 3892 }, { "epoch": 0.25, "learning_rate": 0.00017575393786032674, "loss": 0.1612, "step": 3893 }, { "epoch": 0.25, "learning_rate": 0.00017574037991681241, "loss": 0.1459, "step": 3894 }, { "epoch": 0.25, "learning_rate": 0.00017572681870693698, "loss": 0.1519, "step": 3895 }, { "epoch": 0.25, "learning_rate": 0.00017571325423128533, "loss": 0.179, "step": 3896 }, { "epoch": 0.25, "learning_rate": 0.00017569968649044243, "loss": 0.1575, "step": 3897 }, { "epoch": 0.25, "learning_rate": 0.00017568611548499335, "loss": 0.1516, "step": 3898 }, { "epoch": 0.25, "learning_rate": 0.00017567254121552344, "loss": 0.1672, "step": 3899 }, { "epoch": 0.25, "learning_rate": 0.00017565896368261803, "loss": 0.1547, "step": 3900 }, { "epoch": 0.25, "learning_rate": 0.00017564538288686267, "loss": 0.1711, "step": 3901 }, { "epoch": 0.25, "learning_rate": 0.00017563179882884307, "loss": 0.1458, "step": 3902 }, { "epoch": 0.25, "learning_rate": 0.00017561821150914502, "loss": 0.158, "step": 3903 }, { "epoch": 0.25, "learning_rate": 0.00017560462092835452, "loss": 0.1842, "step": 3904 }, { "epoch": 0.25, "learning_rate": 0.00017559102708705765, "loss": 0.1561, "step": 3905 }, { "epoch": 0.25, "learning_rate": 0.00017557742998584064, "loss": 0.1581, "step": 3906 }, { "epoch": 0.25, "learning_rate": 0.0001755638296252899, "loss": 0.157, "step": 3907 }, { "epoch": 0.25, "learning_rate": 0.00017555022600599198, "loss": 0.1532, "step": 3908 }, { "epoch": 0.25, "learning_rate": 0.00017553661912853347, "loss": 0.1468, "step": 3909 }, { "epoch": 0.25, "learning_rate": 0.00017552300899350128, "loss": 0.1393, "step": 3910 }, { "epoch": 0.25, "learning_rate": 0.00017550939560148228, "loss": 0.1413, "step": 3911 }, { "epoch": 0.25, "learning_rate": 0.00017549577895306358, "loss": 0.1468, "step": 3912 }, { "epoch": 0.25, "learning_rate": 0.00017548215904883237, "loss": 0.1531, "step": 3913 }, { "epoch": 0.25, "learning_rate": 0.00017546853588937608, "loss": 0.1513, "step": 3914 }, { "epoch": 0.25, "learning_rate": 0.0001754549094752822, "loss": 0.1442, "step": 3915 }, { "epoch": 0.25, "learning_rate": 0.00017544127980713837, "loss": 0.1659, "step": 3916 }, { "epoch": 0.25, "learning_rate": 0.0001754276468855324, "loss": 0.1596, "step": 3917 }, { "epoch": 0.25, "learning_rate": 0.0001754140107110522, "loss": 0.2121, "step": 3918 }, { "epoch": 0.25, "learning_rate": 0.00017540037128428582, "loss": 0.2211, "step": 3919 }, { "epoch": 0.25, "learning_rate": 0.00017538672860582151, "loss": 0.1956, "step": 3920 }, { "epoch": 0.25, "learning_rate": 0.00017537308267624762, "loss": 0.1718, "step": 3921 }, { "epoch": 0.25, "learning_rate": 0.00017535943349615262, "loss": 0.1844, "step": 3922 }, { "epoch": 0.25, "learning_rate": 0.00017534578106612512, "loss": 0.1828, "step": 3923 }, { "epoch": 0.25, "learning_rate": 0.00017533212538675395, "loss": 0.1923, "step": 3924 }, { "epoch": 0.25, "learning_rate": 0.00017531846645862798, "loss": 0.1958, "step": 3925 }, { "epoch": 0.25, "learning_rate": 0.00017530480428233625, "loss": 0.1769, "step": 3926 }, { "epoch": 0.25, "learning_rate": 0.00017529113885846802, "loss": 0.1632, "step": 3927 }, { "epoch": 0.25, "learning_rate": 0.00017527747018761253, "loss": 0.1793, "step": 3928 }, { "epoch": 0.25, "learning_rate": 0.0001752637982703593, "loss": 0.1635, "step": 3929 }, { "epoch": 0.25, "learning_rate": 0.000175250123107298, "loss": 0.1612, "step": 3930 }, { "epoch": 0.25, "learning_rate": 0.00017523644469901828, "loss": 0.149, "step": 3931 }, { "epoch": 0.25, "learning_rate": 0.00017522276304611005, "loss": 0.1646, "step": 3932 }, { "epoch": 0.25, "learning_rate": 0.0001752090781491634, "loss": 0.1606, "step": 3933 }, { "epoch": 0.25, "learning_rate": 0.00017519539000876842, "loss": 0.1585, "step": 3934 }, { "epoch": 0.25, "learning_rate": 0.0001751816986255155, "loss": 0.1716, "step": 3935 }, { "epoch": 0.25, "learning_rate": 0.00017516800399999507, "loss": 0.1798, "step": 3936 }, { "epoch": 0.25, "learning_rate": 0.0001751543061327977, "loss": 0.153, "step": 3937 }, { "epoch": 0.25, "learning_rate": 0.00017514060502451414, "loss": 0.1719, "step": 3938 }, { "epoch": 0.25, "learning_rate": 0.00017512690067573522, "loss": 0.1462, "step": 3939 }, { "epoch": 0.25, "learning_rate": 0.00017511319308705198, "loss": 0.1791, "step": 3940 }, { "epoch": 0.25, "learning_rate": 0.00017509948225905562, "loss": 0.1608, "step": 3941 }, { "epoch": 0.25, "learning_rate": 0.00017508576819233733, "loss": 0.1737, "step": 3942 }, { "epoch": 0.25, "learning_rate": 0.0001750720508874886, "loss": 0.1715, "step": 3943 }, { "epoch": 0.25, "learning_rate": 0.00017505833034510104, "loss": 0.1697, "step": 3944 }, { "epoch": 0.25, "learning_rate": 0.00017504460656576627, "loss": 0.1625, "step": 3945 }, { "epoch": 0.25, "learning_rate": 0.00017503087955007615, "loss": 0.1495, "step": 3946 }, { "epoch": 0.25, "learning_rate": 0.00017501714929862275, "loss": 0.1522, "step": 3947 }, { "epoch": 0.25, "learning_rate": 0.0001750034158119981, "loss": 0.1549, "step": 3948 }, { "epoch": 0.25, "learning_rate": 0.00017498967909079452, "loss": 0.1675, "step": 3949 }, { "epoch": 0.25, "learning_rate": 0.00017497593913560442, "loss": 0.1413, "step": 3950 }, { "epoch": 0.25, "learning_rate": 0.0001749621959470203, "loss": 0.145, "step": 3951 }, { "epoch": 0.25, "learning_rate": 0.00017494844952563492, "loss": 0.1671, "step": 3952 }, { "epoch": 0.25, "learning_rate": 0.00017493469987204103, "loss": 0.1745, "step": 3953 }, { "epoch": 0.25, "learning_rate": 0.00017492094698683163, "loss": 0.1633, "step": 3954 }, { "epoch": 0.25, "learning_rate": 0.00017490719087059984, "loss": 0.1625, "step": 3955 }, { "epoch": 0.25, "learning_rate": 0.00017489343152393886, "loss": 0.1567, "step": 3956 }, { "epoch": 0.25, "learning_rate": 0.0001748796689474421, "loss": 0.1518, "step": 3957 }, { "epoch": 0.25, "learning_rate": 0.0001748659031417031, "loss": 0.1629, "step": 3958 }, { "epoch": 0.25, "learning_rate": 0.00017485213410731545, "loss": 0.1745, "step": 3959 }, { "epoch": 0.25, "learning_rate": 0.00017483836184487304, "loss": 0.1616, "step": 3960 }, { "epoch": 0.25, "learning_rate": 0.00017482458635496975, "loss": 0.1726, "step": 3961 }, { "epoch": 0.25, "learning_rate": 0.00017481080763819966, "loss": 0.1468, "step": 3962 }, { "epoch": 0.25, "learning_rate": 0.00017479702569515705, "loss": 0.1468, "step": 3963 }, { "epoch": 0.25, "learning_rate": 0.00017478324052643617, "loss": 0.1507, "step": 3964 }, { "epoch": 0.25, "learning_rate": 0.0001747694521326316, "loss": 0.1717, "step": 3965 }, { "epoch": 0.25, "learning_rate": 0.00017475566051433793, "loss": 0.1628, "step": 3966 }, { "epoch": 0.25, "learning_rate": 0.00017474186567215, "loss": 0.1757, "step": 3967 }, { "epoch": 0.25, "learning_rate": 0.00017472806760666262, "loss": 0.1523, "step": 3968 }, { "epoch": 0.25, "learning_rate": 0.00017471426631847096, "loss": 0.1845, "step": 3969 }, { "epoch": 0.25, "learning_rate": 0.0001747004618081701, "loss": 0.1864, "step": 3970 }, { "epoch": 0.25, "learning_rate": 0.00017468665407635543, "loss": 0.1408, "step": 3971 }, { "epoch": 0.25, "learning_rate": 0.00017467284312362242, "loss": 0.1595, "step": 3972 }, { "epoch": 0.25, "learning_rate": 0.00017465902895056668, "loss": 0.1557, "step": 3973 }, { "epoch": 0.25, "learning_rate": 0.0001746452115577839, "loss": 0.1602, "step": 3974 }, { "epoch": 0.25, "learning_rate": 0.00017463139094587004, "loss": 0.1724, "step": 3975 }, { "epoch": 0.25, "learning_rate": 0.0001746175671154211, "loss": 0.1759, "step": 3976 }, { "epoch": 0.26, "learning_rate": 0.00017460374006703322, "loss": 0.1618, "step": 3977 }, { "epoch": 0.26, "learning_rate": 0.00017458990980130271, "loss": 0.1537, "step": 3978 }, { "epoch": 0.26, "learning_rate": 0.00017457607631882603, "loss": 0.1858, "step": 3979 }, { "epoch": 0.26, "learning_rate": 0.00017456223962019975, "loss": 0.1704, "step": 3980 }, { "epoch": 0.26, "learning_rate": 0.00017454839970602057, "loss": 0.1402, "step": 3981 }, { "epoch": 0.26, "learning_rate": 0.00017453455657688536, "loss": 0.166, "step": 3982 }, { "epoch": 0.26, "learning_rate": 0.00017452071023339114, "loss": 0.1584, "step": 3983 }, { "epoch": 0.26, "learning_rate": 0.00017450686067613502, "loss": 0.1327, "step": 3984 }, { "epoch": 0.26, "learning_rate": 0.00017449300790571426, "loss": 0.1581, "step": 3985 }, { "epoch": 0.26, "learning_rate": 0.00017447915192272628, "loss": 0.1491, "step": 3986 }, { "epoch": 0.26, "learning_rate": 0.00017446529272776865, "loss": 0.182, "step": 3987 }, { "epoch": 0.26, "learning_rate": 0.00017445143032143905, "loss": 0.1732, "step": 3988 }, { "epoch": 0.26, "learning_rate": 0.0001744375647043353, "loss": 0.1551, "step": 3989 }, { "epoch": 0.26, "learning_rate": 0.00017442369587705535, "loss": 0.1539, "step": 3990 }, { "epoch": 0.26, "learning_rate": 0.00017440982384019733, "loss": 0.1545, "step": 3991 }, { "epoch": 0.26, "learning_rate": 0.00017439594859435946, "loss": 0.1612, "step": 3992 }, { "epoch": 0.26, "learning_rate": 0.00017438207014014014, "loss": 0.1947, "step": 3993 }, { "epoch": 0.26, "learning_rate": 0.0001743681884781379, "loss": 0.1482, "step": 3994 }, { "epoch": 0.26, "learning_rate": 0.00017435430360895137, "loss": 0.1655, "step": 3995 }, { "epoch": 0.26, "learning_rate": 0.00017434041553317933, "loss": 0.1734, "step": 3996 }, { "epoch": 0.26, "learning_rate": 0.00017432652425142074, "loss": 0.1823, "step": 3997 }, { "epoch": 0.26, "learning_rate": 0.00017431262976427467, "loss": 0.1511, "step": 3998 }, { "epoch": 0.26, "learning_rate": 0.00017429873207234033, "loss": 0.1602, "step": 3999 }, { "epoch": 0.26, "learning_rate": 0.00017428483117621707, "loss": 0.1658, "step": 4000 }, { "epoch": 0.26, "learning_rate": 0.00017427092707650437, "loss": 0.1637, "step": 4001 }, { "epoch": 0.26, "learning_rate": 0.00017425701977380185, "loss": 0.1482, "step": 4002 }, { "epoch": 0.26, "learning_rate": 0.00017424310926870932, "loss": 0.1343, "step": 4003 }, { "epoch": 0.26, "learning_rate": 0.0001742291955618266, "loss": 0.1542, "step": 4004 }, { "epoch": 0.26, "learning_rate": 0.00017421527865375382, "loss": 0.1667, "step": 4005 }, { "epoch": 0.26, "learning_rate": 0.0001742013585450911, "loss": 0.1393, "step": 4006 }, { "epoch": 0.26, "learning_rate": 0.00017418743523643873, "loss": 0.1806, "step": 4007 }, { "epoch": 0.26, "learning_rate": 0.0001741735087283972, "loss": 0.1798, "step": 4008 }, { "epoch": 0.26, "learning_rate": 0.00017415957902156714, "loss": 0.1425, "step": 4009 }, { "epoch": 0.26, "learning_rate": 0.0001741456461165492, "loss": 0.168, "step": 4010 }, { "epoch": 0.26, "learning_rate": 0.00017413171001394432, "loss": 0.167, "step": 4011 }, { "epoch": 0.26, "learning_rate": 0.00017411777071435345, "loss": 0.1676, "step": 4012 }, { "epoch": 0.26, "learning_rate": 0.00017410382821837775, "loss": 0.1454, "step": 4013 }, { "epoch": 0.26, "learning_rate": 0.00017408988252661848, "loss": 0.142, "step": 4014 }, { "epoch": 0.26, "learning_rate": 0.00017407593363967712, "loss": 0.166, "step": 4015 }, { "epoch": 0.26, "learning_rate": 0.0001740619815581552, "loss": 0.1804, "step": 4016 }, { "epoch": 0.26, "learning_rate": 0.00017404802628265435, "loss": 0.1563, "step": 4017 }, { "epoch": 0.26, "learning_rate": 0.00017403406781377648, "loss": 0.155, "step": 4018 }, { "epoch": 0.26, "learning_rate": 0.0001740201061521235, "loss": 0.1759, "step": 4019 }, { "epoch": 0.26, "learning_rate": 0.00017400614129829762, "loss": 0.1352, "step": 4020 }, { "epoch": 0.26, "learning_rate": 0.00017399217325290097, "loss": 0.1577, "step": 4021 }, { "epoch": 0.26, "learning_rate": 0.00017397820201653602, "loss": 0.168, "step": 4022 }, { "epoch": 0.26, "learning_rate": 0.0001739642275898052, "loss": 0.1401, "step": 4023 }, { "epoch": 0.26, "learning_rate": 0.00017395024997331123, "loss": 0.1694, "step": 4024 }, { "epoch": 0.26, "learning_rate": 0.00017393626916765688, "loss": 0.1527, "step": 4025 }, { "epoch": 0.26, "learning_rate": 0.00017392228517344513, "loss": 0.1323, "step": 4026 }, { "epoch": 0.26, "learning_rate": 0.000173908297991279, "loss": 0.1597, "step": 4027 }, { "epoch": 0.26, "learning_rate": 0.00017389430762176168, "loss": 0.1615, "step": 4028 }, { "epoch": 0.26, "learning_rate": 0.00017388031406549657, "loss": 0.1407, "step": 4029 }, { "epoch": 0.26, "learning_rate": 0.00017386631732308716, "loss": 0.1557, "step": 4030 }, { "epoch": 0.26, "learning_rate": 0.000173852317395137, "loss": 0.1691, "step": 4031 }, { "epoch": 0.26, "learning_rate": 0.0001738383142822499, "loss": 0.1316, "step": 4032 }, { "epoch": 0.26, "learning_rate": 0.00017382430798502975, "loss": 0.1515, "step": 4033 }, { "epoch": 0.26, "learning_rate": 0.00017381029850408058, "loss": 0.1529, "step": 4034 }, { "epoch": 0.26, "learning_rate": 0.00017379628584000655, "loss": 0.1839, "step": 4035 }, { "epoch": 0.26, "learning_rate": 0.00017378226999341197, "loss": 0.1853, "step": 4036 }, { "epoch": 0.26, "learning_rate": 0.00017376825096490127, "loss": 0.1548, "step": 4037 }, { "epoch": 0.26, "learning_rate": 0.00017375422875507904, "loss": 0.1565, "step": 4038 }, { "epoch": 0.26, "learning_rate": 0.00017374020336455003, "loss": 0.1711, "step": 4039 }, { "epoch": 0.26, "learning_rate": 0.00017372617479391903, "loss": 0.173, "step": 4040 }, { "epoch": 0.26, "learning_rate": 0.00017371214304379113, "loss": 0.1433, "step": 4041 }, { "epoch": 0.26, "learning_rate": 0.00017369810811477136, "loss": 0.1581, "step": 4042 }, { "epoch": 0.26, "learning_rate": 0.00017368407000746503, "loss": 0.1673, "step": 4043 }, { "epoch": 0.26, "learning_rate": 0.00017367002872247758, "loss": 0.1548, "step": 4044 }, { "epoch": 0.26, "learning_rate": 0.00017365598426041448, "loss": 0.1599, "step": 4045 }, { "epoch": 0.26, "learning_rate": 0.00017364193662188147, "loss": 0.1535, "step": 4046 }, { "epoch": 0.26, "learning_rate": 0.0001736278858074843, "loss": 0.1664, "step": 4047 }, { "epoch": 0.26, "learning_rate": 0.00017361383181782898, "loss": 0.1701, "step": 4048 }, { "epoch": 0.26, "learning_rate": 0.00017359977465352156, "loss": 0.1582, "step": 4049 }, { "epoch": 0.26, "learning_rate": 0.00017358571431516832, "loss": 0.1445, "step": 4050 }, { "epoch": 0.26, "learning_rate": 0.00017357165080337557, "loss": 0.1633, "step": 4051 }, { "epoch": 0.26, "learning_rate": 0.00017355758411874983, "loss": 0.1741, "step": 4052 }, { "epoch": 0.26, "learning_rate": 0.00017354351426189773, "loss": 0.1404, "step": 4053 }, { "epoch": 0.26, "learning_rate": 0.00017352944123342603, "loss": 0.1375, "step": 4054 }, { "epoch": 0.26, "learning_rate": 0.00017351536503394167, "loss": 0.156, "step": 4055 }, { "epoch": 0.26, "learning_rate": 0.00017350128566405172, "loss": 0.1583, "step": 4056 }, { "epoch": 0.26, "learning_rate": 0.00017348720312436325, "loss": 0.1613, "step": 4057 }, { "epoch": 0.26, "learning_rate": 0.0001734731174154837, "loss": 0.167, "step": 4058 }, { "epoch": 0.26, "learning_rate": 0.00017345902853802046, "loss": 0.1626, "step": 4059 }, { "epoch": 0.26, "learning_rate": 0.00017344493649258118, "loss": 0.1538, "step": 4060 }, { "epoch": 0.26, "learning_rate": 0.00017343084127977357, "loss": 0.161, "step": 4061 }, { "epoch": 0.26, "learning_rate": 0.00017341674290020543, "loss": 0.1539, "step": 4062 }, { "epoch": 0.26, "learning_rate": 0.00017340264135448487, "loss": 0.151, "step": 4063 }, { "epoch": 0.26, "learning_rate": 0.00017338853664321992, "loss": 0.1537, "step": 4064 }, { "epoch": 0.26, "learning_rate": 0.00017337442876701892, "loss": 0.1608, "step": 4065 }, { "epoch": 0.26, "learning_rate": 0.00017336031772649033, "loss": 0.1645, "step": 4066 }, { "epoch": 0.26, "learning_rate": 0.0001733462035222426, "loss": 0.1461, "step": 4067 }, { "epoch": 0.26, "learning_rate": 0.00017333208615488447, "loss": 0.1479, "step": 4068 }, { "epoch": 0.26, "learning_rate": 0.00017331796562502475, "loss": 0.1609, "step": 4069 }, { "epoch": 0.26, "learning_rate": 0.0001733038419332724, "loss": 0.1579, "step": 4070 }, { "epoch": 0.26, "learning_rate": 0.00017328971508023653, "loss": 0.1844, "step": 4071 }, { "epoch": 0.26, "learning_rate": 0.00017327558506652636, "loss": 0.1441, "step": 4072 }, { "epoch": 0.26, "learning_rate": 0.00017326145189275124, "loss": 0.1678, "step": 4073 }, { "epoch": 0.26, "learning_rate": 0.00017324731555952067, "loss": 0.1489, "step": 4074 }, { "epoch": 0.26, "learning_rate": 0.00017323317606744436, "loss": 0.1478, "step": 4075 }, { "epoch": 0.26, "learning_rate": 0.00017321903341713203, "loss": 0.133, "step": 4076 }, { "epoch": 0.26, "learning_rate": 0.00017320488760919355, "loss": 0.1699, "step": 4077 }, { "epoch": 0.26, "learning_rate": 0.00017319073864423902, "loss": 0.1746, "step": 4078 }, { "epoch": 0.26, "learning_rate": 0.00017317658652287867, "loss": 0.1489, "step": 4079 }, { "epoch": 0.26, "learning_rate": 0.00017316243124572273, "loss": 0.1684, "step": 4080 }, { "epoch": 0.26, "learning_rate": 0.0001731482728133817, "loss": 0.1918, "step": 4081 }, { "epoch": 0.26, "learning_rate": 0.0001731341112264662, "loss": 0.1619, "step": 4082 }, { "epoch": 0.26, "learning_rate": 0.0001731199464855869, "loss": 0.1669, "step": 4083 }, { "epoch": 0.26, "learning_rate": 0.0001731057785913547, "loss": 0.1308, "step": 4084 }, { "epoch": 0.26, "learning_rate": 0.00017309160754438064, "loss": 0.1551, "step": 4085 }, { "epoch": 0.26, "learning_rate": 0.00017307743334527578, "loss": 0.1476, "step": 4086 }, { "epoch": 0.26, "learning_rate": 0.00017306325599465141, "loss": 0.1454, "step": 4087 }, { "epoch": 0.26, "learning_rate": 0.000173049075493119, "loss": 0.1566, "step": 4088 }, { "epoch": 0.26, "learning_rate": 0.00017303489184129002, "loss": 0.1637, "step": 4089 }, { "epoch": 0.26, "learning_rate": 0.00017302070503977617, "loss": 0.1596, "step": 4090 }, { "epoch": 0.26, "learning_rate": 0.0001730065150891893, "loss": 0.1398, "step": 4091 }, { "epoch": 0.26, "learning_rate": 0.00017299232199014135, "loss": 0.1678, "step": 4092 }, { "epoch": 0.26, "learning_rate": 0.00017297812574324439, "loss": 0.1616, "step": 4093 }, { "epoch": 0.26, "learning_rate": 0.00017296392634911064, "loss": 0.1581, "step": 4094 }, { "epoch": 0.26, "learning_rate": 0.0001729497238083525, "loss": 0.1526, "step": 4095 }, { "epoch": 0.26, "learning_rate": 0.0001729355181215824, "loss": 0.1548, "step": 4096 }, { "epoch": 0.26, "learning_rate": 0.00017292130928941304, "loss": 0.1225, "step": 4097 }, { "epoch": 0.26, "learning_rate": 0.00017290709731245714, "loss": 0.1415, "step": 4098 }, { "epoch": 0.26, "learning_rate": 0.00017289288219132762, "loss": 0.1361, "step": 4099 }, { "epoch": 0.26, "learning_rate": 0.00017287866392663752, "loss": 0.167, "step": 4100 }, { "epoch": 0.26, "learning_rate": 0.000172864442519, "loss": 0.1692, "step": 4101 }, { "epoch": 0.26, "learning_rate": 0.00017285021796902838, "loss": 0.1511, "step": 4102 }, { "epoch": 0.26, "learning_rate": 0.00017283599027733613, "loss": 0.1876, "step": 4103 }, { "epoch": 0.26, "learning_rate": 0.00017282175944453675, "loss": 0.1518, "step": 4104 }, { "epoch": 0.26, "learning_rate": 0.00017280752547124405, "loss": 0.1552, "step": 4105 }, { "epoch": 0.26, "learning_rate": 0.00017279328835807182, "loss": 0.179, "step": 4106 }, { "epoch": 0.26, "learning_rate": 0.00017277904810563408, "loss": 0.1674, "step": 4107 }, { "epoch": 0.26, "learning_rate": 0.00017276480471454492, "loss": 0.1287, "step": 4108 }, { "epoch": 0.26, "learning_rate": 0.00017275055818541862, "loss": 0.1529, "step": 4109 }, { "epoch": 0.26, "learning_rate": 0.00017273630851886954, "loss": 0.1743, "step": 4110 }, { "epoch": 0.26, "learning_rate": 0.00017272205571551225, "loss": 0.1596, "step": 4111 }, { "epoch": 0.26, "learning_rate": 0.0001727077997759614, "loss": 0.1494, "step": 4112 }, { "epoch": 0.26, "learning_rate": 0.00017269354070083179, "loss": 0.1527, "step": 4113 }, { "epoch": 0.26, "learning_rate": 0.00017267927849073836, "loss": 0.1626, "step": 4114 }, { "epoch": 0.26, "learning_rate": 0.00017266501314629616, "loss": 0.1696, "step": 4115 }, { "epoch": 0.26, "learning_rate": 0.0001726507446681204, "loss": 0.1577, "step": 4116 }, { "epoch": 0.26, "learning_rate": 0.0001726364730568264, "loss": 0.1541, "step": 4117 }, { "epoch": 0.26, "learning_rate": 0.00017262219831302967, "loss": 0.1503, "step": 4118 }, { "epoch": 0.26, "learning_rate": 0.00017260792043734578, "loss": 0.1425, "step": 4119 }, { "epoch": 0.26, "learning_rate": 0.00017259363943039052, "loss": 0.1376, "step": 4120 }, { "epoch": 0.26, "learning_rate": 0.00017257935529277972, "loss": 0.1847, "step": 4121 }, { "epoch": 0.26, "learning_rate": 0.00017256506802512943, "loss": 0.1417, "step": 4122 }, { "epoch": 0.26, "learning_rate": 0.0001725507776280558, "loss": 0.1532, "step": 4123 }, { "epoch": 0.26, "learning_rate": 0.00017253648410217508, "loss": 0.1448, "step": 4124 }, { "epoch": 0.26, "learning_rate": 0.00017252218744810374, "loss": 0.1534, "step": 4125 }, { "epoch": 0.26, "learning_rate": 0.0001725078876664583, "loss": 0.1571, "step": 4126 }, { "epoch": 0.26, "learning_rate": 0.00017249358475785544, "loss": 0.1746, "step": 4127 }, { "epoch": 0.26, "learning_rate": 0.000172479278722912, "loss": 0.1675, "step": 4128 }, { "epoch": 0.26, "learning_rate": 0.00017246496956224494, "loss": 0.1607, "step": 4129 }, { "epoch": 0.26, "learning_rate": 0.00017245065727647139, "loss": 0.1499, "step": 4130 }, { "epoch": 0.26, "learning_rate": 0.0001724363418662085, "loss": 0.1542, "step": 4131 }, { "epoch": 0.26, "learning_rate": 0.00017242202333207364, "loss": 0.1872, "step": 4132 }, { "epoch": 0.27, "learning_rate": 0.00017240770167468438, "loss": 0.1569, "step": 4133 }, { "epoch": 0.27, "learning_rate": 0.00017239337689465832, "loss": 0.1617, "step": 4134 }, { "epoch": 0.27, "learning_rate": 0.0001723790489926132, "loss": 0.1493, "step": 4135 }, { "epoch": 0.27, "learning_rate": 0.00017236471796916694, "loss": 0.1268, "step": 4136 }, { "epoch": 0.27, "learning_rate": 0.0001723503838249376, "loss": 0.1521, "step": 4137 }, { "epoch": 0.27, "learning_rate": 0.0001723360465605433, "loss": 0.1738, "step": 4138 }, { "epoch": 0.27, "learning_rate": 0.00017232170617660238, "loss": 0.1717, "step": 4139 }, { "epoch": 0.27, "learning_rate": 0.00017230736267373326, "loss": 0.1507, "step": 4140 }, { "epoch": 0.27, "learning_rate": 0.00017229301605255453, "loss": 0.175, "step": 4141 }, { "epoch": 0.27, "learning_rate": 0.0001722786663136849, "loss": 0.1633, "step": 4142 }, { "epoch": 0.27, "learning_rate": 0.00017226431345774323, "loss": 0.149, "step": 4143 }, { "epoch": 0.27, "learning_rate": 0.00017224995748534844, "loss": 0.1432, "step": 4144 }, { "epoch": 0.27, "learning_rate": 0.00017223559839711974, "loss": 0.1603, "step": 4145 }, { "epoch": 0.27, "learning_rate": 0.00017222123619367626, "loss": 0.1551, "step": 4146 }, { "epoch": 0.27, "learning_rate": 0.00017220687087563742, "loss": 0.1678, "step": 4147 }, { "epoch": 0.27, "learning_rate": 0.0001721925024436228, "loss": 0.1587, "step": 4148 }, { "epoch": 0.27, "learning_rate": 0.00017217813089825197, "loss": 0.1341, "step": 4149 }, { "epoch": 0.27, "learning_rate": 0.00017216375624014478, "loss": 0.1572, "step": 4150 }, { "epoch": 0.27, "learning_rate": 0.0001721493784699211, "loss": 0.1681, "step": 4151 }, { "epoch": 0.27, "learning_rate": 0.00017213499758820096, "loss": 0.1568, "step": 4152 }, { "epoch": 0.27, "learning_rate": 0.0001721206135956046, "loss": 0.1495, "step": 4153 }, { "epoch": 0.27, "learning_rate": 0.00017210622649275234, "loss": 0.1625, "step": 4154 }, { "epoch": 0.27, "learning_rate": 0.0001720918362802646, "loss": 0.1557, "step": 4155 }, { "epoch": 0.27, "learning_rate": 0.00017207744295876198, "loss": 0.163, "step": 4156 }, { "epoch": 0.27, "learning_rate": 0.00017206304652886523, "loss": 0.1608, "step": 4157 }, { "epoch": 0.27, "learning_rate": 0.00017204864699119516, "loss": 0.15, "step": 4158 }, { "epoch": 0.27, "learning_rate": 0.0001720342443463728, "loss": 0.1615, "step": 4159 }, { "epoch": 0.27, "learning_rate": 0.00017201983859501928, "loss": 0.1615, "step": 4160 }, { "epoch": 0.27, "learning_rate": 0.0001720054297377558, "loss": 0.1437, "step": 4161 }, { "epoch": 0.27, "learning_rate": 0.0001719910177752038, "loss": 0.161, "step": 4162 }, { "epoch": 0.27, "learning_rate": 0.00017197660270798482, "loss": 0.1784, "step": 4163 }, { "epoch": 0.27, "learning_rate": 0.0001719621845367205, "loss": 0.1612, "step": 4164 }, { "epoch": 0.27, "learning_rate": 0.00017194776326203258, "loss": 0.1704, "step": 4165 }, { "epoch": 0.27, "learning_rate": 0.0001719333388845431, "loss": 0.1455, "step": 4166 }, { "epoch": 0.27, "learning_rate": 0.00017191891140487403, "loss": 0.1562, "step": 4167 }, { "epoch": 0.27, "learning_rate": 0.00017190448082364765, "loss": 0.1445, "step": 4168 }, { "epoch": 0.27, "learning_rate": 0.0001718900471414862, "loss": 0.1619, "step": 4169 }, { "epoch": 0.27, "learning_rate": 0.0001718756103590122, "loss": 0.1435, "step": 4170 }, { "epoch": 0.27, "learning_rate": 0.00017186117047684825, "loss": 0.1417, "step": 4171 }, { "epoch": 0.27, "learning_rate": 0.00017184672749561705, "loss": 0.1798, "step": 4172 }, { "epoch": 0.27, "learning_rate": 0.00017183228141594143, "loss": 0.1558, "step": 4173 }, { "epoch": 0.27, "learning_rate": 0.00017181783223844447, "loss": 0.155, "step": 4174 }, { "epoch": 0.27, "learning_rate": 0.0001718033799637493, "loss": 0.1948, "step": 4175 }, { "epoch": 0.27, "learning_rate": 0.00017178892459247908, "loss": 0.1651, "step": 4176 }, { "epoch": 0.27, "learning_rate": 0.00017177446612525734, "loss": 0.1638, "step": 4177 }, { "epoch": 0.27, "learning_rate": 0.00017176000456270752, "loss": 0.1343, "step": 4178 }, { "epoch": 0.27, "learning_rate": 0.00017174553990545333, "loss": 0.135, "step": 4179 }, { "epoch": 0.27, "learning_rate": 0.00017173107215411854, "loss": 0.1737, "step": 4180 }, { "epoch": 0.27, "learning_rate": 0.00017171660130932713, "loss": 0.1564, "step": 4181 }, { "epoch": 0.27, "learning_rate": 0.00017170212737170311, "loss": 0.1652, "step": 4182 }, { "epoch": 0.27, "learning_rate": 0.00017168765034187073, "loss": 0.1359, "step": 4183 }, { "epoch": 0.27, "learning_rate": 0.00017167317022045428, "loss": 0.1737, "step": 4184 }, { "epoch": 0.27, "learning_rate": 0.00017165868700807826, "loss": 0.1705, "step": 4185 }, { "epoch": 0.27, "learning_rate": 0.00017164420070536727, "loss": 0.1504, "step": 4186 }, { "epoch": 0.27, "learning_rate": 0.000171629711312946, "loss": 0.1608, "step": 4187 }, { "epoch": 0.27, "learning_rate": 0.00017161521883143934, "loss": 0.1724, "step": 4188 }, { "epoch": 0.27, "learning_rate": 0.0001716007232614723, "loss": 0.1478, "step": 4189 }, { "epoch": 0.27, "learning_rate": 0.00017158622460367, "loss": 0.1432, "step": 4190 }, { "epoch": 0.27, "learning_rate": 0.00017157172285865773, "loss": 0.1576, "step": 4191 }, { "epoch": 0.27, "learning_rate": 0.00017155721802706087, "loss": 0.1355, "step": 4192 }, { "epoch": 0.27, "learning_rate": 0.00017154271010950492, "loss": 0.1654, "step": 4193 }, { "epoch": 0.27, "learning_rate": 0.00017152819910661558, "loss": 0.1548, "step": 4194 }, { "epoch": 0.27, "learning_rate": 0.00017151368501901865, "loss": 0.1959, "step": 4195 }, { "epoch": 0.27, "learning_rate": 0.0001714991678473401, "loss": 0.1434, "step": 4196 }, { "epoch": 0.27, "learning_rate": 0.0001714846475922059, "loss": 0.1575, "step": 4197 }, { "epoch": 0.27, "learning_rate": 0.0001714701242542423, "loss": 0.1558, "step": 4198 }, { "epoch": 0.27, "learning_rate": 0.00017145559783407564, "loss": 0.1637, "step": 4199 }, { "epoch": 0.27, "learning_rate": 0.00017144106833233232, "loss": 0.14, "step": 4200 }, { "epoch": 0.27, "learning_rate": 0.00017142653574963904, "loss": 0.1448, "step": 4201 }, { "epoch": 0.27, "learning_rate": 0.00017141200008662249, "loss": 0.1535, "step": 4202 }, { "epoch": 0.27, "learning_rate": 0.00017139746134390947, "loss": 0.1513, "step": 4203 }, { "epoch": 0.27, "learning_rate": 0.000171382919522127, "loss": 0.1446, "step": 4204 }, { "epoch": 0.27, "learning_rate": 0.00017136837462190227, "loss": 0.1628, "step": 4205 }, { "epoch": 0.27, "learning_rate": 0.00017135382664386248, "loss": 0.1744, "step": 4206 }, { "epoch": 0.27, "learning_rate": 0.00017133927558863504, "loss": 0.1424, "step": 4207 }, { "epoch": 0.27, "learning_rate": 0.00017132472145684746, "loss": 0.142, "step": 4208 }, { "epoch": 0.27, "learning_rate": 0.0001713101642491274, "loss": 0.1318, "step": 4209 }, { "epoch": 0.27, "learning_rate": 0.0001712956039661027, "loss": 0.148, "step": 4210 }, { "epoch": 0.27, "learning_rate": 0.00017128104060840123, "loss": 0.136, "step": 4211 }, { "epoch": 0.27, "learning_rate": 0.00017126647417665107, "loss": 0.1523, "step": 4212 }, { "epoch": 0.27, "learning_rate": 0.0001712519046714804, "loss": 0.1698, "step": 4213 }, { "epoch": 0.27, "learning_rate": 0.00017123733209351754, "loss": 0.1608, "step": 4214 }, { "epoch": 0.27, "learning_rate": 0.00017122275644339093, "loss": 0.1606, "step": 4215 }, { "epoch": 0.27, "learning_rate": 0.00017120817772172916, "loss": 0.1647, "step": 4216 }, { "epoch": 0.27, "learning_rate": 0.00017119359592916098, "loss": 0.1686, "step": 4217 }, { "epoch": 0.27, "learning_rate": 0.0001711790110663152, "loss": 0.1516, "step": 4218 }, { "epoch": 0.27, "learning_rate": 0.0001711644231338208, "loss": 0.1517, "step": 4219 }, { "epoch": 0.27, "learning_rate": 0.000171149832132307, "loss": 0.1725, "step": 4220 }, { "epoch": 0.27, "learning_rate": 0.00017113523806240287, "loss": 0.137, "step": 4221 }, { "epoch": 0.27, "learning_rate": 0.00017112064092473795, "loss": 0.157, "step": 4222 }, { "epoch": 0.27, "learning_rate": 0.00017110604071994166, "loss": 0.1504, "step": 4223 }, { "epoch": 0.27, "learning_rate": 0.0001710914374486437, "loss": 0.1616, "step": 4224 }, { "epoch": 0.27, "learning_rate": 0.00017107683111147376, "loss": 0.1363, "step": 4225 }, { "epoch": 0.27, "learning_rate": 0.00017106222170906185, "loss": 0.1573, "step": 4226 }, { "epoch": 0.27, "learning_rate": 0.00017104760924203797, "loss": 0.1482, "step": 4227 }, { "epoch": 0.27, "learning_rate": 0.0001710329937110323, "loss": 0.1641, "step": 4228 }, { "epoch": 0.27, "learning_rate": 0.00017101837511667513, "loss": 0.1527, "step": 4229 }, { "epoch": 0.27, "learning_rate": 0.0001710037534595969, "loss": 0.1564, "step": 4230 }, { "epoch": 0.27, "learning_rate": 0.0001709891287404282, "loss": 0.17, "step": 4231 }, { "epoch": 0.27, "learning_rate": 0.00017097450095979975, "loss": 0.1578, "step": 4232 }, { "epoch": 0.27, "learning_rate": 0.0001709598701183423, "loss": 0.1596, "step": 4233 }, { "epoch": 0.27, "learning_rate": 0.00017094523621668693, "loss": 0.1551, "step": 4234 }, { "epoch": 0.27, "learning_rate": 0.0001709305992554646, "loss": 0.1444, "step": 4235 }, { "epoch": 0.27, "learning_rate": 0.0001709159592353067, "loss": 0.1678, "step": 4236 }, { "epoch": 0.27, "learning_rate": 0.00017090131615684446, "loss": 0.1686, "step": 4237 }, { "epoch": 0.27, "learning_rate": 0.0001708866700207094, "loss": 0.164, "step": 4238 }, { "epoch": 0.27, "learning_rate": 0.00017087202082753325, "loss": 0.1629, "step": 4239 }, { "epoch": 0.27, "learning_rate": 0.00017085736857794763, "loss": 0.1467, "step": 4240 }, { "epoch": 0.27, "learning_rate": 0.00017084271327258448, "loss": 0.1672, "step": 4241 }, { "epoch": 0.27, "learning_rate": 0.0001708280549120759, "loss": 0.1639, "step": 4242 }, { "epoch": 0.27, "learning_rate": 0.0001708133934970539, "loss": 0.1629, "step": 4243 }, { "epoch": 0.27, "learning_rate": 0.00017079872902815084, "loss": 0.1694, "step": 4244 }, { "epoch": 0.27, "learning_rate": 0.00017078406150599916, "loss": 0.1641, "step": 4245 }, { "epoch": 0.27, "learning_rate": 0.00017076939093123137, "loss": 0.15, "step": 4246 }, { "epoch": 0.27, "learning_rate": 0.00017075471730448015, "loss": 0.1469, "step": 4247 }, { "epoch": 0.27, "learning_rate": 0.00017074004062637833, "loss": 0.1508, "step": 4248 }, { "epoch": 0.27, "learning_rate": 0.00017072536089755884, "loss": 0.1701, "step": 4249 }, { "epoch": 0.27, "learning_rate": 0.00017071067811865476, "loss": 0.1526, "step": 4250 }, { "epoch": 0.27, "learning_rate": 0.00017069599229029928, "loss": 0.1505, "step": 4251 }, { "epoch": 0.27, "learning_rate": 0.0001706813034131258, "loss": 0.1461, "step": 4252 }, { "epoch": 0.27, "learning_rate": 0.00017066661148776766, "loss": 0.1569, "step": 4253 }, { "epoch": 0.27, "learning_rate": 0.00017065191651485862, "loss": 0.1506, "step": 4254 }, { "epoch": 0.27, "learning_rate": 0.00017063721849503227, "loss": 0.1426, "step": 4255 }, { "epoch": 0.27, "learning_rate": 0.00017062251742892256, "loss": 0.1593, "step": 4256 }, { "epoch": 0.27, "learning_rate": 0.00017060781331716348, "loss": 0.1667, "step": 4257 }, { "epoch": 0.27, "learning_rate": 0.0001705931061603891, "loss": 0.1656, "step": 4258 }, { "epoch": 0.27, "learning_rate": 0.00017057839595923375, "loss": 0.1413, "step": 4259 }, { "epoch": 0.27, "learning_rate": 0.00017056368271433175, "loss": 0.1612, "step": 4260 }, { "epoch": 0.27, "learning_rate": 0.00017054896642631766, "loss": 0.145, "step": 4261 }, { "epoch": 0.27, "learning_rate": 0.00017053424709582613, "loss": 0.1414, "step": 4262 }, { "epoch": 0.27, "learning_rate": 0.00017051952472349195, "loss": 0.1615, "step": 4263 }, { "epoch": 0.27, "learning_rate": 0.00017050479930995, "loss": 0.1346, "step": 4264 }, { "epoch": 0.27, "learning_rate": 0.00017049007085583533, "loss": 0.1464, "step": 4265 }, { "epoch": 0.27, "learning_rate": 0.00017047533936178315, "loss": 0.1573, "step": 4266 }, { "epoch": 0.27, "learning_rate": 0.0001704606048284287, "loss": 0.1378, "step": 4267 }, { "epoch": 0.27, "learning_rate": 0.00017044586725640756, "loss": 0.1454, "step": 4268 }, { "epoch": 0.27, "learning_rate": 0.0001704311266463551, "loss": 0.1418, "step": 4269 }, { "epoch": 0.27, "learning_rate": 0.00017041638299890717, "loss": 0.146, "step": 4270 }, { "epoch": 0.27, "learning_rate": 0.00017040163631469954, "loss": 0.1606, "step": 4271 }, { "epoch": 0.27, "learning_rate": 0.0001703868865943682, "loss": 0.188, "step": 4272 }, { "epoch": 0.27, "learning_rate": 0.0001703721338385492, "loss": 0.1541, "step": 4273 }, { "epoch": 0.27, "learning_rate": 0.00017035737804787883, "loss": 0.1619, "step": 4274 }, { "epoch": 0.27, "learning_rate": 0.00017034261922299336, "loss": 0.1461, "step": 4275 }, { "epoch": 0.27, "learning_rate": 0.00017032785736452937, "loss": 0.1375, "step": 4276 }, { "epoch": 0.27, "learning_rate": 0.0001703130924731234, "loss": 0.1582, "step": 4277 }, { "epoch": 0.27, "learning_rate": 0.0001702983245494122, "loss": 0.1467, "step": 4278 }, { "epoch": 0.27, "learning_rate": 0.00017028355359403273, "loss": 0.1549, "step": 4279 }, { "epoch": 0.27, "learning_rate": 0.0001702687796076219, "loss": 0.1808, "step": 4280 }, { "epoch": 0.27, "learning_rate": 0.00017025400259081694, "loss": 0.1434, "step": 4281 }, { "epoch": 0.27, "learning_rate": 0.000170239222544255, "loss": 0.174, "step": 4282 }, { "epoch": 0.27, "learning_rate": 0.0001702244394685736, "loss": 0.1479, "step": 4283 }, { "epoch": 0.27, "learning_rate": 0.0001702096533644102, "loss": 0.1463, "step": 4284 }, { "epoch": 0.27, "learning_rate": 0.0001701948642324025, "loss": 0.155, "step": 4285 }, { "epoch": 0.27, "learning_rate": 0.00017018007207318827, "loss": 0.1521, "step": 4286 }, { "epoch": 0.27, "learning_rate": 0.00017016527688740545, "loss": 0.1779, "step": 4287 }, { "epoch": 0.27, "learning_rate": 0.0001701504786756921, "loss": 0.1343, "step": 4288 }, { "epoch": 0.28, "learning_rate": 0.00017013567743868633, "loss": 0.1632, "step": 4289 }, { "epoch": 0.28, "learning_rate": 0.00017012087317702656, "loss": 0.1322, "step": 4290 }, { "epoch": 0.28, "learning_rate": 0.00017010606589135116, "loss": 0.1469, "step": 4291 }, { "epoch": 0.28, "learning_rate": 0.00017009125558229874, "loss": 0.1427, "step": 4292 }, { "epoch": 0.28, "learning_rate": 0.000170076442250508, "loss": 0.1582, "step": 4293 }, { "epoch": 0.28, "learning_rate": 0.0001700616258966178, "loss": 0.1464, "step": 4294 }, { "epoch": 0.28, "learning_rate": 0.00017004680652126704, "loss": 0.1803, "step": 4295 }, { "epoch": 0.28, "learning_rate": 0.00017003198412509488, "loss": 0.1787, "step": 4296 }, { "epoch": 0.28, "learning_rate": 0.00017001715870874052, "loss": 0.1509, "step": 4297 }, { "epoch": 0.28, "learning_rate": 0.00017000233027284332, "loss": 0.1367, "step": 4298 }, { "epoch": 0.28, "learning_rate": 0.00016998749881804278, "loss": 0.1514, "step": 4299 }, { "epoch": 0.28, "learning_rate": 0.00016997266434497848, "loss": 0.1595, "step": 4300 }, { "epoch": 0.28, "learning_rate": 0.0001699578268542902, "loss": 0.1665, "step": 4301 }, { "epoch": 0.28, "learning_rate": 0.00016994298634661785, "loss": 0.1729, "step": 4302 }, { "epoch": 0.28, "learning_rate": 0.0001699281428226014, "loss": 0.1665, "step": 4303 }, { "epoch": 0.28, "learning_rate": 0.00016991329628288093, "loss": 0.1456, "step": 4304 }, { "epoch": 0.28, "learning_rate": 0.0001698984467280968, "loss": 0.1729, "step": 4305 }, { "epoch": 0.28, "learning_rate": 0.0001698835941588894, "loss": 0.1364, "step": 4306 }, { "epoch": 0.28, "learning_rate": 0.0001698687385758992, "loss": 0.1654, "step": 4307 }, { "epoch": 0.28, "learning_rate": 0.00016985387997976693, "loss": 0.1483, "step": 4308 }, { "epoch": 0.28, "learning_rate": 0.0001698390183711333, "loss": 0.1521, "step": 4309 }, { "epoch": 0.28, "learning_rate": 0.00016982415375063926, "loss": 0.1464, "step": 4310 }, { "epoch": 0.28, "learning_rate": 0.0001698092861189259, "loss": 0.1577, "step": 4311 }, { "epoch": 0.28, "learning_rate": 0.00016979441547663435, "loss": 0.1535, "step": 4312 }, { "epoch": 0.28, "learning_rate": 0.00016977954182440595, "loss": 0.1373, "step": 4313 }, { "epoch": 0.28, "learning_rate": 0.00016976466516288209, "loss": 0.1462, "step": 4314 }, { "epoch": 0.28, "learning_rate": 0.00016974978549270443, "loss": 0.1643, "step": 4315 }, { "epoch": 0.28, "learning_rate": 0.00016973490281451455, "loss": 0.18, "step": 4316 }, { "epoch": 0.28, "learning_rate": 0.0001697200171289543, "loss": 0.1514, "step": 4317 }, { "epoch": 0.28, "learning_rate": 0.0001697051284366657, "loss": 0.153, "step": 4318 }, { "epoch": 0.28, "learning_rate": 0.00016969023673829082, "loss": 0.1665, "step": 4319 }, { "epoch": 0.28, "learning_rate": 0.00016967534203447184, "loss": 0.1352, "step": 4320 }, { "epoch": 0.28, "learning_rate": 0.00016966044432585115, "loss": 0.1528, "step": 4321 }, { "epoch": 0.28, "learning_rate": 0.00016964554361307118, "loss": 0.1349, "step": 4322 }, { "epoch": 0.28, "learning_rate": 0.00016963063989677455, "loss": 0.1351, "step": 4323 }, { "epoch": 0.28, "learning_rate": 0.000169615733177604, "loss": 0.1365, "step": 4324 }, { "epoch": 0.28, "learning_rate": 0.0001696008234562024, "loss": 0.1541, "step": 4325 }, { "epoch": 0.28, "learning_rate": 0.00016958591073321273, "loss": 0.1427, "step": 4326 }, { "epoch": 0.28, "learning_rate": 0.0001695709950092781, "loss": 0.1656, "step": 4327 }, { "epoch": 0.28, "learning_rate": 0.0001695560762850418, "loss": 0.161, "step": 4328 }, { "epoch": 0.28, "learning_rate": 0.00016954115456114716, "loss": 0.164, "step": 4329 }, { "epoch": 0.28, "learning_rate": 0.00016952622983823777, "loss": 0.1623, "step": 4330 }, { "epoch": 0.28, "learning_rate": 0.00016951130211695717, "loss": 0.163, "step": 4331 }, { "epoch": 0.28, "learning_rate": 0.0001694963713979492, "loss": 0.1544, "step": 4332 }, { "epoch": 0.28, "learning_rate": 0.00016948143768185774, "loss": 0.1418, "step": 4333 }, { "epoch": 0.28, "learning_rate": 0.00016946650096932682, "loss": 0.1585, "step": 4334 }, { "epoch": 0.28, "learning_rate": 0.00016945156126100055, "loss": 0.1598, "step": 4335 }, { "epoch": 0.28, "learning_rate": 0.0001694366185575233, "loss": 0.1539, "step": 4336 }, { "epoch": 0.28, "learning_rate": 0.00016942167285953946, "loss": 0.16, "step": 4337 }, { "epoch": 0.28, "learning_rate": 0.00016940672416769354, "loss": 0.1577, "step": 4338 }, { "epoch": 0.28, "learning_rate": 0.0001693917724826302, "loss": 0.1552, "step": 4339 }, { "epoch": 0.28, "learning_rate": 0.00016937681780499432, "loss": 0.1359, "step": 4340 }, { "epoch": 0.28, "learning_rate": 0.00016936186013543077, "loss": 0.1499, "step": 4341 }, { "epoch": 0.28, "learning_rate": 0.00016934689947458465, "loss": 0.1429, "step": 4342 }, { "epoch": 0.28, "learning_rate": 0.0001693319358231011, "loss": 0.1509, "step": 4343 }, { "epoch": 0.28, "learning_rate": 0.00016931696918162547, "loss": 0.141, "step": 4344 }, { "epoch": 0.28, "learning_rate": 0.00016930199955080322, "loss": 0.1523, "step": 4345 }, { "epoch": 0.28, "learning_rate": 0.00016928702693127989, "loss": 0.1941, "step": 4346 }, { "epoch": 0.28, "learning_rate": 0.00016927205132370122, "loss": 0.1617, "step": 4347 }, { "epoch": 0.28, "learning_rate": 0.000169257072728713, "loss": 0.1507, "step": 4348 }, { "epoch": 0.28, "learning_rate": 0.00016924209114696127, "loss": 0.1335, "step": 4349 }, { "epoch": 0.28, "learning_rate": 0.00016922710657909205, "loss": 0.1696, "step": 4350 }, { "epoch": 0.28, "learning_rate": 0.00016921211902575157, "loss": 0.1642, "step": 4351 }, { "epoch": 0.28, "learning_rate": 0.0001691971284875862, "loss": 0.1408, "step": 4352 }, { "epoch": 0.28, "learning_rate": 0.00016918213496524247, "loss": 0.1876, "step": 4353 }, { "epoch": 0.28, "learning_rate": 0.0001691671384593669, "loss": 0.1693, "step": 4354 }, { "epoch": 0.28, "learning_rate": 0.00016915213897060623, "loss": 0.1603, "step": 4355 }, { "epoch": 0.28, "learning_rate": 0.00016913713649960738, "loss": 0.1739, "step": 4356 }, { "epoch": 0.28, "learning_rate": 0.0001691221310470173, "loss": 0.1566, "step": 4357 }, { "epoch": 0.28, "learning_rate": 0.00016910712261348313, "loss": 0.1624, "step": 4358 }, { "epoch": 0.28, "learning_rate": 0.0001690921111996521, "loss": 0.1391, "step": 4359 }, { "epoch": 0.28, "learning_rate": 0.00016907709680617162, "loss": 0.1526, "step": 4360 }, { "epoch": 0.28, "learning_rate": 0.0001690620794336892, "loss": 0.1645, "step": 4361 }, { "epoch": 0.28, "learning_rate": 0.00016904705908285243, "loss": 0.1512, "step": 4362 }, { "epoch": 0.28, "learning_rate": 0.00016903203575430913, "loss": 0.1574, "step": 4363 }, { "epoch": 0.28, "learning_rate": 0.00016901700944870713, "loss": 0.1276, "step": 4364 }, { "epoch": 0.28, "learning_rate": 0.00016900198016669448, "loss": 0.1516, "step": 4365 }, { "epoch": 0.28, "learning_rate": 0.00016898694790891936, "loss": 0.1382, "step": 4366 }, { "epoch": 0.28, "learning_rate": 0.00016897191267603, "loss": 0.1753, "step": 4367 }, { "epoch": 0.28, "learning_rate": 0.00016895687446867485, "loss": 0.1858, "step": 4368 }, { "epoch": 0.28, "learning_rate": 0.0001689418332875024, "loss": 0.1615, "step": 4369 }, { "epoch": 0.28, "learning_rate": 0.00016892678913316137, "loss": 0.1853, "step": 4370 }, { "epoch": 0.28, "learning_rate": 0.00016891174200630047, "loss": 0.1544, "step": 4371 }, { "epoch": 0.28, "learning_rate": 0.00016889669190756868, "loss": 0.1451, "step": 4372 }, { "epoch": 0.28, "learning_rate": 0.000168881638837615, "loss": 0.1464, "step": 4373 }, { "epoch": 0.28, "learning_rate": 0.00016886658279708866, "loss": 0.1543, "step": 4374 }, { "epoch": 0.28, "learning_rate": 0.00016885152378663895, "loss": 0.1691, "step": 4375 }, { "epoch": 0.28, "learning_rate": 0.00016883646180691525, "loss": 0.1538, "step": 4376 }, { "epoch": 0.28, "learning_rate": 0.00016882139685856718, "loss": 0.1491, "step": 4377 }, { "epoch": 0.28, "learning_rate": 0.00016880632894224442, "loss": 0.163, "step": 4378 }, { "epoch": 0.28, "learning_rate": 0.00016879125805859674, "loss": 0.1567, "step": 4379 }, { "epoch": 0.28, "learning_rate": 0.00016877618420827413, "loss": 0.1685, "step": 4380 }, { "epoch": 0.28, "learning_rate": 0.00016876110739192664, "loss": 0.1561, "step": 4381 }, { "epoch": 0.28, "learning_rate": 0.00016874602761020446, "loss": 0.1513, "step": 4382 }, { "epoch": 0.28, "learning_rate": 0.00016873094486375795, "loss": 0.1569, "step": 4383 }, { "epoch": 0.28, "learning_rate": 0.00016871585915323757, "loss": 0.1426, "step": 4384 }, { "epoch": 0.28, "learning_rate": 0.00016870077047929382, "loss": 0.1365, "step": 4385 }, { "epoch": 0.28, "learning_rate": 0.00016868567884257752, "loss": 0.1491, "step": 4386 }, { "epoch": 0.28, "learning_rate": 0.0001686705842437394, "loss": 0.1454, "step": 4387 }, { "epoch": 0.28, "learning_rate": 0.00016865548668343054, "loss": 0.1789, "step": 4388 }, { "epoch": 0.28, "learning_rate": 0.00016864038616230196, "loss": 0.1311, "step": 4389 }, { "epoch": 0.28, "learning_rate": 0.00016862528268100487, "loss": 0.1616, "step": 4390 }, { "epoch": 0.28, "learning_rate": 0.00016861017624019068, "loss": 0.1716, "step": 4391 }, { "epoch": 0.28, "learning_rate": 0.0001685950668405108, "loss": 0.1552, "step": 4392 }, { "epoch": 0.28, "learning_rate": 0.00016857995448261689, "loss": 0.1561, "step": 4393 }, { "epoch": 0.28, "learning_rate": 0.00016856483916716065, "loss": 0.1374, "step": 4394 }, { "epoch": 0.28, "learning_rate": 0.00016854972089479396, "loss": 0.1514, "step": 4395 }, { "epoch": 0.28, "learning_rate": 0.00016853459966616877, "loss": 0.1386, "step": 4396 }, { "epoch": 0.28, "learning_rate": 0.00016851947548193727, "loss": 0.1583, "step": 4397 }, { "epoch": 0.28, "learning_rate": 0.00016850434834275163, "loss": 0.1353, "step": 4398 }, { "epoch": 0.28, "learning_rate": 0.00016848921824926422, "loss": 0.1666, "step": 4399 }, { "epoch": 0.28, "learning_rate": 0.00016847408520212758, "loss": 0.1532, "step": 4400 }, { "epoch": 0.28, "learning_rate": 0.00016845894920199429, "loss": 0.149, "step": 4401 }, { "epoch": 0.28, "learning_rate": 0.0001684438102495171, "loss": 0.1496, "step": 4402 }, { "epoch": 0.28, "learning_rate": 0.000168428668345349, "loss": 0.1478, "step": 4403 }, { "epoch": 0.28, "learning_rate": 0.00016841352349014285, "loss": 0.1523, "step": 4404 }, { "epoch": 0.28, "learning_rate": 0.00016839837568455182, "loss": 0.1575, "step": 4405 }, { "epoch": 0.28, "learning_rate": 0.00016838322492922924, "loss": 0.1531, "step": 4406 }, { "epoch": 0.28, "learning_rate": 0.00016836807122482843, "loss": 0.1526, "step": 4407 }, { "epoch": 0.28, "learning_rate": 0.00016835291457200293, "loss": 0.1653, "step": 4408 }, { "epoch": 0.28, "learning_rate": 0.0001683377549714064, "loss": 0.1551, "step": 4409 }, { "epoch": 0.28, "learning_rate": 0.00016832259242369253, "loss": 0.1566, "step": 4410 }, { "epoch": 0.28, "learning_rate": 0.0001683074269295153, "loss": 0.1563, "step": 4411 }, { "epoch": 0.28, "learning_rate": 0.0001682922584895287, "loss": 0.1541, "step": 4412 }, { "epoch": 0.28, "learning_rate": 0.00016827708710438692, "loss": 0.1673, "step": 4413 }, { "epoch": 0.28, "learning_rate": 0.00016826191277474414, "loss": 0.1341, "step": 4414 }, { "epoch": 0.28, "learning_rate": 0.00016824673550125486, "loss": 0.1461, "step": 4415 }, { "epoch": 0.28, "learning_rate": 0.0001682315552845736, "loss": 0.1473, "step": 4416 }, { "epoch": 0.28, "learning_rate": 0.000168216372125355, "loss": 0.1416, "step": 4417 }, { "epoch": 0.28, "learning_rate": 0.00016820118602425382, "loss": 0.1682, "step": 4418 }, { "epoch": 0.28, "learning_rate": 0.000168185996981925, "loss": 0.1418, "step": 4419 }, { "epoch": 0.28, "learning_rate": 0.0001681708049990236, "loss": 0.1483, "step": 4420 }, { "epoch": 0.28, "learning_rate": 0.00016815561007620472, "loss": 0.1461, "step": 4421 }, { "epoch": 0.28, "learning_rate": 0.0001681404122141237, "loss": 0.1509, "step": 4422 }, { "epoch": 0.28, "learning_rate": 0.00016812521141343595, "loss": 0.1431, "step": 4423 }, { "epoch": 0.28, "learning_rate": 0.000168110007674797, "loss": 0.1558, "step": 4424 }, { "epoch": 0.28, "learning_rate": 0.0001680948009988626, "loss": 0.1692, "step": 4425 }, { "epoch": 0.28, "learning_rate": 0.00016807959138628845, "loss": 0.142, "step": 4426 }, { "epoch": 0.28, "learning_rate": 0.00016806437883773052, "loss": 0.135, "step": 4427 }, { "epoch": 0.28, "learning_rate": 0.00016804916335384486, "loss": 0.1509, "step": 4428 }, { "epoch": 0.28, "learning_rate": 0.00016803394493528765, "loss": 0.1695, "step": 4429 }, { "epoch": 0.28, "learning_rate": 0.00016801872358271518, "loss": 0.1634, "step": 4430 }, { "epoch": 0.28, "learning_rate": 0.00016800349929678391, "loss": 0.1422, "step": 4431 }, { "epoch": 0.28, "learning_rate": 0.0001679882720781504, "loss": 0.1432, "step": 4432 }, { "epoch": 0.28, "learning_rate": 0.0001679730419274713, "loss": 0.149, "step": 4433 }, { "epoch": 0.28, "learning_rate": 0.0001679578088454034, "loss": 0.1592, "step": 4434 }, { "epoch": 0.28, "learning_rate": 0.00016794257283260375, "loss": 0.1525, "step": 4435 }, { "epoch": 0.28, "learning_rate": 0.00016792733388972932, "loss": 0.1612, "step": 4436 }, { "epoch": 0.28, "learning_rate": 0.0001679120920174373, "loss": 0.1492, "step": 4437 }, { "epoch": 0.28, "learning_rate": 0.0001678968472163851, "loss": 0.1557, "step": 4438 }, { "epoch": 0.28, "learning_rate": 0.00016788159948723003, "loss": 0.1465, "step": 4439 }, { "epoch": 0.28, "learning_rate": 0.00016786634883062974, "loss": 0.1707, "step": 4440 }, { "epoch": 0.28, "learning_rate": 0.00016785109524724193, "loss": 0.1524, "step": 4441 }, { "epoch": 0.28, "learning_rate": 0.0001678358387377244, "loss": 0.1554, "step": 4442 }, { "epoch": 0.28, "learning_rate": 0.00016782057930273507, "loss": 0.1704, "step": 4443 }, { "epoch": 0.28, "learning_rate": 0.00016780531694293206, "loss": 0.1552, "step": 4444 }, { "epoch": 0.29, "learning_rate": 0.00016779005165897359, "loss": 0.1435, "step": 4445 }, { "epoch": 0.29, "learning_rate": 0.00016777478345151794, "loss": 0.1437, "step": 4446 }, { "epoch": 0.29, "learning_rate": 0.00016775951232122358, "loss": 0.1644, "step": 4447 }, { "epoch": 0.29, "learning_rate": 0.00016774423826874908, "loss": 0.1586, "step": 4448 }, { "epoch": 0.29, "learning_rate": 0.00016772896129475315, "loss": 0.1421, "step": 4449 }, { "epoch": 0.29, "learning_rate": 0.0001677136813998946, "loss": 0.1432, "step": 4450 }, { "epoch": 0.29, "learning_rate": 0.00016769839858483246, "loss": 0.1689, "step": 4451 }, { "epoch": 0.29, "learning_rate": 0.00016768311285022573, "loss": 0.1486, "step": 4452 }, { "epoch": 0.29, "learning_rate": 0.00016766782419673368, "loss": 0.1768, "step": 4453 }, { "epoch": 0.29, "learning_rate": 0.00016765253262501559, "loss": 0.1682, "step": 4454 }, { "epoch": 0.29, "learning_rate": 0.00016763723813573096, "loss": 0.1622, "step": 4455 }, { "epoch": 0.29, "learning_rate": 0.00016762194072953937, "loss": 0.157, "step": 4456 }, { "epoch": 0.29, "learning_rate": 0.00016760664040710047, "loss": 0.1581, "step": 4457 }, { "epoch": 0.29, "learning_rate": 0.00016759133716907424, "loss": 0.1688, "step": 4458 }, { "epoch": 0.29, "learning_rate": 0.0001675760310161205, "loss": 0.1512, "step": 4459 }, { "epoch": 0.29, "learning_rate": 0.00016756072194889943, "loss": 0.1505, "step": 4460 }, { "epoch": 0.29, "learning_rate": 0.00016754540996807117, "loss": 0.1591, "step": 4461 }, { "epoch": 0.29, "learning_rate": 0.00016753009507429614, "loss": 0.1469, "step": 4462 }, { "epoch": 0.29, "learning_rate": 0.00016751477726823476, "loss": 0.1605, "step": 4463 }, { "epoch": 0.29, "learning_rate": 0.00016749945655054763, "loss": 0.141, "step": 4464 }, { "epoch": 0.29, "learning_rate": 0.00016748413292189547, "loss": 0.1581, "step": 4465 }, { "epoch": 0.29, "learning_rate": 0.00016746880638293912, "loss": 0.1374, "step": 4466 }, { "epoch": 0.29, "learning_rate": 0.0001674534769343395, "loss": 0.1593, "step": 4467 }, { "epoch": 0.29, "learning_rate": 0.00016743814457675785, "loss": 0.148, "step": 4468 }, { "epoch": 0.29, "learning_rate": 0.00016742280931085523, "loss": 0.1784, "step": 4469 }, { "epoch": 0.29, "learning_rate": 0.00016740747113729306, "loss": 0.1519, "step": 4470 }, { "epoch": 0.29, "learning_rate": 0.0001673921300567328, "loss": 0.1401, "step": 4471 }, { "epoch": 0.29, "learning_rate": 0.00016737678606983607, "loss": 0.1526, "step": 4472 }, { "epoch": 0.29, "learning_rate": 0.00016736143917726453, "loss": 0.1626, "step": 4473 }, { "epoch": 0.29, "learning_rate": 0.00016734608937968008, "loss": 0.1636, "step": 4474 }, { "epoch": 0.29, "learning_rate": 0.00016733073667774467, "loss": 0.1675, "step": 4475 }, { "epoch": 0.29, "learning_rate": 0.00016731538107212039, "loss": 0.1697, "step": 4476 }, { "epoch": 0.29, "learning_rate": 0.0001673000225634695, "loss": 0.1401, "step": 4477 }, { "epoch": 0.29, "learning_rate": 0.00016728466115245428, "loss": 0.1553, "step": 4478 }, { "epoch": 0.29, "learning_rate": 0.00016726929683973725, "loss": 0.1418, "step": 4479 }, { "epoch": 0.29, "learning_rate": 0.00016725392962598106, "loss": 0.1688, "step": 4480 }, { "epoch": 0.29, "learning_rate": 0.00016723855951184834, "loss": 0.1495, "step": 4481 }, { "epoch": 0.29, "learning_rate": 0.00016722318649800192, "loss": 0.1551, "step": 4482 }, { "epoch": 0.29, "learning_rate": 0.00016720781058510487, "loss": 0.1507, "step": 4483 }, { "epoch": 0.29, "learning_rate": 0.00016719243177382024, "loss": 0.1607, "step": 4484 }, { "epoch": 0.29, "learning_rate": 0.00016717705006481123, "loss": 0.137, "step": 4485 }, { "epoch": 0.29, "learning_rate": 0.00016716166545874123, "loss": 0.1534, "step": 4486 }, { "epoch": 0.29, "learning_rate": 0.0001671462779562737, "loss": 0.1375, "step": 4487 }, { "epoch": 0.29, "learning_rate": 0.0001671308875580722, "loss": 0.1686, "step": 4488 }, { "epoch": 0.29, "learning_rate": 0.00016711549426480052, "loss": 0.1492, "step": 4489 }, { "epoch": 0.29, "learning_rate": 0.00016710009807712246, "loss": 0.1582, "step": 4490 }, { "epoch": 0.29, "learning_rate": 0.000167084698995702, "loss": 0.1674, "step": 4491 }, { "epoch": 0.29, "learning_rate": 0.00016706929702120323, "loss": 0.1769, "step": 4492 }, { "epoch": 0.29, "learning_rate": 0.00016705389215429043, "loss": 0.1656, "step": 4493 }, { "epoch": 0.29, "learning_rate": 0.00016703848439562785, "loss": 0.1402, "step": 4494 }, { "epoch": 0.29, "learning_rate": 0.00016702307374588004, "loss": 0.163, "step": 4495 }, { "epoch": 0.29, "learning_rate": 0.00016700766020571157, "loss": 0.1572, "step": 4496 }, { "epoch": 0.29, "learning_rate": 0.00016699224377578713, "loss": 0.1609, "step": 4497 }, { "epoch": 0.29, "learning_rate": 0.0001669768244567716, "loss": 0.161, "step": 4498 }, { "epoch": 0.29, "learning_rate": 0.00016696140224932996, "loss": 0.1514, "step": 4499 }, { "epoch": 0.29, "learning_rate": 0.00016694597715412727, "loss": 0.1484, "step": 4500 }, { "epoch": 0.29, "learning_rate": 0.00016693054917182878, "loss": 0.1488, "step": 4501 }, { "epoch": 0.29, "learning_rate": 0.0001669151183030998, "loss": 0.1554, "step": 4502 }, { "epoch": 0.29, "learning_rate": 0.00016689968454860582, "loss": 0.1384, "step": 4503 }, { "epoch": 0.29, "learning_rate": 0.00016688424790901242, "loss": 0.1578, "step": 4504 }, { "epoch": 0.29, "learning_rate": 0.00016686880838498536, "loss": 0.1657, "step": 4505 }, { "epoch": 0.29, "learning_rate": 0.00016685336597719043, "loss": 0.1704, "step": 4506 }, { "epoch": 0.29, "learning_rate": 0.0001668379206862936, "loss": 0.1481, "step": 4507 }, { "epoch": 0.29, "learning_rate": 0.000166822472512961, "loss": 0.1594, "step": 4508 }, { "epoch": 0.29, "learning_rate": 0.00016680702145785878, "loss": 0.1315, "step": 4509 }, { "epoch": 0.29, "learning_rate": 0.00016679156752165333, "loss": 0.1422, "step": 4510 }, { "epoch": 0.29, "learning_rate": 0.00016677611070501112, "loss": 0.1466, "step": 4511 }, { "epoch": 0.29, "learning_rate": 0.00016676065100859868, "loss": 0.1534, "step": 4512 }, { "epoch": 0.29, "learning_rate": 0.00016674518843308277, "loss": 0.1482, "step": 4513 }, { "epoch": 0.29, "learning_rate": 0.0001667297229791302, "loss": 0.1513, "step": 4514 }, { "epoch": 0.29, "learning_rate": 0.00016671425464740797, "loss": 0.1712, "step": 4515 }, { "epoch": 0.29, "learning_rate": 0.00016669878343858309, "loss": 0.1377, "step": 4516 }, { "epoch": 0.29, "learning_rate": 0.00016668330935332287, "loss": 0.137, "step": 4517 }, { "epoch": 0.29, "learning_rate": 0.00016666783239229453, "loss": 0.1611, "step": 4518 }, { "epoch": 0.29, "learning_rate": 0.0001666523525561656, "loss": 0.1414, "step": 4519 }, { "epoch": 0.29, "learning_rate": 0.00016663686984560365, "loss": 0.1634, "step": 4520 }, { "epoch": 0.29, "learning_rate": 0.00016662138426127635, "loss": 0.1527, "step": 4521 }, { "epoch": 0.29, "learning_rate": 0.00016660589580385154, "loss": 0.1681, "step": 4522 }, { "epoch": 0.29, "learning_rate": 0.0001665904044739972, "loss": 0.1824, "step": 4523 }, { "epoch": 0.29, "learning_rate": 0.00016657491027238138, "loss": 0.1522, "step": 4524 }, { "epoch": 0.29, "learning_rate": 0.0001665594131996723, "loss": 0.1501, "step": 4525 }, { "epoch": 0.29, "learning_rate": 0.00016654391325653824, "loss": 0.1422, "step": 4526 }, { "epoch": 0.29, "learning_rate": 0.0001665284104436477, "loss": 0.1681, "step": 4527 }, { "epoch": 0.29, "learning_rate": 0.0001665129047616692, "loss": 0.1526, "step": 4528 }, { "epoch": 0.29, "learning_rate": 0.00016649739621127146, "loss": 0.1821, "step": 4529 }, { "epoch": 0.29, "learning_rate": 0.00016648188479312332, "loss": 0.165, "step": 4530 }, { "epoch": 0.29, "learning_rate": 0.0001664663705078937, "loss": 0.1484, "step": 4531 }, { "epoch": 0.29, "learning_rate": 0.00016645085335625166, "loss": 0.1375, "step": 4532 }, { "epoch": 0.29, "learning_rate": 0.00016643533333886637, "loss": 0.1483, "step": 4533 }, { "epoch": 0.29, "learning_rate": 0.0001664198104564072, "loss": 0.1671, "step": 4534 }, { "epoch": 0.29, "learning_rate": 0.00016640428470954353, "loss": 0.1543, "step": 4535 }, { "epoch": 0.29, "learning_rate": 0.00016638875609894494, "loss": 0.1574, "step": 4536 }, { "epoch": 0.29, "learning_rate": 0.00016637322462528112, "loss": 0.1612, "step": 4537 }, { "epoch": 0.29, "learning_rate": 0.00016635769028922187, "loss": 0.1334, "step": 4538 }, { "epoch": 0.29, "learning_rate": 0.00016634215309143714, "loss": 0.1447, "step": 4539 }, { "epoch": 0.29, "learning_rate": 0.00016632661303259691, "loss": 0.1533, "step": 4540 }, { "epoch": 0.29, "learning_rate": 0.00016631107011337145, "loss": 0.1286, "step": 4541 }, { "epoch": 0.29, "learning_rate": 0.00016629552433443102, "loss": 0.1492, "step": 4542 }, { "epoch": 0.29, "learning_rate": 0.00016627997569644605, "loss": 0.1366, "step": 4543 }, { "epoch": 0.29, "learning_rate": 0.00016626442420008706, "loss": 0.1499, "step": 4544 }, { "epoch": 0.29, "learning_rate": 0.00016624886984602477, "loss": 0.1489, "step": 4545 }, { "epoch": 0.29, "learning_rate": 0.0001662333126349299, "loss": 0.1605, "step": 4546 }, { "epoch": 0.29, "learning_rate": 0.00016621775256747348, "loss": 0.1466, "step": 4547 }, { "epoch": 0.29, "learning_rate": 0.00016620218964432643, "loss": 0.1576, "step": 4548 }, { "epoch": 0.29, "learning_rate": 0.00016618662386615998, "loss": 0.1501, "step": 4549 }, { "epoch": 0.29, "learning_rate": 0.00016617105523364538, "loss": 0.1297, "step": 4550 }, { "epoch": 0.29, "learning_rate": 0.00016615548374745407, "loss": 0.172, "step": 4551 }, { "epoch": 0.29, "learning_rate": 0.00016613990940825754, "loss": 0.1469, "step": 4552 }, { "epoch": 0.29, "learning_rate": 0.0001661243322167275, "loss": 0.1521, "step": 4553 }, { "epoch": 0.29, "learning_rate": 0.00016610875217353574, "loss": 0.1442, "step": 4554 }, { "epoch": 0.29, "learning_rate": 0.0001660931692793541, "loss": 0.1674, "step": 4555 }, { "epoch": 0.29, "learning_rate": 0.00016607758353485462, "loss": 0.1476, "step": 4556 }, { "epoch": 0.29, "learning_rate": 0.00016606199494070945, "loss": 0.1399, "step": 4557 }, { "epoch": 0.29, "learning_rate": 0.00016604640349759085, "loss": 0.1495, "step": 4558 }, { "epoch": 0.29, "learning_rate": 0.0001660308092061713, "loss": 0.1493, "step": 4559 }, { "epoch": 0.29, "learning_rate": 0.00016601521206712318, "loss": 0.1476, "step": 4560 }, { "epoch": 0.29, "learning_rate": 0.00016599961208111922, "loss": 0.1585, "step": 4561 }, { "epoch": 0.29, "learning_rate": 0.00016598400924883216, "loss": 0.1542, "step": 4562 }, { "epoch": 0.29, "learning_rate": 0.00016596840357093484, "loss": 0.1557, "step": 4563 }, { "epoch": 0.29, "learning_rate": 0.00016595279504810035, "loss": 0.1397, "step": 4564 }, { "epoch": 0.29, "learning_rate": 0.00016593718368100173, "loss": 0.1478, "step": 4565 }, { "epoch": 0.29, "learning_rate": 0.0001659215694703123, "loss": 0.1496, "step": 4566 }, { "epoch": 0.29, "learning_rate": 0.0001659059524167054, "loss": 0.1431, "step": 4567 }, { "epoch": 0.29, "learning_rate": 0.00016589033252085455, "loss": 0.153, "step": 4568 }, { "epoch": 0.29, "learning_rate": 0.00016587470978343335, "loss": 0.1711, "step": 4569 }, { "epoch": 0.29, "learning_rate": 0.00016585908420511556, "loss": 0.1647, "step": 4570 }, { "epoch": 0.29, "learning_rate": 0.00016584345578657503, "loss": 0.1635, "step": 4571 }, { "epoch": 0.29, "learning_rate": 0.00016582782452848575, "loss": 0.1682, "step": 4572 }, { "epoch": 0.29, "learning_rate": 0.0001658121904315218, "loss": 0.1709, "step": 4573 }, { "epoch": 0.29, "learning_rate": 0.00016579655349635748, "loss": 0.1644, "step": 4574 }, { "epoch": 0.29, "learning_rate": 0.00016578091372366712, "loss": 0.1575, "step": 4575 }, { "epoch": 0.29, "learning_rate": 0.00016576527111412516, "loss": 0.1556, "step": 4576 }, { "epoch": 0.29, "learning_rate": 0.00016574962566840625, "loss": 0.145, "step": 4577 }, { "epoch": 0.29, "learning_rate": 0.00016573397738718507, "loss": 0.1759, "step": 4578 }, { "epoch": 0.29, "learning_rate": 0.00016571832627113647, "loss": 0.1325, "step": 4579 }, { "epoch": 0.29, "learning_rate": 0.00016570267232093544, "loss": 0.181, "step": 4580 }, { "epoch": 0.29, "learning_rate": 0.00016568701553725706, "loss": 0.172, "step": 4581 }, { "epoch": 0.29, "learning_rate": 0.00016567135592077654, "loss": 0.1557, "step": 4582 }, { "epoch": 0.29, "learning_rate": 0.0001656556934721692, "loss": 0.1478, "step": 4583 }, { "epoch": 0.29, "learning_rate": 0.00016564002819211048, "loss": 0.1574, "step": 4584 }, { "epoch": 0.29, "learning_rate": 0.000165624360081276, "loss": 0.1836, "step": 4585 }, { "epoch": 0.29, "learning_rate": 0.00016560868914034145, "loss": 0.1731, "step": 4586 }, { "epoch": 0.29, "learning_rate": 0.0001655930153699826, "loss": 0.1586, "step": 4587 }, { "epoch": 0.29, "learning_rate": 0.00016557733877087548, "loss": 0.1331, "step": 4588 }, { "epoch": 0.29, "learning_rate": 0.0001655616593436961, "loss": 0.1529, "step": 4589 }, { "epoch": 0.29, "learning_rate": 0.00016554597708912065, "loss": 0.1564, "step": 4590 }, { "epoch": 0.29, "learning_rate": 0.00016553029200782544, "loss": 0.1553, "step": 4591 }, { "epoch": 0.29, "learning_rate": 0.0001655146041004869, "loss": 0.1693, "step": 4592 }, { "epoch": 0.29, "learning_rate": 0.00016549891336778162, "loss": 0.1543, "step": 4593 }, { "epoch": 0.29, "learning_rate": 0.00016548321981038622, "loss": 0.1736, "step": 4594 }, { "epoch": 0.29, "learning_rate": 0.00016546752342897752, "loss": 0.1663, "step": 4595 }, { "epoch": 0.29, "learning_rate": 0.00016545182422423246, "loss": 0.1148, "step": 4596 }, { "epoch": 0.29, "learning_rate": 0.00016543612219682805, "loss": 0.1468, "step": 4597 }, { "epoch": 0.29, "learning_rate": 0.00016542041734744144, "loss": 0.1569, "step": 4598 }, { "epoch": 0.29, "learning_rate": 0.00016540470967674995, "loss": 0.12, "step": 4599 }, { "epoch": 0.29, "learning_rate": 0.00016538899918543094, "loss": 0.1567, "step": 4600 }, { "epoch": 0.3, "learning_rate": 0.000165373285874162, "loss": 0.1653, "step": 4601 }, { "epoch": 0.3, "learning_rate": 0.00016535756974362075, "loss": 0.1798, "step": 4602 }, { "epoch": 0.3, "learning_rate": 0.00016534185079448494, "loss": 0.1781, "step": 4603 }, { "epoch": 0.3, "learning_rate": 0.00016532612902743246, "loss": 0.1738, "step": 4604 }, { "epoch": 0.3, "learning_rate": 0.00016531040444314136, "loss": 0.1505, "step": 4605 }, { "epoch": 0.3, "learning_rate": 0.00016529467704228977, "loss": 0.1538, "step": 4606 }, { "epoch": 0.3, "learning_rate": 0.0001652789468255559, "loss": 0.1349, "step": 4607 }, { "epoch": 0.3, "learning_rate": 0.00016526321379361816, "loss": 0.1527, "step": 4608 }, { "epoch": 0.3, "learning_rate": 0.00016524747794715506, "loss": 0.1558, "step": 4609 }, { "epoch": 0.3, "learning_rate": 0.0001652317392868452, "loss": 0.1487, "step": 4610 }, { "epoch": 0.3, "learning_rate": 0.00016521599781336733, "loss": 0.1499, "step": 4611 }, { "epoch": 0.3, "learning_rate": 0.0001652002535274003, "loss": 0.1294, "step": 4612 }, { "epoch": 0.3, "learning_rate": 0.00016518450642962314, "loss": 0.1733, "step": 4613 }, { "epoch": 0.3, "learning_rate": 0.00016516875652071485, "loss": 0.1522, "step": 4614 }, { "epoch": 0.3, "learning_rate": 0.0001651530038013548, "loss": 0.1602, "step": 4615 }, { "epoch": 0.3, "learning_rate": 0.00016513724827222227, "loss": 0.1447, "step": 4616 }, { "epoch": 0.3, "learning_rate": 0.00016512148993399665, "loss": 0.1456, "step": 4617 }, { "epoch": 0.3, "learning_rate": 0.00016510572878735764, "loss": 0.1416, "step": 4618 }, { "epoch": 0.3, "learning_rate": 0.00016508996483298493, "loss": 0.176, "step": 4619 }, { "epoch": 0.3, "learning_rate": 0.0001650741980715583, "loss": 0.1538, "step": 4620 }, { "epoch": 0.3, "learning_rate": 0.00016505842850375778, "loss": 0.1663, "step": 4621 }, { "epoch": 0.3, "learning_rate": 0.0001650426561302634, "loss": 0.1514, "step": 4622 }, { "epoch": 0.3, "learning_rate": 0.00016502688095175537, "loss": 0.1658, "step": 4623 }, { "epoch": 0.3, "learning_rate": 0.000165011102968914, "loss": 0.1557, "step": 4624 }, { "epoch": 0.3, "learning_rate": 0.0001649953221824197, "loss": 0.1589, "step": 4625 }, { "epoch": 0.3, "learning_rate": 0.00016497953859295308, "loss": 0.1547, "step": 4626 }, { "epoch": 0.3, "learning_rate": 0.00016496375220119476, "loss": 0.1527, "step": 4627 }, { "epoch": 0.3, "learning_rate": 0.0001649479630078256, "loss": 0.1579, "step": 4628 }, { "epoch": 0.3, "learning_rate": 0.00016493217101352649, "loss": 0.1752, "step": 4629 }, { "epoch": 0.3, "learning_rate": 0.00016491637621897847, "loss": 0.1595, "step": 4630 }, { "epoch": 0.3, "learning_rate": 0.0001649005786248627, "loss": 0.139, "step": 4631 }, { "epoch": 0.3, "learning_rate": 0.00016488477823186048, "loss": 0.1675, "step": 4632 }, { "epoch": 0.3, "learning_rate": 0.0001648689750406532, "loss": 0.1484, "step": 4633 }, { "epoch": 0.3, "learning_rate": 0.0001648531690519224, "loss": 0.1576, "step": 4634 }, { "epoch": 0.3, "learning_rate": 0.00016483736026634972, "loss": 0.1751, "step": 4635 }, { "epoch": 0.3, "learning_rate": 0.00016482154868461692, "loss": 0.1628, "step": 4636 }, { "epoch": 0.3, "learning_rate": 0.00016480573430740586, "loss": 0.1835, "step": 4637 }, { "epoch": 0.3, "learning_rate": 0.0001647899171353986, "loss": 0.156, "step": 4638 }, { "epoch": 0.3, "learning_rate": 0.00016477409716927725, "loss": 0.15, "step": 4639 }, { "epoch": 0.3, "learning_rate": 0.000164758274409724, "loss": 0.1913, "step": 4640 }, { "epoch": 0.3, "learning_rate": 0.00016474244885742134, "loss": 0.1426, "step": 4641 }, { "epoch": 0.3, "learning_rate": 0.00016472662051305163, "loss": 0.1535, "step": 4642 }, { "epoch": 0.3, "learning_rate": 0.0001647107893772976, "loss": 0.1318, "step": 4643 }, { "epoch": 0.3, "learning_rate": 0.00016469495545084185, "loss": 0.164, "step": 4644 }, { "epoch": 0.3, "learning_rate": 0.00016467911873436732, "loss": 0.1568, "step": 4645 }, { "epoch": 0.3, "learning_rate": 0.00016466327922855695, "loss": 0.1736, "step": 4646 }, { "epoch": 0.3, "learning_rate": 0.00016464743693409386, "loss": 0.1644, "step": 4647 }, { "epoch": 0.3, "learning_rate": 0.0001646315918516612, "loss": 0.1724, "step": 4648 }, { "epoch": 0.3, "learning_rate": 0.00016461574398194236, "loss": 0.1507, "step": 4649 }, { "epoch": 0.3, "learning_rate": 0.0001645998933256208, "loss": 0.1872, "step": 4650 }, { "epoch": 0.3, "learning_rate": 0.00016458403988338005, "loss": 0.1665, "step": 4651 }, { "epoch": 0.3, "learning_rate": 0.00016456818365590377, "loss": 0.1716, "step": 4652 }, { "epoch": 0.3, "learning_rate": 0.00016455232464387587, "loss": 0.1555, "step": 4653 }, { "epoch": 0.3, "learning_rate": 0.0001645364628479802, "loss": 0.1801, "step": 4654 }, { "epoch": 0.3, "learning_rate": 0.00016452059826890084, "loss": 0.1797, "step": 4655 }, { "epoch": 0.3, "learning_rate": 0.00016450473090732194, "loss": 0.151, "step": 4656 }, { "epoch": 0.3, "learning_rate": 0.00016448886076392784, "loss": 0.1614, "step": 4657 }, { "epoch": 0.3, "learning_rate": 0.0001644729878394029, "loss": 0.1703, "step": 4658 }, { "epoch": 0.3, "learning_rate": 0.00016445711213443172, "loss": 0.1466, "step": 4659 }, { "epoch": 0.3, "learning_rate": 0.00016444123364969885, "loss": 0.1719, "step": 4660 }, { "epoch": 0.3, "learning_rate": 0.00016442535238588918, "loss": 0.1758, "step": 4661 }, { "epoch": 0.3, "learning_rate": 0.0001644094683436875, "loss": 0.1636, "step": 4662 }, { "epoch": 0.3, "learning_rate": 0.00016439358152377887, "loss": 0.1455, "step": 4663 }, { "epoch": 0.3, "learning_rate": 0.00016437769192684845, "loss": 0.1318, "step": 4664 }, { "epoch": 0.3, "learning_rate": 0.00016436179955358143, "loss": 0.1624, "step": 4665 }, { "epoch": 0.3, "learning_rate": 0.00016434590440466323, "loss": 0.1467, "step": 4666 }, { "epoch": 0.3, "learning_rate": 0.00016433000648077926, "loss": 0.1632, "step": 4667 }, { "epoch": 0.3, "learning_rate": 0.00016431410578261524, "loss": 0.1387, "step": 4668 }, { "epoch": 0.3, "learning_rate": 0.0001642982023108568, "loss": 0.1671, "step": 4669 }, { "epoch": 0.3, "learning_rate": 0.00016428229606618987, "loss": 0.1413, "step": 4670 }, { "epoch": 0.3, "learning_rate": 0.00016426638704930037, "loss": 0.1459, "step": 4671 }, { "epoch": 0.3, "learning_rate": 0.0001642504752608744, "loss": 0.1601, "step": 4672 }, { "epoch": 0.3, "learning_rate": 0.00016423456070159816, "loss": 0.1695, "step": 4673 }, { "epoch": 0.3, "learning_rate": 0.00016421864337215803, "loss": 0.1385, "step": 4674 }, { "epoch": 0.3, "learning_rate": 0.00016420272327324038, "loss": 0.123, "step": 4675 }, { "epoch": 0.3, "learning_rate": 0.00016418680040553181, "loss": 0.1454, "step": 4676 }, { "epoch": 0.3, "learning_rate": 0.000164170874769719, "loss": 0.144, "step": 4677 }, { "epoch": 0.3, "learning_rate": 0.00016415494636648877, "loss": 0.1568, "step": 4678 }, { "epoch": 0.3, "learning_rate": 0.00016413901519652803, "loss": 0.1501, "step": 4679 }, { "epoch": 0.3, "learning_rate": 0.00016412308126052385, "loss": 0.1674, "step": 4680 }, { "epoch": 0.3, "learning_rate": 0.00016410714455916337, "loss": 0.153, "step": 4681 }, { "epoch": 0.3, "learning_rate": 0.00016409120509313385, "loss": 0.1914, "step": 4682 }, { "epoch": 0.3, "learning_rate": 0.00016407526286312272, "loss": 0.138, "step": 4683 }, { "epoch": 0.3, "learning_rate": 0.00016405931786981755, "loss": 0.1571, "step": 4684 }, { "epoch": 0.3, "learning_rate": 0.00016404337011390586, "loss": 0.1588, "step": 4685 }, { "epoch": 0.3, "learning_rate": 0.00016402741959607552, "loss": 0.1567, "step": 4686 }, { "epoch": 0.3, "learning_rate": 0.0001640114663170143, "loss": 0.173, "step": 4687 }, { "epoch": 0.3, "learning_rate": 0.0001639955102774103, "loss": 0.1631, "step": 4688 }, { "epoch": 0.3, "learning_rate": 0.00016397955147795163, "loss": 0.1758, "step": 4689 }, { "epoch": 0.3, "learning_rate": 0.00016396358991932646, "loss": 0.1699, "step": 4690 }, { "epoch": 0.3, "learning_rate": 0.00016394762560222322, "loss": 0.1518, "step": 4691 }, { "epoch": 0.3, "learning_rate": 0.00016393165852733027, "loss": 0.177, "step": 4692 }, { "epoch": 0.3, "learning_rate": 0.00016391568869533632, "loss": 0.1891, "step": 4693 }, { "epoch": 0.3, "learning_rate": 0.00016389971610693, "loss": 0.1367, "step": 4694 }, { "epoch": 0.3, "learning_rate": 0.0001638837407628002, "loss": 0.1924, "step": 4695 }, { "epoch": 0.3, "learning_rate": 0.00016386776266363583, "loss": 0.1516, "step": 4696 }, { "epoch": 0.3, "learning_rate": 0.00016385178181012598, "loss": 0.1553, "step": 4697 }, { "epoch": 0.3, "learning_rate": 0.00016383579820295982, "loss": 0.1406, "step": 4698 }, { "epoch": 0.3, "learning_rate": 0.00016381981184282663, "loss": 0.1539, "step": 4699 }, { "epoch": 0.3, "learning_rate": 0.00016380382273041593, "loss": 0.1799, "step": 4700 }, { "epoch": 0.3, "learning_rate": 0.00016378783086641713, "loss": 0.1548, "step": 4701 }, { "epoch": 0.3, "learning_rate": 0.00016377183625152002, "loss": 0.1486, "step": 4702 }, { "epoch": 0.3, "learning_rate": 0.00016375583888641427, "loss": 0.1775, "step": 4703 }, { "epoch": 0.3, "learning_rate": 0.00016373983877178986, "loss": 0.1477, "step": 4704 }, { "epoch": 0.3, "learning_rate": 0.00016372383590833675, "loss": 0.1343, "step": 4705 }, { "epoch": 0.3, "learning_rate": 0.00016370783029674512, "loss": 0.1642, "step": 4706 }, { "epoch": 0.3, "learning_rate": 0.00016369182193770517, "loss": 0.1711, "step": 4707 }, { "epoch": 0.3, "learning_rate": 0.00016367581083190735, "loss": 0.1259, "step": 4708 }, { "epoch": 0.3, "learning_rate": 0.0001636597969800421, "loss": 0.1617, "step": 4709 }, { "epoch": 0.3, "learning_rate": 0.0001636437803828, "loss": 0.1509, "step": 4710 }, { "epoch": 0.3, "learning_rate": 0.00016362776104087185, "loss": 0.1852, "step": 4711 }, { "epoch": 0.3, "learning_rate": 0.00016361173895494845, "loss": 0.1535, "step": 4712 }, { "epoch": 0.3, "learning_rate": 0.0001635957141257208, "loss": 0.1504, "step": 4713 }, { "epoch": 0.3, "learning_rate": 0.00016357968655387992, "loss": 0.1396, "step": 4714 }, { "epoch": 0.3, "learning_rate": 0.00016356365624011705, "loss": 0.1633, "step": 4715 }, { "epoch": 0.3, "learning_rate": 0.00016354762318512354, "loss": 0.1541, "step": 4716 }, { "epoch": 0.3, "learning_rate": 0.00016353158738959076, "loss": 0.1695, "step": 4717 }, { "epoch": 0.3, "learning_rate": 0.00016351554885421035, "loss": 0.1616, "step": 4718 }, { "epoch": 0.3, "learning_rate": 0.0001634995075796739, "loss": 0.1748, "step": 4719 }, { "epoch": 0.3, "learning_rate": 0.00016348346356667324, "loss": 0.1446, "step": 4720 }, { "epoch": 0.3, "learning_rate": 0.00016346741681590028, "loss": 0.1688, "step": 4721 }, { "epoch": 0.3, "learning_rate": 0.00016345136732804705, "loss": 0.1502, "step": 4722 }, { "epoch": 0.3, "learning_rate": 0.00016343531510380567, "loss": 0.1636, "step": 4723 }, { "epoch": 0.3, "learning_rate": 0.00016341926014386846, "loss": 0.1446, "step": 4724 }, { "epoch": 0.3, "learning_rate": 0.00016340320244892777, "loss": 0.1686, "step": 4725 }, { "epoch": 0.3, "learning_rate": 0.0001633871420196761, "loss": 0.1633, "step": 4726 }, { "epoch": 0.3, "learning_rate": 0.00016337107885680606, "loss": 0.1582, "step": 4727 }, { "epoch": 0.3, "learning_rate": 0.00016335501296101037, "loss": 0.1475, "step": 4728 }, { "epoch": 0.3, "learning_rate": 0.00016333894433298194, "loss": 0.1735, "step": 4729 }, { "epoch": 0.3, "learning_rate": 0.0001633228729734137, "loss": 0.173, "step": 4730 }, { "epoch": 0.3, "learning_rate": 0.00016330679888299873, "loss": 0.1564, "step": 4731 }, { "epoch": 0.3, "learning_rate": 0.0001632907220624303, "loss": 0.1618, "step": 4732 }, { "epoch": 0.3, "learning_rate": 0.00016327464251240166, "loss": 0.1481, "step": 4733 }, { "epoch": 0.3, "learning_rate": 0.00016325856023360628, "loss": 0.1549, "step": 4734 }, { "epoch": 0.3, "learning_rate": 0.00016324247522673773, "loss": 0.1518, "step": 4735 }, { "epoch": 0.3, "learning_rate": 0.00016322638749248969, "loss": 0.124, "step": 4736 }, { "epoch": 0.3, "learning_rate": 0.00016321029703155596, "loss": 0.1601, "step": 4737 }, { "epoch": 0.3, "learning_rate": 0.0001631942038446304, "loss": 0.1644, "step": 4738 }, { "epoch": 0.3, "learning_rate": 0.00016317810793240711, "loss": 0.1538, "step": 4739 }, { "epoch": 0.3, "learning_rate": 0.0001631620092955802, "loss": 0.1464, "step": 4740 }, { "epoch": 0.3, "learning_rate": 0.000163145907934844, "loss": 0.14, "step": 4741 }, { "epoch": 0.3, "learning_rate": 0.00016312980385089278, "loss": 0.1559, "step": 4742 }, { "epoch": 0.3, "learning_rate": 0.0001631136970444211, "loss": 0.1546, "step": 4743 }, { "epoch": 0.3, "learning_rate": 0.00016309758751612357, "loss": 0.1736, "step": 4744 }, { "epoch": 0.3, "learning_rate": 0.00016308147526669498, "loss": 0.1541, "step": 4745 }, { "epoch": 0.3, "learning_rate": 0.00016306536029683013, "loss": 0.1428, "step": 4746 }, { "epoch": 0.3, "learning_rate": 0.00016304924260722396, "loss": 0.1696, "step": 4747 }, { "epoch": 0.3, "learning_rate": 0.00016303312219857166, "loss": 0.1524, "step": 4748 }, { "epoch": 0.3, "learning_rate": 0.0001630169990715683, "loss": 0.1579, "step": 4749 }, { "epoch": 0.3, "learning_rate": 0.00016300087322690932, "loss": 0.1634, "step": 4750 }, { "epoch": 0.3, "learning_rate": 0.00016298474466529007, "loss": 0.1499, "step": 4751 }, { "epoch": 0.3, "learning_rate": 0.00016296861338740618, "loss": 0.1635, "step": 4752 }, { "epoch": 0.3, "learning_rate": 0.00016295247939395328, "loss": 0.1327, "step": 4753 }, { "epoch": 0.3, "learning_rate": 0.0001629363426856272, "loss": 0.1496, "step": 4754 }, { "epoch": 0.3, "learning_rate": 0.00016292020326312377, "loss": 0.1329, "step": 4755 }, { "epoch": 0.3, "learning_rate": 0.00016290406112713906, "loss": 0.172, "step": 4756 }, { "epoch": 0.31, "learning_rate": 0.00016288791627836927, "loss": 0.1548, "step": 4757 }, { "epoch": 0.31, "learning_rate": 0.00016287176871751058, "loss": 0.1683, "step": 4758 }, { "epoch": 0.31, "learning_rate": 0.00016285561844525938, "loss": 0.1468, "step": 4759 }, { "epoch": 0.31, "learning_rate": 0.00016283946546231218, "loss": 0.1422, "step": 4760 }, { "epoch": 0.31, "learning_rate": 0.00016282330976936558, "loss": 0.1657, "step": 4761 }, { "epoch": 0.31, "learning_rate": 0.00016280715136711633, "loss": 0.1214, "step": 4762 }, { "epoch": 0.31, "learning_rate": 0.00016279099025626122, "loss": 0.1613, "step": 4763 }, { "epoch": 0.31, "learning_rate": 0.00016277482643749726, "loss": 0.1199, "step": 4764 }, { "epoch": 0.31, "learning_rate": 0.0001627586599115215, "loss": 0.1975, "step": 4765 }, { "epoch": 0.31, "learning_rate": 0.0001627424906790312, "loss": 0.1734, "step": 4766 }, { "epoch": 0.31, "learning_rate": 0.00016272631874072354, "loss": 0.1447, "step": 4767 }, { "epoch": 0.31, "learning_rate": 0.00016271014409729605, "loss": 0.1599, "step": 4768 }, { "epoch": 0.31, "learning_rate": 0.00016269396674944623, "loss": 0.1615, "step": 4769 }, { "epoch": 0.31, "learning_rate": 0.00016267778669787174, "loss": 0.1324, "step": 4770 }, { "epoch": 0.31, "learning_rate": 0.0001626616039432704, "loss": 0.1389, "step": 4771 }, { "epoch": 0.31, "learning_rate": 0.0001626454184863401, "loss": 0.1414, "step": 4772 }, { "epoch": 0.31, "learning_rate": 0.0001626292303277788, "loss": 0.1553, "step": 4773 }, { "epoch": 0.31, "learning_rate": 0.00016261303946828463, "loss": 0.1661, "step": 4774 }, { "epoch": 0.31, "learning_rate": 0.0001625968459085559, "loss": 0.1473, "step": 4775 }, { "epoch": 0.31, "learning_rate": 0.00016258064964929088, "loss": 0.1646, "step": 4776 }, { "epoch": 0.31, "learning_rate": 0.00016256445069118814, "loss": 0.1589, "step": 4777 }, { "epoch": 0.31, "learning_rate": 0.00016254824903494616, "loss": 0.1454, "step": 4778 }, { "epoch": 0.31, "learning_rate": 0.00016253204468126374, "loss": 0.154, "step": 4779 }, { "epoch": 0.31, "learning_rate": 0.00016251583763083967, "loss": 0.1574, "step": 4780 }, { "epoch": 0.31, "learning_rate": 0.00016249962788437294, "loss": 0.2019, "step": 4781 }, { "epoch": 0.31, "learning_rate": 0.0001624834154425625, "loss": 0.151, "step": 4782 }, { "epoch": 0.31, "learning_rate": 0.00016246720030610763, "loss": 0.1843, "step": 4783 }, { "epoch": 0.31, "learning_rate": 0.00016245098247570758, "loss": 0.149, "step": 4784 }, { "epoch": 0.31, "learning_rate": 0.00016243476195206173, "loss": 0.1444, "step": 4785 }, { "epoch": 0.31, "learning_rate": 0.00016241853873586967, "loss": 0.1759, "step": 4786 }, { "epoch": 0.31, "learning_rate": 0.00016240231282783096, "loss": 0.1487, "step": 4787 }, { "epoch": 0.31, "learning_rate": 0.00016238608422864542, "loss": 0.1639, "step": 4788 }, { "epoch": 0.31, "learning_rate": 0.00016236985293901284, "loss": 0.1772, "step": 4789 }, { "epoch": 0.31, "learning_rate": 0.00016235361895963329, "loss": 0.1349, "step": 4790 }, { "epoch": 0.31, "learning_rate": 0.00016233738229120683, "loss": 0.1514, "step": 4791 }, { "epoch": 0.31, "learning_rate": 0.0001623211429344337, "loss": 0.1992, "step": 4792 }, { "epoch": 0.31, "learning_rate": 0.00016230490089001422, "loss": 0.1716, "step": 4793 }, { "epoch": 0.31, "learning_rate": 0.00016228865615864883, "loss": 0.1724, "step": 4794 }, { "epoch": 0.31, "learning_rate": 0.00016227240874103815, "loss": 0.1466, "step": 4795 }, { "epoch": 0.31, "learning_rate": 0.00016225615863788278, "loss": 0.1597, "step": 4796 }, { "epoch": 0.31, "learning_rate": 0.00016223990584988358, "loss": 0.1403, "step": 4797 }, { "epoch": 0.31, "learning_rate": 0.00016222365037774143, "loss": 0.1598, "step": 4798 }, { "epoch": 0.31, "learning_rate": 0.00016220739222215738, "loss": 0.1759, "step": 4799 }, { "epoch": 0.31, "learning_rate": 0.00016219113138383258, "loss": 0.2096, "step": 4800 }, { "epoch": 0.31, "learning_rate": 0.00016217486786346827, "loss": 0.149, "step": 4801 }, { "epoch": 0.31, "learning_rate": 0.00016215860166176585, "loss": 0.1794, "step": 4802 }, { "epoch": 0.31, "learning_rate": 0.00016214233277942676, "loss": 0.1781, "step": 4803 }, { "epoch": 0.31, "learning_rate": 0.00016212606121715267, "loss": 0.1702, "step": 4804 }, { "epoch": 0.31, "learning_rate": 0.00016210978697564528, "loss": 0.1658, "step": 4805 }, { "epoch": 0.31, "learning_rate": 0.00016209351005560643, "loss": 0.1561, "step": 4806 }, { "epoch": 0.31, "learning_rate": 0.00016207723045773807, "loss": 0.1759, "step": 4807 }, { "epoch": 0.31, "learning_rate": 0.00016206094818274229, "loss": 0.1529, "step": 4808 }, { "epoch": 0.31, "learning_rate": 0.00016204466323132123, "loss": 0.1487, "step": 4809 }, { "epoch": 0.31, "learning_rate": 0.00016202837560417725, "loss": 0.1387, "step": 4810 }, { "epoch": 0.31, "learning_rate": 0.0001620120853020127, "loss": 0.1499, "step": 4811 }, { "epoch": 0.31, "learning_rate": 0.0001619957923255302, "loss": 0.1404, "step": 4812 }, { "epoch": 0.31, "learning_rate": 0.00016197949667543228, "loss": 0.1632, "step": 4813 }, { "epoch": 0.31, "learning_rate": 0.00016196319835242183, "loss": 0.1428, "step": 4814 }, { "epoch": 0.31, "learning_rate": 0.00016194689735720162, "loss": 0.1643, "step": 4815 }, { "epoch": 0.31, "learning_rate": 0.0001619305936904747, "loss": 0.1637, "step": 4816 }, { "epoch": 0.31, "learning_rate": 0.00016191428735294413, "loss": 0.1641, "step": 4817 }, { "epoch": 0.31, "learning_rate": 0.00016189797834531322, "loss": 0.1639, "step": 4818 }, { "epoch": 0.31, "learning_rate": 0.00016188166666828522, "loss": 0.1351, "step": 4819 }, { "epoch": 0.31, "learning_rate": 0.00016186535232256364, "loss": 0.1591, "step": 4820 }, { "epoch": 0.31, "learning_rate": 0.00016184903530885202, "loss": 0.1666, "step": 4821 }, { "epoch": 0.31, "learning_rate": 0.00016183271562785406, "loss": 0.1431, "step": 4822 }, { "epoch": 0.31, "learning_rate": 0.00016181639328027354, "loss": 0.161, "step": 4823 }, { "epoch": 0.31, "learning_rate": 0.00016180006826681438, "loss": 0.1498, "step": 4824 }, { "epoch": 0.31, "learning_rate": 0.00016178374058818063, "loss": 0.1393, "step": 4825 }, { "epoch": 0.31, "learning_rate": 0.0001617674102450764, "loss": 0.1903, "step": 4826 }, { "epoch": 0.31, "learning_rate": 0.00016175107723820595, "loss": 0.1687, "step": 4827 }, { "epoch": 0.31, "learning_rate": 0.0001617347415682737, "loss": 0.1626, "step": 4828 }, { "epoch": 0.31, "learning_rate": 0.00016171840323598409, "loss": 0.1483, "step": 4829 }, { "epoch": 0.31, "learning_rate": 0.0001617020622420417, "loss": 0.1418, "step": 4830 }, { "epoch": 0.31, "learning_rate": 0.00016168571858715132, "loss": 0.1643, "step": 4831 }, { "epoch": 0.31, "learning_rate": 0.00016166937227201776, "loss": 0.1577, "step": 4832 }, { "epoch": 0.31, "learning_rate": 0.00016165302329734595, "loss": 0.1558, "step": 4833 }, { "epoch": 0.31, "learning_rate": 0.00016163667166384094, "loss": 0.1553, "step": 4834 }, { "epoch": 0.31, "learning_rate": 0.00016162031737220796, "loss": 0.1469, "step": 4835 }, { "epoch": 0.31, "learning_rate": 0.00016160396042315224, "loss": 0.1524, "step": 4836 }, { "epoch": 0.31, "learning_rate": 0.0001615876008173792, "loss": 0.1296, "step": 4837 }, { "epoch": 0.31, "learning_rate": 0.0001615712385555944, "loss": 0.1572, "step": 4838 }, { "epoch": 0.31, "learning_rate": 0.00016155487363850342, "loss": 0.1594, "step": 4839 }, { "epoch": 0.31, "learning_rate": 0.00016153850606681202, "loss": 0.1476, "step": 4840 }, { "epoch": 0.31, "learning_rate": 0.00016152213584122614, "loss": 0.1525, "step": 4841 }, { "epoch": 0.31, "learning_rate": 0.00016150576296245165, "loss": 0.1317, "step": 4842 }, { "epoch": 0.31, "learning_rate": 0.0001614893874311947, "loss": 0.1592, "step": 4843 }, { "epoch": 0.31, "learning_rate": 0.00016147300924816146, "loss": 0.1391, "step": 4844 }, { "epoch": 0.31, "learning_rate": 0.00016145662841405832, "loss": 0.1617, "step": 4845 }, { "epoch": 0.31, "learning_rate": 0.00016144024492959166, "loss": 0.1558, "step": 4846 }, { "epoch": 0.31, "learning_rate": 0.00016142385879546804, "loss": 0.1499, "step": 4847 }, { "epoch": 0.31, "learning_rate": 0.00016140747001239415, "loss": 0.1481, "step": 4848 }, { "epoch": 0.31, "learning_rate": 0.00016139107858107675, "loss": 0.155, "step": 4849 }, { "epoch": 0.31, "learning_rate": 0.0001613746845022227, "loss": 0.1563, "step": 4850 }, { "epoch": 0.31, "learning_rate": 0.00016135828777653902, "loss": 0.1515, "step": 4851 }, { "epoch": 0.31, "learning_rate": 0.0001613418884047329, "loss": 0.1393, "step": 4852 }, { "epoch": 0.31, "learning_rate": 0.00016132548638751152, "loss": 0.1507, "step": 4853 }, { "epoch": 0.31, "learning_rate": 0.0001613090817255822, "loss": 0.1574, "step": 4854 }, { "epoch": 0.31, "learning_rate": 0.00016129267441965248, "loss": 0.1535, "step": 4855 }, { "epoch": 0.31, "learning_rate": 0.00016127626447042984, "loss": 0.1766, "step": 4856 }, { "epoch": 0.31, "learning_rate": 0.00016125985187862208, "loss": 0.1486, "step": 4857 }, { "epoch": 0.31, "learning_rate": 0.0001612434366449369, "loss": 0.1897, "step": 4858 }, { "epoch": 0.31, "learning_rate": 0.0001612270187700823, "loss": 0.1527, "step": 4859 }, { "epoch": 0.31, "learning_rate": 0.0001612105982547663, "loss": 0.1525, "step": 4860 }, { "epoch": 0.31, "learning_rate": 0.000161194175099697, "loss": 0.1739, "step": 4861 }, { "epoch": 0.31, "learning_rate": 0.00016117774930558273, "loss": 0.1496, "step": 4862 }, { "epoch": 0.31, "learning_rate": 0.00016116132087313177, "loss": 0.1775, "step": 4863 }, { "epoch": 0.31, "learning_rate": 0.00016114488980305272, "loss": 0.1639, "step": 4864 }, { "epoch": 0.31, "learning_rate": 0.00016112845609605412, "loss": 0.168, "step": 4865 }, { "epoch": 0.31, "learning_rate": 0.00016111201975284466, "loss": 0.1839, "step": 4866 }, { "epoch": 0.31, "learning_rate": 0.00016109558077413323, "loss": 0.1304, "step": 4867 }, { "epoch": 0.31, "learning_rate": 0.00016107913916062875, "loss": 0.1414, "step": 4868 }, { "epoch": 0.31, "learning_rate": 0.00016106269491304026, "loss": 0.1591, "step": 4869 }, { "epoch": 0.31, "learning_rate": 0.00016104624803207695, "loss": 0.1291, "step": 4870 }, { "epoch": 0.31, "learning_rate": 0.0001610297985184481, "loss": 0.151, "step": 4871 }, { "epoch": 0.31, "learning_rate": 0.0001610133463728631, "loss": 0.1657, "step": 4872 }, { "epoch": 0.31, "learning_rate": 0.00016099689159603146, "loss": 0.1536, "step": 4873 }, { "epoch": 0.31, "learning_rate": 0.00016098043418866283, "loss": 0.1634, "step": 4874 }, { "epoch": 0.31, "learning_rate": 0.00016096397415146693, "loss": 0.1351, "step": 4875 }, { "epoch": 0.31, "learning_rate": 0.0001609475114851536, "loss": 0.1711, "step": 4876 }, { "epoch": 0.31, "learning_rate": 0.00016093104619043283, "loss": 0.1492, "step": 4877 }, { "epoch": 0.31, "learning_rate": 0.00016091457826801465, "loss": 0.1573, "step": 4878 }, { "epoch": 0.31, "learning_rate": 0.00016089810771860932, "loss": 0.1438, "step": 4879 }, { "epoch": 0.31, "learning_rate": 0.0001608816345429271, "loss": 0.1688, "step": 4880 }, { "epoch": 0.31, "learning_rate": 0.0001608651587416784, "loss": 0.1622, "step": 4881 }, { "epoch": 0.31, "learning_rate": 0.00016084868031557384, "loss": 0.143, "step": 4882 }, { "epoch": 0.31, "learning_rate": 0.00016083219926532394, "loss": 0.1544, "step": 4883 }, { "epoch": 0.31, "learning_rate": 0.00016081571559163952, "loss": 0.1465, "step": 4884 }, { "epoch": 0.31, "learning_rate": 0.00016079922929523145, "loss": 0.1534, "step": 4885 }, { "epoch": 0.31, "learning_rate": 0.00016078274037681072, "loss": 0.1529, "step": 4886 }, { "epoch": 0.31, "learning_rate": 0.0001607662488370884, "loss": 0.1447, "step": 4887 }, { "epoch": 0.31, "learning_rate": 0.0001607497546767757, "loss": 0.131, "step": 4888 }, { "epoch": 0.31, "learning_rate": 0.000160733257896584, "loss": 0.1648, "step": 4889 }, { "epoch": 0.31, "learning_rate": 0.00016071675849722468, "loss": 0.1684, "step": 4890 }, { "epoch": 0.31, "learning_rate": 0.0001607002564794093, "loss": 0.1257, "step": 4891 }, { "epoch": 0.31, "learning_rate": 0.0001606837518438495, "loss": 0.1586, "step": 4892 }, { "epoch": 0.31, "learning_rate": 0.0001606672445912571, "loss": 0.151, "step": 4893 }, { "epoch": 0.31, "learning_rate": 0.00016065073472234395, "loss": 0.1428, "step": 4894 }, { "epoch": 0.31, "learning_rate": 0.0001606342222378221, "loss": 0.152, "step": 4895 }, { "epoch": 0.31, "learning_rate": 0.00016061770713840363, "loss": 0.1687, "step": 4896 }, { "epoch": 0.31, "learning_rate": 0.00016060118942480076, "loss": 0.1376, "step": 4897 }, { "epoch": 0.31, "learning_rate": 0.00016058466909772584, "loss": 0.1519, "step": 4898 }, { "epoch": 0.31, "learning_rate": 0.0001605681461578913, "loss": 0.1577, "step": 4899 }, { "epoch": 0.31, "learning_rate": 0.0001605516206060097, "loss": 0.1643, "step": 4900 }, { "epoch": 0.31, "learning_rate": 0.00016053509244279382, "loss": 0.1491, "step": 4901 }, { "epoch": 0.31, "learning_rate": 0.0001605185616689563, "loss": 0.1683, "step": 4902 }, { "epoch": 0.31, "learning_rate": 0.00016050202828521014, "loss": 0.157, "step": 4903 }, { "epoch": 0.31, "learning_rate": 0.00016048549229226832, "loss": 0.1425, "step": 4904 }, { "epoch": 0.31, "learning_rate": 0.000160468953690844, "loss": 0.1474, "step": 4905 }, { "epoch": 0.31, "learning_rate": 0.00016045241248165035, "loss": 0.1563, "step": 4906 }, { "epoch": 0.31, "learning_rate": 0.0001604358686654008, "loss": 0.1555, "step": 4907 }, { "epoch": 0.31, "learning_rate": 0.00016041932224280877, "loss": 0.1517, "step": 4908 }, { "epoch": 0.31, "learning_rate": 0.00016040277321458784, "loss": 0.1518, "step": 4909 }, { "epoch": 0.31, "learning_rate": 0.0001603862215814517, "loss": 0.1543, "step": 4910 }, { "epoch": 0.31, "learning_rate": 0.00016036966734411418, "loss": 0.1482, "step": 4911 }, { "epoch": 0.31, "learning_rate": 0.00016035311050328915, "loss": 0.1472, "step": 4912 }, { "epoch": 0.32, "learning_rate": 0.00016033655105969065, "loss": 0.1775, "step": 4913 }, { "epoch": 0.32, "learning_rate": 0.00016031998901403286, "loss": 0.1442, "step": 4914 }, { "epoch": 0.32, "learning_rate": 0.00016030342436702997, "loss": 0.1548, "step": 4915 }, { "epoch": 0.32, "learning_rate": 0.00016028685711939636, "loss": 0.1542, "step": 4916 }, { "epoch": 0.32, "learning_rate": 0.00016027028727184654, "loss": 0.1348, "step": 4917 }, { "epoch": 0.32, "learning_rate": 0.00016025371482509507, "loss": 0.156, "step": 4918 }, { "epoch": 0.32, "learning_rate": 0.00016023713977985664, "loss": 0.1591, "step": 4919 }, { "epoch": 0.32, "learning_rate": 0.00016022056213684609, "loss": 0.1546, "step": 4920 }, { "epoch": 0.32, "learning_rate": 0.0001602039818967783, "loss": 0.1746, "step": 4921 }, { "epoch": 0.32, "learning_rate": 0.00016018739906036836, "loss": 0.1479, "step": 4922 }, { "epoch": 0.32, "learning_rate": 0.0001601708136283314, "loss": 0.1475, "step": 4923 }, { "epoch": 0.32, "learning_rate": 0.00016015422560138263, "loss": 0.1445, "step": 4924 }, { "epoch": 0.32, "learning_rate": 0.00016013763498023747, "loss": 0.1535, "step": 4925 }, { "epoch": 0.32, "learning_rate": 0.0001601210417656114, "loss": 0.136, "step": 4926 }, { "epoch": 0.32, "learning_rate": 0.00016010444595821998, "loss": 0.1529, "step": 4927 }, { "epoch": 0.32, "learning_rate": 0.000160087847558779, "loss": 0.1451, "step": 4928 }, { "epoch": 0.32, "learning_rate": 0.00016007124656800418, "loss": 0.1631, "step": 4929 }, { "epoch": 0.32, "learning_rate": 0.00016005464298661151, "loss": 0.136, "step": 4930 }, { "epoch": 0.32, "learning_rate": 0.000160038036815317, "loss": 0.1274, "step": 4931 }, { "epoch": 0.32, "learning_rate": 0.00016002142805483685, "loss": 0.135, "step": 4932 }, { "epoch": 0.32, "learning_rate": 0.00016000481670588727, "loss": 0.1641, "step": 4933 }, { "epoch": 0.32, "learning_rate": 0.00015998820276918465, "loss": 0.1562, "step": 4934 }, { "epoch": 0.32, "learning_rate": 0.00015997158624544552, "loss": 0.1479, "step": 4935 }, { "epoch": 0.32, "learning_rate": 0.0001599549671353864, "loss": 0.1517, "step": 4936 }, { "epoch": 0.32, "learning_rate": 0.0001599383454397241, "loss": 0.1603, "step": 4937 }, { "epoch": 0.32, "learning_rate": 0.0001599217211591754, "loss": 0.1716, "step": 4938 }, { "epoch": 0.32, "learning_rate": 0.00015990509429445718, "loss": 0.1495, "step": 4939 }, { "epoch": 0.32, "learning_rate": 0.00015988846484628656, "loss": 0.1488, "step": 4940 }, { "epoch": 0.32, "learning_rate": 0.00015987183281538066, "loss": 0.1504, "step": 4941 }, { "epoch": 0.32, "learning_rate": 0.00015985519820245681, "loss": 0.1586, "step": 4942 }, { "epoch": 0.32, "learning_rate": 0.00015983856100823228, "loss": 0.161, "step": 4943 }, { "epoch": 0.32, "learning_rate": 0.00015982192123342468, "loss": 0.1575, "step": 4944 }, { "epoch": 0.32, "learning_rate": 0.0001598052788787515, "loss": 0.1452, "step": 4945 }, { "epoch": 0.32, "learning_rate": 0.00015978863394493055, "loss": 0.1591, "step": 4946 }, { "epoch": 0.32, "learning_rate": 0.0001597719864326796, "loss": 0.1675, "step": 4947 }, { "epoch": 0.32, "learning_rate": 0.00015975533634271662, "loss": 0.145, "step": 4948 }, { "epoch": 0.32, "learning_rate": 0.00015973868367575962, "loss": 0.179, "step": 4949 }, { "epoch": 0.32, "learning_rate": 0.00015972202843252679, "loss": 0.1549, "step": 4950 }, { "epoch": 0.32, "learning_rate": 0.00015970537061373637, "loss": 0.1506, "step": 4951 }, { "epoch": 0.32, "learning_rate": 0.00015968871022010677, "loss": 0.1691, "step": 4952 }, { "epoch": 0.32, "learning_rate": 0.00015967204725235649, "loss": 0.1605, "step": 4953 }, { "epoch": 0.32, "learning_rate": 0.0001596553817112041, "loss": 0.1604, "step": 4954 }, { "epoch": 0.32, "learning_rate": 0.00015963871359736832, "loss": 0.1642, "step": 4955 }, { "epoch": 0.32, "learning_rate": 0.000159622042911568, "loss": 0.1486, "step": 4956 }, { "epoch": 0.32, "learning_rate": 0.00015960536965452205, "loss": 0.1431, "step": 4957 }, { "epoch": 0.32, "learning_rate": 0.00015958869382694954, "loss": 0.1689, "step": 4958 }, { "epoch": 0.32, "learning_rate": 0.0001595720154295696, "loss": 0.1263, "step": 4959 }, { "epoch": 0.32, "learning_rate": 0.0001595553344631015, "loss": 0.1658, "step": 4960 }, { "epoch": 0.32, "learning_rate": 0.0001595386509282647, "loss": 0.1686, "step": 4961 }, { "epoch": 0.32, "learning_rate": 0.00015952196482577852, "loss": 0.171, "step": 4962 }, { "epoch": 0.32, "learning_rate": 0.00015950527615636272, "loss": 0.1647, "step": 4963 }, { "epoch": 0.32, "learning_rate": 0.00015948858492073696, "loss": 0.1674, "step": 4964 }, { "epoch": 0.32, "learning_rate": 0.00015947189111962105, "loss": 0.1557, "step": 4965 }, { "epoch": 0.32, "learning_rate": 0.00015945519475373492, "loss": 0.153, "step": 4966 }, { "epoch": 0.32, "learning_rate": 0.0001594384958237986, "loss": 0.1366, "step": 4967 }, { "epoch": 0.32, "learning_rate": 0.00015942179433053232, "loss": 0.1306, "step": 4968 }, { "epoch": 0.32, "learning_rate": 0.00015940509027465626, "loss": 0.1517, "step": 4969 }, { "epoch": 0.32, "learning_rate": 0.00015938838365689084, "loss": 0.1491, "step": 4970 }, { "epoch": 0.32, "learning_rate": 0.00015937167447795653, "loss": 0.1303, "step": 4971 }, { "epoch": 0.32, "learning_rate": 0.0001593549627385739, "loss": 0.1401, "step": 4972 }, { "epoch": 0.32, "learning_rate": 0.0001593382484394637, "loss": 0.1575, "step": 4973 }, { "epoch": 0.32, "learning_rate": 0.00015932153158134673, "loss": 0.1697, "step": 4974 }, { "epoch": 0.32, "learning_rate": 0.00015930481216494392, "loss": 0.1409, "step": 4975 }, { "epoch": 0.32, "learning_rate": 0.0001592880901909763, "loss": 0.1556, "step": 4976 }, { "epoch": 0.32, "learning_rate": 0.00015927136566016504, "loss": 0.1543, "step": 4977 }, { "epoch": 0.32, "learning_rate": 0.00015925463857323138, "loss": 0.1681, "step": 4978 }, { "epoch": 0.32, "learning_rate": 0.00015923790893089667, "loss": 0.1487, "step": 4979 }, { "epoch": 0.32, "learning_rate": 0.00015922117673388244, "loss": 0.1431, "step": 4980 }, { "epoch": 0.32, "learning_rate": 0.00015920444198291023, "loss": 0.1592, "step": 4981 }, { "epoch": 0.32, "learning_rate": 0.0001591877046787017, "loss": 0.1519, "step": 4982 }, { "epoch": 0.32, "learning_rate": 0.0001591709648219788, "loss": 0.1351, "step": 4983 }, { "epoch": 0.32, "learning_rate": 0.00015915422241346333, "loss": 0.1674, "step": 4984 }, { "epoch": 0.32, "learning_rate": 0.00015913747745387736, "loss": 0.1531, "step": 4985 }, { "epoch": 0.32, "learning_rate": 0.000159120729943943, "loss": 0.145, "step": 4986 }, { "epoch": 0.32, "learning_rate": 0.00015910397988438252, "loss": 0.1495, "step": 4987 }, { "epoch": 0.32, "learning_rate": 0.0001590872272759183, "loss": 0.1702, "step": 4988 }, { "epoch": 0.32, "learning_rate": 0.00015907047211927277, "loss": 0.1352, "step": 4989 }, { "epoch": 0.32, "learning_rate": 0.00015905371441516855, "loss": 0.1572, "step": 4990 }, { "epoch": 0.32, "learning_rate": 0.0001590369541643283, "loss": 0.145, "step": 4991 }, { "epoch": 0.32, "learning_rate": 0.0001590201913674748, "loss": 0.1831, "step": 4992 }, { "epoch": 0.32, "learning_rate": 0.000159003426025331, "loss": 0.1395, "step": 4993 }, { "epoch": 0.32, "learning_rate": 0.0001589866581386199, "loss": 0.1582, "step": 4994 }, { "epoch": 0.32, "learning_rate": 0.00015896988770806462, "loss": 0.1482, "step": 4995 }, { "epoch": 0.32, "learning_rate": 0.0001589531147343884, "loss": 0.172, "step": 4996 }, { "epoch": 0.32, "learning_rate": 0.0001589363392183146, "loss": 0.1567, "step": 4997 }, { "epoch": 0.32, "learning_rate": 0.00015891956116056668, "loss": 0.1536, "step": 4998 }, { "epoch": 0.32, "learning_rate": 0.00015890278056186817, "loss": 0.1902, "step": 4999 }, { "epoch": 0.32, "learning_rate": 0.00015888599742294282, "loss": 0.1387, "step": 5000 }, { "epoch": 0.32, "learning_rate": 0.0001588692117445143, "loss": 0.1692, "step": 5001 }, { "epoch": 0.32, "learning_rate": 0.0001588524235273066, "loss": 0.1653, "step": 5002 }, { "epoch": 0.32, "learning_rate": 0.0001588356327720437, "loss": 0.1612, "step": 5003 }, { "epoch": 0.32, "learning_rate": 0.00015881883947944974, "loss": 0.1672, "step": 5004 }, { "epoch": 0.32, "learning_rate": 0.0001588020436502489, "loss": 0.1647, "step": 5005 }, { "epoch": 0.32, "learning_rate": 0.00015878524528516552, "loss": 0.1505, "step": 5006 }, { "epoch": 0.32, "learning_rate": 0.00015876844438492404, "loss": 0.1567, "step": 5007 }, { "epoch": 0.32, "learning_rate": 0.000158751640950249, "loss": 0.1473, "step": 5008 }, { "epoch": 0.32, "learning_rate": 0.00015873483498186515, "loss": 0.1268, "step": 5009 }, { "epoch": 0.32, "learning_rate": 0.00015871802648049715, "loss": 0.1576, "step": 5010 }, { "epoch": 0.32, "learning_rate": 0.0001587012154468699, "loss": 0.1533, "step": 5011 }, { "epoch": 0.32, "learning_rate": 0.00015868440188170847, "loss": 0.1541, "step": 5012 }, { "epoch": 0.32, "learning_rate": 0.00015866758578573785, "loss": 0.1481, "step": 5013 }, { "epoch": 0.32, "learning_rate": 0.0001586507671596833, "loss": 0.1717, "step": 5014 }, { "epoch": 0.32, "learning_rate": 0.0001586339460042701, "loss": 0.1562, "step": 5015 }, { "epoch": 0.32, "learning_rate": 0.00015861712232022375, "loss": 0.1678, "step": 5016 }, { "epoch": 0.32, "learning_rate": 0.00015860029610826972, "loss": 0.1558, "step": 5017 }, { "epoch": 0.32, "learning_rate": 0.00015858346736913367, "loss": 0.1484, "step": 5018 }, { "epoch": 0.32, "learning_rate": 0.00015856663610354138, "loss": 0.1521, "step": 5019 }, { "epoch": 0.32, "learning_rate": 0.00015854980231221869, "loss": 0.1756, "step": 5020 }, { "epoch": 0.32, "learning_rate": 0.00015853296599589153, "loss": 0.14, "step": 5021 }, { "epoch": 0.32, "learning_rate": 0.00015851612715528604, "loss": 0.1491, "step": 5022 }, { "epoch": 0.32, "learning_rate": 0.00015849928579112835, "loss": 0.1562, "step": 5023 }, { "epoch": 0.32, "learning_rate": 0.0001584824419041448, "loss": 0.143, "step": 5024 }, { "epoch": 0.32, "learning_rate": 0.00015846559549506184, "loss": 0.1387, "step": 5025 }, { "epoch": 0.32, "learning_rate": 0.00015844874656460587, "loss": 0.1792, "step": 5026 }, { "epoch": 0.32, "learning_rate": 0.00015843189511350358, "loss": 0.1481, "step": 5027 }, { "epoch": 0.32, "learning_rate": 0.00015841504114248167, "loss": 0.1686, "step": 5028 }, { "epoch": 0.32, "learning_rate": 0.00015839818465226703, "loss": 0.1787, "step": 5029 }, { "epoch": 0.32, "learning_rate": 0.0001583813256435866, "loss": 0.1662, "step": 5030 }, { "epoch": 0.32, "learning_rate": 0.00015836446411716736, "loss": 0.1446, "step": 5031 }, { "epoch": 0.32, "learning_rate": 0.00015834760007373658, "loss": 0.1506, "step": 5032 }, { "epoch": 0.32, "learning_rate": 0.0001583307335140215, "loss": 0.1803, "step": 5033 }, { "epoch": 0.32, "learning_rate": 0.00015831386443874946, "loss": 0.1664, "step": 5034 }, { "epoch": 0.32, "learning_rate": 0.000158296992848648, "loss": 0.168, "step": 5035 }, { "epoch": 0.32, "learning_rate": 0.0001582801187444447, "loss": 0.1572, "step": 5036 }, { "epoch": 0.32, "learning_rate": 0.00015826324212686726, "loss": 0.1559, "step": 5037 }, { "epoch": 0.32, "learning_rate": 0.00015824636299664353, "loss": 0.1615, "step": 5038 }, { "epoch": 0.32, "learning_rate": 0.00015822948135450142, "loss": 0.1467, "step": 5039 }, { "epoch": 0.32, "learning_rate": 0.00015821259720116897, "loss": 0.172, "step": 5040 }, { "epoch": 0.32, "learning_rate": 0.00015819571053737427, "loss": 0.1789, "step": 5041 }, { "epoch": 0.32, "learning_rate": 0.00015817882136384562, "loss": 0.1459, "step": 5042 }, { "epoch": 0.32, "learning_rate": 0.00015816192968131138, "loss": 0.1714, "step": 5043 }, { "epoch": 0.32, "learning_rate": 0.00015814503549050002, "loss": 0.1704, "step": 5044 }, { "epoch": 0.32, "learning_rate": 0.00015812813879214005, "loss": 0.1366, "step": 5045 }, { "epoch": 0.32, "learning_rate": 0.00015811123958696025, "loss": 0.1701, "step": 5046 }, { "epoch": 0.32, "learning_rate": 0.00015809433787568935, "loss": 0.1722, "step": 5047 }, { "epoch": 0.32, "learning_rate": 0.00015807743365905627, "loss": 0.1369, "step": 5048 }, { "epoch": 0.32, "learning_rate": 0.00015806052693779, "loss": 0.1635, "step": 5049 }, { "epoch": 0.32, "learning_rate": 0.00015804361771261965, "loss": 0.1524, "step": 5050 }, { "epoch": 0.32, "learning_rate": 0.00015802670598427448, "loss": 0.1486, "step": 5051 }, { "epoch": 0.32, "learning_rate": 0.00015800979175348382, "loss": 0.1725, "step": 5052 }, { "epoch": 0.32, "learning_rate": 0.00015799287502097706, "loss": 0.1535, "step": 5053 }, { "epoch": 0.32, "learning_rate": 0.00015797595578748378, "loss": 0.1788, "step": 5054 }, { "epoch": 0.32, "learning_rate": 0.00015795903405373363, "loss": 0.1539, "step": 5055 }, { "epoch": 0.32, "learning_rate": 0.00015794210982045636, "loss": 0.1341, "step": 5056 }, { "epoch": 0.32, "learning_rate": 0.0001579251830883819, "loss": 0.1578, "step": 5057 }, { "epoch": 0.32, "learning_rate": 0.00015790825385824013, "loss": 0.1458, "step": 5058 }, { "epoch": 0.32, "learning_rate": 0.0001578913221307612, "loss": 0.1668, "step": 5059 }, { "epoch": 0.32, "learning_rate": 0.0001578743879066753, "loss": 0.1518, "step": 5060 }, { "epoch": 0.32, "learning_rate": 0.0001578574511867127, "loss": 0.1387, "step": 5061 }, { "epoch": 0.32, "learning_rate": 0.00015784051197160386, "loss": 0.1966, "step": 5062 }, { "epoch": 0.32, "learning_rate": 0.00015782357026207925, "loss": 0.1529, "step": 5063 }, { "epoch": 0.32, "learning_rate": 0.00015780662605886951, "loss": 0.1612, "step": 5064 }, { "epoch": 0.32, "learning_rate": 0.0001577896793627054, "loss": 0.1432, "step": 5065 }, { "epoch": 0.32, "learning_rate": 0.00015777273017431772, "loss": 0.1493, "step": 5066 }, { "epoch": 0.32, "learning_rate": 0.00015775577849443738, "loss": 0.1279, "step": 5067 }, { "epoch": 0.32, "learning_rate": 0.00015773882432379552, "loss": 0.1508, "step": 5068 }, { "epoch": 0.33, "learning_rate": 0.00015772186766312328, "loss": 0.1542, "step": 5069 }, { "epoch": 0.33, "learning_rate": 0.00015770490851315188, "loss": 0.1545, "step": 5070 }, { "epoch": 0.33, "learning_rate": 0.00015768794687461273, "loss": 0.1546, "step": 5071 }, { "epoch": 0.33, "learning_rate": 0.00015767098274823733, "loss": 0.1373, "step": 5072 }, { "epoch": 0.33, "learning_rate": 0.00015765401613475726, "loss": 0.1538, "step": 5073 }, { "epoch": 0.33, "learning_rate": 0.00015763704703490418, "loss": 0.154, "step": 5074 }, { "epoch": 0.33, "learning_rate": 0.00015762007544940993, "loss": 0.1499, "step": 5075 }, { "epoch": 0.33, "learning_rate": 0.00015760310137900643, "loss": 0.1628, "step": 5076 }, { "epoch": 0.33, "learning_rate": 0.0001575861248244257, "loss": 0.1734, "step": 5077 }, { "epoch": 0.33, "learning_rate": 0.00015756914578639983, "loss": 0.1587, "step": 5078 }, { "epoch": 0.33, "learning_rate": 0.00015755216426566112, "loss": 0.1605, "step": 5079 }, { "epoch": 0.33, "learning_rate": 0.00015753518026294185, "loss": 0.1532, "step": 5080 }, { "epoch": 0.33, "learning_rate": 0.00015751819377897447, "loss": 0.1649, "step": 5081 }, { "epoch": 0.33, "learning_rate": 0.00015750120481449156, "loss": 0.1323, "step": 5082 }, { "epoch": 0.33, "learning_rate": 0.00015748421337022576, "loss": 0.1665, "step": 5083 }, { "epoch": 0.33, "learning_rate": 0.00015746721944690986, "loss": 0.1742, "step": 5084 }, { "epoch": 0.33, "learning_rate": 0.00015745022304527675, "loss": 0.1641, "step": 5085 }, { "epoch": 0.33, "learning_rate": 0.00015743322416605937, "loss": 0.1633, "step": 5086 }, { "epoch": 0.33, "learning_rate": 0.00015741622280999087, "loss": 0.1541, "step": 5087 }, { "epoch": 0.33, "learning_rate": 0.00015739921897780436, "loss": 0.1633, "step": 5088 }, { "epoch": 0.33, "learning_rate": 0.0001573822126702332, "loss": 0.1437, "step": 5089 }, { "epoch": 0.33, "learning_rate": 0.0001573652038880108, "loss": 0.1396, "step": 5090 }, { "epoch": 0.33, "learning_rate": 0.00015734819263187063, "loss": 0.1632, "step": 5091 }, { "epoch": 0.33, "learning_rate": 0.0001573311789025464, "loss": 0.1456, "step": 5092 }, { "epoch": 0.33, "learning_rate": 0.00015731416270077178, "loss": 0.1578, "step": 5093 }, { "epoch": 0.33, "learning_rate": 0.00015729714402728062, "loss": 0.1482, "step": 5094 }, { "epoch": 0.33, "learning_rate": 0.00015728012288280683, "loss": 0.1638, "step": 5095 }, { "epoch": 0.33, "learning_rate": 0.00015726309926808452, "loss": 0.1428, "step": 5096 }, { "epoch": 0.33, "learning_rate": 0.0001572460731838478, "loss": 0.1347, "step": 5097 }, { "epoch": 0.33, "learning_rate": 0.00015722904463083098, "loss": 0.1448, "step": 5098 }, { "epoch": 0.33, "learning_rate": 0.00015721201360976834, "loss": 0.1514, "step": 5099 }, { "epoch": 0.33, "learning_rate": 0.00015719498012139446, "loss": 0.1581, "step": 5100 }, { "epoch": 0.33, "learning_rate": 0.00015717794416644386, "loss": 0.1639, "step": 5101 }, { "epoch": 0.33, "learning_rate": 0.00015716090574565124, "loss": 0.1426, "step": 5102 }, { "epoch": 0.33, "learning_rate": 0.0001571438648597514, "loss": 0.174, "step": 5103 }, { "epoch": 0.33, "learning_rate": 0.00015712682150947923, "loss": 0.1612, "step": 5104 }, { "epoch": 0.33, "learning_rate": 0.00015710977569556976, "loss": 0.1497, "step": 5105 }, { "epoch": 0.33, "learning_rate": 0.0001570927274187581, "loss": 0.1576, "step": 5106 }, { "epoch": 0.33, "learning_rate": 0.00015707567667977944, "loss": 0.1476, "step": 5107 }, { "epoch": 0.33, "learning_rate": 0.00015705862347936914, "loss": 0.1687, "step": 5108 }, { "epoch": 0.33, "learning_rate": 0.0001570415678182626, "loss": 0.1655, "step": 5109 }, { "epoch": 0.33, "learning_rate": 0.00015702450969719539, "loss": 0.1528, "step": 5110 }, { "epoch": 0.33, "learning_rate": 0.00015700744911690314, "loss": 0.1406, "step": 5111 }, { "epoch": 0.33, "learning_rate": 0.00015699038607812162, "loss": 0.1535, "step": 5112 }, { "epoch": 0.33, "learning_rate": 0.00015697332058158664, "loss": 0.1521, "step": 5113 }, { "epoch": 0.33, "learning_rate": 0.00015695625262803422, "loss": 0.17, "step": 5114 }, { "epoch": 0.33, "learning_rate": 0.0001569391822182004, "loss": 0.1598, "step": 5115 }, { "epoch": 0.33, "learning_rate": 0.0001569221093528213, "loss": 0.1631, "step": 5116 }, { "epoch": 0.33, "learning_rate": 0.00015690503403263328, "loss": 0.139, "step": 5117 }, { "epoch": 0.33, "learning_rate": 0.0001568879562583727, "loss": 0.1438, "step": 5118 }, { "epoch": 0.33, "learning_rate": 0.00015687087603077609, "loss": 0.1532, "step": 5119 }, { "epoch": 0.33, "learning_rate": 0.00015685379335057996, "loss": 0.1733, "step": 5120 }, { "epoch": 0.33, "learning_rate": 0.0001568367082185211, "loss": 0.1534, "step": 5121 }, { "epoch": 0.33, "learning_rate": 0.00015681962063533625, "loss": 0.1788, "step": 5122 }, { "epoch": 0.33, "learning_rate": 0.00015680253060176236, "loss": 0.1381, "step": 5123 }, { "epoch": 0.33, "learning_rate": 0.00015678543811853643, "loss": 0.1672, "step": 5124 }, { "epoch": 0.33, "learning_rate": 0.00015676834318639564, "loss": 0.1571, "step": 5125 }, { "epoch": 0.33, "learning_rate": 0.00015675124580607718, "loss": 0.1344, "step": 5126 }, { "epoch": 0.33, "learning_rate": 0.00015673414597831836, "loss": 0.1775, "step": 5127 }, { "epoch": 0.33, "learning_rate": 0.0001567170437038567, "loss": 0.1646, "step": 5128 }, { "epoch": 0.33, "learning_rate": 0.0001566999389834297, "loss": 0.1342, "step": 5129 }, { "epoch": 0.33, "learning_rate": 0.00015668283181777498, "loss": 0.1331, "step": 5130 }, { "epoch": 0.33, "learning_rate": 0.00015666572220763037, "loss": 0.1652, "step": 5131 }, { "epoch": 0.33, "learning_rate": 0.0001566486101537337, "loss": 0.178, "step": 5132 }, { "epoch": 0.33, "learning_rate": 0.00015663149565682292, "loss": 0.151, "step": 5133 }, { "epoch": 0.33, "learning_rate": 0.00015661437871763612, "loss": 0.1464, "step": 5134 }, { "epoch": 0.33, "learning_rate": 0.00015659725933691156, "loss": 0.1675, "step": 5135 }, { "epoch": 0.33, "learning_rate": 0.0001565801375153874, "loss": 0.1459, "step": 5136 }, { "epoch": 0.33, "learning_rate": 0.00015656301325380208, "loss": 0.1559, "step": 5137 }, { "epoch": 0.33, "learning_rate": 0.0001565458865528941, "loss": 0.1668, "step": 5138 }, { "epoch": 0.33, "learning_rate": 0.0001565287574134021, "loss": 0.1784, "step": 5139 }, { "epoch": 0.33, "learning_rate": 0.00015651162583606474, "loss": 0.1617, "step": 5140 }, { "epoch": 0.33, "learning_rate": 0.00015649449182162086, "loss": 0.1689, "step": 5141 }, { "epoch": 0.33, "learning_rate": 0.00015647735537080933, "loss": 0.1722, "step": 5142 }, { "epoch": 0.33, "learning_rate": 0.00015646021648436923, "loss": 0.166, "step": 5143 }, { "epoch": 0.33, "learning_rate": 0.00015644307516303964, "loss": 0.1427, "step": 5144 }, { "epoch": 0.33, "learning_rate": 0.00015642593140755982, "loss": 0.1476, "step": 5145 }, { "epoch": 0.33, "learning_rate": 0.00015640878521866914, "loss": 0.1375, "step": 5146 }, { "epoch": 0.33, "learning_rate": 0.00015639163659710693, "loss": 0.1415, "step": 5147 }, { "epoch": 0.33, "learning_rate": 0.00015637448554361286, "loss": 0.1292, "step": 5148 }, { "epoch": 0.33, "learning_rate": 0.00015635733205892653, "loss": 0.1522, "step": 5149 }, { "epoch": 0.33, "learning_rate": 0.00015634017614378772, "loss": 0.1584, "step": 5150 }, { "epoch": 0.33, "learning_rate": 0.00015632301779893622, "loss": 0.1699, "step": 5151 }, { "epoch": 0.33, "learning_rate": 0.0001563058570251121, "loss": 0.1493, "step": 5152 }, { "epoch": 0.33, "learning_rate": 0.00015628869382305533, "loss": 0.1598, "step": 5153 }, { "epoch": 0.33, "learning_rate": 0.00015627152819350617, "loss": 0.1693, "step": 5154 }, { "epoch": 0.33, "learning_rate": 0.00015625436013720487, "loss": 0.1539, "step": 5155 }, { "epoch": 0.33, "learning_rate": 0.0001562371896548918, "loss": 0.1498, "step": 5156 }, { "epoch": 0.33, "learning_rate": 0.0001562200167473075, "loss": 0.1672, "step": 5157 }, { "epoch": 0.33, "learning_rate": 0.00015620284141519246, "loss": 0.1488, "step": 5158 }, { "epoch": 0.33, "learning_rate": 0.00015618566365928747, "loss": 0.1704, "step": 5159 }, { "epoch": 0.33, "learning_rate": 0.00015616848348033334, "loss": 0.1347, "step": 5160 }, { "epoch": 0.33, "learning_rate": 0.00015615130087907092, "loss": 0.1511, "step": 5161 }, { "epoch": 0.33, "learning_rate": 0.00015613411585624126, "loss": 0.146, "step": 5162 }, { "epoch": 0.33, "learning_rate": 0.00015611692841258545, "loss": 0.1608, "step": 5163 }, { "epoch": 0.33, "learning_rate": 0.00015609973854884473, "loss": 0.1473, "step": 5164 }, { "epoch": 0.33, "learning_rate": 0.00015608254626576048, "loss": 0.1411, "step": 5165 }, { "epoch": 0.33, "learning_rate": 0.000156065351564074, "loss": 0.1566, "step": 5166 }, { "epoch": 0.33, "learning_rate": 0.00015604815444452693, "loss": 0.165, "step": 5167 }, { "epoch": 0.33, "learning_rate": 0.0001560309549078609, "loss": 0.1657, "step": 5168 }, { "epoch": 0.33, "learning_rate": 0.00015601375295481762, "loss": 0.1385, "step": 5169 }, { "epoch": 0.33, "learning_rate": 0.00015599654858613893, "loss": 0.1441, "step": 5170 }, { "epoch": 0.33, "learning_rate": 0.00015597934180256685, "loss": 0.1607, "step": 5171 }, { "epoch": 0.33, "learning_rate": 0.0001559621326048433, "loss": 0.1433, "step": 5172 }, { "epoch": 0.33, "learning_rate": 0.00015594492099371059, "loss": 0.1678, "step": 5173 }, { "epoch": 0.33, "learning_rate": 0.0001559277069699109, "loss": 0.167, "step": 5174 }, { "epoch": 0.33, "learning_rate": 0.0001559104905341866, "loss": 0.1508, "step": 5175 }, { "epoch": 0.33, "learning_rate": 0.00015589327168728022, "loss": 0.1598, "step": 5176 }, { "epoch": 0.33, "learning_rate": 0.00015587605042993429, "loss": 0.1331, "step": 5177 }, { "epoch": 0.33, "learning_rate": 0.00015585882676289145, "loss": 0.1675, "step": 5178 }, { "epoch": 0.33, "learning_rate": 0.00015584160068689458, "loss": 0.1733, "step": 5179 }, { "epoch": 0.33, "learning_rate": 0.00015582437220268647, "loss": 0.1383, "step": 5180 }, { "epoch": 0.33, "learning_rate": 0.00015580714131101017, "loss": 0.154, "step": 5181 }, { "epoch": 0.33, "learning_rate": 0.0001557899080126088, "loss": 0.1409, "step": 5182 }, { "epoch": 0.33, "learning_rate": 0.0001557726723082255, "loss": 0.1531, "step": 5183 }, { "epoch": 0.33, "learning_rate": 0.00015575543419860357, "loss": 0.1365, "step": 5184 }, { "epoch": 0.33, "learning_rate": 0.00015573819368448646, "loss": 0.1342, "step": 5185 }, { "epoch": 0.33, "learning_rate": 0.00015572095076661765, "loss": 0.1273, "step": 5186 }, { "epoch": 0.33, "learning_rate": 0.00015570370544574076, "loss": 0.1479, "step": 5187 }, { "epoch": 0.33, "learning_rate": 0.0001556864577225995, "loss": 0.1367, "step": 5188 }, { "epoch": 0.33, "learning_rate": 0.00015566920759793774, "loss": 0.1319, "step": 5189 }, { "epoch": 0.33, "learning_rate": 0.00015565195507249936, "loss": 0.1414, "step": 5190 }, { "epoch": 0.33, "learning_rate": 0.0001556347001470284, "loss": 0.1616, "step": 5191 }, { "epoch": 0.33, "learning_rate": 0.00015561744282226896, "loss": 0.1425, "step": 5192 }, { "epoch": 0.33, "learning_rate": 0.00015560018309896534, "loss": 0.1549, "step": 5193 }, { "epoch": 0.33, "learning_rate": 0.0001555829209778618, "loss": 0.1586, "step": 5194 }, { "epoch": 0.33, "learning_rate": 0.00015556565645970282, "loss": 0.1511, "step": 5195 }, { "epoch": 0.33, "learning_rate": 0.000155548389545233, "loss": 0.1328, "step": 5196 }, { "epoch": 0.33, "learning_rate": 0.0001555311202351969, "loss": 0.1449, "step": 5197 }, { "epoch": 0.33, "learning_rate": 0.0001555138485303393, "loss": 0.1602, "step": 5198 }, { "epoch": 0.33, "learning_rate": 0.0001554965744314051, "loss": 0.1548, "step": 5199 }, { "epoch": 0.33, "learning_rate": 0.0001554792979391392, "loss": 0.1565, "step": 5200 }, { "epoch": 0.33, "learning_rate": 0.00015546201905428667, "loss": 0.1403, "step": 5201 }, { "epoch": 0.33, "learning_rate": 0.00015544473777759273, "loss": 0.1281, "step": 5202 }, { "epoch": 0.33, "learning_rate": 0.00015542745410980258, "loss": 0.1568, "step": 5203 }, { "epoch": 0.33, "learning_rate": 0.00015541016805166162, "loss": 0.1645, "step": 5204 }, { "epoch": 0.33, "learning_rate": 0.00015539287960391533, "loss": 0.1608, "step": 5205 }, { "epoch": 0.33, "learning_rate": 0.00015537558876730926, "loss": 0.1407, "step": 5206 }, { "epoch": 0.33, "learning_rate": 0.00015535829554258913, "loss": 0.1865, "step": 5207 }, { "epoch": 0.33, "learning_rate": 0.00015534099993050072, "loss": 0.1477, "step": 5208 }, { "epoch": 0.33, "learning_rate": 0.00015532370193178986, "loss": 0.1486, "step": 5209 }, { "epoch": 0.33, "learning_rate": 0.00015530640154720263, "loss": 0.1659, "step": 5210 }, { "epoch": 0.33, "learning_rate": 0.00015528909877748505, "loss": 0.1618, "step": 5211 }, { "epoch": 0.33, "learning_rate": 0.00015527179362338333, "loss": 0.1486, "step": 5212 }, { "epoch": 0.33, "learning_rate": 0.00015525448608564378, "loss": 0.1576, "step": 5213 }, { "epoch": 0.33, "learning_rate": 0.0001552371761650128, "loss": 0.1496, "step": 5214 }, { "epoch": 0.33, "learning_rate": 0.0001552198638622369, "loss": 0.1484, "step": 5215 }, { "epoch": 0.33, "learning_rate": 0.00015520254917806265, "loss": 0.1261, "step": 5216 }, { "epoch": 0.33, "learning_rate": 0.00015518523211323682, "loss": 0.1579, "step": 5217 }, { "epoch": 0.33, "learning_rate": 0.00015516791266850612, "loss": 0.1674, "step": 5218 }, { "epoch": 0.33, "learning_rate": 0.0001551505908446176, "loss": 0.1575, "step": 5219 }, { "epoch": 0.33, "learning_rate": 0.00015513326664231815, "loss": 0.1689, "step": 5220 }, { "epoch": 0.33, "learning_rate": 0.000155115940062355, "loss": 0.1566, "step": 5221 }, { "epoch": 0.33, "learning_rate": 0.00015509861110547528, "loss": 0.1395, "step": 5222 }, { "epoch": 0.33, "learning_rate": 0.00015508127977242636, "loss": 0.1499, "step": 5223 }, { "epoch": 0.33, "learning_rate": 0.00015506394606395568, "loss": 0.14, "step": 5224 }, { "epoch": 0.34, "learning_rate": 0.00015504660998081073, "loss": 0.1502, "step": 5225 }, { "epoch": 0.34, "learning_rate": 0.00015502927152373914, "loss": 0.1894, "step": 5226 }, { "epoch": 0.34, "learning_rate": 0.0001550119306934887, "loss": 0.1741, "step": 5227 }, { "epoch": 0.34, "learning_rate": 0.0001549945874908072, "loss": 0.1501, "step": 5228 }, { "epoch": 0.34, "learning_rate": 0.00015497724191644257, "loss": 0.155, "step": 5229 }, { "epoch": 0.34, "learning_rate": 0.00015495989397114289, "loss": 0.141, "step": 5230 }, { "epoch": 0.34, "learning_rate": 0.00015494254365565628, "loss": 0.1295, "step": 5231 }, { "epoch": 0.34, "learning_rate": 0.00015492519097073097, "loss": 0.1549, "step": 5232 }, { "epoch": 0.34, "learning_rate": 0.00015490783591711534, "loss": 0.1397, "step": 5233 }, { "epoch": 0.34, "learning_rate": 0.00015489047849555782, "loss": 0.145, "step": 5234 }, { "epoch": 0.34, "learning_rate": 0.00015487311870680698, "loss": 0.1544, "step": 5235 }, { "epoch": 0.34, "learning_rate": 0.00015485575655161147, "loss": 0.1538, "step": 5236 }, { "epoch": 0.34, "learning_rate": 0.00015483839203072, "loss": 0.1712, "step": 5237 }, { "epoch": 0.34, "learning_rate": 0.00015482102514488154, "loss": 0.1469, "step": 5238 }, { "epoch": 0.34, "learning_rate": 0.0001548036558948449, "loss": 0.1354, "step": 5239 }, { "epoch": 0.34, "learning_rate": 0.00015478628428135924, "loss": 0.1557, "step": 5240 }, { "epoch": 0.34, "learning_rate": 0.00015476891030517372, "loss": 0.1544, "step": 5241 }, { "epoch": 0.34, "learning_rate": 0.00015475153396703756, "loss": 0.1468, "step": 5242 }, { "epoch": 0.34, "learning_rate": 0.00015473415526770019, "loss": 0.1463, "step": 5243 }, { "epoch": 0.34, "learning_rate": 0.00015471677420791102, "loss": 0.131, "step": 5244 }, { "epoch": 0.34, "learning_rate": 0.00015469939078841964, "loss": 0.1338, "step": 5245 }, { "epoch": 0.34, "learning_rate": 0.00015468200500997576, "loss": 0.155, "step": 5246 }, { "epoch": 0.34, "learning_rate": 0.0001546646168733291, "loss": 0.1632, "step": 5247 }, { "epoch": 0.34, "learning_rate": 0.00015464722637922957, "loss": 0.1539, "step": 5248 }, { "epoch": 0.34, "learning_rate": 0.00015462983352842712, "loss": 0.1414, "step": 5249 }, { "epoch": 0.34, "learning_rate": 0.00015461243832167186, "loss": 0.1494, "step": 5250 }, { "epoch": 0.34, "learning_rate": 0.000154595040759714, "loss": 0.1308, "step": 5251 }, { "epoch": 0.34, "learning_rate": 0.00015457764084330375, "loss": 0.1691, "step": 5252 }, { "epoch": 0.34, "learning_rate": 0.00015456023857319154, "loss": 0.1663, "step": 5253 }, { "epoch": 0.34, "learning_rate": 0.0001545428339501278, "loss": 0.1656, "step": 5254 }, { "epoch": 0.34, "learning_rate": 0.0001545254269748632, "loss": 0.13, "step": 5255 }, { "epoch": 0.34, "learning_rate": 0.00015450801764814838, "loss": 0.1601, "step": 5256 }, { "epoch": 0.34, "learning_rate": 0.00015449060597073417, "loss": 0.1565, "step": 5257 }, { "epoch": 0.34, "learning_rate": 0.00015447319194337143, "loss": 0.1467, "step": 5258 }, { "epoch": 0.34, "learning_rate": 0.0001544557755668111, "loss": 0.1476, "step": 5259 }, { "epoch": 0.34, "learning_rate": 0.00015443835684180436, "loss": 0.1475, "step": 5260 }, { "epoch": 0.34, "learning_rate": 0.0001544209357691024, "loss": 0.1566, "step": 5261 }, { "epoch": 0.34, "learning_rate": 0.00015440351234945646, "loss": 0.1568, "step": 5262 }, { "epoch": 0.34, "learning_rate": 0.00015438608658361797, "loss": 0.1592, "step": 5263 }, { "epoch": 0.34, "learning_rate": 0.00015436865847233845, "loss": 0.1436, "step": 5264 }, { "epoch": 0.34, "learning_rate": 0.00015435122801636948, "loss": 0.1553, "step": 5265 }, { "epoch": 0.34, "learning_rate": 0.00015433379521646276, "loss": 0.1517, "step": 5266 }, { "epoch": 0.34, "learning_rate": 0.0001543163600733701, "loss": 0.151, "step": 5267 }, { "epoch": 0.34, "learning_rate": 0.00015429892258784337, "loss": 0.1465, "step": 5268 }, { "epoch": 0.34, "learning_rate": 0.00015428148276063462, "loss": 0.1292, "step": 5269 }, { "epoch": 0.34, "learning_rate": 0.00015426404059249596, "loss": 0.146, "step": 5270 }, { "epoch": 0.34, "learning_rate": 0.00015424659608417955, "loss": 0.1569, "step": 5271 }, { "epoch": 0.34, "learning_rate": 0.00015422914923643772, "loss": 0.1403, "step": 5272 }, { "epoch": 0.34, "learning_rate": 0.0001542117000500229, "loss": 0.1363, "step": 5273 }, { "epoch": 0.34, "learning_rate": 0.00015419424852568757, "loss": 0.161, "step": 5274 }, { "epoch": 0.34, "learning_rate": 0.00015417679466418435, "loss": 0.1726, "step": 5275 }, { "epoch": 0.34, "learning_rate": 0.00015415933846626593, "loss": 0.156, "step": 5276 }, { "epoch": 0.34, "learning_rate": 0.0001541418799326852, "loss": 0.1359, "step": 5277 }, { "epoch": 0.34, "learning_rate": 0.000154124419064195, "loss": 0.1683, "step": 5278 }, { "epoch": 0.34, "learning_rate": 0.00015410695586154832, "loss": 0.1466, "step": 5279 }, { "epoch": 0.34, "learning_rate": 0.00015408949032549835, "loss": 0.1526, "step": 5280 }, { "epoch": 0.34, "learning_rate": 0.00015407202245679825, "loss": 0.1534, "step": 5281 }, { "epoch": 0.34, "learning_rate": 0.00015405455225620137, "loss": 0.1378, "step": 5282 }, { "epoch": 0.34, "learning_rate": 0.0001540370797244611, "loss": 0.1527, "step": 5283 }, { "epoch": 0.34, "learning_rate": 0.00015401960486233093, "loss": 0.1758, "step": 5284 }, { "epoch": 0.34, "learning_rate": 0.00015400212767056456, "loss": 0.1384, "step": 5285 }, { "epoch": 0.34, "learning_rate": 0.00015398464814991565, "loss": 0.1705, "step": 5286 }, { "epoch": 0.34, "learning_rate": 0.000153967166301138, "loss": 0.1639, "step": 5287 }, { "epoch": 0.34, "learning_rate": 0.00015394968212498555, "loss": 0.1546, "step": 5288 }, { "epoch": 0.34, "learning_rate": 0.00015393219562221235, "loss": 0.1654, "step": 5289 }, { "epoch": 0.34, "learning_rate": 0.00015391470679357248, "loss": 0.1449, "step": 5290 }, { "epoch": 0.34, "learning_rate": 0.0001538972156398202, "loss": 0.1353, "step": 5291 }, { "epoch": 0.34, "learning_rate": 0.00015387972216170977, "loss": 0.1425, "step": 5292 }, { "epoch": 0.34, "learning_rate": 0.00015386222635999563, "loss": 0.1696, "step": 5293 }, { "epoch": 0.34, "learning_rate": 0.0001538447282354323, "loss": 0.1306, "step": 5294 }, { "epoch": 0.34, "learning_rate": 0.00015382722778877443, "loss": 0.155, "step": 5295 }, { "epoch": 0.34, "learning_rate": 0.0001538097250207767, "loss": 0.1568, "step": 5296 }, { "epoch": 0.34, "learning_rate": 0.00015379221993219396, "loss": 0.1639, "step": 5297 }, { "epoch": 0.34, "learning_rate": 0.00015377471252378113, "loss": 0.1567, "step": 5298 }, { "epoch": 0.34, "learning_rate": 0.0001537572027962932, "loss": 0.1628, "step": 5299 }, { "epoch": 0.34, "learning_rate": 0.00015373969075048533, "loss": 0.1559, "step": 5300 }, { "epoch": 0.34, "learning_rate": 0.0001537221763871127, "loss": 0.1461, "step": 5301 }, { "epoch": 0.34, "learning_rate": 0.0001537046597069306, "loss": 0.1385, "step": 5302 }, { "epoch": 0.34, "learning_rate": 0.0001536871407106946, "loss": 0.1307, "step": 5303 }, { "epoch": 0.34, "learning_rate": 0.00015366961939916008, "loss": 0.1496, "step": 5304 }, { "epoch": 0.34, "learning_rate": 0.00015365209577308266, "loss": 0.134, "step": 5305 }, { "epoch": 0.34, "learning_rate": 0.00015363456983321814, "loss": 0.1722, "step": 5306 }, { "epoch": 0.34, "learning_rate": 0.0001536170415803223, "loss": 0.1491, "step": 5307 }, { "epoch": 0.34, "learning_rate": 0.000153599511015151, "loss": 0.1495, "step": 5308 }, { "epoch": 0.34, "learning_rate": 0.00015358197813846036, "loss": 0.1501, "step": 5309 }, { "epoch": 0.34, "learning_rate": 0.00015356444295100644, "loss": 0.1358, "step": 5310 }, { "epoch": 0.34, "learning_rate": 0.00015354690545354548, "loss": 0.1686, "step": 5311 }, { "epoch": 0.34, "learning_rate": 0.0001535293656468338, "loss": 0.1353, "step": 5312 }, { "epoch": 0.34, "learning_rate": 0.0001535118235316278, "loss": 0.137, "step": 5313 }, { "epoch": 0.34, "learning_rate": 0.000153494279108684, "loss": 0.1588, "step": 5314 }, { "epoch": 0.34, "learning_rate": 0.000153476732378759, "loss": 0.1683, "step": 5315 }, { "epoch": 0.34, "learning_rate": 0.00015345918334260956, "loss": 0.1592, "step": 5316 }, { "epoch": 0.34, "learning_rate": 0.00015344163200099246, "loss": 0.1372, "step": 5317 }, { "epoch": 0.34, "learning_rate": 0.00015342407835466465, "loss": 0.1339, "step": 5318 }, { "epoch": 0.34, "learning_rate": 0.0001534065224043831, "loss": 0.1435, "step": 5319 }, { "epoch": 0.34, "learning_rate": 0.000153388964150905, "loss": 0.1558, "step": 5320 }, { "epoch": 0.34, "learning_rate": 0.00015337140359498745, "loss": 0.1522, "step": 5321 }, { "epoch": 0.34, "learning_rate": 0.0001533538407373878, "loss": 0.1589, "step": 5322 }, { "epoch": 0.34, "learning_rate": 0.00015333627557886353, "loss": 0.1688, "step": 5323 }, { "epoch": 0.34, "learning_rate": 0.0001533187081201721, "loss": 0.137, "step": 5324 }, { "epoch": 0.34, "learning_rate": 0.00015330113836207107, "loss": 0.1499, "step": 5325 }, { "epoch": 0.34, "learning_rate": 0.00015328356630531825, "loss": 0.159, "step": 5326 }, { "epoch": 0.34, "learning_rate": 0.00015326599195067143, "loss": 0.1528, "step": 5327 }, { "epoch": 0.34, "learning_rate": 0.0001532484152988884, "loss": 0.1545, "step": 5328 }, { "epoch": 0.34, "learning_rate": 0.0001532308363507273, "loss": 0.1468, "step": 5329 }, { "epoch": 0.34, "learning_rate": 0.0001532132551069462, "loss": 0.1459, "step": 5330 }, { "epoch": 0.34, "learning_rate": 0.00015319567156830329, "loss": 0.1535, "step": 5331 }, { "epoch": 0.34, "learning_rate": 0.00015317808573555683, "loss": 0.1593, "step": 5332 }, { "epoch": 0.34, "learning_rate": 0.00015316049760946533, "loss": 0.1599, "step": 5333 }, { "epoch": 0.34, "learning_rate": 0.00015314290719078725, "loss": 0.1631, "step": 5334 }, { "epoch": 0.34, "learning_rate": 0.00015312531448028106, "loss": 0.1407, "step": 5335 }, { "epoch": 0.34, "learning_rate": 0.00015310771947870565, "loss": 0.1527, "step": 5336 }, { "epoch": 0.34, "learning_rate": 0.00015309012218681973, "loss": 0.1518, "step": 5337 }, { "epoch": 0.34, "learning_rate": 0.00015307252260538217, "loss": 0.1336, "step": 5338 }, { "epoch": 0.34, "learning_rate": 0.00015305492073515206, "loss": 0.1688, "step": 5339 }, { "epoch": 0.34, "learning_rate": 0.0001530373165768884, "loss": 0.1547, "step": 5340 }, { "epoch": 0.34, "learning_rate": 0.0001530197101313504, "loss": 0.1578, "step": 5341 }, { "epoch": 0.34, "learning_rate": 0.00015300210139929736, "loss": 0.1387, "step": 5342 }, { "epoch": 0.34, "learning_rate": 0.0001529844903814887, "loss": 0.1608, "step": 5343 }, { "epoch": 0.34, "learning_rate": 0.0001529668770786839, "loss": 0.1468, "step": 5344 }, { "epoch": 0.34, "learning_rate": 0.00015294926149164248, "loss": 0.1543, "step": 5345 }, { "epoch": 0.34, "learning_rate": 0.00015293164362112423, "loss": 0.1412, "step": 5346 }, { "epoch": 0.34, "learning_rate": 0.00015291402346788885, "loss": 0.1329, "step": 5347 }, { "epoch": 0.34, "learning_rate": 0.00015289640103269625, "loss": 0.1622, "step": 5348 }, { "epoch": 0.34, "learning_rate": 0.00015287877631630641, "loss": 0.1699, "step": 5349 }, { "epoch": 0.34, "learning_rate": 0.00015286114931947946, "loss": 0.1609, "step": 5350 }, { "epoch": 0.34, "learning_rate": 0.0001528435200429755, "loss": 0.1407, "step": 5351 }, { "epoch": 0.34, "learning_rate": 0.0001528258884875548, "loss": 0.1289, "step": 5352 }, { "epoch": 0.34, "learning_rate": 0.00015280825465397783, "loss": 0.1748, "step": 5353 }, { "epoch": 0.34, "learning_rate": 0.000152790618543005, "loss": 0.141, "step": 5354 }, { "epoch": 0.34, "learning_rate": 0.00015277298015539686, "loss": 0.1591, "step": 5355 }, { "epoch": 0.34, "learning_rate": 0.0001527553394919141, "loss": 0.1549, "step": 5356 }, { "epoch": 0.34, "learning_rate": 0.0001527376965533175, "loss": 0.1501, "step": 5357 }, { "epoch": 0.34, "learning_rate": 0.0001527200513403679, "loss": 0.153, "step": 5358 }, { "epoch": 0.34, "learning_rate": 0.00015270240385382628, "loss": 0.1487, "step": 5359 }, { "epoch": 0.34, "learning_rate": 0.00015268475409445373, "loss": 0.1441, "step": 5360 }, { "epoch": 0.34, "learning_rate": 0.00015266710206301136, "loss": 0.137, "step": 5361 }, { "epoch": 0.34, "learning_rate": 0.0001526494477602604, "loss": 0.1277, "step": 5362 }, { "epoch": 0.34, "learning_rate": 0.00015263179118696232, "loss": 0.1756, "step": 5363 }, { "epoch": 0.34, "learning_rate": 0.0001526141323438785, "loss": 0.1649, "step": 5364 }, { "epoch": 0.34, "learning_rate": 0.00015259647123177044, "loss": 0.1431, "step": 5365 }, { "epoch": 0.34, "learning_rate": 0.00015257880785139987, "loss": 0.1516, "step": 5366 }, { "epoch": 0.34, "learning_rate": 0.00015256114220352855, "loss": 0.153, "step": 5367 }, { "epoch": 0.34, "learning_rate": 0.00015254347428891825, "loss": 0.1505, "step": 5368 }, { "epoch": 0.34, "learning_rate": 0.00015252580410833093, "loss": 0.1634, "step": 5369 }, { "epoch": 0.34, "learning_rate": 0.00015250813166252866, "loss": 0.1479, "step": 5370 }, { "epoch": 0.34, "learning_rate": 0.0001524904569522736, "loss": 0.1301, "step": 5371 }, { "epoch": 0.34, "learning_rate": 0.00015247277997832787, "loss": 0.1632, "step": 5372 }, { "epoch": 0.34, "learning_rate": 0.00015245510074145396, "loss": 0.1664, "step": 5373 }, { "epoch": 0.34, "learning_rate": 0.0001524374192424142, "loss": 0.1536, "step": 5374 }, { "epoch": 0.34, "learning_rate": 0.0001524197354819711, "loss": 0.1711, "step": 5375 }, { "epoch": 0.34, "learning_rate": 0.00015240204946088737, "loss": 0.1669, "step": 5376 }, { "epoch": 0.34, "learning_rate": 0.00015238436117992569, "loss": 0.1417, "step": 5377 }, { "epoch": 0.34, "learning_rate": 0.00015236667063984885, "loss": 0.15, "step": 5378 }, { "epoch": 0.34, "learning_rate": 0.00015234897784141983, "loss": 0.1333, "step": 5379 }, { "epoch": 0.34, "learning_rate": 0.00015233128278540158, "loss": 0.1509, "step": 5380 }, { "epoch": 0.35, "learning_rate": 0.0001523135854725573, "loss": 0.1446, "step": 5381 }, { "epoch": 0.35, "learning_rate": 0.00015229588590365007, "loss": 0.1356, "step": 5382 }, { "epoch": 0.35, "learning_rate": 0.00015227818407944333, "loss": 0.1444, "step": 5383 }, { "epoch": 0.35, "learning_rate": 0.0001522604800007004, "loss": 0.1473, "step": 5384 }, { "epoch": 0.35, "learning_rate": 0.0001522427736681848, "loss": 0.1504, "step": 5385 }, { "epoch": 0.35, "learning_rate": 0.00015222506508266015, "loss": 0.1602, "step": 5386 }, { "epoch": 0.35, "learning_rate": 0.00015220735424489016, "loss": 0.1496, "step": 5387 }, { "epoch": 0.35, "learning_rate": 0.0001521896411556386, "loss": 0.1403, "step": 5388 }, { "epoch": 0.35, "learning_rate": 0.0001521719258156693, "loss": 0.1405, "step": 5389 }, { "epoch": 0.35, "learning_rate": 0.00015215420822574632, "loss": 0.1533, "step": 5390 }, { "epoch": 0.35, "learning_rate": 0.00015213648838663377, "loss": 0.1398, "step": 5391 }, { "epoch": 0.35, "learning_rate": 0.00015211876629909576, "loss": 0.1716, "step": 5392 }, { "epoch": 0.35, "learning_rate": 0.0001521010419638966, "loss": 0.1424, "step": 5393 }, { "epoch": 0.35, "learning_rate": 0.00015208331538180069, "loss": 0.1608, "step": 5394 }, { "epoch": 0.35, "learning_rate": 0.00015206558655357245, "loss": 0.1533, "step": 5395 }, { "epoch": 0.35, "learning_rate": 0.00015204785547997648, "loss": 0.1378, "step": 5396 }, { "epoch": 0.35, "learning_rate": 0.00015203012216177749, "loss": 0.1392, "step": 5397 }, { "epoch": 0.35, "learning_rate": 0.00015201238659974013, "loss": 0.1745, "step": 5398 }, { "epoch": 0.35, "learning_rate": 0.00015199464879462932, "loss": 0.1677, "step": 5399 }, { "epoch": 0.35, "learning_rate": 0.00015197690874721003, "loss": 0.1426, "step": 5400 }, { "epoch": 0.35, "learning_rate": 0.00015195916645824737, "loss": 0.124, "step": 5401 }, { "epoch": 0.35, "learning_rate": 0.00015194142192850635, "loss": 0.1507, "step": 5402 }, { "epoch": 0.35, "learning_rate": 0.00015192367515875233, "loss": 0.1346, "step": 5403 }, { "epoch": 0.35, "learning_rate": 0.0001519059261497506, "loss": 0.1539, "step": 5404 }, { "epoch": 0.35, "learning_rate": 0.0001518881749022666, "loss": 0.1499, "step": 5405 }, { "epoch": 0.35, "learning_rate": 0.00015187042141706587, "loss": 0.1476, "step": 5406 }, { "epoch": 0.35, "learning_rate": 0.00015185266569491405, "loss": 0.1533, "step": 5407 }, { "epoch": 0.35, "learning_rate": 0.0001518349077365769, "loss": 0.1416, "step": 5408 }, { "epoch": 0.35, "learning_rate": 0.0001518171475428202, "loss": 0.1479, "step": 5409 }, { "epoch": 0.35, "learning_rate": 0.0001517993851144099, "loss": 0.1634, "step": 5410 }, { "epoch": 0.35, "learning_rate": 0.00015178162045211197, "loss": 0.1296, "step": 5411 }, { "epoch": 0.35, "learning_rate": 0.00015176385355669255, "loss": 0.1676, "step": 5412 }, { "epoch": 0.35, "learning_rate": 0.0001517460844289179, "loss": 0.1365, "step": 5413 }, { "epoch": 0.35, "learning_rate": 0.00015172831306955428, "loss": 0.1487, "step": 5414 }, { "epoch": 0.35, "learning_rate": 0.00015171053947936807, "loss": 0.1808, "step": 5415 }, { "epoch": 0.35, "learning_rate": 0.00015169276365912583, "loss": 0.1542, "step": 5416 }, { "epoch": 0.35, "learning_rate": 0.00015167498560959413, "loss": 0.1465, "step": 5417 }, { "epoch": 0.35, "learning_rate": 0.00015165720533153964, "loss": 0.1689, "step": 5418 }, { "epoch": 0.35, "learning_rate": 0.00015163942282572914, "loss": 0.1379, "step": 5419 }, { "epoch": 0.35, "learning_rate": 0.00015162163809292957, "loss": 0.1499, "step": 5420 }, { "epoch": 0.35, "learning_rate": 0.00015160385113390792, "loss": 0.1633, "step": 5421 }, { "epoch": 0.35, "learning_rate": 0.00015158606194943115, "loss": 0.1467, "step": 5422 }, { "epoch": 0.35, "learning_rate": 0.00015156827054026652, "loss": 0.1559, "step": 5423 }, { "epoch": 0.35, "learning_rate": 0.0001515504769071813, "loss": 0.1623, "step": 5424 }, { "epoch": 0.35, "learning_rate": 0.00015153268105094284, "loss": 0.1641, "step": 5425 }, { "epoch": 0.35, "learning_rate": 0.0001515148829723186, "loss": 0.1433, "step": 5426 }, { "epoch": 0.35, "learning_rate": 0.0001514970826720761, "loss": 0.1385, "step": 5427 }, { "epoch": 0.35, "learning_rate": 0.0001514792801509831, "loss": 0.156, "step": 5428 }, { "epoch": 0.35, "learning_rate": 0.0001514614754098072, "loss": 0.1664, "step": 5429 }, { "epoch": 0.35, "learning_rate": 0.00015144366844931635, "loss": 0.1556, "step": 5430 }, { "epoch": 0.35, "learning_rate": 0.00015142585927027845, "loss": 0.1371, "step": 5431 }, { "epoch": 0.35, "learning_rate": 0.00015140804787346153, "loss": 0.1479, "step": 5432 }, { "epoch": 0.35, "learning_rate": 0.00015139023425963372, "loss": 0.1507, "step": 5433 }, { "epoch": 0.35, "learning_rate": 0.00015137241842956328, "loss": 0.1696, "step": 5434 }, { "epoch": 0.35, "learning_rate": 0.0001513546003840185, "loss": 0.152, "step": 5435 }, { "epoch": 0.35, "learning_rate": 0.00015133678012376777, "loss": 0.1445, "step": 5436 }, { "epoch": 0.35, "learning_rate": 0.00015131895764957968, "loss": 0.1644, "step": 5437 }, { "epoch": 0.35, "learning_rate": 0.00015130113296222275, "loss": 0.1302, "step": 5438 }, { "epoch": 0.35, "learning_rate": 0.00015128330606246576, "loss": 0.1757, "step": 5439 }, { "epoch": 0.35, "learning_rate": 0.0001512654769510774, "loss": 0.1822, "step": 5440 }, { "epoch": 0.35, "learning_rate": 0.0001512476456288267, "loss": 0.1707, "step": 5441 }, { "epoch": 0.35, "learning_rate": 0.0001512298120964826, "loss": 0.1517, "step": 5442 }, { "epoch": 0.35, "learning_rate": 0.00015121197635481414, "loss": 0.1583, "step": 5443 }, { "epoch": 0.35, "learning_rate": 0.00015119413840459055, "loss": 0.1455, "step": 5444 }, { "epoch": 0.35, "learning_rate": 0.0001511762982465811, "loss": 0.1391, "step": 5445 }, { "epoch": 0.35, "learning_rate": 0.00015115845588155509, "loss": 0.1505, "step": 5446 }, { "epoch": 0.35, "learning_rate": 0.00015114061131028206, "loss": 0.1622, "step": 5447 }, { "epoch": 0.35, "learning_rate": 0.00015112276453353157, "loss": 0.1598, "step": 5448 }, { "epoch": 0.35, "learning_rate": 0.00015110491555207328, "loss": 0.1463, "step": 5449 }, { "epoch": 0.35, "learning_rate": 0.0001510870643666769, "loss": 0.1556, "step": 5450 }, { "epoch": 0.35, "learning_rate": 0.00015106921097811227, "loss": 0.1459, "step": 5451 }, { "epoch": 0.35, "learning_rate": 0.00015105135538714937, "loss": 0.165, "step": 5452 }, { "epoch": 0.35, "learning_rate": 0.0001510334975945582, "loss": 0.1691, "step": 5453 }, { "epoch": 0.35, "learning_rate": 0.00015101563760110893, "loss": 0.1644, "step": 5454 }, { "epoch": 0.35, "learning_rate": 0.00015099777540757177, "loss": 0.1379, "step": 5455 }, { "epoch": 0.35, "learning_rate": 0.000150979911014717, "loss": 0.1442, "step": 5456 }, { "epoch": 0.35, "learning_rate": 0.00015096204442331513, "loss": 0.1512, "step": 5457 }, { "epoch": 0.35, "learning_rate": 0.0001509441756341366, "loss": 0.1668, "step": 5458 }, { "epoch": 0.35, "learning_rate": 0.000150926304647952, "loss": 0.1758, "step": 5459 }, { "epoch": 0.35, "learning_rate": 0.0001509084314655321, "loss": 0.1391, "step": 5460 }, { "epoch": 0.35, "learning_rate": 0.00015089055608764765, "loss": 0.1464, "step": 5461 }, { "epoch": 0.35, "learning_rate": 0.0001508726785150695, "loss": 0.1454, "step": 5462 }, { "epoch": 0.35, "learning_rate": 0.0001508547987485687, "loss": 0.1348, "step": 5463 }, { "epoch": 0.35, "learning_rate": 0.00015083691678891633, "loss": 0.1622, "step": 5464 }, { "epoch": 0.35, "learning_rate": 0.00015081903263688352, "loss": 0.1583, "step": 5465 }, { "epoch": 0.35, "learning_rate": 0.0001508011462932415, "loss": 0.1597, "step": 5466 }, { "epoch": 0.35, "learning_rate": 0.00015078325775876175, "loss": 0.1473, "step": 5467 }, { "epoch": 0.35, "learning_rate": 0.00015076536703421565, "loss": 0.1781, "step": 5468 }, { "epoch": 0.35, "learning_rate": 0.00015074747412037478, "loss": 0.164, "step": 5469 }, { "epoch": 0.35, "learning_rate": 0.00015072957901801076, "loss": 0.1409, "step": 5470 }, { "epoch": 0.35, "learning_rate": 0.00015071168172789534, "loss": 0.1587, "step": 5471 }, { "epoch": 0.35, "learning_rate": 0.00015069378225080032, "loss": 0.1585, "step": 5472 }, { "epoch": 0.35, "learning_rate": 0.00015067588058749769, "loss": 0.1555, "step": 5473 }, { "epoch": 0.35, "learning_rate": 0.00015065797673875943, "loss": 0.156, "step": 5474 }, { "epoch": 0.35, "learning_rate": 0.00015064007070535767, "loss": 0.1615, "step": 5475 }, { "epoch": 0.35, "learning_rate": 0.0001506221624880647, "loss": 0.1545, "step": 5476 }, { "epoch": 0.35, "learning_rate": 0.00015060425208765268, "loss": 0.1881, "step": 5477 }, { "epoch": 0.35, "learning_rate": 0.00015058633950489406, "loss": 0.1576, "step": 5478 }, { "epoch": 0.35, "learning_rate": 0.0001505684247405614, "loss": 0.1584, "step": 5479 }, { "epoch": 0.35, "learning_rate": 0.00015055050779542718, "loss": 0.1479, "step": 5480 }, { "epoch": 0.35, "learning_rate": 0.00015053258867026421, "loss": 0.161, "step": 5481 }, { "epoch": 0.35, "learning_rate": 0.00015051466736584515, "loss": 0.156, "step": 5482 }, { "epoch": 0.35, "learning_rate": 0.00015049674388294295, "loss": 0.1584, "step": 5483 }, { "epoch": 0.35, "learning_rate": 0.00015047881822233053, "loss": 0.1362, "step": 5484 }, { "epoch": 0.35, "learning_rate": 0.00015046089038478095, "loss": 0.1591, "step": 5485 }, { "epoch": 0.35, "learning_rate": 0.00015044296037106735, "loss": 0.1256, "step": 5486 }, { "epoch": 0.35, "learning_rate": 0.000150425028181963, "loss": 0.1488, "step": 5487 }, { "epoch": 0.35, "learning_rate": 0.0001504070938182413, "loss": 0.1517, "step": 5488 }, { "epoch": 0.35, "learning_rate": 0.00015038915728067556, "loss": 0.165, "step": 5489 }, { "epoch": 0.35, "learning_rate": 0.00015037121857003937, "loss": 0.1551, "step": 5490 }, { "epoch": 0.35, "learning_rate": 0.00015035327768710634, "loss": 0.1515, "step": 5491 }, { "epoch": 0.35, "learning_rate": 0.0001503353346326502, "loss": 0.1497, "step": 5492 }, { "epoch": 0.35, "learning_rate": 0.00015031738940744473, "loss": 0.1718, "step": 5493 }, { "epoch": 0.35, "learning_rate": 0.00015029944201226387, "loss": 0.1797, "step": 5494 }, { "epoch": 0.35, "learning_rate": 0.00015028149244788162, "loss": 0.1617, "step": 5495 }, { "epoch": 0.35, "learning_rate": 0.000150263540715072, "loss": 0.1468, "step": 5496 }, { "epoch": 0.35, "learning_rate": 0.00015024558681460924, "loss": 0.152, "step": 5497 }, { "epoch": 0.35, "learning_rate": 0.0001502276307472676, "loss": 0.1693, "step": 5498 }, { "epoch": 0.35, "learning_rate": 0.00015020967251382146, "loss": 0.152, "step": 5499 }, { "epoch": 0.35, "learning_rate": 0.0001501917121150453, "loss": 0.1624, "step": 5500 }, { "epoch": 0.35, "learning_rate": 0.00015017374955171364, "loss": 0.1454, "step": 5501 }, { "epoch": 0.35, "learning_rate": 0.00015015578482460116, "loss": 0.1551, "step": 5502 }, { "epoch": 0.35, "learning_rate": 0.00015013781793448257, "loss": 0.1625, "step": 5503 }, { "epoch": 0.35, "learning_rate": 0.00015011984888213274, "loss": 0.1572, "step": 5504 }, { "epoch": 0.35, "learning_rate": 0.00015010187766832658, "loss": 0.1467, "step": 5505 }, { "epoch": 0.35, "learning_rate": 0.00015008390429383908, "loss": 0.1464, "step": 5506 }, { "epoch": 0.35, "learning_rate": 0.00015006592875944542, "loss": 0.1286, "step": 5507 }, { "epoch": 0.35, "learning_rate": 0.00015004795106592078, "loss": 0.1478, "step": 5508 }, { "epoch": 0.35, "learning_rate": 0.00015002997121404046, "loss": 0.144, "step": 5509 }, { "epoch": 0.35, "learning_rate": 0.00015001198920457985, "loss": 0.1599, "step": 5510 }, { "epoch": 0.35, "learning_rate": 0.00014999400503831444, "loss": 0.1682, "step": 5511 }, { "epoch": 0.35, "learning_rate": 0.00014997601871601982, "loss": 0.1252, "step": 5512 }, { "epoch": 0.35, "learning_rate": 0.00014995803023847164, "loss": 0.1552, "step": 5513 }, { "epoch": 0.35, "learning_rate": 0.0001499400396064457, "loss": 0.1544, "step": 5514 }, { "epoch": 0.35, "learning_rate": 0.00014992204682071785, "loss": 0.1445, "step": 5515 }, { "epoch": 0.35, "learning_rate": 0.00014990405188206401, "loss": 0.13, "step": 5516 }, { "epoch": 0.35, "learning_rate": 0.00014988605479126027, "loss": 0.1441, "step": 5517 }, { "epoch": 0.35, "learning_rate": 0.00014986805554908274, "loss": 0.191, "step": 5518 }, { "epoch": 0.35, "learning_rate": 0.00014985005415630765, "loss": 0.1616, "step": 5519 }, { "epoch": 0.35, "learning_rate": 0.00014983205061371138, "loss": 0.1655, "step": 5520 }, { "epoch": 0.35, "learning_rate": 0.00014981404492207025, "loss": 0.1568, "step": 5521 }, { "epoch": 0.35, "learning_rate": 0.00014979603708216083, "loss": 0.1554, "step": 5522 }, { "epoch": 0.35, "learning_rate": 0.0001497780270947597, "loss": 0.1624, "step": 5523 }, { "epoch": 0.35, "learning_rate": 0.0001497600149606436, "loss": 0.1584, "step": 5524 }, { "epoch": 0.35, "learning_rate": 0.00014974200068058928, "loss": 0.1659, "step": 5525 }, { "epoch": 0.35, "learning_rate": 0.0001497239842553736, "loss": 0.1436, "step": 5526 }, { "epoch": 0.35, "learning_rate": 0.00014970596568577356, "loss": 0.1593, "step": 5527 }, { "epoch": 0.35, "learning_rate": 0.00014968794497256623, "loss": 0.1491, "step": 5528 }, { "epoch": 0.35, "learning_rate": 0.00014966992211652873, "loss": 0.1516, "step": 5529 }, { "epoch": 0.35, "learning_rate": 0.00014965189711843831, "loss": 0.1557, "step": 5530 }, { "epoch": 0.35, "learning_rate": 0.0001496338699790724, "loss": 0.1769, "step": 5531 }, { "epoch": 0.35, "learning_rate": 0.00014961584069920835, "loss": 0.1748, "step": 5532 }, { "epoch": 0.35, "learning_rate": 0.00014959780927962369, "loss": 0.1382, "step": 5533 }, { "epoch": 0.35, "learning_rate": 0.00014957977572109605, "loss": 0.1564, "step": 5534 }, { "epoch": 0.35, "learning_rate": 0.00014956174002440315, "loss": 0.1508, "step": 5535 }, { "epoch": 0.35, "learning_rate": 0.00014954370219032282, "loss": 0.1607, "step": 5536 }, { "epoch": 0.36, "learning_rate": 0.00014952566221963288, "loss": 0.1476, "step": 5537 }, { "epoch": 0.36, "learning_rate": 0.0001495076201131114, "loss": 0.1619, "step": 5538 }, { "epoch": 0.36, "learning_rate": 0.00014948957587153637, "loss": 0.1467, "step": 5539 }, { "epoch": 0.36, "learning_rate": 0.00014947152949568605, "loss": 0.1499, "step": 5540 }, { "epoch": 0.36, "learning_rate": 0.00014945348098633864, "loss": 0.1459, "step": 5541 }, { "epoch": 0.36, "learning_rate": 0.00014943543034427254, "loss": 0.1535, "step": 5542 }, { "epoch": 0.36, "learning_rate": 0.00014941737757026617, "loss": 0.1571, "step": 5543 }, { "epoch": 0.36, "learning_rate": 0.00014939932266509807, "loss": 0.1534, "step": 5544 }, { "epoch": 0.36, "learning_rate": 0.00014938126562954692, "loss": 0.1528, "step": 5545 }, { "epoch": 0.36, "learning_rate": 0.00014936320646439136, "loss": 0.1294, "step": 5546 }, { "epoch": 0.36, "learning_rate": 0.00014934514517041028, "loss": 0.1607, "step": 5547 }, { "epoch": 0.36, "learning_rate": 0.00014932708174838252, "loss": 0.1648, "step": 5548 }, { "epoch": 0.36, "learning_rate": 0.00014930901619908716, "loss": 0.1767, "step": 5549 }, { "epoch": 0.36, "learning_rate": 0.00014929094852330317, "loss": 0.1549, "step": 5550 }, { "epoch": 0.36, "learning_rate": 0.00014927287872180988, "loss": 0.174, "step": 5551 }, { "epoch": 0.36, "learning_rate": 0.00014925480679538647, "loss": 0.1524, "step": 5552 }, { "epoch": 0.36, "learning_rate": 0.0001492367327448123, "loss": 0.1655, "step": 5553 }, { "epoch": 0.36, "learning_rate": 0.00014921865657086685, "loss": 0.1354, "step": 5554 }, { "epoch": 0.36, "learning_rate": 0.00014920057827432973, "loss": 0.1429, "step": 5555 }, { "epoch": 0.36, "learning_rate": 0.00014918249785598046, "loss": 0.1366, "step": 5556 }, { "epoch": 0.36, "learning_rate": 0.00014916441531659884, "loss": 0.1458, "step": 5557 }, { "epoch": 0.36, "learning_rate": 0.00014914633065696474, "loss": 0.1614, "step": 5558 }, { "epoch": 0.36, "learning_rate": 0.00014912824387785798, "loss": 0.157, "step": 5559 }, { "epoch": 0.36, "learning_rate": 0.00014911015498005859, "loss": 0.1508, "step": 5560 }, { "epoch": 0.36, "learning_rate": 0.0001490920639643467, "loss": 0.1324, "step": 5561 }, { "epoch": 0.36, "learning_rate": 0.00014907397083150248, "loss": 0.1403, "step": 5562 }, { "epoch": 0.36, "learning_rate": 0.0001490558755823062, "loss": 0.1357, "step": 5563 }, { "epoch": 0.36, "learning_rate": 0.0001490377782175383, "loss": 0.1522, "step": 5564 }, { "epoch": 0.36, "learning_rate": 0.00014901967873797914, "loss": 0.1528, "step": 5565 }, { "epoch": 0.36, "learning_rate": 0.0001490015771444093, "loss": 0.1273, "step": 5566 }, { "epoch": 0.36, "learning_rate": 0.0001489834734376095, "loss": 0.1591, "step": 5567 }, { "epoch": 0.36, "learning_rate": 0.0001489653676183604, "loss": 0.1348, "step": 5568 }, { "epoch": 0.36, "learning_rate": 0.00014894725968744285, "loss": 0.1471, "step": 5569 }, { "epoch": 0.36, "learning_rate": 0.00014892914964563774, "loss": 0.1515, "step": 5570 }, { "epoch": 0.36, "learning_rate": 0.00014891103749372614, "loss": 0.1308, "step": 5571 }, { "epoch": 0.36, "learning_rate": 0.0001488929232324891, "loss": 0.1359, "step": 5572 }, { "epoch": 0.36, "learning_rate": 0.00014887480686270782, "loss": 0.1757, "step": 5573 }, { "epoch": 0.36, "learning_rate": 0.00014885668838516362, "loss": 0.1597, "step": 5574 }, { "epoch": 0.36, "learning_rate": 0.00014883856780063782, "loss": 0.1526, "step": 5575 }, { "epoch": 0.36, "learning_rate": 0.0001488204451099119, "loss": 0.1498, "step": 5576 }, { "epoch": 0.36, "learning_rate": 0.00014880232031376744, "loss": 0.1695, "step": 5577 }, { "epoch": 0.36, "learning_rate": 0.00014878419341298607, "loss": 0.121, "step": 5578 }, { "epoch": 0.36, "learning_rate": 0.0001487660644083495, "loss": 0.1744, "step": 5579 }, { "epoch": 0.36, "learning_rate": 0.0001487479333006396, "loss": 0.1439, "step": 5580 }, { "epoch": 0.36, "learning_rate": 0.00014872980009063825, "loss": 0.1637, "step": 5581 }, { "epoch": 0.36, "learning_rate": 0.0001487116647791275, "loss": 0.1489, "step": 5582 }, { "epoch": 0.36, "learning_rate": 0.00014869352736688937, "loss": 0.141, "step": 5583 }, { "epoch": 0.36, "learning_rate": 0.00014867538785470617, "loss": 0.1729, "step": 5584 }, { "epoch": 0.36, "learning_rate": 0.0001486572462433601, "loss": 0.1571, "step": 5585 }, { "epoch": 0.36, "learning_rate": 0.00014863910253363352, "loss": 0.1504, "step": 5586 }, { "epoch": 0.36, "learning_rate": 0.00014862095672630893, "loss": 0.1605, "step": 5587 }, { "epoch": 0.36, "learning_rate": 0.0001486028088221689, "loss": 0.1457, "step": 5588 }, { "epoch": 0.36, "learning_rate": 0.00014858465882199598, "loss": 0.1382, "step": 5589 }, { "epoch": 0.36, "learning_rate": 0.000148566506726573, "loss": 0.1363, "step": 5590 }, { "epoch": 0.36, "learning_rate": 0.00014854835253668274, "loss": 0.1582, "step": 5591 }, { "epoch": 0.36, "learning_rate": 0.00014853019625310813, "loss": 0.1511, "step": 5592 }, { "epoch": 0.36, "learning_rate": 0.00014851203787663212, "loss": 0.1348, "step": 5593 }, { "epoch": 0.36, "learning_rate": 0.00014849387740803788, "loss": 0.1486, "step": 5594 }, { "epoch": 0.36, "learning_rate": 0.00014847571484810857, "loss": 0.1575, "step": 5595 }, { "epoch": 0.36, "learning_rate": 0.00014845755019762743, "loss": 0.146, "step": 5596 }, { "epoch": 0.36, "learning_rate": 0.00014843938345737785, "loss": 0.1532, "step": 5597 }, { "epoch": 0.36, "learning_rate": 0.00014842121462814328, "loss": 0.1608, "step": 5598 }, { "epoch": 0.36, "learning_rate": 0.0001484030437107073, "loss": 0.1499, "step": 5599 }, { "epoch": 0.36, "learning_rate": 0.0001483848707058535, "loss": 0.1572, "step": 5600 }, { "epoch": 0.36, "learning_rate": 0.00014836669561436557, "loss": 0.1432, "step": 5601 }, { "epoch": 0.36, "learning_rate": 0.0001483485184370274, "loss": 0.1539, "step": 5602 }, { "epoch": 0.36, "learning_rate": 0.00014833033917462291, "loss": 0.163, "step": 5603 }, { "epoch": 0.36, "learning_rate": 0.000148312157827936, "loss": 0.154, "step": 5604 }, { "epoch": 0.36, "learning_rate": 0.00014829397439775082, "loss": 0.1511, "step": 5605 }, { "epoch": 0.36, "learning_rate": 0.00014827578888485153, "loss": 0.1462, "step": 5606 }, { "epoch": 0.36, "learning_rate": 0.00014825760129002238, "loss": 0.1503, "step": 5607 }, { "epoch": 0.36, "learning_rate": 0.00014823941161404774, "loss": 0.1513, "step": 5608 }, { "epoch": 0.36, "learning_rate": 0.00014822121985771204, "loss": 0.1408, "step": 5609 }, { "epoch": 0.36, "learning_rate": 0.00014820302602179983, "loss": 0.156, "step": 5610 }, { "epoch": 0.36, "learning_rate": 0.0001481848301070957, "loss": 0.1526, "step": 5611 }, { "epoch": 0.36, "learning_rate": 0.00014816663211438447, "loss": 0.1407, "step": 5612 }, { "epoch": 0.36, "learning_rate": 0.0001481484320444508, "loss": 0.1712, "step": 5613 }, { "epoch": 0.36, "learning_rate": 0.00014813022989807964, "loss": 0.1398, "step": 5614 }, { "epoch": 0.36, "learning_rate": 0.00014811202567605597, "loss": 0.155, "step": 5615 }, { "epoch": 0.36, "learning_rate": 0.00014809381937916488, "loss": 0.1697, "step": 5616 }, { "epoch": 0.36, "learning_rate": 0.0001480756110081915, "loss": 0.1493, "step": 5617 }, { "epoch": 0.36, "learning_rate": 0.0001480574005639211, "loss": 0.147, "step": 5618 }, { "epoch": 0.36, "learning_rate": 0.00014803918804713904, "loss": 0.1437, "step": 5619 }, { "epoch": 0.36, "learning_rate": 0.0001480209734586307, "loss": 0.1505, "step": 5620 }, { "epoch": 0.36, "learning_rate": 0.00014800275679918161, "loss": 0.1571, "step": 5621 }, { "epoch": 0.36, "learning_rate": 0.0001479845380695774, "loss": 0.1574, "step": 5622 }, { "epoch": 0.36, "learning_rate": 0.00014796631727060375, "loss": 0.148, "step": 5623 }, { "epoch": 0.36, "learning_rate": 0.00014794809440304646, "loss": 0.1501, "step": 5624 }, { "epoch": 0.36, "learning_rate": 0.00014792986946769135, "loss": 0.1516, "step": 5625 }, { "epoch": 0.36, "learning_rate": 0.00014791164246532449, "loss": 0.1531, "step": 5626 }, { "epoch": 0.36, "learning_rate": 0.00014789341339673185, "loss": 0.1697, "step": 5627 }, { "epoch": 0.36, "learning_rate": 0.0001478751822626996, "loss": 0.1613, "step": 5628 }, { "epoch": 0.36, "learning_rate": 0.00014785694906401393, "loss": 0.1806, "step": 5629 }, { "epoch": 0.36, "learning_rate": 0.0001478387138014612, "loss": 0.1489, "step": 5630 }, { "epoch": 0.36, "learning_rate": 0.00014782047647582785, "loss": 0.1642, "step": 5631 }, { "epoch": 0.36, "learning_rate": 0.0001478022370879003, "loss": 0.1394, "step": 5632 }, { "epoch": 0.36, "learning_rate": 0.00014778399563846525, "loss": 0.1447, "step": 5633 }, { "epoch": 0.36, "learning_rate": 0.00014776575212830927, "loss": 0.1499, "step": 5634 }, { "epoch": 0.36, "learning_rate": 0.00014774750655821914, "loss": 0.1749, "step": 5635 }, { "epoch": 0.36, "learning_rate": 0.00014772925892898175, "loss": 0.144, "step": 5636 }, { "epoch": 0.36, "learning_rate": 0.00014771100924138402, "loss": 0.161, "step": 5637 }, { "epoch": 0.36, "learning_rate": 0.000147692757496213, "loss": 0.1414, "step": 5638 }, { "epoch": 0.36, "learning_rate": 0.00014767450369425583, "loss": 0.1668, "step": 5639 }, { "epoch": 0.36, "learning_rate": 0.00014765624783629963, "loss": 0.1453, "step": 5640 }, { "epoch": 0.36, "learning_rate": 0.0001476379899231318, "loss": 0.1621, "step": 5641 }, { "epoch": 0.36, "learning_rate": 0.00014761972995553968, "loss": 0.1545, "step": 5642 }, { "epoch": 0.36, "learning_rate": 0.00014760146793431076, "loss": 0.1655, "step": 5643 }, { "epoch": 0.36, "learning_rate": 0.00014758320386023255, "loss": 0.1568, "step": 5644 }, { "epoch": 0.36, "learning_rate": 0.00014756493773409275, "loss": 0.1672, "step": 5645 }, { "epoch": 0.36, "learning_rate": 0.00014754666955667912, "loss": 0.139, "step": 5646 }, { "epoch": 0.36, "learning_rate": 0.00014752839932877948, "loss": 0.1489, "step": 5647 }, { "epoch": 0.36, "learning_rate": 0.00014751012705118172, "loss": 0.1432, "step": 5648 }, { "epoch": 0.36, "learning_rate": 0.00014749185272467386, "loss": 0.1755, "step": 5649 }, { "epoch": 0.36, "learning_rate": 0.00014747357635004396, "loss": 0.1747, "step": 5650 }, { "epoch": 0.36, "learning_rate": 0.00014745529792808024, "loss": 0.1596, "step": 5651 }, { "epoch": 0.36, "learning_rate": 0.000147437017459571, "loss": 0.1499, "step": 5652 }, { "epoch": 0.36, "learning_rate": 0.0001474187349453045, "loss": 0.1405, "step": 5653 }, { "epoch": 0.36, "learning_rate": 0.00014740045038606932, "loss": 0.1449, "step": 5654 }, { "epoch": 0.36, "learning_rate": 0.00014738216378265388, "loss": 0.1556, "step": 5655 }, { "epoch": 0.36, "learning_rate": 0.00014736387513584686, "loss": 0.2033, "step": 5656 }, { "epoch": 0.36, "learning_rate": 0.00014734558444643695, "loss": 0.153, "step": 5657 }, { "epoch": 0.36, "learning_rate": 0.000147327291715213, "loss": 0.1274, "step": 5658 }, { "epoch": 0.36, "learning_rate": 0.00014730899694296383, "loss": 0.1578, "step": 5659 }, { "epoch": 0.36, "learning_rate": 0.00014729070013047845, "loss": 0.1313, "step": 5660 }, { "epoch": 0.36, "learning_rate": 0.00014727240127854592, "loss": 0.1549, "step": 5661 }, { "epoch": 0.36, "learning_rate": 0.0001472541003879554, "loss": 0.1518, "step": 5662 }, { "epoch": 0.36, "learning_rate": 0.0001472357974594961, "loss": 0.1608, "step": 5663 }, { "epoch": 0.36, "learning_rate": 0.00014721749249395737, "loss": 0.1542, "step": 5664 }, { "epoch": 0.36, "learning_rate": 0.00014719918549212864, "loss": 0.1827, "step": 5665 }, { "epoch": 0.36, "learning_rate": 0.00014718087645479937, "loss": 0.1422, "step": 5666 }, { "epoch": 0.36, "learning_rate": 0.00014716256538275922, "loss": 0.1675, "step": 5667 }, { "epoch": 0.36, "learning_rate": 0.00014714425227679777, "loss": 0.1603, "step": 5668 }, { "epoch": 0.36, "learning_rate": 0.00014712593713770488, "loss": 0.1301, "step": 5669 }, { "epoch": 0.36, "learning_rate": 0.00014710761996627035, "loss": 0.1569, "step": 5670 }, { "epoch": 0.36, "learning_rate": 0.00014708930076328417, "loss": 0.1376, "step": 5671 }, { "epoch": 0.36, "learning_rate": 0.0001470709795295363, "loss": 0.1735, "step": 5672 }, { "epoch": 0.36, "learning_rate": 0.0001470526562658169, "loss": 0.1617, "step": 5673 }, { "epoch": 0.36, "learning_rate": 0.0001470343309729162, "loss": 0.141, "step": 5674 }, { "epoch": 0.36, "learning_rate": 0.00014701600365162442, "loss": 0.1583, "step": 5675 }, { "epoch": 0.36, "learning_rate": 0.000146997674302732, "loss": 0.1602, "step": 5676 }, { "epoch": 0.36, "learning_rate": 0.0001469793429270294, "loss": 0.1468, "step": 5677 }, { "epoch": 0.36, "learning_rate": 0.00014696100952530717, "loss": 0.1425, "step": 5678 }, { "epoch": 0.36, "learning_rate": 0.00014694267409835591, "loss": 0.1736, "step": 5679 }, { "epoch": 0.36, "learning_rate": 0.0001469243366469664, "loss": 0.1448, "step": 5680 }, { "epoch": 0.36, "learning_rate": 0.00014690599717192946, "loss": 0.1202, "step": 5681 }, { "epoch": 0.36, "learning_rate": 0.00014688765567403599, "loss": 0.1438, "step": 5682 }, { "epoch": 0.36, "learning_rate": 0.0001468693121540769, "loss": 0.1447, "step": 5683 }, { "epoch": 0.36, "learning_rate": 0.00014685096661284336, "loss": 0.1508, "step": 5684 }, { "epoch": 0.36, "learning_rate": 0.00014683261905112653, "loss": 0.1609, "step": 5685 }, { "epoch": 0.36, "learning_rate": 0.0001468142694697176, "loss": 0.1735, "step": 5686 }, { "epoch": 0.36, "learning_rate": 0.000146795917869408, "loss": 0.1598, "step": 5687 }, { "epoch": 0.36, "learning_rate": 0.00014677756425098907, "loss": 0.1663, "step": 5688 }, { "epoch": 0.36, "learning_rate": 0.00014675920861525237, "loss": 0.1716, "step": 5689 }, { "epoch": 0.36, "learning_rate": 0.0001467408509629895, "loss": 0.1676, "step": 5690 }, { "epoch": 0.36, "learning_rate": 0.00014672249129499214, "loss": 0.159, "step": 5691 }, { "epoch": 0.36, "learning_rate": 0.00014670412961205207, "loss": 0.1542, "step": 5692 }, { "epoch": 0.37, "learning_rate": 0.00014668576591496112, "loss": 0.1385, "step": 5693 }, { "epoch": 0.37, "learning_rate": 0.00014666740020451128, "loss": 0.1573, "step": 5694 }, { "epoch": 0.37, "learning_rate": 0.00014664903248149457, "loss": 0.1539, "step": 5695 }, { "epoch": 0.37, "learning_rate": 0.0001466306627467031, "loss": 0.1493, "step": 5696 }, { "epoch": 0.37, "learning_rate": 0.00014661229100092907, "loss": 0.1744, "step": 5697 }, { "epoch": 0.37, "learning_rate": 0.00014659391724496486, "loss": 0.1428, "step": 5698 }, { "epoch": 0.37, "learning_rate": 0.00014657554147960272, "loss": 0.1511, "step": 5699 }, { "epoch": 0.37, "learning_rate": 0.00014655716370563522, "loss": 0.1506, "step": 5700 }, { "epoch": 0.37, "learning_rate": 0.00014653878392385488, "loss": 0.17, "step": 5701 }, { "epoch": 0.37, "learning_rate": 0.00014652040213505434, "loss": 0.1496, "step": 5702 }, { "epoch": 0.37, "learning_rate": 0.00014650201834002632, "loss": 0.1787, "step": 5703 }, { "epoch": 0.37, "learning_rate": 0.00014648363253956363, "loss": 0.1543, "step": 5704 }, { "epoch": 0.37, "learning_rate": 0.0001464652447344592, "loss": 0.1478, "step": 5705 }, { "epoch": 0.37, "learning_rate": 0.00014644685492550604, "loss": 0.1515, "step": 5706 }, { "epoch": 0.37, "learning_rate": 0.00014642846311349716, "loss": 0.1683, "step": 5707 }, { "epoch": 0.37, "learning_rate": 0.00014641006929922576, "loss": 0.1713, "step": 5708 }, { "epoch": 0.37, "learning_rate": 0.00014639167348348505, "loss": 0.1273, "step": 5709 }, { "epoch": 0.37, "learning_rate": 0.00014637327566706845, "loss": 0.1662, "step": 5710 }, { "epoch": 0.37, "learning_rate": 0.00014635487585076926, "loss": 0.1564, "step": 5711 }, { "epoch": 0.37, "learning_rate": 0.00014633647403538107, "loss": 0.1545, "step": 5712 }, { "epoch": 0.37, "learning_rate": 0.00014631807022169745, "loss": 0.1417, "step": 5713 }, { "epoch": 0.37, "learning_rate": 0.00014629966441051208, "loss": 0.1623, "step": 5714 }, { "epoch": 0.37, "learning_rate": 0.00014628125660261874, "loss": 0.1494, "step": 5715 }, { "epoch": 0.37, "learning_rate": 0.00014626284679881125, "loss": 0.1432, "step": 5716 }, { "epoch": 0.37, "learning_rate": 0.00014624443499988354, "loss": 0.154, "step": 5717 }, { "epoch": 0.37, "learning_rate": 0.00014622602120662967, "loss": 0.1821, "step": 5718 }, { "epoch": 0.37, "learning_rate": 0.00014620760541984374, "loss": 0.156, "step": 5719 }, { "epoch": 0.37, "learning_rate": 0.00014618918764031991, "loss": 0.1719, "step": 5720 }, { "epoch": 0.37, "learning_rate": 0.00014617076786885248, "loss": 0.1361, "step": 5721 }, { "epoch": 0.37, "learning_rate": 0.00014615234610623588, "loss": 0.1328, "step": 5722 }, { "epoch": 0.37, "learning_rate": 0.00014613392235326446, "loss": 0.1719, "step": 5723 }, { "epoch": 0.37, "learning_rate": 0.0001461154966107328, "loss": 0.1483, "step": 5724 }, { "epoch": 0.37, "learning_rate": 0.00014609706887943553, "loss": 0.1853, "step": 5725 }, { "epoch": 0.37, "learning_rate": 0.00014607863916016737, "loss": 0.1416, "step": 5726 }, { "epoch": 0.37, "learning_rate": 0.0001460602074537231, "loss": 0.1669, "step": 5727 }, { "epoch": 0.37, "learning_rate": 0.00014604177376089763, "loss": 0.1637, "step": 5728 }, { "epoch": 0.37, "learning_rate": 0.00014602333808248584, "loss": 0.1498, "step": 5729 }, { "epoch": 0.37, "learning_rate": 0.0001460049004192829, "loss": 0.1656, "step": 5730 }, { "epoch": 0.37, "learning_rate": 0.00014598646077208387, "loss": 0.1644, "step": 5731 }, { "epoch": 0.37, "learning_rate": 0.000145968019141684, "loss": 0.1477, "step": 5732 }, { "epoch": 0.37, "learning_rate": 0.00014594957552887858, "loss": 0.1401, "step": 5733 }, { "epoch": 0.37, "learning_rate": 0.000145931129934463, "loss": 0.1503, "step": 5734 }, { "epoch": 0.37, "learning_rate": 0.00014591268235923283, "loss": 0.1636, "step": 5735 }, { "epoch": 0.37, "learning_rate": 0.00014589423280398354, "loss": 0.1528, "step": 5736 }, { "epoch": 0.37, "learning_rate": 0.00014587578126951076, "loss": 0.1331, "step": 5737 }, { "epoch": 0.37, "learning_rate": 0.0001458573277566103, "loss": 0.1315, "step": 5738 }, { "epoch": 0.37, "learning_rate": 0.00014583887226607797, "loss": 0.144, "step": 5739 }, { "epoch": 0.37, "learning_rate": 0.00014582041479870966, "loss": 0.1613, "step": 5740 }, { "epoch": 0.37, "learning_rate": 0.00014580195535530133, "loss": 0.1521, "step": 5741 }, { "epoch": 0.37, "learning_rate": 0.00014578349393664915, "loss": 0.1617, "step": 5742 }, { "epoch": 0.37, "learning_rate": 0.00014576503054354917, "loss": 0.1261, "step": 5743 }, { "epoch": 0.37, "learning_rate": 0.0001457465651767977, "loss": 0.1596, "step": 5744 }, { "epoch": 0.37, "learning_rate": 0.0001457280978371911, "loss": 0.1405, "step": 5745 }, { "epoch": 0.37, "learning_rate": 0.00014570962852552573, "loss": 0.1429, "step": 5746 }, { "epoch": 0.37, "learning_rate": 0.0001456911572425981, "loss": 0.1496, "step": 5747 }, { "epoch": 0.37, "learning_rate": 0.0001456726839892048, "loss": 0.1535, "step": 5748 }, { "epoch": 0.37, "learning_rate": 0.00014565420876614258, "loss": 0.1466, "step": 5749 }, { "epoch": 0.37, "learning_rate": 0.00014563573157420808, "loss": 0.1556, "step": 5750 }, { "epoch": 0.37, "learning_rate": 0.00014561725241419822, "loss": 0.1422, "step": 5751 }, { "epoch": 0.37, "learning_rate": 0.0001455987712869099, "loss": 0.1549, "step": 5752 }, { "epoch": 0.37, "learning_rate": 0.00014558028819314011, "loss": 0.1703, "step": 5753 }, { "epoch": 0.37, "learning_rate": 0.000145561803133686, "loss": 0.1275, "step": 5754 }, { "epoch": 0.37, "learning_rate": 0.00014554331610934474, "loss": 0.1435, "step": 5755 }, { "epoch": 0.37, "learning_rate": 0.00014552482712091357, "loss": 0.1503, "step": 5756 }, { "epoch": 0.37, "learning_rate": 0.00014550633616918984, "loss": 0.1398, "step": 5757 }, { "epoch": 0.37, "learning_rate": 0.00014548784325497103, "loss": 0.1409, "step": 5758 }, { "epoch": 0.37, "learning_rate": 0.00014546934837905462, "loss": 0.1478, "step": 5759 }, { "epoch": 0.37, "learning_rate": 0.00014545085154223824, "loss": 0.1526, "step": 5760 }, { "epoch": 0.37, "learning_rate": 0.00014543235274531955, "loss": 0.1661, "step": 5761 }, { "epoch": 0.37, "learning_rate": 0.00014541385198909635, "loss": 0.1627, "step": 5762 }, { "epoch": 0.37, "learning_rate": 0.0001453953492743665, "loss": 0.1463, "step": 5763 }, { "epoch": 0.37, "learning_rate": 0.0001453768446019279, "loss": 0.1574, "step": 5764 }, { "epoch": 0.37, "learning_rate": 0.00014535833797257866, "loss": 0.154, "step": 5765 }, { "epoch": 0.37, "learning_rate": 0.00014533982938711682, "loss": 0.1627, "step": 5766 }, { "epoch": 0.37, "learning_rate": 0.00014532131884634062, "loss": 0.1454, "step": 5767 }, { "epoch": 0.37, "learning_rate": 0.00014530280635104828, "loss": 0.1452, "step": 5768 }, { "epoch": 0.37, "learning_rate": 0.00014528429190203824, "loss": 0.1427, "step": 5769 }, { "epoch": 0.37, "learning_rate": 0.00014526577550010893, "loss": 0.196, "step": 5770 }, { "epoch": 0.37, "learning_rate": 0.00014524725714605885, "loss": 0.1637, "step": 5771 }, { "epoch": 0.37, "learning_rate": 0.00014522873684068665, "loss": 0.1409, "step": 5772 }, { "epoch": 0.37, "learning_rate": 0.00014521021458479102, "loss": 0.1465, "step": 5773 }, { "epoch": 0.37, "learning_rate": 0.00014519169037917074, "loss": 0.1332, "step": 5774 }, { "epoch": 0.37, "learning_rate": 0.0001451731642246247, "loss": 0.1353, "step": 5775 }, { "epoch": 0.37, "learning_rate": 0.0001451546361219519, "loss": 0.1508, "step": 5776 }, { "epoch": 0.37, "learning_rate": 0.00014513610607195122, "loss": 0.1386, "step": 5777 }, { "epoch": 0.37, "learning_rate": 0.00014511757407542194, "loss": 0.1424, "step": 5778 }, { "epoch": 0.37, "learning_rate": 0.0001450990401331632, "loss": 0.1549, "step": 5779 }, { "epoch": 0.37, "learning_rate": 0.0001450805042459743, "loss": 0.1653, "step": 5780 }, { "epoch": 0.37, "learning_rate": 0.00014506196641465459, "loss": 0.1548, "step": 5781 }, { "epoch": 0.37, "learning_rate": 0.0001450434266400036, "loss": 0.173, "step": 5782 }, { "epoch": 0.37, "learning_rate": 0.0001450248849228208, "loss": 0.1597, "step": 5783 }, { "epoch": 0.37, "learning_rate": 0.00014500634126390585, "loss": 0.1592, "step": 5784 }, { "epoch": 0.37, "learning_rate": 0.00014498779566405844, "loss": 0.1498, "step": 5785 }, { "epoch": 0.37, "learning_rate": 0.00014496924812407836, "loss": 0.1586, "step": 5786 }, { "epoch": 0.37, "learning_rate": 0.00014495069864476552, "loss": 0.143, "step": 5787 }, { "epoch": 0.37, "learning_rate": 0.00014493214722691982, "loss": 0.1539, "step": 5788 }, { "epoch": 0.37, "learning_rate": 0.0001449135938713414, "loss": 0.1504, "step": 5789 }, { "epoch": 0.37, "learning_rate": 0.00014489503857883034, "loss": 0.1363, "step": 5790 }, { "epoch": 0.37, "learning_rate": 0.0001448764813501868, "loss": 0.1647, "step": 5791 }, { "epoch": 0.37, "learning_rate": 0.0001448579221862111, "loss": 0.1434, "step": 5792 }, { "epoch": 0.37, "learning_rate": 0.0001448393610877037, "loss": 0.1323, "step": 5793 }, { "epoch": 0.37, "learning_rate": 0.00014482079805546493, "loss": 0.1727, "step": 5794 }, { "epoch": 0.37, "learning_rate": 0.0001448022330902954, "loss": 0.1619, "step": 5795 }, { "epoch": 0.37, "learning_rate": 0.00014478366619299576, "loss": 0.1521, "step": 5796 }, { "epoch": 0.37, "learning_rate": 0.00014476509736436675, "loss": 0.1504, "step": 5797 }, { "epoch": 0.37, "learning_rate": 0.00014474652660520905, "loss": 0.1336, "step": 5798 }, { "epoch": 0.37, "learning_rate": 0.0001447279539163236, "loss": 0.1547, "step": 5799 }, { "epoch": 0.37, "learning_rate": 0.0001447093792985114, "loss": 0.1538, "step": 5800 }, { "epoch": 0.37, "learning_rate": 0.00014469080275257342, "loss": 0.1792, "step": 5801 }, { "epoch": 0.37, "learning_rate": 0.00014467222427931088, "loss": 0.1435, "step": 5802 }, { "epoch": 0.37, "learning_rate": 0.00014465364387952488, "loss": 0.1288, "step": 5803 }, { "epoch": 0.37, "learning_rate": 0.0001446350615540168, "loss": 0.14, "step": 5804 }, { "epoch": 0.37, "learning_rate": 0.000144616477303588, "loss": 0.1482, "step": 5805 }, { "epoch": 0.37, "learning_rate": 0.00014459789112903992, "loss": 0.1671, "step": 5806 }, { "epoch": 0.37, "learning_rate": 0.00014457930303117411, "loss": 0.1519, "step": 5807 }, { "epoch": 0.37, "learning_rate": 0.00014456071301079216, "loss": 0.1614, "step": 5808 }, { "epoch": 0.37, "learning_rate": 0.00014454212106869584, "loss": 0.1443, "step": 5809 }, { "epoch": 0.37, "learning_rate": 0.00014452352720568693, "loss": 0.1539, "step": 5810 }, { "epoch": 0.37, "learning_rate": 0.00014450493142256727, "loss": 0.1547, "step": 5811 }, { "epoch": 0.37, "learning_rate": 0.00014448633372013886, "loss": 0.1511, "step": 5812 }, { "epoch": 0.37, "learning_rate": 0.00014446773409920372, "loss": 0.1675, "step": 5813 }, { "epoch": 0.37, "learning_rate": 0.00014444913256056392, "loss": 0.1526, "step": 5814 }, { "epoch": 0.37, "learning_rate": 0.00014443052910502176, "loss": 0.159, "step": 5815 }, { "epoch": 0.37, "learning_rate": 0.00014441192373337947, "loss": 0.1419, "step": 5816 }, { "epoch": 0.37, "learning_rate": 0.00014439331644643946, "loss": 0.1633, "step": 5817 }, { "epoch": 0.37, "learning_rate": 0.0001443747072450041, "loss": 0.1982, "step": 5818 }, { "epoch": 0.37, "learning_rate": 0.00014435609612987603, "loss": 0.1655, "step": 5819 }, { "epoch": 0.37, "learning_rate": 0.0001443374831018578, "loss": 0.1476, "step": 5820 }, { "epoch": 0.37, "learning_rate": 0.00014431886816175212, "loss": 0.1417, "step": 5821 }, { "epoch": 0.37, "learning_rate": 0.00014430025131036183, "loss": 0.152, "step": 5822 }, { "epoch": 0.37, "learning_rate": 0.00014428163254848975, "loss": 0.1681, "step": 5823 }, { "epoch": 0.37, "learning_rate": 0.0001442630118769388, "loss": 0.1488, "step": 5824 }, { "epoch": 0.37, "learning_rate": 0.00014424438929651204, "loss": 0.1562, "step": 5825 }, { "epoch": 0.37, "learning_rate": 0.0001442257648080126, "loss": 0.1534, "step": 5826 }, { "epoch": 0.37, "learning_rate": 0.00014420713841224365, "loss": 0.1503, "step": 5827 }, { "epoch": 0.37, "learning_rate": 0.00014418851011000847, "loss": 0.1318, "step": 5828 }, { "epoch": 0.37, "learning_rate": 0.00014416987990211044, "loss": 0.157, "step": 5829 }, { "epoch": 0.37, "learning_rate": 0.00014415124778935297, "loss": 0.1543, "step": 5830 }, { "epoch": 0.37, "learning_rate": 0.00014413261377253962, "loss": 0.1652, "step": 5831 }, { "epoch": 0.37, "learning_rate": 0.00014411397785247398, "loss": 0.1382, "step": 5832 }, { "epoch": 0.37, "learning_rate": 0.00014409534002995974, "loss": 0.142, "step": 5833 }, { "epoch": 0.37, "learning_rate": 0.00014407670030580066, "loss": 0.14, "step": 5834 }, { "epoch": 0.37, "learning_rate": 0.00014405805868080056, "loss": 0.1451, "step": 5835 }, { "epoch": 0.37, "learning_rate": 0.00014403941515576344, "loss": 0.1624, "step": 5836 }, { "epoch": 0.37, "learning_rate": 0.0001440207697314933, "loss": 0.1482, "step": 5837 }, { "epoch": 0.37, "learning_rate": 0.0001440021224087942, "loss": 0.1529, "step": 5838 }, { "epoch": 0.37, "learning_rate": 0.00014398347318847037, "loss": 0.157, "step": 5839 }, { "epoch": 0.37, "learning_rate": 0.00014396482207132606, "loss": 0.1347, "step": 5840 }, { "epoch": 0.37, "learning_rate": 0.00014394616905816555, "loss": 0.1417, "step": 5841 }, { "epoch": 0.37, "learning_rate": 0.00014392751414979332, "loss": 0.1489, "step": 5842 }, { "epoch": 0.37, "learning_rate": 0.0001439088573470139, "loss": 0.1642, "step": 5843 }, { "epoch": 0.37, "learning_rate": 0.00014389019865063187, "loss": 0.1572, "step": 5844 }, { "epoch": 0.37, "learning_rate": 0.00014387153806145187, "loss": 0.1621, "step": 5845 }, { "epoch": 0.37, "learning_rate": 0.00014385287558027864, "loss": 0.1744, "step": 5846 }, { "epoch": 0.37, "learning_rate": 0.00014383421120791706, "loss": 0.1649, "step": 5847 }, { "epoch": 0.37, "learning_rate": 0.000143815544945172, "loss": 0.1799, "step": 5848 }, { "epoch": 0.38, "learning_rate": 0.00014379687679284852, "loss": 0.127, "step": 5849 }, { "epoch": 0.38, "learning_rate": 0.00014377820675175166, "loss": 0.1677, "step": 5850 }, { "epoch": 0.38, "learning_rate": 0.00014375953482268655, "loss": 0.15, "step": 5851 }, { "epoch": 0.38, "learning_rate": 0.00014374086100645846, "loss": 0.1608, "step": 5852 }, { "epoch": 0.38, "learning_rate": 0.00014372218530387271, "loss": 0.1693, "step": 5853 }, { "epoch": 0.38, "learning_rate": 0.0001437035077157347, "loss": 0.1376, "step": 5854 }, { "epoch": 0.38, "learning_rate": 0.00014368482824284992, "loss": 0.1682, "step": 5855 }, { "epoch": 0.38, "learning_rate": 0.00014366614688602396, "loss": 0.1486, "step": 5856 }, { "epoch": 0.38, "learning_rate": 0.00014364746364606244, "loss": 0.1493, "step": 5857 }, { "epoch": 0.38, "learning_rate": 0.00014362877852377106, "loss": 0.1641, "step": 5858 }, { "epoch": 0.38, "learning_rate": 0.0001436100915199557, "loss": 0.1489, "step": 5859 }, { "epoch": 0.38, "learning_rate": 0.0001435914026354222, "loss": 0.1566, "step": 5860 }, { "epoch": 0.38, "learning_rate": 0.00014357271187097654, "loss": 0.1627, "step": 5861 }, { "epoch": 0.38, "learning_rate": 0.00014355401922742478, "loss": 0.1529, "step": 5862 }, { "epoch": 0.38, "learning_rate": 0.00014353532470557306, "loss": 0.1569, "step": 5863 }, { "epoch": 0.38, "learning_rate": 0.00014351662830622757, "loss": 0.1352, "step": 5864 }, { "epoch": 0.38, "learning_rate": 0.00014349793003019465, "loss": 0.1384, "step": 5865 }, { "epoch": 0.38, "learning_rate": 0.00014347922987828065, "loss": 0.1846, "step": 5866 }, { "epoch": 0.38, "learning_rate": 0.00014346052785129204, "loss": 0.1576, "step": 5867 }, { "epoch": 0.38, "learning_rate": 0.00014344182395003533, "loss": 0.1523, "step": 5868 }, { "epoch": 0.38, "learning_rate": 0.00014342311817531716, "loss": 0.1529, "step": 5869 }, { "epoch": 0.38, "learning_rate": 0.0001434044105279442, "loss": 0.1469, "step": 5870 }, { "epoch": 0.38, "learning_rate": 0.00014338570100872334, "loss": 0.1764, "step": 5871 }, { "epoch": 0.38, "learning_rate": 0.00014336698961846134, "loss": 0.1616, "step": 5872 }, { "epoch": 0.38, "learning_rate": 0.00014334827635796511, "loss": 0.152, "step": 5873 }, { "epoch": 0.38, "learning_rate": 0.0001433295612280418, "loss": 0.1653, "step": 5874 }, { "epoch": 0.38, "learning_rate": 0.0001433108442294984, "loss": 0.1549, "step": 5875 }, { "epoch": 0.38, "learning_rate": 0.00014329212536314217, "loss": 0.1411, "step": 5876 }, { "epoch": 0.38, "learning_rate": 0.0001432734046297803, "loss": 0.1451, "step": 5877 }, { "epoch": 0.38, "learning_rate": 0.00014325468203022024, "loss": 0.1748, "step": 5878 }, { "epoch": 0.38, "learning_rate": 0.00014323595756526934, "loss": 0.1688, "step": 5879 }, { "epoch": 0.38, "learning_rate": 0.0001432172312357351, "loss": 0.1487, "step": 5880 }, { "epoch": 0.38, "learning_rate": 0.00014319850304242516, "loss": 0.1481, "step": 5881 }, { "epoch": 0.38, "learning_rate": 0.0001431797729861471, "loss": 0.1471, "step": 5882 }, { "epoch": 0.38, "learning_rate": 0.00014316104106770874, "loss": 0.1578, "step": 5883 }, { "epoch": 0.38, "learning_rate": 0.00014314230728791795, "loss": 0.1655, "step": 5884 }, { "epoch": 0.38, "learning_rate": 0.00014312357164758253, "loss": 0.1633, "step": 5885 }, { "epoch": 0.38, "learning_rate": 0.00014310483414751058, "loss": 0.1608, "step": 5886 }, { "epoch": 0.38, "learning_rate": 0.00014308609478851008, "loss": 0.1775, "step": 5887 }, { "epoch": 0.38, "learning_rate": 0.00014306735357138916, "loss": 0.1501, "step": 5888 }, { "epoch": 0.38, "learning_rate": 0.00014304861049695616, "loss": 0.1457, "step": 5889 }, { "epoch": 0.38, "learning_rate": 0.0001430298655660193, "loss": 0.1659, "step": 5890 }, { "epoch": 0.38, "learning_rate": 0.000143011118779387, "loss": 0.153, "step": 5891 }, { "epoch": 0.38, "learning_rate": 0.00014299237013786772, "loss": 0.1352, "step": 5892 }, { "epoch": 0.38, "learning_rate": 0.00014297361964227002, "loss": 0.1573, "step": 5893 }, { "epoch": 0.38, "learning_rate": 0.00014295486729340253, "loss": 0.1564, "step": 5894 }, { "epoch": 0.38, "learning_rate": 0.00014293611309207393, "loss": 0.1701, "step": 5895 }, { "epoch": 0.38, "learning_rate": 0.00014291735703909306, "loss": 0.1512, "step": 5896 }, { "epoch": 0.38, "learning_rate": 0.00014289859913526874, "loss": 0.1506, "step": 5897 }, { "epoch": 0.38, "learning_rate": 0.00014287983938140997, "loss": 0.1577, "step": 5898 }, { "epoch": 0.38, "learning_rate": 0.00014286107777832573, "loss": 0.1395, "step": 5899 }, { "epoch": 0.38, "learning_rate": 0.00014284231432682516, "loss": 0.1522, "step": 5900 }, { "epoch": 0.38, "learning_rate": 0.00014282354902771742, "loss": 0.1441, "step": 5901 }, { "epoch": 0.38, "learning_rate": 0.00014280478188181177, "loss": 0.1544, "step": 5902 }, { "epoch": 0.38, "learning_rate": 0.00014278601288991762, "loss": 0.1442, "step": 5903 }, { "epoch": 0.38, "learning_rate": 0.00014276724205284434, "loss": 0.1604, "step": 5904 }, { "epoch": 0.38, "learning_rate": 0.00014274846937140144, "loss": 0.1595, "step": 5905 }, { "epoch": 0.38, "learning_rate": 0.00014272969484639857, "loss": 0.1305, "step": 5906 }, { "epoch": 0.38, "learning_rate": 0.0001427109184786453, "loss": 0.1439, "step": 5907 }, { "epoch": 0.38, "learning_rate": 0.00014269214026895144, "loss": 0.1482, "step": 5908 }, { "epoch": 0.38, "learning_rate": 0.0001426733602181268, "loss": 0.1617, "step": 5909 }, { "epoch": 0.38, "learning_rate": 0.00014265457832698125, "loss": 0.1501, "step": 5910 }, { "epoch": 0.38, "learning_rate": 0.0001426357945963248, "loss": 0.1625, "step": 5911 }, { "epoch": 0.38, "learning_rate": 0.00014261700902696753, "loss": 0.1691, "step": 5912 }, { "epoch": 0.38, "learning_rate": 0.0001425982216197196, "loss": 0.1478, "step": 5913 }, { "epoch": 0.38, "learning_rate": 0.00014257943237539118, "loss": 0.1466, "step": 5914 }, { "epoch": 0.38, "learning_rate": 0.00014256064129479254, "loss": 0.1381, "step": 5915 }, { "epoch": 0.38, "learning_rate": 0.00014254184837873414, "loss": 0.1587, "step": 5916 }, { "epoch": 0.38, "learning_rate": 0.00014252305362802642, "loss": 0.151, "step": 5917 }, { "epoch": 0.38, "learning_rate": 0.0001425042570434799, "loss": 0.1466, "step": 5918 }, { "epoch": 0.38, "learning_rate": 0.00014248545862590516, "loss": 0.1437, "step": 5919 }, { "epoch": 0.38, "learning_rate": 0.000142466658376113, "loss": 0.1302, "step": 5920 }, { "epoch": 0.38, "learning_rate": 0.0001424478562949141, "loss": 0.1468, "step": 5921 }, { "epoch": 0.38, "learning_rate": 0.00014242905238311934, "loss": 0.1563, "step": 5922 }, { "epoch": 0.38, "learning_rate": 0.00014241024664153967, "loss": 0.151, "step": 5923 }, { "epoch": 0.38, "learning_rate": 0.0001423914390709861, "loss": 0.1253, "step": 5924 }, { "epoch": 0.38, "learning_rate": 0.00014237262967226969, "loss": 0.1443, "step": 5925 }, { "epoch": 0.38, "learning_rate": 0.00014235381844620164, "loss": 0.1489, "step": 5926 }, { "epoch": 0.38, "learning_rate": 0.0001423350053935932, "loss": 0.1386, "step": 5927 }, { "epoch": 0.38, "learning_rate": 0.00014231619051525564, "loss": 0.1719, "step": 5928 }, { "epoch": 0.38, "learning_rate": 0.00014229737381200045, "loss": 0.1465, "step": 5929 }, { "epoch": 0.38, "learning_rate": 0.00014227855528463907, "loss": 0.1463, "step": 5930 }, { "epoch": 0.38, "learning_rate": 0.00014225973493398305, "loss": 0.162, "step": 5931 }, { "epoch": 0.38, "learning_rate": 0.00014224091276084406, "loss": 0.1516, "step": 5932 }, { "epoch": 0.38, "learning_rate": 0.00014222208876603384, "loss": 0.1492, "step": 5933 }, { "epoch": 0.38, "learning_rate": 0.00014220326295036415, "loss": 0.1362, "step": 5934 }, { "epoch": 0.38, "learning_rate": 0.00014218443531464686, "loss": 0.1332, "step": 5935 }, { "epoch": 0.38, "learning_rate": 0.00014216560585969396, "loss": 0.1398, "step": 5936 }, { "epoch": 0.38, "learning_rate": 0.00014214677458631745, "loss": 0.1556, "step": 5937 }, { "epoch": 0.38, "learning_rate": 0.0001421279414953295, "loss": 0.1476, "step": 5938 }, { "epoch": 0.38, "learning_rate": 0.00014210910658754222, "loss": 0.1308, "step": 5939 }, { "epoch": 0.38, "learning_rate": 0.00014209026986376796, "loss": 0.1442, "step": 5940 }, { "epoch": 0.38, "learning_rate": 0.000142071431324819, "loss": 0.1522, "step": 5941 }, { "epoch": 0.38, "learning_rate": 0.0001420525909715078, "loss": 0.1415, "step": 5942 }, { "epoch": 0.38, "learning_rate": 0.00014203374880464688, "loss": 0.1902, "step": 5943 }, { "epoch": 0.38, "learning_rate": 0.0001420149048250488, "loss": 0.1541, "step": 5944 }, { "epoch": 0.38, "learning_rate": 0.00014199605903352622, "loss": 0.1394, "step": 5945 }, { "epoch": 0.38, "learning_rate": 0.00014197721143089186, "loss": 0.1406, "step": 5946 }, { "epoch": 0.38, "learning_rate": 0.0001419583620179586, "loss": 0.1368, "step": 5947 }, { "epoch": 0.38, "learning_rate": 0.0001419395107955393, "loss": 0.1571, "step": 5948 }, { "epoch": 0.38, "learning_rate": 0.00014192065776444688, "loss": 0.154, "step": 5949 }, { "epoch": 0.38, "learning_rate": 0.00014190180292549445, "loss": 0.1714, "step": 5950 }, { "epoch": 0.38, "learning_rate": 0.00014188294627949516, "loss": 0.1554, "step": 5951 }, { "epoch": 0.38, "learning_rate": 0.00014186408782726218, "loss": 0.1594, "step": 5952 }, { "epoch": 0.38, "learning_rate": 0.00014184522756960878, "loss": 0.1677, "step": 5953 }, { "epoch": 0.38, "learning_rate": 0.0001418263655073484, "loss": 0.1337, "step": 5954 }, { "epoch": 0.38, "learning_rate": 0.00014180750164129435, "loss": 0.1474, "step": 5955 }, { "epoch": 0.38, "learning_rate": 0.00014178863597226027, "loss": 0.174, "step": 5956 }, { "epoch": 0.38, "learning_rate": 0.0001417697685010597, "loss": 0.1692, "step": 5957 }, { "epoch": 0.38, "learning_rate": 0.00014175089922850633, "loss": 0.1396, "step": 5958 }, { "epoch": 0.38, "learning_rate": 0.0001417320281554139, "loss": 0.1767, "step": 5959 }, { "epoch": 0.38, "learning_rate": 0.00014171315528259623, "loss": 0.1513, "step": 5960 }, { "epoch": 0.38, "learning_rate": 0.00014169428061086727, "loss": 0.1522, "step": 5961 }, { "epoch": 0.38, "learning_rate": 0.00014167540414104093, "loss": 0.1559, "step": 5962 }, { "epoch": 0.38, "learning_rate": 0.00014165652587393136, "loss": 0.1648, "step": 5963 }, { "epoch": 0.38, "learning_rate": 0.00014163764581035266, "loss": 0.1333, "step": 5964 }, { "epoch": 0.38, "learning_rate": 0.000141618763951119, "loss": 0.1715, "step": 5965 }, { "epoch": 0.38, "learning_rate": 0.00014159988029704477, "loss": 0.1648, "step": 5966 }, { "epoch": 0.38, "learning_rate": 0.00014158099484894428, "loss": 0.1602, "step": 5967 }, { "epoch": 0.38, "learning_rate": 0.00014156210760763198, "loss": 0.1476, "step": 5968 }, { "epoch": 0.38, "learning_rate": 0.00014154321857392239, "loss": 0.1426, "step": 5969 }, { "epoch": 0.38, "learning_rate": 0.00014152432774863015, "loss": 0.158, "step": 5970 }, { "epoch": 0.38, "learning_rate": 0.00014150543513256992, "loss": 0.1658, "step": 5971 }, { "epoch": 0.38, "learning_rate": 0.00014148654072655646, "loss": 0.164, "step": 5972 }, { "epoch": 0.38, "learning_rate": 0.00014146764453140459, "loss": 0.1531, "step": 5973 }, { "epoch": 0.38, "learning_rate": 0.0001414487465479293, "loss": 0.1361, "step": 5974 }, { "epoch": 0.38, "learning_rate": 0.00014142984677694543, "loss": 0.1649, "step": 5975 }, { "epoch": 0.38, "learning_rate": 0.00014141094521926816, "loss": 0.1661, "step": 5976 }, { "epoch": 0.38, "learning_rate": 0.00014139204187571265, "loss": 0.1559, "step": 5977 }, { "epoch": 0.38, "learning_rate": 0.00014137313674709404, "loss": 0.1558, "step": 5978 }, { "epoch": 0.38, "learning_rate": 0.0001413542298342277, "loss": 0.161, "step": 5979 }, { "epoch": 0.38, "learning_rate": 0.0001413353211379289, "loss": 0.1778, "step": 5980 }, { "epoch": 0.38, "learning_rate": 0.00014131641065901328, "loss": 0.1616, "step": 5981 }, { "epoch": 0.38, "learning_rate": 0.00014129749839829616, "loss": 0.1246, "step": 5982 }, { "epoch": 0.38, "learning_rate": 0.00014127858435659328, "loss": 0.1491, "step": 5983 }, { "epoch": 0.38, "learning_rate": 0.00014125966853472026, "loss": 0.139, "step": 5984 }, { "epoch": 0.38, "learning_rate": 0.0001412407509334929, "loss": 0.1449, "step": 5985 }, { "epoch": 0.38, "learning_rate": 0.000141221831553727, "loss": 0.1401, "step": 5986 }, { "epoch": 0.38, "learning_rate": 0.00014120291039623848, "loss": 0.1635, "step": 5987 }, { "epoch": 0.38, "learning_rate": 0.00014118398746184338, "loss": 0.1393, "step": 5988 }, { "epoch": 0.38, "learning_rate": 0.00014116506275135767, "loss": 0.1433, "step": 5989 }, { "epoch": 0.38, "learning_rate": 0.00014114613626559755, "loss": 0.1508, "step": 5990 }, { "epoch": 0.38, "learning_rate": 0.00014112720800537928, "loss": 0.134, "step": 5991 }, { "epoch": 0.38, "learning_rate": 0.00014110827797151907, "loss": 0.16, "step": 5992 }, { "epoch": 0.38, "learning_rate": 0.00014108934616483335, "loss": 0.1396, "step": 5993 }, { "epoch": 0.38, "learning_rate": 0.00014107041258613857, "loss": 0.1546, "step": 5994 }, { "epoch": 0.38, "learning_rate": 0.00014105147723625117, "loss": 0.1385, "step": 5995 }, { "epoch": 0.38, "learning_rate": 0.0001410325401159879, "loss": 0.1486, "step": 5996 }, { "epoch": 0.38, "learning_rate": 0.00014101360122616532, "loss": 0.1537, "step": 5997 }, { "epoch": 0.38, "learning_rate": 0.0001409946605676002, "loss": 0.1596, "step": 5998 }, { "epoch": 0.38, "learning_rate": 0.0001409757181411094, "loss": 0.1701, "step": 5999 }, { "epoch": 0.38, "learning_rate": 0.0001409567739475098, "loss": 0.1703, "step": 6000 }, { "epoch": 0.38, "learning_rate": 0.00014093782798761843, "loss": 0.1521, "step": 6001 }, { "epoch": 0.38, "learning_rate": 0.0001409188802622523, "loss": 0.143, "step": 6002 }, { "epoch": 0.38, "learning_rate": 0.00014089993077222857, "loss": 0.147, "step": 6003 }, { "epoch": 0.38, "learning_rate": 0.00014088097951836444, "loss": 0.1463, "step": 6004 }, { "epoch": 0.39, "learning_rate": 0.0001408620265014772, "loss": 0.1508, "step": 6005 }, { "epoch": 0.39, "learning_rate": 0.00014084307172238417, "loss": 0.1559, "step": 6006 }, { "epoch": 0.39, "learning_rate": 0.0001408241151819029, "loss": 0.1476, "step": 6007 }, { "epoch": 0.39, "learning_rate": 0.00014080515688085082, "loss": 0.1511, "step": 6008 }, { "epoch": 0.39, "learning_rate": 0.0001407861968200455, "loss": 0.1555, "step": 6009 }, { "epoch": 0.39, "learning_rate": 0.0001407672350003047, "loss": 0.1262, "step": 6010 }, { "epoch": 0.39, "learning_rate": 0.00014074827142244607, "loss": 0.1692, "step": 6011 }, { "epoch": 0.39, "learning_rate": 0.00014072930608728746, "loss": 0.1521, "step": 6012 }, { "epoch": 0.39, "learning_rate": 0.0001407103389956468, "loss": 0.1619, "step": 6013 }, { "epoch": 0.39, "learning_rate": 0.00014069137014834202, "loss": 0.1587, "step": 6014 }, { "epoch": 0.39, "learning_rate": 0.00014067239954619116, "loss": 0.1485, "step": 6015 }, { "epoch": 0.39, "learning_rate": 0.00014065342719001236, "loss": 0.1578, "step": 6016 }, { "epoch": 0.39, "learning_rate": 0.00014063445308062385, "loss": 0.1343, "step": 6017 }, { "epoch": 0.39, "learning_rate": 0.00014061547721884384, "loss": 0.1543, "step": 6018 }, { "epoch": 0.39, "learning_rate": 0.0001405964996054907, "loss": 0.1498, "step": 6019 }, { "epoch": 0.39, "learning_rate": 0.0001405775202413829, "loss": 0.1714, "step": 6020 }, { "epoch": 0.39, "learning_rate": 0.00014055853912733887, "loss": 0.1548, "step": 6021 }, { "epoch": 0.39, "learning_rate": 0.00014053955626417723, "loss": 0.1352, "step": 6022 }, { "epoch": 0.39, "learning_rate": 0.00014052057165271658, "loss": 0.1508, "step": 6023 }, { "epoch": 0.39, "learning_rate": 0.00014050158529377573, "loss": 0.1449, "step": 6024 }, { "epoch": 0.39, "learning_rate": 0.0001404825971881734, "loss": 0.154, "step": 6025 }, { "epoch": 0.39, "learning_rate": 0.0001404636073367285, "loss": 0.1561, "step": 6026 }, { "epoch": 0.39, "learning_rate": 0.00014044461574026, "loss": 0.1408, "step": 6027 }, { "epoch": 0.39, "learning_rate": 0.00014042562239958688, "loss": 0.1716, "step": 6028 }, { "epoch": 0.39, "learning_rate": 0.00014040662731552828, "loss": 0.146, "step": 6029 }, { "epoch": 0.39, "learning_rate": 0.0001403876304889034, "loss": 0.1873, "step": 6030 }, { "epoch": 0.39, "learning_rate": 0.00014036863192053143, "loss": 0.1608, "step": 6031 }, { "epoch": 0.39, "learning_rate": 0.00014034963161123175, "loss": 0.1555, "step": 6032 }, { "epoch": 0.39, "learning_rate": 0.00014033062956182369, "loss": 0.1514, "step": 6033 }, { "epoch": 0.39, "learning_rate": 0.00014031162577312676, "loss": 0.1567, "step": 6034 }, { "epoch": 0.39, "learning_rate": 0.00014029262024596058, "loss": 0.1549, "step": 6035 }, { "epoch": 0.39, "learning_rate": 0.00014027361298114474, "loss": 0.1594, "step": 6036 }, { "epoch": 0.39, "learning_rate": 0.00014025460397949893, "loss": 0.1607, "step": 6037 }, { "epoch": 0.39, "learning_rate": 0.0001402355932418429, "loss": 0.1528, "step": 6038 }, { "epoch": 0.39, "learning_rate": 0.00014021658076899653, "loss": 0.1322, "step": 6039 }, { "epoch": 0.39, "learning_rate": 0.00014019756656177975, "loss": 0.1751, "step": 6040 }, { "epoch": 0.39, "learning_rate": 0.00014017855062101257, "loss": 0.1482, "step": 6041 }, { "epoch": 0.39, "learning_rate": 0.00014015953294751508, "loss": 0.1507, "step": 6042 }, { "epoch": 0.39, "learning_rate": 0.00014014051354210735, "loss": 0.1463, "step": 6043 }, { "epoch": 0.39, "learning_rate": 0.0001401214924056097, "loss": 0.1565, "step": 6044 }, { "epoch": 0.39, "learning_rate": 0.0001401024695388424, "loss": 0.1599, "step": 6045 }, { "epoch": 0.39, "learning_rate": 0.0001400834449426258, "loss": 0.14, "step": 6046 }, { "epoch": 0.39, "learning_rate": 0.00014006441861778038, "loss": 0.1372, "step": 6047 }, { "epoch": 0.39, "learning_rate": 0.00014004539056512667, "loss": 0.1949, "step": 6048 }, { "epoch": 0.39, "learning_rate": 0.00014002636078548525, "loss": 0.1742, "step": 6049 }, { "epoch": 0.39, "learning_rate": 0.00014000732927967676, "loss": 0.1485, "step": 6050 }, { "epoch": 0.39, "learning_rate": 0.000139988296048522, "loss": 0.1501, "step": 6051 }, { "epoch": 0.39, "learning_rate": 0.00013996926109284183, "loss": 0.1454, "step": 6052 }, { "epoch": 0.39, "learning_rate": 0.00013995022441345704, "loss": 0.1484, "step": 6053 }, { "epoch": 0.39, "learning_rate": 0.00013993118601118867, "loss": 0.1581, "step": 6054 }, { "epoch": 0.39, "learning_rate": 0.00013991214588685775, "loss": 0.1613, "step": 6055 }, { "epoch": 0.39, "learning_rate": 0.00013989310404128542, "loss": 0.1437, "step": 6056 }, { "epoch": 0.39, "learning_rate": 0.00013987406047529283, "loss": 0.1441, "step": 6057 }, { "epoch": 0.39, "learning_rate": 0.0001398550151897013, "loss": 0.1514, "step": 6058 }, { "epoch": 0.39, "learning_rate": 0.0001398359681853321, "loss": 0.1352, "step": 6059 }, { "epoch": 0.39, "learning_rate": 0.00013981691946300675, "loss": 0.1462, "step": 6060 }, { "epoch": 0.39, "learning_rate": 0.00013979786902354666, "loss": 0.1762, "step": 6061 }, { "epoch": 0.39, "learning_rate": 0.00013977881686777342, "loss": 0.1592, "step": 6062 }, { "epoch": 0.39, "learning_rate": 0.00013975976299650865, "loss": 0.15, "step": 6063 }, { "epoch": 0.39, "learning_rate": 0.0001397407074105741, "loss": 0.1516, "step": 6064 }, { "epoch": 0.39, "learning_rate": 0.00013972165011079155, "loss": 0.1651, "step": 6065 }, { "epoch": 0.39, "learning_rate": 0.00013970259109798277, "loss": 0.1671, "step": 6066 }, { "epoch": 0.39, "learning_rate": 0.00013968353037296984, "loss": 0.1399, "step": 6067 }, { "epoch": 0.39, "learning_rate": 0.00013966446793657467, "loss": 0.15, "step": 6068 }, { "epoch": 0.39, "learning_rate": 0.0001396454037896194, "loss": 0.1727, "step": 6069 }, { "epoch": 0.39, "learning_rate": 0.00013962633793292614, "loss": 0.1624, "step": 6070 }, { "epoch": 0.39, "learning_rate": 0.00013960727036731714, "loss": 0.1521, "step": 6071 }, { "epoch": 0.39, "learning_rate": 0.0001395882010936147, "loss": 0.1642, "step": 6072 }, { "epoch": 0.39, "learning_rate": 0.00013956913011264118, "loss": 0.1293, "step": 6073 }, { "epoch": 0.39, "learning_rate": 0.00013955005742521902, "loss": 0.1763, "step": 6074 }, { "epoch": 0.39, "learning_rate": 0.00013953098303217082, "loss": 0.1434, "step": 6075 }, { "epoch": 0.39, "learning_rate": 0.0001395119069343191, "loss": 0.1444, "step": 6076 }, { "epoch": 0.39, "learning_rate": 0.00013949282913248658, "loss": 0.1648, "step": 6077 }, { "epoch": 0.39, "learning_rate": 0.00013947374962749597, "loss": 0.1436, "step": 6078 }, { "epoch": 0.39, "learning_rate": 0.0001394546684201701, "loss": 0.1537, "step": 6079 }, { "epoch": 0.39, "learning_rate": 0.00013943558551133186, "loss": 0.1588, "step": 6080 }, { "epoch": 0.39, "learning_rate": 0.00013941650090180423, "loss": 0.1656, "step": 6081 }, { "epoch": 0.39, "learning_rate": 0.00013939741459241021, "loss": 0.1389, "step": 6082 }, { "epoch": 0.39, "learning_rate": 0.00013937832658397298, "loss": 0.1583, "step": 6083 }, { "epoch": 0.39, "learning_rate": 0.00013935923687731567, "loss": 0.1544, "step": 6084 }, { "epoch": 0.39, "learning_rate": 0.00013934014547326153, "loss": 0.1587, "step": 6085 }, { "epoch": 0.39, "learning_rate": 0.0001393210523726339, "loss": 0.1671, "step": 6086 }, { "epoch": 0.39, "learning_rate": 0.00013930195757625623, "loss": 0.1417, "step": 6087 }, { "epoch": 0.39, "learning_rate": 0.00013928286108495197, "loss": 0.145, "step": 6088 }, { "epoch": 0.39, "learning_rate": 0.00013926376289954463, "loss": 0.1753, "step": 6089 }, { "epoch": 0.39, "learning_rate": 0.00013924466302085787, "loss": 0.1567, "step": 6090 }, { "epoch": 0.39, "learning_rate": 0.00013922556144971542, "loss": 0.1592, "step": 6091 }, { "epoch": 0.39, "learning_rate": 0.00013920645818694098, "loss": 0.1615, "step": 6092 }, { "epoch": 0.39, "learning_rate": 0.00013918735323335842, "loss": 0.1629, "step": 6093 }, { "epoch": 0.39, "learning_rate": 0.0001391682465897917, "loss": 0.1671, "step": 6094 }, { "epoch": 0.39, "learning_rate": 0.00013914913825706474, "loss": 0.1713, "step": 6095 }, { "epoch": 0.39, "learning_rate": 0.00013913002823600162, "loss": 0.1437, "step": 6096 }, { "epoch": 0.39, "learning_rate": 0.0001391109165274265, "loss": 0.1547, "step": 6097 }, { "epoch": 0.39, "learning_rate": 0.00013909180313216363, "loss": 0.1386, "step": 6098 }, { "epoch": 0.39, "learning_rate": 0.00013907268805103714, "loss": 0.1608, "step": 6099 }, { "epoch": 0.39, "learning_rate": 0.0001390535712848715, "loss": 0.1312, "step": 6100 }, { "epoch": 0.39, "learning_rate": 0.00013903445283449113, "loss": 0.1644, "step": 6101 }, { "epoch": 0.39, "learning_rate": 0.00013901533270072053, "loss": 0.1637, "step": 6102 }, { "epoch": 0.39, "learning_rate": 0.00013899621088438418, "loss": 0.1453, "step": 6103 }, { "epoch": 0.39, "learning_rate": 0.00013897708738630686, "loss": 0.1559, "step": 6104 }, { "epoch": 0.39, "learning_rate": 0.00013895796220731322, "loss": 0.1558, "step": 6105 }, { "epoch": 0.39, "learning_rate": 0.00013893883534822797, "loss": 0.1504, "step": 6106 }, { "epoch": 0.39, "learning_rate": 0.0001389197068098761, "loss": 0.1402, "step": 6107 }, { "epoch": 0.39, "learning_rate": 0.00013890057659308246, "loss": 0.1718, "step": 6108 }, { "epoch": 0.39, "learning_rate": 0.0001388814446986721, "loss": 0.1503, "step": 6109 }, { "epoch": 0.39, "learning_rate": 0.00013886231112747005, "loss": 0.1561, "step": 6110 }, { "epoch": 0.39, "learning_rate": 0.00013884317588030155, "loss": 0.1323, "step": 6111 }, { "epoch": 0.39, "learning_rate": 0.0001388240389579917, "loss": 0.1508, "step": 6112 }, { "epoch": 0.39, "learning_rate": 0.00013880490036136583, "loss": 0.166, "step": 6113 }, { "epoch": 0.39, "learning_rate": 0.0001387857600912494, "loss": 0.1722, "step": 6114 }, { "epoch": 0.39, "learning_rate": 0.00013876661814846773, "loss": 0.1531, "step": 6115 }, { "epoch": 0.39, "learning_rate": 0.00013874747453384643, "loss": 0.1439, "step": 6116 }, { "epoch": 0.39, "learning_rate": 0.000138728329248211, "loss": 0.1603, "step": 6117 }, { "epoch": 0.39, "learning_rate": 0.00013870918229238715, "loss": 0.1397, "step": 6118 }, { "epoch": 0.39, "learning_rate": 0.00013869003366720058, "loss": 0.1671, "step": 6119 }, { "epoch": 0.39, "learning_rate": 0.00013867088337347704, "loss": 0.1363, "step": 6120 }, { "epoch": 0.39, "learning_rate": 0.0001386517314120425, "loss": 0.1328, "step": 6121 }, { "epoch": 0.39, "learning_rate": 0.00013863257778372288, "loss": 0.1622, "step": 6122 }, { "epoch": 0.39, "learning_rate": 0.00013861342248934412, "loss": 0.1409, "step": 6123 }, { "epoch": 0.39, "learning_rate": 0.0001385942655297324, "loss": 0.161, "step": 6124 }, { "epoch": 0.39, "learning_rate": 0.00013857510690571382, "loss": 0.1561, "step": 6125 }, { "epoch": 0.39, "learning_rate": 0.00013855594661811463, "loss": 0.1842, "step": 6126 }, { "epoch": 0.39, "learning_rate": 0.0001385367846677611, "loss": 0.1439, "step": 6127 }, { "epoch": 0.39, "learning_rate": 0.0001385176210554797, "loss": 0.1702, "step": 6128 }, { "epoch": 0.39, "learning_rate": 0.00013849845578209677, "loss": 0.1599, "step": 6129 }, { "epoch": 0.39, "learning_rate": 0.00013847928884843885, "loss": 0.1631, "step": 6130 }, { "epoch": 0.39, "learning_rate": 0.0001384601202553326, "loss": 0.1747, "step": 6131 }, { "epoch": 0.39, "learning_rate": 0.00013844095000360457, "loss": 0.1352, "step": 6132 }, { "epoch": 0.39, "learning_rate": 0.00013842177809408155, "loss": 0.1462, "step": 6133 }, { "epoch": 0.39, "learning_rate": 0.00013840260452759035, "loss": 0.1382, "step": 6134 }, { "epoch": 0.39, "learning_rate": 0.00013838342930495783, "loss": 0.1507, "step": 6135 }, { "epoch": 0.39, "learning_rate": 0.0001383642524270109, "loss": 0.1358, "step": 6136 }, { "epoch": 0.39, "learning_rate": 0.00013834507389457666, "loss": 0.1453, "step": 6137 }, { "epoch": 0.39, "learning_rate": 0.00013832589370848218, "loss": 0.1452, "step": 6138 }, { "epoch": 0.39, "learning_rate": 0.00013830671186955456, "loss": 0.1511, "step": 6139 }, { "epoch": 0.39, "learning_rate": 0.00013828752837862103, "loss": 0.1461, "step": 6140 }, { "epoch": 0.39, "learning_rate": 0.000138268343236509, "loss": 0.152, "step": 6141 }, { "epoch": 0.39, "learning_rate": 0.00013824915644404572, "loss": 0.15, "step": 6142 }, { "epoch": 0.39, "learning_rate": 0.00013822996800205869, "loss": 0.155, "step": 6143 }, { "epoch": 0.39, "learning_rate": 0.00013821077791137545, "loss": 0.13, "step": 6144 }, { "epoch": 0.39, "learning_rate": 0.00013819158617282357, "loss": 0.1487, "step": 6145 }, { "epoch": 0.39, "learning_rate": 0.00013817239278723066, "loss": 0.1538, "step": 6146 }, { "epoch": 0.39, "learning_rate": 0.0001381531977554245, "loss": 0.1632, "step": 6147 }, { "epoch": 0.39, "learning_rate": 0.0001381340010782329, "loss": 0.1428, "step": 6148 }, { "epoch": 0.39, "learning_rate": 0.00013811480275648368, "loss": 0.1569, "step": 6149 }, { "epoch": 0.39, "learning_rate": 0.00013809560279100483, "loss": 0.1347, "step": 6150 }, { "epoch": 0.39, "learning_rate": 0.00013807640118262434, "loss": 0.1569, "step": 6151 }, { "epoch": 0.39, "learning_rate": 0.00013805719793217032, "loss": 0.1533, "step": 6152 }, { "epoch": 0.39, "learning_rate": 0.00013803799304047085, "loss": 0.1515, "step": 6153 }, { "epoch": 0.39, "learning_rate": 0.00013801878650835422, "loss": 0.1409, "step": 6154 }, { "epoch": 0.39, "learning_rate": 0.00013799957833664873, "loss": 0.1624, "step": 6155 }, { "epoch": 0.39, "learning_rate": 0.00013798036852618275, "loss": 0.1618, "step": 6156 }, { "epoch": 0.39, "learning_rate": 0.00013796115707778465, "loss": 0.1342, "step": 6157 }, { "epoch": 0.39, "learning_rate": 0.00013794194399228304, "loss": 0.1504, "step": 6158 }, { "epoch": 0.39, "learning_rate": 0.00013792272927050644, "loss": 0.1459, "step": 6159 }, { "epoch": 0.39, "learning_rate": 0.00013790351291328347, "loss": 0.1349, "step": 6160 }, { "epoch": 0.4, "learning_rate": 0.0001378842949214429, "loss": 0.1538, "step": 6161 }, { "epoch": 0.4, "learning_rate": 0.00013786507529581354, "loss": 0.1371, "step": 6162 }, { "epoch": 0.4, "learning_rate": 0.00013784585403722415, "loss": 0.1262, "step": 6163 }, { "epoch": 0.4, "learning_rate": 0.00013782663114650379, "loss": 0.1464, "step": 6164 }, { "epoch": 0.4, "learning_rate": 0.0001378074066244814, "loss": 0.1646, "step": 6165 }, { "epoch": 0.4, "learning_rate": 0.000137788180471986, "loss": 0.1408, "step": 6166 }, { "epoch": 0.4, "learning_rate": 0.00013776895268984682, "loss": 0.169, "step": 6167 }, { "epoch": 0.4, "learning_rate": 0.00013774972327889302, "loss": 0.1529, "step": 6168 }, { "epoch": 0.4, "learning_rate": 0.0001377304922399539, "loss": 0.1454, "step": 6169 }, { "epoch": 0.4, "learning_rate": 0.00013771125957385882, "loss": 0.1535, "step": 6170 }, { "epoch": 0.4, "learning_rate": 0.00013769202528143718, "loss": 0.123, "step": 6171 }, { "epoch": 0.4, "learning_rate": 0.00013767278936351854, "loss": 0.1423, "step": 6172 }, { "epoch": 0.4, "learning_rate": 0.00013765355182093235, "loss": 0.1536, "step": 6173 }, { "epoch": 0.4, "learning_rate": 0.00013763431265450835, "loss": 0.1603, "step": 6174 }, { "epoch": 0.4, "learning_rate": 0.00013761507186507615, "loss": 0.1554, "step": 6175 }, { "epoch": 0.4, "learning_rate": 0.00013759582945346564, "loss": 0.1409, "step": 6176 }, { "epoch": 0.4, "learning_rate": 0.00013757658542050654, "loss": 0.1419, "step": 6177 }, { "epoch": 0.4, "learning_rate": 0.00013755733976702884, "loss": 0.1504, "step": 6178 }, { "epoch": 0.4, "learning_rate": 0.00013753809249386254, "loss": 0.1674, "step": 6179 }, { "epoch": 0.4, "learning_rate": 0.0001375188436018376, "loss": 0.1257, "step": 6180 }, { "epoch": 0.4, "learning_rate": 0.00013749959309178423, "loss": 0.1461, "step": 6181 }, { "epoch": 0.4, "learning_rate": 0.00013748034096453256, "loss": 0.1275, "step": 6182 }, { "epoch": 0.4, "learning_rate": 0.00013746108722091294, "loss": 0.1636, "step": 6183 }, { "epoch": 0.4, "learning_rate": 0.0001374418318617556, "loss": 0.1494, "step": 6184 }, { "epoch": 0.4, "learning_rate": 0.000137422574887891, "loss": 0.1189, "step": 6185 }, { "epoch": 0.4, "learning_rate": 0.00013740331630014962, "loss": 0.1374, "step": 6186 }, { "epoch": 0.4, "learning_rate": 0.00013738405609936197, "loss": 0.137, "step": 6187 }, { "epoch": 0.4, "learning_rate": 0.0001373647942863587, "loss": 0.1557, "step": 6188 }, { "epoch": 0.4, "learning_rate": 0.0001373455308619704, "loss": 0.1268, "step": 6189 }, { "epoch": 0.4, "learning_rate": 0.00013732626582702793, "loss": 0.1574, "step": 6190 }, { "epoch": 0.4, "learning_rate": 0.00013730699918236203, "loss": 0.1563, "step": 6191 }, { "epoch": 0.4, "learning_rate": 0.00013728773092880364, "loss": 0.1192, "step": 6192 }, { "epoch": 0.4, "learning_rate": 0.00013726846106718371, "loss": 0.1396, "step": 6193 }, { "epoch": 0.4, "learning_rate": 0.00013724918959833325, "loss": 0.1446, "step": 6194 }, { "epoch": 0.4, "learning_rate": 0.00013722991652308336, "loss": 0.1527, "step": 6195 }, { "epoch": 0.4, "learning_rate": 0.00013721064184226523, "loss": 0.1298, "step": 6196 }, { "epoch": 0.4, "learning_rate": 0.00013719136555671, "loss": 0.1544, "step": 6197 }, { "epoch": 0.4, "learning_rate": 0.0001371720876672491, "loss": 0.148, "step": 6198 }, { "epoch": 0.4, "learning_rate": 0.00013715280817471384, "loss": 0.1576, "step": 6199 }, { "epoch": 0.4, "learning_rate": 0.0001371335270799357, "loss": 0.158, "step": 6200 }, { "epoch": 0.4, "learning_rate": 0.00013711424438374615, "loss": 0.1497, "step": 6201 }, { "epoch": 0.4, "learning_rate": 0.0001370949600869768, "loss": 0.1553, "step": 6202 }, { "epoch": 0.4, "learning_rate": 0.00013707567419045925, "loss": 0.1693, "step": 6203 }, { "epoch": 0.4, "learning_rate": 0.00013705638669502527, "loss": 0.1645, "step": 6204 }, { "epoch": 0.4, "learning_rate": 0.00013703709760150664, "loss": 0.1681, "step": 6205 }, { "epoch": 0.4, "learning_rate": 0.00013701780691073524, "loss": 0.1525, "step": 6206 }, { "epoch": 0.4, "learning_rate": 0.0001369985146235429, "loss": 0.1423, "step": 6207 }, { "epoch": 0.4, "learning_rate": 0.00013697922074076173, "loss": 0.1512, "step": 6208 }, { "epoch": 0.4, "learning_rate": 0.00013695992526322373, "loss": 0.1433, "step": 6209 }, { "epoch": 0.4, "learning_rate": 0.00013694062819176103, "loss": 0.1629, "step": 6210 }, { "epoch": 0.4, "learning_rate": 0.00013692132952720587, "loss": 0.132, "step": 6211 }, { "epoch": 0.4, "learning_rate": 0.00013690202927039047, "loss": 0.155, "step": 6212 }, { "epoch": 0.4, "learning_rate": 0.00013688272742214724, "loss": 0.1503, "step": 6213 }, { "epoch": 0.4, "learning_rate": 0.00013686342398330848, "loss": 0.1477, "step": 6214 }, { "epoch": 0.4, "learning_rate": 0.00013684411895470677, "loss": 0.1597, "step": 6215 }, { "epoch": 0.4, "learning_rate": 0.0001368248123371746, "loss": 0.1679, "step": 6216 }, { "epoch": 0.4, "learning_rate": 0.00013680550413154455, "loss": 0.1467, "step": 6217 }, { "epoch": 0.4, "learning_rate": 0.00013678619433864938, "loss": 0.134, "step": 6218 }, { "epoch": 0.4, "learning_rate": 0.0001367668829593218, "loss": 0.1459, "step": 6219 }, { "epoch": 0.4, "learning_rate": 0.00013674756999439464, "loss": 0.1687, "step": 6220 }, { "epoch": 0.4, "learning_rate": 0.0001367282554447008, "loss": 0.1734, "step": 6221 }, { "epoch": 0.4, "learning_rate": 0.00013670893931107316, "loss": 0.1718, "step": 6222 }, { "epoch": 0.4, "learning_rate": 0.00013668962159434482, "loss": 0.1551, "step": 6223 }, { "epoch": 0.4, "learning_rate": 0.00013667030229534883, "loss": 0.1485, "step": 6224 }, { "epoch": 0.4, "learning_rate": 0.00013665098141491837, "loss": 0.1532, "step": 6225 }, { "epoch": 0.4, "learning_rate": 0.00013663165895388664, "loss": 0.1599, "step": 6226 }, { "epoch": 0.4, "learning_rate": 0.000136612334913087, "loss": 0.1705, "step": 6227 }, { "epoch": 0.4, "learning_rate": 0.00013659300929335277, "loss": 0.1894, "step": 6228 }, { "epoch": 0.4, "learning_rate": 0.00013657368209551738, "loss": 0.1505, "step": 6229 }, { "epoch": 0.4, "learning_rate": 0.0001365543533204143, "loss": 0.1511, "step": 6230 }, { "epoch": 0.4, "learning_rate": 0.00013653502296887715, "loss": 0.1502, "step": 6231 }, { "epoch": 0.4, "learning_rate": 0.00013651569104173954, "loss": 0.1494, "step": 6232 }, { "epoch": 0.4, "learning_rate": 0.0001364963575398352, "loss": 0.1733, "step": 6233 }, { "epoch": 0.4, "learning_rate": 0.00013647702246399787, "loss": 0.1546, "step": 6234 }, { "epoch": 0.4, "learning_rate": 0.0001364576858150614, "loss": 0.1652, "step": 6235 }, { "epoch": 0.4, "learning_rate": 0.0001364383475938597, "loss": 0.1427, "step": 6236 }, { "epoch": 0.4, "learning_rate": 0.00013641900780122677, "loss": 0.1433, "step": 6237 }, { "epoch": 0.4, "learning_rate": 0.0001363996664379966, "loss": 0.1563, "step": 6238 }, { "epoch": 0.4, "learning_rate": 0.00013638032350500335, "loss": 0.1231, "step": 6239 }, { "epoch": 0.4, "learning_rate": 0.0001363609790030812, "loss": 0.1603, "step": 6240 }, { "epoch": 0.4, "learning_rate": 0.00013634163293306433, "loss": 0.1602, "step": 6241 }, { "epoch": 0.4, "learning_rate": 0.0001363222852957871, "loss": 0.1571, "step": 6242 }, { "epoch": 0.4, "learning_rate": 0.0001363029360920839, "loss": 0.146, "step": 6243 }, { "epoch": 0.4, "learning_rate": 0.0001362835853227892, "loss": 0.1395, "step": 6244 }, { "epoch": 0.4, "learning_rate": 0.00013626423298873744, "loss": 0.1677, "step": 6245 }, { "epoch": 0.4, "learning_rate": 0.00013624487909076328, "loss": 0.1558, "step": 6246 }, { "epoch": 0.4, "learning_rate": 0.00013622552362970133, "loss": 0.1347, "step": 6247 }, { "epoch": 0.4, "learning_rate": 0.0001362061666063863, "loss": 0.1428, "step": 6248 }, { "epoch": 0.4, "learning_rate": 0.00013618680802165303, "loss": 0.1373, "step": 6249 }, { "epoch": 0.4, "learning_rate": 0.00013616744787633633, "loss": 0.1578, "step": 6250 }, { "epoch": 0.4, "learning_rate": 0.00013614808617127113, "loss": 0.1679, "step": 6251 }, { "epoch": 0.4, "learning_rate": 0.0001361287229072924, "loss": 0.1426, "step": 6252 }, { "epoch": 0.4, "learning_rate": 0.00013610935808523527, "loss": 0.1753, "step": 6253 }, { "epoch": 0.4, "learning_rate": 0.00013608999170593476, "loss": 0.1474, "step": 6254 }, { "epoch": 0.4, "learning_rate": 0.00013607062377022613, "loss": 0.1545, "step": 6255 }, { "epoch": 0.4, "learning_rate": 0.0001360512542789446, "loss": 0.1309, "step": 6256 }, { "epoch": 0.4, "learning_rate": 0.0001360318832329255, "loss": 0.1344, "step": 6257 }, { "epoch": 0.4, "learning_rate": 0.00013601251063300422, "loss": 0.1426, "step": 6258 }, { "epoch": 0.4, "learning_rate": 0.00013599313648001625, "loss": 0.1428, "step": 6259 }, { "epoch": 0.4, "learning_rate": 0.0001359737607747971, "loss": 0.1258, "step": 6260 }, { "epoch": 0.4, "learning_rate": 0.00013595438351818233, "loss": 0.1513, "step": 6261 }, { "epoch": 0.4, "learning_rate": 0.00013593500471100763, "loss": 0.1285, "step": 6262 }, { "epoch": 0.4, "learning_rate": 0.0001359156243541087, "loss": 0.1619, "step": 6263 }, { "epoch": 0.4, "learning_rate": 0.00013589624244832136, "loss": 0.1383, "step": 6264 }, { "epoch": 0.4, "learning_rate": 0.00013587685899448147, "loss": 0.1499, "step": 6265 }, { "epoch": 0.4, "learning_rate": 0.00013585747399342496, "loss": 0.1586, "step": 6266 }, { "epoch": 0.4, "learning_rate": 0.0001358380874459878, "loss": 0.1402, "step": 6267 }, { "epoch": 0.4, "learning_rate": 0.00013581869935300604, "loss": 0.1521, "step": 6268 }, { "epoch": 0.4, "learning_rate": 0.00013579930971531584, "loss": 0.1537, "step": 6269 }, { "epoch": 0.4, "learning_rate": 0.0001357799185337534, "loss": 0.1416, "step": 6270 }, { "epoch": 0.4, "learning_rate": 0.0001357605258091549, "loss": 0.1528, "step": 6271 }, { "epoch": 0.4, "learning_rate": 0.00013574113154235675, "loss": 0.1524, "step": 6272 }, { "epoch": 0.4, "learning_rate": 0.0001357217357341953, "loss": 0.1443, "step": 6273 }, { "epoch": 0.4, "learning_rate": 0.00013570233838550703, "loss": 0.1555, "step": 6274 }, { "epoch": 0.4, "learning_rate": 0.00013568293949712848, "loss": 0.1319, "step": 6275 }, { "epoch": 0.4, "learning_rate": 0.00013566353906989623, "loss": 0.1426, "step": 6276 }, { "epoch": 0.4, "learning_rate": 0.0001356441371046469, "loss": 0.1458, "step": 6277 }, { "epoch": 0.4, "learning_rate": 0.00013562473360221726, "loss": 0.1294, "step": 6278 }, { "epoch": 0.4, "learning_rate": 0.00013560532856344404, "loss": 0.1385, "step": 6279 }, { "epoch": 0.4, "learning_rate": 0.0001355859219891642, "loss": 0.1373, "step": 6280 }, { "epoch": 0.4, "learning_rate": 0.00013556651388021458, "loss": 0.1511, "step": 6281 }, { "epoch": 0.4, "learning_rate": 0.0001355471042374322, "loss": 0.1364, "step": 6282 }, { "epoch": 0.4, "learning_rate": 0.0001355276930616541, "loss": 0.1357, "step": 6283 }, { "epoch": 0.4, "learning_rate": 0.00013550828035371738, "loss": 0.1555, "step": 6284 }, { "epoch": 0.4, "learning_rate": 0.0001354888661144593, "loss": 0.1727, "step": 6285 }, { "epoch": 0.4, "learning_rate": 0.00013546945034471707, "loss": 0.1732, "step": 6286 }, { "epoch": 0.4, "learning_rate": 0.00013545003304532802, "loss": 0.1419, "step": 6287 }, { "epoch": 0.4, "learning_rate": 0.0001354306142171295, "loss": 0.136, "step": 6288 }, { "epoch": 0.4, "learning_rate": 0.000135411193860959, "loss": 0.1637, "step": 6289 }, { "epoch": 0.4, "learning_rate": 0.00013539177197765403, "loss": 0.1401, "step": 6290 }, { "epoch": 0.4, "learning_rate": 0.00013537234856805214, "loss": 0.1606, "step": 6291 }, { "epoch": 0.4, "learning_rate": 0.00013535292363299104, "loss": 0.1438, "step": 6292 }, { "epoch": 0.4, "learning_rate": 0.00013533349717330842, "loss": 0.1618, "step": 6293 }, { "epoch": 0.4, "learning_rate": 0.00013531406918984202, "loss": 0.1454, "step": 6294 }, { "epoch": 0.4, "learning_rate": 0.00013529463968342976, "loss": 0.1324, "step": 6295 }, { "epoch": 0.4, "learning_rate": 0.0001352752086549095, "loss": 0.1577, "step": 6296 }, { "epoch": 0.4, "learning_rate": 0.00013525577610511922, "loss": 0.1295, "step": 6297 }, { "epoch": 0.4, "learning_rate": 0.00013523634203489696, "loss": 0.1497, "step": 6298 }, { "epoch": 0.4, "learning_rate": 0.00013521690644508087, "loss": 0.1499, "step": 6299 }, { "epoch": 0.4, "learning_rate": 0.0001351974693365091, "loss": 0.1425, "step": 6300 }, { "epoch": 0.4, "learning_rate": 0.00013517803071001985, "loss": 0.1518, "step": 6301 }, { "epoch": 0.4, "learning_rate": 0.00013515859056645156, "loss": 0.1473, "step": 6302 }, { "epoch": 0.4, "learning_rate": 0.00013513914890664244, "loss": 0.149, "step": 6303 }, { "epoch": 0.4, "learning_rate": 0.00013511970573143095, "loss": 0.1417, "step": 6304 }, { "epoch": 0.4, "learning_rate": 0.00013510026104165566, "loss": 0.147, "step": 6305 }, { "epoch": 0.4, "learning_rate": 0.00013508081483815513, "loss": 0.1553, "step": 6306 }, { "epoch": 0.4, "learning_rate": 0.00013506136712176798, "loss": 0.1385, "step": 6307 }, { "epoch": 0.4, "learning_rate": 0.00013504191789333284, "loss": 0.1578, "step": 6308 }, { "epoch": 0.4, "learning_rate": 0.0001350224671536886, "loss": 0.1412, "step": 6309 }, { "epoch": 0.4, "learning_rate": 0.000135003014903674, "loss": 0.1444, "step": 6310 }, { "epoch": 0.4, "learning_rate": 0.00013498356114412791, "loss": 0.1346, "step": 6311 }, { "epoch": 0.4, "learning_rate": 0.00013496410587588939, "loss": 0.1638, "step": 6312 }, { "epoch": 0.4, "learning_rate": 0.0001349446490997974, "loss": 0.1534, "step": 6313 }, { "epoch": 0.4, "learning_rate": 0.00013492519081669094, "loss": 0.1528, "step": 6314 }, { "epoch": 0.4, "learning_rate": 0.00013490573102740934, "loss": 0.1536, "step": 6315 }, { "epoch": 0.4, "learning_rate": 0.00013488626973279173, "loss": 0.1371, "step": 6316 }, { "epoch": 0.41, "learning_rate": 0.00013486680693367738, "loss": 0.1567, "step": 6317 }, { "epoch": 0.41, "learning_rate": 0.00013484734263090564, "loss": 0.1645, "step": 6318 }, { "epoch": 0.41, "learning_rate": 0.00013482787682531596, "loss": 0.1494, "step": 6319 }, { "epoch": 0.41, "learning_rate": 0.0001348084095177478, "loss": 0.1702, "step": 6320 }, { "epoch": 0.41, "learning_rate": 0.00013478894070904064, "loss": 0.1773, "step": 6321 }, { "epoch": 0.41, "learning_rate": 0.00013476947040003422, "loss": 0.1297, "step": 6322 }, { "epoch": 0.41, "learning_rate": 0.00013474999859156809, "loss": 0.1486, "step": 6323 }, { "epoch": 0.41, "learning_rate": 0.00013473052528448201, "loss": 0.1562, "step": 6324 }, { "epoch": 0.41, "learning_rate": 0.00013471105047961585, "loss": 0.1594, "step": 6325 }, { "epoch": 0.41, "learning_rate": 0.00013469157417780943, "loss": 0.1593, "step": 6326 }, { "epoch": 0.41, "learning_rate": 0.00013467209637990266, "loss": 0.1403, "step": 6327 }, { "epoch": 0.41, "learning_rate": 0.00013465261708673552, "loss": 0.131, "step": 6328 }, { "epoch": 0.41, "learning_rate": 0.00013463313629914818, "loss": 0.1378, "step": 6329 }, { "epoch": 0.41, "learning_rate": 0.00013461365401798062, "loss": 0.1829, "step": 6330 }, { "epoch": 0.41, "learning_rate": 0.00013459417024407313, "loss": 0.1484, "step": 6331 }, { "epoch": 0.41, "learning_rate": 0.0001345746849782659, "loss": 0.148, "step": 6332 }, { "epoch": 0.41, "learning_rate": 0.00013455519822139927, "loss": 0.1507, "step": 6333 }, { "epoch": 0.41, "learning_rate": 0.00013453570997431363, "loss": 0.1686, "step": 6334 }, { "epoch": 0.41, "learning_rate": 0.00013451622023784944, "loss": 0.1529, "step": 6335 }, { "epoch": 0.41, "learning_rate": 0.00013449672901284722, "loss": 0.1734, "step": 6336 }, { "epoch": 0.41, "learning_rate": 0.00013447723630014745, "loss": 0.1396, "step": 6337 }, { "epoch": 0.41, "learning_rate": 0.00013445774210059085, "loss": 0.1464, "step": 6338 }, { "epoch": 0.41, "learning_rate": 0.00013443824641501813, "loss": 0.1692, "step": 6339 }, { "epoch": 0.41, "learning_rate": 0.00013441874924427, "loss": 0.1404, "step": 6340 }, { "epoch": 0.41, "learning_rate": 0.0001343992505891873, "loss": 0.1274, "step": 6341 }, { "epoch": 0.41, "learning_rate": 0.000134379750450611, "loss": 0.139, "step": 6342 }, { "epoch": 0.41, "learning_rate": 0.000134360248829382, "loss": 0.1733, "step": 6343 }, { "epoch": 0.41, "learning_rate": 0.00013434074572634126, "loss": 0.1394, "step": 6344 }, { "epoch": 0.41, "learning_rate": 0.00013432124114232997, "loss": 0.1691, "step": 6345 }, { "epoch": 0.41, "learning_rate": 0.00013430173507818926, "loss": 0.1672, "step": 6346 }, { "epoch": 0.41, "learning_rate": 0.0001342822275347603, "loss": 0.1492, "step": 6347 }, { "epoch": 0.41, "learning_rate": 0.0001342627185128844, "loss": 0.1453, "step": 6348 }, { "epoch": 0.41, "learning_rate": 0.00013424320801340287, "loss": 0.1566, "step": 6349 }, { "epoch": 0.41, "learning_rate": 0.00013422369603715718, "loss": 0.1515, "step": 6350 }, { "epoch": 0.41, "learning_rate": 0.00013420418258498873, "loss": 0.1602, "step": 6351 }, { "epoch": 0.41, "learning_rate": 0.00013418466765773907, "loss": 0.1474, "step": 6352 }, { "epoch": 0.41, "learning_rate": 0.00013416515125624982, "loss": 0.1564, "step": 6353 }, { "epoch": 0.41, "learning_rate": 0.00013414563338136264, "loss": 0.1377, "step": 6354 }, { "epoch": 0.41, "learning_rate": 0.0001341261140339192, "loss": 0.1315, "step": 6355 }, { "epoch": 0.41, "learning_rate": 0.00013410659321476132, "loss": 0.1564, "step": 6356 }, { "epoch": 0.41, "learning_rate": 0.0001340870709247309, "loss": 0.1748, "step": 6357 }, { "epoch": 0.41, "learning_rate": 0.00013406754716466978, "loss": 0.1356, "step": 6358 }, { "epoch": 0.41, "learning_rate": 0.00013404802193541998, "loss": 0.1473, "step": 6359 }, { "epoch": 0.41, "learning_rate": 0.00013402849523782353, "loss": 0.1617, "step": 6360 }, { "epoch": 0.41, "learning_rate": 0.0001340089670727225, "loss": 0.1426, "step": 6361 }, { "epoch": 0.41, "learning_rate": 0.00013398943744095907, "loss": 0.1437, "step": 6362 }, { "epoch": 0.41, "learning_rate": 0.00013396990634337557, "loss": 0.1616, "step": 6363 }, { "epoch": 0.41, "learning_rate": 0.0001339503737808141, "loss": 0.1518, "step": 6364 }, { "epoch": 0.41, "learning_rate": 0.0001339308397541172, "loss": 0.1482, "step": 6365 }, { "epoch": 0.41, "learning_rate": 0.00013391130426412723, "loss": 0.17, "step": 6366 }, { "epoch": 0.41, "learning_rate": 0.0001338917673116866, "loss": 0.1551, "step": 6367 }, { "epoch": 0.41, "learning_rate": 0.0001338722288976379, "loss": 0.1536, "step": 6368 }, { "epoch": 0.41, "learning_rate": 0.0001338526890228238, "loss": 0.1525, "step": 6369 }, { "epoch": 0.41, "learning_rate": 0.00013383314768808694, "loss": 0.1646, "step": 6370 }, { "epoch": 0.41, "learning_rate": 0.00013381360489427, "loss": 0.1509, "step": 6371 }, { "epoch": 0.41, "learning_rate": 0.00013379406064221582, "loss": 0.1667, "step": 6372 }, { "epoch": 0.41, "learning_rate": 0.00013377451493276727, "loss": 0.1479, "step": 6373 }, { "epoch": 0.41, "learning_rate": 0.00013375496776676726, "loss": 0.1509, "step": 6374 }, { "epoch": 0.41, "learning_rate": 0.00013373541914505876, "loss": 0.142, "step": 6375 }, { "epoch": 0.41, "learning_rate": 0.00013371586906848486, "loss": 0.1466, "step": 6376 }, { "epoch": 0.41, "learning_rate": 0.00013369631753788866, "loss": 0.1523, "step": 6377 }, { "epoch": 0.41, "learning_rate": 0.00013367676455411327, "loss": 0.1646, "step": 6378 }, { "epoch": 0.41, "learning_rate": 0.00013365721011800204, "loss": 0.1556, "step": 6379 }, { "epoch": 0.41, "learning_rate": 0.00013363765423039816, "loss": 0.1481, "step": 6380 }, { "epoch": 0.41, "learning_rate": 0.00013361809689214503, "loss": 0.1525, "step": 6381 }, { "epoch": 0.41, "learning_rate": 0.00013359853810408613, "loss": 0.1471, "step": 6382 }, { "epoch": 0.41, "learning_rate": 0.0001335789778670649, "loss": 0.1248, "step": 6383 }, { "epoch": 0.41, "learning_rate": 0.0001335594161819249, "loss": 0.1578, "step": 6384 }, { "epoch": 0.41, "learning_rate": 0.00013353985304950973, "loss": 0.1537, "step": 6385 }, { "epoch": 0.41, "learning_rate": 0.00013352028847066306, "loss": 0.158, "step": 6386 }, { "epoch": 0.41, "learning_rate": 0.00013350072244622867, "loss": 0.1432, "step": 6387 }, { "epoch": 0.41, "learning_rate": 0.00013348115497705028, "loss": 0.1495, "step": 6388 }, { "epoch": 0.41, "learning_rate": 0.00013346158606397182, "loss": 0.1523, "step": 6389 }, { "epoch": 0.41, "learning_rate": 0.0001334420157078372, "loss": 0.1527, "step": 6390 }, { "epoch": 0.41, "learning_rate": 0.00013342244390949042, "loss": 0.1533, "step": 6391 }, { "epoch": 0.41, "learning_rate": 0.00013340287066977547, "loss": 0.1334, "step": 6392 }, { "epoch": 0.41, "learning_rate": 0.00013338329598953653, "loss": 0.1413, "step": 6393 }, { "epoch": 0.41, "learning_rate": 0.00013336371986961775, "loss": 0.1525, "step": 6394 }, { "epoch": 0.41, "learning_rate": 0.00013334414231086332, "loss": 0.1589, "step": 6395 }, { "epoch": 0.41, "learning_rate": 0.00013332456331411761, "loss": 0.1398, "step": 6396 }, { "epoch": 0.41, "learning_rate": 0.00013330498288022493, "loss": 0.1468, "step": 6397 }, { "epoch": 0.41, "learning_rate": 0.00013328540101002973, "loss": 0.1455, "step": 6398 }, { "epoch": 0.41, "learning_rate": 0.00013326581770437644, "loss": 0.1743, "step": 6399 }, { "epoch": 0.41, "learning_rate": 0.0001332462329641097, "loss": 0.1623, "step": 6400 }, { "epoch": 0.41, "learning_rate": 0.00013322664679007402, "loss": 0.1422, "step": 6401 }, { "epoch": 0.41, "learning_rate": 0.0001332070591831141, "loss": 0.1542, "step": 6402 }, { "epoch": 0.41, "learning_rate": 0.00013318747014407468, "loss": 0.1571, "step": 6403 }, { "epoch": 0.41, "learning_rate": 0.0001331678796738006, "loss": 0.1518, "step": 6404 }, { "epoch": 0.41, "learning_rate": 0.00013314828777313663, "loss": 0.1433, "step": 6405 }, { "epoch": 0.41, "learning_rate": 0.0001331286944429277, "loss": 0.1443, "step": 6406 }, { "epoch": 0.41, "learning_rate": 0.00013310909968401887, "loss": 0.1505, "step": 6407 }, { "epoch": 0.41, "learning_rate": 0.00013308950349725504, "loss": 0.1696, "step": 6408 }, { "epoch": 0.41, "learning_rate": 0.00013306990588348144, "loss": 0.1388, "step": 6409 }, { "epoch": 0.41, "learning_rate": 0.00013305030684354315, "loss": 0.1256, "step": 6410 }, { "epoch": 0.41, "learning_rate": 0.00013303070637828542, "loss": 0.1342, "step": 6411 }, { "epoch": 0.41, "learning_rate": 0.00013301110448855357, "loss": 0.1364, "step": 6412 }, { "epoch": 0.41, "learning_rate": 0.00013299150117519287, "loss": 0.1432, "step": 6413 }, { "epoch": 0.41, "learning_rate": 0.00013297189643904882, "loss": 0.1543, "step": 6414 }, { "epoch": 0.41, "learning_rate": 0.00013295229028096678, "loss": 0.1537, "step": 6415 }, { "epoch": 0.41, "learning_rate": 0.0001329326827017924, "loss": 0.1424, "step": 6416 }, { "epoch": 0.41, "learning_rate": 0.00013291307370237116, "loss": 0.1391, "step": 6417 }, { "epoch": 0.41, "learning_rate": 0.0001328934632835488, "loss": 0.1555, "step": 6418 }, { "epoch": 0.41, "learning_rate": 0.00013287385144617096, "loss": 0.1365, "step": 6419 }, { "epoch": 0.41, "learning_rate": 0.0001328542381910835, "loss": 0.1795, "step": 6420 }, { "epoch": 0.41, "learning_rate": 0.00013283462351913215, "loss": 0.1478, "step": 6421 }, { "epoch": 0.41, "learning_rate": 0.00013281500743116294, "loss": 0.1591, "step": 6422 }, { "epoch": 0.41, "learning_rate": 0.0001327953899280217, "loss": 0.145, "step": 6423 }, { "epoch": 0.41, "learning_rate": 0.00013277577101055454, "loss": 0.1433, "step": 6424 }, { "epoch": 0.41, "learning_rate": 0.0001327561506796075, "loss": 0.1486, "step": 6425 }, { "epoch": 0.41, "learning_rate": 0.00013273652893602673, "loss": 0.1429, "step": 6426 }, { "epoch": 0.41, "learning_rate": 0.00013271690578065844, "loss": 0.1432, "step": 6427 }, { "epoch": 0.41, "learning_rate": 0.00013269728121434882, "loss": 0.1374, "step": 6428 }, { "epoch": 0.41, "learning_rate": 0.00013267765523794432, "loss": 0.1377, "step": 6429 }, { "epoch": 0.41, "learning_rate": 0.00013265802785229126, "loss": 0.1738, "step": 6430 }, { "epoch": 0.41, "learning_rate": 0.0001326383990582361, "loss": 0.1441, "step": 6431 }, { "epoch": 0.41, "learning_rate": 0.0001326187688566253, "loss": 0.1743, "step": 6432 }, { "epoch": 0.41, "learning_rate": 0.00013259913724830548, "loss": 0.1452, "step": 6433 }, { "epoch": 0.41, "learning_rate": 0.00013257950423412325, "loss": 0.1571, "step": 6434 }, { "epoch": 0.41, "learning_rate": 0.0001325598698149253, "loss": 0.1636, "step": 6435 }, { "epoch": 0.41, "learning_rate": 0.00013254023399155836, "loss": 0.1316, "step": 6436 }, { "epoch": 0.41, "learning_rate": 0.0001325205967648693, "loss": 0.1474, "step": 6437 }, { "epoch": 0.41, "learning_rate": 0.00013250095813570494, "loss": 0.1373, "step": 6438 }, { "epoch": 0.41, "learning_rate": 0.0001324813181049122, "loss": 0.1396, "step": 6439 }, { "epoch": 0.41, "learning_rate": 0.00013246167667333808, "loss": 0.132, "step": 6440 }, { "epoch": 0.41, "learning_rate": 0.00013244203384182968, "loss": 0.1716, "step": 6441 }, { "epoch": 0.41, "learning_rate": 0.00013242238961123402, "loss": 0.1368, "step": 6442 }, { "epoch": 0.41, "learning_rate": 0.0001324027439823984, "loss": 0.1435, "step": 6443 }, { "epoch": 0.41, "learning_rate": 0.0001323830969561699, "loss": 0.1404, "step": 6444 }, { "epoch": 0.41, "learning_rate": 0.00013236344853339595, "loss": 0.1519, "step": 6445 }, { "epoch": 0.41, "learning_rate": 0.0001323437987149238, "loss": 0.1615, "step": 6446 }, { "epoch": 0.41, "learning_rate": 0.00013232414750160096, "loss": 0.1586, "step": 6447 }, { "epoch": 0.41, "learning_rate": 0.0001323044948942748, "loss": 0.1475, "step": 6448 }, { "epoch": 0.41, "learning_rate": 0.00013228484089379295, "loss": 0.1509, "step": 6449 }, { "epoch": 0.41, "learning_rate": 0.00013226518550100294, "loss": 0.1407, "step": 6450 }, { "epoch": 0.41, "learning_rate": 0.00013224552871675245, "loss": 0.1472, "step": 6451 }, { "epoch": 0.41, "learning_rate": 0.00013222587054188917, "loss": 0.147, "step": 6452 }, { "epoch": 0.41, "learning_rate": 0.0001322062109772609, "loss": 0.1337, "step": 6453 }, { "epoch": 0.41, "learning_rate": 0.00013218655002371547, "loss": 0.1541, "step": 6454 }, { "epoch": 0.41, "learning_rate": 0.00013216688768210073, "loss": 0.1416, "step": 6455 }, { "epoch": 0.41, "learning_rate": 0.0001321472239532647, "loss": 0.1794, "step": 6456 }, { "epoch": 0.41, "learning_rate": 0.00013212755883805537, "loss": 0.1389, "step": 6457 }, { "epoch": 0.41, "learning_rate": 0.0001321078923373208, "loss": 0.1485, "step": 6458 }, { "epoch": 0.41, "learning_rate": 0.0001320882244519091, "loss": 0.1641, "step": 6459 }, { "epoch": 0.41, "learning_rate": 0.00013206855518266855, "loss": 0.1504, "step": 6460 }, { "epoch": 0.41, "learning_rate": 0.0001320488845304473, "loss": 0.1553, "step": 6461 }, { "epoch": 0.41, "learning_rate": 0.0001320292124960937, "loss": 0.1532, "step": 6462 }, { "epoch": 0.41, "learning_rate": 0.00013200953908045615, "loss": 0.1439, "step": 6463 }, { "epoch": 0.41, "learning_rate": 0.00013198986428438305, "loss": 0.1482, "step": 6464 }, { "epoch": 0.41, "learning_rate": 0.00013197018810872288, "loss": 0.1552, "step": 6465 }, { "epoch": 0.41, "learning_rate": 0.00013195051055432422, "loss": 0.17, "step": 6466 }, { "epoch": 0.41, "learning_rate": 0.0001319308316220357, "loss": 0.1841, "step": 6467 }, { "epoch": 0.41, "learning_rate": 0.0001319111513127059, "loss": 0.1587, "step": 6468 }, { "epoch": 0.41, "learning_rate": 0.00013189146962718363, "loss": 0.146, "step": 6469 }, { "epoch": 0.41, "learning_rate": 0.00013187178656631768, "loss": 0.137, "step": 6470 }, { "epoch": 0.41, "learning_rate": 0.00013185210213095683, "loss": 0.1497, "step": 6471 }, { "epoch": 0.41, "learning_rate": 0.00013183241632195, "loss": 0.14, "step": 6472 }, { "epoch": 0.42, "learning_rate": 0.00013181272914014627, "loss": 0.1778, "step": 6473 }, { "epoch": 0.42, "learning_rate": 0.0001317930405863945, "loss": 0.1555, "step": 6474 }, { "epoch": 0.42, "learning_rate": 0.00013177335066154387, "loss": 0.145, "step": 6475 }, { "epoch": 0.42, "learning_rate": 0.0001317536593664435, "loss": 0.136, "step": 6476 }, { "epoch": 0.42, "learning_rate": 0.00013173396670194257, "loss": 0.147, "step": 6477 }, { "epoch": 0.42, "learning_rate": 0.0001317142726688904, "loss": 0.1426, "step": 6478 }, { "epoch": 0.42, "learning_rate": 0.00013169457726813625, "loss": 0.1288, "step": 6479 }, { "epoch": 0.42, "learning_rate": 0.00013167488050052953, "loss": 0.1611, "step": 6480 }, { "epoch": 0.42, "learning_rate": 0.0001316551823669197, "loss": 0.1337, "step": 6481 }, { "epoch": 0.42, "learning_rate": 0.0001316354828681562, "loss": 0.1403, "step": 6482 }, { "epoch": 0.42, "learning_rate": 0.0001316157820050886, "loss": 0.1497, "step": 6483 }, { "epoch": 0.42, "learning_rate": 0.00013159607977856656, "loss": 0.1361, "step": 6484 }, { "epoch": 0.42, "learning_rate": 0.00013157637618943968, "loss": 0.1403, "step": 6485 }, { "epoch": 0.42, "learning_rate": 0.0001315566712385578, "loss": 0.1509, "step": 6486 }, { "epoch": 0.42, "learning_rate": 0.0001315369649267706, "loss": 0.1509, "step": 6487 }, { "epoch": 0.42, "learning_rate": 0.00013151725725492797, "loss": 0.1435, "step": 6488 }, { "epoch": 0.42, "learning_rate": 0.00013149754822387982, "loss": 0.1568, "step": 6489 }, { "epoch": 0.42, "learning_rate": 0.00013147783783447613, "loss": 0.1588, "step": 6490 }, { "epoch": 0.42, "learning_rate": 0.00013145812608756694, "loss": 0.1591, "step": 6491 }, { "epoch": 0.42, "learning_rate": 0.00013143841298400227, "loss": 0.1343, "step": 6492 }, { "epoch": 0.42, "learning_rate": 0.00013141869852463232, "loss": 0.1549, "step": 6493 }, { "epoch": 0.42, "learning_rate": 0.0001313989827103073, "loss": 0.1614, "step": 6494 }, { "epoch": 0.42, "learning_rate": 0.00013137926554187738, "loss": 0.1312, "step": 6495 }, { "epoch": 0.42, "learning_rate": 0.00013135954702019298, "loss": 0.1422, "step": 6496 }, { "epoch": 0.42, "learning_rate": 0.00013133982714610444, "loss": 0.168, "step": 6497 }, { "epoch": 0.42, "learning_rate": 0.0001313201059204622, "loss": 0.1571, "step": 6498 }, { "epoch": 0.42, "learning_rate": 0.00013130038334411667, "loss": 0.1399, "step": 6499 }, { "epoch": 0.42, "learning_rate": 0.00013128065941791857, "loss": 0.164, "step": 6500 }, { "epoch": 0.42, "learning_rate": 0.00013126093414271835, "loss": 0.1573, "step": 6501 }, { "epoch": 0.42, "learning_rate": 0.00013124120751936675, "loss": 0.1608, "step": 6502 }, { "epoch": 0.42, "learning_rate": 0.0001312214795487145, "loss": 0.1368, "step": 6503 }, { "epoch": 0.42, "learning_rate": 0.00013120175023161235, "loss": 0.1519, "step": 6504 }, { "epoch": 0.42, "learning_rate": 0.0001311820195689112, "loss": 0.1517, "step": 6505 }, { "epoch": 0.42, "learning_rate": 0.00013116228756146188, "loss": 0.1475, "step": 6506 }, { "epoch": 0.42, "learning_rate": 0.0001311425542101154, "loss": 0.1395, "step": 6507 }, { "epoch": 0.42, "learning_rate": 0.0001311228195157227, "loss": 0.1456, "step": 6508 }, { "epoch": 0.42, "learning_rate": 0.00013110308347913496, "loss": 0.1807, "step": 6509 }, { "epoch": 0.42, "learning_rate": 0.00013108334610120324, "loss": 0.1624, "step": 6510 }, { "epoch": 0.42, "learning_rate": 0.00013106360738277879, "loss": 0.1517, "step": 6511 }, { "epoch": 0.42, "learning_rate": 0.00013104386732471274, "loss": 0.1367, "step": 6512 }, { "epoch": 0.42, "learning_rate": 0.00013102412592785653, "loss": 0.1566, "step": 6513 }, { "epoch": 0.42, "learning_rate": 0.00013100438319306148, "loss": 0.1474, "step": 6514 }, { "epoch": 0.42, "learning_rate": 0.00013098463912117892, "loss": 0.1547, "step": 6515 }, { "epoch": 0.42, "learning_rate": 0.00013096489371306047, "loss": 0.1728, "step": 6516 }, { "epoch": 0.42, "learning_rate": 0.00013094514696955755, "loss": 0.1495, "step": 6517 }, { "epoch": 0.42, "learning_rate": 0.00013092539889152183, "loss": 0.1314, "step": 6518 }, { "epoch": 0.42, "learning_rate": 0.00013090564947980493, "loss": 0.1313, "step": 6519 }, { "epoch": 0.42, "learning_rate": 0.00013088589873525857, "loss": 0.1591, "step": 6520 }, { "epoch": 0.42, "learning_rate": 0.0001308661466587345, "loss": 0.1575, "step": 6521 }, { "epoch": 0.42, "learning_rate": 0.0001308463932510845, "loss": 0.1466, "step": 6522 }, { "epoch": 0.42, "learning_rate": 0.00013082663851316054, "loss": 0.1537, "step": 6523 }, { "epoch": 0.42, "learning_rate": 0.00013080688244581452, "loss": 0.1364, "step": 6524 }, { "epoch": 0.42, "learning_rate": 0.00013078712504989843, "loss": 0.1518, "step": 6525 }, { "epoch": 0.42, "learning_rate": 0.0001307673663262643, "loss": 0.1534, "step": 6526 }, { "epoch": 0.42, "learning_rate": 0.0001307476062757643, "loss": 0.1368, "step": 6527 }, { "epoch": 0.42, "learning_rate": 0.00013072784489925055, "loss": 0.1392, "step": 6528 }, { "epoch": 0.42, "learning_rate": 0.0001307080821975753, "loss": 0.1495, "step": 6529 }, { "epoch": 0.42, "learning_rate": 0.0001306883181715908, "loss": 0.1407, "step": 6530 }, { "epoch": 0.42, "learning_rate": 0.0001306685528221494, "loss": 0.144, "step": 6531 }, { "epoch": 0.42, "learning_rate": 0.00013064878615010348, "loss": 0.1615, "step": 6532 }, { "epoch": 0.42, "learning_rate": 0.00013062901815630556, "loss": 0.1537, "step": 6533 }, { "epoch": 0.42, "learning_rate": 0.00013060924884160814, "loss": 0.1265, "step": 6534 }, { "epoch": 0.42, "learning_rate": 0.00013058947820686367, "loss": 0.1346, "step": 6535 }, { "epoch": 0.42, "learning_rate": 0.00013056970625292488, "loss": 0.1591, "step": 6536 }, { "epoch": 0.42, "learning_rate": 0.00013054993298064447, "loss": 0.1777, "step": 6537 }, { "epoch": 0.42, "learning_rate": 0.00013053015839087508, "loss": 0.1418, "step": 6538 }, { "epoch": 0.42, "learning_rate": 0.00013051038248446958, "loss": 0.1383, "step": 6539 }, { "epoch": 0.42, "learning_rate": 0.00013049060526228078, "loss": 0.158, "step": 6540 }, { "epoch": 0.42, "learning_rate": 0.00013047082672516163, "loss": 0.1731, "step": 6541 }, { "epoch": 0.42, "learning_rate": 0.00013045104687396504, "loss": 0.1747, "step": 6542 }, { "epoch": 0.42, "learning_rate": 0.00013043126570954407, "loss": 0.1308, "step": 6543 }, { "epoch": 0.42, "learning_rate": 0.0001304114832327518, "loss": 0.1396, "step": 6544 }, { "epoch": 0.42, "learning_rate": 0.00013039169944444135, "loss": 0.1294, "step": 6545 }, { "epoch": 0.42, "learning_rate": 0.0001303719143454659, "loss": 0.156, "step": 6546 }, { "epoch": 0.42, "learning_rate": 0.0001303521279366787, "loss": 0.1384, "step": 6547 }, { "epoch": 0.42, "learning_rate": 0.00013033234021893312, "loss": 0.1484, "step": 6548 }, { "epoch": 0.42, "learning_rate": 0.0001303125511930824, "loss": 0.1697, "step": 6549 }, { "epoch": 0.42, "learning_rate": 0.0001302927608599801, "loss": 0.1647, "step": 6550 }, { "epoch": 0.42, "learning_rate": 0.00013027296922047952, "loss": 0.1392, "step": 6551 }, { "epoch": 0.42, "learning_rate": 0.00013025317627543431, "loss": 0.1318, "step": 6552 }, { "epoch": 0.42, "learning_rate": 0.00013023338202569805, "loss": 0.1663, "step": 6553 }, { "epoch": 0.42, "learning_rate": 0.00013021358647212437, "loss": 0.144, "step": 6554 }, { "epoch": 0.42, "learning_rate": 0.00013019378961556696, "loss": 0.1395, "step": 6555 }, { "epoch": 0.42, "learning_rate": 0.00013017399145687957, "loss": 0.1415, "step": 6556 }, { "epoch": 0.42, "learning_rate": 0.000130154191996916, "loss": 0.165, "step": 6557 }, { "epoch": 0.42, "learning_rate": 0.00013013439123653016, "loss": 0.1476, "step": 6558 }, { "epoch": 0.42, "learning_rate": 0.00013011458917657588, "loss": 0.1496, "step": 6559 }, { "epoch": 0.42, "learning_rate": 0.00013009478581790725, "loss": 0.1554, "step": 6560 }, { "epoch": 0.42, "learning_rate": 0.0001300749811613783, "loss": 0.1671, "step": 6561 }, { "epoch": 0.42, "learning_rate": 0.000130055175207843, "loss": 0.1465, "step": 6562 }, { "epoch": 0.42, "learning_rate": 0.00013003536795815561, "loss": 0.1509, "step": 6563 }, { "epoch": 0.42, "learning_rate": 0.0001300155594131703, "loss": 0.152, "step": 6564 }, { "epoch": 0.42, "learning_rate": 0.0001299957495737413, "loss": 0.1486, "step": 6565 }, { "epoch": 0.42, "learning_rate": 0.00012997593844072297, "loss": 0.1467, "step": 6566 }, { "epoch": 0.42, "learning_rate": 0.00012995612601496967, "loss": 0.1466, "step": 6567 }, { "epoch": 0.42, "learning_rate": 0.00012993631229733582, "loss": 0.1293, "step": 6568 }, { "epoch": 0.42, "learning_rate": 0.00012991649728867586, "loss": 0.1386, "step": 6569 }, { "epoch": 0.42, "learning_rate": 0.0001298966809898444, "loss": 0.1811, "step": 6570 }, { "epoch": 0.42, "learning_rate": 0.000129876863401696, "loss": 0.1483, "step": 6571 }, { "epoch": 0.42, "learning_rate": 0.00012985704452508527, "loss": 0.162, "step": 6572 }, { "epoch": 0.42, "learning_rate": 0.00012983722436086699, "loss": 0.1716, "step": 6573 }, { "epoch": 0.42, "learning_rate": 0.00012981740290989586, "loss": 0.1521, "step": 6574 }, { "epoch": 0.42, "learning_rate": 0.00012979758017302672, "loss": 0.1329, "step": 6575 }, { "epoch": 0.42, "learning_rate": 0.00012977775615111443, "loss": 0.1491, "step": 6576 }, { "epoch": 0.42, "learning_rate": 0.00012975793084501393, "loss": 0.1711, "step": 6577 }, { "epoch": 0.42, "learning_rate": 0.0001297381042555802, "loss": 0.1777, "step": 6578 }, { "epoch": 0.42, "learning_rate": 0.00012971827638366822, "loss": 0.1497, "step": 6579 }, { "epoch": 0.42, "learning_rate": 0.00012969844723013317, "loss": 0.1438, "step": 6580 }, { "epoch": 0.42, "learning_rate": 0.00012967861679583016, "loss": 0.1772, "step": 6581 }, { "epoch": 0.42, "learning_rate": 0.0001296587850816144, "loss": 0.1395, "step": 6582 }, { "epoch": 0.42, "learning_rate": 0.00012963895208834112, "loss": 0.1573, "step": 6583 }, { "epoch": 0.42, "learning_rate": 0.00012961911781686567, "loss": 0.1496, "step": 6584 }, { "epoch": 0.42, "learning_rate": 0.00012959928226804337, "loss": 0.1559, "step": 6585 }, { "epoch": 0.42, "learning_rate": 0.00012957944544272969, "loss": 0.148, "step": 6586 }, { "epoch": 0.42, "learning_rate": 0.0001295596073417801, "loss": 0.1471, "step": 6587 }, { "epoch": 0.42, "learning_rate": 0.00012953976796605012, "loss": 0.1518, "step": 6588 }, { "epoch": 0.42, "learning_rate": 0.00012951992731639533, "loss": 0.1437, "step": 6589 }, { "epoch": 0.42, "learning_rate": 0.00012950008539367139, "loss": 0.1474, "step": 6590 }, { "epoch": 0.42, "learning_rate": 0.00012948024219873398, "loss": 0.1641, "step": 6591 }, { "epoch": 0.42, "learning_rate": 0.00012946039773243889, "loss": 0.1408, "step": 6592 }, { "epoch": 0.42, "learning_rate": 0.00012944055199564186, "loss": 0.1507, "step": 6593 }, { "epoch": 0.42, "learning_rate": 0.00012942070498919882, "loss": 0.1481, "step": 6594 }, { "epoch": 0.42, "learning_rate": 0.0001294008567139657, "loss": 0.1368, "step": 6595 }, { "epoch": 0.42, "learning_rate": 0.00012938100717079838, "loss": 0.1396, "step": 6596 }, { "epoch": 0.42, "learning_rate": 0.00012936115636055296, "loss": 0.1595, "step": 6597 }, { "epoch": 0.42, "learning_rate": 0.0001293413042840855, "loss": 0.1467, "step": 6598 }, { "epoch": 0.42, "learning_rate": 0.00012932145094225213, "loss": 0.1289, "step": 6599 }, { "epoch": 0.42, "learning_rate": 0.00012930159633590908, "loss": 0.1357, "step": 6600 }, { "epoch": 0.42, "learning_rate": 0.00012928174046591251, "loss": 0.1529, "step": 6601 }, { "epoch": 0.42, "learning_rate": 0.00012926188333311884, "loss": 0.1808, "step": 6602 }, { "epoch": 0.42, "learning_rate": 0.00012924202493838432, "loss": 0.1421, "step": 6603 }, { "epoch": 0.42, "learning_rate": 0.00012922216528256538, "loss": 0.1644, "step": 6604 }, { "epoch": 0.42, "learning_rate": 0.00012920230436651853, "loss": 0.1457, "step": 6605 }, { "epoch": 0.42, "learning_rate": 0.0001291824421911002, "loss": 0.159, "step": 6606 }, { "epoch": 0.42, "learning_rate": 0.00012916257875716705, "loss": 0.1188, "step": 6607 }, { "epoch": 0.42, "learning_rate": 0.0001291427140655757, "loss": 0.1452, "step": 6608 }, { "epoch": 0.42, "learning_rate": 0.00012912284811718275, "loss": 0.149, "step": 6609 }, { "epoch": 0.42, "learning_rate": 0.00012910298091284503, "loss": 0.1438, "step": 6610 }, { "epoch": 0.42, "learning_rate": 0.00012908311245341927, "loss": 0.1401, "step": 6611 }, { "epoch": 0.42, "learning_rate": 0.00012906324273976233, "loss": 0.1613, "step": 6612 }, { "epoch": 0.42, "learning_rate": 0.00012904337177273105, "loss": 0.1313, "step": 6613 }, { "epoch": 0.42, "learning_rate": 0.00012902349955318246, "loss": 0.1476, "step": 6614 }, { "epoch": 0.42, "learning_rate": 0.00012900362608197355, "loss": 0.1481, "step": 6615 }, { "epoch": 0.42, "learning_rate": 0.00012898375135996136, "loss": 0.1282, "step": 6616 }, { "epoch": 0.42, "learning_rate": 0.000128963875388003, "loss": 0.1622, "step": 6617 }, { "epoch": 0.42, "learning_rate": 0.00012894399816695565, "loss": 0.1504, "step": 6618 }, { "epoch": 0.42, "learning_rate": 0.00012892411969767654, "loss": 0.1551, "step": 6619 }, { "epoch": 0.42, "learning_rate": 0.0001289042399810229, "loss": 0.1453, "step": 6620 }, { "epoch": 0.42, "learning_rate": 0.00012888435901785213, "loss": 0.1767, "step": 6621 }, { "epoch": 0.42, "learning_rate": 0.0001288644768090215, "loss": 0.1298, "step": 6622 }, { "epoch": 0.42, "learning_rate": 0.00012884459335538856, "loss": 0.1385, "step": 6623 }, { "epoch": 0.42, "learning_rate": 0.00012882470865781075, "loss": 0.1536, "step": 6624 }, { "epoch": 0.42, "learning_rate": 0.00012880482271714563, "loss": 0.1633, "step": 6625 }, { "epoch": 0.42, "learning_rate": 0.00012878493553425071, "loss": 0.1371, "step": 6626 }, { "epoch": 0.42, "learning_rate": 0.00012876504710998375, "loss": 0.1708, "step": 6627 }, { "epoch": 0.42, "learning_rate": 0.00012874515744520243, "loss": 0.1491, "step": 6628 }, { "epoch": 0.43, "learning_rate": 0.0001287252665407645, "loss": 0.1539, "step": 6629 }, { "epoch": 0.43, "learning_rate": 0.00012870537439752772, "loss": 0.1416, "step": 6630 }, { "epoch": 0.43, "learning_rate": 0.00012868548101635003, "loss": 0.1759, "step": 6631 }, { "epoch": 0.43, "learning_rate": 0.0001286655863980893, "loss": 0.1412, "step": 6632 }, { "epoch": 0.43, "learning_rate": 0.00012864569054360349, "loss": 0.1535, "step": 6633 }, { "epoch": 0.43, "learning_rate": 0.00012862579345375064, "loss": 0.1405, "step": 6634 }, { "epoch": 0.43, "learning_rate": 0.00012860589512938888, "loss": 0.1474, "step": 6635 }, { "epoch": 0.43, "learning_rate": 0.00012858599557137628, "loss": 0.147, "step": 6636 }, { "epoch": 0.43, "learning_rate": 0.000128566094780571, "loss": 0.1527, "step": 6637 }, { "epoch": 0.43, "learning_rate": 0.00012854619275783134, "loss": 0.1673, "step": 6638 }, { "epoch": 0.43, "learning_rate": 0.00012852628950401555, "loss": 0.1373, "step": 6639 }, { "epoch": 0.43, "learning_rate": 0.000128506385019982, "loss": 0.1604, "step": 6640 }, { "epoch": 0.43, "learning_rate": 0.00012848647930658905, "loss": 0.1446, "step": 6641 }, { "epoch": 0.43, "learning_rate": 0.0001284665723646952, "loss": 0.1605, "step": 6642 }, { "epoch": 0.43, "learning_rate": 0.0001284466641951589, "loss": 0.1852, "step": 6643 }, { "epoch": 0.43, "learning_rate": 0.00012842675479883875, "loss": 0.1636, "step": 6644 }, { "epoch": 0.43, "learning_rate": 0.0001284068441765933, "loss": 0.1405, "step": 6645 }, { "epoch": 0.43, "learning_rate": 0.00012838693232928127, "loss": 0.1678, "step": 6646 }, { "epoch": 0.43, "learning_rate": 0.00012836701925776135, "loss": 0.152, "step": 6647 }, { "epoch": 0.43, "learning_rate": 0.00012834710496289228, "loss": 0.1407, "step": 6648 }, { "epoch": 0.43, "learning_rate": 0.00012832718944553296, "loss": 0.1354, "step": 6649 }, { "epoch": 0.43, "learning_rate": 0.00012830727270654214, "loss": 0.1483, "step": 6650 }, { "epoch": 0.43, "learning_rate": 0.00012828735474677884, "loss": 0.1621, "step": 6651 }, { "epoch": 0.43, "learning_rate": 0.00012826743556710202, "loss": 0.1481, "step": 6652 }, { "epoch": 0.43, "learning_rate": 0.00012824751516837067, "loss": 0.1463, "step": 6653 }, { "epoch": 0.43, "learning_rate": 0.0001282275935514439, "loss": 0.1562, "step": 6654 }, { "epoch": 0.43, "learning_rate": 0.00012820767071718088, "loss": 0.1601, "step": 6655 }, { "epoch": 0.43, "learning_rate": 0.0001281877466664407, "loss": 0.1585, "step": 6656 }, { "epoch": 0.43, "learning_rate": 0.0001281678214000827, "loss": 0.1603, "step": 6657 }, { "epoch": 0.43, "learning_rate": 0.00012814789491896614, "loss": 0.1718, "step": 6658 }, { "epoch": 0.43, "learning_rate": 0.00012812796722395036, "loss": 0.1446, "step": 6659 }, { "epoch": 0.43, "learning_rate": 0.00012810803831589474, "loss": 0.1572, "step": 6660 }, { "epoch": 0.43, "learning_rate": 0.00012808810819565876, "loss": 0.1547, "step": 6661 }, { "epoch": 0.43, "learning_rate": 0.00012806817686410188, "loss": 0.1747, "step": 6662 }, { "epoch": 0.43, "learning_rate": 0.00012804824432208366, "loss": 0.1613, "step": 6663 }, { "epoch": 0.43, "learning_rate": 0.00012802831057046377, "loss": 0.1716, "step": 6664 }, { "epoch": 0.43, "learning_rate": 0.00012800837561010184, "loss": 0.1621, "step": 6665 }, { "epoch": 0.43, "learning_rate": 0.0001279884394418575, "loss": 0.1571, "step": 6666 }, { "epoch": 0.43, "learning_rate": 0.00012796850206659062, "loss": 0.1577, "step": 6667 }, { "epoch": 0.43, "learning_rate": 0.00012794856348516095, "loss": 0.1786, "step": 6668 }, { "epoch": 0.43, "learning_rate": 0.0001279286236984284, "loss": 0.1516, "step": 6669 }, { "epoch": 0.43, "learning_rate": 0.00012790868270725283, "loss": 0.1382, "step": 6670 }, { "epoch": 0.43, "learning_rate": 0.00012788874051249425, "loss": 0.1687, "step": 6671 }, { "epoch": 0.43, "learning_rate": 0.00012786879711501273, "loss": 0.1372, "step": 6672 }, { "epoch": 0.43, "learning_rate": 0.00012784885251566823, "loss": 0.1653, "step": 6673 }, { "epoch": 0.43, "learning_rate": 0.00012782890671532096, "loss": 0.1296, "step": 6674 }, { "epoch": 0.43, "learning_rate": 0.00012780895971483108, "loss": 0.1557, "step": 6675 }, { "epoch": 0.43, "learning_rate": 0.0001277890115150588, "loss": 0.1289, "step": 6676 }, { "epoch": 0.43, "learning_rate": 0.00012776906211686443, "loss": 0.1286, "step": 6677 }, { "epoch": 0.43, "learning_rate": 0.0001277491115211083, "loss": 0.1701, "step": 6678 }, { "epoch": 0.43, "learning_rate": 0.00012772915972865073, "loss": 0.1349, "step": 6679 }, { "epoch": 0.43, "learning_rate": 0.00012770920674035226, "loss": 0.15, "step": 6680 }, { "epoch": 0.43, "learning_rate": 0.0001276892525570733, "loss": 0.1525, "step": 6681 }, { "epoch": 0.43, "learning_rate": 0.0001276692971796744, "loss": 0.1459, "step": 6682 }, { "epoch": 0.43, "learning_rate": 0.0001276493406090162, "loss": 0.1528, "step": 6683 }, { "epoch": 0.43, "learning_rate": 0.00012762938284595931, "loss": 0.1524, "step": 6684 }, { "epoch": 0.43, "learning_rate": 0.0001276094238913644, "loss": 0.1516, "step": 6685 }, { "epoch": 0.43, "learning_rate": 0.00012758946374609223, "loss": 0.1531, "step": 6686 }, { "epoch": 0.43, "learning_rate": 0.00012756950241100365, "loss": 0.133, "step": 6687 }, { "epoch": 0.43, "learning_rate": 0.00012754953988695943, "loss": 0.1444, "step": 6688 }, { "epoch": 0.43, "learning_rate": 0.0001275295761748205, "loss": 0.1289, "step": 6689 }, { "epoch": 0.43, "learning_rate": 0.0001275096112754478, "loss": 0.1758, "step": 6690 }, { "epoch": 0.43, "learning_rate": 0.0001274896451897024, "loss": 0.1385, "step": 6691 }, { "epoch": 0.43, "learning_rate": 0.00012746967791844524, "loss": 0.1432, "step": 6692 }, { "epoch": 0.43, "learning_rate": 0.00012744970946253746, "loss": 0.1436, "step": 6693 }, { "epoch": 0.43, "learning_rate": 0.00012742973982284028, "loss": 0.1562, "step": 6694 }, { "epoch": 0.43, "learning_rate": 0.00012740976900021484, "loss": 0.1366, "step": 6695 }, { "epoch": 0.43, "learning_rate": 0.0001273897969955224, "loss": 0.1533, "step": 6696 }, { "epoch": 0.43, "learning_rate": 0.0001273698238096243, "loss": 0.1375, "step": 6697 }, { "epoch": 0.43, "learning_rate": 0.0001273498494433819, "loss": 0.1418, "step": 6698 }, { "epoch": 0.43, "learning_rate": 0.00012732987389765658, "loss": 0.1582, "step": 6699 }, { "epoch": 0.43, "learning_rate": 0.00012730989717330978, "loss": 0.1595, "step": 6700 }, { "epoch": 0.43, "learning_rate": 0.00012728991927120307, "loss": 0.1447, "step": 6701 }, { "epoch": 0.43, "learning_rate": 0.00012726994019219797, "loss": 0.1384, "step": 6702 }, { "epoch": 0.43, "learning_rate": 0.00012724995993715612, "loss": 0.1507, "step": 6703 }, { "epoch": 0.43, "learning_rate": 0.00012722997850693916, "loss": 0.145, "step": 6704 }, { "epoch": 0.43, "learning_rate": 0.00012720999590240886, "loss": 0.158, "step": 6705 }, { "epoch": 0.43, "learning_rate": 0.0001271900121244269, "loss": 0.1491, "step": 6706 }, { "epoch": 0.43, "learning_rate": 0.00012717002717385514, "loss": 0.1492, "step": 6707 }, { "epoch": 0.43, "learning_rate": 0.00012715004105155544, "loss": 0.1348, "step": 6708 }, { "epoch": 0.43, "learning_rate": 0.00012713005375838973, "loss": 0.1472, "step": 6709 }, { "epoch": 0.43, "learning_rate": 0.00012711006529521994, "loss": 0.1378, "step": 6710 }, { "epoch": 0.43, "learning_rate": 0.00012709007566290813, "loss": 0.1386, "step": 6711 }, { "epoch": 0.43, "learning_rate": 0.0001270700848623164, "loss": 0.1589, "step": 6712 }, { "epoch": 0.43, "learning_rate": 0.00012705009289430675, "loss": 0.1362, "step": 6713 }, { "epoch": 0.43, "learning_rate": 0.00012703009975974144, "loss": 0.1413, "step": 6714 }, { "epoch": 0.43, "learning_rate": 0.0001270101054594827, "loss": 0.131, "step": 6715 }, { "epoch": 0.43, "learning_rate": 0.00012699010999439275, "loss": 0.1345, "step": 6716 }, { "epoch": 0.43, "learning_rate": 0.0001269701133653339, "loss": 0.1378, "step": 6717 }, { "epoch": 0.43, "learning_rate": 0.00012695011557316857, "loss": 0.1415, "step": 6718 }, { "epoch": 0.43, "learning_rate": 0.00012693011661875914, "loss": 0.144, "step": 6719 }, { "epoch": 0.43, "learning_rate": 0.00012691011650296812, "loss": 0.1283, "step": 6720 }, { "epoch": 0.43, "learning_rate": 0.000126890115226658, "loss": 0.1434, "step": 6721 }, { "epoch": 0.43, "learning_rate": 0.00012687011279069132, "loss": 0.1577, "step": 6722 }, { "epoch": 0.43, "learning_rate": 0.00012685010919593077, "loss": 0.1416, "step": 6723 }, { "epoch": 0.43, "learning_rate": 0.00012683010444323897, "loss": 0.1271, "step": 6724 }, { "epoch": 0.43, "learning_rate": 0.00012681009853347873, "loss": 0.1339, "step": 6725 }, { "epoch": 0.43, "learning_rate": 0.00012679009146751265, "loss": 0.1451, "step": 6726 }, { "epoch": 0.43, "learning_rate": 0.00012677008324620367, "loss": 0.1372, "step": 6727 }, { "epoch": 0.43, "learning_rate": 0.00012675007387041466, "loss": 0.1422, "step": 6728 }, { "epoch": 0.43, "learning_rate": 0.00012673006334100852, "loss": 0.1483, "step": 6729 }, { "epoch": 0.43, "learning_rate": 0.00012671005165884816, "loss": 0.1379, "step": 6730 }, { "epoch": 0.43, "learning_rate": 0.0001266900388247967, "loss": 0.1298, "step": 6731 }, { "epoch": 0.43, "learning_rate": 0.0001266700248397172, "loss": 0.1345, "step": 6732 }, { "epoch": 0.43, "learning_rate": 0.00012665000970447267, "loss": 0.1382, "step": 6733 }, { "epoch": 0.43, "learning_rate": 0.00012662999341992638, "loss": 0.1878, "step": 6734 }, { "epoch": 0.43, "learning_rate": 0.00012660997598694154, "loss": 0.1796, "step": 6735 }, { "epoch": 0.43, "learning_rate": 0.00012658995740638136, "loss": 0.136, "step": 6736 }, { "epoch": 0.43, "learning_rate": 0.00012656993767910915, "loss": 0.1365, "step": 6737 }, { "epoch": 0.43, "learning_rate": 0.00012654991680598838, "loss": 0.1856, "step": 6738 }, { "epoch": 0.43, "learning_rate": 0.0001265298947878824, "loss": 0.1525, "step": 6739 }, { "epoch": 0.43, "learning_rate": 0.00012650987162565465, "loss": 0.1408, "step": 6740 }, { "epoch": 0.43, "learning_rate": 0.00012648984732016867, "loss": 0.1241, "step": 6741 }, { "epoch": 0.43, "learning_rate": 0.00012646982187228804, "loss": 0.1447, "step": 6742 }, { "epoch": 0.43, "learning_rate": 0.0001264497952828763, "loss": 0.1655, "step": 6743 }, { "epoch": 0.43, "learning_rate": 0.00012642976755279723, "loss": 0.1502, "step": 6744 }, { "epoch": 0.43, "learning_rate": 0.00012640973868291447, "loss": 0.144, "step": 6745 }, { "epoch": 0.43, "learning_rate": 0.00012638970867409177, "loss": 0.1329, "step": 6746 }, { "epoch": 0.43, "learning_rate": 0.00012636967752719294, "loss": 0.1539, "step": 6747 }, { "epoch": 0.43, "learning_rate": 0.0001263496452430819, "loss": 0.1374, "step": 6748 }, { "epoch": 0.43, "learning_rate": 0.0001263296118226225, "loss": 0.1416, "step": 6749 }, { "epoch": 0.43, "learning_rate": 0.00012630957726667867, "loss": 0.175, "step": 6750 }, { "epoch": 0.43, "learning_rate": 0.0001262895415761145, "loss": 0.1454, "step": 6751 }, { "epoch": 0.43, "learning_rate": 0.00012626950475179397, "loss": 0.1413, "step": 6752 }, { "epoch": 0.43, "learning_rate": 0.00012624946679458122, "loss": 0.1606, "step": 6753 }, { "epoch": 0.43, "learning_rate": 0.0001262294277053404, "loss": 0.1395, "step": 6754 }, { "epoch": 0.43, "learning_rate": 0.00012620938748493572, "loss": 0.163, "step": 6755 }, { "epoch": 0.43, "learning_rate": 0.00012618934613423138, "loss": 0.165, "step": 6756 }, { "epoch": 0.43, "learning_rate": 0.00012616930365409173, "loss": 0.1452, "step": 6757 }, { "epoch": 0.43, "learning_rate": 0.00012614926004538108, "loss": 0.1627, "step": 6758 }, { "epoch": 0.43, "learning_rate": 0.00012612921530896388, "loss": 0.1393, "step": 6759 }, { "epoch": 0.43, "learning_rate": 0.00012610916944570452, "loss": 0.1454, "step": 6760 }, { "epoch": 0.43, "learning_rate": 0.00012608912245646754, "loss": 0.1555, "step": 6761 }, { "epoch": 0.43, "learning_rate": 0.00012606907434211743, "loss": 0.1322, "step": 6762 }, { "epoch": 0.43, "learning_rate": 0.0001260490251035188, "loss": 0.1396, "step": 6763 }, { "epoch": 0.43, "learning_rate": 0.00012602897474153627, "loss": 0.1546, "step": 6764 }, { "epoch": 0.43, "learning_rate": 0.0001260089232570346, "loss": 0.1629, "step": 6765 }, { "epoch": 0.43, "learning_rate": 0.00012598887065087846, "loss": 0.1433, "step": 6766 }, { "epoch": 0.43, "learning_rate": 0.00012596881692393262, "loss": 0.1397, "step": 6767 }, { "epoch": 0.43, "learning_rate": 0.00012594876207706199, "loss": 0.1802, "step": 6768 }, { "epoch": 0.43, "learning_rate": 0.0001259287061111314, "loss": 0.1364, "step": 6769 }, { "epoch": 0.43, "learning_rate": 0.0001259086490270057, "loss": 0.1603, "step": 6770 }, { "epoch": 0.43, "learning_rate": 0.00012588859082555005, "loss": 0.1591, "step": 6771 }, { "epoch": 0.43, "learning_rate": 0.0001258685315076293, "loss": 0.1561, "step": 6772 }, { "epoch": 0.43, "learning_rate": 0.00012584847107410864, "loss": 0.1442, "step": 6773 }, { "epoch": 0.43, "learning_rate": 0.00012582840952585315, "loss": 0.1599, "step": 6774 }, { "epoch": 0.43, "learning_rate": 0.00012580834686372797, "loss": 0.1511, "step": 6775 }, { "epoch": 0.43, "learning_rate": 0.00012578828308859835, "loss": 0.174, "step": 6776 }, { "epoch": 0.43, "learning_rate": 0.00012576821820132957, "loss": 0.1567, "step": 6777 }, { "epoch": 0.43, "learning_rate": 0.00012574815220278688, "loss": 0.1376, "step": 6778 }, { "epoch": 0.43, "learning_rate": 0.0001257280850938357, "loss": 0.1534, "step": 6779 }, { "epoch": 0.43, "learning_rate": 0.00012570801687534147, "loss": 0.1371, "step": 6780 }, { "epoch": 0.43, "learning_rate": 0.00012568794754816954, "loss": 0.1563, "step": 6781 }, { "epoch": 0.43, "learning_rate": 0.0001256678771131855, "loss": 0.1484, "step": 6782 }, { "epoch": 0.43, "learning_rate": 0.00012564780557125488, "loss": 0.1671, "step": 6783 }, { "epoch": 0.43, "learning_rate": 0.0001256277329232433, "loss": 0.1438, "step": 6784 }, { "epoch": 0.44, "learning_rate": 0.00012560765917001633, "loss": 0.1469, "step": 6785 }, { "epoch": 0.44, "learning_rate": 0.00012558758431243974, "loss": 0.1656, "step": 6786 }, { "epoch": 0.44, "learning_rate": 0.0001255675083513793, "loss": 0.1347, "step": 6787 }, { "epoch": 0.44, "learning_rate": 0.00012554743128770073, "loss": 0.1598, "step": 6788 }, { "epoch": 0.44, "learning_rate": 0.0001255273531222699, "loss": 0.1534, "step": 6789 }, { "epoch": 0.44, "learning_rate": 0.00012550727385595266, "loss": 0.1346, "step": 6790 }, { "epoch": 0.44, "learning_rate": 0.000125487193489615, "loss": 0.1636, "step": 6791 }, { "epoch": 0.44, "learning_rate": 0.00012546711202412287, "loss": 0.1521, "step": 6792 }, { "epoch": 0.44, "learning_rate": 0.00012544702946034234, "loss": 0.1666, "step": 6793 }, { "epoch": 0.44, "learning_rate": 0.0001254269457991394, "loss": 0.1453, "step": 6794 }, { "epoch": 0.44, "learning_rate": 0.00012540686104138023, "loss": 0.1441, "step": 6795 }, { "epoch": 0.44, "learning_rate": 0.00012538677518793103, "loss": 0.1821, "step": 6796 }, { "epoch": 0.44, "learning_rate": 0.00012536668823965794, "loss": 0.1286, "step": 6797 }, { "epoch": 0.44, "learning_rate": 0.00012534660019742727, "loss": 0.149, "step": 6798 }, { "epoch": 0.44, "learning_rate": 0.00012532651106210535, "loss": 0.145, "step": 6799 }, { "epoch": 0.44, "learning_rate": 0.0001253064208345585, "loss": 0.1422, "step": 6800 }, { "epoch": 0.44, "learning_rate": 0.0001252863295156532, "loss": 0.1744, "step": 6801 }, { "epoch": 0.44, "learning_rate": 0.00012526623710625576, "loss": 0.1434, "step": 6802 }, { "epoch": 0.44, "learning_rate": 0.00012524614360723282, "loss": 0.1518, "step": 6803 }, { "epoch": 0.44, "learning_rate": 0.00012522604901945084, "loss": 0.156, "step": 6804 }, { "epoch": 0.44, "learning_rate": 0.00012520595334377646, "loss": 0.1623, "step": 6805 }, { "epoch": 0.44, "learning_rate": 0.00012518585658107632, "loss": 0.139, "step": 6806 }, { "epoch": 0.44, "learning_rate": 0.00012516575873221708, "loss": 0.1643, "step": 6807 }, { "epoch": 0.44, "learning_rate": 0.00012514565979806553, "loss": 0.142, "step": 6808 }, { "epoch": 0.44, "learning_rate": 0.00012512555977948838, "loss": 0.14, "step": 6809 }, { "epoch": 0.44, "learning_rate": 0.00012510545867735248, "loss": 0.1498, "step": 6810 }, { "epoch": 0.44, "learning_rate": 0.00012508535649252472, "loss": 0.128, "step": 6811 }, { "epoch": 0.44, "learning_rate": 0.00012506525322587207, "loss": 0.1513, "step": 6812 }, { "epoch": 0.44, "learning_rate": 0.0001250451488782614, "loss": 0.1389, "step": 6813 }, { "epoch": 0.44, "learning_rate": 0.00012502504345055982, "loss": 0.1571, "step": 6814 }, { "epoch": 0.44, "learning_rate": 0.00012500493694363432, "loss": 0.1761, "step": 6815 }, { "epoch": 0.44, "learning_rate": 0.000124984829358352, "loss": 0.1532, "step": 6816 }, { "epoch": 0.44, "learning_rate": 0.00012496472069558007, "loss": 0.1411, "step": 6817 }, { "epoch": 0.44, "learning_rate": 0.00012494461095618572, "loss": 0.1687, "step": 6818 }, { "epoch": 0.44, "learning_rate": 0.00012492450014103617, "loss": 0.1428, "step": 6819 }, { "epoch": 0.44, "learning_rate": 0.00012490438825099877, "loss": 0.1487, "step": 6820 }, { "epoch": 0.44, "learning_rate": 0.0001248842752869408, "loss": 0.1441, "step": 6821 }, { "epoch": 0.44, "learning_rate": 0.00012486416124972966, "loss": 0.1553, "step": 6822 }, { "epoch": 0.44, "learning_rate": 0.0001248440461402328, "loss": 0.1495, "step": 6823 }, { "epoch": 0.44, "learning_rate": 0.00012482392995931768, "loss": 0.1353, "step": 6824 }, { "epoch": 0.44, "learning_rate": 0.00012480381270785185, "loss": 0.1419, "step": 6825 }, { "epoch": 0.44, "learning_rate": 0.00012478369438670284, "loss": 0.1488, "step": 6826 }, { "epoch": 0.44, "learning_rate": 0.00012476357499673833, "loss": 0.136, "step": 6827 }, { "epoch": 0.44, "learning_rate": 0.00012474345453882595, "loss": 0.1402, "step": 6828 }, { "epoch": 0.44, "learning_rate": 0.00012472333301383342, "loss": 0.1724, "step": 6829 }, { "epoch": 0.44, "learning_rate": 0.00012470321042262842, "loss": 0.1464, "step": 6830 }, { "epoch": 0.44, "learning_rate": 0.00012468308676607886, "loss": 0.1274, "step": 6831 }, { "epoch": 0.44, "learning_rate": 0.00012466296204505256, "loss": 0.1465, "step": 6832 }, { "epoch": 0.44, "learning_rate": 0.0001246428362604174, "loss": 0.1544, "step": 6833 }, { "epoch": 0.44, "learning_rate": 0.00012462270941304132, "loss": 0.1321, "step": 6834 }, { "epoch": 0.44, "learning_rate": 0.00012460258150379227, "loss": 0.1303, "step": 6835 }, { "epoch": 0.44, "learning_rate": 0.00012458245253353832, "loss": 0.1505, "step": 6836 }, { "epoch": 0.44, "learning_rate": 0.00012456232250314755, "loss": 0.1306, "step": 6837 }, { "epoch": 0.44, "learning_rate": 0.0001245421914134881, "loss": 0.1567, "step": 6838 }, { "epoch": 0.44, "learning_rate": 0.00012452205926542808, "loss": 0.1504, "step": 6839 }, { "epoch": 0.44, "learning_rate": 0.00012450192605983578, "loss": 0.1095, "step": 6840 }, { "epoch": 0.44, "learning_rate": 0.0001244817917975794, "loss": 0.1223, "step": 6841 }, { "epoch": 0.44, "learning_rate": 0.00012446165647952723, "loss": 0.1583, "step": 6842 }, { "epoch": 0.44, "learning_rate": 0.00012444152010654768, "loss": 0.1405, "step": 6843 }, { "epoch": 0.44, "learning_rate": 0.00012442138267950911, "loss": 0.1531, "step": 6844 }, { "epoch": 0.44, "learning_rate": 0.00012440124419928, "loss": 0.1384, "step": 6845 }, { "epoch": 0.44, "learning_rate": 0.00012438110466672878, "loss": 0.1411, "step": 6846 }, { "epoch": 0.44, "learning_rate": 0.000124360964082724, "loss": 0.1482, "step": 6847 }, { "epoch": 0.44, "learning_rate": 0.0001243408224481343, "loss": 0.1702, "step": 6848 }, { "epoch": 0.44, "learning_rate": 0.0001243206797638282, "loss": 0.1455, "step": 6849 }, { "epoch": 0.44, "learning_rate": 0.00012430053603067448, "loss": 0.1426, "step": 6850 }, { "epoch": 0.44, "learning_rate": 0.00012428039124954174, "loss": 0.1419, "step": 6851 }, { "epoch": 0.44, "learning_rate": 0.0001242602454212988, "loss": 0.1477, "step": 6852 }, { "epoch": 0.44, "learning_rate": 0.0001242400985468145, "loss": 0.1367, "step": 6853 }, { "epoch": 0.44, "learning_rate": 0.00012421995062695756, "loss": 0.1406, "step": 6854 }, { "epoch": 0.44, "learning_rate": 0.00012419980166259706, "loss": 0.1536, "step": 6855 }, { "epoch": 0.44, "learning_rate": 0.00012417965165460179, "loss": 0.1292, "step": 6856 }, { "epoch": 0.44, "learning_rate": 0.00012415950060384079, "loss": 0.1774, "step": 6857 }, { "epoch": 0.44, "learning_rate": 0.0001241393485111831, "loss": 0.1491, "step": 6858 }, { "epoch": 0.44, "learning_rate": 0.00012411919537749774, "loss": 0.1356, "step": 6859 }, { "epoch": 0.44, "learning_rate": 0.00012409904120365392, "loss": 0.1423, "step": 6860 }, { "epoch": 0.44, "learning_rate": 0.00012407888599052068, "loss": 0.1521, "step": 6861 }, { "epoch": 0.44, "learning_rate": 0.00012405872973896738, "loss": 0.1518, "step": 6862 }, { "epoch": 0.44, "learning_rate": 0.00012403857244986315, "loss": 0.1504, "step": 6863 }, { "epoch": 0.44, "learning_rate": 0.00012401841412407733, "loss": 0.1406, "step": 6864 }, { "epoch": 0.44, "learning_rate": 0.0001239982547624793, "loss": 0.1447, "step": 6865 }, { "epoch": 0.44, "learning_rate": 0.00012397809436593836, "loss": 0.1672, "step": 6866 }, { "epoch": 0.44, "learning_rate": 0.00012395793293532405, "loss": 0.1492, "step": 6867 }, { "epoch": 0.44, "learning_rate": 0.00012393777047150573, "loss": 0.1427, "step": 6868 }, { "epoch": 0.44, "learning_rate": 0.00012391760697535306, "loss": 0.1431, "step": 6869 }, { "epoch": 0.44, "learning_rate": 0.0001238974424477355, "loss": 0.1422, "step": 6870 }, { "epoch": 0.44, "learning_rate": 0.00012387727688952264, "loss": 0.1311, "step": 6871 }, { "epoch": 0.44, "learning_rate": 0.00012385711030158422, "loss": 0.1495, "step": 6872 }, { "epoch": 0.44, "learning_rate": 0.00012383694268478993, "loss": 0.1396, "step": 6873 }, { "epoch": 0.44, "learning_rate": 0.00012381677404000942, "loss": 0.1479, "step": 6874 }, { "epoch": 0.44, "learning_rate": 0.00012379660436811262, "loss": 0.149, "step": 6875 }, { "epoch": 0.44, "learning_rate": 0.00012377643366996927, "loss": 0.1346, "step": 6876 }, { "epoch": 0.44, "learning_rate": 0.00012375626194644922, "loss": 0.1458, "step": 6877 }, { "epoch": 0.44, "learning_rate": 0.0001237360891984225, "loss": 0.1748, "step": 6878 }, { "epoch": 0.44, "learning_rate": 0.00012371591542675896, "loss": 0.134, "step": 6879 }, { "epoch": 0.44, "learning_rate": 0.0001236957406323287, "loss": 0.1376, "step": 6880 }, { "epoch": 0.44, "learning_rate": 0.00012367556481600168, "loss": 0.1589, "step": 6881 }, { "epoch": 0.44, "learning_rate": 0.00012365538797864812, "loss": 0.1194, "step": 6882 }, { "epoch": 0.44, "learning_rate": 0.00012363521012113802, "loss": 0.1518, "step": 6883 }, { "epoch": 0.44, "learning_rate": 0.00012361503124434168, "loss": 0.1416, "step": 6884 }, { "epoch": 0.44, "learning_rate": 0.00012359485134912928, "loss": 0.1542, "step": 6885 }, { "epoch": 0.44, "learning_rate": 0.0001235746704363711, "loss": 0.1317, "step": 6886 }, { "epoch": 0.44, "learning_rate": 0.00012355448850693746, "loss": 0.1715, "step": 6887 }, { "epoch": 0.44, "learning_rate": 0.0001235343055616987, "loss": 0.1466, "step": 6888 }, { "epoch": 0.44, "learning_rate": 0.00012351412160152526, "loss": 0.1448, "step": 6889 }, { "epoch": 0.44, "learning_rate": 0.00012349393662728757, "loss": 0.1571, "step": 6890 }, { "epoch": 0.44, "learning_rate": 0.0001234737506398561, "loss": 0.1498, "step": 6891 }, { "epoch": 0.44, "learning_rate": 0.00012345356364010143, "loss": 0.1592, "step": 6892 }, { "epoch": 0.44, "learning_rate": 0.00012343337562889411, "loss": 0.136, "step": 6893 }, { "epoch": 0.44, "learning_rate": 0.00012341318660710475, "loss": 0.1456, "step": 6894 }, { "epoch": 0.44, "learning_rate": 0.00012339299657560408, "loss": 0.1401, "step": 6895 }, { "epoch": 0.44, "learning_rate": 0.00012337280553526277, "loss": 0.1379, "step": 6896 }, { "epoch": 0.44, "learning_rate": 0.00012335261348695155, "loss": 0.1703, "step": 6897 }, { "epoch": 0.44, "learning_rate": 0.00012333242043154126, "loss": 0.1377, "step": 6898 }, { "epoch": 0.44, "learning_rate": 0.00012331222636990268, "loss": 0.1517, "step": 6899 }, { "epoch": 0.44, "learning_rate": 0.0001232920313029068, "loss": 0.1585, "step": 6900 }, { "epoch": 0.44, "learning_rate": 0.0001232718352314244, "loss": 0.1465, "step": 6901 }, { "epoch": 0.44, "learning_rate": 0.00012325163815632659, "loss": 0.1457, "step": 6902 }, { "epoch": 0.44, "learning_rate": 0.00012323144007848436, "loss": 0.1285, "step": 6903 }, { "epoch": 0.44, "learning_rate": 0.00012321124099876865, "loss": 0.1499, "step": 6904 }, { "epoch": 0.44, "learning_rate": 0.0001231910409180507, "loss": 0.1572, "step": 6905 }, { "epoch": 0.44, "learning_rate": 0.0001231708398372016, "loss": 0.1317, "step": 6906 }, { "epoch": 0.44, "learning_rate": 0.0001231506377570925, "loss": 0.1387, "step": 6907 }, { "epoch": 0.44, "learning_rate": 0.00012313043467859468, "loss": 0.1352, "step": 6908 }, { "epoch": 0.44, "learning_rate": 0.0001231102306025794, "loss": 0.1591, "step": 6909 }, { "epoch": 0.44, "learning_rate": 0.000123090025529918, "loss": 0.1539, "step": 6910 }, { "epoch": 0.44, "learning_rate": 0.00012306981946148178, "loss": 0.1424, "step": 6911 }, { "epoch": 0.44, "learning_rate": 0.0001230496123981422, "loss": 0.1496, "step": 6912 }, { "epoch": 0.44, "learning_rate": 0.0001230294043407707, "loss": 0.1449, "step": 6913 }, { "epoch": 0.44, "learning_rate": 0.0001230091952902387, "loss": 0.1372, "step": 6914 }, { "epoch": 0.44, "learning_rate": 0.00012298898524741779, "loss": 0.1598, "step": 6915 }, { "epoch": 0.44, "learning_rate": 0.0001229687742131796, "loss": 0.1594, "step": 6916 }, { "epoch": 0.44, "learning_rate": 0.00012294856218839558, "loss": 0.1505, "step": 6917 }, { "epoch": 0.44, "learning_rate": 0.00012292834917393755, "loss": 0.1775, "step": 6918 }, { "epoch": 0.44, "learning_rate": 0.0001229081351706771, "loss": 0.1468, "step": 6919 }, { "epoch": 0.44, "learning_rate": 0.00012288792017948607, "loss": 0.1342, "step": 6920 }, { "epoch": 0.44, "learning_rate": 0.00012286770420123617, "loss": 0.1315, "step": 6921 }, { "epoch": 0.44, "learning_rate": 0.00012284748723679929, "loss": 0.1464, "step": 6922 }, { "epoch": 0.44, "learning_rate": 0.00012282726928704727, "loss": 0.1521, "step": 6923 }, { "epoch": 0.44, "learning_rate": 0.000122807050352852, "loss": 0.1448, "step": 6924 }, { "epoch": 0.44, "learning_rate": 0.00012278683043508544, "loss": 0.1215, "step": 6925 }, { "epoch": 0.44, "learning_rate": 0.00012276660953461967, "loss": 0.1609, "step": 6926 }, { "epoch": 0.44, "learning_rate": 0.00012274638765232665, "loss": 0.1393, "step": 6927 }, { "epoch": 0.44, "learning_rate": 0.00012272616478907846, "loss": 0.1431, "step": 6928 }, { "epoch": 0.44, "learning_rate": 0.0001227059409457473, "loss": 0.1564, "step": 6929 }, { "epoch": 0.44, "learning_rate": 0.00012268571612320529, "loss": 0.1537, "step": 6930 }, { "epoch": 0.44, "learning_rate": 0.0001226654903223246, "loss": 0.1518, "step": 6931 }, { "epoch": 0.44, "learning_rate": 0.00012264526354397755, "loss": 0.183, "step": 6932 }, { "epoch": 0.44, "learning_rate": 0.00012262503578903638, "loss": 0.1293, "step": 6933 }, { "epoch": 0.44, "learning_rate": 0.0001226048070583735, "loss": 0.1358, "step": 6934 }, { "epoch": 0.44, "learning_rate": 0.00012258457735286123, "loss": 0.1373, "step": 6935 }, { "epoch": 0.44, "learning_rate": 0.000122564346673372, "loss": 0.144, "step": 6936 }, { "epoch": 0.44, "learning_rate": 0.0001225441150207783, "loss": 0.1401, "step": 6937 }, { "epoch": 0.44, "learning_rate": 0.00012252388239595262, "loss": 0.1498, "step": 6938 }, { "epoch": 0.44, "learning_rate": 0.00012250364879976748, "loss": 0.1557, "step": 6939 }, { "epoch": 0.44, "learning_rate": 0.0001224834142330955, "loss": 0.1332, "step": 6940 }, { "epoch": 0.45, "learning_rate": 0.0001224631786968093, "loss": 0.1599, "step": 6941 }, { "epoch": 0.45, "learning_rate": 0.00012244294219178158, "loss": 0.1538, "step": 6942 }, { "epoch": 0.45, "learning_rate": 0.00012242270471888501, "loss": 0.1475, "step": 6943 }, { "epoch": 0.45, "learning_rate": 0.00012240246627899238, "loss": 0.1466, "step": 6944 }, { "epoch": 0.45, "learning_rate": 0.0001223822268729765, "loss": 0.1433, "step": 6945 }, { "epoch": 0.45, "learning_rate": 0.00012236198650171016, "loss": 0.1406, "step": 6946 }, { "epoch": 0.45, "learning_rate": 0.00012234174516606626, "loss": 0.1672, "step": 6947 }, { "epoch": 0.45, "learning_rate": 0.00012232150286691773, "loss": 0.1429, "step": 6948 }, { "epoch": 0.45, "learning_rate": 0.00012230125960513755, "loss": 0.1432, "step": 6949 }, { "epoch": 0.45, "learning_rate": 0.0001222810153815987, "loss": 0.1688, "step": 6950 }, { "epoch": 0.45, "learning_rate": 0.00012226077019717422, "loss": 0.1365, "step": 6951 }, { "epoch": 0.45, "learning_rate": 0.00012224052405273724, "loss": 0.1455, "step": 6952 }, { "epoch": 0.45, "learning_rate": 0.00012222027694916088, "loss": 0.1594, "step": 6953 }, { "epoch": 0.45, "learning_rate": 0.00012220002888731828, "loss": 0.1579, "step": 6954 }, { "epoch": 0.45, "learning_rate": 0.00012217977986808264, "loss": 0.1507, "step": 6955 }, { "epoch": 0.45, "learning_rate": 0.00012215952989232728, "loss": 0.1604, "step": 6956 }, { "epoch": 0.45, "learning_rate": 0.0001221392789609255, "loss": 0.1526, "step": 6957 }, { "epoch": 0.45, "learning_rate": 0.00012211902707475052, "loss": 0.1587, "step": 6958 }, { "epoch": 0.45, "learning_rate": 0.00012209877423467584, "loss": 0.1327, "step": 6959 }, { "epoch": 0.45, "learning_rate": 0.00012207852044157484, "loss": 0.1528, "step": 6960 }, { "epoch": 0.45, "learning_rate": 0.00012205826569632093, "loss": 0.1449, "step": 6961 }, { "epoch": 0.45, "learning_rate": 0.0001220380099997877, "loss": 0.1477, "step": 6962 }, { "epoch": 0.45, "learning_rate": 0.00012201775335284864, "loss": 0.1657, "step": 6963 }, { "epoch": 0.45, "learning_rate": 0.00012199749575637733, "loss": 0.1305, "step": 6964 }, { "epoch": 0.45, "learning_rate": 0.00012197723721124743, "loss": 0.1543, "step": 6965 }, { "epoch": 0.45, "learning_rate": 0.00012195697771833256, "loss": 0.1576, "step": 6966 }, { "epoch": 0.45, "learning_rate": 0.00012193671727850647, "loss": 0.1788, "step": 6967 }, { "epoch": 0.45, "learning_rate": 0.00012191645589264283, "loss": 0.157, "step": 6968 }, { "epoch": 0.45, "learning_rate": 0.0001218961935616155, "loss": 0.1484, "step": 6969 }, { "epoch": 0.45, "learning_rate": 0.00012187593028629831, "loss": 0.1449, "step": 6970 }, { "epoch": 0.45, "learning_rate": 0.00012185566606756511, "loss": 0.1595, "step": 6971 }, { "epoch": 0.45, "learning_rate": 0.0001218354009062898, "loss": 0.1635, "step": 6972 }, { "epoch": 0.45, "learning_rate": 0.00012181513480334634, "loss": 0.1306, "step": 6973 }, { "epoch": 0.45, "learning_rate": 0.00012179486775960874, "loss": 0.174, "step": 6974 }, { "epoch": 0.45, "learning_rate": 0.00012177459977595095, "loss": 0.1581, "step": 6975 }, { "epoch": 0.45, "learning_rate": 0.00012175433085324714, "loss": 0.1425, "step": 6976 }, { "epoch": 0.45, "learning_rate": 0.00012173406099237139, "loss": 0.1669, "step": 6977 }, { "epoch": 0.45, "learning_rate": 0.00012171379019419787, "loss": 0.1506, "step": 6978 }, { "epoch": 0.45, "learning_rate": 0.00012169351845960073, "loss": 0.1363, "step": 6979 }, { "epoch": 0.45, "learning_rate": 0.00012167324578945423, "loss": 0.1711, "step": 6980 }, { "epoch": 0.45, "learning_rate": 0.00012165297218463261, "loss": 0.1462, "step": 6981 }, { "epoch": 0.45, "learning_rate": 0.00012163269764601025, "loss": 0.157, "step": 6982 }, { "epoch": 0.45, "learning_rate": 0.00012161242217446145, "loss": 0.1495, "step": 6983 }, { "epoch": 0.45, "learning_rate": 0.00012159214577086062, "loss": 0.1448, "step": 6984 }, { "epoch": 0.45, "learning_rate": 0.00012157186843608222, "loss": 0.1464, "step": 6985 }, { "epoch": 0.45, "learning_rate": 0.00012155159017100071, "loss": 0.1627, "step": 6986 }, { "epoch": 0.45, "learning_rate": 0.00012153131097649057, "loss": 0.1572, "step": 6987 }, { "epoch": 0.45, "learning_rate": 0.0001215110308534264, "loss": 0.1421, "step": 6988 }, { "epoch": 0.45, "learning_rate": 0.0001214907498026828, "loss": 0.1413, "step": 6989 }, { "epoch": 0.45, "learning_rate": 0.00012147046782513436, "loss": 0.138, "step": 6990 }, { "epoch": 0.45, "learning_rate": 0.00012145018492165581, "loss": 0.1533, "step": 6991 }, { "epoch": 0.45, "learning_rate": 0.00012142990109312182, "loss": 0.1548, "step": 6992 }, { "epoch": 0.45, "learning_rate": 0.00012140961634040717, "loss": 0.1525, "step": 6993 }, { "epoch": 0.45, "learning_rate": 0.00012138933066438667, "loss": 0.155, "step": 6994 }, { "epoch": 0.45, "learning_rate": 0.00012136904406593507, "loss": 0.16, "step": 6995 }, { "epoch": 0.45, "learning_rate": 0.00012134875654592736, "loss": 0.1852, "step": 6996 }, { "epoch": 0.45, "learning_rate": 0.00012132846810523841, "loss": 0.148, "step": 6997 }, { "epoch": 0.45, "learning_rate": 0.00012130817874474317, "loss": 0.1666, "step": 6998 }, { "epoch": 0.45, "learning_rate": 0.0001212878884653166, "loss": 0.1473, "step": 6999 }, { "epoch": 0.45, "learning_rate": 0.00012126759726783379, "loss": 0.144, "step": 7000 }, { "epoch": 0.45, "learning_rate": 0.00012124730515316978, "loss": 0.1447, "step": 7001 }, { "epoch": 0.45, "learning_rate": 0.00012122701212219968, "loss": 0.1383, "step": 7002 }, { "epoch": 0.45, "learning_rate": 0.00012120671817579867, "loss": 0.1455, "step": 7003 }, { "epoch": 0.45, "learning_rate": 0.00012118642331484194, "loss": 0.1397, "step": 7004 }, { "epoch": 0.45, "learning_rate": 0.00012116612754020469, "loss": 0.1473, "step": 7005 }, { "epoch": 0.45, "learning_rate": 0.00012114583085276221, "loss": 0.1468, "step": 7006 }, { "epoch": 0.45, "learning_rate": 0.0001211255332533898, "loss": 0.1415, "step": 7007 }, { "epoch": 0.45, "learning_rate": 0.00012110523474296281, "loss": 0.1397, "step": 7008 }, { "epoch": 0.45, "learning_rate": 0.00012108493532235666, "loss": 0.1633, "step": 7009 }, { "epoch": 0.45, "learning_rate": 0.00012106463499244673, "loss": 0.1561, "step": 7010 }, { "epoch": 0.45, "learning_rate": 0.00012104433375410853, "loss": 0.1588, "step": 7011 }, { "epoch": 0.45, "learning_rate": 0.00012102403160821753, "loss": 0.1714, "step": 7012 }, { "epoch": 0.45, "learning_rate": 0.0001210037285556493, "loss": 0.1367, "step": 7013 }, { "epoch": 0.45, "learning_rate": 0.00012098342459727943, "loss": 0.1449, "step": 7014 }, { "epoch": 0.45, "learning_rate": 0.00012096311973398348, "loss": 0.1605, "step": 7015 }, { "epoch": 0.45, "learning_rate": 0.00012094281396663722, "loss": 0.1404, "step": 7016 }, { "epoch": 0.45, "learning_rate": 0.00012092250729611627, "loss": 0.1394, "step": 7017 }, { "epoch": 0.45, "learning_rate": 0.00012090219972329639, "loss": 0.1421, "step": 7018 }, { "epoch": 0.45, "learning_rate": 0.00012088189124905338, "loss": 0.1682, "step": 7019 }, { "epoch": 0.45, "learning_rate": 0.00012086158187426304, "loss": 0.1406, "step": 7020 }, { "epoch": 0.45, "learning_rate": 0.00012084127159980123, "loss": 0.1556, "step": 7021 }, { "epoch": 0.45, "learning_rate": 0.00012082096042654382, "loss": 0.173, "step": 7022 }, { "epoch": 0.45, "learning_rate": 0.0001208006483553668, "loss": 0.1725, "step": 7023 }, { "epoch": 0.45, "learning_rate": 0.00012078033538714611, "loss": 0.1557, "step": 7024 }, { "epoch": 0.45, "learning_rate": 0.00012076002152275774, "loss": 0.1475, "step": 7025 }, { "epoch": 0.45, "learning_rate": 0.0001207397067630778, "loss": 0.1475, "step": 7026 }, { "epoch": 0.45, "learning_rate": 0.00012071939110898234, "loss": 0.1626, "step": 7027 }, { "epoch": 0.45, "learning_rate": 0.00012069907456134746, "loss": 0.1441, "step": 7028 }, { "epoch": 0.45, "learning_rate": 0.0001206787571210494, "loss": 0.1655, "step": 7029 }, { "epoch": 0.45, "learning_rate": 0.0001206584387889643, "loss": 0.1398, "step": 7030 }, { "epoch": 0.45, "learning_rate": 0.00012063811956596845, "loss": 0.138, "step": 7031 }, { "epoch": 0.45, "learning_rate": 0.0001206177994529381, "loss": 0.1681, "step": 7032 }, { "epoch": 0.45, "learning_rate": 0.00012059747845074956, "loss": 0.1285, "step": 7033 }, { "epoch": 0.45, "learning_rate": 0.00012057715656027923, "loss": 0.1531, "step": 7034 }, { "epoch": 0.45, "learning_rate": 0.00012055683378240347, "loss": 0.1372, "step": 7035 }, { "epoch": 0.45, "learning_rate": 0.00012053651011799876, "loss": 0.1547, "step": 7036 }, { "epoch": 0.45, "learning_rate": 0.0001205161855679415, "loss": 0.1674, "step": 7037 }, { "epoch": 0.45, "learning_rate": 0.0001204958601331083, "loss": 0.1446, "step": 7038 }, { "epoch": 0.45, "learning_rate": 0.00012047553381437562, "loss": 0.1329, "step": 7039 }, { "epoch": 0.45, "learning_rate": 0.0001204552066126201, "loss": 0.1659, "step": 7040 }, { "epoch": 0.45, "learning_rate": 0.00012043487852871835, "loss": 0.1481, "step": 7041 }, { "epoch": 0.45, "learning_rate": 0.000120414549563547, "loss": 0.1548, "step": 7042 }, { "epoch": 0.45, "learning_rate": 0.00012039421971798282, "loss": 0.1582, "step": 7043 }, { "epoch": 0.45, "learning_rate": 0.00012037388899290252, "loss": 0.1553, "step": 7044 }, { "epoch": 0.45, "learning_rate": 0.00012035355738918284, "loss": 0.1576, "step": 7045 }, { "epoch": 0.45, "learning_rate": 0.00012033322490770067, "loss": 0.1445, "step": 7046 }, { "epoch": 0.45, "learning_rate": 0.00012031289154933281, "loss": 0.1424, "step": 7047 }, { "epoch": 0.45, "learning_rate": 0.00012029255731495613, "loss": 0.1547, "step": 7048 }, { "epoch": 0.45, "learning_rate": 0.00012027222220544762, "loss": 0.1521, "step": 7049 }, { "epoch": 0.45, "learning_rate": 0.00012025188622168422, "loss": 0.1595, "step": 7050 }, { "epoch": 0.45, "learning_rate": 0.00012023154936454293, "loss": 0.1561, "step": 7051 }, { "epoch": 0.45, "learning_rate": 0.00012021121163490078, "loss": 0.1397, "step": 7052 }, { "epoch": 0.45, "learning_rate": 0.00012019087303363493, "loss": 0.1623, "step": 7053 }, { "epoch": 0.45, "learning_rate": 0.00012017053356162238, "loss": 0.1452, "step": 7054 }, { "epoch": 0.45, "learning_rate": 0.00012015019321974034, "loss": 0.1405, "step": 7055 }, { "epoch": 0.45, "learning_rate": 0.00012012985200886602, "loss": 0.168, "step": 7056 }, { "epoch": 0.45, "learning_rate": 0.00012010950992987662, "loss": 0.1634, "step": 7057 }, { "epoch": 0.45, "learning_rate": 0.00012008916698364942, "loss": 0.127, "step": 7058 }, { "epoch": 0.45, "learning_rate": 0.00012006882317106174, "loss": 0.1703, "step": 7059 }, { "epoch": 0.45, "learning_rate": 0.0001200484784929909, "loss": 0.1349, "step": 7060 }, { "epoch": 0.45, "learning_rate": 0.0001200281329503143, "loss": 0.1384, "step": 7061 }, { "epoch": 0.45, "learning_rate": 0.00012000778654390931, "loss": 0.1565, "step": 7062 }, { "epoch": 0.45, "learning_rate": 0.00011998743927465344, "loss": 0.1592, "step": 7063 }, { "epoch": 0.45, "learning_rate": 0.00011996709114342417, "loss": 0.1603, "step": 7064 }, { "epoch": 0.45, "learning_rate": 0.00011994674215109898, "loss": 0.1423, "step": 7065 }, { "epoch": 0.45, "learning_rate": 0.00011992639229855549, "loss": 0.137, "step": 7066 }, { "epoch": 0.45, "learning_rate": 0.00011990604158667131, "loss": 0.1699, "step": 7067 }, { "epoch": 0.45, "learning_rate": 0.00011988569001632403, "loss": 0.1502, "step": 7068 }, { "epoch": 0.45, "learning_rate": 0.00011986533758839136, "loss": 0.1515, "step": 7069 }, { "epoch": 0.45, "learning_rate": 0.00011984498430375101, "loss": 0.1261, "step": 7070 }, { "epoch": 0.45, "learning_rate": 0.00011982463016328074, "loss": 0.1776, "step": 7071 }, { "epoch": 0.45, "learning_rate": 0.00011980427516785829, "loss": 0.1516, "step": 7072 }, { "epoch": 0.45, "learning_rate": 0.00011978391931836156, "loss": 0.1698, "step": 7073 }, { "epoch": 0.45, "learning_rate": 0.00011976356261566834, "loss": 0.1764, "step": 7074 }, { "epoch": 0.45, "learning_rate": 0.00011974320506065657, "loss": 0.1422, "step": 7075 }, { "epoch": 0.45, "learning_rate": 0.00011972284665420417, "loss": 0.1474, "step": 7076 }, { "epoch": 0.45, "learning_rate": 0.00011970248739718912, "loss": 0.1474, "step": 7077 }, { "epoch": 0.45, "learning_rate": 0.00011968212729048943, "loss": 0.1361, "step": 7078 }, { "epoch": 0.45, "learning_rate": 0.00011966176633498312, "loss": 0.1233, "step": 7079 }, { "epoch": 0.45, "learning_rate": 0.00011964140453154833, "loss": 0.1581, "step": 7080 }, { "epoch": 0.45, "learning_rate": 0.00011962104188106311, "loss": 0.1587, "step": 7081 }, { "epoch": 0.45, "learning_rate": 0.00011960067838440562, "loss": 0.1504, "step": 7082 }, { "epoch": 0.45, "learning_rate": 0.00011958031404245411, "loss": 0.1211, "step": 7083 }, { "epoch": 0.45, "learning_rate": 0.00011955994885608677, "loss": 0.151, "step": 7084 }, { "epoch": 0.45, "learning_rate": 0.00011953958282618186, "loss": 0.1537, "step": 7085 }, { "epoch": 0.45, "learning_rate": 0.00011951921595361766, "loss": 0.143, "step": 7086 }, { "epoch": 0.45, "learning_rate": 0.00011949884823927258, "loss": 0.1636, "step": 7087 }, { "epoch": 0.45, "learning_rate": 0.0001194784796840249, "loss": 0.1452, "step": 7088 }, { "epoch": 0.45, "learning_rate": 0.00011945811028875311, "loss": 0.1393, "step": 7089 }, { "epoch": 0.45, "learning_rate": 0.00011943774005433562, "loss": 0.1491, "step": 7090 }, { "epoch": 0.45, "learning_rate": 0.0001194173689816509, "loss": 0.1646, "step": 7091 }, { "epoch": 0.45, "learning_rate": 0.00011939699707157748, "loss": 0.1457, "step": 7092 }, { "epoch": 0.45, "learning_rate": 0.0001193766243249939, "loss": 0.1344, "step": 7093 }, { "epoch": 0.45, "learning_rate": 0.00011935625074277884, "loss": 0.1399, "step": 7094 }, { "epoch": 0.45, "learning_rate": 0.00011933587632581075, "loss": 0.1582, "step": 7095 }, { "epoch": 0.45, "learning_rate": 0.00011931550107496846, "loss": 0.143, "step": 7096 }, { "epoch": 0.46, "learning_rate": 0.00011929512499113058, "loss": 0.1346, "step": 7097 }, { "epoch": 0.46, "learning_rate": 0.00011927474807517587, "loss": 0.1455, "step": 7098 }, { "epoch": 0.46, "learning_rate": 0.00011925437032798306, "loss": 0.1383, "step": 7099 }, { "epoch": 0.46, "learning_rate": 0.00011923399175043103, "loss": 0.1337, "step": 7100 }, { "epoch": 0.46, "learning_rate": 0.0001192136123433986, "loss": 0.1408, "step": 7101 }, { "epoch": 0.46, "learning_rate": 0.00011919323210776459, "loss": 0.144, "step": 7102 }, { "epoch": 0.46, "learning_rate": 0.00011917285104440796, "loss": 0.147, "step": 7103 }, { "epoch": 0.46, "learning_rate": 0.00011915246915420768, "loss": 0.1546, "step": 7104 }, { "epoch": 0.46, "learning_rate": 0.00011913208643804267, "loss": 0.1377, "step": 7105 }, { "epoch": 0.46, "learning_rate": 0.00011911170289679199, "loss": 0.1413, "step": 7106 }, { "epoch": 0.46, "learning_rate": 0.00011909131853133473, "loss": 0.1604, "step": 7107 }, { "epoch": 0.46, "learning_rate": 0.00011907093334254993, "loss": 0.1482, "step": 7108 }, { "epoch": 0.46, "learning_rate": 0.00011905054733131668, "loss": 0.1418, "step": 7109 }, { "epoch": 0.46, "learning_rate": 0.00011903016049851424, "loss": 0.1484, "step": 7110 }, { "epoch": 0.46, "learning_rate": 0.00011900977284502176, "loss": 0.1431, "step": 7111 }, { "epoch": 0.46, "learning_rate": 0.00011898938437171842, "loss": 0.1395, "step": 7112 }, { "epoch": 0.46, "learning_rate": 0.00011896899507948358, "loss": 0.1473, "step": 7113 }, { "epoch": 0.46, "learning_rate": 0.00011894860496919654, "loss": 0.1365, "step": 7114 }, { "epoch": 0.46, "learning_rate": 0.00011892821404173653, "loss": 0.1492, "step": 7115 }, { "epoch": 0.46, "learning_rate": 0.00011890782229798305, "loss": 0.1583, "step": 7116 }, { "epoch": 0.46, "learning_rate": 0.00011888742973881543, "loss": 0.1458, "step": 7117 }, { "epoch": 0.46, "learning_rate": 0.00011886703636511316, "loss": 0.1511, "step": 7118 }, { "epoch": 0.46, "learning_rate": 0.00011884664217775567, "loss": 0.1622, "step": 7119 }, { "epoch": 0.46, "learning_rate": 0.00011882624717762252, "loss": 0.1362, "step": 7120 }, { "epoch": 0.46, "learning_rate": 0.00011880585136559328, "loss": 0.1439, "step": 7121 }, { "epoch": 0.46, "learning_rate": 0.00011878545474254746, "loss": 0.1337, "step": 7122 }, { "epoch": 0.46, "learning_rate": 0.00011876505730936474, "loss": 0.1615, "step": 7123 }, { "epoch": 0.46, "learning_rate": 0.00011874465906692473, "loss": 0.1506, "step": 7124 }, { "epoch": 0.46, "learning_rate": 0.00011872426001610717, "loss": 0.1448, "step": 7125 }, { "epoch": 0.46, "learning_rate": 0.00011870386015779172, "loss": 0.1648, "step": 7126 }, { "epoch": 0.46, "learning_rate": 0.00011868345949285822, "loss": 0.169, "step": 7127 }, { "epoch": 0.46, "learning_rate": 0.00011866305802218642, "loss": 0.1682, "step": 7128 }, { "epoch": 0.46, "learning_rate": 0.00011864265574665612, "loss": 0.1352, "step": 7129 }, { "epoch": 0.46, "learning_rate": 0.00011862225266714723, "loss": 0.1533, "step": 7130 }, { "epoch": 0.46, "learning_rate": 0.00011860184878453964, "loss": 0.159, "step": 7131 }, { "epoch": 0.46, "learning_rate": 0.00011858144409971326, "loss": 0.1485, "step": 7132 }, { "epoch": 0.46, "learning_rate": 0.00011856103861354809, "loss": 0.1644, "step": 7133 }, { "epoch": 0.46, "learning_rate": 0.0001185406323269241, "loss": 0.1671, "step": 7134 }, { "epoch": 0.46, "learning_rate": 0.00011852022524072135, "loss": 0.1427, "step": 7135 }, { "epoch": 0.46, "learning_rate": 0.0001184998173558199, "loss": 0.1387, "step": 7136 }, { "epoch": 0.46, "learning_rate": 0.00011847940867309985, "loss": 0.1511, "step": 7137 }, { "epoch": 0.46, "learning_rate": 0.00011845899919344136, "loss": 0.1403, "step": 7138 }, { "epoch": 0.46, "learning_rate": 0.00011843858891772457, "loss": 0.14, "step": 7139 }, { "epoch": 0.46, "learning_rate": 0.00011841817784682974, "loss": 0.1564, "step": 7140 }, { "epoch": 0.46, "learning_rate": 0.00011839776598163707, "loss": 0.1682, "step": 7141 }, { "epoch": 0.46, "learning_rate": 0.00011837735332302684, "loss": 0.153, "step": 7142 }, { "epoch": 0.46, "learning_rate": 0.0001183569398718794, "loss": 0.152, "step": 7143 }, { "epoch": 0.46, "learning_rate": 0.00011833652562907504, "loss": 0.1493, "step": 7144 }, { "epoch": 0.46, "learning_rate": 0.00011831611059549419, "loss": 0.147, "step": 7145 }, { "epoch": 0.46, "learning_rate": 0.00011829569477201721, "loss": 0.1539, "step": 7146 }, { "epoch": 0.46, "learning_rate": 0.00011827527815952459, "loss": 0.1728, "step": 7147 }, { "epoch": 0.46, "learning_rate": 0.00011825486075889683, "loss": 0.1556, "step": 7148 }, { "epoch": 0.46, "learning_rate": 0.00011823444257101438, "loss": 0.1285, "step": 7149 }, { "epoch": 0.46, "learning_rate": 0.00011821402359675781, "loss": 0.1569, "step": 7150 }, { "epoch": 0.46, "learning_rate": 0.00011819360383700778, "loss": 0.1394, "step": 7151 }, { "epoch": 0.46, "learning_rate": 0.00011817318329264477, "loss": 0.1461, "step": 7152 }, { "epoch": 0.46, "learning_rate": 0.00011815276196454956, "loss": 0.1485, "step": 7153 }, { "epoch": 0.46, "learning_rate": 0.00011813233985360277, "loss": 0.1401, "step": 7154 }, { "epoch": 0.46, "learning_rate": 0.00011811191696068515, "loss": 0.1492, "step": 7155 }, { "epoch": 0.46, "learning_rate": 0.0001180914932866774, "loss": 0.1272, "step": 7156 }, { "epoch": 0.46, "learning_rate": 0.00011807106883246038, "loss": 0.1462, "step": 7157 }, { "epoch": 0.46, "learning_rate": 0.00011805064359891487, "loss": 0.1393, "step": 7158 }, { "epoch": 0.46, "learning_rate": 0.00011803021758692167, "loss": 0.158, "step": 7159 }, { "epoch": 0.46, "learning_rate": 0.00011800979079736176, "loss": 0.1696, "step": 7160 }, { "epoch": 0.46, "learning_rate": 0.00011798936323111604, "loss": 0.168, "step": 7161 }, { "epoch": 0.46, "learning_rate": 0.00011796893488906545, "loss": 0.1504, "step": 7162 }, { "epoch": 0.46, "learning_rate": 0.00011794850577209093, "loss": 0.1489, "step": 7163 }, { "epoch": 0.46, "learning_rate": 0.00011792807588107357, "loss": 0.1407, "step": 7164 }, { "epoch": 0.46, "learning_rate": 0.00011790764521689442, "loss": 0.1636, "step": 7165 }, { "epoch": 0.46, "learning_rate": 0.00011788721378043451, "loss": 0.1484, "step": 7166 }, { "epoch": 0.46, "learning_rate": 0.00011786678157257502, "loss": 0.1526, "step": 7167 }, { "epoch": 0.46, "learning_rate": 0.0001178463485941971, "loss": 0.1349, "step": 7168 }, { "epoch": 0.46, "learning_rate": 0.0001178259148461819, "loss": 0.1443, "step": 7169 }, { "epoch": 0.46, "learning_rate": 0.00011780548032941067, "loss": 0.158, "step": 7170 }, { "epoch": 0.46, "learning_rate": 0.00011778504504476467, "loss": 0.1769, "step": 7171 }, { "epoch": 0.46, "learning_rate": 0.00011776460899312514, "loss": 0.1533, "step": 7172 }, { "epoch": 0.46, "learning_rate": 0.00011774417217537344, "loss": 0.1533, "step": 7173 }, { "epoch": 0.46, "learning_rate": 0.00011772373459239092, "loss": 0.1588, "step": 7174 }, { "epoch": 0.46, "learning_rate": 0.00011770329624505898, "loss": 0.1722, "step": 7175 }, { "epoch": 0.46, "learning_rate": 0.00011768285713425901, "loss": 0.1449, "step": 7176 }, { "epoch": 0.46, "learning_rate": 0.00011766241726087249, "loss": 0.1483, "step": 7177 }, { "epoch": 0.46, "learning_rate": 0.00011764197662578086, "loss": 0.1673, "step": 7178 }, { "epoch": 0.46, "learning_rate": 0.00011762153522986568, "loss": 0.1579, "step": 7179 }, { "epoch": 0.46, "learning_rate": 0.00011760109307400849, "loss": 0.1342, "step": 7180 }, { "epoch": 0.46, "learning_rate": 0.00011758065015909085, "loss": 0.1638, "step": 7181 }, { "epoch": 0.46, "learning_rate": 0.00011756020648599443, "loss": 0.1508, "step": 7182 }, { "epoch": 0.46, "learning_rate": 0.00011753976205560085, "loss": 0.1365, "step": 7183 }, { "epoch": 0.46, "learning_rate": 0.00011751931686879177, "loss": 0.1449, "step": 7184 }, { "epoch": 0.46, "learning_rate": 0.00011749887092644891, "loss": 0.1291, "step": 7185 }, { "epoch": 0.46, "learning_rate": 0.00011747842422945403, "loss": 0.1492, "step": 7186 }, { "epoch": 0.46, "learning_rate": 0.00011745797677868891, "loss": 0.1541, "step": 7187 }, { "epoch": 0.46, "learning_rate": 0.00011743752857503538, "loss": 0.1526, "step": 7188 }, { "epoch": 0.46, "learning_rate": 0.00011741707961937525, "loss": 0.1634, "step": 7189 }, { "epoch": 0.46, "learning_rate": 0.00011739662991259042, "loss": 0.1709, "step": 7190 }, { "epoch": 0.46, "learning_rate": 0.00011737617945556276, "loss": 0.1577, "step": 7191 }, { "epoch": 0.46, "learning_rate": 0.00011735572824917427, "loss": 0.1564, "step": 7192 }, { "epoch": 0.46, "learning_rate": 0.00011733527629430688, "loss": 0.1349, "step": 7193 }, { "epoch": 0.46, "learning_rate": 0.0001173148235918426, "loss": 0.1628, "step": 7194 }, { "epoch": 0.46, "learning_rate": 0.00011729437014266351, "loss": 0.1294, "step": 7195 }, { "epoch": 0.46, "learning_rate": 0.00011727391594765163, "loss": 0.1461, "step": 7196 }, { "epoch": 0.46, "learning_rate": 0.00011725346100768908, "loss": 0.1582, "step": 7197 }, { "epoch": 0.46, "learning_rate": 0.000117233005323658, "loss": 0.1453, "step": 7198 }, { "epoch": 0.46, "learning_rate": 0.00011721254889644054, "loss": 0.1391, "step": 7199 }, { "epoch": 0.46, "learning_rate": 0.00011719209172691892, "loss": 0.1443, "step": 7200 }, { "epoch": 0.46, "learning_rate": 0.00011717163381597538, "loss": 0.1714, "step": 7201 }, { "epoch": 0.46, "learning_rate": 0.00011715117516449214, "loss": 0.1503, "step": 7202 }, { "epoch": 0.46, "learning_rate": 0.00011713071577335156, "loss": 0.153, "step": 7203 }, { "epoch": 0.46, "learning_rate": 0.0001171102556434359, "loss": 0.1384, "step": 7204 }, { "epoch": 0.46, "learning_rate": 0.00011708979477562754, "loss": 0.1483, "step": 7205 }, { "epoch": 0.46, "learning_rate": 0.00011706933317080885, "loss": 0.1296, "step": 7206 }, { "epoch": 0.46, "learning_rate": 0.00011704887082986232, "loss": 0.1444, "step": 7207 }, { "epoch": 0.46, "learning_rate": 0.00011702840775367034, "loss": 0.1436, "step": 7208 }, { "epoch": 0.46, "learning_rate": 0.00011700794394311544, "loss": 0.1461, "step": 7209 }, { "epoch": 0.46, "learning_rate": 0.00011698747939908012, "loss": 0.1632, "step": 7210 }, { "epoch": 0.46, "learning_rate": 0.00011696701412244691, "loss": 0.1236, "step": 7211 }, { "epoch": 0.46, "learning_rate": 0.0001169465481140984, "loss": 0.1405, "step": 7212 }, { "epoch": 0.46, "learning_rate": 0.0001169260813749172, "loss": 0.1495, "step": 7213 }, { "epoch": 0.46, "learning_rate": 0.00011690561390578595, "loss": 0.1346, "step": 7214 }, { "epoch": 0.46, "learning_rate": 0.00011688514570758737, "loss": 0.1467, "step": 7215 }, { "epoch": 0.46, "learning_rate": 0.0001168646767812041, "loss": 0.14, "step": 7216 }, { "epoch": 0.46, "learning_rate": 0.00011684420712751896, "loss": 0.161, "step": 7217 }, { "epoch": 0.46, "learning_rate": 0.00011682373674741462, "loss": 0.1367, "step": 7218 }, { "epoch": 0.46, "learning_rate": 0.00011680326564177394, "loss": 0.161, "step": 7219 }, { "epoch": 0.46, "learning_rate": 0.00011678279381147975, "loss": 0.1543, "step": 7220 }, { "epoch": 0.46, "learning_rate": 0.00011676232125741491, "loss": 0.1472, "step": 7221 }, { "epoch": 0.46, "learning_rate": 0.00011674184798046231, "loss": 0.139, "step": 7222 }, { "epoch": 0.46, "learning_rate": 0.00011672137398150482, "loss": 0.1584, "step": 7223 }, { "epoch": 0.46, "learning_rate": 0.00011670089926142553, "loss": 0.153, "step": 7224 }, { "epoch": 0.46, "learning_rate": 0.0001166804238211073, "loss": 0.158, "step": 7225 }, { "epoch": 0.46, "learning_rate": 0.0001166599476614332, "loss": 0.1653, "step": 7226 }, { "epoch": 0.46, "learning_rate": 0.0001166394707832863, "loss": 0.1469, "step": 7227 }, { "epoch": 0.46, "learning_rate": 0.00011661899318754965, "loss": 0.1371, "step": 7228 }, { "epoch": 0.46, "learning_rate": 0.00011659851487510637, "loss": 0.1452, "step": 7229 }, { "epoch": 0.46, "learning_rate": 0.0001165780358468396, "loss": 0.1479, "step": 7230 }, { "epoch": 0.46, "learning_rate": 0.00011655755610363252, "loss": 0.1426, "step": 7231 }, { "epoch": 0.46, "learning_rate": 0.00011653707564636833, "loss": 0.1781, "step": 7232 }, { "epoch": 0.46, "learning_rate": 0.00011651659447593024, "loss": 0.1406, "step": 7233 }, { "epoch": 0.46, "learning_rate": 0.00011649611259320157, "loss": 0.1311, "step": 7234 }, { "epoch": 0.46, "learning_rate": 0.00011647562999906558, "loss": 0.154, "step": 7235 }, { "epoch": 0.46, "learning_rate": 0.0001164551466944056, "loss": 0.1655, "step": 7236 }, { "epoch": 0.46, "learning_rate": 0.00011643466268010503, "loss": 0.1389, "step": 7237 }, { "epoch": 0.46, "learning_rate": 0.00011641417795704718, "loss": 0.1625, "step": 7238 }, { "epoch": 0.46, "learning_rate": 0.00011639369252611552, "loss": 0.1413, "step": 7239 }, { "epoch": 0.46, "learning_rate": 0.00011637320638819349, "loss": 0.1509, "step": 7240 }, { "epoch": 0.46, "learning_rate": 0.00011635271954416461, "loss": 0.1236, "step": 7241 }, { "epoch": 0.46, "learning_rate": 0.0001163322319949123, "loss": 0.16, "step": 7242 }, { "epoch": 0.46, "learning_rate": 0.00011631174374132016, "loss": 0.1499, "step": 7243 }, { "epoch": 0.46, "learning_rate": 0.00011629125478427178, "loss": 0.1436, "step": 7244 }, { "epoch": 0.46, "learning_rate": 0.00011627076512465074, "loss": 0.1705, "step": 7245 }, { "epoch": 0.46, "learning_rate": 0.00011625027476334063, "loss": 0.1591, "step": 7246 }, { "epoch": 0.46, "learning_rate": 0.00011622978370122519, "loss": 0.1531, "step": 7247 }, { "epoch": 0.46, "learning_rate": 0.00011620929193918806, "loss": 0.1629, "step": 7248 }, { "epoch": 0.46, "learning_rate": 0.00011618879947811296, "loss": 0.1435, "step": 7249 }, { "epoch": 0.46, "learning_rate": 0.00011616830631888367, "loss": 0.1599, "step": 7250 }, { "epoch": 0.46, "learning_rate": 0.00011614781246238398, "loss": 0.1645, "step": 7251 }, { "epoch": 0.46, "learning_rate": 0.00011612731790949767, "loss": 0.1588, "step": 7252 }, { "epoch": 0.47, "learning_rate": 0.00011610682266110858, "loss": 0.1405, "step": 7253 }, { "epoch": 0.47, "learning_rate": 0.00011608632671810063, "loss": 0.1366, "step": 7254 }, { "epoch": 0.47, "learning_rate": 0.00011606583008135768, "loss": 0.1617, "step": 7255 }, { "epoch": 0.47, "learning_rate": 0.00011604533275176366, "loss": 0.1284, "step": 7256 }, { "epoch": 0.47, "learning_rate": 0.00011602483473020256, "loss": 0.1432, "step": 7257 }, { "epoch": 0.47, "learning_rate": 0.00011600433601755838, "loss": 0.1362, "step": 7258 }, { "epoch": 0.47, "learning_rate": 0.00011598383661471513, "loss": 0.1412, "step": 7259 }, { "epoch": 0.47, "learning_rate": 0.0001159633365225568, "loss": 0.1584, "step": 7260 }, { "epoch": 0.47, "learning_rate": 0.00011594283574196758, "loss": 0.151, "step": 7261 }, { "epoch": 0.47, "learning_rate": 0.00011592233427383152, "loss": 0.1477, "step": 7262 }, { "epoch": 0.47, "learning_rate": 0.00011590183211903276, "loss": 0.1497, "step": 7263 }, { "epoch": 0.47, "learning_rate": 0.00011588132927845552, "loss": 0.1565, "step": 7264 }, { "epoch": 0.47, "learning_rate": 0.00011586082575298394, "loss": 0.1536, "step": 7265 }, { "epoch": 0.47, "learning_rate": 0.00011584032154350223, "loss": 0.1667, "step": 7266 }, { "epoch": 0.47, "learning_rate": 0.00011581981665089475, "loss": 0.1632, "step": 7267 }, { "epoch": 0.47, "learning_rate": 0.00011579931107604571, "loss": 0.1448, "step": 7268 }, { "epoch": 0.47, "learning_rate": 0.00011577880481983945, "loss": 0.1643, "step": 7269 }, { "epoch": 0.47, "learning_rate": 0.0001157582978831603, "loss": 0.144, "step": 7270 }, { "epoch": 0.47, "learning_rate": 0.0001157377902668927, "loss": 0.1481, "step": 7271 }, { "epoch": 0.47, "learning_rate": 0.000115717281971921, "loss": 0.1676, "step": 7272 }, { "epoch": 0.47, "learning_rate": 0.00011569677299912962, "loss": 0.1242, "step": 7273 }, { "epoch": 0.47, "learning_rate": 0.00011567626334940307, "loss": 0.1385, "step": 7274 }, { "epoch": 0.47, "learning_rate": 0.00011565575302362581, "loss": 0.1441, "step": 7275 }, { "epoch": 0.47, "learning_rate": 0.00011563524202268241, "loss": 0.1487, "step": 7276 }, { "epoch": 0.47, "learning_rate": 0.00011561473034745735, "loss": 0.1444, "step": 7277 }, { "epoch": 0.47, "learning_rate": 0.0001155942179988353, "loss": 0.1684, "step": 7278 }, { "epoch": 0.47, "learning_rate": 0.00011557370497770081, "loss": 0.1307, "step": 7279 }, { "epoch": 0.47, "learning_rate": 0.00011555319128493851, "loss": 0.1565, "step": 7280 }, { "epoch": 0.47, "learning_rate": 0.0001155326769214331, "loss": 0.1337, "step": 7281 }, { "epoch": 0.47, "learning_rate": 0.00011551216188806928, "loss": 0.1515, "step": 7282 }, { "epoch": 0.47, "learning_rate": 0.00011549164618573174, "loss": 0.1532, "step": 7283 }, { "epoch": 0.47, "learning_rate": 0.00011547112981530531, "loss": 0.1524, "step": 7284 }, { "epoch": 0.47, "learning_rate": 0.0001154506127776747, "loss": 0.1428, "step": 7285 }, { "epoch": 0.47, "learning_rate": 0.00011543009507372473, "loss": 0.1694, "step": 7286 }, { "epoch": 0.47, "learning_rate": 0.00011540957670434027, "loss": 0.1529, "step": 7287 }, { "epoch": 0.47, "learning_rate": 0.0001153890576704062, "loss": 0.1263, "step": 7288 }, { "epoch": 0.47, "learning_rate": 0.00011536853797280737, "loss": 0.1271, "step": 7289 }, { "epoch": 0.47, "learning_rate": 0.00011534801761242872, "loss": 0.1642, "step": 7290 }, { "epoch": 0.47, "learning_rate": 0.00011532749659015527, "loss": 0.1505, "step": 7291 }, { "epoch": 0.47, "learning_rate": 0.00011530697490687194, "loss": 0.1532, "step": 7292 }, { "epoch": 0.47, "learning_rate": 0.00011528645256346375, "loss": 0.1566, "step": 7293 }, { "epoch": 0.47, "learning_rate": 0.00011526592956081575, "loss": 0.1574, "step": 7294 }, { "epoch": 0.47, "learning_rate": 0.00011524540589981301, "loss": 0.1473, "step": 7295 }, { "epoch": 0.47, "learning_rate": 0.00011522488158134064, "loss": 0.1451, "step": 7296 }, { "epoch": 0.47, "learning_rate": 0.00011520435660628373, "loss": 0.1724, "step": 7297 }, { "epoch": 0.47, "learning_rate": 0.0001151838309755275, "loss": 0.1428, "step": 7298 }, { "epoch": 0.47, "learning_rate": 0.00011516330468995708, "loss": 0.146, "step": 7299 }, { "epoch": 0.47, "learning_rate": 0.00011514277775045768, "loss": 0.1482, "step": 7300 }, { "epoch": 0.47, "learning_rate": 0.00011512225015791455, "loss": 0.1544, "step": 7301 }, { "epoch": 0.47, "learning_rate": 0.000115101721913213, "loss": 0.145, "step": 7302 }, { "epoch": 0.47, "learning_rate": 0.00011508119301723824, "loss": 0.1459, "step": 7303 }, { "epoch": 0.47, "learning_rate": 0.00011506066347087568, "loss": 0.1305, "step": 7304 }, { "epoch": 0.47, "learning_rate": 0.00011504013327501064, "loss": 0.1634, "step": 7305 }, { "epoch": 0.47, "learning_rate": 0.00011501960243052847, "loss": 0.1579, "step": 7306 }, { "epoch": 0.47, "learning_rate": 0.00011499907093831463, "loss": 0.1518, "step": 7307 }, { "epoch": 0.47, "learning_rate": 0.0001149785387992545, "loss": 0.1569, "step": 7308 }, { "epoch": 0.47, "learning_rate": 0.0001149580060142336, "loss": 0.161, "step": 7309 }, { "epoch": 0.47, "learning_rate": 0.00011493747258413738, "loss": 0.1643, "step": 7310 }, { "epoch": 0.47, "learning_rate": 0.00011491693850985139, "loss": 0.1567, "step": 7311 }, { "epoch": 0.47, "learning_rate": 0.00011489640379226116, "loss": 0.149, "step": 7312 }, { "epoch": 0.47, "learning_rate": 0.00011487586843225227, "loss": 0.1554, "step": 7313 }, { "epoch": 0.47, "learning_rate": 0.00011485533243071032, "loss": 0.1506, "step": 7314 }, { "epoch": 0.47, "learning_rate": 0.00011483479578852093, "loss": 0.1545, "step": 7315 }, { "epoch": 0.47, "learning_rate": 0.00011481425850656978, "loss": 0.1248, "step": 7316 }, { "epoch": 0.47, "learning_rate": 0.00011479372058574252, "loss": 0.1453, "step": 7317 }, { "epoch": 0.47, "learning_rate": 0.00011477318202692493, "loss": 0.1538, "step": 7318 }, { "epoch": 0.47, "learning_rate": 0.0001147526428310027, "loss": 0.1405, "step": 7319 }, { "epoch": 0.47, "learning_rate": 0.0001147321029988616, "loss": 0.1415, "step": 7320 }, { "epoch": 0.47, "learning_rate": 0.00011471156253138744, "loss": 0.1454, "step": 7321 }, { "epoch": 0.47, "learning_rate": 0.00011469102142946603, "loss": 0.1519, "step": 7322 }, { "epoch": 0.47, "learning_rate": 0.00011467047969398321, "loss": 0.1484, "step": 7323 }, { "epoch": 0.47, "learning_rate": 0.0001146499373258249, "loss": 0.1455, "step": 7324 }, { "epoch": 0.47, "learning_rate": 0.00011462939432587698, "loss": 0.129, "step": 7325 }, { "epoch": 0.47, "learning_rate": 0.00011460885069502538, "loss": 0.1513, "step": 7326 }, { "epoch": 0.47, "learning_rate": 0.00011458830643415608, "loss": 0.1595, "step": 7327 }, { "epoch": 0.47, "learning_rate": 0.00011456776154415502, "loss": 0.1677, "step": 7328 }, { "epoch": 0.47, "learning_rate": 0.00011454721602590828, "loss": 0.1697, "step": 7329 }, { "epoch": 0.47, "learning_rate": 0.0001145266698803018, "loss": 0.1489, "step": 7330 }, { "epoch": 0.47, "learning_rate": 0.00011450612310822177, "loss": 0.1582, "step": 7331 }, { "epoch": 0.47, "learning_rate": 0.00011448557571055423, "loss": 0.1502, "step": 7332 }, { "epoch": 0.47, "learning_rate": 0.00011446502768818528, "loss": 0.1429, "step": 7333 }, { "epoch": 0.47, "learning_rate": 0.00011444447904200109, "loss": 0.1486, "step": 7334 }, { "epoch": 0.47, "learning_rate": 0.00011442392977288783, "loss": 0.1606, "step": 7335 }, { "epoch": 0.47, "learning_rate": 0.00011440337988173173, "loss": 0.1548, "step": 7336 }, { "epoch": 0.47, "learning_rate": 0.00011438282936941896, "loss": 0.1583, "step": 7337 }, { "epoch": 0.47, "learning_rate": 0.00011436227823683583, "loss": 0.1386, "step": 7338 }, { "epoch": 0.47, "learning_rate": 0.00011434172648486861, "loss": 0.1355, "step": 7339 }, { "epoch": 0.47, "learning_rate": 0.00011432117411440359, "loss": 0.1298, "step": 7340 }, { "epoch": 0.47, "learning_rate": 0.00011430062112632716, "loss": 0.1562, "step": 7341 }, { "epoch": 0.47, "learning_rate": 0.00011428006752152563, "loss": 0.148, "step": 7342 }, { "epoch": 0.47, "learning_rate": 0.0001142595133008854, "loss": 0.1617, "step": 7343 }, { "epoch": 0.47, "learning_rate": 0.00011423895846529286, "loss": 0.1532, "step": 7344 }, { "epoch": 0.47, "learning_rate": 0.00011421840301563453, "loss": 0.1491, "step": 7345 }, { "epoch": 0.47, "learning_rate": 0.00011419784695279683, "loss": 0.1452, "step": 7346 }, { "epoch": 0.47, "learning_rate": 0.00011417729027766624, "loss": 0.1568, "step": 7347 }, { "epoch": 0.47, "learning_rate": 0.00011415673299112932, "loss": 0.1429, "step": 7348 }, { "epoch": 0.47, "learning_rate": 0.00011413617509407259, "loss": 0.1541, "step": 7349 }, { "epoch": 0.47, "learning_rate": 0.0001141156165873826, "loss": 0.1579, "step": 7350 }, { "epoch": 0.47, "learning_rate": 0.00011409505747194604, "loss": 0.1576, "step": 7351 }, { "epoch": 0.47, "learning_rate": 0.00011407449774864947, "loss": 0.1692, "step": 7352 }, { "epoch": 0.47, "learning_rate": 0.00011405393741837955, "loss": 0.1459, "step": 7353 }, { "epoch": 0.47, "learning_rate": 0.00011403337648202296, "loss": 0.1517, "step": 7354 }, { "epoch": 0.47, "learning_rate": 0.00011401281494046645, "loss": 0.1463, "step": 7355 }, { "epoch": 0.47, "learning_rate": 0.00011399225279459669, "loss": 0.1567, "step": 7356 }, { "epoch": 0.47, "learning_rate": 0.00011397169004530044, "loss": 0.1457, "step": 7357 }, { "epoch": 0.47, "learning_rate": 0.00011395112669346452, "loss": 0.1346, "step": 7358 }, { "epoch": 0.47, "learning_rate": 0.00011393056273997573, "loss": 0.157, "step": 7359 }, { "epoch": 0.47, "learning_rate": 0.00011390999818572093, "loss": 0.1577, "step": 7360 }, { "epoch": 0.47, "learning_rate": 0.00011388943303158693, "loss": 0.1512, "step": 7361 }, { "epoch": 0.47, "learning_rate": 0.00011386886727846065, "loss": 0.1341, "step": 7362 }, { "epoch": 0.47, "learning_rate": 0.00011384830092722898, "loss": 0.133, "step": 7363 }, { "epoch": 0.47, "learning_rate": 0.00011382773397877888, "loss": 0.1534, "step": 7364 }, { "epoch": 0.47, "learning_rate": 0.00011380716643399731, "loss": 0.1233, "step": 7365 }, { "epoch": 0.47, "learning_rate": 0.00011378659829377127, "loss": 0.1406, "step": 7366 }, { "epoch": 0.47, "learning_rate": 0.00011376602955898778, "loss": 0.1703, "step": 7367 }, { "epoch": 0.47, "learning_rate": 0.00011374546023053388, "loss": 0.1677, "step": 7368 }, { "epoch": 0.47, "learning_rate": 0.00011372489030929661, "loss": 0.1342, "step": 7369 }, { "epoch": 0.47, "learning_rate": 0.00011370431979616306, "loss": 0.1522, "step": 7370 }, { "epoch": 0.47, "learning_rate": 0.00011368374869202038, "loss": 0.132, "step": 7371 }, { "epoch": 0.47, "learning_rate": 0.00011366317699775573, "loss": 0.1389, "step": 7372 }, { "epoch": 0.47, "learning_rate": 0.00011364260471425626, "loss": 0.1544, "step": 7373 }, { "epoch": 0.47, "learning_rate": 0.00011362203184240914, "loss": 0.1273, "step": 7374 }, { "epoch": 0.47, "learning_rate": 0.00011360145838310162, "loss": 0.1479, "step": 7375 }, { "epoch": 0.47, "learning_rate": 0.00011358088433722092, "loss": 0.1438, "step": 7376 }, { "epoch": 0.47, "learning_rate": 0.00011356030970565433, "loss": 0.1455, "step": 7377 }, { "epoch": 0.47, "learning_rate": 0.00011353973448928914, "loss": 0.1564, "step": 7378 }, { "epoch": 0.47, "learning_rate": 0.0001135191586890127, "loss": 0.128, "step": 7379 }, { "epoch": 0.47, "learning_rate": 0.00011349858230571233, "loss": 0.1685, "step": 7380 }, { "epoch": 0.47, "learning_rate": 0.0001134780053402754, "loss": 0.1357, "step": 7381 }, { "epoch": 0.47, "learning_rate": 0.0001134574277935893, "loss": 0.1679, "step": 7382 }, { "epoch": 0.47, "learning_rate": 0.00011343684966654149, "loss": 0.1415, "step": 7383 }, { "epoch": 0.47, "learning_rate": 0.00011341627096001937, "loss": 0.1682, "step": 7384 }, { "epoch": 0.47, "learning_rate": 0.00011339569167491042, "loss": 0.147, "step": 7385 }, { "epoch": 0.47, "learning_rate": 0.00011337511181210218, "loss": 0.157, "step": 7386 }, { "epoch": 0.47, "learning_rate": 0.00011335453137248215, "loss": 0.15, "step": 7387 }, { "epoch": 0.47, "learning_rate": 0.00011333395035693786, "loss": 0.1425, "step": 7388 }, { "epoch": 0.47, "learning_rate": 0.00011331336876635689, "loss": 0.1505, "step": 7389 }, { "epoch": 0.47, "learning_rate": 0.00011329278660162684, "loss": 0.1605, "step": 7390 }, { "epoch": 0.47, "learning_rate": 0.00011327220386363533, "loss": 0.1525, "step": 7391 }, { "epoch": 0.47, "learning_rate": 0.00011325162055327001, "loss": 0.1545, "step": 7392 }, { "epoch": 0.47, "learning_rate": 0.00011323103667141855, "loss": 0.1638, "step": 7393 }, { "epoch": 0.47, "learning_rate": 0.00011321045221896865, "loss": 0.1392, "step": 7394 }, { "epoch": 0.47, "learning_rate": 0.00011318986719680799, "loss": 0.1604, "step": 7395 }, { "epoch": 0.47, "learning_rate": 0.00011316928160582438, "loss": 0.1438, "step": 7396 }, { "epoch": 0.47, "learning_rate": 0.00011314869544690554, "loss": 0.1323, "step": 7397 }, { "epoch": 0.47, "learning_rate": 0.00011312810872093928, "loss": 0.1358, "step": 7398 }, { "epoch": 0.47, "learning_rate": 0.00011310752142881344, "loss": 0.1472, "step": 7399 }, { "epoch": 0.47, "learning_rate": 0.00011308693357141582, "loss": 0.1801, "step": 7400 }, { "epoch": 0.47, "learning_rate": 0.00011306634514963433, "loss": 0.16, "step": 7401 }, { "epoch": 0.47, "learning_rate": 0.00011304575616435682, "loss": 0.1586, "step": 7402 }, { "epoch": 0.47, "learning_rate": 0.00011302516661647123, "loss": 0.1414, "step": 7403 }, { "epoch": 0.47, "learning_rate": 0.00011300457650686547, "loss": 0.1621, "step": 7404 }, { "epoch": 0.47, "learning_rate": 0.00011298398583642756, "loss": 0.1649, "step": 7405 }, { "epoch": 0.47, "learning_rate": 0.00011296339460604545, "loss": 0.1463, "step": 7406 }, { "epoch": 0.47, "learning_rate": 0.00011294280281660717, "loss": 0.1634, "step": 7407 }, { "epoch": 0.47, "learning_rate": 0.00011292221046900074, "loss": 0.1582, "step": 7408 }, { "epoch": 0.48, "learning_rate": 0.00011290161756411421, "loss": 0.1341, "step": 7409 }, { "epoch": 0.48, "learning_rate": 0.00011288102410283567, "loss": 0.1404, "step": 7410 }, { "epoch": 0.48, "learning_rate": 0.00011286043008605327, "loss": 0.1484, "step": 7411 }, { "epoch": 0.48, "learning_rate": 0.00011283983551465511, "loss": 0.1663, "step": 7412 }, { "epoch": 0.48, "learning_rate": 0.00011281924038952933, "loss": 0.1555, "step": 7413 }, { "epoch": 0.48, "learning_rate": 0.00011279864471156414, "loss": 0.1497, "step": 7414 }, { "epoch": 0.48, "learning_rate": 0.00011277804848164774, "loss": 0.1444, "step": 7415 }, { "epoch": 0.48, "learning_rate": 0.00011275745170066834, "loss": 0.1366, "step": 7416 }, { "epoch": 0.48, "learning_rate": 0.0001127368543695142, "loss": 0.1679, "step": 7417 }, { "epoch": 0.48, "learning_rate": 0.0001127162564890736, "loss": 0.1399, "step": 7418 }, { "epoch": 0.48, "learning_rate": 0.00011269565806023483, "loss": 0.1488, "step": 7419 }, { "epoch": 0.48, "learning_rate": 0.00011267505908388625, "loss": 0.1566, "step": 7420 }, { "epoch": 0.48, "learning_rate": 0.00011265445956091613, "loss": 0.1434, "step": 7421 }, { "epoch": 0.48, "learning_rate": 0.00011263385949221295, "loss": 0.146, "step": 7422 }, { "epoch": 0.48, "learning_rate": 0.00011261325887866498, "loss": 0.1352, "step": 7423 }, { "epoch": 0.48, "learning_rate": 0.00011259265772116072, "loss": 0.16, "step": 7424 }, { "epoch": 0.48, "learning_rate": 0.00011257205602058858, "loss": 0.1475, "step": 7425 }, { "epoch": 0.48, "learning_rate": 0.00011255145377783704, "loss": 0.1586, "step": 7426 }, { "epoch": 0.48, "learning_rate": 0.00011253085099379459, "loss": 0.1694, "step": 7427 }, { "epoch": 0.48, "learning_rate": 0.00011251024766934973, "loss": 0.1389, "step": 7428 }, { "epoch": 0.48, "learning_rate": 0.00011248964380539096, "loss": 0.1302, "step": 7429 }, { "epoch": 0.48, "learning_rate": 0.00011246903940280689, "loss": 0.142, "step": 7430 }, { "epoch": 0.48, "learning_rate": 0.00011244843446248605, "loss": 0.1515, "step": 7431 }, { "epoch": 0.48, "learning_rate": 0.00011242782898531712, "loss": 0.1619, "step": 7432 }, { "epoch": 0.48, "learning_rate": 0.00011240722297218867, "loss": 0.1566, "step": 7433 }, { "epoch": 0.48, "learning_rate": 0.00011238661642398933, "loss": 0.1552, "step": 7434 }, { "epoch": 0.48, "learning_rate": 0.00011236600934160787, "loss": 0.1327, "step": 7435 }, { "epoch": 0.48, "learning_rate": 0.00011234540172593287, "loss": 0.1799, "step": 7436 }, { "epoch": 0.48, "learning_rate": 0.00011232479357785308, "loss": 0.1355, "step": 7437 }, { "epoch": 0.48, "learning_rate": 0.0001123041848982573, "loss": 0.1519, "step": 7438 }, { "epoch": 0.48, "learning_rate": 0.00011228357568803423, "loss": 0.1567, "step": 7439 }, { "epoch": 0.48, "learning_rate": 0.00011226296594807271, "loss": 0.1559, "step": 7440 }, { "epoch": 0.48, "learning_rate": 0.00011224235567926149, "loss": 0.1475, "step": 7441 }, { "epoch": 0.48, "learning_rate": 0.00011222174488248949, "loss": 0.1433, "step": 7442 }, { "epoch": 0.48, "learning_rate": 0.00011220113355864547, "loss": 0.1377, "step": 7443 }, { "epoch": 0.48, "learning_rate": 0.00011218052170861835, "loss": 0.1462, "step": 7444 }, { "epoch": 0.48, "learning_rate": 0.00011215990933329705, "loss": 0.1531, "step": 7445 }, { "epoch": 0.48, "learning_rate": 0.00011213929643357048, "loss": 0.1506, "step": 7446 }, { "epoch": 0.48, "learning_rate": 0.00011211868301032758, "loss": 0.1316, "step": 7447 }, { "epoch": 0.48, "learning_rate": 0.0001120980690644573, "loss": 0.1441, "step": 7448 }, { "epoch": 0.48, "learning_rate": 0.00011207745459684871, "loss": 0.1649, "step": 7449 }, { "epoch": 0.48, "learning_rate": 0.00011205683960839075, "loss": 0.1406, "step": 7450 }, { "epoch": 0.48, "learning_rate": 0.00011203622409997245, "loss": 0.1645, "step": 7451 }, { "epoch": 0.48, "learning_rate": 0.00011201560807248293, "loss": 0.1404, "step": 7452 }, { "epoch": 0.48, "learning_rate": 0.00011199499152681123, "loss": 0.1588, "step": 7453 }, { "epoch": 0.48, "learning_rate": 0.00011197437446384644, "loss": 0.1501, "step": 7454 }, { "epoch": 0.48, "learning_rate": 0.00011195375688447773, "loss": 0.1484, "step": 7455 }, { "epoch": 0.48, "learning_rate": 0.00011193313878959427, "loss": 0.1422, "step": 7456 }, { "epoch": 0.48, "learning_rate": 0.00011191252018008512, "loss": 0.1432, "step": 7457 }, { "epoch": 0.48, "learning_rate": 0.00011189190105683958, "loss": 0.152, "step": 7458 }, { "epoch": 0.48, "learning_rate": 0.00011187128142074681, "loss": 0.1483, "step": 7459 }, { "epoch": 0.48, "learning_rate": 0.00011185066127269607, "loss": 0.1519, "step": 7460 }, { "epoch": 0.48, "learning_rate": 0.00011183004061357661, "loss": 0.1616, "step": 7461 }, { "epoch": 0.48, "learning_rate": 0.00011180941944427774, "loss": 0.1465, "step": 7462 }, { "epoch": 0.48, "learning_rate": 0.00011178879776568871, "loss": 0.1428, "step": 7463 }, { "epoch": 0.48, "learning_rate": 0.00011176817557869887, "loss": 0.148, "step": 7464 }, { "epoch": 0.48, "learning_rate": 0.0001117475528841976, "loss": 0.1251, "step": 7465 }, { "epoch": 0.48, "learning_rate": 0.00011172692968307424, "loss": 0.1515, "step": 7466 }, { "epoch": 0.48, "learning_rate": 0.00011170630597621819, "loss": 0.162, "step": 7467 }, { "epoch": 0.48, "learning_rate": 0.00011168568176451881, "loss": 0.1517, "step": 7468 }, { "epoch": 0.48, "learning_rate": 0.00011166505704886563, "loss": 0.1485, "step": 7469 }, { "epoch": 0.48, "learning_rate": 0.00011164443183014803, "loss": 0.1491, "step": 7470 }, { "epoch": 0.48, "learning_rate": 0.00011162380610925552, "loss": 0.1497, "step": 7471 }, { "epoch": 0.48, "learning_rate": 0.00011160317988707759, "loss": 0.14, "step": 7472 }, { "epoch": 0.48, "learning_rate": 0.00011158255316450376, "loss": 0.1406, "step": 7473 }, { "epoch": 0.48, "learning_rate": 0.00011156192594242359, "loss": 0.1274, "step": 7474 }, { "epoch": 0.48, "learning_rate": 0.00011154129822172663, "loss": 0.1506, "step": 7475 }, { "epoch": 0.48, "learning_rate": 0.0001115206700033025, "loss": 0.1401, "step": 7476 }, { "epoch": 0.48, "learning_rate": 0.00011150004128804072, "loss": 0.1429, "step": 7477 }, { "epoch": 0.48, "learning_rate": 0.00011147941207683101, "loss": 0.1394, "step": 7478 }, { "epoch": 0.48, "learning_rate": 0.00011145878237056299, "loss": 0.1639, "step": 7479 }, { "epoch": 0.48, "learning_rate": 0.00011143815217012633, "loss": 0.1538, "step": 7480 }, { "epoch": 0.48, "learning_rate": 0.00011141752147641069, "loss": 0.1488, "step": 7481 }, { "epoch": 0.48, "learning_rate": 0.00011139689029030585, "loss": 0.1714, "step": 7482 }, { "epoch": 0.48, "learning_rate": 0.00011137625861270151, "loss": 0.1556, "step": 7483 }, { "epoch": 0.48, "learning_rate": 0.00011135562644448741, "loss": 0.1408, "step": 7484 }, { "epoch": 0.48, "learning_rate": 0.00011133499378655335, "loss": 0.155, "step": 7485 }, { "epoch": 0.48, "learning_rate": 0.00011131436063978911, "loss": 0.1455, "step": 7486 }, { "epoch": 0.48, "learning_rate": 0.00011129372700508453, "loss": 0.1446, "step": 7487 }, { "epoch": 0.48, "learning_rate": 0.00011127309288332945, "loss": 0.1605, "step": 7488 }, { "epoch": 0.48, "learning_rate": 0.00011125245827541372, "loss": 0.1469, "step": 7489 }, { "epoch": 0.48, "learning_rate": 0.00011123182318222726, "loss": 0.1711, "step": 7490 }, { "epoch": 0.48, "learning_rate": 0.00011121118760465992, "loss": 0.1523, "step": 7491 }, { "epoch": 0.48, "learning_rate": 0.00011119055154360165, "loss": 0.1735, "step": 7492 }, { "epoch": 0.48, "learning_rate": 0.00011116991499994241, "loss": 0.1533, "step": 7493 }, { "epoch": 0.48, "learning_rate": 0.00011114927797457216, "loss": 0.1345, "step": 7494 }, { "epoch": 0.48, "learning_rate": 0.00011112864046838083, "loss": 0.1596, "step": 7495 }, { "epoch": 0.48, "learning_rate": 0.0001111080024822585, "loss": 0.1586, "step": 7496 }, { "epoch": 0.48, "learning_rate": 0.00011108736401709518, "loss": 0.1504, "step": 7497 }, { "epoch": 0.48, "learning_rate": 0.00011106672507378093, "loss": 0.1552, "step": 7498 }, { "epoch": 0.48, "learning_rate": 0.0001110460856532058, "loss": 0.1186, "step": 7499 }, { "epoch": 0.48, "learning_rate": 0.00011102544575625988, "loss": 0.1319, "step": 7500 }, { "epoch": 0.48, "learning_rate": 0.00011100480538383327, "loss": 0.1396, "step": 7501 }, { "epoch": 0.48, "learning_rate": 0.00011098416453681613, "loss": 0.1552, "step": 7502 }, { "epoch": 0.48, "learning_rate": 0.00011096352321609862, "loss": 0.1734, "step": 7503 }, { "epoch": 0.48, "learning_rate": 0.00011094288142257086, "loss": 0.1457, "step": 7504 }, { "epoch": 0.48, "learning_rate": 0.0001109222391571231, "loss": 0.1514, "step": 7505 }, { "epoch": 0.48, "learning_rate": 0.0001109015964206455, "loss": 0.1446, "step": 7506 }, { "epoch": 0.48, "learning_rate": 0.00011088095321402836, "loss": 0.1589, "step": 7507 }, { "epoch": 0.48, "learning_rate": 0.00011086030953816187, "loss": 0.1466, "step": 7508 }, { "epoch": 0.48, "learning_rate": 0.00011083966539393633, "loss": 0.1455, "step": 7509 }, { "epoch": 0.48, "learning_rate": 0.00011081902078224207, "loss": 0.1343, "step": 7510 }, { "epoch": 0.48, "learning_rate": 0.00011079837570396931, "loss": 0.149, "step": 7511 }, { "epoch": 0.48, "learning_rate": 0.00011077773016000849, "loss": 0.1266, "step": 7512 }, { "epoch": 0.48, "learning_rate": 0.00011075708415124993, "loss": 0.1403, "step": 7513 }, { "epoch": 0.48, "learning_rate": 0.00011073643767858395, "loss": 0.1407, "step": 7514 }, { "epoch": 0.48, "learning_rate": 0.000110715790742901, "loss": 0.1461, "step": 7515 }, { "epoch": 0.48, "learning_rate": 0.0001106951433450915, "loss": 0.1432, "step": 7516 }, { "epoch": 0.48, "learning_rate": 0.00011067449548604587, "loss": 0.1393, "step": 7517 }, { "epoch": 0.48, "learning_rate": 0.00011065384716665455, "loss": 0.1632, "step": 7518 }, { "epoch": 0.48, "learning_rate": 0.00011063319838780804, "loss": 0.1375, "step": 7519 }, { "epoch": 0.48, "learning_rate": 0.0001106125491503968, "loss": 0.1686, "step": 7520 }, { "epoch": 0.48, "learning_rate": 0.00011059189945531134, "loss": 0.162, "step": 7521 }, { "epoch": 0.48, "learning_rate": 0.00011057124930344227, "loss": 0.1448, "step": 7522 }, { "epoch": 0.48, "learning_rate": 0.00011055059869568007, "loss": 0.1469, "step": 7523 }, { "epoch": 0.48, "learning_rate": 0.00011052994763291533, "loss": 0.1628, "step": 7524 }, { "epoch": 0.48, "learning_rate": 0.00011050929611603867, "loss": 0.1346, "step": 7525 }, { "epoch": 0.48, "learning_rate": 0.00011048864414594066, "loss": 0.1386, "step": 7526 }, { "epoch": 0.48, "learning_rate": 0.00011046799172351197, "loss": 0.1582, "step": 7527 }, { "epoch": 0.48, "learning_rate": 0.0001104473388496432, "loss": 0.1734, "step": 7528 }, { "epoch": 0.48, "learning_rate": 0.00011042668552522506, "loss": 0.1605, "step": 7529 }, { "epoch": 0.48, "learning_rate": 0.00011040603175114826, "loss": 0.1424, "step": 7530 }, { "epoch": 0.48, "learning_rate": 0.00011038537752830347, "loss": 0.1389, "step": 7531 }, { "epoch": 0.48, "learning_rate": 0.00011036472285758144, "loss": 0.1385, "step": 7532 }, { "epoch": 0.48, "learning_rate": 0.00011034406773987289, "loss": 0.1357, "step": 7533 }, { "epoch": 0.48, "learning_rate": 0.00011032341217606863, "loss": 0.1413, "step": 7534 }, { "epoch": 0.48, "learning_rate": 0.0001103027561670594, "loss": 0.1605, "step": 7535 }, { "epoch": 0.48, "learning_rate": 0.00011028209971373605, "loss": 0.1606, "step": 7536 }, { "epoch": 0.48, "learning_rate": 0.00011026144281698938, "loss": 0.1556, "step": 7537 }, { "epoch": 0.48, "learning_rate": 0.00011024078547771023, "loss": 0.1378, "step": 7538 }, { "epoch": 0.48, "learning_rate": 0.00011022012769678952, "loss": 0.1673, "step": 7539 }, { "epoch": 0.48, "learning_rate": 0.00011019946947511804, "loss": 0.1308, "step": 7540 }, { "epoch": 0.48, "learning_rate": 0.00011017881081358672, "loss": 0.1394, "step": 7541 }, { "epoch": 0.48, "learning_rate": 0.00011015815171308655, "loss": 0.1282, "step": 7542 }, { "epoch": 0.48, "learning_rate": 0.00011013749217450839, "loss": 0.1524, "step": 7543 }, { "epoch": 0.48, "learning_rate": 0.00011011683219874323, "loss": 0.1298, "step": 7544 }, { "epoch": 0.48, "learning_rate": 0.00011009617178668203, "loss": 0.169, "step": 7545 }, { "epoch": 0.48, "learning_rate": 0.00011007551093921581, "loss": 0.1506, "step": 7546 }, { "epoch": 0.48, "learning_rate": 0.00011005484965723559, "loss": 0.1412, "step": 7547 }, { "epoch": 0.48, "learning_rate": 0.00011003418794163236, "loss": 0.1305, "step": 7548 }, { "epoch": 0.48, "learning_rate": 0.0001100135257932972, "loss": 0.1486, "step": 7549 }, { "epoch": 0.48, "learning_rate": 0.0001099928632131212, "loss": 0.1618, "step": 7550 }, { "epoch": 0.48, "learning_rate": 0.00010997220020199541, "loss": 0.1415, "step": 7551 }, { "epoch": 0.48, "learning_rate": 0.00010995153676081096, "loss": 0.1424, "step": 7552 }, { "epoch": 0.48, "learning_rate": 0.00010993087289045898, "loss": 0.1651, "step": 7553 }, { "epoch": 0.48, "learning_rate": 0.00010991020859183061, "loss": 0.151, "step": 7554 }, { "epoch": 0.48, "learning_rate": 0.00010988954386581698, "loss": 0.1348, "step": 7555 }, { "epoch": 0.48, "learning_rate": 0.00010986887871330934, "loss": 0.1612, "step": 7556 }, { "epoch": 0.48, "learning_rate": 0.00010984821313519884, "loss": 0.1454, "step": 7557 }, { "epoch": 0.48, "learning_rate": 0.00010982754713237671, "loss": 0.143, "step": 7558 }, { "epoch": 0.48, "learning_rate": 0.0001098068807057342, "loss": 0.159, "step": 7559 }, { "epoch": 0.48, "learning_rate": 0.00010978621385616255, "loss": 0.1414, "step": 7560 }, { "epoch": 0.48, "learning_rate": 0.00010976554658455303, "loss": 0.1586, "step": 7561 }, { "epoch": 0.48, "learning_rate": 0.00010974487889179692, "loss": 0.1352, "step": 7562 }, { "epoch": 0.48, "learning_rate": 0.00010972421077878559, "loss": 0.1316, "step": 7563 }, { "epoch": 0.48, "learning_rate": 0.0001097035422464103, "loss": 0.1509, "step": 7564 }, { "epoch": 0.49, "learning_rate": 0.00010968287329556246, "loss": 0.1575, "step": 7565 }, { "epoch": 0.49, "learning_rate": 0.00010966220392713338, "loss": 0.1378, "step": 7566 }, { "epoch": 0.49, "learning_rate": 0.00010964153414201446, "loss": 0.1434, "step": 7567 }, { "epoch": 0.49, "learning_rate": 0.00010962086394109705, "loss": 0.1678, "step": 7568 }, { "epoch": 0.49, "learning_rate": 0.00010960019332527267, "loss": 0.1581, "step": 7569 }, { "epoch": 0.49, "learning_rate": 0.00010957952229543271, "loss": 0.1526, "step": 7570 }, { "epoch": 0.49, "learning_rate": 0.0001095588508524686, "loss": 0.1523, "step": 7571 }, { "epoch": 0.49, "learning_rate": 0.00010953817899727183, "loss": 0.1508, "step": 7572 }, { "epoch": 0.49, "learning_rate": 0.00010951750673073389, "loss": 0.1423, "step": 7573 }, { "epoch": 0.49, "learning_rate": 0.00010949683405374628, "loss": 0.1335, "step": 7574 }, { "epoch": 0.49, "learning_rate": 0.00010947616096720053, "loss": 0.1712, "step": 7575 }, { "epoch": 0.49, "learning_rate": 0.00010945548747198819, "loss": 0.1343, "step": 7576 }, { "epoch": 0.49, "learning_rate": 0.00010943481356900083, "loss": 0.126, "step": 7577 }, { "epoch": 0.49, "learning_rate": 0.00010941413925913, "loss": 0.1376, "step": 7578 }, { "epoch": 0.49, "learning_rate": 0.00010939346454326728, "loss": 0.1519, "step": 7579 }, { "epoch": 0.49, "learning_rate": 0.00010937278942230435, "loss": 0.145, "step": 7580 }, { "epoch": 0.49, "learning_rate": 0.00010935211389713275, "loss": 0.1781, "step": 7581 }, { "epoch": 0.49, "learning_rate": 0.00010933143796864421, "loss": 0.1263, "step": 7582 }, { "epoch": 0.49, "learning_rate": 0.00010931076163773035, "loss": 0.1464, "step": 7583 }, { "epoch": 0.49, "learning_rate": 0.00010929008490528287, "loss": 0.1661, "step": 7584 }, { "epoch": 0.49, "learning_rate": 0.00010926940777219348, "loss": 0.1542, "step": 7585 }, { "epoch": 0.49, "learning_rate": 0.00010924873023935387, "loss": 0.1564, "step": 7586 }, { "epoch": 0.49, "learning_rate": 0.00010922805230765578, "loss": 0.1287, "step": 7587 }, { "epoch": 0.49, "learning_rate": 0.00010920737397799095, "loss": 0.1494, "step": 7588 }, { "epoch": 0.49, "learning_rate": 0.00010918669525125118, "loss": 0.1619, "step": 7589 }, { "epoch": 0.49, "learning_rate": 0.00010916601612832824, "loss": 0.1453, "step": 7590 }, { "epoch": 0.49, "learning_rate": 0.00010914533661011394, "loss": 0.1323, "step": 7591 }, { "epoch": 0.49, "learning_rate": 0.0001091246566975001, "loss": 0.1398, "step": 7592 }, { "epoch": 0.49, "learning_rate": 0.00010910397639137853, "loss": 0.1556, "step": 7593 }, { "epoch": 0.49, "learning_rate": 0.00010908329569264111, "loss": 0.1492, "step": 7594 }, { "epoch": 0.49, "learning_rate": 0.00010906261460217968, "loss": 0.1425, "step": 7595 }, { "epoch": 0.49, "learning_rate": 0.0001090419331208862, "loss": 0.1458, "step": 7596 }, { "epoch": 0.49, "learning_rate": 0.0001090212512496525, "loss": 0.1464, "step": 7597 }, { "epoch": 0.49, "learning_rate": 0.00010900056898937055, "loss": 0.1514, "step": 7598 }, { "epoch": 0.49, "learning_rate": 0.00010897988634093224, "loss": 0.1632, "step": 7599 }, { "epoch": 0.49, "learning_rate": 0.00010895920330522956, "loss": 0.1601, "step": 7600 }, { "epoch": 0.49, "learning_rate": 0.00010893851988315446, "loss": 0.1503, "step": 7601 }, { "epoch": 0.49, "learning_rate": 0.00010891783607559896, "loss": 0.1385, "step": 7602 }, { "epoch": 0.49, "learning_rate": 0.00010889715188345504, "loss": 0.1601, "step": 7603 }, { "epoch": 0.49, "learning_rate": 0.00010887646730761472, "loss": 0.15, "step": 7604 }, { "epoch": 0.49, "learning_rate": 0.00010885578234897003, "loss": 0.1437, "step": 7605 }, { "epoch": 0.49, "learning_rate": 0.00010883509700841307, "loss": 0.1271, "step": 7606 }, { "epoch": 0.49, "learning_rate": 0.00010881441128683589, "loss": 0.1509, "step": 7607 }, { "epoch": 0.49, "learning_rate": 0.00010879372518513053, "loss": 0.1389, "step": 7608 }, { "epoch": 0.49, "learning_rate": 0.00010877303870418916, "loss": 0.1406, "step": 7609 }, { "epoch": 0.49, "learning_rate": 0.00010875235184490384, "loss": 0.1344, "step": 7610 }, { "epoch": 0.49, "learning_rate": 0.00010873166460816677, "loss": 0.1639, "step": 7611 }, { "epoch": 0.49, "learning_rate": 0.00010871097699487005, "loss": 0.1621, "step": 7612 }, { "epoch": 0.49, "learning_rate": 0.0001086902890059059, "loss": 0.1436, "step": 7613 }, { "epoch": 0.49, "learning_rate": 0.00010866960064216644, "loss": 0.1368, "step": 7614 }, { "epoch": 0.49, "learning_rate": 0.00010864891190454391, "loss": 0.1512, "step": 7615 }, { "epoch": 0.49, "learning_rate": 0.00010862822279393054, "loss": 0.1477, "step": 7616 }, { "epoch": 0.49, "learning_rate": 0.00010860753331121854, "loss": 0.1609, "step": 7617 }, { "epoch": 0.49, "learning_rate": 0.00010858684345730017, "loss": 0.1359, "step": 7618 }, { "epoch": 0.49, "learning_rate": 0.00010856615323306769, "loss": 0.171, "step": 7619 }, { "epoch": 0.49, "learning_rate": 0.00010854546263941339, "loss": 0.1554, "step": 7620 }, { "epoch": 0.49, "learning_rate": 0.00010852477167722957, "loss": 0.1278, "step": 7621 }, { "epoch": 0.49, "learning_rate": 0.00010850408034740849, "loss": 0.1296, "step": 7622 }, { "epoch": 0.49, "learning_rate": 0.00010848338865084256, "loss": 0.1285, "step": 7623 }, { "epoch": 0.49, "learning_rate": 0.00010846269658842407, "loss": 0.1479, "step": 7624 }, { "epoch": 0.49, "learning_rate": 0.0001084420041610454, "loss": 0.1409, "step": 7625 }, { "epoch": 0.49, "learning_rate": 0.00010842131136959894, "loss": 0.1288, "step": 7626 }, { "epoch": 0.49, "learning_rate": 0.00010840061821497709, "loss": 0.1399, "step": 7627 }, { "epoch": 0.49, "learning_rate": 0.00010837992469807218, "loss": 0.127, "step": 7628 }, { "epoch": 0.49, "learning_rate": 0.00010835923081977673, "loss": 0.1628, "step": 7629 }, { "epoch": 0.49, "learning_rate": 0.00010833853658098312, "loss": 0.1343, "step": 7630 }, { "epoch": 0.49, "learning_rate": 0.00010831784198258383, "loss": 0.1601, "step": 7631 }, { "epoch": 0.49, "learning_rate": 0.00010829714702547132, "loss": 0.132, "step": 7632 }, { "epoch": 0.49, "learning_rate": 0.0001082764517105381, "loss": 0.1539, "step": 7633 }, { "epoch": 0.49, "learning_rate": 0.00010825575603867666, "loss": 0.1659, "step": 7634 }, { "epoch": 0.49, "learning_rate": 0.00010823506001077945, "loss": 0.1496, "step": 7635 }, { "epoch": 0.49, "learning_rate": 0.00010821436362773912, "loss": 0.1329, "step": 7636 }, { "epoch": 0.49, "learning_rate": 0.00010819366689044812, "loss": 0.1531, "step": 7637 }, { "epoch": 0.49, "learning_rate": 0.00010817296979979905, "loss": 0.1352, "step": 7638 }, { "epoch": 0.49, "learning_rate": 0.00010815227235668448, "loss": 0.1531, "step": 7639 }, { "epoch": 0.49, "learning_rate": 0.00010813157456199705, "loss": 0.1592, "step": 7640 }, { "epoch": 0.49, "learning_rate": 0.00010811087641662931, "loss": 0.1356, "step": 7641 }, { "epoch": 0.49, "learning_rate": 0.00010809017792147386, "loss": 0.1498, "step": 7642 }, { "epoch": 0.49, "learning_rate": 0.00010806947907742341, "loss": 0.1687, "step": 7643 }, { "epoch": 0.49, "learning_rate": 0.00010804877988537059, "loss": 0.1502, "step": 7644 }, { "epoch": 0.49, "learning_rate": 0.00010802808034620804, "loss": 0.1427, "step": 7645 }, { "epoch": 0.49, "learning_rate": 0.00010800738046082848, "loss": 0.1346, "step": 7646 }, { "epoch": 0.49, "learning_rate": 0.0001079866802301246, "loss": 0.1447, "step": 7647 }, { "epoch": 0.49, "learning_rate": 0.00010796597965498908, "loss": 0.1713, "step": 7648 }, { "epoch": 0.49, "learning_rate": 0.00010794527873631468, "loss": 0.1543, "step": 7649 }, { "epoch": 0.49, "learning_rate": 0.00010792457747499414, "loss": 0.1345, "step": 7650 }, { "epoch": 0.49, "learning_rate": 0.00010790387587192021, "loss": 0.1533, "step": 7651 }, { "epoch": 0.49, "learning_rate": 0.00010788317392798563, "loss": 0.1535, "step": 7652 }, { "epoch": 0.49, "learning_rate": 0.00010786247164408328, "loss": 0.1366, "step": 7653 }, { "epoch": 0.49, "learning_rate": 0.0001078417690211059, "loss": 0.1445, "step": 7654 }, { "epoch": 0.49, "learning_rate": 0.00010782106605994626, "loss": 0.129, "step": 7655 }, { "epoch": 0.49, "learning_rate": 0.00010780036276149728, "loss": 0.1613, "step": 7656 }, { "epoch": 0.49, "learning_rate": 0.00010777965912665175, "loss": 0.1656, "step": 7657 }, { "epoch": 0.49, "learning_rate": 0.00010775895515630257, "loss": 0.1567, "step": 7658 }, { "epoch": 0.49, "learning_rate": 0.00010773825085134255, "loss": 0.1363, "step": 7659 }, { "epoch": 0.49, "learning_rate": 0.00010771754621266466, "loss": 0.1937, "step": 7660 }, { "epoch": 0.49, "learning_rate": 0.00010769684124116175, "loss": 0.1468, "step": 7661 }, { "epoch": 0.49, "learning_rate": 0.00010767613593772672, "loss": 0.1447, "step": 7662 }, { "epoch": 0.49, "learning_rate": 0.00010765543030325256, "loss": 0.1391, "step": 7663 }, { "epoch": 0.49, "learning_rate": 0.00010763472433863217, "loss": 0.1728, "step": 7664 }, { "epoch": 0.49, "learning_rate": 0.00010761401804475855, "loss": 0.1355, "step": 7665 }, { "epoch": 0.49, "learning_rate": 0.00010759331142252462, "loss": 0.1372, "step": 7666 }, { "epoch": 0.49, "learning_rate": 0.00010757260447282345, "loss": 0.1588, "step": 7667 }, { "epoch": 0.49, "learning_rate": 0.00010755189719654795, "loss": 0.1413, "step": 7668 }, { "epoch": 0.49, "learning_rate": 0.0001075311895945912, "loss": 0.1429, "step": 7669 }, { "epoch": 0.49, "learning_rate": 0.00010751048166784623, "loss": 0.138, "step": 7670 }, { "epoch": 0.49, "learning_rate": 0.00010748977341720605, "loss": 0.145, "step": 7671 }, { "epoch": 0.49, "learning_rate": 0.00010746906484356372, "loss": 0.1683, "step": 7672 }, { "epoch": 0.49, "learning_rate": 0.00010744835594781233, "loss": 0.17, "step": 7673 }, { "epoch": 0.49, "learning_rate": 0.000107427646730845, "loss": 0.1367, "step": 7674 }, { "epoch": 0.49, "learning_rate": 0.00010740693719355477, "loss": 0.1768, "step": 7675 }, { "epoch": 0.49, "learning_rate": 0.00010738622733683479, "loss": 0.1324, "step": 7676 }, { "epoch": 0.49, "learning_rate": 0.00010736551716157817, "loss": 0.1536, "step": 7677 }, { "epoch": 0.49, "learning_rate": 0.00010734480666867808, "loss": 0.1444, "step": 7678 }, { "epoch": 0.49, "learning_rate": 0.0001073240958590276, "loss": 0.1598, "step": 7679 }, { "epoch": 0.49, "learning_rate": 0.00010730338473352001, "loss": 0.1646, "step": 7680 }, { "epoch": 0.49, "learning_rate": 0.00010728267329304846, "loss": 0.149, "step": 7681 }, { "epoch": 0.49, "learning_rate": 0.00010726196153850607, "loss": 0.1616, "step": 7682 }, { "epoch": 0.49, "learning_rate": 0.00010724124947078614, "loss": 0.143, "step": 7683 }, { "epoch": 0.49, "learning_rate": 0.00010722053709078184, "loss": 0.1287, "step": 7684 }, { "epoch": 0.49, "learning_rate": 0.00010719982439938643, "loss": 0.1634, "step": 7685 }, { "epoch": 0.49, "learning_rate": 0.00010717911139749314, "loss": 0.1427, "step": 7686 }, { "epoch": 0.49, "learning_rate": 0.00010715839808599529, "loss": 0.1499, "step": 7687 }, { "epoch": 0.49, "learning_rate": 0.00010713768446578612, "loss": 0.1417, "step": 7688 }, { "epoch": 0.49, "learning_rate": 0.00010711697053775887, "loss": 0.1686, "step": 7689 }, { "epoch": 0.49, "learning_rate": 0.0001070962563028069, "loss": 0.1648, "step": 7690 }, { "epoch": 0.49, "learning_rate": 0.00010707554176182355, "loss": 0.1394, "step": 7691 }, { "epoch": 0.49, "learning_rate": 0.00010705482691570207, "loss": 0.1574, "step": 7692 }, { "epoch": 0.49, "learning_rate": 0.00010703411176533588, "loss": 0.161, "step": 7693 }, { "epoch": 0.49, "learning_rate": 0.00010701339631161833, "loss": 0.1464, "step": 7694 }, { "epoch": 0.49, "learning_rate": 0.00010699268055544276, "loss": 0.1458, "step": 7695 }, { "epoch": 0.49, "learning_rate": 0.00010697196449770255, "loss": 0.152, "step": 7696 }, { "epoch": 0.49, "learning_rate": 0.00010695124813929111, "loss": 0.1421, "step": 7697 }, { "epoch": 0.49, "learning_rate": 0.00010693053148110185, "loss": 0.1344, "step": 7698 }, { "epoch": 0.49, "learning_rate": 0.00010690981452402817, "loss": 0.121, "step": 7699 }, { "epoch": 0.49, "learning_rate": 0.00010688909726896353, "loss": 0.1766, "step": 7700 }, { "epoch": 0.49, "learning_rate": 0.0001068683797168014, "loss": 0.146, "step": 7701 }, { "epoch": 0.49, "learning_rate": 0.00010684766186843516, "loss": 0.1597, "step": 7702 }, { "epoch": 0.49, "learning_rate": 0.00010682694372475834, "loss": 0.1534, "step": 7703 }, { "epoch": 0.49, "learning_rate": 0.00010680622528666444, "loss": 0.168, "step": 7704 }, { "epoch": 0.49, "learning_rate": 0.00010678550655504693, "loss": 0.1415, "step": 7705 }, { "epoch": 0.49, "learning_rate": 0.0001067647875307993, "loss": 0.1636, "step": 7706 }, { "epoch": 0.49, "learning_rate": 0.00010674406821481512, "loss": 0.1442, "step": 7707 }, { "epoch": 0.49, "learning_rate": 0.00010672334860798794, "loss": 0.1548, "step": 7708 }, { "epoch": 0.49, "learning_rate": 0.00010670262871121122, "loss": 0.156, "step": 7709 }, { "epoch": 0.49, "learning_rate": 0.00010668190852537858, "loss": 0.1426, "step": 7710 }, { "epoch": 0.49, "learning_rate": 0.00010666118805138362, "loss": 0.1504, "step": 7711 }, { "epoch": 0.49, "learning_rate": 0.00010664046729011987, "loss": 0.1436, "step": 7712 }, { "epoch": 0.49, "learning_rate": 0.00010661974624248097, "loss": 0.1517, "step": 7713 }, { "epoch": 0.49, "learning_rate": 0.00010659902490936052, "loss": 0.134, "step": 7714 }, { "epoch": 0.49, "learning_rate": 0.00010657830329165211, "loss": 0.1516, "step": 7715 }, { "epoch": 0.49, "learning_rate": 0.00010655758139024942, "loss": 0.1531, "step": 7716 }, { "epoch": 0.49, "learning_rate": 0.00010653685920604608, "loss": 0.1461, "step": 7717 }, { "epoch": 0.49, "learning_rate": 0.00010651613673993576, "loss": 0.1452, "step": 7718 }, { "epoch": 0.49, "learning_rate": 0.00010649541399281211, "loss": 0.1537, "step": 7719 }, { "epoch": 0.49, "learning_rate": 0.00010647469096556883, "loss": 0.1627, "step": 7720 }, { "epoch": 0.5, "learning_rate": 0.00010645396765909961, "loss": 0.1297, "step": 7721 }, { "epoch": 0.5, "learning_rate": 0.00010643324407429818, "loss": 0.1703, "step": 7722 }, { "epoch": 0.5, "learning_rate": 0.00010641252021205823, "loss": 0.1475, "step": 7723 }, { "epoch": 0.5, "learning_rate": 0.0001063917960732735, "loss": 0.1718, "step": 7724 }, { "epoch": 0.5, "learning_rate": 0.00010637107165883775, "loss": 0.1243, "step": 7725 }, { "epoch": 0.5, "learning_rate": 0.0001063503469696447, "loss": 0.1385, "step": 7726 }, { "epoch": 0.5, "learning_rate": 0.00010632962200658815, "loss": 0.1527, "step": 7727 }, { "epoch": 0.5, "learning_rate": 0.00010630889677056189, "loss": 0.1294, "step": 7728 }, { "epoch": 0.5, "learning_rate": 0.00010628817126245967, "loss": 0.1398, "step": 7729 }, { "epoch": 0.5, "learning_rate": 0.00010626744548317534, "loss": 0.1252, "step": 7730 }, { "epoch": 0.5, "learning_rate": 0.00010624671943360267, "loss": 0.1533, "step": 7731 }, { "epoch": 0.5, "learning_rate": 0.0001062259931146355, "loss": 0.1516, "step": 7732 }, { "epoch": 0.5, "learning_rate": 0.00010620526652716769, "loss": 0.1398, "step": 7733 }, { "epoch": 0.5, "learning_rate": 0.00010618453967209309, "loss": 0.1483, "step": 7734 }, { "epoch": 0.5, "learning_rate": 0.00010616381255030554, "loss": 0.1403, "step": 7735 }, { "epoch": 0.5, "learning_rate": 0.0001061430851626989, "loss": 0.1226, "step": 7736 }, { "epoch": 0.5, "learning_rate": 0.0001061223575101671, "loss": 0.147, "step": 7737 }, { "epoch": 0.5, "learning_rate": 0.00010610162959360401, "loss": 0.1339, "step": 7738 }, { "epoch": 0.5, "learning_rate": 0.0001060809014139035, "loss": 0.1521, "step": 7739 }, { "epoch": 0.5, "learning_rate": 0.00010606017297195956, "loss": 0.1339, "step": 7740 }, { "epoch": 0.5, "learning_rate": 0.00010603944426866608, "loss": 0.1451, "step": 7741 }, { "epoch": 0.5, "learning_rate": 0.00010601871530491698, "loss": 0.1354, "step": 7742 }, { "epoch": 0.5, "learning_rate": 0.00010599798608160627, "loss": 0.1504, "step": 7743 }, { "epoch": 0.5, "learning_rate": 0.00010597725659962788, "loss": 0.1336, "step": 7744 }, { "epoch": 0.5, "learning_rate": 0.00010595652685987575, "loss": 0.1444, "step": 7745 }, { "epoch": 0.5, "learning_rate": 0.00010593579686324394, "loss": 0.1663, "step": 7746 }, { "epoch": 0.5, "learning_rate": 0.00010591506661062638, "loss": 0.1723, "step": 7747 }, { "epoch": 0.5, "learning_rate": 0.00010589433610291713, "loss": 0.1514, "step": 7748 }, { "epoch": 0.5, "learning_rate": 0.00010587360534101017, "loss": 0.1506, "step": 7749 }, { "epoch": 0.5, "learning_rate": 0.00010585287432579956, "loss": 0.1433, "step": 7750 }, { "epoch": 0.5, "learning_rate": 0.0001058321430581793, "loss": 0.1513, "step": 7751 }, { "epoch": 0.5, "learning_rate": 0.00010581141153904348, "loss": 0.1786, "step": 7752 }, { "epoch": 0.5, "learning_rate": 0.00010579067976928614, "loss": 0.1584, "step": 7753 }, { "epoch": 0.5, "learning_rate": 0.00010576994774980138, "loss": 0.1522, "step": 7754 }, { "epoch": 0.5, "learning_rate": 0.00010574921548148326, "loss": 0.1533, "step": 7755 }, { "epoch": 0.5, "learning_rate": 0.00010572848296522588, "loss": 0.139, "step": 7756 }, { "epoch": 0.5, "learning_rate": 0.00010570775020192337, "loss": 0.152, "step": 7757 }, { "epoch": 0.5, "learning_rate": 0.00010568701719246978, "loss": 0.1544, "step": 7758 }, { "epoch": 0.5, "learning_rate": 0.00010566628393775928, "loss": 0.1379, "step": 7759 }, { "epoch": 0.5, "learning_rate": 0.00010564555043868604, "loss": 0.1486, "step": 7760 }, { "epoch": 0.5, "learning_rate": 0.00010562481669614417, "loss": 0.1335, "step": 7761 }, { "epoch": 0.5, "learning_rate": 0.00010560408271102782, "loss": 0.143, "step": 7762 }, { "epoch": 0.5, "learning_rate": 0.00010558334848423117, "loss": 0.1323, "step": 7763 }, { "epoch": 0.5, "learning_rate": 0.00010556261401664842, "loss": 0.1604, "step": 7764 }, { "epoch": 0.5, "learning_rate": 0.00010554187930917372, "loss": 0.1641, "step": 7765 }, { "epoch": 0.5, "learning_rate": 0.00010552114436270129, "loss": 0.1549, "step": 7766 }, { "epoch": 0.5, "learning_rate": 0.00010550040917812533, "loss": 0.157, "step": 7767 }, { "epoch": 0.5, "learning_rate": 0.00010547967375634009, "loss": 0.139, "step": 7768 }, { "epoch": 0.5, "learning_rate": 0.00010545893809823979, "loss": 0.1518, "step": 7769 }, { "epoch": 0.5, "learning_rate": 0.00010543820220471866, "loss": 0.1508, "step": 7770 }, { "epoch": 0.5, "learning_rate": 0.00010541746607667096, "loss": 0.1393, "step": 7771 }, { "epoch": 0.5, "learning_rate": 0.00010539672971499093, "loss": 0.1678, "step": 7772 }, { "epoch": 0.5, "learning_rate": 0.00010537599312057283, "loss": 0.1608, "step": 7773 }, { "epoch": 0.5, "learning_rate": 0.00010535525629431099, "loss": 0.1378, "step": 7774 }, { "epoch": 0.5, "learning_rate": 0.00010533451923709969, "loss": 0.1443, "step": 7775 }, { "epoch": 0.5, "learning_rate": 0.00010531378194983321, "loss": 0.1566, "step": 7776 }, { "epoch": 0.5, "learning_rate": 0.00010529304443340588, "loss": 0.1595, "step": 7777 }, { "epoch": 0.5, "learning_rate": 0.000105272306688712, "loss": 0.1605, "step": 7778 }, { "epoch": 0.5, "learning_rate": 0.00010525156871664589, "loss": 0.1362, "step": 7779 }, { "epoch": 0.5, "learning_rate": 0.00010523083051810197, "loss": 0.1387, "step": 7780 }, { "epoch": 0.5, "learning_rate": 0.0001052100920939745, "loss": 0.1543, "step": 7781 }, { "epoch": 0.5, "learning_rate": 0.00010518935344515788, "loss": 0.1511, "step": 7782 }, { "epoch": 0.5, "learning_rate": 0.00010516861457254648, "loss": 0.1613, "step": 7783 }, { "epoch": 0.5, "learning_rate": 0.00010514787547703466, "loss": 0.1473, "step": 7784 }, { "epoch": 0.5, "learning_rate": 0.00010512713615951683, "loss": 0.117, "step": 7785 }, { "epoch": 0.5, "learning_rate": 0.00010510639662088737, "loss": 0.1647, "step": 7786 }, { "epoch": 0.5, "learning_rate": 0.00010508565686204073, "loss": 0.137, "step": 7787 }, { "epoch": 0.5, "learning_rate": 0.00010506491688387127, "loss": 0.1426, "step": 7788 }, { "epoch": 0.5, "learning_rate": 0.00010504417668727346, "loss": 0.1505, "step": 7789 }, { "epoch": 0.5, "learning_rate": 0.00010502343627314173, "loss": 0.1355, "step": 7790 }, { "epoch": 0.5, "learning_rate": 0.00010500269564237051, "loss": 0.1641, "step": 7791 }, { "epoch": 0.5, "learning_rate": 0.00010498195479585427, "loss": 0.1324, "step": 7792 }, { "epoch": 0.5, "learning_rate": 0.00010496121373448741, "loss": 0.1328, "step": 7793 }, { "epoch": 0.5, "learning_rate": 0.00010494047245916452, "loss": 0.1271, "step": 7794 }, { "epoch": 0.5, "learning_rate": 0.00010491973097078003, "loss": 0.1306, "step": 7795 }, { "epoch": 0.5, "learning_rate": 0.00010489898927022838, "loss": 0.1441, "step": 7796 }, { "epoch": 0.5, "learning_rate": 0.00010487824735840417, "loss": 0.1411, "step": 7797 }, { "epoch": 0.5, "learning_rate": 0.00010485750523620184, "loss": 0.1717, "step": 7798 }, { "epoch": 0.5, "learning_rate": 0.00010483676290451591, "loss": 0.1664, "step": 7799 }, { "epoch": 0.5, "learning_rate": 0.00010481602036424094, "loss": 0.1597, "step": 7800 }, { "epoch": 0.5, "learning_rate": 0.00010479527761627146, "loss": 0.1659, "step": 7801 }, { "epoch": 0.5, "learning_rate": 0.00010477453466150203, "loss": 0.1333, "step": 7802 }, { "epoch": 0.5, "learning_rate": 0.00010475379150082714, "loss": 0.1446, "step": 7803 }, { "epoch": 0.5, "learning_rate": 0.00010473304813514146, "loss": 0.1635, "step": 7804 }, { "epoch": 0.5, "learning_rate": 0.00010471230456533947, "loss": 0.1465, "step": 7805 }, { "epoch": 0.5, "learning_rate": 0.00010469156079231579, "loss": 0.1502, "step": 7806 }, { "epoch": 0.5, "learning_rate": 0.000104670816816965, "loss": 0.1436, "step": 7807 }, { "epoch": 0.5, "learning_rate": 0.00010465007264018172, "loss": 0.1497, "step": 7808 }, { "epoch": 0.5, "learning_rate": 0.00010462932826286056, "loss": 0.1532, "step": 7809 }, { "epoch": 0.5, "learning_rate": 0.00010460858368589608, "loss": 0.1528, "step": 7810 }, { "epoch": 0.5, "learning_rate": 0.000104587838910183, "loss": 0.1478, "step": 7811 }, { "epoch": 0.5, "learning_rate": 0.0001045670939366159, "loss": 0.1476, "step": 7812 }, { "epoch": 0.5, "learning_rate": 0.00010454634876608937, "loss": 0.149, "step": 7813 }, { "epoch": 0.5, "learning_rate": 0.00010452560339949815, "loss": 0.1526, "step": 7814 }, { "epoch": 0.5, "learning_rate": 0.0001045048578377369, "loss": 0.1513, "step": 7815 }, { "epoch": 0.5, "learning_rate": 0.00010448411208170021, "loss": 0.1543, "step": 7816 }, { "epoch": 0.5, "learning_rate": 0.00010446336613228282, "loss": 0.1549, "step": 7817 }, { "epoch": 0.5, "learning_rate": 0.00010444261999037939, "loss": 0.1432, "step": 7818 }, { "epoch": 0.5, "learning_rate": 0.00010442187365688458, "loss": 0.1633, "step": 7819 }, { "epoch": 0.5, "learning_rate": 0.00010440112713269319, "loss": 0.1344, "step": 7820 }, { "epoch": 0.5, "learning_rate": 0.00010438038041869984, "loss": 0.1402, "step": 7821 }, { "epoch": 0.5, "learning_rate": 0.00010435963351579929, "loss": 0.1496, "step": 7822 }, { "epoch": 0.5, "learning_rate": 0.00010433888642488623, "loss": 0.1336, "step": 7823 }, { "epoch": 0.5, "learning_rate": 0.00010431813914685544, "loss": 0.1485, "step": 7824 }, { "epoch": 0.5, "learning_rate": 0.00010429739168260164, "loss": 0.1425, "step": 7825 }, { "epoch": 0.5, "learning_rate": 0.00010427664403301956, "loss": 0.1445, "step": 7826 }, { "epoch": 0.5, "learning_rate": 0.00010425589619900398, "loss": 0.143, "step": 7827 }, { "epoch": 0.5, "learning_rate": 0.00010423514818144968, "loss": 0.144, "step": 7828 }, { "epoch": 0.5, "learning_rate": 0.0001042143999812514, "loss": 0.1441, "step": 7829 }, { "epoch": 0.5, "learning_rate": 0.00010419365159930392, "loss": 0.15, "step": 7830 }, { "epoch": 0.5, "learning_rate": 0.00010417290303650211, "loss": 0.1569, "step": 7831 }, { "epoch": 0.5, "learning_rate": 0.00010415215429374068, "loss": 0.1287, "step": 7832 }, { "epoch": 0.5, "learning_rate": 0.00010413140537191442, "loss": 0.1488, "step": 7833 }, { "epoch": 0.5, "learning_rate": 0.0001041106562719182, "loss": 0.1311, "step": 7834 }, { "epoch": 0.5, "learning_rate": 0.00010408990699464684, "loss": 0.1563, "step": 7835 }, { "epoch": 0.5, "learning_rate": 0.00010406915754099518, "loss": 0.1491, "step": 7836 }, { "epoch": 0.5, "learning_rate": 0.00010404840791185797, "loss": 0.1345, "step": 7837 }, { "epoch": 0.5, "learning_rate": 0.00010402765810813015, "loss": 0.1518, "step": 7838 }, { "epoch": 0.5, "learning_rate": 0.00010400690813070655, "loss": 0.1542, "step": 7839 }, { "epoch": 0.5, "learning_rate": 0.00010398615798048194, "loss": 0.1548, "step": 7840 }, { "epoch": 0.5, "learning_rate": 0.00010396540765835131, "loss": 0.1682, "step": 7841 }, { "epoch": 0.5, "learning_rate": 0.00010394465716520949, "loss": 0.1549, "step": 7842 }, { "epoch": 0.5, "learning_rate": 0.00010392390650195132, "loss": 0.1447, "step": 7843 }, { "epoch": 0.5, "learning_rate": 0.00010390315566947174, "loss": 0.1464, "step": 7844 }, { "epoch": 0.5, "learning_rate": 0.00010388240466866563, "loss": 0.1592, "step": 7845 }, { "epoch": 0.5, "learning_rate": 0.00010386165350042784, "loss": 0.1293, "step": 7846 }, { "epoch": 0.5, "learning_rate": 0.00010384090216565339, "loss": 0.1628, "step": 7847 }, { "epoch": 0.5, "learning_rate": 0.00010382015066523709, "loss": 0.1694, "step": 7848 }, { "epoch": 0.5, "learning_rate": 0.00010379939900007393, "loss": 0.1595, "step": 7849 }, { "epoch": 0.5, "learning_rate": 0.0001037786471710588, "loss": 0.1466, "step": 7850 }, { "epoch": 0.5, "learning_rate": 0.00010375789517908668, "loss": 0.1199, "step": 7851 }, { "epoch": 0.5, "learning_rate": 0.00010373714302505251, "loss": 0.1353, "step": 7852 }, { "epoch": 0.5, "learning_rate": 0.00010371639070985117, "loss": 0.1372, "step": 7853 }, { "epoch": 0.5, "learning_rate": 0.00010369563823437772, "loss": 0.1437, "step": 7854 }, { "epoch": 0.5, "learning_rate": 0.00010367488559952704, "loss": 0.1463, "step": 7855 }, { "epoch": 0.5, "learning_rate": 0.00010365413280619418, "loss": 0.1598, "step": 7856 }, { "epoch": 0.5, "learning_rate": 0.00010363337985527405, "loss": 0.1501, "step": 7857 }, { "epoch": 0.5, "learning_rate": 0.00010361262674766173, "loss": 0.1508, "step": 7858 }, { "epoch": 0.5, "learning_rate": 0.0001035918734842521, "loss": 0.138, "step": 7859 }, { "epoch": 0.5, "learning_rate": 0.00010357112006594021, "loss": 0.1325, "step": 7860 }, { "epoch": 0.5, "learning_rate": 0.00010355036649362109, "loss": 0.1502, "step": 7861 }, { "epoch": 0.5, "learning_rate": 0.00010352961276818974, "loss": 0.159, "step": 7862 }, { "epoch": 0.5, "learning_rate": 0.00010350885889054117, "loss": 0.1552, "step": 7863 }, { "epoch": 0.5, "learning_rate": 0.00010348810486157039, "loss": 0.1532, "step": 7864 }, { "epoch": 0.5, "learning_rate": 0.0001034673506821725, "loss": 0.1526, "step": 7865 }, { "epoch": 0.5, "learning_rate": 0.00010344659635324245, "loss": 0.165, "step": 7866 }, { "epoch": 0.5, "learning_rate": 0.00010342584187567534, "loss": 0.1571, "step": 7867 }, { "epoch": 0.5, "learning_rate": 0.00010340508725036622, "loss": 0.1553, "step": 7868 }, { "epoch": 0.5, "learning_rate": 0.00010338433247821015, "loss": 0.1558, "step": 7869 }, { "epoch": 0.5, "learning_rate": 0.00010336357756010215, "loss": 0.1505, "step": 7870 }, { "epoch": 0.5, "learning_rate": 0.00010334282249693737, "loss": 0.134, "step": 7871 }, { "epoch": 0.5, "learning_rate": 0.00010332206728961085, "loss": 0.1443, "step": 7872 }, { "epoch": 0.5, "learning_rate": 0.00010330131193901762, "loss": 0.1366, "step": 7873 }, { "epoch": 0.5, "learning_rate": 0.00010328055644605285, "loss": 0.1663, "step": 7874 }, { "epoch": 0.5, "learning_rate": 0.00010325980081161161, "loss": 0.1408, "step": 7875 }, { "epoch": 0.51, "learning_rate": 0.000103239045036589, "loss": 0.155, "step": 7876 }, { "epoch": 0.51, "learning_rate": 0.0001032182891218801, "loss": 0.1374, "step": 7877 }, { "epoch": 0.51, "learning_rate": 0.00010319753306838007, "loss": 0.1619, "step": 7878 }, { "epoch": 0.51, "learning_rate": 0.00010317677687698404, "loss": 0.1594, "step": 7879 }, { "epoch": 0.51, "learning_rate": 0.00010315602054858706, "loss": 0.1609, "step": 7880 }, { "epoch": 0.51, "learning_rate": 0.00010313526408408436, "loss": 0.1632, "step": 7881 }, { "epoch": 0.51, "learning_rate": 0.00010311450748437099, "loss": 0.1505, "step": 7882 }, { "epoch": 0.51, "learning_rate": 0.00010309375075034214, "loss": 0.1523, "step": 7883 }, { "epoch": 0.51, "learning_rate": 0.00010307299388289296, "loss": 0.1433, "step": 7884 }, { "epoch": 0.51, "learning_rate": 0.00010305223688291858, "loss": 0.1391, "step": 7885 }, { "epoch": 0.51, "learning_rate": 0.00010303147975131422, "loss": 0.1291, "step": 7886 }, { "epoch": 0.51, "learning_rate": 0.00010301072248897498, "loss": 0.1349, "step": 7887 }, { "epoch": 0.51, "learning_rate": 0.00010298996509679606, "loss": 0.1389, "step": 7888 }, { "epoch": 0.51, "learning_rate": 0.00010296920757567266, "loss": 0.1558, "step": 7889 }, { "epoch": 0.51, "learning_rate": 0.00010294844992649988, "loss": 0.1562, "step": 7890 }, { "epoch": 0.51, "learning_rate": 0.00010292769215017302, "loss": 0.16, "step": 7891 }, { "epoch": 0.51, "learning_rate": 0.00010290693424758722, "loss": 0.1646, "step": 7892 }, { "epoch": 0.51, "learning_rate": 0.0001028861762196377, "loss": 0.1334, "step": 7893 }, { "epoch": 0.51, "learning_rate": 0.00010286541806721964, "loss": 0.143, "step": 7894 }, { "epoch": 0.51, "learning_rate": 0.00010284465979122824, "loss": 0.1541, "step": 7895 }, { "epoch": 0.51, "learning_rate": 0.00010282390139255878, "loss": 0.1502, "step": 7896 }, { "epoch": 0.51, "learning_rate": 0.00010280314287210637, "loss": 0.1541, "step": 7897 }, { "epoch": 0.51, "learning_rate": 0.00010278238423076636, "loss": 0.1659, "step": 7898 }, { "epoch": 0.51, "learning_rate": 0.00010276162546943395, "loss": 0.1529, "step": 7899 }, { "epoch": 0.51, "learning_rate": 0.0001027408665890043, "loss": 0.1577, "step": 7900 }, { "epoch": 0.51, "learning_rate": 0.00010272010759037276, "loss": 0.1622, "step": 7901 }, { "epoch": 0.51, "learning_rate": 0.0001026993484744345, "loss": 0.1569, "step": 7902 }, { "epoch": 0.51, "learning_rate": 0.00010267858924208478, "loss": 0.1267, "step": 7903 }, { "epoch": 0.51, "learning_rate": 0.0001026578298942189, "loss": 0.1399, "step": 7904 }, { "epoch": 0.51, "learning_rate": 0.0001026370704317321, "loss": 0.1444, "step": 7905 }, { "epoch": 0.51, "learning_rate": 0.00010261631085551964, "loss": 0.1309, "step": 7906 }, { "epoch": 0.51, "learning_rate": 0.00010259555116647681, "loss": 0.1455, "step": 7907 }, { "epoch": 0.51, "learning_rate": 0.00010257479136549889, "loss": 0.131, "step": 7908 }, { "epoch": 0.51, "learning_rate": 0.0001025540314534811, "loss": 0.1333, "step": 7909 }, { "epoch": 0.51, "learning_rate": 0.00010253327143131879, "loss": 0.1632, "step": 7910 }, { "epoch": 0.51, "learning_rate": 0.00010251251129990725, "loss": 0.1387, "step": 7911 }, { "epoch": 0.51, "learning_rate": 0.00010249175106014178, "loss": 0.1385, "step": 7912 }, { "epoch": 0.51, "learning_rate": 0.00010247099071291764, "loss": 0.15, "step": 7913 }, { "epoch": 0.51, "learning_rate": 0.00010245023025913016, "loss": 0.1465, "step": 7914 }, { "epoch": 0.51, "learning_rate": 0.00010242946969967467, "loss": 0.1318, "step": 7915 }, { "epoch": 0.51, "learning_rate": 0.00010240870903544644, "loss": 0.1431, "step": 7916 }, { "epoch": 0.51, "learning_rate": 0.00010238794826734081, "loss": 0.1525, "step": 7917 }, { "epoch": 0.51, "learning_rate": 0.00010236718739625313, "loss": 0.1639, "step": 7918 }, { "epoch": 0.51, "learning_rate": 0.0001023464264230787, "loss": 0.1577, "step": 7919 }, { "epoch": 0.51, "learning_rate": 0.00010232566534871286, "loss": 0.1476, "step": 7920 }, { "epoch": 0.51, "learning_rate": 0.00010230490417405094, "loss": 0.1315, "step": 7921 }, { "epoch": 0.51, "learning_rate": 0.0001022841428999883, "loss": 0.1403, "step": 7922 }, { "epoch": 0.51, "learning_rate": 0.00010226338152742026, "loss": 0.1365, "step": 7923 }, { "epoch": 0.51, "learning_rate": 0.00010224262005724217, "loss": 0.1473, "step": 7924 }, { "epoch": 0.51, "learning_rate": 0.0001022218584903494, "loss": 0.1847, "step": 7925 }, { "epoch": 0.51, "learning_rate": 0.00010220109682763732, "loss": 0.1605, "step": 7926 }, { "epoch": 0.51, "learning_rate": 0.00010218033507000129, "loss": 0.1363, "step": 7927 }, { "epoch": 0.51, "learning_rate": 0.00010215957321833664, "loss": 0.1305, "step": 7928 }, { "epoch": 0.51, "learning_rate": 0.00010213881127353878, "loss": 0.1517, "step": 7929 }, { "epoch": 0.51, "learning_rate": 0.00010211804923650303, "loss": 0.1267, "step": 7930 }, { "epoch": 0.51, "learning_rate": 0.00010209728710812485, "loss": 0.1652, "step": 7931 }, { "epoch": 0.51, "learning_rate": 0.00010207652488929958, "loss": 0.1734, "step": 7932 }, { "epoch": 0.51, "learning_rate": 0.00010205576258092257, "loss": 0.1466, "step": 7933 }, { "epoch": 0.51, "learning_rate": 0.00010203500018388929, "loss": 0.1451, "step": 7934 }, { "epoch": 0.51, "learning_rate": 0.00010201423769909506, "loss": 0.1389, "step": 7935 }, { "epoch": 0.51, "learning_rate": 0.00010199347512743533, "loss": 0.1494, "step": 7936 }, { "epoch": 0.51, "learning_rate": 0.00010197271246980545, "loss": 0.1301, "step": 7937 }, { "epoch": 0.51, "learning_rate": 0.00010195194972710086, "loss": 0.1559, "step": 7938 }, { "epoch": 0.51, "learning_rate": 0.00010193118690021699, "loss": 0.1405, "step": 7939 }, { "epoch": 0.51, "learning_rate": 0.00010191042399004922, "loss": 0.1338, "step": 7940 }, { "epoch": 0.51, "learning_rate": 0.00010188966099749297, "loss": 0.143, "step": 7941 }, { "epoch": 0.51, "learning_rate": 0.00010186889792344367, "loss": 0.1577, "step": 7942 }, { "epoch": 0.51, "learning_rate": 0.00010184813476879673, "loss": 0.1474, "step": 7943 }, { "epoch": 0.51, "learning_rate": 0.00010182737153444758, "loss": 0.1383, "step": 7944 }, { "epoch": 0.51, "learning_rate": 0.00010180660822129164, "loss": 0.1359, "step": 7945 }, { "epoch": 0.51, "learning_rate": 0.00010178584483022438, "loss": 0.1757, "step": 7946 }, { "epoch": 0.51, "learning_rate": 0.0001017650813621412, "loss": 0.1312, "step": 7947 }, { "epoch": 0.51, "learning_rate": 0.00010174431781793754, "loss": 0.1633, "step": 7948 }, { "epoch": 0.51, "learning_rate": 0.00010172355419850888, "loss": 0.1347, "step": 7949 }, { "epoch": 0.51, "learning_rate": 0.00010170279050475062, "loss": 0.1549, "step": 7950 }, { "epoch": 0.51, "learning_rate": 0.00010168202673755825, "loss": 0.16, "step": 7951 }, { "epoch": 0.51, "learning_rate": 0.00010166126289782721, "loss": 0.1468, "step": 7952 }, { "epoch": 0.51, "learning_rate": 0.00010164049898645296, "loss": 0.1533, "step": 7953 }, { "epoch": 0.51, "learning_rate": 0.00010161973500433093, "loss": 0.1398, "step": 7954 }, { "epoch": 0.51, "learning_rate": 0.00010159897095235661, "loss": 0.1439, "step": 7955 }, { "epoch": 0.51, "learning_rate": 0.00010157820683142549, "loss": 0.1292, "step": 7956 }, { "epoch": 0.51, "learning_rate": 0.00010155744264243296, "loss": 0.1404, "step": 7957 }, { "epoch": 0.51, "learning_rate": 0.00010153667838627456, "loss": 0.1484, "step": 7958 }, { "epoch": 0.51, "learning_rate": 0.00010151591406384578, "loss": 0.1564, "step": 7959 }, { "epoch": 0.51, "learning_rate": 0.00010149514967604202, "loss": 0.1732, "step": 7960 }, { "epoch": 0.51, "learning_rate": 0.0001014743852237588, "loss": 0.1714, "step": 7961 }, { "epoch": 0.51, "learning_rate": 0.00010145362070789164, "loss": 0.1336, "step": 7962 }, { "epoch": 0.51, "learning_rate": 0.00010143285612933598, "loss": 0.1405, "step": 7963 }, { "epoch": 0.51, "learning_rate": 0.00010141209148898727, "loss": 0.1576, "step": 7964 }, { "epoch": 0.51, "learning_rate": 0.00010139132678774108, "loss": 0.1547, "step": 7965 }, { "epoch": 0.51, "learning_rate": 0.00010137056202649289, "loss": 0.1616, "step": 7966 }, { "epoch": 0.51, "learning_rate": 0.00010134979720613813, "loss": 0.1397, "step": 7967 }, { "epoch": 0.51, "learning_rate": 0.0001013290323275724, "loss": 0.1353, "step": 7968 }, { "epoch": 0.51, "learning_rate": 0.00010130826739169113, "loss": 0.1533, "step": 7969 }, { "epoch": 0.51, "learning_rate": 0.00010128750239938983, "loss": 0.1452, "step": 7970 }, { "epoch": 0.51, "learning_rate": 0.00010126673735156402, "loss": 0.1486, "step": 7971 }, { "epoch": 0.51, "learning_rate": 0.00010124597224910921, "loss": 0.1426, "step": 7972 }, { "epoch": 0.51, "learning_rate": 0.00010122520709292089, "loss": 0.1412, "step": 7973 }, { "epoch": 0.51, "learning_rate": 0.00010120444188389462, "loss": 0.1546, "step": 7974 }, { "epoch": 0.51, "learning_rate": 0.00010118367662292587, "loss": 0.1596, "step": 7975 }, { "epoch": 0.51, "learning_rate": 0.00010116291131091016, "loss": 0.1498, "step": 7976 }, { "epoch": 0.51, "learning_rate": 0.00010114214594874303, "loss": 0.146, "step": 7977 }, { "epoch": 0.51, "learning_rate": 0.00010112138053731998, "loss": 0.152, "step": 7978 }, { "epoch": 0.51, "learning_rate": 0.00010110061507753657, "loss": 0.1468, "step": 7979 }, { "epoch": 0.51, "learning_rate": 0.00010107984957028829, "loss": 0.1561, "step": 7980 }, { "epoch": 0.51, "learning_rate": 0.00010105908401647071, "loss": 0.1497, "step": 7981 }, { "epoch": 0.51, "learning_rate": 0.00010103831841697931, "loss": 0.1645, "step": 7982 }, { "epoch": 0.51, "learning_rate": 0.00010101755277270963, "loss": 0.131, "step": 7983 }, { "epoch": 0.51, "learning_rate": 0.00010099678708455722, "loss": 0.1508, "step": 7984 }, { "epoch": 0.51, "learning_rate": 0.00010097602135341764, "loss": 0.1583, "step": 7985 }, { "epoch": 0.51, "learning_rate": 0.00010095525558018637, "loss": 0.1348, "step": 7986 }, { "epoch": 0.51, "learning_rate": 0.00010093448976575902, "loss": 0.1368, "step": 7987 }, { "epoch": 0.51, "learning_rate": 0.00010091372391103107, "loss": 0.1665, "step": 7988 }, { "epoch": 0.51, "learning_rate": 0.0001008929580168981, "loss": 0.1332, "step": 7989 }, { "epoch": 0.51, "learning_rate": 0.0001008721920842556, "loss": 0.131, "step": 7990 }, { "epoch": 0.51, "learning_rate": 0.0001008514261139992, "loss": 0.1631, "step": 7991 }, { "epoch": 0.51, "learning_rate": 0.0001008306601070244, "loss": 0.1519, "step": 7992 }, { "epoch": 0.51, "learning_rate": 0.00010080989406422674, "loss": 0.1591, "step": 7993 }, { "epoch": 0.51, "learning_rate": 0.00010078912798650177, "loss": 0.156, "step": 7994 }, { "epoch": 0.51, "learning_rate": 0.00010076836187474512, "loss": 0.1246, "step": 7995 }, { "epoch": 0.51, "learning_rate": 0.00010074759572985227, "loss": 0.141, "step": 7996 }, { "epoch": 0.51, "learning_rate": 0.00010072682955271875, "loss": 0.1386, "step": 7997 }, { "epoch": 0.51, "learning_rate": 0.00010070606334424018, "loss": 0.1342, "step": 7998 }, { "epoch": 0.51, "learning_rate": 0.00010068529710531209, "loss": 0.1323, "step": 7999 }, { "epoch": 0.51, "learning_rate": 0.00010066453083683006, "loss": 0.1229, "step": 8000 }, { "epoch": 0.51, "learning_rate": 0.0001006437645396896, "loss": 0.1297, "step": 8001 }, { "epoch": 0.51, "learning_rate": 0.00010062299821478636, "loss": 0.1543, "step": 8002 }, { "epoch": 0.51, "learning_rate": 0.0001006022318630158, "loss": 0.1466, "step": 8003 }, { "epoch": 0.51, "learning_rate": 0.00010058146548527357, "loss": 0.1455, "step": 8004 }, { "epoch": 0.51, "learning_rate": 0.00010056069908245517, "loss": 0.1736, "step": 8005 }, { "epoch": 0.51, "learning_rate": 0.00010053993265545622, "loss": 0.1581, "step": 8006 }, { "epoch": 0.51, "learning_rate": 0.00010051916620517228, "loss": 0.1388, "step": 8007 }, { "epoch": 0.51, "learning_rate": 0.00010049839973249888, "loss": 0.1423, "step": 8008 }, { "epoch": 0.51, "learning_rate": 0.00010047763323833164, "loss": 0.1431, "step": 8009 }, { "epoch": 0.51, "learning_rate": 0.00010045686672356612, "loss": 0.1574, "step": 8010 }, { "epoch": 0.51, "learning_rate": 0.00010043610018909782, "loss": 0.1536, "step": 8011 }, { "epoch": 0.51, "learning_rate": 0.0001004153336358224, "loss": 0.1298, "step": 8012 }, { "epoch": 0.51, "learning_rate": 0.00010039456706463539, "loss": 0.1457, "step": 8013 }, { "epoch": 0.51, "learning_rate": 0.0001003738004764324, "loss": 0.1389, "step": 8014 }, { "epoch": 0.51, "learning_rate": 0.000100353033872109, "loss": 0.1329, "step": 8015 }, { "epoch": 0.51, "learning_rate": 0.00010033226725256071, "loss": 0.1345, "step": 8016 }, { "epoch": 0.51, "learning_rate": 0.00010031150061868313, "loss": 0.1375, "step": 8017 }, { "epoch": 0.51, "learning_rate": 0.00010029073397137187, "loss": 0.1545, "step": 8018 }, { "epoch": 0.51, "learning_rate": 0.00010026996731152252, "loss": 0.1517, "step": 8019 }, { "epoch": 0.51, "learning_rate": 0.00010024920064003059, "loss": 0.1256, "step": 8020 }, { "epoch": 0.51, "learning_rate": 0.00010022843395779169, "loss": 0.1424, "step": 8021 }, { "epoch": 0.51, "learning_rate": 0.00010020766726570143, "loss": 0.1222, "step": 8022 }, { "epoch": 0.51, "learning_rate": 0.00010018690056465537, "loss": 0.1533, "step": 8023 }, { "epoch": 0.51, "learning_rate": 0.00010016613385554903, "loss": 0.128, "step": 8024 }, { "epoch": 0.51, "learning_rate": 0.00010014536713927807, "loss": 0.1448, "step": 8025 }, { "epoch": 0.51, "learning_rate": 0.00010012460041673806, "loss": 0.1245, "step": 8026 }, { "epoch": 0.51, "learning_rate": 0.00010010383368882455, "loss": 0.1617, "step": 8027 }, { "epoch": 0.51, "learning_rate": 0.0001000830669564331, "loss": 0.1237, "step": 8028 }, { "epoch": 0.51, "learning_rate": 0.00010006230022045939, "loss": 0.1459, "step": 8029 }, { "epoch": 0.51, "learning_rate": 0.0001000415334817989, "loss": 0.1449, "step": 8030 }, { "epoch": 0.51, "learning_rate": 0.00010002076674134723, "loss": 0.1425, "step": 8031 }, { "epoch": 0.52, "learning_rate": 0.0001, "loss": 0.1361, "step": 8032 }, { "epoch": 0.52, "learning_rate": 9.99792332586528e-05, "loss": 0.1398, "step": 8033 }, { "epoch": 0.52, "learning_rate": 9.995846651820111e-05, "loss": 0.143, "step": 8034 }, { "epoch": 0.52, "learning_rate": 9.993769977954064e-05, "loss": 0.1668, "step": 8035 }, { "epoch": 0.52, "learning_rate": 9.991693304356688e-05, "loss": 0.1342, "step": 8036 }, { "epoch": 0.52, "learning_rate": 9.989616631117548e-05, "loss": 0.1427, "step": 8037 }, { "epoch": 0.52, "learning_rate": 9.987539958326195e-05, "loss": 0.1408, "step": 8038 }, { "epoch": 0.52, "learning_rate": 9.985463286072194e-05, "loss": 0.1697, "step": 8039 }, { "epoch": 0.52, "learning_rate": 9.983386614445099e-05, "loss": 0.1485, "step": 8040 }, { "epoch": 0.52, "learning_rate": 9.981309943534466e-05, "loss": 0.1307, "step": 8041 }, { "epoch": 0.52, "learning_rate": 9.97923327342986e-05, "loss": 0.141, "step": 8042 }, { "epoch": 0.52, "learning_rate": 9.977156604220831e-05, "loss": 0.1367, "step": 8043 }, { "epoch": 0.52, "learning_rate": 9.975079935996942e-05, "loss": 0.1301, "step": 8044 }, { "epoch": 0.52, "learning_rate": 9.973003268847749e-05, "loss": 0.1275, "step": 8045 }, { "epoch": 0.52, "learning_rate": 9.970926602862814e-05, "loss": 0.144, "step": 8046 }, { "epoch": 0.52, "learning_rate": 9.96884993813169e-05, "loss": 0.1379, "step": 8047 }, { "epoch": 0.52, "learning_rate": 9.966773274743931e-05, "loss": 0.1809, "step": 8048 }, { "epoch": 0.52, "learning_rate": 9.964696612789106e-05, "loss": 0.1491, "step": 8049 }, { "epoch": 0.52, "learning_rate": 9.962619952356761e-05, "loss": 0.1497, "step": 8050 }, { "epoch": 0.52, "learning_rate": 9.960543293536462e-05, "loss": 0.1564, "step": 8051 }, { "epoch": 0.52, "learning_rate": 9.958466636417761e-05, "loss": 0.1526, "step": 8052 }, { "epoch": 0.52, "learning_rate": 9.95638998109022e-05, "loss": 0.141, "step": 8053 }, { "epoch": 0.52, "learning_rate": 9.954313327643395e-05, "loss": 0.1417, "step": 8054 }, { "epoch": 0.52, "learning_rate": 9.952236676166837e-05, "loss": 0.1425, "step": 8055 }, { "epoch": 0.52, "learning_rate": 9.950160026750115e-05, "loss": 0.176, "step": 8056 }, { "epoch": 0.52, "learning_rate": 9.948083379482774e-05, "loss": 0.1325, "step": 8057 }, { "epoch": 0.52, "learning_rate": 9.94600673445438e-05, "loss": 0.1709, "step": 8058 }, { "epoch": 0.52, "learning_rate": 9.943930091754482e-05, "loss": 0.1582, "step": 8059 }, { "epoch": 0.52, "learning_rate": 9.941853451472647e-05, "loss": 0.1433, "step": 8060 }, { "epoch": 0.52, "learning_rate": 9.939776813698422e-05, "loss": 0.1407, "step": 8061 }, { "epoch": 0.52, "learning_rate": 9.937700178521368e-05, "loss": 0.1433, "step": 8062 }, { "epoch": 0.52, "learning_rate": 9.935623546031043e-05, "loss": 0.146, "step": 8063 }, { "epoch": 0.52, "learning_rate": 9.933546916316997e-05, "loss": 0.1455, "step": 8064 }, { "epoch": 0.52, "learning_rate": 9.931470289468793e-05, "loss": 0.1387, "step": 8065 }, { "epoch": 0.52, "learning_rate": 9.929393665575982e-05, "loss": 0.1425, "step": 8066 }, { "epoch": 0.52, "learning_rate": 9.927317044728129e-05, "loss": 0.1429, "step": 8067 }, { "epoch": 0.52, "learning_rate": 9.925240427014779e-05, "loss": 0.1468, "step": 8068 }, { "epoch": 0.52, "learning_rate": 9.92316381252549e-05, "loss": 0.1554, "step": 8069 }, { "epoch": 0.52, "learning_rate": 9.921087201349824e-05, "loss": 0.151, "step": 8070 }, { "epoch": 0.52, "learning_rate": 9.919010593577329e-05, "loss": 0.1422, "step": 8071 }, { "epoch": 0.52, "learning_rate": 9.916933989297563e-05, "loss": 0.1464, "step": 8072 }, { "epoch": 0.52, "learning_rate": 9.914857388600085e-05, "loss": 0.1318, "step": 8073 }, { "epoch": 0.52, "learning_rate": 9.912780791574442e-05, "loss": 0.177, "step": 8074 }, { "epoch": 0.52, "learning_rate": 9.910704198310196e-05, "loss": 0.1531, "step": 8075 }, { "epoch": 0.52, "learning_rate": 9.908627608896895e-05, "loss": 0.1486, "step": 8076 }, { "epoch": 0.52, "learning_rate": 9.906551023424101e-05, "loss": 0.1359, "step": 8077 }, { "epoch": 0.52, "learning_rate": 9.904474441981362e-05, "loss": 0.1414, "step": 8078 }, { "epoch": 0.52, "learning_rate": 9.90239786465824e-05, "loss": 0.1317, "step": 8079 }, { "epoch": 0.52, "learning_rate": 9.90032129154428e-05, "loss": 0.1478, "step": 8080 }, { "epoch": 0.52, "learning_rate": 9.898244722729039e-05, "loss": 0.1309, "step": 8081 }, { "epoch": 0.52, "learning_rate": 9.896168158302075e-05, "loss": 0.159, "step": 8082 }, { "epoch": 0.52, "learning_rate": 9.894091598352933e-05, "loss": 0.1457, "step": 8083 }, { "epoch": 0.52, "learning_rate": 9.892015042971172e-05, "loss": 0.1615, "step": 8084 }, { "epoch": 0.52, "learning_rate": 9.889938492246344e-05, "loss": 0.1663, "step": 8085 }, { "epoch": 0.52, "learning_rate": 9.887861946268003e-05, "loss": 0.1543, "step": 8086 }, { "epoch": 0.52, "learning_rate": 9.885785405125701e-05, "loss": 0.1436, "step": 8087 }, { "epoch": 0.52, "learning_rate": 9.883708868908985e-05, "loss": 0.1454, "step": 8088 }, { "epoch": 0.52, "learning_rate": 9.881632337707418e-05, "loss": 0.1449, "step": 8089 }, { "epoch": 0.52, "learning_rate": 9.87955581161054e-05, "loss": 0.1565, "step": 8090 }, { "epoch": 0.52, "learning_rate": 9.877479290707913e-05, "loss": 0.1524, "step": 8091 }, { "epoch": 0.52, "learning_rate": 9.87540277508908e-05, "loss": 0.1365, "step": 8092 }, { "epoch": 0.52, "learning_rate": 9.8733262648436e-05, "loss": 0.1546, "step": 8093 }, { "epoch": 0.52, "learning_rate": 9.871249760061021e-05, "loss": 0.1542, "step": 8094 }, { "epoch": 0.52, "learning_rate": 9.869173260830889e-05, "loss": 0.1567, "step": 8095 }, { "epoch": 0.52, "learning_rate": 9.867096767242764e-05, "loss": 0.1621, "step": 8096 }, { "epoch": 0.52, "learning_rate": 9.865020279386186e-05, "loss": 0.1314, "step": 8097 }, { "epoch": 0.52, "learning_rate": 9.862943797350715e-05, "loss": 0.1318, "step": 8098 }, { "epoch": 0.52, "learning_rate": 9.860867321225892e-05, "loss": 0.134, "step": 8099 }, { "epoch": 0.52, "learning_rate": 9.858790851101275e-05, "loss": 0.1335, "step": 8100 }, { "epoch": 0.52, "learning_rate": 9.856714387066409e-05, "loss": 0.1597, "step": 8101 }, { "epoch": 0.52, "learning_rate": 9.85463792921084e-05, "loss": 0.1642, "step": 8102 }, { "epoch": 0.52, "learning_rate": 9.852561477624122e-05, "loss": 0.1415, "step": 8103 }, { "epoch": 0.52, "learning_rate": 9.850485032395799e-05, "loss": 0.1419, "step": 8104 }, { "epoch": 0.52, "learning_rate": 9.848408593615426e-05, "loss": 0.1591, "step": 8105 }, { "epoch": 0.52, "learning_rate": 9.846332161372542e-05, "loss": 0.1479, "step": 8106 }, { "epoch": 0.52, "learning_rate": 9.844255735756706e-05, "loss": 0.1445, "step": 8107 }, { "epoch": 0.52, "learning_rate": 9.842179316857456e-05, "loss": 0.1432, "step": 8108 }, { "epoch": 0.52, "learning_rate": 9.84010290476434e-05, "loss": 0.1526, "step": 8109 }, { "epoch": 0.52, "learning_rate": 9.83802649956691e-05, "loss": 0.1443, "step": 8110 }, { "epoch": 0.52, "learning_rate": 9.835950101354707e-05, "loss": 0.1326, "step": 8111 }, { "epoch": 0.52, "learning_rate": 9.83387371021728e-05, "loss": 0.1285, "step": 8112 }, { "epoch": 0.52, "learning_rate": 9.831797326244174e-05, "loss": 0.135, "step": 8113 }, { "epoch": 0.52, "learning_rate": 9.829720949524939e-05, "loss": 0.1262, "step": 8114 }, { "epoch": 0.52, "learning_rate": 9.827644580149116e-05, "loss": 0.1537, "step": 8115 }, { "epoch": 0.52, "learning_rate": 9.825568218206247e-05, "loss": 0.1458, "step": 8116 }, { "epoch": 0.52, "learning_rate": 9.823491863785885e-05, "loss": 0.1512, "step": 8117 }, { "epoch": 0.52, "learning_rate": 9.821415516977563e-05, "loss": 0.1389, "step": 8118 }, { "epoch": 0.52, "learning_rate": 9.819339177870839e-05, "loss": 0.1425, "step": 8119 }, { "epoch": 0.52, "learning_rate": 9.817262846555243e-05, "loss": 0.1641, "step": 8120 }, { "epoch": 0.52, "learning_rate": 9.815186523120328e-05, "loss": 0.1696, "step": 8121 }, { "epoch": 0.52, "learning_rate": 9.813110207655637e-05, "loss": 0.1457, "step": 8122 }, { "epoch": 0.52, "learning_rate": 9.811033900250705e-05, "loss": 0.1396, "step": 8123 }, { "epoch": 0.52, "learning_rate": 9.808957600995079e-05, "loss": 0.1386, "step": 8124 }, { "epoch": 0.52, "learning_rate": 9.8068813099783e-05, "loss": 0.1486, "step": 8125 }, { "epoch": 0.52, "learning_rate": 9.804805027289915e-05, "loss": 0.1664, "step": 8126 }, { "epoch": 0.52, "learning_rate": 9.802728753019455e-05, "loss": 0.1722, "step": 8127 }, { "epoch": 0.52, "learning_rate": 9.800652487256469e-05, "loss": 0.146, "step": 8128 }, { "epoch": 0.52, "learning_rate": 9.798576230090497e-05, "loss": 0.1491, "step": 8129 }, { "epoch": 0.52, "learning_rate": 9.796499981611073e-05, "loss": 0.1371, "step": 8130 }, { "epoch": 0.52, "learning_rate": 9.794423741907744e-05, "loss": 0.1426, "step": 8131 }, { "epoch": 0.52, "learning_rate": 9.792347511070043e-05, "loss": 0.1566, "step": 8132 }, { "epoch": 0.52, "learning_rate": 9.790271289187518e-05, "loss": 0.1371, "step": 8133 }, { "epoch": 0.52, "learning_rate": 9.788195076349696e-05, "loss": 0.1413, "step": 8134 }, { "epoch": 0.52, "learning_rate": 9.786118872646125e-05, "loss": 0.139, "step": 8135 }, { "epoch": 0.52, "learning_rate": 9.784042678166341e-05, "loss": 0.1407, "step": 8136 }, { "epoch": 0.52, "learning_rate": 9.781966492999875e-05, "loss": 0.1337, "step": 8137 }, { "epoch": 0.52, "learning_rate": 9.77989031723627e-05, "loss": 0.1382, "step": 8138 }, { "epoch": 0.52, "learning_rate": 9.77781415096506e-05, "loss": 0.1532, "step": 8139 }, { "epoch": 0.52, "learning_rate": 9.775737994275787e-05, "loss": 0.1535, "step": 8140 }, { "epoch": 0.52, "learning_rate": 9.773661847257976e-05, "loss": 0.1644, "step": 8141 }, { "epoch": 0.52, "learning_rate": 9.771585710001173e-05, "loss": 0.1279, "step": 8142 }, { "epoch": 0.52, "learning_rate": 9.76950958259491e-05, "loss": 0.13, "step": 8143 }, { "epoch": 0.52, "learning_rate": 9.767433465128716e-05, "loss": 0.1465, "step": 8144 }, { "epoch": 0.52, "learning_rate": 9.765357357692133e-05, "loss": 0.1463, "step": 8145 }, { "epoch": 0.52, "learning_rate": 9.763281260374688e-05, "loss": 0.1365, "step": 8146 }, { "epoch": 0.52, "learning_rate": 9.761205173265921e-05, "loss": 0.1698, "step": 8147 }, { "epoch": 0.52, "learning_rate": 9.759129096455358e-05, "loss": 0.1439, "step": 8148 }, { "epoch": 0.52, "learning_rate": 9.757053030032536e-05, "loss": 0.1455, "step": 8149 }, { "epoch": 0.52, "learning_rate": 9.754976974086988e-05, "loss": 0.1529, "step": 8150 }, { "epoch": 0.52, "learning_rate": 9.752900928708238e-05, "loss": 0.1667, "step": 8151 }, { "epoch": 0.52, "learning_rate": 9.750824893985826e-05, "loss": 0.1418, "step": 8152 }, { "epoch": 0.52, "learning_rate": 9.748748870009275e-05, "loss": 0.1527, "step": 8153 }, { "epoch": 0.52, "learning_rate": 9.746672856868123e-05, "loss": 0.1606, "step": 8154 }, { "epoch": 0.52, "learning_rate": 9.744596854651893e-05, "loss": 0.1572, "step": 8155 }, { "epoch": 0.52, "learning_rate": 9.742520863450115e-05, "loss": 0.1369, "step": 8156 }, { "epoch": 0.52, "learning_rate": 9.740444883352324e-05, "loss": 0.149, "step": 8157 }, { "epoch": 0.52, "learning_rate": 9.738368914448037e-05, "loss": 0.1602, "step": 8158 }, { "epoch": 0.52, "learning_rate": 9.736292956826792e-05, "loss": 0.1596, "step": 8159 }, { "epoch": 0.52, "learning_rate": 9.734217010578111e-05, "loss": 0.1438, "step": 8160 }, { "epoch": 0.52, "learning_rate": 9.732141075791524e-05, "loss": 0.129, "step": 8161 }, { "epoch": 0.52, "learning_rate": 9.730065152556555e-05, "loss": 0.1238, "step": 8162 }, { "epoch": 0.52, "learning_rate": 9.727989240962728e-05, "loss": 0.1509, "step": 8163 }, { "epoch": 0.52, "learning_rate": 9.725913341099571e-05, "loss": 0.141, "step": 8164 }, { "epoch": 0.52, "learning_rate": 9.723837453056607e-05, "loss": 0.1549, "step": 8165 }, { "epoch": 0.52, "learning_rate": 9.721761576923366e-05, "loss": 0.1435, "step": 8166 }, { "epoch": 0.52, "learning_rate": 9.719685712789361e-05, "loss": 0.1564, "step": 8167 }, { "epoch": 0.52, "learning_rate": 9.717609860744126e-05, "loss": 0.1438, "step": 8168 }, { "epoch": 0.52, "learning_rate": 9.71553402087718e-05, "loss": 0.1539, "step": 8169 }, { "epoch": 0.52, "learning_rate": 9.71345819327804e-05, "loss": 0.1332, "step": 8170 }, { "epoch": 0.52, "learning_rate": 9.711382378036233e-05, "loss": 0.1468, "step": 8171 }, { "epoch": 0.52, "learning_rate": 9.709306575241277e-05, "loss": 0.1431, "step": 8172 }, { "epoch": 0.52, "learning_rate": 9.707230784982699e-05, "loss": 0.1408, "step": 8173 }, { "epoch": 0.52, "learning_rate": 9.705155007350011e-05, "loss": 0.1462, "step": 8174 }, { "epoch": 0.52, "learning_rate": 9.703079242432738e-05, "loss": 0.1498, "step": 8175 }, { "epoch": 0.52, "learning_rate": 9.701003490320398e-05, "loss": 0.134, "step": 8176 }, { "epoch": 0.52, "learning_rate": 9.698927751102504e-05, "loss": 0.1555, "step": 8177 }, { "epoch": 0.52, "learning_rate": 9.696852024868582e-05, "loss": 0.1458, "step": 8178 }, { "epoch": 0.52, "learning_rate": 9.69477631170814e-05, "loss": 0.1383, "step": 8179 }, { "epoch": 0.52, "learning_rate": 9.692700611710707e-05, "loss": 0.1362, "step": 8180 }, { "epoch": 0.52, "learning_rate": 9.690624924965787e-05, "loss": 0.1406, "step": 8181 }, { "epoch": 0.52, "learning_rate": 9.688549251562903e-05, "loss": 0.1497, "step": 8182 }, { "epoch": 0.52, "learning_rate": 9.686473591591569e-05, "loss": 0.1701, "step": 8183 }, { "epoch": 0.52, "learning_rate": 9.684397945141295e-05, "loss": 0.1596, "step": 8184 }, { "epoch": 0.52, "learning_rate": 9.6823223123016e-05, "loss": 0.1376, "step": 8185 }, { "epoch": 0.52, "learning_rate": 9.680246693161991e-05, "loss": 0.1438, "step": 8186 }, { "epoch": 0.52, "learning_rate": 9.678171087811991e-05, "loss": 0.1486, "step": 8187 }, { "epoch": 0.53, "learning_rate": 9.676095496341102e-05, "loss": 0.1636, "step": 8188 }, { "epoch": 0.53, "learning_rate": 9.67401991883884e-05, "loss": 0.145, "step": 8189 }, { "epoch": 0.53, "learning_rate": 9.671944355394719e-05, "loss": 0.1378, "step": 8190 }, { "epoch": 0.53, "learning_rate": 9.66986880609824e-05, "loss": 0.1424, "step": 8191 }, { "epoch": 0.53, "learning_rate": 9.66779327103892e-05, "loss": 0.1534, "step": 8192 }, { "epoch": 0.53, "learning_rate": 9.665717750306263e-05, "loss": 0.1536, "step": 8193 }, { "epoch": 0.53, "learning_rate": 9.663642243989787e-05, "loss": 0.1408, "step": 8194 }, { "epoch": 0.53, "learning_rate": 9.661566752178987e-05, "loss": 0.1648, "step": 8195 }, { "epoch": 0.53, "learning_rate": 9.659491274963379e-05, "loss": 0.1809, "step": 8196 }, { "epoch": 0.53, "learning_rate": 9.65741581243247e-05, "loss": 0.1634, "step": 8197 }, { "epoch": 0.53, "learning_rate": 9.655340364675757e-05, "loss": 0.1399, "step": 8198 }, { "epoch": 0.53, "learning_rate": 9.653264931782755e-05, "loss": 0.1412, "step": 8199 }, { "epoch": 0.53, "learning_rate": 9.65118951384296e-05, "loss": 0.1563, "step": 8200 }, { "epoch": 0.53, "learning_rate": 9.649114110945887e-05, "loss": 0.1581, "step": 8201 }, { "epoch": 0.53, "learning_rate": 9.647038723181026e-05, "loss": 0.1528, "step": 8202 }, { "epoch": 0.53, "learning_rate": 9.644963350637892e-05, "loss": 0.1747, "step": 8203 }, { "epoch": 0.53, "learning_rate": 9.642887993405984e-05, "loss": 0.1493, "step": 8204 }, { "epoch": 0.53, "learning_rate": 9.640812651574793e-05, "loss": 0.1548, "step": 8205 }, { "epoch": 0.53, "learning_rate": 9.638737325233834e-05, "loss": 0.1669, "step": 8206 }, { "epoch": 0.53, "learning_rate": 9.636662014472595e-05, "loss": 0.1414, "step": 8207 }, { "epoch": 0.53, "learning_rate": 9.634586719380583e-05, "loss": 0.1423, "step": 8208 }, { "epoch": 0.53, "learning_rate": 9.632511440047295e-05, "loss": 0.1562, "step": 8209 }, { "epoch": 0.53, "learning_rate": 9.630436176562232e-05, "loss": 0.1462, "step": 8210 }, { "epoch": 0.53, "learning_rate": 9.628360929014885e-05, "loss": 0.1618, "step": 8211 }, { "epoch": 0.53, "learning_rate": 9.626285697494753e-05, "loss": 0.149, "step": 8212 }, { "epoch": 0.53, "learning_rate": 9.624210482091335e-05, "loss": 0.1506, "step": 8213 }, { "epoch": 0.53, "learning_rate": 9.62213528289412e-05, "loss": 0.1515, "step": 8214 }, { "epoch": 0.53, "learning_rate": 9.620060099992609e-05, "loss": 0.1653, "step": 8215 }, { "epoch": 0.53, "learning_rate": 9.617984933476291e-05, "loss": 0.1317, "step": 8216 }, { "epoch": 0.53, "learning_rate": 9.615909783434665e-05, "loss": 0.1561, "step": 8217 }, { "epoch": 0.53, "learning_rate": 9.613834649957216e-05, "loss": 0.1439, "step": 8218 }, { "epoch": 0.53, "learning_rate": 9.61175953313344e-05, "loss": 0.1493, "step": 8219 }, { "epoch": 0.53, "learning_rate": 9.609684433052829e-05, "loss": 0.1473, "step": 8220 }, { "epoch": 0.53, "learning_rate": 9.607609349804869e-05, "loss": 0.1498, "step": 8221 }, { "epoch": 0.53, "learning_rate": 9.605534283479055e-05, "loss": 0.1633, "step": 8222 }, { "epoch": 0.53, "learning_rate": 9.603459234164868e-05, "loss": 0.1528, "step": 8223 }, { "epoch": 0.53, "learning_rate": 9.601384201951807e-05, "loss": 0.1763, "step": 8224 }, { "epoch": 0.53, "learning_rate": 9.59930918692935e-05, "loss": 0.1424, "step": 8225 }, { "epoch": 0.53, "learning_rate": 9.597234189186986e-05, "loss": 0.1322, "step": 8226 }, { "epoch": 0.53, "learning_rate": 9.595159208814205e-05, "loss": 0.157, "step": 8227 }, { "epoch": 0.53, "learning_rate": 9.593084245900486e-05, "loss": 0.1493, "step": 8228 }, { "epoch": 0.53, "learning_rate": 9.591009300535317e-05, "loss": 0.144, "step": 8229 }, { "epoch": 0.53, "learning_rate": 9.588934372808182e-05, "loss": 0.1611, "step": 8230 }, { "epoch": 0.53, "learning_rate": 9.58685946280856e-05, "loss": 0.1257, "step": 8231 }, { "epoch": 0.53, "learning_rate": 9.584784570625938e-05, "loss": 0.1379, "step": 8232 }, { "epoch": 0.53, "learning_rate": 9.582709696349792e-05, "loss": 0.1316, "step": 8233 }, { "epoch": 0.53, "learning_rate": 9.58063484006961e-05, "loss": 0.1652, "step": 8234 }, { "epoch": 0.53, "learning_rate": 9.578560001874863e-05, "loss": 0.1431, "step": 8235 }, { "epoch": 0.53, "learning_rate": 9.576485181855035e-05, "loss": 0.1499, "step": 8236 }, { "epoch": 0.53, "learning_rate": 9.574410380099606e-05, "loss": 0.1303, "step": 8237 }, { "epoch": 0.53, "learning_rate": 9.572335596698047e-05, "loss": 0.1282, "step": 8238 }, { "epoch": 0.53, "learning_rate": 9.57026083173984e-05, "loss": 0.1385, "step": 8239 }, { "epoch": 0.53, "learning_rate": 9.568186085314457e-05, "loss": 0.1547, "step": 8240 }, { "epoch": 0.53, "learning_rate": 9.566111357511381e-05, "loss": 0.1384, "step": 8241 }, { "epoch": 0.53, "learning_rate": 9.564036648420073e-05, "loss": 0.1439, "step": 8242 }, { "epoch": 0.53, "learning_rate": 9.561961958130019e-05, "loss": 0.162, "step": 8243 }, { "epoch": 0.53, "learning_rate": 9.559887286730686e-05, "loss": 0.1548, "step": 8244 }, { "epoch": 0.53, "learning_rate": 9.557812634311543e-05, "loss": 0.129, "step": 8245 }, { "epoch": 0.53, "learning_rate": 9.555738000962066e-05, "loss": 0.1321, "step": 8246 }, { "epoch": 0.53, "learning_rate": 9.55366338677172e-05, "loss": 0.1474, "step": 8247 }, { "epoch": 0.53, "learning_rate": 9.551588791829982e-05, "loss": 0.1351, "step": 8248 }, { "epoch": 0.53, "learning_rate": 9.549514216226311e-05, "loss": 0.1452, "step": 8249 }, { "epoch": 0.53, "learning_rate": 9.547439660050186e-05, "loss": 0.1454, "step": 8250 }, { "epoch": 0.53, "learning_rate": 9.545365123391065e-05, "loss": 0.1566, "step": 8251 }, { "epoch": 0.53, "learning_rate": 9.543290606338413e-05, "loss": 0.1568, "step": 8252 }, { "epoch": 0.53, "learning_rate": 9.541216108981704e-05, "loss": 0.1587, "step": 8253 }, { "epoch": 0.53, "learning_rate": 9.539141631410391e-05, "loss": 0.1718, "step": 8254 }, { "epoch": 0.53, "learning_rate": 9.537067173713947e-05, "loss": 0.1633, "step": 8255 }, { "epoch": 0.53, "learning_rate": 9.534992735981828e-05, "loss": 0.1384, "step": 8256 }, { "epoch": 0.53, "learning_rate": 9.532918318303502e-05, "loss": 0.1366, "step": 8257 }, { "epoch": 0.53, "learning_rate": 9.530843920768424e-05, "loss": 0.1538, "step": 8258 }, { "epoch": 0.53, "learning_rate": 9.528769543466054e-05, "loss": 0.1394, "step": 8259 }, { "epoch": 0.53, "learning_rate": 9.526695186485859e-05, "loss": 0.1517, "step": 8260 }, { "epoch": 0.53, "learning_rate": 9.524620849917286e-05, "loss": 0.1334, "step": 8261 }, { "epoch": 0.53, "learning_rate": 9.5225465338498e-05, "loss": 0.1555, "step": 8262 }, { "epoch": 0.53, "learning_rate": 9.520472238372852e-05, "loss": 0.1382, "step": 8263 }, { "epoch": 0.53, "learning_rate": 9.518397963575907e-05, "loss": 0.1335, "step": 8264 }, { "epoch": 0.53, "learning_rate": 9.516323709548411e-05, "loss": 0.1612, "step": 8265 }, { "epoch": 0.53, "learning_rate": 9.514249476379817e-05, "loss": 0.147, "step": 8266 }, { "epoch": 0.53, "learning_rate": 9.512175264159587e-05, "loss": 0.1369, "step": 8267 }, { "epoch": 0.53, "learning_rate": 9.510101072977163e-05, "loss": 0.1669, "step": 8268 }, { "epoch": 0.53, "learning_rate": 9.508026902922001e-05, "loss": 0.1596, "step": 8269 }, { "epoch": 0.53, "learning_rate": 9.505952754083546e-05, "loss": 0.1496, "step": 8270 }, { "epoch": 0.53, "learning_rate": 9.50387862655126e-05, "loss": 0.1533, "step": 8271 }, { "epoch": 0.53, "learning_rate": 9.501804520414578e-05, "loss": 0.1414, "step": 8272 }, { "epoch": 0.53, "learning_rate": 9.499730435762951e-05, "loss": 0.1303, "step": 8273 }, { "epoch": 0.53, "learning_rate": 9.497656372685832e-05, "loss": 0.1448, "step": 8274 }, { "epoch": 0.53, "learning_rate": 9.495582331272655e-05, "loss": 0.1496, "step": 8275 }, { "epoch": 0.53, "learning_rate": 9.493508311612874e-05, "loss": 0.1532, "step": 8276 }, { "epoch": 0.53, "learning_rate": 9.491434313795927e-05, "loss": 0.1494, "step": 8277 }, { "epoch": 0.53, "learning_rate": 9.489360337911264e-05, "loss": 0.1467, "step": 8278 }, { "epoch": 0.53, "learning_rate": 9.487286384048319e-05, "loss": 0.1426, "step": 8279 }, { "epoch": 0.53, "learning_rate": 9.485212452296535e-05, "loss": 0.1546, "step": 8280 }, { "epoch": 0.53, "learning_rate": 9.483138542745357e-05, "loss": 0.1509, "step": 8281 }, { "epoch": 0.53, "learning_rate": 9.481064655484214e-05, "loss": 0.1641, "step": 8282 }, { "epoch": 0.53, "learning_rate": 9.478990790602552e-05, "loss": 0.1415, "step": 8283 }, { "epoch": 0.53, "learning_rate": 9.476916948189804e-05, "loss": 0.1246, "step": 8284 }, { "epoch": 0.53, "learning_rate": 9.474843128335412e-05, "loss": 0.141, "step": 8285 }, { "epoch": 0.53, "learning_rate": 9.472769331128804e-05, "loss": 0.1486, "step": 8286 }, { "epoch": 0.53, "learning_rate": 9.470695556659415e-05, "loss": 0.1454, "step": 8287 }, { "epoch": 0.53, "learning_rate": 9.468621805016682e-05, "loss": 0.1384, "step": 8288 }, { "epoch": 0.53, "learning_rate": 9.466548076290032e-05, "loss": 0.1476, "step": 8289 }, { "epoch": 0.53, "learning_rate": 9.464474370568903e-05, "loss": 0.1485, "step": 8290 }, { "epoch": 0.53, "learning_rate": 9.462400687942717e-05, "loss": 0.1625, "step": 8291 }, { "epoch": 0.53, "learning_rate": 9.46032702850091e-05, "loss": 0.1287, "step": 8292 }, { "epoch": 0.53, "learning_rate": 9.45825339233291e-05, "loss": 0.1816, "step": 8293 }, { "epoch": 0.53, "learning_rate": 9.456179779528135e-05, "loss": 0.1349, "step": 8294 }, { "epoch": 0.53, "learning_rate": 9.454106190176024e-05, "loss": 0.1381, "step": 8295 }, { "epoch": 0.53, "learning_rate": 9.452032624365989e-05, "loss": 0.1524, "step": 8296 }, { "epoch": 0.53, "learning_rate": 9.449959082187468e-05, "loss": 0.1444, "step": 8297 }, { "epoch": 0.53, "learning_rate": 9.447885563729872e-05, "loss": 0.1572, "step": 8298 }, { "epoch": 0.53, "learning_rate": 9.44581206908263e-05, "loss": 0.152, "step": 8299 }, { "epoch": 0.53, "learning_rate": 9.443738598335164e-05, "loss": 0.1486, "step": 8300 }, { "epoch": 0.53, "learning_rate": 9.441665151576885e-05, "loss": 0.1233, "step": 8301 }, { "epoch": 0.53, "learning_rate": 9.43959172889722e-05, "loss": 0.1308, "step": 8302 }, { "epoch": 0.53, "learning_rate": 9.437518330385584e-05, "loss": 0.1351, "step": 8303 }, { "epoch": 0.53, "learning_rate": 9.435444956131398e-05, "loss": 0.1454, "step": 8304 }, { "epoch": 0.53, "learning_rate": 9.43337160622407e-05, "loss": 0.1486, "step": 8305 }, { "epoch": 0.53, "learning_rate": 9.431298280753024e-05, "loss": 0.1575, "step": 8306 }, { "epoch": 0.53, "learning_rate": 9.42922497980767e-05, "loss": 0.156, "step": 8307 }, { "epoch": 0.53, "learning_rate": 9.427151703477413e-05, "loss": 0.1552, "step": 8308 }, { "epoch": 0.53, "learning_rate": 9.425078451851676e-05, "loss": 0.135, "step": 8309 }, { "epoch": 0.53, "learning_rate": 9.423005225019862e-05, "loss": 0.1291, "step": 8310 }, { "epoch": 0.53, "learning_rate": 9.420932023071389e-05, "loss": 0.1311, "step": 8311 }, { "epoch": 0.53, "learning_rate": 9.418858846095654e-05, "loss": 0.1345, "step": 8312 }, { "epoch": 0.53, "learning_rate": 9.416785694182071e-05, "loss": 0.1504, "step": 8313 }, { "epoch": 0.53, "learning_rate": 9.41471256742005e-05, "loss": 0.1434, "step": 8314 }, { "epoch": 0.53, "learning_rate": 9.412639465898984e-05, "loss": 0.1665, "step": 8315 }, { "epoch": 0.53, "learning_rate": 9.410566389708289e-05, "loss": 0.1773, "step": 8316 }, { "epoch": 0.53, "learning_rate": 9.408493338937362e-05, "loss": 0.1549, "step": 8317 }, { "epoch": 0.53, "learning_rate": 9.40642031367561e-05, "loss": 0.1472, "step": 8318 }, { "epoch": 0.53, "learning_rate": 9.404347314012427e-05, "loss": 0.1578, "step": 8319 }, { "epoch": 0.53, "learning_rate": 9.402274340037215e-05, "loss": 0.1568, "step": 8320 }, { "epoch": 0.53, "learning_rate": 9.400201391839377e-05, "loss": 0.1503, "step": 8321 }, { "epoch": 0.53, "learning_rate": 9.398128469508303e-05, "loss": 0.142, "step": 8322 }, { "epoch": 0.53, "learning_rate": 9.396055573133396e-05, "loss": 0.1713, "step": 8323 }, { "epoch": 0.53, "learning_rate": 9.393982702804045e-05, "loss": 0.139, "step": 8324 }, { "epoch": 0.53, "learning_rate": 9.391909858609653e-05, "loss": 0.1264, "step": 8325 }, { "epoch": 0.53, "learning_rate": 9.389837040639604e-05, "loss": 0.1491, "step": 8326 }, { "epoch": 0.53, "learning_rate": 9.387764248983291e-05, "loss": 0.1629, "step": 8327 }, { "epoch": 0.53, "learning_rate": 9.385691483730112e-05, "loss": 0.1298, "step": 8328 }, { "epoch": 0.53, "learning_rate": 9.383618744969448e-05, "loss": 0.1373, "step": 8329 }, { "epoch": 0.53, "learning_rate": 9.381546032790693e-05, "loss": 0.1491, "step": 8330 }, { "epoch": 0.53, "learning_rate": 9.379473347283229e-05, "loss": 0.1394, "step": 8331 }, { "epoch": 0.53, "learning_rate": 9.377400688536451e-05, "loss": 0.1372, "step": 8332 }, { "epoch": 0.53, "learning_rate": 9.375328056639737e-05, "loss": 0.1471, "step": 8333 }, { "epoch": 0.53, "learning_rate": 9.373255451682469e-05, "loss": 0.1241, "step": 8334 }, { "epoch": 0.53, "learning_rate": 9.371182873754034e-05, "loss": 0.1336, "step": 8335 }, { "epoch": 0.53, "learning_rate": 9.369110322943812e-05, "loss": 0.1368, "step": 8336 }, { "epoch": 0.53, "learning_rate": 9.367037799341187e-05, "loss": 0.1642, "step": 8337 }, { "epoch": 0.53, "learning_rate": 9.36496530303553e-05, "loss": 0.1508, "step": 8338 }, { "epoch": 0.53, "learning_rate": 9.362892834116228e-05, "loss": 0.1344, "step": 8339 }, { "epoch": 0.53, "learning_rate": 9.360820392672653e-05, "loss": 0.1566, "step": 8340 }, { "epoch": 0.53, "learning_rate": 9.35874797879418e-05, "loss": 0.1451, "step": 8341 }, { "epoch": 0.53, "learning_rate": 9.356675592570184e-05, "loss": 0.1521, "step": 8342 }, { "epoch": 0.53, "learning_rate": 9.354603234090039e-05, "loss": 0.1447, "step": 8343 }, { "epoch": 0.54, "learning_rate": 9.352530903443119e-05, "loss": 0.1395, "step": 8344 }, { "epoch": 0.54, "learning_rate": 9.35045860071879e-05, "loss": 0.142, "step": 8345 }, { "epoch": 0.54, "learning_rate": 9.348386326006425e-05, "loss": 0.1504, "step": 8346 }, { "epoch": 0.54, "learning_rate": 9.346314079395395e-05, "loss": 0.1423, "step": 8347 }, { "epoch": 0.54, "learning_rate": 9.344241860975059e-05, "loss": 0.1423, "step": 8348 }, { "epoch": 0.54, "learning_rate": 9.34216967083479e-05, "loss": 0.1418, "step": 8349 }, { "epoch": 0.54, "learning_rate": 9.340097509063949e-05, "loss": 0.1552, "step": 8350 }, { "epoch": 0.54, "learning_rate": 9.338025375751907e-05, "loss": 0.1455, "step": 8351 }, { "epoch": 0.54, "learning_rate": 9.335953270988014e-05, "loss": 0.1431, "step": 8352 }, { "epoch": 0.54, "learning_rate": 9.33388119486164e-05, "loss": 0.1423, "step": 8353 }, { "epoch": 0.54, "learning_rate": 9.331809147462145e-05, "loss": 0.1455, "step": 8354 }, { "epoch": 0.54, "learning_rate": 9.32973712887888e-05, "loss": 0.1721, "step": 8355 }, { "epoch": 0.54, "learning_rate": 9.327665139201211e-05, "loss": 0.1387, "step": 8356 }, { "epoch": 0.54, "learning_rate": 9.325593178518489e-05, "loss": 0.1283, "step": 8357 }, { "epoch": 0.54, "learning_rate": 9.323521246920073e-05, "loss": 0.1283, "step": 8358 }, { "epoch": 0.54, "learning_rate": 9.321449344495309e-05, "loss": 0.1486, "step": 8359 }, { "epoch": 0.54, "learning_rate": 9.319377471333558e-05, "loss": 0.1528, "step": 8360 }, { "epoch": 0.54, "learning_rate": 9.317305627524169e-05, "loss": 0.161, "step": 8361 }, { "epoch": 0.54, "learning_rate": 9.315233813156487e-05, "loss": 0.1542, "step": 8362 }, { "epoch": 0.54, "learning_rate": 9.313162028319865e-05, "loss": 0.1701, "step": 8363 }, { "epoch": 0.54, "learning_rate": 9.311090273103647e-05, "loss": 0.1432, "step": 8364 }, { "epoch": 0.54, "learning_rate": 9.309018547597185e-05, "loss": 0.1348, "step": 8365 }, { "epoch": 0.54, "learning_rate": 9.306946851889817e-05, "loss": 0.1531, "step": 8366 }, { "epoch": 0.54, "learning_rate": 9.30487518607089e-05, "loss": 0.1498, "step": 8367 }, { "epoch": 0.54, "learning_rate": 9.302803550229749e-05, "loss": 0.166, "step": 8368 }, { "epoch": 0.54, "learning_rate": 9.300731944455726e-05, "loss": 0.1429, "step": 8369 }, { "epoch": 0.54, "learning_rate": 9.29866036883817e-05, "loss": 0.143, "step": 8370 }, { "epoch": 0.54, "learning_rate": 9.29658882346641e-05, "loss": 0.1607, "step": 8371 }, { "epoch": 0.54, "learning_rate": 9.294517308429794e-05, "loss": 0.1612, "step": 8372 }, { "epoch": 0.54, "learning_rate": 9.292445823817646e-05, "loss": 0.1472, "step": 8373 }, { "epoch": 0.54, "learning_rate": 9.290374369719312e-05, "loss": 0.1448, "step": 8374 }, { "epoch": 0.54, "learning_rate": 9.288302946224118e-05, "loss": 0.1556, "step": 8375 }, { "epoch": 0.54, "learning_rate": 9.286231553421392e-05, "loss": 0.1306, "step": 8376 }, { "epoch": 0.54, "learning_rate": 9.284160191400475e-05, "loss": 0.1386, "step": 8377 }, { "epoch": 0.54, "learning_rate": 9.282088860250687e-05, "loss": 0.1491, "step": 8378 }, { "epoch": 0.54, "learning_rate": 9.280017560061359e-05, "loss": 0.1356, "step": 8379 }, { "epoch": 0.54, "learning_rate": 9.277946290921816e-05, "loss": 0.133, "step": 8380 }, { "epoch": 0.54, "learning_rate": 9.275875052921389e-05, "loss": 0.1289, "step": 8381 }, { "epoch": 0.54, "learning_rate": 9.273803846149395e-05, "loss": 0.1349, "step": 8382 }, { "epoch": 0.54, "learning_rate": 9.271732670695157e-05, "loss": 0.1417, "step": 8383 }, { "epoch": 0.54, "learning_rate": 9.269661526648e-05, "loss": 0.1547, "step": 8384 }, { "epoch": 0.54, "learning_rate": 9.267590414097238e-05, "loss": 0.15, "step": 8385 }, { "epoch": 0.54, "learning_rate": 9.265519333132195e-05, "loss": 0.1507, "step": 8386 }, { "epoch": 0.54, "learning_rate": 9.263448283842187e-05, "loss": 0.1589, "step": 8387 }, { "epoch": 0.54, "learning_rate": 9.261377266316524e-05, "loss": 0.1428, "step": 8388 }, { "epoch": 0.54, "learning_rate": 9.259306280644525e-05, "loss": 0.147, "step": 8389 }, { "epoch": 0.54, "learning_rate": 9.257235326915501e-05, "loss": 0.1454, "step": 8390 }, { "epoch": 0.54, "learning_rate": 9.255164405218768e-05, "loss": 0.1304, "step": 8391 }, { "epoch": 0.54, "learning_rate": 9.25309351564363e-05, "loss": 0.1416, "step": 8392 }, { "epoch": 0.54, "learning_rate": 9.251022658279398e-05, "loss": 0.1606, "step": 8393 }, { "epoch": 0.54, "learning_rate": 9.248951833215382e-05, "loss": 0.1442, "step": 8394 }, { "epoch": 0.54, "learning_rate": 9.246881040540882e-05, "loss": 0.1482, "step": 8395 }, { "epoch": 0.54, "learning_rate": 9.244810280345206e-05, "loss": 0.1674, "step": 8396 }, { "epoch": 0.54, "learning_rate": 9.242739552717656e-05, "loss": 0.1448, "step": 8397 }, { "epoch": 0.54, "learning_rate": 9.24066885774754e-05, "loss": 0.1405, "step": 8398 }, { "epoch": 0.54, "learning_rate": 9.238598195524146e-05, "loss": 0.1488, "step": 8399 }, { "epoch": 0.54, "learning_rate": 9.236527566136785e-05, "loss": 0.1488, "step": 8400 }, { "epoch": 0.54, "learning_rate": 9.234456969674748e-05, "loss": 0.1404, "step": 8401 }, { "epoch": 0.54, "learning_rate": 9.23238640622733e-05, "loss": 0.1402, "step": 8402 }, { "epoch": 0.54, "learning_rate": 9.230315875883831e-05, "loss": 0.1357, "step": 8403 }, { "epoch": 0.54, "learning_rate": 9.228245378733537e-05, "loss": 0.1412, "step": 8404 }, { "epoch": 0.54, "learning_rate": 9.226174914865749e-05, "loss": 0.1426, "step": 8405 }, { "epoch": 0.54, "learning_rate": 9.224104484369747e-05, "loss": 0.148, "step": 8406 }, { "epoch": 0.54, "learning_rate": 9.222034087334826e-05, "loss": 0.1695, "step": 8407 }, { "epoch": 0.54, "learning_rate": 9.219963723850277e-05, "loss": 0.1254, "step": 8408 }, { "epoch": 0.54, "learning_rate": 9.217893394005376e-05, "loss": 0.1558, "step": 8409 }, { "epoch": 0.54, "learning_rate": 9.215823097889415e-05, "loss": 0.1382, "step": 8410 }, { "epoch": 0.54, "learning_rate": 9.213752835591673e-05, "loss": 0.1319, "step": 8411 }, { "epoch": 0.54, "learning_rate": 9.211682607201438e-05, "loss": 0.1403, "step": 8412 }, { "epoch": 0.54, "learning_rate": 9.209612412807982e-05, "loss": 0.1573, "step": 8413 }, { "epoch": 0.54, "learning_rate": 9.20754225250059e-05, "loss": 0.1215, "step": 8414 }, { "epoch": 0.54, "learning_rate": 9.205472126368537e-05, "loss": 0.1428, "step": 8415 }, { "epoch": 0.54, "learning_rate": 9.203402034501095e-05, "loss": 0.1558, "step": 8416 }, { "epoch": 0.54, "learning_rate": 9.201331976987546e-05, "loss": 0.1542, "step": 8417 }, { "epoch": 0.54, "learning_rate": 9.199261953917153e-05, "loss": 0.1477, "step": 8418 }, { "epoch": 0.54, "learning_rate": 9.1971919653792e-05, "loss": 0.1566, "step": 8419 }, { "epoch": 0.54, "learning_rate": 9.19512201146294e-05, "loss": 0.1513, "step": 8420 }, { "epoch": 0.54, "learning_rate": 9.19305209225766e-05, "loss": 0.1329, "step": 8421 }, { "epoch": 0.54, "learning_rate": 9.190982207852615e-05, "loss": 0.1797, "step": 8422 }, { "epoch": 0.54, "learning_rate": 9.188912358337072e-05, "loss": 0.1372, "step": 8423 }, { "epoch": 0.54, "learning_rate": 9.186842543800299e-05, "loss": 0.1561, "step": 8424 }, { "epoch": 0.54, "learning_rate": 9.184772764331552e-05, "loss": 0.1617, "step": 8425 }, { "epoch": 0.54, "learning_rate": 9.182703020020096e-05, "loss": 0.1476, "step": 8426 }, { "epoch": 0.54, "learning_rate": 9.180633310955187e-05, "loss": 0.163, "step": 8427 }, { "epoch": 0.54, "learning_rate": 9.178563637226092e-05, "loss": 0.1655, "step": 8428 }, { "epoch": 0.54, "learning_rate": 9.176493998922056e-05, "loss": 0.1472, "step": 8429 }, { "epoch": 0.54, "learning_rate": 9.174424396132336e-05, "loss": 0.1426, "step": 8430 }, { "epoch": 0.54, "learning_rate": 9.172354828946193e-05, "loss": 0.1421, "step": 8431 }, { "epoch": 0.54, "learning_rate": 9.170285297452868e-05, "loss": 0.1501, "step": 8432 }, { "epoch": 0.54, "learning_rate": 9.168215801741618e-05, "loss": 0.1377, "step": 8433 }, { "epoch": 0.54, "learning_rate": 9.166146341901687e-05, "loss": 0.1336, "step": 8434 }, { "epoch": 0.54, "learning_rate": 9.16407691802233e-05, "loss": 0.1563, "step": 8435 }, { "epoch": 0.54, "learning_rate": 9.162007530192784e-05, "loss": 0.1282, "step": 8436 }, { "epoch": 0.54, "learning_rate": 9.159938178502294e-05, "loss": 0.145, "step": 8437 }, { "epoch": 0.54, "learning_rate": 9.157868863040108e-05, "loss": 0.1457, "step": 8438 }, { "epoch": 0.54, "learning_rate": 9.155799583895461e-05, "loss": 0.1415, "step": 8439 }, { "epoch": 0.54, "learning_rate": 9.153730341157595e-05, "loss": 0.1351, "step": 8440 }, { "epoch": 0.54, "learning_rate": 9.151661134915745e-05, "loss": 0.1487, "step": 8441 }, { "epoch": 0.54, "learning_rate": 9.149591965259154e-05, "loss": 0.1663, "step": 8442 }, { "epoch": 0.54, "learning_rate": 9.14752283227705e-05, "loss": 0.1521, "step": 8443 }, { "epoch": 0.54, "learning_rate": 9.145453736058663e-05, "loss": 0.1402, "step": 8444 }, { "epoch": 0.54, "learning_rate": 9.143384676693235e-05, "loss": 0.1345, "step": 8445 }, { "epoch": 0.54, "learning_rate": 9.141315654269985e-05, "loss": 0.1359, "step": 8446 }, { "epoch": 0.54, "learning_rate": 9.139246668878147e-05, "loss": 0.1457, "step": 8447 }, { "epoch": 0.54, "learning_rate": 9.137177720606947e-05, "loss": 0.1315, "step": 8448 }, { "epoch": 0.54, "learning_rate": 9.13510880954561e-05, "loss": 0.1509, "step": 8449 }, { "epoch": 0.54, "learning_rate": 9.13303993578336e-05, "loss": 0.1313, "step": 8450 }, { "epoch": 0.54, "learning_rate": 9.130971099409413e-05, "loss": 0.1302, "step": 8451 }, { "epoch": 0.54, "learning_rate": 9.128902300512999e-05, "loss": 0.1681, "step": 8452 }, { "epoch": 0.54, "learning_rate": 9.126833539183325e-05, "loss": 0.1452, "step": 8453 }, { "epoch": 0.54, "learning_rate": 9.124764815509617e-05, "loss": 0.1366, "step": 8454 }, { "epoch": 0.54, "learning_rate": 9.122696129581085e-05, "loss": 0.1499, "step": 8455 }, { "epoch": 0.54, "learning_rate": 9.120627481486949e-05, "loss": 0.1367, "step": 8456 }, { "epoch": 0.54, "learning_rate": 9.118558871316416e-05, "loss": 0.1458, "step": 8457 }, { "epoch": 0.54, "learning_rate": 9.116490299158694e-05, "loss": 0.162, "step": 8458 }, { "epoch": 0.54, "learning_rate": 9.114421765102999e-05, "loss": 0.1346, "step": 8459 }, { "epoch": 0.54, "learning_rate": 9.112353269238528e-05, "loss": 0.1533, "step": 8460 }, { "epoch": 0.54, "learning_rate": 9.1102848116545e-05, "loss": 0.1475, "step": 8461 }, { "epoch": 0.54, "learning_rate": 9.108216392440105e-05, "loss": 0.14, "step": 8462 }, { "epoch": 0.54, "learning_rate": 9.106148011684555e-05, "loss": 0.1484, "step": 8463 }, { "epoch": 0.54, "learning_rate": 9.10407966947705e-05, "loss": 0.1599, "step": 8464 }, { "epoch": 0.54, "learning_rate": 9.102011365906779e-05, "loss": 0.1457, "step": 8465 }, { "epoch": 0.54, "learning_rate": 9.099943101062949e-05, "loss": 0.1239, "step": 8466 }, { "epoch": 0.54, "learning_rate": 9.097874875034751e-05, "loss": 0.152, "step": 8467 }, { "epoch": 0.54, "learning_rate": 9.095806687911384e-05, "loss": 0.166, "step": 8468 }, { "epoch": 0.54, "learning_rate": 9.093738539782033e-05, "loss": 0.1868, "step": 8469 }, { "epoch": 0.54, "learning_rate": 9.091670430735892e-05, "loss": 0.1439, "step": 8470 }, { "epoch": 0.54, "learning_rate": 9.089602360862152e-05, "loss": 0.1424, "step": 8471 }, { "epoch": 0.54, "learning_rate": 9.087534330249994e-05, "loss": 0.147, "step": 8472 }, { "epoch": 0.54, "learning_rate": 9.085466338988608e-05, "loss": 0.145, "step": 8473 }, { "epoch": 0.54, "learning_rate": 9.083398387167176e-05, "loss": 0.1691, "step": 8474 }, { "epoch": 0.54, "learning_rate": 9.081330474874885e-05, "loss": 0.1524, "step": 8475 }, { "epoch": 0.54, "learning_rate": 9.079262602200907e-05, "loss": 0.1508, "step": 8476 }, { "epoch": 0.54, "learning_rate": 9.077194769234425e-05, "loss": 0.1538, "step": 8477 }, { "epoch": 0.54, "learning_rate": 9.075126976064618e-05, "loss": 0.147, "step": 8478 }, { "epoch": 0.54, "learning_rate": 9.073059222780654e-05, "loss": 0.1316, "step": 8479 }, { "epoch": 0.54, "learning_rate": 9.070991509471714e-05, "loss": 0.1399, "step": 8480 }, { "epoch": 0.54, "learning_rate": 9.068923836226963e-05, "loss": 0.1579, "step": 8481 }, { "epoch": 0.54, "learning_rate": 9.06685620313558e-05, "loss": 0.1567, "step": 8482 }, { "epoch": 0.54, "learning_rate": 9.064788610286726e-05, "loss": 0.1361, "step": 8483 }, { "epoch": 0.54, "learning_rate": 9.062721057769567e-05, "loss": 0.1514, "step": 8484 }, { "epoch": 0.54, "learning_rate": 9.060653545673274e-05, "loss": 0.1552, "step": 8485 }, { "epoch": 0.54, "learning_rate": 9.058586074087003e-05, "loss": 0.1417, "step": 8486 }, { "epoch": 0.54, "learning_rate": 9.056518643099919e-05, "loss": 0.1436, "step": 8487 }, { "epoch": 0.54, "learning_rate": 9.05445125280118e-05, "loss": 0.1524, "step": 8488 }, { "epoch": 0.54, "learning_rate": 9.052383903279948e-05, "loss": 0.1666, "step": 8489 }, { "epoch": 0.54, "learning_rate": 9.050316594625374e-05, "loss": 0.172, "step": 8490 }, { "epoch": 0.54, "learning_rate": 9.048249326926612e-05, "loss": 0.1565, "step": 8491 }, { "epoch": 0.54, "learning_rate": 9.04618210027282e-05, "loss": 0.1383, "step": 8492 }, { "epoch": 0.54, "learning_rate": 9.044114914753143e-05, "loss": 0.1477, "step": 8493 }, { "epoch": 0.54, "learning_rate": 9.042047770456731e-05, "loss": 0.1329, "step": 8494 }, { "epoch": 0.54, "learning_rate": 9.039980667472733e-05, "loss": 0.1424, "step": 8495 }, { "epoch": 0.54, "learning_rate": 9.037913605890296e-05, "loss": 0.1451, "step": 8496 }, { "epoch": 0.54, "learning_rate": 9.03584658579856e-05, "loss": 0.1538, "step": 8497 }, { "epoch": 0.54, "learning_rate": 9.033779607286666e-05, "loss": 0.1337, "step": 8498 }, { "epoch": 0.54, "learning_rate": 9.031712670443759e-05, "loss": 0.1403, "step": 8499 }, { "epoch": 0.55, "learning_rate": 9.02964577535897e-05, "loss": 0.141, "step": 8500 }, { "epoch": 0.55, "learning_rate": 9.027578922121444e-05, "loss": 0.1469, "step": 8501 }, { "epoch": 0.55, "learning_rate": 9.025512110820307e-05, "loss": 0.1399, "step": 8502 }, { "epoch": 0.55, "learning_rate": 9.023445341544701e-05, "loss": 0.1464, "step": 8503 }, { "epoch": 0.55, "learning_rate": 9.02137861438375e-05, "loss": 0.1507, "step": 8504 }, { "epoch": 0.55, "learning_rate": 9.019311929426583e-05, "loss": 0.1745, "step": 8505 }, { "epoch": 0.55, "learning_rate": 9.017245286762334e-05, "loss": 0.1319, "step": 8506 }, { "epoch": 0.55, "learning_rate": 9.015178686480117e-05, "loss": 0.1302, "step": 8507 }, { "epoch": 0.55, "learning_rate": 9.013112128669069e-05, "loss": 0.147, "step": 8508 }, { "epoch": 0.55, "learning_rate": 9.011045613418301e-05, "loss": 0.1407, "step": 8509 }, { "epoch": 0.55, "learning_rate": 9.008979140816941e-05, "loss": 0.1405, "step": 8510 }, { "epoch": 0.55, "learning_rate": 9.006912710954107e-05, "loss": 0.1381, "step": 8511 }, { "epoch": 0.55, "learning_rate": 9.004846323918906e-05, "loss": 0.1705, "step": 8512 }, { "epoch": 0.55, "learning_rate": 9.002779979800461e-05, "loss": 0.1241, "step": 8513 }, { "epoch": 0.55, "learning_rate": 9.000713678687881e-05, "loss": 0.1355, "step": 8514 }, { "epoch": 0.55, "learning_rate": 8.998647420670282e-05, "loss": 0.1466, "step": 8515 }, { "epoch": 0.55, "learning_rate": 8.996581205836765e-05, "loss": 0.1317, "step": 8516 }, { "epoch": 0.55, "learning_rate": 8.994515034276442e-05, "loss": 0.1521, "step": 8517 }, { "epoch": 0.55, "learning_rate": 8.992448906078422e-05, "loss": 0.1582, "step": 8518 }, { "epoch": 0.55, "learning_rate": 8.990382821331798e-05, "loss": 0.1326, "step": 8519 }, { "epoch": 0.55, "learning_rate": 8.98831678012568e-05, "loss": 0.1609, "step": 8520 }, { "epoch": 0.55, "learning_rate": 8.986250782549162e-05, "loss": 0.1385, "step": 8521 }, { "epoch": 0.55, "learning_rate": 8.984184828691349e-05, "loss": 0.1543, "step": 8522 }, { "epoch": 0.55, "learning_rate": 8.982118918641327e-05, "loss": 0.1414, "step": 8523 }, { "epoch": 0.55, "learning_rate": 8.980053052488199e-05, "loss": 0.1352, "step": 8524 }, { "epoch": 0.55, "learning_rate": 8.977987230321055e-05, "loss": 0.1362, "step": 8525 }, { "epoch": 0.55, "learning_rate": 8.975921452228978e-05, "loss": 0.1525, "step": 8526 }, { "epoch": 0.55, "learning_rate": 8.973855718301064e-05, "loss": 0.1441, "step": 8527 }, { "epoch": 0.55, "learning_rate": 8.971790028626395e-05, "loss": 0.1356, "step": 8528 }, { "epoch": 0.55, "learning_rate": 8.969724383294062e-05, "loss": 0.1222, "step": 8529 }, { "epoch": 0.55, "learning_rate": 8.967658782393139e-05, "loss": 0.1464, "step": 8530 }, { "epoch": 0.55, "learning_rate": 8.965593226012712e-05, "loss": 0.145, "step": 8531 }, { "epoch": 0.55, "learning_rate": 8.963527714241861e-05, "loss": 0.1498, "step": 8532 }, { "epoch": 0.55, "learning_rate": 8.961462247169656e-05, "loss": 0.1375, "step": 8533 }, { "epoch": 0.55, "learning_rate": 8.959396824885176e-05, "loss": 0.1596, "step": 8534 }, { "epoch": 0.55, "learning_rate": 8.957331447477493e-05, "loss": 0.1642, "step": 8535 }, { "epoch": 0.55, "learning_rate": 8.955266115035683e-05, "loss": 0.1553, "step": 8536 }, { "epoch": 0.55, "learning_rate": 8.953200827648805e-05, "loss": 0.1584, "step": 8537 }, { "epoch": 0.55, "learning_rate": 8.951135585405936e-05, "loss": 0.138, "step": 8538 }, { "epoch": 0.55, "learning_rate": 8.949070388396138e-05, "loss": 0.1591, "step": 8539 }, { "epoch": 0.55, "learning_rate": 8.947005236708468e-05, "loss": 0.1572, "step": 8540 }, { "epoch": 0.55, "learning_rate": 8.944940130431995e-05, "loss": 0.1503, "step": 8541 }, { "epoch": 0.55, "learning_rate": 8.942875069655774e-05, "loss": 0.1483, "step": 8542 }, { "epoch": 0.55, "learning_rate": 8.940810054468867e-05, "loss": 0.1443, "step": 8543 }, { "epoch": 0.55, "learning_rate": 8.938745084960323e-05, "loss": 0.1422, "step": 8544 }, { "epoch": 0.55, "learning_rate": 8.9366801612192e-05, "loss": 0.1422, "step": 8545 }, { "epoch": 0.55, "learning_rate": 8.93461528333455e-05, "loss": 0.1239, "step": 8546 }, { "epoch": 0.55, "learning_rate": 8.932550451395414e-05, "loss": 0.1534, "step": 8547 }, { "epoch": 0.55, "learning_rate": 8.930485665490853e-05, "loss": 0.1401, "step": 8548 }, { "epoch": 0.55, "learning_rate": 8.928420925709899e-05, "loss": 0.1435, "step": 8549 }, { "epoch": 0.55, "learning_rate": 8.926356232141606e-05, "loss": 0.1474, "step": 8550 }, { "epoch": 0.55, "learning_rate": 8.924291584875013e-05, "loss": 0.1642, "step": 8551 }, { "epoch": 0.55, "learning_rate": 8.922226983999152e-05, "loss": 0.1317, "step": 8552 }, { "epoch": 0.55, "learning_rate": 8.92016242960307e-05, "loss": 0.1477, "step": 8553 }, { "epoch": 0.55, "learning_rate": 8.918097921775796e-05, "loss": 0.1394, "step": 8554 }, { "epoch": 0.55, "learning_rate": 8.916033460606369e-05, "loss": 0.1417, "step": 8555 }, { "epoch": 0.55, "learning_rate": 8.913969046183814e-05, "loss": 0.1536, "step": 8556 }, { "epoch": 0.55, "learning_rate": 8.911904678597167e-05, "loss": 0.1387, "step": 8557 }, { "epoch": 0.55, "learning_rate": 8.909840357935453e-05, "loss": 0.144, "step": 8558 }, { "epoch": 0.55, "learning_rate": 8.907776084287693e-05, "loss": 0.1244, "step": 8559 }, { "epoch": 0.55, "learning_rate": 8.905711857742916e-05, "loss": 0.1536, "step": 8560 }, { "epoch": 0.55, "learning_rate": 8.90364767839014e-05, "loss": 0.1268, "step": 8561 }, { "epoch": 0.55, "learning_rate": 8.90158354631839e-05, "loss": 0.1403, "step": 8562 }, { "epoch": 0.55, "learning_rate": 8.899519461616674e-05, "loss": 0.1449, "step": 8563 }, { "epoch": 0.55, "learning_rate": 8.897455424374016e-05, "loss": 0.1502, "step": 8564 }, { "epoch": 0.55, "learning_rate": 8.895391434679426e-05, "loss": 0.1537, "step": 8565 }, { "epoch": 0.55, "learning_rate": 8.893327492621909e-05, "loss": 0.1424, "step": 8566 }, { "epoch": 0.55, "learning_rate": 8.891263598290484e-05, "loss": 0.1344, "step": 8567 }, { "epoch": 0.55, "learning_rate": 8.88919975177415e-05, "loss": 0.1543, "step": 8568 }, { "epoch": 0.55, "learning_rate": 8.88713595316192e-05, "loss": 0.1424, "step": 8569 }, { "epoch": 0.55, "learning_rate": 8.885072202542788e-05, "loss": 0.1348, "step": 8570 }, { "epoch": 0.55, "learning_rate": 8.88300850000576e-05, "loss": 0.1192, "step": 8571 }, { "epoch": 0.55, "learning_rate": 8.880944845639838e-05, "loss": 0.1494, "step": 8572 }, { "epoch": 0.55, "learning_rate": 8.878881239534009e-05, "loss": 0.1594, "step": 8573 }, { "epoch": 0.55, "learning_rate": 8.876817681777276e-05, "loss": 0.1429, "step": 8574 }, { "epoch": 0.55, "learning_rate": 8.874754172458627e-05, "loss": 0.1364, "step": 8575 }, { "epoch": 0.55, "learning_rate": 8.872690711667057e-05, "loss": 0.1268, "step": 8576 }, { "epoch": 0.55, "learning_rate": 8.870627299491548e-05, "loss": 0.1378, "step": 8577 }, { "epoch": 0.55, "learning_rate": 8.86856393602109e-05, "loss": 0.1435, "step": 8578 }, { "epoch": 0.55, "learning_rate": 8.866500621344671e-05, "loss": 0.1278, "step": 8579 }, { "epoch": 0.55, "learning_rate": 8.864437355551263e-05, "loss": 0.1535, "step": 8580 }, { "epoch": 0.55, "learning_rate": 8.862374138729853e-05, "loss": 0.1612, "step": 8581 }, { "epoch": 0.55, "learning_rate": 8.860310970969416e-05, "loss": 0.1451, "step": 8582 }, { "epoch": 0.55, "learning_rate": 8.858247852358932e-05, "loss": 0.1597, "step": 8583 }, { "epoch": 0.55, "learning_rate": 8.856184782987368e-05, "loss": 0.1274, "step": 8584 }, { "epoch": 0.55, "learning_rate": 8.854121762943703e-05, "loss": 0.1503, "step": 8585 }, { "epoch": 0.55, "learning_rate": 8.852058792316902e-05, "loss": 0.1449, "step": 8586 }, { "epoch": 0.55, "learning_rate": 8.84999587119593e-05, "loss": 0.1505, "step": 8587 }, { "epoch": 0.55, "learning_rate": 8.847932999669755e-05, "loss": 0.156, "step": 8588 }, { "epoch": 0.55, "learning_rate": 8.845870177827336e-05, "loss": 0.1391, "step": 8589 }, { "epoch": 0.55, "learning_rate": 8.843807405757643e-05, "loss": 0.1448, "step": 8590 }, { "epoch": 0.55, "learning_rate": 8.841744683549623e-05, "loss": 0.1317, "step": 8591 }, { "epoch": 0.55, "learning_rate": 8.839682011292242e-05, "loss": 0.1331, "step": 8592 }, { "epoch": 0.55, "learning_rate": 8.837619389074455e-05, "loss": 0.1568, "step": 8593 }, { "epoch": 0.55, "learning_rate": 8.835556816985199e-05, "loss": 0.1584, "step": 8594 }, { "epoch": 0.55, "learning_rate": 8.833494295113441e-05, "loss": 0.1647, "step": 8595 }, { "epoch": 0.55, "learning_rate": 8.83143182354812e-05, "loss": 0.131, "step": 8596 }, { "epoch": 0.55, "learning_rate": 8.829369402378185e-05, "loss": 0.1516, "step": 8597 }, { "epoch": 0.55, "learning_rate": 8.827307031692577e-05, "loss": 0.1647, "step": 8598 }, { "epoch": 0.55, "learning_rate": 8.825244711580241e-05, "loss": 0.1417, "step": 8599 }, { "epoch": 0.55, "learning_rate": 8.823182442130114e-05, "loss": 0.1372, "step": 8600 }, { "epoch": 0.55, "learning_rate": 8.82112022343113e-05, "loss": 0.1563, "step": 8601 }, { "epoch": 0.55, "learning_rate": 8.81905805557223e-05, "loss": 0.1324, "step": 8602 }, { "epoch": 0.55, "learning_rate": 8.816995938642339e-05, "loss": 0.1362, "step": 8603 }, { "epoch": 0.55, "learning_rate": 8.814933872730395e-05, "loss": 0.1594, "step": 8604 }, { "epoch": 0.55, "learning_rate": 8.812871857925318e-05, "loss": 0.1436, "step": 8605 }, { "epoch": 0.55, "learning_rate": 8.810809894316046e-05, "loss": 0.148, "step": 8606 }, { "epoch": 0.55, "learning_rate": 8.80874798199149e-05, "loss": 0.147, "step": 8607 }, { "epoch": 0.55, "learning_rate": 8.806686121040577e-05, "loss": 0.1639, "step": 8608 }, { "epoch": 0.55, "learning_rate": 8.804624311552229e-05, "loss": 0.1444, "step": 8609 }, { "epoch": 0.55, "learning_rate": 8.802562553615357e-05, "loss": 0.1338, "step": 8610 }, { "epoch": 0.55, "learning_rate": 8.80050084731888e-05, "loss": 0.1648, "step": 8611 }, { "epoch": 0.55, "learning_rate": 8.798439192751707e-05, "loss": 0.1647, "step": 8612 }, { "epoch": 0.55, "learning_rate": 8.796377590002756e-05, "loss": 0.1394, "step": 8613 }, { "epoch": 0.55, "learning_rate": 8.794316039160929e-05, "loss": 0.1308, "step": 8614 }, { "epoch": 0.55, "learning_rate": 8.79225454031513e-05, "loss": 0.1812, "step": 8615 }, { "epoch": 0.55, "learning_rate": 8.790193093554271e-05, "loss": 0.1384, "step": 8616 }, { "epoch": 0.55, "learning_rate": 8.788131698967243e-05, "loss": 0.1427, "step": 8617 }, { "epoch": 0.55, "learning_rate": 8.786070356642955e-05, "loss": 0.1437, "step": 8618 }, { "epoch": 0.55, "learning_rate": 8.784009066670295e-05, "loss": 0.1386, "step": 8619 }, { "epoch": 0.55, "learning_rate": 8.781947829138167e-05, "loss": 0.1232, "step": 8620 }, { "epoch": 0.55, "learning_rate": 8.779886644135456e-05, "loss": 0.1377, "step": 8621 }, { "epoch": 0.55, "learning_rate": 8.777825511751054e-05, "loss": 0.1566, "step": 8622 }, { "epoch": 0.55, "learning_rate": 8.775764432073853e-05, "loss": 0.1674, "step": 8623 }, { "epoch": 0.55, "learning_rate": 8.773703405192731e-05, "loss": 0.1363, "step": 8624 }, { "epoch": 0.55, "learning_rate": 8.771642431196578e-05, "loss": 0.1364, "step": 8625 }, { "epoch": 0.55, "learning_rate": 8.769581510174275e-05, "loss": 0.1379, "step": 8626 }, { "epoch": 0.55, "learning_rate": 8.767520642214694e-05, "loss": 0.1331, "step": 8627 }, { "epoch": 0.55, "learning_rate": 8.765459827406718e-05, "loss": 0.1443, "step": 8628 }, { "epoch": 0.55, "learning_rate": 8.763399065839217e-05, "loss": 0.1318, "step": 8629 }, { "epoch": 0.55, "learning_rate": 8.76133835760107e-05, "loss": 0.1356, "step": 8630 }, { "epoch": 0.55, "learning_rate": 8.759277702781135e-05, "loss": 0.1551, "step": 8631 }, { "epoch": 0.55, "learning_rate": 8.75721710146829e-05, "loss": 0.1383, "step": 8632 }, { "epoch": 0.55, "learning_rate": 8.755156553751398e-05, "loss": 0.1295, "step": 8633 }, { "epoch": 0.55, "learning_rate": 8.753096059719314e-05, "loss": 0.1567, "step": 8634 }, { "epoch": 0.55, "learning_rate": 8.751035619460908e-05, "loss": 0.1473, "step": 8635 }, { "epoch": 0.55, "learning_rate": 8.748975233065031e-05, "loss": 0.1437, "step": 8636 }, { "epoch": 0.55, "learning_rate": 8.746914900620544e-05, "loss": 0.1398, "step": 8637 }, { "epoch": 0.55, "learning_rate": 8.744854622216295e-05, "loss": 0.1347, "step": 8638 }, { "epoch": 0.55, "learning_rate": 8.742794397941145e-05, "loss": 0.139, "step": 8639 }, { "epoch": 0.55, "learning_rate": 8.740734227883932e-05, "loss": 0.1775, "step": 8640 }, { "epoch": 0.55, "learning_rate": 8.738674112133503e-05, "loss": 0.1373, "step": 8641 }, { "epoch": 0.55, "learning_rate": 8.73661405077871e-05, "loss": 0.1579, "step": 8642 }, { "epoch": 0.55, "learning_rate": 8.734554043908388e-05, "loss": 0.1578, "step": 8643 }, { "epoch": 0.55, "learning_rate": 8.732494091611377e-05, "loss": 0.1418, "step": 8644 }, { "epoch": 0.55, "learning_rate": 8.730434193976516e-05, "loss": 0.1275, "step": 8645 }, { "epoch": 0.55, "learning_rate": 8.728374351092642e-05, "loss": 0.1326, "step": 8646 }, { "epoch": 0.55, "learning_rate": 8.726314563048583e-05, "loss": 0.1414, "step": 8647 }, { "epoch": 0.55, "learning_rate": 8.724254829933167e-05, "loss": 0.1376, "step": 8648 }, { "epoch": 0.55, "learning_rate": 8.722195151835231e-05, "loss": 0.1459, "step": 8649 }, { "epoch": 0.55, "learning_rate": 8.720135528843587e-05, "loss": 0.1399, "step": 8650 }, { "epoch": 0.55, "learning_rate": 8.718075961047069e-05, "loss": 0.1405, "step": 8651 }, { "epoch": 0.55, "learning_rate": 8.71601644853449e-05, "loss": 0.1532, "step": 8652 }, { "epoch": 0.55, "learning_rate": 8.713956991394674e-05, "loss": 0.152, "step": 8653 }, { "epoch": 0.55, "learning_rate": 8.711897589716435e-05, "loss": 0.144, "step": 8654 }, { "epoch": 0.55, "learning_rate": 8.70983824358858e-05, "loss": 0.1448, "step": 8655 }, { "epoch": 0.56, "learning_rate": 8.707778953099931e-05, "loss": 0.1623, "step": 8656 }, { "epoch": 0.56, "learning_rate": 8.705719718339285e-05, "loss": 0.1446, "step": 8657 }, { "epoch": 0.56, "learning_rate": 8.703660539395457e-05, "loss": 0.1476, "step": 8658 }, { "epoch": 0.56, "learning_rate": 8.701601416357244e-05, "loss": 0.1464, "step": 8659 }, { "epoch": 0.56, "learning_rate": 8.699542349313454e-05, "loss": 0.1713, "step": 8660 }, { "epoch": 0.56, "learning_rate": 8.697483338352881e-05, "loss": 0.1402, "step": 8661 }, { "epoch": 0.56, "learning_rate": 8.69542438356432e-05, "loss": 0.1467, "step": 8662 }, { "epoch": 0.56, "learning_rate": 8.693365485036572e-05, "loss": 0.1457, "step": 8663 }, { "epoch": 0.56, "learning_rate": 8.69130664285842e-05, "loss": 0.1501, "step": 8664 }, { "epoch": 0.56, "learning_rate": 8.689247857118658e-05, "loss": 0.1498, "step": 8665 }, { "epoch": 0.56, "learning_rate": 8.687189127906072e-05, "loss": 0.1406, "step": 8666 }, { "epoch": 0.56, "learning_rate": 8.68513045530945e-05, "loss": 0.1315, "step": 8667 }, { "epoch": 0.56, "learning_rate": 8.683071839417566e-05, "loss": 0.1475, "step": 8668 }, { "epoch": 0.56, "learning_rate": 8.681013280319202e-05, "loss": 0.1573, "step": 8669 }, { "epoch": 0.56, "learning_rate": 8.678954778103142e-05, "loss": 0.1494, "step": 8670 }, { "epoch": 0.56, "learning_rate": 8.676896332858148e-05, "loss": 0.157, "step": 8671 }, { "epoch": 0.56, "learning_rate": 8.674837944673001e-05, "loss": 0.1497, "step": 8672 }, { "epoch": 0.56, "learning_rate": 8.672779613636468e-05, "loss": 0.1571, "step": 8673 }, { "epoch": 0.56, "learning_rate": 8.670721339837319e-05, "loss": 0.152, "step": 8674 }, { "epoch": 0.56, "learning_rate": 8.668663123364314e-05, "loss": 0.1448, "step": 8675 }, { "epoch": 0.56, "learning_rate": 8.666604964306215e-05, "loss": 0.1398, "step": 8676 }, { "epoch": 0.56, "learning_rate": 8.664546862751789e-05, "loss": 0.1465, "step": 8677 }, { "epoch": 0.56, "learning_rate": 8.662488818789781e-05, "loss": 0.1392, "step": 8678 }, { "epoch": 0.56, "learning_rate": 8.66043083250896e-05, "loss": 0.1392, "step": 8679 }, { "epoch": 0.56, "learning_rate": 8.658372903998064e-05, "loss": 0.1448, "step": 8680 }, { "epoch": 0.56, "learning_rate": 8.656315033345854e-05, "loss": 0.1551, "step": 8681 }, { "epoch": 0.56, "learning_rate": 8.654257220641074e-05, "loss": 0.1431, "step": 8682 }, { "epoch": 0.56, "learning_rate": 8.652199465972463e-05, "loss": 0.1371, "step": 8683 }, { "epoch": 0.56, "learning_rate": 8.65014176942877e-05, "loss": 0.1519, "step": 8684 }, { "epoch": 0.56, "learning_rate": 8.64808413109873e-05, "loss": 0.1256, "step": 8685 }, { "epoch": 0.56, "learning_rate": 8.646026551071088e-05, "loss": 0.1411, "step": 8686 }, { "epoch": 0.56, "learning_rate": 8.643969029434568e-05, "loss": 0.1434, "step": 8687 }, { "epoch": 0.56, "learning_rate": 8.64191156627791e-05, "loss": 0.1334, "step": 8688 }, { "epoch": 0.56, "learning_rate": 8.639854161689845e-05, "loss": 0.1504, "step": 8689 }, { "epoch": 0.56, "learning_rate": 8.63779681575909e-05, "loss": 0.1268, "step": 8690 }, { "epoch": 0.56, "learning_rate": 8.635739528574378e-05, "loss": 0.1592, "step": 8691 }, { "epoch": 0.56, "learning_rate": 8.633682300224427e-05, "loss": 0.1446, "step": 8692 }, { "epoch": 0.56, "learning_rate": 8.631625130797964e-05, "loss": 0.1421, "step": 8693 }, { "epoch": 0.56, "learning_rate": 8.629568020383695e-05, "loss": 0.1511, "step": 8694 }, { "epoch": 0.56, "learning_rate": 8.627510969070343e-05, "loss": 0.1429, "step": 8695 }, { "epoch": 0.56, "learning_rate": 8.625453976946618e-05, "loss": 0.157, "step": 8696 }, { "epoch": 0.56, "learning_rate": 8.623397044101223e-05, "loss": 0.1471, "step": 8697 }, { "epoch": 0.56, "learning_rate": 8.621340170622874e-05, "loss": 0.1445, "step": 8698 }, { "epoch": 0.56, "learning_rate": 8.619283356600269e-05, "loss": 0.1302, "step": 8699 }, { "epoch": 0.56, "learning_rate": 8.617226602122115e-05, "loss": 0.1514, "step": 8700 }, { "epoch": 0.56, "learning_rate": 8.615169907277103e-05, "loss": 0.1369, "step": 8701 }, { "epoch": 0.56, "learning_rate": 8.613113272153937e-05, "loss": 0.1385, "step": 8702 }, { "epoch": 0.56, "learning_rate": 8.611056696841312e-05, "loss": 0.1658, "step": 8703 }, { "epoch": 0.56, "learning_rate": 8.60900018142791e-05, "loss": 0.1549, "step": 8704 }, { "epoch": 0.56, "learning_rate": 8.606943726002428e-05, "loss": 0.1434, "step": 8705 }, { "epoch": 0.56, "learning_rate": 8.604887330653549e-05, "loss": 0.1521, "step": 8706 }, { "epoch": 0.56, "learning_rate": 8.60283099546996e-05, "loss": 0.1672, "step": 8707 }, { "epoch": 0.56, "learning_rate": 8.600774720540336e-05, "loss": 0.1448, "step": 8708 }, { "epoch": 0.56, "learning_rate": 8.598718505953358e-05, "loss": 0.1548, "step": 8709 }, { "epoch": 0.56, "learning_rate": 8.596662351797706e-05, "loss": 0.1404, "step": 8710 }, { "epoch": 0.56, "learning_rate": 8.594606258162046e-05, "loss": 0.1473, "step": 8711 }, { "epoch": 0.56, "learning_rate": 8.592550225135055e-05, "loss": 0.1445, "step": 8712 }, { "epoch": 0.56, "learning_rate": 8.590494252805397e-05, "loss": 0.1654, "step": 8713 }, { "epoch": 0.56, "learning_rate": 8.58843834126174e-05, "loss": 0.1493, "step": 8714 }, { "epoch": 0.56, "learning_rate": 8.586382490592746e-05, "loss": 0.1343, "step": 8715 }, { "epoch": 0.56, "learning_rate": 8.584326700887071e-05, "loss": 0.1544, "step": 8716 }, { "epoch": 0.56, "learning_rate": 8.58227097223338e-05, "loss": 0.1345, "step": 8717 }, { "epoch": 0.56, "learning_rate": 8.580215304720321e-05, "loss": 0.1477, "step": 8718 }, { "epoch": 0.56, "learning_rate": 8.57815969843655e-05, "loss": 0.1545, "step": 8719 }, { "epoch": 0.56, "learning_rate": 8.576104153470713e-05, "loss": 0.1319, "step": 8720 }, { "epoch": 0.56, "learning_rate": 8.574048669911463e-05, "loss": 0.165, "step": 8721 }, { "epoch": 0.56, "learning_rate": 8.571993247847442e-05, "loss": 0.1618, "step": 8722 }, { "epoch": 0.56, "learning_rate": 8.569937887367287e-05, "loss": 0.1497, "step": 8723 }, { "epoch": 0.56, "learning_rate": 8.567882588559642e-05, "loss": 0.1623, "step": 8724 }, { "epoch": 0.56, "learning_rate": 8.565827351513139e-05, "loss": 0.1209, "step": 8725 }, { "epoch": 0.56, "learning_rate": 8.563772176316419e-05, "loss": 0.1532, "step": 8726 }, { "epoch": 0.56, "learning_rate": 8.561717063058104e-05, "loss": 0.1474, "step": 8727 }, { "epoch": 0.56, "learning_rate": 8.55966201182683e-05, "loss": 0.1514, "step": 8728 }, { "epoch": 0.56, "learning_rate": 8.55760702271122e-05, "loss": 0.1619, "step": 8729 }, { "epoch": 0.56, "learning_rate": 8.555552095799894e-05, "loss": 0.1386, "step": 8730 }, { "epoch": 0.56, "learning_rate": 8.553497231181475e-05, "loss": 0.1443, "step": 8731 }, { "epoch": 0.56, "learning_rate": 8.551442428944578e-05, "loss": 0.1523, "step": 8732 }, { "epoch": 0.56, "learning_rate": 8.549387689177824e-05, "loss": 0.1568, "step": 8733 }, { "epoch": 0.56, "learning_rate": 8.547333011969818e-05, "loss": 0.1475, "step": 8734 }, { "epoch": 0.56, "learning_rate": 8.545278397409175e-05, "loss": 0.1333, "step": 8735 }, { "epoch": 0.56, "learning_rate": 8.5432238455845e-05, "loss": 0.1608, "step": 8736 }, { "epoch": 0.56, "learning_rate": 8.541169356584394e-05, "loss": 0.1549, "step": 8737 }, { "epoch": 0.56, "learning_rate": 8.539114930497463e-05, "loss": 0.1362, "step": 8738 }, { "epoch": 0.56, "learning_rate": 8.537060567412302e-05, "loss": 0.1473, "step": 8739 }, { "epoch": 0.56, "learning_rate": 8.535006267417511e-05, "loss": 0.1532, "step": 8740 }, { "epoch": 0.56, "learning_rate": 8.532952030601679e-05, "loss": 0.1281, "step": 8741 }, { "epoch": 0.56, "learning_rate": 8.530897857053399e-05, "loss": 0.1615, "step": 8742 }, { "epoch": 0.56, "learning_rate": 8.528843746861262e-05, "loss": 0.1362, "step": 8743 }, { "epoch": 0.56, "learning_rate": 8.526789700113843e-05, "loss": 0.1431, "step": 8744 }, { "epoch": 0.56, "learning_rate": 8.524735716899733e-05, "loss": 0.1452, "step": 8745 }, { "epoch": 0.56, "learning_rate": 8.522681797307508e-05, "loss": 0.1676, "step": 8746 }, { "epoch": 0.56, "learning_rate": 8.520627941425749e-05, "loss": 0.1502, "step": 8747 }, { "epoch": 0.56, "learning_rate": 8.518574149343023e-05, "loss": 0.1615, "step": 8748 }, { "epoch": 0.56, "learning_rate": 8.516520421147908e-05, "loss": 0.1494, "step": 8749 }, { "epoch": 0.56, "learning_rate": 8.514466756928973e-05, "loss": 0.126, "step": 8750 }, { "epoch": 0.56, "learning_rate": 8.512413156774777e-05, "loss": 0.1513, "step": 8751 }, { "epoch": 0.56, "learning_rate": 8.510359620773886e-05, "loss": 0.1533, "step": 8752 }, { "epoch": 0.56, "learning_rate": 8.508306149014862e-05, "loss": 0.1467, "step": 8753 }, { "epoch": 0.56, "learning_rate": 8.506252741586264e-05, "loss": 0.1258, "step": 8754 }, { "epoch": 0.56, "learning_rate": 8.504199398576642e-05, "loss": 0.1479, "step": 8755 }, { "epoch": 0.56, "learning_rate": 8.50214612007455e-05, "loss": 0.1735, "step": 8756 }, { "epoch": 0.56, "learning_rate": 8.500092906168542e-05, "loss": 0.1651, "step": 8757 }, { "epoch": 0.56, "learning_rate": 8.498039756947154e-05, "loss": 0.1321, "step": 8758 }, { "epoch": 0.56, "learning_rate": 8.495986672498938e-05, "loss": 0.1564, "step": 8759 }, { "epoch": 0.56, "learning_rate": 8.493933652912432e-05, "loss": 0.1729, "step": 8760 }, { "epoch": 0.56, "learning_rate": 8.491880698276177e-05, "loss": 0.161, "step": 8761 }, { "epoch": 0.56, "learning_rate": 8.489827808678701e-05, "loss": 0.1484, "step": 8762 }, { "epoch": 0.56, "learning_rate": 8.487774984208546e-05, "loss": 0.1418, "step": 8763 }, { "epoch": 0.56, "learning_rate": 8.485722224954237e-05, "loss": 0.1627, "step": 8764 }, { "epoch": 0.56, "learning_rate": 8.483669531004294e-05, "loss": 0.1436, "step": 8765 }, { "epoch": 0.56, "learning_rate": 8.481616902447254e-05, "loss": 0.1363, "step": 8766 }, { "epoch": 0.56, "learning_rate": 8.479564339371626e-05, "loss": 0.1416, "step": 8767 }, { "epoch": 0.56, "learning_rate": 8.477511841865937e-05, "loss": 0.1517, "step": 8768 }, { "epoch": 0.56, "learning_rate": 8.475459410018698e-05, "loss": 0.1666, "step": 8769 }, { "epoch": 0.56, "learning_rate": 8.473407043918426e-05, "loss": 0.1601, "step": 8770 }, { "epoch": 0.56, "learning_rate": 8.471354743653627e-05, "loss": 0.1387, "step": 8771 }, { "epoch": 0.56, "learning_rate": 8.469302509312807e-05, "loss": 0.1479, "step": 8772 }, { "epoch": 0.56, "learning_rate": 8.467250340984477e-05, "loss": 0.1282, "step": 8773 }, { "epoch": 0.56, "learning_rate": 8.465198238757126e-05, "loss": 0.1656, "step": 8774 }, { "epoch": 0.56, "learning_rate": 8.463146202719264e-05, "loss": 0.1535, "step": 8775 }, { "epoch": 0.56, "learning_rate": 8.461094232959381e-05, "loss": 0.1463, "step": 8776 }, { "epoch": 0.56, "learning_rate": 8.459042329565974e-05, "loss": 0.1539, "step": 8777 }, { "epoch": 0.56, "learning_rate": 8.45699049262753e-05, "loss": 0.1472, "step": 8778 }, { "epoch": 0.56, "learning_rate": 8.454938722232531e-05, "loss": 0.1372, "step": 8779 }, { "epoch": 0.56, "learning_rate": 8.452887018469473e-05, "loss": 0.1592, "step": 8780 }, { "epoch": 0.56, "learning_rate": 8.450835381426825e-05, "loss": 0.1296, "step": 8781 }, { "epoch": 0.56, "learning_rate": 8.448783811193073e-05, "loss": 0.1501, "step": 8782 }, { "epoch": 0.56, "learning_rate": 8.446732307856694e-05, "loss": 0.1478, "step": 8783 }, { "epoch": 0.56, "learning_rate": 8.444680871506152e-05, "loss": 0.1515, "step": 8784 }, { "epoch": 0.56, "learning_rate": 8.442629502229924e-05, "loss": 0.1576, "step": 8785 }, { "epoch": 0.56, "learning_rate": 8.440578200116471e-05, "loss": 0.168, "step": 8786 }, { "epoch": 0.56, "learning_rate": 8.438526965254267e-05, "loss": 0.1278, "step": 8787 }, { "epoch": 0.56, "learning_rate": 8.436475797731762e-05, "loss": 0.1529, "step": 8788 }, { "epoch": 0.56, "learning_rate": 8.43442469763742e-05, "loss": 0.135, "step": 8789 }, { "epoch": 0.56, "learning_rate": 8.432373665059698e-05, "loss": 0.1402, "step": 8790 }, { "epoch": 0.56, "learning_rate": 8.430322700087041e-05, "loss": 0.131, "step": 8791 }, { "epoch": 0.56, "learning_rate": 8.428271802807904e-05, "loss": 0.1438, "step": 8792 }, { "epoch": 0.56, "learning_rate": 8.42622097331073e-05, "loss": 0.1371, "step": 8793 }, { "epoch": 0.56, "learning_rate": 8.424170211683972e-05, "loss": 0.1416, "step": 8794 }, { "epoch": 0.56, "learning_rate": 8.422119518016056e-05, "loss": 0.1531, "step": 8795 }, { "epoch": 0.56, "learning_rate": 8.420068892395432e-05, "loss": 0.1363, "step": 8796 }, { "epoch": 0.56, "learning_rate": 8.41801833491053e-05, "loss": 0.1367, "step": 8797 }, { "epoch": 0.56, "learning_rate": 8.415967845649777e-05, "loss": 0.1523, "step": 8798 }, { "epoch": 0.56, "learning_rate": 8.413917424701612e-05, "loss": 0.1407, "step": 8799 }, { "epoch": 0.56, "learning_rate": 8.411867072154452e-05, "loss": 0.1477, "step": 8800 }, { "epoch": 0.56, "learning_rate": 8.409816788096727e-05, "loss": 0.1666, "step": 8801 }, { "epoch": 0.56, "learning_rate": 8.40776657261685e-05, "loss": 0.1457, "step": 8802 }, { "epoch": 0.56, "learning_rate": 8.405716425803244e-05, "loss": 0.1471, "step": 8803 }, { "epoch": 0.56, "learning_rate": 8.403666347744324e-05, "loss": 0.1604, "step": 8804 }, { "epoch": 0.56, "learning_rate": 8.40161633852849e-05, "loss": 0.1374, "step": 8805 }, { "epoch": 0.56, "learning_rate": 8.399566398244166e-05, "loss": 0.1464, "step": 8806 }, { "epoch": 0.56, "learning_rate": 8.397516526979745e-05, "loss": 0.1655, "step": 8807 }, { "epoch": 0.56, "learning_rate": 8.395466724823635e-05, "loss": 0.1469, "step": 8808 }, { "epoch": 0.56, "learning_rate": 8.393416991864233e-05, "loss": 0.1459, "step": 8809 }, { "epoch": 0.56, "learning_rate": 8.391367328189941e-05, "loss": 0.1586, "step": 8810 }, { "epoch": 0.56, "learning_rate": 8.389317733889145e-05, "loss": 0.1766, "step": 8811 }, { "epoch": 0.57, "learning_rate": 8.387268209050235e-05, "loss": 0.1448, "step": 8812 }, { "epoch": 0.57, "learning_rate": 8.385218753761606e-05, "loss": 0.1578, "step": 8813 }, { "epoch": 0.57, "learning_rate": 8.383169368111634e-05, "loss": 0.1592, "step": 8814 }, { "epoch": 0.57, "learning_rate": 8.381120052188706e-05, "loss": 0.1475, "step": 8815 }, { "epoch": 0.57, "learning_rate": 8.379070806081195e-05, "loss": 0.1578, "step": 8816 }, { "epoch": 0.57, "learning_rate": 8.377021629877484e-05, "loss": 0.1461, "step": 8817 }, { "epoch": 0.57, "learning_rate": 8.374972523665938e-05, "loss": 0.1573, "step": 8818 }, { "epoch": 0.57, "learning_rate": 8.372923487534927e-05, "loss": 0.1388, "step": 8819 }, { "epoch": 0.57, "learning_rate": 8.370874521572824e-05, "loss": 0.1435, "step": 8820 }, { "epoch": 0.57, "learning_rate": 8.368825625867985e-05, "loss": 0.1455, "step": 8821 }, { "epoch": 0.57, "learning_rate": 8.36677680050877e-05, "loss": 0.1641, "step": 8822 }, { "epoch": 0.57, "learning_rate": 8.36472804558354e-05, "loss": 0.138, "step": 8823 }, { "epoch": 0.57, "learning_rate": 8.362679361180652e-05, "loss": 0.1453, "step": 8824 }, { "epoch": 0.57, "learning_rate": 8.36063074738845e-05, "loss": 0.1415, "step": 8825 }, { "epoch": 0.57, "learning_rate": 8.358582204295283e-05, "loss": 0.1539, "step": 8826 }, { "epoch": 0.57, "learning_rate": 8.356533731989502e-05, "loss": 0.1308, "step": 8827 }, { "epoch": 0.57, "learning_rate": 8.354485330559441e-05, "loss": 0.154, "step": 8828 }, { "epoch": 0.57, "learning_rate": 8.352437000093443e-05, "loss": 0.1422, "step": 8829 }, { "epoch": 0.57, "learning_rate": 8.350388740679842e-05, "loss": 0.1432, "step": 8830 }, { "epoch": 0.57, "learning_rate": 8.348340552406978e-05, "loss": 0.1321, "step": 8831 }, { "epoch": 0.57, "learning_rate": 8.34629243536317e-05, "loss": 0.1563, "step": 8832 }, { "epoch": 0.57, "learning_rate": 8.34424438963675e-05, "loss": 0.1442, "step": 8833 }, { "epoch": 0.57, "learning_rate": 8.342196415316044e-05, "loss": 0.1523, "step": 8834 }, { "epoch": 0.57, "learning_rate": 8.340148512489365e-05, "loss": 0.1174, "step": 8835 }, { "epoch": 0.57, "learning_rate": 8.338100681245036e-05, "loss": 0.152, "step": 8836 }, { "epoch": 0.57, "learning_rate": 8.33605292167137e-05, "loss": 0.1562, "step": 8837 }, { "epoch": 0.57, "learning_rate": 8.334005233856681e-05, "loss": 0.1318, "step": 8838 }, { "epoch": 0.57, "learning_rate": 8.331957617889272e-05, "loss": 0.1434, "step": 8839 }, { "epoch": 0.57, "learning_rate": 8.32991007385745e-05, "loss": 0.152, "step": 8840 }, { "epoch": 0.57, "learning_rate": 8.327862601849519e-05, "loss": 0.155, "step": 8841 }, { "epoch": 0.57, "learning_rate": 8.325815201953773e-05, "loss": 0.1367, "step": 8842 }, { "epoch": 0.57, "learning_rate": 8.323767874258513e-05, "loss": 0.146, "step": 8843 }, { "epoch": 0.57, "learning_rate": 8.321720618852025e-05, "loss": 0.136, "step": 8844 }, { "epoch": 0.57, "learning_rate": 8.319673435822608e-05, "loss": 0.1462, "step": 8845 }, { "epoch": 0.57, "learning_rate": 8.317626325258541e-05, "loss": 0.1466, "step": 8846 }, { "epoch": 0.57, "learning_rate": 8.315579287248106e-05, "loss": 0.1474, "step": 8847 }, { "epoch": 0.57, "learning_rate": 8.313532321879591e-05, "loss": 0.1609, "step": 8848 }, { "epoch": 0.57, "learning_rate": 8.311485429241263e-05, "loss": 0.1492, "step": 8849 }, { "epoch": 0.57, "learning_rate": 8.309438609421406e-05, "loss": 0.1725, "step": 8850 }, { "epoch": 0.57, "learning_rate": 8.30739186250828e-05, "loss": 0.1456, "step": 8851 }, { "epoch": 0.57, "learning_rate": 8.305345188590162e-05, "loss": 0.1415, "step": 8852 }, { "epoch": 0.57, "learning_rate": 8.303298587755314e-05, "loss": 0.1636, "step": 8853 }, { "epoch": 0.57, "learning_rate": 8.30125206009199e-05, "loss": 0.1428, "step": 8854 }, { "epoch": 0.57, "learning_rate": 8.299205605688457e-05, "loss": 0.1594, "step": 8855 }, { "epoch": 0.57, "learning_rate": 8.297159224632964e-05, "loss": 0.1512, "step": 8856 }, { "epoch": 0.57, "learning_rate": 8.295112917013769e-05, "loss": 0.1454, "step": 8857 }, { "epoch": 0.57, "learning_rate": 8.293066682919114e-05, "loss": 0.1403, "step": 8858 }, { "epoch": 0.57, "learning_rate": 8.291020522437248e-05, "loss": 0.1466, "step": 8859 }, { "epoch": 0.57, "learning_rate": 8.288974435656416e-05, "loss": 0.1498, "step": 8860 }, { "epoch": 0.57, "learning_rate": 8.286928422664848e-05, "loss": 0.1445, "step": 8861 }, { "epoch": 0.57, "learning_rate": 8.284882483550788e-05, "loss": 0.1439, "step": 8862 }, { "epoch": 0.57, "learning_rate": 8.282836618402463e-05, "loss": 0.1532, "step": 8863 }, { "epoch": 0.57, "learning_rate": 8.28079082730811e-05, "loss": 0.1643, "step": 8864 }, { "epoch": 0.57, "learning_rate": 8.278745110355949e-05, "loss": 0.151, "step": 8865 }, { "epoch": 0.57, "learning_rate": 8.276699467634201e-05, "loss": 0.1296, "step": 8866 }, { "epoch": 0.57, "learning_rate": 8.274653899231096e-05, "loss": 0.1463, "step": 8867 }, { "epoch": 0.57, "learning_rate": 8.27260840523484e-05, "loss": 0.1389, "step": 8868 }, { "epoch": 0.57, "learning_rate": 8.270562985733652e-05, "loss": 0.1423, "step": 8869 }, { "epoch": 0.57, "learning_rate": 8.26851764081574e-05, "loss": 0.145, "step": 8870 }, { "epoch": 0.57, "learning_rate": 8.266472370569316e-05, "loss": 0.1368, "step": 8871 }, { "epoch": 0.57, "learning_rate": 8.264427175082576e-05, "loss": 0.147, "step": 8872 }, { "epoch": 0.57, "learning_rate": 8.262382054443725e-05, "loss": 0.1395, "step": 8873 }, { "epoch": 0.57, "learning_rate": 8.260337008740963e-05, "loss": 0.1359, "step": 8874 }, { "epoch": 0.57, "learning_rate": 8.258292038062477e-05, "loss": 0.1615, "step": 8875 }, { "epoch": 0.57, "learning_rate": 8.256247142496464e-05, "loss": 0.1379, "step": 8876 }, { "epoch": 0.57, "learning_rate": 8.254202322131109e-05, "loss": 0.1287, "step": 8877 }, { "epoch": 0.57, "learning_rate": 8.2521575770546e-05, "loss": 0.1489, "step": 8878 }, { "epoch": 0.57, "learning_rate": 8.250112907355112e-05, "loss": 0.1262, "step": 8879 }, { "epoch": 0.57, "learning_rate": 8.248068313120825e-05, "loss": 0.1384, "step": 8880 }, { "epoch": 0.57, "learning_rate": 8.246023794439922e-05, "loss": 0.1355, "step": 8881 }, { "epoch": 0.57, "learning_rate": 8.243979351400559e-05, "loss": 0.142, "step": 8882 }, { "epoch": 0.57, "learning_rate": 8.241934984090916e-05, "loss": 0.1632, "step": 8883 }, { "epoch": 0.57, "learning_rate": 8.239890692599152e-05, "loss": 0.1457, "step": 8884 }, { "epoch": 0.57, "learning_rate": 8.237846477013436e-05, "loss": 0.1292, "step": 8885 }, { "epoch": 0.57, "learning_rate": 8.235802337421919e-05, "loss": 0.1323, "step": 8886 }, { "epoch": 0.57, "learning_rate": 8.233758273912755e-05, "loss": 0.1329, "step": 8887 }, { "epoch": 0.57, "learning_rate": 8.231714286574104e-05, "loss": 0.1287, "step": 8888 }, { "epoch": 0.57, "learning_rate": 8.229670375494105e-05, "loss": 0.1595, "step": 8889 }, { "epoch": 0.57, "learning_rate": 8.22762654076091e-05, "loss": 0.1507, "step": 8890 }, { "epoch": 0.57, "learning_rate": 8.225582782462656e-05, "loss": 0.1698, "step": 8891 }, { "epoch": 0.57, "learning_rate": 8.223539100687487e-05, "loss": 0.1577, "step": 8892 }, { "epoch": 0.57, "learning_rate": 8.22149549552354e-05, "loss": 0.1512, "step": 8893 }, { "epoch": 0.57, "learning_rate": 8.219451967058935e-05, "loss": 0.1688, "step": 8894 }, { "epoch": 0.57, "learning_rate": 8.217408515381812e-05, "loss": 0.1253, "step": 8895 }, { "epoch": 0.57, "learning_rate": 8.215365140580291e-05, "loss": 0.146, "step": 8896 }, { "epoch": 0.57, "learning_rate": 8.2133218427425e-05, "loss": 0.1199, "step": 8897 }, { "epoch": 0.57, "learning_rate": 8.211278621956549e-05, "loss": 0.1432, "step": 8898 }, { "epoch": 0.57, "learning_rate": 8.20923547831056e-05, "loss": 0.14, "step": 8899 }, { "epoch": 0.57, "learning_rate": 8.207192411892646e-05, "loss": 0.1418, "step": 8900 }, { "epoch": 0.57, "learning_rate": 8.205149422790909e-05, "loss": 0.1713, "step": 8901 }, { "epoch": 0.57, "learning_rate": 8.203106511093459e-05, "loss": 0.1364, "step": 8902 }, { "epoch": 0.57, "learning_rate": 8.201063676888397e-05, "loss": 0.1359, "step": 8903 }, { "epoch": 0.57, "learning_rate": 8.199020920263825e-05, "loss": 0.1319, "step": 8904 }, { "epoch": 0.57, "learning_rate": 8.196978241307832e-05, "loss": 0.1468, "step": 8905 }, { "epoch": 0.57, "learning_rate": 8.194935640108517e-05, "loss": 0.1619, "step": 8906 }, { "epoch": 0.57, "learning_rate": 8.192893116753966e-05, "loss": 0.1493, "step": 8907 }, { "epoch": 0.57, "learning_rate": 8.19085067133226e-05, "loss": 0.1479, "step": 8908 }, { "epoch": 0.57, "learning_rate": 8.188808303931487e-05, "loss": 0.1446, "step": 8909 }, { "epoch": 0.57, "learning_rate": 8.186766014639723e-05, "loss": 0.1453, "step": 8910 }, { "epoch": 0.57, "learning_rate": 8.184723803545047e-05, "loss": 0.1554, "step": 8911 }, { "epoch": 0.57, "learning_rate": 8.182681670735523e-05, "loss": 0.1376, "step": 8912 }, { "epoch": 0.57, "learning_rate": 8.180639616299226e-05, "loss": 0.1322, "step": 8913 }, { "epoch": 0.57, "learning_rate": 8.178597640324221e-05, "loss": 0.1575, "step": 8914 }, { "epoch": 0.57, "learning_rate": 8.176555742898565e-05, "loss": 0.1566, "step": 8915 }, { "epoch": 0.57, "learning_rate": 8.174513924110322e-05, "loss": 0.1463, "step": 8916 }, { "epoch": 0.57, "learning_rate": 8.172472184047542e-05, "loss": 0.1382, "step": 8917 }, { "epoch": 0.57, "learning_rate": 8.170430522798282e-05, "loss": 0.1375, "step": 8918 }, { "epoch": 0.57, "learning_rate": 8.168388940450584e-05, "loss": 0.1473, "step": 8919 }, { "epoch": 0.57, "learning_rate": 8.166347437092498e-05, "loss": 0.1556, "step": 8920 }, { "epoch": 0.57, "learning_rate": 8.164306012812065e-05, "loss": 0.1438, "step": 8921 }, { "epoch": 0.57, "learning_rate": 8.162264667697317e-05, "loss": 0.1413, "step": 8922 }, { "epoch": 0.57, "learning_rate": 8.160223401836295e-05, "loss": 0.1533, "step": 8923 }, { "epoch": 0.57, "learning_rate": 8.158182215317027e-05, "loss": 0.1745, "step": 8924 }, { "epoch": 0.57, "learning_rate": 8.156141108227544e-05, "loss": 0.1425, "step": 8925 }, { "epoch": 0.57, "learning_rate": 8.154100080655865e-05, "loss": 0.1347, "step": 8926 }, { "epoch": 0.57, "learning_rate": 8.152059132690017e-05, "loss": 0.156, "step": 8927 }, { "epoch": 0.57, "learning_rate": 8.150018264418015e-05, "loss": 0.1404, "step": 8928 }, { "epoch": 0.57, "learning_rate": 8.147977475927867e-05, "loss": 0.1393, "step": 8929 }, { "epoch": 0.57, "learning_rate": 8.145936767307593e-05, "loss": 0.15, "step": 8930 }, { "epoch": 0.57, "learning_rate": 8.143896138645192e-05, "loss": 0.1461, "step": 8931 }, { "epoch": 0.57, "learning_rate": 8.141855590028678e-05, "loss": 0.1622, "step": 8932 }, { "epoch": 0.57, "learning_rate": 8.139815121546036e-05, "loss": 0.1379, "step": 8933 }, { "epoch": 0.57, "learning_rate": 8.137774733285278e-05, "loss": 0.1492, "step": 8934 }, { "epoch": 0.57, "learning_rate": 8.135734425334393e-05, "loss": 0.1552, "step": 8935 }, { "epoch": 0.57, "learning_rate": 8.133694197781359e-05, "loss": 0.1487, "step": 8936 }, { "epoch": 0.57, "learning_rate": 8.131654050714182e-05, "loss": 0.1337, "step": 8937 }, { "epoch": 0.57, "learning_rate": 8.129613984220828e-05, "loss": 0.1389, "step": 8938 }, { "epoch": 0.57, "learning_rate": 8.127573998389286e-05, "loss": 0.1448, "step": 8939 }, { "epoch": 0.57, "learning_rate": 8.12553409330753e-05, "loss": 0.1544, "step": 8940 }, { "epoch": 0.57, "learning_rate": 8.12349426906353e-05, "loss": 0.1583, "step": 8941 }, { "epoch": 0.57, "learning_rate": 8.121454525745256e-05, "loss": 0.1309, "step": 8942 }, { "epoch": 0.57, "learning_rate": 8.119414863440674e-05, "loss": 0.1611, "step": 8943 }, { "epoch": 0.57, "learning_rate": 8.117375282237749e-05, "loss": 0.1383, "step": 8944 }, { "epoch": 0.57, "learning_rate": 8.115335782224434e-05, "loss": 0.1307, "step": 8945 }, { "epoch": 0.57, "learning_rate": 8.113296363488687e-05, "loss": 0.1317, "step": 8946 }, { "epoch": 0.57, "learning_rate": 8.11125702611846e-05, "loss": 0.1793, "step": 8947 }, { "epoch": 0.57, "learning_rate": 8.109217770201698e-05, "loss": 0.1372, "step": 8948 }, { "epoch": 0.57, "learning_rate": 8.107178595826348e-05, "loss": 0.1851, "step": 8949 }, { "epoch": 0.57, "learning_rate": 8.105139503080349e-05, "loss": 0.1407, "step": 8950 }, { "epoch": 0.57, "learning_rate": 8.103100492051645e-05, "loss": 0.1442, "step": 8951 }, { "epoch": 0.57, "learning_rate": 8.101061562828159e-05, "loss": 0.1551, "step": 8952 }, { "epoch": 0.57, "learning_rate": 8.099022715497828e-05, "loss": 0.1507, "step": 8953 }, { "epoch": 0.57, "learning_rate": 8.096983950148581e-05, "loss": 0.13, "step": 8954 }, { "epoch": 0.57, "learning_rate": 8.094945266868334e-05, "loss": 0.1402, "step": 8955 }, { "epoch": 0.57, "learning_rate": 8.092906665745012e-05, "loss": 0.1582, "step": 8956 }, { "epoch": 0.57, "learning_rate": 8.090868146866529e-05, "loss": 0.1413, "step": 8957 }, { "epoch": 0.57, "learning_rate": 8.088829710320803e-05, "loss": 0.1288, "step": 8958 }, { "epoch": 0.57, "learning_rate": 8.086791356195735e-05, "loss": 0.145, "step": 8959 }, { "epoch": 0.57, "learning_rate": 8.084753084579235e-05, "loss": 0.1437, "step": 8960 }, { "epoch": 0.57, "learning_rate": 8.082714895559207e-05, "loss": 0.1535, "step": 8961 }, { "epoch": 0.57, "learning_rate": 8.080676789223543e-05, "loss": 0.1709, "step": 8962 }, { "epoch": 0.57, "learning_rate": 8.078638765660145e-05, "loss": 0.1549, "step": 8963 }, { "epoch": 0.57, "learning_rate": 8.076600824956898e-05, "loss": 0.148, "step": 8964 }, { "epoch": 0.57, "learning_rate": 8.074562967201695e-05, "loss": 0.1626, "step": 8965 }, { "epoch": 0.57, "learning_rate": 8.072525192482417e-05, "loss": 0.1328, "step": 8966 }, { "epoch": 0.57, "learning_rate": 8.070487500886946e-05, "loss": 0.1524, "step": 8967 }, { "epoch": 0.58, "learning_rate": 8.06844989250316e-05, "loss": 0.1318, "step": 8968 }, { "epoch": 0.58, "learning_rate": 8.066412367418926e-05, "loss": 0.1462, "step": 8969 }, { "epoch": 0.58, "learning_rate": 8.064374925722123e-05, "loss": 0.1595, "step": 8970 }, { "epoch": 0.58, "learning_rate": 8.062337567500608e-05, "loss": 0.1469, "step": 8971 }, { "epoch": 0.58, "learning_rate": 8.060300292842256e-05, "loss": 0.1476, "step": 8972 }, { "epoch": 0.58, "learning_rate": 8.058263101834911e-05, "loss": 0.1355, "step": 8973 }, { "epoch": 0.58, "learning_rate": 8.05622599456644e-05, "loss": 0.1589, "step": 8974 }, { "epoch": 0.58, "learning_rate": 8.054188971124694e-05, "loss": 0.1579, "step": 8975 }, { "epoch": 0.58, "learning_rate": 8.052152031597512e-05, "loss": 0.1579, "step": 8976 }, { "epoch": 0.58, "learning_rate": 8.050115176072747e-05, "loss": 0.1418, "step": 8977 }, { "epoch": 0.58, "learning_rate": 8.048078404638234e-05, "loss": 0.1389, "step": 8978 }, { "epoch": 0.58, "learning_rate": 8.046041717381817e-05, "loss": 0.1554, "step": 8979 }, { "epoch": 0.58, "learning_rate": 8.044005114391324e-05, "loss": 0.1515, "step": 8980 }, { "epoch": 0.58, "learning_rate": 8.041968595754592e-05, "loss": 0.1556, "step": 8981 }, { "epoch": 0.58, "learning_rate": 8.03993216155944e-05, "loss": 0.119, "step": 8982 }, { "epoch": 0.58, "learning_rate": 8.037895811893691e-05, "loss": 0.1471, "step": 8983 }, { "epoch": 0.58, "learning_rate": 8.035859546845172e-05, "loss": 0.1283, "step": 8984 }, { "epoch": 0.58, "learning_rate": 8.033823366501688e-05, "loss": 0.1467, "step": 8985 }, { "epoch": 0.58, "learning_rate": 8.031787270951058e-05, "loss": 0.1572, "step": 8986 }, { "epoch": 0.58, "learning_rate": 8.029751260281087e-05, "loss": 0.1386, "step": 8987 }, { "epoch": 0.58, "learning_rate": 8.027715334579584e-05, "loss": 0.1474, "step": 8988 }, { "epoch": 0.58, "learning_rate": 8.025679493934346e-05, "loss": 0.1577, "step": 8989 }, { "epoch": 0.58, "learning_rate": 8.023643738433167e-05, "loss": 0.1372, "step": 8990 }, { "epoch": 0.58, "learning_rate": 8.021608068163849e-05, "loss": 0.1397, "step": 8991 }, { "epoch": 0.58, "learning_rate": 8.019572483214171e-05, "loss": 0.148, "step": 8992 }, { "epoch": 0.58, "learning_rate": 8.017536983671928e-05, "loss": 0.1422, "step": 8993 }, { "epoch": 0.58, "learning_rate": 8.015501569624898e-05, "loss": 0.1497, "step": 8994 }, { "epoch": 0.58, "learning_rate": 8.013466241160865e-05, "loss": 0.1297, "step": 8995 }, { "epoch": 0.58, "learning_rate": 8.011430998367599e-05, "loss": 0.1703, "step": 8996 }, { "epoch": 0.58, "learning_rate": 8.00939584133287e-05, "loss": 0.1416, "step": 8997 }, { "epoch": 0.58, "learning_rate": 8.007360770144452e-05, "loss": 0.1399, "step": 8998 }, { "epoch": 0.58, "learning_rate": 8.005325784890103e-05, "loss": 0.1551, "step": 8999 }, { "epoch": 0.58, "learning_rate": 8.003290885657586e-05, "loss": 0.1454, "step": 9000 }, { "epoch": 0.58, "learning_rate": 8.001256072534656e-05, "loss": 0.1558, "step": 9001 }, { "epoch": 0.58, "learning_rate": 7.999221345609071e-05, "loss": 0.1633, "step": 9002 }, { "epoch": 0.58, "learning_rate": 7.997186704968575e-05, "loss": 0.1335, "step": 9003 }, { "epoch": 0.58, "learning_rate": 7.995152150700911e-05, "loss": 0.1482, "step": 9004 }, { "epoch": 0.58, "learning_rate": 7.99311768289383e-05, "loss": 0.1551, "step": 9005 }, { "epoch": 0.58, "learning_rate": 7.991083301635059e-05, "loss": 0.1591, "step": 9006 }, { "epoch": 0.58, "learning_rate": 7.98904900701234e-05, "loss": 0.1565, "step": 9007 }, { "epoch": 0.58, "learning_rate": 7.987014799113397e-05, "loss": 0.1285, "step": 9008 }, { "epoch": 0.58, "learning_rate": 7.984980678025968e-05, "loss": 0.1389, "step": 9009 }, { "epoch": 0.58, "learning_rate": 7.982946643837765e-05, "loss": 0.1568, "step": 9010 }, { "epoch": 0.58, "learning_rate": 7.98091269663651e-05, "loss": 0.1511, "step": 9011 }, { "epoch": 0.58, "learning_rate": 7.978878836509924e-05, "loss": 0.1474, "step": 9012 }, { "epoch": 0.58, "learning_rate": 7.976845063545708e-05, "loss": 0.1383, "step": 9013 }, { "epoch": 0.58, "learning_rate": 7.97481137783158e-05, "loss": 0.1617, "step": 9014 }, { "epoch": 0.58, "learning_rate": 7.972777779455239e-05, "loss": 0.1582, "step": 9015 }, { "epoch": 0.58, "learning_rate": 7.97074426850439e-05, "loss": 0.1511, "step": 9016 }, { "epoch": 0.58, "learning_rate": 7.968710845066724e-05, "loss": 0.157, "step": 9017 }, { "epoch": 0.58, "learning_rate": 7.966677509229936e-05, "loss": 0.1382, "step": 9018 }, { "epoch": 0.58, "learning_rate": 7.96464426108172e-05, "loss": 0.1396, "step": 9019 }, { "epoch": 0.58, "learning_rate": 7.962611100709749e-05, "loss": 0.1479, "step": 9020 }, { "epoch": 0.58, "learning_rate": 7.96057802820172e-05, "loss": 0.1489, "step": 9021 }, { "epoch": 0.58, "learning_rate": 7.958545043645303e-05, "loss": 0.1135, "step": 9022 }, { "epoch": 0.58, "learning_rate": 7.956512147128168e-05, "loss": 0.1444, "step": 9023 }, { "epoch": 0.58, "learning_rate": 7.954479338737995e-05, "loss": 0.1598, "step": 9024 }, { "epoch": 0.58, "learning_rate": 7.952446618562438e-05, "loss": 0.1461, "step": 9025 }, { "epoch": 0.58, "learning_rate": 7.950413986689173e-05, "loss": 0.1529, "step": 9026 }, { "epoch": 0.58, "learning_rate": 7.948381443205848e-05, "loss": 0.1481, "step": 9027 }, { "epoch": 0.58, "learning_rate": 7.946348988200128e-05, "loss": 0.1476, "step": 9028 }, { "epoch": 0.58, "learning_rate": 7.944316621759655e-05, "loss": 0.1403, "step": 9029 }, { "epoch": 0.58, "learning_rate": 7.942284343972079e-05, "loss": 0.167, "step": 9030 }, { "epoch": 0.58, "learning_rate": 7.940252154925048e-05, "loss": 0.1406, "step": 9031 }, { "epoch": 0.58, "learning_rate": 7.938220054706194e-05, "loss": 0.1312, "step": 9032 }, { "epoch": 0.58, "learning_rate": 7.936188043403159e-05, "loss": 0.1419, "step": 9033 }, { "epoch": 0.58, "learning_rate": 7.93415612110357e-05, "loss": 0.1593, "step": 9034 }, { "epoch": 0.58, "learning_rate": 7.932124287895064e-05, "loss": 0.1363, "step": 9035 }, { "epoch": 0.58, "learning_rate": 7.930092543865257e-05, "loss": 0.1292, "step": 9036 }, { "epoch": 0.58, "learning_rate": 7.92806088910177e-05, "loss": 0.1656, "step": 9037 }, { "epoch": 0.58, "learning_rate": 7.926029323692225e-05, "loss": 0.1355, "step": 9038 }, { "epoch": 0.58, "learning_rate": 7.923997847724227e-05, "loss": 0.149, "step": 9039 }, { "epoch": 0.58, "learning_rate": 7.921966461285393e-05, "loss": 0.1527, "step": 9040 }, { "epoch": 0.58, "learning_rate": 7.91993516446332e-05, "loss": 0.1429, "step": 9041 }, { "epoch": 0.58, "learning_rate": 7.91790395734562e-05, "loss": 0.149, "step": 9042 }, { "epoch": 0.58, "learning_rate": 7.915872840019882e-05, "loss": 0.1197, "step": 9043 }, { "epoch": 0.58, "learning_rate": 7.913841812573698e-05, "loss": 0.1354, "step": 9044 }, { "epoch": 0.58, "learning_rate": 7.911810875094665e-05, "loss": 0.1378, "step": 9045 }, { "epoch": 0.58, "learning_rate": 7.909780027670362e-05, "loss": 0.1361, "step": 9046 }, { "epoch": 0.58, "learning_rate": 7.907749270388374e-05, "loss": 0.1501, "step": 9047 }, { "epoch": 0.58, "learning_rate": 7.905718603336279e-05, "loss": 0.1506, "step": 9048 }, { "epoch": 0.58, "learning_rate": 7.903688026601653e-05, "loss": 0.1386, "step": 9049 }, { "epoch": 0.58, "learning_rate": 7.901657540272062e-05, "loss": 0.1542, "step": 9050 }, { "epoch": 0.58, "learning_rate": 7.899627144435071e-05, "loss": 0.1462, "step": 9051 }, { "epoch": 0.58, "learning_rate": 7.89759683917825e-05, "loss": 0.1333, "step": 9052 }, { "epoch": 0.58, "learning_rate": 7.89556662458915e-05, "loss": 0.1381, "step": 9053 }, { "epoch": 0.58, "learning_rate": 7.893536500755329e-05, "loss": 0.1674, "step": 9054 }, { "epoch": 0.58, "learning_rate": 7.891506467764335e-05, "loss": 0.1445, "step": 9055 }, { "epoch": 0.58, "learning_rate": 7.889476525703722e-05, "loss": 0.1377, "step": 9056 }, { "epoch": 0.58, "learning_rate": 7.887446674661024e-05, "loss": 0.1491, "step": 9057 }, { "epoch": 0.58, "learning_rate": 7.885416914723782e-05, "loss": 0.1492, "step": 9058 }, { "epoch": 0.58, "learning_rate": 7.883387245979536e-05, "loss": 0.1502, "step": 9059 }, { "epoch": 0.58, "learning_rate": 7.88135766851581e-05, "loss": 0.1415, "step": 9060 }, { "epoch": 0.58, "learning_rate": 7.879328182420135e-05, "loss": 0.1598, "step": 9061 }, { "epoch": 0.58, "learning_rate": 7.877298787780033e-05, "loss": 0.1439, "step": 9062 }, { "epoch": 0.58, "learning_rate": 7.875269484683025e-05, "loss": 0.1521, "step": 9063 }, { "epoch": 0.58, "learning_rate": 7.873240273216624e-05, "loss": 0.143, "step": 9064 }, { "epoch": 0.58, "learning_rate": 7.871211153468341e-05, "loss": 0.1498, "step": 9065 }, { "epoch": 0.58, "learning_rate": 7.869182125525687e-05, "loss": 0.1625, "step": 9066 }, { "epoch": 0.58, "learning_rate": 7.867153189476159e-05, "loss": 0.1246, "step": 9067 }, { "epoch": 0.58, "learning_rate": 7.865124345407266e-05, "loss": 0.1259, "step": 9068 }, { "epoch": 0.58, "learning_rate": 7.863095593406491e-05, "loss": 0.1703, "step": 9069 }, { "epoch": 0.58, "learning_rate": 7.861066933561336e-05, "loss": 0.1588, "step": 9070 }, { "epoch": 0.58, "learning_rate": 7.859038365959286e-05, "loss": 0.1492, "step": 9071 }, { "epoch": 0.58, "learning_rate": 7.85700989068782e-05, "loss": 0.1413, "step": 9072 }, { "epoch": 0.58, "learning_rate": 7.854981507834421e-05, "loss": 0.1638, "step": 9073 }, { "epoch": 0.58, "learning_rate": 7.852953217486562e-05, "loss": 0.1637, "step": 9074 }, { "epoch": 0.58, "learning_rate": 7.850925019731723e-05, "loss": 0.1625, "step": 9075 }, { "epoch": 0.58, "learning_rate": 7.84889691465736e-05, "loss": 0.1557, "step": 9076 }, { "epoch": 0.58, "learning_rate": 7.846868902350944e-05, "loss": 0.1567, "step": 9077 }, { "epoch": 0.58, "learning_rate": 7.844840982899934e-05, "loss": 0.1267, "step": 9078 }, { "epoch": 0.58, "learning_rate": 7.842813156391779e-05, "loss": 0.1556, "step": 9079 }, { "epoch": 0.58, "learning_rate": 7.840785422913939e-05, "loss": 0.1547, "step": 9080 }, { "epoch": 0.58, "learning_rate": 7.838757782553856e-05, "loss": 0.1544, "step": 9081 }, { "epoch": 0.58, "learning_rate": 7.836730235398979e-05, "loss": 0.1546, "step": 9082 }, { "epoch": 0.58, "learning_rate": 7.83470278153674e-05, "loss": 0.1425, "step": 9083 }, { "epoch": 0.58, "learning_rate": 7.83267542105458e-05, "loss": 0.146, "step": 9084 }, { "epoch": 0.58, "learning_rate": 7.830648154039932e-05, "loss": 0.1584, "step": 9085 }, { "epoch": 0.58, "learning_rate": 7.828620980580217e-05, "loss": 0.144, "step": 9086 }, { "epoch": 0.58, "learning_rate": 7.826593900762862e-05, "loss": 0.1654, "step": 9087 }, { "epoch": 0.58, "learning_rate": 7.824566914675284e-05, "loss": 0.16, "step": 9088 }, { "epoch": 0.58, "learning_rate": 7.822540022404906e-05, "loss": 0.1277, "step": 9089 }, { "epoch": 0.58, "learning_rate": 7.82051322403913e-05, "loss": 0.1499, "step": 9090 }, { "epoch": 0.58, "learning_rate": 7.818486519665368e-05, "loss": 0.1396, "step": 9091 }, { "epoch": 0.58, "learning_rate": 7.816459909371023e-05, "loss": 0.1446, "step": 9092 }, { "epoch": 0.58, "learning_rate": 7.814433393243492e-05, "loss": 0.139, "step": 9093 }, { "epoch": 0.58, "learning_rate": 7.81240697137017e-05, "loss": 0.1341, "step": 9094 }, { "epoch": 0.58, "learning_rate": 7.810380643838449e-05, "loss": 0.1389, "step": 9095 }, { "epoch": 0.58, "learning_rate": 7.80835441073572e-05, "loss": 0.1354, "step": 9096 }, { "epoch": 0.58, "learning_rate": 7.80632827214936e-05, "loss": 0.1485, "step": 9097 }, { "epoch": 0.58, "learning_rate": 7.804302228166746e-05, "loss": 0.1553, "step": 9098 }, { "epoch": 0.58, "learning_rate": 7.802276278875262e-05, "loss": 0.1657, "step": 9099 }, { "epoch": 0.58, "learning_rate": 7.800250424362269e-05, "loss": 0.1349, "step": 9100 }, { "epoch": 0.58, "learning_rate": 7.798224664715139e-05, "loss": 0.1431, "step": 9101 }, { "epoch": 0.58, "learning_rate": 7.796199000021231e-05, "loss": 0.169, "step": 9102 }, { "epoch": 0.58, "learning_rate": 7.794173430367908e-05, "loss": 0.1305, "step": 9103 }, { "epoch": 0.58, "learning_rate": 7.792147955842521e-05, "loss": 0.151, "step": 9104 }, { "epoch": 0.58, "learning_rate": 7.790122576532417e-05, "loss": 0.1388, "step": 9105 }, { "epoch": 0.58, "learning_rate": 7.788097292524951e-05, "loss": 0.1603, "step": 9106 }, { "epoch": 0.58, "learning_rate": 7.786072103907453e-05, "loss": 0.1448, "step": 9107 }, { "epoch": 0.58, "learning_rate": 7.784047010767273e-05, "loss": 0.1361, "step": 9108 }, { "epoch": 0.58, "learning_rate": 7.782022013191736e-05, "loss": 0.1394, "step": 9109 }, { "epoch": 0.58, "learning_rate": 7.779997111268174e-05, "loss": 0.138, "step": 9110 }, { "epoch": 0.58, "learning_rate": 7.777972305083917e-05, "loss": 0.1495, "step": 9111 }, { "epoch": 0.58, "learning_rate": 7.775947594726277e-05, "loss": 0.1423, "step": 9112 }, { "epoch": 0.58, "learning_rate": 7.77392298028258e-05, "loss": 0.1354, "step": 9113 }, { "epoch": 0.58, "learning_rate": 7.771898461840131e-05, "loss": 0.1398, "step": 9114 }, { "epoch": 0.58, "learning_rate": 7.769874039486249e-05, "loss": 0.1499, "step": 9115 }, { "epoch": 0.58, "learning_rate": 7.767849713308228e-05, "loss": 0.1511, "step": 9116 }, { "epoch": 0.58, "learning_rate": 7.765825483393376e-05, "loss": 0.1453, "step": 9117 }, { "epoch": 0.58, "learning_rate": 7.763801349828989e-05, "loss": 0.1548, "step": 9118 }, { "epoch": 0.58, "learning_rate": 7.761777312702353e-05, "loss": 0.1592, "step": 9119 }, { "epoch": 0.58, "learning_rate": 7.759753372100763e-05, "loss": 0.1594, "step": 9120 }, { "epoch": 0.58, "learning_rate": 7.757729528111498e-05, "loss": 0.1341, "step": 9121 }, { "epoch": 0.58, "learning_rate": 7.755705780821845e-05, "loss": 0.1446, "step": 9122 }, { "epoch": 0.58, "learning_rate": 7.753682130319068e-05, "loss": 0.1522, "step": 9123 }, { "epoch": 0.59, "learning_rate": 7.75165857669045e-05, "loss": 0.1468, "step": 9124 }, { "epoch": 0.59, "learning_rate": 7.749635120023256e-05, "loss": 0.158, "step": 9125 }, { "epoch": 0.59, "learning_rate": 7.747611760404742e-05, "loss": 0.135, "step": 9126 }, { "epoch": 0.59, "learning_rate": 7.745588497922172e-05, "loss": 0.1414, "step": 9127 }, { "epoch": 0.59, "learning_rate": 7.743565332662799e-05, "loss": 0.1312, "step": 9128 }, { "epoch": 0.59, "learning_rate": 7.741542264713881e-05, "loss": 0.1468, "step": 9129 }, { "epoch": 0.59, "learning_rate": 7.739519294162652e-05, "loss": 0.1465, "step": 9130 }, { "epoch": 0.59, "learning_rate": 7.737496421096363e-05, "loss": 0.1352, "step": 9131 }, { "epoch": 0.59, "learning_rate": 7.735473645602251e-05, "loss": 0.1331, "step": 9132 }, { "epoch": 0.59, "learning_rate": 7.733450967767543e-05, "loss": 0.1547, "step": 9133 }, { "epoch": 0.59, "learning_rate": 7.731428387679476e-05, "loss": 0.1302, "step": 9134 }, { "epoch": 0.59, "learning_rate": 7.729405905425272e-05, "loss": 0.1347, "step": 9135 }, { "epoch": 0.59, "learning_rate": 7.727383521092156e-05, "loss": 0.1497, "step": 9136 }, { "epoch": 0.59, "learning_rate": 7.725361234767338e-05, "loss": 0.1566, "step": 9137 }, { "epoch": 0.59, "learning_rate": 7.723339046538035e-05, "loss": 0.1406, "step": 9138 }, { "epoch": 0.59, "learning_rate": 7.721316956491458e-05, "loss": 0.1444, "step": 9139 }, { "epoch": 0.59, "learning_rate": 7.719294964714804e-05, "loss": 0.1274, "step": 9140 }, { "epoch": 0.59, "learning_rate": 7.717273071295278e-05, "loss": 0.1384, "step": 9141 }, { "epoch": 0.59, "learning_rate": 7.715251276320072e-05, "loss": 0.1414, "step": 9142 }, { "epoch": 0.59, "learning_rate": 7.713229579876384e-05, "loss": 0.1328, "step": 9143 }, { "epoch": 0.59, "learning_rate": 7.711207982051394e-05, "loss": 0.1478, "step": 9144 }, { "epoch": 0.59, "learning_rate": 7.70918648293229e-05, "loss": 0.15, "step": 9145 }, { "epoch": 0.59, "learning_rate": 7.70716508260625e-05, "loss": 0.1479, "step": 9146 }, { "epoch": 0.59, "learning_rate": 7.705143781160444e-05, "loss": 0.1497, "step": 9147 }, { "epoch": 0.59, "learning_rate": 7.703122578682046e-05, "loss": 0.1386, "step": 9148 }, { "epoch": 0.59, "learning_rate": 7.70110147525822e-05, "loss": 0.157, "step": 9149 }, { "epoch": 0.59, "learning_rate": 7.699080470976132e-05, "loss": 0.1471, "step": 9150 }, { "epoch": 0.59, "learning_rate": 7.69705956592293e-05, "loss": 0.1505, "step": 9151 }, { "epoch": 0.59, "learning_rate": 7.695038760185781e-05, "loss": 0.1444, "step": 9152 }, { "epoch": 0.59, "learning_rate": 7.693018053851823e-05, "loss": 0.1348, "step": 9153 }, { "epoch": 0.59, "learning_rate": 7.6909974470082e-05, "loss": 0.125, "step": 9154 }, { "epoch": 0.59, "learning_rate": 7.688976939742061e-05, "loss": 0.1458, "step": 9155 }, { "epoch": 0.59, "learning_rate": 7.686956532140533e-05, "loss": 0.1324, "step": 9156 }, { "epoch": 0.59, "learning_rate": 7.684936224290751e-05, "loss": 0.1435, "step": 9157 }, { "epoch": 0.59, "learning_rate": 7.682916016279842e-05, "loss": 0.138, "step": 9158 }, { "epoch": 0.59, "learning_rate": 7.680895908194933e-05, "loss": 0.1385, "step": 9159 }, { "epoch": 0.59, "learning_rate": 7.678875900123137e-05, "loss": 0.1418, "step": 9160 }, { "epoch": 0.59, "learning_rate": 7.676855992151568e-05, "loss": 0.1532, "step": 9161 }, { "epoch": 0.59, "learning_rate": 7.674836184367344e-05, "loss": 0.1521, "step": 9162 }, { "epoch": 0.59, "learning_rate": 7.672816476857559e-05, "loss": 0.1643, "step": 9163 }, { "epoch": 0.59, "learning_rate": 7.670796869709325e-05, "loss": 0.1532, "step": 9164 }, { "epoch": 0.59, "learning_rate": 7.66877736300973e-05, "loss": 0.1858, "step": 9165 }, { "epoch": 0.59, "learning_rate": 7.666757956845878e-05, "loss": 0.1416, "step": 9166 }, { "epoch": 0.59, "learning_rate": 7.664738651304848e-05, "loss": 0.1625, "step": 9167 }, { "epoch": 0.59, "learning_rate": 7.662719446473725e-05, "loss": 0.1388, "step": 9168 }, { "epoch": 0.59, "learning_rate": 7.660700342439595e-05, "loss": 0.1521, "step": 9169 }, { "epoch": 0.59, "learning_rate": 7.658681339289524e-05, "loss": 0.1409, "step": 9170 }, { "epoch": 0.59, "learning_rate": 7.656662437110591e-05, "loss": 0.1352, "step": 9171 }, { "epoch": 0.59, "learning_rate": 7.654643635989858e-05, "loss": 0.1708, "step": 9172 }, { "epoch": 0.59, "learning_rate": 7.652624936014394e-05, "loss": 0.1515, "step": 9173 }, { "epoch": 0.59, "learning_rate": 7.650606337271248e-05, "loss": 0.144, "step": 9174 }, { "epoch": 0.59, "learning_rate": 7.648587839847476e-05, "loss": 0.1338, "step": 9175 }, { "epoch": 0.59, "learning_rate": 7.646569443830134e-05, "loss": 0.1423, "step": 9176 }, { "epoch": 0.59, "learning_rate": 7.644551149306258e-05, "loss": 0.1285, "step": 9177 }, { "epoch": 0.59, "learning_rate": 7.642532956362894e-05, "loss": 0.1359, "step": 9178 }, { "epoch": 0.59, "learning_rate": 7.640514865087077e-05, "loss": 0.1407, "step": 9179 }, { "epoch": 0.59, "learning_rate": 7.638496875565835e-05, "loss": 0.16, "step": 9180 }, { "epoch": 0.59, "learning_rate": 7.6364789878862e-05, "loss": 0.1327, "step": 9181 }, { "epoch": 0.59, "learning_rate": 7.634461202135192e-05, "loss": 0.1261, "step": 9182 }, { "epoch": 0.59, "learning_rate": 7.632443518399834e-05, "loss": 0.1449, "step": 9183 }, { "epoch": 0.59, "learning_rate": 7.630425936767134e-05, "loss": 0.1349, "step": 9184 }, { "epoch": 0.59, "learning_rate": 7.628408457324106e-05, "loss": 0.1387, "step": 9185 }, { "epoch": 0.59, "learning_rate": 7.626391080157756e-05, "loss": 0.1587, "step": 9186 }, { "epoch": 0.59, "learning_rate": 7.624373805355079e-05, "loss": 0.1422, "step": 9187 }, { "epoch": 0.59, "learning_rate": 7.622356633003078e-05, "loss": 0.1439, "step": 9188 }, { "epoch": 0.59, "learning_rate": 7.62033956318874e-05, "loss": 0.1493, "step": 9189 }, { "epoch": 0.59, "learning_rate": 7.618322595999059e-05, "loss": 0.1501, "step": 9190 }, { "epoch": 0.59, "learning_rate": 7.616305731521008e-05, "loss": 0.1486, "step": 9191 }, { "epoch": 0.59, "learning_rate": 7.61428896984158e-05, "loss": 0.1721, "step": 9192 }, { "epoch": 0.59, "learning_rate": 7.61227231104774e-05, "loss": 0.1464, "step": 9193 }, { "epoch": 0.59, "learning_rate": 7.610255755226454e-05, "loss": 0.1454, "step": 9194 }, { "epoch": 0.59, "learning_rate": 7.608239302464699e-05, "loss": 0.1289, "step": 9195 }, { "epoch": 0.59, "learning_rate": 7.606222952849427e-05, "loss": 0.1438, "step": 9196 }, { "epoch": 0.59, "learning_rate": 7.604206706467598e-05, "loss": 0.1395, "step": 9197 }, { "epoch": 0.59, "learning_rate": 7.602190563406162e-05, "loss": 0.1399, "step": 9198 }, { "epoch": 0.59, "learning_rate": 7.600174523752073e-05, "loss": 0.1284, "step": 9199 }, { "epoch": 0.59, "learning_rate": 7.598158587592268e-05, "loss": 0.1447, "step": 9200 }, { "epoch": 0.59, "learning_rate": 7.596142755013686e-05, "loss": 0.1372, "step": 9201 }, { "epoch": 0.59, "learning_rate": 7.594127026103267e-05, "loss": 0.144, "step": 9202 }, { "epoch": 0.59, "learning_rate": 7.59211140094793e-05, "loss": 0.1524, "step": 9203 }, { "epoch": 0.59, "learning_rate": 7.590095879634611e-05, "loss": 0.1431, "step": 9204 }, { "epoch": 0.59, "learning_rate": 7.588080462250224e-05, "loss": 0.1668, "step": 9205 }, { "epoch": 0.59, "learning_rate": 7.586065148881694e-05, "loss": 0.1464, "step": 9206 }, { "epoch": 0.59, "learning_rate": 7.584049939615924e-05, "loss": 0.1265, "step": 9207 }, { "epoch": 0.59, "learning_rate": 7.582034834539822e-05, "loss": 0.1517, "step": 9208 }, { "epoch": 0.59, "learning_rate": 7.580019833740298e-05, "loss": 0.1466, "step": 9209 }, { "epoch": 0.59, "learning_rate": 7.578004937304242e-05, "loss": 0.1537, "step": 9210 }, { "epoch": 0.59, "learning_rate": 7.575990145318554e-05, "loss": 0.1583, "step": 9211 }, { "epoch": 0.59, "learning_rate": 7.57397545787012e-05, "loss": 0.1565, "step": 9212 }, { "epoch": 0.59, "learning_rate": 7.57196087504583e-05, "loss": 0.1321, "step": 9213 }, { "epoch": 0.59, "learning_rate": 7.569946396932557e-05, "loss": 0.1451, "step": 9214 }, { "epoch": 0.59, "learning_rate": 7.56793202361718e-05, "loss": 0.1361, "step": 9215 }, { "epoch": 0.59, "learning_rate": 7.565917755186575e-05, "loss": 0.1254, "step": 9216 }, { "epoch": 0.59, "learning_rate": 7.5639035917276e-05, "loss": 0.1361, "step": 9217 }, { "epoch": 0.59, "learning_rate": 7.561889533327125e-05, "loss": 0.154, "step": 9218 }, { "epoch": 0.59, "learning_rate": 7.559875580072001e-05, "loss": 0.1466, "step": 9219 }, { "epoch": 0.59, "learning_rate": 7.557861732049091e-05, "loss": 0.1561, "step": 9220 }, { "epoch": 0.59, "learning_rate": 7.555847989345235e-05, "loss": 0.1585, "step": 9221 }, { "epoch": 0.59, "learning_rate": 7.553834352047278e-05, "loss": 0.1466, "step": 9222 }, { "epoch": 0.59, "learning_rate": 7.551820820242065e-05, "loss": 0.1519, "step": 9223 }, { "epoch": 0.59, "learning_rate": 7.549807394016425e-05, "loss": 0.1431, "step": 9224 }, { "epoch": 0.59, "learning_rate": 7.547794073457193e-05, "loss": 0.1476, "step": 9225 }, { "epoch": 0.59, "learning_rate": 7.545780858651191e-05, "loss": 0.1436, "step": 9226 }, { "epoch": 0.59, "learning_rate": 7.543767749685247e-05, "loss": 0.1406, "step": 9227 }, { "epoch": 0.59, "learning_rate": 7.54175474664617e-05, "loss": 0.1528, "step": 9228 }, { "epoch": 0.59, "learning_rate": 7.539741849620775e-05, "loss": 0.1516, "step": 9229 }, { "epoch": 0.59, "learning_rate": 7.537729058695875e-05, "loss": 0.1411, "step": 9230 }, { "epoch": 0.59, "learning_rate": 7.535716373958263e-05, "loss": 0.1491, "step": 9231 }, { "epoch": 0.59, "learning_rate": 7.533703795494747e-05, "loss": 0.1415, "step": 9232 }, { "epoch": 0.59, "learning_rate": 7.531691323392112e-05, "loss": 0.1545, "step": 9233 }, { "epoch": 0.59, "learning_rate": 7.529678957737159e-05, "loss": 0.1215, "step": 9234 }, { "epoch": 0.59, "learning_rate": 7.527666698616664e-05, "loss": 0.1294, "step": 9235 }, { "epoch": 0.59, "learning_rate": 7.525654546117407e-05, "loss": 0.1318, "step": 9236 }, { "epoch": 0.59, "learning_rate": 7.523642500326168e-05, "loss": 0.1396, "step": 9237 }, { "epoch": 0.59, "learning_rate": 7.521630561329714e-05, "loss": 0.1437, "step": 9238 }, { "epoch": 0.59, "learning_rate": 7.519618729214818e-05, "loss": 0.1414, "step": 9239 }, { "epoch": 0.59, "learning_rate": 7.517607004068232e-05, "loss": 0.1605, "step": 9240 }, { "epoch": 0.59, "learning_rate": 7.515595385976721e-05, "loss": 0.1523, "step": 9241 }, { "epoch": 0.59, "learning_rate": 7.513583875027037e-05, "loss": 0.1446, "step": 9242 }, { "epoch": 0.59, "learning_rate": 7.511572471305922e-05, "loss": 0.1268, "step": 9243 }, { "epoch": 0.59, "learning_rate": 7.509561174900125e-05, "loss": 0.1413, "step": 9244 }, { "epoch": 0.59, "learning_rate": 7.50754998589638e-05, "loss": 0.1326, "step": 9245 }, { "epoch": 0.59, "learning_rate": 7.50553890438143e-05, "loss": 0.1506, "step": 9246 }, { "epoch": 0.59, "learning_rate": 7.503527930441993e-05, "loss": 0.1522, "step": 9247 }, { "epoch": 0.59, "learning_rate": 7.5015170641648e-05, "loss": 0.1358, "step": 9248 }, { "epoch": 0.59, "learning_rate": 7.499506305636575e-05, "loss": 0.1525, "step": 9249 }, { "epoch": 0.59, "learning_rate": 7.497495654944022e-05, "loss": 0.1558, "step": 9250 }, { "epoch": 0.59, "learning_rate": 7.495485112173861e-05, "loss": 0.157, "step": 9251 }, { "epoch": 0.59, "learning_rate": 7.493474677412794e-05, "loss": 0.128, "step": 9252 }, { "epoch": 0.59, "learning_rate": 7.491464350747527e-05, "loss": 0.1581, "step": 9253 }, { "epoch": 0.59, "learning_rate": 7.489454132264755e-05, "loss": 0.1465, "step": 9254 }, { "epoch": 0.59, "learning_rate": 7.487444022051164e-05, "loss": 0.1332, "step": 9255 }, { "epoch": 0.59, "learning_rate": 7.485434020193453e-05, "loss": 0.1415, "step": 9256 }, { "epoch": 0.59, "learning_rate": 7.483424126778293e-05, "loss": 0.1415, "step": 9257 }, { "epoch": 0.59, "learning_rate": 7.481414341892372e-05, "loss": 0.1452, "step": 9258 }, { "epoch": 0.59, "learning_rate": 7.479404665622354e-05, "loss": 0.1393, "step": 9259 }, { "epoch": 0.59, "learning_rate": 7.47739509805492e-05, "loss": 0.139, "step": 9260 }, { "epoch": 0.59, "learning_rate": 7.475385639276724e-05, "loss": 0.1259, "step": 9261 }, { "epoch": 0.59, "learning_rate": 7.473376289374425e-05, "loss": 0.1461, "step": 9262 }, { "epoch": 0.59, "learning_rate": 7.471367048434688e-05, "loss": 0.1578, "step": 9263 }, { "epoch": 0.59, "learning_rate": 7.46935791654415e-05, "loss": 0.1331, "step": 9264 }, { "epoch": 0.59, "learning_rate": 7.467348893789467e-05, "loss": 0.1504, "step": 9265 }, { "epoch": 0.59, "learning_rate": 7.465339980257273e-05, "loss": 0.1542, "step": 9266 }, { "epoch": 0.59, "learning_rate": 7.463331176034209e-05, "loss": 0.1419, "step": 9267 }, { "epoch": 0.59, "learning_rate": 7.461322481206901e-05, "loss": 0.1617, "step": 9268 }, { "epoch": 0.59, "learning_rate": 7.459313895861977e-05, "loss": 0.1361, "step": 9269 }, { "epoch": 0.59, "learning_rate": 7.457305420086064e-05, "loss": 0.1383, "step": 9270 }, { "epoch": 0.59, "learning_rate": 7.45529705396577e-05, "loss": 0.1289, "step": 9271 }, { "epoch": 0.59, "learning_rate": 7.453288797587714e-05, "loss": 0.1585, "step": 9272 }, { "epoch": 0.59, "learning_rate": 7.451280651038499e-05, "loss": 0.1413, "step": 9273 }, { "epoch": 0.59, "learning_rate": 7.449272614404736e-05, "loss": 0.139, "step": 9274 }, { "epoch": 0.59, "learning_rate": 7.447264687773015e-05, "loss": 0.1559, "step": 9275 }, { "epoch": 0.59, "learning_rate": 7.445256871229928e-05, "loss": 0.1412, "step": 9276 }, { "epoch": 0.59, "learning_rate": 7.443249164862074e-05, "loss": 0.135, "step": 9277 }, { "epoch": 0.59, "learning_rate": 7.441241568756024e-05, "loss": 0.1346, "step": 9278 }, { "epoch": 0.59, "learning_rate": 7.439234082998369e-05, "loss": 0.1292, "step": 9279 }, { "epoch": 0.6, "learning_rate": 7.437226707675673e-05, "loss": 0.1252, "step": 9280 }, { "epoch": 0.6, "learning_rate": 7.435219442874513e-05, "loss": 0.1444, "step": 9281 }, { "epoch": 0.6, "learning_rate": 7.433212288681453e-05, "loss": 0.1417, "step": 9282 }, { "epoch": 0.6, "learning_rate": 7.431205245183047e-05, "loss": 0.1383, "step": 9283 }, { "epoch": 0.6, "learning_rate": 7.429198312465857e-05, "loss": 0.134, "step": 9284 }, { "epoch": 0.6, "learning_rate": 7.427191490616429e-05, "loss": 0.1326, "step": 9285 }, { "epoch": 0.6, "learning_rate": 7.425184779721313e-05, "loss": 0.1322, "step": 9286 }, { "epoch": 0.6, "learning_rate": 7.423178179867045e-05, "loss": 0.142, "step": 9287 }, { "epoch": 0.6, "learning_rate": 7.421171691140166e-05, "loss": 0.1482, "step": 9288 }, { "epoch": 0.6, "learning_rate": 7.419165313627207e-05, "loss": 0.1423, "step": 9289 }, { "epoch": 0.6, "learning_rate": 7.417159047414688e-05, "loss": 0.162, "step": 9290 }, { "epoch": 0.6, "learning_rate": 7.415152892589138e-05, "loss": 0.1436, "step": 9291 }, { "epoch": 0.6, "learning_rate": 7.413146849237068e-05, "loss": 0.1678, "step": 9292 }, { "epoch": 0.6, "learning_rate": 7.411140917444999e-05, "loss": 0.131, "step": 9293 }, { "epoch": 0.6, "learning_rate": 7.409135097299427e-05, "loss": 0.1375, "step": 9294 }, { "epoch": 0.6, "learning_rate": 7.407129388886865e-05, "loss": 0.1499, "step": 9295 }, { "epoch": 0.6, "learning_rate": 7.405123792293804e-05, "loss": 0.152, "step": 9296 }, { "epoch": 0.6, "learning_rate": 7.403118307606738e-05, "loss": 0.129, "step": 9297 }, { "epoch": 0.6, "learning_rate": 7.401112934912156e-05, "loss": 0.143, "step": 9298 }, { "epoch": 0.6, "learning_rate": 7.399107674296541e-05, "loss": 0.1627, "step": 9299 }, { "epoch": 0.6, "learning_rate": 7.397102525846374e-05, "loss": 0.1413, "step": 9300 }, { "epoch": 0.6, "learning_rate": 7.395097489648122e-05, "loss": 0.1508, "step": 9301 }, { "epoch": 0.6, "learning_rate": 7.39309256578826e-05, "loss": 0.1272, "step": 9302 }, { "epoch": 0.6, "learning_rate": 7.391087754353253e-05, "loss": 0.1504, "step": 9303 }, { "epoch": 0.6, "learning_rate": 7.38908305542955e-05, "loss": 0.1391, "step": 9304 }, { "epoch": 0.6, "learning_rate": 7.387078469103614e-05, "loss": 0.1686, "step": 9305 }, { "epoch": 0.6, "learning_rate": 7.38507399546189e-05, "loss": 0.1482, "step": 9306 }, { "epoch": 0.6, "learning_rate": 7.38306963459083e-05, "loss": 0.1318, "step": 9307 }, { "epoch": 0.6, "learning_rate": 7.381065386576864e-05, "loss": 0.1539, "step": 9308 }, { "epoch": 0.6, "learning_rate": 7.379061251506431e-05, "loss": 0.1413, "step": 9309 }, { "epoch": 0.6, "learning_rate": 7.377057229465964e-05, "loss": 0.1423, "step": 9310 }, { "epoch": 0.6, "learning_rate": 7.37505332054188e-05, "loss": 0.1507, "step": 9311 }, { "epoch": 0.6, "learning_rate": 7.373049524820605e-05, "loss": 0.1385, "step": 9312 }, { "epoch": 0.6, "learning_rate": 7.371045842388552e-05, "loss": 0.1394, "step": 9313 }, { "epoch": 0.6, "learning_rate": 7.369042273332136e-05, "loss": 0.1379, "step": 9314 }, { "epoch": 0.6, "learning_rate": 7.367038817737753e-05, "loss": 0.1448, "step": 9315 }, { "epoch": 0.6, "learning_rate": 7.365035475691812e-05, "loss": 0.1309, "step": 9316 }, { "epoch": 0.6, "learning_rate": 7.363032247280708e-05, "loss": 0.1414, "step": 9317 }, { "epoch": 0.6, "learning_rate": 7.361029132590827e-05, "loss": 0.144, "step": 9318 }, { "epoch": 0.6, "learning_rate": 7.359026131708556e-05, "loss": 0.1389, "step": 9319 }, { "epoch": 0.6, "learning_rate": 7.357023244720278e-05, "loss": 0.1397, "step": 9320 }, { "epoch": 0.6, "learning_rate": 7.35502047171237e-05, "loss": 0.1331, "step": 9321 }, { "epoch": 0.6, "learning_rate": 7.353017812771197e-05, "loss": 0.14, "step": 9322 }, { "epoch": 0.6, "learning_rate": 7.351015267983135e-05, "loss": 0.147, "step": 9323 }, { "epoch": 0.6, "learning_rate": 7.349012837434536e-05, "loss": 0.1289, "step": 9324 }, { "epoch": 0.6, "learning_rate": 7.347010521211761e-05, "loss": 0.1483, "step": 9325 }, { "epoch": 0.6, "learning_rate": 7.345008319401163e-05, "loss": 0.1455, "step": 9326 }, { "epoch": 0.6, "learning_rate": 7.343006232089082e-05, "loss": 0.137, "step": 9327 }, { "epoch": 0.6, "learning_rate": 7.341004259361867e-05, "loss": 0.1485, "step": 9328 }, { "epoch": 0.6, "learning_rate": 7.339002401305848e-05, "loss": 0.1283, "step": 9329 }, { "epoch": 0.6, "learning_rate": 7.337000658007363e-05, "loss": 0.1431, "step": 9330 }, { "epoch": 0.6, "learning_rate": 7.334999029552734e-05, "loss": 0.1201, "step": 9331 }, { "epoch": 0.6, "learning_rate": 7.332997516028282e-05, "loss": 0.1457, "step": 9332 }, { "epoch": 0.6, "learning_rate": 7.33099611752033e-05, "loss": 0.1477, "step": 9333 }, { "epoch": 0.6, "learning_rate": 7.328994834115182e-05, "loss": 0.1337, "step": 9334 }, { "epoch": 0.6, "learning_rate": 7.326993665899152e-05, "loss": 0.1513, "step": 9335 }, { "epoch": 0.6, "learning_rate": 7.324992612958538e-05, "loss": 0.1264, "step": 9336 }, { "epoch": 0.6, "learning_rate": 7.322991675379633e-05, "loss": 0.1519, "step": 9337 }, { "epoch": 0.6, "learning_rate": 7.320990853248737e-05, "loss": 0.1569, "step": 9338 }, { "epoch": 0.6, "learning_rate": 7.318990146652131e-05, "loss": 0.1525, "step": 9339 }, { "epoch": 0.6, "learning_rate": 7.316989555676104e-05, "loss": 0.1479, "step": 9340 }, { "epoch": 0.6, "learning_rate": 7.314989080406925e-05, "loss": 0.141, "step": 9341 }, { "epoch": 0.6, "learning_rate": 7.312988720930869e-05, "loss": 0.1368, "step": 9342 }, { "epoch": 0.6, "learning_rate": 7.310988477334207e-05, "loss": 0.1497, "step": 9343 }, { "epoch": 0.6, "learning_rate": 7.308988349703191e-05, "loss": 0.1705, "step": 9344 }, { "epoch": 0.6, "learning_rate": 7.306988338124089e-05, "loss": 0.1367, "step": 9345 }, { "epoch": 0.6, "learning_rate": 7.304988442683145e-05, "loss": 0.1587, "step": 9346 }, { "epoch": 0.6, "learning_rate": 7.302988663466614e-05, "loss": 0.1506, "step": 9347 }, { "epoch": 0.6, "learning_rate": 7.300989000560729e-05, "loss": 0.1465, "step": 9348 }, { "epoch": 0.6, "learning_rate": 7.298989454051733e-05, "loss": 0.1544, "step": 9349 }, { "epoch": 0.6, "learning_rate": 7.29699002402586e-05, "loss": 0.1589, "step": 9350 }, { "epoch": 0.6, "learning_rate": 7.294990710569328e-05, "loss": 0.1567, "step": 9351 }, { "epoch": 0.6, "learning_rate": 7.292991513768366e-05, "loss": 0.1292, "step": 9352 }, { "epoch": 0.6, "learning_rate": 7.290992433709186e-05, "loss": 0.1458, "step": 9353 }, { "epoch": 0.6, "learning_rate": 7.288993470478009e-05, "loss": 0.1428, "step": 9354 }, { "epoch": 0.6, "learning_rate": 7.28699462416103e-05, "loss": 0.145, "step": 9355 }, { "epoch": 0.6, "learning_rate": 7.284995894844458e-05, "loss": 0.1426, "step": 9356 }, { "epoch": 0.6, "learning_rate": 7.282997282614492e-05, "loss": 0.1596, "step": 9357 }, { "epoch": 0.6, "learning_rate": 7.280998787557313e-05, "loss": 0.1513, "step": 9358 }, { "epoch": 0.6, "learning_rate": 7.279000409759119e-05, "loss": 0.1535, "step": 9359 }, { "epoch": 0.6, "learning_rate": 7.277002149306085e-05, "loss": 0.1472, "step": 9360 }, { "epoch": 0.6, "learning_rate": 7.275004006284392e-05, "loss": 0.1551, "step": 9361 }, { "epoch": 0.6, "learning_rate": 7.273005980780202e-05, "loss": 0.1459, "step": 9362 }, { "epoch": 0.6, "learning_rate": 7.271008072879695e-05, "loss": 0.1452, "step": 9363 }, { "epoch": 0.6, "learning_rate": 7.269010282669026e-05, "loss": 0.1447, "step": 9364 }, { "epoch": 0.6, "learning_rate": 7.267012610234345e-05, "loss": 0.1315, "step": 9365 }, { "epoch": 0.6, "learning_rate": 7.265015055661815e-05, "loss": 0.1593, "step": 9366 }, { "epoch": 0.6, "learning_rate": 7.263017619037571e-05, "loss": 0.1363, "step": 9367 }, { "epoch": 0.6, "learning_rate": 7.26102030044776e-05, "loss": 0.1703, "step": 9368 }, { "epoch": 0.6, "learning_rate": 7.259023099978517e-05, "loss": 0.1531, "step": 9369 }, { "epoch": 0.6, "learning_rate": 7.257026017715976e-05, "loss": 0.1609, "step": 9370 }, { "epoch": 0.6, "learning_rate": 7.255029053746255e-05, "loss": 0.1385, "step": 9371 }, { "epoch": 0.6, "learning_rate": 7.253032208155478e-05, "loss": 0.1236, "step": 9372 }, { "epoch": 0.6, "learning_rate": 7.251035481029765e-05, "loss": 0.1412, "step": 9373 }, { "epoch": 0.6, "learning_rate": 7.24903887245522e-05, "loss": 0.131, "step": 9374 }, { "epoch": 0.6, "learning_rate": 7.247042382517952e-05, "loss": 0.1377, "step": 9375 }, { "epoch": 0.6, "learning_rate": 7.245046011304058e-05, "loss": 0.1314, "step": 9376 }, { "epoch": 0.6, "learning_rate": 7.243049758899638e-05, "loss": 0.1407, "step": 9377 }, { "epoch": 0.6, "learning_rate": 7.241053625390777e-05, "loss": 0.1491, "step": 9378 }, { "epoch": 0.6, "learning_rate": 7.23905761086356e-05, "loss": 0.1671, "step": 9379 }, { "epoch": 0.6, "learning_rate": 7.237061715404072e-05, "loss": 0.1525, "step": 9380 }, { "epoch": 0.6, "learning_rate": 7.235065939098381e-05, "loss": 0.1495, "step": 9381 }, { "epoch": 0.6, "learning_rate": 7.23307028203256e-05, "loss": 0.1364, "step": 9382 }, { "epoch": 0.6, "learning_rate": 7.23107474429267e-05, "loss": 0.159, "step": 9383 }, { "epoch": 0.6, "learning_rate": 7.229079325964778e-05, "loss": 0.1301, "step": 9384 }, { "epoch": 0.6, "learning_rate": 7.22708402713493e-05, "loss": 0.1339, "step": 9385 }, { "epoch": 0.6, "learning_rate": 7.225088847889174e-05, "loss": 0.1521, "step": 9386 }, { "epoch": 0.6, "learning_rate": 7.223093788313562e-05, "loss": 0.1527, "step": 9387 }, { "epoch": 0.6, "learning_rate": 7.221098848494123e-05, "loss": 0.165, "step": 9388 }, { "epoch": 0.6, "learning_rate": 7.219104028516896e-05, "loss": 0.1275, "step": 9389 }, { "epoch": 0.6, "learning_rate": 7.217109328467905e-05, "loss": 0.1628, "step": 9390 }, { "epoch": 0.6, "learning_rate": 7.21511474843318e-05, "loss": 0.1353, "step": 9391 }, { "epoch": 0.6, "learning_rate": 7.213120288498731e-05, "loss": 0.1302, "step": 9392 }, { "epoch": 0.6, "learning_rate": 7.211125948750574e-05, "loss": 0.1517, "step": 9393 }, { "epoch": 0.6, "learning_rate": 7.209131729274721e-05, "loss": 0.1484, "step": 9394 }, { "epoch": 0.6, "learning_rate": 7.207137630157163e-05, "loss": 0.1572, "step": 9395 }, { "epoch": 0.6, "learning_rate": 7.205143651483906e-05, "loss": 0.1529, "step": 9396 }, { "epoch": 0.6, "learning_rate": 7.203149793340938e-05, "loss": 0.139, "step": 9397 }, { "epoch": 0.6, "learning_rate": 7.20115605581425e-05, "loss": 0.1494, "step": 9398 }, { "epoch": 0.6, "learning_rate": 7.199162438989822e-05, "loss": 0.1466, "step": 9399 }, { "epoch": 0.6, "learning_rate": 7.197168942953623e-05, "loss": 0.1405, "step": 9400 }, { "epoch": 0.6, "learning_rate": 7.195175567791635e-05, "loss": 0.152, "step": 9401 }, { "epoch": 0.6, "learning_rate": 7.193182313589815e-05, "loss": 0.1226, "step": 9402 }, { "epoch": 0.6, "learning_rate": 7.191189180434128e-05, "loss": 0.139, "step": 9403 }, { "epoch": 0.6, "learning_rate": 7.189196168410525e-05, "loss": 0.1417, "step": 9404 }, { "epoch": 0.6, "learning_rate": 7.187203277604968e-05, "loss": 0.122, "step": 9405 }, { "epoch": 0.6, "learning_rate": 7.185210508103388e-05, "loss": 0.1611, "step": 9406 }, { "epoch": 0.6, "learning_rate": 7.183217859991728e-05, "loss": 0.1469, "step": 9407 }, { "epoch": 0.6, "learning_rate": 7.181225333355933e-05, "loss": 0.1361, "step": 9408 }, { "epoch": 0.6, "learning_rate": 7.179232928281914e-05, "loss": 0.1507, "step": 9409 }, { "epoch": 0.6, "learning_rate": 7.177240644855612e-05, "loss": 0.1454, "step": 9410 }, { "epoch": 0.6, "learning_rate": 7.175248483162935e-05, "loss": 0.1434, "step": 9411 }, { "epoch": 0.6, "learning_rate": 7.1732564432898e-05, "loss": 0.1253, "step": 9412 }, { "epoch": 0.6, "learning_rate": 7.17126452532212e-05, "loss": 0.1454, "step": 9413 }, { "epoch": 0.6, "learning_rate": 7.169272729345787e-05, "loss": 0.14, "step": 9414 }, { "epoch": 0.6, "learning_rate": 7.167281055446708e-05, "loss": 0.1386, "step": 9415 }, { "epoch": 0.6, "learning_rate": 7.16528950371077e-05, "loss": 0.1314, "step": 9416 }, { "epoch": 0.6, "learning_rate": 7.163298074223868e-05, "loss": 0.1637, "step": 9417 }, { "epoch": 0.6, "learning_rate": 7.161306767071875e-05, "loss": 0.1595, "step": 9418 }, { "epoch": 0.6, "learning_rate": 7.159315582340671e-05, "loss": 0.1426, "step": 9419 }, { "epoch": 0.6, "learning_rate": 7.157324520116128e-05, "loss": 0.1485, "step": 9420 }, { "epoch": 0.6, "learning_rate": 7.155333580484111e-05, "loss": 0.1418, "step": 9421 }, { "epoch": 0.6, "learning_rate": 7.153342763530482e-05, "loss": 0.1548, "step": 9422 }, { "epoch": 0.6, "learning_rate": 7.151352069341095e-05, "loss": 0.1392, "step": 9423 }, { "epoch": 0.6, "learning_rate": 7.149361498001803e-05, "loss": 0.1304, "step": 9424 }, { "epoch": 0.6, "learning_rate": 7.147371049598449e-05, "loss": 0.1498, "step": 9425 }, { "epoch": 0.6, "learning_rate": 7.145380724216868e-05, "loss": 0.1443, "step": 9426 }, { "epoch": 0.6, "learning_rate": 7.143390521942903e-05, "loss": 0.1375, "step": 9427 }, { "epoch": 0.6, "learning_rate": 7.141400442862376e-05, "loss": 0.1442, "step": 9428 }, { "epoch": 0.6, "learning_rate": 7.139410487061115e-05, "loss": 0.1185, "step": 9429 }, { "epoch": 0.6, "learning_rate": 7.137420654624934e-05, "loss": 0.1598, "step": 9430 }, { "epoch": 0.6, "learning_rate": 7.135430945639654e-05, "loss": 0.1687, "step": 9431 }, { "epoch": 0.6, "learning_rate": 7.133441360191076e-05, "loss": 0.1597, "step": 9432 }, { "epoch": 0.6, "learning_rate": 7.131451898365e-05, "loss": 0.1391, "step": 9433 }, { "epoch": 0.6, "learning_rate": 7.12946256024723e-05, "loss": 0.1411, "step": 9434 }, { "epoch": 0.6, "learning_rate": 7.127473345923554e-05, "loss": 0.15, "step": 9435 }, { "epoch": 0.61, "learning_rate": 7.125484255479758e-05, "loss": 0.147, "step": 9436 }, { "epoch": 0.61, "learning_rate": 7.123495289001624e-05, "loss": 0.1358, "step": 9437 }, { "epoch": 0.61, "learning_rate": 7.121506446574931e-05, "loss": 0.1238, "step": 9438 }, { "epoch": 0.61, "learning_rate": 7.119517728285442e-05, "loss": 0.1532, "step": 9439 }, { "epoch": 0.61, "learning_rate": 7.117529134218927e-05, "loss": 0.1455, "step": 9440 }, { "epoch": 0.61, "learning_rate": 7.115540664461146e-05, "loss": 0.1396, "step": 9441 }, { "epoch": 0.61, "learning_rate": 7.11355231909785e-05, "loss": 0.1375, "step": 9442 }, { "epoch": 0.61, "learning_rate": 7.111564098214792e-05, "loss": 0.1323, "step": 9443 }, { "epoch": 0.61, "learning_rate": 7.109576001897708e-05, "loss": 0.1367, "step": 9444 }, { "epoch": 0.61, "learning_rate": 7.107588030232349e-05, "loss": 0.1473, "step": 9445 }, { "epoch": 0.61, "learning_rate": 7.105600183304437e-05, "loss": 0.1431, "step": 9446 }, { "epoch": 0.61, "learning_rate": 7.1036124611997e-05, "loss": 0.1367, "step": 9447 }, { "epoch": 0.61, "learning_rate": 7.101624864003868e-05, "loss": 0.1282, "step": 9448 }, { "epoch": 0.61, "learning_rate": 7.099637391802645e-05, "loss": 0.1279, "step": 9449 }, { "epoch": 0.61, "learning_rate": 7.097650044681756e-05, "loss": 0.1397, "step": 9450 }, { "epoch": 0.61, "learning_rate": 7.095662822726896e-05, "loss": 0.1318, "step": 9451 }, { "epoch": 0.61, "learning_rate": 7.093675726023772e-05, "loss": 0.1363, "step": 9452 }, { "epoch": 0.61, "learning_rate": 7.09168875465808e-05, "loss": 0.1435, "step": 9453 }, { "epoch": 0.61, "learning_rate": 7.0897019087155e-05, "loss": 0.1367, "step": 9454 }, { "epoch": 0.61, "learning_rate": 7.087715188281726e-05, "loss": 0.1311, "step": 9455 }, { "epoch": 0.61, "learning_rate": 7.085728593442431e-05, "loss": 0.1302, "step": 9456 }, { "epoch": 0.61, "learning_rate": 7.083742124283296e-05, "loss": 0.1385, "step": 9457 }, { "epoch": 0.61, "learning_rate": 7.081755780889977e-05, "loss": 0.1311, "step": 9458 }, { "epoch": 0.61, "learning_rate": 7.07976956334815e-05, "loss": 0.1315, "step": 9459 }, { "epoch": 0.61, "learning_rate": 7.077783471743465e-05, "loss": 0.1637, "step": 9460 }, { "epoch": 0.61, "learning_rate": 7.07579750616157e-05, "loss": 0.1332, "step": 9461 }, { "epoch": 0.61, "learning_rate": 7.073811666688119e-05, "loss": 0.1466, "step": 9462 }, { "epoch": 0.61, "learning_rate": 7.071825953408745e-05, "loss": 0.1403, "step": 9463 }, { "epoch": 0.61, "learning_rate": 7.069840366409096e-05, "loss": 0.1291, "step": 9464 }, { "epoch": 0.61, "learning_rate": 7.067854905774787e-05, "loss": 0.1393, "step": 9465 }, { "epoch": 0.61, "learning_rate": 7.06586957159145e-05, "loss": 0.1365, "step": 9466 }, { "epoch": 0.61, "learning_rate": 7.063884363944708e-05, "loss": 0.1427, "step": 9467 }, { "epoch": 0.61, "learning_rate": 7.061899282920163e-05, "loss": 0.1368, "step": 9468 }, { "epoch": 0.61, "learning_rate": 7.059914328603434e-05, "loss": 0.1363, "step": 9469 }, { "epoch": 0.61, "learning_rate": 7.057929501080116e-05, "loss": 0.1397, "step": 9470 }, { "epoch": 0.61, "learning_rate": 7.055944800435815e-05, "loss": 0.1234, "step": 9471 }, { "epoch": 0.61, "learning_rate": 7.053960226756114e-05, "loss": 0.1583, "step": 9472 }, { "epoch": 0.61, "learning_rate": 7.051975780126603e-05, "loss": 0.1344, "step": 9473 }, { "epoch": 0.61, "learning_rate": 7.049991460632865e-05, "loss": 0.1402, "step": 9474 }, { "epoch": 0.61, "learning_rate": 7.04800726836047e-05, "loss": 0.1456, "step": 9475 }, { "epoch": 0.61, "learning_rate": 7.046023203394991e-05, "loss": 0.1493, "step": 9476 }, { "epoch": 0.61, "learning_rate": 7.044039265821991e-05, "loss": 0.1499, "step": 9477 }, { "epoch": 0.61, "learning_rate": 7.042055455727034e-05, "loss": 0.1359, "step": 9478 }, { "epoch": 0.61, "learning_rate": 7.040071773195663e-05, "loss": 0.1458, "step": 9479 }, { "epoch": 0.61, "learning_rate": 7.038088218313435e-05, "loss": 0.1414, "step": 9480 }, { "epoch": 0.61, "learning_rate": 7.036104791165892e-05, "loss": 0.1493, "step": 9481 }, { "epoch": 0.61, "learning_rate": 7.034121491838563e-05, "loss": 0.1428, "step": 9482 }, { "epoch": 0.61, "learning_rate": 7.032138320416986e-05, "loss": 0.1529, "step": 9483 }, { "epoch": 0.61, "learning_rate": 7.030155276986683e-05, "loss": 0.1481, "step": 9484 }, { "epoch": 0.61, "learning_rate": 7.02817236163318e-05, "loss": 0.1368, "step": 9485 }, { "epoch": 0.61, "learning_rate": 7.026189574441984e-05, "loss": 0.132, "step": 9486 }, { "epoch": 0.61, "learning_rate": 7.02420691549861e-05, "loss": 0.1473, "step": 9487 }, { "epoch": 0.61, "learning_rate": 7.022224384888562e-05, "loss": 0.1481, "step": 9488 }, { "epoch": 0.61, "learning_rate": 7.02024198269733e-05, "loss": 0.1689, "step": 9489 }, { "epoch": 0.61, "learning_rate": 7.018259709010418e-05, "loss": 0.1482, "step": 9490 }, { "epoch": 0.61, "learning_rate": 7.016277563913302e-05, "loss": 0.1531, "step": 9491 }, { "epoch": 0.61, "learning_rate": 7.014295547491476e-05, "loss": 0.1534, "step": 9492 }, { "epoch": 0.61, "learning_rate": 7.012313659830404e-05, "loss": 0.1269, "step": 9493 }, { "epoch": 0.61, "learning_rate": 7.010331901015562e-05, "loss": 0.142, "step": 9494 }, { "epoch": 0.61, "learning_rate": 7.008350271132417e-05, "loss": 0.1536, "step": 9495 }, { "epoch": 0.61, "learning_rate": 7.006368770266421e-05, "loss": 0.1555, "step": 9496 }, { "epoch": 0.61, "learning_rate": 7.004387398503036e-05, "loss": 0.1415, "step": 9497 }, { "epoch": 0.61, "learning_rate": 7.002406155927704e-05, "loss": 0.1147, "step": 9498 }, { "epoch": 0.61, "learning_rate": 7.000425042625871e-05, "loss": 0.1488, "step": 9499 }, { "epoch": 0.61, "learning_rate": 6.998444058682974e-05, "loss": 0.1514, "step": 9500 }, { "epoch": 0.61, "learning_rate": 6.996463204184441e-05, "loss": 0.1507, "step": 9501 }, { "epoch": 0.61, "learning_rate": 6.994482479215702e-05, "loss": 0.118, "step": 9502 }, { "epoch": 0.61, "learning_rate": 6.992501883862174e-05, "loss": 0.1356, "step": 9503 }, { "epoch": 0.61, "learning_rate": 6.990521418209276e-05, "loss": 0.1378, "step": 9504 }, { "epoch": 0.61, "learning_rate": 6.988541082342411e-05, "loss": 0.1573, "step": 9505 }, { "epoch": 0.61, "learning_rate": 6.986560876346989e-05, "loss": 0.1391, "step": 9506 }, { "epoch": 0.61, "learning_rate": 6.984580800308403e-05, "loss": 0.1546, "step": 9507 }, { "epoch": 0.61, "learning_rate": 6.982600854312045e-05, "loss": 0.1633, "step": 9508 }, { "epoch": 0.61, "learning_rate": 6.980621038443306e-05, "loss": 0.1733, "step": 9509 }, { "epoch": 0.61, "learning_rate": 6.978641352787562e-05, "loss": 0.1263, "step": 9510 }, { "epoch": 0.61, "learning_rate": 6.976661797430195e-05, "loss": 0.1436, "step": 9511 }, { "epoch": 0.61, "learning_rate": 6.974682372456567e-05, "loss": 0.1516, "step": 9512 }, { "epoch": 0.61, "learning_rate": 6.972703077952048e-05, "loss": 0.1373, "step": 9513 }, { "epoch": 0.61, "learning_rate": 6.970723914001998e-05, "loss": 0.1292, "step": 9514 }, { "epoch": 0.61, "learning_rate": 6.96874488069176e-05, "loss": 0.1379, "step": 9515 }, { "epoch": 0.61, "learning_rate": 6.966765978106692e-05, "loss": 0.1513, "step": 9516 }, { "epoch": 0.61, "learning_rate": 6.964787206332129e-05, "loss": 0.1611, "step": 9517 }, { "epoch": 0.61, "learning_rate": 6.962808565453413e-05, "loss": 0.1435, "step": 9518 }, { "epoch": 0.61, "learning_rate": 6.960830055555867e-05, "loss": 0.1638, "step": 9519 }, { "epoch": 0.61, "learning_rate": 6.958851676724823e-05, "loss": 0.139, "step": 9520 }, { "epoch": 0.61, "learning_rate": 6.956873429045596e-05, "loss": 0.1456, "step": 9521 }, { "epoch": 0.61, "learning_rate": 6.954895312603497e-05, "loss": 0.1603, "step": 9522 }, { "epoch": 0.61, "learning_rate": 6.952917327483841e-05, "loss": 0.1587, "step": 9523 }, { "epoch": 0.61, "learning_rate": 6.950939473771922e-05, "loss": 0.1359, "step": 9524 }, { "epoch": 0.61, "learning_rate": 6.948961751553048e-05, "loss": 0.1493, "step": 9525 }, { "epoch": 0.61, "learning_rate": 6.946984160912493e-05, "loss": 0.1324, "step": 9526 }, { "epoch": 0.61, "learning_rate": 6.945006701935557e-05, "loss": 0.1474, "step": 9527 }, { "epoch": 0.61, "learning_rate": 6.943029374707514e-05, "loss": 0.1392, "step": 9528 }, { "epoch": 0.61, "learning_rate": 6.941052179313634e-05, "loss": 0.1541, "step": 9529 }, { "epoch": 0.61, "learning_rate": 6.939075115839191e-05, "loss": 0.1396, "step": 9530 }, { "epoch": 0.61, "learning_rate": 6.937098184369442e-05, "loss": 0.1279, "step": 9531 }, { "epoch": 0.61, "learning_rate": 6.935121384989651e-05, "loss": 0.1316, "step": 9532 }, { "epoch": 0.61, "learning_rate": 6.933144717785062e-05, "loss": 0.1543, "step": 9533 }, { "epoch": 0.61, "learning_rate": 6.931168182840923e-05, "loss": 0.1412, "step": 9534 }, { "epoch": 0.61, "learning_rate": 6.929191780242477e-05, "loss": 0.1303, "step": 9535 }, { "epoch": 0.61, "learning_rate": 6.927215510074945e-05, "loss": 0.1474, "step": 9536 }, { "epoch": 0.61, "learning_rate": 6.925239372423573e-05, "loss": 0.1498, "step": 9537 }, { "epoch": 0.61, "learning_rate": 6.92326336737357e-05, "loss": 0.1494, "step": 9538 }, { "epoch": 0.61, "learning_rate": 6.921287495010159e-05, "loss": 0.1555, "step": 9539 }, { "epoch": 0.61, "learning_rate": 6.919311755418549e-05, "loss": 0.1348, "step": 9540 }, { "epoch": 0.61, "learning_rate": 6.917336148683948e-05, "loss": 0.1416, "step": 9541 }, { "epoch": 0.61, "learning_rate": 6.915360674891552e-05, "loss": 0.1233, "step": 9542 }, { "epoch": 0.61, "learning_rate": 6.913385334126555e-05, "loss": 0.1438, "step": 9543 }, { "epoch": 0.61, "learning_rate": 6.911410126474148e-05, "loss": 0.1303, "step": 9544 }, { "epoch": 0.61, "learning_rate": 6.909435052019509e-05, "loss": 0.1495, "step": 9545 }, { "epoch": 0.61, "learning_rate": 6.907460110847819e-05, "loss": 0.1304, "step": 9546 }, { "epoch": 0.61, "learning_rate": 6.905485303044243e-05, "loss": 0.1422, "step": 9547 }, { "epoch": 0.61, "learning_rate": 6.903510628693957e-05, "loss": 0.1679, "step": 9548 }, { "epoch": 0.61, "learning_rate": 6.901536087882108e-05, "loss": 0.1428, "step": 9549 }, { "epoch": 0.61, "learning_rate": 6.899561680693856e-05, "loss": 0.1584, "step": 9550 }, { "epoch": 0.61, "learning_rate": 6.89758740721435e-05, "loss": 0.1443, "step": 9551 }, { "epoch": 0.61, "learning_rate": 6.895613267528724e-05, "loss": 0.1362, "step": 9552 }, { "epoch": 0.61, "learning_rate": 6.893639261722125e-05, "loss": 0.1411, "step": 9553 }, { "epoch": 0.61, "learning_rate": 6.891665389879674e-05, "loss": 0.1403, "step": 9554 }, { "epoch": 0.61, "learning_rate": 6.889691652086505e-05, "loss": 0.148, "step": 9555 }, { "epoch": 0.61, "learning_rate": 6.88771804842773e-05, "loss": 0.1528, "step": 9556 }, { "epoch": 0.61, "learning_rate": 6.885744578988463e-05, "loss": 0.1386, "step": 9557 }, { "epoch": 0.61, "learning_rate": 6.883771243853816e-05, "loss": 0.1495, "step": 9558 }, { "epoch": 0.61, "learning_rate": 6.881798043108883e-05, "loss": 0.1444, "step": 9559 }, { "epoch": 0.61, "learning_rate": 6.879824976838764e-05, "loss": 0.1437, "step": 9560 }, { "epoch": 0.61, "learning_rate": 6.87785204512855e-05, "loss": 0.1467, "step": 9561 }, { "epoch": 0.61, "learning_rate": 6.875879248063326e-05, "loss": 0.1656, "step": 9562 }, { "epoch": 0.61, "learning_rate": 6.873906585728168e-05, "loss": 0.1484, "step": 9563 }, { "epoch": 0.61, "learning_rate": 6.871934058208146e-05, "loss": 0.1272, "step": 9564 }, { "epoch": 0.61, "learning_rate": 6.869961665588334e-05, "loss": 0.1675, "step": 9565 }, { "epoch": 0.61, "learning_rate": 6.867989407953785e-05, "loss": 0.1458, "step": 9566 }, { "epoch": 0.61, "learning_rate": 6.866017285389558e-05, "loss": 0.1417, "step": 9567 }, { "epoch": 0.61, "learning_rate": 6.864045297980702e-05, "loss": 0.1509, "step": 9568 }, { "epoch": 0.61, "learning_rate": 6.862073445812263e-05, "loss": 0.1229, "step": 9569 }, { "epoch": 0.61, "learning_rate": 6.860101728969275e-05, "loss": 0.165, "step": 9570 }, { "epoch": 0.61, "learning_rate": 6.858130147536767e-05, "loss": 0.1514, "step": 9571 }, { "epoch": 0.61, "learning_rate": 6.856158701599775e-05, "loss": 0.1378, "step": 9572 }, { "epoch": 0.61, "learning_rate": 6.854187391243308e-05, "loss": 0.1458, "step": 9573 }, { "epoch": 0.61, "learning_rate": 6.852216216552387e-05, "loss": 0.1441, "step": 9574 }, { "epoch": 0.61, "learning_rate": 6.850245177612022e-05, "loss": 0.1468, "step": 9575 }, { "epoch": 0.61, "learning_rate": 6.848274274507206e-05, "loss": 0.148, "step": 9576 }, { "epoch": 0.61, "learning_rate": 6.846303507322945e-05, "loss": 0.1383, "step": 9577 }, { "epoch": 0.61, "learning_rate": 6.844332876144223e-05, "loss": 0.1349, "step": 9578 }, { "epoch": 0.61, "learning_rate": 6.842362381056033e-05, "loss": 0.1348, "step": 9579 }, { "epoch": 0.61, "learning_rate": 6.840392022143345e-05, "loss": 0.174, "step": 9580 }, { "epoch": 0.61, "learning_rate": 6.838421799491142e-05, "loss": 0.1378, "step": 9581 }, { "epoch": 0.61, "learning_rate": 6.836451713184387e-05, "loss": 0.1483, "step": 9582 }, { "epoch": 0.61, "learning_rate": 6.834481763308032e-05, "loss": 0.1428, "step": 9583 }, { "epoch": 0.61, "learning_rate": 6.83251194994705e-05, "loss": 0.1319, "step": 9584 }, { "epoch": 0.61, "learning_rate": 6.830542273186376e-05, "loss": 0.1571, "step": 9585 }, { "epoch": 0.61, "learning_rate": 6.828572733110963e-05, "loss": 0.1614, "step": 9586 }, { "epoch": 0.61, "learning_rate": 6.826603329805741e-05, "loss": 0.1365, "step": 9587 }, { "epoch": 0.61, "learning_rate": 6.824634063355654e-05, "loss": 0.1246, "step": 9588 }, { "epoch": 0.61, "learning_rate": 6.822664933845617e-05, "loss": 0.1613, "step": 9589 }, { "epoch": 0.61, "learning_rate": 6.820695941360552e-05, "loss": 0.1514, "step": 9590 }, { "epoch": 0.61, "learning_rate": 6.81872708598538e-05, "loss": 0.1304, "step": 9591 }, { "epoch": 0.62, "learning_rate": 6.816758367804998e-05, "loss": 0.1354, "step": 9592 }, { "epoch": 0.62, "learning_rate": 6.81478978690432e-05, "loss": 0.122, "step": 9593 }, { "epoch": 0.62, "learning_rate": 6.812821343368234e-05, "loss": 0.1301, "step": 9594 }, { "epoch": 0.62, "learning_rate": 6.810853037281639e-05, "loss": 0.1362, "step": 9595 }, { "epoch": 0.62, "learning_rate": 6.808884868729412e-05, "loss": 0.1471, "step": 9596 }, { "epoch": 0.62, "learning_rate": 6.806916837796434e-05, "loss": 0.128, "step": 9597 }, { "epoch": 0.62, "learning_rate": 6.80494894456758e-05, "loss": 0.1707, "step": 9598 }, { "epoch": 0.62, "learning_rate": 6.802981189127713e-05, "loss": 0.139, "step": 9599 }, { "epoch": 0.62, "learning_rate": 6.801013571561699e-05, "loss": 0.1428, "step": 9600 }, { "epoch": 0.62, "learning_rate": 6.799046091954387e-05, "loss": 0.1445, "step": 9601 }, { "epoch": 0.62, "learning_rate": 6.797078750390633e-05, "loss": 0.1564, "step": 9602 }, { "epoch": 0.62, "learning_rate": 6.795111546955274e-05, "loss": 0.1745, "step": 9603 }, { "epoch": 0.62, "learning_rate": 6.793144481733148e-05, "loss": 0.1576, "step": 9604 }, { "epoch": 0.62, "learning_rate": 6.791177554809091e-05, "loss": 0.1473, "step": 9605 }, { "epoch": 0.62, "learning_rate": 6.789210766267923e-05, "loss": 0.1479, "step": 9606 }, { "epoch": 0.62, "learning_rate": 6.787244116194467e-05, "loss": 0.1419, "step": 9607 }, { "epoch": 0.62, "learning_rate": 6.78527760467353e-05, "loss": 0.145, "step": 9608 }, { "epoch": 0.62, "learning_rate": 6.783311231789928e-05, "loss": 0.1334, "step": 9609 }, { "epoch": 0.62, "learning_rate": 6.781344997628458e-05, "loss": 0.1443, "step": 9610 }, { "epoch": 0.62, "learning_rate": 6.779378902273912e-05, "loss": 0.1502, "step": 9611 }, { "epoch": 0.62, "learning_rate": 6.777412945811088e-05, "loss": 0.1359, "step": 9612 }, { "epoch": 0.62, "learning_rate": 6.775447128324759e-05, "loss": 0.1353, "step": 9613 }, { "epoch": 0.62, "learning_rate": 6.773481449899708e-05, "loss": 0.1393, "step": 9614 }, { "epoch": 0.62, "learning_rate": 6.771515910620705e-05, "loss": 0.1382, "step": 9615 }, { "epoch": 0.62, "learning_rate": 6.769550510572521e-05, "loss": 0.134, "step": 9616 }, { "epoch": 0.62, "learning_rate": 6.767585249839907e-05, "loss": 0.1457, "step": 9617 }, { "epoch": 0.62, "learning_rate": 6.765620128507619e-05, "loss": 0.1495, "step": 9618 }, { "epoch": 0.62, "learning_rate": 6.763655146660409e-05, "loss": 0.1557, "step": 9619 }, { "epoch": 0.62, "learning_rate": 6.76169030438301e-05, "loss": 0.1382, "step": 9620 }, { "epoch": 0.62, "learning_rate": 6.759725601760165e-05, "loss": 0.1301, "step": 9621 }, { "epoch": 0.62, "learning_rate": 6.757761038876596e-05, "loss": 0.1471, "step": 9622 }, { "epoch": 0.62, "learning_rate": 6.755796615817034e-05, "loss": 0.1351, "step": 9623 }, { "epoch": 0.62, "learning_rate": 6.753832332666194e-05, "loss": 0.1387, "step": 9624 }, { "epoch": 0.62, "learning_rate": 6.751868189508783e-05, "loss": 0.1534, "step": 9625 }, { "epoch": 0.62, "learning_rate": 6.74990418642951e-05, "loss": 0.1486, "step": 9626 }, { "epoch": 0.62, "learning_rate": 6.74794032351307e-05, "loss": 0.1287, "step": 9627 }, { "epoch": 0.62, "learning_rate": 6.745976600844163e-05, "loss": 0.1607, "step": 9628 }, { "epoch": 0.62, "learning_rate": 6.744013018507471e-05, "loss": 0.1564, "step": 9629 }, { "epoch": 0.62, "learning_rate": 6.742049576587677e-05, "loss": 0.1312, "step": 9630 }, { "epoch": 0.62, "learning_rate": 6.740086275169455e-05, "loss": 0.1306, "step": 9631 }, { "epoch": 0.62, "learning_rate": 6.738123114337472e-05, "loss": 0.1491, "step": 9632 }, { "epoch": 0.62, "learning_rate": 6.736160094176394e-05, "loss": 0.1529, "step": 9633 }, { "epoch": 0.62, "learning_rate": 6.734197214770875e-05, "loss": 0.1479, "step": 9634 }, { "epoch": 0.62, "learning_rate": 6.73223447620557e-05, "loss": 0.1449, "step": 9635 }, { "epoch": 0.62, "learning_rate": 6.730271878565117e-05, "loss": 0.1432, "step": 9636 }, { "epoch": 0.62, "learning_rate": 6.72830942193416e-05, "loss": 0.1531, "step": 9637 }, { "epoch": 0.62, "learning_rate": 6.726347106397332e-05, "loss": 0.148, "step": 9638 }, { "epoch": 0.62, "learning_rate": 6.724384932039252e-05, "loss": 0.1474, "step": 9639 }, { "epoch": 0.62, "learning_rate": 6.722422898944548e-05, "loss": 0.1412, "step": 9640 }, { "epoch": 0.62, "learning_rate": 6.720461007197829e-05, "loss": 0.1497, "step": 9641 }, { "epoch": 0.62, "learning_rate": 6.71849925688371e-05, "loss": 0.1297, "step": 9642 }, { "epoch": 0.62, "learning_rate": 6.716537648086784e-05, "loss": 0.1386, "step": 9643 }, { "epoch": 0.62, "learning_rate": 6.714576180891654e-05, "loss": 0.1528, "step": 9644 }, { "epoch": 0.62, "learning_rate": 6.712614855382908e-05, "loss": 0.1362, "step": 9645 }, { "epoch": 0.62, "learning_rate": 6.710653671645123e-05, "loss": 0.144, "step": 9646 }, { "epoch": 0.62, "learning_rate": 6.708692629762887e-05, "loss": 0.1331, "step": 9647 }, { "epoch": 0.62, "learning_rate": 6.706731729820763e-05, "loss": 0.1415, "step": 9648 }, { "epoch": 0.62, "learning_rate": 6.704770971903324e-05, "loss": 0.1452, "step": 9649 }, { "epoch": 0.62, "learning_rate": 6.702810356095123e-05, "loss": 0.1326, "step": 9650 }, { "epoch": 0.62, "learning_rate": 6.700849882480714e-05, "loss": 0.1411, "step": 9651 }, { "epoch": 0.62, "learning_rate": 6.69888955114465e-05, "loss": 0.154, "step": 9652 }, { "epoch": 0.62, "learning_rate": 6.69692936217146e-05, "loss": 0.1367, "step": 9653 }, { "epoch": 0.62, "learning_rate": 6.694969315645687e-05, "loss": 0.1423, "step": 9654 }, { "epoch": 0.62, "learning_rate": 6.693009411651859e-05, "loss": 0.145, "step": 9655 }, { "epoch": 0.62, "learning_rate": 6.691049650274499e-05, "loss": 0.1476, "step": 9656 }, { "epoch": 0.62, "learning_rate": 6.68909003159812e-05, "loss": 0.1332, "step": 9657 }, { "epoch": 0.62, "learning_rate": 6.687130555707231e-05, "loss": 0.1346, "step": 9658 }, { "epoch": 0.62, "learning_rate": 6.685171222686341e-05, "loss": 0.1505, "step": 9659 }, { "epoch": 0.62, "learning_rate": 6.683212032619943e-05, "loss": 0.1567, "step": 9660 }, { "epoch": 0.62, "learning_rate": 6.681252985592533e-05, "loss": 0.1443, "step": 9661 }, { "epoch": 0.62, "learning_rate": 6.67929408168859e-05, "loss": 0.1219, "step": 9662 }, { "epoch": 0.62, "learning_rate": 6.677335320992602e-05, "loss": 0.1677, "step": 9663 }, { "epoch": 0.62, "learning_rate": 6.675376703589036e-05, "loss": 0.1399, "step": 9664 }, { "epoch": 0.62, "learning_rate": 6.673418229562356e-05, "loss": 0.1415, "step": 9665 }, { "epoch": 0.62, "learning_rate": 6.671459898997032e-05, "loss": 0.1316, "step": 9666 }, { "epoch": 0.62, "learning_rate": 6.669501711977508e-05, "loss": 0.1283, "step": 9667 }, { "epoch": 0.62, "learning_rate": 6.667543668588242e-05, "loss": 0.1349, "step": 9668 }, { "epoch": 0.62, "learning_rate": 6.665585768913668e-05, "loss": 0.1317, "step": 9669 }, { "epoch": 0.62, "learning_rate": 6.663628013038227e-05, "loss": 0.1311, "step": 9670 }, { "epoch": 0.62, "learning_rate": 6.66167040104635e-05, "loss": 0.1574, "step": 9671 }, { "epoch": 0.62, "learning_rate": 6.659712933022453e-05, "loss": 0.1414, "step": 9672 }, { "epoch": 0.62, "learning_rate": 6.657755609050961e-05, "loss": 0.1468, "step": 9673 }, { "epoch": 0.62, "learning_rate": 6.65579842921628e-05, "loss": 0.145, "step": 9674 }, { "epoch": 0.62, "learning_rate": 6.65384139360282e-05, "loss": 0.1571, "step": 9675 }, { "epoch": 0.62, "learning_rate": 6.651884502294973e-05, "loss": 0.1653, "step": 9676 }, { "epoch": 0.62, "learning_rate": 6.649927755377136e-05, "loss": 0.1268, "step": 9677 }, { "epoch": 0.62, "learning_rate": 6.647971152933698e-05, "loss": 0.1486, "step": 9678 }, { "epoch": 0.62, "learning_rate": 6.64601469504903e-05, "loss": 0.1452, "step": 9679 }, { "epoch": 0.62, "learning_rate": 6.644058381807512e-05, "loss": 0.1445, "step": 9680 }, { "epoch": 0.62, "learning_rate": 6.64210221329351e-05, "loss": 0.1391, "step": 9681 }, { "epoch": 0.62, "learning_rate": 6.640146189591387e-05, "loss": 0.1416, "step": 9682 }, { "epoch": 0.62, "learning_rate": 6.638190310785495e-05, "loss": 0.1536, "step": 9683 }, { "epoch": 0.62, "learning_rate": 6.636234576960185e-05, "loss": 0.1363, "step": 9684 }, { "epoch": 0.62, "learning_rate": 6.634278988199802e-05, "loss": 0.1402, "step": 9685 }, { "epoch": 0.62, "learning_rate": 6.632323544588674e-05, "loss": 0.1648, "step": 9686 }, { "epoch": 0.62, "learning_rate": 6.630368246211138e-05, "loss": 0.1517, "step": 9687 }, { "epoch": 0.62, "learning_rate": 6.628413093151516e-05, "loss": 0.1348, "step": 9688 }, { "epoch": 0.62, "learning_rate": 6.626458085494127e-05, "loss": 0.1277, "step": 9689 }, { "epoch": 0.62, "learning_rate": 6.624503223323277e-05, "loss": 0.1555, "step": 9690 }, { "epoch": 0.62, "learning_rate": 6.622548506723276e-05, "loss": 0.1477, "step": 9691 }, { "epoch": 0.62, "learning_rate": 6.620593935778422e-05, "loss": 0.1493, "step": 9692 }, { "epoch": 0.62, "learning_rate": 6.618639510573003e-05, "loss": 0.1425, "step": 9693 }, { "epoch": 0.62, "learning_rate": 6.616685231191311e-05, "loss": 0.1553, "step": 9694 }, { "epoch": 0.62, "learning_rate": 6.61473109771762e-05, "loss": 0.1391, "step": 9695 }, { "epoch": 0.62, "learning_rate": 6.61277711023621e-05, "loss": 0.1418, "step": 9696 }, { "epoch": 0.62, "learning_rate": 6.610823268831343e-05, "loss": 0.152, "step": 9697 }, { "epoch": 0.62, "learning_rate": 6.608869573587283e-05, "loss": 0.152, "step": 9698 }, { "epoch": 0.62, "learning_rate": 6.606916024588283e-05, "loss": 0.1451, "step": 9699 }, { "epoch": 0.62, "learning_rate": 6.60496262191859e-05, "loss": 0.1356, "step": 9700 }, { "epoch": 0.62, "learning_rate": 6.603009365662449e-05, "loss": 0.159, "step": 9701 }, { "epoch": 0.62, "learning_rate": 6.601056255904091e-05, "loss": 0.128, "step": 9702 }, { "epoch": 0.62, "learning_rate": 6.599103292727753e-05, "loss": 0.1455, "step": 9703 }, { "epoch": 0.62, "learning_rate": 6.597150476217651e-05, "loss": 0.1169, "step": 9704 }, { "epoch": 0.62, "learning_rate": 6.595197806458004e-05, "loss": 0.1139, "step": 9705 }, { "epoch": 0.62, "learning_rate": 6.593245283533025e-05, "loss": 0.1361, "step": 9706 }, { "epoch": 0.62, "learning_rate": 6.591292907526913e-05, "loss": 0.1518, "step": 9707 }, { "epoch": 0.62, "learning_rate": 6.589340678523867e-05, "loss": 0.1331, "step": 9708 }, { "epoch": 0.62, "learning_rate": 6.587388596608081e-05, "loss": 0.147, "step": 9709 }, { "epoch": 0.62, "learning_rate": 6.58543666186374e-05, "loss": 0.139, "step": 9710 }, { "epoch": 0.62, "learning_rate": 6.583484874375017e-05, "loss": 0.1583, "step": 9711 }, { "epoch": 0.62, "learning_rate": 6.581533234226095e-05, "loss": 0.1543, "step": 9712 }, { "epoch": 0.62, "learning_rate": 6.57958174150113e-05, "loss": 0.1415, "step": 9713 }, { "epoch": 0.62, "learning_rate": 6.577630396284284e-05, "loss": 0.157, "step": 9714 }, { "epoch": 0.62, "learning_rate": 6.575679198659714e-05, "loss": 0.1368, "step": 9715 }, { "epoch": 0.62, "learning_rate": 6.573728148711562e-05, "loss": 0.1369, "step": 9716 }, { "epoch": 0.62, "learning_rate": 6.571777246523972e-05, "loss": 0.1359, "step": 9717 }, { "epoch": 0.62, "learning_rate": 6.569826492181075e-05, "loss": 0.1497, "step": 9718 }, { "epoch": 0.62, "learning_rate": 6.567875885767002e-05, "loss": 0.1448, "step": 9719 }, { "epoch": 0.62, "learning_rate": 6.565925427365876e-05, "loss": 0.1352, "step": 9720 }, { "epoch": 0.62, "learning_rate": 6.563975117061803e-05, "loss": 0.142, "step": 9721 }, { "epoch": 0.62, "learning_rate": 6.562024954938903e-05, "loss": 0.1448, "step": 9722 }, { "epoch": 0.62, "learning_rate": 6.560074941081268e-05, "loss": 0.1419, "step": 9723 }, { "epoch": 0.62, "learning_rate": 6.558125075573002e-05, "loss": 0.1346, "step": 9724 }, { "epoch": 0.62, "learning_rate": 6.556175358498188e-05, "loss": 0.1458, "step": 9725 }, { "epoch": 0.62, "learning_rate": 6.554225789940915e-05, "loss": 0.1379, "step": 9726 }, { "epoch": 0.62, "learning_rate": 6.552276369985257e-05, "loss": 0.1674, "step": 9727 }, { "epoch": 0.62, "learning_rate": 6.550327098715281e-05, "loss": 0.1493, "step": 9728 }, { "epoch": 0.62, "learning_rate": 6.548377976215057e-05, "loss": 0.1311, "step": 9729 }, { "epoch": 0.62, "learning_rate": 6.546429002568635e-05, "loss": 0.1349, "step": 9730 }, { "epoch": 0.62, "learning_rate": 6.544480177860073e-05, "loss": 0.1332, "step": 9731 }, { "epoch": 0.62, "learning_rate": 6.542531502173414e-05, "loss": 0.138, "step": 9732 }, { "epoch": 0.62, "learning_rate": 6.54058297559269e-05, "loss": 0.1243, "step": 9733 }, { "epoch": 0.62, "learning_rate": 6.53863459820194e-05, "loss": 0.1322, "step": 9734 }, { "epoch": 0.62, "learning_rate": 6.536686370085186e-05, "loss": 0.1516, "step": 9735 }, { "epoch": 0.62, "learning_rate": 6.534738291326449e-05, "loss": 0.1489, "step": 9736 }, { "epoch": 0.62, "learning_rate": 6.532790362009736e-05, "loss": 0.1408, "step": 9737 }, { "epoch": 0.62, "learning_rate": 6.530842582219058e-05, "loss": 0.1429, "step": 9738 }, { "epoch": 0.62, "learning_rate": 6.528894952038418e-05, "loss": 0.1453, "step": 9739 }, { "epoch": 0.62, "learning_rate": 6.526947471551798e-05, "loss": 0.1296, "step": 9740 }, { "epoch": 0.62, "learning_rate": 6.525000140843194e-05, "loss": 0.145, "step": 9741 }, { "epoch": 0.62, "learning_rate": 6.523052959996581e-05, "loss": 0.1368, "step": 9742 }, { "epoch": 0.62, "learning_rate": 6.521105929095936e-05, "loss": 0.1473, "step": 9743 }, { "epoch": 0.62, "learning_rate": 6.519159048225223e-05, "loss": 0.1389, "step": 9744 }, { "epoch": 0.62, "learning_rate": 6.517212317468407e-05, "loss": 0.1397, "step": 9745 }, { "epoch": 0.62, "learning_rate": 6.51526573690944e-05, "loss": 0.1423, "step": 9746 }, { "epoch": 0.62, "learning_rate": 6.513319306632264e-05, "loss": 0.1423, "step": 9747 }, { "epoch": 0.63, "learning_rate": 6.51137302672083e-05, "loss": 0.1501, "step": 9748 }, { "epoch": 0.63, "learning_rate": 6.509426897259066e-05, "loss": 0.1217, "step": 9749 }, { "epoch": 0.63, "learning_rate": 6.507480918330906e-05, "loss": 0.1367, "step": 9750 }, { "epoch": 0.63, "learning_rate": 6.505535090020263e-05, "loss": 0.1374, "step": 9751 }, { "epoch": 0.63, "learning_rate": 6.503589412411064e-05, "loss": 0.1449, "step": 9752 }, { "epoch": 0.63, "learning_rate": 6.501643885587212e-05, "loss": 0.1357, "step": 9753 }, { "epoch": 0.63, "learning_rate": 6.499698509632602e-05, "loss": 0.1359, "step": 9754 }, { "epoch": 0.63, "learning_rate": 6.497753284631142e-05, "loss": 0.1321, "step": 9755 }, { "epoch": 0.63, "learning_rate": 6.495808210666714e-05, "loss": 0.128, "step": 9756 }, { "epoch": 0.63, "learning_rate": 6.493863287823205e-05, "loss": 0.1466, "step": 9757 }, { "epoch": 0.63, "learning_rate": 6.491918516184486e-05, "loss": 0.1441, "step": 9758 }, { "epoch": 0.63, "learning_rate": 6.489973895834435e-05, "loss": 0.134, "step": 9759 }, { "epoch": 0.63, "learning_rate": 6.488029426856907e-05, "loss": 0.1685, "step": 9760 }, { "epoch": 0.63, "learning_rate": 6.48608510933576e-05, "loss": 0.1257, "step": 9761 }, { "epoch": 0.63, "learning_rate": 6.48414094335485e-05, "loss": 0.1519, "step": 9762 }, { "epoch": 0.63, "learning_rate": 6.482196928998012e-05, "loss": 0.114, "step": 9763 }, { "epoch": 0.63, "learning_rate": 6.480253066349091e-05, "loss": 0.1253, "step": 9764 }, { "epoch": 0.63, "learning_rate": 6.478309355491912e-05, "loss": 0.1526, "step": 9765 }, { "epoch": 0.63, "learning_rate": 6.476365796510305e-05, "loss": 0.1329, "step": 9766 }, { "epoch": 0.63, "learning_rate": 6.474422389488081e-05, "loss": 0.1268, "step": 9767 }, { "epoch": 0.63, "learning_rate": 6.472479134509052e-05, "loss": 0.1534, "step": 9768 }, { "epoch": 0.63, "learning_rate": 6.470536031657028e-05, "loss": 0.1472, "step": 9769 }, { "epoch": 0.63, "learning_rate": 6.468593081015799e-05, "loss": 0.1279, "step": 9770 }, { "epoch": 0.63, "learning_rate": 6.466650282669161e-05, "loss": 0.1398, "step": 9771 }, { "epoch": 0.63, "learning_rate": 6.464707636700897e-05, "loss": 0.1436, "step": 9772 }, { "epoch": 0.63, "learning_rate": 6.462765143194788e-05, "loss": 0.1377, "step": 9773 }, { "epoch": 0.63, "learning_rate": 6.460822802234602e-05, "loss": 0.1265, "step": 9774 }, { "epoch": 0.63, "learning_rate": 6.458880613904103e-05, "loss": 0.1479, "step": 9775 }, { "epoch": 0.63, "learning_rate": 6.456938578287054e-05, "loss": 0.129, "step": 9776 }, { "epoch": 0.63, "learning_rate": 6.454996695467201e-05, "loss": 0.1386, "step": 9777 }, { "epoch": 0.63, "learning_rate": 6.453054965528295e-05, "loss": 0.1387, "step": 9778 }, { "epoch": 0.63, "learning_rate": 6.45111338855407e-05, "loss": 0.1417, "step": 9779 }, { "epoch": 0.63, "learning_rate": 6.449171964628262e-05, "loss": 0.1547, "step": 9780 }, { "epoch": 0.63, "learning_rate": 6.447230693834594e-05, "loss": 0.1396, "step": 9781 }, { "epoch": 0.63, "learning_rate": 6.445289576256784e-05, "loss": 0.1447, "step": 9782 }, { "epoch": 0.63, "learning_rate": 6.443348611978546e-05, "loss": 0.1573, "step": 9783 }, { "epoch": 0.63, "learning_rate": 6.441407801083583e-05, "loss": 0.1254, "step": 9784 }, { "epoch": 0.63, "learning_rate": 6.439467143655595e-05, "loss": 0.1397, "step": 9785 }, { "epoch": 0.63, "learning_rate": 6.437526639778275e-05, "loss": 0.1183, "step": 9786 }, { "epoch": 0.63, "learning_rate": 6.435586289535314e-05, "loss": 0.1513, "step": 9787 }, { "epoch": 0.63, "learning_rate": 6.433646093010382e-05, "loss": 0.1573, "step": 9788 }, { "epoch": 0.63, "learning_rate": 6.431706050287151e-05, "loss": 0.1345, "step": 9789 }, { "epoch": 0.63, "learning_rate": 6.429766161449298e-05, "loss": 0.1456, "step": 9790 }, { "epoch": 0.63, "learning_rate": 6.42782642658047e-05, "loss": 0.1399, "step": 9791 }, { "epoch": 0.63, "learning_rate": 6.425886845764328e-05, "loss": 0.1453, "step": 9792 }, { "epoch": 0.63, "learning_rate": 6.423947419084511e-05, "loss": 0.1496, "step": 9793 }, { "epoch": 0.63, "learning_rate": 6.422008146624664e-05, "loss": 0.1363, "step": 9794 }, { "epoch": 0.63, "learning_rate": 6.420069028468419e-05, "loss": 0.1256, "step": 9795 }, { "epoch": 0.63, "learning_rate": 6.418130064699397e-05, "loss": 0.129, "step": 9796 }, { "epoch": 0.63, "learning_rate": 6.416191255401223e-05, "loss": 0.1474, "step": 9797 }, { "epoch": 0.63, "learning_rate": 6.414252600657504e-05, "loss": 0.1374, "step": 9798 }, { "epoch": 0.63, "learning_rate": 6.412314100551854e-05, "loss": 0.1413, "step": 9799 }, { "epoch": 0.63, "learning_rate": 6.410375755167863e-05, "loss": 0.1579, "step": 9800 }, { "epoch": 0.63, "learning_rate": 6.40843756458913e-05, "loss": 0.1355, "step": 9801 }, { "epoch": 0.63, "learning_rate": 6.406499528899242e-05, "loss": 0.1437, "step": 9802 }, { "epoch": 0.63, "learning_rate": 6.404561648181769e-05, "loss": 0.1581, "step": 9803 }, { "epoch": 0.63, "learning_rate": 6.402623922520293e-05, "loss": 0.1335, "step": 9804 }, { "epoch": 0.63, "learning_rate": 6.400686351998375e-05, "loss": 0.1367, "step": 9805 }, { "epoch": 0.63, "learning_rate": 6.398748936699579e-05, "loss": 0.1295, "step": 9806 }, { "epoch": 0.63, "learning_rate": 6.396811676707453e-05, "loss": 0.1603, "step": 9807 }, { "epoch": 0.63, "learning_rate": 6.394874572105543e-05, "loss": 0.1459, "step": 9808 }, { "epoch": 0.63, "learning_rate": 6.392937622977392e-05, "loss": 0.1464, "step": 9809 }, { "epoch": 0.63, "learning_rate": 6.391000829406527e-05, "loss": 0.1354, "step": 9810 }, { "epoch": 0.63, "learning_rate": 6.389064191476476e-05, "loss": 0.1535, "step": 9811 }, { "epoch": 0.63, "learning_rate": 6.387127709270757e-05, "loss": 0.1285, "step": 9812 }, { "epoch": 0.63, "learning_rate": 6.385191382872888e-05, "loss": 0.1415, "step": 9813 }, { "epoch": 0.63, "learning_rate": 6.383255212366369e-05, "loss": 0.137, "step": 9814 }, { "epoch": 0.63, "learning_rate": 6.381319197834696e-05, "loss": 0.1238, "step": 9815 }, { "epoch": 0.63, "learning_rate": 6.379383339361371e-05, "loss": 0.1401, "step": 9816 }, { "epoch": 0.63, "learning_rate": 6.377447637029869e-05, "loss": 0.134, "step": 9817 }, { "epoch": 0.63, "learning_rate": 6.375512090923674e-05, "loss": 0.1483, "step": 9818 }, { "epoch": 0.63, "learning_rate": 6.373576701126255e-05, "loss": 0.1376, "step": 9819 }, { "epoch": 0.63, "learning_rate": 6.371641467721083e-05, "loss": 0.1616, "step": 9820 }, { "epoch": 0.63, "learning_rate": 6.369706390791612e-05, "loss": 0.1406, "step": 9821 }, { "epoch": 0.63, "learning_rate": 6.36777147042129e-05, "loss": 0.1335, "step": 9822 }, { "epoch": 0.63, "learning_rate": 6.36583670669357e-05, "loss": 0.134, "step": 9823 }, { "epoch": 0.63, "learning_rate": 6.363902099691885e-05, "loss": 0.1665, "step": 9824 }, { "epoch": 0.63, "learning_rate": 6.361967649499667e-05, "loss": 0.1439, "step": 9825 }, { "epoch": 0.63, "learning_rate": 6.360033356200339e-05, "loss": 0.1666, "step": 9826 }, { "epoch": 0.63, "learning_rate": 6.358099219877327e-05, "loss": 0.1327, "step": 9827 }, { "epoch": 0.63, "learning_rate": 6.356165240614031e-05, "loss": 0.1602, "step": 9828 }, { "epoch": 0.63, "learning_rate": 6.354231418493861e-05, "loss": 0.1442, "step": 9829 }, { "epoch": 0.63, "learning_rate": 6.352297753600217e-05, "loss": 0.1398, "step": 9830 }, { "epoch": 0.63, "learning_rate": 6.350364246016483e-05, "loss": 0.1347, "step": 9831 }, { "epoch": 0.63, "learning_rate": 6.348430895826047e-05, "loss": 0.167, "step": 9832 }, { "epoch": 0.63, "learning_rate": 6.346497703112286e-05, "loss": 0.133, "step": 9833 }, { "epoch": 0.63, "learning_rate": 6.344564667958574e-05, "loss": 0.1483, "step": 9834 }, { "epoch": 0.63, "learning_rate": 6.342631790448269e-05, "loss": 0.1532, "step": 9835 }, { "epoch": 0.63, "learning_rate": 6.340699070664726e-05, "loss": 0.1548, "step": 9836 }, { "epoch": 0.63, "learning_rate": 6.338766508691303e-05, "loss": 0.1365, "step": 9837 }, { "epoch": 0.63, "learning_rate": 6.336834104611333e-05, "loss": 0.1483, "step": 9838 }, { "epoch": 0.63, "learning_rate": 6.334901858508167e-05, "loss": 0.1464, "step": 9839 }, { "epoch": 0.63, "learning_rate": 6.332969770465118e-05, "loss": 0.1438, "step": 9840 }, { "epoch": 0.63, "learning_rate": 6.33103784056552e-05, "loss": 0.1473, "step": 9841 }, { "epoch": 0.63, "learning_rate": 6.329106068892687e-05, "loss": 0.1522, "step": 9842 }, { "epoch": 0.63, "learning_rate": 6.327174455529923e-05, "loss": 0.1424, "step": 9843 }, { "epoch": 0.63, "learning_rate": 6.325243000560537e-05, "loss": 0.1358, "step": 9844 }, { "epoch": 0.63, "learning_rate": 6.323311704067818e-05, "loss": 0.1467, "step": 9845 }, { "epoch": 0.63, "learning_rate": 6.321380566135063e-05, "loss": 0.1361, "step": 9846 }, { "epoch": 0.63, "learning_rate": 6.319449586845545e-05, "loss": 0.1448, "step": 9847 }, { "epoch": 0.63, "learning_rate": 6.317518766282543e-05, "loss": 0.143, "step": 9848 }, { "epoch": 0.63, "learning_rate": 6.315588104529327e-05, "loss": 0.1355, "step": 9849 }, { "epoch": 0.63, "learning_rate": 6.313657601669154e-05, "loss": 0.1198, "step": 9850 }, { "epoch": 0.63, "learning_rate": 6.31172725778528e-05, "loss": 0.1386, "step": 9851 }, { "epoch": 0.63, "learning_rate": 6.309797072960952e-05, "loss": 0.1445, "step": 9852 }, { "epoch": 0.63, "learning_rate": 6.307867047279417e-05, "loss": 0.15, "step": 9853 }, { "epoch": 0.63, "learning_rate": 6.305937180823898e-05, "loss": 0.131, "step": 9854 }, { "epoch": 0.63, "learning_rate": 6.304007473677629e-05, "loss": 0.1688, "step": 9855 }, { "epoch": 0.63, "learning_rate": 6.30207792592383e-05, "loss": 0.139, "step": 9856 }, { "epoch": 0.63, "learning_rate": 6.30014853764571e-05, "loss": 0.1589, "step": 9857 }, { "epoch": 0.63, "learning_rate": 6.29821930892648e-05, "loss": 0.136, "step": 9858 }, { "epoch": 0.63, "learning_rate": 6.296290239849336e-05, "loss": 0.1451, "step": 9859 }, { "epoch": 0.63, "learning_rate": 6.294361330497474e-05, "loss": 0.1387, "step": 9860 }, { "epoch": 0.63, "learning_rate": 6.292432580954075e-05, "loss": 0.1376, "step": 9861 }, { "epoch": 0.63, "learning_rate": 6.290503991302324e-05, "loss": 0.1275, "step": 9862 }, { "epoch": 0.63, "learning_rate": 6.288575561625388e-05, "loss": 0.132, "step": 9863 }, { "epoch": 0.63, "learning_rate": 6.28664729200643e-05, "loss": 0.143, "step": 9864 }, { "epoch": 0.63, "learning_rate": 6.284719182528615e-05, "loss": 0.1385, "step": 9865 }, { "epoch": 0.63, "learning_rate": 6.282791233275089e-05, "loss": 0.1478, "step": 9866 }, { "epoch": 0.63, "learning_rate": 6.280863444329002e-05, "loss": 0.15, "step": 9867 }, { "epoch": 0.63, "learning_rate": 6.278935815773482e-05, "loss": 0.125, "step": 9868 }, { "epoch": 0.63, "learning_rate": 6.277008347691666e-05, "loss": 0.1422, "step": 9869 }, { "epoch": 0.63, "learning_rate": 6.275081040166679e-05, "loss": 0.161, "step": 9870 }, { "epoch": 0.63, "learning_rate": 6.273153893281631e-05, "loss": 0.1258, "step": 9871 }, { "epoch": 0.63, "learning_rate": 6.271226907119636e-05, "loss": 0.1556, "step": 9872 }, { "epoch": 0.63, "learning_rate": 6.269300081763795e-05, "loss": 0.1638, "step": 9873 }, { "epoch": 0.63, "learning_rate": 6.26737341729721e-05, "loss": 0.1493, "step": 9874 }, { "epoch": 0.63, "learning_rate": 6.26544691380296e-05, "loss": 0.151, "step": 9875 }, { "epoch": 0.63, "learning_rate": 6.263520571364134e-05, "loss": 0.1447, "step": 9876 }, { "epoch": 0.63, "learning_rate": 6.261594390063808e-05, "loss": 0.1421, "step": 9877 }, { "epoch": 0.63, "learning_rate": 6.25966836998504e-05, "loss": 0.1486, "step": 9878 }, { "epoch": 0.63, "learning_rate": 6.257742511210902e-05, "loss": 0.1438, "step": 9879 }, { "epoch": 0.63, "learning_rate": 6.25581681382444e-05, "loss": 0.1507, "step": 9880 }, { "epoch": 0.63, "learning_rate": 6.253891277908707e-05, "loss": 0.1196, "step": 9881 }, { "epoch": 0.63, "learning_rate": 6.251965903546741e-05, "loss": 0.1435, "step": 9882 }, { "epoch": 0.63, "learning_rate": 6.25004069082158e-05, "loss": 0.138, "step": 9883 }, { "epoch": 0.63, "learning_rate": 6.248115639816242e-05, "loss": 0.1279, "step": 9884 }, { "epoch": 0.63, "learning_rate": 6.246190750613748e-05, "loss": 0.1555, "step": 9885 }, { "epoch": 0.63, "learning_rate": 6.244266023297118e-05, "loss": 0.1521, "step": 9886 }, { "epoch": 0.63, "learning_rate": 6.242341457949345e-05, "loss": 0.1591, "step": 9887 }, { "epoch": 0.63, "learning_rate": 6.240417054653438e-05, "loss": 0.1399, "step": 9888 }, { "epoch": 0.63, "learning_rate": 6.238492813492387e-05, "loss": 0.1376, "step": 9889 }, { "epoch": 0.63, "learning_rate": 6.236568734549168e-05, "loss": 0.1373, "step": 9890 }, { "epoch": 0.63, "learning_rate": 6.234644817906767e-05, "loss": 0.139, "step": 9891 }, { "epoch": 0.63, "learning_rate": 6.232721063648148e-05, "loss": 0.1519, "step": 9892 }, { "epoch": 0.63, "learning_rate": 6.230797471856283e-05, "loss": 0.1489, "step": 9893 }, { "epoch": 0.63, "learning_rate": 6.228874042614119e-05, "loss": 0.1262, "step": 9894 }, { "epoch": 0.63, "learning_rate": 6.226950776004612e-05, "loss": 0.1495, "step": 9895 }, { "epoch": 0.63, "learning_rate": 6.225027672110702e-05, "loss": 0.165, "step": 9896 }, { "epoch": 0.63, "learning_rate": 6.223104731015321e-05, "loss": 0.1439, "step": 9897 }, { "epoch": 0.63, "learning_rate": 6.221181952801403e-05, "loss": 0.1411, "step": 9898 }, { "epoch": 0.63, "learning_rate": 6.219259337551863e-05, "loss": 0.131, "step": 9899 }, { "epoch": 0.63, "learning_rate": 6.217336885349625e-05, "loss": 0.1359, "step": 9900 }, { "epoch": 0.63, "learning_rate": 6.215414596277584e-05, "loss": 0.1377, "step": 9901 }, { "epoch": 0.63, "learning_rate": 6.21349247041865e-05, "loss": 0.1205, "step": 9902 }, { "epoch": 0.63, "learning_rate": 6.211570507855713e-05, "loss": 0.1432, "step": 9903 }, { "epoch": 0.64, "learning_rate": 6.209648708671654e-05, "loss": 0.1536, "step": 9904 }, { "epoch": 0.64, "learning_rate": 6.20772707294936e-05, "loss": 0.1645, "step": 9905 }, { "epoch": 0.64, "learning_rate": 6.205805600771697e-05, "loss": 0.1273, "step": 9906 }, { "epoch": 0.64, "learning_rate": 6.203884292221536e-05, "loss": 0.1553, "step": 9907 }, { "epoch": 0.64, "learning_rate": 6.201963147381726e-05, "loss": 0.1644, "step": 9908 }, { "epoch": 0.64, "learning_rate": 6.200042166335127e-05, "loss": 0.1538, "step": 9909 }, { "epoch": 0.64, "learning_rate": 6.19812134916458e-05, "loss": 0.1496, "step": 9910 }, { "epoch": 0.64, "learning_rate": 6.196200695952919e-05, "loss": 0.1454, "step": 9911 }, { "epoch": 0.64, "learning_rate": 6.194280206782974e-05, "loss": 0.1535, "step": 9912 }, { "epoch": 0.64, "learning_rate": 6.192359881737569e-05, "loss": 0.143, "step": 9913 }, { "epoch": 0.64, "learning_rate": 6.190439720899521e-05, "loss": 0.1374, "step": 9914 }, { "epoch": 0.64, "learning_rate": 6.188519724351634e-05, "loss": 0.1245, "step": 9915 }, { "epoch": 0.64, "learning_rate": 6.186599892176713e-05, "loss": 0.1225, "step": 9916 }, { "epoch": 0.64, "learning_rate": 6.184680224457553e-05, "loss": 0.132, "step": 9917 }, { "epoch": 0.64, "learning_rate": 6.182760721276937e-05, "loss": 0.1452, "step": 9918 }, { "epoch": 0.64, "learning_rate": 6.180841382717647e-05, "loss": 0.141, "step": 9919 }, { "epoch": 0.64, "learning_rate": 6.178922208862455e-05, "loss": 0.1352, "step": 9920 }, { "epoch": 0.64, "learning_rate": 6.177003199794132e-05, "loss": 0.1423, "step": 9921 }, { "epoch": 0.64, "learning_rate": 6.175084355595427e-05, "loss": 0.1344, "step": 9922 }, { "epoch": 0.64, "learning_rate": 6.173165676349103e-05, "loss": 0.15, "step": 9923 }, { "epoch": 0.64, "learning_rate": 6.171247162137899e-05, "loss": 0.1436, "step": 9924 }, { "epoch": 0.64, "learning_rate": 6.169328813044545e-05, "loss": 0.1437, "step": 9925 }, { "epoch": 0.64, "learning_rate": 6.167410629151786e-05, "loss": 0.1411, "step": 9926 }, { "epoch": 0.64, "learning_rate": 6.165492610542334e-05, "loss": 0.1445, "step": 9927 }, { "epoch": 0.64, "learning_rate": 6.163574757298908e-05, "loss": 0.1584, "step": 9928 }, { "epoch": 0.64, "learning_rate": 6.161657069504218e-05, "loss": 0.13, "step": 9929 }, { "epoch": 0.64, "learning_rate": 6.159739547240968e-05, "loss": 0.1511, "step": 9930 }, { "epoch": 0.64, "learning_rate": 6.157822190591848e-05, "loss": 0.1405, "step": 9931 }, { "epoch": 0.64, "learning_rate": 6.155904999639546e-05, "loss": 0.1392, "step": 9932 }, { "epoch": 0.64, "learning_rate": 6.153987974466746e-05, "loss": 0.1392, "step": 9933 }, { "epoch": 0.64, "learning_rate": 6.152071115156115e-05, "loss": 0.1327, "step": 9934 }, { "epoch": 0.64, "learning_rate": 6.150154421790324e-05, "loss": 0.1403, "step": 9935 }, { "epoch": 0.64, "learning_rate": 6.14823789445203e-05, "loss": 0.1412, "step": 9936 }, { "epoch": 0.64, "learning_rate": 6.146321533223889e-05, "loss": 0.153, "step": 9937 }, { "epoch": 0.64, "learning_rate": 6.14440533818854e-05, "loss": 0.137, "step": 9938 }, { "epoch": 0.64, "learning_rate": 6.142489309428619e-05, "loss": 0.1454, "step": 9939 }, { "epoch": 0.64, "learning_rate": 6.140573447026763e-05, "loss": 0.1349, "step": 9940 }, { "epoch": 0.64, "learning_rate": 6.138657751065589e-05, "loss": 0.1423, "step": 9941 }, { "epoch": 0.64, "learning_rate": 6.136742221627714e-05, "loss": 0.1414, "step": 9942 }, { "epoch": 0.64, "learning_rate": 6.134826858795748e-05, "loss": 0.1389, "step": 9943 }, { "epoch": 0.64, "learning_rate": 6.132911662652297e-05, "loss": 0.174, "step": 9944 }, { "epoch": 0.64, "learning_rate": 6.130996633279947e-05, "loss": 0.131, "step": 9945 }, { "epoch": 0.64, "learning_rate": 6.129081770761288e-05, "loss": 0.1307, "step": 9946 }, { "epoch": 0.64, "learning_rate": 6.127167075178904e-05, "loss": 0.1565, "step": 9947 }, { "epoch": 0.64, "learning_rate": 6.125252546615358e-05, "loss": 0.1347, "step": 9948 }, { "epoch": 0.64, "learning_rate": 6.123338185153226e-05, "loss": 0.1356, "step": 9949 }, { "epoch": 0.64, "learning_rate": 6.121423990875061e-05, "loss": 0.1739, "step": 9950 }, { "epoch": 0.64, "learning_rate": 6.119509963863416e-05, "loss": 0.1511, "step": 9951 }, { "epoch": 0.64, "learning_rate": 6.117596104200834e-05, "loss": 0.1456, "step": 9952 }, { "epoch": 0.64, "learning_rate": 6.115682411969849e-05, "loss": 0.1328, "step": 9953 }, { "epoch": 0.64, "learning_rate": 6.113768887252998e-05, "loss": 0.145, "step": 9954 }, { "epoch": 0.64, "learning_rate": 6.111855530132793e-05, "loss": 0.1578, "step": 9955 }, { "epoch": 0.64, "learning_rate": 6.109942340691756e-05, "loss": 0.1426, "step": 9956 }, { "epoch": 0.64, "learning_rate": 6.108029319012392e-05, "loss": 0.152, "step": 9957 }, { "epoch": 0.64, "learning_rate": 6.106116465177205e-05, "loss": 0.1551, "step": 9958 }, { "epoch": 0.64, "learning_rate": 6.104203779268684e-05, "loss": 0.131, "step": 9959 }, { "epoch": 0.64, "learning_rate": 6.102291261369315e-05, "loss": 0.1618, "step": 9960 }, { "epoch": 0.64, "learning_rate": 6.100378911561582e-05, "loss": 0.1472, "step": 9961 }, { "epoch": 0.64, "learning_rate": 6.09846672992795e-05, "loss": 0.1318, "step": 9962 }, { "epoch": 0.64, "learning_rate": 6.0965547165508875e-05, "loss": 0.1307, "step": 9963 }, { "epoch": 0.64, "learning_rate": 6.0946428715128524e-05, "loss": 0.1682, "step": 9964 }, { "epoch": 0.64, "learning_rate": 6.0927311948962885e-05, "loss": 0.15, "step": 9965 }, { "epoch": 0.64, "learning_rate": 6.0908196867836446e-05, "loss": 0.1591, "step": 9966 }, { "epoch": 0.64, "learning_rate": 6.088908347257349e-05, "loss": 0.1439, "step": 9967 }, { "epoch": 0.64, "learning_rate": 6.086997176399838e-05, "loss": 0.1266, "step": 9968 }, { "epoch": 0.64, "learning_rate": 6.085086174293526e-05, "loss": 0.1452, "step": 9969 }, { "epoch": 0.64, "learning_rate": 6.083175341020833e-05, "loss": 0.1279, "step": 9970 }, { "epoch": 0.64, "learning_rate": 6.08126467666416e-05, "loss": 0.1349, "step": 9971 }, { "epoch": 0.64, "learning_rate": 6.079354181305904e-05, "loss": 0.1376, "step": 9972 }, { "epoch": 0.64, "learning_rate": 6.077443855028463e-05, "loss": 0.1474, "step": 9973 }, { "epoch": 0.64, "learning_rate": 6.0755336979142144e-05, "loss": 0.1411, "step": 9974 }, { "epoch": 0.64, "learning_rate": 6.0736237100455395e-05, "loss": 0.152, "step": 9975 }, { "epoch": 0.64, "learning_rate": 6.0717138915048045e-05, "loss": 0.1409, "step": 9976 }, { "epoch": 0.64, "learning_rate": 6.069804242374379e-05, "loss": 0.1352, "step": 9977 }, { "epoch": 0.64, "learning_rate": 6.0678947627366114e-05, "loss": 0.1426, "step": 9978 }, { "epoch": 0.64, "learning_rate": 6.065985452673848e-05, "loss": 0.1398, "step": 9979 }, { "epoch": 0.64, "learning_rate": 6.064076312268438e-05, "loss": 0.1458, "step": 9980 }, { "epoch": 0.64, "learning_rate": 6.062167341602704e-05, "loss": 0.1335, "step": 9981 }, { "epoch": 0.64, "learning_rate": 6.060258540758979e-05, "loss": 0.1343, "step": 9982 }, { "epoch": 0.64, "learning_rate": 6.0583499098195775e-05, "loss": 0.1316, "step": 9983 }, { "epoch": 0.64, "learning_rate": 6.0564414488668165e-05, "loss": 0.1504, "step": 9984 }, { "epoch": 0.64, "learning_rate": 6.0545331579829935e-05, "loss": 0.1346, "step": 9985 }, { "epoch": 0.64, "learning_rate": 6.0526250372504036e-05, "loss": 0.1541, "step": 9986 }, { "epoch": 0.64, "learning_rate": 6.050717086751345e-05, "loss": 0.1616, "step": 9987 }, { "epoch": 0.64, "learning_rate": 6.048809306568091e-05, "loss": 0.1529, "step": 9988 }, { "epoch": 0.64, "learning_rate": 6.046901696782921e-05, "loss": 0.1479, "step": 9989 }, { "epoch": 0.64, "learning_rate": 6.044994257478096e-05, "loss": 0.1579, "step": 9990 }, { "epoch": 0.64, "learning_rate": 6.0430869887358864e-05, "loss": 0.1471, "step": 9991 }, { "epoch": 0.64, "learning_rate": 6.041179890638535e-05, "loss": 0.136, "step": 9992 }, { "epoch": 0.64, "learning_rate": 6.039272963268288e-05, "loss": 0.1399, "step": 9993 }, { "epoch": 0.64, "learning_rate": 6.03736620670739e-05, "loss": 0.1359, "step": 9994 }, { "epoch": 0.64, "learning_rate": 6.035459621038062e-05, "loss": 0.1305, "step": 9995 }, { "epoch": 0.64, "learning_rate": 6.0335532063425325e-05, "loss": 0.1387, "step": 9996 }, { "epoch": 0.64, "learning_rate": 6.031646962703016e-05, "loss": 0.1173, "step": 9997 }, { "epoch": 0.64, "learning_rate": 6.029740890201723e-05, "loss": 0.1341, "step": 9998 }, { "epoch": 0.64, "learning_rate": 6.027834988920851e-05, "loss": 0.1267, "step": 9999 }, { "epoch": 0.64, "learning_rate": 6.025929258942591e-05, "loss": 0.1327, "step": 10000 }, { "epoch": 0.64, "learning_rate": 6.0240237003491376e-05, "loss": 0.137, "step": 10001 }, { "epoch": 0.64, "learning_rate": 6.02211831322266e-05, "loss": 0.1532, "step": 10002 }, { "epoch": 0.64, "learning_rate": 6.020213097645336e-05, "loss": 0.1416, "step": 10003 }, { "epoch": 0.64, "learning_rate": 6.0183080536993256e-05, "loss": 0.1574, "step": 10004 }, { "epoch": 0.64, "learning_rate": 6.01640318146679e-05, "loss": 0.1587, "step": 10005 }, { "epoch": 0.64, "learning_rate": 6.014498481029873e-05, "loss": 0.1625, "step": 10006 }, { "epoch": 0.64, "learning_rate": 6.0125939524707176e-05, "loss": 0.1418, "step": 10007 }, { "epoch": 0.64, "learning_rate": 6.0106895958714626e-05, "loss": 0.1249, "step": 10008 }, { "epoch": 0.64, "learning_rate": 6.008785411314224e-05, "loss": 0.1394, "step": 10009 }, { "epoch": 0.64, "learning_rate": 6.006881398881135e-05, "loss": 0.1509, "step": 10010 }, { "epoch": 0.64, "learning_rate": 6.004977558654297e-05, "loss": 0.1176, "step": 10011 }, { "epoch": 0.64, "learning_rate": 6.0030738907158204e-05, "loss": 0.1262, "step": 10012 }, { "epoch": 0.64, "learning_rate": 6.0011703951478013e-05, "loss": 0.1396, "step": 10013 }, { "epoch": 0.64, "learning_rate": 5.999267072032325e-05, "loss": 0.14, "step": 10014 }, { "epoch": 0.64, "learning_rate": 5.99736392145148e-05, "loss": 0.1423, "step": 10015 }, { "epoch": 0.64, "learning_rate": 5.9954609434873344e-05, "loss": 0.1341, "step": 10016 }, { "epoch": 0.64, "learning_rate": 5.9935581382219644e-05, "loss": 0.1762, "step": 10017 }, { "epoch": 0.64, "learning_rate": 5.9916555057374216e-05, "loss": 0.1411, "step": 10018 }, { "epoch": 0.64, "learning_rate": 5.9897530461157624e-05, "loss": 0.1417, "step": 10019 }, { "epoch": 0.64, "learning_rate": 5.987850759439033e-05, "loss": 0.1267, "step": 10020 }, { "epoch": 0.64, "learning_rate": 5.9859486457892655e-05, "loss": 0.1636, "step": 10021 }, { "epoch": 0.64, "learning_rate": 5.984046705248496e-05, "loss": 0.1542, "step": 10022 }, { "epoch": 0.64, "learning_rate": 5.982144937898743e-05, "loss": 0.1597, "step": 10023 }, { "epoch": 0.64, "learning_rate": 5.980243343822026e-05, "loss": 0.1306, "step": 10024 }, { "epoch": 0.64, "learning_rate": 5.978341923100348e-05, "loss": 0.1368, "step": 10025 }, { "epoch": 0.64, "learning_rate": 5.976440675815711e-05, "loss": 0.1468, "step": 10026 }, { "epoch": 0.64, "learning_rate": 5.974539602050112e-05, "loss": 0.1444, "step": 10027 }, { "epoch": 0.64, "learning_rate": 5.972638701885527e-05, "loss": 0.1423, "step": 10028 }, { "epoch": 0.64, "learning_rate": 5.9707379754039415e-05, "loss": 0.1573, "step": 10029 }, { "epoch": 0.64, "learning_rate": 5.968837422687321e-05, "loss": 0.1482, "step": 10030 }, { "epoch": 0.64, "learning_rate": 5.966937043817635e-05, "loss": 0.1308, "step": 10031 }, { "epoch": 0.64, "learning_rate": 5.9650368388768295e-05, "loss": 0.1396, "step": 10032 }, { "epoch": 0.64, "learning_rate": 5.9631368079468596e-05, "loss": 0.1666, "step": 10033 }, { "epoch": 0.64, "learning_rate": 5.961236951109665e-05, "loss": 0.1297, "step": 10034 }, { "epoch": 0.64, "learning_rate": 5.959337268447173e-05, "loss": 0.1471, "step": 10035 }, { "epoch": 0.64, "learning_rate": 5.957437760041313e-05, "loss": 0.1621, "step": 10036 }, { "epoch": 0.64, "learning_rate": 5.955538425974001e-05, "loss": 0.1487, "step": 10037 }, { "epoch": 0.64, "learning_rate": 5.9536392663271514e-05, "loss": 0.1684, "step": 10038 }, { "epoch": 0.64, "learning_rate": 5.9517402811826605e-05, "loss": 0.1579, "step": 10039 }, { "epoch": 0.64, "learning_rate": 5.9498414706224295e-05, "loss": 0.1247, "step": 10040 }, { "epoch": 0.64, "learning_rate": 5.947942834728344e-05, "loss": 0.1458, "step": 10041 }, { "epoch": 0.64, "learning_rate": 5.94604437358228e-05, "loss": 0.1481, "step": 10042 }, { "epoch": 0.64, "learning_rate": 5.9441460872661156e-05, "loss": 0.1263, "step": 10043 }, { "epoch": 0.64, "learning_rate": 5.942247975861711e-05, "loss": 0.1397, "step": 10044 }, { "epoch": 0.64, "learning_rate": 5.94035003945093e-05, "loss": 0.1252, "step": 10045 }, { "epoch": 0.64, "learning_rate": 5.938452278115618e-05, "loss": 0.1371, "step": 10046 }, { "epoch": 0.64, "learning_rate": 5.9365546919376166e-05, "loss": 0.1575, "step": 10047 }, { "epoch": 0.64, "learning_rate": 5.934657280998765e-05, "loss": 0.128, "step": 10048 }, { "epoch": 0.64, "learning_rate": 5.932760045380885e-05, "loss": 0.1451, "step": 10049 }, { "epoch": 0.64, "learning_rate": 5.9308629851658014e-05, "loss": 0.1395, "step": 10050 }, { "epoch": 0.64, "learning_rate": 5.9289661004353217e-05, "loss": 0.1394, "step": 10051 }, { "epoch": 0.64, "learning_rate": 5.927069391271256e-05, "loss": 0.1345, "step": 10052 }, { "epoch": 0.64, "learning_rate": 5.9251728577553966e-05, "loss": 0.1324, "step": 10053 }, { "epoch": 0.64, "learning_rate": 5.9232764999695337e-05, "loss": 0.1391, "step": 10054 }, { "epoch": 0.64, "learning_rate": 5.921380317995451e-05, "loss": 0.1569, "step": 10055 }, { "epoch": 0.64, "learning_rate": 5.91948431191492e-05, "loss": 0.1528, "step": 10056 }, { "epoch": 0.64, "learning_rate": 5.917588481809713e-05, "loss": 0.1438, "step": 10057 }, { "epoch": 0.64, "learning_rate": 5.915692827761581e-05, "loss": 0.1582, "step": 10058 }, { "epoch": 0.64, "learning_rate": 5.913797349852283e-05, "loss": 0.1376, "step": 10059 }, { "epoch": 0.65, "learning_rate": 5.91190204816356e-05, "loss": 0.1239, "step": 10060 }, { "epoch": 0.65, "learning_rate": 5.910006922777145e-05, "loss": 0.1408, "step": 10061 }, { "epoch": 0.65, "learning_rate": 5.9081119737747724e-05, "loss": 0.1418, "step": 10062 }, { "epoch": 0.65, "learning_rate": 5.906217201238158e-05, "loss": 0.145, "step": 10063 }, { "epoch": 0.65, "learning_rate": 5.904322605249022e-05, "loss": 0.1268, "step": 10064 }, { "epoch": 0.65, "learning_rate": 5.902428185889062e-05, "loss": 0.1271, "step": 10065 }, { "epoch": 0.65, "learning_rate": 5.900533943239981e-05, "loss": 0.1611, "step": 10066 }, { "epoch": 0.65, "learning_rate": 5.898639877383474e-05, "loss": 0.147, "step": 10067 }, { "epoch": 0.65, "learning_rate": 5.896745988401213e-05, "loss": 0.1498, "step": 10068 }, { "epoch": 0.65, "learning_rate": 5.8948522763748826e-05, "loss": 0.1356, "step": 10069 }, { "epoch": 0.65, "learning_rate": 5.8929587413861454e-05, "loss": 0.1482, "step": 10070 }, { "epoch": 0.65, "learning_rate": 5.8910653835166676e-05, "loss": 0.1385, "step": 10071 }, { "epoch": 0.65, "learning_rate": 5.889172202848093e-05, "loss": 0.1472, "step": 10072 }, { "epoch": 0.65, "learning_rate": 5.8872791994620746e-05, "loss": 0.1293, "step": 10073 }, { "epoch": 0.65, "learning_rate": 5.885386373440247e-05, "loss": 0.1412, "step": 10074 }, { "epoch": 0.65, "learning_rate": 5.883493724864235e-05, "loss": 0.1548, "step": 10075 }, { "epoch": 0.65, "learning_rate": 5.881601253815667e-05, "loss": 0.1402, "step": 10076 }, { "epoch": 0.65, "learning_rate": 5.8797089603761524e-05, "loss": 0.1313, "step": 10077 }, { "epoch": 0.65, "learning_rate": 5.877816844627304e-05, "loss": 0.1457, "step": 10078 }, { "epoch": 0.65, "learning_rate": 5.875924906650713e-05, "loss": 0.157, "step": 10079 }, { "epoch": 0.65, "learning_rate": 5.874033146527976e-05, "loss": 0.131, "step": 10080 }, { "epoch": 0.65, "learning_rate": 5.872141564340677e-05, "loss": 0.1471, "step": 10081 }, { "epoch": 0.65, "learning_rate": 5.8702501601703866e-05, "loss": 0.1402, "step": 10082 }, { "epoch": 0.65, "learning_rate": 5.868358934098678e-05, "loss": 0.1324, "step": 10083 }, { "epoch": 0.65, "learning_rate": 5.8664678862071076e-05, "loss": 0.1274, "step": 10084 }, { "epoch": 0.65, "learning_rate": 5.864577016577235e-05, "loss": 0.1411, "step": 10085 }, { "epoch": 0.65, "learning_rate": 5.8626863252905984e-05, "loss": 0.1581, "step": 10086 }, { "epoch": 0.65, "learning_rate": 5.8607958124287384e-05, "loss": 0.1391, "step": 10087 }, { "epoch": 0.65, "learning_rate": 5.858905478073187e-05, "loss": 0.1628, "step": 10088 }, { "epoch": 0.65, "learning_rate": 5.857015322305459e-05, "loss": 0.1494, "step": 10089 }, { "epoch": 0.65, "learning_rate": 5.855125345207076e-05, "loss": 0.1274, "step": 10090 }, { "epoch": 0.65, "learning_rate": 5.853235546859541e-05, "loss": 0.144, "step": 10091 }, { "epoch": 0.65, "learning_rate": 5.851345927344358e-05, "loss": 0.1376, "step": 10092 }, { "epoch": 0.65, "learning_rate": 5.849456486743008e-05, "loss": 0.1359, "step": 10093 }, { "epoch": 0.65, "learning_rate": 5.847567225136987e-05, "loss": 0.1207, "step": 10094 }, { "epoch": 0.65, "learning_rate": 5.8456781426077644e-05, "loss": 0.1404, "step": 10095 }, { "epoch": 0.65, "learning_rate": 5.843789239236803e-05, "loss": 0.15, "step": 10096 }, { "epoch": 0.65, "learning_rate": 5.841900515105576e-05, "loss": 0.127, "step": 10097 }, { "epoch": 0.65, "learning_rate": 5.840011970295525e-05, "loss": 0.135, "step": 10098 }, { "epoch": 0.65, "learning_rate": 5.838123604888102e-05, "loss": 0.1452, "step": 10099 }, { "epoch": 0.65, "learning_rate": 5.836235418964737e-05, "loss": 0.1377, "step": 10100 }, { "epoch": 0.65, "learning_rate": 5.834347412606864e-05, "loss": 0.1383, "step": 10101 }, { "epoch": 0.65, "learning_rate": 5.832459585895908e-05, "loss": 0.1322, "step": 10102 }, { "epoch": 0.65, "learning_rate": 5.830571938913275e-05, "loss": 0.1571, "step": 10103 }, { "epoch": 0.65, "learning_rate": 5.82868447174038e-05, "loss": 0.1467, "step": 10104 }, { "epoch": 0.65, "learning_rate": 5.826797184458612e-05, "loss": 0.1483, "step": 10105 }, { "epoch": 0.65, "learning_rate": 5.824910077149371e-05, "loss": 0.1266, "step": 10106 }, { "epoch": 0.65, "learning_rate": 5.8230231498940314e-05, "loss": 0.1574, "step": 10107 }, { "epoch": 0.65, "learning_rate": 5.8211364027739724e-05, "loss": 0.1608, "step": 10108 }, { "epoch": 0.65, "learning_rate": 5.819249835870566e-05, "loss": 0.1361, "step": 10109 }, { "epoch": 0.65, "learning_rate": 5.817363449265163e-05, "loss": 0.1342, "step": 10110 }, { "epoch": 0.65, "learning_rate": 5.8154772430391235e-05, "loss": 0.1388, "step": 10111 }, { "epoch": 0.65, "learning_rate": 5.813591217273784e-05, "loss": 0.1357, "step": 10112 }, { "epoch": 0.65, "learning_rate": 5.8117053720504864e-05, "loss": 0.1575, "step": 10113 }, { "epoch": 0.65, "learning_rate": 5.8098197074505544e-05, "loss": 0.1333, "step": 10114 }, { "epoch": 0.65, "learning_rate": 5.807934223555311e-05, "loss": 0.1385, "step": 10115 }, { "epoch": 0.65, "learning_rate": 5.8060489204460744e-05, "loss": 0.1409, "step": 10116 }, { "epoch": 0.65, "learning_rate": 5.8041637982041405e-05, "loss": 0.1362, "step": 10117 }, { "epoch": 0.65, "learning_rate": 5.802278856910815e-05, "loss": 0.1392, "step": 10118 }, { "epoch": 0.65, "learning_rate": 5.80039409664738e-05, "loss": 0.1278, "step": 10119 }, { "epoch": 0.65, "learning_rate": 5.798509517495124e-05, "loss": 0.1333, "step": 10120 }, { "epoch": 0.65, "learning_rate": 5.796625119535314e-05, "loss": 0.1328, "step": 10121 }, { "epoch": 0.65, "learning_rate": 5.7947409028492194e-05, "loss": 0.1574, "step": 10122 }, { "epoch": 0.65, "learning_rate": 5.792856867518101e-05, "loss": 0.1424, "step": 10123 }, { "epoch": 0.65, "learning_rate": 5.7909730136232064e-05, "loss": 0.125, "step": 10124 }, { "epoch": 0.65, "learning_rate": 5.78908934124578e-05, "loss": 0.1475, "step": 10125 }, { "epoch": 0.65, "learning_rate": 5.787205850467052e-05, "loss": 0.1641, "step": 10126 }, { "epoch": 0.65, "learning_rate": 5.785322541368257e-05, "loss": 0.1649, "step": 10127 }, { "epoch": 0.65, "learning_rate": 5.783439414030606e-05, "loss": 0.139, "step": 10128 }, { "epoch": 0.65, "learning_rate": 5.781556468535314e-05, "loss": 0.1238, "step": 10129 }, { "epoch": 0.65, "learning_rate": 5.779673704963588e-05, "loss": 0.1474, "step": 10130 }, { "epoch": 0.65, "learning_rate": 5.7777911233966165e-05, "loss": 0.136, "step": 10131 }, { "epoch": 0.65, "learning_rate": 5.775908723915595e-05, "loss": 0.1303, "step": 10132 }, { "epoch": 0.65, "learning_rate": 5.774026506601695e-05, "loss": 0.156, "step": 10133 }, { "epoch": 0.65, "learning_rate": 5.7721444715360984e-05, "loss": 0.1438, "step": 10134 }, { "epoch": 0.65, "learning_rate": 5.7702626187999576e-05, "loss": 0.1496, "step": 10135 }, { "epoch": 0.65, "learning_rate": 5.768380948474437e-05, "loss": 0.1527, "step": 10136 }, { "epoch": 0.65, "learning_rate": 5.766499460640687e-05, "loss": 0.1616, "step": 10137 }, { "epoch": 0.65, "learning_rate": 5.764618155379839e-05, "loss": 0.1356, "step": 10138 }, { "epoch": 0.65, "learning_rate": 5.762737032773036e-05, "loss": 0.1508, "step": 10139 }, { "epoch": 0.65, "learning_rate": 5.7608560929013946e-05, "loss": 0.1321, "step": 10140 }, { "epoch": 0.65, "learning_rate": 5.758975335846033e-05, "loss": 0.1406, "step": 10141 }, { "epoch": 0.65, "learning_rate": 5.757094761688069e-05, "loss": 0.1387, "step": 10142 }, { "epoch": 0.65, "learning_rate": 5.755214370508592e-05, "loss": 0.1551, "step": 10143 }, { "epoch": 0.65, "learning_rate": 5.7533341623887036e-05, "loss": 0.1296, "step": 10144 }, { "epoch": 0.65, "learning_rate": 5.751454137409483e-05, "loss": 0.1438, "step": 10145 }, { "epoch": 0.65, "learning_rate": 5.7495742956520135e-05, "loss": 0.1392, "step": 10146 }, { "epoch": 0.65, "learning_rate": 5.747694637197361e-05, "loss": 0.1308, "step": 10147 }, { "epoch": 0.65, "learning_rate": 5.745815162126584e-05, "loss": 0.1561, "step": 10148 }, { "epoch": 0.65, "learning_rate": 5.743935870520747e-05, "loss": 0.1407, "step": 10149 }, { "epoch": 0.65, "learning_rate": 5.742056762460886e-05, "loss": 0.1356, "step": 10150 }, { "epoch": 0.65, "learning_rate": 5.7401778380280446e-05, "loss": 0.1414, "step": 10151 }, { "epoch": 0.65, "learning_rate": 5.7382990973032454e-05, "loss": 0.1239, "step": 10152 }, { "epoch": 0.65, "learning_rate": 5.736420540367522e-05, "loss": 0.1484, "step": 10153 }, { "epoch": 0.65, "learning_rate": 5.7345421673018775e-05, "loss": 0.1481, "step": 10154 }, { "epoch": 0.65, "learning_rate": 5.7326639781873224e-05, "loss": 0.1479, "step": 10155 }, { "epoch": 0.65, "learning_rate": 5.73078597310486e-05, "loss": 0.149, "step": 10156 }, { "epoch": 0.65, "learning_rate": 5.72890815213547e-05, "loss": 0.1443, "step": 10157 }, { "epoch": 0.65, "learning_rate": 5.7270305153601476e-05, "loss": 0.1366, "step": 10158 }, { "epoch": 0.65, "learning_rate": 5.725153062859856e-05, "loss": 0.1564, "step": 10159 }, { "epoch": 0.65, "learning_rate": 5.72327579471557e-05, "loss": 0.1546, "step": 10160 }, { "epoch": 0.65, "learning_rate": 5.72139871100824e-05, "loss": 0.137, "step": 10161 }, { "epoch": 0.65, "learning_rate": 5.719521811818822e-05, "loss": 0.1403, "step": 10162 }, { "epoch": 0.65, "learning_rate": 5.717645097228261e-05, "loss": 0.1601, "step": 10163 }, { "epoch": 0.65, "learning_rate": 5.715768567317486e-05, "loss": 0.1214, "step": 10164 }, { "epoch": 0.65, "learning_rate": 5.7138922221674305e-05, "loss": 0.1411, "step": 10165 }, { "epoch": 0.65, "learning_rate": 5.7120160618590045e-05, "loss": 0.1324, "step": 10166 }, { "epoch": 0.65, "learning_rate": 5.710140086473129e-05, "loss": 0.1345, "step": 10167 }, { "epoch": 0.65, "learning_rate": 5.708264296090696e-05, "loss": 0.1416, "step": 10168 }, { "epoch": 0.65, "learning_rate": 5.706388690792607e-05, "loss": 0.135, "step": 10169 }, { "epoch": 0.65, "learning_rate": 5.70451327065975e-05, "loss": 0.1348, "step": 10170 }, { "epoch": 0.65, "learning_rate": 5.7026380357730005e-05, "loss": 0.1337, "step": 10171 }, { "epoch": 0.65, "learning_rate": 5.7007629862132325e-05, "loss": 0.1561, "step": 10172 }, { "epoch": 0.65, "learning_rate": 5.698888122061303e-05, "loss": 0.1447, "step": 10173 }, { "epoch": 0.65, "learning_rate": 5.697013443398076e-05, "loss": 0.1505, "step": 10174 }, { "epoch": 0.65, "learning_rate": 5.6951389503043884e-05, "loss": 0.1347, "step": 10175 }, { "epoch": 0.65, "learning_rate": 5.6932646428610845e-05, "loss": 0.137, "step": 10176 }, { "epoch": 0.65, "learning_rate": 5.691390521148998e-05, "loss": 0.1426, "step": 10177 }, { "epoch": 0.65, "learning_rate": 5.689516585248946e-05, "loss": 0.134, "step": 10178 }, { "epoch": 0.65, "learning_rate": 5.687642835241749e-05, "loss": 0.1428, "step": 10179 }, { "epoch": 0.65, "learning_rate": 5.685769271208208e-05, "loss": 0.1468, "step": 10180 }, { "epoch": 0.65, "learning_rate": 5.683895893229127e-05, "loss": 0.1599, "step": 10181 }, { "epoch": 0.65, "learning_rate": 5.682022701385288e-05, "loss": 0.1486, "step": 10182 }, { "epoch": 0.65, "learning_rate": 5.6801496957574884e-05, "loss": 0.1561, "step": 10183 }, { "epoch": 0.65, "learning_rate": 5.678276876426494e-05, "loss": 0.1477, "step": 10184 }, { "epoch": 0.65, "learning_rate": 5.6764042434730704e-05, "loss": 0.1492, "step": 10185 }, { "epoch": 0.65, "learning_rate": 5.67453179697798e-05, "loss": 0.1661, "step": 10186 }, { "epoch": 0.65, "learning_rate": 5.672659537021969e-05, "loss": 0.1371, "step": 10187 }, { "epoch": 0.65, "learning_rate": 5.670787463685785e-05, "loss": 0.1385, "step": 10188 }, { "epoch": 0.65, "learning_rate": 5.668915577050158e-05, "loss": 0.119, "step": 10189 }, { "epoch": 0.65, "learning_rate": 5.667043877195821e-05, "loss": 0.1571, "step": 10190 }, { "epoch": 0.65, "learning_rate": 5.66517236420349e-05, "loss": 0.1383, "step": 10191 }, { "epoch": 0.65, "learning_rate": 5.66330103815387e-05, "loss": 0.1348, "step": 10192 }, { "epoch": 0.65, "learning_rate": 5.661429899127671e-05, "loss": 0.1556, "step": 10193 }, { "epoch": 0.65, "learning_rate": 5.6595589472055785e-05, "loss": 0.1481, "step": 10194 }, { "epoch": 0.65, "learning_rate": 5.657688182468285e-05, "loss": 0.1381, "step": 10195 }, { "epoch": 0.65, "learning_rate": 5.6558176049964676e-05, "loss": 0.1345, "step": 10196 }, { "epoch": 0.65, "learning_rate": 5.6539472148707984e-05, "loss": 0.1549, "step": 10197 }, { "epoch": 0.65, "learning_rate": 5.6520770121719394e-05, "loss": 0.1379, "step": 10198 }, { "epoch": 0.65, "learning_rate": 5.650206996980536e-05, "loss": 0.1363, "step": 10199 }, { "epoch": 0.65, "learning_rate": 5.648337169377247e-05, "loss": 0.1456, "step": 10200 }, { "epoch": 0.65, "learning_rate": 5.646467529442697e-05, "loss": 0.1319, "step": 10201 }, { "epoch": 0.65, "learning_rate": 5.6445980772575234e-05, "loss": 0.1507, "step": 10202 }, { "epoch": 0.65, "learning_rate": 5.6427288129023496e-05, "loss": 0.1379, "step": 10203 }, { "epoch": 0.65, "learning_rate": 5.640859736457782e-05, "loss": 0.1485, "step": 10204 }, { "epoch": 0.65, "learning_rate": 5.638990848004435e-05, "loss": 0.1558, "step": 10205 }, { "epoch": 0.65, "learning_rate": 5.637122147622895e-05, "loss": 0.1392, "step": 10206 }, { "epoch": 0.65, "learning_rate": 5.635253635393762e-05, "loss": 0.1428, "step": 10207 }, { "epoch": 0.65, "learning_rate": 5.633385311397607e-05, "loss": 0.1572, "step": 10208 }, { "epoch": 0.65, "learning_rate": 5.6315171757150087e-05, "loss": 0.1447, "step": 10209 }, { "epoch": 0.65, "learning_rate": 5.629649228426533e-05, "loss": 0.1577, "step": 10210 }, { "epoch": 0.65, "learning_rate": 5.6277814696127316e-05, "loss": 0.1293, "step": 10211 }, { "epoch": 0.65, "learning_rate": 5.625913899354158e-05, "loss": 0.1511, "step": 10212 }, { "epoch": 0.65, "learning_rate": 5.624046517731348e-05, "loss": 0.1393, "step": 10213 }, { "epoch": 0.65, "learning_rate": 5.622179324824841e-05, "loss": 0.1424, "step": 10214 }, { "epoch": 0.65, "learning_rate": 5.620312320715151e-05, "loss": 0.1447, "step": 10215 }, { "epoch": 0.66, "learning_rate": 5.618445505482799e-05, "loss": 0.1411, "step": 10216 }, { "epoch": 0.66, "learning_rate": 5.616578879208296e-05, "loss": 0.1293, "step": 10217 }, { "epoch": 0.66, "learning_rate": 5.614712441972136e-05, "loss": 0.1503, "step": 10218 }, { "epoch": 0.66, "learning_rate": 5.612846193854817e-05, "loss": 0.1699, "step": 10219 }, { "epoch": 0.66, "learning_rate": 5.610980134936815e-05, "loss": 0.1343, "step": 10220 }, { "epoch": 0.66, "learning_rate": 5.609114265298612e-05, "loss": 0.1341, "step": 10221 }, { "epoch": 0.66, "learning_rate": 5.6072485850206645e-05, "loss": 0.1457, "step": 10222 }, { "epoch": 0.66, "learning_rate": 5.6053830941834464e-05, "loss": 0.1277, "step": 10223 }, { "epoch": 0.66, "learning_rate": 5.603517792867401e-05, "loss": 0.1442, "step": 10224 }, { "epoch": 0.66, "learning_rate": 5.601652681152965e-05, "loss": 0.1359, "step": 10225 }, { "epoch": 0.66, "learning_rate": 5.599787759120584e-05, "loss": 0.1267, "step": 10226 }, { "epoch": 0.66, "learning_rate": 5.597923026850673e-05, "loss": 0.1495, "step": 10227 }, { "epoch": 0.66, "learning_rate": 5.596058484423656e-05, "loss": 0.1535, "step": 10228 }, { "epoch": 0.66, "learning_rate": 5.594194131919943e-05, "loss": 0.1433, "step": 10229 }, { "epoch": 0.66, "learning_rate": 5.592329969419938e-05, "loss": 0.1473, "step": 10230 }, { "epoch": 0.66, "learning_rate": 5.5904659970040307e-05, "loss": 0.1395, "step": 10231 }, { "epoch": 0.66, "learning_rate": 5.588602214752604e-05, "loss": 0.1457, "step": 10232 }, { "epoch": 0.66, "learning_rate": 5.586738622746041e-05, "loss": 0.1494, "step": 10233 }, { "epoch": 0.66, "learning_rate": 5.584875221064704e-05, "loss": 0.1457, "step": 10234 }, { "epoch": 0.66, "learning_rate": 5.583012009788956e-05, "loss": 0.1314, "step": 10235 }, { "epoch": 0.66, "learning_rate": 5.5811489889991506e-05, "loss": 0.1553, "step": 10236 }, { "epoch": 0.66, "learning_rate": 5.5792861587756364e-05, "loss": 0.1493, "step": 10237 }, { "epoch": 0.66, "learning_rate": 5.577423519198743e-05, "loss": 0.1501, "step": 10238 }, { "epoch": 0.66, "learning_rate": 5.5755610703487974e-05, "loss": 0.1436, "step": 10239 }, { "epoch": 0.66, "learning_rate": 5.573698812306125e-05, "loss": 0.1313, "step": 10240 }, { "epoch": 0.66, "learning_rate": 5.571836745151029e-05, "loss": 0.155, "step": 10241 }, { "epoch": 0.66, "learning_rate": 5.569974868963816e-05, "loss": 0.1428, "step": 10242 }, { "epoch": 0.66, "learning_rate": 5.5681131838247836e-05, "loss": 0.1657, "step": 10243 }, { "epoch": 0.66, "learning_rate": 5.566251689814221e-05, "loss": 0.1351, "step": 10244 }, { "epoch": 0.66, "learning_rate": 5.564390387012401e-05, "loss": 0.1714, "step": 10245 }, { "epoch": 0.66, "learning_rate": 5.562529275499591e-05, "loss": 0.1356, "step": 10246 }, { "epoch": 0.66, "learning_rate": 5.56066835535606e-05, "loss": 0.1287, "step": 10247 }, { "epoch": 0.66, "learning_rate": 5.5588076266620546e-05, "loss": 0.1333, "step": 10248 }, { "epoch": 0.66, "learning_rate": 5.5569470894978246e-05, "loss": 0.1417, "step": 10249 }, { "epoch": 0.66, "learning_rate": 5.555086743943606e-05, "loss": 0.1573, "step": 10250 }, { "epoch": 0.66, "learning_rate": 5.553226590079632e-05, "loss": 0.1396, "step": 10251 }, { "epoch": 0.66, "learning_rate": 5.551366627986119e-05, "loss": 0.1239, "step": 10252 }, { "epoch": 0.66, "learning_rate": 5.5495068577432744e-05, "loss": 0.1304, "step": 10253 }, { "epoch": 0.66, "learning_rate": 5.5476472794313115e-05, "loss": 0.1442, "step": 10254 }, { "epoch": 0.66, "learning_rate": 5.545787893130418e-05, "loss": 0.1564, "step": 10255 }, { "epoch": 0.66, "learning_rate": 5.543928698920784e-05, "loss": 0.1481, "step": 10256 }, { "epoch": 0.66, "learning_rate": 5.54206969688259e-05, "loss": 0.1442, "step": 10257 }, { "epoch": 0.66, "learning_rate": 5.54021088709601e-05, "loss": 0.1338, "step": 10258 }, { "epoch": 0.66, "learning_rate": 5.538352269641204e-05, "loss": 0.1412, "step": 10259 }, { "epoch": 0.66, "learning_rate": 5.53649384459832e-05, "loss": 0.1493, "step": 10260 }, { "epoch": 0.66, "learning_rate": 5.534635612047514e-05, "loss": 0.1339, "step": 10261 }, { "epoch": 0.66, "learning_rate": 5.5327775720689154e-05, "loss": 0.1463, "step": 10262 }, { "epoch": 0.66, "learning_rate": 5.530919724742657e-05, "loss": 0.1221, "step": 10263 }, { "epoch": 0.66, "learning_rate": 5.5290620701488594e-05, "loss": 0.1201, "step": 10264 }, { "epoch": 0.66, "learning_rate": 5.52720460836764e-05, "loss": 0.1329, "step": 10265 }, { "epoch": 0.66, "learning_rate": 5.5253473394790986e-05, "loss": 0.1516, "step": 10266 }, { "epoch": 0.66, "learning_rate": 5.5234902635633304e-05, "loss": 0.1341, "step": 10267 }, { "epoch": 0.66, "learning_rate": 5.521633380700426e-05, "loss": 0.1349, "step": 10268 }, { "epoch": 0.66, "learning_rate": 5.519776690970457e-05, "loss": 0.1579, "step": 10269 }, { "epoch": 0.66, "learning_rate": 5.51792019445351e-05, "loss": 0.1142, "step": 10270 }, { "epoch": 0.66, "learning_rate": 5.5160638912296327e-05, "loss": 0.1559, "step": 10271 }, { "epoch": 0.66, "learning_rate": 5.514207781378891e-05, "loss": 0.1381, "step": 10272 }, { "epoch": 0.66, "learning_rate": 5.512351864981326e-05, "loss": 0.1473, "step": 10273 }, { "epoch": 0.66, "learning_rate": 5.5104961421169706e-05, "loss": 0.1472, "step": 10274 }, { "epoch": 0.66, "learning_rate": 5.508640612865861e-05, "loss": 0.124, "step": 10275 }, { "epoch": 0.66, "learning_rate": 5.5067852773080144e-05, "loss": 0.1553, "step": 10276 }, { "epoch": 0.66, "learning_rate": 5.5049301355234496e-05, "loss": 0.1437, "step": 10277 }, { "epoch": 0.66, "learning_rate": 5.503075187592167e-05, "loss": 0.1384, "step": 10278 }, { "epoch": 0.66, "learning_rate": 5.501220433594159e-05, "loss": 0.1409, "step": 10279 }, { "epoch": 0.66, "learning_rate": 5.499365873609421e-05, "loss": 0.1642, "step": 10280 }, { "epoch": 0.66, "learning_rate": 5.4975115077179226e-05, "loss": 0.1446, "step": 10281 }, { "epoch": 0.66, "learning_rate": 5.4956573359996424e-05, "loss": 0.1228, "step": 10282 }, { "epoch": 0.66, "learning_rate": 5.493803358534538e-05, "loss": 0.1475, "step": 10283 }, { "epoch": 0.66, "learning_rate": 5.491949575402573e-05, "loss": 0.1222, "step": 10284 }, { "epoch": 0.66, "learning_rate": 5.4900959866836845e-05, "loss": 0.145, "step": 10285 }, { "epoch": 0.66, "learning_rate": 5.4882425924578085e-05, "loss": 0.1559, "step": 10286 }, { "epoch": 0.66, "learning_rate": 5.4863893928048826e-05, "loss": 0.1295, "step": 10287 }, { "epoch": 0.66, "learning_rate": 5.484536387804817e-05, "loss": 0.1186, "step": 10288 }, { "epoch": 0.66, "learning_rate": 5.4826835775375285e-05, "loss": 0.1261, "step": 10289 }, { "epoch": 0.66, "learning_rate": 5.4808309620829234e-05, "loss": 0.1528, "step": 10290 }, { "epoch": 0.66, "learning_rate": 5.4789785415208985e-05, "loss": 0.1467, "step": 10291 }, { "epoch": 0.66, "learning_rate": 5.477126315931338e-05, "loss": 0.1409, "step": 10292 }, { "epoch": 0.66, "learning_rate": 5.475274285394116e-05, "loss": 0.1358, "step": 10293 }, { "epoch": 0.66, "learning_rate": 5.473422449989112e-05, "loss": 0.1464, "step": 10294 }, { "epoch": 0.66, "learning_rate": 5.471570809796178e-05, "loss": 0.1504, "step": 10295 }, { "epoch": 0.66, "learning_rate": 5.4697193648951714e-05, "loss": 0.1429, "step": 10296 }, { "epoch": 0.66, "learning_rate": 5.46786811536594e-05, "loss": 0.1423, "step": 10297 }, { "epoch": 0.66, "learning_rate": 5.46601706128832e-05, "loss": 0.1358, "step": 10298 }, { "epoch": 0.66, "learning_rate": 5.464166202742139e-05, "loss": 0.1343, "step": 10299 }, { "epoch": 0.66, "learning_rate": 5.462315539807211e-05, "loss": 0.1364, "step": 10300 }, { "epoch": 0.66, "learning_rate": 5.460465072563357e-05, "loss": 0.1607, "step": 10301 }, { "epoch": 0.66, "learning_rate": 5.458614801090369e-05, "loss": 0.1376, "step": 10302 }, { "epoch": 0.66, "learning_rate": 5.4567647254680464e-05, "loss": 0.1347, "step": 10303 }, { "epoch": 0.66, "learning_rate": 5.4549148457761765e-05, "loss": 0.1708, "step": 10304 }, { "epoch": 0.66, "learning_rate": 5.4530651620945395e-05, "loss": 0.1491, "step": 10305 }, { "epoch": 0.66, "learning_rate": 5.451215674502901e-05, "loss": 0.1581, "step": 10306 }, { "epoch": 0.66, "learning_rate": 5.4493663830810174e-05, "loss": 0.1704, "step": 10307 }, { "epoch": 0.66, "learning_rate": 5.4475172879086476e-05, "loss": 0.1222, "step": 10308 }, { "epoch": 0.66, "learning_rate": 5.4456683890655256e-05, "loss": 0.1375, "step": 10309 }, { "epoch": 0.66, "learning_rate": 5.4438196866314005e-05, "loss": 0.1292, "step": 10310 }, { "epoch": 0.66, "learning_rate": 5.4419711806859876e-05, "loss": 0.1297, "step": 10311 }, { "epoch": 0.66, "learning_rate": 5.440122871309014e-05, "loss": 0.1373, "step": 10312 }, { "epoch": 0.66, "learning_rate": 5.438274758580184e-05, "loss": 0.1273, "step": 10313 }, { "epoch": 0.66, "learning_rate": 5.436426842579195e-05, "loss": 0.1368, "step": 10314 }, { "epoch": 0.66, "learning_rate": 5.4345791233857455e-05, "loss": 0.1346, "step": 10315 }, { "epoch": 0.66, "learning_rate": 5.432731601079517e-05, "loss": 0.1301, "step": 10316 }, { "epoch": 0.66, "learning_rate": 5.4308842757401926e-05, "loss": 0.1483, "step": 10317 }, { "epoch": 0.66, "learning_rate": 5.4290371474474286e-05, "loss": 0.1513, "step": 10318 }, { "epoch": 0.66, "learning_rate": 5.427190216280893e-05, "loss": 0.1411, "step": 10319 }, { "epoch": 0.66, "learning_rate": 5.425343482320233e-05, "loss": 0.1496, "step": 10320 }, { "epoch": 0.66, "learning_rate": 5.4234969456450856e-05, "loss": 0.1605, "step": 10321 }, { "epoch": 0.66, "learning_rate": 5.421650606335088e-05, "loss": 0.1282, "step": 10322 }, { "epoch": 0.66, "learning_rate": 5.419804464469864e-05, "loss": 0.1352, "step": 10323 }, { "epoch": 0.66, "learning_rate": 5.4179585201290364e-05, "loss": 0.1462, "step": 10324 }, { "epoch": 0.66, "learning_rate": 5.416112773392204e-05, "loss": 0.1297, "step": 10325 }, { "epoch": 0.66, "learning_rate": 5.4142672243389716e-05, "loss": 0.1281, "step": 10326 }, { "epoch": 0.66, "learning_rate": 5.412421873048927e-05, "loss": 0.1484, "step": 10327 }, { "epoch": 0.66, "learning_rate": 5.410576719601652e-05, "loss": 0.1363, "step": 10328 }, { "epoch": 0.66, "learning_rate": 5.40873176407672e-05, "loss": 0.1427, "step": 10329 }, { "epoch": 0.66, "learning_rate": 5.406887006553697e-05, "loss": 0.1075, "step": 10330 }, { "epoch": 0.66, "learning_rate": 5.405042447112144e-05, "loss": 0.1432, "step": 10331 }, { "epoch": 0.66, "learning_rate": 5.403198085831602e-05, "loss": 0.139, "step": 10332 }, { "epoch": 0.66, "learning_rate": 5.401353922791616e-05, "loss": 0.1393, "step": 10333 }, { "epoch": 0.66, "learning_rate": 5.3995099580717154e-05, "loss": 0.1635, "step": 10334 }, { "epoch": 0.66, "learning_rate": 5.397666191751418e-05, "loss": 0.1353, "step": 10335 }, { "epoch": 0.66, "learning_rate": 5.3958226239102404e-05, "loss": 0.1426, "step": 10336 }, { "epoch": 0.66, "learning_rate": 5.3939792546276904e-05, "loss": 0.1554, "step": 10337 }, { "epoch": 0.66, "learning_rate": 5.3921360839832644e-05, "loss": 0.1518, "step": 10338 }, { "epoch": 0.66, "learning_rate": 5.390293112056446e-05, "loss": 0.1443, "step": 10339 }, { "epoch": 0.66, "learning_rate": 5.388450338926723e-05, "loss": 0.1375, "step": 10340 }, { "epoch": 0.66, "learning_rate": 5.38660776467356e-05, "loss": 0.1353, "step": 10341 }, { "epoch": 0.66, "learning_rate": 5.384765389376417e-05, "loss": 0.1452, "step": 10342 }, { "epoch": 0.66, "learning_rate": 5.3829232131147525e-05, "loss": 0.1387, "step": 10343 }, { "epoch": 0.66, "learning_rate": 5.381081235968008e-05, "loss": 0.1633, "step": 10344 }, { "epoch": 0.66, "learning_rate": 5.3792394580156305e-05, "loss": 0.1318, "step": 10345 }, { "epoch": 0.66, "learning_rate": 5.377397879337034e-05, "loss": 0.1382, "step": 10346 }, { "epoch": 0.66, "learning_rate": 5.375556500011649e-05, "loss": 0.1487, "step": 10347 }, { "epoch": 0.66, "learning_rate": 5.373715320118882e-05, "loss": 0.1491, "step": 10348 }, { "epoch": 0.66, "learning_rate": 5.37187433973813e-05, "loss": 0.1314, "step": 10349 }, { "epoch": 0.66, "learning_rate": 5.3700335589487925e-05, "loss": 0.1406, "step": 10350 }, { "epoch": 0.66, "learning_rate": 5.3681929778302544e-05, "loss": 0.1426, "step": 10351 }, { "epoch": 0.66, "learning_rate": 5.366352596461894e-05, "loss": 0.1418, "step": 10352 }, { "epoch": 0.66, "learning_rate": 5.3645124149230744e-05, "loss": 0.1239, "step": 10353 }, { "epoch": 0.66, "learning_rate": 5.36267243329316e-05, "loss": 0.155, "step": 10354 }, { "epoch": 0.66, "learning_rate": 5.360832651651495e-05, "loss": 0.1392, "step": 10355 }, { "epoch": 0.66, "learning_rate": 5.3589930700774246e-05, "loss": 0.1417, "step": 10356 }, { "epoch": 0.66, "learning_rate": 5.357153688650287e-05, "loss": 0.1315, "step": 10357 }, { "epoch": 0.66, "learning_rate": 5.355314507449397e-05, "loss": 0.1568, "step": 10358 }, { "epoch": 0.66, "learning_rate": 5.3534755265540815e-05, "loss": 0.1253, "step": 10359 }, { "epoch": 0.66, "learning_rate": 5.3516367460436415e-05, "loss": 0.1304, "step": 10360 }, { "epoch": 0.66, "learning_rate": 5.349798165997373e-05, "loss": 0.1308, "step": 10361 }, { "epoch": 0.66, "learning_rate": 5.3479597864945696e-05, "loss": 0.1583, "step": 10362 }, { "epoch": 0.66, "learning_rate": 5.3461216076145124e-05, "loss": 0.1417, "step": 10363 }, { "epoch": 0.66, "learning_rate": 5.3442836294364794e-05, "loss": 0.1421, "step": 10364 }, { "epoch": 0.66, "learning_rate": 5.342445852039728e-05, "loss": 0.1551, "step": 10365 }, { "epoch": 0.66, "learning_rate": 5.340608275503518e-05, "loss": 0.1245, "step": 10366 }, { "epoch": 0.66, "learning_rate": 5.338770899907095e-05, "loss": 0.1509, "step": 10367 }, { "epoch": 0.66, "learning_rate": 5.3369337253296934e-05, "loss": 0.1354, "step": 10368 }, { "epoch": 0.66, "learning_rate": 5.335096751850546e-05, "loss": 0.1383, "step": 10369 }, { "epoch": 0.66, "learning_rate": 5.3332599795488725e-05, "loss": 0.1333, "step": 10370 }, { "epoch": 0.66, "learning_rate": 5.3314234085038894e-05, "loss": 0.128, "step": 10371 }, { "epoch": 0.67, "learning_rate": 5.3295870387947944e-05, "loss": 0.1347, "step": 10372 }, { "epoch": 0.67, "learning_rate": 5.3277508705007885e-05, "loss": 0.1327, "step": 10373 }, { "epoch": 0.67, "learning_rate": 5.325914903701055e-05, "loss": 0.1271, "step": 10374 }, { "epoch": 0.67, "learning_rate": 5.324079138474765e-05, "loss": 0.1616, "step": 10375 }, { "epoch": 0.67, "learning_rate": 5.3222435749010937e-05, "loss": 0.1387, "step": 10376 }, { "epoch": 0.67, "learning_rate": 5.3204082130592e-05, "loss": 0.1498, "step": 10377 }, { "epoch": 0.67, "learning_rate": 5.31857305302824e-05, "loss": 0.141, "step": 10378 }, { "epoch": 0.67, "learning_rate": 5.316738094887348e-05, "loss": 0.1375, "step": 10379 }, { "epoch": 0.67, "learning_rate": 5.314903338715667e-05, "loss": 0.1548, "step": 10380 }, { "epoch": 0.67, "learning_rate": 5.3130687845923145e-05, "loss": 0.1437, "step": 10381 }, { "epoch": 0.67, "learning_rate": 5.311234432596408e-05, "loss": 0.147, "step": 10382 }, { "epoch": 0.67, "learning_rate": 5.309400282807055e-05, "loss": 0.1516, "step": 10383 }, { "epoch": 0.67, "learning_rate": 5.307566335303358e-05, "loss": 0.1582, "step": 10384 }, { "epoch": 0.67, "learning_rate": 5.3057325901644095e-05, "loss": 0.1561, "step": 10385 }, { "epoch": 0.67, "learning_rate": 5.303899047469285e-05, "loss": 0.1427, "step": 10386 }, { "epoch": 0.67, "learning_rate": 5.3020657072970616e-05, "loss": 0.1423, "step": 10387 }, { "epoch": 0.67, "learning_rate": 5.300232569726804e-05, "loss": 0.1536, "step": 10388 }, { "epoch": 0.67, "learning_rate": 5.298399634837561e-05, "loss": 0.1398, "step": 10389 }, { "epoch": 0.67, "learning_rate": 5.296566902708383e-05, "loss": 0.1483, "step": 10390 }, { "epoch": 0.67, "learning_rate": 5.2947343734183085e-05, "loss": 0.1383, "step": 10391 }, { "epoch": 0.67, "learning_rate": 5.292902047046373e-05, "loss": 0.1375, "step": 10392 }, { "epoch": 0.67, "learning_rate": 5.291069923671586e-05, "loss": 0.1234, "step": 10393 }, { "epoch": 0.67, "learning_rate": 5.289238003372967e-05, "loss": 0.1418, "step": 10394 }, { "epoch": 0.67, "learning_rate": 5.287406286229517e-05, "loss": 0.1569, "step": 10395 }, { "epoch": 0.67, "learning_rate": 5.285574772320222e-05, "loss": 0.1489, "step": 10396 }, { "epoch": 0.67, "learning_rate": 5.283743461724082e-05, "loss": 0.1577, "step": 10397 }, { "epoch": 0.67, "learning_rate": 5.2819123545200624e-05, "loss": 0.131, "step": 10398 }, { "epoch": 0.67, "learning_rate": 5.2800814507871397e-05, "loss": 0.1455, "step": 10399 }, { "epoch": 0.67, "learning_rate": 5.2782507506042634e-05, "loss": 0.1294, "step": 10400 }, { "epoch": 0.67, "learning_rate": 5.276420254050394e-05, "loss": 0.1346, "step": 10401 }, { "epoch": 0.67, "learning_rate": 5.2745899612044625e-05, "loss": 0.1388, "step": 10402 }, { "epoch": 0.67, "learning_rate": 5.272759872145407e-05, "loss": 0.1364, "step": 10403 }, { "epoch": 0.67, "learning_rate": 5.270929986952157e-05, "loss": 0.1372, "step": 10404 }, { "epoch": 0.67, "learning_rate": 5.269100305703617e-05, "loss": 0.1367, "step": 10405 }, { "epoch": 0.67, "learning_rate": 5.267270828478703e-05, "loss": 0.1586, "step": 10406 }, { "epoch": 0.67, "learning_rate": 5.265441555356303e-05, "loss": 0.145, "step": 10407 }, { "epoch": 0.67, "learning_rate": 5.263612486415317e-05, "loss": 0.1402, "step": 10408 }, { "epoch": 0.67, "learning_rate": 5.261783621734613e-05, "loss": 0.1566, "step": 10409 }, { "epoch": 0.67, "learning_rate": 5.259954961393069e-05, "loss": 0.1472, "step": 10410 }, { "epoch": 0.67, "learning_rate": 5.2581265054695494e-05, "loss": 0.1548, "step": 10411 }, { "epoch": 0.67, "learning_rate": 5.256298254042902e-05, "loss": 0.1267, "step": 10412 }, { "epoch": 0.67, "learning_rate": 5.254470207191978e-05, "loss": 0.1673, "step": 10413 }, { "epoch": 0.67, "learning_rate": 5.252642364995606e-05, "loss": 0.1318, "step": 10414 }, { "epoch": 0.67, "learning_rate": 5.2508147275326204e-05, "loss": 0.1332, "step": 10415 }, { "epoch": 0.67, "learning_rate": 5.2489872948818305e-05, "loss": 0.1445, "step": 10416 }, { "epoch": 0.67, "learning_rate": 5.247160067122052e-05, "loss": 0.126, "step": 10417 }, { "epoch": 0.67, "learning_rate": 5.245333044332088e-05, "loss": 0.1411, "step": 10418 }, { "epoch": 0.67, "learning_rate": 5.243506226590722e-05, "loss": 0.146, "step": 10419 }, { "epoch": 0.67, "learning_rate": 5.241679613976747e-05, "loss": 0.1379, "step": 10420 }, { "epoch": 0.67, "learning_rate": 5.239853206568928e-05, "loss": 0.1341, "step": 10421 }, { "epoch": 0.67, "learning_rate": 5.2380270044460355e-05, "loss": 0.1586, "step": 10422 }, { "epoch": 0.67, "learning_rate": 5.236201007686822e-05, "loss": 0.144, "step": 10423 }, { "epoch": 0.67, "learning_rate": 5.234375216370035e-05, "loss": 0.1538, "step": 10424 }, { "epoch": 0.67, "learning_rate": 5.2325496305744214e-05, "loss": 0.1305, "step": 10425 }, { "epoch": 0.67, "learning_rate": 5.2307242503787e-05, "loss": 0.1381, "step": 10426 }, { "epoch": 0.67, "learning_rate": 5.228899075861601e-05, "loss": 0.1256, "step": 10427 }, { "epoch": 0.67, "learning_rate": 5.227074107101828e-05, "loss": 0.1477, "step": 10428 }, { "epoch": 0.67, "learning_rate": 5.225249344178091e-05, "loss": 0.1447, "step": 10429 }, { "epoch": 0.67, "learning_rate": 5.223424787169078e-05, "loss": 0.1381, "step": 10430 }, { "epoch": 0.67, "learning_rate": 5.221600436153478e-05, "loss": 0.1576, "step": 10431 }, { "epoch": 0.67, "learning_rate": 5.21977629120997e-05, "loss": 0.1573, "step": 10432 }, { "epoch": 0.67, "learning_rate": 5.217952352417217e-05, "loss": 0.1264, "step": 10433 }, { "epoch": 0.67, "learning_rate": 5.216128619853882e-05, "loss": 0.1437, "step": 10434 }, { "epoch": 0.67, "learning_rate": 5.214305093598608e-05, "loss": 0.1343, "step": 10435 }, { "epoch": 0.67, "learning_rate": 5.212481773730047e-05, "loss": 0.1393, "step": 10436 }, { "epoch": 0.67, "learning_rate": 5.210658660326818e-05, "loss": 0.1482, "step": 10437 }, { "epoch": 0.67, "learning_rate": 5.208835753467551e-05, "loss": 0.1437, "step": 10438 }, { "epoch": 0.67, "learning_rate": 5.207013053230865e-05, "loss": 0.1366, "step": 10439 }, { "epoch": 0.67, "learning_rate": 5.205190559695357e-05, "loss": 0.1322, "step": 10440 }, { "epoch": 0.67, "learning_rate": 5.203368272939628e-05, "loss": 0.127, "step": 10441 }, { "epoch": 0.67, "learning_rate": 5.201546193042261e-05, "loss": 0.1483, "step": 10442 }, { "epoch": 0.67, "learning_rate": 5.199724320081838e-05, "loss": 0.1256, "step": 10443 }, { "epoch": 0.67, "learning_rate": 5.197902654136933e-05, "loss": 0.1381, "step": 10444 }, { "epoch": 0.67, "learning_rate": 5.1960811952860975e-05, "loss": 0.134, "step": 10445 }, { "epoch": 0.67, "learning_rate": 5.194259943607891e-05, "loss": 0.1384, "step": 10446 }, { "epoch": 0.67, "learning_rate": 5.192438899180849e-05, "loss": 0.1613, "step": 10447 }, { "epoch": 0.67, "learning_rate": 5.190618062083516e-05, "loss": 0.1483, "step": 10448 }, { "epoch": 0.67, "learning_rate": 5.188797432394404e-05, "loss": 0.1559, "step": 10449 }, { "epoch": 0.67, "learning_rate": 5.1869770101920365e-05, "loss": 0.1305, "step": 10450 }, { "epoch": 0.67, "learning_rate": 5.185156795554924e-05, "loss": 0.1771, "step": 10451 }, { "epoch": 0.67, "learning_rate": 5.1833367885615556e-05, "loss": 0.1414, "step": 10452 }, { "epoch": 0.67, "learning_rate": 5.1815169892904294e-05, "loss": 0.1424, "step": 10453 }, { "epoch": 0.67, "learning_rate": 5.179697397820018e-05, "loss": 0.1352, "step": 10454 }, { "epoch": 0.67, "learning_rate": 5.177878014228799e-05, "loss": 0.1368, "step": 10455 }, { "epoch": 0.67, "learning_rate": 5.176058838595228e-05, "loss": 0.1276, "step": 10456 }, { "epoch": 0.67, "learning_rate": 5.174239870997764e-05, "loss": 0.1478, "step": 10457 }, { "epoch": 0.67, "learning_rate": 5.172421111514851e-05, "loss": 0.1334, "step": 10458 }, { "epoch": 0.67, "learning_rate": 5.17060256022492e-05, "loss": 0.1452, "step": 10459 }, { "epoch": 0.67, "learning_rate": 5.1687842172064035e-05, "loss": 0.1228, "step": 10460 }, { "epoch": 0.67, "learning_rate": 5.166966082537712e-05, "loss": 0.1359, "step": 10461 }, { "epoch": 0.67, "learning_rate": 5.1651481562972615e-05, "loss": 0.1347, "step": 10462 }, { "epoch": 0.67, "learning_rate": 5.1633304385634426e-05, "loss": 0.1638, "step": 10463 }, { "epoch": 0.67, "learning_rate": 5.1615129294146515e-05, "loss": 0.1385, "step": 10464 }, { "epoch": 0.67, "learning_rate": 5.159695628929273e-05, "loss": 0.1397, "step": 10465 }, { "epoch": 0.67, "learning_rate": 5.157878537185671e-05, "loss": 0.142, "step": 10466 }, { "epoch": 0.67, "learning_rate": 5.156061654262217e-05, "loss": 0.1394, "step": 10467 }, { "epoch": 0.67, "learning_rate": 5.154244980237258e-05, "loss": 0.1216, "step": 10468 }, { "epoch": 0.67, "learning_rate": 5.152428515189147e-05, "loss": 0.1411, "step": 10469 }, { "epoch": 0.67, "learning_rate": 5.1506122591962126e-05, "loss": 0.1661, "step": 10470 }, { "epoch": 0.67, "learning_rate": 5.148796212336787e-05, "loss": 0.1615, "step": 10471 }, { "epoch": 0.67, "learning_rate": 5.146980374689192e-05, "loss": 0.1493, "step": 10472 }, { "epoch": 0.67, "learning_rate": 5.1451647463317274e-05, "loss": 0.1517, "step": 10473 }, { "epoch": 0.67, "learning_rate": 5.1433493273427035e-05, "loss": 0.126, "step": 10474 }, { "epoch": 0.67, "learning_rate": 5.141534117800403e-05, "loss": 0.1318, "step": 10475 }, { "epoch": 0.67, "learning_rate": 5.1397191177831164e-05, "loss": 0.1357, "step": 10476 }, { "epoch": 0.67, "learning_rate": 5.1379043273691085e-05, "loss": 0.1387, "step": 10477 }, { "epoch": 0.67, "learning_rate": 5.136089746636648e-05, "loss": 0.1327, "step": 10478 }, { "epoch": 0.67, "learning_rate": 5.134275375663994e-05, "loss": 0.1314, "step": 10479 }, { "epoch": 0.67, "learning_rate": 5.132461214529384e-05, "loss": 0.1228, "step": 10480 }, { "epoch": 0.67, "learning_rate": 5.1306472633110634e-05, "loss": 0.1522, "step": 10481 }, { "epoch": 0.67, "learning_rate": 5.128833522087253e-05, "loss": 0.1346, "step": 10482 }, { "epoch": 0.67, "learning_rate": 5.127019990936175e-05, "loss": 0.135, "step": 10483 }, { "epoch": 0.67, "learning_rate": 5.125206669936042e-05, "loss": 0.1231, "step": 10484 }, { "epoch": 0.67, "learning_rate": 5.123393559165051e-05, "loss": 0.1422, "step": 10485 }, { "epoch": 0.67, "learning_rate": 5.121580658701397e-05, "loss": 0.1401, "step": 10486 }, { "epoch": 0.67, "learning_rate": 5.119767968623257e-05, "loss": 0.1186, "step": 10487 }, { "epoch": 0.67, "learning_rate": 5.117955489008812e-05, "loss": 0.1323, "step": 10488 }, { "epoch": 0.67, "learning_rate": 5.11614321993622e-05, "loss": 0.1437, "step": 10489 }, { "epoch": 0.67, "learning_rate": 5.114331161483639e-05, "loss": 0.1505, "step": 10490 }, { "epoch": 0.67, "learning_rate": 5.112519313729218e-05, "loss": 0.1349, "step": 10491 }, { "epoch": 0.67, "learning_rate": 5.11070767675109e-05, "loss": 0.1322, "step": 10492 }, { "epoch": 0.67, "learning_rate": 5.1088962506273884e-05, "loss": 0.1541, "step": 10493 }, { "epoch": 0.67, "learning_rate": 5.107085035436225e-05, "loss": 0.1422, "step": 10494 }, { "epoch": 0.67, "learning_rate": 5.10527403125572e-05, "loss": 0.1273, "step": 10495 }, { "epoch": 0.67, "learning_rate": 5.103463238163963e-05, "loss": 0.1427, "step": 10496 }, { "epoch": 0.67, "learning_rate": 5.1016526562390507e-05, "loss": 0.1288, "step": 10497 }, { "epoch": 0.67, "learning_rate": 5.09984228555907e-05, "loss": 0.1317, "step": 10498 }, { "epoch": 0.67, "learning_rate": 5.098032126202088e-05, "loss": 0.1181, "step": 10499 }, { "epoch": 0.67, "learning_rate": 5.096222178246175e-05, "loss": 0.116, "step": 10500 }, { "epoch": 0.67, "learning_rate": 5.09441244176938e-05, "loss": 0.1446, "step": 10501 }, { "epoch": 0.67, "learning_rate": 5.092602916849756e-05, "loss": 0.1598, "step": 10502 }, { "epoch": 0.67, "learning_rate": 5.0907936035653324e-05, "loss": 0.1415, "step": 10503 }, { "epoch": 0.67, "learning_rate": 5.088984501994142e-05, "loss": 0.1526, "step": 10504 }, { "epoch": 0.67, "learning_rate": 5.087175612214208e-05, "loss": 0.1479, "step": 10505 }, { "epoch": 0.67, "learning_rate": 5.0853669343035305e-05, "loss": 0.1264, "step": 10506 }, { "epoch": 0.67, "learning_rate": 5.083558468340117e-05, "loss": 0.1283, "step": 10507 }, { "epoch": 0.67, "learning_rate": 5.081750214401956e-05, "loss": 0.1509, "step": 10508 }, { "epoch": 0.67, "learning_rate": 5.0799421725670335e-05, "loss": 0.1514, "step": 10509 }, { "epoch": 0.67, "learning_rate": 5.0781343429133155e-05, "loss": 0.1392, "step": 10510 }, { "epoch": 0.67, "learning_rate": 5.076326725518771e-05, "loss": 0.1452, "step": 10511 }, { "epoch": 0.67, "learning_rate": 5.074519320461357e-05, "loss": 0.14, "step": 10512 }, { "epoch": 0.67, "learning_rate": 5.072712127819015e-05, "loss": 0.1641, "step": 10513 }, { "epoch": 0.67, "learning_rate": 5.0709051476696846e-05, "loss": 0.1578, "step": 10514 }, { "epoch": 0.67, "learning_rate": 5.0690983800912884e-05, "loss": 0.119, "step": 10515 }, { "epoch": 0.67, "learning_rate": 5.067291825161751e-05, "loss": 0.1505, "step": 10516 }, { "epoch": 0.67, "learning_rate": 5.0654854829589757e-05, "loss": 0.1474, "step": 10517 }, { "epoch": 0.67, "learning_rate": 5.0636793535608637e-05, "loss": 0.1532, "step": 10518 }, { "epoch": 0.67, "learning_rate": 5.061873437045312e-05, "loss": 0.1503, "step": 10519 }, { "epoch": 0.67, "learning_rate": 5.060067733490192e-05, "loss": 0.141, "step": 10520 }, { "epoch": 0.67, "learning_rate": 5.0582622429733864e-05, "loss": 0.1419, "step": 10521 }, { "epoch": 0.67, "learning_rate": 5.0564569655727475e-05, "loss": 0.1265, "step": 10522 }, { "epoch": 0.67, "learning_rate": 5.054651901366139e-05, "loss": 0.1628, "step": 10523 }, { "epoch": 0.67, "learning_rate": 5.052847050431397e-05, "loss": 0.1683, "step": 10524 }, { "epoch": 0.67, "learning_rate": 5.051042412846363e-05, "loss": 0.1425, "step": 10525 }, { "epoch": 0.67, "learning_rate": 5.049237988688865e-05, "loss": 0.1264, "step": 10526 }, { "epoch": 0.67, "learning_rate": 5.0474337780367145e-05, "loss": 0.1429, "step": 10527 }, { "epoch": 0.68, "learning_rate": 5.0456297809677234e-05, "loss": 0.1338, "step": 10528 }, { "epoch": 0.68, "learning_rate": 5.043825997559686e-05, "loss": 0.1428, "step": 10529 }, { "epoch": 0.68, "learning_rate": 5.042022427890395e-05, "loss": 0.1543, "step": 10530 }, { "epoch": 0.68, "learning_rate": 5.0402190720376344e-05, "loss": 0.1314, "step": 10531 }, { "epoch": 0.68, "learning_rate": 5.038415930079168e-05, "loss": 0.1491, "step": 10532 }, { "epoch": 0.68, "learning_rate": 5.0366130020927624e-05, "loss": 0.132, "step": 10533 }, { "epoch": 0.68, "learning_rate": 5.0348102881561675e-05, "loss": 0.1251, "step": 10534 }, { "epoch": 0.68, "learning_rate": 5.033007788347131e-05, "loss": 0.1279, "step": 10535 }, { "epoch": 0.68, "learning_rate": 5.031205502743381e-05, "loss": 0.1448, "step": 10536 }, { "epoch": 0.68, "learning_rate": 5.0294034314226445e-05, "loss": 0.1332, "step": 10537 }, { "epoch": 0.68, "learning_rate": 5.0276015744626426e-05, "loss": 0.135, "step": 10538 }, { "epoch": 0.68, "learning_rate": 5.0257999319410733e-05, "loss": 0.1391, "step": 10539 }, { "epoch": 0.68, "learning_rate": 5.023998503935643e-05, "loss": 0.1506, "step": 10540 }, { "epoch": 0.68, "learning_rate": 5.022197290524029e-05, "loss": 0.1348, "step": 10541 }, { "epoch": 0.68, "learning_rate": 5.02039629178392e-05, "loss": 0.137, "step": 10542 }, { "epoch": 0.68, "learning_rate": 5.018595507792977e-05, "loss": 0.1296, "step": 10543 }, { "epoch": 0.68, "learning_rate": 5.0167949386288636e-05, "loss": 0.1475, "step": 10544 }, { "epoch": 0.68, "learning_rate": 5.014994584369236e-05, "loss": 0.1526, "step": 10545 }, { "epoch": 0.68, "learning_rate": 5.0131944450917265e-05, "loss": 0.1333, "step": 10546 }, { "epoch": 0.68, "learning_rate": 5.011394520873977e-05, "loss": 0.1313, "step": 10547 }, { "epoch": 0.68, "learning_rate": 5.009594811793601e-05, "loss": 0.1443, "step": 10548 }, { "epoch": 0.68, "learning_rate": 5.00779531792822e-05, "loss": 0.1285, "step": 10549 }, { "epoch": 0.68, "learning_rate": 5.005996039355432e-05, "loss": 0.1508, "step": 10550 }, { "epoch": 0.68, "learning_rate": 5.004196976152836e-05, "loss": 0.1235, "step": 10551 }, { "epoch": 0.68, "learning_rate": 5.002398128398021e-05, "loss": 0.1463, "step": 10552 }, { "epoch": 0.68, "learning_rate": 5.000599496168558e-05, "loss": 0.1561, "step": 10553 }, { "epoch": 0.68, "learning_rate": 4.998801079542019e-05, "loss": 0.1504, "step": 10554 }, { "epoch": 0.68, "learning_rate": 4.997002878595957e-05, "loss": 0.1376, "step": 10555 }, { "epoch": 0.68, "learning_rate": 4.995204893407927e-05, "loss": 0.1271, "step": 10556 }, { "epoch": 0.68, "learning_rate": 4.99340712405546e-05, "loss": 0.1444, "step": 10557 }, { "epoch": 0.68, "learning_rate": 4.991609570616093e-05, "loss": 0.1185, "step": 10558 }, { "epoch": 0.68, "learning_rate": 4.9898122331673466e-05, "loss": 0.1526, "step": 10559 }, { "epoch": 0.68, "learning_rate": 4.9880151117867294e-05, "loss": 0.13, "step": 10560 }, { "epoch": 0.68, "learning_rate": 4.986218206551747e-05, "loss": 0.1468, "step": 10561 }, { "epoch": 0.68, "learning_rate": 4.984421517539887e-05, "loss": 0.1269, "step": 10562 }, { "epoch": 0.68, "learning_rate": 4.9826250448286405e-05, "loss": 0.1489, "step": 10563 }, { "epoch": 0.68, "learning_rate": 4.980828788495473e-05, "loss": 0.1551, "step": 10564 }, { "epoch": 0.68, "learning_rate": 4.9790327486178545e-05, "loss": 0.1365, "step": 10565 }, { "epoch": 0.68, "learning_rate": 4.977236925273243e-05, "loss": 0.1578, "step": 10566 }, { "epoch": 0.68, "learning_rate": 4.9754413185390794e-05, "loss": 0.146, "step": 10567 }, { "epoch": 0.68, "learning_rate": 4.973645928492805e-05, "loss": 0.1533, "step": 10568 }, { "epoch": 0.68, "learning_rate": 4.9718507552118424e-05, "loss": 0.1484, "step": 10569 }, { "epoch": 0.68, "learning_rate": 4.9700557987736117e-05, "loss": 0.1397, "step": 10570 }, { "epoch": 0.68, "learning_rate": 4.968261059255523e-05, "loss": 0.1193, "step": 10571 }, { "epoch": 0.68, "learning_rate": 4.9664665367349804e-05, "loss": 0.1553, "step": 10572 }, { "epoch": 0.68, "learning_rate": 4.964672231289368e-05, "loss": 0.1472, "step": 10573 }, { "epoch": 0.68, "learning_rate": 4.962878142996065e-05, "loss": 0.1293, "step": 10574 }, { "epoch": 0.68, "learning_rate": 4.961084271932448e-05, "loss": 0.1225, "step": 10575 }, { "epoch": 0.68, "learning_rate": 4.9592906181758736e-05, "loss": 0.1671, "step": 10576 }, { "epoch": 0.68, "learning_rate": 4.9574971818036976e-05, "loss": 0.1309, "step": 10577 }, { "epoch": 0.68, "learning_rate": 4.955703962893262e-05, "loss": 0.1393, "step": 10578 }, { "epoch": 0.68, "learning_rate": 4.953910961521907e-05, "loss": 0.1555, "step": 10579 }, { "epoch": 0.68, "learning_rate": 4.9521181777669525e-05, "loss": 0.144, "step": 10580 }, { "epoch": 0.68, "learning_rate": 4.950325611705707e-05, "loss": 0.1437, "step": 10581 }, { "epoch": 0.68, "learning_rate": 4.948533263415488e-05, "loss": 0.1341, "step": 10582 }, { "epoch": 0.68, "learning_rate": 4.946741132973582e-05, "loss": 0.1403, "step": 10583 }, { "epoch": 0.68, "learning_rate": 4.94494922045728e-05, "loss": 0.156, "step": 10584 }, { "epoch": 0.68, "learning_rate": 4.943157525943861e-05, "loss": 0.1463, "step": 10585 }, { "epoch": 0.68, "learning_rate": 4.941366049510595e-05, "loss": 0.133, "step": 10586 }, { "epoch": 0.68, "learning_rate": 4.939574791234738e-05, "loss": 0.1384, "step": 10587 }, { "epoch": 0.68, "learning_rate": 4.9377837511935344e-05, "loss": 0.1303, "step": 10588 }, { "epoch": 0.68, "learning_rate": 4.9359929294642336e-05, "loss": 0.1277, "step": 10589 }, { "epoch": 0.68, "learning_rate": 4.934202326124058e-05, "loss": 0.1363, "step": 10590 }, { "epoch": 0.68, "learning_rate": 4.932411941250231e-05, "loss": 0.1452, "step": 10591 }, { "epoch": 0.68, "learning_rate": 4.9306217749199665e-05, "loss": 0.1393, "step": 10592 }, { "epoch": 0.68, "learning_rate": 4.9288318272104696e-05, "loss": 0.1304, "step": 10593 }, { "epoch": 0.68, "learning_rate": 4.9270420981989294e-05, "loss": 0.1397, "step": 10594 }, { "epoch": 0.68, "learning_rate": 4.925252587962524e-05, "loss": 0.1434, "step": 10595 }, { "epoch": 0.68, "learning_rate": 4.923463296578438e-05, "loss": 0.1339, "step": 10596 }, { "epoch": 0.68, "learning_rate": 4.9216742241238264e-05, "loss": 0.1466, "step": 10597 }, { "epoch": 0.68, "learning_rate": 4.919885370675848e-05, "loss": 0.1428, "step": 10598 }, { "epoch": 0.68, "learning_rate": 4.918096736311654e-05, "loss": 0.1445, "step": 10599 }, { "epoch": 0.68, "learning_rate": 4.916308321108369e-05, "loss": 0.1221, "step": 10600 }, { "epoch": 0.68, "learning_rate": 4.914520125143133e-05, "loss": 0.1343, "step": 10601 }, { "epoch": 0.68, "learning_rate": 4.912732148493051e-05, "loss": 0.1482, "step": 10602 }, { "epoch": 0.68, "learning_rate": 4.910944391235241e-05, "loss": 0.132, "step": 10603 }, { "epoch": 0.68, "learning_rate": 4.9091568534467925e-05, "loss": 0.1303, "step": 10604 }, { "epoch": 0.68, "learning_rate": 4.907369535204799e-05, "loss": 0.1544, "step": 10605 }, { "epoch": 0.68, "learning_rate": 4.905582436586343e-05, "loss": 0.1349, "step": 10606 }, { "epoch": 0.68, "learning_rate": 4.903795557668488e-05, "loss": 0.1287, "step": 10607 }, { "epoch": 0.68, "learning_rate": 4.902008898528301e-05, "loss": 0.1334, "step": 10608 }, { "epoch": 0.68, "learning_rate": 4.9002224592428256e-05, "loss": 0.132, "step": 10609 }, { "epoch": 0.68, "learning_rate": 4.898436239889111e-05, "loss": 0.1446, "step": 10610 }, { "epoch": 0.68, "learning_rate": 4.896650240544179e-05, "loss": 0.136, "step": 10611 }, { "epoch": 0.68, "learning_rate": 4.894864461285066e-05, "loss": 0.1358, "step": 10612 }, { "epoch": 0.68, "learning_rate": 4.893078902188778e-05, "loss": 0.1375, "step": 10613 }, { "epoch": 0.68, "learning_rate": 4.891293563332314e-05, "loss": 0.141, "step": 10614 }, { "epoch": 0.68, "learning_rate": 4.889508444792677e-05, "loss": 0.1295, "step": 10615 }, { "epoch": 0.68, "learning_rate": 4.887723546646844e-05, "loss": 0.1385, "step": 10616 }, { "epoch": 0.68, "learning_rate": 4.885938868971793e-05, "loss": 0.1265, "step": 10617 }, { "epoch": 0.68, "learning_rate": 4.884154411844489e-05, "loss": 0.1333, "step": 10618 }, { "epoch": 0.68, "learning_rate": 4.8823701753418935e-05, "loss": 0.1376, "step": 10619 }, { "epoch": 0.68, "learning_rate": 4.880586159540948e-05, "loss": 0.1288, "step": 10620 }, { "epoch": 0.68, "learning_rate": 4.878802364518587e-05, "loss": 0.1322, "step": 10621 }, { "epoch": 0.68, "learning_rate": 4.877018790351743e-05, "loss": 0.1578, "step": 10622 }, { "epoch": 0.68, "learning_rate": 4.875235437117329e-05, "loss": 0.1358, "step": 10623 }, { "epoch": 0.68, "learning_rate": 4.8734523048922565e-05, "loss": 0.1344, "step": 10624 }, { "epoch": 0.68, "learning_rate": 4.871669393753424e-05, "loss": 0.1266, "step": 10625 }, { "epoch": 0.68, "learning_rate": 4.869886703777726e-05, "loss": 0.1628, "step": 10626 }, { "epoch": 0.68, "learning_rate": 4.868104235042037e-05, "loss": 0.1196, "step": 10627 }, { "epoch": 0.68, "learning_rate": 4.866321987623224e-05, "loss": 0.1357, "step": 10628 }, { "epoch": 0.68, "learning_rate": 4.864539961598155e-05, "loss": 0.131, "step": 10629 }, { "epoch": 0.68, "learning_rate": 4.862758157043674e-05, "loss": 0.1328, "step": 10630 }, { "epoch": 0.68, "learning_rate": 4.860976574036627e-05, "loss": 0.1533, "step": 10631 }, { "epoch": 0.68, "learning_rate": 4.859195212653846e-05, "loss": 0.1536, "step": 10632 }, { "epoch": 0.68, "learning_rate": 4.857414072972156e-05, "loss": 0.1759, "step": 10633 }, { "epoch": 0.68, "learning_rate": 4.8556331550683684e-05, "loss": 0.139, "step": 10634 }, { "epoch": 0.68, "learning_rate": 4.8538524590192814e-05, "loss": 0.1523, "step": 10635 }, { "epoch": 0.68, "learning_rate": 4.852071984901696e-05, "loss": 0.1474, "step": 10636 }, { "epoch": 0.68, "learning_rate": 4.8502917327923906e-05, "loss": 0.1288, "step": 10637 }, { "epoch": 0.68, "learning_rate": 4.848511702768142e-05, "loss": 0.1428, "step": 10638 }, { "epoch": 0.68, "learning_rate": 4.846731894905715e-05, "loss": 0.1281, "step": 10639 }, { "epoch": 0.68, "learning_rate": 4.8449523092818704e-05, "loss": 0.1449, "step": 10640 }, { "epoch": 0.68, "learning_rate": 4.843172945973351e-05, "loss": 0.1353, "step": 10641 }, { "epoch": 0.68, "learning_rate": 4.8413938050568875e-05, "loss": 0.1377, "step": 10642 }, { "epoch": 0.68, "learning_rate": 4.8396148866092165e-05, "loss": 0.136, "step": 10643 }, { "epoch": 0.68, "learning_rate": 4.8378361907070443e-05, "loss": 0.1405, "step": 10644 }, { "epoch": 0.68, "learning_rate": 4.836057717427085e-05, "loss": 0.1375, "step": 10645 }, { "epoch": 0.68, "learning_rate": 4.834279466846037e-05, "loss": 0.1565, "step": 10646 }, { "epoch": 0.68, "learning_rate": 4.8325014390405895e-05, "loss": 0.1539, "step": 10647 }, { "epoch": 0.68, "learning_rate": 4.830723634087421e-05, "loss": 0.1686, "step": 10648 }, { "epoch": 0.68, "learning_rate": 4.8289460520631935e-05, "loss": 0.1169, "step": 10649 }, { "epoch": 0.68, "learning_rate": 4.827168693044577e-05, "loss": 0.1814, "step": 10650 }, { "epoch": 0.68, "learning_rate": 4.825391557108212e-05, "loss": 0.1776, "step": 10651 }, { "epoch": 0.68, "learning_rate": 4.823614644330744e-05, "loss": 0.1454, "step": 10652 }, { "epoch": 0.68, "learning_rate": 4.8218379547888024e-05, "loss": 0.1319, "step": 10653 }, { "epoch": 0.68, "learning_rate": 4.820061488559014e-05, "loss": 0.1489, "step": 10654 }, { "epoch": 0.68, "learning_rate": 4.818285245717984e-05, "loss": 0.135, "step": 10655 }, { "epoch": 0.68, "learning_rate": 4.816509226342313e-05, "loss": 0.1096, "step": 10656 }, { "epoch": 0.68, "learning_rate": 4.8147334305085945e-05, "loss": 0.1356, "step": 10657 }, { "epoch": 0.68, "learning_rate": 4.812957858293412e-05, "loss": 0.1316, "step": 10658 }, { "epoch": 0.68, "learning_rate": 4.8111825097733424e-05, "loss": 0.1395, "step": 10659 }, { "epoch": 0.68, "learning_rate": 4.809407385024941e-05, "loss": 0.132, "step": 10660 }, { "epoch": 0.68, "learning_rate": 4.80763248412477e-05, "loss": 0.1341, "step": 10661 }, { "epoch": 0.68, "learning_rate": 4.805857807149369e-05, "loss": 0.163, "step": 10662 }, { "epoch": 0.68, "learning_rate": 4.804083354175267e-05, "loss": 0.1412, "step": 10663 }, { "epoch": 0.68, "learning_rate": 4.802309125278994e-05, "loss": 0.1224, "step": 10664 }, { "epoch": 0.68, "learning_rate": 4.800535120537065e-05, "loss": 0.137, "step": 10665 }, { "epoch": 0.68, "learning_rate": 4.798761340025989e-05, "loss": 0.145, "step": 10666 }, { "epoch": 0.68, "learning_rate": 4.7969877838222544e-05, "loss": 0.1294, "step": 10667 }, { "epoch": 0.68, "learning_rate": 4.795214452002352e-05, "loss": 0.1343, "step": 10668 }, { "epoch": 0.68, "learning_rate": 4.7934413446427585e-05, "loss": 0.1459, "step": 10669 }, { "epoch": 0.68, "learning_rate": 4.791668461819934e-05, "loss": 0.1307, "step": 10670 }, { "epoch": 0.68, "learning_rate": 4.789895803610339e-05, "loss": 0.1382, "step": 10671 }, { "epoch": 0.68, "learning_rate": 4.788123370090423e-05, "loss": 0.1366, "step": 10672 }, { "epoch": 0.68, "learning_rate": 4.7863511613366254e-05, "loss": 0.1454, "step": 10673 }, { "epoch": 0.68, "learning_rate": 4.78457917742537e-05, "loss": 0.1569, "step": 10674 }, { "epoch": 0.68, "learning_rate": 4.782807418433072e-05, "loss": 0.1436, "step": 10675 }, { "epoch": 0.68, "learning_rate": 4.781035884436147e-05, "loss": 0.1481, "step": 10676 }, { "epoch": 0.68, "learning_rate": 4.779264575510988e-05, "loss": 0.1528, "step": 10677 }, { "epoch": 0.68, "learning_rate": 4.777493491733984e-05, "loss": 0.1422, "step": 10678 }, { "epoch": 0.68, "learning_rate": 4.7757226331815184e-05, "loss": 0.1414, "step": 10679 }, { "epoch": 0.68, "learning_rate": 4.773951999929962e-05, "loss": 0.1311, "step": 10680 }, { "epoch": 0.68, "learning_rate": 4.7721815920556724e-05, "loss": 0.1409, "step": 10681 }, { "epoch": 0.68, "learning_rate": 4.770411409634995e-05, "loss": 0.1326, "step": 10682 }, { "epoch": 0.68, "learning_rate": 4.768641452744277e-05, "loss": 0.1379, "step": 10683 }, { "epoch": 0.69, "learning_rate": 4.766871721459843e-05, "loss": 0.1345, "step": 10684 }, { "epoch": 0.69, "learning_rate": 4.7651022158580196e-05, "loss": 0.1413, "step": 10685 }, { "epoch": 0.69, "learning_rate": 4.7633329360151134e-05, "loss": 0.1344, "step": 10686 }, { "epoch": 0.69, "learning_rate": 4.761563882007434e-05, "loss": 0.152, "step": 10687 }, { "epoch": 0.69, "learning_rate": 4.759795053911267e-05, "loss": 0.131, "step": 10688 }, { "epoch": 0.69, "learning_rate": 4.758026451802891e-05, "loss": 0.1252, "step": 10689 }, { "epoch": 0.69, "learning_rate": 4.7562580757585864e-05, "loss": 0.1435, "step": 10690 }, { "epoch": 0.69, "learning_rate": 4.754489925854607e-05, "loss": 0.1298, "step": 10691 }, { "epoch": 0.69, "learning_rate": 4.752722002167211e-05, "loss": 0.1762, "step": 10692 }, { "epoch": 0.69, "learning_rate": 4.750954304772641e-05, "loss": 0.1577, "step": 10693 }, { "epoch": 0.69, "learning_rate": 4.749186833747135e-05, "loss": 0.1615, "step": 10694 }, { "epoch": 0.69, "learning_rate": 4.74741958916691e-05, "loss": 0.1295, "step": 10695 }, { "epoch": 0.69, "learning_rate": 4.745652571108178e-05, "loss": 0.14, "step": 10696 }, { "epoch": 0.69, "learning_rate": 4.7438857796471504e-05, "loss": 0.1603, "step": 10697 }, { "epoch": 0.69, "learning_rate": 4.742119214860009e-05, "loss": 0.1584, "step": 10698 }, { "epoch": 0.69, "learning_rate": 4.740352876822955e-05, "loss": 0.1611, "step": 10699 }, { "epoch": 0.69, "learning_rate": 4.7385867656121505e-05, "loss": 0.1532, "step": 10700 }, { "epoch": 0.69, "learning_rate": 4.736820881303769e-05, "loss": 0.1419, "step": 10701 }, { "epoch": 0.69, "learning_rate": 4.735055223973961e-05, "loss": 0.1335, "step": 10702 }, { "epoch": 0.69, "learning_rate": 4.7332897936988674e-05, "loss": 0.1405, "step": 10703 }, { "epoch": 0.69, "learning_rate": 4.731524590554628e-05, "loss": 0.1598, "step": 10704 }, { "epoch": 0.69, "learning_rate": 4.72975961461737e-05, "loss": 0.1446, "step": 10705 }, { "epoch": 0.69, "learning_rate": 4.727994865963211e-05, "loss": 0.1347, "step": 10706 }, { "epoch": 0.69, "learning_rate": 4.726230344668251e-05, "loss": 0.1288, "step": 10707 }, { "epoch": 0.69, "learning_rate": 4.7244660508085926e-05, "loss": 0.1391, "step": 10708 }, { "epoch": 0.69, "learning_rate": 4.7227019844603195e-05, "loss": 0.1408, "step": 10709 }, { "epoch": 0.69, "learning_rate": 4.720938145699504e-05, "loss": 0.1319, "step": 10710 }, { "epoch": 0.69, "learning_rate": 4.7191745346022175e-05, "loss": 0.135, "step": 10711 }, { "epoch": 0.69, "learning_rate": 4.717411151244517e-05, "loss": 0.1275, "step": 10712 }, { "epoch": 0.69, "learning_rate": 4.715647995702452e-05, "loss": 0.1254, "step": 10713 }, { "epoch": 0.69, "learning_rate": 4.7138850680520554e-05, "loss": 0.1408, "step": 10714 }, { "epoch": 0.69, "learning_rate": 4.7121223683693596e-05, "loss": 0.1453, "step": 10715 }, { "epoch": 0.69, "learning_rate": 4.710359896730379e-05, "loss": 0.1374, "step": 10716 }, { "epoch": 0.69, "learning_rate": 4.7085976532111184e-05, "loss": 0.1521, "step": 10717 }, { "epoch": 0.69, "learning_rate": 4.70683563788758e-05, "loss": 0.1492, "step": 10718 }, { "epoch": 0.69, "learning_rate": 4.705073850835751e-05, "loss": 0.1384, "step": 10719 }, { "epoch": 0.69, "learning_rate": 4.703312292131614e-05, "loss": 0.1352, "step": 10720 }, { "epoch": 0.69, "learning_rate": 4.70155096185113e-05, "loss": 0.1244, "step": 10721 }, { "epoch": 0.69, "learning_rate": 4.699789860070265e-05, "loss": 0.1426, "step": 10722 }, { "epoch": 0.69, "learning_rate": 4.698028986864965e-05, "loss": 0.1472, "step": 10723 }, { "epoch": 0.69, "learning_rate": 4.696268342311164e-05, "loss": 0.1396, "step": 10724 }, { "epoch": 0.69, "learning_rate": 4.6945079264847966e-05, "loss": 0.1488, "step": 10725 }, { "epoch": 0.69, "learning_rate": 4.69274773946178e-05, "loss": 0.1436, "step": 10726 }, { "epoch": 0.69, "learning_rate": 4.690987781318029e-05, "loss": 0.1293, "step": 10727 }, { "epoch": 0.69, "learning_rate": 4.689228052129435e-05, "loss": 0.1452, "step": 10728 }, { "epoch": 0.69, "learning_rate": 4.6874685519718945e-05, "loss": 0.1385, "step": 10729 }, { "epoch": 0.69, "learning_rate": 4.6857092809212845e-05, "loss": 0.1395, "step": 10730 }, { "epoch": 0.69, "learning_rate": 4.6839502390534694e-05, "loss": 0.1537, "step": 10731 }, { "epoch": 0.69, "learning_rate": 4.6821914264443156e-05, "loss": 0.1308, "step": 10732 }, { "epoch": 0.69, "learning_rate": 4.680432843169672e-05, "loss": 0.1365, "step": 10733 }, { "epoch": 0.69, "learning_rate": 4.6786744893053824e-05, "loss": 0.1486, "step": 10734 }, { "epoch": 0.69, "learning_rate": 4.6769163649272694e-05, "loss": 0.1586, "step": 10735 }, { "epoch": 0.69, "learning_rate": 4.675158470111162e-05, "loss": 0.1296, "step": 10736 }, { "epoch": 0.69, "learning_rate": 4.673400804932865e-05, "loss": 0.1675, "step": 10737 }, { "epoch": 0.69, "learning_rate": 4.671643369468177e-05, "loss": 0.135, "step": 10738 }, { "epoch": 0.69, "learning_rate": 4.6698861637928927e-05, "loss": 0.1614, "step": 10739 }, { "epoch": 0.69, "learning_rate": 4.668129187982791e-05, "loss": 0.1575, "step": 10740 }, { "epoch": 0.69, "learning_rate": 4.66637244211365e-05, "loss": 0.123, "step": 10741 }, { "epoch": 0.69, "learning_rate": 4.66461592626122e-05, "loss": 0.1602, "step": 10742 }, { "epoch": 0.69, "learning_rate": 4.6628596405012604e-05, "loss": 0.1476, "step": 10743 }, { "epoch": 0.69, "learning_rate": 4.661103584909504e-05, "loss": 0.1454, "step": 10744 }, { "epoch": 0.69, "learning_rate": 4.659347759561688e-05, "loss": 0.1159, "step": 10745 }, { "epoch": 0.69, "learning_rate": 4.657592164533536e-05, "loss": 0.1279, "step": 10746 }, { "epoch": 0.69, "learning_rate": 4.6558367999007525e-05, "loss": 0.1511, "step": 10747 }, { "epoch": 0.69, "learning_rate": 4.654081665739045e-05, "loss": 0.1492, "step": 10748 }, { "epoch": 0.69, "learning_rate": 4.652326762124099e-05, "loss": 0.1311, "step": 10749 }, { "epoch": 0.69, "learning_rate": 4.650572089131604e-05, "loss": 0.1542, "step": 10750 }, { "epoch": 0.69, "learning_rate": 4.648817646837222e-05, "loss": 0.1455, "step": 10751 }, { "epoch": 0.69, "learning_rate": 4.647063435316621e-05, "loss": 0.1492, "step": 10752 }, { "epoch": 0.69, "learning_rate": 4.645309454645453e-05, "loss": 0.1277, "step": 10753 }, { "epoch": 0.69, "learning_rate": 4.643555704899356e-05, "loss": 0.1426, "step": 10754 }, { "epoch": 0.69, "learning_rate": 4.641802186153966e-05, "loss": 0.1207, "step": 10755 }, { "epoch": 0.69, "learning_rate": 4.640048898484904e-05, "loss": 0.1381, "step": 10756 }, { "epoch": 0.69, "learning_rate": 4.638295841967776e-05, "loss": 0.1319, "step": 10757 }, { "epoch": 0.69, "learning_rate": 4.636543016678189e-05, "loss": 0.1354, "step": 10758 }, { "epoch": 0.69, "learning_rate": 4.6347904226917336e-05, "loss": 0.1529, "step": 10759 }, { "epoch": 0.69, "learning_rate": 4.633038060083996e-05, "loss": 0.1398, "step": 10760 }, { "epoch": 0.69, "learning_rate": 4.631285928930541e-05, "loss": 0.1366, "step": 10761 }, { "epoch": 0.69, "learning_rate": 4.629534029306938e-05, "loss": 0.1394, "step": 10762 }, { "epoch": 0.69, "learning_rate": 4.627782361288735e-05, "loss": 0.1365, "step": 10763 }, { "epoch": 0.69, "learning_rate": 4.626030924951471e-05, "loss": 0.1396, "step": 10764 }, { "epoch": 0.69, "learning_rate": 4.624279720370681e-05, "loss": 0.1284, "step": 10765 }, { "epoch": 0.69, "learning_rate": 4.622528747621887e-05, "loss": 0.1384, "step": 10766 }, { "epoch": 0.69, "learning_rate": 4.620778006780604e-05, "loss": 0.1406, "step": 10767 }, { "epoch": 0.69, "learning_rate": 4.6190274979223294e-05, "loss": 0.1442, "step": 10768 }, { "epoch": 0.69, "learning_rate": 4.617277221122559e-05, "loss": 0.1417, "step": 10769 }, { "epoch": 0.69, "learning_rate": 4.615527176456773e-05, "loss": 0.134, "step": 10770 }, { "epoch": 0.69, "learning_rate": 4.6137773640004404e-05, "loss": 0.16, "step": 10771 }, { "epoch": 0.69, "learning_rate": 4.612027783829026e-05, "loss": 0.1315, "step": 10772 }, { "epoch": 0.69, "learning_rate": 4.610278436017981e-05, "loss": 0.1186, "step": 10773 }, { "epoch": 0.69, "learning_rate": 4.6085293206427525e-05, "loss": 0.1387, "step": 10774 }, { "epoch": 0.69, "learning_rate": 4.6067804377787635e-05, "loss": 0.1254, "step": 10775 }, { "epoch": 0.69, "learning_rate": 4.6050317875014445e-05, "loss": 0.1424, "step": 10776 }, { "epoch": 0.69, "learning_rate": 4.6032833698862044e-05, "loss": 0.1238, "step": 10777 }, { "epoch": 0.69, "learning_rate": 4.60153518500844e-05, "loss": 0.1377, "step": 10778 }, { "epoch": 0.69, "learning_rate": 4.599787232943545e-05, "loss": 0.128, "step": 10779 }, { "epoch": 0.69, "learning_rate": 4.5980395137669055e-05, "loss": 0.1351, "step": 10780 }, { "epoch": 0.69, "learning_rate": 4.596292027553893e-05, "loss": 0.1235, "step": 10781 }, { "epoch": 0.69, "learning_rate": 4.5945447743798656e-05, "loss": 0.1377, "step": 10782 }, { "epoch": 0.69, "learning_rate": 4.5927977543201784e-05, "loss": 0.1282, "step": 10783 }, { "epoch": 0.69, "learning_rate": 4.591050967450171e-05, "loss": 0.1409, "step": 10784 }, { "epoch": 0.69, "learning_rate": 4.589304413845167e-05, "loss": 0.1248, "step": 10785 }, { "epoch": 0.69, "learning_rate": 4.587558093580504e-05, "loss": 0.1397, "step": 10786 }, { "epoch": 0.69, "learning_rate": 4.5858120067314815e-05, "loss": 0.1347, "step": 10787 }, { "epoch": 0.69, "learning_rate": 4.584066153373408e-05, "loss": 0.1265, "step": 10788 }, { "epoch": 0.69, "learning_rate": 4.582320533581566e-05, "loss": 0.1302, "step": 10789 }, { "epoch": 0.69, "learning_rate": 4.5805751474312464e-05, "loss": 0.1545, "step": 10790 }, { "epoch": 0.69, "learning_rate": 4.578829994997712e-05, "loss": 0.1554, "step": 10791 }, { "epoch": 0.69, "learning_rate": 4.577085076356228e-05, "loss": 0.1441, "step": 10792 }, { "epoch": 0.69, "learning_rate": 4.575340391582048e-05, "loss": 0.1315, "step": 10793 }, { "epoch": 0.69, "learning_rate": 4.573595940750406e-05, "loss": 0.1324, "step": 10794 }, { "epoch": 0.69, "learning_rate": 4.57185172393654e-05, "loss": 0.1454, "step": 10795 }, { "epoch": 0.69, "learning_rate": 4.5701077412156636e-05, "loss": 0.1242, "step": 10796 }, { "epoch": 0.69, "learning_rate": 4.568363992662995e-05, "loss": 0.1309, "step": 10797 }, { "epoch": 0.69, "learning_rate": 4.566620478353726e-05, "loss": 0.1405, "step": 10798 }, { "epoch": 0.69, "learning_rate": 4.5648771983630525e-05, "loss": 0.1289, "step": 10799 }, { "epoch": 0.69, "learning_rate": 4.563134152766156e-05, "loss": 0.1471, "step": 10800 }, { "epoch": 0.69, "learning_rate": 4.5613913416382026e-05, "loss": 0.1321, "step": 10801 }, { "epoch": 0.69, "learning_rate": 4.559648765054357e-05, "loss": 0.1556, "step": 10802 }, { "epoch": 0.69, "learning_rate": 4.557906423089763e-05, "loss": 0.1385, "step": 10803 }, { "epoch": 0.69, "learning_rate": 4.5561643158195667e-05, "loss": 0.1475, "step": 10804 }, { "epoch": 0.69, "learning_rate": 4.554422443318891e-05, "loss": 0.151, "step": 10805 }, { "epoch": 0.69, "learning_rate": 4.55268080566286e-05, "loss": 0.1418, "step": 10806 }, { "epoch": 0.69, "learning_rate": 4.550939402926586e-05, "loss": 0.1464, "step": 10807 }, { "epoch": 0.69, "learning_rate": 4.549198235185161e-05, "loss": 0.1363, "step": 10808 }, { "epoch": 0.69, "learning_rate": 4.5474573025136813e-05, "loss": 0.151, "step": 10809 }, { "epoch": 0.69, "learning_rate": 4.5457166049872205e-05, "loss": 0.1468, "step": 10810 }, { "epoch": 0.69, "learning_rate": 4.5439761426808516e-05, "loss": 0.1421, "step": 10811 }, { "epoch": 0.69, "learning_rate": 4.542235915669627e-05, "loss": 0.139, "step": 10812 }, { "epoch": 0.69, "learning_rate": 4.5404959240286006e-05, "loss": 0.1491, "step": 10813 }, { "epoch": 0.69, "learning_rate": 4.538756167832814e-05, "loss": 0.1423, "step": 10814 }, { "epoch": 0.69, "learning_rate": 4.537016647157287e-05, "loss": 0.1385, "step": 10815 }, { "epoch": 0.69, "learning_rate": 4.535277362077046e-05, "loss": 0.1395, "step": 10816 }, { "epoch": 0.69, "learning_rate": 4.533538312667092e-05, "loss": 0.119, "step": 10817 }, { "epoch": 0.69, "learning_rate": 4.5317994990024285e-05, "loss": 0.1389, "step": 10818 }, { "epoch": 0.69, "learning_rate": 4.530060921158037e-05, "loss": 0.1488, "step": 10819 }, { "epoch": 0.69, "learning_rate": 4.528322579208899e-05, "loss": 0.1357, "step": 10820 }, { "epoch": 0.69, "learning_rate": 4.5265844732299844e-05, "loss": 0.1289, "step": 10821 }, { "epoch": 0.69, "learning_rate": 4.524846603296244e-05, "loss": 0.1608, "step": 10822 }, { "epoch": 0.69, "learning_rate": 4.523108969482632e-05, "loss": 0.1267, "step": 10823 }, { "epoch": 0.69, "learning_rate": 4.521371571864077e-05, "loss": 0.1416, "step": 10824 }, { "epoch": 0.69, "learning_rate": 4.519634410515514e-05, "loss": 0.1434, "step": 10825 }, { "epoch": 0.69, "learning_rate": 4.5178974855118505e-05, "loss": 0.1436, "step": 10826 }, { "epoch": 0.69, "learning_rate": 4.5161607969279985e-05, "loss": 0.1364, "step": 10827 }, { "epoch": 0.69, "learning_rate": 4.514424344838856e-05, "loss": 0.1492, "step": 10828 }, { "epoch": 0.69, "learning_rate": 4.5126881293193026e-05, "loss": 0.1609, "step": 10829 }, { "epoch": 0.69, "learning_rate": 4.51095215044422e-05, "loss": 0.1638, "step": 10830 }, { "epoch": 0.69, "learning_rate": 4.509216408288467e-05, "loss": 0.1432, "step": 10831 }, { "epoch": 0.69, "learning_rate": 4.507480902926903e-05, "loss": 0.1482, "step": 10832 }, { "epoch": 0.69, "learning_rate": 4.505745634434375e-05, "loss": 0.1273, "step": 10833 }, { "epoch": 0.69, "learning_rate": 4.5040106028857124e-05, "loss": 0.1565, "step": 10834 }, { "epoch": 0.69, "learning_rate": 4.502275808355746e-05, "loss": 0.1511, "step": 10835 }, { "epoch": 0.69, "learning_rate": 4.500541250919283e-05, "loss": 0.1242, "step": 10836 }, { "epoch": 0.69, "learning_rate": 4.498806930651134e-05, "loss": 0.1307, "step": 10837 }, { "epoch": 0.69, "learning_rate": 4.497072847626087e-05, "loss": 0.1325, "step": 10838 }, { "epoch": 0.69, "learning_rate": 4.4953390019189276e-05, "loss": 0.1413, "step": 10839 }, { "epoch": 0.7, "learning_rate": 4.4936053936044356e-05, "loss": 0.1226, "step": 10840 }, { "epoch": 0.7, "learning_rate": 4.491872022757364e-05, "loss": 0.1401, "step": 10841 }, { "epoch": 0.7, "learning_rate": 4.490138889452474e-05, "loss": 0.138, "step": 10842 }, { "epoch": 0.7, "learning_rate": 4.4884059937645016e-05, "loss": 0.1306, "step": 10843 }, { "epoch": 0.7, "learning_rate": 4.4866733357681867e-05, "loss": 0.1243, "step": 10844 }, { "epoch": 0.7, "learning_rate": 4.484940915538243e-05, "loss": 0.1382, "step": 10845 }, { "epoch": 0.7, "learning_rate": 4.4832087331493864e-05, "loss": 0.1322, "step": 10846 }, { "epoch": 0.7, "learning_rate": 4.481476788676323e-05, "loss": 0.1427, "step": 10847 }, { "epoch": 0.7, "learning_rate": 4.4797450821937356e-05, "loss": 0.1321, "step": 10848 }, { "epoch": 0.7, "learning_rate": 4.478013613776314e-05, "loss": 0.1399, "step": 10849 }, { "epoch": 0.7, "learning_rate": 4.476282383498721e-05, "loss": 0.1476, "step": 10850 }, { "epoch": 0.7, "learning_rate": 4.4745513914356253e-05, "loss": 0.1299, "step": 10851 }, { "epoch": 0.7, "learning_rate": 4.4728206376616686e-05, "loss": 0.151, "step": 10852 }, { "epoch": 0.7, "learning_rate": 4.4710901222514954e-05, "loss": 0.1234, "step": 10853 }, { "epoch": 0.7, "learning_rate": 4.469359845279739e-05, "loss": 0.1389, "step": 10854 }, { "epoch": 0.7, "learning_rate": 4.467629806821012e-05, "loss": 0.1497, "step": 10855 }, { "epoch": 0.7, "learning_rate": 4.465900006949931e-05, "loss": 0.1455, "step": 10856 }, { "epoch": 0.7, "learning_rate": 4.464170445741087e-05, "loss": 0.1482, "step": 10857 }, { "epoch": 0.7, "learning_rate": 4.462441123269077e-05, "loss": 0.143, "step": 10858 }, { "epoch": 0.7, "learning_rate": 4.46071203960847e-05, "loss": 0.1287, "step": 10859 }, { "epoch": 0.7, "learning_rate": 4.458983194833839e-05, "loss": 0.1386, "step": 10860 }, { "epoch": 0.7, "learning_rate": 4.457254589019746e-05, "loss": 0.1425, "step": 10861 }, { "epoch": 0.7, "learning_rate": 4.4555262222407304e-05, "loss": 0.1417, "step": 10862 }, { "epoch": 0.7, "learning_rate": 4.4537980945713354e-05, "loss": 0.1246, "step": 10863 }, { "epoch": 0.7, "learning_rate": 4.452070206086083e-05, "loss": 0.1261, "step": 10864 }, { "epoch": 0.7, "learning_rate": 4.450342556859495e-05, "loss": 0.1394, "step": 10865 }, { "epoch": 0.7, "learning_rate": 4.448615146966072e-05, "loss": 0.1415, "step": 10866 }, { "epoch": 0.7, "learning_rate": 4.446887976480312e-05, "loss": 0.1179, "step": 10867 }, { "epoch": 0.7, "learning_rate": 4.445161045476704e-05, "loss": 0.1457, "step": 10868 }, { "epoch": 0.7, "learning_rate": 4.443434354029718e-05, "loss": 0.1278, "step": 10869 }, { "epoch": 0.7, "learning_rate": 4.4417079022138244e-05, "loss": 0.1573, "step": 10870 }, { "epoch": 0.7, "learning_rate": 4.43998169010347e-05, "loss": 0.1392, "step": 10871 }, { "epoch": 0.7, "learning_rate": 4.4382557177731046e-05, "loss": 0.1387, "step": 10872 }, { "epoch": 0.7, "learning_rate": 4.436529985297163e-05, "loss": 0.1279, "step": 10873 }, { "epoch": 0.7, "learning_rate": 4.434804492750064e-05, "loss": 0.1416, "step": 10874 }, { "epoch": 0.7, "learning_rate": 4.4330792402062285e-05, "loss": 0.1463, "step": 10875 }, { "epoch": 0.7, "learning_rate": 4.431354227740049e-05, "loss": 0.1465, "step": 10876 }, { "epoch": 0.7, "learning_rate": 4.429629455425926e-05, "loss": 0.1456, "step": 10877 }, { "epoch": 0.7, "learning_rate": 4.427904923338237e-05, "loss": 0.1503, "step": 10878 }, { "epoch": 0.7, "learning_rate": 4.426180631551355e-05, "loss": 0.1372, "step": 10879 }, { "epoch": 0.7, "learning_rate": 4.4244565801396456e-05, "loss": 0.1571, "step": 10880 }, { "epoch": 0.7, "learning_rate": 4.422732769177452e-05, "loss": 0.1403, "step": 10881 }, { "epoch": 0.7, "learning_rate": 4.4210091987391234e-05, "loss": 0.1393, "step": 10882 }, { "epoch": 0.7, "learning_rate": 4.419285868898983e-05, "loss": 0.1202, "step": 10883 }, { "epoch": 0.7, "learning_rate": 4.417562779731355e-05, "loss": 0.1499, "step": 10884 }, { "epoch": 0.7, "learning_rate": 4.415839931310545e-05, "loss": 0.1493, "step": 10885 }, { "epoch": 0.7, "learning_rate": 4.414117323710853e-05, "loss": 0.136, "step": 10886 }, { "epoch": 0.7, "learning_rate": 4.412394957006575e-05, "loss": 0.1503, "step": 10887 }, { "epoch": 0.7, "learning_rate": 4.410672831271978e-05, "loss": 0.1393, "step": 10888 }, { "epoch": 0.7, "learning_rate": 4.4089509465813406e-05, "loss": 0.1366, "step": 10889 }, { "epoch": 0.7, "learning_rate": 4.4072293030089106e-05, "loss": 0.1281, "step": 10890 }, { "epoch": 0.7, "learning_rate": 4.4055079006289444e-05, "loss": 0.1541, "step": 10891 }, { "epoch": 0.7, "learning_rate": 4.4037867395156696e-05, "loss": 0.153, "step": 10892 }, { "epoch": 0.7, "learning_rate": 4.4020658197433184e-05, "loss": 0.1435, "step": 10893 }, { "epoch": 0.7, "learning_rate": 4.400345141386108e-05, "loss": 0.1407, "step": 10894 }, { "epoch": 0.7, "learning_rate": 4.3986247045182395e-05, "loss": 0.1394, "step": 10895 }, { "epoch": 0.7, "learning_rate": 4.396904509213913e-05, "loss": 0.148, "step": 10896 }, { "epoch": 0.7, "learning_rate": 4.395184555547307e-05, "loss": 0.139, "step": 10897 }, { "epoch": 0.7, "learning_rate": 4.393464843592603e-05, "loss": 0.1419, "step": 10898 }, { "epoch": 0.7, "learning_rate": 4.3917453734239566e-05, "loss": 0.1281, "step": 10899 }, { "epoch": 0.7, "learning_rate": 4.390026145115526e-05, "loss": 0.1325, "step": 10900 }, { "epoch": 0.7, "learning_rate": 4.388307158741457e-05, "loss": 0.1449, "step": 10901 }, { "epoch": 0.7, "learning_rate": 4.386588414375876e-05, "loss": 0.1585, "step": 10902 }, { "epoch": 0.7, "learning_rate": 4.384869912092913e-05, "loss": 0.1321, "step": 10903 }, { "epoch": 0.7, "learning_rate": 4.383151651966669e-05, "loss": 0.1437, "step": 10904 }, { "epoch": 0.7, "learning_rate": 4.3814336340712556e-05, "loss": 0.1192, "step": 10905 }, { "epoch": 0.7, "learning_rate": 4.3797158584807566e-05, "loss": 0.1294, "step": 10906 }, { "epoch": 0.7, "learning_rate": 4.3779983252692545e-05, "loss": 0.1419, "step": 10907 }, { "epoch": 0.7, "learning_rate": 4.376281034510823e-05, "loss": 0.1421, "step": 10908 }, { "epoch": 0.7, "learning_rate": 4.374563986279515e-05, "loss": 0.1415, "step": 10909 }, { "epoch": 0.7, "learning_rate": 4.3728471806493856e-05, "loss": 0.1424, "step": 10910 }, { "epoch": 0.7, "learning_rate": 4.371130617694468e-05, "loss": 0.1284, "step": 10911 }, { "epoch": 0.7, "learning_rate": 4.369414297488792e-05, "loss": 0.146, "step": 10912 }, { "epoch": 0.7, "learning_rate": 4.36769822010638e-05, "loss": 0.1469, "step": 10913 }, { "epoch": 0.7, "learning_rate": 4.365982385621231e-05, "loss": 0.1262, "step": 10914 }, { "epoch": 0.7, "learning_rate": 4.36426679410735e-05, "loss": 0.1537, "step": 10915 }, { "epoch": 0.7, "learning_rate": 4.362551445638715e-05, "loss": 0.1281, "step": 10916 }, { "epoch": 0.7, "learning_rate": 4.360836340289309e-05, "loss": 0.1696, "step": 10917 }, { "epoch": 0.7, "learning_rate": 4.35912147813309e-05, "loss": 0.1287, "step": 10918 }, { "epoch": 0.7, "learning_rate": 4.357406859244018e-05, "loss": 0.147, "step": 10919 }, { "epoch": 0.7, "learning_rate": 4.3556924836960386e-05, "loss": 0.1566, "step": 10920 }, { "epoch": 0.7, "learning_rate": 4.353978351563079e-05, "loss": 0.1626, "step": 10921 }, { "epoch": 0.7, "learning_rate": 4.35226446291907e-05, "loss": 0.1445, "step": 10922 }, { "epoch": 0.7, "learning_rate": 4.3505508178379176e-05, "loss": 0.1526, "step": 10923 }, { "epoch": 0.7, "learning_rate": 4.348837416393529e-05, "loss": 0.1495, "step": 10924 }, { "epoch": 0.7, "learning_rate": 4.3471242586597915e-05, "loss": 0.1235, "step": 10925 }, { "epoch": 0.7, "learning_rate": 4.3454113447105884e-05, "loss": 0.1376, "step": 10926 }, { "epoch": 0.7, "learning_rate": 4.3436986746197937e-05, "loss": 0.1297, "step": 10927 }, { "epoch": 0.7, "learning_rate": 4.341986248461262e-05, "loss": 0.154, "step": 10928 }, { "epoch": 0.7, "learning_rate": 4.3402740663088494e-05, "loss": 0.1287, "step": 10929 }, { "epoch": 0.7, "learning_rate": 4.3385621282363865e-05, "loss": 0.143, "step": 10930 }, { "epoch": 0.7, "learning_rate": 4.336850434317712e-05, "loss": 0.1683, "step": 10931 }, { "epoch": 0.7, "learning_rate": 4.335138984626633e-05, "loss": 0.1495, "step": 10932 }, { "epoch": 0.7, "learning_rate": 4.3334277792369635e-05, "loss": 0.1539, "step": 10933 }, { "epoch": 0.7, "learning_rate": 4.331716818222504e-05, "loss": 0.1363, "step": 10934 }, { "epoch": 0.7, "learning_rate": 4.3300061016570325e-05, "loss": 0.147, "step": 10935 }, { "epoch": 0.7, "learning_rate": 4.328295629614334e-05, "loss": 0.1341, "step": 10936 }, { "epoch": 0.7, "learning_rate": 4.326585402168164e-05, "loss": 0.1406, "step": 10937 }, { "epoch": 0.7, "learning_rate": 4.3248754193922866e-05, "loss": 0.1459, "step": 10938 }, { "epoch": 0.7, "learning_rate": 4.323165681360438e-05, "loss": 0.1373, "step": 10939 }, { "epoch": 0.7, "learning_rate": 4.321456188146357e-05, "loss": 0.1547, "step": 10940 }, { "epoch": 0.7, "learning_rate": 4.319746939823768e-05, "loss": 0.139, "step": 10941 }, { "epoch": 0.7, "learning_rate": 4.3180379364663784e-05, "loss": 0.1422, "step": 10942 }, { "epoch": 0.7, "learning_rate": 4.316329178147896e-05, "loss": 0.1305, "step": 10943 }, { "epoch": 0.7, "learning_rate": 4.314620664942006e-05, "loss": 0.1404, "step": 10944 }, { "epoch": 0.7, "learning_rate": 4.312912396922397e-05, "loss": 0.1322, "step": 10945 }, { "epoch": 0.7, "learning_rate": 4.311204374162731e-05, "loss": 0.124, "step": 10946 }, { "epoch": 0.7, "learning_rate": 4.309496596736672e-05, "loss": 0.1323, "step": 10947 }, { "epoch": 0.7, "learning_rate": 4.307789064717873e-05, "loss": 0.1153, "step": 10948 }, { "epoch": 0.7, "learning_rate": 4.3060817781799646e-05, "loss": 0.1346, "step": 10949 }, { "epoch": 0.7, "learning_rate": 4.304374737196582e-05, "loss": 0.1205, "step": 10950 }, { "epoch": 0.7, "learning_rate": 4.302667941841337e-05, "loss": 0.1218, "step": 10951 }, { "epoch": 0.7, "learning_rate": 4.300961392187842e-05, "loss": 0.1409, "step": 10952 }, { "epoch": 0.7, "learning_rate": 4.299255088309684e-05, "loss": 0.1606, "step": 10953 }, { "epoch": 0.7, "learning_rate": 4.297549030280461e-05, "loss": 0.1365, "step": 10954 }, { "epoch": 0.7, "learning_rate": 4.295843218173742e-05, "loss": 0.1383, "step": 10955 }, { "epoch": 0.7, "learning_rate": 4.294137652063087e-05, "loss": 0.124, "step": 10956 }, { "epoch": 0.7, "learning_rate": 4.29243233202206e-05, "loss": 0.1397, "step": 10957 }, { "epoch": 0.7, "learning_rate": 4.290727258124193e-05, "loss": 0.1446, "step": 10958 }, { "epoch": 0.7, "learning_rate": 4.289022430443024e-05, "loss": 0.1531, "step": 10959 }, { "epoch": 0.7, "learning_rate": 4.287317849052075e-05, "loss": 0.1468, "step": 10960 }, { "epoch": 0.7, "learning_rate": 4.285613514024861e-05, "loss": 0.1299, "step": 10961 }, { "epoch": 0.7, "learning_rate": 4.28390942543488e-05, "loss": 0.1415, "step": 10962 }, { "epoch": 0.7, "learning_rate": 4.282205583355616e-05, "loss": 0.1364, "step": 10963 }, { "epoch": 0.7, "learning_rate": 4.280501987860559e-05, "loss": 0.1275, "step": 10964 }, { "epoch": 0.7, "learning_rate": 4.278798639023167e-05, "loss": 0.1381, "step": 10965 }, { "epoch": 0.7, "learning_rate": 4.2770955369169055e-05, "loss": 0.1428, "step": 10966 }, { "epoch": 0.7, "learning_rate": 4.275392681615219e-05, "loss": 0.1359, "step": 10967 }, { "epoch": 0.7, "learning_rate": 4.27369007319155e-05, "loss": 0.1456, "step": 10968 }, { "epoch": 0.7, "learning_rate": 4.2719877117193196e-05, "loss": 0.1398, "step": 10969 }, { "epoch": 0.7, "learning_rate": 4.2702855972719414e-05, "loss": 0.1355, "step": 10970 }, { "epoch": 0.7, "learning_rate": 4.268583729922826e-05, "loss": 0.1511, "step": 10971 }, { "epoch": 0.7, "learning_rate": 4.266882109745361e-05, "loss": 0.1422, "step": 10972 }, { "epoch": 0.7, "learning_rate": 4.265180736812936e-05, "loss": 0.1257, "step": 10973 }, { "epoch": 0.7, "learning_rate": 4.2634796111989204e-05, "loss": 0.1545, "step": 10974 }, { "epoch": 0.7, "learning_rate": 4.261778732976682e-05, "loss": 0.171, "step": 10975 }, { "epoch": 0.7, "learning_rate": 4.260078102219569e-05, "loss": 0.1462, "step": 10976 }, { "epoch": 0.7, "learning_rate": 4.2583777190009176e-05, "loss": 0.126, "step": 10977 }, { "epoch": 0.7, "learning_rate": 4.256677583394066e-05, "loss": 0.147, "step": 10978 }, { "epoch": 0.7, "learning_rate": 4.2549776954723284e-05, "loss": 0.1412, "step": 10979 }, { "epoch": 0.7, "learning_rate": 4.2532780553090145e-05, "loss": 0.1546, "step": 10980 }, { "epoch": 0.7, "learning_rate": 4.251578662977423e-05, "loss": 0.1397, "step": 10981 }, { "epoch": 0.7, "learning_rate": 4.249879518550847e-05, "loss": 0.1236, "step": 10982 }, { "epoch": 0.7, "learning_rate": 4.248180622102558e-05, "loss": 0.1346, "step": 10983 }, { "epoch": 0.7, "learning_rate": 4.246481973705819e-05, "loss": 0.125, "step": 10984 }, { "epoch": 0.7, "learning_rate": 4.244783573433893e-05, "loss": 0.1486, "step": 10985 }, { "epoch": 0.7, "learning_rate": 4.243085421360018e-05, "loss": 0.1417, "step": 10986 }, { "epoch": 0.7, "learning_rate": 4.241387517557431e-05, "loss": 0.1351, "step": 10987 }, { "epoch": 0.7, "learning_rate": 4.239689862099359e-05, "loss": 0.1179, "step": 10988 }, { "epoch": 0.7, "learning_rate": 4.237992455059008e-05, "loss": 0.1477, "step": 10989 }, { "epoch": 0.7, "learning_rate": 4.236295296509586e-05, "loss": 0.1436, "step": 10990 }, { "epoch": 0.7, "learning_rate": 4.234598386524278e-05, "loss": 0.1444, "step": 10991 }, { "epoch": 0.7, "learning_rate": 4.2329017251762716e-05, "loss": 0.1402, "step": 10992 }, { "epoch": 0.7, "learning_rate": 4.23120531253873e-05, "loss": 0.1257, "step": 10993 }, { "epoch": 0.7, "learning_rate": 4.2295091486848134e-05, "loss": 0.1315, "step": 10994 }, { "epoch": 0.7, "learning_rate": 4.227813233687677e-05, "loss": 0.1326, "step": 10995 }, { "epoch": 0.71, "learning_rate": 4.22611756762045e-05, "loss": 0.1561, "step": 10996 }, { "epoch": 0.71, "learning_rate": 4.2244221505562655e-05, "loss": 0.1266, "step": 10997 }, { "epoch": 0.71, "learning_rate": 4.222726982568233e-05, "loss": 0.139, "step": 10998 }, { "epoch": 0.71, "learning_rate": 4.221032063729463e-05, "loss": 0.1426, "step": 10999 }, { "epoch": 0.71, "learning_rate": 4.2193373941130475e-05, "loss": 0.1432, "step": 11000 }, { "epoch": 0.71, "learning_rate": 4.217642973792076e-05, "loss": 0.1604, "step": 11001 }, { "epoch": 0.71, "learning_rate": 4.215948802839618e-05, "loss": 0.1425, "step": 11002 }, { "epoch": 0.71, "learning_rate": 4.214254881328731e-05, "loss": 0.1491, "step": 11003 }, { "epoch": 0.71, "learning_rate": 4.212561209332474e-05, "loss": 0.1312, "step": 11004 }, { "epoch": 0.71, "learning_rate": 4.210867786923882e-05, "loss": 0.1209, "step": 11005 }, { "epoch": 0.71, "learning_rate": 4.209174614175988e-05, "loss": 0.1427, "step": 11006 }, { "epoch": 0.71, "learning_rate": 4.207481691161811e-05, "loss": 0.1409, "step": 11007 }, { "epoch": 0.71, "learning_rate": 4.205789017954364e-05, "loss": 0.1478, "step": 11008 }, { "epoch": 0.71, "learning_rate": 4.20409659462664e-05, "loss": 0.1499, "step": 11009 }, { "epoch": 0.71, "learning_rate": 4.202404421251623e-05, "loss": 0.1365, "step": 11010 }, { "epoch": 0.71, "learning_rate": 4.200712497902297e-05, "loss": 0.1212, "step": 11011 }, { "epoch": 0.71, "learning_rate": 4.1990208246516205e-05, "loss": 0.1383, "step": 11012 }, { "epoch": 0.71, "learning_rate": 4.197329401572551e-05, "loss": 0.1237, "step": 11013 }, { "epoch": 0.71, "learning_rate": 4.1956382287380316e-05, "loss": 0.1861, "step": 11014 }, { "epoch": 0.71, "learning_rate": 4.193947306221001e-05, "loss": 0.1342, "step": 11015 }, { "epoch": 0.71, "learning_rate": 4.1922566340943756e-05, "loss": 0.1334, "step": 11016 }, { "epoch": 0.71, "learning_rate": 4.190566212431066e-05, "loss": 0.1382, "step": 11017 }, { "epoch": 0.71, "learning_rate": 4.188876041303978e-05, "loss": 0.1385, "step": 11018 }, { "epoch": 0.71, "learning_rate": 4.1871861207859955e-05, "loss": 0.138, "step": 11019 }, { "epoch": 0.71, "learning_rate": 4.1854964509499994e-05, "loss": 0.1282, "step": 11020 }, { "epoch": 0.71, "learning_rate": 4.1838070318688604e-05, "loss": 0.1335, "step": 11021 }, { "epoch": 0.71, "learning_rate": 4.1821178636154376e-05, "loss": 0.1556, "step": 11022 }, { "epoch": 0.71, "learning_rate": 4.1804289462625766e-05, "loss": 0.1356, "step": 11023 }, { "epoch": 0.71, "learning_rate": 4.1787402798831065e-05, "loss": 0.1281, "step": 11024 }, { "epoch": 0.71, "learning_rate": 4.1770518645498613e-05, "loss": 0.1334, "step": 11025 }, { "epoch": 0.71, "learning_rate": 4.175363700335648e-05, "loss": 0.1446, "step": 11026 }, { "epoch": 0.71, "learning_rate": 4.173675787313273e-05, "loss": 0.1239, "step": 11027 }, { "epoch": 0.71, "learning_rate": 4.171988125555529e-05, "loss": 0.1405, "step": 11028 }, { "epoch": 0.71, "learning_rate": 4.170300715135201e-05, "loss": 0.1295, "step": 11029 }, { "epoch": 0.71, "learning_rate": 4.168613556125057e-05, "loss": 0.1312, "step": 11030 }, { "epoch": 0.71, "learning_rate": 4.166926648597853e-05, "loss": 0.1453, "step": 11031 }, { "epoch": 0.71, "learning_rate": 4.165239992626345e-05, "loss": 0.1379, "step": 11032 }, { "epoch": 0.71, "learning_rate": 4.1635535882832646e-05, "loss": 0.1367, "step": 11033 }, { "epoch": 0.71, "learning_rate": 4.161867435641342e-05, "loss": 0.1453, "step": 11034 }, { "epoch": 0.71, "learning_rate": 4.1601815347732964e-05, "loss": 0.1473, "step": 11035 }, { "epoch": 0.71, "learning_rate": 4.158495885751834e-05, "loss": 0.1294, "step": 11036 }, { "epoch": 0.71, "learning_rate": 4.1568104886496473e-05, "loss": 0.1343, "step": 11037 }, { "epoch": 0.71, "learning_rate": 4.155125343539417e-05, "loss": 0.1478, "step": 11038 }, { "epoch": 0.71, "learning_rate": 4.153440450493823e-05, "loss": 0.1278, "step": 11039 }, { "epoch": 0.71, "learning_rate": 4.151755809585517e-05, "loss": 0.1526, "step": 11040 }, { "epoch": 0.71, "learning_rate": 4.150071420887165e-05, "loss": 0.1289, "step": 11041 }, { "epoch": 0.71, "learning_rate": 4.148387284471397e-05, "loss": 0.1584, "step": 11042 }, { "epoch": 0.71, "learning_rate": 4.1467034004108483e-05, "loss": 0.1238, "step": 11043 }, { "epoch": 0.71, "learning_rate": 4.1450197687781365e-05, "loss": 0.1187, "step": 11044 }, { "epoch": 0.71, "learning_rate": 4.1433363896458635e-05, "loss": 0.1528, "step": 11045 }, { "epoch": 0.71, "learning_rate": 4.141653263086631e-05, "loss": 0.169, "step": 11046 }, { "epoch": 0.71, "learning_rate": 4.139970389173026e-05, "loss": 0.1394, "step": 11047 }, { "epoch": 0.71, "learning_rate": 4.138287767977625e-05, "loss": 0.1449, "step": 11048 }, { "epoch": 0.71, "learning_rate": 4.136605399572988e-05, "loss": 0.1331, "step": 11049 }, { "epoch": 0.71, "learning_rate": 4.134923284031673e-05, "loss": 0.1664, "step": 11050 }, { "epoch": 0.71, "learning_rate": 4.1332414214262204e-05, "loss": 0.1419, "step": 11051 }, { "epoch": 0.71, "learning_rate": 4.1315598118291576e-05, "loss": 0.136, "step": 11052 }, { "epoch": 0.71, "learning_rate": 4.12987845531301e-05, "loss": 0.1346, "step": 11053 }, { "epoch": 0.71, "learning_rate": 4.128197351950286e-05, "loss": 0.1374, "step": 11054 }, { "epoch": 0.71, "learning_rate": 4.126516501813488e-05, "loss": 0.144, "step": 11055 }, { "epoch": 0.71, "learning_rate": 4.124835904975097e-05, "loss": 0.1321, "step": 11056 }, { "epoch": 0.71, "learning_rate": 4.123155561507598e-05, "loss": 0.1238, "step": 11057 }, { "epoch": 0.71, "learning_rate": 4.121475471483453e-05, "loss": 0.1669, "step": 11058 }, { "epoch": 0.71, "learning_rate": 4.119795634975113e-05, "loss": 0.1445, "step": 11059 }, { "epoch": 0.71, "learning_rate": 4.1181160520550266e-05, "loss": 0.1498, "step": 11060 }, { "epoch": 0.71, "learning_rate": 4.116436722795627e-05, "loss": 0.1526, "step": 11061 }, { "epoch": 0.71, "learning_rate": 4.11475764726934e-05, "loss": 0.1419, "step": 11062 }, { "epoch": 0.71, "learning_rate": 4.1130788255485686e-05, "loss": 0.1278, "step": 11063 }, { "epoch": 0.71, "learning_rate": 4.111400257705722e-05, "loss": 0.143, "step": 11064 }, { "epoch": 0.71, "learning_rate": 4.109721943813186e-05, "loss": 0.1348, "step": 11065 }, { "epoch": 0.71, "learning_rate": 4.108043883943335e-05, "loss": 0.1352, "step": 11066 }, { "epoch": 0.71, "learning_rate": 4.106366078168541e-05, "loss": 0.1459, "step": 11067 }, { "epoch": 0.71, "learning_rate": 4.10468852656116e-05, "loss": 0.1583, "step": 11068 }, { "epoch": 0.71, "learning_rate": 4.1030112291935406e-05, "loss": 0.1226, "step": 11069 }, { "epoch": 0.71, "learning_rate": 4.101334186138015e-05, "loss": 0.1378, "step": 11070 }, { "epoch": 0.71, "learning_rate": 4.099657397466903e-05, "loss": 0.1469, "step": 11071 }, { "epoch": 0.71, "learning_rate": 4.0979808632525254e-05, "loss": 0.1427, "step": 11072 }, { "epoch": 0.71, "learning_rate": 4.0963045835671746e-05, "loss": 0.1354, "step": 11073 }, { "epoch": 0.71, "learning_rate": 4.094628558483147e-05, "loss": 0.143, "step": 11074 }, { "epoch": 0.71, "learning_rate": 4.092952788072722e-05, "loss": 0.141, "step": 11075 }, { "epoch": 0.71, "learning_rate": 4.091277272408171e-05, "loss": 0.1491, "step": 11076 }, { "epoch": 0.71, "learning_rate": 4.089602011561751e-05, "loss": 0.1391, "step": 11077 }, { "epoch": 0.71, "learning_rate": 4.087927005605703e-05, "loss": 0.1431, "step": 11078 }, { "epoch": 0.71, "learning_rate": 4.0862522546122694e-05, "loss": 0.154, "step": 11079 }, { "epoch": 0.71, "learning_rate": 4.0845777586536704e-05, "loss": 0.1493, "step": 11080 }, { "epoch": 0.71, "learning_rate": 4.082903517802121e-05, "loss": 0.1243, "step": 11081 }, { "epoch": 0.71, "learning_rate": 4.081229532129827e-05, "loss": 0.1401, "step": 11082 }, { "epoch": 0.71, "learning_rate": 4.079555801708981e-05, "loss": 0.1456, "step": 11083 }, { "epoch": 0.71, "learning_rate": 4.0778823266117606e-05, "loss": 0.1433, "step": 11084 }, { "epoch": 0.71, "learning_rate": 4.076209106910335e-05, "loss": 0.1292, "step": 11085 }, { "epoch": 0.71, "learning_rate": 4.0745361426768624e-05, "loss": 0.1276, "step": 11086 }, { "epoch": 0.71, "learning_rate": 4.0728634339834946e-05, "loss": 0.1434, "step": 11087 }, { "epoch": 0.71, "learning_rate": 4.0711909809023696e-05, "loss": 0.1263, "step": 11088 }, { "epoch": 0.71, "learning_rate": 4.069518783505607e-05, "loss": 0.1427, "step": 11089 }, { "epoch": 0.71, "learning_rate": 4.0678468418653284e-05, "loss": 0.1326, "step": 11090 }, { "epoch": 0.71, "learning_rate": 4.0661751560536335e-05, "loss": 0.1452, "step": 11091 }, { "epoch": 0.71, "learning_rate": 4.0645037261426124e-05, "loss": 0.1255, "step": 11092 }, { "epoch": 0.71, "learning_rate": 4.0628325522043495e-05, "loss": 0.1237, "step": 11093 }, { "epoch": 0.71, "learning_rate": 4.0611616343109174e-05, "loss": 0.1386, "step": 11094 }, { "epoch": 0.71, "learning_rate": 4.0594909725343753e-05, "loss": 0.1416, "step": 11095 }, { "epoch": 0.71, "learning_rate": 4.0578205669467687e-05, "loss": 0.1403, "step": 11096 }, { "epoch": 0.71, "learning_rate": 4.0561504176201405e-05, "loss": 0.1278, "step": 11097 }, { "epoch": 0.71, "learning_rate": 4.0544805246265125e-05, "loss": 0.1338, "step": 11098 }, { "epoch": 0.71, "learning_rate": 4.052810888037899e-05, "loss": 0.1298, "step": 11099 }, { "epoch": 0.71, "learning_rate": 4.051141507926306e-05, "loss": 0.1418, "step": 11100 }, { "epoch": 0.71, "learning_rate": 4.0494723843637273e-05, "loss": 0.122, "step": 11101 }, { "epoch": 0.71, "learning_rate": 4.047803517422149e-05, "loss": 0.1337, "step": 11102 }, { "epoch": 0.71, "learning_rate": 4.0461349071735335e-05, "loss": 0.1332, "step": 11103 }, { "epoch": 0.71, "learning_rate": 4.044466553689851e-05, "loss": 0.1367, "step": 11104 }, { "epoch": 0.71, "learning_rate": 4.0427984570430445e-05, "loss": 0.1514, "step": 11105 }, { "epoch": 0.71, "learning_rate": 4.041130617305049e-05, "loss": 0.1369, "step": 11106 }, { "epoch": 0.71, "learning_rate": 4.0394630345477956e-05, "loss": 0.1651, "step": 11107 }, { "epoch": 0.71, "learning_rate": 4.0377957088431996e-05, "loss": 0.1353, "step": 11108 }, { "epoch": 0.71, "learning_rate": 4.0361286402631695e-05, "loss": 0.1429, "step": 11109 }, { "epoch": 0.71, "learning_rate": 4.034461828879591e-05, "loss": 0.135, "step": 11110 }, { "epoch": 0.71, "learning_rate": 4.032795274764354e-05, "loss": 0.1413, "step": 11111 }, { "epoch": 0.71, "learning_rate": 4.031128977989326e-05, "loss": 0.1352, "step": 11112 }, { "epoch": 0.71, "learning_rate": 4.029462938626365e-05, "loss": 0.1342, "step": 11113 }, { "epoch": 0.71, "learning_rate": 4.027797156747323e-05, "loss": 0.1375, "step": 11114 }, { "epoch": 0.71, "learning_rate": 4.026131632424038e-05, "loss": 0.1348, "step": 11115 }, { "epoch": 0.71, "learning_rate": 4.024466365728341e-05, "loss": 0.1694, "step": 11116 }, { "epoch": 0.71, "learning_rate": 4.02280135673204e-05, "loss": 0.144, "step": 11117 }, { "epoch": 0.71, "learning_rate": 4.0211366055069474e-05, "loss": 0.1219, "step": 11118 }, { "epoch": 0.71, "learning_rate": 4.019472112124853e-05, "loss": 0.1447, "step": 11119 }, { "epoch": 0.71, "learning_rate": 4.017807876657537e-05, "loss": 0.1444, "step": 11120 }, { "epoch": 0.71, "learning_rate": 4.016143899176772e-05, "loss": 0.1456, "step": 11121 }, { "epoch": 0.71, "learning_rate": 4.014480179754321e-05, "loss": 0.1453, "step": 11122 }, { "epoch": 0.71, "learning_rate": 4.0128167184619345e-05, "loss": 0.1235, "step": 11123 }, { "epoch": 0.71, "learning_rate": 4.011153515371344e-05, "loss": 0.1403, "step": 11124 }, { "epoch": 0.71, "learning_rate": 4.009490570554285e-05, "loss": 0.161, "step": 11125 }, { "epoch": 0.71, "learning_rate": 4.007827884082468e-05, "loss": 0.1332, "step": 11126 }, { "epoch": 0.71, "learning_rate": 4.0061654560275896e-05, "loss": 0.1486, "step": 11127 }, { "epoch": 0.71, "learning_rate": 4.0045032864613605e-05, "loss": 0.1265, "step": 11128 }, { "epoch": 0.71, "learning_rate": 4.002841375455451e-05, "loss": 0.1525, "step": 11129 }, { "epoch": 0.71, "learning_rate": 4.0011797230815374e-05, "loss": 0.1292, "step": 11130 }, { "epoch": 0.71, "learning_rate": 3.999518329411275e-05, "loss": 0.1532, "step": 11131 }, { "epoch": 0.71, "learning_rate": 3.997857194516319e-05, "loss": 0.1344, "step": 11132 }, { "epoch": 0.71, "learning_rate": 3.9961963184683005e-05, "loss": 0.1408, "step": 11133 }, { "epoch": 0.71, "learning_rate": 3.9945357013388495e-05, "loss": 0.1381, "step": 11134 }, { "epoch": 0.71, "learning_rate": 3.992875343199584e-05, "loss": 0.1361, "step": 11135 }, { "epoch": 0.71, "learning_rate": 3.9912152441221017e-05, "loss": 0.1333, "step": 11136 }, { "epoch": 0.71, "learning_rate": 3.989555404178003e-05, "loss": 0.1375, "step": 11137 }, { "epoch": 0.71, "learning_rate": 3.987895823438862e-05, "loss": 0.1474, "step": 11138 }, { "epoch": 0.71, "learning_rate": 3.986236501976256e-05, "loss": 0.1402, "step": 11139 }, { "epoch": 0.71, "learning_rate": 3.9845774398617384e-05, "loss": 0.1364, "step": 11140 }, { "epoch": 0.71, "learning_rate": 3.982918637166863e-05, "loss": 0.1454, "step": 11141 }, { "epoch": 0.71, "learning_rate": 3.981260093963166e-05, "loss": 0.1348, "step": 11142 }, { "epoch": 0.71, "learning_rate": 3.979601810322169e-05, "loss": 0.1309, "step": 11143 }, { "epoch": 0.71, "learning_rate": 3.977943786315394e-05, "loss": 0.139, "step": 11144 }, { "epoch": 0.71, "learning_rate": 3.976286022014339e-05, "loss": 0.137, "step": 11145 }, { "epoch": 0.71, "learning_rate": 3.974628517490496e-05, "loss": 0.1384, "step": 11146 }, { "epoch": 0.71, "learning_rate": 3.9729712728153465e-05, "loss": 0.1359, "step": 11147 }, { "epoch": 0.71, "learning_rate": 3.971314288060363e-05, "loss": 0.1342, "step": 11148 }, { "epoch": 0.71, "learning_rate": 3.9696575632970056e-05, "loss": 0.1489, "step": 11149 }, { "epoch": 0.71, "learning_rate": 3.968001098596717e-05, "loss": 0.1576, "step": 11150 }, { "epoch": 0.71, "learning_rate": 3.966344894030937e-05, "loss": 0.1224, "step": 11151 }, { "epoch": 0.72, "learning_rate": 3.9646889496710906e-05, "loss": 0.1355, "step": 11152 }, { "epoch": 0.72, "learning_rate": 3.963033265588587e-05, "loss": 0.1501, "step": 11153 }, { "epoch": 0.72, "learning_rate": 3.961377841854832e-05, "loss": 0.135, "step": 11154 }, { "epoch": 0.72, "learning_rate": 3.9597226785412166e-05, "loss": 0.1186, "step": 11155 }, { "epoch": 0.72, "learning_rate": 3.958067775719126e-05, "loss": 0.1511, "step": 11156 }, { "epoch": 0.72, "learning_rate": 3.956413133459921e-05, "loss": 0.1516, "step": 11157 }, { "epoch": 0.72, "learning_rate": 3.954758751834966e-05, "loss": 0.1293, "step": 11158 }, { "epoch": 0.72, "learning_rate": 3.953104630915605e-05, "loss": 0.1299, "step": 11159 }, { "epoch": 0.72, "learning_rate": 3.95145077077317e-05, "loss": 0.1484, "step": 11160 }, { "epoch": 0.72, "learning_rate": 3.949797171478986e-05, "loss": 0.1221, "step": 11161 }, { "epoch": 0.72, "learning_rate": 3.9481438331043686e-05, "loss": 0.161, "step": 11162 }, { "epoch": 0.72, "learning_rate": 3.946490755720621e-05, "loss": 0.139, "step": 11163 }, { "epoch": 0.72, "learning_rate": 3.9448379393990276e-05, "loss": 0.1244, "step": 11164 }, { "epoch": 0.72, "learning_rate": 3.9431853842108734e-05, "loss": 0.126, "step": 11165 }, { "epoch": 0.72, "learning_rate": 3.941533090227423e-05, "loss": 0.1486, "step": 11166 }, { "epoch": 0.72, "learning_rate": 3.9398810575199285e-05, "loss": 0.1503, "step": 11167 }, { "epoch": 0.72, "learning_rate": 3.9382292861596393e-05, "loss": 0.1495, "step": 11168 }, { "epoch": 0.72, "learning_rate": 3.9365777762177904e-05, "loss": 0.1366, "step": 11169 }, { "epoch": 0.72, "learning_rate": 3.9349265277656055e-05, "loss": 0.146, "step": 11170 }, { "epoch": 0.72, "learning_rate": 3.933275540874292e-05, "loss": 0.1375, "step": 11171 }, { "epoch": 0.72, "learning_rate": 3.931624815615054e-05, "loss": 0.1363, "step": 11172 }, { "epoch": 0.72, "learning_rate": 3.9299743520590735e-05, "loss": 0.1309, "step": 11173 }, { "epoch": 0.72, "learning_rate": 3.9283241502775334e-05, "loss": 0.1328, "step": 11174 }, { "epoch": 0.72, "learning_rate": 3.926674210341603e-05, "loss": 0.142, "step": 11175 }, { "epoch": 0.72, "learning_rate": 3.9250245323224285e-05, "loss": 0.1629, "step": 11176 }, { "epoch": 0.72, "learning_rate": 3.9233751162911625e-05, "loss": 0.1412, "step": 11177 }, { "epoch": 0.72, "learning_rate": 3.9217259623189294e-05, "loss": 0.1294, "step": 11178 }, { "epoch": 0.72, "learning_rate": 3.920077070476858e-05, "loss": 0.1654, "step": 11179 }, { "epoch": 0.72, "learning_rate": 3.9184284408360486e-05, "loss": 0.1452, "step": 11180 }, { "epoch": 0.72, "learning_rate": 3.9167800734676054e-05, "loss": 0.146, "step": 11181 }, { "epoch": 0.72, "learning_rate": 3.9151319684426194e-05, "loss": 0.1438, "step": 11182 }, { "epoch": 0.72, "learning_rate": 3.9134841258321565e-05, "loss": 0.1303, "step": 11183 }, { "epoch": 0.72, "learning_rate": 3.9118365457072916e-05, "loss": 0.152, "step": 11184 }, { "epoch": 0.72, "learning_rate": 3.9101892281390685e-05, "loss": 0.1567, "step": 11185 }, { "epoch": 0.72, "learning_rate": 3.908542173198537e-05, "loss": 0.1588, "step": 11186 }, { "epoch": 0.72, "learning_rate": 3.90689538095672e-05, "loss": 0.1494, "step": 11187 }, { "epoch": 0.72, "learning_rate": 3.9052488514846405e-05, "loss": 0.1525, "step": 11188 }, { "epoch": 0.72, "learning_rate": 3.903602584853311e-05, "loss": 0.1556, "step": 11189 }, { "epoch": 0.72, "learning_rate": 3.9019565811337176e-05, "loss": 0.1504, "step": 11190 }, { "epoch": 0.72, "learning_rate": 3.900310840396857e-05, "loss": 0.1435, "step": 11191 }, { "epoch": 0.72, "learning_rate": 3.898665362713693e-05, "loss": 0.1557, "step": 11192 }, { "epoch": 0.72, "learning_rate": 3.897020148155195e-05, "loss": 0.1426, "step": 11193 }, { "epoch": 0.72, "learning_rate": 3.895375196792308e-05, "loss": 0.1316, "step": 11194 }, { "epoch": 0.72, "learning_rate": 3.8937305086959754e-05, "loss": 0.1415, "step": 11195 }, { "epoch": 0.72, "learning_rate": 3.892086083937129e-05, "loss": 0.1404, "step": 11196 }, { "epoch": 0.72, "learning_rate": 3.890441922586678e-05, "loss": 0.1321, "step": 11197 }, { "epoch": 0.72, "learning_rate": 3.888798024715537e-05, "loss": 0.1316, "step": 11198 }, { "epoch": 0.72, "learning_rate": 3.8871543903945904e-05, "loss": 0.1342, "step": 11199 }, { "epoch": 0.72, "learning_rate": 3.885511019694732e-05, "loss": 0.1452, "step": 11200 }, { "epoch": 0.72, "learning_rate": 3.883867912686823e-05, "loss": 0.1302, "step": 11201 }, { "epoch": 0.72, "learning_rate": 3.882225069441728e-05, "loss": 0.1381, "step": 11202 }, { "epoch": 0.72, "learning_rate": 3.880582490030301e-05, "loss": 0.1332, "step": 11203 }, { "epoch": 0.72, "learning_rate": 3.878940174523371e-05, "loss": 0.1378, "step": 11204 }, { "epoch": 0.72, "learning_rate": 3.877298122991771e-05, "loss": 0.1294, "step": 11205 }, { "epoch": 0.72, "learning_rate": 3.87565633550631e-05, "loss": 0.1755, "step": 11206 }, { "epoch": 0.72, "learning_rate": 3.8740148121377975e-05, "loss": 0.1603, "step": 11207 }, { "epoch": 0.72, "learning_rate": 3.872373552957017e-05, "loss": 0.1454, "step": 11208 }, { "epoch": 0.72, "learning_rate": 3.870732558034754e-05, "loss": 0.138, "step": 11209 }, { "epoch": 0.72, "learning_rate": 3.869091827441782e-05, "loss": 0.1525, "step": 11210 }, { "epoch": 0.72, "learning_rate": 3.867451361248851e-05, "loss": 0.1353, "step": 11211 }, { "epoch": 0.72, "learning_rate": 3.8658111595267134e-05, "loss": 0.1655, "step": 11212 }, { "epoch": 0.72, "learning_rate": 3.864171222346098e-05, "loss": 0.142, "step": 11213 }, { "epoch": 0.72, "learning_rate": 3.8625315497777314e-05, "loss": 0.1592, "step": 11214 }, { "epoch": 0.72, "learning_rate": 3.860892141892329e-05, "loss": 0.1478, "step": 11215 }, { "epoch": 0.72, "learning_rate": 3.8592529987605864e-05, "loss": 0.1441, "step": 11216 }, { "epoch": 0.72, "learning_rate": 3.857614120453198e-05, "loss": 0.1324, "step": 11217 }, { "epoch": 0.72, "learning_rate": 3.855975507040835e-05, "loss": 0.1338, "step": 11218 }, { "epoch": 0.72, "learning_rate": 3.854337158594171e-05, "loss": 0.159, "step": 11219 }, { "epoch": 0.72, "learning_rate": 3.8526990751838545e-05, "loss": 0.139, "step": 11220 }, { "epoch": 0.72, "learning_rate": 3.8510612568805314e-05, "loss": 0.1266, "step": 11221 }, { "epoch": 0.72, "learning_rate": 3.8494237037548384e-05, "loss": 0.1431, "step": 11222 }, { "epoch": 0.72, "learning_rate": 3.847786415877389e-05, "loss": 0.1332, "step": 11223 }, { "epoch": 0.72, "learning_rate": 3.8461493933188e-05, "loss": 0.1489, "step": 11224 }, { "epoch": 0.72, "learning_rate": 3.84451263614966e-05, "loss": 0.1424, "step": 11225 }, { "epoch": 0.72, "learning_rate": 3.842876144440565e-05, "loss": 0.1486, "step": 11226 }, { "epoch": 0.72, "learning_rate": 3.841239918262082e-05, "loss": 0.1229, "step": 11227 }, { "epoch": 0.72, "learning_rate": 3.839603957684778e-05, "loss": 0.1404, "step": 11228 }, { "epoch": 0.72, "learning_rate": 3.837968262779207e-05, "loss": 0.1315, "step": 11229 }, { "epoch": 0.72, "learning_rate": 3.8363328336159054e-05, "loss": 0.1395, "step": 11230 }, { "epoch": 0.72, "learning_rate": 3.834697670265407e-05, "loss": 0.1315, "step": 11231 }, { "epoch": 0.72, "learning_rate": 3.833062772798224e-05, "loss": 0.1448, "step": 11232 }, { "epoch": 0.72, "learning_rate": 3.83142814128487e-05, "loss": 0.1653, "step": 11233 }, { "epoch": 0.72, "learning_rate": 3.8297937757958304e-05, "loss": 0.1333, "step": 11234 }, { "epoch": 0.72, "learning_rate": 3.8281596764015924e-05, "loss": 0.1376, "step": 11235 }, { "epoch": 0.72, "learning_rate": 3.8265258431726337e-05, "loss": 0.1318, "step": 11236 }, { "epoch": 0.72, "learning_rate": 3.824892276179406e-05, "loss": 0.1465, "step": 11237 }, { "epoch": 0.72, "learning_rate": 3.8232589754923644e-05, "loss": 0.1274, "step": 11238 }, { "epoch": 0.72, "learning_rate": 3.82162594118194e-05, "loss": 0.1472, "step": 11239 }, { "epoch": 0.72, "learning_rate": 3.819993173318566e-05, "loss": 0.1374, "step": 11240 }, { "epoch": 0.72, "learning_rate": 3.8183606719726485e-05, "loss": 0.1325, "step": 11241 }, { "epoch": 0.72, "learning_rate": 3.816728437214595e-05, "loss": 0.1425, "step": 11242 }, { "epoch": 0.72, "learning_rate": 3.815096469114801e-05, "loss": 0.1597, "step": 11243 }, { "epoch": 0.72, "learning_rate": 3.8134647677436366e-05, "loss": 0.1196, "step": 11244 }, { "epoch": 0.72, "learning_rate": 3.8118333331714807e-05, "loss": 0.1263, "step": 11245 }, { "epoch": 0.72, "learning_rate": 3.81020216546868e-05, "loss": 0.1226, "step": 11246 }, { "epoch": 0.72, "learning_rate": 3.8085712647055895e-05, "loss": 0.1329, "step": 11247 }, { "epoch": 0.72, "learning_rate": 3.8069406309525336e-05, "loss": 0.1442, "step": 11248 }, { "epoch": 0.72, "learning_rate": 3.80531026427984e-05, "loss": 0.1349, "step": 11249 }, { "epoch": 0.72, "learning_rate": 3.8036801647578214e-05, "loss": 0.1533, "step": 11250 }, { "epoch": 0.72, "learning_rate": 3.802050332456771e-05, "loss": 0.1722, "step": 11251 }, { "epoch": 0.72, "learning_rate": 3.800420767446985e-05, "loss": 0.1563, "step": 11252 }, { "epoch": 0.72, "learning_rate": 3.79879146979873e-05, "loss": 0.139, "step": 11253 }, { "epoch": 0.72, "learning_rate": 3.797162439582279e-05, "loss": 0.141, "step": 11254 }, { "epoch": 0.72, "learning_rate": 3.795533676867877e-05, "loss": 0.1484, "step": 11255 }, { "epoch": 0.72, "learning_rate": 3.793905181725772e-05, "loss": 0.1295, "step": 11256 }, { "epoch": 0.72, "learning_rate": 3.792276954226194e-05, "loss": 0.1339, "step": 11257 }, { "epoch": 0.72, "learning_rate": 3.7906489944393574e-05, "loss": 0.1433, "step": 11258 }, { "epoch": 0.72, "learning_rate": 3.789021302435474e-05, "loss": 0.1347, "step": 11259 }, { "epoch": 0.72, "learning_rate": 3.787393878284734e-05, "loss": 0.1344, "step": 11260 }, { "epoch": 0.72, "learning_rate": 3.785766722057323e-05, "loss": 0.1399, "step": 11261 }, { "epoch": 0.72, "learning_rate": 3.784139833823419e-05, "loss": 0.1353, "step": 11262 }, { "epoch": 0.72, "learning_rate": 3.7825132136531735e-05, "loss": 0.1339, "step": 11263 }, { "epoch": 0.72, "learning_rate": 3.7808868616167445e-05, "loss": 0.1529, "step": 11264 }, { "epoch": 0.72, "learning_rate": 3.779260777784263e-05, "loss": 0.1425, "step": 11265 }, { "epoch": 0.72, "learning_rate": 3.77763496222586e-05, "loss": 0.1471, "step": 11266 }, { "epoch": 0.72, "learning_rate": 3.7760094150116446e-05, "loss": 0.1498, "step": 11267 }, { "epoch": 0.72, "learning_rate": 3.7743841362117225e-05, "loss": 0.1487, "step": 11268 }, { "epoch": 0.72, "learning_rate": 3.772759125896189e-05, "loss": 0.141, "step": 11269 }, { "epoch": 0.72, "learning_rate": 3.771134384135116e-05, "loss": 0.1367, "step": 11270 }, { "epoch": 0.72, "learning_rate": 3.769509910998581e-05, "loss": 0.1237, "step": 11271 }, { "epoch": 0.72, "learning_rate": 3.767885706556632e-05, "loss": 0.153, "step": 11272 }, { "epoch": 0.72, "learning_rate": 3.76626177087932e-05, "loss": 0.1299, "step": 11273 }, { "epoch": 0.72, "learning_rate": 3.764638104036673e-05, "loss": 0.133, "step": 11274 }, { "epoch": 0.72, "learning_rate": 3.7630147060987164e-05, "loss": 0.1374, "step": 11275 }, { "epoch": 0.72, "learning_rate": 3.7613915771354636e-05, "loss": 0.1283, "step": 11276 }, { "epoch": 0.72, "learning_rate": 3.759768717216906e-05, "loss": 0.1521, "step": 11277 }, { "epoch": 0.72, "learning_rate": 3.758146126413038e-05, "loss": 0.1366, "step": 11278 }, { "epoch": 0.72, "learning_rate": 3.756523804793828e-05, "loss": 0.1316, "step": 11279 }, { "epoch": 0.72, "learning_rate": 3.7549017524292465e-05, "loss": 0.1232, "step": 11280 }, { "epoch": 0.72, "learning_rate": 3.7532799693892384e-05, "loss": 0.1334, "step": 11281 }, { "epoch": 0.72, "learning_rate": 3.751658455743749e-05, "loss": 0.1393, "step": 11282 }, { "epoch": 0.72, "learning_rate": 3.75003721156271e-05, "loss": 0.1518, "step": 11283 }, { "epoch": 0.72, "learning_rate": 3.748416236916032e-05, "loss": 0.1363, "step": 11284 }, { "epoch": 0.72, "learning_rate": 3.746795531873628e-05, "loss": 0.1481, "step": 11285 }, { "epoch": 0.72, "learning_rate": 3.745175096505386e-05, "loss": 0.1378, "step": 11286 }, { "epoch": 0.72, "learning_rate": 3.743554930881192e-05, "loss": 0.1407, "step": 11287 }, { "epoch": 0.72, "learning_rate": 3.741935035070914e-05, "loss": 0.1301, "step": 11288 }, { "epoch": 0.72, "learning_rate": 3.740315409144413e-05, "loss": 0.1288, "step": 11289 }, { "epoch": 0.72, "learning_rate": 3.738696053171539e-05, "loss": 0.15, "step": 11290 }, { "epoch": 0.72, "learning_rate": 3.737076967222122e-05, "loss": 0.1623, "step": 11291 }, { "epoch": 0.72, "learning_rate": 3.7354581513659933e-05, "loss": 0.1388, "step": 11292 }, { "epoch": 0.72, "learning_rate": 3.7338396056729595e-05, "loss": 0.1493, "step": 11293 }, { "epoch": 0.72, "learning_rate": 3.732221330212827e-05, "loss": 0.1366, "step": 11294 }, { "epoch": 0.72, "learning_rate": 3.73060332505538e-05, "loss": 0.1395, "step": 11295 }, { "epoch": 0.72, "learning_rate": 3.728985590270397e-05, "loss": 0.127, "step": 11296 }, { "epoch": 0.72, "learning_rate": 3.72736812592765e-05, "loss": 0.1341, "step": 11297 }, { "epoch": 0.72, "learning_rate": 3.7257509320968856e-05, "loss": 0.1544, "step": 11298 }, { "epoch": 0.72, "learning_rate": 3.7241340088478515e-05, "loss": 0.1325, "step": 11299 }, { "epoch": 0.72, "learning_rate": 3.722517356250276e-05, "loss": 0.1546, "step": 11300 }, { "epoch": 0.72, "learning_rate": 3.720900974373878e-05, "loss": 0.1485, "step": 11301 }, { "epoch": 0.72, "learning_rate": 3.7192848632883705e-05, "loss": 0.1522, "step": 11302 }, { "epoch": 0.72, "learning_rate": 3.7176690230634426e-05, "loss": 0.1381, "step": 11303 }, { "epoch": 0.72, "learning_rate": 3.716053453768785e-05, "loss": 0.149, "step": 11304 }, { "epoch": 0.72, "learning_rate": 3.714438155474064e-05, "loss": 0.1427, "step": 11305 }, { "epoch": 0.72, "learning_rate": 3.7128231282489465e-05, "loss": 0.1469, "step": 11306 }, { "epoch": 0.72, "learning_rate": 3.711208372163075e-05, "loss": 0.1329, "step": 11307 }, { "epoch": 0.73, "learning_rate": 3.709593887286092e-05, "loss": 0.1206, "step": 11308 }, { "epoch": 0.73, "learning_rate": 3.707979673687626e-05, "loss": 0.1505, "step": 11309 }, { "epoch": 0.73, "learning_rate": 3.7063657314372834e-05, "loss": 0.1637, "step": 11310 }, { "epoch": 0.73, "learning_rate": 3.704752060604675e-05, "loss": 0.1413, "step": 11311 }, { "epoch": 0.73, "learning_rate": 3.703138661259383e-05, "loss": 0.1481, "step": 11312 }, { "epoch": 0.73, "learning_rate": 3.7015255334709956e-05, "loss": 0.1532, "step": 11313 }, { "epoch": 0.73, "learning_rate": 3.699912677309072e-05, "loss": 0.1365, "step": 11314 }, { "epoch": 0.73, "learning_rate": 3.69830009284317e-05, "loss": 0.1343, "step": 11315 }, { "epoch": 0.73, "learning_rate": 3.696687780142839e-05, "loss": 0.1292, "step": 11316 }, { "epoch": 0.73, "learning_rate": 3.6950757392776036e-05, "loss": 0.1332, "step": 11317 }, { "epoch": 0.73, "learning_rate": 3.693463970316991e-05, "loss": 0.1366, "step": 11318 }, { "epoch": 0.73, "learning_rate": 3.6918524733305026e-05, "loss": 0.1376, "step": 11319 }, { "epoch": 0.73, "learning_rate": 3.6902412483876435e-05, "loss": 0.1394, "step": 11320 }, { "epoch": 0.73, "learning_rate": 3.6886302955578923e-05, "loss": 0.1261, "step": 11321 }, { "epoch": 0.73, "learning_rate": 3.687019614910724e-05, "loss": 0.147, "step": 11322 }, { "epoch": 0.73, "learning_rate": 3.685409206515605e-05, "loss": 0.1142, "step": 11323 }, { "epoch": 0.73, "learning_rate": 3.68379907044198e-05, "loss": 0.1449, "step": 11324 }, { "epoch": 0.73, "learning_rate": 3.682189206759291e-05, "loss": 0.1444, "step": 11325 }, { "epoch": 0.73, "learning_rate": 3.680579615536961e-05, "loss": 0.1488, "step": 11326 }, { "epoch": 0.73, "learning_rate": 3.67897029684441e-05, "loss": 0.1347, "step": 11327 }, { "epoch": 0.73, "learning_rate": 3.677361250751034e-05, "loss": 0.1537, "step": 11328 }, { "epoch": 0.73, "learning_rate": 3.6757524773262274e-05, "loss": 0.1233, "step": 11329 }, { "epoch": 0.73, "learning_rate": 3.674143976639375e-05, "loss": 0.1213, "step": 11330 }, { "epoch": 0.73, "learning_rate": 3.6725357487598367e-05, "loss": 0.1406, "step": 11331 }, { "epoch": 0.73, "learning_rate": 3.670927793756974e-05, "loss": 0.1433, "step": 11332 }, { "epoch": 0.73, "learning_rate": 3.669320111700128e-05, "loss": 0.1458, "step": 11333 }, { "epoch": 0.73, "learning_rate": 3.667712702658635e-05, "loss": 0.1313, "step": 11334 }, { "epoch": 0.73, "learning_rate": 3.666105566701809e-05, "loss": 0.1318, "step": 11335 }, { "epoch": 0.73, "learning_rate": 3.664498703898963e-05, "loss": 0.1329, "step": 11336 }, { "epoch": 0.73, "learning_rate": 3.6628921143193984e-05, "loss": 0.1452, "step": 11337 }, { "epoch": 0.73, "learning_rate": 3.661285798032392e-05, "loss": 0.1399, "step": 11338 }, { "epoch": 0.73, "learning_rate": 3.6596797551072256e-05, "loss": 0.1391, "step": 11339 }, { "epoch": 0.73, "learning_rate": 3.658073985613155e-05, "loss": 0.1506, "step": 11340 }, { "epoch": 0.73, "learning_rate": 3.656468489619435e-05, "loss": 0.1519, "step": 11341 }, { "epoch": 0.73, "learning_rate": 3.654863267195294e-05, "loss": 0.1382, "step": 11342 }, { "epoch": 0.73, "learning_rate": 3.6532583184099735e-05, "loss": 0.1365, "step": 11343 }, { "epoch": 0.73, "learning_rate": 3.6516536433326796e-05, "loss": 0.1588, "step": 11344 }, { "epoch": 0.73, "learning_rate": 3.6500492420326126e-05, "loss": 0.1506, "step": 11345 }, { "epoch": 0.73, "learning_rate": 3.64844511457897e-05, "loss": 0.1267, "step": 11346 }, { "epoch": 0.73, "learning_rate": 3.646841261040925e-05, "loss": 0.1395, "step": 11347 }, { "epoch": 0.73, "learning_rate": 3.645237681487648e-05, "loss": 0.1308, "step": 11348 }, { "epoch": 0.73, "learning_rate": 3.6436343759882926e-05, "loss": 0.1354, "step": 11349 }, { "epoch": 0.73, "learning_rate": 3.642031344612009e-05, "loss": 0.1507, "step": 11350 }, { "epoch": 0.73, "learning_rate": 3.640428587427925e-05, "loss": 0.1258, "step": 11351 }, { "epoch": 0.73, "learning_rate": 3.638826104505157e-05, "loss": 0.1363, "step": 11352 }, { "epoch": 0.73, "learning_rate": 3.637223895912818e-05, "loss": 0.1475, "step": 11353 }, { "epoch": 0.73, "learning_rate": 3.6356219617200004e-05, "loss": 0.1378, "step": 11354 }, { "epoch": 0.73, "learning_rate": 3.634020301995792e-05, "loss": 0.1458, "step": 11355 }, { "epoch": 0.73, "learning_rate": 3.6324189168092635e-05, "loss": 0.153, "step": 11356 }, { "epoch": 0.73, "learning_rate": 3.630817806229482e-05, "loss": 0.1149, "step": 11357 }, { "epoch": 0.73, "learning_rate": 3.629216970325492e-05, "loss": 0.135, "step": 11358 }, { "epoch": 0.73, "learning_rate": 3.6276164091663254e-05, "loss": 0.1381, "step": 11359 }, { "epoch": 0.73, "learning_rate": 3.6260161228210174e-05, "loss": 0.1449, "step": 11360 }, { "epoch": 0.73, "learning_rate": 3.624416111358574e-05, "loss": 0.145, "step": 11361 }, { "epoch": 0.73, "learning_rate": 3.6228163748480005e-05, "loss": 0.143, "step": 11362 }, { "epoch": 0.73, "learning_rate": 3.621216913358285e-05, "loss": 0.1449, "step": 11363 }, { "epoch": 0.73, "learning_rate": 3.61961772695841e-05, "loss": 0.1539, "step": 11364 }, { "epoch": 0.73, "learning_rate": 3.618018815717338e-05, "loss": 0.133, "step": 11365 }, { "epoch": 0.73, "learning_rate": 3.616420179704021e-05, "loss": 0.1478, "step": 11366 }, { "epoch": 0.73, "learning_rate": 3.614821818987407e-05, "loss": 0.1468, "step": 11367 }, { "epoch": 0.73, "learning_rate": 3.61322373363642e-05, "loss": 0.1381, "step": 11368 }, { "epoch": 0.73, "learning_rate": 3.611625923719981e-05, "loss": 0.1296, "step": 11369 }, { "epoch": 0.73, "learning_rate": 3.6100283893069985e-05, "loss": 0.1344, "step": 11370 }, { "epoch": 0.73, "learning_rate": 3.608431130466371e-05, "loss": 0.1369, "step": 11371 }, { "epoch": 0.73, "learning_rate": 3.606834147266976e-05, "loss": 0.1627, "step": 11372 }, { "epoch": 0.73, "learning_rate": 3.605237439777683e-05, "loss": 0.1475, "step": 11373 }, { "epoch": 0.73, "learning_rate": 3.603641008067358e-05, "loss": 0.1497, "step": 11374 }, { "epoch": 0.73, "learning_rate": 3.602044852204839e-05, "loss": 0.1499, "step": 11375 }, { "epoch": 0.73, "learning_rate": 3.6004489722589686e-05, "loss": 0.1646, "step": 11376 }, { "epoch": 0.73, "learning_rate": 3.598853368298567e-05, "loss": 0.1565, "step": 11377 }, { "epoch": 0.73, "learning_rate": 3.5972580403924514e-05, "loss": 0.1393, "step": 11378 }, { "epoch": 0.73, "learning_rate": 3.595662988609417e-05, "loss": 0.1736, "step": 11379 }, { "epoch": 0.73, "learning_rate": 3.594068213018249e-05, "loss": 0.1376, "step": 11380 }, { "epoch": 0.73, "learning_rate": 3.5924737136877305e-05, "loss": 0.1479, "step": 11381 }, { "epoch": 0.73, "learning_rate": 3.590879490686617e-05, "loss": 0.1392, "step": 11382 }, { "epoch": 0.73, "learning_rate": 3.589285544083665e-05, "loss": 0.1451, "step": 11383 }, { "epoch": 0.73, "learning_rate": 3.5876918739476186e-05, "loss": 0.1566, "step": 11384 }, { "epoch": 0.73, "learning_rate": 3.586098480347198e-05, "loss": 0.1389, "step": 11385 }, { "epoch": 0.73, "learning_rate": 3.584505363351126e-05, "loss": 0.1311, "step": 11386 }, { "epoch": 0.73, "learning_rate": 3.582912523028101e-05, "loss": 0.1384, "step": 11387 }, { "epoch": 0.73, "learning_rate": 3.58131995944682e-05, "loss": 0.148, "step": 11388 }, { "epoch": 0.73, "learning_rate": 3.579727672675962e-05, "loss": 0.1349, "step": 11389 }, { "epoch": 0.73, "learning_rate": 3.5781356627842e-05, "loss": 0.1456, "step": 11390 }, { "epoch": 0.73, "learning_rate": 3.576543929840186e-05, "loss": 0.1269, "step": 11391 }, { "epoch": 0.73, "learning_rate": 3.574952473912562e-05, "loss": 0.1206, "step": 11392 }, { "epoch": 0.73, "learning_rate": 3.573361295069967e-05, "loss": 0.1538, "step": 11393 }, { "epoch": 0.73, "learning_rate": 3.5717703933810165e-05, "loss": 0.1354, "step": 11394 }, { "epoch": 0.73, "learning_rate": 3.57017976891432e-05, "loss": 0.1495, "step": 11395 }, { "epoch": 0.73, "learning_rate": 3.568589421738477e-05, "loss": 0.1414, "step": 11396 }, { "epoch": 0.73, "learning_rate": 3.5669993519220744e-05, "loss": 0.1236, "step": 11397 }, { "epoch": 0.73, "learning_rate": 3.565409559533684e-05, "loss": 0.1555, "step": 11398 }, { "epoch": 0.73, "learning_rate": 3.563820044641859e-05, "loss": 0.135, "step": 11399 }, { "epoch": 0.73, "learning_rate": 3.562230807315159e-05, "loss": 0.155, "step": 11400 }, { "epoch": 0.73, "learning_rate": 3.560641847622113e-05, "loss": 0.1586, "step": 11401 }, { "epoch": 0.73, "learning_rate": 3.55905316563125e-05, "loss": 0.1397, "step": 11402 }, { "epoch": 0.73, "learning_rate": 3.5574647614110825e-05, "loss": 0.1356, "step": 11403 }, { "epoch": 0.73, "learning_rate": 3.5558766350301144e-05, "loss": 0.133, "step": 11404 }, { "epoch": 0.73, "learning_rate": 3.554288786556833e-05, "loss": 0.1381, "step": 11405 }, { "epoch": 0.73, "learning_rate": 3.552701216059711e-05, "loss": 0.1327, "step": 11406 }, { "epoch": 0.73, "learning_rate": 3.551113923607221e-05, "loss": 0.1376, "step": 11407 }, { "epoch": 0.73, "learning_rate": 3.5495269092678083e-05, "loss": 0.1171, "step": 11408 }, { "epoch": 0.73, "learning_rate": 3.5479401731099196e-05, "loss": 0.1159, "step": 11409 }, { "epoch": 0.73, "learning_rate": 3.546353715201982e-05, "loss": 0.1355, "step": 11410 }, { "epoch": 0.73, "learning_rate": 3.544767535612417e-05, "loss": 0.1336, "step": 11411 }, { "epoch": 0.73, "learning_rate": 3.543181634409626e-05, "loss": 0.1608, "step": 11412 }, { "epoch": 0.73, "learning_rate": 3.5415960116619996e-05, "loss": 0.1273, "step": 11413 }, { "epoch": 0.73, "learning_rate": 3.540010667437925e-05, "loss": 0.1383, "step": 11414 }, { "epoch": 0.73, "learning_rate": 3.538425601805765e-05, "loss": 0.1219, "step": 11415 }, { "epoch": 0.73, "learning_rate": 3.536840814833879e-05, "loss": 0.1344, "step": 11416 }, { "epoch": 0.73, "learning_rate": 3.535256306590614e-05, "loss": 0.1474, "step": 11417 }, { "epoch": 0.73, "learning_rate": 3.5336720771443054e-05, "loss": 0.138, "step": 11418 }, { "epoch": 0.73, "learning_rate": 3.532088126563271e-05, "loss": 0.1435, "step": 11419 }, { "epoch": 0.73, "learning_rate": 3.530504454915816e-05, "loss": 0.144, "step": 11420 }, { "epoch": 0.73, "learning_rate": 3.528921062270246e-05, "loss": 0.1376, "step": 11421 }, { "epoch": 0.73, "learning_rate": 3.5273379486948375e-05, "loss": 0.1444, "step": 11422 }, { "epoch": 0.73, "learning_rate": 3.525755114257867e-05, "loss": 0.1531, "step": 11423 }, { "epoch": 0.73, "learning_rate": 3.524172559027596e-05, "loss": 0.1447, "step": 11424 }, { "epoch": 0.73, "learning_rate": 3.522590283072277e-05, "loss": 0.1402, "step": 11425 }, { "epoch": 0.73, "learning_rate": 3.521008286460142e-05, "loss": 0.1418, "step": 11426 }, { "epoch": 0.73, "learning_rate": 3.519426569259414e-05, "loss": 0.145, "step": 11427 }, { "epoch": 0.73, "learning_rate": 3.5178451315383124e-05, "loss": 0.1313, "step": 11428 }, { "epoch": 0.73, "learning_rate": 3.516263973365027e-05, "loss": 0.1175, "step": 11429 }, { "epoch": 0.73, "learning_rate": 3.5146830948077604e-05, "loss": 0.1337, "step": 11430 }, { "epoch": 0.73, "learning_rate": 3.5131024959346784e-05, "loss": 0.1699, "step": 11431 }, { "epoch": 0.73, "learning_rate": 3.511522176813953e-05, "loss": 0.1436, "step": 11432 }, { "epoch": 0.73, "learning_rate": 3.509942137513733e-05, "loss": 0.1852, "step": 11433 }, { "epoch": 0.73, "learning_rate": 3.5083623781021556e-05, "loss": 0.1435, "step": 11434 }, { "epoch": 0.73, "learning_rate": 3.506782898647353e-05, "loss": 0.1393, "step": 11435 }, { "epoch": 0.73, "learning_rate": 3.5052036992174395e-05, "loss": 0.1357, "step": 11436 }, { "epoch": 0.73, "learning_rate": 3.503624779880525e-05, "loss": 0.1403, "step": 11437 }, { "epoch": 0.73, "learning_rate": 3.5020461407046936e-05, "loss": 0.144, "step": 11438 }, { "epoch": 0.73, "learning_rate": 3.500467781758032e-05, "loss": 0.1373, "step": 11439 }, { "epoch": 0.73, "learning_rate": 3.498889703108605e-05, "loss": 0.1402, "step": 11440 }, { "epoch": 0.73, "learning_rate": 3.497311904824466e-05, "loss": 0.1579, "step": 11441 }, { "epoch": 0.73, "learning_rate": 3.4957343869736604e-05, "loss": 0.1415, "step": 11442 }, { "epoch": 0.73, "learning_rate": 3.49415714962422e-05, "loss": 0.1458, "step": 11443 }, { "epoch": 0.73, "learning_rate": 3.492580192844169e-05, "loss": 0.1391, "step": 11444 }, { "epoch": 0.73, "learning_rate": 3.4910035167015075e-05, "loss": 0.1285, "step": 11445 }, { "epoch": 0.73, "learning_rate": 3.489427121264237e-05, "loss": 0.1348, "step": 11446 }, { "epoch": 0.73, "learning_rate": 3.487851006600339e-05, "loss": 0.1235, "step": 11447 }, { "epoch": 0.73, "learning_rate": 3.4862751727777797e-05, "loss": 0.1265, "step": 11448 }, { "epoch": 0.73, "learning_rate": 3.484699619864522e-05, "loss": 0.1403, "step": 11449 }, { "epoch": 0.73, "learning_rate": 3.483124347928513e-05, "loss": 0.149, "step": 11450 }, { "epoch": 0.73, "learning_rate": 3.48154935703769e-05, "loss": 0.1693, "step": 11451 }, { "epoch": 0.73, "learning_rate": 3.479974647259969e-05, "loss": 0.1291, "step": 11452 }, { "epoch": 0.73, "learning_rate": 3.478400218663269e-05, "loss": 0.1365, "step": 11453 }, { "epoch": 0.73, "learning_rate": 3.4768260713154844e-05, "loss": 0.1249, "step": 11454 }, { "epoch": 0.73, "learning_rate": 3.475252205284496e-05, "loss": 0.1408, "step": 11455 }, { "epoch": 0.73, "learning_rate": 3.4736786206381845e-05, "loss": 0.1235, "step": 11456 }, { "epoch": 0.73, "learning_rate": 3.47210531744441e-05, "loss": 0.1395, "step": 11457 }, { "epoch": 0.73, "learning_rate": 3.470532295771025e-05, "loss": 0.1471, "step": 11458 }, { "epoch": 0.73, "learning_rate": 3.468959555685864e-05, "loss": 0.1513, "step": 11459 }, { "epoch": 0.73, "learning_rate": 3.4673870972567545e-05, "loss": 0.1362, "step": 11460 }, { "epoch": 0.73, "learning_rate": 3.46581492055151e-05, "loss": 0.1337, "step": 11461 }, { "epoch": 0.73, "learning_rate": 3.4642430256379276e-05, "loss": 0.1584, "step": 11462 }, { "epoch": 0.73, "learning_rate": 3.4626714125838e-05, "loss": 0.1373, "step": 11463 }, { "epoch": 0.74, "learning_rate": 3.461100081456904e-05, "loss": 0.1495, "step": 11464 }, { "epoch": 0.74, "learning_rate": 3.459529032325008e-05, "loss": 0.1446, "step": 11465 }, { "epoch": 0.74, "learning_rate": 3.457958265255861e-05, "loss": 0.14, "step": 11466 }, { "epoch": 0.74, "learning_rate": 3.4563877803171996e-05, "loss": 0.134, "step": 11467 }, { "epoch": 0.74, "learning_rate": 3.45481757757676e-05, "loss": 0.1214, "step": 11468 }, { "epoch": 0.74, "learning_rate": 3.4532476571022473e-05, "loss": 0.1356, "step": 11469 }, { "epoch": 0.74, "learning_rate": 3.4516780189613794e-05, "loss": 0.1219, "step": 11470 }, { "epoch": 0.74, "learning_rate": 3.4501086632218394e-05, "loss": 0.1303, "step": 11471 }, { "epoch": 0.74, "learning_rate": 3.448539589951311e-05, "loss": 0.1371, "step": 11472 }, { "epoch": 0.74, "learning_rate": 3.44697079921746e-05, "loss": 0.1379, "step": 11473 }, { "epoch": 0.74, "learning_rate": 3.445402291087938e-05, "loss": 0.1411, "step": 11474 }, { "epoch": 0.74, "learning_rate": 3.443834065630391e-05, "loss": 0.1514, "step": 11475 }, { "epoch": 0.74, "learning_rate": 3.442266122912452e-05, "loss": 0.1473, "step": 11476 }, { "epoch": 0.74, "learning_rate": 3.44069846300174e-05, "loss": 0.1388, "step": 11477 }, { "epoch": 0.74, "learning_rate": 3.439131085965857e-05, "loss": 0.1285, "step": 11478 }, { "epoch": 0.74, "learning_rate": 3.4375639918724024e-05, "loss": 0.1449, "step": 11479 }, { "epoch": 0.74, "learning_rate": 3.4359971807889566e-05, "loss": 0.1399, "step": 11480 }, { "epoch": 0.74, "learning_rate": 3.434430652783085e-05, "loss": 0.1384, "step": 11481 }, { "epoch": 0.74, "learning_rate": 3.4328644079223495e-05, "loss": 0.1417, "step": 11482 }, { "epoch": 0.74, "learning_rate": 3.431298446274295e-05, "loss": 0.1411, "step": 11483 }, { "epoch": 0.74, "learning_rate": 3.429732767906458e-05, "loss": 0.1216, "step": 11484 }, { "epoch": 0.74, "learning_rate": 3.428167372886354e-05, "loss": 0.132, "step": 11485 }, { "epoch": 0.74, "learning_rate": 3.426602261281496e-05, "loss": 0.123, "step": 11486 }, { "epoch": 0.74, "learning_rate": 3.4250374331593804e-05, "loss": 0.1307, "step": 11487 }, { "epoch": 0.74, "learning_rate": 3.4234728885874865e-05, "loss": 0.1303, "step": 11488 }, { "epoch": 0.74, "learning_rate": 3.421908627633289e-05, "loss": 0.1339, "step": 11489 }, { "epoch": 0.74, "learning_rate": 3.42034465036425e-05, "loss": 0.157, "step": 11490 }, { "epoch": 0.74, "learning_rate": 3.41878095684782e-05, "loss": 0.1587, "step": 11491 }, { "epoch": 0.74, "learning_rate": 3.417217547151427e-05, "loss": 0.1414, "step": 11492 }, { "epoch": 0.74, "learning_rate": 3.4156544213425e-05, "loss": 0.1242, "step": 11493 }, { "epoch": 0.74, "learning_rate": 3.414091579488449e-05, "loss": 0.1673, "step": 11494 }, { "epoch": 0.74, "learning_rate": 3.4125290216566676e-05, "loss": 0.1444, "step": 11495 }, { "epoch": 0.74, "learning_rate": 3.410966747914546e-05, "loss": 0.1351, "step": 11496 }, { "epoch": 0.74, "learning_rate": 3.40940475832946e-05, "loss": 0.1585, "step": 11497 }, { "epoch": 0.74, "learning_rate": 3.407843052968772e-05, "loss": 0.1372, "step": 11498 }, { "epoch": 0.74, "learning_rate": 3.406281631899827e-05, "loss": 0.1347, "step": 11499 }, { "epoch": 0.74, "learning_rate": 3.404720495189969e-05, "loss": 0.1548, "step": 11500 }, { "epoch": 0.74, "learning_rate": 3.4031596429065194e-05, "loss": 0.1458, "step": 11501 }, { "epoch": 0.74, "learning_rate": 3.401599075116789e-05, "loss": 0.1285, "step": 11502 }, { "epoch": 0.74, "learning_rate": 3.4000387918880794e-05, "loss": 0.1482, "step": 11503 }, { "epoch": 0.74, "learning_rate": 3.3984787932876814e-05, "loss": 0.1551, "step": 11504 }, { "epoch": 0.74, "learning_rate": 3.3969190793828734e-05, "loss": 0.1298, "step": 11505 }, { "epoch": 0.74, "learning_rate": 3.3953596502409124e-05, "loss": 0.1656, "step": 11506 }, { "epoch": 0.74, "learning_rate": 3.3938005059290564e-05, "loss": 0.1391, "step": 11507 }, { "epoch": 0.74, "learning_rate": 3.3922416465145435e-05, "loss": 0.154, "step": 11508 }, { "epoch": 0.74, "learning_rate": 3.390683072064594e-05, "loss": 0.1477, "step": 11509 }, { "epoch": 0.74, "learning_rate": 3.389124782646428e-05, "loss": 0.1257, "step": 11510 }, { "epoch": 0.74, "learning_rate": 3.387566778327248e-05, "loss": 0.1486, "step": 11511 }, { "epoch": 0.74, "learning_rate": 3.3860090591742466e-05, "loss": 0.1522, "step": 11512 }, { "epoch": 0.74, "learning_rate": 3.3844516252545944e-05, "loss": 0.1286, "step": 11513 }, { "epoch": 0.74, "learning_rate": 3.3828944766354656e-05, "loss": 0.1517, "step": 11514 }, { "epoch": 0.74, "learning_rate": 3.3813376133840094e-05, "loss": 0.1307, "step": 11515 }, { "epoch": 0.74, "learning_rate": 3.3797810355673584e-05, "loss": 0.1387, "step": 11516 }, { "epoch": 0.74, "learning_rate": 3.378224743252656e-05, "loss": 0.1623, "step": 11517 }, { "epoch": 0.74, "learning_rate": 3.376668736507009e-05, "loss": 0.1423, "step": 11518 }, { "epoch": 0.74, "learning_rate": 3.375113015397526e-05, "loss": 0.1371, "step": 11519 }, { "epoch": 0.74, "learning_rate": 3.3735575799912946e-05, "loss": 0.1503, "step": 11520 }, { "epoch": 0.74, "learning_rate": 3.372002430355399e-05, "loss": 0.1362, "step": 11521 }, { "epoch": 0.74, "learning_rate": 3.3704475665568993e-05, "loss": 0.1322, "step": 11522 }, { "epoch": 0.74, "learning_rate": 3.368892988662854e-05, "loss": 0.1411, "step": 11523 }, { "epoch": 0.74, "learning_rate": 3.3673386967403095e-05, "loss": 0.1365, "step": 11524 }, { "epoch": 0.74, "learning_rate": 3.3657846908562885e-05, "loss": 0.1317, "step": 11525 }, { "epoch": 0.74, "learning_rate": 3.364230971077816e-05, "loss": 0.1375, "step": 11526 }, { "epoch": 0.74, "learning_rate": 3.362677537471889e-05, "loss": 0.1381, "step": 11527 }, { "epoch": 0.74, "learning_rate": 3.361124390105509e-05, "loss": 0.1435, "step": 11528 }, { "epoch": 0.74, "learning_rate": 3.35957152904565e-05, "loss": 0.1362, "step": 11529 }, { "epoch": 0.74, "learning_rate": 3.358018954359281e-05, "loss": 0.1322, "step": 11530 }, { "epoch": 0.74, "learning_rate": 3.3564666661133645e-05, "loss": 0.1434, "step": 11531 }, { "epoch": 0.74, "learning_rate": 3.354914664374835e-05, "loss": 0.1498, "step": 11532 }, { "epoch": 0.74, "learning_rate": 3.3533629492106335e-05, "loss": 0.1566, "step": 11533 }, { "epoch": 0.74, "learning_rate": 3.3518115206876686e-05, "loss": 0.1341, "step": 11534 }, { "epoch": 0.74, "learning_rate": 3.350260378872855e-05, "loss": 0.1223, "step": 11535 }, { "epoch": 0.74, "learning_rate": 3.348709523833081e-05, "loss": 0.1209, "step": 11536 }, { "epoch": 0.74, "learning_rate": 3.347158955635231e-05, "loss": 0.1408, "step": 11537 }, { "epoch": 0.74, "learning_rate": 3.345608674346178e-05, "loss": 0.1388, "step": 11538 }, { "epoch": 0.74, "learning_rate": 3.344058680032772e-05, "loss": 0.1182, "step": 11539 }, { "epoch": 0.74, "learning_rate": 3.342508972761864e-05, "loss": 0.1416, "step": 11540 }, { "epoch": 0.74, "learning_rate": 3.3409595526002845e-05, "loss": 0.1302, "step": 11541 }, { "epoch": 0.74, "learning_rate": 3.339410419614849e-05, "loss": 0.1375, "step": 11542 }, { "epoch": 0.74, "learning_rate": 3.337861573872367e-05, "loss": 0.1308, "step": 11543 }, { "epoch": 0.74, "learning_rate": 3.3363130154396374e-05, "loss": 0.1444, "step": 11544 }, { "epoch": 0.74, "learning_rate": 3.3347647443834416e-05, "loss": 0.1423, "step": 11545 }, { "epoch": 0.74, "learning_rate": 3.333216760770548e-05, "loss": 0.1381, "step": 11546 }, { "epoch": 0.74, "learning_rate": 3.3316690646677175e-05, "loss": 0.1346, "step": 11547 }, { "epoch": 0.74, "learning_rate": 3.330121656141694e-05, "loss": 0.137, "step": 11548 }, { "epoch": 0.74, "learning_rate": 3.328574535259207e-05, "loss": 0.1294, "step": 11549 }, { "epoch": 0.74, "learning_rate": 3.32702770208698e-05, "loss": 0.1474, "step": 11550 }, { "epoch": 0.74, "learning_rate": 3.325481156691723e-05, "loss": 0.1451, "step": 11551 }, { "epoch": 0.74, "learning_rate": 3.323934899140134e-05, "loss": 0.1312, "step": 11552 }, { "epoch": 0.74, "learning_rate": 3.3223889294988896e-05, "loss": 0.1282, "step": 11553 }, { "epoch": 0.74, "learning_rate": 3.3208432478346684e-05, "loss": 0.1241, "step": 11554 }, { "epoch": 0.74, "learning_rate": 3.319297854214126e-05, "loss": 0.1286, "step": 11555 }, { "epoch": 0.74, "learning_rate": 3.3177527487039004e-05, "loss": 0.1262, "step": 11556 }, { "epoch": 0.74, "learning_rate": 3.3162079313706405e-05, "loss": 0.1632, "step": 11557 }, { "epoch": 0.74, "learning_rate": 3.314663402280957e-05, "loss": 0.1299, "step": 11558 }, { "epoch": 0.74, "learning_rate": 3.3131191615014666e-05, "loss": 0.1464, "step": 11559 }, { "epoch": 0.74, "learning_rate": 3.3115752090987575e-05, "loss": 0.1442, "step": 11560 }, { "epoch": 0.74, "learning_rate": 3.3100315451394205e-05, "loss": 0.1214, "step": 11561 }, { "epoch": 0.74, "learning_rate": 3.308488169690021e-05, "loss": 0.1303, "step": 11562 }, { "epoch": 0.74, "learning_rate": 3.3069450828171233e-05, "loss": 0.1345, "step": 11563 }, { "epoch": 0.74, "learning_rate": 3.305402284587275e-05, "loss": 0.1317, "step": 11564 }, { "epoch": 0.74, "learning_rate": 3.303859775067005e-05, "loss": 0.1407, "step": 11565 }, { "epoch": 0.74, "learning_rate": 3.302317554322841e-05, "loss": 0.1466, "step": 11566 }, { "epoch": 0.74, "learning_rate": 3.300775622421288e-05, "loss": 0.1174, "step": 11567 }, { "epoch": 0.74, "learning_rate": 3.2992339794288476e-05, "loss": 0.1343, "step": 11568 }, { "epoch": 0.74, "learning_rate": 3.297692625411998e-05, "loss": 0.117, "step": 11569 }, { "epoch": 0.74, "learning_rate": 3.296151560437214e-05, "loss": 0.1202, "step": 11570 }, { "epoch": 0.74, "learning_rate": 3.2946107845709596e-05, "loss": 0.1277, "step": 11571 }, { "epoch": 0.74, "learning_rate": 3.293070297879675e-05, "loss": 0.1148, "step": 11572 }, { "epoch": 0.74, "learning_rate": 3.291530100429802e-05, "loss": 0.144, "step": 11573 }, { "epoch": 0.74, "learning_rate": 3.2899901922877554e-05, "loss": 0.1459, "step": 11574 }, { "epoch": 0.74, "learning_rate": 3.288450573519951e-05, "loss": 0.1491, "step": 11575 }, { "epoch": 0.74, "learning_rate": 3.28691124419278e-05, "loss": 0.1381, "step": 11576 }, { "epoch": 0.74, "learning_rate": 3.285372204372631e-05, "loss": 0.1376, "step": 11577 }, { "epoch": 0.74, "learning_rate": 3.283833454125878e-05, "loss": 0.1264, "step": 11578 }, { "epoch": 0.74, "learning_rate": 3.2822949935188777e-05, "loss": 0.1137, "step": 11579 }, { "epoch": 0.74, "learning_rate": 3.28075682261798e-05, "loss": 0.1354, "step": 11580 }, { "epoch": 0.74, "learning_rate": 3.2792189414895146e-05, "loss": 0.1277, "step": 11581 }, { "epoch": 0.74, "learning_rate": 3.2776813501998105e-05, "loss": 0.1311, "step": 11582 }, { "epoch": 0.74, "learning_rate": 3.2761440488151705e-05, "loss": 0.1577, "step": 11583 }, { "epoch": 0.74, "learning_rate": 3.2746070374018964e-05, "loss": 0.1365, "step": 11584 }, { "epoch": 0.74, "learning_rate": 3.273070316026274e-05, "loss": 0.1382, "step": 11585 }, { "epoch": 0.74, "learning_rate": 3.271533884754572e-05, "loss": 0.1397, "step": 11586 }, { "epoch": 0.74, "learning_rate": 3.269997743653054e-05, "loss": 0.1454, "step": 11587 }, { "epoch": 0.74, "learning_rate": 3.268461892787962e-05, "loss": 0.1378, "step": 11588 }, { "epoch": 0.74, "learning_rate": 3.266926332225536e-05, "loss": 0.1444, "step": 11589 }, { "epoch": 0.74, "learning_rate": 3.265391062031994e-05, "loss": 0.1308, "step": 11590 }, { "epoch": 0.74, "learning_rate": 3.2638560822735475e-05, "loss": 0.1345, "step": 11591 }, { "epoch": 0.74, "learning_rate": 3.2623213930163964e-05, "loss": 0.1213, "step": 11592 }, { "epoch": 0.74, "learning_rate": 3.26078699432672e-05, "loss": 0.1473, "step": 11593 }, { "epoch": 0.74, "learning_rate": 3.259252886270696e-05, "loss": 0.1449, "step": 11594 }, { "epoch": 0.74, "learning_rate": 3.257719068914479e-05, "loss": 0.1663, "step": 11595 }, { "epoch": 0.74, "learning_rate": 3.256185542324219e-05, "loss": 0.1359, "step": 11596 }, { "epoch": 0.74, "learning_rate": 3.2546523065660483e-05, "loss": 0.1249, "step": 11597 }, { "epoch": 0.74, "learning_rate": 3.25311936170609e-05, "loss": 0.1538, "step": 11598 }, { "epoch": 0.74, "learning_rate": 3.251586707810457e-05, "loss": 0.1362, "step": 11599 }, { "epoch": 0.74, "learning_rate": 3.2500543449452395e-05, "loss": 0.1436, "step": 11600 }, { "epoch": 0.74, "learning_rate": 3.2485222731765274e-05, "loss": 0.1445, "step": 11601 }, { "epoch": 0.74, "learning_rate": 3.246990492570387e-05, "loss": 0.1382, "step": 11602 }, { "epoch": 0.74, "learning_rate": 3.2454590031928824e-05, "loss": 0.1277, "step": 11603 }, { "epoch": 0.74, "learning_rate": 3.243927805110061e-05, "loss": 0.1325, "step": 11604 }, { "epoch": 0.74, "learning_rate": 3.242396898387951e-05, "loss": 0.1327, "step": 11605 }, { "epoch": 0.74, "learning_rate": 3.24086628309258e-05, "loss": 0.1291, "step": 11606 }, { "epoch": 0.74, "learning_rate": 3.239335959289952e-05, "loss": 0.1597, "step": 11607 }, { "epoch": 0.74, "learning_rate": 3.237805927046068e-05, "loss": 0.1687, "step": 11608 }, { "epoch": 0.74, "learning_rate": 3.2362761864269066e-05, "loss": 0.1139, "step": 11609 }, { "epoch": 0.74, "learning_rate": 3.234746737498442e-05, "loss": 0.1208, "step": 11610 }, { "epoch": 0.74, "learning_rate": 3.233217580326635e-05, "loss": 0.1335, "step": 11611 }, { "epoch": 0.74, "learning_rate": 3.231688714977427e-05, "loss": 0.1324, "step": 11612 }, { "epoch": 0.74, "learning_rate": 3.230160141516757e-05, "loss": 0.1333, "step": 11613 }, { "epoch": 0.74, "learning_rate": 3.228631860010539e-05, "loss": 0.1474, "step": 11614 }, { "epoch": 0.74, "learning_rate": 3.227103870524688e-05, "loss": 0.1619, "step": 11615 }, { "epoch": 0.74, "learning_rate": 3.225576173125095e-05, "loss": 0.1188, "step": 11616 }, { "epoch": 0.74, "learning_rate": 3.2240487678776434e-05, "loss": 0.1634, "step": 11617 }, { "epoch": 0.74, "learning_rate": 3.2225216548482085e-05, "loss": 0.1461, "step": 11618 }, { "epoch": 0.74, "learning_rate": 3.220994834102642e-05, "loss": 0.1293, "step": 11619 }, { "epoch": 0.75, "learning_rate": 3.219468305706794e-05, "loss": 0.1277, "step": 11620 }, { "epoch": 0.75, "learning_rate": 3.217942069726494e-05, "loss": 0.1408, "step": 11621 }, { "epoch": 0.75, "learning_rate": 3.216416126227565e-05, "loss": 0.1376, "step": 11622 }, { "epoch": 0.75, "learning_rate": 3.214890475275809e-05, "loss": 0.146, "step": 11623 }, { "epoch": 0.75, "learning_rate": 3.213365116937026e-05, "loss": 0.1328, "step": 11624 }, { "epoch": 0.75, "learning_rate": 3.2118400512769996e-05, "loss": 0.1334, "step": 11625 }, { "epoch": 0.75, "learning_rate": 3.210315278361493e-05, "loss": 0.1401, "step": 11626 }, { "epoch": 0.75, "learning_rate": 3.2087907982562705e-05, "loss": 0.1541, "step": 11627 }, { "epoch": 0.75, "learning_rate": 3.207266611027069e-05, "loss": 0.1177, "step": 11628 }, { "epoch": 0.75, "learning_rate": 3.205742716739628e-05, "loss": 0.1586, "step": 11629 }, { "epoch": 0.75, "learning_rate": 3.204219115459659e-05, "loss": 0.1354, "step": 11630 }, { "epoch": 0.75, "learning_rate": 3.202695807252871e-05, "loss": 0.1262, "step": 11631 }, { "epoch": 0.75, "learning_rate": 3.201172792184963e-05, "loss": 0.1433, "step": 11632 }, { "epoch": 0.75, "learning_rate": 3.1996500703216095e-05, "loss": 0.1305, "step": 11633 }, { "epoch": 0.75, "learning_rate": 3.198127641728483e-05, "loss": 0.1304, "step": 11634 }, { "epoch": 0.75, "learning_rate": 3.1966055064712364e-05, "loss": 0.1519, "step": 11635 }, { "epoch": 0.75, "learning_rate": 3.195083664615517e-05, "loss": 0.143, "step": 11636 }, { "epoch": 0.75, "learning_rate": 3.19356211622695e-05, "loss": 0.1357, "step": 11637 }, { "epoch": 0.75, "learning_rate": 3.192040861371156e-05, "loss": 0.1332, "step": 11638 }, { "epoch": 0.75, "learning_rate": 3.1905199001137434e-05, "loss": 0.1458, "step": 11639 }, { "epoch": 0.75, "learning_rate": 3.188999232520298e-05, "loss": 0.128, "step": 11640 }, { "epoch": 0.75, "learning_rate": 3.187478858656408e-05, "loss": 0.1347, "step": 11641 }, { "epoch": 0.75, "learning_rate": 3.185958778587632e-05, "loss": 0.154, "step": 11642 }, { "epoch": 0.75, "learning_rate": 3.1844389923795294e-05, "loss": 0.1276, "step": 11643 }, { "epoch": 0.75, "learning_rate": 3.1829195000976456e-05, "loss": 0.1499, "step": 11644 }, { "epoch": 0.75, "learning_rate": 3.181400301807501e-05, "loss": 0.1383, "step": 11645 }, { "epoch": 0.75, "learning_rate": 3.179881397574621e-05, "loss": 0.1483, "step": 11646 }, { "epoch": 0.75, "learning_rate": 3.1783627874645014e-05, "loss": 0.1374, "step": 11647 }, { "epoch": 0.75, "learning_rate": 3.176844471542643e-05, "loss": 0.1346, "step": 11648 }, { "epoch": 0.75, "learning_rate": 3.175326449874513e-05, "loss": 0.1442, "step": 11649 }, { "epoch": 0.75, "learning_rate": 3.173808722525584e-05, "loss": 0.1323, "step": 11650 }, { "epoch": 0.75, "learning_rate": 3.172291289561312e-05, "loss": 0.142, "step": 11651 }, { "epoch": 0.75, "learning_rate": 3.17077415104713e-05, "loss": 0.1291, "step": 11652 }, { "epoch": 0.75, "learning_rate": 3.1692573070484723e-05, "loss": 0.1554, "step": 11653 }, { "epoch": 0.75, "learning_rate": 3.1677407576307484e-05, "loss": 0.1319, "step": 11654 }, { "epoch": 0.75, "learning_rate": 3.1662245028593664e-05, "loss": 0.1274, "step": 11655 }, { "epoch": 0.75, "learning_rate": 3.164708542799709e-05, "loss": 0.1569, "step": 11656 }, { "epoch": 0.75, "learning_rate": 3.163192877517157e-05, "loss": 0.1371, "step": 11657 }, { "epoch": 0.75, "learning_rate": 3.161677507077078e-05, "loss": 0.1301, "step": 11658 }, { "epoch": 0.75, "learning_rate": 3.1601624315448166e-05, "loss": 0.1382, "step": 11659 }, { "epoch": 0.75, "learning_rate": 3.158647650985718e-05, "loss": 0.1284, "step": 11660 }, { "epoch": 0.75, "learning_rate": 3.157133165465103e-05, "loss": 0.1455, "step": 11661 }, { "epoch": 0.75, "learning_rate": 3.155618975048289e-05, "loss": 0.1237, "step": 11662 }, { "epoch": 0.75, "learning_rate": 3.154105079800572e-05, "loss": 0.142, "step": 11663 }, { "epoch": 0.75, "learning_rate": 3.1525914797872437e-05, "loss": 0.1299, "step": 11664 }, { "epoch": 0.75, "learning_rate": 3.151078175073581e-05, "loss": 0.1518, "step": 11665 }, { "epoch": 0.75, "learning_rate": 3.1495651657248395e-05, "loss": 0.1444, "step": 11666 }, { "epoch": 0.75, "learning_rate": 3.148052451806276e-05, "loss": 0.1424, "step": 11667 }, { "epoch": 0.75, "learning_rate": 3.146540033383122e-05, "loss": 0.136, "step": 11668 }, { "epoch": 0.75, "learning_rate": 3.145027910520607e-05, "loss": 0.1494, "step": 11669 }, { "epoch": 0.75, "learning_rate": 3.143516083283936e-05, "loss": 0.1425, "step": 11670 }, { "epoch": 0.75, "learning_rate": 3.142004551738311e-05, "loss": 0.1525, "step": 11671 }, { "epoch": 0.75, "learning_rate": 3.1404933159489214e-05, "loss": 0.1477, "step": 11672 }, { "epoch": 0.75, "learning_rate": 3.138982375980934e-05, "loss": 0.1412, "step": 11673 }, { "epoch": 0.75, "learning_rate": 3.137471731899515e-05, "loss": 0.1312, "step": 11674 }, { "epoch": 0.75, "learning_rate": 3.135961383769807e-05, "loss": 0.1364, "step": 11675 }, { "epoch": 0.75, "learning_rate": 3.1344513316569504e-05, "loss": 0.1386, "step": 11676 }, { "epoch": 0.75, "learning_rate": 3.132941575626061e-05, "loss": 0.121, "step": 11677 }, { "epoch": 0.75, "learning_rate": 3.13143211574225e-05, "loss": 0.1487, "step": 11678 }, { "epoch": 0.75, "learning_rate": 3.12992295207062e-05, "loss": 0.1417, "step": 11679 }, { "epoch": 0.75, "learning_rate": 3.1284140846762456e-05, "loss": 0.1339, "step": 11680 }, { "epoch": 0.75, "learning_rate": 3.1269055136242063e-05, "loss": 0.1309, "step": 11681 }, { "epoch": 0.75, "learning_rate": 3.1253972389795536e-05, "loss": 0.1241, "step": 11682 }, { "epoch": 0.75, "learning_rate": 3.123889260807339e-05, "loss": 0.1716, "step": 11683 }, { "epoch": 0.75, "learning_rate": 3.1223815791725855e-05, "loss": 0.1297, "step": 11684 }, { "epoch": 0.75, "learning_rate": 3.120874194140326e-05, "loss": 0.1368, "step": 11685 }, { "epoch": 0.75, "learning_rate": 3.119367105775561e-05, "loss": 0.1155, "step": 11686 }, { "epoch": 0.75, "learning_rate": 3.1178603141432826e-05, "loss": 0.1356, "step": 11687 }, { "epoch": 0.75, "learning_rate": 3.116353819308478e-05, "loss": 0.1322, "step": 11688 }, { "epoch": 0.75, "learning_rate": 3.1148476213361077e-05, "loss": 0.139, "step": 11689 }, { "epoch": 0.75, "learning_rate": 3.113341720291134e-05, "loss": 0.138, "step": 11690 }, { "epoch": 0.75, "learning_rate": 3.111836116238498e-05, "loss": 0.1498, "step": 11691 }, { "epoch": 0.75, "learning_rate": 3.110330809243134e-05, "loss": 0.1379, "step": 11692 }, { "epoch": 0.75, "learning_rate": 3.108825799369956e-05, "loss": 0.1297, "step": 11693 }, { "epoch": 0.75, "learning_rate": 3.1073210866838665e-05, "loss": 0.147, "step": 11694 }, { "epoch": 0.75, "learning_rate": 3.105816671249763e-05, "loss": 0.137, "step": 11695 }, { "epoch": 0.75, "learning_rate": 3.104312553132517e-05, "loss": 0.1443, "step": 11696 }, { "epoch": 0.75, "learning_rate": 3.102808732396999e-05, "loss": 0.1415, "step": 11697 }, { "epoch": 0.75, "learning_rate": 3.101305209108065e-05, "loss": 0.1197, "step": 11698 }, { "epoch": 0.75, "learning_rate": 3.0998019833305514e-05, "loss": 0.1413, "step": 11699 }, { "epoch": 0.75, "learning_rate": 3.09829905512929e-05, "loss": 0.1523, "step": 11700 }, { "epoch": 0.75, "learning_rate": 3.096796424569089e-05, "loss": 0.128, "step": 11701 }, { "epoch": 0.75, "learning_rate": 3.09529409171476e-05, "loss": 0.1326, "step": 11702 }, { "epoch": 0.75, "learning_rate": 3.093792056631082e-05, "loss": 0.15, "step": 11703 }, { "epoch": 0.75, "learning_rate": 3.092290319382837e-05, "loss": 0.1282, "step": 11704 }, { "epoch": 0.75, "learning_rate": 3.09078888003479e-05, "loss": 0.1371, "step": 11705 }, { "epoch": 0.75, "learning_rate": 3.089287738651688e-05, "loss": 0.1407, "step": 11706 }, { "epoch": 0.75, "learning_rate": 3.0877868952982726e-05, "loss": 0.142, "step": 11707 }, { "epoch": 0.75, "learning_rate": 3.086286350039263e-05, "loss": 0.1486, "step": 11708 }, { "epoch": 0.75, "learning_rate": 3.08478610293938e-05, "loss": 0.134, "step": 11709 }, { "epoch": 0.75, "learning_rate": 3.083286154063314e-05, "loss": 0.145, "step": 11710 }, { "epoch": 0.75, "learning_rate": 3.081786503475754e-05, "loss": 0.1473, "step": 11711 }, { "epoch": 0.75, "learning_rate": 3.080287151241379e-05, "loss": 0.1298, "step": 11712 }, { "epoch": 0.75, "learning_rate": 3.078788097424843e-05, "loss": 0.1358, "step": 11713 }, { "epoch": 0.75, "learning_rate": 3.077289342090799e-05, "loss": 0.1358, "step": 11714 }, { "epoch": 0.75, "learning_rate": 3.075790885303875e-05, "loss": 0.1642, "step": 11715 }, { "epoch": 0.75, "learning_rate": 3.074292727128703e-05, "loss": 0.1334, "step": 11716 }, { "epoch": 0.75, "learning_rate": 3.072794867629881e-05, "loss": 0.1333, "step": 11717 }, { "epoch": 0.75, "learning_rate": 3.071297306872013e-05, "loss": 0.1421, "step": 11718 }, { "epoch": 0.75, "learning_rate": 3.0698000449196815e-05, "loss": 0.1288, "step": 11719 }, { "epoch": 0.75, "learning_rate": 3.068303081837455e-05, "loss": 0.1307, "step": 11720 }, { "epoch": 0.75, "learning_rate": 3.0668064176898935e-05, "loss": 0.13, "step": 11721 }, { "epoch": 0.75, "learning_rate": 3.065310052541538e-05, "loss": 0.1295, "step": 11722 }, { "epoch": 0.75, "learning_rate": 3.063813986456926e-05, "loss": 0.1433, "step": 11723 }, { "epoch": 0.75, "learning_rate": 3.062318219500569e-05, "loss": 0.1362, "step": 11724 }, { "epoch": 0.75, "learning_rate": 3.0608227517369794e-05, "loss": 0.1326, "step": 11725 }, { "epoch": 0.75, "learning_rate": 3.05932758323065e-05, "loss": 0.1294, "step": 11726 }, { "epoch": 0.75, "learning_rate": 3.057832714046056e-05, "loss": 0.138, "step": 11727 }, { "epoch": 0.75, "learning_rate": 3.0563381442476724e-05, "loss": 0.1325, "step": 11728 }, { "epoch": 0.75, "learning_rate": 3.054843873899945e-05, "loss": 0.1328, "step": 11729 }, { "epoch": 0.75, "learning_rate": 3.053349903067321e-05, "loss": 0.1446, "step": 11730 }, { "epoch": 0.75, "learning_rate": 3.051856231814226e-05, "loss": 0.1363, "step": 11731 }, { "epoch": 0.75, "learning_rate": 3.0503628602050816e-05, "loss": 0.1338, "step": 11732 }, { "epoch": 0.75, "learning_rate": 3.048869788304286e-05, "loss": 0.144, "step": 11733 }, { "epoch": 0.75, "learning_rate": 3.047377016176227e-05, "loss": 0.1351, "step": 11734 }, { "epoch": 0.75, "learning_rate": 3.0458845438852866e-05, "loss": 0.1371, "step": 11735 }, { "epoch": 0.75, "learning_rate": 3.0443923714958235e-05, "loss": 0.1377, "step": 11736 }, { "epoch": 0.75, "learning_rate": 3.042900499072191e-05, "loss": 0.1305, "step": 11737 }, { "epoch": 0.75, "learning_rate": 3.041408926678728e-05, "loss": 0.1584, "step": 11738 }, { "epoch": 0.75, "learning_rate": 3.0399176543797625e-05, "loss": 0.1325, "step": 11739 }, { "epoch": 0.75, "learning_rate": 3.0384266822396035e-05, "loss": 0.128, "step": 11740 }, { "epoch": 0.75, "learning_rate": 3.0369360103225474e-05, "loss": 0.1216, "step": 11741 }, { "epoch": 0.75, "learning_rate": 3.0354456386928864e-05, "loss": 0.1497, "step": 11742 }, { "epoch": 0.75, "learning_rate": 3.0339555674148878e-05, "loss": 0.1285, "step": 11743 }, { "epoch": 0.75, "learning_rate": 3.0324657965528157e-05, "loss": 0.1634, "step": 11744 }, { "epoch": 0.75, "learning_rate": 3.0309763261709167e-05, "loss": 0.1284, "step": 11745 }, { "epoch": 0.75, "learning_rate": 3.0294871563334292e-05, "loss": 0.135, "step": 11746 }, { "epoch": 0.75, "learning_rate": 3.027998287104572e-05, "loss": 0.1409, "step": 11747 }, { "epoch": 0.75, "learning_rate": 3.0265097185485492e-05, "loss": 0.1266, "step": 11748 }, { "epoch": 0.75, "learning_rate": 3.0250214507295637e-05, "loss": 0.1486, "step": 11749 }, { "epoch": 0.75, "learning_rate": 3.0235334837117913e-05, "loss": 0.1142, "step": 11750 }, { "epoch": 0.75, "learning_rate": 3.0220458175594058e-05, "loss": 0.1472, "step": 11751 }, { "epoch": 0.75, "learning_rate": 3.0205584523365626e-05, "loss": 0.1391, "step": 11752 }, { "epoch": 0.75, "learning_rate": 3.0190713881074105e-05, "loss": 0.1312, "step": 11753 }, { "epoch": 0.75, "learning_rate": 3.017584624936075e-05, "loss": 0.1424, "step": 11754 }, { "epoch": 0.75, "learning_rate": 3.0160981628866726e-05, "loss": 0.1686, "step": 11755 }, { "epoch": 0.75, "learning_rate": 3.0146120020233127e-05, "loss": 0.1377, "step": 11756 }, { "epoch": 0.75, "learning_rate": 3.0131261424100822e-05, "loss": 0.1568, "step": 11757 }, { "epoch": 0.75, "learning_rate": 3.0116405841110617e-05, "loss": 0.1371, "step": 11758 }, { "epoch": 0.75, "learning_rate": 3.010155327190318e-05, "loss": 0.1381, "step": 11759 }, { "epoch": 0.75, "learning_rate": 3.008670371711908e-05, "loss": 0.1322, "step": 11760 }, { "epoch": 0.75, "learning_rate": 3.0071857177398665e-05, "loss": 0.152, "step": 11761 }, { "epoch": 0.75, "learning_rate": 3.005701365338217e-05, "loss": 0.1594, "step": 11762 }, { "epoch": 0.75, "learning_rate": 3.0042173145709817e-05, "loss": 0.1319, "step": 11763 }, { "epoch": 0.75, "learning_rate": 3.0027335655021537e-05, "loss": 0.1461, "step": 11764 }, { "epoch": 0.75, "learning_rate": 3.0012501181957243e-05, "loss": 0.1349, "step": 11765 }, { "epoch": 0.75, "learning_rate": 2.9997669727156685e-05, "loss": 0.1416, "step": 11766 }, { "epoch": 0.75, "learning_rate": 2.99828412912595e-05, "loss": 0.1449, "step": 11767 }, { "epoch": 0.75, "learning_rate": 2.9968015874905165e-05, "loss": 0.143, "step": 11768 }, { "epoch": 0.75, "learning_rate": 2.9953193478732977e-05, "loss": 0.153, "step": 11769 }, { "epoch": 0.75, "learning_rate": 2.9938374103382262e-05, "loss": 0.1305, "step": 11770 }, { "epoch": 0.75, "learning_rate": 2.9923557749491992e-05, "loss": 0.1564, "step": 11771 }, { "epoch": 0.75, "learning_rate": 2.9908744417701275e-05, "loss": 0.1341, "step": 11772 }, { "epoch": 0.75, "learning_rate": 2.9893934108648836e-05, "loss": 0.163, "step": 11773 }, { "epoch": 0.75, "learning_rate": 2.987912682297347e-05, "loss": 0.1275, "step": 11774 }, { "epoch": 0.76, "learning_rate": 2.9864322561313696e-05, "loss": 0.1332, "step": 11775 }, { "epoch": 0.76, "learning_rate": 2.984952132430795e-05, "loss": 0.142, "step": 11776 }, { "epoch": 0.76, "learning_rate": 2.9834723112594565e-05, "loss": 0.1369, "step": 11777 }, { "epoch": 0.76, "learning_rate": 2.981992792681172e-05, "loss": 0.1449, "step": 11778 }, { "epoch": 0.76, "learning_rate": 2.9805135767597502e-05, "loss": 0.1302, "step": 11779 }, { "epoch": 0.76, "learning_rate": 2.9790346635589818e-05, "loss": 0.1467, "step": 11780 }, { "epoch": 0.76, "learning_rate": 2.9775560531426415e-05, "loss": 0.1444, "step": 11781 }, { "epoch": 0.76, "learning_rate": 2.9760777455745015e-05, "loss": 0.1545, "step": 11782 }, { "epoch": 0.76, "learning_rate": 2.9745997409183102e-05, "loss": 0.1532, "step": 11783 }, { "epoch": 0.76, "learning_rate": 2.973122039237809e-05, "loss": 0.1467, "step": 11784 }, { "epoch": 0.76, "learning_rate": 2.9716446405967267e-05, "loss": 0.1553, "step": 11785 }, { "epoch": 0.76, "learning_rate": 2.9701675450587784e-05, "loss": 0.1363, "step": 11786 }, { "epoch": 0.76, "learning_rate": 2.968690752687663e-05, "loss": 0.1244, "step": 11787 }, { "epoch": 0.76, "learning_rate": 2.9672142635470658e-05, "loss": 0.1433, "step": 11788 }, { "epoch": 0.76, "learning_rate": 2.965738077700666e-05, "loss": 0.1326, "step": 11789 }, { "epoch": 0.76, "learning_rate": 2.96426219521212e-05, "loss": 0.1374, "step": 11790 }, { "epoch": 0.76, "learning_rate": 2.96278661614508e-05, "loss": 0.1304, "step": 11791 }, { "epoch": 0.76, "learning_rate": 2.9613113405631798e-05, "loss": 0.1506, "step": 11792 }, { "epoch": 0.76, "learning_rate": 2.9598363685300466e-05, "loss": 0.1275, "step": 11793 }, { "epoch": 0.76, "learning_rate": 2.9583617001092868e-05, "loss": 0.1386, "step": 11794 }, { "epoch": 0.76, "learning_rate": 2.9568873353644923e-05, "loss": 0.1475, "step": 11795 }, { "epoch": 0.76, "learning_rate": 2.955413274359251e-05, "loss": 0.1453, "step": 11796 }, { "epoch": 0.76, "learning_rate": 2.9539395171571304e-05, "loss": 0.1428, "step": 11797 }, { "epoch": 0.76, "learning_rate": 2.9524660638216872e-05, "loss": 0.1233, "step": 11798 }, { "epoch": 0.76, "learning_rate": 2.950992914416467e-05, "loss": 0.1414, "step": 11799 }, { "epoch": 0.76, "learning_rate": 2.949520069005003e-05, "loss": 0.1409, "step": 11800 }, { "epoch": 0.76, "learning_rate": 2.94804752765081e-05, "loss": 0.162, "step": 11801 }, { "epoch": 0.76, "learning_rate": 2.9465752904173882e-05, "loss": 0.1397, "step": 11802 }, { "epoch": 0.76, "learning_rate": 2.9451033573682374e-05, "loss": 0.146, "step": 11803 }, { "epoch": 0.76, "learning_rate": 2.943631728566827e-05, "loss": 0.1438, "step": 11804 }, { "epoch": 0.76, "learning_rate": 2.942160404076627e-05, "loss": 0.1453, "step": 11805 }, { "epoch": 0.76, "learning_rate": 2.9406893839610884e-05, "loss": 0.1317, "step": 11806 }, { "epoch": 0.76, "learning_rate": 2.9392186682836543e-05, "loss": 0.1207, "step": 11807 }, { "epoch": 0.76, "learning_rate": 2.9377482571077465e-05, "loss": 0.1302, "step": 11808 }, { "epoch": 0.76, "learning_rate": 2.936278150496774e-05, "loss": 0.1487, "step": 11809 }, { "epoch": 0.76, "learning_rate": 2.9348083485141432e-05, "loss": 0.1346, "step": 11810 }, { "epoch": 0.76, "learning_rate": 2.933338851223234e-05, "loss": 0.1353, "step": 11811 }, { "epoch": 0.76, "learning_rate": 2.9318696586874238e-05, "loss": 0.1391, "step": 11812 }, { "epoch": 0.76, "learning_rate": 2.930400770970071e-05, "loss": 0.1363, "step": 11813 }, { "epoch": 0.76, "learning_rate": 2.9289321881345254e-05, "loss": 0.1523, "step": 11814 }, { "epoch": 0.76, "learning_rate": 2.9274639102441194e-05, "loss": 0.1353, "step": 11815 }, { "epoch": 0.76, "learning_rate": 2.9259959373621692e-05, "loss": 0.1299, "step": 11816 }, { "epoch": 0.76, "learning_rate": 2.9245282695519848e-05, "loss": 0.1409, "step": 11817 }, { "epoch": 0.76, "learning_rate": 2.9230609068768632e-05, "loss": 0.1396, "step": 11818 }, { "epoch": 0.76, "learning_rate": 2.921593849400085e-05, "loss": 0.1443, "step": 11819 }, { "epoch": 0.76, "learning_rate": 2.920127097184915e-05, "loss": 0.1387, "step": 11820 }, { "epoch": 0.76, "learning_rate": 2.9186606502946123e-05, "loss": 0.1396, "step": 11821 }, { "epoch": 0.76, "learning_rate": 2.917194508792416e-05, "loss": 0.1317, "step": 11822 }, { "epoch": 0.76, "learning_rate": 2.915728672741552e-05, "loss": 0.153, "step": 11823 }, { "epoch": 0.76, "learning_rate": 2.914263142205238e-05, "loss": 0.146, "step": 11824 }, { "epoch": 0.76, "learning_rate": 2.9127979172466756e-05, "loss": 0.1283, "step": 11825 }, { "epoch": 0.76, "learning_rate": 2.9113329979290583e-05, "loss": 0.1475, "step": 11826 }, { "epoch": 0.76, "learning_rate": 2.909868384315555e-05, "loss": 0.1501, "step": 11827 }, { "epoch": 0.76, "learning_rate": 2.9084040764693343e-05, "loss": 0.1604, "step": 11828 }, { "epoch": 0.76, "learning_rate": 2.9069400744535425e-05, "loss": 0.1331, "step": 11829 }, { "epoch": 0.76, "learning_rate": 2.9054763783313122e-05, "loss": 0.1453, "step": 11830 }, { "epoch": 0.76, "learning_rate": 2.904012988165771e-05, "loss": 0.144, "step": 11831 }, { "epoch": 0.76, "learning_rate": 2.9025499040200267e-05, "loss": 0.1466, "step": 11832 }, { "epoch": 0.76, "learning_rate": 2.9010871259571804e-05, "loss": 0.15, "step": 11833 }, { "epoch": 0.76, "learning_rate": 2.8996246540403095e-05, "loss": 0.1521, "step": 11834 }, { "epoch": 0.76, "learning_rate": 2.8981624883324886e-05, "loss": 0.1158, "step": 11835 }, { "epoch": 0.76, "learning_rate": 2.8967006288967745e-05, "loss": 0.1542, "step": 11836 }, { "epoch": 0.76, "learning_rate": 2.8952390757962056e-05, "loss": 0.1572, "step": 11837 }, { "epoch": 0.76, "learning_rate": 2.8937778290938154e-05, "loss": 0.1472, "step": 11838 }, { "epoch": 0.76, "learning_rate": 2.8923168888526232e-05, "loss": 0.13, "step": 11839 }, { "epoch": 0.76, "learning_rate": 2.8908562551356343e-05, "loss": 0.1538, "step": 11840 }, { "epoch": 0.76, "learning_rate": 2.8893959280058348e-05, "loss": 0.1315, "step": 11841 }, { "epoch": 0.76, "learning_rate": 2.887935907526208e-05, "loss": 0.1325, "step": 11842 }, { "epoch": 0.76, "learning_rate": 2.8864761937597163e-05, "loss": 0.161, "step": 11843 }, { "epoch": 0.76, "learning_rate": 2.8850167867693057e-05, "loss": 0.1318, "step": 11844 }, { "epoch": 0.76, "learning_rate": 2.8835576866179192e-05, "loss": 0.1242, "step": 11845 }, { "epoch": 0.76, "learning_rate": 2.8820988933684812e-05, "loss": 0.1479, "step": 11846 }, { "epoch": 0.76, "learning_rate": 2.8806404070839056e-05, "loss": 0.1455, "step": 11847 }, { "epoch": 0.76, "learning_rate": 2.8791822278270853e-05, "loss": 0.1508, "step": 11848 }, { "epoch": 0.76, "learning_rate": 2.877724355660911e-05, "loss": 0.157, "step": 11849 }, { "epoch": 0.76, "learning_rate": 2.8762667906482522e-05, "loss": 0.1311, "step": 11850 }, { "epoch": 0.76, "learning_rate": 2.874809532851963e-05, "loss": 0.132, "step": 11851 }, { "epoch": 0.76, "learning_rate": 2.8733525823348937e-05, "loss": 0.1406, "step": 11852 }, { "epoch": 0.76, "learning_rate": 2.8718959391598753e-05, "loss": 0.1416, "step": 11853 }, { "epoch": 0.76, "learning_rate": 2.8704396033897294e-05, "loss": 0.1327, "step": 11854 }, { "epoch": 0.76, "learning_rate": 2.8689835750872607e-05, "loss": 0.164, "step": 11855 }, { "epoch": 0.76, "learning_rate": 2.8675278543152552e-05, "loss": 0.1279, "step": 11856 }, { "epoch": 0.76, "learning_rate": 2.8660724411365004e-05, "loss": 0.1424, "step": 11857 }, { "epoch": 0.76, "learning_rate": 2.8646173356137516e-05, "loss": 0.1289, "step": 11858 }, { "epoch": 0.76, "learning_rate": 2.8631625378097748e-05, "loss": 0.1529, "step": 11859 }, { "epoch": 0.76, "learning_rate": 2.8617080477872983e-05, "loss": 0.1435, "step": 11860 }, { "epoch": 0.76, "learning_rate": 2.8602538656090562e-05, "loss": 0.1317, "step": 11861 }, { "epoch": 0.76, "learning_rate": 2.858799991337757e-05, "loss": 0.1296, "step": 11862 }, { "epoch": 0.76, "learning_rate": 2.8573464250360972e-05, "loss": 0.1213, "step": 11863 }, { "epoch": 0.76, "learning_rate": 2.8558931667667666e-05, "loss": 0.1435, "step": 11864 }, { "epoch": 0.76, "learning_rate": 2.854440216592438e-05, "loss": 0.1269, "step": 11865 }, { "epoch": 0.76, "learning_rate": 2.8529875745757718e-05, "loss": 0.1476, "step": 11866 }, { "epoch": 0.76, "learning_rate": 2.8515352407794116e-05, "loss": 0.1592, "step": 11867 }, { "epoch": 0.76, "learning_rate": 2.850083215265994e-05, "loss": 0.1469, "step": 11868 }, { "epoch": 0.76, "learning_rate": 2.8486314980981366e-05, "loss": 0.1323, "step": 11869 }, { "epoch": 0.76, "learning_rate": 2.8471800893384437e-05, "loss": 0.1399, "step": 11870 }, { "epoch": 0.76, "learning_rate": 2.8457289890495098e-05, "loss": 0.13, "step": 11871 }, { "epoch": 0.76, "learning_rate": 2.8442781972939146e-05, "loss": 0.1489, "step": 11872 }, { "epoch": 0.76, "learning_rate": 2.8428277141342285e-05, "loss": 0.1413, "step": 11873 }, { "epoch": 0.76, "learning_rate": 2.841377539632999e-05, "loss": 0.1432, "step": 11874 }, { "epoch": 0.76, "learning_rate": 2.8399276738527714e-05, "loss": 0.127, "step": 11875 }, { "epoch": 0.76, "learning_rate": 2.8384781168560693e-05, "loss": 0.1418, "step": 11876 }, { "epoch": 0.76, "learning_rate": 2.8370288687054037e-05, "loss": 0.1456, "step": 11877 }, { "epoch": 0.76, "learning_rate": 2.8355799294632758e-05, "loss": 0.1445, "step": 11878 }, { "epoch": 0.76, "learning_rate": 2.834131299192173e-05, "loss": 0.1473, "step": 11879 }, { "epoch": 0.76, "learning_rate": 2.8326829779545727e-05, "loss": 0.1693, "step": 11880 }, { "epoch": 0.76, "learning_rate": 2.8312349658129266e-05, "loss": 0.1494, "step": 11881 }, { "epoch": 0.76, "learning_rate": 2.8297872628296896e-05, "loss": 0.1281, "step": 11882 }, { "epoch": 0.76, "learning_rate": 2.8283398690672914e-05, "loss": 0.1492, "step": 11883 }, { "epoch": 0.76, "learning_rate": 2.8268927845881477e-05, "loss": 0.1365, "step": 11884 }, { "epoch": 0.76, "learning_rate": 2.8254460094546687e-05, "loss": 0.1248, "step": 11885 }, { "epoch": 0.76, "learning_rate": 2.823999543729249e-05, "loss": 0.1458, "step": 11886 }, { "epoch": 0.76, "learning_rate": 2.8225533874742695e-05, "loss": 0.1449, "step": 11887 }, { "epoch": 0.76, "learning_rate": 2.8211075407520914e-05, "loss": 0.1429, "step": 11888 }, { "epoch": 0.76, "learning_rate": 2.8196620036250755e-05, "loss": 0.1514, "step": 11889 }, { "epoch": 0.76, "learning_rate": 2.818216776155557e-05, "loss": 0.1369, "step": 11890 }, { "epoch": 0.76, "learning_rate": 2.816771858405859e-05, "loss": 0.1332, "step": 11891 }, { "epoch": 0.76, "learning_rate": 2.8153272504383e-05, "loss": 0.1364, "step": 11892 }, { "epoch": 0.76, "learning_rate": 2.8138829523151765e-05, "loss": 0.1371, "step": 11893 }, { "epoch": 0.76, "learning_rate": 2.8124389640987813e-05, "loss": 0.1244, "step": 11894 }, { "epoch": 0.76, "learning_rate": 2.81099528585138e-05, "loss": 0.1472, "step": 11895 }, { "epoch": 0.76, "learning_rate": 2.809551917635238e-05, "loss": 0.1454, "step": 11896 }, { "epoch": 0.76, "learning_rate": 2.8081088595125994e-05, "loss": 0.1278, "step": 11897 }, { "epoch": 0.76, "learning_rate": 2.8066661115456927e-05, "loss": 0.1348, "step": 11898 }, { "epoch": 0.76, "learning_rate": 2.805223673796742e-05, "loss": 0.1347, "step": 11899 }, { "epoch": 0.76, "learning_rate": 2.8037815463279526e-05, "loss": 0.1456, "step": 11900 }, { "epoch": 0.76, "learning_rate": 2.8023397292015207e-05, "loss": 0.1306, "step": 11901 }, { "epoch": 0.76, "learning_rate": 2.8008982224796208e-05, "loss": 0.1418, "step": 11902 }, { "epoch": 0.76, "learning_rate": 2.7994570262244236e-05, "loss": 0.1293, "step": 11903 }, { "epoch": 0.76, "learning_rate": 2.7980161404980754e-05, "loss": 0.1347, "step": 11904 }, { "epoch": 0.76, "learning_rate": 2.79657556536272e-05, "loss": 0.1358, "step": 11905 }, { "epoch": 0.76, "learning_rate": 2.795135300880486e-05, "loss": 0.1422, "step": 11906 }, { "epoch": 0.76, "learning_rate": 2.7936953471134787e-05, "loss": 0.1263, "step": 11907 }, { "epoch": 0.76, "learning_rate": 2.792255704123804e-05, "loss": 0.146, "step": 11908 }, { "epoch": 0.76, "learning_rate": 2.7908163719735415e-05, "loss": 0.152, "step": 11909 }, { "epoch": 0.76, "learning_rate": 2.789377350724769e-05, "loss": 0.1438, "step": 11910 }, { "epoch": 0.76, "learning_rate": 2.7879386404395404e-05, "loss": 0.1299, "step": 11911 }, { "epoch": 0.76, "learning_rate": 2.7865002411799034e-05, "loss": 0.1374, "step": 11912 }, { "epoch": 0.76, "learning_rate": 2.7850621530078936e-05, "loss": 0.1237, "step": 11913 }, { "epoch": 0.76, "learning_rate": 2.783624375985523e-05, "loss": 0.1295, "step": 11914 }, { "epoch": 0.76, "learning_rate": 2.7821869101748034e-05, "loss": 0.1358, "step": 11915 }, { "epoch": 0.76, "learning_rate": 2.7807497556377206e-05, "loss": 0.137, "step": 11916 }, { "epoch": 0.76, "learning_rate": 2.7793129124362583e-05, "loss": 0.1296, "step": 11917 }, { "epoch": 0.76, "learning_rate": 2.777876380632376e-05, "loss": 0.1418, "step": 11918 }, { "epoch": 0.76, "learning_rate": 2.7764401602880285e-05, "loss": 0.1338, "step": 11919 }, { "epoch": 0.76, "learning_rate": 2.7750042514651552e-05, "loss": 0.1398, "step": 11920 }, { "epoch": 0.76, "learning_rate": 2.773568654225678e-05, "loss": 0.1545, "step": 11921 }, { "epoch": 0.76, "learning_rate": 2.772133368631511e-05, "loss": 0.1475, "step": 11922 }, { "epoch": 0.76, "learning_rate": 2.7706983947445465e-05, "loss": 0.1473, "step": 11923 }, { "epoch": 0.76, "learning_rate": 2.769263732626677e-05, "loss": 0.1377, "step": 11924 }, { "epoch": 0.76, "learning_rate": 2.7678293823397638e-05, "loss": 0.1668, "step": 11925 }, { "epoch": 0.76, "learning_rate": 2.7663953439456715e-05, "loss": 0.1376, "step": 11926 }, { "epoch": 0.76, "learning_rate": 2.7649616175062433e-05, "loss": 0.1257, "step": 11927 }, { "epoch": 0.76, "learning_rate": 2.7635282030833063e-05, "loss": 0.1333, "step": 11928 }, { "epoch": 0.76, "learning_rate": 2.762095100738682e-05, "loss": 0.1246, "step": 11929 }, { "epoch": 0.76, "learning_rate": 2.76066231053417e-05, "loss": 0.1535, "step": 11930 }, { "epoch": 0.77, "learning_rate": 2.7592298325315636e-05, "loss": 0.1407, "step": 11931 }, { "epoch": 0.77, "learning_rate": 2.7577976667926364e-05, "loss": 0.1595, "step": 11932 }, { "epoch": 0.77, "learning_rate": 2.7563658133791524e-05, "loss": 0.1415, "step": 11933 }, { "epoch": 0.77, "learning_rate": 2.754934272352866e-05, "loss": 0.1392, "step": 11934 }, { "epoch": 0.77, "learning_rate": 2.7535030437755062e-05, "loss": 0.1281, "step": 11935 }, { "epoch": 0.77, "learning_rate": 2.7520721277088024e-05, "loss": 0.1357, "step": 11936 }, { "epoch": 0.77, "learning_rate": 2.75064152421446e-05, "loss": 0.1494, "step": 11937 }, { "epoch": 0.77, "learning_rate": 2.749211233354174e-05, "loss": 0.129, "step": 11938 }, { "epoch": 0.77, "learning_rate": 2.7477812551896286e-05, "loss": 0.1593, "step": 11939 }, { "epoch": 0.77, "learning_rate": 2.7463515897824922e-05, "loss": 0.1304, "step": 11940 }, { "epoch": 0.77, "learning_rate": 2.7449222371944227e-05, "loss": 0.1383, "step": 11941 }, { "epoch": 0.77, "learning_rate": 2.743493197487058e-05, "loss": 0.1398, "step": 11942 }, { "epoch": 0.77, "learning_rate": 2.742064470722031e-05, "loss": 0.1358, "step": 11943 }, { "epoch": 0.77, "learning_rate": 2.7406360569609545e-05, "loss": 0.1374, "step": 11944 }, { "epoch": 0.77, "learning_rate": 2.7392079562654226e-05, "loss": 0.1456, "step": 11945 }, { "epoch": 0.77, "learning_rate": 2.737780168697037e-05, "loss": 0.1403, "step": 11946 }, { "epoch": 0.77, "learning_rate": 2.7363526943173624e-05, "loss": 0.1456, "step": 11947 }, { "epoch": 0.77, "learning_rate": 2.734925533187964e-05, "loss": 0.1331, "step": 11948 }, { "epoch": 0.77, "learning_rate": 2.733498685370386e-05, "loss": 0.139, "step": 11949 }, { "epoch": 0.77, "learning_rate": 2.732072150926167e-05, "loss": 0.1355, "step": 11950 }, { "epoch": 0.77, "learning_rate": 2.7306459299168206e-05, "loss": 0.1472, "step": 11951 }, { "epoch": 0.77, "learning_rate": 2.7292200224038588e-05, "loss": 0.1329, "step": 11952 }, { "epoch": 0.77, "learning_rate": 2.7277944284487754e-05, "loss": 0.1395, "step": 11953 }, { "epoch": 0.77, "learning_rate": 2.726369148113046e-05, "loss": 0.1331, "step": 11954 }, { "epoch": 0.77, "learning_rate": 2.7249441814581424e-05, "loss": 0.1403, "step": 11955 }, { "epoch": 0.77, "learning_rate": 2.7235195285455107e-05, "loss": 0.1209, "step": 11956 }, { "epoch": 0.77, "learning_rate": 2.7220951894365963e-05, "loss": 0.1323, "step": 11957 }, { "epoch": 0.77, "learning_rate": 2.7206711641928197e-05, "loss": 0.138, "step": 11958 }, { "epoch": 0.77, "learning_rate": 2.7192474528755963e-05, "loss": 0.1347, "step": 11959 }, { "epoch": 0.77, "learning_rate": 2.7178240555463264e-05, "loss": 0.1511, "step": 11960 }, { "epoch": 0.77, "learning_rate": 2.7164009722663897e-05, "loss": 0.1353, "step": 11961 }, { "epoch": 0.77, "learning_rate": 2.7149782030971637e-05, "loss": 0.1268, "step": 11962 }, { "epoch": 0.77, "learning_rate": 2.7135557481000006e-05, "loss": 0.1474, "step": 11963 }, { "epoch": 0.77, "learning_rate": 2.7121336073362512e-05, "loss": 0.1319, "step": 11964 }, { "epoch": 0.77, "learning_rate": 2.710711780867239e-05, "loss": 0.1393, "step": 11965 }, { "epoch": 0.77, "learning_rate": 2.7092902687542865e-05, "loss": 0.1398, "step": 11966 }, { "epoch": 0.77, "learning_rate": 2.7078690710586986e-05, "loss": 0.1276, "step": 11967 }, { "epoch": 0.77, "learning_rate": 2.70644818784176e-05, "loss": 0.1393, "step": 11968 }, { "epoch": 0.77, "learning_rate": 2.705027619164754e-05, "loss": 0.1435, "step": 11969 }, { "epoch": 0.77, "learning_rate": 2.703607365088937e-05, "loss": 0.1367, "step": 11970 }, { "epoch": 0.77, "learning_rate": 2.702187425675564e-05, "loss": 0.156, "step": 11971 }, { "epoch": 0.77, "learning_rate": 2.7007678009858673e-05, "loss": 0.1198, "step": 11972 }, { "epoch": 0.77, "learning_rate": 2.6993484910810697e-05, "loss": 0.1516, "step": 11973 }, { "epoch": 0.77, "learning_rate": 2.6979294960223845e-05, "loss": 0.1466, "step": 11974 }, { "epoch": 0.77, "learning_rate": 2.6965108158709995e-05, "loss": 0.1415, "step": 11975 }, { "epoch": 0.77, "learning_rate": 2.6950924506881038e-05, "loss": 0.1364, "step": 11976 }, { "epoch": 0.77, "learning_rate": 2.6936744005348592e-05, "loss": 0.1426, "step": 11977 }, { "epoch": 0.77, "learning_rate": 2.6922566654724256e-05, "loss": 0.1178, "step": 11978 }, { "epoch": 0.77, "learning_rate": 2.690839245561939e-05, "loss": 0.1349, "step": 11979 }, { "epoch": 0.77, "learning_rate": 2.689422140864528e-05, "loss": 0.1295, "step": 11980 }, { "epoch": 0.77, "learning_rate": 2.6880053514413116e-05, "loss": 0.1359, "step": 11981 }, { "epoch": 0.77, "learning_rate": 2.6865888773533822e-05, "loss": 0.1337, "step": 11982 }, { "epoch": 0.77, "learning_rate": 2.6851727186618315e-05, "loss": 0.1498, "step": 11983 }, { "epoch": 0.77, "learning_rate": 2.6837568754277288e-05, "loss": 0.1397, "step": 11984 }, { "epoch": 0.77, "learning_rate": 2.6823413477121374e-05, "loss": 0.1193, "step": 11985 }, { "epoch": 0.77, "learning_rate": 2.6809261355760984e-05, "loss": 0.1511, "step": 11986 }, { "epoch": 0.77, "learning_rate": 2.6795112390806467e-05, "loss": 0.1358, "step": 11987 }, { "epoch": 0.77, "learning_rate": 2.6780966582868027e-05, "loss": 0.15, "step": 11988 }, { "epoch": 0.77, "learning_rate": 2.676682393255565e-05, "loss": 0.1165, "step": 11989 }, { "epoch": 0.77, "learning_rate": 2.6752684440479337e-05, "loss": 0.1338, "step": 11990 }, { "epoch": 0.77, "learning_rate": 2.6738548107248774e-05, "loss": 0.131, "step": 11991 }, { "epoch": 0.77, "learning_rate": 2.672441493347365e-05, "loss": 0.1428, "step": 11992 }, { "epoch": 0.77, "learning_rate": 2.6710284919763486e-05, "loss": 0.137, "step": 11993 }, { "epoch": 0.77, "learning_rate": 2.6696158066727594e-05, "loss": 0.1332, "step": 11994 }, { "epoch": 0.77, "learning_rate": 2.668203437497526e-05, "loss": 0.1406, "step": 11995 }, { "epoch": 0.77, "learning_rate": 2.6667913845115544e-05, "loss": 0.1436, "step": 11996 }, { "epoch": 0.77, "learning_rate": 2.6653796477757432e-05, "loss": 0.1541, "step": 11997 }, { "epoch": 0.77, "learning_rate": 2.6639682273509692e-05, "loss": 0.1174, "step": 11998 }, { "epoch": 0.77, "learning_rate": 2.6625571232981063e-05, "loss": 0.1351, "step": 11999 }, { "epoch": 0.77, "learning_rate": 2.6611463356780096e-05, "loss": 0.1386, "step": 12000 }, { "epoch": 0.77, "learning_rate": 2.6597358645515168e-05, "loss": 0.153, "step": 12001 }, { "epoch": 0.77, "learning_rate": 2.658325709979459e-05, "loss": 0.1233, "step": 12002 }, { "epoch": 0.77, "learning_rate": 2.6569158720226462e-05, "loss": 0.1365, "step": 12003 }, { "epoch": 0.77, "learning_rate": 2.6555063507418842e-05, "loss": 0.1253, "step": 12004 }, { "epoch": 0.77, "learning_rate": 2.6540971461979535e-05, "loss": 0.1246, "step": 12005 }, { "epoch": 0.77, "learning_rate": 2.65268825845163e-05, "loss": 0.1518, "step": 12006 }, { "epoch": 0.77, "learning_rate": 2.651279687563677e-05, "loss": 0.1307, "step": 12007 }, { "epoch": 0.77, "learning_rate": 2.6498714335948316e-05, "loss": 0.1313, "step": 12008 }, { "epoch": 0.77, "learning_rate": 2.6484634966058353e-05, "loss": 0.116, "step": 12009 }, { "epoch": 0.77, "learning_rate": 2.647055876657397e-05, "loss": 0.1343, "step": 12010 }, { "epoch": 0.77, "learning_rate": 2.645648573810231e-05, "loss": 0.1332, "step": 12011 }, { "epoch": 0.77, "learning_rate": 2.6442415881250192e-05, "loss": 0.1433, "step": 12012 }, { "epoch": 0.77, "learning_rate": 2.6428349196624434e-05, "loss": 0.1401, "step": 12013 }, { "epoch": 0.77, "learning_rate": 2.641428568483171e-05, "loss": 0.1249, "step": 12014 }, { "epoch": 0.77, "learning_rate": 2.6400225346478434e-05, "loss": 0.1404, "step": 12015 }, { "epoch": 0.77, "learning_rate": 2.6386168182171045e-05, "loss": 0.1532, "step": 12016 }, { "epoch": 0.77, "learning_rate": 2.6372114192515718e-05, "loss": 0.131, "step": 12017 }, { "epoch": 0.77, "learning_rate": 2.635806337811858e-05, "loss": 0.1446, "step": 12018 }, { "epoch": 0.77, "learning_rate": 2.6344015739585535e-05, "loss": 0.1581, "step": 12019 }, { "epoch": 0.77, "learning_rate": 2.6329971277522435e-05, "loss": 0.1424, "step": 12020 }, { "epoch": 0.77, "learning_rate": 2.6315929992534982e-05, "loss": 0.1673, "step": 12021 }, { "epoch": 0.77, "learning_rate": 2.6301891885228656e-05, "loss": 0.1385, "step": 12022 }, { "epoch": 0.77, "learning_rate": 2.628785695620891e-05, "loss": 0.1311, "step": 12023 }, { "epoch": 0.77, "learning_rate": 2.6273825206080972e-05, "loss": 0.1506, "step": 12024 }, { "epoch": 0.77, "learning_rate": 2.6259796635450017e-05, "loss": 0.1406, "step": 12025 }, { "epoch": 0.77, "learning_rate": 2.624577124492098e-05, "loss": 0.1114, "step": 12026 }, { "epoch": 0.77, "learning_rate": 2.623174903509875e-05, "loss": 0.1336, "step": 12027 }, { "epoch": 0.77, "learning_rate": 2.621773000658808e-05, "loss": 0.1337, "step": 12028 }, { "epoch": 0.77, "learning_rate": 2.6203714159993475e-05, "loss": 0.126, "step": 12029 }, { "epoch": 0.77, "learning_rate": 2.618970149591946e-05, "loss": 0.1408, "step": 12030 }, { "epoch": 0.77, "learning_rate": 2.6175692014970266e-05, "loss": 0.1233, "step": 12031 }, { "epoch": 0.77, "learning_rate": 2.6161685717750095e-05, "loss": 0.1462, "step": 12032 }, { "epoch": 0.77, "learning_rate": 2.614768260486301e-05, "loss": 0.149, "step": 12033 }, { "epoch": 0.77, "learning_rate": 2.6133682676912852e-05, "loss": 0.1246, "step": 12034 }, { "epoch": 0.77, "learning_rate": 2.6119685934503437e-05, "loss": 0.1458, "step": 12035 }, { "epoch": 0.77, "learning_rate": 2.6105692378238322e-05, "loss": 0.1317, "step": 12036 }, { "epoch": 0.77, "learning_rate": 2.6091702008721043e-05, "loss": 0.1407, "step": 12037 }, { "epoch": 0.77, "learning_rate": 2.607771482655489e-05, "loss": 0.1438, "step": 12038 }, { "epoch": 0.77, "learning_rate": 2.6063730832343103e-05, "loss": 0.1489, "step": 12039 }, { "epoch": 0.77, "learning_rate": 2.604975002668879e-05, "loss": 0.1295, "step": 12040 }, { "epoch": 0.77, "learning_rate": 2.6035772410194813e-05, "loss": 0.1415, "step": 12041 }, { "epoch": 0.77, "learning_rate": 2.6021797983464026e-05, "loss": 0.1422, "step": 12042 }, { "epoch": 0.77, "learning_rate": 2.6007826747099027e-05, "loss": 0.1382, "step": 12043 }, { "epoch": 0.77, "learning_rate": 2.5993858701702413e-05, "loss": 0.1516, "step": 12044 }, { "epoch": 0.77, "learning_rate": 2.597989384787648e-05, "loss": 0.1505, "step": 12045 }, { "epoch": 0.77, "learning_rate": 2.596593218622353e-05, "loss": 0.1346, "step": 12046 }, { "epoch": 0.77, "learning_rate": 2.595197371734568e-05, "loss": 0.1573, "step": 12047 }, { "epoch": 0.77, "learning_rate": 2.5938018441844837e-05, "loss": 0.1522, "step": 12048 }, { "epoch": 0.77, "learning_rate": 2.5924066360322908e-05, "loss": 0.13, "step": 12049 }, { "epoch": 0.77, "learning_rate": 2.5910117473381524e-05, "loss": 0.1227, "step": 12050 }, { "epoch": 0.77, "learning_rate": 2.589617178162229e-05, "loss": 0.1272, "step": 12051 }, { "epoch": 0.77, "learning_rate": 2.5882229285646588e-05, "loss": 0.125, "step": 12052 }, { "epoch": 0.77, "learning_rate": 2.5868289986055693e-05, "loss": 0.1462, "step": 12053 }, { "epoch": 0.77, "learning_rate": 2.585435388345081e-05, "loss": 0.1323, "step": 12054 }, { "epoch": 0.77, "learning_rate": 2.584042097843288e-05, "loss": 0.1385, "step": 12055 }, { "epoch": 0.77, "learning_rate": 2.5826491271602805e-05, "loss": 0.1208, "step": 12056 }, { "epoch": 0.77, "learning_rate": 2.5812564763561275e-05, "loss": 0.1373, "step": 12057 }, { "epoch": 0.77, "learning_rate": 2.5798641454908944e-05, "loss": 0.1274, "step": 12058 }, { "epoch": 0.77, "learning_rate": 2.57847213462462e-05, "loss": 0.1427, "step": 12059 }, { "epoch": 0.77, "learning_rate": 2.577080443817338e-05, "loss": 0.1428, "step": 12060 }, { "epoch": 0.77, "learning_rate": 2.5756890731290696e-05, "loss": 0.1391, "step": 12061 }, { "epoch": 0.77, "learning_rate": 2.5742980226198132e-05, "loss": 0.14, "step": 12062 }, { "epoch": 0.77, "learning_rate": 2.5729072923495646e-05, "loss": 0.123, "step": 12063 }, { "epoch": 0.77, "learning_rate": 2.5715168823782943e-05, "loss": 0.1324, "step": 12064 }, { "epoch": 0.77, "learning_rate": 2.57012679276597e-05, "loss": 0.1433, "step": 12065 }, { "epoch": 0.77, "learning_rate": 2.568737023572535e-05, "loss": 0.145, "step": 12066 }, { "epoch": 0.77, "learning_rate": 2.5673475748579278e-05, "loss": 0.1327, "step": 12067 }, { "epoch": 0.77, "learning_rate": 2.5659584466820718e-05, "loss": 0.1313, "step": 12068 }, { "epoch": 0.77, "learning_rate": 2.5645696391048667e-05, "loss": 0.164, "step": 12069 }, { "epoch": 0.77, "learning_rate": 2.5631811521862146e-05, "loss": 0.1276, "step": 12070 }, { "epoch": 0.77, "learning_rate": 2.561792985985987e-05, "loss": 0.1312, "step": 12071 }, { "epoch": 0.77, "learning_rate": 2.5604051405640572e-05, "loss": 0.1263, "step": 12072 }, { "epoch": 0.77, "learning_rate": 2.559017615980267e-05, "loss": 0.1481, "step": 12073 }, { "epoch": 0.77, "learning_rate": 2.5576304122944663e-05, "loss": 0.1383, "step": 12074 }, { "epoch": 0.77, "learning_rate": 2.556243529566473e-05, "loss": 0.1316, "step": 12075 }, { "epoch": 0.77, "learning_rate": 2.5548569678560964e-05, "loss": 0.1384, "step": 12076 }, { "epoch": 0.77, "learning_rate": 2.5534707272231372e-05, "loss": 0.1388, "step": 12077 }, { "epoch": 0.77, "learning_rate": 2.5520848077273727e-05, "loss": 0.138, "step": 12078 }, { "epoch": 0.77, "learning_rate": 2.550699209428574e-05, "loss": 0.1552, "step": 12079 }, { "epoch": 0.77, "learning_rate": 2.5493139323864977e-05, "loss": 0.1327, "step": 12080 }, { "epoch": 0.77, "learning_rate": 2.547928976660887e-05, "loss": 0.13, "step": 12081 }, { "epoch": 0.77, "learning_rate": 2.5465443423114654e-05, "loss": 0.1261, "step": 12082 }, { "epoch": 0.77, "learning_rate": 2.545160029397945e-05, "loss": 0.1586, "step": 12083 }, { "epoch": 0.77, "learning_rate": 2.543776037980029e-05, "loss": 0.1334, "step": 12084 }, { "epoch": 0.77, "learning_rate": 2.5423923681173988e-05, "loss": 0.1321, "step": 12085 }, { "epoch": 0.77, "learning_rate": 2.54100901986973e-05, "loss": 0.1383, "step": 12086 }, { "epoch": 0.78, "learning_rate": 2.539625993296678e-05, "loss": 0.1601, "step": 12087 }, { "epoch": 0.78, "learning_rate": 2.5382432884578923e-05, "loss": 0.1188, "step": 12088 }, { "epoch": 0.78, "learning_rate": 2.5368609054129987e-05, "loss": 0.1289, "step": 12089 }, { "epoch": 0.78, "learning_rate": 2.5354788442216105e-05, "loss": 0.1478, "step": 12090 }, { "epoch": 0.78, "learning_rate": 2.534097104943336e-05, "loss": 0.1467, "step": 12091 }, { "epoch": 0.78, "learning_rate": 2.532715687637759e-05, "loss": 0.128, "step": 12092 }, { "epoch": 0.78, "learning_rate": 2.531334592364457e-05, "loss": 0.1245, "step": 12093 }, { "epoch": 0.78, "learning_rate": 2.5299538191829918e-05, "loss": 0.1344, "step": 12094 }, { "epoch": 0.78, "learning_rate": 2.5285733681529067e-05, "loss": 0.1388, "step": 12095 }, { "epoch": 0.78, "learning_rate": 2.5271932393337393e-05, "loss": 0.147, "step": 12096 }, { "epoch": 0.78, "learning_rate": 2.5258134327850026e-05, "loss": 0.1297, "step": 12097 }, { "epoch": 0.78, "learning_rate": 2.5244339485662093e-05, "loss": 0.1447, "step": 12098 }, { "epoch": 0.78, "learning_rate": 2.5230547867368426e-05, "loss": 0.1375, "step": 12099 }, { "epoch": 0.78, "learning_rate": 2.5216759473563846e-05, "loss": 0.1307, "step": 12100 }, { "epoch": 0.78, "learning_rate": 2.520297430484301e-05, "loss": 0.1358, "step": 12101 }, { "epoch": 0.78, "learning_rate": 2.518919236180035e-05, "loss": 0.1381, "step": 12102 }, { "epoch": 0.78, "learning_rate": 2.517541364503029e-05, "loss": 0.1331, "step": 12103 }, { "epoch": 0.78, "learning_rate": 2.5161638155126987e-05, "loss": 0.1539, "step": 12104 }, { "epoch": 0.78, "learning_rate": 2.5147865892684575e-05, "loss": 0.1379, "step": 12105 }, { "epoch": 0.78, "learning_rate": 2.513409685829694e-05, "loss": 0.1377, "step": 12106 }, { "epoch": 0.78, "learning_rate": 2.5120331052557912e-05, "loss": 0.1325, "step": 12107 }, { "epoch": 0.78, "learning_rate": 2.5106568476061167e-05, "loss": 0.1358, "step": 12108 }, { "epoch": 0.78, "learning_rate": 2.5092809129400186e-05, "loss": 0.1213, "step": 12109 }, { "epoch": 0.78, "learning_rate": 2.5079053013168398e-05, "loss": 0.136, "step": 12110 }, { "epoch": 0.78, "learning_rate": 2.506530012795899e-05, "loss": 0.1233, "step": 12111 }, { "epoch": 0.78, "learning_rate": 2.5051550474365128e-05, "loss": 0.1521, "step": 12112 }, { "epoch": 0.78, "learning_rate": 2.5037804052979685e-05, "loss": 0.1537, "step": 12113 }, { "epoch": 0.78, "learning_rate": 2.5024060864395605e-05, "loss": 0.1433, "step": 12114 }, { "epoch": 0.78, "learning_rate": 2.501032090920551e-05, "loss": 0.1397, "step": 12115 }, { "epoch": 0.78, "learning_rate": 2.4996584188001914e-05, "loss": 0.1335, "step": 12116 }, { "epoch": 0.78, "learning_rate": 2.49828507013773e-05, "loss": 0.1561, "step": 12117 }, { "epoch": 0.78, "learning_rate": 2.4969120449923857e-05, "loss": 0.1489, "step": 12118 }, { "epoch": 0.78, "learning_rate": 2.4955393434233754e-05, "loss": 0.1379, "step": 12119 }, { "epoch": 0.78, "learning_rate": 2.494166965489897e-05, "loss": 0.1438, "step": 12120 }, { "epoch": 0.78, "learning_rate": 2.4927949112511396e-05, "loss": 0.1365, "step": 12121 }, { "epoch": 0.78, "learning_rate": 2.491423180766269e-05, "loss": 0.14, "step": 12122 }, { "epoch": 0.78, "learning_rate": 2.4900517740944408e-05, "loss": 0.1265, "step": 12123 }, { "epoch": 0.78, "learning_rate": 2.4886806912948035e-05, "loss": 0.1325, "step": 12124 }, { "epoch": 0.78, "learning_rate": 2.4873099324264803e-05, "loss": 0.1479, "step": 12125 }, { "epoch": 0.78, "learning_rate": 2.4859394975485885e-05, "loss": 0.134, "step": 12126 }, { "epoch": 0.78, "learning_rate": 2.4845693867202302e-05, "loss": 0.1406, "step": 12127 }, { "epoch": 0.78, "learning_rate": 2.483199600000494e-05, "loss": 0.1449, "step": 12128 }, { "epoch": 0.78, "learning_rate": 2.481830137448451e-05, "loss": 0.1167, "step": 12129 }, { "epoch": 0.78, "learning_rate": 2.4804609991231576e-05, "loss": 0.1274, "step": 12130 }, { "epoch": 0.78, "learning_rate": 2.4790921850836644e-05, "loss": 0.149, "step": 12131 }, { "epoch": 0.78, "learning_rate": 2.477723695388997e-05, "loss": 0.1314, "step": 12132 }, { "epoch": 0.78, "learning_rate": 2.4763555300981745e-05, "loss": 0.1392, "step": 12133 }, { "epoch": 0.78, "learning_rate": 2.474987689270202e-05, "loss": 0.1359, "step": 12134 }, { "epoch": 0.78, "learning_rate": 2.4736201729640694e-05, "loss": 0.1443, "step": 12135 }, { "epoch": 0.78, "learning_rate": 2.4722529812387496e-05, "loss": 0.1385, "step": 12136 }, { "epoch": 0.78, "learning_rate": 2.4708861141532012e-05, "loss": 0.1336, "step": 12137 }, { "epoch": 0.78, "learning_rate": 2.4695195717663776e-05, "loss": 0.1438, "step": 12138 }, { "epoch": 0.78, "learning_rate": 2.468153354137206e-05, "loss": 0.149, "step": 12139 }, { "epoch": 0.78, "learning_rate": 2.4667874613246077e-05, "loss": 0.1253, "step": 12140 }, { "epoch": 0.78, "learning_rate": 2.4654218933874883e-05, "loss": 0.1284, "step": 12141 }, { "epoch": 0.78, "learning_rate": 2.4640566503847418e-05, "loss": 0.1355, "step": 12142 }, { "epoch": 0.78, "learning_rate": 2.4626917323752418e-05, "loss": 0.1385, "step": 12143 }, { "epoch": 0.78, "learning_rate": 2.4613271394178506e-05, "loss": 0.1428, "step": 12144 }, { "epoch": 0.78, "learning_rate": 2.4599628715714218e-05, "loss": 0.1491, "step": 12145 }, { "epoch": 0.78, "learning_rate": 2.458598928894783e-05, "loss": 0.1269, "step": 12146 }, { "epoch": 0.78, "learning_rate": 2.4572353114467616e-05, "loss": 0.1351, "step": 12147 }, { "epoch": 0.78, "learning_rate": 2.4558720192861617e-05, "loss": 0.1287, "step": 12148 }, { "epoch": 0.78, "learning_rate": 2.4545090524717808e-05, "loss": 0.1403, "step": 12149 }, { "epoch": 0.78, "learning_rate": 2.4531464110623935e-05, "loss": 0.1361, "step": 12150 }, { "epoch": 0.78, "learning_rate": 2.451784095116765e-05, "loss": 0.1369, "step": 12151 }, { "epoch": 0.78, "learning_rate": 2.4504221046936483e-05, "loss": 0.1334, "step": 12152 }, { "epoch": 0.78, "learning_rate": 2.4490604398517756e-05, "loss": 0.1287, "step": 12153 }, { "epoch": 0.78, "learning_rate": 2.447699100649874e-05, "loss": 0.1468, "step": 12154 }, { "epoch": 0.78, "learning_rate": 2.4463380871466503e-05, "loss": 0.1168, "step": 12155 }, { "epoch": 0.78, "learning_rate": 2.444977399400804e-05, "loss": 0.1435, "step": 12156 }, { "epoch": 0.78, "learning_rate": 2.443617037471012e-05, "loss": 0.1224, "step": 12157 }, { "epoch": 0.78, "learning_rate": 2.4422570014159372e-05, "loss": 0.1276, "step": 12158 }, { "epoch": 0.78, "learning_rate": 2.4408972912942397e-05, "loss": 0.1445, "step": 12159 }, { "epoch": 0.78, "learning_rate": 2.4395379071645485e-05, "loss": 0.1275, "step": 12160 }, { "epoch": 0.78, "learning_rate": 2.4381788490854995e-05, "loss": 0.1558, "step": 12161 }, { "epoch": 0.78, "learning_rate": 2.4368201171156933e-05, "loss": 0.1445, "step": 12162 }, { "epoch": 0.78, "learning_rate": 2.4354617113137346e-05, "loss": 0.1398, "step": 12163 }, { "epoch": 0.78, "learning_rate": 2.4341036317382017e-05, "loss": 0.1251, "step": 12164 }, { "epoch": 0.78, "learning_rate": 2.4327458784476588e-05, "loss": 0.1411, "step": 12165 }, { "epoch": 0.78, "learning_rate": 2.4313884515006648e-05, "loss": 0.1528, "step": 12166 }, { "epoch": 0.78, "learning_rate": 2.4300313509557583e-05, "loss": 0.1282, "step": 12167 }, { "epoch": 0.78, "learning_rate": 2.428674576871468e-05, "loss": 0.1238, "step": 12168 }, { "epoch": 0.78, "learning_rate": 2.427318129306304e-05, "loss": 0.1471, "step": 12169 }, { "epoch": 0.78, "learning_rate": 2.4259620083187606e-05, "loss": 0.1603, "step": 12170 }, { "epoch": 0.78, "learning_rate": 2.4246062139673274e-05, "loss": 0.1398, "step": 12171 }, { "epoch": 0.78, "learning_rate": 2.423250746310469e-05, "loss": 0.1236, "step": 12172 }, { "epoch": 0.78, "learning_rate": 2.4218956054066434e-05, "loss": 0.1307, "step": 12173 }, { "epoch": 0.78, "learning_rate": 2.420540791314291e-05, "loss": 0.1133, "step": 12174 }, { "epoch": 0.78, "learning_rate": 2.4191863040918437e-05, "loss": 0.1274, "step": 12175 }, { "epoch": 0.78, "learning_rate": 2.4178321437977104e-05, "loss": 0.1373, "step": 12176 }, { "epoch": 0.78, "learning_rate": 2.416478310490289e-05, "loss": 0.1343, "step": 12177 }, { "epoch": 0.78, "learning_rate": 2.4151248042279695e-05, "loss": 0.1432, "step": 12178 }, { "epoch": 0.78, "learning_rate": 2.4137716250691166e-05, "loss": 0.1413, "step": 12179 }, { "epoch": 0.78, "learning_rate": 2.4124187730720917e-05, "loss": 0.1242, "step": 12180 }, { "epoch": 0.78, "learning_rate": 2.411066248295235e-05, "loss": 0.1411, "step": 12181 }, { "epoch": 0.78, "learning_rate": 2.409714050796881e-05, "loss": 0.1363, "step": 12182 }, { "epoch": 0.78, "learning_rate": 2.4083621806353385e-05, "loss": 0.1283, "step": 12183 }, { "epoch": 0.78, "learning_rate": 2.4070106378689062e-05, "loss": 0.1373, "step": 12184 }, { "epoch": 0.78, "learning_rate": 2.405659422555877e-05, "loss": 0.1372, "step": 12185 }, { "epoch": 0.78, "learning_rate": 2.404308534754517e-05, "loss": 0.129, "step": 12186 }, { "epoch": 0.78, "learning_rate": 2.4029579745230866e-05, "loss": 0.1414, "step": 12187 }, { "epoch": 0.78, "learning_rate": 2.4016077419198302e-05, "loss": 0.1327, "step": 12188 }, { "epoch": 0.78, "learning_rate": 2.4002578370029805e-05, "loss": 0.1331, "step": 12189 }, { "epoch": 0.78, "learning_rate": 2.398908259830749e-05, "loss": 0.1349, "step": 12190 }, { "epoch": 0.78, "learning_rate": 2.3975590104613354e-05, "loss": 0.1408, "step": 12191 }, { "epoch": 0.78, "learning_rate": 2.3962100889529338e-05, "loss": 0.1446, "step": 12192 }, { "epoch": 0.78, "learning_rate": 2.3948614953637104e-05, "loss": 0.1211, "step": 12193 }, { "epoch": 0.78, "learning_rate": 2.393513229751827e-05, "loss": 0.1461, "step": 12194 }, { "epoch": 0.78, "learning_rate": 2.392165292175429e-05, "loss": 0.1548, "step": 12195 }, { "epoch": 0.78, "learning_rate": 2.390817682692651e-05, "loss": 0.1408, "step": 12196 }, { "epoch": 0.78, "learning_rate": 2.3894704013616052e-05, "loss": 0.1125, "step": 12197 }, { "epoch": 0.78, "learning_rate": 2.3881234482403926e-05, "loss": 0.1315, "step": 12198 }, { "epoch": 0.78, "learning_rate": 2.386776823387107e-05, "loss": 0.1258, "step": 12199 }, { "epoch": 0.78, "learning_rate": 2.3854305268598132e-05, "loss": 0.1294, "step": 12200 }, { "epoch": 0.78, "learning_rate": 2.3840845587165838e-05, "loss": 0.1481, "step": 12201 }, { "epoch": 0.78, "learning_rate": 2.3827389190154548e-05, "loss": 0.1206, "step": 12202 }, { "epoch": 0.78, "learning_rate": 2.381393607814465e-05, "loss": 0.1389, "step": 12203 }, { "epoch": 0.78, "learning_rate": 2.3800486251716293e-05, "loss": 0.1211, "step": 12204 }, { "epoch": 0.78, "learning_rate": 2.3787039711449465e-05, "loss": 0.1207, "step": 12205 }, { "epoch": 0.78, "learning_rate": 2.3773596457924095e-05, "loss": 0.1398, "step": 12206 }, { "epoch": 0.78, "learning_rate": 2.3760156491719942e-05, "loss": 0.137, "step": 12207 }, { "epoch": 0.78, "learning_rate": 2.3746719813416628e-05, "loss": 0.1516, "step": 12208 }, { "epoch": 0.78, "learning_rate": 2.373328642359358e-05, "loss": 0.16, "step": 12209 }, { "epoch": 0.78, "learning_rate": 2.3719856322830158e-05, "loss": 0.1296, "step": 12210 }, { "epoch": 0.78, "learning_rate": 2.3706429511705543e-05, "loss": 0.125, "step": 12211 }, { "epoch": 0.78, "learning_rate": 2.3693005990798733e-05, "loss": 0.1279, "step": 12212 }, { "epoch": 0.78, "learning_rate": 2.367958576068865e-05, "loss": 0.1506, "step": 12213 }, { "epoch": 0.78, "learning_rate": 2.366616882195407e-05, "loss": 0.1378, "step": 12214 }, { "epoch": 0.78, "learning_rate": 2.3652755175173625e-05, "loss": 0.1456, "step": 12215 }, { "epoch": 0.78, "learning_rate": 2.3639344820925724e-05, "loss": 0.1353, "step": 12216 }, { "epoch": 0.78, "learning_rate": 2.3625937759788775e-05, "loss": 0.1242, "step": 12217 }, { "epoch": 0.78, "learning_rate": 2.3612533992340922e-05, "loss": 0.157, "step": 12218 }, { "epoch": 0.78, "learning_rate": 2.3599133519160188e-05, "loss": 0.1359, "step": 12219 }, { "epoch": 0.78, "learning_rate": 2.3585736340824515e-05, "loss": 0.1325, "step": 12220 }, { "epoch": 0.78, "learning_rate": 2.3572342457911666e-05, "loss": 0.1254, "step": 12221 }, { "epoch": 0.78, "learning_rate": 2.355895187099928e-05, "loss": 0.1451, "step": 12222 }, { "epoch": 0.78, "learning_rate": 2.354556458066478e-05, "loss": 0.1376, "step": 12223 }, { "epoch": 0.78, "learning_rate": 2.3532180587485576e-05, "loss": 0.1296, "step": 12224 }, { "epoch": 0.78, "learning_rate": 2.3518799892038823e-05, "loss": 0.1405, "step": 12225 }, { "epoch": 0.78, "learning_rate": 2.350542249490154e-05, "loss": 0.13, "step": 12226 }, { "epoch": 0.78, "learning_rate": 2.349204839665068e-05, "loss": 0.1332, "step": 12227 }, { "epoch": 0.78, "learning_rate": 2.3478677597862996e-05, "loss": 0.1333, "step": 12228 }, { "epoch": 0.78, "learning_rate": 2.3465310099115145e-05, "loss": 0.1389, "step": 12229 }, { "epoch": 0.78, "learning_rate": 2.3451945900983564e-05, "loss": 0.1412, "step": 12230 }, { "epoch": 0.78, "learning_rate": 2.3438585004044645e-05, "loss": 0.1225, "step": 12231 }, { "epoch": 0.78, "learning_rate": 2.342522740887456e-05, "loss": 0.1472, "step": 12232 }, { "epoch": 0.78, "learning_rate": 2.3411873116049333e-05, "loss": 0.14, "step": 12233 }, { "epoch": 0.78, "learning_rate": 2.3398522126144907e-05, "loss": 0.1183, "step": 12234 }, { "epoch": 0.78, "learning_rate": 2.3385174439737046e-05, "loss": 0.1693, "step": 12235 }, { "epoch": 0.78, "learning_rate": 2.3371830057401435e-05, "loss": 0.1325, "step": 12236 }, { "epoch": 0.78, "learning_rate": 2.3358488979713466e-05, "loss": 0.1467, "step": 12237 }, { "epoch": 0.78, "learning_rate": 2.3345151207248573e-05, "loss": 0.1477, "step": 12238 }, { "epoch": 0.78, "learning_rate": 2.33318167405819e-05, "loss": 0.1448, "step": 12239 }, { "epoch": 0.78, "learning_rate": 2.3318485580288506e-05, "loss": 0.1265, "step": 12240 }, { "epoch": 0.78, "learning_rate": 2.3305157726943327e-05, "loss": 0.1351, "step": 12241 }, { "epoch": 0.78, "learning_rate": 2.3291833181121116e-05, "loss": 0.1452, "step": 12242 }, { "epoch": 0.79, "learning_rate": 2.3278511943396563e-05, "loss": 0.1528, "step": 12243 }, { "epoch": 0.79, "learning_rate": 2.326519401434407e-05, "loss": 0.1412, "step": 12244 }, { "epoch": 0.79, "learning_rate": 2.3251879394538066e-05, "loss": 0.1207, "step": 12245 }, { "epoch": 0.79, "learning_rate": 2.3238568084552683e-05, "loss": 0.1285, "step": 12246 }, { "epoch": 0.79, "learning_rate": 2.322526008496201e-05, "loss": 0.1314, "step": 12247 }, { "epoch": 0.79, "learning_rate": 2.3211955396340002e-05, "loss": 0.1256, "step": 12248 }, { "epoch": 0.79, "learning_rate": 2.319865401926037e-05, "loss": 0.1511, "step": 12249 }, { "epoch": 0.79, "learning_rate": 2.3185355954296795e-05, "loss": 0.1459, "step": 12250 }, { "epoch": 0.79, "learning_rate": 2.3172061202022753e-05, "loss": 0.1228, "step": 12251 }, { "epoch": 0.79, "learning_rate": 2.3158769763011546e-05, "loss": 0.1292, "step": 12252 }, { "epoch": 0.79, "learning_rate": 2.314548163783643e-05, "loss": 0.1252, "step": 12253 }, { "epoch": 0.79, "learning_rate": 2.313219682707044e-05, "loss": 0.1353, "step": 12254 }, { "epoch": 0.79, "learning_rate": 2.311891533128653e-05, "loss": 0.1504, "step": 12255 }, { "epoch": 0.79, "learning_rate": 2.3105637151057426e-05, "loss": 0.1226, "step": 12256 }, { "epoch": 0.79, "learning_rate": 2.3092362286955805e-05, "loss": 0.1181, "step": 12257 }, { "epoch": 0.79, "learning_rate": 2.3079090739554133e-05, "loss": 0.1396, "step": 12258 }, { "epoch": 0.79, "learning_rate": 2.3065822509424716e-05, "loss": 0.134, "step": 12259 }, { "epoch": 0.79, "learning_rate": 2.3052557597139813e-05, "loss": 0.1262, "step": 12260 }, { "epoch": 0.79, "learning_rate": 2.303929600327146e-05, "loss": 0.1322, "step": 12261 }, { "epoch": 0.79, "learning_rate": 2.3026037728391604e-05, "loss": 0.1335, "step": 12262 }, { "epoch": 0.79, "learning_rate": 2.301278277307196e-05, "loss": 0.1405, "step": 12263 }, { "epoch": 0.79, "learning_rate": 2.299953113788422e-05, "loss": 0.1248, "step": 12264 }, { "epoch": 0.79, "learning_rate": 2.2986282823399852e-05, "loss": 0.1485, "step": 12265 }, { "epoch": 0.79, "learning_rate": 2.2973037830190158e-05, "loss": 0.1423, "step": 12266 }, { "epoch": 0.79, "learning_rate": 2.2959796158826364e-05, "loss": 0.1323, "step": 12267 }, { "epoch": 0.79, "learning_rate": 2.294655780987953e-05, "loss": 0.1601, "step": 12268 }, { "epoch": 0.79, "learning_rate": 2.2933322783920607e-05, "loss": 0.1252, "step": 12269 }, { "epoch": 0.79, "learning_rate": 2.29200910815203e-05, "loss": 0.1398, "step": 12270 }, { "epoch": 0.79, "learning_rate": 2.290686270324929e-05, "loss": 0.1377, "step": 12271 }, { "epoch": 0.79, "learning_rate": 2.2893637649678047e-05, "loss": 0.1392, "step": 12272 }, { "epoch": 0.79, "learning_rate": 2.2880415921376863e-05, "loss": 0.1315, "step": 12273 }, { "epoch": 0.79, "learning_rate": 2.2867197518915984e-05, "loss": 0.1384, "step": 12274 }, { "epoch": 0.79, "learning_rate": 2.2853982442865453e-05, "loss": 0.1283, "step": 12275 }, { "epoch": 0.79, "learning_rate": 2.2840770693795198e-05, "loss": 0.1384, "step": 12276 }, { "epoch": 0.79, "learning_rate": 2.282756227227495e-05, "loss": 0.1251, "step": 12277 }, { "epoch": 0.79, "learning_rate": 2.281435717887438e-05, "loss": 0.1334, "step": 12278 }, { "epoch": 0.79, "learning_rate": 2.2801155414162934e-05, "loss": 0.1402, "step": 12279 }, { "epoch": 0.79, "learning_rate": 2.2787956978709923e-05, "loss": 0.1356, "step": 12280 }, { "epoch": 0.79, "learning_rate": 2.277476187308458e-05, "loss": 0.1399, "step": 12281 }, { "epoch": 0.79, "learning_rate": 2.2761570097855935e-05, "loss": 0.1185, "step": 12282 }, { "epoch": 0.79, "learning_rate": 2.274838165359292e-05, "loss": 0.1474, "step": 12283 }, { "epoch": 0.79, "learning_rate": 2.2735196540864267e-05, "loss": 0.1376, "step": 12284 }, { "epoch": 0.79, "learning_rate": 2.2722014760238618e-05, "loss": 0.1264, "step": 12285 }, { "epoch": 0.79, "learning_rate": 2.2708836312284443e-05, "loss": 0.1296, "step": 12286 }, { "epoch": 0.79, "learning_rate": 2.2695661197569995e-05, "loss": 0.1365, "step": 12287 }, { "epoch": 0.79, "learning_rate": 2.26824894166636e-05, "loss": 0.1348, "step": 12288 }, { "epoch": 0.79, "learning_rate": 2.266932097013319e-05, "loss": 0.1198, "step": 12289 }, { "epoch": 0.79, "learning_rate": 2.2656155858546746e-05, "loss": 0.1439, "step": 12290 }, { "epoch": 0.79, "learning_rate": 2.2642994082471946e-05, "loss": 0.1467, "step": 12291 }, { "epoch": 0.79, "learning_rate": 2.262983564247646e-05, "loss": 0.1254, "step": 12292 }, { "epoch": 0.79, "learning_rate": 2.261668053912771e-05, "loss": 0.1415, "step": 12293 }, { "epoch": 0.79, "learning_rate": 2.2603528772993043e-05, "loss": 0.1472, "step": 12294 }, { "epoch": 0.79, "learning_rate": 2.2590380344639663e-05, "loss": 0.142, "step": 12295 }, { "epoch": 0.79, "learning_rate": 2.2577235254634554e-05, "loss": 0.1427, "step": 12296 }, { "epoch": 0.79, "learning_rate": 2.256409350354467e-05, "loss": 0.1245, "step": 12297 }, { "epoch": 0.79, "learning_rate": 2.2550955091936687e-05, "loss": 0.1365, "step": 12298 }, { "epoch": 0.79, "learning_rate": 2.2537820020377276e-05, "loss": 0.1282, "step": 12299 }, { "epoch": 0.79, "learning_rate": 2.252468828943285e-05, "loss": 0.1497, "step": 12300 }, { "epoch": 0.79, "learning_rate": 2.251155989966973e-05, "loss": 0.1257, "step": 12301 }, { "epoch": 0.79, "learning_rate": 2.2498434851654126e-05, "loss": 0.1217, "step": 12302 }, { "epoch": 0.79, "learning_rate": 2.2485313145952025e-05, "loss": 0.1311, "step": 12303 }, { "epoch": 0.79, "learning_rate": 2.247219478312934e-05, "loss": 0.147, "step": 12304 }, { "epoch": 0.79, "learning_rate": 2.2459079763751766e-05, "loss": 0.127, "step": 12305 }, { "epoch": 0.79, "learning_rate": 2.244596808838496e-05, "loss": 0.1395, "step": 12306 }, { "epoch": 0.79, "learning_rate": 2.2432859757594316e-05, "loss": 0.1295, "step": 12307 }, { "epoch": 0.79, "learning_rate": 2.2419754771945156e-05, "loss": 0.1276, "step": 12308 }, { "epoch": 0.79, "learning_rate": 2.2406653132002687e-05, "loss": 0.1549, "step": 12309 }, { "epoch": 0.79, "learning_rate": 2.2393554838331855e-05, "loss": 0.1323, "step": 12310 }, { "epoch": 0.79, "learning_rate": 2.2380459891497608e-05, "loss": 0.1378, "step": 12311 }, { "epoch": 0.79, "learning_rate": 2.2367368292064605e-05, "loss": 0.1433, "step": 12312 }, { "epoch": 0.79, "learning_rate": 2.2354280040597485e-05, "loss": 0.1477, "step": 12313 }, { "epoch": 0.79, "learning_rate": 2.2341195137660643e-05, "loss": 0.1422, "step": 12314 }, { "epoch": 0.79, "learning_rate": 2.23281135838184e-05, "loss": 0.142, "step": 12315 }, { "epoch": 0.79, "learning_rate": 2.231503537963494e-05, "loss": 0.1207, "step": 12316 }, { "epoch": 0.79, "learning_rate": 2.2301960525674205e-05, "loss": 0.1365, "step": 12317 }, { "epoch": 0.79, "learning_rate": 2.2288889022500125e-05, "loss": 0.1307, "step": 12318 }, { "epoch": 0.79, "learning_rate": 2.2275820870676346e-05, "loss": 0.1456, "step": 12319 }, { "epoch": 0.79, "learning_rate": 2.2262756070766512e-05, "loss": 0.1288, "step": 12320 }, { "epoch": 0.79, "learning_rate": 2.2249694623333994e-05, "loss": 0.1413, "step": 12321 }, { "epoch": 0.79, "learning_rate": 2.2236636528942102e-05, "loss": 0.1415, "step": 12322 }, { "epoch": 0.79, "learning_rate": 2.2223581788154013e-05, "loss": 0.1299, "step": 12323 }, { "epoch": 0.79, "learning_rate": 2.2210530401532647e-05, "loss": 0.1281, "step": 12324 }, { "epoch": 0.79, "learning_rate": 2.2197482369640932e-05, "loss": 0.1465, "step": 12325 }, { "epoch": 0.79, "learning_rate": 2.2184437693041505e-05, "loss": 0.1267, "step": 12326 }, { "epoch": 0.79, "learning_rate": 2.2171396372296993e-05, "loss": 0.1322, "step": 12327 }, { "epoch": 0.79, "learning_rate": 2.215835840796975e-05, "loss": 0.1292, "step": 12328 }, { "epoch": 0.79, "learning_rate": 2.2145323800622063e-05, "loss": 0.1473, "step": 12329 }, { "epoch": 0.79, "learning_rate": 2.2132292550816113e-05, "loss": 0.144, "step": 12330 }, { "epoch": 0.79, "learning_rate": 2.2119264659113824e-05, "loss": 0.1375, "step": 12331 }, { "epoch": 0.79, "learning_rate": 2.210624012607707e-05, "loss": 0.1411, "step": 12332 }, { "epoch": 0.79, "learning_rate": 2.2093218952267504e-05, "loss": 0.1474, "step": 12333 }, { "epoch": 0.79, "learning_rate": 2.2080201138246692e-05, "loss": 0.133, "step": 12334 }, { "epoch": 0.79, "learning_rate": 2.206718668457607e-05, "loss": 0.1342, "step": 12335 }, { "epoch": 0.79, "learning_rate": 2.205417559181685e-05, "loss": 0.1338, "step": 12336 }, { "epoch": 0.79, "learning_rate": 2.2041167860530175e-05, "loss": 0.1232, "step": 12337 }, { "epoch": 0.79, "learning_rate": 2.2028163491276987e-05, "loss": 0.1514, "step": 12338 }, { "epoch": 0.79, "learning_rate": 2.2015162484618146e-05, "loss": 0.1525, "step": 12339 }, { "epoch": 0.79, "learning_rate": 2.200216484111428e-05, "loss": 0.1557, "step": 12340 }, { "epoch": 0.79, "learning_rate": 2.1989170561325957e-05, "loss": 0.1347, "step": 12341 }, { "epoch": 0.79, "learning_rate": 2.1976179645813576e-05, "loss": 0.1531, "step": 12342 }, { "epoch": 0.79, "learning_rate": 2.1963192095137343e-05, "loss": 0.1389, "step": 12343 }, { "epoch": 0.79, "learning_rate": 2.195020790985739e-05, "loss": 0.1343, "step": 12344 }, { "epoch": 0.79, "learning_rate": 2.193722709053364e-05, "loss": 0.1253, "step": 12345 }, { "epoch": 0.79, "learning_rate": 2.1924249637725936e-05, "loss": 0.1285, "step": 12346 }, { "epoch": 0.79, "learning_rate": 2.1911275551993892e-05, "loss": 0.1364, "step": 12347 }, { "epoch": 0.79, "learning_rate": 2.1898304833897066e-05, "loss": 0.1649, "step": 12348 }, { "epoch": 0.79, "learning_rate": 2.188533748399483e-05, "loss": 0.1358, "step": 12349 }, { "epoch": 0.79, "learning_rate": 2.1872373502846378e-05, "loss": 0.1391, "step": 12350 }, { "epoch": 0.79, "learning_rate": 2.1859412891010832e-05, "loss": 0.1418, "step": 12351 }, { "epoch": 0.79, "learning_rate": 2.1846455649047092e-05, "loss": 0.1391, "step": 12352 }, { "epoch": 0.79, "learning_rate": 2.1833501777513975e-05, "loss": 0.1304, "step": 12353 }, { "epoch": 0.79, "learning_rate": 2.1820551276970102e-05, "loss": 0.1385, "step": 12354 }, { "epoch": 0.79, "learning_rate": 2.180760414797399e-05, "loss": 0.146, "step": 12355 }, { "epoch": 0.79, "learning_rate": 2.179466039108401e-05, "loss": 0.1349, "step": 12356 }, { "epoch": 0.79, "learning_rate": 2.178172000685833e-05, "loss": 0.1369, "step": 12357 }, { "epoch": 0.79, "learning_rate": 2.1768782995855052e-05, "loss": 0.1355, "step": 12358 }, { "epoch": 0.79, "learning_rate": 2.1755849358632063e-05, "loss": 0.1468, "step": 12359 }, { "epoch": 0.79, "learning_rate": 2.174291909574717e-05, "loss": 0.1556, "step": 12360 }, { "epoch": 0.79, "learning_rate": 2.1729992207757966e-05, "loss": 0.1397, "step": 12361 }, { "epoch": 0.79, "learning_rate": 2.1717068695221944e-05, "loss": 0.1325, "step": 12362 }, { "epoch": 0.79, "learning_rate": 2.170414855869647e-05, "loss": 0.1084, "step": 12363 }, { "epoch": 0.79, "learning_rate": 2.1691231798738677e-05, "loss": 0.1492, "step": 12364 }, { "epoch": 0.79, "learning_rate": 2.1678318415905674e-05, "loss": 0.13, "step": 12365 }, { "epoch": 0.79, "learning_rate": 2.1665408410754306e-05, "loss": 0.1298, "step": 12366 }, { "epoch": 0.79, "learning_rate": 2.165250178384137e-05, "loss": 0.1401, "step": 12367 }, { "epoch": 0.79, "learning_rate": 2.1639598535723426e-05, "loss": 0.1357, "step": 12368 }, { "epoch": 0.79, "learning_rate": 2.1626698666956967e-05, "loss": 0.1314, "step": 12369 }, { "epoch": 0.79, "learning_rate": 2.1613802178098342e-05, "loss": 0.1254, "step": 12370 }, { "epoch": 0.79, "learning_rate": 2.1600909069703645e-05, "loss": 0.1239, "step": 12371 }, { "epoch": 0.79, "learning_rate": 2.1588019342328968e-05, "loss": 0.1523, "step": 12372 }, { "epoch": 0.79, "learning_rate": 2.1575132996530146e-05, "loss": 0.1283, "step": 12373 }, { "epoch": 0.79, "learning_rate": 2.156225003286293e-05, "loss": 0.1377, "step": 12374 }, { "epoch": 0.79, "learning_rate": 2.1549370451882935e-05, "loss": 0.1261, "step": 12375 }, { "epoch": 0.79, "learning_rate": 2.1536494254145544e-05, "loss": 0.142, "step": 12376 }, { "epoch": 0.79, "learning_rate": 2.1523621440206122e-05, "loss": 0.1441, "step": 12377 }, { "epoch": 0.79, "learning_rate": 2.151075201061975e-05, "loss": 0.1363, "step": 12378 }, { "epoch": 0.79, "learning_rate": 2.1497885965941488e-05, "loss": 0.1557, "step": 12379 }, { "epoch": 0.79, "learning_rate": 2.148502330672614e-05, "loss": 0.1234, "step": 12380 }, { "epoch": 0.79, "learning_rate": 2.1472164033528453e-05, "loss": 0.1109, "step": 12381 }, { "epoch": 0.79, "learning_rate": 2.1459308146903013e-05, "loss": 0.1475, "step": 12382 }, { "epoch": 0.79, "learning_rate": 2.1446455647404184e-05, "loss": 0.1256, "step": 12383 }, { "epoch": 0.79, "learning_rate": 2.1433606535586292e-05, "loss": 0.1469, "step": 12384 }, { "epoch": 0.79, "learning_rate": 2.1420760812003426e-05, "loss": 0.1434, "step": 12385 }, { "epoch": 0.79, "learning_rate": 2.140791847720961e-05, "loss": 0.1247, "step": 12386 }, { "epoch": 0.79, "learning_rate": 2.139507953175862e-05, "loss": 0.1328, "step": 12387 }, { "epoch": 0.79, "learning_rate": 2.1382243976204175e-05, "loss": 0.1464, "step": 12388 }, { "epoch": 0.79, "learning_rate": 2.136941181109986e-05, "loss": 0.1393, "step": 12389 }, { "epoch": 0.79, "learning_rate": 2.1356583036999e-05, "loss": 0.1357, "step": 12390 }, { "epoch": 0.79, "learning_rate": 2.134375765445491e-05, "loss": 0.124, "step": 12391 }, { "epoch": 0.79, "learning_rate": 2.1330935664020625e-05, "loss": 0.1349, "step": 12392 }, { "epoch": 0.79, "learning_rate": 2.1318117066249176e-05, "loss": 0.1428, "step": 12393 }, { "epoch": 0.79, "learning_rate": 2.1305301861693316e-05, "loss": 0.1216, "step": 12394 }, { "epoch": 0.79, "learning_rate": 2.1292490050905734e-05, "loss": 0.1396, "step": 12395 }, { "epoch": 0.79, "learning_rate": 2.1279681634438974e-05, "loss": 0.1696, "step": 12396 }, { "epoch": 0.79, "learning_rate": 2.1266876612845355e-05, "loss": 0.1417, "step": 12397 }, { "epoch": 0.79, "learning_rate": 2.1254074986677153e-05, "loss": 0.1354, "step": 12398 }, { "epoch": 0.8, "learning_rate": 2.1241276756486416e-05, "loss": 0.1455, "step": 12399 }, { "epoch": 0.8, "learning_rate": 2.1228481922825106e-05, "loss": 0.1488, "step": 12400 }, { "epoch": 0.8, "learning_rate": 2.1215690486244966e-05, "loss": 0.1217, "step": 12401 }, { "epoch": 0.8, "learning_rate": 2.1202902447297658e-05, "loss": 0.1586, "step": 12402 }, { "epoch": 0.8, "learning_rate": 2.1190117806534715e-05, "loss": 0.1429, "step": 12403 }, { "epoch": 0.8, "learning_rate": 2.117733656450741e-05, "loss": 0.1321, "step": 12404 }, { "epoch": 0.8, "learning_rate": 2.1164558721767026e-05, "loss": 0.1433, "step": 12405 }, { "epoch": 0.8, "learning_rate": 2.115178427886454e-05, "loss": 0.1543, "step": 12406 }, { "epoch": 0.8, "learning_rate": 2.1139013236350925e-05, "loss": 0.1119, "step": 12407 }, { "epoch": 0.8, "learning_rate": 2.1126245594776873e-05, "loss": 0.1338, "step": 12408 }, { "epoch": 0.8, "learning_rate": 2.1113481354693043e-05, "loss": 0.1312, "step": 12409 }, { "epoch": 0.8, "learning_rate": 2.110072051664993e-05, "loss": 0.1247, "step": 12410 }, { "epoch": 0.8, "learning_rate": 2.1087963081197782e-05, "loss": 0.1345, "step": 12411 }, { "epoch": 0.8, "learning_rate": 2.1075209048886844e-05, "loss": 0.1589, "step": 12412 }, { "epoch": 0.8, "learning_rate": 2.1062458420267085e-05, "loss": 0.1366, "step": 12413 }, { "epoch": 0.8, "learning_rate": 2.1049711195888434e-05, "loss": 0.1367, "step": 12414 }, { "epoch": 0.8, "learning_rate": 2.1036967376300575e-05, "loss": 0.1355, "step": 12415 }, { "epoch": 0.8, "learning_rate": 2.102422696205312e-05, "loss": 0.147, "step": 12416 }, { "epoch": 0.8, "learning_rate": 2.101148995369554e-05, "loss": 0.1339, "step": 12417 }, { "epoch": 0.8, "learning_rate": 2.099875635177707e-05, "loss": 0.1416, "step": 12418 }, { "epoch": 0.8, "learning_rate": 2.0986026156846917e-05, "loss": 0.1339, "step": 12419 }, { "epoch": 0.8, "learning_rate": 2.0973299369454015e-05, "loss": 0.1459, "step": 12420 }, { "epoch": 0.8, "learning_rate": 2.0960575990147257e-05, "loss": 0.1314, "step": 12421 }, { "epoch": 0.8, "learning_rate": 2.0947856019475364e-05, "loss": 0.1369, "step": 12422 }, { "epoch": 0.8, "learning_rate": 2.093513945798684e-05, "loss": 0.1296, "step": 12423 }, { "epoch": 0.8, "learning_rate": 2.092242630623016e-05, "loss": 0.1459, "step": 12424 }, { "epoch": 0.8, "learning_rate": 2.0909716564753523e-05, "loss": 0.1435, "step": 12425 }, { "epoch": 0.8, "learning_rate": 2.0897010234105108e-05, "loss": 0.1425, "step": 12426 }, { "epoch": 0.8, "learning_rate": 2.088430731483283e-05, "loss": 0.1398, "step": 12427 }, { "epoch": 0.8, "learning_rate": 2.087160780748454e-05, "loss": 0.1455, "step": 12428 }, { "epoch": 0.8, "learning_rate": 2.0858911712607932e-05, "loss": 0.1298, "step": 12429 }, { "epoch": 0.8, "learning_rate": 2.084621903075049e-05, "loss": 0.1314, "step": 12430 }, { "epoch": 0.8, "learning_rate": 2.0833529762459658e-05, "loss": 0.1505, "step": 12431 }, { "epoch": 0.8, "learning_rate": 2.0820843908282594e-05, "loss": 0.1258, "step": 12432 }, { "epoch": 0.8, "learning_rate": 2.0808161468766453e-05, "loss": 0.1275, "step": 12433 }, { "epoch": 0.8, "learning_rate": 2.0795482444458115e-05, "loss": 0.1149, "step": 12434 }, { "epoch": 0.8, "learning_rate": 2.078280683590441e-05, "loss": 0.1303, "step": 12435 }, { "epoch": 0.8, "learning_rate": 2.0770134643651994e-05, "loss": 0.1324, "step": 12436 }, { "epoch": 0.8, "learning_rate": 2.0757465868247328e-05, "loss": 0.1316, "step": 12437 }, { "epoch": 0.8, "learning_rate": 2.0744800510236806e-05, "loss": 0.1445, "step": 12438 }, { "epoch": 0.8, "learning_rate": 2.073213857016657e-05, "loss": 0.1365, "step": 12439 }, { "epoch": 0.8, "learning_rate": 2.0719480048582753e-05, "loss": 0.1338, "step": 12440 }, { "epoch": 0.8, "learning_rate": 2.070682494603119e-05, "loss": 0.1316, "step": 12441 }, { "epoch": 0.8, "learning_rate": 2.069417326305767e-05, "loss": 0.1396, "step": 12442 }, { "epoch": 0.8, "learning_rate": 2.0681525000207835e-05, "loss": 0.1229, "step": 12443 }, { "epoch": 0.8, "learning_rate": 2.0668880158027105e-05, "loss": 0.1424, "step": 12444 }, { "epoch": 0.8, "learning_rate": 2.065623873706085e-05, "loss": 0.1384, "step": 12445 }, { "epoch": 0.8, "learning_rate": 2.0643600737854175e-05, "loss": 0.1299, "step": 12446 }, { "epoch": 0.8, "learning_rate": 2.063096616095217e-05, "loss": 0.1496, "step": 12447 }, { "epoch": 0.8, "learning_rate": 2.0618335006899657e-05, "loss": 0.1268, "step": 12448 }, { "epoch": 0.8, "learning_rate": 2.0605707276241382e-05, "loss": 0.1415, "step": 12449 }, { "epoch": 0.8, "learning_rate": 2.059308296952196e-05, "loss": 0.1247, "step": 12450 }, { "epoch": 0.8, "learning_rate": 2.058046208728577e-05, "loss": 0.1289, "step": 12451 }, { "epoch": 0.8, "learning_rate": 2.0567844630077137e-05, "loss": 0.1481, "step": 12452 }, { "epoch": 0.8, "learning_rate": 2.0555230598440156e-05, "loss": 0.1355, "step": 12453 }, { "epoch": 0.8, "learning_rate": 2.0542619992918876e-05, "loss": 0.1329, "step": 12454 }, { "epoch": 0.8, "learning_rate": 2.0530012814057075e-05, "loss": 0.144, "step": 12455 }, { "epoch": 0.8, "learning_rate": 2.0517409062398484e-05, "loss": 0.1452, "step": 12456 }, { "epoch": 0.8, "learning_rate": 2.050480873848667e-05, "loss": 0.1471, "step": 12457 }, { "epoch": 0.8, "learning_rate": 2.0492211842864973e-05, "loss": 0.1204, "step": 12458 }, { "epoch": 0.8, "learning_rate": 2.0479618376076705e-05, "loss": 0.1257, "step": 12459 }, { "epoch": 0.8, "learning_rate": 2.0467028338664906e-05, "loss": 0.1406, "step": 12460 }, { "epoch": 0.8, "learning_rate": 2.045444173117257e-05, "loss": 0.14, "step": 12461 }, { "epoch": 0.8, "learning_rate": 2.04418585541425e-05, "loss": 0.1346, "step": 12462 }, { "epoch": 0.8, "learning_rate": 2.0429278808117368e-05, "loss": 0.143, "step": 12463 }, { "epoch": 0.8, "learning_rate": 2.0416702493639685e-05, "loss": 0.1381, "step": 12464 }, { "epoch": 0.8, "learning_rate": 2.0404129611251764e-05, "loss": 0.1325, "step": 12465 }, { "epoch": 0.8, "learning_rate": 2.039156016149588e-05, "loss": 0.1234, "step": 12466 }, { "epoch": 0.8, "learning_rate": 2.0378994144914055e-05, "loss": 0.1428, "step": 12467 }, { "epoch": 0.8, "learning_rate": 2.0366431562048228e-05, "loss": 0.154, "step": 12468 }, { "epoch": 0.8, "learning_rate": 2.0353872413440155e-05, "loss": 0.133, "step": 12469 }, { "epoch": 0.8, "learning_rate": 2.034131669963152e-05, "loss": 0.137, "step": 12470 }, { "epoch": 0.8, "learning_rate": 2.0328764421163737e-05, "loss": 0.1535, "step": 12471 }, { "epoch": 0.8, "learning_rate": 2.0316215578578125e-05, "loss": 0.122, "step": 12472 }, { "epoch": 0.8, "learning_rate": 2.0303670172415913e-05, "loss": 0.1238, "step": 12473 }, { "epoch": 0.8, "learning_rate": 2.029112820321808e-05, "loss": 0.134, "step": 12474 }, { "epoch": 0.8, "learning_rate": 2.0278589671525517e-05, "loss": 0.1414, "step": 12475 }, { "epoch": 0.8, "learning_rate": 2.026605457787898e-05, "loss": 0.1374, "step": 12476 }, { "epoch": 0.8, "learning_rate": 2.0253522922819067e-05, "loss": 0.138, "step": 12477 }, { "epoch": 0.8, "learning_rate": 2.0240994706886195e-05, "loss": 0.1275, "step": 12478 }, { "epoch": 0.8, "learning_rate": 2.022846993062063e-05, "loss": 0.15, "step": 12479 }, { "epoch": 0.8, "learning_rate": 2.021594859456255e-05, "loss": 0.1407, "step": 12480 }, { "epoch": 0.8, "learning_rate": 2.0203430699251914e-05, "loss": 0.1378, "step": 12481 }, { "epoch": 0.8, "learning_rate": 2.0190916245228575e-05, "loss": 0.1391, "step": 12482 }, { "epoch": 0.8, "learning_rate": 2.017840523303224e-05, "loss": 0.1373, "step": 12483 }, { "epoch": 0.8, "learning_rate": 2.0165897663202472e-05, "loss": 0.1198, "step": 12484 }, { "epoch": 0.8, "learning_rate": 2.0153393536278653e-05, "loss": 0.1214, "step": 12485 }, { "epoch": 0.8, "learning_rate": 2.01408928528e-05, "loss": 0.1377, "step": 12486 }, { "epoch": 0.8, "learning_rate": 2.0128395613305673e-05, "loss": 0.1354, "step": 12487 }, { "epoch": 0.8, "learning_rate": 2.011590181833456e-05, "loss": 0.144, "step": 12488 }, { "epoch": 0.8, "learning_rate": 2.0103411468425515e-05, "loss": 0.1349, "step": 12489 }, { "epoch": 0.8, "learning_rate": 2.0090924564117197e-05, "loss": 0.1474, "step": 12490 }, { "epoch": 0.8, "learning_rate": 2.0078441105948066e-05, "loss": 0.1277, "step": 12491 }, { "epoch": 0.8, "learning_rate": 2.006596109445653e-05, "loss": 0.1609, "step": 12492 }, { "epoch": 0.8, "learning_rate": 2.0053484530180768e-05, "loss": 0.1579, "step": 12493 }, { "epoch": 0.8, "learning_rate": 2.0041011413658874e-05, "loss": 0.1431, "step": 12494 }, { "epoch": 0.8, "learning_rate": 2.0028541745428708e-05, "loss": 0.1455, "step": 12495 }, { "epoch": 0.8, "learning_rate": 2.0016075526028065e-05, "loss": 0.1488, "step": 12496 }, { "epoch": 0.8, "learning_rate": 2.0003612755994584e-05, "loss": 0.1356, "step": 12497 }, { "epoch": 0.8, "learning_rate": 1.9991153435865683e-05, "loss": 0.1422, "step": 12498 }, { "epoch": 0.8, "learning_rate": 1.9978697566178727e-05, "loss": 0.1312, "step": 12499 }, { "epoch": 0.8, "learning_rate": 1.996624514747084e-05, "loss": 0.1316, "step": 12500 }, { "epoch": 0.8, "learning_rate": 1.995379618027908e-05, "loss": 0.1406, "step": 12501 }, { "epoch": 0.8, "learning_rate": 1.9941350665140257e-05, "loss": 0.1246, "step": 12502 }, { "epoch": 0.8, "learning_rate": 1.992890860259119e-05, "loss": 0.1395, "step": 12503 }, { "epoch": 0.8, "learning_rate": 1.9916469993168397e-05, "loss": 0.1213, "step": 12504 }, { "epoch": 0.8, "learning_rate": 1.9904034837408282e-05, "loss": 0.1346, "step": 12505 }, { "epoch": 0.8, "learning_rate": 1.989160313584717e-05, "loss": 0.1303, "step": 12506 }, { "epoch": 0.8, "learning_rate": 1.9879174889021136e-05, "loss": 0.1202, "step": 12507 }, { "epoch": 0.8, "learning_rate": 1.986675009746618e-05, "loss": 0.1434, "step": 12508 }, { "epoch": 0.8, "learning_rate": 1.9854328761718133e-05, "loss": 0.1373, "step": 12509 }, { "epoch": 0.8, "learning_rate": 1.9841910882312697e-05, "loss": 0.1232, "step": 12510 }, { "epoch": 0.8, "learning_rate": 1.982949645978539e-05, "loss": 0.1315, "step": 12511 }, { "epoch": 0.8, "learning_rate": 1.981708549467155e-05, "loss": 0.1371, "step": 12512 }, { "epoch": 0.8, "learning_rate": 1.980467798750647e-05, "loss": 0.1236, "step": 12513 }, { "epoch": 0.8, "learning_rate": 1.979227393882518e-05, "loss": 0.1378, "step": 12514 }, { "epoch": 0.8, "learning_rate": 1.9779873349162647e-05, "loss": 0.1287, "step": 12515 }, { "epoch": 0.8, "learning_rate": 1.9767476219053638e-05, "loss": 0.1392, "step": 12516 }, { "epoch": 0.8, "learning_rate": 1.9755082549032832e-05, "loss": 0.1584, "step": 12517 }, { "epoch": 0.8, "learning_rate": 1.9742692339634684e-05, "loss": 0.1279, "step": 12518 }, { "epoch": 0.8, "learning_rate": 1.9730305591393495e-05, "loss": 0.1436, "step": 12519 }, { "epoch": 0.8, "learning_rate": 1.9717922304843515e-05, "loss": 0.1545, "step": 12520 }, { "epoch": 0.8, "learning_rate": 1.9705542480518736e-05, "loss": 0.1444, "step": 12521 }, { "epoch": 0.8, "learning_rate": 1.9693166118953065e-05, "loss": 0.1283, "step": 12522 }, { "epoch": 0.8, "learning_rate": 1.968079322068024e-05, "loss": 0.1565, "step": 12523 }, { "epoch": 0.8, "learning_rate": 1.966842378623389e-05, "loss": 0.1369, "step": 12524 }, { "epoch": 0.8, "learning_rate": 1.965605781614741e-05, "loss": 0.1167, "step": 12525 }, { "epoch": 0.8, "learning_rate": 1.964369531095408e-05, "loss": 0.1449, "step": 12526 }, { "epoch": 0.8, "learning_rate": 1.9631336271187096e-05, "loss": 0.1238, "step": 12527 }, { "epoch": 0.8, "learning_rate": 1.9618980697379398e-05, "loss": 0.1361, "step": 12528 }, { "epoch": 0.8, "learning_rate": 1.9606628590063835e-05, "loss": 0.1274, "step": 12529 }, { "epoch": 0.8, "learning_rate": 1.959427994977313e-05, "loss": 0.1367, "step": 12530 }, { "epoch": 0.8, "learning_rate": 1.9581934777039846e-05, "loss": 0.1373, "step": 12531 }, { "epoch": 0.8, "learning_rate": 1.956959307239633e-05, "loss": 0.1243, "step": 12532 }, { "epoch": 0.8, "learning_rate": 1.9557254836374837e-05, "loss": 0.14, "step": 12533 }, { "epoch": 0.8, "learning_rate": 1.9544920069507488e-05, "loss": 0.139, "step": 12534 }, { "epoch": 0.8, "learning_rate": 1.9532588772326177e-05, "loss": 0.1495, "step": 12535 }, { "epoch": 0.8, "learning_rate": 1.9520260945362746e-05, "loss": 0.1525, "step": 12536 }, { "epoch": 0.8, "learning_rate": 1.950793658914882e-05, "loss": 0.1353, "step": 12537 }, { "epoch": 0.8, "learning_rate": 1.9495615704215953e-05, "loss": 0.1453, "step": 12538 }, { "epoch": 0.8, "learning_rate": 1.948329829109543e-05, "loss": 0.1331, "step": 12539 }, { "epoch": 0.8, "learning_rate": 1.9470984350318444e-05, "loss": 0.169, "step": 12540 }, { "epoch": 0.8, "learning_rate": 1.945867388241609e-05, "loss": 0.1451, "step": 12541 }, { "epoch": 0.8, "learning_rate": 1.944636688791921e-05, "loss": 0.1314, "step": 12542 }, { "epoch": 0.8, "learning_rate": 1.943406336735858e-05, "loss": 0.1372, "step": 12543 }, { "epoch": 0.8, "learning_rate": 1.9421763321264807e-05, "loss": 0.1244, "step": 12544 }, { "epoch": 0.8, "learning_rate": 1.9409466750168358e-05, "loss": 0.1317, "step": 12545 }, { "epoch": 0.8, "learning_rate": 1.939717365459952e-05, "loss": 0.1274, "step": 12546 }, { "epoch": 0.8, "learning_rate": 1.938488403508839e-05, "loss": 0.1352, "step": 12547 }, { "epoch": 0.8, "learning_rate": 1.937259789216501e-05, "loss": 0.1387, "step": 12548 }, { "epoch": 0.8, "learning_rate": 1.9360315226359227e-05, "loss": 0.1321, "step": 12549 }, { "epoch": 0.8, "learning_rate": 1.9348036038200766e-05, "loss": 0.1244, "step": 12550 }, { "epoch": 0.8, "learning_rate": 1.933576032821912e-05, "loss": 0.1286, "step": 12551 }, { "epoch": 0.8, "learning_rate": 1.9323488096943754e-05, "loss": 0.1385, "step": 12552 }, { "epoch": 0.8, "learning_rate": 1.9311219344903875e-05, "loss": 0.1353, "step": 12553 }, { "epoch": 0.8, "learning_rate": 1.9298954072628562e-05, "loss": 0.1464, "step": 12554 }, { "epoch": 0.81, "learning_rate": 1.928669228064679e-05, "loss": 0.132, "step": 12555 }, { "epoch": 0.81, "learning_rate": 1.927443396948736e-05, "loss": 0.1287, "step": 12556 }, { "epoch": 0.81, "learning_rate": 1.9262179139678937e-05, "loss": 0.1464, "step": 12557 }, { "epoch": 0.81, "learning_rate": 1.9249927791749987e-05, "loss": 0.149, "step": 12558 }, { "epoch": 0.81, "learning_rate": 1.92376799262289e-05, "loss": 0.1579, "step": 12559 }, { "epoch": 0.81, "learning_rate": 1.9225435543643845e-05, "loss": 0.1541, "step": 12560 }, { "epoch": 0.81, "learning_rate": 1.9213194644522848e-05, "loss": 0.1319, "step": 12561 }, { "epoch": 0.81, "learning_rate": 1.9200957229393836e-05, "loss": 0.1359, "step": 12562 }, { "epoch": 0.81, "learning_rate": 1.918872329878455e-05, "loss": 0.1565, "step": 12563 }, { "epoch": 0.81, "learning_rate": 1.9176492853222617e-05, "loss": 0.1575, "step": 12564 }, { "epoch": 0.81, "learning_rate": 1.9164265893235457e-05, "loss": 0.1368, "step": 12565 }, { "epoch": 0.81, "learning_rate": 1.9152042419350346e-05, "loss": 0.1386, "step": 12566 }, { "epoch": 0.81, "learning_rate": 1.9139822432094477e-05, "loss": 0.1366, "step": 12567 }, { "epoch": 0.81, "learning_rate": 1.9127605931994797e-05, "loss": 0.1378, "step": 12568 }, { "epoch": 0.81, "learning_rate": 1.911539291957818e-05, "loss": 0.1438, "step": 12569 }, { "epoch": 0.81, "learning_rate": 1.9103183395371304e-05, "loss": 0.1519, "step": 12570 }, { "epoch": 0.81, "learning_rate": 1.9090977359900763e-05, "loss": 0.1457, "step": 12571 }, { "epoch": 0.81, "learning_rate": 1.9078774813692924e-05, "loss": 0.1402, "step": 12572 }, { "epoch": 0.81, "learning_rate": 1.9066575757273987e-05, "loss": 0.1321, "step": 12573 }, { "epoch": 0.81, "learning_rate": 1.905438019117011e-05, "loss": 0.1348, "step": 12574 }, { "epoch": 0.81, "learning_rate": 1.9042188115907178e-05, "loss": 0.1366, "step": 12575 }, { "epoch": 0.81, "learning_rate": 1.9029999532011023e-05, "loss": 0.1387, "step": 12576 }, { "epoch": 0.81, "learning_rate": 1.9017814440007274e-05, "loss": 0.1508, "step": 12577 }, { "epoch": 0.81, "learning_rate": 1.900563284042145e-05, "loss": 0.1384, "step": 12578 }, { "epoch": 0.81, "learning_rate": 1.899345473377886e-05, "loss": 0.1386, "step": 12579 }, { "epoch": 0.81, "learning_rate": 1.8981280120604684e-05, "loss": 0.139, "step": 12580 }, { "epoch": 0.81, "learning_rate": 1.8969109001424e-05, "loss": 0.129, "step": 12581 }, { "epoch": 0.81, "learning_rate": 1.895694137676166e-05, "loss": 0.1228, "step": 12582 }, { "epoch": 0.81, "learning_rate": 1.894477724714241e-05, "loss": 0.1225, "step": 12583 }, { "epoch": 0.81, "learning_rate": 1.893261661309085e-05, "loss": 0.1282, "step": 12584 }, { "epoch": 0.81, "learning_rate": 1.8920459475131423e-05, "loss": 0.149, "step": 12585 }, { "epoch": 0.81, "learning_rate": 1.8908305833788408e-05, "loss": 0.1286, "step": 12586 }, { "epoch": 0.81, "learning_rate": 1.8896155689585915e-05, "loss": 0.1331, "step": 12587 }, { "epoch": 0.81, "learning_rate": 1.888400904304798e-05, "loss": 0.1296, "step": 12588 }, { "epoch": 0.81, "learning_rate": 1.8871865894698336e-05, "loss": 0.1346, "step": 12589 }, { "epoch": 0.81, "learning_rate": 1.8859726245060794e-05, "loss": 0.1574, "step": 12590 }, { "epoch": 0.81, "learning_rate": 1.8847590094658796e-05, "loss": 0.1487, "step": 12591 }, { "epoch": 0.81, "learning_rate": 1.883545744401578e-05, "loss": 0.1115, "step": 12592 }, { "epoch": 0.81, "learning_rate": 1.8823328293654952e-05, "loss": 0.1398, "step": 12593 }, { "epoch": 0.81, "learning_rate": 1.8811202644099356e-05, "loss": 0.131, "step": 12594 }, { "epoch": 0.81, "learning_rate": 1.8799080495871955e-05, "loss": 0.1387, "step": 12595 }, { "epoch": 0.81, "learning_rate": 1.878696184949552e-05, "loss": 0.1228, "step": 12596 }, { "epoch": 0.81, "learning_rate": 1.87748467054927e-05, "loss": 0.134, "step": 12597 }, { "epoch": 0.81, "learning_rate": 1.8762735064385917e-05, "loss": 0.129, "step": 12598 }, { "epoch": 0.81, "learning_rate": 1.8750626926697555e-05, "loss": 0.1211, "step": 12599 }, { "epoch": 0.81, "learning_rate": 1.8738522292949767e-05, "loss": 0.1348, "step": 12600 }, { "epoch": 0.81, "learning_rate": 1.872642116366453e-05, "loss": 0.1337, "step": 12601 }, { "epoch": 0.81, "learning_rate": 1.8714323539363744e-05, "loss": 0.1317, "step": 12602 }, { "epoch": 0.81, "learning_rate": 1.870222942056914e-05, "loss": 0.1282, "step": 12603 }, { "epoch": 0.81, "learning_rate": 1.8690138807802304e-05, "loss": 0.1379, "step": 12604 }, { "epoch": 0.81, "learning_rate": 1.86780517015846e-05, "loss": 0.1332, "step": 12605 }, { "epoch": 0.81, "learning_rate": 1.866596810243735e-05, "loss": 0.1313, "step": 12606 }, { "epoch": 0.81, "learning_rate": 1.8653888010881637e-05, "loss": 0.1288, "step": 12607 }, { "epoch": 0.81, "learning_rate": 1.8641811427438394e-05, "loss": 0.1293, "step": 12608 }, { "epoch": 0.81, "learning_rate": 1.862973835262848e-05, "loss": 0.1246, "step": 12609 }, { "epoch": 0.81, "learning_rate": 1.861766878697253e-05, "loss": 0.137, "step": 12610 }, { "epoch": 0.81, "learning_rate": 1.860560273099109e-05, "loss": 0.1475, "step": 12611 }, { "epoch": 0.81, "learning_rate": 1.8593540185204472e-05, "loss": 0.1285, "step": 12612 }, { "epoch": 0.81, "learning_rate": 1.858148115013293e-05, "loss": 0.1418, "step": 12613 }, { "epoch": 0.81, "learning_rate": 1.8569425626296477e-05, "loss": 0.1345, "step": 12614 }, { "epoch": 0.81, "learning_rate": 1.855737361421501e-05, "loss": 0.1367, "step": 12615 }, { "epoch": 0.81, "learning_rate": 1.854532511440831e-05, "loss": 0.1316, "step": 12616 }, { "epoch": 0.81, "learning_rate": 1.8533280127395967e-05, "loss": 0.1265, "step": 12617 }, { "epoch": 0.81, "learning_rate": 1.852123865369746e-05, "loss": 0.14, "step": 12618 }, { "epoch": 0.81, "learning_rate": 1.8509200693832028e-05, "loss": 0.1502, "step": 12619 }, { "epoch": 0.81, "learning_rate": 1.8497166248318876e-05, "loss": 0.1277, "step": 12620 }, { "epoch": 0.81, "learning_rate": 1.8485135317676972e-05, "loss": 0.1454, "step": 12621 }, { "epoch": 0.81, "learning_rate": 1.8473107902425135e-05, "loss": 0.1322, "step": 12622 }, { "epoch": 0.81, "learning_rate": 1.8461084003082073e-05, "loss": 0.1511, "step": 12623 }, { "epoch": 0.81, "learning_rate": 1.8449063620166327e-05, "loss": 0.1301, "step": 12624 }, { "epoch": 0.81, "learning_rate": 1.843704675419633e-05, "loss": 0.1516, "step": 12625 }, { "epoch": 0.81, "learning_rate": 1.842503340569025e-05, "loss": 0.1268, "step": 12626 }, { "epoch": 0.81, "learning_rate": 1.841302357516622e-05, "loss": 0.1235, "step": 12627 }, { "epoch": 0.81, "learning_rate": 1.840101726314215e-05, "loss": 0.128, "step": 12628 }, { "epoch": 0.81, "learning_rate": 1.8389014470135813e-05, "loss": 0.1205, "step": 12629 }, { "epoch": 0.81, "learning_rate": 1.8377015196664836e-05, "loss": 0.1366, "step": 12630 }, { "epoch": 0.81, "learning_rate": 1.8365019443246724e-05, "loss": 0.1334, "step": 12631 }, { "epoch": 0.81, "learning_rate": 1.83530272103988e-05, "loss": 0.1353, "step": 12632 }, { "epoch": 0.81, "learning_rate": 1.8341038498638206e-05, "loss": 0.1468, "step": 12633 }, { "epoch": 0.81, "learning_rate": 1.8329053308482025e-05, "loss": 0.1383, "step": 12634 }, { "epoch": 0.81, "learning_rate": 1.8317071640447048e-05, "loss": 0.1318, "step": 12635 }, { "epoch": 0.81, "learning_rate": 1.8305093495050042e-05, "loss": 0.1347, "step": 12636 }, { "epoch": 0.81, "learning_rate": 1.829311887280759e-05, "loss": 0.1328, "step": 12637 }, { "epoch": 0.81, "learning_rate": 1.828114777423605e-05, "loss": 0.1451, "step": 12638 }, { "epoch": 0.81, "learning_rate": 1.8269180199851755e-05, "loss": 0.1438, "step": 12639 }, { "epoch": 0.81, "learning_rate": 1.825721615017074e-05, "loss": 0.1394, "step": 12640 }, { "epoch": 0.81, "learning_rate": 1.8245255625709035e-05, "loss": 0.1455, "step": 12641 }, { "epoch": 0.81, "learning_rate": 1.8233298626982375e-05, "loss": 0.1381, "step": 12642 }, { "epoch": 0.81, "learning_rate": 1.8221345154506464e-05, "loss": 0.1381, "step": 12643 }, { "epoch": 0.81, "learning_rate": 1.8209395208796808e-05, "loss": 0.1451, "step": 12644 }, { "epoch": 0.81, "learning_rate": 1.819744879036873e-05, "loss": 0.1399, "step": 12645 }, { "epoch": 0.81, "learning_rate": 1.8185505899737466e-05, "loss": 0.1353, "step": 12646 }, { "epoch": 0.81, "learning_rate": 1.817356653741803e-05, "loss": 0.1401, "step": 12647 }, { "epoch": 0.81, "learning_rate": 1.8161630703925292e-05, "loss": 0.1469, "step": 12648 }, { "epoch": 0.81, "learning_rate": 1.814969839977403e-05, "loss": 0.1382, "step": 12649 }, { "epoch": 0.81, "learning_rate": 1.8137769625478828e-05, "loss": 0.1737, "step": 12650 }, { "epoch": 0.81, "learning_rate": 1.812584438155416e-05, "loss": 0.1406, "step": 12651 }, { "epoch": 0.81, "learning_rate": 1.8113922668514228e-05, "loss": 0.1341, "step": 12652 }, { "epoch": 0.81, "learning_rate": 1.810200448687325e-05, "loss": 0.1351, "step": 12653 }, { "epoch": 0.81, "learning_rate": 1.8090089837145174e-05, "loss": 0.1218, "step": 12654 }, { "epoch": 0.81, "learning_rate": 1.8078178719843787e-05, "loss": 0.1561, "step": 12655 }, { "epoch": 0.81, "learning_rate": 1.8066271135482806e-05, "loss": 0.1276, "step": 12656 }, { "epoch": 0.81, "learning_rate": 1.805436708457574e-05, "loss": 0.1363, "step": 12657 }, { "epoch": 0.81, "learning_rate": 1.8042466567636008e-05, "loss": 0.1365, "step": 12658 }, { "epoch": 0.81, "learning_rate": 1.803056958517676e-05, "loss": 0.1431, "step": 12659 }, { "epoch": 0.81, "learning_rate": 1.801867613771111e-05, "loss": 0.158, "step": 12660 }, { "epoch": 0.81, "learning_rate": 1.800678622575196e-05, "loss": 0.138, "step": 12661 }, { "epoch": 0.81, "learning_rate": 1.7994899849812042e-05, "loss": 0.1317, "step": 12662 }, { "epoch": 0.81, "learning_rate": 1.798301701040399e-05, "loss": 0.1425, "step": 12663 }, { "epoch": 0.81, "learning_rate": 1.7971137708040254e-05, "loss": 0.1556, "step": 12664 }, { "epoch": 0.81, "learning_rate": 1.7959261943233175e-05, "loss": 0.1249, "step": 12665 }, { "epoch": 0.81, "learning_rate": 1.7947389716494833e-05, "loss": 0.1449, "step": 12666 }, { "epoch": 0.81, "learning_rate": 1.793552102833731e-05, "loss": 0.1408, "step": 12667 }, { "epoch": 0.81, "learning_rate": 1.7923655879272393e-05, "loss": 0.1216, "step": 12668 }, { "epoch": 0.81, "learning_rate": 1.7911794269811778e-05, "loss": 0.1307, "step": 12669 }, { "epoch": 0.81, "learning_rate": 1.7899936200467006e-05, "loss": 0.1461, "step": 12670 }, { "epoch": 0.81, "learning_rate": 1.7888081671749478e-05, "loss": 0.1476, "step": 12671 }, { "epoch": 0.81, "learning_rate": 1.787623068417047e-05, "loss": 0.1238, "step": 12672 }, { "epoch": 0.81, "learning_rate": 1.7864383238240988e-05, "loss": 0.1613, "step": 12673 }, { "epoch": 0.81, "learning_rate": 1.785253933447202e-05, "loss": 0.1312, "step": 12674 }, { "epoch": 0.81, "learning_rate": 1.7840698973374326e-05, "loss": 0.1513, "step": 12675 }, { "epoch": 0.81, "learning_rate": 1.7828862155458482e-05, "loss": 0.1203, "step": 12676 }, { "epoch": 0.81, "learning_rate": 1.7817028881235042e-05, "loss": 0.1478, "step": 12677 }, { "epoch": 0.81, "learning_rate": 1.780519915121427e-05, "loss": 0.1223, "step": 12678 }, { "epoch": 0.81, "learning_rate": 1.7793372965906362e-05, "loss": 0.1284, "step": 12679 }, { "epoch": 0.81, "learning_rate": 1.7781550325821305e-05, "loss": 0.1343, "step": 12680 }, { "epoch": 0.81, "learning_rate": 1.776973123146899e-05, "loss": 0.1488, "step": 12681 }, { "epoch": 0.81, "learning_rate": 1.7757915683359083e-05, "loss": 0.1167, "step": 12682 }, { "epoch": 0.81, "learning_rate": 1.774610368200117e-05, "loss": 0.1415, "step": 12683 }, { "epoch": 0.81, "learning_rate": 1.7734295227904664e-05, "loss": 0.1314, "step": 12684 }, { "epoch": 0.81, "learning_rate": 1.7722490321578757e-05, "loss": 0.1303, "step": 12685 }, { "epoch": 0.81, "learning_rate": 1.771068896353262e-05, "loss": 0.1473, "step": 12686 }, { "epoch": 0.81, "learning_rate": 1.769889115427513e-05, "loss": 0.152, "step": 12687 }, { "epoch": 0.81, "learning_rate": 1.7687096894315124e-05, "loss": 0.1298, "step": 12688 }, { "epoch": 0.81, "learning_rate": 1.7675306184161188e-05, "loss": 0.1226, "step": 12689 }, { "epoch": 0.81, "learning_rate": 1.766351902432185e-05, "loss": 0.1489, "step": 12690 }, { "epoch": 0.81, "learning_rate": 1.7651735415305436e-05, "loss": 0.1304, "step": 12691 }, { "epoch": 0.81, "learning_rate": 1.7639955357620096e-05, "loss": 0.1248, "step": 12692 }, { "epoch": 0.81, "learning_rate": 1.762817885177389e-05, "loss": 0.1335, "step": 12693 }, { "epoch": 0.81, "learning_rate": 1.7616405898274646e-05, "loss": 0.1481, "step": 12694 }, { "epoch": 0.81, "learning_rate": 1.760463649763012e-05, "loss": 0.164, "step": 12695 }, { "epoch": 0.81, "learning_rate": 1.7592870650347848e-05, "loss": 0.1567, "step": 12696 }, { "epoch": 0.81, "learning_rate": 1.758110835693525e-05, "loss": 0.1383, "step": 12697 }, { "epoch": 0.81, "learning_rate": 1.756934961789961e-05, "loss": 0.1491, "step": 12698 }, { "epoch": 0.81, "learning_rate": 1.755759443374798e-05, "loss": 0.1438, "step": 12699 }, { "epoch": 0.81, "learning_rate": 1.754584280498738e-05, "loss": 0.123, "step": 12700 }, { "epoch": 0.81, "learning_rate": 1.7534094732124528e-05, "loss": 0.1279, "step": 12701 }, { "epoch": 0.81, "learning_rate": 1.752235021566615e-05, "loss": 0.1396, "step": 12702 }, { "epoch": 0.81, "learning_rate": 1.751060925611866e-05, "loss": 0.138, "step": 12703 }, { "epoch": 0.81, "learning_rate": 1.749887185398843e-05, "loss": 0.1383, "step": 12704 }, { "epoch": 0.81, "learning_rate": 1.7487138009781677e-05, "loss": 0.1252, "step": 12705 }, { "epoch": 0.81, "learning_rate": 1.7475407724004378e-05, "loss": 0.1421, "step": 12706 }, { "epoch": 0.81, "learning_rate": 1.7463680997162458e-05, "loss": 0.139, "step": 12707 }, { "epoch": 0.81, "learning_rate": 1.745195782976159e-05, "loss": 0.1508, "step": 12708 }, { "epoch": 0.81, "learning_rate": 1.7440238222307402e-05, "loss": 0.1431, "step": 12709 }, { "epoch": 0.81, "learning_rate": 1.742852217530525e-05, "loss": 0.1361, "step": 12710 }, { "epoch": 0.82, "learning_rate": 1.7416809689260427e-05, "loss": 0.1383, "step": 12711 }, { "epoch": 0.82, "learning_rate": 1.7405100764678072e-05, "loss": 0.1427, "step": 12712 }, { "epoch": 0.82, "learning_rate": 1.7393395402063086e-05, "loss": 0.1449, "step": 12713 }, { "epoch": 0.82, "learning_rate": 1.7381693601920312e-05, "loss": 0.1514, "step": 12714 }, { "epoch": 0.82, "learning_rate": 1.7369995364754377e-05, "loss": 0.137, "step": 12715 }, { "epoch": 0.82, "learning_rate": 1.7358300691069797e-05, "loss": 0.1245, "step": 12716 }, { "epoch": 0.82, "learning_rate": 1.7346609581370888e-05, "loss": 0.135, "step": 12717 }, { "epoch": 0.82, "learning_rate": 1.7334922036161838e-05, "loss": 0.1576, "step": 12718 }, { "epoch": 0.82, "learning_rate": 1.7323238055946724e-05, "loss": 0.1335, "step": 12719 }, { "epoch": 0.82, "learning_rate": 1.731155764122938e-05, "loss": 0.1411, "step": 12720 }, { "epoch": 0.82, "learning_rate": 1.729988079251357e-05, "loss": 0.14, "step": 12721 }, { "epoch": 0.82, "learning_rate": 1.7288207510302813e-05, "loss": 0.1274, "step": 12722 }, { "epoch": 0.82, "learning_rate": 1.7276537795100565e-05, "loss": 0.1369, "step": 12723 }, { "epoch": 0.82, "learning_rate": 1.726487164741012e-05, "loss": 0.1404, "step": 12724 }, { "epoch": 0.82, "learning_rate": 1.7253209067734522e-05, "loss": 0.1281, "step": 12725 }, { "epoch": 0.82, "learning_rate": 1.7241550056576794e-05, "loss": 0.1482, "step": 12726 }, { "epoch": 0.82, "learning_rate": 1.722989461443969e-05, "loss": 0.1405, "step": 12727 }, { "epoch": 0.82, "learning_rate": 1.7218242741825897e-05, "loss": 0.137, "step": 12728 }, { "epoch": 0.82, "learning_rate": 1.7206594439237865e-05, "loss": 0.1382, "step": 12729 }, { "epoch": 0.82, "learning_rate": 1.7194949707177966e-05, "loss": 0.1309, "step": 12730 }, { "epoch": 0.82, "learning_rate": 1.7183308546148413e-05, "loss": 0.1368, "step": 12731 }, { "epoch": 0.82, "learning_rate": 1.7171670956651186e-05, "loss": 0.1379, "step": 12732 }, { "epoch": 0.82, "learning_rate": 1.7160036939188218e-05, "loss": 0.1333, "step": 12733 }, { "epoch": 0.82, "learning_rate": 1.7148406494261184e-05, "loss": 0.143, "step": 12734 }, { "epoch": 0.82, "learning_rate": 1.7136779622371703e-05, "loss": 0.1425, "step": 12735 }, { "epoch": 0.82, "learning_rate": 1.712515632402114e-05, "loss": 0.1285, "step": 12736 }, { "epoch": 0.82, "learning_rate": 1.71135365997108e-05, "loss": 0.1344, "step": 12737 }, { "epoch": 0.82, "learning_rate": 1.71019204499418e-05, "loss": 0.1379, "step": 12738 }, { "epoch": 0.82, "learning_rate": 1.709030787521504e-05, "loss": 0.1289, "step": 12739 }, { "epoch": 0.82, "learning_rate": 1.7078698876031395e-05, "loss": 0.163, "step": 12740 }, { "epoch": 0.82, "learning_rate": 1.7067093452891436e-05, "loss": 0.1368, "step": 12741 }, { "epoch": 0.82, "learning_rate": 1.7055491606295726e-05, "loss": 0.1485, "step": 12742 }, { "epoch": 0.82, "learning_rate": 1.7043893336744554e-05, "loss": 0.1322, "step": 12743 }, { "epoch": 0.82, "learning_rate": 1.703229864473811e-05, "loss": 0.1487, "step": 12744 }, { "epoch": 0.82, "learning_rate": 1.702070753077646e-05, "loss": 0.1292, "step": 12745 }, { "epoch": 0.82, "learning_rate": 1.7009119995359422e-05, "loss": 0.1315, "step": 12746 }, { "epoch": 0.82, "learning_rate": 1.6997536038986785e-05, "loss": 0.1338, "step": 12747 }, { "epoch": 0.82, "learning_rate": 1.698595566215806e-05, "loss": 0.1371, "step": 12748 }, { "epoch": 0.82, "learning_rate": 1.6974378865372686e-05, "loss": 0.1415, "step": 12749 }, { "epoch": 0.82, "learning_rate": 1.696280564912991e-05, "loss": 0.127, "step": 12750 }, { "epoch": 0.82, "learning_rate": 1.695123601392883e-05, "loss": 0.1408, "step": 12751 }, { "epoch": 0.82, "learning_rate": 1.6939669960268423e-05, "loss": 0.1417, "step": 12752 }, { "epoch": 0.82, "learning_rate": 1.692810748864746e-05, "loss": 0.121, "step": 12753 }, { "epoch": 0.82, "learning_rate": 1.6916548599564595e-05, "loss": 0.1415, "step": 12754 }, { "epoch": 0.82, "learning_rate": 1.690499329351829e-05, "loss": 0.1431, "step": 12755 }, { "epoch": 0.82, "learning_rate": 1.6893441571006918e-05, "loss": 0.1383, "step": 12756 }, { "epoch": 0.82, "learning_rate": 1.6881893432528605e-05, "loss": 0.1408, "step": 12757 }, { "epoch": 0.82, "learning_rate": 1.68703488785814e-05, "loss": 0.1384, "step": 12758 }, { "epoch": 0.82, "learning_rate": 1.6858807909663188e-05, "loss": 0.1369, "step": 12759 }, { "epoch": 0.82, "learning_rate": 1.684727052627165e-05, "loss": 0.1249, "step": 12760 }, { "epoch": 0.82, "learning_rate": 1.683573672890437e-05, "loss": 0.1432, "step": 12761 }, { "epoch": 0.82, "learning_rate": 1.6824206518058727e-05, "loss": 0.1498, "step": 12762 }, { "epoch": 0.82, "learning_rate": 1.6812679894231976e-05, "loss": 0.1327, "step": 12763 }, { "epoch": 0.82, "learning_rate": 1.680115685792124e-05, "loss": 0.1415, "step": 12764 }, { "epoch": 0.82, "learning_rate": 1.6789637409623416e-05, "loss": 0.1435, "step": 12765 }, { "epoch": 0.82, "learning_rate": 1.6778121549835336e-05, "loss": 0.1267, "step": 12766 }, { "epoch": 0.82, "learning_rate": 1.676660927905357e-05, "loss": 0.1677, "step": 12767 }, { "epoch": 0.82, "learning_rate": 1.6755100597774666e-05, "loss": 0.1393, "step": 12768 }, { "epoch": 0.82, "learning_rate": 1.674359550649487e-05, "loss": 0.1563, "step": 12769 }, { "epoch": 0.82, "learning_rate": 1.6732094005710387e-05, "loss": 0.1249, "step": 12770 }, { "epoch": 0.82, "learning_rate": 1.672059609591724e-05, "loss": 0.1555, "step": 12771 }, { "epoch": 0.82, "learning_rate": 1.6709101777611257e-05, "loss": 0.1355, "step": 12772 }, { "epoch": 0.82, "learning_rate": 1.6697611051288164e-05, "loss": 0.1508, "step": 12773 }, { "epoch": 0.82, "learning_rate": 1.6686123917443485e-05, "loss": 0.1313, "step": 12774 }, { "epoch": 0.82, "learning_rate": 1.6674640376572638e-05, "loss": 0.1519, "step": 12775 }, { "epoch": 0.82, "learning_rate": 1.666316042917081e-05, "loss": 0.1433, "step": 12776 }, { "epoch": 0.82, "learning_rate": 1.665168407573312e-05, "loss": 0.1174, "step": 12777 }, { "epoch": 0.82, "learning_rate": 1.6640211316754517e-05, "loss": 0.1396, "step": 12778 }, { "epoch": 0.82, "learning_rate": 1.6628742152729728e-05, "loss": 0.1301, "step": 12779 }, { "epoch": 0.82, "learning_rate": 1.6617276584153398e-05, "loss": 0.1243, "step": 12780 }, { "epoch": 0.82, "learning_rate": 1.6605814611519954e-05, "loss": 0.1253, "step": 12781 }, { "epoch": 0.82, "learning_rate": 1.659435623532375e-05, "loss": 0.1532, "step": 12782 }, { "epoch": 0.82, "learning_rate": 1.6582901456058886e-05, "loss": 0.135, "step": 12783 }, { "epoch": 0.82, "learning_rate": 1.657145027421939e-05, "loss": 0.1226, "step": 12784 }, { "epoch": 0.82, "learning_rate": 1.6560002690299115e-05, "loss": 0.1285, "step": 12785 }, { "epoch": 0.82, "learning_rate": 1.6548558704791705e-05, "loss": 0.1263, "step": 12786 }, { "epoch": 0.82, "learning_rate": 1.6537118318190736e-05, "loss": 0.1307, "step": 12787 }, { "epoch": 0.82, "learning_rate": 1.6525681530989533e-05, "loss": 0.132, "step": 12788 }, { "epoch": 0.82, "learning_rate": 1.6514248343681372e-05, "loss": 0.1436, "step": 12789 }, { "epoch": 0.82, "learning_rate": 1.6502818756759276e-05, "loss": 0.1329, "step": 12790 }, { "epoch": 0.82, "learning_rate": 1.649139277071615e-05, "loss": 0.1387, "step": 12791 }, { "epoch": 0.82, "learning_rate": 1.6479970386044806e-05, "loss": 0.1224, "step": 12792 }, { "epoch": 0.82, "learning_rate": 1.646855160323777e-05, "loss": 0.1289, "step": 12793 }, { "epoch": 0.82, "learning_rate": 1.6457136422787545e-05, "loss": 0.1446, "step": 12794 }, { "epoch": 0.82, "learning_rate": 1.644572484518637e-05, "loss": 0.1194, "step": 12795 }, { "epoch": 0.82, "learning_rate": 1.6434316870926435e-05, "loss": 0.1408, "step": 12796 }, { "epoch": 0.82, "learning_rate": 1.6422912500499654e-05, "loss": 0.1456, "step": 12797 }, { "epoch": 0.82, "learning_rate": 1.6411511734397878e-05, "loss": 0.1326, "step": 12798 }, { "epoch": 0.82, "learning_rate": 1.640011457311279e-05, "loss": 0.1348, "step": 12799 }, { "epoch": 0.82, "learning_rate": 1.6388721017135865e-05, "loss": 0.1317, "step": 12800 }, { "epoch": 0.82, "learning_rate": 1.637733106695851e-05, "loss": 0.1438, "step": 12801 }, { "epoch": 0.82, "learning_rate": 1.636594472307188e-05, "loss": 0.1358, "step": 12802 }, { "epoch": 0.82, "learning_rate": 1.635456198596702e-05, "loss": 0.1466, "step": 12803 }, { "epoch": 0.82, "learning_rate": 1.6343182856134866e-05, "loss": 0.1314, "step": 12804 }, { "epoch": 0.82, "learning_rate": 1.6331807334066085e-05, "loss": 0.1297, "step": 12805 }, { "epoch": 0.82, "learning_rate": 1.6320435420251324e-05, "loss": 0.1374, "step": 12806 }, { "epoch": 0.82, "learning_rate": 1.6309067115180932e-05, "loss": 0.1264, "step": 12807 }, { "epoch": 0.82, "learning_rate": 1.629770241934525e-05, "loss": 0.1321, "step": 12808 }, { "epoch": 0.82, "learning_rate": 1.6286341333234324e-05, "loss": 0.1489, "step": 12809 }, { "epoch": 0.82, "learning_rate": 1.6274983857338143e-05, "loss": 0.1369, "step": 12810 }, { "epoch": 0.82, "learning_rate": 1.6263629992146524e-05, "loss": 0.1537, "step": 12811 }, { "epoch": 0.82, "learning_rate": 1.6252279738149067e-05, "loss": 0.1462, "step": 12812 }, { "epoch": 0.82, "learning_rate": 1.624093309583531e-05, "loss": 0.1457, "step": 12813 }, { "epoch": 0.82, "learning_rate": 1.6229590065694534e-05, "loss": 0.1301, "step": 12814 }, { "epoch": 0.82, "learning_rate": 1.621825064821596e-05, "loss": 0.1394, "step": 12815 }, { "epoch": 0.82, "learning_rate": 1.6206914843888565e-05, "loss": 0.1166, "step": 12816 }, { "epoch": 0.82, "learning_rate": 1.6195582653201236e-05, "loss": 0.1556, "step": 12817 }, { "epoch": 0.82, "learning_rate": 1.6184254076642713e-05, "loss": 0.1443, "step": 12818 }, { "epoch": 0.82, "learning_rate": 1.6172929114701497e-05, "loss": 0.1291, "step": 12819 }, { "epoch": 0.82, "learning_rate": 1.6161607767866026e-05, "loss": 0.1288, "step": 12820 }, { "epoch": 0.82, "learning_rate": 1.6150290036624505e-05, "loss": 0.1408, "step": 12821 }, { "epoch": 0.82, "learning_rate": 1.613897592146507e-05, "loss": 0.1321, "step": 12822 }, { "epoch": 0.82, "learning_rate": 1.6127665422875594e-05, "loss": 0.1472, "step": 12823 }, { "epoch": 0.82, "learning_rate": 1.6116358541343868e-05, "loss": 0.1287, "step": 12824 }, { "epoch": 0.82, "learning_rate": 1.610505527735755e-05, "loss": 0.1386, "step": 12825 }, { "epoch": 0.82, "learning_rate": 1.609375563140405e-05, "loss": 0.1225, "step": 12826 }, { "epoch": 0.82, "learning_rate": 1.6082459603970712e-05, "loss": 0.1281, "step": 12827 }, { "epoch": 0.82, "learning_rate": 1.6071167195544657e-05, "loss": 0.1416, "step": 12828 }, { "epoch": 0.82, "learning_rate": 1.6059878406612893e-05, "loss": 0.1454, "step": 12829 }, { "epoch": 0.82, "learning_rate": 1.6048593237662248e-05, "loss": 0.1586, "step": 12830 }, { "epoch": 0.82, "learning_rate": 1.6037311689179403e-05, "loss": 0.1407, "step": 12831 }, { "epoch": 0.82, "learning_rate": 1.6026033761650917e-05, "loss": 0.1331, "step": 12832 }, { "epoch": 0.82, "learning_rate": 1.60147594555631e-05, "loss": 0.1381, "step": 12833 }, { "epoch": 0.82, "learning_rate": 1.600348877140223e-05, "loss": 0.1389, "step": 12834 }, { "epoch": 0.82, "learning_rate": 1.599222170965431e-05, "loss": 0.1606, "step": 12835 }, { "epoch": 0.82, "learning_rate": 1.598095827080528e-05, "loss": 0.135, "step": 12836 }, { "epoch": 0.82, "learning_rate": 1.596969845534084e-05, "loss": 0.1385, "step": 12837 }, { "epoch": 0.82, "learning_rate": 1.595844226374661e-05, "loss": 0.1405, "step": 12838 }, { "epoch": 0.82, "learning_rate": 1.5947189696508035e-05, "loss": 0.1284, "step": 12839 }, { "epoch": 0.82, "learning_rate": 1.5935940754110347e-05, "loss": 0.1394, "step": 12840 }, { "epoch": 0.82, "learning_rate": 1.5924695437038717e-05, "loss": 0.1336, "step": 12841 }, { "epoch": 0.82, "learning_rate": 1.591345374577805e-05, "loss": 0.1379, "step": 12842 }, { "epoch": 0.82, "learning_rate": 1.5902215680813214e-05, "loss": 0.1403, "step": 12843 }, { "epoch": 0.82, "learning_rate": 1.5890981242628778e-05, "loss": 0.147, "step": 12844 }, { "epoch": 0.82, "learning_rate": 1.587975043170933e-05, "loss": 0.1236, "step": 12845 }, { "epoch": 0.82, "learning_rate": 1.5868523248539157e-05, "loss": 0.1332, "step": 12846 }, { "epoch": 0.82, "learning_rate": 1.5857299693602435e-05, "loss": 0.1325, "step": 12847 }, { "epoch": 0.82, "learning_rate": 1.5846079767383214e-05, "loss": 0.1301, "step": 12848 }, { "epoch": 0.82, "learning_rate": 1.5834863470365326e-05, "loss": 0.1511, "step": 12849 }, { "epoch": 0.82, "learning_rate": 1.5823650803032498e-05, "loss": 0.145, "step": 12850 }, { "epoch": 0.82, "learning_rate": 1.5812441765868292e-05, "loss": 0.1403, "step": 12851 }, { "epoch": 0.82, "learning_rate": 1.5801236359356132e-05, "loss": 0.1363, "step": 12852 }, { "epoch": 0.82, "learning_rate": 1.5790034583979218e-05, "loss": 0.1558, "step": 12853 }, { "epoch": 0.82, "learning_rate": 1.5778836440220635e-05, "loss": 0.131, "step": 12854 }, { "epoch": 0.82, "learning_rate": 1.5767641928563348e-05, "loss": 0.1366, "step": 12855 }, { "epoch": 0.82, "learning_rate": 1.5756451049490072e-05, "loss": 0.1281, "step": 12856 }, { "epoch": 0.82, "learning_rate": 1.574526380348347e-05, "loss": 0.1314, "step": 12857 }, { "epoch": 0.82, "learning_rate": 1.5734080191025968e-05, "loss": 0.1413, "step": 12858 }, { "epoch": 0.82, "learning_rate": 1.572290021259992e-05, "loss": 0.1364, "step": 12859 }, { "epoch": 0.82, "learning_rate": 1.5711723868687443e-05, "loss": 0.128, "step": 12860 }, { "epoch": 0.82, "learning_rate": 1.570055115977048e-05, "loss": 0.1276, "step": 12861 }, { "epoch": 0.82, "learning_rate": 1.568938208633094e-05, "loss": 0.1503, "step": 12862 }, { "epoch": 0.82, "learning_rate": 1.567821664885043e-05, "loss": 0.1383, "step": 12863 }, { "epoch": 0.82, "learning_rate": 1.566705484781049e-05, "loss": 0.1516, "step": 12864 }, { "epoch": 0.82, "learning_rate": 1.5655896683692495e-05, "loss": 0.1492, "step": 12865 }, { "epoch": 0.82, "learning_rate": 1.5644742156977666e-05, "loss": 0.1302, "step": 12866 }, { "epoch": 0.83, "learning_rate": 1.563359126814703e-05, "loss": 0.1352, "step": 12867 }, { "epoch": 0.83, "learning_rate": 1.562244401768144e-05, "loss": 0.1309, "step": 12868 }, { "epoch": 0.83, "learning_rate": 1.561130040606169e-05, "loss": 0.1356, "step": 12869 }, { "epoch": 0.83, "learning_rate": 1.5600160433768317e-05, "loss": 0.1235, "step": 12870 }, { "epoch": 0.83, "learning_rate": 1.5589024101281747e-05, "loss": 0.1313, "step": 12871 }, { "epoch": 0.83, "learning_rate": 1.5577891409082247e-05, "loss": 0.149, "step": 12872 }, { "epoch": 0.83, "learning_rate": 1.5566762357649956e-05, "loss": 0.1485, "step": 12873 }, { "epoch": 0.83, "learning_rate": 1.5555636947464792e-05, "loss": 0.1239, "step": 12874 }, { "epoch": 0.83, "learning_rate": 1.5544515179006513e-05, "loss": 0.133, "step": 12875 }, { "epoch": 0.83, "learning_rate": 1.5533397052754816e-05, "loss": 0.1285, "step": 12876 }, { "epoch": 0.83, "learning_rate": 1.552228256918914e-05, "loss": 0.1363, "step": 12877 }, { "epoch": 0.83, "learning_rate": 1.5511171728788797e-05, "loss": 0.1235, "step": 12878 }, { "epoch": 0.83, "learning_rate": 1.550006453203301e-05, "loss": 0.1458, "step": 12879 }, { "epoch": 0.83, "learning_rate": 1.548896097940071e-05, "loss": 0.137, "step": 12880 }, { "epoch": 0.83, "learning_rate": 1.5477861071370803e-05, "loss": 0.1173, "step": 12881 }, { "epoch": 0.83, "learning_rate": 1.546676480842194e-05, "loss": 0.1354, "step": 12882 }, { "epoch": 0.83, "learning_rate": 1.54556721910327e-05, "loss": 0.1376, "step": 12883 }, { "epoch": 0.83, "learning_rate": 1.5444583219681408e-05, "loss": 0.127, "step": 12884 }, { "epoch": 0.83, "learning_rate": 1.5433497894846306e-05, "loss": 0.1482, "step": 12885 }, { "epoch": 0.83, "learning_rate": 1.5422416217005498e-05, "loss": 0.1369, "step": 12886 }, { "epoch": 0.83, "learning_rate": 1.5411338186636815e-05, "loss": 0.1304, "step": 12887 }, { "epoch": 0.83, "learning_rate": 1.5400263804218072e-05, "loss": 0.1226, "step": 12888 }, { "epoch": 0.83, "learning_rate": 1.5389193070226816e-05, "loss": 0.145, "step": 12889 }, { "epoch": 0.83, "learning_rate": 1.537812598514049e-05, "loss": 0.1353, "step": 12890 }, { "epoch": 0.83, "learning_rate": 1.5367062549436385e-05, "loss": 0.1187, "step": 12891 }, { "epoch": 0.83, "learning_rate": 1.5356002763591636e-05, "loss": 0.1326, "step": 12892 }, { "epoch": 0.83, "learning_rate": 1.534494662808317e-05, "loss": 0.1521, "step": 12893 }, { "epoch": 0.83, "learning_rate": 1.5333894143387783e-05, "loss": 0.1415, "step": 12894 }, { "epoch": 0.83, "learning_rate": 1.532284530998217e-05, "loss": 0.1355, "step": 12895 }, { "epoch": 0.83, "learning_rate": 1.531180012834277e-05, "loss": 0.1349, "step": 12896 }, { "epoch": 0.83, "learning_rate": 1.530075859894593e-05, "loss": 0.138, "step": 12897 }, { "epoch": 0.83, "learning_rate": 1.5289720722267832e-05, "loss": 0.1279, "step": 12898 }, { "epoch": 0.83, "learning_rate": 1.527868649878451e-05, "loss": 0.1351, "step": 12899 }, { "epoch": 0.83, "learning_rate": 1.5267655928971814e-05, "loss": 0.1513, "step": 12900 }, { "epoch": 0.83, "learning_rate": 1.5256629013305402e-05, "loss": 0.1427, "step": 12901 }, { "epoch": 0.83, "learning_rate": 1.5245605752260883e-05, "loss": 0.1194, "step": 12902 }, { "epoch": 0.83, "learning_rate": 1.523458614631359e-05, "loss": 0.1286, "step": 12903 }, { "epoch": 0.83, "learning_rate": 1.5223570195938764e-05, "loss": 0.1302, "step": 12904 }, { "epoch": 0.83, "learning_rate": 1.5212557901611491e-05, "loss": 0.1396, "step": 12905 }, { "epoch": 0.83, "learning_rate": 1.5201549263806713e-05, "loss": 0.1296, "step": 12906 }, { "epoch": 0.83, "learning_rate": 1.5190544282999142e-05, "loss": 0.1365, "step": 12907 }, { "epoch": 0.83, "learning_rate": 1.5179542959663373e-05, "loss": 0.1442, "step": 12908 }, { "epoch": 0.83, "learning_rate": 1.5168545294273873e-05, "loss": 0.1293, "step": 12909 }, { "epoch": 0.83, "learning_rate": 1.5157551287304895e-05, "loss": 0.1374, "step": 12910 }, { "epoch": 0.83, "learning_rate": 1.514656093923058e-05, "loss": 0.1353, "step": 12911 }, { "epoch": 0.83, "learning_rate": 1.5135574250524897e-05, "loss": 0.1414, "step": 12912 }, { "epoch": 0.83, "learning_rate": 1.5124591221661678e-05, "loss": 0.1116, "step": 12913 }, { "epoch": 0.83, "learning_rate": 1.511361185311454e-05, "loss": 0.1239, "step": 12914 }, { "epoch": 0.83, "learning_rate": 1.5102636145356975e-05, "loss": 0.1268, "step": 12915 }, { "epoch": 0.83, "learning_rate": 1.5091664098862356e-05, "loss": 0.1432, "step": 12916 }, { "epoch": 0.83, "learning_rate": 1.508069571410381e-05, "loss": 0.1528, "step": 12917 }, { "epoch": 0.83, "learning_rate": 1.5069730991554376e-05, "loss": 0.1261, "step": 12918 }, { "epoch": 0.83, "learning_rate": 1.5058769931686922e-05, "loss": 0.1305, "step": 12919 }, { "epoch": 0.83, "learning_rate": 1.5047812534974181e-05, "loss": 0.1472, "step": 12920 }, { "epoch": 0.83, "learning_rate": 1.5036858801888676e-05, "loss": 0.1452, "step": 12921 }, { "epoch": 0.83, "learning_rate": 1.502590873290276e-05, "loss": 0.1346, "step": 12922 }, { "epoch": 0.83, "learning_rate": 1.5014962328488713e-05, "loss": 0.1297, "step": 12923 }, { "epoch": 0.83, "learning_rate": 1.5004019589118568e-05, "loss": 0.1407, "step": 12924 }, { "epoch": 0.83, "learning_rate": 1.4993080515264258e-05, "loss": 0.1222, "step": 12925 }, { "epoch": 0.83, "learning_rate": 1.4982145107397526e-05, "loss": 0.1402, "step": 12926 }, { "epoch": 0.83, "learning_rate": 1.4971213365990021e-05, "loss": 0.1237, "step": 12927 }, { "epoch": 0.83, "learning_rate": 1.4960285291513132e-05, "loss": 0.1286, "step": 12928 }, { "epoch": 0.83, "learning_rate": 1.4949360884438136e-05, "loss": 0.1409, "step": 12929 }, { "epoch": 0.83, "learning_rate": 1.4938440145236188e-05, "loss": 0.1481, "step": 12930 }, { "epoch": 0.83, "learning_rate": 1.4927523074378202e-05, "loss": 0.1491, "step": 12931 }, { "epoch": 0.83, "learning_rate": 1.4916609672335058e-05, "loss": 0.1437, "step": 12932 }, { "epoch": 0.83, "learning_rate": 1.4905699939577344e-05, "loss": 0.1416, "step": 12933 }, { "epoch": 0.83, "learning_rate": 1.4894793876575597e-05, "loss": 0.1338, "step": 12934 }, { "epoch": 0.83, "learning_rate": 1.4883891483800127e-05, "loss": 0.1359, "step": 12935 }, { "epoch": 0.83, "learning_rate": 1.4872992761721083e-05, "loss": 0.1362, "step": 12936 }, { "epoch": 0.83, "learning_rate": 1.4862097710808509e-05, "loss": 0.1307, "step": 12937 }, { "epoch": 0.83, "learning_rate": 1.4851206331532252e-05, "loss": 0.1351, "step": 12938 }, { "epoch": 0.83, "learning_rate": 1.4840318624362038e-05, "loss": 0.132, "step": 12939 }, { "epoch": 0.83, "learning_rate": 1.4829434589767366e-05, "loss": 0.1426, "step": 12940 }, { "epoch": 0.83, "learning_rate": 1.4818554228217662e-05, "loss": 0.1408, "step": 12941 }, { "epoch": 0.83, "learning_rate": 1.480767754018213e-05, "loss": 0.1437, "step": 12942 }, { "epoch": 0.83, "learning_rate": 1.4796804526129793e-05, "loss": 0.1339, "step": 12943 }, { "epoch": 0.83, "learning_rate": 1.4785935186529609e-05, "loss": 0.14, "step": 12944 }, { "epoch": 0.83, "learning_rate": 1.4775069521850316e-05, "loss": 0.122, "step": 12945 }, { "epoch": 0.83, "learning_rate": 1.4764207532560515e-05, "loss": 0.1364, "step": 12946 }, { "epoch": 0.83, "learning_rate": 1.475334921912861e-05, "loss": 0.1652, "step": 12947 }, { "epoch": 0.83, "learning_rate": 1.4742494582022903e-05, "loss": 0.1306, "step": 12948 }, { "epoch": 0.83, "learning_rate": 1.47316436217115e-05, "loss": 0.148, "step": 12949 }, { "epoch": 0.83, "learning_rate": 1.4720796338662324e-05, "loss": 0.117, "step": 12950 }, { "epoch": 0.83, "learning_rate": 1.4709952733343203e-05, "loss": 0.1382, "step": 12951 }, { "epoch": 0.83, "learning_rate": 1.469911280622177e-05, "loss": 0.121, "step": 12952 }, { "epoch": 0.83, "learning_rate": 1.4688276557765523e-05, "loss": 0.1425, "step": 12953 }, { "epoch": 0.83, "learning_rate": 1.4677443988441752e-05, "loss": 0.121, "step": 12954 }, { "epoch": 0.83, "learning_rate": 1.466661509871765e-05, "loss": 0.1372, "step": 12955 }, { "epoch": 0.83, "learning_rate": 1.4655789889060213e-05, "loss": 0.1436, "step": 12956 }, { "epoch": 0.83, "learning_rate": 1.4644968359936251e-05, "loss": 0.1254, "step": 12957 }, { "epoch": 0.83, "learning_rate": 1.4634150511812483e-05, "loss": 0.1307, "step": 12958 }, { "epoch": 0.83, "learning_rate": 1.4623336345155436e-05, "loss": 0.1436, "step": 12959 }, { "epoch": 0.83, "learning_rate": 1.4612525860431492e-05, "loss": 0.1386, "step": 12960 }, { "epoch": 0.83, "learning_rate": 1.4601719058106845e-05, "loss": 0.1428, "step": 12961 }, { "epoch": 0.83, "learning_rate": 1.459091593864752e-05, "loss": 0.149, "step": 12962 }, { "epoch": 0.83, "learning_rate": 1.458011650251947e-05, "loss": 0.1412, "step": 12963 }, { "epoch": 0.83, "learning_rate": 1.4569320750188365e-05, "loss": 0.1534, "step": 12964 }, { "epoch": 0.83, "learning_rate": 1.455852868211981e-05, "loss": 0.1332, "step": 12965 }, { "epoch": 0.83, "learning_rate": 1.4547740298779223e-05, "loss": 0.1343, "step": 12966 }, { "epoch": 0.83, "learning_rate": 1.4536955600631874e-05, "loss": 0.149, "step": 12967 }, { "epoch": 0.83, "learning_rate": 1.452617458814286e-05, "loss": 0.1354, "step": 12968 }, { "epoch": 0.83, "learning_rate": 1.4515397261777076e-05, "loss": 0.1208, "step": 12969 }, { "epoch": 0.83, "learning_rate": 1.4504623621999357e-05, "loss": 0.1323, "step": 12970 }, { "epoch": 0.83, "learning_rate": 1.449385366927427e-05, "loss": 0.14, "step": 12971 }, { "epoch": 0.83, "learning_rate": 1.448308740406632e-05, "loss": 0.1308, "step": 12972 }, { "epoch": 0.83, "learning_rate": 1.447232482683979e-05, "loss": 0.146, "step": 12973 }, { "epoch": 0.83, "learning_rate": 1.4461565938058863e-05, "loss": 0.1372, "step": 12974 }, { "epoch": 0.83, "learning_rate": 1.4450810738187492e-05, "loss": 0.1513, "step": 12975 }, { "epoch": 0.83, "learning_rate": 1.4440059227689473e-05, "loss": 0.1334, "step": 12976 }, { "epoch": 0.83, "learning_rate": 1.4429311407028512e-05, "loss": 0.1319, "step": 12977 }, { "epoch": 0.83, "learning_rate": 1.4418567276668104e-05, "loss": 0.1259, "step": 12978 }, { "epoch": 0.83, "learning_rate": 1.4407826837071635e-05, "loss": 0.1401, "step": 12979 }, { "epoch": 0.83, "learning_rate": 1.4397090088702236e-05, "loss": 0.1137, "step": 12980 }, { "epoch": 0.83, "learning_rate": 1.438635703202298e-05, "loss": 0.1378, "step": 12981 }, { "epoch": 0.83, "learning_rate": 1.4375627667496739e-05, "loss": 0.1331, "step": 12982 }, { "epoch": 0.83, "learning_rate": 1.4364901995586166e-05, "loss": 0.1278, "step": 12983 }, { "epoch": 0.83, "learning_rate": 1.435418001675387e-05, "loss": 0.132, "step": 12984 }, { "epoch": 0.83, "learning_rate": 1.4343461731462226e-05, "loss": 0.1396, "step": 12985 }, { "epoch": 0.83, "learning_rate": 1.4332747140173496e-05, "loss": 0.1254, "step": 12986 }, { "epoch": 0.83, "learning_rate": 1.43220362433497e-05, "loss": 0.1336, "step": 12987 }, { "epoch": 0.83, "learning_rate": 1.4311329041452826e-05, "loss": 0.1431, "step": 12988 }, { "epoch": 0.83, "learning_rate": 1.4300625534944578e-05, "loss": 0.1395, "step": 12989 }, { "epoch": 0.83, "learning_rate": 1.4289925724286535e-05, "loss": 0.1468, "step": 12990 }, { "epoch": 0.83, "learning_rate": 1.427922960994017e-05, "loss": 0.1287, "step": 12991 }, { "epoch": 0.83, "learning_rate": 1.426853719236676e-05, "loss": 0.1409, "step": 12992 }, { "epoch": 0.83, "learning_rate": 1.4257848472027435e-05, "loss": 0.1403, "step": 12993 }, { "epoch": 0.83, "learning_rate": 1.4247163449383117e-05, "loss": 0.1399, "step": 12994 }, { "epoch": 0.83, "learning_rate": 1.4236482124894646e-05, "loss": 0.1322, "step": 12995 }, { "epoch": 0.83, "learning_rate": 1.4225804499022643e-05, "loss": 0.1389, "step": 12996 }, { "epoch": 0.83, "learning_rate": 1.4215130572227564e-05, "loss": 0.1222, "step": 12997 }, { "epoch": 0.83, "learning_rate": 1.420446034496975e-05, "loss": 0.1414, "step": 12998 }, { "epoch": 0.83, "learning_rate": 1.4193793817709377e-05, "loss": 0.1384, "step": 12999 }, { "epoch": 0.83, "learning_rate": 1.4183130990906445e-05, "loss": 0.1296, "step": 13000 }, { "epoch": 0.83, "learning_rate": 1.4172471865020775e-05, "loss": 0.1176, "step": 13001 }, { "epoch": 0.83, "learning_rate": 1.4161816440512087e-05, "loss": 0.1241, "step": 13002 }, { "epoch": 0.83, "learning_rate": 1.4151164717839871e-05, "loss": 0.1559, "step": 13003 }, { "epoch": 0.83, "learning_rate": 1.414051669746349e-05, "loss": 0.1329, "step": 13004 }, { "epoch": 0.83, "learning_rate": 1.4129872379842145e-05, "loss": 0.1451, "step": 13005 }, { "epoch": 0.83, "learning_rate": 1.4119231765434904e-05, "loss": 0.1167, "step": 13006 }, { "epoch": 0.83, "learning_rate": 1.4108594854700651e-05, "loss": 0.139, "step": 13007 }, { "epoch": 0.83, "learning_rate": 1.4097961648098079e-05, "loss": 0.1342, "step": 13008 }, { "epoch": 0.83, "learning_rate": 1.4087332146085808e-05, "loss": 0.16, "step": 13009 }, { "epoch": 0.83, "learning_rate": 1.4076706349122193e-05, "loss": 0.1276, "step": 13010 }, { "epoch": 0.83, "learning_rate": 1.406608425766549e-05, "loss": 0.144, "step": 13011 }, { "epoch": 0.83, "learning_rate": 1.4055465872173778e-05, "loss": 0.1283, "step": 13012 }, { "epoch": 0.83, "learning_rate": 1.4044851193105002e-05, "loss": 0.1326, "step": 13013 }, { "epoch": 0.83, "learning_rate": 1.403424022091694e-05, "loss": 0.1206, "step": 13014 }, { "epoch": 0.83, "learning_rate": 1.4023632956067167e-05, "loss": 0.1168, "step": 13015 }, { "epoch": 0.83, "learning_rate": 1.4013029399013156e-05, "loss": 0.1279, "step": 13016 }, { "epoch": 0.83, "learning_rate": 1.4002429550212182e-05, "loss": 0.1354, "step": 13017 }, { "epoch": 0.83, "learning_rate": 1.3991833410121324e-05, "loss": 0.1259, "step": 13018 }, { "epoch": 0.83, "learning_rate": 1.398124097919763e-05, "loss": 0.1384, "step": 13019 }, { "epoch": 0.83, "learning_rate": 1.3970652257897853e-05, "loss": 0.1223, "step": 13020 }, { "epoch": 0.83, "learning_rate": 1.3960067246678688e-05, "loss": 0.1212, "step": 13021 }, { "epoch": 0.83, "learning_rate": 1.3949485945996555e-05, "loss": 0.1455, "step": 13022 }, { "epoch": 0.84, "learning_rate": 1.3938908356307845e-05, "loss": 0.1397, "step": 13023 }, { "epoch": 0.84, "learning_rate": 1.3928334478068672e-05, "loss": 0.1363, "step": 13024 }, { "epoch": 0.84, "learning_rate": 1.3917764311735071e-05, "loss": 0.1293, "step": 13025 }, { "epoch": 0.84, "learning_rate": 1.39071978577629e-05, "loss": 0.1342, "step": 13026 }, { "epoch": 0.84, "learning_rate": 1.3896635116607803e-05, "loss": 0.14, "step": 13027 }, { "epoch": 0.84, "learning_rate": 1.3886076088725363e-05, "loss": 0.1341, "step": 13028 }, { "epoch": 0.84, "learning_rate": 1.3875520774570882e-05, "loss": 0.1437, "step": 13029 }, { "epoch": 0.84, "learning_rate": 1.3864969174599629e-05, "loss": 0.1329, "step": 13030 }, { "epoch": 0.84, "learning_rate": 1.3854421289266595e-05, "loss": 0.1282, "step": 13031 }, { "epoch": 0.84, "learning_rate": 1.3843877119026693e-05, "loss": 0.1431, "step": 13032 }, { "epoch": 0.84, "learning_rate": 1.3833336664334661e-05, "loss": 0.1385, "step": 13033 }, { "epoch": 0.84, "learning_rate": 1.3822799925645036e-05, "loss": 0.1617, "step": 13034 }, { "epoch": 0.84, "learning_rate": 1.3812266903412252e-05, "loss": 0.1455, "step": 13035 }, { "epoch": 0.84, "learning_rate": 1.3801737598090537e-05, "loss": 0.1406, "step": 13036 }, { "epoch": 0.84, "learning_rate": 1.3791212010133947e-05, "loss": 0.1194, "step": 13037 }, { "epoch": 0.84, "learning_rate": 1.3780690139996433e-05, "loss": 0.1413, "step": 13038 }, { "epoch": 0.84, "learning_rate": 1.3770171988131775e-05, "loss": 0.1483, "step": 13039 }, { "epoch": 0.84, "learning_rate": 1.3759657554993565e-05, "loss": 0.141, "step": 13040 }, { "epoch": 0.84, "learning_rate": 1.3749146841035232e-05, "loss": 0.1317, "step": 13041 }, { "epoch": 0.84, "learning_rate": 1.3738639846710088e-05, "loss": 0.1322, "step": 13042 }, { "epoch": 0.84, "learning_rate": 1.3728136572471229e-05, "loss": 0.1369, "step": 13043 }, { "epoch": 0.84, "learning_rate": 1.37176370187716e-05, "loss": 0.1334, "step": 13044 }, { "epoch": 0.84, "learning_rate": 1.3707141186064031e-05, "loss": 0.1343, "step": 13045 }, { "epoch": 0.84, "learning_rate": 1.369664907480116e-05, "loss": 0.1207, "step": 13046 }, { "epoch": 0.84, "learning_rate": 1.3686160685435478e-05, "loss": 0.1329, "step": 13047 }, { "epoch": 0.84, "learning_rate": 1.367567601841927e-05, "loss": 0.1502, "step": 13048 }, { "epoch": 0.84, "learning_rate": 1.366519507420474e-05, "loss": 0.1288, "step": 13049 }, { "epoch": 0.84, "learning_rate": 1.3654717853243858e-05, "loss": 0.1319, "step": 13050 }, { "epoch": 0.84, "learning_rate": 1.3644244355988445e-05, "loss": 0.1358, "step": 13051 }, { "epoch": 0.84, "learning_rate": 1.3633774582890202e-05, "loss": 0.124, "step": 13052 }, { "epoch": 0.84, "learning_rate": 1.3623308534400648e-05, "loss": 0.1469, "step": 13053 }, { "epoch": 0.84, "learning_rate": 1.3612846210971154e-05, "loss": 0.1351, "step": 13054 }, { "epoch": 0.84, "learning_rate": 1.360238761305288e-05, "loss": 0.1212, "step": 13055 }, { "epoch": 0.84, "learning_rate": 1.3591932741096892e-05, "loss": 0.143, "step": 13056 }, { "epoch": 0.84, "learning_rate": 1.3581481595554046e-05, "loss": 0.1173, "step": 13057 }, { "epoch": 0.84, "learning_rate": 1.3571034176875042e-05, "loss": 0.1591, "step": 13058 }, { "epoch": 0.84, "learning_rate": 1.356059048551046e-05, "loss": 0.1362, "step": 13059 }, { "epoch": 0.84, "learning_rate": 1.3550150521910664e-05, "loss": 0.1511, "step": 13060 }, { "epoch": 0.84, "learning_rate": 1.353971428652594e-05, "loss": 0.1317, "step": 13061 }, { "epoch": 0.84, "learning_rate": 1.3529281779806291e-05, "loss": 0.1275, "step": 13062 }, { "epoch": 0.84, "learning_rate": 1.3518853002201693e-05, "loss": 0.1316, "step": 13063 }, { "epoch": 0.84, "learning_rate": 1.3508427954161829e-05, "loss": 0.1329, "step": 13064 }, { "epoch": 0.84, "learning_rate": 1.3498006636136307e-05, "loss": 0.1345, "step": 13065 }, { "epoch": 0.84, "learning_rate": 1.34875890485746e-05, "loss": 0.1298, "step": 13066 }, { "epoch": 0.84, "learning_rate": 1.3477175191925918e-05, "loss": 0.1344, "step": 13067 }, { "epoch": 0.84, "learning_rate": 1.3466765066639408e-05, "loss": 0.155, "step": 13068 }, { "epoch": 0.84, "learning_rate": 1.3456358673163972e-05, "loss": 0.1396, "step": 13069 }, { "epoch": 0.84, "learning_rate": 1.3445956011948436e-05, "loss": 0.1534, "step": 13070 }, { "epoch": 0.84, "learning_rate": 1.3435557083441374e-05, "loss": 0.1277, "step": 13071 }, { "epoch": 0.84, "learning_rate": 1.3425161888091287e-05, "loss": 0.1239, "step": 13072 }, { "epoch": 0.84, "learning_rate": 1.3414770426346479e-05, "loss": 0.1467, "step": 13073 }, { "epoch": 0.84, "learning_rate": 1.3404382698655061e-05, "loss": 0.1346, "step": 13074 }, { "epoch": 0.84, "learning_rate": 1.339399870546505e-05, "loss": 0.1377, "step": 13075 }, { "epoch": 0.84, "learning_rate": 1.338361844722421e-05, "loss": 0.1373, "step": 13076 }, { "epoch": 0.84, "learning_rate": 1.3373241924380254e-05, "loss": 0.1304, "step": 13077 }, { "epoch": 0.84, "learning_rate": 1.3362869137380618e-05, "loss": 0.1292, "step": 13078 }, { "epoch": 0.84, "learning_rate": 1.3352500086672682e-05, "loss": 0.1444, "step": 13079 }, { "epoch": 0.84, "learning_rate": 1.3342134772703618e-05, "loss": 0.1431, "step": 13080 }, { "epoch": 0.84, "learning_rate": 1.3331773195920416e-05, "loss": 0.1243, "step": 13081 }, { "epoch": 0.84, "learning_rate": 1.332141535676995e-05, "loss": 0.1311, "step": 13082 }, { "epoch": 0.84, "learning_rate": 1.3311061255698876e-05, "loss": 0.1407, "step": 13083 }, { "epoch": 0.84, "learning_rate": 1.3300710893153756e-05, "loss": 0.1257, "step": 13084 }, { "epoch": 0.84, "learning_rate": 1.3290364269580935e-05, "loss": 0.1346, "step": 13085 }, { "epoch": 0.84, "learning_rate": 1.3280021385426622e-05, "loss": 0.1319, "step": 13086 }, { "epoch": 0.84, "learning_rate": 1.3269682241136882e-05, "loss": 0.1233, "step": 13087 }, { "epoch": 0.84, "learning_rate": 1.325934683715757e-05, "loss": 0.1314, "step": 13088 }, { "epoch": 0.84, "learning_rate": 1.3249015173934454e-05, "loss": 0.1184, "step": 13089 }, { "epoch": 0.84, "learning_rate": 1.3238687251913028e-05, "loss": 0.1354, "step": 13090 }, { "epoch": 0.84, "learning_rate": 1.3228363071538752e-05, "loss": 0.1371, "step": 13091 }, { "epoch": 0.84, "learning_rate": 1.3218042633256811e-05, "loss": 0.1454, "step": 13092 }, { "epoch": 0.84, "learning_rate": 1.3207725937512316e-05, "loss": 0.1443, "step": 13093 }, { "epoch": 0.84, "learning_rate": 1.319741298475019e-05, "loss": 0.1262, "step": 13094 }, { "epoch": 0.84, "learning_rate": 1.3187103775415156e-05, "loss": 0.1299, "step": 13095 }, { "epoch": 0.84, "learning_rate": 1.3176798309951843e-05, "loss": 0.139, "step": 13096 }, { "epoch": 0.84, "learning_rate": 1.3166496588804644e-05, "loss": 0.131, "step": 13097 }, { "epoch": 0.84, "learning_rate": 1.315619861241787e-05, "loss": 0.1378, "step": 13098 }, { "epoch": 0.84, "learning_rate": 1.3145904381235574e-05, "loss": 0.1288, "step": 13099 }, { "epoch": 0.84, "learning_rate": 1.3135613895701736e-05, "loss": 0.1442, "step": 13100 }, { "epoch": 0.84, "learning_rate": 1.3125327156260159e-05, "loss": 0.1193, "step": 13101 }, { "epoch": 0.84, "learning_rate": 1.3115044163354428e-05, "loss": 0.1332, "step": 13102 }, { "epoch": 0.84, "learning_rate": 1.3104764917428048e-05, "loss": 0.1556, "step": 13103 }, { "epoch": 0.84, "learning_rate": 1.309448941892426e-05, "loss": 0.1372, "step": 13104 }, { "epoch": 0.84, "learning_rate": 1.3084217668286246e-05, "loss": 0.1297, "step": 13105 }, { "epoch": 0.84, "learning_rate": 1.3073949665956998e-05, "loss": 0.1197, "step": 13106 }, { "epoch": 0.84, "learning_rate": 1.3063685412379278e-05, "loss": 0.128, "step": 13107 }, { "epoch": 0.84, "learning_rate": 1.3053424907995792e-05, "loss": 0.1252, "step": 13108 }, { "epoch": 0.84, "learning_rate": 1.3043168153248997e-05, "loss": 0.1158, "step": 13109 }, { "epoch": 0.84, "learning_rate": 1.3032915148581249e-05, "loss": 0.1381, "step": 13110 }, { "epoch": 0.84, "learning_rate": 1.3022665894434682e-05, "loss": 0.1204, "step": 13111 }, { "epoch": 0.84, "learning_rate": 1.3012420391251312e-05, "loss": 0.1553, "step": 13112 }, { "epoch": 0.84, "learning_rate": 1.3002178639473028e-05, "loss": 0.1397, "step": 13113 }, { "epoch": 0.84, "learning_rate": 1.2991940639541467e-05, "loss": 0.1418, "step": 13114 }, { "epoch": 0.84, "learning_rate": 1.2981706391898175e-05, "loss": 0.1413, "step": 13115 }, { "epoch": 0.84, "learning_rate": 1.2971475896984475e-05, "loss": 0.1189, "step": 13116 }, { "epoch": 0.84, "learning_rate": 1.2961249155241628e-05, "loss": 0.1514, "step": 13117 }, { "epoch": 0.84, "learning_rate": 1.2951026167110603e-05, "loss": 0.1282, "step": 13118 }, { "epoch": 0.84, "learning_rate": 1.2940806933032302e-05, "loss": 0.1224, "step": 13119 }, { "epoch": 0.84, "learning_rate": 1.293059145344746e-05, "loss": 0.1243, "step": 13120 }, { "epoch": 0.84, "learning_rate": 1.2920379728796584e-05, "loss": 0.1446, "step": 13121 }, { "epoch": 0.84, "learning_rate": 1.2910171759520106e-05, "loss": 0.1349, "step": 13122 }, { "epoch": 0.84, "learning_rate": 1.2899967546058212e-05, "loss": 0.1313, "step": 13123 }, { "epoch": 0.84, "learning_rate": 1.2889767088851001e-05, "loss": 0.1358, "step": 13124 }, { "epoch": 0.84, "learning_rate": 1.2879570388338347e-05, "loss": 0.141, "step": 13125 }, { "epoch": 0.84, "learning_rate": 1.2869377444959997e-05, "loss": 0.1343, "step": 13126 }, { "epoch": 0.84, "learning_rate": 1.2859188259155564e-05, "loss": 0.1392, "step": 13127 }, { "epoch": 0.84, "learning_rate": 1.2849002831364409e-05, "loss": 0.1366, "step": 13128 }, { "epoch": 0.84, "learning_rate": 1.2838821162025838e-05, "loss": 0.1067, "step": 13129 }, { "epoch": 0.84, "learning_rate": 1.2828643251578887e-05, "loss": 0.1499, "step": 13130 }, { "epoch": 0.84, "learning_rate": 1.2818469100462549e-05, "loss": 0.1157, "step": 13131 }, { "epoch": 0.84, "learning_rate": 1.280829870911554e-05, "loss": 0.1321, "step": 13132 }, { "epoch": 0.84, "learning_rate": 1.2798132077976477e-05, "loss": 0.1522, "step": 13133 }, { "epoch": 0.84, "learning_rate": 1.2787969207483841e-05, "loss": 0.1227, "step": 13134 }, { "epoch": 0.84, "learning_rate": 1.277781009807586e-05, "loss": 0.1455, "step": 13135 }, { "epoch": 0.84, "learning_rate": 1.2767654750190705e-05, "loss": 0.1536, "step": 13136 }, { "epoch": 0.84, "learning_rate": 1.2757503164266283e-05, "loss": 0.1288, "step": 13137 }, { "epoch": 0.84, "learning_rate": 1.2747355340740441e-05, "loss": 0.1508, "step": 13138 }, { "epoch": 0.84, "learning_rate": 1.2737211280050753e-05, "loss": 0.1261, "step": 13139 }, { "epoch": 0.84, "learning_rate": 1.2727070982634714e-05, "loss": 0.1453, "step": 13140 }, { "epoch": 0.84, "learning_rate": 1.2716934448929662e-05, "loss": 0.1337, "step": 13141 }, { "epoch": 0.84, "learning_rate": 1.2706801679372704e-05, "loss": 0.1577, "step": 13142 }, { "epoch": 0.84, "learning_rate": 1.2696672674400856e-05, "loss": 0.1322, "step": 13143 }, { "epoch": 0.84, "learning_rate": 1.2686547434450891e-05, "loss": 0.1399, "step": 13144 }, { "epoch": 0.84, "learning_rate": 1.2676425959959527e-05, "loss": 0.1371, "step": 13145 }, { "epoch": 0.84, "learning_rate": 1.2666308251363202e-05, "loss": 0.1238, "step": 13146 }, { "epoch": 0.84, "learning_rate": 1.265619430909829e-05, "loss": 0.1234, "step": 13147 }, { "epoch": 0.84, "learning_rate": 1.2646084133600955e-05, "loss": 0.1322, "step": 13148 }, { "epoch": 0.84, "learning_rate": 1.263597772530719e-05, "loss": 0.1238, "step": 13149 }, { "epoch": 0.84, "learning_rate": 1.2625875084652871e-05, "loss": 0.1517, "step": 13150 }, { "epoch": 0.84, "learning_rate": 1.261577621207365e-05, "loss": 0.1345, "step": 13151 }, { "epoch": 0.84, "learning_rate": 1.2605681108005052e-05, "loss": 0.1387, "step": 13152 }, { "epoch": 0.84, "learning_rate": 1.2595589772882465e-05, "loss": 0.1324, "step": 13153 }, { "epoch": 0.84, "learning_rate": 1.2585502207141054e-05, "loss": 0.1278, "step": 13154 }, { "epoch": 0.84, "learning_rate": 1.2575418411215889e-05, "loss": 0.1358, "step": 13155 }, { "epoch": 0.84, "learning_rate": 1.2565338385541792e-05, "loss": 0.1485, "step": 13156 }, { "epoch": 0.84, "learning_rate": 1.2555262130553513e-05, "loss": 0.1159, "step": 13157 }, { "epoch": 0.84, "learning_rate": 1.2545189646685574e-05, "loss": 0.1199, "step": 13158 }, { "epoch": 0.84, "learning_rate": 1.2535120934372357e-05, "loss": 0.1282, "step": 13159 }, { "epoch": 0.84, "learning_rate": 1.2525055994048119e-05, "loss": 0.1337, "step": 13160 }, { "epoch": 0.84, "learning_rate": 1.2514994826146865e-05, "loss": 0.1517, "step": 13161 }, { "epoch": 0.84, "learning_rate": 1.2504937431102547e-05, "loss": 0.1395, "step": 13162 }, { "epoch": 0.84, "learning_rate": 1.2494883809348846e-05, "loss": 0.1278, "step": 13163 }, { "epoch": 0.84, "learning_rate": 1.2484833961319364e-05, "loss": 0.143, "step": 13164 }, { "epoch": 0.84, "learning_rate": 1.2474787887447492e-05, "loss": 0.1447, "step": 13165 }, { "epoch": 0.84, "learning_rate": 1.246474558816647e-05, "loss": 0.1154, "step": 13166 }, { "epoch": 0.84, "learning_rate": 1.245470706390942e-05, "loss": 0.1365, "step": 13167 }, { "epoch": 0.84, "learning_rate": 1.2444672315109207e-05, "loss": 0.1333, "step": 13168 }, { "epoch": 0.84, "learning_rate": 1.2434641342198639e-05, "loss": 0.1402, "step": 13169 }, { "epoch": 0.84, "learning_rate": 1.2424614145610259e-05, "loss": 0.1189, "step": 13170 }, { "epoch": 0.84, "learning_rate": 1.241459072577653e-05, "loss": 0.1493, "step": 13171 }, { "epoch": 0.84, "learning_rate": 1.2404571083129701e-05, "loss": 0.1378, "step": 13172 }, { "epoch": 0.84, "learning_rate": 1.2394555218101889e-05, "loss": 0.1366, "step": 13173 }, { "epoch": 0.84, "learning_rate": 1.2384543131125048e-05, "loss": 0.1161, "step": 13174 }, { "epoch": 0.84, "learning_rate": 1.2374534822630924e-05, "loss": 0.1528, "step": 13175 }, { "epoch": 0.84, "learning_rate": 1.2364530293051157e-05, "loss": 0.1414, "step": 13176 }, { "epoch": 0.84, "learning_rate": 1.2354529542817184e-05, "loss": 0.1358, "step": 13177 }, { "epoch": 0.84, "learning_rate": 1.2344532572360323e-05, "loss": 0.1297, "step": 13178 }, { "epoch": 0.85, "learning_rate": 1.2334539382111665e-05, "loss": 0.1306, "step": 13179 }, { "epoch": 0.85, "learning_rate": 1.2324549972502185e-05, "loss": 0.1302, "step": 13180 }, { "epoch": 0.85, "learning_rate": 1.231456434396272e-05, "loss": 0.1142, "step": 13181 }, { "epoch": 0.85, "learning_rate": 1.2304582496923844e-05, "loss": 0.1338, "step": 13182 }, { "epoch": 0.85, "learning_rate": 1.2294604431816092e-05, "loss": 0.1281, "step": 13183 }, { "epoch": 0.85, "learning_rate": 1.2284630149069731e-05, "loss": 0.1519, "step": 13184 }, { "epoch": 0.85, "learning_rate": 1.227465964911495e-05, "loss": 0.1269, "step": 13185 }, { "epoch": 0.85, "learning_rate": 1.2264692932381682e-05, "loss": 0.1143, "step": 13186 }, { "epoch": 0.85, "learning_rate": 1.2254729999299785e-05, "loss": 0.1329, "step": 13187 }, { "epoch": 0.85, "learning_rate": 1.2244770850298926e-05, "loss": 0.1329, "step": 13188 }, { "epoch": 0.85, "learning_rate": 1.223481548580857e-05, "loss": 0.1418, "step": 13189 }, { "epoch": 0.85, "learning_rate": 1.2224863906258088e-05, "loss": 0.1305, "step": 13190 }, { "epoch": 0.85, "learning_rate": 1.2214916112076602e-05, "loss": 0.1339, "step": 13191 }, { "epoch": 0.85, "learning_rate": 1.2204972103693146e-05, "loss": 0.119, "step": 13192 }, { "epoch": 0.85, "learning_rate": 1.219503188153659e-05, "loss": 0.1618, "step": 13193 }, { "epoch": 0.85, "learning_rate": 1.2185095446035555e-05, "loss": 0.1416, "step": 13194 }, { "epoch": 0.85, "learning_rate": 1.2175162797618611e-05, "loss": 0.1304, "step": 13195 }, { "epoch": 0.85, "learning_rate": 1.216523393671406e-05, "loss": 0.116, "step": 13196 }, { "epoch": 0.85, "learning_rate": 1.2155308863750137e-05, "loss": 0.1409, "step": 13197 }, { "epoch": 0.85, "learning_rate": 1.2145387579154832e-05, "loss": 0.1463, "step": 13198 }, { "epoch": 0.85, "learning_rate": 1.2135470083356027e-05, "loss": 0.1346, "step": 13199 }, { "epoch": 0.85, "learning_rate": 1.2125556376781444e-05, "loss": 0.1346, "step": 13200 }, { "epoch": 0.85, "learning_rate": 1.2115646459858577e-05, "loss": 0.1589, "step": 13201 }, { "epoch": 0.85, "learning_rate": 1.2105740333014826e-05, "loss": 0.1373, "step": 13202 }, { "epoch": 0.85, "learning_rate": 1.2095837996677372e-05, "loss": 0.1512, "step": 13203 }, { "epoch": 0.85, "learning_rate": 1.2085939451273298e-05, "loss": 0.1586, "step": 13204 }, { "epoch": 0.85, "learning_rate": 1.2076044697229461e-05, "loss": 0.1294, "step": 13205 }, { "epoch": 0.85, "learning_rate": 1.2066153734972573e-05, "loss": 0.1498, "step": 13206 }, { "epoch": 0.85, "learning_rate": 1.205626656492922e-05, "loss": 0.1233, "step": 13207 }, { "epoch": 0.85, "learning_rate": 1.2046383187525767e-05, "loss": 0.1328, "step": 13208 }, { "epoch": 0.85, "learning_rate": 1.2036503603188464e-05, "loss": 0.1421, "step": 13209 }, { "epoch": 0.85, "learning_rate": 1.202662781234335e-05, "loss": 0.1413, "step": 13210 }, { "epoch": 0.85, "learning_rate": 1.2016755815416347e-05, "loss": 0.1328, "step": 13211 }, { "epoch": 0.85, "learning_rate": 1.2006887612833174e-05, "loss": 0.1321, "step": 13212 }, { "epoch": 0.85, "learning_rate": 1.1997023205019419e-05, "loss": 0.1385, "step": 13213 }, { "epoch": 0.85, "learning_rate": 1.1987162592400503e-05, "loss": 0.1285, "step": 13214 }, { "epoch": 0.85, "learning_rate": 1.1977305775401626e-05, "loss": 0.1387, "step": 13215 }, { "epoch": 0.85, "learning_rate": 1.1967452754447939e-05, "loss": 0.136, "step": 13216 }, { "epoch": 0.85, "learning_rate": 1.195760352996429e-05, "loss": 0.1368, "step": 13217 }, { "epoch": 0.85, "learning_rate": 1.1947758102375495e-05, "loss": 0.1368, "step": 13218 }, { "epoch": 0.85, "learning_rate": 1.1937916472106104e-05, "loss": 0.123, "step": 13219 }, { "epoch": 0.85, "learning_rate": 1.1928078639580553e-05, "loss": 0.1454, "step": 13220 }, { "epoch": 0.85, "learning_rate": 1.1918244605223128e-05, "loss": 0.1181, "step": 13221 }, { "epoch": 0.85, "learning_rate": 1.1908414369457899e-05, "loss": 0.1282, "step": 13222 }, { "epoch": 0.85, "learning_rate": 1.1898587932708838e-05, "loss": 0.1564, "step": 13223 }, { "epoch": 0.85, "learning_rate": 1.1888765295399674e-05, "loss": 0.1329, "step": 13224 }, { "epoch": 0.85, "learning_rate": 1.1878946457954055e-05, "loss": 0.1287, "step": 13225 }, { "epoch": 0.85, "learning_rate": 1.1869131420795388e-05, "loss": 0.1362, "step": 13226 }, { "epoch": 0.85, "learning_rate": 1.1859320184346978e-05, "loss": 0.1401, "step": 13227 }, { "epoch": 0.85, "learning_rate": 1.1849512749031966e-05, "loss": 0.126, "step": 13228 }, { "epoch": 0.85, "learning_rate": 1.1839709115273245e-05, "loss": 0.1312, "step": 13229 }, { "epoch": 0.85, "learning_rate": 1.1829909283493667e-05, "loss": 0.1203, "step": 13230 }, { "epoch": 0.85, "learning_rate": 1.1820113254115806e-05, "loss": 0.1381, "step": 13231 }, { "epoch": 0.85, "learning_rate": 1.1810321027562154e-05, "loss": 0.1465, "step": 13232 }, { "epoch": 0.85, "learning_rate": 1.1800532604254965e-05, "loss": 0.1454, "step": 13233 }, { "epoch": 0.85, "learning_rate": 1.1790747984616445e-05, "loss": 0.1285, "step": 13234 }, { "epoch": 0.85, "learning_rate": 1.1780967169068524e-05, "loss": 0.137, "step": 13235 }, { "epoch": 0.85, "learning_rate": 1.1771190158032986e-05, "loss": 0.1281, "step": 13236 }, { "epoch": 0.85, "learning_rate": 1.1761416951931514e-05, "loss": 0.1412, "step": 13237 }, { "epoch": 0.85, "learning_rate": 1.1751647551185541e-05, "loss": 0.1642, "step": 13238 }, { "epoch": 0.85, "learning_rate": 1.1741881956216405e-05, "loss": 0.1362, "step": 13239 }, { "epoch": 0.85, "learning_rate": 1.1732120167445248e-05, "loss": 0.1178, "step": 13240 }, { "epoch": 0.85, "learning_rate": 1.172236218529308e-05, "loss": 0.1285, "step": 13241 }, { "epoch": 0.85, "learning_rate": 1.1712608010180703e-05, "loss": 0.1541, "step": 13242 }, { "epoch": 0.85, "learning_rate": 1.1702857642528741e-05, "loss": 0.1114, "step": 13243 }, { "epoch": 0.85, "learning_rate": 1.1693111082757746e-05, "loss": 0.137, "step": 13244 }, { "epoch": 0.85, "learning_rate": 1.1683368331287992e-05, "loss": 0.1222, "step": 13245 }, { "epoch": 0.85, "learning_rate": 1.1673629388539665e-05, "loss": 0.1264, "step": 13246 }, { "epoch": 0.85, "learning_rate": 1.1663894254932762e-05, "loss": 0.1269, "step": 13247 }, { "epoch": 0.85, "learning_rate": 1.1654162930887136e-05, "loss": 0.1357, "step": 13248 }, { "epoch": 0.85, "learning_rate": 1.1644435416822452e-05, "loss": 0.1484, "step": 13249 }, { "epoch": 0.85, "learning_rate": 1.1634711713158174e-05, "loss": 0.1425, "step": 13250 }, { "epoch": 0.85, "learning_rate": 1.1624991820313701e-05, "loss": 0.1316, "step": 13251 }, { "epoch": 0.85, "learning_rate": 1.1615275738708176e-05, "loss": 0.1361, "step": 13252 }, { "epoch": 0.85, "learning_rate": 1.1605563468760616e-05, "loss": 0.1311, "step": 13253 }, { "epoch": 0.85, "learning_rate": 1.1595855010889866e-05, "loss": 0.131, "step": 13254 }, { "epoch": 0.85, "learning_rate": 1.1586150365514658e-05, "loss": 0.1377, "step": 13255 }, { "epoch": 0.85, "learning_rate": 1.1576449533053469e-05, "loss": 0.1277, "step": 13256 }, { "epoch": 0.85, "learning_rate": 1.156675251392464e-05, "loss": 0.1256, "step": 13257 }, { "epoch": 0.85, "learning_rate": 1.1557059308546402e-05, "loss": 0.1278, "step": 13258 }, { "epoch": 0.85, "learning_rate": 1.1547369917336747e-05, "loss": 0.1384, "step": 13259 }, { "epoch": 0.85, "learning_rate": 1.153768434071355e-05, "loss": 0.1364, "step": 13260 }, { "epoch": 0.85, "learning_rate": 1.152800257909451e-05, "loss": 0.1425, "step": 13261 }, { "epoch": 0.85, "learning_rate": 1.1518324632897192e-05, "loss": 0.138, "step": 13262 }, { "epoch": 0.85, "learning_rate": 1.150865050253893e-05, "loss": 0.1371, "step": 13263 }, { "epoch": 0.85, "learning_rate": 1.1498980188436914e-05, "loss": 0.1449, "step": 13264 }, { "epoch": 0.85, "learning_rate": 1.1489313691008219e-05, "loss": 0.1274, "step": 13265 }, { "epoch": 0.85, "learning_rate": 1.1479651010669678e-05, "loss": 0.1364, "step": 13266 }, { "epoch": 0.85, "learning_rate": 1.1469992147838037e-05, "loss": 0.1332, "step": 13267 }, { "epoch": 0.85, "learning_rate": 1.146033710292983e-05, "loss": 0.1228, "step": 13268 }, { "epoch": 0.85, "learning_rate": 1.1450685876361467e-05, "loss": 0.1317, "step": 13269 }, { "epoch": 0.85, "learning_rate": 1.1441038468549136e-05, "loss": 0.1416, "step": 13270 }, { "epoch": 0.85, "learning_rate": 1.1431394879908863e-05, "loss": 0.1305, "step": 13271 }, { "epoch": 0.85, "learning_rate": 1.142175511085659e-05, "loss": 0.1352, "step": 13272 }, { "epoch": 0.85, "learning_rate": 1.1412119161807988e-05, "loss": 0.1336, "step": 13273 }, { "epoch": 0.85, "learning_rate": 1.1402487033178632e-05, "loss": 0.1498, "step": 13274 }, { "epoch": 0.85, "learning_rate": 1.1392858725383948e-05, "loss": 0.1323, "step": 13275 }, { "epoch": 0.85, "learning_rate": 1.1383234238839113e-05, "loss": 0.118, "step": 13276 }, { "epoch": 0.85, "learning_rate": 1.137361357395924e-05, "loss": 0.1071, "step": 13277 }, { "epoch": 0.85, "learning_rate": 1.1363996731159188e-05, "loss": 0.1478, "step": 13278 }, { "epoch": 0.85, "learning_rate": 1.135438371085369e-05, "loss": 0.1574, "step": 13279 }, { "epoch": 0.85, "learning_rate": 1.1344774513457335e-05, "loss": 0.128, "step": 13280 }, { "epoch": 0.85, "learning_rate": 1.1335169139384538e-05, "loss": 0.1344, "step": 13281 }, { "epoch": 0.85, "learning_rate": 1.1325567589049534e-05, "loss": 0.1447, "step": 13282 }, { "epoch": 0.85, "learning_rate": 1.131596986286635e-05, "loss": 0.1337, "step": 13283 }, { "epoch": 0.85, "learning_rate": 1.1306375961248961e-05, "loss": 0.1488, "step": 13284 }, { "epoch": 0.85, "learning_rate": 1.129678588461105e-05, "loss": 0.1402, "step": 13285 }, { "epoch": 0.85, "learning_rate": 1.128719963336623e-05, "loss": 0.134, "step": 13286 }, { "epoch": 0.85, "learning_rate": 1.1277617207927915e-05, "loss": 0.1364, "step": 13287 }, { "epoch": 0.85, "learning_rate": 1.1268038608709364e-05, "loss": 0.1338, "step": 13288 }, { "epoch": 0.85, "learning_rate": 1.1258463836123645e-05, "loss": 0.1343, "step": 13289 }, { "epoch": 0.85, "learning_rate": 1.1248892890583673e-05, "loss": 0.1321, "step": 13290 }, { "epoch": 0.85, "learning_rate": 1.1239325772502218e-05, "loss": 0.1355, "step": 13291 }, { "epoch": 0.85, "learning_rate": 1.122976248229185e-05, "loss": 0.1412, "step": 13292 }, { "epoch": 0.85, "learning_rate": 1.1220203020364994e-05, "loss": 0.1438, "step": 13293 }, { "epoch": 0.85, "learning_rate": 1.1210647387133921e-05, "loss": 0.1279, "step": 13294 }, { "epoch": 0.85, "learning_rate": 1.1201095583010756e-05, "loss": 0.1266, "step": 13295 }, { "epoch": 0.85, "learning_rate": 1.1191547608407382e-05, "loss": 0.1412, "step": 13296 }, { "epoch": 0.85, "learning_rate": 1.1182003463735557e-05, "loss": 0.132, "step": 13297 }, { "epoch": 0.85, "learning_rate": 1.1172463149406919e-05, "loss": 0.1239, "step": 13298 }, { "epoch": 0.85, "learning_rate": 1.1162926665832863e-05, "loss": 0.1486, "step": 13299 }, { "epoch": 0.85, "learning_rate": 1.1153394013424667e-05, "loss": 0.1286, "step": 13300 }, { "epoch": 0.85, "learning_rate": 1.1143865192593438e-05, "loss": 0.1409, "step": 13301 }, { "epoch": 0.85, "learning_rate": 1.1134340203750149e-05, "loss": 0.1196, "step": 13302 }, { "epoch": 0.85, "learning_rate": 1.1124819047305534e-05, "loss": 0.1429, "step": 13303 }, { "epoch": 0.85, "learning_rate": 1.1115301723670179e-05, "loss": 0.1315, "step": 13304 }, { "epoch": 0.85, "learning_rate": 1.1105788233254578e-05, "loss": 0.1278, "step": 13305 }, { "epoch": 0.85, "learning_rate": 1.1096278576468966e-05, "loss": 0.1511, "step": 13306 }, { "epoch": 0.85, "learning_rate": 1.1086772753723463e-05, "loss": 0.1472, "step": 13307 }, { "epoch": 0.85, "learning_rate": 1.1077270765428027e-05, "loss": 0.1211, "step": 13308 }, { "epoch": 0.85, "learning_rate": 1.1067772611992444e-05, "loss": 0.1354, "step": 13309 }, { "epoch": 0.85, "learning_rate": 1.105827829382633e-05, "loss": 0.1307, "step": 13310 }, { "epoch": 0.85, "learning_rate": 1.10487878113391e-05, "loss": 0.1395, "step": 13311 }, { "epoch": 0.85, "learning_rate": 1.1039301164940075e-05, "loss": 0.147, "step": 13312 }, { "epoch": 0.85, "learning_rate": 1.1029818355038346e-05, "loss": 0.1459, "step": 13313 }, { "epoch": 0.85, "learning_rate": 1.1020339382042888e-05, "loss": 0.1493, "step": 13314 }, { "epoch": 0.85, "learning_rate": 1.1010864246362461e-05, "loss": 0.1209, "step": 13315 }, { "epoch": 0.85, "learning_rate": 1.1001392948405743e-05, "loss": 0.1208, "step": 13316 }, { "epoch": 0.85, "learning_rate": 1.0991925488581156e-05, "loss": 0.1375, "step": 13317 }, { "epoch": 0.85, "learning_rate": 1.0982461867296967e-05, "loss": 0.1277, "step": 13318 }, { "epoch": 0.85, "learning_rate": 1.0973002084961337e-05, "loss": 0.1554, "step": 13319 }, { "epoch": 0.85, "learning_rate": 1.0963546141982184e-05, "loss": 0.1313, "step": 13320 }, { "epoch": 0.85, "learning_rate": 1.095409403876737e-05, "loss": 0.116, "step": 13321 }, { "epoch": 0.85, "learning_rate": 1.094464577572447e-05, "loss": 0.1379, "step": 13322 }, { "epoch": 0.85, "learning_rate": 1.0935201353260994e-05, "loss": 0.1207, "step": 13323 }, { "epoch": 0.85, "learning_rate": 1.0925760771784199e-05, "loss": 0.1542, "step": 13324 }, { "epoch": 0.85, "learning_rate": 1.0916324031701209e-05, "loss": 0.1282, "step": 13325 }, { "epoch": 0.85, "learning_rate": 1.0906891133419017e-05, "loss": 0.1362, "step": 13326 }, { "epoch": 0.85, "learning_rate": 1.0897462077344411e-05, "loss": 0.1418, "step": 13327 }, { "epoch": 0.85, "learning_rate": 1.0888036863884054e-05, "loss": 0.1336, "step": 13328 }, { "epoch": 0.85, "learning_rate": 1.0878615493444377e-05, "loss": 0.1294, "step": 13329 }, { "epoch": 0.85, "learning_rate": 1.086919796643172e-05, "loss": 0.1306, "step": 13330 }, { "epoch": 0.85, "learning_rate": 1.0859784283252205e-05, "loss": 0.1457, "step": 13331 }, { "epoch": 0.85, "learning_rate": 1.0850374444311784e-05, "loss": 0.1272, "step": 13332 }, { "epoch": 0.85, "learning_rate": 1.0840968450016276e-05, "loss": 0.1329, "step": 13333 }, { "epoch": 0.85, "learning_rate": 1.0831566300771323e-05, "loss": 0.1402, "step": 13334 }, { "epoch": 0.86, "learning_rate": 1.0822167996982423e-05, "loss": 0.1229, "step": 13335 }, { "epoch": 0.86, "learning_rate": 1.0812773539054843e-05, "loss": 0.1396, "step": 13336 }, { "epoch": 0.86, "learning_rate": 1.080338292739378e-05, "loss": 0.1211, "step": 13337 }, { "epoch": 0.86, "learning_rate": 1.0793996162404162e-05, "loss": 0.1304, "step": 13338 }, { "epoch": 0.86, "learning_rate": 1.0784613244490816e-05, "loss": 0.1436, "step": 13339 }, { "epoch": 0.86, "learning_rate": 1.0775234174058368e-05, "loss": 0.1298, "step": 13340 }, { "epoch": 0.86, "learning_rate": 1.0765858951511332e-05, "loss": 0.1293, "step": 13341 }, { "epoch": 0.86, "learning_rate": 1.0756487577254026e-05, "loss": 0.1449, "step": 13342 }, { "epoch": 0.86, "learning_rate": 1.0747120051690562e-05, "loss": 0.137, "step": 13343 }, { "epoch": 0.86, "learning_rate": 1.073775637522495e-05, "loss": 0.1414, "step": 13344 }, { "epoch": 0.86, "learning_rate": 1.0728396548261011e-05, "loss": 0.1243, "step": 13345 }, { "epoch": 0.86, "learning_rate": 1.0719040571202343e-05, "loss": 0.1416, "step": 13346 }, { "epoch": 0.86, "learning_rate": 1.070968844445247e-05, "loss": 0.1307, "step": 13347 }, { "epoch": 0.86, "learning_rate": 1.0700340168414701e-05, "loss": 0.1324, "step": 13348 }, { "epoch": 0.86, "learning_rate": 1.0690995743492215e-05, "loss": 0.1225, "step": 13349 }, { "epoch": 0.86, "learning_rate": 1.0681655170087956e-05, "loss": 0.1243, "step": 13350 }, { "epoch": 0.86, "learning_rate": 1.0672318448604778e-05, "loss": 0.1264, "step": 13351 }, { "epoch": 0.86, "learning_rate": 1.0662985579445317e-05, "loss": 0.128, "step": 13352 }, { "epoch": 0.86, "learning_rate": 1.0653656563012038e-05, "loss": 0.1424, "step": 13353 }, { "epoch": 0.86, "learning_rate": 1.0644331399707286e-05, "loss": 0.124, "step": 13354 }, { "epoch": 0.86, "learning_rate": 1.0635010089933206e-05, "loss": 0.1276, "step": 13355 }, { "epoch": 0.86, "learning_rate": 1.0625692634091823e-05, "loss": 0.1333, "step": 13356 }, { "epoch": 0.86, "learning_rate": 1.0616379032584922e-05, "loss": 0.1248, "step": 13357 }, { "epoch": 0.86, "learning_rate": 1.0607069285814154e-05, "loss": 0.1281, "step": 13358 }, { "epoch": 0.86, "learning_rate": 1.0597763394181037e-05, "loss": 0.1244, "step": 13359 }, { "epoch": 0.86, "learning_rate": 1.0588461358086843e-05, "loss": 0.1278, "step": 13360 }, { "epoch": 0.86, "learning_rate": 1.0579163177932805e-05, "loss": 0.1341, "step": 13361 }, { "epoch": 0.86, "learning_rate": 1.0569868854119858e-05, "loss": 0.1395, "step": 13362 }, { "epoch": 0.86, "learning_rate": 1.056057838704887e-05, "loss": 0.1461, "step": 13363 }, { "epoch": 0.86, "learning_rate": 1.0551291777120464e-05, "loss": 0.1308, "step": 13364 }, { "epoch": 0.86, "learning_rate": 1.0542009024735122e-05, "loss": 0.1502, "step": 13365 }, { "epoch": 0.86, "learning_rate": 1.0532730130293188e-05, "loss": 0.1567, "step": 13366 }, { "epoch": 0.86, "learning_rate": 1.0523455094194823e-05, "loss": 0.1349, "step": 13367 }, { "epoch": 0.86, "learning_rate": 1.0514183916840036e-05, "loss": 0.1406, "step": 13368 }, { "epoch": 0.86, "learning_rate": 1.0504916598628623e-05, "loss": 0.1442, "step": 13369 }, { "epoch": 0.86, "learning_rate": 1.0495653139960282e-05, "loss": 0.1316, "step": 13370 }, { "epoch": 0.86, "learning_rate": 1.0486393541234474e-05, "loss": 0.1348, "step": 13371 }, { "epoch": 0.86, "learning_rate": 1.0477137802850512e-05, "loss": 0.1299, "step": 13372 }, { "epoch": 0.86, "learning_rate": 1.0467885925207587e-05, "loss": 0.1442, "step": 13373 }, { "epoch": 0.86, "learning_rate": 1.0458637908704682e-05, "loss": 0.1347, "step": 13374 }, { "epoch": 0.86, "learning_rate": 1.0449393753740655e-05, "loss": 0.148, "step": 13375 }, { "epoch": 0.86, "learning_rate": 1.0440153460714109e-05, "loss": 0.148, "step": 13376 }, { "epoch": 0.86, "learning_rate": 1.0430917030023601e-05, "loss": 0.1425, "step": 13377 }, { "epoch": 0.86, "learning_rate": 1.0421684462067427e-05, "loss": 0.144, "step": 13378 }, { "epoch": 0.86, "learning_rate": 1.0412455757243722e-05, "loss": 0.1193, "step": 13379 }, { "epoch": 0.86, "learning_rate": 1.0403230915950512e-05, "loss": 0.1377, "step": 13380 }, { "epoch": 0.86, "learning_rate": 1.0394009938585625e-05, "loss": 0.1519, "step": 13381 }, { "epoch": 0.86, "learning_rate": 1.0384792825546741e-05, "loss": 0.1247, "step": 13382 }, { "epoch": 0.86, "learning_rate": 1.0375579577231309e-05, "loss": 0.1265, "step": 13383 }, { "epoch": 0.86, "learning_rate": 1.03663701940367e-05, "loss": 0.1289, "step": 13384 }, { "epoch": 0.86, "learning_rate": 1.0357164676360053e-05, "loss": 0.1268, "step": 13385 }, { "epoch": 0.86, "learning_rate": 1.0347963024598362e-05, "loss": 0.1305, "step": 13386 }, { "epoch": 0.86, "learning_rate": 1.033876523914845e-05, "loss": 0.1262, "step": 13387 }, { "epoch": 0.86, "learning_rate": 1.0329571320406983e-05, "loss": 0.1552, "step": 13388 }, { "epoch": 0.86, "learning_rate": 1.0320381268770496e-05, "loss": 0.121, "step": 13389 }, { "epoch": 0.86, "learning_rate": 1.031119508463525e-05, "loss": 0.1404, "step": 13390 }, { "epoch": 0.86, "learning_rate": 1.0302012768397461e-05, "loss": 0.1318, "step": 13391 }, { "epoch": 0.86, "learning_rate": 1.0292834320453105e-05, "loss": 0.1222, "step": 13392 }, { "epoch": 0.86, "learning_rate": 1.0283659741197992e-05, "loss": 0.1409, "step": 13393 }, { "epoch": 0.86, "learning_rate": 1.027448903102779e-05, "loss": 0.138, "step": 13394 }, { "epoch": 0.86, "learning_rate": 1.026532219033799e-05, "loss": 0.1309, "step": 13395 }, { "epoch": 0.86, "learning_rate": 1.0256159219523953e-05, "loss": 0.1395, "step": 13396 }, { "epoch": 0.86, "learning_rate": 1.0247000118980799e-05, "loss": 0.1426, "step": 13397 }, { "epoch": 0.86, "learning_rate": 1.0237844889103553e-05, "loss": 0.1435, "step": 13398 }, { "epoch": 0.86, "learning_rate": 1.0228693530287015e-05, "loss": 0.15, "step": 13399 }, { "epoch": 0.86, "learning_rate": 1.0219546042925843e-05, "loss": 0.1433, "step": 13400 }, { "epoch": 0.86, "learning_rate": 1.0210402427414545e-05, "loss": 0.1284, "step": 13401 }, { "epoch": 0.86, "learning_rate": 1.0201262684147428e-05, "loss": 0.1298, "step": 13402 }, { "epoch": 0.86, "learning_rate": 1.0192126813518687e-05, "loss": 0.1397, "step": 13403 }, { "epoch": 0.86, "learning_rate": 1.0182994815922275e-05, "loss": 0.129, "step": 13404 }, { "epoch": 0.86, "learning_rate": 1.0173866691752043e-05, "loss": 0.1326, "step": 13405 }, { "epoch": 0.86, "learning_rate": 1.0164742441401643e-05, "loss": 0.126, "step": 13406 }, { "epoch": 0.86, "learning_rate": 1.0155622065264526e-05, "loss": 0.1243, "step": 13407 }, { "epoch": 0.86, "learning_rate": 1.0146505563734066e-05, "loss": 0.142, "step": 13408 }, { "epoch": 0.86, "learning_rate": 1.0137392937203394e-05, "loss": 0.1277, "step": 13409 }, { "epoch": 0.86, "learning_rate": 1.0128284186065529e-05, "loss": 0.1362, "step": 13410 }, { "epoch": 0.86, "learning_rate": 1.0119179310713244e-05, "loss": 0.1414, "step": 13411 }, { "epoch": 0.86, "learning_rate": 1.011007831153924e-05, "loss": 0.1304, "step": 13412 }, { "epoch": 0.86, "learning_rate": 1.0100981188935954e-05, "loss": 0.1423, "step": 13413 }, { "epoch": 0.86, "learning_rate": 1.0091887943295741e-05, "loss": 0.1287, "step": 13414 }, { "epoch": 0.86, "learning_rate": 1.0082798575010777e-05, "loss": 0.1426, "step": 13415 }, { "epoch": 0.86, "learning_rate": 1.007371308447299e-05, "loss": 0.1413, "step": 13416 }, { "epoch": 0.86, "learning_rate": 1.0064631472074248e-05, "loss": 0.1249, "step": 13417 }, { "epoch": 0.86, "learning_rate": 1.005555373820617e-05, "loss": 0.1582, "step": 13418 }, { "epoch": 0.86, "learning_rate": 1.0046479883260262e-05, "loss": 0.1184, "step": 13419 }, { "epoch": 0.86, "learning_rate": 1.0037409907627815e-05, "loss": 0.121, "step": 13420 }, { "epoch": 0.86, "learning_rate": 1.0028343811700002e-05, "loss": 0.1455, "step": 13421 }, { "epoch": 0.86, "learning_rate": 1.0019281595867813e-05, "loss": 0.1226, "step": 13422 }, { "epoch": 0.86, "learning_rate": 1.0010223260522034e-05, "loss": 0.1394, "step": 13423 }, { "epoch": 0.86, "learning_rate": 1.000116880605334e-05, "loss": 0.1262, "step": 13424 }, { "epoch": 0.86, "learning_rate": 9.992118232852177e-06, "loss": 0.1478, "step": 13425 }, { "epoch": 0.86, "learning_rate": 9.983071541308897e-06, "loss": 0.1333, "step": 13426 }, { "epoch": 0.86, "learning_rate": 9.974028731813611e-06, "loss": 0.145, "step": 13427 }, { "epoch": 0.86, "learning_rate": 9.96498980475632e-06, "loss": 0.1053, "step": 13428 }, { "epoch": 0.86, "learning_rate": 9.955954760526831e-06, "loss": 0.1276, "step": 13429 }, { "epoch": 0.86, "learning_rate": 9.94692359951478e-06, "loss": 0.133, "step": 13430 }, { "epoch": 0.86, "learning_rate": 9.937896322109653e-06, "loss": 0.1389, "step": 13431 }, { "epoch": 0.86, "learning_rate": 9.928872928700761e-06, "loss": 0.1431, "step": 13432 }, { "epoch": 0.86, "learning_rate": 9.919853419677217e-06, "loss": 0.1442, "step": 13433 }, { "epoch": 0.86, "learning_rate": 9.910837795428007e-06, "loss": 0.1406, "step": 13434 }, { "epoch": 0.86, "learning_rate": 9.901826056341935e-06, "loss": 0.1244, "step": 13435 }, { "epoch": 0.86, "learning_rate": 9.892818202807674e-06, "loss": 0.1392, "step": 13436 }, { "epoch": 0.86, "learning_rate": 9.883814235213652e-06, "loss": 0.1167, "step": 13437 }, { "epoch": 0.86, "learning_rate": 9.874814153948198e-06, "loss": 0.1502, "step": 13438 }, { "epoch": 0.86, "learning_rate": 9.865817959399436e-06, "loss": 0.1213, "step": 13439 }, { "epoch": 0.86, "learning_rate": 9.856825651955315e-06, "loss": 0.1458, "step": 13440 }, { "epoch": 0.86, "learning_rate": 9.847837232003654e-06, "loss": 0.1467, "step": 13441 }, { "epoch": 0.86, "learning_rate": 9.838852699932078e-06, "loss": 0.1401, "step": 13442 }, { "epoch": 0.86, "learning_rate": 9.829872056128087e-06, "loss": 0.1342, "step": 13443 }, { "epoch": 0.86, "learning_rate": 9.820895300978927e-06, "loss": 0.1449, "step": 13444 }, { "epoch": 0.86, "learning_rate": 9.811922434871767e-06, "loss": 0.1409, "step": 13445 }, { "epoch": 0.86, "learning_rate": 9.802953458193554e-06, "loss": 0.1277, "step": 13446 }, { "epoch": 0.86, "learning_rate": 9.793988371331042e-06, "loss": 0.1367, "step": 13447 }, { "epoch": 0.86, "learning_rate": 9.785027174670936e-06, "loss": 0.1256, "step": 13448 }, { "epoch": 0.86, "learning_rate": 9.776069868599624e-06, "loss": 0.1341, "step": 13449 }, { "epoch": 0.86, "learning_rate": 9.767116453503455e-06, "loss": 0.1259, "step": 13450 }, { "epoch": 0.86, "learning_rate": 9.758166929768498e-06, "loss": 0.1274, "step": 13451 }, { "epoch": 0.86, "learning_rate": 9.749221297780754e-06, "loss": 0.132, "step": 13452 }, { "epoch": 0.86, "learning_rate": 9.740279557925969e-06, "loss": 0.1427, "step": 13453 }, { "epoch": 0.86, "learning_rate": 9.73134171058977e-06, "loss": 0.1611, "step": 13454 }, { "epoch": 0.86, "learning_rate": 9.722407756157658e-06, "loss": 0.1281, "step": 13455 }, { "epoch": 0.86, "learning_rate": 9.71347769501485e-06, "loss": 0.1431, "step": 13456 }, { "epoch": 0.86, "learning_rate": 9.704551527546502e-06, "loss": 0.1329, "step": 13457 }, { "epoch": 0.86, "learning_rate": 9.69562925413754e-06, "loss": 0.151, "step": 13458 }, { "epoch": 0.86, "learning_rate": 9.686710875172778e-06, "loss": 0.1473, "step": 13459 }, { "epoch": 0.86, "learning_rate": 9.677796391036775e-06, "loss": 0.1428, "step": 13460 }, { "epoch": 0.86, "learning_rate": 9.668885802114003e-06, "loss": 0.1311, "step": 13461 }, { "epoch": 0.86, "learning_rate": 9.659979108788763e-06, "loss": 0.1174, "step": 13462 }, { "epoch": 0.86, "learning_rate": 9.651076311445118e-06, "loss": 0.138, "step": 13463 }, { "epoch": 0.86, "learning_rate": 9.642177410467046e-06, "loss": 0.1262, "step": 13464 }, { "epoch": 0.86, "learning_rate": 9.633282406238276e-06, "loss": 0.1306, "step": 13465 }, { "epoch": 0.86, "learning_rate": 9.624391299142466e-06, "loss": 0.1308, "step": 13466 }, { "epoch": 0.86, "learning_rate": 9.615504089563e-06, "loss": 0.147, "step": 13467 }, { "epoch": 0.86, "learning_rate": 9.606620777883168e-06, "loss": 0.1272, "step": 13468 }, { "epoch": 0.86, "learning_rate": 9.597741364486078e-06, "loss": 0.1409, "step": 13469 }, { "epoch": 0.86, "learning_rate": 9.588865849754647e-06, "loss": 0.1254, "step": 13470 }, { "epoch": 0.86, "learning_rate": 9.579994234071655e-06, "loss": 0.1424, "step": 13471 }, { "epoch": 0.86, "learning_rate": 9.571126517819672e-06, "loss": 0.1594, "step": 13472 }, { "epoch": 0.86, "learning_rate": 9.562262701381164e-06, "loss": 0.1319, "step": 13473 }, { "epoch": 0.86, "learning_rate": 9.553402785138332e-06, "loss": 0.1214, "step": 13474 }, { "epoch": 0.86, "learning_rate": 9.544546769473317e-06, "loss": 0.135, "step": 13475 }, { "epoch": 0.86, "learning_rate": 9.535694654768023e-06, "loss": 0.1326, "step": 13476 }, { "epoch": 0.86, "learning_rate": 9.526846441404203e-06, "loss": 0.138, "step": 13477 }, { "epoch": 0.86, "learning_rate": 9.518002129763458e-06, "loss": 0.1401, "step": 13478 }, { "epoch": 0.86, "learning_rate": 9.509161720227167e-06, "loss": 0.1298, "step": 13479 }, { "epoch": 0.86, "learning_rate": 9.50032521317663e-06, "loss": 0.1263, "step": 13480 }, { "epoch": 0.86, "learning_rate": 9.491492608992891e-06, "loss": 0.1517, "step": 13481 }, { "epoch": 0.86, "learning_rate": 9.482663908056865e-06, "loss": 0.1542, "step": 13482 }, { "epoch": 0.86, "learning_rate": 9.473839110749328e-06, "loss": 0.1317, "step": 13483 }, { "epoch": 0.86, "learning_rate": 9.465018217450816e-06, "loss": 0.1489, "step": 13484 }, { "epoch": 0.86, "learning_rate": 9.456201228541772e-06, "loss": 0.1453, "step": 13485 }, { "epoch": 0.86, "learning_rate": 9.447388144402403e-06, "loss": 0.136, "step": 13486 }, { "epoch": 0.86, "learning_rate": 9.438578965412803e-06, "loss": 0.1338, "step": 13487 }, { "epoch": 0.86, "learning_rate": 9.429773691952858e-06, "loss": 0.1364, "step": 13488 }, { "epoch": 0.86, "learning_rate": 9.420972324402311e-06, "loss": 0.1317, "step": 13489 }, { "epoch": 0.86, "learning_rate": 9.412174863140744e-06, "loss": 0.1312, "step": 13490 }, { "epoch": 0.87, "learning_rate": 9.40338130854751e-06, "loss": 0.1278, "step": 13491 }, { "epoch": 0.87, "learning_rate": 9.394591661001895e-06, "loss": 0.1111, "step": 13492 }, { "epoch": 0.87, "learning_rate": 9.385805920882907e-06, "loss": 0.1452, "step": 13493 }, { "epoch": 0.87, "learning_rate": 9.377024088569452e-06, "loss": 0.1429, "step": 13494 }, { "epoch": 0.87, "learning_rate": 9.368246164440297e-06, "loss": 0.1345, "step": 13495 }, { "epoch": 0.87, "learning_rate": 9.359472148873927e-06, "loss": 0.1246, "step": 13496 }, { "epoch": 0.87, "learning_rate": 9.350702042248793e-06, "loss": 0.149, "step": 13497 }, { "epoch": 0.87, "learning_rate": 9.341935844943051e-06, "loss": 0.1594, "step": 13498 }, { "epoch": 0.87, "learning_rate": 9.333173557334806e-06, "loss": 0.1262, "step": 13499 }, { "epoch": 0.87, "learning_rate": 9.324415179801882e-06, "loss": 0.1317, "step": 13500 }, { "epoch": 0.87, "learning_rate": 9.31566071272203e-06, "loss": 0.1362, "step": 13501 }, { "epoch": 0.87, "learning_rate": 9.306910156472803e-06, "loss": 0.144, "step": 13502 }, { "epoch": 0.87, "learning_rate": 9.298163511431545e-06, "loss": 0.1403, "step": 13503 }, { "epoch": 0.87, "learning_rate": 9.289420777975477e-06, "loss": 0.1338, "step": 13504 }, { "epoch": 0.87, "learning_rate": 9.280681956481619e-06, "loss": 0.1266, "step": 13505 }, { "epoch": 0.87, "learning_rate": 9.271947047326879e-06, "loss": 0.1352, "step": 13506 }, { "epoch": 0.87, "learning_rate": 9.263216050887902e-06, "loss": 0.1208, "step": 13507 }, { "epoch": 0.87, "learning_rate": 9.254488967541253e-06, "loss": 0.1424, "step": 13508 }, { "epoch": 0.87, "learning_rate": 9.245765797663298e-06, "loss": 0.115, "step": 13509 }, { "epoch": 0.87, "learning_rate": 9.237046541630201e-06, "loss": 0.1379, "step": 13510 }, { "epoch": 0.87, "learning_rate": 9.22833119981803e-06, "loss": 0.1335, "step": 13511 }, { "epoch": 0.87, "learning_rate": 9.219619772602594e-06, "loss": 0.1327, "step": 13512 }, { "epoch": 0.87, "learning_rate": 9.210912260359628e-06, "loss": 0.141, "step": 13513 }, { "epoch": 0.87, "learning_rate": 9.202208663464595e-06, "loss": 0.1319, "step": 13514 }, { "epoch": 0.87, "learning_rate": 9.193508982292887e-06, "loss": 0.1545, "step": 13515 }, { "epoch": 0.87, "learning_rate": 9.18481321721968e-06, "loss": 0.1307, "step": 13516 }, { "epoch": 0.87, "learning_rate": 9.176121368619961e-06, "loss": 0.1411, "step": 13517 }, { "epoch": 0.87, "learning_rate": 9.1674334368686e-06, "loss": 0.1391, "step": 13518 }, { "epoch": 0.87, "learning_rate": 9.15874942234024e-06, "loss": 0.1482, "step": 13519 }, { "epoch": 0.87, "learning_rate": 9.150069325409427e-06, "loss": 0.1209, "step": 13520 }, { "epoch": 0.87, "learning_rate": 9.14139314645045e-06, "loss": 0.1409, "step": 13521 }, { "epoch": 0.87, "learning_rate": 9.13272088583751e-06, "loss": 0.1303, "step": 13522 }, { "epoch": 0.87, "learning_rate": 9.124052543944606e-06, "loss": 0.1214, "step": 13523 }, { "epoch": 0.87, "learning_rate": 9.11538812114553e-06, "loss": 0.1258, "step": 13524 }, { "epoch": 0.87, "learning_rate": 9.106727617813993e-06, "loss": 0.1507, "step": 13525 }, { "epoch": 0.87, "learning_rate": 9.098071034323441e-06, "loss": 0.1433, "step": 13526 }, { "epoch": 0.87, "learning_rate": 9.089418371047232e-06, "loss": 0.1351, "step": 13527 }, { "epoch": 0.87, "learning_rate": 9.080769628358476e-06, "loss": 0.1229, "step": 13528 }, { "epoch": 0.87, "learning_rate": 9.072124806630189e-06, "loss": 0.157, "step": 13529 }, { "epoch": 0.87, "learning_rate": 9.063483906235193e-06, "loss": 0.1359, "step": 13530 }, { "epoch": 0.87, "learning_rate": 9.054846927546101e-06, "loss": 0.1346, "step": 13531 }, { "epoch": 0.87, "learning_rate": 9.046213870935426e-06, "loss": 0.115, "step": 13532 }, { "epoch": 0.87, "learning_rate": 9.037584736775428e-06, "loss": 0.1219, "step": 13533 }, { "epoch": 0.87, "learning_rate": 9.028959525438285e-06, "loss": 0.1194, "step": 13534 }, { "epoch": 0.87, "learning_rate": 9.020338237295966e-06, "loss": 0.1447, "step": 13535 }, { "epoch": 0.87, "learning_rate": 9.011720872720241e-06, "loss": 0.1265, "step": 13536 }, { "epoch": 0.87, "learning_rate": 9.00310743208277e-06, "loss": 0.1234, "step": 13537 }, { "epoch": 0.87, "learning_rate": 8.994497915754995e-06, "loss": 0.1436, "step": 13538 }, { "epoch": 0.87, "learning_rate": 8.985892324108226e-06, "loss": 0.1248, "step": 13539 }, { "epoch": 0.87, "learning_rate": 8.97729065751355e-06, "loss": 0.1257, "step": 13540 }, { "epoch": 0.87, "learning_rate": 8.968692916341948e-06, "loss": 0.1509, "step": 13541 }, { "epoch": 0.87, "learning_rate": 8.960099100964215e-06, "loss": 0.1404, "step": 13542 }, { "epoch": 0.87, "learning_rate": 8.95150921175094e-06, "loss": 0.1251, "step": 13543 }, { "epoch": 0.87, "learning_rate": 8.942923249072588e-06, "loss": 0.1252, "step": 13544 }, { "epoch": 0.87, "learning_rate": 8.934341213299414e-06, "loss": 0.1186, "step": 13545 }, { "epoch": 0.87, "learning_rate": 8.925763104801554e-06, "loss": 0.1335, "step": 13546 }, { "epoch": 0.87, "learning_rate": 8.917188923948916e-06, "loss": 0.1286, "step": 13547 }, { "epoch": 0.87, "learning_rate": 8.90861867111128e-06, "loss": 0.1355, "step": 13548 }, { "epoch": 0.87, "learning_rate": 8.900052346658261e-06, "loss": 0.1338, "step": 13549 }, { "epoch": 0.87, "learning_rate": 8.891489950959253e-06, "loss": 0.1361, "step": 13550 }, { "epoch": 0.87, "learning_rate": 8.88293148438355e-06, "loss": 0.1278, "step": 13551 }, { "epoch": 0.87, "learning_rate": 8.874376947300222e-06, "loss": 0.1463, "step": 13552 }, { "epoch": 0.87, "learning_rate": 8.865826340078199e-06, "loss": 0.1266, "step": 13553 }, { "epoch": 0.87, "learning_rate": 8.857279663086226e-06, "loss": 0.1427, "step": 13554 }, { "epoch": 0.87, "learning_rate": 8.84873691669288e-06, "loss": 0.1291, "step": 13555 }, { "epoch": 0.87, "learning_rate": 8.840198101266595e-06, "loss": 0.1461, "step": 13556 }, { "epoch": 0.87, "learning_rate": 8.83166321717559e-06, "loss": 0.135, "step": 13557 }, { "epoch": 0.87, "learning_rate": 8.823132264787959e-06, "loss": 0.1237, "step": 13558 }, { "epoch": 0.87, "learning_rate": 8.814605244471575e-06, "loss": 0.1425, "step": 13559 }, { "epoch": 0.87, "learning_rate": 8.80608215659422e-06, "loss": 0.1285, "step": 13560 }, { "epoch": 0.87, "learning_rate": 8.797563001523412e-06, "loss": 0.1434, "step": 13561 }, { "epoch": 0.87, "learning_rate": 8.789047779626559e-06, "loss": 0.1395, "step": 13562 }, { "epoch": 0.87, "learning_rate": 8.780536491270919e-06, "loss": 0.1475, "step": 13563 }, { "epoch": 0.87, "learning_rate": 8.77202913682349e-06, "loss": 0.149, "step": 13564 }, { "epoch": 0.87, "learning_rate": 8.763525716651222e-06, "loss": 0.1388, "step": 13565 }, { "epoch": 0.87, "learning_rate": 8.755026231120766e-06, "loss": 0.1238, "step": 13566 }, { "epoch": 0.87, "learning_rate": 8.74653068059873e-06, "loss": 0.1292, "step": 13567 }, { "epoch": 0.87, "learning_rate": 8.738039065451442e-06, "loss": 0.1411, "step": 13568 }, { "epoch": 0.87, "learning_rate": 8.729551386045131e-06, "loss": 0.1125, "step": 13569 }, { "epoch": 0.87, "learning_rate": 8.721067642745851e-06, "loss": 0.1183, "step": 13570 }, { "epoch": 0.87, "learning_rate": 8.712587835919427e-06, "loss": 0.1443, "step": 13571 }, { "epoch": 0.87, "learning_rate": 8.704111965931604e-06, "loss": 0.1328, "step": 13572 }, { "epoch": 0.87, "learning_rate": 8.695640033147867e-06, "loss": 0.1363, "step": 13573 }, { "epoch": 0.87, "learning_rate": 8.687172037933622e-06, "loss": 0.121, "step": 13574 }, { "epoch": 0.87, "learning_rate": 8.678707980654e-06, "loss": 0.1271, "step": 13575 }, { "epoch": 0.87, "learning_rate": 8.670247861674075e-06, "loss": 0.1438, "step": 13576 }, { "epoch": 0.87, "learning_rate": 8.661791681358677e-06, "loss": 0.1317, "step": 13577 }, { "epoch": 0.87, "learning_rate": 8.653339440072462e-06, "loss": 0.1431, "step": 13578 }, { "epoch": 0.87, "learning_rate": 8.644891138179977e-06, "loss": 0.1251, "step": 13579 }, { "epoch": 0.87, "learning_rate": 8.636446776045527e-06, "loss": 0.1259, "step": 13580 }, { "epoch": 0.87, "learning_rate": 8.628006354033291e-06, "loss": 0.1562, "step": 13581 }, { "epoch": 0.87, "learning_rate": 8.619569872507283e-06, "loss": 0.1353, "step": 13582 }, { "epoch": 0.87, "learning_rate": 8.611137331831331e-06, "loss": 0.1447, "step": 13583 }, { "epoch": 0.87, "learning_rate": 8.602708732369091e-06, "loss": 0.1319, "step": 13584 }, { "epoch": 0.87, "learning_rate": 8.594284074484038e-06, "loss": 0.1254, "step": 13585 }, { "epoch": 0.87, "learning_rate": 8.585863358539515e-06, "loss": 0.1353, "step": 13586 }, { "epoch": 0.87, "learning_rate": 8.577446584898652e-06, "loss": 0.1494, "step": 13587 }, { "epoch": 0.87, "learning_rate": 8.569033753924427e-06, "loss": 0.1444, "step": 13588 }, { "epoch": 0.87, "learning_rate": 8.560624865979671e-06, "loss": 0.1443, "step": 13589 }, { "epoch": 0.87, "learning_rate": 8.552219921427007e-06, "loss": 0.1323, "step": 13590 }, { "epoch": 0.87, "learning_rate": 8.543818920628921e-06, "loss": 0.1351, "step": 13591 }, { "epoch": 0.87, "learning_rate": 8.53542186394769e-06, "loss": 0.1344, "step": 13592 }, { "epoch": 0.87, "learning_rate": 8.527028751745469e-06, "loss": 0.1428, "step": 13593 }, { "epoch": 0.87, "learning_rate": 8.51863958438418e-06, "loss": 0.1152, "step": 13594 }, { "epoch": 0.87, "learning_rate": 8.510254362225633e-06, "loss": 0.1432, "step": 13595 }, { "epoch": 0.87, "learning_rate": 8.501873085631474e-06, "loss": 0.1534, "step": 13596 }, { "epoch": 0.87, "learning_rate": 8.49349575496311e-06, "loss": 0.1548, "step": 13597 }, { "epoch": 0.87, "learning_rate": 8.485122370581856e-06, "loss": 0.14, "step": 13598 }, { "epoch": 0.87, "learning_rate": 8.476752932848774e-06, "loss": 0.124, "step": 13599 }, { "epoch": 0.87, "learning_rate": 8.468387442124848e-06, "loss": 0.1344, "step": 13600 }, { "epoch": 0.87, "learning_rate": 8.460025898770818e-06, "loss": 0.1199, "step": 13601 }, { "epoch": 0.87, "learning_rate": 8.451668303147286e-06, "loss": 0.1336, "step": 13602 }, { "epoch": 0.87, "learning_rate": 8.443314655614708e-06, "loss": 0.1319, "step": 13603 }, { "epoch": 0.87, "learning_rate": 8.434964956533297e-06, "loss": 0.1318, "step": 13604 }, { "epoch": 0.87, "learning_rate": 8.426619206263175e-06, "loss": 0.1327, "step": 13605 }, { "epoch": 0.87, "learning_rate": 8.418277405164232e-06, "loss": 0.1311, "step": 13606 }, { "epoch": 0.87, "learning_rate": 8.409939553596247e-06, "loss": 0.126, "step": 13607 }, { "epoch": 0.87, "learning_rate": 8.401605651918753e-06, "loss": 0.1178, "step": 13608 }, { "epoch": 0.87, "learning_rate": 8.393275700491187e-06, "loss": 0.1166, "step": 13609 }, { "epoch": 0.87, "learning_rate": 8.384949699672795e-06, "loss": 0.1369, "step": 13610 }, { "epoch": 0.87, "learning_rate": 8.376627649822599e-06, "loss": 0.1354, "step": 13611 }, { "epoch": 0.87, "learning_rate": 8.368309551299536e-06, "loss": 0.1459, "step": 13612 }, { "epoch": 0.87, "learning_rate": 8.359995404462306e-06, "loss": 0.1222, "step": 13613 }, { "epoch": 0.87, "learning_rate": 8.35168520966948e-06, "loss": 0.146, "step": 13614 }, { "epoch": 0.87, "learning_rate": 8.343378967279413e-06, "loss": 0.1379, "step": 13615 }, { "epoch": 0.87, "learning_rate": 8.335076677650344e-06, "loss": 0.1411, "step": 13616 }, { "epoch": 0.87, "learning_rate": 8.326778341140318e-06, "loss": 0.146, "step": 13617 }, { "epoch": 0.87, "learning_rate": 8.318483958107182e-06, "loss": 0.1362, "step": 13618 }, { "epoch": 0.87, "learning_rate": 8.310193528908671e-06, "loss": 0.1532, "step": 13619 }, { "epoch": 0.87, "learning_rate": 8.30190705390228e-06, "loss": 0.1363, "step": 13620 }, { "epoch": 0.87, "learning_rate": 8.293624533445377e-06, "loss": 0.1238, "step": 13621 }, { "epoch": 0.87, "learning_rate": 8.285345967895175e-06, "loss": 0.1142, "step": 13622 }, { "epoch": 0.87, "learning_rate": 8.27707135760869e-06, "loss": 0.1348, "step": 13623 }, { "epoch": 0.87, "learning_rate": 8.26880070294277e-06, "loss": 0.1294, "step": 13624 }, { "epoch": 0.87, "learning_rate": 8.260534004254062e-06, "loss": 0.1402, "step": 13625 }, { "epoch": 0.87, "learning_rate": 8.252271261899114e-06, "loss": 0.139, "step": 13626 }, { "epoch": 0.87, "learning_rate": 8.244012476234219e-06, "loss": 0.1256, "step": 13627 }, { "epoch": 0.87, "learning_rate": 8.235757647615583e-06, "loss": 0.1419, "step": 13628 }, { "epoch": 0.87, "learning_rate": 8.227506776399185e-06, "loss": 0.147, "step": 13629 }, { "epoch": 0.87, "learning_rate": 8.219259862940865e-06, "loss": 0.1325, "step": 13630 }, { "epoch": 0.87, "learning_rate": 8.211016907596269e-06, "loss": 0.1666, "step": 13631 }, { "epoch": 0.87, "learning_rate": 8.20277791072086e-06, "loss": 0.1238, "step": 13632 }, { "epoch": 0.87, "learning_rate": 8.194542872669986e-06, "loss": 0.1422, "step": 13633 }, { "epoch": 0.87, "learning_rate": 8.186311793798751e-06, "loss": 0.1374, "step": 13634 }, { "epoch": 0.87, "learning_rate": 8.17808467446215e-06, "loss": 0.1416, "step": 13635 }, { "epoch": 0.87, "learning_rate": 8.169861515014976e-06, "loss": 0.1302, "step": 13636 }, { "epoch": 0.87, "learning_rate": 8.161642315811879e-06, "loss": 0.1209, "step": 13637 }, { "epoch": 0.87, "learning_rate": 8.15342707720731e-06, "loss": 0.1338, "step": 13638 }, { "epoch": 0.87, "learning_rate": 8.145215799555528e-06, "loss": 0.1438, "step": 13639 }, { "epoch": 0.87, "learning_rate": 8.137008483210696e-06, "loss": 0.1216, "step": 13640 }, { "epoch": 0.87, "learning_rate": 8.128805128526717e-06, "loss": 0.1344, "step": 13641 }, { "epoch": 0.87, "learning_rate": 8.12060573585739e-06, "loss": 0.1127, "step": 13642 }, { "epoch": 0.87, "learning_rate": 8.112410305556307e-06, "loss": 0.1479, "step": 13643 }, { "epoch": 0.87, "learning_rate": 8.10421883797694e-06, "loss": 0.1193, "step": 13644 }, { "epoch": 0.87, "learning_rate": 8.09603133347252e-06, "loss": 0.1291, "step": 13645 }, { "epoch": 0.87, "learning_rate": 8.08784779239613e-06, "loss": 0.1388, "step": 13646 }, { "epoch": 0.88, "learning_rate": 8.079668215100723e-06, "loss": 0.1185, "step": 13647 }, { "epoch": 0.88, "learning_rate": 8.071492601939013e-06, "loss": 0.1309, "step": 13648 }, { "epoch": 0.88, "learning_rate": 8.063320953263587e-06, "loss": 0.1364, "step": 13649 }, { "epoch": 0.88, "learning_rate": 8.055153269426874e-06, "loss": 0.1338, "step": 13650 }, { "epoch": 0.88, "learning_rate": 8.046989550781125e-06, "loss": 0.1358, "step": 13651 }, { "epoch": 0.88, "learning_rate": 8.038829797678381e-06, "loss": 0.1492, "step": 13652 }, { "epoch": 0.88, "learning_rate": 8.030674010470518e-06, "loss": 0.1372, "step": 13653 }, { "epoch": 0.88, "learning_rate": 8.022522189509307e-06, "loss": 0.156, "step": 13654 }, { "epoch": 0.88, "learning_rate": 8.014374335146258e-06, "loss": 0.1239, "step": 13655 }, { "epoch": 0.88, "learning_rate": 8.006230447732766e-06, "loss": 0.1474, "step": 13656 }, { "epoch": 0.88, "learning_rate": 7.99809052762005e-06, "loss": 0.1385, "step": 13657 }, { "epoch": 0.88, "learning_rate": 7.989954575159175e-06, "loss": 0.1425, "step": 13658 }, { "epoch": 0.88, "learning_rate": 7.981822590700983e-06, "loss": 0.1374, "step": 13659 }, { "epoch": 0.88, "learning_rate": 7.973694574596146e-06, "loss": 0.129, "step": 13660 }, { "epoch": 0.88, "learning_rate": 7.965570527195243e-06, "loss": 0.1297, "step": 13661 }, { "epoch": 0.88, "learning_rate": 7.957450448848569e-06, "loss": 0.1339, "step": 13662 }, { "epoch": 0.88, "learning_rate": 7.949334339906378e-06, "loss": 0.1511, "step": 13663 }, { "epoch": 0.88, "learning_rate": 7.941222200718634e-06, "loss": 0.139, "step": 13664 }, { "epoch": 0.88, "learning_rate": 7.933114031635202e-06, "loss": 0.1357, "step": 13665 }, { "epoch": 0.88, "learning_rate": 7.925009833005759e-06, "loss": 0.133, "step": 13666 }, { "epoch": 0.88, "learning_rate": 7.916909605179767e-06, "loss": 0.1362, "step": 13667 }, { "epoch": 0.88, "learning_rate": 7.908813348506583e-06, "loss": 0.1553, "step": 13668 }, { "epoch": 0.88, "learning_rate": 7.900721063335358e-06, "loss": 0.1301, "step": 13669 }, { "epoch": 0.88, "learning_rate": 7.892632750015095e-06, "loss": 0.1427, "step": 13670 }, { "epoch": 0.88, "learning_rate": 7.884548408894598e-06, "loss": 0.1256, "step": 13671 }, { "epoch": 0.88, "learning_rate": 7.876468040322482e-06, "loss": 0.1402, "step": 13672 }, { "epoch": 0.88, "learning_rate": 7.868391644647267e-06, "loss": 0.1412, "step": 13673 }, { "epoch": 0.88, "learning_rate": 7.860319222217205e-06, "loss": 0.1446, "step": 13674 }, { "epoch": 0.88, "learning_rate": 7.852250773380454e-06, "loss": 0.124, "step": 13675 }, { "epoch": 0.88, "learning_rate": 7.844186298484968e-06, "loss": 0.1406, "step": 13676 }, { "epoch": 0.88, "learning_rate": 7.836125797878546e-06, "loss": 0.1261, "step": 13677 }, { "epoch": 0.88, "learning_rate": 7.828069271908788e-06, "loss": 0.1275, "step": 13678 }, { "epoch": 0.88, "learning_rate": 7.82001672092313e-06, "loss": 0.1442, "step": 13679 }, { "epoch": 0.88, "learning_rate": 7.811968145268878e-06, "loss": 0.1521, "step": 13680 }, { "epoch": 0.88, "learning_rate": 7.803923545293079e-06, "loss": 0.1617, "step": 13681 }, { "epoch": 0.88, "learning_rate": 7.7958829213427e-06, "loss": 0.1423, "step": 13682 }, { "epoch": 0.88, "learning_rate": 7.787846273764487e-06, "loss": 0.1383, "step": 13683 }, { "epoch": 0.88, "learning_rate": 7.77981360290505e-06, "loss": 0.1293, "step": 13684 }, { "epoch": 0.88, "learning_rate": 7.771784909110791e-06, "loss": 0.1419, "step": 13685 }, { "epoch": 0.88, "learning_rate": 7.76376019272792e-06, "loss": 0.1364, "step": 13686 }, { "epoch": 0.88, "learning_rate": 7.755739454102562e-06, "loss": 0.1202, "step": 13687 }, { "epoch": 0.88, "learning_rate": 7.747722693580573e-06, "loss": 0.1348, "step": 13688 }, { "epoch": 0.88, "learning_rate": 7.739709911507697e-06, "loss": 0.1414, "step": 13689 }, { "epoch": 0.88, "learning_rate": 7.731701108229505e-06, "loss": 0.1348, "step": 13690 }, { "epoch": 0.88, "learning_rate": 7.723696284091385e-06, "loss": 0.1385, "step": 13691 }, { "epoch": 0.88, "learning_rate": 7.715695439438541e-06, "loss": 0.1275, "step": 13692 }, { "epoch": 0.88, "learning_rate": 7.707698574615984e-06, "loss": 0.1309, "step": 13693 }, { "epoch": 0.88, "learning_rate": 7.69970568996865e-06, "loss": 0.1405, "step": 13694 }, { "epoch": 0.88, "learning_rate": 7.691716785841175e-06, "loss": 0.1458, "step": 13695 }, { "epoch": 0.88, "learning_rate": 7.683731862578115e-06, "loss": 0.162, "step": 13696 }, { "epoch": 0.88, "learning_rate": 7.675750920523828e-06, "loss": 0.1431, "step": 13697 }, { "epoch": 0.88, "learning_rate": 7.667773960022507e-06, "loss": 0.1509, "step": 13698 }, { "epoch": 0.88, "learning_rate": 7.659800981418152e-06, "loss": 0.1269, "step": 13699 }, { "epoch": 0.88, "learning_rate": 7.651831985054591e-06, "loss": 0.1306, "step": 13700 }, { "epoch": 0.88, "learning_rate": 7.643866971275527e-06, "loss": 0.1367, "step": 13701 }, { "epoch": 0.88, "learning_rate": 7.635905940424414e-06, "loss": 0.1281, "step": 13702 }, { "epoch": 0.88, "learning_rate": 7.6279488928445945e-06, "loss": 0.1321, "step": 13703 }, { "epoch": 0.88, "learning_rate": 7.619995828879234e-06, "loss": 0.1289, "step": 13704 }, { "epoch": 0.88, "learning_rate": 7.612046748871327e-06, "loss": 0.1337, "step": 13705 }, { "epoch": 0.88, "learning_rate": 7.604101653163664e-06, "loss": 0.1365, "step": 13706 }, { "epoch": 0.88, "learning_rate": 7.596160542098863e-06, "loss": 0.1286, "step": 13707 }, { "epoch": 0.88, "learning_rate": 7.588223416019413e-06, "loss": 0.1259, "step": 13708 }, { "epoch": 0.88, "learning_rate": 7.5802902752676095e-06, "loss": 0.1235, "step": 13709 }, { "epoch": 0.88, "learning_rate": 7.572361120185589e-06, "loss": 0.1377, "step": 13710 }, { "epoch": 0.88, "learning_rate": 7.564435951115267e-06, "loss": 0.1366, "step": 13711 }, { "epoch": 0.88, "learning_rate": 7.556514768398459e-06, "loss": 0.1328, "step": 13712 }, { "epoch": 0.88, "learning_rate": 7.548597572376748e-06, "loss": 0.1196, "step": 13713 }, { "epoch": 0.88, "learning_rate": 7.54068436339157e-06, "loss": 0.1114, "step": 13714 }, { "epoch": 0.88, "learning_rate": 7.532775141784188e-06, "loss": 0.124, "step": 13715 }, { "epoch": 0.88, "learning_rate": 7.524869907895693e-06, "loss": 0.1294, "step": 13716 }, { "epoch": 0.88, "learning_rate": 7.516968662067025e-06, "loss": 0.1446, "step": 13717 }, { "epoch": 0.88, "learning_rate": 7.509071404638912e-06, "loss": 0.1407, "step": 13718 }, { "epoch": 0.88, "learning_rate": 7.5011781359519366e-06, "loss": 0.1325, "step": 13719 }, { "epoch": 0.88, "learning_rate": 7.4932888563465035e-06, "loss": 0.1382, "step": 13720 }, { "epoch": 0.88, "learning_rate": 7.485403566162829e-06, "loss": 0.1391, "step": 13721 }, { "epoch": 0.88, "learning_rate": 7.477522265740977e-06, "loss": 0.1266, "step": 13722 }, { "epoch": 0.88, "learning_rate": 7.46964495542084e-06, "loss": 0.1449, "step": 13723 }, { "epoch": 0.88, "learning_rate": 7.461771635542148e-06, "loss": 0.1336, "step": 13724 }, { "epoch": 0.88, "learning_rate": 7.453902306444405e-06, "loss": 0.1243, "step": 13725 }, { "epoch": 0.88, "learning_rate": 7.446036968467029e-06, "loss": 0.126, "step": 13726 }, { "epoch": 0.88, "learning_rate": 7.438175621949195e-06, "loss": 0.1273, "step": 13727 }, { "epoch": 0.88, "learning_rate": 7.430318267229908e-06, "loss": 0.133, "step": 13728 }, { "epoch": 0.88, "learning_rate": 7.4224649046480525e-06, "loss": 0.1288, "step": 13729 }, { "epoch": 0.88, "learning_rate": 7.414615534542291e-06, "loss": 0.1363, "step": 13730 }, { "epoch": 0.88, "learning_rate": 7.406770157251164e-06, "loss": 0.1408, "step": 13731 }, { "epoch": 0.88, "learning_rate": 7.398928773112979e-06, "loss": 0.1283, "step": 13732 }, { "epoch": 0.88, "learning_rate": 7.39109138246592e-06, "loss": 0.1451, "step": 13733 }, { "epoch": 0.88, "learning_rate": 7.383257985647973e-06, "loss": 0.141, "step": 13734 }, { "epoch": 0.88, "learning_rate": 7.375428582996946e-06, "loss": 0.1175, "step": 13735 }, { "epoch": 0.88, "learning_rate": 7.367603174850502e-06, "loss": 0.1223, "step": 13736 }, { "epoch": 0.88, "learning_rate": 7.359781761546114e-06, "loss": 0.1195, "step": 13737 }, { "epoch": 0.88, "learning_rate": 7.3519643434211135e-06, "loss": 0.1299, "step": 13738 }, { "epoch": 0.88, "learning_rate": 7.344150920812576e-06, "loss": 0.1463, "step": 13739 }, { "epoch": 0.88, "learning_rate": 7.336341494057519e-06, "loss": 0.1314, "step": 13740 }, { "epoch": 0.88, "learning_rate": 7.328536063492708e-06, "loss": 0.1272, "step": 13741 }, { "epoch": 0.88, "learning_rate": 7.320734629454728e-06, "loss": 0.1491, "step": 13742 }, { "epoch": 0.88, "learning_rate": 7.312937192280045e-06, "loss": 0.1296, "step": 13743 }, { "epoch": 0.88, "learning_rate": 7.305143752304944e-06, "loss": 0.131, "step": 13744 }, { "epoch": 0.88, "learning_rate": 7.297354309865523e-06, "loss": 0.114, "step": 13745 }, { "epoch": 0.88, "learning_rate": 7.289568865297691e-06, "loss": 0.1552, "step": 13746 }, { "epoch": 0.88, "learning_rate": 7.281787418937191e-06, "loss": 0.136, "step": 13747 }, { "epoch": 0.88, "learning_rate": 7.274009971119644e-06, "loss": 0.1455, "step": 13748 }, { "epoch": 0.88, "learning_rate": 7.266236522180392e-06, "loss": 0.1428, "step": 13749 }, { "epoch": 0.88, "learning_rate": 7.258467072454744e-06, "loss": 0.128, "step": 13750 }, { "epoch": 0.88, "learning_rate": 7.250701622277712e-06, "loss": 0.1176, "step": 13751 }, { "epoch": 0.88, "learning_rate": 7.242940171984225e-06, "loss": 0.1439, "step": 13752 }, { "epoch": 0.88, "learning_rate": 7.235182721908984e-06, "loss": 0.1243, "step": 13753 }, { "epoch": 0.88, "learning_rate": 7.227429272386521e-06, "loss": 0.1275, "step": 13754 }, { "epoch": 0.88, "learning_rate": 7.219679823751213e-06, "loss": 0.1246, "step": 13755 }, { "epoch": 0.88, "learning_rate": 7.2119343763372704e-06, "loss": 0.1419, "step": 13756 }, { "epoch": 0.88, "learning_rate": 7.204192930478737e-06, "loss": 0.1327, "step": 13757 }, { "epoch": 0.88, "learning_rate": 7.1964554865094456e-06, "loss": 0.1219, "step": 13758 }, { "epoch": 0.88, "learning_rate": 7.1887220447630964e-06, "loss": 0.1224, "step": 13759 }, { "epoch": 0.88, "learning_rate": 7.1809926055731894e-06, "loss": 0.1377, "step": 13760 }, { "epoch": 0.88, "learning_rate": 7.1732671692730345e-06, "loss": 0.144, "step": 13761 }, { "epoch": 0.88, "learning_rate": 7.165545736195844e-06, "loss": 0.1514, "step": 13762 }, { "epoch": 0.88, "learning_rate": 7.157828306674575e-06, "loss": 0.1348, "step": 13763 }, { "epoch": 0.88, "learning_rate": 7.150114881042091e-06, "loss": 0.1457, "step": 13764 }, { "epoch": 0.88, "learning_rate": 7.142405459630996e-06, "loss": 0.1362, "step": 13765 }, { "epoch": 0.88, "learning_rate": 7.13470004277379e-06, "loss": 0.1438, "step": 13766 }, { "epoch": 0.88, "learning_rate": 7.126998630802772e-06, "loss": 0.1385, "step": 13767 }, { "epoch": 0.88, "learning_rate": 7.119301224050045e-06, "loss": 0.1342, "step": 13768 }, { "epoch": 0.88, "learning_rate": 7.111607822847577e-06, "loss": 0.1385, "step": 13769 }, { "epoch": 0.88, "learning_rate": 7.103918427527179e-06, "loss": 0.145, "step": 13770 }, { "epoch": 0.88, "learning_rate": 7.0962330384204414e-06, "loss": 0.1402, "step": 13771 }, { "epoch": 0.88, "learning_rate": 7.0885516558588e-06, "loss": 0.1222, "step": 13772 }, { "epoch": 0.88, "learning_rate": 7.080874280173533e-06, "loss": 0.1441, "step": 13773 }, { "epoch": 0.88, "learning_rate": 7.0732009116957324e-06, "loss": 0.1303, "step": 13774 }, { "epoch": 0.88, "learning_rate": 7.0655315507562884e-06, "loss": 0.1383, "step": 13775 }, { "epoch": 0.88, "learning_rate": 7.0578661976859695e-06, "loss": 0.1362, "step": 13776 }, { "epoch": 0.88, "learning_rate": 7.0502048528153455e-06, "loss": 0.1469, "step": 13777 }, { "epoch": 0.88, "learning_rate": 7.042547516474851e-06, "loss": 0.1376, "step": 13778 }, { "epoch": 0.88, "learning_rate": 7.034894188994656e-06, "loss": 0.1359, "step": 13779 }, { "epoch": 0.88, "learning_rate": 7.027244870704874e-06, "loss": 0.1386, "step": 13780 }, { "epoch": 0.88, "learning_rate": 7.019599561935352e-06, "loss": 0.1464, "step": 13781 }, { "epoch": 0.88, "learning_rate": 7.0119582630157835e-06, "loss": 0.1319, "step": 13782 }, { "epoch": 0.88, "learning_rate": 7.0043209742757356e-06, "loss": 0.1523, "step": 13783 }, { "epoch": 0.88, "learning_rate": 6.996687696044568e-06, "loss": 0.1343, "step": 13784 }, { "epoch": 0.88, "learning_rate": 6.9890584286514845e-06, "loss": 0.1352, "step": 13785 }, { "epoch": 0.88, "learning_rate": 6.981433172425467e-06, "loss": 0.1389, "step": 13786 }, { "epoch": 0.88, "learning_rate": 6.973811927695406e-06, "loss": 0.14, "step": 13787 }, { "epoch": 0.88, "learning_rate": 6.96619469478994e-06, "loss": 0.1125, "step": 13788 }, { "epoch": 0.88, "learning_rate": 6.958581474037562e-06, "loss": 0.1475, "step": 13789 }, { "epoch": 0.88, "learning_rate": 6.9509722657666085e-06, "loss": 0.1482, "step": 13790 }, { "epoch": 0.88, "learning_rate": 6.94336707030524e-06, "loss": 0.1339, "step": 13791 }, { "epoch": 0.88, "learning_rate": 6.935765887981438e-06, "loss": 0.1471, "step": 13792 }, { "epoch": 0.88, "learning_rate": 6.928168719122996e-06, "loss": 0.1492, "step": 13793 }, { "epoch": 0.88, "learning_rate": 6.920575564057574e-06, "loss": 0.1366, "step": 13794 }, { "epoch": 0.88, "learning_rate": 6.9129864231126e-06, "loss": 0.1158, "step": 13795 }, { "epoch": 0.88, "learning_rate": 6.905401296615366e-06, "loss": 0.1288, "step": 13796 }, { "epoch": 0.88, "learning_rate": 6.897820184893011e-06, "loss": 0.1399, "step": 13797 }, { "epoch": 0.88, "learning_rate": 6.8902430882724525e-06, "loss": 0.1324, "step": 13798 }, { "epoch": 0.88, "learning_rate": 6.882670007080472e-06, "loss": 0.1298, "step": 13799 }, { "epoch": 0.88, "learning_rate": 6.875100941643653e-06, "loss": 0.1304, "step": 13800 }, { "epoch": 0.88, "learning_rate": 6.867535892288435e-06, "loss": 0.1288, "step": 13801 }, { "epoch": 0.88, "learning_rate": 6.859974859341034e-06, "loss": 0.1218, "step": 13802 }, { "epoch": 0.89, "learning_rate": 6.852417843127557e-06, "loss": 0.1443, "step": 13803 }, { "epoch": 0.89, "learning_rate": 6.844864843973897e-06, "loss": 0.1423, "step": 13804 }, { "epoch": 0.89, "learning_rate": 6.837315862205773e-06, "loss": 0.1391, "step": 13805 }, { "epoch": 0.89, "learning_rate": 6.829770898148757e-06, "loss": 0.12, "step": 13806 }, { "epoch": 0.89, "learning_rate": 6.822229952128212e-06, "loss": 0.1245, "step": 13807 }, { "epoch": 0.89, "learning_rate": 6.814693024469376e-06, "loss": 0.1498, "step": 13808 }, { "epoch": 0.89, "learning_rate": 6.8071601154972465e-06, "loss": 0.137, "step": 13809 }, { "epoch": 0.89, "learning_rate": 6.7996312255367065e-06, "loss": 0.1411, "step": 13810 }, { "epoch": 0.89, "learning_rate": 6.792106354912464e-06, "loss": 0.1293, "step": 13811 }, { "epoch": 0.89, "learning_rate": 6.784585503948993e-06, "loss": 0.1612, "step": 13812 }, { "epoch": 0.89, "learning_rate": 6.777068672970666e-06, "loss": 0.1561, "step": 13813 }, { "epoch": 0.89, "learning_rate": 6.769555862301635e-06, "loss": 0.1495, "step": 13814 }, { "epoch": 0.89, "learning_rate": 6.76204707226592e-06, "loss": 0.1241, "step": 13815 }, { "epoch": 0.89, "learning_rate": 6.754542303187316e-06, "loss": 0.1317, "step": 13816 }, { "epoch": 0.89, "learning_rate": 6.747041555389478e-06, "loss": 0.1126, "step": 13817 }, { "epoch": 0.89, "learning_rate": 6.73954482919591e-06, "loss": 0.1439, "step": 13818 }, { "epoch": 0.89, "learning_rate": 6.732052124929866e-06, "loss": 0.1338, "step": 13819 }, { "epoch": 0.89, "learning_rate": 6.724563442914522e-06, "loss": 0.1299, "step": 13820 }, { "epoch": 0.89, "learning_rate": 6.717078783472785e-06, "loss": 0.1298, "step": 13821 }, { "epoch": 0.89, "learning_rate": 6.7095981469274875e-06, "loss": 0.152, "step": 13822 }, { "epoch": 0.89, "learning_rate": 6.702121533601191e-06, "loss": 0.1325, "step": 13823 }, { "epoch": 0.89, "learning_rate": 6.6946489438163505e-06, "loss": 0.1545, "step": 13824 }, { "epoch": 0.89, "learning_rate": 6.687180377895241e-06, "loss": 0.1186, "step": 13825 }, { "epoch": 0.89, "learning_rate": 6.6797158361599275e-06, "loss": 0.1243, "step": 13826 }, { "epoch": 0.89, "learning_rate": 6.672255318932341e-06, "loss": 0.1343, "step": 13827 }, { "epoch": 0.89, "learning_rate": 6.664798826534224e-06, "loss": 0.1431, "step": 13828 }, { "epoch": 0.89, "learning_rate": 6.6573463592871085e-06, "loss": 0.1232, "step": 13829 }, { "epoch": 0.89, "learning_rate": 6.6498979175124156e-06, "loss": 0.1265, "step": 13830 }, { "epoch": 0.89, "learning_rate": 6.642453501531365e-06, "loss": 0.1282, "step": 13831 }, { "epoch": 0.89, "learning_rate": 6.635013111665011e-06, "loss": 0.1278, "step": 13832 }, { "epoch": 0.89, "learning_rate": 6.62757674823421e-06, "loss": 0.1214, "step": 13833 }, { "epoch": 0.89, "learning_rate": 6.62014441155967e-06, "loss": 0.1239, "step": 13834 }, { "epoch": 0.89, "learning_rate": 6.6127161019619136e-06, "loss": 0.1179, "step": 13835 }, { "epoch": 0.89, "learning_rate": 6.605291819761261e-06, "loss": 0.1461, "step": 13836 }, { "epoch": 0.89, "learning_rate": 6.597871565277946e-06, "loss": 0.1213, "step": 13837 }, { "epoch": 0.89, "learning_rate": 6.590455338831936e-06, "loss": 0.1361, "step": 13838 }, { "epoch": 0.89, "learning_rate": 6.583043140743084e-06, "loss": 0.152, "step": 13839 }, { "epoch": 0.89, "learning_rate": 6.575634971331024e-06, "loss": 0.1291, "step": 13840 }, { "epoch": 0.89, "learning_rate": 6.568230830915268e-06, "loss": 0.1485, "step": 13841 }, { "epoch": 0.89, "learning_rate": 6.560830719815081e-06, "loss": 0.1375, "step": 13842 }, { "epoch": 0.89, "learning_rate": 6.55343463834962e-06, "loss": 0.1408, "step": 13843 }, { "epoch": 0.89, "learning_rate": 6.546042586837875e-06, "loss": 0.1337, "step": 13844 }, { "epoch": 0.89, "learning_rate": 6.53865456559859e-06, "loss": 0.1256, "step": 13845 }, { "epoch": 0.89, "learning_rate": 6.531270574950421e-06, "loss": 0.1244, "step": 13846 }, { "epoch": 0.89, "learning_rate": 6.523890615211758e-06, "loss": 0.1327, "step": 13847 }, { "epoch": 0.89, "learning_rate": 6.516514686700914e-06, "loss": 0.1385, "step": 13848 }, { "epoch": 0.89, "learning_rate": 6.509142789735945e-06, "loss": 0.1311, "step": 13849 }, { "epoch": 0.89, "learning_rate": 6.5017749246347955e-06, "loss": 0.1267, "step": 13850 }, { "epoch": 0.89, "learning_rate": 6.494411091715202e-06, "loss": 0.123, "step": 13851 }, { "epoch": 0.89, "learning_rate": 6.4870512912947324e-06, "loss": 0.1315, "step": 13852 }, { "epoch": 0.89, "learning_rate": 6.4796955236907985e-06, "loss": 0.1478, "step": 13853 }, { "epoch": 0.89, "learning_rate": 6.472343789220592e-06, "loss": 0.1299, "step": 13854 }, { "epoch": 0.89, "learning_rate": 6.464996088201192e-06, "loss": 0.1197, "step": 13855 }, { "epoch": 0.89, "learning_rate": 6.457652420949456e-06, "loss": 0.127, "step": 13856 }, { "epoch": 0.89, "learning_rate": 6.4503127877820865e-06, "loss": 0.134, "step": 13857 }, { "epoch": 0.89, "learning_rate": 6.442977189015631e-06, "loss": 0.1349, "step": 13858 }, { "epoch": 0.89, "learning_rate": 6.435645624966413e-06, "loss": 0.1205, "step": 13859 }, { "epoch": 0.89, "learning_rate": 6.428318095950647e-06, "loss": 0.1322, "step": 13860 }, { "epoch": 0.89, "learning_rate": 6.420994602284292e-06, "loss": 0.1389, "step": 13861 }, { "epoch": 0.89, "learning_rate": 6.413675144283226e-06, "loss": 0.1407, "step": 13862 }, { "epoch": 0.89, "learning_rate": 6.406359722263067e-06, "loss": 0.1324, "step": 13863 }, { "epoch": 0.89, "learning_rate": 6.399048336539326e-06, "loss": 0.1607, "step": 13864 }, { "epoch": 0.89, "learning_rate": 6.3917409874273085e-06, "loss": 0.1351, "step": 13865 }, { "epoch": 0.89, "learning_rate": 6.3844376752421275e-06, "loss": 0.1252, "step": 13866 }, { "epoch": 0.89, "learning_rate": 6.377138400298776e-06, "loss": 0.1442, "step": 13867 }, { "epoch": 0.89, "learning_rate": 6.369843162912004e-06, "loss": 0.1505, "step": 13868 }, { "epoch": 0.89, "learning_rate": 6.362551963396468e-06, "loss": 0.1339, "step": 13869 }, { "epoch": 0.89, "learning_rate": 6.355264802066563e-06, "loss": 0.1387, "step": 13870 }, { "epoch": 0.89, "learning_rate": 6.34798167923657e-06, "loss": 0.1483, "step": 13871 }, { "epoch": 0.89, "learning_rate": 6.340702595220604e-06, "loss": 0.141, "step": 13872 }, { "epoch": 0.89, "learning_rate": 6.333427550332538e-06, "loss": 0.1477, "step": 13873 }, { "epoch": 0.89, "learning_rate": 6.326156544886153e-06, "loss": 0.1243, "step": 13874 }, { "epoch": 0.89, "learning_rate": 6.318889579194975e-06, "loss": 0.1378, "step": 13875 }, { "epoch": 0.89, "learning_rate": 6.311626653572444e-06, "loss": 0.1299, "step": 13876 }, { "epoch": 0.89, "learning_rate": 6.304367768331742e-06, "loss": 0.1539, "step": 13877 }, { "epoch": 0.89, "learning_rate": 6.29711292378593e-06, "loss": 0.1315, "step": 13878 }, { "epoch": 0.89, "learning_rate": 6.289862120247891e-06, "loss": 0.1201, "step": 13879 }, { "epoch": 0.89, "learning_rate": 6.282615358030286e-06, "loss": 0.1395, "step": 13880 }, { "epoch": 0.89, "learning_rate": 6.2753726374456755e-06, "loss": 0.1287, "step": 13881 }, { "epoch": 0.89, "learning_rate": 6.268133958806377e-06, "loss": 0.1475, "step": 13882 }, { "epoch": 0.89, "learning_rate": 6.260899322424574e-06, "loss": 0.1465, "step": 13883 }, { "epoch": 0.89, "learning_rate": 6.253668728612294e-06, "loss": 0.1349, "step": 13884 }, { "epoch": 0.89, "learning_rate": 6.246442177681311e-06, "loss": 0.1341, "step": 13885 }, { "epoch": 0.89, "learning_rate": 6.239219669943319e-06, "loss": 0.1481, "step": 13886 }, { "epoch": 0.89, "learning_rate": 6.232001205709758e-06, "loss": 0.1381, "step": 13887 }, { "epoch": 0.89, "learning_rate": 6.22478678529197e-06, "loss": 0.1379, "step": 13888 }, { "epoch": 0.89, "learning_rate": 6.217576409001036e-06, "loss": 0.1379, "step": 13889 }, { "epoch": 0.89, "learning_rate": 6.210370077147931e-06, "loss": 0.1328, "step": 13890 }, { "epoch": 0.89, "learning_rate": 6.2031677900434625e-06, "loss": 0.1325, "step": 13891 }, { "epoch": 0.89, "learning_rate": 6.195969547998182e-06, "loss": 0.1333, "step": 13892 }, { "epoch": 0.89, "learning_rate": 6.188775351322562e-06, "loss": 0.1197, "step": 13893 }, { "epoch": 0.89, "learning_rate": 6.1815852003268225e-06, "loss": 0.1182, "step": 13894 }, { "epoch": 0.89, "learning_rate": 6.174399095321082e-06, "loss": 0.1269, "step": 13895 }, { "epoch": 0.89, "learning_rate": 6.167217036615214e-06, "loss": 0.1497, "step": 13896 }, { "epoch": 0.89, "learning_rate": 6.16003902451896e-06, "loss": 0.1235, "step": 13897 }, { "epoch": 0.89, "learning_rate": 6.152865059341894e-06, "loss": 0.1194, "step": 13898 }, { "epoch": 0.89, "learning_rate": 6.145695141393371e-06, "loss": 0.1308, "step": 13899 }, { "epoch": 0.89, "learning_rate": 6.138529270982629e-06, "loss": 0.1202, "step": 13900 }, { "epoch": 0.89, "learning_rate": 6.131367448418668e-06, "loss": 0.1177, "step": 13901 }, { "epoch": 0.89, "learning_rate": 6.124209674010373e-06, "loss": 0.1373, "step": 13902 }, { "epoch": 0.89, "learning_rate": 6.117055948066419e-06, "loss": 0.1211, "step": 13903 }, { "epoch": 0.89, "learning_rate": 6.109906270895305e-06, "loss": 0.1525, "step": 13904 }, { "epoch": 0.89, "learning_rate": 6.102760642805394e-06, "loss": 0.1501, "step": 13905 }, { "epoch": 0.89, "learning_rate": 6.095619064104818e-06, "loss": 0.1366, "step": 13906 }, { "epoch": 0.89, "learning_rate": 6.088481535101598e-06, "loss": 0.1306, "step": 13907 }, { "epoch": 0.89, "learning_rate": 6.081348056103498e-06, "loss": 0.1479, "step": 13908 }, { "epoch": 0.89, "learning_rate": 6.074218627418205e-06, "loss": 0.1374, "step": 13909 }, { "epoch": 0.89, "learning_rate": 6.067093249353128e-06, "loss": 0.1296, "step": 13910 }, { "epoch": 0.89, "learning_rate": 6.05997192221559e-06, "loss": 0.1321, "step": 13911 }, { "epoch": 0.89, "learning_rate": 6.0528546463127e-06, "loss": 0.1453, "step": 13912 }, { "epoch": 0.89, "learning_rate": 6.04574142195139e-06, "loss": 0.1261, "step": 13913 }, { "epoch": 0.89, "learning_rate": 6.038632249438425e-06, "loss": 0.131, "step": 13914 }, { "epoch": 0.89, "learning_rate": 6.031527129080395e-06, "loss": 0.1624, "step": 13915 }, { "epoch": 0.89, "learning_rate": 6.024426061183719e-06, "loss": 0.1365, "step": 13916 }, { "epoch": 0.89, "learning_rate": 6.0173290460546096e-06, "loss": 0.1391, "step": 13917 }, { "epoch": 0.89, "learning_rate": 6.010236083999154e-06, "loss": 0.1462, "step": 13918 }, { "epoch": 0.89, "learning_rate": 6.003147175323243e-06, "loss": 0.1223, "step": 13919 }, { "epoch": 0.89, "learning_rate": 5.996062320332585e-06, "loss": 0.1521, "step": 13920 }, { "epoch": 0.89, "learning_rate": 5.988981519332726e-06, "loss": 0.1341, "step": 13921 }, { "epoch": 0.89, "learning_rate": 5.98190477262901e-06, "loss": 0.1414, "step": 13922 }, { "epoch": 0.89, "learning_rate": 5.974832080526649e-06, "loss": 0.1288, "step": 13923 }, { "epoch": 0.89, "learning_rate": 5.967763443330654e-06, "loss": 0.1274, "step": 13924 }, { "epoch": 0.89, "learning_rate": 5.960698861345859e-06, "loss": 0.1435, "step": 13925 }, { "epoch": 0.89, "learning_rate": 5.953638334876943e-06, "loss": 0.146, "step": 13926 }, { "epoch": 0.89, "learning_rate": 5.946581864228373e-06, "loss": 0.1397, "step": 13927 }, { "epoch": 0.89, "learning_rate": 5.939529449704495e-06, "loss": 0.1358, "step": 13928 }, { "epoch": 0.89, "learning_rate": 5.932481091609421e-06, "loss": 0.14, "step": 13929 }, { "epoch": 0.89, "learning_rate": 5.925436790247118e-06, "loss": 0.1371, "step": 13930 }, { "epoch": 0.89, "learning_rate": 5.918396545921412e-06, "loss": 0.1372, "step": 13931 }, { "epoch": 0.89, "learning_rate": 5.91136035893588e-06, "loss": 0.1379, "step": 13932 }, { "epoch": 0.89, "learning_rate": 5.90432822959398e-06, "loss": 0.1443, "step": 13933 }, { "epoch": 0.89, "learning_rate": 5.897300158198971e-06, "loss": 0.1418, "step": 13934 }, { "epoch": 0.89, "learning_rate": 5.890276145053963e-06, "loss": 0.1382, "step": 13935 }, { "epoch": 0.89, "learning_rate": 5.8832561904618385e-06, "loss": 0.1366, "step": 13936 }, { "epoch": 0.89, "learning_rate": 5.8762402947253546e-06, "loss": 0.1441, "step": 13937 }, { "epoch": 0.89, "learning_rate": 5.869228458147103e-06, "loss": 0.1302, "step": 13938 }, { "epoch": 0.89, "learning_rate": 5.862220681029429e-06, "loss": 0.1318, "step": 13939 }, { "epoch": 0.89, "learning_rate": 5.855216963674592e-06, "loss": 0.1419, "step": 13940 }, { "epoch": 0.89, "learning_rate": 5.848217306384584e-06, "loss": 0.1356, "step": 13941 }, { "epoch": 0.89, "learning_rate": 5.8412217094613195e-06, "loss": 0.1273, "step": 13942 }, { "epoch": 0.89, "learning_rate": 5.834230173206445e-06, "loss": 0.1317, "step": 13943 }, { "epoch": 0.89, "learning_rate": 5.827242697921509e-06, "loss": 0.1186, "step": 13944 }, { "epoch": 0.89, "learning_rate": 5.820259283907847e-06, "loss": 0.1329, "step": 13945 }, { "epoch": 0.89, "learning_rate": 5.8132799314665975e-06, "loss": 0.1271, "step": 13946 }, { "epoch": 0.89, "learning_rate": 5.8063046408987965e-06, "loss": 0.1517, "step": 13947 }, { "epoch": 0.89, "learning_rate": 5.799333412505215e-06, "loss": 0.1444, "step": 13948 }, { "epoch": 0.89, "learning_rate": 5.792366246586511e-06, "loss": 0.1403, "step": 13949 }, { "epoch": 0.89, "learning_rate": 5.785403143443146e-06, "loss": 0.1435, "step": 13950 }, { "epoch": 0.89, "learning_rate": 5.7784441033754025e-06, "loss": 0.1386, "step": 13951 }, { "epoch": 0.89, "learning_rate": 5.771489126683427e-06, "loss": 0.1202, "step": 13952 }, { "epoch": 0.89, "learning_rate": 5.764538213667104e-06, "loss": 0.1236, "step": 13953 }, { "epoch": 0.89, "learning_rate": 5.757591364626258e-06, "loss": 0.1385, "step": 13954 }, { "epoch": 0.89, "learning_rate": 5.750648579860418e-06, "loss": 0.128, "step": 13955 }, { "epoch": 0.89, "learning_rate": 5.743709859669044e-06, "loss": 0.1221, "step": 13956 }, { "epoch": 0.89, "learning_rate": 5.73677520435133e-06, "loss": 0.1334, "step": 13957 }, { "epoch": 0.89, "learning_rate": 5.72984461420637e-06, "loss": 0.1276, "step": 13958 }, { "epoch": 0.9, "learning_rate": 5.722918089533058e-06, "loss": 0.1546, "step": 13959 }, { "epoch": 0.9, "learning_rate": 5.715995630630078e-06, "loss": 0.1205, "step": 13960 }, { "epoch": 0.9, "learning_rate": 5.70907723779599e-06, "loss": 0.1331, "step": 13961 }, { "epoch": 0.9, "learning_rate": 5.702162911329123e-06, "loss": 0.1379, "step": 13962 }, { "epoch": 0.9, "learning_rate": 5.695252651527705e-06, "loss": 0.1317, "step": 13963 }, { "epoch": 0.9, "learning_rate": 5.688346458689697e-06, "loss": 0.1399, "step": 13964 }, { "epoch": 0.9, "learning_rate": 5.681444333112984e-06, "loss": 0.1423, "step": 13965 }, { "epoch": 0.9, "learning_rate": 5.674546275095205e-06, "loss": 0.1375, "step": 13966 }, { "epoch": 0.9, "learning_rate": 5.667652284933822e-06, "loss": 0.1351, "step": 13967 }, { "epoch": 0.9, "learning_rate": 5.660762362926186e-06, "loss": 0.1493, "step": 13968 }, { "epoch": 0.9, "learning_rate": 5.653876509369393e-06, "loss": 0.1243, "step": 13969 }, { "epoch": 0.9, "learning_rate": 5.646994724560406e-06, "loss": 0.1341, "step": 13970 }, { "epoch": 0.9, "learning_rate": 5.64011700879602e-06, "loss": 0.1321, "step": 13971 }, { "epoch": 0.9, "learning_rate": 5.633243362372853e-06, "loss": 0.1348, "step": 13972 }, { "epoch": 0.9, "learning_rate": 5.626373785587324e-06, "loss": 0.1509, "step": 13973 }, { "epoch": 0.9, "learning_rate": 5.619508278735663e-06, "loss": 0.1508, "step": 13974 }, { "epoch": 0.9, "learning_rate": 5.612646842114e-06, "loss": 0.1367, "step": 13975 }, { "epoch": 0.9, "learning_rate": 5.605789476018186e-06, "loss": 0.1444, "step": 13976 }, { "epoch": 0.9, "learning_rate": 5.598936180743975e-06, "loss": 0.1413, "step": 13977 }, { "epoch": 0.9, "learning_rate": 5.5920869565869285e-06, "loss": 0.1254, "step": 13978 }, { "epoch": 0.9, "learning_rate": 5.585241803842434e-06, "loss": 0.1299, "step": 13979 }, { "epoch": 0.9, "learning_rate": 5.578400722805677e-06, "loss": 0.1406, "step": 13980 }, { "epoch": 0.9, "learning_rate": 5.571563713771666e-06, "loss": 0.1314, "step": 13981 }, { "epoch": 0.9, "learning_rate": 5.564730777035288e-06, "loss": 0.1246, "step": 13982 }, { "epoch": 0.9, "learning_rate": 5.557901912891184e-06, "loss": 0.1445, "step": 13983 }, { "epoch": 0.9, "learning_rate": 5.551077121633874e-06, "loss": 0.1495, "step": 13984 }, { "epoch": 0.9, "learning_rate": 5.544256403557691e-06, "loss": 0.133, "step": 13985 }, { "epoch": 0.9, "learning_rate": 5.537439758956764e-06, "loss": 0.1334, "step": 13986 }, { "epoch": 0.9, "learning_rate": 5.530627188125092e-06, "loss": 0.1396, "step": 13987 }, { "epoch": 0.9, "learning_rate": 5.52381869135643e-06, "loss": 0.1301, "step": 13988 }, { "epoch": 0.9, "learning_rate": 5.517014268944443e-06, "loss": 0.1205, "step": 13989 }, { "epoch": 0.9, "learning_rate": 5.51021392118255e-06, "loss": 0.1344, "step": 13990 }, { "epoch": 0.9, "learning_rate": 5.503417648364018e-06, "loss": 0.1456, "step": 13991 }, { "epoch": 0.9, "learning_rate": 5.496625450781978e-06, "loss": 0.1455, "step": 13992 }, { "epoch": 0.9, "learning_rate": 5.489837328729297e-06, "loss": 0.1471, "step": 13993 }, { "epoch": 0.9, "learning_rate": 5.4830532824987735e-06, "loss": 0.1185, "step": 13994 }, { "epoch": 0.9, "learning_rate": 5.476273312382918e-06, "loss": 0.1314, "step": 13995 }, { "epoch": 0.9, "learning_rate": 5.469497418674163e-06, "loss": 0.1341, "step": 13996 }, { "epoch": 0.9, "learning_rate": 5.4627256016647085e-06, "loss": 0.1336, "step": 13997 }, { "epoch": 0.9, "learning_rate": 5.455957861646588e-06, "loss": 0.1506, "step": 13998 }, { "epoch": 0.9, "learning_rate": 5.449194198911689e-06, "loss": 0.1446, "step": 13999 }, { "epoch": 0.9, "learning_rate": 5.442434613751668e-06, "loss": 0.1481, "step": 14000 }, { "epoch": 0.9, "learning_rate": 5.435679106458069e-06, "loss": 0.1414, "step": 14001 }, { "epoch": 0.9, "learning_rate": 5.428927677322193e-06, "loss": 0.1344, "step": 14002 }, { "epoch": 0.9, "learning_rate": 5.42218032663524e-06, "loss": 0.1377, "step": 14003 }, { "epoch": 0.9, "learning_rate": 5.415437054688144e-06, "loss": 0.1349, "step": 14004 }, { "epoch": 0.9, "learning_rate": 5.408697861771772e-06, "loss": 0.1245, "step": 14005 }, { "epoch": 0.9, "learning_rate": 5.401962748176737e-06, "loss": 0.1234, "step": 14006 }, { "epoch": 0.9, "learning_rate": 5.395231714193461e-06, "loss": 0.1456, "step": 14007 }, { "epoch": 0.9, "learning_rate": 5.388504760112278e-06, "loss": 0.1535, "step": 14008 }, { "epoch": 0.9, "learning_rate": 5.381781886223247e-06, "loss": 0.1465, "step": 14009 }, { "epoch": 0.9, "learning_rate": 5.375063092816313e-06, "loss": 0.13, "step": 14010 }, { "epoch": 0.9, "learning_rate": 5.368348380181243e-06, "loss": 0.1296, "step": 14011 }, { "epoch": 0.9, "learning_rate": 5.361637748607606e-06, "loss": 0.1242, "step": 14012 }, { "epoch": 0.9, "learning_rate": 5.354931198384805e-06, "loss": 0.1178, "step": 14013 }, { "epoch": 0.9, "learning_rate": 5.3482287298020515e-06, "loss": 0.1413, "step": 14014 }, { "epoch": 0.9, "learning_rate": 5.3415303431484155e-06, "loss": 0.1203, "step": 14015 }, { "epoch": 0.9, "learning_rate": 5.3348360387127425e-06, "loss": 0.123, "step": 14016 }, { "epoch": 0.9, "learning_rate": 5.328145816783747e-06, "loss": 0.1365, "step": 14017 }, { "epoch": 0.9, "learning_rate": 5.321459677649943e-06, "loss": 0.1327, "step": 14018 }, { "epoch": 0.9, "learning_rate": 5.314777621599698e-06, "loss": 0.1422, "step": 14019 }, { "epoch": 0.9, "learning_rate": 5.308099648921172e-06, "loss": 0.1321, "step": 14020 }, { "epoch": 0.9, "learning_rate": 5.301425759902323e-06, "loss": 0.1436, "step": 14021 }, { "epoch": 0.9, "learning_rate": 5.29475595483101e-06, "loss": 0.1497, "step": 14022 }, { "epoch": 0.9, "learning_rate": 5.288090233994847e-06, "loss": 0.1245, "step": 14023 }, { "epoch": 0.9, "learning_rate": 5.2814285976813036e-06, "loss": 0.1318, "step": 14024 }, { "epoch": 0.9, "learning_rate": 5.274771046177685e-06, "loss": 0.128, "step": 14025 }, { "epoch": 0.9, "learning_rate": 5.268117579771092e-06, "loss": 0.1424, "step": 14026 }, { "epoch": 0.9, "learning_rate": 5.261468198748454e-06, "loss": 0.1197, "step": 14027 }, { "epoch": 0.9, "learning_rate": 5.254822903396528e-06, "loss": 0.1378, "step": 14028 }, { "epoch": 0.9, "learning_rate": 5.248181694001919e-06, "loss": 0.1205, "step": 14029 }, { "epoch": 0.9, "learning_rate": 5.241544570851009e-06, "loss": 0.1309, "step": 14030 }, { "epoch": 0.9, "learning_rate": 5.234911534230036e-06, "loss": 0.1392, "step": 14031 }, { "epoch": 0.9, "learning_rate": 5.22828258442506e-06, "loss": 0.1371, "step": 14032 }, { "epoch": 0.9, "learning_rate": 5.221657721721973e-06, "loss": 0.1245, "step": 14033 }, { "epoch": 0.9, "learning_rate": 5.21503694640646e-06, "loss": 0.1433, "step": 14034 }, { "epoch": 0.9, "learning_rate": 5.208420258764035e-06, "loss": 0.1611, "step": 14035 }, { "epoch": 0.9, "learning_rate": 5.2018076590800826e-06, "loss": 0.1392, "step": 14036 }, { "epoch": 0.9, "learning_rate": 5.19519914763974e-06, "loss": 0.1274, "step": 14037 }, { "epoch": 0.9, "learning_rate": 5.1885947247280135e-06, "loss": 0.1418, "step": 14038 }, { "epoch": 0.9, "learning_rate": 5.18199439062973e-06, "loss": 0.1395, "step": 14039 }, { "epoch": 0.9, "learning_rate": 5.1753981456295504e-06, "loss": 0.1356, "step": 14040 }, { "epoch": 0.9, "learning_rate": 5.1688059900119265e-06, "loss": 0.1339, "step": 14041 }, { "epoch": 0.9, "learning_rate": 5.16221792406113e-06, "loss": 0.1354, "step": 14042 }, { "epoch": 0.9, "learning_rate": 5.155633948061323e-06, "loss": 0.1365, "step": 14043 }, { "epoch": 0.9, "learning_rate": 5.149054062296388e-06, "loss": 0.1313, "step": 14044 }, { "epoch": 0.9, "learning_rate": 5.142478267050122e-06, "loss": 0.1301, "step": 14045 }, { "epoch": 0.9, "learning_rate": 5.135906562606119e-06, "loss": 0.1333, "step": 14046 }, { "epoch": 0.9, "learning_rate": 5.129338949247775e-06, "loss": 0.129, "step": 14047 }, { "epoch": 0.9, "learning_rate": 5.122775427258331e-06, "loss": 0.1403, "step": 14048 }, { "epoch": 0.9, "learning_rate": 5.116215996920815e-06, "loss": 0.1412, "step": 14049 }, { "epoch": 0.9, "learning_rate": 5.109660658518145e-06, "loss": 0.1379, "step": 14050 }, { "epoch": 0.9, "learning_rate": 5.103109412332985e-06, "loss": 0.1369, "step": 14051 }, { "epoch": 0.9, "learning_rate": 5.096562258647919e-06, "loss": 0.1427, "step": 14052 }, { "epoch": 0.9, "learning_rate": 5.090019197745244e-06, "loss": 0.1252, "step": 14053 }, { "epoch": 0.9, "learning_rate": 5.083480229907178e-06, "loss": 0.1246, "step": 14054 }, { "epoch": 0.9, "learning_rate": 5.076945355415686e-06, "loss": 0.1462, "step": 14055 }, { "epoch": 0.9, "learning_rate": 5.070414574552595e-06, "loss": 0.1446, "step": 14056 }, { "epoch": 0.9, "learning_rate": 5.06388788759955e-06, "loss": 0.1308, "step": 14057 }, { "epoch": 0.9, "learning_rate": 5.057365294838035e-06, "loss": 0.134, "step": 14058 }, { "epoch": 0.9, "learning_rate": 5.0508467965493355e-06, "loss": 0.1393, "step": 14059 }, { "epoch": 0.9, "learning_rate": 5.0443323930145616e-06, "loss": 0.1378, "step": 14060 }, { "epoch": 0.9, "learning_rate": 5.037822084514643e-06, "loss": 0.1236, "step": 14061 }, { "epoch": 0.9, "learning_rate": 5.031315871330356e-06, "loss": 0.1422, "step": 14062 }, { "epoch": 0.9, "learning_rate": 5.024813753742274e-06, "loss": 0.1324, "step": 14063 }, { "epoch": 0.9, "learning_rate": 5.018315732030821e-06, "loss": 0.1195, "step": 14064 }, { "epoch": 0.9, "learning_rate": 5.0118218064762135e-06, "loss": 0.1209, "step": 14065 }, { "epoch": 0.9, "learning_rate": 5.005331977358529e-06, "loss": 0.1517, "step": 14066 }, { "epoch": 0.9, "learning_rate": 4.998846244957633e-06, "loss": 0.1417, "step": 14067 }, { "epoch": 0.9, "learning_rate": 4.992364609553213e-06, "loss": 0.1326, "step": 14068 }, { "epoch": 0.9, "learning_rate": 4.985887071424822e-06, "loss": 0.1575, "step": 14069 }, { "epoch": 0.9, "learning_rate": 4.97941363085177e-06, "loss": 0.137, "step": 14070 }, { "epoch": 0.9, "learning_rate": 4.972944288113268e-06, "loss": 0.1239, "step": 14071 }, { "epoch": 0.9, "learning_rate": 4.966479043488293e-06, "loss": 0.149, "step": 14072 }, { "epoch": 0.9, "learning_rate": 4.960017897255675e-06, "loss": 0.1404, "step": 14073 }, { "epoch": 0.9, "learning_rate": 4.95356084969405e-06, "loss": 0.1373, "step": 14074 }, { "epoch": 0.9, "learning_rate": 4.947107901081871e-06, "loss": 0.1267, "step": 14075 }, { "epoch": 0.9, "learning_rate": 4.940659051697449e-06, "loss": 0.1282, "step": 14076 }, { "epoch": 0.9, "learning_rate": 4.9342143018188736e-06, "loss": 0.1441, "step": 14077 }, { "epoch": 0.9, "learning_rate": 4.927773651724088e-06, "loss": 0.1525, "step": 14078 }, { "epoch": 0.9, "learning_rate": 4.921337101690848e-06, "loss": 0.1448, "step": 14079 }, { "epoch": 0.9, "learning_rate": 4.914904651996755e-06, "loss": 0.1192, "step": 14080 }, { "epoch": 0.9, "learning_rate": 4.908476302919196e-06, "loss": 0.1223, "step": 14081 }, { "epoch": 0.9, "learning_rate": 4.902052054735373e-06, "loss": 0.116, "step": 14082 }, { "epoch": 0.9, "learning_rate": 4.895631907722387e-06, "loss": 0.1456, "step": 14083 }, { "epoch": 0.9, "learning_rate": 4.88921586215707e-06, "loss": 0.1311, "step": 14084 }, { "epoch": 0.9, "learning_rate": 4.8828039183161365e-06, "loss": 0.1245, "step": 14085 }, { "epoch": 0.9, "learning_rate": 4.876396076476109e-06, "loss": 0.1329, "step": 14086 }, { "epoch": 0.9, "learning_rate": 4.869992336913331e-06, "loss": 0.1385, "step": 14087 }, { "epoch": 0.9, "learning_rate": 4.863592699903974e-06, "loss": 0.1345, "step": 14088 }, { "epoch": 0.9, "learning_rate": 4.8571971657240036e-06, "loss": 0.1225, "step": 14089 }, { "epoch": 0.9, "learning_rate": 4.850805734649255e-06, "loss": 0.154, "step": 14090 }, { "epoch": 0.9, "learning_rate": 4.844418406955331e-06, "loss": 0.1386, "step": 14091 }, { "epoch": 0.9, "learning_rate": 4.838035182917733e-06, "loss": 0.1387, "step": 14092 }, { "epoch": 0.9, "learning_rate": 4.831656062811718e-06, "loss": 0.1246, "step": 14093 }, { "epoch": 0.9, "learning_rate": 4.825281046912411e-06, "loss": 0.1272, "step": 14094 }, { "epoch": 0.9, "learning_rate": 4.818910135494714e-06, "loss": 0.1366, "step": 14095 }, { "epoch": 0.9, "learning_rate": 4.812543328833385e-06, "loss": 0.1309, "step": 14096 }, { "epoch": 0.9, "learning_rate": 4.806180627202994e-06, "loss": 0.1418, "step": 14097 }, { "epoch": 0.9, "learning_rate": 4.799822030877932e-06, "loss": 0.1429, "step": 14098 }, { "epoch": 0.9, "learning_rate": 4.793467540132457e-06, "loss": 0.1354, "step": 14099 }, { "epoch": 0.9, "learning_rate": 4.7871171552405615e-06, "loss": 0.14, "step": 14100 }, { "epoch": 0.9, "learning_rate": 4.780770876476148e-06, "loss": 0.1337, "step": 14101 }, { "epoch": 0.9, "learning_rate": 4.774428704112899e-06, "loss": 0.1331, "step": 14102 }, { "epoch": 0.9, "learning_rate": 4.768090638424283e-06, "loss": 0.1457, "step": 14103 }, { "epoch": 0.9, "learning_rate": 4.761756679683682e-06, "loss": 0.1363, "step": 14104 }, { "epoch": 0.9, "learning_rate": 4.7554268281642335e-06, "loss": 0.1304, "step": 14105 }, { "epoch": 0.9, "learning_rate": 4.74910108413893e-06, "loss": 0.1213, "step": 14106 }, { "epoch": 0.9, "learning_rate": 4.742779447880552e-06, "loss": 0.1464, "step": 14107 }, { "epoch": 0.9, "learning_rate": 4.7364619196617495e-06, "loss": 0.129, "step": 14108 }, { "epoch": 0.9, "learning_rate": 4.73014849975496e-06, "loss": 0.1327, "step": 14109 }, { "epoch": 0.9, "learning_rate": 4.723839188432444e-06, "loss": 0.1475, "step": 14110 }, { "epoch": 0.9, "learning_rate": 4.7175339859662936e-06, "loss": 0.1201, "step": 14111 }, { "epoch": 0.9, "learning_rate": 4.711232892628448e-06, "loss": 0.1225, "step": 14112 }, { "epoch": 0.9, "learning_rate": 4.704935908690644e-06, "loss": 0.1375, "step": 14113 }, { "epoch": 0.9, "learning_rate": 4.698643034424421e-06, "loss": 0.136, "step": 14114 }, { "epoch": 0.91, "learning_rate": 4.6923542701011955e-06, "loss": 0.1316, "step": 14115 }, { "epoch": 0.91, "learning_rate": 4.68606961599215e-06, "loss": 0.1303, "step": 14116 }, { "epoch": 0.91, "learning_rate": 4.679789072368313e-06, "loss": 0.1441, "step": 14117 }, { "epoch": 0.91, "learning_rate": 4.673512639500543e-06, "loss": 0.1218, "step": 14118 }, { "epoch": 0.91, "learning_rate": 4.667240317659527e-06, "loss": 0.1268, "step": 14119 }, { "epoch": 0.91, "learning_rate": 4.660972107115758e-06, "loss": 0.1303, "step": 14120 }, { "epoch": 0.91, "learning_rate": 4.654708008139552e-06, "loss": 0.1417, "step": 14121 }, { "epoch": 0.91, "learning_rate": 4.648448021001073e-06, "loss": 0.1363, "step": 14122 }, { "epoch": 0.91, "learning_rate": 4.642192145970259e-06, "loss": 0.1183, "step": 14123 }, { "epoch": 0.91, "learning_rate": 4.635940383316906e-06, "loss": 0.1245, "step": 14124 }, { "epoch": 0.91, "learning_rate": 4.629692733310631e-06, "loss": 0.1417, "step": 14125 }, { "epoch": 0.91, "learning_rate": 4.623449196220875e-06, "loss": 0.1388, "step": 14126 }, { "epoch": 0.91, "learning_rate": 4.6172097723168995e-06, "loss": 0.1449, "step": 14127 }, { "epoch": 0.91, "learning_rate": 4.6109744618677564e-06, "loss": 0.1453, "step": 14128 }, { "epoch": 0.91, "learning_rate": 4.604743265142386e-06, "loss": 0.1203, "step": 14129 }, { "epoch": 0.91, "learning_rate": 4.598516182409485e-06, "loss": 0.1393, "step": 14130 }, { "epoch": 0.91, "learning_rate": 4.592293213937604e-06, "loss": 0.1258, "step": 14131 }, { "epoch": 0.91, "learning_rate": 4.586074359995119e-06, "loss": 0.1311, "step": 14132 }, { "epoch": 0.91, "learning_rate": 4.579859620850213e-06, "loss": 0.1385, "step": 14133 }, { "epoch": 0.91, "learning_rate": 4.573648996770941e-06, "loss": 0.1273, "step": 14134 }, { "epoch": 0.91, "learning_rate": 4.567442488025086e-06, "loss": 0.1508, "step": 14135 }, { "epoch": 0.91, "learning_rate": 4.561240094880348e-06, "loss": 0.1375, "step": 14136 }, { "epoch": 0.91, "learning_rate": 4.5550418176041776e-06, "loss": 0.1542, "step": 14137 }, { "epoch": 0.91, "learning_rate": 4.548847656463906e-06, "loss": 0.1441, "step": 14138 }, { "epoch": 0.91, "learning_rate": 4.542657611726664e-06, "loss": 0.1336, "step": 14139 }, { "epoch": 0.91, "learning_rate": 4.5364716836593715e-06, "loss": 0.1441, "step": 14140 }, { "epoch": 0.91, "learning_rate": 4.530289872528848e-06, "loss": 0.1176, "step": 14141 }, { "epoch": 0.91, "learning_rate": 4.524112178601647e-06, "loss": 0.1289, "step": 14142 }, { "epoch": 0.91, "learning_rate": 4.517938602144189e-06, "loss": 0.1255, "step": 14143 }, { "epoch": 0.91, "learning_rate": 4.511769143422728e-06, "loss": 0.1289, "step": 14144 }, { "epoch": 0.91, "learning_rate": 4.5056038027033284e-06, "loss": 0.1348, "step": 14145 }, { "epoch": 0.91, "learning_rate": 4.499442580251878e-06, "loss": 0.126, "step": 14146 }, { "epoch": 0.91, "learning_rate": 4.493285476334075e-06, "loss": 0.1492, "step": 14147 }, { "epoch": 0.91, "learning_rate": 4.487132491215451e-06, "loss": 0.1429, "step": 14148 }, { "epoch": 0.91, "learning_rate": 4.480983625161362e-06, "loss": 0.14, "step": 14149 }, { "epoch": 0.91, "learning_rate": 4.474838878436971e-06, "loss": 0.1326, "step": 14150 }, { "epoch": 0.91, "learning_rate": 4.46869825130728e-06, "loss": 0.1416, "step": 14151 }, { "epoch": 0.91, "learning_rate": 4.462561744037108e-06, "loss": 0.1346, "step": 14152 }, { "epoch": 0.91, "learning_rate": 4.456429356891123e-06, "loss": 0.1303, "step": 14153 }, { "epoch": 0.91, "learning_rate": 4.4503010901337346e-06, "loss": 0.1253, "step": 14154 }, { "epoch": 0.91, "learning_rate": 4.444176944029288e-06, "loss": 0.1184, "step": 14155 }, { "epoch": 0.91, "learning_rate": 4.43805691884186e-06, "loss": 0.1356, "step": 14156 }, { "epoch": 0.91, "learning_rate": 4.4319410148353725e-06, "loss": 0.1374, "step": 14157 }, { "epoch": 0.91, "learning_rate": 4.425829232273582e-06, "loss": 0.1384, "step": 14158 }, { "epoch": 0.91, "learning_rate": 4.4197215714200765e-06, "loss": 0.1414, "step": 14159 }, { "epoch": 0.91, "learning_rate": 4.413618032538269e-06, "loss": 0.1375, "step": 14160 }, { "epoch": 0.91, "learning_rate": 4.407518615891337e-06, "loss": 0.1204, "step": 14161 }, { "epoch": 0.91, "learning_rate": 4.401423321742359e-06, "loss": 0.1466, "step": 14162 }, { "epoch": 0.91, "learning_rate": 4.395332150354192e-06, "loss": 0.1442, "step": 14163 }, { "epoch": 0.91, "learning_rate": 4.389245101989503e-06, "loss": 0.1285, "step": 14164 }, { "epoch": 0.91, "learning_rate": 4.383162176910804e-06, "loss": 0.1253, "step": 14165 }, { "epoch": 0.91, "learning_rate": 4.377083375380442e-06, "loss": 0.1151, "step": 14166 }, { "epoch": 0.91, "learning_rate": 4.371008697660572e-06, "loss": 0.1391, "step": 14167 }, { "epoch": 0.91, "learning_rate": 4.364938144013153e-06, "loss": 0.1488, "step": 14168 }, { "epoch": 0.91, "learning_rate": 4.358871714699997e-06, "loss": 0.1263, "step": 14169 }, { "epoch": 0.91, "learning_rate": 4.352809409982705e-06, "loss": 0.162, "step": 14170 }, { "epoch": 0.91, "learning_rate": 4.346751230122737e-06, "loss": 0.1307, "step": 14171 }, { "epoch": 0.91, "learning_rate": 4.340697175381336e-06, "loss": 0.1278, "step": 14172 }, { "epoch": 0.91, "learning_rate": 4.3346472460195966e-06, "loss": 0.1364, "step": 14173 }, { "epoch": 0.91, "learning_rate": 4.328601442298452e-06, "loss": 0.1394, "step": 14174 }, { "epoch": 0.91, "learning_rate": 4.322559764478596e-06, "loss": 0.1261, "step": 14175 }, { "epoch": 0.91, "learning_rate": 4.316522212820595e-06, "loss": 0.1393, "step": 14176 }, { "epoch": 0.91, "learning_rate": 4.310488787584832e-06, "loss": 0.1283, "step": 14177 }, { "epoch": 0.91, "learning_rate": 4.304459489031476e-06, "loss": 0.132, "step": 14178 }, { "epoch": 0.91, "learning_rate": 4.298434317420574e-06, "loss": 0.1272, "step": 14179 }, { "epoch": 0.91, "learning_rate": 4.2924132730119524e-06, "loss": 0.1524, "step": 14180 }, { "epoch": 0.91, "learning_rate": 4.286396356065281e-06, "loss": 0.1396, "step": 14181 }, { "epoch": 0.91, "learning_rate": 4.280383566840029e-06, "loss": 0.1282, "step": 14182 }, { "epoch": 0.91, "learning_rate": 4.274374905595524e-06, "loss": 0.1166, "step": 14183 }, { "epoch": 0.91, "learning_rate": 4.268370372590868e-06, "loss": 0.1278, "step": 14184 }, { "epoch": 0.91, "learning_rate": 4.262369968085034e-06, "loss": 0.1348, "step": 14185 }, { "epoch": 0.91, "learning_rate": 4.25637369233679e-06, "loss": 0.1386, "step": 14186 }, { "epoch": 0.91, "learning_rate": 4.250381545604709e-06, "loss": 0.1503, "step": 14187 }, { "epoch": 0.91, "learning_rate": 4.2443935281472395e-06, "loss": 0.128, "step": 14188 }, { "epoch": 0.91, "learning_rate": 4.238409640222596e-06, "loss": 0.1284, "step": 14189 }, { "epoch": 0.91, "learning_rate": 4.2324298820888505e-06, "loss": 0.145, "step": 14190 }, { "epoch": 0.91, "learning_rate": 4.2264542540038646e-06, "loss": 0.1326, "step": 14191 }, { "epoch": 0.91, "learning_rate": 4.2204827562253525e-06, "loss": 0.1457, "step": 14192 }, { "epoch": 0.91, "learning_rate": 4.214515389010865e-06, "loss": 0.1421, "step": 14193 }, { "epoch": 0.91, "learning_rate": 4.208552152617717e-06, "loss": 0.1209, "step": 14194 }, { "epoch": 0.91, "learning_rate": 4.202593047303094e-06, "loss": 0.1403, "step": 14195 }, { "epoch": 0.91, "learning_rate": 4.1966380733239665e-06, "loss": 0.1283, "step": 14196 }, { "epoch": 0.91, "learning_rate": 4.190687230937173e-06, "loss": 0.1337, "step": 14197 }, { "epoch": 0.91, "learning_rate": 4.184740520399333e-06, "loss": 0.1218, "step": 14198 }, { "epoch": 0.91, "learning_rate": 4.178797941966895e-06, "loss": 0.1151, "step": 14199 }, { "epoch": 0.91, "learning_rate": 4.172859495896176e-06, "loss": 0.1385, "step": 14200 }, { "epoch": 0.91, "learning_rate": 4.166925182443227e-06, "loss": 0.1339, "step": 14201 }, { "epoch": 0.91, "learning_rate": 4.160995001863999e-06, "loss": 0.1425, "step": 14202 }, { "epoch": 0.91, "learning_rate": 4.15506895441422e-06, "loss": 0.1206, "step": 14203 }, { "epoch": 0.91, "learning_rate": 4.149147040349477e-06, "loss": 0.1259, "step": 14204 }, { "epoch": 0.91, "learning_rate": 4.14322925992513e-06, "loss": 0.1291, "step": 14205 }, { "epoch": 0.91, "learning_rate": 4.137315613396409e-06, "loss": 0.1188, "step": 14206 }, { "epoch": 0.91, "learning_rate": 4.1314061010183425e-06, "loss": 0.126, "step": 14207 }, { "epoch": 0.91, "learning_rate": 4.125500723045761e-06, "loss": 0.1205, "step": 14208 }, { "epoch": 0.91, "learning_rate": 4.119599479733371e-06, "loss": 0.138, "step": 14209 }, { "epoch": 0.91, "learning_rate": 4.113702371335637e-06, "loss": 0.1408, "step": 14210 }, { "epoch": 0.91, "learning_rate": 4.107809398106898e-06, "loss": 0.1394, "step": 14211 }, { "epoch": 0.91, "learning_rate": 4.101920560301276e-06, "loss": 0.1316, "step": 14212 }, { "epoch": 0.91, "learning_rate": 4.096035858172742e-06, "loss": 0.1334, "step": 14213 }, { "epoch": 0.91, "learning_rate": 4.090155291975095e-06, "loss": 0.1333, "step": 14214 }, { "epoch": 0.91, "learning_rate": 4.0842788619618985e-06, "loss": 0.1397, "step": 14215 }, { "epoch": 0.91, "learning_rate": 4.078406568386617e-06, "loss": 0.1521, "step": 14216 }, { "epoch": 0.91, "learning_rate": 4.072538411502469e-06, "loss": 0.1494, "step": 14217 }, { "epoch": 0.91, "learning_rate": 4.066674391562553e-06, "loss": 0.1366, "step": 14218 }, { "epoch": 0.91, "learning_rate": 4.060814508819721e-06, "loss": 0.1141, "step": 14219 }, { "epoch": 0.91, "learning_rate": 4.054958763526706e-06, "loss": 0.1179, "step": 14220 }, { "epoch": 0.91, "learning_rate": 4.049107155936061e-06, "loss": 0.1322, "step": 14221 }, { "epoch": 0.91, "learning_rate": 4.043259686300094e-06, "loss": 0.1222, "step": 14222 }, { "epoch": 0.91, "learning_rate": 4.037416354871037e-06, "loss": 0.1262, "step": 14223 }, { "epoch": 0.91, "learning_rate": 4.031577161900835e-06, "loss": 0.141, "step": 14224 }, { "epoch": 0.91, "learning_rate": 4.025742107641339e-06, "loss": 0.1325, "step": 14225 }, { "epoch": 0.91, "learning_rate": 4.0199111923441946e-06, "loss": 0.1304, "step": 14226 }, { "epoch": 0.91, "learning_rate": 4.014084416260832e-06, "loss": 0.1295, "step": 14227 }, { "epoch": 0.91, "learning_rate": 4.008261779642575e-06, "loss": 0.1444, "step": 14228 }, { "epoch": 0.91, "learning_rate": 4.0024432827404975e-06, "loss": 0.1327, "step": 14229 }, { "epoch": 0.91, "learning_rate": 3.996628925805557e-06, "loss": 0.1405, "step": 14230 }, { "epoch": 0.91, "learning_rate": 3.990818709088473e-06, "loss": 0.1375, "step": 14231 }, { "epoch": 0.91, "learning_rate": 3.985012632839824e-06, "loss": 0.1214, "step": 14232 }, { "epoch": 0.91, "learning_rate": 3.97921069731001e-06, "loss": 0.1256, "step": 14233 }, { "epoch": 0.91, "learning_rate": 3.97341290274923e-06, "loss": 0.1389, "step": 14234 }, { "epoch": 0.91, "learning_rate": 3.9676192494075395e-06, "loss": 0.1328, "step": 14235 }, { "epoch": 0.91, "learning_rate": 3.961829737534772e-06, "loss": 0.1264, "step": 14236 }, { "epoch": 0.91, "learning_rate": 3.956044367380629e-06, "loss": 0.1221, "step": 14237 }, { "epoch": 0.91, "learning_rate": 3.9502631391945765e-06, "loss": 0.132, "step": 14238 }, { "epoch": 0.91, "learning_rate": 3.944486053225949e-06, "loss": 0.13, "step": 14239 }, { "epoch": 0.91, "learning_rate": 3.9387131097239126e-06, "loss": 0.1258, "step": 14240 }, { "epoch": 0.91, "learning_rate": 3.932944308937381e-06, "loss": 0.1264, "step": 14241 }, { "epoch": 0.91, "learning_rate": 3.9271796511151874e-06, "loss": 0.1225, "step": 14242 }, { "epoch": 0.91, "learning_rate": 3.921419136505899e-06, "loss": 0.1252, "step": 14243 }, { "epoch": 0.91, "learning_rate": 3.915662765357975e-06, "loss": 0.1321, "step": 14244 }, { "epoch": 0.91, "learning_rate": 3.909910537919625e-06, "loss": 0.1346, "step": 14245 }, { "epoch": 0.91, "learning_rate": 3.904162454438942e-06, "loss": 0.1227, "step": 14246 }, { "epoch": 0.91, "learning_rate": 3.8984185151638265e-06, "loss": 0.1291, "step": 14247 }, { "epoch": 0.91, "learning_rate": 3.892678720341969e-06, "loss": 0.1335, "step": 14248 }, { "epoch": 0.91, "learning_rate": 3.886943070220916e-06, "loss": 0.139, "step": 14249 }, { "epoch": 0.91, "learning_rate": 3.8812115650480045e-06, "loss": 0.1428, "step": 14250 }, { "epoch": 0.91, "learning_rate": 3.875484205070434e-06, "loss": 0.155, "step": 14251 }, { "epoch": 0.91, "learning_rate": 3.8697609905351875e-06, "loss": 0.1452, "step": 14252 }, { "epoch": 0.91, "learning_rate": 3.864041921689076e-06, "loss": 0.1397, "step": 14253 }, { "epoch": 0.91, "learning_rate": 3.858326998778761e-06, "loss": 0.1497, "step": 14254 }, { "epoch": 0.91, "learning_rate": 3.852616222050675e-06, "loss": 0.1293, "step": 14255 }, { "epoch": 0.91, "learning_rate": 3.846909591751124e-06, "loss": 0.1268, "step": 14256 }, { "epoch": 0.91, "learning_rate": 3.841207108126188e-06, "loss": 0.1413, "step": 14257 }, { "epoch": 0.91, "learning_rate": 3.8355087714218255e-06, "loss": 0.1444, "step": 14258 }, { "epoch": 0.91, "learning_rate": 3.8298145818837396e-06, "loss": 0.1265, "step": 14259 }, { "epoch": 0.91, "learning_rate": 3.824124539757512e-06, "loss": 0.1237, "step": 14260 }, { "epoch": 0.91, "learning_rate": 3.818438645288558e-06, "loss": 0.1179, "step": 14261 }, { "epoch": 0.91, "learning_rate": 3.8127568987220476e-06, "loss": 0.1426, "step": 14262 }, { "epoch": 0.91, "learning_rate": 3.8070793003030292e-06, "loss": 0.1194, "step": 14263 }, { "epoch": 0.91, "learning_rate": 3.8014058502763407e-06, "loss": 0.1255, "step": 14264 }, { "epoch": 0.91, "learning_rate": 3.7957365488866634e-06, "loss": 0.1312, "step": 14265 }, { "epoch": 0.91, "learning_rate": 3.790071396378503e-06, "loss": 0.1234, "step": 14266 }, { "epoch": 0.91, "learning_rate": 3.7844103929961515e-06, "loss": 0.1344, "step": 14267 }, { "epoch": 0.91, "learning_rate": 3.7787535389837593e-06, "loss": 0.1393, "step": 14268 }, { "epoch": 0.91, "learning_rate": 3.7731008345852635e-06, "loss": 0.145, "step": 14269 }, { "epoch": 0.91, "learning_rate": 3.76745228004447e-06, "loss": 0.1258, "step": 14270 }, { "epoch": 0.92, "learning_rate": 3.76180787560495e-06, "loss": 0.1355, "step": 14271 }, { "epoch": 0.92, "learning_rate": 3.7561676215101314e-06, "loss": 0.1357, "step": 14272 }, { "epoch": 0.92, "learning_rate": 3.750531518003264e-06, "loss": 0.1304, "step": 14273 }, { "epoch": 0.92, "learning_rate": 3.7448995653273866e-06, "loss": 0.1359, "step": 14274 }, { "epoch": 0.92, "learning_rate": 3.739271763725416e-06, "loss": 0.145, "step": 14275 }, { "epoch": 0.92, "learning_rate": 3.733648113440025e-06, "loss": 0.1354, "step": 14276 }, { "epoch": 0.92, "learning_rate": 3.7280286147137523e-06, "loss": 0.1348, "step": 14277 }, { "epoch": 0.92, "learning_rate": 3.7224132677889266e-06, "loss": 0.1319, "step": 14278 }, { "epoch": 0.92, "learning_rate": 3.716802072907732e-06, "loss": 0.1396, "step": 14279 }, { "epoch": 0.92, "learning_rate": 3.7111950303121644e-06, "loss": 0.1412, "step": 14280 }, { "epoch": 0.92, "learning_rate": 3.7055921402439963e-06, "loss": 0.1279, "step": 14281 }, { "epoch": 0.92, "learning_rate": 3.69999340294489e-06, "loss": 0.1418, "step": 14282 }, { "epoch": 0.92, "learning_rate": 3.694398818656275e-06, "loss": 0.1491, "step": 14283 }, { "epoch": 0.92, "learning_rate": 3.688808387619447e-06, "loss": 0.1264, "step": 14284 }, { "epoch": 0.92, "learning_rate": 3.683222110075457e-06, "loss": 0.1358, "step": 14285 }, { "epoch": 0.92, "learning_rate": 3.677639986265247e-06, "loss": 0.1591, "step": 14286 }, { "epoch": 0.92, "learning_rate": 3.672062016429556e-06, "loss": 0.1388, "step": 14287 }, { "epoch": 0.92, "learning_rate": 3.666488200808915e-06, "loss": 0.1398, "step": 14288 }, { "epoch": 0.92, "learning_rate": 3.6609185396437205e-06, "loss": 0.1452, "step": 14289 }, { "epoch": 0.92, "learning_rate": 3.655353033174147e-06, "loss": 0.1385, "step": 14290 }, { "epoch": 0.92, "learning_rate": 3.6497916816402358e-06, "loss": 0.137, "step": 14291 }, { "epoch": 0.92, "learning_rate": 3.6442344852817946e-06, "loss": 0.1223, "step": 14292 }, { "epoch": 0.92, "learning_rate": 3.6386814443384985e-06, "loss": 0.154, "step": 14293 }, { "epoch": 0.92, "learning_rate": 3.6331325590498344e-06, "loss": 0.1334, "step": 14294 }, { "epoch": 0.92, "learning_rate": 3.627587829655088e-06, "loss": 0.137, "step": 14295 }, { "epoch": 0.92, "learning_rate": 3.6220472563934016e-06, "loss": 0.1353, "step": 14296 }, { "epoch": 0.92, "learning_rate": 3.616510839503673e-06, "loss": 0.1471, "step": 14297 }, { "epoch": 0.92, "learning_rate": 3.610978579224722e-06, "loss": 0.1355, "step": 14298 }, { "epoch": 0.92, "learning_rate": 3.6054504757950802e-06, "loss": 0.1507, "step": 14299 }, { "epoch": 0.92, "learning_rate": 3.599926529453168e-06, "loss": 0.1444, "step": 14300 }, { "epoch": 0.92, "learning_rate": 3.5944067404372394e-06, "loss": 0.146, "step": 14301 }, { "epoch": 0.92, "learning_rate": 3.588891108985293e-06, "loss": 0.1216, "step": 14302 }, { "epoch": 0.92, "learning_rate": 3.5833796353352266e-06, "loss": 0.1337, "step": 14303 }, { "epoch": 0.92, "learning_rate": 3.5778723197247067e-06, "loss": 0.1181, "step": 14304 }, { "epoch": 0.92, "learning_rate": 3.572369162391265e-06, "loss": 0.138, "step": 14305 }, { "epoch": 0.92, "learning_rate": 3.5668701635722004e-06, "loss": 0.1539, "step": 14306 }, { "epoch": 0.92, "learning_rate": 3.5613753235046676e-06, "loss": 0.128, "step": 14307 }, { "epoch": 0.92, "learning_rate": 3.5558846424256663e-06, "loss": 0.1271, "step": 14308 }, { "epoch": 0.92, "learning_rate": 3.5503981205719407e-06, "loss": 0.1413, "step": 14309 }, { "epoch": 0.92, "learning_rate": 3.544915758180145e-06, "loss": 0.1307, "step": 14310 }, { "epoch": 0.92, "learning_rate": 3.5394375554866686e-06, "loss": 0.1413, "step": 14311 }, { "epoch": 0.92, "learning_rate": 3.533963512727789e-06, "loss": 0.1517, "step": 14312 }, { "epoch": 0.92, "learning_rate": 3.5284936301395844e-06, "loss": 0.1403, "step": 14313 }, { "epoch": 0.92, "learning_rate": 3.523027907957932e-06, "loss": 0.1295, "step": 14314 }, { "epoch": 0.92, "learning_rate": 3.5175663464185436e-06, "loss": 0.1242, "step": 14315 }, { "epoch": 0.92, "learning_rate": 3.5121089457569645e-06, "loss": 0.1382, "step": 14316 }, { "epoch": 0.92, "learning_rate": 3.5066557062085394e-06, "loss": 0.1286, "step": 14317 }, { "epoch": 0.92, "learning_rate": 3.5012066280084466e-06, "loss": 0.1446, "step": 14318 }, { "epoch": 0.92, "learning_rate": 3.4957617113916763e-06, "loss": 0.135, "step": 14319 }, { "epoch": 0.92, "learning_rate": 3.490320956593063e-06, "loss": 0.1197, "step": 14320 }, { "epoch": 0.92, "learning_rate": 3.4848843638472295e-06, "loss": 0.157, "step": 14321 }, { "epoch": 0.92, "learning_rate": 3.4794519333886334e-06, "loss": 0.1559, "step": 14322 }, { "epoch": 0.92, "learning_rate": 3.4740236654515536e-06, "loss": 0.1402, "step": 14323 }, { "epoch": 0.92, "learning_rate": 3.4685995602700915e-06, "loss": 0.1351, "step": 14324 }, { "epoch": 0.92, "learning_rate": 3.4631796180781495e-06, "loss": 0.1362, "step": 14325 }, { "epoch": 0.92, "learning_rate": 3.4577638391094845e-06, "loss": 0.1229, "step": 14326 }, { "epoch": 0.92, "learning_rate": 3.452352223597666e-06, "loss": 0.1323, "step": 14327 }, { "epoch": 0.92, "learning_rate": 3.4469447717760396e-06, "loss": 0.1374, "step": 14328 }, { "epoch": 0.92, "learning_rate": 3.4415414838778415e-06, "loss": 0.1382, "step": 14329 }, { "epoch": 0.92, "learning_rate": 3.436142360136063e-06, "loss": 0.1315, "step": 14330 }, { "epoch": 0.92, "learning_rate": 3.4307474007835737e-06, "loss": 0.1495, "step": 14331 }, { "epoch": 0.92, "learning_rate": 3.4253566060530096e-06, "loss": 0.1179, "step": 14332 }, { "epoch": 0.92, "learning_rate": 3.4199699761768624e-06, "loss": 0.1473, "step": 14333 }, { "epoch": 0.92, "learning_rate": 3.414587511387446e-06, "loss": 0.1299, "step": 14334 }, { "epoch": 0.92, "learning_rate": 3.409209211916864e-06, "loss": 0.1336, "step": 14335 }, { "epoch": 0.92, "learning_rate": 3.403835077997075e-06, "loss": 0.1284, "step": 14336 }, { "epoch": 0.92, "learning_rate": 3.3984651098598275e-06, "loss": 0.1311, "step": 14337 }, { "epoch": 0.92, "learning_rate": 3.3930993077367246e-06, "loss": 0.1144, "step": 14338 }, { "epoch": 0.92, "learning_rate": 3.3877376718591592e-06, "loss": 0.1397, "step": 14339 }, { "epoch": 0.92, "learning_rate": 3.382380202458346e-06, "loss": 0.1544, "step": 14340 }, { "epoch": 0.92, "learning_rate": 3.3770268997653564e-06, "loss": 0.1406, "step": 14341 }, { "epoch": 0.92, "learning_rate": 3.3716777640110276e-06, "loss": 0.1361, "step": 14342 }, { "epoch": 0.92, "learning_rate": 3.366332795426064e-06, "loss": 0.1579, "step": 14343 }, { "epoch": 0.92, "learning_rate": 3.3609919942409695e-06, "loss": 0.1189, "step": 14344 }, { "epoch": 0.92, "learning_rate": 3.355655360686061e-06, "loss": 0.1328, "step": 14345 }, { "epoch": 0.92, "learning_rate": 3.3503228949914865e-06, "loss": 0.1218, "step": 14346 }, { "epoch": 0.92, "learning_rate": 3.3449945973872075e-06, "loss": 0.1373, "step": 14347 }, { "epoch": 0.92, "learning_rate": 3.33967046810304e-06, "loss": 0.1376, "step": 14348 }, { "epoch": 0.92, "learning_rate": 3.3343505073685554e-06, "loss": 0.131, "step": 14349 }, { "epoch": 0.92, "learning_rate": 3.329034715413204e-06, "loss": 0.1287, "step": 14350 }, { "epoch": 0.92, "learning_rate": 3.3237230924662132e-06, "loss": 0.1323, "step": 14351 }, { "epoch": 0.92, "learning_rate": 3.318415638756667e-06, "loss": 0.1361, "step": 14352 }, { "epoch": 0.92, "learning_rate": 3.313112354513448e-06, "loss": 0.1268, "step": 14353 }, { "epoch": 0.92, "learning_rate": 3.307813239965274e-06, "loss": 0.1336, "step": 14354 }, { "epoch": 0.92, "learning_rate": 3.302518295340673e-06, "loss": 0.1203, "step": 14355 }, { "epoch": 0.92, "learning_rate": 3.2972275208679625e-06, "loss": 0.1358, "step": 14356 }, { "epoch": 0.92, "learning_rate": 3.29194091677536e-06, "loss": 0.153, "step": 14357 }, { "epoch": 0.92, "learning_rate": 3.2866584832908055e-06, "loss": 0.121, "step": 14358 }, { "epoch": 0.92, "learning_rate": 3.2813802206421275e-06, "loss": 0.1297, "step": 14359 }, { "epoch": 0.92, "learning_rate": 3.2761061290569663e-06, "loss": 0.1376, "step": 14360 }, { "epoch": 0.92, "learning_rate": 3.2708362087627732e-06, "loss": 0.1281, "step": 14361 }, { "epoch": 0.92, "learning_rate": 3.2655704599868e-06, "loss": 0.1324, "step": 14362 }, { "epoch": 0.92, "learning_rate": 3.2603088829561424e-06, "loss": 0.1279, "step": 14363 }, { "epoch": 0.92, "learning_rate": 3.255051477897708e-06, "loss": 0.1355, "step": 14364 }, { "epoch": 0.92, "learning_rate": 3.249798245038227e-06, "loss": 0.1452, "step": 14365 }, { "epoch": 0.92, "learning_rate": 3.244549184604251e-06, "loss": 0.1278, "step": 14366 }, { "epoch": 0.92, "learning_rate": 3.2393042968221543e-06, "loss": 0.1331, "step": 14367 }, { "epoch": 0.92, "learning_rate": 3.234063581918123e-06, "loss": 0.1269, "step": 14368 }, { "epoch": 0.92, "learning_rate": 3.2288270401181765e-06, "loss": 0.1499, "step": 14369 }, { "epoch": 0.92, "learning_rate": 3.2235946716481113e-06, "loss": 0.1497, "step": 14370 }, { "epoch": 0.92, "learning_rate": 3.2183664767336253e-06, "loss": 0.1357, "step": 14371 }, { "epoch": 0.92, "learning_rate": 3.213142455600138e-06, "loss": 0.1257, "step": 14372 }, { "epoch": 0.92, "learning_rate": 3.2079226084729795e-06, "loss": 0.1338, "step": 14373 }, { "epoch": 0.92, "learning_rate": 3.2027069355772264e-06, "loss": 0.1466, "step": 14374 }, { "epoch": 0.92, "learning_rate": 3.197495437137854e-06, "loss": 0.1333, "step": 14375 }, { "epoch": 0.92, "learning_rate": 3.1922881133795825e-06, "loss": 0.1186, "step": 14376 }, { "epoch": 0.92, "learning_rate": 3.187084964526965e-06, "loss": 0.1375, "step": 14377 }, { "epoch": 0.92, "learning_rate": 3.181885990804423e-06, "loss": 0.1241, "step": 14378 }, { "epoch": 0.92, "learning_rate": 3.176691192436143e-06, "loss": 0.1264, "step": 14379 }, { "epoch": 0.92, "learning_rate": 3.1715005696461687e-06, "loss": 0.1178, "step": 14380 }, { "epoch": 0.92, "learning_rate": 3.1663141226583537e-06, "loss": 0.1373, "step": 14381 }, { "epoch": 0.92, "learning_rate": 3.1611318516963527e-06, "loss": 0.1292, "step": 14382 }, { "epoch": 0.92, "learning_rate": 3.155953756983665e-06, "loss": 0.1414, "step": 14383 }, { "epoch": 0.92, "learning_rate": 3.1507798387435894e-06, "loss": 0.1367, "step": 14384 }, { "epoch": 0.92, "learning_rate": 3.1456100971992807e-06, "loss": 0.1622, "step": 14385 }, { "epoch": 0.92, "learning_rate": 3.14044453257365e-06, "loss": 0.1248, "step": 14386 }, { "epoch": 0.92, "learning_rate": 3.135283145089485e-06, "loss": 0.1369, "step": 14387 }, { "epoch": 0.92, "learning_rate": 3.1301259349693968e-06, "loss": 0.1508, "step": 14388 }, { "epoch": 0.92, "learning_rate": 3.124972902435752e-06, "loss": 0.1305, "step": 14389 }, { "epoch": 0.92, "learning_rate": 3.1198240477108176e-06, "loss": 0.129, "step": 14390 }, { "epoch": 0.92, "learning_rate": 3.1146793710166046e-06, "loss": 0.1258, "step": 14391 }, { "epoch": 0.92, "learning_rate": 3.109538872575013e-06, "loss": 0.1245, "step": 14392 }, { "epoch": 0.92, "learning_rate": 3.1044025526076994e-06, "loss": 0.138, "step": 14393 }, { "epoch": 0.92, "learning_rate": 3.099270411336208e-06, "loss": 0.1277, "step": 14394 }, { "epoch": 0.92, "learning_rate": 3.0941424489818518e-06, "loss": 0.148, "step": 14395 }, { "epoch": 0.92, "learning_rate": 3.0890186657657638e-06, "loss": 0.1159, "step": 14396 }, { "epoch": 0.92, "learning_rate": 3.0838990619089347e-06, "loss": 0.1341, "step": 14397 }, { "epoch": 0.92, "learning_rate": 3.078783637632121e-06, "loss": 0.1304, "step": 14398 }, { "epoch": 0.92, "learning_rate": 3.0736723931559464e-06, "loss": 0.1436, "step": 14399 }, { "epoch": 0.92, "learning_rate": 3.0685653287008454e-06, "loss": 0.1477, "step": 14400 }, { "epoch": 0.92, "learning_rate": 3.063462444487053e-06, "loss": 0.1418, "step": 14401 }, { "epoch": 0.92, "learning_rate": 3.058363740734649e-06, "loss": 0.1302, "step": 14402 }, { "epoch": 0.92, "learning_rate": 3.0532692176634903e-06, "loss": 0.1365, "step": 14403 }, { "epoch": 0.92, "learning_rate": 3.048178875493324e-06, "loss": 0.1235, "step": 14404 }, { "epoch": 0.92, "learning_rate": 3.0430927144436294e-06, "loss": 0.1219, "step": 14405 }, { "epoch": 0.92, "learning_rate": 3.038010734733765e-06, "loss": 0.1372, "step": 14406 }, { "epoch": 0.92, "learning_rate": 3.0329329365829105e-06, "loss": 0.141, "step": 14407 }, { "epoch": 0.92, "learning_rate": 3.027859320210058e-06, "loss": 0.1357, "step": 14408 }, { "epoch": 0.92, "learning_rate": 3.0227898858339875e-06, "loss": 0.1356, "step": 14409 }, { "epoch": 0.92, "learning_rate": 3.0177246336733135e-06, "loss": 0.1484, "step": 14410 }, { "epoch": 0.92, "learning_rate": 3.012663563946505e-06, "loss": 0.1154, "step": 14411 }, { "epoch": 0.92, "learning_rate": 3.0076066768718102e-06, "loss": 0.131, "step": 14412 }, { "epoch": 0.92, "learning_rate": 3.00255397266731e-06, "loss": 0.1333, "step": 14413 }, { "epoch": 0.92, "learning_rate": 2.997505451550919e-06, "loss": 0.1377, "step": 14414 }, { "epoch": 0.92, "learning_rate": 2.9924611137403525e-06, "loss": 0.1296, "step": 14415 }, { "epoch": 0.92, "learning_rate": 2.987420959453158e-06, "loss": 0.1387, "step": 14416 }, { "epoch": 0.92, "learning_rate": 2.9823849889066613e-06, "loss": 0.1318, "step": 14417 }, { "epoch": 0.92, "learning_rate": 2.9773532023180896e-06, "loss": 0.1292, "step": 14418 }, { "epoch": 0.92, "learning_rate": 2.972325599904413e-06, "loss": 0.1243, "step": 14419 }, { "epoch": 0.92, "learning_rate": 2.9673021818824585e-06, "loss": 0.1291, "step": 14420 }, { "epoch": 0.92, "learning_rate": 2.962282948468853e-06, "loss": 0.1433, "step": 14421 }, { "epoch": 0.92, "learning_rate": 2.9572678998800894e-06, "loss": 0.1302, "step": 14422 }, { "epoch": 0.92, "learning_rate": 2.9522570363324178e-06, "loss": 0.1356, "step": 14423 }, { "epoch": 0.92, "learning_rate": 2.947250358041931e-06, "loss": 0.1302, "step": 14424 }, { "epoch": 0.92, "learning_rate": 2.9422478652245678e-06, "loss": 0.1463, "step": 14425 }, { "epoch": 0.92, "learning_rate": 2.9372495580960446e-06, "loss": 0.1296, "step": 14426 }, { "epoch": 0.93, "learning_rate": 2.9322554368719226e-06, "loss": 0.1325, "step": 14427 }, { "epoch": 0.93, "learning_rate": 2.9272655017675732e-06, "loss": 0.1377, "step": 14428 }, { "epoch": 0.93, "learning_rate": 2.9222797529982137e-06, "loss": 0.123, "step": 14429 }, { "epoch": 0.93, "learning_rate": 2.9172981907788388e-06, "loss": 0.1481, "step": 14430 }, { "epoch": 0.93, "learning_rate": 2.9123208153242767e-06, "loss": 0.146, "step": 14431 }, { "epoch": 0.93, "learning_rate": 2.9073476268491995e-06, "loss": 0.1379, "step": 14432 }, { "epoch": 0.93, "learning_rate": 2.9023786255680586e-06, "loss": 0.1211, "step": 14433 }, { "epoch": 0.93, "learning_rate": 2.8974138116951487e-06, "loss": 0.1421, "step": 14434 }, { "epoch": 0.93, "learning_rate": 2.8924531854445993e-06, "loss": 0.1258, "step": 14435 }, { "epoch": 0.93, "learning_rate": 2.8874967470303382e-06, "loss": 0.1349, "step": 14436 }, { "epoch": 0.93, "learning_rate": 2.8825444966661063e-06, "loss": 0.1436, "step": 14437 }, { "epoch": 0.93, "learning_rate": 2.877596434565455e-06, "loss": 0.1323, "step": 14438 }, { "epoch": 0.93, "learning_rate": 2.872652560941802e-06, "loss": 0.1322, "step": 14439 }, { "epoch": 0.93, "learning_rate": 2.867712876008355e-06, "loss": 0.137, "step": 14440 }, { "epoch": 0.93, "learning_rate": 2.862777379978132e-06, "loss": 0.1377, "step": 14441 }, { "epoch": 0.93, "learning_rate": 2.8578460730639744e-06, "loss": 0.1153, "step": 14442 }, { "epoch": 0.93, "learning_rate": 2.8529189554785673e-06, "loss": 0.1343, "step": 14443 }, { "epoch": 0.93, "learning_rate": 2.8479960274343744e-06, "loss": 0.1335, "step": 14444 }, { "epoch": 0.93, "learning_rate": 2.843077289143714e-06, "loss": 0.1397, "step": 14445 }, { "epoch": 0.93, "learning_rate": 2.838162740818695e-06, "loss": 0.1435, "step": 14446 }, { "epoch": 0.93, "learning_rate": 2.8332523826712807e-06, "loss": 0.1436, "step": 14447 }, { "epoch": 0.93, "learning_rate": 2.8283462149132356e-06, "loss": 0.1394, "step": 14448 }, { "epoch": 0.93, "learning_rate": 2.8234442377561232e-06, "loss": 0.1272, "step": 14449 }, { "epoch": 0.93, "learning_rate": 2.8185464514113635e-06, "loss": 0.1202, "step": 14450 }, { "epoch": 0.93, "learning_rate": 2.8136528560901653e-06, "loss": 0.1324, "step": 14451 }, { "epoch": 0.93, "learning_rate": 2.808763452003571e-06, "loss": 0.1256, "step": 14452 }, { "epoch": 0.93, "learning_rate": 2.8038782393624343e-06, "loss": 0.1198, "step": 14453 }, { "epoch": 0.93, "learning_rate": 2.7989972183774303e-06, "loss": 0.1285, "step": 14454 }, { "epoch": 0.93, "learning_rate": 2.7941203892590915e-06, "loss": 0.1457, "step": 14455 }, { "epoch": 0.93, "learning_rate": 2.789247752217705e-06, "loss": 0.1292, "step": 14456 }, { "epoch": 0.93, "learning_rate": 2.784379307463392e-06, "loss": 0.1398, "step": 14457 }, { "epoch": 0.93, "learning_rate": 2.77951505520615e-06, "loss": 0.1456, "step": 14458 }, { "epoch": 0.93, "learning_rate": 2.774654995655712e-06, "loss": 0.1394, "step": 14459 }, { "epoch": 0.93, "learning_rate": 2.7697991290216885e-06, "loss": 0.1457, "step": 14460 }, { "epoch": 0.93, "learning_rate": 2.7649474555135e-06, "loss": 0.1222, "step": 14461 }, { "epoch": 0.93, "learning_rate": 2.7600999753403913e-06, "loss": 0.1299, "step": 14462 }, { "epoch": 0.93, "learning_rate": 2.755256688711383e-06, "loss": 0.1266, "step": 14463 }, { "epoch": 0.93, "learning_rate": 2.7504175958353527e-06, "loss": 0.1364, "step": 14464 }, { "epoch": 0.93, "learning_rate": 2.745582696921001e-06, "loss": 0.1267, "step": 14465 }, { "epoch": 0.93, "learning_rate": 2.7407519921768266e-06, "loss": 0.1334, "step": 14466 }, { "epoch": 0.93, "learning_rate": 2.7359254818111526e-06, "loss": 0.1499, "step": 14467 }, { "epoch": 0.93, "learning_rate": 2.731103166032145e-06, "loss": 0.1229, "step": 14468 }, { "epoch": 0.93, "learning_rate": 2.7262850450477607e-06, "loss": 0.1437, "step": 14469 }, { "epoch": 0.93, "learning_rate": 2.7214711190657883e-06, "loss": 0.1315, "step": 14470 }, { "epoch": 0.93, "learning_rate": 2.7166613882938174e-06, "loss": 0.1509, "step": 14471 }, { "epoch": 0.93, "learning_rate": 2.7118558529392937e-06, "loss": 0.1242, "step": 14472 }, { "epoch": 0.93, "learning_rate": 2.7070545132094284e-06, "loss": 0.131, "step": 14473 }, { "epoch": 0.93, "learning_rate": 2.702257369311312e-06, "loss": 0.123, "step": 14474 }, { "epoch": 0.93, "learning_rate": 2.697464421451812e-06, "loss": 0.1237, "step": 14475 }, { "epoch": 0.93, "learning_rate": 2.69267566983763e-06, "loss": 0.1318, "step": 14476 }, { "epoch": 0.93, "learning_rate": 2.6878911146753007e-06, "loss": 0.1474, "step": 14477 }, { "epoch": 0.93, "learning_rate": 2.6831107561711255e-06, "loss": 0.1325, "step": 14478 }, { "epoch": 0.93, "learning_rate": 2.6783345945312952e-06, "loss": 0.1283, "step": 14479 }, { "epoch": 0.93, "learning_rate": 2.6735626299617457e-06, "loss": 0.1457, "step": 14480 }, { "epoch": 0.93, "learning_rate": 2.6687948626683225e-06, "loss": 0.153, "step": 14481 }, { "epoch": 0.93, "learning_rate": 2.664031292856606e-06, "loss": 0.1583, "step": 14482 }, { "epoch": 0.93, "learning_rate": 2.6592719207320426e-06, "loss": 0.1186, "step": 14483 }, { "epoch": 0.93, "learning_rate": 2.65451674649988e-06, "loss": 0.1364, "step": 14484 }, { "epoch": 0.93, "learning_rate": 2.6497657703651755e-06, "loss": 0.1425, "step": 14485 }, { "epoch": 0.93, "learning_rate": 2.6450189925328218e-06, "loss": 0.1509, "step": 14486 }, { "epoch": 0.93, "learning_rate": 2.6402764132075432e-06, "loss": 0.1314, "step": 14487 }, { "epoch": 0.93, "learning_rate": 2.6355380325938763e-06, "loss": 0.1364, "step": 14488 }, { "epoch": 0.93, "learning_rate": 2.630803850896135e-06, "loss": 0.1234, "step": 14489 }, { "epoch": 0.93, "learning_rate": 2.6260738683185016e-06, "loss": 0.1401, "step": 14490 }, { "epoch": 0.93, "learning_rate": 2.6213480850649675e-06, "loss": 0.1366, "step": 14491 }, { "epoch": 0.93, "learning_rate": 2.616626501339314e-06, "loss": 0.1418, "step": 14492 }, { "epoch": 0.93, "learning_rate": 2.6119091173451683e-06, "loss": 0.1295, "step": 14493 }, { "epoch": 0.93, "learning_rate": 2.6071959332859886e-06, "loss": 0.1342, "step": 14494 }, { "epoch": 0.93, "learning_rate": 2.602486949365024e-06, "loss": 0.1319, "step": 14495 }, { "epoch": 0.93, "learning_rate": 2.597782165785345e-06, "loss": 0.1204, "step": 14496 }, { "epoch": 0.93, "learning_rate": 2.5930815827498678e-06, "loss": 0.1536, "step": 14497 }, { "epoch": 0.93, "learning_rate": 2.5883852004613074e-06, "loss": 0.1331, "step": 14498 }, { "epoch": 0.93, "learning_rate": 2.5836930191221685e-06, "loss": 0.117, "step": 14499 }, { "epoch": 0.93, "learning_rate": 2.579005038934823e-06, "loss": 0.1345, "step": 14500 }, { "epoch": 0.93, "learning_rate": 2.574321260101453e-06, "loss": 0.1259, "step": 14501 }, { "epoch": 0.93, "learning_rate": 2.569641682824053e-06, "loss": 0.145, "step": 14502 }, { "epoch": 0.93, "learning_rate": 2.564966307304417e-06, "loss": 0.1355, "step": 14503 }, { "epoch": 0.93, "learning_rate": 2.560295133744184e-06, "loss": 0.1294, "step": 14504 }, { "epoch": 0.93, "learning_rate": 2.555628162344792e-06, "loss": 0.135, "step": 14505 }, { "epoch": 0.93, "learning_rate": 2.550965393307514e-06, "loss": 0.1442, "step": 14506 }, { "epoch": 0.93, "learning_rate": 2.5463068268334335e-06, "loss": 0.1245, "step": 14507 }, { "epoch": 0.93, "learning_rate": 2.541652463123456e-06, "loss": 0.1182, "step": 14508 }, { "epoch": 0.93, "learning_rate": 2.537002302378322e-06, "loss": 0.1412, "step": 14509 }, { "epoch": 0.93, "learning_rate": 2.5323563447985366e-06, "loss": 0.136, "step": 14510 }, { "epoch": 0.93, "learning_rate": 2.5277145905844957e-06, "loss": 0.1384, "step": 14511 }, { "epoch": 0.93, "learning_rate": 2.5230770399363613e-06, "loss": 0.1459, "step": 14512 }, { "epoch": 0.93, "learning_rate": 2.5184436930541177e-06, "loss": 0.1598, "step": 14513 }, { "epoch": 0.93, "learning_rate": 2.513814550137605e-06, "loss": 0.1245, "step": 14514 }, { "epoch": 0.93, "learning_rate": 2.509189611386442e-06, "loss": 0.1425, "step": 14515 }, { "epoch": 0.93, "learning_rate": 2.504568877000113e-06, "loss": 0.1296, "step": 14516 }, { "epoch": 0.93, "learning_rate": 2.4999523471778475e-06, "loss": 0.1342, "step": 14517 }, { "epoch": 0.93, "learning_rate": 2.495340022118764e-06, "loss": 0.1372, "step": 14518 }, { "epoch": 0.93, "learning_rate": 2.490731902021781e-06, "loss": 0.1172, "step": 14519 }, { "epoch": 0.93, "learning_rate": 2.486127987085585e-06, "loss": 0.1297, "step": 14520 }, { "epoch": 0.93, "learning_rate": 2.48152827750876e-06, "loss": 0.1273, "step": 14521 }, { "epoch": 0.93, "learning_rate": 2.4769327734896707e-06, "loss": 0.1276, "step": 14522 }, { "epoch": 0.93, "learning_rate": 2.4723414752264918e-06, "loss": 0.1386, "step": 14523 }, { "epoch": 0.93, "learning_rate": 2.46775438291722e-06, "loss": 0.1405, "step": 14524 }, { "epoch": 0.93, "learning_rate": 2.4631714967596974e-06, "loss": 0.1378, "step": 14525 }, { "epoch": 0.93, "learning_rate": 2.458592816951544e-06, "loss": 0.1256, "step": 14526 }, { "epoch": 0.93, "learning_rate": 2.4540183436902233e-06, "loss": 0.1361, "step": 14527 }, { "epoch": 0.93, "learning_rate": 2.449448077173033e-06, "loss": 0.1355, "step": 14528 }, { "epoch": 0.93, "learning_rate": 2.4448820175970388e-06, "loss": 0.1368, "step": 14529 }, { "epoch": 0.93, "learning_rate": 2.440320165159171e-06, "loss": 0.1361, "step": 14530 }, { "epoch": 0.93, "learning_rate": 2.4357625200561617e-06, "loss": 0.118, "step": 14531 }, { "epoch": 0.93, "learning_rate": 2.4312090824845755e-06, "loss": 0.1358, "step": 14532 }, { "epoch": 0.93, "learning_rate": 2.426659852640756e-06, "loss": 0.1236, "step": 14533 }, { "epoch": 0.93, "learning_rate": 2.4221148307209117e-06, "loss": 0.1458, "step": 14534 }, { "epoch": 0.93, "learning_rate": 2.4175740169210426e-06, "loss": 0.1313, "step": 14535 }, { "epoch": 0.93, "learning_rate": 2.4130374114369804e-06, "loss": 0.1505, "step": 14536 }, { "epoch": 0.93, "learning_rate": 2.4085050144643685e-06, "loss": 0.1327, "step": 14537 }, { "epoch": 0.93, "learning_rate": 2.4039768261986616e-06, "loss": 0.1262, "step": 14538 }, { "epoch": 0.93, "learning_rate": 2.3994528468351486e-06, "loss": 0.1391, "step": 14539 }, { "epoch": 0.93, "learning_rate": 2.394933076568928e-06, "loss": 0.1432, "step": 14540 }, { "epoch": 0.93, "learning_rate": 2.3904175155949116e-06, "loss": 0.1291, "step": 14541 }, { "epoch": 0.93, "learning_rate": 2.3859061641078538e-06, "loss": 0.1442, "step": 14542 }, { "epoch": 0.93, "learning_rate": 2.3813990223022885e-06, "loss": 0.139, "step": 14543 }, { "epoch": 0.93, "learning_rate": 2.376896090372616e-06, "loss": 0.1402, "step": 14544 }, { "epoch": 0.93, "learning_rate": 2.372397368513013e-06, "loss": 0.1423, "step": 14545 }, { "epoch": 0.93, "learning_rate": 2.367902856917481e-06, "loss": 0.1394, "step": 14546 }, { "epoch": 0.93, "learning_rate": 2.3634125557798535e-06, "loss": 0.1325, "step": 14547 }, { "epoch": 0.93, "learning_rate": 2.358926465293787e-06, "loss": 0.1317, "step": 14548 }, { "epoch": 0.93, "learning_rate": 2.3544445856527487e-06, "loss": 0.1248, "step": 14549 }, { "epoch": 0.93, "learning_rate": 2.349966917050017e-06, "loss": 0.1498, "step": 14550 }, { "epoch": 0.93, "learning_rate": 2.3454934596786935e-06, "loss": 0.1304, "step": 14551 }, { "epoch": 0.93, "learning_rate": 2.341024213731713e-06, "loss": 0.135, "step": 14552 }, { "epoch": 0.93, "learning_rate": 2.336559179401787e-06, "loss": 0.1442, "step": 14553 }, { "epoch": 0.93, "learning_rate": 2.332098356881496e-06, "loss": 0.1377, "step": 14554 }, { "epoch": 0.93, "learning_rate": 2.3276417463632073e-06, "loss": 0.1345, "step": 14555 }, { "epoch": 0.93, "learning_rate": 2.323189348039134e-06, "loss": 0.1515, "step": 14556 }, { "epoch": 0.93, "learning_rate": 2.318741162101257e-06, "loss": 0.1485, "step": 14557 }, { "epoch": 0.93, "learning_rate": 2.3142971887414435e-06, "loss": 0.1485, "step": 14558 }, { "epoch": 0.93, "learning_rate": 2.3098574281513185e-06, "loss": 0.1365, "step": 14559 }, { "epoch": 0.93, "learning_rate": 2.3054218805223404e-06, "loss": 0.1314, "step": 14560 }, { "epoch": 0.93, "learning_rate": 2.3009905460458227e-06, "loss": 0.1292, "step": 14561 }, { "epoch": 0.93, "learning_rate": 2.2965634249128564e-06, "loss": 0.1446, "step": 14562 }, { "epoch": 0.93, "learning_rate": 2.2921405173143673e-06, "loss": 0.1446, "step": 14563 }, { "epoch": 0.93, "learning_rate": 2.2877218234410913e-06, "loss": 0.1433, "step": 14564 }, { "epoch": 0.93, "learning_rate": 2.2833073434836095e-06, "loss": 0.1215, "step": 14565 }, { "epoch": 0.93, "learning_rate": 2.2788970776322807e-06, "loss": 0.134, "step": 14566 }, { "epoch": 0.93, "learning_rate": 2.2744910260772747e-06, "loss": 0.1353, "step": 14567 }, { "epoch": 0.93, "learning_rate": 2.270089189008662e-06, "loss": 0.1394, "step": 14568 }, { "epoch": 0.93, "learning_rate": 2.265691566616235e-06, "loss": 0.1193, "step": 14569 }, { "epoch": 0.93, "learning_rate": 2.2612981590896644e-06, "loss": 0.1447, "step": 14570 }, { "epoch": 0.93, "learning_rate": 2.2569089666183985e-06, "loss": 0.126, "step": 14571 }, { "epoch": 0.93, "learning_rate": 2.252523989391753e-06, "loss": 0.1525, "step": 14572 }, { "epoch": 0.93, "learning_rate": 2.248143227598809e-06, "loss": 0.1286, "step": 14573 }, { "epoch": 0.93, "learning_rate": 2.2437666814284942e-06, "loss": 0.1167, "step": 14574 }, { "epoch": 0.93, "learning_rate": 2.239394351069568e-06, "loss": 0.1248, "step": 14575 }, { "epoch": 0.93, "learning_rate": 2.235026236710569e-06, "loss": 0.1412, "step": 14576 }, { "epoch": 0.93, "learning_rate": 2.2306623385398904e-06, "loss": 0.1502, "step": 14577 }, { "epoch": 0.93, "learning_rate": 2.2263026567457156e-06, "loss": 0.1213, "step": 14578 }, { "epoch": 0.93, "learning_rate": 2.221947191516072e-06, "loss": 0.1449, "step": 14579 }, { "epoch": 0.93, "learning_rate": 2.217595943038786e-06, "loss": 0.138, "step": 14580 }, { "epoch": 0.93, "learning_rate": 2.2132489115014976e-06, "loss": 0.1449, "step": 14581 }, { "epoch": 0.93, "learning_rate": 2.2089060970917008e-06, "loss": 0.1464, "step": 14582 }, { "epoch": 0.94, "learning_rate": 2.204567499996668e-06, "loss": 0.1348, "step": 14583 }, { "epoch": 0.94, "learning_rate": 2.2002331204035053e-06, "loss": 0.1438, "step": 14584 }, { "epoch": 0.94, "learning_rate": 2.1959029584991298e-06, "loss": 0.116, "step": 14585 }, { "epoch": 0.94, "learning_rate": 2.191577014470292e-06, "loss": 0.1344, "step": 14586 }, { "epoch": 0.94, "learning_rate": 2.1872552885035537e-06, "loss": 0.1303, "step": 14587 }, { "epoch": 0.94, "learning_rate": 2.1829377807852768e-06, "loss": 0.136, "step": 14588 }, { "epoch": 0.94, "learning_rate": 2.1786244915016796e-06, "loss": 0.1381, "step": 14589 }, { "epoch": 0.94, "learning_rate": 2.174315420838757e-06, "loss": 0.135, "step": 14590 }, { "epoch": 0.94, "learning_rate": 2.1700105689823615e-06, "loss": 0.1442, "step": 14591 }, { "epoch": 0.94, "learning_rate": 2.1657099361181098e-06, "loss": 0.1335, "step": 14592 }, { "epoch": 0.94, "learning_rate": 2.1614135224315102e-06, "loss": 0.1296, "step": 14593 }, { "epoch": 0.94, "learning_rate": 2.157121328107814e-06, "loss": 0.1438, "step": 14594 }, { "epoch": 0.94, "learning_rate": 2.15283335333214e-06, "loss": 0.1365, "step": 14595 }, { "epoch": 0.94, "learning_rate": 2.1485495982894176e-06, "loss": 0.128, "step": 14596 }, { "epoch": 0.94, "learning_rate": 2.1442700631643776e-06, "loss": 0.1242, "step": 14597 }, { "epoch": 0.94, "learning_rate": 2.1399947481415827e-06, "loss": 0.1422, "step": 14598 }, { "epoch": 0.94, "learning_rate": 2.135723653405397e-06, "loss": 0.1315, "step": 14599 }, { "epoch": 0.94, "learning_rate": 2.13145677914004e-06, "loss": 0.1355, "step": 14600 }, { "epoch": 0.94, "learning_rate": 2.1271941255294967e-06, "loss": 0.133, "step": 14601 }, { "epoch": 0.94, "learning_rate": 2.1229356927576103e-06, "loss": 0.1311, "step": 14602 }, { "epoch": 0.94, "learning_rate": 2.1186814810080335e-06, "loss": 0.1353, "step": 14603 }, { "epoch": 0.94, "learning_rate": 2.1144314904642195e-06, "loss": 0.1331, "step": 14604 }, { "epoch": 0.94, "learning_rate": 2.1101857213094655e-06, "loss": 0.1278, "step": 14605 }, { "epoch": 0.94, "learning_rate": 2.1059441737268594e-06, "loss": 0.1264, "step": 14606 }, { "epoch": 0.94, "learning_rate": 2.101706847899332e-06, "loss": 0.1379, "step": 14607 }, { "epoch": 0.94, "learning_rate": 2.097473744009615e-06, "loss": 0.1448, "step": 14608 }, { "epoch": 0.94, "learning_rate": 2.093244862240273e-06, "loss": 0.1422, "step": 14609 }, { "epoch": 0.94, "learning_rate": 2.0890202027736727e-06, "loss": 0.135, "step": 14610 }, { "epoch": 0.94, "learning_rate": 2.0847997657920116e-06, "loss": 0.1372, "step": 14611 }, { "epoch": 0.94, "learning_rate": 2.080583551477289e-06, "loss": 0.1317, "step": 14612 }, { "epoch": 0.94, "learning_rate": 2.0763715600113366e-06, "loss": 0.146, "step": 14613 }, { "epoch": 0.94, "learning_rate": 2.0721637915757985e-06, "loss": 0.141, "step": 14614 }, { "epoch": 0.94, "learning_rate": 2.0679602463521518e-06, "loss": 0.1312, "step": 14615 }, { "epoch": 0.94, "learning_rate": 2.0637609245216628e-06, "loss": 0.1593, "step": 14616 }, { "epoch": 0.94, "learning_rate": 2.0595658262654305e-06, "loss": 0.1325, "step": 14617 }, { "epoch": 0.94, "learning_rate": 2.055374951764377e-06, "loss": 0.1409, "step": 14618 }, { "epoch": 0.94, "learning_rate": 2.0511883011992363e-06, "loss": 0.121, "step": 14619 }, { "epoch": 0.94, "learning_rate": 2.0470058747505516e-06, "loss": 0.1351, "step": 14620 }, { "epoch": 0.94, "learning_rate": 2.0428276725987017e-06, "loss": 0.1374, "step": 14621 }, { "epoch": 0.94, "learning_rate": 2.0386536949238864e-06, "loss": 0.1215, "step": 14622 }, { "epoch": 0.94, "learning_rate": 2.0344839419060845e-06, "loss": 0.1279, "step": 14623 }, { "epoch": 0.94, "learning_rate": 2.0303184137251517e-06, "loss": 0.1462, "step": 14624 }, { "epoch": 0.94, "learning_rate": 2.0261571105607e-06, "loss": 0.1259, "step": 14625 }, { "epoch": 0.94, "learning_rate": 2.0220000325922086e-06, "loss": 0.1436, "step": 14626 }, { "epoch": 0.94, "learning_rate": 2.017847179998944e-06, "loss": 0.1231, "step": 14627 }, { "epoch": 0.94, "learning_rate": 2.013698552959997e-06, "loss": 0.1326, "step": 14628 }, { "epoch": 0.94, "learning_rate": 2.0095541516543025e-06, "loss": 0.1331, "step": 14629 }, { "epoch": 0.94, "learning_rate": 2.0054139762605615e-06, "loss": 0.1283, "step": 14630 }, { "epoch": 0.94, "learning_rate": 2.001278026957354e-06, "loss": 0.1313, "step": 14631 }, { "epoch": 0.94, "learning_rate": 1.997146303923014e-06, "loss": 0.1292, "step": 14632 }, { "epoch": 0.94, "learning_rate": 1.9930188073357448e-06, "loss": 0.1309, "step": 14633 }, { "epoch": 0.94, "learning_rate": 1.9888955373735474e-06, "loss": 0.122, "step": 14634 }, { "epoch": 0.94, "learning_rate": 1.9847764942142245e-06, "loss": 0.1391, "step": 14635 }, { "epoch": 0.94, "learning_rate": 1.9806616780354336e-06, "loss": 0.1265, "step": 14636 }, { "epoch": 0.94, "learning_rate": 1.976551089014622e-06, "loss": 0.1193, "step": 14637 }, { "epoch": 0.94, "learning_rate": 1.9724447273290593e-06, "loss": 0.1387, "step": 14638 }, { "epoch": 0.94, "learning_rate": 1.9683425931558364e-06, "loss": 0.1356, "step": 14639 }, { "epoch": 0.94, "learning_rate": 1.9642446866718567e-06, "loss": 0.1284, "step": 14640 }, { "epoch": 0.94, "learning_rate": 1.9601510080538565e-06, "loss": 0.1467, "step": 14641 }, { "epoch": 0.94, "learning_rate": 1.9560615574783614e-06, "loss": 0.1297, "step": 14642 }, { "epoch": 0.94, "learning_rate": 1.9519763351217635e-06, "loss": 0.12, "step": 14643 }, { "epoch": 0.94, "learning_rate": 1.9478953411601997e-06, "loss": 0.1329, "step": 14644 }, { "epoch": 0.94, "learning_rate": 1.943818575769696e-06, "loss": 0.1243, "step": 14645 }, { "epoch": 0.94, "learning_rate": 1.939746039126056e-06, "loss": 0.1282, "step": 14646 }, { "epoch": 0.94, "learning_rate": 1.9356777314049166e-06, "loss": 0.139, "step": 14647 }, { "epoch": 0.94, "learning_rate": 1.9316136527817164e-06, "loss": 0.1338, "step": 14648 }, { "epoch": 0.94, "learning_rate": 1.9275538034317253e-06, "loss": 0.152, "step": 14649 }, { "epoch": 0.94, "learning_rate": 1.923498183530037e-06, "loss": 0.1173, "step": 14650 }, { "epoch": 0.94, "learning_rate": 1.9194467932515336e-06, "loss": 0.1383, "step": 14651 }, { "epoch": 0.94, "learning_rate": 1.9153996327709536e-06, "loss": 0.1228, "step": 14652 }, { "epoch": 0.94, "learning_rate": 1.911356702262823e-06, "loss": 0.1306, "step": 14653 }, { "epoch": 0.94, "learning_rate": 1.9073180019015036e-06, "loss": 0.1349, "step": 14654 }, { "epoch": 0.94, "learning_rate": 1.9032835318611665e-06, "loss": 0.135, "step": 14655 }, { "epoch": 0.94, "learning_rate": 1.8992532923157836e-06, "loss": 0.1231, "step": 14656 }, { "epoch": 0.94, "learning_rate": 1.8952272834391938e-06, "loss": 0.1219, "step": 14657 }, { "epoch": 0.94, "learning_rate": 1.8912055054049916e-06, "loss": 0.1385, "step": 14658 }, { "epoch": 0.94, "learning_rate": 1.8871879583866492e-06, "loss": 0.1265, "step": 14659 }, { "epoch": 0.94, "learning_rate": 1.8831746425573948e-06, "loss": 0.1271, "step": 14660 }, { "epoch": 0.94, "learning_rate": 1.8791655580903233e-06, "loss": 0.1284, "step": 14661 }, { "epoch": 0.94, "learning_rate": 1.8751607051583297e-06, "loss": 0.1324, "step": 14662 }, { "epoch": 0.94, "learning_rate": 1.8711600839341203e-06, "loss": 0.1266, "step": 14663 }, { "epoch": 0.94, "learning_rate": 1.8671636945902237e-06, "loss": 0.1171, "step": 14664 }, { "epoch": 0.94, "learning_rate": 1.8631715372990022e-06, "loss": 0.1225, "step": 14665 }, { "epoch": 0.94, "learning_rate": 1.8591836122326067e-06, "loss": 0.142, "step": 14666 }, { "epoch": 0.94, "learning_rate": 1.855199919563011e-06, "loss": 0.122, "step": 14667 }, { "epoch": 0.94, "learning_rate": 1.8512204594620331e-06, "loss": 0.1212, "step": 14668 }, { "epoch": 0.94, "learning_rate": 1.8472452321012912e-06, "loss": 0.1284, "step": 14669 }, { "epoch": 0.94, "learning_rate": 1.843274237652204e-06, "loss": 0.1349, "step": 14670 }, { "epoch": 0.94, "learning_rate": 1.839307476286034e-06, "loss": 0.1422, "step": 14671 }, { "epoch": 0.94, "learning_rate": 1.8353449481738449e-06, "loss": 0.1408, "step": 14672 }, { "epoch": 0.94, "learning_rate": 1.831386653486533e-06, "loss": 0.1258, "step": 14673 }, { "epoch": 0.94, "learning_rate": 1.8274325923947954e-06, "loss": 0.129, "step": 14674 }, { "epoch": 0.94, "learning_rate": 1.823482765069151e-06, "loss": 0.1254, "step": 14675 }, { "epoch": 0.94, "learning_rate": 1.8195371716799525e-06, "loss": 0.1211, "step": 14676 }, { "epoch": 0.94, "learning_rate": 1.8155958123973305e-06, "loss": 0.1415, "step": 14677 }, { "epoch": 0.94, "learning_rate": 1.8116586873913044e-06, "loss": 0.133, "step": 14678 }, { "epoch": 0.94, "learning_rate": 1.8077257968316164e-06, "loss": 0.1364, "step": 14679 }, { "epoch": 0.94, "learning_rate": 1.8037971408879084e-06, "loss": 0.1531, "step": 14680 }, { "epoch": 0.94, "learning_rate": 1.7998727197295784e-06, "loss": 0.1404, "step": 14681 }, { "epoch": 0.94, "learning_rate": 1.795952533525902e-06, "loss": 0.1266, "step": 14682 }, { "epoch": 0.94, "learning_rate": 1.7920365824459219e-06, "loss": 0.139, "step": 14683 }, { "epoch": 0.94, "learning_rate": 1.7881248666585138e-06, "loss": 0.1222, "step": 14684 }, { "epoch": 0.94, "learning_rate": 1.7842173863323873e-06, "loss": 0.1383, "step": 14685 }, { "epoch": 0.94, "learning_rate": 1.7803141416360414e-06, "loss": 0.1261, "step": 14686 }, { "epoch": 0.94, "learning_rate": 1.7764151327378187e-06, "loss": 0.1192, "step": 14687 }, { "epoch": 0.94, "learning_rate": 1.7725203598058515e-06, "loss": 0.1292, "step": 14688 }, { "epoch": 0.94, "learning_rate": 1.7686298230081167e-06, "loss": 0.1373, "step": 14689 }, { "epoch": 0.94, "learning_rate": 1.764743522512391e-06, "loss": 0.1508, "step": 14690 }, { "epoch": 0.94, "learning_rate": 1.760861458486285e-06, "loss": 0.1239, "step": 14691 }, { "epoch": 0.94, "learning_rate": 1.7569836310971978e-06, "loss": 0.1445, "step": 14692 }, { "epoch": 0.94, "learning_rate": 1.7531100405123846e-06, "loss": 0.125, "step": 14693 }, { "epoch": 0.94, "learning_rate": 1.7492406868988675e-06, "loss": 0.1171, "step": 14694 }, { "epoch": 0.94, "learning_rate": 1.7453755704235463e-06, "loss": 0.1354, "step": 14695 }, { "epoch": 0.94, "learning_rate": 1.741514691253099e-06, "loss": 0.1296, "step": 14696 }, { "epoch": 0.94, "learning_rate": 1.737658049554014e-06, "loss": 0.1205, "step": 14697 }, { "epoch": 0.94, "learning_rate": 1.7338056454926255e-06, "loss": 0.1337, "step": 14698 }, { "epoch": 0.94, "learning_rate": 1.729957479235067e-06, "loss": 0.1264, "step": 14699 }, { "epoch": 0.94, "learning_rate": 1.7261135509472949e-06, "loss": 0.1289, "step": 14700 }, { "epoch": 0.94, "learning_rate": 1.7222738607950762e-06, "loss": 0.1426, "step": 14701 }, { "epoch": 0.94, "learning_rate": 1.718438408944012e-06, "loss": 0.1318, "step": 14702 }, { "epoch": 0.94, "learning_rate": 1.7146071955595034e-06, "loss": 0.1392, "step": 14703 }, { "epoch": 0.94, "learning_rate": 1.7107802208067735e-06, "loss": 0.1522, "step": 14704 }, { "epoch": 0.94, "learning_rate": 1.7069574848508574e-06, "loss": 0.149, "step": 14705 }, { "epoch": 0.94, "learning_rate": 1.7031389878566228e-06, "loss": 0.1382, "step": 14706 }, { "epoch": 0.94, "learning_rate": 1.6993247299887382e-06, "loss": 0.1258, "step": 14707 }, { "epoch": 0.94, "learning_rate": 1.6955147114116942e-06, "loss": 0.1241, "step": 14708 }, { "epoch": 0.94, "learning_rate": 1.691708932289815e-06, "loss": 0.1656, "step": 14709 }, { "epoch": 0.94, "learning_rate": 1.6879073927872135e-06, "loss": 0.1466, "step": 14710 }, { "epoch": 0.94, "learning_rate": 1.6841100930678477e-06, "loss": 0.1342, "step": 14711 }, { "epoch": 0.94, "learning_rate": 1.6803170332954642e-06, "loss": 0.1417, "step": 14712 }, { "epoch": 0.94, "learning_rate": 1.6765282136336547e-06, "loss": 0.1286, "step": 14713 }, { "epoch": 0.94, "learning_rate": 1.6727436342457992e-06, "loss": 0.1174, "step": 14714 }, { "epoch": 0.94, "learning_rate": 1.6689632952951228e-06, "loss": 0.1406, "step": 14715 }, { "epoch": 0.94, "learning_rate": 1.665187196944662e-06, "loss": 0.1286, "step": 14716 }, { "epoch": 0.94, "learning_rate": 1.6614153393572418e-06, "loss": 0.1215, "step": 14717 }, { "epoch": 0.94, "learning_rate": 1.6576477226955544e-06, "loss": 0.14, "step": 14718 }, { "epoch": 0.94, "learning_rate": 1.6538843471220477e-06, "loss": 0.1554, "step": 14719 }, { "epoch": 0.94, "learning_rate": 1.6501252127990473e-06, "loss": 0.1379, "step": 14720 }, { "epoch": 0.94, "learning_rate": 1.646370319888657e-06, "loss": 0.1345, "step": 14721 }, { "epoch": 0.94, "learning_rate": 1.6426196685528029e-06, "loss": 0.1315, "step": 14722 }, { "epoch": 0.94, "learning_rate": 1.6388732589532662e-06, "loss": 0.1378, "step": 14723 }, { "epoch": 0.94, "learning_rate": 1.635131091251574e-06, "loss": 0.116, "step": 14724 }, { "epoch": 0.94, "learning_rate": 1.6313931656091296e-06, "loss": 0.136, "step": 14725 }, { "epoch": 0.94, "learning_rate": 1.6276594821871382e-06, "loss": 0.1212, "step": 14726 }, { "epoch": 0.94, "learning_rate": 1.623930041146604e-06, "loss": 0.1436, "step": 14727 }, { "epoch": 0.94, "learning_rate": 1.6202048426483651e-06, "loss": 0.148, "step": 14728 }, { "epoch": 0.94, "learning_rate": 1.616483886853082e-06, "loss": 0.1459, "step": 14729 }, { "epoch": 0.94, "learning_rate": 1.6127671739212257e-06, "loss": 0.126, "step": 14730 }, { "epoch": 0.94, "learning_rate": 1.6090547040130689e-06, "loss": 0.137, "step": 14731 }, { "epoch": 0.94, "learning_rate": 1.6053464772887273e-06, "loss": 0.1407, "step": 14732 }, { "epoch": 0.94, "learning_rate": 1.6016424939081065e-06, "loss": 0.139, "step": 14733 }, { "epoch": 0.94, "learning_rate": 1.5979427540309566e-06, "loss": 0.1185, "step": 14734 }, { "epoch": 0.94, "learning_rate": 1.5942472578168167e-06, "loss": 0.13, "step": 14735 }, { "epoch": 0.94, "learning_rate": 1.5905560054250812e-06, "loss": 0.1296, "step": 14736 }, { "epoch": 0.94, "learning_rate": 1.5868689970149231e-06, "loss": 0.1261, "step": 14737 }, { "epoch": 0.94, "learning_rate": 1.5831862327453483e-06, "loss": 0.1326, "step": 14738 }, { "epoch": 0.95, "learning_rate": 1.5795077127751856e-06, "loss": 0.1215, "step": 14739 }, { "epoch": 0.95, "learning_rate": 1.5758334372630523e-06, "loss": 0.1181, "step": 14740 }, { "epoch": 0.95, "learning_rate": 1.572163406367433e-06, "loss": 0.1282, "step": 14741 }, { "epoch": 0.95, "learning_rate": 1.5684976202465784e-06, "loss": 0.1541, "step": 14742 }, { "epoch": 0.95, "learning_rate": 1.564836079058607e-06, "loss": 0.1242, "step": 14743 }, { "epoch": 0.95, "learning_rate": 1.5611787829613922e-06, "loss": 0.1277, "step": 14744 }, { "epoch": 0.95, "learning_rate": 1.5575257321126745e-06, "loss": 0.1196, "step": 14745 }, { "epoch": 0.95, "learning_rate": 1.5538769266699948e-06, "loss": 0.1361, "step": 14746 }, { "epoch": 0.95, "learning_rate": 1.550232366790705e-06, "loss": 0.1326, "step": 14747 }, { "epoch": 0.95, "learning_rate": 1.5465920526319788e-06, "loss": 0.1312, "step": 14748 }, { "epoch": 0.95, "learning_rate": 1.5429559843508024e-06, "loss": 0.1309, "step": 14749 }, { "epoch": 0.95, "learning_rate": 1.5393241621040167e-06, "loss": 0.1191, "step": 14750 }, { "epoch": 0.95, "learning_rate": 1.5356965860482075e-06, "loss": 0.1344, "step": 14751 }, { "epoch": 0.95, "learning_rate": 1.5320732563398388e-06, "loss": 0.1392, "step": 14752 }, { "epoch": 0.95, "learning_rate": 1.5284541731351631e-06, "loss": 0.1195, "step": 14753 }, { "epoch": 0.95, "learning_rate": 1.5248393365902447e-06, "loss": 0.1336, "step": 14754 }, { "epoch": 0.95, "learning_rate": 1.5212287468609809e-06, "loss": 0.1301, "step": 14755 }, { "epoch": 0.95, "learning_rate": 1.5176224041031028e-06, "loss": 0.1346, "step": 14756 }, { "epoch": 0.95, "learning_rate": 1.5140203084721194e-06, "loss": 0.1258, "step": 14757 }, { "epoch": 0.95, "learning_rate": 1.510422460123373e-06, "loss": 0.1352, "step": 14758 }, { "epoch": 0.95, "learning_rate": 1.5068288592120283e-06, "loss": 0.138, "step": 14759 }, { "epoch": 0.95, "learning_rate": 1.5032395058930615e-06, "loss": 0.1531, "step": 14760 }, { "epoch": 0.95, "learning_rate": 1.4996544003212487e-06, "loss": 0.1325, "step": 14761 }, { "epoch": 0.95, "learning_rate": 1.4960735426512217e-06, "loss": 0.1224, "step": 14762 }, { "epoch": 0.95, "learning_rate": 1.4924969330374127e-06, "loss": 0.1373, "step": 14763 }, { "epoch": 0.95, "learning_rate": 1.4889245716340539e-06, "loss": 0.1304, "step": 14764 }, { "epoch": 0.95, "learning_rate": 1.4853564585951995e-06, "loss": 0.1475, "step": 14765 }, { "epoch": 0.95, "learning_rate": 1.481792594074738e-06, "loss": 0.1355, "step": 14766 }, { "epoch": 0.95, "learning_rate": 1.4782329782263571e-06, "loss": 0.1346, "step": 14767 }, { "epoch": 0.95, "learning_rate": 1.4746776112035787e-06, "loss": 0.1336, "step": 14768 }, { "epoch": 0.95, "learning_rate": 1.4711264931597134e-06, "loss": 0.1518, "step": 14769 }, { "epoch": 0.95, "learning_rate": 1.4675796242479167e-06, "loss": 0.1371, "step": 14770 }, { "epoch": 0.95, "learning_rate": 1.464037004621155e-06, "loss": 0.1473, "step": 14771 }, { "epoch": 0.95, "learning_rate": 1.4604986344321947e-06, "loss": 0.1343, "step": 14772 }, { "epoch": 0.95, "learning_rate": 1.4569645138336364e-06, "loss": 0.1425, "step": 14773 }, { "epoch": 0.95, "learning_rate": 1.4534346429779023e-06, "loss": 0.1154, "step": 14774 }, { "epoch": 0.95, "learning_rate": 1.449909022017204e-06, "loss": 0.1087, "step": 14775 }, { "epoch": 0.95, "learning_rate": 1.4463876511035869e-06, "loss": 0.135, "step": 14776 }, { "epoch": 0.95, "learning_rate": 1.4428705303889177e-06, "loss": 0.1302, "step": 14777 }, { "epoch": 0.95, "learning_rate": 1.439357660024876e-06, "loss": 0.1418, "step": 14778 }, { "epoch": 0.95, "learning_rate": 1.4358490401629733e-06, "loss": 0.135, "step": 14779 }, { "epoch": 0.95, "learning_rate": 1.4323446709544885e-06, "loss": 0.1257, "step": 14780 }, { "epoch": 0.95, "learning_rate": 1.4288445525505678e-06, "loss": 0.1477, "step": 14781 }, { "epoch": 0.95, "learning_rate": 1.4253486851021569e-06, "loss": 0.1323, "step": 14782 }, { "epoch": 0.95, "learning_rate": 1.421857068760013e-06, "loss": 0.1353, "step": 14783 }, { "epoch": 0.95, "learning_rate": 1.4183697036747267e-06, "loss": 0.1393, "step": 14784 }, { "epoch": 0.95, "learning_rate": 1.414886589996678e-06, "loss": 0.1256, "step": 14785 }, { "epoch": 0.95, "learning_rate": 1.4114077278760908e-06, "loss": 0.1206, "step": 14786 }, { "epoch": 0.95, "learning_rate": 1.4079331174629783e-06, "loss": 0.1327, "step": 14787 }, { "epoch": 0.95, "learning_rate": 1.404462758907199e-06, "loss": 0.1361, "step": 14788 }, { "epoch": 0.95, "learning_rate": 1.4009966523584105e-06, "loss": 0.1366, "step": 14789 }, { "epoch": 0.95, "learning_rate": 1.3975347979661047e-06, "loss": 0.1419, "step": 14790 }, { "epoch": 0.95, "learning_rate": 1.3940771958795506e-06, "loss": 0.1322, "step": 14791 }, { "epoch": 0.95, "learning_rate": 1.3906238462478738e-06, "loss": 0.1586, "step": 14792 }, { "epoch": 0.95, "learning_rate": 1.3871747492200105e-06, "loss": 0.1351, "step": 14793 }, { "epoch": 0.95, "learning_rate": 1.3837299049446861e-06, "loss": 0.1366, "step": 14794 }, { "epoch": 0.95, "learning_rate": 1.3802893135704708e-06, "loss": 0.1242, "step": 14795 }, { "epoch": 0.95, "learning_rate": 1.376852975245746e-06, "loss": 0.1226, "step": 14796 }, { "epoch": 0.95, "learning_rate": 1.373420890118715e-06, "loss": 0.1277, "step": 14797 }, { "epoch": 0.95, "learning_rate": 1.3699930583373821e-06, "loss": 0.1318, "step": 14798 }, { "epoch": 0.95, "learning_rate": 1.366569480049562e-06, "loss": 0.124, "step": 14799 }, { "epoch": 0.95, "learning_rate": 1.3631501554029148e-06, "loss": 0.1312, "step": 14800 }, { "epoch": 0.95, "learning_rate": 1.3597350845448998e-06, "loss": 0.1352, "step": 14801 }, { "epoch": 0.95, "learning_rate": 1.3563242676227772e-06, "loss": 0.1311, "step": 14802 }, { "epoch": 0.95, "learning_rate": 1.3529177047836627e-06, "loss": 0.146, "step": 14803 }, { "epoch": 0.95, "learning_rate": 1.3495153961744612e-06, "loss": 0.133, "step": 14804 }, { "epoch": 0.95, "learning_rate": 1.3461173419418994e-06, "loss": 0.1434, "step": 14805 }, { "epoch": 0.95, "learning_rate": 1.3427235422325158e-06, "loss": 0.1187, "step": 14806 }, { "epoch": 0.95, "learning_rate": 1.339333997192671e-06, "loss": 0.1332, "step": 14807 }, { "epoch": 0.95, "learning_rate": 1.3359487069685484e-06, "loss": 0.1296, "step": 14808 }, { "epoch": 0.95, "learning_rate": 1.3325676717061308e-06, "loss": 0.1431, "step": 14809 }, { "epoch": 0.95, "learning_rate": 1.3291908915512353e-06, "loss": 0.1312, "step": 14810 }, { "epoch": 0.95, "learning_rate": 1.3258183666495006e-06, "loss": 0.1301, "step": 14811 }, { "epoch": 0.95, "learning_rate": 1.322450097146355e-06, "loss": 0.1349, "step": 14812 }, { "epoch": 0.95, "learning_rate": 1.3190860831870378e-06, "loss": 0.1446, "step": 14813 }, { "epoch": 0.95, "learning_rate": 1.3157263249166662e-06, "loss": 0.1257, "step": 14814 }, { "epoch": 0.95, "learning_rate": 1.3123708224801023e-06, "loss": 0.1353, "step": 14815 }, { "epoch": 0.95, "learning_rate": 1.3090195760220637e-06, "loss": 0.1271, "step": 14816 }, { "epoch": 0.95, "learning_rate": 1.3056725856870788e-06, "loss": 0.1298, "step": 14817 }, { "epoch": 0.95, "learning_rate": 1.3023298516194881e-06, "loss": 0.1419, "step": 14818 }, { "epoch": 0.95, "learning_rate": 1.298991373963443e-06, "loss": 0.1332, "step": 14819 }, { "epoch": 0.95, "learning_rate": 1.295657152862928e-06, "loss": 0.1288, "step": 14820 }, { "epoch": 0.95, "learning_rate": 1.2923271884617283e-06, "loss": 0.1403, "step": 14821 }, { "epoch": 0.95, "learning_rate": 1.289001480903429e-06, "loss": 0.1206, "step": 14822 }, { "epoch": 0.95, "learning_rate": 1.2856800303315041e-06, "loss": 0.1235, "step": 14823 }, { "epoch": 0.95, "learning_rate": 1.2823628368891505e-06, "loss": 0.1303, "step": 14824 }, { "epoch": 0.95, "learning_rate": 1.2790499007194424e-06, "loss": 0.1338, "step": 14825 }, { "epoch": 0.95, "learning_rate": 1.2757412219652655e-06, "loss": 0.1197, "step": 14826 }, { "epoch": 0.95, "learning_rate": 1.2724368007692721e-06, "loss": 0.1309, "step": 14827 }, { "epoch": 0.95, "learning_rate": 1.2691366372739932e-06, "loss": 0.1337, "step": 14828 }, { "epoch": 0.95, "learning_rate": 1.265840731621748e-06, "loss": 0.1446, "step": 14829 }, { "epoch": 0.95, "learning_rate": 1.2625490839546784e-06, "loss": 0.1365, "step": 14830 }, { "epoch": 0.95, "learning_rate": 1.259261694414726e-06, "loss": 0.1447, "step": 14831 }, { "epoch": 0.95, "learning_rate": 1.2559785631436894e-06, "loss": 0.121, "step": 14832 }, { "epoch": 0.95, "learning_rate": 1.2526996902831212e-06, "loss": 0.135, "step": 14833 }, { "epoch": 0.95, "learning_rate": 1.249425075974442e-06, "loss": 0.1296, "step": 14834 }, { "epoch": 0.95, "learning_rate": 1.2461547203588719e-06, "loss": 0.132, "step": 14835 }, { "epoch": 0.95, "learning_rate": 1.242888623577454e-06, "loss": 0.1359, "step": 14836 }, { "epoch": 0.95, "learning_rate": 1.2396267857710308e-06, "loss": 0.1351, "step": 14837 }, { "epoch": 0.95, "learning_rate": 1.236369207080268e-06, "loss": 0.1317, "step": 14838 }, { "epoch": 0.95, "learning_rate": 1.2331158876456638e-06, "loss": 0.155, "step": 14839 }, { "epoch": 0.95, "learning_rate": 1.2298668276075176e-06, "loss": 0.1391, "step": 14840 }, { "epoch": 0.95, "learning_rate": 1.2266220271059393e-06, "loss": 0.1422, "step": 14841 }, { "epoch": 0.95, "learning_rate": 1.2233814862808613e-06, "loss": 0.1462, "step": 14842 }, { "epoch": 0.95, "learning_rate": 1.2201452052720498e-06, "loss": 0.1585, "step": 14843 }, { "epoch": 0.95, "learning_rate": 1.2169131842190706e-06, "loss": 0.16, "step": 14844 }, { "epoch": 0.95, "learning_rate": 1.2136854232612793e-06, "loss": 0.138, "step": 14845 }, { "epoch": 0.95, "learning_rate": 1.21046192253792e-06, "loss": 0.1422, "step": 14846 }, { "epoch": 0.95, "learning_rate": 1.20724268218797e-06, "loss": 0.1272, "step": 14847 }, { "epoch": 0.95, "learning_rate": 1.2040277023502743e-06, "loss": 0.1398, "step": 14848 }, { "epoch": 0.95, "learning_rate": 1.2008169831634885e-06, "loss": 0.1429, "step": 14849 }, { "epoch": 0.95, "learning_rate": 1.1976105247660685e-06, "loss": 0.1269, "step": 14850 }, { "epoch": 0.95, "learning_rate": 1.1944083272963035e-06, "loss": 0.1197, "step": 14851 }, { "epoch": 0.95, "learning_rate": 1.1912103908922945e-06, "loss": 0.1383, "step": 14852 }, { "epoch": 0.95, "learning_rate": 1.188016715691931e-06, "loss": 0.1229, "step": 14853 }, { "epoch": 0.95, "learning_rate": 1.1848273018329691e-06, "loss": 0.1121, "step": 14854 }, { "epoch": 0.95, "learning_rate": 1.1816421494529328e-06, "loss": 0.1231, "step": 14855 }, { "epoch": 0.95, "learning_rate": 1.1784612586891897e-06, "loss": 0.1353, "step": 14856 }, { "epoch": 0.95, "learning_rate": 1.17528462967893e-06, "loss": 0.1309, "step": 14857 }, { "epoch": 0.95, "learning_rate": 1.172112262559144e-06, "loss": 0.1365, "step": 14858 }, { "epoch": 0.95, "learning_rate": 1.1689441574666448e-06, "loss": 0.1356, "step": 14859 }, { "epoch": 0.95, "learning_rate": 1.1657803145380454e-06, "loss": 0.1296, "step": 14860 }, { "epoch": 0.95, "learning_rate": 1.1626207339098138e-06, "loss": 0.1391, "step": 14861 }, { "epoch": 0.95, "learning_rate": 1.159465415718175e-06, "loss": 0.1357, "step": 14862 }, { "epoch": 0.95, "learning_rate": 1.156314360099231e-06, "loss": 0.1271, "step": 14863 }, { "epoch": 0.95, "learning_rate": 1.1531675671888619e-06, "loss": 0.1319, "step": 14864 }, { "epoch": 0.95, "learning_rate": 1.1500250371227817e-06, "loss": 0.1239, "step": 14865 }, { "epoch": 0.95, "learning_rate": 1.1468867700365148e-06, "loss": 0.1223, "step": 14866 }, { "epoch": 0.95, "learning_rate": 1.1437527660653869e-06, "loss": 0.1432, "step": 14867 }, { "epoch": 0.95, "learning_rate": 1.1406230253445782e-06, "loss": 0.1346, "step": 14868 }, { "epoch": 0.95, "learning_rate": 1.1374975480090367e-06, "loss": 0.1462, "step": 14869 }, { "epoch": 0.95, "learning_rate": 1.1343763341935765e-06, "loss": 0.1282, "step": 14870 }, { "epoch": 0.95, "learning_rate": 1.131259384032779e-06, "loss": 0.1254, "step": 14871 }, { "epoch": 0.95, "learning_rate": 1.128146697661081e-06, "loss": 0.134, "step": 14872 }, { "epoch": 0.95, "learning_rate": 1.1250382752127197e-06, "loss": 0.1401, "step": 14873 }, { "epoch": 0.95, "learning_rate": 1.1219341168217212e-06, "loss": 0.1393, "step": 14874 }, { "epoch": 0.95, "learning_rate": 1.1188342226219895e-06, "loss": 0.1284, "step": 14875 }, { "epoch": 0.95, "learning_rate": 1.1157385927471843e-06, "loss": 0.1275, "step": 14876 }, { "epoch": 0.95, "learning_rate": 1.1126472273308319e-06, "loss": 0.1289, "step": 14877 }, { "epoch": 0.95, "learning_rate": 1.1095601265062262e-06, "loss": 0.134, "step": 14878 }, { "epoch": 0.95, "learning_rate": 1.1064772904065158e-06, "loss": 0.1378, "step": 14879 }, { "epoch": 0.95, "learning_rate": 1.10339871916465e-06, "loss": 0.1319, "step": 14880 }, { "epoch": 0.95, "learning_rate": 1.1003244129133782e-06, "loss": 0.1388, "step": 14881 }, { "epoch": 0.95, "learning_rate": 1.0972543717852947e-06, "loss": 0.1307, "step": 14882 }, { "epoch": 0.95, "learning_rate": 1.094188595912804e-06, "loss": 0.1388, "step": 14883 }, { "epoch": 0.95, "learning_rate": 1.091127085428112e-06, "loss": 0.1395, "step": 14884 }, { "epoch": 0.95, "learning_rate": 1.088069840463235e-06, "loss": 0.1568, "step": 14885 }, { "epoch": 0.95, "learning_rate": 1.0850168611500456e-06, "loss": 0.1309, "step": 14886 }, { "epoch": 0.95, "learning_rate": 1.0819681476201937e-06, "loss": 0.1381, "step": 14887 }, { "epoch": 0.95, "learning_rate": 1.0789237000051522e-06, "loss": 0.1219, "step": 14888 }, { "epoch": 0.95, "learning_rate": 1.0758835184362271e-06, "loss": 0.1403, "step": 14889 }, { "epoch": 0.95, "learning_rate": 1.0728476030445022e-06, "loss": 0.1195, "step": 14890 }, { "epoch": 0.95, "learning_rate": 1.0698159539609399e-06, "loss": 0.1524, "step": 14891 }, { "epoch": 0.95, "learning_rate": 1.0667885713162684e-06, "loss": 0.1357, "step": 14892 }, { "epoch": 0.95, "learning_rate": 1.0637654552410281e-06, "loss": 0.1363, "step": 14893 }, { "epoch": 0.95, "learning_rate": 1.0607466058656258e-06, "loss": 0.1374, "step": 14894 }, { "epoch": 0.96, "learning_rate": 1.0577320233202126e-06, "loss": 0.1438, "step": 14895 }, { "epoch": 0.96, "learning_rate": 1.0547217077348294e-06, "loss": 0.1241, "step": 14896 }, { "epoch": 0.96, "learning_rate": 1.0517156592392717e-06, "loss": 0.1286, "step": 14897 }, { "epoch": 0.96, "learning_rate": 1.0487138779632034e-06, "loss": 0.145, "step": 14898 }, { "epoch": 0.96, "learning_rate": 1.0457163640360646e-06, "loss": 0.1258, "step": 14899 }, { "epoch": 0.96, "learning_rate": 1.0427231175871189e-06, "loss": 0.1287, "step": 14900 }, { "epoch": 0.96, "learning_rate": 1.039734138745474e-06, "loss": 0.1337, "step": 14901 }, { "epoch": 0.96, "learning_rate": 1.0367494276400048e-06, "loss": 0.1323, "step": 14902 }, { "epoch": 0.96, "learning_rate": 1.0337689843994414e-06, "loss": 0.123, "step": 14903 }, { "epoch": 0.96, "learning_rate": 1.0307928091523145e-06, "loss": 0.1485, "step": 14904 }, { "epoch": 0.96, "learning_rate": 1.027820902026988e-06, "loss": 0.1367, "step": 14905 }, { "epoch": 0.96, "learning_rate": 1.024853263151615e-06, "loss": 0.1335, "step": 14906 }, { "epoch": 0.96, "learning_rate": 1.021889892654182e-06, "loss": 0.1217, "step": 14907 }, { "epoch": 0.96, "learning_rate": 1.0189307906624867e-06, "loss": 0.1548, "step": 14908 }, { "epoch": 0.96, "learning_rate": 1.0159759573041272e-06, "loss": 0.135, "step": 14909 }, { "epoch": 0.96, "learning_rate": 1.0130253927065458e-06, "loss": 0.1247, "step": 14910 }, { "epoch": 0.96, "learning_rate": 1.0100790969969964e-06, "loss": 0.132, "step": 14911 }, { "epoch": 0.96, "learning_rate": 1.007137070302533e-06, "loss": 0.1342, "step": 14912 }, { "epoch": 0.96, "learning_rate": 1.0041993127500205e-06, "loss": 0.1346, "step": 14913 }, { "epoch": 0.96, "learning_rate": 1.0012658244661798e-06, "loss": 0.1331, "step": 14914 }, { "epoch": 0.96, "learning_rate": 9.98336605577488e-07, "loss": 0.1312, "step": 14915 }, { "epoch": 0.96, "learning_rate": 9.954116562102878e-07, "loss": 0.1353, "step": 14916 }, { "epoch": 0.96, "learning_rate": 9.924909764907231e-07, "loss": 0.1379, "step": 14917 }, { "epoch": 0.96, "learning_rate": 9.895745665447376e-07, "loss": 0.1158, "step": 14918 }, { "epoch": 0.96, "learning_rate": 9.866624264981084e-07, "loss": 0.1319, "step": 14919 }, { "epoch": 0.96, "learning_rate": 9.83754556476435e-07, "loss": 0.1309, "step": 14920 }, { "epoch": 0.96, "learning_rate": 9.808509566051061e-07, "loss": 0.1449, "step": 14921 }, { "epoch": 0.96, "learning_rate": 9.779516270093436e-07, "loss": 0.1445, "step": 14922 }, { "epoch": 0.96, "learning_rate": 9.75056567814192e-07, "loss": 0.1372, "step": 14923 }, { "epoch": 0.96, "learning_rate": 9.72165779144496e-07, "loss": 0.1511, "step": 14924 }, { "epoch": 0.96, "learning_rate": 9.692792611249224e-07, "loss": 0.1294, "step": 14925 }, { "epoch": 0.96, "learning_rate": 9.663970138799605e-07, "loss": 0.1389, "step": 14926 }, { "epoch": 0.96, "learning_rate": 9.63519037533911e-07, "loss": 0.1175, "step": 14927 }, { "epoch": 0.96, "learning_rate": 9.606453322108744e-07, "loss": 0.139, "step": 14928 }, { "epoch": 0.96, "learning_rate": 9.57775898034796e-07, "loss": 0.1239, "step": 14929 }, { "epoch": 0.96, "learning_rate": 9.549107351294106e-07, "loss": 0.1226, "step": 14930 }, { "epoch": 0.96, "learning_rate": 9.520498436182967e-07, "loss": 0.1344, "step": 14931 }, { "epoch": 0.96, "learning_rate": 9.491932236248113e-07, "loss": 0.133, "step": 14932 }, { "epoch": 0.96, "learning_rate": 9.46340875272178e-07, "loss": 0.1489, "step": 14933 }, { "epoch": 0.96, "learning_rate": 9.434927986833764e-07, "loss": 0.1293, "step": 14934 }, { "epoch": 0.96, "learning_rate": 9.406489939812413e-07, "loss": 0.1407, "step": 14935 }, { "epoch": 0.96, "learning_rate": 9.378094612884192e-07, "loss": 0.1337, "step": 14936 }, { "epoch": 0.96, "learning_rate": 9.349742007273566e-07, "loss": 0.1113, "step": 14937 }, { "epoch": 0.96, "learning_rate": 9.321432124203444e-07, "loss": 0.1309, "step": 14938 }, { "epoch": 0.96, "learning_rate": 9.293164964894629e-07, "loss": 0.1425, "step": 14939 }, { "epoch": 0.96, "learning_rate": 9.264940530566036e-07, "loss": 0.1334, "step": 14940 }, { "epoch": 0.96, "learning_rate": 9.236758822435132e-07, "loss": 0.1286, "step": 14941 }, { "epoch": 0.96, "learning_rate": 9.208619841716947e-07, "loss": 0.1362, "step": 14942 }, { "epoch": 0.96, "learning_rate": 9.180523589625179e-07, "loss": 0.123, "step": 14943 }, { "epoch": 0.96, "learning_rate": 9.152470067371521e-07, "loss": 0.1335, "step": 14944 }, { "epoch": 0.96, "learning_rate": 9.124459276165787e-07, "loss": 0.1347, "step": 14945 }, { "epoch": 0.96, "learning_rate": 9.096491217215896e-07, "loss": 0.1394, "step": 14946 }, { "epoch": 0.96, "learning_rate": 9.068565891728109e-07, "loss": 0.1518, "step": 14947 }, { "epoch": 0.96, "learning_rate": 9.040683300906683e-07, "loss": 0.1372, "step": 14948 }, { "epoch": 0.96, "learning_rate": 9.012843445953878e-07, "loss": 0.1297, "step": 14949 }, { "epoch": 0.96, "learning_rate": 8.985046328070512e-07, "loss": 0.1308, "step": 14950 }, { "epoch": 0.96, "learning_rate": 8.957291948455405e-07, "loss": 0.1352, "step": 14951 }, { "epoch": 0.96, "learning_rate": 8.929580308305374e-07, "loss": 0.1298, "step": 14952 }, { "epoch": 0.96, "learning_rate": 8.901911408815577e-07, "loss": 0.1341, "step": 14953 }, { "epoch": 0.96, "learning_rate": 8.87428525117917e-07, "loss": 0.1327, "step": 14954 }, { "epoch": 0.96, "learning_rate": 8.846701836587534e-07, "loss": 0.1148, "step": 14955 }, { "epoch": 0.96, "learning_rate": 8.819161166230383e-07, "loss": 0.1273, "step": 14956 }, { "epoch": 0.96, "learning_rate": 8.791663241295323e-07, "loss": 0.1381, "step": 14957 }, { "epoch": 0.96, "learning_rate": 8.764208062968182e-07, "loss": 0.1284, "step": 14958 }, { "epoch": 0.96, "learning_rate": 8.736795632433015e-07, "loss": 0.128, "step": 14959 }, { "epoch": 0.96, "learning_rate": 8.709425950872097e-07, "loss": 0.1345, "step": 14960 }, { "epoch": 0.96, "learning_rate": 8.682099019465595e-07, "loss": 0.1412, "step": 14961 }, { "epoch": 0.96, "learning_rate": 8.654814839392233e-07, "loss": 0.1504, "step": 14962 }, { "epoch": 0.96, "learning_rate": 8.627573411828405e-07, "loss": 0.1274, "step": 14963 }, { "epoch": 0.96, "learning_rate": 8.600374737949057e-07, "loss": 0.124, "step": 14964 }, { "epoch": 0.96, "learning_rate": 8.573218818927143e-07, "loss": 0.133, "step": 14965 }, { "epoch": 0.96, "learning_rate": 8.546105655933945e-07, "loss": 0.1497, "step": 14966 }, { "epoch": 0.96, "learning_rate": 8.519035250138307e-07, "loss": 0.1271, "step": 14967 }, { "epoch": 0.96, "learning_rate": 8.492007602708074e-07, "loss": 0.1142, "step": 14968 }, { "epoch": 0.96, "learning_rate": 8.465022714808646e-07, "loss": 0.1401, "step": 14969 }, { "epoch": 0.96, "learning_rate": 8.43808058760387e-07, "loss": 0.1337, "step": 14970 }, { "epoch": 0.96, "learning_rate": 8.411181222255482e-07, "loss": 0.1431, "step": 14971 }, { "epoch": 0.96, "learning_rate": 8.384324619923667e-07, "loss": 0.133, "step": 14972 }, { "epoch": 0.96, "learning_rate": 8.35751078176672e-07, "loss": 0.1258, "step": 14973 }, { "epoch": 0.96, "learning_rate": 8.330739708940716e-07, "loss": 0.1451, "step": 14974 }, { "epoch": 0.96, "learning_rate": 8.304011402600509e-07, "loss": 0.1216, "step": 14975 }, { "epoch": 0.96, "learning_rate": 8.277325863898511e-07, "loss": 0.1299, "step": 14976 }, { "epoch": 0.96, "learning_rate": 8.250683093985689e-07, "loss": 0.1393, "step": 14977 }, { "epoch": 0.96, "learning_rate": 8.224083094011015e-07, "loss": 0.1322, "step": 14978 }, { "epoch": 0.96, "learning_rate": 8.19752586512168e-07, "loss": 0.1424, "step": 14979 }, { "epoch": 0.96, "learning_rate": 8.171011408462881e-07, "loss": 0.1454, "step": 14980 }, { "epoch": 0.96, "learning_rate": 8.144539725178146e-07, "loss": 0.1353, "step": 14981 }, { "epoch": 0.96, "learning_rate": 8.118110816409119e-07, "loss": 0.1367, "step": 14982 }, { "epoch": 0.96, "learning_rate": 8.091724683295443e-07, "loss": 0.1442, "step": 14983 }, { "epoch": 0.96, "learning_rate": 8.065381326975097e-07, "loss": 0.1247, "step": 14984 }, { "epoch": 0.96, "learning_rate": 8.039080748584282e-07, "loss": 0.1119, "step": 14985 }, { "epoch": 0.96, "learning_rate": 8.012822949256982e-07, "loss": 0.1355, "step": 14986 }, { "epoch": 0.96, "learning_rate": 7.986607930125844e-07, "loss": 0.1254, "step": 14987 }, { "epoch": 0.96, "learning_rate": 7.960435692321188e-07, "loss": 0.1431, "step": 14988 }, { "epoch": 0.96, "learning_rate": 7.934306236971889e-07, "loss": 0.1463, "step": 14989 }, { "epoch": 0.96, "learning_rate": 7.908219565204711e-07, "loss": 0.1393, "step": 14990 }, { "epoch": 0.96, "learning_rate": 7.882175678144643e-07, "loss": 0.1189, "step": 14991 }, { "epoch": 0.96, "learning_rate": 7.856174576915009e-07, "loss": 0.1358, "step": 14992 }, { "epoch": 0.96, "learning_rate": 7.830216262636802e-07, "loss": 0.123, "step": 14993 }, { "epoch": 0.96, "learning_rate": 7.804300736429903e-07, "loss": 0.1443, "step": 14994 }, { "epoch": 0.96, "learning_rate": 7.77842799941153e-07, "loss": 0.1355, "step": 14995 }, { "epoch": 0.96, "learning_rate": 7.752598052697679e-07, "loss": 0.1562, "step": 14996 }, { "epoch": 0.96, "learning_rate": 7.726810897402459e-07, "loss": 0.1303, "step": 14997 }, { "epoch": 0.96, "learning_rate": 7.701066534637535e-07, "loss": 0.1399, "step": 14998 }, { "epoch": 0.96, "learning_rate": 7.675364965513577e-07, "loss": 0.1458, "step": 14999 }, { "epoch": 0.96, "learning_rate": 7.649706191138695e-07, "loss": 0.1434, "step": 15000 }, { "epoch": 0.96, "learning_rate": 7.624090212619562e-07, "loss": 0.138, "step": 15001 }, { "epoch": 0.96, "learning_rate": 7.598517031060848e-07, "loss": 0.1227, "step": 15002 }, { "epoch": 0.96, "learning_rate": 7.572986647565339e-07, "loss": 0.1344, "step": 15003 }, { "epoch": 0.96, "learning_rate": 7.547499063234264e-07, "loss": 0.1294, "step": 15004 }, { "epoch": 0.96, "learning_rate": 7.522054279166635e-07, "loss": 0.1262, "step": 15005 }, { "epoch": 0.96, "learning_rate": 7.496652296459905e-07, "loss": 0.1284, "step": 15006 }, { "epoch": 0.96, "learning_rate": 7.471293116209421e-07, "loss": 0.1337, "step": 15007 }, { "epoch": 0.96, "learning_rate": 7.445976739508865e-07, "loss": 0.1177, "step": 15008 }, { "epoch": 0.96, "learning_rate": 7.420703167449916e-07, "loss": 0.1376, "step": 15009 }, { "epoch": 0.96, "learning_rate": 7.395472401122705e-07, "loss": 0.1317, "step": 15010 }, { "epoch": 0.96, "learning_rate": 7.37028444161525e-07, "loss": 0.1298, "step": 15011 }, { "epoch": 0.96, "learning_rate": 7.345139290013791e-07, "loss": 0.1483, "step": 15012 }, { "epoch": 0.96, "learning_rate": 7.320036947402798e-07, "loss": 0.1328, "step": 15013 }, { "epoch": 0.96, "learning_rate": 7.294977414864624e-07, "loss": 0.1399, "step": 15014 }, { "epoch": 0.96, "learning_rate": 7.269960693480293e-07, "loss": 0.148, "step": 15015 }, { "epoch": 0.96, "learning_rate": 7.244986784328389e-07, "loss": 0.1223, "step": 15016 }, { "epoch": 0.96, "learning_rate": 7.220055688486049e-07, "loss": 0.1229, "step": 15017 }, { "epoch": 0.96, "learning_rate": 7.195167407028525e-07, "loss": 0.1577, "step": 15018 }, { "epoch": 0.96, "learning_rate": 7.170321941028956e-07, "loss": 0.1564, "step": 15019 }, { "epoch": 0.96, "learning_rate": 7.145519291558934e-07, "loss": 0.1337, "step": 15020 }, { "epoch": 0.96, "learning_rate": 7.120759459688153e-07, "loss": 0.127, "step": 15021 }, { "epoch": 0.96, "learning_rate": 7.096042446484319e-07, "loss": 0.1437, "step": 15022 }, { "epoch": 0.96, "learning_rate": 7.071368253013244e-07, "loss": 0.1395, "step": 15023 }, { "epoch": 0.96, "learning_rate": 7.0467368803393e-07, "loss": 0.1348, "step": 15024 }, { "epoch": 0.96, "learning_rate": 7.022148329524525e-07, "loss": 0.1318, "step": 15025 }, { "epoch": 0.96, "learning_rate": 6.997602601629404e-07, "loss": 0.1276, "step": 15026 }, { "epoch": 0.96, "learning_rate": 6.973099697712537e-07, "loss": 0.1392, "step": 15027 }, { "epoch": 0.96, "learning_rate": 6.94863961883041e-07, "loss": 0.1356, "step": 15028 }, { "epoch": 0.96, "learning_rate": 6.92422236603818e-07, "loss": 0.1387, "step": 15029 }, { "epoch": 0.96, "learning_rate": 6.89984794038867e-07, "loss": 0.1195, "step": 15030 }, { "epoch": 0.96, "learning_rate": 6.875516342933042e-07, "loss": 0.1183, "step": 15031 }, { "epoch": 0.96, "learning_rate": 6.851227574720675e-07, "loss": 0.1196, "step": 15032 }, { "epoch": 0.96, "learning_rate": 6.826981636799068e-07, "loss": 0.14, "step": 15033 }, { "epoch": 0.96, "learning_rate": 6.802778530213827e-07, "loss": 0.125, "step": 15034 }, { "epoch": 0.96, "learning_rate": 6.778618256008673e-07, "loss": 0.1333, "step": 15035 }, { "epoch": 0.96, "learning_rate": 6.754500815225551e-07, "loss": 0.1308, "step": 15036 }, { "epoch": 0.96, "learning_rate": 6.730426208904517e-07, "loss": 0.1405, "step": 15037 }, { "epoch": 0.96, "learning_rate": 6.706394438083963e-07, "loss": 0.132, "step": 15038 }, { "epoch": 0.96, "learning_rate": 6.682405503800171e-07, "loss": 0.1414, "step": 15039 }, { "epoch": 0.96, "learning_rate": 6.658459407087536e-07, "loss": 0.1548, "step": 15040 }, { "epoch": 0.96, "learning_rate": 6.634556148979121e-07, "loss": 0.1428, "step": 15041 }, { "epoch": 0.96, "learning_rate": 6.610695730505323e-07, "loss": 0.1186, "step": 15042 }, { "epoch": 0.96, "learning_rate": 6.58687815269543e-07, "loss": 0.1369, "step": 15043 }, { "epoch": 0.96, "learning_rate": 6.563103416576621e-07, "loss": 0.1345, "step": 15044 }, { "epoch": 0.96, "learning_rate": 6.539371523173965e-07, "loss": 0.1557, "step": 15045 }, { "epoch": 0.96, "learning_rate": 6.515682473511198e-07, "loss": 0.1223, "step": 15046 }, { "epoch": 0.96, "learning_rate": 6.492036268609725e-07, "loss": 0.1232, "step": 15047 }, { "epoch": 0.96, "learning_rate": 6.468432909489397e-07, "loss": 0.1163, "step": 15048 }, { "epoch": 0.96, "learning_rate": 6.444872397168067e-07, "loss": 0.1438, "step": 15049 }, { "epoch": 0.96, "learning_rate": 6.42135473266181e-07, "loss": 0.128, "step": 15050 }, { "epoch": 0.97, "learning_rate": 6.397879916984928e-07, "loss": 0.1397, "step": 15051 }, { "epoch": 0.97, "learning_rate": 6.374447951149609e-07, "loss": 0.1475, "step": 15052 }, { "epoch": 0.97, "learning_rate": 6.351058836166601e-07, "loss": 0.1424, "step": 15053 }, { "epoch": 0.97, "learning_rate": 6.32771257304443e-07, "loss": 0.1352, "step": 15054 }, { "epoch": 0.97, "learning_rate": 6.304409162789959e-07, "loss": 0.1301, "step": 15055 }, { "epoch": 0.97, "learning_rate": 6.281148606408161e-07, "loss": 0.1464, "step": 15056 }, { "epoch": 0.97, "learning_rate": 6.257930904902121e-07, "loss": 0.1462, "step": 15057 }, { "epoch": 0.97, "learning_rate": 6.234756059273261e-07, "loss": 0.1313, "step": 15058 }, { "epoch": 0.97, "learning_rate": 6.211624070520894e-07, "loss": 0.1293, "step": 15059 }, { "epoch": 0.97, "learning_rate": 6.188534939642665e-07, "loss": 0.1377, "step": 15060 }, { "epoch": 0.97, "learning_rate": 6.165488667634223e-07, "loss": 0.1397, "step": 15061 }, { "epoch": 0.97, "learning_rate": 6.14248525548955e-07, "loss": 0.1444, "step": 15062 }, { "epoch": 0.97, "learning_rate": 6.11952470420063e-07, "loss": 0.1388, "step": 15063 }, { "epoch": 0.97, "learning_rate": 6.096607014757561e-07, "loss": 0.1337, "step": 15064 }, { "epoch": 0.97, "learning_rate": 6.073732188148995e-07, "loss": 0.1445, "step": 15065 }, { "epoch": 0.97, "learning_rate": 6.050900225361034e-07, "loss": 0.1311, "step": 15066 }, { "epoch": 0.97, "learning_rate": 6.028111127378666e-07, "loss": 0.1272, "step": 15067 }, { "epoch": 0.97, "learning_rate": 6.00536489518444e-07, "loss": 0.1254, "step": 15068 }, { "epoch": 0.97, "learning_rate": 5.98266152975946e-07, "loss": 0.1563, "step": 15069 }, { "epoch": 0.97, "learning_rate": 5.960001032082718e-07, "loss": 0.1354, "step": 15070 }, { "epoch": 0.97, "learning_rate": 5.937383403131546e-07, "loss": 0.1568, "step": 15071 }, { "epoch": 0.97, "learning_rate": 5.914808643881387e-07, "loss": 0.1487, "step": 15072 }, { "epoch": 0.97, "learning_rate": 5.89227675530557e-07, "loss": 0.1369, "step": 15073 }, { "epoch": 0.97, "learning_rate": 5.869787738376098e-07, "loss": 0.135, "step": 15074 }, { "epoch": 0.97, "learning_rate": 5.847341594062528e-07, "loss": 0.1352, "step": 15075 }, { "epoch": 0.97, "learning_rate": 5.824938323333195e-07, "loss": 0.1345, "step": 15076 }, { "epoch": 0.97, "learning_rate": 5.802577927153995e-07, "loss": 0.134, "step": 15077 }, { "epoch": 0.97, "learning_rate": 5.780260406489269e-07, "loss": 0.1337, "step": 15078 }, { "epoch": 0.97, "learning_rate": 5.757985762301577e-07, "loss": 0.1301, "step": 15079 }, { "epoch": 0.97, "learning_rate": 5.735753995551485e-07, "loss": 0.132, "step": 15080 }, { "epoch": 0.97, "learning_rate": 5.713565107197783e-07, "loss": 0.1506, "step": 15081 }, { "epoch": 0.97, "learning_rate": 5.691419098197259e-07, "loss": 0.1133, "step": 15082 }, { "epoch": 0.97, "learning_rate": 5.669315969505151e-07, "loss": 0.1476, "step": 15083 }, { "epoch": 0.97, "learning_rate": 5.647255722074585e-07, "loss": 0.1472, "step": 15084 }, { "epoch": 0.97, "learning_rate": 5.625238356857022e-07, "loss": 0.1387, "step": 15085 }, { "epoch": 0.97, "learning_rate": 5.603263874801811e-07, "loss": 0.1312, "step": 15086 }, { "epoch": 0.97, "learning_rate": 5.581332276856643e-07, "loss": 0.1288, "step": 15087 }, { "epoch": 0.97, "learning_rate": 5.559443563967426e-07, "loss": 0.1261, "step": 15088 }, { "epoch": 0.97, "learning_rate": 5.537597737078071e-07, "loss": 0.1297, "step": 15089 }, { "epoch": 0.97, "learning_rate": 5.515794797130825e-07, "loss": 0.1351, "step": 15090 }, { "epoch": 0.97, "learning_rate": 5.494034745065712e-07, "loss": 0.143, "step": 15091 }, { "epoch": 0.97, "learning_rate": 5.472317581821318e-07, "loss": 0.1297, "step": 15092 }, { "epoch": 0.97, "learning_rate": 5.450643308334225e-07, "loss": 0.13, "step": 15093 }, { "epoch": 0.97, "learning_rate": 5.42901192553913e-07, "loss": 0.1362, "step": 15094 }, { "epoch": 0.97, "learning_rate": 5.407423434368842e-07, "loss": 0.1393, "step": 15095 }, { "epoch": 0.97, "learning_rate": 5.385877835754394e-07, "loss": 0.1375, "step": 15096 }, { "epoch": 0.97, "learning_rate": 5.364375130624932e-07, "loss": 0.143, "step": 15097 }, { "epoch": 0.97, "learning_rate": 5.342915319907937e-07, "loss": 0.1333, "step": 15098 }, { "epoch": 0.97, "learning_rate": 5.321498404528668e-07, "loss": 0.1586, "step": 15099 }, { "epoch": 0.97, "learning_rate": 5.300124385410943e-07, "loss": 0.1271, "step": 15100 }, { "epoch": 0.97, "learning_rate": 5.278793263476356e-07, "loss": 0.14, "step": 15101 }, { "epoch": 0.97, "learning_rate": 5.257505039645061e-07, "loss": 0.1282, "step": 15102 }, { "epoch": 0.97, "learning_rate": 5.236259714834768e-07, "loss": 0.1293, "step": 15103 }, { "epoch": 0.97, "learning_rate": 5.215057289961966e-07, "loss": 0.1323, "step": 15104 }, { "epoch": 0.97, "learning_rate": 5.193897765941036e-07, "loss": 0.147, "step": 15105 }, { "epoch": 0.97, "learning_rate": 5.172781143684357e-07, "loss": 0.1462, "step": 15106 }, { "epoch": 0.97, "learning_rate": 5.151707424102759e-07, "loss": 0.1467, "step": 15107 }, { "epoch": 0.97, "learning_rate": 5.130676608104845e-07, "loss": 0.1407, "step": 15108 }, { "epoch": 0.97, "learning_rate": 5.109688696597781e-07, "loss": 0.1515, "step": 15109 }, { "epoch": 0.97, "learning_rate": 5.088743690486619e-07, "loss": 0.1427, "step": 15110 }, { "epoch": 0.97, "learning_rate": 5.067841590674638e-07, "loss": 0.1353, "step": 15111 }, { "epoch": 0.97, "learning_rate": 5.046982398063227e-07, "loss": 0.1403, "step": 15112 }, { "epoch": 0.97, "learning_rate": 5.026166113551889e-07, "loss": 0.1236, "step": 15113 }, { "epoch": 0.97, "learning_rate": 5.005392738038572e-07, "loss": 0.1357, "step": 15114 }, { "epoch": 0.97, "learning_rate": 4.984662272418894e-07, "loss": 0.1287, "step": 15115 }, { "epoch": 0.97, "learning_rate": 4.963974717587027e-07, "loss": 0.1294, "step": 15116 }, { "epoch": 0.97, "learning_rate": 4.943330074435038e-07, "loss": 0.1303, "step": 15117 }, { "epoch": 0.97, "learning_rate": 4.922728343853322e-07, "loss": 0.145, "step": 15118 }, { "epoch": 0.97, "learning_rate": 4.902169526730393e-07, "loss": 0.137, "step": 15119 }, { "epoch": 0.97, "learning_rate": 4.881653623952654e-07, "loss": 0.1245, "step": 15120 }, { "epoch": 0.97, "learning_rate": 4.861180636405061e-07, "loss": 0.1406, "step": 15121 }, { "epoch": 0.97, "learning_rate": 4.840750564970464e-07, "loss": 0.1358, "step": 15122 }, { "epoch": 0.97, "learning_rate": 4.820363410529938e-07, "loss": 0.1336, "step": 15123 }, { "epoch": 0.97, "learning_rate": 4.800019173962556e-07, "loss": 0.1268, "step": 15124 }, { "epoch": 0.97, "learning_rate": 4.779717856145949e-07, "loss": 0.1252, "step": 15125 }, { "epoch": 0.97, "learning_rate": 4.7594594579554174e-07, "loss": 0.1216, "step": 15126 }, { "epoch": 0.97, "learning_rate": 4.739243980264707e-07, "loss": 0.1487, "step": 15127 }, { "epoch": 0.97, "learning_rate": 4.719071423945676e-07, "loss": 0.1266, "step": 15128 }, { "epoch": 0.97, "learning_rate": 4.698941789868072e-07, "loss": 0.1396, "step": 15129 }, { "epoch": 0.97, "learning_rate": 4.678855078900202e-07, "loss": 0.1272, "step": 15130 }, { "epoch": 0.97, "learning_rate": 4.6588112919082605e-07, "loss": 0.1292, "step": 15131 }, { "epoch": 0.97, "learning_rate": 4.6388104297566683e-07, "loss": 0.1313, "step": 15132 }, { "epoch": 0.97, "learning_rate": 4.618852493307957e-07, "loss": 0.146, "step": 15133 }, { "epoch": 0.97, "learning_rate": 4.5989374834228825e-07, "loss": 0.1303, "step": 15134 }, { "epoch": 0.97, "learning_rate": 4.5790654009600917e-07, "loss": 0.1279, "step": 15135 }, { "epoch": 0.97, "learning_rate": 4.559236246776788e-07, "loss": 0.1202, "step": 15136 }, { "epoch": 0.97, "learning_rate": 4.5394500217280646e-07, "loss": 0.1381, "step": 15137 }, { "epoch": 0.97, "learning_rate": 4.5197067266672387e-07, "loss": 0.1484, "step": 15138 }, { "epoch": 0.97, "learning_rate": 4.500006362445741e-07, "loss": 0.1243, "step": 15139 }, { "epoch": 0.97, "learning_rate": 4.480348929913225e-07, "loss": 0.1327, "step": 15140 }, { "epoch": 0.97, "learning_rate": 4.460734429917235e-07, "loss": 0.1335, "step": 15141 }, { "epoch": 0.97, "learning_rate": 4.4411628633038717e-07, "loss": 0.1403, "step": 15142 }, { "epoch": 0.97, "learning_rate": 4.4216342309170155e-07, "loss": 0.1317, "step": 15143 }, { "epoch": 0.97, "learning_rate": 4.4021485335988823e-07, "loss": 0.1284, "step": 15144 }, { "epoch": 0.97, "learning_rate": 4.38270577218991e-07, "loss": 0.1298, "step": 15145 }, { "epoch": 0.97, "learning_rate": 4.363305947528651e-07, "loss": 0.1216, "step": 15146 }, { "epoch": 0.97, "learning_rate": 4.3439490604514354e-07, "loss": 0.1345, "step": 15147 }, { "epoch": 0.97, "learning_rate": 4.324635111793374e-07, "loss": 0.1314, "step": 15148 }, { "epoch": 0.97, "learning_rate": 4.3053641023871325e-07, "loss": 0.1461, "step": 15149 }, { "epoch": 0.97, "learning_rate": 4.2861360330639365e-07, "loss": 0.1162, "step": 15150 }, { "epoch": 0.97, "learning_rate": 4.2669509046529e-07, "loss": 0.1183, "step": 15151 }, { "epoch": 0.97, "learning_rate": 4.247808717981472e-07, "loss": 0.1386, "step": 15152 }, { "epoch": 0.97, "learning_rate": 4.228709473875214e-07, "loss": 0.1349, "step": 15153 }, { "epoch": 0.97, "learning_rate": 4.209653173157801e-07, "loss": 0.1517, "step": 15154 }, { "epoch": 0.97, "learning_rate": 4.19063981665091e-07, "loss": 0.1301, "step": 15155 }, { "epoch": 0.97, "learning_rate": 4.171669405174772e-07, "loss": 0.1295, "step": 15156 }, { "epoch": 0.97, "learning_rate": 4.1527419395470667e-07, "loss": 0.1349, "step": 15157 }, { "epoch": 0.97, "learning_rate": 4.133857420584475e-07, "loss": 0.1337, "step": 15158 }, { "epoch": 0.97, "learning_rate": 4.115015849101234e-07, "loss": 0.1242, "step": 15159 }, { "epoch": 0.97, "learning_rate": 4.0962172259098043e-07, "loss": 0.1227, "step": 15160 }, { "epoch": 0.97, "learning_rate": 4.0774615518210934e-07, "loss": 0.1321, "step": 15161 }, { "epoch": 0.97, "learning_rate": 4.058748827643899e-07, "loss": 0.1417, "step": 15162 }, { "epoch": 0.97, "learning_rate": 4.040079054185131e-07, "loss": 0.1325, "step": 15163 }, { "epoch": 0.97, "learning_rate": 4.0214522322499225e-07, "loss": 0.1341, "step": 15164 }, { "epoch": 0.97, "learning_rate": 4.002868362641743e-07, "loss": 0.1383, "step": 15165 }, { "epoch": 0.97, "learning_rate": 3.9843274461618397e-07, "loss": 0.1164, "step": 15166 }, { "epoch": 0.97, "learning_rate": 3.9658294836099064e-07, "loss": 0.1312, "step": 15167 }, { "epoch": 0.97, "learning_rate": 3.9473744757837496e-07, "loss": 0.1226, "step": 15168 }, { "epoch": 0.97, "learning_rate": 3.9289624234790656e-07, "loss": 0.1366, "step": 15169 }, { "epoch": 0.97, "learning_rate": 3.9105933274899976e-07, "loss": 0.1432, "step": 15170 }, { "epoch": 0.97, "learning_rate": 3.892267188608689e-07, "loss": 0.1121, "step": 15171 }, { "epoch": 0.97, "learning_rate": 3.8739840076256196e-07, "loss": 0.1387, "step": 15172 }, { "epoch": 0.97, "learning_rate": 3.855743785329158e-07, "loss": 0.1349, "step": 15173 }, { "epoch": 0.97, "learning_rate": 3.837546522505786e-07, "loss": 0.1282, "step": 15174 }, { "epoch": 0.97, "learning_rate": 3.819392219940543e-07, "loss": 0.1497, "step": 15175 }, { "epoch": 0.97, "learning_rate": 3.801280878416136e-07, "loss": 0.1227, "step": 15176 }, { "epoch": 0.97, "learning_rate": 3.7832124987136063e-07, "loss": 0.156, "step": 15177 }, { "epoch": 0.97, "learning_rate": 3.7651870816123316e-07, "loss": 0.1363, "step": 15178 }, { "epoch": 0.97, "learning_rate": 3.747204627889689e-07, "loss": 0.1515, "step": 15179 }, { "epoch": 0.97, "learning_rate": 3.7292651383210584e-07, "loss": 0.1361, "step": 15180 }, { "epoch": 0.97, "learning_rate": 3.711368613680155e-07, "loss": 0.1332, "step": 15181 }, { "epoch": 0.97, "learning_rate": 3.693515054738694e-07, "loss": 0.1347, "step": 15182 }, { "epoch": 0.97, "learning_rate": 3.675704462266727e-07, "loss": 0.1163, "step": 15183 }, { "epoch": 0.97, "learning_rate": 3.6579368370323054e-07, "loss": 0.1211, "step": 15184 }, { "epoch": 0.97, "learning_rate": 3.640212179801705e-07, "loss": 0.1377, "step": 15185 }, { "epoch": 0.97, "learning_rate": 3.6225304913393153e-07, "loss": 0.1308, "step": 15186 }, { "epoch": 0.97, "learning_rate": 3.604891772407637e-07, "loss": 0.1268, "step": 15187 }, { "epoch": 0.97, "learning_rate": 3.5872960237672835e-07, "loss": 0.1212, "step": 15188 }, { "epoch": 0.97, "learning_rate": 3.569743246177204e-07, "loss": 0.125, "step": 15189 }, { "epoch": 0.97, "learning_rate": 3.552233440394348e-07, "loss": 0.1245, "step": 15190 }, { "epoch": 0.97, "learning_rate": 3.5347666071737785e-07, "loss": 0.1412, "step": 15191 }, { "epoch": 0.97, "learning_rate": 3.5173427472688924e-07, "loss": 0.1356, "step": 15192 }, { "epoch": 0.97, "learning_rate": 3.499961861430978e-07, "loss": 0.1473, "step": 15193 }, { "epoch": 0.97, "learning_rate": 3.482623950409658e-07, "loss": 0.1283, "step": 15194 }, { "epoch": 0.97, "learning_rate": 3.4653290149526674e-07, "loss": 0.1372, "step": 15195 }, { "epoch": 0.97, "learning_rate": 3.448077055805854e-07, "loss": 0.1374, "step": 15196 }, { "epoch": 0.97, "learning_rate": 3.4308680737130673e-07, "loss": 0.1329, "step": 15197 }, { "epoch": 0.97, "learning_rate": 3.4137020694166024e-07, "loss": 0.1201, "step": 15198 }, { "epoch": 0.97, "learning_rate": 3.3965790436568665e-07, "loss": 0.1266, "step": 15199 }, { "epoch": 0.97, "learning_rate": 3.379498997172159e-07, "loss": 0.1265, "step": 15200 }, { "epoch": 0.97, "learning_rate": 3.3624619306990014e-07, "loss": 0.1336, "step": 15201 }, { "epoch": 0.97, "learning_rate": 3.34546784497225e-07, "loss": 0.1562, "step": 15202 }, { "epoch": 0.97, "learning_rate": 3.328516740724763e-07, "loss": 0.13, "step": 15203 }, { "epoch": 0.97, "learning_rate": 3.3116086186875115e-07, "loss": 0.1441, "step": 15204 }, { "epoch": 0.97, "learning_rate": 3.2947434795898015e-07, "loss": 0.1274, "step": 15205 }, { "epoch": 0.97, "learning_rate": 3.277921324158828e-07, "loss": 0.1287, "step": 15206 }, { "epoch": 0.98, "learning_rate": 3.2611421531201225e-07, "loss": 0.1474, "step": 15207 }, { "epoch": 0.98, "learning_rate": 3.244405967197328e-07, "loss": 0.1327, "step": 15208 }, { "epoch": 0.98, "learning_rate": 3.22771276711209e-07, "loss": 0.1278, "step": 15209 }, { "epoch": 0.98, "learning_rate": 3.2110625535843874e-07, "loss": 0.1335, "step": 15210 }, { "epoch": 0.98, "learning_rate": 3.194455327332313e-07, "loss": 0.1211, "step": 15211 }, { "epoch": 0.98, "learning_rate": 3.1778910890719604e-07, "loss": 0.1181, "step": 15212 }, { "epoch": 0.98, "learning_rate": 3.161369839517758e-07, "loss": 0.1328, "step": 15213 }, { "epoch": 0.98, "learning_rate": 3.1448915793822474e-07, "loss": 0.1285, "step": 15214 }, { "epoch": 0.98, "learning_rate": 3.128456309375971e-07, "loss": 0.1284, "step": 15215 }, { "epoch": 0.98, "learning_rate": 3.1120640302076953e-07, "loss": 0.1299, "step": 15216 }, { "epoch": 0.98, "learning_rate": 3.09571474258441e-07, "loss": 0.1375, "step": 15217 }, { "epoch": 0.98, "learning_rate": 3.0794084472111073e-07, "loss": 0.1302, "step": 15218 }, { "epoch": 0.98, "learning_rate": 3.063145144791224e-07, "loss": 0.1463, "step": 15219 }, { "epoch": 0.98, "learning_rate": 3.046924836025866e-07, "loss": 0.1308, "step": 15220 }, { "epoch": 0.98, "learning_rate": 3.0307475216146963e-07, "loss": 0.1534, "step": 15221 }, { "epoch": 0.98, "learning_rate": 3.0146132022553785e-07, "loss": 0.1248, "step": 15222 }, { "epoch": 0.98, "learning_rate": 2.9985218786435786e-07, "loss": 0.1151, "step": 15223 }, { "epoch": 0.98, "learning_rate": 2.9824735514732974e-07, "loss": 0.1422, "step": 15224 }, { "epoch": 0.98, "learning_rate": 2.966468221436758e-07, "loss": 0.1295, "step": 15225 }, { "epoch": 0.98, "learning_rate": 2.9505058892241867e-07, "loss": 0.1394, "step": 15226 }, { "epoch": 0.98, "learning_rate": 2.934586555523811e-07, "loss": 0.1319, "step": 15227 }, { "epoch": 0.98, "learning_rate": 2.918710221022303e-07, "loss": 0.135, "step": 15228 }, { "epoch": 0.98, "learning_rate": 2.902876886404227e-07, "loss": 0.1375, "step": 15229 }, { "epoch": 0.98, "learning_rate": 2.8870865523525915e-07, "loss": 0.1345, "step": 15230 }, { "epoch": 0.98, "learning_rate": 2.8713392195481856e-07, "loss": 0.1409, "step": 15231 }, { "epoch": 0.98, "learning_rate": 2.855634888670133e-07, "loss": 0.1274, "step": 15232 }, { "epoch": 0.98, "learning_rate": 2.8399735603957814e-07, "loss": 0.1095, "step": 15233 }, { "epoch": 0.98, "learning_rate": 2.824355235400589e-07, "loss": 0.1347, "step": 15234 }, { "epoch": 0.98, "learning_rate": 2.808779914358017e-07, "loss": 0.1477, "step": 15235 }, { "epoch": 0.98, "learning_rate": 2.793247597939752e-07, "loss": 0.1397, "step": 15236 }, { "epoch": 0.98, "learning_rate": 2.77775828681559e-07, "loss": 0.1281, "step": 15237 }, { "epoch": 0.98, "learning_rate": 2.762311981653554e-07, "loss": 0.1253, "step": 15238 }, { "epoch": 0.98, "learning_rate": 2.7469086831198866e-07, "loss": 0.1243, "step": 15239 }, { "epoch": 0.98, "learning_rate": 2.731548391878835e-07, "loss": 0.141, "step": 15240 }, { "epoch": 0.98, "learning_rate": 2.7162311085927596e-07, "loss": 0.1318, "step": 15241 }, { "epoch": 0.98, "learning_rate": 2.7009568339222414e-07, "loss": 0.1345, "step": 15242 }, { "epoch": 0.98, "learning_rate": 2.685725568525976e-07, "loss": 0.1249, "step": 15243 }, { "epoch": 0.98, "learning_rate": 2.6705373130608836e-07, "loss": 0.147, "step": 15244 }, { "epoch": 0.98, "learning_rate": 2.655392068181883e-07, "loss": 0.1387, "step": 15245 }, { "epoch": 0.98, "learning_rate": 2.640289834542231e-07, "loss": 0.1418, "step": 15246 }, { "epoch": 0.98, "learning_rate": 2.6252306127931836e-07, "loss": 0.1328, "step": 15247 }, { "epoch": 0.98, "learning_rate": 2.610214403584221e-07, "loss": 0.1416, "step": 15248 }, { "epoch": 0.98, "learning_rate": 2.595241207562715e-07, "loss": 0.1258, "step": 15249 }, { "epoch": 0.98, "learning_rate": 2.580311025374704e-07, "loss": 0.136, "step": 15250 }, { "epoch": 0.98, "learning_rate": 2.5654238576637843e-07, "loss": 0.1247, "step": 15251 }, { "epoch": 0.98, "learning_rate": 2.550579705072109e-07, "loss": 0.1357, "step": 15252 }, { "epoch": 0.98, "learning_rate": 2.5357785682399437e-07, "loss": 0.1372, "step": 15253 }, { "epoch": 0.98, "learning_rate": 2.5210204478054445e-07, "loss": 0.1544, "step": 15254 }, { "epoch": 0.98, "learning_rate": 2.506305344405213e-07, "loss": 0.1309, "step": 15255 }, { "epoch": 0.98, "learning_rate": 2.4916332586736315e-07, "loss": 0.1243, "step": 15256 }, { "epoch": 0.98, "learning_rate": 2.477004191243526e-07, "loss": 0.1263, "step": 15257 }, { "epoch": 0.98, "learning_rate": 2.462418142745948e-07, "loss": 0.1306, "step": 15258 }, { "epoch": 0.98, "learning_rate": 2.447875113809728e-07, "loss": 0.1451, "step": 15259 }, { "epoch": 0.98, "learning_rate": 2.4333751050621413e-07, "loss": 0.1426, "step": 15260 }, { "epoch": 0.98, "learning_rate": 2.418918117128577e-07, "loss": 0.1346, "step": 15261 }, { "epoch": 0.98, "learning_rate": 2.4045041506324253e-07, "loss": 0.1404, "step": 15262 }, { "epoch": 0.98, "learning_rate": 2.390133206195189e-07, "loss": 0.145, "step": 15263 }, { "epoch": 0.98, "learning_rate": 2.3758052844368162e-07, "loss": 0.121, "step": 15264 }, { "epoch": 0.98, "learning_rate": 2.3615203859751466e-07, "loss": 0.1235, "step": 15265 }, { "epoch": 0.98, "learning_rate": 2.347278511426132e-07, "loss": 0.1397, "step": 15266 }, { "epoch": 0.98, "learning_rate": 2.3330796614040585e-07, "loss": 0.1346, "step": 15267 }, { "epoch": 0.98, "learning_rate": 2.318923836521325e-07, "loss": 0.1363, "step": 15268 }, { "epoch": 0.98, "learning_rate": 2.3048110373883324e-07, "loss": 0.1293, "step": 15269 }, { "epoch": 0.98, "learning_rate": 2.290741264613594e-07, "loss": 0.1316, "step": 15270 }, { "epoch": 0.98, "learning_rate": 2.2767145188040684e-07, "loss": 0.1256, "step": 15271 }, { "epoch": 0.98, "learning_rate": 2.2627308005644943e-07, "loss": 0.1388, "step": 15272 }, { "epoch": 0.98, "learning_rate": 2.2487901104981668e-07, "loss": 0.1107, "step": 15273 }, { "epoch": 0.98, "learning_rate": 2.2348924492060497e-07, "loss": 0.131, "step": 15274 }, { "epoch": 0.98, "learning_rate": 2.2210378172876633e-07, "loss": 0.1296, "step": 15275 }, { "epoch": 0.98, "learning_rate": 2.207226215340419e-07, "loss": 0.121, "step": 15276 }, { "epoch": 0.98, "learning_rate": 2.1934576439599508e-07, "loss": 0.1343, "step": 15277 }, { "epoch": 0.98, "learning_rate": 2.1797321037400066e-07, "loss": 0.1457, "step": 15278 }, { "epoch": 0.98, "learning_rate": 2.1660495952726677e-07, "loss": 0.1368, "step": 15279 }, { "epoch": 0.98, "learning_rate": 2.1524101191477962e-07, "loss": 0.1436, "step": 15280 }, { "epoch": 0.98, "learning_rate": 2.1388136759536992e-07, "loss": 0.1394, "step": 15281 }, { "epoch": 0.98, "learning_rate": 2.1252602662767963e-07, "loss": 0.1392, "step": 15282 }, { "epoch": 0.98, "learning_rate": 2.1117498907013978e-07, "loss": 0.1344, "step": 15283 }, { "epoch": 0.98, "learning_rate": 2.098282549810371e-07, "loss": 0.1276, "step": 15284 }, { "epoch": 0.98, "learning_rate": 2.0848582441843624e-07, "loss": 0.1327, "step": 15285 }, { "epoch": 0.98, "learning_rate": 2.0714769744023532e-07, "loss": 0.1301, "step": 15286 }, { "epoch": 0.98, "learning_rate": 2.0581387410414377e-07, "loss": 0.1327, "step": 15287 }, { "epoch": 0.98, "learning_rate": 2.044843544676822e-07, "loss": 0.1351, "step": 15288 }, { "epoch": 0.98, "learning_rate": 2.031591385881826e-07, "loss": 0.1295, "step": 15289 }, { "epoch": 0.98, "learning_rate": 2.0183822652281026e-07, "loss": 0.143, "step": 15290 }, { "epoch": 0.98, "learning_rate": 2.0052161832850856e-07, "loss": 0.1452, "step": 15291 }, { "epoch": 0.98, "learning_rate": 1.9920931406207655e-07, "loss": 0.1286, "step": 15292 }, { "epoch": 0.98, "learning_rate": 1.9790131378010225e-07, "loss": 0.1307, "step": 15293 }, { "epoch": 0.98, "learning_rate": 1.9659761753898497e-07, "loss": 0.1402, "step": 15294 }, { "epoch": 0.98, "learning_rate": 1.9529822539495757e-07, "loss": 0.1467, "step": 15295 }, { "epoch": 0.98, "learning_rate": 1.9400313740405297e-07, "loss": 0.1353, "step": 15296 }, { "epoch": 0.98, "learning_rate": 1.9271235362213757e-07, "loss": 0.1226, "step": 15297 }, { "epoch": 0.98, "learning_rate": 1.9142587410484468e-07, "loss": 0.1241, "step": 15298 }, { "epoch": 0.98, "learning_rate": 1.9014369890769657e-07, "loss": 0.1173, "step": 15299 }, { "epoch": 0.98, "learning_rate": 1.88865828085949e-07, "loss": 0.1511, "step": 15300 }, { "epoch": 0.98, "learning_rate": 1.8759226169473565e-07, "loss": 0.1395, "step": 15301 }, { "epoch": 0.98, "learning_rate": 1.8632299978896817e-07, "loss": 0.1265, "step": 15302 }, { "epoch": 0.98, "learning_rate": 1.8505804242340274e-07, "loss": 0.129, "step": 15303 }, { "epoch": 0.98, "learning_rate": 1.8379738965256244e-07, "loss": 0.1502, "step": 15304 }, { "epoch": 0.98, "learning_rate": 1.8254104153083707e-07, "loss": 0.1251, "step": 15305 }, { "epoch": 0.98, "learning_rate": 1.812889981123944e-07, "loss": 0.133, "step": 15306 }, { "epoch": 0.98, "learning_rate": 1.8004125945123574e-07, "loss": 0.1357, "step": 15307 }, { "epoch": 0.98, "learning_rate": 1.7879782560117352e-07, "loss": 0.1551, "step": 15308 }, { "epoch": 0.98, "learning_rate": 1.7755869661582047e-07, "loss": 0.1546, "step": 15309 }, { "epoch": 0.98, "learning_rate": 1.763238725486338e-07, "loss": 0.1306, "step": 15310 }, { "epoch": 0.98, "learning_rate": 1.7509335345284872e-07, "loss": 0.1223, "step": 15311 }, { "epoch": 0.98, "learning_rate": 1.7386713938154496e-07, "loss": 0.1258, "step": 15312 }, { "epoch": 0.98, "learning_rate": 1.726452303875914e-07, "loss": 0.1278, "step": 15313 }, { "epoch": 0.98, "learning_rate": 1.7142762652369025e-07, "loss": 0.1409, "step": 15314 }, { "epoch": 0.98, "learning_rate": 1.7021432784235514e-07, "loss": 0.117, "step": 15315 }, { "epoch": 0.98, "learning_rate": 1.6900533439591081e-07, "loss": 0.1295, "step": 15316 }, { "epoch": 0.98, "learning_rate": 1.6780064623648228e-07, "loss": 0.1371, "step": 15317 }, { "epoch": 0.98, "learning_rate": 1.6660026341603908e-07, "loss": 0.1438, "step": 15318 }, { "epoch": 0.98, "learning_rate": 1.6540418598633977e-07, "loss": 0.1246, "step": 15319 }, { "epoch": 0.98, "learning_rate": 1.6421241399897647e-07, "loss": 0.1402, "step": 15320 }, { "epoch": 0.98, "learning_rate": 1.6302494750533026e-07, "loss": 0.1216, "step": 15321 }, { "epoch": 0.98, "learning_rate": 1.6184178655661576e-07, "loss": 0.1284, "step": 15322 }, { "epoch": 0.98, "learning_rate": 1.6066293120385878e-07, "loss": 0.134, "step": 15323 }, { "epoch": 0.98, "learning_rate": 1.5948838149790758e-07, "loss": 0.1381, "step": 15324 }, { "epoch": 0.98, "learning_rate": 1.5831813748939938e-07, "loss": 0.1429, "step": 15325 }, { "epoch": 0.98, "learning_rate": 1.571521992288161e-07, "loss": 0.1465, "step": 15326 }, { "epoch": 0.98, "learning_rate": 1.5599056676642855e-07, "loss": 0.1373, "step": 15327 }, { "epoch": 0.98, "learning_rate": 1.5483324015234114e-07, "loss": 0.1358, "step": 15328 }, { "epoch": 0.98, "learning_rate": 1.536802194364695e-07, "loss": 0.1449, "step": 15329 }, { "epoch": 0.98, "learning_rate": 1.525315046685294e-07, "loss": 0.1582, "step": 15330 }, { "epoch": 0.98, "learning_rate": 1.513870958980479e-07, "loss": 0.1296, "step": 15331 }, { "epoch": 0.98, "learning_rate": 1.502469931743966e-07, "loss": 0.1419, "step": 15332 }, { "epoch": 0.98, "learning_rate": 1.4911119654674731e-07, "loss": 0.1332, "step": 15333 }, { "epoch": 0.98, "learning_rate": 1.4797970606407198e-07, "loss": 0.1104, "step": 15334 }, { "epoch": 0.98, "learning_rate": 1.4685252177516483e-07, "loss": 0.1211, "step": 15335 }, { "epoch": 0.98, "learning_rate": 1.457296437286315e-07, "loss": 0.1306, "step": 15336 }, { "epoch": 0.98, "learning_rate": 1.4461107197292212e-07, "loss": 0.1177, "step": 15337 }, { "epoch": 0.98, "learning_rate": 1.4349680655624253e-07, "loss": 0.1361, "step": 15338 }, { "epoch": 0.98, "learning_rate": 1.4238684752667653e-07, "loss": 0.1482, "step": 15339 }, { "epoch": 0.98, "learning_rate": 1.412811949320747e-07, "loss": 0.1403, "step": 15340 }, { "epoch": 0.98, "learning_rate": 1.4017984882012113e-07, "loss": 0.1262, "step": 15341 }, { "epoch": 0.98, "learning_rate": 1.3908280923831118e-07, "loss": 0.1388, "step": 15342 }, { "epoch": 0.98, "learning_rate": 1.3799007623396254e-07, "loss": 0.133, "step": 15343 }, { "epoch": 0.98, "learning_rate": 1.3690164985419306e-07, "loss": 0.1437, "step": 15344 }, { "epoch": 0.98, "learning_rate": 1.3581753014595412e-07, "loss": 0.1317, "step": 15345 }, { "epoch": 0.98, "learning_rate": 1.3473771715598604e-07, "loss": 0.131, "step": 15346 }, { "epoch": 0.98, "learning_rate": 1.3366221093085164e-07, "loss": 0.1288, "step": 15347 }, { "epoch": 0.98, "learning_rate": 1.3259101151694708e-07, "loss": 0.1449, "step": 15348 }, { "epoch": 0.98, "learning_rate": 1.315241189604688e-07, "loss": 0.1247, "step": 15349 }, { "epoch": 0.98, "learning_rate": 1.304615333074133e-07, "loss": 0.1231, "step": 15350 }, { "epoch": 0.98, "learning_rate": 1.2940325460362168e-07, "loss": 0.1402, "step": 15351 }, { "epoch": 0.98, "learning_rate": 1.2834928289472416e-07, "loss": 0.1294, "step": 15352 }, { "epoch": 0.98, "learning_rate": 1.272996182261732e-07, "loss": 0.1409, "step": 15353 }, { "epoch": 0.98, "learning_rate": 1.2625426064323264e-07, "loss": 0.1391, "step": 15354 }, { "epoch": 0.98, "learning_rate": 1.2521321019099975e-07, "loss": 0.1334, "step": 15355 }, { "epoch": 0.98, "learning_rate": 1.241764669143497e-07, "loss": 0.1379, "step": 15356 }, { "epoch": 0.98, "learning_rate": 1.2314403085801342e-07, "loss": 0.1422, "step": 15357 }, { "epoch": 0.98, "learning_rate": 1.221159020664997e-07, "loss": 0.1466, "step": 15358 }, { "epoch": 0.98, "learning_rate": 1.21092080584162e-07, "loss": 0.1327, "step": 15359 }, { "epoch": 0.98, "learning_rate": 1.2007256645514276e-07, "loss": 0.1374, "step": 15360 }, { "epoch": 0.98, "learning_rate": 1.1905735972340682e-07, "loss": 0.1251, "step": 15361 }, { "epoch": 0.98, "learning_rate": 1.1804646043274136e-07, "loss": 0.1326, "step": 15362 }, { "epoch": 0.99, "learning_rate": 1.1703986862674487e-07, "loss": 0.1459, "step": 15363 }, { "epoch": 0.99, "learning_rate": 1.1603758434882705e-07, "loss": 0.1298, "step": 15364 }, { "epoch": 0.99, "learning_rate": 1.1503960764220889e-07, "loss": 0.1337, "step": 15365 }, { "epoch": 0.99, "learning_rate": 1.1404593854992263e-07, "loss": 0.1333, "step": 15366 }, { "epoch": 0.99, "learning_rate": 1.1305657711482287e-07, "loss": 0.1423, "step": 15367 }, { "epoch": 0.99, "learning_rate": 1.120715233795977e-07, "loss": 0.1462, "step": 15368 }, { "epoch": 0.99, "learning_rate": 1.1109077738670204e-07, "loss": 0.1328, "step": 15369 }, { "epoch": 0.99, "learning_rate": 1.1011433917843539e-07, "loss": 0.1259, "step": 15370 }, { "epoch": 0.99, "learning_rate": 1.091422087969196e-07, "loss": 0.1355, "step": 15371 }, { "epoch": 0.99, "learning_rate": 1.0817438628407673e-07, "loss": 0.1326, "step": 15372 }, { "epoch": 0.99, "learning_rate": 1.0721087168162891e-07, "loss": 0.1515, "step": 15373 }, { "epoch": 0.99, "learning_rate": 1.0625166503114292e-07, "loss": 0.1438, "step": 15374 }, { "epoch": 0.99, "learning_rate": 1.0529676637398567e-07, "loss": 0.1328, "step": 15375 }, { "epoch": 0.99, "learning_rate": 1.0434617575133532e-07, "loss": 0.1406, "step": 15376 }, { "epoch": 0.99, "learning_rate": 1.0339989320418131e-07, "loss": 0.1254, "step": 15377 }, { "epoch": 0.99, "learning_rate": 1.0245791877334654e-07, "loss": 0.1316, "step": 15378 }, { "epoch": 0.99, "learning_rate": 1.0152025249943187e-07, "loss": 0.1427, "step": 15379 }, { "epoch": 0.99, "learning_rate": 1.0058689442289382e-07, "loss": 0.1363, "step": 15380 }, { "epoch": 0.99, "learning_rate": 9.965784458397798e-08, "loss": 0.1259, "step": 15381 }, { "epoch": 0.99, "learning_rate": 9.873310302275229e-08, "loss": 0.1364, "step": 15382 }, { "epoch": 0.99, "learning_rate": 9.781266977909598e-08, "loss": 0.1546, "step": 15383 }, { "epoch": 0.99, "learning_rate": 9.689654489269951e-08, "loss": 0.1441, "step": 15384 }, { "epoch": 0.99, "learning_rate": 9.59847284030757e-08, "loss": 0.1381, "step": 15385 }, { "epoch": 0.99, "learning_rate": 9.507722034954869e-08, "loss": 0.1499, "step": 15386 }, { "epoch": 0.99, "learning_rate": 9.41740207712538e-08, "loss": 0.1282, "step": 15387 }, { "epoch": 0.99, "learning_rate": 9.32751297071377e-08, "loss": 0.1393, "step": 15388 }, { "epoch": 0.99, "learning_rate": 9.238054719596934e-08, "loss": 0.126, "step": 15389 }, { "epoch": 0.99, "learning_rate": 9.149027327632898e-08, "loss": 0.1294, "step": 15390 }, { "epoch": 0.99, "learning_rate": 9.060430798660813e-08, "loss": 0.1184, "step": 15391 }, { "epoch": 0.99, "learning_rate": 8.972265136502068e-08, "loss": 0.1275, "step": 15392 }, { "epoch": 0.99, "learning_rate": 8.884530344958064e-08, "loss": 0.1107, "step": 15393 }, { "epoch": 0.99, "learning_rate": 8.797226427813554e-08, "loss": 0.1503, "step": 15394 }, { "epoch": 0.99, "learning_rate": 8.710353388832194e-08, "loss": 0.1473, "step": 15395 }, { "epoch": 0.99, "learning_rate": 8.623911231760984e-08, "loss": 0.143, "step": 15396 }, { "epoch": 0.99, "learning_rate": 8.537899960328055e-08, "loss": 0.1321, "step": 15397 }, { "epoch": 0.99, "learning_rate": 8.452319578242662e-08, "loss": 0.1384, "step": 15398 }, { "epoch": 0.99, "learning_rate": 8.367170089195187e-08, "loss": 0.1351, "step": 15399 }, { "epoch": 0.99, "learning_rate": 8.282451496859356e-08, "loss": 0.1551, "step": 15400 }, { "epoch": 0.99, "learning_rate": 8.198163804885584e-08, "loss": 0.1204, "step": 15401 }, { "epoch": 0.99, "learning_rate": 8.11430701691207e-08, "loss": 0.1315, "step": 15402 }, { "epoch": 0.99, "learning_rate": 8.030881136552592e-08, "loss": 0.1311, "step": 15403 }, { "epoch": 0.99, "learning_rate": 7.947886167406493e-08, "loss": 0.1265, "step": 15404 }, { "epoch": 0.99, "learning_rate": 7.86532211305313e-08, "loss": 0.1469, "step": 15405 }, { "epoch": 0.99, "learning_rate": 7.78318897705188e-08, "loss": 0.1238, "step": 15406 }, { "epoch": 0.99, "learning_rate": 7.701486762945464e-08, "loss": 0.1217, "step": 15407 }, { "epoch": 0.99, "learning_rate": 7.620215474258841e-08, "loss": 0.1246, "step": 15408 }, { "epoch": 0.99, "learning_rate": 7.539375114493653e-08, "loss": 0.1467, "step": 15409 }, { "epoch": 0.99, "learning_rate": 7.45896568713933e-08, "loss": 0.1258, "step": 15410 }, { "epoch": 0.99, "learning_rate": 7.378987195661991e-08, "loss": 0.141, "step": 15411 }, { "epoch": 0.99, "learning_rate": 7.299439643511097e-08, "loss": 0.1366, "step": 15412 }, { "epoch": 0.99, "learning_rate": 7.220323034117238e-08, "loss": 0.1198, "step": 15413 }, { "epoch": 0.99, "learning_rate": 7.141637370892129e-08, "loss": 0.1367, "step": 15414 }, { "epoch": 0.99, "learning_rate": 7.063382657229723e-08, "loss": 0.1458, "step": 15415 }, { "epoch": 0.99, "learning_rate": 6.985558896503986e-08, "loss": 0.1273, "step": 15416 }, { "epoch": 0.99, "learning_rate": 6.908166092072232e-08, "loss": 0.1355, "step": 15417 }, { "epoch": 0.99, "learning_rate": 6.831204247270684e-08, "loss": 0.1265, "step": 15418 }, { "epoch": 0.99, "learning_rate": 6.754673365418906e-08, "loss": 0.1365, "step": 15419 }, { "epoch": 0.99, "learning_rate": 6.678573449818703e-08, "loss": 0.127, "step": 15420 }, { "epoch": 0.99, "learning_rate": 6.602904503749674e-08, "loss": 0.1424, "step": 15421 }, { "epoch": 0.99, "learning_rate": 6.527666530476984e-08, "loss": 0.1375, "step": 15422 }, { "epoch": 0.99, "learning_rate": 6.452859533244703e-08, "loss": 0.1278, "step": 15423 }, { "epoch": 0.99, "learning_rate": 6.378483515278033e-08, "loss": 0.139, "step": 15424 }, { "epoch": 0.99, "learning_rate": 6.304538479785515e-08, "loss": 0.1303, "step": 15425 }, { "epoch": 0.99, "learning_rate": 6.231024429956822e-08, "loss": 0.1374, "step": 15426 }, { "epoch": 0.99, "learning_rate": 6.15794136896053e-08, "loss": 0.1331, "step": 15427 }, { "epoch": 0.99, "learning_rate": 6.08528929994856e-08, "loss": 0.1297, "step": 15428 }, { "epoch": 0.99, "learning_rate": 6.013068226055074e-08, "loss": 0.1338, "step": 15429 }, { "epoch": 0.99, "learning_rate": 5.941278150395357e-08, "loss": 0.1224, "step": 15430 }, { "epoch": 0.99, "learning_rate": 5.869919076063601e-08, "loss": 0.1495, "step": 15431 }, { "epoch": 0.99, "learning_rate": 5.798991006137344e-08, "loss": 0.1244, "step": 15432 }, { "epoch": 0.99, "learning_rate": 5.7284939436774707e-08, "loss": 0.1433, "step": 15433 }, { "epoch": 0.99, "learning_rate": 5.658427891721552e-08, "loss": 0.1387, "step": 15434 }, { "epoch": 0.99, "learning_rate": 5.588792853292724e-08, "loss": 0.1354, "step": 15435 }, { "epoch": 0.99, "learning_rate": 5.5195888313941404e-08, "loss": 0.1412, "step": 15436 }, { "epoch": 0.99, "learning_rate": 5.450815829010081e-08, "loss": 0.1334, "step": 15437 }, { "epoch": 0.99, "learning_rate": 5.382473849105951e-08, "loss": 0.1462, "step": 15438 }, { "epoch": 0.99, "learning_rate": 5.314562894629394e-08, "loss": 0.1394, "step": 15439 }, { "epoch": 0.99, "learning_rate": 5.247082968510286e-08, "loss": 0.1275, "step": 15440 }, { "epoch": 0.99, "learning_rate": 5.180034073656303e-08, "loss": 0.1168, "step": 15441 }, { "epoch": 0.99, "learning_rate": 5.1134162129606866e-08, "loss": 0.1251, "step": 15442 }, { "epoch": 0.99, "learning_rate": 5.0472293892955825e-08, "loss": 0.1195, "step": 15443 }, { "epoch": 0.99, "learning_rate": 4.981473605516485e-08, "loss": 0.1447, "step": 15444 }, { "epoch": 0.99, "learning_rate": 4.916148864457793e-08, "loss": 0.1377, "step": 15445 }, { "epoch": 0.99, "learning_rate": 4.851255168937252e-08, "loss": 0.141, "step": 15446 }, { "epoch": 0.99, "learning_rate": 4.786792521753736e-08, "loss": 0.1274, "step": 15447 }, { "epoch": 0.99, "learning_rate": 4.722760925687242e-08, "loss": 0.1415, "step": 15448 }, { "epoch": 0.99, "learning_rate": 4.659160383497785e-08, "loss": 0.1329, "step": 15449 }, { "epoch": 0.99, "learning_rate": 4.5959908979309463e-08, "loss": 0.1227, "step": 15450 }, { "epoch": 0.99, "learning_rate": 4.533252471708993e-08, "loss": 0.1245, "step": 15451 }, { "epoch": 0.99, "learning_rate": 4.4709451075375386e-08, "loss": 0.1233, "step": 15452 }, { "epoch": 0.99, "learning_rate": 4.409068808103323e-08, "loss": 0.1307, "step": 15453 }, { "epoch": 0.99, "learning_rate": 4.347623576076432e-08, "loss": 0.1425, "step": 15454 }, { "epoch": 0.99, "learning_rate": 4.286609414105858e-08, "loss": 0.1476, "step": 15455 }, { "epoch": 0.99, "learning_rate": 4.2260263248228296e-08, "loss": 0.1206, "step": 15456 }, { "epoch": 0.99, "learning_rate": 4.1658743108397013e-08, "loss": 0.1298, "step": 15457 }, { "epoch": 0.99, "learning_rate": 4.106153374751065e-08, "loss": 0.1303, "step": 15458 }, { "epoch": 0.99, "learning_rate": 4.0468635191326375e-08, "loss": 0.1398, "step": 15459 }, { "epoch": 0.99, "learning_rate": 3.988004746540153e-08, "loss": 0.1528, "step": 15460 }, { "epoch": 0.99, "learning_rate": 3.9295770595138004e-08, "loss": 0.1214, "step": 15461 }, { "epoch": 0.99, "learning_rate": 3.871580460571567e-08, "loss": 0.1155, "step": 15462 }, { "epoch": 0.99, "learning_rate": 3.814014952215894e-08, "loss": 0.1539, "step": 15463 }, { "epoch": 0.99, "learning_rate": 3.756880536928131e-08, "loss": 0.1212, "step": 15464 }, { "epoch": 0.99, "learning_rate": 3.7001772171740836e-08, "loss": 0.148, "step": 15465 }, { "epoch": 0.99, "learning_rate": 3.643904995396241e-08, "loss": 0.1342, "step": 15466 }, { "epoch": 0.99, "learning_rate": 3.5880638740248826e-08, "loss": 0.1301, "step": 15467 }, { "epoch": 0.99, "learning_rate": 3.532653855464752e-08, "loss": 0.1342, "step": 15468 }, { "epoch": 0.99, "learning_rate": 3.477674942108378e-08, "loss": 0.133, "step": 15469 }, { "epoch": 0.99, "learning_rate": 3.4231271363249776e-08, "loss": 0.1422, "step": 15470 }, { "epoch": 0.99, "learning_rate": 3.369010440467113e-08, "loss": 0.1327, "step": 15471 }, { "epoch": 0.99, "learning_rate": 3.3153248568695835e-08, "loss": 0.1211, "step": 15472 }, { "epoch": 0.99, "learning_rate": 3.2620703878460946e-08, "loss": 0.13, "step": 15473 }, { "epoch": 0.99, "learning_rate": 3.2092470356948066e-08, "loss": 0.1336, "step": 15474 }, { "epoch": 0.99, "learning_rate": 3.1568548026927877e-08, "loss": 0.128, "step": 15475 }, { "epoch": 0.99, "learning_rate": 3.1048936911004523e-08, "loss": 0.1289, "step": 15476 }, { "epoch": 0.99, "learning_rate": 3.05336370315823e-08, "loss": 0.1348, "step": 15477 }, { "epoch": 0.99, "learning_rate": 3.002264841087676e-08, "loss": 0.131, "step": 15478 }, { "epoch": 0.99, "learning_rate": 2.9515971070925852e-08, "loss": 0.1424, "step": 15479 }, { "epoch": 0.99, "learning_rate": 2.9013605033589853e-08, "loss": 0.1477, "step": 15480 }, { "epoch": 0.99, "learning_rate": 2.8515550320529216e-08, "loss": 0.1352, "step": 15481 }, { "epoch": 0.99, "learning_rate": 2.802180695322676e-08, "loss": 0.1301, "step": 15482 }, { "epoch": 0.99, "learning_rate": 2.7532374952965457e-08, "loss": 0.1179, "step": 15483 }, { "epoch": 0.99, "learning_rate": 2.7047254340850646e-08, "loss": 0.1212, "step": 15484 }, { "epoch": 0.99, "learning_rate": 2.656644513782114e-08, "loss": 0.1258, "step": 15485 }, { "epoch": 0.99, "learning_rate": 2.6089947364604793e-08, "loss": 0.1465, "step": 15486 }, { "epoch": 0.99, "learning_rate": 2.561776104174074e-08, "loss": 0.1398, "step": 15487 }, { "epoch": 0.99, "learning_rate": 2.514988618960157e-08, "loss": 0.1512, "step": 15488 }, { "epoch": 0.99, "learning_rate": 2.468632282836003e-08, "loss": 0.1505, "step": 15489 }, { "epoch": 0.99, "learning_rate": 2.4227070978011245e-08, "loss": 0.1394, "step": 15490 }, { "epoch": 0.99, "learning_rate": 2.3772130658372695e-08, "loss": 0.1232, "step": 15491 }, { "epoch": 0.99, "learning_rate": 2.332150188903981e-08, "loss": 0.1323, "step": 15492 }, { "epoch": 0.99, "learning_rate": 2.28751846894526e-08, "loss": 0.1335, "step": 15493 }, { "epoch": 0.99, "learning_rate": 2.2433179078873436e-08, "loss": 0.119, "step": 15494 }, { "epoch": 0.99, "learning_rate": 2.199548507635374e-08, "loss": 0.1425, "step": 15495 }, { "epoch": 0.99, "learning_rate": 2.1562102700767305e-08, "loss": 0.1358, "step": 15496 }, { "epoch": 0.99, "learning_rate": 2.113303197081029e-08, "loss": 0.1249, "step": 15497 }, { "epoch": 0.99, "learning_rate": 2.070827290497901e-08, "loss": 0.1366, "step": 15498 }, { "epoch": 0.99, "learning_rate": 2.0287825521603243e-08, "loss": 0.1369, "step": 15499 }, { "epoch": 0.99, "learning_rate": 1.9871689838801833e-08, "loss": 0.1364, "step": 15500 }, { "epoch": 0.99, "learning_rate": 1.9459865874527084e-08, "loss": 0.1266, "step": 15501 }, { "epoch": 0.99, "learning_rate": 1.9052353646542566e-08, "loss": 0.1393, "step": 15502 }, { "epoch": 0.99, "learning_rate": 1.8649153172423105e-08, "loss": 0.1194, "step": 15503 }, { "epoch": 0.99, "learning_rate": 1.8250264469543697e-08, "loss": 0.1317, "step": 15504 }, { "epoch": 0.99, "learning_rate": 1.78556875551239e-08, "loss": 0.1205, "step": 15505 }, { "epoch": 0.99, "learning_rate": 1.7465422446161228e-08, "loss": 0.1276, "step": 15506 }, { "epoch": 0.99, "learning_rate": 1.7079469159508865e-08, "loss": 0.1364, "step": 15507 }, { "epoch": 0.99, "learning_rate": 1.6697827711797953e-08, "loss": 0.1609, "step": 15508 }, { "epoch": 0.99, "learning_rate": 1.6320498119481998e-08, "loss": 0.1399, "step": 15509 }, { "epoch": 0.99, "learning_rate": 1.594748039884797e-08, "loss": 0.1301, "step": 15510 }, { "epoch": 0.99, "learning_rate": 1.5578774565971898e-08, "loss": 0.1392, "step": 15511 }, { "epoch": 0.99, "learning_rate": 1.521438063676328e-08, "loss": 0.1285, "step": 15512 }, { "epoch": 0.99, "learning_rate": 1.485429862692067e-08, "loss": 0.1667, "step": 15513 }, { "epoch": 0.99, "learning_rate": 1.4498528551998292e-08, "loss": 0.1456, "step": 15514 }, { "epoch": 0.99, "learning_rate": 1.4147070427317221e-08, "loss": 0.1386, "step": 15515 }, { "epoch": 0.99, "learning_rate": 1.3799924268032006e-08, "loss": 0.1411, "step": 15516 }, { "epoch": 0.99, "learning_rate": 1.3457090089130653e-08, "loss": 0.1292, "step": 15517 }, { "epoch": 0.99, "learning_rate": 1.3118567905390233e-08, "loss": 0.1344, "step": 15518 }, { "epoch": 1.0, "learning_rate": 1.2784357731410179e-08, "loss": 0.1213, "step": 15519 }, { "epoch": 1.0, "learning_rate": 1.2454459581612287e-08, "loss": 0.1548, "step": 15520 }, { "epoch": 1.0, "learning_rate": 1.2128873470196312e-08, "loss": 0.1279, "step": 15521 }, { "epoch": 1.0, "learning_rate": 1.1807599411239878e-08, "loss": 0.1393, "step": 15522 }, { "epoch": 1.0, "learning_rate": 1.1490637418565264e-08, "loss": 0.1353, "step": 15523 }, { "epoch": 1.0, "learning_rate": 1.117798750586152e-08, "loss": 0.1371, "step": 15524 }, { "epoch": 1.0, "learning_rate": 1.0869649686617855e-08, "loss": 0.1354, "step": 15525 }, { "epoch": 1.0, "learning_rate": 1.0565623974101434e-08, "loss": 0.1316, "step": 15526 }, { "epoch": 1.0, "learning_rate": 1.02659103814573e-08, "loss": 0.1241, "step": 15527 }, { "epoch": 1.0, "learning_rate": 9.970508921597343e-09, "loss": 0.1411, "step": 15528 }, { "epoch": 1.0, "learning_rate": 9.679419607244722e-09, "loss": 0.1564, "step": 15529 }, { "epoch": 1.0, "learning_rate": 9.392642450978261e-09, "loss": 0.143, "step": 15530 }, { "epoch": 1.0, "learning_rate": 9.110177465154746e-09, "loss": 0.1405, "step": 15531 }, { "epoch": 1.0, "learning_rate": 8.832024661964423e-09, "loss": 0.1228, "step": 15532 }, { "epoch": 1.0, "learning_rate": 8.558184053386598e-09, "loss": 0.1264, "step": 15533 }, { "epoch": 1.0, "learning_rate": 8.288655651234045e-09, "loss": 0.131, "step": 15534 }, { "epoch": 1.0, "learning_rate": 8.023439467141902e-09, "loss": 0.1379, "step": 15535 }, { "epoch": 1.0, "learning_rate": 7.762535512534363e-09, "loss": 0.1485, "step": 15536 }, { "epoch": 1.0, "learning_rate": 7.505943798680193e-09, "loss": 0.1251, "step": 15537 }, { "epoch": 1.0, "learning_rate": 7.25366433662611e-09, "loss": 0.129, "step": 15538 }, { "epoch": 1.0, "learning_rate": 7.005697137263401e-09, "loss": 0.1415, "step": 15539 }, { "epoch": 1.0, "learning_rate": 6.762042211283514e-09, "loss": 0.1312, "step": 15540 }, { "epoch": 1.0, "learning_rate": 6.522699569200263e-09, "loss": 0.1336, "step": 15541 }, { "epoch": 1.0, "learning_rate": 6.2876692213165166e-09, "loss": 0.1417, "step": 15542 }, { "epoch": 1.0, "learning_rate": 6.056951177790815e-09, "loss": 0.1412, "step": 15543 }, { "epoch": 1.0, "learning_rate": 5.8305454485596545e-09, "loss": 0.1312, "step": 15544 }, { "epoch": 1.0, "learning_rate": 5.608452043392998e-09, "loss": 0.1373, "step": 15545 }, { "epoch": 1.0, "learning_rate": 5.390670971860967e-09, "loss": 0.1255, "step": 15546 }, { "epoch": 1.0, "learning_rate": 5.1772022433560495e-09, "loss": 0.1553, "step": 15547 }, { "epoch": 1.0, "learning_rate": 4.968045867104198e-09, "loss": 0.1178, "step": 15548 }, { "epoch": 1.0, "learning_rate": 4.7632018520982206e-09, "loss": 0.1289, "step": 15549 }, { "epoch": 1.0, "learning_rate": 4.562670207186592e-09, "loss": 0.1214, "step": 15550 }, { "epoch": 1.0, "learning_rate": 4.36645094100685e-09, "loss": 0.1433, "step": 15551 }, { "epoch": 1.0, "learning_rate": 4.1745440620410965e-09, "loss": 0.1171, "step": 15552 }, { "epoch": 1.0, "learning_rate": 3.986949578549393e-09, "loss": 0.1395, "step": 15553 }, { "epoch": 1.0, "learning_rate": 3.803667498625263e-09, "loss": 0.1298, "step": 15554 }, { "epoch": 1.0, "learning_rate": 3.624697830173496e-09, "loss": 0.1197, "step": 15555 }, { "epoch": 1.0, "learning_rate": 3.450040580910141e-09, "loss": 0.119, "step": 15556 }, { "epoch": 1.0, "learning_rate": 3.279695758373613e-09, "loss": 0.1388, "step": 15557 }, { "epoch": 1.0, "learning_rate": 3.1136633699024863e-09, "loss": 0.1183, "step": 15558 }, { "epoch": 1.0, "learning_rate": 2.951943422657699e-09, "loss": 0.1232, "step": 15559 }, { "epoch": 1.0, "learning_rate": 2.7945359236225545e-09, "loss": 0.1489, "step": 15560 }, { "epoch": 1.0, "learning_rate": 2.6414408795805145e-09, "loss": 0.1323, "step": 15561 }, { "epoch": 1.0, "learning_rate": 2.4926582971263045e-09, "loss": 0.1325, "step": 15562 }, { "epoch": 1.0, "learning_rate": 2.348188182688116e-09, "loss": 0.145, "step": 15563 }, { "epoch": 1.0, "learning_rate": 2.2080305424831972e-09, "loss": 0.1279, "step": 15564 }, { "epoch": 1.0, "learning_rate": 2.072185382573366e-09, "loss": 0.1312, "step": 15565 }, { "epoch": 1.0, "learning_rate": 1.9406527087983963e-09, "loss": 0.1226, "step": 15566 }, { "epoch": 1.0, "learning_rate": 1.813432526842629e-09, "loss": 0.1391, "step": 15567 }, { "epoch": 1.0, "learning_rate": 1.690524842190566e-09, "loss": 0.1266, "step": 15568 }, { "epoch": 1.0, "learning_rate": 1.571929660149074e-09, "loss": 0.1382, "step": 15569 }, { "epoch": 1.0, "learning_rate": 1.4576469858140761e-09, "loss": 0.1256, "step": 15570 }, { "epoch": 1.0, "learning_rate": 1.3476768241260651e-09, "loss": 0.1234, "step": 15571 }, { "epoch": 1.0, "learning_rate": 1.2420191798367953e-09, "loss": 0.1556, "step": 15572 }, { "epoch": 1.0, "learning_rate": 1.140674057487079e-09, "loss": 0.1317, "step": 15573 }, { "epoch": 1.0, "learning_rate": 1.0436414614511947e-09, "loss": 0.1316, "step": 15574 }, { "epoch": 1.0, "learning_rate": 9.509213959146835e-10, "loss": 0.1242, "step": 15575 }, { "epoch": 1.0, "learning_rate": 8.625138648743481e-10, "loss": 0.1351, "step": 15576 }, { "epoch": 1.0, "learning_rate": 7.784188721493557e-10, "loss": 0.1326, "step": 15577 }, { "epoch": 1.0, "learning_rate": 6.986364213701357e-10, "loss": 0.1263, "step": 15578 }, { "epoch": 1.0, "learning_rate": 6.231665159672773e-10, "loss": 0.1293, "step": 15579 }, { "epoch": 1.0, "learning_rate": 5.520091591937337e-10, "loss": 0.1401, "step": 15580 }, { "epoch": 1.0, "learning_rate": 4.851643541248229e-10, "loss": 0.126, "step": 15581 }, { "epoch": 1.0, "learning_rate": 4.226321036360226e-10, "loss": 0.1354, "step": 15582 }, { "epoch": 1.0, "learning_rate": 3.6441241043627674e-10, "loss": 0.1369, "step": 15583 }, { "epoch": 1.0, "learning_rate": 3.1050527702358724e-10, "loss": 0.1297, "step": 15584 }, { "epoch": 1.0, "learning_rate": 2.6091070572942246e-10, "loss": 0.1428, "step": 15585 }, { "epoch": 1.0, "learning_rate": 2.156286986965128e-10, "loss": 0.1125, "step": 15586 }, { "epoch": 1.0, "learning_rate": 1.7465925786774862e-10, "loss": 0.1425, "step": 15587 }, { "epoch": 1.0, "learning_rate": 1.380023850194867e-10, "loss": 0.125, "step": 15588 }, { "epoch": 1.0, "learning_rate": 1.0565808172824376e-10, "loss": 0.1398, "step": 15589 }, { "epoch": 1.0, "learning_rate": 7.762634938179858e-11, "loss": 0.1403, "step": 15590 }, { "epoch": 1.0, "learning_rate": 5.390718920139648e-11, "loss": 0.1249, "step": 15591 }, { "epoch": 1.0, "learning_rate": 3.450060220844265e-11, "loss": 0.1299, "step": 15592 }, { "epoch": 1.0, "learning_rate": 1.9406589224502114e-11, "loss": 0.1446, "step": 15593 }, { "epoch": 1.0, "learning_rate": 8.625150926810933e-12, "loss": 0.1182, "step": 15594 }, { "epoch": 1.0, "learning_rate": 2.1562877594583085e-12, "loss": 0.1236, "step": 15595 }, { "epoch": 1.0, "learning_rate": 0.0, "loss": 0.1155, "step": 15596 }, { "epoch": 1.0, "step": 15596, "total_flos": 0.0, "train_loss": 0.04023431732994558, "train_runtime": 27864.628, "train_samples_per_second": 17.911, "train_steps_per_second": 0.56 } ], "logging_steps": 1.0, "max_steps": 15596, "num_input_tokens_seen": 0, "num_train_epochs": 1, "save_steps": 999, "total_flos": 0.0, "train_batch_size": 32, "trial_name": null, "trial_params": null }